Redress and Conscientiousness in the Attribution of Responsibility for Accidents1. &LLY. G. SHAVER. The College of William and Mary,. Wi&msburg,. Virginia.
of Responsibility &LLY The
for Accidents1 G.
Two experiments were conducted to test the possibility that greater attributed responsibility to persons potentially at fault for severe accidents reflects a greater perceived necessity for compensation of the victim of the accident, Experiment I was a 2 X 2 factorial design in which the severity of the accident’s consequences and whether or not the stimulus person bad insurance that compensated the victim were varied. Contrary to previous &dings (Walster, 1966), there were no differences in attributed responsibility based on the severity of the consequences. Regardless of the severity, less responsibility was attributed when the stimulus person had insurance that provided redress than when he carried no insurance. Experiment II showed that this latter finding was not due to provision of compensation, but rather to the fact that the stimulus person had carried insurance. Possible reasons for the failure to replicate findings of severity-dependent attributed responsibility were discussed. When one person hears of a serious accident that might have been caused by another person, what factors affect the former’s attribution of responsibility for the mishap? Surely one of the important considerations, especially in a culture that places a high value on both property and life, is the severity of the accident’s consequences. Thus it is not surprising that a recent study by Walster (1966) showed more responsibility to be assigned to a person potentially at fault for an automobile accident when the consequences of the accident were severe than when they were mild. In that study subjects evaluated the responsibility of a young man whose parked and empty automobile rolled down a hill, resulting in damage as inconsequential as a dented fender or as grave as demolishing the car and injuring two persons. Walster interpreted her results as evidence for the existence of a tendency to avoid admitting that a serious accident is “the kind of thing that could happen to anyone.” She reasoned that for a person to make that admis‘This research was supported in part by NSF Grant G-8857 to Duke University, Edward E. Jones, principal investigator. 100
sion implied that an accident of equal gravity could happen to him. Presumably the subject would feel protected from such an eventuality if he could assert that the stimulus person was in some unique way to blame; implying that had he, the subject, been confronted with the same circumstances, he could have avoided the occurrence. Walster considers this reassurance essentially a two-component process, First, responsibility for the accident is assigned to the person involved, then there are attempts to differentiate the self from that person. If this characterization of the process is correct, it would seem particularly important for the subject to assert that he is much more careful than person potentially responsible for the accident. Assuming that the subjects’ self-perceptions were relatively constant across conditions, the stimulus person should have been seen as much less careful in the case of the severe accident than in the case of the trivial accident. The results, however, showed no such differences in attributed carefulness. The subjects in the severe conditions did appear to be using more stringent “moral standards” in judging the stimulus person’s behavior prior to the XX dent, and this finding, coupled with the lack of differences in attribute care, suggests an alternative interpretation of her findings. The need to attribute responsibility for an accident might not be so much of an irrational defense against accidents as a rational reaction based on moral and legal tradition. When the consequences of an accident are slight, it matters veiy little who pays for the damage. When the consequences of an accident are severe, however, assignment of legal and psychological responsibility becomes of paramount importance, an is the first step in providing restitution for whatever damage was incurred. If this interpretation is correct, then where it is clear that redress has been provided, the attributed responsibility should be greatly decreased. Specifically, subjects in Walster’s (1966) study were told that the young owner of the car “didn’t have any insurance at the time.” Quite possibly this statement alone was sufficient to engage responsibility attribution directed toward providing restitution. To test this possibility, Experiment I was conducted as an altere replication of the Walster experiment in which severity of consequences (MILD-SEVERE) and presence or absence of compensatio insurance (INS-NO INS) were varied in a 2 x 2 factorial keep the stimulus material similar to that used by Walster, the compensation-through-insurance manipulation was accomplished by stating either that the stimulus person had no insurance at the time ( as Walster stated), or that the stimulus person’s insurance had paid all the costs involved, Assuming that Walster’s results could be explained adequately by the notion of redress-oriented attribution of responsibility, two predictions
were made for Experiment I. First, regardless of the severity of the consequences of the accident, less responsibility should be assigned to the person potentially at fault when there is restitution than when there is not. Second, there should be an interaction between presence or absence of insurance compensation and consequence severity such that Walster’s finding of severity-dependent attribution of responsibility should be replicated in the NO INS conditions, but should disappear in the INS conditions. EXPERIMENT
