RUNNING HEAD: Differential Safety Behavior Effects



Abstract

Exposure therapy is one of the most powerful and generalizable treatments available to cognitive-behavioral psychologists. During exposure, subjects often engage in behaviors that reduce their anxiety. Some studies indicate that these safety behaviors reduce the effectiveness of exposure treatment. This paper reviews the evidence for two major models: distraction, in which safety behaviors draw attention away from disconfirming evidence, and misattribution, in which subjects believe that their own actions prevented the feared outcome, causing them to reconceptualize safe (fear-disconfirming) situations as dangerous (fear-confirming). I advance a third hypothesis, that safety behaviors are detrimental only when they restrict access to disconfirmatory information.

Spider-phobic subjects were given 15 minutes of exposure to a live tarantula. One group was instructed to trap the spider in a small space. Another was given a shield but did not restrain the tarantula. They were compared to each other and to no-safety-behavior controls. The number of subjects was too small to analyze the data on fear reduction and fear reduction maintenance, but the safety behaviors were found to be similar enough that they control for all factors other than access to disconfirmatory evidence. A larger study is planned with minor modifications.

Acknowledgements

My work and I are dedicated to:

Christine Hoff, who helps me think, for mutual support.

John Kraemer, who always has something to show me, for more mutual support.

Steven Hazel and Liza Ferneyhough, ever-tolerant friends, for yet more support – mostly in my direction.

Barry and Linda, my parents, for believing.

Tarantula Brad and Tarantula Skywise – very cooperative, for giant spiders.

A Note on the text

My thesis was submitted with only cosmetic changes to the psychology department as a psychology honors project. The language and style are not what I would choose for conveying the vitality and excitement of this work to my Plan II colleagues. I also didn’t think up the big ugly “Insert figure 1 about here” notes. However, I have preserved them for authenticity’s sake.

Plan II readers are encouraged to see the personal note in the afterward. This thesis may also be read online at . Comments and questions are welcome. Please send email to mikethecrow@azrael.dyn. or traiben@.

Table of Contents

Abstract 2

Acknowledgements 3

A Note on the text 4

Table of Contents 5

Introduction 6

Cognitive and behavioral etiology of safety and anxiety 7

Safety behaviors in exposure therapy 10

Theoretical explanations for safety behavior effects 12

Weaknesses of current theories 16

Escape vs. coping: an alternate distinction 19

Reconceptualization of past studies 21

The current study 24

Methods 27

Results 36

Manipulation Checks 36

Outcome measures: 37

Discussion 38

Manipulation Checks 40

Outcome Measures 41

Changes to the Protocol 42

References 43

Tables 46

Appendix A – Instructions and instruments 53

Afterward 64

About the Author 66

Introduction

Cognitive-behavioral therapy is the current treatment of choice for anxiety disorders, and its paradigm of exposing subjects to their feared stimulus has been widely successful in reducing anxiety and avoidance in panic disorder (Craske & Rodriguez, 1994; Marks, et al., 1993), claustrophobia (Thorpe & Burns, 1983; Sloan, 2000), and simple phobias (Rachman, 1990b). The clear effectiveness of this general method has led to more nuanced explorations of its differential effects across subjects and situations. Currently, much research is focused on how subjects respond when placed in feared circumstances, and how these responses can act as moderators to treatment effectiveness.

One promising avenue is the use of safety behaviors, actions that fearful individuals use to reduce anxiety. Safety behaviors can take the form of overt actions, covert behaviors (thoughts), or comforting or protective objects. For example, an individual with panic disorder might react to the fear of having a heart attack by moving very slowly, by concentrating hard on relaxing thoughts, or by fingering an alprazolam tablet. It appears that these safety behaviors can potentially interfere with the benefits of exposure therapy, but the evidence is far from conclusive. This paper reviews the cognitive mechanisms of safety behaviors, their ability to reduce fear, and theories explaining their effects on exposure, culminating with a proposed revision to current categories of safety behaviors and their expected effects. Initially, our examination of safety behaviors must begin with the origins of anxiety.

Cognitive and behavioral etiology of safety and anxiety

Procedures for experimentally creating a fear response to a neutral stimulus are well-established and satisfactorily explained by classical conditioning: when a conditioned stimulus (CS) is followed by a punishing unconditioned stimulus (UCS), the negative reaction to the UCS becomes associated with the CS as well. Subjects, both animal and human, will then attempt to avoid the CS, or respond to its presence with anxiety and escape behavior (Gray 1987). Often this escape behavior is highly successful, and the feared UCS is rarely or never encountered. With the negative stimulus gone, one might expect the avoidance or escape behavior to self-terminate through extinction (i.e., the CS triggering escape is repeatedly encountered without the aversive UCS). On the contrary, escape behaviors are notoriously persistent. Dogs will expend great effort to flee a negative CS, continuing for hundreds of trials during which no UCS is ever administered (Kamin, 1956). Analogously, humans with anxiety disorders will continue to avoid situations they perceive as dangerous, even when those situations do not lead to negative experiences (because the situation is inherently benign, or because they flee from it). Why do these natural outcomes differ so markedly from the successes of exposure therapy noted above?

