Some comments on therapy analogue research with small animal “phobias”

1. Bchav.27m. It &q~. Psych&u.Vol. 2, up. 225-231.Pergmon Rcas, 1971.F’rintedin Great Britain.

SOME COMMENTS

ON THERAPY

WITH SMALL ANIMAL DOUGLAS A. BERNsTElNt Psychological

ANALOGUE

RESEARCH

“PHOBIAS”*

and GORDON L. PAUL

Clinic, University of Illinois at Urbana-Champaign

Summary-Although laboratory analogues may play an important role in the development of behavior modification techniques, their utility is limited by the degree to which the variables studied in the laboratory share certain essential characteristics of the variables in the clinical context. The recent popularity of small animal “phobia” analogue studies has been accompanied by a geometric increase in analogue research errors involving: subjects and selection procedures; target behaviors and assessment procedures; treatment techniques and parametric variattons; and therapist characteristics and treatment environments. With a few notable exceptions, the majority of these “snake, rat and spider” studies involve sufficient analogue research errors to limit severely the cogency of their findings to the circumstances in which data were collected, without making significant contributions to clinical practice, or to knowledge of the mechanisms involved in the operation or treatment of anxiety. Common analogue research errors are described, along with suggestions for identifying and overcoming the major errors of this type. BECAUSE of its potential precision in the isolation, manipulation, and control of relevant variables, the standard of scientific research has always been the laboratory-based study. Because of ethical considerations, and the number and complexity of variables involved in behavior modification, more clarity may be achieved at various stages in the development of knowledge relevant to clinical problems and procedures by laboratory experimentation on variables which are less complex, but analogous to those involved in the clinical context than by less well controlled studies in the “real life” clinical setting. Experimental analogues may be especially useful in narrowing parameters of influence, determining probable mechanisms of change, developing techniques and explanatory principles, and determining parameters from which new therapeutic techniques may be derived. Once a treatment “package”, such as systematic desensitization, has been found to be effective (Paul, 19696, c), the more stringent and

economical conditions of the experimental analogue may be particularly useful in determining probable mechanisms of change, different parameters of influence within classes of variables, and in refinement of specific aspects of treatment technology (see Paul (1969a) for a general discussion of the place of laboratory analogues in behavior modification research). Theoretically, treatment techniques such as systematic desensitization, implosion, aversive conditioning, and live or symbolic modeling should, in fact, lend themselves especially well to laboratory analogy. The major principles upon which they are based come directly from the laboratory, thus making it logical to employ laboratory-based procedures to evaluate and analyze them. The basic operations involved in behavior modification procedures derived from principles of learning are relatively clear and specifiable. Therefore procedures employed across experiments, clinics, and/or laboratories should be easily compared. Finally, because

*Preparation of this manuscript was supported in part by Grant MH-17376 from the National Institute of Mental Health, United States Public Health Service. TRequests for reprints should be addressed to Douglas A. Bernstein, Psychological Clinic, University of Illinois, 51 E. Gerty Drive, Champaign, Illinois 61820. 225

226

DOUGLAS

A. BERNSTEIN

treatment targets should be specified in advance in behavioral terms, outcome evaluation and assessment of the relationship between manipulation of parameters of treatment techniques and change in subjects’ distressing behavior should be accomplished with little ambiguity. Analogue research activity-the result of such available specified treatment techniques, and the models for researching them-is a positive development because new techniques need to be systematically evaluated, both before and after clinical application. Analogue reserach is one of the important steps in the complicated process of doing so. It should be recognized, however, that not all laboratory analogues have the same relevance. Therapy analogue studies become relevant for clinical practice only when generalization from the laboratory to the clinical setting is possible. As Paul (1969a) has previously stated, “ . . . the generality of particular conclusions from experimental analogues to the real-life behavior modification context must be a function of the degree to which the variables studied in the laboratory share the essential characteristics of the variables in the clinical context” (p. 54). Analogues which do not closely approximate the variables involved in the clinical situation are not totally irrelevant, however, since they are capable of (a) contributing to knowledge about behavior change in general (without reference to clinical behavior modification) and (b) providing a basis for the selection of variables and the formation of hypotheses which can then be employed and tested in more clinically relevant research (Levis, 1970; Paul, 1969u). Establishing an experimental analogue which has clinical relevance is not easy. In attempting it for desensitization, or any procedure designed to reduce anxiety, seemingly small methodological details or outright “corner-cutting” or oversimplification can seriously vitiate the strength of the analogy. Any attempt to create a strong analogy for treatment of anxiety-based problems must deal carefully with the major variables of importance in behavior modifica-

and GORDON

L. PAUL

tion research: subjects, their characteristics, and social environments; target behaviors and the nature of assessment; treatment techniques, procedures, and their parameters; and therapists, their characteristics and the physical-social environment in which they operate. A “perfect” laboratory analogue would employ as Ss only persons who seek help with clinically distressing problems showing clear anxiety responses, and as Es only practicing professionals who uniformly administer treatments identical to clinical procedures. Unfortunately, the problems involved in setting up a “perfect” analogy are equivalent to those faced in controlled research in a clinical setting so that the rationale for the analogue design, if not the “analogue” itself, would be eliminated. Thus, each investigator must construct his analogue in a less than “perfect” way, but in a way that will allow answers to the research questions in which he is most interested. Any analogue study which is internally valid may make a research contribution, but the nature of that contribution (e.g. its clinical relevance) will be a function of the way in which the analogue is set up. Researchers need to be aware of this point so that both prior expectations regarding the clinical applicability of data and the statements made about those data in introductory and discussion sections are consonant with the level of product possible from the design employed. At present, the relationship between design and possible level of product in therapy analogue research is apparently not well understood since conclusions regarding the clinical use of systematic desensitization and other techniques are being drawn on the basis of analogue designs whose variables and procedures are not generalizable to clinical settings. The publication of Lang and Lazovik’s (1963) well known analogue desensitization study of snake phobias marked an important and timely development in the history of knowledge in systematic desensitization (Paul, 1969c). It also provided a prototype for analogue studies on systematic desensitization, and other anxiety-

