\

hr1plici1t Mea!:•ures of Alttitudt~~s

Edited by

BERND WITTENBRIINK NORBERT SCHWAHZ

2007

THE GUILFORD PRESS New 'ri:>rk London

11 VVhat Do \Ne Know about Implicit Attitude Me1asures and \Nhat Do \Ne Have to Learn?

Betrtram Gawronski Galen V. Bodenhausen

I cannot totally grasp all that I am. , .. For that darkness is lamentable in which the possibilities in me are hidden from myseJf: so that my mind, questioning itself upon its own powers, feels that it cannot rightly trust its own report.

-ST. AUGUST.!NE, Confessions

In his influential Confessions, St. Augustine lamented the evasiveness of full self-understanding, questioning the reliability of our intuitions about our own minds. Although introspection and self-report formed the foundation of the earliest approach.,; to study­ing psychology, the development of psychological research over the course of the 20th century led social psychologists steadily back to the view espoused by St. Augustine in the early 5th century. In partic­ular, demonstrations of respondents' lack of introspective access to the causes of their own judgments and behavior (see, e.g., Nisbett & Wilson, 19'77) forever shattered the illusion that self-reports would be a sufficient means for illuminating the workings of the mind. Al­ternatives to self-report began to be formula1ted, and at the dawn of the 21st century an explosion of research is exploring the usefulness of these new, implicit measures of mental coments and processes (see,

265

266 CHIT'CAL PERSPECTIVES

e.g.: L~nc .. B:nwji, .'Jm,ck, & (~rccn\\-·ald 1 Chapter .1, this \·olumc~ \X:'ntcJtbrink, Ch<Iptcr 2, th1s \'Ohmt<.:L \Xlhile ,mdcni<Jhly cxc·iting, this plL·thorct uf JH.'W techniques for peering into our inner mental lin:s rc1iscs a number of import<lnr qucstionh l'hat will have t·o be thoroughly E'-.,c,ndwd Jnd :--aci-.faLtorily am.wcrcd Ill or(h~r for the prom1sc ol thc~L' llC\\: meJsures w he fully rcalitcd. \\'hat 1s the ap­propnatc theoretical cunstruc1. corresponding, tu each measmc? Do lhC<>l' measures really provide J.(Ces~ to unconsuous mcnral pro­(('<.;st·s? \Xihat do the chffcn:nt measures h<W(' in comnH.m, and what is uniqnc to cad1 particuLn mc-;:.surc? f I ow ar1: the Jiffcrcnt t_ncasures related 1o ph~·>iolog.ictl corr(·Lncs? lt is tO\V!lrLI rhcsc quc~uons lhar we turn our :1ttcnnon.

CONCEPTUAL CLAR:ITY

The ini tla 1 cnthusiasn1 for implicit mca.<.,Ul'es st~emed to be accom pa :1ied hv rclaLivclv loose J.n~l <.:.hifrir.g cmh:cptuah:t.ations nf their nw:1r.ing. P~nr ol the. confu~ion \HH.ior:htcdly <Hiscs +rom more fundcHlKntal amhic;u it 1rs Jlld dcb.a tcs .1 brw t the llle;1 ning o+ rcleY3Ht dworetica I :.:on struc~~. l·or example, mnch of the work on implicit mccbl1fCS hJs fo(:tl'~cd nn the ;1s.scssmcn"L of :ttlitudcs, However, fund.1menul dis­agreements <:.till exist concerning the appropnatl" (.onceprualiz.ation of "attirudL·," Some theorist'~ V!CW atlitudcs .<:~.<:.enduring structures m loruz,~tcm.l memory 1:c.~., Fazin 1 '1995). wherc;1~ others vie\:o..·· them as .mo;·nenr-:1ril v <:on::.lructcd evalualinn~ that i ntcgr ate current ~:ontL·xnw 1 infonnatior~ with sclect:i ve su bscts of long ·term memory { c .g.., Sch \"~/<.Hi'. & Holmer. ~~001 ). ~/hcrc<l~· the tormc-r .1pproach conceives ol ~1tritudcs :h n:LHivelv <.;talic and defined by fix.ed stru(:tural prorcnics, the bttL·r {-'niphJ.si:r.c~ a dynam1c {)tTJcc·~s with tninmtal structural ass_l~~nplion,r.,. l"hcorists hom thcs(· different camp'-> rnigbr very well make dttkrcnt as­sumptions about \vbat an itnp\icit mca:..uf{' of attitudes is capturing-.. fhe .strength of a st:-1 hlc <1SHKicuion in lnng-ter111 memory on one hand, or t!w cJ,;lc:cgcnt ner c\·c1luarive implications o+ cnntcxually ~1ctivatcJ knowledge and situarion<.dly :1vailab\c input on the other. Thu~, Libel­ing something as an "attitude measure"_ ?(H.\'i not.hing to ~.larity lhe nhttter, because the tcrrn attiiudc mo::nns ddl'erc11t dung~ to chttcrent rc­

'->carchcrs. ~:t \'-.au open question whclher imp] kit me:;1sures nup well on to~' p:1nicubr conceptualization of ·'attirllde/' a questmn that must be JdJres ... .:.-d by sysu:-nntic empirical investigation.

The importance of thi~ issue becomes even more apparent in th_e t.:outext of earlv theorizing t'tal implicit attitude measures pronde di­n:~(."t <KCC<.:..s to ~tahle ev::t\uative reprcscntalions that have their roots

What Do \Ne Have to Learn? :!67

in long-term SOl~ializ;Hion cxpcrii..'nccs (Sl'C 1 e.g., Dovidio~ Kawakami, &. Reach, 2001; (~rcrmvJld &. Banaji, 199S; Rudman, 2004; \'ll:iilson, l.mdscy, C.\. Schooler~ 2UOO). This as:-.ulllption has been L:halknl!:cd over the Llst· few years hy ;tccumuLning cvidcn~.:c that implicit IT;;.,l sut-es of arritudcs arc hi~hly susceptible to cnntcxtualtnflucncc..:. 1Jor a review, see Bla1r, 2002). However, even though such findillgs J'tl<.l\'

he interpreted as evidence for the attituclcs-;~s-\"onstruL:tions a~:cnun.t (Sdrwar;: & Bohner, 2001 i:, it 1:. not entirely dear how llllplicit atti­tude l!lc;burts arc inihlUKcd hr l"hl' co!lll'Xl. DcK·s the context lead to J >hift in thv rncasurement ot. :m otherwise :-.ublc attitude~, or doc<.; dw context influence ho\v the attitude itself i.<> constructed on rhc spot? Ag<Jm, the specific allS\VC'r to these qucstioJL'I dcp(·nds on tht:' preferred COJin:ptii::.JI i;:at inu nf au itudcs as mon 1<:nta rv L·onstrucriuns (SdtVV<1l'Z & Hohner, 2001) or Stahle ~>;,-·a\u;Jtivc ;TJlr<.'\t::JII.ati\·l!1S {Fazio, 1')95). ~ot\\'Ith,tanding th(\'>C atnhigultics, the fact t.hat irnplicit ~atitude meas11rc;; show a stroll[!. su:;cepribility to LOlltexi~Ltal intlucJKcs poses a serious challenge w Lhe original cxpn:t:1tiun 1ltat these measures prov.ide din:cl: c11d unbiased :lcccss to st::1hlc c\·al­uanvc rcprcscmations in mcmorv.

:\nothcr fundamentalt.·onfusion, TJ{>tcd bv ::.ever a! S(holars !'sec c:-.­

pe(iJlly De I Touwer, 2006)) concerns rhe mear.ting of the term IJU~a.~ure, wh.t...:h {.:an he use-d to refer w ;1 me:l.-.tliTIIlcnt fnocedure or w rhc out­tome nf ;:1 mcasuJ-cmcm proccd ure. As I k l Tmnvcr (2006) noted, :1

nteasureJrwn-r proccd ll rc cau he dire~~r or i ndircct. \vhercas dl(' Le rm im /Jfit:it is mec1ningfnl only with r{'f_cud ro the mltcc.;mc of a mc~burcmeni [lroccdun·. Conversely, it Jm:'s not make sense to call the outcome of mcasun:lfJCnL pro~...~edurc direct or indirect, just as it docs nnt Jllclke

sense ro 1:-:;hci: a EllC<lSurement procedure explicit or implicit .. An ilmpor­

tant considcr:1tio11 is that v:herea.s rhr direcr vnsu:.;. indirect natutrc o~ a rrH:a~urcnK·nr procedure can be determmed ;1 prion hy the objective properti1.>i nf the ta.sk 1 the explicit verslL~ impJJ.cit nam rc of the ou tcoJn(· of a mc<1surcmcnt procedure need:-. to he cstahli·~hed empirically·.

ln rhts cont'cxt, De H.ouwer (2006) pointed out another, n·latcd problem in the w;.Jy rhc term implicit is used. In particular, he argued th;lt rc:.carcher.s often tail to ~pccify in what sense a P<'Hticuhr mea­sure shonld he regarded as !mplk;r. A-. onginallv used Hl the mcmorv iiLer~Hure on patients vvith amnesia {St'f', e.g.~ \Van:ingto11 &. \Vei.skr~nt;~ l<.J(~R), "implici.t'' n1cntJl processes referred specific1lly to pnKcsses tlur operated in lhe abseiKc nf conscious cnvareness. Implicit mem­ory thu;;. wa~ evident in task p(-;rfnrmancc t"ct'lccting the residue of previous expcncncc in the Jbsence of :1ny explicit: JllL'IIlOry fnr that experience. l.n keeping wirh thi~ precedent, it se('ms to he ....:omn1nnlv the ~:Js.c in the :-i.ocai psychology· literature tlut when{'ver smYtcthmg

268 CRITICAL PERSPECTIVES

is called "implicit," it is as\umed to he consciously inaccessible. But as De Houwcr (2006) noted, this is an empirical assumption~ and it is usua!!y an unrestcd one.

