A LABORATORY INVESTIGATION OF THE EFFECTS OF TREATMENT INTEGRITY FAILURES ON DIFFERENT IAL REINFORCEMENT PROCEDURES By CLAIRE CATHLEEN ST. PETER PIPKIN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006
Copyright 2006 by Claire Cathleen St. Peter Pipkin
iii ACKNOWLEDGMENTS I would like express my sincere apprecia tion to my mentors throughout graduate school, Drs. Timothy Vollmer and Henry Pennypacker, Jr., for all of their help and support throughout this process. I would also like to thank the memb ers of my committee, Drs. Lise Abrams, Stephen Boggs, Jesse Dall ery, and Brian Iwata, for their comments, and my brother-in-law, David Pipkin, for hi s help with programming. Finally, I would like to thank my husband, Michael Pipkin, and my parents, Norman and Donna St. Peter, for their unwavering love, support, and encouragement.
iv TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iii LIST OF TABLES............................................................................................................vii LIST OF FIGURES.........................................................................................................viii ABSTRACT....................................................................................................................... ..x CHAPTER 1 INTRODUCTION........................................................................................................1 Overview.......................................................................................................................1 Treatment Integrity and Behavioral Treatments...........................................................3 Effects of Treatment Integrity on Problem Behavior...................................................4 Purpose........................................................................................................................ .8 2 GENERAL METHODS FOR EXPE RIMENTS I THROUGH IV............................13 Participants and Setting..............................................................................................13 Method........................................................................................................................13 3 EFFECTS OF SIGNALED DE CREASES IN INTEGRITY.....................................17 Procedure....................................................................................................................17 Results and Discussion...............................................................................................19 4 EFFECTS OF UNSIGNALED DECREASES IN INTEGRITY...............................27 Procedure....................................................................................................................27 Results and Discussion...............................................................................................28 5 EFFECTS OF MULTIPLE EXEMPLARS ON RESPONDING DURING ERRORS OF COMMISSION....................................................................................34 Procedure....................................................................................................................34 Results and Discussion...............................................................................................35
v 6 EFFECTS OF INTEGRITY FADING ON RESPONDING DURING ERRORS OF COMMISSION.....................................................................................................40 Procedure....................................................................................................................40 Results and Discussion...............................................................................................41 7 GENERAL METHODS FOR EX PERIMENTS V AND VI.....................................46 Participants and Setting..............................................................................................46 Method........................................................................................................................46 8 EFFECTS OF TREATMENT INTEGRITY FAILURES ON DRA TREATMENTS..........................................................................................................50 Procedure....................................................................................................................50 Results and Discussion...............................................................................................51 9 EFFECTS OF CONDITION SEQUENCE ON RESPONDING DURING DRA TREATMENT INTEGRITY FAILURES : HUMAN OPERANT WITH BRIEF EXPOSURES.............................................................................................................59 Procedure....................................................................................................................59 Results and Discussion...............................................................................................60 10 EFFECTS OF CONDITION SEQUENCE ON RESPONDING DURING DRA TREATMENT INTEGRITY FAIL URES: HUMAN OPERANT WITH EXTENDED EXPOSURES.......................................................................................65 Purpose.......................................................................................................................65 Participants and Setting..............................................................................................65 Method........................................................................................................................66 Results and Discussion...............................................................................................68 11 EFFECTS OF CONDITION SEQUENCE ON RESPONDING DURING DRA TREATMENT INTEGRITY FAILUR ES: APPLIED REPLICATION....................73 Purpose.......................................................................................................................73 Methods......................................................................................................................73 Participant and Setting.........................................................................................73 Data Collection and Interobserver Agreement....................................................73 Functional Analysis.............................................................................................75 Baseline...............................................................................................................76 Differential Reinforcement..................................................................................76 Treatment Integrity Failure (50/50).....................................................................77 Results and Discussion...............................................................................................77 12 GENERAL DISCUSSION.........................................................................................81
vi LIST OF REFERENCES...................................................................................................91 BIOGRAPHICAL SKETCH.............................................................................................95
vii LIST OF TABLES Table page 1 Error level groups during Experiment I...................................................................17 2 Order of conditions during Experiment I.................................................................18 3 Order of conditions during Experiment III..............................................................35 4 Order of conditions during Experiment IV..............................................................40
viii LIST OF FIGURES Figure page 1: Screen shot of computer program used for Experiments I-IV......................................16 2: Participants exposed to 90% a nd 80% integrity with signaled errors...........................25 3: Participants exposed to 70% a nd 60% integrity with signaled errors...........................26 4: Mean rates of responding for participants with signaled errors....................................26 5: Participants disqualified from furthe r analysis. Both participants experienced signaled changes between FI and DRO....................................................................29 6: Participants who experienced no changes in background color...................................30 7: Participants who experienced change s in background only between FI and DRO......31 8: Participants for whom multiple exem plars were sufficient to reduce responding........36 9: Participants for whom multiple exempl ars were insufficient to suppress responding.38 10: Participants whose response rate incr eased during the first signaled exposure to 80% integrity............................................................................................................44 11: Participants who did not respond dur ing the initial exposur e to 80% integrity..........45 12: Results for Group I, who were exposed to errors of omission only...........................52 13: Results for Group II, who experi enced only errors of commission............................54 14: Results for Group III, who experienced both errors of commission and errors of omission...................................................................................................................55 15: Participants showing carr yover from the previous phase...........................................61 16: Participants showing some sw itching from the previous phase.................................64 17: Session-by-session response rates for all 3 participants in Experiment VII...............71
ix 18: Last five baseline or DRA sessions in a phase and first five 50/50 sessions in a phase.........................................................................................................................72 19: Functional analysis results for Jake............................................................................78 20: Treatment analysis for Jake........................................................................................80
x Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy A LABORATORY INVESTIGATION OF THE EFFECTS OF TREATMENT INTEGRITY FAILURES ON COMMON BEHAVIORAL TREATMENTS By Claire Cathleen St. Peter Pipkin August 2006 Chair: Timothy R. Vollmer Major Department: Psychology Differential reinforcement procedures ar e frequently used as a treatment for problem behavior. Two common differential re inforcement procedures involve providing reinforcers following a period of time in whic h no instances of unde sired behavior occur (known as differential reinforcement of other behavior, or DRO) or providing reinforcers following an appropriate alternative behavi or but not following undesired behavior (known as differential reinforcement of alternative behavior, or DRA). Although differential reinforcement procedures are re latively straightforward, they may not be consistently implemented as designed, partic ularly by caregivers w ith little training. The degree of appropriate treatment implementation is often called â€œtreatment integrity.â€ The following experiments evaluate the effects of treatment integrity failures on DRO and DRA treatments. Because of the potential da nger associated with providing at-risk participants with rela tively low-quality, long-term care, the first seven experiments were conducted with non-clinical populations on an analog task. Experiments I and II
xi examined the effects of signaled and unsi gnaled failures to appropriately deliver a reinforcer (errors of omission) and of in correct reinforcer deliveries (errors of commission) on a DRO treatment. In both st udies, errors of commission were more detrimental to the treatment than were e rrors of omission. Experiments III and IV examined means of attenuating the detrim ental effects of errors of commission. Experiments V through VIII examined the effects of treatment integrity failures on DRA treatments. Experiment V evaluated different levels and types of integrity failures on a DRA treatment and found that errors were most detrimental when some mixture of errors of omission and commission was in place. E xperiments VI and VII evaluated sequence effects during integrity failures phases, using brief (Experiment VI) or extended (Experiment VII) exposures. Experiment VIII attempted to replicate the findings of Experiments VI and VII in a school setti ng with a student who was referred for the assessment and treatment of aggression. The results of Experiments VI through VIII demonstrated that condition sequence infl uences responding during integrity failures: failures were generally more detrimental when they followed baseline than when they followed full treatment phases. Implications of the results for research and clinical practice are discussed.
1 CHAPTER 1 INTRODUCTION Overview Appropriate and effective implementati on of behavior-reduction procedures by caregivers is often a goal of behavioral treatment. Despite cliniciansâ€™ efforts to train caregivers to exacting standards, caregiver implementation of treatments may either begin relatively low (Wickstrom, Jones, La Fleur, & Witt, 1998) or decrease over time (Greene, Norman, Searle, Daniels, & Lubec k, 1995; Witt, Noell, LaFleur, & Mortenson, 1997). For example, Wickstrom et al. observe d teachers implementing several different kinds of behavioral interventions, and showed that the teachers delivered the programmed consequence for target behavior an average of 4% of the time. The degree to which a treatment is imp lemented as designed has been called treatment integrity (Peterson, Homer, & Wonderlich, 1982; Gresham, 1989), treatment fidelity, or procedural reliability (Gresham , 2005). Treatment integrity is important for research purposes because causal relationshi ps between treatment implementation and changes in problem behavior are not conc lusive when treatments are implemented inconsistently. In other words, without hi gh levels of treatment integrity it becomes impossible to determine if the changes in be havior were due to the treatment as it was designed. Treatment integrity is also an important issu e for clinical practice, as higher levels of treatment integrity may be associated with better treatment outcomes (Yeaton & Sechrest, 1981). Although delete rious effects of treatment in tegrity have been shown in
2 many studies, effect sizes are generally only moderate. For example, Gresham and colleagues (1993) reviewed studi es reporting treatment integr ity of behaviorally-based interventions in the school psychology literature and f ound only a moderate relation between the level of treatment integrity a nd the outcome of the intervention. Other research has found similar re lations (Noell et al., 2005). It is possible that low integrity on certain non-critical components of the treatment may result in a low integrity score while the effectiveness of the treatment is retained. Therefore, overall scores of treatment in tegrity may not be truly representative of treatment implementation, because integr ity may actually be low only on certain components of the treatment. Variation in treatment impl ementation across components could occur because of the time or effort re quired to implement certain components, or because different skill sets are required to appropriately implement various intervention components. For example, Codding, Feinber g, Dunn, and Pace (2005) demonstrated that teachers initially implemented antecedent-b ased intervention components at a much higher level of integrity than they impl emented components involving manipulation of consequences. Integrity was also lower on interventions that involved multiple components or procedures that were not t ypically part of the teachersâ€™ routines. The particular procedures used in a tr eatment may influence the degree to which caregivers implement the treatment as designe d. For example, it is possible that some treatment procedures are more difficult to im plement than others, and thus may be more prone to integrity failures. Higher levels of treatment integrity are more frequently associated with reinforcement-based treatments (Tingstrom, 1989; Witt, Martens, & Elliot, 1984) and treatments for which followup or consultation with professionals is
3 available (Codding, Feinberg, Dunn, & Pace, 2005; Noell et al., 2000; Northup, George, Jones, Broussard, & Vollmer, 1996; Sterli ng-Turner, 2002; Sterling-Turner, Watson, & Moore, 2002; Sterling-Turner, Watson, Wild mon, Watkins, & Lit tle, 2001; Taylor & Miller, 1997; Witt et al., 1997). Treatment Integrity and Behavioral Treatments Differential reinforcement procedures are commonly used as a treatment for problem behavior. Although several types of differential reinforcement procedures have been developed, two frequently used treatment procedures are differential reinforcement of other behavior (DRO) and differential rein forcement of alternative behavior (DRA). The DRO procedures typically involve the deliv ery of a reinforcer following a period of time in which no undesired (target) responses occur. For example, if screaming was reinforced by access to attention, a DRO tr eatment might involve providing attention following each 5-min period during which th e child did not scr eam. Appropriately implementing some DRO procedures involves c onstant monitoring of the client (to reset the interval when the target behavior occu rs) and frequent delivery of reinforcers. Therefore, this treatment could become labor-i ntensive and yield a high rate of errors. Differential reinforcement of alternat ive behavior (DRA) typically involves withholding reinforcers following problem be havior (extinction, EXT) and providing reinforcers contingent on some appropriate, alternative response . Using the screaming example described above, a DRA treatment might involve ignoring screaming and providing attention following some appropriate behavior, such as sa ying hello. This type of treatment may be implemented with rela tively low levels of integrity by some caregivers because those individuals may ha ve a long history of reinforcing problem
4 behavior and may therefore find EXT procedur es difficult, or because caregivers may have difficulty providing c onsistent reinforcement follo wing alternative responses. The effects of treatment integrity failure s on differential reinforcement procedures have not been carefully examined. Most prio r research on treatment integrity has focused on whether or not integrity wa s reported in experimental studies (Armstrong, Ehrhardt, Cool, & Poling, 1997; Gresham, Gansle, & Noel l, 1993; Gresham, Gansle, Noell, Cohen, & Rosenblum, 1993; Gresham, MacMillan, Beebe-Frankenberger, & Bocian, 2000; Peterson, Homer, & Wonderlich, 1982), factors in fluencing integrity (Hargett & Webster, 1996; Northup et al., 1996; Sasso et al., 1992; Sterling-Turner, 2002; Sterling-Turner et al., 2001), or recommendations to increas e integrity (Gresham, 1989; Noell, Duhon, Gatti, & Connell, 2002; Noell et al., 2000; W itt et al., 1997). Research focusing on the effects of treatment integrity as an independe nt variable have primarily used procedures to develop academic performance (Noell et al., 2002) or social skills (Peterson & McConnell, 1996; Peterson et al., 1982), not response-reduction pr ocedures like those described above. Because of this relative pauc ity of research focusing on the effects of low levels of treatment integrity, the level of integrity adequate fo r achieving intervention effects during differential reinforcemen t procedures remains largely unknown. Effects of Treatment Integrity on Problem Behavior Two (of many) possible types of treatment integrity failures involve either the failure to deliver an earned reinforcer accordi ng to the treatment schedule (here termed an error of omission) or the delivery of a re inforcer following problem behavior (here termed an error of commission). An error of omission would occu r when the caregiver failed to deliver a reinforcer at the correct time. Studies examining the effects of fading reinforcers in treatment procedures may be ak in to increasing errors of omission. In these
5 studies, which typically focus on treatments involving response-independent reinforcer deliveries (Hagopian, Fisher, & Legac y, 1994; Marcus & Vollmer, 1996; Vollmer, Borrero, Lalli, & Daniel, 1999; Vollmer, Iwata, Zarcone, Smith, & Mazaleski, 1993; Worsdell, Iwata, Hanley, Thompson, & Ka hng, 2000) or DRA (or a variation of DRA known as functional communication training or FCT), the reinforcement schedule is gradually thinned, resulting in fewer reinforcer deliveries and a more extinction-like schedule. In one example of a reinforcement thinning study, Hanley, Iwata, and Thompson (2001) examined four means of thinning th e reinforcement schedule for appropriate behavior in a DRA treatment: delaying the delivery of reinforcement, reinforcing behavior on a fixed interval (FI) schedule, reinforcing behavior on a multiple FR1/EXT schedule, and reinforcing behavior on a mi xed FR1/EXT schedule. During the delay to reinforcement condition, the time between th e emission of the appropriate response and the delivery of the reinforcer gradually increased. During the FI schedule, the first appropriate response after a fixed period of time was reinforcedâ€”responses that occurred before the interval elapsed produced no pr ogrammed consequences. During the FR1/EXT schedules, adjusting periods of time were designated as a reinforcement component, during which every response re sulted in reinforcement, a nd an extinction component, during which no reinforcers were available. These periods were not signaled during the mixed schedule, but were correlated with different colored cards during the multiple schedule. All of these manipulations are si milar to errors of omission because they involve failure (albeit programme d failure) to reinforce approp riate behavior according to the initial treatment plan.
