Conditioned Subjective Responses to Socially Relevant Stimuli in Social Anxiety Disorder and Subclinical Social Anxiety

Conditioned Subjective Responses to SociallyRelevant Stimuli in Social Anxiety Disorderand Subclinical Social Anxiety

Daniella Tinoco-González,1 Miquel Angel Fullana,1,2,3* David Torrents-Rodas,1Albert Bonillo,4 Bram Vervliet,5 Guillem Pailhez,2 Magí Farré,6 Oscar Andión,7,8Víctor Perez1,2,9 and Rafael Torrubia11Department of Psychiatry and Forensic Medicine, Institute of Neurosciences, School of Medicine, UniversitatAutònoma de Barcelona, Bellaterra, Spain2Anxiety Unit, Institute of Neuropsychiatry and Addiction (INAD), Hospital del Mar, Parc de Salut MAR,Barcelona, Spain3Department of Psychological Medicine, King’s College Institute of Psychiatry, London, UK4Department of Psychobiology and Methodology of Health Sciences, School of Psychology, Universitat Autònoma deBarcelona, Bellaterra, Spain5University of Leuven, Leuven, Belgium6Human Pharmacology and Clinical Neurosciences Research Group, Neuroscience Research Program, Hospital delMar Research Institute (IMIM), Parc de Salut MAR, Barcelona, Spain7Department of Psychiatry, Vall d’Hebron University Hospital, Barcelona, Spain8Vall d’Hebron Research Institute (VHIR), Universitat Autònoma de Barcelona, Barcelona, Spain9Centro de Investigación Biomédica en Red de Salud Mental (CIBERSAM), Madrid, Spain

Although enhanced fear conditioning has been implicated in the origins of social anxiety disorder(SAD), laboratory evidence in support of this association is limited. Using a paradigm employingsocially relevant unconditioned stimuli, we conducted two separate studies to asses fear conditioningin individuals with SAD and non-clinical individuals with high social anxiety (subclinical socialanxiety [SSA]). They were compared with age-matched and gender-matched individuals with anotheranxiety disorder (panic disorder with agoraphobia) and healthy controls (Study 1) and with individualswith low social anxiety (Study 2). Contrary to our expectations, in both studies, self-report measures(ratings of anxiety, unpleasantness and arousal to the conditioned stimuli) of fear conditioning failedto discriminate between SAD or SSA and the other participant groups. Our results suggest thatenhanced fear conditioning does not play a major role in pathological social anxiety. Copyright ©2014 John Wiley & Sons, Ltd.

Key Practitioner Message• We used a social conditioning paradigm to study fear conditioning in clinical and subclinical

social anxiety.• We found no evidence of enhanced fear conditioning in social anxiety individuals.• Enhanced fear conditioning may not be a hallmark of pathological social anxiety.

Keywords: Fear Conditioning, Social Anxiety Disorder, Panic Disorder, Fear-Potentiated Startle,Anxiety Disorders

INTRODUCTION

Fear conditioning is a form of associative learning in whichan aversive unconditioned stimulus (US) is repeatedly

paired with a neutral conditioned stimulus (CS), resultingin a conditioned fear response. Fear conditioning has longbeen considered a central pathogenic pathway in anxietydisorders (Lissek et al., 2005; Mineka & Zinbarg, 2006).Social anxiety disorder (SAD) is a common anxiety

disorder characterized by a persistent irrational fear ofsocial/performance situations in which the person isexposed to unfamiliar people or to possible scrutiny by

*Correspondence to: Miquel A. Fullana, Anxiety Unit, Institute ofNeuropsychiatry and Addiction (INAD), Hospital del Mar, Parc deSalut MAR, Passeig Marítim 25/29, 08003 Barcelona, Spain.E-mail: [email protected]

Clinical Psychology and PsychotherapyClin. Psychol. Psychother. (2014)Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/cpp.1883

Copyright © 2014 John Wiley & Sons, Ltd.

others (American Psychiatric Association, 2000). Previousresearch suggests that fear conditioning may play asignificant role in SAD. For example, individuals withSAD often remember traumatic aversive experiences as aprecipitant of the disorder (Lissek et al., 2008). The fearconditioning model of SAD suggests that once associatedwith an aversive outcome, social situations may evokeanxiety that persists in the absence of the original aversiveexperience (Öhman & Mineka, 2001). For example, ahighly humiliating public speaking experience (US) couldbe associated with certain people or places (CS) andcontribute to the persistence of social anxiety beyondthe presence of the US in SAD individuals. Accordingto this model, SAD individuals could be characterizedby abnormalities in fear conditioning such as en-hanced conditionability (i.e., an increased facility toassociate a US with a CS) or increased fear generaliza-tion (i.e., transferring fear to other CSs that resemblethe originally feared CS).As regards the possible enhanced conditionability, at

least three laboratory studies have failed to show thisenhanced conditionability in SAD patients compared withhealthy controls (HC). For example, Schneider et al. (1999)used a differential aversive conditioning paradigm in 12generalized SAD individuals and 12 HC. Heart rate andsubjective ratings were used to measure fear conditioning.Both groups showed evidence of conditioning as reflectedby subjective measures but not heart rate, with patientsshowing enhanced fear conditioning compared with HC.Using similar procedures, Hermann, Ziegler, Birbaumer,and Flor (2002) assessed the role of fear conditioning in14 SAD individuals and 19 HC. Both groups showedevidence of conditioning as reflected in subjectivemeasures, skin conductance response (SCR) and fear-potentiated startle (FPS). However, there was no evidencefor an enhanced fear conditioning among SAD subjectsrelative to HC in any of these measures. Finally, Veitet al. (2002) compared four SAD individuals and sevenHC using SCR and subjective ratings to asses fearconditioning. Surprisingly, the authors found enhancedconditioning in HC but not in SAD individuals asmeasured by SCR. Conditioning effects but not between-group differences were found for subjective ratings.In sum, all three studies failed to show enhancedconditionability in SAD patients compared with HC,whether the conditioning response was SCR, FPS or heartrate. Although the use of different paradigms andmeasures preclude definitive conclusions, the results ofthese studies pose a major problem to the fear condition-ing model of SAD based on a ‘general’ enhancedconditionability of SAD individuals.However, in one study using an ecologically enhanced

social conditioning paradigm with socially relevantstimuli as the US, SAD individuals did show enhancedfear conditioning. In this study, Lissek et al. (2008)

compared individuals with SAD and HC in a differentialfear conditioning paradigm using negative insults withcritical faces as the US and neutral faces as the CS.Only individuals with SAD displayed evidence of fearconditioning to the CS.More evidence for the ‘enhanced conditionability’ of

