IMPLICIT AND EXPLICIT ATTITUDE CHANGE 709 of thie of EC If people rely on th attitudinal behavior(and thus changed their explici ve judgments of the target. ond.participants may use the pendent evidence for the validity of the measure used.thus imply on of the target via p ositional re evaluations would be more ave eval This dua king versus oking s e and Thu even though the APE del predicts nul Even though all bu es.such as by n correlational patterns (e.g..Gawr ne tha rcal-life ndcpliei negativity toward spiders Discussion with such a he are (e.g..pict videos.real nge.the APE model.and (b) vide an intee ttacks)In a simila ein.Rudman A ch 001 the av an d ex ided bu the the Dlicit and at implicit and explic As with Teachm f their underlying menta nd whethe given factor leads to n implicit or explici with tha hav nf d in the fnla studies we have dis ssed is of process mediates changes in the other.Thisc involve many kind plicit but no Ape model may provide a useful fram Hilton.2001) changes in both explicit and implicit attitudes (eg M A Olson F zio.2001).As such.the APE model goes s beyon g..Chaiken et al.1989:Kruglanski&Thom n.1999:Petty which a iven factor shoud in etation of null effects in long-ter idio lations.For instance.with inability to explain any kind of di g th ognitive d ld obi that th een low thus d explicit attitudes.In the remainder of this articl we provide asis of th such that tion for the propos account direction repeatedly presenting a person (CS) with the label child molester (US) should result in a negative associative evaluation of this person via processes of EC. If people rely on their associative evaluations when making an evaluative judgment, this direct in￾fluence on associative evaluations should indirectly influence eval￾uative judgments of the target. Second, participants may use the provided negative information about the target to infer a negative evaluation of the target via processes of propositional reasoning. The deliberate retrieval of relevant category information in the course of propositional reasoning, in turn, may influence the particular pattern of associations that is activated (J. P. Mitchell et al., 2003), thus leading to an indirect effect on associative evalu￾ations. This dual pathway implies that evaluative judgments may be indirectly affected via associative evaluations even when par￾ticipants have no explicit memory for the category information (see Case 1; e.g., Castelli et al., 2004, Experiments 4 and 5). Moreover, associative evaluations may be indirectly affected via processes of propositional reasoning even when the provided in￾formation is insufficient to result in a direct EC effect (see Case 4). Mixed influences and null effects. Even though all but one of the cases discussed so far have their empirical counterparts in the literature, it is reasonable to assume that many real-life situations involve multiple, mixed influences. Teachman and Woody (2003), for example, demonstrated treatment-related reductions in both implicit and explicit negativity toward spiders in people with spider phobia. This treatment involved a cognitive–behavioral therapy with several components, such as gradually increasing exposure (e.g., pictures, videos, real spiders) and education about the harmlessness of spiders (e.g., analysis of evidence for spider attacks). In a similar vein, Rudman, Ashmore, and Gary (2001) found a reduction of both implicit and explicit prejudice against African Americans in White students after a 14-week seminar on prejudice and conflict. As with Teachman and Woody’s (2003) cognitive–behavioral therapy, this seminar included multiple com￾ponents (e.g., education about socially determined inequalities, contact with an African American professor) that might have contributed to the final outcome. The crucial difference between these field studies and the laboratory studies we have discussed is that field studies often involve many kinds of simultaneous or sequential influences that are quite difficult to disentangle. Nev￾ertheless, the APE model may provide a useful framework for investigating the contribution of particular processes, which may then be tested outside of the laboratory for their real-life implications. A final issue concerns null effects of experimental manipula￾tions on explicit and implicit attitudes. In the preceding sections, we discussed several cases in which a given factor should influ￾ence explicit but not implicit attitudes or implicit but not explicit attitudes. Even though the interpretation of null effects poses several problems (e.g., lack of statistical power, beta error), the APE model circumvents this problem by predicting specific pat￾terns of correlations. For instance, with regard to Gawronski and Strack’s (2004) finding that cognitive dissonance changed only explicit but not implicit attitudes, one could object that the reli￾ability of the implicit measure might have been low, thus under￾mining significant effects on this measure. However, this interpre￾tation can be ruled out on the basis of the predicted correlational patterns, such that explicit and implicit attitudes were highly correlated when participants had a situational explanation for their counterattitudinal behavior as well as under control conditions but not when participants did not have a situational explanation for their counterattitudinal behavior (and thus changed their explicit attitudes). Notwithstanding these findings, there are a few studies in the literature that obtained null effects without providing inde￾pendent evidence for the validity of the measure used, thus imply￾ing all problems typically associated with the interpretation of null effects. For instance, Glen and Banse (2004) found that interviews focusing on personal strengths versus personal deficits did not lead to the predicted changes in self-evaluations, such that self￾evaluations would be more negative after participants focused on personal weaknesses than on personal strengths. This null effect held for both explicit and implicit self-evaluations. In a similar vein, Huijding, de Jong, Wiers, and Verkooijen (2005) found that smoking versus nonsmoking settings did not influence smokers’ attitudes toward smoking. Again, this null effect was obtained for both explicit and implicit attitudes. Thus, even though the APE model predicts null effects on either explicit or implicit attitudes under certain conditions, it seems important to independently establish the validity of the respective measures, such as by means of predicted correlational patterns (e.g., Gawronski & Strack, 2004). Discussion 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 theoretical assumptions of the APE model. In particular, we argue that implicit and explicit attitudes should be understood in terms of their underlying mental processes, which are associative and propositional processes. Thus, whether a given factor leads to changes in implicit or explicit attitudes should depend on (a) which of the two processes is influenced in the first place and (b) whether a change in one kind of process mediates changes in the other. This conceptualization provides an integration of a broad range of seemingly inconsistent patterns of attitude change, such as changes in explicit but not implicit attitudes (e.g., Gawronski & Strack, 2004), changes in implicit but not explicit attitudes (e.g., Karpinski & Hilton, 2001), and changes in both explicit and implicit attitudes (e.g., M. A. Olson & Fazio, 2001). As such, the APE model goes beyond previous models that either imply only a single attitude construct (e.g., Chaiken et al., 1989; Kruglanski & Thompson, 1999; Petty & Cacioppo, 1986; Van Overwalle & Siebler, 2005) or claim that implicit attitudes reflect highly robust attitudes that have their origin in long-term socialization experiences (e.g., Dovidio et al., 2001; Petty et al., 2006; Rudman, 2004; Wilson et al., 2000). Whereas the former class of models is limited in its explanatory power by its inability to explain any kind of dissociation between explicit and implicit attitude change, the latter class of models is limited by its inability to explain changes in implicit but not explicit attitudes. In the remainder of this article, we provide a final discussion of (a) the relation of the APE model to persuasion theories of attitude change, (b) the relative stability of attitude changes, (c) limits of a single-process approach as compared with the proposed dual-process account, (d) directions for future re￾IMPLICIT AND EXPLICIT ATTITUDE CHANGE 709