DOES GRIEF COUNSELING WORK?
JOHN R. JORDAN
The Family Loss Project, Sherborn, Massachusetts, USA
The University of Memphis, Memphis, Tennessee, USA
Most bereavement caregivers accept as a truism that their interventions are helpful. However; an examination of the bereavement intervention literature suggests that the scientific basis for accepting the efficacy of grief counseling may be quite weak. This article summarizes the findings of four recent qualitative and quantitative reviews of the bereavement intervention literature. It then discusses three possible explanations for these surprising findings and concludes with recommendations for both researchers and clinicians in thanatology that could help to focus efforts to answer the questions of when and for whom grief counseling is helpful.
It is a part of the assumptive world of most bereavement caregivers that their interventions work. Most practitioners believe that what they do is helpful and necessary, and that it does no harm. Moreover, most grief counselors would probably argue that most mourners would benefit from the services they offer. Understandably, caregivers in the helping professions want to believe that their work is valuable and their competence is recognized. These assumptions notwithstanding, several recent literature reviews and meta-analyses of bereavement intervention studies have suggested that these services may be surprisingly ineffective. Just as Wortman and Silver’s seminal article challenged the field to reconsider and expand its “mythology” about the normal trajectory of bereavement (Wortman & Silver, 1989; also see Wortman & Silver, 2001, for an update), there may be similar unquestioned assumptions about assisting the bereaved that require more careful scrutiny by caregivers. In part, this may reffect a significant gap that exists between researchers who study interventions for the bereaved, and the majority of caregivers who provide direct services to them (Jordan, 2000; Neimeyer, 2000; Silverman, 2000).
We have several goals for this article. First, we will review these challenging studies and summarize their conclusions for the reader. Then we will offer several possible interpretations of these findings and their implications for future research in thanatology and for caregivers who are on the “front lines” of providing support. Lastly, we hope to promote more relevant research and thoughtful clinical practice in the field of grief counseling. We hope that this article will be both provocative and stimulating, encouraging the reader to think more deeply about the implications of the research literature for their own work, and about the importance of bridging the gap that often divides those who study and those who practice.
Research on Interventions for the Bereaved
Recent years have seen a growing number of studies of the effectiveness of services for people seeking help with grief-related problems. A review of all bereavement-related intervention outcome studies is beyond the scope of this article. Instead, we will discuss and summarize several recent qualitative and quantitative reviews of the literature on the effectiveness of individual and group interventions for bereavement-related problems, primarily for adults. For those who may be unfamiliar with this topic, a qualitative (or narrative) literature review involves a scholarly analysis of the existing literature, from which the reviewer draws reasoned conclusions about the state of knowledge in a given area of inquiry. In contrast, a quantitative review (or meta-analysis) is a statistical technique (or group of techniques) that allows objective data from many different studies to be combined to produce a numerical answer to the question, “How effective is a particular type of treatment?” (Hedges & Olkin, 1985). The resulting measure, commonly referred to as the effect size of a treatment, is a standardized way of assessing to what degree people who receive the treatment do better on outcome measures than people in control groups who do not receive the treatment (Kazdin, 1998; Wampold, 2001). The general effect size for many psychotherapeutic treatments for a broad range of problems is on the order of .8 (Wampold, 2001), which is commonly considered a large effect in the social sciences (Cohen, 1988). An effect size of .8 indicates that approximately 79% of people who receive the treatment are better off after treatment than those who do not receive treatment. This is the rough standard against which we might compare the effectiveness of grief counseling. Accordingly, we will summarize below four recent major literature reviews and/or meta-analyses of grief counseling.
Meta-analyses and Literature Reviews
Allumbaugh and Hoyt (1999) reviewed 35 bereavement intervention studies that included both single-group (pre- and posttest) and two- group (treatment and control) designs. Thus, they did not limit the analysis to studies that included random assignment to treatment/control conditions. Rather, the authors used a method of meta-analysis that corrected for the pre-treatment status of the groups on such variables as levels of distress, and allowing for inclusion of single-group designs, which typically yield larger effect sizes than controlled studies. Overall, Allumbaugh and Hoyt found an effect size for bereavement interventions of .43. They suggested that this relatively small effect may be due to a general ineffectiveness of grief counseling, to the low statistical power of many of the studies, or to one or more intervening variables that masked real effects of the interventions. The authors examined 12 of these potential moderator variables, including categorical variables such as characteristics of the treatment (e.g, level of practitioner training) and treatment modality (e.g., group vs. individual; number of sessions), as well as client characteristics, such as age, gender, time since loss, and relationship to the deceased. They concluded that more highly trained practitioners produced a better result (particularly when compared to non-professional therapists), and individual therapy produced better results than group treatment. However, these two variables were confounded because studies using individual treatment also tended to use professionally trained therapists rather than paraprofessional volunteers.
Allumbaugh and Hoyt (1999) also found that more effective interventions included a greater number of sessions and began closer to the time of death of the loved one, although the mean length of time since death across the studies was over 2 years. The authors noted that this is a relatively long time after death for an intervention to begin, and may account for the failure to find expected improvement over time in the control groups, because much of the change in functioning may already have been accomplished by two years. Importantly, Allumbaugh and Hoyt also noted that studies using clients who were self-identified and specifically seeking help for their bereavement had much larger effect sizes than studies where participants were recruited by the investigators. Despite this, they found only a marginally significant trend for clients defined as high risk to benefit more from the interventions. They speculated that this unexpected finding might have been a result of the great inconsistency in definitions of high risk from one study to another.
Kato and Mann (1999) provided a combined qualitative and quantitative review of bereavement intervention studies. They used selection criteria that required random assignment to treatment and control groups, similar recruitment procedures for both groups, and initiation of the intervention only after the loss had occurred. Using this stricter set of criteria, the authors reviewed 13 articles, breaking the sample into studies that used individual, family, or group interventions. They noted that three of the four studies using individual therapy interventions produced only slight changes in physical health, and one found improvement in stress reactions of the participants. Kato and Mann further concluded that one family therapy study and six of the eight group studies reviewed found almost no beneficial effects of the interventions. Using meta-analytic computations for combined data from the thirteen studies, the authors reported an overall effect size of .052, .272, and .095 for the reduction of depressive symptoms, somatic symptoms, and all other psychological symptoms, respectively. They also found a global average effect size across all types of outcome variables of .114 and concluded that, “The effect sizes for these studies suggest that psychological interventions for bereavement are not effective interventions” (p. 293).
Kato and Mann (1999) offered three possible explanations for these findings. These included the possibility that psychological interventions for the bereaved are simply not helpful, that they are not powerful enough as delivered in the studies (e.g., too few sessions), and that the positive effects of the interventions have been masked by methodological problems such as small sample sizes, unreliable and invalid measures, and high drop-out rates. Importantly, Kato and Mann also noted that most of the studies failed to analyze the data separately by gender and by expectedness of the loss, two factors that have sometimes been shown to differentially affect bereavement outcome. Finally, the authors noted that, in the majority of studies, participants in both the treatment and control groups tended to improve. This supports the widely recognized observation that most bereavement is self-limiting without formal intervention (Stroebe, Hansson, Stroebe, & Schut, 2001). This, in turn, would tend to decrease the contrast between control and treatment groups in these bereavement intervention studies. Indeed, it is this tendency for many bereaved individuals to improve with or without professional intervention that may account for the larger apparent effects reported by Allumbaugh and Hoyt (1999), who included studies simply reporting pre–post changes without ensuring that such improvement exceeded that of untreated controls.
Neimeyer (2000) summarized an unpublished meta-analytic study of bereavement interventions performed by Fortner and Neimeyer. The authors searched the literature from 1975 through 1998, including in their research only studies that met acceptable scientific criteria of random assignment to treatment and control groups. They included studies involving both children and adults across all types of losses. To assess the effectiveness of the interventions, they used both a more traditional measure of effect size, Cohen’s d (Cohen, 1988), and a more novel measure of “treatment induced deterioration” (p. 544). The latter indicated the percentage of people who, on a statistical basis, would have been better off in the no-treatment condition. They found an overall effect size of .13 across the sample of 23 studies, which closely corresponded to Kato and Mann’s (1999) findings for a smaller sample of studies. They also found that approximately 38% of participants would have had a better outcome had they been assigned to the control, rather than the treatment condition. This contrasts with most psychotherapy outcome studies, which showed an average deterioration of only 5% (Neimeyer, 2000). In an attempt to further understand these findings, the authors investigated the differential effectiveness of variables such as length of therapy, credentials of the therapist (professional vs. non-professional), modality of treatment (individual vs. group), and theoretical approach used by the therapist. None of these variables was correlated with effect size. However, Neimeyer and Fortner did find that a greater length of time since the death, younger age of the participant, and higher levels of risk (sudden violent death or evidence of chronic grief) were related to increased effect size for the interventions.
Schut, Stroebe, van den Bout, and Terheggen (2001) have published the most recent review of bereavement intervention research. These authors offered a general methodological critique of existing research, noting such common problems as a lack of control groups and random assignment of participants, low initial participation rates, and high levels of non-adherence to and attrition from intervention protocols. They also questioned whether there is a sufficient number of methodologically comparable and rigorous studies to use meta-analytic techniques at this point in time. Accordingly, Schut and his associates limited their review to a qualitative summary of three categories of studies: primary, secondary, and tertiary interventions after loss.
With regard to interventions designed to prevent the development of problems in the general population of bereaved persons (primary prevention), Schut et al. (2001) reviewed 16 studies involving either children and/or adults. They concluded that “primary preventive interventions receive hardly any empirical support for their effectiveness” (p. 720), although they noted that there is somewhat more support for the efficacy of interventions with children than with adults.
The authors also evaluated seven studies that focused on bereaved persons who were defined as being at high risk for developing bereavement related-problems (secondary prevention). These included populations who had experienced the sudden, traumatic death of a loved one, those who were in a high-risk category (e.g., bereaved parents), and those who showed high levels of symptomatic distress on pre-intervention measures or clinical assessment. Schut and his colleagues concluded that although there is more evidence of intervention efficacy for this population, the effects are still quite modest in comparison to traditional psychotherapy outcome studies. They also emphasized the importance of doing gender specific analyses, because several of the studies showed differential effectiveness of the interventions for men and women. Importantly, they also found that studies that specifically screened for high levels of distress (rather than simply selecting on the basis of membership in a high-risk category, such as bereaved mothers) tended to show better results for the intervention.
Finally, Schut and his colleagues examined interventions for people who had already developed a complicated mourning response (tertiary prevention). This group included seven studies in which the participants were already suffering from clinical levels of depression, anxiety, and other bereavement-induced disorders at the time of entry into the study.
Participants also tended to be self-referred for help (as opposed to being recruited to participate), and the interventions were typically delivered a longer time after the death than in secondary and primary prevention interventions. Despite some methodological limitations of the research, Schut’s group found that this type of tertiary intervention was generally successful, as indicated by reductions in levels of psychiatric symptoms and grief-related distress when compared with control group participants.
Schut et al. (2001) concluded their review by stating that “the general pattern emerging from this review is that the more complicated the grief process appears to be, the better the chances of interventions leading to positive results” (p. 731). They suggest that this finding may be a result of the fact that many primary and secondary interventions involve outreach recruitment of participants, rather than help seeking on the part of the bereaved. This may result in participants in tertiary studies who are both more distressed (and therefore have greater room to improve) and are more motivated to accept and make good use of psychotherapeutic help. In addition, Schut and his group noted that the most successful interventions tended to be delivered later in the bereavement process. It is possible that this apparent time effect also reffects the presence of greater complication, insofar as the most recent criteria for the diagnosis of complicated grief require marked symptomatology that persists for six months or more following loss (Neimeyer, Prigerson & Davies, 2002; Prigerson & Jacobs, 2001).
