《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_P6-P10

Summary Figure 1 provides a schematic depiction of our assumptions regarding the interplay of association activation and propositional reasoning in evaluation. We argue that implicit attitudes reflect automatic affective reactions resulting from the particular associations that are activated automatically when a person encounters a relevant stimulus. Explicit attitudes, in contrast, should be conceived as evaluative judgments about an attitude object that have their roots in processes of propositional reasoning. The crucial difference between associative and propositional processes is their dependency on truth values. Whereas the activation of associations can occur regardless of whether a person considers these associations as true or false, processes of propositional reasoning are generally concerned with the validation of propositions. Thus, even though evaluative judgments are usually based on automatic affective reactions, they can also be independent of associative evaluations when the propositional implication of an automatic affective reaction is considered invalid. Such rejections may occur when the propositional evaluation implied by an automatic affective reaction is inconsistent with other propositions that are considered relevant for the evaluative judgment. In the following sections, we discuss several factors that can influence association activation and propositional reasoning, thus leading to implicit or explicit attitude change. In this context, we also provide an integrative, exhaustive review of the available evidence on implicit and explicit attitude change, which is organized according to the various mechanisms implied by the APE model. Implicit and Explicit Attitude Change The conceptualization of implicit and explicit attitudes proposed by the APE model has a number of implications for attitude change. First, changes in each kind of evaluation can be due to several processes. Second, the particular interplay of associative and propositional processes may lead to various patterns of explicit and implicit attitude change. We start our review by discussing different causes of attitude change and then provide an overview of the potential interplay of implicit and explicit attitude changes. Causes of Implicit Attitude Change From the perspective of the APE model, changes in associative evaluations may reflect either (a) an incremental change in the associative structure or (b) a temporal change in the activation of preexisting patterns (Smith, 1996). Whereas the first case involves the learning of a new evaluation, the latter case reflects the differential activation of old evaluations that are already stored in associative memory. Changes in associative structure. The prototypical case for implicit attitude changes resulting from changes in associative structure is evaluative conditioning (EC; for reviews, see De Houwer, Baeyens, & Field, 2005; De Houwer, Thomas, & Bayens, 2001; Walther, Nagengast, & Trasselli, 2005). Consistent with this notion, several studies demonstrated changes in implicit attitudes resulting from repeated pairings of an attitude object with positive or negative stimuli (e.g., Baccus, Baldwin, & Packer, 2004; Dijksterhuis, 2004; Hermans, Baeyens, Lamote, Spruyt, & Eelen, 2005; Hermans, Vansteenwegen, Crombez, Baeyens, & Eelen, 2002; Karpinski & Hilton, 2001; C. J. Mitchell, Anderson, & Lovibond, 2003; M. A. Olson & Fazio, 2001, 2002, 2006; Petty et al., 2006). M. A. Olson and Fazio (2001), for example, developed a conditioning paradigm in which several hundred randomly presented words and images were interspersed with critical pairings of positive or negative unconditioned stimuli (USs) with neutral conditioned stimuli (CSs). Results indicated that CSs paired with positive USs exhibited a more positive valence in an Implicit Association Test than CSs paired with negative USs. This finding was replicated in several follow-up studies that used a subliminal affective priming task as a dependent measure (M. A. Olson & Fazio, 2002) and pictures of Black and White individuals as CSs (M. A. Olson & Fazio, 2006). Similar results were obtained by C. J. Mitchell et al. (2003), who found that the ostensible evaluative meaning randomly assigned to meaningless nonwords was subsequently reflected in implicit evaluations of these nonwords assessed with an Implicit Association Test. That is, nonwords that were randomly associated with a positive word showed a more positive implicit valence than nonwords that were randomly associated with a negative word. With regard to social stimuli, Hermans et al. (2002) showed that formerly neutral faces (CS) that were associated with negative electrocutaneous stimulation (US) exhibited a more negative implicit valence in an affective priming task than when these faces were not associated with electrocutaneous stimulation. It is interesting to note that EC effects resulting from aversive stimulation did not differ from EC effects resulting from pairings with acoustically presented negative words. Hermans et