Available online at www.sciencedirect.com

ScienceDirectBehavior Therapy 44 (2013) 662–673

www.elsevier.com/locate/bt

A Randomized Trial of Attention Training for Generalized SocialPhobia: Does Attention Training Change Social Behavior?

Brian E. BunnellDeborah C. Beidel

Franklin MesaUniversity of Central Florida

The use of attention training protocols for the treatment ofgeneralized social anxiety disorder (SAD) is undergoingincreased examination. Initial investigations were positivebut more recent investigations have been less supportive ofthe treatment paradigm. One significant limitation ofcurrent investigations is overreliance on self-report. In thisinvestigation, we expanded on initial investigations byusing a multimodal assessment of patient functioning(i.e., including behavioral assessment). Patients with aprimary diagnosis of SAD (n = 31) were randomly assignedto eight sessions of attention training (n = 15) or placebo/control (n = 16). Participants were assessed at pre- andposttreatment via self- and clinician-report of social anxietyas well as anxious and behavioral response to two in vivosocial interactions. Results revealed no differences betweengroups at posttreatment for all study outcome variables,suggesting a lack of effect for the attention training condition.The results are concordantwith recent investigations finding alack of support for the use of attention training as anefficacious treatment for patients with SAD.

Keywords: attention training; social anxiety; bias

INDIVIDUALS PRESENTING WITH SOCIAL ANXIETY DISORDER

(SAD; also termed social phobia) experience anintense fear and apprehension of social situations

Address correspondence to Brian E. Bunnell, M.S., Department ofPsychology, University of Central Florida, 4000 Central FloridaBoulevard,Orlando, FL32816; e-mail: Brian.Bunnell@knights.ucf.edu.0005-7894/44/662-673/$1.00/0© 2013 Association for Behavioral and Cognitive Therapies. Published byElsevier Ltd. All rights reserved.

during which they might be evaluated by others(DSM-IV-TR; American Psychiatric Association[APA], 2000). In addition to social distress, SADis frequently marked by avoidance of the anxiety-provoking situation (e.g., avoiding conversationsand/or giving presentations in a class or at work;Turner, Beidel, Dancu, & Keys, 1986). The physio-logical and psychological distress, anticipatory re-sponse, and behavioral avoidance create significantfunctional impairment (Davidson, Hughes, George,&Blazer, 1993; Edelman&Baker, 2002;Heimberg,Hope, Dodge, & Becker, 1990; Turner et al., 1986;Wittchen & Beloch, 1996). Two subtypes have beenidentified for people with SAD. These include thenongeneralized or specific (social fears are limited tofew, specific social situations) and the generalized(social fears are present in most social situations)subtypes (APA, 2000). SAD is a common disorder,with prevalence rates ranging from 1 to 15% of thegeneral population (APA, 2000; Costello, Egger, &Angold, 2004; Costello, Egger, & Angold, 2005;Heimberg, Stein, Hiripi, & Kessler, 2000) and thetypical age of onset is early to mid-adolescence (ages11 to 15; DeWit, Ogborne, Offord, & MacDonald,1999; Silverman et al., 1999; Weiss & Last, 2001).Earlier-onset SAD is associated with more negativeoutcomes including comorbid anxiety disorders,depression, substance use, and conduct problemslater in life (Beidel & Turner, 1998; Grant et al.,2005; Kessler, 2003; Lecrubier, 1998).The cognitive elements of SAD have received

increased attention recently. According to cognitivetheories, anxious individuals direct their attentiontoward threatening stimuli pertinent to their specificconcerns (e.g., Beck, Emery, & Greenberg, 1985;Eysenck, 1997; Mogg & Bradley, 1998; Rapee &

663attent ion tra in ing for general i z ed soc i al phob ia

Heimberg, 1997; Williams, Watts, MacLeod, &Mathews, 1997). This attentional bias is theorized tobe influential in the etiological and maintenanceaspects of anxiety disorders, particularly SAD, andhas therefore translated into research examiningattentional biases toward symptom-specific stimuliassociated with SAD.Support for the existence of attentional biases in

socially anxious participants has been found using avariety of measurement methods. Examples of theseinclude heightened activation of the anterior cingu-late cortex while viewing socially evaluative faces(Amir et al., 2005), elevated duration of eye gazetoward faces with emotional valence as measuredby infrared eye-tracking systems (Bradley, Mogg,& Millar, 1999), and biased responding towardsocially threatening/related words during modifiedStroop (1935) Color-Naming Tasks (Becker, Rinck,Margraf, & Roth, 2001; Holle, Neely, &Heimberg,1997; Hope, Rapee, Heimberg, & Dombeck, 1990;Maidenberg, Chen, Craske, Bohn, & Bystritsky,1996; Mattia, Heimberg, & Hope, 1993; McNeilet al., 1995; Spector, Pecknold, & Libman, 2003).Computerized dot-probe paradigms (MacLeod,Mathews, & Tata, 1986) have also been used tomeasure attentional biases in this population.Briefly, this paradigm includes the short presentation(typically 500 ms) of two stimuli (e.g., neutral/threatening words or pictures). Upon the disappear-ance of the stimuli, a dot (also referred to as a probe)appears in the place of one of the stimuli. Concep-tually, a participant is biased toward attending to aparticular stimulus type (e.g., a picture of a sociallythreatening face) if he or she recognizes the presenceof the dot with increased speed and accuracy when ittakes the place of that particular stimulus type. Basedon the use of this paradigm, attentional biases insocially anxious samples have been observed towardsocially threatening words (Amir, Elias, Klumpp, &Przeworski, 2003; Asmundson & Stein, 1994;Ononaiye, Turpin,&Reidy, 2007) and later picturesof socially threatening faces (Mogg&Bradley, 2002;Mogg, Philippot, & Bradley, 2004; Pishyar, Harris,&Menzies, 2004; Sposari & Rapee, 2007), as singlewords are unlikely to be the source of threat whileparticipants with SAD engage in social situations(Bradley et al., 1997).Researchers also have examined themanipulability

of attentional biases. For example, some investiga-tions have observed decreases in the severity ofattentional biases following cognitive-behavioraltherapy (Calamaras, Tone, & Anderson, 2012;Mattia et al., 1993; Pishyar, Harris, & Menzies,2008; Price, Tone,&Anderson, 2011). The potentialmalleability of attentional biases prompted thedesigning of specific attention training paradigms

for participants with social anxiety. Studies examin-ing the utility of attention training (also attentionbias modification) treatments in people with SADare emerging. Two initial randomized controltrials (RCTs) used an identical eight-session attentiontraining protocol (i.e., Amir et al., 2009; Schmidt,Richey, Buckner, & Timpano, 2009), and bothconcluded that treatment gains were observed in thetreatment condition above and beyond that of aplacebo control. However, in one trial (i.e., Schmidtet al., 2009), significant between-group differences atposttreatment were observed for only one of the twoclinician-report measures administered, and nodifferences were observed for self-report measuresat this time point.Two infrequently discussed, but nonetheless im-

