The Use of Self-Service Technologies in Stress Management: A Pilot Project

Carissa Morris

St. Catherine University

Recommended Citation

Morris, Carissa, «The Use of Self-Service Technologies in Stress Management: A Pilot Project» (2012). Master of Social Work Clinical

Research Papers. Paper 61.

This Clinical research paper is brought to you for free and open access by the School of Social Work at SOPHIA. It has been accepted for inclusion in Master of Social Work Clinical Research Papers by an authorized administrator of SOPHIA. For more information, please contact ejasch(o)

Running head: Use Of Technology In Managing Stress

The Use of Self-Service Technologies in Stress Management: A Pilot Project

Submitted by Carissa Morris May 2012

MSW Clinical Research Paper

The Clinical Research Project is a graduation requirement for MSW students at St. Catherine University/University of St. Thomas School of Social Work in St. Paul, Minnesota and is conducted within a nine-month time frame to demonstrate facility with basic social research methods. Students must independently conceptualize a research problem, formulate a research design that is approved by a research committee and the university Institutional Review Board, implement the project, and publicly present their findings. This project is not a dissertation.

School of Social Work

St Catherine University & University of St. Thomas St. Paul, Minnesota

Committee Members:

Philip AuClaire, PhD, Committee Chair Janet Dahlem, MA, Associate Professor in MA Holistic Health Studies

April Brandt, MS


I would like to express my appreciation to the Committee Chair, Philip AuClaire, PhD for his guidance throughout the length of this project. I would like to extend my heartfelt gratitude to my committee members Janet Dahlem, MA and April Brandt for their ongoing support and encouragement both personally, as well as with bringing their holistic lens and critical eye to the research.

This research would have been impossible if not for the support of my family and friends. In particular, to the father of my children, Alex DeCamp and our always loving and supportive children, Myles and Kaia DeCamp who have also made many personal sacrifices to help make this work possible.



This research describes a pilot program created to help St. Catherine University’s Associate nursing students receive education regarding stress management and practice relaxation techniques. The program was developed using three key elements: 1) practicing a variety of relaxation and mindfulness techniques 2) practicing these techniques through the use of self-service technologies, and 3) participating in group sessions where psychoeducation regarding stress, depression and anxiety was provided.

Six participants completed the stress management program. They practiced breathwork, mindfulness and guided imagery exercises through podcasts and apps on internet-capable, mobile devices. Results showed decreases in stress and anxiety levels among participants. Additional studies, and replications of this type of program using integrative practices, in a group setting and utilizing self-service technologies as a means to reduce stress is encouraged.

Keywords: Stress, Depression, Anxiety, Podcasts, Apps, Self-Service Technologies


Table of Contents

Acknowledgements i

Abstract ii

List of Tables v

Introduction 1

Literature Review 3

Nature of Stress. 3

The Stress Response 4

Impact of Stress 6

Gender Differences in Managing Stress 7

Stress among College Students 8

Mental Health Needs at Saint Catherine University 9

How Stress Relates to Depression and Anxiety 12

College Students and Test Anxiety 13

Western Approach to the Treatment of Stress and Anxiety 14

Integrative Psychotherapy 15

Benefits of Stress Management 16

Use of Mind-Body Techniques in Managing Stress 17

College Student Satisfaction in Participating in Stress Management Programs. 18

Autogenics and Use with College Students. 19

Breathwork and Use by College Students 20

Meditation and Use with College Students. 21

Visualization/Guided Imagery and Use with College Students 22

Efficacy of Mind-Body Techniques in Stress Management Research 23

Use of Technology in Stress Management 24

Challenges/Limitations in Using Technology 28

Anticipated Benefits of the Stress Management Program 29

Reason for Research 29

Conceptual Framework 32

Program Description 35

Methods 37

Research Design 3 7

Sample 38

Protection of Human Subjects. 39

Recruitment Process & Agency and Institutional Support 39

Data Collection Instrument and Process. 40


Data Analysis Plan 40

Findings 41

Demographic Data 41

DASS-21 Results 42

Qualitative Results. 48

Discussion 48

Strengths and Limitations 51

Implications for Practice 53

Implications for Research 54

References 56

Appendixes 60

Appendix A: Informed Consent 61

Appendix B: DASS-21 Scale (Pre-test & Post-test) 65

Appendix C: DASS-21 Scoring Form 66

Appendix D: Intake Form 67

Appendix E: Practice Log 68

Appendix F: List of Podcasts, Apps, and Websites 69

Appendix G: Exit Survey 70


List of Tables

Number Page Number

Table 1. Average DASS Scores Between Pretests and Posttests p.42

Table 2. Resequenced Participant’s Scores vs. Non-Resequenced Students p.43

Table 3. Resequenced Participants Score Differences p.44

Table 4. DASS Scoring of Individual Participants by Sub-Category p.44

Table 5. DASS Items Showing Decreased Scores p.45

Table 6. DASS Items Showing Increased Scores p.46

Table 7. Frequency Distribution of DASS Items p.47

Table 8. Data from Participant’s Practice Log p.48



Over the past four decades, institutions of higher education have been working with students to address and mitigate the effects of stress while in college. Certain areas of study notoriously known for academic pressure and competiveness are those professions within the medical field. Healthcare professions have a reputation for having high academic expectations, as well as requiring healthcare professions to provide patient-centered care when interacting with their patients. The combination of being able to execute critical thinking skills while maintaining the dignity and respect of their patients creates a challenging educational experience for those just beginning in the profession. On a macro level, MacLaughlin (2001) notes that the impending healthcare personnel shortage, due to the aging baby boom generation is only likely to increase stress within the medical profession. The increasing demand for qualified health professionals places the need for higher education institutions to help meet the demand. Tipton (2008) reports that entry-level nursing programs across the country are challenged in preparing students to pass the nursing licensure exam that expects graduates to demonstrate no less than 1,185 competencies in all areas of healthcare. Therefore, when students are not successful it impacts the student, the nursing program and the healthcare community. In order to address this need, several colleges and universities have developed stress management programs to support students pursuing education in healthcare.

One way in which colleges are supporting their students is through the promotion of personal well-being. Educating students on how to practice good self-care is equally important in retaining highly trained professionals and providing good patient care (MacLaughlin, 2001). In particular, student nurses are an appropriate group to target because they are a readily available population with significant stress-related problems and are often faced with deficiencies in how to manage work/life related stress (Jones, 2000). Institutions of higher education have created programs aimed at helping students managing stress in order to support academic excellence, as well as to foster coping skills in a profession known for risk of “burn out.”

One of the paradoxes that exist with this population is motivating the student to take time to practice stress management techniques. Cost college students are short on time and the idea of integrating a new self-care habit must be affordable and easily accessible. Technology has the potential to help meet the needs of students in a variety of ways. Today’s modern college student is well-acquainted and equipped with technological devices, Computers, online learning environments, and mobile phones allow access to a variety of technological capabilities. These technologies, such as podcasts, email, computer-aided presentations, etc., are intended to enrich the student’s learning environment. It is also important to exploring how today’s technological devices and programs can also help students in managing their stress, as well.

Application software and podcasts are designed to help users to perform specific tasks. Application software is commonly known as “apps”. An app is a piece of software that can be downloaded and installed on internet-capable devices. These apps can mn on the Internet, Mp3 players, computers, phones and other electronic devices. Apps can be used to perform a variety of services and functions. Another popular technology is podcasts. Podcasts are video and audio broadcasts that have been converted to MP3 or other audio file format for playback in a digital music player.

Podcasts may be offered as downloads from a website, or they may be made available via a syndication format.

Podcasts and apps can assist people in managing their stress. People can access a variety of stress relieving and mindfulness practices through personal computers, such as IPads and Internet capable smartphones, at any location. Although students have been using these types of technologies in their daily lives, they are now beginning to be used in academic settings as learning tools. The popularity of mobile devices is rising at a rapid pace and research on the efficacy of using these devices for learning is just beginning.

The questions examined in this study is: can participation in stress management classes coupled with the utilization of self-service technologies, such as podcasts and downloadable ‘apps’ utilizing guided imagery, breathwork and mindfulness, help associate nursing students reduce their stress, depression, and anxiety levels?

Literature Review

Nature of Stress

“Stress was originally defined by Selye (1976) as the non-specific response of the body to any demand. In current usage, the term typically refers to both physiological and psychological responses to increased demands on the organism” (Giuseffi, 2011, p. e269). Stress is a normal part of the human experience. Stress can be both positive and negative. Exposure to stress can be beneficial. For example, it can increase human performance. The stress response is a fundamental human experience whose primary goal is to ensure safety and protection. However, when we are exposed to stress for an extended duration of time and/or if there are too many stressors occurring concurrently, it can produce physiological dysfunction.

Seward (2012) acknowledges that there are three types of stress: eustress, neutress and distress. Eustress is the type that is experienced by people as motivating or inspiring, such as the stress that accompanies getting married. Neustress is considered neither good nor bad. Distress is the most common and identifiable type that is considered bad. People experience stress as either being acute, which is generally abbreviated, or chronic which continues over a prolonged period of time.

The Stress Response

When a stress event occurs, our bodies have a physiological response. The stress reaction perceives the stressor as negative and a danger to our well-being. When this happens, the sympathetic nervous system is activated. Our bodies respond with increased heart rate, sweating, blood vessel constriction, and muscle tension, as this is our body’s way to prepare itself to begin to cope with the stressor.

These physical signs initiate what is known as the “fight or flight response”.

Also, often referred to as the “fight, flight or freeze response”. When humans experience an overwhelming threat to their well-being, they are motivated to either fight or mn away. “In 1914 Harvard physiologist Walter Cannon first coined the term fight or flight response to describe the dynamics involved in the body’s physiological arousal to survive a threat” (Seward, 2012, p. 6). This response is known as our stress response.

The fight aspect can involve a physical argument, verbal assaults, and increasingly more common, the use of technologies such as email and text messaging. Whereas, the flight response includes physically escaping the stress, as well as through escapism such as

surfing the web, playing video games and the use of drugs or alcohol. Freezing is often seen in cases involving young children. For example children, when exposed to an acute stressor, will freeze as a means of coping with the stressor. It’s important to understand our bodies stress response in order to help recognize the effects of stress on humans.

In a healthy stress response, once the stressor is no longer a threat, the body begins to recover and regain homeostasis. An unhealthy stress response occurs when the individual is unable to recover from the stressor or when the stressor is ongoing and prolonged. As a result, the body and mind is uncomfortable with the thoughts and sensations it’s experiencing and quickly works to internalize and/or inhibit the stress reaction. Individuals can develop maladaptive coping behaviors in order to avoid their uncomfortable thoughts and feelings. However over time, this often leads to dysregulation in the body, which can manifest itself in problems like depression and anxiety disorders.

Experiencing stress is inevitable. However, people’s ability to manage stress varies. Some individuals are more susceptible to the effects of stress, whereas others may be more resilient. Those who are sensitive to the effects of stress can practice ways of building their resiliency. A common practice includes developing mindfulness to a means of coping with a stressful event. Mindfulness is an essential practice in caring for self.

Stressors are often thought of as being external. However, Vogel & Bower (1991) explain that stress can also manifest from within the mind of the individual. “We create our own stress, make our own stressful events, and cause our own diseases. These processes are based on specific genetic vulnerabilities, individual experiences, and environmental circumstances” (Plotnikoff, 1999, p. 169). Therefore, not only are we responding to stressors as part of everyday life, but we also are responsible for manifesting stress based on how we perceive and interpret our experiences. The degree to which an individual is susceptible to the effects of stress and how it manifests in a person’s life is variable by the nature and extent in which it is experienced (Plotnikoff, 1999).

Experiencing stressors is unavoidable and it often occurs unexpectedly.

However, stressors are also anticipated, particularly when we are on the cusp of major life transitions. Attending college is one of those times. For traditional age students it is the first time they are living independently of their parents, and it is a large financial investment into their future for students who are the first in their family to attend college. The pressure to succeed is great. The latter often lacks the support upon entering college on how to navigate higher education. In addition, today’s college students are often balancing the additional responsibilities of work and family responsibilities.

Impact of Stress

Exposure to stress can have detrimental effects on one’s health. Stress is a major contributor to leading causes of death, such as heart disease and stroke. Selye (1976) states stress is a high-prevalence universal problem. On a global level there is a high- prevalence of stress. “A 2006 international poll showed that nearly 75% of residents of developed countries reported feeling stressed on a daily basis (Associated Press-Ipsos, 2006)” (Giuseffi, 2011, p. e269). Of those surveyed, 53% felt concerned about their current level of stress (Giuseffi, 2011).

Gender Differences in Managing Stress

Research indicates that there may be some fundamental differences about how women and men experience and cope with stress. A theory supporting the differences between how women compared with men deal with stressors is termed ‘tend and befriend’. In 2000, Shelly Taylor developed the theory which asserts that women are more likely to rely on their friends and socials networks during unpleasant events and circumstances, whereas, men are more likely to engage the flight or fight response (Seward, 2012). Recognizing the need for adequate social support for women to help in managing stress is an important aspect to consider in approaching stress management.

“Social support has been found to mitigate the impact of stress (Cohen & Wills, 1985). However, women attending college reported being dissatisfied with their available social support. Even though most women (70%) had a close confidante, women of all ages reported being uncomfortable asking for help. Feeling uncomfortable imposing on others was also universal” (Giuseffi, 2011, p. e279). Guiseffi (2011) also found that non-traditional aged women in college are more likely to experience feelings of isolation due to the fact that many women are balancing family and/or work demands. Although a growing minority among the college population, their needs are unlike the traditional aged (18-24 year old) college students of their peers.

According to Giuseffi (2011), research has shown that women: 1) report higher levels of daily and chronic stress; 2) are more likely to report home and family life events and caregiving roles as stressful; 3) have reported higher rates of negative


emotional reactions (e.g. irritability, sadness) and physical symptoms (e.g. headache, gastrointestinal complaints) in response to stress, and as a result; 4) women were shown to be more likely to use prescription medications, and over/under eat in attempting to cope with stress (Giuseffi, 2011). Female students feel overwhelmed and pressured for time. Self-doubts and the perceived need of students to focus virtually exclusively on classes may conflict with attempts to meet family demands, compounding stress. Therefore, it is critically important to help women integrate stress management techniques that encourage self-care to not only help manage the multiple and often conflicting roles and responsibilities, but to also help foster emotional, physical and mental well-being.

In a 2009 College Counseling Center Directors Annual Survey, asserts “this shift is especially alarming for female students, as women in general suffer disproportionately from depression and anxiety (Arehart-Triechel, 2002) and most campus counseling clientele are female (Beattie, 2010). Additionally, women are often more likely to be balancing family needs and responsibilities, as well as managing their personal welfare. Stress, anxiety and depression have very similar symptoms, making it difficult to discern their differences. Stress and anxiety are linked to feeling hopeless, while depression is linked to an inability to function and, in extreme cases, suicide (Barr, Rando, Krylowicz, & Winfield, 2010). Useful intervention strategies are needed in order to address anxiety and stress because it is essential to the success of female university students (Margolin, 2011).

Stress among College Students

Stress among college students is commonly known. College students find


themselves in a new environment with new responsibilities. It’s a major life milestone and transition in one’s life. College students take on new financial responsibilities and academic pressures. They are exposed to new people, concepts and temptations. The growing diversity among the student population brings students with unique needs. Generally speaking, they are required to make decisions on a higher level than what they are used to. “In a 2009 College Counseling Center Directors Annual Survey, 94% of Directors reported an increase in the number of students with significant psychological problems (Barr, Rando, Krylowicz, & Winfield, 2010). According to one Canadian counseling center, over the past decade, students’ concerns have shifted from normal developmental issues, (75% in 2001; 35% in 2008) to more serious and severe mental health difficulties (25% in 2001; 35% in 2008; Beattie, 2010)” (Margolin, 2011, pp. 234- 235). Healthy stress management is critical for students to be able to perform well academically.

Mental Health Needs at Saint Catherine University

A 2008 survey of St. Catherine University students conducted by the University of Minnesota Boynton Health Services found that approximately 25% of students report that they are unable to manage their stress level, and 40% of students report having adequate sleep only three or fewer days in the previous seven (Boynton, 2010). Without the ability to receive adequate sleep and manage stress, academic performance is negatively affected.

Peterson (2011) reported in a survey of St. Catherine University, that associate- degree students reported anxiety and depression as the two most frequently reported mental health diagnoses for both within their lifetime and experienced within the past 12 months. Among the associate student population, 43.3% report being diagnosed with mental illness in their lifetime, 24.8% of two-year students report that they are unable to manage their stress level, 16.8% of two-year students with unmanaged stress report being diagnosed with anxiety within the past year, and 12.9% of two-year students report taking medication for depression. The reported experiences of St. Catherine University students, of which 90% are women, are in alignment with national trends indicating a prevalence of mental health issues, particularly, anxiety and depression.

National Affairs Administrators in Higher Education (NASPA) reports an 80% increase in elective serotonin re-uptake inhibitors (SSRIs) use in the last five years. SSRI’s are a class of compounds typically used as antidepressants in the treatment of depression, anxiety disorders, and some personality disorders. Ten years ago 16% of counseling center students had a major mental health diagnosis, and today that has increased to 44%. NASPA predicts that mental health needs amongst college students will continue to increase for the next ten to twelve years (Peterson, 2011).

Stress experienced during college years can be detrimental to one’s health and well-being. MacLaughlin (2001), reports the stress among medical professionals begins prior to the commencement of their professional careers. Medical professionals report having begun experiencing high levels of academic stress during their academic and clinical trainings while in college. “Medical education has deleterious consequences. Trainees (students, interns, and residents) suffer high levels of stress, Medical students have mean anxiety scores one standard deviation above those of non-patients, and their depression levels increase significantly throughout the first year of medical school” (Shapiro S. S., 2000, p. 748). In the length of time it takes for college students to earn

their degree, students experience both chronic and acute stress. Brennan (2010) addresses this as the inherent challenge of adjusting to college life. There is pressure related to both academic and performance expectations.

Research related to college students in medical and nursing programs has shown stressors unique to their profession. Students pursing degrees in the medical field must meet rigorous standards and competencies in order to be certified or licensed to practice. Generally speaking, the medical profession is known for its hierarchy of educational levels, as well as expertise within specified areas related to illness, health and recovery. “A review by Dyrbe et al. highlighted the many stressors unique to medical school, which range from the high volume workload to ethical conflicts, and exposure to human suffering and death” (MacLaughlin, 2001, p. 1).

Brennan (2010) reports that women in the medical field may experience additional stress due to the fact that it is a traditionally male-dominated field. Additionally, medical students report feelings of powerlessness contribute to their overall stress. “Although there is an individual optimal stress level that can enhance one’s performance, too much stress or test anxiety can also hinder an individual’s performance. Studies have revealed that being a student, not just a medical student, is stressful and that students’ main concerns are academic performance and/or test anxiety” (Paul, 2007, p. 287). There are several ways in which stress can compromise the health and well being of students in college. In addition to impacting academic performance, De Kooker (2008) identified that exam-related distress also impacts immunity, decreased attention and concentration, poor decision-making ability, alcohol and drug abuse, depression and anxiety, relationship difficulties, and suicide (Shapiro S. S.,



There are several short-term symptoms of stress. Shapiro (2000) and MacLaughlin (2001) discuss the negative effects that include decreased attention and concentration, difficulty with decision-making, and challenges with interpersonal relationships. Typical behaviors include increased cynicism, academic dishonesty, substance abuse, depression and anxiety, and even suicide.

Long-term consequences of stress can lead to physiological dysfunction resulting

in increasing symptoms of anxiety and/or depression. Overtime, these symptoms can

lead to burnout. Brennan (2010) reported,

Anxiety, depression, and a general sense of “burnout” may all result from the increased stress experienced by medical students. Dyrbye and colleagues report that about 50% of medical students have experienced ‘burnout’ and about 10% have experienced suicidal ideation at some point during medical school. Although many students are habitual exercisers in college, many decrease or eliminate ‘self care activities’ such as exercise and monitoring their diet and general health in response to the time demands of medical school. Increased use of alcohol and drugs may also occur. (Brennan, 2010, p. 15)

Mental health support and stress management programs are important aspects of supporting the social and emotional needs of students in college, particularly when these students enter a profession known for having high levels of stress and potential for professional burnout.

How Stress Relates to Depression and Anxiety

According to Grobeman (2001) stress is a result of particular situations that arise during a person’s life as they deal with certain daily pressures. These pressures elicit a stress response within the body and cause adrenaline to be released. This creates a cascade of bio-chemical changes that can lead to depression, raised blood pressure and


other negative health side effects if it occurs for too long. Stress can also cause anxiety, and depression. This is often why depression and anxiety are closely linked. Anxiety is the process during which a person becomes scared and apprehensive of what may lay ahead. Anxiety symptoms often manifest in physical problems like physical pains, dizziness, and panic attacks.

While the triggering of a stress-inducing factor known as a stressor causes stress, anxiety is what happens when someone gets stressed out and has no reasonable root stressor that can simply be removed. This is precisely why while anxiety is considered a legitimate mental disorder, stress is not.

(Grobeman, 2001)

Anxiety symptoms that extend beyond 6 months are the period of time that qualifies it as a clinical diagnosis.

College Students and Test Anxiety

Test anxiety is a common experience among college students, particularly in programs that are competency based and are intrinsically competitive. According to Paul (2009) test anxiety is a type of distress, which has a physiological component. “Research studies have found that elevations in corticosteroid levels (hormones released during times of distress) can impair declarative memory, concentration, and learning. Thus, high levels of stress can make it more difficult for students to concentrate and comprehend information” (Paul, 2007, p. 287). The body’s fight or flight’s response can inhibit their ability to access complex information and critical thinking skills. “The worry or emotional arousal that can accompany test anxiety causes the student to become centered on the “self’ during an exam rather than the task, thus possibly negatively influencing future academic performance if no coping techniques are adopted” (Paul, 2007, p. 287).

Western Approach to the Treatment of Stress and Anxiety

General practitioner and family practice doctors see the majority of people currently being diagnosed and treated for anxiety and depression. “These doctors, including internists, pediatricians and obstetrician/gynecologists, prescribe more than 60 percent of anti-depressants and anti-anxiety drugs” (Solomon, 2011). As a result, doctors need to seek continuing education on diagnostic assessment of mental health issues. Seven percent of all visits to primary care doctors result in prescriptions for antidepressants. This is up from three percent in 1997, according to a study published recently in Health Affairs. Of the mental health diagnoses by family physicians, only vague complaints of «fatigue, pain and malaise,» are noted, and often without a formal clinical diagnosis (Solomon, 2011).

When mental health professionals for clinical assessment see patients, patients participate in a psychosocial assessment as part of their diagnostic assessment. In order for a mental health practitioner to be able to make a diagnosis of anxiety the patient must fit the following criteria as defined by the American Psychiatric Association. The patient must demonstrate excessive anxiety and worry, occurring more days than not for at least six months, about a number of events or activities. The person finds it difficult to control the worry. The anxiety and worry are associated with three (or more) of the following six symptoms: 1) restlessness or feeling keyed up or on edge; 2) being easily fatigued; 3) difficulty concentrating or mind going blank; 4) irritability; 5) muscle tension; 6) sleep disturbance and; 7) The symptoms cause impairment in social, occupational, or other important areas of functioning. Whereas, for a diagnosis of depression a patient must report at least five of the following symptoms, 1) depressed

mood, most of the day; 2) diminished interest/pleasure; 3) weight gain or loss; 4) insomnia; 5) fatigue; 6) feeling worthless and; 7) difficulty concentrating, as defined by the American Psychiatric Association (2000) in the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV).

In a clinical counseling session, De Kooker (2008) reports that a therapist helps patients manage their stressors by helping them recognize and deal with psychosocial factors that may interact negatively with existing conditions. Goals include broadening the individual’s scope of attention, cognition and action, and the building of physical, intellectual and social resources that facilitate self-efficacy, optimism, resilience and health. By understanding oneself in an environment in a complex, and multi-faceted way, the individual develops a wide range of personal coping skills in managing stress, and therefore reduces their vulnerability to anxiety, depression and immune dysfunction.

When patients are seen by a mental health professional for anxiety, in addition to the prescription of medications, the following counseling treatment components are included in a treatment plan which may involve psychoeducation, relaxation training, gradual exposure, cognitive restructuring, study skills training, and relapse prevention. Clinician-administered treatments tend to be effective in both individual and group modalities (Pless, 2010). In addition to traditional approaches to addressing anxiety and depression, the integration of holistic approaches have gained in popularity and efficacy in treatment outcomes.

Integrative Psychotherapy

Integrative psychotherapy refers to the bringing together of the affective, cognitive, behavioral, and physiological systems within a person, with an awareness of

the social and transpersonal aspects of the systems surrounding the person. In the context of anxiety and depression, it’s important to address the spiritual practices and beliefs, nutrition, sleep hygiene and social support.

German (2003) illustrates a case study of an integrative approach with a college student named Mark. He began integrative therapy with a psychologist for treatment of his symptoms of anxiety and depression, as well as suicidal ideation. Mark reported high levels of stress and perfectionist tendencies. In conjunction with conventional clinical treatment goals, the therapist used relaxation techniques including progressive muscle relaxation and deep breathing techniques in session. Additionally, the therapist utilized other mental strategies such as self-hypnosis, mindfulness, and meditation to assist in managing stress. Results of the case study resulted in a reduction of both anxiety and particularly depression to where it would be considered in the “normal” range. At termination, the client was able to administer self-hypnosis as a way of increasing self -awareness through a mindfulness practice.

Benefits of Stress Management

As mentioned earlier, practicing stress management techniques can help prevent professional burnout. Research has also shown that participants who practice stress management reap many personal benefits. Shapiro (2000) performed a meta-analysis study of stress management programs in medical schools and reported medical trainees participating in stress-management programs demonstrated: (1) improved immunologic functioning; (2) decreases in depression and anxiety; (3) increased spirituality and empathy; (4) enhanced knowledge of alternative therapies for future referrals; (5) improved knowledge of the effects of stress; (6) greater use of positive coping skills; and

(7) the ability to resolve role conflicts. Four of the 24 stress management programs reviewed found no difference between experimental and control groups on standardized measures of psychological functioning, immune functioning, or health at postassessment. However, participants did report that they felt positively regarding their participation in the stress management programs (Shapiro S. S., 2000).

Use of Mind-Body Techniques in Managing Stress

In stress management programs, the use of several types of mind and body techniques is practiced to initiate the relaxation response. Using mind-body techniques acknowledges that our thoughts impact our body, as our body impacts our mind. MacLaughlin (2001) states our well-being is enhanced by the promotion of self- awareness and self-care through using integrative techniques. “Mind body medicine takes into account the connectedness between the mind and body, and its effect on overall health. It embraces several well-defined strategies such as relaxation, meditation, yoga, biofeedback, imagery, autogenic training, tai chi, qigong, hypnosis and spirituality” (MacLaughlin, 2001, p.2).

Regular practice of relaxation techniques reinforces what is learned during the session and facilitates generalization of the relaxation response. In this way individuals leam to heighten body awareness and reduce physiological arousal leading to more automatic relaxation (Andrasik, 1990). Thus, individuals are able to develop valuable self-regulation skills that help to relax the central nervous system, decrease sympathetic arousal, and help retrain the autonomic nervous system – producing homeostasis and supporting general health and well-being.

College Student Satisfaction in Participating in Stress Management Programs

It is expected that participants will learn new coping skills and techniques to assist with managing stress. For example, in Beauchemin’s (2008) research adolescents diagnosed with learning disabilities participated in a 5-week mindfulness meditation program. After the 5-week pilot, outcomes demonstrated a reduction in feelings of anxiety by the participants. Additionally, students expressed positive attitudes regarding the experience. One hundred percent of participants reported positive feelings about the meditation and expressed that the mediation led to feelings of calm, quiet, relaxation, peacefulness, or better overall feelings. Also, students demonstrated improved social skills and improved academic performance.

Brennen (2010) reported that students were most satisfied in sessions when they could practice small behavior changes that did not take much time out of their daily schedule. For example, deep breathing, identifying and then reducing tension in their body, changing the foods that they eat and staying mindful were found to be easily integrated into one’s life without much hassle. Whereas, sessions that required students to do more complex tasks and may have required more time and energy, such as guided imagery, balancing, coping, and changing their thinking, seemed more difficult for them to integrate into their lives.

Another benefit of the program was that students became more aware of their stress level and therefore more able to proactively address it. A few participants asked about referrals for mental health treatment, who most likely would not have done so without this program. Thus, such a program is an effective way to help students improve their self-assessment of their own mental state (Brennan, 2010). Stress management

techniques frequently cited in the literature include autogenics, breathwork, meditation and guided imagery. It will be important to further explore each of the modality’s application and potential benefits.

Autogenics and Use with College Students

Autogenic training is a mind-body technique that is a self-relaxation procedure based on passive concentration on functioning of the body. For example, imagining heaviness and warmth of arms, legs and abdomen or concentrating on the rhythm of breathing and heartbeat. Autogenic training is particularly helpful when practiced by individuals at the end of the day to help induce sleep. Jorm (2001) performed a metaanalysis of clinical outcome studies that assessed the effectiveness of this treatment in reducing anxiety. According to Jorm:

Three studies found that, compared with control groups, autogenic training significantly reduced anxiety for individuals with a diagnosed anxiety disorder, and those who described themselves as being anxious or having high levels of stress. One study found that student nurses given autogenic training for 6 weeks to reduce their risk of stress took significantly fewer days off compared with a control group. Two other studies found that autogenic training alone or autogenic training with other treatments have so significant effect on individuals’ levels of test anxiety. The remaining study found that study participants who experienced ‘tension’ reported that their levels of tension reduced by similar amounts after receiving no treatment, after receiving autogenic training only, or after receiving autogenic training and other treatments. (Jorm, 2004, p. S35)

Although the remaining study did not show improvement in regards to the reduction of tension, it did not produce any harm to the participants. Research has shown that autogenic training can also help with overcoming addictions, change unwanted behaviors, resolve anxieties, and mitigate symptoms of physical ailments.

Breathwork and Use by College Students

Breathwork is a specific body-mind therapy that utilizes consciousness and breathing in an intentional way. Breathwork utilizes the diaphragm when drawing in deep, slow breaths. The breaths can be connected to where there is no pause between the inhalation and exhalation. This is sometimes called circular breathing. Breathwork is a therapeutic tool that has the capacity to affect not only the physical aspects of our bodies but also the emotional, mental and spiritual aspects as well.

Paul (2007) conducted a study with 64 college students, utilizing breathwork. As a result, students reported decreased perceptions of test anxiety, nervousness, self-doubt, and concentration loss. They used the technique outside of two classes, and believed it helped them academically, and that it would help them in the future in their professional role. The initial benefits of using breathwork allow the body to get into a state of relaxation and release feelings of stress. As our bodies and our breath become deeper, we leam to physically relax the exhale. Once this becomes well practiced, we find that this type of breathing has the capacity to release deeper emotional and mental patterns as well.

Diaphragmatic breathing is also often used as part of meditation practices. Paul’s (2007) research showed students who meditated or used diaphragmatic breathing resulted in significant increases in students’ academic learning and achievement. According to Paul:

Diaphragmatic breathing is known to counteract the fight or flight response symptoms that are often associated with anxiety. When a student perceives the exam to be a challenge or has low self-confidence from previous testing experiences, he or she may experience the physiological symptoms previously discussed in addition to the symptoms associated with the fight or flight response that include increased heartbeat, respiration, blood pressure, muscle tension, and gastric discomfort.

Meditation can also be used to counteract stressful situations, as it is a technique to develop concentration and awareness to produce a calming effect; diaphragmatic breathing is central to any meditation practice. (Paul,

2007, p. 288).

Meditation and Use with College Students

Meditation practice encompasses a variety of elements that can vary by persons practicing meditation. Meditation can be practiced, sitting up, lying down or in an inverted position. Some people who meditate may use chanting, and breathwork as part of their practice. Some meditations, such as loving kindness meditation, can be used to express and nurture gratitude within and between oneself and others. Tang (2007) cites meditative techniques that are easier for novices to begin are: concentration meditation, mantra, and mindfulness meditation, which rely on mind control or thought work which can include focusing on an object and/or paying attention to the present moment. It is easier to begin with concentrating on something instead of trying to be void of thought. Tang (2007) also states that using compact discs for practice instruction occupies the novice’s wandering mind through continuous sensory input, maintaining and facilitating the mindful state. Many meditation-training methods use audiotapes or compact discs to help beginners.

Mental training methods also share several key components, such as body relaxation, breathing practice, mental imagery, and mindfulness, which can facilitate a meditative state. Research by Shapiro (2008) and Tang (2007) shows that integrative mind-body training improves emotional and cognitive performance and social behavior. Tang’s (2007) research goes on to say that meditative practices have been shown to be effective after only a few days of practice and the combination of breathing, mental imagery and mindfulness work together to amplify the training effect versus the use of only one of these components.

Integrating a mindfulness practice while in college can improve both cognitive and academic performance. Shapiro (2008) conducted a meta-analysis of stress management programs and concluded that meditation has a positive impact on academic performance, psychological well-being, and interpersonal experience for students in college, medical school, and other higher education settings. “Mindfulness meditation may improve ability to maintain preparedness and orient attention. Mindfulness meditation may improve ability to process information quickly and accurately” (Shapiro S. W., 2008, p.

4). Shapiro (2008) cites that concentration-based meditation, practiced over a sustained period of time, may show a positive impact on academic achievement and decrease stress, anxiety and depression levels.

Meditation helps to develop the whole person. Shapiro (2008) states that meditation supports the development of creativity. It supports and enhances the development of skills needed for interpersonal relationships, as well as, cultivating compassion of self and others.

Visualization/Guided Imagery and Use with College Students

Guided imagery is a program of directed thoughts and suggestions that guide your imagination toward a relaxed, focused state. “Mediation and visualization are ancient helping strategies well established in Hinduism, Judeo-Christianity, and traditional Chinese and Aboriginal medicine. These alternative interventions are resurfacing in contemporary North America and show promise for increased self-awareness, concentration, improved mood and sleep, reduced fear and anxiety, and pain reduction” (Margolin, 2011, p. 235). Practicing visualization is popular approach to helping initiate the relaxation response in the body by using the mind to affect the body and vice versa.

“Visualization is the practice of conscious control of mental imagery. ‘Mental imagery’ refers to the perceptual information that is brought to the mind from memory and imagination, rather than arising from activation of the sense organs” (Margolin, 2011, p. 241). Visualization has been practiced for centuries and is used in many hospitals to help cope with pain, as well as in the promotion of healing. “Guided imagery is the form of visualization most studied and practiced in relation to health and wellness in recent years. Its aim is to employ the mind’s language to communicate with the self and make sense of both inner and outer experiences (Kabat-Zinn, 1990). Margolin (2011) found that most female university students in their study suffered from anxiety concerning exams and feeling overwhelmed with coursework. Many chose to explore the effects of creative visualization on the psychological distress they were experiencing at university. Using creative visualization as a strategy, participants reported a newly found awareness in their ability to curb their performance-anxiety concerning university exams and coursework.

Efficacy of Mind-Body Techniques in Stress Management Research

Tang (2005) found that participants practicing mindfulness and meditation with only 5 days of practice showed significantly better attention and control of stress, compared with the control group. The research study was comprised of 40 undergraduate Chinese students, who were given five days of 20-minutes a day to practice mind-body training techniques. The results showed greater improvement in conflict scores on the Attention Network Test, lower anxiety, depression, anger, and fatigue, and higher vigor on the Profile of Mood States scale. Physiologically, they showed a significant decrease in stress-related cortisol, and an increase in immunoreactivity (Tang, 2007, p. 17152).

Dr. Jon Kabat-Zinn, a well-known medical doctor is known for developing a Mindfulness-Based Stress Reduction (MBSR) program that has been shown to combat stress, psychologic distress, and job burnout that health care professionals frequently experience. Other studies show similar stress-reducing effects of mindfulness meditation in primary care physicians and in nursing students (Horowitz, 2010).

Use of Technology in Stress Management

Technology has become integrated into every aspect of modem society. The use of media to teach stress management techniques has been on the rise. For example, mobile web applications, known as “apps”, harness the power of the Internet with the simplicity of multi-touch technology on a small screen and have seen exponential growth in the last few years (Pulman, 2010). IPhone and iPod touch customers have now downloaded billions of apps (Apple, 2009) and can choose from a range of 20 different categories including health. Pulman (2010), reported health-related apps that offer tremendous potential for users and could be specifically designed for particular conditions and purposes. For example, Mayo Clinic Meditation (Apple App Store, 2010) is a clinically validated method of meditation developed by the Mayo Clinic, which offers a 5 and 15-minute meditation aimed to help users feel more focused and relaxed throughout their day.

Research shows that 18-26 year olds are likely to use the Internet and self-service technologies as a source for health information. Cotton & Gupta (2004) studies

computer usage for health information. Results indicated that individuals who seek health information online tend to be younger and college-educated. They are also more likely to spend time using the Internet for activities other than email. Furthermore, Escoffery’s (2005) study of undergraduate college students indicated 74% of respondents indicated having received health information using the Internet. Johnson (2011) reported that by 2015, 80% of people accessing the Internet would be doing so from mobile devices. Users access the Internet through the use of applications for many reasons, and in particular for education.

Video and audio podcasts, as well as online self-help programs have been developed to help people practice stress management techniques. Sethi (2010) conducted research utilizing Computerized Cognitive-Behavioral Therapy (CBT) adults suffering from depression. The internet-delivered self-help program used minimal therapist contact for patients suffering from mild to moderate depression. Participants who received the online self-help intervention improved significantly on measures of depression compared to those in the control group. “The authors concluded that Internet self-help can have positive effects on mild to moderate depression and is more effective in terms of both time and cost than traditional face-to-face CBT” (Sethi, 2010, p. 145). “The results detailed indicate differences in symptoms between participants in the experimental and control conditions. Both computerized cognitive-behavioral therapy (CCBT) and standard face-to-face cognitive-behavioral therapy (CBT) produced significant reductions in symptoms of participants suffering from mild to moderate depression and anxiety when compared with a no-treatment control group” (Sethi, 2010, pp. 155-156). In a similar research study, Marks (2003) reported, “Individuals with anxiety and depression improved significantly and clinically meaningfully, and were

fairly satisfied with computer-aided CBT despite a marginal preference for face-to-face care” (Marks, 2003, p. 62).

Van Vilet (2009) used web-based curriculum with students teaching stress management and coping techniques. Results showed a small but significant increase in their knowledge about stress and coping. Students reported increased use of supportseeking coping behaviors, which was an aspect in every lesson. Psychological distress decreased and life satisfaction increased, consistent with the intervention having a beneficial effect on mood. The web-based curriculum provided mental health education that was financially feasible and demonstrated efficacious results at the level expected from a short-term intervention(Van Vilet, 2009).

Brennan (2010) offers additional considerations when utilizing technology as an intervention. Sending email or text messages as reminders to practice, as well as providing relaxation exercises on the web or in mp3 format for participants to use as a supportive measure is helpful in prompting participants to engage with the technology. Brennan also supports the idea of integrating these sessions into the medical school curriculum as well as including some individual coaching or mentoring to optimize the effectiveness of the program.

In a similar study, Proudfoot and colleagues (2003) examined the efficacy of a commercially available computerized CBT program called Beating the Blues (BtB). For the study, BtB was administered to a patient who reported feeling anxious and depressed and was compared with general practitioner care. Results demonstrated significant improvements across measures of anxiety, depression, and social adjustment, in favor of the computerized CBT program. Treatment gains were maintained at a six-month follow-up” (Sethi, 2010, pp. 145-146). “These findings have been replicated in studies

by Cavanagh and colleagues (2006), larger decreases in symptoms of depression and/or anxiety were observed following computerized CBT” (Sethi, 2010, pp. 145-146).

An Australian-developed Internet intervention for depression (MoodGYM) has shown promise in trials with adults (Christensen, Griffiths, & Korten, 2002; Christensen et al., 2004). MoodGYM provides information about depression and teaches skills to address symptoms based primarily on cognitive-behavioral principles, including cognitive restructuring, pleasant activity scheduling, and interpersonal problem solving. It also provides animated demonstrations, quizzes, and ‘homework’ exercises for users to learn more about proneness to depression and to practice their skills” (Sethi, 2010, pp. 145-146).

Research is being done not only on treatments with technology as the intervention but also as a supplement to augment skills presented during therapy sessions. Sethi (2010) reported that face-to-face sessions followed by a computerized treatment had the largest impact on decreasing symptoms of depression and anxiety. The rationale behind the efficacy of this treatment is the belief that meeting with a therapist in real time motivated individuals to “to do a better job” on MoodGYM and to engage more with the program. Following individual therapy session practicing Cognitive Behavioral Therapy (CBT) with the participant, and having built a rapport with the therapist, participants were able to cement their learning with computerized CBT modules and demonstrations. “Despite these findings, results suggest that receiving treatment solely through MoodGYM is effective in reducing symptoms of anxiety, distress, and automatic negative thoughts in comparison to a control group. This is consistent with previous findings (Christensen et al., 2004) and is especially important for those in remote areas who are unable to access face-to-face therapists due to geographical barriers. This may also be an important finding for those who either cannot afford therapy or for those who find it difficult to speak ‘one on one’ about their depression or anxiety” (Sethi, 2010, pp. 156-157). Technology-based, confidential solutions like digital health coaching or other self-help technologies may reach women who may not otherwise seek or receive help for stress management (Giuseffi, 2011). In addition, accessibility and affordability is more cost-effective than traditional, in-office therapy sessions. In web-based education, fidelity is assured, scalability is simple and costs are minimal (Van Vilet, 2009).

Sethi (2010) cites some of the inherent challenges of seeking support from mental health practitioners that includes a shortage of skilled therapists, long waiting lists, and affordability. As mentioned earlier, people seeking treatment for symptoms of depression and anxiety are more likely to seek treatment through a doctor of medicine.

Challenges/Limitations in Using Technology

There are some concerns regarding the use of technology in helping treat persons with depression and anxiety. Pinnock et al. (2007) raised concerns about the risk of engendering dependence and impeding an individual’s ability to self-manage (Pulman, 2010). Secondly, individuals who have limited access to computer or are unable to type, face barriers in utilizing the online programming. Therefore, it can make it more complicated than showing up for a face-to-face meeting with a counselor” (Pulman, 2010, p. 117). Accessibility to technology, as well as having clients who are literate and technologically savvy is essential components to this using this approach. Additionally, “many researchers working within the field of computerized self-help interventions highlight the challenges associated with participants dropping out of studies” (Farvolden et al., 2005; Eysenbach 2005; Christensen et al., 2004). Using technology as the intervention can make it difficult for participants to be motivated to continue the work. In a traditional counseling situation utilizing CBT, the therapist’ ability to establish tmst and rapport is a key component to retaining the client in the therapy. Sethi (2010) cites the major challenges in using technology in providing self-help programs, therefore the goal is to maximize engagement, as well as strategies to encourage retention.

Anticipated Benefits of the Stress Management Program

Studies have indicated that nurses and nursing-students alike benefit from programs designed to enhance their self-care abilities (Anstead, 2009, p. 50). The primary goal of the stress management sessions is to help participants to lower their stress, depression and anxiety levels, through practicing relaxation and mind-body techniques.

Participants will receive psychoeducation regarding stress, the stress response, anxiety and depression, as well as how the use of mind-body techniques helps initiate the body’s relaxation response. Through psychoeducation and self-reflections, participants will increase their understanding of their environmental stressors and triggers, as well as a better understanding of their relationship with stress and ability to cope.

Reason for Research

Stress management programs exist in medical schools because high levels of stress negatively affect student performance. Medical schools rely on maintaining high board passage rates in order to maintain their accreditation. However, the prevalence and impact of stress on students puts those rates at risk. If students are

unable to be successful academically, it impacts the student, the nursing program, and the healthcare community. According to Jones (2000), student nurses are appropriate to target for stress management since they are a readily available population with significant stress-related problems. By supporting a student’s personal well-being through providing stress management, it also supports the goals of the institution to foster academic achievement, and to retain highly trained professionals who are better able to provide good patient care.

Persons susceptible to the negative effects of stress make them vulnerable to developing symptoms of depression and anxiety. According to the research previously cited in the 2008 survey of St. Catherine Associate students in healthcare programs, conducted by the University of Minnesota Boynton Health Services, it identifies the need for further support of college students managing stress, depression and anxiety. These needs are forecasted to grow in the years ahead. By addressing one’s stress in college, the student is better equipped to handle the stress that will carry over into their professional career.

For over four decades, stress management programs in higher education have been helping medical students. These programs are known to utilize the practices of meditation, guided imagery and breathwork. These practices can be effective in helping people manage stress, depression and anxiety. In today’s modern world, the convenience and integration of technology into the academic and personal lives of students has grown significantly. It is important to examine how technology could assist students in managing their stress through its utilization. Self-service technologies are commonly used and it’s important to examine ways in which they can support the user’s well-being.

Several reputable institutions such as the University of Minnesota’s Center for Spirituality and Healing and the Mayo Clinic have developed apps that facilitate meditation and guided imagery for stress management. Research shows that self-help technologies can be as effective as traditional methods for treatment of stress, anxiety and depression making a convincing argument for serving as a catalyst for moving through therapeutic goals more quickly, but also to greater therapeutic depth. This research is intended to use technology as a means by which nursing students can practice relaxation techniques while participating in group stress management sessions and assessing how it affects stress, anxiety and depression levels.

The advancement of technology has dramatically changed how people work and live. The consequences of which, have been both positive and negative. It is important to examine to what extent technology could assist people in managing stress. Shapiro (2000) identified several positive impacts stress management programs have demonstrated. In addition, Sethi (2010), Marks (2009), Brennen (2010) and Giuseffi (2011) cited research demonstrating that self-help technologies can be as effective as traditional methods for treatment of stress, anxiety and depression.

Paul (2007) recommends that medical students be provided with educational interventions, which teach them effective coping strategies as a proactive measure to counteract distress. “Students must be provided with regular opportunities to develop and practice a strategy as it takes time to abandon an ineffective behavior and replace it with a behavior that is empowering. Our findings suggest that significant behavior changes occurred when students were given a continuous opportunity to practice a 5- minute stress reduction technique meant to reduce the physiological and psychological effects that can be associated with academic stress” (Paul, 2007, p. 290).

Using technology as part of a stress management program to practice relaxation techniques, while participating in group sessions providing an opportunity for students to receive psychoeducation regarding stress and mental illness, as well to acknowledge and understand the consequences of stress, and to acquire a variety of techniques to utilize during future stressful situations.

There is an established need for a stress management program for nursing students. Balancing the demands on a college student’s time often challenges him/her and is a consistently reported concern. Additionally, Guiseffi (2011) reported women of all ages are uncomfortable asking for help. It’s also important to consider the previously cited theory of a women’s need to “tend and befriend” while experiencing a stress event. Taking this theory into consideration, including a group work component in the development of the stress management program is key, particularly because the anticipated majority of participants will be female. Additionally, previously stated findings suggest that technology-based solutions may reach women who may not otherwise seek or receive help for stress management.

The purpose of the research is to see if the combination of group support, self- service technology utilization in practicing mind-body practices can help decrease stress, depression, and anxiety levels among program participants?

Conceptual Framework

Behavioral theory states that people adapt to their environment through conditioning (Urdang, 2008). Behavioral techniques are still widely used in therapeutic settings to help clients leam new skills, reframe thoughts and change behaviors.

Through this research, the practice of mind-body techniques, as well as the

psychoeducation in the group sessions will provide an opportunity for participants to modify their thoughts and behaviors.

This research helps participants increase self-awareness by practicing mindfulness and relaxation techniques. When people experience an increase in self- awareness, they are more likely to change behaviors that do not align with their personal values. People who experience mindfulness become aware of their automatic thoughts and subsequent behaviors. Individuals develop through cognitive functioning and learn through acting in their environment. Behaviors are shaped through the experience of practice and exposure. Knowledge is constructed through experience. Participants’ thoughts can impact their state of relaxation and impact behavior. It is also tme that our psychoemotional state of being impacts our thoughts and cognitive abilities.

Participants will be intentional about experiencing and dedicating time and energy to practicing mindfulness and relaxation techniques. Participants will increase their awareness of the feedback loops in their interactions with their environment. They will also become aware of the “trigger points” when they feel the need to react to their environment.

The group sessions provide an opportunity for participants to benefit not only on a personal level, but also as part of an academic community. Being part of a group has additional benefit to seeking help individually. Konopka (1963) defines group work as a method of social work that helps individuals to enhance their social functioning through purposeful group experiences. Group sessions help participants cope more effectively with personal, group and community problems (Harte, 1999-2001). Setting up a series of stress management group sessions for the Associate nursing students should not only help them decrease their feelings of stress, but also help establish a community of support among fellow students. Group work provides an opportunity for individuals to help themselves, as well as each other. It can influence personal, group, organizational and community problems (Harte, 1999-2001).

One advantage of support group therapy is in helping a participant realize that he or she is not alone, and that there is some level of shared experience amongst those within the group. Being in a support group can also help you develop new skills to relate to others. The dynamics of a group often mirror those of society in general, and learning how to interact with the other members can help one in relationships outside of therapy. In addition, the members of the group who have the same problems can support each other, and may suggest new ways of dealing with a particular problem.

Powell (2012) cites the National Center for Complementary and Alternative Medicine of the National Institutes of Health, which classifies Complementary and Alternative Medicine (CAM) therapies into five categories, one of which is mind-body interventions. Mind-body interventions consist of a variety of techniques used to enhance the mind’s capacity to affect the body. The integration of mind-body practices as a means to reduce stress, depression and anxiety levels are a core concept of this research. By combining mind-body therapies with traditional mental health practices, such as psychoeducation and group work, it can provide additional benefits to the participants. Mind-body practices support health and well-being beyond just the physical. Integrating mindfulness and spiritual practices into one’s life can positively impact both emotional and spiritual aspects of well-being as well. CAM therapies acknowledge that our brain can impact our body and how it functions, and our body can impact our brain. Likewise, systems theory addresses the relationship between ourselves and our environment.

Systems theory is a fundamental theory on which the research is based. System theory acknowledges that there is ongoing interaction between a person and their environment. What impacts one part of the system affects the whole, as well as the other subsystems. Therefore, when a participant integrates a new practice into their lives, such as incorporating mind-body techniques, it affects the student as well as all other parts of the system such as their social and academic systems. All systems within an individual’s life are interrelated and part of a whole. This research focuses on how an individual is acting within their environment. Through this research, we will evaluate what effect, if any, the practice of mind-body techniques has on the individuals stress, depression and anxiety levels (Urdang, 2008).

Program Description

The goals of the group session are threefold: (1) The sessions will provide psychoeducation pertaining to stress, anxiety and depression, (2) it will engage the participants in a variety of stress management techniques including meditation, guided imagery and breathwork, and (3) the group process will aid in supporting the participants in managing stress their stress.

There is very little risk involved with participating in the project. The hope is that participants will be able to use some relaxation techniques that they find helpful to them both personally and professionally. However, practicing mind-body techniques can impact people in ways they were not expecting. Mind-body practices can increase one’s

awareness of thoughts, feeling and actions, and as a result, it can elicit an emotional response. In preparation to any potential adverse effects, mental health resources, both on and off campus, will be provided to participants during the first group session.

The stress management program was conducted on the Minneapolis campus of St. Catherine University with Associate degree nursing students. Participants were recruited via email and flyers posted in public areas on campus. The project consisted of three, 1-hour sessions occurring weekly during the month of February. Students were required to complete an intake form. Interested participants did this when they met with the facilitator. During this time the facilitator provided a brief overview of the stress management sessions, including: (1) The purpose of the pilot project, (2) the receipt of psychoeducation regarding stress, depression, anxiety and specifically test anxiety (3) the instmment/s to be utilized (DASS survey instmment, podcasts, apps), and (4) an explanation of the relaxation technique/s selected. During the second session, there was further discussion of how the relaxation techniques help alleviate stress-related symptoms. At the third and final session, the participants will practice some relaxation techniques and discuss movement meditations and yoga as a means for stress reduction. At the end of the session the DASS-21 (See Appendix C: DASS-21 Form) and the exit interview were given to participants to complete.

Students experienced guided imagery, meditation, and breathwork to help build coping skills. Participants will be expected to access, practice and record their utilization of self-service technologies during the period between the first and third group sessions.



Research Design

Nursing students participated in three group sessions over the course of a 3-week period. Participants received psychoeducation regarding stress, anxiety and depression, as well as practiced relaxation techniques via self-service technologies. The study used a mixed-method design, using both quantitative and qualitative approaches. The study included group work as a means of using the shared experience of being in nursing school and to help reduce feelings of social isolation among the participants. The study also required independent practice of mindfulness and relaxation practices. Participants were asked to practice as much as they prefer and for how long, however, participants were asked to practice a minimum of three times a week, for a minimum of three minutes each time using podcast, apps and other self-service technologies as their medium.

Participants attended an intake interview. During this time, potential participants were given an overview of the research study, and required to sign the informed consent form (See Appendix A: Informed Consent Form). Participants were asked to fill out an intake form. The intake form captures demographic information of the participants, as well as some general background regarding their general health and wellness.

The DASS-21 scale was used as the primary basis for determining if stress, anxiety, and depression levels of the participants were impacted as a result of participating in the stress management program. (See Appendix B: DASS-21 Scale) Participants completed the DASS-21 during the first group session, and again at the end of the third group session. Individual comparison as well as group comparison on the

DASS-21 scores will indicate what impact, if any, occurred as a result of the group sessions, and relaxation practices (See Appendix C: DASS-21 Scoring Form). The demographic information collected from the Intake Form and DASS-21 scores were analyzed to examine if any correlations exist (See Appendix D: Intake Form).

Throughout the 3-week period, participants were asked to practice relaxation techniques using self-service technologies. They were asked to maintain a practice log (See Appendix E: Practice Log Form). Again, the goal for frequency and length of time was to practice a minimum of three times a week, for a minimum of three minutes each time. The participant was exposed to a variety of practices, such as breathwork, autogenics, mindfulness mediation, and guided imagery. The participants were able to choose which practices they would prefer among a variety of options (See Appendix F: List of Podcast, Apps and Websites). This log was used as a means to gauge level of participation and compliance with stated expectation.

Lastly, participants were asked to fill out a three-question exit interview (See Appendix G: Exit Survey). These questions are open-ended. The intention of the exit interview was to allow participants an opportunity to reflect on their experience of the stress management series and to report what benefits, if any, they received from their participation in the stress management program and the use of technology.


Participants were recruited through email and from posted flyers on the St. Catherine University’s Minneapolis campus. Students were given a $10 Target gift card for each group session the participants attended. The group size of this research project


began with seven associate nursing students. Four of the students were in their first year of the nursing program and three were in their second and final year. All participants were women. Of the seven who began the program, six successfully completed all aspects of the program. Of the six who completed the program, half of the students had previously failed a nursing class during their time at St. Catherine.

Protection of Human Subjects

All written information was kept confidential. All paperwork was stored in a locked file cabinet, in an office that was also locked when not being occupied. Information of the participants was coded and stored on researcher’s personal laptop, which was also password protected. The data collected was explicitly used for the purpose of this research. The researcher is a licensed social worker, and abides by the ethical standards set by the profession. All raw data collected will be confidentially destroyed as of May 30th, 2012.

Recruitment Process & Agency and institutional Support

Approval was garnered from University of St. Thomas and St. Catherine University Institutional Review Boards (Ш.В). In addition, letters of support were received from the Dean of Associate Healthcare programs at St. Catherine University, as well as from the Director of the associate nursing program that manages the program from which the participants were solicited. Emails were sent to all Associate nursing students. Lastly, flyers were posted on the Minneapolis campus on which the research was conducted.


Data Collection Instrument and Process

Participants filled out an intake form regarding basic demographic data in addition to some variables that address some of their general health and well-being. Participants were asked to fill out a 21-question scale that assesses stress, depression and anxiety levels. This inventory is known as the DASS-21 (Depression, Anxiety, and Stress Scale). The participants completed this form as a pre-test during the first group session, and completed the scale again as a post-test at the end of the third group session. The pre-test and post-test scores were compared between each individual as well as in the aggregate. In between the group sessions, participants were asked to keep a practice log of the types of relaxation practice they use and to record the length of time. Students were asked to practice at least 3 relaxation techniques each week, for a minimum of three minutes each time. The participants turned in their practice log at the end of the third session. Additionally, each participant was asked to take an exit interview that consisted of the few questions to acquire some basic feedback regarding the participant’s experience of the program.

Data Analysis Plan

A key question of the research is the difference in stress, anxiety and depression levels between the pre-tests and post-tests. It’s important to not only examine individual scores, but also those in the aggregate. Once the pre-test and post-test data was known, it was important to examine if any independent variables acquired from the participant’s intake interview, such as race or year in nursing, showed any correlation with stress, anxiety and depressions scores.

Additionally, it is important to examine whether the frequency of practice using self-service technologies, as well as total length of time, had any obvious impact on stress, depression and anxiety levels.

Finally, the exit interview will be important in gauging the participants’ perceptions on their experience of the stress management sessions, the use of technology and what benefits, if any, they report from their personal experience.


Demographic Data

Six females participated in the stress management program and attended all three, group sessions. Three of the participants identified their race as being Caucasian, two as Black, and one as African American. Two participants were in their first year of their nursing program, and four were second year nursing students. Half of the students had previously failed nursing, and as a result, needed to resequence. Students are required to resequence after failing a nursing exam, which means the student is required to petition the program to be readmitted to the nursing program, then retake the nursing class that the student had previously failed. The nursing program at St. Catherine University permanently dismisses students who have scored below 78% more than once during the time they are taking nursing program classes.

In regards to general health habits, three out of six believed they were eating nutritiously. Three participants report getting adequate sleep, whereas two reported they did not, and one reported only sometimes. None of the students reported having regular exercise. All students reported having an adequate social support system.

DASS-21 Results

The DASS-21 form is made up of 21 statements that participants answered on a Likert scale. The DASS-21 was used as a pre-test and a post-test that participants fill-out at the beginning and end of the group sessions. Seven participants started the program and six of the seven finished the program.

Table 1 shows the sum of the scores related to anxiety, stress and depression between their pre-test and post-tests, along with the standard deviation. The scores show a decrease between pre-test and post-test scores for both anxiety and stress, however the standard deviation is large enough that there is no statistical significance.

Table 1. Average PASS Scores Between Pretests and Posttests

  Pretest Stan. Dev. Posttest Stan. Dev. Difference
Anxiety 6.5 4.8 3.67 2.3 2.83
Stress 7.5 2.4 5.5 3.8 2
Depression 4.2 2.8 4.2 3.7 0

There was a significant range in scores which accounts for the reason why the high standard deviation. Upon closer examination, the high scores correlated with the participants who had previously failed a nursing class, thus requiring them to resequence. Table 2 shows the resequenced participants (A-l, A-5, and A-6) and their high DASS score, where it was scored (pre-test and post-test), under which category (stress, anxiety or depression) compared to the three participants who had not resequenced (A-2, A-3, and A-4) by combining their scores to get an average score, as well as the standard deviation. There is a remarkable difference between the scores, and the standard deviation indicates that the group of students

who had not resequenced had scores in close proximity to one another with the exception of the post-test for the stress sub-scale.

Table 2. Resequenced Participant’s Scores vs. Non-Resequenced Students



DASS high score Pretest or Posttest Average score of participants who have not resequenced [A-2, A-3, and A-4] Standard


A-l 9 Pre Anxiety 4.3 1.5
A-5 15 Pre Anxiety 4.3 1.5
A-5 10 Pre Stress 6 1
A-6 11 Pre Stress 6 1
A-6 11 Post Stress 4.7 4
A-6 10 Post Depress 3.7 2.5

Although the participants who had resequenced had the highest DASS scores, they also demonstrated the greatest decreases in their anxiety and stress scores between the pre-tests and post-tests, as compared to the other students. In Table 3, participant A-5 made the greatest statistical change between their anxiety and stress pre-test and post-test scores.

Table 3. Resequenced Participants Score Differences


Pre high score Post Score Difference
A-l [anxiety] 9 6 -3U
A-5 [anxiety] 15 3 -12U
A-5 [stress] 10 3 -7-U-
A-6 [stress] 11 11 0



Table 4, A-l through A-6 are coded to represent the six participants that complete the stress management program. The numbers indicate the difference in scores of the individual participants between their pre-test and post-tests by category (stress, anxiety, depression]. The downward facing arrows indicate decreases in their scores, which is the desired outcome.

Table 4. PASS Scoring of Individual Participants by Sub-Category



Difference between pre/post stress score Difference between pre/post anxiety score Difference between pre/post depression score
A-l -lU -3U 1
A-2 -4U -3U -lU
A-3 -3U 0 1
A-4 3 -3U -3U
A-5 -7-U- -12U -3U
A-6 0 4 5

In Table 4, four of the six participants showed decreases in their stress scores, as well as in their anxiety scores. Three of the six participants showed decreased in their depression scores. In Table 5, it shows which DASS items showed a decrease in scores among the group as a whole, by comparing participant’s scores between their pre-tests and post-tests.

Table 5 accounts for the DASS items that showed a decrease in score. For example, the group of participants scored a total of 8 on the DASS pretest for item number 8, which stated «I felt I was using a lot of nervous energy». The total score on the post-test, and was five, for a difference of three. DASS #8, as well as DASS #1 demonstrated the largest difference in scores. DASS #1 being «I found it hard to wind down».


Table 5. DASS Items Showing Decreased Scores

DASS Item Category Pre-test




Difference in Score

I felt I was using a lot of nervous energy

Stress 8 5 -3U

I found it hard to wind down

Stress 6 3 -3U

I found it difficult to relax

Stress 7 5 -2U

I tended to over-react to situations

Stress 5 4 -1U

I found myself getting agitated

Stress 6 5 -1U

I was intolerant of anything that kept me from getting on with what I was doing

Stress 3 2 -1U

I felt that I was rather touchy

Stress 4 3 -1U

I experienced trembling (e.g. in the hands)

Anxiety 3 1 -2U

I was aware of the action of my heart in the absence of physical exertion (e.g. sense of heart rate increase, heart missing a beat)

Anxiety 5 3 -2U

I was worried about situations in which I might panic and make a fool of myself

Anxiety 5 4 -1U

I felt I was close to panic

Anxiety 4 3 -1U

I felt scared without any good reason

Anxiety 3 2 -1U

I couldn’t seem to experience any positive feelings at all

Depression 3 2 -1U
#17 Depression 3 2 -1ft
I felt I wasn’t worth much as a        

It is also important to note in Table 5, that although the largest differences were seen in the stress and anxiety sub-categories, participants on the pretests also rated them the higher. Whereas, the depression pre-test scores were low to begin with. Table 6, illustrates the DASS items that demonstrated an increase in scores. Three of the 4 DASS items were in the depression sub-category.

Table 6. DASS Items Showing Increased Scores    
DASS Item Category Pre-test




Difference in Score

I felt I had nothing to look forward to

Depression 0 2 2ft

I was unable to become enthusiastic about anything

Depression 2 3 1ft

I felt that life was meaningless

Depression 1 2 1ft

I was aware of dryness in my mouth

Anxiety 1 2 1ft

Table 7 shows the DASS items that were most frequently identified by participants and were scored highly. For example, on DASS item number twelve, “I found it difficult to relax”, seven of seven participants identified this to some degree as having experienced. The total score for this DASS item is a total score of eight.

Table 7. Frequency Distribution of DASS Items

DASS Item DASS SubCategory Most frequently scored DASS-21 line item (n=7) Highest scored DASS Item
#8 Stress 7 8


I felt that I was using a lot of nervous energy      

I found it difficult to relax

Stress 7 8

I was worried about situations in which I might panic and make a fool of myself

Anxiety 5 8

I found it difficult to work up the initiative to do things

Depression 7 8

I felt down-hearted and blue

Depression 5 8
*Numbers based on a N=7 of pre-tests, a N=6 for post-test

Table 8 shows how frequently each participant reported practicing a relaxation technique along with the total length of time spent practicing. Results showed that there was a vast discrepancy between the time participants logged as practicing. Four out of the six participants met the minimum requirement for a total of three times a week, and all six participants met the time requirement.

Table 8. Data from Participant’s Practice Log

Participants Number of Times Spent Practicing Number of Minutes Participants Practiced Relaxation Techniques using Self-Service Technologies
A-1 4 21
A-2 3 44
A-3 8 19
A-4 7 207
A-5 7 105
A-6 9 29

*Students practiced for a total of 2 weeks



Qualitative Results

All six participants acknowledged some benefit from participating the program. The common themes between the respondents was that they were better able to recognize when they were feeling stress and to what extent. They reported they found new ways of coping and managing their stress through the exposure of different mindfulness practices. Lastly, they acknowledged technology as resource for helping practice relaxation techniques. DASS scores did show a decrease in stress, anxiety and depression though it is not significantly significant. (See Table 1)

All participants stated they intend to continue practicing stress management techniques. A majority of the students expressed gratitude for the opportunity to participate in the stress management sessions. Several students commented on the need for these practices to be integrated into the nursing program. A couple students expressed a strong desire to continue the practice at home and integrate into family life.


It is important to note that inferential statistics are unable to be used as a result of small sample size. Of the participants who attended each of the sessions, and practiced relaxation techniques using self-service technologies, six students completed the requirements of the study. Of the students who participated, Table 1 showed that the group demonstrated a decrease in stress and anxiety scores between the pretest and posttests from the DASS-21. Depression levels remained the same. Due to the fact that half of the participants had some high stress, anxiety and depression scores, it resulted in a high standard deviation among the group data. Despite this fact, it is important to note that the participants who had the highest scores, all were students who had failed a nursing class in the past and subsequently resequenced. Of the students who had resequenced, these students scored higher stress, anxiety and depression levels than those who have not resequenced. Table 2 shows the resequenced students and line items in which they scored highest on. This table compared these individual high scores against the aggregate scores of the students who had not failed a nursing class. The table demonstrates a significant difference between the scores. As a result, it demonstrates that students who have resequenced are more likely to have high stress, anxiety and depression scores.

Of the participants who had not resequenced, one participant (A-4) did demonstrate a few marginally high numbers. This participant scored a nine on the pretest on the depression scale, as well as scoring a nine on the post-test, stress sub-scale. It is important to note this student did report during the week preceding the final stress management session, that she had experienced a traumatic event and was unable to practice any relaxation techniques for several days.

The participants experienced the greatest reduction of the following stress and anxiety symptoms from participation in the stress management program. Table 7 showed the DASS line items displaying the biggest decrease among the participants in comparing pre-tests and post-tests were related to stress and anxiety. Specifically, participants reported: 1) Experiencing less nervous energy, 2) finding it easier to wind down, 3) less trembling, and 4) less awareness of the action of their heart in absence of physical exertion.

The depression sub-scale results did not show the expected decreases, as well as stress and anxiety sub-scale results did. Table 5 highlights the DASS-21 items that showed decreased levels between the pre-tests and post-tests. Where as, Table 6 shows the DASS-21 items that had increased between pre-test and post-test. It is important to explore a possible reason why only depression scores increased. Could it be possible that participants in becoming more aware of their stress demonstrate higher depression scores? It is important to acknowledge that on Table 6, the depression scores were notably lower on the pre-tests, compared to the pretest scores on the stress and anxiety scores highlighted in Table 5.

There are limitations in western psychotherapy and the paradigm and academic training programs. There is often a focus on helping clients managing external forces rather than tmsting one’s self to help heal oneself and conventional psychological programs are slow to change curriculum. The perceptions of healing oneself versus fixing one’s suffering because something (externally) is getting in the way of the healing process is one of the fundamental differences between mindfulness practices and psychotherapy. Integrative healing practices often draw from the position that we are allowing ourselves the opportunity to do what our body naturally knows to do.

It’s important to acknowledge that this program was developed uniquely developed to meet the needs of women, in nursing school. This program emerged as a result of the literature review. The three key elements include: 1) practicing a variety of relaxation and mindfulness techniques 2) practicing these techniques through the use of self-service technologies, and 3) participating in group sessions where psychoeducation regarding stress, depression and anxiety was provided. To date there is not currently a program set up to meet the needs of this specific population. Due to the small sample size, and due to the fact that the research is set up as a pilot project, it is important to continue to evaluate how these components, working together, aid in helps lower stress, anxiety and depression levels in college students. In this research, it is impossible to discern if the positive trends of the data are related to the use of podcasts, the support of the group process or the mindfulness practice itself.

All six participants acknowledged benefits from participating the program. Notably, several reported being better able to recognize when they were feeling stress and knowing new ways of coping and managing their stress through the exposure of different mindfulness practices. This is in accordance with Shapiro’s (2000) metaanalysis of stress management programs in higher education, which concluded that students felt positively regarding their participation in stress management programs. Additionally, these students specifically recommended that this program become integrated as part of the Associate nursing program at St. Catherine University.

Strengths and Limitations

One of the strengths of the research is the mixed-method design. It provides an opportunity for the collection of evidence to be reported in multiple ways, which can add to the validity of the research. Using quantitative and qualitative data together has a complementary effect. This provides stronger evidence for a conclusion through the convergence and corroboration of findings. If the results of the qualitative and quantitative indicate similar outcomes, triangulation occurs and helps validate results. This can add insights and understanding that may be missed when only a single method is used. Using both methods together can increase the generalizability of the results.

Additionally, this approach produces more complete knowledge necessary to inform theory and practice.

The following can be seen as both a strength and weaknesses of the research study, which is the level of subjectivity and variation between participants in their practice of relaxation techniques, as well as the time commitment spent practicing. The participants have some flexibility in deciding which type of self-service technologies to use, frequency of use and length of time. The research design only used a minimum number of three times a week, for a minimum of three minutes. Due to the relative lack of structured time and practice, there is the potential for wide variation between participants, making it difficult to draw conclusions. However, the requirement of three times a week and for a minimum of three minutes is not too much of a burden, time wise. Additionally, it is difficult to see of certain practices, such as breathwork were more efficacious than other types of relaxation practices. It is also challenging to decipher if an app is more effective than a podcast. The variations between length of time, types of practice and technological medium are impossible to discern, particularly with a small sample size.

Another limitation of the study has to do with the fact that the stress management sessions are not designed within the nursing program. Aligning the information and practices to coincide with the curriculum and timing of quizzes, exams and clinical experiences could be a more effective integration of relaxation practices and psychoeducational content. Extending the duration of the stress management program may show better results and be more supportive of helping student integrate relaxation techniques. Lastly, the stress management sessions began and ended during the nursing

semester, which was not in sync with the ebb and flow of the nursing exams and clinical experiences.

The number of participants was less than ten. It is not a sufficient number to run inferential statistics using the quantitative data. Therefore, statistics can only be reported descriptively. The data is only indicative of trends. Any associations between variables are not well grounded and cannot infer direct correlations between independent and dependent variables.

Implications for Practice

Integrative practices such as breathwork and mindfulness have been shown to be an effective approach to managing stress and helping managing symptoms of depression and anxiety. Mind-body techniques help facilitate healing and complements western mental health practices. Continued integration of mind-body practices into counseling has several therapeutic benefits.

The potential of using technology as a means to help facilitate health and wellbeing has yet to be explored to it’s fullest. Technologies, such as podcasts and apps are both a cost-effective, and accessible means for providing opportunities for health education and health practices. However, its benefits compared to traditional methods are unclear. Supplementing the use of self-service technologies as part of a more traditional practice, like group work may provide adequate support. For women in particular, the need to connect socially may be a critical part of their healing process.

Examining the integration of alternative therapies, in combination with traditional modes of therapy and the use of technology requires careful consideration and further experimentation. Combining these practices has the potential help create health changes that are both accessible and affordable to people.

Participants reported several benefits from the program. Participants suggested that the stress management program become an integrated part of the nursing program at St. Catherine University. Stress management programs in higher education support acquiring stress management skills to students entering professions where people’s lives depend on them.

Implications for Research

This pilot program has shown reasonable evidence that it is a worthwhile endeavor to continue to provide students. Further data collections are necessary in order to further collect and analyze data with a greater total number of participants. This would further assist in discovering the efficacy of the program, as well as substantiate whether there are correlations between variables.

One of the strengths of the research is the mixed-method design. It provides an opportunity for the collection of evidence to be reported in multiple ways, which can add to the validity of the research. Using quantitative and qualitative data together has a complementary effect. This provides stronger evidence for a conclusion through the convergence and corroboration of findings. If the results of the qualitative and quantitative indicate similar outcomes, triangulation occurs and helps validate results. This can add insights and understanding that may be missed when only a single method is used. Using both methods together can increase the generalizability of the results. Additionally, this approach produces more complete knowledge necessary to inform theory and practice. In addition to the mixed-method design it is also important to do research that combines participant in group sessions, using integrative relaxing practices with self-service technologies as the medium in which they are practiced.

It is also important to assess the potential health habits, and environmental stressors that may be affecting participants in stress management program. In this research, the variable of whether the students had previously failed a nursing class seems to have had a significant effect on the mental health of the participants. One questions I would add to future studies, particularly at St. Catherine University would be to assess participants time spend participating in faith practices. St. Catherine University is a Catholic college, and should have been assess on the intake form of participants, and this may impact DASS-21 scores.

Further suggestions for future research on this subject would be to examine whether certain types of relaxation practices are more effective than others. Is there a time or frequency threshold that is most effective?

Lastly, although none of the participants directly mentioned the benefit of the group work and social support between the participants. It would be important in future studies to capture participant’s feelings regarding the group process.



Appendix A: Informed Consent

Appendix B: DASS-21 Scale (Pre-test/Posttest)

Appendix C: DASS-21 Scoring)

Appendix D: Intake Form Appendix E: Practice Log

Appendix F: List of Podcasts, Apps and Websites Appendix G: Exit Survey


Appendix A: Informed Consent

Using self-service technology for stress management: A pilot project RESEARCH INFORMATION AND CONSENT FORM


You are invited to participate in a research study investigating the use of technology and psychoeducation in helping manage stress, depression and anxiety levels. This study is being conducted by Carissa Morris, LSW at St. Catherine University. You were selected as a possible participant in this research because you are currently enrolled in the associate nursing program at St Catherine University. Please read this form and ask questions before you decide whether to participate in the study.

Background Information:

The purpose of this study is to provide education on stress, anxiety and depression and for participants to utilize self-service technologies in helping nursing students manage stress. No more than 12 nursing students are expected to participate in this research.


When a potential participant makes contact with researcher regarding their interest in participating in the research project, the researcher will verify that the nursing student has not had more than 3 substantive one-to-one meetings with the principal investigator of the research project through her role as staff with the Access and Success program at St Catherine University. The principal investigator will verify this by looking up client records to access potential history as a client in professional role. If substantive contact has been made (3 or more one-to-one contacts], the nursing student will be informed that they are ineligible to participate and referrals to mental health resources will be made. Those who are eligible to participate will be given an informed consent form outlining the extent of the research. Principal researcher will be ask questions to ensure their understanding of the research as well as ask if he/she had any further questions regarding the research or their participation in it Participants will be enrolled in the research on a first come, first serve basis, until the maximum of 12 participants is reached.

In order for a nursing student to take part in the research, he/she will need to set up an initial meeting with the researcher. The researcher will have mental health resources available during this meeting.

If you decide to participate, you will be asked to participate in an intake interview, attend the 3 group session, currently scheduled for Feb. 10th, 17th and 24th 2012. As a participant in the stress management program, you will need to attend three, 1-hour group sessions occurring in February 2012, as well as practice and record a minimum of 3 relaxation exercises a week, throughout the duration of the program. The first session will be held Friday, Feb. 10th, second session on Feb. 17th and the final session on Feb. 24th . I am limiting the number of participants to no more than 12 associate-level nursing students.

The goals of the group session are threefold, 1] the sessions will provide psychoeducation pertaining to stress, anxiety and depression. 2) It will engage the participants in a variety of stress management techniques including meditation, guided imagery and breathwork, 3) and the group process will aid in supporting the participants in managing stress their stress.

«Homework» will be assigned in-between group sessions. Satisfactory completion of this «homework» involves using technology as a means to practice stress relaxation techniques. Technology may include using a computer, smartphone, iPad or similar equipment Students will use online videos, podcasts or apps that provide guided meditation, guided imagery or breathwork. Participants will be expected to engage in the stress management technique for a minimum of 3 minutes, 3 times a week. As the facilitator, I will provide you with several options from which to choose from.

Because it is a research project, I will be collecting some very basic information about you before beginning. All participants will be asked to answer a 21-question stress, anxiety and depression scale both before and after participating in the program. The initial DASS- 21 will be filled out during the 1st group session. The final DASS-21 will be completed during the 3rd session. All participants will be provided with mental health resources both on campus, and in the community at each group session. During the 1st and 3rd group sessions, when the participants complete the DASS-21, they will be given information on how to evaluate how they scored. Participants will not be asked to share their results publicly. The researcher will use the DASS-21 as a reference while providing psychoeducation on the stress response, depression and anxiety, in particular among college students. Participants will also be practicing stress management skills during the group sessions. Participants will be asked to write down the dates and length of time they practice their «homework» and to answer a 3 question, exit interview during the 3rd and final group session. All identifying information will be seen by me and me only. Confidentiality statements will be made at the beginning of each group session. A confidentiality statement will be made by the researcher assuring that all information discussed in the group will be held confidentially by the facilitator and it is also expected among the participants. Once the data is collected, it will be publicly shared, but no individual names will be identified.

This study will take no more than 5 hours total. Attending the 3 group sessions will take 3 hours. The initial intake interview will likely take approximately 30 minutes. Students will be asked to practice meditation and/or guided imagery outside of scheduled meetings that must be a minimum of 9 minutes each week for a total of 3 weeks.

Risks and Benefits:

The study has minimal risks. The hope is that participants will be able to use some relaxation techniques that they find helpful to them both personally and professionally. However, those who practice mind-body techniques may gain some personal insight or awareness that they were not expecting and it can elicit an emotional response. If that were to happen, I could assist you in referring you for additional professsional support

Participation in stress management classes are completely voluntary.

Although a confidentiality statement will be made at each sesssion, it is important to note that the researcher can not guarantee what other participants may do in regards to maintaining confidentiality.

The only direct benefit participants may receive by participating in the study is the $50 Target gift card. Of which, 2 will be raffled off to those who attended all 3 group sessions.


Any information obtained in connection with this research study that could identify you will be kept confidential. In any written reports or publications, no one will be identified or identifiable and only group data will be presented. (If it applies to your study, include ways in which you will maintain confidentiality, e.g., «No one in the daycare center will know your child’s results.” If you release information to anyone for any reason, you must state the persons or agencies to whom the information will be furnished, the nature of the information to be furnished, and the purpose of the disclosure.)

I will keep the research results in a password protected computer and a locked file cabinet in researchers home and only I and my advisor will have access to the records while I work on this project I will finish analyzing the data by May 31st, 2012. I will then destroy all original reports and identifying information that can be linked back to you.

Voluntary nature of the study:

Participation in this research study is voluntary. Your decision whether or not to participate will not affect your future relations with St. Catherine University in any way. If you decide to participate, you are free to stop at any time without affecting these relationships, and no further data will be collected.

Contacts and questions:

If you have any questions, please feel free to contact me, Carissa Morris, at 651-497-8518. You may ask questions now, or if you have any additional questions later, the faculty advisor, Philip Auclaire, PhD 612-752-8181, or I will be happy to answer them. If you have other questions or concerns regarding the study and would like to talk to someone other than the researcher(s), you may also contact John Schmitt, PhD, Chair of the St. Catherine University Institutional Review Board, at (651) 690-7739.

You may keep a copy of this form for your records.

Statement of Consent:

You are making a decision whether or not to participate. Your signature indicates that you have read this information and your questions have been answered. Even after signing this form, please know that you may withdraw from the study at any time and no further data will be collected.

Signature of Participant


Signature of Parent, Legal Guardian, or Witness Date

(if applicable, otherwise delete this line)


Signature of Researcher


Appendix В: DASS-21 Scale (Pre-test & Post-test)

DASS21 Name: Date:      
Please read each statement and circle a number 0, 1, 2 or 3, which indicates how much the statement applied to you over the past week. There are no right or wrong answers. Do not spend too much time on any statement.

The rating scale is as follows:

0 Did not apply to me at all

1 Applied to me to some degree, or some of the time

2 Applied to me to a considerable degree, or a good part of time

3 Applied to me very much, or most of the time

1 I found it hard to wind down 0 1 2 3
2 I was aware of dryness of my mouth 0 1 2 3
3 I couldn’t seem to experience any positive feeling at all 0 1 2 3
4 I experienced breathing difficulty (e.g., excessively rapid breathing, breathlessness in the absence of physical exertion) 0 1 2 3
5 I found it difficult to work up the initiative to do things 0 1 2 3
6 I tended to over-react to situations 0 1 2 3
7 I experienced trembling (egg, in the hands) 0 1 2 3
8 I felt that I was using a lot of nervous energy 0 1 2 3
9 I was worried about situations in which I might panic and make a fool of myself 0 1 2 3
10 I felt that I had nothing to look forward to 0 1 2 3
11 I found myself getting agitated 0 1 2 3
12 I found it difficult to relax 0 1 2 3
13 I felt down-hearted and blue 0 1 2 3
14 I was intolerant of anything that kept me from getting on with what I was doing 0 1 2 3
15 I felt I was close to panic 0 1 2 3
16 I was unable to become enthusiastic about anything 0 1 2 3
17 I felt I wasn’t worth much as a person 0 1 2 3
18 I felt that I was rather touchy 0 1 2 3
19 I was aware of the action of my heart in the absence of physical exertion (e.g., sense of heart rate increase, heart missing a beat) 0 1 2 3
20 I felt scared without any good reason 0 1 2 3
21 I felt that life was meaningless 0 1 2 3


Pre-packaged guided imagery for stress reduction: Initial results

Elizabeth Carter Registered Psychologist

Author Contact:

E.D.Carter, P.O. Box 224, Sans Souci, NSW, 2219


A study into the use of pre-packaged compact discs (CDs) which incorporate Guided Imagery (GI) with suggestions and affirmations, indicates that the use of these CDs results in quick reduction of stress-related issues for all participants. Notable improvements were identified in general feelings of well-being (91% of participants), positive thoughts (82%) and ability to cope in stressful situations (73%). Decreases in incidence ratings were greatest for insomnia, anger and negative thoughts. Most commonly the first benefits people noticed were increased relaxation, decreased negative thoughts, and decreased stress. Despite the effectiveness of the CDs, improvements tended to be short-lived due to diminishing use of CDs (and relaxation techniques) over time. There was no long term commitment to the regular practice of the relaxation technique. Rather participants apparently sought speedy issue resolution from the CDs. Consequently not all improvements endured, with some participants needing to revisit the CDs each time symptoms returned.


Guided imagery; relaxation; stress reduction; self-hypnosis; well-being; positive thoughts; compact discs (CDs)



Many people suffer symptoms of stress manifesting in a range of physical symptoms, of different levels of severity, such as headaches, bodily aches and pains, anxiety, and inability to cope with life generally (Jackson: 1993). However, faced with these symptoms, what do we do? Initially, we may attempt to heal ourselves by taking self-prescribed medication (analgesic, tonic, multivitamin), improving our diet, decreasing alcohol intake, etc. and/or we may visit our general practitioner (GP). All these options involve treating the body rather than addressing the root cause, the stress, or, the state of the mind.

There appears no question that stress-related symptoms are real, and can be caused by an emotional imbalance, rather than a physical one (Clarke & Smith:2000), which is not easily treated in an effective manner within a traditional Western medical model based on biomedical science, which is evidence-based as well as being fundamentally analytical and reductive. The wide acceptance of this fact is evidenced by the increase in the availability of alternative therapies, eg relaxation techniques, yoga, Pilates, massage therapy, and self- hypnosis, in many forms and mediums such as therapy sessions, classes, videos, cassettes, CDs, DVDs etc.

The fundamental element in any therapy aimed at reducing stress and stress related symptoms is relaxation, and relaxation has been shown to be beneficial for a range of symptoms and health problems such as shorter hospital stays; decreases in pain along with the promotion of more rapid recovery (Blankfield: 1991); quicker wound healing (Kiecolt- Glaser:2001); reduced depression, headaches, fatigue, and improved job satisfaction/performance; better relationships with supervisors/co-workers; and improved quality of sleep (Bellarosa:1997). Benefits also noted include reduction of nervousness; promotion of performance; enhanced selective attention and memory recall; promotion of self-control/self-actualisation and reductions in psychosomatic complaints (Krampen:1996). These techniques have been seen to assist in, (i) alleviation of insomnia and gastrointestinal disorders (McCubbin:1996); (ii) anger reduction (Deffenbacher:1995); (iii) decreases in irrational beliefs and increases of self-efficacy, (iv) reduction in hot and cold spells, along with (v) tension headaches, insomnia, anxiety, stress and excessive fidgeting (Kiselica:1994). Widespread benefits are noted in holistic terms, and may positively influence (i) improvements in mixed psycho-physiologic disorders (Blanchard: 1991); (ii) reduced negative outcomes associated with ageing, (iii) reduced hospitalisation and admission rates; (iv) improved mental health and (v) increases in longevity (Alexander, Langer, Newman Chandler & Davies: 1989). In addition to providing assistance for people with cancer and other terminal illness, stress-reduction approaches are known to generally help people cope with illness and chronic pain (Cupal & Brewer:2001).

The use of imagery and guided imagery (GI) in psychotherapy to induce relaxation is widely adopted and accepted (Ackerman & Turhoshi:2000; Bazzo & Moeller: 1999; Complementary & Alternative Medicine:Guided Imagery:2004; Eller: 1999; Gruzelier:2002; Hudetz, Hudetz, & Klayman:2000; Johnston:2000; Peck, Bray & Kehle:2003; Rossman:2000; Shames 1996; Syrjala & Abrams:2002; and Varlas:2001). Guided Imagery is a therapeutic technique allowing individuals to use their own imagination to connect body and mind to achieve desirable outcomes (Ackerman & Turkoski:2000). GI induces a relaxed state and facilitates cognitive restructuring when ‘suggestion’ is included as part of the therapy (Syrjala & Abrams:2002). There is debate however that imagery and the associated suggestions are most beneficial when their utilisation is individualised. Evidence for and against individualisation has had mixed outcomes with a key element of success being mastery or competence of imagery technique (Enns:2001; Syrjala & Abrams:2002).

Arguments against individualisation for some professionals has popularised mass produced therapies available as tapes, CDs etc., incorporating self-hypnosis techniques, imagery and GI, with the same induction and training being presented to everyone using the CD or tape. This type of easily accessible and relatively inexpensive therapy is becoming more popular and is also being recommended to clients in therapy, to help clients relax (Bourne, 2001; Blanchard: 1991).

The author conducted a research project on the effect of non-individualised GI CDs, or ‘prepackaged guided imagery” (PPGI) on life stress and stress related symptoms. The PPGI used in this instance was a series of CDs with associated written material (Guidebook), developed for a range of personal issues by practicing professionals (a psychotherapist and a peak performance coach). Although the CDs are for a range of issues, they all incorporate the same induction technique of relaxation training using GI followed by imagery incorporating suggestions and affirmations, i.e. 90% of the content of the CD’s was the same except for the focus of the affirmations and suggestions.

Theoretically, the development of such CD’s is in according with Jackson (1993), in that if there is an overall increase of positive thoughts through this technique and with additional positivity from affirmations in this instance, there should be a generalised increase of positive feelings and increase in good health. Thus, it is hypothesised that because of the broadness of the approach, regardless of the underlying issue, there will be an overall improvement across a range of issues.

The aims and objectives of the project were to assess by qualitative and quantitative methods, the following:

  • perceived role of non-individualised PPGI (i.e. readily available CDs)
  • expected outcomes versus actual outcomes of PPGI CDs
  • associated benefits after undertaking GI training, and to
  • identify any surprising or unexpected benefits.



From previous research already mentioned, a list of potentially stress-related issues and related questions was developed as the basis for a survey of people who had used the PPGI which incorporated specific issue-related suggestions and affirmations. Participants were obtained from a list compiled at the time of acquiring the product, of people willing to participate in a study relating to outcomes. Participants acquired at least one of the CDs from the set of 10 (titled respectively, ‘Absolute Relaxation; Achieving Sensational Grades; Attracting People Magnetically; Busting Away Depression; Creating Abundant Wealth; Discovering Past Lives; Quick & Easy Weight Loss; Sleeping Like a Baby; Stop Smoking Permanently; Meditating Easily’, Briggs & Green:2002). In all, 22 participants completed questionnaires either by phone or self-completion. The questionnaire consisted of open and closed questions. All participants received copies of the questions whilst engaging with the product.


Just over half of the participants were female 59%, and 41% were male. Participants ranged in ages from 19 to 67. Overall a wide range of CDs in the selection were used by the participants. The most commonly used CD was ‘Absolute Relaxation’ (27%), followed by ‘Creating Abundant Wealth’, and ‘Sleeping Like a Baby’ (18% each). Only 4 respondents had used more than one of the CDs individually. At the time of their interview, participants had begun using their CDs from between 2 to 14 months prior, with average usage period of

6.6 months.

From qualitative responses, 45% of participants were classified as being in “Good Health ”, describing themselves as “active”, “healthy”, “feeling great” etc., prior to starting the PPGI. The others (55%) were classified (by the author) as in “Poor Health”, reporting a range of health issues such as “depression”, “being overweight”, “lethargic”, “having poor eating habits”, “suffering from headaches”, “experiencing mood swings”, being “anxious”,

“weepy”, “uptight”, “out of balance”, “generally unwell”, or suffering from “mental stress”. All but one in this group reported multiple symptoms. Across all ages respondents reported health issues. Female participants reported more health issues than did males (see Table 1).


Table 1. Overall Percentages of Participants Age and Sex by Health Classification

  Overall Good


Poor Health/Multiple issues
Age Ranges        
Less than 30 yrs 18 9   9
30-39 yrs 32 14   18
40-49 yrs 27 14   14
50 or more yrs 23 9   14
Male 41 27   14
Female 59 18   41
Overall 100 45   55

* Percentages have been rounded.

Perceived role of PPGI

Almost all participants had high expectations of success, believing that the CD would work. The reasons given for taking up the PPGI could be classified into 3 broad categories:

  • a specific/identifiable problem targeted by one of the CDs (eg weight loss, insomnia) (50%)
  • a general feeling of depression/anxiety manifesting in multiple health concerns (32%)
  • self-development (18%)

All participants perceived that their problems were partly psychological or emotional, and for those experiencing symptoms, that these physical symptoms were caused by stress. Less than half (41%) of all participants had attended their GP to discuss their concerns. Those who were prescribed medication discontinued shortly afterwards, saying there were too many side effects or they just didn’t want to take drugs (antidepressants, sleeping pills, etc.). Although all participants were optimistic about expected outcomes, only 9% of them followed the instructions as prescribed, even though many believed they had. A large proportion (73%) said that they followed the instructions of the program, but when qualitative information was analysed more specifically, it was found that only 14% initially practiced twice a day, as recommended. Most commonly, participants initially listened to the CD each day for one to four weeks. Nor did participants read the additional material at least once a week, as recommended. In fact, the written instructions and additional material were hardly ever referred to by 59% of participants.

From analysis of qualitative responses relating to initial interaction with the CDs, participants were grouped into 3 groups for further investigation: Initial Interaction with CD/material High {High 77); Initial Interaction with CD/material Moderate {Moderate /7); Initial Interaction with CD/materials Low {Low 77).

Expected versus Actual Outcomes

The expectations of participants were almost all directly in line with the titles of the CDs, i.e. “lose weight”, “get better grades”, “get a good night’s sleep”; one participant who purchased multiple CDs expected “a new me”. However, there was no evidence from the qualitative information that any participant expected any other improvements in their health or wellbeing.

When asked how they felt while they were listening to the CD, most often participants mentioned being “more calm” and “relaxed”, others mentioned being “de-stressed”, “switching off’, “feeling detached from the daily grind”, “sleepy”, and “nurturing to myself’. All participants reported an improvement with their concerns, with only one reporting a minimal change for the better. Regardless of the level of Initial Interaction in listening to the CDs, the improvements reported were moderate, with 36% of participants noticing improvements after listening to the CD for the first time, and 59% noticing improvements within 1 to 4 weeks. The remaining respondents were unable to determine how long before noticing an improvement. Improvement timeframe (speed of change) was as expected for 27%, faster for 36% and the remaining 36% had no expectations of speed of change.

Qualitatively, participants individually reported many benefits including:

  1. feeling calmer/more relaxed;
  2. sleeping better;
  3. being more positive/having more positive thoughts;
  4. having more energy/energised;
  5. losing weight;
  6. being more focused;
  7. being more in control;
  8. having less headaches;
  9. stopping smoking;
  10. having increased patience;
  11. feeling more balanced;
  12. having normalised blood pressure;
  13. not feeling afraid;
  14. being able to turn-off (i.e. no “busy brain”);
  15. having reduced negative thoughts;
  16. experiencing personality improvements (more understanding of other people);
  17. improved listening skills;
  18. experiencing diminished mood swings.

Participants were asked if they had noticed improvements in some specific areas. Not surprisingly, “Poor Health ” participants were more likely to have noticed benefits for a range of issues (see Table 2).

Specific Area Percentage of participants with a noticed improvement.
  Overall Good


Poor Health
General feeling of well-being 91 80 100
Positive thoughts about the future 82 60 100
Ability to cope in stressful situations 73 60 83
Feeling of optimism about the future 64 40 83
Confidence 64 40 83
Sleep 55 50 58
Concentration 55 50 58
Memory/recall of information 50 50 50
Reaction to situations where you would normally get angry 45 20 67
General health 41 20 58
Luck 36 10 58
Tendency to procrastinate 32 0 58
Weight (loss or gain) 23 10 33
Feelings of attractiveness 23 10 33
Finances 23 0 42

Improvement in “general feeling of well-being” and “positive thoughts” appears least likely for those in the following groups: How II’ and ‘Used the CDs for less than 6 months Participants were also asked to rate the incidence of certain common medical issues prior to listening to the CD and after listening to the CD, on the following broad 4-point scale: Often

(4), Regularly (3), Sometimes (2), Never (1). All of the issues yielded an overall positive difference in mean scores for before and after experiencing the PPGI, as reported in Table 3,

i.e. all conditions improved. The highest differences, (i.e. greatest improvements), were noted for Insomnia, Anger and Negative Thinking. Even higher score differences were recorded for those in the “Poor Health” group. Interestingly, some of the biggest differences were noted for participants who were grouped into the “Moderate IF group.

