- Email: [email protected]
Analog baselines: A critical review of the methodology
Pergamon
Research in Developmental Disabilities, Vol. 16, No. 4, pp. 269-284, 1995 Copyright © 1995 Elsevier Science Ltd Printed in the USA. All rights resen,ed 0891-4222/95 $9.50 + .00
0891-4222(95)00014-3
Analog Baselines: A Critical Review of the Methodology Peter Sturmey San Antonio State School
Analog baselines are an experimental methodology for identifying the functions of maladaptive behavior in the naturally occurring environment (lwata, Dorsey, Slifer, Bauman, & Richman, 1982; lwata et at, 1994). This article identifies a number of potential limitations in this methodology. These include: (a) procedural problems inherent in the use of multielement designs, (b) the fidelity of analog baseline design conditions, (c) the relation of the analog conditions to the naturally occurring environment, (d) a narrow analysis o f behavior limited by an implicit adherence to an ABC model of behavior, (e) a limited acknowledgement of midtifunction and idiosyncratically motivated behaviors, (JO problems in the definition of response classes, and (g) difficulties in the use of analog baselines to design interventions. Future research should attend to three main questions. First, the convergent validity of different assessment methodologies, including analog baselines, should be evaluated. Future research should attend to procedures that can integrate the entire clinical process of referral, identifying the functions of the target behavior, including other methods of identifying the functions of behavior, treatment design, and implementation. Second, assessment failures couM be examined carefully to identify ways of developing this methodology further. Third, the process of designing an intervention depends upon input from many sources of information. The use of analog baselines will be enhanced by a greater understanding of tbe process of clinical decision making.
The functional analysis of individual persons' behavior is the hallmark of the applied behavior analysis of maladaptive behaviors in persons with developmental disabilities (Baer, Wolf, & Risley, 1968; Iwata, Vollmer, & Zarcone, 1990). Requests for reprints should be sent to P. Sturmey, Department of Psychology, San Antonio State School, San Antonio, TX 78214, USA.
269
270
P. Sturmey
Determining the function(s) of target behavior is hypothesized to be essential for the correct design of behavioral interventions (Iwata et al., 1994; Repp, Felce, & Barton, 1988). Functional analysis should identify the reinforcers that maintain target behaviors and that need to be rescheduled, antecedent stimuli that may provoke the behavior, and problematic situations that intervention must address. A functional analysis can identify those interventions that are indicated and contraindicated treatments (Iwata et al., 1994- Fig. 6). A range of assessment technologies has been developed to evaluate the functions of malaclaptive behaviors, including direct observation (Repp et al., 1988; Touchette, MacDonald, & Langer, 1985), structured interviews (O'Neill, Homer, Albin, Storey, & Sprague 1990), psychometric instruments (Sturmey, 1994), and experimental observational methods (Iwata et al., 1982, 1994). One such methodology is analog baselines (Iwata et al., 1982). Using a multielemerit design, three experimental conditions are systematically compared. The three experimental conditions are named Social Disapproval, Academic Demand, and Alone. In Social Disapproval the experimenter appears to be reading and does not interact with the subject unless the subject displays the target behavior. If the subject does display the target behavior then the experimenter should express concern and disapproval and make brief, nonpunitive physical contact with the subject. In the Academic Demand condition a teaching task is taught using least-to-most intrusive prompting. If the target behavior is displayed, teaching is briefly terminated. If the target behavior is still being displayed then a further 30-s delay in task presentation is added. In the Alone condition the person has no access to materials or people. An additional condition, Unstructured Play, is also included to control for the presence of materials and social interaction. In this condition play materials are made available and the experimenter remains close to the subject and interacts contingent upon 30-s absence of the target behavior (Iwata et al., 1982, pp. 10-12). These conditions were developed to simulate the environments that could typically maintain the target behaviors. If high rates of the target behavior are associated with only one condition it is inferred that the function of the target behavior in the natural environment has been identified. Thus, if high rates occur predominantly in the Academic Demand condition it is inferred that the function of the behavior is escapeavoidance. If high rates occur predominantly in the Social Disapproval condition it is inferred that the behavior is maintained by social reinforcement. If high rates occur predominantly in Alone it is inferred that the behavior is reinforced automatically. Thus, the results of analog baselines are used to make inferences concerning the naturally occurring environmental conditions. Analog baselines have become highly influential. Originally developed for the analysis of self-injury, they have also been used to assess topographies such as stereotyped behaviors (Sturmey, Carlsen, Crisp, & Newton, 1988; Vollmer, Marcus, & LeBlanc, 1994), dangerous pill taking (Cfiapman, Fisher, Piazza, &
Analog Baselines
271
Kurtz, 1993), disruptive behavior in persons with acquired brain damage (Pace, Ivancic, & Jefferson, 1994), and aggression, property destruction (Grace, Kahng, & Fisher, 1994), and multiple behavior problems (Derby et al., 1994; Sturmey et al., 1988). The procedure has been modified by including the development of very brief assessments (Wacker et al., 1990), extending the number of sessions required to analyze the functions of the behaviors, analyzing the within-session pattern of responding (Vollmer, Iwata, Zarcone, Smith, & Mazaleski, 1993), and modifying the condition's idiosyncratic ways (Northup et al., 1994; Sturmey, in press-a). Analog baselines have also been used to validate nonexperimental methods of functional assessment (Durand & Crimmins, 1988; Lerman & Iwata, 1993; Taylor & Romanczyk, 1994). Analog baselines have been evaluated as one of the most productive and influential methods of functional analysis (Homer, 1994; Mace, 1994) that rejuvenated interest in the analytic roots of behavioral intervention. Analog baselines have also formed the basis of several new assessment methodologies, including brief, experimental assessments (Wacker et al., 1994) and new naturalistic, correlational methods of assessment (Repp, 1994; Taylor & Can', 1992a, 1992b; Taylor & Romanczyk, 1994). This review critically reviews the role of analog baselines as a method to identify the function of maladaptive behaviors in persons with developmental disabilities and as a methodology to design treatments. PROCEDURAL PROBLEMS
Experimental Design Analog baselines use a multielement design in which several experimental conditions rapidly alternate. Each session typically lasts 10 or 15 min, with breaks in between each session. Blocks of four sessions are scheduled once or twice a day. Thus, analog baselines might take 1 or 2 weeks to complete (Iwata et al., 1994). Others have abbreviated the procedure into a 90 rain schedule of brief reversal designs in which the most promising experimental conditions are compared (Wacker et al., 1990, 1994). The choice of this design contains several assumptions about the nature of the target behavior. First, the behavior must occur at a fairly frequent rate, probably at least several times an hour although more frequent behavior, in at least some of the conditions, would be needed to interpret the graphs. Second, the behavior must respond to rapidly changing experimental conditions. Third, the conditions must be highly discriminable to the subject. Finally, the subject must be able to learn the changing contingencies rapidly. Although many maladaptive behaviors meet these criteria, many important clinically significant behaviors do not. Whitaker (1993) noted that many behavior problems that occur in community settings occur at rates of less than one per day. Stunney (1995) reported a case of functional assessment and intervention in a case of suicidal threats and behavior. In this case analog baselines could not
272
P. Sturmey
be implemented because the behavior was too dangerous and too infrequent. In some cases the severity of the behavior will preclude any extended period in which the person can be permitted to freely engage in the behavior. To address the problem of discrimination of experimental conditions, for example, some authors have attempted to enhance the discriminability of the conditions by wearing different colored coats for each condition (Iwata et al., 1994). In other reports several consecutive sessions of the same condition have been conducted in a reversal design in order to facilitate learning of the contingency (Vollmer, Iwata, Duncan, & Lerman, 1993). In multielement designs the outcome of one condition may depend upon the condition that preceded it. For example, the sequence Social Attention-Social Attention may not be equivalent to the sequence Alone--Social Attention. This is because social attention may be a more powerful reinforcer after the Alone condition in which the person is deprived of social interaction. If such effects are of small magnitude they may not be important. However, when they are large, then considerable error variance is introduced into the procedure, which may render the results uninterpretable. The relation of session length to the schedule of reinforcement maintaining the behavior in the natural environment might critically affect the results of analog baselines. For example, in the Alone condition dramatic extinction might occur if the behavior was maintained on a very rich schedule of reinforcement. However, if the behavior is maintained on a lean schedule, then extinction may never occur during a 15 min period. Results such as these can only be interpreted with a knowledge of the naturally occurring schedules of reinforcement. Choice of Reinforcers The original protocol specifies social praise as a consequence of no target behavior. This may represent a major limitation for some people (Green et al., 1985) and may even introduce artifacts into the assessment procedure. If reinforcing edibles were usually available to the subject in natural environments similar to Academic Demand, then extinction might occur in this condition. This might lead analog baselines to incorrectly identify escape as the function of the target behavior. Some recent articles have begun to incorporate empirical assessment of stimuli as reinforcers for part of their protocol (Taylor & Romanczyk, 1994). Responding to Other Behavior Problems Many individuals present with multiple rather than single behavior problems (Kobe, Mulick, Rash, & Martin, 1994). However, almost all reports of analog baselines, with a few exceptions (Derby et al., 1994; Sturmey et al., 1988), have reported only a single topography. During analog baselines the
Analog Baselines
273