portant, caveats exist regarding the outcomes ofthese investigations. First is the extent of clinicallysignificant, as opposed to statistically significant,change observed at posttreatment. Specifically,although statistical differences in self- and clinician-reported social anxiety were observed between thetreatment conditions, at posttreatment averagescores on these measures still fell far above the scoresdelineating the clinical range for SAD, as recom-mended by prior literature. For example, averageposttreatment scores on the Social Phobia andAnxiety Inventory (SPAI; Turner, Beidel, & Dancu,1996) were 99.1 (Amir et al., 2009) and 92.47(Schmidt et al., 2009), exceeding established cutoffscores of 60 for probable SAD. Despite self-reportedsocial anxiety, large percentages of treated partici-pants no longer met diagnostic criteria for SAD(50 and 72% for Amir et al., 2009, and Schmidtet al., 2009, respectively). Moreover, neither inves-tigation attempted to examine changes in partici-pants’ social behavior during social interactions,previously identified as a common deficit in patientswith SAD (Beidel, Rao, Scharfstein, Wong, &Alfano, 2010; Herbert et al., 2005; Turner, Beidel,Dancu, & Stanley, 1989). Since the time of theinitial publication, subsequent RCTs have failedto replicate the initial positive outcome (Boettcher,Berger, & Renneberg, 2011; Carlbring et al.,2012; Heeren, Reese, McNally, & Philippot, 2012;Neubauer et al., 2013) whether delivered in personor over the Internet.In addition to the controversy regarding the

efficacy of these interventions for decreasing socialdistress, to date, no study has addressed how theseinterventions affect other aspects of impaired socialbehavior, often found among individuals with SAD.Of particular importance, relative to impairment, isthat an early history of SAD may also result indysfunctional social skills (Beidel et al., 2010; Turneret al., 1989). As children and adolescents with

664 bunnell et al .

SAD progressively avoid social situations, they missopportunities to develop the social skills necessaryfor effective social interaction. Determining treat-ment efficacy must include more than self-report ofdecreased distress. This suggests the importance ofdeveloping treatments that target not only symptomsof anxiety but also social skill deficits. Cognitivetheories of social interaction propose that individualspossess adequate social skills but are unable to usethose skills due to overwhelming distress. Thus,decreasing social anxiety should allow positive socialbehaviors to emerge and assessing social behaviorduring treatment trials may provide an incrementaladvancement in our understanding of the effects ofparticular interventions aswell as provide assessmentof clinically significant change in “real-world”functioning, as recommended by Kazdin (1999).Behavioral assessments of social anxiety and socialbehavior provide just such an opportunity (Norton& Hope, 2001).Therefore, this study had two specific purposes.

First, as noted above, more recent investigations ofthis paradigm have not replicated the initial positiveoutcome. Because these subsequent investigationshave not used the exact treatment paradigm, it isdifficult to know if the failure to replicate is due totreatment variations or simply a failure to replicate.Thus, this investigation constitutes a replication ofthe initial studies of Amir et al. (2009) and Schmidtet al. (2009) in order to assess its effect on socialphobia. Second, consistent with Kazdin (1999), thedependent variables in this investigation wereexpanded to include aspects of social behavior,to determine if changes in social distress lead tochanges in social interactions. The specific hypoth-eses are as follows: (a) at posttreatment the per-centage of participantsmeeting diagnostic criteria forSAD in the treatment condition will be significantlylower than the percentage meeting criteria in theplacebo condition, (b) at posttreatment the meanSPAI score for the treatment condition will besignificantly lower than that of the control condition,and (c) at posttreatment, significant decreases insocial distress will lead to significant improvement insocial behavior during in vivo social interactions, asreflected by increases in duration of eye contact andtime spent verbally engaged in social interactions,relative to the control condition.

Methodprocedure

Following informed consent, participants wereassessed via a clinician-administered diagnosticinterview, self- and clinician-report measures, and abehavioral assessment of social skills at our AnxietyDisorders Clinic in the southeastern United States.

As in previous investigations, participants wereinformed that they would be randomly assigned toone of two groups: one group would receive theanxiety treatment and the other group wouldparticipate in the nontreatment condition. Partici-pants were informed that the purpose of the studywould be to evaluate the usefulness of new computer-based treatments for anxiety. They were thenrandomly assigned to either the attention training(AT; n = 15) or attention control (AC; n = 16)condition using a Microsoft Excel random numbergenerator formula. Participants completed a total ofeight biweekly treatment sessions during which theywere instructed to attempt to identify the letter probe(E or F) as rapidly as possible without sacrificingaccuracy. Following the completion of eight sessions,participants completed the posttreatment assess-ment, which was identical to the pretreatmentassessment. All assessments and treatment sessionswere administered by senior doctoral students inclinical psychology (the first and third authors). Boththe clinicians and participants were blinded totreatment condition until the conclusion of the study.

participants

Participants were recruited via community adver-tisements targeting “shy adults.” Participants com-pleted a telephone screen, followed by an in-persondiagnostic interview, self- and clinician-reportmeasures, and a behavioral assessment. Partici-pants who met DSM-IV-TR criteria for a primarydiagnosis of generalized SAD were invited toparticipate in the study. Exclusionary criteria(from Amir et al., 2009; Schmidt et al., 2009)included (a) evidence of severe depression orsuicidal intent; (b) evidence of current substanceabuse or dependence; (c) evidence of current orpast schizophrenia, bipolar disorder, or organicmental disorder; (d) any concurrent psychotherapy;(e) change in pharmacological treatments duringthe 12 weeks prior to study entry; and (f) cognitive-behavioral therapy within the previous 6 months(see Figure 1 for Consort Diagram).A total of 31 adults participated in the study.

Participants in the AT condition ranged from 18 to45 years of age (M = 24.20, SD = 7.99) and thosein the AC condition ranged from 18 to 44 years ofage (M = 24.44, SD = 6.96). The two groups didnot differ significantly in age, F(1, 29) = .008, p =.930, ηp

2 = .000. There were slightly more males(62.5%) than females (37.5%) in the AC condition,whereas the opposite was true in the AT condition(46.7 and 53.3%, respectively) but the differencewas not statistically significant, χ2(1, 31) = .784,p = .376, Φ = .159. Similarly, there were no groupdifferences in race/ethnicity, χ2(4, 31) = 2.372,

Assessed for eligibility (n = 39)

Excluded (n = 6) Not meeting inclusion criteria (n = 2) Declined to participate (n = 5)

Analyzed (n = 15) Excluded from analysis (n = 0)

Lost to posttreatment (substance use) (n = 1) Discontinued intervention (n = 0)

Allocated to intervention (n = 16) Received allocated intervention (n = 16)Did not receive allocated intervention (n = 0)

Lost to posttreatment (n = 0) Discontinued intervention (n = 0)

Allocated to control (n = 16) Received allocated intervention (n = 16)Did not receive allocated intervention (n = 0)

Analyzed (n = 16) Excluded from analysis (n = 0)

Randomized (n = 32)

FIGURE 1 Consort Diagram.

665attent ion tra in ing for general i z ed soc i al phob ia

p = .668, Φ = .277. The AT condition comprisedmostly Caucasians (53.3%), but also includedHispanic/Latino (26.7%), Asian/Asian Indian(13.3%), and African American (6.7%) partici-pants. Caucasians (56.2%) made up the majority ofthe AC condition, which also included Asian/AsianIndian (18.8%), Hispanic/Latino (18.8%), andbiracial (6.3%) participants. There were comorbiddiagnoses within both conditions. Specifically,33.3% of the AT condition and 18.8% of the ACcondition met criteria for a secondary diagnosis,although this difference was not statistically signif-icant, χ2(6, 31) = 8.368, p = .212, Φ = .520.Comorbid diagnoses in the AT condition includedmajor depressive disorder (in partial remission;6.7%), dysthymic disorder (6.7%), panic disorderwithout agoraphobia (6.7%), and specific phobia(13.3%). Comorbid diagnoses in the AC conditionincluded major depressive disorder (in either partialor full remission; 18.8%). See Table 1 for partic-ipant demographics.

assessmentDiagnosis and Severity

Diagnostic interview. The Anxiety DisordersInterview Schedule (ADIS; Brown, Di Nardo, &Barlow, 1994) was used to assess for DSM-IV-TRAxis I diagnoses. The ADIS was administered bytrained doctoral student clinicians who demonstratedboth aptitude and interrater reliability in performanceprior to the start of the study. Training to proficiencyfollowed similar methods used by Schmidt andcolleagues (2009; reviewing ADIS training tapes,observing taped ADIS administration, observing liveADIS administration, and conducting ADIS inter-views with a trained interviewer) in order to facilitatean accurate replication. Diagnoses were determinedduring weekly clinical meetings under the direction ofa licensed clinical psychologist (the second author).Twenty percent of all diagnostic interviews conductedduring the study were audio recorded and rated by ablinded clinician to establish interrater reliability,which was excellent (κ = 1.0).