6 The results of the study by Hanley et al. (2001) suggest that errors of omission may have variable, detrimental effects on a ppropriate behavior during DRA treatments. Appropriate responding was eliminated when th e reinforcer was pres ented after a delay. This suggests that failures to deliver rein forcers within close te mporal proximity of appropriate behavior may lead to treatment failure. During the FI schedule, appropriate behavior occurred too frequently, rendering it no longer appropriate. For example, one participant engaged in â€œappropr iateâ€ behavior at a rate exceeding 12 responses per min, which could interrupt ongoing activities or annoy caregivers. Again, this may suggest that periodic failure to reinforce appropriate be havior, as occurred for responding during the FI interval, could also lead to undesirabl e outcomes. Finally, a ppropriate responding was more efficient (that is, occurred during the FR1 component and not during the EXT component) during the multiple schedule than during the mixed schedule. Of the fading procedures tested by Hanl ey et al. (2001), the mixed FR1/EXT is similar to errors of omission as they may occur in the natural environment. That is, caregivers may fail to implement the treatmen t appropriately periodically throughout the day, while implementing the procedure with high integrity at other times. Therefore, it is important to note that, while appropriate behavior occurred regularly throughout the mixed schedule evaluation, rates of problem be havior also increased, suggesting another possible effect of errors of omission. Research examining the effects of ongoi ng reinforcement for problem behavior during behavioral treatments may provide some evidence of the possible effects of errors of commission. This research has been conduc ted primarily with DRA treatments, which typically involve continuous reinforcement for appropriate behavi or and extinction for
7 problem behavior. Studies in which extinction is not implemented as a component of the DRA have resulted in decreased efficacy of the intervention (Kelley, Lerman, & Van Camp, 2002; Shirley, Iwata, Kahng, Mazaleski, & Lerman, 1997). Often, the failure to implement the extinction component of the DRA causes the treatment to fail entirely, with rates of problem behavior remaining hi gh and rates of appropria te behavior low or zero. This may be due to the participants having a long history of reinforcement for problem behavior, which leads to the pers istence of that be havior even when reinforcement becomes available for an appropr iate alternative res ponse. Alternatively, it is possible that the undesired response requires less effort to emit than does the appropriate response. In contrast, errors of commission involve the delivery of a reinforcer following problem behavior. In a study examining the e ffects of these errors on DRA, Worsdell et al. (2000) varied the reinforcement schedul e for problem behavior from FR1 to FR20 while keeping the schedule for appropriate be havior at FR1. For 1 participant, response allocation shifted to appropria te behavior when both responses were reinforced on an FR1 schedule. For the other 2 participants, re sponse allocation gra dually shifted toward appropriate behavior as the schedule for problem behavior became thinner. However, 1 participant continued to engage in problem behavior at low rates even when the schedule of reinforcement was FR20. These results sugg est that errors of commissionâ€”accidental reinforcement of problem behaviorâ€”may be highly detrimental to treatments like DRA. Similarly, research examining the effects of a contingent reinforcement schedule operating concurrently with a response-indepe ndent schedule showed that the rate of responding changed based on the proportion of response-contingent to response-
8 independent reinforcers (Williams & Lattal, 1999), which may have implications for the effects of errors of commission during FT tr eatments. Taken collectiv ely, the results of these studies suggest that not only may errors of commission affect rates of responding, but also that different levels of integrity failure may be correlated with different behavioral effects. Although the above mentioned studies provide some evidence of possible effects of integrity failures on behavioral treatmen ts, none directly focused on the manipulation of procedural integrity at a variety of levels . In fact, few studies have directly examined the effect of treatment integrity failures on treatment outcomes. One exception examined the effects of imperfect implementation of a DRA procedure (Vollmer, Roane, Ringdahl, & Marcus, 1999), which involved periodic reinforcement of pr oblem behavior (errors of commission) and failures to reinforce appropr iate behavior (errors of omission). One feature that distinguis hes this study from those discussed above is the manipulation of both errors of omission and commission. The re sults showed that, overall, the effects of the DRA procedure were resistant to decremen ts in integrity level. However, rates of undesirable behavior increased somewhat when both desira ble and undesirable responses were reinforced. These increases could be pr oblematic when dealing with severe problem behavior, such as self-injury, aggression, or property destruction, and again underscore the potential negative effects of treatment integrity failures. Purpose One potential reason for the paucity of experimental research on treatment integrity is the ethical limitations associated with providing people who engage in severe problem behavior with relati vely long-term, low-quality care. Indeed, intentional exposure of participants with behavior disorders to degrad ed treatments is a possible
9 limitation of conducting preliminary treatmen t integrity research with clinical populations. As an alternative, translational re search methods could be used to determine initial effects of treatment integrity failures. Translational research typical ly begins with controlled laboratory studies that are later replicated with clini cal populations (Lerman, 2003). In addition to insulating at-risk populations from potentially damaging manipula tions, translational research may afford more control over variables th an is usually available in application, such as precise delivery of antecedent stimuli and reinforcers. Isolation of particular variables may illuminate the most influential factors, wh ich can later be examined in application. Translational research may also allow for more rapid manipulation of variables than typically afforded with clinical populations, where the influence of extraneous variables such as therapist differences, limited sessi on times, or the participantsâ€™ histories may affect the outcome of studies, particularly those that use relatively brief experimental phases. Through pilot research, we demonstrated th at rates of responding on an arbitrary task typically changed in accordance with ch anges in the reinforcement schedule, such that higher rates were obtained during inte rval and ratio schedules than during common â€œtreatmentâ€ schedules, such as FT and DR O (St. Peter & Vollmer, 2004). The general method involved the use of a computer pr ogram designed as an analog for common baseline and treatment conditions, which al lowed for automated, precise delivery of antecedent stimuli and reinforcers. Underg raduate students engaged in an arbitrary response, clicking a computer mouse, which served as an analog for problem behavior. Points were used as a reinforcer. This prepar ation typically resulted in rapid changes in
10 responding consistent with schedule contro l, permitting the evaluation of multiple variables over a shor t period of time. The purpose of the current experiments was to study the effects of treatment integrity failures on commonly used behavi oral treatments, using a laboratory preparation. Experiments I through IV evaluated the effects of treatment integrity failures on a DRO treatment. In the first two experi ments, we examined the effects of two possible types of treatment integrity failures. In one type of failure, an arbitrarily defined â€œproblem behaviorâ€ was periodically reinfor ced during the treatment condition (error of commission). The second type of failure i nvolved periodic failure s to deliver earned points (error of omission). It seemed likel y that errors of commission would produce higher rates of problem behavior because th ey create a conjoint schedule in which the rate of reinforcement for the problem beha vior may be higher than the rate of reinforcement available from treatment schedu le. Conversely, errors of omission may be less detrimental because the extinction compone nt of the treatment remains intactâ€”the reinforcement schedule simply becomes thinner as the degree of integrity failure increases. Yet, because treatment integrity is often reported as a single statistic, equal weight is given to omission and commission errors, perhaps incorrectly. Experiment I examined the effects of signaled changes in tr eatment integrity; the signals were removed in Experiment II to examine unsignaled decr eases in integrity. E xperiments III and IV attempted to attenuate the detrimental effects of treatment integrity failures in DRO treatments by providing participants with a hi story of â€œmultiple exemplarsâ€ associated with perfect integrity, or by gr adually decreasing the level of treatment integrity. During Experiment III, participants were exposed to several perfect integrity phases, each
11 associated with different stimulus cond itions, before errors of commission were introduced. During Experiment IV, the level of treatment integrity was decreased by small percentages across conditions, such that participants had expe rience with gradually lower levels of integrity rather than immediate drops in integrity level. Experiments V through VIII examined the effects of treatment integrity failures on a DRA treatment. Experiment V examined the effects of errors of omission alone, errors of commission alone, and combined e rrors of omission and commission, at four different levels of failure. Experiments VI through VIII evaluated whether the condition sequence affected responding during integrity failures. Specifically, these experiments examined whether responding during an error condition would differ based on the preceding condition. This series of experiment s moved from a rapid, highly controlled assessment to a school-based evaluation al ong the basic-to-applie d continuum. In Experiment VI, participants had only a brief exposure to each of the conditions (10min per exposure), similar to Experiments I thr ough V. Experiment VII used similar methods to Experiment VI, but participants were paid based on their point earnings, and conditions were run to stabil ity over longer periods of time (months). Experiment VIII was an applied replication of Experiment s VI and VII. One adolescent boy with developmental disabilities, who was referred for the treatment of aggressive behavior, participated. This experiment allowed assessmen t of the clinical valid ity of the results of the prior human operant research. The present studies extend prior research in three ways. First, they examine the effects of both errors of om ission and errors of commissi on on differential reinforcement procedures. Second, they demonstrate that the e ffects of these errors depend in part on the
12 history of the organism, and that these hi stories can be programmed in a way that attenuates the disruptive effects of integrit y failures. Third, Experiments VI through VIII provide an example of translational resear ch spanning the basicto-applied continuum.
13 CHAPTER 2 GENERAL METHODS FOR EXPE RIMENTS I THROUGH IV Participants and Setting All participants were undergraduate students enrolled in an Introductory Psychology course at the Univer sity of Florida. Students r eceived course credit for completing the experiment, but this credit was not dependent on pe rformance during the experimental sessions. Students registered for participation based on minimal information provided on a website, which included only the session times available, the amount of time required for completion of the study, the location of the st udy, and the number a ssigned to the study by the Institutional Review Board. Each partic ipant attended a single 2-hr experimental session. All sessions were conducted in a labo ratory room equipped with a computer desk, a computer, and a chair. Method When students arrived at their appointmen ts, they were asked to read and sign an informed consent which stated that they were being asked to participate in an experiment on the effects of different contingencies of re inforcement. Participants were also given the following instructions: The purpose of this experiment is to examine how people learn to respond to earn points. If you agree to participate, you will be asked to sit at that computer [point to computer] and play a game. During the game, you should use only the mouse to earn as many points as you can. Although this is technically a 2 hour experiment, we have divided up your time into sessions. You will be able to ta ke a 2-3 minute break between sessions. If you are interested in participating, you should pr int your name on the first page of the consent form and sign the second page.
14 If participants signed the consent, they we re asked to leave all electronic devices, including watches, cell phones, pe rsonal digital assistants (PDAs ), and pagers in the outer room of the experimental area. The purpose of removing these devi ces was twofold: to prevent participants from using these devices instead of attending to the experiment and to prevent participants from overtly timing interval-based schedules. Participants were then escorted to the room that served as the primary experimental area. Participants were asked to take a seat at the desk and were given the following instructions: This is where you will be working. Use only the mouse to earn as many points as you can. Your points will be displayed at the bottom of the screen as a score. The instructions that you see [â€œClick OK to Beginâ€] are the onl y instructions you will receive. Itâ€™s up to you to figure out how to earn as many points as possible. When you click â€˜ok,â€™ the program will start. At the end of the session, a thank you message will appear on the screen. When that happens, please come out [to the outer room] and get me. Good luck! When the participants clicked â€œOK,â€ the program began running, and the programmed schedules of reinforcement took effect. The computer screen was divided into 4 equally sized quadrants by a horizontal and a vertical line as shown in Figure 1. Programmed schedules of reinforcement were in effect only for clicks occurring in the upper left quadrant (hereafter called the â€œ operative quadrantâ€). C licks in all other quadrants were not reinforced. The program recorded clicks made anywhere on the computer screen, the quadrant in which thes e clicks occurred, and point deliveries, as well as the time in the session associat ed with each of these events. Participants earned points by either c licking or not clicking in the upper left quadrant, according to the rein forcement schedules in effect . Fixed interval (FI) and DRO schedules were used. All schedules had a va lue of 15 s. Although this schedule value was chosen arbitrarily, it seemed to provide pa rticipants with sufficient contact with the contingency to permit rapid changes in re sponding when the reinforcement schedules
15 changed. During the FI schedules, the first c lick in the operative quadrant after 15 s had elapsed was followed by a point. During th e DRO schedules, a point was delivered following 15-s periods in which no clicks in th e operative quadrant occurred; any clicks in the operative quadrant reset the interval. Two types of treatment integrity failures were examined: errors of commission and errors of omission. Errors of commission i nvolved the intermittent delivery of points following clicks in the operative quadrant. Th is type of error was designed as an analog of caregivers who may inadvert ently reinforce problem behavi or instead of adhering to the DRO treatment. Errors of omission involve d the intermittent failu re to deliver an earned point. During omission phases, particip ants who refrained from clicking in the operative quadrant for the durat ion of the interval would so metimes not receive a point. This type of error was designed as an analog of caregivers who may â€œforgetâ€ to deliver reinforcers in accordance with the DRO treatment. Each level of integrity was associated with a correlated probability of errors, such that 90% treatment integrity was associat ed with an error probability of 0.1, 80% integrity was associated with a probability of 0.2, 70% integrity was associated with a probability of 0.3, and 60% integrity was asso ciated with a probability of 0.4. During errors of commission, this crea ted a conjoint schedule by whic h participants could earn points either following a period of no res ponding, according to the DRO schedule, or following a response, according to the probabi listic error schedule. During errors of omission, participants could earn points only according to the DRO schedule (the extinction component of the schedule was preserved). However, there was a certain probability that earned points would not be delivered.
16 Although participants were instructed to earn as many points as possible, these points were not exchangeable for any back-up re inforcers. Prior research in our lab (St. Peter & Vollmer, 2003) and others (e.g., Wein er, 1969) has shown that points alone will function as reinforcers for college students. As an additional assura nce that points alone would reinforce behavior, clear reinforcement effects and appropriate changes in rates of responding following contingency changes betw een baseline (FI schedule) and treatment (DRO schedule with perfect inte grity) were required for a participant to be included in the experiments. Participants who did not s how differentiation in response rates between baseline and DRO phases were excluded of 45 participants failed to meet the inclusion criteria. At the end of the 2-hr participation time, participants were granted their course credits and were asked if they had any questi ons about the study, or if they wanted a copy of the informed consent. Score: 1 Figure 1: Screen shot of computer program used for Experiments I-IV
17 CHAPTER 3 EFFECTS OF SIGNALED DE CREASES IN INTEGRITY Procedure The purpose of Experiment I was to determ ine the effects of signaled decreases in treatment integrity on DRO treatments. This type of signaled decrease in treatment integrity may be analogous to different caregivers implementing the treatment with different levels of integrity. Some caregivers may implement the treatment with perfect integrity whereas others make e rrors of omission or commission. Eleven undergraduates participated in Experi ment I. Each student participated in a two-hour experimental session. Participants took a three-min break following every 27min of participation. One participant withdr ew from the study af ter the second break. Data from another participant were ex cluded because of a computer malfunction following the third break, resu lting in an incomplete da ta set. Thus, data from 9 participants were in cluded in the study. The participants were divided into 4 gr oups, each exposed to a different level of integrity failure. The groups, number of pa rticipants in each group, and associated integrity levels are listed in Table 1. Each participant was exposed to the same order of conditions, regardless of the treatme nt integrity level grouping. Table 1: Error level groups during Experiment I Treatment Integrity Probability of Errors Number of Participants 90% 0.1 2 80% 0.2 3 70% 0.3 2 60% 0.4 2
18 The order of conditions and the stimulus associated with each condition are listed in Table 2. Participants were first exposed to an FI 15-s sc hedule to establish responding, during which the first click in the operative qua drant after 15s resulted in a point delivery. Following exposure to the FI schedule, a DRO 15-s schedule was implemented, during which responses in the operative quadrant resu lted in resetting of the 15-s timer, and an absence of responding in the operative quadr ant for 15 s resulted in the delivery of a point. After responding had been established and reduced by the FI and DRO schedules, respectively, errors of omission and comm ission were evaluated. A multiple schedule was in place throughout this experiment; each type of condition was associated with a distinct background colo r of the computer screen (see Ta ble 2 for the specific stimuli associated with each condition). Table 2: Order of conditions during Experiment I Reinforcement Schedule Background Color FI 15 s Green DRO 15 s (perfect integrity) Red DRO 15 s (errors of commission: DROc) Orange DRO 15 s (errors of omission: DROo) Yellow DRO 15 s (errors of commission: DROc) Orange DRO 15 s (errors of omission: DROo) Yellow BREAK FI 15 s Green DRO 15 s (perfect integrity) Red DRO 15 s (errors of omission: DROo) Yellow DRO 15 s (errors of commission: DROc) Orange DRO 15 s (errors of omission: DROo) Yellow DRO 15 s (errors of commission: DROc) Orange During errors of commission, participan ts could earn points by not responding, according to the DRO schedule, or after a response in the operative quadrant on an intermittent schedule, as determined by the programmed integrity level. Points earned
19 based on the integrity failures were deliver ed in a manner similar to a random ratio schedule, based on an experimenter-specified probability. During errors of omission, participants could only earn points by not responding, according to the DRO schedule. However, some earned points were not delivered, as determined by the programmed integrity le vel. Whether points were delivered or withheld was automatically calculated by the computer, as described above for commission points. All phases were 9-min in length, except for participant 90TI1, who was exposed to 10-min phases (this caused the experiment to run considerably over the allotted 2-hr time limit because of break time, so exposures were decreased for subsequent participants). Results and Discussion Results for participants exposed to 90% and 80% integrity are shown in Figure 2. Results for participants exposed to 70% a nd 60% integrity are shown in Figure 3. The level of integrity for each participant is denoted by the first two numbers in the participant code. For example, participant 90TI1 was expos ed to 90% integrity, while participant 60TI1 was exposed to 60% integrit y. In both figures, mouse clicks (responses) in the operative quadrant are along the y-axis , and consecutive minut es in the session are along the x-axis. Error conditi ons are labeled as either commission or omission by the letters C or O, followed by the proba bility of integrity failure.