experimenter may have to respond to incontinence, other behavior problems, or the subject's attempts to leave the experimental room. This issue is partially covered in the protocol (Iwata et al., 1982). For example, in the Social Disapproval condition the experimenter is instructed to ignore all other behaviors. However, how the experimenter responds may critically affect results. Sprague and Homer (1992) demonstrated that when one member of a response class was reduced another member of the response class increased. If data are only collected on a single target behavior, such phenomena may go unreported. This issue is also highlighted by the problem of how to manage restraint. The protocol specifies that a medical opinion is sought and a stopping criterion is specified, at which point restraint is applied. The management of restraint may be crucially important because its use may be intimately related to the function of the behavior. In some situations the behavior may be so severe that only partial release from restraint is possible. In other situations the parent/ guardian may decline to give consent for the removal of restraint. Thus, in these situations analogue baselines may be invalidated as they do not approximate to the naturally occurring environment. The stopping criterion selected may also introduce confounds. If the stopping criterion is conservative, then extinction bursts may be shaped up; if a liberal stopping criterion is used, then a complete extinction burst may occur in a way that never does in the natural environment. In either case, knowledge of the use of the restraint device used in the natural environment may be crucial in the interpretation of the results. ASSESSMENT FIDELITY Treatment fidelity has long been recognized as a key methodological requirement for adequate treatment. In the case of analog baselines faithful implementation of the protocol is also required. Sturmey et al. (1988) used a detailed observational measure of the experimenter's behavior for each of the four experimental conditions to ensure that the assessments were correctly implemented. Northup et al. (1994) recorded the percentage of correct consequences delivered by teachers during analog assessments and demonstrated that teachers correctly changed their behavior during each of the different conditions. Documenting treatment integrity can assist in ensuring assessment integrity. However, other more subtle issues in treatment integrity remain. There is considerable latitude in the design and implementation of the four conditions. In the case of choosing a demand, tasks could be selected from: (a) the subject's daily schedule items that are associated with behavior problems; (b) the training objectives, which might be difficult because of a low rate of success on that objective, but which is not necessarily associated with behavior problems; and (c) tasks that are rarely presented to the individual because they are associated with intense behavior problems. There is a wide range of different tasks
274
P. Sturmey
within the pool of potential demand tasks that could be selected. Some tasks might be highly aversive because of high motor effort; other tasks may be aversive because of the extended periods of work involved or the lack of task variability. Other tasks may be aversive because they require physical confinement in one place or temporary loss of movement. Others have suggested that some tasks are avoided because they are associated with social interaction (Carr, 1994; Cart & Durand, 1985). For some subjects with very limited skills, Unstructured Play could approximate to Demand. For example, some individuals have little or no independent leisure skills and have objectives to learn these skills. For these subjects Academic Demand and Unstructured Play may be functionally equivalent. The Alone condition should simulate an unstimulating environment. However, if the subject is alone in a novel assessment environment this may not be equivalent to being ignored in the familiar environment. The function of Alone as an unstimulating environment may depend upon the ambient level of reinforcement prior to the assessment procedure as well as the level of reinforcement in the experimental condition immediately preceding the Alone condition. Some services do not have one-way screens or rooms; thus, it may not be possible to truly implement the Alone condition. Similarly, some self-injury may be so severe that the person can be permitted to freely engage in the behavior (Northup et al., 1994). This situation might better be described as Ignore. In these circumstances it may be difficult to truly implement a condition that validly approximates the removal of all stimulation and social consequences. During Social Disapproval the unacceptable interactions may be mildly reinforcing to some individuals and punishing to others (Fisher, Piazza, Bowman, Hagopian, & Langdon, 1994). It is possible that some experimenters might use disapproval that is sufficiently intense that it approximates to a punishing verbal reprimand. Some subjects may show very high or more or less continuous rates of the target behavior. In such circumstances it is hard to know how to implement a contingency between the target behavior and the experimenter's social disapproval. This situation can make the experimental conditions hard to discriminate for the subject. It may compromise the validity of the condition as an analog of socially reinforced behavior. Analog baselines include the Unstructured Play condition as a condition to control for the presence of materials and high rates of interaction. However, if a subject behaves in certain ways, it may not be possible to learn the new contingencies in a 15 rain session. For example, if the subject shows high behavior rates during Unstructured Play, they may never experience the DRO component embedded in this condition. C O N V E R G E N T VALIDITY Analog baselines were intended as an analog of events believed to be occurring in the natural environment. They are typically conducted in clinic or hos-
Analog Baselines
275
pital settings (Iwata et al., 1994; Wacker et al., 1994). Thus, the issue of their validity is a legitimate one. Whereas there appears to be a growing consensus that questionnaire measures to assess the function of maladaptive behaviors are limited and often unreliable (Sturmey, 1994; Zarcone, Rodgers, Iwata, Rourke, & Dorsey, 1991), there is no such consensus on the status of experimental and naturalistic observation assessment methods. Some authors have elevated analog baselines to be the criterion by which other methods such as questionnaires (Dnrand & Crimmins, 1988) and naturalistic observation methods (Lerman & Iwata, 1993; Taylor & Romanczyk, 1993) are to be validated. Others have noted the limitations of experimental methods of studying behavior (Tinbergen, 1963). Mace (1994) states that " . . . experimental analyses... may not generalize outside the analog conditions" (p. 388), whereas Baer et al. (1968) note that " . . . the analysis of a b e h a v i o r . . , requires a believable demonstration of the events that can be responsible for the occurrence or nonoccurrence of that behavior. An experimenter has achieved an analysis of a behavior when he can exercise control over it" (pp. 93-94). Cart (1994) noted that the context of behavior may be crucial in understanding the specific settings in which behavior occurs, the setting events that set the occasion for a consequence to become highly motivating, and the educational, training, and social environment that forms the context of behavior (Taylor & Carr, 1992a, 1992b; Taylor & Romanczyk, 1994). Analog baselines evaluate the effect of environmental events that occur very close to the target behavior. Thus, analog baselines emphasize the antecedents and consequences of behavior rather than a more broad-based evaluation of environmental determinants. Internal States Internal states, such as discomfort, tension, illness, drug-related states, and arousal (Carr, 1994; O'Neill et al., 1990) are major elements in any comprehensive functional analysis (Skinner, 1989). Internal states can enter a functional analysis of behavior because they can reliably alter the three term contingency. For example, a subject may show demand-related behavior when sleepy but not when alert. Analog baselines explicitly exclude these elements of a functional analysis by randomizing the order of the experimental conditions. If they are a crucial part of the functional analysis, then analog baselines will not yield information about them. Multiple and Idiosyncratic Functions Analog baselines have been used to identify multifunction behaviors on a small number of occasions. Smith, Iwata, Vollmer, and Zarcone (1993) demonstrated that two of three subjects showed multifunction self-injury maintained
276
P. Sturmey
by both attention and demand avoidance. They also demonstrated that the selfinjury was reduced by the treatment indicated by the behavior's two functions. Others have now reported similar results (Day, Hornet, & O'Neill, 1994; Derby et al., 1994). Thus, analog baselines can contribute to the assessment and treatment of multifunction behavior. However, multifunction behavior and idiosyncratically motivated behavior still presents a serious challenge to this methodology. Consider the difficulty in identifying the functions of behavior when it appears that the individual may display the behavior for some of the following reasons: (a) demand-related behavior during aversive internal states, including prior to going to the bathroom, gas, feeling sleepy; (b) escaping aversive confinement such as experienced during taking medicine; and, (c) lack of immediate access to reinforcers, such as waiting for meals and snacks (Sturmey, 1995). This issue is only important if multifunction behaviors are common or clinically significant, lwata et al. (1994) reported that only 5% of their series demonstrated two functions of their behavior. It may be possible that the 5% of cases where no function could be determined from the analog baselines might also contain complex multi function behavior. In cases where there are multiple functions and one function predominates initially, the importance of secondary functions will vary within as treatment proceeds. If treatment of the primary function is implemented successfully, then the proportion of the variance accounted for by secondary functions increases. Although the overall frequency of the target behavior decreases, the management of these secondary functions then becomes more important. Thus, the treatment indicated by the initial functional analysis may be less important as intervention procedes. Response Classes The correct definition of response classes is an issue for any method of functional analysis. However, almost all papers using analog baselines have limited themselves to the analysis of a single topography. However, many persons display whole ecologies of maladaptive responses (Wahler, 1975). Multiple maladaptive responses may also be organized as response classes in which different topographies serve the same function and have the same relationship to environmental events as each other. Responses may be organized as hierarchies or chains of responses (Zlutnick, Mayville, & Moffat, 1975). Pharmacological Treatments Many persons with maladaptive behavior take psychotropic medication for their behavior, including medications for seizures that may also have behavioral effects. Prevalence estimates have been of the order of 30% to 60%, depending on the time of the study, setting, and population studied (Schaal &
Analog Baselines
277