SOME COMMENTS

ON THERAPY

reduction procedures. This study appealed strongly to the interests of clinical researchers whose work is conducted in academic settings and, perhaps even more strongly, to degree candidates whose interests and goals are divided between making a contribution to the field and completing a thesis project in a year or less. The popularity of this analogue approach is readily apparent from the fact that over 50 analogue studies on small animal “phobias” appeared between January, 1968 and June, 1971 in only five professional journals. The vast majority of these papers were often interesting and imaginative, but unfortunately, relevant only to artificial laboratory settings largely unrelated to anxiety problems. Findings from most such studies may serve, mainly, to refine laboratory procedures and suggest the kinds of variables and hypotheses which might be manipulated and tested in later, more clinically meaningful designs. Unfortunately, the majority of the investigators involved pay lip-service to the fact that subject responses or treatment procedures differed from those observed in clinical practice, and yet draw conclusions and make recommendations as if identities existed. Knowledge concerning particularly systematic desensitization has progressed sufficiently to make analytic laboratory analogue research especially relevant (Lang, 1969; Paul, 1969c). But with the exception of the work of Lang (e.g. Lang and Lazovik, 1963; Lang, Lazovik and Reynolds, 1965), Davison (e.g. Davison, 1968), and Bandura (e.g. Bandura, Grusec and Menlove, 1967; Bandura and Menlove, 1968; Bandura, Blanchard and Ritter, 1969) few analogue studies exist on small animal “phobias” which can be generalized or be relevant to clinical situations beyond the provision of suggestive hypotheses. In fact, the following statements, abstracted from editorial comments on innumerable rejected manuscripts, could as well apply to the majority of the published papers on small animal “phobias”: “Of particular concern is the enduring significance and importance of such studies. Specifically, it

ANALOGUE

RESEARCH

227

is unlikely that selection procedures obtained subjects with a sufficient degree of stress response in the presence of the eliciting stimulus to allow investigation of ‘anxiety’ related problems, or of therapeutic techniques designed to reduce anxiety responses. Equation of a low-demand avoidance test performance with ‘fear’ or ‘anxiety’ in the absence of considerably greater additional validating assessment operations is empirically and theoretically unjustified. In the absence of strong anxiety responses, as typified in such studies, subtle differences in manner, procedure, presentation, etc., are likely to effect performance on a low demand avoidance test by shifting the ‘demand characteristics’ experienced by the subjects. Most of these studies involve procedures sufficiently removed from those used clinically, and in controlled investigations with subjects showing demonstrable physiological arousal, to be ‘analogues of analogues.’ None of the studies typified above really appear to make a significant contribution to clinical practice, or to our knowledge of the mechanisms involved in the operation or treatment of anxiety responses.” COMMON ANALOGUE RESEARCH ERRORS Although the above quotation is descriptive of most published studies, investigators tend to build upon published works. We shall, therefore, attempt to elaborate some of the more common pitfalls in these analogue studies in the hope of helping investigators become aware of them, and thereby avoid receipt of such editorial comments on future manuscripts. When specific studies are cited below as examples of what are considered to be analogue research errors, it should be understood that each reference is meant only to illustrate a problem, not single out a particular individual or group. The studies cited were selected only because they represent the research tactics employed by many other investigators. We shall not discuss genera1 aspects of design relating to internal validity and adequate reporting. Many analogue studies do fail to provide adequate

228

DOUGLAS A. BERNSTEIN

pretreatment equation of groups: statistical analyses; reports of procedures; reports of prior findings; reports of the reliability of instruments and data reduction; or control groups run concurrently with treatment groups. However, these problems are not restricted to analogue studies, and have been discussed elsewhere (Paul, 1969a). Cooper, Furst and Bridger (1969) have recently argued that research aimed at studying the effectiveness of treatments for eliminating fear of small animals, such as snakes and rats, may be “ . . . irrelevant to a knowledge of the treatment of clinical phobias (p. 414)” We do not believe their argument to be a valid indictment of analogue research because fear of snakes and rats can be “clinical phobias” and because development of procedures to eliminate such phobic responses may be just as relevant as research on anxiety and avoidance behavior in response to other classes of eliciting stimuli. There is, in fact, no sound empirical evidence to indicate that the nature of anxiety responses or secondary escape and avoidance behaviors differ from one class of eliciting stimuli to another. Rather, both theoretical and empirical support exists for considering the nature and operation of anxiety reactions, anxiety levels, and secondary escape and avoidance behaviors to be similar, not only among classifications of people (e.g. “normal”, “neurotic”, “psychotic”), but among species as well (Paul and Bernstein, in press; Ullman and Krasner, 1969). However, while it makes sense to conduct research on anxiety elicited by snakes, rats or spiders, it is incumbent upon E to employ as Ss only persons who can be shown to display significant and therefore clinically relevant increases in physiological arousal and cognitive distress (i.e. anxiety) as a result of the presence of the presumed eliciting stimulus object. Even if direct clinical

application is not of major concern to the investigator, he still must demonstrate that his Ss exhibit a measurable anxiety response which is not a function of situational artifacts if findings are to have any relevance to clinical problems or treatment. Analogue research

