You are on page 1of 6

Implicit Attitude Formation through Classical Conditioning

Author(s): Michael A. Olson and Russell H. Fazio


Source: Psychological Science, Vol. 12, No. 5 (Sep., 2001), pp. 413-417
Published by: Sage Publications, Inc. on behalf of the Association for Psychological Science
Stable URL: http://www.jstor.org/stable/40063658
Accessed: 07-05-2015 23:39 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

Sage Publications, Inc. and Association for Psychological Science are collaborating with JSTOR to digitize, preserve and
extend access to Psychological Science.

http://www.jstor.org

This content downloaded from 101.255.89.163 on Thu, 07 May 2015 23:39:33 UTC
All use subject to JSTOR Terms and Conditions
PSYCHOLOGICALSCIENCE

Research Article
IMPLICIT ATTITUDE FORMATION THROUGH
CLASSICAL CONDITIONING
Michael A. Olson and Russell H. Fazio
Indiana University

Abstract- We sought to demonstrate that attitudes can develop PRIOR RESEARCH ON THE CLASSICAL
throughimplicitcovariationdetection in a new classical conditioning CONDITIONING OF ATTITUDES
paradigm. In two experimentspurportedly about surveillance and Razran (1938) read various political slogans to people while pro-
vigilance, participants viewed several hundred randomlypresented
wordsand images interspersedwith critical pairings of valenced un- viding some of them with a free meal, and found that agreement with
the slogans was greater when people received the free meal than when
conditionedstimuli (USs) with novel conditionedstimuli (CSs). Atti-
tudes towardthe novel objects were influencedby the paired USs: In they did not. Though myriad alternative explanations have been pro-
a surprise evaluation task, the CS paired with positive items was posed to account for Razran 's findings, this was probably the first ex-
evaluated more positively than the CS paired with negative items. perimental attempt to affect attitudes toward various objects via
This attitudinalconditioningeffect was found using both an explicit pairing an attitude object (CS, a political slogan) with other valenced
measure(Experiments1 and 2) and an implicitmeasure(Experiment objects (US, a free lunch). Staats and Staats (1958), in a more tightly
controlled experiment, paired each of two national names ("Swedish"
2). In a covariationestimationtask involvingthe stimulipresented in and "Dutch") with either 18 positive or 18 negative words by having
the conditioningprocedure,participants displayed no explicit mem-
subjects read the CS terms as they appeared on a screen (along with
oryfor the pairings. four filler national names) while the experimenter read aloud the US
terms. The nation paired with positively valenced terms was later eval-
Attitude formation, how people come to evaluate objects in the en- uated more favorably than the one paired with negatively valenced
vironment positively and negatively, is a long-standing issue in social terms.
psychology. A fundamental form of attitude acquisition, classical con- The attitudinal conditioning effect reported by Staats and Staats
ditioning, has struck the curiosities of not only social psychologists (1958) was presumed to occur without awareness of the CS-US pair-
(e.g., Cacioppo, Marshall-Goodell, Tassinary, & Petty, 1992; Kros- ings on the part of participants. This is a critical point, because if par-
nick, Betz, Jussim, Lynn, & Stephens, 1992; Zanna, Kiesler, & Pilko- ticipants had noticed that, say, "Sweden" was always followed by
nis, 1970), but marketing and advertising researchers (e.g., Allen & "good," then the attitudinal conditioning effect might be explained as a
Janiszewski, 1989; Kim, Allen, & Kardes, 1996; Shimp, Stuart, & En- demand artifact (i.e., participants may have been simply reporting
gle, 1991), human learning theorists (e.g., Baeyens, Eelen, & Van den what they assumed the experimenter wanted them to report). An alter-
Bergh, 1990; Hammerl & Grabitz, 1996a, 1996b; Levey & Martin, native explanation, had subjects noticed the CS-US pairings, is that
1975), and cognitive psychologists (e.g., Lewicki, 1986) as well. they may have formed attitudes on the basis of deliberate, expectancy-
Attitudes are thought to develop via classical conditioning value reasoning. As Fishbein and his colleagues have argued, the pair-
through repeated pairings of potential attitude objects (conditioned ings may have induced participants to consciously infer correspondent
stimuli, CSs) with positively and negatively valenced stimuli (uncon- beliefs about the attitude object (Fishbein & Ajzen, 1975; Fishbein &
ditioned stimuli, USs), and intuitively, one would expect this to be a Middlestadt, 1995).
ubiquitous means of attitude formation. From the development of ra- Although Staats and Staats (1958) argued that an associative
cial prejudice through repeated media portrayals of minority-group bond was created between the US and CS through the conditioning
members in a negative fashion, to the creation of brand preferences procedure without awareness on the part of participants, Page (1969)
via pairings of a target product with supermodels, the case might ap- contended that conditioning effects were simply demand artifacts
pear closed based on intuition alone. Yet, although early research based on participants' deliberate guesswork regarding the experi-
seemed to support the contention that classical conditioning is a pri- menter's hypothesis. He used a more sensitive, funneled debriefing
mary origin of attitudes (e.g., Razran, 1938; Staats & Staats, 1958), measure of awareness (in which participants respond to progres-
research on the classical conditioning of attitudes has since suffered a sively revealing questions about the nature of the experiment), and
sporadic and troubled history. found that conditioning effects occurred only when participants were
Through the present research, we hoped to provide more solid evi- aware of the experimenter's hypothesis. Page conducted several sub-
dence that attitudes can develop implicitly via classical conditioning. sequent studies (e.g., Page, 1974) that repeatedly reinforced his view
First, we touch on some of the problems of past research, and mention that attitudinal conditioning was the result of demand awareness;
some noteworthy attempts at solving them. The present research de- these studies surely contributed to the current neglect of condition-
veloped a new paradigm to help answer the question of whether atti- ing processes in social psychology.
tudes can develop implicitly via classical conditioning. Concerns about the possible role of demand characteristics were
abated by research that separated collection of the final dependent
measure from the conditioning phase of the experiment, for example,
by having the attitude measure administered by a different experi-
Addresscorrespondenceto MichaelA. Olson or Russell H. Fazio, Depart- menter in the context of an ostensibly unrelated experiment (e.g.,
ment of Psychology, 1885 Neil Ave., Ohio State University,Columbus, OH Zanna et al., 1970). However, the role of awareness of the CS-US con-
43210-1222; e-mail: micolson@indiana.eduorfazio@psy.ohio-state.edu. tingencies remains a highly contentious issue (e.g., Field, 2000). In

