Abstract
Although numerous studies have shown that response interruption and redirection (RIRD) can decrease vocal stereotypy displayed by children with autism spectrum disorder, relatively few have evaluated the subsequent effects of RIRD. We evaluated the immediate (when a change agent implemented RIRD) and subsequent (after a change agent discontinued RIRD) effects of RIRD on two participants’ vocal stereotypy using a nonconcurrent multiple baseline design with an embedded reversal design combined with a multiple schedule. As a part of the analysis, we conducted the fourth component 25 min after the third component. In addition, we compared the effects of RIRD with 5-min and 20-min components on both participants’ vocal stereotypy. Results show that (a) RIRD during either 5-min or 20-min components decreased each participant’s immediate engagement in vocal stereotypy, (b) only RIRD during 20-min components decreased each participant’s subsequent engagement in vocal stereotypy relative to the no intervention component, and (c) the subsequent effects of RIRD did not extend into the fourth component for either participant.
Rapp and Vollmer (2005) defined motor stereotypy as repetitive, invariant noninjurious behavior that persists in the absence of social consequences. As an extension of that definition, Lanovaz and Sladeczek (2012) defined vocal stereotypy (VS) “as any repetitive sounds or words produced by an individual’s vocal apparatus that are maintained by nonsocial reinforcement” (p. 148). Notably, these definitions include both structural (i.e., movement or sound repetition) and functional components (persistence of the repetitive behavior without social consequences; Rapp & Lanovaz, 2016). In a review of the literature, Chebli, Martin, and Lanovaz (2016) reported that 88% of individuals with autism spectrum disorder (ASD) engaged in at least one form of stereotypy.
During the past decade, response interruption and redirection (RIRD), originally described by Ahearn, Clark, MacDonald, and Chung (2007), has become one of the most researched interventions for VS (Rapp & Lanovaz, 2016). As an intervention for VS, RIRD involves a change agent providing demands or questions that require the participant to engage in appropriate vocal responses. The change agent continues to provide the demands until the participant responds independently to three consecutive questions without engaging in VS. As a whole, most studies have found that RIRD decreases vocal or motor stereotypy (e.g., Ahrens, Lerman, Kodak, Worsdell, & Keegan, 2011; Cassella, Sidener, Sidener, & Progar, 2011; Liu-Gitz & Banda, 2009; Shawler & Miguel, 2015).
Despite abundant empirical support for RIRD, not all aspects of the intervention are clear. For example, some studies suggest that the effects of RIRD on vocal or motor stereotypy may be overstated if researchers do not account for either the RIRD implementation time during each session (Carroll & Kodak, 2014; Wunderlich & Vollmer, 2015) or the number of times RIRD is implemented each session (Pastrana, Rapp, & Frewing, 2013). These studies also suggest that RIRD does not function as punisher for some participants’ stereotypy. Relatedly, although researchers suspect that the additional effort imposed by RIRD contributes to a punishment effect for stereotypy, the minimum effective parameter of RIRD is not yet clear. To this end, Saini, Gregory, Uran, and Fanetti (2015) have shown that providing RIRD with one demand (RIRD-1) per instance of stereotypy was as effective as providing RIRD with three demands (RIRD-3) per instance of stereotypy.
The context for which RIRD is best suited is also not yet clear. Toper-Korkmaz, Lerman, and Tsami (2018) recently evaluated the extent to which response contingent loss of toys contributed to treatment effects with RIRD-1 and RIRD-3 during a play or leisure context for three children with ASD. Consistent with Saini et al. (2015), Toper-Korkmaz et al. found that the RIRD-1 condition was as effective as the RIRD-3 condition; however, they also found that toy removal alone reduced stereotypy to low levels for two of three participants. These findings suggest that (a) less time-intensive variations of RIRD are a viable treatment option for VS, and (b) RIRD may not be the first-line intervention in all contexts. Toward this end, Cook, Rapp, and Brogan (2018) suggested that practitioners provide context-specific intervention for stereotypy. Specifically, Cook et al. suggested that RIRD was most appropriate during academic periods or other activities wherein an instructor is typically in close proximity to the learner who displays stereotypy.
Another practical feature that has not received as much attention from researchers is the subsequent effects of RIRD on vocal or motor stereotypy. Lanovaz, Rapp, and Fletcher (2010) described a three-component methodology for evaluating the immediate (when intervention is provided) and subsequent (after intervention is discontinued) effects of interventions for automatically reinforced behavior such as stereotypy. Accordingly, a researcher exposes each participant to a series of no-interaction (NI) control and test sequences; each sequence comprised three 5-min or 10-min components. During all three components of the NI sequence, the researcher does not provide consequences for the participant’s stereotypy or any other behavior. During the test sequence, a researcher provides a specific intervention for the targeted form of stereotypy during only the second component of the sequence; the researcher conducts the first and third components in the same manner as in the NI sequence.