and GORDON L. PAUL designs typically have not met this first minimum requirement. It is perfectly reasonable to suggest that the S population in such studies is comprised mainly of persons whose “fear” may be the result of social and other situational factors and is of an intensity which does not approximate clinical proportions. In such cases, Cooper et al.‘s conclusions would be valid, not because of inherent differences in the nature of stimulusresponse relationships involving small animals as a stimulus class or “normals” as a subject population, but because of an analogue error in failing to obtain Ss who demonstrate responses with the essential characteristics of the responses seen in the clinical context. Subjects in the vast majority of small animal “phobia” analogue studies are college students who participate as the result of an invitation from the experimenter. Sometimes the invitation is extended selectively (e.g. only to Ss who have indicated fear of the target object on a questionnaire) but, in other cases, a blanket invitation is offered to anyone who considers himself to be fearful of the available stimulus object. Payment for participation, either in terms of money or credit toward fulfilling a course requirement is often included as a reward for participation. Considering the social nature of decisions regarding labels of abnormality and actions leading to application for clinical services (Ullman and Krasner, 1969, pp. 20-24), there is no logical reason why people volunteering or requesting treatment for anxiety-related problems in response to an announcement of availability of services established for research should differ from those who apply for existing servicesunless the nature of the announcement, invitation, and/or other situational contingencies draws people to volunteer because of some other than obtaining anticipated “payoff” relief from a problem in which inappropriate anxiety is central. The potential analogue research error is, again, the likelihood that the anxiety response found in the latter volunteers will not be comparable to the anxiety response found in populations to which the investigator wishes to generalize results.

SOME

COMMENTS

ON THERAPY

Two things should be made clear regarding typical subject recruitment in analogue studies of small animal “phobias”. First, the probability of finding persons who dispIay a clinically relevant degree of anxiety in the presence of a snake or rat is relatively low since the number of individuals in a college population who experience sufficient anxiety in response to a particular “small animal” stimulus to increase arousal and disrupt performance is likely to be very small indeed. In fact, Michael Evans, under the supervision of the senior author, recently found only 42 Ss who met such a criterion from “snake phobic” response volunteers in the entire city of Chicago. Lang (1968) estimated the figure for college populations to be at only 2-3 per cent of enrollment. Second, the typical S recruitment procedure usually involves presentation of the project as experimental research as well as or rather than clinical treatment, in which another “payoff” in the form of money or meeting course requirements for research participation is offered. Thus, it is not unreasonable to assume that many Ss volunteer more out of curiosity, a desire to contribute to science, or to obtain the other payoff than because of significant discomfort over snakes, rats, or whatever. Further, the experimental nature of the project and the many rigorous procedures involved make it likely that demand strong experimental characteristics (Orne, 1962) are brought to bear on Ss before, during, and after treatment. Thus, the volunteer S obtained in most small animal “phobia” analogue studies may very well be a person who experiences little or no anxiety in relation to the target stimulus object and who, as a college student participating in an experiment conducted by a professor, or other authority figure is likely to be sensitive to and strongly influenced by social cues regarding appropriate behavior as they are communicated to him throughout the study. In light of these considerations, it is not terribly surprising (though very disturbing) that, in one study (Robinson and Suinn, 1969), 70 per cent (40 out of 57) of female college

ANALOGUE

RESEARCH

229

student volunteers were classified as “spiderphobic” on the basis of their response to a fear rating form. Of the 22 “phobic” Ss selected for therapy (on the basis of scheduling convenience), none were eliminated by later screening measures. Thus about 38 per cent of the original tested sample was employed as a “phobic” population. It is obvious that, in order to overcome the negative effects of typical recruitment procedures, potential Ss must be screened further to eliminate those persons who do not respond with strong anxiety to the “eliciting” target object, or who do so mainly as a function of variables such as payment, social pressure, or other situational factors. Target behavior assessment and/or sequential screening in most small animal “phobia” analogue research has relied, mainly, on some form of a Behavioral Avoidance Test (BAT), which is based upon the procedures pioneered by Lang and Lazovik (1963). The basic characteristic of these BATS consists of asking Ss to approach, touch, and handle the feared target object. Those Ss who make physical contact with the object are usually dropped from the study as not being sufficiently “phobic”, and, for those remaining, the degree of physical approach is taken as an index of anxiety. The BAT is almost always repeated after treatment for all Ss as a measure of improvement. The crucial question regarding BATS, as they have been employed, is whether or not they actually serve their purpose, i.e. to eliminate inappropriate Ss from the originally recruited population, and to serve as an index of anxiety responses for retained Ss. Their use as an unsupplemented diagnostic device can be justified only if it can be shown that they meet a minimum requirement by serving that function. The major problem with BATS is that they are an indirect measure of anxiety, since escape or avoidance behaviors are secondary operant responses. For Ss whose anxiety response to the target stimulus is extreme, little difficulty would be found, since all three direct channels of measurement of anxiety (verbal, physiological, observable motoric) and indirect, observable

230

DOUGLAS A. BERNSTEIN

avoidance behavior would likely coincide. However, with even moderate degrees of anxiety, operant escape and avoidance responses are more likely to be a function of other discriminative and reinforcing stimuli in the assessment context than of the degree of anxiety elicited by the target object functioning as a conditioned stimulus (see Paul and Bernstein, in press, for a discussion of classes of assessment instruments and their interrelationships at different anxiety levels.) Failure to touch a target object (such as a snake) on a therapy pretest could be due to a variety of factors other than S’s spontaneous, conditioned, and broadly generalized tendency to respond with strong anxiety to that object. For example, Ss are likely to be unfamiliar with organisms such as snakes, rats, and spiders, and may simply not know how to go about handling them. In the senior author’s laboratory, it was discovered that Ss given instructions for handling snakes do so significantly more quickly (and look less fearful) than uninstructed Ss. Therefore, some Ss are likely to be classified as “phobic” because of uncertainty regarding handling procedures, especially if “phobia” is defined as not having touched the target within a given time. Another potential influence upon S behavior during BATS is the content and manner of presentation of test instructions. It is possible, for example, that Ss told to approach, touch, and handle the target (e.g. Cotler, 1970) may be more likely to do so than those given similar instructions, but asked only to do what they want to do or only what is comfortable for them (e.g. Lick and Bootzin, 1970; Hodgson and Rachman, 1970). Moreover, there may be an effect of whether instructions are given as a “one-shot” presentation (e.g. McGlynn and Wiliams, 1970) or progressively administered while S attempts each aspect of the BAT task (e.g. Rimm and Mahoney, 1969). While specific factors such as instructions and target familiarity probably play their role in determination of BAT performance, they may be less potent than other, more general social factors best summarized as situational context

