Teaching mands for actions to children with autism spectrum disorder using systematic instruction, behavior chain interruption, and a speech-generating device

Abstract

Objective: Evaluate the use of procedures to teach three nonvocal children with autism spectrum disorder to use a speech-generating device to mand for actions.

Method: A multiple probe across participants design was used to evaluate the effectiveness systematic instruction and the use of a behavior chain interruption.

Results: All three participants acquired the target mand. However, generalization to an untrained item only occurred for one participant.

Conclusion: Results suggest that mands for actions expressed with a speech-generating device can be taught to children with autism spectrum disorder using systematic instruction and behavior chain interruption strategies.

Communication might be an end in its own right, but a well-established verbal behavior repertoire can also enable individuals to precisely contact and engage with others and their environments (Skinner 1957; Sundberg and Michael 2001). Perhaps due to a core deficit, the verbal behavior repertoires of individuals with autism spectrum disorder (ASD) often do not develop or are underdeveloped. It is estimated that about 25–30% of individuals with ASD do not fully develop spoken language and may benefit from the use of augmentative or alternative communication systems (AAC; Ganz et al. 2012; Tager-Flusberg and Kasari 2013; Wodka et al. 2013). Several studies have evaluated procedures for teaching basic mands (i.e. requesting preferred objects) repertoire to individuals with ASD who use various types of AAC systems (e.g. picture exchange, speech-generating devices, or manual sign).

In a review of the literature on interventions involving speech-generating devices (SGDs) for children with ASD, van der Meer and Rispoli (2010) found that 69.5% of the participants were taught to use the SGD for basic manding (i.e. activating a single icon from the screen to request access to a preferred object or activity). The results of these studies suggested that systematic instructional procedures could be effective in teaching children with ASD to use SGDs to acquire this basic type of manding skill. Specifically, 87% of the 23 included studies reported that the participants had successfully acquired the targeted mand(s).

After a child learns to use to use an SGD for basic mands, the next logical step would be to teach more advanced manding skills, such as manding for multiple items, manding for actions, and/or manding for information (Lechago and Low 2015; Raulston et al. 2013). In some cases, a mand for action, when compared to a basic mand, may be more precise and thus more likely to lead to a reinforcing outcome for the speaker. For example, if a speaker produced the mand Door, the function of that mand may be unclear to the listener as compared to mand for a specific action (e.g. Please open the door.). To reduce the probability of misinterpretations, and to enable the listener to provide the corresponding reinforcement for the speaker’s mands, there may be benefits to teaching advanced mands (e.g. mands for specific actions). Simply state, teaching a child to make a request that specifies a precise action that he or she wants the communication partner to perform (e.g. Please open the door).

Various procedures have been evaluated for teaching these types of advanced mands. One such procedure makes use of the behavior chain interruption strategy (BCIS). The BCIS is a naturalistic approach to creating opportunities for teaching communication skills. This procedure involves creating opportunities to mand by interrupting a chain of behavior (Carter and Grunsell 2001; Goetz et al. 1985; Hall and Sundberg 1987; Hunt and Goetz 1988). As demonstrated in a seminal study by Hall and Sundberg (1987) the BCIS may be a useful approach in part because it can effectively create the need, or motivation, for communication. For example, in this study, the authors first taught a behavior chain (e.g. how to make coffee or soup) that resulted in access to a corresponding terminal reinforcer (e.g. drinking coffee, eating soup). Once this chain was established, opportunities for teaching a mand were created by interrupting the behavior chain by hiding the needing materials (e.g. cup, water). These interruptions were used to create an opportunity to teach a mand for a missing item. More technically, the authors demonstrated that the need for the mand could be addressed by contriving the motivating operation (MO) through the interruption the behavior chain. Once the mand is reinforced by the listener taking the appropriate action, the speaker will have the necessary materials that enable the continuation of the chain and thus access the terminal reinforcer (Albert et al. 2012; Hall and Sundberg 1987).