SAD individuals comes from two recent neuroimagingstudies. Davis, Johnstone, Mazzulla, Oler, and Whalen(2010) assessed differences in amygdala activation toexplore the neural correlates of social conditioning in 47HC using neutral facial expressions as CSs and self-relevant sentences as USs. Results showed an increasedblood oxygen level dependent (BOLD) response in thelateral ventral region of the amygdala during facespredicting negative social outcomes, suggesting that socialanxiety is associated with an enhanced acquisition of CRs.Pejic, Hermann, Vaitl, and Stark (2011) investigated theneural basis of fear acquisition in 49 HC within a 2-dayfunctional magnetic resonance imaging (fMRI) differentialconditioning protocol. They found an enhanced activationin the left amygdala and trends in the right dorsal anteriorcingulate cortex and the left hippocampus while observingneutral facial expressions (CSs) paired with film clips ofcritical comments (USs), demonstrating that social anxietyis related to altered social threat learning. Overall, bothstudies suggest a relationship between social anxiety andconditionability to socially threatening stimuli.The aforementioned study by Lissek et al. (2008) also

highlighted the importance of using disorder-specific USin the study of fear conditioning in anxiety disorders. It isrelatively well established that the qualitative nature of theCR depends on the qualitative nature of the CS and, there-fore, that the closer and more similar the CS is to the US,the stronger the CR will be (Bouton, Mineka, & Barlow,2001). This ‘specificity hypothesis’ implies that usingsocially relevant stimuli as the US should elicit ‘higherconditioning’ in SAD compared with other anxiety patients(in the same manner, interoceptive cues should elicit higherconditioning in panic disorder versus other anxiety disor-ders; see De Cort, Griez, Buchler, & Schruers, 2012). How-ever, determining whether this enhanced conditionabilityto socially relevant US is specific to SAD in comparisonwithother anxiety disorders remains to be tested.A limitation of the use of clinical populations for study-

ing fear conditioning in patients with anxiety disorders isthe frequent co-morbidity among the different anxietydisorders and between anxiety disorders and other mentaldisorders such as depression (Merikangas & Swanson,2010). Previous research on specific anxiety disorderssuggests that the presence of additional anxiety disorders(e.g., Borkovec, Alcaine, & Behar, 2004) and/or co-morbiddepression (Melzig, Weike, Zimmermann, & Hamm, 2007)may have a significant impact on physiological andsubjective indices of fear (Melzig et al., 2007; McTeagueet al., 2009; McTeague & Lang, 2012). For example,

D. Tinoco-González et al.

Copyright © 2014 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. (2014)

McTeague and Lang (2012) observed that generalizedsocial phobia patients with co-morbid depressionpresented an attenuated startle response compared withpatients without depression.However, studying non-co-morbid clinical samples can

be difficult. In fact, all four studies to date assessing fearconditioning in SAD used samples with significant co-morbidity. For example, almost 40% of the SAD individ-uals in Lissek et al. (2008) had a current co-morbid anxietydisorder. The other three studies that have assessed fearconditioning in SAD also included patients with co-morbidity (Schneider et al., 1999; Hermann et al., 2002;Veit et al., 2002).One possibility for circumventing this limitation is the

use of subclinical samples (e.g., ‘at-risk’ individuals withsignificant social anxiety symptoms but who do not fulfilthe criteria for SAD or any other mental disorder). Thisseems justified given epidemiological data showing thatSAD exists along a continuum (Bögels et al., 2010; Stein,Torgrud, & Walker, 2000).To address these questions, we conducted two separate

studies using the paradigm developed by Lissek et al.(2008). In Study 1, we compared fear conditioning acrosstwo clinical groups (SAD and panic disorder withagoraphobia [PDA]) and HC. We expected SAD individ-uals to show greater fear conditioning than individualswith PDA and HC. To our knowledge, this is the firstattempt to replicate the study of Lissek et al. (differencesbetween SAD patients and HC) and the first test of thespecificity hypothesis (differences between SAD andPDA patients). In Study 2, we tested whether enhancedfear conditioning was also evident in individuals at-riskfor SAD. To this end, we compared social conditioningacross individuals with subclinical social anxiety (SSA)and HC. We expected individuals with SSA to showgreater levels of fear conditioning than the HC.

METHOD

Participants

Participants in Study 1 included 16 individuals with SAD,16 individuals with PDA and 16 age-matched and gender-matched HC with no current mental disorder. Of the 16participants with SAD, 12 met the criteria for generalizedSAD and four met the criteria for non-generalized SAD.1

Participants in Study 2 included 20 individuals with nocurrent mental disorder and with high (>50) or low

(<20) scores on the self-report version of the LiebowitzSocial Anxiety Scale (Bobes, 1999).All individuals were interviewed by a licensed clinical

psychologist not involved in the experimental phase usingthe Spanish version of the Mini International Neuropsy-chiatric Interview (MINI, Sheehan et al., 1998). The MINIhas shown to have good concordance with other diagnos-tic measures and good inter-rater and test–retest reliability(Sheehan et al., 1998; Lecrubier et al., 1997). In a previousstudy in a Spanish sample, the sensitivity of SAD diagno-sis was 100, and its specificity was 84.2 (Bobes, 1998).The exclusion criteria for all participants included the

following: (1) use of pharmacological medication (exceptfor the PDA group, see below) or presence of medicalpathology capable of interfering with the study objectives(e.g., neurological disorders or hearing/visual impair-ments); (2) use of illicit drugs; (3) pregnancy; and (4) notspeaking Spanish. Additional exclusion criteria for theSAD and PDA subjects included the following: (1) currentor past history of bipolar disorder, psychosis or delusionaldisorder; (2) current major depressive disorder, post-traumatic stress disorder (PTSD) or suicidal ideation; (3)alcohol or substance abuse or dependence (other thannicotine) during the last 3months; and (4) psychomotordelay during the last 2weeks (item A3c of the MINI).Participants in the PDA group could be taking antidepres-sants (selective serotonin reuptake inhibitors) providedthat they had been on a stable dose for at least 3monthsprior to the study. All participants were Caucasian.