al. (2005) obtained similar results using food stimuli as CSs and pleasant versus unpleasant odors as USs. With regard to implicit self-evaluations, Dijksterhuis (2004) demonstrated that subliminal EC of self-related words can change implicit evaluations of the self. In particular, Dijksterhuis found that participants exhibited a higher level of implicit self-esteem Figure 1. Schematic depiction of the interplay of association activation and propositional reasoning in explicit and implicit attitudes. Input stimuli are assumed to determine the pattern of associations that gets activated in memory (bold lines connecting dots), which in turn influences affective reactions reflected in implicit attitude measures. Processes of propositional reasoning are assumed to assess the validity of activated associations, which provides the basis for evaluative judgments reflected in explicit attitude measures. IMPLICIT AND EXPLICIT ATTITUDE CHANGE 697

when self-related words (CS) were subliminally paired with positive adjectives (US) than when self-related words were subliminally paired with neutral words. These effects were consistent across different measures of implicit self-esteem (e.g., IAT, name letter effect). Baccus et al. (2004) obtained similar effects using a computer game that repeatedly paired self-relevant words (CS) with smiling, frowning, or neutral faces (US). The notion of EC is also reflected in a study by Karpinski and Hilton (2001). These researchers found that implicit prejudice against older adults was influenced by repeated pairings of the words youth and elderly (CS) with positive and negative words (US). In particular, participants exhibited a lower level of implicit prejudice against older people in an Implicit Association Test when youth was repeatedly paired with negative words and elderly was repeatedly paired with positive words than when youth was repeatedly paired with positive words and elderly was repeatedly paired with negative words. A special case of EC is represented by the creation of new associations between an attitude object and the self. Consistent with this notion, Walther et al. (2005) recently argued that the self can function as a US when it is “paired” with a neutral CS. Given that most people show positive evaluations of themselves (Bosson, Swann, & Pennebaker, 2000; Greenwald & Farnham, 2000; Koole, Dijksterhuis, & Van Knippenberg, 2001), the creation of a new association between an attitude object and the self should lead to more positive evaluations of the attitude object (e.g., Walther & Trasselli, 2003). It is important to note that, as the proposed conditioning process involves a mere associative transfer of selfevaluations to the attitude object, the resulting implicit attitude should depend on people’s implicit self-evaluation such that implicit positivity toward the attitude object should increase as a function of implicit positivity toward the self. Evidence for these assumptions can be found in a series of studies by Gawronski, Bodenhausen, and Becker (in press). Gawronski et al. found that choosing between two equally attractive pictures changed participants’ implicit evaluations of these pictures, such that they evaluated chosen pictures more positively after than before the decision (see Brehm, 1956). Moreover, postdecisional associative evaluations were moderated by implicit selfevaluations, such that postdecisional positivity toward chosen pictures increased as function of participants’ implicit positivity toward the self. It is important to note that this effect was independent of whether ownership resulted from participants’ decision or from random assignment by the experimenter. In terms of the APE model, these results suggest that mere ownership can create an association between the owned object and the self, which then leads to an associative transfer of implicit self-evaluations to the object. The notion of EC effects arising from new associations to the self is also consistent with research on implicit in-group favoritism. Using the minimal group paradigm (Tajfel, Billig, Bundy, & Flament, 1971), several researchers have found that minimal group settings are sufficient to induce an implicit preference for ingroups over out-groups (e.g., Ashburn-Nardo, Voils, & Monteith, 2001; Castelli, Zogmeister, Smith, & Arcuri, 2004; DeSteno, Dasgupta, Bartlett, & Cajdric, 2004; Otten & Wentura, 1999; Pratto & Shih, 2000). Given that participants usually have little or no declarative knowledge regarding such minimal groups, this finding may appear somewhat surprising. From the perspective of EC, however, one could argue that minimal group settings are sufficient to create an association between the new in-group and the self, which, in turn, should lead to an associative transfer of implicit self-evaluations to the new in-group (see also Cadinu & Rothbart, 1996; Gramzow & Gaertner, 2005; Greenwald & Banaji, 1995; Otten & Wentura, 2001; for a review, see Otten, 2003). Thus, given that most people’s implicit self-evaluation is highly positive (Bosson et al., 2000; Greenwald & Farnham, 