and GORDON L. PAUL

variables. These include mainly the rationale for Ss’ presence at the BAT and the demand characteristics operating within it. Most Ss have participated in a BAT in an “honest context” (Bernstein, in press), i.e. they know that they are there because of their verbally reported fear, that their fear is being measured by the test and, possibly, that failure to demonstrate some degree of fear will eliminate them from the project. Given this context, in which S has already reported himself to be phobic, other for positive consequences are anticipated participation, and in which both S and E have invested some time and effort, it seems clear that a display of avoidance behavior, (or at least, reluctant approach) is the most socially appropriate and likely to be reinforced behavior in the situation. In some studies (e.g. Hodgson and Rachman, 1970) Ss are told prior to the pretest that they will have the opportunity to try the BAT again after treatment, thus making it clear that some improvement is expected and making pretest avoidance even more obviously expected. BATS conducted in such contexts, unsupplemented by other appropriate measures, are likely to result in the selection of a large proportion of Ss whose behavior during the avoidance test is more a function of social pressure than of the target object. If this occurs, the S population will be inappropriate for either process or outcome research on behavior modification techniques relating to anxiety reduction, and a lethal analogue research error will have occurred. When one reads that, on the basis of a fear survey and unsupplemented BAT, 18 to 22 per cent of the originally contacted population were found to be “phobic” (e.g. Krapfi and Nawas, 1969; Lick and Bootzin, 1970; Miller and Nawas, 1970; Nawas, Welsch and Fishman, 1970; Nawas, Fishman and Pucel, 1970) or that, once Ss reach the BAT stage in sequential screening, 50 to 100 per cent of them qualify as “phobic”, one is forced to ask to what extent these results reflect validity of the earlier verbal report or social pressure in the BAT situation. Demand and other social factors are, of

SOME

COMMENTS

ON THERAPY

course, also operative at the posttreatment B.4T and one must ask about their effects at that point as well. It seems likely that after treatment, (especially if it has been unusual, effortful, or time consuming) there would be strong pressure exerted on Ss to touch and/or handle the previously avoided object and, if the original recruitment and screening of Ss resulted in selection of a large proportion of persons who are strongly affected by such pressure, they could be expected to respond to it again and become “cured”. Investigators frequently add the S’s assessment of his own anxiety during BAT performance in an attempt to obtain a measure of anxiety through another channel. An example is the “fear thermometer” (e.g. Lang and Lazovik, 1963). While multiple measurement is, in fact, necessary, a single fear thermometer at the point of termination of approach in the BAT is subject to all of the invalidating influences discussed for the BAT itself when severe anxiety responses are not elicited. Furthermore, self-report measures such as the fear thermometer are typically obtained during posttest BATS at the point of posttest termination of approach. If the self-report measure and the BAT measure were both valid, there should be no change on the self-report measure from pretreatment to posttreatment, since Ss should on each occasion approach the eliciting stimulus to the point where the anxiety response became sufficiently uncomfortable to terminate the approach. Thus, a reduction in anxiety should lead to an increase in approach to the point where the anxiety level equalled that of the pretreatment BAT termination. Appropriate direct assessment of anxiety within BATS requires direct assessment through all three channels at both pretests and posttests at the point of pretest BAT termination (see Paul and Bernstein, in press). If such multiple assessments are carried out, investigators are likely to find high-demand-for-approach conditions to be necessary in the BAT to demonstrate substantial increases in physiological arousal and cognitive distress in the typical S populations employed.

ANALOGUE

RESEARCH

231

The major consequence of what we are saying here is that the recruitment and “diagnostic” practices employed in most therapy analogue research involving small animal “phobias” may, to a large extent, have created “target problems” which are “cured” by procedures that are effective or ineffective only by virtue of their manipulation of expectancy variables and social pressure (see Bernstein, in press, for a fuller discussion of the issues and some preliminary data bearing on them). Treatment techniques, or specific components of treatment packages are typically the independent variables of interest in analogue research. In order to make meaningful statements regarding the value of a particular technique and its parameters, an accurate replica of the techniques as used clinically must be employed in the laboratory with at least one group of Ss. This seems to be an obvious minimum requirement but, in many analogue studies, important aspects of treatments such as systematic desensitization or implosion have been altered in the laboratory in order to meet the dictates of convenience, ease of administration or other factors (e.g. so that a degree candidate can compIete a thesis before the end of the academic year). Even though the changes vitiate the analogy to the clinical treatment, the laboratory version is referred to in published reports as “systematic desensitization”, “implosion” or whatever. Some procedural alterations make both practical and theoretical sense, such as when a group rather than individual hierarchy (geared to the most anxious S) is used to reduce inter-S variance in terms of the nature of items presented. All Ss in such a group can legitimately be said to have received systematic desensitization. However, some variants violate the major principles upon which the original treatment rests without providing data from appropriate comparison groups within the study, or from independent controlled research to show that the changes made do not seriously alter its effectiveness. A common feature in analogue research on

