Introduction

Social deficits are diagnostic characteristics for autism spectrum disorder (ASD) and individualized and intensive instruction using evidence-based practices are recommended for all students with ASD (National Research Council 2001; Wong et al. 2014). Although these practices can be implemented in a number of settings, current social, legal, educational, and ethical zeitgeists often result in the call for these practices to be implemented in inclusive settings, with typical peers present (e.g., Council for Exceptional Children 2010). One difficulty in implementing EBPs is that teachers of students with ASD often report inadequate knowledge about these practices and inadequate support for implementing them in typical settings (Brock et al. 2014; Morrier et al. 2011). Moreover, teaching children with ASD alongside their non-ASD peers requires an additional set of skills, including the ability to manage groups of students with varying needs while simultaneously providing effective instruction; this skill is particularly problematic for teachers (for a review, see Oliver and Reschly 2007). This may result in a situation in which children with autism, even those served in general education settings, are often taught in one-to-one instructional arrangements. In addition to difficulties with behavior management, this emphasis on individual instruction may also be due in part to what Strain et al. (2011) call the “Tutorial Instruction Myth”—that students with ASD can only benefit from individual instruction. Although we have evidence that individual instruction can be beneficial for teaching skills to students with ASD, one drawback is that this arrangement may further isolate children with social difficulties from peers who do not have these problems.

On the other hand, small group instruction, which doesn’t require the same isolation, has also been shown to be effective in many studies (for a review, see Ledford et al. 2012; Morrison et al. 2001). Social interactions between students with and without disabilities may be more likely in small group instructional formats when compared with large group, one-to-one, independent, and non-instructional arrangements (Carter et al. 2008; Dugan et al. 1995; Kamps et al. 1994). Instructors can also purposefully plan opportunities for social interactions during these small group activities; one way to do this is to train peers to engage in pro-social behaviors directed towards each other and the child with ASD (c.f., Ledford and Wolery 2013, 2014). These pro-social interactions may be important, since teachers spend relatively little time encouraging peer relationships (Irvin et al. 2014). Providing repeated and salient opportunities to observe peers engaging in pro-social behaviors might increase the likelihood that children with ASD will attend to and imitate peer models (when compared with pro-social behaviors occurring in more complex and less frequent contexts). Because children with ASD are likely to have deficits in the ability to learn information from their peers (Taylor and DeQuinzio 2012), increased salience and frequency may be crucial. Thus, teaching children with ASD in small groups may result in (1) effective instruction on individualized academic targets, (2) incidental peer interactions, and (3) planned and guided pro-social interactions that may result in increases in peer-related pro-social behavior.

The purpose of this study was to determine whether children with and without ASD could be simultaneously taught pro-social and academic behaviors during small group instructional sessions. Specific questions were: (a) Do students without disabilities (but who are at risk for academic failure; hereafter referred to as “peers”) learn to engage in specific pro-social behaviors during instruction given during a single social training session? (b) To what extent do peers continue to engage in these behaviors without prompting from instructors? (c) Do students with ASD learn to engage in prosocial behaviors after watching their peers engage in these behaviors during small group instruction? (d) Do all students engage in these behaviors during an unstructured generalization sessions? and (e) Do all students simultaneously learn academic material during small group activities?