2000; Koole et al., 2001), the association between self and in-group may be sufficient to produce an implicit preference for in-groups over out-groups in minimal group settings. Changes in pattern activation. A second cause for changes in associative evaluations is represented by changes in pattern activation. Such changes should occur when simple context cues are sufficient to influence which pattern of associations is activated for an already familiar attitude object. These cases differ from incremental changes in associative structure (e.g., by means of EC), such that no pairing with evaluative information is required to change implicit attitudes. Instead, changes in pattern activation presuppose (a) that the attitude object is already represented in a multifaceted manner and (b) that the presence of particular context cues activates different associative patterns reflecting different subsets of this representation. A first set of studies that is consistent with the notion of differential pattern activation is concerned with the effects of momentarily accessible group members. Dasgupta and Greenwald (2001), for example, found that implicit prejudice against African Americans was lower when participants were presented with pictures of admired Blacks and disliked Whites before they completed an implicit prejudice task than when they were presented with pictures of disliked Blacks and admired Whites. Drawing on basic assumptions by the APE model, we argue that the presentation of familiar exemplars activates different patterns of the preexisting associative structure related to Black and White people, which, in turn, influences participants’ associative evaluation of Black and White people in general. Because the pictures used in Dasgupta and Greenwald’s studies generally depicted highly familiar individuals (e.g., Michael Jordan, Tom Cruise), it seems likely that these pictures led to changes in the activation of already existing patterns rather than to changes in the underlying associative structure. This interpretation can also be applied to findings by Rudman and Lee (2002), who found that exposure to violent rap music led White participants to show more negative associative evaluations of Black people. In terms of the APE model, exposure to violent rap music may activate a particular associative pattern of participants’ representation of Black people. If this pattern involves a more negative evaluation of Black people, mere exposure to violent rap music may be sufficient to enhance implicit prejudice against African Americans. Similar findings were obtained by J. P. Mitchell, Nosek, and Banaji (2003). These researchers found that associative evaluations elicited by well-known individuals depended on whether these individuals were categorized in terms of their race or other applicable categories (e.g., occupation). For example, Michael Jordan elicited a positive associative evaluation when he was categorized as an athlete but a negative associative evaluation when he was categorized as African American. These effects were obtained when categorization was manipulated by the category labels in an IAT (Greenwald et al., 1998) or by the salience of 698 GAWRONSKI AND BODENHAUSEN

categories in a go/no-go association task (Nosek & Banaji, 2001). It is important to note that the targets used by J. P. Mitchell et al. were all highly familiar individuals. In other words, participants already had a representation of these individuals in their associative memory, but the particular category applied to these individuals influenced which of the associative patterns got activated. Particularly interesting with regard to the present question is research showing that simply increasing the salience of group categories can influence subsequent associative evaluations (e.g., Ku¨hnen et al., 2001; Pratto & Shih, 2000; J. R. Steele & Ambady, 2006). In a study by Ku¨hnen et al. (2001), for example, German participants were primed with the two subcategories East German and West German and then completed an IAT designed to assess implicit preference for West Germans over East Germans. Overall, the social stereotype of East Germans tends to be negative, whereas the social stereotype of West Germans tends to be positive. Results indicate that participants exhibited a stronger preference for West Germans over East Germans when they were primed with the two subcategories than when they were not primed with the two subcategories. It is interesting that this effect emerged for both East German and West German participants. Hence, activation of the in-group and out-group categories led to more or less favorable associative evaluations of the in-group, depending on the particular valence of in-group-related associations. Because Ku¨hnen et al.’s priming procedure involved a neutral category prime rather than direct priming of positive or negative stereotypes, these results are consistent with the notion of differential pattern activation. If the two subcategories were not already associated with positive or negative stereotypes in associative memory, simple category