Threat Facilitates Subsequent Executive Control During Anxious Mood

14 Pages·2011·0.5 MB·English

Emotion ©2011AmericanPsychologicalAssociation
2011,Vol.11,No.6,1291–1304 1528-3542/11/$12.00 DOI:10.1037/a0026152
Threat Facilitates Subsequent Executive Control During Anxious Mood
Jeffrey L. Birk Tracy A. Dennis
TuftsUniversity CityUniversityofNewYork
Lisa M. Shin Heather L. Urry
TuftsUniversityandMassachusettsGeneralHospital/Harvard TuftsUniversity
MedicalSchool
Dualcompetitionframework(DCF)positsthatlow-levelthreatmayfacilitatebehavioralperformanceby
influencingexecutivecontrolfunctions.Anxietyisthoughttostrengthenthiseffectbyenhancingthreat’s
affective significance. To test these ideas directly, we examined the effects of low-level threat and
experimentallyinducedanxietyononeexecutivecontrolfunction,theefficiencyofresponseinhibition.
In Study 1, briefly presented stimuli that were mildly threatening (i.e., fearful faces) relative to
nonthreatening(i.e.,neutralfaces)ledtofacilitatedexecutivecontrolefficiencyduringexperimentally
inducedanxiety.Nosucheffectwasobservedduringanequallyarousing,experimentallyinducedhappy
moodstate.InStudy2,weassessedtheeffectsoflow-levelthreat,experimentallyinducedanxiety,and
individualdifferencesintraitanxietyonexecutivecontrolefficiency.ConsistentwithStudy1,fearful
relative to neutral faces led to facilitated executive control efficiency during experimentally induced
anxiety.Nosucheffectwasobservedduringanexperimentallyinducedneutralmoodstate.Moreover,
individual differences in trait anxiety did not moderate the effects of threat and anxiety on executive
control efficiency. The findings are partially consistent with the predictions of DCF in that low-level
threatimprovedexecutivecontrol,atleastduringastateofanxiety.
Keywords:low-levelthreat,executivecontrol,stateanxiety,traitanxiety,dualcompetitionframework
Thehumanattentionalsystemisbiasedtowardthedetectionof scribed(LoBue,2010),thespecificattentionalmechanismsunder-
threatening information. Indeed, the presence of threat-related lyingthreat’seffectonbehaviorremainlargelyunexplained.
information (e.g., fearful faces) leads to enhancement at an early One candidate attentional mechanism is executive control,
stage of visual processing (Phelps, Ling, & Carrasco, 2006) and which refers to the ability to organize cognitive and sensory
capturesattentionsoefficientlythatthreatscanbedetectedevenin processing in a goal-directed way in order to guide appropriate
the absence of conscious awareness (Dijksterhuis & Aarts, 2003; action (Miller & Cohen, 2001). An essential component of exec-
Marcos&Redondo,2005).Althoughtheevolutionaryrootsofthis utive control is response inhibition, namely the ability to inhibit
remarkable prioritization of threat detection are fairly well de- dominant or prepotent responses as needed (Miyake, Friedman,
Emerson,Witzki,&Howerter,2000).Responseinhibitionmaybe
critical for promoting adaptive functions following threat experi-
ThisarticlewaspublishedOnlineFirstNovember7,2011. ences. Imagine, for example, that you are having a conversation
JeffreyL.BirkandHeatherL.Urry,DepartmentofPsychology,Tufts with a friend during a hike in the woods when suddenly you
University;TracyA.Dennis,DepartmentofPsychology,HunterCollege, encounterapoisonoussnake.Themostadvantageousresponsein
CityUniversityofNewYork;LisaM.Shin,DepartmentofPsychology,
thisscenariowouldinvolverapidlydetectingthesnakewhilealso
Tufts University and Psychiatric Neuroimaging Group, Massachusetts
inhibiting previously planned actions related to the conversation.
GeneralHospital/HarvardMedicalSchool.
To explain how the executive control of attention influences the
WethankTuftsUniversityfortheirassistanceinfundingthisworkvia
effectofthreatonbehavior,Pessoa(2009)proposeddualcompe-
two Graduate Student Research Awards to J. L. Birk. We also thank S.
Bansil,A.Bellet,P.Bene,K.Brethel,E.Brown,C.Callahan,R.Citron,E. titionframework(DCF).
Davidowitz,M.DeMatteo,S.DelDonno,R.Gavrielov,J.Laks,B.Meller, AccordingtoDCF,threat-relatedstimulicarryaffectivesignif-
D.Millstein,M.Patlingrao,V.Peisch,P.Pensuwan,P.Pop,C.Rucinski, icance,whichaltersperformancebystrengtheningsensoryrepre-
M. Santarsieri, S. Sloley, R. Trumball, V. Tran, A. Wei, and J. Yih for sentationsattheperceptuallevelandbyprioritizingattentionatthe
assistancewithdatacollectionandprocessing.Finally,wethankP.Opitz, executive level. The model predicts that, although threat consis-
S. Cavanagh, and J. DiCorcia for valuable feedback and technical assis- tently leads to prioritized perceptual processing, its effect on
tance,andF.WilhelmandP.PeykformakingANSLAB,asuiteofopen
executivecontroldependsdramaticallyonthelevelofthreat(e.g.,
source Matlab routines used to process physiological data, available as
high/extreme,low/mild).Specifically,DCFleadstotheprediction
freeware in the Society for Psychophysiological Research Software Re-
thatahigh-threatstimulusplacesapremiumonprocessingofthe
pository(http://www.sprweb.org/repository).
CorrespondenceconcerningthisarticleshouldbeaddressedtoHeather threatand,thus,reducestheexecutiveresourcesthatareavailable
L.Urry,DepartmentofPsychology,TuftsUniversity,490BostonAvenue, to manage responses to stimuli other than the threat itself (i.e.,
Medford,MA02155.E-mail:[email protected] “hard prioritization”). A low-threat stimulus, on the other hand,
1291
1292 BIRK,DENNIS,SHIN,ANDURRY
placesasmallerpremiumonprocessingofthethreatand,thus,has during experimentally manipulated happiness. In Study 2, we
minoreffectsontheavailabilityofexecutiveresourcestomanage compare the effect of threat during experimentally manipulated
responses to stimuli other than the threat (i.e., “soft prioritiza- anxiety to the effect of threat during an experimentally manipu-
tion”). In fact, according to DCF, low-level threats prioritize lated neutral mood state. Because there is some indication that
subsequent processing at the location of the threat “as part of experimental demonstrations of threat-related attentional capture
additionalinformationgathering”(p.161)andthusshouldenhance may reflect an interaction between state and trait anxiety (for a
subsequenttaskperformanceatthatsamelocation. review, see Williams, Mathews, & MacLeod, 1996), we also
Several studies have provided evidence for the conclusion that examinewhethereffectsoflow-levelthreatandmanipulatedanx-
threat can improve executive control function (Blanchette, 2006; ietydependonindividualdifferencesintraitanxiety.
