Skinner (1953) defined generalization as occurring when stimuli controlling behavior in the training environment transfer to stimuli with shared properties outside of the original training context. Stokes and Baer (1977) oriented applied behavior analysts to a more deliberate process of generalization by reviewing the extant literature and classifying generalization into nine different guidelines: train and hope, sequential modification, introduce to natural maintaining contingencies, train sufficient exemplars, train loosely, use indiscriminable contingencies, mediate generalization, train to generalize, and program common stimuli. Despite the socially significant implications around adequately programming for generalization, and behavior analysts’ recognition of the importance of what Stokes and Baer lamented, generalization strategies remain an understudied area of applied behavior analytic research and practice (Osnes & Leiblein, 2003).

Applied behavior analysts face the challenge of planning for behavior change in relevant contexts, with the necessary behavior-change agents, materials, and other person and setting-specific variables (e.g., caregivers, classroom stimuli). Consideration of these variables during generalization programming can be laborious and time-consuming, so practitioners should strategize efforts to promote generalization as early as possible (Stokes & Baer, 1977; Stokes & Osnes, 1989).

One area of focus should be on the individuals responsible for implementing the behavior-change procedures: in particular, caregivers. For example, Gianoumis and Sturmey (2012) investigated the prevalence of generalization programming when conducting staff and caregiver training according to the guidelines described by Stokes and Baer (1977). The authors reported generalization programming in 89% of studies, with 30% of studies specifically targeting caregiver training. However, limited information was available on the process of programming for generalization with specific reference to caregiver behaviors. Few studies focus on strategies aimed specifically at caregiver generalization of behavior plans to reduce child challenging behavior beyond the training setting (i.e., generalization of caregiver implementation; Cordisco et al., 1988; Fryling et al., 2012; Sanders & James, 1983).

Additional research is needed to help guide behavior analysts’ strategies for targeting the generalization of caregiver behavior. At the start of the COVID-19 pandemic, telehealth was a means to continue delivering effective behavior interventions in the timeliest manner possible. Fortunately, the increasing prevalence of telehealth services opens a new pathway to programming for generalization through increased opportunities to observe and train caregivers in relevant treatment contexts. Thus, the provision of telehealth services and increased opportunities for observing caregivers and children in their natural environment highlighted useful tactics related to the generalization of caregiver behavior.

The Application of Telehealth to Promote Generalization

A growing area of research and practice with the potential for increased opportunities to solidify generalization programming for caregiver behavior is the provision of applied behavior analysis (ABA) therapy through telehealth (e.g., Ferguson et al., 2019; Leyser et al., 2021; Neely et al., 2021; Schieltz et al., 2022; Schieltz & Wacker, 2020). In general, telehealth includes substantial caregiver involvement so that ABA practitioners can provide in-the-moment coaching on how to conduct certain skill acquisition or behavior reduction procedures for children with developmental disabilities such as autism spectrum disorder (for a review, see Ferguson et al., 2019). Telehealth also offers a variety of advantages that can facilitate successful generalization and behavior change, including access to natural environmental contingencies and relevant behavior-change agents, contexts, and stimuli.

Although telehealth as a service model provides a unique opportunity to promote generalization because training is more likely to occur in the client’s natural environment, surprisingly little focus on generalization outcomes is available within the telehealth literature. Ferguson et al. (2019) completed a systematic review of studies conducted via telehealth within ABA. In the sample of studies that reported using a single-case design, generalization was included in only 40% of investigations. Thus, these findings suggest a paucity of research on the generalization of treatment outcomes within behavior analytic telehealth treatment studies.

The purpose of this article is to describe how telehealth facilitated generalization programming for caregiver behavior, informed by the guidelines set forth by Stokes and Baer (1977). The goal of the services described herein was to develop a function-based treatment to decrease challenging behavior and increase appropriate alternative behavior for an adolescent male who demonstrated severe challenging behavior. The admission occurred early in the global pandemic when onsite services were limited, and the family lived over an hour away from the clinic. Therefore, all assessment and treatment sessions were completed via telehealth in the family’s home, with the caregiver conducting all sessions. The behavior therapy team provided coaching and feedback from a remote clinic or home location. The team conducted the typical assessment and treatment process, including a caregiver-conducted functional analysis (FA) and subsequent function-based treatment with a focus on generalization programming within the family’s natural environment.

