Visual Supports to Increase Conversation Engagement for Preschoolers With Autism Spectrum Disorder During Mealtimes: An Initial Investigation

Abstract

The diversity of children within the preschool classroom is dramatically changing as children with autism spectrum disorder are increasingly included within it. To engage in the benefits of inclusion, social skills are needed. Yet, children with autism commonly experience difficulties in this area. Extant literature indicates that social skills are more successfully acquired when taught through naturalistic and embedded instruction in established routines. A commonly occurring routine in most classroom, home, and community settings is mealtime. The purpose of this study was to investigate the effectiveness of Snack Talk, a visual communication support, for increasing the communication engagement of five preschool children with autism. A reversal design across participants was used to analyze the relation between Snack Talk and conversation engagement. Results from the maintenance probes show that conversation engagement increased across all participants when compared to baseline. Furthermore, a functional relation was established between the teaching phase (baseline and intervention data collection phases) and the maintenance phase. Limitations and directions for further research are also discussed.

Ensuring that all young children, regardless of their disability, language spoken, or communication needs, have the skills to engage in meaningful conversations, develop relationships, and be in community with their same-aged peers is a priority for all-inclusive early learning programs. Communication skills needed to interact successfully with others have long-term implications for children throughout their educational career, as students with positive classroom relationships are more likely to be academically engaged (Estell & Perdue, 2013; Perdue et al., 2009) and demonstrate higher levels of academic achievement (Wentzel, 2009). Furthermore, social interactions with peers are strongly associated with children’s social–emotional, language, and cognitive development. However, engaging in peer interactions and building relationships can be challenging for some children with disabilities, particularly, for some children with Autism Spectrum Disorder (ASD). As described by the American Psychiatric Association (APA), children with ASD demonstrate persistent difficulties with social communication have restricted interests and engage in repetitive behaviors (American Psychiatric Association [APA], 2013). These behaviors and ways of communicating often result in children with ASD having fewer friendships, higher rates of loneliness, and less social integration in their classroom or learning community (Chamberlain et al., 2007; Kasari et al., 2011). Given the significant amount of time children spend in school compared with other settings, communication and social skills interventions and support in schools have the potential to support children with ASD in developing relationships and having meaningful interactions with their same-aged peers.

Social Skills in the Classroom

Strong social and communication skills are critical across the school-age years as they enable children to develop social relationships, become integral members of a classroom’s social learning community (Vygotsky, 1978), and access desired outcomes such as academic engagement and achievement (Gresham, 2014; Wentzel, 2009). The classroom, as a social learning community, is a micro-culture in which students share routines that guide engagement in classroom activities. For example, in early childhood classrooms, knowing where to sit during a floor-based activity or how to work cooperatively during group play requires the usage of a wide range of communication and social skills. Programs that support positive relationships with peers and teachers are tied to improved academic engagement (Estell & Perdue, 2013; Perdue et al., 2009) and higher levels of academic achievement (Wentzel, 2009).

Social skills also foster community membership or a sense of belonging to a group of peers and their teachers (Schwartz, 2000). Community membership has been identified as a significant outcome (Schwartz, 2000; Staub et al., 1994) as current research indicates that students with disabilities who do not experience community membership or strong social connections may be more likely to experience detrimental outcomes. These outcomes include higher rates of isolation, exclusion, and bullying as well as an increased susceptibility for anxiety, depression, and suicidal ideation (e.g., Kasari et al., 2011; Locke et al., 2017). Conversely, effective social engagement has been linked to an increase in quality of life (Howlin et al., 2004; Locke et al., 2017; Marriage et al., 2009; Strain & Schwartz, 2001). As such, early intervention to support the social and communication skills development of children with disabilities is important for their overall school experience and trajectory. Effective techniques to target social skills for children with ASD include using peer-mediated programs such as Stay Play and Talk (Maich et al., 2018), an intervention in which students are directly taught to play with peers with ASD as well as video modeling programs (Kouo, 2019) to demonstrate and reinforce interactive play skills. Yet, despite these examples, extant literature suggests teachers experience a myriad of challenges in providing effective communication and social skills instruction. Further research in this area is warranted.

One challenge is that teachers do not consistently receive sufficient training to support the social and communication development of children or implement social skills instruction (Dobbins et al., 2010). When educators do implement social interventions, they are often more likely to implement class-wide, programmatic interventions versus individualized interventions (Brown et al., 2001). Teacher-facilitated programmatic classroom strategies often include a purchased curriculum in which students read, discuss, and role-play a particular social skills target, (e.g., PEERS; Laugeson et al., 2014), class-wide behavior monitoring systems for individual behavior (Schneider & Goldstein, 2009), and the use of group contingencies (e.g., good behavior game; Embry, 2002).

Social Skills Interventions

Researchers also have documented the lack of generalization of many of these social skills interventions (Bellini et al., 2007). Researchers noted that weak outcomes, such as poor generalization and generalization of skills, from some social skills interventions, may be attributed to the fact that they take place in “contrived, restricted, and decontextualized” (Gresham et al., 2001, p. 340) environments, such as pull-out settings or separate social skills groups. Conversely, individualized social skills interventions that take place within the ongoing classroom routines and within the child’s natural classroom setting lead to higher intervention, generalization, and generalization effects (Bellini et al., 2007). Furthermore, routines and activities in which the child is already interested promote naturalistic teaching opportunities, which have been shown to be beneficial (Brown & Odom, 1995). Critical features of a natural teaching environment found to enhance the likelihood of social interactions among children with ASD include activities and materials that promote opportunities for social interactions, such as books in a book area and snacks or mealtimes (Reszka et al., 2012). The significance of setting up an environment for success was highlighted in recommendations from the Division for Early Childhood (DEC) which supported social supports and strategies within established, natural routines (Division for Early Childhood [DEC], 2014). Similarly, early childhood teachers are encouraged to use evidence-based strategies such as scripts, video modeling, and opportunities for choice in natural settings to promote social skills (Barnett, 2018). One such natural activity in the traditional classroom is mealtime.

