2007 social cognition mcm

8/8/2019 2007 Social Cognition MCM

1/30


2/30

one common thread across the years is that attitudes were assumed tohave an evaluative component. Indeed, today attitudes commonly referto peoples evaluations of a wide variety of objects, issues, and people, in-cluding the self. Our approach is compatible with this view.

Our primary goal in this article is to present our MetaCognitiveModel (MCM) of attitudes(see also Petty, 2006; Petty& Briol, 2006).Be-fore describing the MCM in some detail, it is useful to outline brieflythreealternative views of attitudesone that is nowclassic andtwooth-ers that are of more recent vintage. After describing these alternativesand the MCM, we compare our approach to the alternatives. Then, wearticulate what the MCMcancontribute to understanding what deliber-ative andautomatic attitude measures reflect, what it means when thesemeasures conflict, what happens when attitudes change, how the MCM

relates to the underlying strength of attitudes, and some other uniqueperspectives offered by the MCM.1

THREE PERSPECTIVES ON ATTITUDE REPRESENTATION

There is one model of attitude representation that has dominatedthe lit-erature for some time, and two that have gained adherents more re-cently. We describe each approach briefly so that we can compare themto our metacognitive framework.

SINGLE ATTITUDE MODEL

The first, and now classic approach to attitude representation, assertsthat attitudes are best conceptualized as an objectevaluation link in

658 PETTY ET AL.

1. We use the terms deliberative and automatic attitudes and attitude measures ratherthan implicit/explicit attitudes/measures because of the ambiguity that surrounds the useof the latter terms in the literature (Petty, Tormala, Briol, & Jarvis, 2006). In general, re-searchers use the term implicit attitude to refer to what is assessed by measures that tapinto relatively quick (even automatic) evaluative associations (i.e., implicit measures),whereas the term explicit attitude refers to what is assessed with more deliberative andundisguised measures (explicit measures). Sometimes, however, researchers use the termimplicittorefertoanattitude(ormeasure)ofwhichpeopleareunaware,ortoabasisoftheattitude of which people are unaware, or to an effect of an attitude of which people are un-aware (see also Fazio & Olson, 2003; Petty, Wheeler, & Tormala, 2003). To allow for moreprecision, we willusetheterms automaticand deliberativeattitudes that are tappedbyautomatic and deliberative measures (though neither measure is process pure, e.g., seeSherman, in press). Inearlierliterature,the termopinionwasoften usedto referto thever-

bal expression of an implicit (internal) attitude (e.g., see Hovland, Janis, & Kelley, 1953;Thurstone, 1928).


3/30

memory (e.g., Fazio, 1995; Fiske & Pavelchak, 1986; see also Fazio, this

issue). This approach, which we call the single attitude model, is de-picted in the top panel of Figure 1 and shows a person who has a posi-tive attitude toward smoking. The most well-articulated andinfluential example of this conceptualization is Fazios MODE Model(Fazio & TowlesSchwen, 1999), which has contributed much to ourunderstanding of attitude structure and expression. In brief, theMODE Model holds that people have stored evaluative associations toattitude objects. In this framework, automatic measures of attitudes(e.g., evaluative priming; Fazio, Jackson, Dunton & Williams, 1995;and the Implicit Association Test [IAT]; Greenwald, McGhee &Schwarz, 1998) tend to assess the stored evaluative association,whereas more deliberative measures (e.g., semantic differential;Osgood, Suci, & Tannenbaum, 1957) tap the retrieved evaluative asso-ciation along with the outcome of any downstream cognitive pro-cesses. Thus, if a person expresses a different attitude on a deliberativethan an automatic measure, it is presumably because he or she has en-gaged in some thought that modifies the initial automatic evaluative

META-COGNITIVE MODEL OF ATTITUDES 659

GOOD

BAD

NO

Smoking

GOODSmoking

BADSmoking

GOODSmoking

Single Attitude Model

Dual Attitudes Model

Meta-cognitive Model

YES

FIGURE 1. Structural Representation of Three Attitude Models.


4/30

reaction that comes to mind (see also, Gawronski & Bodenhausen,2006, for similar assumptions). This thought can reflect additionalmentalcontents that arebrought to mind or activatedby thecontext, orit can stem from impression management or correction motives (seeFazio, this issue).

DUAL ATTITUDES MODEL

A second approach that has captured the attention of social psycholo-gists more recently argues that people can hold separate explicit (con-scious, deliberative) and implicit (unconscious, automatic) attitudes(e.g., Greenwald & Banaji, 1995; Wilson, Lindsey, & Schooler, 2000)

which can take on different values. Although there are various assump-tions that have been made about dual attitudesby the different theoristswho advocate this view, three assumptions are fairly common.

Perhaps the most important assumption shared by many dual atti-tudes proponents is that the dual attitudes (implicit and explicit) haveseparate mental representations (e.g., see Wilson et al., 2000). As de-picted in the middle panel of Figure 1, implicit and explicit attitudes areviewed as distinct mental entities that are stored separately in differentareas of the brain (e.g., see DeCoster, Banner, Smith, & Semin, 2006).Thus, a person might have a deliberative (explicit) attitude toward anobject of one valence but an automatic (implicit) attitude of a differentvalence.2

A second common assumption is that the two attitudes stem from dis-tinct mental processes. Implicit attitudes are said to stem from associa-tive processes such as evaluative conditioning, whereas explicitattitudesstem from propositional processessuch as thinking aboutmes-sage arguments (e.g., Rydell, McConnell, Mackie, & Strain, 2006). Thedifferentmental processes responsible for explicit and implicitattitudesare often characterized as being governed by separate mental systems(e.g., reflective/impulsive, fast/slow; DeCoster et al., 2006; Rydell &McConnell, 2006).

Third, implicit and explicit attitudes are postulated to be relatively in-dependent and to operate in different situations. Thus, these attitudesare not expected to be in conflict (DeCoster et al., 2006), but instead are

660 PETTY ET AL.

2. In the single attitude approach, deliberative and automatic measures canshow differ-entattitudes,but it is notassumed thatthere areseparate mental representations. Rather, adifference in the measures can occur when a deliberative attitude is constructed onlinefrom considerationof the automaticevaluation along with other thingsthat come to mind(Fazio & Towles-Schwen, 1999).


5/30

postulated to work in different arenas. In particular, implicit attitudesarepostulated to guide behavior inspontaneous situations when peopleare not engaged in much thought, whereas explicit attitudes are said toguide behavior when people are being reflective (see Dovidio,Kawakami, Johnson, Johnson, & Howard, 1997, Davidio, Kawakami,Smoak, & Gaertner, in press). In this sense, then, the manner of measur-ing implicit and explicit attitudes (i.e., with automatic versus delibera-tive measures) matches the situations in which they guide behavior(Vargas, 2004).

When considering all of these assumptions together, the dual atti-tudes framework suggests that attitudes assessed with automatic anddeliberative measures are quite different. These attitudes have separaterepresentations, are the resultof different processes from separate men-

talsystems, andoperate in differentsituations. There is notmuch, if any,interaction between them. Indeed, they have been characterized as twoships passing in the night (Cohen & Reed, 2006, p. 9).

PROCESS MODELS OF ATTITUDES

One additional perspective thattakes a radically differentapproach to at-titudes holds that there are no stored evaluations in memory. Instead, ac-cording to this constructivist perspective, attitudes are formed, as needed,

basedoncurrentlysalient beliefs, feelings,andbehaviors (e.g.,Schwarz&Bohner, 2001; Wilson & Hodges, 1992; see also Schwarz, this issue). Ac-

cording to this approach, different contexts will make different knowl-edge accessible, resulting in changes in peoples reported evaluations.Any consistency in attitudinal reports, according to this perspective, co-mes from the same set of building blocks being retrieved each time and

being usedincomputing the current evaluation.In our view, it isnotclearwhy a mental system would develop such that itwould linkparticular at-tributes to evaluations (e.g., being lazy is bad) but notlink the overall atti-tude object to an evaluation (e.g., Fred is bad; see Petty et al., 2003).Indeed, researchsuggests thatpeople often do notretrieve attribute infor-mation when they have a previously formed and relevant attitude toguide decisions (see Lingle & Ostrom, 1981; Lichtenstein & Srull, 1985).

