Functional communication training (FCT) is a well-established, evidence-based practice for the reduction of challenging behavior in individuals with autism, intellectual disability, other health impairments, and multiple disabilities (Gerow, Davis, Radhakrishnan, Gregori, & Rivera, 2018; Kurtz, Boelter, Jarmolowicz, Chin, & Hagopian, 2011). There are multiple variations of FCT (Rooker, Jessel, Kurtz, & Hagopian, 2013), and best practice guidelines have described how to implement FCT effectively (Tiger, Hanley, & Bruzek, 2008). However, researchers have called for additional research to identify variables that practitioners can manipulate, such as the reinforcement of mand variability (Adami, Falcomata, Muething, & Hoffman, 2017; Falcomata et al., 2018), to mitigate the resurgence of challenging behavior during or after successful FCT treatment (e.g., Bloom & Lambert, 2015; Falcomata & Wacker, 2012; Heath, Ganz, Parker, Burke, & Ninci, 2015; Neely, Garcia, Bankston, & Green, 2018).

Operant variability can be conceptualized as “a continuum ranging from repetitive at one end to stochastic at the other” (Neuringer, 2002, p. 672) and can be reinforced using lag schedules (Page & Neuringer, 1985). Under a lag schedule, a response or response sequence is reinforced if it differs from N preceding responses or response sequences on one or more dimensions (e.g., topography or stimuli selected), with N equal to the value of the lag (e.g., Neuringer, 2002). Under a Lag 1 schedule, each response is reinforced when it differs from the preceding response. For example, under a Lag 1 schedule of positive reinforcement, Silbaugh and Falcomata (2016) presented a participant with a plate of multiple foods and, on each trial, provided the participant with 30-s access to high-preferred toys contingent on the participant selecting and consuming a food that differed from the food consumed in the preceding trial. The seminal applied lag-schedule study by Lee, McComas, and Jawor (2002) demonstrated that differential reinforcement of appropriate responding with a lag schedule increased variability in appropriate answers to social questions in individuals with autism. Researchers have since demonstrated the clinical and educational utility of lag schedules with and without prompts for increasing the variability or novelty in verbal and nonverbal behavior such as tacting (Heldt & Schlinger, 2012), responding to interview questions (O’Neill & Rehfeldt, 2014), mand frames (Brodhead, Higbee, Gerencser, & Akers, 2016), and play skills (e.g., Baruni, Rapp, Lipe, & Novotny, 2014).

Two studies evaluated the effects of FCT with lag schedules on mand variability and challenging behavior. Adami et al. (2017) demonstrated that FCT with a Lag 1 schedule of reinforcement replaced challenging behavior with elevated mand variability across manding modalities (e.g., tablet, picture exchange, microswitch) in adult and adolescent males with autism. In a partial replication by Falcomata et al. (2018), the same group subsequently demonstrated that as the value of the lag schedule increased from one to five across phases, mand variability generally increased and challenging behavior remained at near-zero rates. Additionally, in a 2-week follow-up for one boy, elevated mand variability maintained for multiple sessions as FCT was implemented with a Lag 0 schedule, suggesting FCT with lag schedules may produce mand variability that persists even when the variability requirement is removed. However, the effects of lag schedules on variability in functionally equivalent mand topographies have not been evaluated.

Accordingly, Silbaugh et al. (2018) evaluated the effects of a Lag 1 schedule of reinforcement with a progressive time delay (TD) on topographical vocal mand variability in two children with autism. The results demonstrated that lag schedules increased topographical vocal mand variability. In a follow-up study, Silbaugh and Falcomata (2018) replicated increased mand variability using similar procedures but across sign mand topographies in a boy with autism and limited vocal skills.

Although FCT with lag schedules has been shown to decrease challenging behavior when nonvocal mand variability is reinforced across modalities (e.g., Adami et al., 2017), whether challenging behavior can be reduced during FCT when topographical variability in vocal manding is reinforced was unknown. Therefore, in the current study we aimed not only to replicate prior research by reinforcing mand variability during FCT but also to extend prior research by evaluating, for the first time, the effects of FCT with lag schedules and prompt-fading procedures on challenging behavior and functionally equivalent vocal mand variability in children with autism.

Method

Participants, Settings, and Materials

Each participant had a diagnosis of autism and demonstrated (a) socially mediated challenging behavior as determined by a functional behavioral assessment, (b) a generalized echoic repertoire, (c) a vocal manding repertoire, and (d) at least one appropriate mand topography exhibited during the functional behavioral assessment that appeared to be part of the functional response class of challenging behavior targeted with FCT.

Zahid was a 5-year-old Pakistani boy who could follow basic one-step instructions; exhibited vocal mands, tacts, intraverbals, and echoics; and received services part time at an applied behavior analysis (ABA) clinic. Zahid’s therapist conducted the sessions in a 3 meters. room in his clinic with coaching from the experimenter, with the exception that the experimenter conducted the multielement functional analysis (FA) and duplic assessment. Materials included toys identified using free-operant stimulus preference assessments (FO-SPA; Roane, Vollmer, Ringdahl, & Marcus, 1998). No instructions were given to clinic staff to withhold access to the toys outside of sessions. The experimenter video recorded the sessions.

William was a 5-year-old Caucasian and Hispanic boy. He followed some one-step instructions and had a minimally vocal verbal repertoire consisting also of signs and selection-based speaker responses (e.g., card exchange or tablet-based augmentative alternative communication systems). He received home-based ABA services prior to the study and throughout the study. Materials included toys included in the preference assessment, as well as an electronic tablet used during the tangible condition of an FA. No materials were used in the treatment evaluation. The experimenter conducted all sessions in William’s living room. His mother video recorded the sessions.

Chris was a 4-year-old Caucasian boy who could follow basic one-step instructions and exhibited vocal mands, tacts, intraverbals, and echoics. He received home-based ABA services prior to the study and throughout the study attended a preschool program for children with disabilities. The experimenter conducted all sessions in Chris’s living room, sitting on a couch. Chris’s father video recorded the sessions and did not interact with him during the study. Materials included a video game on an electronic tablet and nonpreferred toys identified by indirect assessment. Chris’s father agreed to withhold access to the tablet for 30 min after the last session of each visit.

