Suppressing and focusing on a negative memory in social anxiety: Effects on unwanted thoughts and mood

ARTICLE IN PRESS

Behaviour Research and Therapy 45 (2007) 2836–2849 www.elsevier.com/locate/brat

Suppressing and focusing on a negative memory in social anxiety: Effects on unwanted thoughts and mood Joshua C. Mageea,, Richard E. Zinbarga,b a

Northwestern University, Evanston, IL, USA The Family Institute at Northwestern University, Evanston, IL, USA

b

Received 27 June 2005; received in revised form 29 April 2007; accepted 8 May 2007

Abstract Researchers have hypothesized that thought suppression contributes to the large volume of unwanted thoughts in anxiety disorders. However, comparisons to both non-suppression and non-anxious groups are necessary for studies on thought suppression in high anxiety. Participants completed a series of thought verbalization periods and a social interaction. During one period, participants were randomly assigned to focus upon a negative social memory, suppress it, or think freely while monitoring the memory. Results indicated that thought suppression and focusing caused a greater rise and subsequent decline in unwanted thoughts than monitoring instructions for both high and low social anxiety groups. Importantly, highly socially anxious participants had more unwanted thoughts overall, but did not respond significantly differently to thinking instructions when compared to the less anxious group. Interestingly, highly socially anxious participants did report more thought suppression attempts in their everyday life. They also appeared to benefit by experiencing less shyness after suppression when compared to focusing, a pattern not evident for the low social anxiety group. r 2007 Elsevier Ltd. All rights reserved. Keywords: Thought suppression; Social anxiety; Intrusive thought; Social interaction

Introduction Researchers have recently become interested in the application of thought suppression models to the understanding of anxiety disorders. It has been well documented that anxious populations report an excessive number of unwanted thoughts and attempts to control these thoughts (Amir, Cashman, & Foa, 1997; Freeston & Ladoucer, 1997; Warda & Bryant, 1998). Consequently, it is important to determine whether anxious populations simply experience more unwanted thoughts, or whether they also address them with ineffective control strategies. In the latter case, unsuccessful control strategies may inadvertently play a causal role in producing further unwanted thoughts (Rassin, Merckelbach, & Muris, 2000).

Corresponding author. Department of Psychology, University of Virginia, P.O. Box 400400, Charlottesville, VA 22904-4400, USA.

E-mail address: [email protected] (J.C. Magee). 0005-7967/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.brat.2007.05.003

ARTICLE IN PRESS J.C. Magee, R.E. Zinbarg / Behaviour Research and Therapy 45 (2007) 2836–2849

2837

Thought suppression is one control strategy that has been frequently found to be counterproductive (Wegner, Schneider, Carter, & White, 1987; Wenzlaff & Wegner, 2000). During thought suppression, it is believed that individuals engage in an effortful search for distracters while simultaneously monitoring automatically for further thought instances (Wegner, 1994). According to this theoretical account, cognitive demands during suppression serve to disable the effortful search for distracters and allow the automatic monitoring process to enhance the accessibility of the thought. This enhancement of unwanted thoughts during suppression is known as an ‘‘initial enhancement’’ effect of suppression. An alternative method by which studies have identified ironic effects of suppression is by examining whether enhancement occurs in thought periods after the suppression period. In this case, the effortful search process ends, while the residual accessibility from the automatic monitoring process leads to an enhancement in thoughts. This enhancement is termed as ‘‘rebound effect’’ because unwanted thoughts rebound once a person no longer attempts to suppress them. Studies have supported both of these ideas, as researchers have found that persons attempting to suppress target thoughts often ironically end up thinking more about those thoughts when simultaneously encountering other cognitive demands (Wegner & Zanakos, 1994; for a review, see Wenzlaff & Wegner, 2000). Additionally, there seems to be a small to moderate rebound effect across studies (Abramowitz, Tolin, & Street, 2001). Wenzlaff and Wegner (2000) speculated that anxiety sufferers’ unwanted thoughts may be characterized by a ‘‘self-loading system’’ based on this model. They proposed that an anxious mood prevents successful suppression by drawing cognitive resources away from the effortful component of suppression attempts. Without the necessary cognitive resources, the individual is left vulnerable to unwanted thoughts primed by the automatic monitoring process described above. The added anxiety of experiencing thought suppression failure results in a feedback loop, increasing the ironic effects of attempted suppression in a self-perpetuating cycle. Social anxiety offers a useful avenue to test this model of thought suppression. It is relatively (a) common, (b) easy to manipulate, and (c) easy to observe in an ecologically valid manner even within an experimental setting (i.e., by arranging for a social interaction). Thus, social anxiety offers a chance to extend models of thought suppression beyond their most obvious applications to obsessive–compulsive disorder. As with other forms of anxiety (and depression), the experience of social anxiety often involves unwanted thoughts or imagery that would be likely targets for thought suppression. Leading cognitive–behavioral models have hypothesized that socially anxious people experience excessively negative self-imagery that encourages the biased processing of social situations (Clark & Wells, 1995), and several studies have supported this hypothesis experimentally (Hirsch, Clark, Mathews, & Williams, 2003; Hirsch, Meynen, & Clark, 2004). Thought suppression models suggest that attempting to suppress such negative self-imagery may have the unintended effect of increasing its salience under some conditions. Given that anxious populations frequently utilize strategies like thought suppression (Amir et al., 1997; Freeston & Ladoucer, 1997; Warda & Bryant, 1998), it is likely that socially anxious people frequently attempt to suppress the negative self-imagery they experience in social interactions. Interestingly, the limited research on thought suppression and social anxiety has suggested that thought suppression may function differently in social anxiety than in other disorders. In one study, researchers found possible indications that under non-cognitively demanding conditions social phobia may be characterized by a more wide-ranging impairment in thought suppression ability than other anxiety disorders (Fehm & Margraf, 2002). Complicating the picture, others have found that highly socially anxious participants may actually suppress more successfully when anticipating a social threat than when not anticipating a threat (Cougle, Smits, Powers, Lee, & Telch, 2005). The current study builds upon these intriguing suggestions by using nonanxious comparison groups as well as several types of non-suppression thinking instructions. We also employ sufficient cognitive demands as to ensure a test of the ‘‘self-loading system’’ proposed by Wenzlaff and Wegner (2000). In short, there is evidence that both anxiety group status and the use of thought suppression can independently contribute to the number of unwanted thoughts experienced (Becker, Rinck, Roth, & Margraf, 1998; Fehm & Margraf, 2002; Harvey & Bryant, 1998; Mathews & Milroy, 1994; Muris, de Jongh, Merckelbach, Postema, & Vet, 1998; Muris, Merckelbach, Horselenberg, Sijsenaar, & Leeuw, 1997), but there seems to be little support for an interaction between the two. A few researchers have found an

