Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=pcem20

Download by: [50.181.76.89] Date: 20 May 2016, At: 07:43

Cognition and Emotion

ISSN: 0269-9931 (Print) 1464-0600 (Online) Journal homepage: http://www.tandfonline.com/loi/pcem20

The effects of emotion priming on visual search insocially anxious adults

Sara A. Haas, Dima Amso & Nathan A. Fox

To cite this article: Sara A. Haas, Dima Amso & Nathan A. Fox (2016): The effects ofemotion priming on visual search in socially anxious adults, Cognition and Emotion, DOI:10.1080/02699931.2016.1180281

To link to this article: http://dx.doi.org/10.1080/02699931.2016.1180281

Published online: 19 May 2016.

Submit your article to this journal

View related articles

View Crossmark data

BRIEF ARTICLE

The effects of emotion priming on visual search in socially anxious adultsSara A. Haasa, Dima Amsob and Nathan A. Foxa

aDepartment of Human Development and Quantitative Methodology, University of Maryland, College Park, MD, USA; bCognitive,Linguistics, and Psychological Sciences, Brown University, Providence, RI, USA

ABSTRACTThis study examined the effects of emotion priming on visual search in participantscharacterised for different levels of social anxiety. Participants were primed withfive facial emotions (angry, fear, happy, neutral, and surprised) and one scrambledface immediately prior to visual search trials involving finding a slanted colouredline amongst distractors, as reaction times and accuracy to target detection wererecorded. Results suggest that for individuals low in social anxiety, being primedwith an angry, surprised, or fearful face facilitated visual search compared to beingprimed with scrambled, neutral or happy faces. However, these same emotionsdegraded visual search in participants with high levels of social anxiety. This studyexpands on previous research on the impact of emotion on attention, finding thatamongst socially anxious individuals, the effects of priming with threat extendbeyond initial attention capture or disengagement, degrading later visual search.

ARTICLE HISTORYReceived 15 September 2015Revised 25 January 2016Accepted 14 April 2016

KEYWORDSThreat; social anxiety; visualsearch; emotion; selectiveattention; facial expressions

The mechanisms underlying the fear system serve anadaptive function, allowing individuals to rapidlydetect threat in the environment and mount anappropriate response (Bar-Haim, Lamy, Pergamin,Bakermans-Kranenburg, & van IJzendoorn, 2007;Öhman, 2005). However, hypervigilant processing ofthreat – a heightened tendency to direct attentionpreferentially to threatening stimuli – is often foundin anxious individuals (MacLeod, Mathews, & Tata,1986; Mathews, Mackintosh, & Fulcher, 1997;Mathews & MacLeod, 1985; Mogg, Mathews, &Eysenck, 1992). This attention bias to threat hasbeen found in both clinically and non-clinicallyanxious individuals in a wide variety of tasks, andmay be involved in both the development and main-tenance of anxiety symptoms (Bar-Haim et al., 2007).

There is, however, evidence to suggest that stress-ful circumstances alter anxious individuals’ pattern ofattention to threat. Bar-Haim et al. (2010), forexample, found that anxious individuals under acutethreat displayed attentional avoidance as measuredwith the dot-probe task, rather than a bias towardsthreat. In addition, Shechner, Pelc, Pine, Fox, andBar-Haim (2012) reported on the plasticity of

attentional bias as a function of context. Individualswho previously received shock in one context dis-played avoidance using the dot probe compared toothers who had not previously received the shock.Emotional context, thus, appears to influence themanner in which threat is detected and responded to.

Two paradigms commonly used to evaluate theeffects of threatening stimuli on attention are thedot-probe task and face-in-the crowd task. In thedot-probe task, participants are instructed to press abutton indicating the location of a target, whichappears in a location previously occupied by one oftwo faces. One of these faces typically expresses aneutral emotion and the other, a negative emotion,such as anger. Studies using the dot-probe paradigmhave noted the presence of an attention bias tothreat in individuals with anxiety disorders (Bar-Haimet al., 2007). Specifically, increased anxiety is associ-ated with an enhanced bias or vigilance to detecttargets that appear in the same location as threaten-ing faces (angry faces) or words compared to non-threatening faces (happy or neutral) compared tonon-anxious participants (Amir, Elias, Klumpp, & Prze-worski, 2003; Mogg & Bradley, 2002; Mogg, Philippot,

© 2016 Informa UK Limited, trading as Taylor & Francis Group

CONTACT Sara A. Haas sahaas@umd.edu

COGNITION AND EMOTION, 2016http://dx.doi.org/10.1080/02699931.2016.1180281

Dow

nloa

ded

by [

50.1

81.7

6.89

] at

07:

43 2

0 M

ay 2

016

& Bradley, 2004). The “face-in-the crowd” paradigm, atype of visual search task, requires participants toidentify faces amidst distractors (e.g. an angry faceamong an array of happy faces). In this task, individ-uals with social anxiety are also quicker duringsearch to detect threatening faces amongst other dis-tracting stimuli compared to happy or neutral faces(Gilboa-Schechtman, Foa, & Amir, 1999; Gilboa-Schechtman, Presburger, Marom, & Hermesh, 2005;Ohman, Flykt, & Esteves, 2001).

These studies generally examine group differencesin attention mechanisms involved in emotion proces-sing. However, a second line of work has focused onthe impact emotion priming may have on subsequentattention. Several studies have demonstrated thatearly visual processes are affected by emotionprimes, fearful primes in particular, and these effectsare lasting. Neural evidence from steady state visualevoked potentials (ssVEP), thought to originate inearly visual cortical regions, has demonstrated that afearful cue facilitates identification of threat (Wieser& Keil, 2014). When a scene containing threat is pre-ceded by a fearful face expression, ssVEP’s to thethreat scene increase, compared to other scenes(Wieser & Keil, 2014). Phelps, Ling, and Carrasco(2006) also demonstrated that fearful expressionshave the capability of altering early vision, specifically,contrast sensitivity. Compared to trials preceded byneutral faces, trials preceded by fearful faceslowered the contrast sensitivity threshold necessarydetect the orientation of the subsequent targetGabor stimuli. The observed advantage for trials pre-ceded by a fearful face in early visual processing, wasparticularly pronounced for the condition in whichthe emotion cues appeared in a single peripherallocation, as opposed to distributed in each of thefour possible target locations (Phelps et al., 2006). Ina subsequent study, Ferneyhough, Kim, Phelps, andCarrasco (2013) replicated the findings of Phelpset al. (2006) in subjects who reported low traitanxiety characteristics. However, subjects whoreported high trait anxiety characteristics displayedcompromised contrast sensitivity on trials where a per-ipheral fearful face was followed by a display in whichthe target Gabor patch was in a different location fromthe original prime. The attentional cost of divertingattention away from the fearful cue in order detectthe orientation of the target Gabor, proved costly forthe visual perception of individuals who scoredhighest on trait anxiety (Ferneyhough et al., 2013).

