Effects of an Augmentative and Alternative Communication Intervention Package on Socio-Communicative Behaviors Between Minimally Speaking Autistic Children and Their Peers

Abstract

Purpose:

The purpose of this study is to investigate the effectiveness of an augmentative and alternative communication (AAC) intervention package consisting of systematic instruction and aided modeling with speech-output technologies on the acquisition, maintenance, and generalization of socio-communicative behaviors—initiating a request for a turn, answering questions, and commenting—in four, minimally speaking (MS) autistic children between the ages of 6 and 9 years.

Method:

A multiple–probe design across behaviors replicated across participants was implemented to evaluate the effects of systematic instruction and aided modeling on initiating requests for a turn, answering questions, and commenting behaviors. Additionally, a pre- and posttreatment multiple-generalization-probes design was used to assess generalization across peers.

Results:

Visual analyses demonstrated experimental control for two participants (i.e., Derek, Ajay) showing a functional relationship between the intervention and outcomes across all social communicative behavior. For one participant (i.e., Matthew), experimental control could not be established because he did not reach the learning criterion for commenting. The fourth participant (i.e., John) transferred to a different school after making some progress on requesting. Effect size indicator analyses corroborated these findings, indicating medium-to-strong effects for initiating requests for a turn strong effects for answering questions, and medium-to-strong effects for commenting. Generalization of socio-communicative behaviors from researcher to a typically developing peer was variable across participants. Participants maintained socio-communicative behaviors 3 weeks after the last intervention session with varying degrees of success.

Conclusion:

The outcomes of this study suggest that aided modeling and systematic instruction using speech-output technologies may lead to gains in socio-communicative behaviors in some MS autistic children.

Supplemental Material:

https://doi.org/10.23641/asha.25799935

Approximately 30%–40% of children diagnosed with autism spectrum disorder (ASD) present with little to no functional speech and, thus, have difficulty developing functional communication (Skwerer et al., 2019). These children often face social and education isolation as well as significant frustration because they are unable to communicate their wants and needs, desires, knowledge, and emotions to their parents, siblings, peers, and teachers (Romski & Sevcik, 2005). To facilitate communication, these individuals often receive explicit instruction in the use of augmentative and alternative communication (AAC) strategies, aids, and techniques. AAC intervention aims to supplement or replace natural speech through unaided (e.g., gestures, manual signs) and/or aided approaches (e.g., speech-output technologies; Koul et al., 2001). Research indicates that minimally speaking (MS) autistic children can make substantial gains in functional communication skills, cognitive/conceptual development, literacy learning, and social participation through low-tech AAC systems (Ganz et al., 2012; Hart & Banda, 2010) and/or high-tech AAC systems (Morin et al., 2018; Sievers et al., 2018; Tincani et al., 2020).

For school-age, MS autistic children, AAC interventions often focus on developing basic communicative skills such as requesting, rather than facilitating acquisition and generalization of social communicative behaviors. Indeed, recent reviews have revealed that majority of AAC intervention studies have focused on teaching MS autistic children to request preferred items from an adult (e.g., Ganz et al., 2017; Lorah et al., 2021; Schlosser & Koul, 2015). For example, even though the last phase of the Picture Exchange Communication Systems (Bondy & Frost, 2001) targets commenting, systematic reviews indicate that most AAC intervention studies primarily focus on targeting requesting skills (Ganz et al., 2012; Gilroy et al., 2017; Logan et al., 2017). Though requesting is a foundational skill for communicators, it is a small percentage of what most individuals communicate daily (Ganz et al., 2017). Previous research has indicated that a substantial amount of communication is classified as small talk (Ball et al., 1999; King et al., 1995). Small talk is a type of conversational exchange used for initiating and maintaining conversation (Beukelman & Light, 2020). For instance, a study conducted by Ball et al. (1999) revealed that approximately half of typically developing preschool children's utterances in both home and school settings were classified as generic small talk, whereas only 18% of their utterances were classified as requests. Furthermore, 39% of utterances produced by typically developing adults in daily conversation were classified as small talk (King et al., 1995). Even though both typically developing children and adults engage in extensive small talk, little research has been conducted to determine the effectiveness of teaching socio-communicative behaviors (e.g., initiating conversation, commenting, answering questions) to MS autistic individuals. However, researchers have begun investigating the effects of aided AAC interventions in promoting symbolic turn taking (Drager et al., 2019; Therrien & Light, 2016) and conducting generic small talk (Chavers et al., 2021) in individuals with autism and other developmental disabilities.

For instance, Therrien and Light (2016) utilized a multiple-probe design (MPD) across partners with one replication to evaluate the effectiveness of an iPad¹ installed with GoTalk NOW² and dyadic turn taking training on the number of communicative turns taken between two preschool children with developmental disabilities and their same-age peers. Results showed that the intervention package consisting of turn taking training and provision of AAC was only effective for increasing symbolic communicative turns for one participant. The results for the other participant, however, did not support the effectiveness of the intervention as initial gains of the intervention were not maintained. Additionally, no generalization data were collected. Therefore, it is not clear as to the generalization of acquired behaviors to other activities, communication partners, or settings.

Additionally, Drager et al. (2019) investigated the effectiveness of an intervention consisting of aided AAC technologies with visual scene displays (VSDs) and “just in time” (JIT) programming on the frequency of symbolic communicative turns in nine students with developmental disabilities (age range: 8–20 years). Results indicated that the utilization of an AAC app that included VSDs and JIT programming is effective in increasing the frequency of communicative turns for all nine participants. However, generalization and maintenance data were not collected. Therefore, it is unknown if acquired behaviors were generalized to new environments or communication partners and if the acquired behaviors were maintained over a period of time.

Lastly, Chavers et al. (2021) evaluated the effects of systematic instruction (i.e., least-to-most prompting, constant time delay, error correction, reinforcement) on the acquisition, maintenance, and generalization of multistep requesting and generic small talk in three MS autistic children between the ages of 7 and 9 years. Results indicated that the intervention was effective in increasing multistep requesting and generic small talk behaviors in MS autistic children. Furthermore, participants generalized the acquired communicative behaviors to request untrained preferred snacks and activities and engage in generic small talk with familiar communication partners who were not a part of the intervention. However, preferred snacks and/or activities used in the generalization phase were not probed during the baseline phase, leading to less-than-conclusive evidence for generalization of target behaviors.

Two of the aforementioned studies (Drager et al., 2019; Therrien & Light, 2016) provide initial evidence that individuals with ASD and other developmental disabilities are able to utilize VSDs to engage in symbolic turn taking with varying degrees of success. However, these studies only evaluated the frequency of symbolic turn taking and failed to present data on the function of the symbolic turn (e.g., requesting, commenting, initiating conversation, asking a question). Successful social interaction encompasses a variety of communicative functions, such as commenting, asking questions, and greeting others. Therefore, research is warranted to determine if individuals with ASD and other developmental disabilities can utilize aided AAC technologies to engage in a range of social communicative behaviors with a variety of communication partners.

Furthermore, two of the three studies did not address generalization (Drager et al., 2019; Therrien & Light, 2016), and one study (Chavers et al., 2021) only collected posttreatment generalization probes. Thus, definitive conclusions on the generalization of social communicative behaviors cannot be drawn. Thus, research devoted to addressing the generalization and maintenance of acquired behaviors across communication partners, settings, or activities will provide greater insight into the strength of the intervention.

