When Arousal Influences Ad Evaluation and Valence Does Not

14 Pages·2001·0.24 MB·English

JOURNAL OF CONSUMER PSYCHOLOGY,11(1),43–55
Copyright © 2001, Lawrence Erlbaum Associates, Inc.
When Arousal Influences Ad Evaluation and Valence
Does Not (and Vice Versa)
Gerald Gorn
Department of Marketing
Hong Kong University of Science and Technology
Michel Tuan Pham
Graduate School of Business
Columbia University
Leo Yatming Sin
Department of Business Administration
Chinese University of Hong Kong
Thisresearchexamines,across2studies,theinterplaybetweenthevalenceandarousalcompo-
nentsofaffectivestatesandtheaffectivetoneofatargetad.Inthefirststudy,musicwasusedtoin-
duceapleasantorunpleasantmood,whilecontrollingforarousal.Participantsweresubsequently
exposedtoanadthateitherhadapositive-affectivetoneorwasambiguousinitsaffectivetone.As
predicted,thevalenceoftheaffectivestatecoloredtheevaluationoftheadinamood-congruent
direction,butthiscoloringeffectoccurredonlywhentheadhadanambiguous-affectivetone.In
thesecondstudy,thetargetadhadaclearpositiveornegativeaffectivetone,andthevalenceand
arousaldimensionsofthemoodstateweremanipulatedindependently.Aspredicted,thearousal
dimension,butnotthevalencedimension,influencedadevaluation.Adevaluationsweremore
polarizedinthedirectionofthead’saffectivetoneunderhigharousalthanunderlowarousal.This
effectwasmorepronouncedforself-referentevaluations(e.g.,“Ilikethead”)thanforobject-ref-
erentevaluations(e.g.,“Theadisgood”),favoringanattributionalexplanation—theexcitation
transferhypothesis—overanattention-narrowingexplanation—thedynamiccomplexityhypoth-
esis.Takentogether,theresultsofthe2studiesstresstheimportantcontingencyoftheaffective
toneofthead,whenexaminingtheeffectsofthevalenceandarousaldimensionsofaperson’saf-
fectivestateonadevaluation.Theresultsalsoprovideadditionalinsightsintohowandwhenaf-
fectservesasinformationinjudgmentprocesses.
Thesourcesofaffectcanbediverse.Inadexposuresettings, of this interplay, formalize its main contingencies, and ex-
onesourceofaffectistheconsumer’spreexistingmoodstate plore underlying processes.
atthetimeofexposure.Asecondsourceliesintheaffective Wesuggestthattoexaminethisinterplay,itisusefultode-
toneof the aditself(e.g.,humorousor fear-inducing).1The composeconsumers’affectivestatesalongtheirtwoprimary
interplaybetweentheconsumer’saffectivestateandtheaf- dimensions(Havlena&Holbrook,1986;Mehrabian&Rus-
fective content of the adcan have intricate effects on judg- sell, 1974): valence (pleasant or unpleasant) and arousal
ments.Thepurposeofthisresearchistodocumentthenature (highorlow).Weproposethateachofthesetwocomponents
of consumers’ affective states interacts differently with the
affectivetoneofatargetad.Priorresearchhassuggestedthat
thevalencecomponenthastheeffectof“coloring”people’s
RequestforreprintsshouldbesenttoMichelTuanPham,515UrisHall, judgmentsinamood-congruentdirection(e.g.,Forgas,1995;
GraduateSchoolofBusiness,ColumbiaUniversity,3022Broadway,New
Mayer,Gashke,Braverman,&Evans,1992).Ceterisparibus,
York, NY 10027. E-mail: [email protected]
thetargetadshouldbeevaluatedmorefavorablywhentheaf-
1Byaffectivetone(e.g.,humorous),wemeanthattheadhasthepotentialto
evokespecificfeelingsrelatedtothetone(e.g.,feelingsofamusement=laughter). fective state is pleasant than when it is unpleasant (e.g.,
44 GORN, PHAM, SIN
Goldberg&Gorn,1987).However,wearguethatthiscolor- biguous.Forinstance,IsenandShalker(1982)observedthat
ingeffectismorelikelytooccurwhentheadisambiguousin the effect of being in a pleasant mood when assessing the
itsaffectivetonethanwhentheaditselfhasaclearlypositive pleasantnessofslidesoflocalsceneswasstrongerwhenthe
ornegativeaffectivetone.Whentheaddoeshaveaclearaf- slides were affectively neutral than when the slides were
fectivetone—andmoodvalencehaslessofaninfluence—it clearly pleasant or unpleasant. Similarly, Miniard, Bhatla,
isthearousalcomponentthathasaneffectonadevaluation. andSirdeshmukh(1992)observedthattheeffectofbeingina
Thiseffectistopolarizejudgmentsbyaccentuatingtheaffec- pleasant or unpleasant mood on participants’ rating of the
tivetoneofthetarget ad.Adsthathavea positive-affective taste of a brand of peanut butter was stronger when the
tone are likely to be evaluated more favorably under high brand’s actual taste was ambiguous than when it clearly
arousalthanunderlowarousal,whereasadsthathaveanega- tastedgoodorbad.Severalmechanisms,whicharenotneces-
tiveaffectivetonearelikelytobeevaluatedlessfavorablyun- sarilymutuallyexclusive,canexplainwhymoodshaveless
der high arousal than under low arousal. In other words, ofaneffectwhenthetargetisaffectivelyunambiguous.First,
dependingontheaffectivetoneofthead,thevalenceofaffec- thetarget’saffectivecontentmayactuallyerasethepreexist-
tivestateshasaneffectwhenarousaldoesnot,andviceversa. ingmoodstate,precludinganymoodinfluenceonjudgment
Thesepredictionsaretestedintwostudies:onefocusingon (Miniardetal.,1992).Second,whenthetargetisaffectively
coloring and the other focusing on polarization. The pro- unambiguous,themoodstatewillloseitsheuristic(informa-
cesses underlying coloring and polarization are discussed, tion) value in evaluative judgments (Clore et al., 1994).
withtwodifferentexplanationsof polarizationexaminedin Third,judgmentsofaffectivelyunambiguoustargetsareless
Study 2. likelytorequireasearchforadditionalinformation.Asare-
sult,mood-congruentassociationswillhavelessofaninflu-
ence on these judgments (Isen & Shalker, 1982).
