Abstract
Joint attention (JA), defined as shared focus on an object or event with another person, is an important skill for social and language development. Autistic children may demonstrate differences in JA, and addressing JA may be included in naturalistic intervention. The study taught three behavior therapists (BTs) to implement the JA teaching procedure using behavioral skills training (BST). In addition, four autistic children learned to respond to a “show” bid for JA using socially based reinforcers and interactive stimuli using a multiple baseline across participants design. BTs demonstrated low accuracy of teaching JA bids during baseline. Following BST, the percentage correct increased to mastery levels and generalized to teaching the child participants to respond to JA bids. When JA training was introduced, performance levels of responding to JA moved from low levels to mastery levels for child participants and generalized to novel stimuli or people. BTs found both training procedures to be socially valid.
Joint attention (JA) is defined as the shared attention and interaction between two people and an object (Mundy et al., 1990). Differences in JA behavior are commonly seen in children who receive an autism diagnosis. In neurotypical children, JA usually emerges between 9 and 12 months of age, making its absence a reliable early marker for autism (Jones & Feeley, 2009; Osterling & Dawson, 1994). There are two types of JA, responding to joint attention (RJA) and initiating joint attention (IJA). RJA is defined as an individual responding to a social partner’s point, gaze shift, or gesture to an item or event in the environment. IJA involves an individual initiating an interaction (a “bid”) via a gesture, gaze shift, and/or comment toward an object or event in the environment for their social partner to respond to. JA behavior can be described as being comprised of multiple, discrete behaviors occurring in response to both environmental stimuli and the social partner’s behavior. For RJA, the component behaviors may include looking in the direction of the bid and back at the social partner. For IJA, the component behavior may include initiating the bid, looking in the direction of the bid, and looking at the social partner. These smaller responses operate in coordination between the two social partners to create dyadic interaction (involving two communication partners) and triadic interaction (involving two communication partners and a referent in the environment; Bakeman & Adamson, 1984). In both types of JA, the series of responses is hypothesized to be maintained by social interaction with the social partner (Dube et al., 2004).
JA, along with joint engagement, are often targeted as skills to measure and teach because they are considered to be foundational for more complex social-communicative development and language acquisition (Schertz & Robb, 2006; Wong & Kasari, 2012). JA measures such as the Early Social Communication Scale (ESCS; Mundy et al., 2003) note RJA as a skill predictive of later language acquisition (Mundy et al., 2007). Bakeman and Adamson (1984) identify supported joint engagement (shared parent-child interactions with an object, without the child referencing the adult) and coordinated joint engagement (the child actively engaged with an object and looking between object an adult, also called IJA) as early social behaviors that promote language development. The absence or reduced occurrence of JA and joint engagement skills likely lead to decreased early language learning for many autistic 1 children (Adamson et al., 2019; Warreyn et al., 2007). The presence of JA and language are positively correlated, making JA a common teaching target for autistic children (Charman et al., 2003; Waddington et al., 2021). Teaching RJA and IJA can also increase an individual’s social initiations, imitation, and independent language (Bottema-Beutel, 2016). Previous research has sought to identify effective teaching methods to establish JA skills.
Naturalistic developmental behavioral intervention (NDBI; Schreibman et al., 2015) includes naturalistic behavioral teaching approaches that have been demonstrated to be effective in establishing a repertoire of JA behavior in autistic children (Crank et al., 2021; Sandbank et al., 2020). NDBI approaches to teaching JA, including prompting and fading strategies, teaching in a familiar environment with naturally occurring social consequences, using highly motivating and interesting items during teaching, and teaching toward generalization and maintenance of skills in the natural environment (Frost et al., 2020), have been effective in teaching both RJA and IJA behavior. Whalen and Schreibman (2003) provided an early example in the literature of teaching RJA and IJA to autistic children. The researchers taught five autistic preschoolers to respond to a series of JA bids, gradually decreasing in intrusiveness from a hand on the toy being played with to a gaze shift. Four of the five children also learned to initiate JA. The participants also demonstrated JA in generalized settings during follow-up conditions.
Isaksen and Holth (2009) sought to increase JA in autistic children by using a JA task list made up of RJA, IJA, and turn taking activities. They first measured the participants’ levels of JA using the ESCS (Mundy et al., 1990). Responding to JA bids were taught first by presenting novel toys to the participant while pointing or gaze shifting. In the next phase, the researcher’s social response of nodding during multiple activities served as a discriminative stimulus for the participant’s behavior. The experimenter placed preferred items on a table and nodded/smiled when the child was granted permission to take the toy, thereby establishing smiling and nodding as conditioned reinforcers. The researchers used IJA and turn taking activities, such as block building or puzzles, to facilitate JA in a naturalistic environment. The participants learned to switch between responding and initiating JA bids while participating in a turn taking activity. All four participants of the study showed an increase in responding and initiating JA bids and maintained demonstration of the skills during follow-up.
