- Email: [email protected]
Cognitive processes in social anxiety: the effects of self-focus, rumination and anticipatory processing
Behaviour Research and Therapy 38 (2000) 243±257
www.elsevier.com/locate/brat
Cognitive processes in social anxiety: the eects of selffocus, rumination and anticipatory processing Tanna M.B. Mellings, Lynn E. Alden* University of British Columbia, Vancouver, BC, Canada V6T 1Z4 Received 4 January 1999
Abstract We examined three cognitive processes hypothesized to contribute to biases in judgments about and memory for social events: self-focused attention, post-event rumination, and anticipatory processing. Socially anxious (N = 58) and nonanxious (N = 58) subjects participated in a social interaction and then completed measures of self-focused attention and anxiety-related physiological sensations and behavior. The next day, subjects completed measures that assessed frequency of post-event processing and recall of the interaction. The results indicated that selective attention to negative self-related information led to biases in social judgments and recollections and that post-event processing contributed to the recall of negative self-related information. No evidence was found for selective retrieval of negative self-related information prior to a second social interaction. The results reconcile inconsistent previous ®ndings related to memory bias in social anxiety. # 2000 Published by Elsevier Science Ltd. All rights reserved. Keywords: Cognitive; Memory; Attention; Social anxiety; Rumination; Self-focus
1. Introduction Cognitive theories of social phobia are based on the general concept of biased processing of social information (e.g., Beck & Emery, 1985; Clark & Wells, 1995). Such models have three central tenets: (a) that patients with social phobia selectively attend to negative information about social situations; (b) that this leads to biases in their judgments and recollections of social events and (c) that these biases exacerbate and perpetuate social fears. Although there is * Corresponding author. Fax: +1-604-8226923. 0005-7967/00/$ - see front matter # 2000 Published by Elsevier Science Ltd. All rights reserved. PII: S 0 0 0 5 - 7 9 6 7 ( 9 9 ) 0 0 0 4 0 - 6
244
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
growing empirical support for this approach, several key features of the cognitive perspective remain to be con®rmed, particularly in regards to the existence of memory bias and the identi®cation of the speci®c cognitive operations that contribute to cognitive distortions. Clark & Wells, (1995) provided a detailed description of the way in which biased processing is believed to occur. According to these writers, in social situations, social phobics focus on monitoring themselves, particularly their anxiety-related internal sensations, thoughts and behaviors. Self-focused attention increases the salience of this negative self-related information at the expense of external information about the social situation. This not only exacerbates anxiety, but leads to negative biases in the person's social judgments both during and after the situation, particularly about the person's own performance. The adoption of safety behaviors contributes to this process by increasing anxiety and self-focused attention and preventing the discon®rmation of negative beliefs. Following social situations, social phobics engage in postevent processing in which they review the interaction in detail. According to Clark and Wells, (1995), this postmortem rumination is likely to center on anxious feelings and negative selfperceptions because they ``were processed in detail . . . and hence were strongly encoded in memory'' (Clark & Wells, 1995, p. 74). The end result of this is that the person recalls the interaction as more negative than it really was. Prior to subsequent social events, the social phobic engages in anticipatory processing in which their thoughts ``tend to be dominated by recollections of past failures'' and predictions of poor performance and thus, the cycle begins again. One unique feature of this model is the delineation of several distinct cognitive operations, most notably self-focused attention and anticipatory and post-event processing, that are proposed to contribute to social anxiety. Self-focused attention occurs during social events and highlights speci®c types of information. Post-event rumination occurs between social events and reactivates memory traces, which results in deeper processing of the information recalled. Anticipatory processing occurs prior to social events and evokes the negative memories and predictions that begin the anxious cycle. Thus, all three operations contribute to biased processing of social information and hence social anxiety, although at dierent points in time and in dierent ways. Some aspects of this model have empirical support. Research has demonstrated that social scrutiny increases self-focused attention in socially anxious and social phobic individuals (e.g., Buss, 1980; Alden, Teshuck & Tee, 1992; Woody, 1996) and it is well established that selffocused attention heightens awareness of negative emotional states (e.g., Buss, 1980). Similarly, socially anxious and social phobic individuals are known to display negative biases in their self and social judgments (e.g., McEwan & Devins, 1983; Lucock & Salkovskis, 1988; Rapee & Lim, 1992; Stopa & Clark, 1993; Alden & Wallace, 1995). However, the extent to which biased social judgments are due to self-focused attention, rather than other factors such as a general negative response set or the other cognitive operations identi®ed by Clark and Wells, has yet to be established. If judgmental biases arise as a result of self-focused attention, one would expect individuals who display more extensive self-directed attention to display greater distortions in their social judgments. Another unanswered question is whether self-focused attention leads to biases in memory for the social event. Although judgment and memory are clearly interrelated, they are not identical processes. For example, one may remember many pieces of information about a social event,
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
245
yet base one's general judgments (e.g., ``How well did I handle that situation?'') on only a few of these (e.g., a single inept comment). Therefore, demonstrating that socially anxious individuals discount their behavior in a social situation does not necessarily mean that they have stored only negative information about the event. Relatively few studies have examined memory biases in socially anxious populations and these have produced inconsistent results. The strongest evidence for such biases comes from a study by Hope, Heimberg, and Klein (1990), which examined recall following a social interaction. These researchers found that socially anxious subjects had less accurate recall of partner- and topic-related information than control subjects did. Support is also provided by Lundh and Ost (1996), who concluded that generalized social phobics displayed selective recall of pictures depicting negative as opposed to positive faces. However, this study did not control for the possibility that social phobics may have a general negative response set that aected their ratings. Weaker support is provided by O'Banion and Arkowitz (1977), who found a statistically nonsigni®cant trend for socially anxious subjects to recall more negative than positive feedback. In contrast to these three studies, Rapee, McCallum, Melville, Ravenscroft and Rodney (1994) conducted four studies that examined implicit and explicit memory on a variety of semantic and other memory retrieval tasks and found no dierences between social phobic and control subjects. It is dicult to compare these studies because they measured recall for dierent types of information (e.g., word lists, experimenter feedback), which also varied in terms of locus (internal versus external), valence (neutral versus positive or negative) and situational complexity during encoding. In addition, with the exception of the autobiographical memory retrieval study by Rapee et al. (1994, study 4), memory was assessed immediately after the interaction or stimulus presentation. It may be that more extensive memory biases would emerge if time was allowed for memory decay, i.e., assessed via delayed recall. It is also worth noting that some researchers have argued that biases in memory are associated with depression as opposed to anxiety (Mathews & MacLeod, 1994). Thus, studies of memory bias in social anxiety must control for the eects of comorbid depression and this has not always been done. Even if memory biases can be identi®ed, it remains to be established whether these stem from self-focused attention, as proposed by cognitive writers. Only one study examined selffocused attention in this regard. Hope et al. (1990) assessed public self-awareness and found a signi®cant, albeit modest, correlation with memory omissions for external social cues. However, public self-awareness was not associated with overall recall accuracy in socially anxious subjects and was actually associated with better recall in nonanxious subjects. Thus, although cognitive perspectives on social anxiety imply that self-focused attention should lead to biased recall of social events, this relationship remains to be established. Clark and Wells, (1995) suggest that both anticipatory and post-event processing play a role in social anxiety. Consistent with this suggestion, socially anxious individuals have been found to engage in negative thoughts and predictions prior to social interactions, which likely contribute to their anxiety (e.g., Dodge, Hope, Heimberg & Becker, 1988). However, the extent to which anticipatory processing also involves recollections of past failure remains to be determined. If this is the case, one would expect that social anticipation would lead to selective retrieval of negative information about past social events. In a similar vein, although socially anxious and social phobic individuals report ruminating after social events (e.g., Fairbrother, Rachman & Mitchell, 1998), no work of which we are aware has examined the relationship
246
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
between post-event processing and judgmental or recall bias. Such a relationship could take one of three forms. First, rumination may simply perpetuate existing biases by keeping memory traces active. On the other hand, further processing may increase the salience of negative self-related information, thereby exacerbating initial biases. Both of these patterns are in keeping with the Clark±Wells' model. In contrast, it is also conceivable that prolonged processing of an anxiety-provoking event might help some individuals resolve their concerns. Overall, the relationships between anticipatory- and post-event processing and cognitive biases require further examination. In this study, socially anxious students participated in a social interaction with an opposite sex confederate and then completed measures of self-focused attention, ratings of anxietyrelated physiological sensations and anxiety-related behavior. At a second session the next day, subjects rated frequency of ruminative thinking, memory for anxiety-related physiological sensations and anxiety-related behavior and two additional measures of recall of self- versus external-information. In this context, we examined the following predictions. Consistent with Clark and Wells' model, we anticipated that socially anxious subjects would report more self-focused attention and less other-focused attention than control subjects. We also anticipated that socially anxious subjects would over-estimate their anxiety-related behaviors relative to the ratings of objective observers, as this discounting eect is well established in the literature. If the cognitive perspective is correct, one would expect that selffocused attention would be associated with distortions in judgments about anxiety-related behavior and this was our prediction. We also predicted that self-focused attention would lead to biased recall of social information. In particular, we expected that socially anxious subjects would recall less partner-related (i.e., external) and more negative self-related information at delayed recall than nonanxious control subjects. Less precise predictions could be made about the eects of post-event- and anticipatoryprocessing because of the absence of previous work. However, if post-event processing plays a role in social anxiety, not only should socially anxious subjects report more rumination about the social interaction than control subjects, but in addition, post-event processing should be associated with either the perpetuation or exacerbation of any initial biases in encoding. In terms of anticipatory processing, we examined whether socially anxious subjects anticipating a second interaction displayed signs of selective retrieval of negative information about the initial interaction.