Best de::. the interpretation of the term implicit as "unconscious,'' implicit attitude measures arc sornetimes assumed to reflect ·'auto­matic'' J.rtitudes. Bargh { J 994) identified four separate senses in whkh infornMtion rn;ccssing can he considered automatic, and lack of av .. -areness is hut one ot these "four horsemen" of automaticity. rvlany ~o~called autornatic processes happen with a\varencss (Moors & De Ilouwcr, 2006). At the level of implicit attitude measures, it is thns important to specify which aspect(s) of automaticit~' one capit;-1]­izcs on. It mav be rh~H avv'areness covaries 'vith measures, such thcH for some mc~surcs, respondents are consciously a\vare of the as·­sessed attitude. ·rhe same holds for other aspects of automaticity (i.e., controllability. spontaneity, anJ resource dependency:!. \'V'e argue that it is important to un(krstJrH.l how these issues ma-p onto the features of a particu!Jr task and ·whether they remain invariant across differ­ent l:Oiltent domains or interact with content. It may be, for example, that the comrollabi!ity of responses varies across different tasks, across content domains~ or as a function of motivation and practice at controlling particular kinds of responses. It is inuunbem upon re­searchers to understand the scope of what rhcy can claim about the implicitness nf their measures and to railnr thur theoretical conclu­sions accordingly.

Another major lesson for researchers that \viii lead to greater conceptual clarity is the loss of innocence regarding our aspiration to create "process-pure" measures. A process-pure measure \vmdJ he one that cleanly and unambiguously indexes a single construct of in­terest. !\'Lmy people using the Implicit Association Test (lAT), for ex­ample~ functionally regard it as a process-pure measure of the strength of a mental associ:.ltion (or, more accurately, a pair of associ­ations). Ho\.vcver, performance on the IAT is clearly influenced by many other bctors, some of \·vhich are theoreti(al\y uninteresting and potcntiallv controllable le.g., general perceptnal-motor skills) and others of \vhich are highly interl·sting m their own right. The application of process-dissociarion techniques (Conrey, Sherman, GJwronski, J-lugcnberg, & Grocnn~ 2005; Payne, 2001) provides a way of SJ'StcmaticJlly decomposing these theoretically relcv;-mt coJn­ponents of performance, and it n~presents one of the most promising and importanr directions for research employing im.plicit attitude measures. For example, Conrey and colleagues 1)005), in their Quad-:vfodel, have shown that performance on the !AT iand many orhcr mcJsures'l reflects not only the automatic associative bias that

What Do We Have to Learn? 269

has been the focus of most IAT' us.crs, hut also on several othc.T J.rro­cess.es, such as the discriminability of the .stimulus, success <~.r 0\'Cr­

coming bias, and general guessing biases. Of the other processes in the Quad-lvlodel, the theoretically most relevant is the process of overcoming I.Jicts. fviultinomial modeling of IAT data has confirmed that the motivation/ability ro ov('rcome the 3lltomaric associative hia.., also influences IAT performance sy:-:tematically. By applying a process-dissocicHion methodology to the TAT 1:among other mc8 -

sures), i~ becomes possible to obtain much more finely tuned esti­mares oJ· the_ a~ltomatic associative bias by pulling out sepcuate pro­cesses thGt wtlucncc overall rerformance. The important general lesson here is thar no measure is proc.ess pure-. and a major path to­:~Jard clarity lies in the application of new techniques for decotnpo!-i­mg task performance into more specific, conceptuaUy meaningful subcomponents. By defining our terms and specifying our constructs as precisely as possible, rapid progress will be- greatly facilitated.

INTROSPECTION AND CONSCIOUSNESS

D<_!cs the outcome of an indirect measurement procedure necessarilv retlecr an implicit attitude in the sense that it is unavaihblc ro lntn;­spc..::tion or self-report? Not necessarily. As noted bv De Hou\ver ::2~){:6:1, whether or not the construct as~essed by a giv~'n task is ''im­p!t~Jt" (nr unconsciouS:! needs to be established empirically. IndceJ, attitudes ass.e~sed \vlth indire-ct measures ;-lre sometimes highly corre­bted with selt-rcporred atritudcs (e.g., !lanse, Sersc, & Zcrbes, 20tll; Dovidio~ Kawakami, Johnson, Johnson, & Hmvard. 1997; ·Icach­rrwn, Gregg~ & \\loody: 2001:1. On one hand~ these findings may incli­C<lte that the attitude reflected in the implicit measure is av<lilable to introspection. On the other hand~ hovvcver, one could object that there might be a separate, implicit aspect to the attitude that just happens to coincide with the explicit aspect in tl1c case of some atli·· tudes. Not\vithstanding these two possible interpretations, it seems more parsimonious to assume that the atritude in such cases is a vail·· able to introspection; yet it may nevertheless influence sornc other automati..:: aspects of information processing thar are picked up via implicit <~ttitude measures. Someone vvho detests spiders may be V/cll a\\'CUe oJ· that fact and may show a rapid, involuntary avoidance response to spiders on an indirect mf'asurc of attitudes .. Should th~H response be labeled an ''implicit" attitude (hat is unavailable tn inuospection, or is it best newed a;;, a spontaueotLS

1 unintentional

consequence of a given attitude?

270 CRITICAL PERSI'ECTIVFS

In ;1 :-.imilar \V;1y, om' could <tsk \:vhcthcr lo\v corrcbtions he tween sclf-rl'pnrtcd ~;nd 1mlire(tly ;p,~essrd attitudec; indiL·,nc rh<H dtL' indircctlv ;1:-.sc:-.~cd <Hl'ltndc 1:-. "irnpliL·it" or unconscious . .:\gam, the ;111:-.\\'L't' t'o thio..; qncsLinn is, not nccr<>saril)'. Corrchtions bct\VL'L'11 sclt­rcportl'd ~md indirccdy assc,.scd attitude:-, would naturally he expected ro lw lovv· if the inJirccdv CJ<>scsscd ,Htitudc IS ll\Konscious [.unless dwre is n:<L'>Oil to JS<;Hilll' <.~spurious relation). Hown·cr~ lu\V corrcLi'·_ tions hcrwccn the t\vo kinJs nf nH.:a~ures CHI <I rise for a lllu!ciwdc ol reasons other t·han a lack of intrO"'PL'ctive <KLcs:-. ((J<lWl"IHJ:-.ki) .Hof.­m;lllll, & \':\/ilbur, 20061. ln fact, there is no\v a large hudy of eYidcnce ~howing .rhat multiple factors determine whether '_'implicit" ~lttitudcs ac,sessed \vith indirrct measures arc related to sdt-rcpuncd "cxplicir" :_tltirudes. It these factors nre controlled, explicit an~l im­p\i..:it arnrudc m<'asun~s typically show quite \uh~t<lnt"i;:!l correlations.

First, it is nften assumed th;H indirC'cr :1ttitude mca.snrcs are lcs~ :dfceled bv individuals·- dclihertHc :1t:tc111pts to control their response!-, th;-\Tl St'lf-1:cport mC<1Surc:.;. Such n1otivationally driven intlnclh . .'CS a.rc p:1rtindarly pronoullet:'d 111 socially sensitive domams where soc1al dcsirabilit\: m;1V ;:d"t"L·ct self reported, hut not indirccdy Jsscs~ed arti rndcs. Co;lSi<;rc.nt with lhis assumption, sever;ll 1..tudic~ h;_lVt: <.h:mon :;rr;-ltcd dwr .;,clt-rcponcd and indirectly asses5ccl <-Ittiwdc)- ro\\:"~rd Lit_-ial minority )~roup~ ~uc highly .:orrcl~11cd when individual llitler CIKCS m the m.otiv,ninn to control pre-judiced reaction' <trc controlled fc,g,, Akrclllli & Fkch,unmer, 200.); Bansc &. Gav-.:ronski~ 2.00.1~ fhuuon & Fat.io, J YY7: Fat.io, .!ackson. Dunton, & \Villia111S~ J '19 S; Cr.:t\','l"Ortsh, Crt'schke~ & lhnsC', 20CH; I----Infnunn~ Cscin:vcndner, t'X Schn1inj 200~). In a simi.L.:n vein~ Nier {2005) drmon . ..;LLHcd th;\t ..:or rcl<1ti<ms hct\VC'Cll sc!f-rcpnrtcd <Uld indin:ctly <l'>'->Csscd att;m(lcs tn­Wdr<..l Afncan Amen~::111s \Vcrc sig11ili(antly higher when partkipanb hc!ic\·ed th:ll· in:\L'CULltC <-idf-repnrts co1dd he detected hy mc~li1S of<\ he detector.

