Cognitive Behaviour Therapy Vol 36, No 4, pp. 193–209, 2007Cognitive Factors that Maintain Social AnxietyDisorder: a Comprehensive Model and its TreatmentImplicationsStefan G. HofmannDepartment of Psychology, Boston University, Boston, MA, USAAbstract. Social anxiety disorder (SAD) is a common, distressing and persistent mental illness.Recent studies have identified a number of psychological factors that could explain the maintenanceof the disorder. These factors are presented here as part of a comprehensive psychologicalmaintenance model of SAD. This model assumes that social apprehension is associated withunrealistic social standards and a deficiency in selecting attainable social goals. When confrontedwith challenging social situations, individuals with SAD shift their attention toward their anxiety,view themselves negatively as a social object, overestimate the negative consequences of a socialencounter, believe that they have little control over their emotional response, and view their socialskills as inadequate to effectively cope with the social situation. In order to avoid social mishaps,individuals with SAD revert to maladaptive coping strategies, including avoidance and safetybehaviors, followed by post-event rumination, which leads to further social apprehension in thefuture. Possible disorder-specific intervention strategies are discussed. Key words: social anxietydisorder; social phobia; maintaining factors; cognitive behavioral therapy; exposure therapy.Received January 23, 2007; Accepted April 26, 2007Correspondence address: Stefan G. Hofmann, PhD, Department of Psychology, Boston University, 648Beacon Street, 6th Floor, Boston, MA 02215-2002, USA. Tel: 617 353 9610; Fax: 617 353 9609.E-mail: [email protected] theories of social anxiety andits clinical expression, social anxiety disorder (SAD), emphasize the role of cognitiveprocesses in the maintenance of the disorder(Clark & Wells, 1995; Leary & Kowalski,1995; Rapee & Heimberg, 1997). The mostpopular and best-researched treatmentapproach is cognitive behavioral group therapy (CBGT) (Heimberg & Becker, 2002). Thistreatment is an adaptation of Beck andEmery’s (1985) cognitive therapy of anxietydisorders. In fact, due to the similaritiesbetween Beck’s cognitive therapy andHeimberg’s treatment of SAD, previousinvestigators have labeled the interventionthat is based on this model ‘‘Beck-HeimbergCBT’’ (Feske & Chambless, 1995, p. 714).This intervention is typically administeredby 2 therapists in 12 weekly 2.5-hour sessions# 2007 Taylor & Francis ISSN 1650-6073DOI 10.1080/16506070701421313to groups of 6 and consists of several distinct,but interwoven, treatment components. In thefirst 2 sessions, patients are taught the BeckianCBT model as applied to SAD, and they areintroduced to cognitive restructuring techniques. Specifically, patients practice identifyingnegative cognitions (automatic thoughts),observing the co-variation between anxiousmood and automatic thoughts, examining theerrors of logic, and formulating rationalalternatives to their automatic thoughts. Inthe remaining 10 sessions of acute treatment,patients confront increasingly difficult fearedsituations (simulated in the therapy group)while applying cognitive restructuring techniques. Behavioral experiments are utilized toconfront specific reactions to exposure experiences. When this process is complete, thepatient and group agree on assignments for

194HofmannCOGNITIVE BEHAVIOUR THERAPYexposure to similar real-life situations duringthe week. Patients complete self-administeredcognitive restructuring exercises before andafter each behavioral homework assignment.Heimberg’s treatment protocol is the mostwidely accepted and disseminated approachto treating SAD. Similar treatment protocolshave been developed and tested by others (e.g.Davidson et al., 2004; Lucock & Salkovskis,1988; Mattick & Peters, 1988; Mersch,Emmelkamp, Bögels, & van der Sleen, 1989).For the remaining discussion, I will refer tothis approach as the ‘‘conventional CBTmodel of SAD’’.change on the Social Phobic DisordersSeverity Change Form (Liebowitz et al.,1992). Patients rated as markedly or moderately improved were classified as responders.Using a stricter improvement criterion,Mattick and Peters (1988) found that only38% of individuals with SAD who completeda treatment very similar to Heimberg’s protocol achieved high end-state functioning. Thecontrolled effect size estimate comparing CBTand educational supportive therapy at posttest based on the Liebowitz Social AnxietyScale (LSAS; Liebowitz, 1987) was in thesmall-to-medium range (see Table 