Received: 27 March 2020 Revised: 15 September 2020 Accepted: 16 September 2020 DOI: 10.1002/bin.1749 RESEARCH ARTICLE WILEY Developing a treatment for hand-clapping maintained by automatic reinforcement using sensory analysis, noncontingent reinforcement, and thinning Sarah K. Slocum 1.2 | Nicolette Yatros | Mindy Scheithauer 1,2 Department of Severe Behavior, Marcus Autism Center, Atlanta, Georgia, USA 2Department of Pediatrics, Emory University School of Medicine, Atlanta, Georgia, USA 3Department of Health Professions, Rollins College, Winter Park, Colorado, USA Correspondence Sarah K. Slocum, Department of Severe Behavior, Marcus Autism Center, 1920 Briarcliff Rd. NE, Atlanta, GA, 30329, USA. Email: sarah.slocum.freeman@choa.org Abstract Two subjects diagnosed with autism spectrum disorder and related disabilities who engaged in hand-clapping main- tained by automatic reinforcement participated in this study. We conducted a sensory analysis to evaluate matched stimuli that functioned as an abolishing operation or extinction for different sources of sensory reinforcement. Finally, we implemented noncontingent reinforcement (NCR) using the stimuli found to compete in the assessment to reduce the target behavior, and we thinned item availability. Results showed a decrease in hand-clapping, and hand-clapping remained low when we thinned the schedule of reinforcement. This research further elucidates how NCR can treat problem behavior maintained by automatic reinforcement. KEYWORDS assessment, autism spectrum disorder, functional analysis, noncontingent reinforcement, stereotypy, treatment 1 INTRODUCTION Problem behavior maintained by automatic reinforcement is prevalent across clinical and classroom settings (Beavers, Iwata, & Lerman, 2013) and is commonly treated using noncontingent reinforcement (NCR; Phillips, Iannaccone, Rooker, & Hagopian, 2017). In an NCR arrangement, the reinforcer is provided on a fixed-time or variable-time schedule rather than being contingent on an individual's behavior. In the treatment of problem behavior maintained by automatic reinforcement, interventionists implement NCR using the delivery or availability of several types of stimuli as follows: (a) arbitrary stimuli (environmental enrichment; Horner, 1980), (b) highly 228 Behavioral Interventions. 2021;36:228-241. wileyonlinelibrary.com/journal/bin 2020 John Wiley & Sons Ltd. SLOCUM ET AL. WILEY 229 preferred stimuli (Lindberg, Iwata, Roscoe, Worsdell, & Hanley, 2003; Phillips et al., 2017), (c) competing stimuli (providing access to stimuli with potentially more reinforcing values than is produced by the target behavior; Kennedy & Souza, 1995; Love, Miguel, Fernand, & LaBrie, 2012; Saini et al., 2016), and (d) matched competing stimuli (stimuli that provide similar stimulation as the target behavior; Piazza, Adelinis, Hanley, Goh, & Delia, 2000; Piazza et al., 1998; Rapp, 2006, 2007). Piazza et al. (1998) presented an assessment of pica maintained by automatic reinforcement comparing matched (i.e., items producing oral stimulation) to unmatched stimuli. For each of the three subjects, Piazza et al. assessed the level of pica and item engagement across 18 to 20 items, and identified clear hierarchies with matched stimuli resulting in more item interaction and less pica. Noncontingent access to both matched and unmatched stimuli found matched stimuli were more likely to compete for two of the subjects. Matched stimuli were also more likely to compete for vocal stereotypy, with one study identifying immediate reductions in vocal stereotypy for 10 of 11 subjects with a matched item compared to only one of 10 when using unmatched stimuli (Rapp et al., 2013). The results of these studies have been replicated across research groups and behavior topographies (e.g., Kennedy & Souza, 1995; Piazza et al., 2000; Rapp, 2006; Rincover, Cook, Peoples, & Packard, 1979). While interventions using noncontingent access to items are often successful alone, for some subjects, it is necessary to combine noncontingent access to items with a procedure designed to implement sensory extinction, such as blocking, protective equipment, or redirection (Gover, Fahmie, & McKeown, 2019; Rooker, Bonner, Dillon, & Zarcone, 2018). For example, Roscoe et al. found substantial decreases in problem behavior maintained by automatic reinforcement when they presented subjects with noncontingent access to leisure items as a sole treatment component for only 7 of 14 subjects. Blocking was necessary to achieve clinically significant reductions in problem behavior for the remaining subjects. However, it is important to consider the limitations of blocking. First, response blocking can evoke other topographies of problem behavior such as aggression (Hagopian & Toole, 2009). Second, the inclusion of response blocking is time intensive in that it requires implementation with a high degree of procedural integrity to be effective (Smith, Russo, & Le, 1999; Wunderlich, Vollmer, & Zabala, 2017). Finally, the use of response blocking may produce stimulus control such that the behavior will be extinguished or punished under conditions in which an adult is present (signaling response blocking is in place); however, the response may continue to occur under alone or covert conditions (see Toussaint & Tiger, 2012 for an example of