RESPONSE-CLASS HIERARCHIES AND RESURGENCE OF SEVERE PROBLEM BEHAVIOR

Transcription

1 The Psychological Record, 24, 54, RESPONSE-CLASS HIERARCHIES AND RESURGENCE OF SEVERE PROBLEM BEHAVIOR GREGORY A. L1EVING, LOUIS P. HAGOPIAN, ETHAN S. LONG, and JULIA O'CONNOR The Kennedy Krieger Institute and Johns Hopkins University School of Medicine Resurgence may be defined generally as the extinctioninduced recurrence of previously learned response patterns. Understanding the conditions under which this phenomenon occurs has theoretical, clinical, and applied implications, particularly with respect to a related area of research on responseclass hierarchies. In the current study, we examined resurgence with 2 participants who exhibited response-class hierarchies consisting of various topographies of severe problem behavior maintained by positive reinforcement. Baseline levels of responding were maintained by brief access to tangible items. In a second condition, reinforcement was produced by an alternative topography of severe problem behavior, and the initial topography was extinguished concurrently. When the reinforcement contingency for alternative behavior was removed, previously reinforced topographies recovered. This resurgence was specific to behavior that recently produced the reinforcer, which suggests that the recovery was not simply extinction-induced variation or emotional responding. The clinical implications of the results are discussed and related to results that have been produced in the laboratory by a variety of methods. The term resurgence has been used to describe the recurrence of previously reinforced operants when another operant is placed on extinction (e.g., Epstein, 1983, 1985; Lieving & Lattal, 23; see also Leitenberg, Rawson, & Bath, 197; Leitenberg, Rawson, & Mulick, 1975; Staddon & Simmelhag, 1971), the recovery of respondents induced by removal of unconditional stimuli (e.g., Epstein & Skinner, 198; Lindblom & Jenkins, 1981), and the recurrence of emergent behavior during periods of nonreinforcement (Wilson & Hayes, 1996). The general resurgence paradigm, as it applies to operant behavior, can be characterized by three phases: (1) a response produces reinforcement, We thank Kristie Arnold for her assistance. Please address correspondence regarding this study to Gregory A. Lieving, PhD, Spectrum of Hope, 1411 Cypress Creek Blvd., Cypress, TX ( drlieving@spectrumofhope.com).

2 622 LlEVING ET AL. (2) that response then is extinguished, while an alternative response is reinforced (serially, as in Epstein, 1983; Lieving & Lattal, 23 Experiment 1, or concurrently as in Epstein, 1985; Leitenberg et ai., 197; Lieving & Lattal, in press, Experiments 2, 3, and 4), and (3) the reinforcement contingency for alternative behavior is removed via conventional extinction. Resurgence is said to occur if the original response recurs when the alternative response is placed on extinction during the third phase. The term resurgence was coined in a study by Epstein and Skinner (198) that demonstrated recovery of autoshaped key pecking with pigeons. In the first condition, key pecks were autoshaped to a moving dot that was followed by food presentations. Key pecks then were eliminated in a second condition by the removal of the respondent contingency between dot movement and food presentations. Food continued to be delivered intermittently, but was uncorrelated with the moving dot. In this second condition, key pecks rapidly extinguished. In the third condition, all food presentations were withheld. Key pecks recovered to high rates, and the effect was termed resurgence. This phenomenon was demonstrated in later work by Epstein (1983, 1985) using operant methods. In the first condition, key pecks were maintained by positive reinforcement. In a second condition, alternative behavior (e.g., wing raises) produced the reinforcer and key pecks were extinguished concurrently (Epstein, 1985) or were extinguished prior to the second condition (Epstein, 1983). In the third condition, all reinforcement was withheld. Resurgence was said to occur when key pecks recovered during this final condition. Similarly, Lieving and Latta I (23) conducted several experiments to examine the resurgence of key pecking with pigeons wherein the alternative class of behavior consisted of treadle presses. Resurgence also has been used to describe effects similar to those found by Epstein (1983, 1985). Leitenberg and colleagues (e.g., Leitenberg et ai., 197, 1975) utilized a procedure similar to that of Epstein's to examine the effects of alternative reinforcement contingencies with rats. Responses on one lever were reinforced in the first condition. In the second condition, responses on a second lever produced reinforcement, while the previously effective lever press was extinguished. In the third condition, all reinforcement was unavailable, and the previously effective lever press response recurred. The initial response, however, was not eliminated prior to the provision of reinforcement for the alternative response as in Epstein (1983). Wilson and Hayes (1996) provided an interesting replication to previous resurgence studies by utilizing a stimulus equivalence procedure. Following stimulus training, subjects were given feedback for relational responding. The emergence of a second equivalence class then was arranged for, and when all feedback was removed, the older class of responses reemerged. Resurgence therefore has been demonstrated with humans exhibiting relatively complex behavioral patterns that are neither clearly operant nor respondent. That is, relational responding "emerged" without explicit training