Although research has been conducted evaluating the use of the BCIS in efforts to teach advanced skills, such as manding for information (Lechago and Low 2015; Raulston et al. 2013), few studies have investigated procedures for teaching children with ASD to mand for specific actions (Choi et al. 2010; Shillingsburg et al. 2013). For example, Choi et al. (2010) evaluated procedures for teaching two types of mands (i.e. requesting and rejecting items) in the context of two BCIS (i.e. the missing-item format and the wrong item format). In this study, four children with developmental disabilities were taught to use an SGD or a picture exchange communication system to make requests, some of which could be viewed as mands for actions (e.g. one child was taught to ask the listener to play a DVD). In another relevant study, Shillingsburg et al. (2013) taught five children with ASD to vocally mand for the listener to move away when that person was blocking the view of the television). Each participant acquired the action mand (i.e. asking the person to play the DVD and asking the person to move away from the television).

While each of these studies reported successful acquisition of mands for actions, there is a need for extending this work because there have been few studies in which participants have been taught to use an SGD to produce actions mands. Given that SGDs are increasingly being used and recommended as an AAC mode for children with ASD (Ganz 2015; Ganz et al. 2014; Schlosser and Wendt 2008), it would be useful to evaluate teaching procedures for teaching advanced manding skills to children who use SGDs.

To date, there is limited research extending the manding literature for children with ASD who use SGDs related to increasing the complexity. For example, Waddington et al. (2014) evaluated procedures for teaching a multi-step communication sequence using an iPad®-based SGD. In this study, systematic instruction was used to teach three boys with ASD to make a general mand for a preferred activity (e.g. ‘I would like a toy please’), then a specific mand (e.g. ‘I would like the Lego®.’ or ‘I would like the drawing pad.’), and finally to engage in a social response upon receiving the requested object (e.g. ‘Thank you.’). Results were mixed in that two of the three participants showed increases in performance of the three-step communication sequence. The third participant had required a procedural modification (i.e. changing the display type from a static to a progressive display) before he acquired the targeted mand sequence. These results suggest that it may be possible to teach more advanced manding skills to children with ASD using systematic instruction and an SGD.

The present study was designed to investigate the effects of implementing systematic instruction within a BCIS format for teaching mands for actions to three children with ASD. The study also aimed to evaluate generalization of the action mand to a novel stimulus within the behavior chain and to assess the maintenance of the newly acquired mand.

Method

Participants

Three children were selected from a pool of children receiving school-based behavioral intervention and attending a university-based clinical program. The children were receiving intervention at school and in the clinic, because they had been diagnosed with of ASD, had little or no functional speech, and had limited augmentative communication skills related to manding for actions.

Ryan was a 10-year-old boy with ASD. He scored a 42.5 on the Childhood Autism Rating Scale (CARS; Schopler et al. 1980), which indicated severe autistic symptoms. On the second edition of the Vineland Adaptive Behavior Scales (Vineland-II; Sparrow et al. 2005), he scored at the 1:11 (year: month) age equivalency level on the receptive domain and 2:1 on the expressive domain, which indicated low adaptive functioning. For written communication, he was scored at the 7:1 age equivalency level, which indicated a moderately low adaptive functioning. On the Verbal Behavior Milestone Assessment Placement Program (VB-MAPP; Sundberg 2008) he was rated at an emerging level two, meaning he was functioning at an age equivalency of about 18 months regarding his expressive language development. Specifically, he could use an SGD to make specific requests for items that were not visually present by naming the item or using its features (e.g. color, size, quantity). He could also use an SGD mand for approximately 15 missing, but needed items and would spontaneously mand for at least 50 different items.

Franny was a 13-year-old girl with diagnoses of ASD and Down syndrome. She scored a 46 on the CARS, which indicated severe autistic symptoms. On the Vineland-II, she scored at the 1:4 age equivalency level on the receptive domain and at the 1:3 age equivalency level on the expressive domain, which indicated low adaptive functioning. For written communication, she scored a 3:10 age equivalence, which indicated low adaptive functioning. On the VB-MAPP, she was at the beginning stages of level two (age equivalency of about 18 months). In particular, she was able to use an SGD and a picture-exchange communication system to mand for items not visually present, and mand for at least 15 different items with discrimination, but had not yet acquired mands for actions.