VOL. 12, NO. 5, SEPTEMBER2001 Copyright© 2001 AmericanPsychological Society 413

This content downloaded from 101.255.89.163 on Thu, 07 May 2015 23:39:33 UTC
All use subject to JSTOR Terms and Conditions
PSYCHOLOGICALSCIENCE

Implicit Attitude Formation

fact, severalearlierstudies (e.g., Cohen, 1964; Insko & Oakes, 1966), THE IMPLICIT LEARNING PARADIGM
as well as more recent ones (e.g., Allen & Janiszewski, 1989; Shimp Some reason for optimism about the validity of attitudinalcondi-
et al., 1991), have found conditioningeffects only among participants
tioning is providedby the literatureon implicit learning.Individuals
who reportawarenessof the contingencies. sometimes show evidence of having learneda rule or associationim-
Some researchers,particularlyBaeyens and his colleagues (e.g.,
plicitly, even though they are unable to articulateany explicit, con-
Baeyens et al., 1990; Baeyens, Eelen, Crombez, & Van den Bergh, scious knowledge of the relevantinformation(for reviews, see Berry,
1992), remainconfidentthatattitudinalconditioningcan occur without 1994; Seger, 1994). Such implicit learninghas been demonstratedin a
awareness.In theirtypical researchparadigm,originallydevelopedby
variety of domains, such as learning an artificial grammar(Reber,
Levey and Martin(1975), picturesthateach participantearlierratedas 1967), learning covariationsbetween facial features and trait ascrip-
neutralare pairedrepeatedlywith picturesthat they earlier indicated tions (Lewicki, 1986; Lewicki, Hill, & Czyzewska, 1992), and learn-
liking or disliking strongly,all underthe guise of an experimentcon- ing such visual covariationas regularitiesbetween targetobjects and
cernedwith physiological responses.The neutralstimulusis typically theirlocationin a spatiallayout(Chun& Jiang, 1999). Forthis reason,
presentedfor 1 s, and 3 s later,the US is presentedfor 1 s. The stimu- we sought to develop an experimentalparadigmmore similar to the
lus pairs are separatedby 8 s. The initially neutralstimuli have been
procedurestypically employed to study implicit learning.We wanted
shownto acquirethe valenceof the repeatedlyassociatedUS. Awareness the paradigmto (a) have minimal likelihood of promotingeither de-
is measuredconcurrently(by asking participantsto predictwhethera mand characteristicsor contingencyawarenesson the partof partici-
neutral,liked, or disliked stimulus will follow the CS) or postcondi- pants and (b) capture the real world's tendency to present natural
tioning (by a recognitiontest in which participantsindicatewhich par- covariationsbetween objects and valenced outcomes that may affect
ticular US followed each CS). Baeyens et al. (1990) observed that,
people's attitudes,despite theirattentionbeing directedelsewhere.
even though some participantsreportawarenessof some contingen- Withthese aims in mind,we chose to employsupraliminal exposures
-
cies, contingency awarenessis unrelatedto the conditioningeffect to the CS andUS andto presentthe CS-US pairssimultaneously, butin-
awareand unaware participantsboth show it. in the context of a nonrhythmic stream of perceptual
terspersed rapid,
Althoughthe studies by Baeyens and his colleagues are encourag- events. The paradigminvolves leading participantsto believe thatthe
ing, theirconclusions remaincontested.Otherresearchers(e.g., Field, experimentis about"attentionandrapidresponding"to particular events.
2000; Field & Davey, 1998; Shanks& St. John, 1994) have expressed We present to them a wide assortmentof images and words, including
concern about the high levels of contingency awarenessdisplayed on variousneutralitems (e.g., an umbrella,a glass of milk, the word"con-
the concurrentmeasure, and about the validity and sensitivity of the crete"),thatcan appearanywhereon the screen,eitheraloneor in pairs.