Second, (orn:lations bcn\'CL'll self-reported and indirectly -<1'>·

sc">srd attitudes h;-wc hccn sho\ovn to depend on the degree of cogm­ti'-T deliberJtion. Consistent with th1.:.; assumption~ Hnrc~ck. Scuabis, and Bless {2(HH) demonstr:.ltcd thclt imlividu,Jis vvah <1 strong dl-.posi · tiotJ;ll tCH(,IcnL'V ro cngag_e in ~:ognitive dchhcr~Hion (i.L'·1 bi~h need for c(l~nition; see, CJt:ioppo, Petty, fcinstcm, & Jarvis~ 1996·: shcnvcd ln\'>/LT cnrrclarions hctwccil .self---reported and indirectly asse:sscd attitudes than individuals with ;l low tendency to cnf,<"tge in delibera­tion. ln <l .s1milar vein, <1 mct:-J-;lllalvsis hy Hofmann) Cawron:.;ki, Co..;ch\v(·ndnn:_ Lc

1 and Schmitt (2UOS! ·round thm correlation.:., hc-tw(Tll

sdf-rqwrtcd <~11d HKlirecxly a:-.sesscd <lttitucles gcncrallr incrc~Jsl' <15 a funcnon of the sponmneily ot :---clf-rtports.

What Do We Have to Loarn? 271

Third, self-reports may or may not correspond to indircn rnea­surcs \\'ith regard tn lhc .:-:.)XTifi...: as peer of the attitude that is heinp as­_-;csscd. As such~ corrL'lati<lllS hct\\.'f'C\1 the two an: \<lmctimes rec(~ccd simply because of such nnderlying conc('ptual differences. l-\<Hl.SC a11d

colleagues (200 1 l~ for cxamplr~ demo!lstrated that indirectl}' a:-.se:-.scd attitudes tnvvard homosexuals show higher correLntons \-vith sc!t­rcponcd attitud{'s when the laucr i11vnlvc self- reports nn affective rc spoHscs (e.g.~ "( feel unuJmfort.1hlc nearby tV•/O Illl'll loso.;ing each other""l th,u1 whC'n thq· invol·vc sdi-rcports 011 normative hdic+-. (e.g.~ "Gay men sho11ld not work v;ith children o1· adolesccnu.;"}. I'hesc rl':-.ults were corroborated ill ITohnCJnn

1 Gan .. ronski, and col­

leagues~ (200.1) meta analysis~ .<,howing that ;Htitucks as~l'sscd with t}~c- 1/\T show higher rorrelations with ::1ffcctive as cowparcd to cog· flltiVT sdt··rcpnrr measures. h1 addirion. Hofm~uw ct al. found th~H lovv corrcLuions (JJ1 also he Jnc to mismato:hes in dinlc-n~ior:;;lJin· fsec also No)ek~ 2.005). The IJ\T, for c:x;:lnl:plc, gcnerallv involve<.; <1 ·G~m p:1rison bcnveen tvvo arttrudc objects, thus n:p~c:.;enting rt.'Luivc rCJther than absolute evaluations. TlntS 1 it is not very snrp~-ising that ~:orrebttons between the TA I' ::~nd expliciL .self-reports :1rc generally higher \-'1--'bcn tb: latter involve the s<1mc rdmivc rather than ab:;olut~ ev <-ll u <l tio11 s.

Fin:c~lly 1 implicit ::Htitude mea~ure~ often exhibit lo\\' imcrunl cun­sistencics (e.g., Banse, 1999; Bosso11. Sw:cwn, & PrnnchaktT, 20fl0; Cawron.ski, :Z.fl02-i Olson&_ Lvio, 200.3). ·rhus, their corrcbtion.-" ro self--reported attitudes arc often reduced hy mea-surement error, C:on -'>istcnt" with this as-'umption, Cunningh~u11, Preacher, and lkmaji (;_~f)O I) fou!ld sub~unti:-1! correl..HitJJlS between sdf.rcportrd tlnd indi­ret'Lly as~L's<,cd attitudes when the in1pau of mcv;un'tnC'Ill error w<1s

cotltrollcd wirh latent v~niahlc <lnalyscr,; 1:sce al;;o Ca\vronsk!, 2002; TlofmJnn~ G,nvronski, ct :d .• 2005).

·_Llken together, tlwst' rr~ults suggrst tlut indirn:tly :1s.sesscd and self-reported attitude~ show 'iuh~tanti<-ll correla.tions undt'r Lcrt:Jin

conditions. If the relevtlnt factors :uc controlled, the 1wo kinds of at tltl](Je measurt:·s typically s.hovv quite substantial correlations. from this perspective, it seems unvvarrantcd lo claim that "implicit'' ~ttti­tudcs ctsscs~cd with indire..:t n1easnrcs reflect unconsciou\ attitudes dJ<lt <liT 1111.1'\-';Jilable to intro:-.pcction. Tn be sure, indirc ... :t attiludc measures ustl<llly do nut require introspenion for the :1ssess.rncm of an attitude. I--lov.-Tvcr~ that does not implv lhat the assessed attitude is unavailahlc to introspection, .

fn addition to the as~umptio11 thcH indirectly -tssc.<.:':lf.'d a[titudes lhemselves ,lrl: unconsuous, there is ;molhcr in.tcrpreLJtion of the term unnmscuJUS that has been proposed in the comcxl" of irnplicir

272 CRITICAL PEF1SPECTIVES

attitude mc_·asurcs (Grecnvv·ald & Banaji, 199.5). Rather than refer· ring to the attitude itself, the term unconscious could also refer ro the influence an attitude has on other mental processes. In other \vords, indirccrlv assessed attitudes may be com;cionsly accessible, but they may influence other psychological processes oursick of conscious aw<Heness (see Ni\bett & \Vilson, 1977). Even though research on this qucsti<;n is rather limited, there seems to be at least some evi·· dcncc for this assumption. for exarnple, Ga\vronski and collcc1gucs (2003} demonstrated that people arc sometimes un~nvar~ of h~)W in­dircctlv assessed attitudes influence their interpretation ot amb1guous inforn{ation. ln this stndy_, German participants \"'-:ere asked to form an impression of either a Cerman or a Turkish individual on t~1c basis of evaluativclv ambiguous behavior. Consistent \vith prev1ous re­search (e.g., barlev & Gross, 1983; Duncan, 1976; Dunning & Sherman, J997; Km;da & Sherman-Williams, 1993; Sagar & Schofield, 19RO), German participants evaluated rhe behavior more negatively when the target \vas Turkish than when he \vas Cerman. 1--Io\vevcr, this effect \Vas moderated by indirectly assessed attitudes, such that the target's category membership influenced the interpretation of ambiguous behavior only for participants vvith negative at_titudcsto­\\'ard Turkish people and not for those with neutral attttudes (:ee also Hugenbcrg & Bodenhauscn, 2003). ;\·lore impor_tant, _the n~flu­cnce of indircctlv assessed attitudes on the intcrprctatHm ot amhlgu­ous behavior w;s nnt moderated by participants' motivation to con­trol prejudiced reactions. Instead, motivation to cm1t.rol_ prejudice affected onlv the relation between self-reported and mclirectly as­sessed attitu~lc.<> toward Turkish people, snch that self-reported and indirectlv assessed attitudes were highly correlated for participants low in ,;lotivation, but not for thos~ high in motivation to control prejudice (_see also Akrarni & Ekehammer, 2005; Banse & G;;nvronski, 2003~ Dunton (X Fazio, 1997; Fazio ct al., _1995; Hofmann, Gschwend­ner, & Schmitt, 2005). Self.wportecl attitudes had no impact on the mterpretation of ambiguous behavior. Thus, given that parti..::ipan~s \Vcrc generally able to control the influence of indirectly assessed a:o­tudcs on their interpretation of ambiguous behavior (1.e., partJCI·· pants \VCre not under time prcs'}urc or otherwi~e _cognitively de­pleted)~ these results arc consistent with the assumptwn that people are sometimes unaware of the impact of "implicit'' attitudes on the interpretation of ambiguous behavior. ln other words, whereas indi­rectlv as<;cssed attitudes seem to he conscious in the sense that they are i;"ltrospcctively accessible to self-report, they. st_ill seem to involve an uncon-,:cious component, such that they c1n mtluence other men-

Wl1at Do We Have to Learn? 273

t~l I:ro..::esses outside of conscious awareness. Notwithstanding this fmdmg, much n:ore research is needed to clarify which particular processes arc intlucnced by indirectly assessed attitudes, and which of these influences do or do not occur outside people's conscious awc1reness.

COMMONALITIES AND DIFFERENCES BETWEEN IMPLICIT MEASURES

Original.ly it was a matter of sorne vexation \Vhcn t\\'O different indi­:·ccr measures of a t:articular attitude (e.g., IAT and sequential prim­mg measures of racul attitudes) were found not t·o correlate particu­larly strongly (or even not at all). After all, it is not U!KOllln1on to f!nd that" a Likert scale and a semantic differential sec-de measuring at­tttu~les toward the .same attitude object correlate extremely strongly. 1f dtfferent self·· report measures of attitudes correlate in this manner \vhy sho~Iid indirect measures not likewise correspond? If the startin~ assurnptton is that there is some fixed entity called an "implicit atti·· tude'' that can be measured in a variety of different vvays, then it would tndeed seem disheartening if different measures of this same entity did not correlate \\'ith one another. Howcvet; as we have ar­gued, if the emphasis is shifted to the idea that it is the outcome of a particular measurement procedure that is implicit, and that the im­plicitness or antomaticity of this outcome may depend on specific tcatures of the measurement p.rocedHre, then the important question becomes, which auromatic aspect(s) of attitude activation can be cap­tured by a particular task? Is it spontaneous approach/avoidance ten­dendc.s or the activation of evaluative associations in mcmorv {Neumann, Hii!senbcck! & Scibt, 2004)? ls it the acrivation of concq;­tual or cva~u~tivc kno\vledgc in memory (Wlittcnbrink, Judd, & Park, 200 I)? Or rs rt something else? ,'vJcasures should be expected to correlate strongly only to the extent that they each tap similar information­processing consequences of clttirude activation. Different self-report me~surcs_ typ1cal!y do assess the very same aspect of processing (i.e., dehberatJve evaluation of the attitude object), \vhile different indirect measures arc likely to assess a broader range of disparate processes. Indeed, recent research has begun to examine these differences, lead­ing toward a typology of indirect attitude mcasUITs that will provide _resea_rchers with a rich arsenal to clrav·/ from in attacking the mvstcr·· trs ot the black box. Ultirnalely, the target and range Jppropriai·e fnr each of these research tools need to be specified.