Cohen, Henik, & Mor, 2011; Öhman, Flykt, & Esteves, 2001; In both studies, consistent with the research on optimizing
Schimmack,2005;butseeKrysko&Rutherford,2009).However, mood-induction procedures (for a review, see Gilet, 2008), we
missing from these studies is a consideration of state-dependent usedanovel,multimodalmood-inductionprocedurethatincorpo-
influences (e.g., mood state). Indeed, DCF proposes that the ex- ratednarrativetext,mood-congruentmusic,emotionalimages,and
ecutiveenhancementseeninresponsetolow-levelthreatshouldbe imagined self-involvement. To measure the executive control of
especially pronounced for people in an elevated state of anxiety attention,participantscompletedamodifiedversionoftheAtten-
becauseelevatedanxietyheightenstheperceivedaffectivesignif- tion Network Test (ANT; Fan, McCandliss, Sommer, Raz, &
icanceofthethreat.Consistentwiththisidea,thepresentationof Posner, 2002) following the mood-induction procedure. In this
low-threat stimuli (i.e., fearful faces) relative to neutral stimuli variantoftheANT(Dennisetal.,2008),eachtrialwaspreceded
(i.e., neutral faces) improved subsequent executive control effi- byathreat-related(fearfulface)ornonthreat-related(neutralface)
ciency for people with high levels of naturally occurring state stimulus.Fearfulfacesconstitutelow-levelthreatstimulibecause
anxiety but not for people with low levels of anxiety (Dennis, they do not directly jeopardize one’s survival, but they neverthe-
Chen,&McCandliss,2008).Similarly,asecondstudyfoundthat lessclearlysignifythepotentialpresenceofdanger.Foreachcell
higher levels of executive control efficiency following presenta- ofthedesign,wecalculatedaconflictscore—abehavioralindexof
tion of fear-evoking stimuli (e.g., pictures of sharks) but not theefficiencywithwhichapersonusesexecutivecontroltoinhibit
neutralstimuli(e.g.,picturesoffish)werecorrelatedwithhigher adominantresponse.Asatestofspecificity,wealsocalculatedan
levels of naturally occurring state anxiety (Finucane & Power, alertingscore,abehavioralindexoftheefficiencywithwhichone
2010). However, because naturally occurring state anxiety might achieves a state of readiness to act. To be certain that our novel
covarysystematicallywithunmeasured/unknownextraneousvari- mood-induction procedures were effective, we measured self-
ables (e.g., arousal, intelligence, personality traits), it is possible reportedaffectandfacialelectromyography(EMG)overthecor-
thatathirdvariablemightexplainthemoderatingeffectsofstate rugator supercilii muscle region, which is active when frowning
anxietythatwereobservedinthesetwostudies. (Cacioppo,Martzke,Petty,&Tassinary,1988).
Finally,asharedlimitationofexistingstudiesistheirfocuson
one, unpleasant arousing mood state, namely anxiety. It may be Study 1
that threat affects subsequent executive control efficiency during
anxiousmoodstatesbecauseanxietyrepresentsastateofheight- In Study 1, we tested the hypothesis that the presentation of a
ened arousal. In that case, any aroused state, even one that is fearful face relative to a neutral face increases executive control
pleasant rather than unpleasant, might similarly influence execu- efficiency in response to target stimuli presented subsequently at
tivecontrolefficiency.Indeed,thereissomeevidencethatastate thesamespatiallocation.Giventhewell-establishedlinkbetween
ofhappiness,whichisanaroused,positivelyvalencedmoodstate, anxiety and threat, it was possible that the facilitative effect of
impairsperformanceefficiencyontasksrequiringexecutiveatten- low-levelthreatonexecutivecontrolefficiencywouldoccuronly
tion, such as the Flanker task (Rowe, Hirsh, & Anderson, 2007) forpeopleinananxiousmoodstateandnotforpeopleinahappy
and the alternating Stroop task (Phillips, Bull, Adams, & Fraser, mood state. However, if the effect of an anxious mood state is
2002),ascomparedwithaneutralstate(butseeFinucane,White- explained by arousal, then a similar effect would be observed
man, & Power, 2009). Finucane (2011) demonstrated that exper- during the equally arousing happy mood state. Finally, although
imentally induced fearful and angry mood states improved exec- our primary focus was to determine whether low-level threat
utivecontrolefficiencyrelativetoaneutralmoodstate,buttoour would facilitate subsequent executive control, we also examined
knowledge, only one study has directly compared effects of ex- whether threat would influence attention more broadly, in which
perimentallyinducedanxiousandpositive(“nonanxious”)moods casewewouldalsoobserveeffectsoflow-levelthreatonalerting.
onexecutivecontrolefficiency(Pacheco–Unguetti,Acosta,Calle-
jas, & Lupia´n˜ez, 2010). In that study, there was no effect of Method
inducedmoodstateonexecutivecontrolefficiency.Itisimportant
that neither Finucane (2011) nor Pacheco–Unguetti et al. (2010)
Participants
evaluatedtheeffectofthreat.
Inwhatfollows,wepresenttheresultsoftwostudies,foreach Sixty-one undergraduate students from Tufts University (35
of which the goal was to determine whether low-level threat female;mean[M] (cid:1)20.10years,standarddeviation[SD] (cid:1)
age age
stimuli would, as predicted by DCF, facilitate subsequent execu- 1.76) participated for course credit or monetary compensation.
tivecontrolefficiencyinthecontextofanexperimentallymanip- Participantswere67.2%Caucasian,19.7%AsianorAsianAmer-
ulatedstateofanxiety.InStudy1,wecomparetheeffectofthreat ican, and 8.2% Black or African American; 4.9% declined to
during experimentally manipulated anxiety to the effect of threat providethisinformation.Ofthetotalsample,9.8%endorsedbeing
THREATFACILITATESSUBSEQUENTEXECUTIVECONTROL 1293
of Hispanic origin. All study procedures were approved by the &Cacioppo,2003).CorrugatorEMGactivitywasmeasuredusing
InstitutionalReviewBoardatTuftsUniversity,andallparticipants shielded 4-mm Ag/AgCl electrodes. Following site preparation
providedinformedconsentpriortoparticipatinginthestudy. with an electrode preparation pad, a pair of sensors was attached
directlyabovetheeyebrowandagroundelectrodewasattachedto
themiddleoftheforeheadconsistentwithguidelinesbyFridlund
Materials
andCacioppo(1986).
Moodmanipulation. Moodwasmanipulatedonabetween- EMG data were collected using MP 150 hardware and Acq-
subjectsbasis.Participantswererandomlyassignedtooneoftwo Knowledge3.8.2software(Biopac,Goleta,CA).Dataweresam-
groups.Thirtyparticipantscompletedananxiousmoodinduction, pled at 1,000 Hz and filtered from 5 Hz to 3 kHz (60-Hz notch
and 31 participants completed an equally arousing happy mood filteron)online.Offline,datawereresampledto400Hz,rectified,
induction. All participants also completed a nonaroused neutral andsmoothedwitha16-Hzlow-passfilter,decimatedto4Hz,and
moodinduction,whichservedasabasisforcomparisoninmood- further smoothed with a 1-s prior moving average filter. These
manipulationchecks.Intheanxiousmoodinduction,participants steps were completed in part with Matlab software (Mathworks,
imagined being a passenger in a drunk-driving car accident and Natick, MA) using ANSLAB routines (Wilhelm & Peyk, 2005).
helpinginjuredpeopleinitsaftermath.Inthehappymoodinduc- Corrugatoractivitywasaveragedacrossthe4-mindurationofeach
tion, participants imagined walking with a friend on a warm day mood-inductionconditionandlog-transformedtoachievenormal-
amid picnicking families, playing children, and running dogs. In ity.
the neutral mood induction, participants imagined a sequence of AttentionNetworkTest. Aftereachofthetwomoodinduc-
mundane activities such as shopping for groceries, doing small tions,participantscompletedamodifiedversionoftheANT(Fan
tasksathome,andcallingafamilymember. etal.,2002)similartothatusedbyDennisandcolleagues(2008).
