《社会心理学》课程教学资源（文献资料）attitude&stereotype——Associative and Propositional Processes in Evaluation：An Integrative Review of Implicit and Explicit Attitude Change.pdf_P11-P14

ators may influence (a) which propositions within a given set of relevant propositions are considered as valid or (b) which other propositions may be considered relevant in addition to the central propositions we have discussed. Interplay of Explicit and Implicit Attitude Change It is important to note that the different kinds of influences on associative and propositional processes do not occur in isolation. Rather, changes in associative and propositional processes may or may not affect each other, thus leading to different patterns of explicit and implicit attitude change. Figure 2 depicts the general model that is used as a framework for discussing different patterns of explicit and implicit attitude change. According to the APE model, a given factor may influence the activation of associations in memory or processes of propositional reasoning (or both). Moreover, propositional reasoning may or may not lead to a rejection of the propositional implication of an associative evaluation, determining whether evaluative judgments will be influenced by automatic affective reactions. Finally, processes of propositional reasoning may or may not influence the activation of associations in memory, thus determining the associative evaluation of an attitude object. In other words, a given factor may influence explicit attitudes either directly or indirectly via changes in implicit attitudes. Conversely, a given factor may influence implicit attitudes either directly or indirectly via changes in explicit attitudes. As such, the APE model implies several patterns of how changes in one kind of evaluation may or may not be mediated by changes in the other kind of evaluation (i.e., no mediation, partial mediation, full mediation; see Baron & Kenny, 1986). In the following sections, we discuss the various patterns of explicit and implicit attitude change implied by the APE model and how these cases are reflected in previous research on explicit and implicit attitude change (see Table 1). In this context, we also outline some new predictions that have not yet been subjected to empirical tests and thus may stimulate further research under the framework of the APE model. For each of the theoretical cases, we present a diagram to convey the key aspects of the case. In these diagrams, the relation between associative and propositional processes is conveyed by arrows; solid arrows reflect a causal influence, whereas open arrows reflect a lack of influence. Case 1. The intuitively simplest case involves a direct influence on associative evaluations, which, in turn, provide the basis for evaluative judgments (see Figure 3). This pattern implies corresponding changes in implicit and explicit attitudes, with changes in explicit attitudes being fully mediated by changes in implicit attitudes. According to the APE model, such patterns should emerge when (a) a given factor leads to a change in pattern activation or associative structure and, additionally, (b) associative evaluations are consistent with the momentarily considered set of subjectively valid propositions. An example fitting this pattern can be found in research on EC. Studies in this area have consistently shown that repeated CS–US pairings influence subsequent CS evaluations (for reviews, see De Houwer et al., 2001, 2005; Walther et al., 2005). Whereas previous studies have shown EC effects primarily for evaluative judgments, recent research using both explicit and implicit measures has demonstrated parallel effects for associative evaluations and evaluative judgments (e.g., Hermans et al., 2002; M. A. Olson & Fazio, 2001). M. A. Olson and Fazio (2001), for example, found corresponding influences of EC on both explicit and implicit attitudes, with the two being highly correlated. Most important, a reanalysis of M. A. Olson and Fazio’s (2001) data revealed that changes in explicit attitudes were fully mediated by corresponding changes in implicit attitudes. That is, EC effects on explicit attitudes decreased to nonsignificance after the researchers controlled for implicit attitudes (M. A. Olson, personal communication, April 5, 2004). Moreover, changes in implicit attitudes were not mediated by explicit attitudes, as suggested by a significant effect of EC on implicit attitudes after the researchers controlled for explicit attitudes (M. A. Olson, personal communication, January 4, 2006). In other words, EC changed the associative representation of the attitude object, which then led to corresponding changes in evaluative judgments. A similar pattern can be assumed for in-group favoritism in minimal group settings (Tajfel et al., 1971). As we have already outlined, previous research has consistently shown that minimal group settings are sufficient to induce a preference for in-groups over out-groups (for a review, see Hewstone, Rubin, & Willis, 2002). As with research on EC, such effects were originally demonstrated for explicit attitudes. Recent research found similar effects for implicit attitudes (e.g., Ashburn-Nardo et al., 2001; Castelli et al., 2004; DeSteno et al., 2004; Otten & Wentura, 1999; Pratto & Shih, 2000). Even though there is no evidence for a particular pattern of mediation available yet, the APE model implies that minimal group situations may change associative evaluations of the in-group, which then provide a basis for evaluative judgments about the in-group. That is, minimal group settings may create an association between the new in-group and the self. The associative evaluation of the self may then transfer to the new in-group (see also Cadinu & Rothbart, 1996; Gramzow & Gaertner, 2005; Greenwald & Banaji, 1995; Otten & Wentura, 2001). Given that most people have positive associative evaluations of the self (Bosson et al., 2000; Greenwald & Farnham, 2000; Koole et al., 2001), this evaluative transfer should lead to more positive associative evaluations of the in-group as compared with the out-group. Moreover, because evaluative judgments are usually based on the propositional implications of associative evaluations, this associative transfer of evaluations should lead to in-group favoritism not only for associative evaluations but also for evaluFigure 2. Processes underlying explicit and implicit attitude change, depicting potential direct and indirect influences on associative evaluations as determinants of implicit attitudes and propositional reasoning as determinants of explicit attitudes. Thin arrows depict direct effects on associative evaluations and propositional reasoning, whereas fat arrows depict indirect effects. 702 GAWRONSKI AND BODENHAUSEN