Seth was a 5-year-old boy with ASD. He scored a 43.5 on the CARS, which indicated severe autistic symptoms. On the Vineland-II, he scored at the 1:1 age equivalency level for the receptive domain and 0:8 for the expressive domain, indicating low adaptive functioning. For written communication, he scored a 3:5 age equivalency, which indicated a moderately low adaptive functioning. On the VB-MAPP, he was at an emerging level two for his manding repertoire (functioning at an age equivalency of about 18 months). Specifically, he was reportedly able to use an SGD to mand for at least five needed, but missing items (related to preferred toys and activities) and mand for at least 15 different items with discrimination. Before the start of the study, he had acquired one request for actions related to items (i.e. open) and used an SGD to perform this action mand across a variety of items that he needed to be opened.

Setting and sessions

Sessions for Ryan and Franny were conducted in a small conference room at their school. The room had a table and chairs, with cabinets and shelves. Sessions for Seth were held in a university-based clinic room, which contained a child-sized table and chairs, cabinet, and a two-way mirror. Each session involved the initiation and completion of one behavior chain (i.e. one activity) and lasted approximately 5 min. During each session, the participant sat across the table from the experimenter. A second person was present to assess the reliability and procedural fidelity and sat a few feet away from the participant. Two or three sessions were conducted per day and sessions were held 2 to 3 days per week.

Identifying preferred activities

A two-step preferences assessment (Kang et al. 2013) was conducted to identify preferred activities. First, each child’s parent and/or teacher was interviewed using an adaptation of the Reinforcer Assessment for Individuals with Severe Disabilities (RAISD; Fisher et al. 1996). The RAISD is a third-party interview protocol intended to identify preferred stimuli for persons with severe communication impairment. Second, preference assessment was conducted using four items or activities that were nominated as most preferred from the RAISD interview. The direct preference consisted of implementing a pairwise assessment protocol described by Fisher et al. (1992). Specifically, the participants were offered nine trials to choose between pairs of the items or activities covering the various combinations of items. The results indicated that Ryan and Seth always selected the activity of watching a video game and Franny always chose to watch a music video.

Behavior chains

Based on the results of the preference assessment, task analysis of the steps that were needed to complete the preferred activity was developed for each participant. The activity for Ryan and Seth was playing a video game. The activity for Franny was watching a music video. A task analysis was then developed for these preferred activities, which outlined the steps that were required to initiate and complete the activity to obtaining the terminal reinforcer. Within each task analysis (i.e. behavioral chain), a point of interruption was identified, which was intended as a way of creating the need for the child to produce the targeted action mand using the SGD. Each behavior chain required the participant to retrieve the device that they need to play a video game or watch a music video (i.e. iPad mini® or a laptop computer). Table 1 lists the steps of each task analysis or behavior chain. It should be noted that the devices used to play video games or watch music videos were not the same device used as the child’s SGD.

Table 1. Description of targeted behavior chains.

Participant	Behavior chain	Description	Terminal reinforcer
Ryan	iPad game	Remove lid from bin	Play game
		Get out iPad
		Open the cover
		*Turn on
		Select game
Seth	iPad game	Remove lid from bin	Play game
		Get out iPad
		Open the cover
		*Turn on
		Select game
Franny	Music video	Gets laptop	Watches video
		Opens case
		Gets laptop from case
		*Turn on laptop
		Selects video

^*Denotes point of interruption.

Speech-generating device

Each participant was taught to use an individually configured SGD to produce the action mand related to either playing a video game (i.e. Ryan and Seth) or watching a music video (i.e. Franny). The SGDs used were Apple iPad® minis equipped with the speech synthesizing application Proloquo2Go® (McNaughton and Light 2013; Sennott and Bowker 2009). The configuration of the SGDs and exact responses required for each participant were individualized based on their previous SGD use. Ryan used the keyboard feature of Proloquo2Go®, which is similar to a standard keyboard on a computer.

The iPad minis® Franny and Seth used as SGDs were configured with a combination of colored picture symbols from the set of icons within the Proloquo2Go® program database and photo images (i.e. laptop computer and iPad). Both of their devices were programmed as a static symbol display, with four category folders (2 × 2 grid).

Targeted responses

Using the keyboard display, Ryan was required to type out his targeted action mand, and then insert the response into the sentence strip by pressing the insert button, which appears at the top of the keyboard (see Fig. 1 for Ryan’s Typing display). He then activates the sentence strip to output the corresponding synthesized speech (i.e. ‘Unlock the iPad.’).

Figure 1.