Tentative Conclusions about Bereavement Interventions
What conclusions can be drawn about the effectiveness of bereavement related interventions from this brief summary of recent qualitative and meta-analytic reviews of the literature? In general, it appears that the scientifically demonstrated efficacy of formal interventions for the bereaved is distressingly low, far below that of most other types of psychotherapeutic interventions. This is a conclusion that runs counter to the professional experience of many clinicians in the field, including us. Upsetting as they may be, however, these findings deserve careful consideration by practitioners in the field, lest we miss a valuable opportunity to improve the practice of grief counseling. We believe the data provide important information for bereavement researchers and practitioners alike, and, in that spirit, we would like to offer several possible interpretations of the studies and then conclude with some recommendations for future research and the practice of bereavement care. More specifically, we would like to offer three possible and interrelated explanations for the general findings that grief counseling has very low efficacy: grief counseling may not be needed by most mourners; grief counseling may not work in the form that it is typically delivered in research studies, and the positive effects of grief counseling may be masked by methodological issues in the design and implementation of the studies.
Formal Intervention May Not Be Needed Much of the Time
One likely explanation for the failure to find significant effects is that most uncomplicated grief is probably naturally self-limiting. Most longitudinal studies of bereavement show naturally occurring declines in bereavement symptoms (Ott & Lueger, 2002; Raphael, Minkov, & Dobson, 2001; Stroebe et al., 2001). With the help of family and friends, apparently most mourners are able to work through and integrate their losses relatively well. This self-healing trend is also evidenced by the fact that control participants in many studies tend to improve without any intervention (a fact that would tend to wash out differences between control and treatment groups). Likewise, Schut et al. (2001), and Allum- baugh and Hoyt (1999) noted that studies showing the greatest efficacy involved designs where mourners sought help for self-identified bereavement-related distress, rather than being recruited into the study simply because they were bereaved. This suggests that most mourners who seek out professional care in non-research settings may be more distressed and not following the more expected trajectory of “healing with time.” It is also plausible that persons who seek out grief counseling on their own are more motivated to deal with the loss than the general population of mourners, although we are aware of no research studies that have specifically examined that proposition.
Neimeyer’s (2000) calculation that 38% of participants would have done better had they not received the intervention also raises the sobering possibility that some services may be actively detrimental for some mourners. As an example, Murphy and her colleagues (Murphy et al., 1998) found that men who participated in a carefully designed group intervention for parents bereaved by the violent death of their children actually worsened on symptoms of PTSD. Moreover, this unexpected outcome occurred even though the group was based on reasonable clinical principles that balanced supportive discussion with coping skills training in each session. Likewise, Farberow (1992) found that participants in a suicide survivor support group worsened on certain types of emotional responses after participation. Researchers and clinicians alike should be very alert to the possibility that for some, interventions may do more harm than good (Jordan, 2000). It may be that people with more avoidant (Bonanno, 1999) and/or instrumental (Martin & Doka, 1999) coping styles are more likely to respond poorly to traditional bereavement interventions that emphasize emotion-focused, rather than problem-focused, coping styles. This is buttressed by studies showing that men may use and respond to bereavement support services differently than women (Mastrogianis & Lumley, 2002).
Of course, by observing that most mourners may not need formal bereavement intervention, we are not suggesting that all people who do not seek services will necessarily do well after the death of a loved one. On the contrary, one of the most important trends in these reviews is the recognition that there are subgroups of mourners who are at elevated risk for dysfunction and who respond well to formal intervention. For example, the research associated with the proposed new diagnostic category of traumatic or complicated grief (Prigerson & Jacobs, 2001) strongly suggests that this disorder is neither self-limiting nor benign in its mental health and medical consequences (Ott, 2003). Moreover, we simply do not know what percentage of at-risk and high-distress mourners actually seek treatment. There may be a group of bereaved persons who are not likely to improve simply with time and informal support, yet who are also unwilling or unable to access professional services. This “hermit” group of mourners is of particular concern, and may require alternative service modalities, including extra outreach (please see recommendations below).
Grief Counseling as Delivered, in Research Studies May Be Ineffective
Most studies represented in the reviews offered 8–12 sessions of intervention, whether group or individual. These sessions were generally held on a weekly basis, meaning that on average the participants had contact for about three months with the caregivers/researchers. It is possible that the dosage (how many sessions) and timing (when they were delivered) of treatment were simply too “weak” to produce measurable effects. Several studies have suggested that for some people, mourning may be a longer process than commonly believed, particularly after certain types of traumatic loss (Lehman, Wortman, &Williams, 1987; Murphy, Johnson, & Lohan, 2003). There is evidence that certain aspects of the experience may worsen rather than improve in the second and third years (Wortman, 2001). Thus, a research intervention consisting of only a few sessions may be considerably below the “therapeutic dosage” level needed to produce a measurable effect.
Likewise, the timing of interventions in the bereavement trajectory may be important, with many interventions being delivered too early or too late in the grieving process to be effective. Allumbaugh and Hoyt (1999) found that interventions delivered closer in time to the loss seemed to produce greater effect sizes. However, their mean time since the death across all reviewed studies was two years, suggesting that “earlier” after the loss may still have been a relatively long time post-death when compared with programs that begin shortly after the loss. The Kato and Mann (1999) review, which primarily had duration times of 3–12 months post-loss (and tighter methodological selection criteria for inclusion of studies), did not find significant effects for time since the loss. In addition, Neimeyer (2000) actually found that interventions that occurred sooner after the death had significantly smaller effect sizes. Similarly, in their qualitative review, Schut et al. (2001) reached a similar conclusion that interventions offered too soon in the mourning process may be less effective, or even counter productive. It is difficult to reconcile these somewhat contradictory findings, unless we consider the possibility that the optimal timing of bereavement interventions may be curvilinear. That is, there may be a critical window of time, neither too soon nor too long after a loss, when mourners are most responsive to and able to use formal support services. One possibility is that services may be most effective when delivered in a 6–18 month period following the loss. This may be the time when complicated grief is both diagnosable and prognostic of later difficulties (Ott, 2003; Prigerson & Jacobs, 2001), but before problematic patterns of adjustment have become entrenched or triggered “empathic failure” in the mourner’s family or community (Neimeyer & Jordan, 2002). This speculation obviously requires further empirical validation before clinical practices are adjusted to reffect this concept.
A third possibility is that the types of support needed early in the mourning trajectory may be dilferent from those required later in the process. For example, young widows and widowers who are early in their grief may need support with grieving and the adjustment to single parenting One to two years after the loss, however, their focus may shift toward finding a new partner and establishing a blended family, concerns that are substantially different from those of early grievers. In any study that mixes mourners who are at different points in their mourning process, a generic intervention may be effective for only some of the participants, making detection of significant effects more difficult. We are aware of no study that has attempted to customize the type of intervention to particular points in the bereavement trajectory, although this would be a valuable contribution.
Finally, the membership composition of the group therapies used in most intervention studies could also limit their effectiveness. Greater identification and support among group members (especially early in the group) has been associated empirically with more favorable outcome (Geron, Ginzburg, & Solomon, 2003; Yalom, 1995), suggesting that clients who do not identify with the other group members may be at risk for negative effects (Neimeyer, Harter & Alexander, 1991). Because support and therapy groups provided to the bereaved rarely prescreen prospective members to ensure their homogeneity on relevant variables (e.g., ethnicity, gender, or level of distress), it is likely that some members might perceive themselves as “outliers,” at risk of feeling alienated from the group. Likewise, the presence of people in the group with markedly different levels of difficulty (e.g., including one or two persons with extremely traumatic losses or complicated grief reactions in a group of mourners with more normative experiences) could lower group cohesion, lead to a preoccupation with (or resentment of) needy members, or frighten more typical grievers with the prospect of their own deterioration. It is even possible that the apparently greater efficacy of interventions for at-risk mourners reffects their greater homogeneity and its implications for greater between-member identification and supportive sense of universality in group interventions, rather than their direct responsiveness to treatment because of their distress. (However, this would not explain the apparently greater effectiveness of treatments employing individual modalities for at risk individuals.) This possibility could be more carefully evaluated in future studies by examining outcomes for complicated grievers in heterogeneous versus homogeneous groups or could be controlled by ensuring sufficient commonality of membership through advance screening of members to minimize the potential problem of group outliers.
To summarize, the duration, timing, precision, and group composition of services in typical outcome studies may be out of synchrony with some mourners participating in the research, thus minimizing the overall positive impact of the interventions and making it appear that grief counseling has little efficacy In fact, interventions of greater “strength” or more customized design might well demonstrate greater effects.
Methodological Problems May Mask True Benefits
Although recent studies have shown improved methodological sophistication, the general quality of research on bereavement interventions leaves much to be desired. The designs of many studies suffer from the common problems found in many psychotherapy intervention studies (see Schut et al., 2001, for a review). These include the lack of appropriate control groups and random assignment to treatment conditions, small sample size (and consequent low statistical power), and lack of a theoretical foundation for and adherence to a clear treatment protocol. An additional problem is the failure to choose psychometrically sound outcome measures that pertain to the grieving process, despite the recent development of such measures (Neimeyer & Hogan, 2001). Most studies have used narrowly defined and simplistic criteria to assess the impact of the intervention, such as psychiatric symptom checklists or global measures of functioning. Many models of bereavement suggest that simplistic formulations of recovery are likely to be of little help in conceptualizing or measuring outcomes in bereavement. For example, Rubin and Malkinson (2001) have proposed a two-track model of bereavement outcome that examines both the return to functioning of the mourner and the ongoing and evolving internal relationship to the deceased (an issue unique to loss-related phenomena). It is possible that many bereavement intervention studies fail to find an effect because they are measuring the wrong outcomes and need to focus on variables that are specific to bereavement.
On a related point, the apparently disappointing outcome that characterizes existing intervention studies could derive in part from their orientation to the pathogenic rather than the salutogenic consequences of profound loss. That is, by focusing preemptively on outcomes defined in terms of depression, anxiety, or grief symptomatology, investigators may fail to observe equally important positive consequences of bereavement that have been well documented (Frantz, Farrell, & Trolley, 2001; Neimeyer, Prigerson, & Davies, 2002). As contemporary theories of posttraumatic growth (Tedeschi, Park, & Calhoun, 1998) and meaning reconstruction (Neimeyer, 2001) emphasize, personal development defined in terms of enhanced empathy for the suffering of others, a revised sense of life priorities, and deepened existential or spiritual attu- nement seems to occur because of residual pain and distress rather than only when it is ameliorated. For this reason, measures that assess meaningful growth as well as bereavement symptomatology (e.g, Hogan, Greenfield & Schmidt, 2001) are worth considering in future studies, as well as in clinical practice.
Finally many studies have failed to investigate a number of interactional or moderator variables that may be masking the positive benefits derived by subsets of participants in intervention studies. A number of these intervening variables have been previously mentioned, including differential responses to an intervention based on gender, risk status, circumstances of the death, and time since the death. The reviews cited in this article suggest that women, high-risk mourners, sudden traumatic loss survivors, and mourners further along in the grieving process generally respond better to the studied interventions than men, low-risk mourners, and expected death and recent loss survivors. Investigations that fail to analyze data for these groups separately are more likely to have “washouts,” wherein real intervention effects for some groups are masked by the apparent lack of impact for the entire subject pool. These are precisely the results that Murphy et al. (1998) and Murray, Terry, Vance, Battistutta, and Connolly (2000) obtained in their carefully designed intervention studies.
Recommendations for the Future
We believe there are trends noted in these four reviews that have important implications for the conduct of grief counseling. Hence, we would like to conclude this article with a discussion of some of the implications for both researchers and clinicians in thanatology and to offer a set of specific recommendations for future research and practice.