20 Although differences in the absolute rate s of responding and degree of change between phases differed across participants, 8 of the 9 pa rticipants showed similar patterns of responding. All 9 par ticipants responded at great er rates during the FI 15-s schedule than during the DRO 15-s schedule with perfect in tegrity, suggesting that the DRO was an effective â€œtreatmentâ€ for the analog problem behavior. Eight of the 9 participants responded at higher rates during one or more phases involving errors of commission relative to both the FI and DRO with pe rfect integrity phas es, suggesting that â€œerroneousâ€ reinforcement of problem behavior may be particularly detrimental to DRO treatments. In contrast, none of the partic ipants responded at hi gher rates during phases involving errors of omission re lative to DRO when implemen ted with perfect integrity. Notably, increases and decreases in respondi ng occurred rapidly after the start of each phase. The rapidity in change of response rates may be due to the distinct stimulus conditions associated with each condition type. One possible treatment implication is that caregivers who implement a treatment with errors of commission may never see treatment effects, even when the child is making a transition to that caregiverâ€™s environment from one in which the treatment was implemented with perfect integrity. Data for the 1 participant for whom e rrors of commission did not decrease the effectiveness of the DRO are depicted in th e graph labeled 70TI1, located in the upper left panel of Figure 3. For this participant, rates of responding were highest in the FI phases, and responding was eliminated during DRO. However, response rates did not increase when errors of commission were in pl ace. It is possible that this participant did not have sufficient exposures to the errors for rates to increase. Because no responses occurred during the errors of commission until the third exposure, this participant did not
21 contact the commission error contingency until the 82nd min of participation. The history with DRO implemented with perfect integrity and with errors of omission may have affected this participantâ€™s responding (although participant 70 TI2 also did not contact the commission contingency until the 82nd min, yet errors of commission increased response rates once the contingency was contacted). The level of treatment integrity to which the participant was exposed did not seem to affect the overall pattern of results: errors of commission were detrimental to the treatment, even when the DRO was implem ented at 90% integrity. Additionally, lower levels of treatment integrity did not seem to be consistently correlated with greater overall increases in rates of responding during commission phases. Figure 4 shows the mean rates of responding during each phase. Mean rates were calculated by summing the number of responses within a condition, acr oss participants, and dividing by the total number of minutes of exposure to that condition. Mean rates of responding were moderate during the FI cond ition, low during DRO with perf ect integrity and DRO with errors of omission, and high during DRO with errors of commission. The only substantial difference in rates of responding occurred during errors of commission, in which 60% treatment integrity resulted in greater rate s of responding than did the other levels of treatment integrity. It is important to note th at this increase in mean rates is due to participant 60TI2 engaging in extremely hi gh response rates during the error conditions (figure in lower right panel of Figure 3), with response rates r eaching over 1000 clicks per minute during the commission phases. The results of Experiment I show that errors of commission may be highly detrimental to DRO treatments, regardless of the integrity level at which those errors
22 occur. Failure to deliver an earned rein forcer is less detrimental to DRO than inappropriately reinforcing problem behavior . The detrimental effects of errors of commission are in line with those of prio r applied research showing that DRO is ineffective when implemented without extinc tion (Mazaleski, Iwata, Vollmer, Zarcone, & Smith, 1993). Notably, commission errors at 90% integrity were far more detrimental than omission errors at 60% integrity. Thus, integrity statistics are only meaningful when the context of the integrity failure is understood. When errors of commission are in place, the reinforcement schedule basically becomes a conjoint DRO RR schedule. That is, participants can earn points based on the DRO schedule by not responding, or can earn poin ts on a probabilistic basis, similar to a random ratio (RR) reinforcement schedule. Pr ior research has shown that organisms will choose the richer reinforcement schedule (Her rnstein, 1970). In this experiment, even participants exposed to 90% integrity c ould earn more points during commission phases by responding than not responding, as long as mo re than 20 responses occurred per 15 s interval (with 20 responses pe r 15-s interval, an average of 2 points would be earned every 15s, but a maximum of 1 point could be earned every 15 s based on the DRO schedule). Because most participants res ponded at rates upwards of 200 responses per min (approximately 3.3 responses per s, or 50 responses in 15 s) during the commission phase, participants were able to maximize point earnings. The suppressive effects of DRO with erro rs of omission are not surprising when considering the actual schedule of reinforcem ent in effect. During errors of omission, the DRO schedule becomes more extinction-like as reinforcers are omitted, causing the DRO to operate like a variable DRO (vDRO) reinfo rcement schedule, in which reinforcers are
23 delivered following varying periods of tim e during which no instances of the target response occur. Because vDRO schedules have been successfully used as a treatment in other studies (Lindberg, Iwat a, Kahng, & DeLeon, 1999), it is re latively unsurprising that a DRO with errors of omission would conti nue to suppress behavi or. However, it is possible that errors of omission may result in negative side eff ects similar to those reported with extinction treatments. Other than extinction bursts (w hich did not occur), these possible side effects were not examined in this series of experiments; future research should examine the possibility of negative side effects. Although this research was conducted in a laboratory setting, the results have implications for application. First, even though in practice â€œmissedâ€ opportunities to present reinforcers are often emphasized as problematic, such errors of omission are probably not too harmful in DRO treatments. S econd, the detrimental effect of errors of commission suggests that the extinction component of DRO treatments should be stressed when training caregivers. Because caregivers often have a long history of reinforcing problem behavior, the extin ction component of DRO treatments may ordinarily be prone to treatment integrity fa ilures. If caregivers â€œgive inâ€ when response rates increase and provide a reinforcer fo llowing problem behavior, treatment failure becomes likely. Third, the results suggest th at there is no absolute level of â€œgoodâ€ treatment integrity. Even when errors of co mmission occurred with a probability of 0.1â€” the treatment was implemented with 90% in tegrityâ€”errors of co mmission were highly detrimental. Experiment I showed that errors of co mmission are highly detrimental to DRO treatments, when the change to a phase invo lving errors of commission are signaled by a
24 stimulus change. This may be analogous to different caregivers implementing the treatment with different levels of integrityâ€”some caregivers implement the treatment perfectly, while others make errors. It is possible, however, that a single caregiver implements the treatment at varying levels of integrity throughout the day. In that case, decreases in treatment integrity may not be associated with stimulus changes. When decreases in treatment integrity are not signale d, the effects of integrity failures may not be as dramatic. Experiment II addressed this possibility by not providing stimulus changes associated with decr eases in integrity level.
25 Figure 2: Participants expos ed to 90% and 80% integrity with signaled errors Participant 90TI1Minutes 0102030405060708090100110120Responses 0 50 100 150 200 250 300 FI15 DRO15DRO15 C 0.1 DRO15 0 0.1 FI15 DRO15 DRO15 C 0.1 DRO15 C 0.1 DRO15 C 0.1 DRO15 0 0.1 DRO15 0 0.1 DRO15 0 0.1 Participant 90TI2Minutes 0102030405060708090100110Responses 0 200 400 600 800 FI15 DRO15DRO15 C=.1 DRO15 O=.1 DRO15 C=.1 DRO15 C=.1 DRO15 C=.1 DRO15 O=.1 DRO15 O=.1 DRO15 O=.1 FI15 DRO15 Participant 80TI5Minutes 0102030405060708090100110Responses 0 50 100 150 200 250 300 FI15 DRO15 DRO15 C=.2 DRO15 O=.2 FI15DRO15 DRO15 C=.2 DRO15 c=.2 DRO15 C=.2 DRO15 O=.2 DRO15 O=.2 DRO15 O=.2 Participant 80TI3Minutes 0102030405060708090100110Responses 0 50 100 150 200 250 300 FI DRO DRO15 C=.2 DRO15 O=.2 FI DRO DRO15 C=.2 DRO15 C=.2 DRO15 C=.2 DRO15 O=.2 DRO15 O=.2 DRO15 O=.2 Participant 80TI4Minutes 0102030405060708090100110Responses 0 100 200 300 400 500 FI15 DRO15DRO15 C=.2 DRO15 O=.2 FI15DRO15 DRO15 C=.2 DRO15 C=.2 DRO15 C=.2 DRO15 O=.2 DRO15 O=.2 DRO15 O=.2
26 Figure 3: Participants expos ed to 70% and 60% integrity with signaled errors Figure 4: Mean rates of responding for participants with signaled errors Participant 70TI2 (CG) Operative QuadrantMinutes 0102030405060708090100110Responses 0 100 200 300 400 500 FI15DRO15 DRO15 c=.3 DRO15 c=.3 DRO15 c=.3 DRO15 c=.3 DRO15 o=.3 DRO15 o=.3 DRO15 o=.3 DRO15 o=.3 FI15DRO15 Participant 70TI1 (JT) Operative QuadrantMinutes 0102030405060708090100110Responses 0 20 40 60 80 100 FI15 DRO15DRO15 C = .3 DRO15 O = .3 DRO15 C = .3 DRO15 C = .3 DRO15 C = .3 DRO15 O = .3 DRO15 O = .3 DRO15 O = .3 FI15 DRO15 Participant 60TI1 (DR) Operative QuadrantMinutes 0102030405060708090100110Responses 0 100 200 300 400 500 FI15DRO15DRO15 C = .4 DRO15 O = .4 DRO15 C = .4 DRO15 C = .4 DRO15 C = .4 DRO15 O = .4 DRO15 O = .4 DRO15 O = .4 FI15DRO15 Participant 60TI2 (JG) Operative QuadrantMinutes 0102030405060708090100110Responses 0 200 400 600 800 1000 1200 FI15DRO15 DRO15 C = .4 DRO15 O = .4 DRO15 C = .4 DRO15 C = .4 DRO15 C = .4 DRO15 O = .4 DRO15 O = .4 DRO15 O = .4 FI15DRO15 Mean Rate of Responding 0 50 100 150 200 250 300 350 400 FIDRODROcDROo ConditionClicks per minute 90% 80% 70% 60%
27 CHAPTER 4 EFFECTS OF UNSIGNALED DECREASES IN INTEGRITY Procedure Experiment II examined the effects of d ecreased treatment integrity levels that were not associated with distinct background co lors of the computer screen. This may be analogous to a single caregiver implementing a treatment with varying levels of integrity. Ten undergraduates participated in Experiment II. The 10 participants were randomly divided into two groups, with 5 pa rticipants in each group. Group I experienced no stimulus changes throughout the experiment â€”the background of the computer screen remained green regardless of the reinforcem ent schedule in effect . Group II experienced stimulus changes between baseline and tr eatment, but not between treatment and subsequent integrity failures. For these pa rticipants, the backgr ound of the computer screen was green during FI and red during a ll DRO phases, regardless of the level of treatment integrity. The second group was cr eated after two participants from Group I failed to show consistent decreases in responding when DRO was implemented with perfect integrity (depicted in Figure 5). Data from these 2 participants were excluded from the study, leaving 3 participants in Gr oup I. One participant, NSD01, participated for only 115 min with an 8-min break because of a computer error during the break that delayed the start of the second participation block. Students in both groups pa rticipated in a 123-min experimental session. Participants took a 3-min break following the fi rst 60-min of particip ation. Participants were exposed to both errors of commission and errors of omission, but at a single level of
28 integrity failure (80%). The baseline and full treatment schedules used were identical to Experiment I (FI15s and DRO15s, respectively) . Participants were first exposed to FI15s to establish responding, then to DRO15s with perfect integrity. Following those initial exposures, errors of commission and omi ssion were alternated , with four total replications of each of the error types. Results and Discussion Figures 5 through 7 show the results fo r Experiment II. For all figures in Experiment II, responses are along the y-axis and consecutive minutes in the session are along the x-axis. Condition labels denote the re inforcement schedule in effect: FI15 (FI), DRO15 (DRO), DRO15 with 80% errors of commission (DROc), and DRO15 with 80% errors of omission (DROo). As mentioned previously, DRO implemente d with perfect integrity was effective for only 8 of the 10 participants. Figure 5 s hows the data for the two participants for whom the DRO was ineffective. Because DR O implemented with perfect integrity was insufficient to produce reliable reductions in responding, data from these 2 participants were excluded from further analysis in Experi ment II. However, it is interesting to note that both participants were in Group I, and, therefore, e xperienced no stimulus changes when reinforcement schedules changed. It is possible that, for these participants, 10-min exposures to the DRO schedule were insufficient to decrea se responding. This could be due to a failure to discriminate between the FI and DRO conditions. Discrimination may be facilitated by changes in stimulus conditions between th e FI and DRO. Indeed, all participants in Group II, who experienced a green background during FI and a red background during DRO, showed a differentia tion in response rates between the two conditions.
29 Figure 5: Participants disqualified from furthe r analysis. Both participants experienced signaled changes between FI and DRO Results for the remaining 8 participants in Groups I and II are shown in Figures 6 and 7, respectively. In both figures, rate s of responding are along the y-axis and consecutive minutes in the session are along th e x-axis. As in Figur e 5, error conditions are labeled as DRO C (DRO with errors of commission) and DRO O (D RO with errors of omission). In general, rates of respondi ng were moderate during FI, low during DRO with perfect integrity and errors of om ission, and high during mo st DRO phases with errors of commission. All 3 participants in Group I responded at greater rates during
30 phases involving errors of commission than during any other condition, as shown in Figure 6. For 1 participant in Group I (NSS04, shown in the uppe r left panel of Figure 6), rates did not increase during errors of comm ission until the final two exposures to that condition. This could be due to limited c ontact with the commissi on contingency during the initial exposures to that condition, as a result of low response rates. Errors of omission were as effective at suppressing re sponse rates during DRO implemented with perfect integrity for all participants in Group I. Figure 6: Participants who experi enced no changes in background color
31 Figure 7: Participants who e xperienced changes in backgr ound only between FI and DRO Results for participants in Group II are de picted in Figure 7. Results similar to those from Group I were obtained for 3 of the 5 participants in Group II (NSD03, NSD05, and NSD07). For these 3 participants, rates of responding were elevated during
32 errors of commission but remained low duri ng errors of omission. Participants NSD01 and NSD09 (upper and lower left panels of Figure 7, respectively) did not respond at higher rates during commission phases. For participant NSD01 (upper left panel), exposure to the initial DRO treatment, impl emented with perfect integrity, decreased response rates to zero within the first two min of the treatment, and responding remained at zero for much of the rest the participan tâ€™s exposure to the red background (all DRO conditions, including treatment integrity failur es). This participant only contacted the commission contingency during the final minute of the thir d commission phase (minute 95), which seemed to result in a brief in crease in responding dur ing the subsequent omission phase. Similarly, the DRO treatment decreased rates of re sponding to 0 within the first 2 min of the treatment for particip ant NSD09 (lower left panel). Following this decrease, responding remained at 0 throughout the rest of the exposures to DRO, including those in which integrity failures were in place. Therefore, this participant never contacted the commission error contingency. The results of Experiment II indicate that errors of commission may be detrimental to DRO treatments when the change s in treatment integr ity are not signaled. This suggests that even a single caregiver who occasionally makes errors may undermine the treatment effects, at least while th e treatment integrity is compromised. However, the results of Experiment II were more mixed than those obtained during Experiment I. Two participants in Group I did not qualify for the study. Additionally, 2 participants fr om Group II did not respond at hi gher rates duri ng errors of commission. It is important to note that th ese 2 participants did not have substantial contact with the commission contingency because the DRO was so effective at
33 suppressing responding. It is also interesting that both of thes e participants were in Group II, who experienced a change in the background color of the computer screen when the reinforcement schedule changed from FI to DR O. It is possible that these 2 participants generated some rule about responding, such as â€œClick during green, do not click during red.â€ Although no formal data were collected on this type of rule-generation, 1 of the 2 participants did informally report such a ru le at the conclusion of the experiment. Both types of resultsâ€”those showing incr eases in rates during commission phases and those showing no increasesâ€”can inform applied research and practice. Those participants for whom errors of commissi on were detrimental suggest that caregivers must remain vigilant about the implemen tation of DRO treatments, as errors of commission undermined the treatment even when they were not signaled. It could be that one way to attenuate the effects of possible integrity failure s is to ensure that the DRO is implemented with perfect integrity until rates of responding consistently remain at zero. When responding is eliminated or greatly reduced, as with part icipants NSD01 and NSD09, errors of commission have li ttle or no opportunity to occur. Unfortunately, decreasing rates of res ponding to zero is not always possible, precluding this type of insulation from the de trimental effects of e rrors of commission. Another possibility is to provi de participants with a hist ory with different stimulus contexts with perfect integrity before intr oducing stimulus contexts with errors. We evaluated this approach, which shares features with procedures using â€œmultiple exemplarsâ€ to teach new responses (Sprague, 1993), in Experiment III.
