issues related to social anxiety among controls in social phobia research

13
BEHAVIOR THERAPY27, 79-91, 1996 Issues Related to Social Anxiety Among Controls in Social Phobia Research STEFAN G. HOFMANN WALTON T. ROTH Stanford University School of Medicine and Veterans Affairs Medical Center, Palo Alto, California Twenty-four social phobic individuals and 22 nonphobic controls participated in an interview and questionnaire study. By applying the same criterion that was used in a previous study, both phobics and controls were retrospectively subdivided into groups with or without generalized social fear, yielding four groups: nongeneralized phobics (n = 9), generalized phobics (n = 15), nongeneralized controls (n = 10), and generalized controls (n = 12). The four groups differed in the severity ratings of their social anxiety. Generalized controls scored as high as nongeneralized phobics. These two groups showed lower scores than generalized phobics and higher scores than nongeneralized controls. A comparison with scores reported in other studies indicated that the nongeneralized controls can be characterized as "supernormal;' and generalized controls as "subclinical" The results illustrate problems related to the use of control groups in social phobia research. According to the Diagnostic and Statistical Manual of Mental Disorders (DSM-III-R, American Psychological Association, 1987) and DSM-IV (APA, 1994), the clinician can specify a generalized subtype of social phobia if the social fears are related to most social situations. Recent research indicates that both the generalized subtype of social phobia and the highly overlapping diagnosis of avoidant personality disorder apply to more severe cases of social phobia rather than to qualitatively different subgroups (e.g., Liebowitz, Gorman, Fyer, & Klein, 1985; Schneier, Spitzer, Gibbon, Fyer, & Liebowitz, 1991; Turner, Beidel, Borden, Stanley, & Jacob, 1991). Differences between social phobia subtypes were found in self-report measures (Herbert, Hope, & BeUack, 1992; Holt, Heimberg & Hope, 1992; Turner, Beidel, & Townsley, 1992), This work was supported in part by the Deutscher Akademischer Austauschdienst and the VA Medical Center. Part of the data from this study was presented at the 15th National Con- ference of the Anxiety Disorders Association of America, Pittsburgh, PA. We thank David H. Barlow and our reviewers for their valuable editorial comments. Please address correspondence concerning this article to Stefan G. Hofmann, Center for Stress and Anxiety Disorders, 1535 Western Avenue, Albany, New York 12203. 79 0005-7894/96/0079-0091 $1.00/0 Copyright 1996 by Associationfor Advancement of BehaviorTherapy All rights of reproduction in any form reserved.

Upload: bu

Post on 29-Nov-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

BEHAVIOR THERAPY 27, 79-91, 1996

Issues Related to Social Anxiety Among Controls in Social Phobia Research

STEFAN G. HOFMANN

WALTON T. ROTH

Stanford University School of Medicine and Veterans Affairs Medical Center, Palo Alto, California

Twenty-four social phobic individuals and 22 nonphobic controls participated in an interview and questionnaire study. By applying the same criterion that was used in a previous study, both phobics and controls were retrospectively subdivided into groups with or without generalized social fear, yielding four groups: nongeneralized phobics (n = 9), generalized phobics (n = 15), nongeneralized controls (n = 10), and generalized controls (n = 12). The four groups differed in the severity ratings of their social anxiety. Generalized controls scored as high as nongeneralized phobics. These two groups showed lower scores than generalized phobics and higher scores than nongeneralized controls. A comparison with scores reported in other studies indicated that the nongeneralized controls can be characterized as "supernormal;' and generalized controls as "subclinical" The results illustrate problems related to the use of control groups in social phobia research.

According to the Diagnostic and Statistical Manual of Mental Disorders (DSM-III-R, American Psychological Association, 1987) and DSM-IV (APA, 1994), the clinician can specify a generalized subtype of social phobia if the social fears are related to most social situations. Recent research indicates that both the generalized subtype of social phobia and the highly overlapping diagnosis of avoidant personality disorder apply to more severe cases of social phobia rather than to qualitatively different subgroups (e.g., Liebowitz, Gorman, Fyer, & Klein, 1985; Schneier, Spitzer, Gibbon, Fyer, & Liebowitz, 1991; Turner, Beidel, Borden, Stanley, & Jacob, 1991). Differences between social phobia subtypes were found in self-report measures (Herbert, Hope, & BeUack, 1992; Holt, Heimberg & Hope, 1992; Turner, Beidel, & Townsley, 1992),

This work was supported in part by the Deutscher Akademischer Austauschdienst and the VA Medical Center. Part of the data from this study was presented at the 15th National Con- ference of the Anxiety Disorders Association of America, Pittsburgh, PA. We thank David H. Barlow and our reviewers for their valuable editorial comments.

Please address correspondence concerning this article to Stefan G. Hofmann, Center for Stress and Anxiety Disorders, 1535 Western Avenue, Albany, New York 12203.

79 0005-7894/96/0079-0091 $1.00/0 Copyright 1996 by Association for Advancement of Behavior Therapy

All rights of reproduction in any form reserved.