Hackenberg, 1994). Of 12 recently published studies using analog baselines (Neef & Iwata, 1994), only 2 reported any information related to psychotropic or antiseizure medication (Day et al., 1994; Mazaleski, Iwata, Rodgers, Vollmer, & Zarcone, 1994). Psychotropic medication may be an important variable in a functional analysis as it may alter internal states, the reinforcing value of stimuli, the ability of subjects to discriminate conditions, and the rate of learning. The general underreporting of psychotropic medication and possible changes in psychotropic medication during assessment limits the interpretation of this literature and to some extent diminishes its clinical validity. CLINICAL APPLICABILITY OF ANALOG BASELINES Analog baselines identified a single function of self-injury in 90% of 152 cases reported in Iwata et al. (1994) and multiple functions in a further 5% of these cases. Using the 90 min abbreviated form of analog baselines, Derby et al. (1992) were able to identify the functions of maladaptive behaviors in about 50% of 79 cases. Thus, analog baselines can be an effective method of identifying the function of maladaptive behaviors. Iwata et al. (1994) evaluated 156 individuals over an ll-year period. This means that on average this specialized service evaluated approximately 14 persons per year. Treatment was also evaluated for 121 people, or approximately 11 per year. Using the abbreviated procedure, Derby et al. (1994) evaluated 79 people in 3 years or 26 persons per year. However, Iwata et al. (1994) were able to identify functions in 95% of the cases, whereas Derby et al. (1994) identified functions in only about 50% of their cases. It is possible that this difference may reflect the different populations assessed rather than the different methodologies used. However, if this difference does reflect differences in the methodology used, there appears to be a trade off of quality of results against resources invested. Analog baselines have been used mostly with moderate to very high frequency, discrete topographies of moderate to high intensity, such as self-injury, as is typical of much behavioral research (Whitaker, 1993). Analog baselines may be inappropriate or unfeasible for other low frequency, high intensity behaviors, such as some forms of aggression, self-injury, inappropriate sexual behavior, suicide, and fire setting, where the experimental manipulation of the target behavior is undesirable or unethical. In such circumstances, alternate, descriptive methods of functional assessment like ABC diaries (Gelfend & Hartman, 1975) or observational methods (Lerman & Iwata, 1993; Repp et al., 1988; Taylor & Romanczyk, 1994; Taylor & Carr 1992a, 1992b) may be appropriate. When Should Analog Baselines be Used? Analog baselines are a relatively expensive methodology in terms of time and expertise. As shown above, abbreviated methods of assessment may be
278
P. Sturmey
effective in assessing the functions of behavior. Psychometric measures, which are very quick and easy to complete, were developed, in part, as a response to this problem (Durand & Crimmins, 1988). Further, several nonexperimental methods of functional assessment can now be used to accurately identify the functions of behavior (Lerman & Iwata, 1993; Taylor & Carr 1992a, 1992b; Repp et al., 1988; Taylor & Romanczyk, 1994). These nonexperimental methods may be quicker to implement and more easily implemented by a wider range of practitioners than analog baselines. The routine use of analog baselines would therefore not seem appropriate. The question then arises as to when analog baselines should be used. Practitioners need decision-making prostheses to assist them in the rational allocation of resources when they have to evaluate which assessment methodologies to use (Sturmey, in press-a). An example is given in Figure 1. If there are clear medical or provocative environmental causes of the maladaptive behavior evident upon initial assessment, these could be addressed and modified quickly without resorting to further assessment. If the behavior is too dangerous or infrequent, then analog baselines are not appropriate and nonexperimental methods of functional assessment are indicated (Sturmey, 1995). In some situations the function of the behavior may be readily apparent or may be identified rapidly through observational methods, staff interviews, and questionnaires. Even though questionnaires have generally been shown to be unreliable, it is worth noting that they do provide reliable data in a minority of cases, which may, perhaps, be a valid measure of the function of the behavior (Sigafoos, Kerr, & Roberts, 1994). If these methods are not effective, then the abbreviated 90 rain procedure may yield results in a day in about 50% of cases. If these assessments have been cow ducted and the function of the behavior is not apparent, then analog baselines are indicated. If analog baselines do not yield a function of the behaviol; then the practitioner has the option of continuing to refine experimental hypotheses, conducting further experimental studies of the behavior, or using Empirically Derived Consequences (Fisher et al., 1994) or methods of environmental enrichment and contingency management (Vollmer et al., 1994). Practitioners have to decide at which juncture to invest the resources necessary to conduct an analog baseline assessment. Decision-making prostheses such as that shown in Figure 1 may assist them in the rational use of their limited resources. TRANSLATING ASSESSMENT INTO TREATMENT Identifying the single function of the behavior is insufficient to design and implement a treatment program. Other considerations, such as resources, existing interventions, and skills of staff and parents, go into designing the treatment package. Analog baselines have a clear advantage over other assessment methods. Because the experimental conditions approximate to treatment procedures they can often be modified into viable treatments.
Analog Baselines
Any provocative medical or environmental causes?
Yes
279
Treat these.
No Is the behavior too dangerous or too frequent for analog baselines?
Yes
Conduct functional assessment with descriptive data.
No Conduct brief 90-minute, experimental, functional analysis. Are functions apparent? No Conduct full analog baselines. Are functions apparent? No Implement empirically denied consequences procedures. Is an effective treatment desired?
Implement treatment.
I
No Conduct individually tailored, experimental analysis. Is an effective treatment desired? No FIGURE 1. A decision-making prosthesis to guide the selection of different methods of assessing the functions of problem behavior.
Even if analog baselines correctly identify the single function, careful attention must be paid to the natural environment. Specific situations may need to be addressed. If a subject clearly shows escapc-avoidant behavior, analysis of the natural environment may show inappropriate or blatantly provocative use of demands. In this situation, compliance is required from staff or parents rather than a program to train the client to be passive in an inhumane environment. Thus, even when analog baselines correctly identify the function of the
280
P. Sturmey
behavior, the nature of the intervention, if any, requires consideration of many other factors. SUMMARY This article has raised a wide number of criticisms and potential limitations in the use of analog baselines. Some of these issues might be viewed as severely limiting in certain circumstances. However, it should be clearly acknowledged that analog baselines have made a major contribution to the evaluation of maladaptive behaviors. Analog baselines are not an inappropriate methodology; rather, like other methodologies, analog baselines require careful interpretation and analysis. They are appropriate in certain, but not all, circumstances. Validation
Analog baselines may sometimes rapidly and validly identify the functions of a behavior. However, sometimes the results may require careful validation. Future research should attempt to address the integration of analog baseline methods with nonexperimental observational methods of evaluating behaviors. For example, an analog baseline assessment might correctly identify the function of the target behavior, but a scatterplot (Touchette et al., 1985) may be required to assist in identifying the situations that a program must address. Further, an analog baseline may yield information that is rather unrefined. For example, a functional analysis could be conducted on specific aspects of each function. In the case of task avoidance, task difficulty, task rate, type and variety of tasks, reinforcement density, and rate could all be analyzed after a regular analog baseline assessment. Analog baselines could be integrated with naturalistic observations by conducting the analogues in the natural environments, using the actual situations, tasks, and staff that might maintain the behavior. Such procedures might lose some of the experimental control of analog baselines but may gain in ecological validity of their results. Such setting-based functional analyses may also have advantages because they might provide greater incidental opportunities for staff and parent training and might suggest better ideas for treatment programs. There is a tension between experimental and nonexperimental methods; analog baselines lie toward the experimental end of this continuum. Thus, the validity of experimental procedures, such as analog baselines, can be compromised for experimental control. Analog baselines can indicate effective and noneffective treatments (Iwata et al., 1994). However, incremental treatment validity is a key aspect of validity. A key question is whether analog baselines assist in designing more effective treatments than otherwise would have been devised using all the other
Analog Baselines
281
information already available. This is a complex and subtle question that research has not hitherto addressed. Assessment and Treatment Failures Although the limitations of analog baselines have been implicitly acknowledged through the various refinements, there have been few reports of assessment failures. Failures in behavior therapy are a fertile area where future developments can occur (Foa & Emmelkamp, 1983). Sturmey (in press-a) reported that one subject exhibited between 2 and 10 million acts of selfinjury, whereas a variety of functional assessments, including analog baselines, were conducted. Thus, conducting functional assessments can, at times, be very costly to the subject. Iwata et al. (1994) reported that functions could not be identified in approximately 10% of their series, and Derby et al. (1992), using the abbreviated method, reported that functions could not be identified in approximately 50% of cases. Others have reported cases of failure of analog baselines to identify the function(s) of a target behavior and have gone on to use methods such as Empirically Derived Consequences (Fisher et al., 1994) or environmental enrichment and contingency management (Vollmer et al., 1994). Clinical Process and Functional Analysis The process by which an intervention is designed and evaluated is complex and not clearly understood (Sturmey, in press-b). The relationship of analog baselines to this broader process is subtle. For example, there are often hints that the manner in which tasks and materials are selected, and the identification of target behaviors(s), depends upon a complex process of information collection and processing. Further, this clinical process is a social one in which professionals, staff, and parents negotiate priorities, set limits on acceptable practice, and, to a certain extent, develop a shared clinical agenda. The identification of the function of the target behavior and the selection of the method(s) by which this should be done are only two elements in this broader process. There are times when other methods of identifying the functions of the target behavior will be quicker, cheaper, and more acceptable than analog baselines. At other times analog baselines may be worthwhile because they solve a clinical problem that other methods cannot solve, or because the use of analog baselines show a clear demonstration of a serious effort to attempt to solve a difficult problem. However, the correct identification of the function of the target behavior is only one element in the design of the treatment. Future research in this area should focus, not only on the methodologies of functional analysis, but also the broader process of treatment design, of which analog baselines may be one part.