34 CHAPTER 5 EFFECTS OF MULTIPLE EXEMPLARS ON RESPONDING DURING ERRORS OF COMMISSION Procedure Twelve undergraduate students participated in Experiment III. One participant withdrew from the experiment after 65 min of participation. Data from 1 additional participant were excluded from data analysis because there was no differentiation between responding during FI and DRO with perfect integrity. Thus, a total of 10 participants remained. Students participated fo r a total of 103 min, which was broken into 2, 50-min blocks separated by a 3-min break. The particip ants were exposed to an FI 15-s schedule for the first and last 10 min of the sessi on. During FI conditions, the background of the computer screen was green. Following the initi al FI exposure, partic ipants were exposed to 10 min of a DRO 15-s schedule, associat ed with a red background. All subsequent exposures to DRO were 5 min a nd were associated with a va riety of different background colors, including blue, orange, cyan, and yello w. These phases provided participants with a history of multiple stimuli associated with perfect integrity, similar to the â€œmultiple exemplarsâ€ approach used in prior researc h. This procedure may be analogous to multiple caregivers implementing a procedure perfectly before introducing a caregiver who makes errors of commission. Phases involving DRO with errors of commission, implemented with 80% integrity and associated with a purple background, were included three times
35 interspersed among the perfect integrity cond itions. The specific order of conditions is listed in Table 3. Table 3: Order of conditions during Experiment III Schedule Component Time % Integrity Background Color FI 15 s 10 min 100 Green DRO 15 s 10 min 100 Red DRO 15 s 5 min 100 Blue DRO 15 s 5 min 100 Orange DRO 15 s 5 min 100 Cyan DRO 15 s 5 min 100 Yellow DRO 15 s 5 min 80 Purple DRO 15 s 5 min 100 Blue BREAK DRO 15 s 10 min 100 Red DRO 15 s 5 min 100 Blue DRO 15 s 5 min 80 Purple DRO 15 s 5 min 100 Yellow DRO 15 s 5 min 100 Cyan DRO 15 s 5 min 100 Orange DRO 15 s 5 min 80 Purple FI 15 s 10 min 100 Green Results and Discussion For 6 participants in Experiment III, e xperience with multiple stimuli correlated with perfect treatment integrity seemed suffi cient to suppress responding when errors of commission were in place. Results for these pa rticipants are depicted in Figure 8. In this figure and all subsequent figures in Experime nt III, response rate is shown along the yaxis and consecutive minutes in the expe riment along the x-axis. Condition labels correspond to the stimulus presented duri ng the condition. As mentioned above, all stimuli were associated with perfect inte grity except for the purple background, which was associated with 80% errors of commi ssion. All 6 participants whose data are depicted in Figure 8 responded at moderate rates during FI, low to zero rates during DRO
36 implemented with perfect integrity, and low to zero rates when errors of commission were in place. Figure 8: Participants for whom multiple ex emplars were sufficient to reduce responding These results have implications for applic ation. For example, it may be possible for several highly trained therapists to impleme nt the procedure perfec tly and to provide a Participant 702Minutes 0102030405060708090100Responses 0 2 4 6 8 10 12 14 16 18 FI Green RedBlue Orange Cyan Yellow Purple BlueRedBlue Purple Yellow Cyan Orange PurpleFI Green Participant 701Minutes 0102030405060708090100Responses 0 20 40 60 80 100 120 140 FI green RedBlue Orange Cyan Yellow Purple BlueRedBlue Purple Yellow Cyan Orange BlueFI green Participant 705Minutes 0102030405060708090100Responses 0 50 100 150 200 FI Green RedBlue Orange Cyan Yellow Purple Blue BlueRedPurple Purple Yellow Cyan OrangeFI Green Participant 707Minutes 0102030405060708090100Responses 0 20 40 60 80 100 120 140 160 FI Green RedBlue Orange Cyan Yellow Purple Blue BlueRedPurple Purple Yellow Cyan OrangeFI Green Participant 710Minutes 0102030405060708090100Responses 0 10 20 30 40 50 FI Green RedBlue Orange Cyan Yellow Purple Blue BlueRedPurple Purple Yellow Cyan OrangeFI Green Participant 712Minutes 0102030405060708090100Responses 0 10 20 30 40 50 60 FI Green RedBlue Orange Cyan Yellow Purple Blue BlueRedPurple Purple Yellow Cyan OrangeFI Green
37 high degree of monitoring with parents or ot her caregivers to ensu re sufficient exposure to perfect integrity associated with a variety of different care providers. This type of history may insulate the treatment from detr imental effects associated with less trained caregivers implementing the treatment imperfectly. Figure 9 shows results for the remaining 4 participants in Experiment III. Three of these participants (708, 706, & 711) engaged in slightly higher rate s of responding during the stimuli associated with perfect integrit y than did the participants whose data are depicted in Figure 8. Usually, these increased rates occurred shortly after a stimulus change occurred. Participant 706 (upper right panel of Figure 8) engaged in slightly elevated rates of responding during the sec ond exposure to errors of commission (from minutes 68 to 70), which seemed to affect responding during late r perfect integrity conditions. This pattern of re sponding may have important implications for application, as it suggests that exposur e to treatment integrity fa ilures may undermine later highintegrity treatment efforts. Only 1 participant engaged in high ra tes of responding during the commission conditions. Data for this part icipant are shown in the lo wer right panel of Figure 8 (participant 709). This participant responded at low to moderate rates during FI and, eventually, low to zero rates during DRO impl emented with perfect integrity. Response rates were highest in the phases with erro rs of commission. Therefore, a history with multiple stimuli associated with perfect in tegrity was insufficient to attenuate the detrimental effects of treatment integrity failur es for this participant. Interestingly, this participant was the only one who responded at least once per minute during all the phases of DRO with perfect integrity. Although further research is ne eded to determine if this
38 variable is responsible for the increase in rates during errors of commission, it suggests that errors of commission may be especially detrimental when the initial DRO does not eliminate responding. Figure 9: Participants for whom multiple exemplars were insufficient to suppress responding Overall, providing a history with several different stimuli associated with perfect integrity was successful at attenuating the detr imental effects of errors of commission. These results suggest that it may be beneficial to have multiple highly trained individuals implement a procedure before introducing in dividuals who are likely to make errors, especially errors of commission. Participant 706Minutes 0102030405060708090100Responses 0 20 40 60 80 100 120 140 160 FI Green RedBlue Orange Cyan Yellow Purple Blue BlueRedPurple Purple Yellow Cyan OrangeFI Green Participant 708Minutes 0102030405060708090100Responses 0 50 100 150 200 250 300 FI Green RedBlue Orange Cyan Yellow Purple Blue BlueRedPurple Purple Yellow Cyan OrangeFI Green Participant 709Minutes 0102030405060708090100Responses 0 100 200 300 400 FI Green RedBlue Orange Cyan Yellow Purple Blue BlueRedPurple Purple Yellow Cyan OrangeFI Green Participant 711Minutes 0102030405060708090100Responses 0 2 4 6 8 10 12 14 16 FI Green RedBlue Orange Cyan Yellow Purple Blue BlueRedPurple Purple Yellow Cyan OrangeFI Green
39 One limitation of this study is that the se quence of conditions did not include an examination of 80% errors of commission befo re the history with perfect integrity was developed. Therefore, there is no evidence that errors of commission would have resulted in high rates of responding for these participan ts. However, the results of Experiments I and II showed that errors of commission re liably produced elevated response rates. Additionally, the condition sequence used in this experiment s howed that the low rate of responding throughout the DRO phases was not cause d by a loss of reinforcer efficacy, as rates of responding increased for all partic ipants during the fi nal FI condition. A second limitation is the lack of c ontrol for duration of exposure to DRO implemented with perfect integrity. Assessing the effect of multiple exemplars of perfect integrity required more overall exposure to th at condition than in previous experiments. This confound could be addressed in future research to ensure that having multiple stimuli associated with the treatment is more critical than simply increasing the amount of exposure to the DRO implemented perfectly. Regardless of the mechanism behind the eff ects, it seems that this procedure was effective in attenuating th e disruptive effects of erro rs of commission, and suggests possibilities for application. In particular, these results s how that DRO treatments can remain effective even when one caregiver implements the treatment with errors of commission. However, it is possible that a single caregiver would initially implement the treatment well but later make errors, similar to Experiment II. We examined the possibility that gradual decrea ses in treatment integrity may attenuate the effects of these errors in Experiment IV.
40 CHAPTER 6 EFFECTS OF INTEGRITY FADING ON RESPONDING DURING ERRORS OF COMMISSION Procedure Twelve undergraduate students participat ed in Experiment IV. None of the participants withdrew or failed to meet the inclusion criteria for the experiment. Students participated for a total of 103 min, which in cluded 2, 50-min blocks in the experiment with a 3-min break between blocks. The seque nce of conditions is listed in Table 4. Table 4: Order of conditi ons during Experiment IV Schedule Component Time % Integrity Background Color FI 15 s 10 min 100 Green DRO 15 s 10 min 100 Red FI 15 s 10 min 100 Green DRO 15 s 10 min 100 Red DRO 15 s 10 min 80 Blue BREAK DRO 15 s 10 min 100 Red DRO 15 s 5 min 97 Red DRO 15 s 5 min 95 Red DRO 15 s 5 min 90 Red DRO 15 s 5 min 84 Red DRO 15 s 5 min 80 Red DRO 15 s 5 min 80 Blue DRO 15 s 10 min 100 Red Participants were first exposed to an FI 15-s schedule, which was correlated with a green background of the computer screen. Fo llowing the FI, the reinforcement schedule was changed to DRO 15-s schedule implem ented with perfect integrity, which was associated with a red background. These conditio ns were replicated before assessing the effects of a sudden decrease in treatment in tegrity to 80% during an error of commission
41 phase, which was associated with a blue background. These phases were all 10-min in duration, and were followed by a 3-min break. After the break, DRO was again implemented with perfect integrity for a 10-min period, which was followed by a gradual decrease in integrity. Specifically, treatm ent integrity decreased in the following sequence: 97%, 95%, 90%, 84%, and 80%, ex cept for the participant F08, who was exposed to 97%, 93%, 90%, 87%, and 84%. The steps in the fading procedure were changed after participant F08 to allow a ssessment of 80% integrity during the fading procedure. For all participan ts, the background of the computer screen remained red during the fading procedure. This may be analogous to one caregiver initially implementing the treatment perfectly, but begi nning to slip or dr ift over time. At the conclusion of the experiment, we evaluated the effects of 80% integrity associated with a stimulus change before returni ng to a full integrity condition. Results and Discussion Figures 10 and 11 depict data from particip ants in Experiment IV. As in previous figures, responses are depicted along th e y-axis and consecutive minutes in the experiment are depicted along the x-axis. In both figures, the numerical condition labels show the percentage of treatment integrity in place during the DRO schedule in that phase. The conditions labeled 80( B) denote the 80% integrity pha ses that were associated with a change in the backgr ound color from red to blue. For all participants, response rates were low to moderate during FI and low to zero during DRO with perfect in tegrity. Results during the er rors of commission varied somewhat across participants. Figure 10 show s results from 4 participants (F08, F25, F26, and F27) for whom response rates incr eased during the initi al exposure to 80% errors of commission, when that condition wa s signaled by a change in the background
42 color of the computer screen. Interestingly, this increase was either not observed or not maintained for 8 of the participants (F 18, F19, F20, F21, F22, F23, F24, and F28, shown in Figure 11). In general, these participants had lim ited or no contact with the commission contingency because low rates of responding seemed to carry over from the previous perfect integrity phase. For 5 of these participants (F21, F22, F23, F24, and F28), the second exposure to the signaled 80% errors phase did result in increased response rate. However, response rates had begun to increase in the previous fading phases for all of these participants. Neverthele ss, the occasional failu re of 80% errors to increase response rate are unusual in light of the previous experiments. They suggest that, for some participants, commission errors may have no detrimental effects, regardless of the programmed level of treatment integrit y, because no responses ever contact the reinforcer. For the 4 participants who responded at high rates during the first commission contingency (Figure 10), the gradual decrease s in integrity were effective at attenuating the disruptive effects of errors of commi ssion for 2 participants (F08 and F26). For participant F08, rates remained at low or zero levels during the entire fading process, but high, variable rates were observed when th e background color changed during the final presentation of the 80% integrity level. Partic ipant F26 responded at high rates during the initial exposure to 80% integrity, and low to zero rates during the remainder of the experiment, including when the background colo r of the computer screen changed during the final presentation of 80% integrity. This participant seemed to be attending to the experiment, so it is possible that the low ra tes of responding were due to the long history
43 with DRO established during the experiment, or to the loss of reinforcing efficacy of the points. For the other 2 participants who responded at high rates during the initial 80% presentation, response rates clim bed steadily as the level of treatment integrity declined. For F27, responses rates began to increase during the expos ure to 95% integrity, while response rates for F25 increased during exposur e to 90% integrity. Similar results were obtained for 2 participants who did not respond at high rates during th e initial exposure to 80% integrity (F21 and F23). For all the par ticipants for whom fading was not successful at attenuating the disruptive eff ects of integrity failures, response rates during the failure phases rose well above those observed in prev ious conditions, sim ilar to the rates of responding during errors of commission observed in Experiments I and II. Two of the participants (F22 and F24, shown in Figur e 11) responded at low to moderate rates during the inte grity fading conditions. For th ese participants, gradually decreasing the level of integrity was not e ffective at suppressing response rates to the same degree as the full integrity condition. Howe ver, the effects of commission errors did not seem as detrimental as they had in previ ous experiments. This suggests that gradually decreasing the level of integrity may still attenuate the detrimental effects of commission errors, even when not completely e ffective at suppressing response rate. Although the gradual decreases in treatment integrity we re not effective for all 12 participants in this experime nt, the increases in response ra te were evident quickly (upon 90% integrity). This suggests that clinicians may be able to rapidly evaluate whether or not gradual decreases in integr ity would attenuate the detrim ental effects of errors of commission. Additionally, these results undersco re the influence of errors of commission
44 on responding during DRO treatments, even at inte grity values that are at â€œhighâ€ levels. Experiments I through IV addressed one commo n behavioral treatment, DRO. In recent years, DRA has become an even more widely used variation of differential reinforcement procedures. Experiments V through VIII involv ed evaluations of treatment integrity failures during DRA treatments. Figure 10: Participants whose response rate increased during the first signaled exposure to 80% integrity Participant F08 (BL)Minutes 0102030405060708090100Responses 0 100 200 300 400 500 FI 100 FI 100 80(B) 100 97 93 90 878480 (B) 100 Participant F25Minutes 0102030405060708090100Responses 0 100 200 300 400 500 600 700 FI 100 FI 100 80(B) 100 979590 868080 (B) 100 Participant F26Minutes 0102030405060708090100Responses 0 50 100 150 200 250 300 350 FI 100 FI 100 80(B) 100 9795 90 8680 80 (B) 100 Participant F27Minutes 0102030405060708090100Responses 0 100 200 300 400 500 600 FI 100 FI 100 80(B) 100 97 95 90 84 80 80 (B) 100
45 Figure 11: Participants who did not respond dur ing the initial exposure to 80% integrity Participant F24Minutes 0102030405060708090100Responses 0 10 20 30 40 50 60 FI 100 FI 10080(B)100 97 95 90 84 80 80 (B) 100 Participant F23Minutes 0102030405060708090100Responses 0 200 400 600 800 1000 FI 100 FI 100 80(B) 10097 95 90 86 80 80 (B) 100 Participant F18 (CL)Minutes 0102030405060708090100Responses 0 20 40 60 80 FI 100FI 10080(B) 100 97 95 90 84 80 80 (B) 100 Participant F19 (AM) 0102030405060708090100Responses 0 5 10 15 20 25 30 FI 100 FI 10080(B) 100 97 95 90 84 80 80 (B) 100Minutes Participant F20 (SC)Minutes 0102030405060708090100Responses 0 10 20 30 40 50 60 70 FI 100FI 100 80(B) 97 95 90 84 80 80 (B) 100 100 Participant F21 (DW)Minutes 0102030405060708090100Responses 0 50 100 150 200 250 300 350 FI100 FI 10080(B) 100 979590848080 (B) 100 Participant F22Minutes 0102030405060708090100Responses 0 100 200 300 400 500 FI 100 FI 10080(B)10097 9590 84 80 80 (B) 100 Participant F28Minutes 0102030405060708090100Responses 0 20 40 60 80 100 120 140 160 180 200 FI 100FI 100 80(B)100 9795 90 84 80 80 (B) 100
46 CHAPTER 7 GENERAL METHODS FOR EX PERIMENTS V AND VI Participants and Setting All participants were undergraduate students enrolled in an Introductory Psychology course at the Univer sity of Florida. Students r eceived course credit for completing the experiment, but this credit was not dependent on pe rformance during the experimental sessions. Each experimental session involved only one participant. Students registered for participation based on minimal information provided on a website, which included only the session times available, the amount of time required for completion of the study, the location of the st udy, and the number a ssigned to the study by the Institutional Review Board. The durat ion of participation was 2 hrs for each participant. All sessions were conducted in a laboratory r oom located in the Psychology Building, which was equipped with a com puter desk, a computer, and a chair. Method When students arrived at their appointed times, they were asked to read and sign an informed consent which stated that they were being asked to participate in an experiment on the effects of different continge ncies of reinforcement. Participants were also given the following instructions: Here is the informed consent, but I will tell y ou what it says in fewer words. This is an experiment on how people learn to respond under different contingencies of reinforcement. In other words, we are studying how people learn to work for rewards. If you agree to participate, you will be seated at a back room and will use only the mouse to earn as many points as possible during the two-hour session. Rather than have you sit there for two hours straight, you will be able to take a break midway through. If you would like to participate, please read over the consent and sign on the second page. Do you have any questions?