8 0 HOFMANN & ROTH

psychophysiological response during exposure (Heimberg, Hope, Dodge, & Becker, 1990; Hofmann, Newman, Ehlers, & Roth, 1995; Levin et al., 1993), and degree of cognitive interference during a modified Stroop color-naming test (McNeil, Ries, Taylor, et al., 1995). A more detailed summary can be found elsewhere (Heimberg, Holt, Schneier, Spitzer, & Liebowitz, 1993; McNeil, Ries, & Turk, 1995).

Researchers typically distinguish between a generalized subtype and a residual category of"nongeneralized" subtype of social phobia (e.g., Holt, Heimberg, & Hope, 1992; Turner et al., 1992). Heimberg et al. (1993) discussed three possible subtypes of social phobia: generalized, nongeneralized, and circum- scribed. According to their definition, the nongeneralized subtype functions in at least one broad social domain without clinically significant anxiety. The circumscribed subtype, on the other hand, only experiences anxiety in a few discrete situations. Due to the small number of individuals with circumscribed social phobia (Tumer, Beidel, Dancu, & Keys, 1986), however, previous studies either did not include circumscribed social phobics (Herbert et al., 1992), or pooled them with the nongeneralized group (Holt, Heimberg, & Hope, 1992).

Social phobia (including all its subtypes) is a very common form of anxiety disorder. Depending on the threshold used to define distress or impairment, epidemiological and community-based studies have found lifetime prevalence rates of social phobia between 3% and 13% (APA, 1994). Setting the clinical threshold, however, is a general problem in the classification of mental dis- orders. Clinical and subclinical symptoms share many common features (e.g., Craske, Rapee, Jackel, & Barlow, 1989) and a number of symptoms are very common in nonclinical samples (e.g., social anxiety, panic attacks, or worries). For example, Pollard and Henderson (1988) examined the point prevalence of four types of social phobia, including public speaking/performing, writing in front of others, eating in restaurants, and use of public restrooms. The re- suits indicated that the total social phobia prevalence was 2.0% when using the DSM-III significant distress criterion. Omitting this criterion, the preva- lence jumped to 22.6%. Public speaking/performing anxiety was by far the most common form (20.6%). The authors indicated that the "significant dis- tress" criterion may prevent accurate prevalence estimates of phobias because some individuals are successfully able to avoid anxiety-provoking situations. Very similar results were reported by Telch, Lucas, and Nelson (1989) in a study on nonclinical panic attacks. The authors found that 12.29% of a large nonclinical sample of college students (n = 2,375) reported one or more un- expected panic attacks in their lifetime. When symptom and frequency criteria for panic disorder were considered, the prevalence decreased to 2.36%. Inter- estingly, individuals with subclinical panic attacks had impairment in perceived physical and emotional health and occupational and financial functioning (Klerman, Weissman, Oulette, Johnson, and Greenwald, 1991). Similarly, in- dividuals with subclinical depressive symptoms manifested higher levels of psychopathology than individuals with no depressive symptoms (Gotlib, Lewin- sohn, & Seeley, 1995).

CONTROLS IN SOCIAL PHOBIA RESEARCH 81

Studies like these raise important questions about the distinction between clinical and nonclinical symptoms and the validity of our current classification system, which splits continuous variables into different categories. The present study reports the results of a post hoc analysis of the control group, which was part of a previous investigation (Hofmann, Newman, et al., 1995). The goal of the study was to identify problems related to social anxiety among controls in social phobia research in order to examine in more detail what controls are methodologically appropriate for social phobia research. Spe- cifically, we sought to determine if it was possible to apply the same criterion to divide individuals into two control groups that we used to divide phobic individuals into those with "generalized" and "nongeneralized ''1 social fears. What are the characteristics of the resultant groups? Are the groups created better labeled "subclinical social phobiC' and "supernormal" rather than simply "control"?

Method Subject Recruitment

As part of a previous study on psychophysiological differences between social phobics with and without avoidant personality disorder (Hofmann, Newman, et al., 1995), 22 controls and 24 public speaking anxious individuals were recruited through advertisements in the classified section of local newspapers. To recruit participants who were fearful of public speaking, the advertisement read as follows: 'Afraid of public speaking in front of other people? Stanford University and Palo Alto VA are offering free evaluations to people who have a fear of speaking in some or all situations. For further information, c a l l . . "' For controls, the advertisement read: "Speech study. Stanford University and Palo Alto VA will pay $50 for participation in study of attitudes and reactions to speaking. We are looking for healthy adults between 18 and 65 who gen- erally are not afraid of speaking. For further information, call . . "

Participants were scheduled for two structured interviews (described below) conducted on the same day by the first author of the present study. Phobic individuals and controls responding to the advertisements were invited to par- ticipate in the study and scheduled for the interviews. On the basis of the two interviews, participants were divided into four groups: nongeneralized controls (n = 10, 45% of controls), generalized controls (n = 12, 55% of controls), nongeneralized phobics (n = 9, 37.5 % of phobic s), and generalized phobics (n = 15, 62.5% of phobics). The interviews also gathered sufficient information for making DSM-III-R diagnoses of anxiety disorders, psycho- active substance use disorder, schizophrenia, and mood disorders.