VALENCE, AROUSAL, AND Therolethataffectiveambiguityofthetargetplaysinmod-
AD EVALUATION eratingthecoloringeffectofmoodvalencecanalsobeinter-
preted in light of Forgas’s (1995) Affect Infusion Model
Valence and Coloring (AIM).Accordingtothismodel,moodismostlikelyto“in-
fuse”(i.e.,color)judgmentswhenpeopleuseagenerativepro-
Aconsiderableamountofresearchhassuggestedthattheva- cessingstrategy—thatis,a heuristic or systematic searchfor
lenceofpeople’saffectivestatesinfluencestheirjudgments additionalinputs—toconstructtheirjudgments.Whenthetar-
in a mood-congruent direction (e.g., Clore, Schwarz, & getisaffectivelyunambiguous,agenerativeprocessingstrat-
Conway,1994;Forgas,1995;Goldberg&Gorn,1987;Gorn, egy—hence, infusion or coloring—is less likelybecause the
Goldberg,&Basu,1993;Mayeretal.,1992).Twoexplana- target’saffectivetoneprovidespotentcuesforthejudgments.
tionshavebeenofferedforthiscoloringeffect.Oneexplana- Consistentwiththisreasoning,wehypothesizethefollowing:
tionisthatmoodstates—pleasantorunpleasant—cue simi-
larlyvalencedmaterialsinmemory,therebybiasingpeople’s H1: The coloringeffect of the valence of people’s
perceptionsofthetargetatthetimeofevaluation(Isenetal., affectivestatewillbemorepronouncedforads
1978).Anotherexplanation,knownasaffect-as-information, thathaveanambiguous-affectivetonethanfor
is that people often inspect their feelings when making ads that have a clear affective tone.
evaluative judgments (Schwarz & Clore, 1988; see also
Pham, 1998). During this process, pleasant or unpleasant Arousal and Polarization
feelingsemanatingfromthemoodstateatthetimeofjudg-
ment may be misattributed to the target (Schwarz & Clore, Asubstantialamountofresearch(seeReisenzein,1983)has
1983). Forgas(1995)suggestedthatthesetwoexplanations suggested that arousal experienced ata given point in time
areinfactcomplementaryandapplytodifferentprocessing may accentuate subsequent affective and evaluative re-
strategies. When the person uses a systematic processing sponses. Consumers’ responses to a given object—for in-
strategy, mood will primarily influence judgments through stance,anad—maybemoreextremeorpolarizedifthecon-
thecuingofmood-congruentthoughts.Whenthepersonuses sumershavebeenrecentlyaroused.Twoexplanations have
a heuristic processing strategy, mood will influence judg- been offered for this effect. One invokes attributional pro-
mentsthroughaffect-as-informationmechanisms.Still,both cesses, and the other invokes processing interferences.
explanations suggest that the valence of people’s affective
stateatthetimeofexposuretoanadwillcoloritsevaluation. Polarization as misattribution. BuildingonSchachter
Research has suggested that this coloring effect may be andSinger’s(1962)work,Zillmann(1971)offeredtheexcita-
contingentonanumberoffactors(Cloreetal.,1994;Martin, tiontransferhypothesis.Itiswellestablishedthatarousal,asa
Abend,Sekides,&Green,1997;Pham,1998).Ofparticular generalized state of activation, tends to be nonspecific (e.g.,
interestisresearchshowingthatmood-congruentevaluations Schachter&Singer,1962).Moreover,arousaldoesnotdissipate
maybemorepronouncedwhenthetargetsareaffectivelyam- immediately after the removal of the arousing condition:
MOOD VALENCE, AROUSAL, AND AFFECTIVE TONE 45
Arousal decays relatively slowly (e.g., Cantor, Zillmann, & many ads have a positive affective tone (e.g., humorous,
Bryant,1975).Asaresult,theresidualarousalexperiencedfol- warmcommercials),thepositivityofthistoneislikelytovary
lowinganeventmaypolarizetheresponsetoasubsequenttarget (e.g., some commercials are warmer than others). Study 1
by intensifying the affect elicited by this target. For instance, teststhehypothesisthatthecoloringeffectofvalencewillbe
White,Fishbein,andRutstein(1981)observedthatarousal,in- morepronouncedwhentheaffectivetoneofthetargetadis
duced through eitherphysical exercise or emotional material, ambiguousthanwhenitisclearlypositive.Thepolarization
subsequentlyincreased(decreased)maleparticipants’likingof effectshouldmanifestitselfinthereverseconditioninwhich
an attractive (unattractive) female target. Apparently, partici- theaffectivetoneofthetargetadisunambiguous(asopposed
pantsmisattributedtheresidualarousalleftbytheprioreventas toambiguous).InStudy2,wetestthehypothesisthatarousal
partoftheirgenuineaffectiveresponsetothefemaletargetand willhaveapolarizationeffectwhenthead’saffectivetoneis
thereforeexpressedmorepolarizedjudgments.Thisintensify- clearlypositiveornegative.Undersuchconditions,valence
ingeffectofresidualarousalhasbeenreplicatedinmanystudies will have little influence.
(forareview,seeReisenzein,1983).
Polarization as reduced cognitive complexity. The STUDY 1
dynamic complexity hypothesis (Paulhus & Lim, 1994)
statesthatarousalpolarizesevaluativejudgmentsbyreduc- Thepurposeofthisstudywastotestthepropositionthatthe
ingthecomplexityofperceptionsaboutatarget.Thishypoth- coloring effect of consumers’ mood states on their evalua-
esiselaboratesonthetheorythatbecausearousalreducespro- tionsofadvertisementswoulddependontheaffectiveambi-
cessing capacity, aroused individuals tend to selectively guityofthead.Becausetheexperimentfocusedonthecolor-
processimportantcuesattheexpenseoflessimportantones ing effect of affect states, participants’ mood was
(Easterbrook, 1959). Because the primary dimensionof so- manipulatedtobe eitherpleasantor unpleasant, controlling
cialperceptionisevaluative(Osgood,Suci,&Tannenbaum, forvariationinarousal.Followingthemoodinduction,par-
1957),selectiveprocessingfocusingonthisdimensionatthe ticipantsweresubsequentlyexposedtothetargetadinasup-
expense of nonevaluative dimensions will polarize judg- posedly unrelated task. In one condition, the ad’s affective
ments(Paulhus&Lim,1994).Inastudyofpeople’sevalua- tonewasclearlypositive.Intheothercondition,thead’saf-
tionoffamousfiguresandsocialacquaintances,Paulhusand fecttonewasmoreambiguous.Itwaspredictedthattheva-
Limfoundthatarousaldidresultinsimplerperceptionsand lence of participants’moodwouldcolortheirevaluationof
thatthesesimplerperceptionsseemedtoleadtomorepolar- the target ad in the ambiguous-ad condition but not in the
ized judgments. Consistent withthis hypothesis, itwas ob- clearly positive-ad condition.
servedinan advertisingcontext that arousal could polarize
brandevaluationby increasing people’sreliance on which- Method
ever cues—good or bad—were perceived to be more diag-
nostic (Pham, 1996). Participants and design. Participants consisted of
Therefore, there is a strong theoretical reason tobelieve 80 business undergraduateswhoparticipatedin thestudy
thatthearousalcomponentofconsumers’affectivestatesat to fulfill a course requirement. They were randomly as-
thetimeofexposuretoanadvertisementshouldpolarizeits signedtooneoffourconditionsina2×2between-subjects
evaluation. This effect should depend on the ad’s affective design.Thefirstfactormanipulatedtwolevelsofaffective
tone.Ifthead’sprimaryaffectivetoneispositive,evaluations state: pleasant versus unpleasant. The second factor ma-
shouldbemorefavorableasarousalincreases.Ifthead’saf- nipulatedtwolevelsoftheaffectivetoneofthead:positive
fectivetoneisnegative,evaluationsshouldbelessfavorable versusambiguous.