Table 1 Patterns of Implicit and Explicit Attitude Change and Their Conditions Case Theoretical description Implicit attitude change Explicit attitude change Mediation pattern Conditions of applicability Example 1 Indirect influence on propositional reasoning mediated by direct influence on associative evaluation Yes Yes Explicit attitude change fully mediated by implicit attitude change Factor leads to change in pattern activation or associative structure; associative evaluations are consistent with momentarily considered set of subjectively valid propositions Evaluative conditioning without contingency awareness 2 Direct influence on associative evaluation with neither direct nor indirect influence on propositional reasoning Yes No No mediation; no relation between explicit and implicit attitudes Factor leads to change in pattern activation or associative structure; associative evaluations are inconsistent with momentarily considered set of subjectively valid propositions Evaluative conditioning with contingency awareness 3 Direct influence on propositional reasoning with neither direct nor indirect influence on associative evaluation No Yes No mediation; no relation between explicit and implicit attitudes Factor provides a basis for new propositions for propositional reasoning; propositional reasoning leads to retroactive negation of (old) associative evaluation Cognitive dissonance resulting from counterattitudinal behavior 4 Indirect influence on associative evaluation mediated by direct influence on propositional reasoning Yes Yes Implicit attitude change fully mediated by explicit attitude change Factor provides a basis for new propositions for propositional reasoning; propositional reasoning leads to proactive construction of (new) associative evaluation Recategorization or reappraisal of attitude object 5 Direct influence on associative evaluation and direct antagonistic influence on propositional reasoning Yes Yes No mediation; no (or negative) relation between explicit and implicit attitudes Factor leads to change in pattern activation or associative structure; associative evaluations are inconsistent with momentarily considered set of subjectively valid propositions; same external factor provides a basis for new propositions that imply an evaluation of the opposite valence Evaluative conditioning with antagonistic, immediate shortterm and delayed long-term consequences 6 Direct influence on propositional reasoning with additional indirect influence on propositional reasoning mediated by direct influence on associative evaluation Yes Yes Explicit attitude change partially mediated by implicit attitude change Factor leads to change in pattern activation or associative structure; associative evaluations are consistent with momentarily considered set of subjectively valid propositions; external factor provides a basis for new propositions that indirectly imply an evaluation of the same valence Postdecisional dissonance with simultaneous evaluative conditioning effect with self as unconditioned stimulus 7 Direct influence on associative evaluation with additional indirect influence on associative evaluation mediated by direct influence on propositional reasoning Yes Yes Implicit attitude change partially mediated by explicit attitude change Factor leads to change in pattern activation or associative structure; associative evaluations are inconsistent with momentarily considered set of subjectively valid propositions; external factor provides a basis for new propositions that directly imply an evaluation of the same valence (None; case highly unlikely according to APE model) 8 Direct influence on associative evaluation and propositional reasoning with mutual indirect influences Yes Yes Explicit attitude change partially mediated by implicit attitude change; implicit attitude change partially mediated by explicit attitude change Factor leads to change in pattern activation or associative structure; associative evaluations are consistent with momentarily considered set of subjectively valid propositions; external factor provides a basis for new propositions for propositional reasoning that directly imply evaluation of same valence Persuasive message leading to evaluative conditioning effects and corresponding influence on propositional reasoning Note. APE associative–propositional evaluation. IMPLICIT AND EXPLICIT ATTITUDE CHANGE 703