Overall, few studies involving speech-output technologies have implemented interventions to improve social communication between MS autistic, school-age children and their typically developing peers. Despite the very limited evidence base, there is potential for AAC to help in this regard. However, to do so comprehensively, further data investigating the most effective instructional strategies for eliciting social communicative behaviors are needed. Thus, the purpose of this study is to investigate the effectiveness of an AAC intervention package consisting of systematic instruction and aided modeling using speech-output technologies on targeted socio-communication behaviors between MS autistic children and their typically developing peers. For the purposes of this study, systematic instruction was composed of aided modeling, least-to-most prompting, positive reinforcement, constant time delay, and error correction. The aided modeling involved the communication partner providing language models using spoken language in tandem with aided AAC (Binger & Light, 2007; Drager, 2009). Specifically, the study addressed the following research questions: (a) What is the effect of systematic instruction and aided modeling using speech-output technologies on the acquisition and maintenance of socio-communicative behaviors—specifically initiating requests, answering questions, and commenting—with a researcher? (b) What is the effect of systematic instruction and aided modeling using speech-output technologies on the generalization of socio-communicative behaviors from the researcher to typically developing peers who will not participate in intervention?

Method

Participants

Four participants between the ages of 6 and 9 years were recruited for this study. Participants met the following inclusion criteria: (a) a diagnosis of ASD as determined by medical professionals or school personnel, (b) nonspeaking or MS (i.e., less than 20 functional, spontaneous words as determined by parent and teacher report), and (c) no physical or sensory impairments that would serve as a barrier to operating a speech-generating device (SGD). All participants were placed in self-contained classrooms in a public elementary school where they received individualized special education services. This study was approved by the institutional review board at the University of Texas at Austin (Study Number: 00002074). Parents of all participants in the study provided written, voluntary, informed consent. To protect confidentiality, each participant was assigned a pseudonym. Table 1 provides detailed information about the participants' clinical and demographic characteristics.

Table 1.

Clinical and demographic characteristics of participants.

Participants	Ajay	Matthew	Derek	John
Age (years;months)	8;4	6;11	6;5	9;7
Gender	Male	Male	Male	Male
Race	Asian Indian	Hispanic	White	White
CARS-2	Raw score: 42.5 Severity: severe	Raw score: 44.5 Severity: severe	Raw score: 44 Severity: severe	Raw score: 51.5 Severity: severe
TONI-4	SS: 84 Percentile: 14	SS: 78 Percentile: 7	SS: 75 Percentile: 5	SS: 61 Percentile: < 1
ROWPVT-4	SS: 63 Percentile: 1	SS: 63 Percentile: 1	SS: 87 Percentile: 19	SS: < 55 Percentile: < 1
ABDS	Conceptual domain: 40 Percentile: < 1	Conceptual domain: 40 Percentile: < 1	Conceptual domain: 57 Percentile: < 1	Conceptual domain: 40 Percentile: < 1
	Social domain: 40 Percentile: < 1	Social domain: 40 Percentile: < 1	Social domain: 40 Percentile: < 1	Social domain: 40 Percentile: < 1
	Practical domain: 65 Percentile: 1	Practical domain: 44 Percentile: < 1	Practical domain: 67 Percentile: 1	Practical domain: 40 Percentile: < 1
	Composite score: 42 Percentile rank: < 1	Composite score: 34 Percentile rank: < 1	Composite score: 49 Percentile: < 1	Composite score: 32 Percentile rank: < 1
Peer/s, age, and gender	Sarah, 8;1, female	Becca, 6;6, female Carlie, 6;9, female	Dana, 6;8, female	Ashley, 9;5, female
Preferred activity	Sticker scene	Bubble gun	Marble run	Bubble gun

Note. CARS-2 = Childhood Autism Rating Scale–Second Edition; TONI-4 = Test of Nonverbal Intelligence–Fourth Edition; SS = standard score; ROWPVT-4 = Receptive One-Word Picture Vocabulary Test–Fourth Edition; ABDS = Adaptive Behavior Diagnostic Scale.

Ajay was a 9-year-old Asian Indian boy with a diagnosis of autism and speech-language impairment. He received a Childhood Autism Rating Scale–Second Edition (CARS-2; Schopler et al., 2010) score of 42.5, indicating severe symptoms of autism. He obtained a standard score of 84, placing him in the 14th percentile (below average) on the Test of Nonverbal Intelligence–Fourth Edition (TONI-4; Brown et al., 2010). His Receptive One-Word Picture Vocabulary Test–Fourth Edition (ROWPVT-4; Martin & Brownell, 2011) standard score was 63, placing him in the 1st percentile. On the Adaptive Behavior Diagnostic Scale (ABDS; Pearson et al., 2022), Ajay received a standard score of 40 with a percentile rank of < 1 in the conceptual domain, a standard score of 40 with a percentile rank of < 1 in the social domain, and a standard score of 65 with a percentile rank of 1 in the practical domain. His composite score on the ABDS was 42 with a percentile rank of < 1. Five months prior to the study, Ajay's school speech-language pathologist (SLP) conducted an AAC assessment and determined that Proloquo2Go best fit Ajay's communication needs. Ajay was reported to navigate between two screens to request preferred items when given a direct model. Parents and his teacher reported that Ajay's spoken communication was primarily echolalic and included idiosyncratic vocalizations and/or one-word utterances. He demonstrated challenging behaviors such as pica (i.e., eating inedible objects) that appeared to be motivated by avoidance and escapism. His preferred activity used throughout the study was an animal sticker scene. His assigned peer was Sarah, an 8-year-old girl.

Matthew was a 6-year-old Hispanic boy diagnosed with ASD and speech-language impairment. He received a CARS-2 score of 44.5, indicating severe symptoms of autism. His TONI-4 standard score was 78 with a percentile rank of 7, indicating poor cognitive ability. His ROWPVT-4 standard score was 63, placing him in the 1st percentile. On the ABDS, Matthew received a standard score of 40 with a percentile rank of < 1 in the conceptual domain, a standard score of 40 with a percentile rank of < 1 in the social domain, and a standard score of 44 with a percentile rank of < 1 in the practical domain. His composite score on the ABDS was 34 with a percentile rank of < 1. Five months prior to the study, Matthew's SLP conducted an AAC assessment and determined that Proloquo2Go best fit Matthew's communication needs. His teacher reported that Matthew could navigate through two pages to make requests when given a direct model. Parent and teacher reported that Matthew primarily communicated by physically leading individuals to what he wanted and grabbing items within reach. He demonstrated challenging behaviors appeared to be motivated by avoidance (i.e., crying, laying on the floor). Matthew's preferred activity was a bubble gun. Matthew's first assigned peer was Becca, a 6-year-old girl. During the generalization phase for initiating a request for a turn, Matthew did not interact with Becca across two sessions. Thus, a new typically developing peer, Carlie (female, age 6 years) was assigned to Matthew for the remainder of the study.