as arousal increases. In principle, thispolarization effect of
arousal shouldbe independent of anycoloring effect of the Procedure. Theexperimentwasintroducedastwoun-
valence component of the affective state. relatedstudies,conductedintwoseparatesmallrooms.The
“first”studywaspurportedlyaboutmusicalappreciation.To
H2: Whenadshaveaclearlypositiveornegativeaf- enhanceexperimentalcontrol,thestudywasconductedone
fectivetone,higharousalwillpolarizeadeval- participant atatime. Participantswere seatedina comfort-
uations in the direction of this tone. ablechairwhileoneofthetwoselectedmusicalpieces(see
Pretest1next)wasplayedthroughahigh-qualitycassettere-
Insummary,thetwomaincomponentsofconsumers’af- corder. Participants were instructed to listen carefully and
fectivestatesatthetimeofexposuretoanadshouldhavedis- formimpressionsof,andopinionsabout,themusic.Afterlis-
tincteffectsonitsevaluation.Thevalencecomponentshould teningtothemusicfor5min,participantswereaskedtoas-
colortheevaluationoftheadinamood-congruentdirection, sess the music on a number of dimensions (e.g., fast–slow,
whereasthearousalcomponentshouldpolarizeitsevaluation quiet–loud,andinteresting–uninteresting).Participantswere
inthedirectionofthead’sprimaryaffectivetone.Although thendirectedtothe“second”study,whichwaspurportedlya
46 GORN, PHAM, SIN
pilotadvertisingtestconductedbythemarketingdepartment smilingchild.Intheambiguous-affective-tonecondition,the
fora private company. Participantswere exposed to one of picturewasamoredistantshotofamanandwomantakinga
thetwotargetads(seePretest2next)for40sec.Theythen walk.Asidefromthepicture,thetwoadswereidenticalinev-
evaluated the ad on four 7-point items: “I like (dislike) the eryrespect. Bothversions of the ad had a headline reading
ad,”“Theadisgood(bad),”“Theadlooks(doesnot)profes- “YOUCANENJOYMOREFROMLIFE!”andashortcopy
sionally made,” and “I feel positive (negative) towards the about the benefits of insurance with the company.
ad.”Theseitems(a=.84)wereaveragedintoasinglemea- Thetwoversionsoftheadwereshowntoseparategroups
sure of ad evaluation. ofparticipants(N=71).Theseparticipantswereaskedtoex-
aminetheadandassesshowitmadethemfeel.Theaffective
toneoftheadwasassessedbyfour7-pointsemanticdifferen-
Pretest 1: Music selection. Becauseitisimportantto
tialitems(a=.92)anchoredat“happy–sad,”“pleased–dis-
separatethevalenceandarousaldimensionsofaffectivestates,
pleased,” “delighted–distressed,” and “joyful–depressed.”
we investigatedmanipulationsof valence that would control
The arousal contentofthe adwasassessedby three7-point
forthelevelofinducedarousal.Weelectedtorelyonmusicfor
semantic differential items anchored at “stimulated–re-
severalreasons.First,musichasbeenrepeatedlyshowntobe
laxed,” “excited–calm,” and “aroused–unaroused.” As in-
aneffectivemanipulationofvalenceofaffectivestates,bothin
tended,theadwiththehappyfamilywasjudgedtobemore
marketing studies (e.g., Alpert & Alpert, 1990; Kellaris &
positiveintoneandelicitedmorepositivefeelings(M=5.09)
Kent,1993;Miniardetal.,1992)andinpsychologicalstudies
thandidtheadwiththecouple walking(M =4.40), t(69) =
(e.g.,Eich&Metcalfe,1989).Second,comparedtomostso-
3.65,p<.01.Thetwoadswerenotdifferentinthefeelingsof
cial manipulations of affective states—such as performance
arousaltheyelicited(positivetonead:M=3.45andambigu-
feedback,giftgiving,andinteractionwithaconfederate—mu-
ous tone ad:M= 3.26),t< 1.
sicisrelativelycontentfree.Finally,thereisanenormousvari-
abilityofmusicavailableinthemarketplace.Thisfacilitated
thesearchforpiecesthatwouldinducedifferentlevelsofva-
Results
lencewithoutinducingdifferentlevelsofarousal.
Ourgoalinpretestingwastofind2 piecesofmusic that
Manipulation checks. To check the perceived affec-
wereequallyarousingbutonethatwasmorepleasingthanthe
tivetoneoftheadinthemainexperiment,participants’rat-
other. In a preliminary pretest (n = 86), 44 pieces of music
ingsofthepleasantnessofthead(astheyagainlookedatit)
wereevaluatedonpleasureandarousal.Participantswererun
weresubmittedtoa2(adtone)×2(valence)analysisofvari-
insmallgroupsof4to6people,witheachgroupevaluating6
ance(ANOVA).Asexpected, theANOVArevealedonly a
piecesofmusic.Basedonthispreliminarypretest,2piecesof
maineffectofadtone,F(1,79)=9.23,p<.01,showingthat
musicwereselectedtobetestedinamoreformalpretestin
theperceivedpleasantnessoftheadwassignificantlyhigher
which 23 participants were run individuallyandlistened to
inthe positive-adcondition(M = 4.88)thanintheambigu-
only1pieceofmusic.Afterlisteningtoeachpiecefor5min,
ous-ad condition (M = 4.30). Although the positive ad’s
participantsratedhowpleasing and arousingthemusicwas
pleasantness was clearly above the midpoint of the scale,
on7-pointscales.Asexpected,thefirstpiece(“EineKleine
t(39) =8.54,p<.001,theambiguousad’spleasantnesswas
Nachtmusik:Allegro”byMozart)wasratedaspleasing(M=
onlymarginallyabovethemidpointofthescale,t(39)=1.91,
5.70)andhighinarousal(M=5.70).Thesecondpiece(anIn-
p=.06. Themaineffectof valenceandtheAd Tone× Va-
dian classical piece by Pandit Dhimsen Joshi) was ratedas
lence interaction did not approach significance,F’s < 1.
muchlesspleasing(M=2.23)butstillhighinarousal(M=
Theaffective-statemanipulationwascheckedwithanin-
5.61).Itwasconfirmedby t-teststhatthe2 piecesof music
dependentsampleofparticipants(N=20)whowererunindi-
weredifferentinhowpleasingtheywere,t(21)=11.71,p<
vidually. Each participant listened for 5 min (the same
.001, but not in arousal,t(21) = .21,p> .10.
amountoftimeasinthemainexperiment)tooneofthetwo
piecesofmusicusedinthemainexperiment.Afterlistening
Pretest 2: Affective tone of the ad. Asecondpretest tothepiece,participantswereaskedtoreportontheiraffec-
was conducted to identifytwo advertisements inwhich the tive state using the Affect Grid (Russell, Weiss, &
content wouldvaryinterms of their affective tone. One ad Mendelsohn,1989),whichisagraphicalscalethatassesses
shouldhaveaclearpositiveaffectivetone,whereastheother bothfeelingsofpleasureandarousalona9-pointscalerang-
shouldhaveamoreambiguousaffectivetone.Awiderange ingfrom1(low)to9(high).Asexpected,thepleasantmusic
ofadswereinitiallyscreenedbeforesettlingonthefollowing inducedhigherpleasure (M =7.10)thandidthe unpleasant
approachfortheformalpretest.Adsforafictitiousinsurance music(M =3.0), t(18)=10.83,p< .001.However,thetwo
companywereprofessionallyproduced.Acentralfeatureof piecesinducedsimilarlevelsofarousal(M =7.60and
pleasant
theadwasa picture,whichvariedacrossconditions.In the M = 7.10), t(18) = 1.41, p >.10. Therefore, the two
unpleasant
positive-affective-tone condition, the picture depicted a piecesinduceddifferentfeelingsofpleasantnessbutnotdif-
happyfamily: a smilingfatherwith a smiling motheranda ferent levels of arousal.
