UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Counterfactual thinking in the wake of trauma Davis, Christopher G. 1995

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1995-059448.pdf [ 331.63kB ]
JSON: 831-1.0088181.json
JSON-LD: 831-1.0088181-ld.json
RDF/XML (Pretty): 831-1.0088181-rdf.xml
RDF/JSON: 831-1.0088181-rdf.json
Turtle: 831-1.0088181-turtle.txt
N-Triples: 831-1.0088181-rdf-ntriples.txt
Original Record: 831-1.0088181-source.json
Full Text

Full Text

COUNTERFACTUAL THINKING IN THE WAKE OF TRAUMA by CHRISTOPHER G. DAVIS B.Sc, Trent University, 1988 M.A., University of British Columbia, 1991 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES DEPARTMENT OF PSYCHOLOGY WE ACCEPT THIS THESIS AS CONFORMING TO THE REQUIRED STANDARD THE UNIVERSITY OF BRITISH COLUMBIA July, 1995 (c) Christopher G. Davis J I^HS In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis "for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. (Signature) Department of The University of British Columbia Vancouver, Canada DE-6 (2/88) ii Abstract Counterfactuals generated by people who have experienced traumatic life events were examined to elucidate their significance for the coping process. In Study 1, 93 respondents were interviewed 4-7 years after the loss of their spouse or child in a motor vehicle accident. In Study 2, 124 respondents were interviewed 3 weeks and 18 months following the death of their child to Sudden Infant Death Syndrome. Across these two studies it was found that (a) counterfactuals that undid the traumatic event were commonly reported; (b) the focus of counterfactuals was typically on one's own (in)actions, rather than on the behavior of others; (c) the more freguently respondents were undoing the event, the more distress they reported; and (d) this relation held even after controlling for more general ruminations. In Study 3, 106 respondents were interviewed one week following their spinal cord injury. In this study, self-implicating counterfactuals were shown to predict ascriptions of self-blame, controlling for causal attributions and foreseeability estimates. Taken together, these field data suggest that counterfactuals play an important role in how people cope with traumatic life events. Possible roles that these counterfactual thoughts might play are discussed. iii Table of Contents Abstract ii Table of Contents iii List of Tables v Preface vi Chapter 1: Introduction 1 Distinguishing Undoing from Other Cognitive Processes 4 Ruminations 4 Causal Attributions and Ascriptions of Blame . . . . 6 The Present Studies 14 Chapter 2: Study 1 17 Method 18 Respondents and Procedure 18 Response Rates 19 Interview Instrument 2 0 Psychological Distress 20 Ruminations 21 Undoing 22 Attributions of Blame 2 3 Results 24 Sample Demographics 24 Prevalence of Undoing 25 Undoing and Distress 27 Content of Undoing Thoughts 28 Blame for the Accident 30 Discussion 30 Chapter 3: Study 2 36 Method 36 Respondents and Procedure 3 6 Response Rates and Attrition 37 Interview Instrument 38 Psychological Distress 38 Undoing 39 Causal Attributions and Judgments of Responsibility!^ Ruminations 41 Results 41 Sample Demographics 41 Mutability of Events Leading up to SIDS 43 Content of Undoing Thoughts 44 Stability of Counterfactuals over Time 46 Undoing, Mutability, and Distress 47 Undoing and Specific Components of Distress . . . . 49 Undoing at 18 Months Postloss 50 Undoing, Causality, and Responsibility Attributions 53 Discussion 54 iv Chapter 4: Study 3 59 Method . 62 Respondents 62 Procedure 63 Interview Instrument 63 Ratings by Coders 65 Results 67 Sample Demographics 67 Counterfactual Descriptives 69 Behavioral Freedom, Enjoyment, Typicality of Pre-accident Activity, and Injury Foreseeability 69 Relations Between Causal Attributions, Counterfactual Thoughts, and Blame Assignments . 70 Respondents Compared with Raters: Causes and Blame . 72 Predicting Self-Blame from Causal Attributions, Foreseeability, and Counterfactual Thoughts . . . 74 Discussion 76 Limitations of Study 3 80 Chapter 5: General Discussion 83 Future Directions 86 Conclusion 92 References 95 • V List of Tables Table 1: Study 2: Correlations between undoing frequencies and negative affect, 3 weeks and 18 months postloss . 48 Table 2: Study 2: Regression of undoing frequencies on guilt and other ABS subscales 51 Table 3: Study 2: Regression of undoing frequency at 18 months postloss on antecedent mutability, SCL distress, and undoing at 3 weeks postloss 52 Table 4: Study 3: Correlations between respondents' counterfactual thoughts, causal attributions, and blame assignments 71 Table 5: Study 3: Respondents versus raters on cause, foreseeability, and blame 73 Table 6: Study 3: Regression of respondents' self-blame on their causal ascriptions, foreseeability estimates, and counterf actuals 75 Table 7: Study 3: Regression of respondents' causal, counterfactual, and foreseeability ratings on rater-respondent self-blame discrepancy 77 vi Preface Studies 1 and 2 have been published under the title, "The undoing of traumatic life events" in Personality and Social Psychology Bulletin, 21 (1995; pp. 109-124). The article was authored by C. G. Davis, D. R. Lehman, C. B. Wortman, R. C. Silver, and S. C. Thompson. This material is included with permission of Sage Publications Ltd. All writing and analyses were conducted by the author of this dissertation, under the supervision of D. R. Lehman. The collection of data reported in Study 1 was conducted by D. R. Lehman and C. B. Wortman, and was funded by a grant from the (U.S.) Insurance Institute for Highway Safety. The collection of data in Study 2 was conducted by C. B. Wortman and R. C. Silver, and was funded by a grant from the U.S. Public Health Service Bureau of Community Health Services (Maternal and Child Health Research). Aside from their roles in the collection of data, the last three authors' contributions were limited to comments on drafts of the article. Study 3 has been accepted for publication in Personality and Social Psycholoay Bulletin under the title "Self-blame following a traumatic event: The role of perceived avoidability," and is authored by C. G. Davis, D. R. Lehman, R. C. Silver, C. B. Wortman, and J. H. Ellard. All writing and analyses were conducted by the author of this dissertation, under the supervision of D. R. Lehman. The collection of data in Study 3 was conducted by R. C. Silver and C. B. Wortman, and was supported by a (U.S.) National Science Foundation grant. The vii coding of data by raters in Study 3 was supervised and coordinated by J. H. Ellard. Permission to include material from Study 3 has been obtained from each of the authors. The opinions, findings, and conclusions expressed in this dissertation do not necessarily reflect the views of the funding agencies. Darrin R. Lehman, Supervisor 1 Chapter 1 Introduction More than three decades ago, Heider (1958) observed that a person who narrowly misses obtaining a desired outcome, even though it is randomly caused, often becomes preoccupied with the wished-for reality and frustrated by its denial. Heider suggested that the closer one comes to attaining a desired outcome, the easier it is to imagine its achievement, and thus the more apparent is the failure to attain it. For example, Heider predicted that a person who misses winning a lottery by one digit would be more preoccupied with the failure to win, and hence would feel greater frustration, than the person whose ticket was several digits away from the winning number. Subsequent research has supported this prediction (Kahneman & Tversky, 1982; Wells, Taylor, & Turtle, 1987). Kahneman and Tversky (1982) demonstrated that when people are told of undesirable events that have befallen others, they tend to simulate mentally the events leading up to the outcome, desiring to "undo" events that appear to play a role in producing the outcome. According to these authors (see also Kahneman & Miller, 1986; Miller & Turnbull, 1990; Miller, Turnbull, & McFarland, 1990), people often compare the reality of the event with hypothetical alternatives by thinking "if only the person had (or had not) done this, it would not have happened." Thus, undoing thoughts, or "counterfactuals," are concerned with how a negative outcome could have been avoided. 2 This research suggests that the extent to which an outcome will be undone depends on the perceived mutability of the antecedent events and circumstances. That is, if incidents in the sequence of events leading up to the outcome are perceived to be immutable, or unchangeable, then research indicates such events will not be targets for undoing (Kahneman & Miller, 1986; Miller & Turnbull, 1990; Wells et al., 1987). In effect, one simply accepts the reality of the occurrence. On the other hand, if an incident in the sequence is perceived to be highly mutable (i.e., it can easily be imagined otherwise), then such an incident is likely to be shifted to its alternative in a mental simulation, thereby undoing the event. Given that the undesirable outcome has, in fact, occurred, this process of mental simulation is thought to amplify negative affect over "what might not have been." This suggests that events that are easier to undo are more distressing because they are likely to leave one thinking that the event need not have happened. According to laboratory research on counterfactual thinking, two factors that affect the mutability of an outcome are: (a) the exceptionality (vs. normality) of the antecedent, and (b) whether the outcome was a result of someone's actions (vs. a failure to act). It has been demonstrated that outcomes that follow exceptional (and thus "abnormal") sequences are more likely to be imagined otherwise than are outcomes of routine sequences (Kahneman & Miller, 1986; Kahneman & Tversky, 1982; Miller & McFarland, 1986; Wells et al., 1987). In addition, research has 3 shown that it is easier to imagine the target avoiding the occurrence of a negative event by not acting (i.e., undoing a committed act) than it is to imagine the target acting when in reality no action was taken (i.e., undoing an omission; e.g., Gleicher et al., 1990; Landman, 1987; Miller et al., 1990; but see Gilovich & Medvec, 1994). Virtually all of the evidence collected in support of these claims has been laboratory-based; all but a few of the studies (e.g., Markman, Gavanski, Sherman, & McMullen, 1993) have employed role-playing subjects responding to hypothetical scenarios.1 In these studies, subjects merely imagine how the "target" victim might think or feel; they are unlikely to experience the distress and specific cognitions that such events typically evoke in individuals who have actually encountered them. As the ultimate goal for most counterfactual researchers is to understand what the undoing process is like in the real-world (e.g., Miller & Turnbull, 1990), it seems important to examine these issues with people who have actually experienced significant life events. It is possible that the counterfactual thoughts of people coping with real-life traumatic events may differ significantly from those of role-playing subjects. *To my knowledge, only one previous study (Bulman & Wortman, 1977) has described data on the effects of undoing a traumatic life event that subjects have actually experienced (see Miller & Turnbull, 1990, and Wells & Gavanski, 1989, for discussions of this study). Anecdotal evidence from Bulman and Wortman suggested that some of their spinal-cord injured respondents reported wishing they had done something differently immediately before their accident. Bulman and Wortman noted that these respondents were among the worst copers in their study. 4 Whereas role-playing subjects clearly assign significance to the degree of mutability (e.g., the extent to which an antecedent was unusual), those who have experienced traumatic events, motivated by the significance of their situation, might be relatively less influenced by subtle differences in degrees of mutability. Instead, they may be relatively more influenced by whether an antecedent event was changeable from their perspective in any conceivable way. The present research sought to extend the literature on undoing by considering counterfactuals outside the laboratory, employing as respondents people coping with traumatic life events. Based on the hypothetical scenario research, I anticipated that counterfactuals would be commonly reported among individuals who experienced such events. Based on this literature, I further anticipated that undoing would be associated with greater negative affect and distress. I also sought to investigate whether unusual circumstances or committed acts were more likely to be undone, and more highly related to distress, relative to usual circumstances or omitted acts. Distinguishing Undoing from Other Cognitive Processes Ruminations. It is important conceptually to distinguish the notion of undoing from similar constructs in related literatures. One related construct to undoing is rumination, or repetitive and intrusive mental images of the traumatic event (Tait & Silver, 1989). A long tradition of psychiatrists and clinical and health psychologists has noted that ruminations are 5 a common response to traumatic experiences, even long after the event (e.g., Freud, 1920/1963; Janis, 1971; Tait & Silver, 1989; Weiss, 1988). Horowitz (1976), for example, has argued that following trauma, people tend to oscillate between periods in which mental images associated with the trauma involuntarily flood one's mind and periods in which memories of the trauma are repressed or denied. Although the periods of event-related ruminations are thought to be highly distressing, Horowitz suggests that by replaying the details of their traumatic episode the individual is performing the important cognitive task of coming to terms with, or working through, the event (see also Weiss, 1988). Much of the research in this literature has focused on ruminations in general (and their significance for the adjustment process) rather than on the specific content of such thoughts. That' ruminations have been so broadly construed (e.g., any thoughts or images that are repetitive) might explain why some theorists (e.g., Horowitz, 1976) have suggested that they serve an adaptive function, whereas others (e.g., Wortman & Silver, 1987) have failed to support this, arguing instead that more frequent ruminations in the initial weeks following trauma tend to be associated with greater long-term distress. It may be that by drawing distinctions among ruminative thoughts according to their content, researchers will be better able to explicate the various roles that the specific ruminations serve. 6 Such a distinction may be drawn between counterfactuals and other ruminative thoughts. A review of the literature on ruminative thoughts following traumatic events suggests that a number of examples cited by researchers can be classified as counterfactuals (see e.g., Tait & Silver, 1989; Weiss, 1988). Although counterfactuals might be characterized as ruminative in nature (i.e., that they are intrusive, involuntary, and repetitive), they are distinct from general ruminations in substance. Specifically, when undoing, individuals are not reviewing the event as it actually occurred (Horowitz, 1976), but rather are simulating it as it might not have happened. Thus, undoing represents a state of continual upward comparison between reality and the better hypothetical alternative in which the negative outcome does not occur (Miller & Turnbull, 1990). It is noteworthy that when people create hypothetical, better scenarios, their actual outcomes appear especially poor by comparison, and hence when one undoes a traumatic outcome, one is likely to feel worse. Causal Attributions and Ascriptions of Blame. Recent research has drawn attention to the similarities between the processes of undoing and making causal inferences. For example, Wells and Gavanski (1989) argued that much of the evidence cited in support of an undoing process may also be explained by invoking well-accepted attribution models. Lipe (1991), in contrast, argued that a counterfactual model can fully account 7 for evidence used to support various attribution models (see also Hilton, 1990). Such interpretations suggest that counterfactual thoughts are mere tools for determining cause. If mutating an antecedent undoes the outcome, then it may be deemed causal (Roese & Olson, in press). Although counterfactual reasoning may underlie the causal attribution process, I will argue that not all counterfactual thoughts imply a causal attribution. In my view, the two processes are related but nevertheless distinct. The extent to which the two processes can be distinguished may depend on the relative mutability of the causes identified in the attribution exercise. Specifically, to the extent that a sufficient cause is perceived to be highly mutable, it is likely to be a target for undoing. In such a case, the attribution of causality and the content of the counterfactual may be identical (Lipe, 1991). However, counterfactuals are not necessarily limited to antecedents that are causally significant. For instance, the two processes may be distinguished when the cause identified by a causal attribution is not perceived to be as mutable as another antecedent. In such cases, one's attribution of causality may implicate (from the actor's point of view) some relatively immutable behavior (e.g., the actions of a thief), whereas the counterfactual process may implicate a much more mutable behavior (e.g., one's failure to bolt one's windows) — one that could have "prevented" (but did not cause) the event's occurrence. In essence, the distinction is between the 8 perception that a given behavior was instrumental in causing an outcome to occur and the perception that one merely could have prevented or avoided the outcome from occurring. The distinction between cause and avoidability is important to the extent that both may independently influence how people assign blame. Although it is often assumed that a causal attribution is a necessary and sufficient precondition for the assignment of blame and responsibility (and hence that "cause" and "blame" are virtually interchangeable terms; e.g., Frazier, 1990; Tennen, Affleck, & Gershman, 1986; Turnquist, Harvey, & Andersen, 1988), some theorists have argued that this assumption is not valid (Fincham & Jaspars, 1980; Shaver, 1985; Shaver & Drown, 1986). For instance, people are occasionally assigned responsibility when they were not causally implicated (e.g., when a parent assumes responsibility for a child's actions) or, conversely, are not blamed when they accidentally or unintentionally caused misfortune. As Fincham and Jaspars (1980) argued, whereas a causal attribution may satisfy one condition for a judgment of responsibility to be made, it is neither a necessary nor sufficient criterion. Shaver (1985) has developed a prescriptive model of the assignment of blame. According to his model, increasing degrees of moral responsibility are assigned following a causal attribution on the basis of fulfillment of the following criteria: that the causal agent (a) was aware of the potential consequences (i.e., foreseeability), (b) intended the 9 consequences, (c) acted freely, and (d) understood the moral wrongfulness of his or her acts. Less responsibility is ascribed if only one of the criteria is met and most responsibility is assigned when all four criteria are met. Shaver further suggested that blame is only assigned when the perceiver does not accept the excuse or justification of the agent responsible for the outcome.2 Although neither Shaver (1985) nor Fincham and Jaspars (198 0) suggested that blame may be ascribed on the basis of avoidability (except to the extent that avoidability implies foreseeability), a number of scenario studies in the counterfactual literature suggest that counterfactuals may influence ascriptions of blame (Kahneman & Miller, 1986; Miller & Gunasegaram, 1990). For example, Miller and Turnbull (1990) asked half of their subjects to estimate how much self-blame a victim would experience following an out-of-character, and thus highly mutable, action and the other half to estimate how much self-blame a victim would experience following a routine action. The former subjects felt that a person shot when he walked in on a robbery in a store he rarely frequented would experience greater self-blame than did the latter subjects responding to a 2Shaver argued that although various sources (e.g., people, weather) may be deemed causally significant, only people may be deemed responsible, and hence blameworthy. This is because non-human entities do not fulfill any of Shaver's criteria for culpability. Thus, outcomes due to mechanical failure or inclement weather would not result in a blame assignment to those sources. Blame and responsibility, however, may be assigned to technicians or officials who, for example, may have failed to maintain equipment (insofar as the outcome was deemed foreseeable). 