SGD Configurations for Ryan (Typing display) and Franny (Selection display). Seth’s device (not displayed) was programmed similar to Franny’s with exception of the ‘Watch video’ folder.

For responding, Franny and Seth were required to make selection-based responses (i.e. selecting the correct folder, selecting the correct symbols, and navigating back to the main screen) after the behavior chain was interrupted. Figure 1 displays Franny’s programmed SGD, Seth’s display was similar, with the exception of a VIDEO GAME folder rather than WATCH VIDEO folder to correspond with his preferred activity. Each targeted response consisted of selected symbols, UNLOCK and IPAD or COMPUTER (i.e. as related to the child behavior chain). The first screen page showed four category folders (e.g. ACTIONS, MATH ACTIVITIES, PLAY WITH TOYS, and WATCH VIDEO). From the home screen, they were required to press the ACTIONS folder. Doing so brought them to screen with four symbols (i.e. BACK, UNLOCK, OPEN, and PUSH). From this screen, Franny and Seth had to press the UNLOCK symbol, and then navigate back to the home screen. Next, the children were to press the folder representing their preferred activity (e.g. WATCH VIDEO), which then brought them a final screen of symbols related to the behavior chain (e.g. COMPUTER). Lastly, they pressed the sentence strip at the top of the screen, where the selected symbols were listed, to activate the digitized speech output (e.g. Unlock iPad or Unlock computer).

Prerequisite assessment and training

After determining preference and developing the task analyses for the behavior chains, each participant was assessed to determine if he or she could correctly identify the device that was to be used in his or her respective behavior chain (i.e. the iPad® mini that Ryan and Seth used to play video games, laptop computer that was used by Franny to watch music videos). Three probe trials were given to each participant using a discrete trial format. Each participant scored 100% accuracy on these receptive probes. For Ryan, we also conducted tact trials to test his accuracy in labeling the items and spelling the names of the devices correctly via discrete trial. Upon typing the item name and activating the voice output, the experimenter looked at his device to also ensure the accuracy of his spelling. Ryan scored 100% accuracy on these tact probes.

Further, each participant was assessed and subsequently taught to complete their respective behavior chain without the presence of an interruption using least-to-most prompting procedure. Training on these steps consisted of the use of least-to-most prompting procedure and verbal praise for each completed steps. Two or three training sessions per day were given 2 or 3 times a week. Each session consisted of one full behavior chain with access to the terminal reinforcer (i.e. playing a video game or watching a music video) to ensure each participant had a history of contacting the terminal reinforcer without an interruption within the behavior chain. The start of each session was initiated by the experimenter saying Let’s play a game on the iPad or Let’s watch a video on the computer (i.e. depending upon the device the participant used for the preferred activity). This pre-training was completed to ensure each participant had a history of contacting the terminal reinforcer without an interruption within the behavior chain. Each participant learned to independently complete each step of his or her respective behavior chain within 12 training sessions.

Behavioral definitions and data collection

During each session, the experimenter recorded the degree to which the participant had independently emitted the targeted mand for the action that was needed to continue the behavior chain and obtain access to the terminal reinforcer (i.e. play the video game or watching a music video). An independent response was defined as the participant emitting the targeted mand for action (e.g. Unlock iPad, Unlock computer.) using their SGD within 5 s of the interruption. Independent mands were coded as having occurred at the IND level to indicate that the mand for action was presumably under the control of the motivating operation that arose when the activity was interrupted. If an independent response did not occur within 5 s, or if an incorrect response occurred during the 5 s interval (i.e. selecting an incorrect symbol from the array of four symbols), then the experimenter used a least-to-most prompting procedure to prompt an occurrence of the mand for action occurred. Prompted responses were recorded as G when a gesture prompt was required (i.e. gesturing toward the device or appropriate symbol), V when a vocal prompt was required (i.e. Press [symbol name]), P when a partial prompt was required (i.e. partial physical guidance to activate the correct symbols), or as F when a full physical prompt was required (i.e. the use of physical guidance with verbal prompt). During baseline, prompting was not provided. Any opportunity that ended without the participant making a response within 5 s of the interruption, or making an incorrect response were coded as IR (i.e. Incorrect Response). During the generalization sessions, data were collected using the same procedures stated above.