2. Method 2.1. Subjects Undergraduate psychology students (43 men, 73 women) were selected on the basis of their scores on the Social Avoidance and Distress scale (SAD; Watson & Friend, 1969)1. The SAD is a 28 item inventory measuring social anxiety and distress that has displayed good internal 1 Of the original 122 subjects, 6 subjects were dropped due to incomplete data (N = 1), language diculties (N = 3), or suspicion of the confederate (N = 2) to yield a total of 116 subjects.
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
247
consistency (KR-20=0.94) and reliability (test±retest=0.68) (Watson & Friend, 1969). Subjects were assigned to either high (SAD scores > 13) or low (SAD scores < 8) social anxiety groups. The high anxious group (26 men, 32 women) had a mean SAD score of 19.06 (S.D.=3.75, range=14±27). The low anxious group (17 men, 41 women) had a mean SAD score of 3.87 (S.D.=1.56, range=0±6). Within each group, subjects were randomly assigned to the high and low state-anxiety conditions. Chi-square analyses revealed no signi®cant dierences in gender between the various cells. 2.2. Personnel Two experimenters (1 woman, 1 man) were trained to follow a scripted protocol to deliver the experimental instructions. Two undergraduate research assistants (1 man, 1 woman) served as experimental confederates. They were trained to engage in consistent behavior across subjects. Four students (2 men, 2 women) served as observers and were trained to rate subject and confederate behavior. All personnel were blind to hypotheses and group assignments. 2.3. Procedure Subjects were assessed at two time points that were separated by one day. At time 1, subjects participated in an unstructured ten-minute interaction with the confederate, who was presented as another research participant. During this interaction, the confederates presented certain key information about themselves, such as their hobbies, interests and so on. Following the interaction, subjects completed measures of self-focused attention, anxiety-related physiological sensations and anxiety-related behaviors. Subjects returned on the following day (time 2) and were randomly assigned to one of two anticipation conditions. Following the delivery of the manipulation, subjects completed an open-ended recall measure and a structured measure of recall for objective (partner, setting) information. They also completed measures of the frequency of ruminative thinking during the interval and recall for anxiety-related physiological sensations and anxiety-related behaviors during the time 1 interaction. The questionnaires were completed in a dierent research room than at time 1. 2.4. Experimental manipulation of social anticipation Prior to completion of the questionnaires at time 2, the experimenter delivered instructions that indicated the subject either would or would not participate in a second interaction. Because some work suggests that memory biases are more likely to emerge when subjects are in a state of heightened aect (e.g., Mathews & MacLeod, 1994), these instructions were designed to heighten anxious apprehension. 2.4.1. Social apprehension condition In a few moments we will move to our large conference room where you will interact with your partner again. This interaction will take place in front of an audience who will be
248
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
rating your performance. Just before you go in, I would like you to think back to the conversation you had with your partner yesterday and answer these questions. 2.4.2. Control condition There is one last part to the study. As soon as you ®nish this part of the study, you may leave. Just before you leave, I would like you to think back to the conversation you had with your partner yesterday and answer these questions. 2.5. Deception During the debrie®ng, the experimenter assessed subject suspicion about the confederate. Two subjects expressed suspicion and were dropped from the analyses. 2.6. Measures: focus of attention, rumination and apprehension 2.6.1. Focus of attention Focus of attention was measured immediately after the interaction with the Focus of Attention Questionnaire (FAQ; Woody, 1996). This 10-item scale includes two 5-item subscales: self-focused attention (sample item: ``I was focusing on what I would say or do next'') and other-focused attention (sample item: ``I was focusing on the other person's appearance or dress''). Five additional self-focused attention items pertaining to anxiety were added for this study2. Each item was rated on a 5-point scale ranging from not at all (one) to very much (®ve). Items were averaged to yield two focus of attention scores. Research suggests that the two scales are independent and they display distinct patterns of correlations (Woody, Chambless & Glass, 1997). High internal consistency is reported for the scale (Woody et al., 1997). In this sample, Cronbach's alpha was 0.87 for the self-focus subscale and 0.49 for the other-focus subscale. 2.6.2. Rumination questionnaire (RQ) At time 2, subjects were asked to report the extent to which they had engaged in post-event processing of information about the interaction on a measure developed for this study. The RQ scale consisted of ®ve items that asked subjects how much they had thought about the interaction over the previous day3. Ratings were made on 7-point Likert-type scales. The ®ve 2 The added self-focused attention items included the following: ``I was focusing on my internal bodily reactions'' (for example, heart rate); ``I was focusing on past social failures''; ``I was focusing on my feelings of discomfort''; ``I was focusing on my shaking or trembling''; ``I was focusing on my shaky speech''. 3 The Rumination Questionnaire asked subjects to ``consider how much you thought about the last conversation in the time since that interaction and answer the following questions''. Speci®c items were as follows: To what extent did you think about the conversation with your partner in the time since you had the conversation? Were your thoughts about the interaction positive, negative or neutral? To what extent did you criticize yourself about not handling the interaction well? How much did you think about other past conversations or interactions? To what extent did you think about the anxiety you felt during the interaction?
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
249
items were summed to yield a total rumination score. Cronbach's alpha was 0.70 for the total score. 2.6.3. State trait anxiety inventory (STAI; Spielberger, Gorsuch, Lushene, Vagg & Jacobs, 1983) The State Anxiety Scale of the STAI was given to check on the eectiveness of the social apprehension manipulation. The State Anxiety Scale is a twenty-item questionnaire that evaluates the subject's current level of anxiety. Research suggests that the scale has good internal consistency and reliability (Spielberger et al., 1983).
2.7. Measures: social judgment and memory 2.7.1. Anxiety-related behavior (self-rating) Subjects rated three anxiety-related behaviors, (a) pauses, (b) eye contact and (c) ®dgeting, on three 7-point Likert-like scales. Ratings were made immediately after the interaction (time 1) and at delayed recall (time 2). At time 2, subjects were instructed to rate their behavior during the initial interaction. These ratings measure their recall of their earlier anxiety-related behavior. At each time point, behavior ratings were averaged to yield a total anxiety-related behavior score. Cronbach's alpha was 0.51 for time one ratings and 0.56 for time two ratings. This moderate level of agreement is typical for ratings of anxiety-related micro-behaviors. 2.7.2. Anxiety-related behavior (objective rating) Experimenters and observers rated subjects on the same three scales of anxiety-related behavior immediately following the interaction (time 1). Pearson correlations computed between the experimenter and observer ratings revealed good interrater reliability, r = 0.93, p < 0.001. Cronbach's alpha for the total score was 0.50, which again indicated moderate agreement between the three ratings. 2.7.3. Open-ended recall At time 2, subjects were asked to provide a written description of the interaction. These descriptions were content analyzed for references to: (a) negative self-related feelings, (b) negative self-behavior, (c) negative partner observations and (d) negative setting-related observations. Two measures were derived from this information: proportion of negative selfrelated comments and proportion of negative partner and setting-related comments. The protocols were rated by two trained judges who assigned each sentence to one of the four categories. Percent agreement on category assignment was high (98%). 2.7.4. Structured recall At time 2, subjects completed a structured measure that assessed their memory for external information about the interaction. The measure was based on Hope et al. (1990) and consisted of 16 questions asking for factual information related to the partner and the setting (e.g., ``What was your partner's major?'' or ``How many lamps were in the room?''). Two scores were calculated: total partner-related items correctly recalled (total partner recall) and total setting-related items correctly recalled (total setting recall). Interrater reliability was established
250
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
by having two raters score each subject's response yielding good interrater reliability (kappa=0.99). 2.7.5. Body Sensations Questionnaire (BSQ; Chambless, Caputo, Bright & Gallagher, 1984) The BSQ was completed at time 1 and again at time 2, when subjects were asked to recall their anxiety-related body sensations during the time 1 interaction. The BSQ is a 17-item questionnaire measuring the experience of autonomic arousal. Each item is rated on a 9-point scale ranging from ``not at all'' (one) to ``very much'' (nine). The BSQ has been shown to have good internal consistency and reliability (Chambless et al., 1984). In the current study, three items of sensations typically experienced by social phobics (e.g. muscle tension, feeling hot in the face (blushing) and trembling hands) were added. Data were analyzed with these items included and not included and there were no dierences in the results, therefore, only the analyses using all 20 items are reported here. Total scores were calculated at each time point. Cronbach's alpha for the current sample was 0.95. 2.8. Supplemental measure 2.8.1. Beck Depression Inventory (BDI; Beck, Rush, Shaw & Emery, 1979) The revised BDI is a 21-item inventory that measures the severity of depression for adolescents and adults during the past week. Research has shown that the BDI has good internal consistency, reliability and validity (Beck, Steer & Garbin, 1988). Dysphoric mood has been linked to judgmental and memory biases. The BDI was included to determine if any results in the current study were due to dysphoria rather than social anxiety. 3. Results 3.1. Preliminary analyses 3.1.1. Confederate behavior The experimenter and observer rated confederate behavior on two scales assessing friendliness and percentage of speech. Pearson correlations revealed adequate interrater reliability (friendliness r = 0.68, p < 0.001; percentage of speech r = 0.94, p < 0.001). A oneway (group) multivariate analysis of variance (MANOVA) conducted on these ratings revealed no signi®cant dierences in confederate behavior between the high and low social anxiety groups, F(2,97)=2.61, p > 0.05. This suggests that confederates engaged in consistent behavior across subjects and conditions. The experimenter also checked that confederates included necessary information in each session. 3.1.2. Social apprehension To examine the eectiveness of the experimental manipulation, total scores on the State Anxiety Scale were analyzed in a 2 (group) 2 (condition) analysis of variance (ANOVA). The group by condition interaction was signi®cant, F(1,112)=3.96, p < 0.05. Follow-up analyses revealed a signi®cant dierence between the two conditions for socially anxious subjects,
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
251
F(1,112)=22.42, p < 0.001, but not for control subjects, F(1,112)=3.69, p > 0.05. Thus, only the socially anxious subjects became apprehensive about the second interaction (See Table 2). 3.1.3. Depression To control for the potential eects of depression on the results, all analyses were repeated using analyses of covariance with BDI scores as the covariate. The results indicated that depression scores did not in¯uence the results, so only the original analyses are reported here.