47 If participants signed the consent, they we re asked to leave all electronic devices, including watches, cell phones, pe rsonal digital assistants (PDAs ), and pagers in the outer room of the experimental area. The purpos e of removing these de vices was to prevent participants from using these devices instead of attending to the experiment. Participants were then escorted to the room that se rved as the primary experimental area. Participants were asked to take a seat at the desk, and were given the following instructions: This is where you will be working. Remember to use only the mouse to earn as many points as possible. The instructions you see on the screen [â€œclick ok to beginâ€] are the only ones that you will get: it is up to you to figure out how to earn points. At the end of the first session, a thank you message will appe ar on the screen. When it appears, please come out and get me. You can then take yo ur break while I reset the computer. Good luck!! When the participants clicked â€œOK,â€ th e program began running, and the programmed schedules of reinforcement took effect. The computer screen was blank except for one red circle and one black circle . The circles were 0.5in in di ameter and moved at a speed of either 25mm per second (Experiment V) or 50mm per second (Experiment VI) in random directions. Programmed schedules of reinforcement differed for each of the circles. We arbitrarily defined clicking on the black circle as engaging in â€œproblem behaviorâ€ and clicking on the red circle as engaging in â€œappropriate behavior.â€ Participants earned points by clicking th e red or black circles, according to the reinforcement schedules in effect. During baseline, clicking on the black circle was reinforced on an FR1 schedule and clicking on the red circle was not reinforced (EXT). During DRA implemented with full integr ity, clicking on the black circle was not reinforced (EXT) and clicking on the red ci rcle was reinforced on an FR1 schedule. These baseline and treatment schedules were chosen because they are commonly used in applied research on DRA. During treatment inte grity failures, the values of the schedules
48 were altered so that one or both of the available responses were reinforced on a random ratio (RR) schedule. Although the use of RR schedules is only one possible type of integrity failure, it seemed possible that care givers may reinforce be havior on ratio-like schedules, and the use of ratio -based schedules was consiste nt with prior research on treatment integrity failures during DRA schedules (Vollmer, Roane et al., 1999; Worsdell et al., 2000). Three types of integrity failu res were examined: errors of commission in isolation, errors of omission in isolation, and combin ed errors of commission and omission. Like Experiments I through IV, errors of commi ssion involved the intermittent delivery of points following clicks on the black circle (â€œproblem beha viorâ€). Errors of omission involved the intermittent failure to deliver a point following a click on the red circle (â€œappropriate behaviorâ€). During the combined errors phases, participants sometimes earned a point following problem behavior and not earn a point following appropriate behavior. As in Experiments I through IV, participan ts were instructed to earn as many points as possible, but points were not exch angeable for any back-up reinforcers. Clear reinforcement effects and appropriate ch anges in rates of responding following contingency changes between baseline (F R1/EXT schedule) and treatment (DRA schedule with perfect integrity, EXT/FR1) were required for a participant to be included in the experiments. Participants who did not show differentiation in response rates between baseline and DRA phases were exclude d of 27 participants failed to meet the inclusion criteria.
49 At the end of the 2-hr participation time, participants were granted their course credits and were asked if they had any questi ons about the study, or if they wanted a copy of the informed consent.
50 CHAPTER 8 EFFECTS OF TREATMENT INTEGRIT Y FAILURES ON DRA TREATMENTS Procedure The purpose of Experiment V was to exam ine the effects of different types and levels of treatment integrity failures on DRA treatments. Specifically, we examined errors of omission in isolati on, errors of commission in isol ation, and combined errors of omission and commission across a wide range of integrity values (80% integrity to 20% integrity). Seventeen undergraduate students agreed to participate in Experiment V. Five participants withdrew midway through the experiment, co mplaining of boredom. Data from an additional 3 participants were excl uded because of comput er errors during the experiment. Therefore, data from 9 remain ing participants were included. Students participated for a total of 123 min, which in cluded a 2, 60-min blocks in the experiment with a 3-min break between blocks. Participants were randomly assigned to one of three groups. Condition sequence remained constant, regardless of group assi gnment. During the first experimental block, participants were exposed to baseline, DR A at full integrity, followed by 4 phases of reduced treatment integrity. E ach participant was exposed to four levels of treatment integrity (80%, 60%, 40%, and 20%), but the ty pe of integrity mani pulation varied across groups. Participants in Group I were exposed only to errors of omission; for these participants, reduced levels of treatment integrity meant that the FR1 schedule for appropriate behavior became thinner. As trea tment integrity levels became lower, these
51 participants could earn fewer reinforcers fo r engaging in appropriate behavior, but could never earn points for engaging in problem behavior. Participants in Group II were exposed to only errors of commission; for th ese participants, reduced levels of treatment integrity meant that some reinforcers were introduced into the EXT schedule for problem behavior. As the treatment integrity levels became lower, the reinforcement schedule for problem behavior became richer , but appropriate be havior was always reinforced on an FR1 schedule. Participants in Group III were exposed to combined e rrors of omission and commission. For these participants, the level of treatment integrity for problem behavior and appropriate behavior were positively correl ated, such that as fewer reinforcers were available for problem behavior (fewer erro rs of commission), more reinforcers were available for appropriate behavior (fewer errors of omission). Results and Discussion The results for participants in Group I are shown in Figure 12. In this and all subsequent figures in Experiment V, re sponses are shown along the y-axis and consecutive minutes in the session are shown al ong the x-axis. The closed circles depict response rate on the black circle (problem behavior), and the open circles depict response rate on the red circle (appr opriate behavior). The numeri cal condition labels show the percent of treatment integrity for appropria te behavior over the percent of treatment integrity for problem behavior. For exampl e, the 60/100 label in Figure 12 denotes a condition in which 60% treatme nt integrity was in place for the FR1 component of the DRA and 100% integrity was in place for the EXT component. In this case, 60% of appropriate responses were reinforced, but no instances of problem behavior were reinforced.
52 Figure 12: Results for Group I, who were exposed to errors of omission only Results for Group I were highly consistent . For all 3 particip ants in this group, rates of problem behavior were high and rate s of appropriate beha vior were low during the baseline phases. When DRA was implemen ted with perfect integrity, participants engaged in high rates of appr opriate behavior and low rate s of problem behavior. During treatment integrity failure phases, participants generally engaged in appropriate behavior at somewhat lower rates as th e treatment inte grity decreased during the first exposure to failure phases. In general, rates of ap propriate responding remained high during the second exposure to failure phases, with the exception of the final failure phase of the Participant 1DRAP01Minutes 0102030405060708090100110120Responses 0 50 100 150 200 250 BL DRA80/10060/100 40/10020/100BL DRA 20/100 40/10060/10080/100 Participant 1DRAP13Minutes 0102030405060708090100110120Responses 0 50 100 150 200 250 BL DRA80/10060/100 40/10020/100BL DRA 20/100 40/100 60/10080/100 Participant 1DRAP26Minutes 0102030405060708090100110120Responses 0 50 100 150 200 BL DRA80/100 80/100 60/100 60/100 40/10040/100 20/10020/100BL DRA
53 experiment (80/100). It is possible that the decrement observed in appropriate responding just before the break and at the end of the experiment was related to participant fatigue instead of the treatment integrity level. This possibility should be explored in future research by providing participants with more frequent breaks dur ing the experimental session or changing the task. Regardless of th e level of erro rs of omission in place, participants in Group I did not engage in elev ated rates of problem behavior. This finding is consistent with those obtai ned in Experiments I through IV, and suggests that errors of omission in isolation may not be high ly detrimental to DRA treatments. Results for participants in Group II are s hown in Figure 13. For these participants, errors of commission did not become detrimen tal until the level of treatment integrity dropped to 40% or lower. For participants 2DRAP06 and 2DRAP22, high rates of appropriate behavior and low rates of problem behavior were observed during DRA at full integrity, and during 80% and 60% treatmen t integrity. However, rates of appropriate behavior decreased, or rates of problem beha vior increased, at 40% and 20% integrity. The increase in problem behavior during 40% integrity and 20% integrity is somewhat surprising. Because participants could consistently earn a point for each appropriate response, it seems counterintuitive that pa rticipants would allo cate responding to a thinner reinforcement schedule (the treatment integrity failures for problem behavior). Overall, however, the DRA treatment seem ed relatively robust when errors of commission occurred at le ss than 60% integrity. Errors of commission at 40% and 20% inte grity were also detrimental during the second exposure to the integrity failure phase s with participant 2D RAP05. However, this participant engaged in elevat ed rates of problem during th e first exposures to 60% and
54 80% integrity failure. It is possible that the sequence of conditions (l ow integrity to high integrity versus high integrity to low integrity) could have influenced these results. This possibility is supported by the data from E xperiment IV, which showed that gradually decreasing the treatment integrity level attenuated the disruptive effects of errors of commission during DRO treatments. The possi bility that respondi ng during treatment integrity failures is affected by condition se quence was further examined in Experiment VI. Figure 13: Results for Group II, who e xperienced only errors of commission Results from participants in Group III are shown in Figure 14. For these participants, errors of omission and comm ission co-varied. For example, during the conditions labeled 20/20, there was 20% inte grity on both the FR1 and EXT components Participant 2DRAP05Minutes 102030405060708090100110120Responses 0 50 100 150 200 250 BL DRA100/20 100/40 100/60 100/80 100/80 100/60 100/40 100/20BL DRA Partcipant 2DRAP06Minutes 102030405060708090100110120Responses 0 50 100 150 200 BL DRA100/20 100/40 100/60 100/80 100/80 100/60 100/40 100/20BL DRA Participant 2DRAP22Minutes 20406080100120Responses 0 50 100 150 200 250 BL DRA100/20 100/40 100/60 100/80 100/80 100/60100/40 100/20BL DRA
55 of the DRA; this resulted in a 20% chance th at appropriate behavior would result in a point and an 80% chance that problem beha vior would result in a point. Results were consistent across the 3 participants in Group III. Participants engaged in high rates of problem behavior and low rates of appropria te behavior when treatment integrity was 20% or 40%. However, once treatment inte grity reached 60%, response allocations switched, with participants engaging in more appropriate behavior than problem behavior. These results did not seem to be highly affected by the sequence in which the conditions were conducted. Figure 14: Results for Group III, who experien ced both errors of commission and errors of omission Participant 3DRAP07Minutes 102030405060708090100110120Responses 0 50 100 150 200 250 BL DRA80/80 60/60 40/40 20/20BL DRA20/20 40/40 60/60 80/80 Participant 3DRAP19Minutes 0102030405060708090100110120Responses 0 50 100 150 200 250 300 BLDRA 20/20 40/40 60/60 80/80 BLDRA80/80 60/60 40/40 20/20 Participant 3DRAP24Minutes 20406080100120Responses 0 50 100 150 200 250 300 BL DRA20/20 40/40 60/60 80/80 80/80 60/6040/40 20/20BL DRA
56 The results obtained from Group III could be explained based on the reinforcement rate available from each res ponse type. When problem behavior was more likely to result in reinforcement than was a ppropriate behavior (during the 20% and 40% integrity phases), participants engaged in more problem behavior than appropriate behavior. When appropriate beha vior was more likely to result in reinforcement than was problem behavior (during the 60% and 80% in tegrity phases), participants were more likely to engage in appropriate be havior than problem behavior. The results from Experiment V demonstr ate that DRA treatments are susceptible to failure based on different kinds of treatmen t integrity failures. Similar to the results obtained with DRO treatments in Experiments I and II, errors of omi ssion in isolation did not seem to be highly detrimental to the DRA treatment. Errors of commission had a greater impact on responding than did errors of omission, but only at re latively low levels of treatment integrity (20% and 40% integrity ). This can be contrasted with the highly detrimental effect that errors of commissi on had on DRO treatments in Experiments I and II, even at high levels of treatment integrit y. The primary reason for these differences in results may be differences in the treatments themselves. During DRO, participants could maximize the number of points that they ear ned by engaging in high rates of problem behavior, even when the overall treatment inte grity level was as high as 90%. In contrast, participants experiencing a DRA treatment with errors of commission only can maximize points by continuing to respond on the rich er reinforcement scheduleâ€”the FR1 for alternative behavior. Thus, DRA is a more â€œrobustâ€ treatment than DRO overall, as generally lower levels of integrity c ontinue to support treatment effects.
57 In addition to extending Experiments I through IV by examining a different type of treatment (DRA instead of DRO), Experi ment V also extended prior research by examining a mixture of errors of omissi on and commission. Of the three types of treatment integrity examined in this expe riment, combined errors of commission and omission were most detrimental to the treatment. This may be due to the change in responding required for maximization of point earnings as the degree of treatment integrity changed. As mentioned previousl y, participants in Group III allocated the majority of their responses to the richer reinforcement schedule. This result may have important implications for application because caregivers probably make both errors of commission and omission when attempting to implement DRA treatments. Particularly, caregivers with a long histor y of reinforcing problem behavior may be prone to reinforcing problem behavior on a relatively rich schedule an d appropriate behavior on a relatively thin schedule, similar to the lower levels of treatment integrity experienced by Group III. If this is the case, those caregive rs may see little to no improvement in the behavior, despite the attempt to implement a treatment. The results from Group III also suggest some directions for additional research. For example, rates of problem behavior d ecreased and appropriate behavior increased when the level of treatment integrity changed from 40% to 60%. As discussed previously, this change in allocation makes sense in th e context of the reinforcement schedules in place. However, it leaves the question: What would happen if both components of the DRA treatment were implemented at 50% integr ity? If this were the case, there would be no difference between the programmed reinfo rcement rates for engaging in problem behavior versus appropriate behavior. For this reason, 50% integrity with both errors of
58 commission and omission may provide an ideal error schedule on which to assess possible sequence effects during treatment integr ity failures. Therefore, the possibility of sequence effects during 50% integrity condi tions was assessed during Experiments VI through VIII.