Procedure

We used the paradigm developed by Lissek et al. (2008),with slight modification (adding a reminder and using ashorter extinction phase). The subjects participated in adifferential fear conditioning procedure in which 8-sneutral facial expressions from three female actors(blonde, brunette and redhead) served as the CS and werepaired with one of three types of a 3-s, 85-dB audiovisualUS: (1) insults and critical facial expressions (USneg); (2)comments and neutral facial expressions (USneu); and (3)compliments and positive facial expressions (USpos). Theconditioned stimuli paired with the negative, neutraland positive USs will be referred to as CSneg, CSneu andCSpos, respectively.The experiment consisted of four phases: preconditioning,

four presentations of each type of CS in the absence of theUS; conditioning, eight presentations of each type of CSimmediately paired with its respective US; reminder, fourpresentations of each type of CS immediately paired withits respective US; and extinction, four presentations of eachtype of CS in the absence of the US. The preconditioningand conditioning phases were identical to those ofLissek et al. (2008). The reminder phase followed a 5-min

1We operationalized ‘generalized SAD’ as having clinically signifi-cant anxiety (fear or avoidance score = 3) in at least four different so-cial situations, as assessed by the Liebowitz Social Anxiety Scale. Theother participants were classified in the ‘nongeneralized SAD’ group.

Conditioned Subjective Responses to Socially Relevant Stimuli


break during which participants filled out the question-naires. This phase was added so that participantscould experience that the experimental contingenciesremained the same after the break (see, e.g., Baeyenset al., 2005). The extinction phase was shorter than thatin Lissek et al. (2008) given our focus on conditioning.The stimuli were presented in a quasi-random order, withthe constraint of a maximum of two equal consecutivepresentations. The assignment of faces to a type of CSwas counterbalanced across subjects.During the experimental session, all the participants

completed the Spanish versions of the State-Trait AnxietyInventory (Spielberger, Gorsuch, & Lushene, 1982), theself-report version of the Liebowitz Social Anxiety Scale,2

the Fear of Negative Evaluation Scale (Villa, 1999) andthe Beck Depression Inventory-II (Sanz, García-Vera,Espinosa, Fortún, & Vázquez, 2005). In addition, thePDA patients completed the Spanish version of the self-report version of the Panic Disorder Severity Scale(Houck, Spiegel, Shear, & Rucci, 2002). The experimentalsession lasted approximately 90mins. Experiments wereconducted between 9AM and 7 PM.All the participants were paid €15 and, after the

experimental procedure was described in detail, wereasked to sign an informed consent form, which waspreviously approved by the corresponding institution’sEthical Research Committee.

Physiological Recordings and Subjective Ratings

ABIOPAC system (BIOPAC,Goleta, CA)Mod.MP150WSWpsychophysiological recording system was used for stimula-tion and recording. The startle probe consisted of a 40-ms,3.5-psi airpuff delivered through a polyethylene tube affixed1 cm from the participants’ foreheads. The startle blink reflexwas recorded electromyographically (EMG) using two 5-mmsilver surface electrodes placed under the left eye. Impedancelevels were maintained below 5KΩ. The EMG signals wereamplified using a 10–500Hz bandpass analog filter andweredigitalized at a rate of 2000Hz.The experimental phase began with nine startle probes.

During all four phases, the startle probes were delivered4 or 5 s after CS onset. An 18- to 25-s interprobe intervalwas maintained throughout the whole experiment.Ratings of anxiety (How anxious do these faces makeyou feel?), unpleasantness (How unpleasant do thesefaces make you feel?) and arousal (How aroused do thesefaces make you feel?) for all CSs were obtained fromparticipants on a 10-point Likert scale (from 1 to 10)following conditioning and extinction. Additionally, be-tween the conditioning and reminder phases, the

participants rated each type of US by reflecting on thelevel of anxiety (How anxious do these faces make youfeel?) and happiness (How happy do these faces makeyou feel?) elicited using a 10-point Likert scale.

Data Analysis

The startle EMG data were rectified and smoothed withina 20-ms moving window. The blink onset latency windowwas 20–100ms, and the peak magnitude was determinedwithin 150ms following response onset. The averageEMG baseline levels for the 50ms preceding the startleprobe were subtracted from the peak magnitude. TheEMG magnitudes were standardized using within-subjectT-score conversions to normalize the data. The condition-ing phase was divided into four blocks (six trials per blockconstituted by two CSneg, two CSpos and two CSneu), andthe preconditioning, reminder and extinction phases weredivided into two blocks each (six trials per blockconstituted by two CSneg, two CSpos and two CSneu). ForFPS measures, fear conditioning was operationalized asthe difference between the CSneg and the CSpos duringconditioning and fear extinction and as the differencebetween the CSneg and the CSpos in early (first block)versus late (last block) extinction.3

In both studies, the EMG data were analysed usinganalyses of variance (general linear model procedure). InStudy 1, we used a 3 × 2× 3× 4 design for the conditioningphases and a 3× 2× 3× 2 design for the preconditioning,reminder and extinction phases, where subject group(SAD, PDA and HC) and gender were the between-subject factors and type of CS (negative, neutral, positive)and time (blocks 1, 2, 3 and 4 for conditioning; blocks 1and 2 for preconditioning, reminder and extinction) werethe within-subject factors. In Study 2, we used a2 × 2× 3× 4 design for the conditioning phase and a2 × 2× 3× 2 design for the preconditioning, reminder andextinction phases. Subject group (SSA and HC) andgender were the between-subject factors and type of CS(negative, neutral, positive) and time (blocks 1, 2, 3 and4 for conditioning; blocks 1 and 2 for preconditioning,reminder and extinction) were the within-subject factors.In both studies, simple contrasts were calculated tospecify the main or interaction effects.Self-report reactions to the conditioned stimuli (anxiety,

unpleasantness and arousal) were also analysed using ageneral linear model procedure with a 3 × 2× 3 (Study 1)or 2 × 2 × 3 (Study 2) design. Subject group and genderwere the between-subject factors, and type of CS was thewithin-subject factor. The level of significance was p< 0.05(two tailed). Greenhouse–Geisser corrections were applied

2Data presented in Tables and correspond to the second administra-tion of the self-report version of the Liebowitz Social Anxiety Scale.

