Social Cognition, Vol. 24, No. 4, 2006, pp.469-495 RVaOlBenINcSe,O ANr oAuNsDa lC, aOnMdP LTaOteNral Spatial Attention THE AUTOMATICITY OF AFFECTIVE REACTIONS: STIMULUS VALENCE, AROUSAL, AND LATERAL SPATIAL ATTENTION Michael D. Robinson North Dakota State University Rebecca J. Compton Haverford College Socialpsychologistshavesoughttounderstandtheautomaticityofaffec- tivereactionsintermsofconsequencesforattention,evaluation,andjudg- ment. Typically, affect is manipulated by stimulus valence, with little considerationoftheroleofstimulusarousalinsuchprimingeffects.The presenttwostudies,involving54right–handedparticipants,soughttoex- tendthecaseforautomaticaffectandsoughttodosobyhighlightingthe uniqueprimingeffectsofstimulusarousalonthedifferentialactivationof theleftandrighthemispheresofthebrain.Usingaffectiveslidesasstimuli, thepresenttwostudiesfoundthathighlevelsofstimulusarousal,irrespec- tiveofstimulusvalence,shiftedattentionleftwardwithinvisualspace.This phenomenonextendsresearchonautomaticprimingeffectstolateralspa- tialattentionandsuggestsanimplicitspatialprobetaskthatmaybeuseful in future studies focused on the arousing effects of social stimuli. People frequently have automatic affective reactions to stimuli thatcanhaveimportantconsequencesforsubsequentcognition, emotion,andbehavior(Bargh,1997).Insightsofthissorthaveoc- curredinavarietyofliteraturesrelatedtoresearchontheself,at- titudes,andintergrouprelations(Wegner&Gilbert,2000).Much of this work has, for good reason, emphasized automatic reac- tions based on stimulus valence (e.g., Bargh, 1997). However, GrantsupportisacknowledgedfromNSF(9817649)andNIMH(068241).Addresscor- respondencetoMichaelRobinson,PsychologyDepartment,NorthDakotaStateUniver- sity, Fargo, ND 58105; E–mail: [email protected] 469 470 ROBINSON AND COMPTON mostmodelsofaffectpositaseconddimensionrelatedtostimu- lusarousal(e.g.,Lang,Bradley,&Cuthbert,1997),whichhastyp- icallybeenneglectedinstudiesofautomaticaffect(forareview, seeKlauer&Musch,2003).Thegoalsofthepresentinvestigation weretosystematicallycompareautomaticprimingeffectsrelated tostimulusvalenceandarousalandtodosoinaparadigmsensi- tive to the differential activation of the brain’s hemispheres. We startbymakingacaseforfocusingonstimulusarousalinsocial cognitive studies of affect. VALENCE AND AROUSAL IN AFFECTIVE REACTIONS Manymodelsofaffectposittwolargelyindependentdimensions relatedtovalenceandarousal(Langetal.,1997;Russell&Barrett, 1999). The valence dimension varies from negative to positive, whereasthearousaldimensionvariesfromdeactivatedtohighly aroused states (Russell & Barrett, 1999). The independence of thesedimensionshasbeensupportedinpsychometricstudiesof self–reported mood (Russell & Barrett, 1999) and in psychophysiologicalstudiesconcernedwiththebody’sreactions to stimuli (Lang et al., 1997). Most studies of automatic affect, whether related to attention (Pratto & John, 1991), evaluation speed (Fazio, Sanbonmatsu, Powell, &Kardes, 1986), orjudgment (Murphy&Zajonc, 1993), havefocusedexclusivelyonthevalencedimensionofexperience. Thisfocusonvalenceisunderstandableforatleasttworeasons. Prominent theories ofattitudes relate to valence butnot arousal (Fazio, 1995). In addition, the use of word stimuli, common to moststudiesinthisarea(Klauer&Musch,2003),typicallyintro- duces a narrow range of stimulus arousal levels. That is, words varyinvalencequiteabitmorethantheyvaryinarousal(Bradley & Lang, 1999). Nevertheless,anexclusivefocusonvalence,inautomaticaffect studies,isnecessarilysomewhatincomplete.Self–reportedaffec- tivestates(Russell&Barrett,1999),aswellasbodilyreactionsto affective stimuli (Lang et al., 1997), vary in both valence and arousal.Itisnotablethatquiteafewprominentsocialpsychology theories, such as those pertaining to cognitive dissonance (Festinger,1957),attribution(Schachter&Singer,1962),andper- VALENCE, AROUSAL, AND LATERAL SPATIAL ATTENTION 471 suasion(Janis&Feshbach,1953),makeheavyuseofthearousal dimension of experience. It is therefore important to systemati- cally manipulate both valence and arousal in automatic affect studies. This is particularly the case in relation to affective priming of lateral spatial attention, the focus of the present studies. AFFECTIVE PRIMING AND LATERAL SPATIAL ATTENTION