In another priming study utilising the dot-probetask, Helfinstein, White, Bar-Haim, and Fox (2008)studied adults selected for high and low self-reportedsocial anxiety symptoms, priming them with wordseither conveying relevant affective meaning (e.g.shy, embarrassed) or neutral words, before each trialon the dot probe task. They found that when subjectswith high reported social anxiety symptoms wereprimed with affective words, they did not display abias towards threat, however, they displayed a threatbias after being priming with neutral words. Conver-sely, when subjects with low reported social anxietysymptoms were primed with affective words, theyshowed an attention bias to threat that was notpresent when primed with neutral words. Thus,emotion priming appears to differentially impact sub-sequent attention processes as a function of anxietysymptoms in sub-clinical samples.

A number of studies have also found that emotionpriming facilitates visual search for targets embeddedin an array of distractors. For example, Becker (2009)found that when individuals are primed with fearfulexpressions prior to completing a non-valencedvisual search task (e.g. looking for a picture of ahouse amid other common stimuli such as planesand cars), search performance improves. The improve-ment relative to neutral face-primed performancebecomes more pronounced as the number of distract-ing stimuli increases in search arrays (see also Olatunji,Ciesielski, Armstrong, & Zald, 2011 with consistentfindings). In the classic visual search literature,adding distractors to search arrays results in increasesin reaction times or search slopes (Weierich, Treat, &Hollingworth, 2008). Seeing a fearful face results in aflattening of search slopes, or faster target detectionthan would be expected in the absence of thefearful face prime. In an oddball visual detection utilis-ing the same variations in number of distractor stimulias Becker (2009), Quinlan and Johnson (2011) foundthat participants are fastest to detect threateningstimuli compared to non-threatening stimuli whentrials are repeatedly preceded by a fearful cue com-pared to being repeatedly preceded by a neutral cue.

The findings of these priming studies can be inter-preted in the context of a two-step threat detectionmodel: first, attention is engaged by threat-cueingstimuli, quickly followed by a second stage in whichthere is increased capacity for rapid orienting and per-ception of other stimuli (Becker, 2009). In line with thismodel, Becker (2009) suggested that fearful facial

2 S. A. HAAS ET AL.

Dow

nloa

ded

by [

50.1

81.7

6.89

] at

07:

43 2

0 M

ay 2

016

expressions, and other threat-cueing stimuli, mightmodulate attentional orientating by reallocatingattention faster to scanning the environment insearch of immediate threat.

Indeed, differences in vigilance have beensuggested as contributing to patterns of anxiety.While there are costs in early visual processing forhigh trait anxious individuals reported by Ferney-hough et al. (2013) when primed with fearful cues,visual search performance, is not clearly facilitated orhindered by fearful cues as a function of higher traitanxiety (Olatunji et al., 2011). Olatunji et al. (2011)replicated Becker’s (2009) findings that fearful primesfacilitate more efficient visual search, however nosuch facilitation nor deterioration was observed forindividuals who displayed higher characteristics oftrait anxiety. It may be that task limitations, specificallylack of a variety of set sizes (each visual search arraycontained 12 items), restricted the ability to detectclear individual differences in priming effects as afunction of anxiety symptoms. In a follow-up studyexamining a group of veterans high in post traumaticstress disorder (PTSD) symptoms, and a control group,Olatunji, Armstrong, Bilsky, and Zhao (2015) found nodifferences in visual search performance between thePTSD and control groups when subjects were primedwith fearful faces, however, the high PTSD group wassignificantly slower to detect the target during themoderately difficult visual search array when primedwith an angry prime, as compared to the controlgroup. While cue impact may vary by anxiety sympto-mology, it appears that threat-related cues across avariety of anxiety measures and priming tasks, havethe capability to have costly effects on visual search.

To add to the current literature examining primingeffects on attention across the anxiety continuum, weexamined the impact of priming on attention as afunction of anxiety by designing a task in whichemotional faces were used as primes for a visualsearch task. Of specific interest was examining howperformance on this new task might vary as a functionof social anxiety characteristics, thus expanding ourunderstanding of attention in the context of threat.Several studies have demonstrated reaction timeand attentional performance differences on emotioncueing visual attention tasks as a function of sub-clini-cal levels of anxiety (Ferneyhough et al., 2013; Fox,Henderson, Rubin, Calkins, & Schmidt, 2001; Helfin-stein et al., 2008; Koster, Crombez, Verschuere, & DeHouwer, 2006; Macleod & Mathews, 1988; Mogg,Holmes, Garner, & Bradley, 2008; Olatunji et al.,

2011). In the initial examination of the paradigmused for the current study, we chose to examinesocial anxiety characteristics within a subclinical popu-lation, as this type of anxiety may be the most sensi-tive to priming effects of facial expressions. Ournovel search task may allow greater insight into thespecific attention components that influenceemotional processing in anxious individuals. That is,while the dot-probe task has captured patterns ofattention bias in anxiety, it is not clear if the reactiontime measures reflect initial orienting to threat, diffi-culty disengaging form threat, or a general vigilanceto threat cues in the environment (Weierich et al.,2008). Because the current task includes a conditioncomparable to the dot-probe task, as well as severalvisual search set sizes, this task allows us to disambig-uate these processes by separating emotion exposurefrom the visual search task. None of the aforemen-tioned visual search priming tasks included such acondition. Moreover, Olatunji et al. (2011) and Olatunjiet al. (2015), presented the priming cue immediatelyprior to each visual search, with no time to allow forthe extinguishing of attention capture, or disengage-ment effects. The current study follows the timing ofBecker’s (2009) and Quinlan and Johnson’s (2011)tasks, which both include 600 milliseconds inter-stimulus intervals (ISI) between the priming cue andsearch arrays. Without both a set-size comparable tothe dot-probe paradigm, as well as in ISI to extinguishattention capturing/disengagement effects, it isimpossible to discern whether differential visual per-formance is due to the emotion prime directly impact-ing attentional orienting, or simply measuring theamount of time it takes the participant to disengagefrom the emotion prime.

In this study, participants were primed with fivefacial emotion expressions (angry, fear, happy, neutral,and surprised) and one scrambled face immediatelyprior to every visual search trial, as reaction times toidentify each target was recorded. Two baseline com-parison conditions were included: neutral faces andscrambled faces. We included the surprised emotionas research indicates that high socially anxious individ-uals interpret surprised faces as threatening (Kim et al.,2004; Richards et al., 2002). This may be due to percep-tual similarities between surprised and fearful facialexpressions (raised eyebrows, wide eyes, and openmouth), as well as the ambiguity of a surprisedexpression. Indeed, studies that have investigated thevalence interpretation of surprised faces have shownthat they are not consistently rated as either positive

COGNITION AND EMOTION 3

Dow

nloa

ded

by [

50.1

81.7

6.89

] at

07:

43 2

0 M

ay 2

016

or negative, and are dually valenced and ambiguous(Neta, Davis, & Whalen, 2011).

