Using a Lag Schedule of Reinforcement to Increase Response Variability in Children With Autism Spectrum Disorders

Abstract

Lag reinforcement schedules have been shown in previous research to be an effective intervention for teaching verbal and nonverbal response variability to individuals with developmental disabilities. In more recent research, variability itself has been considered a reinforceable behavior in its own right (Susa & Schlinger, The Analysis of Verbal Behavior, 18, 125–130, 2012). Lag x schedules of reinforcement can be used to teach variability by using contingencies that require responses to differ from previous responses. The present study extended Susa and Schlinger’s, The Analysis of Verbal Behavior, 18, 125–130, (2012) research by using 3 social questions instead of 1 in a random rotation and included probes to test for generality. A changing-criterion design was used to evaluate the results with one 11-year-old female participant diagnosed with autism. During baseline, the participant provided little variability, with rote responses. During the Lag 1 and Lag 2 phases, appropriate variable verbal responding increased with the use of echoic prompts, visual aids, and an error correction procedure. Further, the results also showed that the participant learned to vary her responses by demonstrating the ability to emit 11 novel prompted responses and 13 spontaneous responses.

In addition, the participant was able to retain the skills learned in a maintenance probe conducted 4 weeks postintervention.

The American Psychiatric Association (APA, 2013) states that individuals diagnosed with autism spectrum disorders (ASD) frequently engage in stereotypic behavior that results in rote responding to stimuli and are more likely to experience deficits in the areas of socialization and communication. Behavioral and lingual variability is a fundamental aspect of daily social interactions. The frequent use of rote and invariant responses by an individual may have a negative impact on the quality of that person’s social interactions. Opportunities to converse with others may be greatly reduced if repetitive language is used with little or no variability. For instance, a child with ASD is approached on a playground and asked, “What do you want to play with?” If the child’s only response is “I like to play with cars,” the behavior of the individual who initiated the social interaction may not be reinforced during this conversation, and he or she may be less likely to approach that child in the future. Jones (1990) suggested that unvaried responding to social questions might not only reduce opportunities for social interactions but also stigmatize the individual in the social community. Additionally, the absence of varied responding may interfere with the acquisition of novel responses and may prohibit problem solving and creativity (Jones, 1990; Neuringer, 2004). Individuals diagnosed with ASD do not typically have the natural sequence of relationship forming and are often unable to read social cues and body language or give eye contact during social interactions. Therefore, novel and varied responses can promote successful social interactions in everyday life.

Review of the Literature

Existing research has explored the use of a lag reinforcement schedule (i.e., Lag x) in promoting both verbal and nonverbal response variability in individuals with developmental disabilities. Cooper, Heron, and Heward (2007) stated that a Lag x is a schedule of reinforcement in which delivery of a reinforcer is contingent on a response being different in some specified way (e.g., different topography) from the previous response (e.g., Lag 1) or a specified number of previous responses (e.g., Lag 2 or more). In order to increase the number of opportunities for meaningful social interactions for individuals with ASD, variability across a variety of skills is essential. There is limited research promoting response diversity using lag schedules of reinforcement with different social skills and behaviors, such as play skills or vocal responses. Furthermore, the existing literature contains limitations that need to be addressed, such as varying the discriminative stimuli (S^Ds) and planning for generalization and maintenance.

Heldt and Schlinger (2012) conducted a study to test the effectiveness of a Lag 3 schedule of reinforcement on novel responses with two participants. The authors used one target question and visual stimuli to demonstrate the effectiveness of the intervention and included a follow-up probe 3 weeks after the Lag 3 schedule was withdrawn. Results indicated that variability in responding increased and was maintained after the withdrawal. Beyond the teaching of a skill, the follow-up probe also demonstrated the retention of increased response variability.

Variability in verbal behavior using a Lag x schedule of reinforcement was also evaluated by Lee, McComas, and Jawor (2002). Using three participants diagnosed with ASD, a Lag 1 schedule of differential reinforcement of alternative behavior was implemented (i.e., Lag 1/DRA). Intervention phases consisted of a paired-choice preference assessment, discrete-trial training (DTT), and a token economy system to reinforce appropriate varied responses to the social questions “What do you like to do?” and “How are you?” Generalization probes were conducted across three settings throughout the study for two of the three participants. The results indicated that the Lag 1/DRA was effective in teaching varied and appropriate verbal responding for two of the three participants. Although this study demonstrated the importance of using generalization probes across settings, the limited variety of social questions and the lack of generalization across people remain important limitations. Prior to this study, no other research had been conducted on the effectiveness of using lag schedules of reinforcement in teaching response variability.

