تعصبات پردازش صورت در اضطراب اجتماعی: مطالعه الکتروفیزیولوژیک

Studies of information processing biases in social anxiety suggest abnormal processing of negative and positive social stimuli. To further investigate these biases, behavioral performance and event-related brain potentials (ERPs) were measured, while high- and low-socially anxious individuals performed a modified version of the Erikson flanker task comprised of negative and positive facial expressions. While no group differences emerged on behavioral measures, ERP results revealed the presence of a negative face bias in socially anxious subjects as indexed by the parietally maximal attention- and memory-related P3/late positive potential. Additionally, non-anxious subjects evidenced the presence of a positive face bias as reflected in the centrally maximal early attention- and emotion-modulated P2 and the frontally maximal response monitoring-related correct response negativity. These results demonstrate the sensitivity of different processing stages to different biases in high- versus low-socially anxious individuals that may prove important in advancing models of anxious pathology.

Current accounts of social anxiety suggest that it is characterized by abnormal processing of social threat information (see Heinrichs and Hofmann, 2001 for a review) as well as social safety and acceptance signals (see Kashdan, 2007 for a review). These two abnormalities or unique tendencies – typically referred to as ‘biases’ – in information processing manifest such that socially anxious individuals display a bias toward negative social stimuli (e.g., angry faces) whereas they fail to show the normal bias toward positive social stimuli (e.g., praising words). Better understanding the nature of information processing biases in social anxiety is essential to elucidating its conceptualization and treatment. Several lines of research support the idea that social anxiety is characterized by a bias towards social threat information. These studies show that while socially anxious individuals demonstrate preferential processing of (i.e., a bias toward) social threat, normal controls do not seem to show any bias at all (cf. Bar-Haim et al., 2007). For instance, behavioral studies have demonstrated facilitated response times (RT) to task-relevant stimuli that replace negative faces in dot probe tasks (Mogg and Bradley, 2002 and Mogg et al., 2004), faster detection of negative faces during visual search (Eastwood et al., 2005) and pop out (Gilboa-Schechtman et al., 1999), as well as slower disengagement from threat words in a Posner task (e.g., Amir et al., 2003) in socially anxious subjects. Further support for this notion comes from functional neuroimaging studies that have demonstrated hyperactive amygdala, extrastriate visual cortex and insula activation to negatively valenced facial stimuli in socially anxious subjects (Stein et al., 2002 and Straube et al., 2005). Evidence for socially anxious individuals’ failure to show a bias towards positive social stimuli is less robust, but still spans a number of different paradigms. Whereas the negative bias reviewed above is evident in the presence of preferential processing of threat information in socially anxious subjects and the lack of preferential processing in normal controls, the lack of positive bias is evident in the lack of preferential processing of positive social information in socially anxious subjects and the presence of preferential processing of positive social stimuli in normal controls. Socially anxious individuals, for example, fail to evince the faster RT advantage to words that complete ambiguous passages in a positive manner ( Hirsch and Mathews, 2000), to positive words that are associated with self-referential words ( Tanner et al., 2006), to positive faces ( Silvia et al., 2006) that normal controls do. Additionally, socially anxious subjects fail to show the bias to associate positive outcomes with positive facial expressions that normal controls do ( Garner et al., 2006). Thus, the behavioral and neuroimaging data reviewed above suggest that socially anxious subjects show a bias toward threat information, while normal controls do not, and normal controls show a bias toward positive information, while socially anxious subjects do not. However, it is unclear whether both biases can occur in a given experiment, as it seems that when a negative bias is shown, a lack of positive bias is not, and vice versa. One possible reason why the studies reviewed above demonstrate different biases is because the experimental paradigms employed might, in fact, tap into different processes. Another possible reason why studies show different biases is because of the measures typically employed, namely RT and hemodynamic activity. Both RT and hemodynamic activity reflect an amalgam of processes and might therefore be less sensitive to detecting multiple biases or biases reflected in multiple processes in the context of a given experiment. Event-related brain potentials (ERPs), on the other hand, are electrophysiological signals that allow for the examination of the sequence of constituent operations involved in processing and acting on incoming information on the order of milliseconds. Specifically, the