Method Subjects. The subjects were 55 undergraduates at Duke University (33 males, 22 females) who participated in the experiment as part of an introductory psychology course requirement. The subjects were run in two large mixed groups on consecutive days, and within each session the four experimental conditions were randomly assigned. One male subject (in the SEVERE-INS condition) was excluded from the analysis, as his questionnaire responses were unscorable. Thus, the results for Experiment I are based on a total sample of 54. Procedure. When the subjects arrived they were seated in alternate seats in a large lecture room. The experimental communications were contained in identicalappearing booklets that had been previously distributed, according to a random order, across these seats. When all subjects had been seated the experimenter explained a research project on jury deliberation in which he was ostensibly engaged. He stated that in order to study the actual process of deliberation, it was necessary to know the effects of all segments of the testimony that would comprise the case presented to future subjects. For this reason, he continued, each subject iwould read and evaluate only one of these portions of testimony, so that the effect of each segment of testimony could be assessed in isolation from the remainder of the case. After this introduction, the subjects were instructed to turn to their booklets, and read and evaluate the case segment presented therein. The case segments contained in the booklets were the mild and severe consequences communications from the Others Also Suffer conditions of the Walster (1966) experiment, with a single communication in each booklet under the heading: COMPOSITE was involved Information
CASE NO. 76: Lennie B., a high school student whose automobile in an accident in the state of Minnesota. based on testimony of Gordon W., a neighbor of Lenme’s.
For the NO INS conditions the case segment present in the booklet was identical corresponding severity condition of Walster’s study. Thus the MILD-NO ~communication was:
to the INS
Lennie had just bought a car-it was about 6 . . . that was late this summer. years old or so. He and his buddy drove up to Duluth and parked at the top of this hill. Lennie’s buddy said Lennie did set the handbrake, but while they were gone the car started, rolling. Some police who checked the car later said the brake cable was pretty badly rusted and must have broken. Anyway, the car started rolling. If the car had run all the way down the hill, it would have crashed into this
store that’s right at the bottom, and probably hurt either a kid or the grocer that were in the store. But the car didn’t go very far at all. It rolled against an old stump that was sticking out a little way into the street and stopped. The car just got a tiny dent in the front bumper and that’s all. Lennie didn’t have any insurance at the time. In the SEVERE-NO changed, as was Walster’s,
INS condition to read:
the last paragraph
The car might have rolled to a stop against an a little way into the street just in front of where car just missed it and went rolling all the way crashed through the window of this store that’s that was standing at the counter and the grocer. but the grocer was hurt pretty badly. He was in didn’t have any insurance at the time.
of the communication
old stump that was sticking out the car was parked. Instead, the down the hill. The car really right at the bottom. It hit a kid The kid was just dazed a little, the hospital all last year. Leimie
For both the MILD and SEVERE consequences, the manipulation of presence of insurance compensation was accomplished by changing the last sentence of the last paragraph in each case from “Lennie didn’t have any insurance at the time,” to “Fortunately Lenme’s insurance paid all the costs involved.” Included in the questionnaire following the case segment were Walster’s questions about Lennie’s responsibility: the subject’s certainty that Lennie set the handbrake, turned the car’s wheels to the curb, and had bad the brakes checked. Subjects were also asked how morally wrong they felt it was not to have automobile insurance, how conscientious they believed Lennie to be, and how likable they thought he was. Finally, subjects were asked to rate the severity of the accident, to assess the similarity between Lenme’s “beliefs and values” and their own, and to estimate the possibility that they might find themselves in Lennie’s circumstances. All of these items were scored as %-unit scales.
Results On the assumption that Walster’s (1966) finding of severity-dependent attribution of responsibility was in fact a manifestation of the nee restitution, it was predicted that much less responsibility would be assigned where Lennie’s insurance paid all the costs involved than where he had no insurance. It was further predicted that there would be an interaction between presence or absence of insurance compensation severity such that in the INS conditions there would be no differ between MILD and SEVERE consequences (bsecause restitution been made), but that in the NO INS conditions Walster’s findings would be replicated. The data for attributed responsibility in Experiment I are shown in Table 1, and provide strong support for the Grst prediction: les sibility was assigned to Lennie when he had insurance that compensation than when he had no insurance (F = 9.591,1 a p < .Ol ) . The predicted interaction between severity and insurance compensation failed to emerge, however (F < l,O), as the means for
TABLE 1 MEAN SCORES FOR ATTRIBUTED RESPONSIBILITY, PRECAUTIONARY MEASURES, AND ADDITIONAL PERCEPTIONS OF THE STIMULUS PERSON BY CONDITIONS -IN EXPERIMENT I No insurance
Insurance Mild N = 10
Measure Responsibility” Brakes checked” Handbrake set Wheels turned Conscientiousnessc Liking Personal similarity Q The higher b The higher c The higher similarity.