Case Background

John was a 17-year-old male diagnosed with Down syndrome, disruptive behavior disorder, and moderate intellectual disability. John communicated vocally using 1–4 word phrases; however, his articulation could be difficult for unfamiliar people to understand. John demonstrated severe aggression and disruptive behavior that previously resulted in suspensions from school, with risk for out-of-district placement, and multiple visits to the emergency room in the year before his admission. Aggression included hitting, kicking, punching, pushing/shoving, pulling on, slapping, choking, grabbing, or attempting to grab items out of another’s hands, scratching, throwing objects at another person, body slamming/charging others/dropping others, limb twisting, or spitting on others. For example, at one point, John attempted to pick up his mother and slam her down on the floor. Disruptive behavior included tearing and swiping items, pushing over furniture, and throwing, kicking, hitting, banging on, or breaking objects. For example, John would sometimes throw preferred leisure items across the room if his mother was not interacting with him as he requested. John previously received some behavior supports at school, but no other forms of behavioral intervention.

John’s mother was a neurotypical 46-year-old female with a master’s degree, but no formal training in ABA, or any experience with interventions that included parent training. She was a single parent and John’s primary caregiver. John’s older brother also lived in the home. John’s mother had significant concerns involving his severe challenging behavior, due to the impact on the family’s daily life. Due to the severity of John’s challenging behavior, he was referred to an intensive outpatient program for the assessment and treatment of severe challenging behavior. Appointments during his admission were 5 hr per day, and the behavior therapy team worked with John and his mother in the intensive format for 15 consecutive weekdays. All appointments were conducted through a web-based conferencing site (Zoom) using a computer or phone, and wireless headphones worn by the mother similar to a bug-in-ear device. At first, sessions were conducted in the family’s basement, but then extended to all areas of the home. The mother received coaching from a behavior therapy team that consisted of four trained staff ranging from bachelor’s level to doctoral-level practitioners, under the supervision of a licensed psychologist and board certified behavior analyst who provided daily supervision of the clinical services, consultation, and recommendations as needed. All team members had a minimum of 6 months of experience in the assessment and treatment of severe challenging behavior.

An FA was conducted based on the procedures described by Iwata et al. (1982/1994), with the addition of a tangible condition (Day et al., 1994). Due to low rates of challenging behavior in the initial multielement FA, the behavior therapy team initiated a pairwise comparison of alternating control and tangible test conditions (data available upon request). John continued to exhibit low rates of challenging behavior across both conditions, so the team completed a record review, discussed previous behavior incidents with his mother, and completed direct observations of John and his mother. Taken together, these findings suggested a possible mands compliance function and led to the completion of a mands analysis based on the procedures described by Bowman et al. (1997).

During the mands analysis, test and control conditions were randomly alternated in a pairwise comparison, using a multielement design. During the test condition, the mother played with John’s highly preferred activities and toys. When John requested that she stop or play another way, she did not respond to appropriate requests (i.e., mands) and continued to play in the manner that she chose. Contingent upon challenging behavior, she immediately complied with his mands and played “his way” for 30 s. She repeated these procedures for each 10 min test condition. During the control condition, the mother complied with all of John’s mands, did not place any demands, and frequently reminded him that they were playing “his way.” His mother was instructed to ignore any challenging behavior during the control condition.

Figure 1 depicts the results of the mands analysis. John only exhibited challenging behavior during the test condition, suggesting that his challenging behavior functioned to increase the probability of caregiver compliance with mands. We then developed a function-based treatment and evaluated the treatment within an ABAB reversal design. During the treatment evaluation, the A phases included baseline (test condition) contingencies, and the B phase included functional communication training (FCT; Falcomata & Wacker, 2013), extinction, a multiple schedule (with schedule thinning in the second B phase), and a visual daily schedule. Baseline sessions included the original test condition from the mands analysis or a replication of the test condition contingencies during the return to baseline. During FCT, John was taught to request “my way,” which served as an entry mand for subsequent requests to be reinforced by his mother for a brief interval. In particular, after John emitted “my way,” his mother reinforced all reasonable requests for 30 s. Extinction was programmed for all challenging behavior.

