Running head: IDENTIFICATION OF FEARFUL FACES IN SOCIAL PHOBIA
Garner, M., Baldwin, D.S., Bradley, B.P. and Mogg, K. (2009) Impaired identification of fearful
faces in Generalised Social Phobia. Journal of Affective Disorders, 115, 460-465.
(doi:10.1016/j.jad.2008.10.020)
Impaired identification of fearful faces in Generalised Social Phobia
Matthew Garner*, David S. Baldwin, Brendan P. Bradley & Karin Mogg
University of Southampton
Corresponding author: Matthew Garner, m.j.garner@soton.ac.uk,
School of Psychology and Division of Clinical Neurosciences, School of Medicine,
University of Southampton, SO17 1BJ. Tel: +44(0)2380595926. Fax: +44(0)23 8059 4597
2
Abstract
Background. Cognitive models and interventions for anxiety assume that socially anxious
individuals interpret ambiguous social information in a threatening manner. However,
experimental evidence for this hypothesised cognitive bias is mixed. The present study is novel in
using a signal detection approach to clarify whether Generalised Social Phobia (GSP) is
associated with biased identification of emotionally ambiguous facial expressions.
Methods. 16 patients with GSP and 17 non-anxious volunteers classified ambiguous
emotional facial expressions, with each face reflecting a blend of two emotions: angry-happy,
fearful-happy and fearful-angry. Discrimination accuracy and response criterion were assessed.
Results. Patients with GSP showed significantly poorer discrimination of ambiguous
emotional facial expressions that contained an element of fear (i.e., fearful-happy and fearfulangry expressions), compared to non-anxious controls. The groups did not significantly differ in
discrimination of faces which lacked fear content (i.e., angry-happy blend), or on measures of
response criterion.
Limitations. Small sample size, coexisting depressive symptoms.
Conclusions. Findings indicate a selective impairment in fear identification in GSP.
Results are discussed with reference to neurocognitive models of anxiety, and research on
serotonergic modulation of emotional face processing.
3
Introduction
Social phobia is characterised by marked and persistent fear of one or more social or performance
situations involving possible scrutiny by others (American Psychiatric Association, 1994).
Effective psychological treatments for social phobia (e.g. cognitive-behaviour therapy, CBT;
Norton & Price, 2007) aim to correct maladaptive biases in information processing emphasised
by cognitive models of anxiety (e.g. Clark & McManus, 2002; Rapee & Heimberg, 1997)
According to these models social anxiety is associated with biases to preferentially attend towards
social threat cues and interpret ambiguous social information in a threatening manner (e.g.
ambiguous expressions as hostile or angry).
However, cognitive models also predict that under certain conditions a lower threshold
for appraising and orienting to threat can be followed by avoidance (if subjective threat value is
high; Mogg & Bradley, 2004) which in turn may impair (rather than enhance) appraisal of threatrelated information. Furthermore, greater self-focus (attention to internal cues and negative selfimagery) in social phobia may further reduce accurate processing of social cues (Clark &
McManus, 2002). It is therefore theoretically and clinically important to clarify whether socially
anxious individuals show a systematic bias in the interpretation of ambiguous social information.
Evidence to date is mixed. Richards et al. (2002) observed enhanced identification of
fearful (but not angry) faces in high non-clinical social anxiety, whereas Joorman and Gotlib
(2007) found enhanced identification of angry (but not fearful) faces in clinical social phobia.
Other studies have reported evidence of impaired identification of happy faces in adults and
children with social anxiety (Silvia et al. 2006; Simonian et al. 2001) while a recent study
reported no difference between patients with social anxiety and controls in their identification of
expressions (Philippot & Douillez, 2005).
One possible explanation for these inconsistent findings is that studies have not
systematically examined whether biased interpretation of ambiguous expressions primarily
reflects a bias in the ability to discriminate emotional expressions, or in response criterion, i.e.
tendency to label/resist labelling particular emotions. Winton et al. (1995) observed a lower
response criterion for reporting negative expressions in non-clinical social anxiety when
classifying briefly presented, unambiguous negative and neutral faces. However, no study has
examined biases in discrimination accuracy and response criterion for emotional faces in clinical
social phobia.
Our aim was to investigate the identification of angry, fearful and happy facial
expressions in patients with Generalised Social Phobia (GSP) compared with healthy non-anxious
controls. Participants classified ambiguous emotional facial expressions; each blended from two
emotions: angry-happy, happy-fearful and fearful-angry. Discrimination accuracy and response
criterion were assessed.