Table 3: Means Score differences of incidence rating of common medical issues, before and after using PPGI.

Issue Mean Score Difference before and after PPGI
  Overall Poor Health Moderate II
Insomnia 0.89 1.09 1.80
Anger 0.79 1.08 1.00
Negative thinking 0.79 1.00 1.16
Fatigue/tiredness 0.69 0.86 1.00
Headaches 0.67 1.00 0.84
Irrational beliefs 0.66 1.09 1.17
Depression 0.64 0.91 0.67
Digestion 0.44 0.72 0.50
Viral infection 0.32 0.41 0.50
Blood pressure 0.28 0.45 0
Obesity 0.28 0.46 0.33
Aches/pains 0.22 0.37 0.17
Hair loss 0.22 0.37 0.33
Alcoholism 0.21 0.34 0.17
Hot/cold spells 0.17 0.30 0.20


0.16 0.28 0
Decreased heart rate 0.05 0.09 0



A large proportion of the participants would like to continue using the CD (96%), and were still using the CD (83%). 46% of participants were still using the CD on a regular basis, 23% using the CD “as needed”, 14% “not often enough”, and the others (18%), “not at all”. However, in continuing to use the CD, 41% expect that this will only be “as needed”, 18% will continue “once a week”, 23% continuing “a few days a week” or “5 times a week”, with the other participants (18%) unable to specifically say how often they will use the CD.

Unexpected benefits

When asked to identify any unexpected benefits, the following were each mentioned by one or more participants: motivation; stronger memory; needing less sleep; benefits applicable to other areas of life (work/family); stress relief; normalisation of blood pressure; decreased back and neck pain; no more fear; positive thoughts; improved communication skills; ability to control anger; ability to laugh again.

Associated Benefits

Coincidentally, use of analgesics was reported to have decreased slightly over time from the 12 months prior to using the PPGI, and since using the PPGI (Mean Scores of 2.21 and 1.79 respectively, using the previously described 4 point relative frequency scale). Participants also reported that they visit the doctor less often, post PPGI, with an overall number of visits annually of 4.18, prior to PPGI, and after PPGI of 2.88 visits annually. However, these results should be interpreted with caution given that the average time since acquiring the CDs is only 6.6 months. Over half (59%) reported that others (friends/family/colleagues) had recognised improvements in performance and or their behaviour.


It seems that for these participants, many aspects of their lives improved because of the use of PPGI at least in the short term. More research is needed to investigate what, if any, benefits may exist in the long term from the use of PPGI. The key finding of this research is that listening to CDs incorporating the techniques of GI, suggestion, together with affirmations tend to assist people in feeling better by increasing their general feeling of wellbeing, increasing positive thoughts, and improving their ability to cope in stressful situations; in decreasing the incidence of insomnia, and in feelings of anger and negative thinking.

These benefits may occur regardless of the reason for acquiring the CD in the first place. However, further investigation of the causal factors is yet to be seen, as it is difficult if not unlikely to facilitate a controlled sample without mitigating circumstances influencing the results of analysis.

This being said, regardless of the suggestion component (i.e. CD used) it appeared that the most common benefits people noticed were increased relaxation, decreased negative thoughts, and decreased stress; with a couple of exceptions, i.e., the first benefit for those using the “Sleeping Like a Baby” CD, was better sleep. Those people, who had only a moderate commitment initially, reported considerable improvements across a range of symptoms.

For current participants however, there was no long term commitment to developing a lifestyle that incorporates the techniques into their daily lives. The book associated with the CD was perceived to be of little importance to the treatment. For a lasting benefit there appears to be a need to be more committed and persistent in engaging with the CD, with a need to understand that relaxation therapy needs to become a way of life to bring about lasting changes. However, this is yet to be seen and requires further study. Also, participants did not appear to be looking for lasting benefits, but rather sought a ‘quick fix’ or speedy improvement of a problem. They saw the CD as the means to an end, with many of them expecting quick results and consequently getting them, but they didn’t last. The influence of prior personal beliefs associated with the treatment modality poses interesting challenges in understanding therapy outcomes. These beliefs were challenged as participants often indicated that they had to keep revisiting the CDs when the problem arose again. The fact that the written material was hardly referred to by any participant indicates that when people choose treatment in the form of PPGI as a CD, they expect just to listen, and are not interested in reading. These factors may also suggest participant capacity and/or will to retain information, ideas, and beliefs associated with the path of recovery across various symptoms.

Also clear from the qualitative information is that participants have very little time to engage in such a therapy. Thus, it can be said that this type of program should be short and sharp. Participants in this study had trouble listening to a CD for 30 minutes once a day, let alone twice a day as was recommended for maximum benefit. Many respondents (96%) indicated that they would be continuing to practice the techniques, however, there is still an inclination to perceive PPGI within the medical model, and to treat it like a pill, i.e. listen to the CDs or to practice the technique as a ‘quick fix’ when things start to go wrong, and our mind needs a ‘top up’. The idea or notion of keeping the mind permanently in good health by practicing regular relaxation is not yet instilled in the collective consciousness, even for those who have recognised the connection of mind/body. This type of therapy may require some sort of self- checking or monitoring component which would enable individuals to gauge their level of relaxation and progress. These mitigating factors including cultural dispositions to holistic health require greater sensitivity in future research design, particularly in a field like counselling, and psychotherapy, where the whole person tends to be the object of analysis.

Results related to frequency of visits to the doctor and use of analgesics were inconclusive given the timeframe between participants having acquired the CDs and the conducting of the research and the fact that most of the respondents were generally relatively healthy.

Some participants reported positive comments from family and friends. This is interesting as the social aspect of increasing positive outlook in people’s lives is often overlooked in similar research, which tends to focus on the individual exclusively. However, this is an area where further investigation could contribute constructively in developing group or family therapy approaches that incorporate some aspect of PPGI. Comprehensive long-term followup with people who have used this type of CD or therapy (both successfully and unsuccessfully) would be beneficial in understanding how they can be used for the greatest benefit.



Edzard Ernst, MD, PhD, FRCP, FRCP(Ed); Max IL Pittler, MD, PhD; Barbara Wider, MA; Kate Boddy, MA


The effectiveness of mind-body therapies is sometimes doubted. The aim of this article is to evaluate trends in the development of the evidence base for autogenic training, hypnotherapy, and relaxation therapy. For this purpose, a comparison of 2 series of systematic reviews was conducted. The first is related to the evidence base in 2000, the second to that in 2005. Both employed virtually the same methodology and criteria for evaluation. The results of our comparisons show considerable changes during the observation period. The weight of the evidence has become stronger for several indications, and the direction of the evidence has been altered in a positive sense in several conditions. Applying the rules of evidence-based medicine, the following mind-body therapies are now supported by strong evidence: hypnotherapy for labor pain and relaxation therapy for anxiety and insomnia, as well as for nausea and vomiting induced by chemotherapy. It is concluded that an evidence-based approach for mind-body therapies is constructive and can generate positive results. (Altern Ther Health Med. 2007;13(5):62-64.)


Edzard Ernst, MD, PhD, FRCP, FRCP(Ed), is a physician and Director of Complementary Medicine, Max H. Pittler, MD, PhD, is a physician and Deputy Director of Complementary Medicine, Barbara Wider, MA, is a research fellow in Complementary Medicine, and Kate Boddy, MA, is an information officer for the Complementary Medicine unit, all at the Peninsula Medical School, Universities of Exeter and Plymouth, UK.


Мind-body therapies can be defined as a “patient- orientated, proactive approach to health and healing that values personal responsibility and self-motivation. Lifestyle and personal attitudes are the focus to bring about personal transformation and gain mastery over the mind and body.”llp2721 Despite the long history of these approaches, the question of whether they are clinically effective remains controversial.

A resolution of the debate is unlikely to come from single studies that often generate some degree of contradiction. The solution could, however, be facilitated through systematic reviews of the clinical trials that summarize and evaluate the totality of the available clinical evidence of a predefined nature and quality.

In 2001, we published The Desktop Guide to Complementary and Alternative Medicine, which contains a series of systematic reviews of the main forms of complementary and alternative medicine (CAM), including several mind-body iherapies.2 These reviews were concluded in 2000. In 2005, we finished our update for a new edition of this book.3 This update allows us l o compare the evidence base (summarized by using virtually I lie same methodology) as ii existed in 2000 with that of 2005. The aim of this article is to define how the evidence base for a selection of mind-body therapies lias changed overtime and lo discuss any emerging trends.


Both versions of the Desktop Guide are based on systematic reviews employing virtually the same methodology; details are provided elsewhere.2,3 In brief, systematic searches were carried out in the databases Medline, EMBASE, AMED, and the Cochrane Database of Systematic Reviews. Each database was searched from its respective inception until March 2000, and update searches were run until June 2005. Systematic reviews and meta-analyses of clinical trials were given priority, and copies of originals were obtained. Where systematic reviews or meta-analyses were not found, the evidence from randomized controlled trials (RCTs) and controlled clinical trials (CCTs) was considered next. Trials were considered for indications that had dedicated chapters in our book. The decision about which indication merited a dedicated chapter was made on the basis of which indications are commonly seen in primary care and frequently treated with CAM and for which indications the most evidence is available.

To evaluate the data, we created a parameter that we called the “weight of the evidence.” This was a compound variable consisting of the level of evidence (eg, related lo a single trial or a meta-analysis), I he quality of the evidence (usually estimated with the Jadad score4), and the volume (ie, total sample size).

The “weight” was graded in categories: low, moderate, and high. In addition, we graded the “direction of the evidence” in 5 categories: clearly positive, tentatively positive, uncertain, tentatively negative, and clearly negative.

The mind-body therapies included in the present analysis are autogenic training, hypnotherapy, and relaxation. We compared the weight and direction of the evidence for treating those conditions included in our book and associated with trial evidence related to the above-named therapies.




There were several changes in the weight and/or direction of evidence: 3 for autogenic training, 3 for hypnotherapy, and 6 for relaxation techniques (Tables 1-3). The following approaches are supported by strong positive evidence (ie, maximum weight and clearly positive direction of evidence): hypnotherapy for

labor pain and relaxation therapy for anxiety and insomnia as well as for nausea and vomiting induced by chemotherapy.

Weight of the Evidence

The weight of the evidence became more solid in the following cases: autogenic training for asthma and for depression, hypnotherapy for irritable bowel syndrome, and relaxation techniques for cancer palliation, drug/alcohol dependence, insomnia, and nausea due to chemotherapy. The weight is now maximal for several treatments: hypnotherapy for irritable bowel syndrome, labor pain, and obesity and relaxation techniques for anxiety, hypertension, insomnia, and chemotherapy-induced nausea.

TABLE 1 Summary of Clinical Evidence for Autogenic Training Condition Weight of Evidence Direction of Evidence

2000 2005 2000 2005

Anxiety OO OO
Asthma 0 OO ■=> ■=>
Atopic eczema 0 0 ■=> ft
Depression 0 OO  
Headache 0 0
Hypertension No entry OO No entry
Insomnia No entry OO No entry

TABLE 2 Summary of Clinical Evidence for Hypnotherapy

Condition Weight of Evidence Direction of Evidence
  2000 2005 2000 2005
Alcohol dependence 0 0 -a -0
Anxiety 0 OO  
Asthma OO OO  
Atopic eczema 0 0  
Cancer palliation 0 0  
Erectile dysfunction 0 0 ft If
Hay fever No entry OO No entry  
Headache OO OO  
Hypertension No entry 0 No entry  
Insomnia OO OO ft  
Irritable bowel syndrome OO 000 ft  
Labor pain No entry 000 No entry If
Nausea and vomiting induced by chemotherapy OO OO  
Postoperative nausea and vomiting No entry 0 No entry  
Overweight/obesity No entry 000 No entry  
Rheumatoid arthritis 0 0 =>
Smoking cessation OO OO -0 -0
Tinnitus OO OO  

TABLE 3 Summary of Clinical Evidence for Relaxation Therapy

Condition Weight of Evidence Direction of Evidence
  2000 2005 2000 2005
Anxiety 000 000 ff ft
Asthma OO OO ■=> ■=>
Cancer palliation 0 OO ff ft
Crohn’s disease No entry 0 No entry  
Depression OO OO ff ft
Drug/alcohol dependence 0 OO ■=> ■=>
Headache OO OO & &
Hypertension No entry 000 No entry &
Insomnia OO 000 ft ft
Menopause OO OO ft ft
Migraine OO OO ft &
Nausea and vomiting induced by chemotherapy OO 000   ft
Premenstrual syndrome 0 0 ft ft
Rheumatoid arthritis OO 00 ft &
Smoking cessation 0 0 ft ft
Tinnitus 0 0   &
Ulcerative colitis No entry 0 No entry ft
Legend for All Tables
000 High weight of evidence
00 Moderate weight of evidence
0 Low weight of evidence
ft Clearly positive evidence
  Tentatively positive evidence
о Uncertain evidence
Sa Tentatively negative evidence
в Clearly negative evidence



Direction of Evidence

The direction of the evidence changed in a positive sense in a few situations: autogenic training for eczema and depression and relaxation techniques for nausea due to chemotherapy. It changed in a negative sense in the following situations: hypnotherapy for insomnia, irritable bowel syndrome, and rheumatoid arthritis. The direction of the evidence is now clearly positive for several situations: autogenic training for eczema, hypnotherapy for erectile dysfunction and laborpain, and relaxation techniques for anxiety, cancer palliation, depression, insomnia, menopause, chemotherapy-induced nausea, premenstrual syndrome, smoking cessation, and ulcerative colitis. In many of these situations, however, I his positive direction of evidence is not accompanied by the maximum weight of evidence.


These comparisons suggest that, within the observation period (2000-2005), the evidence base of mind-body therapies has changed considerably. In our update,3 we lisl numerous situations where the evidence for CAM has become more solid during the preceding years. Mind-body therapies feature prominenlly in that list. The present analysis shows that several new clinical trials have been conducted in this period. Considering the financial and methodological difficulties in conducting trials of mind-body therapies, I his analysis seems a notable achievement. The emerging data have led to a situation where several indications can be considered to be supported by strong evidence: labor pain (hypnotherapy), anxiety, insomnia, and nausea and vomiting induced by chemotherapy (relaxation techniques).

Our analysis might influence the future research agenda of inind-body therapies. For instance, several indications can be identified for which the direction of evidence is “positive” but the weight of the evidence is not maximal. This seems to suggest that these are areas in which future research might be fruitful.

Our comparison also suggests that the application of evidence-based medicine (EBM) to mind-body therapies is both possible and constructive. It has been argued repeatedly that EBM is not applicable to areas of “alternative medicine.”5 Our analysis seems to contradict this view. It is conceivable that EBM will eventually generate a list of indications for which mind-body therapies are demonstrably effective. As science is not a good tool for proving a negative, EBM will not easily yield a list of indications for which mind-body therapies are definitively ineffective. It will, however, be possible to provide indications for which effectiveness is less likely. In turn, this knowledge could be applied in clinical practice and for educational purposes.6

Several limitations of our comparison should be mentioned. Even though our approach in producing the 2 editions of our Desktop Guide was systematic, we cannot exclude a degree of bias in evaluating the published trial data. As the emphasis of this article was on comparing changes over time, one would hope that such bias is minimized: we used the same approach for both editions. Moreover, there was both an internal and an external review process, which should have minimized bias.3:! Another potential drawback is that we included only a limited number of indications (ie, those conditions for which there was the most CAM evidence) and evaluated only the data from CCTs. Thus our comparisons fail to encompass all conditions for which mind-body therapies have been tested and clinical evidence may be available. Finally, our searches ended in June 2005. New evidence may have become available since then. For instance, a recent randomized clinical trial of autogenic training strengthened the view that it is effective in reducing anxiety,5 and the effectiveness of hypnotherapy for irritable bowel syndrome was supported by additional trial data.’ Similarly, a new systematic review demonstrated that the adjunctive use of relaxation therapy reduces the need for medication in patients suffering from asthma.8

In conclusion, our comparison of the current and past evidence indicates that the research activity relating to CCTs of mind- body therapies has been considerable. The major changes seem to indicate that the trial evidence is increasingly supporting the effectiveness of mind-body therapies for a wide range of conditions.




Fear Behavior, Fear Imagery, and the Psychophysiology of Emotion: The Problem of Affective Response Integration

Peter J. Lang, Daniel N. Levin, Gregory A. Miller, and Michael J. Kozak

‘ University of Wisconsin–Madison

This article explores relationships among behavioral, verbal report, and physiological responses in human fear. The approach taken here is based on a bio- informational theory of emotional imagery (Lang, 1977a, 1979, 1983), which serves here as a more general framework for the understanding of affective behavior. Experiment 1 compared the psychophysiological reactions of subjects with two different fears, focal phobia (N = 12) and social-performance anxiety (V =

13), who were determined by questionnaire and interview to be at the high extreme of their respective fear reference groups. Each group was exposed to both its own and the other group’s primary fear stress (i.e., a snake-exposure test and a public speaking performance). These same subjects were also instructed to imagine both types of fear situations as well as control scenes.

The results indicate a different psychophysiological response for the two fear groups across the two different fear contents. Thus, snake-phobic subjects clearly showed greater arousal (e.g., in verbal report and heart rate increase) when exposed to a live snake than did socially anxious subjects. However, the groups were less clearly differentiated in speech performance. Despite significantly greater verbal reports of fear and arousal by socially anxious subjects, both fear groups showed a similar marked increase in physiological arousal during speech performance. In contrast to these exposure responses, neither group generated a significant physiological reaction to either fear content during imagery assessment.

Experiment 2 was a more intensive examination of emotional imagery with new samples from the same two fear populations (focal phobia, N = 20; social- performance anxiety, N = 20). An imagery pretraining program, based on the reinforcement of verbal report of somatic response content in imagery, led to a significant visceral arousal response during fear imagery. Furthermore, response- trained subjects showed a pattern of heart rate change during imagery that varied between subjects and fear contents in a way that paralleled the results for exposure obtained in the first experiment. Response-trained subjects also showed relatively greater concordance between verbal report and visceral measures than did untrained subjects. A control-training procedure, based on the reinforcement of stimulus information, failed to produce any of these effects on test imagery. Together, the two experiments elucidate theoretical and methodological problems in the laboratory analysis of human fear, how these problems have impeded previous analysis, and how the current approach guides selection of dependent measures and explains sources of discordance in multiple-response assessment.


It has frequently been stated (Borkovec, 1976; Lang, 1964, 1968, 1971, 1977b, 1978; Rachman, 1978) that fear and anxiety can be pragmatically studied as constellations of language behavior, motor acts, and physiological events. Intense emotional states are characterized by high amplitude and temporal contiguity of responses from all three data sources. However, under most conditions of affective stimulation, responding systems appear to be only loosely coupled (e.g., Lang, 1968; Leitenberg, Agras, Butz, & Wincze, 1971). Furthermore, the observed response patterns do not seem identical for all fearful subjects, for all fear stimuli, or for the same stimuli under different conditions of administration (e.g., imagery and exposure). Although there is agreement that fear assessment requires a converging, opera-, tional, multiple-response analysis, the absence of a widely accepted organizing framework that can account for these varied data has made the approach difficult to implement.

The research in this article was guided by the theory of imagery and emotion presented by Lang (1977a, 1979, 1983). This view presumes that an emotion is an action set, defined by a specific information structure in long-term memory, that, when accessed, is processed as both a conceptual and a motor program. The information structure on which an affective response is based is conceived of as an associative, network of propositions. These networks are of the general sort first used by Quillian (1966) to describe semantic knowledge and later adapted to accommodate other types of information (e.g., Anderson & Bower, 1974; Kieras, 1978). The data in the network include (a) stimulus infor-. mation (i.e., propositions about prompting external stimuli and the context in which they occur), (b) response information (i.e., propositions that define the affective language, overt actions, and visceral support responses occasioned by this context), and (c) relational meaning information (i.e., propositions that elaborate relationships and mediate between input and output information).

An affective network is accessed and run as a program when a sufficient number of its propositions are instigated by external input (or presumably also by the spontaneous prompt of associative links in the brain). Network processing, as in perception or imagery, always involves measurable efferent outflow. This occurs because response information is double coded, both as semantic knowledge and as motor programs identified with the specific response propositions of the network. Thus, when subjects imagine or anticipate exposure to an affective context, efferent outflow can be measured even if the designated overt action is inhibited.

Networks vary in coherence. That is, the average associative strength among the propositions varies for different networks and with it varies the probability that a network will be activated as a unit. Phobia networks have high coherence and can be called emotion prototypes. They require fewer matching input propositions for activation and control relatively stable affective response dispositions. Thus, regardless of the medium of external instigation (i.e., imagery, a film model, anticipation, or exposure), the same phobic- information structure in memory is accessed, and a broadly consistent pattern of efferent outflow should be observed across these varied contexts of stimulation.

In the following experiments, in addition to measuring avoidance and performance deficit, verbal report of arousal and physiological activity are assessed in two kinds of stimulus settings. These are object exposure and instruction to imagine the focal context of fear. The goal of this research is to examine the consistency of response patterns as they vary with different input information, subject populations, type of fear, and task context. By way of introduction to these experiments, we will first consider the results of previous research in which two fear types, socially anxious and focal-phobic subjects, were studied under conditions of exposure and imagery. This brief review both highlights the positive results that appear to be consistent with our view and points out inconsistencies in the data that we feel the present approach can logically resolve and for which our experiments provide pertinent results.

Imagery and Physiological Response

In developing systematic desensitization therapy, Wolpe (1958) assumed that imagining fearful situations would prompt patterns of physiological responses essentially identical to those that occur when fear objects are presented. A variety of results suggested such a parallel between physiological responding during the performance of perceptual tasks and during instructed imagery of the same stimuli. For example, concordance between efferent activity in imagery and performance has been found for a weight-judging task (Shaw, 1940), speaking (McGuigan, 1973), and visual tracking (Deckert, 1964; Weerts, Simons, & Lang, Note 1). More relevant to the therapeutic context, imagining apparently fearful or anger-inducing situations (Roberts & Weerts, 1976) elicited visceral and somatic response patterns similar to those found in a previous study of objective, affect-inducing situations (Ax, 1953). Furthermore, the amplitude of skin-conductance and heart rate responses has been shown to increase progressively as subjects imagine scenes higher on a list scaled from least to most fearful (Lang, Melamed, & Hart, 1970; Van Egeren, Feather, & Hein, 1971). Few experiments, however, have directly compared the response patterns of fearful subjects during the imagination of fear scenes with the patterns observed in the specific, referent, actual fear settings.

Fear Group Differences


Limited evidence is already available suggesting that various clinical populations differ in fear-response profiles, particularly during fear-relevant imagery. Lang et al. (1970) compared focal-phobic subjects and socially anxious subjects during visualization of items from randomly ordered, fear desensitization hierarchies. The focal-phobic group rated their scenes more vivid and more anxiety arousing and showed greater heart rate and skin-conductance responses to items high in the hierarchy than did the socially anxious group. Weerts and Lang (1978) replicated these group differences in verbal report of vividness of imagery and arousal and found a similar trend in heart rate and skin conductance. Butullo (Note 2) also reported that imagery vividness was greater for patients with specific fear than for anxiety-state patients.


Borkovec (1973) and associates (Borkovec, Stone, O’Brien, & Kaloupek, 1974; Borkovec, Wall, & Stone, 1974) studied the physiological responding of focal-phobic and socially anxious individuals during fear-relevant behavior tests. In contrast to the findings from the imagery studies, where small-animal-phobic subjects appeared to be more responsive than socially anxious subjects, these investigators found much greater increases in heart rate when socially anxious subjects delivered a speech than when snake-phobic subjects confronted a snake. Borkovec (1976) concluded from this series of studies that snake-phobic subjects were “low physiological reactors” and that socially anxious subjects were “high physiological reactors.” However, the stress tasks performed by the two fear groups may have differed substantially in their requirement for cognitive and motor activation; that is, the physical demands of preparing for and giving a speech may exceed those of viewing a small animal, independent of any difference in affective quality or subject population.


The differential responsiveness of phobic populations in imagined and actual contexts has important treatment implications. It is generally assumed that vividness of affective imagery is a prerequisite for successful imagery therapy (e.g., see Wolpe, 1958). Lang et al. (1970) found that phobic subjects who showed heart rate increases when they reported verbally that a scene was frightening were more likely to benefit from desensitization therapy than subjects whose physiological responses did not match their verbal reports. Thus, ability to generate a more complete imagery response was consistent with good therapy prognosis. Imagery ability differences may also account for varying outcomes in clinical studies of imagery therapies. Boulougouris, Marks, and Marset (1971) found desensitization therapy and flooding equally effective with specific-phobic patients, whereas flooding was more effective with generalized-anxiety patients. It can be argued that the minimal imagery instructions characteristic of desensitization sufficed for specific-phobic subjects, a clinical population found to report vivid imagery. In contrast, the elaborate emotional image scripts used in flooding were required for general-anxiety patients, the groups with less imagery facility.

The Research Problem

The present research assesses affective behavior in two groups of fearful subjects, the socially anxious and the focal phobic. In Experiment 1, tests were undertaken when each group was exposed to actual fear stimuli, both to their own primary stress and to the context that was a primary stress for the other group. Sample measures from each of the three responding systems are examined: (a) verbal report of arousal, (b) avoidance and performance deficit associated with the fear context, and (c) visceral activation. The main hypothesis to be tested is that focal-phobic and socially anxious subjects may be more physiologically reactive to relevant fear stress. This same question is further examined in the context of fear imagery. The verbal reports of arousal and physiological reactivity of each group are evaluated when subjects are under instruction to imagine their own relevant primary-fear stress, to imagine the other group’s fear stress, and to imagine neutral and arousing control scenes that are irrelevant to group assignment. By studying all types of fear situations with both fear groups, in vivo and in imago, the present design permits a resolution of the conflict between overt-behavioral studies and imagery studies concerning response differences between focal phobia and social anxiety.

Experiment 2 represents a more thorough examination of fear imagery in a new sample of socially anxious aiid focal-phobic subjects. As we stated earlier, an emotional image is understood to be a conceptual network in the brain, integrating perceptual, meaning, and somato-visceral propositions. The phobic image expresses a response prototype or template for behavior. For this reason, efferent, physiological components of the image may be critical to its therapeutic effectiveness (Lang et al„ 1970). Only if these response elements are processed in therapy can we expect to alter motor preparation and affective behavior in objective settings. However, subjects vary in their ability to generate imagery, and even severely phobic patients may not show physiological response patterns to phobic imagery, that are discriminable from those occasioned by neutral scenes (Marks & Hu- son, 1973).

This latter problem is specifically addressed in Experiment 2. An effort was made to enhance subjects’ imagery. Thus, prior to evaluation of their physiological response to fear and control scenes, these subjects were given training in imagery processing according to new methods developed by Lang, Kozak, Miller, Levin, and McLean (1980). Half of each fear group received stimulus training designed to focus the imagining subjects on the specific context and objects in the scene (i.e., the fear stimuli); the other half of the group received response training intended to focus image processing on the subject’s active role in the scene. Stimulus training was used here as a control procedure in the sense that it involved a degree of task motivation and social interaction with the experimenter similar to that involved in response training. However, our previous results (Lang et al.,

1980) have shown that stimulus training does not enhance fear-network activation as evi-, denced by an increase in appropriate physiological responding during imagery. A general set to process stimulus detail appears to access information that is either redundant or irrelevant to the fear prototype. On the other hand, response training has been a potent manipulation. An acquired set to process action information (i.e., emphasizing the response side of the conceptual network) has the effect of amplifying physiological patterns that are part of the fear image. This places in relief those differences between scene contents that might otherwise be obscured by subject variance in imagery ability. Experiment 2 was designed to discover if these effects, already found with Unselected normal subjects, could also be obtained for focal- phobic and socially anxious subjects.

Experiment 1

The design of Experiment 1 permits a comparison of the two groups’ reactions under five different conditions of fear evocation: (a) Imagery 1 (preexposure)–subjects are given standard imagery scripts describing fearful situations in the natural setting; (b) anticipation–subjects are told that they may anticipate a direct confrontation with the fear stimulus in the very near future; (c) antici- pation/math–subjects waiting to confront a fear stimulus are given an arithmetic task; (d) exposure–subjects are either to confront a live snake or to give a speech before an audience; (e) Imagery 2 (postexposure)–subjects are given imagery scripts based on the immediately preceding exposure situation.


Figure 1. Frequency distributions based on scores obtained from approximately 3,600 college students. (The position of the 70th percentile for both the male and female distributions is indicated by the male and female symbols on the abcissa.)

Both subject groups (focal phobia and social anxiety) were given all experimental conditions with both snake and speech content. Furthermore, all three components of the emotional response were sampled during exposure and anticipation/math conditions. Exposure permits the assessment of behavior directly pertinent to the subjects’ fears, whereas anticipation/math involves the assessment of performance on a mental task in which deficits have been obtained under similar conditions of anticipatory stress (Geer, 1966; Losen, 1959). An advantage of this procedure is that it permits a performance assessment of both groups while each is under stress from its own dominant fear; however, unlike the exposure situation, the comparison of fear groups is unconfounded by different task demands or different conditions of external stimulation.

In anticipation and the two imagery conditions, objective fear stimuli were not presented, and only verbal report and physiological reactivity were measured. The first imagery assessment permits the evaluation of subjects’ responses to realistic fear situations that cannot be duplicated in the laboratory; the second imagery assessment, in which the scripts are based on the immediately preceding laboratory exposure task, permits the evaluation of a memory image that is based on a recent experience common to all of the subjects. This latter paradigm is one that is commonly used in the study of sensory imagery (e.g., Sheehan, 1972).

The broad aim of this experiment was to determine whether focal phobia and social- performance anxiety involve persistent and differentiated fear-response dispositions. If this is so, then situations that are consistent in thematic content although varied in behavioral context (e.g., imagery and exposure) should show a similar pattern of psycho- physiological response. This further implies that populations can be described that show a reliable response pattern to one fear theme and not to the other.





Subjects were 25 University of Wisconsin volunteers selected from the introductory psychology class and the Psychology Department’s paid subjects pool. Individuals scoring in the top 10% of the distribution of scores on a snake-fear questionnaire, Snake Anxiety Questionnaire (SNAQ; Klorman, Weerts, Hastings, Melamed, & Lang, 1974) and a speech-fear questionnaire, Personal Report of Confidence as a Speaker (PRCS; Paul, 1966), were invited to the laboratory for a structured interview. The interview was used to select subjects who showed the following characteristics: (a) high salience of the fear in daily life, (b) reports of physiological involvement in the fear, and (c) other indications of genuine anxiety. If the student met both questionnaire and interview criteria for anxiety about snakes or public speaking and reported no current drug use or history of cardiovascular abnormality, he or she was invited to participate in the experiment.1 This procedure yielded 12 snake-phobic and 13 speech-anxious subjects. Seven subjects in each fear group were female. Figure 1 shows the frequency distribution of the SNAQ and PROS scales for a college population (nearly 3,600 University of Wisconsin undergraduates). Shaded areas represent the distribution from which subjects who met the high-fear criteria were taken. Mean , scores for the two fear groups are presented in Table 1. It can be seen that all subjects scored quite high on the fear-relevant questionnaire: All group means were above the 90th percentile. These scores are in the same range as those that were obtained from clinic patients seeking treatment for the same fears and anxieties (see Figure 1). As a check on subject-selection procedures, SNAQ and PRCS scores were submitted to separate analyses of variance, which confirmed that subjects in the phobic group had higher SNAQ scores, F(l, 21) = 12.77, p < .002, and that the socially anxious group had higher PRCS scores, F(l, 21) = 42.73, p < .0001. Although some subjects scored relatively high on both scales (group assignment was always determined by the greater fear), none were excluded for this reason, so that the naturally occurring relationship of the two fears was not misrepresented. The trade-off that this entailed was most apparent among male speech-anxious subjects whose mean SNAQ score was nearly as high as their PRCS score.

The specificity of the two groups’ fears was further demonstrated by their similar scores on general anxiety questionnaires. Snake and speech groups did not differ on the Stress subscale of the Differential Personality Questionnaire (DPQ; Tellegen, Note 3), on the Fenz-Ep- stein Anxiety Scale (Fienz & Epstein, 1965), or on any of the latter’s three subscales: Muscle Tfension (MT), Autonomic Arousal (AA), and Feelings of Insecurity (FI). Snake and speech groups also did not differ in self- report “trait” measures of imagery–the Absorption subscale of the DPQ, the Control of Visual Imagery Scale (CVI; Gordon, 1949), total score on Sheehan’s modification of the Questionnaire Upon Mental Imagery (QMI; Betts, 1909; Sheehan, 1967), or on any of the latter’s seven subscales.


Note. The Fenz-Epstein Anxiety Scale (Fenz & Epstein, 1965) includes items from the Manifest Anxiety Scale (Taylor, 1953) and has been factor analyzed into three symptom clusters that make up the three subscales: Autonomic Arousal (AA), Muscle Tension (MT), and Feelings of Inferiority (FI). SNAQ = Snake Anxiety Questionnaire; PRCS = Personal Report of Confidence as a Speaker; QMI = Questionnaire upon Mental Imagery.

Table 1

Questionnaire Score Means and Standard Deviations

    Snake fear (SNAQ) Social anxiety (PRCS)   Fenz-Epstein Anxiety scales   Imagery


Group N AA MT И
Focal phobic M 12 18.1 Experiment 1 14.3 37.4 32.5 52.2 93.6
SD   4.0 7.9 6.9 5.2 9,3 32.4
Socially anxious M 13 10.6 26.1 34.0 34.5 46.8 93.9
SD   6.4 1.6 8.3 8.5 10.6 21.3
Focal phobic M 20 18.2 Experiment 2 15.1 33.9 33.6 50.0 90.4
SD   3.8 7.9 8.6 9.2 11.2 17.9
Socially anxious M 20 12.0 26.0 35.8 33,9 58.4 98.6.
SD   6.6 2.5 6.0 7.4 12.1 22.7

The subject sat in a reclining chair in a room next to the equipment room. During imagery trials, the chair was put in the reclined position and the room light was turned down. During the behavior tests, the subject sat upright with the room light on. The presentation of stimuli, their timing and sequence, and the collection of self- report judgments and physiological data were accomplished under the control of a Digital Equipment Corporation PDP-12 computer. Relaxation and imagery script instructions were delivered through headphones by a cartridge tape recorder. This device formed part of an experimenter-subject intercom. During behavioral procedures (i.e., those other than imagery), subjects received instructions over a loudspeaker and did not wear headphones. After each imagery trial, the subject rated the emotional arousal and vividness of the image. A computer-driven graphics display facing the subject presented two animated displays with visual properties that matched the dimension to be rated. The subject adjusted these displays with a potentiometer. The arousal display consisted of a dynamic, random pattern of dots. The size and rate of change of the pattern increased to indicate greater arousal. The vividness display consisted of a pattern of dots that gradually coalesced into a sharply de-

fined rectangle to indicate greater vividness. The computer converted each rating as it was made to a 64-point scale.

Physiological activity was monitored via a Beckman- type RM dynograph. Lead 2 electrocardiogram (EKG) was obtained from Beckman miniature silver-silver- chloride electrodes attached to the lower rib cage. The Schmitt trigger of a Tektronix 503 oscilloscope interrupted the computer to measure heart period each time it detected a cardiac R wave. Respiration was measured by two 3-inch mercury strain gauges taped in series to the chest and abdomen, forming a Wheatstone bridge with a Beckman 9875B coupler. Amplitude, period, and ratio of inspiratory to expiratory period were obtained by computer sampling of the analog signal. To monitor palmar sweat activity, the volar surface of the middle segment of the first and second digits of the nondominant hand was lightly rubbed with distilled water, then with Johnson & Johnson K-Y jelly, and was then wiped dry. Beckman standard silver-silver-chloride electrodes filled with K-Y jelly were then attached (the rating potentiometer was operated by the other hand). A Beckman 9842 GSR coupler supplied a constant .5 volt to obtain skin conductance (SC). To quantify activity in the frontalis and semispinalis capitus muscles’, electromyogram (EMG) signals from Beckman miniature electrode pairs on the forehead and neck were rectified and integrated separately for the two sites. A second-stage, resetting integrator combined these two signals under the control of the computer, which measured reset period. EMG electrodes were attached during the first imagery procedure, but not subsequently, to permit the subject more freedom of movement during exposure tests.

The snake used in the snake-avoidance procedure was a light brown, 76-cm boa constrictor that was about 3.5 cm in diameter. For testing, the snake was maintained in a state of mild hypothermia so that it typically moved little during the procedure. The snake was contained in a clear plastic box (60 X 30 X 30 cm) that was set on a small rolling table so that the snake was approximately at the seated subject’s chest level.

The audience for the speech test consisted of four observers plus the experimenter. At least one male and one female were included in every audience. The five people stood in silent attention in the hall at the open door of the subject room throughout the speech test. A Sony video camera on a tripod in the subject room recorded part of the subject’s speech.


After the screening interview, subjects participated in three individual laboratory sessions. Lab Session 1 consisted of familiarizing the subjects with the physiological recording procedures ahd the administration of questionnaires. Lab Session 2 consisted of 11 imagery trials, during which physiological and self-report data were collected. Lab Session 3 consisted of snake and speech anticipation and exposure tasks, followed by six imagery trials, during all of which physiological and self-report data were collected.

Lab Session 1. Physiological transducers were attached to the subject, and a series of simple tones was presented.2 The subject then completed Form В of an arithmetic questionnaire (Losen, 1959) so that his or her performance on this task could be assessed under non- stressful conditions approximating those when other forms of the arithmetic test were administered in Lab Session 3. The experimenter instructed the subject to work as quickly and as accurately as possible and allowed 90 sec for the subject to complete as many items as t possible. Physiological sensors were removed, and the subject then completed the CVI, the QMI, the Fenz-Ep- stein Anxiety Scale, and the Stress and Absorption subscales from the DPQ.

Lab Session 2. Imagery trial apparatus, materials, procedures, and data reduction and analysis closely followed those reported elsewhere (Lang et al., 1980). After the electrodes were attached, subjects were told that they would hear taped scene presentations, that they were to imagine each scene while it was described, and that upon termination of the narration they were to continue to imagine the scene starting from the beginning of the description. A relaxation tape was played, then physiological measures were obtained during a resting baseline period. Subjects then received a series of scenes to imagine, each presented according to the following format: a 30-sec rest period, a 50-sec read period during which the script was played to the subject, a 30-sec image period, a 30-sec recover period during which subjects stopped imagining and relaxed, and untimed arousal and vividness ratings. The first, second, and seventh scenes presented in the series were emotionally neutral in content. They involved pleasant, relaxed situations with no physical activity. The first neutral scene was regarded as practice and the data were not used. Data were analyzed for the other two neutral scenes and the other five scene contents: two fear scenes (giving a speech, encountering a snake), one general danger scene (either theatre fire or nighttime intruder), and two action scenes (sexual encounter, physical exercise). The order of presentation for nonneutral scenes was approximately random. The fear, danger, and action scripts each included a single reference to increased cardiovascular, skeletal muscle, sweat gland, and respiratory activity, which were the physiological systems actually being monitored.

Lab Session 3. Following EKG, SC, and respiration transducer placement, each subject was given a series of snake and speech tasks, followed by six imagery trials. Fear-relevant procedures were always given first to each subject to avoid contamination of these procedures by earlier anxiety induction. Each test began with a 3-min base period, which was followed by instructions from the experimenter to anticipate the task and by a subsequent 1 -min wait. The experimenter then instructed the subject to complete arithmetic Form C (Form-D in the second, fear-group-irrelevant behavior test) with instructions identical to those used in the Lab Session 1 pretest. This 90-sec procedure measured cognitive performance deficit during anticipation of fear exposure. Next, the experimenter read instructions for the confrontation task itself.

The snake and speech exposure tasks each consisted of seven 30-sec steps. Prior to each step, the experimenter asked the subject’s permission to proceed. When the subject agreed, the experimenter started the computer’s 30- sec sampling of physiological measures and implemented the step. When the subject refused or when he or she completed the seventh step, the experimenter told the subject that the task was over, withdrew with the snake or audience, and initiated 3-min posttest sampling of physiological responses. Then the 3-min baseline at die beginning of the second (fear-group-irrelevant) task set was begun. At four times during each task set, the subject rated current emotional arousal–alter initial base, after anticipation, after the arithmetic test, and after the last confrontation step that he or she completed.

The snake-exposure test consisted of moving the snake toward the subject’s chair from a point 4 m away in the hall at the open door to the subject’s room. Each step was preceded by the experimenter’s queries, “Would you please permit me to (1) remove the cloth, (2) move the cage just inside the door, (3) move the cage halfway to your chair, (4) move the cage next to your chair, (5) remove the lid and touch the snake, (6) remove the snake from the cage and hold it?” and (7) “Would you please lean forward and touch the snake?” For Step 6, the experimenter grasped the snake behind the head in one hand and lifted the anterior third of the snake from the box. For Step 7, the experimenter moved his hand a few inches toward the subject, rotating it to turn the snake’s head away from the subject.

The speech-exposure test consisted of two speeches, each of which was preceded by a preparation period. The first speech was simply the reading aloud of a typewritten paragraph. The second speech was an extemporaneous presentation based on some suggestions that the subject was given on a typewritten page. The five- member audience observed the entire speech procedure. Videotaping was used as an added stressor at the end of the task. Five discrete steps covered seven measurement periods (Steps 1-3 covered the first three measurement periods, Step 4 covered periods four and five, Step 5 covered six and seven): Step 1–“Would you please look this over for 30 seconds in preparation for reading it aloud?» Step 2–“Would you please read the passage to the audience now?” Step 3–“There is the topic for a 2- minute speech. In addition to the live audience, we will videotape the final minute for later analysis. Would you please take 30 seconds to prepare the speech?” Step 4–

, “Would you please begin your speech now”? Step 5– Experimenter aims the camera at the subject and starts the video recorder.

Following the two exposure tests, additional imagery trials were presented. The procedure for each trial was like the one described’for Lab Session 2; however, subjects received two neutral scenes that were different from the three scenes that were used in the previous imagery session, followed by two scenes (one snake, one speech) referring specifically to the content of the actual exposure tasks just performed. These two scripts contained arousal response’ references, balanced as before with respect to physiological systems. Presentation order of snake and speech scenes was roughly counterbalanced across group and sex. A flow chart of the entire procedure is presented in Table 2.

Data reduction

Physiological and self-report rating data were acquired on-line by the computer, which constructed a histogram for each physiological channel during each measurement period. These data included: heart period, EMG-inte- grator reset period, respiration period, skin conductance, and respiration amplitude (the former three measures timed on a 250-Hz clock, the latter two measures digitized 10 samples per second). For each respiratory cycle, the ratio of inspiratory to expiratory period was computed. From each histogram, median and interquartile range (IQR) were extracted off-line and converted to seconds for heart and respiration period, volts/minute of integrator output for EMG, or micromhos for conductance (resolution better than .02 pmho). Median heart period was then converted to heart rate. Respiration amplitude could not be calibrated in absolute units and remained in A/D units, but between-subjects differences were minimized by calibrating the channel at the beginning of the session to a standard arc of pen excursion for each subject. To reduce between-subjects variance because of basal differences and basal changes over time, median and IQR values for each measure were converted to change scores by subtracting the imagery-trial rest- period value from the read-, image-, and recover-period scores, and by subtracting initial baseline levels (prior to anticipation) from each data collection cell for each exposure-task period.

Table 2

Procedure Flowchart

Event Procedure
Mass testing Abbreviated snake and speech-anxiety questionnaires


Snake-phobia scale (SNAQ) and speech-anxiety scale (PRCS) Clinical interview Consent form
Session 1 Lab familiarization No-stress math performance test Questionnaires General imagery General anxiety
Session 2 Neutral, fear, and arousal imagery trials
Session 3 Primary stress test Anticipation Math performance test Avoidance test Secondary stress test Anticipation Math performance test Avoidance test Fear imagery trials

Note. SNAQ = Snake Anxiety Questionnaire; PRCS = Personal Report of Confidence as a Speaker.

Statistical analysis

The main statistics are univariate analyses of variance, accomplished for each dependent measure. The analyses of the exposure context included between-subjects factors of fear group and sex, and the within-subjects factor of stress (primary vs. secondary, as defined by fear-group assignment and order). Tasks (anticipation, math, and exposure) were analyzed separately. For some analyses, sample period within the task was included as a variable. Where warranted by initial hypothesis and/or significant statistical interaction, subsequent tests were applied both to differences between groups within stressors and differences between groups within exposure tasks. The imagery analysis included the same between-subjects factors of group and sex, with scene content and image period as within-subjects factors. Preexposure (Session 2) and postexposure (Session 3) were analyzed separately. Subject attrition rates during exposure were analyzed with a survival-analysis program (bmdpil), which provides approximate chi-square tests of group differences.


The anticipation, math, and exposure results from Session 3 (verbal report, performance, and physiology) are presented first to establish the pattern of the primary-fear data. The imagery findings from Session 2 (preexposure) and Session 3 (postexposure) follow.

The Psychophysiology of Anticipation

Arousal ratings. Subject reports of arousal during exposure tasks are presented in Figure

  1. At the beginning of the experiment (Base
  2. I) , the two fear groups were not different on this measure. Furthermore, socially anxious and snake-phobic subjects increased by similar amounts in reported arousal when informed (anticipation) that they would shortly be exposed to their primary stressor. That is, snake-phobic subjects waiting to view a snake and socially anxious subjects waiting to give a speech reported the same level of arousal. It can also be seen in Figure 2 that the arousal reports of subjects anticipating their primary stressor are significantly greater than reports obtained from these same subjects prior to each groups’ secondary stress: stress, F(l, 21)=21, p< .007.

During the math test, snake-phobic subjects reported the same high arousal level (greater than Base 1 but somewhat less than the anticipation period) prior to both snake and speech exposure. However, socially anxious subjects tended to be more aroused during the math test before the speech than during the math test before snake exposure: stress, F(l, 21) = 4.34, p < .05; Group X Stress, F( 1, 21) = 3.55, p < .08; Stress 1 versus Stress 2 for the social-anxiety group, F( 1,

  1. II) = 9.67, p < .01; and Stress 1 versus Stress 2 for the snake-phobic group, F < 1,

Figure 2. Arousal ratings obtained at the completion of each stage of behavior assessment. (The continuous line extending from Base 1 to Base 2 shows ratings obtained to the first exposure sequence, in which each fear group confronted their own, relevant fear situation [Stress 1]. The lines unconnected to base values show arousal ratings subsequently given by each fear group to the other group’s relevant fear situation [Stress 2]. Antic = anticipation.)

High arousal level was reported at exposure to the primary stressor for both groups.

Furthermore, snake-phobic subjects indicated a comparable high arousal when exposed to their secondary stressor (i.e., the speech). However, socially anxious subjects tended to show a difference in degree of arousal experienced with speech and snake exposure, responding strongly only to the former: stress, F( 1,21) = 7.24, p < .02, Group X Stress, F(l, 21) = 3.17, p < .09; Stress 1 versus Stress 2 for the social-anxiety group, F(l, 11) = 7.43, p < .02; and Stress 1 versus Stress 2 for the snake group, F < l, ns. Subjects showed higher arousal reports at Base 2 than at Base 1: stress, F(l, 21) = 20.7l,p < .0002. Furthermore, the socially anxious subjects tended to report more arousal at this second base than the phobic subjects: Group X Stress, Д1, 21) = 4.30, p < .06.

Physiological pattern. The base heart rate and skin-conductance measures obtained prior to imagery and prior to anticipation/ math/exposure suggested a trend toward higher values for snake-phobic subjects. However, none of the individual Group X Sex analyses achieved significance; thus, uncorrected change scores are used in subsequent statistical analyses. Unlike the verbal report data, there is no clear tendency for heart rate to be higher at the anticipation period of the primary stressor than of the secondary stressor. Although mean heart rate response was higher for snake-phobic subjects than socially anxious subjects at all measurement points during anticipation, separate tests of group differences for anticipation of snake exposure and speech test did not approach statistical significance.

Subjects showed higher math heart rate prior to their primary stressor than prior to their secondary stressor: stress, F{ 1, 21) = 6.52, p < .02. The snake-phobic subjects had a mean , math heart rate response prior to snake exposure that was nearly twice that preceding the speech (13 vs, 7 beats/minute [B/M]). The difference in stress type heart rate for the speech-anxious subjects was somewhat less (11 B/M for the speech test and 6 B/M for snake exposure). The Group X Stress interaction term fell short of an acceptable confidence level, F( 1, 10) = 2.80, p< .11.

The skin-conductance response was concordant with verbal report of arousal during the anticipation period. That is, both fear groups responded more in anticipation of their primary stressor than of their secondary stressor: stress, F{\, 21) = 9.37, p < .0006. This result was also observed for the math test, although somewhat less emphatically: stress, F(l, 21) = 7.24, p < .02. Math skin conductance was unlike verbal report in that there was no trend toward differences between fear groups for one exposure task or the other.

In summary, no clear difference between stressors was found for the heart-rate response during anticipation. However, with the subsequent math test, greater heart rate activity was observed before primary than before secondary stressors. Electrodermal activity prior to the primary stressor was significantly greater than prior to the secondary stressor, and this effect persisted into the math test; that is, anticipated snake exposure prompted the phobic subjects to give the greater response relative to socially anxious subjects, whereas the latter group responded more when they were told about the speech.

Performance Measures of Fear

Math test. No overall differences between groups in math performance were observed. Furthermore, subjects improved their performance over the three presentations of the math test: test, F(2, 42) = 15.59, p < .0001. Thus, no broad effect of stress on math performance was apparent. However, the snake- phobic subjects tended to improve more than the socially anxious subjects over successive test administrations–Group X Test, F(2, 42) = 2.49, p<.10; Group X Test (linear trend), F( 1, 21) = 4.30, p < .06–which prompts the hypothesis that socially anxious subjects may suffer a relative performance deficit with repeated testing under stress.

Exposure. The behavioral measure of exposure recorded here was the step level achieved just prior to refusal. The subject population is small and differences between groups were generally not statistically significant. However, the trends in the data are interesting. First, there was little difference in the refusal pattern for primary and secondary stress. Subjects were almost equally likely to refuse a step, regardless of whether the stress situation involved their primary fear or a secondary stress situation. Second, the two tasks ultimately prompted the same level of refusal but tended to differ in pattern over steps. Thus, snake exposure prompted a monotonic, almost linear attrition, whereas the speech task showed almost no refusals at the outset, increased abruptly in the middle range of the task, and had fewer subjects who refused near the end of the exposure session. The only variable that occasioned clear differences in avoidance was sex: Females, regardless of stress task, showed more refusals than males at all step levels (Breslow x2 =

5.06, p < .03). Finally, again regardless of task, the snake-phobic subjects tended to be more avoidant than the socially anxious subjects in that they were more likely to refuse at an earlier stage of the test. On the average, female snake-phobic subjects were the earliest and most probable avoiders. However, neither of the latter trends was statistically significant.

The Psychophysiology of Exposure

The two exposure tasks differed greatly in the physical and mental requirements of their performance. Thus, a Group X Stressor analysis, such as that accomplished for anticipation or anticipation/math, would have little relevance to affect in exposure, except perhaps for the first test step. At that point, most subjects are present and, regardless of task, subjects had not yet begun an overt, active participation in events. Therefore, a Group X Sex X Stress analysis of heart rate and skin conductance is presented only for this first step; differences between groups over succeeding steps are considered separately for each exposure situation.

Analysis of first step heart rate clearly showed a greater response in both groups to the speech task. Preparation for a speech prompted a heart rate change score of nearly 15 B/M, whereas the beginning of snake exposure led to only a 4 B/M average change for both fear groups, F(l, 20) = 13.40, p < .002.3 The mean response of snake-phobic subjects was roughly 2 B/M faster than that of the socially anxious subjects for both tasks, but the simple group difference did not achieve statistical significance.

An overall difference in amplitude of response to the two task effects was not obtained for skin conductance. However, the skin-conductance response of both groups was, as expected, greatest to their primary stressor: stress, F(l, 20) = 7.80, p < .02. Furthermore, the socially anxious subjects generated the largest responses when they began the speech-exposure task (4.47 vs. 2.35 /umhos for phobic subjects); these anxious subjects were profoundly less responsive to snake exposure (1.89 jitmhos), whereas snake-phobic subjects responded in the same range as to the speech (2.64 /umhos). This interaction was confirmed by statistical test: Group X Stress, F(l, 20) = 4.71, p < .05.

The separate patterns of heart rate and skin-conductance change over successive steps of the two exposure tests are presented in Figures 3 and 4. Two subsamples of each group are illustrated and analyzed: subjects who completed four or more steps on the exposure test and subjects who completed six or more steps. The former analysis permitted a systematic inspection of the subjects who were most frightened according to a behavioral criterion (e.g., refusal to continue); however, the sample is small and data on the complete test is perforce unavailable. The latter analysis excluded some of the most frightened subjects but permitted the inspection of nearly the entire test period.

Snake exposure. At the onset of the exposure cpndition, heart rate was reduced for the average subject relative to his or her response at the end of the math test. This reduced heart rate continued into the second step. Over subsequent steps, heart rate increased linearly for snake-phobic subjects, with a maximum response at the final step. Socially anxious subjects showed no such heart rate increase, presenting an essentially flat curve over all steps of the exposure test. Analysis of the four-step data yielded the following: group, F\l, 16) = 2.93, /><.11. The six-step analysis confirmed the greater heart rate response of snake-phobic subjects– group, F(l, 11)= 12.39, p < .005–and the pattern of increasing difference over steps– Group X Steps, F{5, 55) = 3.96, p < .004.


The pattern of skin-conductance response was the same for all groups over the course of the snake-exposure tests. Subjects showed an initial reduction in conductance until Steps 3 or 4, followed by a subsequent increase. Although die phobic subjects appeared to show a greater response than the Speech subjects, statistical analyses do not provide clear support for this conclusion: four-step group, Д1,16) = 2.58,/? < .13, and six-step group, Д1, 11) = 2.05, p < .18.

Figure 3. Snake-phobic and socially anxioits groups’ average heart rate and skin-conductance change scores over successive steps of the snake-exposure test. (Data are presented separately for all subjects who completed at least four steps [i.e., the snake is in a closed cage directly in front of die subject) and for the reduced sample of subjects who continued on to the sixth step [i.e., the experimenter removes the snake from the cage and holds it in front of the subject]. Physiological responses during the last period of the math test that just preceded exposure are also shown. B/M = beats per minute.)


Speech exposure. Like the data on attrition pattern, subjects’ heart-rate response over the steps of speech exposure do not suggest a linear, progressive increase in fear. Maximum response is found for the average subject at Step 4, rather than at the end of the test. The other two most obvious facts are that (a) the overall increase in heart rate is very large, with a mean response for speech subjects twice that obtained for snake-phobic subjects during snake (fear-relevant) exposure and (b) the mean response of snake-phobic subjects, contrary to prediction, is actually slightly greater than that for the socially anxious group. Group Fs were not significant: four step, Д1, 14) = 2.37, p < .14, and six step, Д1, 11) = .39.



Figure 4. Snake-phobic and socially anxious groups’ average heart rate and skin-conductance change scores over successive steps of the speech-exposure test. (Data are presented for all subjects who completed at least four steps [i.e., subject agrees to present a speech and performs during the early part of the first speech period] and for all subjects completing six steps [i.e., subject presents speech with the television camera on during the early part of the second speech period]. Physiological responses during the last period of the math test that just preceded exposure are also shown. B/M = beats per minute.)



The mean skin-conductance response of socially anxious subjects to speech exposure was larger than for either group at snake exposure. Furthermore, the results for skin conductance during public speaking were more consistent with stress relevance in that the socially anxious subjects tended to yield the greater response. However, the statistical support for this conclusion is weak: four-step group, F(l, 14) = 4.49, p < .06, and six-step group, F( 1, 11) = 2.12, p < .18.

Exposure and Anticipation: Summary

Snake exposure. In general, socially anxious subjects gave relatively little evidence of psychophysiological arousal prior to or during snake exposure as assessed by verbal report, heart rate, and skin conductance. Snake- phobic subjects did give considerable evidence of arousal on these measures, but the pattern of response varied from anticipation to actual confrontation. Thus, snake-phobic subjects showed high relative verbal report of arousal and high skin conductance at anticipation. The elevated skin conductance continued into the math test, although verbal report was no more elevated than for nonphobic subjects. Heart rate was not significantly higher for phobic subjects than for anxious subjects at any point prior to exposure or even at the first exposure steps, but it then showed a pattern of significant progressive increase to the end of the tests. Skin conductance was also elevated for phobic subjects during actual exposure but was not significantly larger than the response of the anxious subjects.

Speech performance. Socially anxious subjects showed high heart rate and elevated skin conductance and reported high arousal prior to the speech. Snake-phobic subjects showed only the anticipatory heart rate response. During the speech, both groups again showed large heart rate responses and both reported similar levels of arousal. Although the mean skin-conductance response during the speech tended to be larger for socially anxious subjects than for phobic subjects, the statistical tests of this difference were below the acceptable confidence level.4

Imagery and Emotional Response

Scene ratings. Vividness and arousal scene ratings for preexposure imagery are presented in Figure 5. Although mean vividness ratings were generally higher for phobic subjects, the difference between groups was not significant–nor was significance found in tests of non-fear-group-relevant scenes for any other vividness variable.

For preexposure imagery, there was a clear relationship between specific fear content and fear group: Group X Content, F(l, 21) = 16.94, p < .001. Each group found its own content more vivid. Socially anxious subjects had marked difficulty in visualizing snake content. Their images of snake scenes produced the lowest vividness ratings for any subcell of the experiment. However, when these scenes were converted to specific memory images in the postexposure imagery task, vividness reports were more homogeneous, and the interaction between fear content and fear group did not approach significance.

The danger and sex scenes generated the highest reported arousal among the general (non-fear-group-relevant) scenes: content, Щ4, 84) = 5.79, p < .0004. We were surprised that subjects reported no more arousal to the exercise scene than to the neutral scenes. The sexes did not differ consistently in arousal report except for danger scenes, which were reported to be significantly less arousing by females than by males: Sex X Content, F(4, 84) = 3.80, p < .007.

Both groups reported more arousal to preexposure images of their own central fear than when they visualized material relevant to the other group’s fears: Group X Content, F(l, 21) = 22.95, p < .001. This effect was perfectly replicated in the postexposure imagery trials: Group X Content, F(l, 21) = 17.82, p < .001. However, the difference between groups on the speech scene was small and insignificant according to subsequent tests (i.e., both groups found the speech scenes relatively arousing). Conversely, the difference in response of the two groups to the snake scene was dramatically large and significant (p < .05): Phobics reported considerable activation, whereas socially anxious subjects were hardly aroused. The fact that these differences appeared for postexposure (when there were no significant differences in vividness between contents) as well as preexposure imagery suggests that the finding is not related to one group’s unfamiliarity with the other’s stimuli but represents a clear difference in the arousal-response disposition in these two populations.

Physiological pattern. The physiological reaction of the two fear groups to the recording session appears to have been different. Thus, preexposure imagery heart rate levels were somewhat higher for focal-phobic subjects than for socially anxious subjects: group, Щ, 21) = 3.40, p < .08. And skin- conductance level was significantly higher for socially anxious subjects than for phobic subjects: group, F{\, 21) = 4.45, p < .05. No differences were found for respiration or muscle-tension level. .

During preexposure scene visualization, subjects tended to show positive increase in heart rate during arousing scenes. However, skin conductance showed a pattern of habituation over the session. Thus, the skin-conductance response of subjects to arousing, scenes is generally a slowing of this habituation (i.e., a less negative score) rather than a positive increase.


Figure S. Average vividness and arousal ratings for the two fear groups over all imagined scenes: neutral (NTL), danger (DNG), sex (SEX), exercise (EXC), speech, (SPH), and snake (SNK).

The mean change scores during imagery tend to parallel those obtained for the exposure condition tests. That is, snake-phobic subjects appear to respond viscerally to both speech and snake content, with the largest mean heart rate change occurring in response to the snake scene. Snake and exercise contents are the only scenes that occasioned a positive skin-conductance response in these phobic subjects. In contrast, socially anxious subjects showed appreciable heart rate acceleration only to the public speaking scenes and less average skin-conductance habituation to speech than to snake contents. Nevertheless, despite this suggestive pattern of mean values, none of the main hypotheses concerning content and group differences was supported by statistical test.

Analysis of postexposure imagery produced a similar pattern of results, with the mean physiological responses of socially anxious subjects somewhat larger in speech- memory imagery than in snake-memory imagery. Again (particularly for heart rate), focal-phobic subjects tended to respond strongly to both types of fear-group-relevant scenes. However, none of the pertinent tests of postexposure imagery means achieved an acceptable confidence level.

Discussion Anticipation and Exposure

The present experiment clarifies inconsistencies in the literature concerning the visceral arousal response of socially anxious and focal-phobic subjects. First, it is now apparent that socially anxious subjects do not show more trait arousal or more reactivity to fear stimuli than do focal-phobic subjects. Second, our findings suggest that the exposure- test arousal differences previously observed (Borkovec, 1973) were entirely attributable to task factors and did not represent differences in emotional responsivity to fear stimuli.

The task of giving a speech involves cognitive load and motor performance requirements that in themselves occasion marked cardiovascular reactivity. Large increases in heart rate both prior to and during public speaking have been found in unselected subjects by Hastings (1968), and Knight and Borden (1979) have recently reported a failure to distinguish between the heart rate responses of speech-anxious subjects and normal subjects during speech performance. Similarly, in the present experiment the two fear groups generated an equally marked heart-rate change during speech exposure. Although it is possible, of course, to conclude from this that everyone is afraid of public speaking, such a devaluation of the fear construct fails to consider the fact that subjects do differ in verbal report of fear, anticipatory skin conductance, and to some extent, in performance deficit. A more conservative (and at the same time heuristic) interpretation is that the cardiovascular demand required by a speech task is so pronounced that it masks any evidence of ah affective component in this system.

Unlike giving a speech, sitting quietly in a chair and observing a snake make few intrinsic demands on the cardiovascular system. Lacey and Lacey (1970), Obrist et al. (1974), and Graham and Clifton (1966) all agree that simple attention to an external stimulus (even one with negative valence) is most often associated with a reduction in heart rate and a diminution of muscle tension. Thus, it is not surprising that we found a smaller heart rate response, across subjects, to snake exposure than to speech exposure. Furthermore, we are encouraged to conclude that any increase in heart rate over base level observed in the snake context is more likely to be related to a conditioned affective response to the stimulus than to intrinsic task demands. In fact, the mainly non-snake-fear- ful, socially anxious subjects displayed zero mean heart-rate change during snake exposure (and also reported low arousal). Snake- phobic subjects, however, showed a significant progressive increase in heart rate over the course of snake exposure (and reported high arousal).

These findings show that different measures held to index fear produce very different results, not just as a function of stimulus- fear relevance but depending on characteristics of the subject population examined (e.g., sex) and the context of fear measurement. This is true both across the three general response systems (language, action, physiology) and for different measures within a system. Thus, heart rate and skin conductance do not constitute interchangeable measures of physiological fear arousal. Subjects reported that they were more aroused when anticipating exposure to their primary stressor than prior to exposure to their secondary stressor. This same result was obtained for the skin-conductance response, but it was not found for heart rate. During the exposure task itself, both fear groups reported high arousal to speech exposure; however, only snake-phobic subjects reported that snake exposure was arousing. This latter pattern of reactivity was also found for heart rate change; that is, snake-phobic subjects acceleration for both fear stressors, but socially anxious subjects showed significant heart rate increase only to the speech. This exposure pattern over groups and stressor was not clearly present for the skin-conductance response.

Performance Measures of Fear

No evidence was obtained suggesting that mathematical reasoning is less adequate when subjects are tested prior to a relevant fear exposure than when they are tested before a presumably less relevant stress. Although there was a tendency for socially anxious subjects to perform less well than phobic subjects over trials, this was not stress specific and may express a broader problem with evaluation anxiety in these subjects.

The tendency to avoidance was also more strongly determined by subject characteristics than by specific fear situation. Thus, consistent with society’s stereotype, females were significantly more avoidant than males. Phobic subjects tended to be somewhat more avoidant than socially anxious subjects. However, if we accept this trend as meaningful, it still represents a broad response disposition embracing both stress situations rather than a specific response to snake exposure.

Why relevance of stress context did not more significantly influence overt performance (as it did physiological and verbal responses) is not clear. However, it is possible that the subjects’ complete freedom to leave at any time robbed these situations of much of their threat. Social demand in avoidance tests can prompt considerable variance in responding, and it is extremely difficult to control. Previous research has shown that subjects admonished to try hard to tolerate a phobic situation (a high-demand condition) show significantly less avoidance than subjects performing without such social pressure, although under both conditions, verbal report of distress is equally high (Hodgson & Rachman, 1974; Miller & Bernstein, 1972). The attempt of the examiner to be neutral in this context may have been a bad strategy: Although both manipulations raise ethical problems, avoidance under social pressure or performance in an inescapable situation may provide more reliable estimates of fear.

Profiles of Fear

The above cavils aside, the trend toward differences in avoidance behavior between snake-phobic and speech-anxious subjects is not inconsistent with the clinical picture of these fears. Social stimuli are not readily avoided, and patients who report themselves to be socially anxious do not usually describe a pattern of social withdrawal. Rather, they report frequent social interactions that they find painful. Treatment more often involves an effort to increase tolerance of necessary performances than an initiation of approach behavior. On the other hand, avoidance is a defining behavior of focal-phobic subjects. In the clinic, these patients, who are more likely to be female than male, usually come to treatment only because some change in life circumstances now renders avoidance ineffective.

The psychophysiological findings also seem consistent with the above model. Only the focal-phobic subjects showed the pattern of significantly increasing heart rate over exposure steps and the increasing probability of step refusal. This suggests cardiovascular mobilization for action-preparation for flight from a stressor. The high heart rate, which was maintained from the beginning of the task by both fear groups, is also consistent with cardiovascular activity as part of response expression–in this case serving the cognitive and motor requirement of public speaking. Thus, these data are certainly consistent with the motor/metabolic explanations of heart rate change under stress put forward by Obrist, Webb, Sutterer, and Howard (1970) and elaborated by Elliot (1970) to include both “current and pre-current action” (Elliot, 1972, p. 451). On the other hand, increase in skin conductance has generally been associated with orienting behavior, alertness, and heightened attention (Pro- kasy & Raskin, 1973)–often in the absence of a motor response. Heightened orienting behavior is certainly appropriate to a setting in which one anticipates a novel stress. It is also a likely reaction from someone trapped in an uncomfortable and demanding social situation (as the socially anxious subjects giving a required speech). In the present experiment’it is in the latter context that group and stress differences in skin conductance approached significance. If emotion is best construed as an action set (Lang, 1977a, 1983), then indeed its physiology is best understood in terms of those behaviors (overt and in preparation) that are demanded by the task context.

There are few systematic data on these issues available from the clinic population. However, we have administered the same procedures used here (exposure test and imagery trials) to a small patient series at the University of Wisconsin Psychology Department clinic. The preliminary results show considerable consistency with the present findings (Levin, Cook, & Lang, 1982). First, the phobic patients tend to be more avoidant than the anxious patients: Five of six socially anxious patients from this group completed a speech test prior to therapy, whereas none of three focal-phobic patients completed an initial fear-exposure test.5 The socially anxious subjects all showed high heart rate while speaking and high skin-conductance values. The focal-phobic subjects showed a progressive increase in heart rate over steps during exposure. The overall skin-conductance response of phobic subjects during snake exposure was less than that of the socially anxious subjects during public speaking. These parallels with the experimental sample are provocative and encourage exploration of this psychophysiological model of fear in the clinical environment.


Analysis of subjects’ imagery ratings indicated that each fear group found their own most relevant fear content to be more vivid and more arousing than the relevant content of the other group. As in previous studies, mean vividness and arousal values for fearrelevant stimuli were slightly higher for focal- phobic subjects than for socially anxious subjects. In the present experiment, however, this difference did not approach statistical significance. Furthermore, although the pattern of heart rate and skin-conductance responding was similar in imagery to that yielded by each group at actual exposure (again, with the focal-phobic subjects yielding the predicted, higher mean responses), statistical tests did not confirm these relationships. Thus, the hypothesized consistency of psychophysiological pattern across contexts of fear measurement was not demonstrated. Although this result is disappointing, a reading of the imagery literature suggests that it could have been expected. Marks and Huson (1973) examined a series of experiments that were conducted with fearful and anxious patients at the Maudsley Hospital in London. They found that heart rate increase discriminated neutral from phobic imagery in only three of five studies, whereas skin conductance was significantly greater for phobic than for neutral imagery in only two of these experiments. This experience with patients’ imagery response is similar to our own. In Figure 6, the average heart rate response occasioned by personally relevant fear imagery, taken from a clinical sample (20 anxious and fearful patients seen at the University of Wisconsin Psychology Department clinic; Levin et al., 1982), is compared with results obtained here for volunteers responding to the standard fearful scenes. Note that in both cases the average response is quite modest and does not differ greatly in amplitude between patient- and college-volunteer samples. Nevertheless, among both volunteers and clinic patients there are individuals whose imagery response is quite large, equaling reactions observed in actual exposure.

Response Stereotypy and High Fear

Although both laboratory and clinical findings confirm that individuals professing fear arousal generally respond with visceral activation when exposed to the relevant stimuli, it is also true that the correlation between these response systems is far from perfect. Furthermore, the strength of the relationship between physiology and fear might be expected to deteriorate still more when these variables are assessed in the imagery context. Nevertheless, it is generally agreed that there should be considerable consistency for subjects who are judged to be very high in fear. Indeed, the ubiquity of the fear response, regardless of the degree of stimulus provocation, is part of the definition of high fear.

An effort to illustrate this phenomenon is presented in Figure 7.6 In preparing this illustration, all of the subjects in Experiment 1, regardless of group assignment, were ranked separatedly on the SNAQ and the PRCS questionnaires. The five subjects who scored the highest and the five subjects who scored the lowest on each questionnaire were then included in the frequency bar graph in Figure 7, if and only if they were also among the five highest scorers on a particular fear measure (avoidance, arousal report at exposure, etc.). This was done separately for high- and low- SNAQ scorers for snake-exposure assessment and for high- and low-PRCS scorers on the speech test.



Figure 6. Frequency distributions of heart rate change during imagery for the Experiment 1 subjects (N = 25) and for a sample of patients from the University of Wisconsin Psychology Department Fear Clinic (N = 20). (The curves use data from the image period of the research sample’s fear-relevant scene [snake and public speaking] and from the image periods of two separate clinical scenes administered to the patients according to the same procedure employed here. The scripts for these scenes were created by the patient and his or her therapist together. They describe fear situations central to the patient’s clinical problem. B/M = beats per minute.)



It is apparent from this figure that the most frightened subjects according to the SNAQ are also likely to score high on performance, other verbal, and physiological measures of phobia, both in exposure and imagery. In fact, the two subjects ranked one and two on the SNAQ were also among the first five ranks on all other measures. Only two instances of marked disparity are apparent–one low- SNAQ subject was among the subjects reporting the highest arousal in anticipation of exposure and another low-SNAQ subject was among the highest skin-conductance reactors during snake imagery.

There is much poorer stereotypy among high-fear subjects on the speech test. High scorers on the PRCS completed as many steps on the speech test as did low scorers. Heart rate similarly fails to discriminate between the high- and low- questionnaire groups under any condition of measurement. One explanation might be that the questionnaire is an inadequate measure of speech anxiety. However, it is highly face valid and consistent with interview data, and clinic patients seeking treatment for social anxiety obtain high scores (see Footnote 5 and Figure 1). Furthermore, the data were no more orderly when organized around any of the other measures. Thus, it may be that this is simply a poor context for assessing the problem. At the least, it is obvious that heart rate as a fear measure is seriously compromised by task demands.



Figure 7. Coincidence of high-fear responses across different contexts of fear measurement for subjects scoring highest and lowest on the snake (SNAQ, upper bar graph) and speech (PRCS, lower bar graph) questionnaires. (SNAQ = Snake Anxiety Questionnaire; PRCS = Personal Report of Confidence as a Speaker; EXPO = exposure; ANTC = anticipation.)



In summary, in the case of snake-fearful subjects, we are encouraged to believe that avoidance, report of arousal, heart rate, and skin-conductance reactivity are all part of a central program, or phobia prototype, that can be activated by a variety of stimulus situations. The dimensions of fear responding for professed speech-anxious subjects appear to be fewer, and the overall pattern is less consistent. However, of the measures assessed here, only felt anxiety about speaking (arousal report or questionnaire) and the sweat-gland response gave any indication of stability across stimulus contexts in high-fear subjects.

The Imagery Task

Like the direction of the overall group means, the data on high-fear subjects in Figure 6 argue that similar patterns may exist between psychophysiological response in imagery and in exposure. The failure of Experiment 1 to set this relationship in clear relief is a likely consequence of the twin problems of high variance of response and the difficulty in obtaining a large sample of suitably frightened subjects. Important contributors to the former problem are the large individual differences between and within subjects in capacity for image generation (Sheehan, 1972). Even though fearful imagery, dreams, and troubling thoughts about fear objects may be part of the clinical picture of phobia and anxiety, it is quite another matter for the patient or subject to evoke these cognitive events on command in the laboratory or clinic. Previous research (McLean, 1981; Miller et al., 1981) has shown a significant relationship between questionnaire reports of subjects’ general tendency to image vividly (QMI) and physiological response to imagery in the laboratory. The present small sample also yielded consistently low, positive correlations between vivid imagery and physiological response (e.g., .20 and .17 for heart rate during the snake and speech scenes, respectively).

We have suggested here and elsewhere (Lang, 1977a, 1979, 1983) that the image is a propositional template for overt responding. We assume physiological pattern identity between the perceptual-affective response in the reality context and in vivid imagery. Thus, modification of either process is hypothesized to alter the other. We assume this to be the mechanism through which imagery functions as a vehicle for therapeutic change. It may also be inferred that imagery therapies will fail if the identity assumption is not met. Lack of identity between perceptual process and image process can occur either because of a primary failure to code the relevant response information (along with information about stimulus context) at the time of initial perception or because of individual differences in capacity for later image generation. In a recent article, we described a methodology for enhancing imagery based on the reinforcement of postimagery verbal; descriptions of familiar scene contents (Lang et al., 1980). This method is presumed to increase imagery vividness, not by adding new code but by training the subject to access response elements already in memory storage, which are part of the fear prototype but are not always elicited by instructional cues. In Experiment 2 we studied the fear imagery of a new sample of socially anxious and focal- phobic subjects under two conditions of imagery enhancement–stimulus and response training. We hypothesized that response- trained, subjects would generate significant, differential patterns of physiological activity in fear imagery and that these patterns would match those obtained in Experiment 1 during exposure (when subjects were objectively confronted with these same fear contents).

Experiment 2

Experiment 2 was designed to examine the effects- of stimulus- and response-training

procedures used by Lang et al. (1980) on the emotional imagery of fearful subjects. This previous research showed that increasing stimulus content in the imagery of normal subjects (i.e., through training designed to add more context-descriptive propositions to the image structure) did not enhance physiological reactivity during image processing. However, training that promoted response- propositional elements in the image led to increased physiological activation in imagery. Furthermore, the resulting pattern of physiological responding was consistent with both the emotional themes (e.g., fear, action, relaxation) and the response descriptors in toe script (e.g., “sweaty palms,” “pounding heart”).

The present experiment examined these same variables with a new sample of snake- phobic and speech-anxious subjects. Half of each fear group underwent the previously developed response-oriented imagery training, and toe other half was administered stimulus training. All of the subjects were tested on toe same scripts (including both stimulus and response propositions), thus providing an unconfounded test of toe effects of training on subsequent imagery responding. The script contents were similar to those studied in Experiment 1: specific snake- and speech-fear scenes, general fear or danger scenes, emotionally neutral arousal (exercise) scenes, and standard neutral (tranquil) scenes, presented sequentially in a semirandomized order. The use of a variety of scripts again provided an opportunity to examine the response of these populations to specific and general emotional content and the interaction of both with toe training and fear-group variables.

As in Lang et al. (1980), we predicted that the physiological response in imagery of stimulus-trained subjects would be no different than that of untrained subjects. Thus, this procedure can be considered a control for subject/experimenter relationship factors or demand characteristics.

On toe other hand, we anticipated that response training would enhance physiological responding to all arousing scenes (but not neutral material). Furthermore, again consistent with Lang et al. (1980), we expected that fear scenes would show patterns of visceral and somatic reaction that were different from those occasioned by arousing, but non- fearful, images (e.g., skin-conductance, heart rate, and respiration response elevation with fear scenes; heart rate and muscle tension salient in exercise imagery).

We further hypothesized that the pattern of physiological responding for the response- trained fear groups would be consistent with the exposure response of these populations in the previous experiment–and with the nonsignificant trends previously shown by Experiment 1 subjects in imagery. More specifically, we expected that response-trained focal-phobic subjects would show a large heart rate response to both snake and speech scenes but that the response-trained socially anxious subjects would show heart rate acceleration only to the public-speaking scenes.

In addition to physiological responses, verbal reports of scene vividness and arousal were examined. Relationships between fear group, physiological response, and questionnaire estimates of anxiety and imagery ability were also assessed.



Subjects were 40 University of Wisconsin volunteers selected from the introductory psychology class. The selection procedure was identical to that of Experiment 1 and yielded sex-balanced groups of 20 snake-phobic subjects and 20 speech-phobic subjects (see Figure 1 and Table 1). Snake-phobic subjects had higher SNAQ scores–group, F{ 1, 32) = 14.05, p < .001–whereas the speech group scored higher on the PRCS–group, F(l, 32) = 32.75, p < .0001.


Setting and equipment were identical to those used for the imagery session in Experiment 1.


After the screening interview, subjects were randomly assigned to one of the two experimental training conditions: stimulus-structured training or response-structured training. Each condition had five male and five female phobic subjects of each fear type, such that training, fear, and sex were completely crossed in this design. There were three more sessions–two group sessions of relaxation and imagery training followed by an individual imagery test session in the laboratory.


Training procedures were those reported by Lang et al. (1980). First, subjects were given progressive relaxation training involving tension-release cycles in eight muscle groups. This was done to reduce background physiological level and variability so that specific imagerelevant responses could be more clearly observed. The remainder of the first training session and all of the second session (1 hour per session, spaced 1 to 3 days apart) were devoted to imagery training.

For the stimulus group, the experimenter instructed subjects to attend to specific stimulus details in the scenes, presented stimulus-proposition-laden scripts to imagine, requested a description of the scenes from each subject, and reinforced inclusion of specific stimulus attributes in the subjects’ reports.

For the response group, the experimenter presented physiological-response-laden scenes, including physio- logical-response propositions as well as a subset of the stimulus material used in training the stimulus group. The experimenter requested a description of the scene from each subject after each image and reinforced reports of physiological involvement during imagery that coincided with physiological-response propositions included in the script material. All subjects were trained by one of two trainers (D.N.L. or G.A.M.) in groups of from two to five.


The laboratory session, which lasted 90 min, followed within a few days of the second training session. Subjects were tested individually by one of four experimenters so that 10 subjects, assigned for maximum balance across experimenters for the training, fear, and sex variables, were tested by each experimenter. With the exception of scene presentation order, the testing procedure was identical for all subjects. The procedure was the same as the one used for the imagery task in Experiment 1, with three exceptions: (a) Subjects completed the QMI a second time, at the beginning of the lab session, (b) subjects were reminded to imagine the scenes as vividly as possible, as practiced in the training sessions, and (c) only data from the final 30 sec of the 50-sec read period were > retained.

All subjects were given the same 11 scripts. The first script presented was emotionally neutral in content and contained no physiological-response instructions. It was a practice trial, excluded from statistical analysis. The remaining 10 trials involved one pair of scripts from each of five content categories: snake, speech, danger, exercise, and neutral. All but the neutral scripts contained both stimulus and response elements, and each script included one reference to each of the four physiological systems measured (cardiovascular, electroder- mal, somatic muscle, respiratory). Response references were physiologically arousing in nature, such as “Your heart races as you . . .” or “Sweat breaks out on your forehead at the . . .” Neutral scripts did not contain these response references.

Twenty pseudorandomized orders of script presentation were used with the following restrictions: (a) The 10 test scripts were divided into two blocks of five scripts, with each block having one scene from each of the five content categories, and (b) the first scene of the second block differed in content category from the immediately preceding scene. The 20 script orders were assigned to subjects so as to achieve maximum counterbalancing of presentation order for all of the eight Group X Training X Sex cells.

Data Reduction and Analysis

Data reduction was identical to the procedure followed in Experiment 1. Univariate analyses of variance were performed on each dependent measure. The experimental design had three between-subjects factors: group (snake vs. speech), sex, and training (stimulus vs. response). Imagery content was a within-subjects factor. For the physiological dependent measures, an additional within-subjects factor was trial period (read, image, recover).



Questionnaire results are presented in Table

  1. Fear groups did not differ significantly in scales assessing reports of physiological symptoms in anxiety but did differ significantly in reported feelings of insecurity, Д1, 32) = 5.43, p < .03. No significant difference was found between fear groups on the QMI; however, as has been observed before, mean values for the QMI were higher for focal-phobic subjects. Subjects reported increased vividness of imagery at QMI readministration following the training and physiological testing procedures, Д1, 32) = 19.53, p < .001.

Table 3

Mean Vividness and Arousal Ratings and Standard Deviations


Fear group Training Snake Speech Danger Exercise Neutral
Snake Stimulus


51.7 Vividness


50.4 38.2 51.7
  SD 10,0 8.5 17.3 20.9 . 14.5


58.8 46.7 50.3 42.6 51.0
  SD 7.1 16.1 16.6 18.9 18.4
Public speaking Stimulus


50.3 *53.4 50.2 45.4 58.8
  SD 15.7 9.2 15.8 18.0 3.5


45.4 54.8 51.2 40.8 50.7
  SD 15.23 9.2 14.1 14.5 13.9
Snake Stimulus


51.8 Arousal


45.1 32.5 24.5
  SD 15.1 12.5 18.9 15.7 18.6


52.8 40.0 42.8 29.4 13.6
  SD 18.0 18.5 20.3 21.6 17.0
Public speaking Stimulus


43.8 50.3 46.5 35.7 23.8
  SD 17.3 14.1 17.7 20.2 21.1


36.4 46.6 41.2 31.0 13.4
  SD 18.5 15.8 17.2 19.4 14.0

Note. The computer scale on which ratings were made is continuous (0-63).

Scene ratings. Mean vividness and arousal ratings for stimulus- and response-trained subjects for all scenes are shown in Table 3. Significant overall differences in content were obtained for both ratings: vividness–content, F(4, 128) = 9.48, p<.00001; and arousal–content, F(4, 128) = 28.34, p < .00001. However, the two training groups did not differ greatly in their subjective reports; only the neutral scenes had a large difference. Perhaps because the neutral scripts contained no response propositions, response-trained subjects rated them as less arousing than did stimulus-trained subjects, which resulted in a significant overall effect: training, F(l, 32) = 4.84, p < .04.

There was no overall difference between focal-phobic subjects and socially anxious subjects in their ratings of scenes other than snake or speech. However, as in the previous experiment, each fear group rated scenes based on their own anxious or phobic content as more vivid, F(l, 32) = 4.48, p < .05, and arousing, F( 1, 32) = 11.82, p < .002, than the nonrelevant fear scenes. Furthermore, response training clearly accentuated this effect for vividness ratings. Thus, for response- trained subjects, each fear group rated their own fear scenes as substantially more vivid than those of the other fear group, whereas stimulus-trained subjects showed the phenomenon in a much more attenuated form: vividness–Training X Fear Group X Content, Д1, 32) = 4.73, p < .04. Although the pattern of mean arousal values was similar, the interaction was not significant for this measure. Note that response-trained subjects showed the same distribution of vividness and arousal ratings across fear scenes as did the untrained sample from Experiment 1. (That is, the snake scene received the highest ratings by snake-phobic subjects and the lowest ratings by socially anxious subjects; the two groups gave a similarly high arousal rating to the speech scenes.)

Physiology of Imagery

Base levels. Analyses of variance of baseline physiology revealed no initial levels difference between stimulus- and response- trained subjects on any of 12 measures. The only differences found between fear groups were for two related measures of respiration: Snake-phobic subjects showed greater depth of respiration than socially anxious subjects, F(l, 32) = 5.63, p < .025, and a somewhat higher inspiration/expiration (I/E) ratio, F( 1, 32) = 7.40, p < .02.

Table 4

Means and Standard Deviations for Heart Rate and Skin-Conductance Change Scores During Read, Image, and Recover Periods for All Subjects


Period Snake Speech Danger Exercise Neutral


2.60 4.37 Heart rate


0.99 .00
SD 8.01 6.10 6.45 4.89 4.89
M 3.94 4.86 1.83 2.99 -.63
SD 8.93 7.65 7.10 6.34 4.32
M .08 1.36 .02 1.29 .36
SD 6.21 4.68 5.37 4.67 5.36
    Skin conductance    
M -.08 -.03 -.34 -.36 -.38
SD 1.04 .74 .50 .74 .47
M -.14 -.22 -.44 -.51 -.61
SD 1.16 .76 .83 .86 .63
M -.30 -.37 -.51 -.61 -.67
SD 1.18 .76 .91 1.00 .87

Imagery content. As may be seen in Table 4, subjects showed the least physiological response during neutral imagery and the greatest response to snake and speech scenes. In general,, phobic scenes occasioned more responding than other arousing material. The heart rate response was highly significant: content, Д3,96) = 5.88, p < .005. The New- man-Keuls test showed that the snake and speech scenes evoked a greater response than the danger scenes but did not differ significantly in heart rate response from the exercise scenes. A significant skin-conductance level difference was also found: content, F(3, 96) = 3.46, p < .02; the Newman-Keuls test showed skin-conductance during snake and speech scenes to be elevated relative to skin- conductance observed both with danger or exercise scenes. The respiration I/E ratio also tended to increase more during snake- and speech-fear scenes, but the F fell short of the 5% confidence level: content, F(3,96) = 2.38, p < .08,

An interaction between imagery content and fear group was found only for heart rate, Д3, 96) = 3.14, p < .03. As predicted, focal- phobic subjects were consistent heart rate responders in arousing imagery equaling the socially anxious subjects on the speech and exercise scenes and clearly exceeding them in heart-rate changes to the snake and danger scenes. When snake and speech scenes were analyzed alone, the F value was even larger: Content X Group, Д1, 32) = 6.75, p < .02.

Imagery training. For most physiological systems, the effect of response training was (a) to increase responding slightly during the reading of the image script, (b) to prompt a further sharp rise during the following image period, and (c) to return toward base values at recovery. This inverted-V curve was either absent or present only in a much attenuated form for stimulus-trained subjects (similar to the curves found in Experiment 1 for untrained subjects). These patterns are illustrated for heart rate in Figure 8. They are consistent in general form with results reported by Lang et al. (1980).

Figure 8. Heart rate change scores over imagery periods for speech and snake scenes. (Responses of snake- phobic and social-anxiety groups are shown separately after stimulus and response training [Experiment 2] and with no training [Experiment 1]. B/M = beats per minute; RECOV = recovery.)

The Training X Period interaction for heart rate was highly significant and attributable to the large image period acceleration of response-trained subjects, both for all arousal contents, F(2, 64) = 6.30, p < .005, and for the snake and speech scenes alone, F(2,32) = 5.77, p < .005. The heart rate responses during the image period for each scene content are presented in Figure 9. The greater heart rate reaction of the response-trained groups is apparent: training, Д1, 32) = 4.47, p < .05; Training X Period, Д2, 64) = 6.53, p < .005. This training effect was negligible for the neutral scenes but consistently large for arousing scenes. Furthermore, among response-trained subjects, the two fear groups showed a significant, differential response to the four arousing contents: Training X Fear Group X Content X Period, F{6,192) = 2.30, p < .04. As with the untrained sample of Experiment 1 (Figure 9), response-trained, socially anxious subjects reacted somewhat more to the danger scenes than to the exercise scenes, whereas focal-phobic subjects showed a reverse pattern. However, response training appeared to produce the sharpest difference between fear groups for the snake and speech scenes. Although the phobic subjects’ mean heart rate response is comparable to that of socially anxious subjects for the speech scenes, it greatly exceeds the latter groups’ reaction to snake scenes and the phobic subjects’ own heart rate change to speech scenes.

A separate analysis of fear scenes yielded a substantial F value: (Training X Fear Group X Content X Period, F(2, 64) = 2.94, p < .06. However, it cannot be concluded that response training alone accounts for these group reactions to the fear material. A comparison of results obtained from both experiments (see Figures 8 and 9) suggests that the above differences between clinical fears are present as trends in the untrained and stimulus-trained subjects and are only accentuated by response training. Analysis of the fear-scene data from both experiments combined yielded a Content X Phobia interaction, F{ 1, 53), = 3.08, p < .09. To test the predictions prompted by the exposure findings, t tests were performed on the grand mean differences in the two fear groups’ imagery heart rate response, separately for speech and snake content. As expected, for the speech scene the statistic did not approach significance. However, as predicted for the snake scene, the snake-phobic subjects’ heart rate increase during imagery was significantly greater than that of socially anxious subjects, f(63) = 2.02, p < .05. Thus, the overall imagery results parallel heart rate differences noted in Experiment 1 between fear groups during snake exposure and during speech performance. .

Figure 9. Image-period average heart rate and skin-conductance change scores for the two fear groups, following stimulus or response training (Experiment 2) and with no training (data from Experiment I). (To present relationships between visceral responses clearly, 1.0 /imho was added to all skin-conductance change scores. B/M = beats per minute; NTL = neutral; DNG = danger, EXC = exercise; SPH = speech; SNK = snake.)

Experiment 2 respiration rate showed a similar pattern of mean values, but the differences were statistically less reliable than

for the heart rate response. Thus, response- trained, socially anxious subjects increased respiration amplitude only to the speech scene, whereas focal-phobic subjects reacted most to the snake scenes and also gave substantial responses in danger and speech imagery: Training X Fear Group X Content X Period–arousal scenes, F(6, 192) = 1.90, p < .09, and snake and speech scenes, F(2, 64) = 2.88, p < .08. A similar trend was observed for respiration rate: Training X Fear Group X Content X Period, F( 6,192) = 2.12, p < .06.

As we found in Experiment 1, skin conductance tended to habituate over imagery periods, and to the extent that there was an effect of emotional scene content, it was an inhibition of that decrease to relevant fear scenes. The major contribution to this trend appears to come from response-trained subjects, whose skin-conductance change to fear- group-relevant scenes were the only clearly positive increases observed. Nevertheless, the training factor did not achieve statistical significance. Analysis of the overall response of fear subjects to fear scenes did yield a Fear Group X Content interaction, F(l, 32) = 3.71, p < .065, a borderline value. Furthermore, as Figure 9 suggests, this trend was present in both experiments. In fact, with all fear-scene data from both studies combined, there was a Content X Phobia interaction, F(l, 53) = 4.06, p < .05. Again, the t test of differences between fear groups on the speech scene did not approach significance, whereas analysis of the snake scene yielded 1(63) = 2.18, p < .05.

The hypothesis, of respiration and muscle- tension differences for exercise and fear scenes received substantial support. Muscle tension increased among response-trained subjects during all arousal scenes: training, F(l, 32) = 5.80, p < .025; and Training X Period, F(2, 64) = 4.36, p < .02. However, as predicted, this response tended. to be greatest overall for the exercise scene (next strongest for danger scenes) and least for the snake, speech, and neutral scenes: Training x Content x Period, F(8, 256) = 1.84, p < .08. Respiration amplitude during arousal scenes also increased more over periods for response-trained subjects, Д2,64) = 4.27, p < .02. This effect was substantial for danger scenes, as well as speech and snake scenes, but, as previously reported (Lang et al., 1980), not for exercise imagery: Training X Period X Content, Щ, 192) = 2.18, p < .05.


Imagery Training

The present results replicate those obtained by Lang et al. (1980) concerning the effects of instructions and training on physiological response in emotional imagery. As we noted previously, efforts to increase the number of stimulus propositions in the image result in a physiology essentially the same as that of untrained subjects; however, response training generates significantly greater efferent activation in imagery. These findings also confirm the primary hypotheses of this study in that the pattern of physiological events in the response-trained group was specific both to the thematic content of the image scripts and to the fear diagnosis of the subjects.

In the present study, response-trained phobic subjects reacted with greater physiological arousal to their own pertinent fear scenes than to any other scenes. Furthermore, response-trained subjects did not show significant activation to relaxing scenes (which contained none or only a few physiological response propositions) but responded strongly to the emotional and action scenes. In addition, for these response-trained subjects, the organization of visceral and somatic reactivity tended to vary with the theme of the image script; for example, the relevant fear scenes occasioned a coupled increase in respiratory and cardiac rate; imagined exercise generated a similar increase in heart rate, but respiration changes were relatively small; and muscle response during exercise imagery was actually greater than it was for the fear scenes.

It is important to note this specificity ‘of response in trained subjects. They were not simply increasing tension indiscriminantly to any imagery command. Furthermore, as we noted before (Lang et al., 1980), this response specificity is not wholly explained by different response propositions in the imagery scripts. Subjects varied their response according to the contextual theme of the image, paralleling patterns we would expect to see in the relevant, real situations. This suggests that the imagery response was not controlled wholly by the script but that its structure was, in significant part, determined by thematically consistent propositional elements that were brought to the image from the subject’s own long-term memory storage. For phobic subjects, it can be argued that this methodology helped access the subjects’ basic fear prototype. Indeed, the imagery differences confirmed by the manipulations of Experiment 2 were present as trends in the mean values recorded in Experiment 1. Furthermore, heart rate differences between groups in response to phobic content were significant in the analysis of data combining subjects from both experiments. Large visceral responses could not be obtained from nonphobic subjects to phobic content. The physiological response of socially anxious subjects to snake-phobic scenes was little enhanced by script structure or response training; on the other hand, snake-fearful subjects reacted strongly, showing, for example, an average imagery heart rate increase of over 8 B/M in snake scenes.