Ourstudiessoughttoinvestigatetheinfluenceofaffectiveprim- ing on lateral spatial attention, an index of hemispheric activa- tion. Many behavioral studies have taken advantage of the anatomicalfactthatinputfromtheleftvisualfield(LVF)isfirst processedbytherighthemisphere,whereasinputfromtheright visualfield(RVF)isfirstprocessedbythelefthemisphere.Such contralateralconnectionshavebeenextensivelydocumentedin anatomicalstudiesofbothnonhuman(Berlucchi,1990)andhu- man(Bogen,1985)organisms.Brainimagingstudiesalsodem- onstrate that when participants attend to images in one visual field,brainactivityisincreasedinthecontralateralhemisphere (Heinzeetal.,1994;Mangunetal.,2001).Furthermore,damage to one hemisphere leads to attentional neglect of stimuli in the opposite visual field (e.g., Posner, Walker, Friedrich, & Rafal, 1987). In sum, anatomical, behavioral, and clinical work sup- portstheideathattheprocessingofunilateralvisualstimulide- pends primarily on the opposite hemisphere. Therefore, like many studies in the laterality literature, our study relies on re- sponses to LVF and RVF spatial probes as an index of relative hemisphericactivation. More specific to affect, prior research indicates that the two hemispheres tend to be differentially involved in affective pro- cessing.Inunderstandingthisliterature,itisimportanttodistin- guish between perceiving an affective stimulus versus reacting to anaffectivestimulus.Withrespecttoencodinganddecodingthe affectivenatureofstimuli,therighthemisphereissuperiortothe left hemisphere. In support of this point, emotional stimuli pre- sented to the LVF are more accurately and quickly encoded in terms of stimulus valence, and damage to the right hemisphere disruptsthesefunctionsmorethandamagetothelefthemisphere 472 ROBINSON AND COMPTON (Borodetal.,1998;Etcoff,1989;Silberman&Weingartner,1986). However,sucheffectsmustbedistinguishedfromthoseinvolv- ing the priming effects of affective stimuli on the two hemispheres. When a centrally presented stimulus or task is specialized to one hemisphere, it primes that hemisphere differentially. As a consequence,participantsaresubsequentlyfastertorespondto probes in the contralateral visual field. For example, tasks en- gaging left-hemisphere processes (e.g., reading) shift attention rightward,whereastasksengagingright–hemisphereprocesses (e.g., mental rotation) shift attention leftward (Hellige, 1993). Our goal was to examine how a centrally presented emotional stimuluswouldinfluenceprobedetectiontimesintheLVFand RVF,therebyinvestigatinghowincidentalexposuretoaffective stimuli influences the differential activity of the two hemispheres. Twogeneralmodelshavesoughttoexplaintheprimingeffects of affective stimuli on lateral spatial attention. According to the firstmodel, the valence ofthe stimulusshould bethe important priming dimension. Although this general model has been of- fered in several variations, the overarching view is that the left hemisphereismoreinvolvedinmediatingpositiveaffectivereac- tions,whereastherighthemisphereismoreinvolvedinmediat- ing negative affective reactions (Davidson, 1998; Tomarken & Keener,1998).ThismodelissupportedbyEEGstudiesindicating greater left hemisphere activation in positive emotional states and traits, and greater right hemisphere activation in negative statesandtraits(e.g.,Davidson,1998;Sutton&Davidson,1997). Suchmodelsmightpredictthatpositivestimulishouldbiasatten- tion toward the RVF, whereas negative stimuli should bias attention toward the LVF. Analternativeviewisthatarousal,ratherthanvalence,creates lateralspatialbiases.Specifically,Heller(1993)hasproposedthat therighthemisphereisespeciallyinvolvedinhigh-arousalemo- tionalstates.Insupportofthispoint,evidenceindicatesthatthe right hemisphere plays a specialized role in modulating cardiac (Wittling,1995)andelectrodermal(Johnsen&Hugdahl,1993)ac- tivityinresponse toemotional stimulation. Damage tothe right VALENCE, AROUSAL, AND LATERAL SPATIAL ATTENTION 473 (versus left) hemisphere has a more pronounced effect on skin conductance responses to emotional stimuli (Caltagirone, Zoccolotti, Originale, Daniele, & Mammucari, 1989). This re- search, then, suggests a particularly close relationship between the right hemisphere and the arousal systems of the body. Arousal–based