Cognitive models make differing predictions. 1. Discrimination accuracy: GSP is
characterised by (i) enhanced perceptual sensitivity when attending to social cues, which would
facilitate discrimination of cues that signal negative evaluation, e.g. anger (Rapee & Heimberg,
1997; Clark & McManus, 2002), or (ii) avoidance of social threat cues, which would impair
discrimination of emotional expressions. 2. Response criterion: If patients with GSP overestimate the threat value of ambiguous expressions, they would show a bias to misclassify them
as negative (i.e. lower response criterion).
4
Method
Participants
Individuals with GSP were recruited from new outpatients attending the Mood Disorders
Service in Southampton and respondents to a press release. Each underwent a structured
diagnostic interview based on DSM-IV criteria (MINI; Sheehan et al. 1998). Sixteen individuals
who met DSM-IV diagnostic criteria (American Psychiatric Association, 1994) for GSP
(confirmed by a consultant psychiatrist), in the absence of psychosis, current substance abuse,
pharmacological treatment or history of previous treatment with CBT, completed the study (5
men, 11 women, mean age 43.1 years).
The non-clinical group was matched closely for age, gender and education, and had no
known history of psychiatric disorder. Two had unusually high scores (> 3 SDs above nonclinical group mean) on the Liebowitz Social Anxiety Scale (Liebowitz, 1987), so were excluded.
Data are reported for 17 control participants (9 men, 8 women, mean age 39.9 years).
Materials
Prototypical angry, happy and fearful expressions of four female and four male models
were selected from the NimStim Face Set (http://www.macbrain.org/faces/index.htm; models: 1f,
3f, 7f, 8f, 21m, 23m, 27m, 34m) and used to create three emotion blends: Angry-Happy, HappyFearful and Fearful-Angry. For each model and emotion blend, two prototype images were
combined using Gryphon-Software (1994) to create two morphed faces with a 60:40 blend; e.g.
the angry and happy face of each model was blended to create one predominantly angry
(60%angry-40% happy) and one predominantly happy (40%angry-60%happy) face. Greyscale
stimuli measuring 90mm x 120mm were presented using MELv2 (Schneider, 1995) and viewed
at 57cm.
Procedure
Stimuli from each emotion blend (i.e., fearful-happy, fearful-angry, angry-happy) were
presented in separate blocks of trials (order counterbalanced across participants). Participants
classified each face using one of six response options (e.g. fearful-happy block: very fearful,
moderately fearful, slightly fearful, slightly happy, moderately happy, very happy). A fixation
cross was presented for 500 ms, followed by a face until response (inter-trial-interval=1000ms).
Faces appeared once in each block in random order.
To assess whether the groups differed in their ability to identify non-ambiguous
expressions, participants rated the 16 prototype images used to create the morphs at the start of
each block. Before the rating task, participants completed a computer task which used different
stimuli and afterwards completed the following questionnaires: Fear of Negative Evaluation and
Social Avoidance and Distress Scales (Watson & Friend, 1969), State-Trait Anxiety Inventory
(Spielberger et al. 1983), Beck Depression Inventory-II (Beck et al. 1996) and short-form Social
Desirability Scale (Strahan & Gerbasi, 1972).
Analyses
Data were transformed into measures of accuracy and response bias according to the twohigh-threshold (2HT) model (Corwin, 1994). Rating data from the six response options were
collapsed into two categories (e.g. very, moderately and slightly fearful collapsed into fearful;
very, moderately and slightly happy collapsed into happy), as not all six response options were
used by all participants. For each emotion blend, discrimination performance was computed from
Hits (e.g. number of 60%fearful-40%happy faces correctly classified as fearful) and False Alarms
(e.g. number of 60%happy-40%fearful faces incorrectly classified as fearful):
Pr=(Hits+0.5)/(number of targets+1)–(false alarms+0.5)/(number of distractors+1). Response bias
scores for each emotion blend, reflected participants’ tendency to select one or other of the two
available response options: Br=[(false alarms + 0.5)/(number of distractors+1 )]/(1-Pr) .
Pr scores were entered into a 2x3 mixed-design analysis of variance (ANOVA), with
group (GSP; non-anxious) and emotion blend (fearful-happy; fearful-angry; angry-happy) as
independent variables. Response bias (Br) scores were analyzed using independent t-tests as
5
(unlike Pr scores) two Br scores can be calculated for each blend, (e.g. fear-happy) which have
the same absolute value, but opposite signs Thus, results of a 2 x 3 ANOVA would vary
depending on which of the two Br scores were entered for each blend.