3.2. Main analyses Means and standard deviations for the relevant dependent measures can be seen in Tables 1 and 2. The signi®cance level for all analyses was set at 0.05. 3.2.1. Self-focused attention The two scores from the FAQ were analyzed in a one-way (group) MANOVA. The group eect was signi®cant, F(2,113)=23.14, p < 0.001. Univariate analyses of variance revealed that, as expected, socially anxious subjects endorsed signi®cantly more self-focused attention items, Table 1 Means and standard deviations for dependent measures. N = 58 in each group Variable Focus of Attention Questionnaire Other-focused attention Self-focused attention Rumination Total rumination Anxiety-related behaviors Time 1 subject ratings Time 1 experimenter ratings Time 2 subject recall ratings Open-ended recall Proportion of negative self-related information Structured recall Total partner recall Total setting recall Judgmental Bias Scores Time 1 bias score Time 2 bias score Body Sensations Questionnaireta Time 1 score Time 2 score a
High socially anxious
Low socially anxious
2.60 (0.55) 2.48 (0.71)
2.69 (0.58) 1.79 (0.48)
18.90 (5.37)
14.80 (4.28)
4.06 (0.99) 3.31 (1.02) 3.85 (0.95)
3.10 (0.80) 2.76 (0.59) 2.98 (0.98)
0.49 (0.12)
0.45 (0.14)
7.04 (1.27) 3.55 (1.68)
7.76 (1.20) 4.04 (1.96)
1.13 (1.05) 0.91 (1.09)
0.73 (1.04) 0.61 (1.17)
52.74 (24.04) 50.62 (25.10)
31.38 (10.15) 30.55 (10.73)
A second analysis was conducted in which four outliers (2 anxious, 2 nonanxious), who scored three standard deviations from the mean, were eliminated. The results of this analysis were the same as the original one. Therefore, original means and standard deviations are reported.
252
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
Table 2 Means and standard deviations for scores on the state anxiety inventory. Groups with dierent subscripts diered signi®cantly. N = 29 in each group Variable
High socially anxious high state
State Anxiety Inventory Total score 48.93 (10.24)a
Low socially anxious low state
high state
low state
36.45 (10.54)b
37.72 (10.15)b
32.66 (9.15)b
F(1,114)=37.61, p < 0.001, than nonanxious subjects. There was no signi®cant between-group dierence for other-focused attention, F(1,114)=0.80, p > 0.10. 3.2.2. Rumination The total rumination score was analyzed in a 2 (group) 2 (condition) ANOVA. Compared to nonanxious subjects, the socially anxious subjects reported signi®cantly more post-event processing in the period between time 1 and 2, F(1,112)=20.18, p < 0.001. The condition, F(1,112)=2.28, p > 0.05 and group by condition interaction eects, F(1,112)=0.09, p > 0.05, were not signi®cant, which indicated that anxious apprehension had no eect on self-reported rumination. This indicates that subjects responded to this measure as intended, i.e., in terms of their thoughts about the previous event, not in preparation for the upcoming interaction. 3.2.3. Judgmental bias Judgmental bias was assessed by comparing subjects' self-rated behavior at time 1 with the objective behavioral ratings at time 1. A 2 (group) 2 (self versus other rater) between-within ANOVA conducted on the behavioral rating scores revealed a signi®cant main eect for group, F(1,97)=25.99, p < 0.001 and for rater, F(1,97)=38.96, p < 0.001. These main eects were modi®ed by a signi®cant group by rater interaction, F(1,97)=4.41, p < 0.05. Follow-up analyses revealed that all subjects overestimated their anxiety-related behavior relative to objective observers and that the socially anxious subjects displayed signi®cantly larger biases than nonanxious subjects. 3.2.4. Recall 3.2.4.1. Open-ended recall. The distribution of scores for both the negative self-related and negative partner/setting-related comments displayed a signi®cant positive skew. To correct this, a proportion score was created by dividing the number of negative self-related comments by the total number of negative comments. This score, which was normally distributed, re¯ected the extent to which subjects preferentially recalled negative self-related information. A 2 (group) 2 (condition) ANOVA conducted on this measure revealed a trend toward signi®cance between groups, F(1,112)=2.85, p < 0.10. An examination of group means revealed that socially anxious subjects tended to report more negative self-related information than did nonanxious subjects. The condition, F(1,112)=1.19, p > 0.05 and group by condition interaction eects, F(1,112)=1.70, p > 0.05, were not signi®cant. The absence of condition eects indi-
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
253
cated that anxious apprehension did not in¯uence subjects' freeform recollections of the interaction. 3.2.4.2. Structured recall. The two total scores re¯ecting partner and setting-related information were analyzed in a 2 (group) 2 (condition) MANOVA. The group eect was signi®cant, F(2,111)=5.10, p < 0.01. Follow-up ANOVA's revealed that anxious subjects recalled signi®cantly less partner-related information, F(1,112)=9.98, p < 0.01, than nonanxious subjects. The eect for total setting recall, F(1,114)=2.05, p > 0.05, was not signi®cant. The condition, F(2,111)=1.90, p > 0.05 and the group by condition interaction eects, F(2,111)=0.42, p > 0.05, were not signi®cant. The absence of condition eects indicated that anxious apprehension did not in¯uence subjects' recollections of external information about the interaction. 3.2.4.3. Recall of anxiety-related behaviors. To assess whether distortions in judgment were reduced, maintained, or increased at time 2, a 2 (group) 2 (condition) 2 (rater) betweenwithin ANCOVA was conducted comparing subjects' time 2 behavior ratings with the time 1 objective ratings, using the subjects' time 1 behavior ratings as covariate. As before, this analysis revealed a signi®cant eect for rater, F(1,94)=10.85, p < 0.001, which indicated that subjects rated their behavior more negatively than the observers. However, there were no signi®cant eects for group, F(1,94)=0.50, p > 0.05, condition, F(1,94)=0.07, p > 0.05, or for any of the interactions (group by condition, F(1,94)=1.35, p > 0.05, group by rater, F(1,94)=0.33, p > 0.05, condition by rater, F(1,94)=1.53, p > 0.05, group by condition by rater, F(1,94)=2.23, p > 0.05). This indicates that no changes occurred in judgmental biases over the delay period and that anxious apprehension did not aect these biases. 3.2.4.4. Recall of anxiety-related sensations. We assessed recall of anxiety-related symptoms between time 1 and time 2 on the BSQ by conducting a 2 (group) 2 (time ) 2 (condition) between-within ANOVA. The group eect was signi®cant F(1,112)=38.38, p < 0.01. As expected, socially anxious subjects reported signi®cantly more anxiety-related symptoms both immediately following the social interaction and at delayed recall. The time, condition, group by time, condition by time and three way interaction eects were not signi®cant, Fs(1,112)=3.75, 1.98, 0.72, 0.05 and 0.09, respectively, all ps > 0.05. This indicated that subjects' self-reported anxiety-related sensations did not change between time 1 and time 2 and that anxious apprehension had no eect on subjects' recollections of their anxiety-related sensations during the interaction.
3.2.5. Self-focused attention and cognitive bias A judgmental bias index was calculated by subtracting objective ratings of subject behavior from subjects' self-ratings of their own behavior both at time 1 and at time 2. Pearson correlation coecients, computed between the self-focused attention subscale of the FAQ and these judgmental bias scores, revealed that self-focused attention during the interaction was associated with a negative bias in self-related judgments at both time 1 and time 2. Pearson correlation coecients computed between self-focused attention and the open-ended and structured recall measures revealed that subjects who displayed greater self-focused attention
254
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
recalled less partner-related factual information. The correlation between self-focused attention and the open-ended recall measure was not signi®cant. (See Table 3.) 3.2.6. Post-event processing and recall bias Pearson correlation coecients, computed to examine the association between the RQ and the time 2 dependent variables, revealed that rumination predicted open-ended recall of negative self-related information, negative bias in self-judgments at time 2 and time 2 recall of anxiety-related sensations on the BSQ. However, rumination did not predict recall on either scale of the structured recall measure. (See Table 3.) To examine whether post-event processing contributed to biases in recall, partial correlation coecients were computed between the rumination scores and the time 2 judgmental bias index and time 2 recall of anxiety-related sensations on the BSQ, controlling for time 1 scores. These correlations failed to reach signi®cance.