Susa and Schlinger (2012) replicated and extended Lee et al.’s (2002) study by increasing the number of responses required by the lag schedule with one participant diagnosed with ASD. During the intervention phase, the therapist corrected incorrect responses by providing echoic prompts, and once the correct response was emitted, reinforcement was delivered to the participant. The results indicated that implementing a Lag 3 reinforcement schedule in conjunction with echoic prompts was successful in promoting topographically different and still relevant responses. Novel responses (e.g., “I’m good,” “I’m OK,” or “I’m super”) were learned and varied according to the reinforcement schedule. This study suggested that the use of a lag schedule of reinforcement promoted language that closely approximated typical conversational skills.

Many children with ASD do not have the appropriate language to answer social questions with sufficient variability. Evaluations of Lag x schedules used to promote variability have been conducted in a wide range of studies, some of which involved nonverbal behaviors. These studies included variability in selecting activities (Cammilleri & Hanley, 2005), enhancing varied responding in block building (Napolitano, Smith, Zarcone, Goodkin, & McAdam, 2010), variability in script fading (Lee & Sturmey, 2014), and promoting variability of toy play (Baruni, Rapp, Lipe, & Novotny, 2014).

There has been limited research available on the use of lag schedules of reinforcement for promoting variability in verbal behavior. To demonstrate diversity, the use of systematic teaching methods is necessary to address the deficit of rote responding in individuals diagnosed with ASD. Varying the S^Ds and the order in which they are presented has not been demonstrated in any research to date.

The purpose of the present study is to extend Susa and Schlinger’s (2012) study by implementing a Lag 2 schedule of reinforcement to promote variability by preventing rote responding from coming under stimulus control of a single social question. In addition, the current study included multiple S^Ds using a variation of three target social questions that were asked in a random order to promote language for conversational skills. The current study incorporated Lee et al.’s (2002) research by including generalization probes across settings and people and using three social questions instead of one. Last, the current study included a maintenance probe postintervention to assess long-term effectiveness.

Method

Participant

Belle was an 11-year-old girl diagnosed with ASD. She was recruited by a referral from the faculty supervisor’s announcements about the project via e-mail. Belle was fully verbal and able to spontaneously request (i.e., mand) for activities, information, and attention. She often utilized intraverbals in social situations, although she did engage in stereotypic vocalizations and often scripted scenes from preferred movies, shows, and commercials. Belle already had intraverbal responses to various social questions in her repertoire, though she often utilized rote and scripted phrases. Sometimes her responses were fixed and static (e.g., when asked, “What do you like to do?” she often responded with “flying an airplane”), whereas on other occasions, Belle’s responses were scripted (e.g., when asked, “What do you do for fun?” she would reply “the white rabbit is late”).

Setting

The training sessions took place in Belle’s home, in either her bedroom or the living room. Sessions varied in duration and always consisted of at least three teaching trials, separated by a 5-min intertrial interval. The duration of sessions was variable due to Belle’s frequent engagement in scripting behavior and high rates of noncompliance. Sessions were conducted between 2 and 3 days a week. The maintenance probe session was conducted 4 weeks after the participant met mastery criterion. The study consisted of a total of 19 sessions over the course of 10 weeks.

Preference Assessment

Prior to the implementation of the intervention, Belle’s parent identified several items and activities that could potentially serve as reinforcers using a questionnaire. After preferred items and activities were identified, a multiple-stimulus with-replacement (MSW) preference assessment was administered. When conducting the MSW, an array of stimuli was presented to the participant and the participant was instructed to choose an item. The items were ranked on how often they were chosen, with the highest ranked items likely to be the most preferred (Ciccone, Graff, & Ahearn, 2006). Belle showed the highest preference for her iPad, a break in her bedroom, and a fidget spinner. At the start of each trial, the lead therapist offered Belle a pretask choice by presenting pictures of her most highly preferred items. The item she chose was the identified reinforcer for that trial; however, if Belle wanted a different item or activity when she earned reinforcement, her request was honored. Belle had access to reinforcement for an average of 3.5 min. Allowing Belle to choose (and change) the item she was working for increased the likelihood of Belle contacting reinforcement. When a correct response was emitted (i.e., a response that correctly met the current lag criteria), the preferred item was presented to Belle and paired with social praise (e.g., “Great job!”).

