Cognitive Dissonance Theory Has Been Reformulated by Some Authors Since the Original Statement (Festinger, 1957) Ap- Peared in T

Japanese Psychological Research 1980, Vol.22, No.1, 32-41

ATTRIBUTION OF PERSONAL RESPONSIBILITY AND DISSONANCE REDUCTION

HARUKI SAKAI AND KIYOSHI ANDOW 1

University of Tokyo

The relationship between personal responsibility and dissonance reduction was investigated in a situation where an individual had to undergo a negative consequence caused by an agent beyond his control, i.e., by chance. Eighty participants received either strong or weak electric shock, depending on the oddness or evenness of spots of a die thrown either by the experimenter (Ex- perimenter-Caused condition) or by the participant himself (Participant- Caused condition). It was found that participants in the Participant-Caused condition evaluated electric shocks less painful, estimated their heart rates to be less faster, and perceived the experimenter more intelligent and favorable than participants in the Experimenter-Caused condition. These results were interpreted as lending support to the newly reformulated dissonance theory.

Cognitive dissonance theory has been tion may be observed only to the degree reformulated by some authors since the that the individual sees himself as re- original statement (Festinger, 1957) ap- sponsible for bringing cognitions into an peared in the late 1950s. Brehm and inconsistent relationship (p. 10). Cohen (1962) stated that dissonance was These reformulations cited above clearly aroused only when an individual was concerned with the necessary conditions bound by a behavioral commitment to for dissonance arousal, and it may be said, one of the inconsistent cognitions. Aron- as Greenwald and Ronis (1978) pointed son (1968) stressed that the self concept or out, that dissonance theory has evolved some other firmly held expectancy must along a direction of convergence with be involved for dissonance to be aroused ideas from the tradition of self theory. (e.g., Nel, Helmreich, & Aronson, 1969). Thus far several studies have investi- Recently Wicklund and Brehm (1976) gated the relationship between personal have refined the earlier revisions in terms responsibility and dissonance reduction. of the concept of personal responsibility: For example, Sogin and Pallak (1976) ... dissonance arousal requires the per- offered subjects high or low choice for ceptions of a strong causal link between their performing a dull random-number- oneself and the potentially dissonance- writing task. Half of the subjects were arousing event. Unless such a connec- told that task usefulness depended on the tion is perceived, the behavior and con- subject and that task results from others sequence are two irrelevant cognitions as had been useful, while the other half were far as the person is concerned, and the told that task usefulness depended on terms consonance and dissonance are not chance and that task results from others applicable (p. 70). ... Dissonance reduc- had also been useless. This information 1 The authors wish to express their thanks to was intended to vary the extent to which subjects thought that they were responsible Associate Professor Toshiro Suenaga for his helpful comments on the earlier manuscript. Requests for for the usefulness of their production. All reprints should be sent to Haruki Sakai, Depart- subjects later learned that their task results ment of Psychology, University of Tokyo, Bunkyo- were useless (i.e., negative consequences). ku, Tokyo, 113 Japan. Postexperimental ratings of the number Responsibility & Dissonance Reduction 33

