Thoughts of agoraphobic people during scary tasks

Acknowledgement: Stephen Todd Harap and Marc Liebmann submitted portions of the current data for their respective honors theses in psychology at Lehigh University, and they contributed equally to this article.

This research was supported by Public Health Service Grant R29 MH43285.

Correspondence concerning this article should be addressed to: S. Lloyd Williams, Department of Psychology, Chandler-Ullmann Hall, Lehigh University, 17 Memorial Drive East, Bethlehem, Pennsylvania 18015-3068, Electronic Mail may be sent to: SLW1@lehigh.edu.

People with phobias act in ways that seem senseless, even to themselves. Indeed, phobia is defined in part by sufferers’ recognition that their fears are unreasonable and far out of proportion to objective dangers. Phobic people usually accept the rational proposition that nothing bad would happen if they actually did what they feared. Yet they persistently avoid doing so, to the point of bringing upon themselves serious impairment and distress in many spheres of living. Several theories explain such seemingly irrational behavior in terms of conscious cognitive processes.

Perhaps the most influential cognitive theory of phobia is the perceived danger theory of Beck (1976; Beck, Emery, & Greenberg, 1985). In Beck's view, people with phobias have a “dual belief system” in which, despite knowing rationally that danger is minimal, they also really believe that their feared object or activity would cause them physical or psychosocial harm. Beck et al. (1985) explained that danger beliefs are activated most in consciousness when the person is in close proximity to phobic stimuli:

At a distance from a phobic situation, for example, an individual may state the probability of harm in that situation is almost zero. As he approaches the situation, the “odds” generally change: they may increase from zero to 10 percent to 50 percent, and finally to 100 percent once the person is actually in the feared situation. (p. 128)

The sense of danger “increases until the idea of threat completely dominates his appraisal of the situation” (Beck et al., 1985, p. 128), and “the phobic may also actually begin to experience, in fantasy, the catastrophic consequences he fears” (p. 129). Beck et al. (1985) also explained,

Although it is true that patients generally do not volunteer much data about their thinking, particularly when they are in the throes of acute anxiety, we find, when we question a patient specifically, that his consciousness is saturated [italics added] with thoughts and images of a threatening nature. (p. 61)

Several other cognitive theories of phobia emphasize rather different possible conscious cognitive processes in phobia. One alternative posits that phobic behavior results from anticipated anxiety or “fear of fear” (Chambless & Gracely, 1989), in which phobic people escape and avoid to prevent expected aversive feelings of anxiety. A variant of fear-of-fear theory points to anticipated panic, in which agoraphobic disability springs from the wish specifically to prevent a panic attack (Barlow, 1988; Chambless & Gracely, 1989). Yet another cognitive approach, self-efficacy theory, holds that agoraphobic avoidance springs from people's strong belief that they cannot carry out the avoided activities (Bandura, 1977, 1988, 1997; Williams, 1995, 1996).

Unlike perceived danger theory, neither anticipated anxiety theory, nor anticipated panic theory, nor self-efficacy theory explicitly predicts people's cognitive preoccupations during phobia-related activities. But it seems reasonable to suppose that cognitive theorists generally would expect their favored causal cognition to be active while phobic people are distressed and dealing with scary activities and settings. Otherwise, the mechanism linking the cognitive causes to anxious feelings and phobic actions would be needlessly indirect. It is clearly of interest to know the extent to which each of the several conscious thought processes is active in phobic people as they try to do what they fear.

Cognition in anxiety states has been studied extensively with information-processing experimental procedures (Mathews & MacLeod, 1994; McNally, 1996). People with and without various anxiety problems observe verbal stimuli related or not to their problem, and their reaction times or other responses are analyzed for biases in perceiving, attending to, or remembering the problem-related verbal stimuli. The information-processing data are not directly relevant to people's thoughts while doing what they fear, because information-processing participants typically are responding to mere words in a safe clinical laboratory environment, far from actual phobic stimuli and activities. People are markedly less afraid in benign than in phobic settings, so the cognitive causes of fear are presumably less active as well. Certainly Beck's theory predicts day-and-night differences in thinking between the two kinds of settings. It is quite risky to assume that phobic people's reactions toward verbal stimuli in the laboratory tell us much about their thoughts in genuine scary environments (Williams, 1985). Moreover, even within its laboratory context, the information-processing research provides little distinctive support for any of the cognitive theories, as we consider, in some detail, in the Discussion section (below).

Several studies have measured thinking in vivo by asking phobic people to rate specific preselected cognitions while in scary settings, using a structured measurement format in which the experimenters specified the particular dangerous outcome, and participants rated its likelihood. People with severe height phobias rated their perceived likelihood of falling as they stood on progressively higher balconies of a 10-story building (Williams & Watson, 1985) or on rungs of a tall ladder (Menzies & Clark, 1995). In another study, social-phobic, height-phobic, and agoraphobic participants rated the likelihood of experimenter-specified dangerous outcomes while they were in the relevant phobic settings (Andrews, Freed, & Teeson, 1994). In all three studies (Andrews et al., 1994; Menzies & Clark, 1995; Williams & Watson, 1985), the rated likelihood of harm averaged below or far below the midpoint of the likelihood scale. For example, Menzies and Clark's (1995) height-phobic participants rated their likelihood of falling, using a 100-point scale, at less than 12. Indeed, in Andrews et al.'s (1994) study, neither the agoraphobic participants nor the height-phobic participants gave higher danger ratings than did an unscreened group of (presumably) normal people who were riding elevators or an aerial gondola on their own initiative. These structured thought measures clearly fail to support Beck's (1976; Beck et al., 1985) prediction that when phobic people are in scary settings, their consciousness will be so filled with danger thoughts that they will be nearly certain calamity is imminent.

Structured measures in which participants rate preselected dangers might not specify the individuals’ particular danger thoughts or might “put” thoughts into participants’ consciousness that otherwise would not be there. An alternative measurement strategy that avoids both potential problems is for phobic people to simply “think aloud” while doing phobia-related tasks. Two such studies coded agoraphobics’ think-aloud data only into a general negative category of “self-defeating” or “phobic” thinking (Mavissakalian, Michelson, Greenwald, Kornblith, & Greenwald, 1983; Williams & Rappoport, 1983). However, two other studies coded think-aloud data for occurrence of “catastrophic” thoughts per se (Kenardy, Oei, Weir, & Evans, 1993; Last, Barlow, & O'Brien, 1985), and both found danger thoughts to be surprisingly rare.