priming should leave associative evaluations unaffected. This interpretation is also applicable to a finding by J. R. Steele and Ambady (2006), who found that women showed more genderstereotypical implicit attitudes toward arts versus math (i.e., stronger preference for arts over math) when they were primed with gender categories than when they were primed with gender-neutral categories. Additional evidence for changes in pattern activation comes from Wittenbrink et al. (2001), who found that automatic affective reactions elicited by African American individuals were less negative when the individuals were presented in a positive context (e.g., family barbeque) than when they were presented in a negative context (e.g., gang incident). These results indicate that the associative pattern that is activated by a given individual can differ as a function of the context in which this individual is encountered. In the present case, one could ague that the associative representation of African Americans is ambivalent, such that African Americans are associated with both positive and negative aspects. However, which of these aspects gets activated depends on the particular context in which an African American individual is encountered. An interesting extension of Wittenbrink et al.’s (2001) research was recently presented by Barden et al. (2004; see also Maddux, Barden, Brewer, & Petty, 2005). These researchers showed that not the context per se but the social role within a particular context led to changes in automatic evaluations. In one study, for example, Barden et al. found that a Black person presented in a prison context elicited automatic negative reactions when this person was presented with dress suggesting the role of a prisoner. However, the same Black person elicited automatic positive reactions when presented with dress suggesting the role of a lawyer. As with Wittenbrink et al.’s (2001) findings, these results indicate that different context cues influence which associative pattern will be activated for a particular attitude object and that such differences in pattern activation can lead to different associative evaluations of the same attitude object. Most important, such changes in pattern activation seem to be capable of reversing the automatic activation commonly attributed to implicit prejudice (see Devine, 1989; Fiske, 1998), such that the same Black person may activate either a negative or a positive associative evaluation depending on the social role in which this person is encountered. Another study that can be interpreted in terms of differences in pattern activation was conducted by Lowery, Hardin, and Sinclair (2001). These researchers found that mere interaction with an African American experimenter was sufficient to reduce negative associative evaluations of African Americans. It is interesting to note that this effect emerged for European Americans but not for Asian Americans. Even though we cannot rule out the possibility that positive interactions with an African American experimenter may be sufficient to change participants’ associative structure (e.g., by means of EC), it is quite difficult to explain why this effect should emerge only for European Americans, not for Asian Americans (given that the affective quality of the interaction was equal across the two groups of participants). An alternative explanation in terms of the present model is that European Americans and Asian Americans differ with regard to their preexisting associative representation regarding African Americans, such that European Americans’ associative representation is more heterogeneous than Asian Americans’ associative representation. Hence, interacting with an African American experimenter may activate an evaluatively different pattern of associations for European Americans, whereas the patterns activated in Asian Americans may be less affected by individual interactions. Additional evidence for changes in pattern activation comes from research on social roles. Richeson and Ambady (2003) demonstrated that anticipated superior and subordinate roles in dyadic interactions with an African American individual influenced participants’ associative evaluations of African Americans in general. In particular, White participants showed more negative associative evaluations of African Americans when they anticipated being in a superior role than when they anticipated a subordinate role (see also Richeson & Ambady, 2001). However, situational roles had no influence on associative evaluations of African Americans when participants expected to interact with a White person. From the perspective of the APE model, one could argue that anticipated social roles in interactions with African Americans increased the salience of particular aspects of participants’ associative representation of African Americans. Because Richeson and Ambady’s (2003) participants only anticipated, rather than actually engaged in, social interaction, it seems unlikely that participants formed completely new associative evaluations. Instead, anticipated roles might have activated those patterns of the preexisting associative representation that provided the best fit to the anticipated role. Several studies have also shown that differences in the interpretation of abstract category labels can change affective responses on the IAT (e.g., Foroni & Mayr, 2005; Govan & Williams, 2004). Govan and Williams (2004), for example, found that the often demonstrated implicit preference for flowers over insects could be reversed when the stimuli in the IAT were unpleasant flowers (e.g., IMPLICIT AND EXPLICIT ATTITUDE CHANGE 699