To date, only three studies involving a total of five participants have evaluated the immediate and subsequent effects for RIRD on either VS alone (Schumacher & Rapp, 2011) or motor and VS (Frewing, Tanner, Bonner, Baxter, & Pastrana, 2015; Pastrana et al., 2013). Schumacher and Rapp (2011) found that RIRD decreased immediate VS displayed by two children with ASD, but it did not alter (either increase or decrease) their subsequent engagement in VS. Similarly, Pastrana et al. (2013) found that RIRD decreased immediate engagement in targeted motor stereotypy for two children with ASD without altering subsequent engagement for either participant. Similarly, for an adolescent with ASD, Frewing et al. (2015) found that RIRD decreased immediate engagement in stereotypy (motor and vocal combined) without altering his subsequent engagement in stereotypy. Importantly, all three studies (a) signaled the availability of RIRD in Component 2 of the RIRD sequence and (b) used 10-min RIRD (second) components in the RIRD sequence. Thus, it is possible that a lengthier RIRD component, wherein the change agent does not signal RIRD implementation, could produce different subsequent effects.
In a brief review paper, Martinez and Betz (2013) noted that most studies had evaluated the effects of RIRD during 5- to 10-min sessions and therefore suggested that researchers evaluate RIRD during lengthier sessions and in naturalistic contexts. Subsequently, Wells, Collier, and Sheehey (2016) and Martinez, Betz, Liddon, and Werle (2016) showed that RIRD decreased VS for two boys with ASD during 20-min and 30-min sessions in naturalistic settings, respectively; however, neither study evaluated the subsequent effects of RIRD.
The purpose of the present study was to replicate and extend the Schumacher and Rapp (2011), Pastrana et al. (2013), and Frewing et al. (2015) studies by evaluating the immediate and subsequent effects of RIRD for VS and appropriate vocalization in two children with ASD. In addition, we compared the immediate and subsequent effects of RIRD on VS when therapists implemented RIRD without a visual signal during 5-min and 20-min components within a multiple schedule. Furthermore, we conducted some sessions for both participants in a naturalistic setting. As a supplement to our evaluation of the subsequent effects of RIRD, we included a fourth component to the three-component method (e.g., Lanovaz et al., 2010).
Method
Participants
Arun, a 7-year-old boy, and Jason, a 5-year-old boy, participated in the study. At the start of the study, each participant had a diagnosis of ASD and each was receiving 5 hr per week of applied behavior analytic services at a clinic in Chennai, India. According to the Verbal Behavior Milestones Assessment and Placement Program (VB-MAPP; Sundberg, 2008), Arun was a Level 2 learner who spoke in two- to three-word mean length utterances. He had a vocabulary of approximately 200 words including nouns, action words, and adjectives. He could make requests using two- to three-word phrases and respond to two-step receptive instructions. Arun engaged in VS in the form of repetitive, noncontextual vocalizations and unintelligible sounds (e.g., “weee”). Jason was a Level 1 learner based on a VB-MAPP assessment. His communication consisted primarily of one-word mands (e.g., “chips”) and tacts (e.g., “bus”). He could imitate vocalizations and fill in sentences with one word (e.g., he would say “spider” when the adult said “itsy bitsy”). His VS consisted of a series of sounds stated repeatedly (e.g., “ma-eh-go-ma-ki”).
Setting
All functional analysis (FA) sessions took place in a 4 m × 4 m treatment room in a clinic. Treatment evaluation sessions took place in either a 4 m × 4 m treatment room within the clinic or a community park. The treatment room contained a table, two chairs, and a wooden shelf to store material. The park contained a swing, a slide, a seesaw, a park bench, and a sandpit. We selected the park as the community setting due to its accessibility for treatment sessions and participants’ preferences (determined anecdotally through parent interviews). The first author conducted each session.
Dependent Variables
Trained observers collected data on participants’ VS and appropriate vocalizing (AV). Observers also collected data on RIRD implementation by the instructor. We defined VS as any noncontextual or repetitive vocalization, which included repetition of a string of sounds or words and movie/song lines. In addition, if a participant repeated any vocalization 3 or more times, observers scored it as VS. Observers scored VS using continuous duration recording with stopwatches. Observers scored VS with an immediate onset criterion and 1 s offset criterion. We converted data on the duration of VS to a percentage of the session measure by dividing the total number of seconds of engagement in VS by the total duration of the session. We defined AV as any contextual vocalization including requests for tangible activities and breaks, as well as spontaneous tacts or comments not directed by the experimenter. Across phases, the experimenter provided a response to each AV (e.g., access to the tangible activity requested, a break, or social praise). If an AV occurred repeatedly, it was scored as VS. We divided the total number of AVs by the session duration in min and reported the responses per min (rpm). Similarly, we converted the frequency of RIRD implementations to a rate measure by dividing the count per session by number of minutes in each session. The total duration of a session was measured using a second stopwatch. This stopwatch was not paused when RIRD was implemented.
