716 GAWRONSKI AND BODENHAUSEN tween cultural and personal ass ciations as problematic.Froma OlsonFazio.2003).Thus.whe 998),the pre s derived from APE model,it is important to conside the attitud diffe a function of the source.That is,the of a nal v al part ries and whether the evaluation level in mplic measure corresponds to the one implied in the expli implausible from the ctive of res Third.manym ures of implicit attitudes involve a notion of memory that suggests an independent representationo di.me research on the sleeper effect,which also sugge that sour ures are assesse epe ever.such these considerations.we argue that the personal character of as nly sociaig s is determined by their endors ion).Thus fact s that direct affect the mediating ot differ as a f lead to misl Rathe some of the to be and 12005) ay b an lead make some tasks more likely to reveal evaluations at are p by mean 005 utsch.&Seidel.2005).wh rey et al 2005)However this per onal character is deter mined by propositional processes rather than by the representation of evalu 2005).Th Conclusion to i d mis The main goals of the present article are (a)to ose a new theoretical model for the study of explicit and implicit attitude iation Test (Greenwald et al.. )for example.sev in IAT s M ssumptions of the ape d W. ith this e Sherma and Gr 200 n 2003.Lieberman et al..2002: 2004).w 10081 able to dise mith&DeCoster.2000 Strack& Deutsch I be of fo y different p are all iations.and (d)ge sing biase Applied t ture and changes in pattern activation.fot evaluative iude nents oth the relevant processes include changes in associative evaluatio nges I vario that can imply olicit a n Test is DE evidence on explicit and implicit attitude change for a large yariet which thus its of areas (e.g.,cognitive dissonance,EC,priming,persuasion).I ociat pe nally endorsed ciation arh and thus mat ine of the unde nalized variant of the IAT that was de ned to reduce the lying dynamics of associative and propositional processes in mpact of cultural associations distinctionbeparticular categories that are applied to these exemplars within the task (e.g., De Houwer, 2001; J. P. Mitchell et al., 2003; M. A. Olson & Fazio, 2003). Thus, when testing hypotheses derived from the APE model, it is important to consider whether the attitude object in question is represented on the exemplar level or on the level of general categories and whether the evaluation level in the implicit measure corresponds to the one implied in the explicit measure. Third, many measures of implicit attitudes involve a notion of response compatibility (De Houwer, 2003b; see also Kornblum, Hasbroucq, & Osman, 1990). Automatic evaluations in these mea￾sures are assessed by means of the (in)compatibility of the re￾sponse tendency elicited by an automatic affective reaction with the accurate response required by the task. However, such re￾sponse compatibility mechanisms tap mental associations only indirectly rather than directly (i.e., by means of response activa￾tion). Thus, factors that directly affect the mediating variable (e.g., response activation) can sometimes lead to misleading results that do not reflect a genuine change in association activation. For instance, Gawronski and Bodenhausen (2005) and Gawronski, Deutsch, and Seidel (2005) recently demonstrated that increasing stimulation of associations in memory can lead to contrast effects (rather than additive effects) on implicit measures that involve a notion of response compatibility (Gawronski & Bodenhausen, 2005; Gawronski, Deutsch, & Seidel, 2005), whereas increasing stimulation leads to the expected additive effects on measures that do not involve a notion of response compatibility (Balota & Paul, 1996; Gawronski & Bodenhausen, 2005). Thus, when testing the impact of external factors on implicit attitude measures, it is generally important to consider the mechanisms underlying these measures to avoid misinterpretations of the obtained results. Fourth, it is important to acknowledge that implicit attitude measures are not process pure. With regard to the Implicit Asso￾ciation Test (Greenwald et al., 1998), for example, several studies indicated that various nonassociative processes may contribute to systematic variance in IAT scores (Brendl, Markman, & Messner, 2001; McFarland & Crouch, 2002; Mierke & Klauer, 2003; Ro￾thermund & Wentura, 2004). Consistent with this claim, Conrey, Sherman, Gawronski, Hugenberg, and Groom (2005) recently presented a multinomial model (see Batchelder & Riefer, 1999; Klauer & Wegener, 1998) that is able to disentangle the contribu￾tion of four qualitatively different processes on implicit task per￾formance: (a) automatic activation of associations, (b) discrim￾inability of the stimulus, (c) success at overcoming automatic associations, and (d) general guessing biases. Applied to the present question, such models are a desirable way to disentangle the genuine contribution of associative evaluations from other nonassociative processes when investigating the mutual interplay between associative evaluations and evaluative judgments. Finally, some researchers have argued that implicit measures differ as to whether they tap cultural or personal associations. M. A. Olson and Fazio (2004), for example, argued that the standard variant of the Implicit Association Test is contaminated by cultural associations (i.e., associations endorsed by other peo￾ple), which thus undermines its usefulness as a measure of per￾sonal associations (i.e., personally endorsed associations). To solve this problem, M. A. Olson and Fazio (2004) proposed a personalized variant of the IAT that was designed to reduce the impact of cultural associations. We consider this distinction be￾tween cultural and personal associations as problematic. From a representational perspective (Smith, 1998), the proposed distinc￾tion implies that the representation of associations in memory can differ as a function of their source. That is, the source of an association (i.e., personal vs. cultural) must be an essential part of the representation of an association. This assumption, however, seems implausible from the perspective of research on source memory that suggests an independent representation of source information in memory (for a review, see M. K. Johnson, Hash￾troudi, & Lindsay, 1993). A similar conclusion can be drawn from research on the sleeper effect, which also suggests that source and content information are stored independently in memory (for a meta-analysis, see Kumkale & Albarracı´n, 2004). On the basis of these considerations, we argue that the personal character of as￾sociations is determined by their endorsement on a propositional level. That is, the representation of associations in memory does not differ as a function of whether they are personal or cultural. Rather, some of these associations may be more likely to be endorsed, whereas others may be rejected as a basis for evaluative judgments. Moreover, procedural differences among tasks may make some tasks more likely to reveal evaluations that are per￾sonally endorsed on a propositional level—for example, by means of a stronger impact of deliberate control on task performance (see Conrey et al., 2005). However, this personal character is deter￾mined by propositional processes rather than by the representation of evaluations in associative memory. Conclusion The main goals of the present article are (a) to propose a new theoretical model for the study of explicit and implicit attitude change, the APE model, and (b) to provide an integrative, exhaus￾tive review of the available evidence regarding implicit and ex￾plicit attitude change that is guided by the assumptions of the APE model. Drawing on the general notion of associative and proposi￾tional processes implied by several dual-systems models of the mind (e.g., Kahneman, 2003; Lieberman et al., 2002; Sloman, 1996; Smith & DeCoster, 2000; Strack & Deutsch, 2004), we argue that explicit and implicit attitudes should be understood in terms of their underlying mental processes. Hence, different kinds of attitude changes may be better understood by the particular way these processes are affected by external influences. For associative evaluations, these processes include changes in associative struc￾ture and changes in pattern activation. For evaluative judgments, the relevant processes include changes in associative evaluations, changes in the considered set of propositions, and changes in the strategy to achieve propositional consistency. These processes may occur in various patterns that can imply corresponding or noncor￾responding changes in implicit and explicit attitudes. Thus, the APE model provides a theoretical integration of the available evidence on explicit and implicit attitude change for a large variety of areas (e.g., cognitive dissonance, EC, priming, persuasion). It also implies a number of new predictions that may stimulate future research and thus may offer a better understanding of the under￾lying dynamics of associative and propositional processes in evaluation. 716 GAWRONSKI AND BODENHAUSEN