1). Similareffect sizes were found in an earlier study(Heimberg, Dodge, Hope, Kennedy, Zollo, &Becker, 1990).Another example of a conventional CBTmodel is comprehensive cognitive behavioraltherapy (CCBT; Foa, 1994). This treatmentprotocol was included as a treatment condition in a recently published clinical trial(Davidson et al., 2004). The treatment protocol is derived in part from CBGT (Heimberg& Becker, 2002) and combines exposuretechniques, Beckian cognitive restructuringtherapy, and social skills training. The intervention differs from CBGT primarily inthat it includes specific social skills trainingin addition to the conventional cognitiverestructuring exercises and exposure tasks.Furthermore, the roleplays are shorter and thetreatment is 2 sessions longer than CBGT.The study by Davidson et al. (2004) suggestsThe conventional CBT modelAlthough the conventional CBT model forSAD has stimulated a great amount ofresearch, the treatment strategies have shownonly modest effects. For example, in a largescale study on the efficacy of CBGT, 133patients with SAD were randomly assigned tophenelzine (Nardil) a monoamine oxidaseinhibitor (MAOI) commonly used to treatSAD, educational support group therapy, apill placebo, or CBGT (Heimberg al., 1998).After 12 weeks, both the phenelzine (65%) andthe CBGT conditions (58%) had higherproportions of responders than pill placebo(33%) or educational support group therapy(27%), which served as a psychotherapyplacebo condition. The criterion for treatmentresponse was based on a 7-point rating ofTable 1. Summary of randomized-controlled studies testing conventional cognitive behavioral therapy forsocial anxiety disorder (SAD).StudyDavidson et al.(2004)Heimberg et al.(1990)Heimberg et al.(1998)Mattick & Peters(1988)Comparison Sample size Number ofgroupof CBTsessionsDuration ofsession(hours)Pill ortiveGuided exposure25122.533122.52562MeasuresControlledeffect size d1BSPSCGI-SFNESADSLSAS (fear)LSAS ef Social Phobia Scale (Davidson, Miner, deVeaugh Geiss, Tupler, Colket, & Potts, 1997); CGIS5Clinical Global Impression Scale, Severity (Guy, 1976); FNE5Fear of Negative Evaluation Scale (FNE;Watson & Friend, 1969); LSAS5Liebowitz Social Anxiety Scale (Liebowitz, 1987); SADS5Social Anxietyand Distress Scale (SADS; Watson & Friend, 1969).1The controlled effect size d was calculated according to the formula: d5(mean of comparison group at post-test– mean of CBT group at post-test)/pooled standard deviation.

VOL 36, NO 4, 2007that Foa’s treatment shows efficacy rates thatare similar to CBGT. Specifically, the studyrandomized 295 patients with generalizedSAD to 1 of 5 groups: (i) fluoxetine, (ii)CCBT, (iii) placebo, (iv) CCBT combinedwith fluoxetine, or (v) CCBT combined withplacebo. The results showed that all activetreatments were superior to placebo, and thecombined treatment was not superior tothe other treatments. The response rates inthe intention-to-treat sample (using theClinical Global Impressions scale) were50.9% (fluoxetine), 51.7% (CCBT), 54.2%(CCBT/fluoxetine), 50.8% (CCBT/placebo)and 31.7% (placebo). These findings arecomparable to other clinical trials, andsuggest that many participants remain symptomatic after conventional CBT. Davidsonand colleagues (2004), therefore, wonderedwhether ‘‘changes in the delivery of CBTwould improve the results’’ (p. 1012). Table 1depicts a summary of the trials that testedtreatments based on the conventional CBTmodel against credible placebo treatments.Disorder-specific CBT modelsPreliminary evidence in support of the notionthat disorder-specific intervention strategiescould lead to improved outcomes comes froma recent study by Clark and colleagues (2003).The treatment used in this trial is based on theClark and Wells’ (1995) model of SAD andfocuses on modifying safety behaviors andself-focused attention, in addition to theconventional CBT strategies. Clark andWells (1995) discuss at least four psychopathological processes that prevent individuals with SAD from disconfirming theirmaladaptive beliefs. First, when individualswith SAD enter a social situation they shifttheir attention to detailed monitoring andobservations of themselves. This attentionalshift produces an enhanced awareness offeared anxiety responses, interferes with processing the situation and other people’sbehavior, and produces interoceptive information that is used to construct a negative selfimpression. Secondly, individuals with SADengage in a variety of safety behaviors toreduce the risk of rejection. These behaviorsprevent them from critically evaluating theirfeared outcomes (e.g. shaking uncontrollably)and