Last et al. (1985) had four agoraphobic participants complete weekly think-aloud assessment sessions walking through a shopping mall and found that catastrophic thoughts occurred in only 2 of 18 assessment sessions. An average of only 1% of the thoughts were catastrophic (specifically, for 16 sessions 0% of the thoughts were catastrophic, for 1 session 7% were catastrophic, and for 1 session 15%). Kenardy et al. (1993) had 20 panic-disordered people think aloud on cue as they rode trains or elevators or walked to a shopping center, and they found that catastrophic thoughts occurred in only 6.4% of the utterances, somewhat more than in Last et al. (1985) , but far fewer than one would expect from people whose consciousness was full of thoughts and images of imminent disaster.

A problem with the Kenardy et al. (1993) study is that participants were accompanied at all times during the think-aloud test by a psychologist or nurse whose presence could have interfered with participants’ thinking or reporting processes. Moreover, in both the Last et al. (1985) and the Kenardy et al. (1993) studies, participants did not appear to always have privacy during the think-aloud procedure. They might, therefore, have felt inhibited to vocally express themselves in shopping malls or trains. And in both studies, the authors measured only one of the four theoretically derived thought types (danger).

The present research sought to measure all four kinds of thinking in a large and diverse sample of agoraphobic people, by examining their thoughts spoken in privacy while they did scary tasks unaccompanied. In part, this was accomplished by reanalyzing the think-aloud protocols from Williams and Rappoport's (1983) study of driving-phobic agoraphobics, to focus on the various specific theoretically derived cognitions. Because Williams and Rappoport's (1983) participants were selected to be extremely incapacitated by driving phobia, and because the treatment they received was a relatively weak form of performance-based therapy in which anxiety was kept low at all times (cf. Williams, Turner, & Peer, 1985), the majority of the participants remained rather severely phobic of driving even after treatment. The present study gathered new think-aloud data from a second group of 28 driving-phobic agoraphobic participants who were less driving disabled on the average than the Williams and Rappoport (1983) participants. In addition, to examine generality across distinct agoraphobic fears, we gathered think-aloud data from a third group of 26 agoraphobics with claustrophobia.

The goal was to determine the frequency with which the various thought patterns occurred when agoraphobic people do the activities they fear and avoid. We also sought to examine changes in thinking with treatment and to explore the patterns of change in thinking, if any, as phobic participants progressed from easier to more difficult driving tasks.

Method

Participants

There were three different samples of agoraphobic participants totaling 74 participants. In addition, there were two small samples of nonphobic participants, together totaling 11.

Agoraphobic participants

Agoraphobic participants in Study 1 were the 20 participants in the Williams and Rappoport (1983) study of driving phobia conducted in Palo Alto, California. Their average age was 42, they had been phobic a mean of 10 years, all were women, and their mean Fear Questionnaire (FQ; Marks & Mathews, 1979) agoraphobia scale score was 25 out of a possible 40 (SD = 12). Participants in Study 1 had to meet the stringent selection criterion of showing severe behavioral disability on both of two pretreatment driving tests (Williams & Rappoport, 1983). Study 1 participants took part in 1979 (Williams & Rappoport, 1980) before the third revised Diagnostic and Statistical Manual of Mental Disorders (DSM–III; American Psychiatric Association, 1980) was published and were not formally diagnosed, but all were severely driving phobic, all had multiple other severe phobias typical of agoraphobia, and all but one had a history of multiple panic attacks.

Participants in Study 2 were 21 female and 7 male driving phobic participants in the Lehigh University Phobia Program in Bethlehem, Pennsylvania. Their mean age was 44 years old, they had been phobic a mean of 13 years, and their mean FQ agoraphobia score was 18 (SD = 12).

Participants in Study 3 were 19 female and 7 male claustrophobic participants in the Lehigh University phobia program, none of whom took part in Study 2. Their mean age was 46 years old, they had been phobic a mean of 16 years, and their mean FQ agoraphobia score was 19 (SD = 12).

We formally diagnosed all participants in Studies 2 and 3 as having agoraphobia according to criteria from the third edition and the revised third edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM–III; DSM–III–R; American Psychiatric Association, 1980, 1987, respectively; i.e., participants either had agoraphobia with a history of panic or, if they took part after 1987, panic disorder with agoraphobia). Participants in Studies 2 and 3 had widely varying pretreatment behavioral test scores, but all gave at least one anxiety rating during the behavioral test of 6 or higher on a 0–10 scale. Prior to their participation in research, 83% of the 74 agoraphobic participants in the three studies had a history of receiving professional treatment for their phobias.

Nonphobic participants

Two different sets of nonphobic participants underwent behavioral testing and thought sampling, the first (n = 6) to provide normative information on thinking while driving, the second (n = 5) to provide normative information on thinking in enclosed places. The normal driving participants were 6 women averaging 44 years of age, and they completed the same driving test in the same settings as did the phobic participants in the Williams and Rappoport (1983) study. The normal participants tested for claustrophobia were 4 women and 1 man averaging 50 years old, and we tested them in the same setting and we administered the same measures as we did the phobic participants in Study 3.

Procedure

The 20 phobic participants of Study 1 each completed four behavioral assessment procedures on the following schedule: Two weeks before treatment, the day before treatment, the day after the 2-week intensive treatment, and at a 3- to 5-month follow-up. Two participants did not complete a follow-up test, and recording equipment failure occurred in two tests, so we gathered data from a total of 76 behavioral driving tests in Study 1.