There is a continuing need for improvement in the quality of research designs in grief counseling interventions. Although recent reports show an awareness of the principles of good research design (Murphy et al., 1998; Murray et al., 2000; Schut, Stroebe, van den Bout, & de Keijser, 1997), the majority of studies still demonstrate poor internal validity, making it difficult to draw confident conclusions from the research. In addition to the basic principles common to the design of all intervention studies (Kazdin, 1998), there are some specific issues that need improvement in bereavement intervention studies. As previously discussed, a broad range of outcome measures needs to be used not just symptom- oriented tools. These measures should be sensitive to changes in domains that are likely to be pathognomic indicators specific to bereavement, such as yearning for the deceased, feelings of guilt and remorse, shattering of the mourner’s assumptive world, and so on, that are unremitting, long term in duration, and disruptive of functioning. As previously mentioned, we also make a plea for the use of measures that assess the positive outcomes that may accrue to mourners after the loss of a loved one.
Likewise, data need to be analyzed taking into account the moderator variables that appear to be influential in determining the effectiveness of an intervention. At a minimum, these include gender, time since the death, risk status of the mourner (including both membership in a high-risk category and elevated distress levels at the start of the intervention), and pre-existing personality and functioning of the mourner. We also strongly support the incorporation of mixed qualitative and quantitative designs in bereavement intervention research (Jordan, 2000; Neimeyer & Hogan, 2001). Well-conceived qualitative data can tap into many of the bereavement specific issues that are frequently missed in the use of standardized self-report measures. Studies that use qualitative methods to explore which aspects of interventions are most helpful to the bereaved would be particularly useful in designing and improving interventions.
Secondly, we would encourage bereavement intervention researchers to concentrate their intervention studies on high-risk mourners. We believe there is now sufficient evidence to conclude that generic interventions, targeted toward the general population of the bereaved, are likely to be unnecessary and largely unproductive. Instead, interventions that are tailored to the problems of mourners in high-risk categories (e.g., bereaved mothers, suicide survivors, etc.), or showing unremitting or increasing levels of distress after a reasonable period of time are likely to be more beneficial. Jacobs and Prigerson (2000) have offered a review of interventions that show promise in treating individuals meeting criteria for the proposed disorder now being called complicated grief. Along with this, there is a great need for researchers to focus on the development of screening measures that have good predictive validity for identifying persons at risk for complicated mourning responses. New measures such as the Inventory of Complicated Grief (Prigerson & Jacobs, 2001) and the Grief Evaluation Measure (Jordan, Baker, Rosenthal, Matteis, & Ware, 2003) show promise in this regard.
Thirdly, we believe there is a need for researchers to spell out in more detail the theoretical foundation for and the operational implementation of their interventions. Most reports in the literature provide only a cursory description of how the treatment procedures were developed, how they were linked to previous research in the field, and how they were delivered during the intervention. The recent work of Murphy (Murphy et. al., 1998) and Sandler (Sandler et al., in press) and their associates are good examples of research interventions that have a solid theoretical foundation and a detailed accounting (through an available manual) of the procedures used in the intervention. This attention to detail in future studies will facilitate the accumulation of knowledge and replication of interventions for both research and clinical purposes.
Fourth, at the risk of contradicting ourselves with regard to the call for more sophisticated research designs, we would also like to suggest more effectiveness-oriented, as well as efficacy-focused, research within bereavement care programs. Although randomized, controlled studies of treatment interventions are still considered the “gold standard” of research design, we believe that there is much to be learned from studying the effectiveness of common interventions as actually delivered in “real world” clinical settings (Borkovec & Castonguay, 1998; Seligman, 1995). As noted previously, there is a distinct possibility that most research-based interventions are too weak and poorly timed to show efficacy, whereas many interventions delivered in clinical settings might prove more effective. Far example, many agencies and organizations offer bereavement support groups with varying structures and leadership formats, from open-ended, peer-facilitated groups such as the Compassionate Friends to time-limited psychoeducational groups run by professionals. While less credible from a strict validity standpoint, a large-scale study focused on the effectiveness of these groups (perhaps when compared with a sample of mourners who do not attend groups) would offer a wealth of information about the relative usefulness of different styles of group interventions for different types of mourners and loss situations.
Lastly, we make a plea for bereavement intervention researchers to pay close attention to the trends in the very large body of studies on psychotherapy outcome research (Nathan & Gorman, 2002; Wampold, 2001). Although there is still debate about this point, there is convincing evidence that the non-specific, relational, and contextual aspects of psychotherapy are probably the most important “active ingredient” in the treatment process, rather than specific techniques or procedures (see Deegear & Lawson, 2003, and Wampold, 2001, for recent reviews). In the current era of managed care, the prevailing model of psychotherapy is based on an analogy with medical treatment, wherein an expert diagnoses and then treats a relatively passive patient for specific disease or injury. We believe this model is likely to have only limited use when thinking about the provision of assistance for bereaved individuals, even those with specific patterns of dysfunctional bereavement response such as complicated grief (Prigerson & Jacobs, 2001). Wofelt (2003) proposes instead that we conceptualize this work as a form of “companioning.” We feel that this term is compatible with a conceptualization of grief counseling as a specialized and concentrated form of skillful social support, rather than the administration of a highly technical and diagnosis-specific medical procedure. This suggestion is compatible with studies indicating that the relational aspects of any psychotherapeutic encounter provide a context for the inspiration of hope and the learning of new coping skills that is crucial to all successful psychosocial interventions (Frank & Frank, 1991).
Note that we are not suggesting that there is no role for the use of well- defined techniques in grief counseling (such as eye movement desensitization and reprocessing (EMDR), journaling, guided visualization, etc.). Rather, we believe that the common and probably most important factor in all bereavement interventions is the encounter with compassionate and empathically attuned caregivers who provide mourners with a healing experience of being understood and supported in their journey of loss. If this assumption is correct, then bereavement intervention researchers should build on the large body of studies from psychotherapy outcome research that identify crucial aspects of the therapeutic relationship (Norcross, 2002). In a parallel fashion, we encourage than- atology researchers to begin to delineate those aspects of the therapeutic relationship in grief counseling (whether individual or group) that provide this crucial interpersonal matrix for healing after a loss. We are unaware of any studies in grief counseling that have examined this important dimension, and we strongly encourage the field to begin looking at this aspect of intervention.
Our suggestions for clinicians generally mirror the comments for researchers, with perhaps a more circumscribed strength of recommendation, because we believe that recommendations for changes in clinical practice require a higher degree of empirical validation than simple suggestions for promising areas of further research. Perhaps the first request is that grief counselors adopt a more critical attitude toward their methods. On the basis of the trends observed in these summaries, we feel that it should no longer be taken for granted that grief counseling is necessary and necessarily helpful for all or most mourners. Indeed, perhaps the central finding of these reviews is that grief counseling does not appear to be very effective, most probably because many of the people who receive it would do just as well (and perhaps in some cases better) without it. The assumption that grief counseling is “naturally” beneficial for everyone fails to recognize the possibility of harmful effects of bereavement interventions for some individuals. This readiness to encourage all individuals to receive treatment needs to be replaced with an effort to customize interventions to the particular gender, personality, background, resources, and perceived needs of individual mourners. Although the empirical basis for this customization is in its infancy, we encourage clinicians to work toward this goal in their daily practice. Again, there is a growing body of research from the general psychotherapy outcome literature that offers empirically based guidelines for the customization of intervention modalities and styles that may serve as a useful starting point for grief counselors (Beutler, 2000). For example, clients who view their problems in terms of distinct symptomatology have been shown to respond well to behavioral or cognitive therapy, whereas those who view their difficulties in more psychological terms seem to favor insight-oriented group therapy (Winter, 1990).
A corollary of this is that clinicians should not assume that “one size fits all” in terms of the types of interventions required by mourners. It is a truism that grief is unique to each individual (Neimeyer, Keesee, & Fortner, 2000), yet this wisdom is rarely reflected in the design and delivery of services for the bereaved. Although some interventions may be customized to the extent that they target specific kinship losses (e.g., death of a child) or cause of death (e.g., homicide survivors), it is rare that the content or process of an intervention is tailored to address the specific problems of the targeted group, although steps are being made in this direction (Rynearson, 2001). Moreover, our impression is that there is typically very little formal assessment done with persons seeking bereavement support. More attention paid to the personality structure, previous loss, trauma and psychiatric history, coping style, and support resources available would help clinicians make more informed judgments about the types of services that are likely to be beneficial for a given individual.
The reviews cited in this article offer a rather confusing picture for the clinician seeking guidance about the timing of interventions, although there seems to be some evidence that services delivered later in the bereavement trajectory are more effective. This finding may be related to the finding that higher risk mourners also respond better to interventions. Schut et al. (2001) noted that the studies classified as tertiary interventions (which by definition presume the development of a diag- nosable disorder) showed the greatest efficacy for their treatments. These authors pointed out that it might take some time for a mourner to develop full-blown complications after a loss, which might account for the observation that later interventions appeared to be more powerful. As we have previously observed, another possibility is that the relationship between the timing of the intervention and efficacy is actually curvilinear, suggesting that providing assistance too early or too late in the bereavement trajectory may reduce its effectiveness. Our own clinical experience suggests that the optimal time period may be somewhere between 6 to 18 months after the death, but this is an issue that must be empirically evaluated. In the meantime, clinicians would probably do well to have early contact with newly bereaved clients to establish a relationship and provide psychoeducation, but they should be cautious about pushing mourners into treatment too early in the process. Likewise, the research suggests that arbitrary cutoffs of one year for bereavement care, as found in many hospice programs, may be poorly thought out. Rather, a program of less frequent but longer-term contact with bereavement caregivers (e.g., 12 sessions beginning at 4–6 months post-loss and spread over 12 months rather than 12 weeks) may be more effective.
Just as we suggested to researchers, we also believe that clinicians and program administrators should concentrate their efforts on identifying and engaging high-risk mourners. On the basis of the available literature (Parkes, 2002; Stroebe & Schut, 2001), this would include men who lose spouses (particularly older and isolated males), mothers who lose children, and survivors of sudden and/or violent traumatizing losses, such as suicide, terrorist attacks, warfare, homicide, and accidental death. Likewise, individuals with previous psychiatric histories (including depression, substance abuse, post-traumatic stress disorder, and psychotic disorders), low self-esteem or coping self-efficacy, high levels of dependency on the deceased, and abuse/trauma histories are likely at elevated risk. Finally, individuals manifesting high-distress grief (e.g., high levels of depressive, anxiety, anger, or rumination symptoms, or who meet diagnostic criteria for complicated grief) early in their bereavement experience are also likely to benefit more from intervention. Bereavement support programs, hospices, and nursing homes that have contact with individuals facing a loss are in an excellent position to identify and perform outreach to these high-risk categories of mourners.
We hope that this “review of reviews” proves to be both provocative and stimulating to our colleagues in thanatological research and practice. With the incorporation of research findings into the practice of grief counseling, and the thoughtful researching of new interventions and services developed in the clinical setting, we envision an improved and more exacting answer to the question “Does grief counseling work?” in the years to come.
SEARCHING FOR THE MEANING OF MEANING: GRIEF THERAPY AND THE PROCESS OF RECONSTRUCTION
ROBERT A. NEIMEYER
University of Memphis, Memphis, Tennessee, USA
А comprehensive quantitative review of published randomized controlled outcome studies of grief counseling and therapy suggests that such interventions are typically ineffective, and perhaps even deleterious, at least for persons experiencing a normal bereavement. On the other hand, there is some evidence that grief therapy is more beneficial and safer for those who have been traumatically bereaved. Beginning with this sobering appraisal, this article considers the findings of C. G. Davis, С. B. Wortman, D. R. Lehman, and R. C. Silver (this issue) and their implications for a meaning reconstruction approach to grief therapy, arguing that an expanded conception of meaning is necessary to provide a stronger basis for clinical intervention.