Interobserver Agreement (IOA)
Sessions were videotaped using a tablet device and were later scored from video. A second observer scored 30% of sessions in each phase throughout the study. Two weeks prior to the start of the study, the first author provided observers with the operational definitions of VS and AV and trained them to identify examples and nonexamples. We calculated IOA scores for VS using a total duration IOA method (Cooper, Heron, & Heward, 2007). With this method, we calculated IOA scores for VS by comparing the total duration of VS scored by the primary observer with the total duration of VS scored by the secondary observer and dividing the smaller value by the larger value and multiplying by 100. The mean IOA scores for VS for Arun and Jason were 92% (range: 86-98%) and 95% (range: 89-97%), respectively. We calculated IOA scores for AV using the same method for total frequency per session. The IOA score was obtained by dividing the smaller value by the larger value and multiplying the quotient by 100. The mean IOA scores for AV for both Arun and Jason were 100%. We did not assess IOA for RIRD implementation.
FA
The first author conducted a FA of each participant’s repetitive vocalizations using demand, attention, NI, and play conditions (Iwata, Dorsey, Slifer, Bauman, & Richman, 1994). Sessions for each condition were 5 min in duration. We evaluated the effects of the conditions using a multielement design. During the demand condition, the experimenter delivered demands every 15 s. The demands consisted of instructions to either (a) pick up objects (e.g., “Pick up the car”) or (b) imitate motor actions modeled by the experimenter. These demands were part of the participant’s academic curriculum. Following each instance of VS, the experimenter provided a 15-s break. Contingent on correct responses, the instructor provided a brief verbal statement (e.g., saying “okay”) and proceeded to the next demand. The demand condition evaluated the extent to which the repetitive vocalizations were sensitive to social negative reinforcement in the form of escape or avoidance of demands. In the attention condition, the participant had continuous access to some toys including blocks, play-doh, a slinky, and toy vehicles. None of the toys produced auditory stimulation. Contingent on VS, the experimenter delivered verbal attention (“Don’t do that,” “No repeating sounds”) that lasted approximately 5 to 8 s. The attention condition evaluated the extent to which the repetitive vocalizations were sensitive to social positive reinforcement in the form of verbal reprimands. The toys used during the attention and play condition were those identified to be of moderate preference in an assessment conducted prior to the commencement of the study. In the play condition, several toys were continuously available in the treatment room, the instructor provided attention (e.g., saying “Nice job building a rocket, it looks so cool”) on a fixed-time (FT) 15-s schedule, and the instructor ignored instances of VS. The play condition served as the control condition. In the NI condition, the experimenter was present in the room, but did not initiate any interaction with the child (i.e., VS was ignored). The NI condition evaluated the extent to which repetitive vocalizations persisted in the absence of social reinforcement.
We considered a test condition differentiated from the play (control) condition if the data path from the test condition did not intersect with or trend toward the control data path for three or more sessions in each condition (Bartlett, Rapp, & Henrickson, 2011). Following the multielement assessment, we conducted a series of consecutive NI sessions for both participants. The purpose of the consecutive NI sessions was to (a) confirm the persistence of repetitive vocalizations in the absence of social contingencies (e.g., Querim et al., 2013) and (b) reestablish a baseline of VS without social consequences.
Procedures and Experimental Design
For each participant, we used a two-tiered nonconcurrent multiple baseline (NMBL) design across settings (treatment room and a park) with an embedded reversal design in the first tier to evaluate the effects of RIRD on VS and AV. Although guidelines for single-subject designs suggest that a two-tiered NMBL design is not sufficiently rigorous (e.g., Horner, Swaminathan, Sugai, & Smolkowski, 2012; Kratochwill et al., 2010), the embedded reversal design in the first tier allows for additional replications of behavior change (Coon & Rapp, 2018). In addition, based on the methodology used in Lanovaz et al. (2010) and Rapp (2007), each session within the baseline and RIRD phases contained a multiple schedule 1 with which we could also demonstrate experimental control of RIRD on VS. Depending on the phase, sessions (including all four components; see below) ranged between 20 and 35 min in length. We measured the number of RIRD implementations in the clinic setting only. We measured appropriate vocalizations only in the second component during all phases. The first author conducted sessions 3 days per week.