282
P. Smrmey
REFERENCES Ba~r, D. M., Wolf, M. M., & Risley, T. R. (1968). Some current dimensions of applied behavior analysis. Journal of Applied Behavior Analysis, 1, 91-97. Can', E. G. (1994). Emerging themes. Journal of Applied Behavior Analysis, 27, 393-399. Cart, E. G., & Durand, V. M. (1985). Reducing behavior problems through functional communication training. Journal of Applied Behavior Analysis, 18, 111-116. Chapman, S., Fisher, W., Piazza, C. C., & Kurtz, P. E (1993). Functional assessment and treatment of life-threatening drug ingestion in a dually diagnosed youth. Journal of Applied Behavior Analysis, 26, 255-256. Day, H. M., Homer, IL M., & O'Neill, R. E. (1994). Multiple functions of problem behaviors: Assessment and intervention. Journal of Applied Behavior Analysis, 27, 279-289. Derby, K. M., Wackcr, D. P., Peck, S., Sasso, G., DcRaad, A., Berg, W., Asmus, J., & Ulrich, S. (1994). Functional analysis of separate topographies of aberrant behavior. Journal of Applied Behavior Analysis, 27, 267-278. Derby, K. M., Wacker, D. P., Sasso, G., Ste~ge, M., Northup, J., Cigrand, K., & Asmus, J. (1992). Brief functional assessment techniques to evaluate behavior in an outpatient setting: A summary of 79 cases. Journal of Applied Behavior Analysis, 25, 71 3-721. Durand, V. M., & Crimmins, D. B. (1988). The Motivation Assessment Scale (MAS) administration guide. Topeka, KS: Monaco & Associates. Durand, V. M., & Crimmins, D. B. (1988). Identifying the variables maintaining self-injurious behavior. Journal of Autism and Developmental Disabilities, 18, 99-117. Fisher, W., Piazza, C. C., Bowman, L. G., Hagopian, L. P., & Langdon, N. A. (1994). Empirically derived consequences: A data-based method for prescribing treatments for destructive behaviors. Research in Developmental Disabilities, 15, 133-149. Foa, E. B., & Emmelkamp, P. M. G. (1983). Failures in behavior therapy. New York: John Wiley. Gelfend, D. M., & Hartman, D. P. (1975). Child behavior analysis and therapy. New York: Pergamon. Grace, N. C., Kahng, S. W., & Fisher, W. W. (1994). Balancing social acceptability with treatment effectiveness of an intrusive procedure: A case report. Journal of Applied Behavior Analysis, 27, 171-172. Green, C. W., Reid, D. H., White, L. K., Halford, R. C., Brittain, D. P., & Gardner, S. M. (1985). Identifying reinforcers for persons with profound handicaps: Staff opinion versus systematic assessment of preferences. Journal of Applied Behavior Analysis, 21, 31-43. Homer, R. H. (1994). Functional assessment: Contributions and future directions. Journal of Applied Behavior Analysis, 27, 401-404. Iwata, B. A., Dorsey, M. E, Slifer, K. J., Bauman, K. E., & Richman, G. S. (1982). Toward a functional analysis of self-injury. Analysis and Intervention in Developmental Disabilities, 2, 1-20. Iwata, B. A., Pace, G. M., Dorsey, M. E, Zarcone, J. E, VoUmer, T. R., Smith, R. G., Rodgers, T. A., Lerman, D. C., Shore, B. A., Mazaleski, J. L., Goh, H.-L., Cowdery, G. E., Kalsher, M. L, McCosh, K. C., & Willis, K. D. (1994). The functional analysis of self-injury: An experimental-cpidemiological analysis. Journal of Applied Behavior Analysis, 27, 215-240. Iwata, B. A., Vollmer, T. R., & Zarcone, J. R. (1990). The experimental (functional) analysis of behavior disorders: Methodology, applications and limitations. In A. C. Repp & Singh, N. N. (Eds.), Perspectives on the use of aversive and nonaversive interventions for persons with developmental disabilities (pp. 301-330). Sycamore, IL: Sycamore Press. Kobe, E H., Mulick, L A., Rash, T. A., & Martin, J. (1994). Nonambulatory persons with profound mental retardation: Physical, developmental, and behavioral characteristics. Research in Developmental Disabilities, 15, 413-423. Lerman, D. C., & Iwata, B. A. (1993). Descriptive and experimental analysis of variables maintaining self-injurious behavior. Journal of Applied Behavior Analysis, 26, 293-319.
Analog Baselines
283
Mace, E C. (1994). The significance and future of functional analysis methodologies. Journal of Applied BehaviorAnalysis, 27, 385-392. Mazaleski, J. L., Iwata, B. A., Rodgers, T. A., Vollmer, T. R., & Zarcone, J. R. (1994). Protective equipment astreatment for stereotypic handmouthing: Sensory extinction or punishment effects? Journal of Applied BehaviorAnalysis, 27, 345-355. Neef, N. A., & Iwata, B. A. (1994). Current research on functional analysis methodologies: An introduction. Journal of Applied BehaviorAnalysis, 27, 2121-214. Northup, J., Wacker, D. P., Berg, W. K., Kelly, L., Sasso, G., & DeRaad, A. (1994). The treatment of severe behavior problem in a school setting using a technical assistance model. Journal of Applied BehaviorAnalysis, 27, 33-47. O'Neill, R. E., Homer, R. H., Albin, R. W., Storey, K. & Sprague, J. R. (1990). Functionalanalysis of problem behavior:A practical assessment guide. Sycamore, IL: Sycamore Press. Pace, G. M., Ivancic, M. T., & Jefferson, G. (1994). Stimulus fading as treatment for obscenity in a brain damaged adult. Journal of Applied BehaviorAnalysis, 27, 301-305. Repp, A. C. (1994). Commentaries on functional analysis procedures for school-based behavior problems. Journal of Applied Behavior Analysis, 27, 409---411. Repp, A. C., Felce, D., & Barton, L. E. (1988). Basing treatment of stereotypic and self-injurious behaviors on hypotheses of their causes. Journal of Applied BehaviorAnalysis, 21, 281-289. Schaal, D. W., & Hackenberg, T. (1994). Toward a functional analysis of drug treatment for behavior of people with developmental disabilities. American Journal of Mental Retardation, 99, 125-140. Sigafoos, J., Kerr, M., & Roberts, D. (1994). Inter-rater reliability of the Motivation Assessment Scale: Failure to replicate with aggressive behavior. Research in Developmental Disabilities, 15, 333-342. Skinner, B. E (1989). Science and human behavior. New York: Macmillan. Smith, R. G., Iwata, B. A., Vollmer, T. R., & Zarcone, J. R. (1993). Experimental analysis of multiply controlled self-injury. Journal of Applied Behavior Analysis, 26, 183-196. Sprague, J. R., & Homer, R. H. (1992). Covariation within functional response classes: Implications for treatment of severe problem behavior. Journal of Applied BehaviorAnalysis, 25, 735-745. Sturmey, P. (1994). Assessing the functions of aberrant behaviors: A review of psychometric instruments. Journal of Autism and Developmental Disabilities, 24, 293-303. Sturmey, P. (in press-a). Assessing the functions of self-injurious behavior: A case of assessment failure. Journal of Behavior Therapy and Experimental Psychiatry, 25. Sturmey, P. (in press-b). Functionalanalysis in clinicalpsychology. New York: Wiley. Sturmey, P. (1995). Suicidal threats and behavior in a person with developmental disabilities: Effective psychiatric monitoring based upon a functional assessment. Behavioral Interventions and Residential Treatment, 9, 235-245. Sturmey, P., Carlsen, A., Crisp, A. G., & Newton, T. (1988). A functional analysis of multiple aberrant responses: A refinement and extension of Iwata et al.'s, methodology. Journal of Mental Deficiency Research, 32, 31-46. Taylor, J. C., & Carr, E. G. (1992a). Severe problem behaviors related to social interaction: I. Attention seeking and social avoidance. Behavior Modification, 16, 305-335. Taylor, J. C., & Cart, E. G. (1992b). Severe problem behaviors related to social interaction: 1I. A system analysis. BehaviorModification, 16, 336-371. Taylor, J. C., & Romanczyk, R. G. (1994). Generating hypotheses about the function of student behavior by observing teacher behavior. Journal of Applied BehaviorAnalysis, 27, 251-265. Tinbergen, N. (1963). On the aims and methodology of ethology. Zeitschrififur ~erpsychologie, 20, 410-432. Touchette, P. E., MacDonald, R. E, & Langer, S. N. (1985). A scatter plot for identifying stimulus control of problem behavior. Journal of Applied BehaviorAnalysis, 18, 343-35 i. Vollmer, T. R., Iwata, B. A., Duncan, B, A., & Lerman, D. (1993). Extension of multielement functional analysis using reversal-type design. Journal of Developmental and PhysicalDevelopment, 5, 311-325.
284
P Sturmey
Vollmer, T. R., Iwata, B. A., Zarcone, J. R., Smith, R. G., & MaT~leski, J. L. (1993, May). Withinsession patterns of self-injury as indicators of behavioralfunction. Paper presented to the 19th annual convention of the Association for Behavior Analysis-International, Chicago. Vollmer, T R., Marcus, B. A., & LeBlanc, L. (1994). Treatment of self-injury and hand mouthing following inconclusive functional analysis. Journal of Applied BehaviorAnalysis, 27, 331-344. Wacker, D., Berg, W., Cooper, L. J., Derby, K. M., Stcege, M. W., Northup, J. H., & Sasso, G. (1994). The impact of functional analysis methodology on outpatient clinic services. Journal of Applied BehaviorAnalysis, 27, 405-407. Wacker, D., Stcege, M., Noahup, J., Reimers, T., Berg, W., & Sasso, G. (1990). Use of functional analysis and acceptability measures to assess and treat severe behavior problems: An outpatient clinic model. In A. C. Repp & N. N. Singh (F__xls.),Perspectives on the use ofaversive andnonaversive interventionsfor persons with developmental disa~lities (pp. 349-359). Sycamore, IL: Sycamore Press. Wahler, R. G. (1975). Some structural aspects of deviant behavior. Journal of Applied Behavior Analysis, 8, 27-42. Whitaker, S. (1993). The reduction of aggression in people with learning disabilities: A review of psychological methods. British Journal of Clinical Psychology,32, 1-37. Zarcone, J. R., Rodgers, T. A., Iwata, B. A., Rourke, D. A., & Dorsey, M. E (1991). Reliability analysis of the Motivational Assessment Scale. Research in Developmental Disabilities, 12, 349-360. Zlutnick, S., MayviUe, N. J., & Moffat, S. (1975). Modification of seizure disorders: The interruption of behavior chains. Journal of Applied BehoviorAnalysis, 8, 1-13.