Table 1Demographics

AT(n = 15)

AC(n = 16)

M (SD) M (SD)

Age 24.20 (7.99) 24.44 (6.96)n (%) n (%)

SexMale 7 (46.7) 10 (62.5)Female 8 (53.36) 6 (37.5)

RaceAfrican American 1 (6.7) 0 (0.0)Asian/Asian Indian 2 (13.3) 3 (18.8)Caucasian 8 (53.3) 9 (56.2)Latino/a 4 (26.7) 3 (18.8)Biracial 0 (0.0) 1 (6.2)

Comorbid diagnosesMajor depressive disorder 1 (6.7) 3 (18.8)Dysthymia 1 (6.7) 0 (0.0)Specific phobia 2 (13.3) 0 (0.0)Panic disorder withoutagoraphobia

1 (6.7) 0 (0.0)

Note. All patients with major depressive disorder were in eitherpartial or full remission. AC = attention control; AT = attentiontraining.

666 bunnell et al .

Clinician-rated measures. The Clinical GlobalImpressions (CGI; Guy, 1976) Severity of Illnessand Improvement Scales are rated on an 8-pointLikert scale (rated 0 to 7; not at all ill/very muchimproved to among the most extremely ill patients/very much worse). The CGI-Severity scale reflectsthe severity of the patient’s condition and theCGI-Improvement scale reflects the degree to whichthe patient improved from pre- to posttreatment,based on the perception of the clinician. Partici-pants with scores of either 2 or 1 (much improvedor very much improved, respectively) on the CGI-Improvement at posttreatment were classified astreatment responders.The Behavioral Avoidance Rating Scale (BARS;

Beidel et al., 2007) is a 7-point Likert scale (rated 0to 6; no avoidance to complete avoidance) devel-oped by the second author and reflects the degree towhich the patient avoids social situations. Twentypercent of all clinician-rated measures were rated bya blinded clinician to establish interrater reliability,which was adequate for the CGI severity (ICC =.91), CGI-Improvement (ICC = 1.00), and BARS(ICC = .79).

Social AnxietyClinician-rated measure. The Brief Social Phobia

Scale (BSPS; Davidson, Potts, Richichi, & Ford,1991) is an 11-item clinician-administered measurethat comprises seven social situations and four

physiological symptoms (i.e., blushing, palpitations,trembling, sweating) commonly experienced byindividuals with SAD. Clinicians rate the patient’sfear and avoidance relative to the social situationsalong with the severity of their physiologicalsymptoms using a 5-point Likert scale (rated 0 to4; none to extreme). Interrater reliability wasadequate for the BSPS (ICC = .88).

Self-report measures. The SPAI (Turner et al.,1989, 1996) is an empirically derived self-reportmeasure that includes 45 items rated on a 7-pointLikert scale (rated 0 to 6; never to always) reflectingthe frequency of the rater’s experiences. The SPAIprovides both a social phobia and agoraphobiasubscale, and a difference score of the two thatindicates a more pure measure of SAD, and whichwas used as the primary outcome measure of thecurrent investigation. The SPAI has demonstratedgood psychometric properties (Beidel, Turner,Stanley, &Dancu, 1989; Bunnell, Joseph, & Beidel,2012; Turner et al., 1989). The internal consistencywas adequate for the SPAI (Cronbach’s α = .97) inthe current sample.The Liebowitz Social Anxiety Scale–Self Report

(LSAS-SR; Liebowitz, 1987) is a 24-item self-reportmeasure that assesses both fear (rated 0 to 3; noneto severe) and avoidance (rated 0 to 3; never tousually) of social interaction and performancesituations. The LSAS-SR has demonstrated goodpsychometric properties (Baker, Heinrichs, Kim, &Hofmann, 2002; Fresco et al., 2001) and scores arecomparable to those of the clinician-rated version(Fresco et al., 2001). The internal consistency wasadequate for the LSAS-SR (Cronbach’s α = .89) inthe current sample.

DepressionThe Beck Depression Inventory–II (BDI-II; Beck,Steer, & Brown, 1996) is a self-report measureof depressive symptoms with 21 items that are ratedon a Likert scale (rated 0 to 3). The BDI-II hasdemonstrated good psychometric properties (Dozois,Dobson,&Ahnberg, 1998). The internal consistencywas adequate for the BDI-II (Cronbach’s α = .91) inthe current sample.

Behavioral Assessment of Social Anxiety andBehaviorA behavioral assessment was used to assess socialbehaviors and anxiety at pre- and posttreatment.The two in vivo social interaction tasks included anadapted Unstructured Conversation Task (UCT;Turner, Beidel, Cooley, Woody, & Messer, 1994)and an Impromptu Speech Task (IST; Beidel et al.,2010). Each participant rated their level of anxietyimmediately following each task using a 9-point

667attent ion tra in ing for general i z ed soc i al phob ia

Likert scale (0 to 8; no distress to extreme distress).Behavioral assessments were video and audiorecorded and behaviors were coded by independentraters (blinded to treatment condition and timeof assessment) using the Noldus Observer XT(Version 10.1; Noldus Information Technology,2010). Briefly, the Noldus Observer XT softwareallows for the coding of behaviors at variousplayback speeds with exact precision (in hun-dredths of seconds). The Observer XT provides anoutput with the duration and frequency of eachbehavior, which can then be used for statisticalanalyses and comparisons.

Unstructured Conversation Task. The UCT(Turner et al., 1994) involved a 5-minute socialinteraction during which participants were given oneof two scenarios, which was randomly selected atpretreatment. The unused scenario was given duringthe posttreatment UCT. The two scenarios were(a) moving into a new house and meeting a newneighbor and (b) meeting someone at a dinner party.Participants were instructed to “get to know [theconfederate] by having a conversation with him/her.”In these unstructured tasks, confederates were trainedto respond to the participant in a pleasantly neutralmanner without leading the conversation, and the sexof the confederate was randomized for each assess-ment. Social behaviors were coded using the NoldusObserverXT as the total duration (in seconds) of eachparticipant’s eye/facial gaze and time spent speakingduring the UCT. Additionally, the participants’ self-reported anxiety during the task was recorded usingthe 9-point Likert scale described above.

Impromptu Speech Task. The IST (Beidel et al.,2010) required participants to prepare a 10-minutespeech and deliver it to an audience of fiveconfederates. Participants were given 3 minutes toprepare the speech using a maximum of three out offive topics provided by the experimenter. Partici-pants were allowed to terminate the speech after3 minutes by holding up a “stop card” if the distressfrom speaking became too great. Social behaviorswere coded, using the Noldus Observer XT, as thetotal duration (in seconds) of each participant’s timespent speaking, eye/facial gaze, and time beforerequesting to end the task during the IST. Escapebehavior (i.e., requesting to end the task early) wasalso coded. Additionally, the participants’ self-reported anxiety during the task was recordedusing the 9-point Likert scale described above.

