《社会心理学》课程教学资源（文献资料）attitude&stereotype——On the propositional nature of cognitive consistency - Dissonance changes explicit, but not implicit attitudesq

Available online at www.sciencedirect.com Journal of ELSEVIER Joumal of Experimental Social Psychology 4(2004)535-542 www..com/ocate/jesp On the propositional nature of cognitive consistency: Dissonance changes explicit,but not implicit attitudes Bertram Gawronski and Fritz Strackb Abstract Two studies investigated the influence of cognitive dissonance on explicit and implicit attitudes.Employing the induced com ceived situational pressure and control conditions. but not when perceived situational pressure was low Introduction an uncc nfortable feeling of cognitive dissonance when motives is the their original attitude to make it consistent with thei state of otive consisteney.people change their per r.However,the reduction of dissonance by at sonal a the pers 1957 essary only people have no per example.in their seminal work on cognitive dissonance such as a high incentive or other situational forces (for Festinger and Carlsmith(1959)found that participants an overview,see Harmon-Jones Mills,1999). Drwin on the be ey ne al of the however.this effect emerged only when present studies was to investigate the attitudinal conse cinrnts were phud a low incenote os s ng this quences of cognitive dissonance on the explicit and the ation, not when they recei icit lev we argue that bo he caus reduction are inherently propositional.Hence.disso nance-related attitude changes were expected to emerge only hank Gakn Bod licit attitudes however.we would like to provide a more detailed Reb C-ma s(B.)

On the propositional nature of cognitive consistency: Dissonance changes explicit, but not implicit attitudesq Bertram Gawronskia,* and Fritz Strackb a Department of Psychology, Northwestern University, 2029 Sheridan Road, Evanston, IL 60208-2710, USA b Department of Psychology, University of Wu€rzburg, Ro€ntgenring 10, 97070 Wu€rzburg, Germany Received 5 January 2003; revised 1 September 2003 Available online 10 December 2003 Abstract Two studies investigated the influence of cognitive dissonance on explicit and implicit attitudes. Employing the induced com￾pliance paradigm, participants wrote a counterattitudinal essay under conditions of either high or low perceived situational pressure; control participants did not write an essay. Consistent with dissonance theory, results indicated a more favorable explicit attitude toward the initially counterattitudinal position when perceived situational pressure was low, but not when it was high. Implicit attitudes, however, were unaffected by dissonance manipulations. Moreover, explicit attitudes were significantly related to implicit attitudes under high perceived situational pressure and control conditions, but not when perceived situational pressure was low. Results are discussed in terms of associative versus propositional modes of information processing.
2003 Elsevier Inc. All rights reserved. Introduction One of the perhaps most basic human motives is the desire for personal consistency. In order to achieve a state of cognitive consistency, people change their per￾sonal attitudes, their behaviors, or the personal impor￾tance of an attitude object (Festinger, 1957). For example, in their seminal work on cognitive dissonance, Festinger and Carlsmith (1959) found that participants reported a more positive attitude toward a genuinely dull experiment when they had agreed to tell another person that the study was actually interesting. Most importantly, however, this effect emerged only when participants were paid a low incentive for giving this false information, but not when they received a high incentive. According to dissonance theory (Festinger, 1957), this result indicates that people usually experience an uncomfortable feeling of cognitive dissonance when they act in a counterattitudinal manner. Hence, in order to reduce this uncomfortable feeling they often change their original attitude to make it consistent with their behavior. However, the reduction of dissonance by at￾titude change is necessary only if people have no per￾sonal justification for their counterattitudinal behavior such as a high incentive or other situational forces (for an overview, see Harmon-Jones & Mills, 1999). Drawing on the recent distinction between explicit versus implicit attitudes (Greenwald & Banaji, 1995; Wilson, Lindsey, & Schooler, 2000), the main goal of the present studies was to investigate the attitudinal conse￾quences of cognitive dissonance on the explicit and the implicit level. In particular, we argue that both the cause of dissonance experiences and the process of dissonance reduction are inherently propositional. Hence, disso￾nance-related attitude changes were expected to emerge only for explicit attitudinal judgments (i.e., explicit at￾titudes), but not for implicit evaluative associations (i.e., implicit attitudes). Before we test these assumptions, however, we would like to provide a more detailed analysis of what we regard as propositional thinking and what we regard as associative processes. q This research was supported by a grant from the German Science Foundation (DFG) to the first author (Ga 747/1-1). We would like to thank Galen Bodenhausen and Jeff Sherman for helpful comments on an earlier version of this article; and Angela Chiong, Cornelia Eibach, Kimmy Van, and Rebecca Wagner for their help in collecting the data. * Corresponding author. Fax: 1-847-491-7859. E-mail address: gawronski@northwestern.edu (B. Gawronski). 0022-1031/$ - see front matter
2003 Elsevier Inc. All rights reserved. doi:10.1016/j.jesp.2003.10.005 Journal of Experimental Social Psychology 40 (2004) 535–542 www.elsevier.com/locate/jesp Journal of Experimental Social Psychology

536 mtal Social Psychology 40(2004)535-542 Associatice ts.propositional processes when both of themare rearded In order to outline the distinction be 19g200 z& e other (Kruglanski and ropositional we refer to a central dis With regard to disso reduction,we argue that tinction recently proposed by several two-systems inconsistency between two or more propositions is re- models in Sloman 199 mith by explicitly reject lar it is a ed that the are tu p ferent modes of information p One of these to Ffestinger and carlsmith's (1959)induced compliance which best charac paradigm,for example part pr in a netw we and iative p esses are fast but inflexible and require chronie evaluative associations or by finding a justif little cognitive capacity.Most importantly, however. cation for their counterattitudinal behavior.Bo th of ve pi are ent the the e pr wever,are erently prop ona es Eo tive st about African America 199D and the other refec erate attributional process(Kruglanski Klar,1987 1989 Zanna Cooper,1974). dis s the tic rules na to or transform declarative knowledge. thinking is slow but flexible and ofte requires a large ons pre amou mportar positional thinking is ally in terms of truth licit attitudes More xnlicit attitudina ie..a propositional thought is necessarily re judgments were predicted to be more favorable towarda garde aas e o Hence,even though proposi usly advocate ciations.and ven the de is usually dinal behavi ce).than wh the acceptance of these associations (Gilbert.1991).the former can in principle be independent from the latte an activated nod he hro c ev 1at11 additional proposition that resolves the inc Propositional consistency i.e., when perceived situational pressure is high),but not wh redu Applied to ssonanc the th when ceived situational n dissonance reduction require a propositional represen low).Accordingly.explicit attitudinal iudgments should tation of their ele ments With regard to the auses of be argu pres re,but no defined only by an as nment of truth values.That is ymhoueghwoamtaeonistcasociiCa activated same time (M cGrego Experiment I corr propositions are regarded as true.In In order to test these assumptions,participants in other words.inconsistency between two propositions Experiment I were asked to write a counterattitudina ner high or low situ ationa The notion of truth values in p ositi nal thinking ach imp meeradtualhif thout writing an essay.Basec a dicho xhibit able pa more