In order to gather information about the validity of the recruitment pro- cedure, participants were asked in the second interview to give a rating of

The term "nongeneralized subtype" will be used to describe the group of individuals who do not qualify for the "generalized subtype."

82 HOFMANN & ROTH

their fear of public speaking on a 0 (not at all) to 10 (very much) Likert scale. The results of the ANOVAs revealed significant group differences, F(3, 42) = 62.98, p < .0001. Post-hoc tests (Fisher PLSD) showed that both phobic sub- groups (generalized and nongeneralized) reported significantly higher fear rat- ings (M = 9.2 and 8.8, respectively) than the generalized controls (M = 4.6), who in turn evidenced significantly higher ratings than the nongeneralized control group (M = 1.5, p < .05).

Group Assignments

Assignment to the phobic versus the nonphobic group was based on the Structured Clinical Interview forDSM-III-R (SCID; Spitzer, Williams, Gibbon, & First, 1990). All public speaking anxious individuals, but none of the con- trois, met DSM-III-R criteria for social phobia. Interviews were audiotaped, and an independent and experienced clinician listened to a random sample of half of the tapes (14 presumed social phobic individuals and 11 presumed controls). The interviewer and the independent rater agreed in 100% of the cases for the diagnosis of social phobia. All social phobics rated public speaking as their most fearful social situation.

Most of the results of the present study will be compared to studies that tested unscreened student samples. In order to preserve heterogeneity within each group and to avoid selection effects, the presence or absence of clinically significant public speaking anxiety was the only criterion that was used to assign individuals to the control or phobia group. Thus, neither phobics nor controls were excluded in case of any comorbid Axis I diagnosis.

Both phobics and controls were assigned to either the generalized or the nongeneralized subtype, depending on whether or not the phobic situation includes most social situations. The assignment to the generalized or non- generalized subtype was made on the basis of subjective fear ratings of specific social situations during the second interview. In this interview, participants were asked to rate their fear of a number of specific social situations from different social phobia domains. When examining the prevalence and overlap of social anxiety across different classes of situations, Holt, Heimberg, Hope, and Liebowitz (1992) were able to identify four different domains: formal speaking/interaction, informal speaking/interaction, assertive interaction, and observation by others. Situations in Holt, Heimberg, Hope, et al's formal speaking/interaction category (acting, giving a report, speaking up at a meeting, participating in small groups) were not considered as a subtyping criterion because, as a result of our recruitment procedure, controls were generally not afraid of speaking, whereas all phobics were afraid of public speaking. Instead, all situations were part of these categories: informal speaking/inter- action, assertive interaction, and observation by others. All participants were asked to give a rating of their fear on a scale from 0 (not at all) to 10 (very much) for the following six social situations: having a first date, using the telephone, being introduced (part of the category informal speaking/inter-

CONTROLS IN SOCIAL PHOBIA RESEARCH 83

action), meeting people in authority, being teased (part of the category as- sertive interaction), and being under observation by others. 2

The cut-off criterion for the generalized subtype of social fear was similar to that of Turner et al. (1992). Phobics and controls were assigned a gener- alized subtype diagnosis if they rated a minimum of four social situations as at least moderately fear-provoking (4 or more) on a 10 point Likert scale. This cut-off score (four out of six possible social situations) was the authors' interpretation of the DSM-III-R criterion "most social situations" and had been used in a previous study (Hofmann, Newman, et al., 1995). By using this criterion it was possible to assign 55% (12/22) of the controls and 62.5% (15/24) of the phobics to the generalized subtype, 22 (1) = .3, p > .5. Forty-four per- cent (4/9) of the nongeneralized controls, but none of the nongeneralized phobics (0%), rated all situations as four or less. Among the nongeneralized phobics, 22% (2/9) rated one situation, 33% (3/9) rated two situations, and 44% (4/9) rated three situations as at least moderately fear provoking (14 or greater). Thus, the sample characteristic of the nongeneralized phobic subgroup did not allow us to define a separate subgroup of individuals with circumscribed social phobia (Heimberg et al., 1993; McNeil, Reis, Taylor, et al., 1995).

The four groups (nongeneralized controls, generalized controls, nongeneral- ized phobics, and generalized phobics) did not differ in age, F(3, 42) = .53, p > .6, or sex distribution, Z 2 (3) = .06, p > .9. The total sample had a mean age of 46.4 (SD = 16.7, range = 20-65). Over half of the participants (59%) were female.