MOOD VALENCE, AROUSAL, AND AFFECTIVE TONE 47
Ad evaluation. Adevaluationsacrossconditionswere It is possible that mood-congruency effects would have
analyzed in a series of planned contrasts (Rosenthal & beenobservedintheunambiguous-adconditionhadweuseda
Rosnow,1985).Afirstcontrastshowedthatintheambigu- stronger mood manipulation. However, our manipulation (a
ous-adcondition,participantsinapleasantaffectivestatein- 4.1difference on a 7-point scale; w2= .85)wasquite strong
deedreportedmorefavorableevaluations(M=3.66)thandid compared to the mood manipulations used in other studies
participantsinanunpleasantaffectivestate(M=3.16),t(76) where moodcongruencywas indeedobserved(e.g., Gorn et
=1.65,p=.05,onetailed(w2=.02).Thiseffectreplicatesthe al.,1993;Schwarz&Clore,1983).3Thissuggeststhat,inreal
standard mood-congruency effect. In contrast, in the posi- consumersettings,mood-congruencyeffectsonadevaluations
tive-adcondition,thevalenceofparticipants’affectivestates maybelesspervasiveandpronouncedthanimpliedbyprevi-
hadnoeffectsonadevaluations,t<1.Participantsinthiscon- ousresearch(e.g.,Goldberg&Gorn,1987).
ditionreportedcomparableevaluationsregardlessofwhether Giventhatvalenceappearstocoloradevaluationonlywhen
theywereinapleasant(M=3.64)orunpleasant(M=3.84)af- theadisaffectivelyambiguous,thequestionariseswhethercon-
fective state. Consistent with the planned contrasts, which sumers’affectivestateshaveanyinfluenceatallonadevalua-
were the focus of the analyses, a 2 (valence) × 2 (ad tone) tions when the ad is not affectively ambiguous (i.e., clearly
ANOVA revealed a marginally significant Ad Tone × Va- positiveorclearlynegative).ThisissueisexaminedinStudy2.
lenceinteraction,F(1,76)=2.66,p=.10(w2=.02).More-
over, as expected, both the main effect of valence and the
maineffectofadtonewerenotsignificant,F<1andF(1,76) STUDY 2
=2.29,p>.13,respectively,suggestingthatvalencedidnot
coloradevaluationsindependentlyofthead’saffectivetone. Themainobjectiveof thisstudywastotesttheproposition
Overall, the results are consistent withourhypothesis (H1) that when ad targets have a clear affective tone, it is the
thatthecoloringeffectofvalencedependsonwhetherthetar- arousalcomponentofconsumers’affectivestates,andnotthe
get ad has a clear affective tone.2 valencecomponent,thatwillaffecthowtheadwillbeevalu-
ated. Participants whose affective state was manipulated in
termsofbothvalenceandarousalweresubsequentlyexposed
Discussion toatargetadthatwaseitherclearlypositive initsaffective
tone or clearly negative. It was predicted that high arousal
TheresultsareconsistentwithH1andwiththegeneralnotion wouldpolarizeadevaluationsinthedirectionofthead’saf-
that ad evaluations may be influenced by an interplay be- fectivetone.Incontrast,assuggestedbytheresultsofStudy
tweenthead’saffectivetoneandconsumers’affectivestates. 1,valencewouldhavelittleinfluenceonhowtheunambigu-
Whenthetargetad’saffectivetonewasambiguous,adevalu- ously positive or negative ad would be evaluated.
ationwasmorefavorableamongparticipantsinapleasantaf- Should a polarization effect of arousal be uncovered, a
fectivestatethanamongparticipantsinanunpleasantaffec- secondobjectiveofthisstudywastoexplorealternativeex-
tive states. Thiscoloringeffect of valence on adevaluation planationsofthiseffect.Accordingtotheexcitationtransfer
replicatesthewell-establishedmoodcongruencyeffect.This hypothesis (e.g., Cantor et al., 1975; Zillmann, 1971), high
coloringeffectdidnotoccur,however,whenthead’saffec- arousalshouldpolarizesubsequentadevaluationsifconsum-
tive tone was clearly positive. Although mood-congruency ersbelievethatthecontext-inducedarousalreflectstheirgen-
effects on evaluation have been considered to be pervasive uineaffectivereactionstotheaditself.Consumersinspecting
androbust(e.g.,Mayeretal.,1992;Schwarz,1990),theyap- “how they feel” about anad(e.g., Pham, 1998; Schwarz &
pear to be sensitive to the affective ambiguity of the target Clore, 1988) may infer from their arousal that they “feel
(e.g.,Isen&Shalker,1982;Miniardetal.,1992).Thiscontin- strongly”abouttheadandhenceevaluateitmoreextremely.
gencyisnoteworthygiventhatmanyadvertisementsarenot We examined this explanation by comparing responses to
affectively ambiguous. In this study, affective ambigu- ad-evaluationitemsphrasedinself-referentialmanner(e.g.,
ity—andtherebythecoloringeffectofvalence—wasreduced “Ilikethead”and“Ireactfavorablytothead”)toresponsesto
by the mere modification of an execution cue. items phrased in object-referent manner (e.g., “The ad is
good” and “The ad is enjoyable”). It was reasoned that
self-referential items should be particularly sensitive to a
transferofexcitationbecausetheseitemsemphasizepartici-
2Alternatively,onecouldinterpretthedataasshowingthattheaffective
toneoftheadhadastrongereffectonadevaluationsamongparticipantsinan
unpleasantmoodthanamongparticipantsinapleasantmood.Thisinterpre-
tationwouldbeconsistentwithpreviousargumentsthatnegativemoodin-
creasesanalyticalprocessingandmessagescrutiny,whereaspositivemood 3InGorn,Goldberg,andBasu(1993),themoodmanipulationinduced
decreasesit(e.g.,Schwarz,Bless,&Bohner,1991).Thisinterpretationdoes differencesof2.38(w2=.29)and3.16(w2=.22)onthe10-pointscalesthat
notseemtobeapplicabletothisstudy.Theadsdifferedonlyonasingleexe- wereusedtoassessmood.InSchwarzandClore(1983;Study1),themood
cutioncharacteristicthatwassimpletoprocess.Therefore,depthofmessage manipulationinducedadifferenceof1.8ona7-pointscale(w2=.22)anda
scrutiny should have made little difference. difference of 2.6 on an 11-point scale (w2= .25).
48 GORN, PHAM, SIN
pants’ reactions tothe ad. In contrast, object-referent items pieceswasplayed(seePretest3next).Participantswerein-
should be less sensitive to a transfer of excitation because structedtolistencarefullyandformimpressionsofthemusic.
they emphasize the ad itself (for a related discussion, see AsinStudy1,5minfollowingthemusiconset,participants
Wyer,Clore,&Isbell,1999).Inotherwords,ifthepolariza- completedaquestionnaire.Thequestionnaireincluded(a)a
tioneffectofarousalisindeedmediatedbyatransferofexci- 22-item scale assessing perceptions of the music (e.g.,
tation,thiseffectshouldbemorepronouncedonself-referent fast–slow, quiet–loud, and simple–complex); (b) a 3-item
items (e.g., “I like–dislike the ad”) than on object-referent scalemeasuringevaluationsofthemusic(e.g.,like–dislike);
items(e.g.,“Theadisgood–bad”).Thedynamiccomplexity and(c)theAffectGrid,whichservedasamanipulationcheck
hypothesiswouldnotpredictthatpolarizationwoulddepend of participant’s states of pleasure and arousal. Participants
on the phrasing of the items. were then directed to the “second” study.