One recent instantiation of a process approach to attitudes is the APE(Associative Propositional Evaluation) model (Gawronski &

Bodenhausen, 2006). This framework holds that people can respondpositively or negatively to some attitude objectbased solelyon theaffectthat is associated with the object, or based on the propositions that cometo mind with respect to theobject. Affectassociated with an object canbetranslated into propositional form (e.g., I like this) and then checked for



6/30

validity by an online process that examines whether the evaluativeproposition is consistent with other salient propositions. In this frame-work, there are no stored evaluations (attitudes) per se, only stored af-fect and beliefs (propositions) that serve as input to an expressedevaluation (see Gawronski & Bodenhausen, this issue).

THE METACOGNITIVE MODEL (MCM)

Our MetaCognitive Model (MCM) shares some features with each ofthe approaches just described, but also has notable differences. In brief,the MCM holds that attitude objects can be linked in memory to bothpositive and negative evaluations that can vary in the degree to which

they are endorsed or not (see bottom panel of Figure 1). The MCM isbased on a number of assumptions that we outline next.

First, in accord with the single attitude approach and in contrast to theconstructivist view, theMCM holds that attitude objects canbe linkedinmemory to global evaluative associations, and these associations canvary in strength (e.g., Fazio & TowlesSchwen, 1999; see also Fazio, thisissue). Also parallel to the single attitude approach, features of the con-text candeterminewhether an activatedevaluation is used. Thus, just asone can store ones birthdate in long-term memory and retrieve it whenneeded to construct how old one is currently (i.e., by considering ones

birthdate along with todays date),peoplecan retrieve evaluations asso-ciated with attitude objects and modify them according to the current

situation (e.g., I generally like cake, but this one smells odd, so I wonthave any).Second, in concert with the dual attitudes approach and recent re-

search and theory on the separability of positivity and negativity (e.g.,Cacioppo, Gardner, & Berntson, 1997), the MCM holds that attitude ob-

jects can sometimes be linked in memory to evaluative associations ofopposite valence (see also, deLiver, van der Pligt, & Wigboldus, 2007).Whether a positive or negative evaluation comes to mind first will de-pend onall of the various factors thatcan affect memory.Thus, the num-

ber of prior positive and negative experiences, the recency of thoseexperiences, and the context in which those experiences took place willmatter. For example, if negative experiences with some object wereformed largely at work but positive associations were formed at home,

then association measures given in these different contexts should re-veal different responses. Similarly, if people have experienced positivereactions to African Americans in a sports context, but negative reac-tions in an urban setting, measures of association that include these con-textual features should show different evaluations (Barden, Maddux,

662 PETTY ET AL.


7/30

Petty, & Brewer, 2004; Wittenbrink, Judd, & Park, 2001). It is importantto note that not all attitude objects are expected to be associated with

both positive and negative associations. Indeed, when an attitude objectis linkedto only one evaluative association, an attitude structure similarto the singleattitude model emerges.

Third, in contrast with a common assumption of dual attitudes mod-els, we do not assume that opposite evaluative associations, if present,necessarily stem from differentunderlying processesthat is, both pos-itive and negative evaluations can stem from associative or from propo-sitional processes. Thus, although it is possible to have one valencedassociation that stems from propositional processes while the otherstems from associative processesas specifiedby some dual attitudesthe-orists, it is also possible for both valenced associations to be based onsimple associative processes or more reflective thought processes orsome combination of each.

Fourth, the feature of the MCM that gives the model its name is the as-sumption that people can tag their evaluative associations as true orfalse, or that they are held with varying degrees of confidence. In thisway, the model builds on existing research on metacognition (Jost,Kruglanski, & Nelson, 1998). Metacognition refers to thoughts aboutthoughts or thought processes. Research in this domain has demon-strated that a person can have an initial or primary thought (e.g., I likecandy) that is further reflected upon by a metacognitive or secondarythought (e.g., I am confident that I like candy; see Petty, Briol,Tormala, & Wegener, 2007). The metacognitive associations in the

MCM can be represented in various ways such as yes/no, confi-dence/doubt, true/false, accept/reject, and so forth.3 Furthermore,these metacognitions can vary in the strength of their association to thelinkedevaluation, andthestrength of this association will determinethelikelihood that the perceived validity of an evaluation will be retrievedalong with the evaluation itself.

Thus, in concert with the APE model (Gawronski & Bodenhausen,2006) and other conceptualizations (e.g., Cohen & Reed, 2006;Kruglanski, 1989; Petty, Briol, & Tormala, 2002), the MCMputs a focuson validity processes. Unlike the APE model, however, we do not as-sume that validity stems from consistencyanalysesalone,but ratherthat


3. Affective validation is also possible wherein peoples attitudes make them happy orsad, comforted or anxious. Although we focus on validity tags in this model, we also ac-knowledge that other tags might exist, andthese tags might alsoexert an impact on attitudi-nalprocesses.For example,a personmighttaga negative racialevaluationas inappropriateto express even though he or she might personally endorse the association.


8/30

confidence can be inferred directly (e.g., from ease of attitude retrieval;seeHolland,Verplanken,& vanKnippenberg,2003), or resultfrom vari-ousinferences basedon theperceivedcontent of onesattitude (e.g.,neg-atively framed attitudes are held with more confidence than positiveones; Bizer & Petty, 2005) or the process by which an attitude wasformed, maintained, or changed (e.g., having difficulty defending onesattitude can lead to less confidence; Petty, Tormala, & Rucker, 2004),among other factors. Most notably, however, the MCMgoes beyond theidea that validation is solely an online process and holds that perceivedvalidities, like the evaluations themselves, can be stored for later re-trieval. In other words, the MCM assumes that just as it is adaptive tostore evaluations to guide decision making and action (Fazio, 1995), sotoo is it adaptive to know if any activated evaluation is a reliable guide.

Thestorage of validity information canpresumably make decision mak-ing and action more efficient. To the extent that the retrieval of validitytags becomes automatic, it even becomes possible for people to quicklycorrect for undesired evaluations that might come to mind. Further-more, just as evaluative associations can be context specific, so too canpeople learn to associate invalidity tags with evaluations more in somecontexts than others (e.g.,see Maddux, Barden, Brewer, & Petty,2005).

Although there is no definitive research on thestorage of validity tags,some evidence consistent with this idea comes from research on the sta-

bility of belief certainty over time. For example, in one study, a manipu-lation of expressed agreement with ones judgment by another person(i.e., social consensus) produced the same increase in judgmental confi-

dence whether that confidence was measured immediately or 48 hourslater (Wells, Olson, & Charman, 2003). In other research, attitude cer-tainty measured at one point in time was shown to predict certaintyre-lated outcomes (i.e., resistance of attitudes to change) at a later time,even when certainty was not made salient at the second occasion (e.g.,Bassili, 1996, study 2).Theseresults areconsistent with theidea that con-fidence can reside in memory. The MCM does not specify exactly howevaluations and validities are stored in memory, however. In otherwords, memory for evaluations and their validities can be conceptual-ized as part of a traditional semantic association network (e.g., Fiske &Pavelchak, 1986) or as linked patterns of activation in a connectionistmodel (e.g., Eiser, Fazio, Stafford, & Prescott, 2003; see also Conrey &Smith, this issue). Either framework can accommodate the postulated

linkages in the MCM (e.g., see Van Overwalle & Siebler, 2005 for aconnectionist model wherein attitude objects are linked to bothpositivity and negativity).

Finally, theMCMconcurs with researchon cognitivenegation thatsug-gests that untagged evaluations are presumed to be true unless evidence

664 PETTY ET AL.


9/30

against them is or has been generated (e.g., see Gilbert, 1991). Further-more, researchonnegationsuggests thatsuccessful negation isquitediffi-cult (e.g., Deutsch, Gawronski, & Strack, 2006). Indeed, overriding onesnegated attitudes will require motivation and ability, at least in the earlystages (Betsch, Haberstroh, Molter, & Glckner, 2004). People could at-tempt to invalidate their previously formed evaluative associations formanyreasons. For example, peoplemight rejectan evaluative association

because they realize that it stems from the culture (e.g., media exposure)andnot frompersonal beliefs (e.g.,Devine, 1989). In addition, theassocia-tion can represent the opinions of others that have been encoded (e.g.,Han, Olson, & Fazio, 2006; Priester & Petty, 2001). Also, the associationcanrepresenta previously accepted personal view that has morerecently

been discredited (e.g., Gregg, Seibt, & Banaji, 2006; Petty et al., 2006).4

When theassociationand negationarepresentedat thesame point intime(e.g., Johnis notclean), peoplecansometimes reverse the association(i.e.,not clean becomes messy; see Mayo, Schul, & Burnstein, 2004), butwhen the negation follows the association in time (e.g., John is clean. . . .False), this is less likely.