Definitions and Measurement of Target Behaviors

The dependent variable was the percentage of correct trials in DTT during intervention sessions. The independent variable was the use of a lag reinforcement schedule. A response was considered correct if it topographically varied from previous responses. Appropriate responses to questions were determined based on interviews with her parent regarding activities that Belle historically engaged in and that were a part of her typical routine. However, a response that was grammatically correct but noncontextual (e.g., “I like to fly airplanes”) was considered incorrect. Likewise, a 5-s delay with no response after the presentation of a question was also considered incorrect. A novel prompted response consisted of learned responses that were previously not a part of Belle’s repertoire. Spontaneous responses were defined as any appropriate responses emitted that were not taught by the therapist. A trial consisted of the lead therapist presenting the three social questions in a random order according to the lag criterion in effect. If Belle responded correctly, the next trial began; if she errored, the error correction procedure was utilized. A prompted trial was defined as the lead therapist providing a verbal model and/or gestural cue to indicate the correct answer for Belle to emit. Trials were indicated by the set of questions presented according to the lag schedule in effect (i.e., three questions being presented in the Lag 2 criteria were counted as one trial). Mastery criterion was achieved when the participant demonstrated 100% correct responding according to the lag criteria in effect across two sessions. Interobserver agreement (IOA) percentages were calculated by computing the number of agreements between the independent observers and then dividing by the total number of agreements plus disagreements, and then multiplying by 100. IOA was taken by the lead therapist for 33% of each session; mean IOA was 93% (range 83%–100%). Treatment integrity was collected for 34% of all sessions and averaged 99% (range 90%–100%).

Procedure

Lag schedule procedure

A changing-criterion design was implemented to evaluate the different lag schedules with interspersed generalization probes and a maintenance probe. “Names in a Hat” is a randomizer application that generates lists in random order (Peters, 2018). This application was used throughout the study to determine the order in which the three social questions would be presented.

To address the complexity of the skills required for Belle to learn the terminal behavior, six sessions of practice trials were run to establish the prerequisite skills of attending and answering social questions. From Session 1 to Session 6, Belle required 32, 13, 6, 17, 14, and 8 trials, respectively, to establish these skills. Frequency data were taken on all responses, including those that were correct, incorrect, or prompted, within baseline, practice sessions, the intervention phases, and the maintenance probe.

Baseline

In the baseline phase, the lead therapist asked the participant each social question three times in a random rotation. The social questions used were “What do you like to do?” “What do you do for fun?” and “What is one of your favorite activities?” The therapist documented if the responses were rote, invariable responses. Two responses that differed by one minor word (e.g., the) were considered to be a rote invariable response (e.g., “I like trucks,” and “I like the trucks”). A neutral statement from the therapist (e.g., “nice,” “cool”) was given for any response given by the participant. If the participant did not respond within 5-s, the therapist presented the next question after a 30-s intertrial interval.

Lag 1 condition

The Lag 1 condition was identical to the baseline phase except that reinforcement was delivered contingent on two topographically different responses. For example, if the randomizer application dictated the use of the social questions “What do you like to do?” and “What is one of your favorite activities?” and Belle responded correctly, reinforcement was delivered with a 30-s intertrial interval. If Belle responded incorrectly to the first question, the therapist would implement the error correction procedure. This consisted of the therapist holding up a picture of a preferred activity and providing a verbal model (e.g., the therapist showed a picture of Belle riding her bike and stated, “I like to ride my bike”). After Belle repeated the verbal model, the therapist implemented a prompted trial. The prompted trial consisted of the therapist restating the same question (e.g., “What do you like to do?”) while holding up the same visual stimulus without giving the verbal model (e.g., the therapist showed a picture of Belle riding her bike but did not give a verbal prompt). After the participant responded correctly in the prompted trial, the therapist presented the same question (e.g., “What do you like to do?”) without the visual stimulus and allowed the participant to respond independently. If the participant responded correctly, the next question dictated by “Names in a Hat” was presented. If the participant responded incorrectly, the error correction and prompted-trial procedures were repeated.

If Belle responded incorrectly to the second social question (e.g., “What is one of your favorite activities?”), the same error correction procedure was utilized until she emitted a correct response independently. Once an independent correct response was emitted, the therapist re-presented the first question (e.g., “What do you like to do?”) and then the second question (e.g., “What is one of your favorite activities?”). For example, if the therapist presented the first question “What do you like to do?” and Belle responded with “I like to ride my bike,” the therapist would then present the second question. If Belle responded to the second question incorrectly, the error correction procedure was implemented. The therapist would hold up a picture of a different preferred activity and provide a verbal model (e.g., the therapist showed a picture of Belle jumping on a trampoline and stated, “I like to jump on the trampoline”). After the participant repeated the verbal model, the therapist implemented a prompted trial. The prompted trial consisted of the therapist restating the same question (e.g., “What is one of your favorite activities?”) while holding up the same visual stimulus without giving the verbal model (e.g., the therapist showed a picture of Belle jumping on the trampoline but did not give a verbal prompt). After the participant responded correctly in the prompted trial, the therapist re-presented the same question (e.g., “What is one of your favorite activities?”) without the visual stimulus and allowed the participant to respond independently. If the participant responded correctly, the therapist re-presented the first question, and if Belle emitted a correct response, the second question was then presented. If she responded correctly to both questions, this was considered one correct trial, and reinforcement was delivered, followed by a 30-s intertrial interval.