Paul was a 4-year-old Caucasian boy with secondary diagnoses of encephalopathy and attention-deficit/hyperactivity disorder. He attended a preschool program for children with disabilities. The experimenter conducted sessions in any room of Paul’s home in which he chose to play, so materials were freely varied across sessions. Paul’s mother participated in all sessions with guidance from the experimenter, including the video recording of sessions.

Response Definitions and Measurement

We measured challenging behavior during the FAs. We measured echoics during the duplic assessment. Topographical mand variability and challenging behavior were the primary dependent variables in the treatment evaluation. We measured dependent variables in the treatment evaluation only when the programmed establishing operation (EO) was present (i.e., responding was not measured during reinforcement). For each participant, we individualized the definition of a mand topography, and whether a mand topography was considered different, based on the participant’s behavior exhibited during the functional behavioral assessment, FA, or baseline reinforcement of challenging behavior in the first phase of the treatment evaluation.

We defined a mand topography as a complete sentence with three or more words for Zahid and Chris, a single word for William, and utterances consisting of one or more words for Paul. We defined independent variant mand topography as an independent vocal mand topography that differed from the last independent mand topography within the session. We defined an independent invariant mand topography as an independent vocal mand topography that was not different from the last independent mand topography within the session. We defined a prompted variant mand topography as a prompted mand topography that differed from the last independent mand topography within the session. Thus, in some sessions we delivered the same prompt across successive trials as long as the topography being prompted differed from the last mand topography that occurred independently. We counted instances of manding each session and converted the total count to responses per minute (RPM).

For Zahid, a mand topography was considered different if the sentence (a) did not contain the same words in comparison to the preceding independent sentence within a session (i.e., changes in verb tense or other similar changes to words were in sufficient), (b) consisted of English words (Zahid also spoke his family’s native language), and (c) appeared relevant to the reinforcer. For example, “Let me play” would be variant if it followed “Will you share?” but invariant if it followed “I’m still playing.” For William, a mand topography was different if the word was different from the last word he used independently. For example, if he said “wait” after saying “rest” across two consecutive trials within the session, “wait” was considered different and therefore was counted as an independent variant mand topography. However, the second instance of saying “wait” on two consecutive opportunities to request the reinforcer would be considered an independent invariant mand topography. For Chris, a mand topography was different if most of the sentence (i.e., two out of three, or four out of five words) differed from the last sentence used independently. The word “please” established a sentence as different only when it occurred in the sentence “my turn, please.” Otherwise, the word “please” did not establish a sentence as different. For example, if he said “Can I have it?” followed by “Please, can I have it?” the latter was considered invariant. Similarly, if he said “My turn, please” and then “My turn, please please please,” the latter was considered invariant and repeated “please” instances were not recorded in the transcript. One- and two-word utterances were not counted for Chris. Also, if the first sentence was “I want it” and the second sentence was “Can I have it?” the second sentence was considered variant, despite the common word “it.” For Paul, a mand topography was considered different if at least one word differed from the last word or words he used when subtracting common words from the current utterance and its preceding utterance. For example, if he said “No” and then said “No, I’m busy” and then “I’m still busy,” the second and third utterances would be considered different and therefore variant. If Paul’s fourth response was “I’m busy,” it would be considered invariant.

We did not measure the first independent mand topography of each session of the treatment evaluation for the purpose of calculating variant or invariant manding rates. We did, however, measure the first independent mand topography of each session to calculate the overall manding rate and measure variability of the second mand topography that occurred within the session. To evaluate the effects of the independent variable on variant responding independent of language-related constraints, with the exception of Zahid, the topography did not need to be one commonly used to request the reinforcer by typically developing children. For example, when the lag schedule was in effect during the treatment evaluation, if instead of saying “wait,” the participant said “water,” the response would have been reinforced.

We defined challenging behavior individually for each participant. For Zahid, it was defined as spitting, scratching, touching the therapist with his hands or feet, pushing, mouthing, hitting, biting, putting an arm around the therapist’s shoulders or neck, pulling the therapist’s clothes, pushing, kicking or tipping furniture over, or hitting the toys in the therapist’s hands. Nonexamples were dropping toys over the room partition and touching the therapist when Zahid’s body was turned away from the experimenter. Each instance was counted when physical contact discontinued (e.g., hand removed) or discrete responses (e.g., hitting furniture) were separated by 2 s. For William, it was defined as crying and falling to the floor, actual or attempted elopement by walking five or more steps away from the adult’s reach, or aggression (pushing the adult with his hands, pushing or throwing himself onto the adult, kicking the adult, and actual or attempted biting of the adult). Crying or whining was not counted if it did not co-occur with falling to the floor. For Chris, it was defined as including either whimpering consisting of nonwords, kicking his feet rapidly, tensing his torso as indicated by bending forward while shaking, tensing his hands into a cramped or fist-like position or flapping his hands, stomping his feet, hitting furniture with his hand, self-injury (hitting his tailbone or buttocks on the floor, or banging his head on the floor), or elopement (running to another room, consisting of at least five steps away from the adult). For Paul, it was defined as actual or attempted physical contact without permission (e.g., kicking, hitting, headbutting, pushing, pulling hair, swinging his arms at or rolling into the adult), hitting himself on the head or face with his hand or fist, dropping to the floor, stomping feet, or pushing or throwing furniture or tangibles not designed to be thrown (e.g., chair, toy lighthouse). Instances of challenging behavior separated by 2 s without the behavior were counted, and the total count per session was converted to RPM for each dependent variable.