3 HIERARCHIES AND RESURGENCE 623 or reinforcement for doing so and yet recovered during subsequent conditions in which feedback was unavailable. Although the procedures used in the laboratory to investigate resurgence have varied, all of the methods have two features in common. First, the initial response (i.e., the response that eventually resurges) is reduced to zero or near-zero rates. Second, resurgence is induced by a terminal extinction condition in which programmed consequences for alternative behavior are unavailable. Recent clinical studies that have examined response-class hierarchies have demonstrated that placing one response on extinction can increase the rate of other responses (e.n., Harding et ai., 21; Richman, Wacker, Asmus, Casey, & Andelman, 1999). Response-class hierarchies involve an operant class comprised of responses that occur typically in ordered sequences from low to high severity. When reinforcement is available for the emission of a variety of different topographies of problem behavior, low-severity responses often are the only responses to occur. Hierarchies then can be revealed through extinction analyses, during which mild or low-severity responses are extinguished, and more severe topographies occur and produce reinforcement. These more severe topographies have occurred historically and often are the responses that precipitated the search for behavioral treatment. It is not clear, however, tile extent to which these increases in response rates of other behavior are instances of resurgence. There are a variety of behavioral mechanisms that potentially account for increases in the rate of one response when the rate of another decreases because of nonreinforcement, includiing generalized matching (for a review, see Davison & McCarthy, 1988), behavioral contrast (e.g., Williams & Wixted, 1986), extinction-induced variability (e.g., Antonitis, 1951; Eckerman & Lanson, 1969), and extinction-induced aggression (e.g., Azrin, Hutchinson, & Hake, 1966), to name a few. Regardless of the potential behavioral mechanisms of responseclass hierarchies, the clinical importance both of response-class hierarchies and resurgence can not be understated. Perhaps the most relevant milieu in which the clinical importance of resurgence and response-class hierarchies can be appreciated is clinical relapse. After behavioral treatments have been developed for severe behavior problems, treatment integrity may become delraded unintentionally by parents, teachers, or other caregivers. Specifically, the integrity of treatments that depend on the use of reinforcement (and most behavioral treatments do) relies on caregiver compliance with the treatment's protocols, such as a particular schedule of minforcer delivery with a particular stimulus or set of stimuli. When care!ivers do not comply, for whatever reason, reinforcers are absent, delayed, or their delivery is otherwise altered. Extinction may be a common form of treatment integrity degradation and can evoke the recurrence (as in resurgence) of previously reinforced classes of behavior or can evoke other members of a response class (as in response-class hierarchies), including problem