Fig. 1
figure 1

Mands Analysis. Results from the mands analysis portion of John’s functional analysis

Full size image

After the initial treatment package produced consistent reductions in challenging behavior, we systematically modified various aspects of the treatment components to program for generalization of caregiver behavior across trained and untrained activities, contexts, and materials within the home. A multiple schedule was implemented to facilitate schedule thinning. The multiple schedule included two components, signaled by a red and green card, that corresponded with the availability of reinforcement for functional communication responses (FCRs; green, discriminative stimulus, SD) and unavailability of reinforcement for FCRs (red, delta stimulus, S∆; Bowman et al., 1997; Saini et al., 2016). Extinction continued to be programmed for challenging behavior across both schedule components. Schedule thinning proceeded by initially introducing the S∆ for 1 min, and eventually increasing the S∆ interval to the terminal goal of 15 min, which was based on caregiver input. At first, leisure materials (as were present in the FA) were the only items incorporated, followed by the systematic introduction of both academic demands and chores. A visual schedule was introduced after the team observed clinically significant reductions in challenging behavior during sessions that included leisure materials, academic demands, and chores. The mother used her discretion to select the order of activities on the visual schedule at the beginning of each morning and updated it for the remainder of the day after lunch. The schedule consisted of different sequences of demands, leisure activities, and meal/break times. Once John demonstrated minimal levels of challenging behavior after the scheduled activities were introduced, we moved treatment sessions to all areas of the home to allow for the completion of a variety of relevant daily activities.

Trained therapists coached the mother on the treatment components using behavioral skills training and in-the-moment feedback during or between each session as needed. However, once the terminal S∆ period of 15 min was active during treatment sessions, the therapists minimized instruction and feedback to the mother to promote more independent problem solving. The team collected data on John’s rates of challenging behavior and FCRs, his percentage of following instructions, and the mother’s percentage of correct implementation of behavioral treatment plan components (i.e., procedural fidelity). Figure 2 depicts the results from John’s treatment evaluation, which show an overall decrease in challenging behavior, with acceptable levels of caregiver fidelity with treatment components throughout the treatment evaluation.

Fig. 2
figure 2

Treatment Evaluation. Results from John’s treatment evaluation conducted via telehealth during a 3-week intensive outpatient program admission and follow-up sessions. ST = schedule thinning began; numbers indicate the number of minutes that the S∆ period was active during ST

Full size image

Two data collectors recorded data simultaneously, but independently, for 33% of the baseline sessions, 57% of the treatment sessions, and 45% of the final treatment extension sessions. Exact-interval agreement (baseline and treatment sessions) and exact agreement (for the final generalization sessions) methods were used to calculate interobserver agreement (IOA). For exact-interval agreement, each session was divided into 10 s intervals and two independent observers recorded the presence or absence of challenging behavior or FCR. An agreement was defined as both observers reporting an occurrence of the target behavior within the same interval (or session). A disagreement occurred if one observer recorded the presence of the target behavior, whereas the other observer did not. The number of agreements was divided by the total number of agreements and disagreements and divided by 100 to obtain a percentage. Overall, the IOA for challenging behavior during baseline was a mean of 95.4% (range, 92.5%–98.3%) and during treatment was a mean of 99.6% (range, 99.3%–99.9%). IOA for FCR was 98.8%. IOA during treatment extension sessions was 100% for challenging behavior, and a mean of 95% (range, 75%–100%) for FCR.

Two independent observers also recorded procedural fidelity on John’s mother’s implementation of procedures. John’s mother maintained high procedural fidelity throughout all phases of the procedures (collected for 100% of sessions, M = 93.5%, with a mean IOA of 87.4%). She also demonstrated successful extension of the treatment plan to novel and untrained situations according to both self-reports and therapist observation. Moreover, the mother found the treatment high in acceptability according to an adapted version of the Intervention Rating Profile (IRP; Martens et al., 1985) with a score of 71 out of 84 (see Table 1). The IRP score suggests a socially valid treatment and outcome. Positive intervention outcomes were maintained during follow-up bimonthly, monthly, and quarterly appointments.