Angry was included as this emotion has been usedcommonly in dot-probe paradigms as the “threaten-ing” emotion, and demonstrated differences in reac-tion time performance on these visual attention tasksas a function of differences in anxiety levels(MacLeod et al., 1986; Mathews et al., 1997; Mathews& MacLeod, 1985; Mogg et al., 1992). While angryfaces and fearful faces both signal threat, fearful facesindicate ambiguity, such that the location of threat isnot known. We included the fearful face in efforts toreplicate similar findings by Becker (2009), Phelpset al. (2006), and Ferneyhough et al. (2013), who eachfound that when primedwith fearful facial expressions,visual attention improved (visual search improvement:Becker, 2009; low level contrast sensitivity improve-ments: Phelps et al., 2006; and Ferneyhough et al.,2013). Moreover, Ferneyhough et al. (2013) foundthat individuals with high trait anxiety experienceddeterioration in contrast sensitivity when primed withfearful faces, while low trait anxious individuals experi-enced enhanced contrast-sensitivity.

A concern for a number of the tasks previouslymentioned, including the dot-probe, face in thecrowd, visual search, and contrast sensitivity tasks, isthe use of neutral faces for the comparison orcontrol condition. The logic behind using a neutralface is that a neutral face contains all of the percep-tual features, as other emotions, but do not conveyvalence. However, there have been discrepant find-ings with respect to how neutral faces modulatevisual attention as a function of anxiety (Cooney,Atlas, Joormann, Eugène, & Gotlib, 2006; Ferneyhoughet al., 2013). Cooney et al. (2006) found that individ-uals with social anxiety displayed a negative interpret-ation bias towards neutral faces, and displayedheightened patterns of amygdala activation toneutral faces compared to a control group (Cooneyet al., 2006). Neutral faces may not be an idealcontrol condition to examine the effects of emotionson visual attention when studying anxious individ-uals. To address this concern, we included ascrambled face as a control. A scrambled facestimuli contains all the components of a face,however do not allow the participant to perceive anactual face or any type of emotion, making it a poten-tially better comparison condition or baseline thanthe commonly used neutral face.

In addition, by manipulating the processing loadvia the distractor presence we can better capture

the impact of emotion on attention functions. WhileBecker (2009), Olatunji, Armstrong, McHugo, & Zald,2013, and Quinlan and Johnson (2011) utilised avariety of set sizes, these set sizes did not exceed 12items, limiting the interpretability of how visualsearch efficiency changes as a function of increasingdemands in various emotional contexts. Moreover,Olatunji et al. (2011) and Olatunji et al. (2015) utilisedreaction time for the participant to navigate his or hermouse cursor to the target location on the screen asthe dependent measure of visual search latency, limit-ing the specificity and sensitivity of the results ofactual subject latency to detect the target location.Therefore, we varied the number of distractors usedfor visual search, ranging from no distractors, up to29 distractors. As the number of distractors increase,an individual must sample more possible targetlocations, leading to longer target detection times. Ifthreatening emotional contexts differentially affectattentional orienting processes in socially anxious indi-viduals, there should be differences in the slope of thevisual search as a function of the emotion prime andsocial anxiety (Treisman, 1986). Slope refers to thereaction time cost to detect the target as a functionof increasing the number of distractors (Treisman,1986). We predicted that similar to the findingsreported by Olatunji et al. (2011), Olatunji et al.(2015), Quinlan and Johnson (2011), Phelps et al.(2006), Ferneyhough et al. (2013), and Becker (2009),that for participants in the normal or low range ofreported in social anxiety symptoms, being primedwith threat-relevant cues would facilitate visualsearch performance (e.g. smaller slope, indicatingreduced reaction time costs when more distractorsare added). Based on the literature demonstratingdeterioration in visual attention across a variety ofattentional tasks and anxiety dimensions (Amir et al.,2003; Ferneyhough et al., 2013; Gilboa-Schechtmanet al., 1999, 2005; Helfinstein et al., 2008; Mogg &Bradley, 2002; Mogg et al., 2004; Ohman et al., 2001;Olatunji et al., 2011, 2015), we predicted that for par-ticipants scoring higher on social anxiety measures,threatening faces would degrade visual search per-formance (e.g. larger slope, indicating increased reac-tion time costs when more distractors are added). Themajor aims of this study were to further clarifywhether previous findings of costly visual attentionalcontrol when primed with threatening images reflectinitial alerting or attention capture, examine whichemotional contexts affect attention similarly toreported threatening emotional contexts, as well as

4 S. A. HAAS ET AL.

Dow

nloa

ded

by [

50.1

81.7

6.89

] at

07:

43 2

0 M

ay 2

016

examine whether these emotional contexts addition-ally affect later visual attentional processes, specificallyattentional orienting during a visual search.

Methods

Participants

Participants were recruited based on their self-reportanswer on the Liebowitz Scale of Social Anxiety(LSAS) (Heimberg et al., 1999) through an onlinesurvey system. Participants received psychologycourse credit and agreed to be contacted for futurestudies based on their scores. From the 776 studentsthat completed the LSAS, mean LSAS score was calcu-lated, and participants were recruited based on ±1standard deviation from the mean (MLSAS = 45). Partici-pants who scored below 22, or above 68 on the LSASwere invited to participate in the study. To achieveample sample size, the study was opened up to all Uni-versity of Maryland students enrolled in Introductionto Psychology, and an additional unselected 11 sub-jects who had not completed any part of the surveyparticipated. For these subjects, the LSAS adminis-tered at the time of testing.

A power analysis conducted in G*Power 3.1 esti-mated that 85 participants were needed to achieve apower of .95 (a moderate effect size, based upon pre-vious studies with d = .5) when carrying out a 2-group,6 within-subjects conditions Linear Mixed Model(LMM). The final sample included 77 adults (24males) between the ages of 18 and 34 (M= 21.62,SD = 3.16). Self-reported race was as follows: 62% Cau-casian, 21% Asian, 8% African American, 5% Hispanicand 4% of participants chose not to answer. Three par-ticipants who did not complete the entire visualsearch task were excluded from all further analyses,resulting in a final sample of 74 participants (22males). The University Institution Review Boardapproved all procedures and participants of thestudy received verbal and written informed consentin accordance with these procedures.

Questionnaires

The LSAS was administered on-line and again whensubjects came to the lab for the study. The meanelapsed time of completion of LSAS between time 1(pre-selection of the participants) and time 2 (theactual study) was M = 39.6 days, SD = 18.28 days, andranged from 3 to 72 days. LSAS scores obtained

through the online survey system and during the lab-oratory visit were highly correlated r (66) = .79, p< .001. The State portion of the State Trait AnxietyInventory (Spielberger, 1977) was also obtained atthe time of testing for use in analyses to control forstate anxiety at the time of the study (see Table 1).The State Trait Anxiety inventory was only adminis-tered during the laboratory visit, not the onlinesurvey utilised for screen purposes. The laboratory col-lected LSAS scores of the entire sample (n = 77), didnot violate normality tests, and fit a normal distri-bution rather than a bimodal distribution (eventhough we did initially attempt to select to acquireextreme groups) (see Figure 1 and Tables 1 and 2).