What role does the “search for meaning” play in the struggle of bereaved persons to adapt to loss, and how might professional therapy assist with this effort when indigenous sources of support fail? In taking up these thorny questions, Davis, Wortman, Lehman, and Silver (this issue) posed a provocative challenge to the sometimes glib assumptions of grief counselors and researchers, offering data that undermine the presumed necessity of meaning making and extrapolating from suggestive research findings to frame recommendations for practice. My intention in this article is in a sense to reverse this emphasis, by offering first some data regarding the efficacy of psychotherapy for bereaved persons and then pondering the prospects for a meaning reconstruction approach to grief counseling in light of these results. As we shall see, these somewhat different peregrinations lead to a surprisingly similar destination–namely, a more discriminating endorsement of grief therapy for some, but by no means all bereaved individuals, and a cautious optimism about the value of “intervening in meaning” for this subset of mourners.
Is Grief Therapy Justified?
A great deal has been written about grief counseling in the last 25 years, giving rise to a burgeoning popular and professional literature proffering assistance to the bereaved, as well as to persons suffering a wide range of additional losses through means other than the death of a loved one. In the face of this proliferating attention, one might assume that grief counseling is a firmly established, demonstrably effective service, which, like psychotherapy in general, seems to have found a secure niche in the health care field, at least in North America.
Ironically, perhaps, this assumption represents at best a half- truth. Grief counseling has indeed proliferated, both in the formal venues of professional conferences, workshops, and publications, and in the countless institutional or community-based programs run by grief therapists, or operated on a mutual support basis by lay leaders. Moreover, scores of uncontrolled descriptive studies indicate that bereaved persons in these programs typically report reduced depressive, anxious, or general psychiatric symptomatology following their participation, reinforcing the impression that grief counseling is indeed effective in assisting with “recovery” from acute grieving.
However, only controlled studies in which bereaved individuals are randomly assigned to treatment and control conditions can yield a clear verdict on the effectiveness of grief therapy. Uncontrolled studies are at best suggestive, as acute grief could simply remit with the passage of time, as a function of “curative factors” (e.g., social support) in the natural environment or as a result of the bereaved person’s own coping efforts. Indeed, when controlled studies of professional interventions are analyzed (as in Rose and Bisson’s, 1998, review of debriefing interventions cited by Davis et al., in press), results are often equivocal, with different studies suggesting positive, negative, and “no difference” conclusions. Such results make it essential to conduct a comprehensive review of all controlled outcome studies of grief counseling to reach confident conclusions about whether grief therapy is indeed effective, and if so, for whom.
My colleagues–Barry Fortner, Adam Anderson, JefF Berman– and I have just completed such a review (Fortner & Neimeyer, 1999).1 In undertaking this project, we were struck by the extent to which recent reviewers of this literature (e.g., Kato & Mann, 1999) analyzed only a small subset of the available studies, and even then, relied primarily on impressionistic evaluations of outcome, supplemented by relatively global application of quantitative review methods. Others (Allumbaugh & Hoyt, 1999) offered more detailed quantitative reviews of the published and unpublished literature but included numerous uncontrolled one-group studies that could have inflated estimates of the effectiveness of the therapies studied. The result was significant discrepancy from one review to the next regarding the efficacy of psychosocial interventions for the bereaved. To remedy these and other shortcomings of previous reviews, my colleagues and I located all scientifically adequate outcome investigations of grief therapy published between 1975 (when the first such research appeared) and 1998, a total of 23 separate studies reported in 28 different papers. As criteria for inclusion, all had to focus on bereaved persons mourning the death of a loved one, who received some form of psychosocial intervention (psychotherapy, counseling, or facilitated group support), and who were randomly assigned to a treatment or control condition. The over 1,600 participants in these studies had experienced a wide range of losses–of spouses, children, and other family members–who had died from a broad spectrum of causes, both sudden and protracted. Professional therapists provided therapy in 19 of these studies, and nonprofessionals conducted the remainder. Finally, it was notable that the majority of studies assessed outcome on generic measures of health, depression, anxiety, or psychiatric distress, while only a few attempted to measure grief per se, and then typically using idiosyncratic or unvalidated measures.
We assessed the efficacy of grief therapy using two statistics, one of which has been widely used in meta-analyses conducted over the last 20 years and one of which represents a recent innovation in quantitative review procedures. The first of these was Cohen’s d, which reffects the posttest difference between treated and untreated groups across a range of outcome measures–a straightforward measure of the degree of benefit associated with participation in therapy (Cohen, 1997). The second, more novel procedure allowed us to estimate treatment-induced deterioration, which represents the proportion of participants who are worse off after treatment than they would have been if they had been assigned to the control group.2 It is important to emphasize that this latter measure did not reffect absolute deterioration–the case of an individual functioning more poorly after therapy than before–because the therapy could still be deemed helpful in this case if the same person would have been even more symptomatic with no treatment. Likewise, therapy clients could actually make some gains but still fall short of where they would be if the treatment held them back relative to where they would be if untreated. Thus, we were interested in treatment- induced deterioration, defined as all instances in which therapy recipients theoretically would have fared better if left alone, irrespective of the absolute direction of change they showed over the course of the study.
Analyzing the 23 randomized controlled studies using these metrics produced some interesting results. To begin with, we found that the mean effect size of . 13 was positive, and reliably different than zero, reflecting the superiority of outcomes for treated relative to untreated persons. However, the effect size was also quite small in absolute terms, when compared with the much more substantial effects associated with psychotherapy for depression and for psychological disorders more generally. Stated in other terms, the average participant in grief therapy was better off than only 55% of bereaved persons who received no treatment at all–hardly an impressive demonstration of the efficacy of grief counseling.
The analysis for treatment-induced deterioration was perhaps more sobering still. When we computed this statistic, we discovered that nearly 38% of recipients of grief counseling theoretically would have fared better if assigned to the no-treatment condition; in strong contrast, only 5% of clients in a broad range of psychotherapies for other problems showed such deterioration (Anderson, 1999). Thus, not only is the tangible benefit of grief therapy small, but its risk of producing iatrogenic worsening of problems is unacceptably high–a troubling pattern that is unique among typically effective and safe psychosocial interventions.
What could account for these disconcerting findings? Unfortunately, simple explanations focusing on the intractability of loss, or the necessity to engage in distressing “grief work” prior to reestablishing an emotional equilibrium, however valid, fail to account for the differential deterioration of treated versus untreated clients. Moreover, spontaneous improvement of treated and untreated subjects alike seems implausible, given the findings of Allumbaugh and Hoyt (1999) suggesting essentially no improvement in the latter category over the brief periods associated with the average treatment study. The brevity of the therapies provided (whose mean number of sessions was 7) might also be argued to mitigate the effectiveness of these interventions, as substantial grief can persist for a period of years. However, such an argument is weakened by the variable length of therapy represented by these two dozen studies, and our finding that effect size was uncorrelated with length of treatment. Nor did categorical distinctions associated with the therapies or therapists account for the poor showing of these therapies, as outcome was also unrelated to type of treatment (individual vs. family vs. group), therapeutic approach, or level of training of therapists (professional vs. nonprofessional). What, then, could explain the limited use and high risk of grief counseling?
In pursuing answers to these questions, we found some promising leads in the differential responses of different clients to the interventions offered. For example, clients varied considerably in the length of time between their loss and enrollment in bereavement programs (M = 6 mos.), with some being offered services immediately after the death had occurred, and others being approached many years later. Interestingly, better outcomes were obtained for clients who were more distant from the death (r = .5 between treatment effect and weeks of bereavement). Likewise, the deterioration effect was strongly correlated with client age (r = –.7), suggesting that younger clients fared better than older ones in such therapies. Perhaps most interesting, however, was the result of a follow-up analysis in which we discriminated between outcomes in those 5 studies offering treatment for persons who were traumatically bereaved (e.g., through violent, sudden, or untimely death, or whose grief was more chronic) and those that focused on “normal” bereavement reactions. Here, the results were especially clear: Counseling for normal grievers had essentially no measurable positive effect on any variable (d = .06), whereas the subset of studies offering therapy for traumatic grief showed a reliable positive effect (d = .38). Equally heartening was the finding that deterioration effects were substantially lower for traumatized clients (17%) than for normal or unselected samples, for whom nearly one in two clients suffered as a result of treatment. Together, these findings point toward an intriguing and consistent conclusion: That grief therapy is appropriately offered to mourners experiencing protracted, traumatic, or complicated grief reactions. Conversely, existing evidence from scientifically credible controlled outcome trials suggests that grief therapy for normal bereavement is difficult to justify.
Grief Therapy as Meaning Reconstruction
Although the evidence reviewed above provides some encouragement for grief therapy as a legitimate treatment for chronic or traumatic bereavement, it remains the case that the effect sizes associated with such therapies are only half as robust as those associated with psychosocial interventions for other problems. It therefore seems critical to pursue the question, “Why are the results of grief therapy so modest, even for potentially traumatic grief reactions?”
Answers could be sought in a variety of factors. At a methodological level, it could simply be that medically oriented researchers are assessing an inappropriate domain of outcome, focusing on psychiatric and physical problems, rather than features distinctive to grief per se (Neimeyer & Hogan, 2Э00). For example, Rubin (1999) has argued that adaptation to loss progresses along two clearly distinguishable tracks, one of which concerns symptomatology (e.g., anxiety and depression) and the other of which centers on the relationship to the deceased. In support of this conceptualization, independent researchers (Byrne & Raphael, 1997) have provided evidence that core features of grieving focusing on disruptions in the attachment relationship to the lost loved one (such as yearning for the deceased) are relatively independent of general depression (see also Jacobs & Prigerson, this issue). Significantly, persistent relational distress also predicts poorer longterm outcome defined in terms of both mental and physical health status (Prigerson et ah, 1997). This raises the possibility that unique goals of grief therapy–such as helping the bereaved transform the concrete relationship to the deceased to a symbolic one– have yet to be assessed by existing outcome studies, despite the existence of validated scales that might reveal such distinctive patterns (Neimeyer & Hogan, 2300).
A more substantive explanation for the generally unimpressive benefits of grief therapy might focus on the nature of the treatment itself. To a remarkable degree, controlled studies of grief counseling fail to describe the conceptual models that underpin their approach to therapy, in sharp contrast to the general psychotherapy outcome literature, which tends to test well-delineated models of treatment. When grief therapy is described in such studies, it tends to be based on suspiciously simplistic models, such as stage theories of grieving that have been largely repudiated by contemporary theorists and researchers (Gorr, 1993; Neimeyer, 1998). Thus, a second possible reason for the weak showing of grief counseling is that it rarely draws on the best available theories regarding the nature of bereavement and its facilitation.
If investigators were interested in designing and testing more promising approaches to grief therapy, on what principles and procedures might they draw? One answer that is suggested by the results of Davis et al. (this issue) would be those deriving from a focus on meaning making processes in the aftermath of bereavement. Such a perspective would argue for a significant shift in the implicit paradigm under which grief therapy is practiced, away from a medical model emphasizing the control of disruptive symptomatology, and beyond the well-intended but vague assumption that a sharing of feelings in a supportive environment will promote “recovery.” Instead, intervention, when indicated, might be informed by the proposition that “meaning reconstruction in response to a loss is the central process in grieving” (Neimeyer, 1998, p. 110). My goal in the present section is to consider what guidance the results of Davis et al. might give in developing such an alternative treatment approach, and how related scholarship might further the synergy of research and practice in this area.
Davis et al. (this issue) discussed two illuminating studies of persons bereaved by sudden infant death syndrome (SIDS) and motor vehicle accidents (MVAs), reporting data on the percentage of those who undertake a “search for meaning” in their loss, and the relation of this search to psychological well-being. In so doing, the authors provocatively focus attention on the important minority of bereaved who do not seek the meaning of the death, and the evidence that they fare as well as–or better than–their counterparts who engage in a protracted search for significance. Their findings therefore serve as a useful corrective to the glib assumption that meaning must be sought in the death and must be found if the loss is to be resolved. However, taken out of the broader context of their findings, this conclusion presents a distorted picture of the relationship of meaning making to a favorable outcome to bereavement. For this reason, it is worthwhile to reiterate and clarify the findings they report in their article. In brief, the data of the SIDS study document the following:
- 86% of parents who lost a child to SIDS undertook a search for the meaning of the death, whereas 14% did not.
