Effects of Peer-Mediated Instruction to Teach Use of Speech-Generating Devices to Students with Autism in Social Game Routines

Abstract

Supporting social interactions between students with autism spectrum disorders (ASDs) and their typically developing peers presents many challenges. The purpose of this study was to investigate the effects of a peer-mediated intervention designed to teach two students with ASD to use speech-generating devices (SGDs) to engage in interactions with peers in a social context at school. Six peer confederates (three from each student with ASD's general education classroom) were taught to support SGD use during game activities. A multiple baseline design was used to examine the relationship between peer-mediated instruction and an increase in total communicative acts (CAs) by the two students with ASD. Results provide evidence that the confederates acquired the skills needed to support SGD use by students with ASD. The results also suggest that the intervention was effective at increasing total appropriate CAs by students with ASD. In addition, social validity ratings by all of the confederates were positive. Results are discussed regarding educational implications, limitations, and future research.

Keywords:

• Autism spectrum disorders
• Speech-generating devices
• Social interactions

Introduction

Numerous augmentative and alternative communication (AAC) approaches have been used to address the core communication impairments that are associated with autism spectrum disorders (ASDs). In recent years, speech-generating devices (SGDs) – programmable digital devices that provide voice output in the form of digitized or synthesized speech when activated – have been increasingly used in this regard. SGDs have a number of advantages over both manual signing and graphic communication displays that do not produce speech. SGDs provide an efficient means for obtaining a communicative partner's attention, which indirectly creates opportunities for the AAC user's needs to be heard and met (Romski & Sevcik, 1996). In addition, research suggests that the auditory output associated with SGDs is associated with more rapid graphic symbol learning in adults with developmental disabilities (Schlosser, Belfiore, Nigam, Blischak, & Hetzroni, 1995) and spelling acquisition in children with ASD (Schlosser & Blischak, 2001). Finally, in addition to providing information in the form of two modalities (both visual and auditory), SDGs provide a familiar, precise, and clear form of output to communicative partners who are able to speak (Romski, Sevcik, & Adamson, 1999; Schlosser, Sigafoos, & Koul, 2009). This may be especially important for interactions in school and community settings, especially when the AAC users are children with ASD in general education classrooms and interact with peers with typical development.

In a recent review of SGD research, van der Meer and Rispoli (2010) examined 23 studies involving a total of 61 individuals with ASD under the age of 18. Five of these studies (22%) utilized non-experimental designs that provided inconclusive evidence of SGD effectiveness. Of the remaining 18 studies, 72.2% focused on requesting (i.e., gaining access to desired items, asking for a break, or asking for more) as the primary communicative function that was taught via the SGD, and 11.1% focused on spelling instruction. Only three of the 18 studies (16.7%) focused on one or more communicative functions in addition to requesting; these included responding to questions (McMillan, 2008; Schepis, Reid, Behrmann, & Sutton, 1998); social commenting (Schepis et al., 1998; Trembath, Balandin, Togher, & Stancliffe, 2009); and rejecting (Trembath et al., 2009). Although all three studies were conducted in preschool or school settings, only one (Trembath et al., 2009) was aimed at increasing social interactions with peers rather than adults. This is unfortunate, as peers with typical development are ideal intervention agents for teaching social interaction skills to children with ASD because they serve as natural discriminative stimuli for social exchanges, thus promoting generalization and maintenance (Odom, Chandler, Ostrosky, McConnell, & Reaney, 1992).

Strain and his colleagues conducted numerous studies involving peer coaching to produce gains in the social behavior of children with ASD who were able to speak (e.g., Odom & Strain, 1986; Strain, Kerr, & Ragland, 1979; Strain, Shores, & Timm, 1977). More recently, Thiemann and Goldstein (2004) taught peers with typical development to increase social communication in five school-age children with ASD. Peers were trained in specific skills (i.e., answering questions, initiating conversation, complimenting) and received feedback regarding their use of these strategies during interactions with playmates with ASD. The overall rate of social interaction between the peers and children with ASD increased as a result of this intervention. Similarly, Harper, Symon, and Frea (2008) taught peers to use a variety of strategies (i.e., gaining attention, narrating play, reinforcing the target child, turn-taking) to initiate and facilitate play with two boys with ASD during recess. Both boys showed significant improvements in social initiations and responses that were maintained after the treatment was withdrawn.

In contrast to these studies of peer-mediated intervention in natural contexts, all of the SGD-ASD research to date – with the exception of the Trembath et al. (2009) study – has utilized adult-mediated instruction in structured contexts. However, no studies to date have involved peer-mediated instruction in school settings other than preschools. In addition, the study by Trembath et al. has a number of methodological shortcomings (see Nigam & Wendt, 2010), including failure to control for the types of activities played by the participants and their peers, and the absence of measures of the appropriateness of the communicative behaviors exhibited by the participants with ASD. The current study was designed to address these gaps by employing peer-mediated instruction with two elementary school children with ASD and to address the following questions: (a) can peers be taught to support SGD use by students with ASD in social play routines?, and (b) is there a functional relationship between peer-mediated support during social game routines and an increase in spontaneous appropriate communicative acts by students with ASD who use SGDs?

Method

Approval for this study was obtained from the Behavioral Research Ethics Board at the University of British Columbia. Potential participants were identified by the second author, who provided their parents with letters of initial contact and consent forms, and also acted as the confederate trainer in the study.