ARTICLE IN PRESS 2838

J.C. Magee, R.E. Zinbarg / Behaviour Research and Therapy 45 (2007) 2836–2849

interaction (i.e., Shipherd & Beck, 1999), but did not control for baseline differences in unwanted thoughts between control and anxiety disorder groups. In general, the literature shows that patients with anxiety disorders seem to have higher rates of unwanted thoughts than non-anxious controls whether they suppress or not, and suppression seems to lead to higher numbers of unwanted thoughts among many people in the general population. Unfortunately, research examining suppression in anxiety is difficult to conduct without encountering interpretive and/or methodological limitations. Perhaps the major methodological limitation is that many suppression studies have not included control groups of highly anxious participants following nonsuppression instructions. Many studies have only compared different groups’ performances under suppression instructions. Because highly anxious individuals generally show increased unwanted thoughts related to their central worries (Borkovec, Ray, & Sto¨ber, 1998; Mathews, 1990), it is likely that the mere instructions to engage in repeated monitoring of an anxious thought will lead to a high level of unwanted thoughts among highly anxious individuals. In this case, the enhanced accessibility of the target thought should be attributed to its monitoring, not the paradoxical effects of attempted suppression. Also, the reliance of suppression studies on self-report methods of recording thoughts is a necessary limitation. We selected a type of self-report (‘‘event marking’’) that has been recommended because it minimizes the influence of monitoring in that it is covert and relatively effortless (see Abramowitz et al., 2001). Evidence has shown that covert event marking can detect paradoxical effects while being less susceptible to experimental demand than other self-report measures (Abramowitz et al., 2001). Finally, we included monitoring control groups to serve as an experimental safeguard against the possibility of the event marking itself causing unwanted thoughts. While existing research has examined anxiety and thought suppression, few studies have considered the role of cognitive demands in the process or attempted to induce concurrent anxiety in participants. On account of these features, it is likely that previous studies have not incurred sufficient cognitive demands to fully test the ‘‘self-loading system’’ model. Therefore, in the current study we took two steps to ensure that sufficient cognitive demands were included. First, we included a cognitive load designed to induce ironic effects in suppression in all participants. Additionally, we included a task designed to be anxiety provoking for a highly socially anxious group. According to the ‘‘self-loading system’’ model, the highly socially anxious group should experience greater ironic effects than less anxious participants. The current study attempted to test socially anxious participants’ vulnerability to ironic effects of thought suppression. We also wanted to compare the effects of thought suppression to the effects of focusing upon a negative social memory on anxiety during an interaction, as such memories are hypothesized to be activated in anticipation of an interaction among highly socially anxious individuals (Clark & Wells, 1995). In this way, we hoped to distinguish the relative merits or costs of choosing suppression versus focusing as a strategy for addressing one’s unwanted thoughts given that many individuals may perceive suppression and focusing as their only two options when an unwanted thought/memory occurs. We tested the effects of thought suppression, focusing, and thought monitoring upon highly socially anxious participants within a paradigm designed to reduce their available cognitive resources. After we collected baseline mood measures, we told participants that they would undergo a videotaped social interaction with a stranger near the end of the experiment. Based on the existing literature using similar paradigms, we expected that placing highly socially anxious participants in a threatening, anticipatory situation combined with a cognitive load would produce the necessary demand on capacity-limited processing resources to test ironic suppression effects with a highly anxious mood (e.g., Mellings & Alden, 2000; Smith, Ingram, & Brehm, 1983). According to a model hypothesizing a general heightened vulnerability to the ironic effects of thought suppression in anxiety (a ‘‘suppression amplification’’ model), one would expect suppression to have extra consequences for participants with high social anxiety. These participants should experience more unwanted thoughts both during and after suppression than those with low social anxiety, and these differences should be greater than those observed at baseline. That is, an interaction between social anxiety and suppression instructions would be expected, with social anxiety amplifying the effects of suppression. On the other hand, highly socially anxious people may be distinguished not by their sensitivity to the ironic effects of suppression but rather simply in the extent to which they utilize suppression (a ‘‘suppression utilization’’ model). In this case, no interaction between social anxiety and suppression instructions would be expected. In other words, a higher number of unwanted thoughts would occur for the high social anxiety

ARTICLE IN PRESS J.C. Magee, R.E. Zinbarg / Behaviour Research and Therapy 45 (2007) 2836–2849