Baseline
Each session in this phase contained three, consecutive 5-min NI components and a fourth 5-min NI component that commenced 25 min after the completion of the third NI component. During each NI component, the experimenter was present but did not interact with the participant. Toys such as puzzles, blocks, play-doh, toy vehicles, animals, a dollhouse, and superhero figurines were available in a small shelf in the treatment room. The experimenter rotated toys randomly across sessions. Prior to the commencement of the study, the experimenter taught participants to request for items on the shelf. Access to these toys was not restricted during other times, and the experimenter did not provide demands during the sessions. If the child made a request for an item that could be delivered, the experimenter responded, saying, “Nice job asking for [item]” and provided the item. If the participant asked for a break, the instructor provided praise for asking, and provided a break. If the participant made a comment (e.g., a spontaneous tact of an audible sound), the instructor provided a verbal response (e.g., “Oh yes, that’s a [item]”). There were no programmed consequences for VS. The NI components served as a comparison against which we evaluated the effects of RIRD. In the 25 min between the termination of the third component and the commencement of the fourth component, therapists conducted regular academic training (e.g., motor imitation trials, block imitation trials, labeling of pictures, visual and auditory matching-to-sample).
The sessions in the community park were identical to those in the clinic setting with the exception that the toys present in the clinic were not available in the park. The park contained a swing, slide, seesaw, and a sandpit with a beach set. The child was allowed to either walk around freely or sit on a bench. If the child made a request, the experimenter provided social praise and led the child to the area (e.g., to the swing or the sandpit); however, the experimenter did not initiate any other interactions.
RIRD 5 min
Sessions in this phase were identical to those in the baseline phase except that an experimenter implemented RIRD during the second component of the RIRD sequence. Specifically, contingent on the occurrence of VS, the experimenter presented three different vocal demands (in no particular order) for the child to repeat a sound/word (e.g., “Say Bottle”). The first author preselected these sounds/words for the vocal demand from a list of 75 words. If the participant did not respond within 3 s, the experimenter repeated the demand once. In all instances, one repetition of the demand resulted in compliance. The experimenter provided vocal demands until the participant responded correctly to three consecutive vocal demands without emitting VS. 2 The toys described in the NI components were also present during the RIRD session. Requests for items resulted in social praise and access to the item. If the child was playing with a toy during the occurrence of VS, the experimenter presented vocal demands but did not remove the toy. The purpose of the second component was to evaluate the immediate effects of the RIRD procedures on VS and AV. We only measured AV in the second component (when RIRD was in effect). The purpose of the third and fourth components was to evaluate the extent to which the effects of RIRD persisted after the therapist discontinued RIRD. All components were 5 min in duration. In the park setting, the RIRD 5-min sessions were identical to those in the clinic.
RIRD 20 min
Sessions in this phase were identical to those in the RIRD 5-min phase except that we increased the duration of the second component to 20 min to evaluate the immediate and subsequent effects of a longer RIRD intervention on VS. All other components lasted 5 min and were identical to those used in the RIRD 5-min phase described above.
Data Analysis
We used visual analysis to evaluate the immediate and subsequent effects of RIRD on each participant’s VS using a between-sequence analysis (Lanovaz et al., 2010). To evaluate the immediate effects, we compared Component 2 of the baseline phases with Component 2 of the RIRD 5-min and RIRD 20-min phases. To evaluate the subsequent effects, we compared Component 3 of the baseline phases to the Component 3 of the RIRD 5-min and RIRD 20-min phases. Similarly, we compared Component 4 of the baseline phases with Component 4 of the RIRD 5-min and RIRD 20-min phases.
If VS is lower in Components 3 or 4 of the RIRD sequence than in Component 3 or 4 of the NI sequence, this would indicate that RIRD produced a subsequent decrease in VS. By contrast, if VS is higher in Component 3 or 4 of the RIRD sequence than in Component 3 or 4 of the NI sequence, this would indicate that RIRD produced a subsequent increase in VS. Alternatively, the pretreatment and posttreatment components may be undifferentiated revealing no subsequent effects of RIRD.
Results
Figure 1 shows the FA results for Arun (top panel) and Jason (bottom panel). For Arun, the data path obtained for VS during the NI condition (M = 91.25%) was higher and differentiated from the data path for the play (M = 66.5%) condition. In addition, the data paths for the attention (M = 16.25%) and demand (M = 28.25%) conditions were lower and differentiated from the play condition. Arun’s VS also persisted at a high level in the absence of social contingencies during the consecutive NI sessions (M = 92.25%). The results of Jason’s FA were similar to those of Arun. The data path obtained for Jason’s VS during the NI condition (M = 80%) was elevated and differentiated from the data path for the play condition (M = 60.75%). Similarly, the data paths for the attention (M = 24%) and demand (M = 14.25%) conditions were lower and differentiated from the play condition. Subsequently, his VS remained high (M = 81%) during the consecutive NI sessions. For both participants, the results show that VS (a) persisted in the absence of social consequences, (b) decreased when an instructor provided attention in the form of a brief reprimand, (c) decreased when an instructor provided academic demands, and (d) decreased, albeit to a lesser degree, when an instructor provided continuous access to toys and FT attention.