task revealed that the task was seen as cepting or rejecting the personal responsi- more enjoyable in the high than in the bility may be used as a mode of dissonance low-choice conditions, but only when sub- reduction, although there are some meth- jects were led to feel responsible for the odological difficulties to test this idea consequence. Thus these results clearly strictly. These hypotheses are:(a) Even demonstrate that personal responsibility when conditions are not right for inducing for negative consequences affects dis- responsibility, if the cognition reflecting a sonance reduction processes. Similarly, negative consequence is highly resistant to Cooper (1971) has suggested that counter- change, personal responsibility may be attitudinal behavior will produce dis- accepted retrospectively so that the nega- sonance if the actor feels personally re- tivity of the event could be ameliorated by sponsible for the consequences of the means of dissonance reduction;(b) even behavior. According to Cooper (1971), when conditions are right for inducing personal responsibility is accepted if the responsibility, if the cognition reflecting a person had free choice in performing the negative consequence is not highly resistant behavior and the person was able to foresee to change, personal responsibility may be its negative consequences. rejected retrospectively to reduce or avoid In some situations, however, a person possible dissonance. In order to help would feel personally responsible for a test the latter two hypotheses, the present consequence, even if there is no sufficient experiment included the variable of" an- ground for that feeling from an objective ticipated duration of the negative con- point of view (see Wortman, 1975). The sequence" which was expected to vary main purpose of the present study was to the resistance-to-change of cognition re- investigate whether the concept of per- flecting the negative consequence. sonal resposibility suggested by Wicklund An additional purpose of the present and Brehm (1976) could be extended to study was to examine the impact of such a situation where an individual has cognitive dissonance on physiological vari- to undergo a negative consequence caused ables. There is now fairly good evidence by an agent beyond his control, i.e., by that the cognitive changes specified by chance. Wicklund and Brehm's (1976) dissonance theory are eventually mani- formulation leads to the following hy- fested in physiological changes as well. pothesis with respect to a negative conse- For example, Zimbardo, Cohen, Weisen- quence caused by chance: If a person berg, Dworkin, and Firestone (1969) has a feeling of personal responsibility for found that a subject who committed him- a negative consequence even if he did not self to receiving electric shock reduced his bring it about objectively, dissonance will dissonance by lowering his felt level of be aroused. Hence the person will try to pain. In addition, they demonstrated that reduce the dissonance mainly by, in this GSR responses were lower for the high- case, underestimating the negative con- dissonance than the low-dissonance sub- sequence itself. Wortman (1975) found jects. Glass and Mayhew (1969) replicated that merely causing a chance outcome Zimbardo et al.'s (1969) findings in a and having foreknowledge about the con- similar situation. Furthermore, Brehm, sequence induces a feeling of personal Back, and Bogdonoff (1964) found that responsibility. Accordingly, the Wort- dissonance altered hunger level, and that man's (1975) procedure can provide a it was reliably related to changes in the necessary basis for testing the hypothesis level of free fatty acids (FFA). In the mentioned above. present study, the participants' pulse rates A second purpose of the present study were monitored throughout the experi- was to examine the possibility that ac- ment to examine whether the changes 34 H. SAKAI AND K. ANDOW noted earlier in cognitions could be ex- METHOD tended to the noncognitive components of pain. Measuring pulse rates was also Participants. Eighty-three male under- expected to make the" cover story" of graduates at the University of Tokyo the experiment more plausible for the participated individually in the experi- participants. ment entitled" a study of the effects of Theoretical expectations described thus tension state on human behavior." A far were investigated by means of a 2•~ total of three participants were not in- 2•~2 factorial design. In the present excluded in the analyses (one because he periment, electric shock was utilized as a couldn't comprehend the instruction, one negative consequence. The first variable, because he rejected to receive electric causal agent, was whether a die was shock, one because of an equipment mal- thrown by the experimenter (Experimenter- function), leaving a final sample of eighty Caused condition) or by the participants participants. All of them were paid themselves (Participant-Caused condition). 500 yen (approx. 2 dollars) for their The second variable, magnitude of nega- participation at the end of the experiment. tive consequence, was manipulated by the Procedure. When the participant arrived level of electric shock determined by for the experiment, the experimenter in- casting a die (Large-Negative-Consequence troduced himself and escorted him to the versus Small-Negative-Consequence con- experimental room. He was seated at a dition). The third variable, anticipated table facing a small screen on which a duration of negative consequence, was " flow diagram of the experiment" was manipulated by telling the participants written. Then the experimenter gave that a negative task would require either him a general introduction: thirty minutes (Long-Duration condition) The purpose of this study is to in- or only a few minutes (Short-Duration vestigate how tension state affects hu- condition). man behavior. There are, of course, The hypotheses were as follows: many studies which investigated its ef- 1. Participants in the Participant-Caused fects on the speed of learning, per- condition would accept more personal formance level, etc. But now, we are responsibility for the negative consequence, going to examine how conflict will be hence manifest greater dissonance reduc- affected by the tension state. Conflict tion (e.g., underestimation of the negative will be measured by the Stroop Color- consequence and self-justificatory attitude Word Test which I'll explain later in change) than participants in the Experi- detail. Please look at the sheet in front menter-Caused condition. of you. It shows the flow diagram of 2. Within the Experimenter-Caused con- the experiment. First, participants are dition, participants in the Long-Duration divided into either a" high-tension condition would attribute more personal group" or a" low-tension group". In responsibility, hence show greater dis- order to assign participants randomly, sonance reduction than participants in the we're going to use a die. This pro- Short-Duration condition. cedure will allow us to examine how 3. Within the Participant-Caused con- the nature of the conflict will differ dition, participants in the Short-Duration between the two groups. The level of condition would accept less personal re- the tension state will be manipulated sponsibility, hence show less dissonance by the strength of electric shocks ad- reduction than participants in the Long- ministered to the participant's finger. Duration condition. That is, those who are assigned to the " low-tension group" will receive very Responsibility & Dissonance Reduction 35