Participants

There were two types of participants in the study: students with ASD and peer confederates.

Students with ASD

Students with ASD were recruited into the study if they met the following criteria: (a) between 8 and 12 years of age; (b) formal diagnosis of an autism spectrum disorder; (c) able to understand basic spoken English; (d) able to speak no more than 20 words in English or another language; (e) able to make requests using at least 10 Picture Communication Symbols (PCS; Mayer-Johnson Co., 1994); (f) able to use an SGD for requesting but not for social interactions with peers; and (g) enrolled in a general education classroom in the participating school district.

Two boys with ASD, Ian and Max (pseudonyms), participated in the study. Consent was secured for a third student with ASD as well, but he engaged in moderate-to-high rates of self-injurious behavior during the initial two baseline sessions so a decision was made to discontinue his involvement.

Ian was 11;4 (years; months) when the study started. He was the youngest child in a Korean family and lived with his mother and older sister. Ian was diagnosed with autism by a multidisciplinary team at 4;5. He was enrolled in a regular Grade 6 classroom where he worked on a modified curriculum with the support of a special education assistant (SEA) who was assigned to him full time. A Peabody Picture Vocabulary Test–IIIA (PPVT; Dunn & Dunn, 1997) was attempted with Ian prior to the start of the study, but he was unable to obtain a basal score; no other standardized assessment results were available for him. The school speech-language pathologist assigned to him reported that he could read at approximately a Grade 1 level and spoke only three words. His social skill repertoire consisted of approaching others and using gestures to initiate interactions. Ian needed assistance with recreation activities, eating, and toileting, but was able to dress himself independently. He engaged in a number of problem behaviors, including bolting, screaming, hitting, and minor self-injury.

Ian had used a communication book with approximately 25 PCS in it for 4 years prior to the study; he used 10 of these symbols to make requests on a regular basis. Ian had learned to use the symbols with instruction according to the Picture Exchange Communication System (PECS; Bondy & Frost, 1994) and had reached Phase 4 of this protocol. Ian's individual education plan (IEP) contained goals related to his use of the device for requesting and social communication with peers.

Max was 11;1 when the study started. He was diagnosed with autism at the age of 2;3 by a multidisciplinary diagnostic team. No family background information was available. Max was enrolled in a regular Grade 6/7 classroom in a different school than Ian. He worked on a modified curriculum with the support of an SEA who was assigned to him full time. As with Ian, Max was unable to obtain a basal score on the PPVT-IIIA prior to the study onset, and no other standardized test data were available for him. He had no functional speech and emitted fewer than 10 echolalic utterances. He was not able to read. He initiated social interactions using gestures and physical direction. He was not independent in self-help, recreation, or leisure activities; and he required one-to-one support at all times. Max engaged in a number of problem behaviors, including hitting other people, hand flapping, jumping, and repetitive vocalizations.

Prior to receiving his SGD, Max had learned to use approximately 10 PCS symbols to request break activities and use of the bathroom, with instruction via PECS in which he had reached Phase 3. Max's IEP goal for the current school year was to learn to use his SGD in social situations.

Confederates

Three typically developing peers were nominated by Ian's and Max's teachers as confederates, using the following criteria: (a) enrolled as students in the same general education classroom as either Ian or Max; and (b) no identified social, cognitive, or behavioral problems that were likely to interfere with the study. After the parents of all six nominated students signed consent forms, the students themselves agreed to participate following an information session conducted by the second author.

Ian's confederates – Michelle, Colin, and Nathan (pseudonyms) – were all 12 years old and had been his classmates for 1 month prior to study initiation. Max's confederates – Megan, Luke, and Ron (pseudonyms) – were all 11 years old. Megan and Ron had been Max's classmates for 1 month and Luke had been his classmate for 1 school year prior to study initiation.

Setting and Trainers

All instructional and probe sessions occurred in rooms located in Ian's and Max's schools, during times selected by their classroom teachers. Both of the rooms included a large table and at least four chairs. All confederate training was provided by the second author (henceforth referred to as the “trainer”) and all videotaping, coding, and data analysis was performed by the first author. Participants' teachers and classroom assistants were not involved in the study.

Materials

Game materials

Different activities (i.e., games) were used with the two students with ASD to accommodate their individual interests and abilities, as per previous SGD studies aimed at promoting social interaction in such students (e.g., Schepis et al., 1998; Trembath et al., 2009). The games were similar with regard to the length of time and the number of player turns required to produce a winner; and both games required visual identity matching, a skill that both students had in their repertoires. Two different Bingo games with PCS were made for Ian using Boardmaker™ software (Mayer-Johnson Co., 2001); these included playground/recess Bingo and 2010 Olympic Games Bingo. The game materials included four bingo cards (one for each player), small paper squares used as bingo markers, and 25 pictures corresponding to the bingo cards. Max's materials included 18 pairs of matching Star Wars cards for a version of Concentration, a matching game. Edible treats were also used as reinforcement for all students with ASD and confederates at the end of each game session and included gummy candies, chips, and chocolate.