- Of the 14% who did not seek existential answers for why the death occurred, 3% seem to have foreclosed on pre-existing meanings for the death (e.g., as God’s will), whereas 11% reported having no such meaning.
- 18% of the parents had discontinued their search for meaning in the first month following loss, over half of them without finding satisfactory answers.
- Those parents who neither sought nor found meaning in the death fared as well psychologically as those parents who had successfully struggled for meaning, and both groups ultimately did better than those who searched for meaning in the death, but found none.
- The search for meaning was ongoing for many parents, even when some sense was made of the loss early in bereavement; in other words, sense making in the early weeks of loss was provisional rather than permanent.
Many of these patterns are reinforced by the results of the MVA study, in which 70-80% of respondents reported having been concerned with the issue of the meaning of their loved one’s death, whereas 20–30% did not. Again, those persons who sought answers to no avail fared worse in their adjustment to the loss than did those who never sought answers in the first place, although the large number of persons who sought and found workable meanings were intermediate in their adjustment on a number of measures.
Taken together, these studies document that the “search for meaning” plays a compelling role in the grief of the great majority (70-85%) of persons experiencing sudden, potentially traumatizing bereavement, although a significant minority apparently copes straightforwardly with their loss, without engaging in deep-going reflection about its significance (Attig, 1996). For those who seek meaning and find none, the loss can be excruciating, and data suggest that they report intense suffering on a variety of outcome measures. Conversely, bereaved persons who find a measure of meaning in the loss fare better, rivaling the adjustment of those who never feel the need to undertake existential questioning in the first place. Even these “finders” are not necessarily “keepers,” however, insofar as many of those who felt they had found answers to why the loss had occurred revisited these answers in the months that followed.
These findings carry important implications for the practice of grief therapy. First, and most obviously, they highlight the fact that a quest for meaning plays a prominent role in grieving, at least for those who are bereaved by the sudden death of a loved one. When a client is struggling for significance in the loss, the counselor would be well advised to facilitate this process, perhaps by drawing on some of the specific meaning-making strategies that have been formulated for this purpose (Neimeyer, 1998). Second, grief counselors should be cautious about instigating a search for meaning in the minority of cases in which clients do not spontaneously undertake such a search, insofar as these individuals might well be coping adaptively using pragmatic, rather than philosophic, strategies. Finally, counselors would do well to remember that meaning-making is more an activity than an achievement, as early, provisional meanings of the death tend to be revisited as the reality of living with loss raises new questions and undermines old answers.
The Meaning of Meaning
Although Davis et al. (this issue) appropriately sensitized practicing therapists to the importance of a search for meaning for most, but not all, bereaved persons, their data are ultimately too global to specify how such a search might be undertaken in the context of therapy. However, findings from another study by Davis, Nolen-Hoeksema, and Larson (1998) extend this work in practically helpful directions.
Unlike the participants in the SIDS and MVA studies, the 205 participants in the 1998 research were bereaved by the slower progressive death of loved ones in home-based hospice settings, chiefly by cancer. Because death was anticipated, the researchers were able to assess pre-loss functioning (an average of 3 months prior to the death), as well as post-loss adjustment across the first 18 months of bereavement. More important, they posed more refined questions about the effort after meaning at each of the post-loss interviews, asking not only whether the survivor had been able to “make sense of the death,” but also whether she or he had “found anything positive in the experience.” In this way the researchers were able to examine two component processes of meaning making, sense-making and benefit-finding, and to demonstrate that these were largely unrelated and had different antecedents and consequences. For example, those persons who were able to make sense of the death 6 months later tended to be those who lost an older relative, who had a pre-existing spiritual or religious framework, and who had displayed less distress in the months preceding the death. Conversely, the ability to find a “silver lining” in the loss was associated only with the personality characteristic of (preloss) optimism-pessimism. Furthermore, while being able to explain the loss was associated with less distress 6 months following the death, only finding benefit in the experience was consistently associated with better adjustment at the 13- and 18-month post loss interviews. More interesting, analysis of participants’ responses indicated that it was not the content of the sense made (e.g., that death was part of the life cycle; that it was the will of God) or the benefit found (e.g., an enhanced perspective, bringing the family together) but simply whether meaning was made of the loss, that predicted adaptation to bereavement.
For grief therapists, these findings suggest that the meaning making processes that are most relevant to facilitate tend to shift over time, from an early emphasis on finding an answer to the question of “why” the death occurred, to a later focus on the positive (albeit unsought) benefits of the loss for survivors. Moreover, the directional change in meaning reconstruction clearly is important to trace, insofar as those persons who gained in sense-making from 6 to 13 months post-loss showed the greatest improvement in psychological well-being, whereas those who lost ground in the quest for meaning showed the greatest deterioration in functioning. Meaning reconstruction therefore appears to be a dynamic process with multiple aspects, whose provisional outcomes predict key features of adaptation to bereavement.
Extending the work of Davis and his collaborators, how might we elaborate the notion of meaning reconstruction to make it still more adequate to the subtleties of this process in the lives of bereaved persons? What seems essential is transcending a simple “cognitive” reading of the concept of meaning, which interprets it as a conscious, intellectual acquisition of individuals, rather than a predominantly tacit, passionate process that unfolds in a social field. Drawing inspiration from recent qualitative research in bereavement (Neimeyer & Hogan, 2300) and broader constructivist theories of psychotherapy (Neimeyer & Mahoney, 1995), meaning reconstruction might be defined more broadly to include the following.
- The attempt to find or create new meaning in the life of the survivor, as well as in the death of the loved one. Because our relationships with intimate others provide a repository of shared memories and a validating context for our most cherished beliefs (Landfield, 1988), the loss of these relationships undermines our self-narrative, and with it, our identity (Neimeyer, 2300b). Cultivating this insight, contemporary grief theorists such as Attig (1996) have construed grieving as a process of “relearning” the world and the self, finding a new existential grounding for one’s self-concept and life direction. Frantz, Farrell, and Trolley (2300) have documented the pervasiveness of this personal reconstruction in the lives of nearly 400 bereaved adults, who reported a year following the loss that they viewed themselves as more mature and independent (3 2%), living more fully in the present (17%), and more compassionate and expressive with others (14%). Significantly, a minority also acknowledged regressive shifts in their sense of self, noting that a part of them had died (10%), that they were more fearful of death (5%), or were hardened by the experience (5%).
2 The integration of meaning, as well as its construction. A more adequate theory of personal knowledge would view any given construction of significance as situated within a unique ecology of meaning (Neimeyer & Harter, 1988), a system of personal constructs that vary in their hierarchical organization (Kelly, 1955/ 1991; Mahoney, 1991). Thus, the “same” meaning might for one person represent a relatively peripheral construction, whereas for another person (or at a later point in bereavement) it might function as a central, organizing frame for living. Something of this kind has been demonstrated in the sophisticated qualitative research of Richards and Folkman (Richards, Acree, & Folkman, 1999; Richards & Folkman, 1997), who have traced the way in which spiritual constructions of the meaning of a death evolved across the course of bereavement for caregivers of gay men who died of AIDS. Initially, spiritual interpretations seemed to provide ad hoc explanations for the death itself, serving as coping resources for the surviving partner. With time, however, spiritual frames of meaning came to pervade the existence of the majority of these men, resulting in a substantially deepened sense of purpose and significance in their lives with others.
- The construction of meaning as an interpersonal, as well as personal, process. Although the meanings we assign to loss are highly idiosyncratic, they are nonetheless negotiated in a social context (Neimeyer, 1998). Nadeau (1997) has studied this highly interactive process in families contending with the death of a member, developing a taxonomy of the strategies by which they collectively seek significance in the event (through the interpretation of meaningful coincidences, “mind reading” the desires of the deceased, and so on). Similarly, Hagemeister and Rosenblatt (1997) have investigated the shared meanings of the sexual relationship between spouses who have lost a child, meanings that promote joint healing (e.g., “this is a way of affirming our love for each other”) or create emotional impasses (e.g., “sex is too painful, because it is how we made this child”). Thus, the social field is vital in the construction of meaning, providing an audience for those accounts by which we attempt to render unwelcome life transitions intelligible (Harvey, 1996).
- The anchoring of meaning making in cultural, as well as intimate, discursive contexts. A fuller appreciation of reconstructive processes following loss must surely take into account frameworks of meaning that are too large to be confined to a single local network of relationships and too enduring to be accumulated in a single generation. Indeed, the very terms in which we construe death and grief are cultural artifacts, as are the social roles we assign to survivors (Neimeyer, 1998). As Klass (1999) has demonstrated in his elegant ethnographic study of bereaved parents, grieving individuals routinely draw on the discourses and rituals of the cultural traditions in which they are situated, and reinterpret these at personal and interpersonal levels. A more extensive appreciation of the uniqueness of grieving among groups that are often conflated in the minds of would-be helpers is beginning to emerge from qualitative studies, such as the focus groups with representatives of various Asian American subcultures conducted by Braun and Nichols (1997).
- Tacit and preverbal, as well as explicit and articulate meanings. A common shortcoming of cognitive accounts of meaning-making is their simplistic assumption that the construction of significance is (or should be) a logical, verbalizable process (Neimeyer, 1995). But more philosophically sophisticated (Polanyi, 1958) and clinically compelling (Guidano & Liotti, 1983) accounts of personal knowledge and its development argue that the “deep structure” of our constructions of reality is in principle tacit, inexpressible in any complete sense in public speech. Stated differently, there are some meanings that are too embedded in our lives, too embodied in our actions, to be amenable to formulation in a set of “beliefs” or “selfstatements.” In the context of grief therapy, this implies that the counselor needs to attend to nuances of client meanings that might be hinted at by vocal tones, gestures, and emphases, as much as communicated in straightforward propositions (Neimeyer, 2300a). The possibility that the most important meanings of loss might elude simple verbal formulation also prompts a variety of more metaphoric, poetic, and narrative strategies for exploring the multiple meanings of a loved one’s life and death, and its relevance to a client’s own (Neimeyer, 1998).
- The processes of meaning reconstruction, as well as its products. A corollary of the predominantly cognitive interpretation of meaning in our field is that it is often regarded as a product, as something that is “searched for” and “found,” rather than created by the persons or groups who seek it. Although it is important to acknowledge that some meanings of loss are indeed discovered as well as invented (Attig, 2000), an emphasis on pre-existing truths obscures the delicate processes by which fresh meanings are typically constructed. For example, one key process entailed in rebuilding an assumptive world decimated by loss might be the ability to tack back and forth among different styles of narrating our experience, between objective, external accounts, subjective, involved narratives, and reflexive self-examination (Goncalves, Korman, & Angus, 2300). Principles and procedures for developing life narratives that can accommodate traumatic experiences have been offered by a number of contemporary constructivist therapists (Neimeyer & Stewart, 1998; Sewell, 1997; Stewart, 1995).
In summary, the perspective I am advocating here argues for our selective engagement in grief therapy with those bereaved persons whose grief is traumatic or prolonged, as well as respectful witnessing of the self-help efforts of those who do not require our well-meaning involvement–and might even be harmed by it. When grief therapy is offered, I believe it must attend to the profound challenges to clients’ (inter )personal systems of meanings brought about by tragic loss and facilitate the survivors’ own struggle to find significance both in the death and in their ongoing lives. Finally, I would advocate a more refined and clinically rich conception of the process of meaning reconstruction, one that accredits its complexity, its social character, and the conditions that facilitate or impede it. The work of Davis and his colleagues, as well as the other contributors to this special series in Death Studies, offer us some useful direction in this effort.