Mood-congruent instrumental music, selected based on pilot The modified ANT allowed us to measure the efficiency of the
testing of effects on self-reported affect, was played during each executive control and alerting attention networks. The task of
scene(selectionsavailableuponrequest).Inaddition,picturesthat participantswastorespondviamousebuttonswiththethumbsof
accompanied the narrative were shown. Pictures with established both hands to indicate whether a centrally located arrow (i.e.,
positive, negative, and neutral valence were selected from the target) pointed to the left or right. Each trial lasted for 4,050 ms
InternationalAffectivePictureSeries1(Lang,Bradley,&Cuthbert, (seeFigure1fortrialstructure).Sincetheeffectsofinducedmood
2005); supplementary pictures were obtained from Shutterstock could be time limited, we excluded trials from the original ANT
(http://www.shutterstock.com) and Wellcome Images (http:// thatmeasuretheefficiencyoftheorientingnetwork.Wereasoned
images.wellcome.ac.uk). that this would minimize the time it would take to complete the
The mood-induction procedure was programmed in E-Prime taskwhileatthesametimemaximizingourabilitytotesthypoth-
1.2.1 software (Psychology Software Tools, Sharpsburg, PA). esized effects on executive control efficiency and to examine
Eachmoodinductioncontained12pictureswithassociatedstory- specificitybyretainingtrialsthatassessalertingefficiency.
linetext,andlasted4min.Eachpicturewaspresentedwithtextat
To assess executive control efficiency, the target was sur-
thetopofthescreenfor12sandthenwithouttextforanother8s.
roundedbyfourstimuli(i.e.,flankers),twoeachontheleftandthe
Participantswereinstructedtoreadthetextfirstandthenviewthe
right. There were three varieties of flankers that differed across
imagewhileimaginingthatthedepictedeventswereoccurringin
trialtypes:congruent(i.e.,arrowspointinginthesamedirectionas
their real lives. The order of the neutral and anxious (or happy)
the target), incongruent (i.e., arrows pointing in the opposite di-
moodinductionswascounterbalancedacrossparticipants.
rection of the target), and neutral (i.e., horizontal lines with no
Moodmeasures. Participantscompletedacomputerizedver-
directional information). Greater conflict is present on trials in
sionofthePositiveandNegativeAffectSchedule(PANAS;Wat-
which the flanking arrows are incongruent with the target arrow,
son,Clark,&Tellegen,1988)beforeandaftereachmoodinduc-
relative to trials in which the flanking arrows are congruent or
tion. Participants reported how much they felt each of 20
neutral. As is standard practice with the ANT (Fan et al., 2002),
adjectives“rightnow,atthisverymoment”ona5-pointscalefrom
conflict scores were generated by subtracting congruent-flanker
1, very slightly or not at all, to 5, extremely. We computed a
reaction time from incongruent-flanker reaction time for correct
positive affect (PA) score by summing scores for enthusiastic,
trials. Smaller conflict scores indicate greater efficiency of exec-
proud,inspired,anddetermined,andanegativeaffect(NA)score
utivecontrol,specificallyresponseinhibition.
by summing scores for distressed, upset, guilty, scared, hostile,
To assess alerting, on each trial the target was preceded by a
irritable,ashamed,nervous,jittery,andafraid.Wealsocomputed
double warning cue (i.e., asterisks above and below the fixation
anarousalscorebysummingscoresforinterested,excited,alert,
cross)ortherewasnocueprecedingthetarget(i.e.,fixationcross
attentive, strong, and active. These latter six items are typically
only). Greater alerting is present on trials in which the target is
includedasitemsonthePAscale(Watsonetal.,1988),butPatrick
precededbythedoubleasterisks,relativetotrialsinwhichthereis
and Lavoro (1997) demonstrated that they index arousal, since
nocue.Again,followingstandardpracticewiththeANT(Fanet
both positive and negative pictures elicit higher ratings on these
al., 2002), alerting scores were generated by subtracting double-
items,relativetoneutralpictures.
cue reaction time from no-cue reaction time for correct trials.
In addition, for a more objective assessment of emotional va-
lence, we measured EMG activity over the corrugator supercilii
muscle region. Relative to a neutral baseline, greater corrugator 1Identification numbers for IAPS pictures: 2570, 2751, 2810, 5760,
activity is observed for unpleasant emotions, whereas lower cor- 5800,7022,7035,7041,7060,7090,7100,7175,7211,7217,7233,7235,
rugatoractivityisobservedforpleasantemotions(Larsen,Norris, and7920.
1294 BIRK,DENNIS,SHIN,ANDURRY
Figure1. TrialstructureforthemodifiedAttentionNetworkTest.Inthisexampletrial,theparticipantsawa
fearfulfacewhichwasfollowedbyafixationcrossandthenadoublecue(whichwilltendtoincreasealerting
efficiencyrelativetotrialswithnocue).Thedoublecuewasfollowedbyanotherfixationcrossandthenthe
targetwaspresented.Thetargetinthiscasewasaright-facingarrowwhichwassurroundedbyleft-facingarrow
flankers.Thus,thiswasanincongruenttrial(whichwilltendtoincreaseconflict,i.e.,decreaseexecutivecontrol
efficiency,relativetotrialswithcongruentarrows).Thetrialendswithavariable-intervalfixationcrossthat
makeseachtriallastatotalof4,050ms.
Higher alerting scores indicate greater efficiency in achieving a presented in four experimental blocks of 48 trials each. Partici-
stateofactionreadiness. pants were instructed to take a break between each block. At the
Toexaminetheinfluenceofthreatonsubsequentefficiencyof endofthesession,participantsprovideddemographicinformation
the executive control and alerting attentional networks, each trial andweredebriefedaboutthenatureofthestudy.
beganwiththepresentationofoneoftwotypesoffaces—fearful
(i.e.,threat-related)orneutral.FollowingDennisetal.(2008),the Results
delay between the offset of the face and the onset of the target
varied randomly between 900 ms and 1800 ms (see Figure 1).
Data Retention and Analysis
Faces from 12 actors were selected from the NimStim collection
(Tottenham et al., 2009). Greater threat is present on trials pre- For each reaction time measure (conflict score and alerting
cededbyfearfulfacesthanbyneutralfaces. score), Mahalanobis distance was computed for each participant
Conflictandalertingscoreswerecalculatedseparatelyforfear- acrosslevelsofthewithin-subjectsindependentvariablestoiden-
fulandneutralfacetrialsforeachmoodgroup.Accuracydatawill tify multivariate outliers. One participant was excluded from the
not be reported for Study 1 or Study 2 because the ANT was analysis of conflict scores and two participants were excluded
designed to measure efficiency, which is our primary dependent fromtheanalysisofalertingscoresonthebasisofthesetests(p(cid:2)
variable of interest. Furthermore, the accuracy data are not nor- .001). In the general linear model (GLM) analyses below, we
mallydistributed. reportresultsofmultivariateFstatistics((cid:3)(cid:1).05)andfollow-up
Fisher’s least significant difference tests as needed to interpret
Procedure significantresults.
Participantscompletedthetasksinthefollowingorder:baseline
Manipulation Checks
PANAS, mood induction 1, PANAS, ANT 1, PANAS, mood
induction 2, PANAS, ANT 2, PANAS. Each ANT began with a Todeterminewhetherthemoodmanipulationsachievedapos-
practiceblockof12trialsthatwereaccompaniedbyreactiontime itively valenced mood state in the happy group and a negatively
andaccuracyfeedback.Atotalof192no-feedbacktrialswerethen valenced mood state in the anxious group, repeated-measures
THREATFACILITATESSUBSEQUENTEXECUTIVECONTROL 1295
GLMswereperformedwithineachgrouptoexaminetheeffectsof buttherewasnosuchdifferenceinpreinductionarousalineither
arousal (arousing, nonarousing) and period (preinduction, postin- thehappy(p(cid:1).954)oranxiousgroup(p(cid:1).890).
duction) on self-reported PA and NA from the PANAS. In addi- Finally, to determine whether the anxious and happy mood
tion, paired samples t tests were performed within each group to stateswereequallyarousing,wedirectlycomparedarousalscores
compare the arousing (anxious, happy) conditions to the nonar- for the happy and anxious groups using independent samples t
ousing (neutral) condition on mean corrugator activity, a second tests. The postinduction arousal score was higher in the anxious
indexofmoodvalence. groupthaninthehappygroup,t(59)(cid:1)2.11,p(cid:1).039.However,
Inthehappygroup,asignificantinteractionofArousal(cid:4)Period the preinduction arousal score was also marginally higher in the
emergedforPAscores,F(1,30)(cid:1)18.61,p(cid:2).001,(cid:5)2(cid:1).383,but anxious group than in the happy group, t(59) (cid:1) 1.80, p (cid:1) .077.