Percentage of session Arun (top panel) and Jason (bottom panel) engaged in vocal stereotypy across sessions in the attention, demand, no interaction, and play conditions of the functional analysis.
Figure 2 shows the results of the treatment evaluation for Arun. During the first baseline phase in the clinic setting (upper panel), the data paths for VS for all four components were relatively undifferentiated (Component 1, M = 97.2%; Component 2, M = 91.6%; Component 3, M = 94.6%; Component 4, M = 95.8%). During the first RIRD 5-min phase, VS became differentiated such that the data path for Component 2 was lower (M = 32.5%) than the data paths for the other components (Component 1, M = 93%; Component 3, M = 93.5%; Component 4, M = 94.3%); the data paths for Components 1, 3, and 4 were undifferentiated. During Component 2, the rate of Arun’s AV in the RIRD 5-min phase (M = 0.36 rpm) was slightly higher than during the baseline phase (M = 0.2 rpm), and the instructor implemented RIRD at a rate of 3.6 to 6 rpm (M = 4.8 rpm) per session. During the second baseline phase, the data obtained were similar to the first baseline phase. The data paths for all four components were undifferentiated (Component 1, M = 91%; Component 2, M = 95%; Component 3, M = 94.5%; Component 4, M = 97%), and the mean rate of AV was 0.4 rpm. In the second RIRD 5-min phase, VS was lower in Component 2 (M = 26.5%) than in the other components (Component 1, M = 97.5%; Component 3, M = 90.5%; Component 4, M = 93%); the data paths for Components 1, 3, and 4 were undifferentiated. During Component 2, the mean rate of AV was 0.5 rpm, and RIRD was implemented at a rate of 3.8 to 4.8 rpm per session (M = 4.3 rpm).
Percentage of session engaged in vocal stereotypy (primary y axis, upper and middle panels) and rate of appropriate vocalizations (secondary y axis, upper and middle panels) for Arun across baseline and response interruption and redirection (RIRD) phases in the clinic (upper panel) and park (lower panel) settings. Rate of RIRD implementation by the instructor during 5-min and 20-min second components in the clinic (middle panel).
In the first RIRD 20-min phase, the data paths for Component 2 (M = 16.2%) and Component 3 (M = 75.2%) were lower than and differentiated from the data paths for other components (Component 1, M = 94.6%; Component 4, M = 94.8%). In addition, the data path for Component 2 was lower and differentiated from the data path for Component 3. Arun emitted AV at a rate of 0.4 to 0.6 rpm (M = 0.48 rpm). The instructor implemented RIRD at a rate of 1.65 to 3.75 rpm per session (M = 2.39 rpm). During the third baseline phase, the data obtained were similar to those of the previous baseline phases (Component 1, M = 91%; Component 2, M = 91%; Component 3, M = 96%; Component 4, M = 91.5%), and the mean rate of AV was 0.2 rpm. In the second RIRD 20-min phase, Component 2 (M = 13.3%) and Component 3 (M = 72.3%) were lower and differentiated from each other and from Component 1 (M = 94.3%) and Component 4 (M = 93%). In addition, during Component 2, the mean rate of AV was 0.47 rpm, and RIRD was implemented at a rate of 1.5 to 1.8 rpm per session (M = 1.67 rpm).
The middle panel (Figure 2) shows the rate of RIRD implementation across sessions for Arun in Component 2 of the RIRD 5-min and RIRD 20-min sessions in this clinic. Results show that rate of implementation gradually decreased across sessions; however, the instructor typically implemented RIRD at least twice per min in each session.
The lower panel (Figure 2) shows the results for Arun in the community park. During baseline phase, Arun exhibited similar levels of VS as in the baseline sessions in the clinic; the data paths for the four components were relatively undifferentiated (Component 1, M = 96.7%; Component 2, M = 95.2%; Component 3, M = 95.5%; Component 4, M = 94.8%). In the RIRD phase, the data path for Component 2 (M = 68.8%) was lower than and differentiated from the data paths for Component 1 (M = 94.1%), Component 3 (M = 95.4%), and Component 4 (M = 93.8%). During Component 2, the rate of AV in this phase (M = 0.2 rpm) was comparable to baseline (M = 0.1 rpm).