6
Results
Group Characteristics
The GSP group had significantly higher scores on all measures of social anxiety and
mood compared to non-anxious participants (see Table 1). Groups did not differ in age, education
or gender (χ2 = 1.59, N = 33, ns).
Classification of Ambiguous Expressions
The ANOVA of accuracy (Pr) scores revealed main effects of emotion blend,
F(2,62)=5.72, p<.01, ηp2=0.156, and group, F(1,31)=3.32, p=.08, ηp2=0.094, which were
subsumed under a significant group x emotion interaction , F(2,62)=3.15, p=.05, ηp2=0.092. The
pattern of means (Figure 1) indicated that, compared with non-anxious participants, GSP patients
showed poorer discrimination of faces expressing fear (i.e. fearful-happy and fearful-angry
blends), but not of faces which lacked fear content (angry-happy). This was confirmed with posthoc Helmert contrasts: a 2x2 ANOVA comparing the groups on their discrimination of faces
expressing fear revealed a significant main effect of group, F(1,31)=5.83, p=.02, ηp2=0.158,
which did not interact with blend type, F<1. In contrast, groups did not differ in their
discrimination of non-fear (angry-happy) faces, F<1.
Post hoc repeated-measures ANOVA within each group showed that accuracy
significantly differed across the three emotion blends in GSP participants, F(2,30)=5.91, p<.01,
with lower accuracy for fearful-angry than angry-happy expressions (p=.01)1; but not controls,
F(2,32)=2.22, p=.13.
The GSP and control groups did not differ on any measure of response bias, ts(31)<1.49,
ps>.15. Mean Br score (SD in brackets) for angry faces in angry-happy blends, happy faces in
happy-fearful blends and fearful faces in fearful-angry blends were 0.46(0.26), 0.50(0.20),
0.57(0.20) respectively in the GSP group, and 0.59(0.28), 0.44(0.13), 0.60(0.19), respectively, in
controls.
Ratings of prototype expressions.
All participants correctly identified all prototypes as angry, happy or fearful. Emotional
intensity ratings were entered into a 2 (group) x 3 (emotion) ANOVA. A significant main effect
of emotion [F(2,62)=6.00, p<.01] reflected greater intensity ratings for angry relative to happy
and fearful faces. Overall mean ratings for angry, happy and fearful prototypes were 2.50(0.30),
2.26(0.35), 2.31(0.33), respectively (scale: 1=slightly, 2=moderately, 3=very). All other results
were non-significant [group x emotion, F<1].
7
Discussion
Patients with GSP were significantly less accurate than controls in classifying faces
containing an element of fear (irrespective of whether blended with anger or happiness). Groups
did not differ in their ability to classify faces not containing fear (i.e. angry-happy expressions).
Individuals with GSP did not differ from controls on any measure of response criterion i.e. there
was no evidence of a systematic tendency to label an expression as fearful (or angry, or happy).
This study is the first to demonstrate a specific deficit in fear-identification in adult
patients with diagnosed GSP. It extends previous research that has reported a similar deficit in
identifying fearful voice cues in non-clinical socially anxious children (McClure & Nowicki,
2001). The present findings contrast with previous evidence of enhanced fear-identification in
non-clinical social anxiety (Richards et al. 2002; but see Joorman & Gotlib, 2006; Philippot &
Douillez, 2005). Notable methodological differences between studies may account for equivocal
findings, e.g. clinical versus non-clinical samples; unclear differentiation of accuracy and
response criterion effects.
It is helpful to consider whether our results reflect emotion-specific effects or some other
effect such as task difficulty. Is fear more difficult to classify than other expressions and are GSP
patients simply impaired on difficult tasks? Results do not appear to support this interpretation as
there was no evidence that controls found morphed fear faces generally harder to classify, nor that
groups differed in their intensity ratings of prototype expressions. Thus, the present results are
most readily interpreted in terms of a specific bias in GSP, namely, reduced sensitivity to fear
when processing ambiguous facial expressions.
This finding was not expected from cognitive models of anxiety which predict either
increased sensitivity to threat (angry faces) or avoidance and impaired processing of threat.
Instead, our results may be better explained within a neurocognitive framework. Social anxiety is
related to dysfunction of the primary fear network, which involves the amygdala (Amaral, 2002),
and is modulated by the serotonergic system (Furmark et al. 2002; 2004; Blanco et al. 2003).
Neuroimaging evidence implicates the amygdala in the evaluation of fearful faces (Adolphs et al.