Within school, home, and community settings, mealtime is a natural time for children to develop social communication skills through conversations (Massey, 2004). Conversations during meals can incorporate discourse on individual and shared community interests and provide opportunities for embedded instruction on conversational and social skills (Snow & Beals, 2006). In early childhood classrooms, breakfast, snack, and lunch times often take up a significant part of the day and are one of the few times when children are consistently seated in close proximity to their peers. While this natural structure encourages neurotypical children to engage in meaningful conversation with their peers (Massey, 2004), children with ASD may not access and fully participate in these interactions without instruction, modifications, and adaptations. Additional supports and instruction, beyond simple proximity, are often necessary for children with ASD to meaningfully engage with their peers (Strain et al., 2011).

Green and colleagues (1984) developed Table-Talk placemats used during mealtimes that included conversational topics and pictorial games in a preschool setting. The authors assessed the efficacy of Table-Talk in increasing the quantity and quality of social and educational dinnertime conversation at restaurants with eight children and their families. Findings indicated that Table-Talk placemats increased the social and educational dialogue among children and families with high social validity. Spohn et al. (1999) extended this research through implementing a structured placemat game during snack with six children with developmental disabilities. The placemats utilized a picture collage of cartoon characters, foods, and animals to occasion social conversation. In the game, children were encouraged to take turns commenting on one aspect of their placemat; results indicated that this intervention increased young children’s verbal initiations during snack times and generalized to lunch. Given the effectiveness of these interventions, a similar approach may be beneficial for addressing the conversation needs of young children with ASD.

Snack Talk

Snack Talk (Gauvreau, 2017) is a visual-based instructional strategy designed to promote social communication through conversation engagement between children during mealtimes. Much like the talking game, the visual supports for Snack Talk are created around children’s interests (e.g., favorite foods, toys, activities, and routines), classroom curricular themes (e.g., insects, sea life, and community helpers), classroom activities (e.g., games, toys, and books), and other age-appropriate conversational topics (e.g., common city landmarks, summer activities, pets, family members, etc.). These visuals are provided alongside a question prompt such as, “What do you like to do at recess?” or “Have you been here?” followed by 9 to 12 associated photos or symbols (see Gauvreau, 2017). Because Snack Talk visuals include pictures of the child’s interests, children are motivated to engage together socially. As with other visual supports, Snack Talk is only effective when teachers model, prompt, and provide feedback on how children can use these cards to engage with one another (Gauvreau & Schwartz, 2013). Snack Talk is a visual support paired with prompting to support ongoing interactions among children of different communication abilities. While PECS is an AAC system for an individual, Snack Talk is a classwide visual-based intervention to support the social communication of all students. Through individualized prompting, children of all language abilities (e.g., those with verbal and non-verbal communication, and multi-language learners) can use them to initiate and respond to comments and questions from their peers by pointing, using a communication device, or speaking at their ability level.

Research Question

Despite its promise as an effective strategy for increasing the conversation skills and engagement for children with ASD during mealtimes, Snack Talk has not been empirically evaluated. To this end, we implemented Snack Talk in a preschool setting to answer the following research question: Does the use of the Snack Talk intervention result in increased conversation engagement for children with ASD during mealtimes?

Method

Participants

Five preschool-aged children were recruited from a self-contained classroom within an early education center, in a Mid-Atlantic state. The inclusion criteria for participants were (a) medical diagnosis of ASD as required for enrollment into the early education center, (b) chronological age between 3 and 4, (c) eligibility for early education services, and (d) identified challenges in social communication, as reported by the teacher during informal teacher interviews. Challenges in social communication included the limited use of vocal and nonvocal means to get needs met, limited vocal labeling in the environment (commenting on the environment so as to attract the attention of others, ex. “I see a butterfly!”), inconsistent direction following, and inconsistent or absent imitation skills. Demographic information pertaining to age, eligibility, and repertoires of skills were collected via teacher report during an individual informal interview. During these informal interviews, teachers were asked how the children generally communicated their needs and about their environment, their listening and direction-following, and their imitation skills.

Participants with ASD.

Corey was a 3-year-old male with a diagnosis of ASD comorbid with a Receptive/Expressive Language Delay. Weekly, he received 32.5 hr of Applied Behavior Analysis (ABA; Baer et al., 1968) services with no additional related services. Within Corey’s home, his family spoke predominantly English. Corey was considered a preemerging speaker and preemerging listener and used his pointer finger to consistently request preferred items. Corey’s teacher reported that he consistently, independently, and vocally requested items he wanted for at least 3 different items, could receptively identify items, and was beginning to vocally label a variety of items and colors in his environment. At the time of the study, he was beginning to imitate the actions of others but did not yet demonstrate this independently (referred to as generalized imitation).

Marie was a 4-year-old female with a diagnosis of ASD. Weekly, she received 32.5 hr of ABA services and 45 min of speech therapy. Within Marie’s home, her family spoke predominantly English. She was considered an emerging speaker and emerging listener and communicated using one- and two-word responses. Marie’s teacher reported that Marie vocally labeled a wide variety of two-dimensional and three-dimensional stimuli and receptively identified many common items. Furthermore, she imitated adult actions, followed many 1-step directions, and was beginning to respond appropriately and correctly when asked personal information questions.

Jonathan was a 4-year-old male with a diagnosis of ASD. Weekly, he received 32.5 hr of ABA services. Within Jonathan’s home, his family spoke predominantly English. He also received occupational therapy once per week outside of the preschool setting. He was considered an emerging speaker and an emerging listener and primarily communicated using single-word responses. He vocally requested a variety of preferred items and activities, and vocally labeled a wide variety of common objects during instructional sessions. However, he did not frequently vocally label items in the environment. At the time of the study, he was beginning to respond to questions, could receptively identify some items in the environment, and had generalized imitation.

Claire was a 4-year-old female with an ASD diagnosis. Weekly, she received 32.5 hr of ABA therapy and 60 min of speech therapy. Within Claire’s home, her family spoke predominantly English. Claire was considered an emerging speaker and emerging listener. According to her teachers, she vocally requested different items and activities using two-word phrases consistently (e.g., “pick up,” “sit here,” and “sing song”) and was beginning to expand these requests to three or more words. She imitated one-step and two-step actions and followed routine instructions. However, Claire had difficulty receptively identifying objects and items in pictures.