232

DOUGLAS A. BERNSTEIN

systematic desensitization, for example, is the use of tape-recorded procedures which do not allow progress to be contingent upon S response. Recorded progressive relaxation instructions, for example, have frequently been used without first showing that they are capable of inhibiting anxiety, and without appropriate comparison groups trained by personal instruction, (e.g. Nawas, Fishman and Pucel, 1970; Nawas, Welsch and Fishman, 1970; Miller and Nawas, 1970), even though relaxation training was not designed to be administered in this way (e.g. Bernstein and Borkovec, in press; Paul, 1966, 19696, c). (In fact, Paul and Trimble, 1970, found recorded progressive relaxation training to be ineffective, in contrast to recorded direct suggestion of relaxation which was effective over time). If, because of such treatment alterations, Ss are not experiencing significantly reduced autonomic arousal, it becomes theoretically and logically impossible to claim that the data generated by the research are relevant for understanding either the process or outcome of systematic desensitization. The results in such cases are generalizable only to other, similarly conducted treatments, none of which should be labeled “systematic desensitization.” Other procedural variations, such as presentation of hierarchy items for standard time intervals (e.g. Boulougouris, Marks and Marset, 1971; Crowder and Thornton, 1970; Nawas, Welsch and Fishman, 1970; McGlynn and Williams, 1970), taped as opposed to “live” therapy (e.g. McGlynn and Williams, 1970; Puce1 and Nawas, 1970), visual rather than or in addition to imaginal item presentation (e.g. McGlynn, Wilson and Linder, 1970; Miller and Nawas, 1970), extremely abbreviated relaxation training (e.g. DeMoor, 1970; Puce1 and Nawas, 1970), and single rather than multiple presentation of each hierarchy item (e.g. Willis and Edwards, 1969) have also been reported in the small animal “phobia” analogue literature on systematic desensitization and are subject to the same criticisms leveled above. Although many of the above variations violate

and GORDON L. PAUL

basic procedural principles for systematic desensitization (see Paul, 19696), such variations could provide useful data if compared with standard procedures which have been found effective. In fact, given the current level of development of systematic desensitization, parametric investigations on duration and number of items, duration and number of exposures and other temporal variables, order and manner of item presentation, and investigations of the theoretical basis for effects are precisely what is called for, and where analogue studies could contribute most (Lang, 1969; Paul, 1969c). However, without comparison groups with standard treatment and psuedotherapy or attention-placebo, such variations will not be comparable to the techniques to which the investigator wishes to generalize. Unfortunately, much of the parametric analogue research designed to investigate these kinds of variations also suffers from analogue research errors described earlier regarding S selection and assessment of target behaviors, as well as in treatment techniques. For example, Nawas and his colleagues have reported that the effectiveness of “systematic desensitization” is unaffected by (a) whether or not Ss are allowed to control item presentation (Nawas, Fishman and Pucel, 1970; Miller and Nawas, 1970), (b) whether desensitization is “live” or tape-recorded (Krapfl and Nawas, 1969) or (c) whether hierarchy items are presented in order of ascending difficulty or in reverse order (Krapfl and Nawas, 1970). Unfortunately, however, the procedures employed in these, and many other parametric studies provide little or no reason to believe either that the Ss employed were appropriate for the research, or that the training” and “desensitization “relaxation proper” administered were related to the clinical treatments to which the data are meant to apply. It is of utmost importance that parametric investigations involve Ss with demonstrably strong anxiety responses, since the basis for the majority of procedural sequences and timing in systematic desensitization is to allow a gradual approach to those stimuli which

SOME COMMENTS

ON THERAPY

elicit extreme anxiety. To the extent that strong anxiety responses are not elicited, the need for lower hierarchy items at all, let alone their order, duration, manner of presentation, and so forth, becomes theoretically irrelevant. The inclusion of the usual BAT assessments, with implausable “attention-placebo” conditions in such studies is IikeIy to show both within and between groups changes in the BAT that are primarily a function of the impressiveness of the procedures and the degree to which they are obviously related to the target stimulus (see Bernstein, in press). Thus, such parametric analogue studies do not provide sufficient grounds for transferring the treatment variations involved to other research or clinical settings. Therapists/experimenters in analogue studies and the physical-social environment in which they operate may contribute a great deal to the impressiveness of treatments and to expectancy effects. Therapists in laboratory analogies should be persons who approximate “real” therapists, at least in terms of possessing nonspecific clinical skills and a thorough grounding in the specific treatment techniques they administer. An experimenter who is either inexperienced or untrained in the use of the treatments involved in an analogue study or who is generally deficient in clinical experience may bias treatment presentation and S’s response to it in such a way that an unambiguous test of the study’s hypotheses cannot be accomplished. Thus, unless one is interested in doing research on the effects of therapy when conducted by differentially experienced or uniformly inexperienced therapists (which is not usually stated as the purpose of therapy analogues) the experimenters should contribute as little variance as possible to the process or outcome of the treatments administered. Unfortunately, most small animal “phobia” analogue research has been conducted by persons who were likely to bias results through direct procedural errors born of inexperience and absence of in-session monitoring, lack of sophistication in handling small but significant in-session problems, or inability to provide a

ANALOGUE

RESEARCH

233

consistent and professional “delivery system” for treatment which keeps all Ss in treatment and positively disposed toward it. In other words, therapists in many analogue experiments may not be sufficiently well trained to know what they are doing (as clinicians or as experimenters) and this may be communicated to Ss with confounding consequences for treatment outcomes. Testing this assertion is a bit difficult because so many analogue researchers do not present information about Es in their reports at all. An examination of 33 recently published analogue studies revealed that only 9 contained any description of therapists whatsoever. In those studies describing therapists, it appeared that Es were often far from ideal therapists, and may have been chosen more on the basis of availability than anything else. For example, Nawas, Fishman and Puce1 (1970) reported that their treatments were conducted by “ . . . relatively naive graduate students who were given some training in SD technique (p. 52).” Miller and Nawas (1970) used as Es “ . . . clinical psychology Ph.D. candidates who had a course in behavior modification . . . (p. 59),” and the Es in the McGiynn, Reynolds and Linder (1971) study were persons who had no prior experience with desensitization nor supervision during its conduct. While the use of Es of this type does not necessarily invalidate an analogue experiment, at least in terms of interval validity, it may have serious negative consequences for the study’s external validity (generalizability) which, in analogue research, is the name of the game. Similar failures to report the physicalsocial context of treatment prevent clear interpretation of many analogue studies. These features are important since BAT performance has been found to differ significantly when conducted in a “laboratory” vs. a “clinic” (Bernstein, in press), and suggestive evidence for differential S cooperation with procedures has appeared for “experimental rooms” vs. “clinics”, as well as experienced vs. inexperienced therapists (Paul, 1969~). The overall result of the common analogue research errors described above is that the data