Figure 3 shows the results of the treatment evaluation for Jason. During the first baseline phase, in the clinic setting (upper panel), the data paths for VS for all four components were relatively undifferentiated (Component 1, M = 80.8%; Component 2, M = 79.4%; Component 3, M = 81.2%; Component 4, M = 81.4%), and the mean rate of AV during Component 2 was 0.08 rpm. During the first RIRD 5-min phase, his VS was lower during Component 2 (M = 15.5%) than during Component 1 (M = 81%), Component 3 (M = 80.3%), and Component 4 (M = 82.5%); the data paths for Components 1, 3, and 4 were undifferentiated. The rate of Jason’s AV during Component 2 averaged 0.13 rpm per session, and RIRD was implemented at the rate of 2 to 5.2 rpm (M = 2.92 rpm). During the second baseline phase, the data paths for all four components were undifferentiated (Component 1, M = 80%; Component 2, M = 76.5%; Component 3, M = 81%; Component 4, M = 82%), and the mean rate of AV was 0.1 rpm. In the second RIRD 5-min phase, Jason’s VS was lower in Component 2 (M = 12.5%) than in Component 1 (M = 80%), Component 3 (M = 81.5%), and Component 4 (M = 82.5%); the data paths for Components 1, 3, and 4 were undifferentiated. The mean rate of AV was 0.3 rpm, and the instructor implemented RIRD at a rate of 1 to 3.2 rpm per session (M = 2.1 rpm).
Percentage of session engaged in vocal stereotypy (primary y-axis, upper and middle panels) and rate of appropriate vocalizations (secondary y axis, upper and middle panel) for Jason across baseline and response interruption and redirection (RIRD) phases in the clinic (upper panel) and park (lower panel) settings. Rate of RIRD implementation by the instructor during 5-min and 20-min second components in the clinic (middle panel).
In the first RIRD 20-min phase, the data paths for Jason’s VS in Component 2 (M = 15%) and Component 3 (M = 59.6%) were lower than and differentiated from the data paths for Component 1 (M = 79.2%) and Component 4 (M = 82.5%). In addition, the data path for Component 2 was lower and differentiated from the data path for Component 3. His AV ranged from 0 to 0.4 rpm (M = 0.2 rpm), and RIRD was implemented at a rate of 0.95 rpm to 3.6 rpm (M = 2.35 rpm) per session. During the third baseline phase, the data obtained for VS were similar to the previous baseline phases (Component 1, M = 84.5%; Component 2, M = 79.5%; Component 3, M = 78.5%; Component 4, M = 84.5%), and the mean rate of AV was 0.1 rpm. In the second RIRD 20-min phase, the data paths for VS in Component 2 (M = 9.3%) and Component 3 (M = 43.7%) were lower and differentiated from Component 1 (M = 94.3%) and Component 4 (M = 93%). As before, the data path for Component 2 was lower than and differentiated from the data path for Component 3. In Component 2, the mean rate of AV was 0.27 rpm, and RIRD was implemented at a rate of 1 to 1.7 rpm per session (M = 1.32 rpm).
The middle panel (Figure 3) shows the rate of RIRD implementation for Jason across sessions in Component 2 of the RIRD 5-min and RIRD 20-min sessions. As with Arun, results show that rate of implementation for Jason typically decreased across sessions, and the instructor typically implemented RIRD at least twice per min in each session.
The lower panel (Figure 3) shows the results for Jason in the community park. During the baseline phase, the data paths for Jason’s VS in the four components were relatively undifferentiated (Component 1, M = 81.7%; Component 2, M = 79.7%; Component 3, M = 80.8%; Component 4, M = 83%), and the mean rate of his AV was 0.07 rpm. In the RIRD phase, the data path for Component 2 (M = 11%) was lower than and differentiated from the data paths for Component 1 (M = 80.1%), Component 3 (M = 80.4%), and Component 4 (M = 82.6%) and was higher in the RIRD phase (M = 0.14 rpm) than in the baseline phase.
Discussion
Results of the FAs indicated that both participants’ VS persisted without consequences. Subsequently, results of the treatment evaluation showed that RIRD with either 5-min or 20-min sessions (a) decreased immediate engagement in VS for both participants, (b) slightly increased AV for Arun, and (c) did not alter AV for Jason. The immediate decreases in VS that were produced with RIRD replicate prior studies (e.g., Ahearn et al., 2007; Ahrens et al., 2011; Schumacher & Rapp, 2011), whereas the mixed outcomes for AV further contribute to the mixed findings in the literature. Results also showed that although RIRD implementation decreased across sessions in the 5-min and 20-min session for both participants, instructors needed to implement RIRD multiple times each session. Specifically, results from the RIRD 20-min phase suggest that RIRD became more efficient over time, as evidenced by the decrease in the rate of implementations across sessions. Moreover, the RIRD 20-min components produced subsequent decreases in VS during the third (but not the fourth) components.