Research in Developmental Disabilities, Vol. 16, No. 4, pp. 269-284, 1995 Copyright © 1995 Elsevier Science Ltd Printed in the USA. All rights resen,ed 0891-4222/95 $9.50 + .00
0891-4222(95)00014-3
Analog Baselines: A Critical Review of the Methodology Peter Sturmey San Antonio State School
Analog baselines are an experimental methodology for identifying the functions of maladaptive behavior in the naturally occurring environment (lwata, Dorsey, Slifer, Bauman, & Richman, 1982; lwata et at, 1994). This article identifies a number of potential limitations in this methodology. These include: (a) procedural problems inherent in the use of multielement designs, (b) the fidelity of analog baseline design conditions, (c) the relation of the analog conditions to the naturally occurring environment, (d) a narrow analysis o f behavior limited by an implicit adherence to an ABC model of behavior, (e) a limited acknowledgement of midtifunction and idiosyncratically motivated behaviors, (JO problems in the definition of response classes, and (g) difficulties in the use of analog baselines to design interventions. Future research should attend to three main questions. First, the convergent validity of different assessment methodologies, including analog baselines, should be evaluated. Future research should attend to procedures that can integrate the entire clinical process of referral, identifying the functions of the target behavior, including other methods of identifying the functions of behavior, treatment design, and implementation. Second, assessment failures couM be examined carefully to identify ways of developing this methodology further. Third, the process of designing an intervention depends upon input from many sources of information. The use of analog baselines will be enhanced by a greater understanding of tbe process of clinical decision making.
The functional analysis of individual persons' behavior is the hallmark of the applied behavior analysis of maladaptive behaviors in persons with developmental disabilities (Baer, Wolf, & Risley, 1968; Iwata, Vollmer, & Zarcone, 1990). Requests for reprints should be sent to P. Sturmey, Department of Psychology, San Antonio State School, San Antonio, TX 78214, USA.
269
270
P. Sturmey
Determining the function(s) of target behavior is hypothesized to be essential for the correct design of behavioral interventions (Iwata et al., 1994; Repp, Felce, & Barton, 1988). Functional analysis should identify the reinforcers that maintain target behaviors and that need to be rescheduled, antecedent stimuli that may provoke the behavior, and problematic situations that intervention must address. A functional analysis can identify those interventions that are indicated and contraindicated treatments (Iwata et al., 1994- Fig. 6). A range of assessment technologies has been developed to evaluate the functions of malaclaptive behaviors, including direct observation (Repp et al., 1988; Touchette, MacDonald, & Langer, 1985), structured interviews (O'Neill, Homer, Albin, Storey, & Sprague 1990), psychometric instruments (Sturmey, 1994), and experimental observational methods (Iwata et al., 1982, 1994). One such methodology is analog baselines (Iwata et al., 1982). Using a multielemerit design, three experimental conditions are systematically compared. The three experimental conditions are named Social Disapproval, Academic Demand, and Alone. In Social Disapproval the experimenter appears to be reading and does not interact with the subject unless the subject displays the target behavior. If the subject does display the target behavior then the experimenter should express concern and disapproval and make brief, nonpunitive physical contact with the subject. In the Academic Demand condition a teaching task is taught using least-to-most intrusive prompting. If the target behavior is displayed, teaching is briefly terminated. If the target behavior is still being displayed then a further 30-s delay in task presentation is added. In the Alone condition the person has no access to materials or people. An additional condition, Unstructured Play, is also included to control for the presence of materials and social interaction. In this condition play materials are made available and the experimenter remains close to the subject and interacts contingent upon 30-s absence of the target behavior (Iwata et al., 1982, pp. 10-12). These conditions were developed to simulate the environments that could typically maintain the target behaviors. If high rates of the target behavior are associated with only one condition it is inferred that the function of the target behavior in the natural environment has been identified. Thus, if high rates occur predominantly in the Academic Demand condition it is inferred that the function of the behavior is escapeavoidance. If high rates occur predominantly in the Social Disapproval condition it is inferred that the behavior is maintained by social reinforcement. If high rates occur predominantly in Alone it is inferred that the behavior is reinforced automatically. Thus, the results of analog baselines are used to make inferences concerning the naturally occurring environmental conditions. Analog baselines have become highly influential. Originally developed for the analysis of self-injury, they have also been used to assess topographies such as stereotyped behaviors (Sturmey, Carlsen, Crisp, & Newton, 1988; Vollmer, Marcus, & LeBlanc, 1994), dangerous pill taking (Cfiapman, Fisher, Piazza, &
Analog Baselines
271
Kurtz, 1993), disruptive behavior in persons with acquired brain damage (Pace, Ivancic, & Jefferson, 1994), and aggression, property destruction (Grace, Kahng, & Fisher, 1994), and multiple behavior problems (Derby et al., 1994; Sturmey et al., 1988). The procedure has been modified by including the development of very brief assessments (Wacker et al., 1990), extending the number of sessions required to analyze the functions of the behaviors, analyzing the within-session pattern of responding (Vollmer, Iwata, Zarcone, Smith, & Mazaleski, 1993), and modifying the condition's idiosyncratic ways (Northup et al., 1994; Sturmey, in press-a). Analog baselines have also been used to validate nonexperimental methods of functional assessment (Durand & Crimmins, 1988; Lerman & Iwata, 1993; Taylor & Romanczyk, 1994). Analog baselines have been evaluated as one of the most productive and influential methods of functional analysis (Homer, 1994; Mace, 1994) that rejuvenated interest in the analytic roots of behavioral intervention. Analog baselines have also formed the basis of several new assessment methodologies, including brief, experimental assessments (Wacker et al., 1994) and new naturalistic, correlational methods of assessment (Repp, 1994; Taylor & Can', 1992a, 1992b; Taylor & Romanczyk, 1994). This review critically reviews the role of analog baselines as a method to identify the function of maladaptive behaviors in persons with developmental disabilities and as a methodology to design treatments. PROCEDURAL PROBLEMS
Experimental Design Analog baselines use a multielement design in which several experimental conditions rapidly alternate. Each session typically lasts 10 or 15 min, with breaks in between each session. Blocks of four sessions are scheduled once or twice a day. Thus, analog baselines might take 1 or 2 weeks to complete (Iwata et al., 1994). Others have abbreviated the procedure into a 90 rain schedule of brief reversal designs in which the most promising experimental conditions are compared (Wacker et al., 1990, 1994). The choice of this design contains several assumptions about the nature of the target behavior. First, the behavior must occur at a fairly frequent rate, probably at least several times an hour although more frequent behavior, in at least some of the conditions, would be needed to interpret the graphs. Second, the behavior must respond to rapidly changing experimental conditions. Third, the conditions must be highly discriminable to the subject. Finally, the subject must be able to learn the changing contingencies rapidly. Although many maladaptive behaviors meet these criteria, many important clinically significant behaviors do not. Whitaker (1993) noted that many behavior problems that occur in community settings occur at rates of less than one per day. Stunney (1995) reported a case of functional assessment and intervention in a case of suicidal threats and behavior. In this case analog baselines could not
272
P. Sturmey
be implemented because the behavior was too dangerous and too infrequent. In some cases the severity of the behavior will preclude any extended period in which the person can be permitted to freely engage in the behavior. To address the problem of discrimination of experimental conditions, for example, some authors have attempted to enhance the discriminability of the conditions by wearing different colored coats for each condition (Iwata et al., 1994). In other reports several consecutive sessions of the same condition have been conducted in a reversal design in order to facilitate learning of the contingency (Vollmer, Iwata, Duncan, & Lerman, 1993). In multielement designs the outcome of one condition may depend upon the condition that preceded it. For example, the sequence Social Attention-Social Attention may not be equivalent to the sequence Alone--Social Attention. This is because social attention may be a more powerful reinforcer after the Alone condition in which the person is deprived of social interaction. If such effects are of small magnitude they may not be important. However, when they are large, then considerable error variance is introduced into the procedure, which may render the results uninterpretable. The relation of session length to the schedule of reinforcement maintaining the behavior in the natural environment might critically affect the results of analog baselines. For example, in the Alone condition dramatic extinction might occur if the behavior was maintained on a very rich schedule of reinforcement. However, if the behavior is maintained on a lean schedule, then extinction may never occur during a 15 min period. Results such as these can only be interpreted with a knowledge of the naturally occurring schedules of reinforcement. Choice of Reinforcers The original protocol specifies social praise as a consequence of no target behavior. This may represent a major limitation for some people (Green et al., 1985) and may even introduce artifacts into the assessment procedure. If reinforcing edibles were usually available to the subject in natural environments similar to Academic Demand, then extinction might occur in this condition. This might lead analog baselines to incorrectly identify escape as the function of the target behavior. Some recent articles have begun to incorporate empirical assessment of stimuli as reinforcers for part of their protocol (Taylor & Romanczyk, 1994). Responding to Other Behavior Problems Many individuals present with multiple rather than single behavior problems (Kobe, Mulick, Rash, & Martin, 1994). However, almost all reports of analog baselines, with a few exceptions (Derby et al., 1994; Sturmey et al., 1988), have reported only a single topography. During analog baselines the
Analog Baselines
273