covert problem behavior). Due to limitations in response blocking, and evidence that matched stimuli might be more likely to compete than unmatched stimuli for some subjects, it is worth additional evaluations of methodologies to identify matched stimuli. One such extension might be to evaluate procedures based on principles of both abolishing operations (i.e., testing items that produce the same hypothesized sensory stimulation) and extinction (i.e., testing items that block the hypothesized sensory stimulation of the behavior). Past assessments have used procedures based on abolishing operations or extinction but have not combined the two. For example, Piazza et al. (2000) provided items with similar oral stimulation to pica as an abolishing operation, but did not evaluate the effect of eliminating the oral stimulation caused by pica (i.e., extinction). Rincover et al. (1979) eliminated the noise produced by plate spinning as extinction, but did not evaluate the impact of a competing item that produced a similar noise (i.e., abolishing operation). A more thorough assessment may be implemented by combining both potential avenues for evaluating the sensory stimuli maintaining the behavior. Another extension involves adjusting the methodology used in competing items assessments. The assessment of matched items in Piazza et al. (1998) included exposure to different items across repeated presentations (10 presentations of each item for 30 s for one subject and one presentation of each item for 5 min for the other two subjects); the authors presented a bar graph depicting average item engagement and problem behavior when necessary, to the different assessment conditions. Additionally, the assessment included only brief exposures to potential competing or matched stimuli (5 min total), thus the assessment results do not guide understanding of whether reductions in problem behavior would maintain for longer durations. While this approach has been replicated several times (e.g., Jennett, Jann, & Hagopian, 2011; Verriden & Roscoe, 2019), there are situations in which items identified in a competing items assessment do not result in sustained reductions in challenging 230 WILEY. SLOCUM ET AL. behavior (e.g., Roscoe, Iwata, & Zhou, 2013). Thus, there may be room for improvement in this assessment methodology. One potential improvement may be to consider the trend and variability across series of the competing items through a multielement evaluation across several exposures as opposed to graphing condition averages after only a few presentations. It is possible that different items may be identified as competing when trend and variability are accounted for, which may subsequently improve treatment results. On the contrary, it is possible that different items may produce variable levels of problem behavior, which may predict individuals who might not respond to treatment in the long term. Another necessary area of future research related to competing items is the maintenance of treatment gains. The studies discussed to this point have not included any efforts to thin the availability of competing stimuli. While one could argue an item could always be available, there are some concerns. First, continuous access to a competing item may interfere with instructions, such as daily living demands, academic instructions, or teaching social skills (Conroy, Asmus, Sellers, & Ladwig, 2005). Second, competing items may not always be available under conditions in which an item breaks or requires cleaning. Finally, these items may lose their ability to compete over time if they are provided continuously, as satiation could occur. This is especially concerning as past research suggests problem behavior maintained by automatic reinforcement may re-emerge after a competing item is removed when it is not systematically thinned (Lanovaz, Fletcher, & Rapp, 2009). It is important to study schedule thinning to determine the likelihood that treatment results will maintain under scenarios when the competing item is not available. One potential approach for schedule thinning is to begin with brief exposures to the unavailability of leisure items, and slowly increase item unavailability over time. This thinning procedure is common in the treatment of problem behavior maintained by social reinforcement. For example, following a NCR treatment in which the re- searchers delivered the functional reinforcer continuously, Slocum, Grauerholz-Fisher, Peters, and Vollmer (2018) implemented a gradual thinning procedure with a multiple schedule. Using a stimulus card and the presence of items to signal either the availability of the functional reinforcer (SD) or the unavailability of the functional rein- forcer (S), the authors began with brief 10-sec exposures to the unavailability of reinforcement. Across sessions, the researchers increased the SA period until subjects did not have access to the functional reinforcer for half of the session. All three subjects displayed low levels of problem behavior across this schedule thinning procedure. The purpose of the current study was to extend the literature on competing items as a treatment for auto- matically-maintained behavior in three ways. First, we incorporated procedures to test for extinction and abolishing operation properties of sensory stimuli that