4 624 LlEVING ET AL. behavior that has not occurred for relatively long periods of time. This may lead to inadvertent reinforcement of these responses and eventually to a complete breakdown in treatment gains. Although the phenomenon of resurgence has received limited empirical attention in both basic and applied behavior analysis, the conditions under which it occurs in the laboratory may resemble situations encountered in clinical settings and has been recognized conceptually as a potentially important mechanism for behavioral variation (see, for example, Epstein, 1991; Shahan & Chase, 22). The purpose of this study was to demonstrate the resurgence effect within response-class hierarchies, in the course of behavioral assessment of severe problem behavior, in a manner consistent with previous laboratory research. In so doing, we attempted both to reproduce the general effect with humans as in Wilson and Hayes (1996) and to examine the generality of the effect to nonlaboratory conditions and clinical populations. Method Subjects Christine, a 7-year-old female diagnosed with moderate to severe mental retardation and autism, was admitted to an outpatient program for severe behavior problems. She engaged in disruption (banging, throwing, and kicking objects), aggression (hitting, kicking, pushing, and chin pressing), and self-injury (head banging, self-biting, body hitting against walls). Christine was nonverbal, ambulatory, and typically communicated with gestures. Sam, a 9-year-old male diagnosed with mild mental retardation, mood disorder, oppositional defiant disorder and attention deficit hyperactivity disorder, was admitted to an inpatient program for severe behavior problems. He engaged in disruption (banging, throwing, and kicking objects), dangerous acts (standard institution-wide definition; standing on furniture, touching electrical outlets, throwing objects to ceiling, and turning over heavy objects), inappropriate language (cursing), and aggression (hitting, kicking, pushing, grabbing and throwing objects within 2 ft of a person). Sam was verbal and ambulatory. Apparatus All sessions were conducted in a padded room behind a one-way mirror. During sessions, one therapist was present in the session room with the client and delivered reinforcement. At least one trained observer recorded problem behavior on laptop computers and was located outside of the room and observed the sessions through the one-way mirror. Two independent observers collected data during 57% of the sessions (only one observer collected data during the remaining sessions). Percentage of interobserver agreement was calculated in the following manner: (# agreements / (# agreements + # disagreements)) X 1. An agreement was defined as both observers recording the exact same number of

5 HIERARCHIES AND RESURGENCE 625 responses during a given 1-s interval. Mean exact agreement of the occurrence of problem behavior for Christine was 94.4% (range of 83.6% to 1%) and for Sam was 92.8% (range of 88.75% to 97.8%). Procedure For each client, reinforcer assessments first were conducted, followed by extinction analyses. Christine and Sam were admitted to the outpatient and inpatient units, respectively, for the treatment of severe aggression, among other concerns. Aggression, however, proved difficult to assess and treat for these two clients because the behavior did not occur during the standard assessment procedures. The clinical determination (made on the basis of direct observation and parent interviews) was that aggression was part of a response class, and perhaps a member of a hierarchy of responses that had produced reinforcement in the natural environment. It was decided, therefore, to conduct extinction analyses both to investigate this clinical hypothesis and to provide an environment in which aggression could occur and thus be subject for treatment development. These extinction analyses were conducted separately and individually; they were not a part of a research protocol, but were simply similar methods used in the normal course of behavioral assessment and treatment development. The methods used were similar enough to warrant inclusion together in the present paper. Reinforcer assessments. Functional analyses were conducted using conditions similar to those described by Iwata, Dorsey, Slifer, Bauman, and Richman (1982/1994). In addition, tangible conditions were included in the functional analysis. Prior to each session clients gained access to a preferred item for 2 min. At the onset of each session, the therapist removed the preferred item and returned the item for 3 s contingent on the occurrence of problem behavior on a fixed-ratio (FR) 1 schedule. Results of the functional analysis suggested that both Christine's and Sam's problem behavior were maintained by positive reinforcement in the form of access to preferred items, although the functional analyses were not sufficient to evoke all forms of problem behavior. All sessions during this and subsequent phases were 1 min in duration, and reinforcement consisted of 3-s access to a preferred item. Extinction analysis (Christine). The extinction analysis was conducted following the functional analysis. During the initiial phase, reinforcement was available on FR 1 schedules for all topographies (FR 1 ALL). During the second phase, disruption was placed on extinction (EXT DIS). The third phase consisted of placing both disruption and aggression on extinction (EXT DIS+AGG). The initial phase (FR 1 ALL) was reimplemented in the final phase. Response rates for Christine were adjusted for reinforcement time, because she neither responded consistently in bursts during reinforcement conditions, nor emitted responses during the reinforcement cycle. The nominal maximum response rate therefore was two responses per min. For some sessions, however, responses rates were slightly