From a three-systems perspective (Lang, 1978), it is also pertinent that response training occasioned greater synchrony between ratings of arousal and visceral activation than was found for other subjects. That is, all groups reported arousal during fear and action scenes, but this was accompanied by modest and nonsignificant visceral responses for the stimulus and untrained groups. It is our assumption that under most conditions of instructed imagery, a substantial number of subjects process only the semantic network of the script with its natural language propositions. Subjects’ reports of arousal are appropriate to the emotional meaning of this information but do not reflect processing of the deep structure with its somatic and visceral programs (see Lang, 1983, for an extended discussion of these concepts). Furthermore, subjects often do not discriminate between making a report of the arousal meaning of the situation (i.e., what it would feel like if it really happened) and what was actually experienced as an active image. Thus, such ratings can be very difficult to interpret, discordance is high, and there is wide variability between subjects. The response-training method, which encouraged subjects to attend to somatic and visceral information, probably had two effects: It increased both the probability that efferent programs would be accessed and the subject’s discrimination of images that actually involved physiological processing. Thus, with response training, physiological arousal and verbal report were similar in that, for the most part, they varied significantly and in parallel over scene contents. Specifically, for different imagined fear contents, the relationship between arousal reports and heart- rate response was basically the same as that found in exposure. Again, these data are concordant with the view that imagery instructions and exposure can access the same phobia prototype; that is, different input media access a common data structure from memory.

Exposure and Imagery as Psychophysiological Tasks

The physiological reaction of the two fear groups across imagery contents is very similar to the pattern obtained in the first experiment from untrained subjects. Furthermore, with the addition of response training, the major group differences are now all at, or near, acceptable levels of statistical significance. Thus, as predicted, phobic subjects tended to be greater heart-rate responders than the socially anxious subjects. They also showed a large response to both group-relevant fear scenes, equaling the reaction of socially anxious subjects to their own speech scenes. On the other hand, socially anxious subjects failed to show any substantial reaction in snake imagery and produced a strong visceral response only to the speech scene. This differential response of the two fear groups, for fear-group-relevant contents, parallels that found with exposure to actual fear situations.

These results support the view that the psychophysiology of fear is consistent within fear-diagnostic group and across different contexts of fear evocation. That is to say, these findings permit us to retain the hypothesis that imagery and exposure physiologies are similar (provided that they are measured under the appropriate conditions). However, these experiments point up some

important methodological problems in this area of study and show why reliability and’ specificity of the fear response have formerly been so difficult to demonstrate.

The physiology of fear is always determined, at least in part, by the task physiology of the setting in which it is evoked. We have already described how the cardiovascular demands of public speaking mask the heart rate response in fear. We now see that this exposure-task physiology appears to be coded in memory and reactivated in imagery. However, a further complication is added: Imagery processing is itself a cognitive task that makes its own special demands on the physiological system. This phenomenon is illustrated by the differing fidelity with which skin-conductance and heart rate change in imagery reflect the strength of these same responses in exposure..

Blood, Sweat, Toil, and Fear

As in our previous research (Lang et al., 1980), we found that the activating effects of emotional script content and imagery training were somewhat greater for cardiovascular than for skin-conductance responses. Using a different imagery paradigm, Johnson and his colleagues (Jones & Johnson, 1978; Leber & Johnson, 1976; May & Johnson, 1973) observed similar differences in the relative amplitude of these two systems. It is possible that this effect is attributable to the specific physiological requirements of the imagery task. The generation of an image is a kind of mental work. It is well-known that mental effort occasions heart rate increase, whereas attention to the external environment promotes cardiac deceleration (e.g., see Lacey, 1967). Skin conductance enjoys some reputation as an index of internal feelings; however, the strongest and most consistent electrodermal responses are occasioned by external stimulation. Thus, heart rate increases (as an efferent reflection of response information in the image) may benefit from a gratuitous concordance with physiological requirements of the image-processing task. On . the other hand, imagery skin-conductance responses are attenuated because» internal, cognitive processing requires subjects to disattend to external events (a primary source of electrodermal stimulation).

There is other evidence for the above view. In nearly all of our experiments, somatic- and visceral-response amplitudes are largest after the reading of the complete imagery script and during the image period (when no external stimuli are administered). The exception to this rule is skin conductance, which shows a larger response during the external, auditory stimulation of script presentation (i.ei, the read period). Results from three very different experiments conducted in our laboratory provide further conveigent support for this interpretation of the skin-conductance/heart rate dissociation. In a study of classical conditioning with a nociceptive unconditioned stimulus (UCS; Cook, Hodes, Ohman, & Lang, Note 4), subjects were instructed to imagine vividly an action context involving the conditioned stimulus (CS) during the CS-UCS interval. Heart rate was little affected, and the extra imagery task did not interfere with the actual formation of skin- conductance conditioned response (CR). However, we found this procedure to greatly attenuate the absolute magnitude of all electrodermal responses.

In a second experiment (McLean, 1981), subjects were instructed to imagine an emotional scene and were presented with information about the fear context in the form of an acted playlet–rather than being read a prose text. In this case, with stimulus information physically presented to the distance receptors, affect-related skin-conductance responses were large, approaching the amplitude found in the exposure situation. May (1977) found a similar differential effect for heart rate and skin conductance in that the latter response to a “phobic slide was greater than to internally generating or hearing a statement about the phobic material,” whereas the opposite result was obtained for heart rate (p. 855).

In yet another study (Miller et al. 1981), we employed the imagery paradigm used here with a sample of preselected «good imagers” and “poor imagers” (as determined by the QMI). For the first time (among response- trained “good imagers”), we did find strong, content-specific electrodermal changes in response to an imagery script. This time, as with the other physiological responses we recorded, skin-conductance change was greater during the imagery period than during the read period. Taken together, these data suggest (a) that the imagery task may indeed attenuate skin-conductance amplitude relative to other visceral and somatic responses we have studied and (b) that this occurs because processing text involves a tuning out of the external environment to which electrodermal events are so responsive. The available studies also suggest that theme-relevant electrodermal information is, nevertheless, explicitly coded in the image and that efferent evidence of this coding can be uncovered if sample sizes are sufficiently large, if visual media are used to prompt the image, or if the research employs subjects with a special talent for internal representation of the external world.

Behavior, Cognition, and the Organization of Affect

The research reported here indicates that the pattern and amplitude of physiological responses in emotion are to a considerable extent determined by formal characteristics of the stimulus context in which they are measured and by the subjects’ task as it is defined by that context. Thus, instructions to give a speech in the presence of an audience prompt cardiovascular reactivity that is wholly in the service of cognitive and motor behavior that permits compliance with task demands. However, in addition to this instructionally and contextually defined program, frightened subjects may have their own hidden agenda based on previous conditioning and/or specific expectations of the test situation. Indeed, the physiology of the socially anxious subjects studied here suggests a hyperalert state, the source of which one can only speculate on (e.g., expectancy of criticism and derision from the audience).

The assessment of focal phobia is similarly influenced by the measurement context. The task of observing a snake requires only a physiology supportive of quiet attention. Thus, the appearance of cardiovascular arousal in subjects who profess fear cannot be understood as a task demand. However, such cardiac acceleration is consistent with both the anticipation of motor behavior (Chase, Graham, & Graham, 1968) and the performance of an overt action. The longer the phobic subject remains in the increasingly threatening exposure situation, the greater are both the increase in cardiac rate and the probability that he or she will fly from the field. This fact has led us to speculate that the heart-rate change observed in this context is a visceral sign of a specific response disposition–a perceptual-motor set to avoid.

We began this article noting the frequent failures of concordance that have troubled the study of fear and other hypothesized states of emotion. At least some of these difficulties may arise from our efforts to organize the psychophysiological data around a traditional, introspectively derived concept of fear. From this viewpoint, fear is an internal state that precedes and motivates affective behavior. Reports of emotional experience are a key data set. They describe the determining internal state and thus are expected to be concordant with other, more peripheral responses in emotion. However, we have found here that verbal report of affect can be quite independent of other behavior. For example, subjective arousal is not closely related to avoidance. Furthermore, individuals can report arousal in the absence of any increase in visceral-response amplitude as in the imagery trials in Experiment 1 (or following stimulus training in Experiment 2). Alternatively, arousal reports are sometimes concordant with one physiological measure and not with another, and the specific measure varies with subject population and context.

We have suggested here that the psychophysiology of emotion can be better understood if the data are organized around the action orientation of subjects rather than around inferred internal experience. From this perspective, affective states are understood primarily as dispositions to approach or avoid, hyperattentive or rejecting postures vis a vis the environment, or a disorganization of behavior prompted by the coincident generation of incompatible response programs. It is expected that the physiology of emotion will be concordant with these response dispositions. However, the experimenter must still infer the correct behavioral program. The task context in which emotions are measured can prompt other than affectively toned responses, and the psychophysiology of a measurement context can represent a blending of different action sets.

Finally, we suggest (see also Lang, 1983) that affective response dispositions are coded in the brain as propositional networks. These information structures contain stimulus, meaning, and response information that in the case of phobia constitute a relatively stable fear prototype. Processing of these fear networks involves the innervation of motor subprograms that define the action set. Prototype processing can be accomplished through exposure to external fear stimuli but also through other media prompts or instructions. Under appropriate conditions, subjects can imagine the fear state, and this imagery includes the initial mobilization of the efferent action pattern. We have suggested elsewhere (Lang et al., 1970) that the generation of this subthreshold fear behavior is required if imagery is to be a vehicle of broader behavior change. One important direction of future research will be to explore this mechanism in the context of the therapeutic modification of phobia and anxiety.


The Effect of Imagining an Event on Expectations for the Event: An Interpretation in Terms of the Availability Heuristic

John S. Carroll

Carnegie-Mellon University Received November 24, 1976

Previous studies have indicated that explaining a hypothetical event makes the event seem more likely through the creation of causal connections. However, such effects could arise through the use of the availability heuristic; that is, subjective likelihood is increased by an event becoming easier to imagine. Two experiments were designed to demonstrate this principle. In Experiment 1, subjects asked to imagine Jimmy Carter winning the presidential election (prior to the election) predicted that he was more likely to win than subjects asked to imagine Gerald Ford winning. In Experiment 2, subjects asked to imagine a good college football season for the previous championship team were more likely to predict a major bowl bid than subjects asked to imagine a bad season, although the effect did not appear in predictions of the season record. In both studies, subjects who were also asked to explain the imaginary event were no different from subjects who only imagined. Several other attributional distortions are interpreted in terms of the availability heuristic.

Attribution theorists are beginning to uncover a growing number of distortions in the supposedly logical attribution process (Fischhoff, 1976; Ross, 1977). These have often been labeled as “biases” or “errors” in the sense that judgments are consistently nonoptimal. This study analyzes one of these distortions in terms of a parsimonious principle of memory and judgment. The mechanism behind several other distortions seems to be based on the same principle.

Ross, Lepper, and Hubbard (1975) found that subjects who had been given false feedback about their performance on a novel task, and who were later told that the feedback had been random and unrelated to their performance, nevertheless changed their expectations about future performance and about estimates of their ability. Thus, discredited manipulations produced enduring effects that Ross et al. (1975) called “perseverance.»’ In follow-up studies, Ross, Lepper, Strack, and Steinmetz (Note 1, reported in Ross, 1977) asked subjects to explain an event in a patient’s life using a clinical case history of the patient. Subjects came to believe that the explained event was more likely than did controls, even when they had known from the beginning that the event they explained was entirely hypothetical.

Ross and his colleagues have explained the reluctance to abandon an impression formed around a hypothetical or subsequently discredited event by proposing that subjects construct ‘ ‘antecedent-consequence linkages’ ’ which connect the event to a set of causal conditions capable of producing the event. These causal conditions are still salient even if the specific event is labeled hypothetical, and their presence creates expectations for the same result.

However, an alternative and more parsimonious hypothesis exists for these results. Imagining an event could, in itself, be sufficient to make the event seem more likely.[1] Tversky and Kahneman (1973) have shown that people judge the frequency or likelihood of events by the ease with which they can remember or construct relevant instances, what Tversky and Kahneman called the “availability heuristic.” Availability is useful because frequent events are typically easier to recall and to imagine than infrequent events. However, reliance on availability can lead to deviations from an optimal statistical judgment because availability also depends on factors such as recency and saliency that are unrelated to true frequency.

Under the availability hypothesis, the perseverance result would be due to the greater ease of generating relevant instances when an instance has been already presented, regardless of whether the event was imaginary or real, inferred or observed, true or later discredited. The lack of discrimination of imaginary from real events is supported not only by the perseverance results but also by Barclay (1973), Dosher and Russo (1976), and Loftus (1975). For example, Loftus found that eyewitnesses to an accident incorrectly remembered seeing a stop sign 1 week after the event г/they had been asked about a stop sign immediately after the event (and correctly replied “no”). Apparently, people can confuse an event representation created by a question with an event representation created by observation.

Two experiments were designed to test the availability formulation in real-world settings with familiar beliefs. In each experiment, subjects were randomly assigned to imagine one of two hypothetical outcomes for an actual future event. The experiments differed in the event utilized, either the 1976 presidential election or the 1977 University of Pittsburgh football season. To examine whether or not antecedent-consequence linkages were important, all the subjects were instructed to imagine, but half were also asked to give explanations. All subjects then gave their own predictions for the actual event, embedded among several plausible filler items.


Subjects in this experiment were randomly assigned to imagine either Jimmy Carter or Gerald Ford winning the 1976 presidential election 1 day before the election. Half of the subjects in each condition also gave explanations of why the election would have had the imagined outcome. A control group made predictions without being given instructions to imagine a specified outcome.

It was possible to predict that subjects in this study would differ in their expectations about the election, depending upon whom they supported. Ross, Greene, and House (1977) found that subjects incorrectly believed their own preferences to be widely shared, an effect they labeled “false consensus.” If subjects in this study perceived a false consensus, they should have believed that their preferred candidate would be more likely to win. In order to control for this source of variability in the study and to examine the possible interactions of imagining or explaining events consonant or dissonant with prior attitudes, candidate preference was also measured.


Subjects were 97 undergraduates in five different introductory-level psychology classes at two different colleges. Questionnaires were filled out during the first or last 10 min of class. Each subject was randomly assigned to one of five conditions consisting of a basic 2×2 factorial design and a control group. The control group answered seven questions about the 1976 presidential election, of which the third question was the key variable of their “own estimate of what the election results will be” on a graphic scale anchored at the left by “Sure Carter will win,” at the center by “Complete Toss-up,” and at the right by “Sure Ford will win.” The fifth question was “Do you favor either Carter or Ford,” with those two alternatives and “Neither” as possible answers. The control group then did another task in order to keep them working while others worked.

The four experimental groups all participated in “an exercise in imagination.” They were asked to picture themselves on Tuesday and Wednesday during the election. One of two hypothetical scenarios was presented to each subject: (a) Ford wins the election as Carter fails to hold some key states and Ford wins much of the Midwest and West. He wins 316 electoral votes to Carter’s 222, and a listing of states and electoral votes under columns for Carter and Ford shows Ford with 32 states and Carter with 18 states and the District of Columbia, (b) Carter wins the election as his strength in the South and East builds an insurmountable lead that Ford’s near sweep of the West cannot overtake. He wins 342 electoral votes to Ford’s 196, with 28 states and the District of Columbia to 22 states for Ford. The scenarios were constructed to be plausible by using the most up-to-date polls. Subjects were asked to imagine these events as if they were “present on Tuesday and Wednesday, watching some of the news on television, perhaps staying up late, or checking the early morning news by television or newspaper.” This included asking them to imagine the speeches conceding defeat and accepting victory.

Half of the subjects in each condition were also asked to «think about why the results occurred this way,” to find the factors behind the results and the sources of support for Carter and Ford. These subjects were asked to list their reasons for why the events occurred before responding to the key dependent measure of their expectations for the election.

As did the control subjects, the experimental subjects answered several questions. Subjects giving explanations answered nine questions; the expectation question was fifth, immediately following a question asking for their reasons, and their own candidate preference question was eighth. Subjects who only imagined answered eight questions, without giving any reasons; the expectation question was fourth and their own candidate preference question was seventh. The overall impact of the set of questions made it virtually impossible for subjects to detect the purpose of the study. No suspicions were voiced by any of the subjects during debriefing.


The “false consensus” effect would predict by analogy that Ford supporters and Carter supporters would each expect their own candidate to have a better chance of winning than did the supporters of the opponent. As shown in Table 1, control subjects who supported Carter believed Carter would win (M = 33.10 on a scale from 0 to 100 with “sure Carter will win” at 0 and “sure Ford will win” at 100), while control subjects who supported Ford believed Ford would win (M = 68.6). Control subjects stating no preference believed Ford would win (M = 85.3), but these results were questionable because only three control subjects expressed no preference. The relationship between candidate preference and election expectations was strong and significant [F(2,17) = 15.25, p < .001]. On this basis, the analysis of the experimental groups was expanded to include candidate preference as a third factor.

The means for all four experimental groups and the control group, broken down by candidate preference, are given in Table 1, along with the number of subjects in each condition. A 2 (imaginary victor) x 2 (request for explanations) x 3 (candidate preference) analysis of variance revealed a significant main effect of the imaginary victor [F(l,65) = 10.50, p < .005]. Averaging across the experimental groups, subjects who imagined Carter winning believed Carter would win (M = 45.9), while subjects who imagined Ford winning believed Ford would win (M = 60.5). There was also a significant main effect of candidate preference [F(2,65) = 16.52, p < .001]. This was due almost entirely to the linear trend with Carter supporters (M = 35.8) and Ford supporters (M = 67.0), each expecting their preference to be realized [F(l,65) = 31.80,p < .001], and with neutrals (M = 56.8) in between [F(1,65) = 1.24, ns]. No other effects were statistically significant.



Mean Election Expectations of Experimental and Control Conditions by Candidate Preference

Ford Carter

Candidate preference Imagine Explain Imagine Explain Control
Carter 47.4 (7) 43.0 (8) 25.5 (6) 27.4 (5) 33.1 (10)
Neither 75.4 (5) 59.6 (5) 61.0 (5) 31.0 (5) 85.3 (3)
Ford 66.9 (7) 70.7 (7) 65.0 (7) 65.4 (10) 68.6 (7)
Average, unweighted 63.2 (19) 57.8 (20) 50.5 (18) 41.3 (20) 62.3 (20)

Note. Responses are on a scale from 0 = sure Carter will win to 100 = sure Ford will win, with 50 = toss-up. Numbers in parentheses are the number of subjects in each cell.

Since the candidate preference variable was measured after the manipulation, it is important to show that it was not affected by the manipulation. This was tested by comparing the candidate preferences of subjects who imagined Ford winning to those who imagined Carter winning. The results were thoroughly insignificant (y2(2) = .88, ns]. Moreover, the experimental subjects did not differ in the distribution of candidate preferences from the control subjects [x2(2) = 2.05, ns]. Finally, a simple 2×2 analysis of variance ignoring candidate preference confirmed the previous results. The imaginary victor variable was nearly significant [F(l,73) = 3.60, p < .07], with no other effects of any magnitude.


This experiment was designed as a replication of Experiment 1 utilizing a different imagined event and a different population of subjects. Subjects were randomly assigned to imagine either a good or bad outcome for the 1977 football season of the University of Pittsburgh (Pitt) team 6 months before the start of the season. In 1976, Pitt was the national champion. Since Pitt undergraduates were used as subjects in this experiment, interest and importance would be high. With the graduation of top players and the loss of the championship coach, there might be some uncertainty as to the expected outcome, just as there had been in the 1976 presidential race. False consensus effects were not examined, since misperception of student support would have no necessary effect on expectations about the team’s success.


Subjects were 93 undergraduates in one introductory-level psychology class at the University of Pittsburgh, a college different from those used in Experiment 1. Questionnaires were filled out during the last 10 min of class. As in Experiment 1, each subject was randomly assigned to one of five conditions consisting of a 2 x 2 factorial design and a control group. The control group was provided with a schedule listing dates, locations, and opponents and was asked a variety of questions about the 1977 football season, of which one was their “own prediction of how many regular season games (out of eleven) the Pitt team will win, lose, or tie in this new season?” Space was left next to the “Win,” “Lose,” and “Tie” labels for their responses. Following this question was a second question, “Will Pitt be invited to a major bowl (Cotton, Sugar, Orange, Rose) or minor bowl in post-season?” Subjects could circle “Major,” “Minor,” or “None.”

The four experimental groups all participated in «an exercise in imagination.” They were asked to picture themselves in September, October, and November as Pitt plays its games. One of two hypothetical scenarios was presented to each subject: (a) Pitt wins all eleven of its season games, with scores given for each game, including close victories over Notre Dame and Penn State. They are ranked second in the nation and defeat Alabama in the Sugar Bowl, (b) Pitt wins seven, ties one, and loses three of its games. Scores for each game were given, but national rankings and postseason play were not mentioned. These two scenarios were constructed to be plausible using predictions made by a pilot group. Subjects were asked to imagine these events as if they were “present, watching some games, going to some, listening to some, reading about them, talking about them, hearing the sports report on TV, or whatever you would probably do ”

Half of the subjects in each condition were also asked to “think about why Pitt’s season turned out this way,” to find the factors that “would explain or allow someone to understand these events.” These subjects were then asked to list their reasons before answering prediction questions, as in Experiment 1. Prediction questions were sixth and seventh of nine questions for subjects who only imagined or seventh and eighth of ten questions for subjects who also gave their reasons.


The dependent measure of the predicted season record in wins, losses, and ties was converted to one score by subtracting losses from wins. A 2 (imaginary season) x 2 (request for explanations) analysis of variance showed no effect for imagining a good vs. a bad season (F < 1), for requesting explanations (F < 1), or for their interaction [F(l,69) = 1.29, ns]. At test between the control subjects and all experimental subjects showed that the trend for the control group to predict a better season was nonsignificant [r(91) = 1.61, p < .15].

The second dependent measure, predictions for a postseason bowl bid, was treated as a binary variable by combining the two subjects who predicted no bowl bid (both imagined a bad season) with subjects predicting a minor bowl bid. These percentages were transformed by the use of arc sines as proposed by Langer and Abelson (1972). Statistical tests revealed a significant effect for imagining a good vs. a bad season (Z = 2.09, p < .05) and no effect for requests for explanations (Z = 1.33, ns) or their interaction (Z = .64, ns). Of those imagining a good season and a major bowl victory, 22 of 35 subjects (62.9%) predicted a major bowl bid in the actual 1977 season as compared to 15 of 38 subjects (39.5%) imagining a bad season with no mention of a bowl bid. Subjects imagining a good season did not differ from control subjects (72.2%; Z = .69, ns), but significantly fewer subjects imagining a bad season (and not instructed to imagine a bowl bid) predicted a major bowl bid than did controls (Z = 2.34, p < .02).

Given the clear results of the bowl bid prediction variable supporting the findings in Experiment 1, it is disappointing that the prediction of the season record did not produce similar results. This can be explained by recognizing that the manipulation of the imagined season record was actually weak, 11-0 vs. 7-3-1, and scores for several games were close enough so that the scenarios could be taken as instructions to imagine some ambiguous and close contests rather than a “good” or a “bad” season. Hence, all experimental subjects seemed somewhat more conservative in their responses than did control subjects. In contrast, the manipulation of the bowl bid presented along with the manipulation of the season record was stronger and less ambiguous, a major bowl bid and victory vs. no mention of a bowl bid. Further, the season record prediction was not sensitive to how well or poorly the team did within each game, only to gross outcomes. For a team of such high quality (the modal prediction was a 9-2 record), a major bowl bid may be a more sensitive measure of season outcome.


The results of these studies demonstrate that instructions to imagine an event are sufficient to increase expectations for that event. Subjects who imagined Jimmy Carter the victor in the presidential election predicted he was more likely to win than subjects who imagined Gerald Ford the winner. Similarly, a higher proportion of subjects who imagined a good Pitt football season capped by a major bowl bid predicted a major bowl bid than subjects who imagined a bad Pitt football season. There was no additional effect on expectations by instructing subjects to explain the imagined events and state their explanations, in either study. Although one cannot categorically state that subjects who imagined the events did not also explain them (a general problem of reasoning from null results), the results are consistent with the proposed principle that manipulating the availability of relevant instances by imagining an event is sufficient to change expectations about the event. This is further supported by the fact that imagination did not interact with prior beliefs about the presidential election.

The effect of the imagination instructions may depend upon the provision of a relatively detailed, vivid, and concrete scenario. Abelson (1976) has proposed that such “scripts” are the basis for a wide variety of attitudes and behaviors. Abelson (Note 2) has also asked subjects in several studies to imagine they lost their wallet, spilled coffee on themselves, and so on, but he failed to get any consistent support for the effect of imagining an event on expectations for the occurrence of that event. It is thus possible that the detailed scenarios in the present studies provided a sequence of connected events which made the final outcome or summary event seem more likely; however, the scenarios could also have served simply to enrich and amplify what subjects would have otherwise imagined.

The results add additional weight to the principle that expectations are generated by evaluating the ease of recalling or constructing relevant events or the number of such events. The objective fact that some events are imaginary, hypothetical, inferred rather than observed, or even factually discredited is poorly coded or not properly used. Thus, the act of posing a problem or asking a question could itself change the beliefs of subjects. The content of different questions, although equivalent in logic, may affect judgments by facilitating the construction or recall of different events (cf. Kanouse, 1972).

The availability heuristic is similarly applicable to other biases or errors in the attribution process. Nisbett and Borgida (1975) found that the presentation of a few concrete case histories of participants in an experiment and their behaviors strongly altered predictions about the expected behaviors of other participants. This was particularly striking when compared to the result that summary statistics about the behaviors of all participants in the experiment did not alter expectations about individual behaviors. Concrete case histories provide recent, salient examples of events that would be highly available to subjects when they make predictions. Abstract base rates apparently do not generate any examples or make information more available, except when well-practiced ways of dealing with abstract base rates are available, as in achievement situations (Nisbett, Borgida, Crandall, & Reed, 1976).

The false consensus effect (Ross et al., 1975) can also be understood as the result of making an event more available by imagining it as the hypothetical response to a conflict situation or by actually performing the response. As in Experiment 1, belief in one’s own candidates’ better chances may be supported not only by a selective sample of opinions, but also by imagining the desired event more often and thus making it more available. Such self-delusions are probably very common, where we increasingly believe what we frequently imagine, to our growing pleasure or displeasure.

Finally, this research has strong implications for how expectations are generated. Although expectations are considered central to theories of motivation (Atkinson, 1957), decision making (Edwards, 1954), attitudes (Fishbein & Ajzen, 1975), and causal attributions (Ajzen & Fishbein, 1975; Jones & McGillis, 1976), research has typically focused on how expectations are combined and utilized in tasks rather than on how expectations are generated. This research indicates that expectations are not simply retrieved from memory but are actively generated when needed. The process of generating expectations seems to involve the construction or recall of relevant events in a manner consistent with the availability heuristic.


A Dual Representation Theory of Posttraumatic Stress Disorder

Chris R. Brewin Tim Dalgleish

University of London Medical Research Council Applied Psychology Unit

Stephen Joseph

University of Essex

A cognitive theor у of posttraumatic stress disorder (PTSD) is proposed that assumes traumas experienced after early childhood give rise In 2 sorts of memory. I verbally accessible and I automatically accessible through appropriate situational cues. These different types of memory are used to explain thecomplex phenomenology of PTSD, including the experiences ol» reliving the traumatic event and of emotionally processing the trauma. The theory considers 3 possible outcomes of the emotional processing of trauma, successful completion, chronic processing, and premature inhibition of processing. We discuss the implications of the theory for research design, clinical practice, and resolving contradictions in die empirical data.


Posttraumatic stress disorder (PTSD) has recently become the focus for a great deal of empirical and theoretical work. The fact that it is one of the few psychiatric disorders to be associated with a relatively specific etiological agent in the form of a traumatic event, along with its distinctive clinical feature of repeated intrusive memories, has proven a magnet for researchers. Already a number of valuable psychological theories of the disorder have been proposed, many couched in terms of an information processing analysis (Chemtob, Roitblal, Hamada, Carlson, & Twentyman, 1988; Creamer, Burgess, & Pattison, 1992; Foa, Steketee, & Rothbaum, 1989; Foa, Zinbarg, & Roth- baum, 1992; Lit/ & Keane, 1989), In this article we attempt to establish explicit criteria for a theory, review existing formulations, and propose a new cognitive model of PTSD.

We believe an adequate theory of PTSD should address a number of issues. First, the theory should account for the clinical characteristics of the disorder, including the range of associated symptoms and the time course of the disorder. Second, it should indicate whether the symptoms of PTSD arc themselves automatically indicative of an abnormal process whenever they occur. If they are not, the theory should explain how normal and abnormal processing differ within a comprehensive mode!. Third, the theory should outline the conditions that are associ- ated wit fi the severity and outcome of PTSD and should explain Chris R. Brewm. Department of Psychology, Royal Holloway, University of London, Egham, England: Tim Dalgleish, Medical Research Council Applied Psychology Unit, Cambridge, England; Stephen Joseph, I H-parlment of Psychology, University of Essex, Essex, England.

We gracefully acknowledge discussions with Edna Foa and John Teas- dale and comments by Richard McNally, Elizabeth Monk, Bill Stiles, and James Thompson on an earlier version of this article.

( orrcspondence concerning this article should be addressed to Chris R. Brewin, Cognition Emotion and Trauma Group, Department of Psychology, Royal Holloway, University of London. Egham, Surrey TW20 OEX, England. Electronic mail may be sent via the Internet to c, b rew i n f® rhb nc .ac, u к.

these relations. Fourth, the theory should discriminate PTSD from other related disorders and account for any evidence of comorbidity. Fifth, it should explain the experimental data on information processing in PTSD. Finally, ideally, the theory should do a better job of accounting for the available data than existing theories and make novel predictions that can be tested empirically.

We will briefly summarize the state of knowledge in the first five areas described above. Then we will describe existing theories and how they address the available knowledge. Finally, we present our own theory, indicate in what respects we believe it goes beyond existing theories, and describe a number of novel predictions that it generates.

Existing Knowledge About PTSD

Cl in ical Char act east ics

The major clinical features of PTSD are described in the Diagnostic and Statistical Manual of Mental Disorders (DSM- IV; American Psychiatric Association. 1994) and include the reexperiencing symptoms, such as intrusive memories and nightmares; the protective reactions, such as emotional numbing, amnesia, and cognitive avoidance; and the arousal symptoms, such as the startle response and hypervigilance. In addition, PTSD is commonly accompanied by a wide range of negative emotions, such as sadness and anger, and by negative cognitions, such as guilt.

Perhaps the hallmark characteristic of PTSD is the alternation between reexperiencing and avoiding trauma-related memories. The memories that are particularly associated with PTSD appear rapidly and spontaneously, often intruding into consciousness with high frequency. In many cases the intrusive memories consist of images accompanied by high levels of physiological arousal and arc experienced as reenactments of the original trauma (“flashbacks”). Other memories, particularly those associated with traumas in the distant past, may be more fragmentary, consisting of isolated visual, auditory, olfactory, or tactile sensations. They may have many of the qualities of flashbacks, without necessarily corresponding to an easily recognizable event.

Flashbacks can be distinguished from memories of the trauma retrievable through a normal search of long-term mem ory. Whereas a great deal of deliberate and spontaneous cognitive activity typically involves individuals recalling aspects of trauma and the accompanying emotions they remember having experienced at the time, these deliberately retrievable memories are qualitatively different from the spontaneous reenactments, in which emotions often appear to be resynthesized in their original intensity (Janet, 1889; Terr, 1991; Van der Kolk & Fisler, 1995). Herman (1992b) has referred to these as frozen memories, a term that captures their repetitive, unchanging quality. It is also evident that, over time, trauma survivors are able to deliberately recall and describe some aspects of their experiences in a dispassionate way, without provoking full-scale flashbacks, whereas flashbacks may still be triggered by recalling certain details of the trauma or by unexpected reminders of the trauma.

Green (1994) concluded from an extensive review of the literature that rates of PTSD, given exposure to a traumatic event, average around 25 to 30% in the general population, although certain stressors, such as rape, are routinely associated with much higher rates. Prevalence rates tend to decrease with time elapsed, although up to one half of the individuals diagnosed as having PTSD may continue to have it for many years.

Longer term studies have shown PTSD lasting up to 40 years posttrauma in World War II combat veterans and POWs (e.g.. Davidson, Kudler, Saunders, & Smith, 1990; Hierholzer, Munson, Peabody, & Rosenberg, 1992) and Jewish survivors of the Holocaust (Kuch & Cox, 1992). However, although epidemiological data suggest that PTSD can be prevalent many years subsequent to an event, survivors often report being only intermittently troubled by their symptoms (Zeiss & Dickman. 1989). In addition to this varied course, initial onset may occasionally be delayed by many years (Blank, 1993; McFarlane, 1988).

Normal or Abnormal Process?

If the symptoms of PTSD do not occur except in the context of a specific diagnosable condition, then it is appropriate for a theory of PTSD to focus on explaining this distinctive pathology. If PTSD symptoms are found more generally, albeit perhaps at a lesser intensity, a more inclusive theory is required that distinguishes between normal and abnormal responses to extreme stress. To date, with few exceptions (e.g., Horowitz, 1986). theories of PTSD have not adopted this broader focus. This may be connected with the claims made in the Diagnostic and Statistical Manual of Mental Disorders, third edition, revised (DSM-HI-R; American Psychiatric Association. 1987) that the hallmark symptoms of PTSD, such as intrusive reexperiencing, are associated with events outside the range of human experience and are not found in other stress-related conditions, such as adjustment disorder. Theories of the stress response that do span normal and abnormal aspects have, however, been developed, particularly applied to bereavement and victimization, and we will review them briefly.

Among influential theories of the bereavement process are those of Bowlby (1980), Parkes( 1971), and Horowitz (1990). In Bowlby’s attachment theory, the typical response to loss of an attachment figure is a phase of protest and anger, which, if it does not succeed in restoring the loss, is followed by a phase of despair. During this period there is an oscillation between conscious preoccupation with the loss and defensive exclusion of related thoughts and reminders. Bowlby described phases of grief involving numbness, pining, disorganization-despair, and reorganization.

This theory was elaborated by Parkes( 1971,1986), who formulated loss in terms of psychosocial transitions. Psychosocial transitions are major life changes that involve disruption to basic assumptions about one’s nature, needs, goals, and sources of support. Bereavement, Parkes suggested, consists of the difficult and painful process of abandoning long-held assumptions and adopting new and more appropriate ones. Similarly, Horowitz (1990) formulated bereavement in terms of changes to person schemas, cognitive structures summarizing information about oneself, close others, and relationships. In addition to the repeated descriptions of anger, denial, and numbness, bereavement researchers have consistently found anxiety symptoms indicative of heightened arousal, such as muscle tension, panic, and insomnia (Maddison&Viola, 1968;Parkes, 1970; Worden. 1982). Preoccupation accompanied by intrusive images of a bereaved spouse are also extremely common (Horowitz. 1990; Rees, 1975).

General theories of response to problematic experiences or severe life events (eg., Janoff-Bulman, 1992; Klinger, 1975; Lazarus & I.aunier. 1978; Rachman, 1980; Shontz, 1965; Stiles et al., 1990) have many similarities to the theories of bereavement outlined above. Most of these theories explicitly recognize that following severe events there can frequently be observed intrusive cognitions or intrapsychic defences such as denial and avoidance that help to mitigate the intensity of distress. It has been argued that a similar syndrome may result from exposure to repeated minor stressors (Prolonged Duration Stress Disorder or PDSD; Scott & Stradling, 1994), and recent research has shown that intrusive memories are a relatively common everyday experience (Brewin, Christodoulides, & Hutchinson, 1996). This considerable degree of overlap indicates that PTSD can be partly understood in terms of general theories of adjust- mem. However, these theories do not describe the characteristic repetitive flashbacks involving startle responses and high degrees of associated physiological arousal and, thus, do not provide a full account of PTSD. For example, the intrusive memories associated with bereavement often involve positive recollections of the lost person.

Variables Associated With Severity and Outcome

Although there is evidence that bereavement is especially associated with more severe and chronic disturbance (e.g., Green, Grace, Lindy, Tichcner. & Lindy, 1983; Joseph, Yule, Williams, & Hodgkinson. 1994; Shore. Tatum, & Vollmer, 1986), bereavement does not fully account for the individual differences in reactions. Exposure variables, such as personal injury and life-threat, have also been found to influence the course of symptomatology in combat veterans (e.g., Fontana, Rosenheck, & Brett, 1992; Foy, Resnick, Sipprelle, & Carroll, 1987; Kuika et al., 1990; 0mer. Lynch, & Seed, 1993: Yehuda, Southwick, & Giller, 1992) as well as in civilian survivors of a variety of events (e.g,, Cluss, Boughton, Frank, Stewart, & West, 1983; Ellis, Atkeson, & Calhoun, 1981; Gleser, Green. & Winget, 1981; Green et al., 1983; Joseph, Yule, Williams, & Hodg- kinson. 1994; Maida. Gordon, Steinberg, & Gordon, 1989; Resnick, Kilpatrick, Best, & Kramer, 1992: Shore et al., 1986; Smith. Robins, Pryzbeck, Goldring, & Solomon, 1986). However, much of this work has relied on the subjective assessment of exposure variables, and other work has attempted to demonstrate an objective exposure-response relationship. For example, Bromet, Parkinson. Schulberg, Dunn, and Gondek (1982 ) showed that mothers living within 5 miles of the Three Mile Island plant had poorer psychological health than those outside this radius at 3 months after the leak. Pynoosand Nader (1988) also demonstrated that following a fatal sniper shooting in a California school, children trapped in the playground had the strongest posttrauniatic reactions: those who did not attend school that day showed the least reaction.

A number of studies have shown a history of psychological or behavioral problems to be predictive of PTSD symptoms in adults after a natural disaster (McFarlane, 1989): personal injury or violent attack (Breslau. Davis. Andreski, & Peterson, 1991; Helzer, Robins, A McF.voy. 19871; and, in women, following rape (Atkeson, Calhoun, Resiek. & Ellis. 1982; Burgess & Holmstrom, 1978; Frank & Anderson, 1987: Frank, Turner. Stewart, Jacob, & West, 1981; Ruch, Chandler, & Harter. 1980). Other researchers have not found prior clinical history to be associated with outcome (Kilpatrick, Veronen, & Best, 1985; Madakasira & O’Brien, 1987; Solkoff, Gray. & Keill. 1986; Speed, Engdahl, Schwartz, & Eberly, 1989). On the other hand, the important role of prior experience of traumatic events in general has been confirmed in numerous studies (Breslau et al., 1991; Burgess & Holmstrom, 1978; Kilpatrick et al., 1985; Roth. Wayland, & Woolsey. 1990; Ruch & Leon, 1983). Thus, prior trauma may account for the association between previous clinical history and the severity of PTSD.

Postdisaster life events have been associated with poorer outcome following disaster (Joseph, Yule, Williams, & Hodg- kinson, 1994; McFarlane, 1988) and rape (Ruch et al., 1980). Higher levels of social support have been found to be adaptive after rape (Burgess A Holmstrom, 1974; Kilpatrick et al., 1985), disaster (Bartone, Lisa no. Wright, & Ingraham, 1989; Cook & Bickman, 1990; Green, Grace, & Gleser, 1985; Joseph, Andrews, Williams, & Yule, 1992; Joseph, Yule, Williams, & Andrew’s, 1993; Madakasira & O’Brien, 1987), toxic exposure (Bromet cl al.. 1982; Fleming, Baum, Gisriel, & Gatchel, 1982). and combat (boy et al.. 1987; Foy, Sipprelle, Rueger, & Carroll, 1984; Frye & Stockton, 1982; Keane, Scott. Chavoya, Lamparski, & Fairbank, 1985).

Wor k with combat veterans has shown an association between better outcome and a more internal locus of control (Frye & Stockton, 1982; Solomon, Mikulincer, & Benbenishty, 1989) and a more internal and controllable attributional style for positive events (McCormick, Taber, & Kruedelbach, 1989; Mikul- i ncet & Solomon, 1988). { (inversely, in survivors of the Herald of Free Enterprise and Jupiter ship disasters, internal or controllable attributions lor disaster-related events were associated with poorer outcome (Joseph, Brewin, Yule, & Williams, 1991, 1993).


PTSD shares a number of clinical characteristics with other psychiatric disorders and is rarely diagnosed in isolation (Davidson & Foa. 1991; McNally, 1992). Typically, around 80% of PTSD sufferers receive an additional diagnosis (McFarlane, 1992), although work on combat veterans has suggested that lifetime comorbidity rates may be as high as 99% (Kulka et al., 1990). Epidemiological surveys have reported that rates of somatization disorder, psychosis, anxiety disorder, and depression are substantially elevated in PTSD sufferers (Davidson, Hughes. Blazer, & George, 1991; Helzer et al., 1987: Shore, Vollmer, & Tatum, 1989). Studies of Vietnam veteran populations indicate that depression, generalized anxiety disorder, and substance abuse are the most frequent codiagnoses (Davidson & Poa, 1991).

Numerous authors, noting the high comorbidity of PTSD with anxiety disorder and depression, have pointed to the substantial symptom overlap. For example. Farmer, Tranah, O’Donnell, and Catalan (1992) and McNally (1992) have noted that many depressive symptoms appear in Sections C and D of the DSM-III-R criteria for PTSD: C4, markedly diminished interest in significant activities; C5, feelings of estrangement or detachment from others; C6. restricted affect; C7, sense of foreshortened future; Dl. difficulty in staying or falling asleep; D2, irritability or outbursts of anger; and D3, difficulty in concentrating. Other symptoms, such as guilt, are commonly found in both conditions.

It would not be fair to conclude, however, that symptom overlap is an adequate explanation of comorbidity. Noting the various similarities between depression and PTSD, McNally (1992) suggested that the distinctive features of PTSD are the exaggerated startle, the reexperiencing symptoms (such as nightmares or intrusive memories of the trauma), and physiological reactivity to trauma-related cues. He also reviewed a number ofdiffferences in biological variables and drew attention to various ways in which PTSD differed from panic disorder and phobia (i.e., in the latter, psychic numbing and reexperiencing phenomena are unusual).

Experimental Data

As w’dl as providing an explanatory framework for the wealth of self-report data, any comprehensive theory of PTSD needs to account for the findings from laboratory-based paradigms that do not rely on subjective reports of conscious states. Such experimental methodologies can potentially demonstrate biases in the underlying cognitive processes of perception, attention, judgment, and memory in PTSD sufferers; each of these cognitive mechanisms is considered in turn.

McNally et al. ( 1987) investigated auditory threshold for combat-related and neutral material hoth in participants with combat-related PTSD and in controls. Although all of the participants exhibited a lower threshold for the combat-related words, relative to the neutral control stimuli, McNally et al. found some suggestive evidence that detection of threat targets was accompanied by enhanced skin conductance responding in PTSD, relative to control, participants.

Several studies of attentional processing in PTSD base used a modified Stroop (Stroop, 1935) color-naming paradigm. However, there is some debate about whether this task is a true measure of attentional bias (Dalgleish, 1994; MacLeod, 1990). In the modified Stroop task, participants have to name (as fast as they can) the ink color in which words of personal or emotional significance are written, while attempting to ignore the words themselves. McNally, Kaspi, Riemann, and Zeitlin (1990) and McNally, English, and Lipke ( 1993) found that Vietnam veterans with PTSD were slower to color-name Vietnam-related words relative both to control words and to the performance of a group of veterans without PTSD. Kaspi. McNally, and Amir (1995 ) found that this Stroop effect was strongest for threatening information with a strong degree of self-relevance, although only m individuals with PTSD. Similar effects have been found in other studies of combat veterans (Vrana, Roodman, & Beckham, 1995), as well as in studies of rape victims with PTSD (Cassiday, McNally, & Zeitlin, 1992; Foa, Feske, Murdock, Kozak, & McCarthy, 1991) and PTSD sufferers who have survived a major disaster (Thrasher, Dalgleish, & Yule, 1994).

Memory biases have been found on an autobiographical memory task with Vietnam war veterans (McNally, Lasko, Macklin, & Pitman, 1995; McNally, Litz, Prassas, Shin, & Weathers, 1994). Veterans with PTSD, especially those who wore combat regalia in daily life, exhibited deficits in retrieving specific autobiographical memories to a set of cue words. Finally, Dalgleish (1993) examined judgmental bias in survivors of a major disaster with and without PTSD. It was shown that only those survivors with PTSD generated elevated judgments of the probability of a range of negative events happening in the future.

Cognitive Theories of PTSD

In addition to biological theories (e.g., Kolb, 1987; Van der Kolk, Boyd, Krystal, & Greenburg, 1984), a number of psychological paradigms provide frameworks for understanding PTSD, for example, the psvehodynamic (Freud, 1919; Kar- diner, 1941; Krystal, 1968), learning theory (Fairbank & Brown, 1987; Keane, Fairbank, Caddell, Zimering, & Bender, 1985; Keane, Zimering, & Caddell, 1985; Kilpatrick et af, 1985), and cognitive (e.g., Horowitz, 1986). Other authors have proposed models that integrate psychological and biological perspectives (Jones & Barlow, 1990). Although all of these approaches encompass theories that offer interesting insights inlo the nature of the disorder, it is the cognitive one that is perhaps the most fully developed and offers the greatest explanatory and predictive power. For this reason, in this review, we will focus on the various cognitive theories of PTSD.

Cognitive theories of PTSD share certain core theoretical assumptions. They propose that individuals bring to the traumatic experience a set of preexisting beliefs and models of the world. The experience of trauma provides information that is both highly salient and incompatible with these preexisting models. The attempt to assimilate this new information with the existing models leads, it is argued, to the various phenomena that characterize the posttraumatic reaction. Successful information processing occurs when the new information is integrated into the existing models (often by virtue of changes in those same models). Unsuccessful information processing occurs when the individual is unable to bring the new’ trauma- related information into accord with the current models of the world. This can lead to pathological posttraumatic reactions such as PTSD.

The strength of the cognitive approach is that each of the various theories addresses the majority of the data about PTSD that we have reviewed in the preceding sections. However, each theory endeavors to explain only a circumscribed set of empirical and clinical findings. Data that fall outside of the theoretical spotlight, although addressed, are often merely referred to or redescribed, with no explication of how the theory in question svould explain the findings.

Cognitive theories of PTSD fall into two distinct camps. Social-cognitive theories, such as those of Horowitz and Janoll- Bulman (see below), emphasize the impact of the trauma on individuals’ lives and highlight the massive readjustments that often need to be made to integrate the traumatic experience into an individual’s preexisting views of the world. By empha- sizi ng the wider impact of the trauma and its consequences, they are able to explain other reactions, such as anger, anxiety, and depression, which often accompany PTSD. In contrast, information-processing theories, such as that of Foa and colleagues (e.g., Foa et al., 1989), focus more specifically on trauma-related threat, on how trauma-related information is represented in the cognitive system, and how it is subsequently processed. In the next section, a number of social-cognitive and informationprocessing theories of PTSD are briefly described and areas are discussed in which the theories have some explanatory power and in which they offer no more than a descriptive or embryonic analysis of the data.

Horowitz’s (1973, 1976, 1979, 1986; Horowitz, Wilner. Kal- treider, & Alvarez, 1980) formulation of stress-response syndromes offers perhaps the most comprehensive and influential social-cognitive theory of PTSD. Although influenced by classical psychodynamic psychology (notably Freud, 1920), Horowitz’s theory is principally concerned with the cognitive processing of traumatic information (i.e., ideas, thoughts, images, affects, and so on). Horowitz (1976) argued that the main impetus for such processing comes from a completion tendency: The psychological need for new information to be integrated with existing cognitive world models or schemata (Horowitz, 1986).

Horowitz (1986) proposed that, subsequent to an experience of trauma, there is an initial crying out or stunned reaction followed by a period of information overload, in which the thoughts, memories, and images of the trauma cannot be reconciled with current schemata. As a result, a variety of psychological defense mechanisms come into operation to keep the traumatic information unconscious and the individual experiences a period of numbing and denial. However, the completion tendency helps maintain the trauma-related information in active memory, causing it to break through these defenses and intrude into consciousness in the form of flashbacks, nightmares, and unwanted thoughts as the individual endeavors to merge the new- information with existing models. This tension between the completion tendency, on the one hand, and the psychological defense mechanisms, on the other, causes the person to oscillate between phases of intrusion and denial-numbing as he or she gradually integrates the traumalic material with his or her longterm schematic representations. According to Horowitz {1986), failures of information processing can mean that the partially processed traumatic information remains in active memory without ever being fully integrated, thus leading to chronic posttraumatic reactions.

]Iorowitz ( 1986) has developed an extremely comprehensive theory of stress response. His discussion of the processes underlying completion, intrusion, and denial has considerable explanatory potential for PTSD phenomenology. The theory indicates clearly the ways in which normal reactions to trauma can become chronic or pathological. However, Horowitz’s theory has a number of limitations that are important to highlight. First, there is little discussion of why some individuals develop PTSD, whereas others, after ostensibly similar traumatic experiences, show little or no symptomatology. Indeed, there is little detail on the nature of existing schema structure and the exact ways in which it fails to accommodate new information from the traumatic experience. Second. Horowitz’s formulation struggles to account for epidemiological data regarding the frequency of late onset, although this could be ascribed to a long period of denial that later breaks down. Third, although Horowitz provides a clear description of the time course of posttraumatic reactions, it is far from certain that all individuals do experience an initial period of denial or later oscillations between denial and intrusion. In fact, Creamer et al. (1992). whose ideas are discussed below, argued for an initial episode of intrusive symptoms. Fourth, although Horowitz highlights processes such as social support, there is little explanation within the theory of how such factors might operate. Finally, Horowitz’s theory is somewhat passive. Little credit is given lo the power of the individual’s attributions and interpretations of the traumatic experience and the effect that these have on outcome.

The cognitive-appraisal theory of Janoff-Bulman (1985, 1992; Janoff-Bulman & Frieze, 1983) focused principally on the nature of the preexisting models of the world that the individual carries into the traumatic situation. Janoff-Bulman argued that PTSD is the result of certain basic assumptions about the world that are shattered: the assumption of personal invulnerability, the perception of the world as meaningful or comprehensible, and the view of the self in a positive light.

Janoff-Bulman’s (1985, 1992; Janoff-Bulman & Frieze, 1983) work is important in that it describes the ways in which trauma-related information is incongruent with the usual models and assumptions about the world that people possess. There is little attempt, however, to explain how such models are represented or what processes are involved when they are shattered. An additional problem is the ubiquitous finding (e.g., Kilpatrick et a)., 1985) that individuals with a premorbid psychiatric history are more likely to develop PTSD after a trauma. Sucli individuals would presumably be characterized by assumptions of personal vulnerability and views of the self in a negative light. Such premorbid assumptions are unlikely to be shattered by a traumatic experience (in fact, they are more likely to be confirmed), and the high incidence of PTSD following a trauma in this population has yet to be addressed by cognitive-appraisal theories, such as that of Janoff-Bulman.

Applying Lang’s (1979, 1985) concept of fear structures, Foa and her colleagues (Foa & Kozak, 1986; Foa & Riggs, 1993; Foa etal., 1989, 1992) have put forward an information-processing theory of PTSD that centers around the formation of a fear network in memory. This network encompasses stimulus information about the traumatic event; information about cognitive, behavioral, and physiological reactions to the trauma; and interoceptive information that links these stimulus and response elements. Activation of this fear network by triggering stimuli (i.e., reminders of the trauma) causes information in the network to enter consciousness (the intrusion symptoms of PTSD). Attempts to avoid and suppress such activation lead to the cluster of avoidance symptoms. Successful resolution of the trauma can occur only by integrating the information in the fear network with existing memory structures. Such integration requires, first, the activation of the fear network so that it becomes accessible for modification and, second, availability of information that is incompatible with the fear network so that the overall memory structure can be modified. A number of factors make such integration problematic; Foa and her colleagues (Foa & Kozak, 1986; Foa & Riggs, 1993; l oa et al., 1989, 1992) argued that the unpredictability and uncontrollability of the traumatic event make it difficult to assimilate into existing models in which the world is controllable and predictable. In addition, factors such as the severity of the event disrupt the cognitive processes of attention and memory at the time of the trauma. They argued that this disruption leads to the formation of a disjointed and fragmented fear network that is consequently difficult to integrate with existing organized models.

By outlining an information-processing architecture, within which some of Horowitz’s and Janoff-Bulman’s social-cognitive ideas can be instantiated, Foa and her colleagues ( Foa & Kozak, 1986; Foa & Riggs, 1993; Foa et al,, 1989, 1992) have made considerable progress toward a greater understanding of how the cognitive processes underlying PTSD operate. Furthermore, in stressing factors such as the predictability and controllability of the trauma, they have highlighted one important role of the individual’s attributions and interpretations of the irau- matic event. Also, the proposal that ihc availability of information incompatible with the trauma is necessary for successful information processing provides a framework for undemanding both the role of social support as a vehicle for the provision of such information and the processes underlying the success of exposure-based treatments for PTSD. What is less clear is whether network theory provides an architecture powerful enough to cope with the range of PTSD phenomenology. For example, Foa and her colleagues described in any detail only a network with a single level of representation. Such an analysis struggles with even basic concepts such as denial, numbing, and psychogenic amnesia, which require a higher level of represen- tation that cannot access the information in memory. The term consciousness is used (Foa et al., 1989) lo refer to this higher stratum, but it is unclear whether consciousness is represented by a network or by some other form of cognitive architecture. Furthermore, there is little discussion of how the existing models of the world, which are such a feature of the social- cognitive theories (Foa & Riggs, 1993), are represented by networks. how integration of new information with such models might take place, or of why fear networks develop in some individuals but not in others.

The cognitive action theory of Chemtob et al. (1988) is based on work with veterans of the Vietnam war. It presents a similar perspective to that of Foa et al. (1989), although with more detailed analysis of the structure of the fear network, which is formulated as a parallel-distributed, hierarchical system. Chemtob et al. argued that, in individuals with PTSD, the fear network is permanently activated, causing them to function in a survival mode that has proved adaptive during the traumatic incident. This permanent activation leads to the symptoms of hyperarousal and intrusion.

The cognitive action theory of Chemtob et al. (1988) provides another potential processing framework within which to understand PTSD. It suffers slightly from being overly narrow in its emphasis on combat-related trauma. Also, as with the other theories discussed above, it offers little insight into why some individuals remain in a survival mode, whereas others do not. Similarly, there is no discussion of other variables of known importance, such as attributions and social support.

Creamer et al.’s (1992) cognitive processing theory of PTSD is presented as a “synthesis and reconceptualization of existing formulations” (p. 453). It combines the central ideas of Horowitz with the network architecture of Foa et al. {1989) and Chemtob et al. (1988). Creamer et al. proposed–as did Foa– that the fear network must be activated for recovery to take place–a mechanism referred to as network resolution processing. This concept is similar to Horowitz’s (1986) completion tendency. However, as noted above, Horowitz and Creamer et al. differed in the details of such resolution or completion. Horowitz ( 1986 ) suggested an initial period of denial-numbing, followed by oscillating experiences of intrusion and denial- numbing. In contrast. Creamer et al. argued for an initial period of intrusion (because of activation of the fear network), which the individual copes with by bringing a range of defensive and avoidant strategies into play. Creamer et al. suggested that the extent of initial intrusive symptomatology is an index of the degree of network resolution processing that is occurring. Thus, high levels of initial intrusion are a predictor of successful recovery, whereas low levels of initial intrusion are a predictor of poor outcome and chronic pathology. Creamer et al.’s longitudinal follow-up of the victims of an office block shooting supports these predictions. However, other data (e.g., McFarlane, 1989 ) has suggested that prior levels of high intrusion are predictive of a poorer outcome.

Creamer et al.’s ( 1992) theory is significant in that it is based on longitudinal data and makes clear predictions about outcome. However, it has a number of drawbacks. First, the theory has limited explanatory value; it is an attempt at presenting some interesting correlational data in a loose theoretical framework. Second, the theory gives no indication as to why some individuals develop PTSD and others do not, nor does it account for ihe effects of factors of known importance, such as social support and the individual’s attributions and interpretations of the event. Finally, the strong predictions concerning outcome are only partially supported by the literature (Horowitz, 1986: McFarlane, 1989).

In summary, cognitive theories of PTSD fall into two distinct groups. The primary focus of one is the impact of the trauma on the person’s general assumptions and goals, the generation of complex emotions, and the sometimes profound changes that will be necessary to adjust to the consequences of the trauma. The primary focus of the other is the intrusion of trauma-related flashbacks, the generation of the same intense emotions felt during the trauma, and the need to habituate to trauma- related images and stimuli. The two groups of theories appear to be focusing on phenomena that, although clearly related, may have different underlying mechanisms.

Dual Representation Theory and Trauma Processing What Is Trauma?

By trauma we mean any experience that by its occurrence has threatened the health or well-being of the individual. Just as physical trauma may extend from minor abrasions to severe tissue damage, there is no implication that psychological trauma must involve an event outside the ordinary range of human experience, although it may do so. As various authors have suggested (e.g., Horowitz, 1986; Janoff-Buiman, 1992; Parkes,

1986), trauma generally involves a violation of basic assumptions connected with survival as a member of a social group. These include assumptions (not necessarily conscious ones) about personal invulnerability from death or disease, status in a social hierarchy, the ability to meet internal moral standards and achieve major life goals, the continued availability and reliability of attachment figures, and the existence of an orderly relation between actions and outcomes. The types of events likely to violate these assumptions will frequently involve indications that the world is uncontrollable or unpredictable (Foa et al., 1992), including major illness or disability; physical or sexual assault; social humiliation: transgression of one’s own moral code; loss of employment; divorce and separation; bereavement; and involvement in actual or potential accidents, conflict, and natural disasters. Where such assumptions have already been violated, events that confirm the violation may also be traumatic.

The Representation of Trauma in Memory

A number of cognitive psychologists researching “flashbulb memories” have proposed that a special mechanism exists for encoding emotion-laden memories of events such as the death of President John F. Kennedy (e.g., Brown & Kulik, 1977; Conway, 1995). In general, however, this research has not considered the special properties of memory for personal trauma, particularly the experience of reliving the event. Clinical researchers have attempted to account for the often severe impact of trauma by proposing that it leads to the formation of particularly durable representations in memory. They have suggested that these emotional memories contain a record of the stimulus elements of the event, the person’s physiological and motor responses, and the individual meaning of the event (Foa & Kozak, 1986; Lang, 1979 ). However, the concept of a single emotional memory does not appear adequate to capture the full range of observed phenomena reviewed earlier in this article, for example. the distinction between verbally retrievable memories and flashbacks.

Research in several areas of cognitive psychology, social psychology. clinical psychology, and neuropsychology supports the view that sensory input is subject to both conscious and non- conscious information processing (see Brewin, 1988, 1989: Epstein, 1994, for reviews). It has been persuasively argued that the characteristics of nonconscious processing (e.g., extreme rapidity. parallel processing of multiple inputs) permit far more detailed and extensive computations than does conscious processing, which is limited by its slowness, serial nature, and our inability to hold more than a small amount of information in memory at one time. Similarly, it has been proposed that the output of these different forms of processing are stored in different locations or different codes (e.g.. Broadbent, FitzGerald. & Broadbent, 1986; Iuiving & Schacter, 1990). Neuroanatomical research has located a variety of pathways that might permit sensory data associated with an emotionally significant even! to be stored in memory without being subjected to cortical processing (LeDoux, 1992).

In applying these ideas for the first time to the experience of trauma, Brewin N989 land Terr (1991) suggested that from the outset there will be more than one type of representation of an experience of single or repeated trauma. This proposal is mirrored in related theoretical work, with theories such as Teas- dale and Barnard’s ( 1993 ) interacting cognitive subsystems and Johnson and Multhaup’s (1992) multiple-entry modular memory system, emphasizing the need for more than one level of r epresentation in order to understand the complex relationship between emotion and cognition. The theory presented here builds on earlier work (Brewin, 1989)by proposing dual representations in memory of traumatic experiences as the minimum cognitive architecture within which the complex data reviewed earlier can be understood. As illustrated in Figure 1, one representation or set of representations will be of the person’s conscious experience of the trauma. Brewin ( 1989) termed this verbally accessible knowledge because it can in principle be deliberately retrieved from the store of autobiographical experiences. These memories, although they are likely to be reasonably detailed, may be highly selective because anxiety increases attentional selectivity and decreases short-term memory capacity ( Eysenck & Keane, 1990 ). The verbally accessible memories (VAMs) will contain some information about the sensory features of the situation, the emotional and physiological reactions experienced, and the perceived meaning of the event.

The output of the more extensive nonconscious processing of the traumatic situation will form a second representation or set of representations that cannot be deliberately accessed, Brewin (1989) termed this sitnattonally accessible knowledge because the representations may be accessed automatically when the person is in a context in which the physical features or meaning are similar to those of the traumatic situation. This context may be internal, such as consciously thinking about the trauma, or external, such as hearing about a similar trauma on television. The particular form that such representations may take (schemas, mental models, distributed networks, etc.) is a matter of controversy and is not specified by the theory. Although the two representations are likely to share a number of similarities, they will also differ in important respects. This is because people are able to manipulate to some degree their autobiographical memories, and also because situationallv accessible memories (SAMs) are not subject to the same processing capacity limitations as verbally accessible knowledge. Moreover, in SAMs, the meaning automatically ascribed to events may not correspond to verbally retrievable meanings in terms of consciously held goals and plans (see below).

According to a number of authors (e.g.. Jacobs & Nadcl, 1985), the hormonal effects of acute trauma may act to diminish neural activity in anatomical structures serving conscious processing and to enhance activity in structures serving nonconscious perceptual and memory processes. We propose that the sensory (visual, auditory, olfactory, etc.), physiological, and motor aspects of the traumatic experience are represented in situationally accessible knowledge in the form of analogical codes that enable the original experience to be recreated. These codes would be part of an overall representation that contained (a) stimulus information automatically coded for its ability to discriminate the trauma from other previous nontraumatic situations, using criteria such as novelty {Gray, 1982); (b) meaning information derived from prior associative learning and from innate, nonconscious appraisal mechanisms concerned with the achievement of universal goals such as attachment to caregivers or the regulation of status within a social hierarchy (e.g., Gilbert, 1989); and (c) information about the person’s state of consciousness, for example, their degree of detachment (dissociation) from the traumatic situation. This account, although elaborated in various respects, is similar to Leventhal’s {1984) description of schematic emotional memories. The person may only become aware that these representations have been accessed when they experience symptoms such as emotional arousal, motor impulses, spontaneous intrusive images, or dissociative states.


Figure / Cognitive processing of stimuli relevant to prior trauma. Adapted from “Cognitive Change Processes in Psychotherapy,” by C. R. Brewin. 1989, Psychological Review. 96. p. 383. Copyright 1989 by the American Psychological Association.

In contrast, verbally accessible knowledge of the trauma consists of a series of autobiographical memories that can be deliberately and progressively edited. For example, autobiographical knowledge can be ordered on a hierarchy of generality from broad lifetime themes, to general descriptions of events, to detailed event-specific knowledge (Conway, 1996; Williams, 1992). Immediately posttrauma, these memories are likely to be dominated by detailed information concerning (he conscious perception of sensory detail and of bodily reactions. There will also be an initial attempt to assign meaning to the trauma in terms of verbally accessible constructs and categories, and to consider the implications for valued life goals. As time goes on. however, more generic memories may be created that encompass the fact of having experienced the trauma but without the same level of detail. Similarly, repeated recall of certain aspects of the experience may give opportunities to focus on and embellish some features, whereas the ability to recall other features may be impaired (e.g., Hashtroudi, Johnson, Vnek, & Ferguson, 1994).

There is persuasive evidence from the clinical literature documenting a variety of other, less common memory phenomena. Although they are not the primary focus of this article, they will be briefly mentioned. Thus, under conditions of extreme stress, atlcntional narrowing may be so great that VAMs contain significant gaps, for example, omitting certain moments in time or information from a particular sensory modality. Herman (1992a) and Terr (1991) have suggested a distinction between Type 1 traumas, usually involving a single episode, and Type 2 traumas, where there is repeated extreme stress. They proposed that Type 2 traumas may lead to the development of dissociative reactions, for example, where the person is able to observe him or herself from an external (out-of-body) perspective. It is likely that these disturbances of consciousness may in some cases create additional barriers to the registration and recall of verbal memories. Finally, there are fugue states in which there is impaired access not just to specific VAMs but to autobiographical memory as a whole. Theoretically, we would predict, however, that SAMs would remain intact, despite the impairment in VAMs. Studies of functional amnesia after trauma confirm that individuals may demonstrate implicit memory for relevant details even though they have little explicit memory (Christianson & Nilsson, 1989).

Emotional Processing of Trauma

The idea that trauma is normally followed by a period of emotional processing is, as we have seen, a characteristic of social-cognitive and information-processing accounts, as well as of more general explanations of how cognitive-behavior therapies work (e.g., Rachman, 1980). Accounts differ, however, in the nature of the intrusive memories they describe and in the mechanisms they put forward to explain what drives emotional processing. In the information-processing theories (e.g., Foa & Kozak, 1986), it is the activation of traumatic memories by appropriate cues, whereas in the social-cognitive theories (e.g., Janoff-Bulman. i 992), it is the discrepancy between the trauma and the prior assumptions. We will attempt to synthesize the accounts of this process that we reviewed earlier and explain them in terms of a dual representation theory.

We use the term emotional processing broadly to denote a largely conscious process in which representations of past and future events, and awareness of associated bodily states, repeatedly enter into and are actively manipulated within working memory. We suggest that this process, which may or may not lead to a satisfactory outcome, has at least two elements. One element, described best by information-processing theories, involves the activation of highly specific SAMs, whose function is to aid the process of cognitive readjustment by supplying detailed sensory and physiological information concerning the event (flashbacks). The second element, described best by social-cognitive theories, is the conscious attempt to accommodate the conflicting information supplied by the trauma by searching for meaning and making judgments of cause and blame. The end point of this process is to reduce negative affect by restoring a sense of safety and control and by making appropriate adjustments to expectations about the self and the world. This process often has the unwanted effect of creating internal cues that tend to trigger the spontaneous intrusion of SAMs into consciousness. Both elements are further facilitated by the information-processing system giving high priority to trauma- related cues in the form of attentional and memory biases.

Emotional processing will therefore involve, we suggest, different kinds of emotional reactions. First, there will be conditioned emotional reactions corresponding to the activation of specific emotional states (predominantly fear but also other emotions such as anger) experienced during the trauma, as recorded in the person’s SAMs of the event. In conjunction with these, we, along with numerous other authors (e.g., Oatley & Johnson-Laird, 1987; Weiner, 1985), suggest that emotions such as sadness, anger, and fear for the future will be generated by ongoing threat to safety, the disruption of plans, and the loss of valued goals. Other emotions, such as guilt and remorse (and also anger), may be generated through responsibility attributions. All of these emotions, which we term secondary emotions, follow from the consequences and implications of the trauma. The increased cognitive accessibility of trauma-related information can account for other features of PTSD, such as startle responses and an elevated subjective probability of aversive events occurring.

The aim of this period of emotional processing may, similarly, be seen to consist of two elements. On the one hand, the person needs to actively reduce the secondary negative affects generated by the implications of the trauma by consciously reasserting perceived control, rcattributing responsibility, and achieving an integration of the new information with preexisting concepts and beliefs. 1 his process, described in detail by the social- cognitive theories of Horowitz (1986) and JanofF-Bulman (1992), may involve substantial editing of autobiographical memory (VAMs) in order to bring perceptions of the event into line with prior expectations. For example, the behavior of an attacker previously believed to be trustworthy may be reinterpreted, excused, or explained away, or aspects of what was done or said may be forgotten. Alternatively, previous expectations may be adjusted in line with the event, for example, behaviors, neighborhoods, or locations believed to be safe may be reclassified as dangerous or life goals may be abandoned in favor of less ambitious ones.

The second aspect to emotional processing is to prevent the continued automatic reactivation of situationally accessible knowledge about the trauma. It has been suggested that alterations in situationally accessible knowledge may be brought about by incorporating new knowledge into the original SAMs (e.g., Foa & Kozak, 1986) or by creating new SAMs that block access to the original ones (Brewin, 1989).1 Following activation and emergence into consciousness. SAMs will be automatically altered or added to whenever some or all of the information they contain happens to be paired with changes in concurrent bodily states or contents of consciousness. Changes in bodily states may consist of states of reduced arousal and reduced negative affect. These affective and arousal changes may be brought about by a number of means, including spontaneous or programmed hahituation to the traumatic images. Similar changes would also be expected to follow the conscious restoration of a sense of safety (reduced fear), the abandonment of now unattainable goals (reduced sadness), the absolution of others from responsibility for the trauma (reduced anger), the absolution of self from responsibility (reduced guilt), and other attempts to integrate the new information into preexisting expectations.

As this process of conscious cognitive restructuring continues, the trauma images can be paired with progressive representations of effective action-outcome sequences and reduced negative affect. Reminders of the trauma will then, hopefully, lead to the accessing of more recent representations rather than the original traumatic memories. In the absence of the negative affect, there will also be a reduction in attentional and memory biases and, hence, in the accessibility of the memory.

Conversely, situational representations of trauma related threat may be strengthened if the person continues to be anxious and fearful about the consequences of the trauma. Threat representations may also be altered as a result of stimulus revaluation (Davey, 1989). Subsequent information that increases the perception of threat–for example, that an attacker has previously or subsequently killed someone-–may exacerbate attentional and memory biases and result in an increased likelihood that SAMs will be activated by relevant cues. Equally, information that decreases the perception of threat may result in a reduction in selective cognitive biases and the decreased accessibility of situational memories.


Table 1

Outcomes of Emotional Processing

Outcome Description
Completion/integration No memory bias No attentional bias
  No symptoms
Chronic emotional processing Memory biases Attentional biases
  Phobic state Depression Panic Anxiety

Substance abuse

Premature inhibition of processing Attentional biases Avoidance schema Impaired memory Phobic state Dissociation Somatization

As has frequently been noted, the affect accompanying trauma can be so overwhelming that processing may be invol untarily suspended and the individual may experience emotional numbing. Alternatively, people may attempt to deliberately interfere with this processing by distracting themselves and systematically avoiding either reminders of the trauma in general or reminders of specific aspects of the trauma. The various processes of information retrieval, conscious reappraisal, emotional numbing, and cognitive and behavioral avoidance allow for a number of quite different outcomes to the processing of any particular trauma.

Three Endpoints of Emotional Processing

A variety of well-known strategies are commonly used for resolving discrepancies between preexisting schemas and new information. These include externalization of responsibility, disengagement of self-evaluative processes, victim blaming, denial of the validity of the new information, and revaluation of previously held goals (see Fiske & Taylor, 1991, for a review). The nature of trauma is such, however, that previously successful strategies may no longer be effective and more radical solutions may be required. As we have seen, the outcome of the period of emotional processing is likely to be influenced by a number of factors, including the severity and length of the trauma, its meaning to the person, accompanying emotions such as guilt or shame, and the availability of appropriate social support that enables the survivor to confide. Three possible outcomes may be distinguished: completion/integration, chronic emotional processing, and premature inhibition of processing (see Table 1). We describe these outcomes in more detail.

Completion/integration. Completion or integration represents the ideal stage in which the memories of the trauma have been fully processed, or worked through, and integrated with the person’s other memories and sense of self in the world. In part, this involves reducing negative affect by restoring a sense of control and by resolving discrepancies with preexisting expectations and goals. There must be sufficient repetition of the incident in memory for the person to accept the reality of what has happened to them and its consequences without being overcome by the accompanying emotions. The threat of recurrence in the future can be assessed realistically, and experiences that induced guilt or shame and have implications for the self-concept lead to adjustments in self-expectations. Horowitz (1979) noted that before major interpersonal loss can be accepted, the survivor has lo adjust his or her internal model of self and of the relationship to the lost person. However, as noted by Janoff- Bulman (1992), as a result of these adjustments, quite profound changes may occur and the person may feel that he or she is no longer the same as before the trauma occurred.

Reductions in negative affect may permit the person to tolerate the intrusion of situational memories and to allow habituation or stimulus revaluation to take place. Habituation is likely to be brought about by repeated exposure to the memories in states of gradually increasing calm and relaxation. Social support has an obvious role to play here, in directly reducing negative affect by offering physical comfort and emotional support, in providing opportunities for the repeated rehearsal of the traumatic memories, and in assisting in the process of cognitive reappraisal.

Theoretically, we would predict that successful integration would be marked by the absence of attentional and memory biases for trauma-related stimuli; in other words, these stimuli would no longer be deemed sufficiently important to be accorded preferential processing. Successful integration should be achieved under the following conditions: small discrepancies between trauma information and prior assumptions (a result of less intense and nonrepeated stressors, less negative self-evaluation, lack of previous trauma, less extreme assumptions, and less rigid goals), adequate cognitive development, good social support, and ability to tolerate the intrusion of SAMs into consciousness.

Chronic emotional processing. For a variety of reasons complete integration may represent an ideal that is not possible to attain. The trauma may have been so severe and so prolonged, or have such profound consequences for the person’s sense of self and of future safety, that the discrepancy between the trauma and the prior assumptions is too great and the memories cannot be integrated. Alternatively, the person may not have been able to process memories of the trauma effectively because of (a) competing demands, (b) aversive secondary emotions,

(c) lack of an appropriate confidant or unwillingness to confide,

(d) being too young to appreciate the meaning and circumstances of the event, and (e) the presence of ongoing trauma or threat that continually reactivates trauma memories.

One possible result is that VAMs and SAMs concerning the trauma are chronically processed, with the result that the person is permanently preoccupied with the consequences of the trauma and with intrusive memories. The processing may be largely repetitive with little or no change being effected to existing representations. There is now considerable evidence (reviewed earlier) that a proportion of war veterans and survivors of rape or of man-made disasters develop chronic PTSD that may last for months or years.

In these circumstances, the characteristic features of acute emotional processing, such as heightened arousal and attentional and memory biases, are likely to be chronically present. Also, secondary reactions such as depression, cognitive and behavioral avoidance, or anxiety and panic may develop, reflecting generalization of the impact of chronic processing. For example, patients may experience anxiety phenomena thought to be associated with defensive processes such as obsessional thoughts (Salkovskis, 1985) or worrying (Borkovec&Lyonfields, 1993). These cognitions may not necessarily have any direct trauma- relevant content. Substance abuse may also develop as a coping response. We suggest, therefore, that in many cases comorbidity of PTSD with depressive, anxiety, or substance abuse disorders is likely to reflect the effect of chronic emotional processing.

Chronic processing is likely under the following conditions: large discrepancies between trauma information and prior assumptions (a result of more intense or repeated stressors, more negative self-evaluation, presence of previous trauma, more extreme assumptions, and more rigid goals), inadequate cognitive development, poor social support, and inability to prevent the intrusion of SAMs into consciousness. This inability may be linked to resolutions of the discrepancy in which previous assumptions and beliefs have changed in the direction of seeing the self and the world in much more negative terms. Ehlers and Steil ( 1995) also presented evidence suggesting that distressing misinterpretations of the PTSD symptoms themselves (e.g„ interpreting the occurrence of intrusive recollections as a sign of brain damage or impending breakdown) lead to more chronic PTSD.

Premature inhibition of processing. The other possible outcome is that emotional processing will be prematurely inhibited. Usually, inhibition is the result of sustained efforts to avoid the reactivation of unpleasant SAMs and VAMs. Trauma victims frequently describe strategies they use for avoiding thinking about the trauma and hence escaping the accompanying emotional arousal. Sufficient repetition of these strategies may result in this process becoming automatic. For example, an avoidance schema may develop that monitors sensory input for trauma-related stimuli and directs conscious attention away from them. Similarly, the person may develop trauma-related scripts that enable him or her to incorporate the autobiographical fact of the trauma into his or her VAMs and to have limited communication about it without reactivating the SAMs.

In this state there will no longer be any active emotional processing, intrusive memories, or deliberate attempts to avoid intrusions, but the SAMs concerning the trauma should still be accessible in the right circumstances. We therefore predict that it should be distinguishable from the state of completion/ integration in the following ways: (a) Attentional biases should still be present, indicating that trauma-related stimuli are still being accorded high priority; (b) the person may have impaired memory for the trauma or for trauma-related stimuli; (c) the person should show phobic avoidance of trauma-related situations; and (d) the person may show evidence of somatization.

The significance of this type of outcome is that, although the person may appear to have recovered from the effects of the trauma, the unprocessed memories remain vulnerable to reactivation later in life. This may occur when the person encounters similar situations or enters similar mood states. Premature inhibition of processing is likely under the following conditions: large discrepancies between trauma information and prior assumptions, inadequate cognitive development, poor social support, and the ability to prevent the intrusion of SAMs into consciousness. This ability may be linked to the use of dissociative defences during the trauma or to a general facility to avoid processing negative information. People with a repressive coping style appear particularly adept at this, showing attentional avoidance (Fox, 1993). impaired memory for negative stimuli (Myers & Brewin, 1994, 1995), and unrealistically optimistic assumptions and beliefs (Myers & Brewin, in press).

A large volume of research has now been carried out on the physiological effects of not expressing deeply felt emotions (e.g., Pennebaker, Kiecolt-Glaser, & Glaser, 1988). It appears that inhibiting the expression of emotion leads to impaired immune function and poorer health status on a variety of indices.

Relation of Dual Representation Theory to Existing Evidence

Although the dual representation theory of PTSD has not been directly tested, it is appropriate to review briefly its relation to existing findings.

Clinical characteristics of PTSD. The theory deals most directly with the intrusions and avoidance that are typical of PTSD. Unlike other theories, however, it distinguishes different types of intrusive phenomena, represented by repeated accessing of VAMs and SAMs. Repeated accessing of VAMs is accompanied by emotional reactions such as anger, sadness, and guilt linked to loss, future threat, and the assignment of blame. VAMs may be edited in various ways to control affect by emphasizing positive or negative aspects or by creating more or less detailed representations of the trauma. In contrast, SAMs tend to be highly detailed, repetitive memories (flashbacks) that are difficult to edit and that arc accompanied by emotional and physiological changes experienced during the trauma. The greatly enhanced accessibility of VAMs and SAMs and their associated emotional reactions lead to the familiar symptoms of arousal.

The idea that conscious emotional processing may be prematurely inhibited is another feature of our dual representation theory that has not previously received much attention. This feature means that the theory can account more easily than other theories for the unpredictable time course of PTSD and with the frequent observation that, with exposure to appropriate cues, emotional processing may resume years after it appealed to have ceased. It also accounts for clinical observations of patients who appear never to have consciously processed a trauma but have subsequently experienced chronic nightmares. This would represent an extreme example of the inhibition of conscious processing.

Symptoms not addressed in detail by the theory include emotional numbing and dissociative reactions. In common with other authors (e.g., Herman, 1992b; Van der Kolk & Fisler, 1994), we view emotional numbing as a kind of psychic analgesia, analogous to the accompanying physical analgesia, that protects the person from overstimulation. Whereas emotional numbing appears to be a preprogrammed response that is not under conscious control, dissociation may be a learned response to prolonged or repeated trauma (Herman, 1992a). Our dual representation theory suggests that both of these responses to a specific trauma may be coded within the corresponding SAMs and, when triggered by an appropriate cue, may be reinstated along with other aspects of situationally accessible knowledge.

Normal or abnormal process? Unlike other theories of PTSD, our dual representation theory explicitly includes consideration both of a general response to trauma and of mechanisms that are specific to PTSD. Symptom duration of 1 month, necessary for a diagnosis of PTSD in DSM-IV, is likely to capture individuals who, in terms of our theory, are in the process of successfully completing their trauma processing, as well as those who will go on to experience chronic emotional processing or who will prematurely inhibit this processing.

In this way the theory can account for, and indeed predicts, the conflicting findings concerning the relation between intrusive memories and subsequent symptoms (Creamer et al., 1992; Foa & Rothbaum’s, 1990. study, as cited in Rothbaum & Foa. 1993; Joseph, Yule, & Williams, 1994; McFarlane, 1992; Shalev, 1992). It suggests that there are two relevant moderator variables, the degree of trauma and the length of time elapsed since the trauma. With mild trauma, only a proportion of individuals are likely to experience intrusive memories for more than a few days, and more prolonged intrusions may therefore indicate an increased probability of later psychiatric disorder. In contrast, the presence of intrusive memories immediately after substantial trauma is a normal reaction that, as Foa and Rothbaum’s (1990) study (as cited in Rothbaum & Foa, 1993) and Shalev (1992) found, will not predict subsequent adjustment. After some weeks or months, however, continuing emotional processing is likely to signal a failure to resolve discrepancies in expectations and goals and an inability to curtail unwanted memory reactivation. The longer the time elapsed since the trauma, therefore, the more likely this symptom is to predict a poor outcome (e.g., Joseph, Yule, et ah, 1994; McFarlane, 1992).

It should lie noted, however, that individuals who, following recent significant trauma, score low on measures of intrusion may have prematurely inhibited emotional processing. This group is predicted to be at enhanced risk of later psychiatric disorder. It is not clear whether such a process can account for Creamer et al.’s (1992) data, in which measures of memory intrusion taken 4 months after an office shooting were related to later outcome, this is because their raw correlations indicated that more intrusions related to a worse outcome, whereas structural equation modeling suggested the opposite relation.

Other predictors of severity. Many factors predictive of severity appear to relate to both of the aspects of emotional processing described in this article. Thus, the intensity of the stressor may impede both appraisal of the implications of the trauma, because of the greater discrepancy with prior assumptions, and the person’s ability to tolerate flashbacks. Prior adversity and prior psychiatric disorder may affect both conscious and nonconscious appraisals of the degree of threat posed by the trauma. The content of SAMs is likely to be strongly influenced by similarities between the content and context of current and prior adversity. Social support, as already indicated, may aid both in the reappraisal of the present and the future and make the occurrence of flashbacks more accepted and tolerable.

Of particular interest to the theory are reports that emotions such as guilt and anger predict the maintenance of PTSD symptoms (Riggs, Dancu, Gershuny, Greenberg, & Foa, 1992). This is consistent with the proposal that aversive secondary emotions interfere with habituation to trauma SAMs and are associated with more chronic emotional processing.

Comorbidity. According to this dual representation theory, PTSD is distinguished from other disorders by the existence and current levels of activation of trauma-related SAMs. Other disorders, such as depression, phobia, or generalized anxiety disorder, do not require that there be trauma-related memories, although these may be frequently present (whether currently more or less accessible). PTSD also requires that these memories are currently activated, as indicated by the presence of spontaneous intrusive phenomena or conscious efforts at avoidance, or both.

It seems likely that the theory can account for recent observations by Kuyken and Brewin (1994) and Brewin, Phillips, Carroll, and Tata (in press) that patients with a diagnosis of major depressive episode also experience frequent intrusive memories, particularly of past stressors, such as childhood trauma. Our theory requires that the processing of these events was prematurely inhibited, allowing emotional memories to remain dormant and to subsequently be reactivated by later depression or later life events. This is plausible in that numerous factors are likely to make it difficult for children to process abusive memories. Even if children disclose abuse (and there are often powerful pressures on them not to disclose), they may not be believed, and even if they are believed there may be no adult sufficiently knowledgeable, empathic, and responsible to facilitate this process.

There are many possible reasons for the high rate of comorbidity with other disorders. For example, familial aggregation of psychiatric disorders in the relatives of patients with PTSD may implicate more general biological and psychological vulnerabilities that are important in the genesis of PTSD (Jones & Barlow, 1990). Unlike other information-processing theories, however, dual representation theory offers additional specific suggestions to explain the high rale оГ comorbidity with other disorders. For example, depression may result as an emotional reaction to the actual or symbolic loss consequent to the trauma. This may consist of the loss of another person or type of relationship, a highly valued role, health, physical and mental capacities, a future goal, or a sense of a good and effective self. Loss of a belief in the self as good (e.g., caring, altruistic, moral) or effective (e.g., strong, self-controlled) may lead to depressions characterized by feelings of guilt or shame. Depression may also develop more slowly as a secondary reaction to prolonged emotional processing and the accompanying feelings of powerlessness and loss of mental capacity.

One of the well-known effects of depression is to enhance access to negative memories and to reduce the accessibility of positive memories (e.g., Brewin, Andrews, & Gollib, 1993; Dalgleish & Watts, 1990). Teasdale ( 1983) has argued that the greater accessibility of negative memories further potentiates depression, leading to a vicious circle. It seems likely that depression preexisting a trauma could also operate on traumatic memories and tend to prolong emotional processing, resulting in more chronic PTSD.

According to the theory, initial anxiety reactions immediately posttrauma reflect both emotional reactions to continuing or subsequent threat and conditioned fear reactions contained within reactivated trauma SAMs. This double dose of anxiety, also noted by Janoff-Bulman (1992), results in extremely high levels of arousal, sleep disturbance, and startle responses, and may automatically elicit emotional numbing. These initial levels of anxiety generally begin to decline as the person develops strategies of cognitive avoidance to titrate their exposure to trauma cues.

After this initial period, anxiety reactions directed at trauma- related stimuli will usually signal unsuccessful emotional processing, whether chronic or prematurely inhibited. Whereas these reactions will tend to take the form of a phobia (or possibly obsessional thoughts or worrying) if the cues are avoidable, somatic anxiety and possibly panic may develop if the person is exposed to unavoidable trauma cues. Depending on the success with winch trauma cues can be cognitively and behaviorally avoided, there may be an increased incidence of substance abuse as the person makes efforts to restore control over his or her mental processes.

Experimental findings. Dual representation theory accounts for the experimental findings using laboratory tasks by- proposing that the content of both VAMs and SAMs is able to bias a wide range of cognitive processes. Perception and attention processes are biased toward information that is congruent with the individual’s active plans, goals, and concerns (Dalgleish, 1994; Eysenck, 1992). According to dual representation theory such information concerning plans, goals, and concerns is represented as both VAMs and SAMs. Consequently, fallowing a traumatic event, trauma-related information would be represented in both V-\M and SAM formats, and both types of information would be able to drive perception and attention processes to information in the environment that is congruent with the trauma. Biases of perception and attention toward trauma-related information would therefore be expected in individuals suffering from PTSD (e.g.. McNally et al„ 1990).

Extant research using experimental tasks to investigate perception and memory has rarely distinguished between different groups of individuals who do not have PTSD. However, dual representation theory, unlike other cognitive theories of the disorder, would make differential predictions dependent on whether an individual had successfully integrated the traumatic- event or had prematurely inhibited processing of the trauma. The theory would predict that those who had successfully integrated the traumatic experience would show no biases in perception anti attention, as memory of the trauma would not be accompanied by high levels of emotion in either VAM or SAM. In contrast, for those individuals who have inhibited emotional processing prematurely, the memory of trauma would not be accompanied by high levels of emotion in VAM, but would be accompanied by high levels of emotion in SAM. In this case, dual representation theory would predict that these individuals would persist in exhibiting biases of attention and perception for trauma-related information.

Dual representation theory proposes that judgment and memory biases for trauma-related information in individuals with PTSD on laboratory tasks (e.g,, Dalgleish. 1994) are a function of the increased accessibility of such information in VAM. Individuals without PTSD. whether they had successfully integrated the trauma-related information or had prematurely inhibited processing, would not experience increased accessibility of trauma-related VAMs. However, those who had prematurely inhibited processing would experience decreased accessibility of trauma-related VAMs.

Predictions and Recommendations

Our dual representation theory makes a number of unique, testable predictions. Some of these have already been outlined: for example, the role of severity of trauma and time elapsed since the trauma in moderating the relations between measures of reexperiencing and subsequent psychiatric adjustment. We will discuss first those additional predictions that are connected with the distinction between VAMs and SAMs. According to the theory, SAMs contain a large amount of detailed information selected partly on criteria that are not available to conscious inspection. Like the implicit memories studied by cognitive psychologists, they cannot be deliberately retrieved and, in addition, are repetitive and difficult to modify. VAMs, on the other hand, like explicit memory, are less detailed and easy to edit, this leads to the prediction that verbal descriptions of the trauma should differ, depending on whether the corresponding SAMs have been activated. Verbal accounts that are unaccompanied by the subjective experience of intense fear or of reliving the trauma should be more variable and should contain less detail than verbal accounts accompanied by the sensation of reliving the trauma. Whereas the former should become more schematic and less specific over time, the latter should remain highly consistent, even after many years.

We would also predict that conditioned emotional reactions (flashbacks) could be elicited by aspects of the traumatic situation that were not part of survivors’ initial verbal descriptions of the trauma. Flashbacks should also be more readily elicited by aspects of the traumatic situation prominent in SAMs than by aspects prominent only in VAMs.

A further prediction is that behaviorally based treatment simply involving exposure to the traumatic memories should be effective only at extinguishing emotional reactions, predominantly fear, experienced during the trauma itself. It should not extinguish secondary emotional reactions arising from subsequent conscious appraisal, which are not expected to habituate and should respond only to cognitive therapy techniques. Thus, exposure treatment should be more effective when recalling the trauma does not give rise to aversive secondary or complex emotions such as anger or guilt. According to the theory, if the trauma SAMs activated in therapy are repeatedly paired with aversive secondary emotions arising from conscious appraisal, there will be little or no overall reduction in negative affect and, hence, the habituation of fear will be blocked. The theory suggests, therefore, that aversive secondary emotions should be addressed using cognitive techniques (e.g., Resick & Schnicke, 1993 } before exposure treatment is used. It is of interest in this regard that patients who reported more anger prior to therapy and displayed less fear during exposure treatment tended to have a poor outcome (Foa, Riggs, Massie, & Yarczower. 1995).

Other specific predictions relate to the idea that emotional processing may be prematurely inhibited. We predict that among people who have formerly had PTSD but now show no evidence of active emotional processing, there will be a substantial subgroup who (a) show’ a preattentional bias to attend to trauma-related stimuli; (b) show strong priming effects in response to such stimuli; (c) avoid elaborative processing of trauma-related stimuli, resulting in impaired memory for this material; (d) show phobic avoidance of trauma-related stimuli;

(e) show enhanced sensitivity to life events: (f) report more dissociation at the time of the trauma; (g) have unrealistically positive assumptions and beliefs: and (h) show evidence of impaired health status.

The most important recommendation following from the theory is that the cessation of active emotional processing cannot be taken as a guarantee that successful integration of the trauma has taken place. Clinically, therefore, apparent improvement may be misleading. The theory implies a need for careful assessment of phobic avoidance and of attentional and memory biases before concluding treatment for PTSD. Equally, however, the theory implies that individuals entering treatment for other disorders may have inhibited the processing of traumatic events. A careful history of traumatic events should therefore be carried out in all cases, and evidence should be sought for successful or unsuccessful emotional processing. Once again, evidence of cognitive or behavioral avoidance may be a useful marker of incomplete processing.

The research implications are equally wide ranging. If correct, the theory implies that use of traditional outcome measures in naturalistic follow-up studies or treatment trials for PTSD may lead to serious underestimation of the extent of pathology and possible false conclusions concerning the efficacy of different therapeutic techniques. For example, some techniques may only be effective in inhibiting active emotional processing and giving the impression of improvement, rather than in leading to a complete integration of the trauma. Equally, prognostic factors thought to be associated with a good outcome may in fact be associated with premature inhibition of trauma processing. This may help to account for the failure to identify the same prognostic factors in different studies.

The theory would also caution against designing theoretical studies in which individuals with current PTSD are compared with individuals exposed to the same trauma but without current PTSD. This group could contain a mixture of individuals who had successfully completed emotional processing and others who had prematurely inhibited it. As each subgroup would be predicted to have different cognitive characteristics, misleading conclusions could be drawn. We suggest that individuals without current PTSD are screened for the presence of the indicators of incomplete processing described earlier,

A final recommendation concerns measures of intrusive memories. The theory suggests that there may be different types of memory, VAMs and SAMs, but existing scales such as the widely used Impact of Event Scale (Horowitz, Wilner, & Alvarez, 1979) do not capture this distinction. We suggest that it may be profitable to examine the phenomenology of intrusive memories in more detail, with a view to developing more comprehensive measures.


In dual representation theory, PTSD is viewed as a particular type of unsuccessful adaptation to trauma. By emphasizing its relation to trauma processing in general, the theory attempts to build on the current understanding of the disorder offered by other theories and to highlight the overlap with other disorders. The two innovatory elements arc a minimum cognitive architecture involving the distinction between verbally and situation- ally accessible knowledge and the idea of prematurely inhibited processing. We believe the first is necessary to explain the different types of memory associated with P I’SD and to integrate existing social-cognitive and information-processing theories. The second idea also appears heuristic in several respects. Not only does it account for a range of clinical and experimental dala and generate novel predictions, but it draws attention to the underlying connection between what may appear to be two unrelated therapeutic tasks. One task is to terminate the distress occasioned by the chronic processing of trauma found in PTSD. The other, often far less obvious, is to detect a history of trauma, encouraging the individual to overcome inhibitory processes and restart the painful processing of traumatic memories. We believe that a focus on trauma processing has the potential to forge interesting links between areas of clinical practice and research that have hitherto been regarded as separate.