catastrophic beliefs. Thirdly, Clark andMaintaining factors of social anxiety disorder195Wells assume that individuals with SAD showan anxiety-induced performance deficit andoverestimate how negatively other peopleevaluate their performance. Fourthly, themodel suggests that prior to and after a socialevent, individuals with SAD think about thesituation in detail and primarily focus on pastfailures, negative images of themselves in thesituation, and other predictions of poor performance and rejection. The model furtherassumes that these anxious feelings and negativeself-perceptions are strongly encoded in memory because they are processed in such detail.Based on this model, Clark and colleagues(Clark et al., 2003) developed an individualtreatment approach consisting of 16 sessions.An abbreviated version of this protocol wasdeveloped earlier by Wells and Papageorgiou(2001). The Clark et al. (2003) trial randomlyassigned 60 patients with generalized SAD toone of 3 conditions: (i) cognitive therapyalone; (ii) fluoxetine combined with selfexposure; and (iii) fluoxetine combined witha pill placebo. Treatment efficacy was measured by calculating a composite score thatwas based on 6 frequently used self-reportmeasures of SAD and a rating based on astructured clinical interview. The results atpost-treatment and 12-month follow-upassessments showed that cognitive therapywas superior to the other 2 conditions, whichdid not differ from one another. The resultsshowed that the uncontrolled effect size of theseverity rating based on the clinical interviewwas 1.41 (pre-test to post-test) and 1.43 (pretest to 12-month follow-up) in the cognitivetherapy group. Even stronger effects werefound for the composite score, which wasassociated with an uncontrolled pre-posteffect size of 2.14.The trial by Clark et al. (2003), however, alsoshowed a number of notable weaknesses. First,the study did not include a method to assessresponder status and most of the results werebased on self-report instruments. Secondly,another recently published study by Stangieret al. (2003) reported a considerably smalleruncontrolled pre-post effect size after administering Clark’s protocol (0.77) and an evensmaller effect size when administering thistreatment in a group format (0.60). Nevertheless, the Clark et al. (2003) trial suggests thatit is possible to improve the treatment effects bytargeting additional cognitive variables that

196HofmannCOGNITIVE BEHAVIOUR THERAPYhave not been systematically addressed inprevious CBT protocols for SAD.The following will describe a comprehensivemaintenance model of SAD that is built uponrecent laboratory findings and results fromclinical trials. This model shows a number ofsimilarities to the cognitive model by Clarkand colleagues (Clark & Wells, 1995; Clark etal., 2003) but also includes a number ofsignificant differences and unique features.Some of its unique features are based on therecent acceptance-based literature, action theory, emotion theories, and studies on selfperception. The goal is to develop a comprehensive, disorder-specific maintenance modelthat is based on a broad psychopathology andtreatment literature.A comprehensive and disorderspecific CBT model for SADBased on the existing literature on themaintaining factors of SAD, a theoreticalmodel was generated (Figure 1). According tothis model, individuals with SAD are apprehensive in social situations in part becausethey perceive the social standard (i.e. expectations and social goals) as being high. Theydesire to make a particular impression onothers, but doubt that they will be able to doso (Leary, 2001), partly because they areunable to define goals and select specificachievable behavioral strategies to reach thesegoals (Hiemisch, Ehlers, & Westermann,2002). This leads to a further increase insocial apprehension and increased self-focusedattention (Clark & McManus, 2002; Heinrichs& Hofmann, 2001; Hirsch & Clark, 2004;Woody, 1996), which triggers a number ofadditional cognitive processes. Specifically,vulnerable individuals exaggerate the probability of a negative outcome of a socialsituation and overestimate the potential socialcosts (Foa, Franklin, Perry, & Herbert, 1996;Hofmann, 2004). This is consistent with themodel by Clark and Wells (1995), whichHigh perceived social standardsand poorly defined social goalsSocialapprehensionPost-event ruminationHeightenedself-focused attentionAvoidance and safetybehaviorsAnticipation ofsocial mishapNegative selfperceptionHigh estimatedsocial costLow perceivedemotional controlPerceived poorsocial skillsFigure. 1. Psychological factors that m