274 CRITICAL F'EF:Sf'ECTIVES

Scycral proposah have been offered to describe and account for the different cmph:1~:es of the different indirect clttiwde. mea<;urcs. Olson ;md Fazio (20011, for example, proposed that wadable mea­sures differ in the extent tn whH.:h they tap general :::atcgorical re­sponse~; versus cxempL.1r~based responses. The attitude vcrsi01~ of t·hc IAT, for example, requires respondents tn categonze 3 set of cxcm-­plan; (e.g., names or face~) in terms of their mcmhershi p in rhe ~el<> vant :1nitudinal ccltcgory {e.g., a ra(·ial group). As. sucL the cognlt!VC'

focus of the task is on tht.· category. This ;.iru~Ition <.:all he contrasted \vith a common version of the <-1ffco..::tJvc priming task. which exam­ines the dfLctS of individutll exemplars {e.g., fa...:c.s) on dw_ processing of :-.uhscqncntly cncoun tercel words. \Xlhile the GHegory to \Vhich th.c {_'xemphH hdon~s is expected to influence thio:., processing, t~l.~ CO[f,lll ..

rive focus of the task is not on the category pn sc. Thus., d1tterenccs in the vvay excmphrs arc evaluated, vcrsLtS evaluation of ~he.categ~)fy :1s a whole, underlie the low corrcspondC'ncc between aftecttve penn­in~ rneasur<·s and the TAT, according to Olson and hl'LO. They tound that by modifying the affective prinnng task .in <1 \\:·ay tlwl ~ncour .. Jgctl cHcgonzation of the exemplars, corrdat1ons wtth the L\f \~'ere greatly enhanced. i\ilitchc\L Nosek, anJ Banaji (2CHUl proy1dcd further evidence thar the tAT docs indeed operate ar rhc l~~vcl of cate­gorical evaluations. Jn dwir smdics, the s,1mc set_ of C'.:t.·mpla~·s {e.g.~ African American <.'l.thletes) prorluccd ev1dence ot both nt'gati\'C anJ positive a:-:.stJCi<ltions, depending on whether they were (~l1:egorizcd hy race or hy occupation. These stuJics make it cleJ.r that a locus on the category per sc, r:?~tlwr than nn individual exemplars c·f the category, can produce marked shifts in impllcit anituJc measure~, . .

De Houwer (201)3) has proposed another usetlll d1stmcnon he­tween variow, indirect measures. A b~1sic structural property ot lll<-lny

i11direcl mcasurcrm'Ilt proccdu!'cs is that they involve examining whrrhcr Information processing is facilitated (speeded) m impaired (slowed:) by rhe pre-~.cntation of an attitude ohjcct (s~·e Lane et <ll., Chapter \ this volume; \X!ittenhrink, Chapter 2, tlus volume). T_hc crucial question revolves around whether or not the pro~esses tn~· gered by the activation of the attitude ar-e cnmpati blc or~ m~ompatJ­hle \Vith other processing requirements. De Houwer (2003.' argued rhat there arc at least two general kinds of compatibility {ser also Kornblum, 1--Iashroucq~. & o~man, 1990) and that different. tasks varv in the extent to which they Jravv upon one or th( other of these tw~ phenomena. Om~ of these, respo11se comt}(l.t;bility, ~as to d? \\•ith how the (natural or induced) response tendenctes asscKtatcd \Vtth the attitude object map onto the response requirements of the task (sec

What Do We Ha,;e to Learn? 27!)

al':so Stroop, 1935). This tssue is important in the TAT, for cxampk (see Lane et al., Chapter 3, this \·olumc). In a race IAT, African Amer­ican cxe1nplars arr supposed to he categorized a~, "Black.'' ·rhus. when :1n Africnn American face (or name) appc;us on the screen: there v/dl be a (task-appropriate) tendency to respond hy pressing the "Bbck·' button. 1-·Io\:vcver, for ;1 respondent who h;:trbors negarivc a~,.socictticJIL\ tm.vard African Americaus, there may .. dso he a sponta .. ncous response tendency unvard the "Unpleasant'' button. For the sp-calkd compatible trial hiock, the ''Unplcasa11t" button and the "Black" button are one ctnd the 1oame, so the two response tendencies arc compatible and performance should be facilit<lted hy the LH.:t of their coinciding. For the incornpcHiblc tTial block., however, the "Un. plcasant't rcspo.1sc i~ uot the same as the "Black'·' response, so these two response l<'nJencies will be in conflict. As such, performance should !-)utfer (i.e., slower reactions\ more errors:~.

A very ::-;imilar logic Js at \'Vork in the affective priming task pro­posed hy "Fazio and colleagues (Fazio ct a!., 199S; c.;ce \X!ittcnhrink, Chapter 2, this volurnc). ln rhi~; t:1sk, the cxempbr prime (e.g., a BLack faccJ may elk.it a spontancou~ evaluative response tendency. /n(·spcctive of this tendency, the GJsk requirement is to categorize the target \'\:'Ords presented after the primes in terms of their va!cnc('. for trials in \Vhich t:1e- spontaneous evaluative rrsponsc to the prime co­iuudcs with the valence of the urget word, performance should he facilitated bcc:1.use both the prime and the target" potentiatt' the same rc~;ponse. ror trials in which the spontaneous evaluative response to the prime is different frorn the valence of the target word, perfor­mance -,:hould he iuhihitcd because of the incompatibility of the re­Sf~onsc rendencics e licitcd by prime and ungct. From the standpoint ot this structural characteristic, the Ii\T and t!w affective priming tc1sk are qnitc similar.

The other ftmn of compatibility that De Houwcr (200l) de­scribed is $limulus comfJatibihty. The fundamental issue here con­::erns how semantically similar two sets of stimuli arc. The lexical decision task used by i.Vittcnbrink, Judd, and Park ( 1 997) is a good •:xamplc of an indirect measure that relies on stimulns compatil~ility :sec \Xlittenbrink. Chapter 2, this volume). In this rask, a prime stim·­ulLis (e.g.~ .an African American face) precedes a standard lexical deci­sion task, in which respondents musl simply categorize a lertcr string as a word or non word. The words used as target stimuli in the lcxicJl decision task var~' jn their stereotypical relation to the prime stimulus (e.g., h(Jstile vs. ji·iendly as target 1.vords that are consistent vs. incon­sistent with the nc~gativc stereotype of African Americans). For in-

276 CRITICAL PERSPECTIVES

stance, if an African American fc1cc activates stcn·otype-consistent concepts in memory, this reaction will either coincide \Vith or contra­dict the semantic notion of the target 1vord. \'\:'hen the semantic con­notations of the rrimc and target word coincide, responses should be faster, but thcv should he slo1ver \'vhen thev are contradictory. It is impon<~nt to 1~ote th::lt as meaningtul targct,\vords require resPonses \Vith the same- key irrespective of their semantic connotation, (in)corn­p~nibillry in the lexical decision task i~ defined on the level of stirnulus features r<-Hhcr than on the level of response teudencics; any response rendency that is created by rhc pnme vvi\1 he lrrelev<tnt to the task­required resronsc~ \Vhich is "\vord" irrespective of \vhether the target word is stereotype consistent nr stereotype inconsistent'.

De Hou\.vcr\; (20(n} structural analysis leads to the interesting precli;:tinn that pcrform~HlCC on the IAT Jnd the affective priming task may look more .simil<!r to one another than performance on affective. versus semantic priin.ing tJsks. Indeed, this prediction "\vas hornr out in 8 series of studies bv Gawnmski and Bndcnhauscn (2005). These studic& also documcJ~tcd another \Vay in which incli­rect mea5-ures differ from one another~in terms of their sensiti·vity to n1Ctth.::ognitive inferences. ln research on self-reported auiludes) there is extensive evidence that indi ... _,'iduals use mctacognitivc infor­mation to inform their _judgments. In this research, tvvu types of in­fluences on judgment are contrasted: the content of activated knowledge vcrsns. the experienced ease with vvhich this knowledge \Vas re­trieved (for a revinv, sec Schwarz, Bless, \\/;:inke, & \Xlinkielman, 2003). ln a typical study~ p<lrticipants are asked to generate a cer­tain number of instancco;; fjtting a parti.l:ular L~atcgory (e.g., likeab.Je Canadi,1n.s}. For instance, the more likeable Canadians one calls to rnind, the morc one's activated knowledge is consistent \Vith a posi­tive evaluation of CanaJians. Ho\vevcc because it is typically harder to think of more instc1nccs than to think of fc\ver, experienc­ing the difficulty of trying to think of many likeable Cauadians Old\' imply that thrrc are not very many of them, consistent with a mo,re negative evaluation of Canadians. [n the realm o.f self­reported 'judgments~ experienced ease is commonly ohservcd ro dc­tct:mine the tone of overall judgments} unless its diagnostkity is explicitly discrcditc•d.