Derek was a 6-year-old Caucasian boy with a diagnosis of ASD and speech-language impairment. He received a CARS-2 score of 44, indicating severe symptoms of autism. His standard score for the TONI-4 was 75 with a percentile rank of 5, indicating poor cognitive ability. His ROWPVT-4 standard score was 87, placing him in the 19th percentile. On the ABDS, Derek received a standard score of 57 with a percentile rank of < 1 in the conceptual domain, a standard score of 40 with a percentile rank of < 1 in the social domain, and a standard score of 67 with a percentile rank of 1 in the practical domain. His composite score on the ABDS was 49 with a percentile rank of < 1. One year prior to the beginning of the study, Derek's school SLP conducted an AAC assessment and determined that Proloquo2Go best fit Derek's communication needs. Teacher and parent reported that Derek had been using an iPad with Proloquo2Go for 1 year, but only used it to request preferred snacks and activities or label objects when given a verbal or gestural prompt. He could navigate through two pages to make a request or label objects when given a direct model. His primary mode of communication was gestures and unintelligible one- to two-word utterances. He demonstrated challenging behaviors such as self-injurious behaviors and aggression appeared to be motivated by avoidance. Derek's preferred activity was a marble run, and his assigned peer was Dana, a 6-year-old girl.

John was a 9-year-old Caucasian boy with a diagnosis of ASD and severe intellectual disability. He received a CARS-2 score of 51.5, indicating severe symptoms of autism. His TONI-4 standard score was 61 with a percentile rank of < 1, indicating poor cognitive ability. His ROWPVT-4 standard score was < 55 with a percentile rank of < 1. On the ABDS, John received a standard score of 40 with a percentile rank of < 1 in the conceptual domain, a standard score of 40 with a percentile rank of < 1 in the social domain, and a standard score of 40 with a percentile rank of < 1 in the practical domain. His composite score on the ABDS was 32 with a percentile rank of < 1. John's school SLP conducted an AAC assessment and determined that Proloquo2Go best fit John's communication needs. His teacher and parent reported that John had been using an iPad with Proloquo2Go for 2 years but only used it to request a preferred item (i.e., Pluto from Mickey Mouse) independently. He could not navigate through pages and relied on hand-over-hand assistance or gestural cues to communicate his wants or needs. His primary mode of communication was gestures and physically leading individuals to what he wanted. He demonstrated challenging behaviors such as self-injurious behaviors and aggression, which appeared to be motivated by avoidance and escapism. His preferred activity was playing a bubble gun, and his assigned peer was Ashley, a 9-year-old girl. Shortly after beginning intervention for initiating requests, John was transferred to a different school and, thus, did not complete the study.

Typically Developing Peers

Additionally, five age-matched, typically developing peers were recruited. Typically developing peers were nominated by teachers based on (a) age-appropriate social skills, (b) consistent school attendance, (c) ability to follow directions in a small group setting, and (d) willingness to participate. Additionally, all typically developing peers had to be enrolled at the same school as the participants. This study was approved by the institutional review board at the University of Texas at Austin (Study Number: 00002074). Parents of all peers in the study provide written, voluntary, informed consent.

Setting

The study took place at each participant's public elementary school. The participants' schools were selected as data collection sites in order to access typically developing peers and to enhance the ecological validity of the study. Sessions were 20–25 min in duration and occurred 3 times a week for approximately 3 months. Data were collected in a private corner of each participant's classroom that their teacher previously designated as a “work area.” This area included a set of lockers, a small table, two chairs, and a play rug. Each participant was positioned beside the peer or researcher. This allowed for the participant to see each icon the researcher or peer selected. The sessions were videotaped to collect treatment integrity and interobserver agreement (IOA) data.

Materials

Speech-output technologies is an umbrella term that includes dedicated devices such as SGDs and mobile technologies with AAC-specific applications that provide speech output (Schlosser & Koul, 2015). In this study, the latter were used as materials. Specifically, two iPad Pro tablets with Proloquo2Go³ were used. One iPad was utilized for aided modeling, and one iPad was used by participants. Proloquo2Go was utilized as each of the participants had prior experience using this AAC app. During the duration of the study, guided access of Proloquo2Go was not utilized. Additionally, SymbolStix (Clark, 1997) representing relevant vocabulary concepts served as materials. For symbol identification tasks, SymbolStix (Clark, 1997) was utilized as it is the default symbol set for Proloquo2Go.

The iPad was configured into a two-level page to target initiating a request for a turn and a two-level page to target the production of social communicative behaviors (i.e., answering questions and commenting). The first page consisted of the symbols for “I want,” “chat,” “core words,” and “quickfires.” The symbols for “core words” and “quickfires” were placed as foils and therefore did not produce speech output upon selection and were not linked to an additional page. The selection of “I want” led to a page consisting of symbols depicting action words related to the participant's preferred activity (e.g., go, throw, catch, stop, play).

The selection of “chat” led to a second page consisting of symbols for following messages: “Yes,” “No,” “I am happy,” “I am sad,” “Cool,” “I did it!”, “That's silly,” “This is fun!”, “Hi, how are you?”, “Do you like the game?”, “Let's play a game,” “Are you having fun?”, and “Is it my turn?” As social responses are dependent on the context of the conversation, foils were naturally in place. Word referents (i.e., gloss) were added beneath each symbol. Figure 1 depicts screenshots of the steps followed by participants to initiate a request, answer a question, or make a comment.

Figure 1.

Screenshots of the steps followed by participants to initiate request for a turn, answer, a question, a comment, respectively. Symbols included are SymbolStix ® Copyright © 2024 SymbolStix, LLC. All rights reserved. Used with permission.

Experimental Design

An MPD across behaviors (i.e., requesting turns, answering questions, commenting) replicated across participants was implemented (Ledford & Gast, 2018) to evaluate the effects of an AAC intervention, consisting of systematic instruction and aided modeling on initiating requests for a turn, answering questions, and commenting. There were four phases: baseline, intervention, generalization, and maintenance. The baseline sessions were implemented concurrently by the researcher and a peer across all behaviors. When the baseline data points were stable (i.e., no more than 5% variability on the dependent measures) for all behaviors, the intervention for initiating requests was implemented one on one between the researcher and the participant. Once the participant reached the acquisition criterion (three consecutive sessions averaging 80% or higher target responses), the intervention was implemented targeting answering questions. After reaching the acquisition criteria for answering questions, the intervention was implemented for commenting. The same procedure was replicated for all participants. One week after a behavior reached the criterion level (i.e., 80% across three consecutive sessions), the generalization phase was implemented.

A pre- and posttreatment multiple-generalization-probes design was used to assess generalization from the researcher to peers (Schlosser & Braun, 1994). Two weeks after the completion of the generalization phase, a maintenance phase was implemented.

Procedure

All participants were administered several standardized tests to ensure that they meet the inclusion criteria and to identify preferred stimuli for the intervention. Additionally, a preference assessment and symbol identification task were administered. Each participant's peer completed a brief training before administering baseline, generalization, and maintenance probes. Peer training is described in detail below.

Standardized Assessments

Participants were administered four formal assessments to determine their nonverbal intelligence, receptive vocabulary, adaptive behavior, and ASD severity. First, the primary researcher administered the TONI-4 (Brown et al., 2010) to assess each participant's nonverbal intelligence. Additionally, each participant was given the ABDS (Pearson et al., 2022) to assess their adaptive behavior. Participants were also given the ROWPVT-4 (Martin & Brownell, 2011) to assess their receptive vocabulary, followed by the administration of CARS-2 (Schopler et al., 2010) to determine autism symptom severity.