Associative vs. propositional processes In order to outline the distinction between associative and propositional processes, we refer to a central dis￾tinction recently proposed by several two-systems models in social cognition (Sloman, 1996; Smith & DeCoster, 2000; Strack & Deutsch, in press). In partic￾ular, it is assumed that there are two qualitatively dif￾ferent modes of information processing. One of these modes is associative processing, which is best charac￾terized as a process of spreading activation in a network formed on the basis of spatio-temporal contiguity. As￾sociative processes are fast but inflexible and require little cognitive capacity. Most importantly, however, associative processes are independent of the assignment of truth values. For example, the activation level of negative stereotypes about African Americans may be high even though an individual may regard these ste￾reotypes as being inadequate or false (Devine, 1989). Propositional thinking, in contrast, can be charac￾terized as the application of syllogistic rules to generate or transform declarative knowledge. Propositional thinking is slow but flexible, and often requires a large amount of cognitive capacity. The most important fea￾ture with regard to the present question, however, is that propositional thinking is generally in terms of truth values, i.e., a propositional thought is necessarily re￾garded as true or false.1 Hence, even though proposi￾tional thinking usually refers to cognitively accessible associations, and even though its default mode is usually the acceptance of these associations (Gilbert, 1991), the former can in principle be independent from the latter when the propositional implication of an activated node is considered as inadequate or false. Propositional consistency Applied to cognitive dissonance, we argue that both the cause of dissonance experiences and the process of dissonance reduction require a propositional represen￾tation of their elements. With regard to the causes of dissonance experiences, we argue that a dissonance producing inconsistency between two propositions is defined only by an assignment of truth values. That is, even though two antagonistic associative nodes may be activated at the same time (McGregor, Newby-Clark, & Zanna, 1999), inconsistency emerges only when their corresponding propositions are regarded as true. In other words, inconsistency between two propositions results when both of them are regarded as true, and one follows from the obverse of the other (Kruglanski & Klar, 1987; Shultz & Lepper, 1996). With regard to dissonance reduction, we argue that inconsistency between two or more propositions is re￾solved either by explicitly rejecting one proposition as being false or by finding an additional proposition that resolves the inconsistency (Kruglanski, 1989). Applied to Festinger and Carlsmith
s (1959) induced compliance paradigm, for example, participants may solve the in￾consistency between their attitudes and their behavior either by rejecting the propositional implications of their chronic evaluative associations or by finding a justifi- cation for their counterattitudinal behavior. Both of these processes, however, are inherently propositional, with one reflecting a negation of formerly ascribed truth values (Gilbert, 1991), and the other reflecting a delib￾erate attributional process (Kruglanski & Klar, 1987; Zanna & Cooper, 1974). Based on these considerations, we expected disso￾nance-related phenomena to emerge only for explicit propositional attitude judgments, but not with respect to implicit evaluative associations. Specifically, we pre￾dicted counterattitudinal behavior under conditions of low situational pressure to affect explicit, but not im￾plicit attitudes. More precisely, explicit attitudinal judgments were predicted to be more favorable toward a previously advocated counterattitudinal position, when there is no external justification for the counterattitu￾dinal behavior (e.g., induced compliance), than when there is a plausible external justification (e.g., situational forces). Moreover, participants were expected to base their evaluative judgments on their chronic evaluative associations when dissonance can be reduced by an additional proposition that resolves the inconsistency (i.e., when perceived situational pressure is high), but not when dissonance is reduced by a rejection of the propositional implications of their chronic evaluative associations (i.e., when perceived situational pressure is low). Accordingly, explicit attitudinal judgments should be positively related to implicit evaluative associations under conditions of high situational pressure, but not under conditions of low situational pressure. Experiment 1 In order to test these assumptions, participants in Experiment 1 were asked to write a counterattitudinal essay under conditions of either high or low situational pressure. Afterwards, all participants completed both an explicit and an implicit attitude measure on the topic in question. Participants in a control group completed the two attitude measures without writing an essay. Based on the assumptions outlined in the Introduction, we expected participants to exhibit a more favorable ex- 1 The notion of truth values in propositional thinking is not committed to a dichotomous conceptualization, but can also be interpreted in terms of a probabilistic approach implying gradual shifts in the subjective likelihood of truth. For the sake of simplicity, however, we adopted a dichotomous conceptualization to outline our theoretical assumptions. 536 B. Gawronski, F. Strack / Journal of Experimental Social Psychology 40 (2004) 535–542

537 plicit attitude toward the advocated counterattitudinal explicit and implicit attitudes toward alcoholic bever position under low situational pressure than under high na pressure or control ns.Implicit dissonance manipulations.Moreover,explicit and im Explicit attitudes plicit attitudes were predicted to be positively correlated Explicit attitudes toward a general prohibition of al- oh olic beverages were as ms.Spec ons low situati pr they ral n of alo on a rating scale rar ging from 1 (absolutely not)to 5 Method (very much),and (b)whether they would students emale 2 male) petition in favor of a ranging (definitely not)to 5(definitely ves). over Participants were pa s of a3 ald, (essay assignment:induced compliance vs.forced posi and Schwartz's(1998)Implicit Association Test(IAT) e measurement: ng Gree consisted of fv 11 Procedure beer)and names of 10 non-alcoholic beverag ；ee On arrival,participants were welcomed and informed alco red by th on-alcono Participant A name of an alcoholic bev eft-h cared on the screen and a right-hand key ("5"of the number pad)in the beverages. of a non- the attribute dis in and 10 ne of alcoholic beverages should be strongly restricted in hate)were presented and had to be classified according European countries aid to intereste ng (right-han e initi task(Bloc n alterating order.Participants had to press the left was the collection of arguments generated by citizens of hand key when either Jnio In noun was pre ected large number of argument against a sented.In the target-concent discrimination prohibition of alcoholic beverages and that we would be task (Block 4).the initial target-concept discrimination pleased if they we arguments in tho was repeated wit a switch of categorization key they the individual ta salience of their “"free choice,.”participants were asked left-hand key when either a non-alcoholic beverage or a eir perfor sented and right-hand kev was eely c 1994 tion for the following task and a request to respond as arguments either in favor or against a general prohibi uickly as possible even if this would lead to errors.The ion and that they had been generate argu hree (Blocks 1.2.and 4)each nts.P 5)each prised 120 trials The sam pants under control conditions did not have to generate andomized order of trials was used for all participants. the argumen The response generation they were administered measures of Wrong responses