A close review of the most authoritative and reliable research currently available leaves us with sobering conclusions about the general effectiveness (or even advisability) of grief counseling and therapy, as well as a few clues as to when professional intervention might be more clearly indicated. These same findings, in combination with the work of researchers like Davis et al. (in press), also suggest possibilities for the refinement of grief therapy and research in a way that can ultimately strengthen both. I hope that some of the results and reflections offered here contribute to this development, and prompt us toward a more adequate theory of the reconstructive processes that permit adaptation to profound loss.
What Has Become of Grief Counseling? An Evaluation of the Empirical
Foundations of the New Pessimism
A pessimistic view of grief counseling has emerged over the last 7 years, exemplified by R. A. Neimeyer’s (2000) oft-cited claim that “such interventions are typically ineffective, and perhaps even deleterious, at least for persons experiencing a normal bereavement” (p. 541). This negative characterization has little or no empirical grounding, however. The claim rests on 2 pieces of evidence. The 1st is an unorthodox analysis of deterioration effects in 10 outcome studies in B. V. Fortner’s (1999) dissertation, usually attributed to Neimeyer (2000). Neither the analysis nor Fortner’s findings have ever been published or subjected to peer review, until now. This review shows that there is no statistical or empirical basis for claims about deterioration effects in grief counseling. The 2nd piece of evidence involves what the authors believe to be ill-informed summaries of conventional meta-analytic findings. This misrepresentation of empirical findings has damaged the reputation of grief counseling in the field and in the popular media and offers lessons for both researchers and research consumers interested in the relationship between science and practice in psychology.
It ain’t what you don’t know that gets you into trouble. It’s what you know for sure that just ain’t so.
–Attributed to Mark Twain
This well-known aphorism provides a humorous reminder of the perils of unwarranted certainty. In this article, we consider an emerging consensus in the literature on grief or bereavement (we use these terms interchangeably) counseling that, in our view, has arisen independently of the scientific literature on this subject but is nonetheless prevalent (indeed, almost omnipresent) in scientific journals and professional summaries of research designed to guide policy and practice. We describe the evolution of this new pessimism regarding the efficacy of bereavement counseling, with attention to two central claims that ground this reading of the literature. We consider these claims in detail in an effort to demonstrate that neither of them can be substantiated relative to traditional scientific standards, and we offer a different summary of findings on grief counseling that is more faithful to the actual research literature. Finally, we consider the implications of this unwarranted certainty in the research community for practitioners striving to conform to the ideal of empirically grounded practice, and we remind practitioners of several key guidelines that may be helpful to nontechnical readers as they evaluate conclusions offered in scientific journals.
The New Pessimism
Strong consensus has emerged over the past 7 years in the bereavement research community: Grief counseling is at best weakly effective and at worst harmful. This new pessimism about the effectiveness of grief interventions is widely reflected in re-
views of literature published in major research journals and also in the 2003 Report on Bereavement and Grief Research published by the Center for Advancement of Health (CFAH, 2003). The CFAH national report concluded that “the evidence from well-conducted studies of interventions . . . challenges the efficacy and effectiveness of grief interventions for those experiencing uncomplicated bereavement. This evidence also indicates that concerns are warranted about the potential of interventions to cause harm to some individuals” (CFAH, 2003, p. 72).
Such conclusions are regularly reiterated in journals targeted at grief researchers and practitioners. For example, in Death Studies, Paletti (2005) cautioned that “the alignment of research and practice is a critical issue in bereavement care; recent evaluations of a number of intervention programs have indicated that such care might be ineffective, and worse, counterproductive” (p. 669). Also in Death Studies, Shapiro (2005) warned, “As bereavement practitioners, we have recently been challenged by careful reviews (Jordan & Neimeyer, 2003) suggesting that one-size-fits all approaches to ‘grief work’ mandating exploration of distressing feelings can intensify distress for a significant number of mourners” (p. 262).
Claims about deterioration effects form perhaps the most persuasive argument for restraint in offering interventions for bereaved individuals and are certainly shocking, when taken at face value, to bereavement practitioners. Our purpose in this article is to trace this new pessimism among bereavement researchers to its empirical roots and to evaluate the basis for claims of iatrogenic effects in particular and for the more global (although less alarming) claim that grief interventions generally are much less efficacious than psychosocial interventions for other emotional and behavior problems.
Deterioration Effects in Grief Counseling
Probably the major impetus for the rash of cautionary messages regarding bereavement interventions has been the striking finding, first published by Neimeyer (2000), that such treatments may be harmful. Neimeyer reported on the findings of a meta-analysis of grief interventions using a “novel procedure [designed] to estimate treatment-induced deterioration, which represents the proportion of participants who are worse off after treatment than they would have been if they had been assigned to the control group” (p. 544). The findings for the treatment-induced deterioration effects (TIDE) were staggering: Averaging over all studies that provided the necessary information, Neimeyer concluded that “nearly 38% of recipients of grief counseling theoretically would have fared better if assigned to the no-treatment condition” (p. 545). In other words, more than one third of grief counseling clients were worse off at the end of treatment than they would have been in the absence of treatment. Neimeyer reported that the results of subsidiary analyses focusing on “normal” grievers (as opposed to those who were “traumatically bereaved”) were even more alarming: In these “normal or unselected samples . . . nearly one in two clients suffered as a result of treatment” (p. 546).
These findings are now treated as established scientific fact and have been widely cited in major professional journals, including American Psychologist (Bonanno, 2004), Journal of Personality and Social Psychology (Bonanno, Moskowitz, Papa, & Folkman, 2005; Bonanno, Wortman, et al., 2002), Psychiatric Annals (Reissman, Klomp, Kent, & Pfefferbaum, 2004), Journal of Clinical Psychiatry (Harkness, Shear, Frank, & Silberman, 2002), Suicide and Life-Threatening Behavior (Jordan & McMenamy, 2004), Applied and Preventative Psychology (Bonanno, Papa, & O’Neill, 2002), and Death Studies (Jordan & Neimeyer, 2003; Ott & Lueger, 2002; Sikkema, Hansen, Kochman, Tate, & Difranciesco, 2004; Wolf & Jordan, 2000). The findings were also reported in great detail in both the Handbook of Bereavement Research: Consequences, Coping, and Care (Stroebe, Hansson, Stroebe, &
Schut, 2001) and the CFAH (2003) Report on Bereavement and Grief Research. They have begun to be cited in popular books (Sommers & Satel, 2005) and newspapers (“Grieving,” 2003) and were even cited in a report on grief and bereavement services for the Department of Human Services in Victoria, Australia (Nucleus Group, 2004).
TIDE Findings: In Search of the Evidence
Certainly, if the TIDE findings are valid, pessimism about the efficacy of bereavement interventions and extreme caution in their application are warranted. But what is the basis for these claims? This question became a pressing concern for Dale G. Larson when he was invited to participate in a scientific panel titled “Grief Counseling: Can It Be Harmful?” at a national professional meeting. Preparing for the panel, he dutifully reviewed the literature on iatrogenic effects of grief counseling, turning first to Neimeyer’s (2000) article because it was frequently cited in journals, presented at conferences, and discussed among colleagues. He began to review the article, assuming that it contained the original empirical work. However, a cursory reading of Neimeyer (2000) revealed that this was not an empirical study but rather a summary of findings of past research, both published and unpublished. Neimeyer reported the TIDE findings as part of a summary of the results of an unpublished meta-analysis of grief interventions attributed to himself, Barry Fortner, Adam Anderson, and Jeff Berman and presented at a conference (Fortner & Neimeyer, 1999). Neimeyer (p. 543) explained in a footnote that the only report of the metaanalysis was in a dissertation completed by Fortner (1999).
When requests for published or unpublished reports of the research confirmed that the Fortner (1999) dissertation was and still is the only available record of this study, the dissertation became the focus of attention. The dissertation is relatively brief– just 28 pages of text, excluding references and tables. In the Method section, Fortner (1999) described a “statistical method for determining the theoretical proportion of participants who were worse off after treatment than they would have been if they had been assigned to the control group” (p. 14). This is the method that yielded the 38% TIDE estimate. Obviously, the validity of this statistical technique is crucial to evaluating the validity of the TIDE findings, so we review this procedure in some detail.
The TIDE statistic has never been published, nor is any statistical justification for this approach available in unpublished form. Fortner (1999) cited two sources for this technique: a presentation by Anderson, Berman, and Fortner (1998) at the annual meeting of the Society for Psychotherapy Research, and a master’s thesis by Anderson (1988) completed at the University of Memphis. Print copies of these sources were not available, although it appears that each of them reports on another application of the TIDE technique (to a broad sample of psychotherapy outcome studies) rather than presenting a rationale for the technique itself. Thus, the critique presented in the present article must be based on the brief description provided by Fortner (1999).
TIDE Findings: Pattern of Citations
The first question raised by the TIDE findings involved the pattern of citations. As Figure 1 shows, this result was cited regularly in the 7 years from 2000 (when Neimeyer, 2000, summarized the findings in Death Studies) to 2006 (when Dale G. Larson sought to understand the basis for the TIDE claims). What is interesting in Figure 1 is that the only publicly available (but not peer-reviewed) account of the data behind these findings (Fortner, 1999) was cited only once–in the Neimeyer (2000) footnote. Most authors have simply cited Neimeyer’s (2000) summary as the source of this finding, with no reference to the actual (Fortner, 1999) study. For example, in a review article on loss and trauma published in American Psychologist, Bonanno (2004) offered the following summary of Fortner’s findings: “In one of these analyses, an alarming 38% of the individuals receiving grief treatments actually got worse relative to no-treatment controls, whereas the most clear benefits were evidenced primarily with bereaved individuals experiencing chronic grief (Neimeyer, 2000)” (p. 22). Similarly, in a multiple case study published in the Journal of Clinical Psychiatry, Harkness et al. (2002) asserted that “a recent meta-analysis found a strikingly low effect size (0.15) [sic] for supportive grief interventions and a significant minority worsened with these treatments. (Neimeyer, 2000)” (p. 1120). At best, such citations represent inadequate scholarship, in that they fail to point readers to the empirical evidence that justifies these summary claims. This failure to cite the Fortner dissertation is also troubling in that it implies that these authors had not read the empirical report and therefore were not in a position to evaluate the validity of its conclusions.
More recently, authors have cited a subsequent review article by Jordan and Neimeyer (2003), or even other writers (who themselves cited Neimeyer, 2000), as the basis for the TIDE claim. For example, in a research review (targeted at clinicians) published in Clinical Neuroscience Research, Craighead and Nemeroff (2005) ignored both Fortner (1999) and Neimeyer (2000):
In a review article published in Psychiatric Annals, Reissman et al.
(2004) cited both Neimeyer (2000) and Bonanno (2004; who, in turn, cited Neimeyer) as the bases for the conclusion that grief work may be “ineffective, and perhaps even deleterious” (p. 631).
This pattern of citations can be misleading in two ways. First, attributing the finding to Neimeyer’s (2000) published article rather than Fortner’s (1999) unpublished dissertation can lead readers to believe that the empirical data and statistical analyses underlying the TIDE results have been subjected to peer review, which is not the case. Second, the citation of more recently published articles, either instead of or in addition to Neimeyer (2000), may give naive readers the impression that this finding has been corroborated in additional investigations, which again is not the case.
TIDE Findings: A Post Hoc “Blind” Peer Review
The TIDE technique represents an unorthodox statistical analysis–one that has never been subjected to peer review. Recognizing that the pattern of citations might have masked this fact and that an evaluation of the validity of this technique would be beyond his own statistical competence, Dale G. Larson contacted Gary R. VandenBos, the American Psychological Association (APA) publisher and the managing editor of the American Psychologist, for guidance about the citation and evaluation of nonpeer-reviewed work. On the basis of the apparently high impact of the TIDE findings and their unconventional citation pattern (including numerous citations in journals published by APA), VandenBos agreed to conduct a post hoc blind peer review of both the TIDE statistic and the findings reported in Fortner (1999), following the procedures that any APA journal would use. The reviewers, both national methodological and statistical experts, were asked to consider whether the formula for TIDE generates a stable, reliable, valid, and meaningful number that can be interpreted as claimed (i.e., as a deterioration effect). Neither the TIDE statistic nor Fortner’s conclusions passed this peer review. The experts conclusively agreed that the TIDE statistic is seriously flawed and that there is no valid basis for the claim that 38% of grief counseling clients suffered deterioration (nor for Neimeyer’s, 2000, statement that one in two normally bereaved clients suffered).