Speech-generating devices

Ian had received a Vantage Lite™1 SGD 2 months prior to the study onset, upon the recommendation of the provincial Special Education Technology team associated with his school district. He did not use Minspeak™2 icons on the Vantage; rather, he used it with PCS symbols programmed on activity-specific pages. For the study, Ian's SGD was programmed with a Bingo game page that included 15 spoken messages with corresponding PCS symbols, including: “Your turn,” “Pick one,” “I need a marker,” “Oh, no!,” “Bingo!,” “I win,” “I pick the treat,” “I almost had it,” “No cheating!,” “My turn,” “This one,” “I'm winning!,” and all confederates' names.

Max had received a Springboard Lite™3SGD 10 months before the study through the provincial Special Education Technology service and used it to make basic requests when he was either highly motivated or when prompted to do so. Like Ian, Max used his SGD with PCS symbols rather than with Minspeak icons, with vocabulary programmed on separate pages for specific activities. For the study, Max's SGD was programmed with a game page that included 17 messages with corresponding PCS symbols, including: “Yay, I pick the treats!,” “You're cheating!,” “No match,” “Your turn,” “my turn,” “How many do you have?,” “I got a match!,” “I saw that!,” “Who's next?,” “This is hard!,” “Oh, oh, I know, I know!,” “Give me a clue,” “I got it!,” “That's my favorite!,” and all confederates' names. None of the symbols or messages provided to either Ian or Max were familiar to them prior to the study. The message displays for both Ian and Max are depicted in Figure 1.

Figure 1. SGD displays for Ian (left) and Max (right).

Measurement

All sessions were videotaped and subsequently coded for the dependent variables. Coding began for Ian and Max when the first game card was drawn by any player and terminated at the natural endpoint of the game. Two sets of dependent variables were included: prompts to confederates by the trainer, and communicative acts produced by the students with ASD. Social validity was also measured at the conclusion of the study.

Prompts to confederates

During intervention Phase I (confederate training), the role of the trainer was to teach the confederates how to prompt SGD use by the students with ASD. Confederates were taught to wait for 3–5 s before providing one or more prompts each time an opportunity arose during Ian's or Max's turn in the game, so that they had time to initiate an SGD activation themselves. If no initiation occurred, confederates were taught to then provide verbal and/or gestural prompts for SGD activation. A verbal prompt was a spoken utterance designed to either get the attention of the student with ASD (e.g., Max!) or direct the student to activate a message on the SGD (e.g., Say ‘my turn’). A gestural prompt was a point or other hand motion toward the SGD that was provided for message activation.

Because confederates often provided more than one prompt in sequence or in combination, data were recorded on the number of prompt sets (rather than individual prompts) per session. A prompt set was defined as a sequence of one or more verbal and/or gestural prompts provided within 5 s of one another, with the goal of eliciting a CA (e.g., an SGD activation). Prompts that were required for game play rather than communication were not included, nor were confederate actions designed to orient the SGD toward a student with ASD. The rate of confederate prompt sets per minute was calculated by dividing the total number of confederate prompt sets by the total number of minutes in a session.

Communicative acts

This was the primary dependent variable for the students with ASD in the study. A communicative act (CA) consisted of a gesture, verbal utterance, vocalization, or SGD activation that was separated by at least 5 s from the previous CA; if multiple CAs occurred within a 5 s period, only the first one was coded. All CAs produced by students with ASD were transcribed from videotapes of each session. CAs were coded according to the mode of communication, as follows: (a) SGD activations that were directed toward a confederate (via eye contact or body orientation); (b) hand or upper extremity gestures that were directed toward a confederate, the activity materials, or the SGD; (c) verbal utterances that were recognizable as English words and were directed toward a confederate (Schepis et al., 1998); and (d) vocalizations that were not recognizable as English words but were directed toward a confederate (Schepis et al., 1998). Vocal stereotypic behaviors were excluded.

All CAs were also coded as either prompted (i.e., produced by a student with ASD as a direct result of corrective or directive attention provided by a confederate, including expectant pauses and verbal and/or gestural prompts) or spontaneous (i.e., produced by a student with ASD independently, without any type of prompt). Prompted CAs were further recorded as induced by either the trainer or by a confederate. The rate per minute of both spontaneous and prompted CAs was calculated by dividing the number of prompted (or spontaneous) CAs by the total number of minutes in a session.

Appropriateness

Data also were recorded regarding the appropriateness of each CA. Appropriate CAs were defined as those that were contextually suitable for the situation or interaction, while inappropriate CAs were those that were not contextually suitable. For example, an appropriate CA occurred if a student with ASD used his SGD to say “Your turn” to indicate a confederate's turn during the game. However, an inappropriate CA was recorded if a student with ASD activated his SGD to say “That's funny!” in the same situation. The rate per minute of both appropriate and inappropriate CAs was calculated by dividing the number of appropriate (or inappropriate CAs) by the total number of minutes in a session.

Social validity

At the end of the intervention, a short Likert-type questionnaire was completed by each of the confederates (Fawcett, 1991). The measure asked about confederates' perceptions of the study, including whether or not they enjoyed participating in the play sessions, their perception of the effectiveness of the play sessions for the students with ASD, and whether or not they would be willing to participate in similar interventions in the future. Responses were provided on a scale of 0 (not at all) to 5 (very much).