We believe that the MCM provides a more complete, integrative, andflexible view of a persons underlying attitude structure than alterna-tives, and provides an explanation for various attitudinal phenomenasuch as how one underlying attitude structure can lead to different out-comes on deliberative versus automatic measures. The MCM also hassome unique implications. We discuss these issues next.

UNDERSTANDING AUTOMATIC VERSUSDELIBERATIVE ATTITUDE ASSESSMENTS

Much attention has been paid recently to discrepancies that can emergebetween attitudes assessed with deliberative versus automatic mea-sures. These differences have emerged both when an attempt is simplymade to assess a persons existing attitudes(e.g., the personscores posi-tive toward minorities on a deliberative measure but negative on an au-tomatic measure; see Greenwald et al., 1998), and when one attempts toassesstheeffectiveness of anattitude changemanipulation (e.g.,the per-son shows attitude change from the perspective of one type of measure

but notthe other;see Gawronski& Bodenhausen,2006). We discuss eachof these kinds of discrepancies in turn.


4. When the negated evaluation is a prior attitude,we have referred to our approach asthe PAST (Prior Attitudes are Still There) Model (Petty & Jarvis, 1998; Petty et al., 2006).That is, the PAST Model is a specialized case of the more general MCM.


10/30

DIVERGENCE IN AUTOMATIC VERSUSDELIBERATIVE ATTITUDE MEASURES

Imagine a situationin whicha personsattitude shows a positive evalua-tion of some attitude object on a measure of automatic evaluation, but anegative evaluation on a deliberative measure. According to the singleattitude model, the underlying structure is best captured by the toppanel of Figure 1that is, at the association level, the attitude object islinked to a positive evaluation and this shows up on the automatic mea-sure. If a negative evaluation is expressed on a deliberativemeasure, ac-cording to this approach, it must be due to downstream cognitiveprocesses. Forexample, theperson couldbring to mind various negative

attributes of the object that override the positive association, or it couldbe that a persons egalitarian motives or desires to be unprejudiced leadto a correction of thenegative underlying attitude (e.g., Dunton& Fazio,1997;Plant& Devine,1998). Accordingto thedualattitudesperspective,it is possible that the person has distinct positive and negative evalua-tions residing in separate memory systems (Figure 1, middle panel) thatare called forth on the different types of measures. The constructivistperspectivedoesnotallow forstoredgeneral evaluations andthus holdsthat different material from memory is retrieved and integrated onautomatic versus deliberative measures.

Although the MCM accepts that the processes outlined by the singleand constructivist positions can sometimes occur (i.e., when only oneevaluative association, or none, is stored), it also points to another possi-

bilitythat is, one possible attitude structure leading to divergence inautomatic versus deliberative measures is represented in the bottompanelof Figure1. TheMCM assumes that contemporarymeasuresof au-tomatic evaluation tap (although not perfectly) into evaluative associa-tions largely without respect to validity tags. Thus, if such a measurereveals an overall positive attitude, the MCM holds that this is becausepositive associations are stronger than negative ones without respect tothevalidity tags that might be attached to those evaluations. Thereverseis the case if such measures reveal an overall negative attitude. Explicitmeasures, in comparison, reflect evaluative associations as modified bystored validity tags as well as other considerations that come to mindprior to responding.

So, in the evaluative structure depicted in the bottom panel of Figure 1,whichattitudewouldbeassessedwith each measure? Anautomatic mea-sure would reveal a somewhat positive attitude toward theobject.This is

because although the object is associated with both good and bad, thestrength of association to good is stronger than to bad. It does not matter

666 PETTY ET AL.


11/30


12/30

than automatic ones. Imagine a person who has favored smoking for along time andwhoseattitude canbe conceptualizedas inthetop panelofFigure 2. This person begins with both deliberative and automatic atti-tudes in synchboth are positive. Then, this person begins to receivenumerous antismoking messages that are convincing. The person nowdevelops an ambivalent attitude in that both positive andnegative asso-ciations are present, and endorsed, as depicted in the second panel ofFigure 2. When measured with a bipolar deliberative measure(goodbad), the person would show a neutral or slightly positive atti-tude that represents theintegrationof theseparatepositiveand negativereactions. However, if desired, the separate positive and negative reac-tions could be assessed with deliberative unipolar measures that askedabout positivity and negativity separately (see Kaplan, 1972). Like the

bipolar deliberative measure, in the realm of automatic attitude assess-ments, the ambivalent person would come out neutral or slightly posi-tive on a measure that collapsed across positivity and negativity (e.g.,IAT; Greenwald et al., 1998). However, if positivity and negativity wereassessed separately, one could see that for the ambivalent person, thereare automatic associations to both positivity and negativity (e.g., deLiver et al., 2007; NewbyClark, McGregor, & Zanna, 2002)

So, at the point of receiving and accepting negative information aboutsmoking, the automatic and deliberative measures are again showingcomparable effects. Over time, however, the person may come to com-pletely reject his or her initial positive evaluation of smoking and pos-sess the attitude structure shown in Panel 3 of Figure 2. Once the person

rejects the initial attitude, at this point the deliberative measure willlikely show more change from its initial value than the automatic mea-sure. Because the automatic measure does not consider the validity ofthe associations (unless the negation itself becomes automatic), it willshow thesame pattern of mixed reactions as before, whereas theexplicitmeasure will now indicate a negative reaction toward smoking.

Evidence fora pattern of greater changein a deliberativethan an auto-matic measure was obtained in a study by Petty et al. (2006, study 1). Inthis study, participants first formed an initial positive or negative atti-tude toward another person via evaluative conditioning. This manipu-lation was effective in modifying both automatic (evaluative priming)anddeliberative (semanticdifferential)measuresof attitudes.Next, par-ticipantsreceivedinformationabout theopinionsof thetarget personon

several important issues that would make theperson appeareitherverylikable (i.e., had similar attitudes to the participant) or dislikable (i.e.,haddissimilar attitudesto theparticipant; seeByrne, 1961). Insomecon-ditions, this information reinforced the initial impression (i.e., no atti-tude change) and in other conditions this information contradicted the

668 PETTY ET AL.


13/30


initial impression(i.e., attitude change). In thereinforcement conditions,both the deliberative and automatic measures showed the same patternof results (i.e., more positive attitudes toward the similar person whowasconditionedpositively than tothedissimilar personwho wascondi-tioned negatively). However, in the contradiction (attitude change) con-ditions, the measures diverged such that attitudes were more sensitiveto thecontradictorysimilarity information about the targeton the delib-erative than on the automatic measure. In this research, the deliberativemeasure reflected thefact that the oldattitude wasrejected,whereas theautomatic measure reflected fast association of the target personto both

the old and the new evaluations. This state of affairs represents whatmight be calledthenormal attitude change situationin which peoplere-

ject their previous attitude and accept a new one. The explicit measuretracks this change quite well, but the implicit measure lags behind

4. Implicit Ambivalence (b)

1. Univalence

GOOD YESSmoking

2. Explicit Ambivalence

GOOD YES

YES

SmokingSmoking

GOOD

BAD

GOOD YES

NO

SmokingSmoking

GOOD

BAD

GOOD NO

YES

SmokingSmoking

GOOD

BAD

3. Implicit Ambivalence (a)

FIGURE 2. Depiction of Univalence, Explicit Ambivalence, and Implicit Ambivalence from

the Perspective of the Meta-Cognitive Model.


14/30

because of its relative insensitivity to the negation (see also Gregg et al.,2006).