TIDE Analysis: Technical Considerations
Meanwhile, William T. Hoyt joined the search for the basis for the TIDE claims. With a background in meta-analytic work, he was intrigued by the claim that summary data (i.e., means and standard deviations for treatment and control groups) could be used to derive an estimate of the proportion of clients whose condition deteriorated as a result of treatment. He was not familiar with any commonly accepted approach to this problem in metaanalysis and was curious about the statistical justification for this conclusion. In an online appendix to this article (available on the Web at http://dx.doi.org/10.1037/0735-7028.38.4.347.supp), we show how the statistic used by Fortner (1999) is computed and raise questions about its interpretation as an index of TIDE.
After searching out the original report of the TIDE finding, reading it carefully, and consulting with statistical experts, we came to an astonishing conclusion: The 38% finding, which has swept over the death studies field like a veritable TIDE-al wave, appears to have no basis in fact. It is based on a meta-analysis that has never been published, using a statistic that itself has never been published or reviewed by the methodological community and that seems on its face to be obviously flawed. A blind post hoc peer review of Fortner’s (1999) study by national experts confirmed these startling conclusions. This is good news for grief counselors who might have been concerned, on the basis of this “finding,” that they could have been unknowingly harming their clients. There appears to be no foundation for this concern (Larson & Hoyt, 2006).
However, recent reviewers have not confined themselves to the 38% TIDE finding as a source for their negative views of grief counseling. What about the second basis for pessimism: the claim that research shows that grief therapy is ineffective, or at any rate so weakly effective that it is not warranted, at least for so-called normal grievers? This conclusion is based on findings of metaanalyses–quantitative reviews of existing research. In our quest to understand the pessimistic turn among grief researchers, we next examined the basis for this claim.
Analysis of Meta-Analyses and Reviews
Meta-Analyses of Grief Counseling Outcome Studies
Allumbaugh and Hoyt (1999)
The most thorough published quantitative review of research on grief interventions was conducted by Allumbaugh and Hoyt (1999). These authors performed an extensive search including both published and unpublished literature. They used an innovative meta-analytic technique developed by Becker (1988) to allow for incorporation of both pre- and posttreatment scores on the dependent variables from both treatment and control groups. This method is advantageous, when most studies collected both pre- and posttreatment data, because it treats change over time (rather than simple posttreatment status) as the outcome of interest and because it estimates change over time separately for treatment and control groups and then compares them. This feature is especially interesting relative to the presenting problem of bereavement, as it is expected that symptoms will abate over time to some extent as a natural process (Bonanno, Wortman, et al., 2002), and Becker’s method quantifies the extent to which this natural recovery was observed in the studies reviewed. It also allows inclusion of data from studies with no control group, because pretreatment– posttreatment gains for the treatment group can be compared meta-analytically with corresponding gains from control group participants in other studies (Becker, 1988).
Data from 35 independent investigations of grief counseling yielded an aggregate effect size of d 0.43, which is smaller than effect sizes typically observed when counseling and psychotherapy are applied to other psychological or emotional problems. (Such effect sizes typically average d 0.80; see Lipsey & Wilson, 1993; Wampold, 2001.) However, homogeneity tests indicated significant variability in effect sizes beyond that expected as a result of chance (i.e., resulting from sampling error). Such heterogeneity among study effect-size estimates can arise when studies differ on characteristics that are systematically related to treatment effectiveness–that is, on moderator variables.
Allumbaugh and Hoyt (1999) found two moderator variables that appeared particularly important for understanding the relatively low aggregate effect size. These were study characteristics on which participants in the modal study in this review differed from typical clients in grief counseling: motivation for counseling and time since loss.
Motivation for counseling. Clients in most of these studies were recruited into treatment rather than electing to seek grief counseling on their own. These research participants may not be typical of clients in actual practice settings. Furthermore, effect sizes for the handful of studies involving self-referred clients were quite large (ranging from d 1.17 to d 3.05), suggesting a substantial benefit for this client group.
Time since loss. Also, the mean time between the loss of a loved one and the start of the grief intervention for these studies was 27 months. This may be related to the participant recruitment procedures just discussed. Researchers recruiting participants who had experienced a particular type of loss (e.g., spouse, child) might have felt compelled to appeal to participants whose losses were relatively remote if they were unable to recruit a sufficient number of recently bereaved participants. On the basis of the moderator findings for time since loss, Allumbaugh and Hoyt (1999) estimated that, with all other study characteristics held constant, the aggregate effect size for a similar set of studies whose participants were recently bereaved (d 0.70) would be comparable to that typically observed in psychotherapy outcome research.
Conclusions. Allumbaugh and Hoyt (1999) concluded that the relatively small aggregate effect size (d 0.43) for grief interventions in these studies “may say more about the nature of the studies than about the effectiveness of grief treatment per se” (p. 378). The findings in most of these studies were based on participants who (a) were actively recruited by researchers rather than seeking help on their own and (b) had been bereaved, on average, several years before treatment. Studies of samples more typical of actual client populations in bereavement counseling settings “are likely to be as effective [as] or possibly even more effective than psychotherapy in general” (p. 378).
Two other meta-analyses have been cited frequently in subsequent literature reviews in grief studies.
Kato and Mann (1999). Kato and Mann (1999) published a detailed review of a small number of studies (k 11) of psychological interventions for the bereaved. Presumably, the sample was small because the search procedures were limited to electronic searches with the keywords bereaved and bereavement. Thus, studies that used related terms, such as grief, griever, or grieving, were not included. In addition, the quantitative portion of Kato and Mann’s review used problematic procedures. For example, when a study included more than one dependent measure (as most of these studies did), the authors computed a separate effect size for each measure. They then treated these estimates as independent effect sizes in the meta-analysis, although they must certainly be correlated (i.e., statistically dependent). This procedure violates the independence assumption of the statistical tests in meta-analysis and (perhaps more important, for our purposes) can bias study findings, because effect sizes from studies with larger numbers of measures are arbitrarily given greater weight in computation of the aggregate effect size. So, although Kato and Mann obtained a substantially lower aggregate effect size (d 0.11) than Allumbaugh and Hoyt (1999), this smaller effect size was not based on a full review of existing studies, nor were accepted meta-analytic procedures used to analyze data from the studies that were identified.
Fortner (1999). The meta-analytic findings that have been most widely cited in the bereavement literature are those of Fortner’s (1999) unpublished dissertation. Fortner used conventional meta-analytic procedures as well as the TIDE procedure to quantify the effects of grief interventions, arriving at a very weak aggregate effect size of d 0.13 for 23 studies. Neimeyer (2000) summarized both the conventional findings and the TIDE findings in detail, and subsequent reviewers citing the TIDE findings often cited the very weak effect size as additional evidence for a pessimistic view of the effectiveness of grief counseling.
Evaluating the merits of Fortner’s (1999) meta-analytic conclusions is difficult. First, this report has never been published, and it is unclear whether it has ever been submitted for publication. In any case, the findings presented in the dissertation, although they have undergone scrutiny by dissertation committee members, have not been vetted by the more rigorous peer review process necessary for publication in a scientific journal. Analytic procedures are described sketchily, and homogeneity tests, which are important for interpreting findings and which provide the warrant for conducting moderator analyses, are not reported in the Results section.
Interpretation of moderator findings is made even more difficult because coding procedures for the moderator variables are not described, nor are statistics reported reflecting the agreement among multiple raters in the coding of studies. For these reasons, it is difficult to know what to make of the finding that grief interventions were significantly more effective for clients experiencing complicated bereavement than for so-called normal grievers. This finding has been much cited in the literature, along with the even more alarming (and inaccurate) TIDE interpretation of iatrogenic effects for nearly one in two normally bereaved clients during grief counseling (Neimeyer, 2000). However, there is no information on how complicated bereavement was coded for these studies or on whether it was reliably coded. Allumbaugh and Hoyt (1999) compared normal and “high-risk” grievers (using designations by the authors of the primary research reports) and found no difference in intervention effectiveness for these two groups. Yet, on the strength of Fortner’s (1999) conclusion and Neimeyer’s (2000) reiteration, it is now common to suggest that counseling is inappropriate and perhaps even dangerous for normally bereaved individuals.
Still, it is puzzling that the aggregate effect sizes reported by Allumbaugh and Hoyt (1999) and Fortner (1999) are rather different, especially given that search procedures for both studies appear to have been thorough and quite closely comparable. In an effort to understand this discrepancy, we examined the reference lists of the two meta-analyses and made a surprising discovery. The two reviews examined a total of 55 research reports, with only 3 studies in common between them. Thus, their samples of studies were almost completely different. This difference is partly attributable to differences in the two studies’ exclusion criteria. Fortner excluded unpublished studies and studies that did not use an experimental design (i.e., random assignment of participants to treatment and control groups). Allumbaugh and Hoyt excluded studies that did not measure pretreatment status on at least one outcome variable. On the face of it, it seems unlikely that any of these differences in study characteristics would lead to systematic differences in study effect sizes. Given the different sets of studies included in the two reviews, however, it seems that the definitive chapter on the findings of grief intervention research requires a more inclusive analysis considering both sets of studies (as well as research conducted after 1998, at least some of which is methodologically superior to much of what came before).
Grief Counseling Meta-Analyses: Patterns of Citations
Although, for the reasons just cited, an impartial reviewer might conclude that the jury is still out on the effectiveness of grief counseling, there is clearly reason for optimism. One might expect that such a reviewer would give the greatest weight to the Allumbaugh and Hoyt (1999) findings. These appeared in a respected journal published by APA and included the largest sample of studies, with carefully documented and sophisticated statistical analyses. Fortner’s (1999) findings are certainly important, especially to a reader who looks closely enough to realize that Fortner examined a substantially different set of research reports than Allumbaugh and Hoyt. However, one might prefer not to assign the same weight to his conclusions, given the lack of peer review and the omissions in reporting of results discussed above. Kato and Mann’s (1999) article includes valuable reflections on the nature of bereavement and on the strengths and weaknesses of individual studies. However, their quantitative findings would not likely be given much weight by a reviewer familiar with meta-analytic principles, because their literature search was far from exhaustive and their analyses were statistically problematic. 
Given this set of findings, how have reviewers found support for the pessimistic view of grief counseling so prevalent in the literature in recent years? Such an interpretation could only arise through authors distorting, dismissing, or even outright ignoring Allumbaugh and Hoyt’s (1999) review. Neimeyer (2000) dismissed their findings with the explanation that Allumbaugh and Hoyt “included numerous uncontrolled one-group studies that could have inflated [effect size] estimates” (p. 543) and went on to present Fortner’s (1999) findings as authoritative. This critique represents a misunderstanding of Becker’s (1988) approach to including pretreatment as well as posttreatment measures in metaanalysis.
Another approach to de-emphasizing the importance of Allumbaugh and Hoyt’s (1999) findings has been distortion or misrepresentation of their conclusions. For example, Bonanno (2004) began his summary of grief intervention research by citing “two recent meta-analyses [that] independently reached the conclusion that grief-specific therapies tend to be relatively inefficacious (Kato & Mann, 1999; Neimeyer, 2000)” (pp. 21–22), then stated that “a third meta-analytic study” (Allumbaugh & Hoyt, 1999) concluded that “grief therapies can be effective but generally to a lesser degree than . . . other forms of psychotherapy” (p. 22). Compare this with Allumbaugh and Hoyt’s own summary of their findings, cited in full above, stating that this moderate aggregate effect size effects unrepresentative sampling procedures associated with most of these studies and that more ecologically valid studies tend to find that grief interventions are “as effective [as] or possibly even more effective than psychotherapy in general” (p. 378). Lalande and Bonanno (2006) have taken this approach one step further by implying that the three meta-analyses have equivalent (and equivalently discouraging) findings: “Unfortunately, recent meta-analyses have shown that grief treatments have generally been ineffective and can possibly do harm to those who participate in them (Allumbaugh & Hoyt, 1999; Jordan & Neimeyer, 2003; Kato & Mann, 1999; Neimeyer, 2000)” (p. 303).