2839

group in both the suppression and control conditions, and the group differences during and after suppression would not exceed that observed in the control condition. Both the suppression amplification model and the suppression utilization model predict increased anxiety for the high social anxiety group compared to the low social anxiety group during the social interaction. To assess these hypotheses, the current study includes both self-report and behavioral observation ratings of anxiety during the social interaction. However, an interesting exploratory question concerns whether the suppression or focusing conditions cause more anxiety during the interaction than the control condition. While previous research has not addressed this possibility for social anxiety, it seems likely that increased frequency of negative social memories—either because one has been focusing on them or trying to suppress them and experiencing paradoxical enhancement or rebound effects—would cause increased anxiety. Method Participants The sample consisted of 58 participants (38 female) recruited through advertising on the Northwestern University campus. Participants were required to be between 18 and 28 years of age and proficient in English. Participants were paid $5 to fill out a screening questionnaire—the Brief Version of the Fear of Negative Evaluation Scale (BFNE; Leary, 1983a). In order to recruit groups with high and low social anxiety, we screened 200 possible participants and invited 70 participants to return for the actual experiment, 35 from both the upper and lower 20% of the screening pool. The mean score of the highly socially anxious group was 49.4 (SD ¼ 5.0) while that of the low group was 23.2 (SD ¼ 4.7). The mean score for the highly socially anxious group was comparable to the mean of 51.5 (SD ¼ 7.3) obtained in the only study reporting BFNE scores for a sample of patients with social phobia (Collins, Westra, Dozois, & Stewart, 2005). Due to the requirement of a social interaction during the experiment, participants were scheduled in pairs to complete the study. Participants did not know of the pairing until the time of the experiment. Of the 70 we invited to return, four people did not show up, resulting in the loss of data from eight participants (including the experimental partner of each participant that did not show up). We discarded two participants’ results because they did not follow instructions during the experiment. Finally, we excluded another two participants’ data because it was discovered after their social interaction that they had known each other prior to the experiment. Materials The BFNE (Leary, 1983a) is a 12-item scale measuring social-evaluative anxiety. It has satisfactory test–retest reliability, good internal consistency, and demonstrates good convergent validity with other measures of social-evaluative anxiety (Leary 1983a, b). The BFNE has also been shown to discriminate successfully between a sample with social phobia and one with panic disorder (Collins et al., 2005). The Positive and Negative Affect Schedule—Expanded Form (PANAS-X; Watson & Clark, 1994) Fear and Shyness Scale is a 60-item scale measuring higher-order positive and negative affect. In addition to the two higher-order dimensions, it contains 11 specific subscales measuring different affective states. We administered only the fear (‘‘afraid’’, ‘‘scared’’, ‘‘frightened’’, ‘‘nervous’’, ‘‘jittery’’, and ‘‘shaky’’) and shyness (‘‘shy’’, ‘‘bashful’’, ‘‘sheepish’’, and ‘‘timid’’) subscales of the measure. Both scales exhibited adequate reliability, and convergence with peer ratings of affect (Watson & Clark, 1994). The 6-item fear subscale has been shown to have adequate convergent validity with the Hopkins Symptom Checklist-Anxiety Scale and the State-Trait Anxiety Inventory-State Anxiety Scale (Watson & Clark, 1994). The 4-item shyness subscale was also judged to be relevant to state anxiety within an experiment employing social threat. The White Bear Suppression Inventory (WBSI; Wegner & Zanakos, 1994) is a 15-item scale that measures individuals’ tendencies to experience and suppress unwanted thoughts in general situations (Muris, Merckelbach, & Horselenberg, 1996). It has demonstrated satisfactory internal consistency and test–retest reliability.

ARTICLE IN PRESS 2840

J.C. Magee, R.E. Zinbarg / Behaviour Research and Therapy 45 (2007) 2836–2849

The State-Trait Anxiety Inventory-Trait Anxiety scale (STAI-TA; Spielberger, Gorsuch, Lushene, Vagg, & Jacobs, 1983) is a widely used scale measuring general, long-standing anxiety. The 20-item scale has good psychometric properties and was used in the current study to measure trait anxiety. The Global Rumination Scale (GRS; McIntosh & Martin, 1992) is a 10-item scale gauging individuals’ tendency to ruminate in general. The psychometrics of this measure have not been studied extensively, but we included it for an exploratory look at the relationships among thought suppression, ruminative thought, and social anxiety. During thought periods, participants verbalized into a small tape recorder with a microphone cord attachment. Participants also monitored unwanted thoughts of a negative social interaction memory with marks on a blank piece of paper. For the social interaction, we videotaped conversations openly using cameras directly opposite each participant. Objective coding of behavioral observation of anxiety: Coders rated videotapes of social interactions according to a system developed by Miller and Zinbarg (2003), which was based on the work of Glass and Furlong (1990). We instructed coders first to make ratings of individual behaviors such as talk time, gesturing, eye contact, and smiling for each minute of the 5-min interaction. Once raters completed this initial scoring, they reviewed the tape to record a global anxiety score based upon their general impression of the participant’s anxiety level. We used these global anxiety ratings for the analysis. Procedure See Fig. 1 for a diagram of the experimental design. Upon entering the laboratory, participants completed the PANAS-X for a baseline mood measure. After the baseline PANAS-X measure, we briefed participants that they would be undergoing four thought verbalization periods lasting 5 min each as well as a videotaped 5min social interaction between the third and fourth periods. After these initial instructions, participants were asked to recall a past negative social interaction with a stranger. We gave participants a brief form in which they spent 5 min writing details for three sections: (1) ‘‘Describe the situation: the participants and a basic summary of the interaction.’’ (2) ‘‘What feelings and sensations did you experience during the negative interaction?’’ (3) ‘‘How do you think others perceived you during the interaction? In other words, how would you have looked to the other person/people in the interaction or to any outside observer?’’. We chose the negative social interaction memory as a target suppression thought due to suggestions by previous researchers. Specifically, it was a ‘‘non-discrete’’ target thought that did not refer to one specific image (i.e., a white bear) but to a broader memory (Abramowitz et al., 2001). Additionally, the memory concerned social evaluation, a topic of central concern to socially anxious individuals. We expected that a negative, personally relevant target thought would reasonably parallel the perception of targets in the majority of real-life suppression attempts. After the negative memory recall, participants began the first thought verbalization periods. Participants completed thought periods individually in enclosed rooms intended to allow a sense of privacy. We asked them to speak aloud into a tape recorder whatever thoughts ran through their mind, no matter what the content or frequency of certain thoughts was. These verbalization instructions were used following the suggestions of Pope (1978) for recording the stream of consciousness. We also asked participants to make a tally mark on a blank piece of paper each time a thought about the negative social interaction memory entered their mind. This ‘‘event marking’’ was used for the dependent measure. During the first thought verbalization period, all participants were given control ‘‘monitoring’’ instructions (Muris, Merckelbach, van den Hout, & de Jong, 1992). To measure baseline unwanted thought rates related to the simple mention and monitoring of the target thought, participants were instructed to ‘‘think about whatever you would like; you can think about that negative social interaction you wrote about or anything else you would like.’’ If participants asked for further clarification, the free nature of the period was stressed, while still referencing the target thought within each clarification. The first period was designed to measure baseline rates in all groups, which could be compared to post-manipulation rates. During the second verbalization period, the high and low social anxiety groups were randomly assigned to one of three thinking instructions (N ¼ 19 for monitoring, N ¼ 20 for focusing, and N ¼ 19 for suppression).