Table 1 Scores from John’s Mother Related to Social Validity
Full size table

In the following sections, we describe the relevant Stokes and Baer (1977) guidelines that helped with the generalization of caregiver behavior. In particular, we focused on how we incorporated: Use of Natural Maintaining Contingencies, Train to “Generalize,” Program Common Stimuli, Train Sufficient Exemplars, Train Loosely, Mediate Generalization, and Indiscriminable Contingencies. We provide conceptual points and empirical findings from the literature and our case example to illustrate how telehealth facilitated the generalization of caregiver behaviors related to the intervention for severe challenging behavior (see Table 2).

Table 2 Caregiver Generalization Guidelines and Programming Examples
Full size table

Generalization Guidelines

Introduce Natural Maintaining Contingencies and Train to “Generalize”

One method of promoting generalization is teaching an individual to exhibit behaviors that are likely to maintain under natural contingencies (Stokes & Baer, 1977). Teaching caregivers how to implement treatment strategies that help them recruit reinforcement for treatment-related behaviors in natural environments is critical for generalization. Likewise, caregivers should be trained to “generalize.” Stokes and Baer (1977) defined training to generalize by stating, “If generalization is considered a response itself, then a reinforcement contingency may be placed on it, the same as with any other operant” (p. 362). Thus, any time an organism is observed engaging in an untrained response, reinforcement should be delivered.

At first, the caregiver’s adherence to the treatment components was likely maintained, at least in part, by both negative and positive reinforcement contingencies from the treatment team. When the mother demonstrated a procedural error, the therapists provided corrective feedback and reminders. Following correct responses, the therapists delivered praise and approval for correct treatment implementation on a dense schedule (Allen & Warzak, 2000). In other words, correct responding on the mother’s behalf may have been maintained by immediately available contingencies related to escape or avoidance of therapist feedback, and access to therapist praise. Likewise, therapists provided praise  any time the mother was observed or self-reported generalized use of any treatment components to an untrained context. For example, the therapists praised the mother’s  creative use of signaled availability outside of the treatment sessions. In particular, the mother visually showed John it was now the “caregiver’s way” (red card) through video chat, and she requested that he return home from his grandmother’s house. She reported that she was prepared to go to the grandmother’s house and follow through with the prompt for him to return home if needed; however, he returned home independently. This specific adaption of signaled availability had never been directly trained or discussed. Nevertheless, the therapists praised the mother’s demonstration of this unique treatment adaptation (i.e., train to “generalize”).

Following the consistent implementation of the treatment with high levels of procedural fidelity, and extension of treatment components across the day, the mother’s behavior was likely reinforced by increases in John’s appropriate alternative behaviors and decreases in challenging behavior. As we extended treatment across various environments and activities during sessions, it is possible that his mother’s adherence to treatment was reinforced by John’s ongoing instruction following of a diverse range of demands. John began independently completing novel tasks such as meal preparation and washing the dishes. He also completed historically challenging tasks, including making his bed or playing nonpreferred leisure activities with his mother. Because of John’s completion of chores, his mother reportedly gained increased opportunities to engage in her own preferred activities (i.e., a naturally occurring positive reinforcement contingency). In addition, the mother was now relieved of required chores and task demands that John was taught to complete, and she began to experience a safer home environment with overall reductions in challenging behavior (i.e., natural occurring negative reinforcement contingencies; Stocco & Thompson, 2015).

Delivering the training through telehealth provided the treatment team regular access to contingency arrangements that were intrinsic to the family’s natural environment. Access to these contingencies allowed the team to integrate the mother’s naturally occurring reinforcers into treatment. For example, instead of John following directions with some approximated version of a chore in a controlled environment (e.g., clinic room), the actual task that the mother reported difficulty with (e.g., making the bed) was used as a task during the session. The inclusion of this type of task may have contributed to the mother’s positive endorsement of the treatment components. John’s completion of a relevant task that was immediately meaningful to his mother was likely a positive reinforcer that supported the mother’s implementation of specific prompting and escape extinction procedures. On the other hand, if the team guided the mother through the completion of a simulated task with John, she may have experienced this as less meaningful and, in turn, the experience would be less reinforcing (or not reinforcing at all) for treatment-related behaviors.