Taylor and Hoch (2008) examined the effectiveness of training and social reinforcement on teaching JA to three autistic children aged 3 to 8 years old. The researchers divided JA into different components: gaze shift between object and adult, vocally responding to JA, and initiating vocal bids for JA. The experimenters placed salient, highly motivating items around the room. During training conditions for responding to bids of JA, the instructor pointed at an item and with appropriate prompting helped the participant orient to it, comment on it, and look back at the experimenter. Performance for each participant improved after training occurred and generalized to novel settings around the school.
Although both Taylor and Hoch (2008) and Isaksen and Holth (2009) utilized adult attending and social engagement as reinforcers for training JA responses, this approach of using naturalistic consequences and social reinforcers has been in the minority, historically. In their meta-analysis of JA teaching literature, Meindl and Cannella-Malone (2011) found that out of 18 studies that at the time had been published on teaching JA, only three had used social interaction as the reinforcer. As Dube et al. (2004) hypothesized that JA is maintained by adult attention and interaction, using social reinforcers to teach and maintain JA behavior would therefore be the best approach to result in behavior contacting naturally occurring contingencies and maintaining in the natural environment following training. Additional studies since 2011 have utilized social interaction as the reinforcer during JA training (e.g., Fredericks et al., 2023; Gomes et al., 2020), however, additional research to further demonstrate the utility of social consequences to establish and maintain JA is warranted.
Identifying socially based stimuli that may function as reinforcers can occur via a social reinforcer assessment. Smaby et al. (2007) identified effective social reinforcers for three autistic children by conducting social reinforcer assessments. They assessed responding (e.g., giving the experimenter a high five) in the absence of social interaction (the baseline condition) and when followed contingently by tickles, head rubs, or praise as social consequences. All participants demonstrated higher responding to social contingencies than during the baseline condition. The most effective social reinforcer varied across participants, indicating that social preference was idiosyncratic to the individual. Kelly et al. (2014) expanded this line of research by including observational and caregiver survey components to the selection of putative social reinforcers. These methods allowed further tailoring of the selection of potentially idiosyncratic social stimuli based on socially valid feedback and identified effective social reinforcers for all five participants.
While the effectiveness of JA teaching procedures is evidenced across a variety of studies (Meindl & Cannella-Malone, 2011), the topography of the JA response taught can vary depending on the study. Initiation of JA bids is often taught as a combination of pointing, gaze shift, and commenting (e.g., Fredericks et al., 2023). Less commonly, showing an object may be taught in place of pointing as an initiation gesture (e.g., Kasari et al., 2010). When teaching RJA, response to a point and gaze shift are the most common topographies taught, with Whalen and Schreibman (2003) and studies based on their methods, such as Ferraioli and Harris (2011), being an exception. In these studies, responding to a toy being shown was included in the shaping hierarchy of RJA responses. As RJA is closely correlated with language levels (Bottema-Beutel, 2016), continuing to examine ways in which researchers can teach RJA in the context of varied topography of bids may be an important continuation of this line of research.
Training behavior therapists (BTs) or caregivers to implement effective procedures to promote JA responses can be beneficial to increase both treatment integrity (TI) during training sessions and opportunities for generalization. Behavioral skills training (BST) involves providing instructions, modeling, role play, and feedback to teach complex skills (Parsons et al., 2013). BST is an effective strategy used to teach complex skills to BTs (e.g., Sarakoff & Sturmey, 2004). In teaching others how to implement JA training, Jones and Feeley (2009) trained three caregivers to teach their preschool-aged child (ages 3–4) to initiate and respond to JA bids. Researchers utilized modeling and feedback of a JA intervention with the caregiver and child dyads. The caregiver led the baseline and intervention procedures, where a highly preferred toy was held up, the child’s name called, and a visual sweep was used to prompt the JA response. Results indicated that the caregivers taught their child to respond to and initiate JA bids, which extended to a generalization probe that included novel toys.
The purposes of this study were to evaluate the effectiveness of behavioral strategies to (a) train BTs to implement RJA teaching opportunities with fidelity and (b) teach autistic children to respond to a “show” bid for JA. We used BST to teach three BTs to implement RJA sessions, testing for generalization of these skills during their implementation of RJA sessions with preschoolers with autism. Two additional children experienced RJA sessions with the BTs who participated in the study but did not need training in RJA sessions. To extend the literature on teaching JA responses using social consequences and addressing the spread of effects following this training, we focused on using naturalistic, social interaction within the context of training and during generalization probes.
Method
Participants
BT Participants
Three BTs who worked with autistic children were recruited as BT participants. BTs were recruited either through an email sent via the agency that was institutional review board (IRB) approved or through word-of-mouth suggestion by their current supervisor. The supervisors had been notified of the study through the email and suggested that the BT may find this research helpful in their clinical career. To participate in the study, BTs needed to be working at a center providing behavioral services to children with autism, have availability to participate, and score below 80% accuracy on the Joint Attention Training Checklist (see Table 1). They did not need to have any previous experience teaching JA skills. BT 1 was a first-year graduate student of behavior analysis and identified as a White female. She had been working in the field of behavior analysis for less than 1 year. BT 2 was a college junior and identified as a White female. She had been working in the field of behavior analysis for 2 months. BT 3 held a bachelor’s degree in a psychology-related field and identified as a Black female. She had been working in the field of behavior analysis for less than 1 year. These participants gave their own informed consent to participate. Two other BTs also participated as RJA trainers for the child participants. They performed with high integrity (above 80%) on the Joint Attention Training Checklist and therefore did not participate in the study as BT participants. We surveyed their views on the social validity of the study (summarized in Results).