Interobserver Agreement

A trained observer independently viewed videos and coded data during 33% of sessions randomly selected across all phases of the treatment evaluation to assess interobserver agreement (IOA) using the exact count-per-interval method (Cooper, Heron, & Heward, 2007). Each session was divided into 10-s intervals, and the number of intervals in which both observers recorded the same number of responses was totaled and divided by the total number of intervals, and the quotient was converted to a percentage. Mean IOA was calculated by summing session IOAs and dividing the total by the number of sessions. For Zahid, mean IOA was 96% (range 88%–100%) for challenging behavior, 98% (range 88%–100%) for independent variant manding, 98% (range 81%–100%) for independent invariant manding, and 99% (range 94%–100%) for prompted variant manding. For William, mean IOA was 98% (range 91%–100%) for independent variant manding, 98% (range 94%–100%) for independent invariant manding, 96% (range 81%–100%) for prompted variant manding, and 93% (range 87%–100%) for challenging behavior. For Chris, mean IOA was 96% (range 74%–100%) for independent variant manding, 93% (range 84%–100%) for independent invariant manding, 98% (range 84%–100%) for prompted variant manding, and 95% (range 65%–100%) for challenging behavior. For Paul, mean IOA was 93% (range 77%–100%) for independent variant manding, 95% (range 87%–100%) for independent invariant manding, 99% (range 90%–100%) for prompted variant manding, and 97% (range 84%–100%) for challenging behavior.

Functional Analysis

We conducted a brief functional behavioral assessment before each FA. An FA (e.g., Iwata, Dorsey, Slifer, Bauman, & Richman, 1982/1994) was conducted to identify function(s) of challenging behavior to be incorporated into the treatment evaluation for each participant. Each FA continued until (a) differentiated data indicated one or more functions of challenging behavior based on at least three data points collected in each test and control condition, (b) the experimenter exhausted a reasonable number of strategies to enhance differentiation, or (c) time constraints did not permit continuing the assessment.

Zahid

We conducted a pairwise FA (Iwata, Duncan, Zarcone, Lerman, & Shore, 1994) across three clinic visits to test the hypothesis that challenging behavior was maintained by socially mediated positive reinforcement in the form of access to tangibles. Sessions were 5 min, test and control conditions contained preferred toys, and the order of test and control conditions was randomized. With the exception of one session in which the experimenter demonstrated the procedures for the therapist during some trials, Zahid’s therapist ran all sessions with coaching from the experimenter. An FO-SPA was conducted to identify preferred toys before the first session on Days 1 and 2 of the FA. When the control condition was initiated, the therapist told Zahid, “It’s playtime,” and provided him with free access to the same toys used in the test condition. The therapist provided attention on a fixed-time 10-s schedule, omitted demands, and ignored challenging behavior. Prior to the test condition, the therapist provided Zahid with 1 min of free access to toys. Throughout the session, the therapist removed access to the toys using errorless physical prompts and waited for challenging behavior to occur. Contingent access to toys was provided for 30 s for every instance of challenging behavior. The therapist did not provide attention or deliver demands. Starting with Session 6, the therapist also said “Give me toys” when initiating all subsequent trials in an attempt to minimize relinquishing issues.

William and Chris

We conducted a multielement design FA with a control condition and four test conditions to identify the function(s) of challenging behavior. Test and control conditions were 5 min in duration. The experimenter wore a discriminative stimulus (plain clothes, or an orange, blue, green, or purple shirt) associated with each condition. During the control condition, the experimenter did not present demands. He gave the participant free access to an electronic tablet, delivered 2–3 s of attention on a fixed-time 15-s schedule, and ignored challenging behavior. During the ignore condition, the experimenter did not make toys or activities available, and he withheld attention for the duration of the session. Prior to the tangible condition, the experimenter gave the participant free access to the tablet for 1 min and then started the session. Throughout the session, the experimenter said “My turn,” obtained the tablet using least-to-most (LTM) prompting, and provided 30-s access to the tablet contingent on each instance of challenging behavior. For William, following the last tangible condition session of a home visit, he was told “I’m going to put the tablet away,” and it was hidden out of sight for at least 30 min. During the attention condition, the tablet was placed out of sight. The participant was provided with attention and 1-min free access to low-preferred toys identified by a paired-stimulus preference assessment (Fisher et al., 1992) for William, and parent report for Chris. Then, the experimenter said that he was busy but that the participant could play with the toys. The experimenter then withdrew attention and only delivered 30 s of attention contingent on each instance of challenging behavior. During the demand condition, the experimenter did not present the participant with toys or activities. For William, demands consisted of one-step instructions related to transitions such as “It’s time to work,” “Come sit down” and “Come here,” and “Stand up.” For Chris, demands consisted of one-step instructions related to familiar household routines, such as assisting with laundry or throwing trash in the trash can. The experimenter presented demands at 2-s ITI using an LTM prompting hierarchy (i.e., verbal, gestural/model, physical) and praised all instances of compliance. Escape was provided in the form of task termination for 30 s contingent on each instance of challenging behavior.

Paul

The results of a records review and the direct assessment steps of the functional behavioral assessment suggested that Paul’s challenging behavior was maintained by termination of activity interruptions (e.g., Fisher, Adelenis, Thompson, Worsdell, & Zarcone, 1998). Accordingly, Paul’s FA utilized a pairwise design (Iwata, Duncan et al., 1994) with a control condition and an interrupt condition. Sessions were 5 min in duration. A discriminative stimulus (yellow shirt) worn by the experimenter was associated with the interrupt condition. Common procedures included the following: (a) sessions were conducted in any room of his home that Paul wanted to play in, (b) Paul’s mother was available to implement all aspects of the protocol except for interruptions, (c) Paul was permitted to bring any toys or foods into sessions, and (d) sessions began after 1 min without challenging behavior. During the control condition, Paul was not interrupted, minor attention (e.g., “OK”) was given if he commented on or described his play, his mother and the experimenter reinforced all mands on a continuous reinforcement schedule, and every 30 s the experimenter issued a brief reminder that he was available if Paul needed anything. If the reinforcer for a mand was unsafe or impossible to provide, the experimenter made a best attempt to provide a substitute reinforcer corresponding to the mand (e.g., Hagopian, Bruzek, Bowman, & Jennett, 2007). For example, on multiple occasions, Paul asked the experimenter to go to his car and drive home, and the experimenter acted and made sound effects as if he were actually doing so. If challenging behavior occurred, interactions with Paul continued as if the challenging behavior did not occur (i.e., no differential consequences for the behavior). During the interrupt condition, the experimenter initiated trials by interrupting Paul’s ongoing activity and verbally prompting a transition such as “OK, let’s go do something else” or “That’s enough Pirate’s Booty; let’s do some work” or “Put that down and sit at the table.” The experimenter used LTM prompting to terminate Paul’s access to tangibles. “Do” and “Don’t” interruptions were not systematically controlled for as in prior interrupt FAs (e.g., Fisher et al., 1998; Hagopian et al., 2007, Study 2). During interruptions, mands were placed on extinction, and contingent on challenging behavior, the experimenter immediately terminated the interruption while saying something such as “OK, you don’t have to.” For a 30-s reinforcement interval, mands were reinforced on a continuous schedule. Any challenging behavior that occurred during the reinforcement interval was ignored.