Preference Assessment

The preference assessment was conducted in two steps. First, the researcher completed an indirect preference assessment. The indirect preference assessment consisted of sending an e-mail to the participant's teacher and/or parents requesting a list of activities in order of most preferred to least preferred. The researcher provided examples of activities that could be included in the list, such as hands-on activities and simple board or card games. The participant's top five preferred games or activities were used as stimuli if they met the following inclusion criteria: (a) indoor activity, (b) restricted access, (c) allowed for two-people play, and (d) could be used in the context of the intervention session setting.

A free-operant procedure (Roane et al., 1998) was used to document each participant's most preferred activities during free time. A total of three sessions was conducted over a 3-day period with each participant. Each session lasted approximately 10 min. At the beginning of the first session, the researcher placed five preferred activities selected from the indirect preference assessment on a table or a carpeted play area. Participants were then instructed to engage in any activity that they like. The second session was conducted using the same procedure with the five remaining preferred activities. After completing the two sessions, the participant's preferred activities were ranked based on the longest duration (in seconds) of engagement. The participant's top five preferred activities were utilized in the last session to determine a final ranking of preferred activities. The researcher did not engage with the participants during preference assessment trials. The activities with the longest duration of engagement were considered as preferred activities.

Symbol Identification Task

SymbolStix (Clark, 1997) symbols were used to depict actions related to each participant's top five preferred activities/games and five nonpreferred or neutral activities as well as social communication phrases (e.g., “Yes,” “This is fun!”). Participants were required to identify these symbols. Each symbol identification trial consisted of a target symbol and three randomly selected foils. The participants were asked to point to the target symbol in response to the researcher's spoken instructions (i.e., “Which one is throw?”). If the participant selected the correct symbol, the researcher stated, “Great job,” and reemphasized that the participant selected the target symbol (i.e., “That is the picture for throw!”). If the participant selected the incorrect symbol, the researcher stated, “Nice try” and showed the participant the correct symbol. The participants were required to identify all target symbols for both requesting and social communication phrases with 100% accuracy across three consecutive trials before proceeding with the intervention.

Peer Training

The researcher provided peer training to each peer separately, during a 30-min session for 3 days in a quiet room. Peer training consisted of two components that focused on skills for using AAC with a friend and how to implement probes. During the using-AAC-with-your-friend component of training, each peer was taught to maintain proximity to the participant and how to use the Proloquo2Go app (e.g., navigating pages on the device, combining symbols, and using the volume button) and implement aided modeling. Before receiving instructions on how to implement the baseline, generalization, and maintenance probes, each participant had to navigate through pages on the device, identify all symbols, and combine symbols independently with 100% accuracy.

In implementing probes component of peer training, peers were taught how to deliver baseline, generalization, and maintenance probes for each dependent variable. For baseline, generalization, and maintenance probes targeting initiating a request for a turn, peers were instructed to state, “Let's play a game” via aided modeling and select a preferred game/activity to play with participant for 5 min. While playing the preferred game/activity, the peer was instructed to wait 5 s to give the participant an opportunity to initiate a request for a turn. After the participant received a turn, the peer was instructed to state, “I want [Name of ACTION]” via aided modeling and then wait for the participant to give them a turn. No cues were provided when administering probes. For baseline probes, the peer was instructed to give the participant a turn irrespective of whether the participant responded to the probe. For generalization or maintenance probes, if the participant responded to the requesting probe, the peer was instructed to give the participant a turn. However, if the participant did not respond correctly to the requesting probe, the peer was instructed to take another turn before administering another probe. After five probes targeting initiating a request were administered, the peer was instructed to conclude the session.

For baseline, generalization, and maintenance probes targeting answering questions, peers were instructed to begin the session by stating, “Hi! How are you?” via aided modeling and wait 5 s for the participant to respond. After waiting 5 s for a response, the peer was instructed to select a preferred game/activity to play and ask the participant, “Do you like this game?” via aided modeling and wait 5 s for the participant to respond. If the participant did not respond to the probe within 5 s, the peer was instructed to begin the game/activity. If the participant answered “yes” to the probe within 5 s, the peer was instructed to begin playing the game. However, if the participant responded “no” to the probe within 5 s, the peer was instructed to select another preferred activity/game and ask, “Do you like this game?” via aided modeling and wait 5 s for the participant to respond. After deciding on a game/activity to play together, the peer was instructed to engage in an activity/game with the participant for 5 min. While playing the game/activity together, the peer was instructed to ask the participant, “Are you having fun?” and “Is it my turn?” via aided modeling and wait 5 s for the participant to respond. No cues were given when administering answering questions probes. After five probes targeting answering questions were administered, the session concluded.

For baseline, generalization, and maintenance probes targeting commenting, peers were instructed to wait 5 s before delivering probes. Each peer was instructed to state, “Let's play a game” via aided modeling and select a preferred game to play with participant for 15 min. While playing the game, the peer was instructed to wait 5 s in between turns to give the participant an opportunity to comment about the activity (i.e., “That's silly,” “I did it!”, “This is fun,” and “Cool”). No cues were provided when administering probes. The session ended after the peer played with the participant for 15 min, and five probes targeting commenting were administered.

Before administering requesting, answering questions, and commenting probes to the participant, the peer had to independently administer 10 treatment probes targeting each dependent variable to the researcher with 100% accuracy. The peer also had the opportunity to discuss the procedures and ask questions to the researcher before delivering probes to the participant. While administering probes to the participant, the participant was given a visual aid and was cued by the researcher when to administer probes. A summary of procedures during baseline, intervention, generalization, and maintenance sessions is provided in Table 2.

Table 2.

Summary of baseline, intervention, generalization, and maintenance phases.

Baseline	Intervention	Generalization	Maintenance
Five probes targeting each communicative behavior (i.e., requesting social interaction, answering questions, commenting) administered by typically developing peer and experimenter No systematic instruction given	Five probes targeting a communicative behavior administered by the experimenter Systematic instruction given	Five probes targeting a communicative behavior administered by a typically developing peer No systematic instruction given	Five probes targeting a communicative behavior administered by a typically developing peer and the experimenter No systematic instruction given

Baseline Probes (Initiating Requests)

During baseline, the participant was given the iPad and Proloquo2go with an open display screen of the home page consisting of icons representing “I want,” “chat,” “core words,” and “quickfires.” The researcher or peer started the session by using aided modeling to state, “Let's play a game” and selecting a preferred game/activity to play with the participant for 15 min. While playing the preferred game/activity, the researcher or peer waited 5 s before giving the participant a turn in order to provide an opportunity for the participant to initiate a request for a turn. Irrespective of whether the participant responded correctly or incorrectly to the probe, the participant received a turn to engage in the game/activity. After the participant took a turn, the researcher or peer stated, “I want [Name of ACTION]” via aided modeling and then waited for the participant to give them a turn. No cues were given during baseline. After five probes targeting initiating a request were administered, the session concluded.