11.30 8.60 19.70 11.50 13.30 14.90 13.70
the mean the greater the mean the greater the mean the greater
Severe N = 15 9.93 7.13 17.93 7.33 12.07 14.07 13.67
Mild iv = 12 15.33 6.08 17.67 9.17 10.33 12.83 8.83
Severe N = 17 15.00 6.82 17.41 6.35 8.65 13.24 12.00
the attributed responsibility. the certainty that the precaution was taken. the rated conscient,iousness, liking, and personal
MILD and SEVERE in the NO INS condition were virtually identical. This failure to replicate Walster’s findings was surprising, because lof the impressive overall MILD-SEVERE difference in the perceived severity of the accident (F = 24&45, 1 and 50 df, p < 901). It is interesting to note, especially in light of the attributed responsibility, that the presence or absence of insurance does not seem to have altered the perceptions of Lennie’s behavior prior to the accident. Thus, Table 1 shows no differences produced by presence or absence of insurance on any of the “precautionary measure” dependent variables-setting the handbrake, turning the wheels toward the curb, or having the brakes checked. The only reliable difference for these three precautionary measures was a main effect for severity on the judgment of whether the car’s wheels were turned toward the curb. Subjects in the SEVERE condition were less certain than those in the MILD condition that Lennie had turned the wheels in toward the curb (F = 5.69, 1 and 50 df, p < ,025). This finding may simply reflect a difference in the objective content of the experimental communications: It is physically impossible for an automobile to roll straight down a hill (as it did in the SEVERE conditions) when its wheels are turned in toward the curb. With respect to additional perceptions of the stimulus person, in the INS conditions, relative to the NO INS conditions, Lennie was judged to be more conscientious (F = 11.93, 1 and 50 d#, p < .Ol); more likable (F = 8.80, 1 and 50 df, p < .Ol); and more similar in personal character-
istics to the subject (F = 12.82, 1 and 50 df, p < .OI). For this last measure, the imputed similarity between Lennie’s “beliefs and values” those of the subject, there was an effect based on the severity of consequences. Within the NO INS conditions considered alone, subjects thought themselves more similar to Lennie when the consequences were SEVERE than when they were MILD (t = 2.24, 27 df, p < -05, twotailed). There were no other effects on any dependent variable for severity of the accident’s consequences, but there was one additional effect for presence or absence of insurance compensation. Subjects in the INS conditions thought it more likely that they might find themselves in Lennie’s circumstances than did the subjects in the NO INS conditiion (F = 9.83, 1 and 51) df, p < .(El). The results of Experiment I appear to confirm the original prediction, showing less responsibility attributed in the presence of redress to the victim. As noted above, however, to preserve the similarity between the replication (NO INS) conditions and Walster’s original materials while at the same time including restitution to the victim (INS conditions), it was stated in the latter that redress had been provided by the stimulus person’s insurance. Because of this construction of the conditions, subjects in the INS conditions received not just one, but two more bits of information than their counterparts in the NO INS conditions: that the stimulus person had carried insurance, and that his insurance had paid for the damage. Thus it is impossible to determine whether the difference between the INS and NO INS conditions in Experiment I was due to fact that the stimulus person had carried insurance, or to the fact that victim had been compensated, Experiment II was conducted to distinguish between these two alternatives and incorporated three conditions-Stimulus Person’s Insurance Pays ( SPIP), Other Insurance Pays (OIP), No Compensation (NC)based on the SEVERE conditions of Experiment I. The first condition ( SPIP) was identical to the SEVERE-INS condition of Experiment I, and the NC condition was identical to the SEVERE-i% INS condition of Experiment I. The OIP condition was constr that the subjects learned that the victim had been compensated, a it was stated that the stimulus person did not carry insurance: “‘Lennie didn’t have any insurance at the time, but the grocer’s own insurance paid for all the costs involved.” Thus if restitution was the important determinant of the results of Experiment I, the attribution of responsibility in the OIP condition should be low and approximately equal to that in the SPIP condition, because in both of these conditions the victim is compensated. If, in co trast, the fact that Lennie carried insurance had produced the results 0
Experiment I, then the attribution of responsibility in the OIP condition should be high and approximately equal to that in the NC condition, since in both of these conditions Lennie is reported to have no insurance. Finally, if both Lennie’s carrying insurance and victim compensation contributed to the results of Experiment I, then the attribution of responsibility in the OIP condition should fall approximately halfway between the levels of attribution for the other two conditions. EXPERIMENT
Method Subjects. The subjects were 48 students (30 females, 18 males) enrolled in an introductory social psychology course at the College of William and Mary. The experiment was conducted during the second meeting of the class, and although the students were primarily juniors and seniors, none had ever before participated in a social psychological experiment involving deception. -Extensive debriefing revealed that no subject had correctly guessed the true nature of the experiment, but two subjects (one male, one female) were eliminated from the analysis because they were graduate students in a related field of study. Procedure. The subjects were run in a single large group, and within this group there was random assignment to the three experimental conditions. The verbal instructions were identical to those of Experiment I, and the case segments contained in the booklets were either the same as (SPIP, NC) or different from (OIP) the SEVERE case segment of Experiment I as described above. The questions in the booklet following the case segment did not include the assessment of imputed similarity, but were otherwise identical to those employed in Experiment I.