3Fear conditioning and fear extinction were also operationalized asthe difference between the CSneg and the CSneu, but the resultsremained unchanged.



when necessary. Eta squared values or Cohen’s d werecalculated as measures of effect size. Gender results arenot presented given that the main and interaction effectsof gender were not significant for either EMG or the self-report data in both experiments. Since the BeckDepression Inventory, State-Trait Anxiety Inventory—Trait and airpuff ratings were not found to interact withany of the dependent measures (data not shown), theseeffects are not reported below.

RESULTS

Study 1

The demographic and clinical characteristics of theparticipants in Study 1 are reported in Table 1.

Startle ElectromyographyPreconditioning. The main effects of type of CS, type of CSby subject group and type of CS by time interaction werenot significant (all Fs< 175.74, all ps> 0.30, all ηp2 < 0.08).

Conditioning. Although a type of CS by time interactionemerged for the whole sample (F(6, 234) = 2.28, p= 0.04,ηp2 = 0.06), follow-up analyses showed that the responses

to the CSneg and CSpos were not different in any blockduring conditioning (all ps> 0.54). Furthermore, the maineffect of type of CS and the interactions type of CS bysubject group and type of CS by time by subject groupwere not significant (all Fs< 1.59, all ps> 0.19, all ηp

2

0.08; Figure 1). As successful conditioning is required todetect the reminder and extinction effects, results fromthe reminder and extinction phases are not reported.

Table 1. Demographic and clinical characteristics of the participants in Study 1

Group

Social anxietydisorder (n= 16)

Panic disorder(n = 16)

Healthy controls(n = 16)

Variable Mean SD Mean SD Mean SD F df p†

Age 25.81 8.50 33.00 8.65 31.06 10.66 2.54 2, 47 NSLSAS total score (0–144) 76.56a 22.72 40.13b 23.29 17.88c 7.70 37.69 2, 47 <0.0001Anxiety/fear (0–72) 41.63a 10.35 20.75b 11.75 10.50c 5.41 43.98 2, 47 <0.0001Avoidance (0–72) 34.94a 13.16 19.38b 13.35 7.38c 3.72 25.11 2, 47 <0.0001FNES (0–30) 25.31a 4.92 18.31b 5.64 8.75c 4.58 43.06 2, 47 <0.0001BDI (0–63) 19.25a 9.57 16.56a 9.22 2.01c 2.51 22.41 2, 47 <0.0001STAI-Trait (0–60) 36.00a 10.62 29.79a 10.71 10.81c 5.94 29.91 2, 43 <0.0001PDSS-SR (0–28) — — 12.63 5.89 — —Airpuff rating (0–10) 7.87 4.13 6.25 1.34 6.88 .81 1.65 2, 47 NS

N % N % N % χ2 dfFemale gender 13 81.3 11 68.8 13 81.3 0.943 2 NSPsychiatric co-morbidityGAD 3 19 0 0 0 0Current medication (SSRIs) 0 0 10 62.5 0 0

LSAS, Liebowitz Social Anxiety Scale; FNES, Fear of Negative Evaluation Scale; BDI, Beck Depression Inventory; STAI, State-Trait Anxiety Inventory.PDSS-SR, Panic Disorder Severity Scale self-report; GAD, Generalized Anxiety Disorder; SSRIs, selective serotonin reuptake inhibitors.NS, not significant; SD, standard deviation; df, degrees of freedom. Means with different subscripts († § ¶) differ significantly at p< 0.05.†One-way analysis of variance (except for gender, which was assessed with a Chi square).

Figure 1. Average, standardized startle blink electromyography magnitudes during preconditioning and reminder phases (Study 1).Bars represent standard errors of the mean



Self-Report Reactions to the Conditioned StimulusPreconditioning. The main effects of type of CS and type ofCS by subject group interaction were not significant for allself-report measures (all Fs< 0.94, all ps> 0.40, allηp2 < 0.04).

Conditioning. We found evidence of conditioning at thesubjective level for the whole sample, as revealed by asignificant main effects of type of CS for the self-reportmeasures of anxiety (F(2, 84) = 8.82, p< 0.001, ηp2 = 0.17),unpleasantness (F(2, 71.35) = 12.89, p< 0.001, ηp

2 = 0.24)and arousal (F(2, 84) =15.29, p=< 0.001, ηp2 =0.27). Therewere linear increases fromCSpos to CSneu to CSneg for anxiety(F(1, 42) = 9.87, p = 0.003, ηp2 = 0.19) and unpleasantness(F(1, 42) = 14.55, p< 0.001, ηp2 = 0.26), and there was a qua-dratic trend for arousal (F(1, 42) = 28.50, p< 0.001, ηp

2 = 0.40;Figure 2). However, conditioning did not differ betweengroups, as indicated by the non-significant interactionsbetween type of CS and subject group for any of thesevariables (all Fs< 1.60, all ps> 0.18, all η

p

2 < 0.07). A posthoc test comparing the CSneg with CSpos scores for eachsubjective variable (anxiety, unpleasantness and arousal)confirmed the absence of differences in conditioning acrossgroups (all Fs< 1.19, all ps> 0.31, all ηp2 < 0.05).