59 CHAPTER 9 EFFECTS OF CONDITION SEQUENCE ON RESPONDING DURING DRA TREATMENT INTEGRITY FAILURES : HUMAN OPERANT WITH BRIEF EXPOSURES Procedure The purpose of Experiment VI was to examine the possible sequence effects on responding during treatment integrity failure s during DRA treatments. To do this, we compared responding on a 50% integrity c ondition with errors of commission and omission, following both baseline and full treatment conditions. Eighteen undergraduate stude nts agreed to participate in Experiment VI. Students participated for a total of 123 min, which in cluded 2, 60-min blocks in the experiment with a 3-min break between blocks. Five participants were excluded because clear differentiation between baseline and DRA were not obtained or were not replicated. Each participant was exposed to baseline and DRA conditions similar to those described in Experiment V. During baseline conditions, e ngaging in problem behavior (clicking on the black circle) was reinforced on an FR1 sche dule and engaging in appropriate behavior (clicking on the red circle) was on EXT. Duri ng DRA, these contingencies were reversed, such that appropriate behavior was reinfor ced on an FR1 and problem behavior was on EXT. Only one value of treatment integrity failure was used during Experiment VI: 50% integrity with errors of omission and co mmission. During this phase, there was a 0.5 probability that problem behavior or appropriate behavior woul d be reinforced. This level and type of integrity failure was chosen ba sed on the results of Experiment V, which suggested that a failure condition with equa l probabilities of errors may be more
60 susceptible to sequence effects than error c onditions that favored problem behavior or appropriate behavior. Results and Discussion The results of Experiment VI are show n in Figures 15 and 16. In both figures, responses are shown along the y-axis and c onsecutive minutes in the session are shown along the x-axis. The closed circles show rate of problem behavior, and the open squares show rate of appropriate behavior. The 50/50 condition label denotes the treatment integrity phases. In general, two patter ns of responding were observed for the 13 participants: responding during th e error phase carried over from the previous phase, or responding â€œswitchedâ€ from the allocation in the previous phase during the error phase. Figure 15 shows the results for participants whose response allocati ons during baseline and DRA carried over into the treatment inte grity phases. All 5 of these participants engaged in higher rates of problem behavi or during the 50/50 conditions that followed baseline than the 50/50 condi tions that followed DRA. These results highlight the possibility that treatment integrity failures, particularly those that do not clearly favor one response ove r another, may be particularly detrimental if implemented following baseline. For example, it is possible that the reinforcement contingencies similar to baseline may be in place before a caregiver is trained to implement DRA. If the caregiver implemen ts the DRA with 50% integrity following training, there may be no improvement in th e problem behavior. Alternatively, if the caregiver can implement the treatment with a high level of integrity, at least initially, later treatment integrity failures may not be as detr imental. These sequence effects suggest that behavior analysts may want to focus traini ng efforts on both the tr aining itself and the
61 initial implementation of the DRA. Once th e DRA has been successful, close monitoring (and high treatment integrity) may be unnecessary. Figure 15: Participants showing car ryover from the previous phase The remaining 8 participants in Experime nt VI did not show consistent carryover during the 50/50 conditions. Results from th ese participants are shown in Figure 16. Participant R-V07Minutes 0102030405060708090100110120Responses 0 20 40 60 80 100 BLDRA 50/50 BL BL BL DRA DRA DRA 50/50 50/50 50/50 Participant RV 11Minutes 0102030405060708090100110120Responses 0 10 20 30 40 50 60 BL DRA50/50 BL BL BL DRA DRA DRA 50/5050/50 50/50 RV15Minutes 0102030405060708090100110120Responses 0 10 20 30 40 50 60 70 BL DRA 50/50 BL BLBL DRADRA DRA 50/50 50/50 50/50 Participant R-V07Minutes 0102030405060708090100110120Responses 0 20 40 60 80 100 BLDRA 50/50 BL BL BL DRA DRA DRA 50/50 50/50 50/50 Participant RV12Minutes 0102030405060708090100110120Responses 0 20 40 60 80 100 120 140 160 BL DRA50/50 BL BL BL DRA DRA DRA 50/5050/50 50/50 Participant RV17Minutes 0102030405060708090100110120Responses 0 20 40 60 80 BL DRA 50/50 BL BLBL DRADRA DRA 50/50 50/50 50/50
62 Carryover was observed in some phases for thes e participants, and seemed particularly likely to occur in the first exposure to the 50/50 condition following DRA (in which carryover was observed for 7 of 8 participants ). However, as the experiment progressed, these participants responded in an un expected manner during the 50/50 conditions: response allocation completely switched from the previous phase. In other words, participants would engage in primarily ap propriate behavior in 50/50 phases that followed baseline, and primarily problem be havior in 50/50 phases that followed DRA. This change in allocation was observed for the second, third, and fourth exposures to 50/50 for RV01 and RV06, the first and fourth exposures for RV19, the second and third exposures for RV05, the second and fourth e xposures for RV13 and RV20, and the fourth exposure for RV03 and RV04. In almost all of these conditions, the participant engaged in high rates of one response and low rates of the other response throughout the phase. These abrupt changes in allocation may have been due to the reinforcement schedules that were used during the baseline and DR A conditions. As mentioned previously, we chose to use FR1 and EXT sc hedules in these conditions because those schedules were most commonly used in prev ious DRA research. However, it is possible that the continuous reinforcement schedule exerted enough control over responding that the introduction of the 50/50 condition signaled participants to change responding. For example, it could be that, once the cond ition changed from DRA to 50/50, the first unreinforced appropriate response resulted in the participant switching to problem behavior. Alternatively, it is possible that part icipants were engaging in very low rates of responding on the response alternative that wa s not currently being reinforced, and that the first reinforcer for this re sponse resulted in a rapid change in response allocation. For
63 example, participants may click on the black circle (problem behavior) just a few times per minute, even during DRA conditions. When one of those responses is reinforced, the participant begins engaging in primarily prob lem behavior (clicking the black circle). The switching of response allocation observed in this experiment may also be due in part to the experimental arrangement it self. Because the ideal response pattern was alternation between problem behavior and appr opriate behavior every 10min for the first four phases, it is possible that participants generated a ru le about responding, such as â€œswitch every 10min.â€ Although th e participants did not have any exteroceptive means of timing the conditions, it is possible that th e passage of time somehow exerted control over responding. We conducted Experiment VII to address the limitations associated with the fixed condition duration a nd the general limited exposur e to the error conditions.
64 Figure 16: Participants showing some switching from the previous phase Participant R-V01Minutes 0102030405060708090100110120Responses 0 20 40 60 80 100 120 140 160 BL DRA 50/50 BL DRA 50/50 50/5050/50 BL BL DRA DRA Participant R-V03Minutes 01020304050607080901001101 2 Responses 0 20 40 60 80 100 BL DRA BLBL BL DRA DRA DRA 50/5050/5050/50 50/50 Participant R-V04Minutes 0102030405060708090100110120Responses 0 20 40 60 80 100 BLDRA 50/50 BL BL BL DRADRA DRA 50/5050/50 50/50 Participant R-V05Minutes 0102030405060708090100110120Responses 0 20 40 60 80 BLDRA 50/50 BL BL DRA DRA DRA 50/50 50/50 50/50 BL Participant R-V07Minutes 0102030405060708090100110120Responses 0 20 40 60 80 100 120 140 160 180 200 BL DRA 50/50 BL BLBL DRA DRA DRA 50/5050/5050/50 Participant RV20Minutes 0102030405060708090100110120Responses 0 20 40 60 80 100 120 BLDRA BLDRA50/50 BL50/50DRA50/50 BL 50/50DRA Participant RV13Minutes 0102030405060708090100110120Responses 0 10 20 30 40 BL DRA 50/50 BLBLBL DRA DRA DRA 50/5050/50 50/50
65 CHAPTER 10 EFFECTS OF CONDITION SEQUENCE ON RESPONDING DURING DRA TREATMENT INTEGRITY FAILURES: HUMAN OPERANT WITH EXTENDED EXPOSURES Purpose The purpose of Experiment VII was to pr ovide participants with more extended exposures to the baseline, DRA, and 50/ 50 conditions. Extended exposure may help determine the effects of treatment integrity failures and condition sequence over lengthier periods of time. Additionally, extending the exposure to conditions allowed the assessment of sequence effects on stable rates of responding. Participants and Setting Three adults, ranging in age from 19 to 22, participated in Experiment VII. Participants were recruited through flyers placed around campus and in the community. In contrast to the previous experiments, participants were required to attend multiple appointments. During the first appointment, an experimenter briefly explained the procedures and approximate timeline of the study. Additionally, participants were told that they would earn money based on their poi nt earnings in the expe riment. Participants were paid either a minimum of $5 per hour, regardless of their point totals, or $1 for every 1000 points earned, whicheve r was greater. Part icipants could ea rn a maximum of $15 per hr based on their point totals. Appoi ntments ranged in duration from 30min to 3hrs. Each experimental session involved only one participant. All sessions were conducted in a laboratory room located in the Psychology Building, which was equipped with a computer desk, a computer, and a chair.
66 Method During their first appointment, particip ants were asked to read and sign an informed consent, which explained that the purpose of the study was to examine the effects of treatment integrity on common behavi oral treatments. Participants were also informed that the study may take from severa l weeks to several months to complete, and that they should be prepared for extended participation and multiple appointments. Payment structure was also reviewed during the first session; participants were paid by check every Monday for the sessions comp leted the preceding week, and generally picked up their checks at the start of their first appointment for the week. After participants read and signed the informed consent, they were asked to sit at the computer, and were read the following instructions: This is where you will be working. Remember to use only the mouse to earn as many points as you can. The more points you earn, the more you will be paid. Your points are displayed at the bottom of the screen as the score. Good luck! Each session was 10min in duration. At th e end of each 10-min session, the participant notified the experimenter, who reset the co mputer while the participant took a 1-min break. The computer program was identical to the one used in Experiments V and VI. During the sessions, a red circle and a black circle moved in random directions around the computer screen. Both circles were 1in in diameter and moved at a speed of 25 mm/s. Clicking on the black circle was arbitrarily defined as engagi ng in problem behavior, and clicking on the red circle was arbitrarily defi ned as engaging in appr opriate behavior. The conditions used in this study were identical to those used in Expe riment VI: baseline, DRA, and 50/50. During baseline conditions , problem behavior was reinforced on an FR1 schedule and appropriate behavior was on EXT. During DRA, appropriate behavior
67 was reinforced on an FR1 schedule and probl em behavior was on EXT. During the 50/50 condition, there was a 0.5 probability of rein forcement for both problem behavior and appropriate behavior. The 50/ 50 condition followed baseline twice and DRA twice to assess possible sequence effects using a reversal design. Each condition was conducted until respondi ng stabilized, or until 30 sessions passed and responding was not stabilizing. St ability was assessed using Kappa, which detects the degree of proportional change in responding (Johnston & Pennypacker, 1980). Kappa was chosen as the measure of stabilit y because it can be used for both high-rate and low-rate responses. When the Kappa value ranged between 0.95 and 1.05, responding was considered stable (perfect stability yields a Kappa value of 1.0). For Ryko and Rojer, Kappa values were initially calculated based on the minute-to-minute response rates during the last 5min of each session. When responding was stab le during the last 5min of a session, the next condition wa s started. However, this measure did not seem to be capturing stable responding, as considerable variability was still evident on a sessionwide basis. Therefore, star ting with sessions 14 for Ryko and Rojer, Kappa values were calculated based on responding during the thr ee most recent sessions for each condition. Kappa values for Regina were calculated base d on the last three sessions throughout the experiment. Although the experimental design was orig inally a reversal design, this was impossible for Regina, who could not complete the experiment due to a move out of the area, which occurred near the completion of the 50/50 condition that followed baseline. Her departure also resulted in termination of that 50/50 condition before the designated 30 sessions required for a phase. Additionall y, an extra 50/50 condition following DRA
68 was conducted for Rojer (the final exposure to 50/50) because the prior 50/50 following DRA (sessions 116-121) was terminated befo re stability was attained due to an experimenter error. However, only 5 sessi ons were conducted in the extra DRA and 50/50 phases because of time constraints. Results and Discussion The results of Experiment VII are shown in Figures 17 and 18. In both figures, clicks per minute are shown along the y-ax is, and sessions are along the x-axis. The closed symbols denote the rate of problem be havior (clicking the bl ack circle), and the open symbols denote the rate of appropriate behavior (clicking th e red circle). All 3 participants engaged in high rates of probl em behavior and low rates of appropriate behavior during baseline phases. As expecte d, this pattern of responding was reversed, with high rates of appropriate behavior a nd low rates of problem behavior, when DRA was implemented with full integrity. Responding became more variable during the 50/50 phases. During most 50/50 sessions , participants allocated re sponding primarily to one of the two available circles, instead of allocati ng about half of the responses to each circle. This pattern of responding was similar to th e allocation of responding to either problem behavior or appropriate behavior observed during Experi ment VI. Another similarity to the results of Experiment VI is that most participants continued to respond on whichever response was reinforced first during the sessi on. This resulted in a modest degree of carryover from the previous baseline or DRA phase in the beginning of the 50/50 conditions. To better show the ini tial carryover in responding during the 50/50 condition, Figure 18 shows the rate of responding during th e last five baseline or DRA sessions in a phase, as well as the first five 50/50 sessions in each treatmen t integrity failure phase, for
69 all participants. For Ryko and Regina, rates of appropriate behavior we re initially greater than rates of problem behavior in both 50/ 50 conditions that followed DRA implemented with perfect integrity. Additionally, overall ra tes of problem behavior were greater than rates of appropriate behavior in the initial 50/50 sessions following baseline. For Rojer, response rates during the 50/50 condition carried over from the previous phase in two of the exposures to 50/50. In the first 50/50 session following DRA (session 32), however, rates of problem behavior were greater than rates of appropriate behavior. This may be due to the reinforcers provide d during the initial 50/50 session. Examination of the first few seconds of session 32 showed that al though the first responses were appropriate behavior, the first reinforcer followed problem behavior. A similar pattern of responding and reinforcer delivery was observed in the final replication of 50/50 following DRA (starting with session 184). Overall, these data may have important implications for clinical practice, as they suggest that the contingencies in place before a treatment is implemented with errors may affect responding during the treatment in tegrity failure. As shown in previous experiments, treatment integrity failures seem less detrimental if the participant has an immediate history with the treatment impl emented without errors. Rojerâ€™s responding during the first 50/50 phase also has impor tant implications, as it suggests that reinforcement schedules can exert consider able control over responding. During this phase, it seems as if Rojer allocated most of his responding to the first response type that produced a point in any given session. This suggests that individuals may revert to engaging in problem behavior, even fo llowing an immediate history with DRA
70 implemented without failures, if the appropr iate response fails to produce reinforcers initially. Like Experiment VI, the results of this experiment underscore the importance of monitoring caregivers to ensure high levels of integrity during the initial implementation of DRA. Implementing the treatment with an initial high level of integrity may reduce the harmful effects of later treatment integrity fa ilures. However, it is important to note that the 50/50 condition was detrimental to the DRA treatment even when that condition followed DRA. None of the participants engage d in as much appropria te behavior, or as little problem behavior, duri ng any 50/50 condition than during the DRA conditions. This suggests that the 50/50 level of treatment in tegrity failures is somewhat detrimental to DRA treatments regardless of the preceding condition. One limitation of this experiment, and the preceding six experiments, is that they were conducted in a highly controlled laborat ory setting, using an analog task with a nonclinical population. Therefore, it is unclea r whether the results of these experiments would be replicated with the populations who typically receive behavi oral treatments. To address this limitation, Experiment VIII partia lly replicated Experiment VII in a school setting, with an adolescent boy referred for the assessment and trea tment of aggression.