The results of the FAs suggested that challenging behavior was maintained by access to tangibles for Zahid and Chris, multiply maintained by escape and access to tangibles for William (only the escape function was addressed in the treatment evaluation), and termination of activity interruptions for Paul.

Duplic Assessment

A duplic is a verbal operant with formal similarity and point-to-point correspondence with its controlling verbal antecedent stimulus, and is maintained by generalized conditioned reinforcement (Michael, 1982). The purpose of the duplic assessment was to identify target vocal mand topographies that participants would emit without challenging behavior in response to vocal models. Procedures were designed to minimize the likelihood that target mand topographies would come under the control of the EO for challenging behavior prior to the treatment evaluation.

Each participant was either seated at a table or on a couch with the experimenter. Prior to the assessment, we selected multiple target vocal mand topographies (“Will you share?” “Not yet, [therapist name],” “Give me toys,” “Let me play,” and “More time, please.”) because they were likely to be recognized and reinforced by caregivers and because they required the same length of utterance. We conducted three trials for each topography and randomly rotated target topographies across all trials. A topography met criteria for inclusion in the treatment evaluation if the participant responded correctly on all three trials for that mand topography. A trial began when the experimenter said “Copy me,” “Say what I say,” or another verbal prompt previously determined to be effective at evoking echoics. Each trial ended when the experimenter praised (e.g., “Awesome! You said it!”) a correct response that occurred within 5 s of the vocal model prompt, or 5 s had passed without a correct response. The experimenter repeated the vocal model on a trial once if the participant said “Huh?” or an equivalent, indicating he may not have clearly heard the prompt. Intertrial intervals were 2–3 s. Praise was the only programmed consequence for correct responding. No toys were visible or accessible to participants. The experimenter continued the assessment until three target vocal mand topographies met criteria. The experimenter praised other forms of compliance such as sitting in the chair. Paul was advanced to the treatment evaluation following completion of his FA because he refused to discontinue ongoing activities, comply with basic instructions to sit at a table, or echo models stated by the experimenter. Target vocal mand topographies that were selected to be prompted in the treatment evaluation were “Will you share?” “Not yet, [therapist name],” and “Let me play” for Zahid; “Wait,” “Stop,” and “Rest” for William; and “Are you done?” “Is it time?” and “My turn, please” for Chris. For Paul, the experimenter selected the topography “I’m busy” because the form of the response was assumed to be considered appropriate and highly likely to produce the maintaining reinforcer in the natural environment from most listeners.

General Procedure

We used reversal designs to evaluate the effects of FCT plus a Lag 1 schedule of reinforcement with prompting procedures on challenging behavior and topographical mand variability. All sessions were 5 min in duration. We conducted an FO-SPA to identify preferred toys before the first session of each day for Zahid. Between sessions, if Zahid asked for a toy not identified in the FO-SPA, we included it in the subsequent session. During phases in which a variant vocal mand was required for reinforcement, novel or untrained sentences (even if grammatically incorrect) were also reinforced. For example, if Zahid said “Say not yet, [therapist name],” “say” was not part of the “Not yet, [therapist name]” target topography identified in the duplic assessment, although we reinforced this variation if all of the words differed from the immediately preceding sentence spoken independently within the session. For Zahid, if the session duration elapsed without a response eligible for reinforcement, the experimenter extended the session until a target response (e.g., challenging behavior or manding, depending on the condition) was emitted. The longest session in his treatment evaluation was 7 min 52 s.

Baseline

Procedures were identical to a test condition that indicated the maintaining reinforcer (i.e., tangible for Zahid and Chris, escape for William, and interruption termination for Paul) in the FA for each participant.

FCT + Lag 0

Paul and Zahid were exposed to this condition. The purpose was to assess the effects of FCT without a requirement to vary on rates of challenging behavior and manding. Sessions were similar to the baseline condition, except that we placed challenging behavior on extinction and continuously reinforced independent manding. For Paul, the experimenter added prompting and prompt fading to two sessions to increase contact between the mand and the programmed reinforcer (i.e., Paul was prompted to say a response within his repertoire: “I’m busy.”).

FCT + Lag 1

This condition assessed the effects of FCT + Lag 1 without prompts on challenging behavior and manding. Procedures were identical to those used during the FCT + Lag 0 condition, with the exception that the experimenter differentially reinforced independent variant mand topographies on a Lag 1 schedule of reinforcement.

FCT + Lag 1 + TD

Zahid, William, and Chris were exposed to this condition. Procedures were identical to those in the FCT + Lag 1 condition, with multiple exceptions based in part on procedures described by Silbaugh et al. (2018). For the first 6 trials of the first session of the first phase, the experimenter or therapist (depending on the participant) prompted a variant vocal mand within 2 s of the onset of the EO. On the seventh trial, the experimenter or therapist introduced a 2-s TD if all prior trials consisted of invariant mand topographies or prompted variant mand topographies. During the 2-s TD, if a variant mand topography did not occur within 2 s of the EO, the experimenter or therapist modeled a target variant mand topography selected quasi-randomly (e.g., “Wait,” “Rest,” or “Stop” for William). The trial was extended, and the experimenter or therapist redelivered the prompt every 2 s until the target variant mand topography was emitted. The length of the TD increased by 2 s every six consecutive trials that a target variant mand topography was not emitted independently. The criteria to increase the length of the TD could only be met within a session. Thus, if the last five mand topographies for a session were invariant, the length of the TD was not increased. Six consecutive invariant mand topographies would be required in a subsequent session for the TD to be increased. The first session of all subsequent FCT + Lag 1 + TD phases beyond the first phase began with the last value of the TD. The experimenter or therapist conducted a probe session without prompts if insufficient independent variant manding occurred for least three consecutive sessions.

