GAWRONSKI AND BODENHAUSEN swhen they ever.particip ted their atic affective bination（eg.aBlc basis for an evaluative judgment n they expecte ste t wond).Hence is not clea Nier's(2005)manipulations influenced explicit attitudes ck people was due to the neeation of the s o the al.'s findi judgm en driven by the of the cou stereotynical information led to a reduction in automatic stere should emerge when (a)a given s ee. ed showing that deliberate att empts to supp affective hether propositi lead to changes i ab of changing affectiv no( of the iooegingoociaioastawceaoatactivaied (20) In a study by individua nc negation is ami.Dovidio. and th Kawakami.Dovidio Kamp.2005).These hers found n the negatio des art inan sho 5 tional claims)From this standpoint.one would expect negation in pr ada given s individilalimnucncedthe important that K ffirmation of counterstere s.In one study.for mple.pa ment exhibited by ap of Black and he ource ind negati (positive ngruent pe ard the eaning of the o valuations This ir tion implies tha the ob Influence 个↓ balance-relate nferences she the tempora en ipants first leam about the ation about the by Gawronski.Walther. Another example for the present case is a recent study by Petty et al.(2006).These researchers first induced a positive or negativmatic affective reactions toward African Americans when they expected the implicit measure to reveal their true feelings. How￾ever, participants seem to have rejected their automatic affective reactions as a basis for an evaluative judgment when they expected the implicit measure to be a poor measure of attitudes. In other words, Nier’s (2005) manipulations influenced explicit attitudes, such that these manipulations determined whether participants used their automatic affective reactions as a basis for an evaluative judgment. Case 4. The fourth case involves a direct influence on prop￾ositional reasoning, which, in turn, influences associative evalua￾tions (see Figure 6). This pattern implies corresponding changes in implicit and explicit attitudes, with changes in implicit attitudes being fully mediated by changes in explicit attitudes. According to the APE model, such patterns should emerge when (a) a given factor provides a basis for new propositions for propositional reasoning (as, e.g., when persuasive arguments are encountered) and (b) these propositions lead to a proactive construction of new associative evaluations. Thus, the crucial difference determining whether propositional processes lead to changes in associative evaluations is whether propositional reasoning leads to a mere negation of evaluative associations that are already activated or to an affirmation of evaluative associations that were not activated before (see also Wegner, 1994). An illustrative example for the differential role of affirmation and negation is Kawakami, Dovidio, Moll, Hermsen, and Russin’s (2000) research on the negation of social stereotypes (see also Kawakami, Dovidio, & van Kamp, 2005). These researchers found that long-term training in the negation of social stereotypes re￾sulted in lower levels in the automatic stereotype activation. This finding might seem problematic for the current conceptualization, in which negation is conceived as a propositional reasoning oper￾ation (because it involves the assignment of truth values to prop￾ositional claims). From this standpoint, one would expect negation training to operate primarily by promoting rejection of the prop￾ositional implications of an association (Deutsch et al., in press). It is important to note, however, that Kawakami et al.’s (2000) negation training confounded the negation of stereotypes with the affirmation of counterstereotypes. In one study, for example, par￾ticipants were presented with pictures of Black and White people and traits that were related either to the stereotype of Blacks or to the stereotype of Whites. Participants’ task was to respond with a NO key each time they saw a stereotype-congruent person–trait combination (e.g., a Black face with a stereotypically Black trait word) and to respond with a YES key each time they saw a stereotype-incongruent person–trait combination (e.g., a Black face with a stereotypically White trait word). Hence, it is not clear whether the resulting reduction in automatic prejudice against Black people was due to the negation of the stereotype or to the affirmation of counterstereotype. Drawing on the considerations we have outlined, we argue that Kawakami et al.’s findings were exclusively driven by the affirmation of the counterstereotype rather than by the negation of the stereotype. This claim was recently confirmed in a study by Gawronski, Deutsch, and Mbirkou (2006), who found that only training in the affirmation of counterstereotypical information, not training in the negation of stereotypical information, led to a reduction in automatic stereo￾type activation. Moreover, the proposed difference between affir￾mation and negation focus is also consistent with research in other areas showing that deliberate attempts to suppress affective reac￾tions (negation) usually leave these reactions unaffected, whereas attempts to attribute a different meaning to the response-eliciting stimulus (affirmation) is indeed capable of changing affective reactions (e.g., Butler et al., 2003; Gross, 1998). Another example of the present case is research on cognitive balance (Heider, 1958). In a study by Gawronski, Walther, and Blank (2005), for example, participants first formed either positive or negative attitudes toward several “source” individuals (source valence) and then learned that these source individuals either liked or disliked another set of neutral “target” individuals (observed sentiment). Consistent with balance theory, participants showed more positive attitudes toward targets who were liked than toward those who were disliked by positive source individuals. In contrast, participants showed more negative attitudes toward targets who were liked than toward those who were disliked by negative source individuals. It is interesting to note that this effect emerged not only for explicit but also for implicit attitudes. From the perspec￾tive of the present framework, one could argue that a priori attitudes toward a given source individual influenced the interpre￾tation of this individual’s relation to another target individual. That is, participants might have interpreted a positive (negative) senti￾ment exhibited by a positively evaluated source individual as positive (negative) information about the target, whereas they might have interpreted a positive (negative) sentiment exhibited by a negatively evaluated source individual as negative (positive) information about the target. In other words, a priori attitudes toward the source individual proactively influenced participants’ inferences about the evaluative meaning of the observed sentiment, which, in turn, affected not only evaluative judgments but also associative evaluations. This interpretation implies that the ob￾tained effect of cognitive balance on associative evaluations should be mediated by processes of propositional reasoning. More￾over, balance-related inferences should leave associative evalua￾tions unaffected if the temporal order of information acquisition would require a retroactive qualification of previously observed sentiment relations (e.g., when participants first learn about the attitude of a neutral source individual toward a neutral target individual and then receive evaluative information about the source). This assumption was confirmed by Gawronski, Walther, and Blank (2005) in a series of three experiments. Another example for the present case is a recent study by Petty et al. (2006). These researchers first induced a positive or negative Figure 6. Case 4: Indirect influence on associative evaluations mediated by a direct influence on propositional reasoning, leading to corresponding changes in explicit and implicit attitudes. Solid arrows reflect a causal influence, whereas open arrows reflect a lack of influence. 706 GAWRONSKI AND BODENHAUSEN