Gawronski and Bodenhausen (200.5) extended this research to the: realm of in(hrcd ~Htitude measures. A common assumption about indirect mea.-.ure.s i.s rl1<1t they simply tJp the activation level of asso·· ciation~ in memory. From this standpoint~ tht more exemplars of a cd"tair1 type one c,.;Us to mind~ the more one's implicit evalnation \vill rnoYe in the Jirecti.on of those exempbrs, hascd on the activated

VI/hat Do We Have !o Learn? 277

knowledge. W'hcthcr or not it \vas ea.sy or difficult to gcncr--:Itc the- ex­emplars should be irrelevant. Thlls, perfornunl'e on a race lAT should reflect more negative evaluations afrer panic! pants have gcn­~rat~d many dislikcable African /\.mericans, as comp;ued to geiK~rar~ mg JUSt ,1. few dislikeablc .African A.mericans. ln clctualitv hn\vcvcr re-sults were :iust the opposite. IAT scores reflected the .l;~ttern tha·; \_vould be expected if retrieval experiences were guiding responses. Subsequent studies replicated this pattern and sho\'ved that it also generalized to .an affective priming task as \\'ell. Considered from the srandp(~it~t. of De 1--Ioll\ver's (2003) distinction bet\vccn response co~npauhdtty and stimulus (·ompatibilit)~ we reasoned that a direct retlenion ol activated kno\vlcdge would he most likclv in ca~ks chat involve stimulus compatihi!ity. Again, this prediction \~/as borne out. In a direct comparison of two 5-Cquenrial priming tasks Js.sessing im·· plicit prciudice against African An1f'ricans, implicit prejudlce de­pended on the experienced ca:sc of retrieving dislikeable Afncan Amcncans \.Vhcn the ta.sk included an evaluative decision task (i.e .. response compatibility). In contrast, implicit prejudice \\cas i1~flu~ enced by the overaH amount of 3(Tivated exemplars \Vhen the task in­cluded a lexical decision task (i.e.) stimulus compatihilitv). . Further evidence for the differential roie of stimulus ~mnpatibil­Ity and response compJtibility in !ndirecc attitude measures 1s implied by research that invesngated the impact of multiple primes on se­quential priming effeCts. Ralota and Paul ( 1996), for e-xample, found that t\vo sequentially presented prime stimuli resulted in addtrlvc effe-cts in a typical semantic priming paradi~m using a lexical deci­s_ion t~sk_. This result is ..::onsist~nt \Vith an interpretation of sequcn-· t1al pnmmg effects in terms of" spreading activ<'ltion (sec Collins & Loftus, 1975), implying that increasing scirnulation should increase the acti·vation level of corresponding a-~sociations in memory. Inter­estingly, such additive effects turn in!o contrast effects when the stim­ulus compmibility structure implied in the lexical decision task is changed inro a rcspono;;e compatibility srructurc. Ca\vronski, Deut:-;.ch, ~1nd_ Sc_idel (200.5), for example~ found that affective priming effect~ m }'aZio and colleagues' {199Sj paradigm were nwte pronounced \Vhcn tht.· CvaJuative rrimc Stimulus \VaS preceded by J context prime of the Of"Jposite valence. I Iowcver, affective priming effects vv-ere less pronounced \vhen th(' prime stimulus \Vas preceded hv a context prime of the same vatence. This findmg stands in contra~l" to the as··· sumption that affective priming effects in evaluative decision tasks have their roots in rlw same .spreadmg acti.v<ltion mechanisms that are responsii.Jle for semantic priming effects in lexical dcci-:;ion tasks (sec, e.g., Fazio, Sanbonmatsu, Powell, & Kardcs, 1986; Hermans,

278 CRITICAL f'EFISPECTIVES

De Houwer, &. Eelcn, 1994 ). Hcnvevcr. it is consistent \Vith claim . ., that affective priming effects arc driven by a response compatibihty mechanism (sec, e.g., De Hom.ver, Hcrn1an:-, Rothermund, & W'cn·· nirJ., 2002; Klauer, RofSnagel, & l\--1usch, 1997; Klinger, Burton~ & Pitts, 2.000; \Ventura, _1999), Implying that Jffcctivc priming effects shodd be influenced hy particip::tnts' ab·dity to ignore rask-irrelevant (ealure::,:l in this case the valence nf 1"he prime (see n~~sner & StolL. J 999; Bcsncr, Stolz. ($( f~outilier, I 997). Applied to liH: present quc<.,­t:ion. the vJlcncc of the second prime l1L1Y be more salient, and rhn:s more difficult to ignore, when it is evaluat:ivcly im..:on~Istent \\1ith the valence o( the firsl prime. l lownrn, the valenn' of the_ serond prime may be less :-.cdicnt, ::llld thus easier tn ignore, whC'n it is evaluativdy coi~~istcut vvith the vMlcncc nf the first prime (Ga\vrom;ki, [h:utsch~ & Strack, 2005 l-

In sllmmary, the available evidence indicates that the w1dcspread cqu~' tion of rhc outconH' of diifercnt measurement pro<.:cdurc.s \vith :1

single ··'implicit attitude" construct :.s quite problematic. Ditfcrcnt mcJ:mrcs arc char~.'l.ctcrized hy very difJcrent task structur('s (e.g., cxemp\.:1r .. vs. c.<ltegory-related rc~pOilSCS.~ stimulus vs. response c<:m··

p<ltibility). As such, measures that rn<ly appear similar on the surtace can produce very diffcrcm results, depending on their underlying task structure. · l'hus, Jluturc rc~c<lrch employing in.:il r-E'ct ani tude me~1:;ures are well ;J.dviscJ to take strul"turr~l diffen~nces bet\vecn task:; into account in order to avoid theoretical misinterpretations of the obtained results.

PHYSIOLOGICAL CORRELATES

So far, our Ji,..,cussion has primanly addressed measur·cs that Jrc haseJ on response la-:encies such as the IAT (sec Lane cr a!., Chapter 3~ this volume} or sequential priming ~:asks (see \X1iUenhrir1k, Chapter 2, this volume). Even though the vast number of srudics using these mca~mres already provide important insights inw their underlying rncch::ntlsms, rtnd thus into the nature of the construct~< assessed \Vith these tclsks, many issues arc still unresolved. some of which \vere dis­cussed in the preceding sections of this chapter. A fruidui comple­ment in this endeavor could be the use of physiological measures. (sec Ito Sc. Cacioppo, ChapterS, this volume). for instance, by searching for physiological correlates of the performance in indirect 8ttltude measures, physiologi•.::al mea~ures could provide further insights into both the connnonalirics and differences hct\vcen different measures or task properties. In fact, several studies have already :investigated

What Do We Hav·S! to learn? 279

rhe relation hct\vecn implicit attitude measnrcs and _physiological re· spo~1scs. One of the f1rst studies in thi.s area, for cx<Hnple, found that pcrforrncmce on the race J.Yf was significantly rcLHed to an1ygdala acri\'ation 111 response to Black faces (Phelps eta!., 2000). In a silllilar vein, Chce, Sriram, Soon, and Lee (2000) demonstrated that pcrfor­mcmcc on a ilovvcr---insccto;; lAT \vas signif.iccHitly related to activ;Hion in the left dor.-.olatt:Tal prefront.1\ cortex. Milne and Crafman (2001) cxtcnckd Lhcse findings by showing that lesions in the vcntromccLal prefrontal cortex climtn<HC Hnp!icit gender stcreotyring in the lAT. /\ tn()re complex p.1ttcrn in d1e relation between implicit attitude mc<l­sitres a11d phy!lJIJiogical response<; was (lemonstrated hy Richeson and colleagues (_~003j. ln their srucly, r;Ke bias in the IAT prcJictcrl Jctivatinn in the right dorsolatna.l prefrontal cort(')i'- in response to Black hces, and this acriv~ttion mediated the rcbtion bcnvccn r<h.'C

bias in the IAT and pcrfonn;mcc impairment~ in the Stroop task re~ sulnng from interattions with a P,lack confederate. finallv, Wheeler and rtske (2005) h~W(' shown that bDth <lUtornatiC stcreot)cpc actiV.J­tion <1nd arnygdc1la activation in response to Black Lu.:cs w;.ls signifi­cantly reduced \Vhcn partiL~ir<uHs were askcJ to fn:.:u.'-: on a nonr<.Ki.:.d category (see ,,]so Mitchell ct ,,j __ 2003).