Dependent Measures After the interviews, participants were given a battery of questionnaires

assessing social anxiety, including the Fear of Negative Evaluation Scale (FNE; Watson & Friend, 1969), the Social Avoidance and Distress Scale (SADS; Watson & Friend), the Social Phobia and Anxiety Inventory (SPAI; Turner, Beidel, Dancu & Stanley, 1989), the Social Interaction and Self-Statement Test (SISST; Glass, Merluzzi, Biever & Larsen, 1982), and the Test Anxiety Scale (TAS; Sarason, 1978). Typically, the SISST is administered after a struc- tured social interaction. In the present study, individuals were asked to fill out the SISST while imagining a fearful social situation without our specifying the type of social situation. By using this simplified version of the original procedure, the SISST (negative self-statements) distinguished controls, social phobics without avoidant personality disorder, and social phobics with avoidant personality disorder in a previous study (Hofmann, Newman, et al., 1995).

Participants were given self-report instruments to assess the severity of gen- eral anxiety and mood pathology, since previous studies reported a positive correlation between severity of social phobia and trait anxiety and depression

2 Note that these situations are very similar, but not identical, to the specific situations re- ported by Holt, Heimberg, Hope, et al. (1992).

84 HOFMANN & ROTH

(e.g., Herbert et al., 1992). The additional questionnaires included the State- Trait Anxiety Inventory (STAI-trait; Spielberger, Gorsuch, & Lushene, 1970), the Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961), the Body Sensations Questionnaire (BSQ; Chambless, Caputo, Bright, & Gallagher, 1984), and the Agoraphobic Cognitions Ques- tionnaire (ACQ; Chambless et al.). The latter two questionnaires were in- cluded because situational panic attacks and fear of bodily sensations seem to play an important role in the psychopathology of public speaking anxious individuals (Hofmann, Ehlers, & Roth, 1995).

Results Additional Diagnoses

At least one additional DSM-III-R diagnosis (other than social phobia) was found in 44% (4/9) of the nongeneralized phobics and 47% (7/15) of the gen- eralized phobics. Interestingly, 50% (6/12) of the generalized controls, but none of the nongeneralized control group (0/10), received an Axis I diagnosis. The overall difference among the four groups did not reach statistical signifi- cance, ~2 (3, n = 46) = 7.56, p = .056. However, a comparison between nongeneralized and generalized controls was statistically significant, ~2 (1, n = 22) = 4.59, p < .04 (with correction for continuity). Generalized controls received DSM-III-R diagnoses of alcohol abuse, specific phobia (claustro- phobia), posttraumatic stress disorder, and panic disorder. Nongeneralized phobics met criteria for specific phobia (flying, heights), dysmorphophobia, and generalized anxiety disorder; generalized phobics received diagnoses of specific phobias (heights, flying, claustrophobia), generalized anxiety disorder, panic disorder, major depressive episode, and alcohol abuse.

Self-Report Measures

ANOVA results of the questionnaires are summarized in Table 1. Due to the small sample size of subjects in each subgroup and the relatively large number of ANOVA comparisons, the statistical significance level was adjusted according to Bonferroni for the following eight social anxiety scales: FNE, SADS, SPAI (total), SPAI (social phobia subscale), SPAI (agoraphobia sub- scale), SISST (negative statements), SISST (positive statements), and TAS. The Bonferroni-corrected alpha level for these eight scales was p < .006. The results showed significant group differences for the FNE, SADS, SISST (nega- tive statements), SPAI (social phobia subscale and total score), STAI (trait form), ACQ, and BSQ. Post-hoc tests (Fisher PLSD, p < .05) indicated that nongeneralized controls showed the lowest, and generalized phobics the highest scores in social severity measures (FNE, SADS, total SPAI score, and the social phobia subscale of the SPAI) and the STAI (trait). Generalized controls and nongeneralized phobics did not differ in any of those measures. In ad- dition, nongeneralized controls scored lower than any of the other groups in the ACQ and BSQ. The general pattern of results indicated that nongeneral-

CONTROLS IN SOCIAL PHOBIA RESEARCH

TABLE 1 COMPARISON BETWEEN THE SUBGROUPS ON SELF-INVENTORIES

85

Nongeneralized Generalized Nongeneralized Generalized Controls Controls Phobics Phobics

Questionnaires (n = 10) (n = 12) (n = 9) (n = 15) F-value

FNE 4.2 (3.9) a 12.3 (6.7) b 14.1 (9.3) 8 22.7 (6.8) c 15.2"** SADS 1.5 (1.3) a 7.9 (6.6) b 9.3 (7.9) b 17.3 (7.5) ~" 12.3"** SPAI

Total 14.5 (12.0) a 41.7 (24.5) 8 60.0 (23.5) 8 90,3 (30.5) c 20.6*** Social phobia 23.1 (13.6) a 62.8 (23.9) 8 77.4 (28.2) 8 113,4 (39.2) c 19.8"** Agoraphobia 8.6 (4.5) a 21.2 (13.3) 8 17.3 (9.2) 23.1 (15.6) b 3.2 *2

SISST negative 23.1 (8.4) a 33.5 (11.0) ~ 33.4 (8.5) ~ 45.6 (15.5) b 7.0** positive 48.2 (12.8) 48.3 (12.7) 40.6 (7.6) 41.9 (10.4) 1.5