weak shocks, whereas those who are as- condition" or a" Small-Negative-Con- signed to the" high-tension group" sequence (SNC) condition". will receive somewhat stronger shocks. Then, the experimenter explained the In order to minimize the induction of experimental task. Simplified version of anxiety for receiving electric shocks and the Stroop Color-Word Test, developed the" evaluation apprehension"(Rosen- by the present authors, consisted of four berg, 1965), the experimenter continued: color nouns (" ao (blue)"," midori You don't need to concern about the (green)"," aka (red)"," kiiro (yellow)") electric shock you are going to receive, printed in incongruent combination of as it is not so strong as to do any harm noun and color ink. Thus, the noun to the tissue of your skin. And as I " aka (red)" was printed in blue or have already mentioned, the purpose of green ink, etc., and participants were the present study is to investigate the required to pronounce the color of ink effects of the tension state by comparing rather than the color noun. One session the" mean conflict scores" of the two of the Stroop task consisted of twelve groups. So I assure you that any data color-pronouncing trials, and the partici- taken from this experiment will be used pant was presented with a sample card for statistical purposes only. Of course, on which four sets of color words were we do not intend to measure your own printed. Participants were informed that personality or intelligence level via they would receive an electric shock which this task. lasted 2 s just before each session of Then, the experimenter told the partici- the Stroop task. Furthermore, they were pant the rationale for attaching the elec- required to start to pronounce the color trodes to his finger: of ink as soon as they felt the electric Well, while you're engaging in the shock. Then, the experimenter gave an task, your pulse rate will be monitored instruction which constituted the mani- by the apparatus you can see over pulation of the anticipated duration of the there. As I need a few minutes to negative consequence. In the Long-Dura- calibrate our recording instruments, will tion (LD) condition, he said: you let me attach the electrodes to your Well, you're required to perform finger right now? thirty-two sessions. It'll take you about After adjusting the equipment, manipula- half an hour, although there are some tions of the causal agent and the magnitude individual differences. of negative consequence were introduced: In the Short-Duration (SD) condition, Let's decide which group you will be participants were informed that the task assigned by casting a die. If the die consisted of four sessions and it would take shows the spot(s) of odd numbers, you them only a few minutes. will be assigned to the low-tension The experimenter then told the partici- group"; in case of even numbers, you pant the rationale for delivering a sample will be assigned to the" high-tension shock, the participant's estimation of which group". constituted a major dependent variable: For half of the participants, the die was Let me give you a sample shock thrown by the experimenter (Experi- before the task. I think it's better for menter-Caused condition), while for the you to get an idea of what the shock other half, it was thrown by the participant will be like, and I'd also like to check himself (Participant-Caused condition). how your pulse rate will change after According to the number of the spot(s), receiving the shock. they were assigned" randomly" to either Following this instruction, he taped the a" Large-Negative-Consequence (LNC) shock electrodes to the participant's finger 36 H. SAKAI AND K. ANDOW and delivered a sample shock of either (" Extremely responsible"). The partici- .70 mA (LNC condition) or .45 mA pants were also asked to indicate how they (SNC condition). The shock lasted 2.0 s. perceived the experimenter on 7-point Then, the participant was presented with semantic differential-type scales anchored a first questionnaire which asked his own by the following bi-polar adjective pairs: estimation of the shock. Next, the partici- Intelligent-unintelligent, rude-polite, pant was required to perform one session warm-cold, trustworthy-untrustworthy, of the Stroop task" in order to acquaint favorable-unfavorable. The other ques- him with the procedure required for the tions asked them to assess the perception task." Following the exercise, the ex- of the Stroop task (" How enjoyable is the perimenter casually mentioned that he task you are going to perform?"), the would like the participant to fill out a perception of the present experiment itself second questionnaire before beginning the (" How much contribution do you think actual pronunciation task. It was de- this experiment can make to the progress signed to assess the effectiveness of the of psychology?"), etc. Responses to responsibility manipulation, perceptions these questions were also made on 7- of the Stroop task, perceptions of the ex- point scales. perimenter, etc., which were expected to be related to dissonance reduction. After RESULTS completing the second questionnaire, the experiment was over. The participant was Responsibility measure. A 2•~2•~2 thoroughly debriefed and sworn to secrecy. ANOVA was performed on the responsi- The debriefing revealed that no partici- bility measure. Contrary to our expecta- pants entertained any suspicions about the tion, the analysis didn't yield a significant experiment and that the" experimental main effect for the Causal Agent factor, realism" (Carlsmith, Ellsworth, & Aron- F(1,72)<1. That is, participants didn't son, 1976) was reliably high. report greater feelings of responsibility in Dependent measures. The first question- the Participant-Caused condition (M= naire contained seven questions which 2.85) than in the Experimenter-Caused were designed to assess the negativeness of condition (M=3.18). It seems too early, the electric shock. For example, Item 1 however, to conclude that the responsi- asked the participants," How painful was bility manipulation was not effective. the shock?" Item 2 asked them," How First, this result might be due to the fact uncomfortable was the shock?" Item 6 that participants were not completely asked them," How much has your heart sure for what they were asked to assess rate got faster after receiving the shock?" the degree of personal responsibility. All responses were made on 7-point scales Actually, some participants asked the ex- labeled from 1 (" Not at all ") to 7 (" Ex- perimenter to clarify the precise meaning tremely"). On each item, the greater the of the question, and some others clearly scale value, the greater the participants miscomprehended it as responsibility for assessed the negativeness of the shock. " accepting to participate in the experi - The second questionnaire contained ment". A second reason, which is related several filler items as well as eight ques- to the first one, may be that the term tions that served as manipulation checks " responsibility" and the corresponding or dependent variables. The responsibility Japanese term "sekinin" (" responsi- manipulation check (" How responsible bility" is commonly translated as such) do you feel you personally are for receiving have somewhat different connotations. the shock of the present level?") ranged The impression of the Japanese term from 1 (" Not at all responsible") to 7 " sekinin" might be too serious in such Responsibility & Dissonance Reduction 37