We gathered Study 2 data from 26 agoraphobic participants who had completed at least one behavioral test of driving in the Lehigh University Phobia Program after thought sampling was introduced as part of its standard driving assessment procedure. We tested participants for driving phobia if they indicated on a preliminary set of Self-Efficacy Scales for Agoraphobia (SESA; Kinney & Williams, 1988) not having high self-efficacy for driving. Seven of the 28 participants in Study 2 provided data for both pre- and posttreatment driving tests; 17 participants had pretreatment data only, because the participant was insufficiently behaviorally disabled for further inclusion in the treatment study being conducted at the time, the participant had dropped out, or the recording equipment had not operated properly. Four participants had only posttreatment data, because of equipment failure on the pretest. We gathered data from 41 behavioral driving tests in Study 2.

The data for Study 3 consisted of a single pretreatment assessment of claustrophobia gathered from each agoraphobic participant who indicated on the SESA low self-efficacy for coping with enclosed places. We only gathered pretreatment data because claustrophobia is not treated experimentally in the Lehigh University Phobia Program.

Measures

We assessed both driving phobia (Studies 1 and 2) and claustrophobia (Study 3) using four kinds of measures: participants first completed (a) paper-and-pencil structured measures of thought, followed by (b) a behavioral approach test, during which subjects (c) rated their subjective anxiety, and (d) completed think-aloud measures. Each of these is described in detail below. The assessment procedures were standardized by using manuals that clearly specified the particular settings to be used for behavioral testing, the precise sequence of procedures and measures, and the verbatim instructions to participants (see, e.g., Williams & Rappoport, 1983; Williams, Kinney, & Falbo, 1989). Assessors were undergraduate student assistants unaware of the hypotheses under investigation.

Structured measures of thought

Participants first completed paper-and-pencil measures of thought in relation to driving or tolerating enclosed places. Participants rated each type of thought separately for each task in a task hierarchy corresponding to the behavioral test. The driving tasks consisted of longer distances along progressively more challenging driving routes; the claustrophobia tasks were progressively longer intervals of time sitting alone in a dark closet (see next section for behavioral test details).

1. Perceived self-efficacy. Participants in all three studies rated their perceived self-efficacy by indicating their confidence that they could execute each task, using a scale from 0 to 100 in 10-point increments. We scored self-efficacy level in the standard way as the percentage of tasks in the hierarchy rated higher than 10.
2. Anticipated anxiety. In all three studies, anticipated anxiety was rated as the level of anxiety, on a scale from 0 to 10, that participants thought they would experience were they to perform each task.
3. Anticipated panic. Participants in Studies 2 and 3 in addition rated their anticipated panic as the perceived likelihood (from 0% to 100%) of having a panic attack during each task.
4. Perceived danger. Participants in Studies 2 and 3 also rated their perceived danger by first indicating whether a harmful event might occur during the test, then rating its likelihood (from 0% to 100%) for each task. If participants perceived no harmful outcome, the score was zero. The score for anticipated anxiety, anticipated panic, and perceived danger was the mean rating over all hierarchy tasks.

Behavioral tests

We assessed driving phobia by measuring participants’ ability to drive unaccompanied along four routes of increasing phobic difficulty. First was a route through a quiet residential neighborhood, which, in Studies 1 and 2 respectively, was 0.8 km long and 1.1 km long. Next was a route along a minor thoroughfare in a predominantly residential neighborhood with several traffic signals, 3.5 km and 2.3 km long in Studies 1 and 2, respectively. Following that was a route along a heavily traveled downtown main thoroughfare with a traffic signal at nearly every cross street, 3.2 km and 1.6 km long in the Studies 1 and 2, respectively. Last was a route consisting of three exits on a busy freeway, a length of 13 km in Study 1 and 7.6 km in Study 2.

The assessor drove participants to the beginning of each route in turn, then had the participant take the driver's seat. The assessor then described the route, told the participant to drive as far along it as she could and to return after completing it or if unable to continue along it. Participants attempted each route in turn until they were unable to complete a route, at which point the behavioral test was terminated. We verified driving performance by odometer readings taken before and after participants attempted each route.

We measured claustrophobia by having participants sit in a chair in a small dark utility closet, alone, with the door closed (but not locked), for as long as they could up to 20 min, with performance verified on the basis of direct observation and timing with a stopwatch. We scored behavioral performance in both the driving and the claustrophobia tests as the percentage of tasks completed, with a partial value added for partial task completion.

We always administered the behavioral tests in a separate session for the testing only, never on the same day that another kind of procedure took place. The tests were never presented to participants as treatment but were labeled explicitly as assessment procedures. Of the total of 143 behavioral tests with phobic participants, 91 were pretreatment tests (39 in Study 1, 26 in Study 2, and all 26 in Study 3). Of the 52 posttreatment tests, 34 followed a brief (1 to 6-session), purely performance-based exposure treatment approach that deliberately avoided attention to participants’ thinking. Thus, 125 of the 143 tests occurred either before any treatment or following an entirely noncognitive treatment. The remaining 18 tests were completed after participants had cognitive treatment consisting of 11 hr of therapist-guided performance-based exposure therapy in which the therapist encouraged participants to focus their thoughts on positive coping self-statements and to distract themselves from scary thoughts. Williams and Rappoport's (1983) analysis of the think-aloud data found no posttreatment difference between the cognitive and noncognitive groups’ general extent of “phobic thinking.”

Performance-related subjective anxiety

As driving participants completed each route, they rated how anxious they had been along it using the 0–10 anxiety scale. Claustrophobia test participants, while sitting in the closet, were cued every 2 min by a dim light lit for 5 s, to rate their anxiety on a form mounted in front of them. The anxiety score for both tests was the highest anxiety rating participants gave during the test.

Think-aloud data during behavioral testing

To collect thought samples during the driving tests, the assessor placed a lapel microphone on participants, and a tape recorder and an electronic beeper device in the car with them during all driving tasks. The same device was placed in the closet with claustrophobic participants. We instructed participants to state aloud whatever they were thinking at the moment the beep sounded. The beep was a pleasant, brief, and distinctly audible high-pitched tone that sounded every 90 s on the average. Each time it beeped it activated the tape recorder in record mode for 20 s.