Justin was a 4-year-old male with an ASD diagnosis. Weekly, he received 32.5 hr of ABA therapy and 60 min of speech therapy. Within Justin’s home, his family spoke predominantly Spanish. At the time of the study, he was considered a preemerging speaker and prelistener. Justin consistently used a pointer finger to request preferred items, vocally requested items using word approximations, and used approximations to repeat (echoed) most of the words he heard others saying. When vocally requesting items, Justin sometimes used Spanish word approximations instead of English (e.g., approximating “agua” for “water”). He vocally labeled certain stimuli in the environment spontaneously but not consistently. According to his teacher, he vocally labeled uppercase letters and numbers 1 to 20 and matched identical two-dimensional and three-dimensional stimuli and was beginning to imitate the actions of others during instruction. Justin often required prompts to follow one-step instructions.

Peers.

In addition to the five study participants, the classroom also included six peers in the early education program who participated in Snack Talk during the generalization phase. Peers were between the ages of 2 and 5, and all had a diagnosis of ASD. Two of the peers were prespeakers/prelisteners meaning they did not yet communicate vocally or verbally and did not consistently use other methods to communicate. Three peers were emerging speakers and emerging listeners who were beginning to mand for and tact items using single words, follow simple 1-step instructions, and receptively identify items in the environment. The final peer was a speaker and listener who engaged in manding and tacting using full sentences, responded to questions intraverbally, and was beginning to emit conversational units and follow multistep instructions.

Intervention implementers.

Implementers of the intervention included classroom teachers, assistant teachers, and Registered Behavior Technicians (RBTs). The classroom had 15 staff members in total. Head teachers/team leaders included two Board-Certified Behavior Analysts. Teaching assistants included 2 graduate students in ABA programs and 10 RBTs. Teaching assistants worked individually with children in the classroom at 1:1 or 1:2 teacher–student ratios. The classroom also included a program support staff member who provided substitute coverage for teacher assistants during the study. Additional information regarding implementers’ demographics and experience can be found in Table 1.

Table 1.

Staff Information.

Role in classroom	Gender	Race	Level of education	Years of experience
Team Leader	F	White	Bachelor’s Degree	4.5 years
Teaching Assistant	F	White	Bachelor’s Degree	1.5 years
Teaching Assistant	F	White	Bachelor’s Degree	2 year
Teaching Assistant	F	Black or African American	High School Diploma	1 year
Teaching Assistant	F	White	High school Diploma	3.5 years
Team Leader	F	Black or African American	Bachelor’s Degree	4 years
Program Support (previously a Teaching Assistant)	F	Black or African American	High School Diploma	4.5 years
Teaching Assistant	F	White	High School Diploma-working on Bachelor’s	2 years
Teaching Assistant	F	Black or African American	High School Diploma	8 months
Teaching Assistant	F	Native Hawaiian or Other Pacific Islander	High School Diploma-working toward Bachelors	1 year
Teaching Assistant	F	White	Bachelor’s Degree	1 year
Teaching Assistant	F	White	High School Diploma	1 year
Teaching Assistant	M	White	Finishing Bachelor’s Degree	10 months
Teaching Assistant	F	White	Bachelor’s Degree	6 months
Teaching Assistant	F	White	Bachelor’s Degree	3.5 years

Setting and Materials

This study took place in an early education classroom providing ABA services to 11 children diagnosed with ASD. Five children participated in the intervention component and sat at one child-sized, rectangular table during lunchtime. Typical preschool lunchtime occurred as the intervention was implemented, and children not participating in the intervention sat and ate at another nearby table. To protect the confidentiality of children who had not consented to this study, those not participating in the intervention were redirected away from the intervention table during lunchtime.

Snack Talk cards (Gauvreau, 2017; Gauvreau & Schwartz, 2013) were used as the primary intervention tool and were created by the researcher. Snack Talk cards were 8½ ×11-inch laminated sheets of paper that included a social question in English (i.e., “What is your favorite snack?”) and 9 to 12 visuals representing possible answers to the target question. These cards were made using a Microsoft Word^® and Google^® Images search. In addition to Snack Talk, a video camera and tripod were used to collect videos of intervention implementation, and a laptop was utilized to review the videos for paper-pencil coding and analyze the collected data. Data collection forms included dependent variable coding sheets, as well as fidelity of implementation checklists.

Behavioral Definitions and Measurement

The first 10 min of each mealtime was video recorded and coded using a 10-s partial interval recording procedure to identify engagement in social conversation. Coders utilized the dependent variable data collection forms to manually record if an instance of conversation engagement occurred at any point during the 10-s interval for the duration of the mealtime video. The percentage of intervals engaged in target behaviors was reported. The dependent variable measured conversation engagement among participants and their teachers and peers during mealtimes. Conversation engagement was defined as any verbal or nonverbal social-communicative act with another individual (i.e., peers or teachers) that also included a bid for attention. Communicative acts that were initiated by the focal child, as well as communicative acts by the focal child in response to a peer or teacher, were both counted as conversation engagement. Therefore, conversation engagement captured the reciprocal, back and forth nature of the conversation. Bids for attention could include verbal (e.g., saying the name of a peer or teacher) and nonverbal bids (e.g., eye gaze or eye contact with peer). The verbal engagement was defined as any verbal initiation or response directed toward peers or teachers in social conversation. An example of this behavior would be a peer saying, “The tire swing is my favorite” and the focal child responding, “I like the slide.” This could also include any engagement in back-and-forth conversation. A nonexample of this behavior would be a child pointing to an item on the table without communicative intent or an attempt at gaining the attention of a peer or teacher. Another nonexample of this behavior included a participant in the study using a symbol from his PECS (Bondy & Frost, 1994) book to request fish crackers, and the teacher holding out the bowl and saying, “Thanks for asking, take what you want.” Manding or requesting meal-related items were not coded (e.g., goldfish, open, more, etc.). Nonverbal engagement was defined as any nonverbal initiation or response directed toward peers or teachers in social conversation. An example of this behavior included a peer asking a study participant about their favorite color by pointing to a color on the Snack Talk card and making no verbal utterances or responding to a peer’s question by pointing to a choice on a Snack Talk card. If a participant did not respond to a communication bid from a teacher or peer within 5 s of the original bid for communication, this was indicated as a “nonresponse” on the data coding sheet and a least-to-most prompting protocol was begun as described in the intervention sessions. The percent of intervals engaged in nonresponse behaviors was recorded. Once the child engaged in a communicative act for conversation, an initiation or response was recorded for the appropriate interval.