Results It will be recalled that less responsibility was attributed in the INS conditions of Experiment I than in the NO INS conditions. Were this finding based on victim compensation, then in Experiment II, equal and low responsibility should be assigned in the SPIP and OIP conditions where redress is provided. In contrast, had the results of Experiment I been produced by Lermie’s currying insurance regardless of redress, then the attribution of responsibility in the OIP condition should be quite high and equal to that in the NC condition. A glance at Table 2 shows the latter interpretation to be correct. The mean responsibility attributed to the stimulus person in the OIP condition is virtually identical to that in the NC condition; both are significantly higher than the responsibility attributed in the SPIP condition (F for the orthogonal comparison = 18.81, 1 and 43 df, p < .OOl). Table 2 f ur th er shows that essentially the same pattern holds for attributed conscientiousness (F for the orthogonal comparison = 9.58, 1 and 43 df, p < .Ol). There were no other significant differences on any dependent variables in this analysis. Preliminary inspection of the data, however, had indicated that there
FOR ATTRIBUTED RESPONSIBILITY, PREC-~UTIONARY MEASURES, AND LIKING BY CONDITIONS IN EXPERIMENT II
Stimulus person’s insurance pays N = 17
Other insurance pays IV = 15
Responsibilitya Brakes checkedb Handbrake set, Wheels turned Conscientiousnesv Liking
9.35 9.16 16.41 9.88 13.88 14.00
17.53 6.53 14.93 6.40 10.27 13.13
Q The higher 6 The higher e The higher
the mean the mean the mean
the greater the greater the greater
compensation N = 14 17.86 7.36 19.50 7.93 9.79 12.50
the attributed responsibility. the certainty that the precaution was taken. the rated conscientiousness and liking.
might be important differences based on the sex of the subjects, and so the data were also analyzed as a 2 x 3 (sex by conditions) factored design. Only the means for dependent variables showing significant elects are presented in Table 3. As was the case for the previous analysis, less responsibility was attributed when Eennie carried insurance than when he did not (F for Conditions = 13.84, 2 and 40 df, p < . but there was also an interaction between Conditions and Sex such this difference was exaggerated for the male subjects (F = 4.32, 2 40 df, p < ,025). For the Conscientiousness-dependent variable, there was only a main effect for Conditions (F = 4.14, 2 and 40 djF, ip < .O25) wi no effects attributable to Sex. TABLE Mh.4~
FOR ATTRIBUTED CONSCIENTIOUSNESS OF SUBJECTS
St,imulus person’s insurance pays
RESPONSIBILITY, BY CONDITIONS EXPERIMENT
BRARITZ CHECKED, AND SEX
Other insurance pays
Responsibility’ Brakes checkedb ConscientiousnesP
3.75 13.25 14.50
11.08 7.92 13.69
19.17 6.50 11.50
16.44 6.56 9.44
a The higher b The higher c The higher
the mean the mean the mean
the greater the greater the greater
the attributed responsibility. the certainty that the precaution the rated conscientiousness.