Joint Attention Training Checklist.
The study was approved through the affiliated university’s IRB. Participants’ parents or guardians gave informed consent for their child to participate prior to the start of the study, and behavioral indicators of assent to participate in the study (e.g., willingness to go with the researcher, positive or neutral affect, absence of tantrum, whining, or leaving the area) were assessed continuously during all sessions of the study.
Child Participants
Four children participated in the study. All participants received a diagnosis of autism from an outside provider, confirmed via parent-provided report, prior to participating in the study. Participants attended a preschool in the Northeast United States of America for 20 to 30 hrs per week where they received 1:1 support services provided by BTs throughout their day. Participants attended integrated preschool classes with same-age peers in an attached child care center on individualized schedules based on their developmental needs. Participants were eligible to participate in the study if they had an autism diagnosis, were enrolled in the preschool, and could sit either at the table or on the floor for 2 to 3 min. For all participants, JA was identified as a potential teaching target by their educational team and/or via parental request.
Caleb was a 3-year-old White male. Caleb requested items that he wanted or needed using 1-word requests via a picture exchange system. Caleb participated in an inclusive classroom for 1 hr a day with support from his BT. He did not demonstrate JA skills during baseline.
Nora was a 6-year-old White female. She communicated vocal-verbally using three-to-four-word sentences. She showed strengths in completing self-help tasks independently. Nora participated in an inclusive classroom during the morning (about 3 hrs per day) with BT support. She was seen by an outside speech and language pathologist who consulted with her education team. She did not demonstrate JA skills during baseline.
Vijay was a 2-year-old male of White and South Asian heritage. He communicated with a combination of vocal approximations and one-to-two-word phrases using an augmentative and alternative communication (AAC) device (Proloquo2Go; AssistiveWare, 2024). His strengths included requesting for desired items and actions. Vijay participated in the inclusion classroom for 1 hr per day with support from his BT. He was seen by an outside speech and language pathologist who consulted with his education team. He did not demonstrate JA skills during baseline.
Shaun was a 4-year-old Black male. Shaun requested his needs and wants using one-to-two-word phrases on his AAC device (ProqoQuo2Go; AssistiveWare, 2024). Shaun participated in an inclusive classroom for 2 hrs a day with the support of a BT. He did not demonstrate JA skills during baseline.
Setting and Materials
Sessions took place in a large, 8 × 12 m classroom in the preschool dedicated to students working with 1:1 behavioral support. The classroom contained three preschool-height tables with chairs, a circle time area, bookshelves, toy shelves, a pretend play area, and a gross motor area. Sessions occurred at a 1 × 2 m table with child-sized chairs in one part of the classroom. Stimuli used in the study were identified individually for each participant and included small cause-and-effect and interactive toys, such as a light up spinner, a pin wheel, a top, a jumping frog toy, an articulated plastic snake, a puppet, and a push-button music book. The researcher kept the items in a bag or box during the session. Data collection materials, Joint Attention Training Checklist, and participants’ AAC systems (for Vijay and Shaun) were also present. Researchers recorded sessions using a handheld recording device positioned in the classroom.
Pre-Assessment: Social Reinforcer Assessment
Social Reinforcer Assessment Responding
Two child participants, Caleb and Shaun, completed a social reinforcer assessment as a part of the current study, prior to JA training, to determine the best reinforcers to use. For Nora and Vijay, social praise was established as an effective reinforcer prior to the current study. The response measured during the social reinforcer assessment for Caleb and Shaun was placing a block into a colored bucket. Placing a block in a bucket was defined as the participant picking up one block with one or both hands and releasing the block such that it fell into one of the buckets placed on the table. An example of placing a block was the participant bringing a block to the red bucket and dropping it inside. A nonexample of placing a block was picking up and dropping a block back on the table. Responding was measured via permanent product at end of session (blocks in bucket).
Dependent Variables
There were two dependent variables in the study: (a) percentage of correctly completed RJA training steps for BT participants, and (b) percentage of correct and independent RJA per opportunity for child participants.
RJA Checklist Completion—BT Participants
The dependent variable for the BT participants was independent completion of RJA training steps on the Joint Attention Training Checklist (shown in Table 1). Checklist accuracy was measured as percentage of correctly completed RJA training steps out of total steps in the checklist. The percentage correct was calculated by dividing the number of steps completed correctly by the total number of steps in the checklist and multiplying by 100.