ARTICLE IN PRESS J.C. Magee, R.E. Zinbarg / Behaviour Research and Therapy 45 (2007) 2836–2849

2841

Mood Measure (PANAS-X)

Recall of Negative Social Interaction Memory

Thinking Period 1 Monitor

Thinking Period 2 (Cognitive Load) Focus / Suppress / Monitor

Thinking Period 3 Monitor

Social Interaction

Mood Measure (PANAS -X)

Thinking Period 4 Monitor

Questionnaires Fig. 1. Diagram of experimental design.

For the monitoring instructions, we gave participants the same monitoring instructions as baseline to guard against any possible confounds introduced by the level of self-focus or other factors during the free-thought verbalization periods. Researchers have shown that self-focus may contribute to increased anticipatory social anxiety in both socially anxious participants and controls (Woody, 1996; Woody & Rodriguez, 2000). These control instructions served as a method to compare the effects of suppression and focusing to those caused solely by the nature of the experiment. For the focusing instructions, participants were asked to ‘‘please focus on the negative social interaction memory. Try to direct your attention toward the negative memory as much as possible.’’ These instructions sought to maximize activation of the target thought to examine the effects of focusing (expressing the thought out loud) as compared to thought suppression. In the suppression condition, participants were instructed to attempt suppression of the target thought: ‘‘think about whatever you would like but please try not to think about the negative social interaction memory.’’ To ensure that both low and high social anxiety groups experienced sufficient cognitive demand to reveal individual differences in thought suppression difficulty, we asked participants to remember a random five-digit number during the second thinking period. Wegner, Erber, and Zanakos (1993) conducted a suppression study of remembering a six-digit number versus a two-digit number while suppressing sufficiently created differences

ARTICLE IN PRESS J.C. Magee, R.E. Zinbarg / Behaviour Research and Therapy 45 (2007) 2836–2849

2842

in suppression ability. By adding the demand of remembering a five-digit number to our paradigm, we hypothesized that both groups would experience sufficient cognitive demand to experience ironic effects of suppression. Subsequently, we believed that any differences in thought suppression ability due to the cognitive demands of social anxiety would emerge, given the sufficient cognitive demands placed on all participants. During the third and fourth thought verbalization periods, all participants were once again given the same monitoring instructions used in the first period. In this way, we could monitor the longer-term effects due to the period two thinking instructions (focusing vs. suppression vs. monitoring). After the third period, each participant underwent a 5-min social interaction with his or her paired participant. We invited participants into a room in which two chairs and two video cameras were set up; each video camera was positioned directly in one participant’s line of sight. All interactions were recorded on videotape for later scoring of anxiety by four independent coders. When participants finished the interaction, they moved back to their original rooms and completed a second PANAS-X form. This measure asked participants to gauge their mood during the social interaction and allowed for examination of any changes from the original baseline PANAS-X measure. After finishing the PANAS-X, participants underwent the fourth verbalization period, following monitoring instructions once again. This period was included to test whether the effects of suppression and focusing continued during the fourth period following the completion of the interaction task, or whether they were solely driven by the anticipation of a social task. Finally, participants completed a battery of personality measures, including the WBSI, the STAI-TA, and the GRS. Results Preliminary analyses There were no significant differences among randomly assigned instruction groups on measures of repetitive thought (F(2,54) ¼ 1.31, p4.10), thought suppression (F(2,55)o1, p4.10), trait anxiety (F(2,55)o1, p4.10), baseline shyness (F(2,55)o1, p4.10), or baseline fear (F(2,55)o1, p4.10). See Table 1 for means and standard deviations; baseline shyness and fear are reported in Table 2. Additionally, the groups did not differ by gender (w2(2, N ¼ 58) ¼ 2.14, p4.10), indicating that randomization was successful. Consistent with previous research, highly socially anxious participants reported more repetitive thought than the low social anxiety group (GRS; F(1,55) ¼ 8.90, po.01), more frequent suppression attempts in their everyday life (WBSI; F(1,56) ¼ 17.11, po.001), and more general, long-standing anxiety (STAI-TA; F(1,56) ¼ 22.29, po.001). Differences in baseline mood between the social anxiety groups were tested with one-way ANOVAs, using the PANAS-X shyness and fear subscales as dependent variables. The high social anxiety group demonstrated significantly more shyness (F(1,56) ¼ 7.59, po.01) and fear (F(1,56) ¼ 6.25, po.02) than the low social anxiety group at baseline.

Table 1 Means and standard deviations for preliminary analyses Full sample

Social anxiety group

(N ¼ 58)

High

Low

Thinking instructions Focus (N ¼ 10) WBSI GRS STAI-TA Female

50.61 (10.65) 50.65 (9.95) 32.54 (12.13) 66%

a

55.70 (7.06) 55.80a (5.31) 39.20a (10.28) 70%

Suppress (N ¼ 9) a

56.56 (6.06) 52.78a (10.27) 39.78a (14.67) 78%

Monitor (N ¼ 9) a

54.89 (9.13) 54.44a (6.71) 38.33a (11.17) 67%

Focus (N ¼ 10) b

46.00 (13.59) 47.70b (11.77) 26.00b (6.88) 70%

Suppress (N ¼ 10) b

46.33 (10.56) 42.33b (13.50) 26.94b (10.98) 30%

Monitor (N ¼ 10) 44.80b (10.14) 50.60b (5.08) 25.70b (9.30) 80%

Note: Means are presented first with standard deviations in parentheses. Differences are noted by unique letter superscripts (i.e., ‘‘a’’ vs. ‘‘b’’) and are all significant at po.01 or below. WBSI refers to the White Bear Suppression Inventory; GRS refers to Global Rumination Scale; STAI-TA refers to the State-Trait Anxiety Inventory, Trait Version; Female refers to the percentage of female participants.