Communicative acts were collapsed into one broader code due to the range of beginning communication skills demonstrated across participants. This ensured that the data collected represented the nonverbal, and developing verbal communication abilities of this group of learners. Because the nature of social conversation may not include a clear beginning and end, partial interval recording was used to ensure data collected accurately captured engagement in this behavior. While this coding procedure tends to overestimate engagement in the target behavior(s) (Ledford et al., 2018), this procedure was used to capture engagement in communicative acts while accounting for the natural lulls and breaks in the conversation. This is especially important for learners with beginning communication skills, as ongoing, fluid communication with minimal lulls in the conversation has not yet been achieved.

Experimental Design

A withdrawal design (Gast & Ledford, 2018; Severini et al., 2018) was utilized for all participants. In accordance with What Works Clearinghouse, this methodology was designed to meet the criteria for single-case research design standards (Kratochwill et al., 2013). In this study, participants moved through baseline and intervention phases together. As this intervention is intended to be implemented during natural routines within a classroom setting serving multiple students at the same time, all participants received the intervention simultaneously. No individual programming decisions regarding phase changes were made due to the group implementation of this intervention. Instead, phase changes occurred after at least 5 days of data collection in each phase in alignment with guidelines for single-case research design (Kratochwill et al., 2013) and took into consideration school breaks and holidays as well as school-wide activities conflicting with intervention. However, due to unforeseen illness-related absences, some participants did not meet the standard of 5 data collection points in each condition.

Procedures

All sessions occurred during the lunch mealtime routine. A video camera was set up at the head of the table for all conditions. Sessions began as soon as all participants were present at the table for mealtime and lasted 10 min in length. All participants and teachers were present during mealtimes. In both conditions, coding procedures continued if a child left the table. Data collection intervals continued to be coded in these instances as nonengagement.

Baseline sessions.

Baseline sessions were business as usual. Participants sat at the table during lunch mealtime with teachers and ate lunch. All five participants sat around the table with an average of four teachers sitting behind the group of five children. Mealtime was a preestablished routine in this classroom and children were familiar with this activity. Snack Talk supports were not provided in this condition. As Snack Talk supports were not in this condition, prompting was also not provided. Instances of nonresponse to the teacher or peer bids for communication were still recorded. Teachers interacted with children as they typically did during this classroom routine, sporadically prompting children to eat during this time.

Teacher training.

Teachers were trained in implementation procedures by the first author prior to beginning intervention sessions during an afterschool staff meeting. The teacher training protocol followed Parsons et al.’s (2012) evidence-based procedures including (a) Describe the target skill, (b) Provide a succinct written description of the target skill, (c) Demonstrate the target skill, (d) Practice the target skill, (e) Provide feedback during practice, and (f) Repeat steps 4 and 5 until skills mastery (Parsons et al., 2012). All teachers demonstrated 100% fidelity on steps of implementation before the onset of intervention.

Intervention sessions.

In intervention sessions, teachers supported students in either a 1:1 or 1:2 ratio. The teacher supporting the target child rotated daily throughout all phases to support generalization across implementers from the onset of the study. Thus, while each child was supported by one adult for each session, all children worked with all teachers during the course of the study. During the intervention sessions, teachers placed Snack Talk cards on the table where they were easily accessible to children. Several Snack Talk cards were made available to each child so that they could engage with a preferred topic and to diminish the likelihood of the child memorizing a single discriminative stimulus. Next, a teacher held a Snack Talk card up and modeled engaging in conversation by identifying what he or she liked on the card. For example, a teacher would state, “I really like eating grilled cheese for lunch!” Following the initial model, the teacher paused for 5 to 10 s to give the children an opportunity to naturally and independently engage with their peers using the visual. If children did not independently engage, the teacher then used the card to ask each child the topic question on the Snack Talk card, providing verbal social reinforcement for each response as he or she moved to the next child at the table. Following the group model and individual engagement with each child, teachers provided individual support for the children to use the Snack Talk cards with their peers. Modeling was provided each session for participating children throughout the intervention.

Teachers used least-to-most prompting during lulls in social conversation to facilitate conversation engagement. Teachers first used gestural prompts (i.e., pointing) followed by verbal prompts (e.g., “She asked what you like to do on the playground” or “Let’s ask a friend which animal they like”) as needed for nonresponse. During all sessions, if a child did not respond within 5 s to the teacher model, a teacher’s bid for communication, or a peer’s bid for communication, the teacher used the least-to-most prompting hierarchy as described and a nonresponse was recorded during data collection. Teachers moved up the hierarchy of prompts until children successfully engaged with the Snack Talk card with teachers or peers. Reinforcement was provided immediately upon conversation engagement, regardless of prompting level. For all participants, verbal social reinforcement was provided. Verbal social reinforcement was identified by teachers as an effective form of reinforcement for all participants prior to the implementation of this study. Procedures were repeated upon the occurrence of natural lulls in the conversation for the entirety of the targeted mealtime. A natural lull in conversation was defined as a period of 10 to 15 s where there was no verbal or nonverbal social interaction.

Post-intervention phase for generalization across peers.

