گسترش مخروطی نگاه در بیماران مبتلا به اختلال اضطراب اجتماعی و طبیعی شدن آن پس از CBT

Abstract Gaze plays a crucial role in social interactions. Social Anxiety Disorder (SAD), which is associated with severe impairment of social interactions, is thus likely to exhibit disturbances of gaze perception. We conducted two experiments with SAD-patients and healthy control participants using a virtual head whose gaze could be interactively manipulated. We determined the subjective area of mutual gaze, the so-called gaze cone, and measured it prior to and after a psychotherapeutic intervention (Exp. 1). Patients exhibited larger gaze cones than control subjects. Exp. 2 varied the emotional expression of the virtual head. These data were validated using a real person (professional actor) as stimulus. Excellent reliability indices were found for our gaze cone measure. After Cognitive Behavioral Therapy, group differences in gaze cone width had disappeared. Emotional expressions were observed to modulate the gaze cone's width. Especially an angry expression caused the gaze cone to widen, possibly mediated by increased arousal. Finally, wider gaze cones in SAD-patients could be demonstrated for virtual and for real human heads confirming the ecological validity of virtual heads. The findings are of relevance for a more fine-grained understanding of perceptual processes in patients with SAD.

Introduction: the cone of gaze Social anxiety disorder (SAD), also called social phobia, is characterized by strong anxiety of acting embarrassingly or showing shame-related symptoms induced by social interaction or performance situations (American Psychiatric Association, 2000). Not surprisingly, patients with SAD tend to avoid such situations altogether (Delong & Pollack, 2008). SAD severely impairs performance in every-day situations (Barrera & Norton, 2009; Chen & Drummond, 2008), it is known to be long lasting (Stangier, Clark, & Ehlers, 2006) and frequently accompanied by comorbid disorders, such as depressive disorders (Beesdo et al., 2007), anxiety disorders and alcohol and substance abuse (Delong & Pollack, 2008). In comparison with other anxiety disorders, SAD showed high persistence rates and high probability of relapse (Bruce et al., 2005). Epidemiological studies showed a wide range of prevalence rates, depending on sample characteristics and the diagnostic criteria that have been applied (Fehm, Beesdo, Jacobi, & Fiedler, 2008). In a German community sample, cumulative incidence rate by the age of 33 years was 11.6% (Beesdo et al., 2007), which is already close to the lifetime prevalence of 12.1% found in the national comorbidity survey (Kessler et al., 2005). Accordingly, SAD represents a frequent disorder with massive negative impact on patients' lives. We sought to gain deeper insights into its consequences by focusing on the cognitive-perceptual correlate of mutual gaze perception. A valid correlate that is easy to measure would be most valuable for diagnosis and treatment. Gaze research has identified some changes of looking behavior in patients suffering from SAD, mainly the avoidance of eye contact in social interaction. The latter is often considered to be an important indicator of social anxiety (Hofmann, Gerlach, Wender, & Roth, 1997). Eves and Marks (1991) demonstrated such gaze aversion in socially anxious participants who had to speak in front of an audience. Other studies found comparable gaze avoidance in social anxiety (Garner, Mogg, & Bradley, 2006; Roelofs et al., 2010), whereas some studies failed to demonstrate differences of gazing behavior between SAD-patients and healthy controls (Hofmann et al., 1997; Schneier, Kent, Star, & Hirsch, 2009). Clark and Wells' (1995) influential cognitive model of social phobia postulates that SAD-patients shift attention towards themselves, thereby monitoring their own bodily symptoms more closely, which is typically associated with negative self-related cognitions and heightened arousal (Zou, Hudson, & Rapee, 2007). This shift of attention in combination with the activation of negative thoughts and the mentioned changes in gaze behavior make it appear likely that SAD-patients also misperceive the area of mutual gaze. Our work focuses on mutual gaze perception of patients suffering from SAD rather than on their active gazing behavior. Gamer and Hecht (2007) developed a psychophysical paradigm to measure the width and direction of the area of mutual gaze as experienced by the person receiving the gaze (the gaze cone resp. cone of gaze). Subjects are confronted with a virtual head displayed on a monitor or a large screen. Their task is to interactively adjust the eyes of the virtual target head such that it seems to look straight in their eyes or such that the target head is gazing at the edge of the area that the subject would consider to establish mutual gaze. The edges define the subjective gaze cone. If the counterpart's gaze falls within this area the person feels looked at. This paradigm was then applied to SAD-patients ( Gamer, Hecht, Seipp, & Hiller, 2011) to measure their potentially altered gaze