Baseline Probes (Answering Questions)

During baseline, the researcher or peer (pre–generalization probes) began the session by stating, “Hi! How are you?” via aided modeling and waiting 5 s for the participant to respond. After waiting 5 s for a response, the researcher or peer selected a preferred game/activity to play and asked the participant, “Do you like this game?” via aided modeling and waited 5 s for the participant to respond. If the participant answered “yes” or did not respond to the probe within 5 s, the researcher or peer began playing the game/activity. However, if the participant responded “no” to the probe within 5 s, the researcher or peer selected another preferred activity/game and asked, “Do you like this game?” via aided modeling and waited 5 s for the participant to respond. After deciding on a game/activity to play together, the peer or researcher and participant engaged in the game/activity for 5 min. While playing the game/activity together, the researcher or peer asked the participant, “Are you having fun?” and “Is it my turn?” via aided modeling and waited 5 s for the participant to respond. No cues were given when administering answering questions probes. After five probes targeting answering questions were administered, the session concluded.

Baseline Probes (Commenting)

During baseline, the researcher or peer stated, “Let's play a game” via aided modeling and selected a preferred game to play with participant. While playing the game, the researcher or peer waited 5 s in between turns to give the participant an opportunity to comment about the activity (i.e., “Good try,” “I did it!”, “This is fun,” and “That's cool!”). No cues were provided when administering probes. The session concluded after the peer or researcher played the game with the participant for 15 min, and five probes targeting commenting were administered.

Intervention

For all dependent variables (i.e., initiating a request for a turn, answering questions, and commenting), the intervention phase was identical to the baseline phase, except only the researcher implemented systematic instruction (i.e., least-to-most prompting, constant-time delay, error correction, reinforcements) and probed the responses. For example, if the participant did not initiate a request for a turn within 5 s, the researcher would begin systematic instruction. Systematic instruction began with the researcher providing the participant with a verbal cue followed by a constant-time delay of 3 s. If the participant did not respond to the constant-time delay within 3 s, the researcher provided both verbal and gestural prompts. Lastly, if the participant did not respond within 3 s of providing the verbal or gestural prompt, the researcher used aided modeling to show the participant how to respond to the probe. The researcher immediately used an error-correction procedure if the participant activated the incorrect icon. This consisted of the researcher providing verbal and gestural prompts to activate correct icons. (see Supplemental Material S1 for an example script of the intervention).

Acquisition Probes

Probes were utilized to measure the participant's acquisition of the targeted communicative behaviors (e.g., initiating a request for a turn, answering personal questions, commenting). Each session was composed of five acquisition probes targeting initiation of requests for a turn, five probes targeting answering questions, or five probes targeting commenting. The acquisition probe for initiating a request for a turn consisted of the researcher waiting 5 s to give the participant an opportunity to initiate a request for a turn. If the participant responded to the probe independently and within 5 s, the researcher gave the participant a turn. If the participant did not initiate a request for a turn, the researcher took another turn before administering another probe. No cues were provided when administering probes targeting initiating requests for a turn. After the participant reached a criterion of 80% accuracy across three intervention sessions for initiating requests for a turn (Tier 1 of the MPD), intervention targeting answering questions began. One week after reaching the acquisition criterion for initiating requests for a turn, the generalization phase was implemented.

Probes targeting answering questions consisted of the researcher stating, “Hi! How are you?” via aided modeling and waiting 5 s for the participant to respond. After waiting 5 s for a response, the researcher selected a preferred game/activity to play and asked the participant, “Do you like this game?” via aided modeling and waited 5 s for the participant to respond. If the participant answered “yes” or did not respond to the probe within 5 s, the researcher began playing the game/activity. However, if the participant responded “no” to the probe within 5 s, the researcher selected another preferred activity/game and asked, “Do you like this game?” via aided modeling and waited 5 s for the participant to respond. After deciding on a game/activity to play together, the researcher and participant engaged in the game/activity for 15 min. While playing the game/activity together, the researcher asked the participant, “Are you having fun” and “Is it my turn?” via aided modeling and waited 5 s for the participant to respond. No cues were given when administering answering questions probes. After the participant reached a criterion of 80% accuracy across three sessions for answering questions (Tier 2 of MPD), intervention targeting commenting began (Tier 3). One week after reaching the acquisition criterion for the answering questions behavior, the generalization phase was implemented.

To administer a commenting probe, the researcher began each session by stating, “Let's play a game” via aided modeling and selecting a preferred game to play with the participant. While playing the game, the researcher waited 5 s in between turns to give the participant an opportunity to comment about the activity (i.e., “That's silly,” “I did it!”, “This is fun,” and “Cool!”). No cues were provided when administering probes. The session concluded after the researcher engaged in the activity/game with the participant for 15 min, and five probes targeting commenting were administered. One week after reaching the acquisition criterion for the commenting behavior, the generalization phase was implemented.

Generalization Probes

The generalization probes were implemented 1 week following the last intervention session for each behavior (requesting a turn, answering questions, and commenting). The procedure for this phase was identical to the procedure in the baseline phase for both requesting, answering questions, and commenting behaviors, except that the peer instead of the researcher administered probes.

Maintenance Probes

Maintenance probes were implemented 2 weeks after the generalization phase. Each maintenance session consisted of five probes. The procedure for this phase was identical to the baseline probes for requesting, answering questions, and commenting, except that the peer and the researcher probed the dependent variables.

Dependent Measures and Definitions

The dependent variables included (a) initiating a request for a turn, (b) answering questions, and (c) commenting. The dependent measures were the percentage of correct responses when initiating a request for a turn, answering questions, and commenting. When initiating a request for a turn, the operational definition for a correct response included independently selecting (a) the folder labeled “I want to” on the first page, (b) the specific symbol that represents the preferred action on the second page, and (c) the message bar to activate speech output within 5 s of the researcher's or peer's probe. For answering questions, the operational definition included the participant selecting (a) the “chat” folder on the first page, (b) the socially appropriate symbol on the second page in response to the communication partner's utterance (e.g., “Yes,” “No,” “I'm happy,” “I'm sad”), and (c) the message bar to activate speech output within 5 s of the researcher's or peer's probe. For commenting, the operational definition included the participant (a) the “chat” folder on the first page, (b) the socially appropriate symbol on the second page in response to the communication partner's utterance (e.g., “That's silly,” “This is fun!”, “Cool,” “I did it”), and (c) the message bar to activate speech output within 5 s of the researcher's or peer's probe.

Responses were counted as correct if the participant used his or her finger to touch the symbol that corresponded to the selected item with enough pressure followed by selecting the message bar to activate speech output. Incorrect responses included (a) pressing at least one incorrect icon within the entirety of the sequence; (b) selecting icons multiple times, resulting in repetitive speech output; (c) selecting the home screen key to exit out of the AAC application; and (d) requiring prompts to select correct icons or taking longer than 5 s to respond to a probe.

IOA and Procedural Integrity

An independent observer, who was an undergraduate student in speech-language pathology, was trained to collect IOA data. The general purpose of the study and the operational definition of the dependent variables were explained to the observer. One video from the baseline phase and one video from the intervention phase for each behavior were randomly selected for training purposes. Training continued until there was 100% agreement between the researcher and the observer. After training was complete, the observer collected data on at least 30% of baseline, intervention, and generalization probes. IOA was calculated by dividing the number of agreements by the number of agreements plus disagreements multiplied by 100. IOA was 100% accuracy across all phases, behaviors, and communication partners.