Access to materials in the family’s natural environment allowed the treatment team to demonstrate the relevance of treatment components during sessions, providing practice on tasks that the mother could integrate into the daily schedule almost immediately. For John, the increased diversity of tasks led to more exposure to the reinforcement and extinction contingencies in the natural environment. Likewise, and perhaps more important for treatment maintenance, John’s mother received multiple exemplar training on delivering a variety of tasks. The multiple exemplar training helped the mother sample the various naturally occurring reinforcement contingencies available in the home, contingent on the implementation of the behavioral treatment plan.

Program Common Stimuli

Stokes and Baer (1977) defined programming common stimuli as a “. . . guarantee that common and salient stimuli will be present in both [training and generalization settings]” (p.360). Common stimuli may include both physical and social stimuli. Stimuli presented during the training context likely acquire discriminative properties over time that can then evoke similar responses in untrained contexts (Stokes & Baer, 1977). The stimuli involved in treatment implementation (e.g., multiple schedules) are available to become common stimuli that evoke caregiver implementation of treatment plan components in the presence of practitioners and, most important, while implementing the treatment in natural settings.

One example of programming common stimuli into behavioral treatment is the use of a multiple schedule arrangement to simultaneously reduce challenging behavior, establish stimulus control over alternative behavior (Saini et al., 2016), and prompt a caregiver’s treatment-related behaviors in natural environments. Saini et al. (2016) described 147 applications of multiple schedules across 31 empirical studies with children and adults with and without developmental disabilities and noted an overall increase in the frequency of multiple schedule use reported in contemporary applied research. Multiple schedules include two or more schedules of reinforcement, extinction, or punishment presented in an alternating sequence with correlated stimuli signaling the schedule that is in effect (Hanley et al., 2001; Saini et al., 2016). The correlated stimuli can signal the availability (i.e., SD) and unavailability (i.e., S∆) of a relevant reinforcer(s), such that learners can come to discriminate when a specific response (e.g., communication) may be reinforced. For example, a green card can function as an SD for a specific response (e.g., a break from work is provided contingent on the appropriate request), whereas a red card functions as an S∆ (e.g., a break from work is withheld even after the appropriate request). After establishing stimulus control over alternative and challenging behaviors, schedule thinning occurs by systematically decreasing the SD and increasing the S∆.

The goal of multiple schedules should be to teach caregivers how to establish stimulus control of their child’s responses by creating clear signals for when their child’s responses will and will not be reinforced. Multiple schedules also provide the parent with clear signals for how and when to change their behavior in accordance with the different schedules. For example, in the current case example, when the mother indicated to John it was, “his way,” reinforcement was programmed for John’s FCRs. John’s mother continued to engage with activities or requested that John complete a task, but the signaled availability of reinforcement alerted the caregiver to be prepared to honor FCRs or implement three-step prompting for task completion if needed. When the mother indicated it was “caregiver’s way,” extinction was programmed for John’s FCRs, and the caregiver proceeded to engage in activities that did not require her availability to reinforce John’s behavior.

Campos et al. (2020) trained three parents using behavioral skills training to implement FCT and a multiple schedule to treat their children’s challenging behavior. Parents continued to utilize multiple schedules in their homes with their children at 100% accuracy post training. All children demonstrated discriminated requests when multiple schedules were implemented and an 80% reduction from baseline was observed for challenging behavior. Caregivers all rated the procedures with high social validity, endorsing the procedures as “very acceptable,” and being “very willing” to implement the procedures, and two parents rated the procedures as “very likely” and “almost very likely” having advantages. Parents also reported using the procedures “every day,” “multiple times a day,” and “often” in their natural environments. Thus, beyond the findings from our case example, a larger body of empirical support suggests that caregivers consider the effects of multiple schedule implementation to be a valuable component of behavioral treatment plans.

Train Sufficient Exemplars

Another generalization programming strategy includes teaching across multiple exemplars. Stokes and Baer (1977) defined training sufficient exemplars as, “. . . teaching another exemplar of the same generalization lesson [following one successful generalization lesson], and then another, and then another, and so on until the induction is formed. . .” (p. 355). By training multiple exemplars, the target individual experiences many variations of stimuli, or responses, and can thereby contact reinforcement in the presence of different stimuli. Thus, the individual is more likely to respond correctly when novel stimuli or responses share similar dimensions to the trained examples, and less likely to emit the desired response when they do not.