skunkweed) and pleasant insects (e.g., butterfly). According to Govan and Williams, the particular stimuli in the IAT determine how respondents interpret the respective categories, thus influencing their affective reactions to flowers and insects in general. This interpretation is also consistent with research by Foroni and Mayr (2005), who found that the preference for flowers over insects could be attenuated by a fictional scenario implying a reversal of the typical category evaluations. In their study, participants were asked to imagine a post–nuclear war scenario in which flowers were generally contaminated and insects were the only kind of harmless food available. From the perspective of the APE model, these results suggest that different exemplars (Govan & Williams, 2004) or fictional scenarios (Foroni & Mayr, 2005) may be sufficient to activate different associative patterns, thus leading to different associative evaluations of flowers and insects. A particularly interesting case of changes in pattern activation comes from research on motivational states. Ferguson and Bargh (2004), for example, demonstrated that automatic associative evaluations of an attitude object differed as a function of the object’s relevance for goal pursuit. Objects were evaluated more positively when they were relevant than when they were irrelevant for goal pursuit (see also Brendl, Markman, & Messner, 2003). Similar findings were reported by Sherman, Rose, Koch, Presson, and Chassin (2003), who showed that nicotine deprivation led to more positive associative evaluations of cigarettes in heavy smokers. However, light smokers showed more positive associative evaluations of cigarettes when they had just smoked a cigarette than when they were deprived. Taken together, these results indicate that associative evaluations depend not only on external input stimuli but also on internal motivational states. Moreover, given the transient nature of motivational states, these influences are likely to result from differences in pattern activation rather than from genuine differences in the underlying associative structure. Another interesting line of research concerns the effect of emotional states on automatic associative evaluations. In a study by DeSteno et al. (2004), for example, anger but not sadness enhanced automatic negative evaluations of out-groups. According to DeSteno et al., these differences were due to the functional relevance of anger, in contrast to sadness, to intergroup conflict. This interpretation is also consistent with findings by Schaller et al. (2003), who showed that ambient darkness enhanced automatic negative evaluations of African Americans for participants with chronic beliefs in a dangerous world but not for participants who did not believe in a dangerous world. Similar differences were reported by Gemar, Segal, Sagrati, and Kennedy (2001), who found that recently recovered depressed patients showed more negative associative self-evaluations under sad mood as compared with control conditions. For never-depressed control participants, in contrast, mood had no effect on associative self-evaluations. Taken together, these results suggest that associative evaluations depend on emotional as well as motivational states. Moreover, as with transient motivational states, these influences are likely to result from differences in pattern activation rather than from genuine differences in the underlying associative structure. Directly related to the notion of emotional states is a finding by Frantz, Cuddy, Burnett, Ray, and Hart (2004). These researchers found that implicit prejudice scores on the IAT were generally higher when the task was introduced as a diagnostic instrument for the assessment of racism. This finding is consistent with the claim that increased arousal enhances dominant responses (Hull, 1943; Zajonc, 1965). That is, one could argue that automatic associations reflect a particular kind of dominant response (e.g., Lambert et al., 2003). Moreover, the diagnosticity instructions used by Frantz et al. (2004) are quite likely to result in evaluation apprehension, which typically increases arousal. This arousal, in turn, may enhance dominant responses, in this case the activation level of automatic associations. If this interpretation is correct, any kind of evaluation apprehension or arousal should enhance IAT effects, and this increase should emerge irrespective of whether the task is designed to assess implicit prejudice or any other kind of dominant response (e.g., implicit preference