After the conclusion of the final intervention phase, participants were integrated with their peers who did not participate in this study, sitting among new children at new tables during mealtimes. Teachers continued daily implementation of Snack Talk with study participants and their peers who had not participated in the study. To collect data on how conversation engagement skills were maintained with new peers, participants who were enrolled in the study returned to the initial intervention table with other study peers and teachers for generalization probes. Due to nonconsent from the families of other students, generalization probes were not able to be collected in vivo during mealtimes where participants and nonparticipants were engaged in Snack Talk together. Data were collected once every 2 weeks for a total of three generalization probes across a duration of 6 weeks. During these probes, Snack Talk supports were implemented with teacher support, identical to implementation in intervention phases. Data were collected in the same manner as previous phases, utilizing 10-s partial interval recording procedures.

Procedural Fidelity

Procedural fidelity was coded via video observations. A procedural fidelity checklist was used to collect and code consistent implementation of intervention. This checklist consisted of 6 steps of intervention: (a) Place Snack Talk on the table during mealtimes, (b) Model by making an on-topic comment and showing the table the Snack Talk card, (c) If students respond, reinforce these interactions with social praise and other reinforcement systems if appropriate, (d) If students do not respond, begin least to most prompting. Continue the prompting hierarchy until the target response occurs, (e) Reinforce as target responses/initiations occur, and (f) As natural lull in conversation occurs, begin prompting hierarchy again. These data were collected during all instances of implementation of Snack Talk procedures and coded daily during all intervention conditions, including generalization. Procedural fidelity was measured for all teacher implementers, and results of indicated procedures were followed to 100% fidelity across all intervention and generalization phases and implementers. Procedural fidelity was measured at 0% during baseline phases, and no part of the intervention protocol (i.e., Snack Talk visual and least to most prompting) was implemented.

Interobserver Agreement

Interobserver Agreement (IOA) was measured independently by two graduate student research assistants who served as IOA coders. IOA coders were trained to 90% mastery criteria prior to coding using interval-by-interval IOA (Cooper et al., 2007) procedures on 25% of the sessions in each phase of the study and with all participants. Sessions were randomly selected from each phase for IOA data collection. The average IOA was 94% across all participants and phases, ranging from 83% to 100%, meeting design standards identified in Kratochwill et al. (2013). Condition and participant-specific IOA are reported in Supplemental File 1.

Social Validity

To identify teachers’ perceptions of the implementation of Snack Talk supports during mealtimes, anonymous, self-report surveys were collected from study implementers. The research team distributed the surveys to the classroom’s supervisor. Next, the classroom supervisor distributed the social validity surveys to the study implementers. Identifying information was not written on the social validity forms. This survey included an opportunity for teachers to report any additional anecdotal feedback regarding intervention. The classroom supervisor collected the completed forms and stored them in a large folder, ensuring anonymity. Completed surveys were returned to the research team who analyzed the results.

Effect Size Estimation

In addition to visual analysis, we used the between-case standardized mean difference (BC-SMD; Hedges et al., 2013; Shadish et al., 2014; Valentine et al., 2016) to estimate the magnitude of intervention effects. The BC-SMD is interpreted using Cohen’s (1988) guidelines, with a BC-SMD < 0.20 indicating a small effect and a BC-SMD > 0.80 indicating a large effect. The BC-SMD has been identified by Shadish and colleagues (2015) as a robust effect size that is consistent with the group standardized mean difference. The BC-SMD assumes (a) the baseline is stable, (b) the intervention elicits an immediate response, (c) the intervention effect is the same across case, (d) the outcome is normally distributed, and (e) deviations from the phase mean levels follows a first-order autoregressive process (Barton et al., 2017, p. 375). The final assumption accounts for potential autocorrelation.

To calculate the BC-SMD, studies must have a minimum of three datum points in all phases and at least three cases. Generalization points were excluded from the analysis. We inputted the data into scdhlm, an open-source, web-based program designed to calculate the BC-SMD effect size (Pustejovsky et al., 2021) and assigned session numbers to each case as the detrending variable. The effect size estimate was calculated using the Restricted Maximum Likelihood estimation method with random effects for both the baseline and the intervention phases.

Results

Conversation Engagement

For all participants, levels of engagement in conversation during baseline data collection phases were low, and increases were seen across all participants during the intervention phases. Furthermore, data from all participants presented a clear level change between the first intervention phase and the second baseline phase, indicating a functional relationship between the intervention and the conversation engagement. Data from generalization probes indicated generalized effects on conversation engagement across all participants. The BC-SMD effect size was 1.50 (SE = 0.37; confidence interval [0.76, 2.25]) indicating a large effect of the Snack Talk intervention on children’s social engagement.

The results of engagement in the social conversation for each participant are shown in Figures 1 to 5. All communicative acts that were verbal were in English. After an instance of non-response during intervention implementation, teachers immediately began the least-to-most prompting hierarchy as described earlier. Rates of nonresponse for each participant are reported below, indicating how frequently participants required additional prompting, beyond the initial bid for conversation, to engage with peers and teachers using intervention materials during intervention phases.

Figure 1.

Percentage of intervals Corey engaged in conversation.

Figure 5.

Percentage of intervals Justin engaged in conversation.

Corey.

Corey engaged in low levels of conversation engagement with teachers and peers during baseline phases, ranging from 2% to 10% of intervals (See Figure 1). During baseline phases, Corey engaged in low rates of nonresponse, ranging between 0% and 3.33% of intervals $\overline{x}=0.55\%$). Upon intervention, Corey showed an immediate increase in the percentage of intervals in which he engaged in the target behavior, ranging from 6% to 20% of intervals. He demonstrated a stable, increasing trend in intervention Phase 2. Minimal overlap occurred between baseline and intervention sessions (2 sessions), demonstrating an overall higher level of engagement in conversation engagement during intervention than preceding baseline data collection phases. During intervention phases, Corey engaged in slightly higher rates of nonresponse, which were apace with the increase in bids for communication from teachers and peers. Rates of nonresponse did decrease, however, between the first intervention phase (0.0%–5.0%; $\overline{x}=1.67\%$) and the second intervention phase (0.0%–3.33%; $\overline{x}=0.83\%$). During generalization probes, Corey engaged in higher rates of conversation engagement compared with baseline and intervention phases, ranging from 20% to 27% of intervals. In comparison to only baseline phases, generalization data did not show any overlap in data points. During generalization probes, however, Corey engaged in higher rates of nonresponse compared with baseline and intervention phases and required more prompting to engage in conversation (17.0%-23.0%; $\overline{x}=20.0\%$). Finally, data collected in generalization probes suggested a generalized effect, demonstrating moderately consistent, increased trend and level of engagement in conversation engagement.