6 626 LlEVING ET AL. higher due to the occasional response "burst" (i.e., a quick succession of responses that were recorded on the laptop before recording the onset of the reinforcer key). Extinction analysis (Sam). The extinction analysis was conducted following the functional analysis. In the initial phase, reinforcement was available on FR 1 schedules for all topographies (FR 1 ALL). In the second phase, disruption was placed on extinction (EXT DIS). In the third phase, dangerous acts were placed on extinction and disruption continued on extinction (EXT DIS+DA). In the fourth phase, cursing was placed on extinction, while disruption and dangerous acts continued on extinction (EXT DIS+DA+CURSE). For Sam, dangerous acts consisted of climbing on top of ledges, sills, and furniture, and also diving under beds, where his head frequently came into sharp contact with the frame, wall, or floor. During experimental sessions, dangerous acts themselves were not blocked unless Sam was judged to be in danger of falling. When he engaged in climbing, the therapist moved into a position to block or catch him from falling if it occurred. Response rates for Sam were not adjusted for reinforcement time, as he tended to respond consistently during the reinforcement cycle. The amount of behavior emitted during reinforcement was thus of clinical significance, and these responses were important to include in the analysis of his problem behavior. Response rates for Sam therefore occasionally varied well above two responses per min during conditions that arranged for reinforcement. Results The results of the extinction analysis for Christine are displayed in Figure 1. Rates of aggression and self-injury were zero during the first phase and disruption occurred at approximately two responses per min. Rates of self-injury remained at zero throughout the extinction analysis except for one response in the first minute of Session 4, although Christine had a long history of emitting that topography. Although reinforcement was available for each topography, only disruption occurred. In the second phase extinction decreased disruption to zero rate and aggression was established. When both disruption and aggression were placed on extinction in the third phase, disruption recurred at high rates and remained above zero for three sessions (resurgence - Sessions 8-1). In addition, rates of aggression increased during the three sessions of th is phase. Reinforcement again was available for all topographies in the fourth phase, and disruption and aggression returned to initial baseline levels (i.e. disruption to approximately two responses per min, and aggression to zero). The results of the extinction analysis for Sam are displayed in Figure 2. Rates of dangerous acts, cursing, and aggression were zero during the first phase and disruption occurred at approximately three per min. When disruption was placed on extinction in the second phase, disruption

7 HIERARCHIES AND RESURGENCE 627 CII 6 -:::l 5... CII 4 FR 1 ALL EXT DIS I EXT DIS I FR 1 +AGG I ALL a. III 3 I li 2 :::l : iii 1 i5 CII 6 -:::l... CII a. III iii III Cl 1 Cl < CII 6 :; 5... CII 4 a. III CII 3.;:: :::l :E 2 -(jj 1 en : / I : Resurgence : I I I Christine I!= I Sessions Figure 1. Responses per minute across all phases of the study for Christine. Rates of disruption, aggression, and self-injury are displayed in the top, middle, and bottom panels, respectively. The occurrence of resurgence is denoted by an arrow. decreased to zero and dangerous acts increased up to approximately six responses per min. Cursing and aggression remained at zero rates during the second phase. When dangerous acts were placed on extinction in the third phase, three effects occurred. First, dlisruption recurred then extinguished (resurgence - Sessions 7-11). Second, dangerous acts decreased to zero rates. Third, cursing occurred and was maintained at approximately 1 responses per min. Aggression remained at zero rates