ERP waveform represents multiple neural processes by discrete changes in voltage observed at the scalp – i.e., components – that offers several opportunities at detecting processing biases. Therefore, ERPs might be more sensitive to detecting the presence of biases. Consistent with this notion, several studies have demonstrated ERP differences between negative affective (anxious and depressed) and control groups in the face of comparable behavioral performance (Fallgatter et al., 2004, Hajcak et al., 2003, Hajcak et al., 2004a, Hajcak et al., 2004b, Hajcak and Simons, 2002 and Shestyuk et al., 2005). To our knowledge, only two recent studies have examined ERP correlates of information processing biases in social anxiety, however. First, Kolassa and Miltner (2006) reported somewhat larger occipito-temporal N170s to angry faces in socially anxious patients during an emotion identification task. More recently, Rossingol et al. (2007) found that high-socially anxious subjects evinced abnormal processing of anger and disgust faces as reflected in the N2b component (with 10 subjects in each group). Although preliminary, these studies suggest that ERPs can detect biases in the processing of facial expressions in social anxiety. In the current study, we intended to extend these recent findings by examining modulations of stimulus- and response-locked ERPs to negative and positive face categorization. We chose facial stimuli because the core feature of social anxiety is fear of negative social evaluation and rejection, and faces convey significant social information (cf. Adolphs, 2002, Bradley et al., 1997, Ekman, 1993, Izard, 1971 and Ohman et al., 2001). Additionally, we used negative and positive facial expressions, as it allowed us to examine biases in the processing of negative and positive social information that both appear to differentiate socially anxious from non-anxious subjects. By measuring both stimulus- and response-locked ERPs, we were able to examine whether socially anxious or non-anxious subjects showed (or lacked) a negative or positive bias at multiple points during information processing. Specifically, we examined the fronto-central P2 and N2 and parietal P3 of the stimulus-locked ERP and the fronto-central correct response negativity (CRN) of the response-locked ERP. Electrophysiological activity in the time window of the P2 and N2 seems to be a good candidate for studying information processing biases in social anxiety, as a recent review of the literature by Eimer and Holmes (2007) showed that emotional facial expressions elicit an enhanced fronto-central positive shift beginning around 150–200 ms post-stimulus. Eimer and Holmes suggested that the fronto-central modulations by facial expressions may reflect rapid representation of emotional significance in prefrontal regions. Additionally, the previously mentioned reports by Kolassa and Miltner (2006) and Rossingol et al. (2007) suggest that ERPs in this time window can detect information processing biases in social anxiety. Following the above-mentioned processes, the brain engages in more detailed analysis of visual information as reflected by the P3/late positive potential (LPP). The P3/LPP is a positive ERP component observed at parietal recording sites between 200 and 800 ms post-stimulus. The P3/LPP also seems to be a good candidate for studying information processing biases in social anxiety, as a large body of literature indicates that it is a neural index of attentional, perceptual and memory updating processes facilitated by motivationally relevant stimuli (Donchin, 1981, Donchin and Coles, 1988, Nieuwenhuis et al., 2005 and Schupp et al., 2000). The P3/LPP was also shown to be responsive to emotional facial expressions in Eimer and Holmes's (2007) review. In addition, the P3/LPP has been reliably responsive to fear-relevant stimuli in PTSD patients (Attias et al., 1996), panic patients (Pauli et al., 1997), spider phobic patients (Kolassa et al., 2005) and animal phobic students (Miltner et al., 2005). At the same time, a reduction in the P3/LPP to flanker stimuli (i.e., fear-irrelevant, task-relevant stimuli) was found when spider phobic subjects were exposed to a spider challenge (i.e., a fear-relevant, task-irrelevant stimulus; Moser et al., 2005). Taken together, it seems that the P3/LPP is a rather robust measure of emotional processing and information processing biases in anxiety. Response-locked ERPs reflect processes that occur around response execution that are essential to the monitoring and control of behavior. ERPs therefore allow for the differentiation of stimulus- and response-related processes that are confounded in RT measures. The CRN is one such ERP that indexes response-related processes and is typically observed as a negative deflection that peaks at fronto-central recording sites approximately 50–100 ms after a correct response is made in a two-choice speeded reaction time task (Bartholow et al., 2005, Vidal et al., 2000 and Vidal et al., 2003). More specifically, the CRN is part of a class of mediofrontal negativities believed to reflect action monitoring activity of the anterior cingulate cortex (ACC; Bartholow et al., 2005, Vidal et al., 2000 and Vidal et al., 2003). The CRN has been shown to be