postconditioningrecognitionmeasure.In short,considerabledisagree- At leastone thirdof all trialsareblankscreens.Thus,participants see an
mentremainsas to whetherthe availableevidence has establishedthat unrelated assortment of words and images, sometimes appearingalone
attitudescan be conditionedin the absence of contingencyawareness andsometimesin pairs,andin varyinglocationson the screen.The trials
(also see Eagly & Chaiken, 1993; Hammerl,2000; Shanks & Dickin- involvinga blankscreenensurethatthe stimuliarepresentedat a seem-
son, 1990). ingly nonsystematic,irregularpace. Participantsare askedto be vigilant
The studies discussed up to this point have approachedthe prob- for a particularitem,andareinstructedto hit a responsekey as quicklyas
lem of eliminating contingency awarenessby relying on misleading
possiblewheneverthe targetitem appearson the screen.CriticalCS-US
cover stories and participants'limited memories. Other researchhas
pairsarepresentedsimultaneouslythroughoutthe presentation.
taken a subliminalroute- instead of hoping participantswill not no- Ourparadigmdepartsfrom most otherattitudinalconditioningpara-
tice CS-US contingencies, these researcherssimply present the USs
digms in severalways. First,thereis no preratingphasein whichpartic-
for subthresholddurations.Krosnicket al. (1992) presentedsubliminal
ipantsevaluatethe items to be used in the conditioningprocedure;we
images of positive or negativeobjects (USs), followed immediatelyby derive our positively and negatively valenced USs normativelyfrom
supraliminalimages of a woman going aboutvariousmundaneactivi- pretestsusing differentparticipants.Second,insteadof exposingpartici-
ties (CSs). This backwardconditioning procedure,in which the CS
pantsto the sameCS-US pairsrepeatedly,each CS is pairedwith several
follows the US, yielded a significantconditioningeffect. However,a
nonrepeating,same-valencedUSs- all in the interestof diminishingthe
posttest indicated that participantscould detect the presence of the likelihoodof participantsconsciouslynotingthe pairings.Third,instead
USs, even though they could not identify their content. Using a of presentingthe CS and US sequentially,and segmentedfrom other
shorterpresentationtime in a second experiment,Krosnicket al. ob- stimuliby a relativelylengthyintertrialinterval,we presenteach CS-US
tained a marginal conditioning effect. Unfortunately,the backward
pair simultaneously,and embeddedwithin a streamof similarpercep-
pairingprocedureleaves open the possibility of an alternativeexpla- tual events. Finally,the cover story and experimentaltask ensurethat
nation in terms of affective priming, as Eagly and Chaiken (1993)
participantsattendto the stimulusitems, but engage them in a task that
have noted. Participantsmay have respondedmore positively (or neg- is entirelyindependentof covariationdetection.
atively) to the target only because the affective primes increasedthe
accessibility of positivity (or negativity), and not because of condi-
tioned changes in their evaluative reactions to the target itself (see EXPERIMENT 1
Niedenthal, 1990). Subliminalpresentationexperimentsutilizing the Experiment1 was an initial demonstrationof the procedure'sef-
more typical forward conditioning procedure have also been con- fectiveness in classically conditioningattitudestowardnovel objects.
ducted(De Houwer,Baeyens, & Eelen, 1994; De Houwer,Hendrickx,
& Baeyens, 1997). A meta-analysisof five such experimentsrevealed
Method
a statistically reliable conditioning effect, but the research clearly
yielded somewhat inconsistent results involving small effect sizes Participants(45 undergraduatefemales) were told that the experi-
(De Houweret al., 1997). ment was about"video surveillance,"and that severalhundredimages