Treatment CredibilityThree questions regarding treatment credibility(adapted from Borkovec & Nau, 1972) were

administered following the second treatment session.These measures assessed confidence in treatment,perception of logicalness of treatment, and confi-dence in recommending the treatment to a friend.Responses were rated on a 3-point Likert scale (0 to3; not at all to very much).

TreatmentMaterials. Faces used for the treatment program

were selected from a standardized set of emotionalexpressions (Matsumoto & Ekman, 1989). The setincludes pictures of faces of eight individuals (fourmen and four women) displaying neutral andnegative/threatening (i.e., disgust) expressions. Pic-tures were centered horizontally 17.5 cm from the leftedge of the screen and 3.0 cm from the top of thescreen, and there was a 1.5 cm gap between thebottom of the top image and the top of the bottomimage. Pictures were presented against a static light-gray background via LCD computer monitor. Trialswere conducted using E-Prime Professional 2.0(Psychology Software Tools, Inc., Sharpsburg, PA).In order to blind clinicians and participants totreatment condition, an independent research assis-tant created desktop links to each condition that weremasked by ambiguous names (e.g., “Skinner” or“Jung”). These ambiguous names were entered nextto the participants’ names on a tracking sheet toensure that their assigned treatmentwas administered.

Attention training. The attention training proto-col mirrored that conducted by Amir et al. (2009)and Schmidt et al. (2009), who used a modifieddot-probe paradigmoriginally designed byMacLeodet al. (1986). Each trial began with the presentationof a fixation cross (“+”) in the center of the monitorfor 500 ms. Two faces of the same individual werethen presented (one on top and one on bottom) for500 ms. Each pair of faces displayed one of twocombinations of emotions (i.e., neutral and disgust, orneutral and neutral). Following this presentation, aprobe (the letter E or F) replaced one of the two faces,and participants pressed the corresponding button(right or left) on the computer mouse to identifyeither a probe of E or F. A new trial began followingthe participants’ response. During each session,participants observed a total of 160 trials, 128(80%) of which included the presentation of theprobe in place of a neutral face: 2 (disgust faceposition: top or bottom) × 2 (probe type: E or F) × 8(person) × 4 (repetition). The remaining 32 (20%)trials included only neutral faces: 2 (probe type: E orF) × 2 (probe position: top or bottom) × 8 (person).

Attention control. The AC condition replicatedthe AT condition with the exception of the frequency

668 bunnell et al .

in which the probe appeared in the place of neutralexpressions rather than those of disgust. A total of160 trials were conducted and included 64 trials(40%; neutral-disgust) where the probe replaced thedisgust face, 64 trials (40%; neutral-disgust) wherethe probe replaced the neutral face, and 32 trials(20%) where only neutral faces were presented.

Resultsstatistical and power analyses

A power analysis was conducted using the averageeffect size (Cohen’s d) found at posttreatment formain outcome measures during the initial twoattention training trials. This effect size was .67.The results of the power analysis with alpha set at .05and a power (1 – β) of .80 indicated that a totalsample size of 29 would be adequate to detectsignificant differences between groups. Preliminaryanalyses indicated that participants in the twotreatment conditions did not differ significantly on

Table 2Means and Standard Deviations of Study Outcome Variables at Pr

Pre

AT AC

M (SD) M (SD)

SeverityCGI-S 5.20 (.86) 5.19 (.75)CGI-I - -BARS 4.27 (.80) 4.13 (.80)

Social anxietyBSPS 48.73 (8.59) 48.43 (8.51)SPAI 106.80 (15.48) 111.24 (23.7LSAS-SR 86.66 (18.25) 76.80 (22.21

DepressionBDI-II 16.00 (10.88) 15.31 (11.06

Behavioral assessmentUCTAnxiety 5.40 (1.59) 5.43 (2.06)Eye gaze (%) 45.22 (18.21) 49.27 (17.83Speaking (%) 38.93 (17.01) 45.15 (14.33ISTAnxiety 6.73 (2.01) 6.56 (1.75)Eye gaze (%) 36.37 (24.83) 40.28 (22.96Speaking (%) 80.51 (19.73) 73.05 (22.64Before escape(s) 272.01 (161.92) 236.62 (109

CredibilityConfidence - -Logicalness - -Recommend - -

Note. d = Cohen’s d. d was calculated as between-group effect size atscores. AC = attention control; AT = attention training; BARS = BehavioBSPS = Brief Social Phobia Scale; CGI-I = Clinical Global ImpressionIST = Impromptu Speech Task; LSAS-SR = Liebowitz Social AnxietyUCT = Unstructured Conversation Task.

any outcomemeasure at pretreatment. Posttreatmentdata were complete for all participants with theexception of behavioral data for one participant inthe AT condition who completed all treatmentsessions and posttreatment assessment measures butrefused to participate in the behavioral assessment atposttreatment. The intention-to-treat approach wasused for this participant when analyzing posttreat-ment behavioral assessment data. Analysis of covari-ance (ANCOVA) was used to compare posttreatmentscores while covarying for pretreatment scores, asrecommended by Van Bruekelen (2006). Chi-squareand analysis of variance (ANOVA) were used toexamine differences between groups for variables thatwere categorical (i.e., treatment responder status,escape during the speech task) or had no pretreat-ment scores (i.e., CGI-Improvement scores, treat-ment credibility). Means, standard deviations, andbetween-group effect sizes for outcome variables aredisplayed in Table 2.

e- and Posttreatment

Post

AT AC

M (SD) M (SD) d

4.67 (.90) 4.81 (.54) .193.67 (.72) 3.75 (.44) .133.87 (.99) 3.88 (.95) .01

43.66 (9.31) 42.93 (11.18) .075) 86.36 (28.88) 95.81 (19.38) .38) 59.93 (20.01) 66.37 (22.67) .30

) 11.93 (10.84) 14.06 (11.33) .19

3.20 (2.24) 3.56 (1.36) .19) 52.32 (19.10) 57.61 (22.03) .25) 55.25 (16.15) 51.90 (16.94) .20

5.20 (2.21) 4.93 (1.95) .13) 32.60 (20.28) 42.96 (29.74) .40) 84.82 (16.54) 89.33 (14.97) .28.92) 346.47 (167.55) 278.73 (132.10) .45

.85 (.66) .80 (.56) .08

.71 (.72) .40 (.50) .50

.93 (.61) .80 (.41) .25

posttreatment and does not account for variation in pretreatmentral Avoidance Rating Scale; BDI-II = BeckDepression Inventory–II;s–Improvement; CGI-S = Clinical Global Impressions–Severity;Scale–Self-Report; SPAI = Social Phobia and Anxiety Inventory;

669attent ion tra in ing for general i z ed soc i al phob ia

diagnosis and severity atposttreatment

All participants met DSM-IV-TR diagnosticcriteria for SAD at posttreatment. Scores on theCGI-Improvement scale were not significantlydifferent between the two conditions, F(1, 29) =.151, p = .701, ηp

2 = .005. One participant in theAT condition (6.7%) was classified as a treatmentresponder (i.e., received a score of 1 or 2 on theCGI-Improvement scale), whereas no participants inthe AC condition were classified as a responder, adifference that was not statistically significant,χ2(1, 31) = 1.102, p = .294, Φ = .189. Consistent-ly, participants in the two treatment conditions didnot differ significantly at posttreatment on CGI-Severity scores,F(1, 28) = .315, p = .579,ηp

2 = .011,or BARS ratings, F(1, 28) = .127, p = .724, ηp

2 =.005.