In total, we obtained 1,754 20-s tape segments from driving participants (including 129 from nonphobic participants), and 273 tape segments from enclosed places participants (including 47 from nonphobic participants). A coding manual was first developed for, and applied to, the driving data, then was adapted for enclosed places. All 20-s tape segments, including those from normal participants, were transcribed, then all were assembled in randomized order, separately for driving and enclosed places. One pair of coders coded the tape segments for driving, and a different pair of coders coded the tape segments for enclosed places. The raters were first trained using a detailed coding manual and practice segments from pilot participants. One member of each pair of coders was designated in advance as the primary coder whose codes would be used to analyze results, and the other coder was to provide information on interrater reliability. The coders had access to no information about the participant or assessment phase of each tape segment, and they were unaware of any research hypotheses or expectations.

The coders independently assigned to each tape segment as many of the following thought codes as were applicable, and they assigned each code no more than once per segment.

1. Perceived danger thoughts referred to any expression related to concern about social, personal, or physical harm, danger, threat to oneself or others, or safety. The coding manual listed numerous well-known agoraphobic themes of physical and psychosocial danger, such as losing control, causing a scene, heart attack, death, injury, traffic accident, going crazy, helplessness, humiliation, embarrassment, acting foolish, choking, fainting, suffocating, losing bowel or bladder control, collapsing, and any mention of danger, threat, harm, or safety. The code was to be applied to any mention of the preceding terms, or their synonyms or cognate terms, or to any other physically or psychosocially dangerous or harmful event. We coded unpleasant sensations, fear, and panic elsewhere, as described below.

The reason for the inclusion of safety-related statements is that in the context of phobic tasks, such statements often imply clearly the possibility of danger. As Beck (1976) stated, “The anxious patient perceives danger and, lacking confidence in his capacity to cope with it, experiences wishes and thoughts about escape to safety” (pp. 39–40; cf. Beck & Clark, 1997). Examples include, “I'm telling myself it's safe here,” “I need to get back to a safe place.”

2. Anticipated anxiety thoughts referred to any mention of anxiety or related feelings in a future context. We gave raters a list of anxiety terms such as tense, afraid, fearful, frightened, nervous, scared, stressed, distressed, terrified, worried, tight, wound up, and so on, and we told them to code any future-oriented mention of such words and their cognate terms and synonyms into this category. Examples include “I wonder if that turn will get me tense,” “It seems like it's going to get real tight inside,” “I hope this anxiety will just go away.” Future mention of terms like relax or calm also were to be coded here, as in “I'll calm down by the time I'm back.” We coded mention of panic, or of present or past anxiety elsewhere, below.

3. Anticipated panic thoughts consisted of any mention of the word panic and cognate words, or anxiety attack or an attack and similar phrases, in a future context, such as “I want to be sure I won't panic” or “I think this route will bring on an attack”.

4. Self-efficacy thoughts referred to the person's ability to drive the routes or to stay in the closet. Comments on confidence or lack of it, on the ease or difficulty of doing the tasks, “trying” to do them, or ability or inability to navigate the route. Examples include “I don't think I can go any farther,” “I'll try to keep going,” “If I did it before I can do it again,” “I got through that light OK because it was green,” “Two blocks to go, a piece of cake,” “I hope I don't get lost,” and “I feel more confident than last time.” As with the other categories, both positive and negative comments about self-efficacy received this code because the purpose was to gauge the extent of preoccupation with self-efficacy, regardless of valence. Moreover, even ostensibly positive statements can have a distinctly negative sense; for example, “I think I will be able to keep going” nevertheless conveys some doubt about doing so.

5. Anxiety thoughts referred to the same kinds of fear expressions as in Category 2 (anticipated anxiety) above, except in a present, past, or indeterminant temporal context.

6. Panic thoughts referred to the same kinds of panic expressions as in Category 3 (anticipated panic) above, except in a present, past, or indeterminant temporal context.

7. Somatic thoughts referred to any mention of body parts (except idiomatic expressions such as “on the other hand”), mention of bodily sensations (dizzy, lightheaded, breathing, fatigue, physical pain, trembling, feeling chilly, etc.), or reference to sensory processes such as vision or hearing.

8. Other phobic thoughts are comments with a phobic sense to them but not codable into any of the Categories 1 through 7. This included statements projecting negativity about the driving or closet situation, as would seem more likely to be expressed by someone who has phobias than someone who does not. Examples include, “Driving on Highway 22 and not liking it,” “This doesn't seem too bad yet,” “I don't think I should go beyond Hamilton Street,” “I've always had a problem with this,” or “Twenty minutes in this closet is twenty minutes too long.” Code 8 could not be assigned with any other code.

9. Nonphobic thoughts were those that contained only phobia-unrelated, problem-unrelated content. Code 9 could not be assigned with any other code.

10. No response was the code given to segments in which the participant did not respond, or did so unintelligibly or inaudibly. Code 10 could not be assigned with any other code.

Interrater reliability

We calculated reliability for each code as percent agreement using the following formula: number of agreements divided by (agreements plus disagreements), times 100. Following are the percent agreement rates for each code, with the first figure referring to the driving data and the second referring to the enclosed places data: perceived danger 99%, 100%; anticipated anxiety 99%, 99%; anticipated panic 100%, 100%; self-efficacy 92%, 94%; anxiety 98%, 96%; panic 100%, 100%; somatic 99%, 100%; other phobic 96%, 90%; nonphobic 94%, 86%; no response 100%, 100%.

Scoring

We calculated the think-aloud scores as follows. Each behavioral test was given a single overall score for each type of thought, namely, the percent of tape segments from that behavioral test that were coded as expressing that type of thought. Then to ensure that participants with multiple tests would not weigh more heavily in any analysis than participants with only a single test, the percentages were averaged within coding categories across behavioral tests, to yield a single score for each code for each participant.

Results

Performance, Anxiety, and Structured Thought Measures

The mean levels of behavioral approach, subjective anxiety during approach, and scores on the structured (paper-and-pencil) cognitive measures by the phobic participants in the three studies are shown in Table 1. We first averaged each participant's scores for each measure across assessment occasions, and we averaged the resulting means across participants to yield the tabled values. These show that the present participants clearly have a sufficient degree and range of phobic disability and distress to permit meaningful and varied sampling of the occurrence of thinking patterns in phobic people doing what they fear. The 20 participants in Study 1 were able to perform an average of only 38% of the driving tasks across their 76 behavioral tests, quite a severe degree of phobic disability, and much more so overall than participants in Studies 2 and 3. Study 1 participants also had markedly elevated average levels of anxiety and anticipated anxiety and markedly diminished self-efficacy. Phobic participants in Studies 2 and 3 were less extremely incapacitated but had some degree of behavioral limitation, as well as substantial subjective anxiety. Every nonphobic participant displayed maximum performance and self-efficacy, zero anxiety, anticipated anxiety, anticipated panic, and perceived danger.