According to the dynamic complexity hypothesis The“second”studywasagainintroducedasapilotadvertis-
(Paulhus & Lim, 1994), high arousal may polarize subse- ing test conducted by the marketing department for a private
quentadevaluationsifconsumersprocessasimplerrepresen- company.Eachparticipantwasexposedfor40sectooneoftwo
tationoftheadandiftheprimarycomponentofthesimplified ads.Oneadhadaclearpositiveaffectivetoneandtheotherhada
representationisevaluative.Participants’representationsof clear negative affective tone (see Pretest 4 next). Participants
thetargetadwereassessedthroughcognitiveresponses.The then completed a questionnaire, which assessed their evalua-
dynamic complexity hypothesis would predict that aroused tionsoftheadandtheircognitiveresponseswhilereadingthead.
consumersshouldreport(a)fewerthoughtsinresponsetothe Attheendofthesession,participantswereorallyaskedwhatthe
adand(b)ahigherproportionofevaluativetononevaluative purposeofthestudywasandtowritedowntheiranswersona
thoughts. The excitation transfer explanation does not hy- separatesheetofpaper.Theywerethendebriefed.
pothesizeanyspecificchangeincognitiverespondingunder
high arousal.
Measures. The main dependent measure, ad evalua-
tion,wasassessedonseven7-pointsemanticallydifferential
Method
items(a=.88):“Theadispleasant–unpleasant,”“Theadis
good–bad,”“Theadisenjoyable,”“Ilike–dislikethead,”“I
Participants and design. Participants in this study
reactfavorably–unfavorablytothead,”“Ifeelnegative–posi-
consistedof128businessundergraduateswhoparticipatedto
tivetowardthead,”“Theadisfun–notfuntoread.”Partici-
earn course credit. They were randomly assigned to one of
pantswerealsoaskedtolistallthethoughtsandfeelingsthat
eightconditionsofa2(arousal)×2(valence)×2(adaffective
cametomindwhilereadingthead.Twojudges,workingin-
tone)between-subjectsdesign.UnlikeinStudy1,theaffec-
dependently,classifiedeachthoughtorfeelingaseitherposi-
tivetoneoftheadwaseitherclearlypositiveorclearlynega-
tive,negative,orneutral(agreement=95%).Disagreements
tive.Furthermore,valenceofaffectivestate(pleasantvs.un-
were resolved by a third judge.
pleasant) was crossed with two levels of arousal (high vs.
low). Crossingthese two factors presents three advantages.
First,itreducesthelikelihoodthateithermanipulationiscon- Pretests 1 and 2: Valence and arousal of affective
founded with the other. One could argue, for instance, that states. Weagainusedmusictomanipulatebothvalenceand
states of high arousal are less pleasant than states of lower arousalofaffectivestates.Althoughmusichasoftenbeenused
arousal.Second,itallowsustotestthenullpredictionthatva- to manipulate affect valence, the use of music to manipulate
lenceoftheaffectivestatewillnothaveaneffectwhentheaf- arousalorthogonallytovalenceisauniquefeatureofthisstudy.
fectivetoneoftheadisunambiguouslypositiveornegative, Inafirst,within-subjectpretest,22participants,ingroups
andtotestitacrossbothhigherandlowerlevelsofarousal. of4to6,wereaskedtolistento16piecesofmusic(asubsetof
Third,previousresearchonarousaleffectsonpersuasionhas the48piecespretestedinStudy1),onepieceatatimefor2
ignored potential interactions between arousal and valence min.Afterlisteningtoeachpiece,participantsreportedtheir
(Pham, 1996; Sanbonmatsu & Kardes, 1988). Crossing the mood states using the Affect Grid. Based on this pretest, 4
twofactorsallowsus toexaminesuchinteractions.In addi- musical pieces were selected. They were: (a) “Eine Kleine
tiontothesebetween-subjectsfactors,awithin-subjectsfac- Nachtmusik:Allegro”byMozart,whichinducedhighplea-
torwasformedbytheinclusionoftwotypesofad-evaluation sure (M = 7.23) and high arousal (M = 7.36); (b)
items (self-referential vs. object referential). “Saraseeruhasana” by Naada Loludai, which induced low
pleasure(M=3.14)andhigharousal(M=6.73);(c)“What-
Procedure. The procedureclosely followedthat used everWeImage”byDavidFoster,whichinducedhighplea-
inStudy1.Thesessionswereagain runoneparticipantata sure (M= 6.91)andlowarousal(M = 4.14); and(d)“Raga
time.Thestudywasintroducedastwounrelatedstudies,con- Bhopali” by Mohanam, which induced low pleasure (M =
ductedindifferentrooms.The“first”studywaspurportedly 3.73) and low arousal (M= 3.50).
about musical appreciation. Participants were seated in a Inasecond,between-subjectspretest(N=31),eachofthe
comfortableloungechairwhileoneoffourselectedmusical fourselectedpieceswasplayedtoaseparategroupofpartici-
MOOD VALENCE, AROUSAL, AND AFFECTIVE TONE 49
pantsinasoundproofroom.Afterlisteningtothepiecefor5 generalpurpose.Theirdatawerediscardedfromtheanalysis,
min,participantsreportedtheirmoodsusingtheAffectGrid. performedon125observations.Toassessthemanipulations
Theseself-reportsofpleasureandarousalweresubmittedto2 of affective states, participants’ self-ratings of arousal and
× 2 ANOVAs in which the two factors were the level of pleasureafterlisteningtothepieceofmusicwereeachsub-
arousal(highorlow)andvalence(pleasantorunpleasant)ex- mittedtoa2(affectdimension)×2(valencemanipulation)×
pectedtobeinducedbyeachpiece.Asexpected,bothfactors 2 (arousal manipulation)mixedANOVA. As expected, the
had strong main effects on the corresponding self-reports. analysisuncoveredastrongAffectDimension×ArousalMa-
Self-reportsofpleasureweregreaterwhenthemusicalpieces nipulation interaction, F(1, 119) = 153.07, p < .0001 (w2=
werepleasant(M=6.43)thanwhentheywereunpleasant(M .55),aswellasastrongAffectDimension×ValenceManipu-
= 3.60), F(1, 27)= 28. 83, p < .0001 (w2= .44). Similarly, lationinteraction,F(119)=99.37,p<.0001(w2=.44).Fol-
self-reportsofarousalweregreaterwhenthemusicalpieces low-upanalysesshowthatthearousalratingswerestrongly
were highly arousing (M = 6.70) than when they were less influencedbythearousalmanipulation(M=3.76vs.6.97),
arousing (M = 4.39), F(1, 27) = 6.92, p < .02 (w2 = .16). F(1,119)=182.06,p<.0001(w2=.58),andtheyweremuch
Self-reportsofpleasurealsoexhibitedasmallbutsignificant lessinfluenced,althoughsignificantlyso,bythevalencema-
Arousal×Valenceinteraction,F(1,27)=7.33,p<.02(w2= nipulation,F(1,119)=5.51,p<.03(w2=.01).Similarly,the
.06), showing that the valence manipulation was slightly pleasureratingsweremostlyaffectedbythevalencemanipu-
strongerforthemorearousingpieces.Noothereffectswere
lation(M=3.61vs.7.01),F(1,119)=358.65,p<.0001,(w2=
significant,allps>.20,suggestingthatthefourpiecesmanip-
.67).TheywerealsoaffectedbyasmallValence×Arousal
ulatedpleasureandarousallargelyindependently.Thisma-
StateManipulationinteraction,F(1,119)=5.31,p<.05(w2=
nipulation was further assessed in the main experiment.