Treatment Fidelity

The first author examined 100% of the videotaped sessions and evaluated the extent to which the trainer implemented confederate training as designed (Moncher & Prinz, 1991). An implementation checklist was used to code the trainer's performance and a total percent correct score was calculated for each session. The checklist examined whether or not the trainer (a) set up the play materials on the table; (b) turned on the SGD, set it on the correct page with the volume turned up, and placed it in front of the participants; (c) provided feedback to the confederates on their performance during the previous session; (d) chose a player to start the game; (e) waited 5–15 s to prompt the confederates to prompt the student with ASD to use his SGD, if necessary; (f) waited 5–15 s to prompt the confederates to use the SGD themselves, when appropriate; (g) refrained from directly prompting communication by the student with ASD during the session; (h) allowed the game to continue to the natural end point; (i) provided an edible reward to the participants; and (j) cleaned up materials after the participants were dismissed. The mean fidelity score was 93% (range = 75–100%), indicating that the trainer adhered closely to the training protocol. Treatment fidelity related to confederate prompting was not examined.

Inter-Observer Agreement

Twenty-nine percent of the study videotapes were selected at random across baseline, intervention Phase I, and intervention Phase II sessions for assessment of inter-observer agreement (IOA). An observer who was blind to the purpose of the study was provided with operational definitions of all dependent variables and was trained until she obtained at least 90% accuracy (compared to researcher codings) across two consecutive study videotapes that were not used to assess IOA. For each dependent variable, IOA was calculated by dividing the number of agreements by the total number of agreements plus disagreements, multiplied by 100. Disagreements were resolved by consensus. Table 1 displays the IOA scores prior to disagreement resolution for the main dependent variables across all phases.

TABLE 1 Summary of Inter-Observer Agreement Data Across Dependent Variables and ASD Participants.

	Ian		Max
Dependent variable	Mean %	Range %	Mean %	Range %
Communicative acts (CAs)
Frequency of CAs*	96.3	90–100	94.7	91–100
Prompted	71.5	0–90	88.7	86–91
Spontaneous	90.6	67–100	91.7	75–100
Appropriate	100	N/A	96.3	89–100
Inappropriate	100	N/A	100	N/A
Prompts
Identity of prompter (trainer vs confederate)	98	88–100	100	N/A
CtoSA prompts	98	88–100	93.7	89–100
TtoC prompts	96.2	88–100	100	N/A

Note: All CAs in the sessions selected for IOA coding were SGD activations.Key: CtoSA = confederate to student with autism; TtoC = trainer to confederate.

The observer also coded 25% of the videotaped game sessions to examine the accuracy of treatment fidelity ratings. The observer was provided with the same implementation checklist used by the researcher to assess fidelity. IOA was calculated by dividing the number of agreements by the total number of agreements plus disagreements, multiplied by 100. The mean IOA for treatment fidelity for both Ian and Max was 96% (range = 92–100%).

Design

A multiple-baseline, multiple-probe design was used across the two students with ASD (Baer, Wolf, & Risley, 1968). The design included three phases: baseline, intervention Phase I (confederate training), and intervention Phase II (independent confederate implementation). Baseline data were collected for both students with ASD during play sessions with confederates. After a stable baseline was established for Ian, Phase I was initiated for him. Once the training criterion was met by Ian's confederates, Phase II commenced with him and Phase I was initiated with Max's confederates. When the training criterion was achieved by Max's confederates, Phase II began for him.

Procedure

Baseline

Baseline observations were conducted during game sessions until the data were stable over at least three sessions for one student with ASD. Baseline sessions varied in length, depending on how long it took for either a student with ASD or one of the confederates to win the game. On average, Ian's baseline sessions were 9 min long (range = 7–11 min.) and Max's were 11.5 min. long (range = 8–14 min.). During baseline, the trainer provided prompts for game set-up, established the rules for the game (when needed), and selected a player to start. The confederates and the students with ASD were then instructed to Play [game]. The SGD was turned on and placed on the table in a vertical position so that all participants could see and access it easily. The trainer remained nearby and provided hand-over-hand assistance to the student with ASD as required to keep the game/activity in play. The game continued to its natural endpoint (i.e., a player got Bingo in Ian's case, or all Star Wars matching pairs were selected in Max's case). The student with ASD and confederates were then provided with small edible treats and dismissed.

Confederate orientation

Following baseline, the trainer conducted one 15-min SGD orientation session with each group of confederates. The confederates were taught how to use the relevant SGD and had opportunities to practice activating it. They were also provided with basic information such as how to navigate to and from the game page, and were shown all messages on the game page and examples of situations in which each of the messages could be used. Confederates were informed that they would be responsible for encouraging SGD use by the student with ASD using verbal and/or gestural prompts, and each type of prompt was discussed and demonstrated. Each confederate then had the opportunity to practice using the SGD on his or her own. Orientation sessions occurred during a time selected by each classroom teacher.

Intervention Phase I

The focus of the first intervention phase was confederate training. Sessions in this Phase took place 2–4 times per week and varied in length, depending on how long it took for someone to win the game. On average, sessions for Ian and his confederates were 13 min long (range = 7–22 min) and sessions for Max and his confederates were 12.5 min long (range = 8–18 min). During this phase, the trainer set up the game materials and, during the first few sessions only, established the rules for the game. The SGD was turned on, placed on the table within reach of all participants, and set to the game page with the volume turned up. At the beginning of each training session, the trainer provided brief feedback to confederates regarding their performance during the previous session. A player was then chosen to start the game and the participants were instructed to Play [game].