The theoretical model to which most modern understanding of avoidance behavior traces back is Mowrer's two-stage theory (Gray, 1987). Classical conditioning is responsible only for the first stage, the initiation of avoidance. In the second stage, the CS elicits anxiety on its own, and escape functions to reduce that anxiety. This negative reinforcement strengthens the escape behavior. At the same time, a successful escape gives the subject no opportunity to observe that the feared situation may actually be benign. Thus, the interpretation of avoidance as a successful flight from danger is preserved. This can explain both the problem of persistence and the seeming contradiction of exposure therapy. In exposure therapy, subjects are required to tolerate the CS rather than experiencing it briefly and "escaping." As a result, they have an opportunity to learn that their perceptions of danger may be incorrect. This cognitive process is known in cognitive-behavioral theory as disconfirmation (Beck, 1976), and it is the source of the therapeutic dictum that exposure must not be terminated until anxiety has decreased. To do otherwise would simply constitute another reinforcement of escape (Rachman, 1986)

Mowrer’s model is internally consistent and can explain the essential origins and maintenance of anxiety disorders. However, it does not account for some common clinical phenomena. Often, an anxiety reaction may develop without a distinct fear-inducing event that could be responsible for the original respondent conditioning. Rachman's (1990a) three-process theory explains many such cases as arising from cognitive events. An individual who has had no direct interaction with snakes may develop a phobia through modeling (e.g., seeing a peer bitten by one) or abstract learning (e.g., finding out that a certain snake which was previously regarded as harmless is actually poisonous). In these cases, the individual uses new information to mentally associate the CS (the snake) with the feared UCS (pain or danger). The resulting fear is the same as that produced by a negative experience in vivo.

Once the initial fear is established, these variations meld seamlessly with the operant-avoidance stage of Mowrer's theory. A more drastic departure from Mowrer is safety-signal theory, which suggests that escape and avoidance should not be modeled as flight from an undesired stimulus but rather as approach towards stimuli that indicate safety. Rachman (1984) cites the significant percentage of individuals who develop agoraphobia without ever having any fear-provoking experience (even an informational one) outside the home. In many of these cases, however, they did suffer the loss of a close relative or spouse who could be conceptualized as a major source of security and reassurance - a safety signal. Additionally, many agoraphobic and panic disorder subjects show significant reduction in anxiety when accompanied by a trusted person (Rachman, 1984). These observations are difficult to fit into a purely avoidance-based model, but accord well with the idea that certain people use cues to infer safety as well as danger.

Additional support for the use of safety information in the generation or suppression of anxiety comes from an animal study by Rescorla and LoLordo (1965). Dogs were shocked in association with a CS, unless a second CS (the safety signal) was also presented. The dogs were then shocked in association with a new CS and taught to escape from it. When this new CS was presented in conjunction with the safety signal, the dogs’ escape behavior decreased even though the two had never been paired before, and the dogs were still being shocked. Evidently the safety signal was interpreted in a generalized fashion, distinct from any particular fear-provoking stimulus

Humans also appear to take advantage of safety information that is not immediately relevant to the feared stimulus. Rapee, Tefler, & Barlow (1991) studied the severity of response to a CO2 challenge in subjects with panic disorder with agoraphobia when the circumstances of administration were varied. Subjects’ impressions of the experimenter's orderliness and professionalism, and the availability of help available were found to correlate inversely with severity of response. However, this result was found only in a retrospective analysis across groups. The experimenters were unable to manipulate subjects' perceptions of the situation, suggesting that the causation may be bidirectional: subjects who are less prone to feel threatened may be more attentive to the information conceptualized as safety signals.

Safety behaviors in exposure therapy

Research on the potential of safety behaviors to reduce anxiety in the short term has generated interest in their effects on therapy for the anxiety disorders, the goal of which is long-term and generalized anxiety reduction. Based on the construct of safety signals, Rachman (1983, 1984) suggested that therapy for agoraphobia incorporate safety-inducing stimuli into exposure treatment. For example, a man might not want to leave the house because he relies on the reassurance provided by his wife’s presence, and going out typically involves leaving her behind. His wife could be asked to wait for him some distance outside the home, allowing him to experience going out in public as movement towards safety. Through both operant conditioning (with travel as the desired behavior and his wife as a positive reinforcer) and by encouraging exposure (with the associated disconfirmation), this method would ultimately help eliminate his avoidance and anxiety in a wider variety of circumstances.

Supporting the proposition that safety behaviors can usefully be integrated with exposure treatment are two studies adding an element of escape to a validated exposure paradigm for agoraphobia. In one (de Silva & Rachman, 1984), subjects were told to terminate exposure as soon as their fear reached a predetermined high level; in another (Rachman, Craske, Tallman, & Solyom, 1986) they were told to leave but return when their fear had decreased to a predetermined low level. These experimental groups showed fear and avoidance reduction equal to or greater than that of exposure-only controls. Smits and Telch (2000) found similar effects for safety behaviors in the realm of simple phobias, using a graded-exposure treatment for spider-phobia. Subjects viewed, approached, and eventually touched a large spider, with the option of wearing various pieces of protective gear or choosing to leave the situation if they felt too anxious. Groups with and without these safety behaviors available showed no difference in reduction of fear or disgust at post-treatment or two-week follow-up.

The studies reviewed above offer theoretical and empirical support for the proposition that safety behaviors may be integrated with exposure therapy. Allowing subjects to reduce their immediate feelings of fear would be beneficial in improving compliance with exposure therapy, which is understandably unpleasant.

However, closely-related procedures have also produced contrasting results. A paradigm similar to Smits and Telch’s (2000) which used claustrophobic subjects produced significantly less fear reduction when subjects were permitted to stand near an escape route, talk to the experimenter, or open a small window (Sloan, 2000). Salkovskis, Clark, Hackmann, Wells, and Gelder (1999) compared exposure in which agoraphobic subjects were permitted to maintain already-present safety behaviors with exposure including specific instructions not to engage in safety behaviors. Eliminating safety behaviors led to a greater reduction in anxiety and catastrophic beliefs. Similar results were achieved by Wells, Clark, Salkovskis, Ludgate, Hackmann, and Gelder (1995) using a within-subjects variation: trials in which subjects were allowed to use their safety behaviors produced less reduction in anxiety than trials in which safety behaviors were prohibited. Soltysik, Wolfe, Nicholas, Wilson, and Garcia-Sanchez (1983), experimented with an animal conditioning model, in which cats learned that several CSs preceded shock. A large number of extinction trials were then presented in which the CSs were not paired with shock. Additionally, one “protected” CS was consistently paired with a new stimulus, a CI (conditioned inhibitor). The cats showed a reduction in physiological response to all of the CSs. However, when the protected CS was presented without the CI, they reacted as if no extinction had taken place. Evidently, the CI kept the absence of shock from being directly related with the CS. In an analogy with human anxiety, the CI is a safety signal, and its use detracts from the disconfirmation one would expect when the feared circumstances (the CS) are experienced without aversive results.