More Change on Automatic than Deliberative Measures. Notably, accord-ing tothe MCM,insome attitudechange situationsit ispossible for anim-plicit (automatic) attitude measure to tap change better than an explicit(deliberative) measure. Imagine our smoker again (Panel 1 of Figure 2).This time the smoker is exposed to an advertising campaign that featuresdisgusting photos of black lungs, cancerous growths, and other negativeconsequences of smoking. The smoker starts to notice an automatic nega-tive feeling whenever opening a cigarette pack. Then, a friend points outthe manipulative advertising campaign and the smoker recognizes thatthis is likely the cause of the negative feeling. As a consequence, he rejectsthe campaign andthe attitude it implies. In fact, thesmoker resentsbeing

targeted by the messages. If an explicit attitude measure toward smokingis administered at this point, it will likely show little changethe personstill reports being quite positive toward smoking. Yet, an implicit mea-sure would likely show a less positive reaction than previously due to thenew negative associations. The structure of this persons attitudeaccording to the MCM is presented in Panel 4 of Figure 2.

This type of attitude change situation is a bit more unusual than theprevious one, buttheliterature on attitude changehas suggesteda num-

ber of situations inwhichitcanarise. Perhaps themost obvious situationthat contains this possibility is work on thesleepereffect(see Kumkale &Albaracn, 2004 for a review). In this paradigm, people are exposed to avery strong persuasive message (counter to a persons initial attitude),

but this strongmessageis then discounted (e.g., claimed to be false or at-tributedto a lowcrediblesource).The presenceof a newevaluativeasso-ciation (from the strong message) should produce change in theautomatic measure, but the negation of the new association (from thediscounting cue) should leave the deliberative measure unchanged.Note that in this situation, it is the automatic measure that taps the new(morerecent) attitude andthe deliberativemeasure that taps theold (ini-tial) attitude. Thus, the MCM does not predict that automatic measuresinvariably tap old attitudes, whereas deliberative measures tap morerecent ones. Rather, it is the strength of the evaluative association thatmatters along with validity tags. When thenormalattitudechangepara-digm is modified so that a strong new attitude is first created and thennegated, the automatic measure reflects the more recent attitude,

whereas the deliberative measure reflects the older attitude (since thenew one is negated).

In this situation, although the deliberative measure does not showchange initially, over time change can emerge (i.e., the sleeper effect).Viewed from the perspective of the MCM, what happens over time is

670 PETTY ET AL.


15/30

that the negation link is weakened (forgotten or dissociated) so that theevaluative association formed at the time of message exposure can havean impacton thedeliberativemeasure. If thenegation link is strongly at-tachedto theevaluative association (e.g., because this link wasactivatedfrequently or with a great deal of thought), however, the sleeper effectmaynotemerge. Indeed, although ithasneverbeen tested, theMCM an-ticipates that in a sleeper effect paradigm, those who show the largestdiscrepancy between automatic and deliberative measures right afterthe message (i.e., large change on automatic and small change on delib-erative) would show the largest sleeper effect. Another paradigm inwhich peoplemay consciously negate an evaluative association that hasformed (allowing for delayed influence on the deliberative measure) isthat of minority influence (see Tormala, DeSensi, & Petty, 2007).

Different Impact on Automatic and Deliberative Measures. As our finalsituation, we turn to research in which one manipulation affects auto-matic and deliberative measures differently. Consider a study byDeCoster et al. (2006) in which participants were presented with photo-graphs of various target individuals accompanied by an evaluative as-sertion regardingthetarget(e.g., Samis smart)or a negation (e.g., Jack isnot smart). Coming from a dual attitudes perspective, the authors ar-gued that the fast learning (propositional) system would be capable ofhandling negations andthus would be able to differentiate thetwoindi-viduals. The slow learning (associative) system, on the other hand,would not be able to process negations and thus would associate bothSam and Jack with intelligence. With respect to automatic versus delib-

erative measures, the dual systems approach predicts that when pre-sented with a negation, an automatic measure would assess Jack asintelligent, buta deliberativemeasure would assess himas notsmart (oras stupid). Because the automatic and deliberative measures used intheir research indicated opposite evaluative associations to Jack, theauthors reasoned that negationscause people to form dual attitudesthatare stored in separate memory systems.

The MCM explains these data differentlythat is, rather than havingtwo representations of Jack as both smart and stupid stored in separatesystems as depicted in the top panel of Figure 3, the MCM holds that theinformation is encoded as depicted in the bottom panel of Figure 3.Therefore, Jack is associated with intelligence, but this is negated. Forthe reasons already outlined, this cognitive structure would lead to the

prediction that an automatic measure would assess Jack as intelligentbecause this is directly linked to Jack, but motivation and ability areneeded to retrieve the negation. Thus, a deliberative measure would as-sess Jack as stupid because on the more thoughtful measure, it is morelikely that the negation would be retrieved.



16/30

ADDITIONAL IMPLICATIONS

OF THE METACOGNITIVE MODEL

It is important for the MCM to be able to account for phenomena thatexist and that have been explained by other theories. If the MCMcould not account for phenomena such as those just reviewed, itwould not be very useful. However, the MCM goes beyond providingan alternative account for empirical findings that other theories canalso explain. It also makes unique predictions and we turn to thesenext.

Earlier in this article we explained how the MCM could account fordiscrepancies between automatic and deliberative measures of attitudesas well as deliberative measures changing more to an influence attemptthan automatic measures and vice versa. The MCMmakes a unique pre-

diction about both of these situations that we consider next. After this,we consider what the MCM says about the strength of deliberative atti-tude reports and their correlation with automatic measures, and wepresent some new insights that the MCM can offer research onprejudice.

672 PETTY ET AL.

Meta-cognitive Model (MCM)

Not smart/

stupid

Associative

System

Propositional

System

Jack NO

Jack

Jack

Dual Representation Model

Smart

Smart

Not

FIGURE 3. Representation of Jack Is Not Smart from the Dual Systems Approach and theMCM.


17/30

AUTOMATICDELIBERATIVE DISCREPANCIES

The MCM notes that automaticdeliberative discrepancies can occurwhen peoplehave both positive and negative associations to an attitudeobject, and one of these associations is rejected. Although people withthis attitude structure (see the bottom panel of Figure 1) would experi-ence no explicit ambivalence because they do not endorse both positiveand negative aspects of the attitude object, there would still be ambiva-lence at the level of evaluative associationsan ambivalence that wehave referred to as implicit ambivalence (Petty et al., 2006). This ambiva-lence is implicit in that it is at the level of automatic associations and thepersondoes not experience it consciously as ambivalencethat is, there

is no explicit conflict because one of the valences is explicitly rejected. Ifthere is some implicit ambivalence, then peoplemight be expected to at-tempt to resolve it because evaluative conflict (even if just at theassociative level) is typically unpleasant.

As explained earlier, according to some versions of the dual attitudesapproach, there shouldnot beanyconflictwhenautomaticand delibera-tive measures diverge because the implicit and explicit attitudes are as-sumed to have separate lives. Either one or the other will guideresponses depending on whether the situation is a deliberative or aspontaneous one (Dovidio et al., in press). The two evaluations shouldnotbe jointly activatedin anygivensituation, whichis a requirement forambivalence (NewbyClarket al., 2002). Accordingto thedual attitudesapproach, the two attitudes are the result of separate mental systemsand as DeCoster and colleagues (2006) noted, the fact that the two sys-tems store their representations in separate areas of thebrain means thatany inconsistencies between them do not have to be resolved (p. 9).However, in accordwith theMCM, wehave obtainedevidence that peo-ple act as if automaticdeliberative divergence produces a state of(implicit) ambivalence.

In a series of studies we have shown that discrepancies between auto-matic and deliberative measures of attitudes can lead to some of thesame consequences as explicit endorsement of positive and negative at-tributes of some attitude issue or object. One wellknown consequenceof explicit attitudinal ambivalence is that it leads to enhanced informa-tion processing in a presumed attempt to resolve the ambivalence. In

one study, for example, Maio, Bell, and Esses (1996) measured partici-pants explicit ambivalence regarding the issue of immigration to Can-ada (i.e., the extent to which they endorsed both positive and negativereactions to the issue), and thenexposed themto a message favoring im-migration from Hong Kong to Canada that contained either strong or



18/30

weak arguments. The degree to which participants processed the mes-sage information was assessed by examining the extent to which thequality of the arguments affected postmessage attitudes toward immi-gration (Petty, Wells, & Brock, 1976). When people are thinking care-fully about information, they should be affected by the quality of thearguments that a message contains (see Petty & Cacioppo, 1986). As hy-pothesized, Maio et al. (1996) found that individuals who had explicitlyambivalent attitudes toward immigration were more influenced byargumentquality thanwereindividuals lowin ambivalence, suggestingthat they engaged in enhanced scrutiny of the information.