During this phase, the trainer prompted confederates to model use of the SGD each time they took a turn and to prompt SGD activations by the students with ASD. Both verbal instructions and gestural prompts (e.g., pointing to a message on the SGD) were used by the trainer in this regard. Trainer prompts to the confederates for SGD modeling were provided following a 5- to 15-s time delay (Liber, Frea, & Symon, 2008) each time an opportunity for modeling arose; for example, the trainer might tell a confederate to activate “My turn” when it was his or her turn, or “Your turn” + a confederate's name at the end of each turn. Trainer prompts to confederates for SGD activation by a student with ASD were also provided a following a 5- to 15-s delay. If necessary, the trainer then provided a verbal prompt informing a confederate how to provide assistance (e.g., Call his name and tap the device, Point to ‘my turn’), and/or a gestural prompt toward the SGD, if needed. For example, if Ian was close to getting a “bingo” the trainer might prompt a confederate to point to the “I almost have it!” button on the SGD; or if Max got a match in the Stars Wars game, she might prompt a confederate to tell him to activate “I got a match.” Trainer prompts were faded gradually until the confederates were able to support SGD use by the students with ASD on their own. No prompts were provided by the trainer directly to a student with ASD at any time, unless required for game play. The game was allowed to continue until its natural end point (i.e., all matching pairs had been selected or someone got a “bingo”). The training phase was completed when the confederates required ≤0.2 trainer prompts/min over two consecutive sessions (i.e., ≤1.0 prompt/5 min).

Intervention Phase II

This phase differed from Phase I in that confederates supported SGD use by the students with ASD with minimal or no trainer assistance. Game sessions during this phase took place 2–4 times per week; Ian's sessions averaged 10.5 min in length (range = 8.5–13 min) and Max's averaged 10 min in length (range = 8–11 min). The initial set-up sequence during intervention was identical to that during Phase I. At the beginning of most Phase II sessions, the trainer provided brief feedback to confederates regarding their performance during the previous session; this feedback was faded and was not provided at all during the last few sessions.

During Phase II, the trainer provided prompts to the confederates only when they failed to model or encourage SGD use when an opportunity presented itself to do so, as described previously. The trainer provided no prompts to the students with ASD unless required for game play. The game was allowed to continue until its natural end point (all matching pairs had been selected or someone got a “bingo”).

Results

Intervention Phase I Results

Data in this section are related to question 1: Can peers be taught to support SGD use by students with ASD in social game routines? Figure 2 displays the results.

Figure 2. Prompts per minute by the trainer to confederates and by the confederates to students with autism.

During baseline, trainer prompts to the confederates (TtoC) were at zero across four sessions for both Ian and Max. Prompts by the confederates to the students with ASD (CtoSA) were at zero for Max and averaged 0.07 for Ian (range = 0–0.18). The prompts in Session 3 for Ian occurred when he began to activate buttons on his SGD at random, and the confederates encouraged him to activate an appropriate button instead. The prompts in Session 4 occurred when two confederates both called out the name of a game card and Ian repeated it.

Ian's confederates

Nine sessions were required for Ian's confederates to complete intervention Phase I. The rate of TtoC prompts showed an increasing trend for the first three training sessions and a decreasing trend for the final six, as the trainer began to fade her prompts to the peers. Overall, the mean rate of TtoC prompts was 0.41/min (range = 0.14–0.8). The training criterion of ≤0.2 prompts/min over two consecutive sessions was met for Ian's confederates in Sessions 12 and 13. During this phase, CtoSA prompts followed a pattern that was similar to TtoC prompts, with an initially increasing trend followed by a decreasing trend. Overall, CtoSA prompts for Ian occurred at a rate of 0.98/min. (range = 0.70–1.6) during Phase I.

During intervention Phase II, ToC prompts for Ian's confederates were low and stable, with a mean of 0.1/min overall (range = 0–0.38). The exception was Session 22, which occurred immediately after the Christmas break and in which the confederates required more prompting by the trainer, presumably because their instructional skills had deteriorated over the break period. Otherwise, the data show that Ian's confederates were able to implement the intervention during this phase with very minimal trainer support and required no trainer prompts at all during the final two sessions (Sessions 24 and 25). CtoSA prompts occurred at a stable rate during Phase II, averaging 0.57/min (range = 0.31–0.82).

Max's confederates

Seven sessions were required for Max's confederates to meet the training criterion for intervention Phase I. The rate of TtoC prompts increased with the initiation of training, remained stable for the first five sessions, and decreased during the final two sessions. Overall, the mean rate of TtoC prompts for Max was 0.34/min. (range = 0.09–0.57). The training criterion of ≤0.2 prompts/min over two consecutive sessions was met for Max's confederates in sessions 21 and 22. During Phase I, CtoSA prompts showed an increasing trend for the first five sessions and a decreasing trend for the final two. Overall, CtoSA prompts for Max occurred at a rate of 1.04/min (range = 0.67–1.34) during this phase.