Experimental design

A concurrent probe multiple baseline design (Gast and Ledford 2009) was used to assess the effectiveness of the intervention on the acquisition of the targeted mands for actions. This design involved the following sequence of phases: (a) baseline, (b) intervention, (c) procedural modification (for Franny and Seth), (d) follow-up, and (e) re-training (Franny only).

Procedures

Within all sessions, the following variables were held constant: (a) Time of day (b) materials with the behavior chain, and (c) presence, location, and display settings on the SGD.

Baseline

In baseline, each session consisted of one opportunity to respond. Each session was initiated by the experimenter giving a relevant instruction (e.g. ‘Let’s play a game on the iPad.’) to initiate the start of the behavior chain. Verbal praise was delivered for each completed each step of the behavior chain before the interruption (e.g. ‘Nice job opening the iPad case.’). When the interruption occurred (i.e. The screen was locked screen and blocking access to the video game or music video), the instructor waited 10 s to determine if the participant would independently use the SGD to produce the required action mand (i.e. Unlock the iPad. or Unlock computer.). The participant was not prompted to produce the action mand, but an independent response would have resulted in the experimenter immediately performing the requested action to enable the continuation of the behavior chain. When the participant did not independently produce the action mand within 10 s, or if he/she produced an incorrect mand during the 10 s interval, then this response was ignored. After the 10 s interval, the experimenter took the necessary steps to end the interruption so that the behavior chain could continue. The non-contingent continuation was implemented to maintain the child’s interest in participating in the activity.

Intervention

Intervention sessions were similar to baseline except that when the behavior chain was interrupted, the participant was given 5 s to respond with the correct action mand. This shorter response/wait time was used to prevent errors. If an independent (IND) response occurred within this 5 s interval, then the experimenter performed the requested action so that the behavior chain could continue. If the required action mand did not occur within 5 s, then the experimenter followed the least-to-most prompting procedure until the participant emitted the action mand. The prompting hierarchy began with the experimenter giving a gesture prompt, followed by a vocal prompt, a partial physical prompt, and lastly a full physical prompt. This sequence was based on a prior evaluation of participants’ reactions to gesture, vocal, and physical prompts during the standardized assessments. For Ryan, the prompting hierarchy was slightly different in that he did not require full physical prompting. Prompted responses resulted in performing the necessary action and providing neutral verbal praise (e.g. ‘Sure, I’ll unlock the iPad.’), whereas independent responses (i.e. a correct action mand occurring at the IND level) responses resulted in performing the necessary action and providing enthusiastic verbal praise (e.g. ‘Nice asking all by yourself!’).

Procedural modifications

For Franny and Seth, additional learning opportunities and procedures were implemented because they showed variable responding during the initial intervention sessions. The modification involved providing five errorless learning trials before the start of each intervention session. Verbal praise was given after each correct response during the practice trials. Differential levels of verbal praise (i.e. neutral tone versus enthusiastic tone) were used for prompted versus independent (IND) level action mands promote prompt fading. After five trials, the participant was given brief access (i.e. approximately 30 s) to the terminal reinforcer associated with the final step of the behavior chain (e.g. playing a video game). The errorless practice trials were discontinued after the participant had three consecutive mands at the IND level.

Follow-up

Three follow-up sessions were conducted one month after the final intervention session for Ryan and Seth and eight weeks later for Franny. Procedures were the same as in baseline in that prompts were not used, incorrect responses were ignored, and after a 10 s interval, the instructor took the necessary steps to end the interruption to continue the behavior chain. Correct responses resulted in the experimenter performing the necessary action to continue the behavior chain and providing enthusiastic verbal praise (e.g. ‘Nice asking all by yourself!’).

Extra teaching phase

Franny received an extra teaching phase after her follow-up sessions. The procedures used were identical to her procedural modification phase.

Generalization probes

During each phase of the study, generalization probes were conducted to assess for generalization of the response to a different (untrained) locked device. Each generation probe was similar to baseline sessions in that the sessions were initiated by the experimenter giving a relevant instruction (e.g. ‘Let’s play a game on the iPod.’) to initiate the start of the behavior chain. For each participant, a different device was used for the activity. Specifically, for Ryan and Seth generalization was assessed within the video game behavior chain using an iPod rather than the original iPad. For Franny, generalization was assessed within the watching a music video behavior chain using an iPad, rather than the original laptop. Generalization sessions were conducted following the same procedures as baseline sessions. A response was correct if it used both the targeted action ‘Unlock’ and the untrained device name (i.e. iPod for Ryan and Seth and iPad for Franny). Thus, generalization sessions evaluated both response generalization (i.e. use of the action ‘Unlock’) and stimulus generalization (i.e. the use of an untrained device).