Social AnxietyClinician-rated measure. Participants in the two

treatment conditions did not differ significantly atposttreatment on BSPS scores, F(1, 28) = .034, p =.856, ηp

2 = .001.

Self-report measures. Participants in the twotreatment conditions did not differ significantly atposttreatment on SPAI, F(1, 28) = .748, p = .394,ηp2 = .026, or LSAS-SR scores, F(1, 28) = 2.343,

p = .138, ηp2 = .080.

DepressionParticipants in the two treatment conditions did notdiffer significantly at posttreatment on BDI-IIscores, F(1, 28) = .654, p = .426, ηp

2 = .023.

Behavioral Assessment of Anxiety andSocial Behavior

Anxiety level during social interactions. Partici-pants in the two treatment conditions did not differsignificantly at posttreatment on self-reportedanxiety during the UCT, F(1, 28) = .315, p = .579,ηp2 = .011, or the IST, F(1, 28) = .058, p = .812,

ηp2 = .002.

Social behaviors during social interactions. For theUCT, there were no significant differences betweentreatment conditions on the percentage of timemaking eye/facial gaze during the task, F(1, 28) =3.452, p = .074, ηp

2 = .110. The two conditionsdiffered significantly on the length of time speakingduring the task. Participants in the AT conditionspoke significantly longer (55.25% of time in thetask) than those in the AC condition (51.90%; d =.20), F(1, 28) = 4.454, p = .044, ηp

2 = .137.For the IST, participants in the two treatment

conditions did not differ significantly at posttreat-

ment on the percentage of time making eye/facialgaze during the task, F(1, 28) = 1.348, p = .255,ηp2 = .046. The two conditions differed significantly

on the length of time speaking during the task.Participants in the AC condition spoke significantlylonger (89.33% of the time in the task) than thosein the AT condition (84.82%, d = .28), F(1, 28) =4.395, p = .045, ηp

2 = .136. Some participants didnot speak at all but stood silently before endingthe task. Escape was defined as the number ofparticipants who requested to end the IST early dueto overwhelming anxiety. Results of the chi-squaretest revealed no group difference in the number ofparticipants who escaped the IST: 86.7% versus100% for the AT and AC conditions, respectively;χ2(1, 31) = 2.280, p = .131,Φ = .271. The groupsalso did not differ significantly on the length oftime they remained in the task prior to escape,F(1, 28) = 1.012, p = .323, ηp

2 = .035.

Treatment CredibilityParticipants in the two treatment conditions did notdiffer significantly on how confident they were intreatment, F(1, 27) = .063, p = .804, ηp

2 = .002, howlogical the treatment seemed to them,F(1, 27) = .127,p = .847, ηp

2 = .064, or how confident they wouldbe in recommending the treatment to a friend,F(1, 27) = .441, p = .512, ηp

2 = .016.

DiscussionThis was the first study to explicitly replicateprevious methodology for attention training forsocial anxiety (Amir et al., 2009; Schmidt et al.,2009) and also expand the assessment of patientoutcome to include behavioral assessment. Theresults of this investigation revealed that all partic-ipants met DSM-IV-TR criteria for SAD at post-treatment despite treatment condition. Consistently,scores on the SPAI did not differ between the twoconditions at posttreatment, even when accountingfor pretreatment differences. Finally, between-groupdifferences on self-reported anxiety and blindedobserver ratings of social behavior during in vivosocial interactions (i.e., the UCT and IST) were notobserved. No differences were found for social skillvariables such as eye/facial gaze, time before escape,or the number of participants who escaped duringthe IST. Only two behavioral indices indicated groupdifferences, although the results did not favor onegroup over the other. Furthermore, the between-group differences represent less than 1 minute ofdifference in talk time and although statisticallydifferent, are not clinically significant.With regard to diagnosis at posttreatment, it

was expected that some participants in the ATcondition would no longer meet criteria for SADgiven rates of recovery reported in prior investigations

Note. AC = attention control; AT = attention training; IST = Impromptu Speech Task;UCT = Unstructured Conversation Task.

0

20

40

60

80

100

UCT IST

Per

cent

of

Tim

e

Task

AT AC

FIGURE 2 Percentage of time speaking during the posttreatmentbehavioral assessment.

670 bunnell et al .

(i.e., Amir et al., 2009; Schmidt et al., 2009), yet thiswas not the case in this investigation. Although weused a different diagnostic interview, diagnoses werereliable and accurate, as reflected by high interraterreliability. During posttreatment diagnostic inter-views, all participants reported experiencing consid-erable distress and avoidance in social situations, fewreported noticing positive change in their condition,and all participants requested additional treatment.Furthermore, data from clinician- and self-reportmeasures revealed no group differences on anymeasures (i.e., the BSPS, SPAI, and LSAS-SR).Average scores on self-report measures at posttreat-ment fell at or above the recommended cutoff criteriafor probable SAD (approximately 60 for the LSAS-SR[Rytwinski et al., 2009] and the SPAI [Turner et al.,1996]). Posttreatment between-group differences onthese measures were smaller than those observed inthe Amir et al. (2009) investigation, as reflected bybetween-group effect sizes at posttreatment (d = .30to .38).A close perusal of Table 2 reveals decreases on all

measures for both treatment conditions. This is tobe expected, though, considering nonspecific treat-ment effects that might occur simply from attendingtherapy (e.g., the development of patient–therapistrelationships; see Patterson, 1985, for a review oftherapist-related variables associated with nonspe-cific treatment effects) and underscores the needfor placebo control conditions in RCTs. In someinstances, decreases in symptomology may shift apatient’s status to the extent that he or she no longerexperience functional impairment even if occasionalfears/worries remain. However, in this investigationdecreases in symptom severity did not result in a lossof diagnosis for any patients, suggesting all partici-pants were still functionally impaired.The addition of a behavioral assessment of social

anxiety and social behavior is unique to this investiga-tion and provides an important clinical perspective onpatient functioning. In particular, rather than exam-ining solely statistically significant changes on self-report measures of patient functioning, clinicallysignificant change in real-world functioning may bemore closely examined (Kazdin, 1999). Significantdifferences between groups on speech duration duringthe two tasks were found at posttreatment, althoughthese differences did not favor one group over theother (i.e., the AT group spoke longer during theUCT and the AC group spoke longer during the IST;see Figure 2). However, between-group effect sizesat posttreatment were small (i.e., d = .20 to .28),suggesting that any gains were minimal.Another unique aspect of this investigation is the

assessment of participants’ perception of treatmentcredibility. In addition to the lackof group difference,

both groups endorsed low ratings with regard totheir confidence in the treatment, the logicalness ofthe treatment, and the likelihood that they wouldrecommend the treatment to a friend. This lowconfidence, of course, could be related to the lack oftreatment outcome, although the similarly low levelsof confidence reported by both groups suggests thatconfidence did not relate to the lack of differentialbetween-group effects for the treatment.

implications

The results of the current investigation donot supportthe efficacy of attention training for the treatment ofSAD and are consistent with other recent investiga-tions finding null effects for modified versions of thistreatment (Boettcher et al., 2011; Carlbring et al.,2012; Heeren et al., 2012; Neubauer et al., 2013).Consistent with Emmelkamp’s (2012) statement that“there is no robust evidence that attention training isof clinical value,” the results of this study providefurther support for this claim. It is unclear why thepositive response of the initial investigations has notwithstood the test of replication, but it is clear thatfurther research is merited. Given the significantprevalence of this disorder, establishing the efficacy ofa range of interventions remains an important goal ofclinical research. Furthermore, efficacious interven-tions that may be administered by sub-doctoral-levelclinicians are necessary. At this time, there arequestions regarding the ability of attentional trainingtherapies to deliver their initial promise.