Face stimuli

80 pictures of facial expressions were selected fromthe NimStim Inventory (Tottenham et al., 2009). Theselection of NimStim faces used for this paradigmwas based on the following criterion: NimStim facesthat had comparable luminance, no visible jewellery,

Table 1. Correlation of self-reported anxiety questionnaires.

Measure LSAS total STAI state STAI trait

STAI-STATE .620** –STAI-TRAIT .776** .661** –PSWQ .583** .406** .755**

**p < .01.

Figure 1. Histogram of LSAS scores.

COGNITION AND EMOTION 5

Dow

nloa

ded

by [

50.1

81.7

6.89

] at

07:

43 2

0 M

ay 2

016

comparable face size, comparable head tilt angle,minimal hair on face/facial hair, and included avariety of races. The pictures depicting the emotionalfaces were all presented at a size of 2 by 3 in. and cen-tered on the computer screen so that the nose of thestimulus replaced the previously presented crosshair.All faces were grey scaled, and cropped to fit with ina 2 by 3 in. oval, thus controlling for variations incolour (Blau, Naurer, Tottenham, & McCandliss, 2007)(see Figure 1(a)). The racial breakdown of the facesfor males included: two black, one Hispanic, and fivewhite faces. For female faces: two black, one Hispanic,one Asian, and four white faces. The selected NimStimactor numbers were as follows: f03, f05, f06, f07, f09,f11, f13, f18, m23, m24, m26, m27, m34, m37, m41,m42. These included five pictures of eight femaleactors with angry, fearful, happy, neutral, and sur-prised expressions and five pictures of eight maleactors expressing the same emotions (see Figure 2(a)). The scrambled face used, designed by Katoni(2012), was a picture of a female actor presenting aneutral face, divided into various small squares andchanging the position of each square so that theface appeared scrambled (see Figure 2(a)). Moreover,these faces never overlapped with the locations ofany of the visual search targets – this was to insurethat no target location was inhibited or primed byprior visual stimuli.

Visual search stimuli

In the conjunction visual search paradigm, participantswere asked to find a black slanted bar amongst a set ofdistractors. The distractors included white vertical bars,white slanted bars, and black vertical bars. Set sizevaried between 1 (the target black diagonal bar), 5,15, and 30 items. The position of the black diagonalbar varied between eight different positions of aradius of 4 in. from the black crosshair that remainedat the centre of the screen during the visual searchparadigm. The black diagonal bar could appear at 0°,45°, 90°, 135°, 180°, 225°, 270°, and 315° along thisradius. Each visual search display was displayed twicefor each of the 5 emotion conditions; one full set for

the 8 female actors, and one full set for the 8 maleactors, as well as twice for the scrambled face condition.This resulted in 64 trials for each of the 6 conditions. Xand Y coordinates of every distractor item in each visualsearch were randomly assigned using randomizer.org(see Figure 2(b)).

In addition to these trials, there were 24 catch trialsin which there was no actual target present, and par-ticipants were told to do nothing (as opposed to press-ing the spacebar). These catch trials were insertedrandomly as an attention check to make sure that par-ticipants were appropriately completing the task. Catchtrials consisted of 8 trials with 5 distractors, 8 trials with15 distractors, and 8 trials with 30 distractors. Partici-pants completed a total of 528 visual search trials.

Visual search task

The task was presented using E-Prime 2.0 stimuluspresentation software on a Dell laptop with anattached 1080 × 1024 resolution monitor and key-board. Each trial consisted of a face stimulus presentedfor 300 milliseconds, followed by a black crosshaircentred on a gray background for 600 milliseconds.The visual search task was then presented for 2000milliseconds, or until the participant responded bypressing the space bar. The trial ended with anotherblack fixation on a gray background presented for500 milliseconds (see Figure 1(c)). The timing of thepresentation was modelled after Becker (2009). Partici-pants were instructed to press the spacebar as soon asthey identified a black slanted target, or let the trialpass by if they did not detect a black slanted bar. Par-ticipants were not given any a priori informationregarding the face primes in an effort to preventbiasing attention towards the faces. Participants com-pleted 20 practice trials.

E-prime software created a random order of trialsper each participant so that no two participants sawthe same presentation of trials. To avoid the sameemotion showing up several times successively, whichcould lead to habituation (Blau et al., 2007), the taskcycled through each of the 6 priming conditions in arandom order, before repeating any of the 6 priming

Table 2. Anxiety questionnaires descriptives.

Questionnaire Z-score min. Z-score max. Raw mean Std. dev. Min. Max. Skewness Std. error Kurtosis Std. error

LSAS total −1.66 2.45 53.39 29.07 5 125 0.37 0.27 −0.71 0.54STAI-STATE −1.62 3.32 38.05 10.47 21 73 0.85 0.27 0.63 0.54STAI-TRAIT −1.78 2.84 41.27 10.78 22 72 0.73 0.27 0.32 0.54PSWQ −1.98 1.65 53.22 13.69 26 76 −0.17 0.27 −1.15 0.54

6 S. A. HAAS ET AL.

Dow

nloa

ded

by [

50.1

81.7

6.89

] at

07:

43 2

0 M

ay 2

016

conditions. Within each priming condition, and withineach of the 4 set sizes, the 16 faces were randomly pre-sented before each visual search to insure the sameface did not show up more than once per set size.For each of the 24 catch trials, one of the 16 faceswere randomly selected to prime each catch array.

Order of presentation of the visual search arrays wascompletely randomised within each priming condition.The total time to complete the visual search task wasapproximately 20 minutes.

Reaction time data were recorded by E-prime soft-ware and measured the length of time from the

Figure 2. (a) Above is an example of one actress’ emotions, as well as the scrambled face control. (b) Visual search set sizes. These are the 4 setsize versions of the visual search for the target location of 90°. (c) Visual search trial progression.

COGNITION AND EMOTION 7

Dow

nloa

ded

by [

50.1

81.7

6.89

] at

07:

43 2

0 M

ay 2

016

beginning of the visual search presentation until theparticipant detected the target indicated by pressingthe space bar. Catch trials in which there was no realtarget and participants did not respond (after deter-mining the absence of the target) were used toverify participants’ attention and understanding ofthe task. For catch trials, correct responses or omis-sions were scored with a 1 and errors were taggedwith a 0 value. Participants who scored below 90%accuracy for catch trial response were excluded fromfurther analyses (N = 4, M = 65%, SD = 36%).