414 VOL. 12, NO. 5, SEPTEMBER2001

This content downloaded from 101.255.89.163 on Thu, 07 May 2015 23:39:33 UTC
All use subject to JSTOR Terms and Conditions
PSYCHOLOGICALSCIENCE

MichaelA. Olson and Russell H. Fazio

would be presentedrandomlyon a computerscreen over the course of Pokemonpairedwith positive USs more positively than the Pokemon
five blocks. Theirtask was to hit a responsekey as quickly as possible pairedwith negativeUSs.
when a targetimage appeared.Eight CS-US pairs were presentedin
each of the five blocks. These pairsconsisted of one Pokemoncartoon Awareness test
character(a color image with the character'sname below it) with one
of a numberof positively valenced words (e.g., "excellent,""awe- The covariationestimationdata were analyzed by first calculating
some") and images (e.g., puppies, a hot fudge sundae) and another the mean of participants'responses to the three USs of a given va-
Pokemon cartoon characterpaired with negative words (e.g., "terri- lence. A 2 (CS counterbalancingcondition) X 2 (Pokemontarget)X 2
ble,""awful")and images (e.g., a cockroach,a man wielding a knife).1 (US valence) analysis of variance(ANOVA),with repeatedmeasures
Across the five blocks, 20 CS-US pairswere presentedfor each of the on the last two variables,was performedon these scores. If partici-
two Pokemon.Which CS Pokemonwas pairedwith positive USs (i.e., pants were able to consciously detect and recall the covariationsbe-
the CS+) and which was paired with negative USs (i.e., the CS-) tween the CSs and USs, we would expect a three-way interaction.
constitutedone independentvariable(for which no effects emerged). That is, the covariationresponses involving PokemonA and the posi-
The targetimage for each block was also a Pokemoncharacter,and a tive USs and those involving PokemonB and the negativeUSs would
differenttargetwas used for each block. The filler images consisted of be higher than the other combinationsfor the condition in which A
otherPokemon,blank screens, and neutrallyvalenced words and im- was the CS+ and B the CS-, whereas the reverse would be true for
ages presentedindividuallyand in pairs. Each block consisted of 86 the other condition. The ANOVA revealed that participantsexhibited
trialspresentedfor 1.5 s each. no explicit awarenessof the pairings (F < 1). Participantswere no
Afterthe conditioningprocedure,participantswere told thatwe were more confidentthat a given Pokemonhad been pairedwith a US of a
concernedthattheiraffectivereactionsto some of the filleritems might given valence than they were that the Pokemon had appearedwith a
have interferedwith theirabilityto respondrapidlyto the targetitems. US of the opposite valence.
Hence, we askedthem to evaluateseveralof the images (includingthe In orderto create a summarymeasureof awareness,we computed
two CS Pokemon)on a scale from -4 (unpleasant)to +4 (pleasant). the differencebetween the mean of participants'covariationestimates
Participantswere also told that some of the images they saw may for the CS-US pairs that were actuallypresentedand the mean for the
have been presentedtogetherwith some degreeof regularity,and were CS-US pairs with the opposite-valencedUSs. Again, no awareness
asked to provide covariationestimates for several pairs of images. was apparent,M = -0.007, SD = 0.84; t < 1. Nor was this summary
Throughthe course of this covariationestimationtask, each CS was index of awarenessrelatedto the magnitudeof the conditioningeffect,
presentednine times, each time followed by a differentitem selected r = .03, n.s.