Crockett et al. (2007) found that some parents require more extensive multiple exemplar training to generalize treatment-related behaviors (e.g., Melissa, p. 34). Conducting sessions through telehealth allows behavior therapy teams to obtain a myriad of materials, example tasks, and activities to include in multiple exemplar training. In the current case example, we used multiple exemplar training frequently throughout the treatment process. Therapists trained the mother to implement three-step hierarchical prompting, reinforcement, and extinction, within a multiple schedule arrangement, across a wide variety of academic tasks and household chores, as well as leisure activities. Training sufficient exemplars appeared to increase the likelihood of caregiver behavior generalization to untrained materials and settings.

In the final treatment package, we added a visual schedule that included a structured sequence of daily activities. The mother delineated various activities into the proper order within the structured schedule each day across leisure activities, academics, and chores. She regularly varied the sequence and type of activity based on daily needs (e.g., daily vs. weekly chores, school day vs. school holiday). Over time, the mother continued to implement the treatment components in untrained contexts both within and outside of the daily therapy appointments. For example, during consultation at the beginning of an appointment, the mother shared that she implemented three-step prompting after requesting that John make his bed during his morning routine. Making his bed was not specifically trained during the daily sessions, but the mother’s behavior generalized to a task that naturally arose over the course of the day. Toward the end of the treatment evaluation, the mother’s behavior related to visual schedule implementation generalized to activities (e.g., visits to relatives, errands, doctor’s appointments) outside of his therapy sessions.

Telehealth provided access to a variety of relevant materials (common stimuli) as well as multiple exemplars of treatment-related relevant stimuli in naturally occurring contexts. As a result, practicing the treatment components across multiple meaningful exemplars likely increased the possibility of generalization. Moreover, extensive practice opportunities may have increased the mother’s fluency in the intervention within the family’s everyday settings. In a traditional clinic setting, therapists would likely start with representative examples of common or relevant stimuli but may be limited by the number and nature of available items. Therefore, providing treatment via telehealth allowed the treatment team to successfully train across sufficient and relevant examples, increasing the breadth of caregiver behavior generalization.

Train Loosely

Training loosely is a programming strategy that promotes flexibility and variability within training. Stokes and Baer (1977) defined this technique as when “. . . teaching is conducted with relatively little control over the stimuli presented and the correct responses allowed, to maximize sampling relevant dimensions for transfer to other situations and other forms of the behavior” (p. 357). Training loosely contrasts with many traditional “controlled” teaching strategies often employed when initially evaluating the efficacy of an intervention. Stokes and Baer noted that rigid teaching could potentially restrict generalization given the tight control over certain variables. On the other hand, by training loosely, it may promote flexibility within teaching, so that transfer to other similar situations and behaviors may occur more efficiently and effectively.

Rose and Beaulieu (2019) trained caregivers to implement a treatment package for two children diagnosed with autism spectrum disorder. The authors exemplified “training loosely” as caregivers were provided unspecific instructions to play with their children. For example, parents could play with their children for varying intervals and then either provide choices of activities or ask the child what they wanted to play with next. Caregivers could then allow the children to play with a toy, but sometimes they would deny access to it. In general, caregivers were instructed to do what felt “natural” to them to approximate their typical daily interactions. Despite these “loose” guidelines, the intervention resulted in decreased rates of challenging behavior and an increase in FCRs and appropriate waiting for both children. In addition, the parents reported high social validity and acceptance of treatment procedures. As such, providing flexible guidelines to parents may provide a successful model to promote the generalization of caregiver behavior.

In the current case, once initial treatment stability was met with the original demand and leisure materials, therapists allowed the mother to select any activity that fit within a general category of activities (i.e., leisure, academics, chores), and she could arrange them in a visual schedule as she preferred. In addition, the behavior team removed explicit instructions for which activities to complete, and the duration of each multiple schedule component (not to exceed a 15 min S∆). The mother eventually implemented treatment-consistent contingencies with John for both the “caregiver’s way” and “John’s way” with different activities and on variable schedules, both within and outside of the treatment sessions prior to the family’s transition into follow-up appointments.