Reaction time data were summarised and mean RTdata (per emotion and set size) were calculated perparticipant. For each participant’s data, trials with reac-tion times ±2 standard deviations from the partici-pants’ mean reaction time were not included inanalysis of each participant’s RT. After removingmissed trials and outliers, the percentage of usabletrials was calculated for each participant. Participantswith less than 90% of usable trials were excludedfrom further analysis. No further participants wereexcluded based on these criteria. The final sample ana-lysed consisted of 70 of the 74 participants who com-pleted the entire visual search task (MLSAS = 53.39,SDLSAS = 29.07).

Results

To confirm that the four set size conditions were suffi-cient to detect the effects of increasing demandsselective measures of attention, as measured by reac-tion time (Treisman, 1986), we ran a LMM with set sizetype as a fixed-effect (no distractors, 4 distractors, 14distractors, 29 distractors) and participant as arandom effect1. A LMM approach was utilised to incor-porate individual differences and experimental manip-ulations into the analyses. A LMM is a more robustmodel compared to repeated-measures analyses ofvariance (rmANOVA), and among other advantages,accounts for random effects due to participants. Forin-depth explanation and application in visual atten-tion research, see Kliegl, Wei, Dambacher, Yan, andZhou (2011), Kliegl, Masson, and Richter (2010), andMathôt, van der Linden, Grainger, and Vitu (2013).This analysis revealed the expected main effect ofset size, F(3, 1035) = 586.52, p < .001, with reactiontimes increasing with increasing number ofdistractors.

Next, to determine whether the 600 millisecondsITI between the face prime and the visual search wasenough time to extinguish any residual effects of

disengagement (see Becker, 2009) we computed aLMM1. Our dependent measure was RT during setsize 1, participant was our random effect, withemotion (surprised, angry, fear, happy, neutral,scrambled) × Social Anxiety, and covariates StateAnxiety and Sex as fixed effects. There were no maineffects, or interactions, indicating that attention didnot differ as a function of condition F(5, 87.06) =1.12, p = .36, social anxiety F(1, 65.71) = .13, p = .73,state anxiety F(1, 65.71) = .77, p = .38, sex F(1, 65.58)= 3.1, p = .08. The lack of reaction time differencesindicated that baseline reaction time did not differas a function of Social or State Anxiety. All further ana-lyses were conducted on the slope of each individual’svisual search function (Weierich et al., 2008).

Slope calculations

A single slope value was calculated per participant andper emotion priming condition. Slopes were calcu-lated as change in reaction time for target detectionas a function of change in number of items in eachof the set sizes. Each participant’s reaction time datafor each emotion priming condition expression wasfitted to linear slopes, using reaction time to detectthe target during set size 1 as the intercept. Theaverage R2 value for visual search functions of theacross the sample was (M = 0.81, SD = 0.09).

Anxiety scale and model selection

To test whether emotions differentially affected visualsearch efficiency (slope) as a function of social anxiety,a LMM1 was computed using slope as the dependentmeasure, participant as a random effect, and emotion(surprised, angry, fear, happy, neutral, scrambled) ×Social Anxiety2 and covariates State Anxiety and Sexas fixed effects. The Scrambled Slope was the refer-ence or “baseline” condition in the model3. In theLMM, a reference for each categorical variable is uti-lised to examine differences between the reference(one level of the categorical variable) and otherlevels of the categorical variable. In our particularmodel, the selected reference condition was thescrambled condition, so that our estimated fixed-effects were comparing each emotion condition tothe scrambled condition. As mentioned in the intro-duction, we chose to use a scrambled face was usedas our reference condition/ baseline comparisonbecause a scrambled face stimuli contains all the com-ponents of a face, however they do not allow the

8 S. A. HAAS ET AL.

Dow

nloa

ded

by [

50.1

81.7

6.89

] at

07:

43 2

0 M

ay 2

016

participant to perceive an actual face or any type ofemotion, making it a potentially better comparisoncondition or baseline than the commonly usedneutral face. There was no main effect of LSAS scoreF(1, 65.387) = .84, p = .36, or State Anxiety F(1, 65.38)= 1.00, p = .32 and there were trend effects for sexF(1, 64.92) = 3.83, p = .055, and emotion F(5, 90.79) =1.88, p = .1.

There were two significant two-way interactions:Emotion by Social Anxiety F(5, 90.79) = 2.61, p < .05,and Emotion by State Anxiety F(5, 90.79) = 3.13,p < .05, Estimates of fixed-effects indicated that as

State Anxiety increased, all emotion primes, neutralt(87.461) =−2.7, p < .01 d =−.57, happy t(105.41) =−1.80, p = .075, d =−.35, fear t(97.23) =−3.20, p < .01,d =−.65, angry t(98.80) =−3.48, p < .001, d =−.70,and surprised t(101.55) =−3.00, p < .01, d =−.59, facili-tated more efficient visual search compared to beingprimed with a scrambled face. Importantly, estimatesof fixed effects indicated that Social Anxiety, asmeasured by LSAS score, moderated the impact offear t(97.23) = 2.69, p < .01, d = .55, angry t(98.80) =2.73, p < .01, d = .55, and surprised t(101.55) = 2.71,p < .01 d = .54, primes on visual search (see Figure 3

Table 3. LMM estimates of Emotion x LSAS score interaction fixed effects: scrambled condition as reference group.

Parameter Estimate Std. error df t Sig. Lower 95% confidence interval Upper 95% confidence interval

Surprised * LSAS score 0.022 0.008 101.55 2.71 0.008** 0.006 0.022Angry * LSAS score 0.023 0.008 98.80 2.73 0.007** 0.006 0.023Fearful * LSAS score 0.021 0.008 97.23 2.61 0.011** 0.005 0.021Happy * LSAS score 0.013 0.009 105.41 1.49 0.139 −0.004 0.013Neutral * LSAS score 0.008 0.008 87.46 0.98 0.33 −0.008 0.008

**p < .01.

Figure 3. Social anxiety moderates the relations between visual search performance and the effect of emotion primes compared to neutral orscrambled baselines.

COGNITION AND EMOTION 9

Dow

nloa

ded

by [

50.1

81.7

6.89

] at

07:

43 2

0 M

ay 2

016

and Table 3). Visual search performance did not signifi-cantly differ after happy t(105.41) = 1.49, p = .14,d = .29, or neutral t(87.46) = .98, p = .33, d = .21,primes, as a function of social anxiety score.

To examine the generalisability of whether thefacial expressions functioned as threatening or non-threatening, we additionally examined groupingthreatening emotions together, and non-threateningemotions together. Fearful, and Angry emotions func-tioned similarly in our original model, and have beencommonly used in emotion priming tasks, wegrouped those together to create an average“threat” prime condition. Because Happy and func-tioned similarly we grouped those together to createan average “no threat” prime condition. AnotherLMM1 was computed using slope as the dependentmeasure, participant as a random effect, andemotion category (threatening, non-threatening, and

scrambled) × Social Anxiety × State Anxiety as fixedeffects. As in the previous model, scrambled conditionwas the reference or “baseline” condition in themodel. There was no main effect of LSAS score F(1,75.39) = .173, p = .68, or State Anxiety F(1, 75.39)= .001, p = .98 and there was a main effect ofEmotion type F(2, 141.83) = 5.02, p < .01. There weretwo significant two-way interactions: Emotion bySocial Anxiety F(2, 141.83) = 6.12, p < .01, andEmotion by State Anxiety F(2, 141.83) = 9.24, p < .001.