Procedural fidelity

During sessions when IOA was assessed, the independent observer also assessed whether the experimenter had correctly implemented the procedures using a checklist that described each step. The percentage of steps implemented correctly was calculated for each session. The mean percentage of correct implementation across sessions was 99% (range 92–100%).

Inter-observer agreement

A second observer independently collected data on the participant’s responses and level of prompting used during each phase of the study. An agreement was scored if the experimenter and independent observed had recorded the same data for each session, whereas any discrepancy was counted as a disagreement. The following formula was used to calculate a percentage for agreement in each session: Agreements/[Agreements + Disagreements] × 100% to determine the percentage of agreement for each session. IOA was collected on 41–95% of the sessions conducted for each participant and each phase of the study. For Ryan, the independent observer collected data on 95% of sessions with a mean agreement of 99% (range 80–100%). For Franny, the independent observer collected data on 87% of sessions with a mean agreement of 98% (range 80–100%). For Seth, the independent observer collected data on 41% of sessions with a mean agreement of 99% (range 80–100%).

Results

Figure 2 shows the type of response and level of prompting recorded by the experimenter during each session. Ryan (upper panel) did not independently produce the target response during baseline. During the intervention, he showed consistent independent responding after eight sessions. He also showed a high level of generalization. During follow-up, his action mands occurred independently (i.e. at the IND level) across all three of the follow-up sessions.

Figure 2.

Displays the action mand response by type (IND=Independent; G=Gesture; V=Verbal; P=Partial; F=Full) sessions for each participant.

Franny (middle panel) did not emit the target response during baseline. During the intervention, she did not reach criterion by Session 19, and therefore her intervention procedures were modified, as explained in the Method section. After 12 sessions with this procedural modification, she demonstrated independent (i.e. IND level) responding. She did not show generalization and did not maintain the targeted mand at the follow-up sessions. Consequently, she was given an extra teaching phase after follow-up during which she regained the targeted action mand at the IND level.

Seth (lower panel) did not emit the target response during baseline. After 40 intervention sessions, procedural modifications were implemented because of the variability of his responding pattern observed. With the procedural modification, he reached independent responding (i.e. performing the targeted action mand at the IND level) after five sessions. During generalization probes, on some occasions, he emitted the previously acquired action mand but did not produce the untrained form of the action mand (i.e. ‘Unlock the iPod’). He maintained independent manding (i.e. performing the targeted action mand at the IND level) during the three follow-up sessions.

Discussion

The results suggest an effective intervention approach for teaching children with ASD to use an SGD to mand for actions that enable them to access preferred activities. Acquisition of the mand was evident by participants showing independent responding (i.e. performing the targeted action mand at the IND level) across at least five consecutive sessions. Ryan showed the fastest acquisition and was the only participant who generalized the mand to a different device. Franny and Seth, however, showed little progress during their initial intervention phases, respectively. However, they reached the acquisition criteria after a slight procedural modification was made. In light of this delayed progress, the use of pre-session practice trials was effective for Franny and Seth. However, it should be noted that although Franny did show progress after the use of the procedural modification, it is not clear if the change in her behavior was due to the modification or her longer learning history. Additionally, maintenance of skills over time was shown in two out of the three participants (i.e. Ryan and Seth). Franny did not show maintenance and therefore received a final (booster-training) phase in which she recovered the targeted mand at the IND level.

These results have a variety of potential implications in regard to using the BCIS to teach action mands to children with ASD who have developed limited or no speech and are candidates for using SGDs. First, the results suggest that the BCIS and the modified least-to-most prompting procedures were effective in teaching the participants to produce the targeted mand. The targeted mand enabled the behavior chain to continue, thus enabling access to the presumed terminal reinforcer for the preferred activity (i.e. playing a video game or watching a music video). The positive outcomes are perhaps not surprising given that the intervention procedures involved the use of well-established instructional tactics. Contingent access to the next step in the chain was assumed to be an effective type of reinforcement and therefore likely to motivate the participants to engage in the sequence of responses to produce their action mand. The present results are consistent with other studies that have used similar procedures to teach SGD-based mands to children with ASD (e.g. Achmadi et al. 2012; Lorah et al. 2013; Sigafoos et al. 2013; van der Meer et al. 2013).