Functional Relaxation and Guided Imagery as Complementary Therapy in Asthma: A Randomized Controlled Clinical Trial

  1. Lahmannae M. Nickelf T. Schusterb N.Sauerd J. Ronela M. Noll-Hussonga K.Tritte D. Nowakc F. Rohricht9 T. Loewe

department of Psychosomatic Medicine and institute for Medical Statistics and Epidemiology, Technische Universitat Munchen, clnstitute and Outpatient Clinic for Occupational and Environmental Medicine, Ludwig Maximilian University, Munich,d Department of Psychosomatic Medicine, University Medical Center Hamburg-Eppendorf and Hamburg-Eilbek (Schon Clinics), Hamburg, and eDepartment of Psychosomatic Medicine, University of Regensburg Medical Center, Regensburg, Germany; department of Psychosomatic Medicine,

Medical University of Graz, Graz, Austria; 9 University of Hertfordshire and Unit for Social and Community Psychiatry, Queen Mary College, University of London, London, UK



Asthma • Functional relaxation • Guided imagery • Body psychotherapy • Randomized controlled trial


Background: Asthma is a frequently disabling and almost invariably distressing disease that has a high overall prevalence. Although relaxation techniques and hypnotherapeu- tic interventions have proven their effectiveness in numerous trials, relaxation therapies are still not recommended in treatment guidelines due to a lack of methodological quality in many of the trials. Therefore, this study aims to investigate the efficacy of the brief relaxation technique of functional relaxation (FR) and guided imagery (Gl) in adult asthmatics in a randomized controlled trial. Methods: 64 patients with extrinsic bronchial asthma were treated over a 4-week period and assessed at baseline, after treatment and after 4 months, for follow-up. 16 patients completed FR, 14 Gl, 15 both FRandGI (FR/GI)and 13 received a placebo relaxation technique as the control intervention (Cl). The forced expiratory volume in the first second (FEV-|) as well as the specific airway resistance (sRaw) were employed as primary outcome measures. Results:Participation in FR,Gland FFi/GI led to increases in FEVt (% predicted) of 7.6 ± 13.2,3.3 ± 9.8, and 8.3 ± 21.0, respectively, as compared to-1.8 ± 11.1 in the Cl group at the end of the therapy. After follow-up, the increases in FEVi were 6.9 ± 10.3 in the FR group, 4.4 ± 7.3 in the Gl and 4.5 ± 8.1 in the FR/GI, compared to -2.8 ± 9.2 in the Cl. Improvements in sRaw (% predicted) were in keeping with the changes in FEVi in all groups. Conclusions: Our study confirms a positive effect of FR on respiratory parameters and suggests a clinically relevant long-term benefit from FRas a nonpharmacological and complementary therapy treatment option.

Copyright © 2009 s. Karger AG, Basel


Asthma is a frequently disabling and almost invariably distressing disease that has a prevalence of between 2 and 8% in adults [1] and up to 21% in children [2]. According to Eder et al. [3], there is evidence that, in some areas of the Western world, the prevalence of asthma may have plateaued at this very high level.

Asthma is a disease of multifactorial etiology [4, 5]. Regarding the modulation of the course of disease, Mathe and Knappe [6] have already found that psychological stress is associated with a decrease in airway resistance in healthy individuals but with an increase in those with asthma. On the one hand this presents a fundamental paradox, in as much as several components of the generic stress response (i.e. glucocorticoid release from the adrenal cortex or increased activity in the sympathetic nervous system) should in principle lead to bronchodilatation rather than bronchoconstriction. On the other hand, it shows that the idea of a generic stress response is an oversimplification and the precise psychophysiological regulation of respiratory airflow remains unclear [7].

However, since the importance of emotional factors in triggering asthma attacks has been recognized [8, 9], it has been suggested that psychological treatment methods may influence airway caliber by balancing the activity of the autonomous nervous system. Several psychological interventions may be employed to ameliorate health problems associated with asthma, as summarized by Yorke et al. [10]. There seem to be differential effects, with some evidence for a reduction in relief medications after relaxation therapy, improvement of asthma-related quality of life following cognitive behavioral therapy and peak expiratory flow outcome being positively influenced in biofeedback therapy [10].

The adequate management of asthmatic complaints relies heavily on the patient’s ability to detect changes in breathing, to assess these changes accurately and to respond to them appropriately [11]. These considerations have led to the use of complementary relaxation techniques and body-psychotherapeutic methods for improving body awareness and better regulation of the autonomic nervous system.

A large body of literature [12] exists regarding the use of relaxation techniques in the treatment of the symptoms of asthma. These techniques include jacobsonian progressive relaxation [13], hypnotherapy [14], autogenic training [15], biofeedback training [16] and approaches based on mindfulness, such as yoga [17]. As shown in a critical review by Brown [18], hypnosis is efficacious in managing emotional states that exacerbate airway obstruction and may possibly stabilize airway hyperresponsiveness in some individuals. Freeman and Welton [19] reported that imagery interventions led to significant symptomatic improvements in adult asthmatics. Promising results have also been shown in several studies using the technique of functional relaxation (FR), a brief relaxation technique based on the principles of psychodynamic body psychotherapy. In two trials [20, 21] FR led to significant improvements in pulmonary function in asthma patients, with effect sizes comparable to those induced by inhaled p2-sympathomimetics.

Despite these findings, there is still a lack of evidence regarding the efficacy of relaxation therapies in the management of asthma, which is largely due to some methodological limitations in previous studies [12]. Many of the previous trials do not fulfill the criteria of randomized controlled trials as defined by the Consort Statement [22]. Information regarding dropouts is often missing and the studies do not include a follow-up period for detecting the long-term stability of achieved short-term effects. Because of the above, relaxation and other complementary techniques are still not recommended in current asthma treatment guidelines [23].

We therefore carried out a randomized controlled trial that aimed to compare the brief relaxation technique of FR with guided imagery (GI) in adult asthmatics. In addition to the methods applied in previous studies that addressed the efficacy of FR [20, 21], the present trial focused on the long-term effects of FR by including a 4- month follow-up assessment.



Patients suffering from allergic bronchial asthma due to dust mite allergy were recruited using a local newspaper advertisement, which was answered by 291 patients. The inclusion criterion was a diagnosis of controlled allergic asthma already established by a pulmonologist/allergologist at a university hospital outpatient clinic for respiratory diseases. Diagnostics included a positive history of dyspnea upon exposure to house dust, physical examination, lung function tests and a positive prick test to a standardized dust mite extract (Dermatophagoides pteronyssinus and/or Dermatophagoides farinae) with a wheal diameter at least 3 mm greater than the negative control. According to this criterion, 75 suitable patients were identified. Exclusion criteria were defined as follows: age under 18 years, severe psychiatric or somatic disease other than allergic asthma, current use of any psychoactive medication or psychotherapy (including regular relaxation training), use of medication (anything other than the usual types of asthma medication prescribed by a chest physician) or modification of asthma medication during the previous 4 weeks or within the course of the trial. Furthermore, basic asthma medication was kept stable throughout the trial.

Current mental illness was excluded using a clinical face-to- face interview. Potential participants received a comprehensive description of the study specifics, and 64 patients gave their written informed consent to participate. Those included were randomized to FR, GI, FR and GI (FR/GI) or a placebo relaxation technique as the control intervention (Cl) using randomized numbers generated by an Excel table (fig. 1). Randomization was carried out in confidence by a study nurse, with allocation concealment using a randomization list created before the study.





Fig. 1. Flow diagram of the trial progress.


With a sample size of 14 patients per group, the study was designed to detect differences in specific airway resistance (sRaw) change between the control and intervention groups of 130 kPa/s with 80% power, assuming a common standard deviation of 100 kPa/s and considering an overall type I error level of 0.05. These assumptions correspond to the results of a small pilot study. 64 patients were randomized, with a loss to follow-up of 5 patients due to incomplete or missing data.


The standard pulmonary function indicators were measured using a spirometer and a body plethysmograph (Jaeger®), always at the same time of the day. These included standard assessments of forced expiratory volume per second (FEV!) as a percentage of the predicted value and the sRaw, also reported as a percentage of the predicted value. The best of 3 satisfactory FEVi and sRaw tests was recorded. Normal values were calculated from the sex, height and age of each patient, in order to compare the measured values with the normal values [24]. The actual value of the lung function test is calculated as a percentage of the nominal value. The evaluation considers the FEVi and the sRaw as products of the respiratory resistance and the intrathoracic gas volume. As it is independent from patient cooperation, the sRaw is considered to be a sensitive and reliable measure of resistance [24]. An increase in the FEVx associated with a decrease in the sRaw indicates a bronchodilatation. The average duration for a complete pulmonary function test was approximately 7 min. The data were analyzed by an online digital computer.


The time structure of the trial was identical for all conditions. After giving informed consent, the patients were randomized to FR, GI, FR/GI or the Cl. The interventions were carried out over a period of 4 weeks in small groups of 3-5 persons, with 1 group session per week. The participants underwent lung function tests directly before (tO) and after finishing the therapies (tl) and again 4 months later (t2). The test was always taken at the same time of the day using a standardized diagnostic procedure, as recommended by the European Respiratory Society [24]. On assessment days, the patients were not allowed to take any short-acting medication before the examination, which prevented false results caused by potential contamination of effects. The study was completed according to the plan.


All interventions were carried out by a physician specialized in psychosomatic medicine and certified in functional relaxation and hypnotherapy, who was assisted by a PhD student.

FR is a somatopsychotherapeutic intervention technique commonly used in Germany, Austria and Switzerland for the treatment of psychosomatic disorders [25,26]. The therapeutic effects are assumed to be delivered by positive stimulation of the autonomic nervous system as well as by facilitation of proprioreceptive awareness [27]. Minute movements of small joints, hardly noticeable to observers, are performed during relaxed expiration, accompanied by a focus on and exploration of the perceived differences of body feelings triggered by these movements. The focus of attention is thereby directed towards the way the person relates to the environment, particularly the floor as the foundation and as the ‘outer support’ to the bony skeleton (referred to in FR as ‘the frame’ or the ‘inner support’), to the interior regions of the body and to the skin as an ‘outer border’. Unlike exercise-based methods, such as progressive muscle relaxation, the bodily perceptions are explicitly verbalized and processed in the therapeutic relationship.

In the FR treatment, the group therapy was guided by a manual which was generated during previous investigations [25, 28]. This concept has been adapted to the special features of asthmatic patients.

The GI intervention consisted of visualizing a set of standardized guided imageries targeting the airways [29]. In the first session, participants were instructed in positive visualization of a situation with deep ‘free’ breathing. Additionally, the patients were encouraged to integrate this positive imagination into their daydreams. The second session started with a short educational section about allergy and the relevance of mast cells, followed by an exercise to imagine the mast cells acting to alleviate the hypersensitive bronchial system. The third session focused on a healthy self-perception, including imagining a better physical condition in the future. The last session provided time for repetition of the aforementioned topics.

In the FR/GI group, the FR units were augmented by elements of imagery in each session. In contrast, the Cl consisted of isotonic exercises, which required – in comparison to FR – an equivalent amount of movement, but which did not involve the chest, shoulders, neck or head. This intervention did not focus on enhancing bodily selfawareness, as in FR, neither did it include imaginative elements, as in GI. During the isotonic exercises the patients were instructed to hold specific postures for several minutes in a relaxed manner. To enhance its credibility, the patients were told that this new technique was called focused body awareness training.

In the interest of minimizing potential therapist bias, all 3 forms of intervention were conducted by the same therapeutic team. The time structure was the same in every group with one 60-min session weekly and a total of 4 sessions over a 4-week-period. The therapy sessions were videotaped and supervised by an independent, certified body psychotherapist, ensuring that the therapists adhered to the manual.

Statistical Analysis

Statistical analyses were performed using SPSS software v.15.0 (SPSS Inc., Chicago, 111., USA). The patients’ basic data were characterized by descriptive statistics. The \2 test and analysis of variance (ANOVA) were used for exploratory comparisons of social demographic data and baseline values between intervention groups. Analysis of covariance (ANCOVA) was employed for analyzing differences in long-term change of the primary outcome parameters FEV! and sRaw between the control and intervention groups. Baseline FEV! and sRaw were included as adjustment covariates in order to control for the impact of initial value levels and different distributions in treatment groups. All statistical analyses were performed two-sided at a 0.05 level of significance. To adjust for multiple group comparisons, Bonferroni correction of p values was conducted. To minimize the required correction level in multiple test procedures, statistical comparisons of intervention groups were solely conducted for the clinically relevant changes of FEV! and sRaw between measurement 2 and baseline. Statistical evaluation was carried out according to the intention- to-treat principle.

Source of Funding and Ethical Considerations

The study was planned and conducted in accordance with the Declaration of Helsinki and ethical laws pertaining to the medical professions. The design was approved by the ethics committee of the Statutory Physician Board. This study was conducted without any institutional influence and was not funded externally.


The study sample consisted of 64 patients. The average age of all participants was 42.9 ± 12.2 years, 64% were female. No statistically significant group differences were found between the 3 treatment groups regarding sociodemographics (age, sex, partnership and professional status) and baseline measurements at tO. Tables 1 and 2 show the explorative analysis of the outcome parameters FEVi and sRaw over the course of the trial. No statistically significant differences were found in mean baseline values. Controller or reliever medication was kept stable in all patients throughout the study. All p values reported here resulted from group comparisons of changes in FEVi and/or sRaw (from baseline to t2) with Cl as a comparison group, and they are adjusted for baseline FEVi and sRaw. respectively. Bonferroni correction of p values was used in order to apply an overall level of significance of 5% in the face of multiple testing.

Participation in the FR, GI, FR/GI and Cl groups led to increases in FEVi (% predicted) of 7.6 ± 13.2, 3.3 ± 9.8 and 8.3 ± 21.0, respectively, as compared to -1.8 ±

11.1 in the Cl group at the end of the therapy phase. At t2, the increases in FEVi were 6.9 ± 10.3 (p = 0.009) in the FR group, 4.4 ± 7.3 (p = 0.072) in the GI, 4.5 ± 8.1 (p =

0.066) in the FR/GI and -2.8 ± 9.2 in the Cl. While the most beneficial effect regarding FEVi at tl could be seen in patients who participated in FR/GI (followed by FR and GI alone, with a slight decrease in the Cl group), a long-lasting effect at t2 could only be observed in the FR group, with a noticeable decrease in the benefit among FR/GI and GI patients and a further decline in the Cl group. The analyses relating to changes in FEVi revealed that only the FR group reached a level of statistical significance when compared to the CL

Regarding changes in sRaw (% predicted), at the end of the therapy phase (tl) the decreases in sRaw were -162.2± 56.6 in the FR group, -207.4 ± 87.9 in the GI, -158.8 ± 93.3 in the FR/GI and -265.3 ± 111.5 in the Cl. After the 4-month follow-up (t2), the decreases in sRaw were -100.9 ± 100.5 (p = 0.018) in the FR group,-51.7 ± 165.2 (p = 0.228) in the GI, -72.4 ± 56.5 (p – 0.024) in the FR/ GI and 54.9 ± 86.2 in the CL At tl the clearest positive effect in sRaw could be seen with FR, followed by FR/GI and GI alone, whereas even a minor aggravation of airway resistance could be observed in the group receiving the Cl. This pattern was equal at t2, with a marginal decrease over time in the FR, FR/GI and GI groups and a slightly exacerbated aggravation in patients receiving the CL Group comparisons revealed that improvements in the FR and FR/GI groups, but not in the GI group, reached a 0.05 level of significance at t2.

FEVi to       tl       t2      
Patients 16 16 16 16 16 16 16 15 16 15 15 13
Mean 80.2 80.8 81.8 82.1 87.8 84.1 84.9 80.8 87.1 85.6 85.7 76.9
Standard deviation 16.1 17.7 13.5 16.2 14.5 15.4 12.2 19.0 14.5 15.5 14.0 19.7
Minimum 54 40 54 51 45 47 59 40 64 55 60 35
Maximum 105 105 106 113 101 104 100 107 112 109 106 108
25th percentile 65.8 72.0 70.0 72.0 84.0 71.3 77.5 70.0 74.0 73.0 73.0 71.5
50th percentile 81.0 82.0 84.0 80.5 91.5 86.0 88.0 83.0 88.5 89.0 89.0 78.0
75th percentile 96.5 94.8 90.0 92.8 97.5 97.0 92.5 97.0 99.5 98.0 96.0 92.0


Table 2. Measurement of sRaw (% predicted) over the course of the trial            
sRaw to       tl       t2      
Patients 16 16 16 16 16 16 16 15 16 15 15 13
Mean 286.3 252.9 249.6 226.1 162.2 207.4 158.8 258.6 185.4 203.1 172.9 267.6
Standard deviation 123.6 149.9 111.9 85.6 56.6 87.9 93.3 112.4 71.3 117.5 91.5 94.3
Minimum 104 100 93 89 70 110 62 119 101 85 57 127
Maximum 512 638 494 443 271 372 450 531 382 517 379 436
25th percentile 193.3 129.5 150.8 166.5 128.5 138.8 92.5 198.0 136.8 112.0 100.0 204.5
50th percentile 243.0 212.5 254.0 209.5 149.5 174.5 145.0 215.0 170.5 185.0 167.0 229.0
75th percentile 405.8 348.8 314.0 285.8 180.5 251.5 195.8 302.0 216.8 216.0 232.0 365.0



No side effects from the interventions were reported by any of the patients, nor did any patients report difficulties with implementing the body-psychotherapeutic technique of FR.


This study confirms earlier findings showing a positive effect from FR on lung parameters in patients with bronchial asthma [20,21].

In the present study we could demonstrate for the first time that effects achieved through FR remain stable at a 4-month follow-up. While improvements in the GI group initially seemed clinically relevant, they dissipated significantly more rapidly during the follow-up observation period. The effects of the combined FR/GI intervention fell in between those achieved in the other treatment groups. One could speculate that this differential effect is due to the active and individual FR approach, which allows the patient to further apply individually modified exercise elements to various everyday situations over the course of the therapy phase and catamnesis.

Little can be said regarding the therapeutic mechanisms involved as no relevant information is available yet regarding the psychophysiological effects of body-oriented therapy or imagery techniques in asthma. As all patients included in the present trial suffered from allergic, and therefore immunoglobulin-E (IgE)-mediated asthma, the effects may be partially due to a modulation of allergic antibody responses. Castes etal. [30] demonstrated that a 6-month psychosocial intervention program, including relaxation, GI and self-esteem training, led not only to a clinically significant improvement, but also to a significant reduction in the specific IgE responses against Ascaris lumbricoides (the most important allergen in the children in that study). Beyond potential modulation of the IgE-mediated inflammatory response, the observed benefits could be partially due to direct bronchodilatatory effects caused by FR or GI. The underlying mechanism might be a mitigation of vagal stimulation, as ‘rebalancing’ of the autonomic nervous system is assumed to be a possible mode of action of FR. Furthermore, the coincident reduction of asthma-associated anxiety presumably accounts for improvements of lung function, as shown by Lolak et al. [31].

The deterioration of sRaw and FEVi over time in the Cl group might be explained by a negative placebo expectation, a so-called nocebo effect. The nocebo hypothesis proposes that expectations of sickness and the affective states associated with such expectations can cause sickness in the expectant patient [32]. All participants were informed about the possibility of being randomized to any 1 of the 3 different treatment settings (FR, GI, FR/GI) or to the control group. As the Cl consisted only of an instruction of specific bodily postures without any specific movements or pictures for imagination, this intervention was potentially unmasked as a placebo intervention by a considerable proportion of participants. This might have led to a more pessimistic attitude towards the effectiveness of the intervention, and pessimists are known to be more likely to display a negative placebo (nocebo) expectation [33].

This study has several methodological limitations. First, the sample size was relatively small, consisting of only 16 patients in each treatment condition. Since the follow-up period of this trial was only 4 months, the dropout rate might have been lower than in trials with a longer follow-up period.

Given that the body-oriented therapy sessions were carefully designed and matched for possible nonspecific effects (e.g. attention given by the trainer or learning effects), it is difficult to claim that the outcome variables were influenced by nonspecific components of the FR treatment. Flowever, one possible influence may be that the study was not conducted in an entirely double-blind setting. However, creating completely double-blind conditions would not have been feasible, since it is impossible to prevent an experienced therapist knowing the kind of technique she/he is applying. However, a potential bias could arise from not checking the blinding of the patients. Although patients were not informed of which intervention they were allocated to, we did not ask the participants which intervention they believed themselves to be receiving. We also did not measure the degree of subjectively perceived relaxation in the FR group or the degree of vividness in the GI group. Although adherence of therapists to the treatment manual was controlled by continuous supervision, the adherence of patients to the given instructions was not measured in the trial.

The present trial focused on the influence of FR, GI and FR/GI on lung parameters in patients with allergic asthma and, therefore, no subjective outcome parameters were measured. It has to be considered that improved physiological outcomes are not necessarily reflected in an improved asthma-related quality of life [34].

Even considering the above limitations, the results of our study are promising because they show that body- oriented therapy, such as FR, can lead to a significant reduction in the objective parameters of pulmonary obstruction in asthma which, furthermore, was shown to be maintained over a period of 4 months. These treatment benefits are the result of only 4 sessions of complementary body-oriented therapy, which can, therefore, be regarded as a feasible intervention that is an efficient use of time. Nevertheless, future research should consider further the feasibility of body-oriented therapy in daily practice. Additional research is needed to ascertain if these results can be replicated, for how long the benefits last beyond a 4-month follow-up and if the regular complementary use of FR leads to clinically relevant improvements in the long-term course, including improvement in health-related quality of life.


The authors thank Silke Witt-Wulff, MD, for her support in the trial, and Mike Cronan, MA, for his linguistic revision of the manuscript.



Wellness Through a Creative Lens: Mediation and Visualization


School of Social Work, University of Northern British Columbia, Prince George, British

Columbia, Canada

This paper will examine literature focused on meditation and visualization, and their impact on mental and physical health. A key aspect of this examination includes understanding the defined meanings of both meditation and visualization. Specifically, the authors explore concepts of human consciousness, including mindfulness, quantum physics realities, and controlling mental imagery in neurology. These ideas provided the conceptual framework to study the effectiveness of meditation and visualization through an 8-week, program for female university students, experiencing stress, anxiety, and sadness. Initial qualitative data analysis employed a thematic analytic approach. Preliminary findings, such as increased self-a wareness and decreased anxiety, are discussed alongside existing literature that confirms the value of meditation and visualization approaches towards improved health.

KEYWORDS alternative methods, anxiety, consciousness, female university students, image rescripting, meditation, mental and physical health, mindfulness, sadness, self-awareness, spiritual, spirituality, stress, visualization, wellness

Depression and anxiety levels are increasing among North American female university students (Arehart-Treichel, 2002). In a 2009 College Counseling Center Directors Annual Survey, 94% of Directors reported an increase in the number of students with significant psychological problems (Barr, Rando, Krylowicz, & Winfield, 2010). According to one Canadian counseling center, over the past decade, students’ concerns have shifted from normal developmental issues, (75% in 2001; 35% in 2008) to more serious and severe mental health difficulties (25% in 2001; 65% in 2008; Beattie, 2010). This shift is especially alarming for female students, as women in general suffer disproportionately from depression and anxiety (Arehart-Treichel, 2002) and most campus counseling clientele are female (Beattie, 2010). Stress and anxiety are linked to feeling hopeless, while depression is linked to an inability to function and, in extreme cases, suicide (Barr, Rando, Krylowicz, & Winfield, 2010). Thus, useful intervention strategies to address anxiety and stress are essential to the success of female university students.

Meditation and visualization are ancient helping strategies well established in Hinduism, Judeo-Christianity, and traditional Chinese and Aboriginal medicine (Utay & Miller, 2006). These alternative interventions are resurfacing in contemporary North America and show promise for increased self-awareness, concentration, improved mood and sleep, reduced fear and anxiety, and pain reduction (Coholic, 2007; Dekro et ah, 2002; Newberg & Waldman, 2009). Research on treatment effectiveness is still in its infancy with many methodological questions. Existing literature does not bring together the wide range of work that investigates meditation and visualization as viable interventions. This article examines how and why these practices have been employed by other research practitioners. In addition, the authors draw upon their recent pilot project (2010) conducted with female students attending a northern university.


Meditation is an umbrella term used to describe a variety of mind-body therapeutic approaches. Ospina et al. (2007) identified five main types of meditation generally used in research studying the effects of meditation practices on health: Mantra meditation (comprised of Transcendental Meditation, Relaxation Response [RR], and Clinically Standardized Meditation), mindfulness meditation (consisting of Vipassana, Zen, Mindfulness-based Stress Reduction [MBSR], and Mindful-based cognitive therapy), Yoga, Tai Chi, and Qi Gong.

Ospina et al. (2007) investigated a working definition of meditation that focuses on the common threads between the various styles by surveying seven experts in the field. Meditation was characterized as a practice that uses a defined technique, involves logical relaxation, and engages a self- induced state or mode. The investigation found that meditation sometimes has an embedded religious/spiritual/philosophical context. In addition, it can involve a state of psychophysical relaxation, mental silence, or an altered state of consciousness. It was found that meditation often employs self-focus skills or the suspension of logical thoughts. It is important to note that the bulk of research conducted on meditation and health in the West focused on mindfulness meditation. As Hickey (2008) identified, “in 2007 and 2008, fully 80 percent of federally funded scientific research on meditation has focused on mindfulness” (p. 3). Defining mindfulness meditation is therefore important in understanding the relationship between meditation and health/wellness in our society.

Mindfulness meditation originated from the tradition of Theravada Buddhism and has evolved to fit Western ways of understanding. It is an exercise in concentration and insight where all stimuli are attended to equally, without any censorship or selection. In 1971, molecular biologist Kabat-Zinn developed MBSR in an attempt to make the mental and physical health benefits of meditation widely accessible in North America (Hickey, 2008; Starks, 2006).

The authors define meditation as the practice and process of concentration on the breath, sound vibration, and inner sensations of the body that arise out of de-identification with negative reactivity in the mind. The objective is to experience an underlying deeper aspect of self that many individuals do not perceive in day-to-day life. To attain this level of meditation, one must divorce him/herself from anxious thought. Consistent identification with anxiety and pain act as barriers to identification with peace and joy. Sen (2005) ascertains that yoga (a Sanskrit word meaning to add or unite) is not possible without beyoga (meaning to subtract or detach). Meditation begins with beyoga-the subtraction, detachment of all that interrupts an integrated and balanced mind-body-soul on the journey toward yoga uniting with the universe and that which one desires.


This study investigated the effects of a mantra meditation/creative visualization program for female university students experiencing stress, anxiety, and sadness entitled Realigning Self: A Pilot Project to Investigate the Effectiveness of the Sen System for Female University Students. The Sen System stems from the business sector and has, until this study, not yet been investigated as a therapeutic intervention. We are interested in exploring the Sen System’s effectiveness for female students in a northern university setting.

The meditation and visualization style utilized is based in an epistemology that human thought, with each image that accompanies it, has untapped creative potential that can be brought into awareness and used to create desired thoughts, feelings, and circumstance. While the notion that thought could create may appear as a childish fantasy, quantum physics theories provide evidence that it may not be such a preposterous view.

In Quantum Reality: Beyond the New Physics, Physicist, Nick Herbert (1985) highlights quantum reality theories that have particular significance for the authors’ current research concerned with connecting the physical world and human consciousness. The first quantum reality (called the Copenhagen Interpretation, Part 1) dictates that there is no deep reality. Bohr (1963), a quantum physicist, put forth the claim that the sensual world we experience “floats on a world that is not as real. Eveiyday phenomena are themselves built not out of phenomena but out of an utterly different kind of being” (as cited in Herbert, 1985, p. 16). It is a wholeness without space and time.

The third quantum reality acknowledges reality as an undivided wholeness. This premise rests on the idea that despite the obvious separations and boundaries humans perceive through the senses, the world in actuality is a seamless and inseparable whole (Capra, 1975). German light/matter physicist Walter Heitler claimed that quantum theory requires that the separation of the world into an objective outside reality and a self-conscious observer can no longer be maintained. “Subject and object have become inseparable from each other as the act of observation dissolves the boundaiy between observer and observed” (as cited in Herbert, 1985, p. 18).

The seventh quantum reality states that Consciousness creates reality. In 1932, von Neumann wrote a German quantum theory text, now translated in English and considered a seminal work among the quantum mechanics community (von Neumann, 1955). Von Neumann was the first to evidence that a physical object’s definable characteristics only exist if a conscious observer watches them. If an object is not under observation, it is represented as a “wave of probability” because there exists a wide range of possible mathematic values for its attributes–bundles of vibratory possibilities (Herbert, 1985). Alternatively, an observed object has definite particular physical properties such as being in one place at one time. According to Herbert (1985), during the act of measurement (observation), the mathematical description abruptly shifts-from a spread-out range of possible attributes to single actual attributes. This measurement-induced switch of descriptions is called the “quantum jump.” Ordinary realists assume that an object possesses definite attributes whether observed or not, however, according to Herbert, the majority of physicists reject this position of ordinary realism and treat objects as definite only when measured (observed). “The Copenhagen interpretation consists of two distinct parts: 1. There is no reality in the absence of observation; 2. Observation creates reality” (p. 17).

Observer-created reality physicists uphold the absolute existence of matter (electrons, photons) but the attributes of that matter, such as the position in space or momentum, are dynamic and do not exist until consciously observed. Physicist Heisenberg (1958) is famous for declaring, “An atom is not a thing.” He insisted that the unmeasured world is merely semireal, and achieves full reality status during the act of observation. The probability wave refers to a potential of reality, which stands between an idea and event (Herbert, 1985). The quantum world holds several unrealized inclinations for action. “During the magic measurement act, one quantum possibility is singled out, abandons its shadowy sisters, and surfaces in our ordinary world as an actual event” (p. 27). Because quantum theory applies to everything, not just to atoms, all objects without exception exist in this partially unreal state of “objective indefiniteness.” Heisenberg’s (1958) strange picture of the quantum world as half-real possibilities that are actualized only during a measurement act is considered by many physicists to be the most reasonable guess as to how the world deep down really operates. Herbert (1985) ascertains that, “At the logical core of our most materialistic science we meet not dead matter but our own lively selves” (p. 25).

What does quantum physics have to do with meditation and visualization? The nature of the quantum jump, when objects suddenly have measurable definable attributes in time and space, is the biggest mystery in quantum physics. Herbert (1985) believes the mystery of the quantum jump can be explained by human consciousness. He envisions a “holistic physics” where he conjoins Heisenberg’s picture of the “way the world really works” with a relationship between mind and matter. He proposes that mind is not a rare phenomenon associated with certain complex biological systems, but is everywhere, universal in nature. The quantum consciousness assumption asserts that consciousness is an integral part of the physical world, not an emergent property of special biological or computational systems. He ascertained that conscious human decision is what brings one vibratory possibility into reality above others. Sen’s (2005) view of human consciousness coincides with quantum physics findings that conscious observation creates reality. Sen argues that conscious visualization, which focuses on one’s own desires and goals, has the power to create feelings and events in the phenomenal world of one’s own life. If the above mentioned quantum realities do in fact exist, Sen offers an invaluable application: to imagine and work with the connected creative forces of the universe to bring into actuality one’s aspirations and desires. The authors attempted to explore these notions in conjunction with cognitive behavior theory for female university students struggling with their thoughts and consciousness as it relates to stress, anxiety, and sadness.

The theory of consciousness, put forth by Sen (2005) is tapped into through the assiduous practice of four proclamations: a. Pragyanaam Brahman/ Consciousness creates-, b. Tat Tavm Asi/Everything is Consciousness-, c. Ay am Atman Brahman/My Consciousness creates; and d. Aham Brahmas mi/1 create (p. 113). Participants were guided through these mantra meditations consecutively. As with mindfulness meditation, the goal is to gain control of mental processes but, with the added dimension of focusing on the radical idea of self as a creator.


Request for participants was extended via posted advertisements throughout a northern Canadian university and included the university counseling center, where the researchers conducted the study. Criteria to participate included: female university student over eighteen, self-identify as struggling with anxiety, stress, and/or sadness, and a desire to try a new meditation/visualization system. A convenience sample was applied for the purposes of this pilot study. Six students from the university were chosen to participate

The principal investigator, along with the coinvestigator, facilitated an 8-week psychoeducational group using a combination of book reading and reflection, breath and mantra meditation, creative visualization, reflective discussion, and writing techniques. Each participant was provided with the Ancient Secrets of Success book (Sen, 2005) and a journal to reflectively write. Participants were invited to read a section of the book in between sessions and practice meditation at home daily for five minutes.

This qualitative inquiry used an arts-based approach and employed thematic analysis as a methodology. Arts-based research concentrates on using artistic elements such as story, poetry, visual art, dance, and/or drama to enrich research relationships, data collection and the representation of the research (Leavy, 2009). Thematic analysis is a method used to identify, analyze, and report patterns (themes) across a data set in rich detail (Braun & Clarke, 2006). “A theme captures something important about the data in relation to the research question, and represents some level of patterned response or meaning within the data set” (p. 81). While most qualitative research is thematic in nature, thematic analysis can be used with different theoretical frameworks so it corresponds with arts-based research. The authors of this pilot study use thematic analysis as a realist method to report experiences, meanings, and the reality of participants (Braun & Clarke, 2006). This flexibility allowed for the identification of initial themes, which will be included in the larger comparative analysis between Phase I and Phase II.

Data collection methods included audio taped preindividual and postindividual interviews, semistructured group interviews, reflective journals, collage, and a focus group. The examination in this article focused on in-group semistructured interviews and preindividual and postindividual interviews from the pilot study.

In one initial theme, authors found that participants frequently referred to sleep disturbance as impacting their health and wellness. Sleep disturbance inhibited participants from optimal daily performance. After nightly meditations, participants reported a reduction in anxiety about falling asleep and an increase in undisturbed sleep time. One participant reported on her meditation experience, “so I did immediately have results because I normally go to sleep only after hours so I was like ‘wow, a neat little trick to fall asleep’ … so I’ve been using it as a sleeping tool.” Another participant, Jerusha explained, “I don’t sleep well and all my classes are at night so my whole schedule is shifted and … I meditated before sleep and I was out.” Participants reported increased levels of breath meditation were directly related to better sleep quality, which resulted in waking less agitated and a decreased level of perceived stress the next day. Caldwell, Harrison, Adams, Quin and Greenson (2010) also found that meditation improves sleep quality for college students.

In addition to breath meditation, the authors of this pilot study employed mantra meditation techniques of repeating a phrase or sound, while attempting to passively ignore distracting thoughts. The authors found that students reported reduced anxiety and stress. Jerusha noted the following:

I was oming in my bedroom and I could feel the anxiety trying to sneak in and so I just omed louder . . . and it worked, it just was like whoosh and I just smashed it all out of there and by the time I was done I felt really good.

Jerusha consciously fought to disconnect with anxiety and connect with a peace inducing sound. She felt the feeling she wanted to gain momentum and in her words “smashed” the anxiety away. This experience allowed her to gain a sense of mastery over her thoughts and feelings.

Dekro et al. (2002) studied the use of mind/body intervention techniques to address college students’ stress. They used mantra meditation to induce RR, which is a set of physiological changes including a decrease in heart rate, oxygen consumption, blood pressure, and rate of respiration. These researchers found that compared to a waiting list control group, students reported reductions in psychological distress, perception of stress, and anxiety.

The authors’ study offered support for silent breath and mantra meditation to decrease feelings of dysphoria. In particular, one female university student Nadine, who self-identified as living with a mental illness, explained:

With my week things had gotten pretty heavy for the first half of the week and I was kinda feeling like I was drowning and then there was a shift, I guess, on Wednesday where, I acknowledged that I needed to take time to take care of myself and I did start doing meditation and breathing exercises a lot more. Since then I’ve noticed quite an improvement in my mood .. .

Nadine changed her awareness when she began changing her focus from overwhelming worry to meditation and the breath. She became more self-aware, noticed a positive change in mood, and reported feeling more in control of her emotions. Broderick (2005) also studied the potential of mindfulness meditation as an intervention with dysphoric moods in undergraduate students and found that meditation decreased dysphoria and was more successful than distraction as an intervention.


Visualization is the practice of conscious control of mental imagery. “Mental imagery refers to the perceptual information that is brought to the mind from memory and imagination, rather than arising from activation of the sense organs” (Serruya & Grant, 2009, p. 792). Images have played a significant role in the development of humanity. Korn (1994) discusses evidence, based on the way children think, that images preceded language and are in fact, foundational to language. Jung (n.d.) states that “all the works of man have their origin in creative imagination . . . image is psyche and if the psyche creates reality, then what creates reality is the image” (as cited in Adams, 2004, p. 5). The image often plays a foundational role in not only the experience of reality, but in the creation of it.

Guided imagery is the form of visualization most studied and practiced in relation to health and wellness in recent years. Its aim is to employ the mind’s language to communicate with the self and make sense of both inner and outer experiences (Kabat-Zinn, 1990). Imagery rescripting is a specific type of guided imagery that has been used as an intervention in response to various clinical mental health disorders such as: posttraumatic stress disorder, depression, psychotic delusions, and social phobia (Holmes, Arntz, & Smucker, 2007; Serruya & Grant, 2009; Wild, Hackman, & Clark, 2008). Imagery rescripting uses visualization techniques to either transform a preexisting negative mental image to be understood as more benign, or to construct a completely new positive image in order to counteract the negative one. Holmes et al. (2007) argued that “imagery has a more powerful impact on positive emotion than verbal processing, and therefore cognitive behavioural techniques used to promote positive change should always employ imagery” (p. 300).

Combining the above concepts of guided imagery, and imagery rescripting, the authors of this paper utilized creative visualization as the guiding technique for their pilot study. Creative visualization is based on the notion that thought, when under the conscious control of an individual, can create reality (Herbert, 1985; Sen, 2005). Creative visualization invites an individual to imagine the unfolding of an aspect of one’s desired life, as if watching a movie in the mind. This conscious imagining, honed through meditation, can lead to the creation of desired circumstances because, according to Herbert (1985) the universe is alive and entirely interconnected at a quantum level. This interconnected atmosphere responds to human consciousness, where observation (through thought and decision) define matter into concrete attributes of an object. The authors pursue this research in an attempt to explore and document the theory that the images in peoples’ minds can manifest into actuality. Gawain (2008) proposed that form can result from thought. Brehrend (2006) and Sen (2005) likened the process of bringing one’s internal images into physical form to the process that initially brought matter into the universe. The power that macrocosmically creates in nature is the same power that humans can access within ourselves to microcosmically create in our own lives. They postulated that there is only one kind of power and suggested that working from the center of the mind, individuals can transfer the universal to the specific.

Sen (2005) suggests that human consciousness, defined as both creative power and awareness, is like a mirror that reflects any thought placed before it. First, the experience occurs in the thought or image and then, as a reflection, in the life of a person. Margolin (2009) asserts, “Humans not only have the seed of an unlimited capacity for creativity, but we are the seed … In the centre of my Self, I create. In the centre of creation, I am my Self” (p. 24-25).


Kabat-Zinn (2003) stated that meditation is a process to work with the mind’s own plasticity. Studies in neuroscience have shown that meditation practices have the capacity not only to stimulate areas of the brain temporarily, but also create permanent changes. Williams (2010) reported that long-term meditation practices change the structure of the brain and only 8 weeks of mindfulness practice brings about changes in the way emotion is processed.

Newberg and Waldman (2009) reported, be it secular or religious, when an individual intensely and consistently focuses spiritual values and goals, the blood flow to the frontal lobes and anterior cingulated is increased, which causes the activity in the emotional areas of the brain to decrease. Newberg and Waldman (2009) suggested the frontal lobes, which are activated in meditation and visualization, are what make us uniquely human and “hold the secret to making our dreams come true” (p. 32). They suggest the frontal lobes host the neurological roots of reasoning and communication, compassion, motivation, and creativity and imagination.

Halsband, Mueller, Hinterberger, and Strickner (2009) found an enhanced capacity for selective attention in their neurological study of intensive meditation training. This enhanced selective attention capacity is associated with an increased ability to focus on a desired goal while suppressing undesired ones, combined with an improvement in memory functions (Halsband et al., 2009; Newberg & Waldman, 2009). Goldin and Gross (2010) and Newberg and Waldman (2009) concluded that mindfulness meditation results in decreased activity in the amygdala which is a part of the brain that is associated with the generation of fear and anxiety. The pilot study highlighted in this paper found initial evidence for decreased fear and anxiety and enhanced selective attention capacity. Stephanie acknowledged, “I have the ability the mornings that I do meditate when I start becoming anxious about something and doing negative self-talk, I stop myself. I’m just more aware of what I’m doing. I can go on with my day.” Additionally, Randy expressed:

The time I do actually meditate in the morning, I have like a stability a confidence for that day. And when I do get into emotional muddles, it’s not a muddle, its just like “oh, it’s a passing thing” Its not like chaos that just stays there and takes control.

Williams, Teasdale, Segal, and Kabat-Zinn (2007) claimed that, “Our reactions to unhappiness can transform what might otherwise be a brief, passing sadness into persistent dissatisfaction and unhappiness. The problem is never the emotion itself that comes to protect us but the mind’s reaction to the emotion,” (p. 34). Randy’s report clearly points to a radical shift in thinking from an inner conflict that commonly results in both inner and outer chaos, and perceived lack of control to “a passing thing” in her day. Stephanie discussed the heightened awareness to “go on with [her] day.” Both participants attribute their “ability” and “confidence” to beginning their days with meditation. This practice of altering one’s reaction to unwanted emotion can transform one’s relationship with anxiety, stress, and/or sadness from an overwhelming persistent experience to a fleeting feeling.

Visualization induces a variety of neurological responses. Doidge (2007) states:

One reason we can change our brains simply by imagining is that, from a neuroscientific point of view, imagining an act and doing it are not as different as they sound. When people close their eyes and visualize a simple object, such as the letter a, the primary visual cortex lights up, just as it would if the subjects were actually looking at tire letter a. Brain scans show that in action and imagination many of tire same parts of the brain are activated, (p. 203-204)

Wild et al. (2008), in response to participants’ very quick reaction rate to imagery rescripting intervention, proposed that imagery works so immediately, because it is like having a concrete experience. Schnitzler, Salenius, Salmelin, Jousmaki, and Hari (1997) found that the imagery of movement uses the same cortical circuitry and results in the same cortical activation as actual movement (as cited in Wild et al., 2008).


When examining the effectiveness of meditation and visualization on mental health and wellness, it is important to examine the connection between the imagination and body. As legendary American author Mark Twain pointed out, “the power over which a man’s imagination has over his body to heal it or make it sick is a force which none of us are born without. The first man had it, the last one will possess it” (as cited in Pettus, 2006, p. 87). Fernros, Furhoff, and Wandell (2008), in their study of mind-body therapies, found that emotional health was correlated with self-assessed general health, which was correlated with health prognosis.

One cause of psychopathology or mental illness is thought to be the overuse of the “doing mode” of the mind, which is a symbolic process employed by the brain in order to be effective at pursuing tasks, solving problems, and encountering situations (Williams, 2010). Mindfulness meditation is not about getting rid of these processes or just simply clearing the mind, but rather:

cultivating an alternative (“being”) mode . . . that teach(es) people how to pay open-hearted attention to objects in the exterior and interior world as they unfold moment by moment. Attention is paid not only to the objects themselves, but to our reactions to them, particularly reactions of wanting positive states to last, negative states to end and neutral states to be less boring. (Williams, 2010, p. 2)

Mindfulness is theorized to be the active agent in cognitive-behavior therapy, the treatment commonly used for clients with anxiety and depression (Broderick, 2005; Teasdale et al., 2000). Pettus (2006) emphasized that although there are differing “depths” of mindfulness which elicit varying neurobiological effects, there are no right or wrong ways to practice visualization or meditation. He proposed that “any attempt to examine your life more fully, regardless of how it is achieved, has healing value” (p. 276).


Research on mindfulness and meditation is expanding in clinical science, social-cognitive-affective neuroscience, and clinical interventions for anxiety and depression disorders (Goldin & Gross, 2010). Starks (2006) conducted research in Northern Canada documenting the effects MBSR on the mental health of adults in the Yukon. She found that MSBR employed as a practice of self-care, resulted in increased self and body awareness and increased ability of clients to cope with stressful situations.

Yunesian, Aslani, Vash, and Yazdi (2008) did a study on the use of transcendental meditation (TM) as a means of enhancing the mental health of an adult Muslim population in Iran. Results from a self-administered General Health Questionnaire (GHQ) (a tool used in Iran to measure anxiety, depression, somatization, and social dysfunction) were compared before and after a 12-week TM course. The overall score of the GHQ test improved significantly enough, after the intervention, to move the average from above to below the cut-off level for clinically discernable mental health distress. The study also found evidence for decreased anxiety in one participant with a self-disclosed diagnosed mental illness. Nadine stated, “My anxiety levels have gone down and [I’m] just starting to feel like I am breathing as opposed to having been holding my breath for a long time . . Although further research is required before claims can be substantiated, breath meditation seems to have an inverse relationship to anxiety.


Research on health and meditation emphasizes the need for a multidimensional approach to health and wellness and places such practices as meditation and visualization as elements of larger wellness programs that deal with a variety of elements including physical, emotional, behavioral, and spiritual. Chan, Chan and Ng (2006) discussed:

to heal bodily symptoms, Western medicine focuses on how to combat the disease, kill tire bacteria, and cut out the defective body parts. In its traditional biomedical model, illness (including mental illness) is considered to be an evil object, a threat to life, and an enemy against which patients must fight in order to survive, (p. 20)

Although this style of medical treatment has its merits, it often neglects to take into consideration the mind body connection and the capacity of the mind for healing. Chan et al. (2006) found that individuals that suffer physical and emotional health issues, whether the issue is cancer or divorce, favor approaches that emphasize the body-mind-spirit connection as opposed to the Western medical model. They concluded that social work interventions to trauma, such as meditation, should take the approach of strengthening the client’s whole system. Restoring balance between the physical, psychosocial, and spiritual dimensions of an individual is the ultimate goal.

The authors found that most of the female university students in their study suffered from anxiety concerning exams and feeling overwhelmed with coursework. Many chose to explore the effects of creative visualization on the psychological distress they were experiencing at university. Using creative visualization as a strategy, participants reported a newly found awareness in their ability to curb their performance-anxiety concerning university exams and coursework. Randy shared her experience of gaining control over her performance anxiety stating:

A couple of days before the midterm exam and then half an hour before the midterm, I’ll start hyperventilating and start crying and just all of these bodily tilings that I don’t have any control over but its obvious in my mind first. When I quickly read over what I had written about what I want to feel in my exam, I felt like “wow,” I knew much more than I thought I knew. I think the fact that I was reading through and imagining that as opposed to being alone with my head actually just helped me not start crying and hyperventilating.

Implementing creative visualization altered not only this participant’s experience of the exam process but the outcome as well, as her assessment was markedly improved. Creative visualization is an approach that focused on and integrated the physical, psychosocial, and spiritual dimensions of these students’ lives.

Meditation can bring an increased wisdom and compassion to the relationships between social workers and clients. The authors, also social workers, found that working with participants in a spiritually oriented manner encouraged personal self-care, awareness, and brought about deepened relationships within the research group. They correlated their ability to be aware and present with participants as directly related to the group meditation practice. Brenner (2009) also pointed out that awareness permitted social workers to avoid distinctions between self and other, which thereby allowed workers to be fully present with the client.


Trackhtenberg (2008) did a critical review on the research on visualization and the immune system and found that guided imagery interventions can reduce stress, which allows the immune system to most effectively function. Donaldson (2000) showed that negative images can produce physiological outcomes, such as subtle muscle movements, changes in respiration and blood flow, and changes in the brain waves, that move an individual towards the “fight or flight” response that leads to negative stress stimulus. Intentional visualization of positive images, on the other hand, can counteract this response and enhance an individual’s immune system.


Psychosocial and psychiatric health is particularly difficult to maintain during periods of physical illness. Mental health and quality of life are extremely important for patients suffering from chronic illness such as cancer. Walker et al. (1999) studied the effects of relaxation combined with the visualization of defenses destroying tumor cells on women with newly diagnosed, locally advanced breast cancer. They found that women who participated in relaxation and visualization were more relaxed and easy going, had few psychological symptoms, and had a self-reported higher quality of life.

Visualization techniques have been shown to offer relief with mental health and psychiatric disorders. Russinova, Wewiorski, and Cash (2002), in a purposive sample of adults in the United States, found that alternative health care practices, especially guided imagery, assisted patients to cope with serious mental illness. Serruya and Grant (2009) conducted a study on the use of mental imagery as an intervention to treat clients who are suffering from delusions as a result of a psychotic disorder. They identified, “because negative images stir up more affect than verbalizations of the same content, patients tend to find the images particularly distressing, making the images the central players in psychopathology of anxiety” (p. 792). Serruya and Grant (2009) conducted a case study with a 25-year-old individual who was suffering from paranoid delusions. After a number of guided imagery sessions, the client was able to engender new ways of thinking and enabled him to return to school. One participant of the authors’ study self-identified with a mental illness. She normally found group work and social gatherings extremely anxiety provoking (claustrophobic tendencies) and outside her ability to function. For this reason, she was uncertain whether she could complete the 8-week pilot study. After 1 week of participating in guided group meditation, she noted the following: “I like how the room expanded. I felt the walls vanish” and subsequently completed all 8 weeks of the project.

Wild et al. (2008) studied the use of image rescripting interventions in the treatment of social phobia. They hypothesized that, because images prevalent in those who suffer from social phobia are linked to memories of unpleasant social experiences, the use of image rescripting techniques would reduce the participant’s negative self-beliefs and anxiety surrounding social situations and decrease distress caused by recurrent images. They found that imagery rescripting did, in fact, reduce anxiety surrounding feared social situations for those with social phobia disorder. They also found that although there were no changes in the frequency of recurrent images of memories, there was a decrease in the distress associated with them.


Increased Self Awareness

Adams (2004) argued that the development and interpretation of imagination could lead to an increase in consciousness. Below is an example from the authors’ study in which Joy revealed her increased self-awareness after one month of visioning:

Whenever I got upset at somebody or angry or sad in general I’d sort of realize that I had envisioned myself being sad and getting sadder . . . When I realized that… I tried to change the vision, to see myself as being happier, forgiving this person or making up and it helped a lot actually. Once I realized that I wanted to be happy I tried to work towards that instead of working towards being sad.

Joy connected her moods and relational interactions with her thoughts by contemplating and visualizing how she wanted to feel. She realized that thoughts (images) of herself as sad or angry preceded events that played out in her life and resembled previously held images in her mind. By linking past thought-images with future feelings and life events, Joy recognized she was unintentionally thinking and thus, observing/creating herself as sad and angry. This recognition motivated her to change the internal images she had of herself from unintentional negative images to intentional positive images. Through the continuous practice of intentional desired self-observation, Joy was transforming herself from a woman controlled by an unwanted self- image to one who can consciously choose who she wants to be. Creative visualization holds promise that everything that can be imagined is possible.

Improved Self-Image

Recall that Doidge (2007) states, . . we can change our brains simply by imagining . . . Brain scans show that in action and imagination many of the same parts of the brain are activated” (pp. 203-204). Wilson and Dunn (2004) discuss that although the pursuit of self-knowledge is not always productive; the development of self-awareness can, in fact, help people to construct positive narratives and escape focusing on an incomplete understanding of self that facilitates the perpetuation of negative self-image. The awareness raised by the practice of meditation is one of nonjudgment, which can be helpful in taking people out of negative mind traps.

This may explain the research findings with one participant who experienced an improved sense of body image through visualization. Stephanie provides an example of improved self-image:

What worked about it is .. . my vision is that I am healthy and that I like my body, that’s my vision. I live in a house that … all the closet doors are mirrored so every time I walk by it’s like ‘ohhhh, here I am again.’ My thighs are too big . . . It’s just everywhere all the time, so every time I went; I was like “No. I am healthy, I look beautiful today” and just saying that was … a real switch in my brain … I felt better about myself all week.

Stephanie began the process of altering the way she perceives her physical self from a woman who does not fit within a normal range of feminine attractiveness to one who is healthy and beautiful. This perceptual shift, which required Stephanie to not believe in the messages her eyes send, ignited a change in behavior, including exercise and a healthy diet. Newberg and Waldman (2009) stated, “when you intensely meditate on a specific goal over an extended period of time, your brain begins to relate to your idea as if it were an actual object in the world by increasing activity in the thalamus, part of the reality making process of the brain” (p. 34).

In the preinterviews of the pilot study, participants expressed a common theme of continuous negative self-talk and self-blame resulting in recurrent feeling of anxiety and stress. For instance, three participants reported, “I blame myself for all of the silly mistakes,” “Yes, I am critical of myself … I can be quite self-conscious,” and “I have the ability sometimes but not the awareness immediately to stop whatever the negative thoughts are.” Participants were guided in visualization activities focused on imagery rescripting by recalling unwanted thoughts and/or feelings about the self or an event. Once participants recalled their unwanted thought-images, they were guided to replace those images with intentional positive images they desire regardless of their belief in the possibility of those realities. Practice focused on the desired image/feeling resulted in mood and thought pattern changes for most participants. For instance, postinterview data analysis revealed the same participants quoted above experienced a reduction in negative self-talk, self-blame and resulting anxiety. In corresponding order to the above quotes, the first participant noted, “that visualization helps me to replace worries, and that I think that this is very important.” The second participant noted, “I would say I am less self-critical than I would have been before the group started.” The third participant noted, “I’m more aware of when I’m doing negative talk to myself, I don’t think I realize how much I did that to myself.” In addition to increased self-awareness, decreased anxiety and sadness, induced positive moods, and the ability to intentionally think-observe desired aspects of life, participants in this (2010) pilot study experienced an increased ability to sleep, reduction in performance anxiety in school, improved marks, and improved communication and outcomes in relationships. Six months after Phase I, the authors are about to commence Phase II of the pilot project with four of the original six participants. Follow up research will be facilitated with a comparative comprehensive analysis over the two phases of the project, including the artistic data collected, to investigate the effects of the Sen meditation/creative visualization system with female university students experiencing stress, anxiety, and sadness.


The literature revealed that various meditation and visualization practices have been successful in increasing immune functioning, inducing the Relaxation Response, improving sleep, and decreasing anxiety and stress among female university students. Research has also aided other populations to cope with dysphoric mood, mental illness, and stress related to physical illness. Further research, with more accurate data collection strategies and longitudinal designs, are required. The authors’ study highlighted the need to continue further research on the Sen system as an intervention for stress, anxiety, and sadness, and will incorporate a longitudinal design. The authors’ (2010) work deepens the literature, suggesting the power of the mind can directly set in motion a version of reality a client strives for, while diametrically opposing that which they fear. Alternatives practices to health and wellness, such as the Sen System, show promise to improve human quality of life.


Cognitive Behavioural Treatment of Social Phobia Bridging the Gap between Research and Practice


zur Erlangung des Doktorgrades der Naturwissenschaften (Dr. rer. nat.)

dem Fachbereich Psychologie der Philipps-Universitat Marburg vorgelegt von

Tania Marie Lincoln

aus Marburg

Marburg/Lahn, Februar 2003

Erstgutachter: Prof. Dr. Winfried Rief

Zweitgutachter: Prof. Dr. Gert Sommer

Tag der mundlichen Prufung: 03. Juli 2003



Preliminary Comments IV

Acknowledgements V

  1. Theoretical Background 1

1.1. Social Phobia: Concept and Classification 1

1.2. Differential Diagnosis 3

1.2.1. Panic Disorder 3

1.2.2. Generalized Anxiety Disorder 4

1.2.3. Depression 4

1.2.4. Avoidant Personality Disorder 5

1.3. Epidemiology 5

1.3.1. Prevalence 5

1.3.2. Developmental Aspects 6

1.3.3. Risk Factors and Socio-demographic Correlates 6

1.3.4. Comorbidity 7

1.4. The Biological Basis of Social Phobia 7

1.4.1. Genetic Factors 7

1.4.2. Neurobiological Factors 8

1.4.3. Evolutionary Factors 9

1.5. Cognitive and Behavioural Models of Explanation 10

1.5.1. Early Theories 10

1.5.2. Integrative Models 13

1.6. Treatment of Social Phobia 15

1.6.1. Cognitive Behavioural Interventions 15

1.6.2. Pharmacological Treatment 18

1.6.3. Present State of Treatment Research 19

  1. Purpose and Summary of the Studies 24

2.1. Purpose and Summary of STUDY I 24

2.2. Purpose and Summary of STUDY II 26

2.3. Purpose and Summary of STUDY III 28

  1. STUDY I 30

3.1. Introduction 30

3.2. Method 32

3.2.1. Retrieval of Studies 32

3.2.2. Study Sample 32

3.2.3. Data Analysis Plan 33

3.3. Results 35

3.3.1. Comparison of Studies According to Sample and Laboratory Characteristics 35

3.3.2. Effects of Accumulative Research Characteristics 35

3.3.3. Effect of Sample Selection on the Standard Deviations 36

3.4. Discussion 36

  1. STUDYII 42

4.1. Introduction 42

4.2. Method 45

4.2.1. Setting 45

4.2.2. Participants 45

4.2.3. Treatment 46

4.2.4. Therapists 48

4.2.5. Measures 49

4.3. Results 51

4.3.1. Comparison of Treatment Completers and Dropouts 51

4.3.2. Preliminary Analyses 52

4.3.3. Treatment Outcome and Consumer Satisfaction 53

4.3.4. Intra Group Effect Sizes, Reliable Change, and Clinical Significance 53

4.3.5. Effects of Sample Selection 55

4.4. Discussion 55

  1. STUDYIII 65

5.1. Introduction 65

5.2. Method 68

5.2.1. Setting 68

5.2.2. Treatment 68

5.2.3. Participants 69

5.2.4. Measures 70

5.2.5. Analysis 72

5.3. Results 73

5.3.1. Preliminary Analyses 73

5.3.2. Predictors of Treatment Refusal and Dropout 74

5.3.3. Predictors of Treatment Outcome 74

5.3.4. Predictors of Dterioration after Treatment 75

5.4. Discussion 75

  1. Summary 88

6.1. Summary 88

6.2. Zusammenfassung 90

  1. References 93


Preliminary Comments

I would like to use to opportunity to comment on a couple of formal aspects of this work.

The first chapter contains an introduction to the theoretical background of the concept, epidemiological aspects, a etiology and treatment of social phobia. Because the emphasize of this work is placed on the evaluation of treatment for social phobia, the theoretical part also focuses largely on the description of a etiological models and the treatment concepts derived from them as well as on the current state of treatment research.

The second chapter gives a short introduction to the intention, methods and results of the three conducted studies. The chapters 3, 4, and 5 are the original versions of the publication-based manuscripts. The second study „Effectiveness of an Empirically Supported Treatment for Social Phobia in the Field” is now in press in the Journal “Behaviour Research and Therapy”.

In the appendix the interested reader will find a table of the studies analysed in STUDY I, a more detailed description of the complete patient sample underlying STUDY Ш, a detailed description of the therapeutic procedure as well as a copy of all assessment measures and formulas used.

Because the publication based manuscripts were submitted in English language it seemed appropriate, for reasons of standardization, to write the complete doctoral dissertation in English. The sole exceptions to this are the German summary as well as the German original questionnaires and formulas depicted in the appendix.



In the course of my research for this doctoral dissertation, I was supported by a number of people to whom I would like to express my gratitude.

First of all I would like to thank Prof. Dr. Winfried Rief and Prof. Dr. Kurt Hahlweg for their constructive guidance and practical assistance and particularly for the suggestion to conduct this work in an accumulated, publication based manner. The structure of combining individual units written for a wider leadership into a continuous whole has had a very motivating effect.

I am especially obliged to the management board of the Christoph-Domier-Foundation for Clinical Psychology and the director of the institute in Marburg, Dr. Monika Frank. The Christoph-Dornier-Foundation has supported this work in many ways. I would like to emphasize the financial resources they made available as well as the free access to all data laboriously compiled by colleagues and assistants. Finally, I have greatly benefited from the opportunity to carry out treatment for social phobic patients and thus develop a personal understanding of the disorder that is the underlying basis of this work. I am particularly indebted to all trainees in the Christoph-Dornier-Foundation, who so patiently supported me in gathering literature and entering data into the computer. I am also grateful for all the personal support and informed advice I received from my colleagues Vera Martin and Thomas Lang. Finally, I would like to mention by name the colleagues who founded the group “Promoventen unterstotzen Promoventen [doctorands support doctorands],” – Dr. Anne Wietasch, Dr. Markus Funke, Dr. Dorte Zickenheiner, Dr. Torsten Eckardt, Christoph Frenken, Thomas Reininger, and Andres Buchenau.

I also appreciate the professional work my mother, Margaret Lincoln, put into correcting my English manuscripts.

Above all, my deep gratitude goes to Peter Leufgen for his encouragement and support in difficult stages and his practical help in matters of everyday life that was so necessary to complete this work.


  1. Theoretical Background

1.1. Social Phobia: Concept and Classification

Anxiety in social situations is neither uncommon nor particularly dysfunctional. About 80% of the general population report having suffered from shyness at some point in their life and about 40% even describe themselves as shy persons (Pilkonis & Zimbardo, 1979). Many well-known artists suffer from stage fright. Pop-idol Robbie Williams even admitted being so shy that he was on medication during the TV-show „Wetten Dass…“. Presumably all of us have experienced a certain degree of exam nerves or feeling nervous in expectation of an important date. However, while low levels of anxiety or nervousness can even boost performance, higher levels are extremely interfering.

The term social phobia is used in the case of marked and persistent fear in one or more social or performance situations, in which the person is exposed to unfamiliar people or to possible scrutiny by others. Individuals with social phobia fear to act in a way that will be embarrassing or degrading and thus be subject to negative evaluation. In many cases an individual may fear that other people could notice physical symptoms of anxiety and be scornful or humiliating towards them (Criterion A, Diagnostic and Statistical Manual of Mental Disorders, DSM-IV, American Psychiatric Association, APA, 1994). Even the expectation of being confronted with the fear situation provokes anxiety, which may be accompanied by a series of somatic anxiety symptoms (Criterion B). Even though the fear is recognized as excessive (Criterion C) a socially phobic individual will try and avoid the feared situations whenever possible. When this is not possible he or she endures them with intense anxiety (Criterion D). The social fear causes marked distress and can interfere significantly with occupational functioning, social activities or relationships (Criterion E). In individuals younger than 18 years the symptoms must have persisted for at least six months (Criterion F). The fear and avoidance is not due to direct physiological effects of a substance or a general medical condition and is not better accounted for by another mental disorder (Criterion G).

Social phobic fear is associated with performance situations, such as public speaking, and everyday social interactions, such as attending a party or speaking to an employer. The fear of public speaking has been found to be the most typical fear, followed by situations such as entering a room, which is already occupied by others, being addressed in front of others and meetings with strangers (Faravelli et al., 2000; Furmark, Tillfors, Stattin, Ekselius, & Fredrikson, 2000; Stein, Torgrud, & Walker, 2000). Typical worries involve being embarrassed or judged anxious, weak, crazy, inadequate or stupid. People diagnosed with social phobia are also often hypersensitive to criticism and negative evaluation and find it difficult to be assertive. Additionally, many social phobics suffer from feelings of inferiority (Clark & Wells, 1995). The anxiety provoked in a social situation is often accompanied by a series of physical anxiety symptoms, which are likely to be visible, such as blushing, sweating, or trembling. In severe cases these symptoms may meet criteria for a panic attack (DSM-IV, 1994, fourth edition).

The criteria for social phobia have evolved considerably over the years. The first definition of social phobia in DSM-III (American Psychiatric Association, 1980, third edition) classified it as a simple phobia limited to the experience in a situation in which the individual is exposed to possible scrutiny by others. In the accompanying text, it was suggested that “generally an individual has only one social phobia” (p. 227). Individuals who experience anxiety in a broad range of social situations were considered as suffering from avoidant personality disorder. In DSM-III-R (American Psychiatric Association, 1987, third edition revised) the definition of the concept of social phobia broadened and included individuals with fears in a range of social situations. Customary classification systems, the tenth edition of the ICD-10 Classification of Mental and Behavioural Disorders (World Health Organization, 1992) and DSM-IV (1994, fourth edition) have moved closer together in the course of their development and now use relatively similar criteria to describe the degree of distress experienced by people suffering from social phobia. DSM-IV describes more generally an immediate anxiety reaction (criterion B) whereas ICD-10 emphasizes specific physical reactions (blushing or trembling, nausea or urge to urinate). The DSM-III-R and DSM-V also offer the possibility of specifying a generalized subtype if the fear involves almost all social situations as opposed to a nongeneralized subtype, when the fear only involves one or a few social situations. Individuals with generalized social phobia and non-generalized social phobia have been significantly differentiated according to a number of demographic and clinical features. Individuals with generalized social phobia have been found to be younger, less educated and more likely to be unemployed (Heimberg, Hope, Dodge, & Becker, 1990). Also, generalized social phobics endorse higher levels of depression, social anxiety, avoidance and fear of negative evaluation on a row of self-report measures (Brown, Heimberg, & Juster, 1995; Turner, Beidel, & Townsley, 1992), are more often single, have an earlier age at onset and higher rates of alcoholism (Mannuzza et al ., 1995). In spite of these differences the subtyping scheme is a subject of controversial debate. The main controversy seems to focus on the question of whether the subtypes differ qualitatively or only quantitatively (Boone et al., 1999; Chambless, Tran, & Glass, 1997; Holt, Heimberg, & Hope, 1992; Heimberg, Hope, et al., 1990; Stein, Torgrud, & Walker, 2000). It is also unclear how the criteria fear in “most situations” can be operationalized. In answer to the introduction of the subtyping scheme in DSM-III-R some researchers have suggested other subtyping schemes (Eng, Heimberg, Coles, Schneier, & Liebowitz, 2000; Heimberg, Holt, Schneier, Spitzer, & Liebowitz, 1993). However, the dichotomous subtyping system was retained in DSM-IV. The lack of an operational definition for the subtypes allows for a variety of interpretations, thereby hindering comparisons across studies (Hazen & Stein, 1995). As a consequence, STUDY II adopted an attempt used in a study by Gerlach, Wilhelm, Gruber, and Roth (2001) to categorize subtypes according to the number of feared situations listed in a reliable and valid structured clinical interview for DSM-III-R (Diagnostisches Interview bei Psychischen Storungen [Diagnostic Interview for Psychological Disorders], Margraf, Schneider, & Ehlers, 1991).

1.2. Differential Diagnosis

The similarity of symptoms within the anxiety and mood disorders may provide a difficulty in arriving at a reliable diagnosis of social phobia. The anxiety disorders share some overlapping features (e g. fear and avoidance), whereas social phobia and depression have the aspect of social withdrawal in common. These similarities make a thorough diagnostic assessment of social phobia in terms of a diagnostic interview (see STUDY U and STUDY Ш) absolutely necessary, if one is to arrive at a reliable diagnosis.

1.2.1. Panic Disorder

Even though individuals suffering from panic disorder with agoraphobia may avoid social and performance situations, they do so for fear of having a panic attack and being unable to obtain help in that situation, and not specifically for fear of negative evaluation, humiliation and embarrassment (Ball, Otto, Pollack, Uccello, & Rosenbaum, 1995; Mannuzza, Fyer, Liebowitz, & Klein, 1990). Hazen and Stein (1995) point out that although both groups may suffer from panic attacks, in social phobia these attacks are situation bound and occur when entering or anticipating a social situation. In contrast, for the diagnosis of panic disorder there must be a history of at least one unexpected attack and subsequent attacks which do not occur exclusively in social situations. Also, in social phobia the content of automatic thoughts revolves around fear of embarrassment and negative evaluation, whereas in panic disorder, the thoughts revolve around catastrophic consequences, such as heart attack, death or loss of control. Both social phobics as well as individuals diagnosed with panic disorder suffer from somatic anxiety symptoms. However, and not surprisingly, it has been found that social phobics are more likely to endorse symptoms that can be observed by others, such as blushing, muscle twitching, dry mouth, trembling or sweating in comparison to individuals with panic disorder, who tend to experience dizziness, palpitations, chest pain, breathing problems, feeling faint and numbness (Amies, Gelder, & Shaw, 1983; Gorman & Gorman, 1987; Reich, Noyes, & Yates, 1988; Hazen & Stein, 1995).

1.2.2. Generalized Anxiety Disorder

Generalized anxiety disorder and social phobia share some clinical features that complicate differential diagnosis. Mennin, Heimberg, and MacAndrew (2000) found 24% of their large sample of social phobic patients to receive an additional diagnosis of generalized anxiety disorder. Rapee, Sanderson, and Barlow (1988) discovered that although social anxiety is also common among people diagnosed with generalized anxiety disorder, the impairment associated with it is much higher for social phobia. The number of social situations that produce fear was considerably greater than the one reported by subjects with any other anxiety disorder and social phobics spend more time worrying about social situations. Turk, Fresco, and Heimberg (1999) point out that the uncontrollable worry that individuals with generalized anxiety disorder experience is not exclusive to social situations. They emphasize that a hallmark feature of generalized anxiety disorder is the heightened focus on possible catastrophic consequences across several domains of life. Also, like with panic disorder somatic symptoms tend to differ, with individuals with generalized anxiety disorder reporting more frequent occurrences of headaches and fear of dying (Reich et al., 1988; Cameron, Thyer, Feckner, Nesse, & Curtis, 1986). It may be questioned though, whether these distinctions are sufficient to reliably differentiate social phobia from generalized anxiety disorder in a clinical setting (Turk et al, 1999).

1.2.3. Depression

To differentiate social phobia from depression, a clinician must be able to determine whether social withdrawal occurs because of low energy or because of fear of negative evaluation (Turk et al., 1999). Another common feature is the hypersensitivity to rejection or criticism and a negative self-concept, which has lead Brunello et al. (2000) to speculate that social phobia and depression may arise from a common vulnerability. They also see support for this idea in the fact that both disorders respond well to monoamine oxidase inhibitors. Clark and Wells (1995) argue that the negative self-schemata of depressed patients are relatively stable and persist throughout depressive episodes. In contrast, social phobics can have a positive view of themselves when they are alone or in situations they do not find threatening.

1.2.4. Avoidant Personality Disorder

The new criteria for the classification of a generalized subtype have brought about some confusion concerning the distinction to avoidant personality disorder. Apart from the fact that the criteria for avoidant personality disorder have become more similar to those of social phobia, the rules in DSM-III-R (1987, third edition revised) were changed so that both diagnoses can be given to the same person. Turk et al. (1999) raise the question of whether the two diagnostic entities represent distinct disorders or the same disorder differing only in degree. Most researchers have come to the conclusion that the distinction tends to be a quantitative one and that the co-occurrence of generalized social phobia and avoidant personality disorder describes individuals with the most severe social phobias and the poorest global and social functioning (Heimberg et al., 1993; Herbert, Hope, & Bellack, 1992; Holt, Heimberg, & Hope, 1992; Feske, Perry, Chambless, Renneberg, & Goldstein, 1996; Rettew, 2000; Turner et al., 1992).

1.3. Epidemiology

1.3.1. Prevalence

Estimates of prevalence of social phobia fluctuate considerably. One reason for this can be seen in different interpretations of the criterion of interference with a person’s life in DSM-IV (1994, fourth edition) and ICD-10 (World Health Organization, 1992). Stein, Walker, and Forde (1994) investigated the effects of different thresholds in the categorization and found fluctuations in rates of prevalence between 1.9% and 18.7%. Further reasons can be assumed in the differences in the classification criteria between DSM-Ш and DSM-III-R as well as in non-uniform interview systems. The establishment of DSM-III-R and DSM-IV criteria that do not differ much and the development of widespread interview systems based on these criteria has led to more uniform as well as higher rates of prevalence. In the National Comorbidity Survey (Magee, Eaton, Wittchen, McGonagle, & Kessler, 1996) in the USA a rate of prevalence of 13.3% was found. In Basel, Switzerland, this rate was 16.1% (Wacker, Mullejans, Klein, & Battegay, 1992), in Sweden 15.6% (Furmark et al., 1999), in France 7.3% (Pelissolo, Andre, Moutard-Martin, Wittchen, & Lepine, 2000), in Italy 6.6% (Faravelli et al., 2000) and in a German sample of men and women aged between 14-24 in Munich 4.9% and 9.5% respectively (Wittchen, Stein, & Kessler, 1999), indicating social phobia to be one of the most frequent chronic psychological disorders.

1.3.2. Developmental Aspects

Social phobia most often has its onset during adolescence, follows a chronic course and tends not to remit spontaneously (Burke, Burke, Regier, & Rae, 1990; Hazen & Stein, 1995; Schneier, Johnson, Hornig, Liebowitz, & Weissman, 1992; Wittchen, Stein, et al., 1999). Research of developmental aspects of social phobia is still in its early stage. Based on the research reviewed by Hudson and Rapee (2000) it seems that the ability to experience self- consciousness or to anticipate negative evaluation is unlikely to occur below the age of eight years. The beginning of adolescence accompanies the onset of increased self-consciousness. Changes in the individual’s social situation in which an individual may have to regain his or her place in a social group open up the possibility of increased social concerns. It appears that the family may be involved in modelling the childs attitude. Child-rearing styles of overprotection or control, rejection and a lack of warmth as well as restricted exposure to social stimuli and parental modelling of socially related concerns might play an important role. Other environmental factors that could be involved are peer rejection, childhood illness, social isolation and birth order. However, most of the reviewed studies used retrospective data and many did not measure actual social phobia, but related constructs, such as shyness, self- consciousness, social anxiety, and audience sensitivity. Thus, further research is necessary to clarify the processes underlying the development of social phobia

1.3.3. Risk Factors and Socio-demographic Correlates

Epidemiological studies are concurrent in coming to the conclusion that women are affected by social phobia more frequently than men (Magee et al., 1996, Faravelli et al., 2000, Schneier et al., 1992; Wittchen, Stein, et al., 1999), nevertheless there are also contradicting findings (Bourdon et al., 1988). Younger persons as well as persons with a lower socioeconomic status and unmarried persons tend to be more often affected than older, married or better educated people (Magee et al., 1996; Schneier et al., 1992; Schneier et al., 1994), although these factors are likely to be significantly inter-correlated. Studies have found proportions of over 50% of individuals who fulfil the criteria for social phobia to be unmarried, or to be divorced or separated (Furmark et al., 1999, Schneier et al., 1992; see also

Appendix C). Social phobia also seems to be a risk factor for weak school performance, truancy, premature termination of school, weak work performance and alcohol misuse (Mullaney & Trippett, 1979; Liebowitz, Gorman, Fyer, & Klein, 1985; Schneier et al., 1994; Schneier, Martin, Liebowitz, Gorman, & Fyer, 1989; Stein & Kean, 2000) as well as smoking and nicotine dependence (Sonntag, Wittchen, Hofler, Kessler, & Stein, 2000). Apart from showing reduced productivity at work, social phobic individuals spend more days out of work because of emotional problems (Stein, McQuaid, Laffaye, & McCahill, 1999; Wittchen, Stein, et al., 1999). A series of studies have shown individuals with social phobia to suffer from a reduced quality of life in various domains (Bech & Angst, 1996; Schneier et al., 1994; Mendlowicz & Stein, 2000; Stein & Kean, 2000; Wittchen, Fuetsch, Sonntag, Muller, & Liebowitz, 1999). In spite of this impairment, social phobia is poorly recognized and rarely treated by the mental health system (Katzelnick & Greist, 2001; Magee et al., 1996; Ross, 1993; Schneier et al., 1992; Wittchen, Fuetsch, et al., 1999; Wittchen, Stein, et al., 1999).

1.3.4. Comorbidity

The clinical picture of social phobia is complicated by the fact that it is often connected to other psychological disorders. In fact, comorbidity seems to be the rule rather than the exception. Den Boer (2000) analysed data from four US epidemiological studies, investigating a total of 361 persons, who fulfilled the diagnostic criteria for social phobia. On average, 80% of these individuals were diagnosed with a further lifetime diagnosis. Other anxiety disorders were found to be the largest category of comorbid disorders, followed by depression (20%) and alcohol misuse (15%). The tendency of social phobia to be related to a row of other psychological disorders is reported in many other clinical (Barlow, 1994; Gelernter et al, 1991; Otto et al., 2000; Turner, Beidel, Borden, Stanley, & Jacob, 1991) and epidemiological studies (Brown & Barlow, 1992; Perugi et al., 1999; Schneier et al., 1992).

1.4. The Biological Basis of Social Phobia

1.4.1. Genetic Factors

There is considerable evidence suggesting that genetic factors play an important role in the development of social phobia (Hudson & Rapee, 2000). The issue of genetics has been studied in a series of adoption, twin, and family studies. Several family studies have shown higher prevalence of social phobia in relatives of probands with social phobia than in relatives of probands with other anxiety disorders or no psychological disorders (Fyer, Mannuzza, Chapman, Liebowitz, & Klein, 1993; Reich 8c Yates, 1988; Stein et al., 1998). One of these family studies (Stein et al, 1998) found an increased risk for generalized social phobia in first- degree relatives of individuals with generalized social phobia, but not in relatives with non- generalized social phobia, which fits in well with the fact that other authors (Boone et al., 1999; Heimberg et al., 1990; Levin et al., 1993) found differences in biological reactions to social situations between generalized and non-generalized social phobics, suggesting differences in the biological basis of the two groups (Bell, Malizia, & Nutt, 1999). The findings of family studies are supported by twin-studies suggesting moderate heritability of social fears (Kendler, Neale, Kessler, Heath, & Eaves, 1992; Skre, Onstad, Torgersen, Lygren, & Kringlen, 2000).

1.4.2. Neurobiological Factors

Various models have been used to study neurobiological features of social phobia, including assessments of neurotransmitter function, response to chemical challenge, and neuroimaging. However den Boer (2000) points out that most studies involved limited numbers of patients and that there is still no clearly defined biological dysfunction in patients with social phobia.

Several findings in studies using different approaches underline the potential role of the dopaminergic system. First, a high comorbidity between Parkinson’s disease and social phobia has been found, generating the idea that dopamine depletion is a possible cause of social phobia (Lauterbach & Duvoisin, 1991; Richard, Schiffer, & Kurlan, 1996; Stein, Heuser, Juncos, & Uhde, 1990). Second, misuse of amphetamines seems to be capable of causing social phobia through dopamine depletion (Williams, Argyropoulos, & Nutt, 2000). Third, clinical observations of the effects of MAOIs (Liebowitz et al., 1992) also suggest a contribution of the domaninergic system in social phobia. Finally, studies using single photon emission computed tomography (SPECT) in patients with social phobia found striatal dopamine reuptake site densities to be markedly lower in social phobics than in matched comparison groups without a mental disorder (Schneier et al., 2000; Tiihonen et al., 1997). However, Bell et al. (1999) argue that in view of the clinical findings on dopamine it is unlikely that this observation is related to an increase in synaptic dopamine but to a decrease in the number of sites. Nevertheless, Stein (1998) concludes that a role for dysfunction within dopaminergic circuits in social phobia seems probable and further efforts in this direction are likely to be fruitful. However, the controversial interpretations (see also Coupland, 2001; den Boer, 2000) underline the necessity of further clarification of the exact role of dopamine.

A number of further findings point to other neurobiological factors that might be promising. Research on neurotransmitter abnormalities suggests that patients with social phobia may exhibit selective hypersensitivity of serotonergic systems (Tancer et al., 1995). Neuroimaging research has demonstrated that the amygdala is involved in the processing of neutral faces in individuals with generalized social phobia. Slides of neutral faces enhanced amygdala activation in social phobics, but not in the healthy controls, who only responded to emotional facial expressions with amygdala activation (Birbaumer et al., 1998). Finally, in experiments on chemical challenges social phobics have been found to react with an increase of anxiety to CO2 and to caffeine, similar to patients with panic disorder (compare Bell et al., 1999; den Boer, 2000).

On the other hand, it must also be pointed out that a number of studies have failed to find significant abnormalities in social phobics. For example, in a study using magnetic resonance imaging no difference could be demonstrated between patients with social phobia and normal control participants with respect to total, caudate, putamen, and thalamic volumes (Potts, Davidson, Krishnan, & Doraiswamy, 1994). Also, in a SPECT-study social phobics revealed no differences in cerebral blood flow in comparison to healthy comparison subjects (Stein & Leslie, 1996).

1.4.3. Evolutionary Factors

It has been suggested that social anxiety occurs as a result of social conflict and acts as a gesture of submissiveness to ward off attack from more dominant members of the same species thus avoiding fights and potential damage. As such, the socially anxious behaviour of some individuals is favourable for group cohesiveness and functioning as a social unit. This idea has led ethological theorists to state that social phobia has its onset in adolescence because that is the time when the individual is searching for his or her place within the social system (Ohman, 1986). In line with this evolutionary view is the assumption of a biological preparedness (Ost & Hugdahl, 1981; Ohman, 1986). The authors found that Pavlovian contingencies involving evolutionary fear relevant unconditioned and conditioned social stimuli (e.g. angry facial expressions) were much more effective in prompting conditioned fear than contingencies of evolutionary arbitrary stimuli. They concluded that there is a basic preparedness to react fearfully to such stimuli.

1.5. Cognitive and Behavioural Models of Explanation

1.5.1. Early Theories Classic Conditioning

Early models focused on classic conditioning and postulated that a traumatic experience, an embarrassing moment in a social situation is responsible for the onset of the phobia (Ost & Hugdahl, 1981; Ohman, 1986). An example of such an experience could be failing at the blackboard in front of the entire school-class and being laughed at, or beginning a flirt and being mocked or pitied. However, Hofmann, Ehlers, and Roth (1995) found that although traumatic experiences have been reported by individuals with public speaking anxiety, in almost all cases these occurred long after the onset of their social phobia. Deficits in Social Skills

Another theory was put forward suggesting that social phobia is the result of a principally reasonable, but exaggerated fear that has become contra productive in the course of time (Ost, Jerremalm, & Johansson, 1981; Trower, Bryant, & Argyle, 1978). This theory states that individuals with social phobia suffer from a lack of social skills, such as not knowing how to give a good speech (how to prepare, how to pronounce, how to dress), how to begin a conversation with a stranger or how to decline an offer etc.. Social skill deficiencies can also reveal themselves in rapid and breathy speech, tensed posture and jerky and poorly controlled gestures that increase the risk of embarrassment. Instead of training and optimising their skills, these individuals react with an increase of avoidance of social situations, which causes existent social competences to degenerate. Lack of social skill, in the sense of emitting fewer actions followed by less respondence, has been found to characterize depressed patients (Libet & Lewinsohn, 1973) and can possibly explain the onset and maintenance of social phobia for a subgroup of social phobics, but the empirical validation as a general model for social phobia has not been successful. Studies examining the social skills of socially anxious individuals have come to different conclusions, with some finding evidence for behavioural deficiencies (Stopa & Clark, 1993; Halford & Foddy, 1982) and others not (Clark & Arkowitz, 1975; Rapee & Lim, 1992). In fact, Trower et al. (1978) state themselves that many outpatient studies have failed to find clear evidence for the behavioural effect of social skills training in comparison to desensitisation. Furthermore, Heimberg (2001) points out that even if behavioural deficits are observed, it is unclear whether they are due to a lack of social knowledge or skill or to behavioural inhibition and avoidance produced by anxiety. Irrational Beliefs

Ellis (1962) formulated irrational beliefs as an explanation of neurotic disorders. He argues that social anxiety can be explained by the irrational belief that one must always make a good impression in order to be loved and accepted by everybody one is in contact with. Another aspect can be that people get hooked to the idea that they must always achieve perfect performances in order to be regarded as valuable, leading to fear of risk and failure. As a consequence, these people tend to be more occupied with themselves than with the task, which results in less enjoyment or actual failure. Even if people managed to achieve this perfectionist and actually unreachable goal, they would have to continuously worry about how much they are loved or whether they are still loved. According to Lazarus (1979) an overgeneralization of the self takes place when people see their whole ego questioned because of an imperfect performance in a social situation. This overgeneralization goes together with an absolutistic way of thinking and a low feeling of self-worth and thus is mainly responsible for anxiety, feelings of guilt and depressive reactions.

The model of self-representation by Schlenker and Leary (1982; Leary & Kowalski, 1995) takes a similar approach by postulating that the socially anxious person is particularly motivated to present a good, socially desired impression, while simultaneously suffering from a low feeling of self-esteem. A person will feel socially anxious to the degree that they doubt whether they are able to make such an impression. Vulnerability

Beck, Emery, and Greenberg (1985) also stress the role of cognitions in their proposal of a model of vulnerability. They state that persons will feel vulnerable in a given situation if they believe they are lacking important skills necessary to cope with it. The perception of insufficient coping skills makes the situation appear dangerous and triggers the “vulnerability mode”. Once this mode is activated, incoming data are processed in terms of the individual’s weaknesses rather than in terms of his or her resources (e.g. What if I can’t remember my next line?). The person will tend to downgrade his own abilities, since the immediate theme is weakness rather than strength. Incongruent, positive or functional information about the self or the situation are suppressed or distorted, because they have to go against the stream of negative ideation. The socially anxious person may determine his or her degree of vulnerability in an evaluative situation by the answers to a network of implicit questions “To what degree is this a test of my competence or acceptability?”, “How much do I have to prove myself to me or others?”, or “What is my status relative to that of my evaluators?” (Beck et al., 1985, p. 147). Public Self-Consciousness

In the centre of a further model is the concept of self-awareness and public self- consciousness. Buss (1980) argued that although everybody is apt to feel more self-aware in a public situation, this applies even more to socially anxious individuals, who tend to be high on the trait of public self-consciousness. Self-consciousness describes the process of observing and evaluating one’s own perception, thoughts, evaluations and somatic and motor processes and continuously checking these self-observations against a standard of social expectations. Drawing on evidence from experiments on self-esteem, Buss comes to the conclusion that public self-consciousness is likely to lead to inhibition of social responsivity and liveliness, as well as to discomfort, embarrassment or anxiety. Also, the perception of a discrepancy between what you are and what you think the social ideal is can diminish selfesteem. Nevertheless, he emphasises that apart from being self-conscious, several other factors may also serve to heighten a person’s motivation to manage impression, such as the characteristics of the other persons involved and the value of the goals in the interaction. Metacognition

According to Hartman (1983), the socially anxious person engages in too much self-focused meta-cognition, which refers to a self-monitoring of one’s thoughts and “involves the direct awareness of one’s behavioural intentions and inputs to motor systems and thus allows the person to edit the production of his or her behavior” (1983, p.440). The person is preoccupied with thoughts about his or her physiological arousal, ongoing performance and other people’s perception of him- or herself as socially incompetent, nervous or inadequate. Excessive focusing of attention on these normally automatic processes leads to a withdrawal of attention from the situation or the other person, resulting in a loss of efficiency and impairment in interpersonal performance. Hartman suggested that a negative sense of self combines with self-monitoring in producing anxiety. The perceptual and processing mechanism involves a feedback system, which results in an escalating anxiety cycle. Hartman (1983) proposes a combination of his model with the assumptions put forward by Schlenker and Leary (1982). However, Hartman assumes that the desire to make a good impression is an important consideration in the development stages of social anxiety. In later stages the selfconceptualisation as being socially anxious and the fear of embarrassment play a more important role than the desire to make a good impression.

In a review of numerous studies, Hope, Gansler, and Heimberg (1989) found self- consciousness, and particularly self-focused attention to be linked to social anxiety, but only when the subject is vulnerable due to another factor such as social evaluation or lack of confidence to perform well. They also come to the conclusion that physiological arousal or awareness of it leads to self-focused attention. They conclude that excessive self-focused attention may be most problematic for social phobics who experience more intense physiological reactions. Social phobics vary in the degree of their arousal (Ost, Jerremalm, & Johansson, 1981), however, with some exceptions (Jerremalm, Jansson, & Ost, 1986; Scholing & Emmelkamp, 1993a), little attention has been directed to differential response to treatment. STUDY III is to our knowledge the first study examining physiological arousal as a predictor for treatment response.

1.5.2. Integrative Models A Cognitive Behavioural Model of Social Phobia

The cognitive behavioural model by Heimberg, Juster, Hope, and Mattia (1995) does not actually present a new attempt at explanation, but aims at integrating various results from research and existing models. The model is based on the assumption of a predisposition to develop social phobia, which may be inherited or produced by factors in the childhood or adolescent environment, which have sensitised the person to threatening aspects of social encounters. Such factors can include a socially anxious parent, perfectionist standards, or overprotection and isolation from social contacts. Negative peer group or heterosexual experiences may also sensitise the child or adolescent to the potential consequences of social situations. This hypothesis is supported by some retrospective and child research (for a review see Hudson & Rapee, 2000). Heimberg et al. (1995) state that these experiences result in a set of beliefs that increase the probability that the person will approach social situations apprehensively or try and avoid them. These beliefs include the assumption that social encounters are dangerous to one’s self-esteem, that the only way to avoid negative outcomes is to perform perfectly, and that he or she does not have what it takes to perform perfectly. As a consequence the person will anticipate humiliation, embarrassment and rejection and experience increased arousal before and during the social situation. The increased arousal then provides the person with further evidence of danger and may lead him or her to feel anxious that the anxiety will become visible to others. The authors provide a feed-back-model, in which the various processes feed into each other and contribute to the escalation of a person’s anxiety and possibly even result in a disruption of behavioural performance. However, even if performance does not objectively suffer, the authors state that the person is likely to decide that it was inadequate, because he or she compares it to a perfectionist standard and expects that others will evaluate it in the same way. In the end the sequence serves to affirm the negative beliefs and predictions and to increase the probability that the next social incidence will be experienced similarly. A Cognitive Model of Social Phobia

Wells & Clark (1997) argue that although Hartman (1983) and others have underlined the pivotal role of self-focused attention in the maintenance of social phobia, the mechanisms they describe linking self-focus to social phobia are likely to operate in other disorders and it is necessary to specify social phobic specific mechanisms. Drawing on the given theories and extensive clinical work, Clark and Wells (1995) advanced a cognitive model of social phobia. In the model, the social phobic is motivated to present a favourable impression but is insecure in his ability to do so in particular situations. This insecurity is explained as a manifestation of negative self-focused processing. It is linked to safety behaviours that are intended to protect self-esteem and prevent negative judgements from others. Safety behaviours differ from simple avoidance of the complete social situation. For example, someone can merely be avoiding eye contact. The avoidance of revealing blushing by wearing a thick layer of makeup or sweating by wearing particularly cool clothes or using deodorant several times a day are also considered safety behaviours. The authors state that some of these safety behaviours can paradoxically inflame problematic symptoms and increase the likelihood of poor performance. They propose that safety behaviours can maintain distorted thinking in social phobia by exacerbation of symptoms, by prevention of disconfirmation, by maintenance of self-attention, or by contamination of the social situation. The negative consequences of safety behaviours as well as somatic symptoms and cognitive interpretations feed back to the self-consciousness and reinforce distorted impressions of the self. The authors distinguish three phases of distorted processing. Dysfunctional processing can occur in the phobic situation itself, in advance of the situation as apprehension and rumination or, finally, after leaving the situation it is likely to continue as a “post mortem”, in which the social phobic goes over the situation, contemplating how it was, how it should have been and what the possible consequences are. However, the authors emphasize that the most important of these phases with regard to problem maintenance is the phase in the actual social situation. Similar to Beck et al. (1985) they state that the social situation activates dysfunctional conditional assumptions (e.g. If I am quiet people will think I’m boring), self-beliefs (e.g. I’m different) or rigid rules for social situations (e.g. I must always sound fluent and intelligent). Schemas of this type make the individual vulnerable to perceiving social situations as potentially dangerous, leading to somatic and cognitive symptoms and inadequate safety behaviours. Also, when the socially anxious individual enters the social situation, there is a shift in his or her focus of attention towards an intensified negative self-processing. This self-focused attention, which is experienced as an increase in self-consciousness, reduces the attention available for processing external information and increases anxiety. The basic components of the model interact with each other in the maintenance of fear through four key feedback cycles. The self-processing can serve to increase danger appraisals. Safety behaviours maintain negative self-beliefs as well as negatively bias the appraisals of others. Finally, anxiety symptoms offer subjective support to distorted self-appraisals.

1.6. Treatment of Social Phobia

So far, research has focused on cognitive behavioural treatment strategies as well as pharmacological treatment. The major classes of cognitive behavioural therapies that have been applied to social phobia include exposure, cognitive restructuring, relaxation training techniques and social skills training (Heimberg, 2001). Many of the strategies have been derived from the biological and psychological models described above. The usefulness of relaxation strategies was concluded from the knowledge of physiological arousal and its possible impact on self-focused attention. Social skills training is delineated from the model of social skill deficits. Cognitive interventions, such as restructuring beliefs and interpretations as well as re-shifting attention are derived from the cognitive theories (Beck et al., 1985; Buss, 1980; Clark and Wells, 1995; Ellis, 1962; Hartman, 1983; Schlenker & Leary, 1982). The expectancy of a positive effect of exposure was rendered from the good results achieved with patients suffering from simple phobia and panic and agoraphobia (Butler, Cullington, Munby, Amies, & Gelder, 1984), who share a number of common features with social phobic individuals. Similarly, many of the psychopharmacologic therapies were tested because of the good results achieved with patients suffering from major depression.

1.6.1. Cognitive Behavioural Interventions Relaxation Techniques

Relaxation techniques aim at helping the patient to learn to attend to and control the degree of physiological arousal experienced during or in anticipation of feared events. Most of the relaxation techniques, including systematic desensitization are derived from the pioneering work of Wolpe (1969). However, research on systematic desensitisation for social anxiety is meagre, yielding contradicting results. Marzillier, Lambert, and Kellett (1976) tested systematic desensitisation in a sample of psychiatric out-patients with social or interpersonal difficulties and found it not to be superior to an untreated control group. Florin and Gurk (1978) developed a program for the treatment of exam anxiety, in which relaxation techniques took up a large part. Of the participants in the program 50% stated that it had helped them very much in overcoming anxiety. Jerremalm et al. (1986) suggested that relaxation techniques might be specifically effective for patients with fear of physical reactions, but could not support this hypothesis in their treatment outcome study. Social Skills Training

The most commonly used techniques in social skills training are therapist modelling, behavioural rehearsal, corrective feedback, social reinforcement and homework assignments (Trower et al., 1978). Studies investigating the effects of social skill training have yielded non-uniform results. Mersch, Emmelkamp, Bogels, and Van der Sleen (1989) compared it to rational emotive therapy and found it to be equally effective. However they did not find it to be more effective for patients who performed weakly in a social interaction test, thus lending no support to the hypothesis that it might be particularly helpful for this subgroup of patients. Wlazlo, Schroeder-Hartwig, Hand, Kaiser, and Munchau (1990) found no significant difference in treatment efficacy between social skills training and exposure. Also, Stravynski, Marks, and Yule (1982) found no superior effect, when social skills training was combined with cognitive modification. On the other hand, Marzillier et al. (1976) found a waiting-list control group to make a comparable progress to a group of patients treated with social skills training over a period of three to four months. Also, Trower et al. (1978) point out themselves that many outpatient studies have failed to find clear evidence for the behavioural effect of social skills training in comparison to desensitisation. As the effectiveness of social skills training alone for social phobia is questionable it is often combined with exposure (Hofmann et al., 1995; Turner, Beidel, Cooley, Woody, & Messer, 1994) yielding satisfying results. Exposure

Exposure requires the patient to imagine (in sensu exposure) or actually confront (in vivo exposure) the feared stimuli. In most cases, the first step is to generate a list of problematic situations with the patient. Such situations frequently concern giving a speech to an audience, serving drinks, being interviewed, asking for a date. The situations are rank-ordered and (mostly) the patient will begin exposure to a moderately feared situation to gain confidence and experience success before addressing more feared situations (Fresco & Heimberg, 2001). In the early stages of the treatment, situations are entered in company of the therapist and the patient is asked to remain in the situation until he or she has experienced a certain degree of habituation to it. After repeated and prolonged exposure and when the situation no longer elicits a distressing level of fear, exposure is continued in the next situation. This process continues until the patient can master all the feared situations with a significantly reduced amount of anxiety.