STAI, trait 27.8 (5.5) a 38.8 (10.2) 8 38.9 (8.4) b 48.0 (10.7) c 9.6*** TAS 6.3 (3.0) a 15.7 (8.8) b 13.7 (7.0) b 19.7 (8.9) b 6.4 *2 ACQ 1 15.0 (7.1) a 22.2 (7.2) 2 22.9 (5.3) 8 24.5 (9.1) b 4.2* BSQ 20.1 (5.7) a 33.1 (14.2) 8 42.0 (11.3) b 41.4 (13.3) 8 7.7** BDI 1.8 (1.7) 5.7 (5.7) 4.1 (3.3) 6.9 (6.2) 2.3

Note. The Table shows means (standard deviations) and the ANOVA Group effects. Different superscripts indicate significant post-hoc group differences (Fisher test, p < .05). FNE: Fear of Negative Evaluation scale; SADS: Social Avoidance and Distress Scale; SPAI: Social Phobia and Anxiety Inventory; STAI: State-Trait Anxiety Inventory; TAS: Test Anxiety Scale; ACQ: Agoraphobic Cognitions Questionnaire; BSQ: Body Sensations Questionnaire; BDI: Beck Depression Inventory. *p < .05; **p < .001; ***p < .01901; J n = 13 in the generalized phobic group; 2 Not statistically significant after Bonferroni correction (adjusted alpha: p < .006).

ized controls were character ized by ex t remely low levels, and genera l ized phobics by e x t r e m e l y h igh levels o f socia l anxiety , gene ra l anxiety , and anx ious appre - hens ion . G e n e r a l i z e d cont ro ls and n o n g e n e r a l i z e d phob i c s were in te rmedia te .

Comparison With Data From the Literature

In o r d e r to e s t ima te h o w rep resen ta t ive o u r con t ro l s ample was, the ques - t ionna i re scores o f the two con t ro l subgroups o f the p resen t study, the non- g e n e r a l i z e d cont ro l s , and the g e n e r a l i z e d con t ro l s w e r e c o m p a r e d wi th scores r epor t ed in o the r s tudies. Sta t is t ical c o m p a r i s o n s were conduc t ed by ca lcu- lat ing unpa i red t tests (unpoo led va r i ances ) based on the repor ted s ample sizes, means , and s tandard dev ia t ions in p rev ious s tudies .

Watson and F r i e n d (1969) tes ted a r a n d o m s a m p l e o f 145 f ema le and 60 m a l e s tudents and a d m i n i s t e r e d the F N E and the SADS. Watson and Fr i end ' s cont ro l g r o u p had h i g h e r scores than the n o n g e n e r a l i z e d cont ro ls in bo th the F N E (Ms = 15.5 vs. 4 .2) , t (213) = 8.21, p < .0001 and the S A D S (Ms =

9.1 vs. 1.5), t (213) = 10.96, p < .0001. H o w e v e r , there w e r e no d i f fe rences b e t w e e n Watson and Fr iend ' s con t ro l g roup and the gene ra l i z ed con t ro l s in

86 HOFMANN & ROTH

TABLE 2 COMPARISON WITH THE STUDY BY TURNER ET AL. (1989)

Present study Turner et al. (1989)

Nongeneralized Generalized Total of Non-socially Controls Controls Control Group Anxious Students

Questionnaires (n = 10) (n = 12) (n = 22) (n = 123)

FNE 4.2 (3.9) a 12.3 (6.7) b 8.6 (6.9) c 7.0 (3.8) c SADS 1.5 (1.3) a 7.9 (6.6) b 5.0 (5.9) c 3.2 (2.4) c SPAI, total 14.5 (12.0) a 41.7 (24.5) b 29.3 (23.8) c 32.7 (SD: not reported) STAI, trait 27.8 (5.5) a 38.8 (10.2) b 33.8 (9.9) c 32.0 (5.7) c

Note. The Table shows means (standard deviations). Different superscripts indicate significant post-hoc group differences (t tests for unpooled variances, p < .05). FNE: Fear of Nega- tive Evaluation scale; SADS: Social Avoidance and Distress Scale; SPAI: Social Phobia and Anxiety Inventory; STAI: State-Trait Anxiety Inventory.

the FNE, t (215) = 1.56, p > .1 or the SADS, t (215) = .61, p > .3. Similar results were found for the SISST and the TAS. Glass et al. (1982) reported that the lowest mean SISST-negative score was found in a subgroup of 32 normal female psychology undergraduates. Nongeneralized controls scored significantly lower in the SISST than Glass et al 's undergraduates (Ms = 35.3 vs. 23.1), t (39) = 23.1, p < .005. No difference, however, was found between Glass et al 's control group and the generalized controls (Ms = 35.1 and 33.5, respectively), t (42) = .58, p > .5. Sarason (1978) administered the TAS to 283 male students. This group had significantly higher TAS scores than the nongeneralized controls (Ms = 16.7 vs. 6.3), t (291) = 10.17, p < .0001, but did not differ f rom the generalized control group (Ms = 16.7 and 15.7, respec- tively), t (293) = .41, p > .4.