2005); while manipulating brain serotonin modulates the perception of fearful faces. (Harmer et
al. 2003a; 2003b; Venn et al. 2006). Thus, the specific deficit in fear-identification in GSP could
reflect dysfunction in serotonergic modulation of amygdala activity. Although speculative, this
could be tested by examining the effect of serotonergic manipulation on amygdala activation in
GSP during a fear-identification task.
Imaging studies also indicate that the amygdala plays a key role in directing attention to
other peoples’ eyes which are important in conveying fear-relevant information (Adolphs et al.
2005; Whalen et al. 2004), while eye-tracking research indicates that individuals with GSP show
attentional avoidance (reduced fixations) of eyes (Horley et al. 2004). Thus avoidance of eyes in
GSP may account for impaired fear-identification. This deficit could be corrected through
attention training
The present study is not without limitations and may be extended in several ways i.e.
assessing the reliability of the present findings in a larger clinical sample, including comparison
clinical groups (e.g. patients with generalized anxiety disorder), examining other relevant
expressions (e.g. disgust) and interactions between trait and state anxiety. Evidence of a fearidentification deficit in adults with GSP may prove useful in guiding future research into the
cognitive and neurobiological mechanisms considered to underlie social phobia.
8
References
Adolphs, R., Gosselin., F, Buchanan, T.W., Tranel, D., Schyns P., Damasio, A.R., 2005. A
mechanism for impaired fear recognition after amygdala damage. Nature, 433, 68-72.
Amaral, D.G., 2002. The Primate Amygdala and the Neurobiology of Social Behavior:
Implications for Understanding Social Anxiety. Biol. Psychiatry, 51,11–17.
American Psychiatry Association 1994. DSM IV Diagnostic and Statistical - Manual, 4th Edition.
American Psychiatric Association: Washington, D.C.
Beck, A.T, Brown, G., Steer, R.A., 1996. Beck Depression Inventory II manual. San Antonio,
TX: The Psychological Corporation.
Blanco, C., Raza, M.S., Schneier, F.R., Liebowitz, M.R., 2003. The evidence-based
pharmacological treatment of social anxiety disorder. Int. J. Neuropsychopharm., 6, 427-442.
Clark, D.M., McManus, F., 2002. Information processing in social phobia. Biol. Psychiatry, 51,
92-100.
Corwin, J.W., 1994. On measuring discrimination and response bias: Unequal numbers of targets
and distractors and two classes of distractors. Neuropsychology, 8, 110-117.
Furmark, T., Tillfors, M., Marteinsdottir, I., Fischer, H., Pissiota, A., Langstrom, B., et al. 2002.
Common changes in cerebral blood flow in patients with social phobia treated with citalopram or
cognitive-behavioral therapy. Arc. Gen. Psychiatry, 59, 425-433.
Furmark, T., Tillfors, M., Garpenstrand, H., Marteinsdottir, I., Langstrom, B., Oreland, L., et al.
2004. Serotonin transporter polymorphism related to amygdala excitability and symptom severity
in patients with social phobia. Neurosci. Lett., 362, 189-192.
Harmer, C.J,, Bhagwagar, Z., Perrett, D.I., Vollm, B.A., Cowen, P.J., Goodwin, G.M., 2003.
Acute SSRI administration affects the processing of social cues in healthy volunteers.
Neuropsychopharmacology, 28, 148-152.
Harmer, C.J., Rogers, R.D., Tunbridge, E., Cowen, P.J., Goodwin, G.M., 2003. Tryptophan
depletion decreases the recognition of fear in female volunteers. Psychopharm., 167, 411-417.
Horley, K., Williams, L.M., Gonsalvez, C, Gordon, E., 2003. Social phobics do not see eye to
eye: A visual scanpath study of emotional expression processing J. Anxiety Disord., 17, 33-44.
Joormann, J., Gotlib, I.H., 2006. Is this happiness I see? Biases in the identification of emotional
facial expressions in depression and social phobia. J. Abnormal Psychol, 115, 705-714.
Liebowitz, M.R., 1987. Social Phobia. Modern Problems of Pharmacopsychiatry, 22,141-173
McClure, E.B., Nowicki, S. Jr., 2001. Associations between social anxiety and nonverbal
processing skill in preadolescent boys and girls. Journal of Nonverbal Behavior, 25, 3-19.
Mogg, K., Bradley, B.P., 2004. A cognitive-motivational perspective on the processing of threat
information and anxiety. Cognition, Emotion and Psychopathology ed J Yiend), pp. 68-85.
Cambridge University Press: Cambridge.
9
Norton, P.J., Price, E.C., 2007. A Meta-Analytic Review of Adult Cognitive-Behavioral
Treatment Outcome Across the Anxiety Disorders. J. Nerv. Ment. Dis. 195, 521-531.