Methods

Participants and Instructors

Each of five groups was composed of one student with ASD and one or two typically-developing peers who was socially competent but at risk for academic problems. Participants in Groups A–D were in kindergarten and were 5 or 6 years old; Participants in Group E were in 1st grade and were 6 or 7 years old. Seven students were African American and seven were Caucasian. Two participants identified as being of Hispanic descent via parent report; both (Ace and Brie; one student with ASD and one peer) spoke Spanish as their first language and qualified for services as English Language Learners. All peers spent 100 % of the school week in general education settings; target participants spent 69–100 % of time in general education with support from a special education resource teacher and special education paraprofessionals. Characteristics for individual participants are shown in Table 1. All target participants were rated as having low social competence and below average scores in the Communication, Assertion, Engagement, and Self-Control subscales of the Social Skills Improvement System (SSIS: Gresham and Elliott 2008), completed collaboratively by the general and special education teachers. As rated via that assessment, participants’ academic competence ranged from very low (1st percentile) to high (78th percentile; face validity of these ratings was confirmed by the special education teacher). See Table 2 for additional assessment information. All participants with ASD, despite receiving services in inclusive settings, received much of their academic instruction in a 1:1 arrangement. This occurred as individual instruction in self-contained or inclusive settings, or as ongoing 1:1 assistance during typical activities in the inclusive classroom (e.g., paraprofessional support to complete a worksheet while other students completed the same task independently after whole-group instruction).

Table 1 Characteristics of participants with ASD (bold text) and peers
Full size table
Table 2 Assessment information for participants with ASD
Full size table

Participants with ASD were nominated for participation by their special education teacher; peers were chosen by the special education teacher in consultation with each child’s general education teacher. Inclusion criterion for peers were: (a) teacher report of average or better social competence, (b) teacher report of below-average academic performance and need for additional small group academic instruction, and (c) parent consent. Inclusion criteria for participants with ASD were: (a) teacher report of low peer-related social competence, (b) social skills ratings in below average range on the SSIS, (c) teacher report of ASD diagnosis or educational eligibility or presence of ASD characteristics, (d) confirmation of ASD characteristics via above average ratings on Autism Spectrum subscale of the SSIS, and (e) parent consent. Within a group, all students were in the same general education setting; students in Groups A and B were in the same general education classroom while other groups were in different settings. The inclusion criterion regarding presence of either diagnosis or eligibility was devised so that children from minority or low-SES backgrounds were not inadvertently excluded, since these children often receive later diagnoses of ASD (Mandell et al. 2009; Mazurek et al. 2014), and because using educational determinations may result in a sample more representative of children with ASD receiving services in public schools.

Small group sessions were typically conducted by one of two first-year graduate students whose foci were related to child development but not special education and who had no experience with instruction. The author, who also infrequently conducted instructional sessions to demonstrate a technique or when graduate students were unavailable, trained both graduate students. All three conducted screening and initial probe sessions, to minimize classroom disruption and time required to conduct this testing.

Settings and Materials

All instructional, probe, peer training, and small group play sessions occurred in the special education classroom attended by all target participants. The room was not in use during the first 2 h of the school day because all students were served in inclusive settings during this time; one of the periods was also the resource teacher’s planning time. Thus, usually researchers and participants were alone in the room; the classroom teacher was sometimes at her desk doing paperwork when Groups D and E were participating. The room was approximately 20 × 40 m and included three tables, a reading area with beanbags, a large rug used for large group instruction, and a teacher’s desk, along with typical instructional materials. Instruction, probe, and peer training sessions occurred at a round table near the back of the room; small group play sessions occurred at a kidney-shaped table near the front of the room. Typical instructional (e.g., notecards), play (e.g., blocks), and reinforcement-related (e.g., tokens) materials were used in sessions; a list of materials used by session and trial type is presented in Table 3. Social statements printed on stimuli used in observational learning (TI + OL) sessions were either (a) suggested by teachers as commonly used by students or (b) used by typically developing students in instructional sessions occurring before social training; these statements are available from the first author on request.

Table 3 Materials used by session and trial type for each group
Full size table

Target Behaviors, Response Definitions, and Data Collection

All academic and social behaviors were chosen based on assessment reports and in collaboration with the special education teacher during an initial meeting.