a context of the present experiment. TABLE I Third, it may simply have been quite Effects of causal agent, magnitude of negative difficult for any participant to answer consequence, anticipated duration of negative consequence on the about personal responsibility for such a estimated heart rate change pure chance event as a result of casting a die. Taking these into consideration, it was determined to conduct further analyses on the dependent variables, although the responsibility manipulation was seem- ingly unsuccessful. Intensity of the electric shock. The actual intensity of the electric shock administered to each participant was checked by the experimenter. It was revealed that the intensity did not differ systematically as a function of the Causal Agent (F<1) or Note. The higher the score, the greater the the Anticipated Duration of Negative estimated increase in heart rate. For all cells, Consequence factor (F<1). n=10. Perceptions of the electric shock. 2•~2•~2 ANOVAs were performed separately on the seven measures of the perceived nega- first hypothesis that individuals who caused tiveness of the electric shock. All analyses a chance negative consequence reduce

yielded significant main effects for the their dissonance by underestimating the Magnitude of Negative Consequence fac- very negativeness of the consequence. tor, Fs(1,72)>6.68, p<.01, which is not ANOVA on the estimated heart rate too surprising. Two out of the seven change measure also yielded a significant measures yielded significant main effects effect for the Anticipated Duration of for the Causal Agent factor. As expected, Negative Consequence •~ Magnitude of