A final strategy for defending a pessimistic view of grief counseling has been to ignore Allumbaugh and Hoyt’s (1999) findings entirely. For example, in a review intended as a definitive synthesis to guide policy and practice in the bereavement area, the Report on Bereavement and Grief Research (CFAH, 2003) did not cite Allumbaugh and Hoyt at all and instead offered a detailed summary of Fortner’s (1999) findings (which were attributed not to Fortner but to Neimeyer, 2000).
Summary: Rebuttal of Evidentiary Basis for the New Pessimism
The recent pessimism about the effects of grief counseling is based on three interrelated claims, each of which has doubtful empirical grounding.
First Claim: A Large Proportion of Clients Are Harmed
To our knowledge, no valid method exists to quantify treatment deterioration from summary data provided in psychotherapy outcome studies. The TIDE findings for grief counseling, usually attributed to Neimeyer (2000), are actually based on a student dissertation (Fortner, 1999) that has never been published and possibly has never been subjected to peer review outside the dissertation committee. The TIDE statistical procedure used by Fortner is attributed to another student’s master’s thesis (Anderson, 1988), which is unpublished and unavailable for examination. Reviewers appointed by Gary R. VandenBos (APA publisher and the managing editor of the American Psychologist) agreed that there is no valid basis for the claim that the 38% figure represents a percentage of clients who deteriorated as a result of treatment and that the statistic is seriously flawed. In the online Appendix (available on the Web at http://dx.doi.org/10.1037/07357028.38.4.347.supp), we describe Fortner’s method in some detail and show that it cannot provide a valid TIDE estimate.
Second Claim: Effect Sizes for Grief Counseling Outcome Studies Are Zero or Trivial
The most thorough and methodologically rigorous metaanalysis to date (Allumbaugh & Hoyt, 1999) paints a relatively optimistic portrait of grief counseling outcomes but acknowledges limitations in the literature on which these conclusions are based. Still unresolved are the reasons for discrepancies between Allumbaugh and Hoyt’s (1999) findings and those of Fortner (1999), who used a largely different set of studies. Unfortunately, Fortner’s (1999) findings have never been published, and the unpublished dissertation lacks some of the information that would be helpful to address this question.
Third Claim: Grief Counseling Is Especially Ineffective
(and Especially Harmful) for “Normal” Grievers
The contention that “normal” (uncomplicated) bereavement is unresponsive to grief counseling is based on a moderator analysis in Fortner’s (1999) dissertation, as summarized by Neimeyer (2000). As noted above, Fortner (1999) did not report homogeneity tests (which normally provide the warrant for conducting moderator analyses in meta-analysis). He also did not explain how he distinguished normal from complicated grief in coding this moderator variable, nor did he report the level of interrater agreement (to show that this study characteristic, as he defined it, could be coded reliably from the information in the research reports). All of these factors make the validity of this conclusion difficult to gauge. Allumbaugh and Hoyt (1999), who coded a similar study characteristic (normal vs. high-risk grievers), found no evidence of a difference in effect size for these two groups. The related claim (Neimeyer, 2000) that nearly one in two normally bereaved clients is actually worse off following counseling is based on an application of the TIDE statistic to Fortner’s (1999) moderator analysis comparing normal and complicated bereavement. Because the TIDE statistic has been shown not to yield valid deterioration estimates, this claim lacks any empirical grounding.
Lessons From the Propagation of the TIDE Findings
Researchers are only human and are susceptible to human foibles, including cognitive biases and sloppy reasoning in the service of cherished beliefs. As Donald Campbell (1986/1999) noted, science is a reliable method of drawing valid conclusions from empirical data only to the extent that scientists evolve social structures that tend to maximize scientific validity. The widespread acceptance of the TIDE findings among researchers and practitioners interested in issues of grief and loss represents a breakdown in this validity-enhancing system and suggests lessons for both groups in the interest of creating a scientific research culture that will inform, rather than misinform, practice.
Recommendations for Researchers, Editors, and Reviewers
Cite Your Sources
Giving credit to the originators of theories or specific knowledge claims is a hallmark of scholarly conduct in all academic disciplines. In empirical disciplines, such as psychology, writers are asked to clarify whether the source is cited as an acknowledgment of an intellectual debt (to a previous theorist who has espoused a similar or identical position) or as a source of empirical evidence in support of the proposition (APA, 2001, p. 28). The former type of citation acknowledges intellectual priority; the latter type claims an empirical basis for the author’s assertion and points the reader to the evidence so that he or she can evaluate its quality and relevance. One of the challenges for readers seeking to evaluate the validity of the TIDE findings is that authors have cited Neimeyer (2000) rather than the actual reference for the research report (Fortner, 1999). This type of scholarly lapse is certainly fair game for reviewers or editors (some or all of whom have strong familiarity with the research area) as they provide comments on manuscripts submitted for publication.
Read the Sources You Cite
Authors citing a source as empirical support for theoretical claims imply that they have read and evaluated the cited article and believe that it makes a plausible case for the veracity of these claims. If Author C cites Publication A without examining it, simply because Author B (whom C trusts) has also cited A as evidence for a similar proposition, then C is participating not in the conventions of scientific scholarship but rather in a sort of grown-up version of the children’s game of “telephone.” In this game, receivers later in the chain of communication never hear directly from the source but only from those just preceding them in the chain. In the grown-up version of the game, just as in the children’s version, there is every likelihood of serious distortion of the original message.
Know Your Limitations
A pernicious factor contributing to the spread of the TIDE findings was the use of a “novel” (in fact, unpublished) statistical procedure. It seems evident that neither Fortner (1999) himself nor those citing his work (or citing Neimeyer’s, 2000, summary of his work) had sufficient grasp of meta-analytic principles to recognize the limitations of the TIDE statistic, and neither Fortner (1999) nor Neimeyer (2000) addressed these limitations so that readers could take them into consideration. The fact that the statistic itself had never been reviewed by methodologists is critical information that was omitted from all published reports of the TIDE findings. Readers, including editors and reviewers, could have noticed that the citation for the TIDE procedure was an unpublished master’s thesis, but this would have required tracing the finding back to Fortner’s (1999) work.
Recommendations for Readers
According to Campbell (1986/1999), scientific validity depends on “disputatious communities of truth-seekers” (p. 191) who keep one another honest by challenging assumptions and identifying plausible rival explanations for research findings (which qualify current findings and suggest directions for future research). According to this view, readers have a role to play in maintaining a healthy research climate. A skeptical attitude and a willingness to question when one does not understand (or when one needs more information to verify one’s understanding) are important traits for editors and reviewers of scientific manuscripts and are also important for consumers of research findings. At the heart of the scientist–practitioner model of training is the goal of empowering applied psychologists to become active and critical evaluators of the research literature.
Where Do We Go From Here?
Although they are less than 10 years old, the erroneous TIDE findings have had a major impact on both the theory and the practice of grief counseling. For practitioners, the widespread acceptance of these findings has been demoralizing. As the TIDE claims have permeated mental health agencies, hospices, and other funding sources, these parties have become skeptical of the value of grief counseling (and concerned about potential harm to clients). The TIDE claims have also been repeated in the popular media (“Grieving,” 2003; Sommers & Satel, 2005; “Therapy”, 2005), and the general pessimism about the efficacy of grief counseling (which we believe has been fueled in large part by alarm about these claims) has been reflected in major media articles as well (Brody, 2004). These echoes in the mainstream media of the unwarranted pessimism that has flourished in the scientific journals could foster negative impressions of these interventions in the general public and discourage distressed bereaved persons from seeking needed counseling.
To begin to reverse the damage and to minimize the future impact of these claims, we recommend three corrective actions. First, to restore the integrity of the scientific study of grief counseling, it is desirable for the principal journals responsible for propagating these findings to publish retractions in print from proponents of the unfounded conclusions. Public retractions, pointing authors and reviewers to criticisms of the TIDE statistic, will help to prevent future citations of the TIDE claims in professional journals. Second, the TIDE episode provides an opportunity to effect on scientific standards in psychology. It is likely that the TIDE claims would not have attained credibility without laxness on the part of authors and editors in standards for critically evaluating and referencing cited work. We have suggested several lessons for authors, editors, reviewers, and even readers of scientific reports, designed to reduce the probability that erroneous findings will be published and repeatedly cited in the future. Finally, to counteract the potential misinformation about TIDE findings from the popular media, it is important to find mechanisms to educate the general public about the problematic nature of these claims.
In conclusion, let us emphasize the good news growing out of our investigation of the basis for pessimistic claims about the efficacy of grief counseling: There is no empirical or statistical foundation for these claims. There is no evidence that bereaved clients are harmed by counseling or that clients who are “normally” bereaved are at special risk. There is not even any strong evidence that grief counseling, as typically practiced, is less efficacious than other forms of counseling and psychotherapy (which have been shown to have strong positive outcomes for many forms of psychological distress; Lipsey & Wilson, 1993; Wampold, 2001). More work can be done to clarify the implications of existing studies (some of which have been conducted since the publication of the most recent meta-analysis of grief counseling), but findings to date indicate that cautious optimism, rather than the recently fashionable dire pessimism, is the attitude most congruent with empirical findings on grief counseling outcomes.
 Technically, the effect sizes for Allumbaugh and Hoyt (1999) were reported as Becker’s (1988) , not Cohen’s (1988) d. However, interpretation of these two parameters is similar, and the two analyses will yield identical results when treatment and control groups have equivalent scores on the outcome measures prior to treatment. (For nonequivalent group designs, Becker’s method is more accurate.) Studies included in the other two meta-analyses randomly assigned participants to treatment and control groups (which is expected to equalize pretreatment means). So, for simplicity, we report Allumbaugh and Hoyt’s findings as d statistics to avoid burdening readers with this technical issue and to emphasize their comparability with the effect sizes from the other meta-analyses.
 As in the case of the TIDE findings, these citations have attributed Fortner’s (1999) conventional meta-analytic findings to Neimeyer (2000) or later authors, not to Fortner (1999) himself.
 Although Fortner (1999) located and analyzed 23 studies for the conventional meta-analysis, only 10 of these studies provided the data needed to compute the TIDE statistic, as described earlier.
 One difference is that Allumbaugh and Hoyt (1999) sought unpublished as well as published studies, whereas Fortner (1999) ignored unpublished research. This difference does not account for the effect size discrepancy, however, because Allumbaugh and Hoyt’s effect size for the subset of published studies was d 0.39 (see the authors’ Table 1, p. 373), still much larger than that obtained by Fortner.
 Occasionally, reviewers have alluded to an additional line of evidence on the ineffectiveness of early intervention for trauma. For example, Bonanno et al. (2005) cited McNally, Bryant, and Ehlers’s (2003) review of psychological debriefing interventions for trauma survivors (as well as Bonanno, 2004, and Jordan & Neimeyer, 2003) for the claim that “indiscriminately encouraging bereaved or traumatized individuals to participate in grief counseling or other forms of crisis intervention is not only ineffective but can actually be harmful” (p. 841). This type of argument seems to conflate two very different populations (bereaved individuals seeking counseling and persons who have suffered trauma and may be mandated by employers—see Groopman, 2004, e.g.—to participate in a “stress debriefing”) and different interventions. We do not believe that research on trauma debriefing as a preventative against development of posttraumatic stress disorder is informative about the effectiveness of grief counseling as typically practiced.