Academic Behaviors

For initial probe conditions (probe trials only) and for instructional conditions (teaching and probe trials), data were collected on correct academic responding, defined as saying the written word (A–C, E) or geometric shape (D) presented by the instructor. Data were collected by instructor(s), during the ongoing session. During probe trials, no prompting was provided. Thus, each response was coded as unprompted correct (correct response within 4 s of question from instructor) or as unprompted error (no response or incorrect response within 4 s of question rom instructor). During teaching trials, two additional responses could also be coded: prompted correct (no response after question, followed by correct response after instructor model) and prompted incorrect (no response after question, followed by incorrect or no response after instructor model). In Figs. 1 and 2, only unprompted correct responses are graphed.

Fig. 1
figure 1

Unprompted correct responses for academic behaviors for Ace, Ben, and Cam

Full size image
Fig. 2
figure 2

Unprompted correct responses for academic behaviors for Dax and Eli

Full size image

Social Behaviors: Instruction

During instructional sessions, students had the opportunity to display pro-social behaviors towards group mates. For all groups, target behaviors were chosen by the special education teacher during a meeting with the first author; at this time, the author shared assessment information and the two adults discussed potential areas of weakness for each student. For Groups B and C, the target behavior was to share tokens given for correct responding. For Groups A and D, it was to provide social feedback on academic responding (e.g., “That’s OK” for an incorrect response). For Group E, it was to initiate comments towards peers (e.g., “I have 6 tokens!”). For Groups A–D, data were collected on a trial-by-trial basis and were calculated as a percentage of trials during which behaviors were independently and correctly emitted. For Groups B and C, the beginning of a social trial and the discriminative stimulus for responding was the placement of three tokens in the child’s hand or near his area on the table. For Groups A and D, the beginning of a trial and the discriminative stimulus for responding was the correct or incorrect responding by a peer or the indication from the instructor that the response was incorrect. For example, a child could have appropriately provided feedback (“That’s ok!”) immediately after the child said an incorrect response or after the instructor indicated that the answer was incorrect (“No, this is a circle”). For Group E, data were collected using timed event recording, using video records (e.g., initiating comments could be done at any time during the session); the number of comments per session is reported. Unprompted correct responses occurred when a student engaged in an appropriate and targeted social behavior with no instructor prompt and prompted correct responses occurred when a student engaged in an appropriate and targeted social behavior within 5 s of any instructor prompt. Unlike measurement for other groups, a specific number of opportunities did not occur for Group E; data were recorded as a number of occurrences rather than a percent of opportunities. No specific criterion was used for performance; consistent with single case design, a consistent and replicated change in behavior across conditions was evaluated. For Groups A–D, an additional criterion for correct responding was that it occurred within 5 s of the beginning of the trial. Unprompted errors (Groups A–D) occurred when a participant failed to engage in an appropriate behavior within 5 s of the start of the trial and no instructor prompts were given. Prompted errors occurred when a participant failed to engage in a correct behavior within 5 s of an instructor prompt. Unprompted correct responses are plotted in Fig. 3.

Fig. 3
figure 3

Unprompted correct social responses for participants with ASD (closed circles) and peers (open triangles) during instructional sessions

Full size image

Social Behaviors: Unstructured Play

A blind observer coded data from generalization sessions via video; these sessions were intended to serve as an assessment of generalization of social skills. The first author randomly ordered generalization videos and provided only general information on training to the coder (e.g., he was unaware as to whether sessions occurred pre- or post-social training and did not know what skills had been taught to participants). The outcome measure was the percentage of 5-s intervals during which the target participant engaged in an interaction that was directed to peers and that was both positive and successful. These included socially directed (looking at peer or saying his or her name) verbal or nonverbal behaviors that, if appropriate, elicited a peer response. Examples of coded behaviors included verbal and non-verbal initiations and responses (e.g., sharing a block, asking a peer a question, giving a compliment). Examples of non-coded behaviors included any verbal or non-verbal initiations or responses to researchers or other adults, negative behaviors like aggression or curse words, and unsuccessful interactions attempts (e.g., asking quietly for a block while looking away from a peer, who ignores the request). Because resources for coding videos were limited, researchers considered this as a stringent measure of inter-group social behaviors (e.g., required a change in interactions likely to be judged as significant by outside observers). For some groups, intervention was conducted only during the final generalization sessions; because adult behaviors changed during the sessions (e.g., prompts for social initiations were provided), coding for these sessions cannot be considered to be blind.