may be maintaining the target behavior in the same assessment. Second, we implemented the assessment and graphed results using a multielement design across several sessions to account for trend and variability in target behavior. Last, we evaluated treatment effects of the identified competing item during thinning of the reinforcer availability. To achieve these aims, we first conducted a functional analysis to determine stereotypy in the form of hand- clapping was maintained by automatic reinforcement. We subsequently conducted a sensory analysis including conditions to evaluate both items that may serve as abolishing operations and those that may serve as various functions of extinction of specific sensory modalities. Following successful treatment of hand-clapping using noncontingent access to the hypothesized matched stimuli demonstrated across multiple sessions, we thinned the NCR schedule using procedures similar to Slocum et al. (2018). 2 | METHOD 2.1 Subjects, setting, and materials We recruited two subjects attending Rollins College practicum clinics for applied behavior analysis therapy in the central Florida area for this study. Sam was a 9-year-old boy diagnosed with autism spectrum disorder SLOCUM ET AL. WILEY 231 (ASD) and Fragile X. Micah was a 15-year-old boy diagnosed with ASD, obsessive compulsive disorder, and epilepsy. Sam communicated using some picture exchange and an augmentative communication device; Micah communicated using hand leading. Neither subject displayed functional vocal behavior. Both subjects engaged in hand-clapping among other stereotypic behavior such as vocalizations and body rocking. Hand-clapping may seem harmless; however, it can be a hindrance to academic achievement as well as disruptive to peers. This behavior also caused damage to both clients' hands and wrists. Sam had permanent callouses, and Micah had broken his wrists prior to the start of the current study (his wrists had healed before we began our assess- ments and intervention). We did not evaluate other forms of stereotypy (beside hand-clapping) in this study. We obtained institutional review board approval for both subjects; further, we obtained signed consent from caregivers of both subjects. We conducted sessions in therapy rooms at the subjects' clinics. Sam's session room was 2.4 m by 2.9 m, and Micah's session room was 4.6 m by 6.1 m. Each room contained a table, two chairs, and stimuli needed for various conditions. Materials for the functional analysis included academic demand materials and leisure items (based on the condition, described in more detail below). Materials for the sensory analysis consisted of a video of hand-clapping, an audio recording of hand-clapping, sound-proof headphones, gloves, and a vibrating hand massager. Treatment included one or more items from the sensory analysis. Due to reactivity to recording devices, Sam's sessions were not video recorded. For Micah, we recorded all sessions on an iPad for subsequent scoring. 2.2 Response measurement, interobserver agreement, and treatment integrity We defined the target behavior, hand-clapping, as the subject opening his hands with a distance of two in or more followed by closing the hands that produces an audible sound. Additionally, for Micah, hand-clapping included audible clapping his thighs, stomach, and feet with an open palm from two in or more. We measured this behavior using response rate (responses per min). Two observers independently scored responses across 50% of Sam's and Micah's sessions to obtain interobserver agreement (IOA). We calculated IOA using a 10-s exact interval-by-interval method. For each interval, if both observers scored the same number of instances of target behavior (e.g., one observer scored 9 and the other scored 9), we considered that interval an agreement. If the two observers scored a different number of instances of target behavior (e.g., one observer scored 9 and the other scored 10), we considered that interval a disagreement. We divided the total number of agreements by the total number of intervals and multiplied by 100 to obtain a percentage for each session. For Sam, IOA was 96% (range, 83% -100%), and for Micah, IOA was 95% (range, 80%-100%). We collected treatment integrity data throughout 50% of treatment sessions for Sam and Micah using a 6-point checklist. A data collector scored "Yes" or "No" for each item in Appendix A. If an item was not appropriate for a session, the data collector scored "N/A," and we altered the denominator accordingly (e.g., one item states the correct stimulus was delivered, but a stimulus was not provided in baseline sessions). We added up the number of items marked "Yes," divided by the total possible items, and multiplied by 100 to obtain a percentage. For Sam, treatment integrity was 97% (range, 80%-100%), and for Micah, treatment integrity was 96% (range, 80%-100%). 2.3 | Procedure We used a multielement design for the functional and sensory analyses; we used a reversal design to evaluate treatment. Sessions were 5 min across all phases. Attempts to engage in hand-clapping had no programmed consequences (i.e., response blocking was not implemented). 