Lag 2 condition

The Lag 2 condition was identical to the Lag 1 condition except that reinforcement was provided for three topographically different responses. Additionally, if the participant’s answer to the second or third question did not vary from the previous answers, the therapist implemented the error correction procedure. For example, if the randomizer application dictated the use of the social questions “What do you like to do?” “What is one of your favorite activities?” and then “What do you do for fun?” and Belle responded correctly, reinforcement was delivered with a 30-s intertrial interval. If Belle responded incorrectly to the first or second question, the therapist would implement the error correction procedures explained in the Lag 1 condition.

If Belle incorrectly responded to the third question (e.g., “What do you do for fun?”), the error correction procedure was implemented. For example, if the therapist presented the first question “What do you like to do?” and Belle responded with “I like to ride my bike,” the therapist would then present the second question. If Belle responded correctly to the second question “What is one of your favorite activities?” by stating “I like to jump on the trampoline,” the therapist would then present the third question. If Belle responded to the third question (e.g., “What do you do for fun?”) incorrectly, the error correction procedure was implemented. The therapist would hold up a picture of a different preferred activity and provide a verbal model (e.g., the therapist showed a picture of Belle playing on her iPad and stated, “I like to play on my iPad”). After the participant repeated the verbal model, the therapist implemented a prompted trial. The prompted trial consisted of the therapist restating the same question (e.g., “What do you do for fun?”) while holding up the same visual stimulus without giving the verbal model (e.g., the therapist showed a picture of Belle playing on her iPad but did not give a verbal prompt). After the participant responded correctly in the prompted trial, the therapist re-presented the same question (e.g., “What do you do for fun?”) without the visual stimulus and allowed the participant to respond independently. If the participant responded correctly, the therapist re-presented the first and second questions, and if Belle emitted correct responses, the third question was then presented. If she responded correctly to all three questions, this was considered one correct trial, and reinforcement was delivered, followed by a 30-s intertrial interval. During the Lag 2 condition, if Belle errored on any of the three questions, the error correction procedure was implemented and the subsequent questions were presented.

Maintenance and generalization

To maintain response variability across therapists and generalize to novel people, generalization probes across settings and people were interspersed throughout the study. Once the mastery criterion was met during the intervention phases, the assisting therapist and participant’s mother asked Belle the target questions in other areas of the home. A maintenance probe was conducted 4 weeks after withdrawal of the intervention to determine if the participant maintained varied responding.

Results

Figure 1 shows the percentage of variable verbal responding throughout each lag criterion, including baseline and maintenance phases, for Belle. Baseline results showed that Belle did not respond appropriately to the social questions asked (i.e., she provided rote, invariable, scripted, and grammatically incorrect responses or did not respond at all).

Fig. 1.

Percentage of correct responding across conditions. Numbers above the data points indicate the number of trials conducted in that session

During the Lag 1 condition, appropriate responding ranged from 23% to 100% with a mean of 66% over the course of 12 sessions. For the first six sessions of this phase, there was highly variable responding with no observable trend (range 23%–70%). However, there was an increasing trend identified in the last six sessions of this phase, where appropriate responding ranged from 64% to 100%. In the final two sessions, Belle demonstrated 100% correct responding.

During the Lag 2 condition, appropriate responding ranged from 73% to 100% with a mean of 86% over the course of five sessions. The data for the first three sessions were slightly variable and ranged from 73% to 81%. In the final two sessions, Belle demonstrated 100% correct responding.

Throughout the study, Belle emitted 11 novel prompted responses, as well as 13 spontaneous responses. Novel prompted responses in the form of “I like to . . .” included “go to Girl Scouts,” “play with my dog,” “play with my sister,” “go to Chuck E. Cheese,” “read books,” “roast marshmallows,” “play with my iPad,” “jump on the trampoline,” “swim in the pool,” “horseback ride,” and “go to dance.” Spontaneous responses in the form of “I like to . . .” included “play baseball,” “play basketball,” “go to Great Wolf Lodge,” “play with Mom,” “eat candy,” “eat cake,” “eat Pez,” “sing,” “dance,” “act silly,” “be kooky,” “play with my crystal ball,” and “play with my kitty.”