Several studies have demonstrated a clear efficacy of exposure for social phobia (Alden, 1989; Butler et al, 1984; Fava, Grandi, & Canestrari, 1989; Newman, Hofmann, Trabert, Roth, & Taylor, 1994; Turner, Beidel, & Jacob, 1994; Mattick & Peters, 1988; Mattick, Peters, & Clarke, 1989; Mersch, 1995). Nevertheless, a number of problems arise when treating social phobia with pure exposure, which have led some authors (e.g. Fresco & Heimberg, 2001) to question its sufficiency for social phobia. Butler (1985) has listed these difficulties, which include the problem of clearly specifying tasks in advance, because social situations are variable and unpredictable, the time limit of many social situations, and the central role of thoughts and attitudes that are difficult to control in the situation. The postmortem processing problem described by Clark and Wells (1995) can be added. Fresco and Heimberg (2001) point out that exposure is maximally effective when patients fully engage in all aspects of the situation in contrast to distracting themselves and focusing on negative evaluations and predictions or applying safety behaviours. Cognitive Restructuring

Cognitive restructuring consists of a set of interventions originating from the cognitive theory and therapies of Beck et al. (1985) and Ellis (1962). Individuals are taught to identify irrational or negative thoughts that occur during the anxiety-provoking situation. Next, they are taught to evaluate the accuracy of those thoughts as compared with objective information, which is derived by repeated questioning or, as an alternative, by behavioural experiments, such as observing others in a social situation or testing the effect of safety behaviours (Ellis, 1962; Clark and Wells, 1995). When dysfunctional thoughts are triggered by general beliefs, the therapist will question these beliefs (Ellis, 1962). Finally, the patient is motivated to develop rational alternative thoughts based on the acquired information.

Recent research on cognitive interventions focuses on a treatment based on the model of Clark and Wells (1995). Treatment consists of deriving an idiosyncratic version of the model, which is used as a point of reference during treatment, identifying safety behaviours and demonstrating their adverse effects via experimental exercises, training patients to shift their attention away from the self to the external situation (as already suggested by Hartman, 1983), video-feedback to modify distorted self-imagery, behavioural experiments and identifying and modifying problematic anticipatory and post-event processing.

Behavioural experiments contain exposure elements, although exposure is not applied as systematically as described above. In the cognitive approach, exposure is less about habituation but more about the opportunity for patients to collect information that will enable them to revise their judgement about the degree of risk in a given situation (Heimberg, 2001). Although most studies have investigated the combined effect of exposure and cognitive restructuring, studies that only evaluated cognitive interventions supply strong evidence for their efficacy, particularly for the rational emotive therapy (Kanter & Goldfried, 1979; Schelver & Gutsch, 1983) but also for the treatment developed by Clark and Wells (1995) (Stangier, Heidenreich, Peitz, Lauterbach, & Clark, 2002). Additionally, Hofmann (2000) found changes in self-focused attention to be highly correlated with pre-post differences in social phobic anxiety. Combination of Exposure and Cognitive Restructuring Heimberg et al. (1995) have presented a specific cognitive-behavioural group treatment (CBGT) for social phobia. The treatment is conducted in 12 weekly sessions that last for approximately 2.5 hours and is typically administered to groups of six patients and conducted by two co-therapists (Fresco & Heimberg, 2001). Treatment consists of developing a cognitive-behavioural explanation of social phobia, training patients in the skill of identifying, analysing, and disputing problematic cognitions, exposure to simulations of feared situations, cognitive restructuring, in vivo exposure as homework assignments and teaching patients to self-administer cognitive restructuring in combination with homework assignments. CBGT has received the most empirical attention and support (Cox, Ross, Swinson, & Direnfeld, 1998; Gruber, Moran, Roth, & Taylor, 2001; Heimberg, Becker, Goldfinger, & Vermilyea, 1985; Heimberg, Dodge, et al., 1990; Heimberg et al., 1998; Hope, Heimberg, & Bruch, 1995; Hope, Herbert, & White, 1995; Otto et al., 2000). STUDY I describes an individualized approach to the combination of exposure and cognitive interventions (see Appendix D).

1.6.2. Pharmacological Treatment

The goals of pharmacotherapy for social phobia aim at ameliorating the target symptoms, such as anticipatory anxiety, socially cued panic, avoidance behaviour and dysphonic ruminations, to address comorbid conditions as well as to achieve remission and recovery. To achieve this, clinicians have been using various chemical agents (Marshall, 1993; Miner & Davidson, 1995; Scott & Heimberg, 2000; Walker & Kjernistedt, 2000). Irreversible, nonspecific monoamine oxidase inhibitors (MAOIs) have been shown to achieve a positive response rate in studies using phenelzine (Gelernter et al., 1991; Liebowitz et al., 1992), but a lower response for atenolol (Turner, Beidel, & Jacob, 1994). However, despite the well- established efficacy, clinicians rarely chose MAOIs as first-line treatment for social phobia, because of the need for a low tyramine diet and diverse side effects. Other studies have supported the efficacy of clonazepan (Davidson et al., 1993; Munjack, Baltazar, Bohn, Cabe, & Appleton, 1990; Otto et al., 2000) as well as aprazolan (Gelernter et ah, 1991). However, the use of benzodiazepines must be questioned, as many social phobic patients suffer from comorbid alcohol dependence, which is a contraindication for the use of benzodiazepines. All existing selective serotonin reuptake inhibitors (SSRIs) have been studied in the treatment of social anxiety and the evidence from controlled studies supports their efficacy, specifically the efficacy of sertraline (Blomhoff et ah, 2001), fluvoxamine (Stein, Fyer, Davidson, Pollack, & Wiita, 1999), and paroxetine (Baldwin, Bobes, Stein, Scharwachter, & Faure, 1999). The SSRIs seem to be emerging as first line pharmacological treatment for social phobia. They are well tolerated in the short- and long term, safe and also effective in treating frequent comorbid disorders, such as depression (Walker & Kjernisted, 2000). In spite of their efficacy in the treatment of major depression, beta-blockers have been proved less effective in the treatment of social phobia (Liebowitz et ah, 1992; Turner, Beidel, & Jacob, 1994).

1.6.3. Present State of Treatment Research

In the area of social phobia a series of meta-analyses have found a high efficacy of cognitive behavioural treatments in the reduction of social phobic anxiety, with mean effect sizes ranging from 0.8 to 1.1 (Fedoroff & Taylor, 2001; Gould, Buckminster, Pollack, Otto, & Yap, 1997; Feske & Chambless, 1995; Ruhmland & Margraf, 2001; Taylor, 1996). However, it should be pointed out that the meta-analyses were based on similar pools of studies for cognitive-behavioural treatments as the amount of studies is limited. Table 1.1. shows the mean pre-post and pre-follow-up effect sizes for the different treatment conditions from the given meta-analyses.

Fedoroff and Taylor (2001) found treatment with benzodiazepines to be significantly more effective than all other strategies with exception of SSRIs. However, they report follow-up studies only for psychotherapy with effect sizes in the range of attention placebo. The studies they analysed used varying follow-up periods (up to six months), but there was no significant effect when the authors controlled for length.

Table 1.1.

Mean Effect Sizes and Number of Trials for Psychological and Pharmacological Interventions in Meta-Analyses.

  Fedoroff & Gould et al. Feske & Ruhmland Taylor
  Taylor (1997) Chambless & Margraf (1996)
  (2001)   (1995) (2001)  
Wait-list control


0.03 (9)     0.03 (5) -0.13 (5)
FU (1-6 months) Attention placebo


0.45 (4)       0.48 (5)
FU (1-6 months) Exposure 0.42(1) 0.89(9)’      
Post 1.08 (7) 0.99 (9) 1.76 (7) 0.82 (8)
FU (1-6 months) Cognitive Therapy 1.31 (7) 0.60 (4)1 1.04 (7) 1.06 (6) 0.93 (8)
Post 0.72 (7)   1.13 (3) 0.63 (5)
FU (1-6 months) EX and CT 0.78 (5) 0.80 (8)1     0.96(5)
Post 0.84(21) 0.90(12) 1.07(17) 1.06(11)
FU (1-6 months) 0.95 (10)   1.10(10) 1.39(13) 1.08 (9)
Social Skill Training 0.64(7) 0.60 (3)1   0.85 (2) 0.65 (4)
Post 0.86(4)       0.99 (3)
FU (1-6 months) Relaxation


0.51 (4)     0.44 (2)  


2.10(5) 0.72 (2)1      
SSRI   1.89 (2)1      
Post 1.70(12)      




1.08(15) 0.64 (5)1 -0.08 (3)1      

Numbers in parenthesis reflecl Ihe number of trials. Post = post-assessment after treatment, FU = follow-up (1-6 months), EX = exposure, CT = cognitive therapy, SSRI = selective serotonin reuptakc inhibitors, MAOIs = monoamine oxidase inhibitors. 1 = controlled pre-post and follow-up effect sizes.

In the meta-analysis by Ruhmland and Margraf (2001) studies investigating social skills training and relaxation strategies achieved significantly lower effect sizes than exposure, cognitive therapy or a combination of both. Nonetheless, all treatment strategies were better than the waiting list control groups. In order to have a better comparison with the other metaanalyses, the 1-6 months follow-up period is reported in table 1. However, follow-up data over longer periods of time for exposure and cognitive behavioural treatment were reported for five studies, yielding effect sizes comparable to those at post-assessment.

In the meta-analysis by Gould et al. (1997) a more conservative approach was chosen, including only controlled studies and excluding open trial pharmacological studies. Followup-data (3-6 months) are only reported for single studies. They found studies reporting follow-up-data to have a mean follow-up effect size .21, suggesting that subjects continued to make modest improvement, with the exception of the only follow-up study investigating pharmacotherapy, which indicated no further treatment gains. Gould et al. also examined the costs of treatments in relation to their efficacy. Cognitive behavioural group treatment was found to be clearly the least costly intervention, and clonazepan the least costly pharmacological intervention, especially by the end of the second year. Individual cognitive behavioural therapy combined with clonazepan and phenelzine totalled about twice the charge of group treatment and treatments with fluvoxamine were clearly the most expensive interventions.

The meta-analysis by Feske and Chambless (1995) concentrated on the comparison of studies testing cognitive behaviour therapy and studies testing exposure treatment. Their results indicated that treatment modalities are equally effective.

Taylor (1996) compared waiting-list control, placebo, exposure, cognitive therapy, a combination of exposure and cognitive restructuring and social skills training. He found all treatment conditions including placebo to differ significantly from the waiting-list-control group and only the combination of exposure and cognitive restructuring to yield a significantly larger effect than placebo. He also found a tendency for the effects of treatment to increase by a 3-month follow-up.

In sum, it seems that cognitive behavioural treatment is an effective and relatively inexpensive treatment that provides stable long-term effects. Although medical treatment (particularly SSRI) tends to be more effective on a short-term basis, long-term effectiveness is questionable and evidence for it has yet to be delivered. The doubt whether medical treatment is capable of producing stable effects is supported by the results of a large comparison study of CBGT and phenelzine (Heimberg et al., 1998; Liebowitz et al., 1999). At post-test both groups had improved comparably, although phenelzine patients had improved more on a subset of measures. Also, many of the phenelzine patients who were classified as responders at post (after 12 weeks) had already achieved gains by the six-week assessment and this was less common among the patients treated with CBGT. However, after a follow-up period of six months, 50% of the previously responding phenelzine patients relapsed, compared to 17% of the CBGT patients.

To date, there are no published studies that have examined the combined effectiveness of cognitive-behavioural-and pharmacological treatments, although there are some being conducted at present (Heimberg, 2001).

Apart from the need for further investigation of long-term effects for medical treatment, I would like to emphasize two further issues arising from the current state of treatment research for the treatment of social phobia.

First, in spite of the effective treatment, social phobia is an under-treated psychological disorder (Katzelnick & Greist, 2001; Magee et ah, 1996; Ross, 1993; Schneier et ah, 1992; Wittchen, Fuetsch, et ah, 1999; Wittchen, Stein, et ah, 1999). Ross (1993) lists a row of barriers for treatment, expressed by people who contacted the Anxiety Disorders Association of America. They include ignorance about social phobic fears on the part of health professionals and the public, trivialization of the problem by family and friends, under diagnosis, the stigma attached to mental disorders in general, the sense of secrecy, shame, and embarrassment that accompanies social phobia in particular and the lack of access to affordable and professional care. Even of those social phobic patients who overcome the first boundaries and are fortunate enough to receive an adequate treatment offer, not all take up that offer and not all profit from treatment or are able to maintain success over a longer period of time. Scott and Heimberg (2000) point out that clinicians should be aware of alternative strategies for the treatment of social phobia because no treatment has been shown effective for all individuals. Thus, further research should focus on the question of which patients might benefit from which treatment. The question of whether there are patient characteristics on the basis of which the clinician is able to predict treatment attrition and success is addressed in STUDY III.

Second, the fact that treatment has been shown to be effective under research conditions does not necessarily mean that it will be equally effective in clinical practice. In fact, many practitioners doubt whether they will be as successful with their patients as researchers are with the patients they investigate and treat. There are numerous differences between research conditions and clinical practice, ranging from characteristics of the sample to the type of building or the training of therapists. The question of whether these have any influence on the size of the effect is attended to in STUDY I. Finally, there is a need to demonstrate that cognitive behavioural treatment will work just as well in clinical practice, by comparing the effects achieved in clinical practice with those achieved in randomised controlled trials. This is done in STUDY II.

In the following chapter, the conducted studies will be introduced at more length, giving a brief description of their purposes and methods as well as a summary of the results. The complete descriptions of the studies are depicted in the chapters 3, 4, and 5.


  1. Purpose and Summary of the Studies

2.1. Purpose and Summary of STUDY I

Although the effectiveness of cognitive behavioural treatment for social phobia has been studied in a large number of outcome studies and re-analysed in a row of meta-analyses, the question of whether this treatment will work in clinical practice remains unanswered. Private practitioners and other psychotherapists working under no research conditions often argue that their patients obviously differ from the research samples and that they therefore do not obtain as good results as those reported in the given literature. Yet very little attention has been given to the question of the generalization of these results to clinical practice. Juster, Heimberg, and Engelberg (1995) investigated self-selection and sample selection in a treatment study of social phobia. They found that although acceptors were found to score higher on only one of 25 pretreatment measures of clinical functioning, they improved significantly more on 3 of 5 posttreatment measures (global improvement, social anxiety and avoidance) than refusers or excluded patients. Weisz, Weiss, and Donenberg found for child and adolescent therapies that “research focusing on more representative treatment of referred clients in clinics has shown more modest effects, in fact, most clinic studies have not shown significant effects” (1992, p.1578). Shadish et al. (1997) conducted a secondary analysis of meta-analytic data and found very few studies that were even remotely clinically representative. However, studies that fulfilled a certain number of the criteria revealed effect sizes that were about 10% smaller than those of the complete sample of therapy studies.

These findings are in line with “a growing recognition that controlled clinical trials may not capture the full richness and variability of actual clinical practice and a concern on the part of some that the very process of randomisation may undermine the representativeness of clinical encounter” (Chambless & Hollon, 1998, p. 14) and underline the importance of more research to answer the question of generalization of treatment effects. It is possible that the selection criteria generally applied in efficacy studies lead to homogenous samples with low standard deviations in the applied measures. A small denominator in the fraction calculating the effect size could result in an overestimation of treatment effects in comparison to typical clinical samples. Thus, the aim of STUDY I was to direct further attention to the possibility that higher effect sizes in the treatment of social phobia are achieved in typical research conditions and that these are not due to the quality of treatment but to sample selection and study characteristics.

To do this, we re-examined the current research on social phobia treatment and selected studies for which pre-post effect sizes could be calculated with the provided means and standard deviations for the outcome measures across treatment. Thirty studies1 (see Appendix A) on cognitive and/or behavioural treatment of social phobia or severe interpersonal anxiety met our criteria for inclusion and were selected for our review. We categorized the studies according to common exclusion criteria, the heterogeneity of the sample and laboratory study characteristics according to the criteria listed by Shadish et al. (1997) and compared the mean effect size (ES) and standard deviation (SD) for each group of studies according to the applied sample and study criteria. We also calculated a laboratory and a restriction score according to the amount of applied typical research criteria a study fulfilled and analysed the correlation of these scores with the effect sizes.

Generally, the results of STUDY I did not offer convincing evidence for the assumption that effect sizes might be explained by the failure to gain typical samples. Two of the direct group comparisons even revealed the opposite effect. Patient samples that included patients with comorbid avoidant personality disorder and patients with prior treatment experience revealed higher effect sizes than samples without. Also, the results indicate that even the accumulation of sample restriction does not have predictive value for the pre-post effect sizes of treatment. However, there were some results in support of the observations made by private practitioners. Samples excluding patients with comorbid psychosis, substance misuse and bipolar disorder were shown to reach higher effects than those including these patients. The same applied for studies that were conducted following a treatment manual. We also found studies working with participants who were homogeneous in the length of their disorder to produce higher treatment effects than the other studies. There was strong evidence for the hypothesis that this relation is moderated by the size of the standard deviations in the applied measures. Finally, and most importantly we found evidence for an influence of accumulated laboratory criteria for research studies on the effect size. There was a significant tendency for studies applying laboratory treatment conditions, such as recruiting patients by adverts, applying treatment in university settings, using specifically trained therapists and following and monitoring treatment manuals to achieve higher effect sizes.


In summary, the results in STUDY I are in line with Shadish et al. (1997) in finding a tendency for studies applying a row of research criteria to reveal slightly lower effect sizes. However, we found that this is not due to sample restriction in typical research studies. It seems that the laboratory characteristics, such as recruiting patients, the place of the study, the training of therapists or the implementation of a treatment manual have more influence on the difference.

2.2. Purpose and Summary of STUDY II

STUDY II also addressed the potential gap between clinical research and practice, by following the recently popular distinction between the efficacy of psychotherapy and its effectiveness (Weisz, Donenberg, Han, & Weiss, 1995). Efficacy (or research therapy) refers to the effects of psychotherapy in randomised, controlled trials, usually conducted in university settings involving recruited patient clients, using a highly structured treatment manual for a narrow problem focus and trying to establish a high degree of internal validity. Effectiveness (or clinical therapy) refers to the effects of natural clinical psychotherapy conducted in the field, which means in private practice or in mental health centres, using quasi-experimental designs and trying to establish a high degree of external validity or generalization of results to various settings. All of the treatment studies carried out so far can be classified as efficacy studies with varying amounts of sample restriction and laboratory study conditions. So far, no study has tested the hypothesis whether treatment for social phobia can be delivered with the same effectiveness in a clinical setting, in which patients are not recruited by adverts, not randomised to treatment groups or preselected in a way typical of research but are part of the usual referral system and medical routine. STUDY II was an attempt to investigate the generalization of an empirically supported treatment for social phobia to a clinical setting. STUDY II also investigated the possibility that the effect-size could be enhanced by restricting the sample of patients according to the criteria employed in research settings.

The effectiveness of exposure combined with cognitive restructuring was examined in four outpatient clinics in the community and a large number of experienced and inexperienced therapists. Participants were 217 patients diagnosed with social phobia as the primary disorder who agreed to undergo treatment in one of four outpatient clinics run by the Christoph- Domier-Foundation for Clinical Psychology (CDS). The patients were treated with high density in vivo exposure, supplemented by cognitive interventions (Appendix D). Patients were assessed before treatment and six weeks after treatment (Appendix B) with a large battery of disorder specific and related self-report measures (Appendix E).

The results of STUDY II provided support for the clinical effectiveness of exposure combined with cognitive interventions for patients with social phobia. Fifty-six percent of the patients had reliably improved on social phobic fears and 57% were more likely to be drawn from a healthy population sample six weeks after the end of therapy. The rate of patients who felt impaired in important areas of their life dropped significantly, indicating that patients succeeded in transferring the effects of therapy into their every-day-life. The mean effect size for the measures of social phobia was 0.82, thus being within, but at the bottom range of the effect sizes reported in the meta-analyses for cognitive behavioural therapy (Fedoroff & Taylor, 2001; Gould et al., 1997; Feske & Chambless, 1995; Ruhmland & Margraf, 2001; Taylor, 1996). No higher effect size was attained when the sample was restricted, applying frequently used selection criteria. Even the comparison of a subgroup of patients, for which a row of restriction criteria was applied (low depression, no prior treatment, a medium age of 20-50 and homogenous in the severity of disorder) did not reveal a higher effect size than the remaining sample. The finding that the effect size was in the bottom range of those found in meta-analyses might be due to the fact that the questionnaire measures applied were less sensitive to change than those in the comparison studies that have been analysed in the metaanalyses. A direct comparison with studies using the same measures provides a different picture, as effect-sizes in these studies tended to be lower.

One shortcoming of STUDY II was that 11% of the patients could not be motivated to complete the follow-up questionnaires at post assessment. Although no significant differences were found between these patients and the ones who completed the follow-up questionnaires, there was a tendency for them to occupy an intermediate position between the completers and the dropouts, who differed significantly from one another on some of the measures. The higher depression scores and comorbidity found for patients who dropped out of therapy also underline the necessity of giving further attention to this group of patients. Nevertheless, an intent-to-treat-analysis also produced highly significant pre-post differences.

Taken together, STUDY II provided convincing evidence that empirically validated treatment for social phobia, the combination of exposure and cognitive restructuring, can be transported into natural field settings. However, it is most likely that these results require not only a thorough diagnostic procedure to assess social phobia as the primary problem but also frequent and maintained supervision of the therapists.

2.3. Purpose and Summary of STUDY III

Today, the social phobic health service user is in the fortunate position of having a range of treatments from which to choose. However, not all patients benefit from the tested treatment approaches. Turner, Beidel, Wolff, Spaulding, and Jacob (1996) calculated treatment success, taking into consideration not only patients who completed treatment but also those who were offered treatment, but refused or dropped out of it. This resulted in an alarmingly low rate of 52% of the patients seeking treatment for social phobia who actually profited from it. Also, there is little information available to indicate which patient with social phobia is more likely to benefit from which treatment (Walker & Kjemistedt, 2000). Awareness of prognostic features can be helpful in indicating treatments of choice, since a variety of effective treatment variations are available. Knowing about factors that are responsible for attrition as well as for failure to benefit from treatment may help to understand the processes underlying treatment and enable the therapist to adapt treatment procedures, delivery and planning accordingly to improve a specific patient’s prognosis (van Minnen, Arntz, & Keijsers, 2002). A series of studies have investigated predictors for treatment response in social phobia (e.g. Chambless et al., 1997; Mersch, Emmelkamp, & Lips, 1991; Salaberia & Echeburua, 1996; Scholing & Emmelkamp, 1999; Turner et al., 1996). Most studies have concentrated on predicting change caused by treatment and end state functioning, focusing on sociodemographic and biographical variables, impairment, severity, subtypes, and comorbidity as potential predictors for change or end state functioning. Little attention has been directed to the questions of treatment dropout, refusal or relapse after treatment. Also, a number of promising variables has not been examined as predictors. Finally, the available studies are limited by the fact that predictors were studied in the context of controlled outcome studies whose inclusion criteria are likely to limit the variability of the factors studied as predictors. Specifically in the case of treatment refusal the question must be raised whether refusal of participating in a study with random assignment can be compared to the refusal to take up a (individualized) treatment offer as such. The aim of STUDY III was therefore to search for predictors of treatment acceptance, attrition, effectiveness, and relapses after treatment in a field treatment outcome study for social phobia and to compare these to variables identified as predictors in the context of controlled efficacy studies.

Patients diagnosed with social phobia seeking treatment in a naturalistic setting (N = 287, for a detailed description of this sample see Appendix C) were classified as refusers prior to treatment (16%), refusers after cognitive preparation (8%), dropouts (6%), and completers (69%). Outcome was assessed by residual gain scores and patient improvement ratings six weeks and one year after the end of treatment. Patients who completed the one-year follow-up were categorized as stable (87%) or deteriorated (13%). Demographic and disorder-related as well as therapist and treatment variables were used as predictors for each classification.

The results of STUDY in indicate that approximately only 43% of the patients seeking treatment actually completed and benefited from it in the end. The only predictor for treatment attrition was comorbidity. Treatment gain was best predicted by satisfaction with health. Also, patients characterized by more generalized social phobia improved less by 1- year-follow-up. Pretreatment depression had no effect on change as assessed by the self-report measures, although more depressed patients reported having improved less. Patients who were more severely impaired at pretreatment found it harder to maintain treatment gain. Three important clinical implications were derived from the results of STUDY III. (1) Treatment refusers are as severely impaired by social phobic symptoms as patients who undergo treatment and additional efforts are needed to motivate these patients to take up treatment. (2) Cognitive preparation and the beginning of treatment should be even more adapted to pretreatment feelings of impairment and comorbid disorders, by restructuring hampering cognitions or conducting specific treatment for comorbid disorders. (3) It seems important to arrange for additional sessions over a specific period of time when patients are more severely impaired or suffer from more generalized social phobia, to enable them to integrate the treatment effects into their everyday life.


  1. STUDY I

How much do Sample Characteristics Affect the Effect Sizes? – An Investigation of Studies Testing the Treatment Effects for Social Phobia. 2

3.1. Introduction

Private practitioners and other psychotherapists working under no research conditions often argue that their patients obviously differ from the research samples and that they therefore do not obtain as good results as those reported in the given literature. Possibly as a reply to this, writers have recently begun to distinguish between the efficacy of psychotherapy and its effectiveness (Weisz, Donenberg, Han, & Weiss, 1995). Efficacy (or research therapy) refers to the effects of psychotherapy in randomised, controlled trials trying to establish a high degree of internal validity. These are usually conducted in university settings, involving recruited patient clients, selected according to inclusion criteria and using a highly structured treatment manual for a narrow problem focus. Effectiveness (or clinic therapy) refers to the effects of natural clinical psychotherapy conducted in the field, which means in private practice or in mental health centres, using quasi-experimental designs and trying to establish a high degree of external validity. While the efficacy of psychotherapy is generally well established, the generalization of efficacy findings can be challenged. Weisz, Weiss, and Donenberg found for child and adolescent therapies that “research focusing on more representative treatment of referred clients in clinics has shown more modest effects, in fact, most clinic studies have not shown significant effects” (1992, p. 1578). Shadish, Matt, Navarro, Siegle, Crits-Christoph, Hazelrigg, et al. (1997) conducted a secondary analysis of meta-analytic data and found very few studies that were even remotely clinically representative. For a study to pass as clinical it had to be carried out in non-university settings, involve patients that were referred through usual clinical routes, involve experienced, professional therapists with regular caseloads and free to use a wide variety of procedures in treatment rather than therapists in training or trained specifically for the purpose. The therapists were also not to have used a treatment manual and the implementation of the treatment should not have been monitored. Finally, the studies were to have used clients who were heterogeneous in personal characteristics as well as in focal presenting problems. Only one study fulfilled the authors’ complete set of criteria for clinic therapy. However, studies that fulfilled a certain number of the criteria revealed effect sizes that were about 10% smaller than those of the complete sample of therapy studies.

In the area of social phobia a series of meta-analyses have found a high efficacy of cognitive behavioural treatments in the reduction of social phobic anxiety, with mean effect sizes ranging from 0.8 to 1.1 (Fedoroff & Taylor, 2001; Gould, Buckminster, Pollack, Otto, & Yap, 1997; Feske & Chambless, 1995; Ruhmland & Margraf, 2001; Taylor, 1996). Yet very little attention has been given to the question of generalization of these results to clinical practice. Juster, Heimberg, and Engelberg (1995) investigated self-selection and sample selection in a treatment study of social phobia. They found that although acceptors were found to score higher on only one of 25 pretreatment measures of clinical functioning, they improved significantly more on 3 of 5 posttreatment measures (global improvement, social anxiety and avoidance) than refusers or excluded patients. In a large clinical practice study (Lincoln, Rief, Hahlweg, Frank, von Witzleben, et al. 2002) we found the effect size for treatment of social phobia to be at the bottom range of those reported in meta-analyses. These findings are in line with “a growing recognition that controlled clinical trials may not capture the full richness and variability of actual clinical practice and a concern on the part of some that the very process of randomisation may undermine the representativeness of clinical encounter” (Chambless & Hollon, 1998, p. 14) and underline the importance of more research to answer the question of generalization of treatment effects. It is possible that the selection criteria generally applied in efficacy studies leads to homogenous samples with low standard deviations in the applied measures. A small denominator in the fraction calculating the effect size could result in an overestimation of treatment effects in comparison to typical clinical samples. In this case, private practitioners would be well advised to reduce their expectations concerning the effects of treatment that has been proved to be successful in the literature. Thus, further attention must be directed to the possibility that higher effect sizes are achieved in typical research conditions and that these are not due to the quality of treatment but to sample selection and study characteristics. In the present study we will re-examine the current research on social phobia treatment to investigate whether sample restriction and laboratory conditions affect the effect sizes.

3.2. Method

3.2.1. Retrieval of Studies

In a first step, we searched for studies investigating exposure or cognitive behavioural treatment effects for patients with social phobia as primary axis I diagnosis. For this purpose we selected studies for which pre-post effect sizes could be calculated with the provided means and standard deviations for the outcome measures across treatment. We located studies by searching through the reference lists of available studies as well as by using the computer based retrieval system PsycLIT (American Psychological Association, 1994). We used the search terms “social phobia treatment” and “social phobia therapy” to search for journal, book and chapter citations from 1996 to the present 2001. Further studies were located on the basis of the meta-analyses cited above that investigated therapy outcome effects for social phobia. Unpublished studies were retrieved through correspondence with contributers in the field of research on social phobia in Germany.

3.2.2. Study Sample

Thirty studies on cognitive and/or behavioural treatment of social phobia or severe interpersonal anxiety met our criteria for inclusion and were selected for our review. Twenty- two of the studies were listed in one of the meta-analyses referred to above. They were supplemented by six further published and two unpublished studies. Most of these studies investigated treatment effects, many of them comparing different treatments or different orders of treatment components to each other. One study investigated the sensitivity of different questionnaires (Cox, Ross, Swinson, & Direnfeld, 1998) and three studies investigated differences in treatment outcome for different subgroups of patients (Jerremalm, Jansson, & Ost, 1986; Hope, Herbert, & White, 1995; Hofmann, Newman, Becker, Taylor, & Roth, 1995).

Only patient-samples treated with cognitive behaviour therapy that included some form of cognitive restructuring or exposure to feared situations were used to calculate the effect sizes. The treatments applied in the studies included cognitive behavioural group therapy (CBGT) developed by Heimberg, Juster, Hope and Mattia (1995) (Cox et al., 1998; Gruber, Moran, Roth, & Taylor, 2001; Heimberg, Becker, Goldfinger, & Vermilyea, 1985; Heimberg, Dodge, Hope, Kennedy, Zollo, et al., 1990; Heimberg, Liebowitz, Hope, Schneier, Holt, et al., 1998; Hope, Heimberg, & Bruch, 1995; Hope, Herbert, et al., 1995; Otto, Pollack, Gould, Worthington, McArdle, & Rosenbaum, 2000), a combination of exposure and cognitive restructuring (Butler, Cullington, Munby, Amies, & Gelder, 1984; Clark & Agras, 1991; Gelernter, Uhde, Cimbolic, Amkoff, Vittone, et al., 1991; Lincoln, et al., 2002; Maffick & Peters, 1988; Maffick, Peters & Clarke, 1989; Mersch, 1995; Scholing& Emmelkamp, 1993a; Scholing & Emmelkamp, 1993b; Taylor, Woody, Koch, McLean, Paterson, & Anderson, 1997) pure exposure (Alden, 1989; Fava, Grandi, & Canestrari, 1989; Newman, Hofmann, Trabert, Roth, & Taylor, 1994; Turner, Beidel, & Jacob, 1994), exposure and social skills training (Hofmann et al., 1995), personal effectiveness therapy and exposure (Wlazlo, Schroeder-Hartwig, Hand, Kaiser, & Munchau, 1990), social effectiveness therapy (Turner, Beidel, Cooley, Woody, & Messer, 1994), self-instructional training (Jerremalm et al., 1986), social skills training combined with cognitive modification (Stravynski, Marks, & Yule, 1982), rational emotive therapy (Kanter & Goldfried, 1979; Schelver & Gutsch, 1983), and cognitive therapy (Stangier, Heidenreich, Peitz, Lauterbach, & Clark, 2002).

3.2.3. Data Analysis Plan Calculation of effect sizes

As it was our intention to investigate effects of the sample characteristics and not the effects of treatment, subsamples of patients treated with different cognitive or behavioural interventions or different formats (group versus individual) within one study were combined into a single sample. This meant that the effect sizes for different treatment conditions as well as for group and individual therapy within the same study were averaged. We justified this by the fact that meta-analyses (Feske & Chambless, 1995; Ruhmland & Margraf, 2001; Taylor, 1996) failed to find significant differences between cognitive behaviour therapy and exposure or between group and individual therapy. In contrast, subsamples of patients with different characteristics within one study were left as distinct subgroups, thus going into the calculations as separate samples.

We applied the criteria chosen by Feske and Chambless (1995) for the calculation of effect sizes. They were calculated using the formula (Mpretest – posttest / A’/lpretest) and averaged in the case of more than one measure to assess social anxiety. Effect sizes were based on questionnaire self-evaluation measures because clinical ratings have shown to result in larger effects and could lead to an overestimation of the effects in studies using them. The following measures of social anxiety were included: the Fear of Negative Evaluation Scale (FNES,

Watson & Friend, 1969), the Personal Report of Confidence as a Speaker (PRCS, Paul, 1966) the Social Avoidance and Distress Scale (SADS, Watson & Friend, 1969) the Social Phobia and Anxiety Inventory (SPAI, Turner, Beidel, Dancu, & Stanley, 1989), the Social Phobia Subscale of the Fear Questionnaire (Marks & Mathews, 1979), the Social Situations Questionnaire (SSQ, Bryant & Trower, 1974), the Social Phobia and Social Interaction Anxiety Scale (SPS/SIAS, Maffick & Clarke, 1998), the Fear Survey Schedule (FSS, Hallam & Hafner, 1978) and the Unsicherheits-Fragebogen [Uncertainty-questionnaire] (Ullrich de Muynck & Ullrich, 1977), a commonly used and well validated scale in Germany. Categorization of studies

In a second step, we categorized all studies that had provided the necessary information following the guidelines set up by Shadish et al. (1997). We operationalized these criteria by categorizing the studies according to their exclusion criteria, the heterogeneity of their sample and their laboratory study characteristics (see Appendix A). Exclusion criteria.

Exclusion of participants with (a) past or present comorbid substance misuse, past psychosis or bipolar disorder, (b) comorbid depression, (c) comorbid further Axis I disorders, (d) comorbid avoidant personality disorder (APD), (e) a low degree of severity (defined by participants having to reach a certain score in one of the questionnaires or on a rating of severity scale) or (f) prior treatment for social phobia. Heterogeneity of the sample.

(g) Were the majority (more than 60%) of the participants students or academics? (h) Were the participants of the sample heterogeneous in the duration of their disorder (defined by the standard deviation of the mean duration of disorder)? (i) Were they heterogeneous in the severity of their disorder on the questionnaire measures (defined by the standard deviation of the Social Phobia subscale of the FQ [Marks & Mathews, 1979], the FNES and the SADS [Watson & Friend, 1969] as these were the most frequently used measures)? (j) Was the age range limited? (k) Were there qualitative sample restrictions (e.g. investigating only musicians, only patients with comorbid avoidant personality disorder, only generalized or only specific subtypes)? Laboratory characteristics.

(1) Was a large part of the sample recruited by adverts made explicitly for the study? (m) Was the study carried out in a university setting? Because 11 studies had not provided explicit information on this aspect, we decided to categorize these studies according to their reference address. We judged 9 of these studies to have been carried out in a university and 2 to have been carried out in a clinic setting, (n) Were the therapists specifically trained doctoral students or researchers or were they therapists working with normal caseloads? (o) Was a treatment manual used? (p) Was the implementation of the manual strictly monitored? This was assumed if it had been pointed out explicitly in the study.

3.3. Results

3.3.1. Comparison of Studies According to Sample and Laboratory Characteristics

Table 1 shows the mean effect size (ES) and standard deviation (SD) for each group of studies according to the applied sample and study criteria. Because of a number of very small sizes and the assumption that they may be more prone to sample error, studies were weighted with the root of n. The effect-sizes were then compared using /-tests with Bonferoni-adjustment for each comparison separately (p = 0.05/13 = .004). A significant difference in mean effect sizes was found for four of the comparisons. In contrast to expectations, two of these comparisons revealed higher effects for studies fulfilling the criteria for clinical therapy. Samples in which comorbid APD had been included as well as samples including patients with prior treatment experience reached higher effect sizes. However, studies that had excluded patients with comorbid psychosis, substance misuse and bipolar disorder or had been carried through following a treatment manual reached higher effect sizes than those who had not.

To test a possible negative relation of the heterogeneity in age as well as duration and severity of disorder with the effect size, a correlation analysis was carried out. The results are presented in table 2. Studies working with patient samples that were more homogenous in the duration of disorder tended to achieve higher effect sizes and the heterogeneity of the sample concerning the severity of disorder was also negatively related to the effect size. The correlations between the age range and the standard deviation of the mean age and mean effect sizes were lower, with only the age range reaching significance.

3.3.2. Effects of Accumulative Research Characteristics

In order to estimate the accumulative effect of typical research characteristics on the effect size, we calculated a “general research score” for each study. To clarify whether a significant effect can be better explained by sample or by laboratory characteristics, we devided the general score into a “sample restriction score” as well as a “laboratory score”. One point was given for each of the sample selection and laboratory criteria listed in table 1 and these points were then added up. The scores were only calculated for studies that had given information on the variables of interest, because too many missing values could possibly have resulted in an underestimation of restriction criteria applied in the study. For this reason, four of the studies (Cox et al., 1998; Fava et al., 1989; Kanter & Goldfried, 1979; Schelver & Gutsch, 1983) were omitted from the calculation. As information on the amount of students had not been specifically mentioned in many of the studies, this variable was also neglected in the calculation. For the 30 remaining samples the analysis of the general research score with the effect sizes revealed a correlation of r = 21 (two-tailed p <= .01, weighted n = 134). The correlation of the sample restriction score with the effect sizes revealed a correlation of r = .09 (two-tailed p = .28). The correlation of the laboratory score with the mean effect sizes was r = .32 (two-tailed p < .01).

3.3.3. Effect of Sample Selection on the Standard Deviations

To test the hypothesis that sample selection results in a lower standard deviation in the questionnaires, we tested the correlation between the sample restriction score, the age range, the standard deviation of age and of duration of disorder with the standard deviation of the Social Phobia subscale of the FQ (Marks & Mathews, 1979), the FNES and the SADS (Watson & Friend, 1969). Table 3 shows this analysis, revealing only the standard deviation of the duration of disorder to be correlated with the standard deviations of the applied measures. The sample restriction score was not related to a lower standard deviation in the questionnaires and only one of the six correlations of sample variety in age and the standard deviations reached significance, whereas the others did not even reveal a definite tendency.

3.4. Discussion

The main aim of this study was to test the hypothesis that sample selection and laboratory study conditions lead to higher treatment effects in comparison to clinical conditions. We hypothesized that sample restriction would produce homogeneous samples and that this would affect the effect size. We did find the standard deviations of the questionnaire measures at pre-treatment to be positively related to the effect sizes. However, with the exception of the standard deviation of the duration of disorder, our sample restriction criteria were not related to these standard deviations. Generally, there was not much evidence for the assumption that effect sizes might be explained by the failure to gain typical samples. Two of the direct group comparisons even revealed the opposite effect. Patient samples that included patients with comorbid avoidant personality disorder and patients with prior treatment experience revealed higher effect sizes than samples without. Also, the results indicate that even the accumulation of sample restriction does not have any predictive value for the pre-post effect sizes of treatment. On the other hand, we found some results to be in support of the observations made by private practitioners and other psychotherapists working under no research conditions. Firstly, samples excluding patients with comorbid psychosis, substance misuse and bipolar disorder were shown to reach higher effects than those including these patients. The same applied for studies that were conducted following a treatment manual. Secondly, we found studies working with participants who were homogeneous in the length of their disorder to produce higher treatment effects than the other studies. There was strong evidence for the hypothesis that this relation is moderated by the size of the standard deviations in the applied measures, which, in turn revealed moderate to high correlations with the standard deviation of the duration of disorder and moderate correlations with the effect size. Thirdly, and more important than these single findings, is the influence of accumulated laboratory criteria for research studies on the effect size. We found a significant tendency for studies applying laboratory treatment conditions, such as recruiting patients by adverts, applying treatment in university settings, using specifically trained therapists and following and monitoring treatment manuals to achieve higher effect sizes.

It must be noted, though, that the current study is characterized by certain difficulties complicating the interpretation of results. One very small study sample (Fava et al., 1989) revealed a mean effect size of 4.75 thus being far out of the range of the other effect sizes, ranging from 0.3 to 1.8 (see Appendix). However, we considered the size of the effect an insufficient reason for omitting a study. Also, we were interested in securing a large variability of studies. By weighting the studies with the root of n we tried to prevent effect sizes from very small samples from having too much influence on the results. The study by Fava et al., which fulfils most of the criteria for a clinical study, still remains responsible for some of the rather large standard deviations of the mean effect sizes in the group comparisons (see table 1). Without it, more of the comparisons would have had a stronger tendency towards significantly higher effect sizes for the studies applying research characteristics, two more (qualitative sample restrictions and recruiting by adverts) even reaching significance.

The interpretation of the single group differences is complicated by the inter-correlations of the laboratory or sample characteristics. For example the laboratory characteristics are all positive ranging from .25 to .65, suggesting the possibility that some of them might be more relevant than others or that they might cancel each others effects. However, a linear regression analysis of these characteristics with the effect size as dependent variable supported the finding that the use of a treatment manual is the most important predictor of the effect sizes. Generally, significance testing in this study has to be interpreted with caution as the case numbers were artificially raised by the weighting procedure. As an alternative, the size of the differences can be considered, ranging between 20 and 35% for the significant findings.

It could also be argued that sample restriction factors could be confounded with other study factors, that affect pre-post effect size, e g. treatment effectiveness, amount of treatment, type of outcome measure and that we cannot necessarily assume that these are equally distributed across all the comparisons made. However, the amount of treatment was fairly similar across studies and the majority of studies had used more than two outcome measures. Also, by choosing studies with similar treatment approaches we tried to rule out large differences in treatment effectiveness.

Another limitation could be seen in the fact that most of the studies were controlled efficacy studies carried out under typical research conditions. Only a minority of the studies fulfilled many of the criteria listed by Shadish et al. (1997) for being clinically representative. None of the samples were heterogeneous in their focal presenting problems, which was one of the criteria Shadish et al. had set up for clinic therapy. All patients suffered from social phobia or severe social inhibition as primary problem. On the other hand, there is no compelling reason hindering practitioners from treating patients according to their primary diagnosis.

Finally the number of study samples was very small and not all authors had given precise or sufficient information on the variables of interest. This resulted in very low case numbers for some of the comparisons. It may also have resulted in some failures to classify studies correctly (e.g. concerning the place in which the treatment was carried out or the monitoring of the treatment manual).

The optimal conditions in testing the hypothesis would obviously have been a set of about 60 studies all applying the same treatment with varying sample selection and study conditions and noting precise information on these conditions. As this was not the case, we had to make the best of the available studies.

In summary, the data are in line with Shadish et al. (1997), in finding a tendency for studies applying a row of research criteria to reveal slightly lower effect sizes. However, it does not seem to be the impossibility of restricting their samples that could hinder private practitioners from achieving equal effects. It is the accumulation of laboratory characteristics, such as recruiting patients, the place of the study, the training of therapists or the implementation of a treatment manual that correlate positively with treatment effects. These findings give reason to hope, because they imply that researchers as well as practitioners can add to bridging the (small) gap between research and clinical practice. Researchers could try and conduct their treatment research under more natural conditions with health service users. On the other hand, therapists working in clinical practice would be well advised to follow treatment manuals and attend regular disorder specific training or supervision.


Table 3.1.

Mean pre- post effect sizes for social phobia treatment according to sample and laboratory characteristics

  applies   does not apply  
ES (SD) n (N) ES (SD) n (N) P
Sample Restriction Criteria
(a) exclusion of comorbid psychosis, 0.94 (.34) 25 (112) 0.77 (.22) 4 (23) ж
substance misuse or bipolar disorder          
(b) exclusion of comorbid depression 0.91 (.33) 17(71) 0.92 (.32) 12 (63)  
(c) exclusion of comorbid axis I 0.93 (.32) 11 (46) 0.91 (.33) 18(88)  
(d) exclusion of comorbid APD 0.75 (24) 6(25) 0.95 (.33) 23 (109) *
(e) exclusion of low severity 1.03 (.38) 12 (49) 1.00 (.70) 21 (103)  
(f) exclusion of prior treatment 0.71 (.39) 6(32) 1.10 (.63) 26(116) ж
(g) majority of sample are students 1.05 (.40) 11 (47) 1.21 (.99) 7 (42)  
(k) qualitative sample restrictions 1.17 (.39) 10 (40) 0.96 (.67) 23 (112)  
Laboratory Characteristics
(1) patients recruited by adverts 1.03 (.40) 19 (89) 0.98 (.83) 14 (63)  
(m) carried out in a university setting 1.02 (.74) 17 (74) 1.00 (.75) 16 (77)  
(n) using specially trained therapists 1.03 (.31) 19(91) 1.10 (.95) 11 (49)  
(o) following a treatment manual 1.05 (.65) 31 (132) 0.79 (.04) 2(20) *
(p) monitoring treatment manual 1.01 (.36) 16 (72) 1.02 (.77) 17 (79)  

* = p < .004; ES = mean effect size, n = number of samples, N = n weighted by the root of the sample size.



Table 3.2.

Inter correlations Between Effect Sizes and the Heterogeneity of the Sample Concerning Age, Duration of Disorder and Severity

Age range SD Age Ж Duration SD FNES SI) FQ, SP SD SADS

ES -.20* -.19 -.33** -.42** -.45** -.43**

(N = 98)* (N=102) (N = 71) (N = 61) (N = 76) (N = 47)

* = p < .05, **; p < .01; N = number of samples weighted by the root of the sample size; FNES = Fear of Negative Evaluation Scale; FQ, SP = Social Phobia subscale of the Fear Questionnaire; SADS = Social Avoidance and Distress Scale.

Table 3 .3.

Intercorrelations Between Sample Restriction, Sample Variance and the Standard Deviations of the Social Phobia Questionnaires

  Restriction Score Age range Ж Age Ж Duration
Ж» FNES Ж FQ, SP Ж SADS -.15 (N= 47) .19 (N=68) .18 (N= 37) -.22 (N= 31) -.19 (N= 46) .22 (N= 32) -.04 (N= 32) .29 (N= 60)* .15 (#=24) .61 (N= 24)** .47 (N= 51)** .94 (N= 13)**

* = p < .05, **; p < .01; A = number of samples weighted by sample size; FNES = Fear of Negative Evaluation Scale; FQ, SP = Social Phobia subscale of the Fear Questionnaire; SADS = Social Avoidance and Distress Scale.




Effectiveness of an Empirically Supported Treatment for

Social Phobia in the Field[2]

4.1. Introduction

How well do the results of empirically supported treatments hold up in actual clinical practice (Wade, Treat, & Stuart, 1998)? It is often argued on behalf of private practitioners and other psychotherapists working under no research conditions that their patients obviously differ from the research samples and that they therefore do not obtain as good results as those reported in the given literature. Chambless & Hollon (1998) point out “a growing recognition that controlled clinical trials may not capture the full richness and variability of actual clinical practice” (p. 14). Writers have recently begun to distinguish between the efficacy of psychotherapy and its effectiveness (Weisz, Donenberg, Han, & Weiss, 1995). Efficacy (or research therapy) refers to the effects of psychotherapy in randomised, controlled trials, usually conducted in university settings involving recruited patient clients, using a highly structured treatment manual for a narrow problem focus and trying to establish a high degree of internal validity. Effectiveness (or clinical therapy) refers to the effects of natural clinical psychotherapy conducted in the field, which means in private practice or in mental health centres, using quasi-experimental designs and trying to establish a high degree of external validity or generalization of results to various settings.

While the efficacy of psychotherapy is generally well established, the generalization of efficacy findings can be challenged. Weisz, Weiss, and Donenberg (1992) found for child and adolescent therapies that “research focused on more representative treatment of referred clients in clinics has shown more modest effects. In fact, most clinic studies have not shown significant effects” (p. 1578). Recently, Shadish, Matt, Navarro, Siegle, Crits-Christoph, Hazelrigg, et al. (1997) conducted a secondary analysis of past meta-analytic data and found very few studies, which were even remotely clinically representative. For a study to pass as clinical, it had to be carried out in non-university settings, involve patients that were referred through usual clinical routes, involve experienced, professional therapists with regular caseloads and free to use a wide variety of procedures in treatment rather than therapists in training or trained specifically for the purpose. The therapists were also not to have used a treatment manual and the implementation of the treatment was not to have been monitored. Finally, the studies were to have involved clients who were heterogeneous in personal characteristics as well as in focal presenting problems. Only one study fulfilled the authors’ complete set of criteria for clinical therapy. However, studies that fulfilled a certain degree of the criteria revealed effect sizes that were about 10% smaller than those of the complete set of therapy studies. This finding seems to support the doubts of practitioners concerning the transferral of research findings and underlines the necessity of further investigation.

In the area of social phobia there is a large body of support for cognitive behavioural therapy. Four meta-analyses have found average uncontrolled pretest-posttest effect sizes for the reduction of social phobic anxiety ranging from 0.80 (Fedoroff & Taylor, 2001), 0.90 (Feske & Chambless, 1995), 1.06 (Taylor, 1996) to 1.07 (Ruhmland & Margraf, 2001). The mean controlled effect size was found to be 0.84 (Gould, Buckminster, Pollack, Otto, & Yap, 1997). Effect sizes were also high for general anxiety (Ruhmland & Margraf, 2001) but slightly lower for the reduction of depressive symptoms (Feske & Chambless, 1995) after treatment for social phobia.

However, most of the reported studies are characterized by sample selection criteria and thus do not reflect usual patient samples in clinical settings. Typically, the researchers had excluded patients with comorbid major depression, patients with prior treatment, patients outside a certain age range (e.g. 20-50 years), and patients with light to moderate impairment, with many studies even excluding patients with further Axis I disorders. Furthermore, several studies only investigated specific subsamples of patients with social phobia, such as physical reactors, specific subtypes, only musicians or only patients without a partner. All studies were conducted following a treatment manual and most of them involved specifically trained doctoral students and monitored the use of a treatment manual. Also, most of the studies involved patients recruited by newspaper advertisements, often offering free treatment in return for agreeing to take part in the study. Many of the studies were carried through in a university setting, involving mainly student participants. However, in a previous study (Lincoln & Rief, 2002), we found that none of the applied sample restriction criteria resulted in higher effect sizes. The data indicated that involving recruited patients and restricting the variety of the sample in order to achieve a high degree of internal validity did not lead to an overestimation of effects in comparison to more clinically oriented studies. A limitation of the study was that the investigated samples were not clinical in the way defined by Shadish et al. (1997). With exception of the study we are going to present in this article, all of the studies were to be classified as efficacy studies with varying amounts of sample restriction and laboratory study conditions. Thus, generalization studies are needed to explore the transportability of empirically supported treatments to the field of outpatient psychotherapy (Wilson, 1996).

Three recent generalization studies were conducted in Germany. Wetzel, Bents, and Florin

(1999) examined exposure therapy with response prevention for obsessive-compulsive disorder and found results to be comparable with those in controlled studies. Tuschen-Caffier, Pook, and Frank (2001) evaluated the effectiveness of cognitive behavioural therapy for bulimia nervosa. The effect sizes were in the range of those found in controlled research. Similarly Hahlweg, Fiegenbaum, Frank, Schroeder, and von Witzleben (2001) evaluated the effectiveness of individual high-density exposure for panic disorder with agoraphobia and also found the effect sizes to be comparable with the average effect sizes reported by meta- analytic studies of controlled efficacy research.

The only study on social phobia partly studying generalization to clinical practice was a study investigating exposure therapy in general practice (Haug, Hellstrom, Blomhoff, Humble, Madsbu, & Wold, 2000). Although this study qualified as being clinical in the sense that it was carried out in and adapted to clinical conditions, a number of laboratory research aspects remained. More than a third of the participants were recruited by newspaper advertising, all comorbid Axis I diagnoses were excluded as well as treatment for social phobia within the previous six months. Finally, it can be assumed that having to give consent to a randomisation to one of the four treatment groups, which also included medical treatment, could have resulted in further sample selection as reported by Juster, Heimberg, and Engelberg (1995), who found differences between patients who agreed to random assignment to treatment conditions and those who did not. However, the groups responded similarly to cognitive behavioural treatment. So far, no study has tested the hypothesis whether treatment for social phobia can be delivered with the same effectiveness in a clinical setting, in which patients are not recruited by adverts, not randomised to treatment groups or preselected in a way typical for research but are part of the usual referral system and medical routine. The current study is an attempt to investigate the generalization of an empirically supported treatment for social phobia to a clinical setting. The effectiveness of exposure combined with cognitive restructuring will be examined in four outpatient clinics in the community and a large number of experienced and inexperienced therapists and will address the following question: Does an effectiveness study of social phobia treatment deliver results comparable to those of efficacy studies?

The study also investigates the possibility that the effect-size could be enhanced by restricting the sample of patients according to the criteria employed in research settings, by addressing a second question: Which effect does sample selection have on the effect sizes in the current sample?

4.2. Method

4.2.1. Setting

The Christoph-Dornier Foundation for Clinical Psychology (CDS) was founded in 1989 with the aim of promoting research and clinical practice in clinical psychology. The CDS runs seven outpatient clinics in Germany, in which patients with a variety of disorders are treated, in particular patients with anxiety disorders. Patients are referred from different sources; for example, general practitioners, psychotherapists, psychiatric hospitals or they come because they have heard about the CDS. In most cases the patient’s health insurance company paid treatment or part of treatment, but patients had to take the trouble of applying for the reimbursement of expenses.

4.2.2. Participants

Participants were 217 patients who agreed to undergo treatment in one of four CDS outpatient clinics in the cities of Marburg (MB; founded in 1989), Dresden (DD, founded in 1994), Braunschweig, (BS, founded in 1995) and Munster (MS, founded in 1993). All patients were diagnosed with social phobia as the primary disorder with a structured interview (see below) according to the criteria listed in the Diagnostic and Statistical Manual of Mental Disorders (DSM-III-R, 3rd ed., revised, American Psychiatric Association, 1987), meaning that social phobia was judged by the patients to be the most severe disorder and the one for which they wished treatment. Patients were not preselected in any way, with the exception of medical conditions not allowing for high-density exposure treatment. The institute in Marburg contributed 45%, Dresden 26%, Braunschweig 16% and Munster 13% of the participants.

The mean age of the sample was 33.7 years (SD = 10.3, range 12 – 65). Fifty-seven percent of the patients were male, 3% had not completed school, 34% completed secondary school, which compromises the two lower schools in the German school system, 32% completed high school and 31% had a university degree. Sixty percent of the patients were employed, 13% were unemployed, housewives or in retirement and 27% were students or in apprenticeship. Forty-eight percent were married or lived together with a partner, 63% were childless.

The mean age at onset of the disorder was 19.8 years (SD = 9.6). Eighteen percent report the disorder having begun before the age of 13, whereas another 32% report the beginning of the disorder having been during adolescence (13-18 years). The mean duration of disorder was

13.6 years (SD = 11.1, range 0-57). Patients were diagnosed with generalized social phobia if they reported anxiety to be at least moderate in three or more from a list of twelve social situations in the clinical interview and if at least two different situational domains (formal speaking and interaction, informal speaking and interaction, observation of behaviour, and assertive interaction) were represented. Each interview protocol was checked by two raters, who agreed in 88% of the cases and came to a joint decision in unclear cases. Ninety percent were classified as generalized subtype, 10% as specific subtype. The ratings of severity were low (1-3) for 3% of the patients, moderate (4-6) for 64% and high (7-8) for 33% of the patients.

Eighty percent had already undergone therapy: sixty-six percent had undergone some form of psychotherapy, 38% had received professional medical treatment for social phobia and 23% had already been hospitalised in an institution for mental health. Sixty-eight percent were using anxiolytic, antidepressive, neuroleptic or another kind of medication for their anxiety. Forty-four percent had at least one comorbid Axis I disorder. Assessment of Axis II comorbidity was not integrated as a regular part of the diagnostic interview. This limitation is due to financial restrictions set by the insurance companies and a different emphasis at the beginning of data collection.

4.2.3. Treatment

Typically, the patients were treated with high density in vivo exposure supplemented by cognitive interventions. The highly individualized treatment consists of three main phases: Psychological and medical assessment

Psychological assessment (4-6 sessions) is described in detail below. A medical check-up is particularly important in the context of exposure since this can be physiologically stressful and may be contraindicated (e.g. for patients with coronary heart disease). Diagnostic feedback and cognitive preparation

Cognitive preparation for therapy takes place about one week later and is necessary to enhance the patient’s motivation for treatment. The patient’s core assumptions about the aetiology of social phobia are integrated into a model that is able to explain the way in which specific patterns engender and maintain social anxiety. Implications for therapy are then delineated on the basis of this model. Detailed information on the strategies of high-density exposure is provided and in this context the precondition of discontinuation of medication is explained. The patient is given 5-10 days to decide whether to participate in the treatment. The preparation phase is described in detail by Tuschen and Fiegenbaum (1997). It is not considered as actual treatment, but as a preparation for treatment. For this reason, patients {n = 24) who discontinued after this stage are considered as refusers rather than dropouts. High-density exposure combined with cognitive interventions

When the patient decides to participate, exposure and cognitive intervention begin. The program is characterized by short treatment duration, usually lasting about five to seven days. The therapist is in close contact with the patient during the first days, during which it is not unusual for treatment to last for six to eight hours. The intensive treatment phase is followed by a self-control phase of six weeks, in which patients are instructed to continue exposing themselves to the feared situations in their everyday life. The self-control-phase is extensively prepared with the patient and additional support in the form of regular telephone contacts or additional treatment sessions is given when necessary. At the end of the self-control phase, the therapist and patient analyse the progress and the patient is motivated to integrate the interventions more and more into everyday life.

Exposure to the feared situations plays a central role in the therapy as it serves several purposes. It is used to experience a certain degree of habituation to the situation. It also serves to assess and correct core amplifying cognitions as well as safety behaviours and failure focused attention. If possible, an audience used for the exposure situation can also function as giving feedback in order to correct dysfunctional self-perception. The exposure situations are chosen depending on the patients’ individual fears and starting with those feared most.

For example, if one of the most feared items is serving drinks while being observed by other people, the therapist will invite an audience in order to confront the patient with this situation. The exposure is extensively prepared with the patient. Expectations about the way the patient feels he or she is going to be perceived are noted and criteria for success are defined. The patient also decides on which aspects he or she would like to have feedback from the audience. The degree of perceived fear is rated on a scale from 0 (no fear) to 10 (maximum fear). The therapist interrupts the performance to assess the amount of perceived anxiety and instructs the patient to continue until habituation has taken place. A co-therapist videotapes the exercise. The audience is then asked to give the specific feedback defined before the exposure. Finally, the exposure situation and the feedback is discussed with the patient, using it as a natural segue into restructuring interventions in which the patient is taught to identify and challenge specific negative thoughts and general cognitive errors (e.g. because I feel bad, I must be performing badly) and perfectionist thinking (e.g. a less-than-perfect performance is a failed performance). The video feedback is used as an objective feedback and helps to detect safety behaviours.

Generally therapists are free to vary the amount of exposure and cognitive therapy as well as the length of the intervention according to the needs of the individual patient. They are also free to use additional specific interventions for the treatment of comorbid disorders.

4.2.4. Therapists

Treatment was conducted by 57 diploma psychologists (roughly equivalent to a master’s degree; 57% were female, 43% were male) with training in behaviour therapy, who are doctoral students of the CDS. The directors of the respective CDS outpatient clinic supervise treatment extensively. Training in high-density exposure was not delivered in a standardized way and was comparable with procedures described by Wade, Treat and Stuart (1998). Training of novice therapists consisted in reading the relevant literature, viewing videotapes of treatment sessions, attending the supervision sessions and participating as a co-therapist to more experienced therapists or the clinic director in the treatment of at least two patients. Therapists differed in experience: inexperienced therapists (total number of patients treated with any disorder 1-10) treated 22% of the patients, therapists with medium experience (11- 20) treated 43% and experienced therapists (> 21, range 21- 60) treated 35% of the patients.

4.2.5. Measures

Patients were assessed before therapy (pre) with aid of a diagnostic interview as well as an extensive self-report assessment battery, which was also completed 6 weeks after the end of treatment (post). Diagnostic Interview

The diagnosis was determined by a reliable and valid structured clinical interview for DSM- III-R. The Diagnostisches Interview bei Psychischen Storungen (DIPS) [Diagnostic Interview for Psychological Disorders] (Margraf, Schneider, & Ehlers, 1991) is the German version of the Anxiety Disorders Interview Schedule – Revised (ADIS-R, DiNardo, Barlow, Cerney, Vermilyea, Vermilyea, Himadi, et al., 1986). The ADIS-R/DIPS is a semi-structured interview with well-established psychometric properties. The therapists, all of whom had received intensive training in the use and scoring of the instrument, conducted the interviews. The clinical director of the respective outpatient clinic reviewed each case. In difficult cases, a consensus diagnosis was derived jointly. General impairment Symptom Checklist-90-Revised (SCL-90-R, Derogatis, 1994; German version: Franke, 1995). The SCL-90-R is a 90-item questionnaire assessing nine primary symptom dimensions and a Global Severity Index (GSI), based on all 90 items. The GSI is used to measure the intensity of the perceived distress. Internal consistency for the German version of the SCL-90-R is .97. It is frequently used as part of psychotherapy evaluation. Questions on Life Satisfaction (FLZ M, Henrich & Herschbach, 2000; German Version: Henrich & Herschbach, 1996). The FLZ M is a short questionnaire for assessing general and health related quality of life. The questionnaire consists of two eight-item modules, “General Life Satisfaction” and “Satisfaction with Health”. The respondent rates each item twice, once for the subjective importance of the aspects of life or health addressed and once for the degree of satisfaction in that area. The two ratings are combined to a weighted satisfaction score. Internal consistency for the German version is .82 for General Life Satisfaction and .89 for Satisfaction with Health. As the FLZ M was not given to patients from the beginning, calculations can only be made for a smaller sample of n = 65 (FLZ-GA) and w = 73 (FLZ-GG). Social phobia measures The subscale Interpersonal Sensitivity of the SCL-90-R. This scale assesses feelings of social uncertainty as well as fears of being observed or judged negatively. Internal consistency for the German version of the subscale is .86. Social Phobia Scale and Social Interaction Anxiety Scale (SPS/SIAS, Mattick & Clarke, 1998; German Version: Stangier, Heidenreich, Berardi, Golbs, & Hoyer, 1999). The SPS/SIAS is a 40-item self-report questionnaire, consisting of two scales assessing the fear of being observed and evaluated by others as well as interaction anxiety. Internal consistency for the German version is .94 for SIAS and .94 for SPS and sufficient validity data are provided. As the SPS/SIAS was not given to patients from the beginning, calculations can only be made for a smaller sample of n = 117 (SPS) and «=116 (SIAS). Self-rating of impairment due to social phobia. Patients rated on a 5-point rating scale to what extent they felt impaired by their social anxiety in their work, their free time and social activities, and in their family life (0 = not at all, 1 = a little, 2 = moderately, 3 =

severely, 4 = extremely). Related fears and avoidance Body Sensation Questionnaire (BSQ, Chambless, Caputo, Bright, & Gallagher, 1984; German version: Ehlers, Margraf, & Chambless, 1993). The BSQ is a 17-item questionnaire to assess anxiety with regard to bodily symptoms, such as sweating or palpitations, which is common in patients with social phobic fears. This is shown by significant correlations (r =.39) with the SPS (Heinrichs, Hahlweg, Fiegenbaum, Frank, Schroeder, & von Witzleben, 2002). The German version has an internal consistency of 0.85. Agoraphobic Cognition Questionnaire (ACQ), Loss of Control subscale (Chambless, et al., 1984; German version: Ehlers et al., 1993). The ACQ is a 14-item questionnaire to assess anxiety/agoraphobic cognitions. The Loss of Control scale contains some items reflecting typical social phobic fears (e.g. I am going to act foolish). Internal consistency for the German version is .75. The subscale Anxiety of the SCL-90-R. This scale describes physical symptoms of anxiety as well as nervousness, tension and worries. Internal consistency for the German version is .88. Depression Beck Depression Inventory (BDI, Beck & Steer, 1987, German version: Hautzinger, Bailer, Worall, & Keller, 1995). The BDI is a 21-item self-report questionnaire used to assess the severity of depression and common cognitive, affective and somatic symptoms of depression. Internal consistency for the German version is .88 and sufficient validity data are provided. Furthermore, the reliability and validity of the BDI have been specifically affirmed in patients with social phobia (Coles, Gibb, & Heimberg, 2001). The subscale Depression of the SCL-90-R. This scale includes feelings of slight depressiveness as well as symptoms of severe depression. Internal consistency for the German version is .89. Rating of improvement

We used a 7-point rating scale (1 = very much better, 2 = much better, 3 = better, 4 = no change, 5 = worse, 6 = much worse, and 7 = very much worse) to assess the subjective improvement due to the therapy. Patients and therapists rated the degree of improvement six weeks after therapy (post).

4.3. Results

Data analysis was performed in a series of steps. First, treatment completers were compared with patients who dropped out during treatment or those who failed to complete the postassessment. Second, in a preliminary analysis, differences between the four outpatient clinics and between inexperienced and experienced therapists were analysed. In a third step, patients who had completed SPS (n = 85), SIAS (n = 84), FLZ-GG in = 73) and FLZ-GA (n = 65) at pre and post were compared with the rest of the sample to test the possibility of generalizing their results to the complete sample and pre-post comparisons and consumer satisfaction were calculated. Fourthly, we considered effect sizes and the percentages of reliably and clinically significantly improved patients (Jacobson, Follette, & Revenstorf, 1984). Finally, in order to answer the question of whether the effect size is influenced by sample restriction, we compared subgroups characterized by the different exclusion criteria or specific sample characteristics as found in the efficacy studies to the remaining sample.

4.3.1. Comparison of Treatment Completers and Dropouts

Of the 217 patients who agreed to undergo treatment after the cognitive preparation phase, 18 (8%) dropped out during treatment. The following reasons were given for dropping out during treatment: The therapy seemed to hard to endure, the patient was transferred to another institution, the insurance refused to cover the costs, there were organizational difficulties or doubts regarding the rational for the treatment. Another 24 patients (11%) completed the treatment but did not send back the follow-up-questionnaires at post-assessment. Table 1 shows a comparison of pretreatment variables for dropouts, treatment completers with missing follow-ups and treatment completers who participated in follow-ups. Univariate analysis of variance (ANOVA) with Tukey-HSD post hoc tests for continuous variables were used to examine differences between the groups. Post-hoc tests revealed no significant differences between treatment completers and patients with missing follow-ups on continuous variables. However, a comparison of treatment completers with dropouts during therapy revealed two significant differences: First, dropouts scored higher on the SCL-GSI. A more detailed analysis found significant post-hoc differences on the subscales depression (p < .05), phobic anxiety (p< .01) and obsessive compulsive (p < .05). Second, dropouts scored highly on the BDI, with a mean score of 22.7 (SD = 11.7), which indicates a severe level of depression (Hautzinger et al., 1995), compared to patients with missing follow-ups (18.3, SD = 11.6) and treatment completers (14.8, SD= 10.1).

Chi-square-tests were used to examine differences between the groups on categorical variables. Dropouts from therapy were diagnosed significantly (Pearson’s Chi-square = 10.8, df = 2, two-tailed, p < .01) more often with at least one comorbid Axis I disorder than treatment completers or patients with missing follow-ups.

No differences were found between the groups concerning the scores in the questionnaires assessing social phobic fears, marital and educational status, gender, age, severity and duration of disorder, or the amount of prior treatment.

4.3.2. Preliminary Analyses

Analysis of covariance testing for differences between the four outpatient clinics with prescores on the SPS/SIAS and the subscale Interpersonal Sensitivity of the SCL-90-R as covariates yielded nonsignificant results for the post-scores on these measures, indicating that treatment was delivered with the same effectiveness despite the differences in setting, therapists, and supervision. There was also no difference in the duration of treatment between the four clinics, with the mean duration for the complete sample being 35 sessions (each lasting for 50 minutes), including the session for the first contact and 6 sessions for the psychological assessment.

Correlation of the amount of therapist experience with the average effect sizes for the social phobia measures showed no significant effect of experience on therapy outcome (r= .01, p = .873, n = 157).

4.3.3. Treatment Outcome and Consumer Satisfaction

The sub-sample of patients who completed the SPS/SIAS and the FLZ did not differ from the rest of the sample who were not given these questionnaires on the SCL-GSI or any of the SCL-subscales in their response to treatment, so we found it reasonable to generalize the SPS results to the complete sample. Pre-post-comparisons on the questionnaire-measures were calculated for 175 patients who completed therapy and took part in the post assessment using paired sample /-tests with Bonferoni-adjustment for each time comparison separately (p = .05/11 = 005). In table 2 the means, standard deviations with the specific /-value, degrees of freedom and level of significance are presented. Patient scores on all variables decreased highly significantly from pre to post. The same results were achieved for an intent-to-treat- analysis with pre-post comparisons including the complete sample and assuming there had been no change in patients who dropped out of treatment or did not complete the post assessment (see table 2). The questionnaires revealed some overlap, with pretreatment correlations ranging from r = -.15 (FLZ-GA and SCL-AN) to r = .81 (SCL-GSI and SCL-IS). The SCL-GSI revealed the highest correlations with other measures. At postassessment intercorrelations were generally higher, but revealed a similar pattern.

After treatment the patient and the therapist rated improvement on a 7-point rating-scale. At post 51% of the patients rated themselves as being much better or very much better, whereas 70% of the therapists rated their patients to be better or very much better. Forty percent of the patients (25% of the therapists) rated being somewhat better. Six percent of the patients (4% of the therapists) rated being unchanged. Finally, 3% of the patients (1% of the therapists) rated being somewhat worse or much worse. The inter-correlation between ratings by therapists and patients was r = .67.

4.3.4. Intra Group Effect Sizes, Reliable Change, and Clinical Significance

We calculated effect-sizes using the formula (Mpretest-Mposttest)/5Dpretest. According to Cohen (1988), effect sizes for /-tests are categorized as follows: low d > 2, medium d > .5, and high d > .8. Jacobson et al. (1984) propose two necessary conditions a patient has to fulfill for being classified as improved: a) he or she must have moved from a dysfunctional range to the functional range during the course of therapy. This criteria was operationalized using the formular (Mdysf* SDvwc +Mfunc*5Ddysf)/5Z)dysf + SDfunc), defining the cut-off as the point from which it is more likely that a patient has ended up in the functional population than in the dysfunctional population. Means for functional populations were looked up in the test-manuals. For the SPS/SIAS, we used the data from a normal population of n = 80 that had been collected in the Christoph-Dornier-Foundation from control-groups in other studies. This comparison group reached a mean of M = 10.69 (SD = 9.01) in the SPS andM= 18.36 (SD = 8.56) in the SIAS. b) there must have been change during the course of therapy. Here, the Reliable Change Index (RCI) was applied, with RCI = (Mpretest – Mposttest)/®1, with SE = 5Y)pretest yjl-rxx’, where rxx’ is the reliability of the measure. According to Jacobson et al. (1984) a patient is categorized as improved if the RCI is higher than 1.96 and as detoriated if the RCI is lower than – 1.96.

Table 3 shows the results for the outcome variables according to the different criteria. At post assessment effect sizes ranged from .71 to .88. on the social phobic measures (SCL- Interpersonal Sensitivity, SPS/SIAS). They ranged from 0.39 to .89 for general impairment (SCL-GSI, FLZ-GA, FLZ-GG). The effect sizes ranged from 0.70 to 0.78 for related fears (SCL-Anxiety, BSQ, ACQ-Loss of Control) and from 0.58 to 0.68 for depression (SCL- Depression, BDI).

Next, using each outcome measure, the percentage of persons demonstrating reliable improvement or detonation was calculated. On average 56% of the patients were reliably improved on social phobic fears, 41% on related fears and avoidance, 48% on general impairment and 41% on depression. However, 2% of the patients deteriorated in their social phobic fears after the treatment.

The percentage of patients more likely to be drawn from a functional population was calculated for each outcome measure before and after treatment for the sample of patients who completed post-assessment (n = 175). Considering social phobic fears 57% were now more likely to be drawn from a healthy sample, the percentages were 66% for depression, 54% for general impairment and 64% for related fears respectively.

The ratings of impairment in important areas of everyday life provide a final source to estimate clinical significance. Thirty percent still rated themselves as being severely or very severely impaired at work (in comparison to 87% before therapy). Twenty percent still felt impaired during their free time (64% before therapy) and 5% remained feeling impaired in their family (33% before therapy).

4.3.5. Effects of Sample Selection

We calculated differences between the mean effect sizes of the social phobia outcome measures (SCL-Interpersonal Sensitivity, SPS and SIAS) of subgroups characterized by exclusion criteria or sample characteristics that had been applied in the investigated outcome studies in contrast to subgroups for which these criteria did not apply. Common criteria consisted of a) excluding comorbid depression, b) excluding patients with prior psychological treatment for social phobia, c) excluding patients with a severity of disorder below 4 in the DIPS 1-8 rating-scale, and d) excluding patients older than 50 or younger than 20. Further frequently found characteristics were e) excluding comorbid Axis I diagnosis, f) using samples consisting mainly of students, g) only treating specific subtypes of social phobia, or h) only cognitive reacting patients (in contrast to physical reactors). Table 4.4. shows the differences in mean effect sizes of subgroups according to the applied criteria. Using two- tailed /-tests we found only one significant difference that was not, however, in support of the hypothesis that exclusion criteria lead to higher effect sizes. The group of patients with a BDI of 18 or above revealed a higher mean effect size than the rest of the sample. Also, the accumulated application of common exclusion criteria did not lead to higher effect sizes. A sample of patients characterized by a BDI-score below 18, no prior treatment experience for social phobia, a severity of at least 4 in the DIPS rating and aged 20 to 50 did not differ significantly from the remaining sample of patients (see table 5) in the way they responded to treatment.

4.4. Discussion

The questions addressed in this study were whether an effectiveness study of social phobia treatment delivers results comparable to those of efficacy studies and whether sample selection and study characteristics would have resulted in higher effect sizes.

To test whether our sample differed from research samples on relevant pretreatment variables, we compared it to samples in 30[3] comparison studies testing cognitive behavioural and exposure therapy that we had investigated in a previous study (Lincoln & Rief, 2002). The mean age in our sample was 34 compared to 35 (SD = 4.3) as mean age of the comparison studies. Forty-three percent were married or with a partner compared to 43% (SD = 19.6) in the comparison studies. There were slightly more men (57%) than in the comparison studies (51%, SD = 13.5). The duration of disorder of 13.6 years is slightly lower than the mean of 17 (SD = 6.4) in the comparison studies. About half of the patients suffered from comorbid disorders, which is characteristic of patients with social phobia (Magee, Eaton, Wittchen, McGonagle, & Kessler, 1996). A majority of patients (66%) had already received some kind of psychological treatment prior to the treatment in the CDS, which is comparably high in comparison to epidemiological findings (Magee et al., 1996). A direct comparison of comorbidity and prior treatment experience with the comparison studies was not possible because of imprecise description in many of the studies and the fact that comorbidity and prior treatment were frequent exclusion criteria.

The mean pretreatment-score on the SPS (M= 38) was higher than the mean score ofM = 31 (SD = 4.9) in comparison studies using SPS or SIAS (Cox, Ross, Swinson, & Direnfeld, 1998; Gruber, Moran, Roth, & Taylor, 2001; Heimberg, Liebowitz, Hope, Schneier, Holt, Welkowitz et al., 1998; Mattick, Peters, & Clarke, 1989; Otto, Pollack, Gould, Worthington, McArdle, Rosenbaum et al., 2000; Stangier, Heidenreich, Peitz, Lauterbach, & Clark, in press). The score in the SIAS (M = 40) was comparable to the SIAS score in the comparison studies (M = 41, SD = 4.3). The same accounts for the mean BDI score (M = 15) compared to a mean of M= 14.5 (SD = 2.9) in comparison studies using the BDI (Cox et al, 1998; Gruber et al., 2001; Heimberg, Dodge, Hope, Kennedy, Zollo, & Becker, 1990; Jerremalm, Jansson, & Ost, 1986; Stangier et al., in press). A specific comparison with German outcome studies (Stangier et al., in press; Wlazlo, Schroeder-Hartwig, Hand, Kaiser & Munchau, 1990) also yielded no major differences.

To summarize, the characteristics of the investigated, unselected group of social phobic patients were similar to treatment-groups reported in the literature, with the exception of a slightly higher percentage of men, a slightly higher score in the SPS, and possibly a higher ratio of patients with generalized social phobia, which might also be due to the rather liberal criteria applied for subtype discrimination. The duration of disorder for the patients in our study was shorter than the mean duration in the comparison studies, but longer than in the German comparison study (Wlazlo et al., 1990). Thus, it is possible that people suffering from social phobia in Germany do not wait as long before they seek help as patients in the United States, where most of the comparison studies were conducted. The average of 28 treatment sessions was higher, but in the range of the average 22 sessions in the comparison studies (SD = 9). It must be noted though, that many of these are group treatments and individual treatments typically consisted of fewer sessions (M = 17, SD = 8), making the number of sessions in our study appear definitely higher. Possibly, additional treatment sessions were needed in our study to attend to comorbid disorders. However, it is important to keep in mind that some patients in controlled outcome studies with fixed numbers of sessions were offered additional treatment after post-assessment, which might have lead to further improvement not reflected in the pre-post effect sizes of these studies.

The present study fulfils most of the criteria for a clinically representative study as defined by Shadish et al. (1997): (a) treatment was conducted in a non-university setting, (b) involved patients referred through usual clinical routes, (c) used patients heterogeneous in personal characteristics, (d) therapists did not use a treatment manual, (e) therapists were free to use a variety of procedures and were not restricted to a fixed number of sessions and (e) implementation of the treatment manual was not monitored.

Two criteria were not met: (f) homogenous patients with regard to primary diagnosis (social phobia) were included instead of patients heterogeneous in focal presenting problems, and (g) only about 50% of the therapists can be regarded as experienced and the majority of therapists were still in their post-graduate 5-year psychotherapy training. Also, the therapists were doctoral students, which is more typical of efficacy studies. However, the therapists did work with regular caseloads and did not receive training specifically for the research study. On top of this, as Hahlweg et al. (2001) also pointed out, using experienced therapists only may not be a valid criteria for clinically representative studies, because there are varying levels of expertise among therapists working in institutions such as community mental health centres or psychiatric in-patient facilities. Therefore, from our point of view, the present study can be regarded as clinically representative.

The outcome results six weeks after the end of treatment for patients completing the intervention provide support for the clinical effectiveness of exposure combined with cognitive interventions for patients with social phobia. Fifty-six percent of the patients were reliably improved on social phobic fears and 57% were more likely to be drawn from a healthy population sample six weeks after the end of therapy. The rate of patients who felt impaired in important areas of their life dropped significantly, indicating that patients succeeded in transferring the effects of therapy into their every-day-life. The mean effect size for the measures of social phobia was 0.82, thus being within, but at the bottom range of the effect sizes reported in the meta-analyses (Fedoroff & Taylor, 2001; Feske & Chambless, 1995; Gould et al, 1997; Ruhmland & Margraf, 2001; Taylor, 1996).

No higher effect sizes were attained when the sample was restricted, applying frequently used selection criteria. In the contrary, more depressed patients profited more. Even the accumulation of common restriction criteria did not result in a higher effect size. Thus, the absence of sample restriction in this study could not be made responsible for the slightly lower effect size in comparison to the meta-analyses. It also seems unlikely that the slight reduction of the effect size can be explained by sample differences. An explanation could be that most comparison studies are based on the Fear of Negative Evaluation Scale (Watson and Friend, 1969) or the Fear Questionnaire (Marks & Mathews, 1979) which have been reported to have larger treatment sensitivity, resulting in larger effect sizes than the SPS/SIAS (Cox et al. 1998). For a direct comparison with studies using the SPS/SIAS, we calculated the effect sizes based on SPS/SIAS using the formular Mpretest – MposttestASDpretest for the six other outcome studies mentioned above that had applied either SPS or SIAS or both measures at pre- and posttreatment. These studies achieved a mean of effect size 0.63 (SD = .35) for SPS and 0.51 (SD = 21) for SIAS which is lower than the ones achieved in the current study, being 0.88 and 0.86 respectively. None of the 29 comparison studies used the Interpersonal Sensitivity scale of the German version of the SCL-90-R, so that a direct comparison was not possible here. However, five studies (Heimberg et al, 1998; Mersch, 1995; Scholing & Emmelkamp, 1993a and 1993b; Stangier et al., in press) did use some form of the SCL-90-R or specific subscales. Effect sizes based on these scales reach a mean effect size of 0.55, which is also lower than the effect size of 0.71 that we found for the SCL-90-GSI. In the light of these findings it seems reasonable to conclude that the effect size found in the present study is comparable with the mean effect sizes found in the meta-analyses.

Given the large sample size and the number of therapists and institutes involved, it also seems justified to conclude that exposure combined with cognitive intervention can be transported to the treatment of patients with social phobia in natural settings, without reducing its effectiveness. Additionally, we found therapist experience to be unrelated to outcome, which is in line with other findings summarized by Bickman (1999), who pointed out the necessity of conducting such studies in a natural environment.

One shortcoming of the present study is the amount of patients (11%) who could not be motivated to complete the follow-up questionnaires at post assessment. We found a tendency for them to occupy an intermediate position between the completers and the dropouts, who differed significantly from one another on some of the measures. Thus, the question can be raised whether this group of patients differs from the sample of completers concerning the effectiveness of the treatment. On the other hand, 9 of the 24 patients with missing follow-ups agreed to give a rating of improvement, with 7 (77%) rating themselves as better or much better (compared to 52% of the completers), which suggests that they improved at least equally. The amount of missing follow-ups can be explained by the fact that in three of the outpatient clinics, there was no financed personnel to organize the follow-ups. Only 8% actually dropped out of treatment, which is low compared to the outcome studies. Reasons for this can be suspected in the cognitive preparation phase, after which some patients with major concerns about the treatment concept decided not to participate and in the higher binding commitment because of the intensive format and the trouble taken for reimbursement of treatment costs. If the rate of missing follow-ups is added to the rate of dropout it sums up to 19%, which is still in the range of the 29 comparison studies, with a mean dropout-rate of 16% (SD = 7.6). The higher depression scores and comorbidity found for patients who drop out of therapy underline the necessity of giving further attention to this group of patients.

Unfortunately, questionnaires defined specifically for the assessment of social phobia (like the SPS and SIAS) as well as the regular assessment of Axis II comorbidity were not part of the diagnostic assessment from the beginning, because of different priorities at the beginning of data collection. Clearly, the Interpersonal Sensitivity scale of the SC1-90-R is not an optimal measure of social phobia as it has not been explicitly validated with social phobic individuals. However, it tends to correlate highly with SPS and SIAS, both in this study (r = .65) as well as in a large validation study for SPS and SIAS including 357 patients (Heinrichs et al., 2002). Another limitation of the study is that it is based entirely on self report measures. Independent blind assessor ratings are missing – and should be included from a methodological point of view. In the current setting as well as in other clinical settings with no extramural funding and depending on the insurance companies, it is impractical and too expensive to provide such ratings with hired experienced raters. Finally, it must also be pointed out that recent data from randomised controlled trials suggest a high placebo response rate in social phobia (Fedoroff & Taylor, 2001; Taylor, 1996) and thus the use of pretreatment expectancy measures might have provided helpful information.

Nevertheless, the present study provides convincing evidence that empirically validated treatment for social phobia, the combination of exposure and cognitive restructuring, can be transported into natural field settings. The results were achieved using a large number of patients and therapists, which underscores the generalization of the results to other settings and can be added to the list of cumulative evidence for the generalization of research therapy to clinical settings. However, it is most likely that these results require not only a thorough diagnostic procedure to assess social phobia as the primary problem but also frequent and intensive supervision of the therapists.


Table 4.1.

Pretreatment Means of Dropouts During Treatment (DT), Patients with Missing Follow-ups

(MF) and Treatment Completers (TC).


// = 18 w = 24 n= 175

  M SD M SD M SD F df P
Age 28.9 9.4 31.6 9.0 34.5 10.4 2.95 2,214 .06
Duration 8.5 5.3 13.1 11.9 14.2 11.3 2.09 2, 201 .13
Severity 6.4 1.1 6.5 1.1 5.8 1.3 3.04 2, 157 .05[4]
SCL-GSI 1.31 0.65 1.14 0.53 0.93 0.60 4.18 2,210 ,02*a
IS, SCL-90-R 1.90 0.90 1.76 1.00 1.51 0.93 2.12 2,210 .10
SPS 44.7 21.3 44.7 17.5 37.1 16.7 1.90 2, 114 .16
SIAS 42.4 16.0 44.0 12.5 39.6 16.1 0.61 2, 113 .54
BDI 22.7 11.7 18.3 11.6 14.8 10.1 5.32 2, 207 .01*6



Table 4.2.

Means, Standard Deviations and paired t Tests for Clinical Outcome Measures



SD Post


SD df t P
Analysis for sample with completed post-assessment (n SCL-GSI 0.94 0.61 0.51 0.49 = 175)


11.83 .000**
FLZ-GA 25.7 33.0 38.6 33.5 64 -4.12 .000**
FLZ-GG 26.5 31.0 54.4 42.7 72 -6.60 .000**
SCL-IS 1.52 0.95 0.85 0.81 159 11.42 .000**
SPS 37.7 17.1 22.2 16.7 84 10.30 .000**
SIAS 40.0 16.5 25.9 15.6 83 10.60 .000**
BSQ 2.20 0.69 1.70 0.56 148 10.32 .000**
ACQ-KV 2.50 0.76 1.91 0.72 146 10.22 .000**
SCL-A 1.23 0.84 0.64 0.67 157 10.60 .000**
BDI 14.8 10.3 7.8 8.3 154 10.88 .000**
SCL-D 1.19 0.89 0.67 0.72 160 9.58 .000**
Intent-to-treat-analysis (n = 217) SCL-IS 1.57 0.94 1.07 0.92 213 10.41 .000**
SPS 38.6 17.2 27.3 18.8 116 8.89 .000**
SIAS 40.4 15.6 30.2 16.5 115 9.01 .000**
BDI 15.9 10.7 10.7 10.35 209 9.94 .000**

** = p < .004; SCL-GSI = Symptom Checklist-90-Revised, Global Severity Index; FLZ-GA = Questions on Life Satisfaction, general life satisfaction; FLZ-GG = Questions on Life Satisfaction, satisfaction with health; SCL-IS = Symptom Checklist-90-Revised, Interpersonal Sensitivity; SPS = Social Phobia Scale; SIAS = Social Interaction Anxiety Scale; BSQ = Body Sensation Questionnaire; ACQ-KV = Agoraphobic Cognition Questionnaire, Loss of Control; SCL-A = Symptom Checklist-90-Revised, Anxiety; BDI = Beck Depression Inventory; SCL-D = Symptom Checklist-90-Revised, Depression.



Table 4.3.

Intragroup Effect Sizes (IGES), Percentage of Patients with Reliable Change (RC), Deterioration (D), Improvement (I) or Maintenance (M) and Clinical Significance Cut-off- Score (CS) with Percentage of Patients in Healthy Population for Clinical Variables

RC-POST Healthy Population

    M in % D in % I in % CS % pre % post
SCL-GSI 0.71 31.6 5.7 62.7 0.51 31 62
FLZ-GA 0.39 66.2 4.6 29.2 43 30 42
FLZ-GG 0.89 41.1 6.8 52.1 48 19 58
SCL-IS 0.71 55.6 1.3 43.1 0.73 24 56
SPS 0.88 28.2 3.5 68.2 20 17 55
SIAS 0.86 42.9 0.0 57.1 26 26 59
BSQ 0.73 73.2 0.7 26.2 2.0 46 74
ACQ – KV 0.78 39.7 6.2 54.1 2.0 34 55
SCL-A 0.70 55.6 1.3 43.1 0.84 41 62
BDI 0.68 51.6 3.2 45.2 9.3 35 69
SCL-D 0.58 57.8 5.0 37.3 0.64 35 62

SCL-GSI = Symptom Checklist-90-Revised, Global Severity Index; FLZ-GA = Questions on Life Satisfaction, general life satisfaction; FLZ-GG = Questions on Life Satisfaction, satisfaction with health; SCL-IS = Symptom Checklist-90-Revised, Interpersonal Sensitivity; SPS = Social Phobia Scale; SIAS = Social Interaction Anxiety Scale; BSQ = Body Sensation Questionnaire; ACQ-KV = Agoraphobic Cognition Questionnaire, Loss of Control; SCL-A = Symptom Checklist-90-Revised, Anxiety; BDI = Beck Depression Inventory; SCL-D = Symptom Checklist-90-Revised, Depression.



Table 4.4.

Criteria applied Remaining


Mean ES n Mean ES n df t p

Mean Pre – Post Effect Sizes for Subsamples with Different Exclusion Criteria or Sample Characteristics

Exclusion of: (a) BDI ^ 18 0.76 (0.72) 103 1.28 (1.01) 53 79.8 3.29 .00
(b) prior treatment 0.70 (0.70) 51 0.84 (0.75) 105 154 1.11 .27
(c) low severity (DIPS 1-3) 0.92 (0.80) 101 0.66 (0.68) 19 118 1.33 .19
(d) age >20 or < 50 years 0.77 (0.73) 142 0.85 (0.62) 19 159 .47 .64
Combined a, b, c, d 0.65 (0.67) 30 0.83 (0.73) 131 159 1.27 .21
(e) further Axis I disorders 0.77 (0.68) 97 0.89 (0.80) 64 159 1.07 .29
Sample consists of: (f) only students 0.74 (0.74) 35 0.85 (0.71) 117 150 .82 .41
(g) only specific subtype 1.07 (1.09) 15 0.76 (0.71) 131 15.4 1.10 .29
(h) only cognitive reactors 0.76 (0.71) 79 0.74 (0.64) 54 131 0.19 .85

p = two-tailed significance




Who Comes, Who Stays, Who Profits? —

Predicting Refusal, Dropout, Success, and Relapse in the

Treatment of Social Phobia5

5.1. Introduction

The absence of data addressing characteristics of patients who refuse treatment, who drop out or who do not improve from treatment is a major limitation of treatment outcome literature. In the area of social phobia a series of meta-analyses has found a high efficacy of cognitive behavioural treatments, with mean effect sizes ranging from 0.8 to 1.1 (Gould, Buckminster, Pollack, Otto, & Yap, 1997; Fedoroff & Taylor, 2001; Feske & Chambless, 1995; Ruhmland & Margraf, 2001; Taylor, 1996). However, not all patients benefit from the tested treatment approaches. Turner, Beidel, Wolff, Spaulding, and Jacob (1996) calculated treatment success taking into consideration not only patients who completed treatment, but also those who were offered treatment, but refused or dropped out of it. This resulted in an alarmingly low rate of 52% of the patients seeking treatment for social phobia who actually profited from it.

Knowing about factors that are responsible for attrition as well as for failure to benefit from treatment may help to understand the processes underlying treatment and enable the therapist to adapt treatment procedures, delivery and planning accordingly to improve a specific patient’s prognosis (van Minnen, Amtz, & Keijsers, 2002). Also, knowledge of prognostic features can be helpful in indicating treatments of choice, since a variety of effective treatment variations are available. Table 1 gives an overview of the findings in 18 studies investigating prognostic factors of refusal, dropout, gain or endstate functioning in the treatment of social phobia. However, the literature review points to a number of limitations in the current state of predictor research for social phobia.

It becomes clear that most attention has been directed to the questions of treatment success, rather than dropout, refusal or relapse after treatment. In fact, refusal and relapse have been thoroughly neglected in prediction research. Research investigating dropout during treatment yields some evidence indicating that higher pretreatment severity and impairment might be causing some patients to drop out of treatment. This hypothesis will have to receive further attention, as it has important implications for treatment delivery.

Refusal. Two studies investigated characteristics of patients who refused to enter the treatment protocol, which meant agreeing to random assignment. Turner et al. (1996) found patients who refused random assignment (15.5%) in a study testing the effects of atenolol, flooding and pill placebo to be less severely impaired, but found no differences on sociodemographic variables, comorbidity or subtype. Juster et al. (1995) found patients who refused random assignment (33%) more likely to be married, not to live alone and to have more income. They found no differences on other socio-demographic variables or in the response to cognitive behavioural treatment.

Dropout. We found four studies that had investigated prediction of dropout, with most of the investigated variables showing no predictive value. Participants with a lower expectancy towards treatment were found to drop out more often as well as more impaired patients.