Procedures

Probe and Screening Sessions

Prior to small group instruction, words (Groups A–C, E) or geometric shapes (Group D) were presented to each participant in a 1:1 arrangement. The instructor presented each stimulus and asked the student to name it (e.g., “What is this?”). Children earned tokens for correct responding (known stimuli were interspersed) and for appropriate attending and participation behaviors (e.g., “Thanks for trying so hard!”).

Typical Instruction (TI)

During typical instruction (TI) sessions, students were taught in small groups using progressive time delay (PTD) procedures. When using PTD, instructional trials begin with a 0-s delay; the instructor provides the task direction and immediately prompts a correct response (e.g., “What shape is this? Circle.”). In later sessions, the instructor waits for gradually longer periods (1, 2, 3, 4 s) between the task direction and the prompt, to give the student a chance to answer independently. Error correction is often provided when a child gives an incorrect response, including reminders to wait (for unprompted errors; Wolery et al. 1992).

The first four (dyad; Group D) or six (triads; Groups A–C, E) trials for each group were probe trials. These trials were differentiated from instructional trials because stimuli were presented on small notecards with words handwritten with black marker; no instruction was provided. For these trials, students were instructed “First we are going to do cards. You may know the answers, but if you don’t, you will learn it later.” Then the instructor called the name of a student and asked “What’s this?” while presenting the first and second cards; then the second and third students were presented with two words/shapes in an identical fashion. These trials were presented in a predictable order [two trials for the student with ASD, two trials for one peer, two trials for the other peer (if applicable)] and tokens were available for correct responding.

Following these probe trials, instruction was presented for four trials per participant (8–12 total trials; two trials per stimuli). For instruction, trials were presented on letter-sized pages in a three-ring binder, with words presented in size 36 font. The instructor turned to the first page, said the corresponding participant’s name, and asked “What’s this?”. For 0-s trials, she then immediately said the answer. For 1–4 s trials, she paused the corresponding amount of time to allow the child to respond; if no response occurred, she said the correct response. A token was presented following correct responses (prompted and unprompted); incorrect responses (prompted and unprompted) were followed by “No, that was correct answer”. The instructor also said “Remember to wait if you don’t know” following unprompted correct responses. After these consequences, the instructor turned the page and read a sentence that corresponded to the target word (e.g., “And this says ‘The boy eats a cookie” or “An octagon has 8 sides”), pointing to each word in the sentence. A corresponding photo was printed below the sentence, but the instructor did not reference the photo. No consequences were provided if students repeated or read the sentence. To allow for structured opportunities for targeted social skills to occur for Groups B and C three tokens were given (rather than a single token) for correct responding during all trials throughout the study.

Peer Training

During a single 5-min session, peers were introduced to procedures to be used during observational learning (TI + OL) sessions. Instructors followed a script (available from the first author) during which they (a) explained that peers were going to be given opportunities to show that they were good friends, (b) introduced the target behaviors, (c) showed students the visuals and practiced reading them (if applicable), (d) modeled behaviors, and (e) provided one practice and feedback opportunity per peer. Students with ASD were not included in training sessions for research rather than practical reasons—because information is needed regarding ways to improve observational learning of peer behavior and increase prosocial peer-related behaviors without the need for adult assistance. Peers were instructed not to prompt their group mate with ASD to engage in social behaviors; this was sufficient for inhibiting peer prompting.