Participants could emit multiple responses during the TD, and every response was recorded. Vocal models were provided if an independent invariant vocal mand topography was emitted or no mand was emitted during the TD. Thus, vocal models were not contingent on invariant mand topographies. The TD was held constant at 2 s for the last three sessions of the first FCT + Lag 1 + TD for Chris to assess the effects of this variation of the TD on manding and challenging behavior.

FCT + Lag 0 + TD

William and Chris were exposed to this condition. Procedures were identical to those used in the FCT + Lag 1 + TD condition, except that reinforcement was contingent on any independent vocal mand topography.

FCT + Lag 1 + LTM

This condition assessed the effects of an alternative prompt-fading procedure on levels of challenging behavior and variant manding for Zahid. This condition was similar to FCT + Lag 1 + TD, with some exceptions. Instead of the TD procedure, the therapist delivered increasingly complete vocal models within each trail to evoke variant vocal mand topographies. During each trial, the therapist presented the EO and immediately delivered a vocal model for the first word of the target sentence (e.g., “will”), followed by modeling the first two words of the target sentence (e.g., “will you”), followed by modeling the full target sentence (e.g., “Will you share?”), with prompts separated by 2-s interprompt intervals, which allowed for Zahid to say the full target sentence. The therapist delivered the reinforcer for independent or prompted variant mand topographies only if the topography consisted of a full sentence (e.g., “Let me play.”).

No-prompt probe

We exposed Zahid, William, and Chris to no-prompt probe sessions. The purpose was to use extended periods of extinction to induce independent variant vocal manding by omitting all prompts during a single session. Procedures were identical to those described for the condition in which each probe was embedded, except we omitted prompts. For example, for William, the no-prompt probe in Session 7 consisted of FCT with a Lag 1 schedule of reinforcement but no prompts (i.e., the probe was not equivalent to baseline).

Results

The results of the FAs are displayed in Fig. 1. For Zahid, levels of challenging behavior were consistently elevated in the tangible condition (M = 0.8 RPM) relative to the control condition (M = 0.1 RPM). For William, levels of challenging behavior were consistently elevated in the escape (M = 1.4 RPM) and tangible (M = 0.5 RPM) conditions relative to the control condition (M = 0 RPM). Levels of challenging behavior in the ignore and attention conditions (M = 0 RPM) were equal to the control condition (M = 0 RPM). For Chris, levels of challenging behavior were consistently elevated in the tangible condition (M = 1.7 RPM) relative to the control condition (M = 0 RPM). Challenging behavior was absent in other test conditions (M = 0 RPM, attention condition; M = 0 RPM, escape condition; M = 0 RPM, ignore condition). For Paul, levels of challenging behavior were consistently elevated in the interrupt condition (M = 1.47 RPM) relative to the control condition (M = 0.08 RPM).

Fig. 1
figure 1

Responses per minute of challenging behavior during test and control conditions of the functional analyses for Zahid (top left), William (top right), Chris (bottom left), and Paul (bottom right)

Full size image

The results of Zahid’s treatment evaluation are displayed in Fig. 2. During baseline, elevated rates of challenging behavior (M = 1.25 RPM) occurred. No manding occurred. Upon introduction of FCT + Lag 1 + TD, rates of challenging behavior reduced to zero levels until progressive TD was introduced. As the TD increased from 2 to 6 s across sessions, rates of challenging behavior (M = 0.24 RPM) approached baseline levels and rates of independent manding remained at zero levels despite an increase in the level of prompted variant manding rates (M = 1.16 RPM). With the phase change to FCT + Lag 1 + LTM, rates of challenging behavior (M = 0.76 RPM) and prompted manding (M = 1.24 RPM) continued at the level observed when prompts were faded using TD, with no changes in rates of independent manding (M = 0 RPM of total manding). Coinciding with the phase change to FCT + Lag 0, the level of challenging behavior rates returned to baseline (M = 1.41 RPM). In addition, rates of independent invariant manding gradually increased across sessions (M = 0.89 RPM), whereas rates of variant manding remained at near-zero levels (M = 0.03 RPM). Upon reintroduction of FCT + Lag 1 + TD, an increasing trend in rates of challenging behavior (M = 0.90 RPM) and a decreasing trend in rates of independent manding (M = 0.26 RPM of total independent manding) occurred. In the next phase, the TD was omitted from FCT + Lag 1 for a no-prompt probe session, which coincided with the first large increase in variant manding, a return to prior elevated levels of invariant manding, and a reduced rate of challenging behavior. As the condition was continued in the absence of prompts (i.e., FCT + Lag 1), continued elevated levels of independent variant (M = 0.78 RPM) and invariant manding (M = 1.56 RPM) occurred, as well as highly variable and wide-ranging rates of challenging behavior (M = 2.67 RPM). A return to FCT + Lag 0 then coincided with an immediate reduction in the rates of challenging behavior (M = 0.2 RPM) and variant manding (M = 0 RPM), with no change in rates of invariant manding (M = 1.28 RPM). Upon reintroduction of the final FCT + Lag 1 phase, rates of challenging behavior gradually increased across sessions (M = 0.44 RPM), coinciding with a continuation of the prior rates of independent invariant manding (M = 1.0 RPM) and a return to elevated rates of independent variant manding (M = 1.14 RPM) observed only during FCT + Lag 1 sessions.