plicit attitude toward the advocated counterattitudinal position under low situational pressure than under high situational pressure or control conditions. Implicit atti￾tudes, in contrast, were expected to be unaffected by dissonance manipulations. Moreover, explicit and im￾plicit attitudes were predicted to be positively correlated under high situational pressure and under control con￾ditions, but not under conditions of low situational pressure. Method Participants and design Sixty-four German students (43 female, 21 male) drawn from a volunteer pool were recruited for a study on a controversial political issue. Participants were paid 6 Euro (approximately US-$6 at the time). Participants were randomly assigned to the six conditions of a 3 (essay assignment: induced compliance vs. forced posi￾tion vs. no essay)
2 (order of attitude measurement: implicit first vs. explicit first) factorial design. Procedure On arrival, participants were welcomed and informed that they were taking part in a study sponsored by the European Union about alcohol prohibition in European countries. Pretests revealed that most German students have a strong negative attitude toward a general pro￾hibition of alcoholic beverages. The experimenter ex￾plained that there is a controversial discussion in the European Union whether or not the legal consumption of alcoholic beverages should be strongly restricted in European countries. For this purpose, the European Union was said to be interested in finding arguments favoring and opposing a general prohibition of alcoholic beverages. Ostensibly, the main goal of the present study was the collection of arguments generated by citizens of the European Union. In the induced compliance con￾dition, participants were told that we already had col￾lected a large number of arguments against a general prohibition of alcoholic beverages and that we would be pleased if they were willing to generate arguments in favor of a general prohibition, even though they were free to choose either position. In order to increase the salience of their ‘‘free choice,’’ participants were asked to sign a release form, indicating that their performance was freely chosen (Elliot & Devine, 1994). In the forced position condition, participants were told that each participant would be randomly assigned to generate arguments either in favor or against a general prohibi￾tion and that they had been assigned to generate argu￾ments in favor of a general prohibition. Participants had a maximum of 15 min to generate arguments. Partici￾pants under control conditions did not have to generate arguments. After participants had finished the argument generation task, they were administered measures of explicit and implicit attitudes toward alcoholic bever￾ages. Finally, participants were debriefed and thanked for their participation. Explicit attitudes Explicit attitudes toward a general prohibition of al￾coholic beverages were assessed with two items. Specif￾ically, participants were asked to indicate: (a) how much they favor a general prohibition of alcoholic beverages on a rating scale ranging from 1 (absolutely not) to 5 (very much), and (b) whether they would be willing to sign a petition in favor of a general prohibition of al￾coholic beverages on a rating scale ranging from 1 (definitely not) to 5 (definitely yes). Implicit attitudes Implicit attitudes toward alcoholic beverages were assessed with an adaptation of Greenwald, McGhee, and Schwartz
s (1998) Implicit Association Test (IAT). Following Greenwald et al., the IAT consisted of five blocks. In the initial target-concept discrimination task (Block 1), names of 10 alcoholic beverages (e.g., gin, beer) and names of 10 non-alcoholic beverages (e.g., juice, soda) had to be assigned to the categories ‘‘alco￾holic’’ or ‘‘non-alcoholic,’’ respectively. Participants were asked to press a left-hand key (‘‘A’’) when the name of an alcoholic beverage appeared on the screen and a right-hand key (‘‘5’’ of the number pad) in the case of a non-alcoholic beverage. In the attribute dis￾crimination task (Block 2), 10 positive nouns (e.g., happiness, paradise) and 10 negative nouns (e.g., bomb, hate) were presented and had to be classified according to the categories positive (left-hand key) and negative (right-hand key). In the initial combined task (Block 3), target and attribute discrimination trials were presented in alternating order. Participants had to press the left￾hand key when either an alcoholic beverage or a positive noun was presented and the right-hand key when either a non-alcoholic beverage or a negative noun was pre￾sented. In the reversed target-concept discrimination task (Block 4), the initial target-concept discrimination was repeated with a switch of the categorization keys. The reversed combined task (Block 5) again combined the two individual tasks. Participants had to press the left-hand key when either a non-alcoholic beverage or a positive noun was presented and the right-hand key when either an alcoholic beverage or a negative noun was presented. Each block started with a brief instruc￾tion for the following task and a request to respond as quickly as possible even if this would lead to errors. The three discrimination tasks (Blocks 1, 2, and 4) each consisted of a total of 20 trials. The two combined tasks (Blocks 3 and 5) each comprised 120 trials. The same randomized order of trials was used for all participants. The response-stimulus interval following correct re￾sponses was 250 ms. Wrong responses were indicated B. Gawronski, F. Strack / Journal of Experimental Social Psychology 40 (2004) 535–542 537