RJA
The child participants completed JA training, in which the dependent variable was completion of response components that constituted responding to a bid for JA. The series of responses included attending to the toy and orienting toward the adult’s face. Attending to the toy was defined as the participant looking in the direction of the stimulus for at least 1 s as the adult held the item. A nonexample of attending was the participant looking at their hands when the toy was presented. Orienting toward the adult was defined as orienting head toward adult and looking in the direction of their face for at least 1 s, within 2 s of the discriminative stimulus being provided. Eye contact was not required as part of this response. A nonexample of orienting was continuing to look at the toy or at the wall after the adult provided the discriminative stimulus. Both components of the RJA response needed to be present for the response to be considered correct. This response was measured as percentage independent (responding correctly without a prompt) out of total opportunities per session. The percentage independent was calculated by dividing the number of correct independent responses by the total number of response opportunities and multiplying by 100.
Independent Variable and Experimental Design
The independent variables were BST for the BT participants and RJA training for child participants. The experimental design was a nonconcurrent multiple baseline across participants (Kennedy, 2022; Slocum et al., 2022). This was used to assess the effects of BST for the BT participants and RJA training for Caleb, Nora, and Vijay. A nonconcurrent multiple baseline design is a type of single case research design used to evaluate whether there is a change in the dependent variable due to the introduction of the independent variable, measured across multiple baseline-treatment comparisons that are not synchronized in real time (Slocum et al., 2022). The benefit of single case design is that the participant may serve as their own control and variability may be detected without combining data as is common when using group designs. Limitations of the nonconcurrent multiple baseline design are that it may not be able to account for events that are occurring at the same time as the independent variable, and participants may be in baseline for longer in the final tiers, delaying intervention (Slocum et al., 2022). An ABCDAB extended reversal design (Cooper et al., 2019) was used for Shaun. This design is a strong single case design that shows a functional relation between baseline and multiple treatment conditions by replication of the most effective treatment (in this case treatment B or least-to-most prompting).
Interobserver Agreement and TI
Interobserver agreement (IOA) and treatment integrity (TI) data were collected via video recording by independent, trained observers approved through the affiliated IRB. Secondary observers were trained to collect IOA and TI through initial instruction, practice, and feedback provided by a primary experimenter until 100% accuracy on a practice sample was reached.
IOA data were collected for baseline and training sessions during BST and RJA conditions. Secondary observers collected IOA data in 33% to 58% of sessions across participants. IOA scores were calculated by dividing the number of agreements by number of agreements plus disagreements and multiplying by 100. IOA data for accurate completion of steps on the Joint Attention Training Checklist was collected in 30% of sessions during the BST condition for BTs 1 and 2. IOA data could not be collected for BT 3 due to lack of video recording. IOA for BTs 1% and 2 was 100%. Mean IOA during the RJA condition for Nora’s sessions was 93.33% (range, 80%–100%), for Vijay’s sessions was 95.6% (range, 80%–100%), for Shaun’s sessions was 92.4% (range, 80%–100%), and for Caleb’s sessions was 97.6% (range, 83.3%–100%).
Secondary observers collected TI data in 30% to 34% of sessions during the RJA condition per child participant. TI data were collected on the experimenter or BT participant’s correct implementation of the steps of the Joint Attention Training Checklist (Table 1) during the RJA condition. TI data were calculated by dividing the number of steps correctly implemented by the total number of steps in the checklist and multiplying by 100. Mean TI for Nora’s sessions was 97.4% (range, 88%–100%), for Vijay’s sessions was 89.6% (range, 79%–98%), for Shaun’s sessions was 99.3% (range, 94%–100%), and for Caleb’s sessions was 100%.
Social Validity
BT participants received a social validity survey after completion of the study. The survey was designed to measure their view on the social importance of the child participant learning to respond to JA bids and the study’s overall success. Two out of three BT participants returned the social validity survey. BTs for Nora and Vijay, who were not participants in the study and did not need to undergo BST before running RJA training sessions, also completed a social validity survey at the end of the study. These BTs were included so that all individuals running the RJA training with child participants had the opportunity to provide feedback on the appropriateness of the teaching targets, procedures, and outcomes of the study. The BTs scored their answers to questions using a 7-point Likert-type scale, with response options ranging from strongly agree (7) to strongly disagree (1). Statements included in the social validity questionnaire were whether the treatment package was easy to implement, whether the intervention was successful in teaching the target skill, whether the steps of the checklist were clear and concise, whether participants would continue using this or similar BST approaches in the future, whether they would recommend this treatment package to families, and whether the child participant benefited from learning RJA.
Procedure
BT participants received BST, without child participants present, first. Once they could perform RJA training accurately during BST roleplays, they provided RJA training to the child participants. RJA training for child participants consisted of (a) Pre-Assessment: Social Reinforcer Assessment (Caleb and Shaun only), (b) RJA Training, (c) Maintenance/Generalization Probes.