Telehealth provided unique opportunities for therapists to “train loosely.” Therapists could remotely “follow” the family in the home to observe and collect data while limiting client reactivity since the therapists were not physically present. Sessions were streamed through the mother’s cell phone so that when moving locations, the treatment team could still observe and hear the parent and child interactions. Once they arrived at a particular location within the home, the mother would set the phone nearby so that both she and John were in view. Therapists turned off their cameras so that John could not see them, and they communicated with the mother directly via wireless headphones so that John could not hear them. These adaptations permitted the family to engage in more natural activities within the home to promote opportunities for treatment generalization. As the mother became more proficient with implementing the treatment and naturally encountered various situations in the home, she would then incorporate them into the daily schedule. For example, the mother might arrange the visual schedule to include leisure, homework, and chores. Naturally, as it became time for lunch, she decided she wanted to have John help prepare his meal. She could then immediately add this novel activity to the daily schedule and extend the treatment components to this new context. In a typical outpatient setting, therapists are often limited by the types of activities that families can practice, which may compromise their opportunities to train loosely. This limitation may result in loose application of the behavior plan “posttreatment”, once families are away from the clinic, since loose training was not practiced in clinic. If generalization is unsuccessful at this time, there may be less opportunities and resources to address implementation problems.

Mediate Generalization and Use Indiscriminable Contingencie s

Mediating generalization includes, “. . . establishing a response as part of the new learning that is likely to be utilized in other problems as well and will constitute sufficient commonality between the original learning and the new problem to result in generalization” (Stokes & Baer, 1977, p. 361). Allen and Warzak (2000) suggested that the generalization of caregiver behavior may occur through the control exerted by rules. That is, treatment-related rules that produce treatment plan consistent rule-following behavior on behalf of a caregiver might transfer from the clinic or training context into the natural environment. Framing behavioral treatment plan components as treatment-related rules is one potential method for mediating generalization.

One example of framing treatment components in the form of rules to mediate generalization involved teaching the mother to use three-step prompting (i.e., instruction, model, physical guidance; Drifke et al., 2017) whenever she prompted John to complete a task demand. Historically, the mother reported repeating her directions multiple times if John did not respond. During the treatment evaluation, we demonstrated that John completed task demands most often after he initially experienced a sequence of verbal prompts followed by a modeled prompt, and then physical guidance. Therefore, the therapists taught the mother to prompt him to follow directions using three-step prompting each time she delivered a task demand. During treatment, the therapists provided the mother with a treatment-related rule related to prompting task demands. The rule was, “If you ask John to complete a task, always follow the prompting sequence.” We observed high procedural fidelity with the prompting procedure and hypothesized that the mother’s rule-following behavior was initially maintained by therapist praise. Over time, as previously described, the mother’s rule-following behavior was likely reinforced by John’s increases in direction following and decreases in challenging behavior. Generalization was mediated when the mother reported implementing the three-step prompting procedure with novel tasks.

Because rule-governed behavior may make behaviors less sensitive to changing contingencies (Rosenfarb et al., 1992), establishing treatment-related rules has the potential to help mediate generalization when the availability of reinforcement contingencies for rule-following behavior is ambiguous, or unsignaled. For example, in the early stages of treatment when the mother introduced a new demand, John periodically ran out of the room, sometimes demonstrated challenging behavior, and at other times followed instructions (i.e., John’s responses to tasks were indiscriminable from the mother’s perspective). Stokes and Baer (1977) referred to indiscriminable contingencies as the “. . . impossibility of discriminating reinforcement occasions from nonreinforcement occasions until after the fact” (p. 358). The mother’s treatment-consistent behavior may have persisted, even in the presence of indiscriminable contingencies, if treatment-consistent rules exerted control over her behavior.

We hypothesized that the mother’s strong rule-following repertoire helped maintain procedural fidelity even in the absence of immediate reinforcement (i.e., John’s responding to directions), and in the presence of punishment (i.e., John’s noncompliance and challenging behavior). Therefore, from a conceptual standpoint, mediating generalization with treatment-related rules has the possible two-fold benefit of transferring control across novel stimulus contexts and making the caregiver’s implementation more resistant to occasional punishment (see Skinner, 1969). Telehealth provided opportunities for developing and testing the durability of treatment-related rules in an environment with more natural responses from John and his mother, as well as uncontrolled changes in contingencies of reinforcement for treatment-related rule-following.