models, such as Heller’s (1993), therefore pre- dictthathigh-arousalemotionalstatesshouldbiasattentionleft- ward. Consistent with this prediction, manipulations of stress have produced lateral biases in favor of left–sided targets (Asbjörnsen, Hugdahl, & Bryden, 1992; Gruzelier & Phelan, 1991). Likewise, lateral biases in favor of LVF stimuli correlate positively with self–reported arousal on measures of dispositionalmood(Heller,Nitschke,&Lindsay,1997).Inaddi- tion,anxiousarousaliscorrelatedwithanincreasedleftwardper- ceptual bias on a task thought to tap parietal lobe asymmetries, whereas depression (presumably a low-arousal state) is associ- ated with decreased biases on the same task (Heller, Etienne, & Miller, 1995). Such data provide evidence for the idea that arousal, rather than valence, differentially activates the right hemisphere, as inferred from lateral spatial attention tasks (Heller, 1993). Although evidence associating the posterior right hemisphere withthearousaldimensionofemotionalexperiencehasbeenac- cumulating,furtherevidenceisnecessarytotestthemodelmore thoroughly. For example, many studies supporting the Heller modelhavereliedonexaminingindividualdifferencesrelatedto clinical variables such as anxiety and depression (Heller et al., 1995).Clinicalvariablesmayormaynotmapcloselyontotheef- fects of emotional valence and arousal within normal popula- tions.Suchconsiderationsareespeciallyimportantinrelationto positive high–arousal states like excitement. Such states may be lesseasilyactivatedamongindividualssufferingfromclinicalor subclinical forms of negative affect (Clark, Watson, & Mineka, 1994). Therefore, a study focusing on affective reactions among nonclinical samples may be especially informative concerning the normative effects of affective primes on differential hemispheric activity. 474 ROBINSON AND COMPTON OVERVIEW OF PRESENT STUDIES Our understanding of automatic affective reactions would be enrichedbyanincreasedfocusonarousal,particularlygiventhe fact that many social psychological theories make heavy refer- encetoarousal(e.g.,Festinger,1957;Schachter&Singer,1962). Consistent with automatic affect paradigms (Bargh, 1997: Klauer&Musch,2003),wepresentedaffectivestimuliinaninci- dentally primed manner (Bargh & Chartrand, 2000). Specifi- cally, participants were not explicitly instructed to evaluate, or appraise the significance of, affective primes. The primes were emotional slides that varied independently in valence and arousal. Wesoughttoexamineanovelconsequenceofautomaticaffec- tivereactionsbyinvestigatingprimingeffectsonlateralspatial attentionwithinaneutraltargettask.Followingtheremovalof affective primes, participants were asked to indicate whether they saw one or two dots in a target slide. The target slide was presented laterally, either within the LVF or RVF. We could therefore examine potential effects of stimulus valence and arousalonlateralspatialattention.Weentertainedtwoalterna- tive hypotheses. The valence hypothesis (e.g., Van Strien & Morpurgo, 1992) predicts that negative stimuli should shift at- tentionleftwardwithinvisualspace,indicatingactivationofthe right hemisphere. By contrast, the arousal hypothesis (Heller, 1993) predicts that arousing stimuli, whether of a positive or negative valence, should shift attention leftward within visual space.Thepresentdesign,involvingorthogonalmanipulations ofvalenceandarousal,isuniquelycapableofsidinginfavorof onehypothesisovertheother. STUDY 1 METHOD Participants.Participants were 38 (15 male and 23 female) right–handed undergraduate volunteers from the University of VALENCE, AROUSAL, AND LATERAL SPATIAL ATTENTION 475 Illinois,Champaign.Theyweregivencoursecreditforparticipat- ing. Apparatus.A DOS–based computer was connected to two Ko- dakEktaproprojectors.Theprojectorswereofprofessional pre- sentationqualityandwerespecificallydesignedtointerfacewith thecomputer.Theprojectorswereplacedbehindandabovethe participant’schairsuchthattheycould,whenangleddownward, project images from behind the participant during the courseof thestudy.Oneprojectordisplayedemotionalslideswhereasthe other projector displayed dot slides. Affective primes were 60 emotional slides taken from Lang’s well–validated set (Lang, Bradley, & Cuthbert, 1999). Thirty