Marie.

Marie engaged in the highest rates of conversation engagement with teachers and peers during baseline phases, ranging from 10% to 20% of intervals (See Figure 2). During baseline phases, Marie did not engage in any instances of nonresponse. Upon intervention, Marie showed an increase in the percentage of intervals in which she engaged in the target behavior, ranging from 11% to 28% of intervals. She demonstrated increases in level and trend across all intervention phases. Overlap occurred between baseline and intervention sessions (8 sessions), demonstrating a wider range of conversation engagement during intervention than the preceding baseline data collection phases. During the first intervention phase, Marie engaged in slightly higher rates of nonresponse (0.0%–20.0%; $\overline{x}=4.04\%$), apace with the increase in bids for communication from teachers and peers. In the second intervention phase, Marie did not engage in any instances of non-response. During generalization probes, Marie engaged in higher rates of conversation engagement, ranging from 26% to 30% of intervals. In comparison to only baseline phases, generalization data did not show any overlap in data points. Within the generalization phase, Marie engaged in higher rates of non-response and required more prompting to engage in conversation (23.0%–27.0%; $\overline{x}=25.0\%$). Finally, data collected in generalization probes suggested a generalized effect, demonstrating moderately consistent, increased trend and level of engagement in conversation engagement.

Figure 2.

Percentage of intervals Marie engaged in conversation.

Jonathan.

Jonathan engaged in very low levels of conversation engagement with teachers and peers during baseline phases, ranging from 0% to 2% of intervals (See Figure 3). During baseline phases, Jonathan engaged in low rates of nonresponse, ranging between 0% and 1.67% of intervals ($\overline{x}=0.19\%$). Upon intervention, Jonathan showed an immediate increase in the percentage of intervals in which he engaged in the target behavior, ranging from 3% to 23% of intervals. A significant change in the level of conversation engagement was demonstrated across both intervention phases. Overlap did not occur between baseline and intervention sessions, demonstrating an overall higher level of engagement in conversation engagement during intervention than the preceding baseline data collection phases. During the first intervention phase, Jonathan engaged in higher rates of nonresponse (6.67%–11.62%; $\overline{x}=8.33\%$), apace with the increase in bids for communication from teachers and peers. In the second intervention phase, Jonathan engaged in far fewer instances of nonresponse (0.0%-5.0%; $\overline{x}=0.83\%$). While an immediate increase in the percentage of intervals in which he engaged in the target behavior was indicated, this data path demonstrated a decreasing trend. We contribute this to the novelty effect of the intervention, as Jonathan was less and less motivated to use Snack Talks as the intervention went on. During generalization probes, Jonathan engaged in high rates of conversation engagement, ranging from 13% to 32% of intervals. Jonathan engaged in higher rates of nonresponse and required more prompting to engage in conversation (5.0%–30.0%; $\overline{x}=16.0\%$) on the generalization probes compared with both baseline and intervention phases. In comparison to only baseline phases, generalization data did not show any overlap in data points. Data collected in generalization probes suggested a generalized effect, demonstrating variable, yet increased trend and level of engagement in conversation engagement. Finally, experimental control and functional relation were demonstrated.

Figure 3.

Percentage of intervals Jonathan engaged in conversation.

Claire.

Claire engaged in conversation engagement with teachers and peers during baseline phases, ranging from 0% to 10% of intervals (See Figure 4). During baseline phases, Claire did not engage in any instances of nonresponse. Upon intervention, Claire showed an immediate increase in the percentage of intervals in which she engaged in the target behavior, beginning intervention at 12% of intervals. Following the first day of implementation of the intervention, Claire demonstrated stability, as data ranged from 2% to 5% of intervals. Claire showed very low rates of engagement in the target behavior in the second baseline phase. These data stand out and may be attributed to the food available in her lunch these days as Claire had highly preferred items in her lunch during this baseline phase, decreasing the opportunity for social engagement as she independently ate her favorite foods. During the first intervention phase, Claire engaged in slightly higher rates of nonresponse (0.0%–3.33.0%; $\overline{x}=2.22\%$), apace with the increase in bids for communication from teachers and peers. In the second intervention phase, while variable, Claire did not engage in any instances of nonresponse. Variability in this phase is attributed again to the highly preferred food available in Claire’s lunch, affecting opportunities for social engagement as she again independently ate her favorite foods. Overlap occurred between baseline and intervention sessions (9 sessions), demonstrating a wide range of conversation engagement during intervention and baseline data collection phases. In comparison to only baseline phases, generalization data did not show any overlap in data points. However, within the generalization phase, Claire engaged in higher rates of non-response and required more prompting to engage in the target behavior (7.0%–20.0%; $\overline{x}=14.0\%$). Finally, data collected in generalization probes suggested a generalized effect, demonstrating consistent, increased level of engagement in conversation engagement.

Figure 4.

Percentage of intervals Claire engaged in conversation.

Justin.

Justin engaged in low levels of conversation engagement with teachers and peers during baseline phases, ranging from 0% to 3% of intervals (See Figure 5). During the first baseline phase, Justin did not engage in any instances of nonresponse. In the second baseline phase, he engaged in minimal instances of nonresponse (1.0%–1.67%; $\overline{x}=0.42\%$). He showed minimal attendance during baseline phases, missing approximately half of data collection days due to illness. Upon intervention, Justin showed an increase, although not immediate, in the percentage of intervals in which he engaged in the target behavior, ranging from 0% to 22% of intervals. He demonstrated a significant change in level and increasing trend in both intervention phases. During the first intervention phase, Justin engaged in higher rates of nonresponse (0.0%–1.67%; $\overline{x}=0.20\%$), apace with the increase in bids for communication from teachers and peers. During the second intervention phase, Justin engaged in slightly higher rates of nonresponse (0.0%–16.67%; $\overline{x}=5.56\%$). Minimal overlap occurred between baseline and intervention sessions (3 sessions), demonstrating an overall higher level of engagement in conversation engagement during intervention than the preceding baseline data collection phases. During generalization probes, Justin engaged in high rates of conversation engagement, ranging from 15% to 20% of intervals. In comparison to only baseline phases, generalization data did not show any overlap in data points, but Justin did engage in higher rates of nonresponse as compared with both baseline and intervention phases (22.0%-25.0%; $\overline{x}=24.0\%$). Finally, data collected in generalization probes suggested a generalized effect, demonstrating consistent, increased trend, and level of engagement in conversation engagement.