Estimates of fixed-effects indicated that as StateAnxiety increased, both threatening primes(t(139.97) =−4.30, p < .001 d =−.73), and non-threa-tening primes (t(122.23) =−2.91, p < .01 d =−.53)facilitated more efficient visual search compared tobeing primed with a scrambled face. Conversely, esti-mates of fixed effects indicated that as Social Anxiety,as measured by LSAS score, increased, only

Figure 4. Social anxiety moderates the relations between visual search performance and the effect of threatening emotion primes compared toscrambled baselines.

10 S. A. HAAS ET AL.

Dow

nloa

ded

by [

50.1

81.7

6.89

] at

07:

43 2

0 M

ay 2

016

threatening degraded visual search efficiency com-pared to being primed with a scrambled face (t(139.97) =−4.30, p < .001 d = .56). Visual search effi-ciency did not differ as a function of LSAS scorewhen primed with non-threatening primes as com-pared to being primed with a scrambled face (t(122.23) = 1.46, p = .15, d = .26) (Figure 4).

Discussion

The current study aimed to build on the extant litera-ture, which has reported difficulties in disengagementfrom threat in socially anxious individuals. The visualsearch design used in this study successfully differen-tiated between an initial alerting attentional bias, orattention capture towards threatening information,and effects on attentional orienting systemsemployed during visual search (Amir et al., 2003;Fan, McCandliss, Sommer, Raz, & Posner, 2002;Heeren, Lievens, & Philippot, 2011). Specifically, theeffects of emotion priming on performance this stan-dard visual search task emerged as function of self-reported social anxiety symptoms, and demonstratedthat individuals who report higher social anxiety, dis-played degraded visual search performance whenprimed with surprise, angry and fearful faces, ormore generally, threatening faces, as compared toneutral and scrambled faces. Conversely, participantswho are on the lower end of the social anxiety conti-nuum displayed a different and opposite pattern:face primes that indicated threat or ambiguity facili-tated visual search. Results from low socially anxiousparticipants are consistent with previous results withsimilar visual search tasks (Becker, 2009; Ferneyhoughet al., 2013; Helfinstein et al., 2008; Olatunji et al., 2011,2015; Phelps et al., 2006; Quinlan & Johnson, 2011;Wieser & Keil, 2014). In all instances, for individualswho are higher on the anxiety continuum, primingwith negative emotion stimuli decreased either atten-tion bias or search efficiency.

There are a number of possible explanations for thepriming effects found in the current study. It is poss-ible that priming with negative emotions increasedtask difficulty in high anxious individuals. The effi-ciency of visual search may be attenuated in anxiousindividuals due to increased effort expended toprocess negative emotional stimuli. Alternatively, orperhaps as well, the initial threat primes may serveas a distractor, suppressing attention even if taskdemands are not increased. However, the lack of

effect in the set size 1 baseline condition, suggeststhat this may not be the case. If attention wasalerted to, and maintained on threateningexpressions, delayed disengagement from theemotion prime would have resulted in systematicdelays of initiation and orienting of visual search,regardless of visual search difficulty. In addition toexamining the set size 1 baseline condition in isolationacross anxiety and emotion conditions, when baselinereaction time for each emotion type was controlled for(using set size 1 as the intercept in each slope calcu-lation), visual search efficiency differences emergedas a function of both social anxiety levels and threat-related prime conditions. Importantly, these resultsindicate that variations in social anxiety symptomsand how visual attention is distributed in responseto threat emerged as a function of increasingdemands on visual attention. State anxiety did notmoderate the link between social anxiety symptomsand visual search performance, indicating that whiletransient anxiety, or general arousal also significantlyand independently affects visual search efficiency,measures of stable social anxiety affect visual searchefficiency independently and in addition to variablestates of arousal.

It may be that in anxious individuals, negativeemotion primes specifically affect the attentionresource allocation system (see Bar-Haim et al., 2007;Helfinstein et al., 2008), utilising finite resources andthus degrading the subject’s ability for subsequentsearch. Finally, it is possible that priming with negativeemotion stimuli may create more general processinginterference, which would be evident in non-atten-tion-based tasks. Subsequent studies will be neededto carefully compare potential mechanisms.

The findings regarding priming with surprise facialexpressions suggest that surprised faces have a similarimpact on visual search, to fearful faces in both highand low socially anxious individuals. These findingsare consistent with previous research indicating thatsurprise expressions are not consistently rated aseither positive or negative, rather are dually valencedand ambiguous (Neta et al., 2011), as well as researchindicating that high socially anxious individuals inter-pret surprised faces as threatening (Kim et al., 2004;Richards et al., 2002).

Our paradigm was able to detect visual search effi-ciency differences as a function of anxiety andemotion, in contrast to other work (Olatunji et al.,2011, 2015). However, there are several importantlimitations to the current study. First, while our

COGNITION AND EMOTION 11

Dow

nloa

ded

by [

50.1

81.7

6.89

] at

07:

43 2

0 M

ay 2

016

reaction time measures were an improvement onpreviously used target detection measures (e.g.mouse cursor), it is likely that our reaction timemeasurements still underestimated the effectsobserved during both tasks. Future work willemploy eye tracking and ERP methods to gainbetter temporal sensitivity, as well as elucidate theneural underpinnings and time course of variablevisual search performance. Secondly, the faces usedin our study do not reflect the average interactionwith faces that most individuals experience. More-over, various types of anxiety symptoms may be dif-ferentially susceptible to a variety of threat-relatedcontexts. For example, Olatunji et al., 2015 foundthat images of threatening scenarios affected visualattention more than facial expressions, when examin-ing individuals with PTSD. Future iterations of thisparadigm will examine various types of threatprimes, in addition to using facial expressions.Lastly, while the questionnaires obtained during thesession are regarded as reliable self-report measures,self-report measures always vary in accuracy.

While attention bias to threat has similarly beenstudied in children suffering from paediatric anxietydisorders, and in children with subclinical anxietysymptoms, these paradigms have not examined theefficiency of attention with various levels of clutter inthe visual environment as the current paradigmdoes (Pine & Klein, 2008). A child’s everyday environ-ment is full of social stimuli, and various extraneousvisual and auditory stimuli. Thus, it is imperative toexamine attentional processing streams in experimen-tal contexts that attempt replicate attentionaldemands that individuals face in their typical environ-ment. Moreover, because emotion processing skillsand attentional biases develop at a young age, andas a function of experience, an understanding ofearly attentional biases existing in children with pae-diatric anxiety can provide methods of earlier identifi-cation of children at risk for paediatric anxiety, as wellas the development of early intervention programmes(Pine & Klein, 2008).