participants in the Participant-Caused Negative Consequence interaction, F(1,72) condition perceived the shock less painful =10.35, p<.01. This is due to the fact than participants in the Experimenter- that the impact of the latter variable was Caused condition (M=2.00 vs. 2.40, re- more pronounced in the LD condition spectively). This result was characterized than in the SD condition. by, though marginally, a significant main The second and third hypotheses, which effect for the Causal Agent factor, F(1,72) concern the mode of dissonance reduction, =2.99, p<.09. The three-way interaction were examined by comparing the means of was also significant, F(1,72)=6.68, p<.05, the SD and the LD conditions in the which was mainly due to the high value Experimenter- or the Participant-Caused in the Experimenter-Caused/SNC/SD cell condition. On the above two measures,

(M=2.70). In addition, as can be seen it was revealed that the differences of in Table 1, participants in the Experi- means between the LD and the SD con- menter-Caused condition estimated that ditions were negligible for the both meas- their heart rates had got more faster after ures: Within the Experimenter-Caused receiving the electric shock than partici- condition, participants in the LD condi-

pants in the Participant-Caused condition tion did not feel the shock less painful

(M=2.15 vs. 1.60, respectively). ANOVA (M=2.30) and did not estimate less in- on this measure yielded a significant main crease in heart rate (M=2.25) than partici-

effect for the Causal Agent factor, F(1,72) pants in the SD condition (Ms=2.50 and =7.41, p<.01. These results support the 2.05, respectively), Fs<1, for the both 38 H. SAKAI AND K, ANDOW

TABLE 2 measures; within the Participant-Caused condition, the differences of means be- Effects of causal agent, magnitude of negative tween the SD and the LD conditions were consequence, anticipated duration of negative consequence on the also insignificant. This pattern of results actual pulse rate change was almost identical among the other five dependent measures. So it can be con- cluded from these analyses that the two hypotheses were not supported by the present data. Perception of the experimenter. 2•~2•~2 ANOVAs were separately performed on the data from the five semantic differential- type scales. As expected, participants in the Participant-Caused condition perceived the experimenter significantly more "intelligent" than participants in the Ex- Note. Positive numbers indicate the increase perimenter-Caused condition (M=5.45 in pulse rate (bpm). vs. 4.80, respectively), F(1,72)=7.28, p< †Two samples in this cell were excludcd from .01 (see Fig. 1). In addition, participants the analysisbecause of the misoperationof the in the Participant-Caused condition per- recordinginstrument. The resultantn was 8. For ceived him significantly more "favorable" the other cells, n=10. than participants in the Experimenter- Caused condition (M=4.68 vs. 4.18, re- were no significant effects of the inde- spectively), F(1, 72)=4.48, p<.05. These pendent variables upon the measures as- results seem to show that participants in sessing the participants' perceptions of the the Participant-Caused condition mani- Stroop task and the "present experiment" fested greater dissonance reduction (i. e., itself, etc. Accordingly, detailed statistical self-justificatory attitude change) than results are not reported here. Physiological measure. The actual pulse participants in the Experimenter-Caused condition. rate data were reduced to make "change The other questionnaire measures. There scores" by the following procedure: First, a base line of the participant's pulse rate was scored over 10 beats immediately preceding the Causal Agent manipulation into the number of beats per minute. Then, a postshock pulse rate was also scored over 10 beats immediately following the onset of the electric shock. Change scores, which were utilized for the statistical analysis, were calculated by simply subtracting the base line scores from the postshock scores. The mean change scores are presented in Table 2. Participants in the Participant-Caused condition showed smaller increase than participants in the Experimenter-Caused FIG. 1. Effects of causal agent, magnitude of condition (M=5.79 vs. 7.50, respectively). negative consequence (MNC), anticipated duration Although the direction of this effect was of negative consequence (ADNC) on the ratings of consistent with that of the estimated heart the experimenter's intelligence. Responsibility & Dissonance Reduction 39