Social Validity

Social Validity results indicated that all teachers found the intervention to be very effective at increasing social conversation for preschoolers diagnosed with ASD (see Supplemental File 2). Teachers also rated the intervention as highly effective in increasing students’ overall engagement during mealtime as well as the focal students’ repertoire of conversational topics. In addition, anecdotal and survey responses indicated teachers identified that implementation of the Snack Talk supports during mealtimes helped them engage with the children and provided teachers with clear procedures to follow during a time in which they previously engaged minimally. While challenging behavior was not a behavior we directly measured in this study, teachers reported the intervention as moderately effective at reducing challenging behaviors during mealtimes which may be due to the low rates of challenging behavior the participants were already demonstrating. Teachers identified challenging behaviors included pouring food and drink items on the floor, elopement from the mealtime table as well as sitting underneath the mealtime table and chairs.

Discussion

The primary purpose of this study was to investigate the effects of using Snack Talk on conversation engagement for five preschool children with ASD during mealtime. This study demonstrated the effectiveness of the Snack Talk procedure (i.e., visual supports with prompting) in increasing conversation engagement for three (i.e., Corey, Jonathan, and Justin) out of five participants of varying communication levels with peers and teachers. For all participants, the Snack Talk intervention produced a large effect size as measured by the BC-SMD (Hedges et al., 2013; Shadish et al., 2014; Valentine et al., 2016). This intervention enabled teachers with limited planning time and instructional resources to deliver an intervention that benefited a range of children during an often-under-utilized instructional routine. Children with varying communication skill sets and abilities were able to participate in the Snack Talk intervention emphasizing the versatility and flexibility of this intervention.

Overall, Snack Talk implementation supported increases in conversation engagement for three children with ASD with their peers and teachers. Although research exists on a variety of social skills strategies for young children (Kouo, 2019; Maich et al., 2018), the preparation and implementation requirements coupled with reports of insufficient teacher training (Dobbins et al., 2010) suggested a need for more naturalistic, simple strategies. For participants, increases in conversation engagement with peers and teachers observed during the intervention and generalization phases were consistent with the individual participants’ developmental levels and their corresponding individualized learning goals and support needs. Participants in this study communicated by pointing, using PECS, and through one- and two-word vocal phrases. These communication patterns are representative of the range of communication modalities demonstrated by many young children with ASD. The moderate changes in conversation engagement resulting from the Snack Talk procedure were not unexpected given that the students were at an early phase of language development (i.e., we did not expect that using Snack Talk would result in participants speaking in full sentences if they were pointing to communicate their needs in baseline phases).

In addition, due to the nature of the implementation of this intervention and its naturalistic structure, children engaged in conversation with their peers, often with teacher support, or in conjunction with a conversation with their teacher. This was expected given the participants’ verbal language development, their need for language models and prompting, and their emerging conversational skills. This is consistent with prior research on using materials and activities to promote social commenting, turn-taking, and reciprocal interactions (Daubert et al., 2015; Lee & Lee, 2015). By incorporating peer and teacher dialogue together naturally, the intervention allowed for reciprocal conversation and communication exchanges across child–teacher dyads, similar to a natural conversation. As a whole, the Snack Talk procedure resulted in opportunities for developmentally similar social exchanges and contributions among the participants with their peers and teachers. It further provided a format for the interactions to be shared across all group members, as evidenced by the similar rates in conversation engagement across participants. In other words, one child was not dominating the entire session, which could have potentially indicated that other children were not able to engage in conversation with their peers or teachers. This observed reciprocal, turn-taking style of communication has been shown to be related to successful language development for children with communication challenges (Hadley & Rice, 1991).

As children began to engage with Snack Talk, it is important to note that nonengagement tended to decrease during intervention for some children. In addition, teachers identified that a decrease in challenging behavior during mealtimes was observed, promoting distal outcomes of intervention that although not directly measured, are significant in a classroom setting.

In addition to positive outcomes indicated for the children in this study, anecdotal remarks made by teachers in the social validity surveys indicated the potential of this intervention to support positive teacher behaviors. This intervention supported teachers’ ability to incorporate an effective strategy for students with diverse communication needs across a classroom and provide opportunities for all children to participate in mealtime conversations. The intervention also provided teachers with strategies to engage with children during classroom routines where staff interactions with students were often low as demonstrated during baseline conditions. Doing so ensured children had more opportunities to engage in reciprocal communication with their teachers and be supported in conversations with their peers. Low rates of interactions during the mealtime routine are documented in classroom research (Spohn et al., 1999) yet engaging with teachers and peers during this routine is important, and a meaningful outcome. Anecdotal feedback obtained on social validity surveys supported these findings as teachers identified that the intervention decreased personal, off-topic conversation among staff during mealtimes and increased supportive interactions with children. Teachers further reported Snack Talk as beneficial to their use of time and in their ability to plan with few resources while taking advantage of the natural feature of mealtime embedded in the school day (Reszka et al., 2012). On the staff Social Validity measure, teachers reported that the Snack Talk procedure was easy to implement in a natural setting, ultimately supporting that mealtimes are a natural routine in which interventions can easily be embedded.