In the maintenance session conducted 4 weeks after the Lag 2 schedule was withdrawn, two of the three responses Belle emitted were spontaneous responses. Belle demonstrated 100% correct responding during the maintenance probe.

The results of this study indicated that variability in responses increased by using a lag schedule of reinforcement and were maintained following a 4-week period after the Lag 2 schedule was no longer in effect.

Discussion

Results from this study support the use of lag schedules as an effective method of increasing response variability. The present study extended Susa and Schlinger’s (2012) research by demonstrating that the participant acquired more than two responses and varied between them depending on which lag criterion was in effect. Prior to implementing the intervention, Belle did not respond appropriately to social questions. Incorporating practice trials to teach the prerequisite skills necessary for answering social questions likely resulted in the higher rate of correct responding seen from baseline to the initial session of the Lag 1 condition (i.e., Session 2). Belle demonstrated the use of 11 novel prompted responses and 13 spontaneous responses. The majority of her spontaneous responses occurred during the Lag 2 condition between Sessions 15 and 18. Belle also displayed retention of the target skills upon the administration of the maintenance probe. Not only did she retain the skills, but two of the three responses were spontaneous and topographically different from any other response given during the intervention phases of the study. Many individuals diagnosed with ASD typically do not answer social questions with much, if any, variability. This study demonstrated that using a lag schedule of reinforcement was successful in teaching response variability to a child with ASD.

Limitations in previous studies included the possibility of response variability coming under the stimulus control of the experimenter or question(s) being asked. In the present study, these limitations were addressed by having more than one person ask the participant questions (i.e., two therapists and the participant’s mother) and by using three social questions rather than one. In doing so, the likelihood of variable verbal responding coming under the stimulus control of the experimenter or questions was reduced. Another limitation not accounted for in previous studies was the lack of maintenance probes to identify if variable verbal responding learned during treatment was maintained after the termination of the intervention. The present study addressed this by running a maintenance probe 4 weeks after the last treatment session, where mastery criterion for Lag 2 was demonstrated.

In the Lee et al. (2002) study referenced earlier, the authors considered one of their limitations to be that the participant was potentially prompted by having a visual display of preferred toys and activities in the training room. Although this occurred in the present study, this is not considered a limitation, as it is natural for individuals to read their environment. For example, if a neurotypical child at recess is asked by a peer, “What do you want to play with?” he or she will likely look around the environment (e.g., playground) and respond appropriately using the visual stimuli available to him or her.

One possible limitation in the present study was how the error correction procedure was implemented. All answers were required to be the same; the participant could not change answers after error correcting or prompting. For example, if the participant changed her answer from “I like to jump on the trampoline” to “I like to go to Girl Scouts” during the independent trial, that trial would be marked as incorrect, and the teaching procedure (i.e., error correction and prompted trials) would then be implemented.

Another possible limitation was the setting in which the trials took place. There were various interruptions during sessions, especially when working with the participant in her living room. The experimenter could not account for these extraneous variables, which included, but are not limited to, the dog barking, her sister talking in the kitchen, and her brother walking through the treatment setting, which were distracting to the participant.

The procedures for the teaching condition in this study were highly structured, which resulted in a high response effort for both the therapist and the participant. Future research may consider exploring a flexible approach to test the efficiency in teaching verbal variability in applied settings. By comparing two different teaching strategies, including a flexible approach (e.g., a no-no prompt or allowing participants to change their responses) versus a more structured prompting strategy (e.g., errorless teaching), researchers can evaluate the differences in the skill acquisition rates of the participants.

Future research could also evaluate the skill acquisition rate of teaching individuals to read their environment by assessing the visual stimuli around them. Typically, people are influenced by their environment; the presence of different visual stimuli may evoke certain behaviors under specific conditions (e.g., whispering in a library versus shouting on a playground). By taking a more naturalistic approach to social behavior, researchers can examine different methods of teaching variable verbal responding to individuals diagnosed with ASD. Finally, future studies could further refine the use of this reinforcement schedule and explore the relationship between the lag schedule of reinforcement and the success it has in alleviating stereotyped responding in individuals with ASD.

Compliance with Ethical Standards

Conflict of Interest

The author declares that she has no conflicts of interest.

Ethical Approval

All procedures performed in studies involving human participants were in accordance with ethical standards of the institutional and/or national research committee.

Informed Consent

Informed consent was obtained from the parent of the participant in this study.