Extinction. During extinction, and similar to conditioning,the main effects of type of CS were found for anxiety,unpleasantness and arousal (all Fs> 9.31, all ps< 0.001,all ηp2 > 0.18), with linear increases from CSpos to CSneu toCSneg for anxiety (F(1, 42) = 9.10, p= 0.004, ηp2 = 0.18) andunpleasantness (F(1, 42) = 18.68, p< 0.001, ηp2 > 0.31) anda quadratic trend for arousal (F(1, 42) = 24.97, p< 0.001,ηp2 = 0.37). The type of CS by subject group interactions

with anxiety and unpleasantness were not significant(both Fs< 1.65, ps> 0.17, ηp2 = 0.07), but significance wasapproached for the arousal ratings (F(4, 84) = 2.28,p= 0.07, ηp

2 = 0.10). Pairwise comparisons revealed thatthe arousal ratings for the CSpos were significantly higher

for SAD (t(47) = 2.44, p= 0.05, d= 0.35) and PDA patients(t(47) = 2.61, p= 0.04, d= 0.37) in comparison with HC.

Self-Report Reactions to the Unconditioned StimulusThe main effects of type of US were found for the self-

report measures of happiness (F(2, 63.09) = 50.67, p< 0.001,ηp2 = 0.55) and anxiety (F(2, 73.35) = 20.53, p< 0.001, ηp2 =0.33), with linear increases from CS

posto CSneu to CSneg

for the anxiety ratings (F(1, 42) = 23.84, p< 0.001, ηp2 = 0.36)and linear decreases from CSpos to CSneu to CSneg for thehappiness ratings (F(1, 42)> 66.79, p< 0.001, ηp

2 = 0.61).However, there were no differences in the happiness oranxiety ratings of the US between groups, as shown bythe non-significant type of US by group interactions (bothFs< 0.89, both ps> 0.47, all ηp

2 < 0.04).

Co-morbidity, Medication and Social Anxiety DisorderSubtypeWe repeated the above analyses excluding the three

participants from the SAD group who had co-morbidgeneralized anxiety disorder, the participants in the PDAgroup who were taking medication and the four SADparticipants who did not meet the criteria for generalizedSAD. Every patient group was excluded separately ineach of these analyses. None of them had any significanteffect on the results (data not shown).

Study 2

The demographic and psychometric characteristics of theparticipants in Study 2 are reported in Table 2.

Startle ElectromyographyPreconditioning. The main effects of type of CS, type of CSby subject group and type of CS by time interactions werenot significant (all Fs< 2.13, all ps> 0.13, all ηp2 < 0.06).

Figure 2. Likert scale scores for self-report measures of anxiety, unpleasantness and arousal during conditioning (Study 1). Barsrepresent standard errors of the mean



Conditioning. The main effects of type of CS and theinteractions between type of CS and subject group andbetween type of CS by time and subject group were notsignificant (all Fs< 1.58, all ps> 0.15, all ηp

2 < 0.04;Figure 3). As in Study 1, given the absence of successfulconditioning in our startle measures, results from thereminder and extinction phases are not reported.

Self-Report Reactions to the Conditioned StimulusPreconditioning. As in Study 1, the main effects of type ofCS and type of CS by subject group interaction were notsignificant for all self-report measures (all Fs< 1.34, all ps0.25, all ηp2 < 0.04).

Conditioning. As in Study 1, the main effects of type of CSwere found for the self-report measures of anxiety(F(2, 72) = 28.27, p< 0.001, ηp

2 = 0.44), unpleasantness(F(2, 72) = 29.78, p< 0.001, ηp

2 = 0.45) and arousal(F(2, 58.59) = 23.97, p=< 0.001, ηp

2 = 0.40), with linearincreases from CSpos to CSneu to CSneg for anxiety(F(1, 36) = 40.16, p< 0.001, ηp2 = 0.53) and unpleasantness(F(1, 36)> 49.29, p< 0.001, ηp

2 > 0.58) and a quadratic

trend for arousal (F(1, 36) = 33.40, p< 0.001, ηp2 = 0.48;

Figure 4). However, as in Study 1, the type of CS bysubject group interactions for anxiety, unpleasantnessand arousal were not significant (all Fs< 1.92, all ps> 0.16,all ηp2 < 0.05). Again, a post hoc test comparing the CSnegwith CSpos scores for each subjective variable (anxiety,unpleasantness and arousal) revealed a significant differ-ence only for the arousal ratings (F(1, 36) = 5.63, p< 0.02,ηp2 = 0.14). Pairwise comparisons indicated higher arousal

ratings for the CSneg versus CSpos in the SSA group(t(39) = 3.09, p= 0.01, d= 0.49) but not in the HC group(t(39) = 0.28, p= 1.00, d= 0.04).

Extinction. The main effects of type of CS were found forthe anxiety (F(2, 72) = 18.44, p< 0.001, ηp2 = 0.34), unpleas-antness (F(2, 72) = 20.36, p< 0.001, ηp2 = 0.36) and arousalratings (F(2, 72) = 13.43, p< 0.001, ηp2 = 0.27), with linearincreases from CSpos to CSneu to CSneg for anxiety(F(1, 36) = 19.75, p< 0.001, ηp

2 = 0.35) and unpleasant-ness (F(1, 36) = 27.10, p< 0.001, ηp

2 = 0.43) and a qua-dratic trend for arousal (F(1, 36) = 17.98, p< 0.001,ηp2 = 0.33). Additionally, the type of CS by subject group

interactions for anxiety, unpleasantness and arousal

Figure 3. Average, standardized startle blink electromyography magnitudes during preconditioning, conditioning and reminderphases (Study 2). Bars represent standard errors of the mean

Table 2. Demographic and psychometric characteristics of the participants in Study 2

Subclinical social anxiety(n = 20)

Healthy controls(n = 20)

Variable Mean SD Mean SD F df p†

Age 20.15 2.23 21.55 2.60 16.92 1, 39 NSLSAS total score (0–144) 67.80 13.49 17.15 6.61 227.25 1, 39 <0.0001Anxiety/fear (0–72) 35.70 7.71 8.85 3.75 196.41 1, 39 <0.0001Avoidance (0–72) 32.10 7.63 8.30 4.52 144.07 1, 39 <0.0001FNES (0–30) 20.25 6.18 7.95 5.12 46.87 1, 39 <0.0001BDI (0–63) 11.47 9.16 4.50 4.80 9.00 1, 39 0.005STAI-Trait (0–60) 23.63 10.28 10.65 4.62 24.60 1, 39 <0.0001Airpuff rating (0–10) 6.45 1.50 4.65 2.56 7.35 1, 39 0.010

N % N % χ2 dfFemale gender 15 75 15 75 0.00 1 NS

LSAS=Liebowitz Social Anxiety Scale. FNES =Fear of Negative Evaluation Scale. BDI =Beck Depression Inventory. STAI = State-Trait AnxietyInventory. NS, not significant. SD= standard deviation. df = degrees of freedom.†Two-tailed t-tests (except for gender, which was assessed with a Chi square).



were all not significant (all Fs< 0.84, all ps> 0.44, allηp2 < 0.02).