limitations

There are several limitations of this investigation.First, the limited sample size raises concerns regardingthe adequacy of power to detect significant differencesacross groups. However, an a priori power analysisusing power of .80 and amedium effect size indicated

671attent ion tra in ing for general i z ed soc i al phob ia

that the current sample would be acceptable forthe planned analyses. Furthermore, the currentstudy’s sample size (n = 31) exceeded that of others(e.g., Schmidt et al., 2009) at follow-up (n = 26), theonly time point at which significant differencesbetween groups were observed. Second, the demo-graphics of the current sample appear to differsomewhat from those of Amir et al. (2009). Particu-larly, the average age of the current sample wassomewhat younger and slightly more diverse withrespect to race/ethnicity. It is possible that thesedifferences may have affected the credibility ofcomputer-based intervention modalities and hencetheir effectiveness, although the interplay of thesefactors has yet to be studied for this intervention. Athird limitation of this investigation was the lack of amanipulation check to assess participants’ ability todeduct which condition they had been assigned to. Amajor implication of this includes participants’confidence in their assigned treatment condition.Although not a substitute for a manipulation check,participants rated their confidence in their assignedtreatment condition and these ratings did not differbetween groups, suggesting similar levels of confi-dence despite whether such deductions occurred.Fourth, the generalizability of this study’s behavioralassessment as an analogue for day-to-day socialinteractions may appear limited. However, decadesof research affirm the ability of these tasks toapproximate in vivo interactions and in this studythe tasks elicited a significant level of anxiety in theparticipants (seeNorton&Hope, 2001, for a detailedreview). Clearly, behavioral assessment is an impor-tant tool in understanding functional impairmentrelative to various psychopathologies, particularlySAD. Another limitation of this investigation was thelack of assessment of participants’ attentional biases atpre- and posttreatment. It would be important toverify that a shift in attentional bias occurred as aresult of the attention training protocol, as a lack ofsuch a shift might suggest that the treatment did notfunction as intended. Of note is that the protocol usedduring this investigation was programmed withexactness to mirror that which was used in priorinvestigations (i.e., Amir et al., 2009; Schmidt et al.,2009) the former of which demonstrated that thisprotocol was able to successfully shift participants’attentional biases toward faces of a more neutralemotional valence.

conclusion

This study was an attempt to replicate previousinvestigations of attention training as a treatment foradult SAD (i.e., Amir et al., 2009; Schmidt et al., 2009)using a multimodal assessment of patient functioning.Consistent with other studies investigating modified

versions of this protocol, we were unable to replicatethe positive results of these initial investigations basedon either patient or clinician report, as well asbehavioral assessment of social functioning. Futureinvestigations should seek to further elucidate theexact mechanisms by which these treatments mayaffect levels of social anxiety in patients with SAD.

Conflict of Interest StatementThe authors declare that there are no conflicts of interest.

ReferencesAmerican Psychiatric Association. (2000). Diagnostic and statis-

tical manual of mental disorders (rev. 4th ed.). Washington,DC: Author.

Amir, N., Beard, C., Taylor, C., Klumpp, H., Elias, J., Burns,M., & Chen, X. (2009). Attention training in individualswith generalized social phobia: A randomized controlledtrial. Journal of Counseling and Clinical Psychology, 77,961–973. http://dx.doi.org/10.1037/a0016685

Amir, N., Elias, J., Klumpp, H., & Przeworski, A. (2003).Attentional bias to threat in social phobia: Facilitatedprocessing of threat or difficulty disengaging attention fromthreat? Behaviour Research and Therapy, 41, 1325–1335.http://dx.doi.org/10.1016/j.brat.2003.08.001

Amir,N.,Klumpp,H., Elias, J., Bedwell, J., Yanasak,N.,&Miller,S. (2005). Increased activation of the anterior cingulate cortexduring processing of disgust faces in individuals with socialphobia. Biological Psychiatry, 57, 975–981. http://dx.doi.org/10.1016/j.biopsych.2005.01.044

Asmundson, G. G., & Stein, M. B. (1994). Selective processingof social threat in patients with generalized social phobia:Evaluation using a dot-probe paradigm. Journal of AnxietyDisorders, 8(2), 107–117. http://dx.doi.org/10.1016/0887-6185(94)90009-4

Baker, S. L., Heinrichs, N., Kim, H. J., & Hofmann, S. G. (2002)TheLiebowitz Social Anxiety Scale as a self-report instrument:A preliminary psychometric analysis.Behaviour Research andTherapy, 40, 701–715. http://dx.doi.org/10.1016/S0005-7967(01)00060-2,

Beck, A. T., Emery, G., & Greenberg, R. (1985). Anxietydisorders and phobias: A cognitive perspective. New York,NY: Basic Books.

Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Manual forBeck Depression Inventory: II. San Antonio, TX: Psycho-logical Corporation.

Becker, E. S., Rinck, M., Margraf, J., & Roth, W. T. (2001).The emotional Stroop effect in anxiety disorders: Generalemotionality or disorder specificity? Journal of AnxietyDisorders, 15, 147–159. http://dx.doi.org/10.1016/S0887-6185(01)00055-X

Beidel, D. C., Rao, P. A., Scharfstein, L. A., Wong, N., & Alfano,C.A. (2010). Social skills and social phobia:An investigation ofDSM-IV subtypes. Behaviour Research and Therapy, 48(10),992–1001. http://dx.doi.org/10.1016/j.brat.2010.06.005

Beidel, D. C., & Turner, S. M. (1998). Shy children, phobicadults: The nature and treatment of social phobia.WashingtonDC: American Psychological Association Books.

Beidel, D. C., Turner, S. M., Sallee, F. R., Ammerman, R. T.,Crosby, L. A., & Pathak, S. (2007). SET-C versus fluoxetinein the treatment of childhood social phobia. Journal of theAmerican Academy of Child and Adolescent Psychiatry,46(12), 1622–1632. http://dx.doi.org/10.1097/chi.0b013e318154bb57

672 bunnell et al .

Beidel, D. C., Turner, S. M., Stanley, M. A., & Dancu, C. V.(1989). The Social Phobia and Anxiety Inventory:Concurrent and external validity. Behavior Therapy, 20,417–427. http://dx.doi.org/10.1016/S0005-7894(89)80060-7

Boettcher, J., Berger, T., & Renneberg, B. (2011). Internet-based attention training for social anxiety: A randomizedcontrolled trial. Cognitive Therapy and Research, 36(5),522–536. http://dx.doi.org/10.1007/s10608-011-9374-y

Borkovec, T. D., & Nau, S. D. (1972). Credibility of analoguetherapy rationales. Journal of Behavior Therapy and Exper-imental Psychiatry, 3(4), 257–260. http://dx.doi.org/10.1016/0005-7916(72)90045-6

Bradley, B. P., Mogg, K., Millar, N., Bonham-Carter, C.,Fergusson, E., Jenkins, J., & Parr, M. (1997). Attentionalbiases for emotional faces. Cognition and Emotion, 11,25–42. http://dx.doi.org/10.1080/026999397380014

Bradley, B. P., Mogg, K., & Millar, N. M. (1999). Covert andovert orienting of attention to emotional faces in anxiety.Cognition and Emotion, 14, 789–808. http://dx.doi.org/10.1080/02699930050156636

Brown, T. A., Di Nardo, P. A., & Barlow, D. H. (1994).Anxiety Disorders Interview Schedule forDSM-IV (ADIS-IV).San Antonio, TX: Psychological Corporation/GraywindPublications.