4. Discussion Our goal was to determine the extent to which three cognitive operations contributed to biased processing of social information. The results suggested that selective attention to negative self-related information during the interaction led to biased judgments about and recollections of the event, a pattern that was more pronounced among socially anxious than nonanxious subjects. Post-event processing also contributed to the tendency to recall negative self-related information. Overall, these results pointed to the pivotal role of selective attention, rather than selective retrieval, in social anxiety. They also provide an explanation for inconsistencies between previous studies of memory bias in socially anxious individuals. Consistent with previous studies, socially anxious individuals focused their attention more on themselves than on their partners during the social interaction and displayed larger negative biases in their self-related judgments than did control subjects (Rapee & Lim, 1992; Stopa & Clark, 1993; Alden & Wallace, 1995; Woody, 1996). As predicted, socially anxious subjects also displayed biases in their recollections of the interaction on the following day. Relative to control subjects, they recalled less partner-related information on the structured recall task and tended to recall more negative self-related information on the unstructured recall task, Table 3 Pearson correlation coecients between self-focused attention, rumination and the social judgment and memory measures. N = 116. Bias=judgmental bias index. BSQ=Body Sensations Questionnaire Time 1 bias
Self-focused attention Rumination
p < 0.05;
p < 0.01;
0.37
p < 0.001.
Time 2 bias
0.32 0.39
Open-ended recall
0.10 0.20
Factual recall partner
setting
ÿ0.26 ÿ0.07
ÿ0.11 0.02
Time 2 BSQ
0.71 0.58
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
255
although this latter ®nding was modest. Thus, relative to nonanxious individuals, socially anxious individuals' memories for social events displayed biases that appeared to favor negative self-related, as opposed to external, information. The unique contribution of this research was to establish that biases in social judgment and memory were related to self-focused attention and post-event-processing. Subjects who focused more on themselves displayed larger biases in their self-related judgments and recalled less information about their partner. This was true for both socially anxious and nonanxious subjects. However, because the socially anxious subjects directed more of their attention to selfrelated information, their judgments displayed more bias and their store of partner-related information was less complete than that of nonanxious subjects. It is possible that greater attention to self-related phenomena limited the availability of cognitive resources necessary to process partner-related information. The overall pattern of results is consistent with the hypothesis that self-monitoring increases the salience of anxiety-related behaviors at the expense of external information, so that this information weighs more heavily in judgments about and recollections of social events. As predicted, compared to nonanxious subjects, socially anxious subjects engaged in more post-event processing in the day following the interaction. One of the more notable ®ndings was that frequency of post-event rumination, not self-focused attention, predicted recall of negative self-related information on the open-ended memory task. Thus, it appeared that ruminative thinking in the period following the interaction may have increased the salience of negative self-related information so that this information was more likely to come to mind when subjects described the event. Frequency of ruminative thinking was also associated with a continuation of the biases in judgment and factual memory observed at time 1, but did not contribute further distortion to either of these measures. Taken together, these results suggested that post-event processing not only perpetuated existing cognitive biases, but may have contributed to a bias in recall that favored negative self-related information. There was no support for the notion that social anticipation activated selective retrieval of negative information about previous social events. No dierences emerged between socially anxious subjects who anticipated a second interaction and those who did not in terms of their recollections of anxiety-related sensations and behaviors or other types of information. This suggested that encoding bias, along with post-event processing, was responsible for any cognitive distortions in the socially anxious subjects' recollections of social events, not selective retrieval of negative information about the previous interaction. This pattern of results is consistent with writings arguing that selective attention plays a greater role than selective retrieval in anxiety states (e.g., Mathews & MacLeod, 1994). The results also suggested an explanation for inconsistencies among earlier memory studies. As noted previously, several studies found evidence of memory bias (e.g., Hope et al., 1990), whereas others did not (Rapee et al., 1994). To reconcile these ®ndings, it is important to consider the time at which bias enters the cognitive system and the complexity of the information being encoded. The Rapee studies explicitly focused on selective retrieval, whereas Hope and her colleagues made no attempt to distinguish between distortions that might occur as the information was being encoded versus distortions that might arise later as a result of selective retrieval. Several of Rapee's studies involved semantic memory tasks in which subjects read lists of words (Rapee et al., 1994). Hope, on the other hand, studied memory in the
256
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
context of a social interaction. It seems likely that the latter task presented subjects with complex, potentially competing pieces of social information and that subjects were forced to select, not necessarily consciously, which information to consider and which to ignore. The current study found evidence to support both of these positions. Like Hope and her colleagues, we found that socially anxious individuals were less able to recall partner-related information and like Rapee, we found no evidence for selective retrieval of negative social information once that information had been encoded. Taken together, these studies indicate that social interaction tasks may lend themselves to encoding biases in socially anxious individuals, whereas semantic memory tasks may not. Once social information is entered into memory, however, selective retrieval of negative information does not appear to further distort recollections of the social event. These results must be tempered with a reminder that the study had some limitations. First, subjects were selected on the basis of a single measure of social distress and avoidance, which may have included subjects whose social distress stemmed from problems other than social anxiety. However, one signi®cant source of social avoidance, depression, was measured and did not aect the results. Second, generalization to patients with social phobia remains to be established. These patients would be expected to experience higher levels of anxiety and this might result in more extensive rumination and higher levels of anticipatory anxiety. At higher levels, these processes might add more distortion to judgmental and recall biases than was observed here. It has also been suggested that social phobics may be less able to override maladaptive cognitive processes than nonclinically anxious individuals. Work examining patients with social phobia is currently underway. Acknowledgements This research was supported by an SSHRC grant to the second author and an UBC University Graduate Fellowship to the ®rst author. We appreciate the assistance provided by Michaela Gelin, Cameron Weynaus, and Kam Brar in conducting this research. We would also like to thank S.J. Rachman, Ph.D. and E. Eich, Ph.D. for their input. References Alden, L. E., & Wallace, S. T. (1995). Social phobia and social appraisal in successful and unsuccessful interactions. Behaviour Research and Therapy, 33, 497±506. Alden, L. E., Teshuck, M., & Tee, K. (1992). Public self-awareness and withdrawal from social interactions. Cognitive Therapy and Research, 16, 249±267. Beck, A. T., & Emery, G. (1985). In Anxiety disorders and phobias: a cognitive perspective (pp. 146±164). New York: Basic Books. Beck, A. T., Rush, A. J., Shaw, B. F., & Emery, G. (1979). Cognitive therapy of depression. New York: Guilford Press. Beck, A. T., Steer, & Garbin (1988). Psychometric properties of the BDI: Twenty-®ve years of evaluation. Clinical Psychology Review, 8, 77±100. Buss, A. H. (1980). Self-consciousness and social anxiety. San Francisco: Freeman & Company. Chambless, D. L., Caputo, G. C., Bright, P., & Gallagher, R. (1984). Assessment of fear of fear in agoraphobics:
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
257
the body sensations questionnaire and the agoraphobic cognitions questionnaire. Journal of Consulting and Clinical Psychology, 52, 1090±1097. Clark, D. M., & Wells, A. (1995). A cognitive model of social phobia. In R. G. Heimberg, M. R. Liebowitz, D. A. Hope, & F. R. Schneier, Social phobia: diagnosis, assessment and treatment. New York: Guilford Press. Dodge, C. S., Hope, D. A., Heimberg, R. G., & Becker, R. E. (1988). Evaluation of the Social Interaction SelfStatement Test with a social phobic population. Cognitive Therapy and Research, 12(1), 211±222. Fairbrother, N., Rachman, S. J., & Mitchell, D. (1998, July). A new measure of social phobics' beliefs about what others think of them. Poster session presented at the World Congress of Behavioural and Cognitive Therapy, Acapulco, MX. Hope, D. A., Heimberg, R. G., & Klein, J. F. (1990). Social anxiety and the recall of interpersonal information. Journal of Cognitive Psychotherapy, 4, 185±195. Lucock, M. P., & Salkovskis, P. M. (1988). Cognitive factors in social anxiety and its treatment. Behaviour Research and Therapy, 26(4), 297±302. Lundh, L., & Ost, L. (1996). Recognition bias for critical faces in social phobics. Behaviour Research and Therapy, 34(10), 787±794. Mathews, A., & MacLeod, C. (1994). Cognitive approaches to emotion and emotional disorders. Annual Review of Psychology, 45, 25±50. McEwan, K. L., & Devins, G. M. (1983). Is increased arousal in social anxiety noticed by others? Journal of Abnormal Psychology, 92, 417±421. O'Banion, K., & Arkowitz, H. (1977). Social anxiety and selective memory for aective information about the self. Social Behavior and Personality, 5(2), 321±328. Rapee, R. M., & Lim, L. (1992). Discrepancy between self and observer ratings of performance in social phobics. Journal of Abnormal Psychology, 101, 728±731. Rapee, R. M., McCallum, S. L., Melville, L. F., Ravenscroft, H., & Rodney, J. M. (1994). Memory bias in social phobia. Behaviour Research and Therapy, 32(1), 89±99. Spielberger, C. D., Gorsuch, R. L., Lushene, R. E., Vagg, P. R., & Jacobs, G. A. (1983). Manual for the state-trait anxiety inventory. Palo Alto, CA: Consulting Psychologists Press. Stopa, L., & Clark, D. M. (1993). Cognitive processes in social phobia. Behaviour Research and Therapy, 31, 255± 267. Watson, D., & Friend, R. (1969). Measurement of social-evaluative anxiety. Journal of Consulting and Clinical Psychology, 33, 448±457. Woody, S. R. (1996). Eects of focus of attention on anxiety levels and social performance of individuals with social phobia. Journal of Abnormal Psychology, 105, 61±69. Woody, S. R., Chambless, D. L., & Glass, C. R. (1997). Self-focused attention in the treatment of social phobia. Behaviour Research and Therapy, 35(2), 117±129.