experimenter may have to respond to incontinence, other behavior problems, or the subject's attempts to leave the experimental room. This issue is partially covered in the protocol (Iwata et al., 1982). For example, in the Social Disapproval condition the experimenter is instructed to ignore all other behaviors. However, how the experimenter responds may critically affect results. Sprague and Homer (1992) demonstrated that when one member of a response class was reduced another member of the response class increased. If data are only collected on a single target behavior, such phenomena may go unreported. This issue is also highlighted by the problem of how to manage restraint. The protocol specifies that a medical opinion is sought and a stopping criterion is specified, at which point restraint is applied. The management of restraint may be crucially important because its use may be intimately related to the function of the behavior. In some situations the behavior may be so severe that only partial release from restraint is possible. In other situations the parent/ guardian may decline to give consent for the removal of restraint. Thus, in these situations analogue baselines may be invalidated as they do not approximate to the naturally occurring environment. The stopping criterion selected may also introduce confounds. If the stopping criterion is conservative, then extinction bursts may be shaped up; if a liberal stopping criterion is used, then a complete extinction burst may occur in a way that never does in the natural environment. In either case, knowledge of the use of the restraint device used in the natural environment may be crucial in the interpretation of the results. ASSESSMENT FIDELITY Treatment fidelity has long been recognized as a key methodological requirement for adequate treatment. In the case of analog baselines faithful implementation of the protocol is also required. Sturmey et al. (1988) used a detailed observational measure of the experimenter's behavior for each of the four experimental conditions to ensure that the assessments were correctly implemented. Northup et al. (1994) recorded the percentage of correct consequences delivered by teachers during analog assessments and demonstrated that teachers correctly changed their behavior during each of the different conditions. Documenting treatment integrity can assist in ensuring assessment integrity. However, other more subtle issues in treatment integrity remain. There is considerable latitude in the design and implementation of the four conditions. In the case of choosing a demand, tasks could be selected from: (a) the subject's daily schedule items that are associated with behavior problems; (b) the training objectives, which might be difficult because of a low rate of success on that objective, but which is not necessarily associated with behavior problems; and (c) tasks that are rarely presented to the individual because they are associated with intense behavior problems. There is a wide range of different tasks
274
P. Sturmey
within the pool of potential demand tasks that could be selected. Some tasks might be highly aversive because of high motor effort; other tasks may be aversive because of the extended periods of work involved or the lack of task variability. Other tasks may be aversive because they require physical confinement in one place or temporary loss of movement. Others have suggested that some tasks are avoided because they are associated with social interaction (Carr, 1994; Cart & Durand, 1985). For some subjects with very limited skills, Unstructured Play could approximate to Demand. For example, some individuals have little or no independent leisure skills and have objectives to learn these skills. For these subjects Academic Demand and Unstructured Play may be functionally equivalent. The Alone condition should simulate an unstimulating environment. However, if the subject is alone in a novel assessment environment this may not be equivalent to being ignored in the familiar environment. The function of Alone as an unstimulating environment may depend upon the ambient level of reinforcement prior to the assessment procedure as well as the level of reinforcement in the experimental condition immediately preceding the Alone condition. Some services do not have one-way screens or rooms; thus, it may not be possible to truly implement the Alone condition. Similarly, some self-injury may be so severe that the person can be permitted to freely engage in the behavior (Northup et al., 1994). This situation might better be described as Ignore. In these circumstances it may be difficult to truly implement a condition that validly approximates the removal of all stimulation and social consequences. During Social Disapproval the unacceptable interactions may be mildly reinforcing to some individuals and punishing to others (Fisher, Piazza, Bowman, Hagopian, & Langdon, 1994). It is possible that some experimenters might use disapproval that is sufficiently intense that it approximates to a punishing verbal reprimand. Some subjects may show very high or more or less continuous rates of the target behavior. In such circumstances it is hard to know how to implement a contingency between the target behavior and the experimenter's social disapproval. This situation can make the experimental conditions hard to discriminate for the subject. It may compromise the validity of the condition as an analog of socially reinforced behavior. Analog baselines include the Unstructured Play condition as a condition to control for the presence of materials and high rates of interaction. However, if a subject behaves in certain ways, it may not be possible to learn the new contingencies in a 15 rain session. For example, if the subject shows high behavior rates during Unstructured Play, they may never experience the DRO component embedded in this condition. C O N V E R G E N T VALIDITY Analog baselines were intended as an analog of events believed to be occurring in the natural environment. They are typically conducted in clinic or hos-
Analog Baselines
275
pital settings (Iwata et al., 1994; Wacker et al., 1994). Thus, the issue of their validity is a legitimate one. Whereas there appears to be a growing consensus that questionnaire measures to assess the function of maladaptive behaviors are limited and often unreliable (Sturmey, 1994; Zarcone, Rodgers, Iwata, Rourke, & Dorsey, 1991), there is no such consensus on the status of experimental and naturalistic observation assessment methods. Some authors have elevated analog baselines to be the criterion by which other methods such as questionnaires (Dnrand & Crimmins, 1988) and naturalistic observation methods (Lerman & Iwata, 1993; Taylor & Romanczyk, 1993) are to be validated. Others have noted the limitations of experimental methods of studying behavior (Tinbergen, 1963). Mace (1994) states that " . . . experimental analyses... may not generalize outside the analog conditions" (p. 388), whereas Baer et al. (1968) note that " . . . the analysis of a b e h a v i o r . . , requires a believable demonstration of the events that can be responsible for the occurrence or nonoccurrence of that behavior. An experimenter has achieved an analysis of a behavior when he can exercise control over it" (pp. 93-94). Cart (1994) noted that the context of behavior may be crucial in understanding the specific settings in which behavior occurs, the setting events that set the occasion for a consequence to become highly motivating, and the educational, training, and social environment that forms the context of behavior (Taylor & Carr, 1992a, 1992b; Taylor & Romanczyk, 1994). Analog baselines evaluate the effect of environmental events that occur very close to the target behavior. Thus, analog baselines emphasize the antecedents and consequences of behavior rather than a more broad-based evaluation of environmental determinants. Internal States Internal states, such as discomfort, tension, illness, drug-related states, and arousal (Carr, 1994; O'Neill et al., 1990) are major elements in any comprehensive functional analysis (Skinner, 1989). Internal states can enter a functional analysis of behavior because they can reliably alter the three term contingency. For example, a subject may show demand-related behavior when sleepy but not when alert. Analog baselines explicitly exclude these elements of a functional analysis by randomizing the order of the experimental conditions. If they are a crucial part of the functional analysis, then analog baselines will not yield information about them. Multiple and Idiosyncratic Functions Analog baselines have been used to identify multifunction behaviors on a small number of occasions. Smith, Iwata, Vollmer, and Zarcone (1993) demonstrated that two of three subjects showed multifunction self-injury maintained
276
P. Sturmey
by both attention and demand avoidance. They also demonstrated that the selfinjury was reduced by the treatment indicated by the behavior's two functions. Others have now reported similar results (Day, Hornet, & O'Neill, 1994; Derby et al., 1994). Thus, analog baselines can contribute to the assessment and treatment of multifunction behavior. However, multifunction behavior and idiosyncratically motivated behavior still presents a serious challenge to this methodology. Consider the difficulty in identifying the functions of behavior when it appears that the individual may display the behavior for some of the following reasons: (a) demand-related behavior during aversive internal states, including prior to going to the bathroom, gas, feeling sleepy; (b) escaping aversive confinement such as experienced during taking medicine; and, (c) lack of immediate access to reinforcers, such as waiting for meals and snacks (Sturmey, 1995). This issue is only important if multifunction behaviors are common or clinically significant, lwata et al. (1994) reported that only 5% of their series demonstrated two functions of their behavior. It may be possible that the 5% of cases where no function could be determined from the analog baselines might also contain complex multi function behavior. In cases where there are multiple functions and one function predominates initially, the importance of secondary functions will vary within as treatment proceeds. If treatment of the primary function is implemented successfully, then the proportion of the variance accounted for by secondary functions increases. Although the overall frequency of the target behavior decreases, the management of these secondary functions then becomes more important. Thus, the treatment indicated by the initial functional analysis may be less important as intervention procedes. Response Classes The correct definition of response classes is an issue for any method of functional analysis. However, almost all papers using analog baselines have limited themselves to the analysis of a single topography. However, many persons display whole ecologies of maladaptive responses (Wahler, 1975). Multiple maladaptive responses may also be organized as response classes in which different topographies serve the same function and have the same relationship to environmental events as each other. Responses may be organized as hierarchies or chains of responses (Zlutnick, Mayville, & Moffat, 1975). Pharmacological Treatments Many persons with maladaptive behavior take psychotropic medication for their behavior, including medications for seizures that may also have behavioral effects. Prevalence estimates have been of the order of 30% to 60%, depending on the time of the study, setting, and population studied (Schaal &
Analog Baselines
277