8 628 LlEVING ET AL. 4 FR 1 ALL EXT DIS, EXT DIS+DA EXT DIS+DA +CURSE Q) c. 2.2 a. :::> iii C Q) :; iii c. U) U «U) :::> iii Cl t1i C :, Resurgence \ : / \ '---. '... : \!=:-: :6-;,, ' I ', :J:; :1 ", ", / Resurgence : : \ : Q) :; 1 iii c. U) Q) U) 5 :; 15 Q) :; iii 1 c. U) o. 5 Cl Cl «Sessions / :/, Figure 2. Responses per minute across all phases of the study for Sam. Rates of disruption and dangerous acts are displayed in the top two panels. Rates of cursing and aggression are displayed in the bottom two panels. Note the difference in scale for the bottom two panels. To the preserve the scale for rates of dangerous acts, the third data point of the second condition (EXT DIS) is not shown and the arrow indicates a rate of six responses per minute for that session. The occurrences of resurgence are denoted by arrows. 14

9 _. HIERARCHIES AND RESURGENCE 629 during all but Session 8 of this phase. During the final phase when reinforcement was unavailable for all topographies except aggression, three effects again occurred. First, both disruption and dangerous acts recurred and rates increased slightly above zero for two sessions before returning to zero (resurgence-sessions 12-14). Second, cursing decreased to zero rates. Third, aggression occurred and was maintained. Within-session response patterns were examined by generating cumulative records from the data displayed in Figures 1 and 2. The cumulative record data for Christine are displayed in Figure 3 and consist en en Q) 1 FR 1 ALL EXT DIS EXT DI.: FR 1 ALL +AGG: C - 75 en Disruption Q) Aggression Self Injury Session Onset Q) a::: 5... > CO 25 :::J E :::J _.._J_. _..-..L":_J_._I.l-. J _ _. i.risline i i i min Bins Figure 3. Christine's cumulative responses per 1-min bins for all sessions of the study. Disruption, aggression, and self-injury are denoted by solid, dotted, and dashed lines, respectively. Vertical tick marks denote the onset of a new session. of cumulative responses per 1-min bin. In the first phase a steady rate of disruption occurred. During the second phase in which disruption was extinguished, rates of disruption increased briefly. Aggression then occurred and was maintained by reinforcement as disruption approached a rate of zero. Disruption recovered during the third phase, during which aggression was extinguished, as aggression continued at high rates and self-injury remained at zero. The cumulative record data for Sam are displayed in Figure 4 and consist of cumulative responses per 1-s bin. For the sake of clarity, two topographies of problem behavior (disruption and dangerous acts) are depicted in the top panel, and the other two topographies (cursing and aggression) are displayed in the bottom panel. In the first phase a fairly steady rate of disruption occurred. During the second phase in which _

10 63 LlEVING ET AL. en 1 FR 1 ALL EXT DIS EXT DIS+DA EXT DlS+DA 8 6 +CURSE Disruption Dangerous Acts Q) Session Onset en 4 C C- 2 en Q) a: Q) 35 > +-' 28 cts E Cursing Aggression Session Onset J. : s Bins Figure 4. Sam's cumulative responses per 1-s bins for all sessions of the study. In the top panel, disruption and dangerous acts are denoted by solid and dotted lines, respectively. In the bottom panel, cursing, and aggression are denoted by solid and dotted lines. Note the difference in scale between the top and bottom panels. Vertical tick marks denote the onset of a new session.