www.elsevier.com/locate/brat
Cognitive processes in social anxiety: the eects of selffocus, rumination and anticipatory processing Tanna M.B. Mellings, Lynn E. Alden* University of British Columbia, Vancouver, BC, Canada V6T 1Z4 Received 4 January 1999
Abstract We examined three cognitive processes hypothesized to contribute to biases in judgments about and memory for social events: self-focused attention, post-event rumination, and anticipatory processing. Socially anxious (N = 58) and nonanxious (N = 58) subjects participated in a social interaction and then completed measures of self-focused attention and anxiety-related physiological sensations and behavior. The next day, subjects completed measures that assessed frequency of post-event processing and recall of the interaction. The results indicated that selective attention to negative self-related information led to biases in social judgments and recollections and that post-event processing contributed to the recall of negative self-related information. No evidence was found for selective retrieval of negative self-related information prior to a second social interaction. The results reconcile inconsistent previous ®ndings related to memory bias in social anxiety. # 2000 Published by Elsevier Science Ltd. All rights reserved. Keywords: Cognitive; Memory; Attention; Social anxiety; Rumination; Self-focus
1. Introduction Cognitive theories of social phobia are based on the general concept of biased processing of social information (e.g., Beck & Emery, 1985; Clark & Wells, 1995). Such models have three central tenets: (a) that patients with social phobia selectively attend to negative information about social situations; (b) that this leads to biases in their judgments and recollections of social events and (c) that these biases exacerbate and perpetuate social fears. Although there is * Corresponding author. Fax: +1-604-8226923. 0005-7967/00/$ - see front matter # 2000 Published by Elsevier Science Ltd. All rights reserved. PII: S 0 0 0 5 - 7 9 6 7 ( 9 9 ) 0 0 0 4 0 - 6
244
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
growing empirical support for this approach, several key features of the cognitive perspective remain to be con®rmed, particularly in regards to the existence of memory bias and the identi®cation of the speci®c cognitive operations that contribute to cognitive distortions. Clark & Wells, (1995) provided a detailed description of the way in which biased processing is believed to occur. According to these writers, in social situations, social phobics focus on monitoring themselves, particularly their anxiety-related internal sensations, thoughts and behaviors. Self-focused attention increases the salience of this negative self-related information at the expense of external information about the social situation. This not only exacerbates anxiety, but leads to negative biases in the person's social judgments both during and after the situation, particularly about the person's own performance. The adoption of safety behaviors contributes to this process by increasing anxiety and self-focused attention and preventing the discon®rmation of negative beliefs. Following social situations, social phobics engage in postevent processing in which they review the interaction in detail. According to Clark and Wells, (1995), this postmortem rumination is likely to center on anxious feelings and negative selfperceptions because they ``were processed in detail . . . and hence were strongly encoded in memory'' (Clark & Wells, 1995, p. 74). The end result of this is that the person recalls the interaction as more negative than it really was. Prior to subsequent social events, the social phobic engages in anticipatory processing in which their thoughts ``tend to be dominated by recollections of past failures'' and predictions of poor performance and thus, the cycle begins again. One unique feature of this model is the delineation of several distinct cognitive operations, most notably self-focused attention and anticipatory and post-event processing, that are proposed to contribute to social anxiety. Self-focused attention occurs during social events and highlights speci®c types of information. Post-event rumination occurs between social events and reactivates memory traces, which results in deeper processing of the information recalled. Anticipatory processing occurs prior to social events and evokes the negative memories and predictions that begin the anxious cycle. Thus, all three operations contribute to biased processing of social information and hence social anxiety, although at dierent points in time and in dierent ways. Some aspects of this model have empirical support. Research has demonstrated that social scrutiny increases self-focused attention in socially anxious and social phobic individuals (e.g., Buss, 1980; Alden, Teshuck & Tee, 1992; Woody, 1996) and it is well established that selffocused attention heightens awareness of negative emotional states (e.g., Buss, 1980). Similarly, socially anxious and social phobic individuals are known to display negative biases in their self and social judgments (e.g., McEwan & Devins, 1983; Lucock & Salkovskis, 1988; Rapee & Lim, 1992; Stopa & Clark, 1993; Alden & Wallace, 1995). However, the extent to which biased social judgments are due to self-focused attention, rather than other factors such as a general negative response set or the other cognitive operations identi®ed by Clark and Wells, has yet to be established. If judgmental biases arise as a result of self-focused attention, one would expect individuals who display more extensive self-directed attention to display greater distortions in their social judgments. Another unanswered question is whether self-focused attention leads to biases in memory for the social event. Although judgment and memory are clearly interrelated, they are not identical processes. For example, one may remember many pieces of information about a social event,
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
245
yet base one's general judgments (e.g., ``How well did I handle that situation?'') on only a few of these (e.g., a single inept comment). Therefore, demonstrating that socially anxious individuals discount their behavior in a social situation does not necessarily mean that they have stored only negative information about the event. Relatively few studies have examined memory biases in socially anxious populations and these have produced inconsistent results. The strongest evidence for such biases comes from a study by Hope, Heimberg, and Klein (1990), which examined recall following a social interaction. These researchers found that socially anxious subjects had less accurate recall of partner- and topic-related information than control subjects did. Support is also provided by Lundh and Ost (1996), who concluded that generalized social phobics displayed selective recall of pictures depicting negative as opposed to positive faces. However, this study did not control for the possibility that social phobics may have a general negative response set that aected their ratings. Weaker support is provided by O'Banion and Arkowitz (1977), who found a statistically nonsigni®cant trend for socially anxious subjects to recall more negative than positive feedback. In contrast to these three studies, Rapee, McCallum, Melville, Ravenscroft and Rodney (1994) conducted four studies that examined implicit and explicit memory on a variety of semantic and other memory retrieval tasks and found no dierences between social phobic and control subjects. It is dicult to compare these studies because they measured recall for dierent types of information (e.g., word lists, experimenter feedback), which also varied in terms of locus (internal versus external), valence (neutral versus positive or negative) and situational complexity during encoding. In addition, with the exception of the autobiographical memory retrieval study by Rapee et al. (1994, study 4), memory was assessed immediately after the interaction or stimulus presentation. It may be that more extensive memory biases would emerge if time was allowed for memory decay, i.e., assessed via delayed recall. It is also worth noting that some researchers have argued that biases in memory are associated with depression as opposed to anxiety (Mathews & MacLeod, 1994). Thus, studies of memory bias in social anxiety must control for the eects of comorbid depression and this has not always been done. Even if memory biases can be identi®ed, it remains to be established whether these stem from self-focused attention, as proposed by cognitive writers. Only one study examined selffocused attention in this regard. Hope et al. (1990) assessed public self-awareness and found a signi®cant, albeit modest, correlation with memory omissions for external social cues. However, public self-awareness was not associated with overall recall accuracy in socially anxious subjects and was actually associated with better recall in nonanxious subjects. Thus, although cognitive perspectives on social anxiety imply that self-focused attention should lead to biased recall of social events, this relationship remains to be established. Clark and Wells, (1995) suggest that both anticipatory and post-event processing play a role in social anxiety. Consistent with this suggestion, socially anxious individuals have been found to engage in negative thoughts and predictions prior to social interactions, which likely contribute to their anxiety (e.g., Dodge, Hope, Heimberg & Becker, 1988). However, the extent to which anticipatory processing also involves recollections of past failure remains to be determined. If this is the case, one would expect that social anticipation would lead to selective retrieval of negative information about past social events. In a similar vein, although socially anxious and social phobic individuals report ruminating after social events (e.g., Fairbrother, Rachman & Mitchell, 1998), no work of which we are aware has examined the relationship
246
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
between post-event processing and judgmental or recall bias. Such a relationship could take one of three forms. First, rumination may simply perpetuate existing biases by keeping memory traces active. On the other hand, further processing may increase the salience of negative self-related information, thereby exacerbating initial biases. Both of these patterns are in keeping with the Clark±Wells' model. In contrast, it is also conceivable that prolonged processing of an anxiety-provoking event might help some individuals resolve their concerns. Overall, the relationships between anticipatory- and post-event processing and cognitive biases require further examination. In this study, socially anxious students participated in a social interaction with an opposite sex confederate and then completed measures of self-focused attention, ratings of anxietyrelated physiological sensations and anxiety-related behavior. At a second session the next day, subjects rated frequency of ruminative thinking, memory for anxiety-related physiological sensations and anxiety-related behavior and two additional measures of recall of self- versus external-information. In this context, we examined the following predictions. Consistent with Clark and Wells' model, we anticipated that socially anxious subjects would report more self-focused attention and less other-focused attention than control subjects. We also anticipated that socially anxious subjects would over-estimate their anxiety-related behaviors relative to the ratings of objective observers, as this discounting eect is well established in the literature. If the cognitive perspective is correct, one would expect that selffocused attention would be associated with distortions in judgments about anxiety-related behavior and this was our prediction. We also predicted that self-focused attention would lead to biased recall of social information. In particular, we expected that socially anxious subjects would recall less partner-related (i.e., external) and more negative self-related information at delayed recall than nonanxious control subjects. Less precise predictions could be made about the eects of post-event- and anticipatoryprocessing because of the absence of previous work. However, if post-event processing plays a role in social anxiety, not only should socially anxious subjects report more rumination about the social interaction than control subjects, but in addition, post-event processing should be associated with either the perpetuation or exacerbation of any initial biases in encoding. In terms of anticipatory processing, we examined whether socially anxious subjects anticipating a second interaction displayed signs of selective retrieval of negative information about the initial interaction.