71 Figure 17: Session-by-session response rates fo r all 3 participants in Experiment VII Ryko Sessions 102030405060708090100Clicks Per Minute 0 50 100 150 200 250 300 350 BL DRA 50/50 BL 50/50DRA50/50 BL50/50 ReginaSessions 102030405060708090100Clicks Per Minute 0 50 100 150 200 250 300 BL DRA 50/50 BL 50/50 RojerSessions 020406080100120140160180200Clicks Per Minute 0 100 200 300 400 500 600 700 BLDRA50/50 BL BL 50/5050/50 50/5050/50DRADRA
72 Figure 18: Last five baseline or DRA sessions in a phase and first fi ve 50/50 sessions in a phase ReginaSessions 246810121416182022242628Response Rate 0 50 100 150 200 250 300 BL DRA 50/50 BL 50/50 RojerSessions 51015202530354045505560Response Rate 0 100 200 300 400 500 600 700 BL DRA 50/50 BL 50/50 DRA 50/50 BL 50/50 DRA 50/50 Ryko Sessions 510152025303540Response Rate 0 50 100 150 200 250 300 350 DRA 50/50 50/50DRA 50/50 BL50/50 BL BL
73 CHAPTER 11 EFFECTS OF CONDITION SEQUENCE ON RESPONDING DURING DRA TREATMENT INTEGRITY FAILUR ES: APPLIED REPLICATION Purpose The purpose of Experiment VIII was to partially replicate the results of Experiments VI and VII in a school setting with a clinical participant population. Thus, this experiment was designed as one means of validating the results from the human operant laboratory. Methods Participant and Setting One male adolescent with disabilities (â€œJa keâ€) participated in Experiment VIII. Jake was classified as trainable mentally handicapped (TMH), and was enrolled in a center school for children with disabilities. He was referr ed to a school-based treatment program for the assessment and treatment of aggression, which consisted of hitting others on the arm, and grabbing the brea sts or abdomens of females. All sessions were conducted in an empty classroom in Jakeâ€™s school. The classroom was equipped with tables, chairs, a nd leisure items appropria te for a variety of different ages and skill levels. Data Collection and Interobserver Agreement Data for all sessions were collected using handheld computers and a program designed for real-time data collection. Ob servers were undergraduates or graduate students in behavior analysis, who had previously attained interobserver agreement (IOA)
74 of 90% or above for three consecutive session s. Observers were typi cally seated in the corner of the classroom, about 5m away from Jake and the therapists. Data were collected on both the studentâ€™ s and therapistâ€™s behavior. Aggression was defined as hitting (physic al contact between Jakeâ€™s open hand and the therapistâ€™s body) and grabbing (Jakeâ€™s fingers closing on any part of the therapistâ€™s body or clothing). Jakeâ€™s appropriat e alternative response was a vocal greeting, saying â€œhi.â€ Because this alternative behavior occurred at very low levels before the start of the study, no new appropriate alternative responses were shaped for the purposes of this experiment. Both aggression and greetings were scored as frequencies. Therapist responses included attention delivery (therapist looked at Jake and made a vocal statement) and escape delivery (therapist st ated â€œtake a break,â€ removed instructional materials, and turned away from Jake). During 32% of functional analysis sessi ons and 44.6% of treatment evaluation sessions, a second observer simultaneously and independently collected data. Interobserver agreement was calculated by dividing each session into 10-s bins and comparing the number of responses scored within each bin across observers by dividing the smaller number of responses in that bin by the larger numbe r of responses and multiplying by 100, yielding a percentage of I OA for that bin (Shirley et al., 1997). The percentages were then averaged across all bins in the session to yield an overall IOA percentage. During the functi onal analysis, the average I OA score was 98.5% (range, 75.3-100%) for aggression, 100% for gree tings, and 97.5% (range, 80.8-100%) for therapist behavior. During the treatment evaluation, the average IOA score was 93.3%
75 (range, 60.4-100%) for aggression, 91.6% (ra nge, 64.5-100%) for greetings, and 84% (range, 72-95%) for therapist behavior. Data were also collected on treatment integrity. Treatment integrity was defined as the degree to which the therapist impl emented the contingencies as designed. During the 50/50 phases, in which treatment integr ity was deliberately manipulated, integrity scores were based on the randomized comput er output, as described below. Treatment integrity for all sessions was 100%, except fo r session 13, during which one instance of problem behavior was reinforced. Functional Analysis A functional analysis was conducted to de termine possible reinforcers for Jakeâ€™s aggression. Initial functional analysis sessions were similar to those described by Iwata, Dorsey, Slifer, Bauman, & Richman, 1994, and consisted of play, attention, and escape conditions. During the play se ssions, Jake had access to leisure items and continuous therapist attention. During attention sessions, Jake had access to le isure items, but the therapist ignored him until he engaged in aggression. Contingent upon aggression, the therapist provided a brief reprimand, such as â€œdonâ€™t hit me, Jake.â€ During escape sessions, the therapist asked Jake to engage in an academic task, such as letter sorting, using a three-prompt instructional sequence (verbal prompt, model, physical guidance, with 5s between prompts). Contingent upon aggr ession, the therapist a llowed Jake to take a 30-s break from the task, signaled by the therapist saying â€œtake a breakâ€ and removing the instructional materials. When low or decreasing rates of proble m behavior were observed in the initial sessions, additional session types were include d. For example, Jakeâ€™s teacher reported that he was most likely to engage in a ggression when she was talking with another
76 person, or during transitions, when people were frequently in close physical proximity to Jake. Based on these reports and classroom observations, three additional functional analysis conditions were included: neutral at tention, proximity, and diverted attention. The neutral attention condition was identical to the attenti on condition described above, except the form of the attention was changed from a reprimand to a neutral statement, such as â€œwhatâ€™s going on, Ja ke?â€ During the proximity cond ition, the therapis t sat within 0.5m of Jake, and moved away from him con tingent on aggression. During the diverted attention condition, two therapists talked with each othe r. Contingent on aggression, one therapist turned to Jake and made a brief ne utral comment, similar to those used in the neutral attention condition. Baseline The baseline conditions were similar to the diverted attention functional analysis condition. Two therapists sat across the table fr om each other, with one therapist seated next to Jake (within 1m). The therapists ta lked with each other and ignored Jake until he engaged in aggression. Contingent on each aggr essive response, one therapist would turn to Jake and made a brief neutral comm ent. Greetings were not reinforced. Differential Reinforcement An evaluation of DRA was conducted af ter the conclusion of the functional analysis. Because Jake had some verbal skills , saying â€œhiâ€ was select ed as his appropriate alternative response. Before the first DRA se ssion, the therapist prompted Jake to greet her by saying, â€œJake, if you want to talk to me, say hi.â€ When Jake said hi, the therapist turned to him and made a brief neutral comm ent. This prompting procedure was repeated a total of ten times before the start of th e initial DRA phase. Prompts were never used after these initial ten trials. During DRA se ssions, two therapists sat talking with each
77 other and ignoring Jake. The second therapist also collected treatment integrity data during all integrity failure sessions. When Jake said â€œhi,â€ one therapist turned to him and make a brief neutral comment, such as â€œw hat are you looking at Ja ke?â€ In contrast, aggressive responses never produced attention. Treatment Integrity Failure (50/50) The 50/50 condition was included to exam ine the effects of sequence effects on responding during treatment integrity failu res. During the 50/50 condition, 50% of aggressive responses and 50% of greetings re sulted in brief therapist attention. During this condition, two therapists talked with each other and ignored Jake. The therapist seated across the table from Jake held a cli pboard with a computer-g enerated sequence of which responses should result in reinforcemen t, and cued the therap ist sitting closest to Jake. Reinforcers consisted of brief, neutral statements, similar to those provided in the baseline and DRA conditions. The 50/50 c ondition followed baseline twice and DRA twice to assess possible sequence effects using a reversal design. Results and Discussion Figure 19 shows the results of Jakeâ€™s f unctional analysis. Sessions are shown along the x-axis and rate of aggression along the y-axis. Although rates of problem behavior were initially elev ated in the play, attention, and escape conditions, rates eventually decreased to n ear-zero levels. Following th e introduction of the three additional test conditions (positive attenti on, proximity, and divert ed attention), Jake engaged in elevated rates of aggression ex clusively in the divert ed attention condition, suggesting that adult attention served as a re inforcer for Jakeâ€™s aggression, particularly when two adults were talking to each other, but ignoring Jake.
78 Figure 19: Functional anal ysis results for Jake Figure 20 shows the results of Jakeâ€™s treatment analysis . Sessions are along the xaxis and responses per minute of aggressi on and greetings are along the y-axis. The closed symbols show the rate of aggressi on and the open symbols show the rate of greetings. The asterisk above session 13 denotes an accidental treatment integrity failure, in which the therapist reinfor ced one instance of aggression. During all baseline phases, rates of aggre ssion occurred at moderate to high rates, and greetings occurred at low or decreasi ng rates. During the first exposure to DRA implemented with full integrity, rates of probl em behavior spiked before reducing to low levels, and rates of greetings increased to moderate, stable levels. During subsequent DRA phases that followed baseline, problem behavior decreased to low rates and appropriate behavior increased to moderate rates. Based on th ese results, it seemed that DRA was an effective treatment for Jakeâ€™s aggression when implemented at full integrity. Jake Functional AnalysisSessions 5101520253035Responses per Minute Aggression (hitting, head butting, & kicking) 0 2 4 6 8 10 12 14 16 18 Play Attention Escape Positive Attention Proximity Diverted Attention
79 During the 50/50 phases that followed DR A, a mixture of problem behavior and appropriate behavior occurred, with some bias toward pr oblem behavior. This result contrasts with previous studies (e.g., Vollmer , Roane et al., 1999), which showed that DRA had relatively robust effects when implem ented at reduced levels of integrity. The bias toward problem behavior observed in the current experiment could be due to Jakeâ€™s extra-experimental history. Although no formal data were collected on the reinforcers available in Jakeâ€™s classroo m just before and after se ssions, therapists reported anecdotally that Jakeâ€™s teacher frequently attended to Jakeâ€™s a ggression shortly after experimental sessions. The results in the 50/50 conditions follo wing DRA can be contrasted with the results of the 50/50 conditions following ba seline. During the 50/50 following baseline, rates of greetings remained low or near ze ro, and rates of aggression remained high and stable. For Jake, treatment integrity failures (at least, the 50/50 condition used in the current experiment) were more detrimental to the treatment when they followed baseline than when they followed full treatment. These results suggest that the initial implementation of the treatment may affect responding during later integrity failures. Specifically, if caregivers initially implement a DRA treatment with a high level of integrity, later integrity failures may not be as detrimental. However, if a caregiver who previously provided a rich reinforcement schedule for problem behavior initially implements a treatment with moderate or poor integrity, problem beha vior is unlikely to improve, as in the 50/50 conditions followi ng baseline in the cu rrent experiment.
80 Figure 20: Treatment analysis for Jake Jake DRA: Treatment IntegritySessions 20406080100120140160180200220240Responses per minute 0 5 10 15 20 25 30 *BLDRADRA 50/50 BL BL 50/50DRABL DRA50/50 BL50/50 BLDRA
81 CHAPTER 12 GENERAL DISCUSSION This series of experiments evaluated two types of differential reinforcement procedures commonly used in the treatmen t of problem behavior, using both human operant and applied preparati ons. The treatment procedures were evaluated both at full treatment integrity and at reduced levels of treatment integrity. Overall, both procedures were highly effective at reduc ing rates of problem behavior when implemented at full integrity. Although many types of possible integr ity failures exist, tw o types of treatment integrity failures were evaluated in these studies. During errors of omission, earned reinforcers were periodically omitted. During errors of commission, problem behavior sometimes resulted in reinforcer delivery. In the current studies, a to tal of only 9 of the 72 part icipants were excluded because of a lack of response differentiati on between baseline and treatment conditions. Most of the excluded participants failed to engage in problem behavior during baseline, so the effects of the treatment were impossibl e to evaluate. Therefore, it is likely that these participants were excluded because of th e relatively weak reinforcer used in the human operant studies (points that were not exchangeable for any back-up reinforcers) instead of inadequacies in the treatment procedures. Overall, the results of all eight experi ments suggest that although differential reinforcement procedures implemented at full integrity might be highly effective treatments, these treatment effects can be co mpromised when treatment integrity failures are made. When treatment integrity failures we re in place, errors of commission were
82 consistently more detrimental than were e rrors of omission, regardless of the type of procedure being evaluated. In other words, it was more detrimental for problem behavior to occasionally result in the delivery of a re inforcer than it was for appropriate behavior (or the non-occurrence of problem behavior ) to occasionally go unreinforced. This outcome is predictable if viewed in light of the actual reinforcement schedules in place during the error phases. Duri ng errors of omission, participants could maximize the number of points that they earned by completely refraining from engaging in problem behavior, which would guarantee loss of a reinfo rcer during the DRO, and never result in a reinforcer during the DRA. Therefore, er rors of omission are akin to a thinned reinforcement schedule. However, during erro rs of commission, part icipants could earn additional points by engaging in problem behavior, making these phases akin to a concurrent schedule. During DRO commission phases, partic ipants could maximize the number of points that they earned by engaging in modera te or high rates of problem behavior, regardless of the treatment integrity level. Errors of co mmission were probably not as detrimental to DRA as DRO b ecause during the DRA commission phases, appropriate behavior was alwa ys reinforced on a richer schedule than was problem behavior. Once more reinforcers were availa ble for problem behavior than appropriate behavior, as during the mixe d omission/commission phases, treatment effects were compromised. The current experiments also demonstrated that the detrimental effects of treatment integrity failures can be attenua ted through the devel opment of specific histories with the participants. For DRO treat ments, the effects of errors of commission were attenuated by providing a history with different stimuli correlated with perfect
83 integrity, and by gradually decreasing the treat ment integrity level. In Experiment III, participants were provided with a history with several different stimuli, all of which were correlated with perfect treatme nt integrity. During this ex periment, rates of problem behavior remained low even when treatment integrity decreased. It is possible that the effects obtained in this experiment were due more to the extended history with perfect integrity, regardless of the stimuli associated with the treatment, than to the history with multiple stimuli. Indeed, it seems that treatment integrity failures are less detrimental once responding is eliminated during the DRO. This possibility was not examined in the current studies, and should be a topic of fu ture research. During Experiment IV, the level of treatment integrity was gradually reduced, wh ich seemed to reduce the effects of errors of commission for several participants, alt hough the results were not as consistent as those obtained in Experiments I through III. Nonetheless, it seemed possible that the sequence in which the conditions occurred co uld impact the degree to which treatment integrity failures were detrimental. When treatment integrity failures were introduced to the DRA treatment, a combination of errors of omission and errors of commission were most detrimental. This was particularly the case when the reinforcem ent contingencies favored problem behavior over appropriate behavior. Prior research on treatment integrity failures in DRA treatments showed that the treatment was ge nerally robust, but that sequence could be a problem (Vollmer, Roane et al., 1999). Based on the results of Expe riment V, described above, it seemed most likely that condition sequence may affect responding during error conditions in which the reinforcement sche dules did not favor problem behavior or appropriate behavior. Therefor e, Experiments VI through VIII evaluated the effects of
84 condition sequence on a treatment integrity failu re in which both problem behavior and appropriate behavior had a 50% chance of resulting in a reinforcer. The results of Experiments VI through VIII showed that cond ition sequence did influence results. In general, treatment integrity failures seemed to have less of an impact on treatment effects when failure phases followed DRA than when they followed baseline. Although unexpected switching of response allocations we re observed in Experiment VI and VII, the changes in responding are probably due to the conti nuous reinforcement schedules used during baseline and DRA at full integr ity. Continuous reinforcement schedules were used in these studies because they are the most common in prior research on DRA, but it is possible that they exerte d strong stimulus control over responding, such that the first unreinforced response resulted in switching to the alternat ive behavior, or the first reinforced alternative behavior resulted in further alternative responding. The likelihood of switching should be examined in future studies using baseline and DRA schedules other than FR1. In addition to the results relating directly to treatment integrity failures, the current studies also have implications fo r the utility of human operant methods in examining problems with applied significance. The human operant procedures used in these studies allowed for the rapid generation of data sets, a high degree of control over relevant variables, and the ab ility to test a wide range of variables without exposing atrisk participants to potentially detrimental experimental manipulati ons, such as reduced levels of treatment integrity. During Experiments I through VI, each individual participantâ€™s data set was completed in just over two hrs, resulting in rapid generation of complete data sets. Although limited by brief ex posures to the contingencies, the results
85 from these experiments can suggest whic h variables are influencing responding, and which are not. The most influential variable s can then be tested using either more extended exposures in the human operant la boratory (as in Experiment VII), or in application (as in Experiment VIII). Another benefit of the human operant prep aration was the high degree of control over the relevant variables, including stimulus presentation of both antecedent stimuli and reinforcers. This allowed for the programmi ng of exact levels of treatment integrity failures. Although programming comparable levels of treatment integrity in application was possible (Experiment VIII), it required tw o therapists to implement the procedures, and procedural integrity wa s still not exactly 100%. Finally, the human operant pr eparation used in these st udies allowed for testing a wide variety of different type s and levels of treatment inte grity failures without exposing at-risk populations to degraded treatments, at least at first. Also, fewer applied replications may be necessary to convincing ly demonstrate effects when a large number of consistent results from the human operant laboratory are available. Experiments VI through VIII provide a model for this type of â€œtranslationalâ€ research methods. In these experiments, similar results were demonstr ated across both huma n operant and applied preparations. However, Experiment VIII contai ned data from only 1 participant referred for the treatment of problem behavior. Data from this participant were consistent with the general pattern of results observed in Experi ments VI and VII, but did not include any examples of the allocation sw itching observed in those experi ments. Therefore, further applied replications of the results in Experi ments I through VII are needed to demonstrate the generality of results obtained in the human operant laboratory.