Percent Successful Performance, Subjective Anxiety During Performance, and Scores on the Structured Cognitive Measures, by Study

Thinking While Doing Phobia-Related Tasks

The percentages of tape segments expressing each of the various kinds of thinking in phobic and nonphobic participants by study are presented in Table 2. Within each study in Table 2, the values for the various codes sum to more than 100% because, although each code could be applied only once to a given tape segment, the segment could receive multiple different codes among Codes 1 through 8. In practice, relatively few of the segments received more than one code, only about 10% of those from driving phobic participants and fewer yet from claustrophobic participants.

Mean Percent of 20-s Tape Segments Containing Each Kind of Thought, by Agoraphobic and Nonphobic Participants, by Study

Phobic participants

Rows 1–4 in Table 2 correspond to the four cognitive theories discussed in the introductory remarks; namely, perceived danger, anticipated anxiety, anticipated panic, and self-efficacy. Participants gave little evidence of being much preoccupied with thoughts of danger or with anticipations of future anxiety or panic. Among the 48 driving phobic participants and their 117 behavioral tests that yielded 1,625 tape segments, only 11 tape segments expressed a danger thought, and 6 of those occurred in a single behavioral test. Similarly, anticipation of panic occurred in only 1 tape segment of the 1,625 and anticipation of anxiety in 7 segments. Claustrophobic participants did not once express a thought of danger or anticipated panic, and only 2 of their 226 tape segments expressed an anticipation of anxiety. Thoughts related to self-efficacy were more common, being expressed in 15% of phobic participants’ tape segments overall.

Line 5 of Table 2 shows that the single most frequent specific theme of phobic participants was thoughts about anxiety, which were expressed in 29% of their tape segments overall. Participants seemed to talk freely about being afraid, suggesting that they were not generally shy or inhibited about expressing their problem on audiotape. Panic thoughts, shown in Line 6, were unusual. Thoughts with a somatic focus (Line 7) or expressing a nonspecific phobic theme (Line 8) were more common. Thoughts of nonphobia related matters (Line 9) were expressed in over one/third of the tape segments of phobic participants in all three studies. Fewer than 5% of phobic participants’ tape segments were scored as “no response” (Line 10), indicating that participants were not notably inhibited about thinking aloud on the tape.

Nonphobic participants

Nonphobic participants had negligible or zero occurrence of most categories of thought, as Table 2 shows, except for nonphobic thoughts, which occurred in 90% of their driving tape segments and in 76% of their enclosed place tape segments. The only unexpected result among the nonphobic participants was from the enclosed places data, reported in Table 2, in which 22% of the tape segments contained a somatic thought. This was due to several participants mentioning that they felt cold in the test closet; all of the normal participants underwent claustrophobia testing during an unusually cold period in midwinter and the closet was not adequately heated.

Phobic and nonphobic participants differed significantly in their frequency of expressing several kinds of thought, as revealed by t tests with separate variance estimates for the two groups (Hays, 1963), calculated for driving phobia and claustrophobia separately. These analyses showed that the phobic participants more frequently than normal participants had thoughts concerning self-efficacy: driving phobia, t(51.6) = 6.79, claustrophobia, t(25) = 3.61, ps < .001; thoughts concerning anxiety: driving, t(37.6) = 6.96, claustrophobia, t(25) = 4.38, ps < .001; thoughts concerning anticipated anxiety: driving, t(47) = 2.22, p < .05, claustrophobia not significant; and other phobic concerns: driving, t(27.6) = 4.52, claustrophobia, t(29.0) = 3.53, ps < .001. The nonphobic participants thought of nonphobic matters significantly more than did phobic participants: driving phobia, t(14.4) = 9.80, p < .001, claustrophobia, t(6.9) = 3.56, p < .01. Phobic drivers more often than nonphobic drivers gave no response, t(47) = 3.70, p < .001. Otherwise, the two groups did not significantly differ in thinking.

Changes From Before to After Treatment

Because we assessed 27 of the driving phobic participants both before and after treatment (20 in Study 1 and 7 in Study 2), it was possible to examine the effects treatment had on what the participants thought about while driving. Treatment involved 3 to 11 hr of helping participants engage in driving activities in the natural community environment. In cases in which a participant had more than one pretreatment test, those data were averaged to yield a single pretreatment score for the participant before being averaged across participants. We followed a similar procedure with posttreatment data. The results are presented in the upper section of Table 3. The lower section of Table 3 gives the results for performance on the behavioral test, anxiety rated during the test, and the structured measures of thought.

Changes With Treatment

Of the five categories of thought that occurred with any substantial frequency, four changed as a function of treatment. Anxiety thoughts, somatic thoughts, and nonspecific phobic thoughts decreased significantly, and nonphobic thoughts increased at a borderline level of significance.

Situational Specificity of Thinking

Because the think-aloud data were gathered in relation to graduated fine-grained tests of behavioral functioning, it is possible to examine whether the difficulty level of the behavioral tasks was related to the content of participants’ thinking. The driving phobia think-aloud data were well-suited to this analysis because they were recorded separately for each of four progressively more difficult test routes (residential street, minor thoroughfare, major thoroughfare, freeway).

For this analysis, assessment occasions were selected in which the participant drove more than one route (thereby ensuring that there were both easier and harder tasks) but did not complete all four routes (thereby ensuring that the last route driven was truly more difficult than the first route, and that the person did not simply find all four routes easy, as sometimes occurred after treatment). These criteria selected 56 individual driving assessments for comparing thinking during the easy residential route to thinking during the more difficult final route driven. Self-efficacy thoughts were expressed in a mean of 11% of the easy route tape segments (SD = 14) but 19% of the difficult routes (SD = 25), a significant difference, t(55) = 2.54, p < .02. Nonphobic thinking was evident in 45% of the easy route segments (SD = 29) but dropped to 34% of the difficult routes (SD = 30), also a significant difference, t(55) = 2.95, p < .01. The differences for the remaining think-aloud categories were not significant.