for flowers over insects). Causes of Explicit Attitude Change As with changes in implicit attitudes, changes in explicit attitudes can be due to a number of processes. According to the APE model, changes in evaluative judgments can be due to (a) a change of the associative evaluation of the attitude object, (b) a change in the set of propositions that are considered to be relevant for an evaluative judgment, or (c) a change in the strategy used to achieve consistency within a given set of propositions. Changes in associative evaluation. The first kind of explicit attitude change involves instances in which incremental changes in the associative structure or temporal changes in the activation of associative patterns lead to a change of the associative evaluation of the attitude object. Because people tend to base their evaluative judgments on their automatic affective reactions, such changes in associative evaluations may, in turn, lead to corresponding changes in evaluative judgments. With regard to changes in associative structure, a number of studies on EC have demonstrated that repeated pairings of CSs and USs influence subsequent evaluative judgments of the CSs. This research has been reviewed extensively elsewhere (De Houwer et al., 2001, 2005; Walther et al., 2005), and thus we refrain from an elaborate discussion of the relevant evidence. From the perspective of the APE model, however, it is important to note that processes of EC should not directly influence evaluative judgments. Rather, the influence of EC on evaluative judgments should be mediated by associative evaluations, which, in turn, influence evaluative judgments (see also De Houwer et al., 2001, 2005). This assumption has a number of important implications, which we discuss in more detail in the context of the interplay between explicit and implicit attitude changes. As with changes in associative structure, temporal changes in the activation of preexisting associative patterns can lead to corresponding changes in evaluative judgments. That is, one could argue that any context stimulus that leads to a change in pattern activation should influence the associative evaluation of an attitude object, which, in turn, may further influence corresponding evaluative judgments. As with the case of EC, however, it is important to note that such changes in evaluative judgments do not reflect a direct influence of context stimuli on evaluative judgments. Rather, such influences should also be mediated by associative evaluations. Evidence for explicit attitude changes resulting from differences in pattern activation comes from studies showing that recently encountered members of a social group influence judgments about the group in general (e.g., Bless, Schwarz, Bodenhausen, & Thiel, 700 GAWRONSKI AND BODENHAUSEN

2001; Bodenhausen, Schwarz, Bless, & Wa¨nke, 1995; Gawronski, Bodenhausen, & Banse, 2005; Henderson-King & Nisbett, 1996; Sia, Lord, Blessum, Thomas, & Lepper, 1999; Wilder, Simon, & Faith, 1996). From a general perspective, this research has demonstrated that participants evaluated a social group more negatively when they had recently encountered a negatively evaluated exemplar of the group. In contrast, participants evaluated the same group more positively when they had recently encountered a positively evaluated exemplar. Findings such as these are consistent with the assumption that recently encountered group members influence the associative pattern that gets activated for the group. Such changes in pattern activation should influence the associative evaluation of the group (e.g., Dasgupta & Greenwald, 2001), which, in turn, may serve as a basis for evaluation judgments about the group. Changes in the set of considered propositions. The second kind of influence on evaluative judgments involves changes in the set of propositions that are considered to be relevant for an evaluative judgment. According to the APE model, this may be the case when either (a) the acquisition of new propositional beliefs about the world or (b) the additional consideration of already familiar propositions implies a change in the evaluation of an attitude object. The first case is probably best reflected in research on persuasion (for reviews, see Chen & Chaiken, 1999; Petty & Wegener, 1999). In studies of persuasion, participants are usually exposed to a persuasive message containing either strong or weak arguments about an attitude object and different kinds of peripheral or heuristic cues, such as the expertise of the source, the likability of the source, or consensus information. A well-replicated finding is that under conditions of low cognitive elaboration, attitudes are more likely to be influenced by peripheral or heuristic cues rather than by the quality of the arguments (but see Kruglanski & Thompson, 1999). Under conditions of high elaboration, in contrast, attitudes are more likely to be influenced by the quality of arguments, whereas the influence of peripheral or heuristic cues is often (but not always) attenuated. From the perspective of the APE model, persuasive arguments