During intervention Phase II, TtoC prompts for Max's confederates were very low and stable, with an average of 0.02/min (range = 0–0.09). Even after the Christmas break, Max's confederates were able to implement the intervention during this phase with very minor trainer support. As was the case for Ian, TtoC prompts for Max were at zero during the last two sessions of Phase II. In contrast, CtoSA prompts were more variable, averaging 0.66/min (range = 0.22–1.1).

Most of the prompted CAs required only a single confederate prompt set during both intervention Phase I (when the trainer was prompting the confederates) and intervention Phase II (when, for the most part, the trainer was not prompting the confederates). For Ian, 90.2% of CAs required only one prompt set during Phase I, and 80.6% required only one prompt set during Phase II; for Max, these proportions were 84.8% and 97.1%, respectively.

Intervention Phase II Results

Data in this section are related to question 2: Is there a functional relationship between peer-mediated support during social game routines and an increase in spontaneous appropriate communicative acts by students with ASD who use SGDs? Figure 3 displays the results.

Figure 3. Total appropriate communicative acts (CAs) by students with ASD.

Communicative acts for Ian

During baseline, Ian emitted no spontaneous appropriate CAs across four sessions. His average rate of prompted appropriate CAs during baseline was .023 CAs/min (range = 0–0.09). The only appropriate prompted CA occurred during the final session, when Ian repeated the name of a game card (“race”) after two confederates called it out.

Upon implementation of intervention Phase I, Ian's rate of prompted appropriate CAs increased immediately, averaging 0.81/min (range = 0.59–1.03). His rate of spontaneous appropriate CAs required three sessions to show change and then increased steadily over the next six Phase I sessions (overall mean = 0.24/min; range 0–0.54). The data during this Phase reflect the fact that, as Ian's confederates became more adept at providing him with support, he began to use his SGD spontaneously.

During intervention Phase II, Ian's spontaneous appropriate CAs continued to increase to an overall mean of 0.88/min (range = 0.78–1.06). Concurrently, his prompted appropriate CAs decreased gradually to a mean of .47/min (range = 0.21–0.70). Ian's data during this phase reflect his increasing ability to activate appropriate messages on his SGD without confederate assistance.

While the emphasis in this study was on appropriate CAs, both students with ASD did occasionally emit inappropriate CAs (i.e., CAs that were contextually inappropriate), especially in the form of SGD activations. Ian's rate of inappropriate spontaneous SGD use was very low during all three phases (baseline M = 0.08/min; Phase I M = 0.01/min; and Phase II M = 0.02/min).

Communicative acts for Max

During baseline, Max emitted no appropriate spontaneous or prompted CAs across four sessions. However, as was the case for Ian, Max's prompted appropriate CAs increased immediately following the initiation of intervention Phase I, to a mean of 0.77/min (range = 0.5–1.0). Spontaneous appropriate CAs also showed a gradual and immediate increase during this Phase, to a mean of .17/min (range = 0.07–0.33).

During intervention Phase II, both types of CAs showed considerable variability for Max, with a mean of 0.53 (range = 0.22–0.74/min) for prompted appropriate CAs and a mean of 0.30/min (range = 0–0.57) for spontaneous appropriate CAs. The variability in Max's data during Phase II reflects an issue that arose beginning in session 20 (Phase I) with regard to “cheating.” Max began cheating (i.e., taking cards out of turn) during this session, and also began to activate the “You're cheating” button on his SGD spontaneously. This was met with considerable confederate laughter and playful jostling, which apparently acted as positive reinforcement for Max, since his cheating and related spontaneous SGD activations increased further during Session 22 (see Figure 3). In Session 23 (the first session in Phase II), the Star Wars game cards were placed out of Max's reach because the cheating behavior was interfering with game play. Although this change reduced Max's attempts to cheat, it also resulted in a decrease in appropriate spontaneous CAs to zero during this session. In Session 24, the Star Wars cards were again moved closer to Max, and he again began to cheat and to activate the “You're cheating” message spontaneously on this device. Because his behavior continued to interfere with game play, his confederates were instructed (in Session 25) to ignore Max when he cheated and to look for opportunities to verbally reinforce appropriate game play instead. This again resulted in a decrease in problem behavior as well as a concurrent decrease in the rate of CAs for the remaining sessions.

With regard to inappropriate CAs, Max had a higher rate of inappropriate SGD activations during baseline than did Ian (M = 0.44/min), but this decreased considerably in Phase I to a mean of 0.1/min and in Phase II to a mean of 0.16/min. Thus, both students with ASD ended the study with lower rates of inappropriate spontaneous SGD use compared to baseline.

Mode of communication

In addition to rate per minute, appropriate CAs for both students with ASD were also coded to reflect the mode of communication by which they were produced: gesture, verbalization, vocalization, or SGD. During intervention Phase I, 97.5% of Ian's appropriate CAs were produced using his SGD; 25.8% of these were spontaneous. During Phase II, 97% of Ian's appropriate CAs were produced using his SGD; 63.4% of these were spontaneous. For Max, 96.2% of appropriate CAs during Phase I were SGD activations (13.8% spontaneous), and 95.2% of appropriate CAs during Phase II were SGD activations (31.7% spontaneous). For both Ian and Max, the remaining appropriate CAs were verbal utterances.