In order to gather psychometric information on the SPAI, Turner et al. (1989) administered a screening battery, including the SADS, the FNE, STAI (trait), and SADS, to 308 introductory psychology students. Based on the SADS scores, 123 individuals were designated as nonsocially anxious and 59 as so- cially anxious. Turner et al. reported that the socially anxious group scored significantly higher than the nonsocially anxious group on the SADS (Ms = 14.1 vs. 3.2), the FNE (Ms = 22.4 vs. 7.0), the STAI (trait; Ms = 14.1 vs. 3.2), "and the SPAI (Ms = 72.2 vs. 32.7). Table 2 shows the means and stan- dard deviations of the total control group, the nongeneralized controls, the generalized controls of the present study, and Turner et al.'s nonsocially anxious group. A comparison between Turner et a l ' s nonanxious control group and the total control group of the present study showed no differences in the FNE, t (148) = 1.06, p > .1, the SADS, t (148) = 1.41, p > .1, or the STAI (trait), t (148) = .83 p > .2. However, Turner et a l ' s control group had sig- nificantly higher scores than the nongeneralized controls in the FNE, t (136) = -2 .19 , p < .05, the SADS, t (136) = -3 .67, p < .001, and the STAI (trait),

CONTROLS IN SOCIAL PHOBIA RESEARCH 87

t (136) = -2 .32, p < .03, and significantly lower scores than the generalized controls in these questionnaires [t (138) = 4.68, p < .005; t (138) = 2.45, p < .02; and t (138) = 2.28, p < .03, respectively].

Whereas Turner et al. (1989) found that 50% of a nonanxious student popu- lation (n = 124) had SPAI scores of more than 34, none of the nongeneralized controls received a SPAI score of more than 34 in the present sample. The mean (standard deviation) of the nongeneralized controls of the present study was 14.5 (12.0), the mode was 15.1, and the median was 16.2. According to Turner et al., a SPAI score of 60 or above "should alert the clinician to assess for the possibility of social phobia" (p. 40). In fact, 3 of 12 generalized controls (25%) showed a SPAI score of 60 or above (the scores of the three individuals were 66.8, 71.1, and 87.0).

Discussion Some of our results were expected, others were surprising. Generalized

phobics scored highest on self-report measures of social anxiety. These results replicate findings from previous studies suggesting that the generalized sub- type of social phobia identifies the more severe cases of social phobic indi- viduals (Heimberg et al., 1990, 1993; Herbert at al., 1992; Holt, Heimberg, & Hope, 1992; Levin et al., 1993; Liebowitz et al., 1985; McNeil, Reis, Taylor, et al., 1995; Schneier et al., 1991; Turner et al., 1992). More surprising was the fact that 55% of the controls met criteria for generalized social fear, al- though none of the controls met criteria for social phobia. Furthermore, gen- eralized controls scored as high as nongeneralized social phobics on social phobia severity scales, whereas nongeneralized controls scored the lowest. Very similar results were reported by Gotlib et al. (1995) in a study on clinical and nonclinical depressive symptoms, in which individuals with high scores on self-report measures, but who did not qualify for a psychiatric disorder, were labeled as "false positives,' False positives are most likely to be found when possible criterion symptoms, like public speaking anxiety, which is the most prevalent fear in both the general population and among social phobic individuals (Pollard and Henderson, 1988), occur frequently. Both Gotlib et al's false-positive individuals and our generalized controls showed an elevated level of symptomatology without meeting diagnostic criteria for the disorder, and both groups showed more additional DSM-III-R diagnoses than the true- negative/nongeneralized participants.

Maybe our generalized controls were simply misdiagnosed phobics who responded to the advertisement for the control group in order to receive the $50 in reimbursement. Various findings, however, refute this hypothesis. First, controls and phobics differed significantly in their public speaking anxiety. Second, the group assignment was found to be highly reliable. Finally, the questionnaire scores of the total sample of controls were comparable to the college sample tested by Turner et al. (1989). Nongeneralized controls scored lower than any of the control samples reported in the literature (Glass et al.,

88 HOFMANN & ROTH

1982; Sarason, 1978; Turner et al. 1989; Watson & Friend, 1969), while gen- eralized controls were not more socially anxious than normal students in the other studies (Glass et al., 1982; Sarason, 1978; Watson & Friend, 1969).

Instead, the results suggest an additive relationship between generalized social fear and phobic avoidance for the prediction of the severity of social anxiety. Individuals who report social anxiety in a variety of common social situations, but who do not consider themselves to be socially phobic due to lack of phobic avoidance and interference, showed social phobia severity scores that were as high as the ones reported by social phobic individuals with a nongeneralized subtype of social fear. These results indicate that some of our controls were fearless (the nongeneralized controls) while others were coura- geous (the generalized controls) because they expose themselves to social sit- uations despite their fear (Rachman, 1990). The first group (the nongener- alized and fearless controls) may, therefore, also be described as "supernormal" and the second group (the generalized and courageous controls) as "sub- clinical" Supernormals are characterized by extremely low levels of social anxiety, test anxiety, agoraphobic behavior (and cognitions), and fear of bodily sensations.