Bunnell, B. E., Joseph, D. L.,&Beidel, D. C. (2012).Measurementinvariance of the Social Phobia and Anxiety Inventory. Journalof Anxiety Disorders, 27(1), 84–91. http://dx.doi.org/10.1016/j.janxdis.2012.09.001

Calamaras, M. R., Tone, E. B., & Anderson, P. L. (2012).A pilot study of attention bias subtypes: Examining theirrelation to cognitive bias and their change followingcognitive behavioral therapy. Journal of Clinical Psychology,68(7), 745–754. http://dx.doi.org/10.1002/jclp.21875

Carlbring, P., Löfqvist, M., Sehlin, H., Amir, N., Rousseau, A.,Hofmann, S., & Andersson, G. (2012). Internet-deliveredattention bias modification training in individuals withsocial anxiety disorder—a double blind randomized con-trolled trial. BMC Psychiatry, 12(66), 1–9. http://dx.doi.org/10.1186/1471-244X-12-66

Costello, E., Egger, H., & Angold, A. (2005). 10-year researchupdate review: The epidemiology of child and adolescentpsychiatric disorders: I. Methods and public health burden.Journal of the American Academy of Child & AdolescentPsychiatry, 44(10), 972–986. http://dx.doi.org/10.1097/01.chi.0000172552.41596.6f

Costello, E. J., Egger, H. L.,&Angold, A. (2004). Developmentalepidemiology of anxiety disorders. In Ollendick, T. H. & J.S.March (Eds.), Phobic and anxiety disorders in children andadolescents: A clinician’s guide to effective psychosocial andpharmacological interventions (pp. 61–91). New York, NY:Oxford University Press.

Davidson, J. R., Hughes, D. L., George, L. K., & Blazer, D. G.(1993). The epidemiology of social phobia: Findings from theDuke Epidemiological Catchment Area Study. PsychologicalMedicine, 23, 709–718. http://dx.doi.org/10.1017/S0033291700025484

Davidson, J. R., Potts, N. L., Richichi, E. A., & Ford, S. M.(1991). The Brief Social Phobia Scale. Journal of ClinicalPsychiatry, 52(Suppl), 48–51.

DeWit, D. J., Ogborne, A., Offord, D. R., & MacDonald, K.(1999). Antecedents of the risk of recovery from DSM-III-Rsocial phobia. Psychological Medicine, 29, 569–582.

Dozois, D. J. A., Dobson, K. S., & Ahnberg, J. L. (1998). Apsychometric evaluation of the Beck Depression Inventory-II.Psychological Assessment, 10(2), 83–89. http://dx.doi.org/10.1037/1040-3590.10.2.83

Edelman, R. J., & Baker, S. R. (2002). Self-reported and actualphysiological responses in social phobia. British Journal ofClinical Psychology, 41, 1–14. http://dx.doi.org/10.1348/014466502163732

Emmelkamp, P. M. G. (2012). Attention bias modification:The emperor’s new suit? BMC Medicine, 10(63), 1–3.http://dx.doi.org/10.1186/1741-7015-10-63

Eysenck, M. (1997). Anxiety and cognition: A unified theory.Hove, UK: Psychology Press.

Fresco, D.M., Coles, M. E., Heimberg, R. G., Liebowitz, M. R.,Hami, S., Stein, M. B., & Goetz, D. (2001). The LiebowitzSocial Anxiety Scale: A comparison of the psychometricproperties of self-report and clinician-administered formats.Psychological Medicine, 31, 1025–1035. http://dx.doi.org/10.1017/S0033291701004056

Grant, B. F., Stinson, F. S., Hasin, D. S., Dawson, D. A., Chou,S., Ruan, W., & Huang, B. (2005). Prevalence, correlates,and comorbidity of pipolar I disorder and Axis I and IIdisorders: Results from the National Epidemiologic Surveyon Alcohol and Related Conditions. The Journal of ClinicalPsychiatry, 66(10), 1205–1215. http://dx.doi.org/10.4088/JCP.v66n1001

Guy, W. (1976). ECDEU assessment manual for psychopharma-cology. Washington, DC: Department of Health, Education,and Welfare.

Heeren, A., Reese, H. E., McNally, R. J., & Philippot, P. (2012).Attention training toward and away from threat in socialphobia: Effects on subjective, behavioral, and physiologicalmeasures of anxiety. Behaviour Research and Therapy, 50,30–39. http://dx.doi.org/10.1016/j.brat.2011.10.005

Heimberg, R. G., Hope, D. A., Dodge, C. S., & Becker, R. E.(1990). DSM-III-R subtypes of social phobia: Comparisonof generalized social phobics and public speaking phobics.Journal of Nervous and Mental Disease, 178(3), 172–179.http://dx.doi.org/10.1097/00005053-199003000-00004

Heimberg, R. G., Stein, M. B., Hiripi, E., & Kessler, R. C.(2000). Trends in the prevalence of social phobia in theUnited States: A synthetic cohort analysis of changes over fourdecades.EuropeanPsychiatry,15(1), 29–37. http://dx.doi.org/10.1016/S0924-9338(00)00213-3

Herbert, J. D., Gaudiano, B. A., Rheingold, A. A., Myers, V. H.,Dalrymple, K., & Nolan, E. M. (2005). Social skills trainingaugments the effectiveness of cognitive behavioral grouptherapy for social anxiety disorder. Behavior Therapy, 36,125–138.

Holle, C., Neely, J. H., & Heimberg, R. G. (1997). The effects ofblocked versus random presentation and semantic relatednessof stimulus words on response to a modified Stroop taskamong social phobics. Cognitive Therapy and Research, 21,681–697. http://dx.doi.org/10.1023/A:1021860324879

Hope, D. A., Rapee, R. M., Heimberg, R. G., & Dombeck, M. J.(1990). Representations of the self in social phobia: Vulner-ability to social threat. Cognitive Therapy and Research, 14,177–189. http://dx.doi.org/10.1007/BF01176208

Kazdin, A. E. (1999). The meanings and measurement ofclinical significance. Journal of Consulting and ClinicalPsychology, 67(3), 332–339. http://dx.doi.org/10.1037/0022-006X.67.3.332

Kessler, R. C. (2003). The impairments caused by social phobia inthe general population: Implications for intervention. ActaPsychiatrica Scandinavica, 108, 19–27. http://dx.doi.org/10.1034/j.1600-0447.108.s417.2.x

Lecrubier, Y. (1998). Comorbidity in social anxiety disorder:Impact on disease burden and management. Journal ofClinical Psychiatry, 59, 33–37.

Liebowitz, M. R. (1987). Social phobia. Modern Problems ofPharmacopsychiatry, 22, 141–173.

673attent ion tra in ing for general i z ed soc i al phob ia

Macleod, C., Mathews, A., & Tata, P. (1986). Attentional biasin emotional disorders. Journal of Abnormal Psychology,95, 15–20. http://dx.doi.org/10.1037/0021-843X.95.1.15

Maidenberg, E., Chen, E., Craske, M., Bohn, P., & Bystritsky, A.(1996). Specificity of attentional bias in panic disorderand social phobia. Journal of Anxiety Disorders, 6, 529–541.http://dx.doi.org/10.1016/S0887-6185(96)00028-X

Matsumoto, D., & Ekman, P. (1989). The Japanese andCaucasian facial expressions of emotion (JACFEE) andneutrals (JACNeuF). San Francisco, CA: San Francisco StateUniversity, Department of Psychology, Intercultural andEmotion Research Laboratory.