BST for RJA Checklist—BT Participants
BST was provided to the three BTs who did not perform the actions on the Joint Attention Training Checklist at 100% during initial baseline sessions. The experimenter acted as the “student” during all BST sessions. During initial baseline, the experimenter provided a rationale for teaching JA and provided the BT with written instructions and the Joint Attention Training Checklist (shown in Table 1). The BT participant was instructed to follow the checklist. If the BT did not complete the actions on the Joint Attention Training Checklist at 100%, BST was conducted. BST sessions included modeling, roleplay, and feedback. During modeling, the experimenter demonstrated (physically and verbally) how to perform JA training. Roleplay included the experimenter and the BT practicing the JA intervention where the experimenter stood in for the child participant and the BT conducted the JA training according to the Joint Attention Training Checklist. Feedback was then delivered immediately after the roleplay sessions. The Joint Attention Training Checklist was available for reference by the BT participant throughout the session. Each BST session was 5 to 10 min. Mastery criterion was set at 100% across two sessions. BT participants needed two to five training sessions to reach mastery. Once the BTs were trained, they conducted RJA training. Mastery of their performance on the Joint Attention Training Checklist was periodically probed, and data were collected during RJA training sessions.
Pre-Assessment: Social Reinforcement Assessment
To ensure social interactions were highly reinforcing, we conducted a social reinforcer assessment for Caleb before the start of the study and for Shaun during his break from the study. This assessment was not included for Nora and Vijay as it was anecdotally indicated they enjoyed a variety of social interaction and social interaction was used as a reinforcer during their daily teaching sessions. The social reinforcer assessment was based on the procedures of Smaby et al. (2007) and Kelly et al. (2014). First, we asked the participant’s educational team for the participant’s top three preferred types of social interaction, from their perspective. Caleb’s and Shaun’s team listed both of their hypothesized top three preferred social interactions to be hugs, praise, and tickles. Therefore, we used these social interactions to assess their reinforcing efficacy in the social reinforcer assessment.
The social reinforcer assessment was conducted by the third and fourth authors. For this assessment, we placed three small buckets and a pile of blocks on a child-sized table with two child-sized chairs present. The different social reinforcers were represented by different colored pieces of paper (blue, red, and green) taped to the buckets. Blue represented tickles, red represented praise, and green represented hugs. The experimenter brought the participant to the table and asked them to sit down. The researcher initially conducted a sampling trial with each potential social reinforcer, in which the researcher gave the participant a block and gave the instruction to place the block in one of the buckets. As soon as the participant put the block in the bucket, the researcher immediately provided the corresponding social interaction to the participant for 3 to 5 s. Following this sampling procedure, the researcher presented the pile of six blocks to the participant and stated, “You can do more blocks if you want to” (or a similar statement). If the participant placed a block into a bucket, the researcher provided the corresponding social reinforcer. Three sessions were conducted. Each session lasted approximately 2 min. Summarized across sessions, Caleb chose hugs on 10 occasions, praise on 6 occasions, and tickles on 2 occasions. Therefore, hugs were selected as the social reinforcer to use in RJA training for Caleb. Summarized across sessions, Shaun chose praise on 10 occasions, tickles on 6 occasions, and hugs on 2 occasions. Therefore, praise was selected as the social reinforcer to use in RJA training for Shaun.
RJA Condition
Baseline
The BT implemented baseline sessions to evaluate the participants’ current RJA behavior. The BT held up a JA toy around their own shoulder height and perpendicular to the child’s eye level and said, “[Name], look!” “Check it out!,” “Wow!,” or a similar phrase. For Nora, Vijay, and Shaun, the verbal instructional cue varied across trials. For Caleb, only “[Name], look!” was included. JA toys were drawn from a bag of pre-selected toys. The toys included in the bag rotated throughout the sessions. If the participant looked at the toy and then oriented to the adult, naturally occurring reinforcers such as smiles, or praise were delivered, and the BT interacted with the toy and the participant for approximately 10 s. If the participant did not look at the toy or orient to the adult, no programmed consequence or prompt was provided, and the JA toy was put away. BTs provided preferred activities or non-JA toys, commented pleasantly on the participants’ play, or provided mastered tasks for 30 s to 1 min in between trials. Five trials were conducted in each baseline session. Participants moved to the intervention phase if they scored below a mean of 80% independence in baseline.
RJA Training
During the intervention phase, the BT or experimenter sat with the child either on the floor or at a table facing one another. The BT held up a toy to the child’s eye level and said, “Look!” or a variation. If the child looked at the toy and oriented to the adult, reinforcement was provided, and the child/adult interacted with the toy together for up to 10 s. If the participant did not independently look at the toy or orient to the adult, we prompted using least-to-most gestural prompts. The prompting hierarchy included “wiggling” the JA toy in place to draw attention to it, pointing to the JA toy, sweeping the JA toy partway toward the BT’s face, and sweeping the JA toy fully to the BT’s face. A point cue was not used for Caleb. For Nora, Vijay, and Shaun, naturally occurring social interaction, including smiles, praise, and interaction with the JA toy were used as reinforcement for both prompted and unprompted responses. For Caleb, these naturally occurring reinforcers were provided in addition to hugs (identified via his social reinforcer assessment) upon correct responding. Five trials were conducted in each training session, with intertrial activities as described in baseline. Each session was 5 to 7 min. Mastery criteria consisted of 80% independence or above, across two consecutive sessions. Participants experienced between 10 and 54 training sessions with a mean number of 23.75 training sessions (including modifications for Shaun, described below).