10 scenario wherein a person was shot in a store he normally frequented. This was the case even though it was clear to both groups of subjects that neither victim caused his injury. Miller and Turnbull suggested that such presumed self-blame would occur in the former case because it would be very easy for that victim to imagine going to his usual store, and hence avoiding the shooting. The ease with which this counterfactual would come to mind would lead him to believe that he ought to have gone to the usual store. That is, the victim will perceive that, by acting out-of-character, he has brought on the tragedy himself, and hence will feel blameworthy. In contrast, the latter victim, because he did not do anything out of the ordinary, presumably can only feel that he might have gone to a different store. The highly mutable condition of the first victim's scenario apparently leads him to confuse what might have been the case with what ought to have been the case. Miller and Turnbull term this the counterfactual fallacy. Coming from a slightly different perspective, Wells and Gavanski (1989) suggest that counterfactuals influence blame assignment via their effect on causal ascriptions. They contend that people assign causal significance, in part, by considering the likelihood of the counterfactual that would undo the outcome. They demonstrated this in a scenario study involving a cab driver who refused to pick up a disabled couple. Subjects were told that following the cab driver's refusal, the couple decided to drive their own car, and were injured when they drove off a 11 collapsed bridge. Subjects were requested to rate the extent to which the cab driver caused, and was responsible for, the accident. Importantly, from a counterfactual perspective, causality and responsibility ratings assigned to the cab driver were attenuated for other subjects who were told that the cab driver also drove off the bridge. This suggests that when the counterfactual, "if only he had picked up the couple," failed to undo the outcome, the cab driver was seen as less causally significant, and was blamed less, even though his behavior was the same.3 Thus, there is some empirical support for the idea that people may assign responsibility or blame on the basis of a counterfactual, although it is unclear whether the effect is direct (as Miller & Turnbull's [1990] data suggest) or via the counterfactual's effect on one's causal attribution (Wells & Gavanski, 1989). This evidence may be limited, however, in that it concerns the perceptions of observers. That is, subjects in these studies either imagined how much they would blame themselves or were asked to judge others' blameworthiness. And, as considerable attribution research suggests that actors and observers may diverge considerably in their perceptions of cause and blame (e.g., Jones & Nisbett, 1972; for a review see Turnquist et al., 1988), it is important to assess whether 3Wells and Gavanski (1989) suggest that counterfactuals influence responsibility ratings via their effect on causal attributions (see also Lipe, 1991). Interestingly, however, the effects they report for responsibility ratings are larger than the effects they report for causal ratings. 12 counterfactuals influence ascriptions of self-blame among people actually experiencing the negative events. In fact, six studies employing as respondents people coping with traumatic life events suggest that counterfactual thoughts (or perceptions of avoidability) may be related to judgments of self-blame (Abbey, 1987; Branscombe, Owen, & Allison, 1994; Bulman & Wortman, 1977; Frazier, 1990; Kiecolt-Glaser & Williams, 1987; Schulz & Decker, 1985). All of these studies indicate that the more individuals perceived that they could have avoided their victimization, the more they blamed themselves. Unfortunately, none of these studies controlled for causal attributions or outcome foreseeability, thus leaving ambiguous the distinction between causal attributions and counterfactuals that imply outcome-avoidability, and thus cannot speak to the issue of whether counterfactuals relate to ascriptions of blame directly or via their effects on causal attributions. In sum, it is not denied that counterfactual thoughts may influence, or are influenced by, causal ascriptions; rather, it is suggested that distinctions can be made between the two processes, and that both may independently lead to ascriptions of blame. Although it is likely true, as Hilton (1990) and Lipe (1991) suggested, that the initial stages of the counterfactual process may be employed to identify possible causes (e.g., if antecedent X is absent, does the outcome still occur?), the means by which these antecedents are weighed, relative to other elements in the set of possible causes, may be different. 13 Specifically, in an attribution of causality, the relative importance of each possible cause is evaluated with reference to the extent to which it was a necessary, but also sufficient, condition (either alone or in conjunction with other conditions) to produce the particular outcome observed (Shaver, 1985). In contrast, the counterfactual process evaluates antecedents in terms of the relative ease with which they could have been undone or avoided (i.e., their mutability). Thus antecedents that are unusual, or personally enacted, likely will be grist for the counterfactual mill, but will tend to be overlooked in the causal attribution process if they are not part of a minimally sufficient subset of possible causes. For example, if one is struck by a drunk driver, one will likely claim that the drunk driver caused the accident, yet still perceive that the accident could have been avoided had one taken a different route, left earlier, etc. Miller and Turnbull's (1990) analysis of the counterfactual fallacy suggests that to the extent that counterfactuals which implicate one's behavior are salient, one is likely to perceive that one ought to have been able to avoid or prevent the unfortunate outcome, and thus may be partly responsible for it. This process may explain why people who have experienced traumatic events, such as people who have been raped, assaulted, or criminally victimized, or people coping with cancer or the loss of a loved one, often assume blame or responsibility for their victimization despite having done little, if anything, to 14 have caused it (see e.g., Abbey, 1987; Abrams & Finesinger, 1953; Bard & Dyk, 1956; Bard & Sangrey, 1979; Burgess & Holmstrom, 1974; Chodoff, Friedman, & Hamburg, 1964; Janoff-Bulman, 1979; Mechanic, 1977; Medea & Thompson, 1974; Taylor, Lichtman, & Wood, 1984) . The Present Studies The purpose of the research to be described is to evaluate the significance of counterfactual thoughts for people coping with traumatic life events. As these studies are the first to systematically examine counterfactual thoughts among people coping with such significant events, it will be important to demonstrate that counterfactuals are commonly reported in such contexts. Second, this research will explore the extent to which counterfactual thoughts relate to feelings of distress. Third, I will consider the role of mutability factors (such as antecedent exceptionality and omitted vs. committed behaviors) in the undoing process. Finally, I will attempt to distinguish (a) undoing from more general ruminative thoughts and (b) counterfactuals that imply outcome-avoidability from causal attributions, and show why such discriminations may be useful. In Study 1, I examine counterfactuals generated by bereaved adults who had lost a spouse or child in a motor vehicle accident. In all cases, neither the bereaved respondent nor the deceased were deemed legally culpable. Indeed, in most cases, the bereaved person was far removed from the scene of the accident. In many of the cases, the cause was clear: the driver 15 of the other car was driving too fast, or was drunk. In Study 2, I examine counterfactuals reported by parents who recently had lost an infant to Sudden Infant Death Syndrome (SIDS). As, prior to the study, parents were told that their baby had died of SIDS (with no known cause), by implication, they had been vindicated of responsibility for the death. Nevertheless, because the cause of SIDS remains unknown, parents in Study 2 were asked if they held any theories or hunches about what might have led their baby to die. In Study 3, I test the hypothesis that self-implicating counterfactual thoughts are implicated in ascriptions of self-blame, independent of causal attributions. This hypothesis is examined in the context of people coping with a spinal cord injury. Unlike the first two studies, in which it was clear to respondents that they did not cause the event, the causes of the respondents' spinal cord injuries in Study 3 followed from a wide range of activities; some were caused by the respondents' own actions, whereas others were caused by the behavior of others, or were due to situational or chance factors. Thus, the data in Study 3 permit the examination of counterfactual thoughts over a broad array of causes. Although assignment of respondents to mutability circumstances (e.g., the exceptionality vs. normality of the antecedents) is impossible in studies of this kind, the data that these designs provide are arguably richer in other respects. Not only do such field studies enable investigation of the findings 16 generated from lab research in ecologically-valid settings, but they may also be useful in identifying other important forces or parameters that might intensify or attenuate the undoing process or the effects of counterfactual thoughts. 17 Chapter 2 Study 1 The bereaved in Study 1 were interviewed 4-7 years following the loss of their spouse or child in a motor vehicle accident. In the interview, the respondent's current level of distress, as well as past and current emotional and cognitive responses, were assessed. Respondents were asked open- and closed-ended questions regarding ruminations and attempts to undo the event. The latter permitted an assessment of the pervasiveness of past (reported retrospectively) and current undoing, as well as the content of these attempts. By obtaining data on both general ruminations and undoing, it was possible to distinguish the unique association of each with distress. I anticipated that undoing the sudden, unexpected death of a loved one would be associated with increased distress, and that this relation would hold even after controlling for general ruminative thoughts. The open-ended data also allowed me to examine (a) the nature of the events that the bereaved were undoing in terms of previously identified mutability factors, and (b) whether such mutability distinctions were related to the levels of distress respondents reported. Would mutability factors found to influence undoing in the lab do so in the context of actual coping with a traumatic life event? Although it would have been ideal to consider these mutability factors for all respondents (whether or not they reported undoing), this was not possible in Study 1. Even though 18 all respondents were asked to describe the accident, their descriptions were typically general and did not lend themselves to mutability coding. Respondents often described the accident in terms of location, time of day, and a factual account of the event itself (rather than the events leading up to the accident or whether these antecedent events were usual or unusual, etc.). Thus, with Study 1, I am unable to address the question, "Are outcomes with exceptional events preceding them more likely to be undone?" The questions I can address, however, concern the extent to which people coping with traumatic life events undo their loss, and whether previously identified mutability factors appear to influence the undoing process in the "real world." Method Respondents and Procedure The respondents in Study 1 were randomly selected from a microfilm file that contained a complete list of all motor vehicle fatalities in Wayne County, MI that had occurred 4-7 years previously. Records from the state of Michigan were used to screen all accidents and select as respondents only those whose spouse or child was an innocent victim (i.e., had died in a crash in which the driver of their motor vehicle was not responsible). Police records for each crash were examined, and with use of the criteria developed by Haddon (1963), determinations of probable responsibility or fault were made on the basis of movements of the vehicles prior to the crash. In this scheme, the 19 driver/vehicle initiating the crash was identified as responsible (e.g., a car crosses the center line and collides with an oncoming car). The classifications were based on information from the diagram and narrative descriptions of the crash recorded by the investigating police officer. Responsibility, as determined by this scheme, does not necessarily indicate driver culpability or legal responsibility. However, in none of the cases included in the sample was the driver of the vehicle occupied by the deceased cited for a violation. An attempt was made to locate the next of kin of all randomly selected fatalities. If the deceased was married at the time of the accident, his or her spouse was asked to participate in the study. If the deceased was less than 18 years of age and still residing with his/her parents at the time of the accident, both parents were asked to participate. A letter describing the study was sent to potential respondents, and they were informed that an interviewer would be contacting them in the near future to set up an appointment. If the respondent agreed to participate, an interview was scheduled in the respondent's home. Response Rates Fifty-one percent of the randomly selected bereaved agreed to participate. This response rate compares very favorably with other studies involving bereaved respondents (see Stroebe & Stroebe, 1989, for a review). In sum, 40 bereaved spouses and 53 bereaved parents (including 17 cases in which both parents agreed 20 to participate) were interviewed by trained interviewers from the Survey Research Center at the University of Michigan. Whereas all of the deceased adults were occupants of a motor vehicle at the time of the accident, some children died as pedestrians (n = 13) or bicyclists (n = 11). Interview Instrument Psychological distress. The interview instrument included previously validated self-report scales of psychological distress. As the goal of this study was to determine the relation between undoing and general distress in terms of the individual's recovery from bereavement, and not to determine specific pathological or social impairments, scales that Lehman, Wortman, & Williams (1987) found to differentiate the bereaved sample from a matched comparison sample were standardized and combined to yield an overall index of distress. The scales that were combined to form the distress index included the global severity index of the SCL-90-R (Derogatis, Rickels, & Rock, 1976), which assesses various psychiatric symptoms including depression and anxiety; a shortened 10-item version of the Center for Epidemiological Studies Depression Scale (CES-D; Radloff, 1977), which focuses on feelings of depression experienced in the past two weeks; a 6-item depression/suicide index from the Schedule for Affective Disorders Scale (SADS; Spitzer & Endicott, 1975); a 9-item version of the Bradburn Affects Balance Scale (Bradburn, 1969) to assess respondents' outlook on life; the parental and marital 21 stress subscales of the Pearlin and Schooler (1978) Role Strain Scales; and a 3-item Worry Scale, developed for this project to assess respondents' apprehensiveness about bad things happening to them or their family members in the future.4 Principal component analysis of the index yielded a single factor with all component loadings greater than .69. These data suggest that the index reflects a general level of distress (higher positive values denote greater distress). Cronbach's alpha for the index, calculated from the seven subscales listed above, was .87. To confirm that undoing was not simply another measure of distress, the principal component analysis was repeated, adding the frequency of undoing question (described below). In this analysis, two clear components emerged (using both a scree test and the eigenvalues > 1 rule), with the second factor containing only the undoing item. Ruminations. Respondents were asked about current ruminations of the deceased with the question, "In the past month, have memories, thoughts, or mental pictures of your (child/spouse) come into your mind?" Responses were made on a five-point scale ranging from No. never (1) to Yes, all the time (5). Respondents who were widowed and not remarried at the time of the interview were not administered the Pearlin and Schooler Marital Stress subscale. Similarly, respondents who were childless at the time of the interview were not administered the Pearlin and Schooler Parental Stress subscale. For this reason, the distress index is presented as a mean of the standardized scores (M = 0, SD = 1.0) for the scales completed by the respondent (minimum 5 of 7 scales). 