232 WILEY SLOCUM ET AL. 2.3.1 Functional analysis We began by conducting a functional analysis of hand-clapping based on procedures described by Iwata, Dorsey, Slifer, Bauman, and Richman (1994a) with some modifications. We included demand, attention, play and no- interaction conditions. In the demand condition, the therapist presented academic demands to the subject continuously using a least-to-most prompting sequence. We selected demands based on current clinical targets for each subject. For Sam, the demand was to match pictures of house hold items, and for Micah, the demand was to clean up task materials. If the subject engaged in hand-clapping, the therapist responded with "ok, you don't have to," and removed the demand for 30 s before representing demands. In the attention condition, the subject had continuous access to a moderately preferred leisure item identified through a free operant preference assessment (Roane, Vollmer, Ringdahl, & Marcus, 1998). The therapist told the subject, "I have work to catch up on; you sit here until I'm finished." If the subject engaged in hand-clapping, the therapist interjected a brief reprimand (e.g., "stop that"). The play condition consisted of the subject having continuous access to attention from the therapist as well as access to a highly preferred item identified through the preference assessment mentioned above. This condition served as our control. Finally, the no-interaction condition consisted of the subject and therapist in a room without any toys or demands. The therapist did not interact with the subject. To further confirm the behavior was maintained by automatic reinforcement, we included an extended no-interaction phase at the end of the functional analysis. 2.3.2 Sensory analysis of matched stimuli In our sensory analysis of matched items, we assessed a subset of potential sources of sensory reinforcement. We included six items total: video of hand-clapping, audio recording of hand-clapping, sound-proof headphones, gloves, vibrating hand massager, and lights off. We included two items for each sensory modality; one item possessed modality-specific reinforcing qualities (e.g., audio recording of hand-clapping at a similar rate to the subject) and the other item possessed potential extinction qualities or prevented the subject from contacting reinforcement (e.g., sound-proof headphones to prevent sound produced by hand-clapping). Before beginning the sensory analysis, we conducted forced exposure to all stimuli to reduce the likelihood of suppression of hand-clapping due to novelty rather than matched sensory stimulation (Rincover et al., 1979). We exposed both subjects to each item three times for 1 min in a random order. Subsequently, we began the sensory analysis to analyze potential sensory modalities maintaining the target behavior. We alternated the presentation of each item in an alternating treatments design to see which condition resulted in the lowest level of hand-clapping. The no-interaction condition was identical to the no-interaction condition from the functional analysis. This condition served as our baseline against which we compared the level of hand-clapping in other conditions. Across all test conditions, unless otherwise noted, sessions were identical to the no-interaction condition except for the inclusion of the item. To test if visual stimulation was the source of automatic reinforcement for hand-clapping, we conducted a lights-off and a video condition. In the lights-off condition, we turned off the lights for the duration of the session. Low light acted as sensory extinction for hand-clapping maintained by visual stimulation. The video condition included a video of a person hand-clapping at a similar rate to the subject without sound from the perspective of the subject (i.e., the video was taken as if the client was looking down at their own hands). This condition resembled a matched stimulus treatment in that a video of hand-clapping might have provided similar reinforcement if visual stimulation was the sensory variable maintaining hand-clapping. The video was recorded from the perspective of the subject, but was not the subject's actual hands. To test for auditory stimulation, we introduced a pair of sound-proof headphones and an audio recording condition. The sound-proof headphones condition involved the subject wearing sound-proof headphones. In this condition, if hand-clapping was maintained by auditory stimulation, this condition resembled a sensory SLOCUM ET AL. WILEY 233 extinction procedure. In the audio recording condition, we played the sound of hand-clapping throughout the session; this resembled a matched stimulus procedure if auditory stimulation was the reinforcer for hand- clapping. To test for tactile sensory stimulation, we introduced a glove and vibrating massager condition. For the glove condition, the subject wore shock-absorbent gloves to eliminate any tactile reinforcement that might have been provided by hand-clapping. Again, this could have resembled sensory extinction if tactile stimulation was the maintaining reinforcer for hand-clapping. Alternatively, a vibrating massager was tested as a matched stimulus procedure where the subject had access to a vibrating massager for the duration of the session. Across conditions, attempts to interrupt the sensory stimulus were blocked (e.g., attempts to remove gloves were blocked). In addition, across sessions we prompted subjects to interact with items using a