This raises the problem of how to define a control group. One might argue that the generalized controls should have been excluded from the study given the relatively high level of clinical psychopathology. In fact, Levin et al. (1993) excluded individuals with any other past or present anxiety disorders or cur- rent Axis I disorder from the control group. Although this procedure seems intuitively appropriate, it poses a methodological problem in that the experi- menter chooses (more or less) arbitrary criteria to select a certain subgroup of the population and defines it as a control group. As a result of using highly selected supernormal controls, group differences can become difficult to inter- pret because the two groups no longer differ in only one variable (e.g., clini- cally significant public speaking anxiety). Instead, other group differences (e.g., general level of psychopathology) could account for the results, unless the same criteria are applied to the phobic group, which is nearly impossible given the high percentage of comorbid diagnoses among anxiety and mood disorders. Depending on the hypothesis, the researcher will have to decide whether to use a supernormal control group or a representative sample of the normal population (community control group). Using a community con- trol group may increase the validity of the study, but may also produce smaller effect sizes (Cohen, 1992).

This study has a number of limitations. First, the results were only based on questionnaire and interview data, which limit the validity of the findings. Moreover, it is unclear whether controls and phobics used the same subjective scale to rate their anxiety. A rating of"very much" for a nonphobic individual might mean "low" for a phobic subject. Behavioral anchors might have been useful to calibrate the participants in their anxiety ratings. Second, the social fear measures that were used in the study may not have had enough discrim- inant validity to differentiate a clinical from a nonclinical sample. The results

CONTROLS IN SOCIAL PHOBIA RESEARCH 89

might have shown differences between generalized controls and nongeneral- ized phobics if other measures had been used. Also, public speaking anxiety is a very common but not very severe type of social phobia, which may have been the reason why the differences between generalized controls and non- generalized phobics were not as great as expected. Third, our control sample was not ideal because it consisted of a community sample that was consid- erably older than the samples of college students from previous studies that were used as a reference population. Fourth, the number of Axis I diagnoses in the generalized control group was very high, and the sample size was very small, which further limits the validity of the study. Also, due to the small number of subjects, individuals with nongeneralized social anxiety included all participants who did not meet criteria for the generalized subtype. Thus, the nongeneralized subtype was not defined as proposed by Heimberg et al. (1993). A replication that includes an adequate number of college students as controls would be desirable to give clearer recommendations.

In general the present study showed that not all normal controls are the same. Pathological anxiety and normal anxiety are different in degree rather than in quality. Moreover, anxiety and fear are probably normally distributed, ranging from fearless through courageous to fearful. A categorical classifica- tion system that splits a continuous variable into "present" and "absent" is al- ways artificial and imperfect. Selecting nonphobics (i.e., controls) is therefore just as difficult as selecting phobics. Neglecting heterogeneity in controls may result in misleading conclusions regarding the nature of a disorder.

References American Psychiatric Association (1987). Diagnostic and statistical manual for mental disorders

(3rd ed. rev.). Washington, DC: Author. American Psychiatric Association (1994). Diagnostic and statistical manual for mental disorders

(4th ed.). Washington, DC: Author. Beck, A. T., War, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory for

measuring depression. Archives of General Psychiatry, 4, 561-571. Chambless, D. L., Caputo, G. C., Bright, P., & Gallagher, R. (1984). Assessment of fear of

fear in agoraphobics: The Body Sensations Questionnaire and the Agoraphobic Cognitions Questionnaire. Journal of Consulting and Clinical Psychology, 52, 1090-1097.

Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155-159. Craske, M. G., Rapee, R. M., Jackel, L., & Barlow, D. H. (1989). Qualitative dimensions

of worry in DSM-III-R generalized anxiety disorder subjects and nonanxious controls. Behaviour Research and Therapy, 27, 397-402.

Glass, C. R., Merluzzi, T. V., Biever, J. I., & Larsen, K. H. (1982). Cognitive assessment of social anxiety: Development and validation of a self-statement questionnaire. Cognitive Therapy and Research, 6, 37-55.

Gotlib, I. H., Lewinsohn, P. M., & Seeley, J. R. (1995). Symptoms versus a diagnosis of de- pression: Differences in psychosocial functioning. Journal of Consulting and Clinical Psy- chology, 63, 90-100.

Heimberg, R. G., Holt, C. S., Schneier, E R., Spitzer, R. L., & Liebowitz, M. R. (1993). The issue of subtypes in the diagnosis of social phobia. Journal of Anxiety Disorders, 7, 249-269.

90 HOFMANN & ROTH

Heimberg, R. G., Hope, D. A., Dodge, C. S., & Becker, R. E. (1990). DSM-III-R subtypes of social phobia: Comparison of generalized social phobics and public speaking phobics. The Journal of Nervous and Mental Disease, 178, 172-179.