Results of this study contribute to the extant literature in two ways. First, results of this study extend the Schumacher and Rapp (2011), Pastrana et al. (2013), and Frewing et al. (2015) studies by showing that RIRD with 20-min components produced subsequent decreases in VS whereas RIRD with 5-min components did not. As previously noted, there is substantial empirical support for immediate decreases in VS using RIRD; however, to our knowledge, no prior study has shown that RIRD alone can lead to a subsequent decrease in VS. To date, subsequent decreases in either vocal or motor stereotypy have previously only been demonstrated when researchers treated stereotypy using noncontingent reinforcement with structurally matched stimulation (Brogan, Rapp, Sennott, Cook, & Swinkels, 2018; Rapp, 2007; Rapp, Cook, McHugh, & Mann, 2017). Although most studies of treatments for stereotypy show no subsequent change in stereotypy, prior studies have shown that response blocking and verbal reprimands decreased immediate engagement in stereotypy but increased subsequent engagement (Rapp, 2006, 2007). Second, results replicate prior studies by Martinez et al. (2016) and Wells et al. (2016) by showing that RIRD can decrease VS in naturalistic settings.
Some specific findings from the FAs also warrant discussion. First, both participants displayed less VS during the attention condition than in the play condition. This finding is consistent with other studies that used mild verbal reprimands as an intervention to decrease nonsocially reinforced behavior (Cook, Rapp, Gomes, Frazer, & Lindblad, 2014; McKenzie, Smith, Simmons, & Soderlund, 2008; Rapp, 2007). Second, both participants displayed less VS in the demand condition compared with the play condition. This finding is consistent with the outcomes from a recent study by Cook and Rapp (2018), which found that some participants’ vocal and motor stereotypy decreased when practitioners provided standard instruction (i.e., prompts to evoke correct responding and praise for correct independent or prompted correct responding). Third, both participants displayed marginally lower VS in the play condition than in the NI condition. This outcome suggests that environmental enrichment with preferred items and FT attention could be a part of a multicomponent intervention for VS in nonacademic contexts. Collectively, the FA results indicate multiple options for treating the participants’ VS.
Although we did not design this study to identify the behavioral mechanism responsible for the subsequent decrease in VS during the RIRD 20-min components, there are at least two plausible accounts. First, it is possible that each RIRD implementation contained social interaction with or attention from the therapist. Previous studies have shown that social interaction involving auditory stimulation (e.g., continuous reading by an instructor) decreased immediate engagement in multiple forms of stereotypy for several children with ASD (Enloe & Rapp, 2014; Rispoli, Brodhead, Wolfe, & Gregori, 2018). In this way, the auditory stimulation produced during RIRD may have produced an abolishing operation (AO; e.g., Laraway, Snycerski, Michael, & Poling, 2003) for the auditory stimulation generated by VS (e.g., Rapp, 2007). However, this account seems unlikely because the subsequent effects became more pronounced while the RIRD implementation decreased across sessions.
Alternatively, RIRD may have functioned as a positive punisher for both participants’ VS. Unlike prior studies (e.g., Frewing et al., 2015; Martinez et al., 2016; Schumacher & Rapp, 2011), the instructor in this study did not provide a visual signal for when RIRD would be in effect. In the absence of an external antecedent stimulus, the conditions under which the instructor would implement RIRD may have been unclear to the participants, thus accounting for lower VS after the termination of the RIRD component. Typically, researchers endeavor to produce inhibitory stimulus control of stereotypy; however, studies have consistently shown that such antecedent control is difficult to produce (e.g., Cook et al., 2014; Martinez et al., 2016; Rapp, Patel, Ghezzi, O’Flaherty, & Titterington, 2009). As an alternative to inhibitory stimulus control by an external antecedent stimulus, Doughty, Anderson, Doughty, Williams, and Saunders (2007) showed that automatically reinforced behavior could come under the stimulus control of the first punisher delivered within a session, which they referred to as stimulus control of punishment (SDP). In this way, the change agent’s recent delivery of RIRD could have functioned as an SDP that inhibited some instances of VS. Given that we did not compare the subsequent effects of a signaled and nonsignaled 20-min RIRD component, this account is only speculative.