22 Respondents were also asked if they ever reviewed in their mind the events that led up to the accident (i.e., ruminations of the event), and if so, what they had thought about in particular. Affirmative responses were later analyzed to determine the frequency with which respondents spontaneously mentioned trying to undo the event, and whose actions they reported trying to undo. Two independent raters agreed on 88% of codable responses. (All of the coding discrepancies in this research were resolved by the author and the thesis supervisor.) To assess the frequency of current ruminative thoughts of the event, respondents were asked how often they found themselves reviewing the events that led up to the accident in the past 3 0 days on a scale from Never (1) to All the time (5). Undoing. To quantify undoing, respondents subsequently were asked if they ever found themselves thinking "If only I had done something differently, my (child/spouse) would still be alive" (Yes vs. No). In affirmative cases, respondents were asked to describe their thoughts. These responses were later coded with respect to whether the respondent reported wishing to change a usual or an exceptional behavior. Two independent judges agreed in their classification in 76% of cases. The open-ended responses were also content analyzed with respect to whether the respondent reported wishing that s/he had acted or had not acted in such a way as to avoid the accident (i.e., omission vs. commission). A review of the data, however, indicated that many respondents reported wishing both that they 23 had and had not acted, even in the same sentence. Moreover, many of the counterfactuals that were stated as omissions could be interpreted as commissions, and vice versa. For example, some bereaved parents reported wishing that they had told their child not to play in the front yard (or near the road). This could be interpreted as wishing one had acted (e.g., "I should have said 'Don't play there'") as easily as it could be interpreted as wishing one had not acted (e.g., "I should not have let her play there"). To overcome this ambiguity, responses were coded literally into four categories: omissions (i.e., wished had acted), commissions (wished had not acted), both, and ambiguous. Thus, the above example was coded an omission. Using this literal coding scheme, two independent judges agreed in 75% of the cases. Finally, to assess the frequency of current undoing thoughts, those who reported ever thinking "if only I had done something differently, my (child/spouse) would still be alive" were asked, on a five-point scale, how often during the past month they had thought "if only..." (Never [1] to All the time [5]). Attributions of Blame. Respondents were asked if they blamed anyone or anything for the accident. If respondents reported that they blamed someone or something for the accident, they were asked to explain who or what they blamed. These responses were later coded into mutually exclusive categories (i.e., other driver, self, society/justice system, and 24 miscellaneous). Two independent raters agreed in 94% of the cases. Results Sample Demographics Consistent with U.S. statistics (Baker, O'Neill, & Karpf, 1984), 85% of the respondents who had lost a spouse in a motor vehicle accident were female. Of the bereaved parents, 63% were mothers and 37% were fathers.5 The mean age of the bereaved spouses and bereaved parents was 43.5 and 41 years, respectively; mean years of education for bereaved spouses and bereaved parents was 12.5 and 12.1, respectively. Of the bereaved spouses, 85% were White and 15% were Black. Of the bereaved parents, 78% were White and 22% were Black. The mean income of bereaved spouses and bereaved parents (in 1982) was $21,000 and $22,000, respectively. With the exception of age of respondent (such that younger respondents tended to undo more frequently), none of the demographic variables described above predicted whether, or how frequently, a respondent undid the event. As well, none of the results described below changed significantly when age was treated as a covariate. With respect to undoing, no differences were noted between those losing a spouse and those losing a 5The inclusion of data from both parents who lost a child raises the possibility that nonindependence of data may distort the results. To investigate this possibility, I (a) looked for common counterfactuals shared by partners; and (b) repeated all analyses reported for mothers only. Interestingly, no counterfactuals were simultaneously reported by both mother and father, and the results did not change when fathers were excluded from the analyses. To maximize statistical power and the external validity of the results, I report the analyses based on both partners. 25 child. On a cautionary note, the lack of significant difference in undoing as a function of categorical demographic variables (e.g., sex, race) should not be interpreted as suggesting that no differences in each group's propensity to undo exist in their larger population. Small cell sizes among some of these variables render the possibility of Type II error considerable. For instance, the power to detect a moderate effect (d = .5; see Cohen, 1977) for sex is .69.6 The relative lack of association between categorical demographic variables and undoing may be interpreted as suggesting that these variables cannot, in themselves, account for the relations of undoing to distress described below. Prevalence of Undoing When asked if they had ever found themselves going over the events leading up to the accident (i.e., ruminative thoughts of the event), 8 0% responded that they had. Of these respondents, 59% reported doing so during the past month (4-7 years later). When asked what they had thought about when they were going over the events, fully 56% of these 75 respondents spontaneously reported wishfully thinking of undoing the events. For example, one respondent said: "I think about the fact that if I had grounded him that night as I wanted to, it might not have happened ..." 6In contrast to the low power of the categorical variables, the power to detect a moderate effect between undoing and continuous demographic variables (e.g., age, income) was respectable (.83). 26 The remaining 44% of the 75 respondents reported going over the events without explicitly mentioning thinking about undoing the event. For example, one respondent recalled: "[I think about how he] looked at the time. A witness told me how carefully he was driving, and that because of his long, blonde hair, the witness thought it was a girl. I wondered whether he realized what was happening and was terrified or frightened." Of the 42 respondents who spontaneously reported trying to undo in response to the question regarding ruminations of the event, 55% reported trying to change their own behavior, whereas the remainder reported trying to change the behavior of the deceased. Interestingly, not one person reported trying to undo the other driver's behavior, despite the fact that in most cases the other driver was negligent or allegedly under the influence of drugs or alcohol at the time of the accident. The question, "Have you ever found yourself going over the events leading up to the accident?" provides a conservative estimate of the prevalence of undoing because respondents had not been primed to mention undoing thoughts and were free to recall any memory. The frequency with which respondents spontaneously reported undoing, however, attests to the prominence of these thoughts in the minds of the bereaved. To assess attempts at undoing more directly, respondents were asked, "Since the accident, have you ever found yourself thinking 'If only I had done something differently, my 27 (child/spouse) would still be alive?'" Fifty-eight percent responded that they had done so. Of these, 48% reported that they still think, 4-7 years later, "if only ..." at least rarely (M = 2.69, SD = .74). Undoing and Distress On the basis of their answers to these closed-ended questions, respondents were divided into three groups: those who reported never trying to undo the experience (n = 39); those who reported thinking "if only ..." at some point since their loved one's death, but not in the past month (n = 28); and those who still had these thoughts at the time of the interview (n = 2 6). An ANOVA on the distress index was significant, F (2, 90) = 3.92, p_ < .05. Post hoc pairwise contrasts (Tukey) to determine the source of the significant effect indicated that those who were still attempting to undo reported more psychological distress (M = .37) than both those who reported never attempting to undo (M = -.11) and those who reported undoing only in the past (M = -.15). The frequency of undoing in the past month was related to respondents' current level of distress, r (91) = .29, p_ < .01. (The corresponding correlation for the group of 54 respondents who had reported undoing since the accident was .37, p_ < .005). This significant correlation held after controlling for both the frequency with which respondents reported ruminating about the event in the past month (correlated .40 with undoing and .09 with distress) and ruminating about the deceased in the past month (correlated .26 with undoing and .24 with distress), partial r 28 (89) = .25, p_ < .05. Conversely, after partialling out the frequency of undoing thoughts, the correlation between ruminations of the deceased and distress was no longer significant, partial r (90) = .17. Content of Undoing Thoughts With respect to the content of the 54 "if only ..." thoughts, responses were coded in terms of whether the respondent wished to change a usual or an exceptional behavior. Sixty-nine percent of these respondents reported undoing an antecedent event that was coded as a usual activity-. For example, several respondents reported thinking something like: "if I hadn't let him go to the store that day...."7 Another respondent said, "If I had talked him into staying home that day — but there was no reason to. ... He was very good about not missing work." Twenty-eight percent reported undoing an antecedent event that was coded as an exceptional activity. For instance, one parent said: "I sometimes think if we hadn't scheduled my daughter's driving test for that day [my husband] might have survived because he would have been driving the big car." 70ften, it was not possible to determine definitively whether the reported behavior was usual or unusual. For example, although unlikely, "going to the store" may have been an unusual act rather than a usual one. However, the reasonably high inter-rater reliability suggests that coders were, at a minimum, consistent in distinguishing usual from unusual actions. Note also that 2 of the 54 responses were too ambiguous to code. 29 The two groups (those undoing usual vs. exceptional events) did not differ in the frequency with which they were currently undoing (t < 1), nor in their level of distress (t < 1). The power of these tests to detect a moderate effect (d = .5, alpha = .10) was .62. With respect to the omission/commission distinction, 41% of those undoing reported wishing that they had done something (i.e., an omission), while 31% reported wishing that they had not acted as they did (i.e., undoing a commission). Seventeen percent reported both omissions and commissions; the remaining 11% of responses were too ambiguous to code. Again, there were no differences in frequency of undoing or distress as a function of whether respondents were undoing omissions or commissions (ts < 1). Assuming a moderate effect size and a liberal alpha criterion, the power of these tests was .59. Suspecting that the null effects for the two mutability factors may be due to the two factors cancelling each other out, I also compared those who reported neither unusual antecedents nor committed acts in their counterfactuals with those who reported either or both. No significant differences were observed (ts < 1.1), suggesting that the mutability factors did not relate to the respondent's level of distress or the extent to which they undid the event. Assuming a moderate effect size and a liberal alpha criterion, the power of these tests was .69. 30 Blame for the Accident The foregoing, representative quotes indicate that the counterfactuals offered by respondents do not imply that they caused the accidents. When asked who or what they blamed for the accident, 3 0% of respondents said they blamed no one. Of those who did blame someone or something (n = 65), only 6% blamed themselves, whereas 69% blamed the other driver, 11% blamed society or the justice system, and the remainder blamed some other source (e.g., God) or declined to identify a source. Although all four respondents who reported blaming themselves had also reported undoing, across the total sample respondents did not differ in terms of distress as a function of who or what (or if) they blamed (F < 1). Finally, the group of respondents who had reported undoing could not be differentiated from those who did not undo the event in the types of attributions of blame they made for the accident (X2[4] = 5.35, ns.). Discussion The data from Study 1 suggest that undoing is a common response to the unexpected loss of a spouse or child from a motor vehicle accident. Of the 80% of respondents who reported ruminating about the events leading up to the accident, more than half spontaneously mentioned that they had thought "if only ...". When asked directly if they had ever thought "if only I had done something differently...", 58% of respondents reported that they had. Of these, nearly half reported that they still had such thoughts, 4-7 years after the loss of their loved one. 31 The data suggest further that recent undoing is associated with greater distress. Those who were still undoing the event were more distressed than those who reported never undoing and those who reported undoing only in the past. The latter group of respondents were no better or worse off than those who never tried to undo the event. Study 1 enabled me to distinguish undoing thoughts from other more general ruminative thoughts concerning the accident or the deceased (i.e., "going over the events leading up to the accident" and "memories, thoughts, or mental pictures of the deceased"). After controlling for the frequencies of these other ruminative thoughts, the frequency of current undoing remained a significant predictor of distress. General ruminations, on the other hand, were not significantly associated with distress, after controlling for undoing. This suggests a unique association between undoing and distress, above and beyond that described by general ruminative thoughts. Stated differently, the association between undoing and distress does not seem to be mediated by the extent to which people continued to ruminate about the event or the deceased. With reference to the content of reported counterfactuals, I was unable to garner much evidence of previously identified mutability factors influencing the undoing process with real-life events. First, it was difficult to differentiate omissions from commissions. Many responses could be interpreted as either. Indeed, in a number of instances, respondents explicitly stated a 32 single counterfactual in terms of both an omission and a commission. The lack of any difference in undoing frequency or distress between those undoing omissions versus commissions further suggests that such a distinction was not critical for respondents. Second, most of the bereaved reported undoing seemingly typical or usual behaviors and circumstances. Furthermore, those who reported undoing unusual antecedents were not more likely to persist in undoing, nor were they more distressed. Given the base rates in the real-world, circumstances leading up to accidents are more likely to be of the usual, than unusual, kind. Thus, one cannot conclude from these data that typical events are more likely than, or even as likely, to be undone as exceptional events. Perhaps the respondents who reported undoing typical circumstances did not have any unusual circumstances to undo. In the absence of highly mutable circumstances, then, they may have settled for less mutable incidents, but incidents nevertheless over which they could have exerted some influence. Corroborating this, when asked to describe what they thought about when they went over the events leading up to the accident, all respondents who mentioned counterfactuals reported trying to undo their own behavior or that of their deceased loved one. None reported trying to remove the other driver from the scene or changing the other driver's behavior, despite the fact that most people blamed the other driver for engaging in an unlawful behavior (e.g., driving while intoxicated, speeding). This result is consistent 33 with some scenario research suggesting that counterfactuals are biased towards the focal actor's behavior (e.g., Dunning & Parpal, 1987; Kahneman & Miller, 1986; Kahneman & Tversky, 1982). That I failed to find support for two of the traditional mutability factors (i.e., exceptionality and commissions) affecting the undoing process raises questions about the relative psychological significance of such factors within the context of real-world traumatic events. Although these tests may have suffered from a lack of statistical power, it is important to note that the t statistics were far from being significant. At a minimum, the data suggest that a number of other factors may influence the undoing process. For example, given an affectively charged event such as the unanticipated death of a loved one, people seem motivated to search for any behavior that they could imagine performing differently, even highly "normal" or "routine" behaviors. This implies that undoing may be, at least partly, a distress-driven cognitive process. That is, the more distressed one is, the more likely one is to undo. If this is the case, then such initial distress may overshadow the role that previously identified mutability factors may play. One way to pursue this possibility is with a longitudinal field study, which allows an examination of the undoing process over time. The distress-driven interpretation would gain plausibility if Time 1 distress predicts Time 2 undoing, controlling for Time 1 undoing and mutability factors. That is, those who are undoing most frequently at Time 2, controlling for 34 the frequency of undoing at Time 1 and the presence or absence of highly mutable antecedent circumstances, should be those who are most distressed at Time 1. Another reason for the lack of a relation between previously identified mutability factors and distress in samples of real-life victims might stem from the fact that reported distress is due to a variety of thoughts, feelings, and life experiences unrelated to counterfactuals. Given the complexity of the coping process, and the strong individual-difference factors at play, it may be more useful to examine specific affective consequences, rather than general distress. Weiner's attribution model of emotion (198 6; Weiner, Graham, & Chandler, 1982) is instructive in this regard. Weiner et al. (1982) have shown that particular attributions of causality have specific affective consequences. For example, we feel anger if we perceive that others have caused us misfortune (through their controllable actions), or guilt if our controllable actions cause misfortune to befall others (Weiner et al., 1982). Although Weiner's theory is couched within an attribution framework, it may be applied to counterfactuals. Indeed, past counterfactual research suggests that people may feel guilty even when their own actions are not causally implicated in another's misfortune. That is, if one's own actions (or inactions) are perceived to be sufficient to undo the outcome, even if such actions did not play a causal role, one is likely to feel regret (e.g., Landman, 1987), and perhaps to some degree, a sense of 35 guilt and responsibility (Abbey, 1987; Wells & Gavanski, 1989). In fact, some researchers view the emotions of guilt and regret as "counterfactual emotions" (Boninger, Gleicher, & Strathman, 1994) . That so little self-blame was reported by respondents in Study 1 may be owing to the specific methodology employed: respondents were asked merely whom or what they blamed, they were not given the opportunity to rate degrees of blame for different sources (e.g., self, other person, God). An additional issue worth pursuing with longitudinal field data concerns the stability of counterfactuals over time. Do people focus and then dwell on a single counterfactual, or do they shift their counterfactual focus from one antecedent to another? It seems reasonable to expect that outcomes with highly mutable antecedents, because they are so salient, may remain stable over time. Conversely, if a single, highly mutable and salient counterfactual is not available (as I have argued is often the case in the real world), one might anticipate that counterfactuals would be less stable over time. 36 Chapter 3 Study 2 In an effort to test these hypotheses, I examined longitudinal data collected from parents who had lost a child to Sudden Infant Death Syndrome (SIDS), interviewed approximately 3 weeks and 18 months following their baby's death.8 SIDS is identified as the official cause of death when a post mortem examination rules out all other potential causes (Bergman, Beckwith, & Ray, 1970). As such, SIDS has no known cause. Like the prior study, death of one's infant to SIDS is sudden, unexpected, and unpredictable. As well, parents were vindicated of blame and responsibility by a public health nurse who informed them that their infant had died of SIDS (prior to our interviews). For these reasons, there was no obvious causal event for parents to undo and no known objective reason to self-blame. Method Respondents and Procedure Respondents were parents from Cook County, IL, and Wayne County, MI, who had lost an infant to SIDS. All mothers who had lost an infant to SIDS within a two-year window of time were sent 8SIDS parents were also interviewed at a third time point (3 months postloss). As this project is concerned only with early and long-term effects of undoing, I have not included the 3-month data in this report. In addition, one-third of the infants who died of SIDS during the study period were assigned to an additional condition in which parents were invited to participate in the research at one time only, at 18 months postloss. As these respondents are not part of the results I report here, they will not be discussed further. 