11 HIERARCHIES AND RESURGENCE 631 disruption was extinguished, brief increases in rates of disruption occurred followed by the occurrence and maint,enance of dangerous acts (particularly after disruption had reached zero rates). Rates of cursing and aggression remained at zero during the second phase. Disruption recovered during the third phase, during which dangerous acts were extinguished. When both disruption and dangerous acts reached zero or near-zero rates, cursing occurred and was maintained at high rates. A few aggressive responses occurred prior to the occurrence of cursing, but were not maintained. During the fourth phase, in which disruption and dangerous acts continued on extinction and cursing was placed on extinction, disruption again recovered briefly, as well as dangerous acts. As cursing approached zero rates, aggression occurred and was maintained. In summary, one instance of resurgence occurred for Christine, and three instances of resurgence occurred for Sam. Following a baseline in which Christine's disruption was reinforced, disruption was extinguished while her aggression produced reinforcement. When aggression then was extinguished, disruption resurged. Following a baseline in which Sam's disruption was reinforced disruption was extinguished while dangerous acts were reinforced. When dangerous acts then were extinguished, disruption resurged. Following a baseline in which cursing was reinforced both disruption and dangerous acts resurged. Interestingly, Christine's self-injury and Sam's aggression (that had long histories of occurring) were not increased reliably by extinction, although Sam engaged in aggression during the third and final extinction condition. Discussion The results demonstrate the resurgence effect within response-class hierarchies in the course of behavioral assessment with 2 subjects from clinical populations. The findings themfore extend previous demonstrations of the resurgence effect to nonlaboratory settings, clinical populations, and replicate previous findings w iith human subjects, as in Wilson and Hayes (1996). For both subjects, thel removal of reinforcement contingencies for one topography of problem behavior induced recovery of recently reinforced topographies. The present study, however, utilized a clinical procedure that deviated from the procedure used by Epstein (1985), and is more analogous to the procedures used by Leitenberg et al. (197, 1975), and the procedure described by Epstein (1983). Because previous topographies were not extinguished explicitly prior to reinforcement of alternative behavior, it therefore is conceivable that the recovery shown in the current study perhaps is a function of a different behavioral mechanism than extinction-induced resurgence. Epstein (1985) employed a control condition prior to the condition in which an alternative response was reinforced. In that control condition, the original response (key pecking) was extinguished for one or more sessions, to ensure that the response was eliminated prior to the provision of reinforcement for alternative bbhavior. Any subsequent

12 632 LlEVING ET AL. recovery therefore could not simply be an artifact of the response's failure to contact extinction during the condition in which reinforcement was available for the alternative response. In the current study, responses were not extinguished explicitly prior to the imposition of reinforcement contingencies for alternative behavior. The results (particularly the withinsession patterns of responding) suggest, however, that previous topographies did in fact contact extinction, and it was this behavioral contact with extinction that induced both previously reinforced topographies and the next topography in the hierarchy, which then contacted reinforcement and was maintained. It is doubtful, therefore, that the recovery shown in the present study is caused by a latent extinction effect, as shown by Leitenberg et al. (197, 1975). Other potential mechanisms for the results include extinction-induced variability and "emotional" responding. The temporary increases in response rates that we describe as resurgence in the present study do represent a type of extinction-induced variability, but not as that term usually is invoked. The recovery does not appear to be due simply to increases in overall behavioral variability, or overall activity level. The recovery also does not appear to be a function of emotional responding. For both subjects, extinction did not produce increases for all topographies of problem behavior. For Christine, self-injurious behavior did not increase as a result of extinction, even though this was a response that she had engaged in frequently, and produced reinforcement for, historically. This topography, however, had not produced reinforcement relatively recently. For Sam, aggression similarly did not increase above zero until the final extinction phase, even though this was a commonly emitted response for him (Sam's severe aggression, in fact, was the reason he was admitted for inpatient treatment). It is therefore doubtful that the resurgence effects shown simply are due to increases in general activity level, behavioral variation, frustration, or other emotional factors. Only responses that had recently produced reinforcement (and had been "extinguished") recovered when a topography was placed on extinction. The resurgence effect in general has several implications for both research and clinical practice, as noted elsewhere (see, for example, Epstein, 1983; Lerman & Iwata, 1996; Shahan & Chase, 22; Wilson & Hayes, 1996). In the treatment of severe problem behavior, resurgence may play an important role in the ultimate success of interventions that are based on reinforcement for alternative behavior, which make up the majority of behavioral treatments for clients with severe problem behavior. After behavioral treatments have been developed that rely on the maintenance of alternative, socially redeemable responses, extinction may be encountered. This extinction may cause old patterns of problem behavior to resurge temporarily, which may lead to the complete loss of treatment gains if not dealt with properly. Furthermore, for those clients who display response-class hierarchies, there is a preexisting set of circumstances that may be fruitful for further examinations of resurgence and other effects of response