86 There are several other limitations of the current experiments that should be noted. In particular, there are several limitati ons associated directly with the use of an analog procedure in the human operant laboratory. First, the response chosen, a mouse click, was a simple response that existed in all partic ipantsâ€™ repertoires. Although participants were not specifically instructed to click the mouse, they were told to â€œuse only the mouse to earn as many points as pos sible.â€ The reinforcement contingencies in place, in conjunction with this instruction, freque ntly resulted in very high response rates, which often exceeded 200 responses per min and occasionally exceeded 1000 responses per min. These rates are well above those typically observed when treating problem behavior, and may have influenced the outcome of the human operant studies. Second, the reinforcer used in Experiment s I through VI may have been relatively weak to the participants. In all six experiment s, points alone were used as the reinforcer for responding. Points were not exchangeable for any back-up reinforcers. As mentioned previously, use of a relativel y weak reinforcer may have ultimately resulted in the exclusion of 9 participants from this seri es of studies. Researchers should examine differences in responding when a relatively mo re valuable reinforcer is used. However, this limitation is probably minor given that po ints alone did seem to maintain responding at high rates throughout the expe riment for most participants, and given that similar results were obtained in Experiment VII, in which points were exchangeable for money, and Experiment VIII, which used the functional reinforcer for problem behavior. Third, the duration of exposure to the c ontingencies in Experiments I through VI may have influenced the outcomes of thos e studies. Specifically, participants were typically exposed to each phase for only five to ten min, and phase changes occurred at
87 set points in time, regardless of the pattern or rate of responding. Therefore, phase changes often occurred when behavior was on an upward or downward trend. Also, the rates of responding captured in these expe riments included transition states, as responding may not have stabilized before the change in phase. Researchers should attempt to replicate these results using mo re extended exposures, and conducting phases until responding stabilizes. However, it should be noted again that results of Experiment VII, which did conduct phases to stability, di d not differ dramatically from those obtained in Experiment VI. A final limitation of the human operant procedure was the inclusion of only college students or college-age participants in Experiments I through VII. It is possible that another participant demographic may re spond differently. For example, treatment integrity failures may affect individuals with developmental disabilities or limited verbal skills differently than the high-functioning pa rticipants included in these studies. In particular, it is possible that many of the par ticipants in the current experiments generated some kind of rule about responding. Therefore, it would be important to replicate these results with different populations. Other limitations are based on the specific procedures used in these experiments, independent of the human operant preparati on. For example, the baseline and differential reinforcement schedules were restricted to only one set of parameters. For the DRO treatments, a value of 15s was used for the DR O, while all baseline sessions were FI 15s. For the DRA treatments, FR1 and EXT schedules were used in baseline and treatment. As mentioned previously re garding the DRA treatment, di fferent parameters of the baseline and treatment schedules may have re sulted in different pa tterns of behavior
88 during the treatment integrity failure phases. For example, intermittent baseline and DRA schedules may have reduced or eliminat ed the switching observed during treatment integrity failures. Similarly, different va lues of the baseline and DRO treatment, including either richer or leaner schedules , may have changed responding during errors of commission and omission. Another limitation of the current experime nts is that only two types of possible errors were examined: errors of omission and commission. In these studies, these errors occurred according to probabilistic schedul es, which resembled random ratio schedules. This frequently resulted in very high re inforcement rates during commission phases, which probably contributed to the detrimental effects of those errors. However, probabilistic schedules are onl y one way that errors of omission and commission might occur. These errors could also occur base d on a variety of different other types of schedules, including interval-based schedules . The use of interval-based treatment integrity failures would limit the degree to wh ich participants could maximize reinforcers by responding at high rates during errors of commission. Indeed, results from preliminary studies now being conducted sugge st that interval-based errors are not as detrimental as probabilistic-based errors. S econd, errors of omission and commission are only two types of possible errors that may occur. Other er rors might include differential delays to reinforcement, differential reinforcer magnitude s, or alternating peri ods of full treatment and no treatment. Future research should carefully examine parametric variables associated with errors of omission and commi ssion, as well as different types of treatment integrity failures.
89 Despite the limitations, the results of the current experiments have important implications for application. First, the eff ects of errors on differential reinforcement procedures could be used to inform caregi ver training. Currently, many caregiver training procedures stress the importance of deliveri ng reinforcers following appropriate behavior, or following some specified period of time in which no instances of problem behavior occurred. However, the results of the current experiments suggest that accidentally failing to deliver an earned reinforcer (an error of om ission) is probably not highly detrimental to the overall treatment effects. Caregiver tr aining procedures may instead focus on the importance of the EXT component to differentia l reinforcement proce dures. If caregivers do not implement this component with a high le vel of integrity (in other words, if they make errors of commission), the treatment effects could be degraded. Unfortunately, EXT is not always possible, such as with dangerous behavior reinforced by social consequences. Thus, evaluations of differentia l reinforcement procedures that are aimed at minimizing the reinforcement for problem behavior, rather than eliminating it, are warranted. Second, the results also imply that initial monitoring of caregiverâ€™s implementation of procedures may improve th e chance that treatments later have more robust effects. Monitoring and feedback dur ing initial implementation could have two overall effects on the treatment. One, it could increase the overall levels of integrity with which the caregiver implements the procedur e over time by providing a solid training foundation. Therefore, treatment integrity fail ures may never become an issue because the caregiver consistently implements the tr eatment with a high level of integrity. Two, monitoring may help ensure a high degree of treatment integrity during the initial
90 implementation of the treatment. Results of E xperiments III, IV, VI, and VII suggest that later treatment integrity failures may not be as detrimental to treatment effects if they follow a period of time in which treatment inte grity was consistently high. Unfortunately, this implies that caregiver training procedures may be initially time-intensive to ensure the best long-term outcome. Future research should examine the types and levels of treatment integrity failures made by careg ivers, following a variety of training procedures. In conclusion, the methods used in the current studies seemed useful for studying ethically complicated questions like those rela ted to treatment integrity. Across all eight experiments, reduced levels of treatment inte grity had at least minor detrimental effects on responding, suggesting that the degree to which a treatment is implemented as designed can have substantial influence ove r responding. The current studies may help inform future treatment integrity resear ch, which could focus more on errors of commission and combined errors th an errors of omission alone. Despite the consistent effects of trea tment integrity on responding, few studies have examined the effect of different types and levels of integrity failure on responding. Future research could examine different t ypes of treatment integrity failures, and parametrically manipulate the values of th e failures. Additionally, more research is needed on the effects of treatment integr ity failures on other types of behavioral treatments, such as noncontingent reinforcemen t procedures. Finally, the types and levels of integrity failures occurring in application should be examined through descriptive studies. The values obtained through these desc riptive analyses could then be replicated and manipulated in the laboratory.
91 LIST OF REFERENCES Armstrong, K. J., Ehrhardt, K. E., Cool, R. T., & Poling, A. (1997). Social validity and treatment integrity data: Reporting in articles published in the Journal of Developmental and Physi cal Disabilities, 1991-1995. Journal of Developmental and Physical Disabilities, 9 (4), 359-367. Codding, R. S., Feinberg, A. B., Dunn, E. K., & Pace, G. M. (2005). Effects of immediate performance feedback on implementa tion of behavior support plans. Journal of Applied Behavior Analysis, 38 (2), 205-219. Greene, B. F., Norman, K. R., Searle, M. S ., Daniels, M., & Lubeck, R. C. (1995). Child abuse and neglect by parents with disabilities: A tale of two families. Journal of Applied Behavior Analysis, 28 (4), 417-434. Gresham, F. M. (1989). Assessment of treat ment integrity in school consultation and prereferral intervention. School Psychology Review, 18 (1), 37-50. Gresham, F. M. (2005). Treatment integr ity and therapeutic change: Commentary on Perepletchikova and Kazdin. Clinical Psychology-Sci ence and Practice, 12 (4), 391-394. Gresham, F. M., Gansle, K. A., & Noell, G. H. (1993). Treatment integrity in applied behavior analysis with children. Journal of Applied Behavior Analysis, 26 (2), 257-263. Gresham, F. M., Gansle, K. A., Noell, G. H., Cohen, S., & Rosenblum, S. (1993). Treatment integrity of school-based be havioral intervention studies 1980-1990. School Psychology Review, 22 (2), 254-272. Gresham, F. M., MacMillan, D. L., BeebeFrankenberger, M. E., & Bocian, K. M. (2000). Treatment integrity in learning disabilities inte rvention research: Do we really know how treatments are implemented? Learning Disabilities Research & Practice, 15 (4), 198-205. Hagopian, L. P., Fisher, W. W., & Leg acy, S. M. (1994). Schedule effects of noncontingent reinforcement on attention-maintained destructive behavior in identical quadruplets. Journal of Applied Behavior Analysis, 27 (2), 317-325.
92 Hanley, G. P., Iwata, B. A., & Thompson, R. H. (2001). Reinforcement schedule thinning following treatment with func tional communication training. Journal of Applied Behavior Analysis, 34 (1), 17-38. Hargett, M. Q., & Webster, R. E. (1996). Treatment integrity a nd acceptability with families: A case study of a child with school refusal. Psychology in the Schools, 33 (4), 319-324. Iwata, B. A., Dorsey, M. F., Slifer, K. J., Bauman, K. E., & Richman, G. S. (1994). Toward a functional-analysis of self-injur y (Reprinted from Analysis Intervention in Developmental-Disabilities, Vol 2, Pg 3-20, 1982). Journal of Applied Behavior Analysis, 27 (2), 197-209. Johnston, J. M., & Pennypacker, H. S. (1980). Strategies and tactics of human behavioral research . Hillsdale, NJ: L. Erlbaum Associates. Kelley, M. E., Lerman, D. C., & Van Camp, C. M. (2002). The effects of competing reinforcement schedules on the acqui sition of functi onal communication. Journal of Applied Behavior Analysis, 35 (1), 59-63. Lerman, D. C. (2003). From the laboratory to community applica tion: Translational research in behavior analysis. Journal of Applied Behavior Analysis, 36 (4), 415419. Lindberg, J. S., Iwata, B. A., Kahng, S. W., & DeLeon, I. G. (1999). DRO contingencies: An analysis of variable-momentary schedules. Journal of Applied Behavior Analysis, 32 (2), 123-136. Marcus, B. A., & Vollmer, T. R. (1996). Combining noncontingent reinforcement and differential reinforcement schedules as treatment for aberrant behavior. Journal of Applied Behavior Analysis, 29 (1), 43-51. Mazaleski, J.L, Iwata, B. A., Vollmer, T. R., Zarcone, J. R, & Smith, R. G. (1993). Analysis of the reinforcement and extinction components of DRO contingencies with self-injury. Journal of Applied Behavior Analysis, 26 , 143-156. Noell, G. H., Duhon, G. J., Gatti, S. L., & Connell, J. E. (2002). Consultation, follow-up, and implementation of behavior management interventions in general education. School Psychology Review, 31 (2), 217-234. Noell, G. H., Witt, J. C., LaFleur, L. H., Mortenson, B. P., Ranier, D. D., & LeVelle, J. (2000). Increasing intervention implementa tion in general education following consultation: A comparison of two follow-up strategies. Journal of Applied Behavior Analysis, 33 (3), 271-284.
93 Noell, G. H., Witt, J. C., Slider, N. J., Connell, J. E., Gatti, S. L., Williams, K. L., Koenig, J. L., & Resetar, J. L. (2005). Treatment implementation following behavioral consultation in schools: A co mparison of three follow-up strategies. School Psychology Review, 34 (1), 87-106. Northup, J., George, T., Jones, K., Broussar d, C., & Vollmer, T. R. (1996). A comparison of reinforcer assessment methods: The ut ility of verbal and pictorial choice procedures. Journal of Applied Behavior Analysis, 29 (2), 201-212. Peterson, C. A., & McConnell, S. R. (1996). F actors related to intervention integrity and child outcome in social skills interventions. Journal of Early Intervention, 20 (2), 146-164. Peterson, L., Homer, A. L., & Wonderlich, S. A. (1982). Th e integrity of independent variables in behavior analysis. Journal of Applied Behavior Analysis, 15 (4), 477492. Sasso, G. M., Reimers, T. M., Cooper, L. J ., Wacker, D., Berg, W., Steege, M., Kelly, L., & Allaire, A. (1992). Use of descriptive a nd experimental analyses to identify the functional-properties of aberrant behavior in Sc hool Settings. Journal of Applied Behavior Analysis, 25 (4), 809-821. Shirley, M. J., Iwata, B. A., Kahng, S. W., Mazaleski, J. L., & Lerman, D. C. (1997). Does functional communication training compete with ongoing contingencies of reinforcement? An analysis during response acquisition and maintenance. Journal of Applied Behavior Analysis, 30 (1), 93-104. Sprague, R. L. (1993). Whistleblow ing: A very unpleasant vocation. Ethics and Behavior, 3 (1), 103-133. St. Peter, C. C., & Vollmer, T. R. (2004, May 2004). A laboratory model to study DRO and NCR. Paper presented at the Associati on for Behavior Analysis, Boston, MA. Sterling-Turner, H. E. (2002). The effects of direct training and treatment integrity on treatment outcomes in school consultation (vol 17, pg 47, 2002). School Psychology Quarterly, 17 (2). Sterling-Turner, H. E., Watson, T. S., & Moor e, J. W. (2002). The effects of direct training and treatment integrity on trea tment outcomes in school consultation. School Psychology Quarterly, 17 (1), 47-77. Sterling-Turner, H. E., Watson, T. S., Wildmo n, M., Watkins, C., & Little, E. (2001). Investigating the relationship between training type and treatment integrity. School Psychology Quarterly, 16 (1), 56-67.
94 Taylor, J., & Miller, M. (1997). When timeout works some of the rime: The importance of treatment integrity and functional assessment. School Psychology Quarterly, 12 (1), 4-22. Tingstrom, D. H. (1989). Increasing acceptability of alternative behavioral interventions through education. Psychology in the Schools, 26 (2), 188-194. Vollmer, T. R., Borrero, J. C., Lalli, J. S ., & Daniel, D. (1999). Evaluating self-control and impulsivity in children with severe behavior disorders. Journal of Applied Behavior Analysis, 32 (4), 451-466. Vollmer, T. R., Iwata, B. A., Zarcone, J. R ., Smith, R. G., & Mazaleski, J. L. (1993). The role of attention in the treatment of attention-maintained self-injurious-behavior-Noncontingent reinforcement and differen tial reinforcement of other behavior. Journal of Applied Behavior Analysis, 26 (1), 9-21. Vollmer, T. R., Roane, H. S., Ringdahl, J. E., & Marcus, B. A. (1999). Evaluating treatment challenges with differential reinforcement of alternative behavior. Journal of Applied Behavior Analysis, 32 (1), 9-23. Wickstrom, K. F., Jones, K. M., LaFleur, L. H., & Witt, J. C. (1998). An analysis of treatment integrity in school-based behavioral consultation. School Psychology Quarterly, 13 (2), 141-154. Williams, A. M., & Lattal, K. A. (1999). The ro le of the response-reinforcer relation in delay-of-reinforcement effects. Journal of the Experimental Analysis of Behavior, 71 (2), 187-194. Witt, J. C., Martens, B. K., & Elliott, S. N. (1984). Factors affecti ng teachersâ€™ judgments of the acceptability of behavioral inte rventions: Time involvements, behavior problem severity, and type of intervention. Behavior Therapy, 15 (2), 204-209. Witt, J. C., Noell, G. H., LaFleur, L. H ., & Mortenson, B. P. (1997). Teacher use of interventions in general education settings: Measurem ent and analysis of the independent variable. Journal of Applied Behavior Analysis, 30 (4), 693-696. Worsdell, A. S., Iwata, B. A., Hanley, G. P., Thompson, R. H., & Kahng, S. W. (2000). Effects of continuous and intermittent re inforcement for problem behavior during functional communication training. Journal of Applied Behavior Analysis, 33 (2), 167-179. Yeaton, W. H., & Sechrest, L. (1981). Critical dimensions in the choice and maintenance of successful treatments: Strengt h, integrity, and effectiveness. Journal of Consulting and Clinical Psychology, 49 , 156-167.
95 BIOGRAPHICAL SKETCH Claire St. Peter Pipkin has been interested in behavi or analysis since taking undergraduate coursework with Hank Pennypack er. She began her graduate studies in behavior analysis at the Un iversity of Florida in 2001, unde r the supervision of Timothy Vollmer. Claireâ€™s research interests cente r around the development and implementation of interventions for children who exhibit pr oblem behavior or w ho have fallen behind academically. Claire is specifically interested in the effects of reinforcement history on current responding, the effects of treatment integrity fail ures on common behavioral treatments, and the use of laboratory se ttings to inform applied research. During her graduate school career, Claire has conducted research using both laboratory and school settings. In 2002, she developed a human operant laboratory to study the effects of treatment integrity failures on behavioral treatments such as DRO and DRA, and is currently extending the findings of that research into applied settings. She is also currently studying the effects of re inforcement history and development on behavioral interventions. Claire would like to contribute to the field of behavior analysis through both research and teaching. She would like to cont inue her line of programmatic research focusing on treatment integrity and behavior al history, includi ng the influence of historical factors on childr enâ€™s current responding.