perception. The gaze cone was found to be widened in SAD-patients compared to healthy control participants when a second head also looking at the subject was present. Gamer et al. (2011) thereby demonstrated that SAD-patients have a broader criterion for mutual gaze than do controls. In the realm of SAD, the gaze cone width therefore most likely represents a psychophysical measure for the “amplified feeling of being looked at” ( Gamer et al., 2011; p. 757). We sought to test whether the gaze cone is generally widened in SAD-patients or whether the additional social pressure constituted by the second head is required. We varied the orientation of this second head and its gaze orientation twofold to explore whether its mere presence suffices to induce changes. We also asked whether the assessment of the gaze cone using a virtual head constitutes a reliable measure. And most importantly, we were interested to find out whether the observed abnormalities in the gaze cone in patients with SAD disappear or decrease with Cognitive Behavior Therapy (CBT). If so, a venue to use the gaze cone measure to monitor therapy success would open up. We conducted two experiments. The first one measured the gaze cone prior to and after a specific CBT for patients with SAD (SAD-Group). A comparison group of healthy participants (Control Group) was likewise measured prior to and after a waiting period of identical length as the duration of CBT for the SAD-patients. In a second experiment we varied the target stimuli. A virtual head was compared to a real person with professional acting experience. We thereby aimed at examining the ecological validity of our findings. Can equivalent differences in gaze-perception in SAD-patients be observed when facing a real onlooker as compared to the virtual head that was used in our previous studies? In addition, the second experiment varied the emotional expression of the target face whose eyes had to be adjusted. The misclassification of emotions in SAD-patients has been well documented. Coles, Heimberg, and Schofield (2008) for example found a tendency of participants with high levels of social anxiety to misinterpret facial expressions of schematic faces as threatening. They also tended to misclassify neutral facial expressions as showing anger (Bell et al., 2011). Foa, Gilboa-Schechtman, Amir, and Freshman (2000) demonstrated a better memory for emotional expressions in anxious compared to non-anxious participants, especially for negative emotions. Consistent with these findings, Gentili et al. (2008) found higher activation levels in brain areas of anxiety patients that process information about emotional expressions, and lower activation in areas processing other facial information. We therefore varied the emotional expression of the stimulus faces in Experiment 2 hypothesizing that highly activating emotional expressions, such as anger, might further widen the gaze cone.

Results We calculated repeated measures ANOVAs for the differences in the questionnaire values between both groups, and t – tests for the change of symptoms from pre- to post-therapy assessment. To judge reliability we calculated Cronbach's α as an indicator of internal consistency and correlated the Control-Groups' values of the gaze cone's width between the Pre- and Post-Test. Furthermore, we calculated a 3-way ANOVA on the width and direction of the gaze cone: Group (SAD, Control-Group) × Assessment (pre-therapy, post-therapy) × number of heads (1-head-condition vs. 2-head-condition), as well as separate ANOVAs for the two conditions (1-head-condition vs. 2-head-condition) on the width of the gaze cone and its direction for both the Pre- and Post-Test, including the head orientation as a factor (rotated left vs. straight vs. rotated right). Partial η2 is reported as an estimate of the effect size. When the sphericity assumption was violated, we used the Greenhouse-Geisser correction. Two outliers were excluded before analysis. One participant in the SAD-Group exhibited values more than two standard deviations above the mean width and was therefore excluded; one participant of the Control Group was excluded from the analysis of the gaze cone's direction for the same reason. To differentiate between merely statistically significant change and clinically relevant change, we calculated the clinical significance, first introduced by Jacobson et al. (1984). This indicated that a change of symptom values, as measured by the SPS and SIAS, would have been considered a significant improvement if the values of the Post-Test fell below the cut-off value of 11.5 in the SPS and below 20.8 in the SIAS. Questionnaire data The ANOVA revealed significantly higher levels of social anxiety (SPS, SIAS) and depression (BDI) in the SAD-Group both in Pre- and Post-Test (see Table 1). Table 1. Experiment 1: Depression and social anxiety scores for the Control Group (CG) and the SAD-Group (SAD) before (pre) and after (post) cognitive behavior therapy, respectively a waiting period (Means and Standard Deviations in Parentheses). Control-group SAD-group P Value and η2:CG vs. SAD Pre Post Pre