Philippot, P., Douilliez, C., 2005. Social phobics do not misinterpret facial expression of emotion.
Behav. Res. and Ther., 43, 639-652.
Rapee, R.M., Heimberg, R.G., 1997. A cognitive-behavioral model of anxiety in social phobia.
Behav. Res. and Ther., 35, 741-756.
Richards, A., French, C.C., Calder, A.J., Webb, B., Fox, R., Young, A.W., 2002. Anxiety-Related
Bias in the Classification of Emotionally Ambiguous Facial Expressions. Emotion, 2, 273–287.
Schneider, W., 1995. MEL Professional User’s Guide. Pittsburgh: Psychology Software Tools.
Sheehan, D.V., Lecrubier, Y., Harnett-Sheehan, K., Amorim, P., Janavs, J., Weiller, E., Hergueta,
T., Baker, R., Dunbar, G., 1998. The Mini International Neuropsychiatric Interview (M.I.N.I.):
The Development and Validation of a Structured Diagnostic Psychiatric Interview. J. Clin
Psychiatry, 59, 22-33.
Silvia, P.J., Allan, W.D., Beuchamp, D.L., Maschauer, E.L., Workman, J.O., 2006. Biased
recognition of happy facial expressions in social anxiety. Journal of Social and Clinical
Psychology, 25, 585-602.
Simonian, S. J., Beidel, D.C., Turner, S.M., Berkes, J.L., Long, J.H., 2001. Recognition of facial
affect by children and adolescents diagnosed with social phobia. Child Psychiatry Hum. Dev., 32,
137-145.
Spielberger, C.D., Gorsuch, R.L., Lushene, R., Vagg, P.R., Jacobs, G.A., 1983. Manual for the
State-Trait Anxiety Inventory. Consulting Psychologists Press: Palo Alto, CA
Strahan, R., Gerbasi, K.C., 1972. Short homogenous versions of the Marlowe-Crowne Social
Desirability Scale, J. Clin. Psychol., 28, 191-193.
Venn, H. R., Watson, S., Gallagher, P., & Young, A. H. (2006). Facial expression perception: an
objective outcome measure for treatment studies in mood disorders? International Journal of
Neuropsychopharmacology, 9(2), 229-245.
Watson, D., Friend, R., 1969. Measurement of social-evaluative anxiety. J. Consult. Clin.
Psychol., 33, 448-457.
Whalen, P.J., Kagan, J., Cook, R.G., Davis, F.C., Kim, H., Polis, S., McLaren, D.G., Somerville,
L.H., McLean, A.A., Maxwell, J.S., Johnstone, T., 2004. Human amygdala responsivity to
masked fearful eye whites. Science, 306, 2061-2061.
Winton, E.C., Clark, D.M., Edelmann, R.J., 1995. Social anxiety, fear of negative evaluation and
the detection of negative emotion in others. Behav. Res. Ther., 33, 193-196.
10
Footnote
1
The effect of depression on Pr scores in the GSP group was examined in a 2 x
3 ANOVA with two independent variables: presence of major depressive
disorder (i.e., GSP without MDD, n=9 vs. GSP + MDD, n=7) and emotion
blend. There were no significant results. Furthermore, there were no significant
correlations between BDI-II depression scores and Pr or Br scores in either
group.
Figure 1. Accuracy (Pr) scores in GSP and non-anxious control (NAC) groups.
11
12
Table 1. Group Characteristics
GSP
Non-anxious
patients
participants
M
SD
M
SD
t(31)
p
LSAS
85.50
19.63
24.29
8.42
11.77
< .001
FNE
25.00
4.163
10.00
5.10
9.22
< .001
SADS
22.13
4.717
2.24
1.95
16.00
< .001
STAI-state
46.25
12.96
28.35
5.47
5.23
< .001
STAI-trait
58.31
12.77
34.47
6.21
6.89
< .001
BDI-II
19.88
16.94
6.41
5.89
3.09
< .01
SDS
4.19
2.14
4.82
2.16
0.85
.40
Age
43.13
12.45
39.94
11.17
0.77
.45
Years in education
13.94
2.59
13.59
2.69
0.38
.71
Gender m:f
5:11
9:8
Note. LSAS = Liebowitz Social Anxiety Scale (self-report); FNE = Fear of Negative
Evaluation Scale; SADS = Social Avoidance and Distress Scale; STAI = State-Trait
Anxiety Inventory; BDI-II = Beck Depression Inventory version II; SDS = Social
Desirability Scale (short form).