least-to-most prompting hierarchy to engage with the item every 5 s if he stopped doing so. The therapist initially used a verbal prompt such as "Play with the hand massager" or "Wear your gloves." If the subject did not engage with the stimulus after 5 s, the therapist used a gestural prompt pointing toward the stimulus. If the subject did not engage with the stimulus again after 5 s, the therapist placed the item in the subject's hand or prompted the subject to look at the visual stimulus. We did not include these prompts with the auditory clapping or the lights-off conditions as we could not know whether the subject was contacting the stimuli. The condition that produced the lowest levels of hand-clapping was the condition selected for treatment. If multiple conditions in the sensory analysis produced low levels of hand-clapping, we allowed the caregivers to select which inter- vention they would prefer. 2.3.3 Treatment After the sensory analysis, we implemented an extended treatment for each subject within a reversal design. We used data from the no-interaction condition of the sensory analysis as our initial baseline. For Sam and Micah, the vibrating hand massager was used for treatment along with the same prompting procedure described above. We also embedded a strategy from the Iwata, Pace, Cowdrey, and Miltenberger (1994b) study evaluating different procedural forms of extinction for self-injurious behavior maintained by different sources of reinforce- ment (see Jack in Iwata et al., 1994b). In their study, in addition to evaluating the form of extinction matched to the function of problem behavior, such as ignoring for self-injurious behavior maintained by attention, they also added another form of extinction across some portion of baseline and treatment phases, such as sensory extinction (a helmet). For a portion of baseline and treatment phases, we added one of the treatments shown to be ineffective in the sensory analysis to validate the assessment. That is, after implementing the initial baseline and treatment phases, demonstrating a reduction in hand-clapping, we then added an unmatched stimulus. Once that stimulus was added in, we left the stimulus in place in the reversal to baseline and reintroduction of the matched stimulus. After exposure to the matched and unmatched stimuli in the treatment reversal, we removed the unmatched stimulus prior to schedule thinning. If the added unmatched stimulus produced reductions in hand-clapping under baseline contingencies, the sensory analysis would have been invalidated. This method of validation has the advantage of avoiding exposure to a hypothesized ineffective treatment alone and instead capitalizes on the already planned return to baseline as a validation period. After we found treatment effects using an ABAB (with A being baseline or no interaction and B being treatment or continuous access to the hand massager) design, we conducted schedule thinning using methods similar to Slocum et al. (2018). We started treatment with signaled continuous access to the stimulus identified in the sensory analysis to produce the lowest levels of hand-clapping. We then gradually increased the time in which the individual was exposed to the signaled unavailability of the stimulus, beginning with 10 s of unavailability per session, 30 s, 60 s, 120 s, 120 s + 60 s, and 120 s + 60 s + 60 s. In the last two phases of schedule thinning, two and then three 234 WILEY 8- Responses per Minute of Hand Clapping 50- 30- 20- 10- No-Interaction Demand -- Attention -- Play 8 12 Sessions Sam Micah 16 SLOCUM ET AL. FIGURE 1 Results for Sam's (top panel) and Micah's (bottom panel) functional analyses including an extended no-interaction phase separate removals of the stimulus occurred within the session. When the hand massager was unavailable, it was removed from the participant and placed across the room. We also included a card with a white side and a red side to signal whether continuous reinforcement was or was not available, respectively, in a multiple schedule. The close presence and absence of the hand massager signaled its availability, but the card was added as an additional signal of that contingency. To thin the schedule of reinforcement, we increased the unavailability of the stimulus based on the subjects' target behavior remaining at or below 80% of baseline levels for Sam and 53% of baseline levels for Micah for two consecutive sessions (Micah's reduction was 53% rather than 80% due to a calculation error). We decreased the unavailability of the stimulus based on subjects' hand-clapping occurring at rates higher than the reduction cri- terion for a single session. We ended schedule thinning when hand-clapping was low and the matched stimulus was available for 1 min and unavailable for 4 min (i.e., 120+ 60+ 60 s) for seven consecutive sessions. For Micah, we also included generalization probes up to three months after the conclusion of schedule thinning in a different location and with different therapists. SLOCUM ET AL. Responses per Minute of Hand Clapping 50- 40- 30- 20- 10- Headphones Audio Video Lights-off Hand massager Gloves No-interaction Sam Micah WILEY 235 2 4 6 8 10 12 14 16 18 20 22 Sessions FIGURE 2 Results for Sam's (top panel) and Micah's (bottom panel) sensory analyses 3 | RESULTS Figure 1 displays the