It is difficult to draw any general conclusions from the results discussed thus far. Some combinations of safety behaviors, specific situations, and subject variables detract from the effectiveness of exposure therapy, but very similar procedures have demonstrated no such interaction. A review of current models for the effects of safety behaviors is necessary to make the finer distinctions that will be necessary to determine which behaviors will interfere and which will not.

Theoretical explanations for safety behavior effects

Two major safety behavior theories are distraction from evidence and misattribution of safety:

The distraction hypothesis suggests that we cannot look merely at the objective exposure scenario; we must evaluate the degree of functional exposure, based on both the available evidence and the receptivity of the subject. As an extreme example, consider two spider-phobic subjects, one alert and one in a coma. An exposure exercise that helps the first disconfirm his fears will obviously have no effect on the second. A more subtle investigation (Grayson, Foa, & Steketee, 1982) exposed OCD subjects with fears of contamination to two sessions with their most feared object. Subjects were forced to concentrate on and talk about the feared object in the first session, and were distracted with video games in the second; a second group received the treatments in the opposite order. Groups showed equivalent within-session anxiety reduction, but the ones who received distraction first returned to a high level of fear on the next session, while the group instructed to focus on the object maintained its gains. Craske and Rodriguez (1994) report similar results for distracted vs. focused exposure in agoraphobia, suggesting that the need for attention as well as mere presence generalizes across types of anxiety and exposure. Distraction from the feared stimulus thus appears to be sufficient, if not necessary, for the detrimental effects of safety behavior utilization to occur.

Additional evidence for the distraction hypothesis comes from manipulations that heighten attention. In addition to a safety behavior and a no-safety-behavior group, Sloan's (2000) claustrophobia study included a “guided threat reappraisal” (GTR) group in which subjects not only had no safety aids, but were explicitly encouraged to "focus on their perceived threat… and to look for evidence that would weaken their belief in the threat” (70). As distraction theory would predict, increased focus on threat produced even greater treatment gains than were observed in the group given no safety aids and no other instructions. Furthermore, these gains showed the same pattern as in the studies above: in-session fear reduction was equivalent among the three groups, but maintenance at follow-up was lowest for the safety aid group and highest for the GTR group. It appears, then, that acute anxiety reduction occurs on a preattentive level, similar to simple sensory habituation. More clinically-relevant long-term anxiety reduction requires cognitive involvement in exposure and suffers when the subject’s attention is divided.

Competing with distraction theory is Salkovskis’s concept of misattribution (Salkovskis, 1991, Salkovskis & Clark, 1991), which also brings the focus from the environment to the cognitive conceptualization of exposure. Subjects who use safety behaviors do not experience disconfirmation because, in their understanding of the situation, they did not experience safety. Rather, they had the experience of being on the verge of a terrible disaster, which their prompt action was able to prevent. A woman with panic disorder who is frightened when her pulse rises might respond by sitting down, which she believes will prevent a heart attack. She is correct that she will not have a heart attack, but rather than reassuring her the experience provides further evidence of the need for constant vigilance and worry. Indeed, subjects with panic disorder report safety behaviors which are plausible, though often technically incorrect, responses to their individual fears: those who fear heart attack tend to request aid and decrease activity, those who fear fainting hold on to things, and so forth. Few report mismatched or magical anxiety-reducing behaviors (Salkovskis, Clark, & Gelder, 1996). It is important to note that there is nothing inherently pathological about these safety behaviors, given the subjects’ understanding of their situations. Their perceptions become destructive as a result of the constant stress and restriction of activity induced when common and innocuous stimuli are perceived as threatening (Salkovskis & Clark, 1991).

Attribution of improvement was probed directly in a study comparing daily alprazolam, exposure, and combined treatment in panic disorder (Basoglu, et al. 1994). Among subjects who improved during the eight weeks of therapy, degree of relapse in both drug and placebo groups was most strongly predicted by the extent to which they attributed their gains to the medication. Consistent with the misattribution hypothesis, subjects will apparently experience disconfirmation only if they do not have some other explanation for safety. However, the generalizability of these results is questionable. Many panic-prone individuals do use anti-anxiety drugs when they feel they may have an attack, but this ad-lib usage differs from the scheduled regimen used in the study. Experiments with more environmentally valid safety behaviors such as ad-lib benzodiazepene usage, sitting still, and deep breathing have not directly probed attribution, something which future studies should incorporate. Additionally, subjects in the Basoglu study were asked to report the attribution for anxiety reduction that took place over several weeks, rather than their safety in a specific situation. Studies investigating misattribution have concentrated on attributions of safety rather than attributions of fear reduction. Because some studies (e.g., Sloan, 2000) have shown that subjects with safety aids may temporarily improve before relapsing, misattribution theory would be significantly strengthened if such individuals also attributed their short-term reductions in fear to having safety aids.

Both distraction and misattribution prevent subjects from being meaningfully exposed to the CS. Under distraction, the subject is present but unable to fully process the sensory information, thereby reducing its impact. Under misattribution, safety behaviors, though subtle, actually have the same effect as escape. When subjects believe they are actively preventing the feared outcome, they never learn that the aversive UCS (e.g., passing out in public) will not follow the CS (symptoms of panic or an actual panic attack) even without the safety behavior. The differences may be subtle: an individual who avoided even tame dogs if they weren't leashed would be misattributing safety to the leash, but one who focused attention on a dog's leash as a way of ignoring the dog would be using distraction.