2. Method 2.1. Subjects Undergraduate psychology students (43 men, 73 women) were selected on the basis of their scores on the Social Avoidance and Distress scale (SAD; Watson & Friend, 1969)1. The SAD is a 28 item inventory measuring social anxiety and distress that has displayed good internal 1 Of the original 122 subjects, 6 subjects were dropped due to incomplete data (N = 1), language diculties (N = 3), or suspicion of the confederate (N = 2) to yield a total of 116 subjects.
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
247
consistency (KR-20=0.94) and reliability (test±retest=0.68) (Watson & Friend, 1969). Subjects were assigned to either high (SAD scores > 13) or low (SAD scores < 8) social anxiety groups. The high anxious group (26 men, 32 women) had a mean SAD score of 19.06 (S.D.=3.75, range=14±27). The low anxious group (17 men, 41 women) had a mean SAD score of 3.87 (S.D.=1.56, range=0±6). Within each group, subjects were randomly assigned to the high and low state-anxiety conditions. Chi-square analyses revealed no signi®cant dierences in gender between the various cells. 2.2. Personnel Two experimenters (1 woman, 1 man) were trained to follow a scripted protocol to deliver the experimental instructions. Two undergraduate research assistants (1 man, 1 woman) served as experimental confederates. They were trained to engage in consistent behavior across subjects. Four students (2 men, 2 women) served as observers and were trained to rate subject and confederate behavior. All personnel were blind to hypotheses and group assignments. 2.3. Procedure Subjects were assessed at two time points that were separated by one day. At time 1, subjects participated in an unstructured ten-minute interaction with the confederate, who was presented as another research participant. During this interaction, the confederates presented certain key information about themselves, such as their hobbies, interests and so on. Following the interaction, subjects completed measures of self-focused attention, anxiety-related physiological sensations and anxiety-related behaviors. Subjects returned on the following day (time 2) and were randomly assigned to one of two anticipation conditions. Following the delivery of the manipulation, subjects completed an open-ended recall measure and a structured measure of recall for objective (partner, setting) information. They also completed measures of the frequency of ruminative thinking during the interval and recall for anxiety-related physiological sensations and anxiety-related behaviors during the time 1 interaction. The questionnaires were completed in a dierent research room than at time 1. 2.4. Experimental manipulation of social anticipation Prior to completion of the questionnaires at time 2, the experimenter delivered instructions that indicated the subject either would or would not participate in a second interaction. Because some work suggests that memory biases are more likely to emerge when subjects are in a state of heightened aect (e.g., Mathews & MacLeod, 1994), these instructions were designed to heighten anxious apprehension. 2.4.1. Social apprehension condition In a few moments we will move to our large conference room where you will interact with your partner again. This interaction will take place in front of an audience who will be
248
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
rating your performance. Just before you go in, I would like you to think back to the conversation you had with your partner yesterday and answer these questions. 2.4.2. Control condition There is one last part to the study. As soon as you ®nish this part of the study, you may leave. Just before you leave, I would like you to think back to the conversation you had with your partner yesterday and answer these questions. 2.5. Deception During the debrie®ng, the experimenter assessed subject suspicion about the confederate. Two subjects expressed suspicion and were dropped from the analyses. 2.6. Measures: focus of attention, rumination and apprehension 2.6.1. Focus of attention Focus of attention was measured immediately after the interaction with the Focus of Attention Questionnaire (FAQ; Woody, 1996). This 10-item scale includes two 5-item subscales: self-focused attention (sample item: ``I was focusing on what I would say or do next'') and other-focused attention (sample item: ``I was focusing on the other person's appearance or dress''). Five additional self-focused attention items pertaining to anxiety were added for this study2. Each item was rated on a 5-point scale ranging from not at all (one) to very much (®ve). Items were averaged to yield two focus of attention scores. Research suggests that the two scales are independent and they display distinct patterns of correlations (Woody, Chambless & Glass, 1997). High internal consistency is reported for the scale (Woody et al., 1997). In this sample, Cronbach's alpha was 0.87 for the self-focus subscale and 0.49 for the other-focus subscale. 2.6.2. Rumination questionnaire (RQ) At time 2, subjects were asked to report the extent to which they had engaged in post-event processing of information about the interaction on a measure developed for this study. The RQ scale consisted of ®ve items that asked subjects how much they had thought about the interaction over the previous day3. Ratings were made on 7-point Likert-type scales. The ®ve 2 The added self-focused attention items included the following: ``I was focusing on my internal bodily reactions'' (for example, heart rate); ``I was focusing on past social failures''; ``I was focusing on my feelings of discomfort''; ``I was focusing on my shaking or trembling''; ``I was focusing on my shaky speech''. 3 The Rumination Questionnaire asked subjects to ``consider how much you thought about the last conversation in the time since that interaction and answer the following questions''. Speci®c items were as follows: To what extent did you think about the conversation with your partner in the time since you had the conversation? Were your thoughts about the interaction positive, negative or neutral? To what extent did you criticize yourself about not handling the interaction well? How much did you think about other past conversations or interactions? To what extent did you think about the anxiety you felt during the interaction?
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
249
items were summed to yield a total rumination score. Cronbach's alpha was 0.70 for the total score. 2.6.3. State trait anxiety inventory (STAI; Spielberger, Gorsuch, Lushene, Vagg & Jacobs, 1983) The State Anxiety Scale of the STAI was given to check on the eectiveness of the social apprehension manipulation. The State Anxiety Scale is a twenty-item questionnaire that evaluates the subject's current level of anxiety. Research suggests that the scale has good internal consistency and reliability (Spielberger et al., 1983).
2.7. Measures: social judgment and memory 2.7.1. Anxiety-related behavior (self-rating) Subjects rated three anxiety-related behaviors, (a) pauses, (b) eye contact and (c) ®dgeting, on three 7-point Likert-like scales. Ratings were made immediately after the interaction (time 1) and at delayed recall (time 2). At time 2, subjects were instructed to rate their behavior during the initial interaction. These ratings measure their recall of their earlier anxiety-related behavior. At each time point, behavior ratings were averaged to yield a total anxiety-related behavior score. Cronbach's alpha was 0.51 for time one ratings and 0.56 for time two ratings. This moderate level of agreement is typical for ratings of anxiety-related micro-behaviors. 2.7.2. Anxiety-related behavior (objective rating) Experimenters and observers rated subjects on the same three scales of anxiety-related behavior immediately following the interaction (time 1). Pearson correlations computed between the experimenter and observer ratings revealed good interrater reliability, r = 0.93, p < 0.001. Cronbach's alpha for the total score was 0.50, which again indicated moderate agreement between the three ratings. 2.7.3. Open-ended recall At time 2, subjects were asked to provide a written description of the interaction. These descriptions were content analyzed for references to: (a) negative self-related feelings, (b) negative self-behavior, (c) negative partner observations and (d) negative setting-related observations. Two measures were derived from this information: proportion of negative selfrelated comments and proportion of negative partner and setting-related comments. The protocols were rated by two trained judges who assigned each sentence to one of the four categories. Percent agreement on category assignment was high (98%). 2.7.4. Structured recall At time 2, subjects completed a structured measure that assessed their memory for external information about the interaction. The measure was based on Hope et al. (1990) and consisted of 16 questions asking for factual information related to the partner and the setting (e.g., ``What was your partner's major?'' or ``How many lamps were in the room?''). Two scores were calculated: total partner-related items correctly recalled (total partner recall) and total setting-related items correctly recalled (total setting recall). Interrater reliability was established
250
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
by having two raters score each subject's response yielding good interrater reliability (kappa=0.99). 2.7.5. Body Sensations Questionnaire (BSQ; Chambless, Caputo, Bright & Gallagher, 1984) The BSQ was completed at time 1 and again at time 2, when subjects were asked to recall their anxiety-related body sensations during the time 1 interaction. The BSQ is a 17-item questionnaire measuring the experience of autonomic arousal. Each item is rated on a 9-point scale ranging from ``not at all'' (one) to ``very much'' (nine). The BSQ has been shown to have good internal consistency and reliability (Chambless et al., 1984). In the current study, three items of sensations typically experienced by social phobics (e.g. muscle tension, feeling hot in the face (blushing) and trembling hands) were added. Data were analyzed with these items included and not included and there were no dierences in the results, therefore, only the analyses using all 20 items are reported here. Total scores were calculated at each time point. Cronbach's alpha for the current sample was 0.95. 2.8. Supplemental measure 2.8.1. Beck Depression Inventory (BDI; Beck, Rush, Shaw & Emery, 1979) The revised BDI is a 21-item inventory that measures the severity of depression for adolescents and adults during the past week. Research has shown that the BDI has good internal consistency, reliability and validity (Beck, Steer & Garbin, 1988). Dysphoric mood has been linked to judgmental and memory biases. The BDI was included to determine if any results in the current study were due to dysphoria rather than social anxiety. 3. Results 3.1. Preliminary analyses 3.1.1. Confederate behavior The experimenter and observer rated confederate behavior on two scales assessing friendliness and percentage of speech. Pearson correlations revealed adequate interrater reliability (friendliness r = 0.68, p < 0.001; percentage of speech r = 0.94, p < 0.001). A oneway (group) multivariate analysis of variance (MANOVA) conducted on these ratings revealed no signi®cant dierences in confederate behavior between the high and low social anxiety groups, F(2,97)=2.61, p > 0.05. This suggests that confederates engaged in consistent behavior across subjects and conditions. The experimenter also checked that confederates included necessary information in each session. 3.1.2. Social apprehension To examine the eectiveness of the experimental manipulation, total scores on the State Anxiety Scale were analyzed in a 2 (group) 2 (condition) analysis of variance (ANOVA). The group by condition interaction was signi®cant, F(1,112)=3.96, p < 0.05. Follow-up analyses revealed a signi®cant dierence between the two conditions for socially anxious subjects,
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
251
F(1,112)=22.42, p < 0.001, but not for control subjects, F(1,112)=3.69, p > 0.05. Thus, only the socially anxious subjects became apprehensive about the second interaction (See Table 2). 3.1.3. Depression To control for the potential eects of depression on the results, all analyses were repeated using analyses of covariance with BDI scores as the covariate. The results indicated that depression scores did not in¯uence the results, so only the original analyses are reported here.