Although the increases in engagement observed appeared appropriate to the level of development for our participants, it is possible that higher levels of engagement would have been seen if peer models with varying language abilities, rather than similar ones were used. Compared with other instances of implementation of the Snack Talk (Gauvreau, 2017) procedure where participants had a range of expressive and receptive communication skills, the participants in this study had similar communication profiles. During the generalization phases, however, participants in this study engaged in Snack Talk with peers of more complex language abilities, thus potentially explaining the increase in conversation engagement recorded during the generalization probes. Given the consistent engagement across participants in generalization data collection, the incorporation of children of different communication abilities into the intervention is meaningful. Therefore, future research may focus more on the diversity of the peer models by including a broader representation of communicative abilities in participants. The pool in this study was limited to those enrolled and consented in the particular preschool setting.

Limitations and Implications for Future Research

Despite this study’s promising results, several limitations exist that warrant consideration in future research and in the generalization of the results. First, as the participating children within this study were all considered preemerging or emerging speakers, conversation engagement was recorded as a single code. Data were not collected on whether the conversation engagement was an initiation or response. Similarly, as we expected children to need prompting throughout the intervention due to their emerging language skills, we did not collect data regarding the prompting level needed for children to engage. Future research on Snack Talks should seek to examine initiations and responses separately as well as incorporate prompting levels into the coding structure and intervention design.

There are some methodological limitations to note when considering generalization of study findings and when considering directions for future research. While this study included 6 weeks of data demonstrating a post-intervention phase for generalization across peers, this study did not include generalization data with peers who did not participate in intervention. This was due to setting and video consent constraints. Future studies should include additional generalization probes to understand the effects of Snack Talk more fully on generalization of conversation engagement skills following completion of intervention to new settings, such as other classroom settings, mealtimes at home, or in a community setting. Research focused on working with families and community partners to gauge Snack Talk success in those settings is suggested. The study’s goals involved understanding how the use of Snack Talk influenced social communication within an existing routine in a preschool classroom serving a diverse range of young children, thus intervention occurring in an applied classroom setting was not only necessary but also crucial in understanding the effects of this class-wide intervention.

Future studies should also include a fade protocol for prompting as well as a fade plan for the intervention. As this was an initial investigation into this intervention, and the participants were all emerging in their communication abilities, it was not expected that prompting would be able to be faded. Similarly, we did not include a fade plan for the intervention, as the intervention is intended to be used in an ongoing manner as classroom support as the participants were still developing their language and social skills. However, due to this, it may be difficult to understand if changes in conversation engagement were attributed to prompting or the visual cue provided through Snack Talks.

Furthermore, due to the group implementation format of this intervention within an applied setting, phase changes were not based on individual programming and performance. Instead, phase changes were set to occur after a minimum of five implementation sessions (i.e., days). Due to unforeseen illness, not all participants met the criteria for this resulting in non-experimental data for Corey and Justin. Future research could consider alternative designs that would account for the characteristics of this intervention and the realities of applied research in a classroom.

Another consideration is whether in some instances the partial interval recording over-represented the behaviors of the participants. Partial interval recording was used due to its ability to capture the target behaviors and its ease of use in the natural classroom setting. A strength of the study is its ability to demonstrate a simple procedure that has the potential to increase interactions during already-occurring classroom activities. The whole interval recording procedure was not a match to the target behaviors and the momentary time sampling procedure would have most likely under-represented the effects. The partial interval recording was the best fit for this reason and adds practicality for teachers who may consider using this procedure in their own classrooms. It must be noted that experimentally, however, the partial interval recording poses a limitation in the results.

As this was an initial investigation, this study occurred in a classroom serving all students with disabilities. Furthermore, this program was designed to provide ABA therapy services within an early intervention educational setting to preschoolers with a medical diagnosis of ASD. As such, the ratio of one teacher for every one or two students could not be changed to mimic a more typical preschool setting. However, the intervention was implemented by a variety of staff, providing increased opportunities for generalization across implementers and the use of multiple exemplars. Future work should examine the effects of a larger student to teacher ratio on intervention efficacy and feasibility in inclusive education settings. Future studies should also incorporate more nuanced procedural fidelity measures to more accurately capture prompting that may occur in baseline conditions. Last, this study used subjective methods of collecting social validity data. In accordance with Gast and Ledford (2018), future studies should collect both objective and subjective data to demonstrate social validity of intervention for consumers.

Implications for Practice

This initial investigation into the implementation of Snack Talk has several classroom-based implications for educators. First, the use of visual support with teacher prompting in the Snack Talk procedure was effective in increasing the target behavior (acts of social communication). This study supports and extends previous research suggesting positive outcomes of this evidence-based practice (Wong et al., 2015). Educators should consider using Snack Talk visuals and prompting to support communication in group classroom settings. Next, this intervention allows for multiple students with varying communication abilities to engage with one another on a topic of conversation. Educators should consider the implementation of this intervention during mealtimes to promote communication among learners of all ability levels. To extend this research, educators should also consider intentionally creating groups of children with varying social and communication abilities to foster development and promote inclusion. Finally, the results of this intervention align with evidence of naturalistic interventions implemented during naturally occurring classroom routines (Snyder et al., 2015). This initial investigation extends this body of literature, focusing on implementation during mealtimes—a commonly underutilized activity for targeted instruction in early childhood settings (Bateman & Wilson, 2021). Educators should consider implementing Snack Talk during mealtimes across the day.

Conclusion

Results from this study support previous research on visual support and conversation engagement development, extending the literature by focusing on implementation during naturally occurring classroom routines, such as mealtimes. Snack Talk is a simple and easily implementable intervention that has demonstrated effectiveness in increasing conversation engagement for children with ASD with high social validity and acceptability from implementers as measured by a social validity survey. Snack Talk is a low-tech, and low or no-cost intervention that can be made without specific software. Minimal planning and training are needed to support staff in implementing Snack Talk as it utilizes least to most prompting and simple implementation procedures, which supports fidelity of implementation and overall sustainability. Furthermore, this classwide intervention is individualized through prompting and demonstrates increases, and more importantly, generalization, in conversation engagement for preschool children with ASD and complex communication needs. This type of intervention can be extremely useful for classrooms where educators must provide individualized intervention to diverse groups of students at the same time. Implementation of Snack Talk supports fostered and promoted social skills used throughout an individual’s life, emphasizing the importance and meaningful outcomes resulting from this intervention.