Discussion

The most striking finding was that severely phobic people, directly confronting their worst fears and experiencing high anxiety near the limit of their capabilities, gave almost no evidence of preoccupation with thoughts of danger or with anticipations of becoming more anxious or panicky. This result is consistent with two previous think-aloud studies in which agoraphobic people expressed few danger thoughts in scary settings (Kenardy et al., 1993; Last et al., 1985). Had the present finding been that danger thoughts were absent in some of the 74 phobic participants some of the time, that finding might be due to individual differences in expressiveness. But the finding is that danger thoughts were absent in nearly all participants nearly all of the time.

This lack of danger content and of anticipated fear are the more intriguing because immediately before and after think-aloud testing, quite a few agoraphobic participants, although not all, indicated on structured rating forms (as in Table 1) their belief that harmful outcomes are possible or even likely (Williams et al., 1989, 1985; Williams & Watson, 1985). And on thought-listing measures after a performance test, some, but not all, agoraphobic participants indicated that catastrophic thoughts occurred (e.g., Last et al., 1985). More uniformly, on structured measures prior to and following a behavioral test, agoraphobic participants anticipate becoming highly anxious and panic-stricken in the test (e.g., Rachman & Lopatka, 1986; Telch, Brouillard, Telch, Agras, & Taylor, 1989; Williams et al., 1989, 1985; Williams, Dooseman, & Kleifield, 1984). Yet even as people are then doing the test, tackling scarier and scarier tasks in an ascending task hierarchy, they show virtually no thinking about how anxious they might become or about being harmed. If fear of fear is what motivates agoraphobic avoidance, why did our agoraphobic participants not show any evidence of thinking about future anxiety or panic even as they progressed through increasingly scary tasks? They spoke often of present anxiety, but essentially never of future anxiety.

The one theoretically derived pattern of thought that was moderately in evidence was perceived self-efficacy. Participants were actively concerned with their ability to do the tasks. The findings that nonphobic people do not think about their self-efficacy and that phobic participants think more about self-efficacy as they near the point of quitting suggest that in phobia, dwelling on one's capabilities means largely doubting one's capabilities (Bandura, 1997).

The rarity of danger thoughts and of anticipations of panic and anxiety is not due to any apparent methodological problem. The research evaluated a large number of agoraphobic people in relation to two common yet distinct domains of agoraphobic dysfunction. Participants ranged from mildly disabled to severely incapacitated and nearly all were panic victims as well. Their thoughts were sampled while they were directly engaged in performing their most feared activities. Because automobiles and closets are relatively private places, and participants were alone in them, participants could express themselves without being overheard by others, and they did not seem inhibited to speak. They almost always responded to the beep with forthright statements, often reporting fear and other phobia-related content. Their responses often filled the 20-s segments of tape, of which there were more than 1,800 from phobic people, thereby providing an ample number of comments in which they could have expressed danger or anticipated fear thoughts had they been thinking them.

The coding system was also straightforward. For the danger category, the coding manual listed essentially all of the danger terms discussed in the clinical and research literature on agoraphobia and panic. The code was to be applied to any comment mentioning these terms or their synonyms or cognate forms, or even to talk of safety as implying consciousness of danger, following Beck's (1976) suggestion. There was more than sufficient basis for coding danger thoughts as such but few occurred.

An issue, then, is whether the absence of danger thoughts can be reconciled with the statements of perceived danger theory (as quoted at greater length earlier), that when a phobic person undertakes a feared activity, “the idea of threat completely dominates his appraisal of the situation” (Beck et al., 1985, p. 128), “the phobic may also actually begin to experience, in fantasy, the catastrophic consequences he fears” (p. 129), and during acute anxiety, patients’ “consciousness is saturated with thoughts and images of a threatening nature” (p. 6, emphasis added). These statements suggest strongly that think-aloud responses, as gathered in the present research, should be replete with instances of danger ideation.

Are danger thoughts absent because they occur as brief automatic thoughts? First, automatic thoughts are conscious thoughts. They are automatic because their appearance, in consciousness, is rapid, routine, repetitive, familiar, and seemingly involuntary. Second, and more important, is that in perceived danger theory, danger thoughts are not limited to fleeting appearances by any means:

While the clinician or the researcher may believe that anxiety is totally dissociated from conscious fears, the person who is anxious does not necessarily agree, and, when questioned, he expresses a strong sense of impending disaster. The anxious patient . . . becomes preoccupied with his own anxiety symptoms and dwells on thoughts of dying or other catastrophes (Beck et al., 1985, pp. 28–29).

The present findings supported Beck's hypothesis about preoccupation with anxiety, but not about dwelling on catastrophes.

Does the mere presence of high anxiety, or of think-aloud comments about anxiety, permit us to infer that one or another of the various proposed cognitive causes is active? No, because virtually every theory of phobia, cognitive and noncognitive alike, assumes that phobic individuals experience high anxiety while doing phobic tasks, and no theorist would be surprised that phobic people also talk about their anxiety during phobic tasks. Therefore, the fact that they do so, does not in any way show the presence of danger thoughts or of any of the other cognitions. To establish that a thought is present in consciousness requires affirmatively demonstrating it.

Might the findings reflect a marked discrepancy between what people mentally experience and what they state aloud? Perhaps danger thoughts are conscious but inexpressible in words, or perhaps even nonconscious altogether. Empirically supporting either position requires a yet undeveloped method of measuring the hidden or inexpressible thoughts. Beck (1976; Beck et al., 1985) suggested that in instances in which automatic thoughts are fleeting and not spontaneously reported, people can be easily trained to attend to and report them. Such training is fraught with risk, of course, and must carefully avoid suggesting ideas to people that they would not otherwise entertain. In any case, Kenardy et al. (1993) instructed participants about reporting thoughts and still observed a low frequency of danger themes.