To address both the developmental trajectory ofvisual attention in this context, as well, as the clinicalimplications, several lines of work can be pursued.Examination of developmental trajectories of threat-vigilance from middle childhood through late adoles-cence using this task, and consideration of how vari-ations in sub clinical anxiety symptoms affects howthese children perform on the present visual searchtask is currently ongoing. Additionally, we are

examining how threat-vigilance may vary as a func-tion of a clinical adolescent anxiety disorder, and uti-lising eye tracking methodologies to obtain moreprecise measures of visual attention during this task.To gain a deeper understanding of the neural signa-tures of visual attention and emotion processingduring the task, electrophysiology measures includ-ing ERPs and EEG will be examined throughout thecourse of the task. The overarching goal of thesestudies is to formulate a more comprehensive anddevelopmental theory of how anxiety interacts withvarious emotional contexts, and subsequentlyaffects attentional processing.

Conclusions

This study builds upon previous literature examiningattention in socially anxious individuals. Using a newlymodified visual search task, we found that primingwith specific facial emotions performance facilitatedor degraded visual search performance as a functionof level of social anxiety. The effect of emotionpriming extended to the processing of non-affectiveperceptual stimuli, indicating that emotion facesimpacted broad, context-independent processingmechanisms. The current findings suggest that theeffortful allocation of attention is affected by socialanxiety symptom severity: individuals who reporthigher social anxiety severity are differentially impactedby exposure to threat and/or ambiguity, such that theirability to subsequently search their environment in thepresence of distracting information is degraded. Thissignificant deterioration in visual search performancedid not occur when primed with Happy or Neutralemotions, and was independent of the individuals’state anxiety or arousal at the time of test. This difficultyin “recovering” from threat exposure may diminishanxious individuals’ ability to flexibly respond toenvironmental demands. This may be evident even inaffect-neutral tasks. Over time, the negative conse-quences that accompany inflexibility may furtherreinforce their initial response to threat/ambiguity, con-tributing to the broad behaviour patterns typicallyobserved in social anxiety: behavioural freezing, rigidity,and withdrawal.

Notes

1. Each Linear Mixed Model was computed using SPSS 23.0,and utilised the diagonal covariance structure for therepeated fixed-effects, and identity covariance structure

12 S. A. HAAS ET AL.

Dow

nloa

ded

by [

50.1

81.7

6.89

] at

07:

43 2

0 M

ay 2

016

for the participant random effect. Each model was fittedusing the restricted maximum likelihood (REML) criterion.The model was estimated using alternative covariancestructures, and while the autoregressive and factor ana-lytic structures also converged, inferences remained thesame for other structures with little or no differences inmodel fit. The final model fit using the diagonal covari-ance structure for the repeated effect, as measured byAkaike’s Information Criterion (AIC), was 1523.74, andfor Schwarz’s Bayesian Criterion (BIC), 1551.70.

2. The same model was run using a median split of LSASscores group analysis and yielded the same results.

3. The same model was run using the Neutral Slope con-dition as the reference, and yielded the same results

Acknowledgements

Wewould like to thank all the individuals who participated in thestudy. We would like to thank Kelley Gunther, Laurel Gordon andLeena Owen for helping with data collection. We would also liketo thank Dr. Koraly Perez-Edgar for her feedback on the manu-script. We would also like to acknowledge Natasha Katoni’swork as an Honors Thesis student in Dima Amso’s lab on a pre-vious iteration of this task.

Funding

The first author received support from the NICHD TrainingProgram in Social Development Grant (NIH T32 HD007542)awarded to the Department of Human Development and Quan-titative Methodology at the University of Maryland by theNational Institutes of Health’s Eunice Kennedy Shriver NationalInstitute of Child Health and Human Development.

Disclosure statement

No potential conflict of interest was reported by the authors.

References

Amir, N., Elias, J., Klumpp, H., & Przeworski, A. (2003). Attentionalbias to threat in social phobia: Facilitated processing of threator difficulty disengaging attention from threat? BehaviourResearch and Therapy, 41(11), 1325–1335.

Bar-Haim, Y., Holoshitz, Y., Eldar, S., Frenkel, T. I., Muller, D.,Charney, D. S.,…Wald, I. (2010). Life-threatening dangerand suppression of attention bias to threat. AmericanJournal of Psychiatry, 167(6), 694–698.

Bar-Haim, Y., Lamy, D., Pergamin, L., Bakermans-Kranenburg, M.J., & van IJzendoorn, M. H. (2007). Threat-related attentionalbias in anxious and nonanxious individuals: A meta-analyticstudy. Psychological Bulletin, 133(1), 1–24.

Becker, M. W. (2009). Panic search: Fear produces efficient visualsearch for nonthreatening objects. Psychological Science, 20(4), 435–437.

Blau, V. C., Naurer, U., Tottenham, N., & McCandliss, B. D. (2007).The face-specific N170 component is modulated by emotional

facial expression. Behavioral and Brain Functions, 3(7). doi:10.1186/1744-9081-3-7

Cooney, R. E., Atlas, L. Y., Joormann, J., Eugène, F., & Gotlib, I. H.(2006). Amygdala activation in the processing of neutralfaces in social anxiety disorder: Is neutral really neutral?Psychiatry Research: Neuroimaging, 148(1), 55–59. doi:10.1016/j.pscychresns.2006.05.003

Fan, J., McCandliss, B. D., Sommer, T., Raz, A., & Posner, M. I.(2002). Testing the efficiency and independence of attentionalnetworks. Journal of Cognitive Neuroscience, 14(3), 340–347.

Ferneyhough, E., Kim, M. K., Phelps, E. A., & Carrasco, M. (2013).Anxiety modulates the effects of emotion and attention onearly vision. Cognition & Emotion, 27(1), 166–176.

Fox, N. A., Henderson, H. A., Rubin, K. H., Calkins, S. D., & Schmidt,L. A. (2001). Continuity and discontinuity of behavioral inhi-bition and exuberance: Psychophysiological and behavioralinfluences across the first four years of life. ChildDevelopment, 72, 1–21. doi:10.1111/1467-8624.00262

Gilboa-Schechtman, E., Foa, E. B., & Amir, N. (1999). Attentionalbiases for facial expressions in social phobia: The face-in-the-crowd paradigm. Cognition & Emotion, 13(3), 305–318.

Gilboa-Schechtman, E., Presburger, G., Marom, S., & Hermesh, H.(2005). The effects of social anxiety and depression on theevaluation of facial crowds. Behaviour Research and Therapy,43(4), 467–474.

Heeren, A., Lievens, L., & Philippot, P. (2011). How does attentiontraining work in social phobia: Disengagement from threat orre-engagement to non-threat? Journal of Anxiety Disorders, 25(8), 1108–1115.