RJA Training Modifications (Shaun)
The experimenters included two modifications to the training protocol for Shaun during the initial training attempt. First, we modified the positioning of Shaun and the experimenter from sitting on two adjacent sides of the table facing one another to sitting in two chairs pulled out from the table and oriented “knee to knee” from one another, as this arrangement can encourage social interaction and increase skill acquisition (Van Houten & Rolider, 1989). Second, we isolated the JA toys used during sessions rather than rotated them, as Shaun seemed more interested in interacting with the BT with some of the JA toys over others; these more preferred toys were used consistently during this part of the study. When Shaun was reintroduced to the study later, we returned to the original training protocol.
Generalization
The approach to generalization probes varied across participants. For Caleb and Shaun, the BT who participated in RJA training conducted the generalization probes 2 weeks following completion of participant’s intervention phase with novel stimuli. For Nora and Vijay, a novel BT conducted generalization probes immediately upon completion of the participant’s intervention phase with the same rotation of training stimuli. For all participants, generalization procedures were identical to baseline. Each participant received one session of five trials during the generalization probe.
Results
BST for RJA Training Checklist—BT Participants
Figure 1 shows the results for the BT participants. The x-axis represents sessions whereas the y-axis represents percentage correct on RJA training checklist. All BT participants scored 0% on the Joint Attention Training Checklist in baseline. During BST, both BT 1 and BT 2 completed training and surpassed mastery criterion with 100% correct responding across the first two training sessions. BT 3 required five BST sessions to reach mastery criterion (M = 78%, range, 60%–100%). Following BST, mastery and generalization data for performance on the Joint Attention Training Checklist were collected during RJA training sessions. All participants maintained independent responding at high levels and generalized their responding to real-life training sessions (M = 100%).
Results for Behavior Therapists 1, 2, and 3.
Mean BT (participant) and BT (non-participant) responses were similar across groups. BTs in both groups agreed or strongly agreed with all statements, including that the training package was easy to implement (mean score, 6.5 for both groups), the steps were clear and concise (mean score, 7 and 6 for BT participants and non-participants, respectively), they felt the intervention was successful (mean score, 7 and 6.5 for BT participants and non-participants, respectively) and that the child participant benefited from learning the skill (mean score, 7 for both groups). They also all agreed or strongly agreed that they would recommend the use of this training to families (mean score, 7 for both groups) and that they would like to participate in similar trainings in the future (mean score, 7 for both groups).
RJA—Caleb, Nora, Vijay
Figure 2 shows the results for three of the four participants. The x-axis represents sessions whereas the y-axis represents percent of independent responding. The top panel of Figure 2 depicts results for Caleb. During baseline, independent RJA occurred at a near-zero level (M = 7%; range, 0%–20%). Immediately after least-to-most prompting was introduced, independent responding increased to a high level (M = 91%, range, 50%–100%) showing a steady increasing trend with mastery criterion met in 10 sessions. We conducted a generalization probe 2 weeks later with novel stimuli, and independent responding maintained at 80%.
Results for Caleb, Nora, and Vijay.
The middle panel of Figure 2 depicts results for Nora. Baseline independent responding occurred at a low to moderate level with moderate variability (M = 14%, range, 0%–40%). When least-to-most prompting was introduced, independent responding immediately increased and maintained at a high level (M = 96%, range, 80%–100%) showing an increasing trend with moderate variability with met mastery criterion met in 14 sessions. During a generalization probe with a novel BT, Nora responded with 100% independence.
The bottom panel of Figure 2 depicts results for Vijay. During baseline, independent responding occurred at a low to moderate level with moderate variability (M = 26%, range, 0%–40%). Vijay responded with 100% accuracy during the first intervention session, but independent responding decreased over the subsequent three sessions. Between sessions 15 and 21, independent responding showed a steep increasing trend. After an unplanned 2-week break in data collection, Vijay’s responding increased to the mastery criterion. During a generalization probe with a novel BT, however, Vijay responded with 60% independence. After another unplanned 2-week break in data collection, Vijay’s responding increased again to the mastery criterion, and maintained at 100% during a generalization probe with a novel BT.
RJA—Shaun
Figure 3 depicts results for Shaun. The x-axis represents sessions whereas the y-axis represents percentage of independent responding. During the initial baseline phase, independent responding occurred at a near-zero level (M = 7%; range, 0%–20%). Between Sessions 4 and 25, independent responding occurred at a moderate level with high variability (M = 38%; range, 0%–80%). It should also be noted that three unplanned, 3-week breaks in data collection occurred during this period. After the first procedural modification (see RJA Training Modifications), independent responding showed a decreasing trend, particularly after another unplanned 3-week break in data collection. After the second procedural modification (isolated stimuli), independent responding initially showed increasing trends, but after the additional break in data collection, showed decreasing trends.