Percentage of Non-Overlapping Data

For single-subject experimental designs, the Percentage of Non-overlapping data (PND) can be applied as an outcome metric in addition to visual inspection of graphic results (Scruggs, Mastropieri, & Castro, 1987). The PND method requires the calculation of non-overlap between baseline and successive intervention phases by identifying the highest data point in baseline and determining the percentage of data points during intervention exceeding this level. PND scores can range from 0–100% and can be interpreted using the conventions set by Scruggs, Mastropieri, Cook, and Escobar (1986). A PND greater than 90% is indicative of a “highly effective” intervention, a PND between 70% and 90% reflects “fair effectiveness,” a PND between 50% and 70% is considered of “questionable effectiveness,” and a PND below 50% reflects an “unreliable or ineffective” treatment.

The PND for spontaneous CAs was 67% for Ian and 100% for Max in Phase I, and 100% for Ian and 80% for Max in Phase II. Overall, these scores suggest a fair to high level of change between baseline and the final phase of intervention. The PND for prompted CAs was 100% for both Ian and Max during both phases, indicating a high level of change overall for this measure.

Social Validity

At the end of the study, all six confederates (three for each student with ASD) completed social validity questionnaires with five statements that were rated on a scale from 0 (not at all) to 5 (very much). Both Ian's and Max's confederates indicated that they liked participating in the games (M = 4.67 for both) and would like to participate in activities of this nature again in the future (M = 5.0 for both). Both groups of confederates also indicated they felt good that they could show their classmate with ASD how to use his SGD (M = 4.0 for Ian and 4.33 for Max). Ian's confederates were less confident than Max's that the intervention would help him to play with other students in the class (M = 3.67 for Ian and 4.67 for Max). Neither group of confederates believed that they made new friends by participating in the study (M = 3.0 for both).

Discussion

This study employed a multiple baseline design to determine the effectiveness of a peer-mediated intervention to support communication by two students with ASD who used SGDs during social games. Results shown in Figure 2 indicate that confederates were successfully trained to support SGD use; however, because only two students with ASD were involved, and because the change between baseline and intervention was not dramatic for one of them (see Figure 3), it was not possible to establish a functional relationship between the intervention and changes in their communicative performance. A functional relationship requires a minimum of three demonstrations of the experimental effect at three different points in time across different participants (Horner et al., 2005). Nonetheless, there was a change in the level of spontaneous appropriate CAs for Max within one session and for Ian within three sessions, and this change did not occur until the Phase I intervention was initiated. These results, combined with PND scores of 80% or better for the primary dependent variables during Phase II, provide suggestive evidence in support of intervention effectiveness. This is only the second study to investigate the use of peer-mediated instruction to teach SGD use for social interaction with peers (see Trembath et al., 2009). It is the first study to control for the type of play activities, to measure both the spontaneity and appropriateness of communicative acts, and to involve school-age children.

Confederate Training

Overall, confederate training (including the initial SGD orientation session and intervention Phase I) required a total of 130 min for Ian and 96 min for Max. This corresponds favorably to previous peer-mediated intervention studies in which confederate training ranged from approximately 40–150 min (Harper et al., 2008; Strain et al., 1979; Thiemann & Goldstein, 2004). This modest investment of time resulted in confederates' ability to maintain high levels of virtually independent prompting throughout the intervention phase, where they required one trainer prompt or less per 5 min in order to support communication by students with ASD. The sole exception occurred for Ian during Session 17, the first session after a 2-week Christmas break, when his confederates required trainer prompting at a rate of 0.38 prompts/min (i.e., approximately one prompt per 3 min). After this “refresher” session, the trainer was able to return to a lower level of prompting. For both Ian and Max, the trainer was able to fade herself out completely during the last two intervention sessions, and the confederates were able to prompt the students with ASD with no support whatsoever.

In general, the prompts delivered by the confederates were very effective across both intervention phases, with only a single prompt set required for 80% or more of all prompted CAs. For Max, prompt effectiveness increased from 84.8% during intervention Phase I to 97.1% during Phase II; however, the opposite trend occurred for Ian, with a decrease in prompt effectiveness from 90.2% to 80.6%. This might have occurred for two reasons. First, Ian often displayed self-stimulatory behaviors that differed in frequency and topography from session-to-session and that sometimes caused him to disengage from the game for brief periods of time, which necessitated more confederate prompting. Second, Ian's confederates were very eager to prompt him during Phase I, and many of the trainer's prompts to them during this phase involved reminders to “wait it out” (i.e., provide a brief time delay) in order to give Ian time to make an SGD activation. During Phase II, when the trainer provided very few confederate prompts, the confederates sometimes failed to “wait it out” and provided more than one prompt to Ian in rapid succession.

Patterns of SGD Use by Students with ASD

During Phase I and Phase II, Ian's communication followed the anticipated pattern in that, as prompted appropriate CAs gradually decreased, spontaneous appropriate CAs gradually increased. Even after a 2-week hiatus for the Christmas break, Ian was able to maintain high rates of spontaneous CAs, executed primarily using his SGD.