from three positive USs, three negative USs, and three neutralfiller As an additionaltest of participants'ability to notice covariations
items. For each pair,participantsrespondedon a scale from -2 ("I'm during the conditioning procedure,two pairs of filler items had ap-
confidentthat the two items never appearedtogether")to 0 ("don't pearedtogether 15 times throughthe course of the 430 trials (e.g., the
know")to +2 ("I'm confidentthat the two items appearedtogetherat PokemonPrimapeand the word "outlet").Participants'mean covaria-
least once"). Severalpairs of filler items, some of which actually ap- tion estimatesfor a given filler image and the item actuallypairedwith
pearedtogetherduringthe conditioningphase, also were included in it were slightly lower, but not significantly so, than their covariation
the covariationestimationtask. The orderin which participantscom- estimates for that same filler image and two fillers with which it had
pleted the evaluationand explicit memorytasks was counterbalanced not been paired,mean difference score = -0.30, SD = 1.34; f(44)=
acrossparticipants;no effects emergedfor this variable. 1.47, n.s. This measure also was unrelatedto the magnitudeof the
conditioningeffect, r = .13, n.s. Thus, our cover story,along with the
vigilance task, was effective in precludingparticipants'explicit mem-
Results and Discussion ory for the pairings.
Conditioning effect
EXPERIMENT 2
Conditioningwas examinedby collapsing across the variabledes-
ignatingwhich Pokemonwas pairedwith positive versusnegativeUSs Experiment2 was designed to replicate and extend the first ex-
and computinga differencescore (ratingof the positive CS - rating periment by using an additional dependent measure. A recently de-
of the negativeCS), for which higher numbersindicatea greatercon- veloped measure of attitudes, the Implicit Association Test (IAT;
= 2.05), Greenwald,McGhee, & Schwartz, 1998), was included as an addi-
ditioning effect. The mean difference score was 0.64 (SD
which differedsignificantlyfrom zero, f(44) = 2.1 1, p < .05. Thus, a tional measureof the conditioningeffect. The IATis a response-map-
reliable conditioningeffect was obtained;participantsevaluated the ping task in which participantsare requiredto categorize four types
of objects as they are presentedon a screen using only two response
keys. Two of the four types of objects are clearly pleasantand clearly
unpleasantitems (e.g., "love,""murder"),and the othertwo represent
1. The PokemoncartooncharactersShelder and Metapodwere chosen as the two attitudeobjects the experimenteris interestedin assessing (in
CSs because we found that althoughcollege studentswere willing to evaluate
our case, the CS+ and the CS-). The meaning of the categorization
them, they were generallyignorantof Pokemoncharacters(with the exception
of Pikachu,who was not employed in the experimentsreportedhere). US im- keys changes throughoutthe course of the IAT, so that during some
ages were selected based on normativedata provided by Ito, Cacioppo, and blocks, participantsmap compatible items to the same key (pleasant
Lang (1998); US words were selected based on pretestingin our lab. In all items and the CS+, unpleasantitems and the CS- ). In the incom-
cases, positive USs were clearly evaluatedmore positively than negative USs patible blocks, participantsmap incompatible items to the same key
by pretestparticipants. (unpleasantitems and the CS+, pleasantitems and the CS-). Partic-