.01).Insummary,thefourmusicalpiecessucceededinma-
nipulatingpleasureandarousalstateslargelyindependently,
Pretests 3 and 4: Affective tone. The purpose of
although not perfectly orthogonally.
thesepretestswastoidentifytwoadvertisementswithaclearly
positiveor negativeaffective tone.In Pretest 3, 64 participants
Effects on ad evaluation. Ad evaluations were sub-
wereshownoneofeightprintadvertisements.AsinStudy1,the
mittedtoa2(arousal)×2(pleasure–valence)×2(affectivetone)
adswereforafictitiousinsurancecompanyandwereprofession-
between-subjectsANOVA.Themeansacrossconditionsarere-
allyproducedtoinducedifferentlevelsofaffectivetone.Thepos-
portedinTable1.Theanalysisrevealedamaineffectofaffec-
itive-affectivetoneadwasthesameastheoneusedinStudy1.It
featuredthepictureofahappyfamily,thecopystressedthebene- tivetone,F(1,117)=79.90,p<.0001(w2=.39).Evaluations
fitsoftheinsurancecompany,andtheheadlineread“YOUCAN weremorefavorableforthepositive-affectivetoneversionof
ENJOYMOREFROMLIFE!”Thegoalinconstructinganega- thead(M=4.17)thanforthenegativeaffectivetoneversion(M
tiveadwastodifferentiateitsaffectivetoneasmuchaspossible =2.89).Asexpected,giventheaffectiveunambiguityoftheads,
fromthatofthepositivead.Thenegativeaffectivetoneadfea- the valence of participants’ affective states did not have any
tured the picture of a distressed-looking girl described as a mainorinteractioneffectontheiradevaluations,F<1.Partici-
5-year-oldorphanwhoseparenthaddiedinacaraccident.The pantswhowereinapleasantaffectivestatewerenotmorefavor-
headlineoftheadread“SHECOULDHAVELIVEDBETTER! able intheiradevaluations(M = 3.49)thanwereparticipants
PROTECTYOURCHILDRENANDTHEIRFUTURE.”The whowereinanunpleasantaffectivestate(M=3.58).Thisresult
affectivetoneoftheadswasassessedbyfour7-pointsemantic isconsistentwiththeresultsofStudy1.
Moreimportant,therewasasignificantArousal×Affec-
differentialitems(a= .95):“Thisadmademe feel happy–un-
tiveToneinteraction,F(1,117)=4.22,p<.05(w2=.03).As
happy,” “This ad made me feel pleased–displeased,” “This ad
predicted, the simple effect of the ad’s affective tone was
mademefeeldelighted–distressed,”“Thisadmademefeeljoy-
greater under high arousal (M = 4.41 vs. M =
ful–depressed.”Asdesired,thetwoadsdifferedinaffectivetone positive negative
2.85),F(1,117)=58.63,p<.0001(w2=.32)thanunderlow
(M=2.69vs.M=5.27),t(16)=5.58,p<.01.Toensurethatthe
arousal (M = 3.94 vs. M = 3.02), F(1, 117) =
twoads were equal in terms of overall quality, 20 participants positive negative
24.29, p<.0001(w2=.16).Thispatternofresultsindicates
(Pretest4)evaluatedtheadsafterlisteningfor5mintoapieceof
thathigharousalpolarizedadevaluationsinthedirectionof
music(selectedonthebasisofthepreliminarypretestreportedin
thead’saffectivetone,whichsupportsH2.4Noothereffects
Experiment1)designedtoinduceaneutralmood.Asexpected,
thetwoversions of the adwere evaluated similarly(M =
positive
5.02andM =4.88),t<1.
negative
4Wealsotestedthesimpleeffectsofarousalwithineachlevelofaffective
Results toneofthead.Consistentwiththepolarizationhypothesis,inthepositive-af-
fectivetonecondition,adevaluationsweresignificantlymorefavorableunder
higharousalthanlowarousal,F(1,117)=5.24,p<.05.Inthenegativeadtone
Demand and manipulation checks. Three partici-
condition,adevaluationsweresomewhatlessfavorableunderhigharousal
pants expressed at least some understanding of the study’s thanunderlowarousal.However,thedifferencewasnotsignificant,F<1.
50 GORN, PHAM, SIN
TABLE1 identicaltotalnumberofthoughts(M =3.21vs.M
higharousal low
AdEvaluationasaFunctionofValence,Arousal,
=3.16),F<1.Asimilaranalysiswasthendoneonthe
andAffectiveTone arousal
proportionofvalencedthoughts(i.e.,[positive+negative]/
PositiveAffectiveTone NegativeAffectiveTone total).Itrevealedamaineffectofaffectivetone,F(1,115)=
13.39,p<.001(w2=.09).Thenegativelytonedversionofthe
Pleasant Unpleasant Pleasant Unpleasant
Affective Affective Affective Affective adpolarizedparticipants’thoughtsmorestrongly(M=86%)
Arousal State State State State than did the positively toned version (M = 66%). This was
mostlyby increasingthenumberof negative thoughts, F(1,
Low 3.89a 4.02a 2.95a 2.98a
115)=5.76,p<.02(w2=.04),anddecreasingthenumberof
High 4.23a 4.56b 2.85b 2.78b
neutral thoughts, F(1, 115) = 17.70, p < .001 (w2 = .12).
an= 16.bn= 15.
Arousal itself did not increase the proportion of valenced
thoughtsaswouldbe predictedby the dynamic complexity
weresignificant.Thetwodifferentexplanationsofthispolar-
hypothesis, F< 1.
ization effect are examined next.
Furtheranalysisshowedthatarousaldidinfluencepartici-
Recallthat,accordingtotheexcitationtransferexplanation,
pants’cognitiveresponses—albeitnotinthemannerpredicted
preexistingarousalpolarizesevaluationsbecausethisarousal
bythedynamiccomplexityhypothesis.Wecomputedthenet
ismisattributedasbeingpartofagenuineaffectiveresponseto
valenceofparticipants’thoughtsbysubtractingthenumberof
thetarget.Shouldthisexplanationbecorrect,thepolarization
negative thoughts from the number of positive thoughts. A
effectofarousalshouldbemorepronouncedonitemsstressing
three-wayANOVAonthisnetvalenceindexrevealedamain
the participants’ reactions to the ad (e.g., “I like–dislike the
effectoftheaffectivetoneofthead,F(1,117)=4.97,p<.03
ad”)thanonitemsstressingtheaditself(e.g.,“Theadisenjoy-
(w2 = .03). Expectedly, the net valence of participants’
able–notenjoyable”).Wetestedthisreasoningbyexamining
thoughtswasmorenegativeinthenegative-adcondition(M=
theArousal×Toneinteractionforeachevaluationitem,sepa-
–1.45)thaninthepositive-adcondition(M=–.70).Moreim-
rately.As canbe seeninTable 2, the polarization effect ap-
portant,therewasasignificantArousal×AffectiveToneinter-
peared to be consistently more pronounced for the items
action,F(1,117)=4.58;p<.04(w2=.03).Although,under
phrasedinaself-referentialformatthanfortheitemsphrasedin
lowarousal,theaffectivetoneoftheadhadnoinfluenceonthe
a object-referent manner. A post hoc contrast of the pooled
net valence of participants’ thoughts (M = –1.31 vs.