Summary and Conclusions

This discussion article described a successful example of using telehealth to actively program and test for generalization based on the guidelines discussed by Stokes and Baer (1977). What is unique about our study is that the focus was largely aimed at programming for the generalization of the mother’s behavior, which directly influenced the success of the intervention. Treatment outcomes were maintained over time, and treatment monitoring was faded to occasional follow-up check-ins and consultations with the caregiver. Ultimately, the mother maintained high levels of procedural fidelity across all phases of treatment, and she reported high satisfaction and acceptability of the treatment outcomes.

The mother initially learned to implement the treatment components under more controlled contexts within the home, allowing for systematic manipulation of relevant variables over the course of the assessment and treatment process. As a result, it is likely that training the caregiver on the relevant treatment components in the home may have helped to promote generalization beyond what is achieved within a typical clinical setting. However, it can be difficult to determine the exact variables that were responsible for the successful generalization of the mother’s treatment implementation. Although the treatment team attempted to systematically manipulate isolated variables to verify the effects on treatment and generalization, it seems reasonable that a cumulative number of factors contributed to the robust treatment effects. In addition, the mother also participated in the treatment development throughout all stages by providing suggestions on various aspects of the intervention. This collaboration increased the likelihood of promoting naturally occurring contingencies and incorporating common, relevant stimuli.

Given the setting and context in which the treatment was conducted, telehealth provided opportunities to include relevant materials and activities from the start of treatment. It is possible that we eliminated the extra steps of incorporating relevant materials at a later stage of treatment when generalization would more typically occur. Likewise, we were in a better position to  program for generalization early and adapt the treatment components based on barriers or other concerns identified moment-to-moment within the home. Completing sessions in a typical outpatient setting often increases some of the difficulty of programming for generalization given the lack of accessibility to sufficient common stimuli, naturally maintaining contingencies, and other relevant schedules and activities. Thus, telehealth may have provided a more representative sampling of how John was likely to respond under natural conditions. Aside from the generalization benefits, telehealth is also a more economical choice, compared to in-home therapy options, when considering therapist reimbursement for travel time and mileage (Lindgren et al., 2016), as well as the logistics of having multiple or distant appointments to travel to in the same day. Nonetheless, an astute practitioner with the necessary resources at their disposal (support staff and access to the community) can certainly replicate the process described within the current discussion without the use of telehealth. Therefore, we suggest that practitioners work with families and institutions to establish procedures for regularly guiding caregivers through behavioral treatment plan implementation in the clinic, home, and community contexts. Although these practices were already established within our clinical programs, the pandemic temporarily limited our availability to access generalization contexts. Telehealth prompted a more deliberate examination and refinement of these strategies so that they can be systematically programmed during telehealth or in-person service delivery.

Finally, it is important to note that although we actively programmed for generalization throughout all sessions, we did not conduct true generalization tests due to the absence of baseline data for John’s or his mother’s behavior in most contexts. The main purpose of John’s admission was to develop a treatment plan that would successfully reduce challenging behavior and improve appropriate behaviors. Thus, providing this clinical service was the main priority for the family. Due to the rapid development of a successful intervention, however, we had many opportunities to expand the treatment into more relevant situations. It is possible that John and his mother may have already possessed some of the relevant skills before starting the intervention. However, John’s responses during the FA, his mother’s self-reports of her typical behavior before treatment, some brief observations of John and his mother during semi-structured direct observations and her lack of familiarity with behavior analytic assessment and treatment make it unlikely that the treatment-related behaviors were part of their repertoires.

During quarterly appointments, the mother maintained high procedural fidelity (> 90% accuracy), suggesting that she continues to implement the treatment between follow-up appointments. John continues to maintain low rates of challenging behavior and an increase in adaptive and functionallly equivalent alternative behavior. Based in part on the sustained behavioral improvements, John was admitted to a day-school program focusing on work skills and community interactions. His mother’s reports indicate that he continues to work successfully in the program. Future research should continue to explore the advantages of telehealth to determine the potential positive impact on treatment and generalization outcomes for both caregivers and individuals with disabilities.