Fig. 2
figure 2

Responses per minute of independent variant, independent invariant, prompted variant, and total independent manding (first through fourth panels) and challenging behavior (fifth panel, bottom) across phases of the treatment evaluation for Zahid

Full size image

The results of William’s treatment evaluation are displayed in Fig. 3. During baseline, steady rates of challenging behavior (M = 1.53 RPM) occurred. No manding occurred. Upon introduction of FCT + Lag 1 + TD, little change in independent manding (M = 0.1 RPM) occurred. The level of prompted variant manding increased (M = 1.1 RPM), and a reduction in the level of challenging behavior (M = 0.8 RPM) occurred. Two no-prompt probes were embedded in this phase. In the first probe, no changes in independent manding rates were observed, and a large increase in the rate of challenging behavior was observed. Similarly, in the second probe, no changes in independent manding rates occurred, but the rate of challenging behavior decreased to zero. Upon a phase change to FCT + Lag 0 + TD, no changes in dependent variables, other than a slight increase in prompted variant manding rates (M = 1.6 RPM), occurred. In the reversal to baseline, increased rates, but a decreasing trend, in invariant manding (M = 1.07 RPM) were observed and were associated with a replication of the steady efficient rates of challenging behavior (M = 1.53 RPM) observed in the first baseline phase. Upon reintroduction of FCT + Lag 0 + TD, we observed a shift to an increasing trend in invariant manding (M = 1.2 RPM), and challenging behavior reduced to zero rates. A phase change to FCT + Lag 1 + TD was associated with a brief transient increase in rates of variant manding (M = 0.1 RPM), a decrease in invariant manding rates (M = 1.07 RPM), elevated rates of prompted variant manding (M = 1.13 RPM), decreased total independent manding (M = 0.63 RPM), and rates of challenging behavior that were near zero. During a no-prompt probe session embedded in this phase, the rate of invariant manding was unchanged, but a slight decrease in total independent manding and a return to elevated levels of challenging behavior occurred.

Fig. 3
figure 3

Responses per minute of independent variant, independent invariant, prompted variant, and total independent manding (first through fourth panels) and challenging behavior (fifth panel, bottom) across phases of the treatment evaluation for William

Full size image

The results of Chris’s treatment evaluation are displayed in Fig. 4. During baseline, steady efficient rates of challenging behavior (M = 1.7 RPM) occurred. No manding occurred. Coinciding with FCT + Lag 1 + TD, increased rates of prompted responses (M = 1.67) and a shift to an increasing trend in challenging behavior rates (M = 1.65 RPM) occurred, but no changes in independent manding rates (M = 0.1 RPM) occurred. Within this phase, two no-prompt probes were conducted. During the first probe, a large increase in the rate of challenging behavior occurred, but manding rates were unchanged. As the length of the TD was held constant, challenging behavior rates decreased (M = 0.7 RPM), prompted manding rates were constant (M = 1.2), and almost no change in the rates of independent manding (M = 0.4 RPM) occurred. During the second probe, no changes in dependent variables occurred. Upon introduction of FCT + Lag 0 + TD, as the length of the TD increased, challenging behavior rates reduced to zero levels, and independent variant (M = 0.36 RPM) and invariant (M = 0.76 RPM) manding rates increased and stabilized at no trend as the rate of prompted variant manding (M = 0.56 RPM) decreased across sessions. Upon the introduction of FCT + Lag 1 + TD, independent variant (M = 0.8 RPM) and invariant (M = 2.05 RPM) manding rates steadily increased across sessions, coinciding with a slight decrease in rates of prompted variant manding (M = 0.75 RPM) as challenging behavior continued at zero rates. However, a large increase in the rate of challenging behavior was observed during a no-prompt probe within this phase. Rates of challenging behavior remained at or near-zero levels for the rest of the treatment evaluation. Upon the withdrawal to FCT + Lag 0 + TD, independent invariant manding rates (M = 2.27 RPM) maintained, as independent variant manding rates (M = 0.53 RPM) gradually decreased, the level of total independent manding decreased (M = 3.0 RPM), and prompted variant manding occurred at zero rates. When FCT + Lag 1 + TD was reintroduced, a replication of increasing rates of independent variant manding (M = 1.12 RPM) across sessions occurred. Additionally, although the level of responding was unchanged, a shift to a decreasing trend in rates of independent invariant manding (M = 3.0 RPM) occurred. A similar pattern for total independent manding (M = 4.32 RPM) and an increase in the level of prompted manding rates (M = 0.4 RPM) occurred. No changes in dependent variables occurred during a no-prompt probe session. When the TD procedure was withdrawn in the final FCT + Lag 1 phase, high rates of independent variant manding (M = 1.7 RPM) continued as independent invariant manding (M = 1.2 RPM) steadily decreased.

Fig. 4
figure 4

Responses per minute of independent variant, independent invariant, prompted variant, and total independent manding (first through fourth panels) and challenging behavior (fifth panel, bottom) across phases of the treatment evaluation for Chris

Full size image

The results of Paul’s treatment evaluation are displayed in Fig. 5. During baseline, rates of independent variant manding gradually decreased (M = 1.4 RPM), invariant manding rates (M = 1.9 RPM) showed a decreasing trend, there were no occasions to engage in prompted variant manding, total independent manding rates (M = 3.5 RPM) gradually decreased, and challenging behavior (M = 1.7 RPM) occurred at high, steady rates. Upon introduction of FCT + Lag 0, a change in trend occurred for variant manding (M = 0.8 RPM), invariant manding rates were constant (M = 1.4 RPM), there were no occasions for prompted variant manding, no changes in the rate or trend in total manding (M = 2.4 RPM) were observed, and rates of challenging behavior remained constant (M = 1.6 RPM). When prompting and prompt fading were added to the condition, an immediate decrease in rates of challenging behavior occurred (M = 0.3 RPM), accompanied by a decrease in independent manding (M = 0.6 RPM) and an increase in prompted manding rates (M = 1.1 RPM). Across subsequent sessions of the condition without prompting, a return to steady rates of variant manding (M = 1.0 RPM) occurred, accompanied by no change in invariant manding rates (M = 0.6 RPM), steady rates of total independent manding (M = 1.9 RPM), and a continuation of reduced challenging behavior rates (M = 0.5 RPM). Introduction of FCT + Lag 1 coincided with a small increase in the level of variant manding rates (M = 1.4 RPM), no change in invariant manding (M = 0.7 RPM), a small increase in the level of total independent manding (M = 2.2 RPM), and no change in rates of challenging behavior (M = 0.6 RPM). Upon return to FCT + Lag 0, rates of variant manding (M = 0.7 RPM) gradually reduced across sessions, coinciding with no changes in invariant manding rates (M = 0.9 RPM), reduced variability in independent manding rates (M = 1.7 RPM), and a reduction of challenging behavior to zero rates. Upon reintroduction of FCT + Lag 1, a return to slightly elevated and steady rates of variant manding occurred (M = 1.4 RPM), accompanied by an increase in the range of invariant manding rates (M = 1.4 RPM), an increase in the level of total independent manding (M = 3.0 RPM), and a slight increase in the level of challenging behavior rates (M = 0.3 RPM). FCT was withdrawn to baseline, and a large immediate increase in the level of variant manding rates occurred (M = 3.9 RPM), coinciding with no changes in invariant manding (M = 1.1 RPM), an increase in total independent manding (M = 5.2 RPM), and a return to baseline rates of challenging behavior (M = 1.7 RPM). FCT + Lag 1 was reintroduced, which coincided with a return to previous slightly elevated but steady rates of variant manding (M = 1.5 RPM) that occurred in the prior FCT + Lag 1 condition, as well as an increase in the range of invariant manding, a slightly elevated level of total independent manding (M = 3.9 RPM) with an expanded range, and a replication of prior rates of challenging behavior (M = 0.6 RPM) that were relative to baseline.