Self-Report Reactions to the Unconditioned StimulusThe main effects of type of US were found for the self-

report measures of happiness (F(2, 57.41) = 56.41, p< 0.001,ηp2 = 0.61) and anxiety (F(2, 72) = 23.33, p< 0.001, ηp

2 = 0.39),with linear increases from CSpos to CSneu to CSneg for theanxiety ratings (F(1, 36) = 34.02, p< 0.001, ηp

2 = 0.49) andlinear decreases from CSpos to CSneu to CSneg for thehappiness ratings (F(1, 36) = 75.12, p< 0.001, ηp2 = 0.68). Asin Study 1, there were no between-group differences in thehappiness or anxiety ratings for the US, as shown by thenon-significant type of US by group interactions (both Fs1.69, both ps> 0.19, all ηp2 < 0.05).

DISCUSSION

We employed a social conditioning paradigm usingsocially relevant US to study fear conditioning in individ-uals with social anxiety in two separate studies.An important limitation of the present study is that we

were not able to demonstrate fear learning as indexed byour physiological measure, the FPS. In human fearconditioning experiments, it is relatively common forconditioning responses to differ across response systems(Beckers, Krypotos, Boddez, Effting, & Kindt, 2013). Inthe present study, we used the FPS as a physiologicalindex of conditioning, and according to Lipp, Sheridan,and Siddle (1994), the startle reflex can be potentiated aslong as the eliciting stimuli is sufficiently aversive. Thismay suggest that the US employed here was notexperienced as aversive enough to elicit FPS. However,our anxiety ratings to the negative US were in fact higherthan those of Lissek et al. (2008), where conditioning wasevident for FPS data.

It is noteworthy that other three studies assessing fearconditioning in SAD (Schneider et al., 1999; Hermannet al., 2002; Veit et al., 2002) have not detected enhancedconditionability in patients compared with HC asmeasured by psychophysiological indexes. Althoughprevious research on anxiety disorders seems to supportthe idea that enhanced conditionability is a characteristicof PTSD (e.g., Orr et al., 2000), results regarding otheranxiety disorders are mixed. For example, at least twoprevious studies did not find evidence of enhancedconditionability in panic disorder (Michael, Blechert,Vriends, Margraf, & Wilhelm, 2007; Del-Ben et al., 2001)or generalized anxiety disorder (Pitman & Orr, 1986),but a recent study found enhanced conditioning in indi-viduals with specific phobias (Vriends, Michael, Schindler,& Margraf, 2011). In any case, methodological differencesbetween studies make it difficult to come to conclusions inthis area (Lissek et al., 2005), and more importantly, theabsence of fear conditioning in our physiological dataprevents its use in the interpretation of group differences.Therefore, for the rest of the discussion, we will focus onsubjective ratings.Contrary to our hypotheses, we found no evidence of

enhanced fear conditioning in socially anxious individ-uals (clinical or subclinical) in comparison with HC orwith another clinical group (PDA patients). The fact thatin our two studies, SAD (or SSA) individuals did notshow enhanced fear conditioning compared with HC asindexed by subjective ratings is consistent with the resultsof Lissek et al. (2008). In this study, conditioning effectsbut not group differences were shown for subjectiveratings, suggesting also that SAD is not characterized byan enhanced ‘conditionability’. In line with these results,most previous studies have shown that SAD individualswere not ‘more conditionable’ when subjective ratingsindexed fear conditioning (Hermann et al., 2002; Veitet al., 2002). Nevertheless, it should be taken into accountthat in our study, we measured subjective ratings onlyonce after each phase, which is not probably a very

Figure 4. Likert scale score for self-report measures of anxiety, unpleasantness and arousal during conditioning (Study 2). Barsrepresent standard errors of the mean



sensitive measure and may have led to not detectinggroup differences.The fact that our results did not demonstrate an

enhanced fear conditioning in any of the measures drivesus to think there may be other mechanisms, rather thanincreased fear acquisition, that could be ‘abnormal’ inSAD. These include other processes related to fearlearning, such as impaired extinction, as well as non-associative processes.Impaired fear extinction seems to characterize individ-

uals with certain anxiety disorders, including PTSD(Blechert, Michael, Vriends, Margraf, & Wilhelm, 2007;Orr et al., 2000; Peri, Ben-Shakhar, Orr, & Shalev, 2000)and panic disorder (Michael et al., 2007). Onepossibility is that SAD is characterized by a deficit infear extinction rather than in fear acquisition. To date,only one study has reported that SAD patients showimpaired (delayed) extinction (Hermann et al., 2002). Inour two studies, we did not find differences in fearextinction between socially anxious individuals (clinicalor subclinical) and HC neither in psychophysiologicalnor in subjective measures. As regards the FPS, the factthat we found no evidence of fear conditioning makesour results on extinction not amenable to interpretation.In any case, in the only study where SAD individualsshowed enhanced fear acquisition using the FPS (Lisseket al., 2008), no differences with controls in terms ofextinction emerged.Our findings on extinction are also in line with Lissek’s

results. In both studies, fear extinction was incompletefor all participants at the end of the extinction phase, asrevealed by higher anxiety and unpleasantness ratingsto the CSneg compared with the CSpos. Additionally, andalso in agreement with Lissek et al. (2008), no significantbetween-group differences (SAD or SSA versus HC)emerged, suggesting that extinction deficits are not typicalof socially anxious patients/individuals. Anyway, the factthat we found no evidence of conditioning in FPScomplicates the interpretation of our extinction data. Wemust also note that our extinction phase was relativelyshort and future applications of the paradigm employedhere should include a larger phase in order to exploreextinction results.Finally, accounts of SAD that are not based on fear