functional analyses for both subjects. Sam's responding is depicted in the top panel, and Micah's responding is depicted in the bottom panel. In the initial assessment, hand-clapping occurred at high rates across all conditions for Sam (M = 5.6 responses per min, rpm) and Micah (M = 26.7 rpm). Responding persisted in an extended no interaction phase (M = 6.0 rpm for Sam and M = 38.8 rpm for Micah). Figure 2 displays the analysis of sensory reinforcement for both subjects. For Sam, the lowest level of hand-clapping occurred when he had access to the hand massager (M = 1.7 rpm) and the video (M = 1.7 rpm). Sam had the highest level of hand-clapping in the no-interaction condition (M = 3.9 rpm). For Micah, the lowest level of hand-clapping occurred when he had access to the hand massager (M = 0.5 rpm). Micah had the highest level of hand-clapping in the glove condition (M = 38.9 rpm). The treatment data are presented in the top panel of Figure 3 for Sam and the bottom panel of Figure 3 for Micah. Again, baseline is the no-interaction condition of the sensory analysis re-plotted. We selected the hand 236 WILEY Responses per Minute of Hand Clapping 30- 25- 20- 15- 10- 5- 5- 4- 3- BL Treatment Arbitrary Treatment Baseline 10 Arbitrary Treatment BL Treatment Baseline Treatment Sam SLOCUM ET AL. T 20 30 40 50 Treatment 10 20 30 Sessions 40 50 Micah FIGURE 3 Results for Sam's (top panel) and Micah's (bottom panel) treatment and schedule thinning (bottom panel). The asterisk indicates where schedule thinning increased to a leaner schedule while the minus sign indicates where the thinning was decreased to the last successful schedule. The arbitrary treatment was in place from sessions 10 to 23 for Sam and 8 through 20 for Micah massager as the competing stimuli treatment for Sam's hand-clapping based on parent preference. Noncontin- gent access to the hand massager resulted in low levels of hand-clapping for Sam (M = 0.3 rpm). After a reversal to a no-interaction baseline, we re-captured treatment effects (M = 0.2 rpm) and began schedule thinning at Session 26. In addition to this reversal design, at Session 10, we added a treatment component that the sensory analysis suggested would be ineffective, an audio recording of hand-clapping, and kept that stimulus in place through Session 23. This additional component had no impact on behavior (i.e., we replicated baseline rates with this component in place). At the completion of schedule thinning, when Sam had access to the matched stimulus SLOCUM ET AL. WILEY 237 for only 1 min of the 5 min sessions, his rate of hand-clapping remained low for seven consecutive sessions (M = 0.1 rpm). For Micah, noncontingent access to the hand massager resulted in lower levels of hand-clapping (M = 2.2 rpm). After a reversal to a no-interaction baseline, we recaptured treatment effects (M = 5.4 rpm) and began schedule thinning at Session 23. In addition, at Session 8, we added the audio recording of hand-clapping, which the sensory analysis suggested would be ineffective at reducing hand-clapping, and kept that stimulus in place through Session 20. As hypothesized, this stimulus had no impact on hand-clapping. At the completion of schedule thinning, when Micah had access to the matched stimulus for only 1 min of the 5 min sessions, his rate of hand-clapping remained low for seven consecutive sessions (M = 5.9). For Micah's probe sessions, low levels of hand-clapping occurred in a new environment (Sessions 57 and 58) and with new therapists (Sessions 59 and 60). We also recorded the frequency of prompts to interact with the matched stimulus across treatment and scheduling thinning sessions. On average, the therapist prompted Sam once and Micah twice per 5-min session. 4 | DISCUSSION The current study extends existing literature on the assessment and treatment of problem behavior maintained by automatic reinforcement. After confirming hand-clapping was maintained by automatic reinforcement for both subjects, we conducted a sensory analysis of competing items to identify the specific sensory modality reinforcing the target behavior. For both subjects, an item that produced tactile stimulation resulted in the lowest level of hand-clapping, suggesting it is possible this was the source of sensory stimulation maintaining hand-clapping. Finally, we demonstrated the item identified in the sensory analysis reduced hand-clapping and systematically thinned the schedule of reinforcement by increasing the amount of time in which the item was unavailable. We conceptualized the stimuli evaluated in the sensory analysis of competing items as either operating through abolishing operations (video of hand-clapping, audio of hand-clapping, and massager) or extinction (lights off, noise canceling head phones, and gloves). Under this hypothesis, one would expect for conditions matched in the targeted sensory modality to have similar results. However, this was not the case for either subject. For Sam, we saw a lower level of hand-clapping when the video (putative visual abolishing operation) and hand massager (putative tactile abolishing operation) were implemented, but did not observe corresponding decreases in the lights-off (putative visual extinction) and gloves (putative tactile extinction) conditions. Similarly, for Micah, the massager resulted in the lowest