ARTICLE IN PRESS J.C. Magee, R.E. Zinbarg / Behaviour Research and Therapy 45 (2007) 2836–2849

2843

Table 2 Mean PANAS-X shyness and fear subscale scores as a function of social anxiety group and thought instructions Thought instructions

Social anxiety group

Baseline shyness

Shyness during the social interaction

Baseline fear

Fear during the social interaction

Focusing

High (N ¼ 10) Low (N ¼ 10)

4.70 (2.06) 5.00 (2.49)

7.10 (4.20) 5.10 (2.60)

10.30 (3.86) 8.90 (2.23)

13.90 (7.34) 10.70 (4.81)

Suppression

High (N ¼ 9) Low (N ¼ 10)

6.78 (3.15) 3.90 (1.10)

6.00 (2.78) 4.50 (1.78)

11.22 (4.12) 8.80 (1.55)

10.70 (4.81) 10.20 (1.75)

Control

High (N ¼ 9) Low (N ¼ 10)

6.89 (1.83) 4.50 (1.78)

6.33 (2.69) 4.70 (1.57)

12.22 (3.87) 9.70 (3.06)

13.11 (3.10) 10.00 (2.49)

5.24 (2.34)

5.60 (2.78)

10.14 (3.31)

11.40

Total

Note: Means are presented first with standard deviations in parentheses. PANAS-X refers to the Positive and Negative Affect Schedule— Expanded Form.

A w2 test indicated that the gender distribution did not differ between the social anxiety groups p4.10). To test whether sex differences interacted with social anxiety group and/or the three conditions of the experiment, we included participant sex as a third between-subjects factor along with instructions and social anxiety group. No interaction effects were found, so we conducted the remainder of analyses without participant sex as a factor.

(w2(1, N ¼ 58)o1,

Effects of instructions and social anxiety on unwanted thoughts To test the main hypotheses of the experiment, we used a 2 (social anxiety group)  3 (instruction)  4 (period) ANOVA.1 Social anxiety group and instructions were between-group variables, while period was a repeated measure that we analyzed via orthogonal polynomial analysis of trend. This analysis allowed us to examine the changes in unwanted thought rates over time according to instruction condition and anxiety group. We did not expect differences between the instruction conditions in the baseline period (before participants undertook differential instructions), but we did expect differences in the patterns of growth and decay thereafter. We expected the focusing and suppression conditions to show greater increases from period 1 to period 2 than the control condition, but then expected the groups to begin returning to baseline after they were no longer asked to focus or suppress (consistent with a quadratic trend). The results are presented in Fig. 2. For the 2 (social anxiety group)  3 (instruction)  4 (period) ANOVA, the three-way interaction was not significant for either the linear (Z2 ¼ .008) or quadratic trend (Z2 ¼ .0008) of period (both F(2,52)o1). The two-way interaction of instruction condition by the quadratic trend of period was significant (F(2,52) ¼ 3.52, po.05, Z2 ¼ .03). None of the other two-way interactions were significant (all F’so1). To break down the instruction condition by the quadratic trend of period interaction further, we selected one instruction condition at a time, and examined each condition’s quadratic trend over a period. We expected that the focusing and suppression conditions would demonstrate quadratic relationships, while the control condition would not. As expected, the quadratic trend was significant for the focusing condition (F(1,18) ¼ 8.48, po.01, Z2 ¼ .12) and for the suppression condition (F(1,17) ¼ 6.05, po.05, Z2 ¼ .04). Finally, the quadratic trend was not significant for the control condition (F(1,17)o1, po.10). Next, we set up two planned contrasts designed to compare the quadratic trends for the instruction conditions. According to our hypotheses, we expected the focusing and suppression conditions to show greater increases and subsequent declines when compared to the control condition. A contrast comparing the combined focusing and suppressing conditions with the control condition was significant (F(1,54) ¼ 6.04, 1 One participant’s unwanted thought frequencies for the first two periods qualified as extreme outliers based on Tukey’s procedures as used in SPSS. Each outlier was more than three standard deviations above the mean for the period. While the outlier did not affect any of the results, we report the results excluding the outlier.

ARTICLE IN PRESS 2844

J.C. Magee, R.E. Zinbarg / Behaviour Research and Therapy 45 (2007) 2836–2849

Mean Unwanted Thoughts

6 Focusing

5

Suppression

Control

4 3 2 1 0

1

2 3 High Social Anxiety

4

Mean Unwanted Thoughts

6 Focusing

5

Suppression

Control

4 3 2 1 0

1

2 3 Low Social Anxiety

4

Fig. 2. Mean unwanted thoughts of a negative social interaction memory as a function of period for the (a) high social anxiety group and (b) the low social anxiety group.

po.05, Z2 ¼ .02), supporting the hypothesis. For the second contrast, we compared the quadratic trends for the focusing and suppression conditions. We intended for this contrast to be a follow-up to the first contrast. If the focusing and suppression conditions were demonstrated to differ in their pattern of growth and decay, then the result of the first contrast would be less meaningful. Importantly, in the second contrast the two conditions did not significantly differ (F(1,35)o1, p4.10). Thus, it appears that the focusing and suppression instructions caused a greater growth and decline in unwanted thoughts when compared to the control condition. After finding that the social anxiety group did not significantly interact with instructions or period, we wanted to examine whether the social anxiety groups differed in their number of unwanted thoughts over the four periods. Indeed, the between-subjects main effect of social anxiety group was significant (F(1,52) ¼ 4.84, po.05, Z2 ¼ .03). The means indicated that the high social anxiety group (M ¼ 3.02, SD ¼ 2.32) had a higher average number of unwanted thoughts over the four periods than the low social anxiety group (M ¼ 1.98, SD ¼ 1.25). Effects of instructions and social anxiety on state anxiety To examine state anxiety changes, we entered baseline and post-social interaction PANAS-X shyness scores as within-subjects variables, and social anxiety group and instruction conditions as between-subject variables in a repeated measures ANOVA. Next, we ran a similar analysis except with the fear subscale of PANAS-X in place of the shyness scores. Table 2 shows the shyness and fear scores associated with each instruction condition and social anxiety group. For shyness, there was a three-way interaction for the 2 (social anxiety group)  3 (instructions)  2 (measurement occasion) ANOVA (F(2,52) ¼ 4.13, po.03, Z2 ¼ .12). Inspection of the cells in Table 2 revealed that for highly socially anxious participants, the focusing instructions were associated with a larger increase in shyness than the two other conditions. Conversely, differing instructions did not seem to cause any divergences for the participants with low anxiety. To test this interpretation, we ran 2 (measurement occasion)  3 (instructions) ANOVAs separately for the low social anxiety group and the high social anxiety