Fig. 5
figure 5

Responses per minute of independent variant, independent invariant, prompted variant, and total independent manding (first through fourth panels) and challenging behavior (fifth panel, bottom) across phases of the treatment evaluation for Paul; P&PF = prompting and prompt fading; BL = baseline

Full size image

A summary of vocal mand topographies used by each participant during the treatment evaluation is displayed in Table 1. For Zahid, all target alternative vocal mand topographies occurred independently. We observed a total of seven different independent vocal mand topographies. For William, all topographies that were prompted ultimately occurred independently. We observed a total of four different independent vocal mand topographies. For Chris, all topographies that were prompted occurred independently. We observed a total of 13 different independent vocal mand topographies. For Paul, the topography “I’m busy” was prompted and subsequently occurred independently, and we observed a total of 60 different independent vocal mand topographies.

Table 1 Counts of Mand Topographies That Occurred During the Treatment Evaluation for Zahid, William, Chris, and Paul
Full size table

Discussion

We evaluated the effects of FCT plus a Lag 1 schedule of reinforcement on topographical vocal mand variability and challenging behavior. We replicated prior research on FCT and lag schedules by demonstrating that lag schedule conditions coincided with the highest levels of independent variant vocal manding for three of four participants. We provide the first demonstration of experimental control over independent variant vocal manding by lag schedules without prompts during FCT for Paul and Zahid, and maintenance of elevated rates for Chris when prompts were eliminated from the lag schedule. Together, these findings suggest lag schedules may be used alone or in combination with response prompt-fading strategies during FCT to expand mand response classes and reinforce functionally equivalent topographical vocal mand variability. Our results further extend the literature on mand variability, lag schedules, and FCT.

Mand training procedures that do not include contingencies to support varied manding from a speaker with language delays or deficits can result in invariant manding (e.g., Carr & Kologinsky, 1983). Few studies have evaluated functional relations between mand variability and environmental variables. Each study demonstrated increases in mand variability or novel instances of manding across concurrently available reinforcers (i.e., across mands) during mand training for individuals with autism (Bernstein & Sturmey, 2008; Betz, Higbee, Kelley, Sellers, & Pollard, 2011; Brodhead et al., 2016; Carr & Kologinsky, 1983; Drasgow, Martin, Chezan, Wolfe, & Halle, 2015; Sellers, Kelley, Higbee, & Wolfe, 2015) or intellectual disability (e.g., Duker & van Lent, 1991). Alternatively, our group demonstrated that a Lag 1 schedule of reinforcement with progressive TD increased variability in the words children with autism used to mand for the same reinforcer (i.e., topographical vocal mand variability; Silbaugh et al., 2018). A limitation of Silbaugh et al. (2018) was that the TD component was not completely withdrawn, and therefore the separate effects of the lag schedule and prompts could not be determined. The data from three participants in the current study addressed this limitation. For Chris, after we demonstrated experimental control over independent variant manding by FCT + Lag 1 + TD, elevated independent vocal mand variability continued during a no-prompt probe and during an FCT + Lag 1 phase in which prompts had been withdrawn. For Zahid, we demonstrated independent variant vocal manding during a no-prompt probe, and subsequent experimental control over independent variant vocal manding by FCT + Lag 1. For Paul, we demonstrated experimental control over independent variant vocal manding by FCT + Lag 1 despite no prior prompts to vary. Although mand variability was reinforced in the current study in the context of FCT, the results may be generalizable to mand variability training in general.

The current study represents an extension of prior studies of FCT plus lag schedules in several ways. First, compared to prior studies (Adami et al., 2017; Falcomata et al., 2018), we included younger participants. Second, Adami et al. (2017) and Falcomata et al. (2018) investigated manding across nonvocal, largely selection-based mand modalities (e.g., tablet, a card to exchange), whereas this study targeted vocal, topography-based manding. Third, Adami et al. (2017) and Falcomata et al. (2018) did not use prompts, whereas we systematically used response prompt fading to establish contact between variant manding and the lag schedule. In the current study, whereas response prompting maximized reinforcement of variant responding, it is possible that prompt-fading procedures facilitated the transfer of antecedent stimulus control over variant responding. Fourth, in the current study, we measured concomitant changes in independent invariant manding, which were not assessed in Adami et al. (2017) or Falcomata et al. (2018). This extension of the literature is important because invariant and variant manding are not mutually exclusive. In one study, increased vocal variant manding produced by a Lag 1 schedule of positive reinforcement coincided with consistently decreased variant manding (Silbaugh et al., 2018), but this same clear relation was not demonstrated in a replication study targeting sign manding (Silbaugh & Falcomata, 2018). In the current study, invariant manding was not consistently lower during lag schedule conditions for any participants. Together these findings suggest that in some cases lag schedules may have the unwanted effect of increasing early trial bursts of invariant manding when the EO remains in effect during invariant manding. Additional research is needed to examine the conditions under which lag schedules consistently produce reductions in invariant manding as variant manding increases. Fifth, in contrast to Adami et al. (2017), we included a participant whose challenging behavior was maintained by termination of activity interruptions.