Even though these findings provide first insiv,hrs into how irn­pllcit ~Htitudc measures arc reb ted to activitv in different arC<lS of th(• hrain, the origin :-md the nature 01: the ohtatned relations remain oh·: s....:ure as long as r',1ey cannot be matched to the specific psychological mechanisms undnlying indirect attitude me~lSures .. For example, the ~1vailahle dam on the IAT raise the question of \vhy IAT perform~nce is rclared to horh amygdah activation (Phelps et al., ~WOO) and acti­-vation in the dorsolat('ral prefrontal cortex (Chcc et ;:JI., 2000: Richeson et al., 280.3), Civen that amygJ<~la activarion reflects <lilt<>

matically activated negativity, whereas activcHion in the dorsolatcrc1l r;refrontal cortex reflects deliberate inhibition, these finding~ suggest that IAT performance is influenced by both automatic and controlled processes. This as-mmption rs conslstent vvith Conrey and colleague.'.' (2005) claim that performance on indit-cl:t attitude measures iS not procc~s pure, but influenccJ by multiple processes. In t:ern1s of Conrey and colle.Jgttcs' Quad··i\1odcl, amygdala activario11 may be rclal"cd to the associative hi as component, whereas activation in the dorsolateral prdronta[ cortex may be related to the comportC'IH reflecting success J.t overcoming associative bias. Hnwcver, the map·· p'm? of the processes proposed by Conrey ct a!. and specific physio·· logicll correlates is still hypothetical at this stage. Tbus, future research on the relation between indirect attitude measures and phys· icloglcai correlate; could provide an even stronger contribution by

280 CRITICAL PERSPECTIVES

considering rhe specific mechanism~ under:ying different kinds of

measures.

WHAT DO WE HAVE TO LEARN?

C~iven the explosion of 1-cscJn.:h on imfJlicit attitudes and dutoowtk e1Nluations {see \Vittcnhrink, Cluptcr 2~ thi~; volume, Figure 2.1; sec also 1v1uscb &: Klauer, 2003i, it is probably nor an overstatement to claim that the development of indirect atrltude measures brought about, if not a scientific revolution, then at lea">t a substantial reori~ entation of priorities and pnspcctivcs. Certainly, th1s change has left us not only 'N.Ith important new insights intu the psychology of eval~ uation h11t al .... o \vith some unrcsolvcJ contrm·crsies. These contro­versies' raisr :1. lot of chalknF:ing quco.;tions thM call for answers. Sor~1c of the qt1cstions \VC consider to he particularly important .:nc dis­cussed in this ch;lprr-r. These questions concern (l) the generJI nature of artitudes as constructions versus stable representations, (2) the "implicitness" of the constructs assessed by indirect a.ttit~Jde mea· _.:;urcs

1 !'3) the parricuL.u processes that <UC mfluenced by mdJrectly as·

scsscd ;.lttitudes ~mel whether dlesc influences do or do not occur out· side of conscious awareness., (4) commonalities and differences between measures as a function of objective tas.k characteristics .. and {S) physiological correlate~ of the processes underlying indirect <.Uti­

rude mcast: ro .. Kotwithsrcmding the importance of these issues, the question

that \Ve reg~ud as the mos.r lmporrant one h.-1~; ~1ot yet been addressed. This question is related to rhe l~ck of connection betw:-en large·-scale theories on the determinants ni' judgments <'.nd behaviOr, and small·· c.;c.alc tbeoric" about \~'hat exactly indirect J.ttiHtde measures assess, hovv thcv work and ho\v they arc related to each other. The problem address;d by tl;is question is not so much that we would la_ck citi:cr Luge-scale theories on human behavior or small·:scale theones on m· direct ilttit·Jdc mea::..urcs. Rather, the probi·'2tn IS that the concepts rroposcd by these t\.vo das,>;cs of theories Jo not map onto ea_ch other, such tha( the conceptual terminology of one class of -rheones could be unatnbiguously translated inw the renninology of the other. For instance, when talking about indirect attitude measures\ large­~cale thcor:.cs typically employ c1bstract theoretical corlCc~ts, sw.:h _as "implicit attitudes" (e.g., \Xiilson ct a\., 200C) or "evaluat~vc .~ssocta­[ions" (e.g., Strack & Deutsch, 2004\. Smal\-:;.cale theones, m con·· trast usuallv refer to concrete operational concepts, such as "re­spon,se comPatibility" (De Hon\ver1 2003) or "salience asymmetries'~

What Do We Have to Learn? 281

(Rothermund & Wcntura, 20(!4). The chaJengc that researchers bee right now i~~ to find ways to combine the t\vo <1pproachcs, su..::h rho1t the relevant concepts can be mapped onto each other and c:. coherent nomological network em h(' constructed. A first step in this Jircction rnay be Conrey and ..:olkagues' (2005) Qm](l· ;\-lodcL 'This model nor only covers some nf lhc theoretical constTIJLts lypi~..:ally proposed by large ~cale-thcories (e.g., automatic attitude acti\'ation, cognitive control), but :1lso addresses the mechanisms tmderlying <1 specific type of indirect ~Htitude mc;:Jsures (i.e., m(~asurcs b._lscd on response compatibility).

A second question that is directly rebred to this first o:.1c is how physiologict=d processes map onto th~ constructs proposed by large­and srnal:!--sc:Jlc theories. As outlined abO\..:c~ there is an accmnubtinv; hody of research that investigated the reLr:ion lx:tween implicit ·cltti­tude measures and phy.siological correlates (e.g., Chee ct a!.) 2000; Milne & Gralman, 2001; Phelps ct al., 2000:; Richeson ct al" 200l; Vanman 1 S:altz, :\Jathan, Ck: \X-'arrcn, 2004; \X:'hccler & Fiske, 2()05). However, compared \Vith the vast number of studies Oil rhc nneclw­nisms underlying itJJirect attitude nH:asui\~S 1 the availal)lc evidence regarding physiological '--:orrclates is still very limited. \Xh· believe thrtr tlu: proposed mapping of brge··scale and ~;n~; .. tll-~<:ale theories cnulcl be substantia!ly enriched i.f this endeavor IS associared vvith a map­ping to phys.iological corr.::-lates. Such a multifocus approach ·would thns include (1) largc··scalc theoric" of affect, c:ognjtion, and hchav· ior, (2) small-scale theories concerning the mechanisms underlying indirect attitude measures, and (3) theories of phy,..,iologic.ll function­ing. The investigation and integration of this fLdl range of conceptual issues (Onstitutes ail ambitious agcnd~l for rhe next gencrati:)n of IT­

search on "implicit attitudes,"

ACKNOWLEDGMENTS

Preparation of this chapter was supported by grams from the CanJcla Research Chairs Progr:un, the Social Sciences and Humanitie$ Re:scarch Council of Canada, and the Academic Development Fund of the University of \\!estern Onwrio to Berrram Ga,vronski.

REFERENCES

Akr<mli, N .. & Ekehammn.r, H. (2005j. Tlw as:.ociatiotl bet\veen implicit J.!ld ex­plicit prejudice: The moderating role of motiv<>rion to control prejudi;,:ed re­actions. Scaudin,wian ]mana! o( Psy(hology, 46, 361-366.

282 CRITICAL PE"SPECliVES

ga]ota, D. A., N 1\wl, S. T. ( 1996i. SummcHion uf <lctivation: Evidence from m.ultiplc pnmt;s th<>t converge and divrrp' within scmarHK ln('tnory.]om··

1!>1! n( hxfwrimental Psychn{,)g)': l.e .. rning, A·1emiH)\ mul Cogmtiou, 22, 82/-845.

Hansc, R. (! 999). Automatic l'valuati,">H of ~elf and significrtm mhcrs: AffectiV(" p1·iming in close rl'latinnship~. }f•rtrn,11 of Socia{ .md l'erso.'la/l~elatinw-:hi[J~,

16, ~03-~21. lbn•.e, R., & c;:n,nonski, B. (2001). J)j(' Skai<J Motivnrion 111 vornrtellsfrc!cm

Verhallcn: ~kalenc.:gcnschaftcn und Vc1lidicrung iTlll' s<:.·alt-: 1nntivation to act wirhout prcjm!Jcc: Psychometric propertit:-~ :Jnd v:-1hd1ty). Di<If!.110~·ti.:a. 49,4-13.

Bam.c, H., Seis1_\ J,, & /(·rh1-'"• N. (100 l ). Implicit ,untudcs wwanb lwmo-;e;..:twl­

ity: Rdiahility, validity, and controllability of the TAT. Zcitschri(t (i:n /·.xpcrj. mente//(' Psycl,nlogte, -+8, t 45-l60.

]);ugh, .1. A. ( 1994). The four hor;;enwn ()f ;lL totn:Jticily: Av.·areil·~ss. intcmion, {:ffi­lWncr, :m~l control in .'>oci:ll cognition. In R, S. \'1'/ycr & T. 1<. Srull :Fds .. l. Han!ilmuk u( sod,d mg11itioo (pp. 1-401. 1 lill'>dnlf', NJ: Lrlh:mm.

Bescer, 0., & Stoll'., .f. A. ( 1909). Unconscicms!y comrol!nl procd-sing: The Stroop dkn recon:-.idcrcd. Psyd}()nomic Hul/etifl and 1\.cuicu·, 6, 449--4.'1 5.

lksr.n, D., Stolz, J A.,.~ H~)tJtilter, C. ( 19q7). The Stroop cftccr :1nd the rnyd1 nf :luUntnt icit v. Psyd.'oi,'OJntc Buffet in and Hr·1mcw. '', 221---2.25

HL1ir, L V. (2002,1 ~ Th~ mal b1 biliry of autnnnt i'-- stt>re( '' ypt·:. and prcjudi..:c. Pcrsnn­altty rmd \'od~d P.'ycholo!{y H.J'I/tew, 6. 242-·-261.

Bos•.on. J. I<., S\v;-mn, \\1. B., &.l'ennehaku, T. \V. (2000). Stalking the p('dcct me.1· '<un· of implicit se\f .. esreem: The blind men and rlw cleph:lnt n:vi<,ited? joun-r.;l uf ferson<~hty m:d Soda! Psychology. 7'J, (d 1-h··U.