Weaknesses of current theories

It would seem a simple task to distinguish between evidence for the distraction model and evidence for the misattribution model. Distraction models hypothesize that anything that moves attention away from the feared stimulus will interfere with exposure, regardless of its relevance to the situation. Misattribution models require no shift in attention, but predict effectiveness only for behaviors that could be conceptualized as somehow addressing the perceived threat. Yet despite the variety of experiments discussed above, it remains difficult to find evidence that clearly and exclusively supports one model or the other. Any evidence supporting misattribution can also be construed as supporting distraction, because safety behaviors offer an alternative attentional focus. Checking for signs of choking or working to move with exaggerated slowness can distract from feared stimuli just as well as something entirely irrelevant. It may be that distraction is the mechanism for interference, but misattribution is required for distraction -- subjects may be unable to draw their focus away from a threat unless they feel they are doing something to address it. Even the fear-behavior congruence found by Salkovskis, Clark, & Gelder (1996) only demonstrates that subjects initially select their safety behaviors because they think they will prevent disaster. The correspondence does not necessarily indicate anything about why these behaviors can interfere with exposure.

Along similar lines, Sloan (2000) argues that any experiment requiring safety behavior withdrawal (e.g., Wells, et al. 1995; Salkovskis, et al. 1999) is inherently manipulating attentional focus: when there are fewer options for focus, subjects are bound to take greater notice of the threatening stimulus. Unfortunately, a similar problem exists in the interpretation of Sloan's experiment, and emphasizes the difficulty of independently manipulating attribution and attention. Although subjects in the control group with neither safety behaviors nor GTR (Guided Threat Reappraisal, the technique used to heighten attention) did not have any safety aids provided, they were left free to use any pre-existing, idiosyncratic techniques, and such use was not assessed. The assumption that they differed from the GTR group only in terms of attention, then, is unfounded. They may have experienced a degree of misattribution, which GTR would have prevented. Similarly, their difference from the safety behavior group would then be in the type of safety behaviors used, rather than their presence or absence. Resolving such confounds will require that subjects in all conditions be prevented from using any safety behaviors except for those provided. This will ensure that only the prescribed safety behaviors influence the results.

An experiment designed to compare misattribution and distraction hypotheses would require a clear distinction between plausible and implausible safety behaviors, which may require more subtlety than is readily apparent. Many panic subjects’ feared catastrophes are entirely mental. A common error is to misinterpret an anxiety-induced feeling of unreality as a sign that one is about to go mad, and to attempt to prevent this by asserting control over one’s thoughts. Focusing on an irrelevant stimulus or cognitive task could then be given credit for continued sanity. Even simple phobias may have such a “fear-of-fear” element; in spider-phobia there is a strong correlation between fear and disgust towards spiders that is independent of the belief that spiders can cause harm of any kind (Armfield & Mattiske, 1996). Many spider-phobics endorse the statement “if I can’t escape from the spider I will go insane” (Salkovskis, 1991), again showing a fear of their reactions which is unrelated to the spider's behavior. Subjects whose fear is primarily of this type might attribute safety to any distracting behavior. A design using behaviors intended to be purely distracting risks having its results obscured by these subjects, whose prevalence in the population is unknown. Nonetheless, the simple phobias are considerably more externally-directed than other anxiety disorders, and thus provide a clearer window into the distinct features of behaviors that are distracting as contrasted with behaviors that seem to provide safety.

A further difficulty in testing the two models of safety behavior interference is that they are by no means mutually exclusive. Both may be active concurrently, and any interaction is not easily predictable from theory. It may be that both attentional focus and lack of an alternate explanation for safety are required for disconfirmation to occur. Alternately, distraction may interfere with gathering disconfirmatory evidence, but interfere equally with gathering the misperceived confirmatory evidence that contributes to continued fear and misattribution of safety. It may also be, as argued above, that aids to which safety can be misattributed have greater potential to distract from the feared stimulus than ones that seem irrelevant.

Finally, the distinction between the two models may not be as strong as it seems. As the GTR evidence (Sloan, 2000) indicates, distraction is just one manifestation of a larger attentional variable, which can be lowered by distraction or raised by intentional concentration. Furthermore, this variable can be reconceptualized as dealing with locus rather than quantity of attention, with distracting behaviors moving focus away from the feared stimulus. But even a "purely misattributional" behavior (i.e., one requiring little thought or effort) may lead to a refocusing of attention on the behavior's ability to forestall the feared disaster. This is consistent with both hypotheses: attention is not focused on the feared stimulus and its failure to cause harm, while its new focus is generating the danger-confirming perceptions which fuel misattribution.

Escape vs. coping: an alternate distinction

Rather than blaming the complex entanglements between misattribution- and distraction-linked behaviors on our experimental methods, it may be that this is not the correct primary distinction to use. Because the basic mechanism of both distraction and misattribution is interference with the gathering of disconfirmatory evidence, we should consider first trying to distinguish between safety behaviors that interfere and those that do not. De Silva and Rachman (1983) found anxiety reduction in agoraphobic subjects who were removed from exposure at high levels of anxiety. They speculate that this lack of interference may have been due to the relatively small number of trials on which subjects actually escaped. That is, in some trials subjects used the safety behavior of escape, which certainly would prevent the observation of disconfirmatory evidence. But in the majority of cases, they simply knew that they might escape. Knowing that one's anxiety will not be allowed to become unbearable clearly does not interfere with exposure to disconfirmatory evidence, but it could easily reduce anxiety. Smits and Telch (2000) gives a similar explanation for his findings of no interference by suggesting that safety behavior utilization subjects might not have attributed the spider's failure to bite to their protective gear. Armored subjects might feel less anxious due to reduced expectation of harm or disgusting contact if they are attacked, but their protection does not stop them from seeing that the spider has no intention of doing so.