positive
self-referential items and the pooled object-referent items
M =–1.34),F<1,underhigharousal,thenetvalenceof
negative
showsthatthesimpletwo-wayinteractionbetweenArousal×
the thoughts was significantly more negative among partici-
Tone was significantly greater for the self-referential items,
pantsexposedtothenegativetonead(M=–1.57)thanamong
F(1,117)=5.71,p<.02(w2=.04)thanfortheobject-referent
participantsexposedtothepositivetone ad(M =–.06), F(1,
items,F(1,117)=1.90,p=.17(w2=.01).Thiscontrastwasre-
117) = 9.32, p < .01 (w2 = .06). Therefore, the influence of
flectedinasignificantthree-wayArousal×AffectiveTone×
arousalonthenetvalenceofparticipants’thoughtslargelypar-
TypeofReferenceinteraction,F(1,117)=4.47,p<.04(w2=
alleleditseffectonadevaluation.5
.03),whichisdepictedinFigure1.Thispatternoffindingsis
consistentwithanexcitationtransferexplanation.Totestthe
Discussion
otherprocessexplanation,wealsoexaminedparticipants’cog-
nitiveresponses.
Theresultsofthisstudyclearlysupportthehypothesis(H2)
that when an ad has an unambiguous affective tone, the
Effects on cognitive responses. Table 3 summa-
arousal component of consumers’affectivestatesmaypo-
rizestheeffectsofarousalandaffectivetoneonparticipants’
larize ad evaluations in the direction of the ad’s affective
cognitiveresponses.Themeansarecollapsedacrosstheva-
tone.Itwasfoundthatparticipantsevaluatedtheadmorefa-
lencefactor,whichdidnothaveanysignificantinfluenceon
vorably when it hada positive-affective tone than whenit
theseresponses.Asdiscussedearlier,thedynamiccomplex-
hadanegative-affective tone.However,thistendencywas
ity hypothesis postulates that high arousal narrows percep-
significantlystrongerinthehigh-arousal conditionthan in
tionstoaprimarydimensionthatisevaluative.Asimplifica-
thelow-arousalcondition.
tion would imply fewer ad-related thoughts, whereas
Althoughexploratory,theresultsalsoshedsomelighton
narrowingtoanevaluativedimensionwouldimplyagreater
theprocessunderlyingthispolarizationeffect.Thereseemed
proportionofvalencedtononvalencedthoughts.Thesecorre-
tobelittleevidencethatthiseffectwascausedbythetypeof
lated predictions were tested as follows. Participants’ total
evaluative attention narrowing postulated by the dynamic
number of thoughts were submitted to a 2 (arousal) × (va-
complexity hypothesis. Although null effects should, of
lence)×2(adtone)ANOVA.Theanalysisdidnotrevealany
significanteffectofthesefactorsonthetotalnumberofcog-
nitiveresponses(smallestp=.19).Inparticular,participants
5A mediation analysis confirmed that the valence of participants’
in the high- and low-arousal conditions reported an almost thoughts mediated the polarization effect of arousal on ad evaluation.
MOOD VALENCE, AROUSAL, AND AFFECTIVE TONE 51
TABLE2 Theevidencewasmorecongenialwiththeexcitationtrans-
Arousal×AdToneInteractiononEachAdEvaluationItem
fer hypothesis. It was found that the polarization effect of
Items(IntheOrderTheyWereAssessed) FValue pValue arousalwasstrongerwhentheadevaluationitemsemphasized
participants’reactionstothead(“I___thead”)thanwhenthe
The ad is pleasant–unpleasanta 0.21 .65
itemsfocusedontheaditself(“Theadis___”).Thispatternof
The ad is bad–gooda 2.78 .10
I like–dislike the adb 3.80 .05 resultsisconsistentwiththemisattributionmechanismpostu-
The ad is enjoyable–not enjoyablea 0.79 .38 latedbyexcitationtransfer.Consumerswhosestateofarousal
I react favorably–unfavorably to the adb 4.97 .03 hasbeenheightenedbyacontextualcause(e.g.,theprogram-
I feel negative–positiveb 4.45 .04 ming context) and who are then exposed to an affectively
The ad is fun–not fun to reada 0.29 .59
valencedadmayexperienceintensifiedaffectiveresponsesto
aObject-referentialitems(beginningwith“Thead…”).bSelf-referential the ad. They may infer from these intensified affective re-
items (beginning with “I … ”). sponsesthatthey“feelstrongly”aboutthead(eitherpositively
or negatively)andtherefore evaluate itmore extremely.The
differentialinfluenceofarousalandadtoneonself-referential
versusobject-referentitemsrulesoutthepossibilitythathighly
arousedparticipantsweresimplymorealertduringadevalua-
tion.Itwasalsofoundthatarousedparticipantshadpolarized
thoughts and feelings in response to the ad and that these
thoughtsandfeelingsmediatedtheextremityoftheirevalua-
tions. This finding suggests that the polarization effect of
arousalwasnotamereresponsescalingeffect,butagenuine
changeinparticipants’subjectiverespondingtothead.
Although the results appear to favor excitation transfer
overdynamiccomplexity,itisneverthelesspossiblethatour
analysesmaynothavebeenequallysensitivetothetwotypes
ofprocesses.Forinstance,theevidenceforexcitationtransfer
comesprimarilyfromad-evaluationscaleresponses.Incon-
FIGURE 1 Study 2: Arousal × Tone × Reference interaction. trast,theevidence(orlackofit)fordynamiccomplexityco-
mesfromopen-endedthoughtlistings,whicharelikelytobe
less sensitive thanclose-ended ratings. It therefore remains
TABLE3
possiblethatthepolarizationeffectofarousalwascausedby
CognitiveResponsesasaFunctionofArousal
andAffectiveTone bothmisattributionandattention-narrowingprocesses(e.g.,
Easterbrook, 1959) of the type postulated by the dynamic
PositiveAffective NegativeAffective
complexityhypothesis.AselaborateintheGeneralDiscus-
Tone Tone
sionsection,itisalsopossiblethatthetwoexplanationsapply
Low High Low High to different ranges of the arousal continuum.
Arousal Arousal Arousal Arousal Finally,theresultsconvergedwiththoseofStudy1insug-
gestingthatwhenthetargetadhasaclearaffectivetone,the
Positive thoughts 0.44 1.10 0.53 0.63
coloringeffectofvalenceislesslikelytobeobserved.This
Negative thoughts 1.75 1.16 1.88 2.20
Neutral thoughts 1.25 1.00 0.47 0.33 lackofavalenceeffectcannotbeattributedtoaweakmanip-
Total 3.44 3.26 2.88 3.17 ulation.Themanipulationcheckindicatesthat,asinStudy1,
Polarizedthoughts(%)a .63 .68 .85 .87 themanipulationofvalencewasstrong—a3.4differenceona
Net valenceb –1.31 –0.06 –1.34 –1.57
9-pointscale(w2=.67).Itsstrengthwas,infact,comparable
a(positive + negative)/total.b(positive – negative). to that of the arousal manipulation (w2= .58).