p
not for NA scores, F(1, 30) (cid:1) 1.10, p (cid:1) .303, (cid:5)2 (cid:1) .035. As Takingthispreinductiondifferenceintoaccount(i.e.,bysubtract-
p
shown in Table 1, postinduction PA was higher for the arousing ing preinduction levels of arousal from postinduction levels), the
(happy) than the nonarousing (neutral) condition (p (cid:2) .001), but groupdifferencewasnolongersignificant,t(59)(cid:1).086,p(cid:1).932.
there was no such difference in preinduction PA (p (cid:1) .291). In Collectively, these arousal results suggest that the happy and
addition, mean corrugator activity was significantly lower (i.e., anxious mood states were achieved with an equivalent degree of
less“frowning”)duringthearousing(happyM(cid:1)0.60,SD(cid:1)0.33) arousalinthetwogroups.
than during the nonarousing (neutral M (cid:1) 0.74, SD (cid:1) 0.35)
condition, t(30) (cid:1) (cid:6)2.42, p (cid:1) .022. Collectively, these PA and
Hypothesis Testing
corrugator results indicate that the happy induction elicited the
intendedpositivemoodstate. Executive control. GLMs were conducted for the arousing
In the anxious group, there was a significant interaction of moodconditionstoassesstheeffectsofthreat(fearfulface,neutral
Arousal(cid:4)PeriodforNAscores,F(1,29)(cid:1)29.26,p(cid:2).001,(cid:5)2(cid:1) face),moodgroup(happy,anxious),andtime(early[blocks1–2],
p
.502,butnotforPAscores,F(1,29)(cid:1)1.82,p(cid:1).187,(cid:5)2(cid:1).059. late[blocks3–4])onconflictscores.Therewerenomaineffects
p
AsshowninTable1,postinductionNAwashigherforthearous- ofthreat,F(1,58)(cid:1)2.75,p(cid:1).103,(cid:5)2(cid:1).045,ormoodgroup,
p
ing(anxious)thanthenonarousing(neutral)condition(p(cid:2).001), F(1, 58) (cid:1) 0.01, p (cid:1) .938, (cid:5)2 (cid:2) .001. There was a significant
p
buttherewasnosuchdifferenceinpreinductionNA(p(cid:1).562). interaction of Time (cid:4) Threat, F(1, 58) (cid:1) 6.07, p (cid:1) .017, (cid:5)2 (cid:1)
p
Inaddition,meancorrugatoractivitywassignificantlygreater(i.e., .095.Forearlytrials,whichoccurredsoonafterthemoodinduc-
more“frowning”)duringthearousing(anxiousM(cid:1)0.69,SD(cid:1) tion,conflictscoresweresignificantlylowerforthefearful(M(cid:1)
0.34)thanduringthenonarousing(neutralM(cid:1)0.60,SD(cid:1)0.32) 88.33, SD (cid:1) 37.03) relative to the neutral (M (cid:1) 107.09, SD (cid:1)
condition, t(29) (cid:1) 2.25, p (cid:1) .032. Collectively, these NA and 43.94)condition,p(cid:1).002.Forlatetrials,however,conflictscores
corrugatorresultsindicatethattheanxiousmoodinductionelicited didnotdifferforthefearful(M(cid:1)106.60,SD(cid:1)42.56)relativeto
theintendednegativemoodstate. the neutral (M (cid:1) 104.70, SD (cid:1) 53.03) condition, p (cid:1) .723.
Next,todeterminewhetherthemoodmanipulationsachieveda Critically,therewasasignificantThreat(cid:4)MoodGroupinterac-
state of arousal in both groups, repeated-measures GLMs were tion,F(1,58)(cid:1)4.37,p(cid:1).041,(cid:5)2(cid:1).070.Intheanxiousgroup,
p
performed within each group to examine the effects of arousal conflictscoresweresignificantlydiminishedfortrialsprecededby
(arousing, nonarousing) and period (preinduction, postinduction) fearful faces relative to neutral faces, p (cid:1) .012, but no such
onself-reportedarousalfromthePANAS.Asexpected,therewere differencewasobservedinthehappygroup,p(cid:1).757(seeFigure
significantArousal(cid:4)Periodinteractionsforthehappy,F(1,30)(cid:1) 2a). There was no significant interaction of Time (cid:4) Threat (cid:4)
7.34, p (cid:1) .011, (cid:5)2 (cid:1) .197, and anxious, F(1, 29) (cid:1) 15.78, p (cid:2) MoodGroup,F(1,58)(cid:2)0.01,p(cid:1).957,(cid:5)2(cid:2).001.
p p
.001, (cid:5)2 (cid:1) .352, groups. As shown in Table 1, postinduction To examine the nature of the conflict effect in the anxious
p
arousal was higher for the arousing (anxious or happy) than the group,afollow-upGLMforonlytheseparticipantswasconducted
nonarousing (neutral) condition in both groups (both ps (cid:2) .001), toexaminetheeffectsofthreatandflanker(incongruent,neutral,
Table1
Self-ReportedAffectBeforeandAftertheMoodInductionsinStudy1
Preinduction Postinduction
Arousingmood Nonarousingmood Arousingmood Nonarousingmood
Anxiousgroup
PA 8.37(3.66) 8.23(3.36) 6.77(2.80) 7.57(3.20)
NA 13.33(3.46) 12.87(4.02) 18.83(6.16)a 11.40(1.48)a
Arousal 15.60(4.98) 15.43(4.66) 17.77(4.22)a 13.70(4.19)a
Happygroup
PA 7.13(3.55) 7.77(3.79) 9.90(3.18)a 7.03(3.39)a
NA 14.00(3.24) 13.71(3.55) 11.84(3.28) 12.45(3.19)
Arousal 13.42(4.48) 13.48(5.09) 15.48(4.23)a 12.23(4.18)a
Note. Mean(SD)self-reportedaffectscoresfromthePositiveandNegativeAffectSchedule(PANAS;Watson,
Clark,&Tellegen,1988).PAandNAdenote,respectively,positiveaffectscoresandnegativeaffectscores.
aMeanssharingthissamesuperscriptwithinarowaresignificantlydifferentatp(cid:2).05.
1296 BIRK,DENNIS,SHIN,ANDURRY
did not differ between the arousing (anxious M (cid:1) 101.34, SD (cid:1)
33.86) and nonarousing (neutral M (cid:1) 104.83, SD (cid:1) 38.34) con-
ditions,t(28)(cid:1)(cid:6)0.56,p(cid:1).577.Similarly,forthehappygroup,
conflict scores did not differ between the arousing (happy M (cid:1)
102.01,SD(cid:1)32.12)andnonarousing(neutralM(cid:1)99.55,SD(cid:1)
29.16)conditions,t(30)(cid:1)0.61,p(cid:1).545.Theseresultsshowthat
arousing mood alone was not sufficient to influence executive
controlefficiencyforeithergroup.
Alerting. GLMswereconductedforthearousingmoodcon-
ditions to assess the effects of threat (fearful face, neutral face),
mood group (happy, anxious), and time (early, late) on alerting
scores.Themaineffectoftimewastheonlysignificantfindingin
this analysis, F(1, 57) (cid:1) 10.31, p (cid:1) .002, (cid:5)2 (cid:1) .153, indicating
p
thatalertingefficiencywashigheroverallinthelaterelativetothe
earlytrials,allotherFs(cid:2)1.89.Thus,therewasnothreat-related
facilitation of alerting efficiency in either group (see Figure 2b),
which suggests specificity in the threat-related facilitation of ex-
ecutivecontrolefficiencyintheanxiousgroup.