ative judgments. This interpretation implies that (a) explicit ingroup favoritism in minimal group settings should be mediated by implicit in-group favoritism and (b) in-group favoritism should be stronger for participants with high rather than low implicit selfesteem (cf. Rubin & Hewstone, 1998). Even though this interpretation of in-group favoritism is obviously post hoc, future research may provide empirical support for this assumption.7 Another instance of indirect influences on explicit attitudes arises when changes in pattern activation lead to corresponding changes in evaluative judgments. An illustrative example for this case is a study by Richeson and Nussbaum (2004). In their study, participants were presented with messages advocating either a color-blind or a multicultural approach to reducing interethnic tension. Participants evaluated African Americans less negatively after reading a multicultural proposal than after reading a colorblind proposal. This effect emerged for both explicit and implicit attitudes, with changes in explicit attitudes being fully mediated by changes in implicit attitudes (J. A. Richeson, personal communication, April 5, 2004). In contrast, changes in implicit attitudes were not mediated by explicit attitudes, as suggested by a significant effect on implicit attitudes after the researchers controlled for explicit attitudes (J. A. Richeson, personal communication, April 5, 2004). Drawing on the APE model, one could argue that the two ideological perspectives activated different patterns of the preexisting associative representation of African Americans. These resulting associative evaluations, in turn, served as a basis for participants’ evaluative judgments about African Americans. Most important, the two messages did not imply different evaluations of African Americans, which could have led to a direct influence on evaluative judgments. In contrast, it seems that the two ideological approaches were associated with different construals of African Americans on the associative level, which, in turn, affected participants’ evaluative judgments. Case 2. The second case involves a direct influence on associative evaluations, with neither a direct nor an indirect influence on evaluative judgments (see Figure 4). This case implies a change in implicit but not explicit attitudes, with the two being generally uncorrelated. According to the APE model, such patterns should emerge when (a) a given factor leads to a change in pattern activation or associative structure and, additionally, (b) other relevant propositions lead to a rejection of associative evaluations as a valid basis for an evaluative judgment. An illustrative example for this pattern is again found in research on EC, which has shown that conditioning effects on evaluative judgments are often reduced when participants become aware of the contingency between CS und US (De Houwer et al., 2001; Fulcher & Hammerl, 2001; Walther, 2002). From the perspective of the APE model, one could argue that contingency awareness reduces participants’ trust in the validity of their associative evaluations, such that actual changes in associative structure are not reflected in evaluative judgments. However, associative evaluations may still be influenced by EC manipulations, even when contingency awareness undermines corresponding effects for evaluative judgments. In other words, even when contingency awareness attenuates the indirect effect of EC on explicit attitudes, implicit attitudes may nevertheless show a direct influence of EC manipulations. Preliminary evidence for this assumption comes from Karpinski and Hilton (2001). In their study, participants were presented with pairings of the words youth and elderly (CS) with positive and negative words (US). Participants’ task was to memorize the specific word pairings and the frequency of their occurrence. Results showed that participants exhibited a lower level of implicit prejudice against older adults when youth was repeatedly paired with negative words and elderly was repeatedly paired with positive words. However, participants exhibited a higher level of implicit prejudice against older people when youth was repeatedly paired with positive words and elderly was repeatedly paired with negative words. It is interesting to note that explicit attitudes were neither affected by conditioning manipulations nor significantly related to implicit attitudes. Drawing on the considerations outlined above, it seems that memorization instructions made participants highly aware of the particular contingencies implied by the word pairings, which, in turn, reduced participants’ trust in the validity of their associative evaluations. However, because contingency awareness may only undermine the influence of associative evaluations on propositional evaluations, the direct influence of EC on associative evaluations may still be unaffected. In other words, EC with contingency awareness may still change implicit attitudes even when explicit attitudes are unaffected. Asymmetrical influences of this sort would, of course, also be expected when external stimuli lead to changes in pattern activation and perceivers reject their associative evaluations as a valid basis for an evaluative judgment. Examples of such effects are implied by research on exemplar accessibility. As previously noted, Dasgupta and Greenwald (2001) have shown that implicit prejudice against African Americans decreased when participants previously had been presented with pictures of admired Black and disliked White individuals. Similarly, Lowery et al. (2001) found that the mere presence of an African American experimenter reduced White participants’ implicit prejudice against African Americans. In both studies, explicit evaluations of African Americans were unaffected by accessible exemplars. Drawing on the APE model, one could argue that the previously encountered exemplars led to a change in pattern activation. However, participants might have rejected their associative evaluations of African Americans as a valid basis for an evaluative judgment. Such 7 Note that this interpretation explains in-group favoritism but not outgroup derogation (Brewer, 1999; Hewstone et al., 2002). In particular, the proposed transfer of associative self-evaluations to the in-group should influence associative evaluations of the in-group but not the out-group. Figure 3. Case 1: Indirect influence on propositional reasoning mediated by a direct influence on associative evaluations, leading to corresponding changes in explicit and implicit attitudes. Solid arrows reflect a causal influence, whereas open arrows reflect a lack of influence. 704 GAWRONSKI AND BODENHAUSEN