B.Govronski,F.Sirack Journal of Experin al Social Psychology 40(2004)535-542 (German equivalent for appearing for 1000ms below the center o and SD of explicit and implicit attitudes toward alcoholi 、2 beverages asa function of essay assignment conditions,Experiment1 Induced Forced Control Results Explicit attitudes SD 6 Explicit attitude ratings were merged into a single Implicit titude index of explicit negative attitudes toward alcoholic bever by ating Note.Higher values indicate more negative attitudes. attitude ording to the procedures described by Greenwald et al.(1998).Response latencies lowe squalified by a sig- 300ms were trial 、1.aa1 two-way mter05o2o_indicatin。CyPe, .13.indicating that im subtracting the mean response time of the reversed ve when 10a ed aft ather than be. fore explicit attitudes Most importantly.there was a latency of the n high significant two-wa e type and es 2. 02 IAT nce F581-44 from-382 to +858 ms (M SD .01,explicit attitudes were most negativ under I(se nduce nce cond ,m0 der lorce combined blocks were each divided into three c conditions F(2.58)=7.74. tive parts of equal length (i.e.,40 trials).The three thirds 21.Plannc contrasts for explicit attitudes revealed significant dif were then use three IA ferences between all of the three conditions (all ps<.05). tency o mplicit Overall.explicit and implicit attitudes were moder- n order to t the nance manipula ately correlated (r =.26,p =.04).Consistent with ou yp we atio driv by and then submitted to a 2(attitude type)x 3(essa signment)2 (or .05).There was no significant correlation between TA r XD and implicit attituc under indu P= between-subiects factors.This analysis revealed a sig from both forced position,=1.86,p=.03,and con- trol co =1.70,p=.04.There was no signifi an orced position and contro 5 of the Discussion i,2002 Results from Ex that IAT in for ou titud. changes are inherently propositional.Consistent with ants have an ed in an ab ofalcohol sign a

with the word ‘‘FEHLER!’’ (German equivalent for ‘‘ERROR!’’) appearing for 1000 ms below the center of the screen.2 Results Preliminary analyses Explicit attitude ratings were merged into a single index of explicit negative attitudes toward alcoholic beverages by calculating mean values (Cronbach
s a ¼ :62, M ¼ 2:59, SD ¼ :91). IAT-scores for implicit attitudes toward alcoholic beverages were calculated according to the procedures described by Greenwald et al. (1998). Response latencies lower than 300 ms were recoded to 300 ms, and latencies higher than 3000 ms were recoded to 3000 ms. Error trials were excluded from analyses. Individual IAT-scores were calculated by subtracting the mean response time of the reversed combined task (Block 5) from the mean latency of the initial combined task (Block 3), with higher values in￾dicating an implicit preference for non-alcoholic as compared to alcoholic beverages. IAT-scores ranged from –382 to +858 ms (M ¼ 244, SD ¼ 234).3 In order to estimate the reliability of the IAT (see Gawronski, Eh￾renberg, Banse, Zukova, & Klauer, 2003a), the two combined blocks were each divided into three consecu￾tive parts of equal length (i.e., 40 trials). The three thirds were then used to calculate three IAT-scores for each participant, which revealed an internal consistency of .90 (Cronbach
s a). Cognitive dissonance In order to test the effects of dissonance manipula￾tions on explicit and implicit attitudes, the respective indices were first standardized by a z-transformation and then submitted to a 2 (attitude type)
3 (essay as￾signment)
2 (order of attitude measurement) mixed￾model analysis of variance (ANOVA) with the first variable as a within-subjects factor and the other two as between-subjects factors. This analysis revealed a sig￾nificant main effect of order, F ð1; 58Þ ¼ 4:97, p ¼ :03, g2 ¼ :08, indicating that attitudes toward alcoholic beverages were more negative when explicit attitudes were assessed first than when implicit attitudes were assessed first. This main effect was qualified by a sig￾nificant two-way interaction of order and attitude type, F ð1; 58Þ ¼ 8:41, p ¼ :005, g2 ¼ :13, indicating that im￾plicit attitudes toward alcoholic beverages were more negative when they were assessed after rather than be￾fore explicit attitudes. Most importantly, there was a significant two-way interaction of attitude type and es￾say assignment, F ð2; 58Þ ¼ 7:23, p ¼ :002, g2 ¼ :20 (see Table 1). Whereas implicit attitudes were unaffected by dissonance manipulations, F ð2; 58Þ ¼ :44, p ¼ :64, g2 ¼ :01, explicit attitudes were most negative under induced compliance conditions, moderate under forced position conditions, and least negative under control conditions, F ð2; 58Þ ¼ 7:74, p ¼ :001, g2 ¼ :21. Planned contrasts for explicit attitudes revealed significant dif￾ferences between all of the three conditions (all ps<.05).4 Implicit–explicit correlations Overall, explicit and implicit attitudes were moder￾ately correlated (r ¼ :26, p ¼ :04). Consistent with our hypotheses, however, this correlation was driven by particularly pronounced correlations under forced po￾sition (r ¼ :46, p ¼ :02) and control conditions (r ¼ :44, p ¼ :05). There was no significant correlation between explicit and implicit attitudes under induced compliance conditions (r ¼ :11, p ¼ :64). The correlation obtained for induced compliance conditions differed significantly from both forced position, z ¼ 1:86, p ¼ :03, and con￾trol conditions, z ¼ 1:70, p ¼ :04. There was no signifi- cant difference between forced position and control conditions, z ¼ :08, p ¼ :53. Discussion Results from Experiment 1 offer preliminary evidence for our assumption that dissonance-related attitude changes are inherently propositional. Consistent with 2 Because one of the main goals of the present studies was to test predictions about the relation between explicit and implicit attitudes, we deliberately decided not to counterbalance the order of the two combined blocks (for a discussion, see Gawronski, 2002). 3 One could argue that IAT-scores in the present study were generally positive, and hence indicate a negative rather than a positive implicit attitude toward alcoholic beverages. However, since the IAT assesses only relative rather than absolute attitudes, a positive IAT￾score indicates only that participants have an implicit preference for non-alcoholic as compared to alcoholic beverages. Moreover, recent research indicates that IAT-scores cannot be interpreted in an absolute manner, because the overall size of IAT-scores is affected not only by category associations, but also by contingent stimulus factors (e.g., Mitchell, Nosek, & Banaji, 2003; Steffens & Plewe, 2001). Table 1 Means and SD of explicit and implicit attitudes toward alcoholic beverages as a function of essay assignment conditions, Experiment 1 Induced compliance Forced position Control condition Explicit attitudes M 3.10 2.58 2.07 SD .90 .76 .83 Implicit attitudes M 211 272 242 SD 139 246 296 Note. Higher values indicate more negative attitudes. 4 Separate analyses for the two explicit items revealed correspond￾ing effects for attitudes toward a general prohibition of alcoholic beverages, F ð2; 58Þ ¼ 3:37, p ¼ :04, g2 ¼ :10, as well as for willingness to sign a petition in favor of a general prohibition of alcoholic beverages, F ð2; 58Þ ¼ 7:65, p ¼ :001, g2 ¼ :21. 538 B. Gawronski, F. Strack / Journal of Experimental Social Psychology 40 (2004) 535–542