The core of Beck's contribution to psychopathology is his focus on the primacy of conscious reflective cognitive processes, what people think “out loud to themselves” so to speak. This general emphasis on reflective conscious thinking, which is shared by other theories examined in the present study (e.g., Bandura, 1997), sets Beck's (1976) theory sharply apart from the psychoanalytic, conditioning, and biological paradigms that preceded it. To invoke nonconscious thinking to explain the present findings is to undermine Beck's central idea and to burden oneself heavily with having to measure the content of nonconscious thinking.

It is possible to suppose that, diametrically contrary to Beck's dual-belief hypothesis, thoughts of danger are most active when outside of a phobic situation, whereas in the situation the thoughts turn to other concerns such as coping. Although such an interpretation is consistent with the present findings, if true it would call into question the whole idea of danger-driven cognitive causation of phobia. Almost any cognitive theory of phobia would expect people's experience of high anxiety, and their action to stop doing a scary task, to acquire impetus directly from the posited causal cognition. Otherwise, the theory would need to specify yet another cognitive mechanism to connect the immediate phobic response to the absent cognitive cause. This seems a roundabout way for a cognitive cause to behave.

Information-processing research also does not clearly support Beck's theory. First, as mentioned earlier, the widely used information-processing methodologies are limited for evaluating phobic thinking because they infer cognitions from molecular or verbal behavioral responses to artificial verbal stimuli in benign laboratory environments. Theories of phobic thinking must be judged, primarily and ultimately, in relation to phobic behavior and emotion in their natural contexts (Williams, 1985).

Second, even within the limitations of their laboratory settings, the information-processing findings have provided little distinctive or consistent support for any of the cognitive theories under scrutiny here, including danger theory. For example, in anxious patients, danger words should affect both attention and memory (Beck et al., 1985), but in fact they seem to not affect memory (Mathews & MacLeod, 1994; McNally, 1996).

Much confusion generally has arisen over the term threat. Many information-processing reviews refer collectively to the various phobia or panic-related verbal stimuli as “threat words.” This is quite misleading if the reader takes threat to mean danger in Beck's sense of “physical or psychosocial harm” (Beck, 1976; Beck et al., 1985). The so-called threat stimuli vary greatly from study to study. They might be exclusively danger words (e.g., Mathews & MacLeod, 1985), or predominantly nondanger words, such as anxiety terms (e.g., afraid), names of phobic stimuli and activities (driving, shopping alone), bodily processes (breathing), and so on. Thus, the term threat would more accurately be expressed as problem-related.

Anxious people show a bias in attending to such problem-related information generally, not just to danger words per se. As Mathews and MacLeod (1994) pointed out, “[biased] attentional effects are greatest when emotional stimuli match the domain of greatest concern to that subject” (p. 37), or as McNally (1996) put it,

interference [bias] for threat words in anxious patients is apparently a special case of a general phenomenon: attention is captured by cues of either positive or negative valence as long as [they are] directly related to the person's current concerns. (p. 22)

In other words, just as ornithologists show Stroop interference for bird-related terms (Dalgleish, 1995), phobic individuals show Stroop interference for phobia-related terms, whether signifying danger or not. Information-processing research thus has not provided clear distinctive support for any of the cognitive theories discussed here. Moreover, as long as information-processing research continues to gather its data outside the context of real phobia-related emotions, behaviors, and stimuli, it will remain inherently quite limited in what it can teach us about phobic thinking.

Phobic people certainly do have an elevated sense of danger on the average, but it is highly variable from one person and setting to another and is not invariably present (Last et al., 1985; Williams et al., 1985; Williams & Watson, 1985). Phobic people differ cognitively from nonphobic people in many ways, so any particular dimension of average difference, by itself, may well not be of causal importance (Williams, 1985). Studies that have taken the required step of relating the various cognitive processes to actual phobic behavior and fear arousal, have found that danger ratings generally relate poorly or at best only moderately to both phobic behavior and anxiety, whereas anticipated anxiety, anticipated panic, and perceived self-efficacy each consistently relate strongly to both phobic behavior and anxiety (Williams, 1996). The weakness of danger ratings and the strength of the alternative cognitive factors in predicting phobic behavior and anxiety holds true whether the specific dangerous outcomes are specified by the experimenter (e.g., Menzies & Clark, 1995; Williams & Watson, 1985) or by participants themselves (Williams et al. 1984, 1989, 1985).

Other lines of evidence cast doubt upon perceived danger theory and its dual-belief hypothesis. Studies in which phobic participants rated the likelihood of harm occurring, and did so both out of and within the phobic setting, found that danger ratings are not especially high in the phobic setting, they do not markedly increase and may even decrease from far to near the phobic object, and in any case, that relative proximity has a generally small effect on danger beliefs (Andrews et al., 1994; Menzies & Clark, 1995; Poulton & Andrews, 1994; Williams & Watson, 1985).

Danger theory began its life with the difficult problem that in phobia, by definition, the person knows that the phobic object is not dangerous. Danger theory tries to handle this problem by the dual-belief hypothesis that danger ideation occurs mainly near phobic settings but less when remote from phobic settings such as in a therapist's office. But this introduces a new serious problem for danger theory, namely, an inability to explain why phobic people so often act from afar to prevent contact with the phobic object or activity (Williams & Watson, 1985). Whatever process drives phobic behavior would have to be strongly active even remote from phobic stimuli to explain the routine distal avoidance so characteristic of phobias. In other words, if danger thoughts are not active from afar, danger theory cannot explain a frequent pattern of phobic avoidance; if danger thoughts are active from afar, danger theory cannot explain why phobic people know their phobias are out of proportion to objective dangers. In contrast, neither self-efficacy theory nor anticipated anxiety theory nor anticipated panic theory has any difficulty in principle explaining avoidance from any proximity to the phobic setting. The conceptual difficulties of the dual-belief hypothesis, and the major empirical failures of danger theory, suggest that danger thoughts are not central to agoraphobic thinking.