13 HIERARCHIES AND RESURGENCE 633 recovery. In the laboratory, specific learning histories are established such that subjects have more than one response in their repertoire that has produced a specific reinforcing stimulus. In the case of response-class hierarchies, this history is already present, and provides an ideal situation to examine response recovery in general and resurgence in particular with clinical populations. Such research may ultimately lead to strategies that prevent relapse of problem behavior and improve the long-term treatment gains wrought by behavioral interventions and treatments. References ANTON ITIS, J. J. (1951). Response variability in the white rat during conditioning, extinction, and reconditioning. Journal of Experimental Psychology, 42, AZRIN, N. H., HUTCHINSON, R. R., & HAKE, D. F. (1966). Extinction-induced aggression. Journal of the Experimental Analysis of Behavior, 9, DAVISON, M., & MCCARTHY, D. (1988). The matching law: A research review. Hillsdale, NJ: Erlbaum. ECKERMAN, D. A., & LANSON, R. N. (1969). Variability of response location for pigeons responding under continuous reinforcement, intermittent reinforcement, and extinction. Journal of the Experimental Analysis of Behavior, 12, EPSTEIN, R. (1983). Resurgence of previously reinforced behavior during extinction. Behaviour Analysis Letters, 3, EPSTEIN, R. (1985). Extinction-induced resurgence: Preliminary investigations and possible applications. The Psychological Record, 35, EPSTEIN, R. (1991). Skinner, creativity, and the problem of spontaneous behavior. Psychological Science, 2, EPSTEIN, R., & SKINNER, B. F. (198). Resurgence of responding during the cessation of response-independent reinforcement. Proceedings of the National Academy of Sciences USA, 77, 625' HARDING, J. w., WACKER, D. P., BERG, W. K., BARRETTO, A., WINBORN, L., & GARDNER, A. (21). Analysis of response class hierarchies with attention-maintained problem behaviors. Journal of Applied Behavior Analysis, 34, IWATA, B. A., DORSEY, M. F., SLIFER, K. J., BAUMAN, K. E., & RICHMAN, G. S. (1994). Toward a functional analysis of self-injury. Journal of Applied Behavior Analysis, 27, (Reprinted from Analysis and Intervention in Developmental Disabilities, 2, 3-2, 1982). LEITENBERG, H., RAWSON, R. A., & BATH, f<. (197). Reinforcement of competing behavior during extinction. SciencE, 169, LEITENBERG, H., RAWSON, R. A., & MULlCK, J. A. (1975). Extinction and reinforcement alternative behavior. Journal of Comparative and Physiological Psychology, 88, LERMAN, D. C., & IWATA, B. A. (1996). Developin a technology for the use of operant extinction in clinical settings: An examination of basic and applied research. Journal of Applied Behavior Analysis, 29,

14 634 LlEVING ET AL. LlEVING, G. A., & LATTAL, K. A. (23). Recency, repeatability, and reinforcer retrenchment: An experimental analysis of resurgence. Journal of the Experimental Analysis of Behavior, 8, LINDBLOM, L. L., & JENKINS, H. M. (1981). Responses eliminated by noncontingent or negatively contingent reinforcement recover in extinction. Journal of Experimental Psychology: Animal Behavior Processes, 7, RICHMAN, D. M., WACKER, D. P, ASMUS, J. M., CASEY, S. D., & ANDELMAN, M. (1999). Further analysis of problem behavior in response class hierarchies. Journal of Applied Behavior Analysis, 32, SHAHAN, T. A., & CHASE, P. N. (22). Novelty, stimulus control, and operant variability. The Behavior Analyst, 25, STADDON, J. E. R., & SIMMELHAG, V. L. (1971). The 'superstition' experiment: A reexamination of its implications for the principles of adaptive behavior. Psychological Review, 78, WILLIAMS, B. A., & WIXTED, J. T. (1986). An equation for behavioral contrast. Journal of the Experimental Analysis of Behavior, 45, WILSON, K. G., & HAYES, S. C. (1996). Resurgence of derived stimulus relations. Journal of the Experimental Analysis of Behavior, 66,