37 a letter seven days after the infant's death describing the study and inviting their participation. Mothers were subsequently contacted by telephone (or by a home visit if parents had no telephone), and, if they agreed to participate, an interview was scheduled. If a mother agreed to be interviewed and the baby's biological father was living with the mother at the time of the death, the father was also invited to participate. The first interview was scheduled 15-3 0 days after the baby's death. The final follow-up interview occurred at approximately 18 months postloss. Mothers and fathers were interviewed separately by different interviewers and, whenever possible, the parent was assigned the same interviewer for the duration of the study. Response Rates and Attrition Contact was attempted with 281 mothers who met the following eligibility requirements: (a) the death was classified as SIDS on the basis of an autopsy, (b) the mother was English-speaking and at least 15 years of age, and (c) a visiting public health nurse had informed the mother prior to our contact that the infant had died of SIDS. Of the 281 mothers, 69 could not be located by the interviewers or could not be scheduled for the first interview during the eligibility period (15-30 days postloss). Only 40 mothers (18.9%) who were located refused to participate, leaving a sample of 172 mothers who participated in the first interview. Of the 172 mothers, 85 were currently living with the infant's 38 biological father. Of these 85 fathers, 56 (65.9%) also agreed to participate. At 18 months postloss, 124 of these parents were re-interviewed (54.3% of those participating at Time 1). Of those not participating at 18 months postloss, 38 refused, 63 could not be located or scheduled within the eligibility period, and 2 infant deaths (involving 3 parents) were reclassified as non-SIDS. Interview Instrument The interview covered several aspects of coping with the loss of one's infant, and lasted approximately two hours. Like the motor vehicle accident interview instrument, the SIDS instrument was a combination of previously validated self-report scales and open- and closed-ended questions. Psychological distress. General distress was assessed by a 32-item version of the SCL-90-R (Derogatis et al., 1976). The items retained from the SCL-90-R included the depression and somatization subscales, as well as the seven additional items assessing generalized distress. As a more specific index of affect, the Affects Balance Scale (ABS; Derogatis, 1975) was employed to measure the frequency of various emotional states during the previous week. The ABS is made up of eight 5-item subscales assessing four negative (i.e., depression, anxiety, guilt, and anger) and four positive (happiness, vigor, affection, and contentment) affective states. Respondents completed a self-39 administered version of the ABS, then the SCL, at the beginning of each interview. Undoing, Undoing was assessed after extensive consultation with SIDS parents and pilot testing in order to avoid inducing guilt. In the first interview, respondents were asked to describe what happened before their baby's death. These open-ended responses were coded in terms of whether respondents mentioned antecedents that seemed atypical or unusual. One quarter of the responses were coded by two independent judges. The two judges' ratings were identical for 90% of the event description responses. Given the high inter-rater reliabilities, the remaining 75% of responses were coded by one of the coders. Parents were then asked two questions specifically on undoing. The first question was: "Even though medical facts tell us that there is nothing parents could have done to avoid their baby's death, parents sometimes report spending time thinking of ways the death could have been avoided. In the past week, have you ever found yourself thinking of ways the death could have been avoided?" Respondents were then asked, "In the past week, have you ever found yourself thinking, 'if only something had been different, my baby would still be alive?'" Both questions were answered on 5-point scales, ranging from No, never (1) to Yes, all the time (5), and they were combined to form an undoing scale. 40 Following the first of the two undoing questions, respondents who reported that they had, at least rarely, thought that the death could have been avoided were asked to describe how it could have been so. The open-ended responses were later coded in three ways. First, responses were coded in terms of who or what parents thought could have prevented the death. Second, the data were coded in terms of whether the death could have been avoided by undoing a committed (vs. an omitted) action. That is, did parents wish to change something that they or someone else did (a commission), or something that they or someone else did not do (an omission)? Finally, responses were coded according to whether the undoing was based on any unusual antecedent event or circumstance. Again, one quarter of the responses were coded by two independent judges, who agreed in 97%, 91%, and 94% of the cases for each of the three coding questions described above, respectively. Given the high reliabilities, the remaining 75% of responses were coded by a single coder. Causal Attributions and Judgments of Responsibility. Causal attributions were assessed by asking respondents if they had "a hunch or theory about what caused [their] baby to die," and if so, to explain it (see Downey, Silver, & Wortman, 1990).9 These 9Downey et al. (1990), using the same dataset, suggested that one other question in the interview reflected a causal attribution (i.e., "Have you ever thought there must have been something you did or did not do to bring about this event?"). This question cannot be used to distinguish counterfactual thoughts from causal attributions, because an affirmative response can imply either that one could have acted to prevent the death or that one's behavior was instrumental in causing the death. Given this ambiguity, I have elected not to use it. 41 open-ended explanations were coded into one of three categories: theories that implicated the self, theories that did not implicate the self, and no theories or hunches. In addition, parents were asked for their beliefs about who or what was responsible for their baby's death. On a 5-point scale ranging from Not at all (1) to Yes, a great deal (5), parents rated the extent to which they felt: (a) they were responsible; (b) someone else was responsible; (c) God was responsible; and (d) that the baby's death was due to chance. Ruminations. The extent to which respondents experienced ruminations about the deceased was assessed with the question, "During the past week, have memories, thoughts, or mental pictures of your baby come into your mind?" Response options ranged from No, never (1) to Yes, all the time (5). The interview conducted at 18 months postloss did not ask respondents to describe the events leading up to the death. Otherwise, all questions outlined above were identical at both interviews. The open-ended undoing question at the 18-month interview was additionally coded in terms of whether or not the respondent was undoing the same antecedent event(s) he or she described in the initial interview. Results Sample Demographics The final sample (those who completed both interviews; N = 124) was 50% Black, 45% White, and 5% other ethnicities. The proportion of Blacks in the present sample is higher than that of 42 the counties from which the sample was taken (3 0% Black), but reflects the higher rate of SIDS deaths among Black infants (National Center for Health Statistics, 1987). Parents' mean age was 25 years (SD = 5.33). Median family income at the time of the interview (1983-1984) was approximately $11,000 (range $1,000 to over $35,000). The mean years of education was 11.6 (range 6 to more than 17 years). Infants' ages at death ranged from 9 days to 11 months, with a mean age of 81 days (SD = 49 days). Fifty percent of the respondents were married, and 15% reported that they were the only adult member of the household. Ninety-eight respondents (79%) were mothers.10 Except for a Time by Sex interaction, such that mothers reported more undoing than fathers at Time 1 (F [1, 122] = 4.28, p < .05), no differences in the life circumstances described above were related to parents' level of distress or frequency of undoing. None of the results 10As with Study 1, I was concerned about the possibility that the results may be affected by nonindependence of data for the 26 cases where both parents were interviewed. To investigate this possibility, I again (a) looked for common counterfactuals shared by partners and (b) repeated all analyses reported for mothers only. As with Study 1, I found that no counterfactuals were simultaneously reported by both mother and father. As well, results did not change when fathers were excluded from the analyses. The fact that so few fathers were included in this study cautions me from making statements about gender differences in undoing. Given that mothers tend to have more significant care-giving roles than fathers, it seems plausible that they would generate more instances where they could have done something different to prevent the loss. Nevertheless, fathers in the study were just as likely to report behaviors that they could have performed (or not performed) that might have prevented the death. Those fathers who were not included in the study (e.g., because they were not living with the mother at the time of the baby's death) may differ from those fathers who were included in terms of their involvement with the infant and hence their likelihood of generating plausible counterfactuals. 43 described below changed when gender was treated as a covariate. The power of these tests was satisfactory, ranging from .78 to .92. Regarding attempts at undoing, SCL distress, and the specific emotions of the ABS, attrition from 3 weeks to 18 months postloss appeared random. That is, those who participated in the 18-month interview were no more or less likely to report undoing, and scored no differently on the SCL and ABS subscales at Time 1 than those who were only interviewed at three weeks postloss (ts < 1.60, p_s > . 10) . Mutability of Events Leading Up to SIPS When asked to describe the events leading up to their baby's death, 67% (n = 82) of the respondents described what appeared to be exclusively usual or normal events. In fact, several respondents spontaneously mentioned that nothing unusual occurred prior to the baby's death. (These mentions, incidentally, might reflect people's intuitive sense of Kahneman & Miller's [1986] norm theory.) The remaining one-third of respondents (n = 41) described events that were coded as unusual. Of these, 2 7% reported unusual events involving the baby's health (e.g., the baby had recently been ill); 15% recounted that the baby was feeling unusually cranky; 15% mentioned that the baby's sleep was different (e.g., the infant slept through a regular feeding time); 10% noted unusual events occurring in or around the house just before the death (e.g., the infant had recently switched bedrooms); 10% explained that the baby was atypically away from 44 home at the time of, or just prior to, the death; 7% reported that they had a strange premonition that something was wrong with the baby; whereas the remainder (22%) described idiosyncratic unusual events. Content of Undoing Thoughts Despite the finding that only one-third of respondents reported unusual events leading up to the death, 76% of the respondents reported undoing the baby's death at three weeks postloss at least rarely (operationalized as a M of 2 or more on the two-item undoing scale). By 18 months, this percentage had dropped, but was still at 42%. When those who reported undoing at three weeks postloss were asked to recount their thoughts, 90% described their own behaviors, or lack thereof, that might have prevented the loss. Only two respondents (2.4%) mentioned that medical professionals might have done something to avoid the death. The remaining responses were either too vague or ambiguous to be coded. In terms of omitted versus committed actions, the responses were similar to those described in Study 1. That is, many responses could be interpreted as wishing one had acted as easily as they could be interpreted as wishing one had not acted. For example, several respondents reported that they could have stayed awake the night the baby died (i.e., an "omitted" act), whereas others reported wishing that "if only I didn't go to sleep ..." (i.e., undoing a "committed" act). Again, to achieve consistency, responses were coded literally (i.e., "had" vs. "had 45 not"). Using this literal coding rule, 66% of respondents who were undoing the event reported wishing they had acted to prevent the death (an omission), whereas only 12% reported wishing they had not done something (a commission). Sixteen percent of respondents reported both; that is, that they could have, as well as might not have, done something to avoid the death. Six percent of responses were too ambiguous to be coded. Despite the fact that 41 parents described unusual events or circumstances prior to the death, only 13 respondents (11 who had described unusual events and two who in fact had described usual events in the event description) reported undoing something that was coded as an unusual circumstance. For instance, one parent said, "If I just woke the baby up when I got up..., because I usually wake him up and feed [him] before I go to school." Another parent said, "I was late getting up for work. If I had gotten up earlier, it seemed that it had just happened...." In contrast, 63 parents reported undoing antecedent events that were coded as normal or typical. For example, one parent reported, "I thought about maybe if I'd stayed awake, or woke up more frequently [to] check on the baby. I keep thinking of different ways .... So many things go through your mind." Another parent said, 46 "I've thought about a lot of those things, like if I'd never set him down; if I'd never gone to school that morning; if I had never been going to school, leaving him in the mornings. I thought about it, checking him every five to ten minutes instead of every 15-2 0 minutes. I thought about all these things I could have done but I didn't do." Stability of Counterfactuals Over Time Interestingly, only 28% of respondents who reported counterfactuals at 18 months postloss described undoing the same antecedent that they were undoing at three weeks postloss. Even so, the percentages of those who reported undoing their own (in)actions (88%) versus others' (in)actions (4%); omissions (65%) versus commissions (20%); and usual (79%) versus unusual (15%) events were essentially the same as those observed at three weeks postloss. Of the 13 respondents who reported a counterfactual involving an unusual antecedent at three weeks postloss, 31% reported the same counterfactual at 18 months postloss, 46% ceased undoing the event and 23% were undoing different (usual) events at 18 months postloss. Among the 63 respondents undoing exclusively usual antecedents at Time 1, 16% reported the same counterfactual at 18 months, 52% ceased undoing by this point, and 32% reported undoing a different event at the later interview. Thus, there is not strong evidence to suggest that counterfactuals involving unusual events are particularly stable 47 over time (X2 [2] = 1.65, ns.). In general, I observed little stability of counterfactuals across time points. Undoing. Mutability, and Distress The frequency with which respondents reported engaging in undoing was significantly correlated with contemporaneous SCL distress scores, such that more undoing was associated with greater distress (see Table 1). The significant bivariate correlations between frequency of undoing and distress held after controlling for the frequency with which respondents reported ruminating about the deceased (partial r = .40 at 3 weeks; partial r = .26 at 18 months). Whereas frequency of ruminations about the deceased was significantly related to distress (after controlling for undoing) at Time 1, it was not at 18 months postloss. To test for the moderating role of highly mutable circumstances preceding the death, I examined whether those who reported unusual antecedents (in their general descriptions of their baby's death; n = 41) were undoing more frequently or reported greater SCL distress at three weeks of 18 months postloss than those who reported exclusively routine antecedents (n = 82). No main effects or interactions involving the antecedent mutability factor were significant (Fs < 2.1, p_ > .15). The power of this test was .80, assuming a moderate effect size. As in Study 1, I also examined whether parents reporting counterfactuals involving unusual antecedents and/or committed Table 1 Study 2: Correlations Between Undoing Frequencies and Negative Affect, 3 Weeks and 18 Months Postloss 3 Weeks: 1 Undoing 2SCL 3 Guilt 4 Anger 5 Depression 6 Anxiety 18 Months: 1 Undoing 2SCL 3 Guilt 4 Anger 5 Depression 6 Anxiety Means Std. Dev. 1 (.58) .44 .49 .31 .37 .20 .39 .25 .24 .21 .26 .22 2.79 1.25 2 (.94) .58 .65 .71 .58 .37 .60 .28 .41 .38 .39 2.18 0.72 3 Weeks Postloss 3 (.87) .62 .61 .53 .43 .43 .37 .35 .40 .37 2.62 1.05 4 (.77) .73 .59 .32 .41 .22 .36 .29 .35 2.83 0.87 5 (.83) .60 .37 .39 .29 .38 .39 .32 3.03 0.88 6 (.69) .25 .40 .22 .33 .34 .34 3.07 0.79 1 (.71) .29 . .35 .37 .34 .29 1.84 0.99 2 (.94) .59 .66 .67 .60 1.67 0.58 18 Months Postloss 3 4 (.88) .72 (.83) .71 .81 .61 .61 1.92 2.27 0.82 0.82 5 (.87) .60 2.23 0.86 6 (.72) 2.61 0.72 Notes, rs > . 