course,beinterpretedwithcaution,higharousaldidnotap- GENERAL DISCUSSION
peartoreducethenumberofthoughtsparticipantshadinre-
sponsetothead,anditdidnotincreasetheproportionofthese Summary
thoughtsthatwerevalenced.Alongwiththefindingthathigh
arousal increased (rather than decreased) the effects of the Adexposuresdonotoccurinavacuum.Foravarietyofrea-
targetad’stone,theabsenceofamaineffectofarousalonthe sons,consumersexposedtoadvertisementsmaybeinvarious
number of ad-related thoughts rules out the argument that affectivestates(e.g.,theyhavebeenrelaxingonthebeachvs.
highly aroused participants were more distracted during ad ridingacrowdedsubwayafterabusydayofwork).Itisthere-
evaluation than less aroused participants were. foreimportanttounderstandhowtheseaffectivestatesmay
52 GORN, PHAM, SIN
influenceadvertisingresponses,includingadevaluations.Of It is noteworthy that the arousal manipulation used in
course, this issue is not new. Several studies have, for in- Study2(a5-minexposuretoamusicpiece)wasrelativelyin-
stance,examinedhowhigharousalmayinfluencepersuasion nocuous—significantlymilderthanthoseusedinearlierre-
(e.g.,Pham,1996; Sanbonmatsu&Kardes,1988).An even searchon arousal effects.Thatsuchamildmanipulationof
greaternumberofstudieshaveexaminedhowthevalenceof arousalhadasignificant influenceonadvertisingresponses
affectivestates(i.e.,goodvs.badmoods)influenceadvertis- suggeststhatsucheffectsmaybepervasiveinrealworldset-
ingresponses(e.g.,Batra&Stayman,1990;Gardner&Wil- tings.Themildnessofourarousalmanipulationhastheoreti-
helm, 1987; Goldberg & Gorn, 1987; Schwarz, Bless, & cal significance as well. In retrospect, it may be precisely
Bohner,1991;Yi,1990).Yetcomparedtopreviousstudies, becausethismanipulationwasrelativelymildthatwefound
this research is unique in two important respects. greatersupportfortheexcitationtransferhypothesisthanfor
First,bymanipulatingarousalandvalenceindependently, thedynamic complexityhypothesis.Ourmanipulationmay
Study2offersthedistinctadvantageofseparatingtheeffects havebeentoomildtoinducethekindofprocessinginterfer-
ofthesetwodimensionsofaffectivestates.Previousstudies encepositedbythedynamiccomplexityhypothesis.Incom-
have generally focused on one dimension and ignored the parison,studiesthathaveshownsuchinterferenceshaveused
other.Itisthereforepossiblethatpartoftheresultsattributed strongermanipulations,suchascaffeine(e.g.,Humphreys&
to valence may reflect arousal, and vice versa.6 Revelle, 1984), physical exercise (e.g., Pham, 1996;
Theseconddistinguishingfeatureofthisresearchliesinits Sanbonmatsu & Kardes, 1988), and loud white noise (e.g.,
investigationofanimportantcontingencyoftheeffectsofaf- Paulhus & Kim, 1994).
fective states on ad evaluations. The results indicate that, Whereas dynamic complexity may require fairly high
whereasvalenceofaffectivestatesmaycoloradevaluationsin arousal,interestingly,excitationtransfermaybemorelikely
amood-congruentdirectionwhentheadisambiguous(Study undermoremoderatearousal.Thisisbecausewhenarousalis
1),thiscoloringeffectislesslikelywhentheadhasacleareraf- very intense (e.g., immediately after a dramatic emotional
fectivetone(Studies1and2).Ourfindingsechotheresultsof event),itsactualsourceislikelytobeverysalient.Asaresult,
researchinothersettingsinwhichmoodcongruencyeffectson intense arousal is less likelyto be misattributedtoa subse-
evaluationaremitigatedwheneverthetargethasaclearintrin- quenttarget(e.g.,Cantoretal.,1975).Therefore,itispossible
sicvalence(Isen&Shalker,1982;Miniardetal.,1992).They that bothmoderate andveryhigh levels of arousal produce
suggestthatmoodcongruencyeffectsmaynotbeaspervasive polarization—theformerviaexcitationtransfermechanisms,
androbustaspreviouslythought(Mayeretal.,1992). andthelatterviadynamiccomplexitymechanisms.Thisis-
Interestingly,itispreciselywhenthecoloringeffectofva- sueisworthyoffurtherinvestigation,usingabroaderrange
lence is less likely that the polarization effect of arousal be- of arousal levels.
comesmorelikely.ItwasfoundinStudy2that,whenthead’s
affectivetonewasclearlypositiveornegative,higharousalpo- Assimilation–Contrast, Elaboration
larizedadevaluationsinthedirectionofthead’saffectivetone. Likelihood, or Affect Infusion?
Thissuggeststhatpreexistingarousalstateswhileexposedto
anadmayinteractwiththeaffectivetoneoftheaditself.Al- Itisinstructivetorelateourfindingstootherpotentialexplana-
thoughbothdynamiccomplexityandexcitationtransfermech- tionsofaffectiveinfluencesonevaluation.Thecoloringeffect
anismscouldhavebeenatwork,ourresultsappeartosupport of valence can be related to several theoretical frameworks.
thelatterexplanation.Consumersmaymisattributetheirpre- Withinanassimilation–contrastframework, one wouldposit
existingarousalasbeingpartoftheiraffectiveresponsetothe that,duringadexposure,thevalenceofthepreexistingmood,
aditself.Arousedconsumersmayreportmoreextremeevalua- along with mood-congruent cognition, becomes integrated
tionsbecausethey“feelstrongly”aboutthead. withtheconsumer’srepresentationofthetargetad—thereby
resultingintheevaluationoftheadbeingassimilatedtoward
the valence of the preexisting mood(e.g., Schwarz & Bless,
1992; Strack, Schwarz, & Gschneidinger, 1985). Within the
6Apotentiallimitationofourstudiesisthatweuseddifferenttypesofmu-
sic(Indianvs.Western)tomanipulatemoodvalence.Onecouldarguethat Elaboration Likelihood Model (ELM) framework (Petty &
thetwotypesofmusicprimeddifferenttypesofassociations,whichthen Cacioppo,1986),onewouldpositthatthevalenceofthepreex-
taintedsubsequentadevaluations.Thisexplanationwouldnotaccountfor istingmoodservesasaperipheralcueinevaluatingthead(e.g.,
ourresults.First,ourcognitiveresponsesinStudy2gavenoindicationofany
Batra & Stayman, 1990; Petty, Schumann, Richman, &
imagerythatcouldbetracedbacktothemusicthatparticipantsheardinthe
Strathman,1993).WithintheAIM(Forgas,1995),onewould
previoustask.Second,adifferencebetweenWesternandIndianmusiccould
atbestexplainmaineffectsandnotinteractions.Finally,themainresultof positthatthevalenceofthepreexistingmoodandmood-con-
Study2pertainstheArousal×AdToneinteraction.BothIndianandWestern gruentcognitionsthatitprimesareusedasadditionalinputsin
musicwereusedforeachlevelofarousal.Weacknowledge,however,that constructiveprocessesofadevaluation.
keepingthetypeofmusicconstantacrossconditionswouldbedesirablede-
Notethatthesethreeexplanationarenotmutuallyexclu-
signproperty.Aninterestingpossibilitywouldbetomanipulatemoodva-
sive. They do differ, however, intheir ability tohandle the
lenceandarousalbyvaryingthetonalityandtempoofasinglepieceofmusic
that remains constant in all other respects (Kellaris & Kent, 1993). finding that the coloring effect of valence—whether called