Mattia, J. I., Heimberg, R. G., &Hope, D. A. (1993). The revisedStroop color-naming task in social phobics. BehaviourResearch and Therapy, 31(3), 305–313. http://dx.doi.org/10.1016/0005-7967(93)90029-T

McNeil, D.W., Ries, B. J., Taylor, L. J., Boone,M. L., Carter, L. E.,Turk, C. L., & Lewin, M. R. (1995). Comparison of socialphobia subtypes using Stroop tests. Journal of AnxietyDisorders, 9, 47–57. http://dx.doi.org/10.1016/0887-6185(94)00027-8

Mogg, K., & Bradley, B. P. (1998). A cognitive-motivationalanalysis of anxiety. Behaviour Research and Therapy, 36,809–848. http://dx.doi.org/10.1016/S0005-7967(98)00063-1

Mogg, K., & Bradley, B. P. (2002). Selective orienting of attentionto masked threat faces in social anxiety. Behaviour andResearchTherapy, 40, 1403–1414. http://dx.doi.org/10.1016/S0005-7967(02)00017-7

Mogg,K., Philippot, P.,&Bradley, B. P. (2004). Selective attentionto angry faces in clinical social phobia. Journal of AbnormalPsychology, 113(1), 160–165. http://dx.doi.org/10.1037/0021-843X.113.1.160

Neubauer, K., von Auer, M., Murray, E., Petermann, F.,Helbig-Lang, S., & Gerlach, A. L. (2013). Internet-deliveredattention modification training as a treatment for socialphobia: A randomized controlled trial. Behaviour Researchand Therapy, 51(2), 87–97. http://dx.doi.org/10.1016/j.bbr.2011.03.031

Noldus Information Technology. (2010). The ObserverXT (Version 10.1) [Computer software]. Wageningen,Netherlands: Author.

Norton, P. J., & Hope, D. A. (2001). Analogue observationalmethods in the assessment of social functioning in adults.Psychological Assessment, 13(1), 59–72. http://dx.doi.org/10.1037/1040-3590.13.1.59.

Ononaiye, M. S. P., Turpin, G., & Reidy, J. G. (2007).Attentional bias in social anxiety: Manipulation of stimulusduration and social-evaluation anxiety. Cognitive Therapyand Research, 31, 727–740. http://dx.doi.org/10.1007/s10608-006-9096-8

Patterson, C. H. (1985). What is the placebo of psychotherapy?Psychotherapy: Theory, Practice, Training, 22(2), 163–169.http://dx.doi.org/10.1037/h0085489

Pishyar, R., Harris, L. M., &Menzies, R. G. (2004). Attentionalbias for words and faces in social anxiety. Anxiety,Stress, and Coping, 1, 23–36. http://dx.doi.org/10.1080/10615800310001601458

Pishyar, R., Harris, L. M., & Menzies, R. G. (2008). Respon-siveness of measures of attentional bias to clinical changein social phobia. Cognition and Emotion, 22, 1209–1227.http://dx.doi.org/10.1080/02699930701686008

Price, M., Tone, E. B., & Anderson, P. L. (2011). Vigilant andavoidance attention biases as predictors of response to cognitivebehavioral therapy for social phobia.Depression and Anxiety,28, 349–353. http://dx.doi.org/10.1002/da.20791

Rapee, R.M.,&Heimberg, R. G. (1997). A cognitive-behavioralmodel of anxiety in social phobia. Behaviour Research and

Therapy, 35(8), 741–756. http://dx.doi.org/10.1016/S0005-7967(97)00022-3

Rytwinski, N. K., Fresco, D. M., Heimberg, R. G., Coles, M. E.,Liebowitz, M. R., Cissell, S., . . . Hofmann, S. G. (2009).Screening for social anxiety disorder with the self-reportversion of the Liebowitz Social Anxiety Scale.Depression andAnxiety, 26(1), 34–38. http://dx.doi.org/10.1002/da.20503

Schmidt, N., Richey, J., Buckner, J., & Timpano, K. (2009).Attention training for generalized social anxiety disorder.Journal of Abnormal Psychology, 118, 5–14. http://dx.doi.org/10.1037/a0013643

Silverman,W. K., Kurtines,W.M., Ginsburg, G. S.,Weems, C. F.,Rabian, B., & Serafini, L. T. (1999). Contingency manage-ment, self-control, and education support in the treatment ofchildhood phobic disorders: A randomized clinical trial.Journal of Consulting and Clinical Psychology, 67(5),675–687. http://dx.doi.org/10.1037/0022-006X.67.5.675

Spector, I. P., Pecknold, J. C., & Libman, E. (2003). Selectiveattentional bias related to the noticeability aspect of anxietysymptoms in generalized social phobia. Journal of AnxietyDisorders, 17, 517–531. http://dx.doi.org/10.1016/S0887-6185(02)00232-3

Sposari, J. A., & Rapee, R. M. (2007). Attentional bias towardfacial stimuli under conditions of social threat in socially phobicand nonclinical participants.Cognitive Therapy and Research,31, 23–37. http://dx.doi.org/10.1007/s10608-006-9073-2

Stroop, J. R. (1935). Studies of interference in serial verbalreactions. Journal of Experimental Psychology, 18(6),643–662. http://dx.doi.org/10.1037/h0054651

Turner, S. M., Beidel, D. C., Cooley, M. R., Woody, S. R., &Messer, S. C. (1994). Amulticomponent behavioral treatmentfor social phobia: Social effectiveness therapy. BehaviourResearch and Therapy, 32(4), 381–390. http://dx.doi.org/10.1016/0005-7967(94)90001-9

Turner, S. M., Beidel, D. C., & Dancu, C. V. (1996). SocialPhobia and Anxiety Inventory: Manual. Toronto, Ontario:Multi-Health Systems.

Turner, S. M., Beidel, D. C., Dancu, C. V., & Keys, D. J. (1986).Psychopathology of social phobia and comparison to avoidantpersonality disorder. Journal of Abnormal Psychology, 95(4),389–394. http://dx.doi.org/10.1037/0021-843X.101.2.326

Turner, S. M., Beidel, D. C., Dancu, C. V., & Stanley, M. A.(1989). An empirically derived inventory to measure socialfears and anxiety: The Social Phobia and Anxiety Inventory.Psychological Assessment, 1, 35–40. http://dx.doi.org/10.1037/1040-3590.1.1.35

Van Bruekelen, G. J. P. (2006). ANCOVA versus change frombaseline: More power in randomized studies, more bias innonrandomized studies. Journal of Clinical Epidemiology,59(12), 920–925. http://dx.doi.org/10.1016/j.jclinepi.2006.02.007

Weiss, D. D., & Last, C. G. (2001). Developmental variations inthe prevalence andmanifestation of anxiety disorders. InM.W.Vasey, & M. R. Dadds (Eds.), The developmental psychopa-thology of anxiety (pp. 27–43). New York, NY: OxfordUniversity Press.

Williams, J. M. G., Watts, F. N., MacLeod, C., & Mathews, A.(1997). Cognitive psychology and emotional disorders.London, UK: Wiley.

Wittchen,H. -U.,&Beloch, E. (1996). The impact of social phobiaon quality of life. International Clinical Psychopharmacology,11(Suppl 3), 15–23. http://dx.doi.org/10.1097/00004850-199606003-00004

RECEIVED: December 11, 2012ACCEPTED: April 22, 2013Available online 30 April 2013