of theslideswerepositiveandthirtywerenegative.Withineachva- lence, we chose slides that were rated both low and high in arousal.Thus,therewere15negativelow-arousalslides,15nega- tive high-arousal slides, 15 positive low-arousal slides, and 15 positive high-arousal slides. The slides were chosen in such as wayastoeliminateanyrelationshipbetweenvalenceandarousal levels.1 Task.The target task was a simple, nonaffective one. Partici- pants were asked to press the 1 key on a button box if one dot waspresentedandtopressthe2keyonabuttonboxiftwodots were presented.2 Dot slides were prepared for the study. Dots were 1 inch × 1 inch when displayed on the screen. Dots were 1.Langetal.(1999)provideslidenumberstofacilitatecommunicationamongresearch labs.Intermsoftheseslidenumbers:negativehigh-arousalslidesconsistedof1050,1120, 1300,1930,3130,3250,6230,6260,6300,6510,6570,7380,9300,9410,and9570;negative low-arousalslidesconsistedof1111,1220,1301,2053,2520,2800,3230,3350,7361,9008, 9290,9320,9415,9421,and9561;positivehigh-arousalslidesconsistedof4599,4607,4608, 4641,4651,4652,4660,5621,8180,8200,8350,8370,8380,8470,and8490;andpositive low-arousalslidesconsistedof1440,1460,1610,1750,1810,2040,2050,2057,2070,2165, 2352, 2550, 2660, 4606, and 7325. 2.Forpurposesthataredistinctfromcurrentconcerns,25%ofthetrialsinvolvedare- sponsecompatibilitymanipulation.Onthesetrials,whichweresignaledbytwofixation pointsratherthanone,participantswereinstructedtohitthe1keyifthereweretwodots presentedandtohitthe2keyiftherewasonedotpresented.Unfortunately,notallpartici- pantsappearedtounderstandtheseinstructions,asaccuracyratesforthesetrialswereun- acceptablylow(M=86.4%),especiallyamongsomeparticipants.Toavoidambiguity,we decidedtodeleteallofthesetrialsinvolvingincompatibleresponses.Thisleft75%ofthe performance data in Study 1. 476 ROBINSON AND COMPTON randomly presented 2.5 degrees right or left of fixation. Re- sponsesweremadewiththeindexandmiddlefingersoftheright hand.3 Precise timing was accomplished in several ways. The com- puter controlling the projectors used DOS–based programming code to trigger the projectors and collect the response data. The projector lenses were equipped with high–speed Uniblitz shut- ters(approximately2msrandomerror).Onsetofeventswascon- trolledbycomputersignalssenttotheshuttersratherthantothe projectors.Toensurethattheprojectorswerereadytodisplaythe particularslideschosenforagiventrial,projectorsweredisplay- ing their images prior to the shutters opening. Responses were made using a response box with less than 1 ms random error. Procedure.Participants were run in individual testing sessions. Whentheparticipantsarrivedatthelaboratory,weexplainedthe generalproceduresofthestudy,obtainedinformedconsent,and turnedoffthelights.A10–minutedarkadaptationperiodensued, duringwhichtimeparticipantsengagedinseveralcomputertasks unrelatedtothestudy.Followingthisdarkadaptationperiod,par- ticipantsweretoldthattheremainingtaskconsistedoftwoparts. First,theywouldseeslides,whichtheyshouldpayattentiontoin preparation for a subsequent memory test. Note that we did not ask participants to evaluate the valence or significance of slides; therefore, priming effects can be viewed as automatic in this re- spect.Second,participantswereinstructedtorespondasquickly andaccuratelyaspossibletoindicatewhetheroneortwodotswere presented.Participantswereremindedtofocustheirattentionto- wardthecenterofthescreentoensuremaximumperformance. Participantssatinastandardchair,equippedwithasmalldesk towhichabuttonboxwasaffixed.Thechairwasplaced3feetin 3.Proceduresweredesignedtoavoidlateralityeffectsduetoresponsehand;thatis,allre- sponsesweremadewiththerighthand.Wechosetoavoidaddingacounterbalancingfactor forresponsehandforthreereasons.One,ourparticipantswereright–handedandthusmore proficientinmakingmotormovementswiththeirrighthands.Two,ourinterestwasinstim- ulusfactors(i.e.,thenatureoftheemotionalslide)ratherthanresponsefactors(i.e.,which handmadeaparticularresponse).Three,wewantedtoavoidaddinganothervariabletothe design,astherewereanumberofvariablesalready.Itisextremelyunlikelythattheseproce- durescompromisethefindingsinanyway,becausethefindingsconcerntheeffectsof within-subjectvariablesthatwouldbeorthogonaltovariationsinresponsehand. VALENCE, AROUSAL, AND LATERAL