18, p < .05 ; rs > .30, p < .001 (two-tailed). Values in parentheses are reliability (alpha) coefficients for scales. 49 acts also reported greater distress or more frequent undoing relative to those parents describing counterfactuals involving only usual antecedents and omitted acts. Although no differences were observed for undoing frequency, those undoing more mutable antecedents (i.e., unusual antecedents or committed acts) were marginally more distressed at Time 1, t (74) = 1.85, p_ < . 05 (one-tailed). As this effect was not replicated at Time 2, I recommend caution in interpreting it. The power of these tests to detect a moderate effect size, however, was poor, particularly when cell sizes were quite disproportional. Even with a liberal alpha (.10, one-tailed), the power of these tests ranged from .59 to .80 at Wave 1 and from .46 to .64 at Wave 2. (The most powerful tests for each wave were those comparing respondents undoing either unusual antecedents or committed acts to respondents undoing exclusively usual antecedents and omitted acts.) Undoing and Specific Components of Distress Undoing was also associated with each of the four negative affect subscales of the ABS at both interviews (see Table 1). As well, frequency of undoing was weakly associated with the positive affect subscales of the ABS, with correlation coefficients ranging from -.12 to -.21. As expected, the Time 1 correlations suggest that, of the four negative affect subscales, undoing was most strongly associated with guilt (Z = 1.71, p_ < .05, one-tailed). To evaluate the magnitude of the unique relation between undoing 50 frequency and feelings of guilt, I regressed the residual undoing frequency (controlling ABS subscale scores for anger, depression, and anxiety) on the ABS subscale assessing guilt. Despite the sizable shared variance of the negative affect subscales (as is evident in Table 1), guilt uniquely accounted for 12% of the variance in undoing frequencies at 3 weeks postloss (Beta = .47, p_ < .001; see Table 2), thus accounting for almost half of the explained variance. Interestingly, at 18 months postloss, the pattern had changed, such that undoing frequency was no longer uniquely associated with guilt after controlling for levels of anger, depression, and anxiety (see Table 2). Even though at Time 1 undoing was most directly related to guilt, there was no evidence to suggest that more guilt was experienced when (a) unusual (vs. usual) events preceded the death or (b) either unusual antecedents or committed acts were undone (ts < 1). Undoing at 18 Months Postloss To address the hypothesis that psychological distress increases the likelihood of future undoing (the distress-driven hypothesis), I regressed undoing frequency at 18 months postloss on prior undoing frequency and the usual/unusual nature of antecedent events, followed by SCL distress (each assessed at 3 weeks postloss). The results, displayed in Table 3, indicate that on the first step undoing frequency at 3 weeks (Beta = .38, E < .001), but not the usual/unusual quality of antecedent events, predicted undoing frequencies at 18 months. Supporting Table 2 Study 2: Regression of Undoing Frequencies on Guilt and other ABS Subscales (a) Step 1 1 1 2 3 Weeks Postloss (Dependent Var = Variable Entered ABS-Depression (3 Wks) ABS-Anxiety (3 Wks) ABS-Anger (3 Wks) ABS-Guilt (3 Wks) (« = 124) = Undoing 3 Wks.) Beta .34 -.07 .10 Stepl 2 .049* .002 .004 R2= 143** Beta .21 -.15 -.05 .47 Step: R* I 2 .017 .013 .001 120** =.263* (b) 18 Months Postloss (Dependent Var - Undoing 18 Mos.) Step 1 Step Variable Entered Beta sr2 1 AB S-Depression (18 Months) 1 ABS-Anxiety (18 Months) 1 ABS-Anger (18 Months) 2 ABS-Guilt (18 Months) (n = 122) .13 .005 .06 .002 .22 .016 R2=.Ul** Beta Step 2 sr" .08 .002 .02 .000 .17 .008 .16 .010 R2=.l5l* *p< .01; **p< .001 52 Table 3 Study 2: Regression of Undoing Frequency at 18 Months Postloss on Antecedent Mutability. SCL Distress. and Undoing at 3 Weeks Postloss Step Variable Entered Beta CT£ 1 Usual/unusual antecedents .07 .005 1 Undoing Frequency (3 wks) .38 .142** 2 SCL Distress (3 wks) .24 .048* R2= .203 ** */><.01; **/?<.001 53 the distress-driven hypothesis, SCL distress, entered on the second step, produced a significant increment in R2 (Beta = .24, p_ < .01) . Undoing, Causality, and Responsibility Attributions In response to the open-ended question concerning theories or hunches that parents had about the cause of their baby's death, only 18.5% implied that something involving themselves may have contributed to the death (e.g., a difficult pregnancy, smoking during pregnancy, complications during birth). Twenty-eight percent of respondents said they had no idea or hunch about why their baby died, whereas the remaining respondents offered hunches or theories that did not implicate themselves (e.g., "It was God's will"). No differences in guilt or distress were found between those who reported a self-implicating hunch and those who did not (ts < 1). With respect to attributions of responsibility for the death, parents, at three weeks postloss, tended to attribute most responsibility to God (M = 3.08), or reported that the death happened by chance (M = 2.36). This is not surprising given that the causes of SIDS are unknown. Notwithstanding this tendency, several parents (n = 52) assumed some degree of personal responsibility (M = 1.92; see Downey et al., 1990)." Although the 18.5% of respondents whose hunches were coded as self-nBlaming another person was rare; only 15% of the sample claimed that another person was at least a little responsible (M = 1.28). 54 implicating also tended to attribute more responsibility to themselves (M = 2.48) compared to those who did not provide a self-implicating hunch (M = 1.80; t [117] = 2.27, p. < .05), it is significant that nearly 60% of those who accepted some personal responsibility for the death did not report a self-implicating hunch. Indeed, excluding those who reported a self-implicating causal theory for their child's death, undoing was still significantly correlated with distress (r [96] = .39, p_ < .001), guilt (r = .44), and degree of self-responsibility ascribed (r = .52). Thus, the observed relations between undoing and guilt and distress do not appear to be due to parents' self-implicating causal attributions. Discussion Thoughts of undoing were as prevalent in Study 2 as they were in Study 1. Seventy-six percent of SIDS parents interviewed three weeks postloss reported that they had counterfactual thoughts concerning their baby's death during the previous week. At 18 months postloss, 42% of the parents reported still having thoughts of undoing the death. Further, the more thoughts of undoing parents had, the more symptoms of distress they reported. This relation held at both time points, and remained significant after controlling for frequency of ruminations about the deceased. One plausible explanation for these findings is that highly distressed people come to undo more frequently, thus perpetuating their distress. Pursuing this distress-driven hypothesis, I 55 found that Time 1 distress significantly predicted later undoing frequency, even after controlling both for how frequently people were undoing at Time 1 and the mutability of antecedent events. Corroborating Study 1, I was unable to garner much evidence that the exceptionality of antecedents or the commission (vs. omission) of actions was associated with more undoing, guilt, or general distress. Although, as with Study 1, these tests suffered from a lack of statistical power, the observed effects were so small (generally ts < 1) that it is unlikely that more data would have rendered significantly different results. However, the failure of the data to yield differences in undoing or distress as a function of traditional mutability factors should not imply that parents were undoing immutable events. Among the respondents who reported undoing only usual antecedents, it should be noted that these were events, nevertheless, that parents could easily, in hindsight, imagine performing differently. That is, they were mutating events over which they had some control — behaviors that they conceivably could have imagined performing differently. Rarely did parents report wishing that medical professionals had done more; never did they report that God could have intervened to save their baby. The focus of their undoing was nearly always on their own behavior, which supports findings obtained with scenario research (e.g., Kahneman & Tversky, 1982). As in Study 1, the omission-commission distinction found in the laboratory (e.g., Gleicher et al., 1990; Landman, 1987) 56 seemed to be an insignificant psychological nuance to the bereaved respondents. Many of the undoing responses given by parents could be stated either as omissions or commissions. Interestingly, though, respondents generally chose to report their counterfactuals as omissions rather than the supposedly more mutable commissions suggested by past research (see Miller et al., 1990; but see Gilovich & Medvec, 1994 for similar findings to our own). This discrepancy may be owing to understandable differences between real-life phenomena and hypothetical scenarios employed in the lab. In the typical role-play scenario, subjects are presented with two descriptions of a negative outcome, differing only in whether the target person performed, or failed to perform, a salient action. In the latter (omission) case, it is clear to the subject that one particular action was omitted (e.g., not switching classes; Landman, 1987). In real life, however, people are likely to consider countless more omissions than commissions in their construals of counterfactuals, because whereas committed acts are limited to what people have in fact done, omitted acts are limited only by one's imagination. This base-rate differential, in and of itself, might account for the preponderance of omissions reported by the present respondents. This possibility is analogous to the notion raised in the discussion of Study 1 that in field research, reported counterfactuals may follow from usual events more frequently than from unusual events because the latter, by definition, occur so infrequently. 57 Of the four negative emotions assessed three weeks after the loss of their infant to SIDS, feelings of guilt were most closely allied with undoing. In fact, of the four negative emotions measured by the ABS (depression, anxiety, anger, and guilt), guilt was the only emotion to possess a unique relation with undoing while the others were statistically controlled. Guilt reported by respondents, meanwhile, was not associated with the parents' hunches or theories that implicated themselves, likely due to the fact that these hunches or theories that did implicate the bereaved parent tended to be of an uncontrollable nature (e.g., difficult pregnancy; see Weiner et al., 1982). In contrast, counterfactuals that respondents reported overwhelmingly involved events or actions over which they had some control. By 18 months postloss, however, respondents reported fewer feelings of guilt than they had at 3 weeks (see Table 1). By this time, those parents who were still undoing were not feeling especially guilty for their (in)actions. Thus, by 18 months, many parents appear to have concluded that the antecedents they were undoing, though controllable, were not foreseeable (see Miller & Turnbull, 1990). Although the lack of foreseeability may, with time, alleviate the sense of guilt felt by parents, it is important to note that undoing (regardless of the content of the counterfactual) retained its association with distress over the study period. 58 It seems unlikely that the counterfactuals reported by parents of SIDS babies led them to feel more causally significant: whereas 68% of parents, when asked how they thought the death could have been avoided, reported a self-implicating counterfactual, less than 20% reported a self-implicating causal theory for their baby's death. Moreover, a review of the counterfactuals that parents reported clearly indicate that for the vast majority of cases the events parents were undoing were not so much causal as they were preventative (e.g., checking on the baby more frequently than was typical). The SIDS study may have limited generalizability, however, in that parents were faced with a situation where cause of death was medically unknown. In fact, as previously noted, SIDS is deemed the "cause of death" when all other possible causes (including causes that might implicate the parent) have been eliminated in a post mortem examination. Lacking a clear causal agent, it is possible that parents may have relied on counterfactuals as a "next best" alternative to helping them understand why the death occurred and where responsibility or blame should lie. It is important, then, to consider an event with causes that are less ambiguous. In this case, would self-implicating counterfactual thoughts still predict ascriptions of personal blame, even after controlling for causal attributions and foreseeability estimates (cf. Shaver, 1985)? This question was addressed in Study 3. 59 Chapter 4 Study 3 Study 3's respondents were individuals coping with spinal cord injury (SCI). This study provides a unique opportunity to examine relations between counterfactuals, attributions of causality, and judgments of blame. The event is severe, unexpected, and has long-term consequences: qualities that previous research suggests will make issues of cause and responsibility important (Turnquist et al., 1988; Wong & Weiner, 1981). As well, unlike the previous SIDS study, respondents in the SCI study generally knew how their injuries occurred. Moreover, because these accidents followed from a wide range of activities, the data permit the examination of counterfactual thoughts over a broad array of causes. The purpose of Study 3 was to determine whether counterfactual thoughts directly relate to ascriptions of blame, as the results from Study 2 suggest (see also Miller & Turnbull, 1990), or whether the relation is mediated by causal attributions (e.g., Lipe, 1991; Wells & Gavanski, 1989). Because causal attributions (and perceptions of foreseeability) are known to be important determinants of blame (Fincham & Jaspars, 1980; Shaver, 1985), and because counterfactuals have been shown to underlie causal attributions (Hilton, 1990; Lipe, 1991), it is necessary to control for these factors in order to test the hypothesis. Unlike in Studies 1 and 2, where cause was controlled in the sense that respondents knew that they did not cause the negative 60 event (or, in the case of Study 2, had been advised to this effect by the Public Health Department), cause and foreseeability were controlled statistically. In line with the results of Study 2, it was anticipated that, irrespective of their causal attributions, respondents would nevertheless feel that they could have done something to avoid the outcome. Moreover, it was hypothesized that assignments of self-blame would be associated with the extent to which respondents felt that they could have avoided their accident, even after statistically controlling for both respondents' causal attributions and foreseeability estimates. In interviews, respondents were asked to describe in as much detail as possible how their accident happened. In addition, respondents were questioned about who or what caused the accident, who was to blame, and whether they thought the accident could have been avoided. To establish an external measure of causality and blameworthiness, a team of raters was given extensive training on the assignment of cause and blame based on Shaver's (1985) model. Briefly, Shaver's (1985) model suggests that increasing degrees of moral responsibility are assigned following a causal attribution on the basis of fulfillment of the following criteria: that the causal agent (a) was aware of the potential consequences (i.e., foreseeability), (b) intended the consequences, (c) acted freely, and (d) understood the moral wrongfulness of his or her acts. Less responsibility is ascribed 61 if only one of the criteria are met and most responsibility is assigned when all four criteria are met. Finally, blame is assigned when there is perceived to be no justification for the behavior that has been deemed responsible for the outcome. Blind to the respondent's diagnostic status, the four raters made their ascriptions from transcribed and carefully edited versions of each respondent's account of how his or her accident occurred. These transcripts were edited so that any mentions by the respondent of blame, responsibility, pain, and distress, were omitted. This was done because it was felt that such references to the extent or permanence of the injury might bias raters' judgments (e.g., Lerner, 1980; Walster, 1966). Raters estimated the degree to which (a) other people, the respondent, the situation, and chance caused the accident; (b) other people and the respondent were to blame for the accident; and (c) the accident was foreseeable from the respondent's point of view.12 This enabled comparisons between data from raters and data from respondents with SCI with respect to their ascriptions of cause, foreseeability, and blame. It was expected that, relative to raters, respondents would blame themselves more without necessarily implicating themselves more causally. It was reasoned that this would be owing to respondents' perceptions that they could have done something to have avoided the outcome. 12Raters were not asked to rate the extent to which the accident "could have been avoided" because this information is not relevant for the ascription of responsibility or blame according to the Shaver (1985) model (or other models; see Fincham & Jaspars, 1980; Hart & Honore, 1959; Heider, 1958). 