TABLE 3 Summary of the analysis of variance on the estimated heart rate change and actual pulse rate change measures

†Data of two samples were excluded from the analysis bccause of the misoperation of the recording lnstrument. ** p< .01 rate change, a 2•~2•~2 ANOVA on this pant-Caused condition. Thus, it is not measure did not reveal a significant main surprising that participants in the Partici- effect for the Causal Agent factor, F(1, 70) pant-Caused condition evaluated the ex- <1. None of the other main or interac- perimenter more positively than did tion effects were also significant (see participants in the Experimenter-Caused Table 3). condition (Lott & Lott, 1974). Although this explanation by a reinforcement model seems plausible as far as the participants' DISCUSSION AND CONCLUSION perception of the experimenter is con- Personal responsibility and dissonance re- cerned, it is almost impossible to extend duction. The first hypothesis was supported it to cover the obtained differences in the by the present results. Participants in the perceptions of the electric shock. Rather, Participant-Caused condition seem to have these findings, taken in conjunction with reduced their dissonance by underestimat- the results of Wortman (1975) and Zim- ing the negativeness of the event and/or bardo et al. (1969) studies, seem to suggest by evaluating the experimenter more that dissonance reduction did occur but

positively. However, as mentioned in the the measuring instruments were not so- earlier section, the feelings of personal phisticated enough to detect the difference responsibility were not affected by the in the felt level of personal responsibility. causal agent manipulation. This fact What is most important should be, then, may lead to the conclusion that the ob- to construct a reliable instrument for tained results were not due to the process measuring personal responsibility, parti- suggested by Wicklund and Brehm (1976). cularly suitable for Japanese participants. For example, assuming that an individual Further, a more sophisticated experi- is motivated to determine his fate by his mental design should be utilized, because own deed, the behavior of the experi- measuring personal responsibility itself menter, i. e., having allowed the partici- seems to lessen the plausibility of the

pants to cast a die, might have been re- experimental situation. For example, warding to the participants in the Partici- future research can employ the Solomon- 4G H. SAKAI AND K. ANDOW

type experimental design where the par- logical change, which in turn results in ticipants' felt responsibility is " estimated " attitude. In the present study, however, from the actually measured responsibility it was not the case. Participants in the of other " participants " who do not Participant-Caused condition did under- proceed to the measurement of dependent estimate the negativeness of the electric variable(s) (Solomon, 1949; Lana, 1969). shock, but it was not accompanied by the Much research should be done, using such actual pulse rate change. However, it a strict procedure, to identify conditions cannot be determined from the present which affect attribution of personal re- study whether this inconsistency is attri- sponsibility for consequences of one's be- butable to the difference in manipulation havior (Eagly & Himmelfarb, 1978). employed or the difference in response There was no evidence supportive of the mode. That is, the reason for this in- second and third hypotheses. Although consistency may be that the responsibility the check of the manipulation of the anti- manipulation employed in the present cipated duration of negative consequence experiment was not so strong as to affect variable was not included in the present the physiological variable or that the experiment, it is probable that this mani- pulse rate measure was less sensitive than pulation was not completely effective. the GSR measure to detect changes in Since most participants seem to have ex- arousal level after receiving the electric pected that the experiment was to last shock. for more than an hour, participants in the To summarize, the conclusion that Long-Duration condition might take the seems warranted is that the present study instruction concerning the time required identified conditions which affect disso- for the experiment as a matter of course.2 nance reduction processes after a person Thus, the resistance-to-change of cogni- has faced a negative event caused under tion of the negative consequence was re- no personal control. Those who caused a latively low even for the participants in chance negative event themselves mani- the Long-Duration condition. Responsi- fested greater dissonance reduction than bility acceptance, one mode of dissonance those whom the other person caused it, reduction, might be employed if a nega- although the attribution of personal re- tive event should be surely and unchange- sponsibility for the negative event did not ably endurable. From an ethical point of differ between the two groups. It should view, it seems very difficult, if not im- be noted, of course, that much research possible, to create such a situation in a needs to be done before precise predic- laboratory. tions can be made concerning the rela- Physiological response. Some researchers tionship between personal responsibility found that the impact of dissonance is and dissonance reduction. strong enough to affect a noncognitive (physiological) response class (Brehm et REFERENCES al., 1964; Zimbardo et al., 1969). Det- weiler and Zanna (1976) found that an ARONSON, E. 1968 Dissonance theory: Pro- external stimulus results in a real physio- gress and problems. In R. P. Abelson, E. Aronson, W. J. McGuire, T. M. Newcomb, 2In the University of Tokyo , students cnrolled M. J. Rosenberg,& P. H. Tannenbaum (Eds.), in an introductory psychology course are required Theories of cognitive consistency: A sourcebook. at the beginning of the term to fill out a form where Chicago: Rand-McNally, Pp. 5-27. they have to check the "hours" in a week when BREHM, J. W.,& CoHEN, A. R. 1962 Explora- they can participate in psychological experiments. tions in cognitivedissonance. New York: Wiley. In the present experiment, participants were drawn BREHM,M. L., BACK,K. W.,& BOGDONOFF,M. D. from this pool. 1964 A physiological effect of cognitive dis- Responsibility & Dissonance Rcduction 41