:lC;opptl,J. T.~ Petty, R. E., Frinsu·in, J. A .. & Jarvi.~, H. \V. ( 1996:1. DispositiPJul diH'crcnccs in co~nitive motivai'Hm: Tlte hfe anJ limes ol individuals YaryinJ!. in the need for cnr;nitimt. /'sychologiu;/ Bu/letm, 119, L97-2.5 3.

Clw:·, J'v\. l ., ~rir:1m, N ... ~uo11, C. S., & Lee, K. ;\:1. (2000). Dor.<,:)Lnc;r:1l prefrontal

cortex :md (he nnplicit association of conlTpt~ and :lttributcc... Nf.'iii'(Ji·ct•nrt

(iH 1\apid Conm:uwcJtion nf Ne1troscienu: !?.r-·::.r-.·arch. II, 135 140. C :ol!ins, :-\. _M., & Luftr.s, E. F. (1975). :\ <,prc:;:ding·acivat;on theory of semantic

processing. Psydmlogicaf Re1'iew, 82, 407---42l-L C0nrc-y, F. R.,- Sherman, J. \XI., Cawronski, B., f-Iugenherg, ::<., & Groom, C.

( ?.005). Sl'parating rnulriple proces.ses 1 n implicit social co,eniriion: The Quad­l'vlodel ot implicit tas·h: performance. Jounwt of Personal1ty mul Social JJs~'­cho!ogy, 89, 469--AS/'.

C:lu·na:gharn, \V. A., Preacher, K. J., & Banaji, tvl. R. i200 I). ln111licit attitude meJ: ~urernent: Consistencv, stability, and covergent validtt~·- Ps:ychulogiarl Sci· a•ce, 12, 16.1-170.

Dar!e~\ J. ivf., & Gro.:;s, P. J--l. (198J). A hypothesJs·confirming btas in Labeling c-f­fcns. Journal of Persoualitv and Social Psychology, 44, 20-.<;_i.

De 1- Iomve;~, J. (2003 ). 1\ structu~al analy:::ois of indirect mea:-.ures o( attimde'>. Jn J. i\-l.mch & K. C. Klau1.~r (Edc;.), The f'syc!Jology n( trJa!uation: Affectil'e pro­cc,~ses in cognition and emotion (pp. 219-244). Mah\vah) NJ Erlbaum.

Del loll\-ver, J. 12006). \'\ih;H arc implicit measures and \vhy are v;c using thnD? In

What Do 'Ne HavE1 to Learn? 2133

R. \Xi. \\.'1cn. (X A, \Xi. Stacy (Ed.<,.), The h,mdbnok o(im(.'ficit mg11itioH~.md (jd­dtclton (pp. 11-.~H). Thou:"and 0:1k<;, CA: Sage,

Dr Houv-.'er, J, I lt:rrnans, D., Rothcrnmnd, K., & \\i'entura, D. (2002). Afh·ctive ~rin11ng of semantic ~·;ucgr1ri~ation responses. Cognit!on tt1ul hJiot 1011, J 6, h4.1-466.

Dovidio, J. f~., IC-lw,1~ami, 1<., & Beach, K. R. (200J). Impli(it and explicit atti­!udcs: l·.xamrn.ttton of the rclauonship betwt't:n mca~urTs of inlcrgn_Hlp h1;.1:-;. In R. H.rovm & S. I. Gaenner (Eds.), nJackwell handhnok o/socw! 1'-~yd;of­(•g)•: inlaf!.rOtt/1 fnoas.;cs (pp. J 7 ~: .. I '-J7). j'vbldt'n, tVlA: Hb(·hvell.

l)o\'idio,J, F, l(avv;Hanli, f\.,.Jnhnson, C., johmon, B., & I lcnvanL /\. ( 1997). Un

the _nature ,o,f p~·c-judic(': AutnrHati(' and controlled prtkTSSes. jounh71 of f-.x­f/1-'l'l!ll(!l!!a! ,\onal P:>yrhninf!,)', 31, SJCI--510.

Dunc1n, B. L._(197fi). Ditferenti<ll pcru~ption and <lttribution of intergroup vi->

!cnce: I cstmg the I DWf'f I ~nuts or SH'f<:( ltypmg of Blacks. lou mal of Pci'Simd­lfy mid Suoal Jls-.,dmfogy, 34, ')(_)(J-5{JS.

Durrn:ng, D.~ & Sherman, D. 1\. ( 1997)_ Stereotypes and uci· in(crcnce.Jozmw/ 1)j

f't'rsmra{iry <-1fl[.l .\c,d;~l Psyrh(lfo,~y. 7 i, 459-4 ·.71.

l)unton, B.(,., & hnio, R.J--T. (lqlJ7). i\n individual diffnence mea<>urc of nwttv:1

-

1'ion to control PH'! udi,:ed I'C<Kt ion<,. Personality and So!·ial Psychulo~"'' Bulle-!m, 23, 31-h .. )16, '~

l':1zio, R. f-l. ('1995). Attitude,.. as objcct--·endnntion associ;Hinrts: Detcrmin<~nts. C('l1Se~u.enccs, ,·tnJ correlates ot' attitude ;Jcccssihility. In R. F. Petty & J 1\~ 1\rn.':dllCk {hi<>. I, A.rtrludr srreng!h (pp. 2·P .. --282}. A;b]l\\';1h, NJ: Lrlbau111.

hni•D, IC fl., Jack-.:oiJ, J. R., Dnnton. B. C., & \Xldliams) C. [. r·I99S). v,1 ri<lbil­ity in :1utomauc :Ktiva.tion ;:tc; an unohtrmive nH:;!stH·e ni- racial atti1udc.s: ,\ bona fide pi~.,rlinc? [o11nwl (_)f Pcrsn11ality <111rl Social Psycho/of!..\', (JI.J,

1013~1027. '-

Evi!J, H. It) Sanhonma1csu, D. M., Powell. ?vL C., & 1\arde:-;. 1·. R. (I YXIl). On the au\omanc acrivoltiun of .tttitudes. _lounzal of lJt•rsonall/1.' and Sodii! J'sv,·:ho.'~ o;.;y, SO, 7-29-L'·X. - -

Horack~ A., Scarahi."., i\1., & Bles~, 1!. (2r_)(J I). When do <l::.soci:-tliom nutter? ·1 he

nsc of atllomauc <-l~sociaticms towards ethnic groups i:1 person judgn-wnt~ .. fuumaf of 1·: ..... -perimenfcll Sod'ai flsyl;ho/ogy, 37, 51 X·-524. ~

Cctwronski, B. (20021. \~/bat does rhe frnplici[ Association Tf:st meetsmc? A test of tlw cmwergent and discrimin:-Hlt validity of prejl!dice-rda1wl IATs. Expcri­mrntvll F'sycfm/tJgy, 49, 171-180.

Cawn:nski, B:, .& B,n,:lcnhausen, C. V. (2I)OS). Accessibility eHects on implicit so­cJal co~nttton: 1 he role of knmvledge :-tctivation versus retrieval experiences.

.. loumwl o.f Pcrsrmality ahd Soctal P.l)!tholoRJ'· 89, 672~6H.5. (

1awronc;;ki, ~-, Dcut.>;~~h, R., & St'idcl, 0. (2005). Contextual infl11cnces on implici-"

evaiuJ.tton: Addttt vc versus contr<Istive effects of ev~1luativc context stimulr in r1hcctive primin.~. Personality and So(ial Psy£"huluxy Bulletin, 31 1226 ... 1236. . •

Gawronski, B., Deutsch, R .. & 5track, F. (:?.00.5'). Approach/avoidance-related mo· tor actions and the proce<;sing of afft;.~·ti\"C stimuli: Incongruencr effects in 80.

t'omatic attention aUocatton. Social Cognitim1, 23, 182--203.

Gawron.5ki, B .. (Jc<;chke, D., & Banse, R. {2003). Implicit bias in impression for-

284 CRITICAL PEHSF'ECTIVES

m<ltion: Associatimts influence the ccnstrual of in~iividuar::ing infnnn;:~tHm. htrofn:an jol/rna! of Social Pn•chnlnp•, 33, S73-5H9.

G<Iwron~ki, B., l-lofmar.n, \Xi., & \X-'ilbut, C.J. (2006L Are "implicit'' attitudes un­conscious? Consciousness an,l ( .·ognitinn. 1 S. 4 k:5---499.

(i-n>::nwald~ A. C., & B<>naii, M. R. ( 199.51. l1nphcit s.oci<ll CO:!;nirinn: :\rtitudes,

sclf-cstc:C'm, and stt:re•ltypes. PsycJ>olc-gzca/ Reuretu, 102,4-2-:-:. ! lcrmans, D., De I louwer, J., Eelen, P. {1994). The dfectivc priming (.ffect: ;\Hto

matic activation of C\'clluative information in memory. CogH/tiun <md t.:mn·

tinn, S, 515--533. I-fnfin:mn, W'., Gawrnn·,;ki, B., Gschwendnn, T., I.e, 1--L, &:. Schmitt. _;vi. (2005). A

meta-;_1tu!vsis on d1{; -.:orre1ation iwtwt~en ~he lmplictt Asso..:iation Test and explicit ;;rit-rept!rt m('asure. Pns(ma!tty and Social Psyd•.Jiogy Jlulletin, 3 l. 1 :;h9-13G'l.