For Max, the rate of spontaneous CAs was highly variable, especially during Phase II. This was primarily due to his “cheating” behavior and related spontaneous SGD activations of the “You're cheating” message, as described previously. It appeared that Max was more motivated to engage in communicative interactions related to cheating than those required for appropriate game play. Social reinforcement from confederates in the form of statements such as Thanks for waiting your turn, Max during appropriate game play was simply not equivalent to the reactions he received from his confederates for problem behavior (i.e., laughter and playful attempts to retrieve the “stolen” cards). In retrospect, a more thorough assessment of social preferences prior to the study might have revealed Max's enjoyment of interactions that generate peer reactions such as these. Had that been the case, joking comments and other messages designed to produce playful peer reactions could have been programmed into his SGD. However, despite the interference caused by Max's problem behaviors, the intervention appeared to be fairly to highly effective at increasing appropriate prompted and spontaneous CAs during social games for both Ian and Max, as reflected in their PND scores.

Limitations and Future Research

Only two students with ASD participated in this study which, as noted previously, constrained the assessment of experimental control. Only six confederates, who varied considerably in their ability to implement the intervention, were also involved. Future research should replicate this study with a greater number of students with ASD and confederates across a wider range of cognitive, social, and academic abilities.

A second limitation involved the placement of the SGD during the study. The confederates were encouraged to use the SGDs during their respective turns in the game, in order to model SGD use for Ian and Max. However, during most sessions, they passed the SGD from person-to-person in order to activate it, which meant that Ian or Max were not always able to see the message models they provided. Instead, it would have been preferable to keep the SGD stationery on the table so that the two students with ASD could view every message activation by their confederates, consistent with aided language modeling (Drager et al., 2006). Future studies of peer-mediated SGD instruction should explore the impact of aided language modeling more systematically by planning instructional arrangements that facilitate this technique.

In this study, the extent to which the trainer followed the instructional protocol related to confederate training and support was examined across a random sample of sessions. Treatment fidelity for the trainer was 93% overall, indicating close adherence to the protocol. However, treatment fidelity data were not calculated with regard to confederates' implementation of the instructional procedures they were taught to use (e.g., expectant delay, etc.). Thus, although the instruction provided by the confederates resulted in an increase in CAs for both students with ASD, it is not clear that this instruction reflected the intended procedures. Future studies should include a measure of confederate treatment fidelity.

Implications for Practice

The results of this study serve to emphasize the importance of focusing on AAC instruction for social interaction in addition to that related to requesting and labeling. Teaching students with ASD to make requests can be accomplished both efficiently and effectively through the use of non-technical AAC systems that require them to point to or exchange symbols representing desired items or activities (Wendt, 2009). However, while basic wants-and-needs instruction via PECS or other instructional approaches is often a good place to start, we need to remember that “there's more to life than cookies!” (Light, Parsons, & Drager, 2002, p. 187). It is interesting to note that the only previous AAC study that aimed at increasing social interaction between students with ASD and their peers (Trembath et al., 2009) also utilized an SGD. The voice output provided by an SGD is more conducive to social interaction than is a non-technical symbol-based system, because it allows for messages that convey excitement and emotions and thus promote communicative partner engagement (e.g., “Oh, oh, I know, I know!” and “You're cheating!”). Additional research is needed to examine the use of SGDs aimed at social communication goals for individuals with ASD in more detail.

Previous studies have documented the large caseloads of school-based speech-language pathologists (SLPs) (ASHA, 2009) as well as the shortage of SLPs with expertise in either AAC (Simpson, Beukelman, & Bird, 1998) or ASD (Schwartz & Drager, 2008). Given these constraints, it is imperative that the time spent by SLPs whose caseloads include students who require AAC be both effective and efficient. In this study, after a modest investment of time, six confederates were trained to use an SGD and to implement teaching strategies that included prompting and time delay with students with ASD. Essentially, this study created six “mini-interventionists” who are now present in the same classrooms as the two students with ASD, 6 hours a day for 5 days a week. While this study did not examine the extent to which the confederates were able to prompt and support SGD use outside of the intervention setting, the potential now exists for this to occur. In fact, anecdotal reports from both Ian's and Max's teachers and SEAs indicate that both boys often seek out the study confederates for interaction in their classrooms, and that they continue to play the study games and others with both the confederates and with other classmates who the confederates have “trained.”

This study provided preliminary evidence that elementary school-aged peers are capable of taking on the role of teaching students with ASD to use an SGD and that such instruction can result in positive communication outcomes. In the only other study to investigate this issue, Trembath et al. (2009) found that peer-mediated teaching using an SGD was functionally related to increases in communication by preschoolers with ASD during play activities. In contrast to Trembath et al., however, this study provided a more thorough examination of the communicative acts that were produced by the two students with ASD, including evaluations of modality, spontaneity, and appropriateness. For example, unlike Trembath et al., this study distinguished between SGD activations that were contextually appropriate and those that were not. In addition, while neither study controlled for the number of prompts per session, this study included documentation of prompt frequency as part of the analysis. Such documentation provides a more comprehensive understanding of the training process and may contribute to future peer training studies. Because educational goals for children with ASD almost always target both communication and social interaction, interventions focused on providing peer-mediated support in these two domains would seem to be especially valuable.

Acknowledgements

This study was completed in partial fulfillment of the requirements for a M.A. degree in special education at the University of British Columbia by the first author under the supervision of the third author. We are very grateful to Ian, Max, and their confederates for participating. We also want to thank the participants' teachers, special education assistants, and speech-language pathologists for facilitating their participation; Kavita Kamat for data coding; and Special Education Technology-BC for funding support.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.