Toexaminetheeffectofarousingmoodalone,separatepaired
t tests were conducted for each mood group to compare alerting
scores for the arousing (anxious or happy) condition to the non-
arousing(neutral)condition.Fortheanxiousgroup,alertingscores
did not differ between the arousing (anxious M (cid:1) 18.66, SD (cid:1)
21.82)andnonarousing(neutralM(cid:1)23.11,SD(cid:1)18.80)condi-
tions, t(28) (cid:1) (cid:6)1.03, p (cid:1) .310. Similarly, for the happy group,
alerting scores did not differ between the arousing (happy M (cid:1)
25.93, SD (cid:1) 18.68) and nonarousing (neutral M (cid:1) 28.17, SD (cid:1)
Figure2. Thesetwopanelsdepictattentionnetworkefficiencyscoresin 20.60) conditions, t(29) (cid:1) (cid:6)0.65, p (cid:1) .521. These results show
Study1forreactiontime(RT)differencesasafunctionofhappy(HAP)
that arousing mood alone was not sufficient to influence alerting
andanxious(ANX)moodstatesfortrialsprecededbyfearfulfaces(black
efficiencyforeithergroup.
bars)andneutralfaces(graybars).Thetopfigureshowsconflictscores
(incongruent—congruent)andthebottomfigureshowsalertingscores(no
cue—doublecue).Notethatconflictscoresareinverselyrelatedtoexec- Discussion
utivecontrolefficiency.Errorbarsrepresent(cid:7)1standarderror.
Usinganovel,multimodalmood-inductionprocedure,wefound
thatthreat-relatedstimulifacilitatedsubsequentexecutivecontrol
congruent) on raw reaction times. An interaction of Threat (cid:4) efficiencyduringexperimentallyinducedstateanxiety.Therewas
Flanker,F(2,27)(cid:1)7.46,p(cid:1).003,(cid:5)2(cid:1).356,indicatedthatRT no parallel effect of threat on alerting efficiency, which suggests
p
werefastertoincongruent-flankertrialsprecededbyfearfulfaces that threat had a specific effect on executive control. Finally,
(M (cid:1) 597.71, SD (cid:1) 61.69) versus neutral faces (M (cid:1) 608.67, threat-related facilitation of executive control efficiency occurred
SD(cid:1)75.74),p(cid:1).023.Inaddition,RTwereslowerforcongruent- only during the anxious mood state and not during the equally
flankertrialsprecededbyfearfulfaces(M(cid:1)505.90,SD(cid:1)55.22) arousing,happymoodstate,whichsuggeststhattheeffectofthreat
versus neutral faces (M (cid:1) 497.79, SD (cid:1) 58.64), but this was a on executive control efficiency during experimentally induced
marginallysignificanteffect,p(cid:1).072.RTwerealsosignificantly stateanxietycannotbeexplainedbyarousal.Thesefindingsindi-
slower for neutral-flanker trials preceded by fearful faces (M (cid:1) cate that state anxiety may potentiate the facilitative effect of
494.15 SD (cid:1) 47.88) versus neutral faces (M (cid:1) 486.11, SD (cid:1) low-levelthreatonsubsequentexecutivecontrolefficiency.
49.74),p(cid:1).044.Thispatternofresultsindicatesthatthethreat- TherewerefourimportantlimitationstoStudy1.First,among
related conflict effect described above for participants in an anx- peoplewithhightraitanxiety,whichisthedispositionaltendency
iousmoodwasexplainedbyathreat-relatedspeedingofRTinthe tobeanxious,threathasbeenshowntohaveaninhibitingeffecton
incongruentconditionandamarginalthreat-relatedslowinginthe executive control (Derakshan & Eysenck, 2009; Derryberry &
congruent condition. Thus, the attentional effect of threat among Reed,2002;Eysenck,Derakshan,Santos,&Calvo,2007;Wood,
peopleinananxiousmoodisnotsimplyathreat-relatedenhance- Mathews, & Dalgleish, 2001). It may be that the threat-related
mentofattentiontothecentrallocationorathreat-relatedincrease facilitation of executive control efficiency observed during state
in general effort, which would be manifested as faster responses anxiety in Study 1 would be absent or perhaps even reversed in
for trials preceded by fearful faces relative to neutral faces irre- participants with high trait anxiety. Second, although unlikely
spectiveofflankertype. given random assignment of participants to conditions, it is pos-
Toexaminetheeffectofarousingmoodalone,separatepaired sible that the observed effect of threat on executive efficiency
t tests were conducted for each mood group to compare conflict mighthaveexistedintheseparticipantspriortoanyexperimental
scores for the arousing (anxious or happy) condition to the non- manipulation of mood. Third, although the anxious and happy
arousing(neutral)condition.Fortheanxiousgroup,conflictscores moods were purposefully matched for arousal, the mood states
THREATFACILITATESSUBSEQUENTEXECUTIVECONTROL 1297
might have differed in other undesirable ways. For example, in the high trait-anxiety group (n (cid:1) 59; M (cid:1) 49.39, SD (cid:1) 6.38,
happiness might be uniquely associated with alterations in self- range:40–65).
focused attention (Green, Sedikes, Saltzberg, Wood, & Forzano, Moodmanipulation. Participantswererandomlyassignedto
2003)oragenerallyreducedcognitivecapacity(Mackie&Worth, oneoftwogroups.Sixtyparticipantscompletedtheanxiousmood
1989). Fourth, although we confirmed that the mood manipula- induction,and61participantscompletedtheneutralmoodinduc-
tions induced arousing negative and positive moods using mea- tion.
sures that were sensitive to valence and arousal, we could not Moodmeasures. Inadditiontorecordingcorrugatoractivity,
specificallyconfirmthatparticipantswereanxiousinresponseto participants completed a computerized version of the State–Trait
theanxiousmoodmanipulation.Toaddressthesefourlimitations Inventory of Cognitive and Somatic Anxiety (STICSA; Grös,
whilereplicatingtheprimaryfinding,weconductedStudy2. Antony, Simms, & McCabe, 2007) before and after each task.
Participants reported how much they felt each of 20 descriptions
“rightnow,atthisverymoment”ona4-pointscalefrom1,notat
Study 2
all,to4,verymuchso.Wecomputedasomaticanxietyscoreand
AsinStudy1,weexaminedtheeffectsoflow-levelthreatand acognitiveanxietyscoreforeachtimepointinthestudy.
experimentally induced state anxiety on executive control effi- Participants also reported how they felt before and after each
ciencyasmeasuredusingourmodifiedversionoftheANT.Each task using mood ratings that were modified from the adjective
trialoftheANTwasprecededbyathreateningornonthreatening tripletsusedbyTamir,John,Srivastava,andGross(2007).These
stimulus(fearfulorneutralface,respectively).Forthisfollow-up ratingscapturedanxiety(e.g.,anxious,worried,fearful)andgen-
study, we added a measure of trait anxiety and a premood- eral arousal (e.g., active, alert, keyed up) (see Nitschke, Heller,
inductionANT.Thisenabledustotestwhethertheeffectsofthreat Palmieri, & Miller, 1999). They also included filler items corre-
andstateanxietydependonleveloftraitanxiety,andtoruleout spondingtoothernegativeandpositiveemotionalstatessoasnot
the possibility that our two mood groups differed in attentional toemphasizenegativemoodcategories(andanxietyspecifically)
efficiencypriortothemoodinduction.Moreover,ratherthanusing and thereby reduce demand characteristics (data not reported).
a happy mood induction as our comparison group, participants Participantsreportedhowstronglytheyfelteachsetofadjectives
were randomly assigned to an anxious or neutral mood group. “rightnow,atthisverymoment”ona7-pointscalefrom0,notat
Finally, we replaced the valence and arousal mood-manipulation all,to6,verymuch.
measuresfromStudy1withspecificself-reportmeasuresofanx- ANT. Beforeandafterthemoodinduction,participantscom-
iety before and after every mood induction. Consistent with the pletedatwo-blockversionoftheANTadministeredinStudy1to
rationaleforandfindingsfromStudy1,wetestedthehypothesis assess the efficiency of executive control and alerting attention
that the presentation of a fearful face relative to a neutral face networks. Since the threat-related effects were strongest early in
wouldincreaseexecutivecontrolefficiencyinanxiouspeople. the task during Study 1, the overall length of the ANT was
shortenedinStudy2tofocusonattentionaleffectsoccurringclose
to the end of the mood induction. Based on tests of internal
Method
consistency reliability of the ANT data from Study 1, we deter-
minedthatthetasklengthcouldbehalvedwhilestillmaintaining
ThemethodsusedforStudy2wereidenticaltothosedescribed
adequate reliability (Cronbach’s alpha (cid:1) .70 for all cells of the
forStudy1,exceptasnotedbelow.
studydesign).