B.,F.StrackI Journal of Expermental Soclal Psychology(04)535-54 539 this assumption.cognitive dissonance affected only ex- Leippe and Eisenstadt (1994).Applying the notion of dissonance-related attitude changes to the domain of pants showed a mor prejudic hese researchers aske pa der low situational pressure than under high situational African American students at the expense of funding for pressure or control】 conditions.Implicit attitudes,in Caucasian students.Consistent with previous findings in sitive whe tions of high situational pressure and under control uational pressures to write the essay were low,than conditions. .but not when situational pressure was lov when they were high. ving on the ated ch assump tive associations when cognitive dissonance could be emerge only for explicit,but not for Most impor .this should be the case even when the unte at prejudic In the followi seem to have rejected the propositional implications of tested by using an IAT designed to assess implici thei chronic cognitive dissonance efor Caucasians over African Americans.and proposi pre ence assumptions about the propositional nature of cognitive sively with regard to their p cesses but directly corre dissonance one may object that the measures pond to one another with regard to the particular ment not respect to thus ruling out e explanations also with respect to their content.Whereas the T and Eisenstadt's (994)dissonance maninulation essed relative attitudes toward alcoholic versus nor were e ected to be pos In othe word the wo attitudes toward related but not necessarily corre- pressure. Acco rdingly,participant is in h the changing their attitudes toward alcoholic ve Particinants and design verages. This pattern,however,could be estern undergraduate f the propos )participated tudy on politica tive attitudes randomly assigned to the four conditions of a (ssa changed regardless of whether they are asse assignment:induced compliance vs.forced position)x2 explicitlyor implicitly Hence.in order to rule out this dr measuremen:implicit first vs.explic iment 2 aim irst)factorial measures that differ only with respect to the asse thus had to be excluded from analyses. ocesses,but not with respect to the particular attitude object edur ned by a white Experiment 2 a survey conducted by the university administration. The experimenter th adm otal ar ences in the attitude obiect rather than to differences in scholarships,the percentage committed to African

this assumption, cognitive dissonance affected only ex￾plicit, but not implicit attitudes. Specifically, partici￾pants showed a more favorable explicit attitude toward a previously advocated counterattitudinal position un￾der low situational pressure than under high situational pressure or control conditions. Implicit attitudes, in contrast, were unaffected by cognitive dissonance ma￾nipulations. Moreover, explicit and implicit attitudes showed a significant positive correlation under condi￾tions of high situational pressure and under control conditions, but not when situational pressure was low. Taken together, these results indicate that participants based their explicit judgments on their chronic evalua￾tive associations when cognitive dissonance could be reduced by finding an additional proposition that solved the inconsistency between their counterattitudinal be￾havior and the propositional implications of their chronic evaluative associations. However, participants seem to have rejected the propositional implications of their chronic associations when cognitive dissonance could not be reduced by an additional proposition. Even though these findings are consistent with our assumptions about the propositional nature of cognitive dissonance, one may object that the measures used in Experiment 1 differed not only with respect to the as￾sessed processes (i.e., associations vs. judgments), but also with respect to their content. Whereas the IAT as￾sessed relative attitudes toward alcoholic versus non￾alcoholic beverages, the employed explicit measure assessed attitudes toward a prohibition of alcoholic beverages. In other words, the two measures assessed attitudes toward related but not necessarily corre￾sponding attitude objects. Accordingly, participants in the present study could have changed their attitudes toward the particular policy in question without changing their attitudes toward alcoholic versus non￾alcoholic beverages. This pattern, however, could be independent of the proposed interpretation in terms of associative and propositional processes, such that rela￾tive attitudes toward alcoholic beverages remain un￾changed regardless of whether they are assessed explicitly or implicitly. Hence, in order to rule out this alternative interpretation, Experiment 2 aimed to repli￾cate the basic findings of Experiment 1 employing measures that differ only with respect to the assessed processes, but not with respect to the particular attitude object. Experiment 2 In order to rule out the alternative interpretation that the results obtained in Experiment 1 are due to differ￾ences in the attitude object rather than to differences in the cognitive processes, Experiment 2 employed a vari￾ant of the induced compliance paradigm developed by Leippe and Eisenstadt (1994). Applying the notion of dissonance-related attitude changes to the domain of prejudice, these researchers asked participants to write an essay in favor of a general increase of scholarships for African American students at the expense of funding for Caucasian students. Consistent with previous findings in the induced compliance paradigm, Leippe and Eisens￾tadt found that White participants exhibited a lower level of prejudice against African Americans when sit￾uational pressures to write the essay were low, than when they were high. Drawing on the present assumptions, we argue that such dissonance-related changes in prejudice should emerge only for explicit, but not for implicit prejudice. Most importantly, this should be the case even when explicit and implicit prejudice measures directly corre￾spond to one another with respect to their attitude ob￾ject. In the following study, these assumptions were tested by using an IAT designed to assess implicit preference for Caucasians over African Americans, and rating scales assessing explicit preference for Caucasians over African Americans. These measures differ exclu￾sively with regard to their processes, but directly corre￾spond to one another with regard to the particular attitude object, thus ruling out alternative explanations in terms of differing contents. Specifically, we expected Leippe and Eisenstadt
s (1994) dissonance manipulation to affect explicit, but not implicit prejudice. Moreover, explicit and implicit prejudice were expected to be pos￾itively correlated under conditions of high situational pressure, but not under conditions of low situational pressure. Method Participants and design Thirty-nine non-Black Northwestern undergraduates (24 female, 15 male) participated in a study on political issues in return for course credit. Participants were randomly assigned to the four conditions of a 2 (essay assignment: induced compliance vs. forced position)
2 (order of attitude measurement: implicit first vs. explicit first) factorial design. Because of a computer error, data from one participant were only partially recorded, and thus had to be excluded from analyses. Procedure On arrival, participants were welcomed by a White experimenter and informed that they were taking part in a survey conducted by the university administration. The experimenter explained that the administration considers a new scholarship policy implying that of the total amount of money the university has available for scholarships, the percentage committed to African American students should at least be doubled (Leippe & Eisenstadt, 1994). The administration was said to B. Gawronski, F. Strack / Journal of Experimental Social Psychology 40 (2004) 535–542 539