VOL. 12, NO. 5, SEPTEMBER2001 415

This content downloaded from 101.255.89.163 on Thu, 07 May 2015 23:39:33 UTC
All use subject to JSTOR Terms and Conditions
PSYCHOLOGICALSCIENCE

Implicit Attitude Formation

ipants should find the compatible task easier than the incompatible
task, which should be reflectedin their response latencies. For exam-
ple, if our conditioning procedure is effective, then participants
should find it relatively easy to map pleasant items and the CS+ to
one response key, and unpleasant items and the CS- to another
(compatibletask). In contrast,they should find the incompatibletask
more challenging.
Because implicit measureslike the IAT are less susceptibleto de-
mandartifactsthanexplicit measures,and do not requireintrospective
access to any unconsciouslyformedattitudes,implicitmeasuresof atti-
tudes may be especially well suited for detectingattitudedevelopment
via a classical conditioningprocedurethatdoes not requireawareness.

Method
Fifty-six undergraduate women participatedin the experiment.The
conditioningprocedure was identicalto thatused in Experiment1. Af-
ter the procedure,participantsratedseveralof the items (includingthe
two CSs) as in Experiment1, andcompletedan IATdesignedto assess
attitudedevelopmentfor the two CSs. The orderin which participants -
Fig. 1. Mean differences(PokemonA PokemonB) in standardized
completedthese two evaluationtasks was counterbalanced(no effects scores on the implicit and explicit attitudemeasuresby valence of the
emerged for this variable).The IAT included two different practice conditionedstimuli (CSs; A = CS+, B = CS- vs. A = CS-, B =
blocks; participantsfirst categorizedpleasantand unpleasantitems in CS+), Experiment2.
one of these blocks, and then categorizedvariouscombinationsof im-
ages and names of the two CS Pokemon in the second block. They
then completed incompatibleand compatibleblocks consisting of 50
trials each.2After the evaluationtasks, participantscompleted a fun- subjectsfactor,was conductedon these data. The analysis revealeda
= =
neled questionnairemeasure of awareness involving a series of in- significantconditioningeffect, F(l, 46) 6.45, p .015, unmoderated
of
by eithertaskorderor type measure, both Fs < 1.3Subsequentt tests
creasingly direct questions. It began with the question, "Did you indicatedthatthe CS+ was evaluatedmorepositivelythanthe CS- on
notice anythingout of the ordinaryin the way the words and pictures
boththe implicitmeasure(IAT),f(48) = 2.59, p = .007, andthe explicit
were presented?"and ended with the question, "Did you notice any-
(paper-and-pencil) measure,f(48) = 1.92, p = .03. The two measures
thing unusual about the words and images that were presentedwith were correlatedsignificantly,r(48) = .39,/? < .01.
the[CS+andCS-]?"

GENERAL DISCUSSION
Results
In Experiment1, attitudeswere conditionedtowardnovel objects
On the final item of the postexperimentalquestionnaire,6 of the 56
using our newly developedconditioningparadigm.Experiment2 rep-
participantsmentionednoticing at least one of the CS-US contingen- licated this finding using an implicit measureof attitudedevelopment
cies. It is importantto keep in mind that these reportsoccurredonly
in addition to the traditionalexplicit measure. In both experiments,
after the questionnaireinformed participantsthat there was more to
the experimentthanthey had initiallybeen led to believe, and only af- participantsremainedunawareof CS-US contingencies.These results,
ter it was suggested that there was something unusualin the way the along with the subliminalconditioningfindings of De Houwer et al.
CSs were presented.To be conservative,we excluded the data from (1997), strongly suggest that attitudescan be conditionedin the ab-
sence of contingencyawareness.
these 6 participantsfrom all analyses, althoughthis omission had no
It is our hope that our successful developmentof a paradigmin
effect on the statisticalsignificanceof the results.
which attitudesare formed implicitly,with no accompanyingexplicit
The latenciesfrom the IAT were first log-transformed.Differences
were computedbetween(a) each participant'smeanlatencyon trialsfor memory for the CS-US pairings,will invigorateresearchon classical
which PokemonA and pleasantwere representedby the same response conditioningas an attitude-formation process.As a whole, the field of
social psychology has attendedmore to questions regardingattitude
key (andPokemonB andunpleasantwere representedby the otherkey)
and (b) the participant'smean latency on trials for which PokemonA change, attitudestructureand function, and influencesof attitudeson
and unpleasantwere mapped onto the same key (and, hence, B and judgments and behavior than it has to attitudeformation.In fact, in
discussing future directions in the study of attitudes, Eagly and
pleasantwere mappedonto the other). The difference in the explicit Chaiken(1993) referredto the field's current"lackof attentionto the
scale ratingfor A minus B also was computedfor each participant.In
order to place these two measures on a common metric, each was developmentalissue of how attitudesform and become strong"as a
"seriousomission and limitation"(p. 681). The implicitlearningpara-
z-transformed. The meansarepresentedin Figure1.A 2 (A = CS+, B =
CS- vs. A = CS-, B = CS+) X 2 (evaluation-taskorder) X 2 (im-
plicit vs. explicit measure)ANOVA,with the last variableas a within-
3. As in Experiment1, additionalanalysesrevealedno effects for the coun-
terbalancingvariable designating which Pokemon was paired with positive
2. No wordthatappearedin the conditioningprocedureappearedin the IAT. versus negativeUSs.

416 VOL. 12, NO. 5, SEPTEMBER2001

This content downloaded from 101.255.89.163 on Thu, 07 May 2015 23:39:33 UTC
All use subject to JSTOR Terms and Conditions
PSYCHOLOGICAL SCIENCE