Participants
Procedure
One hundred twenty-one undergraduate students from Tufts
Participants completed the tasks in the following order: a dot
University (69 female; M (cid:1) 20.55 years, SD (cid:1) 4.45 years)
age age probe task (data not presented as they are not the focus of this
participated for course credit or monetary compensation. Partici-
report), demographic questionnaire and STAI, STICSA/triplets,
pants were 71.1% Caucasian, 20.7% Asian or Asian American,
baselineANT,STICSA/triplets,moodinduction(ANXorNEU),
4.1% Black or African American, and 0.8% American Indian or
STICSA/triplets,experimentalANT,STICSA/triplets.
AlaskaNative;3.3%declinedtoprovidethisinformation.Ofthe
totalsample,14.9%endorsedbeingofHispanicorigin.
Results
Materials
Data Retention and Analysis
Trait Anxiety Groups Using Mahalanobis distance, three participants were excluded
from the analysis of conflict scores and four participants were
Participants completed the trait form (Y-2) of the State–Trait excludedfromtheanalysisofalertingscoresonthebasisofthese
AnxietyInventoryforAdults(STAI;Spielberger,1983).Thetotal tests (p (cid:2) .001). Additionally, two participants were excluded
scoresfortheSTAIcoveredawidespectrum(M(cid:1)40.79,SD(cid:1) because they represented age outliers within their respective
10.21, median (cid:1) 39, min (cid:1) 21, max (cid:1) 65). Participants with groups(i.e.,greaterthan35yearsold).Finally,oneparticipantwas
scores less than or equal to the median were included in the low automaticallyexcludedfromtheanalysisofconflictscoresdueto
traitanxietygroup(n(cid:1)62;M(cid:1)32.50,SD(cid:1)4.79,range:21–39), missingdata.Onehundredfifteenparticipantswereincludedinthe
andparticipantswithscoresgreaterthanthemedianwereincluded manipulation checks that follow (34 anxious mood, low trait
1298 BIRK,DENNIS,SHIN,ANDURRY
anxiety;24anxiousmood,hightraitanxiety;26neutralmood,low anxious and neutral mood groups for low TA, p (cid:1) .223, or high
trait anxiety; 31 neutral mood, high trait anxiety). There was no TA,p(cid:1).496,participants.Therewasalsonomaineffectoftrait
significant difference in level of trait anxiety for participants anxiety, F(1, 112) (cid:1) 0.810, p (cid:1) .370, (cid:5)2 (cid:1) .007. Therefore,
p
assigned to the anxious (M (cid:1) 39.84, SD (cid:1) 10.25) and neutral despitetheevidenceoftheself-reportmeasuresandunlikeStudy
(M (cid:1) 41.47, SD (cid:1) 10.08) mood conditions, t(114) (cid:1) 0.86, p (cid:1) 1, participants in the anxious mood induction did not exhibit
.392,suggestingthatstateandtraitanxietywerenotconfounded. greaterexpressive“frowning”behavior,ascomparedwithpartic-
ipantsintheneutralinduction.
Manipulation Checks
Accounting for Individual Differences in the
Todeterminewhetherthemoodmanipulationsweresuccessful, Effectiveness of Induced Mood
GLMswereperformedtoassesstheeffectsofmoodgroup(anx-
ious, neutral), trait anxiety group (low TA, high TA), and period Given the great importance of the corrugator measure as an
(preinduction,postinduction)ontheSTICSAandontheanxiety- index of the mood manipulation success that is relatively free of
relevant adjective triplets from the mood ratings. In addition, a demandcharacteristics,theabsenceofamoodeffectoncorrugator
GLM was also performed to examine the effects of mood group activity suggests that anxious mood may not have been success-
(anxious, neutral) and trait anxiety group (low TA, high TA) on fullyinducedinallparticipants.Wethereforeretainedonlythose
meancorrugatoractivityduringthemoodmanipulations. participants for whom the target mood was manipulated to a
There was a trend toward a main effect of mood group, F(1, sufficientdegreeinsubsequentanalyses.Specifically,participants
112)(cid:1)2.99,p(cid:1).087,(cid:5)2(cid:1).026,andasignificantmaineffectof intheanxiousmoodgroupwereretainedonlyiftheydisplayedat
p
periodonSTICSAsomaticanxietyscores,F(1,112)(cid:1)15.96,p(cid:2) least a 50% increase in corrugator activity. Participants in the
.001, (cid:5)2 (cid:1) .125. These effects were qualified by a significant neutralmoodgroupwereretainedonlyiftheydisplayedlessthan
p
MoodGroup(cid:4)Periodinteraction,F(1,112)(cid:1)34.82,p(cid:2).001, a 50% increase in corrugator activity. Proportional change in
(cid:5)2 (cid:1) .237, which revealed that preinduction scores were not corrugator activity was computed by subtracting the average ac-
p
different for the anxious (M (cid:1) 14.52, SD (cid:1) 3.41) relative to the tivityduringthe5-speriodatthestartoftheinduction(baseline)
neutral(M(cid:1)14.85,SD(cid:1)3.41)group,p(cid:1).620,butpostinduc- from the average activity during the 5-s period at the end of the
tionscoresweregreaterfortheanxious(M(cid:1)16.94,SD(cid:1)4.18) inductionanddividingthisdifferencebythebaseline.
relativetotheneutral(M(cid:1)14.30,SD(cid:1)4.07)group,p(cid:1).001.As Intotal,70participantswereretainedforthefinalanalyses(17
anticipated,therewasamaineffectoftraitanxietysuchthathigh anxious mood, low trait anxiety; 11 anxious mood, high trait
TAparticipantshadoverallhigherscores(M(cid:1)16.06,SD(cid:1)3.52) anxiety;19neutralmood,lowtraitanxiety;23neutralmood,high
than low TA participants (M (cid:1) 14.19, SD (cid:1) 3.63), F(1, 112) (cid:1) trait anxiety). As intended, the STAI scores for retained partici-
8.34,p(cid:1).005,(cid:5)2(cid:1).069.Therewerenoothersignificanteffects pantsweresimilartoscoresinthefullsample(lowTAgroup:n(cid:1)
involvingtraitanxpiety,however,indicatingthattheimpactofthe 36;M(cid:1)32.56,SD(cid:1)4.78,range:22–39;highTAgroup:n(cid:1)34;
moodinductionsonsomaticanxietydidnotdifferasafunctionof M(cid:1)49.65,SD(cid:1)6.46,range:40–65).Inaddition,therewasno
traitanxiety. significant difference in TA level for participants assigned to the
TheanalysisofSTICSAcognitiveanxietyscoresrevealedonly anxious(M(cid:1)39.68,SD(cid:1)10.29)andneutral(M(cid:1)41.64,SD(cid:1)
theexpectedmaineffectoftraitanxietygroupsuchthat,asabove, 10.31)moodconditions,t(68)(cid:1)0.781,p(cid:1).437,suggestingagain
highTAparticipantshadoverallhigherscores(M(cid:1)17.00,SD(cid:1) thatstateandtraitanxietywerenotconfounded.Participantswho
4.50) than low TA participants (M (cid:1) 12.70, SD (cid:1) 4.50), F(1, were retained based on the criteria above did not differ from
112) (cid:1) p (cid:2) .001, (cid:5)2 (cid:1) .186. There were no other significant participantswhowerenotretainedinTA,overallconflictscores,
effectsorinteractionsp,allotherFs(cid:2)0.91. oroverallalertingscores,allps(cid:1).220.