Typical Instruction Plus Opportunities for Observational Learning (TI + OL)

During TI + OL sessions, students were taught in a small group consistent with TI procedures. The only difference between TI and TI + OL sessions were the presence of visual cues and timing (TI + OL sessions occurred after peer training). Thus, in these sessions, peers modeled target social behaviors, which made it possible for children with autism to learn them observationally, but academic instruction was identical to the TI condition. Descriptions of visuals are below; they available from the first author. During the TI + OL sessions, teachers placed the visual cue materials beside each peer (between the peer and the participant with ASD). During the first TI + OL session for each group, she provided nonverbal prompts to engage in the target behaviors (pointing to the visual) only to the peers; following this session, pointing was used as error correction (when peers did not engage in the target behavior within the specified interval; this was scored as a prompted response). Only independent correct responses are shown in figures; because visual cues and peer modeling were a part of the intervention package, we cannot be certain to what extent each of those treatment components was response for behavior change.

  • Groups B and C (sharing): A clip art picture depicting three students holding hands placed near each peers token board. Students were given three tokens for correct responding (consistent with baseline) and were expected to share one token with each peer.

  • Groups A and D (saying nice things to your friends): A list of possible compliments and sympathetic feedback statements was provided near each peer’s token board and peers were expected to provide feedback to group mates. The list was presented in two columns, one in red text and one in green text. If needed, the instructor pointed to a red or green square to provide peers information about whether they should provide sympathetic (“That’s OK!”) or complimentary (“Nice!”) feedback.

  • Group E (talking to your friends): A list of comments that could be used to initiate conversational exchanges was generated for Group D.

Generalization Sessions (Unstructured Play)

These sessions were devised to determine whether increased social interactions would occur during an unstructured activity immediately following TI or TI + OL sessions. For the first three sessions across groups, each group was assigned to play with three different sets of materials (building bricks, building logs, and magnetic blocks). In subsequent sessions, a member of the group was asked to choose a play set for each session; each child was provided an opportunity to choose the play set every third session and choices were provided based on alphabetic order of students’ first names. During these sessions, instructors (a) provided a choice to one student; (b) assisted in opening the container; (c) set a 5-min timer; (d) did not prompt social interaction behaviors; and (e) helped and prompted clean-up of materials when the timer indicated 5 min had passed. In addition, instructors answered any questions or responded to any requests during the session. Students sat around a kidney-shaped table; the child with ASD sat in the middle. We understood that the natural consequence of this arrangement was to artificially deflate social exchanges between peers but felt it could facilitate social exchanges with the child with ASD.

Planned Modifications: TI + OL + Prompting

In order to facilitate behavior change while maintaining integrity of research questions, modifications to the TI + OL and generalization conditions were planned. For the TI + OL conditions, the modification was nonverbal teacher prompting added when five consecutive sessions of TI + OL did not result in behavior change for the participant with ASD. For Groups B and C, this consisted of graduated guidance for sharing tokens. Modifications were needed for Group B; they were not needed for the remaining groups. An additional unplanned modification was made for Group E when Eli began using feedback statements but did not consistently use appropriate statements (e.g., said “That’s OK” when Ella gave a correct response). Thus, after three sessions of TI + OL, a verbal explanation was given to Eli. Aside from these modifications, students with ASD did not receive direct instruction for engaging in the target social behaviors.

Planned Modifications: Intervention During Unstructured Play Sessions

Researchers planned a priori to do social skills instruction during generalization sessions in the final 3 weeks of the study if no apparent changes had occurred between sessions occurring immediately after TI and those occurring after TI + OL sessions. During instructional sessions, researchers made the same visuals available as used in instruction (different across groups), and told students they should also be “good friends” and “say and do nice things” during play. Examples given to each group included sharing toys, making comments, giving feedback, and asking questions. Praise was given when students engaged in any pro-social interaction. A 30-s interval timer (on a smart phone) was used to provide cues for engaging in prosocial behaviors. At the beginning of the session, the instructor told students that the timer would beep to remind them to use their “good friend” behaviors. No other prompts were given. Note that once instruction began in this condition, authors no longer considered it to be a context for the measurement of generalization.