Another precursor to this alternate distinction appears in discussions of misattribution (e.g., Salkovskis, Wells, & Gelder, 1996), which frequently note that subjects may be permitted to retain comforting behaviors as long as they acknowledge that they provide a feeling of safety as opposed to actual protection from harm. However, these discussions frame the distinction as entirely cognitive, related to subjects' understanding of their own actions. They do not deal with the possibility that some behaviors are more amenable to anxiety reduction and others to misattribution of safety. It is true that panic subjects might view breathing deeply in response to anxiety as either a chance to collect their thoughts or as a necessary response to prevent themselves from going berserk. However, subjects who respond by taking an alprazolam tablet are in a similar position to our comatose spider-phobic. Completely eliminating their anxiety also eliminates the relevance of the potentially disconfirmatory evidence. If the use of medication offers a clear and correct rationale for the absence of fear, they have no reason to consider the possibility that the situation might be inherently benign. They can reassure themselves that they were never in any real danger, and indeed such educational interventions can help immensely (Salkovskis & Clark, 1991). In terms of effects on simple exposure, though, this type of behavior is clearly counterproductive.

Interference with the gathering of disconfirmatory evidence is a critical determinant of whether a behavior will interfere with exposure. Whether this will occur is affected by both the nature of the behavior and the subject's understanding of the behavior's function. This study seeks to formalize this new distinction between safety behaviors: escape behaviors are ones that hinder or prevent the acquisition of disconfirmatory evidence; coping behaviors do not. Thus, a spider-phobic who leaves the room and one who shuts his eyes and ears until the experimenter removes the spider would both be employing an escape behavior; one who wore a suit of armor or observed from behind a glass barrier could see that attack was not imminent, and so would be employing a coping behavior. The term "escape" is something of a misnomer because it is quite possible for subjects to remain in the presence of the feared stimulus; it is used to emphasize the fact that, whether physically or cognitively, they are no longer exposed to the specific situation they fear.

Most distraction behaviors can be reframed as mild escape behaviors, which prevent the subject from attending to disconfirmatory evidence. However, misattributional behaviors may be either escape or coping. The protective gear in the previous paragraph would be considered a misattribution-type aid, because it would keep the subject safe in the event of an attack. Asking the experimenter to restrain the spider, which is also misattributional, is an escape behavior because it does not provide any opportunity for disconfirmation. The escape-coping distinction is between behaviors that provide safety contingent on the non-occurring disaster and those that seem to actually forestall the disaster.

Reconceptualization of past studies

These new categories lead to a more complex hypothesis regarding safety behavior effects on exposure: escape behaviors should hinder long-term fear reduction; coping behaviors should not. By dividing behaviors this way, the apparently contradictory results of previous studies can be reconciled. Sloan (2000) separated attention and attribution of safety, but the safety-behaviors group had several options, among which were both escape and coping behaviors. Subjects who stood near the door were still in the situation and disconfirming their fears of suffocation; those who opened a window actually let in air, changing the situation so that it was no longer the one they feared. Smits and Telch’s (2000) safety behavior subjects who chose to leave the room were obviously escaping, while those who opted to wear galoshes and goggles were using coping behaviors. There may have been some interference with exposure among subjects who primarily or exclusively used escape behaviors. However, analysis of differential effects of safety behaviors was not provided and, because subjects had many opportunities to utilize different behaviors, may not be possible.

Rachman’s findings of improvement in agoraphobic subjects who were allowed to escape when anxiety was high (1986) may be due to the fact that his “escape” was actually a coping behavior. Subjects retreated to a safe distance that did not entirely remove them from the (individualized) exposure situation. They may have viewed this experience as remaining exposed with a reduced level of anxiety. Another potentially significant element is the manner in which escape was triggered: subjects were regularly asked to report their fear on a SUDS (subjective units of distress) scale ranging from zero to 100, and told to “escape” after reporting a 70. These levels were predetermined and applied across the board to all subjects. For many, the time required for anxiety to reach this level may have provided ample opportunity to gather disconfirmatory evidence. Recalling the evidence that there is a strong cognitive, between-sessions element to long-term fear reduction (Grayson, Foa, & Steketee, 1982; Sloan, 2000; Craske & Rodriguez, 1994), there is no reason to assume that failure to reduce in-situation anxiety on a single exposure trial indicates a failure of exposure to operate on the more general level.

Studies that involved the cessation of already-present safety behaviors (Salkovskis et al., 1999; Wells, et al., 1995) require little reinterpretation to fit into this model. Indiscriminate removal of safety behaviors clearly includes removal of all deleterious ones. The success of these methods merely provides confirmation for the unsurprising proposition that people who have long-term anxiety disorders tend to use safety behaviors that do not contribute to their resolution.

This perspective also offers an opportunity for a rapprochement with Rachman's work on safety-signal theory (1983, 1984). If a safety signal is conceptualized as similar to a safety aid, then neither misattribution nor distraction would predict any benefits from its use. Rachman's suggestion that allowing subjects to travel towards home or loved ones would be either harmful or irrelevant. But if the knowledge that a feeling of safety awaits at the end of the exposure task merely reduces anxiety, then having this knowledge is a coping safety behavior, and such an arrangement can be seen as a positive aid to exposure.

One remaining complication is that if subjects fear their own reaction to anxiety, any behavior that reduces anxiety will constitute an escape. The need to carefully distinguish fear of external disasters from fear of fear remains. Subjects must still be probed to determine the exact nature of their concerns and the function of their safety behaviors, and studying simple phobias is still preferable to studying panic disorder, which is fundamentally a fear of physiological fear reactions.