Heimberg, R. G., Horner, K. J., Juster, H. R., Safren, S. A., Brown, E. J.,Schneider, F. R., & Liebowitz, M. R. (1999). Psychometric prop-erties of the liebowitz social anxiety scale. PsychologicalMedicine, 29, 199–212.

Helfinstein, S. M., White, L. K., Bar-Haim, Y., & Fox, N. A. (2008).Affective primes suppress attention bias to threat in sociallyanxious individuals. Behaviour Research and Therapy, 46(7),799–810.

Katoni, N. (2012). The effect of emotion on attention and visualsearch (Senior Honors Thesis). Brown University, Providence, RI.

Kim, H., Somerville, L. H., Johnstone, T., Polis, S., Alexander, A. L.,Shin, L. M., & Whalen, P. J. (2004). Contextual modulation ofamygdala responsivity to surprised faces. Journal ofCognitive Neuroscience, 16(10), 1730–1745.

Kliegl, R., Masson, M. E., & Richter, E. M. (2010). A linear mixedmodel analysis of masked repetition priming. VisualCognition, 18(5), 655–681.

Kliegl, R., Wei, P., Dambacher, M., Yan, M., & Zhou, X. (2011).Experimental effects and individual differences in linearmixed models: Estimating the relationship between spatial,object, and attraction effects in visual attention. Frontiers inPsychology, 1(238), 1–12.

Koster, E. W., Crombez, G., Verschuere, B., & De Houwer, J. (2006).Attention to threat in anxiety-prone individuals: Mechanismsunderlying attentional bias. Cognitive Therapy And Research,30(5), 635–643. doi:10.1007/s10608-006-9042-9

MacLeod, C., & Mathews, A. (1988). Anxiety and the allocation ofattention to threat. Quarterly Journal of ExperimentalPsychology, 40, 653–670.

MacLeod, C., Mathews, A., & Tata, P. (1986). Attentional bias inemotionaldisorders. JournalofAbnormalPsychology,95(1), 15–20.

COGNITION AND EMOTION 13

Dow

nloa

ded

by [

50.1

81.7

6.89

] at

07:

43 2

0 M

ay 2

016

Mathews, A., Mackintosh, B., & Fulcher, E. P. (1997). Cognitivebiases in anxiety and attention to threat. Trends in CognitiveSciences, 1(9), 340–345.

Mathews, A., & MacLeod, C. (1985). Selective processing of threatcues in anxiety states. Behaviour Research and Therapy, 23(5),563–569.

Mathôt, S., van der Linden, L., Grainger, J., & Vitu, F. (2013). Thepupillary light response reveals the focus of covert visualattention. PLoS ONE, 8(10), e78168. doi:10.1371/journal.pone.0078168

Mogg, K., & Bradley, B. P. (2002). Selective orienting of attentionto masked threat faces in social anxiety. Behaviour Researchand Therapy, 40(12), 1403–1414.

Mogg, K., Holmes, A., Garner, M., & Bradley, B. P. (2008).Effects of threat cues on attentional shifting, disengagementand response slowing in anxious individuals. BehaviourResearch and Therapy, 46(5), 656–667. doi:10.1016/j.brat.2008.02.011

Mogg, K., Mathews, A., & Eysenck, M. (1992). Attentional bias tothreat in clinical anxiety states. Cognition & Emotion, 6(2),149–159.

Mogg, K., Philippot, P., & Bradley, B. P. (2004). Selective attentionto angry faces in clinical social phobia. Journal of AbnormalPsychology, 113(1), 160–165.

Neta, M., Davis, F. C., & Whalen, P. J. (2011). Valence resolution ofambiguous facial expressions using an emotional oddballtask. Emotion, 11(6), 1425–1433.

Öhman, A. (2005). The role of the amygdala in human fear:Automatic detection of threat. Psychoneuroendocrinology, 30(10), 953–958.

Ohman, A., Flykt, A., & Esteves, F. (2001). Emotion drives atten-tion: Detecting the snake in the grass. Journal ofExperimental Psychology: General, 130(3), 466–478.

Olatunji, B. O., Armstrong, T., Bilsky, S. A., & Zhao, M. (2015).Threat modulation of visual search efficiency in PTSD: A com-parison of distinct stimulus categories. Psychiatry Research,229(3), 975–982.

Olatunji, B. O., Armstrong, T., McHugo, M., & Zald, D. H. (2013).Heightened attentional capture by threat in veterans withPTSD. Journal of Abnormal psychology, 122(2), 397–405.

Olatunji, B. O., Ciesielski, B. G., Armstrong, T., & Zald, D. H. (2011).Emotional expressions and visual search efficiency: Specificityand effects of anxiety symptoms. Emotion, 11(5), 1073–1079.

Phelps, E. A., Ling, S., & Carrasco, M. (2006). Emotion facilitatesperception and potentiates the perceptual benefits of atten-tion. Psychological Science, 17(4), 292–299.

Pine, D. S., & Klein, R. G. (2008). Anxiety disorders. In M. Rutter, D.Bishop, D. Pine, S. Scott, J. Stevenson, E. Taylor, & A. Thapar(Eds.), Rutter’s child and adolescent psychiatry (5th edn, pp.628–648). London: Blackwell.

Quinlan, P. T., & Johnson, D. P. (2011). The effect of inducingpanic search on the detection of fear-relevant and neutralimages. Visual Cognition, 19(6), 762–784.

Richards, A., French, C. C., Calder, A. J., Webb, B., Fox, R., & Young,A. W. (2002). Anxiety-related bias in the classification ofemotionally ambiguous facial expressions. Emotion, 2(3),273–287.

Shechner, T., Pelc, T., Pine, D. S., Fox, N. A., & Bar-Haim, Y. (2012).Flexible attention deployment in threatening contexts:An instructed fear conditioning study. Emotion, 12(5), 1041–1049.

Spielberger, C. D. (1977). State-trait anxiety inventory for adults:Self-evaluation questionnaire. Redwood City: Mindgarden.

Tottenham, N., Tanaka, J. W., Leon, A. C., McCarry, T., Nurse, M.,Hare, T. A.,… Nelson, C. (2009). The NimStim set of facialexpressions: Judgments from untrained research participants.Psychiatry Research, 168(3), 242–249.

Treisman, A. (1986). Features and objects in visual processing.Scientific American, 255, 114–125.

Weierich, M. R., Treat, T. A., & Hollingworth, A. (2008). Theoriesand measurement of visual attentional processing inanxiety. Cognition and Emotion, 22(6), 985–1018.

Wieser, M. J., & Keil, A. (2014). Fearful faces heighten the corticalrepresentation of contextual threat. NeuroImage, 86, 317–325.

14 S. A. HAAS ET AL.

Dow

nloa

ded

by [

50.1

81.7

6.89

] at

07:

43 2

0 M

ay 2

016

The author has requested enhancement of the downloaded file. All in-text references underlined in blue are linked to publications on ResearchGate.The author has requested enhancement of the downloaded file. All in-text references underlined in blue are linked to publications on ResearchGate.