Theparallelanalysisonscoresfortheanxious,worried,fearful
and active, alert, keyed up triplets revealed significant Mood Hypothesis Testing
Group(cid:4)Periodinteractions,F(1,112)(cid:1)39.21,p(cid:2).001,(cid:5)2(cid:1)
p Executivecontrol. AGLMtestedtheeffectsofthreat(fear-
.259 and F(1, 112) (cid:1) 28.59, p (cid:2) .001, (cid:5)2 (cid:1) .203, respectively.
p fulface,neutralface),moodgroup(neutral,anxious),traitanxiety
Preinduction scores were not different for the anxious relative to
group(lowTA,highTA),andperiod(preinduction,postinduction)
theneutralgroup,ps(cid:8).05butpostinductionscoresweregreater
onconflictscores.2AsinStudy1,therewasamaineffectofthreat,
fortheanxiousrelativetotheneutralgroup,ps(cid:2).05.Therewere
F(1,66)(cid:1)19.91,p(cid:2).001,(cid:5)2(cid:1).204,suchthatconflictscores
no interactive effects involving trait anxiety. However, high TA p
fortrialsprecededbyfearfulfaces(M(cid:1)88.02,SD(cid:1)43.09)were
participantsendorsedfeelingmoreanxious,worried,fearful(M(cid:1)
lowerthanconflictscoresfortrialsprecededbyneutralfaces(M(cid:1)
2.16,SD(cid:1)0.91)thanthelowTAparticipants(M(cid:1)1.56,SD(cid:1)
106.94,SD(cid:1)45.68).Therewasalsoamaineffectofperiod,F(1,
0.91), F(1, 112) (cid:1) 12.47, p (cid:1) .001, (cid:5)2 (cid:1) .100. High TA partic-
p 66) (cid:1) 7.44, p (cid:1) .008, (cid:5)2 (cid:1) .101, such that conflict scores were
ipantsalsoendorsedfeelinglessactive,alert,keyedup(M(cid:1)3.11, p
lowerduringthepostinductionANTthanduringthepreinduction
SD(cid:1)1.37)thanthelowTAparticipants(M(cid:1)3.92,SD(cid:1)1.36),
ANT. No other main effects or interactions were significant, all
F(1,112)(cid:1)10.05,p(cid:1).002,(cid:5)2(cid:1).082.
p
In the analysis of mean corrugator activity during the mood
manipulations, contrary to expectations, there was no effect of 2Unlike in Study 1, a factor of Time (i.e., early vs. late trials) is not
mood group, F(1, 112) (cid:1) 0.13, p (cid:1) .719, (cid:5)p2 (cid:1) .001, and no includedintheanalysesbecauseeachANTconsistsonlyofblocks1and
interactionofmoodgroupandtraitanxiety,F(1,112)(cid:1)1.80,p(cid:1) 2,whicharetogetherequivalenttotheearlytrialsforwhichthefacilitation
.182, (cid:5)2 (cid:1) .016. No significant difference existed between the effectwasdiscoveredinStudy1.
p
THREATFACILITATESSUBSEQUENTEXECUTIVECONTROL 1299
Fs (cid:2) 1.78, all ps (cid:1) .187. However, the critical Threat (cid:4) Mood threatandflanker(incongruent,neutral,congruent)onrawRT.An
Group(cid:4)Periodinteractionwassignificant,F(1,66)(cid:1)8.61,p(cid:1) interaction of Threat (cid:4) Flanker emerged, F(2, 26) (cid:1) 12.64, p (cid:2)
.005,(cid:5)2(cid:1).115. .001, (cid:5)2 (cid:1) .493. As in Study 1, RT were faster to incongruent-
p p
Tointerpretthisinteraction,weranseparateGLMstestingthe flanker trials preceded by fearful faces (M (cid:1) 561.12, SD (cid:1)
effectsofthreat,moodgroup,andtraitanxietygrouponpreinduc- 105.20)versusneutralfaces(M(cid:1)585.79,SD(cid:1)96.20),p(cid:1).007.
tionandpostinductionconflictscores.Asexpected,thepreinduc- RT were marginally slower for congruent-flanker trials preceded
tion analysis revealed no main effects or interactions, all Fs (cid:2) by fearful faces (M (cid:1) 490.30, SD (cid:1) 92.81) versus neutral faces
1.67,allps(cid:1).200.Incontrast,thepostinductionanalysisrevealed (M(cid:1)473.29,SD(cid:1)83.92),p(cid:1).077.RTwerealsosignificantly
amaineffectofthreat,F(1,66)(cid:1)14.02,p(cid:2).001,(cid:5)2(cid:1).175,and, slower for neutral-flanker trials preceded by fearful faces (M (cid:1)
p
most importantly, a significant interaction of Threat (cid:4) Mood 485.53, SD (cid:1) 81.59) versus neutral faces (M (cid:1) 466.71, SD (cid:1)
Group, F(1, 66) (cid:1) 9.37, p (cid:1) .003, (cid:5)2 (cid:1) .124. For the anxious 78.68),p(cid:1).019.AsinStudy1,thispatternofresultssuggeststhat
p
mood group, conflict scores for trials preceded by fearful faces theattentionaleffectofthreatamongpeopleinananxiousmoodis
(M(cid:1)68.34,SD(cid:1)41.96)werelowerthanconflictscorespreceded notsimplyathreat-relatedenhancementofattentiontothecentral
byneutralfaces(M(cid:1)111.90,SD(cid:1)53.44),p(cid:2).001.However, locationorathreat-relatedincreaseingeneraleffort,whichwould
fortheneutralmoodgroup,conflictscoresfortrialsprecededby manifest as faster responses for trials preceded by fearful faces
fearful faces (M (cid:1) 90.52, SD (cid:1) 41.22) were not significantly relativetoneutralfacesforallthreeflankertypes.
different from conflict scores for trials preceded by neutral faces Alerting. A GLM was conducted to examine the effects of
(M (cid:1) 94.89, SD (cid:1) 52.43), p (cid:1) .587. Again, there were no threat (fearful face, neutral face), mood group (neutral, anxious),
significanteffectsinvolvingtraitanxietygroup(seeFigure3a). trait anxiety group (low TA, high TA), and period (preinduction,
Toexaminethenatureoftheconflicteffectintheanxiousmood postinduction)onalertingscores.Therewerenosignificantmain
group, we computed a follow-up GLM to examine the effect of effectsorinteractions,allFs(cid:2)2.27,allps(cid:1).137.Consistentwith
Study1,thissuggestsspecificityinthethreat-relatedfacilitationof
executive control efficiency in the anxious group. See Figure 3b
for a presentation of the alerting scores for all levels of threat,
moodgroup,andtraitanxietygroupforthepostinductionANT.3
Discussion
Consistentwithourhypothesis,participantseffectivelyinduced
into an anxious mood state demonstrated facilitated efficiency of
executive control following stimuli that were mildly threatening
relative to nonthreatening. No such effect was present for partic-
ipantseffectivelyinducedintoaneutralmoodstateandtherewere
no threat-related or overall effects involving state anxiety on
alertingefficiency.Moreover,therewerenoeffectsoftraitanxiety
onexecutivecontrolefficiency.Infact,notonlydidthefacilitative
effect of threat as a function of experimentally manipulated state
anxietynotdependontraitanxiety,traitanxietyalsodidnotaffect
executive control as a main effect. In sum, these results suggest
that an attentional boost follows threat during an anxious state,
regardless of one’s level of anxious disposition (at least within a
normal range); this attentional benefit specifically takes the form
ofimprovedefficiencyofexecutivecontrol.