The current study

In summary, a study designed to investigate this new model should supplement the standard manipulations of safety behaviors with the following:

1) Use of simple phobias to reduce the escape properties of anxiety reduction,

2) Probing of attention to ensure that ostensible coping behaviors do not constitute an escape from the feared stimulus via distraction,

3) Probing of covert or idiosyncratic safety behaviors used in all conditions,

4) Probing of subjects' attribution of safety immediately following exposure,

5) Distinction between fears of external disasters and fears of reactions (e.g., going mad in response to anxiety).

The current experiment combines these refinements in a study of spider-phobic subjects, comparing long-term reductions in fear and avoidance among three groups: escape behavior utilization, coping behavior utilization, and no-safety-behavior control. All subjects were exposed to spiders for a period of time sufficient for their fear to decrease. Their anxiety on approaching a spider was assessed with a behavioral approach test (BAT) prior to treatment, post-treatment, and at two-week follow-up. On all three assessment occasions, subjects also completed questionnaires dealing with specific fears regarding spiders to provide additional information about their anxiety and test the accuracy of their knowledge about spider behavior.

During exposure, but not during the BAT, the subjects in the coping behavior group were given a shield to keep the spider from touching them. Thus they were protected from the potential of attack, but were free to observe the spider’s behavior. Escape behavior subjects were instructed to use a long-handled plastic cube to trap the spider from a distance. Control subjects were exposed under the same conditions but had no safety behavior prescribed. All three groups were instructed to focus their attention on the fact that a spider was present and not told to take any actions other than those explicitly mentioned.

That the subject is holding a shield should have no plausible effect on the spider's behavior. Thus, subjects’ perceived sense of safety should not have interfered with their ability to see that the spiders were peaceful and lethargic. The two shields should offer equal degrees of protection, making subjective assessment of their protective quality equal. Subjects in both conditions engaged in a safety behavior that could be credited with keeping them safe from harm, so the potential degree of misattribution was constant. The instruction to focus on the spider's presence (along with probing of attention to ensure that it was effective) equalized attention as well. If subjects in the escape condition show any less reduction in fear, the critical factor must be their inability to observe the spider’s natural behavior-- that is, the ability of the safety behavior to block disconfirmatory evidence.

It was hypothesized that:

1) The groups would show no significant differences in attentional focus or utilization of covert safety behaviors,

2) the two experimental groups (coping and escape) would show no differences in attribution of safety to their safety behaviors or sense of threat from the spider.

3) subjects in the escape behavior group would show less reduction in fear of spiders at follow-up than subjects in the other two groups,

4) subjects in the escape behavior group would show less improvement in their knowledge of how spiders can be expected to act than subjects in the other two groups, and

5) if subjects in the coping behavior group showed less long-term fear reduction than subjects in the control group, their fear reduction would still be greater than that in the escape behavior group.

If the hypothesized differences between groups bear out, they will support the current conceptualization and provide evidence against the distraction and misattribution hypotheses.

Methods

Design:

Spider-phobic subjects received 15 minutes of self-guided graduated exposure to a large tarantula in one of three conditions: escape behavior utilization, coping behavior utilization, and no-safety-behavior control. Subjects' spider-related anxiety was assessed with a Behavioral Approach Test (BAT) prior to treatment, at post-treatment, and at two-week follow-up. Prior to each BAT, subjects completed questionnaires dealing with specific fears and beliefs regarding spiders in order to provide additional information about their anxiety, expectations, and degree of disconfirmation.

Subjects:

Subjects (n=10) were drawn from a large (n=~500) pool of introductory psychology students at the University of Texas at Austin, and received course credit for participation. In the current pilot study, all subjects were white females. A larger spider-phobia study at the same university indicates that in a larger sample, all subjects will be female but ~50% will be non-white (Smits & Telch, 2000).

Due to failure to attend all sessions of the study, only eight subjects provided data on treatment, and only seven provided data at follow-up. It would be inappropriate to attempt a full analysis on a population of this size, but they can be used for manipulation checks and to tentatively assess the effectiveness of treatment.

Measures:

All non-standard measures are reproduced in Appendix A.

Self-report questionnaires, BAT

Armfield and Mattiske’s Disgust Scale (Armfield and Mattiske, 1996): This test presents 8 statements about the disgust-provoking aspects of spider and spider behavior. The subject rates each statement on a scale from 0 (strong disagreement) to 6 (strong agreement). Scores range from 0 to 48, with higher scores indicating a greater feeling of disgust towards spiders.

Spider Belief Questionnaire (SBQ): The SBQ (Arntz, Lavy, Van den Berg, & Van Rijsoort, 1993) measures the strength of various beliefs related to spiders. Subjects rate their belief in 42 spider-related statements dealing with qualities such as unpredictability and potential to cause harm. They also rate 36 statements about their own predicted behavior, involving responses such as panic and paralysis. All ratings are on a scale from 0% (no belief at all) to 100% (complete belief). The total score ranges from 0 to 100, with higher scores indicating a greater fear of spiders and of one’s responses to spiders.

Fear of Spider Questionnaire (FSQ): The FSQ (Szymanski & O’Donohue, 1995) focuses on current thoughts and reactions to spiders. Questions include " currently, I sometimes think about getting bit by a spider" and "I would feel very nervous if I saw a spider now." Subjects rate 18 statements about spider fear on a scale from 0 (strongly disagree) to 6 (strongly agree). The total score ranges from 0 to 108, with higher scores indicating a greater fear of spiders.

Spider Phobia Questionnaire (SPQ): The SPQ (Watts & Sharrock, 1984) probes subjects' reactions to spider fear on a number of subscales: vigilance, preoccupation, avoidance-coping, and cognitive-behavioral. Questions include "do you make very sure there are no spiders around before taking a bath" (vigilance) and "are you sometimes distracted by thoughts about spiders" (preoccupation). Subjects answer 33 questions with “yes” or “no,” earning one point for each “yes” response. Scores range from 0 to 32, with higher scores indicating a greater fear of spiders.