sensitive to response and strategy conflict (Bartholow et al., 2005), as well as the combination of cognitive conflict and affective context (Simon-Thomas and Knight, 2005). Based on such studies, it has been suggested that the CRN, and more generally activity of the ACC, is responsible for signaling other frontal brain structures such as the prefrontal cortex (PFC) that processing conflict has occurred and that increased cognitive control (e.g., a change in processing approach) is needed on subsequent trials to maximize performance goals (Bartholow et al., 2005; Bush et al., 2000; Simon-Thomas and Knight, 2005). As models of selective attention (e.g., Desimone and Duncan, 1995) and anxiety (e.g., Mathews and Mackintosh, 1998) posit that resource competition is a necessary condition for observing information processing biases, the CRN is an ideal measure because of the information it reflects about the demands on the frontal system imposed by cognitive and affective load. The current study, then, involved measuring the P2, N2 and P3/LPP of the stimulus-locked ERP and the CRN of the response-locked ERP, while high- and low-socially anxious individuals performed a modified version of the Eriksen flanker task (Eriksen and Eriksen, 1974) comprised of faces expressing negative (anger and disgust) and positive (happiness and surprise) emotions (hereafter the face flanker task). The face flanker task requires subjects to categorize the emotion of a centrally presented face while ignoring flanking distractor faces. Based on current conceptualizations and previous research, we hypothesized that socially anxious subjects would show a negative bias while controls would not (i.e., a negative bias in social anxiety), and controls would show a positive bias while socially anxious subjects would not (i.e., a lack of positive bias in social anxiety). The extant literature, however, does not support particularly strong predictions about exactly which ERPs should reflect these biases. With regard to the early ERPs – the P2 and N2 – previous results suggesting that early attentional and perceptual processes reflect a negative bias in social anxiety (Kolassa and Miltner, 2006 and Rossingol et al., 2007) are not particularly strong. Thus, it seemed unclear whether these ERPs would demonstrate reliable biases in the high- and low-anxious subjects. On the other hand, a number of previous ERP studies have shown that the P3/LPP is larger to threat-relevant stimuli in anxiety (e.g., Attias et al., 1996). We therefore felt confident predicting that the P3/LPP would demonstrate a negative bias in the high-socially anxious group and not in the low-socially anxious subjects. Likewise, as several studies of ERP modulations to affective pictures in unselected populations have shown that the P3/LPP is equally large to negative and positive images (e.g., Schupp et al., 2000), we predicted that the low-anxious group would show no bias in P3/LPP. Last, as no previous studies have used the CRN as a measure of information processing biases to emotional stimuli in anxious populations, it was again unclear what to expect. However, given that the face flanker task employed in the current study is most similar to the task used in the Silvia et al. (2006) study such that both are emotion categorization tasks, and Silvia et al. found a lack of positive bias in socially anxious subjects as reflected in RT, we hypothesized that the CRN would similarly reflect a lack of positive bias as it is an index of response-related processes. Specifically, we predicted that the low-anxious subjects would show a positive bias while the high-anxious subjects would show no bias. 1. Method 1.1. Participants Participants were recruited through the University of Delaware Psychology Department subject pool. Over 1000 undergraduate students completed the Social Phobia Inventory (SPIN; Connor et al., 2000) at a preliminary testing session as partial fulfillment of course requirements. The SPIN is a self-report measure of social phobia comprised of 17 questions that are rated on a Likert scale ranging from 0 (not at all) to 4 (extremely). Higher scores indicate more severe symptoms of social phobia. Following the preliminary testing session, participants were rank-ordered on the basis of their score on the SPIN. Twenty-one students (15 females) from the top 10% of the SPIN distribution comprised the high-socially anxious (High-SA) group and 21 students (11 females) from the bottom 10% of the SPIN distribution comprised the low-socially anxious (Low-SA) group. The groups did not differ with respect to gender ratio, χ2 (1, N = 42) = 1.62, p > 0.20. Retesting at the experimental session revealed that the high-socially anxious group remained significantly higher on the SPIN (M = 38.19, S.D. = 7.87) than the low-socially anxious group (M = 4.48, S.D. = 4.34; t(40) = 17.20, p < 0.001). Although it was not formally established that the college students in the current study met diagnostic criteria for social anxiety disorder, their scores on the SPIN are highly comparable to patients with clinical social anxiety reported previously (range 32.6–43 in Connor et al., 2000, Randall et al., 2001 and Stein et al., 2001) and well above the established cut score of 19 for the