In one study testing the notion that automaticdeliberative attitudediscrepancy could lead to enhanced information processing (Briol,

Petty, & Wheeler, 2006), we assessed undergraduates selfevaluationwith both automatic (IAT; Greenwald et al., 1998) and deliberative(Rosenberg, 1965) measures and then calculated the absolute value ofthe difference between the two standardized measures as our index ofdiscrepancy (see also Kehr, 2004). Next, participants were exposed to ei-ther a strong or weak message about eating vegetables that was framedas selfrelevant (i.e., relevant to ones personal lives andthus relevant tothe discrepancy) or not. The results are displayed in Figure 4. When themessage was framed as selfrelevant, the extent of automaticdelibera-tive discrepancy interacted with argument quality to affect attitudes.The greater the automaticdeliberative discrepancy, the more partici-pantsdifferentiatedstrongfromweakarguments (toppanel). However,when the same strong and weak messages were framed as irrelevant to

the self (i.e., the message was said to be about the properties of vegeta-bles), discrepancy did not interact with argument quality to predict atti-tudes (bottom panel). This suggests that automaticdeliberativediscrepancies do notlead to motivation to process all informationonlythat relevant to the discrepancy.

Finally, we have also attempted to provide more direct evidence forthe ideathat automaticdeliberative discrepancies produce implicit am-

bivalence (Briol, Petty, & Wheeler, 2003). Specifically, in one study wefound that as discrepancy in automatic versus deliberative selfesteem(as assessed using the absolute value of the difference between partici-pants standardized automatic and deliberative selfesteem scores) in-creased, the strength with which participants automatically associated

doubtwords with selfwords on an IAT also increased. However, in-creased discrepancy was not associated with explicit reports of selfun-certainty. This suggests that the selfdoubt that accompaniesautomaticdeliberative discrepancy is either not open to consciousawareness, or is explicitly denied or discounted.

674 PETTY ET AL.


19/30

ATTITUDE CHANGE

As noted in our earlier discussion of persuasion, the MCM holds that insituations of changing attitudesfrom one valence to another, there is thepotentialfor implicitambivalence. Thisis becausewhenthevaliditytagsare ignored, both positive and negative evaluations are associated withthe attitude object (see Panels 3 and 4 in Figure 2). Is there any evidence

that attitude change situations can produce implicit ambivalence in theabsence of explicit ambivalence? In recent research we have shown thatsuch situations can lead to the same information processing conse-quences as having discrepant automatic and deliberative attitudes.

Inonestudy examining this notion (Pettyetal.,2006, Experiment 2), us-


4

5

6

7

8

Strong

message

Weak

message

Discrepancy-Unrelated Frame

A

t

t

i

t

u

d

e

4

5

6

7

8 Strong

message

Weak

message

Discrepancy-Related Frame

Low High

Automatic-Deliberative Discrepancy

A

t

t

i

t

u

d

e

Low High

Automatic-Deliberative DiscrepancyFIGURE 4. Interaction of Frame, Implicit-Explicit Self-esteem Discrepancy, and ArgumentQuality on Attitudes (Data from Briol, Petty, & Wheeler, 2006, Experiment 4).


20/30

ing a procedure described earlier, undergraduate students were initiallyconditioned tolike or dislikea target individual.Then, theparticipantsre-ceived information about the target individuals attitudes on several im-portant topics. The attitudinal information was designed to either get theperson to like or dislike the target by having the target agree or disagreewith the participant on several important issues. In some conditions, thisinformation was in the same direction as the conditioning manipulationso that no attitude change would occur, and in other conditions the infor-mation wasopposite in valence to theconditioning. In thelattersituation,individuals rejected their earlier evaluations based on conditioning andadopted new evaluations based on the similarity information. Accordingto the MCM, this confluence of factors produces a situation in which de-

liberative measures would be more sensitive to the similarity inductionthan would automatic measures (which will still show some evidence ofthe old and now rejected evaluation). However, rather than measuringautomatic and deliberative attitudes following attitude change as de-scribed earlier (Petty et al., 2006, Experiment 1), in this study participantswere told that the target person was a candidate for a job at their univer-sity. To evaluate thecandidate, they were provided with eithera strong ora weak rsum to examine. The key result was that attitudes toward thetarget asa jobcandidateweremore influencedbythe quality of the candi-dates rsum in the condition where attitudes were changed than incon-ditions where attitudes were notchanged. In other words, when attitudeswere changed, people engaged in greater information processing as ifthey were attempting to resolve some ambivalence.

In other studies we have shown that when explicit attitudes changefrom one valence to another, people do not necessarily report feelingmoreambivalent,nordotheyindicatethattheyendorsebothsidesoftheissue. Nevertheless, when explicit attitudes have changed in valence,people still act as if they are ambivalent by processing information ontheissueto a greater extentthan when attitudesarenotchanged. To pro-vide more direct evidence regarding implicit ambivalence, in anotherstudy (Pettyet al., 2006, Study 3) we changed thevalence of participantsattitudes about a target individual or not, and then gave them an IAT tosee if the target individual was more associated with doubt than whenattitudes were not changed. We also measured explicit reports of con-flict in this study. Consistent with the idea that an explicit change in atti-

tudinal valence can produce implicit ambivalence (due to conflictbetweenoldrejectedevaluations andnewly acceptedones), participantswhose attitudes were changed did not report any more explicit doubtabout the target individual, but they did show more doubt on an IATcompared to when attitudes were not changed.

676 PETTY ET AL.


21/30

ATTITUDE STRENGTH

Another important feature of the MCM is the manner in which it ac-counts for the strength of explicit attitudes. Considerable research overthe past few decades has shown that all attitudes of the same valencewhen assessed with common explicit measures are not the same instrength, where strength is defined by the extent to which the attitudesare durable (persist over time and resist change) and impactful (affectother judgments and behavior; Krosnick & Petty, 1995). There are twoaspects of the MCM that relate to the strength of explicit attitudes. Thefirst is the strength of the objectevaluation link(s) that can be approxi-mated with measures of evaluation accessibility (see Fazio, 1995)that

is, the more accessible an objectevaluation link is, the more it will cometo mind and thereby have the potential to be tapped by a deliberative at-titude measure. The second is the validity tag that can be approximatedwith measures of evaluation confidence or certainty (see Gross et al.,1995). The more confident that people are in the validity of an ob-

jectevaluation link (depicted as a yes tag in Figures 1 and 2) and themore accessible this confidence is, the more they are likely to report theattitude on a deliberative measure, and the more likely this attitude is to

be durable and impactful. According to the MCM, then, anything thatincreases the accessibility of evaluations or the confidence that one hasin them will increase attitude strength.5

One of the most studied determinants of attitude strength empha-sized by prominent attitude change theories such as the Elaboration

Likelihood Model (Petty & Cacioppo, 1986), the HeuristicSystematicModel (Chaiken, Liberman, & Eagly, 1989) and the Unimodel(Kruglanski & Thompson, 1999), is theextentof elaboration that theatti-tude object has received (i.e., how much thinking a person has doneabout it). We noted earlier that if people are doing online (downstream)thinking after an initial evaluation is accessed, it could affectthecorrela-tion between automatic and deliberative measures of attitudes because


5.Like theMCM,othertheorieshave proposed that some factorin addition to accessibil-ity determines the impact of ones stored attitudes. However, in these alternative frame-works, these factors typically have been defined with respect to some other concept. Forexample, in the accessibility-diagnosticity approach, the diagnosticity of an attitude is de-fined as the extent to which a respondent believes the attitude correctly indicates how aparticular second response should be made (Feldman & Lynch, 1988). Similarly, the rele-vance of an attitude (which can impact its use) is defined only with respect to some otherparticularoutcome variable (e.g., a particularbehavior;Snyder, 1982). Thus,one speaks ofthe diagnosticity or relevance of an attitude for a particular outcome. In contrast, per-ceivedvalidityis moregeneraland canbe assessedwithout reference to a specific outcomevariable. Thus, it is more useful to store in memory.