are—by definition—propositional statements and thus should influence attitudes primarily by processes of propositional reasoning. In other words, exposing participants to persuasive arguments may add new propositions to the set of propositions that are considered to be relevant for an evaluative judgment. If such changes in the set of considered propositions imply a different evaluation of a given attitude object, exposure to persuasive arguments is quite likely to lead to explicit attitude change. However, if the changes in the set of considered propositions do not imply a different evaluation, exposure to persuasive arguments should not lead to explicit attitude change. We come back to this question in more detail when we compare the APE model with persuasion models of attitude change. The second case is represented by instances in which the additional consideration of already familiar propositions implies a change in evaluative judgments about an attitude object (see Judd & Lusk, 1984). Evidence for this kind of influence comes from research on mere thought (Tesser, 1978) or introspection (Wilson, Dunn, Kraft, & Lisle, 1989). Studies in these areas have demonstrated that merely thinking about an attitude object or introspecting about the reasons for an evaluation can change evaluative judgments about the object. From the perspective of the APE model, mere thought or introspection about reasons is likely to add new propositions to the set of propositions that are considered to be relevant for an evaluative judgment. If such additionally considered propositions imply a different evaluation, mere thought and introspection should change evaluative judgments about the attitude object. If, however, additionally considered propositions confirm the original evaluation, enhanced thinking about an attitude object should leave evaluative judgments unaffected (or lead to attitude polarization; see Tesser, 1978). These assumptions are consistent with research showing that introspection about the reasons, but not introspection on feelings, leads to changes in evaluative judgments about an attitude object (e.g., Millar & Tesser, 1986; Wilson & Dunn, 1986). Changes in the strategy to achieve consistency. The third kind of change in evaluative judgments involves changes in the strategy to achieve consistency. This kind of influence is prototypically reflected in research on cognitive dissonance. According to Festinger (1957), two cognitions are dissonant when, considered by themselves, one of them follows from the opposite of the other. Conversely, two cognitions are consonant when one of them does not follow from the opposite of the other. Because the notion of logical implication presupposed in this definition requires an assignment of truth values, cognitive dissonance can be regarded as an inherently propositional phenomenon. Consistent with this claim, Gawronski and Strack (2004) recently argued that both the causes of cognitive dissonance and the process of dissonance reduction are inherently propositional. With regard to the causes of cognitive dissonance, Gawronski and Strack argued that cognitive inconsistency—and thus cognitive dissonance—arises when two propositions are regarded as true and one follows from the opposite of the other. With regard to the process of dissonance reduction, Gawronski and Strack argued that people resolve cognitive inconsistency—and thus cognitive dissonance—either by explicitly rejecting one of the inconsistent propositions as false or by finding an additional proposition that resolves the inconsistency (see also Gawronski, Strack, & Bodenhausen, in press; Kruglanski, 1989). The latter process of resolving the inconsistency between propositions is particularly relevant for the present question of explicit attitude change. Changes in evaluative judgments may occur when people resolve the inconsistency within a set of relevant propositions by rejecting the propositional implications of an associative evaluation. However, people may not change their evaluative judgments about an attitude object when they find an additional proposition that resolves the inconsistency. These cases are well reflected in Festinger and Carlsmith’s (1959) induced compliance paradigm, in which counterattitudinal behavior leads to explicit attitude change only when participants do not have a situational explanation for their counterattitudinal behavior (i.e., when they reject the propositional implications of their associative evaluations), not when they can justify their counterattitudinal behavior with situational factors (i.e., when they find an additional proposition that resolves the inconsistency between their counterattitudinal behavior and the propositional implication of their associative evaluations). Moreover, these processes of achieving consistency may be affected by various moderators (e.g., HarmonJones, Brehm, Greenberg, Simon, & Nelson, 1996; Stalder & Baron, 1997; C. M. Steele & Lui, 1983; Stone & Cooper, 2003; for a review, see J. M. Olson & Stone, 2005), such that these moderIMPLICIT AND EXPLICIT ATTITUDE CHANGE 701