General Discussion
Taken together, and consistent with DCF, the present studies
demonstrate a replicable, threat-driven improvement in executive
3In order to test for possible effects of trait anxiety as a continuous
measure, the five analyses for the manipulation check were conducted
Figure3. Thesetwopanelsdepictattentionnetworkefficiencyscoresin againasGLMswithSTAIscoreasacovariateinsteadofTAGroupasa
Study2forreactiontime(RT)differencesasafunctionofneutral(NEU) two-level factor. Similarly, for the final set of retained participants, the
andanxious(ANX)moodstatesandlowandhightraitanxiety(TA)for three GLMs for conflict reaction time scores and the GLM for alerting
trialsprecededbyfearfulfaces(blackbars)andneutralfaces(graybars). reactiontimescoreswererepeatedwithSTAIscoreasacovariate.Apart
Theseresultspertaintoparticipantsforwhomthemoodinductionswere from the same four main effects of trait anxiety reported in the original
effective.Thetopfigureshowsconflictscores(incongruent—congruent) analyses (i.e., STICSA somatic; STICSA cognitive, anxious, worried,
andthebottomfigureshowsalertingscores(nocue—doublecue).Note fearfultriplet;active,alert,keyeduptriplet),therewerenoothersignificant
that conflict scores are inversely related to executive control efficiency. effectsinvolvingtraitanxietyinanyofthesenineGLMs,allFs(cid:2)1.64,all
Errorbarsrepresent(cid:7)1standarderror. ps(cid:1).204.
1300 BIRK,DENNIS,SHIN,ANDURRY
control efficiency as a function of low-level threat during state sponds to one or more brain regions (Pessoa, 2009). Substantial
anxiety. Study 1 reveals that this effect is not simply due to the evidence shows that the registration of affective significance is
arousingnatureofanxiety;Study2revealsthatthiseffectdoesnot carriedoutbytheamygdala(Vuilleumier,2005).Perceptualcom-
depend on the level of trait anxiety. Both studies suggest that petition in the visual domain is purported to occur in extrastriate
threat-related facilitation of attention during an anxious state is regions(Pinsk,Doniger,&Kastner,2004).Executivecompetition
specific to executive control efficiency as no such effect was ispurportedtooccurinfrontalregionssuchasanteriorcingulate
observedforalertingefficiency. cortex (ACC; Westlye, Grydeland, Walhovd, & Fjell, 2011),
Althoughpreviousstudieshaveshownthatlow-levelthreatcan which is known to operate in conjunction with dorsolateral pre-
causeprioritizedattentionalprocessinginanxiouspeople,presum- frontal cortex (DLPFC; Silton et al., 2010). We cannot confirm
ablyduetothreat’seffectsonexecutiveattention(e.g.,Fox,Russo, these neural mechanisms in this behavioral study; future studies
Bowles, & Dutton, 2001), few have addressed the combined ef- mustinvestigatewhetherthefacilitatedbehavioraleffectisaccom-
fects of threat and state anxiety on the efficiency of executive panied by an amygdalar response to threat stimuli that correlates
attentionspecifically,andtodatethosestudieshaveonlyexamined withactivationinDLPFCorACC.Ifso,theaccumulatedevidence
individual differences in anxious mood (Dennis et al., 2008; Fi- wouldsuggestthatfunctioninginaprefrontal–amygdalacircuitis
nucane & Power, 2010). In the present studies we randomly themechanismunderlyingtheimprovedexecutivecontrolfollow-
assignedparticipantstomoodconditionsinordertoeliminatethe ingtheperceptionofthreatduringanxiety.
influence of potentially confounding extraneous variables (e.g., Ithasbeensuggestedthatfearfulfaces(unlikeangryfaces)are
enduringpersonalitycharacteristicsthatsystematicallypredispose ambiguous,inthesensethattheysignalthepresenceofthreatbut
particular people to experience state anxiety). This point is espe- notthesource(Whalenetal.,2001).Itmaybethatambiguityas
cially important when attempting, as we did, to disentangle the to source of threat in the context of anxiety was important in
effectsofstateandtraitanxiety. increasing perceived affective significance, which then signaled
It is interesting that experimentally manipulated anxiety failed the need for executive control to resolve the ambiguity. Whalen
toindependentlyfacilitateexecutivecontrolinthesestudies,which (1998)hassuggestedthatthistypeofambiguityismediatedbythe
is consistent with findings of Pacheco–Unguetti and colleagues amygdala, which is consistent with DCF’s suggestion that the
(2010). This null result for anxiety seems at odds with studies amygdalatracksaffectivesignificance.Ifthataccountisaccurate,
suggesting that other negative emotions, specifically fear and onemightaskwhywedidnotobserveacorrespondingincreasein
anger,canfacilitateexecutivecontrolevenintheabsenceofthreat alerting efficiency in conjunction with the increase in executive
(Finucane, 2011). However, Gable and Harmon–Jones (2010a; controlefficiency.Oneexplanationisthattheasteriskcuesusedin
2010b) have suggested that any emotion that engenders suffi- the modified ANT may have been too subtle. Pacheco–Unguetti
ciently high levels of motivational intensity—that is, “impetus to andcolleagues(2010)usedanauditorytonetoinducealertingin
act”—should enhance attentional focusing (Gable & Harmon– their version of the ANT, and observed greater alerting in the
Jones, 2010a; Gable & Harmon–Jones, 2010b). Although atten- anxious compared to the positive group. A stronger alerting ma-
tionalfocusingisnotisomorphicwithexecutivecontrol,itmaybe nipulation may have been necessary to prompt a threat-related
that our experimental manipulation of anxiety only engendered boostinalertinginthepresentstudy.
sufficient motivational intensity to enhance executive control Notably, we did not observe facilitated executive control effi-
when coupled with threat, a notion that is reminiscent of DCF’s ciencyinresponsetothreatinthehappyandneutralmoodstates.
considerationofaffectivesignificance. This result is not fully consistent with the predictions of DCF or
with previous studies (e.g., Cohen et al., 2011). Although the
modelpositsthattheeffectoflow-levelthreatonexecutivecontrol
Implications for Dual Competition Framework
should be greatest during anxiety, it also suggests that threat
AccordingtoDCF,threat-relatedstimulicarryaffectivesignif- shouldinfluenceexecutivecontrol,thoughtoalesserdegree,even
icance, which alters behavior by strengthening sensory represen- forpeoplenotinananxiousstate.Anxiety-relatedvariationsinthe
tations at the perceptual level and by prioritizing attention at the evaluationofaffectivesignificancemayhaveplayedthekeyrole
executive level. The present findings are entirely consistent with in modulating executive control in the present studies; affective
this formulation, at least with regard to manipulated anxiety. We significance may have simply been too low for people in happy
speculate that the mechanism for threat-related facilitation of ex- andneutralmoodstoproduceaneffectonexecutiveattention.
ecutive control during state anxiety was a soft prioritization of
perceptual processing, specifically processing the spatial location
The Role of Conflict Monitoring and Adaptation
ofthreat.Inotherwords,experimentallymanipulatedstateanxiety
led participants to pay greater attention to the spatial location of In line with the conflict-monitoring hypothesis of the ACC
fearfulfacesthanneutralfaces,whichfacilitatedexecutivecontrol (Botvinick, Cohen, & Carter, 2004), we propose that reduced
efficiency.However,wehavenodirectmeasureofattentiontothreat conflict scores in the present studies reflect the outcome of en-
in the ANT. In future work, it would be useful to manipulate the hancedconflictmonitoringandresolutionprocesses.Wespeculate
validitywithwhichthefacespredictthespatiallocationofthetarget. thatthisenhancementisinitiatedbytheamygdala,whichsignals
Iffearfulfaceswereinvalidcuesforthespatiallocationofthetarget, the high affective significance of threat stimuli in the context of
DCF would predict threat-related hampering (certainly not threat- anxiousmood.Next,theamygdalacommunicateswiththeACC,
relatedfacilitation)ofexecutivecontrol. which signals the need to recruit executive control such that
DCF is founded largely on neuroanatomical and functional DLPFC is engaged to facilitate subsequent conflict resolution.
neuroimaging findings such that each aspect of the model corre- When the target finally appears, conflict is reduced, response