540 al Social Psychology 40(2004)535-542 n ther e i against African participants were told that we already had Induced Forced rat Explicit prejudice SD 1.01 158 163 forced position condition,participants were told that Note.Higher values indicate higher levels of prejudice ate or or ag A F nd tha they had been assigned to generate arguments in favor within-subiects factor and the second as between-sub jects factor.This analysis revealed a significant two-way of an increase.Afterwards,participants wer adminis attit measures of explic type and assignmen and thanked for their part affected by dissonance manipulations,F(1,36)=.01 Explici .92. ess negat tha ask ants we F1,36)=8.02,p=008.2-18Po10 can Am ericans,Cauca sian Americans)on scales ranging from I (very negative)to 7(very positive) Implicit- ion Implicit attitudes Implicit attitudes were assessed with an IAT designed ever,replicating the pattern obtained in Experiment 1 this correlation was driven by a relatively pronounce conditi 35 each consisted of a total of 20 trials.The two combined tions.in contrast.explicit and implicit attitudes showed tasks each comprised 80 trials.Order of trials was ran mo 25,p= Even tho dons did not he cor participan The respsoms.W responses were indicated with the word "ERROR =1.75.n=04. appearing for 1000ms below the center of the screen. Results Discussion Preliminary analyses Results from Experiment 2 corroborate the assump ce changes explic culated by subtracting A ans from rati ng denobLuno for Caucas ans over african Americans explicit prejudice against African Americans after writ. 95).IAT-scores for implicit attitudes an essay in favor of a general increase of scholar mgto the p ships for African Americ stude ts und condition fron-365to+476msM=160SD_15 ores range trast wa unafected by dissonance manipulations.Furthermore men titudes vere firs st standardized by a z-transformation and o.J .Dunton.Wlam 195:Gawrons

employ the psychology department to gather arguments on both sides of this issue, which it will then use in making an educated decision. In the induced compliance condition, participants were told that we already had collected enough arguments against a policy change and that we would be pleased if they were willing to generate arguments in favor of a policy change, even though they were of course free to choose either position. In the forced position condition, participants were told that each participant would be randomly assigned to gener￾ate arguments either in favor or against an increase of scholarships for African American students and that they had been assigned to generate arguments in favor of an increase. Afterwards, participants were adminis￾tered measures of explicit and implicit prejudice against African Americans. Finally, participants were debriefed and thanked for their participation. Explicit attitudes Explicit attitudes were assessed with rating scales, asking participants to indicate their personal feelings toward a number of different ethnic groups (e.g., Afri￾can Americans, Caucasian Americans) on scales ranging from 1 (very negative) to 7 (very positive). Implicit attitudes Implicit attitudes were assessed with an IAT designed to assess implicit preference for Caucasians over African Americans. The IAT was conceptually identical to Greenwald et al. (1998). The three discrimination tasks each consisted of a total of 20 trials. The two combined tasks each comprised 80 trials. Order of trials was ran￾domized for each participant. The response-stimulus interval following correct responses was 250 ms. Wrong responses were indicated with the word ‘‘ERROR!’’ appearing for 1000 ms below the center of the screen. Results Preliminary analyses Explicit attitude scores were calculated by subtracting ratings for African Americans from ratings for Cauca￾sian Americans, with higher scores indicating a stronger preference for Caucasians over African Americans (M ¼ :42, SD ¼ :95). IAT-scores for implicit attitudes were calculated according to the procedures described in Experiment 1 (Cronbach
s a ¼ :80). IAT-scores ranged from )365 to +476 ms (M ¼ 160, SD ¼ 156). Cognitive dissonance Order of assessment did not affect any of the present dependent measures, and was thus excluded from the following analyses. Indices for explicit and implicit at￾titudes were first standardized by a z-transformation and then submitted to a 2 (attitude type)
2 (essay assign￾ment) mixed-model ANOVA with the first variable as a within-subjects factor and the second as between-sub￾jects factor. This analysis revealed a significant two-way interaction of attitude type and essay assignment, F ð1; 36Þ ¼ 4:10, p ¼ :05, g2 ¼ :10 (see Table 2). Whereas implicit attitudes toward African Americans were un￾affected by dissonance manipulations, F ð1; 36Þ ¼ :01, p ¼ :92, g2 ¼ :00, explicit attitudes toward African Americans were less negative under induced compliance conditions than under forced position conditions, F ð1; 36Þ ¼ 8:02, p ¼ :008, g2 ¼ :18. Implicit–explicit correlations Overall, explicit and implicit attitudes revealed a non￾significant positive correlation (r ¼ :18, p ¼ :29). How￾ever, replicating the pattern obtained in Experiment 1, this correlation was driven by a relatively pronounced positive correlation under forced position conditions (r ¼ :35, p ¼ :13).5 Under induced compliance condi￾tions, in contrast, explicit and implicit attitudes showed a moderate negative correlation (r ¼ :25, p ¼ :31). Even though these correlations did not reach the con￾ventional level of statistical significance, the difference between the two correlations was statistically significant, z ¼ 1:75, p ¼ :04. Discussion Results from Experiment 2 corroborate the assump￾tion that cognitive dissonance changes explicit, but not implicit attitudes. Replicating the pattern obtained in Experiment 1, participants exhibited a lower level of explicit prejudice against African Americans after writ￾ing an essay in favor of a general increase of scholar￾ships for African American students under conditions of low situational pressure than under conditions of high situational pressure. Implicit prejudice, in contrast, was unaffected by dissonance manipulations. Furthermore, Table 2 Means and SD of explicit and implicit prejudice against African Americans as a function of essay assignment conditions, Experiment 2 Induced compliance Forced position Explicit prejudice M .00 .80 SD .69 1.01 Implicit prejudice M 158 163 SD 114 190 Note. Higher values indicate higher levels of prejudice. 5 The somewhat lower correlation under forced position conditions obtained in Experiment 2 might be due to the fact that only individuals low in motivation to control prejudiced reactions exhibit a high correlation between explicit and implicit prejudice, but not people high in motivation to control (e.g., Banse & Gawronski, 2003; Dunton & Fazio, 1997; Fazio, Jackson, Dunton, & Williams, 1995; Gawronski, Geschke, & Banse, 2003b). 540 B. Gawronski, F. Strack / Journal of Experimental Social Psychology 40 (2004) 535–542