conditioning mechanisms exist. An important body ofresearch suggests that cognitive and attentional processesplay a major role in the origins or maintenance ofpathological social anxiety. For example, cognitive–behavioural models of SAD suggest that distortions inthe processing of social information lead to heightenedanxiety in social situations, which, in turn, helps tomaintain social anxiety (Rapee & Heimberg, 1997; Clark& Wells, 1995). Other theoretical models focus on theattentional bias for threat that seems to characterizeSAD patients (Mogg, Philippot, & Bradley, 2004). Overall,

the results of the present study suggest that fearconditioning is not enhanced in social anxiety and thatalternative mechanisms may play a role in understandingsocial anxiety. Anyway, studying these mechanisms indetail was beyond the scope of the present study.The present study has several limitations. First, our

sample sizes were relatively small, albeit similar toprevious studies in the field. Second, our participantsin the PDA group could be using selective serotoninreuptake inhibitors, but previous data suggest that thisclass of medication does not affect the FPS (Grillon,Chavis, Covington, & Pine, 2009). Finally, as notedabove, our paradigm was not specially suited to thestudy of extinction processes.

CONCLUSIONS

Social anxiety does not seem to be related to abnormal fearconditioning. Other associative or non-associative pro-cesses seem to play a greater role in explaining socialanxiety.

CONFLICT OF INTEREST

The authors declare that they have no conflicts of interestconcerning this article.

ACKNOWLEDGEMENTS

We thank Shmuel Lissek, Ph.D., for giving us the stimuliused in the present study, as well as for his general advicefor the implementation of the paradigm in our laboratory.

This research was financially supported by theMinisterio de Ciencia e Innovación, Spanish government(PS09/00307) and by the Comissionat per a Universitatsi Recerca, Generalitat de Catalunya (2009SGR51). D. T-G.and D. T-R. are recipients of a Ph.D. fellowship from theGeneralitat de Catalunya (FI2010).

REFERENCESAmerican Psychiatric Association. (2000). Diagnostic and

statistical manual of mental disorders. (4th edn, text revision(DSM-IV-TR)). Washington, DC: American Psychiatric Press.

Baeyens, F., Vansteenwegen, D., Beckers, T., Hermans, D.,Kerkhof, I., & De Ceulaer, A. (2005). Extinction and renewalof Pavlovian modulation in human sequential feature positivediscrimination learning. Learning and Memory, 12, 178–192.

Beckers, T., Krypotos, A. M., Boddez, Y., Effting, M., & Kindt, M.(2013). What’s wrong with fear conditioning? BiologicalPsychology, 92, 90–96.



Blechert, J., Michael, T., Vriends, N., Margraf, J., & Wilhelm, F. H.(2007). Fear conditioning in posttraumatic stress disorder:Evidence for delayed extinction of autonomic, experiential,and behavioural responses. Behaviour Research and Therapy, 45,2019–2033.

Bobes, J. (1998). A Spanish validation study of the MINIinternational neuropsychiatric interview. European Psychiatry,13, 198–199.

Bobes, J. (1999). Validación de las versiones en español de loscuestionarios Liebowitz Social Anxiety Scale, Social Anxietyand Distress Scale y Sheehan Disability Inventory para laevaluación de la fobia social. [Validation of the Spanish versionof the Liebowitz Social Anxiety Scale, Social Anxiety andDistress Scale and Sheehan Disability Inventory for the assess-ment of social phobia]. Medicina Clínica, 112, 530–538.

Bögels, S., Alden, L., Beidel, D. C., Clark, L. A., Pine, D. S.,Stein, M. B., & Voncken, M. (2010). Social anxiety disorder:Questions and answers for the DSM-V. Depression and Anxiety,27, 168–189.

Borkovec, T. D., Alcaine, O., & Behar, E. (2004). Avoidance theoryof worry and generalized anxiety disorder. In R. G. Heimberg,C. L. Turk, & D. S. Mennin (Eds.), Generalized anxietydisorder: Advances in research and practice (pp. 77–108).New York: Guilford Press.

Bouton, M.E., Mineka, S., & Barlow, D. H. (2001). A modernlearning theory perspective on the etiology of panic disorder.Psychological Review, 108, 4–32.

Clark, D. M., & Wells, A. (1995). A cognitive model of socialphobia. In R. G. Heimberg, M. R. Liebowitz, D. A. Hope, & F.R. Schneier (Eds), Social phobia: Diagnosis, assessment andtreatment (pp. 69–93). New York: Guilford Press.

Davis, F.C., Johnstone, T., Mazzulla, E.C., Oler, J.A., & Whalen, P.J. (2010). Regional response differences across the humanamygdaloid complex during social conditioning. CerebralCortex, 20, 612–621.

De Cort, K., Griez, E., Buchler, M., & Schruers, K. (2012). The roleof “interoceptive” fear conditioning in the development ofpanic disorder. Behavior Therapy, 43, 203–215.

Del-Ben, C., Vilela, J., Hettema, L., Guimaraes, F., Graeff, F., &Zuardi, W. (2001). Do panic patients process unconditionedfear vs. conditioned anxiety differently than normal subjects?Psychiatry Research, 104, 227–237.

Grillon, C., Chavis, C., Covington, M.F., & Pine, D. S. (2009).Two-week treatment with the selective serotonin reuptakeinhibitor citalopram reduces contextual anxiety but not cuedfear in healthy volunteers: A fear-potentiated startle study.Neuropsychopharmacology, 34, 964–971.

Hermann, C., Ziegler, S., Birbaumer, N., & Flor, H. (2002).Psychophysiological and subjective indicators of aversivePavlovian conditioning in generalized social phobia. BiologicalPsychiatry, 52, 328–337.

Houck, P. R., Spiegel, D. A., Shear, M. K., & Rucci, P. (2002).Reliability of the self-report version of the panic disorderseverity scale. Depression and Anxiety, 15, 183–185.

Lecrubier, Y., Sheehan, D. V., Weiller, E., Amorim, P., Bonora, I.,Sheehan, K. H., Janavs, J., & Dunbar, G. C. (1997). The MiniInternational Neuropsychiatric Interview (MINI). A shortdiagnostic structured interview: reliability and validityaccording to the CIDI. European Psychiatry, 12, 224-231.