Experimental Designs

Social and academic data were collected simultaneously, but in the context of two different designs. Social data from instructional sessions were collected in the context of a multiple baseline design across groups (Gast et al. 2014) with five tiers and four start dates (e.g., four potential demonstrations of effect). Social training for peers occurred on the same date for Groups C and D; this was done for logistical rather than experimental reasons, to reduce time in baseline for groups and because the end of the school year was near. Intervention began for Group A when data were stable; training was conducted for subsequent groups when data from the previous tiers showed consistent changes in level and trend.

Academic data were collected in the context of multiple probe designs across behaviors for each of 14 participants; for each participant, instruction began for behaviors assigned to the first tier when baseline data were stable and usually began in subsequent tiers when a pre-determined criterion was reached (90 % average over 3 consecutive days). In some instances, students met criterion at about the same time; in these cases, intervention for one participant continued in a tier for one or two extra days. This was done to simplify tasks for instructors (e.g., changing stimulus materials was minimized and prompt delays were consistent across participants). Summary data for academic behaviors for peers are described and shown in the “Results” section. Graphed data for peers are not included due to publication space constraints; they are available from the author.

Data from the two design types were not dependent; changes in academic behaviors were used to make decisions about condition changes in the context of multiple probe designs across behaviors and changes in social skills were used to make decisions about condition changes in the context of the multiple baseline design across groups. A group criterion was used for social behaviors—behavior changes by all participants were required before instruction started for the subsequent tier. For acquisition of academic behaviors, each participant served as his own control, with multiple potential demonstrations of effect within each participant. Therefore, individual criteria were used in independently operating designs—students all started instruction at the same time but moved to new academic material when they mastered an instructional set.

Interobserver Agreement and Procedural Fidelity

Interobserver agreement (IOA) and procedural fidelity (PF) data were collected during 33 % of initial probe sessions, 24–26 % of instructional sessions (including probe trials), and 29–38 % of generalization sessions for all participants in each group. IOA data were collected in vivo by the first author or via video by the first author or the graduate student who had not collected the primary data and were calculated on a trial-by-trial basis (Billingsley et al. 1980). PF data were collected via video by the first author using direct systematic recording (e.g., each occurrence or non-occurrence was recorded rather than a single yes/no checklist; Ledford et al. 2014). For instructional sessions, procedural steps included presenting the stimulus, gaining attention, asking the question, waiting, providing a prompt, giving the correct consequence (praise plus token, reminder to wait, or error correction), prompting target social behaviors, and praising target social behaviors. For generalization sessions, procedural steps included giving a choice of materials, seating the child with ASD between two peers, providing no prompts for social interactions, responding to child initiations, contacting children once per session to praise engagement, and providing a clean up direction at the end of the session. Average IOA was 99.6 % (range across groups: 97.9–100 %) for academic behaviors and 97.5 % for social behaviors (range across groups: 96.0–100 %) during small group instructional sessions. Average IOA for positive and successful social interactions during generalization sessions was 95.4, 98.5, 98.8, 97.4, and 98.4 % for Groups A–E. Average fidelity was 99.4 % for probe trials (range across procedural steps 97.5–99.9 %), 98.5 % for instructional trials (range across procedural steps: 97.4–100 %), and 98.9 % for generalization sessions (93.3–100 % by procedural step).

Results

Academic Behaviors

Thirteen participants learned all three sets of target behaviors; one peer (without ASD; Dani) learned more slowly than anticipated and was only taught two sets of target behaviors. As shown in Fig. 1 and Table 4, participants with ASD had an average number of sessions to criterion of 8.9, which is approximately 40 instructional minutes (average session length was approximately 4.5 min). Each participant with ASD learned all target behaviors, although patterns varied; Ace, Dax, and Eli had consistent and moderate to quick acquisition of stimuli to criterion levels, with limited variability. Ben and Cam had more variable data in at least one tier; anecdotally, both participants confused at least one target stimulus with previously taught sight words. According to teacher report, these students had the smallest repertoire of sight words. For peers, average number of sessions to criterion ranged from 5.7 to 20 (excluding Dani, range of 5.7–10.7). The average was 9.4 (including Dani) or 8.1 (excluding Dani). Dani appeared to be somewhat distracted by her partner, who often engaged in stereotypic, problematic, or off-task behaviors.