MichaelA. Olson and Russell H. Fazio

digm should prove useful in redressing this imbalance, illuminating Chun,M.M., & Jiang,Y. (1999). Top-downattentionalguidancebased on implicitlearning
of visual covariation.Psychological Science, 10, 360-365.
and testing some fundamental principles regarding the development of
Cohen, B.H. (1964). Role of awarenessin meaningestablishedby classical conditioning.
evaluative associations. Journalof ExperimentalPsychology,67, 373-378.
If classical conditioning is the pervasive source of attitudes that we De Houwer,J., Baeyens, F, & Eelen, P. (1994). Verbalevaluativeconditioningwith unde-
tected US presentations.BehaviourResearchand Therapy,32, 629-633.
believe it is, then psychologists' understanding of attitude formation De Houwer,J., Hendrickx,H., & Baeyens, F. (1997). Evaluativelearningwith "sublimi-
and change stands to gain much from further work on the implicit for- nally"presentedstimuli. Consciousnessand Cognition,6, 87-107.
mation of attitudes. Instead of concerning itself with the question of Eagly,A.H., & Chaiken,S. (1993). Thepsychologyof attitudes.New York:HarcourtBrace
JovanovichCollege Publishers.
whether attitudes can be classically conditioned, research can turn to Field, A.P. (2000). I like it, but I'm not sure why: Can evaluativeconditioningoccur with-
questions regarding the conditions that moderate such implicit attitude out conscious awareness?Consciousnessand Cognition,9, 13-36.
Field, A.P, & Davey, G.C.L. (1998). Evaluativeconditioning:Arte-factor fiction?A reply
formation. Some work has already been done on this front (see Baey- to Baeyens, De Houwer,Vansteenwegen,& Eelen (1998). Learningand Motivation,
ens et al, 1992; Hammerl & Grabitz, 1996a, 1996b; Kim et al., 1996). 29,475^91.
For example, one such question centers on CS-US relationships: What Fishbein,M., & Ajzen, I. (1975). Belief, attitude,intention,and behavior:An introduction
to theoryand research.Reading,MA: Addison-Wesley.
makes the most effective US for a given CS? Seligman's (Seligman & Fishbein, M., & Middlestadt,S. (1995). Noncognitive effects on attitudeformationand
Hagar, 1972) preparedness theory suggests that humans may more change:Fact or artifact?Journalof ConsumerPsychology,4, 181-202.
Greenwald,A.G., McGhee, D., & Schwartz,J.L.K. (1998). Measuringindividualdiffer-
easily detect some environmental covariations than others (e.g., Ohman ences in cognition:The ImplicitAssociationTask.Journalof Personalityand Social
& Dimberg, 1978). Work by Hammerl and Grabitz has begun address- Psychology,74, 1469-1480.
ing other relevant issues, such as the effects of sensory precondition- Hammerl,M. (2000). I like it, but only when I'm not sure why: Evaluativeconditioning
and the awarenessissue. Consciousnessand Cognition,9, 37-40.
ing on evaluative conditioning (1996b) and the possibility of evaluative Hammerl,M., & Grabitz,H.-J. (1996a). Evaluativeconditioningwith hapticstimuli.Inter-
conditioning using haptic stimuli (1996a). nationalJournalof Psychology,31, 375.
The field has seen more than 40 years of debate regarding whether Hammerl,M., & Grabitz,H.-J. (1996b). Humanevaluativeconditioning without experi-
encing a valuedevent. Learningand Motivation,27, 278-293.
attitudes can form via classical conditioning without awareness. With Insko,C.A., & Oakes,W.F (1966). Awarenessand the "conditioning"of attitudes.Journal
an affirmative answer, much needed progress now can be made toward of Personalityand Social Psychology,4, 487-496.
Ito, T.A., Cacioppo, J.T., & Lang, P.J. (1998). Eliciting affect using the InternationalAf-
understanding the processes underlying attitudinal conditioning, as fective PictureSystem: Trajectoriesthroughevaluative space. Personalityand So-
well as toward designing more effective marketing, health-promotion, cial PsychologyBulletin,24, 855-879.
and prejudice-reduction campaigns that utilize some form of classical Kim, J., Allen, C.T., & Kardes,FR. (1996). An investigationof the mediationalmecha-
nismsunderlyingattitudinalconditioning.Journalof MarketingResearch,33, 318-328.
conditioning. Krosnick,J.A., Betz, A.L., Jussim, L.J., Lynn, A.R., & Stephens, L. (1992). Subliminal