DOUGLAS

234

A. BERNSTEIN

from most laboratory analogue research studies on small animal “phobics” is difficult or impossible to interpret. The findings reported could be due to the effects of inappropriately recruited Ss, reactive and demand-laden assessment techniques, unrepresentative or invalid treatment techniques, and less than competent Es, just as well as to the effects of the independent variables which the experimenter has sought to manipulate. Obviously, this state of affairs retards the accumulation of systematic knowledge about the effectiveness of behavior modification techniques and their modes of operation. In addition, it tends to encourage the proliferation of similar research. Finally, the publication of data which misleadingly validates unusual variations of treatment techniques, can easily lead to inappropriate and unsuccessful treatment by clinicians, generating negative evaluations of the clinical use of behavioral approaches in general.

SUMMARY FOR FUTURE

RECOMMENDATIONS ANALOGUE

RESEARCH

In order to reestablish the reciprocal relationship between laboratory and clinic which is ultimately essential to progress in clinical behavior modification, laboratory analogue research methodology must be improved. The most obvious way to do this is to reduce to a minimum the amount of “noise” in each of the areas of the research “system” discussed earlier. No “checklist” can be provided for therapy research which will insure unambiguous findings, since each study must be designed to answer the specific questions asked (see Paul, 1969a). However, some suggestions may help investigators to avoid analogue research errors.* In terms of S recruitment, it is unrealistic to expect to find, on a “one-shot” basis, 100 or 150 students on a college campus displaying a strong anxiety reaction to a particular small animal. *Bernstein

and GORDON

L. PAUL

It seems that analogue researchers have sometimes been guilty of “case-making” (Ullmann and Krasner, 1969). The quality of research in the area will improve if fewer attempts are made to run large N factorial studies in which all Ss are run during the same period of time. Progressive experiments where cells are filled as appropriate Ss become available, running Ss from all treatment and control groups in sequential waves, would seem to provide an attractive alternative (see Campbell and Stanley, 1963). The potential population of Ss with sufficient anxiety response for meaningful participation might also be expanded by inclusion of more than one eliciting stimulus object; provided that comparability of response were documented in independent research, or controlled by factorial representation in the design. Larger populations of essentially “normal” individuals who respond with clinical levels of anxiety to specific stimulus conditions may, in fact, be found by abandoning the “small animal phobia,” and analyzing potentially stressful situations which people cannot avoid without suffering some loss. For example, institutional requirements place relatively large numbers of individuals in potentially anxiety eliciting situati0ns-e.g. tests, oral exams, speeches, recitals, plays, physical and dental examinations, dating, airplane rides, observation of surgery or autopsies-in which our culture dictates that more than “2 to 3 per cent” will respond with clinical levels of anxiety (Goldstein, Heller and Sechrist, 1966; Paul, 1966). These larger populations would enable the conduct of factorial studies with Ss run concurrently rather than sequentially. Whatever stimulus class might be focussed upon, prospective analogue researchers should also consider the possibility of presenting their projects as “treatment services being made available”, in which evaluation is included, rather than experiments. This would reduce the study’s appeal to simply curious Ss, and would allow for and justify some of the more careful

(in press) discusses some of the issues covered here in greater detail.

SOME COMMENTS

ON THERAPY

assessment procedures discussed below. Of course, such a strategy would require subsequent treatment of Ss in various control groups, but this should probably be done in any context and would provide a source of additional data. Potential “payoffs” other than relief of experienced anxiety should also be avoided. It would appear desirable explicitly not give to either course credit or payment for participation. In fact, more relevant motivation might be insured by charging Ss a nominal amount, which could be returned on completion of final assessments. Once an initial pool of potential Ss has been identified, the investigator is faced with a clear decision. He can either seek a close analogy to a clinical population by rigidly screening out all Ss who fail to display extreme anxiety (and thus find relatively few appropriate Ss in each group tested if using small animal “phobics”), or he can relax his diagnostic restrictions (thus obtaining more Ss in less time) and design his study to control for the effects of the situational factors discussed earlier. If a clinical population is desired, all Ss should be carefully assessed to determine the degree to which their anxiety is currently distressing or bothersome. Then multiple channel assessment of anxiety in the presence of eliciting stimulus conditions should be made. If small animals or other concrete physical stimuli are selected as target objects, and a BAT is to form part of assessment, Ss should be informed how to handle the stimulus object to reduce artifacts. In addition, the BAT should be a “high-demandfor-approach” test, rather than a “legitimizeavoidance” test (see Bernstein, in press) in order to reduce the influence of context variables. In any case, whether extreme or moderate anxiety responses are acceptable, a minimum requirement for S inclusion for any study in this area should be a demonstrably significant increase in anxiety, shown through direct multiple measurement of cognitive, physiological, and observable motoric behavior, as a result of the presence of the eliciting stimuli (see Bernstein, in press; Paul and Bernstein, in