Self-report Questionnaires, Treatment

Reaction to Treatment Questionnaire (RTQ): After learning the treatment rationale and GTR procedure, subjects will be asked four questions regarding their confidence in the proposed treatment. Ratings will be on a scale from 0 (no confidence) to 10 (extreme confidence).

Attentional Focus Questionnaire (AFQ) (Smits & Telch, 2000): Following the exposure, subjects will be asked to rate the degree to which they were focused on the presence of the spider, on their use of the shield, and on other distractors. Attention will be reported on a scale from 0 (complete ignorance) to 4 (complete focus).

Spider Threat Questionnaire (STQ): Composed predominantly of items from the FSQ, the STQ asks subjects to rate their fear of 12 possible outcomes of exposure, including both extrinsic (“the spider will bite me”), intrinsic physical (“I will lose control of my bladder”), and intrinsic emotional (“I will panic”) consequences.

Safety Behavior Attribution (SBAI and SBAS): At the end of the exposure session, subjects will be asked to report any idiosyncratic safety behaviors they used and to what extent they believe these behaviors prevented each of the outcomes in the STQ. Additionally, subjects in the two experimental conditions will be asked the same questions about their use of the shield. Ratings will be on a scale from 0 (no belief) to 6 (complete belief).

Subjective indices, BAT and treatment

Self-Efficacy: Subjects will be asked to report their confidence that they can carry out the exposure tasks immediately before entering the room with the spider (see procedure, below). Ratings will be on a scale from 0 (no confidence) to 10 (complete confidence).

Fear: Subjects will be asked to report the highest level of fear they expect to feel before beginning a task. Afterwards they will be asked for the highest level of fear they actually experienced, and the level of fear they experienced at the end of the trial. Fear will be reported on a scale from 0 (no fear) to 10 (most extreme fear possible).

Disgust: Subjects will be asked to report the highest level of disgust they expect to feel before beginning a task. Afterwards they will be asked for the highest level of disgust they actually experienced, and the level of disgust at the end of the trial. Disgust will be reported on a scale from 0 (no disgust) to 10 (most extreme disgust possible).

Other measures, BAT and treatment

Duration: The length of time the subject is able to remain in the room during the BAT or exposure trial.

Heart Rate Reactivity: While the subject is exposed to the spider, heart rate will be measured by an ambulatory heart rate monitor (Polar™ heart monitor, model M21, Polar Electro Oy, Finland). Heart rate reactivity is defined as average heart rate during exposure minus a baseline reading taken after the exposure and a period of rest.

Procedure:

Screening

All potential subjects completed the FSQ and answered the extra question "Could you walk into a room containing a tarantula in an aquarium, remove the lid from the aquarium, and touch the tarantula." Subjects with a 54 or higher on the FSQ and a 3 (severe anxiety) or 4 (could not do) on the tarantula question were invited for further screening.

In the second stage, subjects were told they would be exposed to a tarantula and shown a picture of the species. They then completed computerized versions of the self-report questionnaires (see measures, above) and took the first behavioral approach test (BAT - baseline; see below). Subjects who chose to terminate exposure on at least one BAT before two minutes elapsed or who rated their peak fear during the BAT as 7 or higher were invited to participate in the treatment phase.

Behavioral Approach Test - baseline

The behavioral approach test exposed subjects to a live Chilean Rose tarantula (species: Grammostola Rosea; body length approximately 4 cm; body width approximately 2.5 cm). The spider was placed in the center of the exposure room (approx. 2m X 3m). Subjects were instructed to enter the room and remain there as long as they could. They were told that the test ended when they asked to leave, or when two minutes had passed. The experimenter observed from outside, through a glass panel in the door.

Before entering the exposure room, subjects reported their Self-Efficacy, Expected Peak Fear, and Expected Peak Disgust (see measures, above). Afterwards, they reported both Peak and End ratings for Fear and Disgust. Heart rate was recorded at 60s intervals and at the beginning and end.

After the first trial, a second test was administered using a Mexican Redknee spider (species: Brachypelma Smithi; body length approximately 2.5 cm; body width approximately 1.5 cm). Subjects were asked to stand at the center of the room, and told that the spider would enter off a ramp through a hole in the door. Their task was to wait until the spider left the ramp, and then maintain a constant distance from the spider for as long as possible. The test ended when subjects asked to leave, or when two minutes had elapsed since the spider reached the floor. Subjects were also instructed that the test would end if they changed the distance between themselves and the spider, although the experimenter actually issued two warnings first. The hole in the door was at ~0.3m in height, and the ramp ended ~0.7m in front of the subject. The spider was placed on the ramp by the experimenter, out of sight of the subject, and squirted with water from a spray bottle to encourage it to walk down.

In both BATs, the experimenter entered at the end of the test and trapped the spider in a box while the subject left the room. This and other steps were taken to ensure that the tests and treatment did not involve any inadvertent modeling therapy.

After the second trial, the subject was told that they would not see any more tarantulas that day, and instructed to sit and relax for one minute. They then stood for thirty seconds, and their baseline heart rate was recorded.

Subjects admitted to the next phase were randomized into one of three treatment groups: escape behavior utilization, coping behavior utilization, and no-safety-behavior control. They returned the next day for treatment.

Exposure treatment took place the day following the baseline BAT. Subjects were given a brief educational handout on spider and tarantula anatomy and behavior. They then read an explanation of exposure treatment based on the disconfirmation model. The rationale specifically discussed the dangers of distraction and misattribution, and instructed subjects to refrain from distracting themselves or doing anything to make themselves feel safe. Subjects in the coping and escape groups were given a description of the shield they would be using, but were assured that they would be completely safe even without it. No mention was made in any condition of safety behaviors or their proposed effects. After reading the rationale, subjects completed the Reaction to Treatment Questionnaire.