Hackenberg, 1994). Of 12 recently published studies using analog baselines (Neef & Iwata, 1994), only 2 reported any information related to psychotropic or antiseizure medication (Day et al., 1994; Mazaleski, Iwata, Rodgers, Vollmer, & Zarcone, 1994). Psychotropic medication may be an important variable in a functional analysis as it may alter internal states, the reinforcing value of stimuli, the ability of subjects to discriminate conditions, and the rate of learning. The general underreporting of psychotropic medication and possible changes in psychotropic medication during assessment limits the interpretation of this literature and to some extent diminishes its clinical validity. CLINICAL APPLICABILITY OF ANALOG BASELINES Analog baselines identified a single function of self-injury in 90% of 152 cases reported in Iwata et al. (1994) and multiple functions in a further 5% of these cases. Using the 90 min abbreviated form of analog baselines, Derby et al. (1992) were able to identify the functions of maladaptive behaviors in about 50% of 79 cases. Thus, analog baselines can be an effective method of identifying the function of maladaptive behaviors. Iwata et al. (1994) evaluated 156 individuals over an ll-year period. This means that on average this specialized service evaluated approximately 14 persons per year. Treatment was also evaluated for 121 people, or approximately 11 per year. Using the abbreviated procedure, Derby et al. (1994) evaluated 79 people in 3 years or 26 persons per year. However, Iwata et al. (1994) were able to identify functions in 95% of the cases, whereas Derby et al. (1994) identified functions in only about 50% of their cases. It is possible that this difference may reflect the different populations assessed rather than the different methodologies used. However, if this difference does reflect differences in the methodology used, there appears to be a trade off of quality of results against resources invested. Analog baselines have been used mostly with moderate to very high frequency, discrete topographies of moderate to high intensity, such as self-injury, as is typical of much behavioral research (Whitaker, 1993). Analog baselines may be inappropriate or unfeasible for other low frequency, high intensity behaviors, such as some forms of aggression, self-injury, inappropriate sexual behavior, suicide, and fire setting, where the experimental manipulation of the target behavior is undesirable or unethical. In such circumstances, alternate, descriptive methods of functional assessment like ABC diaries (Gelfend & Hartman, 1975) or observational methods (Lerman & Iwata, 1993; Repp et al., 1988; Taylor & Romanczyk, 1994; Taylor & Carr 1992a, 1992b) may be appropriate. When Should Analog Baselines be Used? Analog baselines are a relatively expensive methodology in terms of time and expertise. As shown above, abbreviated methods of assessment may be
278
P. Sturmey
effective in assessing the functions of behavior. Psychometric measures, which are very quick and easy to complete, were developed, in part, as a response to this problem (Durand & Crimmins, 1988). Further, several nonexperimental methods of functional assessment can now be used to accurately identify the functions of behavior (Lerman & Iwata, 1993; Taylor & Carr 1992a, 1992b; Repp et al., 1988; Taylor & Romanczyk, 1994). These nonexperimental methods may be quicker to implement and more easily implemented by a wider range of practitioners than analog baselines. The routine use of analog baselines would therefore not seem appropriate. The question then arises as to when analog baselines should be used. Practitioners need decision-making prostheses to assist them in the rational allocation of resources when they have to evaluate which assessment methodologies to use (Sturmey, in press-a). An example is given in Figure 1. If there are clear medical or provocative environmental causes of the maladaptive behavior evident upon initial assessment, these could be addressed and modified quickly without resorting to further assessment. If the behavior is too dangerous or infrequent, then analog baselines are not appropriate and nonexperimental methods of functional assessment are indicated (Sturmey, 1995). In some situations the function of the behavior may be readily apparent or may be identified rapidly through observational methods, staff interviews, and questionnaires. Even though questionnaires have generally been shown to be unreliable, it is worth noting that they do provide reliable data in a minority of cases, which may, perhaps, be a valid measure of the function of the behavior (Sigafoos, Kerr, & Roberts, 1994). If these methods are not effective, then the abbreviated 90 rain procedure may yield results in a day in about 50% of cases. If these assessments have been cow ducted and the function of the behavior is not apparent, then analog baselines are indicated. If analog baselines do not yield a function of the behaviol; then the practitioner has the option of continuing to refine experimental hypotheses, conducting further experimental studies of the behavior, or using Empirically Derived Consequences (Fisher et al., 1994) or methods of environmental enrichment and contingency management (Vollmer et al., 1994). Practitioners have to decide at which juncture to invest the resources necessary to conduct an analog baseline assessment. Decision-making prostheses such as that shown in Figure 1 may assist them in the rational use of their limited resources. TRANSLATING ASSESSMENT INTO TREATMENT Identifying the single function of the behavior is insufficient to design and implement a treatment program. Other considerations, such as resources, existing interventions, and skills of staff and parents, go into designing the treatment package. Analog baselines have a clear advantage over other assessment methods. Because the experimental conditions approximate to treatment procedures they can often be modified into viable treatments.
Analog Baselines
Any provocative medical or environmental causes?
Yes
279
Treat these.
No Is the behavior too dangerous or too frequent for analog baselines?
Yes
Conduct functional assessment with descriptive data.
No Conduct brief 90-minute, experimental, functional analysis. Are functions apparent? No Conduct full analog baselines. Are functions apparent? No Implement empirically denied consequences procedures. Is an effective treatment desired?
Implement treatment.
I
No Conduct individually tailored, experimental analysis. Is an effective treatment desired? No FIGURE 1. A decision-making prosthesis to guide the selection of different methods of assessing the functions of problem behavior.
Even if analog baselines correctly identify the single function, careful attention must be paid to the natural environment. Specific situations may need to be addressed. If a subject clearly shows escapc-avoidant behavior, analysis of the natural environment may show inappropriate or blatantly provocative use of demands. In this situation, compliance is required from staff or parents rather than a program to train the client to be passive in an inhumane environment. Thus, even when analog baselines correctly identify the function of the
280
P. Sturmey
behavior, the nature of the intervention, if any, requires consideration of many other factors. SUMMARY This article has raised a wide number of criticisms and potential limitations in the use of analog baselines. Some of these issues might be viewed as severely limiting in certain circumstances. However, it should be clearly acknowledged that analog baselines have made a major contribution to the evaluation of maladaptive behaviors. Analog baselines are not an inappropriate methodology; rather, like other methodologies, analog baselines require careful interpretation and analysis. They are appropriate in certain, but not all, circumstances. Validation
Analog baselines may sometimes rapidly and validly identify the functions of a behavior. However, sometimes the results may require careful validation. Future research should attempt to address the integration of analog baseline methods with nonexperimental observational methods of evaluating behaviors. For example, an analog baseline assessment might correctly identify the function of the target behavior, but a scatterplot (Touchette et al., 1985) may be required to assist in identifying the situations that a program must address. Further, an analog baseline may yield information that is rather unrefined. For example, a functional analysis could be conducted on specific aspects of each function. In the case of task avoidance, task difficulty, task rate, type and variety of tasks, reinforcement density, and rate could all be analyzed after a regular analog baseline assessment. Analog baselines could be integrated with naturalistic observations by conducting the analogues in the natural environments, using the actual situations, tasks, and staff that might maintain the behavior. Such procedures might lose some of the experimental control of analog baselines but may gain in ecological validity of their results. Such setting-based functional analyses may also have advantages because they might provide greater incidental opportunities for staff and parent training and might suggest better ideas for treatment programs. There is a tension between experimental and nonexperimental methods; analog baselines lie toward the experimental end of this continuum. Thus, the validity of experimental procedures, such as analog baselines, can be compromised for experimental control. Analog baselines can indicate effective and noneffective treatments (Iwata et al., 1994). However, incremental treatment validity is a key aspect of validity. A key question is whether analog baselines assist in designing more effective treatments than otherwise would have been devised using all the other
Analog Baselines
281
information already available. This is a complex and subtle question that research has not hitherto addressed. Assessment and Treatment Failures Although the limitations of analog baselines have been implicitly acknowledged through the various refinements, there have been few reports of assessment failures. Failures in behavior therapy are a fertile area where future developments can occur (Foa & Emmelkamp, 1983). Sturmey (in press-a) reported that one subject exhibited between 2 and 10 million acts of selfinjury, whereas a variety of functional assessments, including analog baselines, were conducted. Thus, conducting functional assessments can, at times, be very costly to the subject. Iwata et al. (1994) reported that functions could not be identified in approximately 10% of their series, and Derby et al. (1992), using the abbreviated method, reported that functions could not be identified in approximately 50% of cases. Others have reported cases of failure of analog baselines to identify the function(s) of a target behavior and have gone on to use methods such as Empirically Derived Consequences (Fisher et al., 1994) or environmental enrichment and contingency management (Vollmer et al., 1994). Clinical Process and Functional Analysis The process by which an intervention is designed and evaluated is complex and not clearly understood (Sturmey, in press-b). The relationship of analog baselines to this broader process is subtle. For example, there are often hints that the manner in which tasks and materials are selected, and the identification of target behaviors(s), depends upon a complex process of information collection and processing. Further, this clinical process is a social one in which professionals, staff, and parents negotiate priorities, set limits on acceptable practice, and, to a certain extent, develop a shared clinical agenda. The identification of the function of the target behavior and the selection of the method(s) by which this should be done are only two elements in this broader process. There are times when other methods of identifying the functions of the target behavior will be quicker, cheaper, and more acceptable than analog baselines. At other times analog baselines may be worthwhile because they solve a clinical problem that other methods cannot solve, or because the use of analog baselines show a clear demonstration of a serious effort to attempt to solve a difficult problem. However, the correct identification of the function of the target behavior is only one element in the design of the treatment. Future research in this area should focus, not only on the methodologies of functional analysis, but also the broader process of treatment design, of which analog baselines may be one part.