ANALOGUE

RESEARCH

235

press). The use of a BAT itself provides an indirect measure (except under the “lowdemand-for-approach” condition) and a context for obtaining the direct measures. If the less rigorous target behavior analogue (i.e. moderate anxiety) is decided upon, E must be prepared to deal with the consequences of his decision; namely, that the probability of selection of Ss whose verbal and BAT behavior is manipulable by situational factors may be greatly increased. In order to assess the degree of behavior change which could be expected solely on the basis of changes in pre-to-posttest demand, a “demand change” control group could be included in the analogue design. This group would supplement rather than replace an attention-placebo group and would not be given formal treatment, but would simply take two BATS which correspond in time to the pre- and posttreatment tests given to other Ss. The first of these would be a “standard” moderate-demand test, identical to the BAT for all Ss, while the second would be a “maximum-high-demand” test with extreme pressure for approach brought to bear, perhaps with an explicit reward offered for compliance. Treatment effects could then be measured against effects produced by “demand changes” in addition to placebo effects, change produced by life experiences, and the like. Devising extremely (and believably) high demand posttests may not be easy, but the task is not impossible and could very well be worth the effort in terms of increasing the quality of data generated about treatment techniques (see Bernstein, in press for examples). As noted above, the nature of many treatment techniques included in analogue research is such that they are sufficiently far removed from the procedures used in clinical practice to make them “analogues of analogues.” This error can be overcome simply by employing procedures in all phases of treatment which are as close as possible to those used in clinical settings and found effective in previous controlled research. This means that a careful assessment of the major principles of any treatment evaluated in

236

DOUGLAS

A. BERNSTEIN

an analogue design must be undertaken, after which the investigator should be satisfied (a) that no violations of the principles upon which the originally proposed treatment is based occur in his laboratory version and (b) that any modifications are supported by independent, parametric data which show them to be inconsequential. If technique parameters are to be manipulated as independent variables, “standard technique” comparisons as well as “attention-placebo” comparisons should be included, either within a single study, or overlapped in a programmatic series of studies in which other classes of variables are held constant (see Paul, 1969a). In this connection, it seems incumbent upon the investigator to specify, in detail sufficient for purposes of interexperiment comparability, the nature of the treatment procedures used in the analogue. Stating that, “Ss were given systematic desensitization (Wolpe, 1958); is neither adequate nor useful in an experimental study. At a minimum, E should provide information about: the nature, duration, and means of relaxation training; the procedures employed in hierarchy construction and the nature of items; the duration of item presentation, the precise consequences of one or more anxiety signals from S; the means through which clarity of item visualization was assessed, and the like (see Paul, 1969c, pp. 150-158). Similar specification should be included for any treatment technique used. In addition, E should provide clear and highly specific information about the way in which a treatment technique is explained to Ss. In other words, what explicit expectations are they given regarding behavior change? In what ways do those expectations differ across experimental conditions? To what extent is treatment embedded in the same experimental context usually present at pretest? Information regarding the experience and training of therapists should also be a part of published reports of analogue research. In the future, data of this kind should appear more frequently and will hopefully reflect an increase in both the level of clinical experience and the

and GORDON

L. PAUL

intensity and duration of treatment-specific therapist training. This does not mean that graduate students are necessarily inappropriate as Es, but it does mean that any investigatior employing them should be aware of the potential bias they (or any inexperienced person) can introduce into the data and take action to increase their level of skill. (In the process, he should also reduce inter-E variability in treatment procedures to a minimum.) Normally, this could be expected to involve at least one year of general clinical experience along with additional training and experience focused upon clinical use of all analogue treatment techniques. The nature of the monitoring should also be specified. The arguments and suggestions presented above are not meant to imply that all analogue research on small animal phobia published to date is completely useless. Some of it allows cause-effect conclusions, most of it demonstrates that therapy analogues are feasible, and all of it provides information about potential errors while providing clues to the development of ways for avoiding them. Laboratory analogue research is a necessary step in the process of behavior accumulating knowledge about modification techniques, but a further necessary step is the validation of results in the clinical setting. There is a vital interplay between the two; and analogue research methodology should not destroy the relationship.

REFERENCES BANDURA A., GRUSEC J. and MENLOVEF. L. (1967) Vicarious extinction of avoidance behavior, J. Pers. Sot. Psychol. 5, 16-23. BANDURAA., BLANCHARDE. B. and RII-TERB. J. (1969) The relattve efficacy of desensitization and modeling therapeutic approaches for inducing behavioral, affective, and attitudinal changes. J. Pers. Sot. Psychof. 13, 173-199. BANDURAA. and MENLOVEF. L. (1968) Factors determining vicarious extinction of avoidance behavior through symbolic modeling. /. Pen. Sot. Psychol. 8,99-108.

SOME COMMENTS

ON THERAPY

BERNSTEIND. A. (in press) Behavioral fear assessment: anxiety or artifact? In Issrte and Trends in Behavior Therupy (Edited by H. A/DAMS and P. UNXEL), Charles C. Thomas, Springfield, Illinois. BERNSTEIND. A. and BORKOVECT. D. (1971) Progressive &?&ion Training: A Manual for Therapists. (in Bou~ouGouars J C MARKS I M. and MARSET P. (1971) Superiohty sf flooding (implosion) to desensitization for reducing pathological fear, Behav. Res. & Therapy

9, 7-16.

CAMPBELLD. T. and STANLEYJ. C. (1963) Experimental and quasi-experimental designs fir reskarcli on teaching. In Handbook of Research on Teeaching (Edited by N. L. GAGE), Rand-McNally, Chicago. CARPERA., FURSTJ. B. and BRIDGERW. (1969) A brief commentary on the usefulness of studying fears of snakes, J. Abn. Psychol. 74, 413-414. COTLERS. B. (1970) Sex differences and generalization of anxiety reduction with automated desensitization and minimal therapist contact, Behov. Res. & Therapy 8,273-285.