Relapse. The least attention has been given to the question of relapse or failure to maintain treatment gain after termination of treatment. Only one study (Mersch et al., 1991) addressed this question and found patients who relapsed after postassessment to be older as well as to have had significantly lower SCL-90 scores at pretest.

Change. Most studies have concentrated on predicting change caused by treatment and endstate functioning, focusing on sociodemographic and biographical variables, impairment, severity, subtypes, and comorbidity as potential predictors for change or endstate functioning. The majority of findings are insignificant and most of the significant effects are low.

It seems that demographic and biographical variables generally have little to offer in the way of predicting treatment outcome. The current research has not fully clarified the prognostic value of pretreatment severity and impairment. Studies investigating the impact on treatment change, rather than endstate-functioning, provide contradictory results. We found no studies considering physical anxiety symptoms as a variable of impairment or severity although these might be more resistant to change than cognitions and thus have a negative impact. Studies examining the predictive value of subtypes generally found patients with generalized social phobia to begin and end treatment with more severe symptoms, but to have a similar rate of improvement to the nongeneralized subtype. One difficulty of testing subtypes as a predictor for treatment response is the absence of clear diagnostic criteria. This has led different authors to use different criteria to distinguish subtypes, making a comparison of results difficult. Some authors point out that the categories are somewhat arbitrarily imposed on a continuum of impairment and suggest using degree of impairment as a continuous measure (Chambless et al., 1997; Stein, Torgrud, & Walker, 2000). Patients without comorbid axis I or axis II disorders have often been found to have lower post-scores on measures of general anxiety and clinical severity but the same rate of improvement as patients without a comorbid disorder. Research on the impact of avoidant personality disorder on treatment outcome has yielded contradicting results which might partly be due to still unsolved conceptional difficulties in the distinction between the generalized subtype of social phobia and avoidant personality disorder. However, there is convincing evidence in support of the hypothesis that comorbid depression is a negative predictor of change. Chambless et al. (1997) found a correlation of r = Al between depression and residual gain by postassessment. Also, in a study with 1027 respondents, DeWit, Ogborne, Offord, and MacDonald (1999) found the probability of recovery without undergoing treatment to be three times as high when participants reported no additional depression. The amount of research on health related predictor variables was meagre, yielding some evidence for a possible negative impact of chronic health problems (De Wit et al., 1999). Also, Mersch et al. (1991) found a tendency for a negative impact of the use of alcohol and medication. None of the studies investigated the effect of therapist variables such as gender or years of treatment experience or treatment duration on response. However, Feske and Chambless (1995) analysed the effect of treatment duration in their meta-analysis and found a larger number of exposure sessions to produce more favourable outcomes.

Apart from the absence of a row of promising variables, the available studies are limited by the fact that predictors were studied in the context of controlled outcome studies whose inclusion criteria are likely to limit the variability of the factors studied as predictors. Steketee and Shapiro (1995, pp. 341) point out, that “to better serve our client populations, research on predictors should be conducted on naturalistic clinical treatments, as well as on controlled trials”. Specifically in the case of treatment refusal the question must be raised whether refusal of participating in a study with random assignment can be compared to the refusal to take up an (individualized) treatment offer as such.

The first aim of this study is therefore to search for predictors of treatment acceptance, attrition, effectiveness, and relapses after treatment for social phobia in a field treatment outcome study in four outpatient clinics and using a large sample of unselected patients. The second aim is to compare these predictors with variables identified as predictors in the context of controlled efficacy studies.

5.2. Method

5.2.1. Setting

The Christoph-Dornier Foundation for Clinical Psychology (CDS) runs seven outpatient clinics in Germany, in which patients with a variety of disorders are treated. Patients were referred from different sources, for example, general practitioners, psychotherapists, or psychiatric hospitals. Most of the treatments were paid by the patient’s insurance company by reimbursement of expenses. This means that invoicement for treatment sessions is directed to the patient, who can apply for reimbursement with his or her health insurance company. The insurance company is free to decide whether they are prepared to cover the expenses for treatment or not. This decision process mostly takes place after diagnostic assessment, as the health insurances usually expect a brief report of the disorder and a treatment plan as a basis for their decision. Additional treatment-expenses, such as accommodation, tickets, etc. were generally not covered by the health insurance. Therapists were 62 diploma psychologists (roughly equivalent to a master’s degree; 58% were female, 42% were male) with training in behaviour therapy.

5.2.2. Treatment

Typically, patients were treated with in vivo exposure combined with cognitive interventions. The intensive treatment program is characterized by a short duration, usually lasting about 5-7 days, during which the patients are expected to confront the feared situations for several hours per day. It consists of three main phases:

Psychological and medical assessment. Psychological assessment (4-6 50-minute sessions) consists of conducting a reliable and valid structured clinical interview according to the criteria listed in the Diagnostic and Statistic Manual of Mental Disorders (DSM-III-R, 3rd ed., revised, American Psychiatric Association, 1987). The “Diagnostisches Interview bei Psychischen Storungen” [Diagnostic Interview for Psychological Disorders] (DIPS; Margraf, Schneider, & Ehlers, 1991) is the German version of the Anxiety Disorders Interview

Schedule – Revised (ADIS-R; DiNardo, Barlow, Cemy, Vermilyea, Vermilyea et al., 1986). The ADIS-R/DIPS is a semi-structured interview with well-established psychometric properties. A medical check-up is particularly important in the context of exposure since this can be physiologically stressful.

Diagnostic feedback and cognitive preparation. Cognitive preparation for therapy takes place about one week after assessment and aims at enhancing the patient’s motivation for treatment. The patient’s core assumptions about the aetiology of social phobia are integrated into a model that is able to explain the way in which specific patterns engender and maintain social anxiety. Implications for therapy are then delineated on the basis of this model and patients are encouraged to discontinue medication. The patient is given 5-10 days to decide whether to participate in the treatment. The preparation phase is described in detail by Tuschen and Fiegenbaum (1997).

High-density exposure combined with cognitive interventions. When the patient decides to participate, exposure and cognitive intervention begin (duration is variable and depends on the individual patient’s needs). Exposure to the feared situations plays a central role in the therapy as it enables the patient to experience a certain degree of habituation and helps the therapist to detect and correct core amplifying cognitions, safety behaviours and failure-focused attention. Exposure is combined with restructuring interventions in which the patient is taught to identify and challenge specific negative thoughts, general cognitive errors and perfectionist thinking. At the end of the intensive treatment-phase patients are instructed to continue exposing themselves to the feared situations in their everyday life and are offered further support if necessary. A more detailed description of the treatment concept is given in a former article (Lincoln, Rief, Hahlweg, Frank, von Witzleben, Schroeder et al., 2002).

5.2.3. Participants

Participants were 287 patients who were diagnosed with social phobia as the primary disorder according to the criteria listed in DSM-III-R (American Psychiatric Association, 1987), meaning that social phobia was judged by the patients to be the most severe disorder and the one for which they wished treatment. Fifty-six percent of the patients were male. The average age was 33.9 years (SD = 10.5) and the average duration of disorder was 13.8 years (SD = 11.7). Eighty-one percent had already undergone some form of psychotherapy or medical treatment, 24% had been hospitalised due to mental problems. Thirty-nine percent were married or living with a partner, 33% had completed secondary school, 33% had a high school degree and 34% a university degree.

Of these 287 patients, 241 came to the cognitive preparation session and 217 decided to begin treatment, of which 199 completed it. Treatment effectiveness has been described in detail in a former study (Lincoln et al., 2002). A total of 175 patients completed the post assessment and 101 completed a one-year follow-up. Figure 1 displays this attrition process. The high number of missing follow-ups is due to financial restrictions. In three of the institutes there was no financed personal to organize follow-ups and patients could not be paid to complete the questionnaires.

5.2.4. Measures Predictors

Demographic and biographical variables. Age, age at onset, duration of disorder, prior treatment experience, gender, marital status (0 = married, 1 = living with partner, 2 = partnership, 3 = single), and educational level (0 = no school degree, 1 = secondary modem school, 2 = advanced secondary school, 3 = A-level, 4 = university degree) were collected with the aid of an application questionnaire, which was completed by all patients.

Severity and Impairment. Patients rated their subjective feeling of impairment on a five-point rating scale (0 = not at all, 1 = a little, 2 = moderately, 3 = severely, 4 = extremely). The intensity of the perceived distress was measured with the SCL-90-R, Global Severity Index (SCL-GSI; Derogatis, 1994; German version: Franke, 1995), which is based on all 90 items of the SCL-90-R assessing nine primary symptom dimensions. Internal consistency for the German version of the SCL-90 is .97. The therapists rated the severity of the disorder on a scale from 0-8 as a result of the diagnostic interview (DIPS). Subjective symptom severity was assessed with the Social Phobia Scale and the Social Interaction Anxiety Scale (SPS/SIAS, Mattick & Clarke, 1998; German Version: Stangier, Heidenreich, Bemardi, Golbs, & Hoyer, 1999). The SPS/SIAS is a 40-item self-report questionnaire, consisting of two scales assessing the fear of being observed and evaluated by others as well as interaction anxiety. Internal consistency for the German version is .94 for the SIAS and .94 for the SPS. As the SPS/SIAS was not given to patients from the beginning of the study, calculations can only be made for a smaller sample of n = 85 (SPS) and n = 84 (SIAS). Physical symptoms a patient generally experienced during a social situation were assessed in the DIPS, the Body Sensation Questionnaire (BSQ; Chambless, Caputo, Bright, & Gallagher, 1984; German version: Ehlers, Margraf, & Chambless, 1993) and the Beck Anxiety Inventory (BAI; Beck & Steer, 1993; German Version: Ehlers & Margraf, in press). The BSQ is a 17-item questionnaire to measure anxiety with regard to bodily symptoms, with an internal consistency of 0.85 for the German version. The BAI was used to assess physical arousal symptoms. Although originally developed to measure symptoms of anxiety in general, recent research supports the view that the BAI is more sensitive to panic related symptoms than to other aspects of anxiety, such as worry and tension (Antony, Purdon, Swinson, & Downie, 1997).

Subtypes. Subtypes were considered on a continuum of the amount of 13 social situations in the DIPS, in which the patient had described anxiety as being at least moderate (0 = no anxiety, 1 = slight anxiety, 2 = moderate anxiety, 3 = severe anxiety, 4 = extremely severe anxiety) as well as the total score of anxiety for all these situations.

Comorbidity. Comorbid disorders were diagnosed based on the information in the DIPS. Additionally, patients completed disorder specific questionnaires. Symptoms and severity of depression were measured with the Beck Depression Inventory (BDI, Beck & Steer, 1987; German version: Hautzinger, Bailer, Worall, & Keller, 1995), a 21-item self-report questionnaire. Obsessive-compulsive symptoms were assessed with a short version of the Hamburg Obsessive-Compulsive Inventory (HZI; Zaworka, Hand, Jauemig, & LiinenschloB, 1983), which includes items on obsessive behaviour as well as ruminations prior to acting. Agoraphobic cognitions were measured with the Agoraphobic Cognition Questionnaire (ACQ; Chambless et al., 1984; German version: Ehlers et al., 1993). Avoidance with regard to common agoraphobic situations was assessed by the Mobility Inventory, subscale Alone (MI-A; Chambless, Caputo, Jasin, Gracely, & Williams, 1985; German version: Ehlers et al., 1993). Hypochondrias was measured with the Whiteley-Index (WI; Pilowsky, 1967; German Version: Rief, Hiller, Geissner, & Fichter, 1994), which assesses disease phobia, bodily preoccupation, and disease conviction.

Health related variables. Chronic health problems were assessed by the application questionnaire and the medical report of the examination before treatment. Satisfaction with health was measured by the “Satisfaction with Health” subscale of Questions on Life Satisfaction (FLZ-GG; Henrich & Herschbach, 2000; German Version: Henrich & Herschbach, 1996). Internal consistency for the German version is .89. As the FLZ-GG was not given to patients from the beginning, calculations can only be made for a smaller sample of n = 65. Alcohol use was assessed by the self-evaluation scale of the Munchner

Alkoholismus-Test [Munich alcoholism test] (MALT-S, Feuerlein, Kiifner, Ringer, & Antons-Volmerg, 1999). The MALT-S scale contains 24 items that assess three relevant aspects of alcoholism: drinking and attitude towards drinking, alcohol related psychological and social impairment, and somatic complaints. It has a split-half reliability of 0.94. The use of benzodiazepines was assessed by the application questionnaire and the DIPS.

Treatment and therapist variables. The experience of the therapists was coded on a 6-point scale, according to the number of patients with any disorder treated so far (1 = 1-10 patients treated with any disorder, 2 = 11-20 patients etc.). Treatment outcome

Symptom Checklist-90-Revised Interpersonal Sensitivity (SCL-IS; Derogatis, 1983; German version: Franke, 1995). This subscale assesses feelings of social uncertainty and fears of being observed or judged negatively. Internal consistency for the German version of the SCL-IS is .86.

Rating of global improvement (RGI). A 7-point rating scale (1 = very much better, 2 = much better, 3 = better, 4 = no change, 5 = worse, 6 = much worse, and 7 = very much worse) was used to assess the subjective perception of improvement. The RGI can be considered as a global consumer satisfaction measure.

5.2.5. Analysis

Analysis was conducted in a series of steps. In a preliminary analysis of treatment attrition, reasons for patient discontinuation and dropout were investigated and patients were classified as refusers, dropouts and treatment completers. Second, in order to find pre-treatment differences between patients who refused treatment and those who completed it ANOVA or chi-square tests were computed.

Third, for prediction of treatment change, the first step was to compute bivariate correlations between potential predictors and SCL-90-Interpersonal Sensitivity “Residual Gain Scores” (RGS) as well as the ratings of global improvement (RGI) at post treatment and one-year- follow-up (FI) for the completers. To compute residual gain, raw scores from pre, post, and FI assessment are first converted into Z scores. Change is calculated by subtracting the Time 1 score, multiplied by the correlation between scores at time 1 and 2 from the time 2 score (RGS = Zpost – Zpre rprepost). Thus, residual gain rescales an individual’s score relative to typical gains made by others at the same initial level. We then regressed each factor on the predictors (method stepwise) to take into account the shared variance of the individual predictors. To safeguard adequate predictive power, we selected only those predictors that related ip < .05) to RGS or RGI and entered them into the equations.

Finally, for the prediction of relapse in the 90 patients that had completed the SCL-IS at post as well as FI, we calculated “Reliable Change Indexes” (RCI) using the formula by Jacobson, Follette, and Revenstorf (1984), with RCI = (Mposttest – Mfi)/SE, and SE = XDposti’kst y/l-rxx’, where rxx’ is the reliability of the measure. Following the authors’ suggestions, we categorized a patient as deteriorated if the RCI was lower than – 1.96. Then we calculated differences in pre-treatment and post-treatment variables between those who had improved further or remained stable from post to FI and those who had deteriorated. Finally, variables that significantly differentiated the two groups were entered into logistic regression.

5.3. Results

5.3.1. Preliminary Analyses

Thirty percent of the group of treatment refusers after diagnostic assessment gave a reason for discontinuation of treatment. Of these, 60% stated that they discontinued because the health insurance refused to cover the costs, another 20% had began treatment elsewhere, 13% had doubts concerning the treatment concept, and 7% reported organizational difficulties. Thus it can be assumed that many of these patients either completed treatment elsewhere or will eventually return to treatment when other problems have been resolved or treatment can be more easily afforded. The group of dropouts after cognitive preparation, who had received an individualized treatment offer presented a different pattern of reasons. Seventy-five percent of this group provided us with a reason for discontinuation. Of these, a far lower percentage of patients discontinued for financial reasons (28%), but many felt that treatment was too difficult to endure (22%) or were sceptical about the treatment rational (17%). Similarly, for the group of dropouts during treatment, of which 63% gave us the reason, 17% felt the treatment to difficult to endure, 8% were sceptical of the rational and 42% marked the rubric “other reasons”, which may have included problems in the therapeutic relationship. As a consequence, refusers after diagnostic procedure must be regarded separately from refusers after cognitive preparation and interpretation of results in this group must be treated with caution. On the basis of this analysis, we decided to categorize the sample as follows: refusers after diagnostic procedure (RD = 16%), refusers after cognitive preparation (RC = 8%), treatment dropouts (TD = 6%) and treatment completers (TC = 69%).

5.3.2. Predictors of Treatment Refusal and Dropout

Table 5.2. shows the pretreatment variables for refusers after diagnostic procedure (RD), refusers after the cognitive preparation phase (RC), dropouts during therapy (TD) and treatment completers (TC). Because of the high number of comparisons we applied Bonferoni-adjustment for each comparison separately (p = .05/32 =.002). On this basis, the groups only differed significantly in their number of comorbid diagnoses and their mean value on the MI-A. TD reached higher scores on the MI-A than any of the other groups.

An additional analysis of group differences, in which all patients giving a financial reason for discontinuing were excluded from the calculation produced the same results, apart from one difference: patients, who refused treatment after diagnostic assessment were using medication significantly less often (23%), than patients who refused after cognitive preparation (56%), dropped out during treatment (75%) or completed treatment (57%), (Chi2 = 16.4, df= 3, p = 001).

5.3.3. Predictors of Treatment Outcome

The results of the two-tailed bivariate correlations between predictors and RGS as well as RGI[5] are shown in table 3. With regard to the RGS at post, the WI was the only significant predictor. Patients revealing more symptoms of hypochondriasis revealed less treatment change at post.

Predictors for RGI at post were the SPS, BSQ, BAI, the number of feared situations as well as the perceived anxiety in these situations, the BDI and the FLZ-GG. Patients who experienced more impairment before treatment on these measures rated themselves as having improved less at post. These seven variables were entered to predict RGI at post. Only the FLZ-GG made a significant contribution with a regression coefficient of В = -0.01 (/3= -.40; p < .01) and explained 16% of the total variance. Due to missing data in one or more of the predictor variables, only 60 patients were entered into the analysis.

By the 1-year follow-up (FI), patients with a higher level of education revealed less change. Also, the number of feared social situations as well as the amount of anxiety in these situations were negatively related to RGS at FI. These three variables were entered into the linear regression analysis with the RGS at FI as dependent variable. The amount of fear in social situations (B = .03; /3 = .32; p < .01) and the level of education (B = .21; /3 = .25; p < .05) both made a significant contribution to predicting treatment change and accounted for 13% of the total variance for N = 89 patients.

Predictors for the RGI at FI were gender, marital status, SCL-GSI, SPS, BAI, the number of feared situations as well as the perceived anxiety in these situations, BDI, ACQ, FLZ-GG and number of treatment sessions. Female as well as married patients tended to rate themselves more improved. On the questionnaires, the severity or impairment before treatment was a negative predictor for perceived improvement. Patients who received a higher number of treatment sessions rated themselves as less improved at FI. Two of the eleven variables entered into linear regression to predict RGI at FI made a significant contribution. The marital status had a regression coefficient of В = 0.24 (/3= .36; p < .05) and the FLZ-GG had a regression coefficient of В = -0.01 (/3 = -.35; p < .05) for N= 31 patients. Together these two variables explained 28% of the total variance.

5.3.4. Predictors of Deterioration after Treatment

The results of the calculation of differences in pretreatment as well as in posttreatment variables between those who had improved further from post to FI and patients who had deteriorated are depicted in table 4. Patients who could not maintain their treatment gain were shown to be significantly younger, to have higher pre-treatment-scores on the SPS, higher pretreatment and posttreatment scores on the SIAS, a larger number of feared social situations as well as higher levels of anxiety in these situations. These variables were entered into logistic regression (forwards, wald). Only the pretreatment score of the SPS reached statistic significance as a predictor, with a coefficient of В = -.12 (wald = 5.4; p < .05; N = 39) and accounted for 24% of the total variance.

5.4. Discussion

The calculation by Turner et al. (1996), in which they estimated 52% of the patients seeking treatment as actually profiting from it, is underlined by our data from the clinical field. If we consider not only patients who refused after cognitive preparation, but also those who discontinued after diagnostic procedure and did not justify refusal with financial difficulties and add the rate of patients who dropped out during treatment, we are left with 80% of the patients who completed treatment. For these, we can consider a rate of 56% reliably improved patients at post, which was calculated in a former study (Lincoln et al., 2002) and optimistically assume that patients who did not send back the follow-up questionnaires equally improved, and we are left with a rate of 43%. Our study was to our knowledge the first field study investigating predictors of refusal, dropout and treatment response for social phobia treatment. In the next section, some of the findings will be discussed in detail.

First, it seems important to point out that treatment refusers are not a less severely impaired group of patients that we do not have to be overly concerned about. Twenty-five percent even of this group have already been hospitalised for mental problems, and they achieve results comparable to treatment completers on all pretreatment questionnaires.

Patients who dropped out during treatment revealed more avoidance behaviour than any of the other groups as indicated by the higher scores on the MI-A. In line with this finding is the significant difference between the groups in the number of comorbid diagnoses. Possibly, the higher comorbidity causes these patients to feel more uncertain about whether the treatment is going to be sufficient. Additionally, dropout might be explained by the tendency for these patients to be characterized by higher rates of depression. A depressive attribution style will tend to be more global and stable (e g. I will always be a total loser) and lead patients to give up more readily, when treatment success does not become visible quickly enough. Thus, in the case of intensive treatment with a large amount of exposure elements it seems to be more important to make sure a depressed patient completes treatment than to worry about treatment response. In spite of slightly contradictory results about the exact way in which depression interferes with treatment, some authors come to the conclusion that it may be wise to spend more time tracing and dealing with pretreatment hampering cognitions or argue for concurrent treatment of anxiety and depression for the more depressed patients (Scholing & Emmelkamp, 1999; Chambless et al., 1997; Rief, Auer, Wambach, & Fichter, in press; Heinrichs et al., 2001).

Although there were a number of significant, but low correlations between pre-treatment variables and change or subjective improvement, there were not many significant predictors once variables were entered into linear regression. The “satisfaction with health” subscale of the FLZ was the most significant predictor in the regression analysis of subjective improvement at post and 1-year follow-up, without, however, predicting actual change. Also, more objective data, such as the presence of chronic disease, as reported in the medical report or stated by patients in the application questionnaire did not show any relationship to improvement. High scores on the FLZ-GG might reflect a positive thinking bias. Patients who reveal less discrepancy between importance of health aspects and satisfaction with these aspects might generally tend to be less sensitive to negative discrepancies in their life.

More generalized social phobia (indicated by a higher amount of fear in social situations) was a negative predictor of change at 1-year-follow up, which stands in contrast to the results of former studies finding no effect of subtype on change (Brown et al., 1995; Hope et al., 1995; Turner et al., 1996). However, most of these studies did not predict change by 1-year followup and all used a dichotomic subtyping scheme. Chambless et al. (1997), using a similar approach (continuum of impairment instead of dichotomised subtypes) found no correlation with change at post symptoms, but a weak correlation after six months. Hope et al. (1995) did not find subtypes to improve unequally by one-year follow-up, but they only had a small sample (N = 16) for their follow-up assessment and the effect is not a very large one. Our finding seems plausible because patients with more generalized social phobia still suffer more from symptoms and avoidance after treatment, possibly leading to fewer positive new experiences in social situations and thus making it harder to maintain treatment gain over a longer period of time.

We also found a higher level of education to be a negative predictor of change at 1-year- follow up. This finding is new as the available studies did not investigate the effect of education on change or endstate functioning. However, the effect is small and definitely needs to be replicated before giving it further attention. Finally, patients, who were married rated themselves as more improved after one year. This finding also stands in contrast to the results in other studies, finding no impact of marital status for social phobic patients (Salaberia & Echebuma,1996), but is in line with findings by Heinrichs, Hahlweg, Fiegenbaum, Frank, & Schroeder (2001) for patients with panic and agoraphobia. It seems, that future research should give more attention to the impact of marriage and partnership.

Although depression added no significant contribution in the regression analyses we would like to point out, that symptoms of depression were related to subjective improvement ratings at post and FI, but not to symptom change. Possibly, the global improvement ratings are vulnerable to depression, because depressed patients tend to evaluate success less optimistically. This explanation also fits in well with the finding of Chambless et al. (1997), who found depressed patients to reveal less change in the self-report measures, but to be rated more positively by observers.

For prediction of deterioration after treatment, only the pretreatment score on the SPS added a significant contribution to the regression analysis. Patients with higher pretreatment social phobia scores were more likely to relapse. This contradicts the finding by Mersch et al. (1991), who found patients who relapsed to have had lower pretreatment severity scores. However, their method of categorization was different, including patients who made no further progress in the category of relapse. Also, our finding is supported by the other significant correlations, indicating that patients with a more generalized form of social phobia deteriorated more often after treatment as well as by the negative predictors of long-term treatment change. In sum, our data indicate that less severely disordered patients tend to find it easier to keep up a stable treatment gain over a longer period of time.

Limitations and Considerations. One limitation of the study is that the Personal Sensitivity subscale of the SCL-90-R was the only outcome variable available for the entire sample. Observer rated outcome as well as more specific measures of social fear and avoidance would have been a better indicator of treatment success. Also, the high percentage of missing followups after one year makes the generalisation of the predictors of long-term change contestable. On the other hand, this high rate of missing data might reflect the reality of field treatment, in which patients are under no obligation to send back follow-ups. A further limitation is that not all variables of interest (e g. Axis П comorbidity, motivation and expectancy) were assessed, which complicates the search for important predictors, as regression coefficients change with every change in the predicting variables. This limitation is due to a long period of data collection and the fact, that at the beginning some variables were considered to be less important.

One problem of long-term follow-up assessments is that it is difficult to control for all the important variables that may influence outcome. For example, Chambless et al. (1997) found medication use and additional treatment between posttest and follow-up to predict outcome and thus controlled for these factors in her study. Although approximately one third of our sample received some form of additional treatment after postassessment, ranging from brief counselling or relaxation to another attempt at cognitive behaviour therapy this revealed no significant relationship to residual gain or reliable change at one-year follow up.

Generally, research on treatment predictors has not led to a great insight in the sense that a particular factor can be seen as mainly responsible for treatment failures. Even if larger effects were found, it is always possible that an unknown third variable moderates the relationship. But, as in the experimental settings, the effect sizes in this study are generally small, suggesting that specific pretreatment variables are of limited value, and that it is more helpful to interpret patterns of predictors. However, some important clinical implications should be emphasised: (1) The group of treatment refusers is as severely impaired by social phobic symptoms as patients who undergo treatment. Additional efforts are needed to motivate these patients to take up treatment. (2) Cognitive preparation and the beginning of treatment should be even more adapted to depression or other comorbid disorders, by restructuring hampering cognitions or conducting disorder-specific additional treatment. (3) It seems important to arrange for additional sessions over a specific period of time when patients are more severely impaired or suffer from more generalized social phobia, to enable them to integrate the treatment effects into their everyday life.


Table 5.1.

Overwiew of Predictors for Treatment Refusal, Dropout, Functioning and Relapse after Treatment for Social Phobia

Treatment Change, Endstate
Predictors Refusal Dropout Change Endstate Relapse
Older Age 9, 18


15, 18


15, 3


+ !2 +!2
Gender 9, 18


15, 18








Marital Status ,9 18 + 0 15, 18




Education 9, 18




Occupation 9






older Age at onset 18




_ 4 _ 4  
Duration of Disorder 9




More family members     _ 4 4  
Medication use     _ 5    
Earlier treatment trials       12  
Severity/Impairment 9 18 0 ~ _ 18 16 3,5 “ 0 12, 13, 15 12
Behavioural impairment 18


+15   12  
Comorbid Axis I 9, 18


18, 11


18, 11


18 11,13

“ 0

Depression     _ 5, 16 13


Generalized Subtype 18




2, 8, 18


_ 2, 8, 18  
Comorbid Axis II 9, 18




_2,5,7, 17, 18 _ 18,7,8  
      1, 3, 16, 18,


High Expectancy   14 15

0 “

+ 5,14 + 15  
Locus of Control       10


Homework Compliance       + 6, 10  

— = negative predictor, + = positive predictor; 0 = no significant effect found, 1 = Alden & Capreol (1993), n = 76; 2 = Brown, Heimberg, & Juster (1995), n = 104 ; 3 = Butler, Cullington, Munby, Amies, & Gelder (1984), n = 49; 4 = Cameron, Thyer, Feckner, Nesse, & Curtis (1986), n = 41 (including specific phobia and agoraphobia); 5 = Chambless, Tran, & Glass (1997), n = 62; 6 = Edelman & Chambless (1995), n = 52; 7 = Feske, Perry, Chambless, Renneberg, & Goldstein (1996), n = 60; 8 = Hope, Herbert, & White (1995), n = 28; 9 = Juster, Heimberg, & Engelberg (1995), n = 70; 10 = Leung & Heimberg (1996), n = 104; 11 = Mennin, Heimberg, & MacAndrew (2000), n = 122; 12 = Mersch, Emmelkamp, & Lips (1991), n = 47; 13 = Otto, Pollack, Gould, Worthington, McArdle, et al. (2000), n = 15; 14 = Safren, Heimberg, & Juster (1997), n = 113;15 = Salaberia & Echebuma (1996), n = 48; 16 = Scholing & Emmelkamp (1999), n = 50; 17 = Turner (1987), n = 13; 18 = Turner, Beidel, Wolff, Spaulding, & Jacob (1996), n = 84.


Table 5.2.

Mean (SD) and Percentages of Variables at Pre-treatment for Refusers after First Session and Diagnostic Procedure (RD), Refusers after Cognitive Preparation (RC), Treatment Dropouts (TD) and Completers (TC).

  RD = 46 RC = 24 TD= 18 TC =199 df Test-Value P
Demographic and biographical variables
Age 33.8(10.4) 35.4(10.4) 29.4 (9.4) 34.1 (10.2) 280 F=\A .241
Male 63% 50% 42% 58% 3 Chi2 = 23 .512
Marital Status         6 Chi2 = 1.6 .266
Married 34% 33% 18% 32%      
Partner 14% 21% 6% 25%      
No partner 46% 46% 77% 43%      
Education         6 Chi2 =3.7 .712
Sec. School 27% 33% 47% 36%      
High School 38% 25% 26% 32%      
University 36% 42% 32% 32%      
Age at onset 20.1 (10.6) 20.4 (9.6) 19.3 (10.7) 19.8 (9.5) 263 F=0.48 .986
Duration 13.4(14.2) 15.2(12.7) 9.8 (7.8) 14.1 (11.4) 260 F= 0.84 .474
Pre-treatment         6 Chi2 = 8.8 .187
None 21% 9% 6% 22%      
Outpatient 52% 57% 50% 57%      
Inpatient 26% 35% 44% 21%      
Medication 37% 48% 71% 57% 3 Chi2 = 8.1 .045
Severity and Impairment
General 3.4 (.73) 3.8 (.52) 3.8 (.43) 3.4 (.67) 207 F= 3.32 .021
SCL-GSI 1.09 (0.59) 1.15 (0.48) 1.32 (0.65) 0.96 (0.59) 264 F = 2.68 .047
SPS 34.9(15.4) 40.4(12.8) 44.7 (21.3) 38.2 (16.9) 149 F= 0.73 .538
SIAS 41.1 (13.4) 42.5 (15.1) 42.4(16.0) 40.3 (15.6) 148 F=0.13 .941
DIPS 5.7 (1.6) 6.2 (1.5) 6.4 (1.2) 5.9 (1.3) 197 F= 1.02 .386
SCL-IS 1.73 (0.94) 1.98 (1.15) 1.96 (0.90) 1.54 (0.94) 265 F = 2.38 .047
Situations 6.4 (2.7) 6.3 (2.9) 7.4 (3.0) 6.7 (3.0) 254 F= 0.50 .683
Anxiety 21.5 (9.0) 22.8 (10.1) 25.8(10.6) 21.8 (9.3) 255 F= 1.09 .353
Diagnoses 0.5 (0.7) 1.0 (0.9) 1.3 (1.3) 0.5 (0.7) 283 F= 8.52 .000
BDI 18.6(8.6) 17.1 (8.5) 22.7(11.7) 15.3 (10.4) 259 F= 3.64 .013
HZI-G 2.5 (2.5) 2.1 (1.3) 3.9 (2.5) 2.4 (2.1) 213 F= 1.86 .137
HZI-H 2.6 (2.3) 2.6 (2.0) 3.3 (1.7) 3.0 (1.9) 213 F= 0.59 .625
ACQ 1.9 (0.6) 2.2 (0.5) 2.3 (0.9) 1.9 (0.5) 260 F= 3.78 .011
MI-A 1.8 (0.7) 1.9 (0.8) 2.6 (1.0)a 1.8 (0.7) 229 F= 6.37 .000
WI 3.9 (3.6) 4.2 (3.2) 4.5 (3.2) 2.8 (2.7) 226 F= 3.29 .022
Physical symptoms
Symptoms 14.7 (7.8) 19.5 (8.3) 18.9(12.5) 17.3 (8.5) 250 F= 1.70 .168
BSQ 2.2 (0.6) 2.4 (0.8) 2.4 (0.7) 2.2 (0.7) 254 F= 0.92 .431
BAI 19.9(11.6) 25.6(11.8) 24.5 (12.2) 21.1 (12.1) 253 F= 1.42 .236

Health related variables

FLZGG 5.4 (28.8) 14.1 (34.8) 15.0(23.5) 23.8(31.4) 137 F= 2.03 .112
Chronic 16% 0% 7% 15% 3 Chi2 =4.5 .213
MALT 3.3 (4.5) 2.9 (4.3) 1.8 (1.6) 2.9 (3.3) 219 F= 0.57 .638
Benzo- 5% 17% 12% 19% 3 Chi2 =5.3 .153
Therapist Variables
Experience 2.6 2.5 2.3 2.2 275 F= 1.83 .142
Male 54% 30% 32% 44% 3 Chi2 =4.4 .218

“ = differences between TD and TC in post hoc Tukey-HSD, Games-Howell or Chi2 Test (p < .01); SCL-GSI = Symptom Checklist-90-Revised, Global Severity Index; SPS = Social Phobia Scale; SIAS = Social Interaction Anxiety Scale; DIPS = Diagnostic Interview Гог Psychological Disorders; SCL-IS = Symptom Checklist-90- Revised, Interpersonal Sensitivity; BDI = Beck Depression Inventory; HZI-G, Hamburg Obsessive Compulsive Inventory – Ruminations; HZI-H; Hamburg Obsessive Compulsive Inventory – Obsessive Behaviour; ACQ = Agoraphobic Cognition Questionnaire; MI-A = Mobility Inventory, Alone; WI = Whiteley Index; BSQ = Body Sensation Questionnaire; BAI = Beck Anxiety Inventory; FLZ-GG = Questions on Life Satisfaction, satisfaction with health; MALT = Munich Alcoholism Test.


Table 5.3.

Bivariate Correlations Between Potential Predictors and Residual Gain Scores (RGS) of Social Phobic Symptoms and Subjective Rating of Global Improvement (RGI) of Completers at Posttreatment (post) and One-year follow-up (F1)

    Post   F1
  SCL-IS (N)   SCL-IS (N)
Demographic and biographical variables    
Age -.00 (160) .10 (166) -.10 (100) -.10 (118)
Gender .05 (160) .05 (167) .02 (100) .28 (119)**
Education -.02 (157) -.07 (164) .20 (99)* .12 (117)
Marital status .02 (157) -.10 (165) -.03 (99) .20 (118)*
Age at onset .04 (149) .04 (156) -.00 (97) -.04 (114)
Duration -.03 (149) .05 (155) -.10 (97) -.04 (113)
Prior treatment .06 (155) .11 (162) .07 (98) .17 (117)
Medication .08 (156) .05 (163) -.05 (99) -.06 (118)
Severity & Impairment      
Impairment .02 (112) -.03 (115) .10 (73) -.20 (81)
SCL-GSI .03 (159) .13 (164) .04 (100) .27 (117)**
DIPS -.03 (120) .15 (129) .06 (81) .19 (95)
SPS .05 (84) .26 (82)* .23 (48) .29 (52)*
SIAS .07 (83) .18 (81) .14 (48) .23 (52)
Physical .01(143) -.04 (150) .02 (90) .16 (109)
BSQ .14(154) .20 (158)* .01 (95) .16 (111)
BAI .10(152) .16 (157)* .03 (95) .23 (111)*
Situations .06 (145) .17 (152)* .26 (91)* .27 (110)**
Anxiety .11 (145) .22 (152)** .27 (91)** .30 (110)**
Diagnoses -.08 (160) .03 (167) -.05 (100) .09 (119)
BDI .09 (155) .23 (160)** .09 (97) .23 (113)*
HZI-G -.01 (139) -.08 (141) .08 (86) .02 (102)
HZI-H .02 (139) .02 (141) -.10 (86) .08 (102)
ACQ .13 (157) .11 (161) .04 (98) .24 (114)**
MI-A -.08 (141) .03 (147) .04 (86) .14 (101)
WI .19 (139)* .15 (139) -.02 (87) .19 (96)
Health related variables      
FLZ-GG -.11 (77) -.34 (74)** -.04 (45) -.37 (48)**
Chronic disease -.06 (82) .01 (86) -.05 (52) -.12 (60)
MALT .12 (133) -.06 (138) .01 (88) .19 (100)
Benzodiazepines .10 (156) .15 (163) .03 (99) .02 (118)
Treatment and therapist variables      
Gender .11 (156) .02 (163) .05 (98) -.05 (117)
Experience .03 (155) -.04 (162) .13 (97) -.10 (116)
No. Sessions .09 (121) .08 (125) .10 (77) .27 (93)**

Sample numbers for correlations are additionally reduced by missing data in the assessment measures. * = p <.05. ** = p <.01 for significant correlations; SCL-GSI = Symptom Checklist-90-Revised, Global Severity Index; DIPS = Diagnostic Interview for Psychological Disorders; SPS = Social Phobia Scale; SIAS = Social Interaction Anxiety Scale; BSQ = Body Sensation Questionnaire; BAI = Beck Anxiety Inventory; BDI = Beck Depression Inventory, HZI-G = Hamburg Obsessive Compulsive Inventory – Ruminations; HZI-H = Hamburg Obsessive Compulsive Inventory – Obsessive Behaviour; ACQ = Agoraphobic Cognition Questionnaire; MI-A = Mobility Inventory, Alone; WI = Whiteley Index; FLZ-GG = Questions on Life Satisfaction, Satisfaction with Health; MALT = Munich Alcoholism Test.


Table 5.4.

Means and Standard Deviations or Percentages of Variables at Pre- and Postassessment for Patients who Deteriorated (DET) or Remained Stable (STAB) between Post- and 1-year


  DET N = 12 STAB N = 78 Test-Value df P
Demographic and biographical variables Age 29.8 (5.4) 34.1 (10.0) t = -2.2 24.7 .037*
Male 75% 53% Chi2 =2.1 1 .145
Marital status 17% married 32% married Chi2 = 1.4 2 .491
Educational level 25% partner 58% solo 0% none 26% partner 42% solo 2% none 00





3 .621
Age at onset 25% sec. school 33% high school 42% university 19.4 (3.3) 37% sec. school 33% high school 28% university 20.8 (9.9) t = -0.9 49.6 .353
Duration 10.4 (3.9) 13.3 (10.7) f = -1.7 44.8 .095
Prior treatment 83% Outpatient 59% Outpatient Chi2= 2.6 1 .109
  25% Inpatient 12 % Inpatient Chi2= 1.5 1 .217
  67% Medication 37% Medication Chi2= 3.8 1 .051
Severity and Impairment SPS pre1 56.4 (10.4) 30.4(16.8) / = 3.4 40 .002**
SPS post 29.4 (19.7) 16.5(13.8) t= 1.9 45 .067
SIAS pre1 58.6(13.1) 36.5(15.7) t= 3.0 40 .005**
SIAS post 36.6(17.9) 21.9(13.8) / = 2.2 45 .034*
DIPS 6.4 (1.1) 5.9(1.1) t= 1.3 73 .214
Impairment 3.5 (0.8) 3.3 (0.8) t= 0.8 64 .441
SCL-GSI pre 1.1 (0.5) 0.9 (0.6) t= 1.5 88 .146
SCL-GSI post 0.5 (0.4) 0.5 (0.5) t = -0.6 87 .951
No. Physical 18.1 (9.6) 16.9(8.0) /=0.5 78 .655
BAI pre 20.3 (7.5) 19.8(11.8) /=0.2 83 .875
BAI post 8.4 (4.5) 11.8(10.4) /=-1.9 37.7 .060
BSQ pre 2.1 (0.5) 2.2 (0.7) /=-0.6 83 .577
BSQ post 1.7 (0.5) 1.6 (0.6) /=0.3 84 .735
Situations 8.4 (2.2) 6.3 (3.0) / = 2.2 79 .030*
Anxiety 26.4 (6.1) 20.5 (8.8) / = 2.1 79 .037*
Diagnoses 0.6 (0.8) 0.51 (0.7) /=0.2 88 .831
BDI pre 18.4(12.8) 13.6(9.6) t= 1.5 85 .129
BDI post 6.6 (6.9) 6.7 (7.1) t = 0.0 14.8 .969
HZI-G pre 3.5 (2.8) 2.6 (2.1) t= 1.3 75 .200
HZI-G post 1.8 (2.3) 1.4 (1.5) t = 0.6 74 .548
HZI-H pre 2.9 (1.3) 3.2 (1.9) t = -0.4 75 .716
HZI-H post 2.5 (1.7) 2.7 (1.8) t = -0.4 73 .710
ACQ pre 1.9 (0.4) 1.9 (0.5) t = 0.4 86 .710
ACQ post 1.5 (0.3) 1.5 (0.5) t = 0.1 85 .935
MI-A pre 1.9 (0.6) 1.8 (0.7) t = 0.4 76 .690
MI-A post 1.3 (0.5) 1.3 (3.6) t = 0.0 79 .998
WI pre 2.4 (3.0) 3.1 (2.9) t = -0.8 77 .405
WIpost 2.3 (2.9) Health related variables 2.0 (2.5) t = 0.3 77 .738
FLZ-GG pre 21.0(17.0) 32.1 (37.4) t = -0.6 37 .565
FLZ-GG post 77.5 (30.7) 59.2(41.2) t = 0.9 42 .395
Chronic Disease 0% 15% СЫ2 = 0.5 1 .470
Benzodiazepine 8.3% 15.6% СЫ2 = 0.4 1 .508
MALT pre 3.5 (2.6) 2.6 (3.0) t = 0.8 76 .424
MALT post Therapist Variables 1.3 (1.6) 2.1 (3.0) t = -0.8 76 .440
Experience 2.1 (1.0) 2.1 (1.0) t= 0.1 85 .922
Male therapists 58% 36% СЫ2 = 2.3 1 .132
No. of sessions 30.7(12.6) 35.4(13.6) t = -0.9 67 .355
* = p < .05, ** = p < .01 for significant differences in t-tcst or Chi Square Test; 1 = SPS = Social Phobia Scale;

SIAS = Social Interaction Anxiety Scale. ‘Calculations for SPS and SIAS were based on smaller samples: for

pre: n = 5 deteriorated and n = 37 stable, for post: n = 5 and n = 42 respectively; DIPS = Diagnostic Interview for Psychological Disorders; SCL-GSI = Symptom Checklist-90-Revised, Global Severity Index; BAI = Beck Anxiety Inventory; BSQ = Body Sensation Questionnaire; BDI = Beck Depression Inventory; HZI-G = Hamburg Obsessive Compulsive Inventory – Ruminations; HZI-H = Hamburg Obsessive Compulsive Inventory – Obsessive Behaviour; ACQ = Agoraphobic Cognition Questionnaire; MI-A = Mobility Inventory, Alone; WI = Whiteley Index; FLZ-GG = Questions on Life Satisfaction, Satisfaction with Health.

Calculations for FLZ-GG were based on smaller samples: for pre: n = 4 deteriorated and n = 35 stable, for post: n = 4 and n = 40 respectively; MALT = Munich Alcoholism Test.


Figure 5.1.

Number of Patients at the Different Stages of Assessment and Treatment



  1. Summary

6.1. Summary

In spite of the success of cognitive behavioural therapy for social phobia found in research, it remains unclear whether interventions will remain successful in the routine of clinical practice, where patients and treatment conditions might differ from those in research samples. Also, response rates make clear that not all patients benefit from the investigated treatment approaches. Almost half of the patients either refuse to undergo treatment after it has been offered, drop out during treatment or do not profit from completing it. In order to adapt treatment conditions better to individual needs, knowledge about variables predicting treatment success or failure is necessary. Studies investigating such predictors have so far yielded some contradictory results, have neglected prediction of refusal and relapse after treatment and have been carried out in typical research conditions, with samples not necessarily representative of clinical practice.

As a consequence, the studies address three basic questions: (1) Do typical research conditions have an affect on the effect sizes achieved? (2) Can the results found in randomised controlled trials be generalized to clinical practice? (3) Which variables can predict treatment attrition and response in clinical practice?

Several approaches were taken in order to answer these questions. First, thirty studies testing treatment effects for social phobia were re-examined by categorizing them according to the quality and amount of applied sample restriction and laboratory study characteristics and comparing their mean effect sizes. Second, 217 unselected patients with a primary diagnosis of social phobia according to DSM-III-R who began treatment in one of four outpatient clinics of the Christoph-Domier Foundation of Clinical Psychology in Germany (CDS) were assessed before and six weeks after treatment, using an extensive assessment battery. Treatment outcome as well as clinical significance were calculated. Both the sample and the treatment outcome were compared to samples and outcome in the 30 efficacy studies and to outcome reported in meta-analyses. Thirdly, it was tested whether a restriction of the sample according to typical exclusion criteria would result in a larger effect size. Finally, the sample was completed by another 70 social phobic patients who were seeking treatment in the CDS but discontinued before treatment started. The 287 patients were then classified as refusers after diagnostic assessment (16%), refusers after cognitive preparation (8%), dropouts (6%), and completers (69%). Outcome was assessed by calculating relative change via residual gain scores and by patient improvement ratings six weeks and one year after the end of treatment. Patients who completed the one-year follow-up (n = 101) were categorized as stable (87%) or deteriorated (13%). Demographic and disorder-related as well as therapist and treatment variables were analysed as predictors for each classification.

The results of the analysis of outcome studies indicate that even the accumulation of sample restriction, such as excluding patients with comorbid disorders or outside a certain age-range does not have any predictive value for treatment effect. However, there was a significant tendency for studies applying several “laboratory treatment conditions”, such as recruiting patients by adverts, applying treatment in university settings, using specifically trained therapists, and following a treatment manual to achieve higher effect sizes.

The sample of patients in the study in clinical practice did not differ considerably from the samples in the comparison studies. The results six weeks after the end of therapy showed significant reductions in social phobic fears and avoidance as well as in general anxiety and symptoms of depression. The effect sizes are comparable with the average effect-sizes reported by meta-analytic studies of controlled efficacy research using selected patients. Restricting the sample according to the selection criteria often applied in research settings did not result in higher effect sizes. Fifty-six percent of the sample changed significantly with regard to social phobic symptoms.

The analysis of response in the sample of 287 patients seeking treatment for social phobia revealed a much lower response rate: only 43% of the patients originally seeking treatment completed and benefited from it in the end. The only significant predictor for treatment attrition was comorbidity. Treatment gain was best predicted by satisfaction with health (FLZ- GG). Also, patients characterized by more generalized social phobia improved less by 1-year- follow-up. Pretreatment depression had no effect on change as assessed by the self report measures, but more depressed patients reported having improved less. Finally, patients who were more severely impaired at pretreatment (as assessed by the SPS) found it harder to maintain treatment gain.

Taken together, it can be concluded that sample selection does not seem to enhance the effects of treatment and that individual cognitive behaviour therapy for social phobia can be transported from research settings to the field of mental health. However, although similar success rates can be achieved in clinical practice, practitioners are well advised to maintain supervision and keep up regular training. Finally, there is hope to further improve the effectiveness of treatment by giving more attention to severely impaired patients or patients with comorbid disorders, who are more prone to dropout or relapse after treatment.

6.2. Zusammenfassung

Trotz der durch eine Vielzahl von Forschungsarbeiten bestatigten Erfolge von kognitiver Verhaltenstherapie fur Soziale Phobie bleibt fraglich, ob die untersuchten Interventionen auch im klinischen Alltag, in dem sich Patienten und Behandlungsbedingungen moglicherweise von denen in der Forschung unterscheiden, erfolgreich sind. Die bisherigen Befunde verdeutlichen zudem, dass bei Weitem nicht alle Patienten von den Behandlungsangeboten profitieren. Fast die Halfte derer, die Behandlung aufsuchen, treten entweder vor Beginn der Behandlung zuriick oder erreichen keine nennenswerte Verbesserung. Um Behandlungs- angebote besser an die individuellen Bedurfnisse anzupassen, ist Wissen iiber Pradiktoren zur Vorhersage von Therapieerfolg unabdingbar. Trotz einer Reihe von Studien, die Therapie- pradiktoren bei der Behandlung von Sozialer Phobie untersuchten, gibt es bisher nur wenige eindeutige Ergebnisse. AuBerdem ist zu bemangeln, dass die Vorhersage von Riicktritten und Rtickfallen vernachlassigt wurde, und dass die Studien unter kontrollierten

Forschungsbedingungen stattfanden, in denen die Bandbreite potentieller Pradiktoren moglicherweise eingeschrankt ist.

Aus diesen Uberlegungen leiten sich drei wesentliche Fragestellungen ab: (1) Haben typische Forschungsbedingungen und selegierte Stichproben Einfluss auf die EffektgroBe? (2) Ist es moglich, die Ergebnisse aus randomisierten und kontrollierten Studien auf die klinische Praxis zu iibertragen? (3) Durch welche Variablen konnen Rucktritte, Dropout und Behandlungserfolg in der klinischen Praxis vorhergesagt werden?

Verschiedene Herangehensweisen dienten der Beantwortung der Fragestellungen. Zunachst wurden 30 Studien, die die Wirksamkeit kognitiver Verhaltenstherapie fur Soziale Phobie untersuchten, re-analysiert, indem sie anhand der Qualitat und Quantitat von Stichprobenselegierung und Laborcharakteristika eingestuft und ihre mittleren Effekte verglichen wurden. Weiterhin wurde eine unselegierte Gruppe von 217 Patienten mit der Primardiagnose Soziale Phobie nach DSM-III-R, die in einem von vier Instituten der Christoph-Domier Stiftung fur Klinische Psychologic (CDS) eine Behandlung aufsuchten, vor und sechs Wochen nach der Behandlung einer ausfuhrlichen Diagnostik anhand einer klinischen Fragebogenbatterie unterzogen. Der Behandlungserfolg und die klinische Signifikanz wurden errechnet. Die Stichprobe und die Ergebnisse wurden mit Stichproben und Ergebnissen der 30 Wirksamkeitsstudien sowie mit den in Metaanalysen berichteten durchschnittlichen Effekten verglichen. Zusatzlich wurde untersucht, ob eine Selegierung der Stichproben anhand forschungsiiblicher Kriterien in einer groBeren Effektstarke resultieren wiirde. SchlieBlich wurde die Stichprobe um weitere 70 Patienten mit Sozialer Phobie erganzt, die zwar Behandlung aufsuchten, aber vor Beginn der Behandlung zuriicktraten. Die Gesamtstichprobe von 287 Patienten wurde in 4 Gruppen unterteilt: Rticktritte nach der Diagnostik (16%), Rticktritte nach Kognitiver Vorbereitung (8%), Abbriiche wahrend der Therapie (6%) und Patienten, die die Therapie abschlossen (69%). Fur die behandelten Patienten wurde der Therapieerfolg als das relative AusmaB der durch die Behandlung erzielten Veranderung („residual gain scores”) sowie durch die subjektive Therapieerfolgsein- schatzung erfasst. Patienten, die an der 1-Jahres Katamnese teilnahmen (n = 101) wurden als stabil (87%) oder verschlechtert (13%) eingestuft. Demographische und storungsbezogene, sowie Therapeuten- und Behandlungsvariablen wurden als Pradiktoren fur jede Klassifikation analysiert.

Die Ergebnisse der Analyse der Therapiestudien deuten darauf hin, dass sogar eine Akkumulation von angewendeten Selektionskriterien, wie z.B. der Ausschluss komorbider Depression oder von Patienten auBerhalb einer bestimmten Altersspanne keinen Einfluss auf die GroBe der Effekte hat. Andererseits fand sich ein Zusammenhang zwischen der Menge zutreffender ,,Laborcharakteristika“, wie z.B. Werbung der Patienten iiber Anzeigen, Durchfiihrung der Behandlung in universitarem Setting, speziell trainierte Therapeuten oder genaues Befolgen eines Therapiemanuals, und der GroBe des Effektes. Studien, auf die mehrere Laborkriterien zutrafen, erzielten etwas hohere Effekte.

Es zeigten sich keine wesentlichen Unterschiede zwischen unserer Stichprobe in der klinischen Praxis und Stichproben in Vergleichsstudien. Die Ergebnisse sechs Wochen nach Beendigung der Therapie zeigten eine deutliche Verringerung sozialphobischer Angste und Vermeidungsverhaltens, wie auch allgemeiner Angstlichkeit und Depressivitat. Die erreichten Effekte entsprechen den mittleren Effektstarken aus Metaanalysen, die uberwiegend kontrollierte Studien mit selegierten Patientengruppen auswerteten. Eine Einschrankung unserer Behandlungsgruppe anhand typischer Selektionskriterien fuhrte nicht zu hoheren Effektstarken. Eine klinisch relevante Verbesserung sozialphobischer Symptome wurde von 56% der Patientenstichprobe erreicht. Die Analyse der Gesamtstichprobe von 287 Patienten, die urspriinglich Behandlung aufsuchten, ergab eine deutlich geringere Erfolgsrate: Lediglich 43% der Patienten vollendeten die Behandlung und profitierten von dieser.

Im Hinblick auf Pradiktoren fur Therapieabbriiche erwies sich eine hohere Komorbiditat als positiver Pradiktor. Die durch die Behandlung erzielte Veranderung konnte am besten durch die subjektive Zufriedenheit mit gesundheitlichen Lebensaspekten (FLZ-GG) vorhergesagt werden. Patienten, die durch eine generalisiertere Sozialphobie gekennzeichnet waren, zeigten weniger Veranderung zur 1-Jahres Katamnese. Komorbide Depression zu Beginn der Behandlung hatte zwar keinen Einfluss auf die Veranderung durch die Behandlung, aber depressivere Patienten hatten den subjektiven Eindruck, weniger profitiert zu haben als nicht depressive. SchlieBlich gelang es Patienten, die bereits vor der Therapie einen hoheren Schweregrad der Sozialen Phobie (in der SPS) aufwiesen, schlechter, den erreichten Behandlungserfolg aufrechtzuerhalten.

Zusammenfassend kann gefolgert werden, dass Stichprobenselegierung die Behandlungs- effekte nicht begiinstigt, und dass individuelle kognitive Verhaltenstherapie fur Soziale Phobie von der Forschung in die klinische Praxis tibertragen werden kann. Obwohl vergleichbare Erfolge erreicht werden konnen, sind Praktiker gut beraten, eine ausfiihrliche Supervision in Anspruch zu nehmen und regelmaBige Fortbildungen aufzusuchen. SchlieBlich bleibt zu hoffen, dass die Effektivitat der Behandlung weiter verbessert werden kann, indem zusatzliche therapeutische Interventionen starkere Beriicksichtigung finden, die der besonderen Belastung von komorbid oder in hoherem AusmaB gestorten Patienten gerecht werden.




Appendix D. Description of Treatment

D.l. Formal Treatment Conditions

During the one or two weeks of intensive treatment patients are normally accommodated in a hotel or a guesthouse situated in close proximity to the Christoph-Dornier-Foundation. At the time of the therapy described in the studies (1990-1999) the cost of treatment was paid for in most cases by the health insurance as «Kostenerstattungsverfahren” [reimbursement of expenses]. This means that invoices for treatment sessions are directed to the patient, who has to apply for reimbursement with his or her health insurance company. The insurance company is free to decide whether or not they are prepared to cover the expenses for treatment. This decision process mostly took place after diagnostic assessment, and the rejection of the application to cover the costs was the most frequent reason for treatment attrition at this stage of treatment. Additional treatment-expenses, such as accommodation, tickets, etc. were not covered by the health insurance.

D.2. First Session

After receiving the application questionnaire, the therapist contacts the patient and arranges a date for a first session. The first session is usually conducted in the rooms of the institute of the Christoph-Dornier-Foundation, but therapists are willing to conduct it in the house of the patient if problems are so disabling they prevent a patient from coming.

The most important aspects of the first session have been described by Frank and Frank

(2000). The first contact offers the patient an opportunity for a first impression of the therapist and the institute, a first description of the problem for which he or she is seeking treatment and clarification of organisational questions.

The therapist tries to gain all necessary information needed for the planning of the diagnostic procedure as to clarify whether he will be able to offer the patient adequate treatment or whether he must refer to another institution. Additionally, the therapist informs about the disorder, offers explanations for symptoms, and gives information on the further procedure of treatment and other organisational questions. At the same time he concentrates on building up a good emotional relationship by assuring that he understands the patient’s suffering, as well as taking the problem seriously and refraining from evaluating or accusing.

D.3. Medical Check-up and Diagnostic Assessment

A medical check-up is carried out by general practitioners or specialists in cooperation, who are well informed about symptoms of social phobia. The doctors complete a medical report, which has been specifically developed for patients with an anxiety disorder. The medical check-up is particularly important in the context of exposure since this can be physiologically stressful and may be contraindicative (e.g. for patients with coronary heart disease). Further, a detailed attempt to clarify the source of specific, particularly impairing symptoms (e.g. extreme trembling) can be of importance with regard to cognitive restructuring interventions as well as for setting realistic goals.

Diagnostic assessment takes place in approximately four to six treatment sessions, usually completed in one day. It consists of several components. One basic component is the diagnostic interview [Diagnostisches Interview fur Psychische Storungen] (DIPS, Margraf, Schneider, & Ehlers, 1991). Apart from gaining a reliable diagnosis the therapist aims at receiving a clear picture of all anxiety provoking situations as well as the amount of fear they provoke. He will also try to gain as much information about avoidance behaviour and safety behaviours, needed for an adequate planning of treatment. The therapist also tries to gather the information required for the model of explanation, the factors that engender and maintain the problem in the past or at present. Finally the patient is asked to complete a series of questionnaires which are depicted in Appendix E.

D.4. Cognitive Preparation

Cognitive preparation for therapy takes place about one week later and is necessary to enhance the patient’s motivation for treatment. The concept of cognitive preparation is based on the explanations given by Battling, Fiegenbaum, and Krause (1980) for the treatment of panic disorder and agoraphobia, but has been adapted to the treatment of social phobia. The length of the cognitive preparation session is variable and depends on the individual problem of a patient, his or her expectations concerning therapy, the motivation for change and the relationship between patient and therapist. The cognitive preparation has four basic goals.

D.4.1. Explanation of cause and maintenance of social phobic behaviour and experiences

According to intellectual abilities, previous experience and own attempts of explaining the problem, different cognitive-behavioural theories can be used. The phobic behaviour is portrayed as normal, learnt behaviour that has developed because of inconvenient learning conditions. A basic aim is to relieve the patient from concerns about being different or something being „fundamentally wrong“ with him or her, as such assumptions lead to devaluations of the self. Instead it is underlined that avoidance and safety behaviour carry the main responsibility for maintenance and intensification of anxiety. The influence of further important factors, as far as they turned out to be of importance in the diagnostic assessment (perfectionism, self-focused attention, one-sided interpretations) is explained. The therapist develops an individualized model with the patient, for example following the model of Clark und Wells (1995), and sketches it for the patient on paper. An example for such a model can be seen in Figure D.l.

Figure D. 1. Simplified Example of an Explanation Model used in Therapy

D.4.2. Deriving the treatment

The sketched model is used as a basis from which the treatment is derived stringently and the treatment procedure is explained in a transparent way. The necessity of a detailed analysis of dysfunctional cognitive schemes responsible for the specific interpretation in the situation as well as the importance of exposure to the feared situations is derived in a way that that is plausible for the patient. The understanding on behalf of the patient is enhanced by using metaphors and encouraging patients to imagine what would happen if they do not avoid the feared situations, using so called “thought-experiments”. Furthermore, information on the duration and course of treatment is given. The therapist points out, that he will not get involved in discussions or distractions during exposure tasks, but will try to prevent avoidance-behaviour whenever possible. The patient is requested to refrain from all precautions that normally serve to reduce fear (sedatives, lucky charms, or not drinking coffee beforehand in order to avoid trembling).

D.4.3. Emphasising the patient’s responsibility

The personal responsibility of the patient is underlined by explicitly giving the patient about a week time to come to their own decision for or against participation in the therapy. The therapist makes absolutely no attempt to persuade the patient to participate. In this time purposely no further interventions take place in order to not disturb the decision process. However, the patient is offered the possibility of consultation if needed.

D.4.4. Development of a trusting relationship

The aim is to build up the relationship between patient and therapist in way that encourages the patient to trust the therapist and perceive him as competent. The therapist makes clear that he takes the patient’s fears seriously and understands how difficult and stressful it must be for the patient to have to confront him- or herself with the feared situations.

D.5. Therapy

When the patient decides to participate, exposure and cognitive intervention begin (duration is variable and depends on the individual patient’s needs). The therapist is in close contact with the patient during the first days, during which it is not unusual for treatment to last for six to eight hours. Exposure to the feared situations plays a central role in the therapy as it serves several purposes, with varying importance in the course of treatment. First, it is used to experience a certain degree of habituation to the situation. Secondly, it helps to assess further anxiety-provoking and maintaining cognitions, safety-behaviour, selective attention as well as self-focused attention. Thirdly, confrontation with the feared situations offers the patient the possibility of testing negative beliefs concerning his or her behaviour or the behaviour of others, by evaluating video-recordings of the exposure situation. If possible, an audience used for the exposure situation can also function as giving feedback in order to correct dysfunctional self-perception. Fourthly, for patients with deficits in social skills the situations can also be used for training.

Exposure always takes place with cognitive restructuring interventions, in which the analysis of fear-relevant cognitive concepts and schemes play a central role. Only after these concepts have been clearly defined the actual restructuring can take place. The basic strategy then consists of “system-immanent” dialoguing, which increases a patient’s motivation to drop and replace dysfunctional concepts. Thus, finally, after restructuring interventions have taken place, the exposure situations give the patient an opportunity to test out alternative concepts (e.g. “It is okay to make a mistake sometimes”), by changing his or her behaviour in the situation (e.g. allowing mistakes to happen or even making a mistake on purpose).

Generally, exposure situations are chosen depending on the patients’ individual fears and starting with those in the top half of an anxiety hierarchy. Examples for exposure situations are giving a short speech in front of an audience, eating soup in a restaurant, serving drinks, chatting with a member of the opposite sex, or keeping eye-contact with other passengers in public transport. The therapist aims at providing the patient with situations that are as natural as possible and contain the specific fear provoking elements.

Further elements of the therapy, that are used when it seems appropriate consist of behavioural experiments, such as experimenting with the effect of safety behaviours using video-feedback or conducting an opinion poll on a theme that is relevant to the patient (e.g. “What goes through people’s mind when someone blushes?”). Apart from being a basis for cognitive restructuring, these interventions also contain exposure elements.

Therapists are free to vary the amount of exposure and cognitive therapy as well as the length of the intervention according to the needs of the individual patient. They are also free to use additional specific interventions for the treatment of co-morbid disorders.

In order to gain a clearer and less abstract picture of the treatment a brief case description will be given. A 25-year old man called Max was one of our patients who was diagnosed with social phobia. Max’s main fear was that other people would see that his hands were shaky and he would be rejected because of this. One of his most feared items was having to serve drinks to his guests, especially while being observed by several people. Thus, one of the first exposure interventions consisted of the therapist inviting an audience and preparing drinks to be served. The therapist made sure that coffee was served in fine china cups, which had to be held by the handle, as mugs would have made it easier to conceal jittering. Also, as Max had reported being particularly concerned about being rejected by men of his age, the therapist made sure that the audience consisted mainly of men in the range of 20 – 30 years of age. The exposure was extensively prepared with Max, noting expectations about the way he felt he was going to be perceived and defining criteria for success. Max feared that at least one person in the audience would be laughing, or that there would always be an awkward silence while he was serving. He defined a successful situation as one in which he did not actually spill any coffee and managed to stay until everybody had been served. For Max it was particularly important to hear from the audience whether and how much they actually perceived him to be trembling and what they had been thinking, when they noticed this.

During the exposure situation a co-therapist videotaped the exercise and the therapist occasionally interrupted the performance to assess the amount of perceived anxiety, rated on a scale from 0 (no fear) to 10 (maximum fear). As fear remained high while serving drinks, the therapist instructed Max to serve another round of coffee. The situation was terminated when Max reported the fear to be at the level of about four. The audience was then asked to give the specific feedback defined before the exposure. The feedback was also videotaped because Max feared that people would not be honest enough to his face. Finally, the therapist discussed the exposure situation and the feedback with Max, using it as a natural segue into restructuring interventions in which Max was taught to identify and challenge specific negative thoughts and general cognitive errors (e g. “Because I feel uncertain, I must be performing badly and trembling extremely.”) and perfectionist thinking (e g. “A less-than- perfect performance is a failed performance”). The video feedback was used as an objective feedback and also helped to detect safety behaviours. For example it could be seen that Max sometimes used both hands when offering the cup to someone.

D.6. Self-Control Phase

In the days after the intensive treatment phase the patient is encouraged to continue exposing himself to the identified situations for several hours each day. During this period further contacts with the therapist (e.g. in form of a telephone contact at the end of the day or a short treatment session) are scheduled. The patient plans the self-control phase in close cooperation with the therapist. Together, they clearly define which situations are to be practiced, how long the patient should remain in the situations and when and how often the patient is to expose him- or herself to feared situations. At the end of the self-control phase patient and therapist analyse the experiences the patient made and derive a supporting program for the next weeks, during which the exposure tasks are integrated more and more into the patient’s every-day life and do not require so much additional time.



A Meta-Analysis of the Effect of Guided Imagery Practice on Outcomes

Debra Van Kuiken, R.N., B.S., B.S.N.

University of Cincinnati

Guided imagery is an intervention used by nurses in a variety of settings. It has been suggested that better outcomes will occur with continued practice. No studies were found that examined the relationship between practice duration and strength of outcomes. The focus of this meta-analysis was the effect size of guided imagery intervention studies with different durations. Statistical findings of 10 studies of various durations were converted to d statistics and plotted against the duration of study. The results show an increase in effect size of guided imagery over the first 5 to 7 weeks; however, the effect was decreased at 18 weeks.

Keywords: meta-analysis; guided imagery; mind-body techniques; self-hypnosis

Imagery, as a healing art, has been around for millennia (Achterberg, 1985). From the use of symbolism, dreams, and visions by shamans to the guided imagery of today, healers have used the connection of mind and body to affect health. Increasingly, researchers have been looking at the efficacy of this low-technology approach to health care. Guided imagery has been studied in a variety of settings including cardiac surgery, cardiology, oncology, stroke rehabilitation, and pain management (Halpin, Speir, CapoBianco, & Barnett, 2002; Klaus et al., 2000; Kolcaba & Fox, 1999; Maguire, 1996; Page, Levine, Sisto, & Johnston, 2001). When Wallace (1997) analyzed the literature on relaxation and imagery, several concerns emerged regarding the delivery of guided imagery as an intervention. Wallace noted that the length, frequency, and details of the intervention were varied and often missing from the research report. Eller’s (1999b) review of the literature noted inconsistencies in content and methods of delivery. Delivery of the intervention has been by audiotape or personal interaction, one on one or in groups. The images used ranged from pleasant environments and memories to symbols and physiological processes. Furthermore, studies ranged in duration from hours to months.


Little research has been done to establish the frequency or duration of guided imagery required to effect change. The purpose of this meta-analysis was to examine the research literature published since 1996 to ascertain the relationship between the duration of guided imagery intervention and the resultant effect size. For purposes of the current study, duration of the intervention was the measure of the time from the beginning of the intervention until the time of outcome measurement.

Literature Review


Imagery is the thought process that takes into account and translates the senses for the nervous system to, in turn, produce healing change throughout the body (Achterberg, 1985; Horrigan, 2002). Guided imagery maybe done through prompting by a live practitioner, an audiotape, or self-prompting. Several theories for this process have been proposed. Achterberg (1985) illustrated pleasant imagery as a method of inducing a deep sense of calm and an avenue of reframing emotions and thus changing the sympathetic-parasympathetic balance. Achterberg also described a neuroanatomic model in which formation of nonverbal images and the processing of emotions occur adjacent to each other in the cerebral right hemisphere. This association is translated to an autonomic response. The left hemisphere’s conscious control of the voluntary nervous system then modulates the emotional-autonomic response pathway. Research in the field of psychoneuroimmunology has established the presence of neurons in areas of white blood cell production and storage, further indication of a direct link between thought and physiological function (Zeller, McCain, & Swanson, 1996).

Dr. Marty Rossman, founder of the Academy for Guided Imagery, stated in an interview that guided imagery is based on the assumption that the body has an innate ability for self-healing and should be used as a complementary therapy to existing medical care, rather than an alternative healing practice (as cited in Horrigan, 2002).

In a review of the literature, four types of guided imagery were identified. The four types are pleasant imagery, physiologically focused imagery, mental rehearsal or reframing, and receptive imagery. Pleasant imagery guides the individual to imagine a calm, comfortable place. This may include images of mountains, oceans, or past memories that generate feelings of calm and nurturance. The individual may also be asked to imagine images of general well-being and health.

Physiologically focused imagery guides the individual to imagine the physiological function of the healing needed. For example, it may include images of lymphocytes and macrophages in the fighting an infection. Images may also be symbolic such as teaching patients undergoing skin grafts to imagine hands of the grafted skin holding hands with the blood vessels of the tissues below. This type of guided imagery requires education about the biological processes involved before the initiation of the guided imagery experience (Achterberg, Dossey, & Kolkmeier, 1994).

Another type of imagery is mental rehearsal or reframing. Mental rehearsal is the process of imagining the performance of a specific task in a relaxed state prior to actually performing the task. One familiar example of mental rehearsal is the athlete’s imaging the physical maneuvers needed before actually attempting those maneuvers physically. Neuronal firing with imagery occurs in the appropriate brain areas and may also reinforce neural pathways (Kosslyn, Ganis, & Thompson, 2001). Reframing imagery involves imagining and reinterpreting an event and the emotions connected to that event.

Finally, receptive imagery involves a scanning of the body (Horrigan, 2002). Because receptive imagery is of a diagnostic or reflective nature, the current study did not include receptive imagery as an intervention.


Study Selection

Computerized searches using the words guided imagery, imagery psychotherapy, and visualization and health on CINAHL (43 hits), MEDLINE (45 hits), and PubMed (10 hits) were done in October 2002. Limits on searches included only published research articles with adult human participants between the years 1996 to 2002. Eller (1999b) and Wallace (1997) provided detailed syntheses of the literature. They described the lack of reported details in many studies of guided imagery. The current meta-analysis attempted to quantify the effect of imagery; therefore, only studies since 1996 were included in the current analysis. Searches were limited to English texts. Thirty- three studies combining imagery with other therapies such as music, journaling, or healing touch were excluded because of the difficulty in isolating the effect of imagery. Studies (25) done for primary psychiatric diagnosis such as post-traumatic stress disorder, anorexia, or depression were also excluded because the confounding emotional and mental factors. Studies using self-hypnosis were included if imagery was the primary component and any induction was described as a simple relaxation technique. After combining the searches, the sample consisted of 16 published studies. Six of these studies did not provide adequate statistical data, such as F or t values, needed for calculating effect size in a meta-analysis. No unpublished studies were used for this analysis.


The method of analysis of the data included the transforming of statistics from each original study to the d statistic. Using the equations set forth by Moody (1990) and Cooper (1998), the d statistic allows for a comparison of the effect size for each of the studies. Effect size represents the difference of the means of the groups (intervention and control) in terms of their pooled standard deviations (Cooper, 1998). If studies had several dependent variables producing more than one effect size per time period, the mean effect size was used to represent each study equally and to prevent violation of independence (Fernandez & Turk, 1989). For the purposes of the current analysis, weighting factors were not calculated. In studies where only means and standard deviation were available, an estimated effect size was obtained by the following formula: ME – Me/ [(SDE – SDC) / 2] (Cooper, 1998).



Figure 1: Guided Imagery Effect Size as a Function of Duration of Intervention.


To determine the relationship between duration of the intervention and the effectiveness of guided imagery, the effect size was plotted against the weeks (duration) of the study (see Figure 1). All studies were represented by only one point, except two studies that had repeated measures at distinct times (Donaldson, 2000; Kolcaba & Fox, 1999). A separate d statistic was calculated for each data collection point. Duration of the intervention was defined as the time in weeks from the onset of the intervention to the time of measuring the outcome.



Ten studies from nursing and medical journals between 1996 and 2002 comprised the sample. The study samples included healthy medical students at exam time (n = 2), persons suffering chronic illness (n = 7), and those experiencing acute illness or invasive procedures (n = 1). Outcomes include physiological measures (immune parameters) and psychosocial measures such as anxiety, locus of control, and coping. Table 1 is a summary of the studies included in the meta-analysis.

Variations in Duration

Plotting the effects of guided imagery against duration of the intervention revealed increasing effects as time progresses (see Figure 1). However, at 18 weeks, the effects declined in strength. Kolcaba and Fox (1999) substantiated this in their findings that the effect size of the intervention at 3 weeks is indeed more than at 18 weeks. Donaldson (2000) found the effect of guided imagery increased with time as evidenced by effect sizes of 4.27, 4.89, and 6.64 for days 30, 60, and 90, respectively (see Figure 2). These results were excluded from Figure 1 as they represented possible outliers. However, the increase in white blood cells (WBCs) of 38.3% over the 90 days of the study is worth noting.

Variation in the Intervention

Unlike previous literature reviews, studies obtained for the current analysis are more consistent in delivery method of guided imagery. All but two studies (Kiecolt-Glaser, Marucha, Atkinson, & Glaser, 2001; Lang et al., 2000) used instruction accompanied by an audiotape for practice at home. Page et al. (2001) reported using three different tapes, providing participants with a variety, whereas Kolcaba and Fox (1999) used one tape for the 18-week intervention.

The content of the imagery for a majority (60%) of the studies was physiologically focused on immune function or disease process. Three studies that used immune parameters as dependent variables found significant results. Twenty people with chronically low WBC counts demonstrated very strong effect size (ranging 4.27 to 6.64), which accumulated over time, after physiologically focused imaging (Donaldson, 2000). In two studies, students under the stress of exams experienced a stable immune function after imagery, while students in the control groups experienced decreases in key immune factors (Gruzelier, Smith, Nagy, & Henderson, 2001; Kiecolt-Glaser et al., 2001). However, it must be noted that these two studies, published in the same year, yielded conflicting results. The markers, natural killer and CD8, that showed stability in the Gruzelier et al. study were the markers least affected by imagery in the Kiecolt-Glaser study.



Studies Used in Meta-Analysis

Author N Sample


Intervention Duration in Weeks Dependent


d Statistic
Donaldson, 2000 20 Dx depressed WBC Physiological imagery of 12 WBCs  
    (< 5000) x 6 months, pre- immune system   1. At 30 days 1.4.27
    and posttest, repeated     2. At 60 days 2.4.89
    measures     3. At 90 days 3. 6.64
Eller, 1999a 69 HIV diagnosis, outpatient Physiological imagery of 6 1. QOL 1. Ns
    clinic RCT immune system   2. Perceived 2. Ns
          health status  
Fors, Sexton, 55 Women with fibromyalgia Pleasant imagery attention 4 Pain  
Gotestam, 2002   RCT focusing   1. With pleasant l.a .41
          2. With attention 2.a-.29
Gruzelier, Smith, 28 Students at exam time RCT Physiological imagery of 4 1. NK cells 1. .98
Nagy, & Henderson,     immune system   2. CD8 2. .70
2001     self-hypnosis   3. Cortisol 3.1.07
          4. CD3,4,&13 4. Ns
Kiecolt-Glaser, 33 Students at exam time RCT Physiological imagery of 2 1.immune assays 1. .83
Marucha, Atkinson,     immune system   2. CD8/NKlysis/ 2. Ns
& Glaser, 2001     self-hypnosis   macrophages  
Kolcaba & Fox, 53 Women with BC RCT Structured-reframing of 18 Comfort  
1999     RT experience   1. At 3 wks l.b .55
          2. At 18 wks 2b .47


Lang et alv 2000 241 Primarily White, during Pleasant imagery hypnosis 0 1. Analgesic used 1. .46
    percutaneous induction   2. Length of 2. .33
    procedures/RCT     procedure  
Maguire, 1996 33 Multiple sclerosis (MS) Physiological imagery 6 1. State anxiety 1. .85
    diagnosis, outpatient RCT     2.internal control 2. .80
          of health  
          3. MS symptoms 3. Ns
Page, Levine, Sisto, 13 Stable poststroke RCT Pleasant imagery, mental 7 1. Motor recovery l.a .85
& Johnston, 2001     rehearsal   2. Arm function 2.a 1.25
Walker et al., 1999 86 Women with advanced Physiological imagery of 18 1. Coping 1. .47
    BC (nonmetastasized) RCT immune system   2. QOL 2. .45
          3 Clinical response to chemotherapy 3. Ns

NOTE: Ns = nonsignificant; RCT = randomized controlled trial; QOL = quality of life; WBC = white blood cells; BC =breast cancer; NK = natural killer; CD3,4, 8, & 13 = types of t-lymphocytes; NKlysis = measure of NK function; RT = radiation therapy.

  1. Effect size estimated using means and SD.
  2. Effect size given by the author.


Figure 2: The Effect Sizes in Donaldson’s Study of a Guided Imagery in Persons

With Chronically Low White Blood Cell Counts.

Fors/ Sexton, and Gotestam (2002) showed surprising results in comparing «attention focused» to «attention distracting» (pleasant) imagery in women with chronic fibromyalgia. Those whose imagery focused on the physiology and pain reported an increase in pain (t = 4.4, p < .001). This was compared to a smaller increase in pain for the control group (f = 2.6, p < .01). However, the group using attention- distracting or pleasant imagery experienced a decrease in pain (t = 3.3, p = .001). Maguire (1996), also using physiologically focused imagery in patients with multiple sclerosis, found no significant decrease in symptoms associated with the disease. Eller (1999a) noted that two participants dropped out of the study citing increased anxiety when imagining «the virus.» This may partially explain the finding that the relaxation (control) group experienced a 5% increase in perceived health compared to an 11% decline with the physiologically focused guided imagery group.

Pleasant imagery was incorporated into four studies, including Fors et al. (2002) mentioned above. Lang et al. (2000) used attentiveness with pleasant imagery during percutaneous procedures that led to a decrease in the amount of pain medication needed as compared to the control group. Page et al. (2001) combined pleasant imagery with mental rehearsal and found significant increase in motor recovery and arm function in patients poststroke. Kolcaba and Fox (1999) combined pleasant imagery with reframing of the radiation treatment in a caring and supportive context (e.g. «the table is a reassuring presence, providing strong support for your body,» p. 68).


Variation in Outcomes

Significant physiological outcomes included increased number of WBC and other immune parameters (Donaldson, 2000; Gruzelier etal.,2001;Kiecolt-Glaser etal., 2001). Physical symptoms of multiple sclerosis and clinical response to chemotherapy were not significantly different than those of controls following physiologically focused imagery (Maguire, 1996; Walker et al., 1999). However, pain, motor recovery, and arm function were improved with pleasant imagery.

Other outcome measures in the study samples include psychosocial feelings of anxiety and control, perception of health, mood, coping, and quality of life. Maguire (1996) reported a significant increase in internal control of health and a decrease in distress. Comfort, during radiation treatments, improved with pleasant refraining imagery (Kolcaba & Fox, 1999).


This meta-analysis included only 10 studies to answer the question of whether the duration of practice influenced the effectiveness of guided imagery. Although the number of studies prevented statistical analysis, it is clear that immediate results from guided imagery are possible, and practice does increase the effectiveness of the intervention. Results prompt the question of sustainability of the intervention. Kolcaba and Fox’s (1999) findings of a decrease in the effect size at 18 weeks may be due to only one tape being used for the duration of the intervention and, consequently, participants losing interest. The sample did not include studies lasting between 8 and 18 weeks, thus there were no data points for those weeks. Increasing the number of studies and including studies with durations evenly distributed may give greater knowledge about a plateau effect and when it might occur.

The need for a specific terminology regarding the type of imagery is evident from this and previous reviews of the literature (Eller, 1999b; Fernandez & Turk, 1989). When reporting on imagery, the type of imagery must be explicitly reported. The importance of this distinction was made evident when Fors et al. (2002) found pleasant imagery effective in reducing fibromyalgia pain, however imagery that focused on the pain was detrimental. This is consistent with Fernandez and Turk’s (1989) findings, in a meta-analysis of cognitive strategies, that pleasant imagery provided a consistent reduction in pain whereas focused imagery increased pain in some studies. This is not to dismiss the use of imagery that focuses on and attempts to transform the pain. Syrjala and Abrams (2002) suggested focused imagery in treating chronic pain and distraction for acute pain. Future studies that are unambiguous in the type of imagery and the types of pain treated are needed to advance the science. Findings indicate that the clinician must monitor the clients’ use of imagery to prevent adverse effects.

Studies with physiologically focused imagery demonstrate effectiveness in maintaining and bolstering immune function. Donaldson (2000) found significant results by having participants visualize an increased WBC count. Randomized controlled trials are needed to support these findings. The reduction of the effect of exam stress on immune function is also noteworthy (Gruzelier et al., 2001; Kiecolt- Glaser et al., 2001). Robinson, Mathews, and Witek-Janusek (2002) pointed out the many confounding factors in immunology studies and questioned the assumption of equating immune parameters with immune competence. The conflicting effects on CD8 cell counts (Gruzelier et al., 2001; Kiecolt-Glaser et al., 2001) demonstrate the need for continued study.

All six studies excluded because of inadequate data used pleasant imagery as the type of intervention (see Table 2). Intervention durations range from 2 weeks to 25 weeks and reflect significance at all time intervals. The studies involving patients of surgical procedures indicated a cost benefit of guided imagery with decreases in length of stay and analgesic use (Halpin et al., 2002; Tusek, Church, & Fazio, 1997). Lang et al. (2000) reported similar findings. Continued work in this area is needed to further develop guidance for safe and effective means of decreasing health care costs.

Richardson et al. (1997), in a study of women with breast cancer, and Klaus et al. (2000), with eight persons with congestive heart failure (CHF), each used pleasant imagery with a focus on well-being for duration of 6 weeks. Both reported positive changes in well-being, but not statistically significant findings. However, Richardson et al. (1997) did report a significant increase in coping in that time.

Another emerging area of research is the use of mental rehearsal in stroke rehabilitation. In a pilot study, two of three patients with hemiparetic stroke displayed increased ability after pleasant imagery combined with a mental rehearsal of tracing a line (Yoo, Park, & Chung, 2001). Page et al. (2001) and Yoo et al. (2001) utilized small samples, and further studies are needed, however this area of investigation appears promising.



Studies Reporting Insufficient Statistics to Calculate a d Statistic

Author N Sample


Intervention Duration in Weeks Findings
Baider, Peretz, Hadani, & Koch, 2001 90 Cancer RCT Pleasant imagery 25 Decrease distress3
Halpin, Speir, CapoBianco, & Barnett, 2002 120 Cardiac surgery, quasi-experimental Pleasant imagery 4 Significant for length of stay,3 procedure, drug cost3
Klaus et al., 2000 8 CHF pilot pre-/post- Pleasant imagery focus on well-being 6 Ns for respiratory and lower extremity function, slight increase in QOL
Richardson et al., 1997 47 BC RCT Pleasant imagery focus on well-being 6 Ns for immune function, slight increase in QOL (Ns), improved coping3
Tusek, Church, & Fazio, 1997 130 Colo-rectal surgery RCT Pleasant imagery 2 Significant for lower analgesic requirement, time to 1st BM,3 length of stay,3 pain and anxiety3
Yoo, Park, & Chung, 2000 3 Hemiparetic stroke, pilot time series Pleasant imagery mental rehearsal 17


2 out of 3 showed better linetracing ability

NOTE: BC = breast cancer; Ns = nonsignificant findings; CHF = congestive heart failure; QOL = quality of life; BM = bowel movement; RCT = randomized controlled trial.

  1. Study cites significant findings, data not reported.



Imaging Ability/Preference

Achterberg et al. (1994) pointed out that a substantial number of people are unable to visualize images. They may be more inclined to have auditory, tactile, or olfactory images, and therefore imagery should include all the senses and be sensitive to individual preferences. Richardson et al. (1997) cited the vividness of image as a factor in determining success with guided imagery. Kwekkeboom, Huseby- Moore, and Ward (1998) examined imaging ability as a function of image vividness and absorption.

Later work by Kwekkeboom (2001) explored the influence of outcome expectancy on the success of guided imagery. Donaldson (2000) found outcomes were independent of the participant’s belief in the efficacy of the intervention. Baider, Peretz, Hadani, and Koch (2001) mentioned that participants’ motivation is a key factor in the success of the intervention being practiced daily and thereby affects change. They also cited that participants with a low initial level of distress had little change in distress. Kolcaba and Fox (1999) indicated similar results of greater efficacy of imagery in the first 3 weeks when the women’s anxiety level is usually highest. Further studies that measure pretreatment anxiety level of the participants may give us important information on the efficacy of guided imagery in individuals with different anxiety levels.

Personal preference of content should also be considered in the selection of the guided imagery. One participant from Kolcaba and Fox’s (1999) study dropped out of the study citing preference of listening to religious tapes. Personal preference of styles may be fundamental to the success of an imagery program and warrants further study.


Meta-analyses of collected studies provide insight into a particular intervention or outcome that is unobtainable from a single study; however, they also have limitations. The sample for this analysis was not a random sample of previous research, and therefore generaliz- ability is limited. Research studies that were either not published, had inadequate statistical data, or were not accessible were excluded from the analysis and may be a source of bias in the sampling (Byers & Stullenbarger, 2003). Nine of the 10 studies were randomized controlled studies; however, the sample sizes may not have provided adequate power. The calculated effect sizes were not weighted for quality or for sample size in this analysis.

Another limitation is the comparison of different outcomes (Polit & Hungler, 1999). Future analysis of studies with similar outcome measures will add rigor to the determination of the effect of practice. Participants were asked to practice daily in all the studies, and practice was recorded by practice log or diary in four studies. The continued need for precise and complete reporting of intervention details and outcomes is evident from the current meta-analysis.


The purpose of this meta-analysis was to determine if there was a relationship between the duration of a guided imagery intervention and the effect size of the outcomes. Plotting the duration of the intervention against the success of that intervention revealed a positive relationship. Although the graph indicated that success does improve with time, the large effect sizes reported by Donaldson (2000) and the decreased effect at 18 weeks are areas for further study. From the results of this meta-analysis, it is difficult to determine a minimal dosing time for significant outcomes. The evidence supports possible moderate to strong results at 4 weeks. However, immediate results seen in Lang et al.’s (2000) study indicated that many weeks of practice might not be needed. The length of time that effectiveness can be maintained is unknown. Because of the small sample size and the diversity of dependent variables, the results of this analysis are not generalizable. Detailed reporting of type of imagery, practice, and outcome measures is needed in future studies. As more studies become available, future meta-analyses using similar outcome measures will reveal a clearer picture of the effect of time on the efficacy of guided imagery.


Images of desire: Cognitive models of craving

Jon May, Jackie Andrade, and Nathalie Panabokke

University of Sheffield, UK

David Kavanagh

University of Queensland, Australia

Cognitive modelling of phenomena in clinical practice allows the operationalisation of otherwise diffuse descriptive terms such as craving or flashbacks. This supports the empirical investigation of the clinical phenomena and the development of targeted treatment interventions. This paper focuses on the cognitive processes underpinning craving, which is recognised as a motivating experience in substance dependence. We use a high-level cognitive architecture, Interacting Cognitive Subsystems (ICS), to compare two theories of craving: Tiffany’s theory, centred on the control of automated action schemata, and our own Elaborated Intrusion theory of craving. Data from a questionnaire study of the subjective aspects of everyday desires experienced by a large non-clinical population are presented. Both the data and the high- level modelling support the central claim of the Elaborated Intrusion theory that imagery is a key element of craving, providing the subjective experience and mediating much of the associated disruption of concurrent cognition.


When I get to work in the morning, I almost always have an intense urge to have a cup of coffee. Sometimes it comes to me almost like I am talking to myself–“I must get a cup of coffee before I do anything else”. Sometimes I look at my coffee cup and imagine it is filled with coffee. Sometimes my mouth feels dry, and I can imagine what it will be like to have a drink. I get these thoughts and feelings whether or not I have already had a cup at home, and I find it hard to concentrate on anything else without getting that cup of coffee first.


Craving is popularly recognised as a powerful subjective experience that motivates people to seek out and use or consume a craved substance. It has long been identified as an important symptom in addictive disorders (Jellinek, 1960), but the subjective experience of craving is not well explained by current conditioning and neurochemical theories. These see craving as a consequence of learned physiological responses to cues in the world or the body, perhaps predisposed by individual differences in neurochemistry, leading to the activation of schematised action plans. By focusing on the antecedent states of the body and the neurophysiological consequences of those states, these models do not address what is happening, mentally, during a craving episode. The subjective aspects are seen as secondary, because consumption and craving are both caused by the same physical events: the feelings that are experienced during a craving episode do not cause people to consume, but are a by-product of the addictive process. On the other hand, the need for a cognitive account of craving that does address the subjective aspects of craving is illustrated by evidence that pharmacological treatments for substance abuse work through their effects on craving (Monti et al., 1999; O’Malley, Krishnan- Sarin, Farren, Sinha, & Kreek, 2002) and do so most effectively when combined with cognitive behavioural therapy (Monti et al., 2001; O’Malley et al., 1996). Understanding the cognitive processes underlying craving may help to improve treatments for substance dependence, by allowing us to influence the mental and subjective aspects of those cravings that occur despite pharmacological intervention. In turn, understanding craving in substance dependence may help to improve our understanding of normal, non-addictive states of desire. In this paper we outline two cognitive accounts of craving, and use an information processing framework called Interacting Cognitive Subsystems (ICS; Barnard, 1985; May, 2001) to model and compare them. We then evaluate the models against data from a questionnaire in which a large sample were asked to describe the subjective nature of an everyday craving.

In general, cognitive theories allow the operationalisation of otherwise diffuse, descriptive terms. By incorporating clinical phenomena into established theories of cognition, this operationalisation supports the empirical investigation of the clinical phenomena, and the development of targeted treatment interventions. Building a cognitive model of a phenomenon requires the theorist to specify what processes are and are not involved, and inspection of the model allows inferences to be drawn concerning processing consequences that follow logically from the model. In this way the model can contribute to the scientific process by making clear whether a theory is self-consistent (i.e., whether it can in fact be modelled), and by identifying empirical tests that might not have been apparent from the theory itself.

We have previously used a cognitive approach to examine Eye Movement Desensitisation- Reprocessing (EMD-R), which has gained some popularity as a treatment for post-traumatic stress disorder (PTSD). In PTSD, vivid, lifelike images of trauma (called flashbacks) are a diagnostic symptom. EMD-R involves the use of eye movements during a type of imaginal desensitisation. While eye movements have not been found to amplify average treatment effects, we predicted from cognitive theory and research that they may have some utility. Using Baddeley and Hitch’s (1974; Baddeley, 1986) model of working memory as a theoretical framework, we hypothesised that tasks loading on either the phonological loop or the visuospatial sketchpad would reduce the vividness of mental imagery in the same modality. We demonstrated that this was the case (Baddeley & Andrade, 2000), and also that there was a concomitant reduction in the reported emotional response to the images and memories (Andrade, Kavanagh, & Baddeley, 1997; Kavanagh, Freese, Andrade, & May, 2001). We have suggested (Kavanagh et al., 2001) that carefully chosen working memory loads may be a useful treatment aid when designing stepwise exposure protocols for the treatment of post-traumatic stress disorder.

The EMD-R studies illustrate the potential that cognitive psychology has to explain the mode of action of clinical treatments, and to refine them to increase their effectiveness. They also illustrate that the development of cognitive explanations of clinical phenomena can be beneficial to cognitive psychology. The finding of reductions in image vividness and emotionality raised theoretical questions about the link between the working memory processes thought to underlie imagery, and emotion. The subjective nature of craving is another topic offering a potentially fruitful overlap between clinical and cognitive approaches. As in the case of traumatic imagery, it raises questions about the relationship between cognition and emotion, and specifically the role that mental imagery plays in craving.


Craving appears to be a major cause of the discomfort experienced by people trying to reduce their substance use, and subjectively is a factor in triggering relapse. While its actual causal role in relapse is disputed, the evidence suggests it is indeed one of several determinants. For example Killen and Fortmann (1997) found highly significant relationships between the craving of 2600 former smokers and their relapse over the subsequent 12 months.

We have proposed the Elaborated Intrusion (El) theory of craving, (Kavanagh, Andrade, & May, 2004) which aims to explain the cognitive processes underlying craving episodes, and to explain the emotional and motivational impact of these processes. In particular, we argue that craving episodes persist because cravers create mental images of the desired substance that are immediately pleasurable but which exacerbate their awareness of deficit. This causes a vicious circle of desire, imagery, and planning to satisfy that desire, followed by greater articulation of the imagery that engages high-level cognitive processes (e.g., working memory), impairs performance on concurrent cognitive tasks, and amplifies the emotional response. Our approach assumes that craving for addictive substances such as nicotine and alcohol is an extreme instance of a range of normal phenomena associated with motivated consumption behaviours.

A craving episode is shown in Figure 1. Any of several triggers can give rise to cognitive activity below the threshold of awareness, which can then trigger other associations. These do not have any specific consequences until they break through into awareness, when they are experienced as intrusive thoughts. These feel spontaneous because the triggers and the prior processing have been acting beneath awareness. Because the thoughts contain information that is linked to the use of some substance, there is an initial immediate positive sensation of reward or relief, as there would be if the substance were actually being used (a conditioned positive affective response). This encourages the individual to elaborate the thought, to enrich it, and to search for further associations, which are in turn also rewarding. We argue that the elaboration process particularly involves the construction of mental images of the substance or of its context of consumption, and it is these images that provide the strongest reward and relief. Semantic and verbal elaborations are also involved, but do not necessarily have the same affective consequences.