measure ( Connor et al., 2000). In addition to the SPIN, subjects completed the 21-Item Depression Anxiety Stress Scale (DASS-21; Lovibond and Lovibond, 1995) at the experimental session. The DASS-21 is a self-report measure comprised of three seven-item subscales, including the depression (DASS-D), anxiety (DASS-A) and stress reactivity (DASS-S) subscales that are rated on a Likert scale ranging from 0 (did not apply to me at all) to 3 (applied to me very much). Higher scores on all subscales indicate more severe symptoms. DASS-21 subscale scores are doubled so that they are comparable to the full 42-item subscale scores ( Lovibond and Lovibond, 1995). The high-socially anxious group scored significantly higher on all three subscales of the DASS-21: DASS-D (high-socially anxious group M = 12.10, S.D. = 8.52; low-socially anxious group M = 2.29, S.D. = 2.85; t(40) = 5.00, p < 0.001); DASS-A (high-socially anxious group M = 10.67, S.D. = 8.59; low-socially anxious group M = 2.29, S.D. = 3.59; t(40) = 4.13, p < 0.001); DASS-S (high-socially anxious group M = 18.67, S.D. = 9.00; low-socially anxious group M = 5.24, S.D. = 5.64; t(40) = 5.80, p < 0.001). The DASS-21 subscale scores reported here for the high-socially anxious group are highly comparable to those previously reported for a group of socially anxious patients (see Antony et al., 1998); in addition, the low-socially anxious group's scores are very similar to those previously reported for a non-clinical group of healthy community volunteers ( Antony et al., 1998). 1.2. Stimuli and task The stimulus set comprised 60 pictures of 30 male and female models each posing anger, disgust, happy and surprise facial expressions taken from Perez-Lopez and Woody's (2001) set of 184 photographs. Perez-Lopez and Woody collapsed the happy and surprise faces into one category that they called ‘reassuring’ and the anger and disgust faces into another category that they called ‘threatening’—throughout the remainder of this paper we will maintain usage of these labels. This stimulus set was chosen because it had independent ratings previously reported for it, it was successful in demonstrating information processing biases in social anxiety in the original study by Perez-Lopez and Woody, and included anger and disgust facial expression that have both been posited to play a prominent role in communicating social rejection that is the core fear of socially anxious individuals (e.g., Amir et al., 2005). In their original study, Perez-Lopez and Woddy (2001) had subjects rate all 184 pictures on a scale ranging from −5 (extremely threatening) to +5 (extremely reassuring) and reported an average rating of −1.94 for the threatening faces and 2.97 for the reassuring faces. In a rating session that followed the current experiment, participants (19 low-socially anxious and 20 high-socially anxious) rated the 60 pictures using an electronic version of the self-assessment manikin (SAM; Bradley and Lang, 1994). The SAM is a language-free measure that assesses the valence and arousal dimensions of emotional stimuli on a 1 (very unpleasant; low arousing) to 9 (very pleasant; highly arousing) scale. An analysis of the valence ratings revealed a significant main effect of face type (F(1, 37) = 169.52, p < 0.001), no main effect for group (F(1, 37) < 1) and no interaction between group and face type (F(1, 37) = 1.92, p > 0.17) indicating that both groups rated the threatening faces as more unpleasant (low-socially anxious M rating for threatening faces = 3.65; low-socially anxious M rating for reassuring faces = 6.74; high-socially anxious M rating for threatening faces = 3.35; high-socially anxious M rating for reassuring faces = 7.18). An analysis of the arousal ratings revealed no main effects or interactions (ps > 0.15) indicating that both groups rated the threatening and reassuring faces as equally arousing (low-socially anxious M rating for threatening faces = 2.3; low-socially anxious M rating for reassuring faces = 2.52; high-socially anxious M rating for threatening faces = 2.46; high-socially anxious M rating for reassuring faces = 2.52). A modified version of the Eriksen Flanker task (Eriksen and Eriksen, 1974) using the facial stimuli described above was administered on a Pentium III class computer, using Presentation software (Neurobehavioral Systems, Inc.) to control the presentation and timing of all stimuli, the determination of response accuracy and the measurement of reaction times. The classic flanker task requires participants to respond to a central target, such as the letter H, while ignoring flanking stimuli that are either associated with the same response (congruent stimuli, e.g., HHHHH) or an alternate response (incongruent stimuli, e.g., SSHSS). Fig. 1 shows that each face flanker stimulus was comprised of three emotional faces oriented horizontally. The three emotional expressions presented were always posed by the same male or female model. The figure also shows examples of congruent (all three express the same emotion) and incongruent (the center face expresses one emotion and the two flanking faces express another) threatening target stimuli (top two, respectively), and congruent and incongruent reassuring target stimuli (bottom two, respectively).