1 Jofrn,mn, \V., ()sc!nvendner, T., & Schmitt~ :-vt (2005). On imp! kit----explicit con ~i:;rencv: The moderating rok of indivt~lual diffcrcnn_·s 1n aw:uene.c,.s ~1nd ad­ju<>tmc;lt. Eumpe,~·n Journal of Pasrmdity, /(), L'l---4-Sl.

I-lugcnl-wrg. K .• &. Bodt-:nh:w::-.en, CT. V. (2003 }. Facing prcj1Hlice: Implicit prejudice ,11 1d the perception of faci,1l threat. Psyd~ologtcol Scienr:e, J<:l, 641)--(d.~.

l<la:wr, K. ( :., RoEnagcl, c~ & Musch, J. { 19971. I .i.-.H:nntext effects tn CVJ.iu<""lriYe priming. Journd r4 Lxju·rimental l'sy(hu/ogy: 1 ,eanung, -~·lerH')/ y, tmd Cog­mtion, 23, JAr~~l:iS.

Klinger, M. R., Burtf,rL P. C., & Pitts, G.~ .. (2.000). ~dcchaHisms of tmcons.~inus priming: L Response compctirinn, not ,c;prf:ading au-ivati:HL ln~mwl C>l Ex­perimcnt(d Psychofogy: Lcarnmg, i\!lemrm:, and Cog1utinn, .-~6 .. 44]-455.

Kornblum, S., 1 hlshnmcq, T, & O.sman, A. f 1990), DimensioiL<llnverlnp: Cogni­tn-'t' basis forst imulu,.-·-re,ponse compatibility-a ntPdel Jnd ra xmHJmy. 1\y­dudughal 1\evicl!', 97,253-270.

K 11Jlda, Z., &: Shennan--\Villiams, B. (19Sl3). ~rl'rcotypt'S and the u~n~trual of indi­viduating inforrnatitv:.l. Personality a11d Snd£11 Psychology Bulletm, 19, 90-

9~~' .Ylilnc, E., 1Sc Crafman, J. {2001). Ventromedial prefrontal cortex lesions in

h111nan-.; dirnin;-~te impliLit gender stereotyping. }omnaf o/}'·.'eurosciena·, 1 I, 1--6.

1'v1ih~hcll, J. P., Nosek, n. A., & Banaji, 1v1. R. (2003 ). Contextual vari~tion.s in irn·· plicit evaluati0n. Jounwl o{ Ex{H'rm·tent.:t! l'.~y,}•oiogy; Cem:ral, 132, 45S~ 469 .

.l\'1oors~ A., & De I-Iouwer., .J. (2006). Atltomat·icity: A theoreti:a! and conceptual analvsis. Psvchnh,;;ical Bulletm, 132, 297---326.

l\llusch,J.~ & Klauer, K. C. (Fds.). (2003). Psychology o{eua!uatiun: A/lectiuc ~·ro­o~sses in cognition and emotion. Mahwah, NJ: Erlbaum.

Ncnmann, R., Hi.llscnbcd:, K., & Seibt, B. ;2004). Attitude5. to"'>1/Md people ""'ith i\IDS and avoid<Jnce behavior: Automatic and reflective ba:,;es of behavior. fournal of Exf;erimentaf Social Psychology. 40, 543-550.

Nicr:l A. (2005). How dissociated arc implicit and explicit racnl attiwdes? Abo­gus pipeline approach. Group Process'.'s a11d 111tcrgrouj; Relattons, 8, 39-52.

Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know: Verbal re" pmts on mental processes. Psychologica/ Reuu-'u-', 84,231-259.

VVhat Do We Have to Learn? 285

Nosek, B. A. (.2(10)) . .l'doderators of the relationship betwt":~('Il implicit 8 nd e;-,.

phc1t arritude:;. journal o( ExfJermu:ntaf Psydmlngy: Cnreral, 134. 56.<-~ 584.

Olson~ _M.A., & F.az:o, R. I I. (20tH_:. Relation" between implicit mtw.;urcs of prej­lldtcl?: \\-''hOlt are \VC mcasunng? l'sychological Sdcna, 14., 6.36···639.

fayne, B. K. \lOOJ). Prejudi~:c and perception: "Jhc role of automauc and con

! rolled prnct~sv·s in mispncciving ;~ \..VC::lpon. Journal of I'ITSOIIa!itv i.llld So­cwlflsycholog_y, 81. un---1()2.

Phdps., LA., O'Connor~ K . .J., Cunningh;_wl, \Xf. 1\.l Funayanu, F. S., (~;_ltenln·, 1. C ... Core . .f. C.,{!" <J.l. (2000 I. Performance on indircu rnea:;ure:- of race cv;_liu;t­<ion predicts arn;'gdah ilctiva.tion. Journal of Cowntin; l'lcurnscie1u-e, L~. 729-7 3 8. '

Riche::-.on, J. A., Hnird, A. A., Cordon, l-1. 1 .. , Hcatherton, T E, Wyh1nd, <:. J .,

Traw:1f~cr, S., eta!. 120()3}. An Hv!Rl invt'stigation of the impact of intcrr::~c·JHl com:Kt on executive function. Nature j\leumscience, (~, _1.12.3-1328.

Rotherrnund, K., & \\?cmura, D. (20CJ<1 ). Underlying processt>-, in the ltnplicir As­~nci;Hion Te:>t: i )rssocJ:.Hing -'>il!icncc from associations.. Journal of Ex{;eri­.1ncntall'sy(holugy: Cenenl, 1 n, 139··· 165.

Rudman. I.. A. (2004), Snurce:. of implicir <Htitudes. CmTe1tl Directions in r 5vcho logical Sdena, U, 79-S2. -

~agaJ. H. A., & Schofi(·ld, J. W. ( 1980). ]bcial and behavioral cues in Qbck ;11 1d

'"·bite children's pet ccptions of ambiguous!~· aggrc:-.:si ve nets. founh11 ol Per­sonality md .\octal hrcholop,y, 1rJ, .590-598.

Srhwarz, N .. Bless, H., W\inke, ivf., & Winkiclm<lll, P. (20031. Accessibilitv rt'vis­;ted. In C. V. P,{){lenhausen & i\. J. L1nrhert ( Eds . .J, 1.-ounlflltif>n~ 0( soc/~1 1 ~:f'f{· nitkm (pp. 51-~?7;•. ;vJalnv,lh, NJ: Erlb<J.tml.

S<..lnvar7., N., & Hohner, G. (2001 ), The construction of attil"ulles. Jn A. Tesscr & N. Schwarz (bk), RillCkwell h,mdhnol: cf socialfJS)Yholor:,v: Intraindi,_,idua-' _t>mresscs (p-p. 436-45'?1. 1\"ialden, \1A: Bbchvdl. '.

Srrack, f., & Dcutst:h, R. (2004). Reflective :md impulsive dctcrrrmwnrs of sociai beha,·Jor. Pers(malil'y and SncLd Ps)'{holotD' Re-uiew, 8, 220---247 .

St··o·)~\ .f. R. { 1935). Studie~ of Interference in serial verbal reactions. journal o/ [:xperimental PsvLhnlr)/!.3'· 28. 613-662.

Teachman, B. A., Gregg, A. P_, & \\ioody S. R. (2001 ). Imphcit assooations for (e:u-rdcv,1m stimllli amoug individuals \Vith snake and spider fears. {ntmwf of Almormal Psychology, 110, 226~235. ·

Vanmctn, F. J.. Saltz, J L,. Nathan, L R., t\.:. Warren, l A. (2.004_1. Racial discrimi­nation by 10\v-prejudiccd Whites: F~.cral movements as implicit measmc of attitude~ related to behavior. Psychofogic.:d Science, 1.5, 711···7!4.

\)/arringron, E. K., & \'i/ciskrantz, L ( 1968). Nnv method of t,:'sting long-term re·· tcntion with spc..:ia! n:-ferC'nce to arnncsi.: patients. Nature, 217, 972-974.

\\lemma, D . .: l99Sli. Activation anJ inhib1tio.D of affective information: FYidcnce for negative priming in the eYaluation task. Cognition and J.imotio11, L3, 65-91.

\X.'hee1!er, M. E., & Fiske, S. T (2005). Controlling raci:1! prejudice: Socia!-cogtlitivr goals atfrct ~unyf_",dala and ~tereotypc <ICtivation. Psychological Science. 16 .• 56-63.

28G CRITICAL PERSPECTIVES

Wilson, T. D., l.i!Hlsey, :).,~X- Schooler. ·r. Y.l21)00). A !nUde] ofdu.1l attitude5. l'sy­cholrJgu:al Rel'ieu•, 10.?, 101---126.

\X!ittenbrink, H., Judd. C. M., & Park, B. 0997). Evidence for ra(i<ll prejudice :lt the impli<--~it level and its re!atiomhips ·with questiiJnnaire tn<'<lSUrts. jounwl o/ flersnn(l/ity olml Sc•(ird Psydmlugy. 7.?, 162-274.

\Vntenhnnk, 1~., Judd, C. M., &. Pilrk, B. (20f)l). Ev;~]ll;-Hin:' ver~us cnTKt:j)\ll;Jl

j udgmcnts in autotn<l tic stcreotypin~ ;_·.nd prejudice . .Jnurna! of Exf!tTimenta! ~oda!J>syeholog_y, 37, 244-252.