62 I was interested to learn if this (potential) difference in blame ascription would be at least partially predicted by the extent to which respondents felt that they could have avoided the outcome, after controlling for cause and foreseeability estimates. Method Respondents All persons who had experienced sudden, traumatic accidents within a 22-month window and who (a) lived within a 60-mile radius of Chicago, (b) were admitted to the SCI unit of Northwestern Memorial Hospital within 7 days of injury, and (c) met eligibility requirements (see below) were invited to participate in the study. This group included all paraplegic, quadriplegic, and neurologically intact patients who met the selection criteria. Although neurologically intact patients (i.e., those with a neck or back injury where the spinal cord is not damaged) do not suffer loss of function or sensation in their limbs, their medical care and early hospitalization parallels that of persons with spinal cord injury. Patients were deemed eligible if they were between the ages of 15 and 65, if they were not experiencing acute psychosis, were not under arrest, or were not injured as a result of a suicide attempt. Approximately 80% of persons meeting the eligibility requirements agreed to participate at Wave 1. In sum, interviews 1 week post injury were conducted with 47 quadriplegic, 22 paraplegic, 31 neurologically intact respondents and a further 6 63 respondents for whom diagnosis was unknown (N = 106).13 Thirty-six percent of respondents with paraplegia and 38% of those with quadriplegia suffered complete lesions (i.e., no sensory or motor function below the injured area); the remainder suffered incomplete lesions (i.e., where some degree of sensory or motor function exists below the injured area). Procedure All respondents were interviewed in their hospital room over two consecutive days. Interviewers were given extensive training in interviewing techniques and were briefed on the acute care unit, hospital routines, and medical and physiological issues in spinal cord injury. Interview Instrument At the beginning of the interview, respondents were asked to describe in detail the circumstances that led up to their accident. Interviewers carefully probed respondents for sufficient information regarding the accident (e.g., why respondents were doing what they were doing, what motivated their actions, whether the activities were planned, spontaneous, chosen, or forced) so that the details were clear. Subsequently, respondents were asked to rate on a 7-point scale (1 = Strongly ^Respondents were also interviewed 3 weeks and 8 weeks post injury. Although I do not report data from the second and third interviews in this thesis, it is worth noting that all findings relating to undoing, self-blame, and causal attributions replicate at Wave 2. The results at Wave 3 are in the expected direction, but tend to fall just short of conventional significance levels due, at least in part, to considerable sample attrition (N at Wave 3 = 55). 64 Disagree. 7 = Strongly Agree) the degree to which they agreed with each of the following statements: (a) "The activity I was participating in at the time of my accident was something I freely chose to do myself," (b) "The activity I was participating in at the time of my accident was an enjoyable one," (c) "My behavior at the time of my accident was typical, that is, something I commonly did," and (d) "I feel I should have been able to foresee my accident." After rating the degree of behavioral freedom, activity enjoyment, typicality, and foreseeability, respondents were asked to rate on the same 7-point scale the degree to which they agreed with each of the following statements: (a) "Another person or other people's behavior caused my accident," (b) "My own behavior caused my accident," (c) "Something in my situation or environment caused my accident," and (d) "Chance or fate caused my accident to occur." Respondents then were asked a series of questions about counterfactual thoughts. First, respondents were asked if they ever thought that their accident could have been avoided (Yes or No). Second, the general frequency of such cognitions was assessed by asking respondents the extent to which they agreed that "I spend a lot of time thinking about how my accident could have been avoided" (1 = Strongly Disagree, 7 = Strongly Agree). Third, the extent to which these avoidability notions were focused on the respondent's own behavior (i.e., self-implicating counterfactuals) was assessed by asking respondents how much they 65 agreed with the statement, "I feel that I could have avoided my accident and therefore my spinal cord injury," using the same 7-point scale. Approximately mid-way through the interview (after the causal attribution and undoing questions), respondents were asked, using the same 7-point scale, the extent to which they agreed that (a) "Another person or other people are to blame for my accident" and (b) "I am to blame for my accident." Ratings by Coders Respondents' descriptions of their accidents were transcribed, edited for references to blame, responsibility, pain, and distress, and given to four trained raters. Raters were blind to the diagnostic status of each respondent. Prior to doing any rating, raters were instructed about the conceptual distinctions between causality and blame as set out by Shaver (1985), and the different criteria needed to infer each one. Specifically, cause was established by the criterion of "minimal sufficiency" (Mackie, 1965; Shaver, 1985). That is, each possible cause identified was evaluated for the extent to which it was sufficient (either alone or in concert with other antecedents) to cause the accident. Responsibility was rated after taking into account the extent to which the respondent's behavior was freely enacted or coerced, the extent to which the accident was foreseeable, and the extent to which the respondent or others were negligent, or in some way intended, the accident. 66 Blame was subsequently rated after taking into account any excuses or justifications that might absolve or reduce blame. Each respondent's description was coded by all four raters at the same time. In order to reduce extraneous sources of variance in the coding (due, for instance, to the possibility that coders might miscategorize an antecedent), coders initially read through the entire transcript and then attempted to achieve consensus on the following points: (a) all of the potential causal factors to be considered in the respondent's account and how they should be categorized (person, situation, chance, etc.), (b) the generally accepted level of risk associated with the activity, and (c) the normative standards of responsibility in the situation (e.g., employers and parents are expected to bear some responsibility for the safety of employees and children, respectively; see Shaver, 1985). This procedure ensured that raters were basing their judgments for each case on the same general understanding of the facts. Raters, however, did not discuss the extent to which any of the potential causal factors were important, and in all other respects, worked independently of each other. From the information given in respondents' accounts of their accident, raters were instructed to rate on the same 7-point scale the extent to which they felt that (a) "Another person or other people's behavior caused the event," (b) "The respondent's own behavior caused the event," (c) "Something in the situation or environment caused the event," and (d) "Chance or fate caused 67 the event." Next, raters were asked to rate the extent to which they felt the respondent should have been able to foresee the possibility of an accident happening in this context. Finally, raters judged the extent to which they felt that "the respondent is to blame for the event," and "another person or other people is/are to blame for the event." As a measure of inter-rater reliability, alpha coefficients were calculated for each of the questions. Alphas for the causal ratings were very high, ranging from .96 (due to chance) to .99 (due to other people). Alphas for the two blame ratings were equally high (.97 for respondent blame, and .99 for other blame). The alpha coefficient for foreseeability was trivially lower (.95). As the alphas were extremely high, ratings for each of these variables were averaged across raters to obtain measures of respondent, other people, situation, and chance causality, respondent and other people blame, and foreseeability. Results Sample Demographics Eighty-one percent of the 106 respondents with spinal cord injuries were male. Eighty percent of respondents were Caucasian, 18% were Black, and the remaining 2% were Latino. The mean age of respondents was 28.4 years (SD = 10.1). These demographic characteristics are consistent with those reported in other large U.S. spinal cord injured samples (e.g., Trieschmann, 1978; Yarkony et al., 1987). 68 ANOVAs were conducted on the extent to which respondents (a) spent time thinking about how their accident could have been avoided and (b) thought that they could have avoided the outcome, as a function of type of injury (quadriplegic, paraplegic, neurologically intact), degree of lesion (complete, incomplete, neurologically intact), sex, and race. The only significant effect was for sex (F [1, 104] = 4.16, p. = .04), such that males (M = 4.17) were more likely to report that they could have avoided their accident than were females (M = 3.05). As with Studies 1 and 2, caution is recommended in generalizing the fact that few categorical demographic and injury variables (e.g., type of injury) related to undoing variables. The lack of association between undoing and these variables should not be interpreted as suggesting that no such differences exist in the population that the sample represents, particularly when cell sizes of categorical variables are largely disproportional (e.g., sex, race). In such cases, statistical power is modest (approximately .68), assuming a moderate effect size and a liberal alpha criteria (.10). The power of the tests involving type of injury and injury severity is .73, assuming a moderate effect size. The relative lack of association between these variables and undoing may be interpreted as suggesting that these variables cannot, in themselves, account for the findings observed. Motor vehicle accidents represented the most common cause of spinal cord injury in this study (43%), followed by sporting accidents (25%; e.g., playing recreational football, diving), 69 falls (15%; e.g., down stairs, off a roof), and violent accidents (9%; e.g., shootings, stabbings). The remainder were injured in other freak accidents (e.g., hit by a falling object). Counterfactual Descriptives Seven days following their spinal cord injury, 85% of respondents said they had thought that their accident could have been avoided. When asked the extent to which they agreed with the statement, "I spend a lot of time thinking about how my accident could have been avoided," the mean response was 3.44 (SD = 2.15). When asked further the extent to which they agreed with the statement, "I feel that I could have avoided my accident and therefore my spinal cord injury" (i.e., self-implicating counterfactuals), the mean rating was 3.96 (SD = 2.26). The correlation between the latter two variables was moderate (r [104] = .35, p < .001), such that the more time respondents spent thinking about how the accident could have been avoided, the more they felt that they, personally, could have avoided it. Behavioral Freedom, Enjoyment, Typicality of Pre-accident Activity, and Injury Foreseeability The vast majority of respondents moderately or strongly agreed that the activity they were engaging in at the time of their injury was freely chosen (M = 6.49 on a 7-point scale, SD = 1.40) and something they enjoyed doing (M = 5.68, SD = 1.82). Given these highly skewed distributions, it is not surprising that neither of these situational aspects correlated significantly with the undoing variables (rs < .10). 70 Activity typicality was also skewed, such that most respondents agreed that their behavior at the time of the accident was typical (M = 5.33, SD = 2.07). Though uncorrelated with self-implicating counterfactuals (r < .10), event typicality was associated with time spent thinking the accident could have been avoided (r = -.21, p_ < .05). Unlike Studies 1 and 2, but corroborating past scenario research in the counterfactual literature (e.g., Kahneman & Miller, 1986), those respondents who reported that their behavior at the time of the accident was atypical tended to spend more time thinking that their accident could have been avoided. As might be expected, injury foreseeability (M = 3.61, SD = 2.29) was directly associated with self-implicating counterfactuals (r = .46, p < .001): Accidents deemed by respondents to be foreseeable were more likely to be seen as personally avoidable. Foreseeability was unrelated to time spent thinking about counterfactuals (r < .10). Relations Between Causal Attributions, Counterfactuals. and Blame Assignments The correlations between respondents' causal attributions, ratings of avoidability, and blame assignments are shown in Table 4. As would be expected, the more respondents thought they caused their accident, the more they tended to blame themselves, and the more they tended to agree that they, personally, could have avoided their accident. Interestingly, when self-blame is partialled from the "self-cause" to "self-avoidability" relation, Table 4 Study 3; Correlations Between Respondents' Counterfactual Thoughts. Causal Attributions, and Blame Assignments Variable (1) (2) (3) (4) (5) (6) (7) (1) Self-implicating counterfactuals (2) Time thinking accident avoidable (3) Other caused (4) Self-caused (5) Situation caused (6) Chance caused (7) Blame others (8) Blame self .35 .19 .40 .11 -.05 .19 .54 — .12 .01 .02 -.22 .12 .06 — -.44 .04 -.17 .73 -.41 — .04 .08 -.43 .66 — .12 .05 -.02 — -.23 -.08 -.48 Notes. All items rated on a 7-point scale (1 = Strongly Disagree. 7 = Strongly Agree). rs > .18, p < .05; rs > .32, p < .001 (two-tailed). N=106. 72 the correlation drops to .07 (ns.). In contrast, when self-cause is partialled from the "self-blame" to "self-avoidability" relation, the correlation remains highly significant (partial r = .41). This pattern of results suggests that the "self-avoidability" to "self-blame" relation is not mediated by "self-cause. " Respondents Compared with Raters: Causes and Blame A series of paired t-tests was conducted comparing respondents' and raters' estimates of the degree to which other people, the respondent, the situation, and chance caused the accident (see Table 5).14 No differences were found in the extent to which respondents and raters attributed cause either to other people or the respondent (ts < 1; rs between respondents and raters for self and other cause = .52 and .71, respectively). However, respondents attributed more causal significance both to the situation and to chance than did raters (see Table 5). Importantly, even though respondents (a) did not attribute more causal significance to themselves than did raters and (b) did attribute more causal significance both to the situation and to chance than did raters, respondents (M = 4.2 0) nevertheless were more likely to blame themselves than were raters (M = 3.19; t [93] = 4.17, p_ < .001; r between respondents and raters = .51), and less likely to blame others (M = 2.65) than were raters "Occasionally, raters were unable to confidently assign cause or blame, typically because insufficient information was provided by the respondent. In other cases, transcripts were unavailable and thus could not be coded. Thus, degrees of freedom range from 99 (for causal attributions) to 93 (for self- and other blame). 73 Table 5 Study 3: Respondents versus Raters on Cause. Foreseeability. and Blame Variable Self-caused Other people caused Situation/Environment caused Chance/Fate caused SCI Respondents' Mean 4.51 2.89 4.10 5.11 Raters' Mean 4.38 2.97 2.86 1.18 t- Statistic t (99)< 1 t (99) < 1 t (97) = 4.24 * t (97) = 20.71 * Accident was Foreseeable 3.61 3.83 f ( 1 0 4 ) < l I could have avoided accident 3.96 Selfistoblame 4.20 3.19 f (93) = 4 . 1 7 * Other people are to blame 2.65 3.07 f (93) = -1.94 + Notes: +p < .06; *p< .001; Means are on a 7-point scale ranging from Strongly Disagree (1) to Strongly Agree (7). 74 (M = 3.07; t [93] = -1.94, p_ < «06; r between respondents and raters = .63). Finally, no significant difference was found between raters' and respondents' estimates of injury foreseeability (t < 1). Predicting Self-Blame from Causal Attributions, Foreseeability, and Counterfactual Thoughts To evaluate the hypothesis that self-implicating counterfactuals contribute to respondents' assignments of self-blame, a hierarchical multiple regression was conducted (see Table 6). On the first step of the regression, respondents' causal ascriptions (i.e., to other people, self, situation, and chance) were entered. Table 6 shows that, as a set, these four variables accounted for 46% of the variance in respondents' ratings of self-blame (p_ < .001). Those who blamed themselves more for their accident were more likely to see themselves as causally implicated (Beta = .58, p_ < .001), and were less likely to see others (Beta = -.18, p < .05) and chance (Beta = -.18, p < .05) as causally significant. On the second step of the regression, respondents' ratings of foreseeability were entered, resulting in a significant R2 increment of 4% (p < .01). Accidents that respondents deemed foreseeable were more likely to leave the respondent feeling blameworthy (Beta = .22). On steps three and four, the two counterfactual variables were entered. To control for the general frequency item, I first entered the extent to which respondents agreed that they spend a Table 6 Study 3: Regression of Respondents' Self-blame on their Causal Ascriptions. Foreseeability Estimates, and Counterfactuals Step 1 2 3 4 Variable self-caused other people caused situation caused chance caused foreseeability time thinking accident avoidable self-implicating counterfactuals Total Regression Beta .58 *** -.18* -.01 -.18* .22 ** .02 .36*** R2 increment .46 .04 .00 .07 .57 F Change F (4,100) = 21.46*** F(l,99) = 7.97** F(l,98) = .06 F (1,97) =16.48*** F (7,97) =18.71 *** * p < . 