541 explicit and implicit prejudice were positively correlated when they did not write a counterattitudinal essav.This. under conditions of high situational pressure,but not however,was not the case.Rather,implicit attitudes under con ons were approxin ely equal regardl s of w Second.if participants based their explicit judgments on sses,but not with respect to their content.Hence activated counterattitudinal associations when there was native interpreta h i0liustincationfo atti in the tude obje ences in the particular processes assessed by the em- attitudes should have exhibited a positive correlation ployed measures. d compliance,but not inder forced pos Th General discussion correlated under forced position and control conditions but not under induced compliance conditions. The main goa of the present studie s was to invest the present concep ma on the ther than disso 1957)Bas 2000S Deutsch,in press their attitudes rom obseh ving their own beha and the of di nces in a he r dissonance-related attitude changes were expected to testing the resulting consistency is exactly wha n the ults e pr presen generally support these assumptions such an interpretation in ter ception may Notwithstanding these findings,however,one may the obta udes of ance for tw to thei pulation uninteresting case of low measurement validity.This titudes (e.g..Bem MeConnell.1970).In the presen ruled out by the pre studies.however pants imp eva ated to and control conditions.but not unde conditions of in- udgments under high situational pressure as well as ess implicit under control conditions,suggesting that participants ac ent validity, ctually ha ve c ous acc to the premanipula and implicit attitudes under t onditions that the s of assigning truth values Another possible objection is that writing a counter. causes an uncomfortable feeling of cognitive dissonance could ctivated counterattitudina as true,and on ind general fored position Moreover narticinants ma change even under conditions of low situational pre base heir explicit judgments on these activate if thes ns when there experiences are attribut ed to causes other was n ons for &C 1930 ior.put they may have relied on their old ttitude when still base their explicit autitudinal judgment on their a situational iustification was available.even though chronic evaluative associations even when self-percep sare theo cally possible,they are no ed and ng participants should have exhibited a more favorable self-perception,the feeling of cognitive dissonance still mplicit attitude toward the counterattitu dinal topi plays a when they actually wrote a counterattitudinal essay thar

explicit and implicit prejudice were positively correlated under conditions of high situational pressure, but not under conditions of low situational pressure. Most im￾portantly, these results were obtained even though the employed measures differed only with respect to their processes, but not with respect to their content. Hence, the present findings rule out the alternative interpreta￾tion that the results obtained in Experiment 1 are due to differences in the attitude object, rather than to differ￾ences in the particular processes assessed by the em￾ployed measures. General discussion The main goal of the present studies was to investi￾gate the impact of cognitive dissonance on explicit and implicit attitudes. Drawing on the distinction between associative and propositional processes (Sloman, 1996; Smith & DeCoster, 2000; Strack & Deutsch, in press), we argued that both the cause of dissonance experiences and the process of dissonance reduction require a propositional representation of their elements. Hence, dissonance-related attitude changes were expected to emerge only for explicit attitudinal judgments, but not for implicit evaluative associations. The present results generally support these assumptions. Notwithstanding these findings, however, one may object that the obtained null effect on implicit attitudes is difficult to interpret, since it could reflect either the pre￾dicted stability of implicit attitudes or the theoretically uninteresting case of low measurement validity. This criticism, however, can easily be ruled out by the pre￾dicted finding that implicit attitudes were significantly related to explicit attitudes under both forced position and control conditions, but not under conditions of in￾duced compliance. If the IATs used to assess implicit attitudes actually had low measurement validity, there should have been no significant relation between explicit and implicit attitudes under these conditions. Another possible objection is that writing a counter￾attitudinal essay could have activated counterattitudinal associations, thus resulting in a general change of im￾plicit attitudes under both induced compliance and forced position conditions. Moreover, participants may have based their explicit judgments on these activated counterattitudinal associations when there was no situ￾ational justification for their counterattitudinal behav￾ior, but they may have relied on their old attitude when a situational justification was available. Even though these assumptions are theoretically possible, they are not consistent with the present data. First, if merely writing a counterattitudinal essay affects implicit attitudes, participants should have exhibited a more favorable implicit attitude toward the counterattitudinal topic when they actually wrote a counterattitudinal essay than when they did not write a counterattitudinal essay. This, however, was not the case. Rather, implicit attitudes were approximately equal regardless of whether partic￾ipants did or did not write a counterattitudinal essay. Second, if participants based their explicit judgments on activated counterattitudinal associations when there was no situational justification for their counterattitudinal behavior, but relied on their old attitude when a situa￾tional justification was available, explicit and implicit attitudes should have exhibited a positive correlation under induced compliance, but not under forced posi￾tion conditions. This, however, was also not the case. In contrast, explicit and implicit attitudes were positively correlated under forced position and control conditions, but not under induced compliance conditions. In some respects, the present conceptualization may sound more like self-perception theory (Bem, 1967) ra￾ther than dissonance theory (Festinger, 1957). Basically, self-perception theory states that individuals deliberately infer their attitudes from observing their own behavior and the circumstances in which it occurs. Hence, applied to the present conceptualization one could argue that the propositional processes of assigning truth values and testing the resulting consistency is exactly what self￾perception theory considers as the process of deliber￾ately inferring one
s attitude. However, even though such an interpretation in terms of self-perception may sound quite plausible, we nevertheless favor a concep￾tualization in terms of cognitive dissonance for two reasons. First, self-perception theory usually argues that individuals have no access to their premanipulation at￾titudes (e.g., Bem & McConnell, 1970). In the present studies, however, premanipulation attitudes were clearly reflected in participants
implicit evaluations. Moreover, implicit evaluations were significantly related to explicit judgments under high situational pressure as well as under control conditions, suggesting that participants actually have conscious access to their premanipulation attitudes. Second, the present conceptualization implies that the propositional process of assigning truth values causes an uncomfortable feeling of cognitive dissonance when two propositions are regarded as true, and one follows from the obverse of the other. However, such dissonance experiences may not result in explicit attitude change—even under conditions of low situational pres￾sure—if these experiences are attributed to causes other than cognitive inconsistency (Fazio, Zanna, & Cooper, 1977; Zanna & Cooper, 1974). In this case, people may still base their explicit attitudinal judgment on their chronic evaluative associations even when self-percep￾tion should lead to attitude change. Hence, even though the process of ascribing of truth values and testing the resulting consistency may be considered as a process of self-perception, the feeling of cognitive dissonance still plays a crucial role for attitude changes caused by counterattitudinal behavior. B. Gawronski, F. Strack / Journal of Experimental Social Psychology 40 (2004) 535–542 541