Figure 1. The Elaborated Intrusion theory (Kavanagh et al.). Antecedents of a craving episode are shown in rounded boxes outside the central box, which represents a craving episode. Cognitive products are indicated by rectangles and processing by an ellipse. Elaborative Processing includes the construction and elaboration of images of the target and its consumption. The consequences of craving (i.e., consumption or avoidance actions) are not shown in this figure.

This process of elaboration following intrusion is the key to our theory, and in normal situations it motivates an individual to seek out some substance that they may need to obtain. In maladaptive situations, it can lead them to seek out substances that they do not need, or are even attempting to avoid consuming. In these cases the elaborated thoughts soon lead to negative affective consequences as the individual realises the conflict with their goal of abstention, and they may attempt to control the elaboration by thought suppression or diversion (with limited success, or even increased elaboration, as in Wegner’s ironic processes, 1994). In other cases, the individual may not want to control their consumption, but may be prevented from obtaining the substance by circumstances or by its non-availability. In this case the rewarding consequences of the elaborated thoughts are soon overcome by the negative realisation of their growing deficit state, which is made increasingly salient by the content of their thoughts.

Memory plays a key role in the El theory of craving, because the processes of elaboration are essentially a goal-driven search through long-term memory for associations of the intrusive thoughts. The mediating mental representations that give rise to the subjective phenomenological aspects of craving are argued to be mental images arising from this elaborative search. The theory was informed by earlier research showing the role of working memory (WM) in imagery (e.g., Bad- deley & Andrade, 2000), but the WM model does not explain the link between imagery and emotion demonstrated by Andrade et al. (1997) and Kavanagh et al. (2001) and assumed in the El theory of craving. This is simply because Baddeley and Hitch’s (1974) WM model was only constructed as a model of short-term memory and its use in complex cognition, and there is no scope within the model for accounts of other phenomena. Emotional consequences of activity with the WM model are unspecified, and so this model does not provide any support for reasoning about the subjective nature of craving.

An information processing approach to cognition that does include an explicit account of the relationship between cognition and emotion is Barnard’s (1985) Interacting Cognitive Subsystems (ICS). Because of this it has been applied to a range of clinical phenomena, and provides a high-level framework for explaining the cognitive processes involved in such conditions. Whereas the El theory is only concerned with craving, ICS provides a way of modelling how craving might interact with other clinical problems. This modelling may help to design treatment interventions for complex problems such as comorbidities of substance abuse and psychological disorders such as schizophrenia or depression. As a general- purpose architecture for cognition, it can also help theorists to relate different accounts of a single phenomenon that have been motivated by different considerations, but which may overlap. Here we attempt an initial interpretation of the El theory within the ICS architecture, to compare it with an influential cognitive theory of craving (Tiffany, 1990), which incorporates the cognitive interference phenomenon but de-emphasises emotional aspects and argues that craving is epi- phenomenal to substance use.


We need first to briefly introduce the essential points of the ICS architecture (more detailed accounts are available in Teasdale & Barnard, 1993, and May, 2001), highlighting key theoretical terms in italics. ICS is a representational theory of mind, in which thought is divided into nine different types of representation, each with a specific quality of informational content (see Figure 2 for descriptions). Cognitive processes transform information from one type of representation into another, qualitatively changing it through interactions with memory. Cognition is a configuration of activity distributed across the overall architecture. The nine subsystems operate upon sensory (visual, acoustic, and body state), effector (articulatory and limb), perceptual (object and morphonolexical), and central (propositional and implicational) levels of representation. The implicational level will be crucial for our models of craving: it consists of high-level, connotative schemata that contain the abstract meaning for the individual of particular situations or events. Its flavour can best be appreciated by contrasting it with the factual and semantic nature of the propositional level. Making a qualitative judgement about something requires the implicational level; making a quantitative or comparative judgement requires the propositional level. The distinction between “hot” implicational and “cold” propositional representations is perhaps the hallmark of ICS.



Figure 2. The nine levels of representation in ICS are shown in the rectangular boxes, each box corresponding to a single subsystem. The arrows indicate processes by which one representation can be transformed into another. The somatovisceral changes caused by the implicational subsystem are physiological rather than cognitive, and so there is no somatovisceral subsystem to process or memorise them; however, they can be detected by the body state subsystem, as indicated by the white arrow.

Despite the different classes of information represented at each level in ICS, the nine subsystems carry out the same generic information processing activities, and share a common architecture. Each consists of an input array, which receives representations produced for it by other subsystems (or the senses); a copy process, which writes the contents of the input array into the subsystem’s image record, which thus acts as a memory for everything the subsystem has ever received; and a set of transformation processes, which operate upon either the contents of the input array or the contents of the image record that have been activated as a result of the copy process (i.e., revived memories). Each subsystem contains two or three transformation processes (indicated by the arrows in Figure 2), meaning that not all subsystems can exchange representations directly.

Memories and processing develop over time so that the transformation processes accrue sets of proceduralised knowledge that allow them to produce an output representation directly from patterns on the input array without accessing memories in the image record. Before this has occurred, however, they do need to access the image record, and only one of the transformations within a subsystem can access its image record at a time. There is thus competition with subsystems for memory access. When the information being processed is particularly uncertain or hard for a transformation to operate upon, a subsystem can enter buffered processing mode, whereby the information being copied into the image record is re-accessed by the transformation a few moments later, providing an integration of the information over the very short term. ICS theory allows only one buffer to be active at any moment, and so there is competition between subsystems for buffered processing mode. While the copy process within each subsystem gives rise to a sense of diffuse awareness for the information it is processing, buffered processing gives rise to a focal sense of awareness of the information. Because there is only one buffer active at any moment, focal awareness has a unitary nature, although its quality can vary depending on the level of representation being buffered.

Memory is a key resource in ICS, in terms of the individual components of the architecture (the image records of the subsystems, and procedural knowledge within the transformation processes, for example), and the dynamics of their interaction (the limited access to the image record within each subsystem, and especially the single focus of buffered processing within a configuration).


An ICS model of the El theory is a process-based account which specifies the memory resources that are required to be active from moment to moment in an episode of craving. The architecture within which any model is built contributes its own constraints and organisation to a less well specified theoretical account of a phenomenon. This specification helps us to be more precise and hence allows us to make detailed predictions about the interaction of craving with other cognitive activities, and in particular, to predict how other activities may interfere with craving. These predictions could perhaps have been made without an ICS, or any other, model of the theory, and once they have been made, they may seem obvious. Nevertheless, the fact that the process- based account did help in their derivation is in itself justification for the modelling process, and it also provides a rationale for the predictions. If the predictions are not supported by subsequent empirical investigation, the rationale can be inspected, and possible explanations examined step by step, to repair either the model (if it was an inaccurate representation of the theory) or the theory (if the model was accurate). Without a model, failure to support a theoretical prediction leaves us none the wiser about how to improve a theory. The triggers postulated as eliciting a craving episode in the El theory of craving could in principle arise at any of the nine levels of representation in ICS. External cues would be sensed and then represented at the visual or acoustic levels. Highly proceduralised knowledge in the visual-object and acoustic-morphono- lexical transformation processes would lead to perceptual representations being formed (at an object and a morphonolexical level respectively); these could then lead to propositional information being generated via object-propositional or morphonolexical-propositional transformations.

Alternatively, the visual-implicational and acoustic-implicational processes might generate information in the high-level schematic representation. This might lead to the implicational- somatovisceral process driving changes in the individual’s physiological state, which could be detected as body state representations and, via the body state-implicational transformation, lead to a change in mood. Similar chains of activity could be triggered by the body state subsystem detecting physiological deficits, or could be triggered in a purely cognitive manner, by processes that are not currently in use as part of the configuration supporting the individual’s primary task.

Processes that are involved in a primary task configuration, but which are momentarily inactive, can be caught up in secondary configurations. Depending on the richness of the representations available for processing, and the degree of pro- ceduralisation, cognitive activity can oscillate between primary and secondary tasks. If the secondary configuration is coherent enough, then a lacuna in the primary configuration can allow it to enter buffered processing and break through into focal awareness, in effect becoming the primary configuration. Subjectively, the individual will have had an intrusion into awareness of a thought not necessarily related to their previous chain of thought. This is the first specific contribution that ICS makes to the El model: the definition of the conditions under which nonconscious cognitive processing can become conscious. This is the moment at which a person becomes aware of a craving, in the El theory.


The next stage in a craving episode is for the intrusive thought to lead to reward or relief due to its similarity to the real substance. This is where the ICS model makes its second specific contribution, by a limitation on the pathways that can give rise to reward and relief. In ICS these affective states come about through the ehcitation of an implicational schema concerned with the substance, and the consequent somatovisceral changes. Intrusive representations will be most likely to result in reward and relief if they directly allow the formation of implicational representations, and only the three sensory subsystems and the propositional subsystem can do this. The object and morphonolexical subsystems cannot produce implicational representations directly, and so intrusions there would need to involve a two-step sequence by first creating a propositional representation, followed by a propositional- implicational transformation. Intrusions in the two effector subsystems would only reach awareness if they conflicted with ongoing behavioural output, resulting in an action slip (Norman, 1981) at the limb level or a Freudian slip at the articulatory level.

A third, and perhaps the most testable, contribution of the ICS model comes from the consequences of the activation of the implicational schema for a substance. If this substance is one that the individual has habitually used, then the existence of procedural knowledge would lead to the elaboration of propositional representations for obtaining and using the substance, followed by the generation of morphonolexical or object level targets controlling these behaviours. These correspond to the generation of verbal and visual images of the substance, respectively. Both are feasible, and which predominated would depend on the nature of the tasks involved in searching for and using the substance. In consequence, it can be predicted that where the task involved physical movement and acquisition of a discrete spatially locatable substance, a triad of implicational, propositional, and object representations would dominate thought, with focal awareness oscillating between mental visual images of the substance and the affective implications of its use. This functionally based emphasis on object-level visual imagery is consistent with the El model, but is not a necessary part of it, and so is a prediction drawn from the model of the theory (and hence from ICS) rather than from the theory itself.

Summarising this, the ICS account sees secondary configurations of cognitive activity arising as a result of unattended activity in the cognitive system. Processes that are not involved in the current primary task can generate a stream of processing that can break into awareness if the proceduralisation of associated processing is strong enough and the primary task is not demanding enough to retain processing resources. Crucial for the intrusion to gain control is the elicitation of an implicational level of representation concerning the craved substance, which will lead to both a somatovisceral response (if this has not already been part of the precursor activity) and mental imagery, which further supports the maintenance of the implicational representation. This activity places heavy demands on the two central subsystems, effectively taking control of thought and precluding other cognitive activity.


Tiffany’s (1990) account of craving is also based on a general distinction between automatic and non-automatic cognitive processing, although he steps back from any particular theoretical definition of automaticity. In general, he points out that non-automatic processes are seen as those that require cognitive effort and control for execution, are strategic, and involve conscious awareness of choices and execution. They are constrained by competition for cognitive resources, and so only a single sequence of non-automatic processing can be carried out at a single moment. In contrast, automatic processes develop through practice, can be carried out without cognitive effort, and are not resource limited. Importantly for Tiffany, they can be executed without requiring conscious control, and are in fact difficult to control, having an aspect of cognitive impenetrability. Repeated drug use leads to the development of action schemata: overlearned, automatic patterns of behaviour directed towards the acquisition and consumption of a particular substance. These schemata are elicited by certain patterns of external and internal sensory stimuli, and contain coordinated sequences of actions, organised with alternative sequences to overcome common obstacles. Included within the schemata are physiological responses that support the motor behaviour required, and the generation of physiological adjustments in anticipation of drug consumption. In practice, the schemata allow the user to obtain and consume their substance with minimum cognitive effort, without distracting them from other cognitive activity, and prepare them for the physiological demands and effects of drug use.

Tiffany’s action schemata for drug use do not inevitably lead to craving. Craving is a consequence of their failure to operate. This can be either because some situation arises in the course of their execution which is not covered by the learned set of alternative action sequences, or because the user is attempting to abstain from drug use, and so inhibiting their execution. Although the subsequent goals and behaviours differ, in both cases non-automatic cognitive processing becomes involved, and the user becomes aware of the situation. They will find that awareness of the physiological responses cued by the action schemata interferes with their nonautomatic, controlled behaviour, giving a subjective characteristic to craving episodes that is not present when other, non-drug related action schemata are blocked or inhibited.

In ICS, the development of action schemata is well specified. Automatic processes correspond to those for which a high level of proceduralised knowledge is available within transformation processes. Given an appropriate input representation, a transformation process can operate immediately and directly to produce an output representation, without needing to access its image record. When sequential transformation processes in different subsystems have been repeatedly engaged in a particular task, the whole sequence can become proceduralised. This has several consequences. First, the absence of image record access means that the schemata can be carried out simultaneously with other configurations of cognitive activity that use different transformations. Second, there is no need for any of the subsystems to enter into buffered processing, and so if other less well proceduralised processing is ongoing, the action schemata will not enter focal awareness. Third, over the course of proceduralisation, loops of cognitive processing will be replaced by direct processing, so that gradually the involvement of the central subsystems will become minimised. For example, a simple visuo-motor task might initially involve a sequence of visual to object, and object to limb transformations, with a loop between the object to propositional levels serving to clarify the identity of the object representation. With practice, the procedural knowledge in the object-limb transformation would develop to the point that it becomes able to operate directly on the representation provided by the visual-object process, and the need for involvement of the propositional subsystem would fade away, leaving a simpler and direct visual-object + object-limb configuration. This would be faster, would not compete with other tasks using the object-propositional transformation or the propositional system, and be even less likely to enter awareness (because there are fewer processes involved), but would be less flexible, and subjectively would be focused more on sensory than semantic attributes of the stimuli.

An action schema for drug use might have several eliciting pathways, from each of the three sensory subsystems, any of which could lead to the behavioural outcome independently. An internal change in body state might lead to a cascade from body state to implicational, to propositional, to object, to limb to make a smoker reach for the cigarettes in their pocket, for example; a visual trigger might lead to a simple visual-object + object-limb sequence as the user reaches for the seen stimulus, coupled with a visual-implicational + implicational-somatovisceral sequence to provide the physiological preparation for use. Internal cognitive triggers provided by other thoughts and associations could activate implicational schemata of drug use and trigger the same sort of cascade from implicational to limb as that cued by the change in body state. Longer and more involved patterns of behaviour involving complex task structures stored in propositional and implicational image records could also be automatically activated, although as Tiffany notes, these are less likely to be completely automated.

So far this is just a restatement of Tiffany’s theory in a particular framework, and little has been added to his account. What can now be done is to examine the consequences and implications of the patterns of cognitive activity that have been described in ICS. One apparent weakness within Tiffany’s theory is the question of how, if craving follows from the controlled inhibition of automatic processes, and those automatic processes can occur without requiring awareness, the controlled inhibition can be initiated. In other words, at what point does the user who is attempting to be abstinent become aware that they are executing an automatic action schema that is leading to drug use, and hence experience craving? The ICS account of Tiffany’s theory allows us to examine the ways in which controlled processing might be triggered.

A subjective sense of the self-control of behaviour arises in ICS when implicational and propositional representations are being alter nately buffered to bring the individual’s motiva- tional/affective state into focal awareness and to select a task structure consistent with the implied goals. During task execution, implicational representations arise from the active propositional representation controlling action, and from the sensory consequences of motor and speech behaviour (as well as from physiological changes in body state). A mismatch between these patterns of implications and the current implicational task schema can lead to an interruption and replanning of behaviour, via an implicational-propositional transformation.

This would allow a user who has the overall personal goal of remaining abstinent to recognise that their own overt behaviour may be leading towards drug use, just as they might notice the behaviour of another person. It also allows them to notice changes in their own physiological state as the schema prepares them for the activity of drug acquisition and the effects of consumption. However, it does not allow the user to notice the internal products of the cognitive processes that are leading to this behaviour. They cannot become implicationally aware of a highly proceduralised visual-object and object-limb sequence, nor of the object and limb representations that are produced, unless those two representations are transformed into implications, and there are no direct processes in ICS to do this. By definition, a proceduralised action schema occurs without requiring focal awareness, and so only the end products of the configuration are available for inspection.

A testable prediction that can be derived from the ICS model, then, is that people could become aware of the implicational representations that are driving changes in their physiological state, if they were engaged in cognitive activity that involved a buffer in the implicational subsystem (e.g., trying to make a qualitative judgement about something, using the implicational-propositional transformation). This is because all information arriving at a subsystem’s input array is copied into its image record, and so even information that is not necessarily relevant for the current primary task can enter awareness. This provides the only way in which the thought processes that are part of an automatic action schema could enter awareness before any overt behavioural or physiological change had taken place. The locus of this awareness is the implicational subsystem, and the content of the information that is entering awareness would be psychophysiological in nature because it is to be used to drive a somatovisceral output. The subjective impression of this trigger for a craving episode would be of a vague feehng of arousal associated with drug acquisition, or a pang of deficit associated with the anticipation of the drug effects.


In Tiffany’s theory and in the El theory, craving only occurs once the individual becomes aware of some aspect of cognitive activity. The ICS account defines what “becoming aware” consists of in processing terms, and locates the focus of subsequent activity at the implicational level of processing in both accounts. Tiffany’s theory hmits the nature of the ehciting information to that which has been procedurahsed as part of a configuration directed towards producing somatovisceral changes and, due to the nature of proceduralisation, one that is probably triggered by sensory cues. The El theory allows any activity that might generate the implicational model to result in a craving episode, but the ICS modelling provides a more explicit account of what happens once the implicational schema has been activated and entered awareness. Like Tiffany’s theory, the El model also relies on the idea of proceduralisation or automation of processing to predict why unattended processing tends to travel down well-worn routes, leading to thoughts of substance use. It goes further than Tiffany’s theory in detailing the content of the craving episode, especially emphasising the role of mental visual imagery as part of the craving episode, which in ICS terms is a sequence of reciprocal activity involving the implicational, propositional, and object levels of representation.

Tiffany emphasised that his theory allowed for absent-minded lapse in the absence of strong craving, that is, when an abstainer’s automated action schemata reached their conclusion without the abstainer noticing their activity and attempting to control their behaviour. The El theory is in accordance here (although it downplays the role of absent-minded lapses). Neither model expects craving to occur in a situation where the underlying automated cognitive activity does not reach awareness. The ICS models of these theories explain how such absent-minded, automatic activity can occur, by delineating the representations that would have to be involved in the lapse (predominantly visual, object, and limb to control motor behaviour) and those that would have to be heavily involved in a primary task (certainly the propositional and implicational, with no body state involvement).

A key advantage of operationalising the two theories within an information processing account is that it allows the processing steps that each theory requires to be specified, and thus gives an indication of the nature of other cognitive tasks that might interfere with craving, prevent it reaching awareness, or ameliorate its effects once it has occurred. ICS shows that the two theories make different predictions about which processes will inhibit craving. In Tiffany’s theory, ICS suggests that craving occurs when implicational representations that are part of an action schema intrude into buffered implicational-propositional processing, or when the user senses their overt or somatovisceral behavioural preparation for drug use. A primary task that requires buffered processing within a different subsystem (e.g., one in which the structure of sounds was in focal awareness), or which avoided the user detecting their preparatory behaviours, would presumably therefore inhibit craving. However, this would only last as long as the user was focused on the inhibiting task. As soon as they stopped attending to it, the craving could resurface. The ICS model of the El theory suggests that once craving has begun, it could be blocked by preventing the operation of the triad of implicational, propositional, and object subsystems. This could be done most directly by involving the propositional- object and object-propositional transformation processes in another task, e.g., a visual search task, or a visual imagery task. Once this inhibiting task ceases, the cycle of craving would have to be restarted by the generation of another intrusive thought from the antecedents. This might be more likely than before, because the recent substance- related cognitive activity will have left an increased level of activation in relevant transformation processes, but it will not be as likely as suggested by Tiffany’s theory.


The two theories differ most notably in the role of visual imagery as a key component of elaborated craving: it is central to the El theory, but not part of Tiffany’s theory. The ICS accounts show this to be due to Tiffany’s lack of emphasis on what happens during a craving episode. Tiffany’s theory, as expressed within ICS, implies that craving episodes would be triggered by the generation of somatovisceral responses to high-level schematic thoughts about a substance, or by an individual noticing their overt behaviour. These should be available for conscious inspection, leading to their attribution as triggering causes of a craving episode. The El theory does not assume that people have any direct access to the processes that triggered an intrusive thought, and therefore suggests that people may often have little insight these triggers. The El theory provides a richer account of what follows the trigger. These differences in emphasis between the theories could, in principle, be noticed without the modelling, but we argue that the modelling makes them readily apparent and provides grounded arguments to justify their reality. Once such differences have been noted, then they can be assessed empirically, leading to comparative assessment of the theories.

Tiffany’s theory is intended to describe the cravings of people dependent on habit-forming drugs, although it is plausible that the obstruction of action schemata could also cause the weaker cravings experienced by people who habitually use other, non-addictive substances. The El theory, on the other hand, explicitly defines addictive cravings as the result of the same set of processes that give rise to everyday desires. It is difficult to compare the models on this point, because little empirical work has been carried out into the nature of everyday desires, nor of the cravings for addictive drugs such as alcohol and tobacco by those who are not substance-depen- dent. We therefore conducted a preliminary survey of everyday cravings using a questionnaire designed to discover more about the subjective aspects of everyday cravings, and to evaluate the differing claims made by the two models.


We designed a simple questionnaire and included it in the University of Sheffield’s mailing to 1500 new students in September 2002. The questionnaire consisted of two sides of A4, and asked recipients to keep it nearby “until you find that you are craving something”. They then had to identify the substance that they were craving by circling one of food, tobacco, alcoholic drink, or soft drink (labelled “non-alcoholic drink” to include tea, coffee, water, etc.), and then specify the substance, and rate the strength of the craving (on a 10-point scale, with the anchors “very slight” and “overwhelming”). There followed a list of 12 potential causes that could have triggered the craving, and 10 statements that could describe the craving (see Table 1). Each had to be rated on a 5- point scale ranging from “not at all” to “definitely” describing the craving. The trigger statements were selected to cover the classes of antecedents identified by the two theories, including bodily sensations, different forms of mental imagery, and intrusive thoughts. The descriptive statements were similarly chosen to contain items that the models suggested should or should not be important.


Within 8 weeks of the mailing (by which time all of the recipients would have moved to Sheffield to start their studies) 361 completed questionnaires had been received, a response rate of 24%. These replies came from 155 males and 201 females (with 5 not specifying their gender). The median age of the 353 respondents who gave their date of birth was 19 years 8 months (ranging from 14 years 6 months to 55 years 4 months).


Percentage of respondents agreeing with potential trigger and descriptive statements as characterising their craving episode, sorted

by mean rating

    Mean ■ rating   Agreement percentage    
Questionnaire items Theory 1 2 3 4 5 z2
What triggered this craving?
I suddenly thought about it E 3.2 15 14 25 26 19  
I felt hungry /thirsty /tired /physical discomfort Both 3.2 23 13 15 17 31 ***
I imagined the smell / taste of it E 3.1 22 14 20 25 20 **
I pictured myself having it E 2.8 27 14 26 19 14  
I had nothing else to do / I was bored T 2.6 36 12 21 18 13  
I saw / heard / smelt it Both 2.1 61 5 8 14 12 **
Other things I was thinking about reminded me of it Both 2.0 51 19 13 12 5  
I felt stressed / anxious / sad E 1.9 59 13 14 9 5 **
I always have it at that time/place T 1.9 59 13 12 8 9 *
I felt happy Neither 1.8 55 19 15 8 3  
I was really busy Neither 1.3 79 11 6 3 1 **
I imagined the sound of myself having it E 1.3 85 6 6 3 1  
Descriptions of craving episode
I want it because I am hungry/thirsty/tired/in physical discomfort Both 3.3 24 8 17 16 34 ***
Having it would feel very comforting right now Both 3.1 18 14 25 24 20  
I am thinking of how much better I will feel after I have had it Both 3.1 18 15 22 25 20 *
I am imagining the taste of it Both 3.1 23 12 19 28 18  
I would feel more relaxed if I had it Both 3.0 22 14 26 21 17 **
I am visualising it E 2.9 23 15 24 23 15 *
КI don’t think about it, my craving will go away E 2.7 26 21 22 15 16  
I am trying to resist having it T 2.5 41 14 16 13 17 ***
I have it with me right now Neither 2.0 60 10 10 7 12 **
I can hear myself having it E 1.4 78 13 5 2 2  

Theory column indicates whether agreement with the item is consistent with the Elaborated Intrusion theory (E), Tiffany’s (T), Both, or Neither theories. Agreement ratings range from 1 (“not at all”) to 5 (“definitely”). Number of responses per row range between 345 and 356; percentages may not sum to 100 due to rounding. %2 tests for each item (df = 20; *p < .05; **p < .01; ***p < .001) indicate contingencies between ratings and craved substance.

Food cravings were reported by 219 people, tobacco cravings by 60, soft drink cravings by 59, and alcoholic drink cravings by 23. The food craving category was broken down by inspecting the specific substance that was reported. This allowed us to identify sub-groups of cravings for chocolate (76 people) and other snacks (75 people), with the remaining (68 people) reporting cravings for main meals, breakfasts, or not specifying any particular substance. The strength of craving scores (reported by 306 people) did not differ significantly between these six craved substance groups, F{5,300) = 0.50, Mse = 4.14, ns, with an overall mean of 5.5 and a standard deviation of 2.0 (every point on the 10-point scale was used; the mode was 7, used by 81 people, with another 59 using 4, giving a somewhat bimodal distribution). It is notable that the alcohol and tobacco cravings were not significantly stronger than the other cravings, despite the fact that usual levels of cigarette smoking involve significant physical dependence and reported craving. This finding supports the notion that everyday craving for targets other than addictive substances may often be of similar intensity to the levels that are produced by at least the everyday experience of cigarette smoking.

Of the 12 potential triggers, 7 received predominantly low ratings, with more than half of the respondents selecting “not at all” (see Table 1). These include external and internal cueing (“I saw/heard/smelt it” and “Other things I was thinking about reminded me of it”), and negative affective mood (“I was stressed / anxious / sad”), as well as a “habit” statement (“I always have it at that time / place”). One other (“bored”) had a modal response of “not at all”, but attracted sufficient higher ratings to produce a mean of 2.6 on the 5-point scale, with 52% giving it a rating of 3 or higher. In order of mean rating score, the four triggers that were typically thought to have caused cravings were that the person “suddenly thought about it”, “felt … discomfort”, “imagined the taste/smell of it”, or “pictured myself having it”. The overall picture is that cravers tend not to attribute their thoughts about a substance to any identifiable cue in the environment, but report them as spontaneous (in accordance with the El theory), or due to somatovisceral sensations (in accord with both theories) and involving olfactory or visual imagery (in accordance with the El theory). Contrary to Tiffany’s theory, cravers do not report experiencing craving in situations where overlearned action schemata might be expected to be activated, such as when they are in habitual usage situations, or when they notice an external or internal cue.

The trigger and descriptive statements all contained response patterns that were highly skewed to one or other end of the 5-point scale for at least one of the substance groups, and so parametric analyses are not suitable. Chi-square tests were therefore used to compare the distribution patterns for each statement to see if there were any differences in response pattern according to craved substance. These showed that of the four highly rated trigger statements, only the ratings for somatovisceral sensations and olfactory imagery differed according to craved substance, with tobacco cravings being rated as triggered by them less often than were soft drinks and the foodstuff cravings (see Figure 3). If the tobacco cravers represent an analogue to a substance-dependent population (which, particularly in the case of nicotine, is not an unreasonable assumption), the ICS model of Tiffany’s theory would imply that the opposite pattern should be found, because the physiological changes preparing the person for acquisition and use of tobacco would be more likely to be a trigger for these substances than others. Instead, the somatovisceral sensations and imagery associated with taste and smell are more commonly seen as reasons for craving non- addictive food, snacks, and soft drinks.

The statements describing the cravings generally received higher levels of affirmation than those describing the triggers of the craving. Only two were rated as “not at all” descriptive by more than half of the respondents: “I have it with me right now” and “I can hear myself having it”. Four of the five receiving the highest ratings all involve some aspect of physiological relief or reward. The three imagery questions produce markedly different patterns of response: while 65% of respondents rated themselves as “imagining the taste” using 3 or higher, and 61% rated themselves as “visualising it”, only 9% rated themselves as “hearing it”. Auditory imagery of the craved substance does not seem to play a role, while visual imagery (which might be involved in locating the substance, according to the ICS model of the El theory) and olfactory imagery (which might be a continuing part of the triggering processing, according to both theories) do emerge as part of everyday cravings.

According to chi-square tests, four statements showed similar patterns of response across all craved substance categories: “Having it would feel very comforting right now”, “I am imagining the taste of it”, “If I don’t think about it, my craving will go away”, and “I can hear myself having it”. The others produced response patterns that tended to depend on the substance being craved, and suggest that somatovisceral sensations are not as important in tobacco and alcohol craving as in cravings for other substances. “I want it because I am hungry / thirsty / tired / in physical discomfort” was rated 3 or higher less often by the alcohol and tobacco cravers, than by the soft drink and foodstuff groups, with a pattern of responses that was understandably almost identical to that for the equivalent trigger statement. In fact, the modal response for these alcohol and tobacco groups is“not at all”, but this may be because of the lack of a specific word in the language (and hence in the questionnaire) describing a deficit for nicotine or alcohol, comparable to “hungry” or “thirsty”. On the other hand, these two groups clearly have positive expectations about the effect that their substance would have on their affective state: “I would feel more relaxed if I had it” was rated 3 or higher by 87% of those craving alcohol and 84% craving tobacco. However, “I am thinking of how much better I will feel after I have had it” was only particularly descriptive of those craving a soft drink (34% giving it the top rating of “definitely”).



’I imagined the smell /taste of it’


I felt hungry/thirsty/tired / physical discomfort’


Figure 3. The percentage of responders giving a rating of 3 or above for the somatovisceral imagery and sensation triggers, showing lower affirmation from tobacco and alcohol cravers.




“I am trying to resist having it” was rated highly by 73% of those craving chocolate, and 66% of those craving tobacco, but by only 31% of those in the other substance groups. “I have it with me right now” was only rated highly by tobacco cravers, of whom 28% answered “definitely”, almost twice as many as the next highest groups. Perversely, tobacco cravers are highly Ukely to be trying to resist using their substance, yet are more likely to carry it with them. Tiffany’s theory predicts that craving occurs when an action schema is blocked, through self-control or situational factors, yet of our sample, 17 cravers gave a rating of “not at all” for “trying to resist” and “definitely” for “have it with me”. Tiffany’s account cannot explain these individuals’ cravings, yet their mean craving rating was no weaker than that of the whole sample (with a modal value of 7).

The only imagery description that showed a contingency on substance craved was “I am visualising it”, and this produced fairly equal ratings in the range 3 or higher of between 54% and 67% for all substances, but with more “definitely” ratings for alcohol, snack, and soft drink cravers, compared to the other groups. The difference here seems to lie in the relative strength of the image, rather than in the presence or absence of visual imagery. This is consistent with the El claim that craving is equivalent to everyday desire, and is not a feature specific to substance dependence.

It must be stressed here that these descriptions of the triggers and nature of craving episodes are the self-reports of a non-clinical sample. As self- reports, they are open to the usual criticisms of introspective data, especially in that they might reflect what people expect their craving to be caused by or feel like, rather than what craving really does feel like; they might also be influenced by the social desirability of giving particular answers. We tried to avoid these possibilities by asking them to wait until they actually were experiencing a craving, so that they would be more likely to focus on the subjective feelings at that moment, rather than on stereotypical assumptions or expectations, and by allowing the responses to be completely anonymous. Nevertheless, such influences will inevitably remain in investigations into the nature of any subjective phenomenon. The use of a non-clinical sample is partly due to convenience, but it is also influenced by our view that clinical cravings are an extreme form of everyday desires. Those who do not accept this can certainly argue that the craving of an addict is phenomenally different from the cravings that our respondents were experiencing, despite the inclusion of tobacco cravers within our sample. It would indeed be interesting to repeat this kind of study on a substance-dependent population, were a large enough sample available.


The information processing approach of ICS has helped us to identify some specific processing steps that could be involved in two models of craving that include cognition as a key explanatory construct for craving. Although Tiffany’s theory does not address the subjective content of a craving episode, the ICS model of his theory allows us to infer what this ought to be. Similarly, while the El theory allows many possible triggers, its ICS model allows us to infer how each of them might operate in practice. For Tiffany’s theory, action schemata are primarily proceduralised processes initiated by a sensory representation, and include an implicational-somatovisceral transformation that induces bodily changes preparatory for substance use, which can be detected by a body state-implicational transformation. For the El theory, the defining character of a craving episode is a reciprocal interaction between the implicational, propositional, and object representations, reflecting the central role of imagery in craving, with the implicational-somatovisceral and body state-implicational transformations providing the affective aspect.

Once these steps were identified, it became possible to evaluate them systematically through an empirical assessment of the nature of everyday cravings. Overall, the questionnaire data suggest that people have little insight into the precursors of a craving episode, reporting them as spontaneous images or somatovisceral sensations that pop into their heads. Once they begin to crave, they experience visual and olfactory images of their desired substance, and anticipate the pleasurable aspects of using it. This is entirely consistent with the El theory. The modelling of Tiffany’s theory is supported by the focus on somatovisceral triggers, but these are reported less strongly by tobacco cravers, which is not what would be expected if Tiffany’s account were correct. The data from this sample of non-substance- dependent people is more consistent with the ICS model of the El theory than with the ICS model of Tiffany’s theory.

This study illustrates the benefits of a cognitive account outlined at the start of this paper: by breaking a mental behaviour down into its component parts, we can not only reason about what might be happening during that behaviour, but can examine different speculations about its causes and effects, and devise specific ways of measuring, intervening in, and redirecting mental behaviour. In the case of craving, if further work substantiates the role of imagery, then interventions targeting the same cognitive processes as those necessary for the construction of mental visual imagery could play a role in supporting those attempting to withstand craving for addictive and non-addictive substances.


Intrusive images and intrusive thoughts as different phenomena: Two experimental studies

Muriel A. Hagenaars

Leiden University, The Netherlands

Chris R. Brewin

University College London, UK

Agnes van Minnen

Overwaal Centre of Anxiety Disorders, Nijmegen, and Radboud University Nijmegen,

The Netherlands

Emily A. Holmes

University of Oxford, UK

Kees A. L. Hoogduin

Radboud University Nijmegen, The Netherlands

According to the dual representation theory of PTSD, intrusive trauma images and intrusive verbal thoughts are produced by separate memory systems. In a previous article it was shown that after watching an aversive film, participants in non-movement conditions reported more intrusive images than participants in a free-to-move control condition (Hagenaars, Van Minnen, Holmes, Brewin, & Hoogduin, 2008). The present study investigates whether the experimental conditions of the Hagenaars et al. study had a different effect on intrusive thoughts than on intrusive images. Experiment 2 further investigated the image-thoughts distinction by manipulating stimulus valence (trauma film versus neutral film) and assessing the subsequent development of intrusive images and thoughts. In addition, both experiments studied the impact of peri-traumatic emotions on subsequent intrusive images and thoughts frequency across conditions. Results showed that experimental manipulations (non-movement and trauma film) caused higher levels of intrusive images relative to control conditions (free movement and neutral film) but they did not affect intrusive thoughts. Peri-traumatic anxiety and horror were associated with subsequent higher levels of intrusive images, but not intrusive thoughts. Correlations were inconclusive for anger and sadness. The results suggest intrusive images and thoughts can be manipulated independently and as such can be considered different phenomena.

Keywords: Intrusions; PTSD; memory.

The dual representation theory of post-traumatic memory systems regulate information processing stress disorder (PTSD) proposes that two separate after trauma (Brewin, 2001; Brewin, Dalgleish, & Joseph, 1996). One memory system (verbally accessible memory or VAM) handles verbally accessible trauma information. To be stored in this system, information requires sufficient conscious processing. In VAM information is placed into context, provided with higher-level meaning, and made available for narration and deliberate cognitive appraisal. VAM is thought to be associated with experiencing memories as belonging to the past, and includes both “primary” and with “secondary” emotions. Primary emotions are considered to occur at the time of the event, whereas secondary emotions, like anger and sadness, would be generated during and after the event, by retrospective cognitive appraisal and reflection of consequences and implications (Brewin & Holmes, 2003). The second memory system (situationally accessible memory or SAM) results from lower- level perceptual processing of both the traumatic scene and the person’s bodily response to it. Trauma memories that are processed this way are therefore difficult to retrieve intentionally and full of sensory impressions. Intrusive images involving reliving originate in this latter memory system, and are thought to be inhibited by corresponding representations in the VAM system. SAM is thought to be associated predominantly with primary emotions, like fear and horror, although secondary emotions may be included if the trauma leaves time for more complex evaluations. Similar distinctions in perceptual and conceptual processing are also found in other models of PTSD (Ehlers & Clark, 2000) and autobiographical memory models (Conway & Pleydell-Pearce, 2000; see also Holmes & Bourne, 2008).

Although the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM- IV, American Psychiatric Association, 1996) states that in PTSD re-experiencing can take the form of “distressing recollections of the event, including images, thoughts or perceptions” (p. 428), there is some uncertainty about what causes these different types of recollection. Some studies indicate that visual intrusions are quite common and intrusive thoughts are relatively rare (Ehlers et al., 2002; Mellman & Davis, 1985), whereas others have found thoughts to be more prominent (Reynolds & Brewin, 1998). However, some people in the latter study experienced intrusions as a combination of thoughts, images, and flashbacks. Assessing the occurrence of intrusive thoughts is complicated, because they are not always distinguished from rumination, worry, and deliberate evaluative thoughts about the trauma. Moreover, images are sometimes included in the assessment of intrusive thoughts. In the present study “intrusive thoughts” refers to intrusive verbal thoughts without images. The distinction between intrusive images and intrusive thoughts is theoretically interesting. That is, although named intrusions, the language-based nature of intrusive thoughts may indicate conscious appraisal at some point and may thus implicate dominant processing in verbally accessible memory.

Lang’s bio-informational theory suggested a special link between mental images and emotion, especially fear (Lang, 1979), and subsequent studies did indeed show that imagining emotional sentences or scripts was related to greater physiological activity (Miller et al., 1987; Vrana, Cuthbert, & Lang, 1986). Unfortunately, however, imagery was not compared to linguistic processing, leaving open whether mental imagery is uniquely linked to emotion. Holmes and Mathews (2005) therefore compared imagining several verbal descriptions with thinking about their meaning. Descriptions in the imagery condition were associated with more anxiety than those in the verbal condition. Convergent evidence for the powerful impact of imagery on emotion has been found using pictorial stimuli (Holmes, Mathews, Mackintosh, & Dalgleish, 2008). However, to our knowledge, the reverse relationship (between emotion and subsequent development of mental images) has not been studied yet.

Recent research exposing non-clinical samples to trauma films (see Holmes & Bourne, 2008, for a review) has shown that increased verbal reports of dissociation and reduced heart rate during the film are associated with an increase in later intrusive memories (Holmes, Brewin, & Hennessy, 2004). These responses may reflect the tendency for some species to freeze in response to danger. Freezing is suggested to enhance attentional orienting towards threat, and is indeed associated with bradycardia (Lang, Bradley, & Cuthbert, 1997). Witnessing a traumatic film under the instruction not to move may be a useful analogue to a dangerous encounter with a predator in which escape is not possible. Despite the advantages from an evolutionary perspective, freezing may become maladaptive, for example by provoking feelings of unpredictability and uncontrollability (important factors in the development of PTSD; Foa, Zinbarg, & Rothbaum, 1992), or by becoming a conditioned response to non-threatening stimuli.

Non-movement may thus cause a shift from VAM towards SAM processing because of enhanced perceptual processing. Consistent with this, a non-movement condition was followed by increased intrusive images compared to a control condition (Hagenaars et al., 2008).

In conclusion, it has been suggested that intrusions may develop as a result of memory disturbing reactions during the trauma, like peri-traumatic dissociation (Ozer, Best, Lipsey, & Weiss, 2003), or non-movement (Elbert & Schauer, 2002). Thus far, peri-traumatic reactions have not been related to intrusion type, like being image based or thought based. In the following two experiments we investigated whether intrusive images and intrusive thoughts are indeed two distinctive types of intrusions that function relatively independently.

In a previous article we reported that participants who did not move while watching a traumatic film experienced more intrusive images in the subsequent week than participants who were allowed to move freely (Hagenaars et al., 2008). That article focused on the effect of non-movement on intrusive images, analogous to re-experiencing in PTSD. We now investigate whether peri-traumatic non-movement in the previous study had a different effect on the development of intrusive images than on intrusive thoughts. Peri-traumatic nonmovement (being related to traumatic events and threat) may enhance storage in the sensory-rich SAM, which could explain the increased frequency of intrusive images (Hagenaars et al., 2008). If non-movement does indeed cause a shift from verbal towards perceptual processing, an imbalance between intrusive images and intrusive thoughts should characterise the non-movement groups. A second experiment was set up to investigate whether intrusive images and intrusive thoughts are also differentially affected by the intensity of the event. In addition, we tested the prediction from the dual representation theory and bio-informational theory that higher levels of SAM-associated peri-traumatic emotions, especially anxiety and horror, would be associated with an increased frequency of intrusive images but not intrusive thoughts.



The data are derived from an experiment (described elsewhere; Hagenaars et al., 2008), set up to investigate the effect of dissociative paralysis on intrusion development, with deliberate nonmovement and free movement as control groups. As non-movement per se, and not the dissociative component, was responsible for the effect, we compared the combined (dissociative f deliberate) non-movement group with the free-to-move group in the present study. For a detailed description of the methodology we therefore refer the reader to that article. In the present article we will describe the methodology only briefly for reasons of clarity.

Participants. A total of 79 healthy students participated, after 10 students were excluded for meeting criteria of a DSM-IV (blood phobia, depressive disorder and drug abuse; n = 9) or failure to comply with the instructions during the experiment (n = 1). Students were randomly assigned to one of three experimental conditions: deliberate non-movement (n = 25), dissociative non-movement (n = 27), and a free-to-move control condition (и = 27). A total of 11 participants (13.9%) were male and 68 (86.1%) were female. The age range was from 18 to 29 years (M = 21 years, SD = 2.2).

Material. A 10-minute film depicting four traumatic scenes of real-life footage of the horrible aftermath of road traffic accidents was used to model a traumatic experience (see also Hagenaars et al., 2008; Holmes et al., 2004). A brief commentary introduced each scene, providing background information about the accident and people involved. To prevent fatigue from sitting still too long, one scene (having the lowest distress ratings as judged by independent raters) was removed from the original five scenes compiled by Steil (1996). A DVD recorder was used, and a projector (3M, type MP8745) to project the film onto a 113 x 88 cm screen.

Measures. For exclusion purposes, psychiatric symptoms were assessed by the Structured Clinical Interview for DSM-III-R (SCID-I; Spitzer, Williams, Gibbon, & First, 1992). The SCID-I is a standardised, semi-structured, diagnostic interview for diagnosing DSM-IV psychiatric disorders, which has good reliability (overall kappas were .61 for current and .68 for lifetime diagnosis). The 12 screening questions of the SCID-I were used initially. Diagnoses were established using the relevant SCID-I sections if participants endorsed symptoms. SCID-I interviews were conducted by licensed clinical psychologists.


At pretest, participants rated anxiety, horror, anger, and sadness on an 11-point scale from 0 (not at all) to 10 (extremely) to show the film’s impact on their emotional state. “Peri-traumatic” mood (anxiety, horror, anger, and sadness during and immediately after the film) was rated postfilm in the same way.

For 7 days after the film, participants recorded every intrusion of the film using a tabular diary (see also Brewin & Saunders, 2001; Davies & Clark, 1998; Holmes et al., 2004). To check the intrusive character, participants had to describe the content of each intrusion, the degree of distress it caused, whether it was image or thought based (or both), and how spontaneous it was. Verbal and written instructions were given about the nature of involuntary intrusions and how to keep the diary. As in previous studies, intrusions were described as unintended, “spontaneously occurring” rather than deliberate memories of the film; they were to interfere with ongoing activity and easily capture attentional resources (Clark & Rhyno, 2005). Participants were also asked to discriminate between evaluative thoughts and rumination versus intrusive thoughts, and to record the latter only. The importance of recording every intrusion was emphasised. They were instructed to carry the diary with them and check whether they had completed their diary at regular times each day.

The total number of intrusive images and the total number of intrusive thoughts were calculated by the experimenter (blind to group membership), by adding up all intrusive images and thoughts respectively. Intrusions that contained images as well as thoughts were scored as image- based intrusions, because we wanted to compare intrusions with a sensory component with intrusions without a sensory component (i.e., intrusive thoughts alone). Following Davies and Clark (1998), to check diary compliance at follow up, participants rated how often they forgot or were unable to record intrusions from 0 (not at all true) to 10 (extremely true).

Procedure. Participants who passed screening gave written informed consent, completed the pre-film questionnaires, and were instructed about the manipulation. Participants in the dissociative non-movement condition were told a special technique would make them unable to move (catalepsy);1 participants in the dehberate non-movement condition were told they were not allowed to move and should sit as still as possible; and participants in the control condition were told they could sit however they wanted to and move as much as they wished. All participants were given standardised instructions on how to watch the film. A video camera recorded the participants discreetly to check for adherence to the instructions. After the film had ended, participants in the dissociative non-movement condition were instructed to shake their arms and legs to get out of their cataleptic state. The dehberate non-movement participants were told they were allowed to move again, and the control group was told the film had ended. All participants then completed “peri-traumatic” mood ratings.2 They were debriefed using a standardised debriefing form, which explained that they might or might not experience intrusions. It was explained what is considered an intrusion, and the difference between intrusive images and intrusive thoughts was defined. The distinction between intrusive thoughts and rumination or evaluative thoughts was thoroughly clarified.


As reported elsewhere (Hagenaars et al., 2008), mood ratings significantly increased from pre to post, i(78) > 5.3 and p < .001 in all cases. In particular horror (PTSD criterion A2 in the DSM-IV) showed a large increase from pre (M = .57, SD = 1.37) to post (M = 5.80, SD = 2.26), indicating that the film was distressing, and quite horrifying. These scores seem somewhat higher than those in other studies using the trauma-film paradigm (e.g., Holmes et al., 2004).

Group differences in intrusion types. ANOVA analyses were conducted to study the frequency of intrusive images and thoughts in the combined non-movement groups versus the control group. These revealed a significant condition x intrusion type interaction effect, F(l, 73) = 9.33; p = .003 (see Figure 1). That is, participants who did not move during the film reported more intrusive images and tended to have fewer intrusive thoughts than participants who could move freely. Tests of simple main effects showed an effect for images, F(1, 73) = 5.23; p = .02, and a trend for thoughts, E(l, 73) = 2.93; p = .09, confirming this difference in intrusion frequency between free-to- move and non-movement groups.

In addition, paired /-tests showed that the number of intrusive images was significantly higher than the number of intrusive thoughts in the non-movement groups, /(48)=–4.11, p = .001. However, the frequency of intrusive images and intrusive thoughts did not differ for participants in the control group, /(25) = 0.82, p = .42.

Intrusion types and emotion. To investigate the relation between intrusion type and emotion, partial correlations were conducted, controlling for the effect of experimental condition (i.e., movement versus non-movement). These revealed that greater peri-traumatic anxiety, horror, and anger were positively related to the frequency of intrusive images (r = .31, p = .008; r = .28, p = .01; and r=.24, p = .03 respectively). Sadness was not significantly related to the frequency of intrusive images (r = .18, p = .12). With respect to intrusive thoughts, there was a non-significant trend towards a negative association with peri-traumatic horror (r= –.21, p = .06). Peri-traumatic anxiety, anger, and sadness were unrelated to the frequency of intrusive thoughts (all rs < – .06 and all ps > .61).


Figure 1. Intrusive thoughts and images in the free-to-move and non-movement conditions.


Our previous study (Hagenaars et al., 2008) showed that non-movement, an experimental condition designed to interfere with the processing of traumatic information, led to an increase in intrusive images. Our further analyses have shown non-movement also tended to lead to a decrease in intrusive thoughts. A closer inspection of the data showed that in the non-movement groups the frequency of intrusive images was significantly higher than the frequency of intrusive thoughts, whereas the frequency of the two types of intrusions was not different in the free- to-move control group.

Our second hypothesis was partly confirmed. That is, peri-traumatic anxiety, horror, and anger were related to an increased frequency of intrusive images. Peri-traumatic sadness was not, although this could be a power problem given that the correlation was moderate and almost reached a trend level in significance. In contrast, none of the emotions was related to the frequency of intrusive thoughts. Moreover, peri-traumatic horror tended to be associated with a decreased frequency of intrusive thoughts.


Experiment 2 was set up to replicate these findings under different circumstances. That is, we now aimed to investigate how stimulus content (traumatic versus neutral) would differentially affect the development of intrusive images and thoughts. In Experiment 2 the film itself and not the manner of watching the film was manipulated. That is, we studied whether a trauma film affected intrusive images and intrusive thoughts differently from a neutral film. As a trauma can cause a shift from verbal towards perceptual processing we hypothesised that, in general, a trauma film would lead to higher levels of intrusive images, but not thoughts, while a neutral film would not provoke either images or thoughts. As in Experiment 1, it was tested whether peri- traumatic emotions were associated with the frequency of intrusive images and thoughts.



Participants. A total of 52 healthy students participated in the second study. Students were randomly assigned to one of two experimental conditions: Neutral film (n = 23), or Trauma film (и = 29). A total of 9 participants (17.3%) were male and 43 (82.7%) were female. The age range was from 18 to 32 years (M = 20.6 years, SD = 2.3).

Material. The film described in Experiment 1, depicting traumatic scenes of real-life footage of the horrible aftermath of road traffic accidents, was used to model a traumatic experience. Another film depicting neutral traffic situations was composed to model a neutral event. The neutral film matched the trauma film in content (traffic- related scenes), number of scenes, duration of scenes, and overall duration. This neutral film consisted of scenes like cars driving by, and traffic crossing a junction. Like the trauma film, a brief commentary introduced each scene, providing background information about the traffic scene. As in Experiment 1 a DVD recorder was used, and a projector (3M, type MP8745) to project the film onto a 113 x 88 cm screen.

Measures. As in Experiment 1, psychiatric symptoms were assessed by the Structured Clinical Interview for DSM-III-R (SCID-I; Spitzer

et al„ 1992).

Pre-film and peri-traumatic mood were assessed by an 11-point scale as in Experiment 1. Intrusions were recorded for 7 days in a tabular diary, as in Experiment 1. Instructions were also similar to Experiment 1.

Procedure. The procedure, including assessments and instructions, was similar to Experiment 1 except for the stimuli (trauma film and neutral film). Because non-movement was associated with higher intrusion frequency, participants in both the trauma and the neutral condition were instructed to sit still during the film.


Mood ratings significantly increased from pre to post, <(51) > 2.7 and p < .01 in all cases. In particular, horror (PTSD criterion A2 in the DSM-IV) showed a large increase from pre (M = .69, SD = 1.14) to post (M = 5.59, SD = 2.13), indicating that, as in Experiment 1, the trauma film was perceived as distressing, and quite horrifying. ANOVA analyses confirmed the neutral film was followed by fewer emotional reactions than the trauma film, F(l, 52) > 15.40 and p < .001 for all emotions.

Group differences in intrusion types. ANOVA analyses were conducted to study the frequency of intrusive images and thoughts in the trauma condition versus the neutral condition. These revealed a significant condition x intrusion type interaction effect, F(l, 50) = 5.98; p = .02 (see Figure 2), and a main effect for intrusion type, F(l, 50) = 5.18; p = .03. Tests of simple main effects showed an effect for intrusive images, F(l, 50) = 5.20; p = .02, but not for intrusive thoughts, F(l, 50) = 0.21; p = .65, showing that groups differed in intrusive images but not thoughts.

In addition, paired г-tests showed that the number of intrusive images was significantly higher than the number of intrusive thoughts in the trauma-film group, t(28) = 3.08, p = .005. Participants who watched the neutral film, on the other hand, did not experience more intrusive images than intrusive thoughts, t(22) = – 0.15, p = .88.

Intrusion types and emotion. To investigate the relation between intrusion type and emotion, partial correlations were conducted, controlling for the effect of experimental condition (i.e., trauma film versus neutral film). These revealed that greater peri-traumatic anxiety, horror, and sadness were positively related to the frequency of intrusive images (r=.26, p = .06; r = .33, p = .02; and r=.34, p = .01 respectively). Anger was not significantly related to the frequency of intrusive images (r=.13, p = .36). With respect to intrusive thoughts, peri-traumatic anxiety, horror, sadness, and anger were all unrelated to the frequency of intrusive thoughts (all rs < .18 and allps> .18).


Figure 2. Intrusive thoughts and images in the trauma film and neutral film conditions.


Using a different manipulation in Experiment 2 we found a similar pattern of findings to Experiment 1 in that stimulus type had a differential effect on the development of intrusive images and intrusive thoughts. Intrusive images were higher in the trauma condition relative to the neutral condition, whereas the frequency of intrusive thoughts did not differ between conditions. As in Experiment 1, heightened levels of peri- traumatic mood were related to the later development of intrusive images and not thoughts. However, this was now the case for anxiety, horror, and sadness, but not for anger.


Both experiments reported here found that intrusive images and intrusive thoughts are developed under different circumstances. That is, nonmovement (versus free movement) provoked intrusive images but not thoughts, and the same was true for a trauma (compared to a neutral) film.

According to the dual representation theory of PTSD, verbal thoughts arise from a verbally accessible memory system that is distinct from a situationally accessible memory system supporting intrusive images (Brewin, 2001; Brewin et al., 1996). Despite their intrusive character, our data suggest that intrusive thoughts and intrusive images do indeed arise from independent memory systems. A similar distinction at the level of processing, proposing that images reflect more data-driven and appraisals more conceptual processing, forms an important part of the cognitive model of PTSD (Ehlers & Clark, 2000) as well as non-clinical models of autobiographical memory (Holmes & Bourne, 2008). These new data further suggest the presence of two independent, although not necessarily competing, memory (or processing) systems. Under normal conditions, here free movement or a neutral event, both systems appear to be operative. However, conditions that favour one system or process do not necessarily affect the operation of the other, resulting in either a dominant encoding of sensory components (hence higher frequency of intrusive images) or a hampered dominance of verbal encoding (hence no increase in the frequency of intrusive thoughts). Our data therefore indicate that imagery-based and verbally based intrusions can be generated independently, and may develop under different circumstances.

The second goal of both experiments was to study the effect of several peri-traumatic emotions on the development of intrusive images and intrusive thoughts. Peri-traumatic anxiety and horror proved to be associated with a subsequent higher frequency of intrusive images, but not intrusive thoughts. The results were inconclusive for peri-traumatic anger and sadness in that these were related to intrusive images in only one of the experiments although, like anxiety and horror, neither was related to intrusive thoughts. Peri- traumatic horror tended to be associated with a decreased frequency of intrusive thoughts, but this was not replicated in Experiment 2. The inconsistency in anger and sadness leaves us with two possibilities: (1) all emotions, including anger and sadness (although possibly not as strongly as anxiety and horror), are associated with mental imagery and this was not consistently found because of a lack of power or differences in emotional intensity, or (2) anxiety and horror, but not anger and sadness, are associated with the formation of intrusive images, which is why the anger and sadness correlations were not replicated.

In considering the first option, it has indeed been suggested that imagery can provoke strong emotions (Holmes & Mathews, 2005). The finding of peri-traumatic emotions being associated with a higher frequency of intrusive images could be an expression of a more general two-way relationship between mental imagery and emotion. It has been suggested that mental imagery is effective in eliciting emotion (Holmes & Matthews, 2005). Our data imply that this relation may also work the other way around: namely, that peri-traumatic emotions have an impact on the subsequent development of intrusive images (and not intrusive thoughts). Any differences between the various emotions in the emotion-intrusive image association could be due to different levels of physiological arousal or to their neural basis. Appropriate studies that would allow these possibilities to be tested have not yet been conducted.

The second option suggests a special link between anxiety and horror and subsequent higher levels of intrusive images. Brewin et al. (1996) indeed suggested such a link, in that intrusive images contain emotional reactions produced immediately by the traumatic situation but not emotions that are the result of subsequent reflection and appraisal, like sadness. Previous research with clinical samples also suggested that there might be a specific relationship between the emotion of fear and re-experiencing (Hellawell & Brewin, 2004; Reynolds & Brewin, 1999), although other studies indicate that flashbacks in patients with PTSD can contain moments of intense sadness or guilt which appear to be congruent with the peri-traumatic experience (Grey & Holmes, 2008). Our results, then, do support previous research suggesting that emotions in PTSD show marked variation in their association with the direct experience of trauma and in their stability over time (Hellawell & Brewin, 2004; Speckens, Ehlers, Hackmann, Ruths, & Clark, 2007). It may thus be that more fundamental properties of emotions–like evoking arousal and indicating self-relevance and action and motivational tendencies–dictate whether they lead to intrusive images, although further research is needed.

Interestingly, only emotions that appear most relevant in dangerous situations (anxiety and horror) were associated with intrusive images. Lang (1979) suggested that mental images include action functions and proposed a special link between mental imagery and fear (relevant to the flight response). Evolutionarily, it may have been relevant for survival to have vivid visual memories of dangerous situations. It has been suggested that emotional systems may be more strongly associated with imagery than language systems because the latter evolved later than basic emotions (Holmes & Matthews, 2005). The fact that anger and sadness were associated with higher levels of intrusive images in only one of the experiments could be due to these not being flight emotions relevant to survival, or may simply be due to the intensity of the emotional experience. An explanation for the absence of a relation between VAM-related emotions (anger, sadness) and intrusive thoughts may be that the participants did not appraise or reflect on the film because of its lack of self-relevance. That is, events seem to be associated with later distress especially if perceived as central to one’s life (Berntsen & Rubin, 2007). As the results on anger and sadness were not consistent across experiments, however, future research should address whether flight-emotions only or emotionality in general causes a shift from conceptual towards perceptual information processing.

Our study was limited by a number of factors, including the assessment of the specific emotions (using Likert scales), and the relative frequency of intrusive images and thoughts recorded. Also, the design was adequate for establishing the effects of condition on intrusion frequencies, but does not allow statements about the mutual influence of intrusive images and intrusive thoughts. For example, the presence of intrusive images could have hindered the development of intrusive thoughts. This seems unlikely however, as intrusive images did not occur during encoding or shortly after the film, but during the entire week afterwards, and they did not occur so frequently that they would have hindered reflection or appraisal processes. However, future research should address this issue.

The results are in need of replication in this and in related paradigms. Nevertheless, the data are supportive of the theoretical proposal that intrusive images and intrusive thoughts are dissociable after an analogue traumatic experience and, importantly, are also differentially associated with emotion. The present study indicates that more intense peri-traumatic emotions are more Ukely to generate intrusive images than thoughts, but of course the design does not allow statements on exactly how information was processed during the film. It would be interesting to explore related issues like emotion-regulation strategies and viewer perspective with respect to the development of intrusive images and thoughts. Understanding these basic processes is likely to be valuable in formulating more effective and more scientifically based treatments for PTSD.

Manuscript received 9 April 2009 Manuscript accepted 4 November 2009


Mental Imagery and Emotion in Treatment across Disorders: Using the Example of Depression

Emily A. Holmes, Tamara J. Lang and Catherine Deeprose

Department of Psychiatry, University of Oxford, Oxford, UK

Abstract. Abnormalities in mental imagery have been implicated in a range of mental health conditions. Imagery has a particularly powerful effect on emotion and as such plays a particularly important role in emotional disorders. In depression, not only is the occurrence of intrusive negative imagery problematic, but also the lack of positive (in particular, future-directed) imagery is important. The authors suggest that, in depression, imagery can exacerbate the effects of interpretation bias. This article outlines an experimental psychopathology subcomponents model of depression that focuses specifically on the role of imagery and interpretation bias in the maintenance of the disorder. The authors propose that negative intrusive imagery, a lack of positive imagery, and negative interpretation bias serve both independently and interactively to maintain depressed mood. Finally, the authors consider the implications of this imagery-based approach for the development of new cognitive treatments in this area. Key words: mental imagery; interpretation bias; depression; intrusive memories; emotion.

Received 15 January, 2009; Accepted 14 April, 2009

Correspondence address: Emily A. Holmes, Department of Psvchiatrv, University of Oxford, Wctrneford Hospital, Oxford 0X3 7JX, UK. Tel: + 44 (0)1865 223912. Fax: + 44 (0)1865 793101. E-mail:


Why consider mental imagery and emotion in treatment across disorders?

Abnormalities in mental imagery are problematic across a range of different psychological disorders. Mental imagery has been described as the experience of “seeing with the mind’s eye,” “hearing with the mind’s ear,” and so on (Kosslyn, Ganis, & Thompson, 2001). Such mental imagery can be of the past or the future and can be either voluntary (deliberately generated) or involuntary (coming to mind spontaneously and “unbidden”).

In cognitive behaviour therapy (CBT), cognitions are assumed to take the form of either verbal thoughts or mental images. The dominant focus has traditionally been on verbal thoughts, although since its inception cognitive therapy has also emphasised the role of mental imagery (Beck, 1976). Figure 1 illustrates how CBT “does exactly what it says on the tin.” That is, it tackles the two types of cognitive “ingredients” that need to be modified: verbal thoughts and mental images. For the best treatment results, clearly we need to target the most toxic cognitions. We have previously argued that mental imagery has a particularly strong impact on emotion and, therefore, provides a particularly important treatment target (Holmes & Mathews, 2005)

In what psychological disorders does negative emotional imagery occur?

Intrusive, affect-laden images constitute a hallmark symptom of posttraumatic stress disorder (PTSD). For example, following an assault, a patient may “reexperience” the event through sensory and affective flashbacks such as “feeling like I am being stabbed in the chest” (Holmes, Grey, & Young, 2005, p. 8).


Figure 1. Cognitive behavioural therapy “does exactly what it says on the tin.” The two cognitive “ingredients” are verbal thoughts and mental imagery.

However, intrusive images cause distress across a range of psychological disorders (Hirsch & Holmes, 2007), from the well- known examples of social phobia and depression to agoraphobia, obsessive-compulsive disorder, spider phobia, bulimia, substance misuse, and suicidality. Interestingly, it is the absence (or “pushing away”) of imagery that characterises some other disorders such as generalised anxiety disorder (Hirsch & Holmes, 2007).

Why might imagery be important in making therapy work across psychological disorders?

Importantly for our distinction between verbal and imagery-based cognitions, evidence shows that imagery has a more powerful impact on emotion than its verbal counterpart (Holmes & Mathews, 2005; Holmes, Mathews, Dalgleish, & Mackintosh, 2006; Holmes, Mathews, Mackintosh, & Dalgleish, 2008). Holmes and Mathews (2005) suggest several reasons why imagery may have such a powerful impact on emotion. Imagery has perceptual correspondence to sensory experience, “as if’ it were really happening. That is, it is possible for imagery to directly provoke emotion in a similar manner as a real percept. For example, the neural representation of visual imagery is similar to that produced by actual visual performance (Sirigu & Duhamed, 2001). Rela- tedly, imagery can be used to access autobiographical memories and their associated emotions. Imagery can “hijack” attention through its highly absorbing nature and sense of “now-ness” and realness, for example, as experienced in flashbacks to a traumatic event in PTSD. Importantly, prospective imagery has been shown to be causal in determining future behaviour; imagining oneself completing a future event leads to significantly greater likelihood of this event being completed in real life (Libby, Shaeffer, Eibach, & Slemmer, 2007).

What is the evidence that imagery has a special relationship with emotion?

Although intrusive imagery has been reported anecdotally by clinicians as being particularly striking in psychopathology, until recently, little experimental research had addressed basic assumptions about mental imagery. The special relationship between mental imagery and emotion that had been observed clinically has now been confirmed in the laboratory using a variety of paradigms (Holmes, Lang, & Shah, 2009; Holmes & Mathews, 2005; Holmes et al., 2006; Holmes, Mathews, et al., 2008). One such paradigm comes from the cognitive bias modification (CBM) literature, which involves training individuals to adopt a particular habit of thought. One version of CBM involving the training of interpretative bias uses repeated exposure to ambiguous scenarios, which are continually resolved either negatively or positively depending on the experimental condition. Holmes and Mathews (2005) compared imagery and verbal processing instructions during negative CBM for interpretation bias and found that imagery processing produced greater increases in negative emotions compared with verbal processing. Holmes et al. (2006, 2009) compared imagery versus verbal processing instructions for positive interpretation CBM and found a greater increase in positive mood associated with imagery processing. An evaluative learning style paradigm has provided convergent evidence for these findings (Holmes, Mathews, et al., 2008). Imagery has thus been shown to have a more powerful effect on increasing both negative and positive emotion and can be considered an “emotional amplifier” in psychopathology (Holmes, Geddes, Colom, & Goodwin, 2008).

If imagery acts an emotional amplifier, what are the overarching implications for “making therapy work”?

Key implications for CBT are that, in addition to considering patients’ verbal cognitions, therapists should be aware of the following:

  1. The importance of assessing negative imagery during assessment across disorders.
  2. The benefits of promoting more positive imagery in treatment.
  3. The potential to develop new, imagery- focused therapy innovations.

We hope that investigating mental imagery in the laboratory will inform developments in cognitive theory, which, in turn, will inform developments in treatment innovation, which should be related back to basic science, thus promoting a continued interweave between experimental psychopathology and clinical application.

What techniques are there across disorders for treating problematic imagery?

A range of successful cognitive therapy interventions address problematic imagery and its treatment at their core, notably in PTSD (Ehlers & Clark, 2000) and in social phobia (Clark et al., 2006). Imagery offers new treatment possibilities to the traditional approaches of working with verbal negative thoughts, one example being “imagery rescripting,” which may be applied across disorders (Holmes, Arntz, & Smucker, 2007). The presence of problematic imagery, however, is not a prerequisite for using imagery techniques; it can also be important to build up more positive and adaptive imagery. The promotion of more positive/adaptive imagery may also be achieved by translating research on CBM paradigms discussed earlier, that is, via computerised programmes aimed at modifying biases (e.g. of interpretation) and promoting positive imagery (Holmes, Coughtrey, & Connor, 2008).

Further thoughts about an imagery approach using the example of depression

Depression is a mood disorder with a range of symptoms. Emotional effects include feelings of extreme sadness and hopelessness. Cognitive effects typically include low self-esteem, guilt, and concentration difficulties. Behavioural effects include agitation and changes in bodily functioning, including sleeping, eating, and sexual problems (American Psychiatric Association, 2000). Rather than consider the whole disorder, in this article we focus on a subset of specific cognitive psychopathological processes: negative interpretation bias, a preponderance of negative imagery, and a lack of positive imagery. These experimental psychopathology subcomponent processes are illustrated in Figure 2 and discussed in further detail next.

Interpretation bias and depression

Everyday we encounter information that is ambiguous in nature, in other words, information that can be interpreted in more than one way. It has long been held that individuals with depression tend to interpret ambiguous information negatively, and this idea is central to traditional cognitive behaviour theories of depression (Beck, 1976). As shown in Figure 2, when presented with an ambiguous event such as “a friend fails to return your phone call,” nondepressed individuals are more likely to display a positive bias and make a benign interpretation, for example, believing their friend was simply preoccupied. In contrast, individuals with depression are more likely to have a negative bias and make a negative interpretation(e.g. suspect their friend is deliberately ignoring them). Greater negative bias has been shown to be predictive of future depressive symptoms (Rude, Valdez, Odom, & Ebrahimi, 2003).



Figure 2. An experimental psychopathology subcomponents model of depression focusing on mental imagery and interpretation bias.


One way in which to resolve ambiguity inherent in all sorts of daily situations is to imagine the outcome, which allows us to mentally simulate the resolution to the situation. Given the powerful effect that imagery has on emotion, this strategy will be particularly toxic when accompanied by a negative interpretation (Holmes & Mathews, 2005). That is, we suggest that when imagining a negative outcome and subsequently mentally simulating it (e.g. seeing oneself as abandoned, lonely and rejected after a friend does not return a phone call), this is likely to exacerbate depressed mood to a greater extent than verbally thinking about the same event.

Can even positive information seem negative?

The confrontation with ambiguous information can reveal a negative interpretation bias. However, even in the face of overtly positive information, a negative bias can emerge. Holmes et al. (2006) found that when patients were given overtly positive material, verbal compared with imagery-based processing produced not only less positive mood but an increase in negative mood and bias. Holmes et al. (2009) suggest that the presentation of overtly positive material may allow participants to make verbal comparisons between their current situation with their own “unachieved standards.” Although it is also possible to make comparisons while imagery processing, this is less likely given the cognitive effort required to mentally switch between images. Experiment 2 in Holmes et al. (2009) concluded that when comparative processing highlights discrepancies among the ideal, ought, and actual selves of participants, it may be partially responsible for the negative effects associated with verbally processing positive information. Current work is exploring this further.

Mental imagery in depression

Depression has traditionally been associated with verbal rather than imagery-based cognitions. A key focus has been on rumination, a predominantly verbal process (Fresco, Frankel, Mennin, Turk, & Heimberg, 2002). However, another clinical feature of depression is the experience of involuntary negative image-based memories. Some studies indicate that up to 90% of depressed patients report experiencing distressing intrusive memories (Birrer, Michael, & Munsch, 2007). It has been proposed that overgeneral autobiographical memory in depression may develop as a protective mechanism adopted by individuals in an attempt to prevent such distressing intrusive memories from coming to mind (Williams et al, 2007).

Negative, maladaptive appraisals of intrusive memories (e.g. “Having this memory means that I am weak”) have been proposed to maintain the occurrence of intrusive memories and, in turn, depressive symptoms (Starr & Moulds, 2006). Maladaptive appraisals of intrusive memories have been shown to be significantly associated with depressive symptoms. Over and above the severity of the memory content and the frequency of the intrusion, maladaptive appraisals were found to be the strongest predictor of depression in these studies.

Negative imagery of the past is indeed a problem in depression; however, a highly neglected area of research is negative imagery of the future. Holmes, Crane, Fennel, and Williams (2007) proposed applying a PTSD perspective to suicidality, asking whether there is a prospective suicidal equivalent to “flashbacks.” Patients with suicidal depression reported highly vivid negative, future-directed imagery of suicide, which the authors termed “flash-forwards” to suicide. These suicidal images may be particularly toxic given the powerful effect of imagery, with its ability to hijack attention and promote behavioural action.

Lack of positive imagery

Positive mental imagery in the context of depressed mood has hitherto also been relatively underexplored. Holmes, Lang, Moulds, and Steele (2008) have shown that people high in dysphoria have a poorer ability to imagine positive future events compared with people low in dysphoria. As is shown in Figure 2, we suggest that a lack of positive imagery will also promote depressed mood.

An experimental psychopathology formulation of mental imagery in depression

We suggest an experimental psychopathology subcomponents model of the processes in depression, focusing on mental imagery and interpretation bias, as presented in Figure 2.

The term “subcomponents” is emphasised because clearly the model does not address all clinical features of depression. However, we believe a subprocesses approach is important in identifying and testing specific hypotheses about psychopathology in the laboratory.

Figure 2 illustrates the key processes associated with the maintenance and exacerbation of depressed mood: interpretation bias, negative intrusive imagery, and lack of positive imagery. First, with interpretation bias (see Figure 2), when faced with an event, one can adopt either a positive or a negative bias. For instance, in the case of the metaphorical “half-filled glass,” if adopting a positive bias, a benign interpretation would follow, such as seeing the glass as “half-full.” This is contrast to a negative bias, which would lead to a negative mental interpretation such as perceiving the glass as “half-empty” and thus promoting depressed mood. Importantly, if the outcome of the negative interpretation takes the form of a mental image (rather than a verbal thought), the powerful effect of imagery on emotion means that depressed mood is likely to be further exacerbated. In contrast, if the event is verbally processed, even in the face of positive information, comparative processing (which would create negative comparisons of the self compared with the positive information) may provoke depressed mood (Holmes et al., 2009). The second process key to this model (see Figure 2) is the preponderance of negative intrusive imagery of the past and future. Again, because of the powerful effect of imagery on emotion, this further lowers depressed mood. For example, in suicidal depression, times of despair can be associated with detailed mental images, for example, of a future suicide attempt (Holmes, Crane, et al., 2007). As illustrated in the model, the interpretation of negative intrusive imagery (e.g. “This means that I am crazy”) also further serves to maintain depressed mood (Starr & Moulds, 2006). Finally, a lack of positive imagery in depression (see Figure 2) contributes to the continuation of depressed mood and absence of healthy optimism that things can improve in the future.

The model, therefore, demonstrates how the key processes of negative intrusive imagery, lack of positive imagery, and negative interpretation bias can function both independently and interactively to maintain depressed mood. This is in line with Hirsch, Clark, and Mathews (2006), who propose a combined cognitive biases hypothesis that highlights the importance of examining cognitive biases in combination as opposed to in isolation.



Figure 3. An experimental psychopathology subcomponents model of depression focusing on mental imagery and interpretation bias: therapeutic directions. (Narrow arrows indicate potential treatment targets.)


Clinical treatments: future directions

There are several implications of this imagery- based approach for the development of new cognitive treatments for depression. Figure 3 is an adaptation of Figure 2 illustrating the suggested potential targets for cognitive therapy. We have highlighted the importance of promoting positive future-oriented imagery in the treatment of depression (Holmes, Lang, et al., 2008; see Figure 3). To do this, computerised CBM techniques hold promise for promoting the habit of creating more positive mental imagery and interpretation biases as a routine part of everyday life (Holmes et al., 2009). In terms of depressive intrusive memories, a computerised CBM task has been developed to specifically modify (or retrain) maladaptive appraisals (Lang, Moulds, & Holmes, 2009). In a nonclinical sample, this technique has been shown to produce increases in a positive appraisal bias and decrease the number of intrusions reported of an analogue negative event (a depressing film).

Imagery rescripting (see Figure 3) offers a cognitive therapy technique to address negative imagery. For example, suicidal imagery could be directly targeted using imagery rescripting to produce an alternative future outcome (e.g. an image of overdosing on pills could be rescripted to an image of disposing of the tablets; Holmes, Crane, et al., 2007).



Our proposal of an experimental psychopathology subcomponents model of processes in depression focuses on the role of imagery and interpretation bias in the maintenance of the disorder. Specifically, this model proposes that negative intrusive imagery, a lack of positive imagery, and negative interpretation bias serve both independently and interactively to maintain depressed mood. Providing a theoretical model by which to test core hypotheses using rigorous experimental techniques opens up new avenues of investigation in the drive for much-needed potential therapeutic targets in this area. Further research is also required to investigate the role of imagery in a range of other psycho- pathological conditions such as schizophrenia and bipolar disorder.


Emily Holmes is supported by a Royal Society Dorothy Hodgkin Fellowship and in part by Economic and Social Research Council Grant RES-061-23-0030 and John Fell OUP Grant PRAC/JF. Tamara J. Lang is supported by the University of Oxford Department of Psychiatry Bursary for Overseas Students. We thank Alan Slater for Figure 1.


Distinct and Overlapping Features of Rumination and Worry: The Relationship of Cognitive Production to Negative Affective States

David M. Fresco,* [6] [7] [8] Ann N. Frankel,1 Douglas S. Mennin,1 Cynthia L. Turk,1 and Richard G. Heimberg12

Worry and rumination are cognitive processes, often represented as verbal or linguistic activities. Despite similarities in definition and description, worry has been most closely examined in relation to anxiety whereas rumination has traditionally been related to depression. This distinction remains in spite of high rates of comorbidity between anxiety and depression. This study sought to belter understand the distinct and overlapping features of worry and rumination as well as their relationship to anxiety and depression. Seven hundred eighty-four unselected college students completed self-report measures of worry, rumination, anxiety, and depression. Items from the respective worry and rumination scales were submitted to factor analysis, which revealed a four-factor solution comprised of 2 worry factors and 2 rumination factors. A Worry Engagement factor as well as a Dwelling on the Negative factor emerged as distilled measures of worry and rumination, respectively. Scores on these factors were highly correlated with each other and demonstrated equally strong relationships to both anxiety and depression. Findings from this study suggest that worry and rumination represent related but distinct cognitive processes that are similarly related to anxiety and depression.


Borkovec and colleagues have been instrumental in furthering our understanding of worry. Borkovec et al. (1983, p. 10) define worry as “a chain of thoughts and images, negatively affect-laden and relatively uncontrollable; it represents an attempt to engage in mental problem-solving on an issue whose outcome is uncertain but contains the possibility of one or more negative outcomes; consequently, worry relates closely to the fear process.” Worry is regarded as the hallmark feature of generalized anxiety disorder (GAD; American Psychiatric Association, 1994), but is common to all anxiety disorders (Barlow, 1988) and to depression (Molina, Borkovec, Peasley, & Person, 1998; Starcevic, 1995). Borkovec’s avoidance theory of worry holds that worry is a predominantly verbal activity that may allow individuals to look away from more emotionally arousing material, which in turn allows them to disengage from emotional pain and regain a sense of emotional and physiological control (Borkovec, 1994). By worrying, individuals successfully avoid aversive images, somatic anxiety, and other negative emotions. Thus, worry serves as an avoidance response that interferes with emotional processing and thus prevents the extinction of fear. Furthermore, by avoiding anxiety-producing stimuli or feelings in the short run, worried individuals prevent themselves from adequately processing situation- ally relevant information, which in turn, may prevent them from deploying their most adaptive coping resources. In the long run, individuals who engage in chronic worry experience all of these negative consequences and may fail to adequately resolve stressors that arise (Borkovec, 1994).

Nolen-Hoeksema (1998, p. 239) defines rumination, or more specifically ruminative responses to depression, as “behaviors and thoughts that passively focus one’s attention on one’s depressive symptoms and on the implications of these symptoms.” Like worry, rumination is described as a largely cognitive or verbal activity. Just as worry is thought to contribute to the maintenance of anxiety by interfering with emotional processing, Nolen-Hoeksema (1998) posits that rumination prolongs and worsens depression. In the absence of an active coping response, Nolen-Hoeksema (1998) suggests that distracting one’s self from thinking about one’s stressors and the symptoms that arise from them is superior to ruminating about them, and a number of empirical studies with clinically depressed and dysphoric populations support this assertion (see Nolen-Hoeksema, 1998, for a review). In contrast, Borkovec, Alcaine, and Behar (in press) suggest that when individuals worry, they tend to think about matters in a superficial fashion, which in turn, serves as a distraction from more distressing thoughts and emotions. However, they regard the accessing of emotional content without avoiding it as more adaptive than engaging in worry or other forms of distraction. Thus, despite substantial similarities between worry and rumination, theorists view the functional relationship of worry to anxiety and of rumination to depression in considerably different ways.

Few studies have examined the relationships among rumination, worry, depression, and anxiety. Segerstrom, Tsao, Alden, and Craske (2000) reported strong zero-order correlations between worry and rumination in both undergraduate and clinical samples. Furthermore, using structural equation modeling, Segerstrom et al. found that a latent variable (repetitive thought) comprised of manifest variables of rumination and worry was significantly associated with both depression and anxiety. In another study (Blagden & Craske, 1996), rumination was associated with greater anxiety while listening to anxiety-evoking music than was distraction. These studies suggest that worry and rumination may be less distinct from one another and less distinctly related to anxiety and depression than often suggested. This study was conducted to further elucidate the similarities and differences between worry and rumination and their relationships to anxiety and depression. Specifically, the goals of this study were to (1) to examine the overlap between and distinctiveness of worry and rumination via factor analysis, and (2) to examine the relationship of worry and rumination factors to anxious and depression symptoms.



Seven hundred eighty-four unselected undergraduate students participated in return for partial course credit in an introductory psychology class. Women comprised the majority of the sample (67.2%). The racial composition of the sample was 39.0% Caucasian, 37.7% African American, 11.7% Asian, 3.5% Latino/Hispanic, 0.4% Middle Eastern, 0.2% Native American, and 7.4% mixed racial heritage. The average age of participants was 20.2 years (SD = 4.2 years).


The Penn State Worry Questionnaire (PSWQ; Meyer, Miller, Metzger, & Borkovec, 1990) is a 16-item inventory designed to assess trait worry and to capture the generality, excessiveness, and uncontrollability characteristics of pathological worry (e.g., “My worries overwhelm me”; “I worry all the time”). Each item is rated on a 1 (not at all typical of me) to 5 (very typical of me) Likert-type scale. Among samples of college undergraduates, the PSWQ has repeatedly demonstrated good internal consistency and good test-retest reliability over intervals as long as 8- 10 weeks (Meyer et al., 1990). In this sample, the PSWQ achieved a Cronbach’s alpha of .90.

The Response Styles Questionnaire (RSQ; Nolen-Hoeksema & Morrow, 1991) is a 71-item self-report measure used to identify four coping strategies in response to depressed mood: problem solving, distraction, engaging in dangerous activities, and rumination. In this study, the entire 71-item measure was administered and, the Ruminative Responses Scale (RRS) was extracted from that measure and utilized in the analyses described below. The RRS consists of 22 items on a Likert-type scale, with values ranging from 1 (almost never) to 4 (almost always). Items such as “go away by yourself and think about why you feel this way” or “try to understand yourself by focusing on your depressed feelings” are used to assess ruminative coping responses to depressed mood. The internal consistency of the RRS is good (a = .89; Nolen-Hoeksema & Morrow, 1991). In this sample, the RRS achieved a Cronbach’s alpha of .90.

The Mood and Anxiety Symptom Questionnaire-Short Form (M ASQ; Watson & Clark, 1991) is a 62-item measure assessing symptoms that commonly occur in the mood and anxiety disorders. The short form is comprised of four subscales. The General Distress Anxious Symptoms subscale includes 11 items that are indicators of anxious mood but do not provide strong differentiation from depressed mood (e.g., “Felt nervous”; “Had an upset stomach”). The General Distress Depressive Symptoms subscale contains 12 items that are indicators of depressed mood but do not provide strong differentiation from anxious mood (e.g., “Felt sad”; “Felt like crying”). The Anxious Arousal subscale is comprised of 17 items assessing symptoms of somatic tension and hyperarousal (e.g., “Startled easily”; “Was trembling or shaking”). The Anhedonic Depression subscale consists of 22 items assessing symptoms relatively specific to depression such as loss of pleasure in usual activities, disinterest, low energy (e.g., “Felt like nothing was very enjoyable”) and reverse-keyed items assessing positive emotional experiences (e.g., “Felt cheerful”). Each item is rated on a 1 (not at all) to 5 (extremely) Likert-type scale.


The PSWQ, RSQ, and MASQ were administered as part of a larger questionnaire battery given to introductory psychology students during the first week of the semester. Participants took the questionnaire packets home to complete, and approximately 76% returned them the following week. Participants who completed the packet received credit toward fulfillment of course requirements. This procedure was followed for two successive semesters (Fall 1997, n = 527; Spring 1998, n = 257). The subsamples for each semester did not differ on sex, race, or age. Thus, all analyses are based on the entire sample.


Exploratory Factor Analysis

The 22 items from the Ruminative Response Scale and the 16 items from the Penn State Worry Questionnaire were submitted to common factor analysis[9] with varimax rotation[10] to investigate whether items from these scales formed factors comprised of elements from each scale. Six factors emerged with eigenvalues greater than one. Using the scree plot method, a four-factor solution was retained. Table I presents the initial communality estimates as well as the rotated factor loadings for all items. An item was considered to load onto a factor if its factor loading exceeded .40. An item was considered to load on more than one factor if the difference between the factor loadings was less than .10.


Table I. Rotated Factor Loadings and Initial Communality Estimates for Common Factor Analysis of the Penn State Worry Questionnaire (PSWQ) and the Ruminative Response Scale of the Response Style Questionnaire (RSQ)

Item and item wording Factor 1 Factor 2 Factor 3 Factor 4 Initial comm, estimate
Factor 1: Worry Engagement (a = .94, M = 28.66, SD = 10.00)          
PSWQ15.1 worry all the time. .816 .247 .024 .110 .748
PSWQ7.1 am always worrying about something. .800 .262 .121 .198 .758
PSWQ14. Once I start worrying, I can’t stop. .788 .282 .059 .076 .699
PSWQ13.1 notice that I have been worrying about things .769 .198 .108 .253 .701
PSWQ12. I’ve been a worrier all my life. .734 .122 .114 .112 .580
PSWQ4. Many situations make me worry. .728 .284 .038 .287 .696
PSWQ5.1 know I shouldn’t worry about things, but I just can’t help it. .690 .180 .079 .350 .661
PSWQ2. My worries overwhelm me. .645 .320 .126 .259 .607
PSWQ9. As soon as I finish one task I start to worry about everything else I have to do. .630 .244 .052 .137 .531
PSWQ6. When I am under pressure, I worry a lot. .621 .201 .075 .335 .562
PSWQ16.1 worry about projects until they are all done. .574 .065 .112 .159 .437
Factor 2: Dwelling on the negative (a = .87, M = 22.26, SD = 6.37)          
RSQ15. Think about how passive and unmotivated you feel. .152 .640 .189 .051 .440
RSQ5. Think about how alone you feel. .105 .599 .225 .191 .447
RSQ43. Think about how sad you feel. .151 .588 .334 .246 .558
RSQ6. Think I won’t be able to do my job/work because I feel so badly. .121 .582 .064 -.026 .369
RSQ7. Think about your feelings of fatigue and achiness. .190 .549 .091 -.077 .353
RSQ44. Think about all your shortcomings, failings, faults, mistakes. .238 .538 .288 .254 .538
RSQ21. Think Why can’t I get going? .169 .532 .214 .016 .413
RSQ8. Think about how hard it is to concentrate. .189 .519 .050 .022 .345
RSQ61. Think about how angry you are with yourself. .176 .505 .246 .174 .421
RSQ40. Think Why do I have problems other people don’t have? .295 .480 .272 .212 .449
RSQ46. Think about how you don’t feel up to doing anything. .194 .480 .237 .060 .368
RSQ19. Think about how you don’t seem to feel anything any more. .136 .421 .218 -.016 .256
Factor 3: Active Cognitive Appraisal (oc = .81, M = 14.32, SD = 4.66)          
RSQ68. Isolate yourself and think about the reasons why you feel sad. .120 .295 .713 .115 .577
RSQ56. Go someplace alone to think about your feelings. .003 .126 .636 .021 .461
RSQ25. Go away by yourself and think about why you feel this way. .020 .134 .597 -.030 .429
RSQ70. Try to understand yourself by focusing on your depressed feelings. .071 .379 .565 .084 .509
RSQ53. Analyze your personality to try to understand why you are depressed. .043 .264 .554 .003 .450
RSQ18. Analyze recent events to try to understand why you are depressed. .001 .337 .464 -.012 .364
RSQ67. Listen to sad music. .137 .224 .439 .120 .326
Factor 4: Absence of Worry (a = .69, M = 9.36, SD = 3.22)          
PSWQ8.1 find it easy to dismiss worrisome thoughts. -.377 -.114 -.005 -.547 .421
PSWQ10.1 never worry about anything. -.377 -.082 -.070 -.514 .397
PSWQ3.1 don’t tend to worry about things. -.325 -.030 .031 -.511 .327
PSWQ1. If I don’t have enough time to do everything, I don’t worry about it -.185 -.030 .007 -.473 .253
Items loading on multiple factors or no factors          
PSWQ11. When there is nothing more I can do about a concern, I don’t worry about it any more. -.253 -.055 .016 -.340 .203
RSQ22. Think Why do I always react this way? .191 .419 .347 .079 .420
RSQ28. Write down what you are thinking about and analyze it .034 .021 .350 -.123 .162
RSQ30. Think about a recent situation, wishing it had gone better. .151 .378 .323 .108 .299

Note. Items were considered to load on a factor if its loading score exceeded 0.40 and the difference between the loadings on other factors was greater than 0.10. Because the Ruminative Response Scale consists of items distributed throughout the 71-item Response Style Questionnaire (RSQ), item numbers reflect their placement in the larger RSQ.


Factor 1 consisted of the 11 positively coded PSWQ items and was labeled Worry Engagement. This factor had an eigenvalue of 11.68 and accounted for 30.73% of the variance. Factor 2 consisted of 12 items from the Ruminative Response Scale whose content reflected a general tendency to dwell on the negative consequences of feeling depressed. This factor was labeled Dwelling on the Negative. It had an eigenvalue of 4.24 and accounted for 11.16% of the variance. Factor 3 consisted of seven RRS items whose general content reflected an active appraisal of one’s situation while depressed and was labeled Active Cognitive Appraisal. Factor 3 had an eigenvalue of 1.72 and accounted for 4.54% of the variance. Finally, Factor 4 consisted of four negatively coded PSWQ items (e.g., “I find it easy to dismiss worrisome thoughts,” and “I never worry about anything.”) and was labeled Absence of Worry. This factor had an eigenvalue of 1.42 and accounted for 3.73% of the variance. Three RRS items and one PSWQ item were dropped for low factor loadings or for loading on more than one factor.[11] Factor scores were computed from this four-factor solution. Sample means and standard deviations as well as Cronbach’s alpha for factor scores were also calculated. These scores are presented in Table I.

Zero-order correlations were computed to examine the relationship among the factor scores. Worry Engagement and Dwelling on the Negative demonstrated a strong positive correlation with each other (r = .46) and negative correlations with Absence of Worry (rs = -.54 and -.22 respectively). Active Cognitive Appraisal was also positively correlated with Worry Engagement (r = .29) and Dwelling on the Negative (r = .57) and modestly negatively correlated with Absence of Worry (r = -.14). All zero-order correlation coefficients were significant at p < .005.

Worry and rumination factor scores were also correlated with the four subscales of the MASQ. As seen in Table II, Dwelling on the Negative and Worry Engagement demonstrated similarly strong correlations with MASQ subscales with coefficients ranging from .39 to .59 and .30 to .51, respectively. Active Cognitive Appraisal demonstrated small to moderately sized correlations with the MASQ subscales–ranging from .14 to .33. Finally, Absence of Worry demonstrated small negative correlations with MASQ subscales–ranging from -.06 to -.26.

To further assess the relationship of the factor scores to measures of depression and anxiety, tests of dependent correlations (Bruning & Kintz, 1987) were conducted. For two of the subscales, MASQ-Anxious Arousal and MASQ-General Distress Depressive Symptoms, the zero-order correlation with Dwelling on the Negative was statistically stronger than that of Worry Engagement although the magnitude of the difference did not exceed Cohen’s criterion for a small effect (Cohen, 1977). Both Dwelling on the Negative and Worry Engagement were generally more highly correlated with MASQ subscales then was Active Cognitive Appraisal. Differences in dependent correlations comparing Dwelling on the Negative and Active Cognitive Appraisal approached or exceeded Cohen’s criterion for a medium effect (Cohen, 1977); differences in dependent correlations between Worry Engagement and Active Cognitive Appraisal approached Cohen’s criterion for a medium effect for three of four MASQ subscales (Cohen, 1977).

  Dwell (a) Worry (b) Appraise (c) Absence (d) Difference a vs. b ((781) d Difference a vs. c ((781) d Difference b vs. c ((781) d
MASQ-AA .39″ .30″ .24″ -.06 2.68** .19 4.91** .35 1.50 .11
MASQ-AD .42″ .38″ .14” -.21** 1.23 .09 9.38** .67 6.09″ .44
MASQ-GDA .46″ .46″ .28” -.18** 0.00 .00 6.12** .44 4.82″ .35
MASQ-GDD .59” .51″ .33” -.26** 2.82** .20 9.70** .69 5.03″ .36
Corrected-AA .26” .18″ .20” .01 2.23** .16 1.88 .13 0.48 .03
Corrected-AD .31″ .30″ .09* -.21** 0.29 .02 7.68** .55 5.66″ .40

Note. Dwell, Dwelling on the Negative; Worry, Worry Engagement; Appraise, Active Cognitive Appraisal; Absence, Absence of Worry; MASQ-AA, Mood and Anxiety Symptom Questionnaire (MASQ)-Anxious Arousal Scale; MASQ-AD, MASQ-Anhedonic Depression Scale; MASQ-GDA, MASQ- General Distress Anxious Symptoms Scale; MASQ-GDD, MASQ-General Distress Depressive Symptoms Scale; Corrected-AA, MASQ-AA controlling for MASQ-AD; Corrected-AD, MASQ-AD controlling for MASQ-AA; d\ Cohen’s d effect size conventions (Cohen, 1977), small = .20, medium = .50, large = .80.

«Tests of dependent correlations between Absence of Worry and the other three factor scores were also computed. Differences between Absence of Worry and the other factor scores were generally in the medium to large range using Cohen’s effect size conventions for d (Cohen, 1977). Differences with Dwelling on the Negative ranged from .66 to 1.41; differences with Worry Engagement ranged from .59 to 1.06; differences with Active Cognitive Appraisal ranged from .46 to .90.

* p < .05 (two-tailed). ** p < .01 (two-tailed).


Although the MASQ was selected as the measure of depressive and anxious symptoms because of its relatively distinct scales of Anxious Arousal and Anhedonic Depression, these two measures were relatively highly correlated (r = .33) in this sample. Consequently, “corrected” measures of Anxious Arousal and Anhedonic Depression were computed by conducting linear regression analyses and saving the unstandardized residuals. In general, compared to correlations with uncorrected measures of anhedonic depression and anxious arousal, correlations between factor scores and corrected measures of anhedonic depression and anxious arousal produced a similar, but attenuated, pattern of findings. The one notable exception was that Dwelling on the Negative differed from Worry Engagement but not Active Cognitive Appraisal in the magnitude of correlation with Anxious Arousal (see Table II).

In summary, Dwelling on the Negative and Worry Engagement demonstrated similarly strong positive correlations with measures of depression and anxiety–with Dwelling on the Negative demonstrating a small, but significantly stronger, relationship than Worry Engagement to two raw measures and one corrected measure. Further, both Dwelling on the Negative and Worry Engagement demonstrated significantly stronger relationships than did Active Cognitive Appraisal to the measures of depression and anxiety. However, the magnitude of this difference was somewhat larger for Dwelling on the Negative (see footnote a in Table II).