Nine separate checklists—three for baseline procedures, three for intervention procedures, and three for generalization procedures—were developed to assess procedural and treatment integrity (see Supplemental Materials S2–S10; Schlosser, 2002). One video from the baseline phase and one video from the intervention phase for each behavior were randomly selected for training purposes. The training continued until the observer reached 100% accuracy in collecting procedural and treatment integrity data. After the training, procedural and treatment integrity data were taken for 30% of baseline, intervention, and generalization sessions. Procedural/treatment integrity was calculated by dividing the number of correctly performed steps by the total number of steps multiplied by 100. Procedural integrity for the baseline phase targeting requesting averaged 99.2% accuracy (range: 96%–100%) across communication partners. During the baseline phase for answering questions and commenting behaviors, procedural integrity averaged accuracy of 100% and 98.9% (range: 96.5%–100%) across communication partners, respectively. Procedural integrity was 100% across intervention and generalization phases for researcher and communication partners.

Data Collection and Analyses

Data were analyzed and interpreted within and across phases for both behaviors for each participant using level (i.e., the data points around the vertical axis), trend (i.e., the direction of the overall data points), and immediacy (i.e., the latency in change of level, trend, and variability after conditional change). To add to visual analysis, descriptive statistics were calculated for each dependent variable across participants. Nonoverlap of all pairs (NAP) with a 95% confidence interval (CI; Parker & Vannest, 2009) was calculated for initiating requests, answering questions, and commenting behaviors for each participant. NAP was selected as an effect size measure due to its strengths related to accuracy as well as its external validation relative to both R² and visual analysis judgments (Parker & Vannest, 2009).

Results

Results for each research question are reported below, beginning with the acquisition and maintenance of socio-communicative behaviors followed by the generalization of behaviors. Within each research question, visual analyses and NAP effect size indicator are reported for each participant. Figures 2–5 show the percentages of correct responses for initiating requests, answering questions, and commenting for each participant.

Figure 2.

The percentage of correct responses for Ajay. NAP = nonoverlap of all pairs.

Figure 3.

The percentage of correct responses for Matthew. NAP = nonoverlap of all pairs.

Figure 4.

The percentage of correct responses for Derek. NAP = nonoverlap of all pairs.

Figure 5.

The percentage of correct responses for John. NAP = nonoverlap of all pairs.

Acquisition and Maintenance of Social Communicative Behaviors

Ajay

During the baseline phase, Ajay did not initiate requests to his typically developing peer or the researcher. Therefore, no level, trend, or variability was observed during the baseline phase. During the intervention phase, Ajay rapidly met the acquisition criterion after three sessions of training during which he averaged 100% accuracy for initiating requests. The intervention data indicated a clear change in level compared with the baseline data, a change in trend to a positive trend, and no variability. His NAP for initiating requests was 1.00 (strong effect with a 95% CI [1.00, 1.00]). The maintenance probe indicated that Ajay maintained his ability to initiate requests to a typically developing peer 3 weeks following his last intervention session.

Baseline data showed that Ajay did not answer questions from his typically developing peer or the researcher. Thus, baseline data showed no level, trend, or variability. During intervention, he met the acquisition criteria after three sessions of training, during which he averaged 100% accuracy for answering questions. Intervention data indicated a clear change in level, upward positive trend, and no variability. His NAP for answering questions was 1.00 (strong effect with a 95% CI [1.00, 1.00]). Three weeks following Ajay's final intervention session, he maintained the ability to answer questions from a typically developing peer with 100% accuracy.

During baseline, Ajay did not make comments with the researcher or peer. That is, baseline data showed no level, trend, or variability. During intervention, he met the acquisition criterion after three sessions of training during which he averaged 93.33% accuracy (range: 80%–100%). The data showed a change in level, upward positive trend, and low variability. His NAP for answering questions was 1.00 (strong effect with a 95% CI [1.00, 1.00]). The maintenance probe with 40% accuracy suggests that he somewhat maintained the ability to make comments 3 weeks following his last intervention session.

Matthew

As depicted in Figure 3, during baseline, Matthew did not initiate requests to the researcher and initiated requests to his typically developing peer with 6.66% accuracy (range: 0%–20%). Therefore, no level, no trend, and low variability were observed. Matthew showed rapid progress when intervention was implemented. He met the acquisition criteria after three sessions of training, during which he averaged 100% accuracy. A change in level, positive trend, and no variability were shown in Matthew's intervention data. His NAP for initiating requests was 1.00 (strong effect with a 95% CI [1.00, 1.00]). Three weeks following intervention, Matthew maintained the ability to initiate requests with a typically developing peer.

Matthew did not answer questions from the researcher or peer during the baseline phase. Therefore, no level, trend, or variability was observed during the baseline phase. During the intervention phase, Matthew met the acquisition criterion after three sessions of training during which he averaged 100% accuracy. The intervention data clearly showed a change in level compared with baseline data, a change in trend to a positive trend, and no variability. His NAP for initiating requests was 1.00 (strong effect with a 95% CI [1.00, 1.00]). Maintenance data (i.e., 20%) indicated that Matthew did not maintain his ability to answer questions from the researcher.

Baseline data indicated that Matthew did not make comments to the researcher or his typically developing peer. Hence, there was no level, trend, or variability. After 11 sessions of training, Matthew did not meet the acquisition criteria (an average of 80% or higher across three sessions). Thus, the intervention phase for commenting was discontinued. Matthew's intervention data showed no level, no trend, and high variability (range: 0%–100%). His NAP for commenting was 0.909 (medium effect with a 95% CI [0.602, 0.983]).

Derek

As shown in the first panel of Figure 4, Derek did not initiate requests to the researcher or his typically developing peer. Thus, no level, trend, or variability was observed. After the intervention was implemented, Derek rapidly met the acquisition criterion after three sessions, during which he averaged 100% accuracy. The intervention phase data indicated an immediate change in level and positive trend, and no variability. His NAP for initiating requests was 1.00 (strong effect) with a 95% CI [1.00, 1.00]. Three weeks post intervention, Derek maintained his ability to initiate requests to a peer with 100% accuracy.

During the baseline phase, Derek did not answer questions from the researcher or his typically developing peer. Hence, there was no level, trend, or variability. After the intervention was implemented, Derek met acquisition criterion after three sessions, during which he averaged 80% accuracy (range: 60%–100%). The intervention phase data indicated a change in level, a change in trend toward a positive trend, and little variability. The data showed no overlap between baseline and intervention phases. His NAP for answering questions was 1.00 (strong effect with a 95% CI [1.00, 1.00]). Maintenance data (i.e., 20%) indicated that Derek did not maintain his ability to answer questions from the researcher.

Derek did not make comments to the researcher or his typically developing peer during baseline. That is, baseline data showed no level, trend, or variability. Derek showed rapid progress when intervention was implemented, meeting the acquisition criterion after three sessions of training with an average of 100% accuracy. A change in level, positive trend, and no variability were shown in Derek's intervention data. His NAP for answering questions was 1.00 (strong effect with a 95% CI [1.00, 1.00]). As indicated by maintenance probes, Derek also sustained the ability to make comments to a peer 3 weeks after the last intervention session.

John

During baseline, John did not initiate requests to the researcher or typically developing peer. Hence, there was no level, trend, or variability. During intervention, John completed seven sessions with an average of 37% (range: 0%–80%) before he transferred to a different school district and discontinued the study. His intervention data showed no level, no trend, and high variability. His NAP for initiating requests was 0.893 (medium effect with a 95% CI [0.570, 0.979]).