References Banse.R..&Gawr G vald.A.G..McGhee.D.E..Schwartz.J.K.L.(1998 act witho Bem.D.() Ha ical As W.(1989).Lay epis Synthe cip onof dis rea McGregor.L Newby-Cark.1.R ().Rer d Social P as psych 07.382-194 ort.In E. Fa y in au. line?Joumal of Personaliry and Fa (07) onski.B.(002) What do ity of ans ,R,Z1 .K.C egory-basednd Strack.F&De 0036) Zann ()Dissonance and the pill:An

References Banse, R., & Gawronski, B. (2003). Die Skala Motivation zu vorurteilsfreiem Verhalten: Psychometrische Eigenschaften und Validit€at [The scale motivation to act without prejudice: Psycho￾metric properties and validity]. Diagnostica, 49, 4–13. Bem, D. J. (1967). Self-perception: An alternative interpretation of cognitive dissonance phenomena. Psychological Review, 74, 183– 200. Bem, D. J., & McConnell, H. K. (1970). Testing the self-perception explanation of dissonance phenomena: On the salience of prem￾anipulation attitudes. Journal of Personality and Social Psychology, 14, 23–31. Devine, P. G. (1989). Stereotypes and prejudice: Their automatic and controlled components. Journal of Personality and Social Psychol￾ogy, 56, 5–18. Dunton, B. C., & Fazio, R. H. (1997). An individual difference measure of motivation to control prejudiced reactions. Personality and Social Psychology Bulletin, 23, 316–326. Elliot, A. J., & Devine, P. G. (1994). On the motivational nature of cognitive dissonance: Dissonance as psychological discomfort. Journal of Personality and Social Psychology, 67, 382–394. Fazio, R. H., Jackson, J. R., Dunton, B. C., & Williams, C. J. (1995). Variability in automatic activation as an unobtrusive measure of racial attitudes: A bona fide pipeline? Journal of Personality and Social Psychology, 69, 1013–1027. Fazio, R. H., Zanna, M. P., & Cooper, J. (1977). Dissonance and self￾perception: An integrative view of each theory
s proper domain of application. Journal of Experimental Social Psychology, 13, 464–479. Festinger, L. (1957). A theory of cognitive dissonance. Evanston, IL: Row Peterson. Festinger, L., & Carlsmith, J. M. (1959). Cognitive consequences of forced compliance. Journal of Abnormal and Social Psychology, 58, 203–210. Gawronski, B. (2002). What does the Implicit Association Test measure? A test of the convergent and discriminant validity of prejudice related IATs. Experimental Psychology, 49, 171–180. Gawronski, B., Ehrenberg, K., Banse, R., Zukova, J., & Klauer, K. C. (2003a). It
s in the mind of the beholder: The impact of stereotypic associations on category-based and individuating impression for￾mation. Journal of Experimental Social Psychology, 39, 16–30. Gawronski, B., Geschke, D., & Banse, R. (2003b). Implicit bias in impression formation: Associations influence the construal of individuating information. European Journal of Social Psychology, 33, 573–589. Gilbert, D. T. (1991). How mental systems believe. American Psychologist, 46, 107–119. Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition: Attitudes, self-esteem, and stereotypes. Psychological Review, 102, 4–27. Greenwald, A. G., McGhee, D. E., & Schwartz, J. K. L. (1998). Measuring individual differences in implicit cognition: The Implicit Association Test. Journal of Personality and Social Psychology, 74, 1464–1480. Harmon-Jones, E., & Mills, J. (Eds.). (1999). Cognitive dissonance: Progress on a pivotal theory in social psychology. Washington, DC: American Psychological Association. Kruglanski, A. W. (1989). Lay epistemics and human knowledge: Cognitive and motivational bases. New York: Plenum Press. Kruglanski, A. W., & Klar, Y. (1987). A view from a bridge: Synthesizing consistency and attribution paradigms from a lay epistemic perspective. European Journal of Social Psychology, 17, 211–241. Leippe, M. R., & Eisenstadt, D. (1994). Generalization of dissonance reduction: Decreasing prejudice through induced compliance. Journal of Personality and Social Psychology, 67, 395–413. McGregor, I., Newby-Clark, I. R., & Zanna, M. P. (1999). Remem￾bering dissonance: Simultaneous accessibility of inconsistent cog￾nitive elements moderates epistemic discomfort. In E. Harmon￾Jones & J. Mills (Eds.), Cognitive dissonance: Progress on a pivotal theory in social psychology (pp. 325–353). Washington, DC: American Psychological Association. Mitchell, J. P., Nosek, B. A., & Banaji, M. R. (2003). Contextual variations in implicit evaluation. Journal of Experimental Psychol￾ogy: General, 132, 455–469. Shultz, T. R., & Lepper, M. R. (1996). Cognitive dissonance reduction as constraint satisfaction. Psychological Review, 103, 219–240. Sloman, S. A. (1996). The empirical case for two systems of reasoning. Psychological Bulletin, 119, 3–22. Smith, E. R., & DeCoster, J. (2000). Dual-process models in social and cognitive psychology: Conceptual integration and links to under￾lying memory systems. Personality and Social Psychology Review, 4, 108–131. Steffens, M., & Plewe, I. (2001). Items
cross-category associations as a confounding factor in the Implicit Association Test. Zeitschrift f€ur Experimentelle Psychologie, 48, 123–134. Strack, F., & Deutsch, R. (in press). Reflective and impulsive determinants of social behavior. Personality and Social Psychology Review. Wilson, T. D., Lindsey, S., & Schooler, T. Y. (2000). A model of dual attitudes. Psychological Review, 107, 101–126. Zanna, M. P., & Cooper, J. (1974). Dissonance and the pill: An attribution approach to studying the arousal properties of disso￾nance. Journal of Personality and Social Psychology, 29, 703–709. 542 B. Gawronski, F. Strack / Journal of Experimental Social Psychology 40 (2004) 535–542