Table 4 Acquisition of targeted academic behaviors
Full size table

Overall, participants with and without autism learned at about the same rate, with the same number of trials per session. This was an expected outcome, since we asked teachers to nominate peers who were socially competent but at risk for learning problems.

Social Behaviors: Instructional Sessions

During instructional sessions conducted prior to implementation of the observational learning procedure, all participants engaged in zero or near-zero levels of target social behaviors, as shown in Fig. 3. During TI + OL sessions (following peer training), all participants with and without ASD increased their use of targeted behaviors. All participants in Group A demonstrated an immediate increase in the percentage of opportunities during which feedback was given, with a variable increasing trend throughout intervention. Peers in Group B showed an immediate increase with an increasing trend and many sessions at or near 100 % correct sharing of tokens. Ben required three sessions of graduated guidance before he demonstrated an increasing trend, which continued after prompting was withdrawn, and remained at 100 % for the final ten intervention sessions. Cam and his peers (Group C) demonstrated increasing trends to 100 % correct sharing behaviors; peers had an immediate change in level, while Cam began sharing after three intervention sessions. For Groups D and E, similar changes in level occurred, with variable and increasing trends for all participants and lower correct performance of commenting and feedback by Dax and Eli than by their peers. In summary, functional relations were exhibited for both the effectiveness of peer training on increasing social behaviors by at-risk students and for the effectiveness of peer modeling on increasing social behaviors by students with ASD.

Social Behaviors: Generalization Sessions

Because a blind coder was used, we did not make data-based decisions about condition changes for data collected during generalization sessions; changes in condition represent a change in the independent variable used during instructional sessions. All generalization sessions during Baseline and Intervention in Academic Groups were procedurally identical. Intervention during generalization sessions was started for Groups A–C; sufficient time was not available to conduct these sessions for Groups D and E.

As shown in Fig. 4, percent of intervals with successful social interactions varied across groups, conditions, and participant type. For Groups A–C, participants with ASD and their peers interacted very little during baseline sessions (sessions immediately following TI). During sessions immediately following TI + OL sessions, interactions increased for peers in all groups, with only small changes for the peers in Group B. For Ace and Cam, there were small increases in successful interactions, and for Ben there was no change. When intervention was implemented in the generalization setting, further increases were seen for Ace and Cam, but there was still no change for Ben. For Groups D and E, it was difficult to assess changes because of high or increasing interactions during the baseline condition which potentially coincide with intervention for other groups. For Dax and Dani, there was an increasing trend in baseline which leveled out and was stable for sessions occurring following TI + OL sessions. For the final group, Eli had minimal interactions in baseline while his peers had variable and increasing interactions. For sessions occurring following TI + OL sessions, peer interaction levels maintained and Eli’s interactions increased slightly, similar to the changes seen for Ace and Cam. Only three sessions occurred post-intervention for Group D; this was primarily due to classroom procedures that prohibited Dax from participating in the group contingent on misbehavior during other parts of the school day.

Fig. 4
figure 4

Successful social interactions (initiations and responses) for participants with ASD (closed circles) and peers (open triangles) during generalization sessions. Baseline sessions occurred immediately following TI sessions; Intervention in Academic Groups sessions occurred immediately following TI + OL sessions; data from Intervention in Generalization sessions cannot be considered true measures of generalization because instruction occurred during these sessions, which also followed TI + OL small group sessions

Full size image

Discussion

This study was designed to assess effects of small group instruction with and without observational learning opportunities on the acquisition of academic and social behaviors for kindergarten and first grade students with and without ASD. Results of this study extend previous research on small group instruction by including peers at-risk for learning problems. In public school settings where resources are limited, these students might benefit from small group academic instruction and may simultaneously serve as appropriate social models. This study also extended social modeling procedures to elementary school students; previous research has been conducted in preschools (c.f., Lane et al. submitted; Ledford and Wolery 2013, 2014). This is also the first study to confirm the effectiveness of simultaneous academic and social intervention in the context of small group direct instruction in public school classrooms. This study improved on generalization measurement by evaluating generalization of social behaviors in less structured activities, with frequent measurement, allowing for assessment of a functional relation.