282
P. Smrmey
REFERENCES Ba~r, D. M., Wolf, M. M., & Risley, T. R. (1968). Some current dimensions of applied behavior analysis. Journal of Applied Behavior Analysis, 1, 91-97. Can', E. G. (1994). Emerging themes. Journal of Applied Behavior Analysis, 27, 393-399. Cart, E. G., & Durand, V. M. (1985). Reducing behavior problems through functional communication training. Journal of Applied Behavior Analysis, 18, 111-116. Chapman, S., Fisher, W., Piazza, C. C., & Kurtz, P. E (1993). Functional assessment and treatment of life-threatening drug ingestion in a dually diagnosed youth. Journal of Applied Behavior Analysis, 26, 255-256. Day, H. M., Homer, IL M., & O'Neill, R. E. (1994). Multiple functions of problem behaviors: Assessment and intervention. Journal of Applied Behavior Analysis, 27, 279-289. Derby, K. M., Wackcr, D. P., Peck, S., Sasso, G., DcRaad, A., Berg, W., Asmus, J., & Ulrich, S. (1994). Functional analysis of separate topographies of aberrant behavior. Journal of Applied Behavior Analysis, 27, 267-278. Derby, K. M., Wacker, D. P., Sasso, G., Ste~ge, M., Northup, J., Cigrand, K., & Asmus, J. (1992). Brief functional assessment techniques to evaluate behavior in an outpatient setting: A summary of 79 cases. Journal of Applied Behavior Analysis, 25, 71 3-721. Durand, V. M., & Crimmins, D. B. (1988). The Motivation Assessment Scale (MAS) administration guide. Topeka, KS: Monaco & Associates. Durand, V. M., & Crimmins, D. B. (1988). Identifying the variables maintaining self-injurious behavior. Journal of Autism and Developmental Disabilities, 18, 99-117. Fisher, W., Piazza, C. C., Bowman, L. G., Hagopian, L. P., & Langdon, N. A. (1994). Empirically derived consequences: A data-based method for prescribing treatments for destructive behaviors. Research in Developmental Disabilities, 15, 133-149. Foa, E. B., & Emmelkamp, P. M. G. (1983). Failures in behavior therapy. New York: John Wiley. Gelfend, D. M., & Hartman, D. P. (1975). Child behavior analysis and therapy. New York: Pergamon. Grace, N. C., Kahng, S. W., & Fisher, W. W. (1994). Balancing social acceptability with treatment effectiveness of an intrusive procedure: A case report. Journal of Applied Behavior Analysis, 27, 171-172. Green, C. W., Reid, D. H., White, L. K., Halford, R. C., Brittain, D. P., & Gardner, S. M. (1985). Identifying reinforcers for persons with profound handicaps: Staff opinion versus systematic assessment of preferences. Journal of Applied Behavior Analysis, 21, 31-43. Homer, R. H. (1994). Functional assessment: Contributions and future directions. Journal of Applied Behavior Analysis, 27, 401-404. Iwata, B. A., Dorsey, M. E, Slifer, K. J., Bauman, K. E., & Richman, G. S. (1982). Toward a functional analysis of self-injury. Analysis and Intervention in Developmental Disabilities, 2, 1-20. Iwata, B. A., Pace, G. M., Dorsey, M. E, Zarcone, J. E, VoUmer, T. R., Smith, R. G., Rodgers, T. A., Lerman, D. C., Shore, B. A., Mazaleski, J. L., Goh, H.-L., Cowdery, G. E., Kalsher, M. L, McCosh, K. C., & Willis, K. D. (1994). The functional analysis of self-injury: An experimental-cpidemiological analysis. Journal of Applied Behavior Analysis, 27, 215-240. Iwata, B. A., Vollmer, T. R., & Zarcone, J. R. (1990). The experimental (functional) analysis of behavior disorders: Methodology, applications and limitations. In A. C. Repp & Singh, N. N. (Eds.), Perspectives on the use of aversive and nonaversive interventions for persons with developmental disabilities (pp. 301-330). Sycamore, IL: Sycamore Press. Kobe, E H., Mulick, L A., Rash, T. A., & Martin, J. (1994). Nonambulatory persons with profound mental retardation: Physical, developmental, and behavioral characteristics. Research in Developmental Disabilities, 15, 413-423. Lerman, D. C., & Iwata, B. A. (1993). Descriptive and experimental analysis of variables maintaining self-injurious behavior. Journal of Applied Behavior Analysis, 26, 293-319.
Analog Baselines
283
Mace, E C. (1994). The significance and future of functional analysis methodologies. Journal of Applied BehaviorAnalysis, 27, 385-392. Mazaleski, J. L., Iwata, B. A., Rodgers, T. A., Vollmer, T. R., & Zarcone, J. R. (1994). Protective equipment astreatment for stereotypic handmouthing: Sensory extinction or punishment effects? Journal of Applied BehaviorAnalysis, 27, 345-355. Neef, N. A., & Iwata, B. A. (1994). Current research on functional analysis methodologies: An introduction. Journal of Applied BehaviorAnalysis, 27, 2121-214. Northup, J., Wacker, D. P., Berg, W. K., Kelly, L., Sasso, G., & DeRaad, A. (1994). The treatment of severe behavior problem in a school setting using a technical assistance model. Journal of Applied BehaviorAnalysis, 27, 33-47. O'Neill, R. E., Homer, R. H., Albin, R. W., Storey, K. & Sprague, J. R. (1990). Functionalanalysis of problem behavior:A practical assessment guide. Sycamore, IL: Sycamore Press. Pace, G. M., Ivancic, M. T., & Jefferson, G. (1994). Stimulus fading as treatment for obscenity in a brain damaged adult. Journal of Applied BehaviorAnalysis, 27, 301-305. Repp, A. C. (1994). Commentaries on functional analysis procedures for school-based behavior problems. Journal of Applied Behavior Analysis, 27, 409---411. Repp, A. C., Felce, D., & Barton, L. E. (1988). Basing treatment of stereotypic and self-injurious behaviors on hypotheses of their causes. Journal of Applied BehaviorAnalysis, 21, 281-289. Schaal, D. W., & Hackenberg, T. (1994). Toward a functional analysis of drug treatment for behavior of people with developmental disabilities. American Journal of Mental Retardation, 99, 125-140. Sigafoos, J., Kerr, M., & Roberts, D. (1994). Inter-rater reliability of the Motivation Assessment Scale: Failure to replicate with aggressive behavior. Research in Developmental Disabilities, 15, 333-342. Skinner, B. E (1989). Science and human behavior. New York: Macmillan. Smith, R. G., Iwata, B. A., Vollmer, T. R., & Zarcone, J. R. (1993). Experimental analysis of multiply controlled self-injury. Journal of Applied Behavior Analysis, 26, 183-196. Sprague, J. R., & Homer, R. H. (1992). Covariation within functional response classes: Implications for treatment of severe problem behavior. Journal of Applied BehaviorAnalysis, 25, 735-745. Sturmey, P. (1994). Assessing the functions of aberrant behaviors: A review of psychometric instruments. Journal of Autism and Developmental Disabilities, 24, 293-303. Sturmey, P. (in press-a). Assessing the functions of self-injurious behavior: A case of assessment failure. Journal of Behavior Therapy and Experimental Psychiatry, 25. Sturmey, P. (in press-b). Functionalanalysis in clinicalpsychology. New York: Wiley. Sturmey, P. (1995). Suicidal threats and behavior in a person with developmental disabilities: Effective psychiatric monitoring based upon a functional assessment. Behavioral Interventions and Residential Treatment, 9, 235-245. Sturmey, P., Carlsen, A., Crisp, A. G., & Newton, T. (1988). A functional analysis of multiple aberrant responses: A refinement and extension of Iwata et al.'s, methodology. Journal of Mental Deficiency Research, 32, 31-46. Taylor, J. C., & Carr, E. G. (1992a). Severe problem behaviors related to social interaction: I. Attention seeking and social avoidance. Behavior Modification, 16, 305-335. Taylor, J. C., & Cart, E. G. (1992b). Severe problem behaviors related to social interaction: 1I. A system analysis. BehaviorModification, 16, 336-371. Taylor, J. C., & Romanczyk, R. G. (1994). Generating hypotheses about the function of student behavior by observing teacher behavior. Journal of Applied BehaviorAnalysis, 27, 251-265. Tinbergen, N. (1963). On the aims and methodology of ethology. Zeitschrififur ~erpsychologie, 20, 410-432. Touchette, P. E., MacDonald, R. E, & Langer, S. N. (1985). A scatter plot for identifying stimulus control of problem behavior. Journal of Applied BehaviorAnalysis, 18, 343-35 i. Vollmer, T. R., Iwata, B. A., Duncan, B, A., & Lerman, D. (1993). Extension of multielement functional analysis using reversal-type design. Journal of Developmental and PhysicalDevelopment, 5, 311-325.
284
P Sturmey
Vollmer, T. R., Iwata, B. A., Zarcone, J. R., Smith, R. G., & MaT~leski, J. L. (1993, May). Withinsession patterns of self-injury as indicators of behavioralfunction. Paper presented to the 19th annual convention of the Association for Behavior Analysis-International, Chicago. Vollmer, T R., Marcus, B. A., & LeBlanc, L. (1994). Treatment of self-injury and hand mouthing following inconclusive functional analysis. Journal of Applied BehaviorAnalysis, 27, 331-344. Wacker, D., Berg, W., Cooper, L. J., Derby, K. M., Stcege, M. W., Northup, J. H., & Sasso, G. (1994). The impact of functional analysis methodology on outpatient clinic services. Journal of Applied BehaviorAnalysis, 27, 405-407. Wacker, D., Stcege, M., Noahup, J., Reimers, T., Berg, W., & Sasso, G. (1990). Use of functional analysis and acceptability measures to assess and treat severe behavior problems: An outpatient clinic model. In A. C. Repp & N. N. Singh (F__xls.),Perspectives on the use ofaversive andnonaversive interventionsfor persons with developmental disa~lities (pp. 349-359). Sycamore, IL: Sycamore Press. Wahler, R. G. (1975). Some structural aspects of deviant behavior. Journal of Applied Behavior Analysis, 8, 27-42. Whitaker, S. (1993). The reduction of aggression in people with learning disabilities: A review of psychological methods. British Journal of Clinical Psychology,32, 1-37. Zarcone, J. R., Rodgers, T. A., Iwata, B. A., Rourke, D. A., & Dorsey, M. E (1991). Reliability analysis of the Motivational Assessment Scale. Research in Developmental Disabilities, 12, 349-360. Zlutnick, S., MayviUe, N. J., & Moffat, S. (1975). Modification of seizure disorders: The interruption of behavior chains. Journal of Applied BehoviorAnalysis, 8, 1-13.