Results for Shaun.
At this point, the research team determined that Shaun was not benefiting from the intervention and placed data collection on hold whereas his clinical team focused on other intervention goals. Nonetheless, this skill remained a priority for Shaun’s caregivers. Eleven months later, after consulting with Shaun’s caregivers and clinical team and determining that JA skills remained a priority, we reconducted baseline. Independent responding showed a decreasing trend (M = 15%, range, 0%––40%). Immediately after least-to-most prompting was introduced, independent responding increased to a high level (M = 96%, range, 80%–100%) showing a steep increasing trend and mastery criterion met in five sessions. We conducted a generalization probe 2 weeks later with novel stimuli, and independent responding maintained at 80%.
Discussion
In the current study, we evaluated the effectiveness of (a) BST in teaching BTs to implement RJA training and (b) the effectiveness of RJA training using social reinforcers to teach autistic children to respond to JA bid. In this study, we successfully taught three BTs to implement RJA training using BST. These findings extend the literature based supporting BST as an effective strategy for teaching complex skills to social service support providers (Slane & Lieberman-Betz, 2021). We also successfully taught four autistic children to respond to a JA “show” bid. The current study replicates findings from previous research (e.g., Gomes et al., 2020; Whalen & Schreibman, 2003) by demonstrating naturalistic teaching approaches to teaching RJA utilizing social stimuli as reinforcers are effective for many autistic learners. These procedures were effective in the current study for all child participants, with the fourth participant’s idiosyncrasies in skill acquisition reviewed below. During baseline in the present study, all participants demonstrated low to moderate levels of responding to JA bids. However, performance levels improved during training for responding to JA bids in three of the four participants initially and then the fourth participant following the second implementation of treatment. Responding also generalized to new BTs or stimuli, depending on the presented generalization condition.
Through the findings of the current study, we also extended the existing literature in two ways. First, experimental examination of teaching responding to “show” as a distinct operant has rarely occurred in the naturalistic teaching literature. Whalen and Schreibman (2003) included responding to show as a step toward shaping RJA responding to a gaze shift. Subsequent research replications of this study (e.g., Ferraioli & Harris, 2011) and training sequences for JA (e.g., Isaksen & Holth, 2009) have included response to a show in their trainings. However, most research studies experimentally evaluating teaching RJA have focused on teaching RJA to a point or gaze shift (e.g., Klein et al., 2009; Taylor & Hoch, 2008), thus leaving room for additional experimental evidence on effective processes for teaching a response to a “show” bid. The current study replicates and extends the findings of Whalen and Schreibman (2003), contributing to the evidence that RJA to a show can be taught using the same teaching strategies as other RJA topographies. These findings strengthen the generality of using naturalistic teaching procedures with socially based reinforcers to teach JA skills.
Second, the findings of the current study contribute to the literature by demonstrating the use of BST in teaching BTs to teach RJA. BST is well-established as a training package for a variety of complex skills (Parsons et al., 2013; Slane & Lieberman-Betz, 2021). We extended the research in this area by demonstrating BST to also be effective in teaching someone how to implement training for RJA skills. Regarding the BTs and their results, baseline conditions showed low levels of teaching JA bids. However, once BST was implemented, the BTs learned how to teach the participants to respond to JA bids. Their responding maintained throughout the study. These results add further validity to the use of BST in teaching complex skills to those working with developmental disabilities. They also add to the literature suggesting instructions alone are often not sufficient to teach new skills (e.g., Guinness et al., 2024). BTs of the child participants in this study viewed the RJA skill as socially valid and the methods of intervention as positive.
The results for Shaun are unusual in that RJA training was implemented twice, once without success and once with success. Training with Shaun was initially unsuccessful with highly variable data, and it was determined he was not benefiting from participating in the study. However, 11 months later, Shaun’s parents and clinical team continued to indicate that they viewed RJA as a valuable skill for Shaun to learn. Therefore, he was reintroduced into the study and reached mastery criteria for the skill within five sessions and demonstrating independent responding under generalization conditions. Teaching procedures during the second introduction of RJA training were the same as those used for Shaun in the initial least-to-most training condition. Maturation alone cannot account for the change in Shaun’s RJA scores, as during the second baseline he did not demonstrate the skill. However, approximately a year had passed since the last intervention attempt, during which Shaun was attending a preschool providing intensive behavior analytic intervention focusing on functional communication, language growth, school readiness, and social interaction. His history of reinforcement for engaging with his teachers and learning a variety of skills via prompting and reinforcement strategies may have been stronger and therefore allowed him to learn new skills more easily in the second introduction of the training.