Major findings from the study include: (a) students with ASD and at-risk peers successfully learned academic behaviors at similar rates, (b) at-risk peers learned to engage in social behaviors and consistently engaged in the behaviors during instructional settings, (c) students with ASD learned to engage in social behaviors during academic sessions (with one student requiring extra prompts that were quickly faded), and (d) variable generalization occurred in an unstructured context, with three of five participants with ASD demonstrating small increases in social behaviors and two of three demonstrating additional increases when intervention was implemented in the generalization setting.

Limitations

Limitations are primarily related to assessment of social behaviors during generalization: (a) due to limited resources, we selected a very stringent dependent variable that may not have been sensitive to behavior changes of participants, (b) limited generalization measurement occurred for Dax and Dani because of Dax’s inappropriate behaviors, and (c) limited assessment of intervention effectiveness during generalization sessions was possible due to the study concluding as the school year ended. The intervention was least efficient for Ben; he required the greatest average number of academic intervention sessions to criterion, required an additional intervention component for increasing social behaviors during academic sessions, and did not demonstrate and increase in social interactions during generalization sessions. Ben was the youngest participant and one of two participants with the lowest social competence (lower than the first percentile). It may be that younger students or those with fewer social skills might require additional instructional components or support compared with other students. Ecological validity of this study may be limited because indigenous implementers were not used and teaching was conducted outside typical contexts. Although social and academic objectives were chosen in collaboration with teachers, no social validity data were collected to assess whether procedures were feasible for use by classroom staff. Because PTD consists of several components (prompting, reinforcement, error correction), we cannot know which components are necessary for learning; the same is true for the peer modeling interventions for social behaviors.

Implications

This study showed that small group instruction for heterogeneous groups of students are possible and may be preferable to individual instruction due to student-to-teacher ratio benefits. In addition it may be possible to embed social opportunities during small group instruction, with no apparent effects on learning (e.g., students did not learn faster during TI sessions when compared with TI + OL sessions). In addition to academic benefits, teaching social behaviors during structured settings may result in some changes in behaviors in other settings, but reminders and reinforcement in typical settings may also be needed. The students with the highest rates of social behaviors during generalization settings were the students with the highest rated social competence (see Table 1). It may be possible that individual and explicit instruction is needed for individuals with lower initial social and academic competence. Alternatively, it may be that carefully structuring an activity more like the generalization context (e.g., teaching peers to consistently comment given a specific cue during play tasks) might result in more changes in unstructured contexts.

Directions for Future Research

Although the results of this study are promising, additional research in the areas of small group academic instruction and increasing social competence in small heterogenous groups is needed. Specifically, studies are needed to: (a) determine whether small group instructional procedures can be implemented as part of an inclusive classroom routine (e.g., while other students are engaging in other tasks) rather than as a “pull out” procedure, (b) determine whether indigenous implementers (teachers, paraprofessionals) can be taught to reliably and effectively implement the procedures, (c) further analyze the features of interventions which increase the generalization of learned skills to typical settings, and (d) determine what child, setting, or resource characteristics lead to the provision of 1:1 instruction.

The results of this study suggest that some students who primarily receive instruction in 1:1 arrangements may benefit from small group instruction, with increases in both academic and social behaviors. Thus, although additional research is needed, practitioners should consider structuring small group academic instruction so that children with ASD have sufficient opportunities to practice pro-social behaviors and to be the recipient of positive social interactions exhibited by peers.