rejections seem particularly likely in the domain of prejudice and stereotyping, in which people may commonly consider other propositions to be more relevant for an evaluative judgment than the propositional implications of their associative evaluations. We come back to this issue in the context of theories of persuasion when we discuss motivational influences on evaluative judgments. The abovementioned interpretation is also applicable to recent findings by M. A. Olson and Fazio (2006). These researchers found that EC influenced only implicit but not explicit prejudice against African Americans, with explicit and implicit prejudice being generally uncorrelated. In terms of the considerations we have outlined, one could argue that for prejudice-related judgments people commonly consider other propositions to be more relevant than the propositional implications of their associative evaluations. Thus, even though Olson and Fazio’s participants were generally unaware of the CS–US pairings of their conditioning manipulation, they might have generally tended to reject their associative evaluations as a basis for evaluative judgments about African Americans. As such, EC should influence only implicit but not explicit prejudice. Case 3. The third case involves a direct influence on propositional reasoning, with neither a direct nor an indirect influence on associative evaluations (see Figure 5). This case implies a change of explicit but not implicit attitudes, with the two being generally uncorrelated. According to the APE model, such a pattern should emerge when (a) a given factor provides a basis for new propositions for propositional reasoning and (b) these propositions lead to a retroactive rejection of already activated associative evaluations. An illustrative example for this case is provided by Gregg et al. (2006). In their research, participants received positive or negative information about two hypothetical groups called “Niffite” and “Luupite.” Information was presented on a computer screen, with one of the two groups being consistently described as negative and the other being consistently described as positive. Afterward, participants completed measures of explicit and implicit attitudes toward the two groups. When participants had completed these measures, the experimenter explained that the particular pairing of positive and negative statements about the two groups was intended to be counterbalanced across participants and that the participant unfortunately had been run in the wrong condition. The experimenter then asked the participant to imagine a reversal of the positive and negative statements about the two groups and to complete the two attitude measures again. Results indicate that reversal instructions clearly affected participants’ evaluative judgments about the two groups. Associative evaluations, however, were generally unaffected, such that groups previously presented with positive statements were evaluated positively and groups previously presented with negative statements were evaluated negatively. In other words, participants changed their evaluative judgments about the two groups after considering the additional proposition that their associative evaluations were not valid. However, this additional proposition simply led to a rejection of the propositional implications of associative evaluations, not to a change of the underlying associative evaluations per se. This interpretation is also applicable to studies showing that instructions to fake a particular attitude influenced only explicit but not implicit attitudes (e.g., Banse, Seise, & Zerbes, 2001; Kim, 2003). Another example of explicit attitude change without corresponding changes in implicit attitudes is implied by research on cognitive dissonance (Festinger, 1957). As previously mentioned, cognitive inconsistency—and thus cognitive dissonance—arises when two propositions are regarded as true and one follows from the opposite of the other (Festinger, 1957). People can resolve this inconsistency either by rejecting one of the inconsistent propositions as false or by adding an additional proposition that resolves the inconsistency. Drawing on these considerations, Gawronski and Strack (2004) predicted that cognitive dissonance arising from counterattitudinal behavior (Festinger & Carlsmith, 1959) should affect only explicit attitudes, not implicit attitudes. Moreover, explicit attitudes should be related to implicit attitudes when cognitive dissonance is reduced by an additional proposition that resolves the inconsistency. However, explicit and implicit attitudes should be uncorrelated when cognitive dissonance is resolved by a rejection of associative evaluations as a valid basis for an evaluative judgment. These predictions were confirmed in two studies in which participants were asked to write an essay in favor of a counterattitudinal position (Gawronski & Strack, 2004). A similar pattern was recently presented by Nier (2005). In this study, explicit and implicit prejudices against African Americans were highly correlated when the implicit measure was introduced as a particular kind of “lie detector” (accuracy condition) but not when the task was described as a poor measure of people’s attitudes (inaccuracy condition). Moreover, explicit prejudice was significantly higher under accuracy than under inaccuracy conditions. Implicit attitudes, in contrast, were generally unaffected by task instructions. In terms of the APE model, these results suggest that participants based their evaluative judgments on their autoFigure 4. Case 2: Direct influence on associative evaluations with neither direct nor indirect influence on propositional reasoning, leading to implicit but not explicit attitude change. Open arrows reflect a lack of influence. Figure 5. Case 3: Direct influence on propositional reasoning with neither direct nor indirect influence on associative evaluations, leading to explicit but not implicit attitude change. Open arrows reflect a lack of influence. IMPLICIT AND EXPLICIT ATTITUDE CHANGE 705