CROWDERJ. E. and THORNTOND. W. (1970) Effects of systematic desensitization, programmed fantasy, and bibliotherapy on a specific fear, Behav. Res. & Therapy

8, 35-41.

DAVISONG. C. (1968) Systematic desensitization as a ;;z;;rconditi&ing process, J. Abnorm. Psycho1 73, D~Mooi W. (1970) Systematic desensitization verses prolonged high intensity stimulation (flooding), J. Bchav.

Ther.

& Exp. Psychiat.

1, 45-52.

GOLDSTEINA. P., HELLERK. and SECHRES~L. B. (1966) Psychotherapy

and the Psychology

of Behavior

Change,

Wiley, New York. HODGSONR. J. and RACHMANS. (1970) An experimental investigation of the implosion technique, Behov. Res. & Therapy

8, 21-27.

KRAPFL J. E. and NAWAS M. M. (1969) Client-therapist relationship factor in systematic desensitization J. Consulting

and Clin. Psycho!.

33, 435-439.

KRAPFL J. E. and NAWAS M. M. (1970) Differential ordering of stimulus presentation in systematic desensitization, J. Abnorm. Psycho/. 75, 333-337. LANG P. J. (1968) Fear reduction and fear behavior: problems in treating a construct. In Research in Psychotherapy. (Edited by J. M. SHLIEN). American Psychological Association, Washington. LANG P. J. (1969) The mechanics of desensitization and the laboratory study of human fear. In Behavior Therapy: Appraisal and Status. (Edited by C. M. FRANKS). McGraw-Hill, New York. LANG P. J. and LAZOV~KA. D. (1963) Experimental desensitization of a phobia, J. Abnorm. Sot. Psycho/. 66, 5 19-525.

LANG P. J., LAZOV~KA. D. and REYNOLDSD. J. (1965) Desensitization, suggestibility and pseudotherapy, /. Abnorm.

Psychol.

70, 395402.

LEVISD. (1970) The phobic test apparatus: An objective measure of human avoidance behavior to small objects, Behov. Res. & Therapy 7, 309-315. LICK J. R. and B~CJTZIN R. R. (1970) Expectancy, demand characteristics, and contact desensitization in behavior change, Behav. Therapy 1, 176-183. M&LYNN F. D. and WILLIAMSC. W. (1970) Systematic desensitization of snake-avoidance under three con~ltr7yl;:suggestion, J. Behav. Ther. & Exp. Psychiat. f (Received

ANALOGUE

RESEARCH

237

MCGLYNN F. D., WILSONA. L. and LINDERL. H. (1970) Systematic desensitization of snake-avoidance with individualized and non-individualized hierarchies, /. Behav. & Exp. Psychiat. 1, 201-204. MILLER H. R. and NAWAS M. M. (1970) Control of aversive stimulus termination in systematic desensitization, Behav. Res. & Therapy 8, 57-61. NAWAS M. M:, FISHMANS. T. and PUCEL J. C. (1970) A standardized desensitization program applicable to group and individual treatments, Behav. Res. & Therapy

8, 49-56.

NAWAS M. M., WEIXH W. V. and FISHMANS. T. (1970) The comparative effectiveness of pairing aversive imagery with relaxation, neutral tasks, and muscular tension in reducing snake phobia, Behav. Res. & Therapy

8, 63-68.

ORNE M. T. (1962) On the social psychology of the psychologicai experiment: with particular reference to demand characteristics and their implications, Amer.

Psychologist

17, 776-783.

PAUL G. L. (1966) Insight vs. Desensitization therapy:

An

Experiment

in

Anxiety

in PsychoReduction,

Stanford University Press, Stanford. PAUL G. L. (1969a) Behavior modification research: design and tactics. In Behavior Therapy: Appraisal and Status. (Edited by C. M. FRANKS). McGraw-Hill, New York. PAUL G. L. (1969b) Outcome of systematic desensitization I: Background procedures, and uncontrolled reports of individual treatment. In Behavior Therapy: Appraisal and Status. (Edited by C. M. FRAXKS). McGraw-Hill. New York. PAUL G. L. (19i9c) Outcome of systematic desensitization II: Controlled investigations of individual treatment, technique variations, and current status. In Behuvior Therapy: Appraisal and Stators. (Edited by C. M. FRANKS). McGraw-Hill, New York. PAUL G. L. and BERNSTEIND. A. (in press) Anxiety and Behavior: Treatment by Systematic Desensitization and Related Techniques, General Learning Press, New

York. PAUL G. L. and TRIMBLER. W. (1970) Recorded vs. “live” relaxation training and hypnotic suggestion: comoarative effectiveness for reducing nhvsioloaical arou’sal and inhibiting stress response,-Bihiv. TheTapy 1. _, 285-302. - __ __-

PUCEL J. C. and NAWAS M. M. (1970) The effects of sex pairings of experimenter and subject on the outcome of systematic desensitization, J. Behav. Ther. & Exp. Psychiat. 1, 103-107. RIMM D. C. and MAHONEYM. J. (1969) The application of reinforcement and participant modeling procedures in the treatment of snake-phobic behavior, Behav. Res. & Therapy 7, 369-376. ROBINSONC. and SUINN R. M. (1969) Group desensitization of a phobia in massed sessions, Behav. Res. & Theraov 7. 3 19-321. ULLMANN E. P.‘and KRASNERL. (1969) A Psychological Approach to Abnormal Behavior. Prentice-Hall,

Englewood Cliffs, New Jersey. WILLIS R. W. and EDWARDSJ. A. (1969) A study of the comparative effectiveness of systematic desensitization and implosive therapy, Behav. Res. & Therapy 7, 387-395.

WOLPE J. (1958) Psychotherapy by Reciprocul Stanford University Press, Stanford.

11 August 197 1)

Inhibition.