No compensation Male N=7
20.57 8.71 9.43
15.14 6.00 10.14
The treatment effects for the Brake Check-dependent measure were significant in this analysis (F for Conditions = 4.25, 2 and 40 cZ~, p < .025) as the error variance was greatly reduced. There was also a significant effect for Sex, with males being more sure that the brakes had been checked (F = 492, 1 and 40 df, p < .05). There were no differences on any of the remaining dependent measures. DISCUSSION
The results of Experiment I showed that less responsibility would be assigned to a person potentially at fault for an accident when he carried insurance that compensated the victim than when he carried no insurance, regardless of the severity of the consequences of the accident. But was this lenience of judgment produced by the provision ~of redress to the victim, or was it simply the result of the person’s having carried insurance? Experiment II indicated, surprisingly, that compensation could not have been the key factor, since as much responsibility was assigned when there was compensation but the stimulus person carried no insurance (OIP condition) as when there was no redress at all (NC condition). The discussion below will first consider possible reasons for the greater attribution when there was no insurance, and then will attempt to interpret the apparent failure to replicate Walster’s (1966) iinding of severity-dependent attributed responsibility. If provision of redress to the victim of an accident is not considered in the attribution of responsibility for the occurrence, why is the fact that the stimulus person carried insurance so crucial? At least two important explanations suggest themselves. First, having insurance may be related to overall conscientiousness: Subjects whombelieve that the stimulus person carried insurance may perceive him to be, in general, a more conscientious and careful person, and thus attribute this accident to factors beyond his control that he could not have anticipated. Indeed, the conscientiousness attributed in both experiments to the insured stimulus person conforms to this expectation. Were overall carefulness the on2y contributor, however, that generalized impression should also have been manifest in the subjects’ certainty that adequate precautions had been taken. It will be recalled that only one such precautionarymeasure difference (Brakes Checked, Sex-by-Conditions analysis of Ex-periment II) was found. The predominant lack of differences in precautions permits speculation that a second contributor to the attribation of responsibility results could -have been a tendency to “punish” the uninsured stimulus person. Subjects in the OIP condition might have believed that, although the victim ‘had been financially compensated, further retribution was necessary to bring the psychological situation into balance. When Lennie’s own in-
surance pays for the damage, this will presumably affect not only his guilt feelings, but also his future insurance payments. When another’s insurance provides compensation, however, Leunie suffers no financial consequences. It could have been in an attempt to augment Lennie’s psychological burden that more responsibility was assigned. Such a process would seem related to Lerner’s (e.g., 1966) notion that observers need to believe in a “just world” where people “get what they deserve--or, after the fact, deserve what they get.” Although Lerner”s work typically has involved restoration of justice through denigration of a sufferer obviously innocent of responsibility for his plight, there is no a priori reason to assume that such restoration could not also be accomplished through punitive attribution, It remains to further research to pursue the implications of this argument. In Experiment I the NO INS conditions were designed to serve as a conceptual, if not empirical, replication of Walster’s (3.966) study. It is possible that the cover story of Experiment I (testimony evaluation) an the written format of the experimental communication induced in the subjects a rational evaluative “set” not produced by Walster’s procedure. In her experiment, subjects were asked simply to “evaluate stimulus materials,” and the experimental communications were delivered as one of a series of five tape-recorded interviews. Enough additional information may have been provided by the other four interviews on the tape that the subjects were not so alerted to the question of determining responsibility for the accident, and were less prepared to weigh carefully all of the evidence in order to reach a rational decision. In contrast, the procedure of the present Experiment I may have led subjects consci to consider the relation between Lennie’s behavior and the accident independently of the accident’s consequences. Although a difference in set is a plausible explanation of the failure to replicate Walster’s (1966) study in Experiment I, tbere are other failures to replicate whose designs do not admit of this explanation (Walster, 1967; Shaver, 1969). These results are surprising because greater attribution of responsibility for severe accidents would seem to be “overdetermined”-by the self-protective tendency, by the belief in a “just world,” and by a moral necessity to affix blame. Since these considerations suggest that severity-dependent attribution is not .to be jetted out of hand, one problem for future research is to specify conditions under which such attribution occurs. REFERENCES LERNER, M. J. The unjust consequences of a need to believe presented at the meeting of the American Psychological September, 1966.
in a just Association,
SHA~M, K. G. The effects of relevance and severity on the attribution of responsibility for accidents. Unpublished doctoral dissertation, Duke University, 1969. WAUTER, E. Assignment of responsibility for an accident. Journal of Personality and Social
(Received November 20, 1968)