ARTICLE IN PRESS J.C. Magee, R.E. Zinbarg / Behaviour Research and Therapy 45 (2007) 2836–2849

2845

Table 3 Standardized mean global anxiety ratings by objective observers of participants’ social interactions as a function of social anxiety group and instructions Thought instructions

Social anxiety group

Global anxiety ratings

Focusing

High (N ¼ 5) Low (N ¼ 8)

.26 (.99) .36 (.73)

Suppression

High (N ¼ 7) Low (N ¼ 8)

.16 (.74) .33 (.53)

Control

High (N ¼ 6) Low (N ¼ 9)

.16 (.74) .40 (.58)

Note: Means are presented first with standard deviations in parentheses.

group. As expected, the simple two-way interaction was significant for the high social anxiety group (F(2,25) ¼ 4.21, po.03, Z2 ¼ .25) but not for the low social anxiety group (F(2,27)o1). Next we computed shyness difference scores for the high social anxiety group, and compared the instruction groups using LSD post hoc tests. Post hoc tests demonstrated that the focusing group significantly differed from the control (po.03) and suppression (po.02) groups. The suppression and control groups were not significantly different. Thus, it seems that the larger increase in shyness for the high anxiety participants in the focusing condition was largely responsible for the interaction effect. The only other significant effect was the main effect of social anxiety (F(1,52) ¼ 8.35, po.01, Z2 ¼ .02), indicating that the high social anxiety group had higher average shyness scores. For fear scores, the three-way interaction for the 2 (social anxiety group)  3 (instructions)  2 (measurement occasion) ANOVA was not significant (F(2,52) ¼ 1.07, p4.10). The only significant effects were that fear rose significantly between measurements (F(1,52) ¼ 4.88, po.04, Z2 ¼ .08), and that the high social anxiety group reported more fear averaged across the two measurements and the three instruction conditions (F(1,52) ¼ 7.11, po.01, Z2 ¼ .009). Effects of instructions and social anxiety on behavior We used the global anxiety ratings of four coders for analysis of the videotaped social interactions. Unfortunately, due to the timing of the experiment the coders were only available to rate 43 out of the 58 participants.2 Codes were standardized within coders to remove variability across coders in their means and standard deviations prior to averaging scores across coders. Inter-rater reliability for coding of global anxiety was good (intraclass correlation ¼ .81) after standardization. Standardized means for global anxiety are presented in Table 3. For the social interaction, there were significant correlations between observer video ratings of anxiety and the self-reports of fear (r ¼ .43, po.01) and shyness (r ¼ .55, po.001) made immediately following the interaction. Looking at the changes from baseline to post-social interaction self-ratings, the correlation between shyness change and coder rating was r ¼ .33, po.03, and between fear change and coder rating was r ¼ .23, p4.10. We analyzed coder ratings of global anxiety with a 2 (social anxiety group)  3 (instruction) ANOVA. None of the effects were significant (all p’s4.10). Discussion The results of the current study suggest that thought suppression is important for understanding social anxiety, but provide more support for what we have called a suppression utilization model than what we have 2

The coders did not complete coding prior to the end of an academic quarter, and ceased participation in the laboratory after this session.

ARTICLE IN PRESS 2846

J.C. Magee, R.E. Zinbarg / Behaviour Research and Therapy 45 (2007) 2836–2849

called a suppression enhancement model. Importantly, the results reported here may also provide some insight regarding why people tend to choose suppression despite its paradoxical effects on unwanted thoughts. For the instruction manipulation, both the focusing and suppression instructions caused an initial rise in and subsequent decline of unwanted thoughts that was not evident for the monitoring instructions. However, the high social anxiety group did not show a significantly different pattern from the low social anxiety group in terms of instruction group differences over periods. In fact, a close inspection of Fig. 2 reveals a nearly constant differential between social anxiety groups as unwanted thought rates grew and decayed. In other words, high social anxiety and thought instructions each had independent but additive effects. This finding was largely consistent with the existing literature on thought suppression and anxiety, and fails to support previous suggestions that social anxiety may be characterized by general deficits in thought suppression ability or a ‘‘self-loading system.’’ At the same time, the alternative framework supported by the results still contends that thought suppression may be a potentially important contributor to social anxiety. However, it is through the increased use of thought suppression among the socially anxious rather than through an enhancement of paradoxical effects that the contributions occur. Specifically, this study found that highly socially anxious participants report more suppression attempts in everyday life than people with lower levels of social anxiety. When combined with the fact that suppression was generally unsuccessful for all participants instructed to undertake it, this implies that highly socially anxious people are at risk for more frequently incurring ironic effects of suppression in everyday life than people with low levels of social anxiety. One possibility may be that highly socially anxious people choose thought suppression for the same proportion of the time as low social anxiety groups. Yet, because highly socially anxious people experience more unwanted thoughts, thought suppression would be applied more frequently as a byproduct of these more common unwanted thoughts. Highly socially anxious people would therefore incur more numerous instances of ironic effects on account of their original problems with unwanted thoughts. Another perhaps more likely scenario is that highly socially anxious people tend to choose thought suppression for a higher percentage of their intrusive thoughts than less anxious people. In this case, they would be incurring complications above and beyond a person with low social anxiety faced with the same initial levels of intrusive thoughts. This idea is similar to one considered by Amir et al. (1997), who found that patients with OCD are more likely to utilize maladaptive thought management strategies than non-patients. Future research could seek to test the suggestion that this tendency reflects social anxiety as well. One interesting dilemma concerns why people tend to choose suppression despite its apparent long-term detriments (Trinder & Salkovskis, 1994). The results of the state anxiety analyses appeared to offer some intriguing hints at why people with high social anxiety may initially benefit from thought suppression. For participants in the high social anxiety group, focusing was associated with a larger increase in self-reported shyness than suppression (if anything, suppression was associated with a reduction in self-reported shyness). Whereas this effect did not emerge in the coders’ ratings of anxious behavior, these ratings were significantly correlated with self-reported shyness. Thus, the self-reported shyness scores have some validity and cannot be entirely attributed to reporting biases. It is possible that this immediate benefit reinforces suppression despite the longer-term negative consequence of increasing unwanted thoughts. Reinforcement would be especially likely to the extent that an individual perceives focusing or suppression as one’s only two options for responding to an unwanted thought. Interestingly, the same benefit was not evident for the low social anxiety group, as neither focusing nor suppression appeared to produce meaningful increases in shyness in this group. This pattern of findings leads us to predict that highly anxious individuals should indeed choose to suppress a greater percentage of unwanted thoughts than less anxious people. One consideration that the results rule out is the idea that the repeated conscious monitoring of a thought might entirely account for the hyperaccessibility that is normally attributed to automatic monitoring processes of thought suppression in Wegner’s and others’ theories. Overall, the suppression groups consistently experienced more thoughts than the group who was simply repeatedly monitored throughout the study. However, it is possible that conscious monitoring of thoughts may serve to amplify the automatic monitoring process proposed by Wegner to be behind the ironic effects of suppression. In this case, one would predict that participants suppressing and consciously monitoring their thoughts would show a larger ironic effect than