22/30

new aspects of the attitude object could come to mind that conflict with(orreinforce)what isrepresentedat thestructural level (see also, Fazio &TowelsSchwen, 1999; Gawronski & Bodenhausen, 2006). However,

justbecause thinkingfollowingretrievalof an evaluation cansometimesreduce automaticdeliberative correlations, this does not mean thatevaluations that are based on high amounts of thinking should showlower automaticdeliberative correlations. To the contrary, whenevalu-ations are based on high amounts of thinking, this will generallyenhance thecorrelation betweenautomaticanddeliberativemeasures.

According to the MCM, attitudes based on high amounts of thinkingshould show higher automaticdeliberative correlations for two rea-sons. First, high amounts of thinkingat least if it is evaluatively con-gruent thinkingshould produce stronger objectevaluation

associations. This should enhance the impact of the automatic compo-nent on deliberative measures. Second, thinking should enhance thesense of perceived validity. This should increase the impact of themetacognitive component on deliberative attitudes. Indeed, the extentof elaboration has been related to attitude accessibility (Bizer &Krosnick, 2001), the extent of attitude confidence (Haugtvedt & Petty,1992), as well as the criteria of attitude durability and impact (see Petty,Haugtvedt, & Smith, 1995, for a review). As might be expected then, themore elaboration an attitude object has received, the stronger the corre-lation between deliberative and automaticmeasures of attitudes tend to

be (Nosek, 2005).

UNDERSTANDING PREJUDICE

As noted earlier, perhaps the most unique aspect of the MCM is theproposition that people can store validity tags. Besides those alreadymentioned, what are some of the insights that this possibility affords forunderstanding social psychological phenomena? Consider the domainof prejudicewhere manyunprejudiced individuals mightbe expectedtohave both positiveandnegativeassociationstosomestigmatizedgroup,

but believe that thenegativelink is wrong.What aretheconsequences ofhaving a wellpracticed wrong tag? Current research suggests thatcontinually practicing a negation would not necessarily stop negativetraitsfrom comingto mind when presentedwitha memberof theminor-

ity group (see Gawronski, Deutsch, Mbirkou, Seibt, & Strack, in press).However, accordingto theMCM,continualpracticeshould lead theper-son to immediately think no or wrong when such stereotypes areactivated. Will peoplewhohave practicednegating stereotypes befasterto reject them (i.e., think wrong after the stereotype comes to mind)

678 PETTY ET AL.


23/30

thereby correcting themselves (and perhaps others) than people whohave not practiced negation? No research has examined this specificissue yet, but the MCM predicts that such an effect is possible.

Staying withinthedomain of prejudice, the MCMoffers a unique per-spective on various conceptualizations postulating thatpeople are oftenmotivated to correct for their internalized prejudice. According to sev-eral formulations (e.g., Dovidio & Gaertner, 2004; Dunton & Fazio,1997), some White individuals have automatic negative reactions toBlacks but have egalitarian values or desires not to be prejudiced thatcause them to discount their activated negativity and report positivityon explicit measures. This analysis assumes that the causal sequence isthat people have preexisting attitudes that are negative, some preexist-ing motive to control these reactions, and these interact to determine aconstructed deliberative positive attitude. Although this is perfectlyplausibleand certainlycan occur,the MCMholdsopenanother possibil-itythat motives can follow from preexisting positive and negative as-sociations, with the latter negated. In other words, some people willrecognize that they have both existing positive and negative associa-tions, with the latter being unwanted. Because they find the latter to beinappropriate or wrong, they develop a motiveto controlthese negativereactions. Thus, rather than a positive constructed attitude followingfrom theinteraction of negative automaticattitudesanda motiveto con-trol them, it could be that a motive to control negative reactions followsfrom the presence of both positive and negative associations to aminority group with the latter being rejected.

SUMMARY AND CONCLUSIONS

In this article we have outlined a MetaCognitive Model (MCM) of atti-tudes. In brief, the model holds that attitudes consist of evaluative asso-ciations (positive and/or negative) along with validity tags. As ourdiscussion above implies, the MCM differs from the single, dual, andprocess models of attitudes in several ways. The MCM offers one inte-grated attitude representation (in contrast to the dual attitudes ap-proach), but postulates that attitude objects can be linked to bothpositive and negative evaluations (in contrast to the single attitude ap-proach). Viewing theattitude representation as an integrated unit rather

than as separate representations activated in different situations (as ad-vanced by some dual attitudes models) allows for joint activation ofpositivity andnegativity in anygiven situationwhere theattitude objectis encountered (assuming that people have both positive and negativeassociations). The possession of both positive and negative associations



24/30

can lead to explicit ambivalence when both valences are endorsed, or toimplicit ambivalence when one valence is accepted and the other is re-

jected. We have also shown in this article how the MCM can account fora variety of attitudinal phenomena in the literature as well as make newpredictions.

Intermsofdefiningattitudes,whenpeoplehavejustonevalenceofas-sociation, the MCM reduces to the single attitude approach (unless thissingle valence is rejected in which case the attitude structure is morecomplex). Because the MCM posits an enduring attitude structure, themodel lies more squarely in the camp postulating that attitudes can bestoredand need notbe constructed anew each time. Indeed, intheMCM,attitudescan be viewedas a persons stored evaluative associations thatare not rejected. However, construction processes are necessarily in-

volved in completing an explicit selfreport if only to translate ones re-trieved sense of positivity and/or negativity onto the attitudeassessmentoffered (e.g., a bipolar scale).6 The MCM holds thatat a mini-mum, when completing a deliberative scale, the attitude report is con-structed from the activated evaluation(s) and any validity tags that areretrieved. At a maximum, theattitude reportcanalso consider anyotherinformation that comes to mind and is activated by the current context.The MCM holds that explicit attitude reports will be more durable,impactful, and correlated with automatic measures (i.e., high attitudestrength), when the evaluative associations are highly accessible andheld with high confidence.

REFERENCES

Barden, J.,Maddux, W. W.,Petty, R. E.,& Brewer, M.B. (2004). Contextual mod-eration ofracialbias: Theimpactofsocial roles oncontrolledandautomat-ically activated attitudes. Journal of Personality and Social Psychology, 87,522.

Bassili, J. N. (1996) Metajudgmental versus operative indices of psychologicalproperties: The case of measures of attitude strength.Journal of Personalityand Social Psychology, 71, 637653.

Betsch, T., Haberstroh, S., Molter, B., & Glckner, A. (2004). Oops, I did itagainrelapse errors in routinized decision making. Organizational Be-havior and Human Decision Processes, 93, 6274.

Bizer, G. Y., & Krosnick, J. A. (2001). Exploringthe structureof strengthrelated

680 PETTY ET AL.

6. With continued experience with bipolar scales, one integrated summary evaluationmight be stored along with any positive and negative links.


25/30

attitude features: The relation between attitude importance and attitudeaccessibility. Journal of Personality and Social Psychology, 81, 566586.

Bizer, G. Y., Petty, R. E., & McCaslin, M. (2005). How we conceptualize our atti-tudes matters: The effects of valence framing on the resistance of politicalattitudes. Political Psychology, 26, 553568.

Briol, P., & Petty, R. E. (in press). Automatic and deliberative attitude changefrom thoughtful andnonthoughtful processes. In R. E. Petty,R. H. Fazio,& P. Briol (Eds.), Attitudes: Insights from the new implicit measures.Mahwah, NJ: Erlbaum.

Briol, P., Petty, R. E. & Wheeler, S. C. (May 2003). Implicit ambivalence: Impli-cations for attitude change. Paper presented at the Attitudinal Incongru-enceand InformationProcessing Symposium.Amsterdam, Netherlands.

Briol, P., Petty, R. E., & Wheeler, S. C. (2006). Discrepancies between explicitand implicit selfconcepts: Consequences for information processing.

Journal of Personality and Social Psychology, 91, 154170.Byrne, D. (1961). Interpersonal attraction and attitude similarity. Journal of Ab-

normal and Social Psychology, 62, 713715.Cacioppo, J.T., Gardner, W.L., & Berntson, G.G. (1997). Beyond bipolar concep-

tualizations andmeasures:The caseof attitudes andevaluative space. Per-sonality and Social Psychology Review, 1, 325.

Chaiken, S.,Liberman,A.,& Eagly,A. H. (1989). Heuristicandsystematic infor-mation processing within and beyond the persuasion context. In J. S.Uleman & J. A. Bargh (Eds.), Unintendedthought (pp. 212252). NewYork:Guilford Press.