Generalization of Social Communicative Behaviors

Ajay

As shown in the first panel in Figure 2, during the baseline phase, Ajay did not initiate requests to his typically developing peer. After acquiring the initiating requests behavior, Ajay generalized the behavior to his typically developing peer with 100% accuracy. The generalization data for the initiating requests behavior indicated a clear change in level from baseline, a positive trend, and no variability.

During baseline, Ajay did not answer questions from his typically developing peer. After acquiring the answering questions behavior, he generalized the behavior to his typically developing peer who was not part of the intervention. Generalization data for the answering questions behavior showed a clear change in level from baseline, a positive trend, and no variability.

Ajay did not comment to his typically developing peer during baseline. After acquiring the commenting behavior, he somewhat generalized the behavior to his typically developing peer during which he averaged 26.66% accuracy (range: 0%–60%). Generalization data for the commenting behavior revealed moderate change in level from baseline, no trend, and high variability.

Matthew

Baseline data revealed that Matthew initiated requests to his typically developing peer with 6.66% accuracy (range: 0%–20%). During the generalization phase, Matthew did not interact with his originally assigned typically developing peer across two sessions. Thus, a new typically developing peer was introduced to Matthew in Session 12 (see first panel in Figure 3). Matthew generalized the initiating requests behaviors across both peers with an average of 36% accuracy (range: 0%–100%). When considering only data with his new typically developing peer, Matthew generalized the targeted behavior with an average 60% accuracy (range: 20%–100%). Overall, a change in level compared to baseline, no trend, and high variability was shown in the generalization data with Matthew's new typically developing peer.

Matthew did not answer questions from his typically developing peer during baseline as shown in the second panel of Figure 3. One week after the intervention phase, Matthew did not generalize the answering questions behavior to his typically developing peer. Thus, generalization data for the answering questions behavior revealed no change in level from baseline, no trend, and no variability.

Derek

As displayed in the first panel of Figure 4, Derek did not initiate requests to his typically developing peer during baseline. After acquiring the initiating requests behavior, he somewhat generalized the behavior to his typically developing peer, during which he averaged 46.66% accuracy (range: 20%–100%). Generalization data revealed a clear change in level from baseline, no trend, and high variability.

During baseline, Derek did not answer questions from his typically developing peer as shown in the second panel of Figure 4. One week following the intervention phase, Derek did not generalize the answering questions behavior to his typically developing peer. Generalization data for the answering questions behavior revealed no change in level from baseline, no trend, and no variability.

Baseline data showed that Derek did not make comments to his typically developing peer. After acquiring the commenting behavior, Derek generalized the behavior to his typically developing peer with 100% accuracy. A clear change in level from baseline to generalization phase, a positive trend, and no variability was shown in Derek's generalization data for commenting.

Discussion

This study investigated the effectiveness of systematic instruction and aided modeling using speech-output technologies on (a) the acquisition and maintenance of socio-communicative behaviors (i.e., initiating requests answering questions, and commenting) with a researcher and (b) the generalization of socio-communicative behaviors from researcher to typically developing peers who did not participate in intervention. Results showed experimental control for two participants (i.e., Derek, Ajay), demonstrating a functional relationship between the intervention and outcomes. For one participant (i.e., Matthew), experimental control could not be established because he did not reach the learning criterion for commenting. Three of the four participants generalized the initiating requests behavior to typically developing peers who were not part of the intervention, whereas only one participant generalized the answering questions behavior to a typically developing peer. Both participants who acquired the commenting behavior generalized the behavior to a typically developing peer. To varying degrees, participants maintained answering questions and commenting 2 and 3 weeks after the last generalization and the last intervention session, respectively. The results of this study address a critical gap in the AAC intervention literature by demonstrating that MS autistic children can engage in advanced socio-communicative behaviors with a researcher and generalize that skill to typically developing peers.

Addressing Gaps in AAC Intervention Research for MS Autistic Children

Reviews on AAC interventions for MS autistic children called for future research with these foci: (a) measuring outcomes with different communication partners, (b) reporting data on the generalization and maintenance of acquired communicative behaviors, (c) methodological rigorous studies, and (d) thorough reporting of participants' clinical and demographic characteristics (Chavers et al., 2022; Logan et al., 2017; Muharib & Alzrayer, 2018; Schlosser & Koul, 2015; Schlosser & Lee, 2000). The current study addresses each of these gaps in the existing literature. First, this study measured the generalization of acquired socio-communicative behaviors from researchers to typically developing peers who did not participate in intervention. Using a pre- and posttreatment multiple-generalization-probes design (Schlosser & Braun, 1994) allowed for repeated measurement of the targeted socio-communicative behaviors between the participant and typically developing peer during the baseline and generalization phase. Thus, this strategy provided evidence that components of the intervention may have resulted in generalization from the researcher to a typically developing peer who was not a part of training. Furthermore, this study was carried out with high fidelity, documenting the integrity across all phases for both the researcher and typically developing peers. Lastly, since the effectiveness of AAC interventions may be related to participant characteristics, each participant's demographic and clinical profiles were thoroughly reported using consistent measures across participants. Specifically, five standardized assessments were given to fully capture each participant's unique learning profile.

Acquisition of Social Communicative Behaviors

The efficacy of aided AAC modeling and systematic instruction to teach socio-communicative behaviors is further supported by the results of the NAP analyses for all participants. As NAP has a strong relationship with R² and visual analyst judgments (Parker & Vannest, 2009), there is strong confidence in the validity of results of this study.

The positive outcomes of the study can be attributed to the utilization of evidence-based instructional strategies, specifically aided modeling (Allen et al., 2017; Biggs et al., 2018) and systematic instruction (Alzrayer et al., 2019; Chavers et al., 2021; Finke et al., 2017). These results were consistent with other studies that utilized aided AAC modeling and/or systematic instruction to teach socio-communicative behaviors (Chavers et al., 2021; Finke et al., 2017; Sennott & Mason, 2016). Regarding aided modeling, participants may have benefited from the visual cues (i.e., SymbolStix) afforded by the speech-output technologies along with consistent auditory cues (e.g., same pitch, rate, intonation), and the motor action of selecting an icon along with the visual and auditory cues (Blischak et al., 2003; Schlosser, 2003; Schlosser & Blischak, 2001; Sterrett et al., 2022). Additionally, systematic instruction (i.e., least-to-most prompting, time delay, reinforcements) provided the participants with the necessary support to successfully engage in advanced socio-communicative interactions.

Generalization of Acquired Social Communicative Behaviors

Despite combined NAP across all participants demonstrating strong effect sizes for each dependent behavior, data collected during the generalization phase showed high variability in transferring learned behaviors to typically developing peers who were not a part of the intervention. These outcomes are noteworthy as they could be attributed to the participant's social motivation and recognition of social cues. Some individuals diagnosed with autism may need supports with (a) orientation to social cues and stimuli, (b) interpersonal maintenance strategies, and (c) response to social rewards (Chevallier et al., 2012; Elias & White, 2020). In the current study, the participants were required to demonstrate high levels of social motivation as they needed to choose between playing a preferred activity or communicating with a peer. Thus, the participants may have prioritized playing a preferred activity over engaging in social communication. Moreover, the peers may not have provided adequate social signals that promoted social bonds with the participants (e.g., eye contact, gaining participant's attention, tone of voice, body language). As social communication is inherently reciprocal, continued exploration of specific approaches that best facilitate social communication between MS autistic children and their typically developing peers is needed.