3.2. Main analyses Means and standard deviations for the relevant dependent measures can be seen in Tables 1 and 2. The signi®cance level for all analyses was set at 0.05. 3.2.1. Self-focused attention The two scores from the FAQ were analyzed in a one-way (group) MANOVA. The group eect was signi®cant, F(2,113)=23.14, p < 0.001. Univariate analyses of variance revealed that, as expected, socially anxious subjects endorsed signi®cantly more self-focused attention items, Table 1 Means and standard deviations for dependent measures. N = 58 in each group Variable Focus of Attention Questionnaire Other-focused attention Self-focused attention Rumination Total rumination Anxiety-related behaviors Time 1 subject ratings Time 1 experimenter ratings Time 2 subject recall ratings Open-ended recall Proportion of negative self-related information Structured recall Total partner recall Total setting recall Judgmental Bias Scores Time 1 bias score Time 2 bias score Body Sensations Questionnaireta Time 1 score Time 2 score a
High socially anxious
Low socially anxious
2.60 (0.55) 2.48 (0.71)
2.69 (0.58) 1.79 (0.48)
18.90 (5.37)
14.80 (4.28)
4.06 (0.99) 3.31 (1.02) 3.85 (0.95)
3.10 (0.80) 2.76 (0.59) 2.98 (0.98)
0.49 (0.12)
0.45 (0.14)
7.04 (1.27) 3.55 (1.68)
7.76 (1.20) 4.04 (1.96)
1.13 (1.05) 0.91 (1.09)
0.73 (1.04) 0.61 (1.17)
52.74 (24.04) 50.62 (25.10)
31.38 (10.15) 30.55 (10.73)
A second analysis was conducted in which four outliers (2 anxious, 2 nonanxious), who scored three standard deviations from the mean, were eliminated. The results of this analysis were the same as the original one. Therefore, original means and standard deviations are reported.
252
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
Table 2 Means and standard deviations for scores on the state anxiety inventory. Groups with dierent subscripts diered signi®cantly. N = 29 in each group Variable
High socially anxious high state
State Anxiety Inventory Total score 48.93 (10.24)a
Low socially anxious low state
high state
low state
36.45 (10.54)b
37.72 (10.15)b
32.66 (9.15)b
F(1,114)=37.61, p < 0.001, than nonanxious subjects. There was no signi®cant between-group dierence for other-focused attention, F(1,114)=0.80, p > 0.10. 3.2.2. Rumination The total rumination score was analyzed in a 2 (group) 2 (condition) ANOVA. Compared to nonanxious subjects, the socially anxious subjects reported signi®cantly more post-event processing in the period between time 1 and 2, F(1,112)=20.18, p < 0.001. The condition, F(1,112)=2.28, p > 0.05 and group by condition interaction eects, F(1,112)=0.09, p > 0.05, were not signi®cant, which indicated that anxious apprehension had no eect on self-reported rumination. This indicates that subjects responded to this measure as intended, i.e., in terms of their thoughts about the previous event, not in preparation for the upcoming interaction. 3.2.3. Judgmental bias Judgmental bias was assessed by comparing subjects' self-rated behavior at time 1 with the objective behavioral ratings at time 1. A 2 (group) 2 (self versus other rater) between-within ANOVA conducted on the behavioral rating scores revealed a signi®cant main eect for group, F(1,97)=25.99, p < 0.001 and for rater, F(1,97)=38.96, p < 0.001. These main eects were modi®ed by a signi®cant group by rater interaction, F(1,97)=4.41, p < 0.05. Follow-up analyses revealed that all subjects overestimated their anxiety-related behavior relative to objective observers and that the socially anxious subjects displayed signi®cantly larger biases than nonanxious subjects. 3.2.4. Recall 3.2.4.1. Open-ended recall. The distribution of scores for both the negative self-related and negative partner/setting-related comments displayed a signi®cant positive skew. To correct this, a proportion score was created by dividing the number of negative self-related comments by the total number of negative comments. This score, which was normally distributed, re¯ected the extent to which subjects preferentially recalled negative self-related information. A 2 (group) 2 (condition) ANOVA conducted on this measure revealed a trend toward signi®cance between groups, F(1,112)=2.85, p < 0.10. An examination of group means revealed that socially anxious subjects tended to report more negative self-related information than did nonanxious subjects. The condition, F(1,112)=1.19, p > 0.05 and group by condition interaction eects, F(1,112)=1.70, p > 0.05, were not signi®cant. The absence of condition eects indi-
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
253
cated that anxious apprehension did not in¯uence subjects' freeform recollections of the interaction. 3.2.4.2. Structured recall. The two total scores re¯ecting partner and setting-related information were analyzed in a 2 (group) 2 (condition) MANOVA. The group eect was signi®cant, F(2,111)=5.10, p < 0.01. Follow-up ANOVA's revealed that anxious subjects recalled signi®cantly less partner-related information, F(1,112)=9.98, p < 0.01, than nonanxious subjects. The eect for total setting recall, F(1,114)=2.05, p > 0.05, was not signi®cant. The condition, F(2,111)=1.90, p > 0.05 and the group by condition interaction eects, F(2,111)=0.42, p > 0.05, were not signi®cant. The absence of condition eects indicated that anxious apprehension did not in¯uence subjects' recollections of external information about the interaction. 3.2.4.3. Recall of anxiety-related behaviors. To assess whether distortions in judgment were reduced, maintained, or increased at time 2, a 2 (group) 2 (condition) 2 (rater) betweenwithin ANCOVA was conducted comparing subjects' time 2 behavior ratings with the time 1 objective ratings, using the subjects' time 1 behavior ratings as covariate. As before, this analysis revealed a signi®cant eect for rater, F(1,94)=10.85, p < 0.001, which indicated that subjects rated their behavior more negatively than the observers. However, there were no signi®cant eects for group, F(1,94)=0.50, p > 0.05, condition, F(1,94)=0.07, p > 0.05, or for any of the interactions (group by condition, F(1,94)=1.35, p > 0.05, group by rater, F(1,94)=0.33, p > 0.05, condition by rater, F(1,94)=1.53, p > 0.05, group by condition by rater, F(1,94)=2.23, p > 0.05). This indicates that no changes occurred in judgmental biases over the delay period and that anxious apprehension did not aect these biases. 3.2.4.4. Recall of anxiety-related sensations. We assessed recall of anxiety-related symptoms between time 1 and time 2 on the BSQ by conducting a 2 (group) 2 (time ) 2 (condition) between-within ANOVA. The group eect was signi®cant F(1,112)=38.38, p < 0.01. As expected, socially anxious subjects reported signi®cantly more anxiety-related symptoms both immediately following the social interaction and at delayed recall. The time, condition, group by time, condition by time and three way interaction eects were not signi®cant, Fs(1,112)=3.75, 1.98, 0.72, 0.05 and 0.09, respectively, all ps > 0.05. This indicated that subjects' self-reported anxiety-related sensations did not change between time 1 and time 2 and that anxious apprehension had no eect on subjects' recollections of their anxiety-related sensations during the interaction.
3.2.5. Self-focused attention and cognitive bias A judgmental bias index was calculated by subtracting objective ratings of subject behavior from subjects' self-ratings of their own behavior both at time 1 and at time 2. Pearson correlation coecients, computed between the self-focused attention subscale of the FAQ and these judgmental bias scores, revealed that self-focused attention during the interaction was associated with a negative bias in self-related judgments at both time 1 and time 2. Pearson correlation coecients computed between self-focused attention and the open-ended and structured recall measures revealed that subjects who displayed greater self-focused attention
254
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
recalled less partner-related factual information. The correlation between self-focused attention and the open-ended recall measure was not signi®cant. (See Table 3.) 3.2.6. Post-event processing and recall bias Pearson correlation coecients, computed to examine the association between the RQ and the time 2 dependent variables, revealed that rumination predicted open-ended recall of negative self-related information, negative bias in self-judgments at time 2 and time 2 recall of anxiety-related sensations on the BSQ. However, rumination did not predict recall on either scale of the structured recall measure. (See Table 3.) To examine whether post-event processing contributed to biases in recall, partial correlation coecients were computed between the rumination scores and the time 2 judgmental bias index and time 2 recall of anxiety-related sensations on the BSQ, controlling for time 1 scores. These correlations failed to reach signi®cance.