Exposure - no-safety-behavior condition

Measures were the same as in the BAT (Self-Efficacy, Peak and End Fear, Peak and End Disgust, and Heart Rate Reactivity). The tarantula and the procedure used were the same as in the second BAT trial, except that the test ended three minutes after the spider reached the end of the ramp. Subjects received a total of fifteen minutes of exposure, over a minimum of five trials. If the subject terminated a trial early, they completed extra trials until they amassed fifteen minutes of exposure. At the end, their baseline heart rate was recorded as in the BAT. They then filled out the AFQ, the STQ, the SBAS (coping and escape conditions only), and the SBAI.

Exposure - coping behavior condition

The coping behavior condition was identical to the no-safety-behavior control, except that subjects were given a wooden "shield" and instructed to use it to stop the spider from touching them. The shield was in an inverted T-shape, ~30cm X 130cm at the top and 100cm X 30cm at the bottom.

Exposure - escape behavior condition

The escape behavior condition was identical to the no-safety-behavior control, except that subjects were given a clear plastic cube (~12cm) with a long wooden handle (~1.7m). They were instructed to wait until the spider left the ramp, and then use the box to trap the spider. They maintained the same distance from the spider as subjects in other conditions. When the trial ended, the experimenter took the handle and kept the spider trapped until the subject left the room.

BAT - post-treatment and follow-up

The BAT - post-treatment was administered on the day following exposure, and the BAT - follow-up at two weeks after exposure. Both were identical to the BAT - baseline. Due to the length of time between the post-treatment and follow-up BATs, subjects were given contact information for the study supervisor at post-treatment and instructed to call if they experienced any continuing discomfort. However, they were not fully debriefed until after the final BAT.

Results

Manipulation Checks

The three conditions in the current study are designed to differ in the access they provide to disconfirmatory information. There is the danger that they also differ in other factors that affect fear reduction. In order to check for confounds, test results were analyzed to check for between-group differences in treatment plausibility, attention, sense of threat, and idiosyncratic safety behaviors. In addition, the two safety behavior groups were compared on attribution of safety to the shield.

Treatment plausibility was assessed with the RTQ, administered just after subjects read the rationale. Analysis of variance finds no significant difference between groups in acceptance of the treatment (F(2,5)=0.84, p=0.48).

Attention was measured using the AFQ. The AFQ was split into two subscales, one dealing with attention to the spider (AFQ-s) and one with attention to irrelevant distractors (AFQ-d). ANOVA shows no difference in attention to distractors (F(2,4)=1.24, p=.38). However, there was a significant difference in attention to the spider between the escape group (M =4.50, SD=0.29) and the control group (M=3.67, SD=0.71) (F(2,4)=10.00, p=0.03).

The SBAI was used to assess attribution of safety to idiosyncratic safety behaviors. Most subjects engaged in some mental behavior, such as distraction or reassurance. None reported any tangible aids or behaviors that they might have credited with their physical safety. ANOVA finds no significant between-groups differences (F(2,5)=0.79, p=0.50). Mean score across groups was 1.77 out of 7 (SD=2.09). Because several questions on the SBAI received scores of 0 or 1 from most subjects, a second ANOVA was done on the subset of questions that received at least two answers greater than 1. This raises the mean score to 2.20 (SD=2.25) but does not reveal any between-group differences.

The two experimental groups were compared on attribution of safety. ANOVA on SBAS scores demonstrates no differences in attribution of safety (F(1,3)=3.68, p=0.60). Mean score across groups was 2.9 out of 7 (SD=1.16). The low-scored questions on the SBAI received similar responses on the SBAS. Again, this subscale showed no difference between groups (F(1,3)=0.01, p=0.95), although it raised the mean score to 3.64 (SD=0.80). Further references to the SBAI and the SBAS will use this subscale.

ANOVA on the STQ (F(2,5)=0.28, p=0.69) reveals that subjects in all three conditions felt equally threatened by the spider (M=2.75, SD=0.55). Extracting the same questions as on the SBAI subscale raises the mean to 3.8 (SD=0.87), but does not change the between-group comparison. Because the response patterns were not as clear as on the safety behavior attribution tests, the entire STQ will be used in further analysis.

Outcome measures:

The basic treatment design showed effectiveness across all groups. Average fear was 6.64 out of 9 (SD=1.38) during the first BAT session and 2.17 (SD=2.30) at follow-up. This is a significant difference (t(5)=3.80, p=0.01). Average disgust also reduced, but was not significant.

Differential treatment effectiveness was analyzed using change in fear and disgust between the beginning and end of treatment and between the pre-test and follow-up BATs. Questionnaire responses were also used. Nearly all subjects were able to stay for 2 minutes during the initial BATs, and all stayed for 2 minutes during every post-treatment BAT. Thus, duration data are not analyzed. End fear and disgust and heart rate data were not collected for most of the pilot and could not be used, but will be included in the analysis of the full experiment.

ANOVA on the fear and disgust measures found significant differences only for treatment disgust (see table 2). Average change between the first and last trials was –0.33 (SD=0.58) for control subjects, and –3.30 (SD=0.99) for coping behavior subjects (Fisher’s PLSD=2.97, p=0.02). A similar but non-significant trend is present for BAT disgust. There is a significant correlation (r(5)=0.76, p=0.05) between treatment disgust and BAT disgust, and a nearly significant correlation (r(5)=0.72, p=0.07) between treatment fear and BAT fear.

ANOVA on SBQ scores showed no significant difference between groups (F=2.45, p=0.20).

None of the tests show significant correlations with one another or with subjective outcome measures (see table 3). The sole exception is that the SPQ correlates very strongly with treatment fear (r(5)=0.95, p ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download