Although the absence of danger thoughts in spontaneous thinking is especially problematic for perceived danger theory, it is also a puzzle for anticipated anxiety or panic and self-efficacy theories. A perception of threat would seem necessary for people to be anxious or panicky in the first place, or what is there to be afraid of? Even self-efficacy theory requires perception of threat to account for why believing oneself unable to manage an activity sometimes results in anxiety rather than sadness or indifference. One may profoundly doubt one's ability to play the piano, yet have no fear whatsoever in relation to pianos. It is the perception of threat that makes low self-efficacy give rise to anxiety: “perceived self-efficacy to exercise control over potentially threatening events [italics added] plays a central role in anxiety arousal” (Bandura, 1988, p. 78).

In short, it is not clear why danger themes or anticipations of fear were so little in evidence in our participants’ think-aloud statements. This is a mystery whose solution should shed new light on the cognitive psychopathology of phobia.

References

American Psychiatric Association. (1980). Diagnostic and statistical manual of mental disorders (3rd ed.). Washington, DC: Author.

American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders (3rd ed., rev). Washington, DC: Author.

Andrews, G., Freed, S., & Teeson, M. (1994). Proximity and anticipation of a negative outcome in phobias. Behaviour Research and Therapy, 32, 643–645.

Bandura, A. (1977). Self-efficacy: Toward a unified theory of behavioral change. Psychological Review, 84, 191–215.

Bandura, A. (1988). Self-efficacy conception of anxiety. Anxiety Research, 1, 77–98.

Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman.

Barlow, D. H. (1988). Anxiety and its disorders. New York: Guilford.

Beck, A. T. (1976). Cognitive therapy and the emotional disorders. New York: International Universities Press.

Beck, A. T., & Clark, D. A. (1997). An information processing model of anxiety: Automatic and strategic processes. Behaviour Research and Therapy, 35, 49–58.

Beck, A. T., Emery, G., & Greenberg, R. L. (1985). Anxiety disorders and phobias: A cognitive perspective. New York: Basic Books.

Chambless, D. L., & Gracely, E. J. (1989). Fear of fear and the anxiety disorders. Cognitive Therapy and Research, 13, 9–20.

Dalgleish, T. (1995). Performance on the emotional stroop task in groups of anxious, expert, and control subjects: A comparison of computer and card presentation formats. Cognition and Emotion, 9, 341–362.

Hays, W. L. (1963). Statistics for psychologists. New York: Holt, Rinehart, & Winston.

Kenardy, J., Oei, T. P. S., Weir, D., & Evans, L. (1993). Phobic anxiety in panic disorder: Cognition, heart rate, and subjective anxiety. Journal of Anxiety Disorders, 7, 359–371.

Kinney, P. J., & Williams, S. L. (1988). Accuracy of fear inventories and self-efficacy scales in predicting agoraphobic behavior. Behaviour Research and Therapy, 26, 513–518.

Last, C. G., Barlow, D. H., & O'Brien, G. T. (1985). Assessing cognitive aspects of anxiety. Behavior Modification, 9, 72–93.

Marks, I., & Mathews, M. (1979). Brief standard self-rating for phobic patients. Behaviour Research and Therapy, 17, 263–267.

Mathews, A., & MacLeod, C. (1985). Selective processing of threat cues in anxiety states. Behaviour Research and Therapy, 23, 563–569.

Mathews, A., & MacLeod, C. (1994). Cognitive approaches to emotion and emotional disorders. Annual Review of Psychology, 45, 25–50.

Mavissakalian, M., Michelson, L., Greenwald, D., Kornblith, S., & Greenwald, M. (1983). Cognitive–behavioral treatment of agoraphobia: Paradoxical intention vs. self-statement training. Behaviour Research and Therapy, 21, 75–86.

McNally, R. J. (1996). Cognitive bias in the anxiety disorders. Nebraska Symposium on Motivation, 43, 211–250.

Menzies, R. G., & Clark, J. C. (1995). Danger expectations and insight in acrophobia. Behaviour Research and Therapy, 33, 215–221.

Poulton, R. G., & Andrews, G. (1994). Appraisal of danger and proximity in social phobics. Behaviour Research and Therapy, 32, 639–642.

Rachman, S., & Lopatka, C. (1986). Match and mismatch in the prediction of fear—I. Behaviour Research and Therapy, 24, 387–393.

Telch, M. J., Brouillard, M., Telch, C. F., Agras, W. S., & Taylor, C. B. (1989). Role of cognitive appraisal in panic-related avoidance. Behaviour Research and Therapy, 27, 373–383.

Williams, S. L. (1985). On the nature and measurement of agoraphobia. Progress in Behavior Modification, 19, 109–144.

Williams, S. L. (1995). Self-efficacy, anxiety, and phobic disorders. In J.Maddux (Ed.), Self-efficacy, adaptation, and adjustment (pp. 69–107). New York: Plenum.

Williams, S. L. (1996). Therapeutic changes in phobic behavior are mediated by changes in perceived self-efficacy. In R.Rapee (Ed.), Current controversies in the anxiety disorders (pp. 373–376). New York: Guilford.

Williams, S. L., Dooseman, G., & Kleifield, E. (1984). Comparative effectiveness of guided mastery and exposure treatments for intractable phobias. Journal of Consulting and Clinical Psychology, 52, 505–518.

Williams, S. L., Kinney, P. J., & Falbo, J. (1989). Generalization of therapeutic changes in agoraphobia: The role of perceived self-efficacy. Journal of Consulting and Clinical Psychology, 57, 436–442.

Williams, S. L., & Rappoport, A. (1980, September). Behavioral practice with and without thought modification for agoraphobics Poster presented at the meeting of the American Psychological Association, Montreal, Canada.

Williams, S. L., & Rappoport, A. (1983). Cognitive treatment in the natural environment for agoraphobics. Behavior Therapy, 14, 299–313.

Williams, S. L., Turner, S. M., & Peer, D. F. (1985). Guided mastery and performance desensitization treatments for severe acrophobia. Journal of Consulting and Clinical Psychology, 53, 237–247.

Williams, S. L., & Watson, N. (1985). Perceived danger and perceived self-efficacy as cognitive determinants of acrophobic behavior. Behavior Therapy, 16, 237–247.

Submitted: January 24, 1995 Revised: October 29, 1996 Accepted: May 14, 1997