Though this study primarily utilized typically developing peers as communication partners rather than agents of intervention, two components of peer-mediated intervention were utilized. First, typically developing peers were trained to be responsive communication partners by providing aided AAC modeling and waiting 3 s for the participant respond to each treatment probe. Both aided modeling and wait time have been shown to be effective strategies to support communication between typically developing peers and MS autistic children (Trembath et al., 2009; Trottier et al., 2011). Additionally, typically developing peers were taught to maintain proximity to the participant for the entirety of each session. Though each peer was instructed to maintain proximity to the participant, explicit instruction was not given for the peer to gain the participant's attention. Teaching peers strategies to gain the participant's attention such as tapping the participant's shoulder or saying the participant's name may have resulted in greater socio-communicative interactions.

Historically, the etiology of development disability has been thought to influence communicative propensities (Ganz et al., 2022). For instance, it has been proposed that autistic children demonstrate a propensity to use communication outside of the context of social interaction (Wetherby, 1986; Wetherby & Prutting, 1984). The current study addressed this disparity by directly targeting advanced socio-communicative behaviors, such as commenting. Results from this study suggest that MS autistic children can make comments to both a researcher and a typically developing peer. Several factors may have influenced the participants' acquisition and generalization of the commenting behavior. First, each participant was given a preference assessment to identify preferred stimuli for the intervention. By using highly preferred activities, participants may have been more motivated to play and socially interact with the researcher and their typically developing peers. In addition to the use of highly preferred activities, the acquisition of the commenting behavior may have been influenced by the participants' preexisting knowledge of symbol–referent relationships as demonstrated by a symbol identification task. Many studies have incorporated symbol identification task and reported strong treatment effects (Alzrayer et al., 2019; Chavers et al., 2021; Lorah, 2016). Additionally, the targeted comments were selected to represent salient features of each participant's preferred activities. For instance, if a participant's preferred activity was playing with a race car, comments, such as “Oh no,” “Woohoo,” “I did it!”, could be targeted. Use of preferred activities and salient comments allowed for the researcher to provide aided modeling in a contextual rich and motivating environment. Lastly, only four symbols targeting commenting were utilized. Thus, the participant was given repeated exposure via aided modeling to the four targeted comments in a socially correct context.

Prior studies suggest that MS autistic children can rapidly acquire requesting behaviors as it results in access to a tangible item (e.g., Alzrayer et al., 2019; Chavers et al., 2021; Schlosser et al., 2020). The current study supports this finding as three participants acquired the initiating requests behavior and generalized the behavior to a typically developing peer who was not part of the intervention. Even though the remaining two dependent variables (i.e., answering questions, commenting) did not result in tangible reinforcement and are therefore often described as more difficult to teach (Schlosser et al., 2020), participants still demonstrated strong effects for the targeted socio-communicative behaviors. This finding is notable as it suggests that participating in an effective social communication exchange may be motivating enough for MS autistic children to learn advanced socio-communicative behaviors.

Clinical Implications

Although the majority of research on AAC interventions for MS autistic individuals has focused on teaching requesting skills (Schlosser & Koul, 2015), practitioners should consider targeting a wider range of communicative behaviors. Outcomes from this study suggest that MS autistic children can engage in socio-communicative behaviors with the researcher and peers. Practitioners should note that results of this study were due to use of evidence-based instructional strategies, rather than use of a specific AAC application. To achieve such outcomes, it is important that practitioners conduct a preference assessment to ensure that highly motivating and preferred activities are utilized during intervention sessions. For some children, the use of highly preferred activities may create a greater desire to communicate with peers (Thiemann-Bourque et al., 2017). Additionally, practitioners should conduct a symbol identification task to ensure that the child understands each referent and is able to discriminate between symbols. Furthermore, practitioners should consider targeting comments that are meaningful to the client and directly related to the activity. For example, if a child's block tower is collapsing, highly contextualized comments such as “It is falling!” or “Oh no” could be targeted. Additionally, since MS autistic children often have difficulty transferring learned behaviors across communication partners and settings (Gunning et al., 2019), practitioners should consider providing inclusion services. Inclusion services can promote generalization by allowing children to learn and practice social communicative behaviors in their natural social environment.

Above all, practitioners should remember that the overarching goal of AAC is to support participation across all communication activities in a wide range of social and physical environments (Williams et al., 2008).

Limitations and Future Directions

The promising results of the study should be considered in light of several limitations. First, this study was only replicated across four participants and only one maintenance session was conducted. Thus, definite conclusions regarding the maintenance of behaviors cannot be drawn. Systematic replication of this work with a larger cohort of MS autistic children and multiple maintenance sessions are needed to further investigate the efficacy of aided modeling and systematic instruction using an SGD on socio-communicative behaviors.

Second, there was not a pre-experimental task solely dedicated to measuring joint attention skills and imitation skills. Joint attention and imitation have been shown to play an important role in the language development of autistic children as it is one of the earliest indicators of social interaction and an important precursor to language (Ganz et al., 2022; Mundy et al., 1987, 1990). Furthermore, acquisition of imitation and joint attention skills leads to less intrusive forms of prompting and further boosts a learner's communicative acquisitions (Ganz et al., 2022). Future studies should consider reporting joint attention and imitation skills at the onset of the intervention as these skills may influence participant's communication outcomes.

An important limitation to this study is the variability observed for generalization of trained behaviors across participants. Variability in the generalization dimension may be partially due to peer training. Though each typically developing peer was taught to maintain proximity to the participant and how to use an SGD and implement aided modeling, typically developing peers were not given strategies on how to gain the participant's attention. Inclusion of strategies to gain attention may have promoted better social communication between the typically developing peer and participant. Additionally, each participant and typically developing peer dyad was assigned based on availability and chronological age. Future studies should consider developing peer and participant dyads based on mutual interest in activities and personality traits (e.g., assigning a participant who is energetic with a peer who is calm). Additionally, future research should consider personalization of displays to enhance ecological validity.

Conclusions

Though social communication and social interaction are core deficits in autism, few studies have explored the effectiveness of AAC interventions in teaching social communicative behaviors beyond requesting (Muharib & Alzrayer, 2018; Schlosser & Koul, 2015). The current study addressed this need by investigating the effectiveness of aided modeling and systematic instruction using speech-output technologies on the acquisition, generalization, and maintenance of socio-communicative behaviors (i.e., initiating requests, answering questions, commenting). Though results across participants are somewhat variable, outcomes of this study suggest that aided modeling and systematic instruction using speech-output technologies may lead to gains in socio-communicative behaviors in MS autistic children. Furthermore, this study revealed generalization, albeit variable, of socio-communicative behaviors from researcher to typically developing peers who did not participate in the intervention. Future research should continue to explore ways to facilitate social communication between MS autistic children and their typically developing peers.

Data Availability Statement

Data presented in this article are available upon request to the corresponding author.

Funding Statement

Research reported in this publication was supported by the National Institute of Child Health & Human Development Reward T32HD007489 (to Tiffany Chavers Edgar).