Cohen, J. B., & Reed, A. (2006). A multiple pathway anchoring and adjustment(MPAA) model of attitude generation and reinforcement. Journal of Con-sumer Research, 33, 115.

Conrey,F.R., & Smith, E.R.(2007).Attituderepresentation: Attitudes as patternsin a distributed, connectionist representational system. Social Cognition,25(5), 718735.

DeCoster,J., Banner,M.J., Smith,E. R.,& Semin,G. R.(2006).Onthe inexplicabil-ityof theimplicit: Differences in theinformation provided by implicitandexplicit tests. Social Cognition, 24, 521.

deLiver, Y.,van der Pligt, J.,& Wigboldus, D. (2007). Positiveand negativeasso-ciations underlying ambivalent attitudes. Journal of Experimental SocialPsychology, 43, 319326.

Deutsch, R.,Gawronski, B.,& Strack,F. (2006). At theboundaries of automaticity:Negation as reflective operation.Journal of Personality and Social Psychology,91, 385405.

Devine, P. G. (1989). Stereotypes and prejudice: Their automatic and controlledcomponents. Journal of Personality and Social Psychology, 56, 518.

Dovidio,J., Kawakami,K.,Johnson,C., Johnson,B., & Howard,A. (1997).The na-ture of prejudice: Automatic and controlled processes. Journal of Experi-mental Social Psychology, 33, 510540.

Dovidio, J. F.,& Gaertner,S. L. (2004). Aversive racism. In M.P. Zanna(Ed.),Ad-



26/30

vances in experimental social psychology (Vol. 36, pp. 151). San Diego: Aca-demic Press.

Dovidio, J. F.,Kawakami, K., Smoak,N., & Gaertner,S. L. (inpress). Theroles ofimplicitand explicitprocessesincontemporaryprejudice. InR.E. Petty,R.H.Fazio,& P.Briol (Eds.),Attitudes:Insightsfrom the newimplicit measures.Hillsdale, NJ: Erlbaum.

Dunton,B. C.,& Fazio.R. H.(1997).Anindividual difference measure ofmotiva-tionto control prejudiced reactions. Personality and Social Psychology Bulle-tin, 15, 543538.

Eagly, A.H., & Chaiken, S. (2007). The advantages of an includsive definition ofattitude. Social Cognition, 25(5), 582602.

Eiser,J. R.,Fazio,R.H., Stafford,T., & Prescott,T. J. (2003). Connectionist simula-tion of attitude learning: Asymmetries in the acquisition of positive andnegative evaluations. Personality and Social Psychology Bulletin, 29 ,12211235.

Fazio, R. H. (1995). Attitudes as objectevaluation associations: Determinants,consequences, andcorrelates of attitude accessibility. In R. E. Petty & J. A.Krosnick (Eds.), Attitude strength: Antecedents and consequences (pp.247283). Hillsdale, NJ: Erlbaum.

Fazio, R. H., Jackson, J. R., Dunton, B. C., & Williams, C. J. (1995). Variability inautomatic activation as an unobtrusive measureof racial attitudes: A bonafide pipeline? Journal of Personality and Social Psychology, 69, 10131027.

Fazio, R. H. (2007). Attitudes as object-evaluation associations of varyingstrength. Social Cognition, 25(5), 603637.

Fazio, R. H., & Olson, M. A. (2003). Implicit measures in social cognition re-search: Their meaning and use. Annual Review of Psychology, 54, 297327.

Fazio, R. H., & TowlesSchwen, T. (1999). The MODE Model of attitudebehav-

ior processes. In S. Chaiken & Y. Trope (Eds.), Dual process theories in socialpsychology (pp. 97116). New York: Guilford Press.

Feldman, J. M., & Lynch,J. G. (1988). Selfgenerated validityandother effects ofmeasurementonbelief, attitude, intention, andbehavior.Journalof AppliedPsychology, 73, 421435.

Fiske, S. T., & Pavelchak, M. A. (1986). Category-based versus piecemeal-basedaffective responses: Developments in schema-driven affect. In R. M.Sorrentino & E. T. Higgins (Eds.), Handbook of motivation and cognition:Foundations of social behavior (pp. 167203). New York: Guilford Press.

Gawronski, B., & Bodenhausen, G. V. (2006). Associative and prepositional pro-cesses in evaluation:An integrative review of implicitand explicitattitudechange. Psychological Bulletin, 132, 692731.

Gawronski, B., & Bodenhausen, G. V. (2007). Unraveling the processes underly-ing evaluation: Attitudes from the perspective of the APE Model. Social

Cognition, 25(5), 687717.Gawronski, B.,Deutsch,R.M.,Mbirkou,S., Seibt,B.,& Strack,F.(in press). When

just say no is not enough: Affirmation versus negation training and the

682 PETTY ET AL.


27/30

reduction of automatic stereotype activation.Journal of Experimental SocialPsychology.

Gilbert, D. T. (1991). How mental systems believe. American Psychologist, 46,107119.

Greenwald, A. G., & Banaji, M. (1995). Implicit social cognition: Attitudes,selfesteem, and stereotypes. Psychological Review, 102, 427.

Greenwald, A.G., McGhee, D.E., & Schwarz, J.L.K. (1998).Measuring individualdifferences in implicit cognition: The Implicit Association Test. Journal ofPersonality and Social Psychology, 74, 14641480.

Gregg,A. P.,Seibt,B.,& Banaji,M. R.(2006). Easier done than undone: Asymme-try in the malleability of implicit preferences. Journal of Personality and So-cial Psychology, 90, 120.

Han, H. A., Olson, M. A., & Fazio, R. H. (2006). The influence of experimentallycreated extrapersonal associations on the Implicit Association Test. Jour-nal of Experimental Social Psychology, 42, 259272

Haugtvedt, C.P., & Petty,R. E. (1992). Personality andpersuasion:Needfor cog-nition moderates the persistence and resistance of attitude changes. Jour-nal of Personality and Social Psychology, 63, 308319.

Holland, R., W., Verplanken, B., & van Knippenberg, A. (2003). From repetitionto conviction: Attitude accessibility as a determinant of attitude certainty.

Journal of Experimental Social Psychology, 39, 594601.Hovland, C. I., Janis, I. L., & Kelley, H. H. (1953). Communication and persuasion:

Psychological studies of opinion change. New Haven, CT: Yale UniversityPress.

Jost, J. T., Kruglanski, A. W., & Nelson, T. O. (1998). Social metacognition: Anexpansionist review. Personality and Social Psychology Review, 2, 137154.

Kaplan,K. J. (1972).Ontheambivalenceindifference problem in attitudetheory

and measurement: A suggested modification of the semantic differentialtechnique. Psychological Bulletin, 77, 361372.

Karpinski, A., & Hilton, J. L. (2001). Attitudes and the Implicit Association Test.Journal of Personality and Social Psychology, 81, 774788.

Kehr, H. M. (2004). Implicit/explicit motive discrepancies and volitional deple-tion among managers. Personality and Social Psychology Bulletin, 30,315327.

Krosnick, J. A., Judd, C. M., & Wittenbrink, B. (2005). The measurement of atti-tudes. InD.Albarracn,B.T. Johnson,& M.P. Zanna(Eds.),Thehandbookofattitudes (pp. 2176). Mahwah, NJ: Erlbaum.

Krosnick, J. A., & Petty, R. E. (1995). Attitude strength: An overview. In R. E.Petty & J. A.Krosnick(Eds.),Attitudestrength:Antecedentsand consequences(pp. 124). Mahwah, NJ: Erlbaum.

Kruglanski,A. W. (1989). Thepsychology of being right: Theproblem of accu-

racyin social perception andcognition.PsychologicalBulletin,106, 395409.Kruglanski, A. W., & Thompson, E. P. (1999). Persuasion by a single route: A

view from the unimodel. Psychological Inquiry, 10, 83109.



28/30

Kumkale, G. T., & Albarracn, D. (2004). The sleeper effect in persuasion: Ametaanalytic review. Psychological Bulletin, 130, 143172.

Lichtenstein, M., & Srull, T. K. (1985). Conceptual and methodological issues inexamining the relationship between consumer memory and judgment. InL. F. Alwitt & A. A. Mitchell (Eds.), Psychological processes in advertising ef-

fects:Theory, researchandapplication (pp.113128).Hillsdale, NJ:Erlbaum.Lingle, J.,& Ostrom, T.M.(1981). Principles ofmemoryand cognitionin attitude

formation. In R. E. Petty, T. M. Ostrom, & T. C. Brock (Eds.), Cognitive re-sponses in persuasion (pp. 399420). Hillsdale, NJ: Erlbaum.