ARTICLE IN PRESS J.C. Magee, R.E. Zinbarg / Behaviour Research and Therapy 45 (2007) 2836–2849

2847

those solely suppressing. Measures such as the Stroop, stem-completion, or lexical decisions tasks (e.g., Tolin, Abramowitz, Przeworski, & Foa, 2002; Wegner & Erber, 1992) that do not necessitate conscious monitoring by participants but that might index accessibility or activation of suppressed representations may offer ways to test this open question. Limitations and conclusion There are a few limitations of the current study that should be noted when interpreting the results. First, the test of the interaction between social anxiety group and thinking instructions should be interpreted with caution due to small sample sizes. The analysis yielded very small effect size point estimates, but it is still possible that the range of possible population effect sizes could have included a meaningful effect. Second, the current high social anxiety sample was not diagnosed with social phobia. While we are confident that the characteristics of this sample were similar to social phobia, it is nonetheless possible that qualitative differences of social phobia might have led to different results. Additionally, the current experiment was not in a position to tease apart the effects of social anxiety versus the interaction of social anxiety with social threat. However, in a practical sense we believe that the social interaction paralleled socially anxious people’s real-life experiences and thus has important implications for understanding social anxiety. It is also possible that there may have been differences between the social anxiety groups in their choice of cognitive processing during the free thought periods (Hinrichsen & Clark, 2003). Although the results suggest that the instructions had immediate effects in the direction one would expect, an unknown proportion of participants may have been spontaneously engaging in suppression or focusing during the other periods. Asking participants to engage in repeated suppression or focusing may have increased confidence in the results. There may have been a different pattern of results if the social interaction had occurred after the fourth thought period instead of before it. There may have been less fear or shyness reduction in the final period if participants were still anticipating a social interaction. Patterns of unwanted thoughts among all social anxiety groups continued along similar trajectories throughout the experiment, but it is unknown whether differences among the groups may have emerged with continued anticipation. Finally, participants were not given detailed instructions about how to demarcate their thoughts and may have used different criteria for determining what qualified as a thought. However, this should have been randomly assigned across the thinking instructions, and we believe that the most likely effect would have been to add to our measurement error and lower power. Thus, the effects that were significant—the main effect of group and the instruction condition by quadratic trend of period interaction—were powerful enough to be detectable despite this noise. Future researchers should consider the influence of participants’ standards for demarcating thoughts when designing thought suppression experiments. The investigation of thought suppression appears promising for social anxiety research. In this study, socially anxious participants reported using thought suppression more frequently in their everyday lives than participants with low levels of social anxiety. Given the current evidence that thought suppression had paradoxical effects independent of anxiety level, it is clear that the choice of thought suppression as a management strategy does matter for people with social anxiety. The current results also suggest that thought suppression may be reinforced among the socially anxious by avoidance of the increase in shyness these people tend to experience when interacting after focusing on a negative social memory. Future research is needed to disentangle the ways in which different thought control strategies influence thoughts, emotions, and behavior. Acknowledgments We thank Emily Durbin, Ph.D., for her careful attention and helpful comments during revision of this manuscript. Also, we would like to thank Elizabeth Miller for developing the behavioral observation of the anxiety coding system used in the experiment.