SPATIAL ATTENTION 477 frontofalarge(5foot×5foot)projection screen. Emotional im- ages, which were approximately 3 feet × 3 feet, were projected ontothecenterofthescreen.Subsequenttothepresentationofan emotionalslide,thecomputertriggeredacentralfixationfroma laserpentoensurethatattentionwascenteredpriortothepresen- tation of the lateral spatial targets. A different random order of emotional slides was chosen for eachparticipant.Dotslideswerealsorandomlypairedwithemo- tionalslides,againseparatelyforeachparticipant.Therewere60 trials,oneforeachoftheemotionalslidesusedinthestudy.Trials consistedofthefollowingsequence.First,anemotionalslidewas shown for 3 seconds. Second, there was a 500 ms delay during whichnoeventsoccurred.Third,alaserpenpresentedacentral fixationpointfor200ms.Fourth,therewasan800msdelaydur- ing which no events occurred. Fifth, one or two dots were pre- sentedleftorrightoffixation,withlateralpositiondeterminedat random. Dotswerepresented foronly100 ms,whichshouldef- fectivelyprecludeeyemovementsinthedirectionofthedotstim- uli (Pashler, 1998). The program then waited for the response, followed by a 5-second delay until the next trial. RESULTS Preliminary Analyses.The accuracy of dot discrimination re- sponseswashigh(M=96.8%).AValence×Arousal×VisualField repeated measures ANOVA on these accuracy rates found one significanteffect:Dotdiscriminationsweremoreaccuratefollow- ingpositiveslides(M=97.8%)thanfollowingnegativeslides(M = 95.9%), F(1, 37) = 4.52, p = .040. This effect may be consistent withresearchshowingthatnegativestimuliaremoredistracting, andthereforecausemoreinterferencewiththeprimarytask,rela- tivetopositivestimuli(Pratto&John,1991).Noneoftheotheref- fects were significant, ps>.10. Priortoanalyzingresponselatencies,wedeletedinaccuratetri- als and then log–transformed times to normalize their distribu- tion. To reduce the impact of outliers, we replaced log latencies 2.5 standard deviations below or above the overall mean with thesecutoffvalues.Consequently, 2.3%ofthelatencieswerere- 478 ROBINSON AND COMPTON placed.Althoughanalysesfocusedonlog–latencyscores,means are reported in milliseconds for ease of interpretation. We used computerized algorithms to assign valence, arousal, andvisualfieldtoindividualtrials.Theuseofsuchalgorithmsis desirable to avoid any systematic assignment tendencies within cognitiveparadigms(Robinson&Neighbors,inpress).However, randomizationalgorithmscanbeproblematicinthecontextofa large within–subject design in combination with a relatively smallnumberoftrials.Accordingly,weobservedthatStudy1did nothavesufficientpowertosimultaneouslyexamineallmainef- fectsandinteractionsinvolvingvalence,arousal,andvisualfield. Specifically,cellsizesforthefullValence×Arousal×VisualField design were low among some participants (range 1–11; mean ~ 5.6).Bycontrast,analysespertainingtotheValence×VisualField and Arousal × Visual Field interactions had far more adequate cell sizes (range 5–19; mean ~ 11.2). A preliminary analysis re- vealedthattherewasnohintofathree–wayinteraction,F<1.We thereforesoughttoaddpowertooursubsequentanalysesbyex- amining valence and arousal separately. This statistical power issueisrectifiedinStudy2andinacombinedanalysisfollowing Study 2. Primary Analyses.The first analysis sought to determine whether the affective valence of slides primed lateral spatial at- tention.Toexaminethisquestion,weperformedaValence×Vi- sual Field ANOVAon the latency data. There was a main effect forValence, F(1, 37) =11.56, p =.002. Responses following posi- tiveslideswerefaster(M=486ms)thanresponsesfollowingneg- ativeslides(M=527ms),aphenomenon thatpreviousresearch has linked to the “freezing” effect of negative affective primes (Wilkowski&Robinson,inpress).ThemaineffectforVisualField wasnotsignificant,F<1.TheValence×VisualFieldinteraction wasalsonotsignificant,F(1,37)=2.19,p=.147.Thus,therewasno indication that the affective valence of stimuli differentially activated the two cerebral hemispheres. In a second analysis, we sought to determine whether the arousallevelofslidesprimedlateralspatialattention.Toexamine this question, we performed an Arousal × Visual Field ANOVA onthelatencydata.NeitherthemaineffectforArousal,F(1,37)= 2.07,p=.158,northemaineffectforVisualField,F<1,wassignifi-