The demonstrated lack of consistent reductions in challenging behavior to clinically significant levels is both a limitation and an important finding because it draws attention to a potential disadvantage of combining FCT with lag schedules. We demonstrated independent variant vocal manding under a lag schedule in three participants. Of these three participants, Chris exhibited zero levels of challenging behavior prior to the observed increase in independent variant vocal manding, and Zahid and Paul continued to exhibit levels of challenging behavior similar to baseline when initial increases in independent variant vocal manding were observed and throughout the remainder of the treatment evaluation. We speculate that temporal proximity between challenging behavior and manding may have allowed for adventitious reinforcement of challenging behavior. Alternatively, if manding and challenging response topographies belonged to the same operant, and the lag schedule embedded in FCT differentially reinforced independent variant vocal mand topographies, response generalization may have occurred in relation to challenging response topographies. Said differently, reinforcement of the variant dimension of the operant (e.g., Page & Neuringer, 1985) with respect to vocal manding may have spread to the variant dimension of the operant with respect to challenging behavior. Additional research is needed to identify variables that can be manipulated during FCT with lag schedules to eliminate challenging behavior during treatment, and studies should examine whether variant challenging behavior is affected by reinforcement of variant manding.

For William, FCT plus lag schedules and TD decreased challenging behavior regardless of the value of the lag. It is unclear why FCT failed to produce steady efficient rates of independent manding. Possible explanations include that the FA may have yielded a false positive, that the effort associated with vocal manding was too high, or that the availability of praise contingent on compliance may have abolished negative reinforcement for challenging behavior. Anecdotal observations were that compliance with instructions was accompanied by gradually increased smiling, laughter, and closeness with the experimenter across FCT sessions. This treatment package may be effective for some young children with autism with a highly limited vocal manding repertoire when (a) caregivers or other stakeholders reject the use of augmentative alternative communication systems, (b) they place more value on compliance, (c) prior attempts to increase independent alternative mands were unsuccessful, (d) previously established alternative mands fail to maintain and/or generalize, or (e) a combination of these factors are present.

All participants emitted a variety of independent variant vocal mand topographies. William only independently exhibited topographies that were previously prompted. However, for participants in whom we demonstrated experimental control over independent variant manding by the lag schedule (Zahid, Chris, and Paul), a variety of topographies were exhibited that were never prompted during the study and that may be considered novel within the context of the study. All topographies exhibited were contextually appropriate in that their structure corresponded to either public stimuli that both speaker and listener (Skinner, 1957) were in contact with, or interactions with the listener, although not all topographies were grammatically correct. This finding is consistent with prior reports in the literature of increased novel verbal responding under lag schedule conditions (e.g., Contreras & Betz, 2016; Koehler-Platten, Grow, Schulze, & Bertone, 2013). Future research could directly evaluate the effects of FCT with lag schedules on the emergence of novel vocal mand topographies.

Multiple other potential limitations of the current study, such as a lack of fidelity data and differences in the conditions across participants, are important to note and suggest opportunities for future research. First, in the current study, we did not use designs that could rule out the contribution of potential sequence effects on changes in the dependent variables due to the sequential introduction of response prompt-fading procedures and lag schedules. Second, some of Chris’s independent variant mand topographies may be considered inappropriate vocalizations. However, Paul’s mother approved of the current procedures, and all vocalizations were considered an improvement over his challenging behavior. Future research could empirically evaluate the social validity of mand variability reinforced by lag schedules during FCT using standardized questionnaires to assess caregivers’ perspectives. At this time, the literature on lag schedules is too limited to speculate on how placing constraints on the “appropriateness” of variant mand topographies might influence mand variability or the clinical implications of such procedures related to “appropriateness” of responding, so future research should also evaluate the effects of FCT plus lag schedules on mand variability when eligible responses are constrained by more conservative qualitative criteria. In practice, it would be clinically prudent to quickly adjust the intervention early in treatment to shape increasingly more socially acceptable alternative mand topographies, but more research is needed to determine the relevant procedures.

Third, we did not attempt to demonstrate experimental control over challenging behavior by FCT for Chris and Zahid. Fourth, the fact that we did not standardize operational definitions of vocal mand topographies across participants may be considered a limitation. However, it is possible that individualized definitions were necessary to achieve increased levels of topographical mand variability with the lag schedule. That is, the degree and manner in which the participants varied manding prior to the treatment evaluation suggested that quickly bringing mand variability in contact with the lag schedule would require operational definitions that most closely reflected observed patterns of manding for each participant. Therefore, the experimenter individually defined “different” topographies based on the manner in which each participant varied vocal topographies under assessment conditions or during baseline reinforcement of challenging behavior under FA conditions or the first baseline in the treatment evaluation. If we used the same operational definition of a variant mand topography for all participants and reinforced only “appropriate” variant topographies, our results may have differed.

Fifth, we used FA designs for William and Zahid that could not rule out multiple functions, and some mands emitted by Paul during reinforcement intervals were impossible to reinforce (e.g., asking the experimenter to leave). To limit the complexity of the procedures, we targeted only a single function of challenging behavior for each participant during treatment evaluations. Therefore, during reinforcement intervals, we could not be certain that consequences, which may have been functionally related to challenging behavior, but which differed from the programmed reinforcer, were completely controlled. To ensure that the graphed data represented instances of behavior controlled by the programmed reinforcer, we only measured dependent variables when the programmed EO was present. A threat to internal validity is unlikely because measurement procedures were held constant across conditions of the treatment evaluation.

In summary, the current study replicated and extended prior research by showing FCT with lag schedules increased vocal mand variability in three of four children with autism. In contrast to prior research, we failed to replicate consistent reductions in challenging behavior. These findings highlight the need for additional research on the effects of lag schedules on mand variability in the context of FCT, the clinical utility of this treatment approach, and its effects on the recurrence of challenging behavior.