Maddux, W. W., Barden, J., Brewer, M. B., & Petty, R. E. (2005). Saying no tonegativity: The effects of context and motivation to control prejudice onautomatic evaluative responses. Journal of Experimental Social Psychology,41, 1935.

Maio, G. R., Bell, D. W., & Esses, V. M. (1996). Ambivalence in persuasion: Theprocessing of messages about immigrant groups. Journal of ExperimentalSocial Psychology, 32, 513536.

Mayo, R., Schul, Y., & Burnstein, E. (2004). I am not guilty versus I am inno-cent: Successful negation may depend on the schema used for its encod-ing. Journal of Experimental Social Psychology, 40, 433449.

NewbyClark, I. R., McGregor, I., & Zanna, M. P. (2002). Thinking and caringabout cognitive inconsistency: When and for whom does attitudinal am-

bivalence feel uncomfortable? Journal of Personality and Social Psychology,82(2), 157166.

Nosek, B. A. (2005).Moderators of the relationshipbetween implicitandexplicitevaluation. Journal of Experimental Psychology: General, 134, 565584.

Osgood, C.E.,Suci, G. J.,& Tanenbaum,P.H. (1957). The measurement of meaning.Urbana, IL: University of Illinois Press.

Petty, R. E. (2006). A metacognitive model of attitudes. Journal of Consumer Re-search, 33, 2224.

Petty, R.E.,& Briol,P. (2006).A metacognitive approachtoimplicit andex-plicit evaluations: Comment on Gawronski and Bodenhausen (2006).Psychological Bulletin, 132, 740744.

Petty, R. E., & Briol, P. (in press). Implicit ambivalence: A metacognitive ap-proach. InR.E.Petty,R. H.Fazio,& P.Briol (Eds.),Attitudes:Insights fromthe new implicit measures. Mahwah, NJ: Erlbaum.

Petty,R. E.,Briol, P.,& Tormala, Z. L. (2002). Thought confidence as a determi-nant of persuasion: The selfvalidation hypothesis. Journal of Personalityand Social Psychology, 82, 722741.

Petty, R. E., Briol, P., Tormala, Z. L., & Wegener, D. T. (2007). The role ofmetacognition in social judgment. In E. T. Higgins & A. W. Kruglanski(Eds.), Social psychology: Handbook of basic principles (2nd ed., pp. 254284).

New York: Guilford Press.Petty,R. E.,& Cacioppo, J.T. (1986). Communicationandpersuasion:Centraland pe-

ripheral routes to attitude change. New York: Springer/Verlag.Petty,R. E.,Haugtvedt,C. P.,& Smith,S. M.(1995). Elaboration as a determinant

684 PETTY ET AL.


29/30

of attitude strength: Creating attitudes that are persistent, resistant, andpredictive of behavior. In R. E. Petty & J. A. Krosnick (Eds.), Attitudestrength:Antecedents andconsequences (pp.93130). Mahwah,NJ: Erlbaum.

Petty,R. E.,& Jarvis, W.B.G. (1998, October). What happens to theold attitude whenattitudeschange? Presentedat theannualmeeting of theSocietyfor Experi-mental Social Psychology, Lexington, KY.

Petty, R. E., Tormala, Z. L., Briol, P., & Jarvis, W. B. G. (2006). Implicit ambiva-lence from attitude change: An exploration of the PAST Model. Journal ofPersonality and Social Psychology, 90, 2141.

Petty, R. E., Tormala, Z. L., & Rucker, D. D. (2004). Resisting persuasion bycounterarguing:Anattitudestrengthperspective. InJ. T. Jost,M. R.Banaji,& D.A. Prentice (Eds.), Perspectivism in socialpsychology: Theyin andyang ofscientific progress (pp. 3751). Washington, DC: American PsychologicalAssociation.

Petty,R. E.,Wells,G. L.,& Brock,T. C. (1976). Distraction canenhance or reduceyielding to propaganda: Thought disruption versus effort justification.

Journal of Personality and Social Psychology, 34, 874884.Petty, R. E., Wheeler, S. C., & Tormala, Z. L. (2003). Persuasion and attitude

change. In T. Millon & M. J. Lerner (Eds.), Handbook of psychology: Vol. 5:Personality and social psychology (pp. 353382). Hoboken, NJ: John Wileyand Sons.

Plant, E. A., & Devine, P. (1998). Internal and external motivation to respondwithoutprejudice.JournalofPersonality andSocial Psychology, 75, 811832.

Priester, J. R., & Petty, R. E. (2001). Extending the bases of subjective attitudinalambivalence: Interpersonal and intrapersonal antecedents of evaluativetension. Journal of Personality and Social Psychology, 80, 1934.

Rosenberg, M. (1965). Society and the adolescent selfimage. Princeton, NJ: Prince-

ton University Press.Rydell, R. J.,& McConnell,A. R. (2006). Understandingimplicit andexplicitatti-

tude change: A systems of reasoning analysis.Journal of Personality andSo-cial Psychology, 91, 9951008.

Rydell, R. J., McConnell, A. R., Mackie, D. M., & Strain, L. M. (2006). Of twominds: Forming and changing valenceinconsistent implicit and explicitattitudes. Psychological Science, 17, 954958.

Schwarz, N. (2007). Attitude construction: Evaluation in context. Social Cogni-tion, 25(5), 638656.

Schwarz,N., & Bohner, G.(2001). Theconstructionof attitudes.In A.Tesser& N.Schwarz (Eds.), Blackwell handbook of social psychology: Intraindividual pro-cesses (pp. 436457). Balden, MA: Blackwell Publishers.

Sherman, J. W. (in press). Controlled influences on implicitmeasures:Confront-ingthe myth of processpurity andtamingthe cognitive monster. In R. E.

Petty,R. H. Fazio,& P. Briol (Eds.),Attitudes:Insights fromthe new implicitmeasures. Hillsdale, NJ: Erlbaum.

Snyder, M. (1982). When believing means doing: Creating links between atti-tudesandbehavior. InM.P. Zanna, E. T. Higgins, & C.P. Hermann (Eds.),



30/30

Consistency in social behavior: The Ontario Symposium (Vol. 2, pp. 105130).Hillsdale, NJ: Erlbaum.

Thurstone, L. L. (1928). Attitudescanbe measured.AmericanJournal of Sociology,33, 529554.

Tormala, Z. L., DeSensi, V. L., & Petty, R. E. (2007). Resisting persuasion by ille-gitimatemeans: A metacognitive perspective on minority influence. Per-sonality and Social Psychology Bulletin, 33, 354367.

VanOverwalle, F.,& Siebler, F. (2005). A connectionist model of attitude forma-tion and change. Personality and Social Psychology Review, 9, 231274.

Vargas, P. T. (2004). The relationship between implicit attitudes and behavior:Some lessons from thepast and directions for thefuture. In G. Haddock &G. R. Maio (Eds.), Contemporary perspectives on the psychology of attitudes(pp. 275297). New York: Psychology Press.

Wells, G. L., Olson, E. A., & Charman, S. D. (2003). Distorted retrospective eye-witness reports as functionsof feedback anddelay.Journal of ExperimentalPsychology: Applied, 9, 4252.

Wilson, T.D.,& Hodges, S.D. (1992). Attitudesas temporaryconstructions. In L.L.Martin & A.Tesser(Eds.),The construction of social judgments (pp.3765).Hillsdale, NJ: Erlbaum.

Wilson, T.D., Lindsey,S.,& Schooler, T.Y. (2000). A modelof dualattitudes.Psy-chological Review, 107, 101126.

Wittenbrink,B., Judd, C. M.,& Park, B. (2001).Spontaneous prejudicein context:Variability in automatically activated attitudes. Journal of Personality andSocial Psychology, 81, 815827.

Zanna, M.P., & Rempel, J. K. (1988). Attitudes: A new look at an old concept. InD.BarTal & A.W.Kruglanski (Eds.),Thesocial psychologyof knowledge (pp.315334). Cambridge, UK: Cambridge University Press.

686 PETTY ET AL.

2007 social cognition mcm

Documents