level of hand-clapping while the gloves resulted in the highest. This is counter to our hypothesis and may be explained in a few ways that require further study. First, it is possible the massager produced preferred stimulation that competed with the automatic rein- forcement produced by hand-clapping, but was not truly matched, meaning it did not actually produce the same tactile stimulation as hand-clapping. This finding is important as it may also be true of other matched stimuli assessment and treatment studies (e.g., Piazza et al., 1998; Roscoe et al., 2013) as they have not included an extinction test. Thus, future research may focus on including both putative extinction and abolishing operation conditions when determining the sensory aspect operating in automatically reinforced problem behavior as our study suggests that assumptions made about the sensory aspect of different competing items could be flawed. Second, the putative extinction procedures may not have sufficiently reduced the sensory stimulation produced by the behavior. For example, there was still very dim light present in the lights-out condition and the gloves lessened but did not completely eliminate the tactile stimulation produced by hand-clapping. Future research should repeat the analysis with different items that may produce different sensations (e.g., banging hands on a table instead of a massager) and more thoroughly eliminate sensory input from the behavior (e.g., using a blind fold or additional padding). Despite limitations, the current study extends prior research in several important ways. First, although most studies use a functional analysis to determine the function of behavior (Hanley, Iwata, & McCord, 2003; Iwata et al., 238 WILEY. SLOCUM ET AL. 1994a), few studies explore which modality of sensory reinforcement maintain behavior when it is automatically reinforced (e.g., tactile, visual, and auditory), which is crucial in developing function-based treatments. The research done in this field has primarily focused on behaviors with one hypothesized sensory stimulation that maintained the behavior, such as oral stimulation for pica and hand mouthing (Piazza et al., 1998; 2000); auditory stimulation for vocal stereotypy and plate spinning (Love et al., 2012; Rapp, 2007; Rapp et al., 2013; Rincover et al., 1979); and kinesthetic consequences for dangerous acts (Piazza et al., 2000). The current study is an extension of this liter- ature as it is targeting a behavior that could be maintained by multiple forms of sensory stimulation, and the sensory analysis included stimuli that isolated each. Second, the current study has some methodological strengths compared to prior work in the assessment and treatment of problem behavior maintained by automatic reinforcement. This study was the first to combine matched-stimulus interventions that act as abolishing operations and those that act as extinction within one analysis. While the results were not congruent with our hypothesis, as described above, the methodology sets the ground work for future replications using this model. Additionally, future studies might also replicate the multi- element design used as it is advantageous in evaluating level, trend, and variability within the assessment phases. Third, the current study evaluated both the item identified in the sensory analysis as a matched stimulus as well as an item identified as an unmatched stimulus using an overlapping approach (Iwata et al., 1994b). That is, we not only exposed subjects to a matched stimulus that produced reductions in hand-clapping; we also exposed them to an unmatched stimulus that did not produce a change in hand-clapping. During times in which the unmatched stimulus was embedded on top of the matched stimulus, we did not see additive treatment effects; during times in which the unmatched stimulus was in place without the unmatched stimulus (in the second baseline phase), we observed original baseline levels of hand-clapping. Therefore, we were able to validate the assessment without exposing the individual to additional phases with hypothesized ineffective contingencies. Finally, although research on thinning NCR schedules has been conducted with problem behavior maintained by social reinforcement (e.g., Hagopian, Crockett, Stone, DeLeon, & Bowman, 2000; Slocum et al., 2018), there has been minimal research applying schedule thinning to NCR interventions for problem behavior maintained by automatic reinforcement. Schedule thinning is an important treatment component because competing stimuli might not always be available, especially if the procedure is implemented for extended periods of time. We were able to thin the schedule of reinforcement such that the matched stimulus was only available for 20% of sessions and the frequency of prompting to engage with the item remained low. There were a few other limitations to the current study that create opportunities for future research. First, we did not program for blocking during the treatment evaluation. However, due to the contingency of redirection to the hand massager after 5 s of no interaction, and the incompatible nature of interacting with the hand massager and hand-clapping, the contingency was inadvertently set-up that 5 s of hand-clapping would result in a redirection to the massager. It is