Our study also provides additional support for the use of a BCIS to create an effective way of contriving the need for communication. This is important as children with ASD may not be as motivated to communicate unless a specific need to do so is created (Sundberg and Michael 2001). Each participant’s learned behavior (i.e. mand for action) was most likely evoked by the interruption of the behavior chain and reinforced by the removal of the interruption to proceed to the next step of the chain. It is likely the behavior was under the control of the MO since it was noted that in each session the participants engaged in the completion of each remaining step of the behavior chain to reach and engage in the terminal reinforcer (e.g. playing the video game or watching the music video).

The present study extends previous research on mand training for children with ASD by its focus on teaching, what is arguably be viewed as a more advanced type of mand (i.e. mands for actions) and by the requirements for participants to progress through a sequence of responses using an iPad®-based SGD. It may also be important to continue the development of these repertoires by training more advanced mand forms since they enable the child to produce a precise response that is less prone to misinterpretation.

Although promising, the present results should be interpreted with caution for several reasons. First, the participants were not subjected to conditions in which an interruption did not occur. Thus, it is not possible to determine if the participants had also acquired the discrimination between when the mand for action should occur versus when it is not needed. In light of this limitation, future research may find it beneficial to assess conditions in which the interruption does not occur to ensure the learner’s mand is differentially sensitive to the relevant MO.

Additionally, the two participants who used picture symbols as their vocabulary mode (Franny and Seth) did not show adequate progress in the initial intervention phase. Therefore, their intervention procedures were modified. There are a few possible explanations as to why these two participants showed little progress during the initial intervention. First, it is possible that the increased response requirements (i.e. multiple symbols with screen navigation) required for the action mand may have accounted for their relatively slower acquisition. That is, perhaps the use of least-to-most prompting procedure was not as effective as the use of errorless prompting used during the practice trials (i.e. procedural modification). This may be due to their history of reinforcement of single symbol mands. It is possible that the new response sequence (i.e. the responses consisting of two symbols) may have caused slower acquisition as suggested by Lovaas (1977). It is important to note that the history of the characteristic consequence of the mand plays a significant role in strengthening the mand in relation to the MO (Skinner 1957). Future interventions might be improved by considering the need to navigate across screens/pages and use multiple symbol responses.

Another limitation is the lack of maintenance for Franny, which may indicate the need to train for greater fluency of acquisition skills for some individuals. Although it is unknown why she did not maintain the skill over time, it should be noted that when the extra teaching phase was implemented, her rate of acquisition was faster than her initial learning. It is possible that booster training was necessary because she needed more exposure to the intervention to become fluent in the mand and thus the booster training was effective because she was able to recoup the target mand, replicating the initial results of the intervention.

These results from the generalization probes were mixed in that Ryan spontaneously showed generalization to an untrained item in the context of the original behavior chain (i.e. manding for an iPod to be unlocked) and Franny and Seth did not show generalization. For some children, it may be that additional training of untrained items would be beneficial to promote generalization of an action mand. Further research related to generalization of action mands may be valuable.

Despite these limitations, the results are promising and suggest there may be value in future research aimed at replicating the intervention with additional participants, different behavior chains, and other types of action mands. Additionally, future research addressing generalization across other activities and communication partners would seem warranted.

Conflict of interest

No potential conflict of interest was reported by the authors.

Funding

The study was supported by a PhD Scholarship awarded to the first author from Victoria University of Wellington.

Notes on contributors

Amarie Carnett, PhD, BCBA-D, is an assistant professor in the Department of Educational Psychology at the University of North Texas. Her research focuses on communication and functional and adaptive skills for individuals with autism spectrum disorder.

Hannah Waddington, PGDipEPP, MEdPsych, is a PhD student at Victoria University of Wellington. Her research focuses on effective early intervention, particularly for children with autism spectrum disorder.

Alicia Bravo, MEd, is a PhD student at Victoria University of Wellington. Her research focuses on communication intervention for children with autism spectrum disorder.