0 5 ; **p_<.01; ***p<.001 76 lot of time thinking how the accident could have been avoided. This did not yield a significant R2 increment. Finally, respondents' ratings of self-implicating counterfactuals were entered. The addition of this variable produced a significant R2 increment of 7% (p. < .001). Those agreeing more with the statement, "I feel that I could have avoided my accident and therefore my spinal cord injury," were more likely to blame themselves (Beta = .36). Finally, to test whether self-implicating counterfactuals could predict the differential ratings of blame ascribed by respondents as compared with trained raters, I simultaneously regressed the difference score of respondents' ratings of self-blame to raters' estimates of respondent blame on respondents' ratings of self-cause, self-avoidability, and foreseeability (see Table 7). The analysis indicated that although self-cause uniquely accounted for some of the difference (Beta = .21, p_ < .06), self-implicating counterfactuals accounted for the most unique variance (Beta = .28, p < .02). The effect for foreseeability was not significant. Discussion Study 3 suggests that attributions of causality and self-implicating counterfactuals can contribute independently to the assignment of personal blame. Moreover, in contrast to past counterfactual research (Wells & Gavanski, 1989), these data imply that notions of avoidability not only influence blame assignments via their effect on ascriptions of causality, but Table 7 Study 3: Regression of Respondents' Causal. Counterfactual. and Foreseeability Ratings on Rater-Respondent Self-blame Discrepancy Variable Beta sA p self-caused ratings .21 .03 p < .06 self-implicating counterfactuals .28 .06 p < .02 foreseeability estimates 02 .00 ns Total Regression .18 p<.00\ N=94 78 also may operate over and above causal explanations. When self-blame was controlled statistically, no relation was observed between self-implicating counterfactuals and ratings of self-cause. Conversely, when self-cause was controlled, self-implicating counterfactuals and ratings of self-blame remained highly related. These results ought not be interpreted as a rejection of past work in the counterfactual literature suggesting that counterfactuals (e.g., perceptions of how the accident could have been avoided) are implicated in causal judgments (e.g., Hilton, 1990; Lipe, 1991). No doubt, counterfactual information is valuable in determining cause. Nevertheless, it is important to note that the counterfactual alternatives that people tend to dwell on frequently have little to do with people's causal ascriptions, yet still seem to affect how they assign blame. These data suggest that it is not sufficient for people to know who or what caused the negative outcome to occur; people also want to know how such outcomes could have been avoided (see also Miller & Turnbull, 1990). To the extent that one's own antecedent (in)actions are perceived as mutable or changeable, suggesting that a negative outcome could have been avoided, one will likely ascribe some degree of blame to oneself. It is also evident from these data that the relation between self-implicating counterfactuals and self-blame is not due to an exaggerated perception of foreseeability on the part of respondents. First, no significant difference in foreseeability 79 was observed between respondents and raters. Second, even after controlling for respondents' ratings of foreseeability (i.e., the extent to which the respondent felt that he or she should have been able to foresee the injury), self-implicating counterfactuals still accounted for a significant amount of variance in personal blame assignments. Rather, the data suggest that respondents may be confusing what might have been the case with what ought to have been the case (i.e., the counterfactual fallacy; Miller & Turnbull, 1990). That is, the more respondents think that they could have avoided their accident, the easier it is for them to imagine the accident not happening. In retrospect, actions that might have been taken become actions that should have been taken. The present study sheds additional light on how those who have experienced traumatic life events may come to blame themselves. The data suggest that, even in the absence of any reasonable causal connection, the more people think that they could have avoided their accidentr the more likely they are to blame themselves. The first two studies have demonstrated that people are more likely to focus on their own behavior when thinking about how negative outcomes might have been avoided than they are to focus on the behavior of others or on the situation. In focusing their counterfactual thoughts on their own behaviors, people appear to be concluding that in some way they may have contributed to their injury. Though they may not have caused their accident, it remains painfully clear to many of them that, 80 for example, had they left minutes earlier, or taken a different route, they would not be disabled. That these personal actions can so easily be mutated in a mental simulation appears to lead people to assume greater personal responsibility and blame for their outcome. It seems important for future research to consider more carefully the (in)actions for which people are blaming themselves. Specifically, this study, and the previous SIDS study, suggest that people may not be blaming themselves so much for being the cause of their injury (or loss, etc.) as much as for not avoiding it (cf. Abbey, 1987; Miller & Porter, 1983). Although SCI respondents' self-blame ratings were higher than those made by trained raters, the two groups did not differ in their ratings of self- and other-cause. Limitations of Study 3 These data challenge the implicit assumption held by many in social psychology (see Shaver & Drown, 1986), as well as in naive psychology, that a causal attribution is a necessary precondition for the assignment of responsibility and blame. I have suggested that people often blame themselves for negative outcomes even if they are not causally implicated. Presumably, they do so because they perceive that they could have done something to avoid the outcome. The data presented in support of this are limited, however, by their cross-sectional nature. I have tended to assume, on logical grounds (and like many past causal attribution theorists), that the direction of causation flows from 81 counterfactual perceptions of avoidability to blame, rather than from blame to counterf actual.15 Although I have controlled for other variables that could account for these results (e.g., causal ascriptions, foreseeability estimates), the possibility remains that some unmeasured third variable drives both blame and perceptions of avoidability. Clearly, further research is needed before definitive causal conclusions can be reached. If one accepts that the causal direction proceeds from counterfactual to blame, then an additional question is why people should accept blame simply because they can imagine that they could have done something differently to avoid the accident. Do self-implicating counterfactuals prime culturally-defined norms that suggest "if you could have avoided it and didn't, then you are blameworthy"? If so, there may be significant cultural variability in the extent of association between self-implicating counterfactuals and blame. For example, in societies where behavior is more readily understood to be role-defined and less a function of individual choice (e.g., in collectivist cultures), the failure to avoid misfortune, even if you can imagine such avoidance, need not imply that you should feel blameworthy. In this study, respondents with spinal cord injuries blamed themselves more than they were blamed by trained raters. I am not in a position, however, to suggest the superiority of one of "Although the data presented are part of a 3-wave panel design, they do not allow me to test the causal direction of cognitions that occur nearly simultaneously, immediately following the accident (and prior to the first interview). 82 the groups over the other in terms of accuracy. Although it is possible that respondents are taking more blame than is their due (as has been argued for other victim populations, see e.g., Bard & Sangrey, 1979; Medea & Thompson, 1974), it is also possible that respondents may have omitted significant details in their accounts of the events that the trained raters read. Although this latter interpretation is inconsistent with the fact that respondents and raters did not differ in their ratings of self-and other-cause, in the absence of any clear, normative standard one must be cautious in suggesting that respondents are "exaggerating" their blame. 83 Chapter 5 General Discussion The three field studies presented offer some much needed ecological validity to the counterfactual literature. They illustrate that undoing is a common response to unexpected traumatic events, and moreover that mentally trying to undo a loss is associated with heightened distress and greater self-blame. The data suggest further that related constructs described in the stress and coping literature, such as causal attributions and ruminations, cannot fully account for the observed relations between counterfactual thinking, distress, and ascriptions of blame. Without arguing that counterfactual thoughts are more important than these other cognitions, I suggest that making these distinctions might help clarify inconsistent results in the literature and further our understanding of how people are affected by traumatic events. For example, these data may shed light on why people victimized by crime or accidents often appear to blame themselves for their predicament (e.g., Bard & Dyk, 1956; Bard & Sangrey, 1979; Bulman & Wortman, 1977; Burgess & Holmstrom, 1974; Janoff-Bulman & Lang-Gunn, 1988; Wortman, 1976). One's desire to understand how the victimization could have been avoided is more likely to focus on even trivial aspects of one's own behavior than on the causally more significant, but less mutable, behavior of the perpetrator. To the extent that a salient and plausible counterfactual suggests that one could have 84 avoided the outcome, one may come to believe that one should have been able to avoid it (Miller & Turnbull, 1990). As a consequence, people may blame themselves (or hold themselves partially responsible), and hence may feel guilty, for not having been able to do something, however unforeseeable, to prevent the victimization (see Abbey, 1987; Miller & Porter, 1983). In support of past counterfactual research (e.g., Kahneman & Miller, 1986), I found that people experiencing traumatic events overwhelmingly tended to focus their undoing on their own behavior, rather than on antecedents that may have been causally more significant, but beyond their control. This self-focus likely stems from perceiving one's own actions as more changeable (mutable) than externally caused events or behaviors. In contrast to the laboratory-based research, however, I failed to find consistent evidence suggesting that unusual antecedents or committed acts are more undoable or distressing than usual antecedents or omitted acts. The rarity of exceptional circumstances in everyday life suggests that additional attention be paid to the process of undoing when highly mutable antecedents are lacking. At a more fundamental level, the present results beg the question, "Why, given the seemingly negative consequences, do people coping with traumatic life events undo as much as they do?" One possibility is that undoing may be an attempt by people to gain some sense of psychological control or understanding over an uncontrollable, senseless event (Wortman, 1976). To the 85 extent that most life events are, in some way, controllable and make sense, undoing processes may have generally adaptive functions. That is, people often can come to believe, through counterfactual thoughts, that they can avoid similar kinds of events in the future (Boninger et al., 1994; Markman et al., 1993; Roese, 1994; Taylor & Schneider, 1989). This argument is strikingly similar to Janoff-Bulman's (1979; Janoff-Bulman & Lang-Gunn, 1988) thesis that self-blame is likely to yield positive consequences if the behavior that is blamed is, from the person's perspective, behaviorally modifiable (i.e., "mutable") and lends a sense of future control. However, among those coping with spinal cord injury, as with the bereaved respondents in Studies 1 and 2, avoidance of a future injury or loss does not appear to be an important psychological issue. Most of the spinal cord injured will never again walk, much less drive a car, play football, or dive into a swimming hole; the bereaved are very unlikely to lose another loved one in similar circumstances. Nevertheless, I hesitate at this time to suggest that the self-blame or counterfactual thoughts reported by these respondents are therefore exclusively dysfunctional. When placed within the larger context of coping with shattered assumptions, counterfactuals, like behavioral self-blame, may yield other less obvious psychological benefits (see Janoff-Bulman, 1992). For instance, they may help people maintain the belief that negative events do not occur randomly and senselessly, and thus may yield some psychological comfort. 86 Perhaps counterfactual thoughts, like behavioral self-blame, help people maintain the belief that negative outcomes are at a minimum potentially controllable, and thus not unpredictable and indiscriminate in their occurrence (Davis & Lehman, in press). An alternative view, analogous to that put forth about cognitive and emotional processing in the face of positive and negative events more generally (see, e.g., Taylor, 1991), is that people simply tend to process in great depth negative life events. Just as people may be driven to understand why negative events happen (e.g., Wong & Weiner, 1981), they may also be compelled to think about how such events might not have happened. Future Directions The present findings open several new avenues for future research. The study of counterfactuals, in contradistinction to causal attributions, offers a new framework within which to study the phenomena of self-blame and guilt. Is the innocent victim's apparent willingness to accept at least some personal responsibility largely due to the salience of a self-implicating counterfactual? If so, then as Downey et al. found (1990; see also Brewin, 1985), causal attributions may be somewhat less important to those experiencing such events than past research has suggested. Further, it seems important from a more clinical standpoint to investigate if, and under what circumstances, counterfactual thoughts should be discouraged. Whether counterfactual thoughts should be confronted (e.g., by exposing the counterfactual fallacy, Miller & Turnbull, 1990) may 87 critically depend on the extent to which the counterfactual (in)action suggests that similar outcomes can be prevented or avoided in the future, or the extent to which counterfactuals represent efforts to maintain fundamental worldviews. The data also suggest that it may be fruitful to examine individual differences in people's propensity to undo. Although the counterfactual literature has focused almost exclusively on the role of situational factors, such as the mutability of antecedents, data presented herein suggest that individual differences also may play a significant role in one's willingness or tendency to consider and dwell on counterfactual alternatives. That is, seemingly irrespective of mutability factors, some people appear particularly unconcerned with counterfactuals whereas others seem especially consumed by such thoughts. Although Study 2 suggests that one's initial level of distress seems to play a role in determining how long one will continue to undo, it is not clear from these data the extent to which one's initial level of distress is a result of individual differences in the ways with which one characteristically responds to negative events, differences in level of distress that pre-date the loss, or unmeasured aspects of the situation that contributed to the sense of loss experienced (e.g., the parent's degree of emotional investment in the child). Nevertheless, some recent work in the counterfactual literature suggests that aspects of neuroticism may relate to both the degree to which people dwell on counterfactuals and the 88 extent to which the focus of their counterfactuals is on their own behavior. For instance, Davis (1991) found that both the general factor of neuroticism (Costa & McCrae, 1989) and a more specific dispositional propensity to ruminate (Trapnell, 1991) related to the persistence of counterfactuals following real-life negative events, as well as the frequency with which subjects imagined they would undo a hypothetical negative event. Similarly, Roese and Olson (1993) have suggested that levels of self-esteem influence the focus of counterfactuals. They found that people with self-esteem deficits were more inclined to focus on their own failures in counterfactuals generated following a role-played negative event relative to people high in self-esteem, who tended to focus their counterfactuals on the behavior of others. If people with high self-esteem are less likely to focus on actions that they might (or might not) have performed to prevent the negative event from occurring, they also may be less likely to experience feelings of guilt and distress. Boninger et al. (1994) have suggested that another individual difference factor may also play a role in counterfactual thinking. Boninger et al. (1994; Strathman, Gleicher, Boninger, & Edwards, 1994) have suggested that some individuals are dispositionally oriented toward focusing on what they can learn from negative events. People high on their "Consideration of Future Consequences" (CFC) Scale are inclined to focus on distant as opposed to immediate consequences of their behavior and tend to be optimistic about their future (Strathman 89 et al., 1994). With respect to counterfactuals, and the emotions they evoke, Boninger et al. (1994) have found that people high on their CFC Scale are likely to report fewer feelings of regret and self-blame following a role-played, hypothetical negative event than people low on their scale. The researchers suggest that although people high on their CFC Scale may be just as likely to think counterfactually as those scoring low on the scale, the former may be more inclined to shift their focus from "what might have been" (which arouses feelings of regret and self-blame) to the more optimistic thoughts of "what may be." In summary, given a negative outcome with mutable antecedents, it seems that most people will think counterfactually of ways the outcome could have been avoided. Nevertheless, the data suggest that whereas some people are rather quick to put these thoughts aside (perhaps by turning their attention to what may be learned from the situation; see Boninger et al., 1994), others persistently dwell on the counterfactual world of "what might have been," and continue even years afterward to experience feelings of guilt, self-blame, and distress. To the extent that this latter group differs from the former group on the basis of aspects of personality, it may be useful to identify the psychological profile of individuals particularly at-risk to suffer the longterm negative emotions associated with counterfactual thinking. Despite the paucity of research focusing on individual differences in counterfactual thinking, particularly among people undoing real-life events, a 90 speculative profile may be advanced. First, the present data suggest that people who are depressed (or with elevated levels of depressive affect) may be particularly susceptible to dwell on counterfactual alternatives to reality. Feelings of negative affect not only hinder one's ability to dispel or disrupt aversive and intrusive thoughts (Wenzlaff, Wegner, & Roper, 1988), such feelings also are believed to lead one to ruminate more frequently on past events and to focus inordinate attention on one's own thoughts and feelings concerning past negative events, including, presumably, counterfactual thoughts (Greenberg & Pyszczynski, 1986; Pyszczynski & Greenberg, 1987; Roese & Olson, 1993; Wood, Saltzberg, Neale, Stone, & Rachmiel, 1990). However, depression may not be the only individual difference factor that identifies those likely to be most consumed by their counterfactual thoughts. For instance, individuals high on some facets of conscientiousness (such as those high in competence and deliberation) may be especially prone to the counterfactual fallacy of perceiving "what could have been" as "what should have been." People who believe that they generally make prudent, competent decisions, or those who believe that they act cautiously and deliberately, may be particularly self-reproachful when a decision they have made is perceived in retrospect to have (unwittingly) set the stage for the event's occurrence. For these people, counterfactual thoughts that make salient personal behaviors or decisions that could have prevented the outcome from occurring may evoke greater 91 self-reproach because such people may believe that they should have known better.16 In general, the present research underscores the need to consider other aspects of the stressful life event, in addition to its degree of mutability. The specific events with which these respondents were coping had unique qualities that differentiate them from many other traumatic events discussed in the stress and coping literature. In the studies reported here, the events were unexpected and, for the most part, unforeseeable. In addition, the events that these respondents were coping with were acute, and with the exception of the neurologically intact respondents in Study 3, had long-term consequences. It remains to be seen, for example, the extent to which undoing thoughts are reported by people coping with more foreseeable traumatic events (e.g., a marital break-up, loss of a loved one following a lengthy illness) and events with only short-term consequences (e.g., a minor motor vehicle accident). Whereas on some levels the losses that these samples experienced were similar, it should be noted that there are potentially important differences between the populations that they represent. For instance, though the losses in Study 1 and Study 2 were particularly senseless, the actual causes (as well 16Though people high in conscientiousness may be more likely to be self-reproachful for not thinking or behaving appropriately to avoid a negative outcome, they also may be less likely (relative to people low in conscientiousness) to put themselves in situations that might result in negative outcomes to the extent that they are more cautious, deliberate, and capable. 