unlikely that this was an active component of treatment because of the low-rates of hand-clapping observed during schedule thinning. During times the massager was unavailable, no redirection occurred. Thus, if this was controlling subjects' behavior, one would expect increased hand-clapping during the signaled times that the massager was unavailable, as this would also represent a signal that the redirection was not in place. Nonetheless, subjects were exposed to this contingency consistently in the early treatment sessions and it is unclear what impact this had on treatment effects. We also do not have information on hand-clapping post-session. There were anecdotal reports from family and staff members that lower levels of hand-clapping would occur for up to 30 min after treatment sessions; however, we have no empirical evidence of this. Future research should measure treatment effects during and following formal experimental sessions similar to Rapp (2006). The study would also be strengthened by maintenance and generalization data for Sam. Unfortunately, this was not possible because of health concerns that caused his family to withdraw him from the clinic for an extended period. Further, we did not conduct any terminal probes of the schedule thinning procedure. In Slocum et al. (2018), one of the three subjects was exposed to terminal-schedule probes to see if the schedule needed to be thinned at each step or if responding would remain low when SLOCUM ET AL. WILEY 239 reinforcement was unavailable for the duration of the terminal goal. They found probes made the evaluation longer for that subject and all steps of schedule thinning were necessary, guiding our decision against terminal probes. However, future research might include terminal-schedule probes to see if time could be saved by skipping steps in the schedule-thinning procedure. Additionally, the study design did not allow an analysis of whether the visual stimuli used to signal the availability of items contributed to the success of schedule thinning or discrimination given the items also changed location in the room when they were available and unavailable. Future research may compare thinning with and without the additional visual stimuli to elucidate any additive effects of this component. More research is needed to verify whether matched stimuli are truly superior to unmatched stimuli as there may be limitations to selecting a matched stimulus in some cases. There may be some topographies of behavior for which a matched stimulus is not available. For instance, for head-targeted self-injurious behavior, the sensory reinforcer may be the pain that results from the behavior, which would be an inappropriate reinforcer to match. Second, clinicians must consider the feasibility of matched stimuli. For instance, some matched stimuli in past research may be difficult to provide in all settings (e.g., shaving cream; Piazza et al., 2000). Last, while past research suggests that some matched stimuli compete better than some unmatched stimuli, this is not necessarily the case for all matched and unmatched items or for all individuals. Especially given that, we cannot be certain that all possible unmatched stimuli that would compete were included in the analyses. In other words, it is possible that different unmatched item may have worked better in treatment than the selected matched stimuli but they were missed in the analysis. In sum, future research should replicate this study, and others that evaluate the use of matched stimuli, with subjects with different presenting concerns and with problem behavior maintained by different sources of sensory stimulation. Future research can work to define what is considered a matched and unmatched stimulus, and to determine which is optimal given the numerous variables to consider in treatment selection (e.g., ease of implementation, maintenance, and generalization). For the time, clinicians should conduct competing items assessment with hypothesized matched (when possible) and highly preferred items (but hypothesized unmatched) items and use the results of the competing items assessment to guide treatment development. The current study demonstrated one approach to assessing various sources of sensory stimulation produced by a behavior maintained by automatic reinforcement. The assessment led to the successful decrease in hand-clapping for two subjects. Furthermore, we applied a method to thin the availability of the matched stimulus to a more feasible schedule. While the current study engendered many more empirical questions, this matched stimulus NCR intervention might be applied to clinical settings in which an effective yet simple treatment is needed to reduce the level of problem behavior maintained by automatic reinforcement. CONFLICT OF INTEREST The authors have no conflict of interests. DATA AVAILABILITY STATEMENT The data that support the findings of this study are available from the corresponding author upon reasonable request. ORCID Sarah K. Slocum https://orcid.org/0000-0002-5597-2216 Mindy Scheithauer https://orcid.org/0000-0002-0775-1013 REFERENCES Beavers, G. A., Iwata, B. A., & Lerman, D. C. (2013). Thirty years of research on the functional analysis of problem behavior. Journal of Applied Behavior Analysis, 46, 1-21. https://doi.org/10.1002/jaba.30