Herbert, J. D., Hope, D. A., & Bellack, A. S. (1992). Validity of the distinction between gen- eralized social phobia and avoidant personality disorder. Journal of Abnormal Psychology, 101,332-339.

Hofmann, S. G., Ehlers, A., & Roth, W. T. (1995). Conditioning theory: A model for the etiology of public speaking anxiety? Behaviour Research and Therapy, 33, 567-571.

Hofmann, S. G., Newman, M. G., Ehlers, A., & Roth, W. T. (1995). Psychophysiological differ- ences between subtypes of social phobics. Journal of Abnormal Psychology, 104, 224- 231.

Holt, C. S., Heimberg, R. G., & Hope, D. A. (1992). Avoidant personality disorder and the generalized subtype of social phobia. Journal of Abnormal Psychology, 101,318-325.

Holt, C. S., Heimberg, R. G., Hope, D. A., & Liebowitz, M. R. (1992). Situational domains of social phobia. Journal of Anxiety Disorders, 6, 63-77.

Klerman, G. L., Weissman, M. M., Oulette, R., Johnson, J., & Greenwald, S. (1991). Panic attacks in the community: Social morbidity and health care utilization. Journal of the American Medical Association, 265, 742-746.

Levin, A. P., Saoud, J. B., Strauman, T., Gorman, J. M., Fyer, A., Crawford, R., & Liebowitz, M. R. (1993). Responses of"generalized" and "discrete" social phobics during public speak- ing. Journal of Anxiety Disorders, 7, 207-221.

Liebowitz, M. R., Gorman, J. M., Fyer, A. J., & Klein, D. F. (1985). Social phobia: Review of a neglected anxiety disorder. Archives of General Psychiatry, 42, 729-736.

McNeil, D. W., Ries, B. J., Taylor, L. J., Boone, M. L., Carter, L. E., Turk, C. L., & Lewin, M. R. (1995). Comparison of social phobia subtypes using Stroop tests. Journal of Anxiety Disorders, 9, 47-57.

McNeil, D. W., Ries, B. J., & Turk, C. L. (1995). Behavioral assessment: Self- and other-report, physiology and overt behavior. In R. G. Heimberg, M. R. Liebowitz, D. A. Hope, & E R. Schneier (Eds.), Social phobia: Diagnosis, assessment and treatment. New York: Guilford Press.

Pollard, C. A., & Henderson, J. G. (1988). Four types of social phobia in a community sample. The Journal of Nervous and Mental Disease, 176, 440-445.

Rachman, S. (1990). Fear and courage. New York: Freeman. Sarason, I. G. (1978). The Test Anxiety Scale: Concept and research. In C. D. Spielberger

& I. G. Sarason (Eds.), Stress andAnxiety (pp. 193-216). New York: John Wiley & Sons. Schneier, E R., Spitzer, R. L., Gibbon, M., Fyer, A. J., & Liebowitz, M. R. (1991). The re-

lationship of social phobia subtypes and avoidant personality disorder. Comprehensive Psy- chiatry. 32, 1-5.

Spielberger, C. D., Gorsuch, R. L., & Lushene, R. E. (1970). State-Trait Anxiety Inventory. Palo Alto, CA: Consulting Psychologists Press.

Spitzer, R. L., Williams, J. B., Gibbon, M., & First, M. B. (1990). Structured Clinicallnterview for DSM-III-R-Sandoz version (SCID). New York: New York State Psychiatric Institute.

Telch, M. J., Lucas, J. A., & Nelson, P. (1989). Nonclinical panic in college students: An in- vestigation of prevalence and symptomatology. Journal of Abnormal Psychology, 98, 300- 306.

Turner, S. M., Beidel, D. C., Borden, J. W., Stanley, M. A., & Jacob, R. G. (1991). Social phobia: Axis I and I! correlates. Journal of Abnormal Psychology, 100, 102-106.

Turner, S. M., Beidel, D. C., Dancu, C. V., & Keys, D. J. (1986). Psychopathology of social phobia and comparison to avoidant personality disorder. Journal of Abnormal Psychology, 95, 389-394.

Turner, S. M., Beidel, D. C., Dancu, C. V., & Stanley, M. A. (1989). An empirically derived

CONTROLS IN SOCIAL PHOBIA RESEARCH 91

inventory to measure social fears and anxiety: The Social Phobia and Anxiety Inventory. Psychology Assessment: A Journal of Consulting and Clinical Psychology, 1, 35-40.

Turner, S. M., Beidel, D. C., & Townsley, R. M. (1992). Social phobia: A comparison of specific and generalized subtypes and avoidant personality disorder. Journal of Abnormal Psychology, 101,326-331.

Watson, D., & Friend, R. (1969). Measurement of social-evaluative anxiety. Journal of Con- suiting and Clinical Psychology, 33, 448-457.

RECEIVED: March 30, 1995 ACCEPTED: October 4, 1995