Post Pre Post BDI 2.8 (3.7) 2.0 (2.6) 10.1 (5.5) 8.7 (7.4) p < .001 p < .001 η2 = .38 η2 = .30 SPS 4.6 (3.9) 4.4 (3.9) 30.0 (15.5) 22.3 (13.4) p < .001 p < .001 η2 = .61 η2 = .48 SIAS 9.4 (5.7) 8.9 (4.3) 40.9 (12.0) 31.8 (13.7) p < .001 p < .001 η2 = .76 η2 = .59 Table options T-tests on the change between Pre- and Post-Test revealed no significant changes of the (initially low) questionnaire values in the Control-Group (BDI: t(21) = 1.56, p = .13; SPS: t(21) = .28, p = .78; SIAS: t(21) = 1.01, p = .32). In the SAD-Group, a different pattern emerged: BDI scores did not change (t(16) = .71, p > .49), whereas the SAD questionnaire values decreased significantly (SPS: t(16) = 2.64, p = .018, d = .57; SIAS: t(16) = 3.376, p = .004, d = .63 (see Fig. 2)). That is, depression values remained relatively stable over time in the SAD-Group whereas the severity of the social anxiety symptoms decreased. However, the RCI indicated that the Post-Test SIAS values failed to reach clinical improvement level, when reliability and standard deviation of the questionnaire were taken into account. The SPS values of 4 participants exceeded the cut-off values. Thus, the amount of change in social anxiety symptoms was not as pronounced as we had hoped for. Levels of social interaction anxiety measured with the social interaction scale ... Fig. 2. Levels of social interaction anxiety measured with the social interaction scale (SIAS) and the social phobia scale (SPS) during the pre- and post- test for the SAD-Group and Controls. Error bars represent standard errors of the mean. Figure options Experimental data Reliability Cronbach's α as an indicator of internal consistency was high for the gaze cone's width in both groups for Pre-Test (α = .99) and Post-Test (α = .99) as well. We correlated the Control-Groups' values of the gaze cone's width of the Pre- and Post-Test, which brought forward highly significant correlations. The correlation of the total width (mean of all conditions and variations) in the one-head-condition was r = .706, p < .001, in the two-head-condition r = .745, p < .001. A t-test yielded no differences for the gaze cone's width between the Pre- and Post-Test (t(21) = .216, p = .83) in the Control-Group. In sum, we demonstrated excellent retest-reliability scores for measuring the gaze cone's width. Differences between SAD-group and control group Target-Head-Orientation showed a large attractor effect on the direction of the gaze cone when taking results of the Pre- and Post-Test together (F(2, 76) = 138.41, p < .001, η2 = .79). As we were not able to demonstrate any significant effects of any other variable (including the Group Factor) on the direction of the gaze cone and had not expected any to emerge in the first place, we concentrated on the results of the width of the gaze cone. The 3-way ANOVA on the width of the gaze cone yielded only a marginally significant medium sized main effect of the Group Factor (F(1, 38) = 3.67, p = .063, η2 = .088), indicating a trend toward a wider gaze cone in the SAD-compared to the Control-Group. No significant main effects were found for the assessment (F(1, 38) = .53, p = .471, η2 = .0148) or the presence of the second head (F(1, 38) = .13, p = .716, η2 = .004). All interactions failed to reach significance (all p-Values >.05). A 2-way ANOVA (Group × Number of Heads present) on the cone's width of the Pre-Test, however, revealed significantly wider gaze cones in the SAD-Group than in the Control-Group both when only the target head was present (F(1, 37) = 4.58, p = .039, η2 = .11), and when it was supplemented with the 2nd head F(1, 37) = 6.73, p = .014, η2 = .15. Although the effect of Group appeared slightly stronger when the 2nd head was present, the interaction Group × Number of Heads present did not reach significance. CBT effects on the gaze cone's width A One-way ANOVA on the width of the gaze cone in the Post-Test no longer revealed a significant influence of the factor Group. That is, the treated patients were no longer different from the control subjects. Neither the target head in isolation (F(1, 37) = 1.19, p = .282, η2 = .03) nor with the 2nd head revealed a significant difference (F(1, 37) = 2.51, p = .122, η2 = .06) (s. Fig. 3). Note, that the CBT effect was small, as a 2-way ANOVA (Group × Assessment), directly testing for the interaction between both factors did not yield significant differences and a t-test testing for a narrowing of the cone within the patient group yielded no significant differences between before and after CBT. In the SAD-Group, however, a significant correlation between the reduction of social anxiety, measured by the SPS, and the reduction of the gaze cone's width emerged (r = −.49, p = .037), and could not be shown for the Control-Group (r = −.14, p = .527). However, this correlation was not observed for the SIAS (r = −.32, p = .2). Width of the gaze cone in the two-heads-condition of the pre- and post-test for ... Fig. 3. Width of the gaze cone in the two-heads-condition of the pre- and post-test for the SAD-Group and Controls. Error bars represent standard errors of the mean.