Some limitations of this study should be noted. First, we compared the effects of a 20-min Component 2 of the RIRD sequence with a 5-min Component 2 in the NI sequence instead of a 20-min Component 2 in the NI sequence. Although this could appear to be an inappropriate comparison, previous research has found that free or prior access to stereotypy may serve as an AO for subsequent engagement in the behavior (Lang et al., 2009; Lang et al., 2010; Rapp, 2004, 2007; Rispoli et al., 2014). Conversely, other studies have shown that simply blocking or restricting stereotypy increases engagement in stereotypy when the behavior is subsequently permitted (Rapp, 2006, 2007; Rapp, Vollmer, St. Peter, Dozier, & Cotnoir, 2004). Due to the effect of prior access to VS, we would expect the establishing operation (EO; e.g., Laraway et al., 2003) for engaging in VS to be higher in the first 5 min than in the last 5 min of a 20-min NI component. Conversely, due to restricted access imposed by RIRD, we would expect the EO for engaging in VS to be highest in the last 5 min than in the first 5 min of the RIRD sequence. Thus, using only a 5-min second component in the NI sequence as a comparison allowed for a conservative evaluation of the extended RIRD condition.
Second, although both participants’ VS decreased during RIRD components, each continued to engage in VS at low levels. Additional treatment components may be required if this low level of VS interfered with engagement in academic or social activities. Prior studies by Love, Miguel, Fernand, and LaBrie (2012) and Gibbs, Tullis, Thomas, and Elkins (2018) have found that combining RIRD with noncontinuous reinforcement (NCR; i.e., continuous access to music) decreased VS and increased on-task behavior for some participants with ASD. To that end, this study is also limited insofar as we did not determine whether RIRD increased academic performance for either participant. Nevertheless, as previously noted, results from the FAs suggest that both participants would have displayed lower levels of VS during demands. Finally, because we collected data on each participant’s VS using a low-tech method (i.e., recording the duration of VS with stopwatches), we were limited to calculating IOA with a less-than-optimal method. Similarly, we did not collect data on either treatment fidelity of RIRD implementation or social validity of the outcomes. Prior studies have socially validated the immediate decreases in VS produced with RIRD (Cassella et al., 2011; Gibbs et al., 2018; McNamara & Cividini-Motta, 2019), whereas the purpose of the present study was to determine if subsequent changes were detectable. To that end, a logical next step would be to determine whether subsequent changes are socially significant (i.e., detectable to stakeholders).
Some potential clinical implications of the study warrant discussion. First, it is important to note that we detected the subsequent decrease in VS for both participants by comparing the third components of the NI and RIRD sequences. Although the differentiation between these two data paths represents a potential benefit of RIRD, it is unlikely that instructors would rate the level of VS in Component 3 of the RIRD sequence as acceptable. Specifically, for both participants, the data path for Component 3 of the RIRD sequence was elevated and separated from the data path for Component 2 of the RIRD sequence. Second, when determining the level of empirical support of a given intervention, practitioners should consider positive and negative findings (e.g., Lanovaz & Rapp, 2016), as well as indirect benefits of the intervention. From a practical perspective, the treatment benefits produced by the subsequent decreases in VS, which is arguably an indirect benefit, may compensate for the extra time required to implement RIRD (nearly twice per min), which detracts somewhat from the overall positive finding.
Given the results from this study and others, researchers and practitioners who treat automatically reinforced behavior should evaluate the immediate and subsequent effects of the targeted as well as untargeted forms of stereotypy (Rapp et al., 2013). Specifically, Rapp et al. (2013) suggested that practitioners evaluate the immediate and subsequent effects of interventions with a three-component multiple schedule before implementing the intervention on a broader basis in instructional or home settings. Given our findings with the 20-min RIRD component, we suggest that practitioners calibrate the duration of the second component to reflect the duration for which change agents will provide the intervention. As suggested by the results, brief sessions may not be sufficient to detect subsequent changes (either increases or decreases) in either targeted or untargeted stereotypy.
Results of this study provide at least three avenues for future research. First, the current study detected subsequent decreases in stereotypy in the 5-min component that immediately followed a 20-min RIRD component but not in the 5-min component conducted 25 min later. Specifically, the data path for Component 4 of the RIRD sequence was undifferentiated from Component 1 of the RIRD sequence and comparable to Component 1 of the NI sequence, indicating that the subsequent effect had completely eroded. Thus, future research should evaluate the time course of the transition from treatment effect to baseline performance. Second, Hagopian, Rooker, and Yenokyan (2018) and Hagopian, Rooker, and Zarcone (2015) identified three subtypes of automatically reinforced self-injurious behavior (SIB) based on the results for the test conditions from FAs. Hagopian and colleagues showed that the FA outcomes served as strong to moderate predictive behavioral markers of the effects of reinforcement-based procedures on the SIB subtypes. Researchers should determine the extent to which this analysis is applicable to the assessment and treatment of stereotypy. Third, future research should evaluate whether signaled RIRD or nonsignaled RIRD produces different immediate and subsequent effects for targeted and nontargeted stereotypy.
Footnotes
Acknowledgements
We thank the families of the children who participated in the study. We also thank Priya and Pravina for their help with data collection.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