sonance under stress and deprivation. Journal attitudes. In T. L. Huston (Ed.), Foundations of Abnormal and Social Psychology,69, 303-310. of interpersonal attraction. New York: Academic CARLSMITH,J. M., ELLSWORTH,P. C.,& ARONSON, Press, Pp. 171-189. E. 1976 Methods of research in social psychology. NEL, E., HELMREICH,R.,& ARONSON,E. 1969 California: Addison-Wesley. Opinion change in the advocate as a function of COOPER, J. 1971 Personal responsibility and the persuasibility of his audience: A clarifica- dissonance: The role of foreseen consequences. tion of the meaning of dissonance. Journal of Journal of Personality and Social Psychology, 18, Personality and Social Psychology, 12, 117-124. 354-363. ROSENBERG,M. J. 1965 When dissonance fails: DETWEILER,R. A.,& ZANNA,M. P. 1976 Physi- On eliminating evaluation apprehension from ological mediation of attitudinal responses. attitude measurement. Journal of Personality and Journal of Personality and Social Psychology, 33, Social Psychology, 1, 28-42. 107-116. SoGIN, S. R.,& PALLAK, M. S. 1976 Bad deci- EAGLY, A. H.,& HIMMELFARB,S. 1978 Atti- sions, responsibility, and attitude change: Ef- tudes and opinions. Annual Review of Psychology, fects of volition, foreseeability, and locus of 29, 517-554. causality of negative consequence. Journal of FESTINGER,L. 1957 A theory of cognitivedissonance. Personality and Social Psychology, 33, 300-306. Evanston, Illinois: Row, Peterson. SOLOMON,R. L. 1949 An extension of control GLASS,D. C.,& MAYHEW, P. 1969 The effects group design. PsychologicalBulletin, 46, 137-150. of cognitive processes on skin conductance re- WICKLUND,R. A.,& BREHM, J. W. 1976 Per- activity to an aversive film. PsychonomicScience, spectives on cognitive dissonance. Hillsdale, New 16, 72-74. Jersey: Lawrence Erlbaum Associates. GREENWALD,A. G.,& RoNIS, D. L. 1978 Twenty WORTMAN, C. B. 1975 Some determinants of years of cognitive dissonance: Case study of the perceived control. Journal of Personality and evolution of a theory. Psychological Review, 85, Social Psychology,31, 282-294. 53-57. ZIMBARDO,P. G., COHEN, A., WEISENBERG,M., LANA, R. E. 1969 Pretest sensitization. In R. DWORKIN, L.,& FIRESTONE, I. 1969 The Rosenthal & R. L. Rosnow (Eds.), Artifact in control of experimental pain. In P. G. Zim- behavioral research. New York: Academic Press, bardo (Ed.), The cognitive control of motivation. Pp. 121-140. Atlanta: Scott, Foresman and Company, Pp. LOTT, A. J.,& LoTT, B. E. 1974 The role of 100-125. reward in the formation of positive interpersonal (Received August 13,1979)