The main purpose of this study was to examine the distinctive and overlapping features of worry and rumination as well as the nature of their relationships to depression and anxiety. Indeed, our findings shed some light on these issues. First, although correlated with one another, worry and rumination also maintained a degree of distinctiveness. The four-factor solution that emerged from a factor analysis of PSWQ and RRS items produced factors consisting of items from just one of the scales. No factors contained items from both scales.

In the case of worry, an 11-item Worry Engagement factor demonstrated a slightly higher Cronbach’s alpha than the 16-item PSWQ. Furthermore, Absence of Worry did not relate as strongly as Worry Engagement to the MASQ measures of depression and anxiety. These findings suggest that the positively worded worry items may form the component of trait worry that is more highly associated with negative affective states. In a separate study of the PSWQ from the same database, Fresco, Heimberg, Mennin, and Turk (in press) used confirmatory factor analysis to demonstrate that Worry Engagement and Absence of Worry were first-order factors that served as indicators for a second-order general worry factor.

The factor analysis revealed two distinct facets of rumination that related somewhat differently to depression and anxiety. Although Dwelling on the Negative and Active Cognitive Appraisal were positively correlated with depression and anxiety, Dwelling on the Negative was more highly correlated with both. Thus, Dwelling on the Negative appears to tap the more maladaptive component of rumination whereas Active Cognitive Appraisal may represent a cognitive process associated with depression and anxiety but that is less destructive than Dwelling on the Negative.

Worry and rumination originated in separate theoretical and research traditions, and their hypothesized role in the onset and maintenance of negative mood states is also relatively distinct. However, given the lack of empirical specificity of worry to anxiety and rumination to depression and the relatively high correlation between them (Segerstrom et al., 2000) and among their components (this study), new conceptualizations are called for. One possible way to understand the similarities between worry and rumination is to consider that they may both serve an avoidance function. For example, Dwelling on the Negative, which seems to be a passive focus on one’s current mood state, was more strongly related to anxiety and depression than Active Cognitive Appraisal, which seems to involve integrating one’s emotional experience with an understanding of one’s self and one’s circumstances. To the extent that one is able to access emotions adaptively to understand what one’s needs are and what actions should be taken to get those needs met, depression and anxiety may be more short-lived. Although future research is needed to bear this out, it may be that Dwelling on the Negative in the short term may blunt some aspects of the affective experience but over the long term inhibit emotional processing and ability to deploy adaptive coping resources. In this way, Dwelling on the Negative would function in a manner very similar to what has been hypothesized for worry.

Dwelling on the Negative and Worry Engagement were highly correlated and related similarly to measures of anxiety and depression. Nevertheless, they loaded on different factors, suggesting that individuals are able to differentiate these two processes. It may be that while anxious and depressive symptomatology do not differentiate these constructs, other indices would. For example, the cognitive process of Worry Engagement may involve more questioning and uncertainty than Dwelling on the Negative.

Alternatively, Dwelling on the Negative and Worry Engagement may occur at different points in time in the same person. For example, a person may dwell on the negative after stress, bringing on depressive or anxious feelings. In turn, she or he may worry about other matters as a means to avoid the painful self-focus of dwelling on the negative. The short-run benefit is a reduction of self-focused negative emotions; however, the long-run cost is inadequate emotional processing of the stressful situation resulting in prolongation of the distress associated with some event rather than tidy, but emotionally evocative, resolution.

Findings from this study clearly indicate that worry and rumination are multidimensional and related but distinct cognitive processes that each have a relationship to both depression and anxiety. Further study of the components of each appears warranted. However, two methodological limitations (use of a cross-sectional design and the use of a nonclinical population) prevent greater generalizability of our findings until additional research has been conducted. Future research utilizing a longitudinal design and including measures of social desirability, coping, and life stress would definitely contribute to our understanding of the components of worry and rumination. An extension of these findings to a clinical population would also be beneficial. In addition, we chose to define worry solely in the way discussed by Borkovec and colleagues and rumination in the way described by Nolen-Hoeksema and colleagues. Replication with other measures (and operationalizations) of worry and rumination would support the validity of the findings from this study.


Seeing Failure in Your Life: Imagery Perspective Determines Whether Self-Esteem Shapes Reactions to Recalled and Imagined Failure

Lisa K. Libby, Greta Valenti, and Alison Pfent Richard P. Eibach

The Ohio State University University of Waterloo

The present research reveals that when it comes to recalling and imagining failure in one’s life, changing how one looks at the event can change its impact on well-being; however, the nature of the effect depends on an aspect of one’s self-concept, namely, self-esteem. Five studies measured or manipulated the visual perspective (internal first-person vs. external third-person) individuals used to mentally image recalled or imagined personal failures. It has been proposed that imagery perspective determines whether people’s reactions to an event are shaped bottom-up by concrete features of the event (first-person) or top-down by their self-concept (third-person; L. K. Libby & R. P. Eibach, 2011b). Evidence suggests that differences in the self-concepts of individuals with low and high self-esteem (LSEs and HSEs) are responsible for self-esteem differences in reaction to failure, leading LSEs to have more negative thoughts and feelings about themselves (e.g., M. H. Kemis, J. Brockner, & B. S. Frankel, 1989). Thus, the authors predicted, and found, that low self-esteem was associated with greater overgeneralization– operationalized as negativity in accessible self-knowledge and feelings of shame–only when participants had pictured failure from the third-person perspective and not from the first-person. Further, picturing failure from the third-person, rather than first-person, perspective, increased shame and the negativity of accessible knowledge among LSEs, whereas it decreased shame among HSEs. Results help to distinguish between different theoretical accounts of how imagery perspective functions and have implications for the study of top-down and bottom-up influences on self-judgment and emotion, as well as for the role of perspective and abstraction in coping.

Keywords: imagery perspective, subjective construal, shame, self-esteem, coping


Failures are like skinned knees, painful but superficial.

–Ross Perot

I failed first grade, which is my biggest problem. You always feel like a failure, like you’re stupid.

–Amy Sedans

Perot and Sedaris would likely agree that the experience of failure is unpleasant; yet the meaning they see failure to have in the broader context of their lives appears to differ. Perot believes that any given failure is insignificant relative to what truly defines him as a person, whereas Sedaris considers the failure she experienced to be self-defining. The subjective nature of these impressions becomes apparent upon considering that one of Perot’s life failures–-an unsuccessful attempt at the U.S. Presidency–should have consequences at least as significant as those of failing first grade. And, despite failing first grade, Sedaris has achieved quite a bit of success as an actor, author, and comedienne, just as Perot has in the world of business. Nevertheless, it seems likely that to the extent that these individuals rely on their impressions of the meaning of failure in their lives, their reactions to thinking about those failures would diverge, and Sedaris would end up worse off.

Given that it is impossible to get through life without experiencing failure, an important determinant of well-being is how one subjectively interprets and copes with such incidents. This is true not only as events unfold in real time (Niiya, Crocker, & Bartmess, 2004) but also as people mentally simulate them in memory and imagination (Taylor, Pham, Rivkin, & Armor, 1998). In the research reported here, participants recalled or imagined failures in their own lives. We were interested in how an aspect of these mental simulations–visual perspective in event imagery–would influence their reactions to thinking about their failures.

Existing accounts make conflicting claims about the effect of imagery perspective on coping with negative life events, either proposing that third-person imagery–in which the self is pictured from an external visual perspective–facilitates adaptive coping (e.g., Kross, Ayduk, & Mischel, 2005) or interferes with it (Kenny, Bryant, Silove, Creamer, O’Donnell, & McFarlane, 2009). We offer an alternative account that predicts that third- person imagery should have the potential to produce either effect, and we test a crucial moderator when it comes to picturing personal failure: self-esteem. Results help to distinguish between different theoretical models of how imagery perspective functions and suggest a new way to conceptualize the role of imagery perspective and abstraction in coping with negative experiences.

Imagery Perspective

As people recall and imagine life events, they often form mental images of those events (Moulton & Kosslyn, 2009) and may do so from different visual perspectives (Nigro & Neisser, 1983). With the first-person perspective, one sees the event from their own vantage point, as an actor in the scene; with the third-person perspective, one sees the event from an external vantage point, watching the self. This subtle phenomenological variable can have powerful effects on responses to pictured events, influencing judgment, emotion, and behavior (e.g., Libby, Shaeffer, Eibach, & Slemmer, 2007; Mclsaac & Eich, 2002; Valenti, Libby, & Eibach, 2011; Vasquez & Buehler, 2007). However, the function of imagery perspective responsible for such effects is still under investigation.

The present experiments contribute to this endeavor by testing predictions from a model (Libby & Eibach, 2011b) that proposes imagery perspective to function in representing the level of meaning in event representations. According to this model, picturing an event from the first-person perspective causes people to understand it bottom-up, in terms of the phenomenology evoked by concrete features of the pictured situation, apart from the broader context. Picturing an event from the third-person perspective causes people to understand it top-down, in terms of abstractions that integrate it with its broader context. Evidence supporting this model demonstrates that actions (e.g., “wiping up a spill”) are understood in terms of constituent aspects (e.g., “using a sponge”) from the first-person perspective but in terms of connections to causes, consequences, traits, goals, and identities from the third- person perspective (e.g., “cleaning up after the kids,” “being a responsible parent”; Libby, Eibach, & Gilovich, 2005; Libby, Shaeffer, & Eibach, 2009; Vasquez & Buehler, 2007). Converging with the idea that visual perspective is the causal element in producing these effects, they emerge not only when visual perspective is manipulated in mental imagery but also when it is manipulated by means of photographs–even when the photographs control the objects in the image and the distance to the action (Libby et al„ 2009).

In considering the implications of these findings for responses to recalled and imagined life events, we invoke the notion of the self as a dual-faceted structure that involves an experiential awareness of the present moment and a conceptual representation of the self as an entity that persists across time. James’s (1890/1950) distinction between the I-self and the me-self exemplifies this notion, and similar models shape the study of the self across diverse traditions of scholarship, from philosophy to neuroscience (e.g., Gallagher, 2000; LeDoux, 1996). The experiential “I” is defined by the phenomenology evoked bottom-up by concrete features of the environment and one’s actions on it. The conceptual “me,” or self-concept, is defined by a rich assemblage of diverse meaningmaking structures including self-schemas (Markus, 1977), selftheories (Hong, Chiu, Dweck, Lin, & Wan, 1999), and selfnarratives (McAdams, 2001; Neisser, 1994; Singer & Salovey, 1993). This diverse array of meaning-making structures organizes and sometimes biases the processing of self-relevant information so as to maintain the conceptual self as a coherent, and often positive, entity (e.g., Ross, 1989; Sanitioso, Kunda, & Fong, 1990; Sedikides & Strube, 1997; Swann, 1983).

The experiential and conceptual facets of the self tend to be interrelated in real time (e.g., Farb et al., 2007); however, the effect of imagery perspective on processing style suggests that imagery perspective may function to differentially highlight these two facets of the self when people picture events in memory or imagination. Specifically, we propose that first-person imagery represents an event in terms of the experiential “I,” and third-person imagery represents the event in relation to the conceptual “me» (Libby & Eibach, 201 lb). If so, the knowledge structures and motivations that define the conceptual self should shape reactions to life events when they are pictured from the third-person perspective but not from the first-person. When it comes to shaping reactions to failure, evidence suggests that the aspect of the self- concept most relevant to consider is self-esteem.


Self-esteem is an aspect of the self-concept that captures people’s global self-evaluations (Conner Christensen, Wood, & Barrett, 2003; Epstein, 1973). One of the most well documented differences between low- and high-self-esteem individuals (LSEs and HSEs) is in how they react to failure: LSEs have more extreme negative reactions (Blaine & Crocker, 1993; Taylor, 1991). In particular, LSEs are prone to overgeneralize, a response style that is characterized by “a tendency to bring thoughts of personal inadequacy to mind and/or experience a reduction in the sense of self-worth” (Carver, Ganellen, & Behar-Mitrani, 1985, p. 727).

For example, in response to a specific failure, LSEs’ thoughts about themselves in general shift in a negative direction: Their weaknesses and shortcomings become more cognitively accessible in memory (Brown & Smart, 1991; Dodgson & Wood, 1998), and they are prone to judge themselves to be a failure as a person in general (Kernis, Brockner, & Frankel, 1989). LSEs’ feelings about themselves in general also shift in a negative direction in response to a specific failure. Shame reflects global negative feelings about the self and can be distinguished from guilt, an emotion based in negative feelings about a particular action (Tangney & Dearing, 2002; Lewis, 1971; Tracy & Robins, 2004). LSEs are prone to experiencing shame in particular (Tangney & Dearing, 2002; Brown & Marshall, 2001).

In contrast to LSEs, HSEs do not show these negative shifts in thoughts and feelings about the self in response to failure. Instead, HSEs appear to respond in ways that counteract the negative impact, focusing on their strengths and positive feelings about themselves in the face of failure (e.g., Brown & Smart, 1991; Dodgson & Wood, 1998; Dutton & Brown, 1997). Thus, LSEs’ reactions to failure are self-defeating, whereas HSEs’ reactions are self-enhancing.

These different reactions appear to be a product of top-down influences of LSEs’ and HSEs’ global self-views, rather than objective differences in the attributes LSEs and HSEs possess or the failure events they experience. LSEs tend to be no worse off, objectively, than HSEs are on dimensions typically considered to define success and social worth (e.g., intelligence, attractiveness; Brown, 1998). And self-esteem differences in response to failure emerge even when the failure event is held constant by experimental control (e.g., Dodgson & Wood, 1998; Tangney & Dearing, 2002). Thus, self-esteem differences in reaction to failure reflect self-esteem differences in the subjective meaning of failure, and these different subjective meanings reflect the top-down influence of the different global self-views that distinguish LSEs from HSEs. The self-concepts of LSEs promote overgeneralization, whereas the self-concepts of HSEs protect against it (Conner Christensen et al„ 2003; Kemis et al., 1989).

Overview of the Present Research

Research documenting self-esteem differences in response to failure has tended to focus on online reactions to failure in the present (cf. Conner Christensen et al., 2003). However, to the extent that these differences result from the influence of the self-concept, our account of imagery perspective’s function predicts that the perspective people use to visually recall or imagine failure should determine whether low self-esteem puts individuals at risk for overgeneralization. If imagery perspective functions to represent an event in terms of the experiential “I” (first-person) or the conceptual “me” (third-person), and if self-esteem differences in overgeneralization reflect the influence of the conceptual “me,” then low self-esteem should predict greater overgeneralization only when individuals picture failures from the third-person perspective and not from the first-person. Further, this account suggests that to the extent perspective influences reactions at high and low levels of self-esteem, third-person imagery should promote overgeneralization among LSEs but protect against it among HSEs.

Testing these predictions has theoretical implications for understanding the function of imagery perspective. According to alternative accounts, picturing an event from the third-person perspective depersonalizes it, detaching it from the self (e.g., Mclsaac & Eich, 2004; Sanitioso, 2008; Williams & Moulds, 2008). Such accounts would not readily predict reactions to be more biased in the direction of the self-concept from the third-person than first- person perspective, as we do. Further, these alternative accounts propose that third-person imagery blunts the emotional response to recalled events, relative to first-person imagery. However, our account predicts that third-person imagery should have the potential to increase or decrease shame, depending on self-esteem. Given that shame interferes with adaptive coping, understanding the effect of perspective on feelings of shame has practical implications for well-being, in addition to theoretical implications for understanding the function of imagery perspective.

Study 1

Study 1 served as an initial exploration of our hypotheses to determine whether the predicted patterns emerged when individuals were allowed to picture a failure from the perspective they spontaneously adopted. Participants recalled and mentally pictured a personal failure. They reported on the visual perspective of their event imagery, and they completed a measure of the extent to which they were overgeneralizing from the failure as they pictured it. We predicted that low self-esteem would be associated with greater overgeneralization from the failure only to the extent that participants pictured it from the third-person perspective. Further, we expected that third-person imagery would be associated with greater overgeneralization among LSEs but less overgeneralization among HSEs. As participants pictured their failure, they also rated the extent to which they were feeling shame, allowing us to assess whether similar patterns would emerge for this emotional component of overgeneralization in particular.


Participants. Eighty-three undergraduates (57 women and 26 men) participated for course credit.

Materials and procedure. Participants sat in individual cubicles to complete a computerized questionnaire. The questionnaire began with two individual difference measures. The first was the Rosenberg (1965) Self-Esteem Scale (RSE). The second was the Attitudes Towards Self Scale (ATS, Carver & Ganellen, 1983), an 18-item inventory with factors indexing chronic tendencies to engage in overgeneralization (e.g., “When even one thing goes wrong I begin to feel bad and wonder if I can do well at anything at all”), self-criticism (e.g., “I am not satisfied with anything less than what I expected of myself’), and the use of high standards (e.g., “I am a perfectionist in my goals”).

Next, participants encountered the failure memory prompt. Specifically, they were directed to recall,

… a time when you failed at something that was important to you. This might be a social or interpersonal failure, an academic failure, a failure in a competitive event, or any time when you failed at something that was important to you.

It was specified that this memory should be from within the past 5 years and should be of an event that occurred at a particular time and place, not a summary of something that happened many times. Participants typed a cue word to identify the event they recalled and then continued to the next screen, which directed them to close their eyes and picture the event for a few moments.

The following screen introduced the distinction between first- person and third-person perspectives using a measure modeled on previous research (D. Cohen & Gunz, 2002; Libby & Eibach, 2002, 2011a). Instructions explained that memories are often accompanied by visual images and that,

With the first-person visual perspective, you see the event from the same visual perspective that you originally did; in other words, in your memory, you are looking out at your surroundings through your own eyes.

With the third-person visual perspective, you see the event from an observer’s visual perspective; in other words, in your memory you, can actually see yourself, as well as your surroundings.

Because participants could have experienced multiple images of the event, and some from each perspective (e.g., Huebner & Fredrickson, 1999), the questionnaire presented an 8-point scale ranging from entirely first-person (coded as 1) to entirely third person (coded as 8) that participants used to indicate the relative proportion of images experienced from each perspective.

Next, participants responded to four items that together indexed the extent to which they were overgeneralizing from the failure as they pictured it. These items were derived from the overgeneralization subscale of the ATS but worded so that they referred to overgeneralization from the recalled failure in particular, rather than overgeneralization as a chronic tendency. Specifically, participants rated how true they felt each of the following statements was: “When I think about this event, I feel like I am a failure”; “Even though this event is a failure, it’s just a one-time occurrence where I did not meet a specific goal” (reverse-scored); “When I think about this event, I wonder if I can do well at anything at all”; and ‘This single event influences how I feel about myself overall.” Participants responded using 5-point fully labeled scales with endpoints anchored at not at all true (coded as 1) and extremely true (coded as 5). These responses (a = .74) were averaged to create an index of event overgeneralization.

The next screen presented instructions for the Positive and Negative Affect Schedule (PANAS, Watson, Clark, & Tellegen, 1988). It was explained that a different feeling or emotion would be shown on successive screens and that for each word participants should “… indicate to what extent you feel that way right now as you picture the event.” Participants responded first to the emotion word, ashamed, and then proceeded to respond to each of the remaining 19 emotion words on the PANAS (10 positive and nine negative), including guilty. Participants responded using 5-point fully labeled scales with endpoints anchored at very slightly or not at all and extremely. Because our hypotheses pertain to shame, unique from guilt, we used a standard method for computing an index of guilt-free shame (Tangney & Dearing, 2002) in this and all subsequent studies involving shame. By this method, shame scores are predicted from guilt scores and the standardized residuals serve as an index of guilt-free shame.

At the end of the session, participants described the event they recalled, estimated the date on which it had occurred, and provided demographic information.

Results and Discussion

Predictor variables: Self-esteem and imagery perspective.

Self-esteem scores ranged from 2.00 to 4.00, out of the possible 1 (lowest self-esteem) to 4 (highest self-esteem), with a mean of 3.36 (SD = 0.46). Imagery perspective spanned the entire range of the scale, entirely first-person (coded as 1) to entirely third-person (coded as 8), with a mean of 3.36 (SD = 2.18). There was no significant association between self-esteem and imagery perspective (r = .04, p = .71).

Overgeneralization from the recalled failure. We predicted that lower self-esteem would be associated with greater overgeneralization from the recalled failure only to the extent that participants pictured it from the third-person perspective. We used linear regression to test this hypothesis. The model included self-esteem, imagery perspective, and their interaction as predictors. So that we could isolate the relation between imagery perspective for the recalled failure and overgeneralization from that particular failure, we controlled for participants’ chronic tendency to overgeneralize, as assessed by the ATS administered at the beginning of the session, b = .36, 3 = .40, t(76) = 2.97, p < .01. And because chronic overgeneralization was correlated with self-esteem (r = -.75, p < .001), one of the variables in our predicted interaction, we also controlled for the interaction between chronic overgeneralization and perspective (Yzerbyt, Muller, & Judd, 2004), b – -.08, 3 = -20, /(76) = 1.57, p = .12.1 The final covariate was memory age, b = -.005, 3 = -.17, /(76) = 2.06, p < .05, which we controlled for in all analyses of memories reported here to isolate our predicted effects from those that are associated with the mere passage of time (e.g., Gilbert, Pinel, Wilson, Blumberg, & Wheatley, 1998). In this regression analysis and all others reported here, we centered continuous predictor variables at the sample mean before computing the main analyses and recentered these variables at one standard deviation above and below the sample mean to test simple slopes at high and low levels of the variables (J. Cohen & Cohen, 1983).

Consistent with our prediction, the relationship between selfesteem and overgeneralization depended on imagery perspective, b = -.23, 3 = -3Q, /(76) = 2.32, p < .05 (see Figure 1). Lower self-esteem was associated with greater overgeneralization from the recalled failure only when it was pictured predominantly from the third-person perspective, b = -.95, 3 = —55, /(76) = 3.58, p < .01, and not when it was pictured predominantly from the first- person perspective, b = .04, 3 = .02, /(76) = 0.12. Thus, although, on average, low self-esteem was associated with greater overgeneralization from the recalled failure, b = -.45, 3 = -.26, /(76) = 2.00, p < .05, this effect depended on the failure being pictured from the third-person perspective.

We also investigated how perspective related to overgeneralization from the recalled failure at low and high levels of self-esteem. As predicted, third-person imagery was associated with marginally greater overgeneralization among LSEs, b = .10, 3 = -28, /(76) = 1.88, p = .06, but significantly less overgeneralization among HSEs, b = –.11, 3 = -.31, /(76) = 1.97, p = .05. Thus, perspective was not associated with overgeneralization directly, b = -.01, 3 = -.01, /(76) = 0.16; the effect depended on self-esteem.

Shame. We predicted that low self-esteem would be associated with greater shame only to the extent that failures were pictured from the third-person perspective. To test this hypothesis, we used linear regression to predict participants’ guilt-free shame scores from self-esteem, imagery perspective, and their interaction, with memory age as a covariate.

Indeed, the relationship between self-esteem and guilt-free shame depended on imagery perspective, b = -.31, 3 = –.31, /(78) = 3.02, p < .01 (see Figure 2).2 Lower self-esteem was associated with greater shame only when failures were pictured predominantly from the third-person perspective, b = -1.06, 3 = -.49, /(78) = 3.70, p < .001, and not when they were pictured predominantly from the first-person perspective, b = .26, 3 = -12, /(78) = 0.76. Thus, although low self-esteem was associated with marginally greater shame on average, b = -.40, 3 = -.18, /(78) = 1.78, p < .08, this trend depended on failures being pictured from the third-person perspective.

We also investigated how perspective related to shame at low and high levels of self-esteem. As predicted, among LSEs, third- person imagery was associated with greater shame, b = .18, 3 = .40, /(78) = 2.76, p < .01. Third-person imagery appeared to be associated with less shame among HSEs, although the effect was not significant, b = -.10, 3 = -.21, /(78) = 1.48, p = .14. Thus, perspective was not associated with shame directly, b = .04, 3 = .10, /(78) = 0.93; the effect depended on self-esteem. 1 2

Figure 1. Overgeneralization from recalled failure, plotted at 1 SD above and below the means of self-esteem and imagery perspective in Study 1, controlling for memory age, chronic overgeneralization, and its interaction with perspective.

Summary. LSEs’ tendency to overgeneralize from failure emerged only to the extent that failures were pictured from the third-person perspective. The same was true regarding shame, an emotional component of overgeneralization. Further, as predicted, picturing failure from the third-person perspective was associated with a greater negative impact among LSEs but a reduced negative impact among HSEs: LSEs overgeneralized more and experienced more shame when picturing their failure from the third-person than from the first-person perspective, whereas HSEs overgeneralized less when picturing from the third-person than from the first- person perspective.

Because Study 1 involved no intervention to manipulate imagery perspective, the results attest to the relevance of the predicted effects for understanding the naturally occurring dynamics of self-esteem, perspective, and overgeneralization as individuals think about failures in their lives. However, the correlational design of Study 1 also limits any conclusions about causation. Previous research suggests that thinking about an event in relation to the self-concept can cause people to adopt the third-person perspective (Libby & Eibach, 201 la); thus, it is possible that such an effect contributed to the pattern of results in Study 1. This would be consistent with the model that informs our predictions. If third-person imagery functions to represent events in relation to the self-concept, and if people shift perspective according to this function, then the relationship between imagery perspective and level of meaning in events should be bidirectional (Libby & Eibach, 2011b). However, because we were specifically interested in the potential power of imagery perspective to change LSEs’ proneness to overgeneralization, the remaining studies manipulated imagery perspective to isolate its causal role in producing the patterns observed in Study 1. Studies 2 and 3 focused on a cognitive component of overgeneralization (negativity of accessible self-knowledge) and Studies 4 and 5 focused on an emotional component (shame).

Study 2

Study 2 manipulated the visual perspective participants used to picture a personal failure and measured the effect on the association between self-esteem and the relative accessibility of additional personal failures versus successes. We predicted that LSEs would show greater negativity in accessible self-knowledge than would HSEs only when failures were pictured from the third-person perspective, not from the first-person. We also investigated the effects of imagery perspective at high and low levels of selfesteem. According to our account, using the third-person as opposed to first-person perspective should produce greater negativity only among LSEs. To the extent that perspective influences accessible self-knowledge among HSEs, third-person imagery should produce less negativity.


Participants. Three hundred seventeen undergraduates (210 women and 107 men) participated for course credit. Self-esteem data were not collected for 43 participants, so they could not be used in analyses. The final sample consisted of 274 participants, with ns per condition ranging from 63 to 72.

Materials and procedure. Upon signing up for the experiment, participants received an e-mail with a link to a secure website where they could complete the RSE. RSE scores ranged from 1.50 to 4.00, out of the possible 1 to 4, with a mean of 3.22 (SD = 0.47).

In the lab, participants sat in individual cubicles to complete a computerized questionnaire that included two tasks. The first was an imagery task that manipulated the perspective (first-person vs. third-person) participants used to picture a personal failure. The second was a timed event-listing task that manipulated the valence of additional events participants were instructed to list (successes vs. failures). These manipulations were crossed to create four versions of the questionnaire, to which participants were assigned randomly with the stipulation that there be approximately equal numbers per condition.

Figure 2. Guilt-free shame (standardized residuals) plotted at 1 SD above and below the means of self-esteem and imagery perspective in Study 1, controlling for memory age.


In all conditions, the questionnaire began with the failure memory prompt used in Study 1. Participants typed a cue word to identify the failure they recalled. Next, instructions introduced the manipulation of imagery perspective, adapted from previous research (e.g., Libby et al., 2005). These instructions varied by condition. In the first-person conditions, participants read,

You should use a FIRST-PERSON visual perspective to picture the failure. With the FIRST-PERSON visual perspective you see the event from the same visual perspective you had when the event originally occurred. That is, you are looking out at your surroundings through your own eyes.

In the third-person conditions, participants read,

You should use a THIRD-PERSON visual perspective to picture the failure. With the THIRD-PERSON visual perspective you see the event from the visual perspective an observer would have had when the event originally occurred. That is, you can see yourself in the image, as well as your surroundings.

Regardless of perspective condition, all participants were directed to close their eyes and form a vivid image of the event from the specified perspective. They were then instructed to hold the image in their minds while they answered the question on the next screen, which served as a manipulation check and varied by condition (first-person condition in italics; third-person condition in brackets): “As you’re picturing it right now, do you see the event [yourself in the event] from the visual perspective you had [an observer would have had] when the event originally occurred?” Participants responded by choosing yes or no. All participants passed this manipulation check.

Next, participants read instructions for the event-listing task designed to measure the relative accessibility of positive and negative self-knowledge. There were two versions of these instructions, either specifying that participants should list personal successes or personal failures. Within each perspective condition participants received one set of instructions or the other (success instructions in italics; failure instructions in brackets):

We are interested in [additional] instances where you [failed] succeeded at something that was important to you. You will have approximately 2 min to think of as many [failures] successes as you can. For each [failure] success (one at a time), please type a cue word or phrase that will identify this event to you.

The 2-min period began when participants advanced to the next screen where there was a text box to type the cue word for the first event they thought of. After typing the cue word, participants pressed return and the text box was cleared for them to type another cue word. They continued in this fashion, listing events at their own pace until the 2 min were up, at which point the computer automatically ended the task.

At the end of the session, participants provided additional information about their responses on the two tasks. They described the initial failure they had pictured and reported the approximate date on which it had occurred. They also classified the events they had listed during the timed task. Specifically, the computer presented the cue word for each event, and participants indicated whether the associated event represented a success, failure, or neither. These classifications allowed us to identify and exclude from the index of accessibility any instances in which participants listed events during the timed task that did not fit the specified category.[12]

Results and Discussion

We hypothesized that low self-esteem would be associated with greater negativity in accessible self-knowledge only when the initial failure had been pictured from the third-person perspective, not from the first-person. To test this hypothesis, we used linear regression to predict the number of events from the specified category that participants listed during the timed task from selfesteem (centered at the sample mean); imagery perspective (first- person = -1, third-person = 1); type of events participants were instructed to list during the timed task (failures = -1, successes = 1); all higher order interactions; and the covariate, memory age.

Does perspective determine the effect of self-esteem on relative negativity of accessible self-knowledge? Consistent with previous research showing an advantage for memory of positive life events (Walker, Skowronski, & Thompson, 2003), participants tended to list more successes than failures, overall, b = 1.75, p = .47, /(265) = 8.66, p < .001. However, in addition, supporting our hypothesis, whether self-esteem predicted the relative accessibility of successes versus failures depended on imagery perspective, b = .85, p = .11, t(265) = 1.95, p = .05 (see Figure 3).

When participants had used the third-person perspective, low self-esteem predicted relatively greater negativity in accessible self-knowledge, b = 1.79, P = .22, t(265) = 2.97, p < .01. In this condition, LSEs listed significantly fewer successes, b = 2.22, P = .27, /(265) = 2.60, p < .05, and marginally more failures, b = -1.36, p = -.17, /(265) = 1.60, p = .11, than did HSEs.

In contrast, when participants had pictured the initial failure from the first-person perspective, self-esteem was not a reliable predictor of negativity, b = .09, P = .01, /(265) = 0.14. In this condition, successes were more accessible than failures, b = 1.90, P = .51, /(265) = 6.54, p < .001, regardless of participants’ level of self-esteem. Thus, although low self-esteem was associated with relatively greater negativity on average, b = .94, p = .12, /(265) = 2.15, p < .05, this effect depended entirely on participants’ picturing the failure from the third-person perspective.

How does perspective influence the relative negativity of accessible self-knowledge among LSEs and HSEs? As predicted, picturing failure from the third-person rather than first- person perspective produced marginally greater negativity of accessible self-knowledge among LSEs, b = -.55, p = -.15, /(265) = 1.92, p < .06. No such negative effect of third-person imagery emerged among HSEs, who showed positive bias, b = 2.19, p = .58, /(265) = 7.56, p < .001, regardless of perspective, b = .25, P = .07, /(265) = 0.87. Thus, perspective did not influence the accessibility of successes versus failures directly, b = -.15, p = -.04, /(265) = 0.74; the effect depended on participants’ self-esteem.

Summary. As predicted, upon recalling personal failure, accessible self-knowledge was relatively more negative among LSEs than HSEs only when picturing the failure from the third- person perspective, not from the first-person perspective. Further, picturing failure from the third-person, rather than first-person, perspective caused accessible self-knowledge to become relatively more negative at low levels of self-esteem. Perspective had no significant effect at high levels of self-esteem. Theoretically, among HSEs, third-person imagery could have been expected to further reduce the negativity of accessible self-knowledge compared with first-person imagery. The fact that this did not occur could be due to the certainty of HSEs’ positive beliefs about their self-attributes (Riketta & Ziegler, 2007). It is also possible that the lack of perspective effect among HSEs reflects ceiling and floor effects on the number of successes and failures that could be generated in the specified time.



Figure 3. Number of successes and failures listed in the timed task of Study 2, depending on self-esteem, and imagery perspective for previous failure memory, controlling for memory age. Values are plotted at 1 SD above and below the sample mean of self-esteem.


What is most important for our hypothesis, however, is that low self-esteem predicted greater negativity only when failures were pictured from the third-person and that imagery perspective had a different effect on negativity of accessible self-knowledge for low- and high-self-esteem individuals, with third-person imagery producing greater negativity among LSEs only. To determine whether the patterns observed in Study 2 were reliable, we sought to replicate them in Study 3 while exercising stricter control over the type of failures participants pictured and the way accessibility of positive and negative self-knowledge was measured.

Study 3

As did Study 2, Study 3 manipulated the perspective participants used to picture personal failure and measured the negativity of accessible self-knowledge. However, whereas in Study 2 participants nominated failures from their own lives to picture, in Study 3 all participants pictured the same hypothetical failure events that we specified. And whereas Study 2 measured the relative accessibility of positive and negative self-knowledge with a task that required participants to explicitly report information about themselves, Study 3 used an implicit measure (Dodgson & Wood, 1998, Experiment 2). Thus, we could isolate the effect of perspective and self-esteem on accessibility of self-knowledge apart from willingness to report it or knowledge that it was being assessed.

Despite these stricter controls in Study 3, we expected to replicate the pattern of results observed in Study 2. Specifically, lower self-esteem would be associated with greater negativity in accessible self-knowledge only when individuals pictured failure from the third-person perspective and not when they used the first- person. And we expected that only among LSEs would picturing failure from the third-person, rather than first-person, perspective cause accessible self-knowledge to become more negative. We expected no such effect of perspective among HSEs: The results of Study 2 suggested that negative information about the self remains relatively inaccessible among HSEs, regardless of imagery perspective.


Participants. One hundred twenty-one undergraduates (60 women, 60 men, and 1 unidentified) participated for course credit.

Materials and procedure.

Overview. Online, before coming into the lab, participants completed the RSE and a self-attributes questionnaire that was used to identify their personal strengths, weaknesses, and neutral attributes. In the lab, participants sat in individual cubicles to complete two allegedly unrelated tasks on a computer. The first was an imagination task that manipulated the imagery perspective participants used to picture personal failure scenarios. The second was a response-time task that was used to measure the accessibility of the personal strengths, weaknesses, and neutral attributes that were identified in the online survey. The online self-attributes questionnaire and the response-time task were the same for all participants and were taken directly from Dodgson and Wood (1998, Experiment 2).[13] The failure scenarios were adapted from the Test of Self-Conscious Affect-3 (TOSCA-3; Tangney & Dearing, 2002) and the Test of Self-Conscious Affect for Adolescents (TOSCA-A; Tangney & Dearing, 2002). The only manipulation was that of imagery perspective, and participants were randomly assigned to condition, with the stipulation that the numbers per condition be approximately equal.

Online questionnaire: Measuring self-attributes and selfesteem. Upon signing up for the experiment, all participants received an e-mail with a link to a secure website where they could complete a modified version of the Self-Attributes Questionnaire (SAQ, Pelham & Swann, 1989) used by Dodgson and Wood (1998). For each of 12 attributes (e.g., intelligence, social skills.


athletic ability) participants used a 10-point fully labeled scale to indicate how their abilities compared with those of «other college students.” The endpoints of the scale were anchored at extremely below average and extremely above average.

Self-esteem was measured using the RSE, which participants completed either as part of the online questionnaire or as part of a mass pretesting survey administered to students in the participant pool at the beginning of the academic quarter. Self-esteem scores of participants in Study 3 ranged from 1.50 to 4.00, out of the possible 1 to 4, with a mean of 3.04 (SD = 0.49).

Lab task 1: Manipulating imagery perspective. There were two versions of this task, which differed only in the imagery perspective instructions. Participants were told that they would hear a variety of scenarios described through headphones, and they should follow specific instructions to picture each scenario as it was described. They then put on the headphones and heard the imagery instructions read aloud as they appeared on the screen. These instructions varied by condition and were the same as those in Study 2, except that they referenced a hypothetical future event rather than a past event (e.g., Libby et al., 2007). In both conditions instructions explained that participants should use the specified perspective to form as clear an image as possible using all of their senses until it felt real, and they should take as much time as they needed to form the image.

Participants then had the chance to familiarize themselves with the procedure by picturing an emotionally neutral practice scenario modeled on the practice scenario in the TOSCA-A (Tangney & Dearing, 2002). When participants had the image in mind from the specified perspective, they opened their eyes and advanced to the next screen, where they responded yes or no to a manipulation-check question, which varied by condition (first-person condition in italics; third-person condition in brackets); “As you’re picturing it right now, do you see the situation from the visual perspective you [an observer] would have if the scenario were actually taking place?” All participants passed this manipulation check.

Next, participants advanced to the main part of the imagination task, where they imagined themselves in 10 failure scenarios spanning the social, academic, and moral domains (e.g., injuring a friend, doing poorly on an exam, breaking a friend’s possession and hiding it). Each one was presented aloud through the headphones and began with instructions for participants to close their eyes and use the specified perspective to form a visual image. After the scenario was described and participants formed an image, they held that image in mind and answered two questions that helped to bolster the cover story that the imagination task served a purpose in and of itself. The first question also constituted a check that participants were actually forming a mental image. Participants were asked to rate the vividness of their image by indicating no image at all or by choosing an option on a fully labeled 4-point continuum that ranged from vague and dim to perfectly clear and as vivid as normal vision (Marks, 1973). Then participants rated how easy it was to picture the situation from the specified perspective. They used a fully labeled 7-point scale with endpoints anchored at extremely easy and extremely difficult. After imagining and rating their images for all 10 scenarios, participants received instructions to take off their headphones and advance to the second task in the session.

Lab task 2: Measuring relative accessibility of strengths, weaknesses and neutral attributes. Instructions explained that the task was made up of a number of trials in which a statement would appear on the screen, and participants were to decide whether that statement .. COULD EVER POSSIBLY be used to describe you, however unlikely you might think it.” Participants were to press the “z” key to indicate “yes” and the “/” key to indicate “no.” It was stressed that participants should be extremely liberal in these judgments. The statement “tall” was used as an example, and instructions explained that even participants who were physically short could answer “yes,” because in certain contexts, such as in relation to children, “tall” could logically be used to describe them. Instructions also directed participants to respond as quickly as possible without sacrificing accuracy. Participants were directed to place one finger over each response key to aid in making quick responses.

After completing 10 practice trials participants went on to complete the main task, which consisted of 239 trials. One hundred twenty of those trials involved statements that were related to the 12 attributes on the SAQ that participants had completed before taking part in the lab portion of the study. There were 10 words related to each of the attributes (e.g., words related to the SAQ attribute “intelligent” included smart, clever, wise). All of these words related to SAQ attributes could logically be used to describe any person given the criterion specified in the instructions. Thus, it was intended that all participants would answer “yes” to all of these items. The remaining 119 words were nondescriptors that could not be logically used to describe any person (e.g., marshy, electronic), and thus, it was intended that all participants would answer “no” to all of these items.

Preparing data for analysis. We followed Dodgson and Wood’s (1998, Experiment 2) procedure to use participants’ response times on the 120 SAQ attribute trials and their self-ratings on the online SAQ to compute an index of accessibility for each SAQ attribute and to identify it as a strength, weakness, or neutral attribute. Excluding response times for items where participants gave incorrect responses (6% of trials) and response times that were more than three standard deviations away from the participant’s mean response time (2% of trials), we computed for each participant the average of the response times for the words associated with each SAQ attribute, thus creating an index of accessibility for each attribute for each participant. Attributes were then defined as personal strengths, weaknesses, or neutral for each participant idiographically. On a within-participant basis, we standardized self-ratings on the 12 attributes of the SAQ. Attributes were considered strengths if they were rated at least .65 SDs above the participant’s mean, weaknesses if they were rated at least .65 SDs below the participant’s mean, and neutral if they were rated less than .65 SDs away from the participant’s mean in either direction.

Eleven participants were excluded from analyses because their personal strengths, weaknesses, and neutral attributes could not be identified–either because they did not complete the SAQ before coming into the lab (five in first-person and three in third-person) or because they provided the same ratings on all of the attributes (two in first-person and one in third-person). Eleven participants were excluded because of errors that occurred during the lab session, either in the computer’s recording of response time (one in first-person and two in third-person) or in participants’ following the instruction to form a visual image of all target scenarios (seven in first-person and one in third-person). Thus, the final sample consisted of 99 participants (first-person n = 51; third-person n = 48).




Results and Discussion

We used multilevel modeling techniques to regress participants’ average response times for each SAQ attribute on attribute category (strengths vs. weakness vs. neutral, coded with weaknesses as the reference group), self-esteem (centered at the sample mean), imagery perspective (first-person = -1, third- person = 1), and all higher order interactions. Because the assignment of attributes to category was idiographic, we also included attribute (e.g., intelligence, social skills, athletic ability) as a categorical covariate (effect coded). This allowed us to test our prediction about the interactive effect of self-esteem and perspective on the relative accessibility of strengths, weaknesses, and neutral attributes, independent of how response time varied according to the attributes themselves or the words used to assess them.

Does perspective determine the effect of self-esteem on relative negativity of accessible self-knowledge? Consistent with our hypothesis and the results of Study 2, whether self-esteem predicted the relative accessibility of strengths, weaknesses, and neutral attributes depended on imagery perspective, F(2, 1070) = 5.17, p < .01. Specifically, the relative accessibility of attributes depended on self-esteem when participants had pictured failure scenarios from the third-person perspective, F(2, 1070) = 7.01, p < .001, but not when they had used the first-person, F(2, 1070) = .49 (see Figure 4).

When participants had pictured failure scenarios from the third- person perspective, LSEs’ weaknesses were more accessible than their neutral attributes, у = 53.28, /(192) = 2.49, p < .05, and marginally more accessible than their strengths, у = 39.63, r(192) = 1.67, p < .10, whereas HSEs’ weaknesses were less accessible than their neutral attributes, у = -59.26, r( 192) = 2.61, p < .01, and no more accessible than their strengths, у = -27.38, /(192) = 1.08, p = .28.

In contrast, when participants had pictured failure scenarios from the first-person perspective, no discernible self-esteem differences emerged. Weaknesses were less accessible than neutral attributes, у = -47.29, /(192) = 3.07, p < .01, or strengths, у = -47.50, /(192) = 2.84, p < .01, regardless of self-esteem level. Thus, self-esteem did not influence the relative negativity of accessible self-knowledge directly, F(l, 95) = .60; the effect depended entirely on scenarios being pictured from the third-person perspective.

How does perspective influence the relative negativity of accessible self-knowledge among LSEs and HSEs? We also investigated the effects of perspective at low and high levels of self-esteem. As predicted, and consistent with results from Study 2, picturing failures from the third-person, rather than first-person, perspective increased the negativity of accessible self-knowledge among LSEs, F(2, 192) = 8.02, p < .001. Also consistent with the results of Study 2, among HSEs, weaknesses were less accessible than their neutral attributes, у = -45.34,

Figure 4. Response times for words associated with personal strengths, weaknesses, and neutral attributes, depending on self-esteem and perspective in Study 3. Values are plotted at 1 SD above and below the sample mean of self-esteem.

/(192) = 2.75, p < .01, and marginally less accessible than their strengths, у = -30.72, /(192) = 1.69, p = .09, regardless of perspective. Thus, perspective did not influence the negativity of accessible self-knowledge directly, F(l, 95) = .35; the effect depended on participants’ self-esteem.

Summary. Previous research suggests that LSEs’ tendency to think negatively about the self in general on the basis of a specific failure reflects the influence of their self-concepts in defining the subjective impact of failure (Conner Christensen et al., 2003; Kemis et al., 1989). Studies 2 and 3 demonstrated that when individuals recalled and imagined failure, a self-esteem difference in this cognitive manifestation of overgeneralization emerged only when individuals pictured failure from the third-person perspective. These findings are consistent with the idea that people’s self-concepts shape reactions to events when they are pictured from the third-person perspective but not from the first-person. Also consistent with that conclusion, negative effects of third- person imagery emerged among LSEs only.

As did Studies 2 and 3, the remaining studies manipulated the perspective individuals used to picture personal failures. However, this time the dependent measures tapped into individuals’ feelings of shame to test whether perspective also has the power to determine whether low self-esteem predicts this emotional manifestation of overgeneralization.

Study 4

Study 4 manipulated the perspective that participants used to picture the failure scenarios from Study 3. However, instead of measuring the accessibility of positive and negative self- knowledge, we measured participants’ feelings of shame in response to each scenario. This allowed us to test for the causal role of perspective in producing the pattern of shame results observed in Study 1. Again, we predicted that LSEs would experience greater shame than HSEs only when scenarios were pictured from the third-person perspective. Further, we expected that to the extent perspective influenced shame at low and high levels of self-esteem, picturing failure from the third-person as opposed to first-person perspective would produce greater shame among LSEs and less shame among HSEs.


Participants. Sixty-eight undergraduates (30 women and 38 men) participated for course credit. Two (one in each perspective condition) were excluded from analyses because they did not complete the online measure of self-esteem. All participants passed all imagery check questions. Thus, the final sample consisted of 66 participants (33 in each perspective condition).

Materials and procedure. As in Studies 2 and 3, self-esteem was measured online using the RSE before participants came into the lab. Scores ranged from 2.50 to 4.00, out of the possible 1 to 4, with a mean of 3.30 (SD = 0.44).

In the lab, participants sat in individual cubicles to complete a computerized imagination task that involved the same scenarios as the task in the first half of the lab session in Study 3. The imagination task in Study 4 followed the same procedure as the one in Study 3, with two exceptions. The first difference was a minor procedural detail: Instead of hearing the imagery perspective instructions and scenarios through headphones as in Study 3, participants in Study 4 heard these aspects of the procedure read aloud by the experimenter.

The second difference between the imagination task in Studies

3 and 4 was the important one. In both studies, for each failure scenario, participants formed an image from the specified perspective and rated this image on ease and vividness. However, in Study

4 participants also then responded to items that allowed us to assess their feelings of shame. The inventory from which the scenarios were adapted, the TOSCA, is designed specifically for this purpose. For each scenario, respondents on the TOSCA rate how likely they would be to have each of four reactions–one corresponds to a behavioral or emotional manifestation of shame; the others correspond to manifestations of guilt, detachment, and extemalization. For example, for the scenario that involves doing poorly on an exam respondents rate how likely they would be to (a) “feel that you should have done better and should have studied more” (guilt); (b) “feel stupid” (shame); (c) “think, ‘It’s only a test’” (detachment); and (d) “think, ‘The teacher must have graded it wrong’” (extemalization).

In the typical administration of the TOSCA, respondents are given no visualization instructions; in Study 4, participants rated the likelihood they would experience each reaction to each scenario as they pictured it from the specified perspective. The position of the shame item among the four responses was varied across scenarios in Study 4 according to the way the items appear in the original TOSCA scales. Participants responded to each item using a 5-point fully labeled scale anchored with endpoints not at all likely (coded as 1) and very likely (coded as 5). For each scenario, we computed a guilt-free shame score for each participant following the same procedure as in Study 1, predicting shame from guilt and using the standardized residuals as an index of guilt-free shame.

Results and Discussion

We predicted that lower self-esteem would be associated with greater shame only when failures were pictured from the third- person perspective, not from the first-person. To find out whether this predicted interaction of self-esteem and perspective would hold up reliably across the different scenarios, we used a mixed linear model to predict participants’ guilt-free shame scores on each scenario from self-esteem (centered at the sample mean), perspective (first-person = -1, third-person = 1), scenario (effect coded), and all higher order interactions.

Does perspective determine the effect of self-esteem on shame? The interaction between perspective and self-esteem fell just short of standard levels of significance, у = -.25, r(62) = 1.88, p = .06, although Figure 5 shows that the results were clearly in line with our predictions and the results of Study 1. The more focused test of our predicted pattern corroborated this interpretation. Lower self-esteem was associated with greater shame only when scenarios were pictured from the third-person perspective, у = -.60, r(62) = 3.03, p < .01, not from the first-person, у = -.11, t(62) = 0.67, p = .50. Thus, low self-esteem was associated with greater shame on average, у = -.36, r(62) = 2.75, p < .01, but this effect appeared to depend on failures being pictured from the third-person perspective.

How does perspective influence shame among LSEs and HSEs? We also investigated the effects of perspective at low and high levels of self-esteem. The patterns were consistent with our hypotheses, with the predicted values of shame being greater with third-person imagery among LSEs, у = .10, f(62) = 1.22, p = .23, and reduced among HSEs, -y = –.11, /(62) = 1.47, p = .15, although neither effect was significant.


Figure 5. Guilt-free shame (standardized residuals) depending on selfesteem and perspective in Study 4. Values are plotted at 1 SD above and below the sample mean of self-esteem.


Summary. As predicted, manipulating imagery perspective revealed its causal power in controlling self-esteem differences in shame upon picturing failure. Low self-esteem predicted greater shame only when individuals pictured the scenario from the third- person perspective, and not from the first-person. Further, the pattern of results was consistent with the idea that third-person imagery has the potential either to increase or to decrease shame, depending on self-esteem. In the final study, we tested the implications of these results for the shame that individuals experience when recalling real failures from their own lives, and we sought converging evidence for our interpretation.

Study 5

Study 5 manipulated the visual perspective participants used to picture a personal failure they recalled and measured the shame they felt as a result. We expected to replicate the pattern of results observed in Studies 1 and 4, where low self-esteem was associated with greater shame only when failures were pictured from the third-person perspective and not from the first-person. According to our account, this pattern reflects the influence of the self- concept in shaping the subjective meaning of life events when pictured from the third-person perspective. An alternative possibility is that the shame LSEs felt when picturing the failure from the third-person perspective was a reaction to the image of themselves, apart from the meaning of the event they were picturing. To distinguish between these two accounts, we included conditions in Study 5 in which participants pictured a success or a neutral incident. According to our account, the interactive effect of selfesteem and perspective on shame should be limited to picturing failure; the alternative would predict the interaction to emerge regardless of event.


Participants. One hundred forty-three undergraduates (69 women, 73 men, and 1 unidentified) participated for course credit. Fifteen were excluded because they did not provide self-esteem data, and 11 were excluded because they either did not follow the instructions for the failure memory prompt (n = 6) or did not pass the perspective manipulation check (three in first-person and two in third-person).

Materials and procedure. As in Studies 2-4, self-esteem was measured online using the RSE scale before participants came into the lab. Scores ranged from 1.90 to 4.00, out of a possible 1 to 4, with a mean of 3.10 (SD = 0.47).

In the lab, participants sat in individual cubicles to complete a computerized questionnaire in which they visualized an event from their lives and then rated their current emotions. Participants were randomly assigned to one of six versions of this questionnaire, created by crossing manipulations of the event type (failure, success, or neutral) and imagery perspective (first-person or third- person).

In the failure conditions, participants received the same failure memory prompt as in Studies 1 and 2. In the success conditions, this memory prompt was modified to refer to a success, rather than a failure. Participants in the failure and success conditions then encountered the same imagery perspective manipulation as in Study 2, either directing them to use the first-person or third- person perspective (depending on condition) to picture the event they recalled. In the neutral conditions participants were also assigned to use either the first-person or third-person perspective to picture an incident in their lives. However, the incident was neutral: tying their shoe.

After participants pictured the specified event, they responded yes or no to the manipulation check question used in Study 2. Participants then completed the PANAS in the same format as in Study 1. The ashamed and guilty items were used to compute an index of guilt-free shame following the same procedure as in Studies 1 and 4.

At the end of the session, participants in the failure and success conditions described the event they recalled, provided the approximate date on which it occurred, and rated its valence on a 7-point fully labeled scale with endpoints anchored at extremely negative and extremely positive. The event descriptions and valence ratings were used as a check on the memory prompt manipulation.

Results and Discussion

We predicted that when participants pictured failure, perspective and self-esteem would interact to predict shame according to the pattern in Studies 1 and 4; further, this pattern would be unique to picturing failure. To test these hypotheses, we submitted the index of guilt-free shame to a linear regression model using self-esteem (centered at the sample mean), imagery perspective (first-person = -1, third-person = 1), and event type (dummy coded with failure as the reference group), along with all higher order interactions as predictors. We conducted follow-up analyses to test simple slopes at one standard deviation above and below the sample mean of self-esteem and within each of the manipulated conditions. Results provided strong support for our predictions.

Failure events. In the failure condition, perspective determined the effect of self-esteem on shame, b = -1.29. |3 = -.60, /(105) = 3.53, p < .01 (see Figure 6). As predicted, lower self-esteem was associated with greater shame only when participants pictured failure from the third-person perspective, b = -1.58, p = -.74, /(105) = 2.71, p < .01, not the first-person, b – 1.00, p = .47, /(105) = 2.26, p < .05.[14]

We also investigated the effects of perspective at low and high levels of self-esteem. As predicted, picturing failure from the third-person rather than first-person perspective increased shame among LSEs, b = .70, p = .70, /(105) = 3.28, p < .01, but decreased shame among HSEs, ft = -.51, p = .52, /(105) = 2.13, p < .05. Thus, perspective did not influence shame directly, ft = .09, p = .09, /(105) = 0.63; the effect depended on self-esteem.

Success and neutral events. Considering the failure condition in relation to the success and neutral conditions revealed that the predicted interactive effect of self-esteem and perspective was unique to failure, as predicted: versus success, ft = 1.33, P = .34, /(105) = 2.71, p < .01; versus neutral, ft = 1.28, p = .38, /(105) = 2.70, p < .01 (see Figure 6). Neither perspective, self-esteem, nor their interaction had any discemable effects on shame in the success or neutral conditions (Irsl < .9, ps >.35). These results are consistent with the idea that the results in the failure conditions reflect the effect of perspective on the role of the self-concept in shaping reactions to failure, rather than an effect of perspective on eliciting self-esteem differences in self-feelings apart from failure.

General Discussion

The present studies demonstrate that when individuals mentally image failures–real or hypothetical–the visual perspective they use determines whether LSEs are prone to overgeneralization. Study 1 demonstrated that the more third-person imagery people spontaneously experienced as they pictured real past failures in their lives, the more that low self-esteem put them at risk for overgeneralization. Further, third-person imagery was associated with greater overgeneralization among LSEs but less among HSEs. The remaining studies manipulated perspective to show its causal influence in producing these patterns. Table 1 summarizes the results. Picturing failure from the third-person perspective caused LSEs to overgeneralize more than HSEs, both in terms of self- knowledge accessibility and feelings of shame. In contrast, no such effects of self-esteem emerged when individuals pictured failure from the first-person perspective. Further, among LSEs, picturing failure from the third-person, as opposed to first-person, perspective produced greater negativity in accessible self-knowledge and greater shame. Among HSEs, no such detrimental effects of third- person imagery occurred–only beneficial effects. Among HSEs, picturing failure from the third-person as opposed to first-person perspective reduced shame. These findings help to distinguish between different accounts of how imagery perspective functions, and they contribute to an understanding of the subjective processes that drive self-esteem differences and the experience of shame. We discuss implications for the study of top-down and bottom-up influences on self-judgment and emotion as well as for the role of perspective and abstraction in coping.




Figure 6. Guilt-free shame (standardized residuals) depending on selfesteem, imagery perspective, and event-type in Study 5. Values are plotted at 1 SD above and below the sample mean of self-esteem.

Table 1

The Interactive Effect of Imagery Perspective and Self-Esteem on Relative Negativity of Accessible Self-Knowledge (Studies 2 and 3) and Shame (Studies 4 and 5), Aggregated Across Studies That Manipulated Imagery Perspective

Effect Average effect size r
Relative negativity of accessible self-knowledge”
Perspective interacts with self-esteem” .15**
Effect of self-esteem  
with third-person: LSE –» greater negativity .20****
with first-person: ns .01
Effect of perspective  
with LSE: third-person –» greater negativity .16**
with HSE: ns .06
Perspective interacts with self-esteem . .36***
Effect of self-esteem  
with third-person: LSE –» greater shame .29***
with first-person: ns -.11
Effect of perspective  
with LSE: third-person –* greater shame .25***
with HSE: third-person –> less shame .20*

Note. LSE = low self-esteem; HSE = high self-esteem. Average effect sizes are weighted for sample sizes (Hedges & Olkin, 1985). Before computing the average, effect sizes in each study were assigned a positive sign if the effect was in the predicted direction and a negative sign if it was not. In the case of the effect of self-esteem with first-person imagery, we expected no significant effect and thus arbitrarily assigned a positive sign when LSE was associated with greater negativity/shame and a negative sign when LSE was associated with less negativity/shame. Significance levels represent combined ps, weighted for sample sizes (Stouffer, Such- man, DeVinney, Star, & Williams, 1949).

“ Relative negativity of accessible self-knowledge reflects the accessibility of negative, relative to positive (Study 2) or relative to positive and neutral (Study 3), self-knowledge. Thus, the interaction of self-esteem and perspective on relative negativity corresponds to the Self-Esteem X Perspective X Self-Knowledge valence interaction terms in Studies 2 and 3; the effect of self-esteem with each perspective corresponds to the Self- Esteem X Self-Knowledge valence interaction terms at each level of the perspective manipulation; the effect of perspective with high and low self-esteem corresponds to the Perspective X Self-Knowledge valence interaction terms at 1 SD above and below the sample mean of self-esteem. b Shame was indexed, apart from guilt, by the residuals left after predicting shame responses from guilt responses (Tangney & Dearing, 2002). The interaction of self-esteem and perspective on shame corresponds to the Self-Esteem X Perspective interaction terms in Studies 4 and 5; the effect of self-esteem with each perspective corresponds to the self-esteem terms at each level of the perspective manipulation; the effect of perspective with high and low self-esteem corresponds to the perspective term at I SD above and below the sample means of self-esteem.

><.05. *><.01. **><.001. ***><.0001.



The Function of Imagery Perspective

Previous research suggests that the self-concepts of LSEs guide their processing of failure in ways that promote overgeneralization, maximizing the failure’s negative impact on the self more broadly. The self-concepts of HSEs guide their processing of failure in ways that minimize its negative impact on the self more broadly (Conner Christensen et al„ 2003; Kemis et al., 1989). The fact that this self-esteem difference emerged in the present studies only when individuals pictured failure from the third-person, and not first-person, perspective provides insight into the function of imagery perspective in representing life events.

By some accounts, imagery perspective functions to determine whether people interpret events in an involved, emotional manner from the first-person perspective versus a detached, unemotional manner from the third-person perspective (e.g., Kenny et al., 2009; Mclsaac & Eich, 2004; Sanitioso, 2008; Williams & Moulds, 2008). However, in the present studies, individuals’ self-beliefs (as indexed by their self-esteem) biased their reactions to failure more from the third-person than from the first-person perspective, not less, as might be expected by a detachment account. In addition, imagery perspective did not influence the emotional experience of shame directly; the effect depended on self-esteem. Whereas HSEs felt less shame with third-person than first-person imagery, LSEs felt more.

By a different account, imagery perspective functions to define the level of meaning in event representations (Libby & Eibach, 2011b). According to this account, picturing events from the first-person perspective involves a bottom-up style of making meaning in which people incorporate information about the concrete features of the pictured situation and define the event in terms of these constituent aspects. Picturing events from the third-person perspective involves a top-down style of making meaning in which people integrate the pictured event with a broader context and define the event in terms of that abstract meaning. The present findings support and further specify this account, highlighting the potential utility of imagery perspective as a methodological tool in studying social cognitive processes.

Imagery perspective moderates the effect of the self-concept on reactions to life events. Consistent with the idea that imagery perspective functions to define the level of meaning in event representations, picturing actions and events from the third-person, rather than first-person, perspective causes people to construe those actions and events more abstractly (Libby et al., 2009; Vasquez & Buehler, 2007). But with any event, abstraction can occur on a variety of dimensions, with divergent implications (Vallacher & Wegner, 1985). For example, a failure could be construed abstractly as “a learning experience” or as “a confirmation of inadequacies.” The fact that the effect of perspective on overgeneralization in the present studies depended on self-esteem suggests that the pictured event, alone, does not shape abstraction. If it did, then third-person imagery should have uniformly produced greater overgeneralization and greater shame. And, if third- person imagery simply activated an abstract evaluation of the self, then LSEs should have felt more shame with third-person than first-person imagery, regardless of what event they pictured. However, LSEs felt more shame with third-person than first-person imagery only when they pictured failure. The fact that self-esteem and perspective interacted, for failure events in particular, suggests that the abstraction that has been demonstrated to occur when people picture life events from the third-person perspective (Libby et al., 2009; Vasquez & Buehler, 2007) is shaped by the self- concept. Thus, as we hypothesize, third-person imagery functions to represent life events in terms of the conceptual “me.” This means that, from the third-person perspective, the event is not necessarily understood as reflective of the seifs internal disposition, as some have suggested (e.g., Frank & Gilovich, 1989). Rather, the event is understood in terms of the broader meaning specified by knowledge structures and motivations that define the self-concept, and this could lead to more or less dispositional attributions, depending on the contents of the person’s self- concept.

Although self-esteem is a particularly relevant aspect of the self-concept with regard to shaping responses to failure, other aspects of the self-concept are relevant in shaping responses to other types of events. For example, internal working models of attachment have been shown to predict reactions to relationship events (Collins, Ford, Guichard, & Allard, 2006), and self-theories of change and stability have been shown to predict judgments of change and stability in the self since an event occurred (Ross, 1989). In both of these contexts, these aspects of the self-concept have been found to influence responses to events when they are pictured from the third-person, but not first-person, perspective (Libby et al., 2005; Marigold, Eibach, Libby, Ross, & Holmes, 2010)–consistent with the effects observed for self-esteem in the present studies. Thus, the present findings contribute to establishing that third-person imagery functions to represent life events in terms of their subjective meaning in the context of one’s life more broadly.

Imagery perspective moderates the effect of top-down versus bottom-up influences.

Common to many existing accounts of how imagery perspective functions in memory is an assumption that picturing events from the first-person perspective promotes reliving of past emotions (e.g., Kenny et al., 2009; Mclsaac & Eich, 2004; Nigro & Neisser, 1983; J. A. Robinson & Swanson, 1993). However, in the studies reported here the patterns that are typically observed in online reactions to failure (e.g., Brown & Marshall, 2001; Dodgson & Wood, 1998; Kemis et al., 1989) are the ones that we observed when participants pictured failure from the third-person perspective, not first-person. The idea that first- person imagery promotes reliving implies an assumption that first-person imagery should produce a more veridical replay of past events. This is intuitively plausible, given that one’s original experience of the event was necessarily encoded from the first- person visual perspective. However, given that memory is a reconstructive process, rather than a replay of the past (Bartlett, 1932; Neisser, 1967), it may be more appropriate to consider how perspective would influence the process by which people reconstruct events, rather than only to consider the match between the visual perspective at encoding and retrieval.


In the online experience of events, both top-down and bottom-up influences can operate to determine reactions (e.g., Bruner & Postman, 1949; Farb et al., 2007; Ochsner, 2007). The present findings suggest that imagery perspective may determine the extent to which each influence contributes as people mentally simulate events in memory (and imagination), with top-down influences exerting a stronger influence from the third-person than from the first-person perspective and bottom-up influences exerting a stronger influence from the first-person than from the third- person perspective (also see Libby, Valenti, Hines, & Eibach, 2011). By this account, first-person imagery may indeed enhance the reliving of aspects of experience that were driven bottom-up by features of the event itself, but third-person imagery may enhance reliving of aspects of experience that reflect the top-down influence of the self-concept.

Apart from whether the reactions that emerged with third-person imagery in the present studies reflect a sense of reliving, the results suggest that manipulating imagery perspective in mental simulation could be a methodological tool with broad applicability. The question of how aspects of the self-concept shape judgment and emotion is of interest not only to those studying self-esteem but also to those studying a variety of self-beliefs (e.g., Dweck & Grant, 2008; Ehrlinger & Dunning, 2003; Ehrlinger, Gilovich, & Ross, 2005; Pronin, Gilovich, & Ross, 2004; M. D. Robinson & Clore, 2002; Ross, 1989). Together with the findings mentioned earlier, the present results suggest the potential utility of imagery perspective manipulations for understanding the role of the self- concept in a wide range of social psychological phenomena.

Relation to Existing Research on Self-Esteem

The present findings are directly relevant to understanding selfesteem differences in response to failure. The idea that subjective perceptions of failure (apart from objective features of the event) are crucial in explaining LSEs’ proneness to overgeneralization comes from research that has focused on immediate reactions to failures in the present (e.g., Brown & Marshall, 2001; Dodgson & Wood, 1998; Kemis et al., 1989). The results of the studies we report here suggest that subjective perceptions also play a key role in explaining reactions to recalled and imagined failure (also see Conner Christensen et al., 2003). Simply varying the visual perspective from which participants pictured failure was enough to turn typical self-esteem differences on (third-person) or off (first- person).

It would be intuitive to hypothesize that the effects of selfesteem on self-knowledge accessibility in the present studies accounted for the effects of self-esteem on shame. However, evidence suggests that self-esteem can affect self-feelings directly, rather than by means of activating specific self-knowledge that supports those feelings (Dutton & Brown, 1997). At the same time, as mentioned earlier, there is evidence that self-esteem can influence self-thoughts in response to failure, thus suggesting that the two effects are independent. This might explain why we found third-person imagery to reduce shame among HSEs, even though we found no effect of imagery perspective on the negativity of HSEs’ accessible self-knowledge (see Table 1). This difference could be interpreted as consistent with the idea that cognitive and affective manifestations of overgeneralization can operate independently, at least for HSEs.

What is central to the purpose of the present experiments is the differential effect of self-esteem with third-person versus first- person imagery. Both components of overgeneralization–selfthoughts and self-feelings–have been proposed to reflect a top- down influence of self-esteem as an element of the self-concept. If third-person imagery functions to integrate specific events with the self-concept, then the previously documented effects of selfesteem on overgeneralization should show up only when failures are pictured from the third-person and not from the first-person, a pattern that emerged consistently and reliably across all five studies reported here. Thus, these findings provide converging evidence for the role of subjective construal processes in accounting for self-esteem differences in reactions to failure.

Relation to Existing Research on Shame

Subjective construal processes have also been proposed as key in understanding the emotion of shame: Feelings of shame appear to be determined not by objective features of actions and events (e.g., type of behavior or presence of others) but, rather, by subjective perceptions of them (Tangney, Miller, Flicker, & Bar- low, 1996; Tracy & Robins, 2004). The present findings are consistent with this conclusion, demonstrating that shame varied according to an interaction between self-esteem and the way people pictured their failures, apart from objectively specified features of those failure events.

Shame has been characterized as a negative evaluation of the self as an entity that transcends any given occasion (e.g., Lewis, 1971; Tangney & Dearing, 2002). Thus, shame is a self-conscious emotion in the sense that it is an emotional reaction to the self, rather than to the environment (Tracy & Robins, 2004). The present results highlight, however, that shame is not a necessary outcome of focusing (visual) attention on the self, even if one is engaged in a potentially shame-worthy action. There was no hint of a main effect of perspective on shame in any of the studies reported here. Third-person imagery increased shame only among LSEs picturing failure. Perspective had no effect on shame when individuals pictured success or neutral incidents, and third-person imagery actually reduced shame among HSEs picturing failure. Thus, the self-concept appears to be an important factor in shaping the subjective construal processes underlying shame.

Imagery Perspective, Abstraction, and Well-being

It has often been assumed that, to the extent imagery perspective plays a role in coping with negative events, it is because third- person imagery reduces the personal connection to events and/or the emotion associated with recall, thus promoting avoidance and dissociation (Kenny et al., 2009; Mclsaac & Eich, 2004; Williams & Moulds, 2008) or facilitating a “cool,” rational reinterpretation (Kross et al., 2005). The present experiments suggest strong caution against assuming that imagery perspective directly influences coping. The fact that imagery perspective had opposite effects on shame depending on self-esteem is particularly relevant in this regard: Shame has been implicated as an impediment to adaptive coping because it promotes withdrawal, aggression, and anger (Tangney, Wagner, & Gramzow, 1992). Third-person imagery decreased shame among HSEs but increased it among LSEs.

The present results do not rule out the possibility that third- person imagery could facilitate a “cool,” rational reinterpretation of negative events, even among those with less positive self-views, provided that instructions to use third-person imagery are accompanied by instructions to personally distance one’s self from a recalled event while analyzing it (as is the case of “self-distancing” manipulations; e.g., Kross & Ayduk, 2009). However, the present results demonstrate that simply leading people to picture an event from the third-person visual perspective, without any added interpretive frame, leads people to draw on their existing self-views to make meaning of the event. This can increase or decrease the negative impact of thinking about that event, depending on the nature of those self-views.

The present experiments demonstrated this pattern when individuals pictured failure, and we expect that future research will show that imagery perspective plays the same role in shaping responses to other types of negative events, such as trauma. Suggestive evidence comes from a study of intrusive memories, a symptom of post-traumatic stress disorder (Williams & Moulds, 2008). Results revealed that spontaneous use of third-person imagery was associated with maladaptive coping only for high dysphoric individuals–individuals who, like those with low selfesteem, tend to apply self-defeating interpretive frameworks to make meaning of life events (Beck, Rush, Shaw, & Emery, 1979; Lyubomirsky & Nolen-Hoeksema, 1995).

Given that third-person imagery promotes abstract construal of actions and events (Libby et al., 2009; Vasquez & Buehler, 2007), research on the role of abstraction in coping is relevant in considering the findings we report here. Rumination is a manner of abstractly analyzing negative events that has been shown to interfere with adaptive coping (e.g., Nolen-Hoeksema, Parker, & Larson, 1994). Abstraction manipulations are particularly harmful for people who tend chronically to engage in this style of thinking (e.g., Moberly & Watkins, 2006; Watkins, 2004). Given that the tendency to ruminate is likely associated with low self-esteem (Lyubomirsky, Tucker, Caldwell, & Berg, 1999), it is possible that a tendency to ruminate explains why third-person imagery was more harmful than first-person imagery for LSEs in our studies. This would suggest a mechanism by which the self-concepts of LSEs guide the processes they use to ascribe meaning to negative life events from the third-person perspective.

The link between perspective and abstraction also suggests implications of the present findings for assessing the role of abstraction in coping. Just as third-person imagery is unlikely to be uniformly helpful or hurtful for coping, abstraction should not be either. Rather, the effect should depend on the framework that guides abstraction. Thus, we predict that results suggesting overall positive effects of abstraction on coping (e.g., Updegraff, Emanuel, Suh, & Gallagher, 2010) may be driven by general tendencies for individuals to apply adaptive meaning-making strategies and that, if variation on this dimension were taken into account, an interaction should emerge. Alternatively, manipulations of abstraction may inadvertently specify a positive framework for abstraction, and positive effects of such manipulations (e.g.. Marigold, Holmes, & Ross, 2007) may depend on this feature (Kille, Eibach, Wood, & Holmes, 2011).


Mishaps are like knives, that either serve us or cut us, as we grasp them by the blade or the handle.

–James Russell Lowell

In the broadest sense, the present research speaks to the role of subjective construal in psychological processes (Asch, 1948; Bruner, 1957; Griffin & Ross, 1991). Like a knife that can be grasped by the blade or the handle, a failure can be thought about in ways that increase or decrease its negative impact. The visual perspective people use to picture failures determines the role of the self-concept in shaping reactions to those events. In this way, imagery perspective and self-esteem interact to guide one’s grasp of failures to determine whether they have the power to cut into well-being.


Guided Imagery

Jane Hart, M.D.

Guided imagery is a mind-body therapy that has been used for decades by individuals and in clinical settings to influence health outcomes. Guided imagery is particularly helpful for pain management and for reducing symptoms related to anxiety, stress, and other mental health conditions in which intruding thoughts play a role in the pathology.

Guided imagery can be an important adjunctive therapy to a conventional treatment approach for various conditions. In patients with chronic pain, for instance, medications and standards of care may not be enough to reduce or eliminate the pain significantly. Complementary therapies, such as guided imagery, when practiced regularly, can decrease pain further and reduce the need for pain medications. Guided imagery can be practiced by individuals on their own, which leads to increased feelings of self-mastery and control, and at very low or no cost to persons, using the therapy. In addition, another benefit is that patients can practice the technique in virtually any location. Individuals may use an imagery practice during stressful medical testing such as getting magnetic resonance imaging (MRI), as well as at home, in the operating waiting room, or on a subway.

Guided imagery has been practiced, written about, and researched for decades and has been used in religious, secular, and diverse cultural settings. From shamanism to psychotherapeutic settings, guided imagery has become a well-known complementary therapy. Increasingly, this imagery is utilized in clinical settings and is often encouraged and implemented by nurses. Blue Shield of California began offering guided imagery tapes to its presurgical patients due to increasing evidence that guided imagery may decrease surgical complications and reduce postoperative pain and pre- and postoperative anxiety.1

This article includes a description of guided imagery, a sampling of available research evidence on the topic, resources, and one woman’s experience with guided imagery.

Definition and Description

Guided imagery is a technique that utilizes stories or narratives to influence the images and patterns that the mind creates. Often, these stories or narratives are combined with background music. Merriam-Webster’s dictionary defines guided imagery as: “any of various techniques (as a series of verbal suggestions) used to guide another person or oneself in imagining sensations and especially in visualizing an image in the mind to bring about a desired physical response (as a reduction in stress, anxiety, or pain).”2 Basically, guided imagery is using the imagination to create images that bring about beneficial emotional and physical effects.

Noted guided imagery expert Belleruth Naparstek,

L.I.S.W., wrote that imagery is “any perception that comes through any of the senses including sight, sound, smells and feel.”3 She also wrote about recalling the smell of the air during a first snowfall and how the recollection of such imagery evokes specific sensations and memories in a person although that event may not be happening in the present moment. Ms. Naparstek wrote that imagery can seem as real as actual events and, thus, therein lies its power. Guided imagery can evoke neurohormonal changes in the body that mimic the changes that occur when an actual event occurs, according to Ms. Naparstek. She wrote that people can intentionally create healthful imagery that has beneficial physical and emotional effects in the body and mind.

Guided imagery may be delivered by a practitioner, a video, or an audio recording, or conducted by an individual. A typical guided-imagery session usually begins with relaxation in which the participant takes some deep breaths and releases tension in his or her mind and body. Then, the participant starts to visualize pleasant or effective imagery that may promote healing.

Boldly Going to Seek and Destroy Cancer Cells

One Woman’s Experience with Guided Imagery During Chemotherapy

Elizabeth McKinley, M.D., M.P.H., was first diagnosed with breast cancer when she was 36 years old. Juggling a busy career in breast-cancer research and internal medicine, along with raising children and going through treatments for her cancer, Dr. McKinley felt overwhelmed by her circumstances. She reached out to a number of different resources to help her get through this stressful time, including a support group at University Hospitals in Cleveland, Ohio, where the invited guest speaker was Belleruth Naparstek, L.I.S.W., a well-known leader in the field of guided imagery. During her presentation in Dr. McKinley’s group, Ms. Naparstek distributed tapes about the use of guided imagery during chemotherapy.

«I listened to the tape and found her voice soothing,» said Dr. McKinley. «There were affirmations on the other side of the tape with statements of fact like,’I’m feeling calm.’She walked you through how you might think about an image that helps you feel more in control with what is happening.» For example, Ms. Naparstek shared how a person might create imagery around cancer cells being destroyed and gave examples in her talk that some people needed images that portrayed an aggressive approach. Dr. McKinley felt it was important for her imagery to be individualized in order to work and to be»hers.»

«After listening to the tapes, I had an image of StarTrek/’said Dr. McKinley. «The captain Jean-Luc Picard is portrayed as an infallible character who I thought I could trust, and so I created an image of him on the bridge of the Starship Enterprise. In my mind, I had him send out shuttle craft to gobble up the cancer cells in my body and ‘seek and destroy’the cancer cells. Every time Picard sent a shuttlecraft out he would say:’Engage!’The idea is that I was confident that he wouldn’t miss anything and that my cancer cells would be destroyed.» Dr. McKinley stated that, over time, she was able to conjure up that image more easily, and this would greatly calm her down during and after chemotherapy.»! felt a little bit more in control when the scary medicine was snaking through my veins,» said Dr. McKinley.

Dr. McKinley noted that her chemotherapy medicine was, in reality, a flame-red color and elaborately covered in materials to protect other people from its toxicity–all of which scared her. Guided imagery helped her move past the frightening aspects of chemotherapy and radiation treatments, which lasted for 8 months. She said the imagery also provided humor. «I would even be laughing during my chemotherapy and people would ask what I was laughing at,» she said. «This imagery helped me keep a sense of humor during something so difficult and scary. My body relaxed and, after I focused on the image a few times, I really felt like it was happening and it was meditative. Sometimes I had to ask people to be quiet so that I could focus on my image.»

The theory behind the use of guided imagery is that if a person can imagine negative or frightening images that increase pain or anxiety, then those images may be counteracted with positive or calming images, and the mind can be habitually trained to focus on healing imagery more often. And, if frightening or negative imagery has the ability to increase pain and other unwanted symptoms, then positive or calming imagery may lessen pain and unwanted symptoms, according to advocates of this approach.

An important point, however, is that guided imagery may include positive thinking but should not be limited to that construct when describing this therapy. For instance, guided imagery may utilize aggressive thinking as in the case of a person who has cancer and visualizes cancer cells in his or her body being destroyed and killed, which can lead to increased feelings of power and control in that individual.4

In a review on the effects of guided imagery on outcomes, Van Kuiken described four types of guided imagery, which include pleasant imagery (imagining a calm place), physiologically focused imagery (focusing on the physiologic function that needs healing), mental rehearsal or reframing (imagining a specific task or performance before the event occurs or reframing a prior event), and receptive imagery (scanning the body to direct healing).5

Guided imagery may be most effective in group settings, with the use of music, and when the imagery matches a persons values and comfort level, according to Ms. Naparstek.3 She also wrote that guided imagery becomes increasingly effective with time and practice. She discriminates guided imagery from meditation and hypnosis, writing that the latter are broader categories under which guided imagery may apply. Hypnosis generally includes verbal suggestions without images, and meditation commonly focuses on one thing, such as the breath or a mantra. But guided imagery, according to Ms. Naparstek, for instance, can be considered a form of meditation.

Weydert and colleagues commented that guided imagery is different than hypnosis in that a person creates individualized imagery rather than a therapist giving a specific suggestion for change.6 They stated that “guided imagery allows for communication with the subliminal part of the mind to create change.”

The exact mechanisms by which guided imagery works are not well-known, but theories include the fact that relaxation and positive imagery attenuate psychoneuroimmunologic and hormonal pathways that affect the stress response.5,6 One study suggested that the mechanism may be through the Gate Control Theory, which proposes that “only one impulse can travel up the spinal cord to the brain at a time” and “if this pathway is occupied with other thoughts, then the sensations of pain cannot be sent to the brain, and therefore the pain is reduced.”7 Guided imagery may also release endorphins, which attenuate the pain response and may reduce pain or increase the pain threshold.’

Research on Guided Imagery

Evidence exists supporting the role of guided imagery for managing stress, anxiety, depression, side-effects related to chemotherapy, pain, and hypertension, preparation for medical procedures, stress during hospital stays, and other states and conditions, as well as for reducing length of hospital stays.8,9

Criticism about existing guided imagery research includes the fact that many studies have heterogenous populations, heterogenous interventions and outcomes measures across studies, low methodological quality, and lack of knowledge about optimum timing and dose of treatment and individual differences, such as outcome expectancy, the ability to create imagery, and perceived credibility of the guided-imagery provider.8,10

There is an abundance of research on the use of guided imagery for treating various conditions.


A patient who receives a diagnosis of cancer is faced with an array of physical and emotional symptoms. Typical emotional symptoms include fear of pain, suffering, treatments and side-effects, recurrence of disease, death, and other events. Guided imagery can be particularly useful in helping patients who have cancer cope with treatments such as chemotherapy and radiation. One study examined the effects of chemotherapy and/or radiation and found that anxiety levels increase significantly at the beginning of, and continue throughout, treatments in some individuals and that higher levels of anxiety are correlated with poorer quality of life.11 As a result, interventions are needed to help reduce unpleas ant symptoms in an already-difficult scenario.

One systematic review on using guided imagery during chemotherapy treatment found that, of six randomized trials, three showed that guided-imagery groups reported beneficial differences in anxiety, comfort, or emotional responses to chemotherapy compared with the responses in control groups.* 12

Another study explored the effect of guided imagery on comfort levels in patients with breast cancer who were beginning radiation therapy.13 Women who used guided imagery reported more comfort compared with the control group. The women in the treatment group listened to a guided-imagery audiotape once per day during the study. The researchers noted that the guided-imagery intervention increased comfort in patients with breast cancer, without an increase in personnel, costs, or time.

Admission to a hospital for pain relief is common among patients who have cancer. Researchers who studied the effects of guided imagery on cancer pain reported that the evidence for benefit is supportive but studies vary in reported outcomes from significant relief to no relief.14 In this study, 26 participants with cancer pain were interviewed about a study in which they were previously enrolled. This study included two trials of progressive muscle relaxation and two trials of guided imagery as an intervention for their pain. Approximately one half of the participants were responders to guided imagery, meaning that their pain scores were improved by 30% or more. Participants who practiced progressive muscle relaxation had similar response rates and scores for pain reduction.

The effectiveness of guided imagery is, in part, dependent on the observations of people who have found the therapy to be useful. Participants in this study[15] reported that the guided- imagery intervention was a source of distraction, provided uninterrupted quiet time, and stimulated relaxation; anesthetic images were helpful; and the intervention did not require active physical involvement. The participants also reported that the pacing of the imagery was important and that they needed enough time under the guidance of the person leading the imagery to create personal images. Certain imagery worked for certain individuals, and other imagery did not help. Some of the nonresponders in this study reported that the type or intensity of their pain prohibited them from being helped by the intervention and that the guided imagery, in fact, increased or reignited their pain or simply did not help.

In another study, 69 hospitalized patients with cancer-related pain used a 12-minute guided-imagery tape with analgesic imagery that offered suggestions to increase patient comfort and provided pleasant nature imagery, including walking along a river among wildflowers.10 Participants filled out pain questionnaires pre- and postintervention.The average pain-intensity score was lower than baseline for 90% of the participants and remained unchanged or increased for 10% of these patients. History of prior imaging use and imaging ability were significant predictors of outcome expectancy and pain outcomes such as pain intensity and control over pain.

Pain Syndromes

Guided imagery has been shown to help alleviate pain and reduce the need for pain medication for various conditions. Guided imagery can also help change the language a person uses to describe pain and the pain experience. One study found that, before using guided imagery, patients described pain in a variety of ways, including: never-ending, explainable, torment, restrictive, and changeable.[16] But after randomization into one of two groups–a guided-imagery group or a control group that was simply monitored–the treatment group reported pain as changeable, and the idea of pain as “never ending” did not reemerge in the treatment group but persisted in the control group.

One study evaluated the role of guided imagery in patients undergoing colorectal surgery.[17] Patients who participated in this trial were randomly assigned to receive standard perioperative care or standard care along with guided imagery. The imagery group listened to a guided-imagery tape 3 days preop- eratively; a music-only tape during induction, during surgery, and postoperatively in the recovery room; and a guided-imagery tape during each of the first 6 postoperative days. Guided imagery reduced pre- and postoperative anxiety, postoperative pain, and the need for opioid medications in treatment subjects, compared with controls.

Guided imagery can be particularly effective in children who tend to be more suggestible than adults and more open to their own creativity and imagination. One study examined the effects of guided imagery on recurrent abdominal pain from functional gastrointestinal (GI) disorders in children/’ In this study, 22 children, ages 5-18, were randomly assigned to breathing exercises alone or to guided imagery with progressive muscle relaxation, for 4 weekly sessions.The imagery group had an initial 1-hour appointment, and then each subsequent appointment was 20-30 minutes. In the first session, the children were guided in relaxation techniques and then asked to create images that represented their pain and then images that would relieve their pain. The children were then sent home with a tape of the guided imagery session. Subsequent appointments were for assessing competence and compliance only.6



At the 1-month follow-up, children in the guided-imagery group had significantly greater decreases in number of days with pain compared with the breathing exercises alone group. The children in the treatment group also had significantly greater decreases in days of missed activities due to pain. After 2 months of follow-up, more children met the desired outcome of 4 days or less per month of abdominal pain and no missed activities compared with children in the breathing exercises alone group.6

Another study evaluated the effects of guided imagery on pain, function, and self-efficacy in people with fibromyalgia.17 In this study, 48 people with fibromyalgia were randomly assigned to receive guided imagery plus usual care or usual care alone. The guided-imagery group listened to one guided-imagery tape a day for a total of 10 weeks. The participants used three tapes, which each lasted 20 minutes, and included relaxation, pleasant imagery, and imagery of practicing and performing behaviors in a way the patients would like to if they were free of pain. Functional status and self-efficacy for managing pain were significantly improved in the guided-imagery group, compared with the usual-care group. There were, however, no significant improvements or differences in pain between the two groups on pain scores.17

Interestingly, in Van Kuiken’s review, she noted another study conducted on fibromyalgia in which patients whose imagery focused on physiology and pain reported increases in pain, compared with patients whose imagery focused on attention-distracting or pleasant imagery, and who reported decreases in pain.5

Interstitial Cystitis

A new study has evaluated the effects of guided imagery on interstitial cystitis (IC), a painful bladder syndrome marked by pain, urgency, and frequency.7 Often, people with IC do not gain adequate relief with medications. Other studies have suggested that complementary therapies, such as biofeedback and hypnosis, have been helpful, but no studies had previously examined the role of guided imagery on IC. For this study, researchers enrolled 30 women with IC and randomly assigned them to either a guided-imagery group or to a control group. The guided-imagery group was instructed to listen to a 25-minute guided-imagery tape twice a day for 8 weeks. The tape emphasized healing the bladder, relaxing pelvic muscles, and calming nerves.The control group was instructed to rest by sitting or lying down for 25 minutes twice a day for 8 weeks. Both groups continued standard care for IC and kept pain and medication diaries.

The researchers found that 45.5% of the participants in the guided-imagery group reported moderate or marked improvement of IC symptoms, compared with 14.3% of women in the control group. Pain scores and urgency significantly decreased in the guided-imagery group as well. The guided-imagery group had significant reductions in pain scores from 5.5 at baseline to 2.57 at the end of the study, compared with the control group, in which pain levels did not achieve significant statistical improvement–4.89 at baseline and 4.39 at the end of the study.


Thoughts are powerful. Experts in guided imagery and research findings suggest that people have the power to change their thoughts to promote health. Patients can move from thoughts of fear and anxiety to thoughts of peace and wholeness. Guided imagery may be more than just positive thinking. The images people conjure up in their minds have lasting physiologic effects on their brains and bodies.

The challenge for clinicians regarding guided imagery and other effective complementary therapies is to remember to recommend and utilize such therapies as adjunctive tools to manage and treat patients’ conditions. Clinicians should inquire about the accessibility of learning such therapies within their own institutions.




Minding the Body …and Mending the Mind

by Joan Borysenko

For this assignment you must read pages 29 thru 46 and answer the questions. The rest you may choose to read on your own. As you read this make notes on the paper. You’ll get more from the text. Then go thru and do the questions. Use your own words.

1 – Two children on the beach

Compare the way that the two children dealt with the inevitability of the waves. How/why did one get frustrated and one cope. Why was the coping more healthy. Finally, identify a situation in your life where you could choose to approach it one of two ways – like the kids.

Part I – the kids

Part II – real life situation

2 – FEAR – Why does it ruin the moment? How does it tense the body?

Read what Borysenko thinks about the subject. Relate her thoughts and your response to those thoughts.


3 – NANCY – Relate her story? How did her fears affect her life?

4 – STEP ONE – Awareness

The author says, «We are often more tuned in to its commentary than we are to what’s actually happening, with the results that we miss the moment. We live in endless variations of old reruns of the mind.» Read around that statement. What is she actually talking about. What does she mean by quieting the mind? Provide an example where it applies to you.

Part II – real life situation

5- What is Meditation?

How does the author define meditation? Think of your life. Using the authors description tell me of a meditation that you already employ. Then tell me what you are like during these activities.

Part I – Definition


Part II – Activity & Explanation

6 – Learning to Meditate

Motivation, effort, determination are the keys to successful use of meditation. Explain each in regards to meditation in your own words.





What is a mantra? Why is a mantra useful? If you used one what might be a good one for you? Why?

Part II – Your Mantra

8 – TRY IT. What Happened?

Try it. Go thru each of the steps that the author suggests. Relate your experience here. What happened?

9 – AWARENESS – the key.

Through all of this awareness is the key? What does it mean…from the author’s perspective as well as your own?

10 – ST FRANCIS – thoughts about thoughts.

Read St Francis’ quote on page 45. What does it mean? How does it relate to meditation?


11 – Reread – And Try it!

Reread the instructions and try meditating for ten minutes. How did it go? Tell me.

12 – PRIMARY GOAL – It is NOT relaxation.

The primary goal of meditation is NOT relaxation. It is only a by-product. What exactly is the primary goal and how do you get there? Explain.

13 – Suggestion for the reader and a place to meditate.

Read the suggestions. Pay special attention to the part about a place to meditate. If you did meditate regularly – where might you do it. Whether or not you choose to do this meditation thing – describe a place that you would.

And read the fable about the Genie and blue bottle – it’s kind of cool.

So what do you think about all of this Meditation stuff.

Be honest – tell me here.


Cognitive Therapy and the Restructuring of Early Memories Through Guided Imagery

David J. A. Edwards

Rhodes University

This article describes the application of a guided imagery psychodrama technique to emotionally charged early memories. Such memories provide access to core schemata about the self and social relationships. Two case studies illustrate how the imagery technique enables the therapist to identify and restructure key cognitions out of which the schemata are constructed. The need for techniques to modify developmentally primitive schemata is discussed. Examples are given of ways to assist the patient in confronting the strong affect that may be aroused and in dealing with cognitions that block the process. Effectiveness is discussed in terms of the contribution of a guided imagery session to the overall process of “learning to learn” that takes place in cognitive psychotherapy.


The Development of Cognitive Therapy

Cognitive therapy as developed by Beck and his colleagues had recently become widely recognized as an effective treatment for depression, and it is also being developed for the treatment of anxiety disorders, (Beck & Emery, 1985; Beck, Rush, Shaw & Emery, 1979). The therapy takes its name from the centrality of cognition in Beck’s formulation of the theory of therapy. Essentially, he argues that people become depressed or anxious becauseof what they believe about themselves, the world, other people, and the future. These beliefs are often inaccurate, and the work of therapy is to identify the beliefs and systematically reevaluate them through logical analysis and reality testing.

The prominence Beck gives to cognition is in stark contrast to its fate in the two major traditions in therapy from which cognitive therapy derives, namely, the psychoanalytic and the behavioral. Early behavior therapy either excluded cognition from its models (Watson’s metaphysical behaviorism), allowed it only as a hypothetical construct (methodological behaviorism), or regarded it as a type of behavior but largely ignored it (Skinner’s radical behaviorism) (Zettle & Hayes, 1982).

In the Freudian tradition, cognition became the “Cinderella of psychoanalysis” (Arieti, 1985). For Freud’s followers, to deal with the cognitive was to take attention from the repressed, irrational unconscious processes of the id. Of Freud’s early colleagues, Adler had the most explicitly cognitive formulation, and today he is sometimes called the father of cognitive therapy (Freeman, 1983). But Freudians rejected his approach as too much involved with ego processes, so much so that Arieti (1985) accused them of systematically repressing cognition as ruthlessly as the rest of society repressed sexuality.

Various reformulations of the traditional Freudian position redressed the balance to some extent. Arieti (1985) explicitly based his approach, which he named Cognitive Psychoanalysis on a cognitive developmental model. Transactional analysis, as developed by Berne and his followers, is heavily cognitive, and an important feature is the identification and critical reevaluation of false beliefs or inappropriate rules (Barnes, 1977: James & Jongeward, 1971). Perls’s Gestalt therapy did the same but in a less explicit way (see Edwards, 1989). Objectrelations theory has developed a detailed cognitive model of the infant ’ s developing schemata for representing herself, others, and the relationships between them (Reppen, 1985). Adler and Homey both contributed to the development of the cognitive aspects of theory and were the major influences on the contemporary leading cognitive theorists Beck and Ellis (Freeman, 1983; Rendon, 1985; Shulman, 1985).

The term cognitive restructuring refers to the alteration of the beliefs, attitudes, and meanings that a person brings to the interpretation of experience. The process can be broken down into the following steps: (1) identification and labeling of emotion; (2) identification of the meaning of the emotion as either an automatic thought, an underlying belief, or a felt meaning; (3) tracing the historical process through which the belief was acquired; (4) testing the belief for rationality (is it logical), accuracy (does it fit the data?) or functionality (does it work for me?); (5) if appropriate, revising the belief; (6) changing behavior to accord with and operationalize the new belief.

Step 3, in which the historical process through which the belief is acquired is examined, is often not employed in short-term treatments. It has been widely used in transactional analysis (Barnes, 1977; Erskine & Zalcman, 1979; James & Jongeward, 1971) and has been recommended by recent writers on cognitive therapy (Beck & Emery, 1985, pp. 296-298; Guidano&Liotti, 1983; Young, 1984). For example, in the context of beliefs that create dependency, Emery (1982) writes:

Discover where your beliefs come from. … By going back and seeing where you adopted your beliefs you can often make them clearer to you