ARTICLE IN PRESS 2848

J.C. Magee, R.E. Zinbarg / Behaviour Research and Therapy 45 (2007) 2836–2849

References Abramowitz, J., Tolin, D., & Street, G. (2001). Paradoxical effects of thought suppression: A meta-analysis of controlled studies. Clinical Psychology Review, 5, 683–703. Amir, N., Cashman, L., & Foa, E. B. (1997). Strategies of thought control in obsessive–compulsive disorder. Behaviour Research and Therapy, 35, 775–777. Becker, E., Rinck, M., Roth, W., & Margraf, J. (1998). Don’t worry and beware of white bears: Thought suppression in anxiety patients. Journal of Anxiety Disorders, 12, 39–55. Borkovec, T. D., Ray, W. J., & Sto¨ber, J. (1998). Worry: A cognitive phenomenon intimately linked to affective, physiological, and interpersonal behavioral processes. Cognitive Therapy and Research, 22, 561–576. Clark, D. M., & Wells, A. (1995). A cognitive model of social phobia. In R. G. Heimberg, M. Liebowitz, D. Hope, & F. Schneier (Eds.), Social phobia: Diagnosis, assessment, and treatment (pp. 69–93). New York: Guilford Press. Collins, K. A., Westra, H. A., Dozois, D. J. A., & Stewart, S. H. (2005). The validity of the brief version of the Fear of Negative Evaluation Scale. Journal of Anxiety Disorders, 19, 345–359. Cougle, J. R., Smits, J. A. J., Powers, M. B., Lee, H., & Telch, M. J. (2005). Singular and combined effects of thought suppression instructions and anxiety induction on frequency of threatening thoughts: An experimental investigation. Cognitive Therapy and Research, 29, 525–539. Fehm, L., & Margraf, J. (2002). Thought suppression: Specificity in agoraphobia versus broad impairment in social phobia? Behaviour Research and Therapy, 40, 57–66. Freeston, M., & Ladoucer, R. (1997). What do patients do with their obsessive thoughts? Behaviour Research and Therapy, 35, 335–348. Glass, C. R., & Furlong, M. (1990). Cognitive assessment of social anxiety: Affective and behavioral correlates. Cognitive Therapy and Research, 14, 365–384. Harvey, A. G., & Bryant, R. A. (1998). The effect of attempted thought suppression in acute stress disorder. Behaviour Research and Therapy, 36, 583–590. Hinrichsen, H., & Clark, D. M. (2003). Anticipatory processing in social anxiety: Two pilot studies. Journal of Behavior Therapy and Experimental Psychiatry, 34, 205–218. Hirsch, C. R., Clark, D. M., Mathews, A., & Williams, R. (2003). Self-images play a causal role in social phobia. Behaviour Research and Therapy, 38, 1183–1192. Hirsch, C. R., Meynen, T., & Clark, D. M. (2004). Negative self-imagery in social anxiety contaminates social interactions. Memory, 12, 496–506. Leary, M. R. (1983a). A brief version of the fear of negative evaluation scale. Personality and Social Psychology Bulletin, 9, 371–375. Leary, M. R. (1983b). Social anxiousness: The construct and its measurement. Journal of Personality Assessment, 47, 66–75. Mathews, A. (1990). Why worry? The cognitive function of anxiety. Behaviour Research and Therapy, 28, 455–468. Mathews, A., & Milroy, R. (1994). Effects of priming and suppression of worry. Behaviour Research and Therapy, 32, 843–850. McIntosh, W. D., & Martin, L. L. (1992). The cybernetics of happiness: The relation of goal attainment, rumination and affect. Review of Personality and Social Psychology, 14, 222–246. Mellings, T. M. B., & Alden, L. E. (2000). Cognitive processes in social anxiety: The effects of self focus, rumination and anticipatory processing. Behaviour Research and Therapy, 38, 243–257. Miller, E., & Zinbarg, R. (2003). Anticipatory processing in social anxiety: The effects of memory and rumination on anxiety during social interaction. Unpublished manuscript, Northwestern University. Muris, P., de Jongh, A., Merckelbach, H., Postema, S., & Vet, M. (1998). Thought suppression in phobic and non-phobic dental patients. Anxiety, Stress, and Coping, 11, 275–287. Muris, P., Merckelbach, H., & Horselenberg, R. (1996). Individual differences in thought suppression. The white bear suppression inventory: Factor structure, reliability, validity, and correlates. Behaviour Research and Therapy, 34, 501–513. Muris, P., Merckelbach, H., Horselenberg, R., Sijsenaar, M., & Leeuw, I. (1997). Thought suppression in spider phobia. Behaviour Research and Therapy, 35, 769–774. Muris, P., Merckelbach, H., van den Hout, M., & de Jong, P. (1992). Suppression of emotional and neutral material. Behaviour Research and Therapy, 30, 639–642. Pope, K. S. (1978). How gender, solitude, and posture influence the stream of consciousness. In K. S. Pope, & J. L. Singer (Eds.), The stream of consciousness (pp. 259–299). New York: Plenum Press. Rassin, E., Merckelbach, H., & Muris, P. (2000). Paradoxical and less paradoxical effects of thought suppression: A critical review. Clinical Psychology Review, 20, 973–995. Shipherd, J. C., & Beck, J. G. (1999). The effects of suppressing trauma-related thoughts on women with rape-related posttraumatic stress disorder. Behaviour Research and Therapy, 37, 99–112. Smith, T. W., Ingram, R. E., & Brehm, S. S. (1983). Social anxiety, anxious self-preoccupation, and recall of self-relevant information. Journal of Personality and Social Psychology, 44, 1276–1283. Spielberger, C. D., Gorsuch, R. L., Lushene, R. E., Vagg, R. E., & Jacobs, G. A. (1983). Manual for the State-Trait Anxiety Inventory. Palo Alto, CA: Consulting Psychologists Press. Tolin, D. F., Abramowitz, J. S., Przeworski, A., & Foa, E. B. (2002). Thought suppression in obsessive–compulsive disorder. Behaviour Research and Therapy, 40, 1255–1274.

ARTICLE IN PRESS J.C. Magee, R.E. Zinbarg / Behaviour Research and Therapy 45 (2007) 2836–2849

2849

Trinder, H., & Salkovskis, P. (1994). Personally relevant intrusions outside the laboratory: Long-term suppression increases intrusion. Behaviour Research and Therapy, 32, 833–842. Warda, G., & Bryant, R. A. (1998). Thought control strategies in acute stress disorder. Behaviour Research and Therapy, 36, 1171–1175. Watson, D., & Clark, L. A. (1994). The PANAS-X: Manual for the positive and negative affect schedule—Expanded form. The University of Iowa. Wegner, D. M. (1994). Ironic processes of mental control. Psychological Review, 101, 34–52. Wegner, D. M., & Erber, R. (1992). The hyperaccessibility of suppressed thoughts. Journal of Personality and Social Psychology, 63, 903–912. Wegner, D. M., Erber, R., & Zanakos, S. (1993). Ironic processes in the mental control of mood and mood-related thought. Journal of Personality and Social Psychology, 65, 1093–1104. Wegner, D. M., Schneider, D. J., Carter, S. R., & White, T. L. (1987). Paradoxical effects of thought suppression. Journal of Personality and Social Psychology, 60, 500–508. Wegner, D., & Zanakos, S. (1994). Chronic thought suppression. Journal of Personality, 62, 615–640. Wenzlaff, R. M., & Wegner, D. M. (2000). Thought suppression. Annual Review of Psychology, 51, 59–91. Woody, S. (1996). Effects of focus of attention on anxiety levels and social performance of individuals with social phobia. Journal of Abnormal Psychology, 105, 61–69. Woody, S., & Rodriguez, B. (2000). Self-focused attention and social anxiety in social phobics and normal controls. Cognitive Therapy and Research, 24, 473–488.