In addition, we conducted a social reinforcer assessment at implementation of RJA training for the second time. This assessment demonstrated that praise functioned as the most effective reinforcer for Shaun. Praise was used as the social reinforcer in the first part of the study for Shaun as well, but it is possible that praise became more reinforcing between the first and second iterations of the training due to the reinforcement history in his school setting. Although unusual, Shaun’s results are encouraging and enlightening in a few ways. First, as Shaun’s parents and clinical team repeatedly noted that RJA was a preferred goal and objective for them, it is encouraging that the researchers reintroduced the training and Shaun succeeded upon the second opportunity. Second, unsuccessful data sets such as Shaun’s first attempt at RJA training are not always published (see Kepes et al., 2012, for a review of the “file drawer effect”). Although unsuccessful data sets may be acknowledged in some publications but not published (e.g., Turan et al., 2012), or partially successful data sets may be published (e.g., Young et al., 1994), it remains unknown how often terminated data sets are not included in final publications, or how often teaching procedures do not produce the intended effects, preventing the research from being reported at all. Omitting participants from final data sets, or the rejection of studies from publication that include “imperfect” examples of skill acquisition, may mislead readers on the generality of the findings to a wider target population than is appropriate or on the universality of the teaching process and expected outcome (see Branch & Pennypacker, 2013, for a review on generality of research findings). Shaun’s data set is therefore enlightening, in that it shows a participant for whom the procedures were initially unsuccessful and then successful. These data show an alternative profile of a participant for whom the teaching process was not linear but was ultimately successful, allowing readers observation of a data set that may not have been published under different circumstances.
There are limitations to the current study. First, BTs 1 and 2 received the same number of BST baseline sessions. This weakens the demonstration of experimental control, in that opportunities for prediction and replication of baseline rates of responding were limited to comparisons of BTs 1 and 2 to BT 3. BT 3 received more BST baseline sessions and independent responding remained at 0% under this extended baseline. Second, due to technical limitations no IOA data were collected for BT 3. In this circumstance, there was a lack of video recording of the sessions. This limits the assurance of accuracy in the data collected for BT 3. However, we can note that both primary data collectors were trained together, and the primary data collector for BT 3 was the secondary observer for BTs 1 and 2, for whom IOA was 100%. It is reasonable, but not assured, that data for BT 3 would be accurate given this set of conditions, which is why we chose to include the data, with the limitation noted here.
Limitations also exist in interpretation of the child participant’s data. Limited maintenance and generalization data were collected due to time constraints on the research project. Additional maintenance and generalization data, collected over longer periods of time and in varied contexts, would provide further evidence of the strength of the training procedures. Future research should include extended generalization and maintenance probes to examine the further spread of effects of RJA training. Aspects of the teaching procedures and generalization contexts varied across participants. In the current study, a compound discriminative stimulus was used throughout the study (show paired with a verbal cue of “look!” or another similar cue). For this reason, we cannot determine with certainty which component of the discriminative stimulus ultimately came to control responding. The compound discriminative stimulus has precedence in JA training (e.g., Jones & Feeley, 2009; Taylor & Hoch, 2008) but does leave questions on the necessary and sufficient components of the discriminative cue that may be best answered with a component analysis of the training procedures (Ward-Horner & Sturmey, 2010). Furthermore, the only instructional cue used in Caleb’s RJA training sessions was “Caleb, look!” This cue functioned as a salient discriminative stimulus for Caleb, and he learned to respond to a show quickly; however, we do not know how Caleb would respond to similar, untrained discriminative cues, limiting the generality of his findings. Further research should evaluate the role of verbal discriminative stimuli in JA bids. As an additional limitation in the current study, not all participants received a social reinforcer assessment to evaluate potent social reinforcers. Although all participants learned the skill using social reinforcement, without a social reinforcer assessment, we do not know if we used the most preferred or effective social reinforcer to teach the skill.
We can also highlight opportunities for future research stemming from the results of this study. The positive effects of teaching JA on vocabulary development are noted in the literature but the effects of teaching an RJA “show” response specifically are unknown (see Manwaring & Stevens, 2017, for a review). Evaluating the collateral effects of teaching responding to a show response in the establishment of new vocabulary could be evaluated next in this line of research. Ways in which RJA may function as a prerequisite to the establishment of a differential observing response (Farber et al., 2017) or increase the salience of environmental items such that instructive feedback may promote learning (Vladescu & Kodak, 2013) could be further examined. Although the use of social reinforcers and naturalistic teaching practices in teaching young autistic children has become more widespread since Whalen and Schreibman’s (2003) JA training article, continued evaluation of the long-lasting effects of a naturalistic teaching approach in teaching and maintaining functional and social skills is an important continued area of research (Frost et al., 2020; Schreibman et al., 2015).
Through the present study, we demonstrate that BST can be implemented to teach BTs who are relatively new to the field to implement RJA skills training for autistic children. We used naturalistic teaching strategies and social reinforcement to teach RJA. Learned skills maintained and generalized for both BTs and child participants. The results support a beneficial intervention that addresses a fundamental area of social communication often targeted for autistic children.
Footnotes
Authors’ Note
This study was completed in partial fulfillment of a Masters in Science in Applied Behavior Analysis by the third and fourth authors.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Ethical Approval
This study complies with standard ethical research practices. This study was approved through the IRB of the affiliated university.
Informed Consent
All participants gave informed consent to participate in the study.
Data Availability
Access to data from this study is available upon request.