92 as the agents responsible) may be much more salient and comprehensible in Study 1. With SIDS, far greater causal ambiguity and uncertainty exists. The fear of negligence, and the implied (if not overt) accusations of others, may lead parents to be more concerned with, and more threatened by, counterfactual possibilities than would be the case if the causes were better understood. On this point, it is worth noting that we know very little about the interpersonal aspects of counterfactual thinking. Although some scenario research suggests that people tend to be sympathetic to individuals whose negative experiences are easily imagined otherwise (Miller & McFarland, 1986), other research suggests that one may be judged more harshly if counterfactuals implicating one's own behavior are made salient (Branscombe, Owen, Garstka, & Coleman, 1993). Future research will need to examine whether people coping with traumatic events routinely share their counterfactuals with others, and with what specific consequences. Conclusion By taking the study of counterfactuals out of the laboratory, where subjects respond to hypothetical scenarios that do not actually involve themselves, and into the real world, where people are actively re-hashing very real counterfactual alternatives, this research has lent support to some laboratory findings yet also identified a number of potential limitations of the literature. This research suggests, for instance, that we in 93 fact still know relatively little about the concept of mutability. How do people select from among antecedents the incident that they will undo? The present research suggests that the selection of antecedents to undo is not merely based on a search for the most significant causal or exceptional antecedent. What this research suggests as an important factor is the extent to which the individual had control over the behavior. Still, the rules by which people select among controllable antecedents have yet to be established. This research also offers evidence that in coping with traumatic circumstances, people are not merely reacting to the obvious causal or factual aspects of the event. It is not enough to know the cause of one's injury; people also want to know how such accidents could have been avoided (Miller & Turnbull, 1990). Although this research demonstrates that, in general, most people are concerned with the counterfactual thoughts that might have prevented their traumatic event, and that such thoughts are associated with greater distress, blame, and feelings of guilt, we still understand very little about why people are so concerned with this hypothetical, or counterfactual, world of "what might have been." Are these thoughts motivated by a need for psychological control or understanding over devastating events, or are such thoughts spontaneous, initially uncontrollable responses to trauma that must be blocked and controlled if one is to cope efficaciously? Finding answers to these important questions will undoubtedly require researchers to continue 94 conducting counterfactual research in the field as well as in the laboratory. 95 References Abbey, A. (1987). Perceptions of personal avoidability versus responsibility: How do they differ? Basic and Applied Social Psychology, 8, 3-19. Abrams, R. D., & Finesinger, J. E. (1953). Guilt reactions in patients with cancer. Cancer, 6, 474-482. Baker, S. P., O'Neill, B., & Karpf, R. S. (1984). The injury fact book. Lexington, MA: Lexington Books. Bard, M., & Dyk, R. B. (1956). The psychodynamic significance of beliefs regarding the cause of serious illness. Psychoanalytic Review, 43. 146-162. Bard, M., & Sangrey, D. (1979). The crime victim's book. New York: Basic Books. Bergman, A. B., Beckwith, J. B., & Ray, C. G. (Eds.). (1970). Sudden Infant Death Syndrome. Proceedings of the Second International Conference on causes of sudden deaths in infants. Seattle, WA: University of Washington Press. Boninger, D. S., Gleicher, F., & Strathman, A. (1994). Counterfactual thinking: From what might have been to what may be. Journal of Personality and Social Psychology, 67, 297-307. Bradburn, N. M. (19 69). The structure of psychological well-being. Chicago: Aldine. Branscombe, N. R., Owen, S., & Allison, J. A. (1994). Effects of counterfactual thinking on adjustment in rape victims. Unpublished manuscript, University of Kansas, Lawrence, KS. Branscombe, N. R., Owen, S., Garstka, T. A., & Coleman, J. (1993). Judgment conseguences of mental simulation: Who might have done otherwise and would it have changed the outcome. Unpublished manuscript, University of Kansas, Lawrence, KS. Brewin, C. R. (1985). Depression and causal attributions: What is their relation. Psychological Bulletin, 98, 297-309. Bulman R. J., & Wortman, C. B. (1977). Attributions of blame and coping in the "real world": Severe accident victims react to their lot. Journal of Personality and Social Psychology, 35, 351-363. Burgess, A. W., & Holmstrom, L. L. (1974). Rape: Victims of crisis. Bowie, MD: Brady. 96 Chodoff, P., Friedman, S. B., & Hamburg, D. A. (1964). Stress, defenses and coping behavior: Observations in parents of children with malignant disease. American Journal of Psychiatry. 120. 743-749. Cohen, J. (1977). Statistical power analysis for the behavioral sciences (Revised Edition). New York: Academic Press. Costa, P. T. Jr., & McCrae, R. R. (1989). NEO-PI/FFI manual supplement for use with the NEO Personality Inventory and the Five-Factor Inventory. Odessa, FL: Psychological Assessment Resources. Davis, C. G. (1991). The undoing experience: Ancedents. consequences, and individual differences. Unpublished Master's Thesis, University of British Columbia, Vancouver, BC. Davis, C. G., & Lehman, D. R. (in press). Counterfactual thinking and coping with traumatic life events. In N. J. Roese & J. M. Olson (Eds.), What might have been: The social psychology of counterfactual thinking. Hillsdale, NJ: Erlbaum. Derogatis, L. R. (1975). The Affects Balance Scale. Baltimore, MD: Clinical Psychometric Research. Derogatis, L. R., Rickels, K., & Rock, A. F. (1976). The SCL-90 and the MMPI: A step in the validation of a new self-report scale. British Journal of Psychiatry, 128, 280-289. Downey, G., Silver, R. C, & Wortman, C. B. (1990). Reconsidering the attribution-adjustment relation following a major negative event: Coping with the loss of a child. Journal of Personality and Social Psychology, 59, 925-940. Dunning, D., & Parpal, M. (1989). Mental addition versus subtraction in counterfactual reasoning: On assessing the impact of personal actions and life events. Journal of Personality and Social Psychology, 57, 5-15. Fincham, F. D., & Jaspars, J. M. (1980). Attribution of responsibility: From man the scientist to man the lawyer. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 13, pp. 81-138). New York: Academic Press. Frazier, P. A. (1990). Victim attributions and post rape trauma. Journal of Personality and Social Psychology, 59., 298-304. Freud, S. (1963). Beyond the pleasure principle. In J. Strachey (Ed.), The standard edition (Vol. 18, pp. 7-64). London: Hogarth Press. (Original work published 1920). 97 Gilovich, T., & Medvec, V. H. (1994). The temporal pattern to the experience of regret. Journal of Personality and Social Psychology, 67, 357-365. Gleicher, F., Kost, K. A., Baker, S. M., Strathman, A. J., Richman, S. A., & Sherman, S. J. (1990). The role of counterfactual thinking in judgments of affect. Personality and Social Psychology Bulletin, 16, 284-295. Greenberg, J., & Pyszczynski, T. (1986). Persistent high self-focus after failure and low self-focus after success: The depressive self-focusing style. Journal of Personality and Social Psychology, 50, 1039-1044. Haddon, W. Jr. (1963) . Alcohol and highway accidents. In Proceedings of the Third International Conference on Alcohol and Road Traffic. London: British Medical Association. Hart, H. L. A., & Honore, A. M. (1959). Causation in the law. London: Oxford University Press. Heider, F. (1958). The psychology of interpersonal relations. New York: Wiley. Hilton, D. J. (1990). Conversational processes and causal explanation. Psychological Bulletin, 107. 65-81. Horowitz, M. J. (1976). Stress response syndromes. New York: Aronson. Janis, I. L. (1971). Stress and frustration. New York: Harcourt Brace Jovanovicn. Janoff-Bulman, R. (1992). Shattered assumptions: Towards a new psychology of trauma. New York: Free Press. Janoff-Bulman, R. (1979). Characterological versus behavioral self-blame: Inquiries into depression and rape. Journal of Personality and Social Psychology. 37, 1798-1809. Janoff-Bulman, R., & Lang-Gunn, L. (1988). Coping with disease, crime, and accidents: The role of self-blame attributions. In L. Y. Abramson (Ed.), Social cognition and clinical psychology: A synthesis. (pp. 116-147). New York: Guilford. Jones, E. E., & Nisbett, R. E. (1972). The actor and the observer: Divergent perceptions of the causes of behavior. In E. E. Jones, D. E. Kanouse, H. H. Kelley, R. E. Nisbett, S. Valins, & B. Weiner (Eds.), Attribution: Perceiving the causes of behavior. Morristown, NJ: General Learning Press. 98 Kahneman, D., & Miller, D. T. (1986). Norm theory: Comparing reality to its alternatives. Psychological Review. 93, 136-153. Kahneman, D., & Tversky, A. (1982). The simulation heuristic. In D. Kahneman, P. Slovic, & A. Tversky (Eds.), Judcrment under uncertainty: Heuristics and biases (pp. 201-208) . New York: Cambridge University Press. Kiecolt-Glaser, J. K., & Williams, D. A. (1987). Self-blame, compliance, and distress among burn patients. Journal of Personality and Social Psychology, 53, 187-193. Landman, J. (1987). Regret and elation following action and inaction: Affective responses to positive versus negative outcomes. Personality and Social Psychology Bulletin. 13. 524-536. Lehman, D. R., Wortman, C. B., & Williams, A. F. (1987). Long-term effects of losing a spouse or child in a motor vehicle crash. Journal of Personality and Social Psychology, 52. 218-231. Lerner, M. J. (1980). The belief in a just world: A fundamental delusion. New York: Plenum. Lipe, M. G. (1991). Counterfactual reasoning as a framework for attributional theories. Psychological Bulletin. 109, 456-471. Mackie, J. L. (1965). Causes and conditions. American Philosophical Quarterly. 2, 245-264. Markman, K. D., Gavanski, I., Sherman, S. J., & McMullen, M. N. (1993). The mental simulation of better and worse possible worlds. Journal of Experimental Social Psychology, 29, 87-109. Mechanic, D. (1977). Illness behavior, social adaptation, and the management of illness: A comparison of educational and medical models. Journal of Nervous and Mental Disease, 165, 79-87. Medea, A., & Thompson, K. (1974). Against rape. New York: Farrar, Straus, & Giroux. Miller, D. T., & Gunasegaram, S. (1990). Temporal order and the perceived mutability of events: Implications for blame assignment. Journal of Personality and Social Psychology, 59, 1111-1118. 99 Miller, D. T., & McFarland, C. (1986). Counterfactual thinking and victim compensation: A test of norm theory. Personality and Social Psychology Bulletin, 12, 513-519. Miller, D. T., & Porter, C. A. (1983). Self-blame in victims of violence. Journal of Social Issues, 3.9(2) , 139-152. Miller, D. T., & Turnbull, W. (1990). The counterfactual fallacy: Confusing what might have been with what ought to have been. Social Justice Research, 4, 1-19. Miller, D. T., Turnbull, W., & McFarland, C. (1990). Counterfactual thinking and social perception: Thinking about what might have been. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 23, pp. 305-331). Orlando, FL: Academic Press. National Center for Health Statistics (1987). Vital statistics of the United States, 1987. McLean, VA: National SIDS Clearinghouse. Pearlin, L. I. & Schooler, C. (1978). The structure of coping. Journal of Health and Social Behavior. 19, 2-21. Pyszczynski, T., & Greenberg, J. (1987). Self-regulatory perseveration and the depressive self-focusing style: A self-awareness theory of the development and maintenance of depression. Psychological Bulletin, 102, 122-138. Radloff, L. S. (1975). Sex differences in depression: The role of occupation and marital status. Sex Roles. 1, 249-265. Roese, N. J. (1994). The functional basis of counterfactual thinking. Journal of Personality and Social Psychology, 66, 805-818. Roese, N. J., & Olson, J. M. (in press). Counterfactual thinking: A critical review. In N. J. Roese & J. M. Olson (Eds.), What might have been: The social psychology of counterfactual thinking. Hillsdale, NJ: Erlbaum. Roese, N. J., & Olson, J. M. (1993). Self-esteem and counterfactual thinking. Journal of Personality and Social Psychology. 65, 199-206. Schulz, R., & Decker, S. (1985). Long-term adjustment to physical disability: The role of social support, perceived control, and self-blame. Journal of Personality and Social Psychology, 48. 1162-1172. Shaver, K. G. (1985). The attribution of blame: Causality, responsibility, and blameworthiness. New York: Springer-Verlag. 100 Shaver, K. G., & Drown, D. (1986). On causality, responsibility, and self-blame: A theoretical note. Journal of Personality and Social Psychology, 50, 697-702. Spitzer R. L. , & Endicott J. (1975) . Schedule for Affective Disorders and Schizophrenia: Lifetime version. New York: New York State Psychiatric Institute. Strathman, A., Gleicher, F., Boninger, D. S., & Edwards, C. S. (1994). The Consideration of Future Consequences: Weighing immediate and distant outcomes of behavior. Journal of Personality and Social Psychology, 66, 742-752. Stroebe, M. S., & Stroebe, W. (1989). Who participates in bereavement research? An empirical study of the impact of health on attrition. Omega, 20, 1-29. Tait, R., & Silver, R. C. (1989). Coming to terms with major negative life events. In J. S. Uleman & J. A. Bargh (Eds.), Unintended thought (pp. 351-382). New York: Guilford. Taylor, S. E. (1991). Asymmetrical effects of positive and negative events: The mobilization-minimization hypothesis. Psychological Bulletin, 110, 67-85. Taylor, S. E., Lichtman, R. R., & Wood, J. V. (1984). Attributions, beliefs about control, and adjustment to breast cancer. Journal of Personality and Social Psychology, 46, 489-502. Taylor, S. E., & Schneider, S. K. (1989). Coping and the simulation of events. Social Cognition. 7, 174-194. Tennen, H., Affleck, G., & Gershman, K. (1986). Self-blame among parents of infants with perinatal complications: The role of self-protective motives. Journal of Personality and Social Psychology, 50, 690-696. Trapnell, P. (1991). [Unpublished Data]. University of British Columbia, Vancouver, BC. Trieschmann, R. B. (1978). The psychological, social, and vocational adjustment in spinal cord injury: A strategy for future research (Final Report #13-P-59011/9-01). Washington, DC: Rehabilitation Services Administration. Turnguist, D. C., Harvey, J. H., & Andersen, B. L. (1988). Attributions and adjustment to life threatening illness. British Journal of Clinical Psychology, 27, 55-65. 101 Walster, E. (1966). The assignment of responsibility for an accident. Journal of Personality and Social Psychology, 3.^  73-79. Weiner, B. (1986). An attributional theory of motivation and emotion. New York: Springer-Verlag. Weiner, B., Graham, S., & Chandler, C. (1982). Pity, anger, and guilt: An attributional analysis. Personality and Social Psychology Bulletin, 8, 226-232. Weiss, R. S. (1988). Loss and recovery. Journal of Social Issues. M(3) , 37-52. Wells, G. L. , & Gavanski, I. (1989). Mental simulation and causality. Journal of Personality and Social Psychology, 56, 161-169. Wells, G. L., Taylor, B. R., & Turtle, J. W. (1987). The undoing of scenarios. Journal of Personality and Social Psychology, 53, 421-430. Wenzlaff, R. M., Wegner, D. M., & Roper, D. (1988). Depression and mental control: The resurgence of unwanted negative thoughts. Journal of Personality and Social Psychology, 55. 882-892. Wong, P. T. P., & Weiner, B. (1981). When people ask "why" questions, and the heuristics of attributional search. Journal of Personality and Social Psychology, 40, 650-663. Wood, J. V., Saltzberg, J. A., Neale, J. M., Stone, A. A., & Rachmiel, T. B. (1990). Self-focused attention, coping responses, and distressed mood in everyday life. Journal of Personality and Social Psychology, 58, 1027-1036. Wortman, C. B. (1976). Causal attributions and personal control. In J. Harvey, W. Ickes, & R. F. Kidd (Eds.), New directions in attribution research (Vol. 2, pp. 23-52). Hillsdale, NJ: Erlbaum. Wortman, C. B., & Silver, R. C. (1987). Coping with irrevocable loss. In G. R. VandenBos & B. K. Bryant, (Eds.), Cataclysms, crises, and catastrophes: Psychology in action (Master Lecture Series, Vol. 6, pp. 189-235). Washington, DC: American Psychological Association. Yarkony, G. M., Roth, E. J., Heinemann, A. W., Wu, Y., Katz, R. T., & Lovell, L. (1987). Benefits of rehabilitation for traumatic spinal cord injury: Multivariate analysis in 711 patients. Archives of Neurology, 44, 93-96. 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items