conditioningof attitudes.Personalityand Social PsychologyBulletin,18, 152-162.
Levey,A.B., & Martin,I. (1975). Classical conditioningof human 'evaluative responses.
Acknowledgments - This research was supported by Senior Scientist BehaviourResearchand Therapy,13, 221-226.
Award MHO1646 and Grant MH38832 from the National Institute of Men- Lewicki, P. (1986). Nonconscious social informationprocessing. New York:Academic
tal Health, and a grant from the Indiana University Graduate School. Por- Press.
tions of this research were presented at the June 2000 annual meeting of the Lewicki, P., Hill, T., & Czyzewska, M. (1992). Nonconscious acquisitionof information.
American Psychological Society in Miami, Florida. We thank Cheryl AmericanPsychologist,47, 796-801.
Niedenthal,P.M. (1990). Implicit perceptionof affective information.Journal of Experi-
Burke, Roxanne Rowers, Katie Hershey, Jeff Johnson, and Danielle Wiser mentalSocial Psychology,26, 505-527.
for assistance in pilot testing the procedure and in collecting data; Suzanne Ohman,A., & Dimberg,U. (1978). Facial expressions as conditionedstimuli for electro-
Miller for many helpful discussions about this research; Daniel and dermalresponses:A case of "preparedness"? Journalof Personalityand Social Psy-
Michael Fazio for help in selecting the stimuli; and Nira Liberman, Jim chology,36, 1251-1258.
Sherman, and Bill Timberlake for valuable feedback on the manuscript. Page, M.M. (1969). Social psychology of a classical conditioningof attitudesexperiment.
Journalof Personalityand Social Psychology,11, 177-186.
Page, M.M. (1974). Demandcharacteristicsand the classical conditioningof attitudesex-
periment.Journalof Personalityand Social Psychology,30, 468^76.
Razran,G.H.S. (1938). Conditioningaway social bias by the luncheontechnique.Psycho-
logical Bulletin,37, 481.
REFERENCES Reber,A.S. (1967). Implicit learningof artificialgrammars.Journal of VerbalLearning
and VerbalBehavior,5, 855-863.
Allen, C.T., & Janiszewski,C.A. (1989). Assessing the role of contingency awarenessin Seger, C.A. (1994). Implicitlearning.Psychological Bulletin, 115, 163-196.
attitudinalconditioningwith implicationsfor advertisingresearch.Journal of Mar- Seligman, M.E.P., & Hagar,J.L. (1972). Biological boundaries of learning. New York:
Meredith.
ketingResearch,26, 30-43.
Shanks,D.R., & Dickinson,A. (1990). Contingencyawarenessin evaluativeconditioning:
Baeyens, F, Eelen, P., Crombez,G., & Vanden Bergh, O. (1992). Humanevaluativecon- A commenton Baeyens, Eelen, & van den Bergh.Cognitionand Emotion,4, 19-30.
ditioning:Acquisitiontrials,presentationschedule,evaluativestyle and contingency Shanks,D.R., & St. John, M.F. (1994). Characteristicsof dissociable humanlearningsys-
awareness.BehaviourResearchand Therapy,30, 133-142.
tems. Behavioraland Brain Sciences, 17, 367-447.
Baeyens, F, Eelen, P., & Vanden Bergh, O. (1990). Contingencyawarenessin evaluative
Shimp,T.A., Stuart,E.W., & Engle, R.W. (1991). A programof classical conditioningex-
conditioning:A case for unawareaffective-evaluativelearning.Cognitionand Emo-
tion, 4,3-18. periments testing variations in the conditioned stimulus and context. Journal of
ConsumerResearch,18, 1-22.
Berry,D.C. (1994). Implicitlearning:Twenty-fiveyears on. A tutonal. In C. Umilta & M. Staats,A.W., & Staats,C.K. (1958). Attitudesestablishedby classical conditioning.Jour-
Moscovitch(Eds.),Attentionand performanceXV:Consciousand nonconsciousin-
nal of Abnormaland Social Psychology,11, 187-192.
formationprocessing(pp. 755-781). Cambridge,MA: MIT Press.
Zanna,M.P.,Kiesler,C.A., & Pilkonis,PA. (1970). Positive and negativeattitudinalaffect
Cacioppo,J.T., Marshall-Goodell,B.S., Tassinary,L.G., & Petty, R.E. (1992). Rudimen- establishedby classical conditioning.Journalof Personalityand Social Psychology,
tary determinantsof attitudes:Classical conditioningis more effective when prior 74,321-328.
knowledgeaboutthe attitudestimulusis low thanhigh.Journalof ExperimentalSo-
cial Psychology,28, 207-233. (Received 10/8/00; Revision accepted 1/2/01)

VOL. 12, NO. 5, SEPTEMBER2001 417

This content downloaded from 101.255.89.163 on Thu, 07 May 2015 23:39:33 UTC
All use subject to JSTOR Terms and Conditions

You might also like