4. Discussion Our goal was to determine the extent to which three cognitive operations contributed to biased processing of social information. The results suggested that selective attention to negative self-related information during the interaction led to biased judgments about and recollections of the event, a pattern that was more pronounced among socially anxious than nonanxious subjects. Post-event processing also contributed to the tendency to recall negative self-related information. Overall, these results pointed to the pivotal role of selective attention, rather than selective retrieval, in social anxiety. They also provide an explanation for inconsistencies between previous studies of memory bias in socially anxious individuals. Consistent with previous studies, socially anxious individuals focused their attention more on themselves than on their partners during the social interaction and displayed larger negative biases in their self-related judgments than did control subjects (Rapee & Lim, 1992; Stopa & Clark, 1993; Alden & Wallace, 1995; Woody, 1996). As predicted, socially anxious subjects also displayed biases in their recollections of the interaction on the following day. Relative to control subjects, they recalled less partner-related information on the structured recall task and tended to recall more negative self-related information on the unstructured recall task, Table 3 Pearson correlation coecients between self-focused attention, rumination and the social judgment and memory measures. N = 116. Bias=judgmental bias index. BSQ=Body Sensations Questionnaire Time 1 bias
Self-focused attention Rumination
p < 0.05;
p < 0.01;
0.37
p < 0.001.
Time 2 bias
0.32 0.39
Open-ended recall
0.10 0.20
Factual recall partner
setting
ÿ0.26 ÿ0.07
ÿ0.11 0.02
Time 2 BSQ
0.71 0.58
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
255
although this latter ®nding was modest. Thus, relative to nonanxious individuals, socially anxious individuals' memories for social events displayed biases that appeared to favor negative self-related, as opposed to external, information. The unique contribution of this research was to establish that biases in social judgment and memory were related to self-focused attention and post-event-processing. Subjects who focused more on themselves displayed larger biases in their self-related judgments and recalled less information about their partner. This was true for both socially anxious and nonanxious subjects. However, because the socially anxious subjects directed more of their attention to selfrelated information, their judgments displayed more bias and their store of partner-related information was less complete than that of nonanxious subjects. It is possible that greater attention to self-related phenomena limited the availability of cognitive resources necessary to process partner-related information. The overall pattern of results is consistent with the hypothesis that self-monitoring increases the salience of anxiety-related behaviors at the expense of external information, so that this information weighs more heavily in judgments about and recollections of social events. As predicted, compared to nonanxious subjects, socially anxious subjects engaged in more post-event processing in the day following the interaction. One of the more notable ®ndings was that frequency of post-event rumination, not self-focused attention, predicted recall of negative self-related information on the open-ended memory task. Thus, it appeared that ruminative thinking in the period following the interaction may have increased the salience of negative self-related information so that this information was more likely to come to mind when subjects described the event. Frequency of ruminative thinking was also associated with a continuation of the biases in judgment and factual memory observed at time 1, but did not contribute further distortion to either of these measures. Taken together, these results suggested that post-event processing not only perpetuated existing cognitive biases, but may have contributed to a bias in recall that favored negative self-related information. There was no support for the notion that social anticipation activated selective retrieval of negative information about previous social events. No dierences emerged between socially anxious subjects who anticipated a second interaction and those who did not in terms of their recollections of anxiety-related sensations and behaviors or other types of information. This suggested that encoding bias, along with post-event processing, was responsible for any cognitive distortions in the socially anxious subjects' recollections of social events, not selective retrieval of negative information about the previous interaction. This pattern of results is consistent with writings arguing that selective attention plays a greater role than selective retrieval in anxiety states (e.g., Mathews & MacLeod, 1994). The results also suggested an explanation for inconsistencies among earlier memory studies. As noted previously, several studies found evidence of memory bias (e.g., Hope et al., 1990), whereas others did not (Rapee et al., 1994). To reconcile these ®ndings, it is important to consider the time at which bias enters the cognitive system and the complexity of the information being encoded. The Rapee studies explicitly focused on selective retrieval, whereas Hope and her colleagues made no attempt to distinguish between distortions that might occur as the information was being encoded versus distortions that might arise later as a result of selective retrieval. Several of Rapee's studies involved semantic memory tasks in which subjects read lists of words (Rapee et al., 1994). Hope, on the other hand, studied memory in the
256
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
context of a social interaction. It seems likely that the latter task presented subjects with complex, potentially competing pieces of social information and that subjects were forced to select, not necessarily consciously, which information to consider and which to ignore. The current study found evidence to support both of these positions. Like Hope and her colleagues, we found that socially anxious individuals were less able to recall partner-related information and like Rapee, we found no evidence for selective retrieval of negative social information once that information had been encoded. Taken together, these studies indicate that social interaction tasks may lend themselves to encoding biases in socially anxious individuals, whereas semantic memory tasks may not. Once social information is entered into memory, however, selective retrieval of negative information does not appear to further distort recollections of the social event. These results must be tempered with a reminder that the study had some limitations. First, subjects were selected on the basis of a single measure of social distress and avoidance, which may have included subjects whose social distress stemmed from problems other than social anxiety. However, one signi®cant source of social avoidance, depression, was measured and did not aect the results. Second, generalization to patients with social phobia remains to be established. These patients would be expected to experience higher levels of anxiety and this might result in more extensive rumination and higher levels of anticipatory anxiety. At higher levels, these processes might add more distortion to judgmental and recall biases than was observed here. It has also been suggested that social phobics may be less able to override maladaptive cognitive processes than nonclinically anxious individuals. Work examining patients with social phobia is currently underway. Acknowledgements This research was supported by an SSHRC grant to the second author and an UBC University Graduate Fellowship to the ®rst author. We appreciate the assistance provided by Michaela Gelin, Cameron Weynaus, and Kam Brar in conducting this research. We would also like to thank S.J. Rachman, Ph.D. and E. Eich, Ph.D. for their input. References Alden, L. E., & Wallace, S. T. (1995). Social phobia and social appraisal in successful and unsuccessful interactions. Behaviour Research and Therapy, 33, 497±506. Alden, L. E., Teshuck, M., & Tee, K. (1992). Public self-awareness and withdrawal from social interactions. Cognitive Therapy and Research, 16, 249±267. Beck, A. T., & Emery, G. (1985). In Anxiety disorders and phobias: a cognitive perspective (pp. 146±164). New York: Basic Books. Beck, A. T., Rush, A. J., Shaw, B. F., & Emery, G. (1979). Cognitive therapy of depression. New York: Guilford Press. Beck, A. T., Steer, & Garbin (1988). Psychometric properties of the BDI: Twenty-®ve years of evaluation. Clinical Psychology Review, 8, 77±100. Buss, A. H. (1980). Self-consciousness and social anxiety. San Francisco: Freeman & Company. Chambless, D. L., Caputo, G. C., Bright, P., & Gallagher, R. (1984). Assessment of fear of fear in agoraphobics:
T.M.B. Mellings, L.E. Alden / Behaviour Research and Therapy 38 (2000) 243±257
257
the body sensations questionnaire and the agoraphobic cognitions questionnaire. Journal of Consulting and Clinical Psychology, 52, 1090±1097. Clark, D. M., & Wells, A. (1995). A cognitive model of social phobia. In R. G. Heimberg, M. R. Liebowitz, D. A. Hope, & F. R. Schneier, Social phobia: diagnosis, assessment and treatment. New York: Guilford Press. Dodge, C. S., Hope, D. A., Heimberg, R. G., & Becker, R. E. (1988). Evaluation of the Social Interaction SelfStatement Test with a social phobic population. Cognitive Therapy and Research, 12(1), 211±222. Fairbrother, N., Rachman, S. J., & Mitchell, D. (1998, July). A new measure of social phobics' beliefs about what others think of them. Poster session presented at the World Congress of Behavioural and Cognitive Therapy, Acapulco, MX. Hope, D. A., Heimberg, R. G., & Klein, J. F. (1990). Social anxiety and the recall of interpersonal information. Journal of Cognitive Psychotherapy, 4, 185±195. Lucock, M. P., & Salkovskis, P. M. (1988). Cognitive factors in social anxiety and its treatment. Behaviour Research and Therapy, 26(4), 297±302. Lundh, L., & Ost, L. (1996). Recognition bias for critical faces in social phobics. Behaviour Research and Therapy, 34(10), 787±794. Mathews, A., & MacLeod, C. (1994). Cognitive approaches to emotion and emotional disorders. Annual Review of Psychology, 45, 25±50. McEwan, K. L., & Devins, G. M. (1983). Is increased arousal in social anxiety noticed by others? Journal of Abnormal Psychology, 92, 417±421. O'Banion, K., & Arkowitz, H. (1977). Social anxiety and selective memory for aective information about the self. Social Behavior and Personality, 5(2), 321±328. Rapee, R. M., & Lim, L. (1992). Discrepancy between self and observer ratings of performance in social phobics. Journal of Abnormal Psychology, 101, 728±731. Rapee, R. M., McCallum, S. L., Melville, L. F., Ravenscroft, H., & Rodney, J. M. (1994). Memory bias in social phobia. Behaviour Research and Therapy, 32(1), 89±99. Spielberger, C. D., Gorsuch, R. L., Lushene, R. E., Vagg, P. R., & Jacobs, G. A. (1983). Manual for the state-trait anxiety inventory. Palo Alto, CA: Consulting Psychologists Press. Stopa, L., & Clark, D. M. (1993). Cognitive processes in social phobia. Behaviour Research and Therapy, 31, 255± 267. Watson, D., & Friend, R. (1969). Measurement of social-evaluative anxiety. Journal of Consulting and Clinical Psychology, 33, 448±457. Woody, S. R. (1996). Eects of focus of attention on anxiety levels and social performance of individuals with social phobia. Journal of Abnormal Psychology, 105, 61±69. Woody, S. R., Chambless, D. L., & Glass, C. R. (1997). Self-focused attention in the treatment of social phobia. Behaviour Research and Therapy, 35(2), 117±129.