Lipp, O. V., Sheridan, J., & Siddle, D. A. (1994). Human blinkstartle during aversive and nonaversive Pavlovian conditioning.Journal of Experimental Psychology-Animal Behavior Processes, 20,380–389.

Lissek, S., Levenson, J., Biggs, A. L., Johnson, L. L., Ameli, R.,Pine, D. S., & Grillon, C. l. (2008). Elevated fear conditioningto socially relevant unconditioned stimuli in social anxietydisorder. American Journal of Psychiatry, 165, 124–132.

Lissek, S., Powers, A. S., McClure, E. B., Phelps, E. A.,Woldehawariat, G., Grillon, C., & Pine, D. S. (2005). Classicalfear conditioning in the anxiety disorders: A meta-analysis.Behaviour Research and Therapy, 43, 1391–1424.

McTeague, L. M., & Lang, P. J. (2012). The anxiety spectrum andthe reflex physiology of defense: From circumscribed fear tobroad distress. Depression and Anxiety, 29, 264–81.

McTeague, L. M., Lang, P. J., Laplante, M. C., Cuthbert, B. N.,Strauss, C. C., & Bradley, M. M. (2009). Fearful imagery insocial phobia: Generalization, comorbidity, and physiologicalreactivity. Biological Psychiatry, 65, 374–382.

Melzig, C. A., Weike, A. I., Zimmermann, J., & Hamm, A. O.(2007). Startle reflex modulation and autonomic respondingduring anxious apprehension in panic disorder patients.Psychophysiology, 44, 846–854.

Merikangas, K. R., & Swanson, S. A. (2010). Comorbidity in anxietydisorders. Current Topics in Behavioral Neurosciences, 2, 37–59.

Michael, T., Blechert, J., Vriends, N., Margraf, J., & Wilhelm, F. H.(2007). Fear conditioning in panic disorder: Enhanced resis-tance to extinction. Journal of Abnormal Psychology, 116, 612–617.

Mineka, S., & Zinbarg, R. (2006). A contemporary learning theoryperspective on the etiology of anxiety disorders: It’s not whatyou thought it was. American Psychologist, 61, 10–26.

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

Öhman, A., & Mineka, S. (2001). Fears, phobias, and prepared-ness: toward an evolved module of fear and fear learning.Psychological Review, 108, 483.

Orr, S. P., Metzger, L. J., Lasko, N. B., Macklin, M. L., Peri, T., &Pitman, R. K. (2000). De novo conditioning in trauma-exposed individuals with and without posttraumatic stressdisorder. Journal of Abnormal Psychology, 109, 290–298.

Pejic, T., Hermann, A., Vaitl, D., & Stark, R. (2011). Social anxietymodulates amygdala activation during social conditioning.Social cognitive and affective neuroscience, advance access.doi:10.1093/scan/nsr095.

Peri, T., Ben-Shakhar, G., Orr, S. P., & Shalev, A. Y. (2000). Psycho-physiologic assessment of aversive conditioning in post-traumatic stress disorder. Biological Psychology, 47, 512–519.

Pitman, R. K., & Orr, S. P. (1986). Test of the conditioning modelof neurosis: Differential aversive conditioning of angry andneutral facial expressions in anxiety disorder patients. Journalof Abnormal Psychology, 95, 208–213.

Rapee, R. M., & Heimberg, R. G. (1997). A cognitive-behavioralmodel of anxiety in social phobia. Behaviour Research andTherapy, 35, 741–756.

Sanz, J., García-Vera, M. P., Espinosa, R., Fortún, M., & Vázquez,C. (2005). Adaptación española del Inventario para laDepresión de Beck-II (BDI-II): 3. Propiedades psicométricasen pacientes con trastornos psicológicos. [Validation of theSpanish version of the Beck Depression Inventory (BDI-II): 3.Psychometric properties in patients with psychological disor-ders]. Clínica y Salud, 16, 121–142.

Schneider, F., Weiss, U., Kessler, C., Müller-Gärtner, H.W., Posse, J.,Salloum, J. B., Grod, W., Himmelmann, F., Gaebel, W., &Birbaumer, M. (1999). Subcortical correlates of differential classi-cal conditioning of aversive emotional reactions in social phobia.Biological Psychiatry, 45, 863–871.



Sheehan, D. V., Lecrubier, Y., Sheehan, K. H., Amorim, P., Janavs, J.,Weiller, E., Herqueta, T., Baker, R., & Dunbar, G. C. (1998). TheMini-InternationalNeuropsychiatric Interview (M.I.N.I.): The devel-opment and validation of a structured diagnostic psychiatric inter-view forDSM-IVand ICD-10. Journal of Clinical Psychiatry, 59, 22–33.

Spielberger, C. D., Gorsuch, R. R., & Lushene, R. E. (1982). STAI.Cuestionario de ansiedad Estado/Rasgo (1st ed.). [Validationof the Spanish version of the State Trait Anxiety Scale (STAI)].Madrid: Técnicos Especialistas Asociados (TEA).

Stein, M. B., Torgrud, L. J., & Walker, J. R. (2000). Social phobiasymptoms, subtypes and severity: Findings from a communitysurvey. Archives of General Psychiatry, 57, 1046–1052.

Veit, R., Flor, H., Erb, M., Hermann, C., Lotze, M., Grodd, W., &Birbaumer, N. (2002). Brain circuits involved in emotionallearning in antisocial behavior and social phobia in humans.Neuroscience Letters, 328, 233–236.

Villa, H. (1999). Validación de un instrumento de medida enfobia social: Miedo a la evaluación negativa. [Validation ofan assessment instrument in social phobia: Fear of NegativeEvaluation]. Unpublished thesis dissertation, UniversitatJaume I.

Vriends, N., Michael, T., Schindler, B., & Margraf, J. (2011).Associative learning in flying phobia. Journal of Behavior Therapyand Experimental Psychiatry, 43, 838–843.



Documents

Conditioned Subjective Responses to Socially Relevant Stimuli in Social Anxiety Disorder and Subclinical Social Anxiety