Generalized Reflexive Responding and Cross-Modal Tactile Transfer of Stimulus Function in Children with Autism

Abstract

We sought to evaluate the efficacy of successive matching training for establishing generalized reflexive matching across 4 children with autism. In Experiment 1, differential reinforcement with delay fading was efficacious in establishing “yes” and “no” matching and nonmatching responses in 2 participants when 2 identical or nonidentical picture stimuli were presented. In addition, emergent visual–visual reflexive relational responses were observed using novel picture stimuli in a transfer test phase. In Experiment 2, differential reinforcement alone was efficacious in establishing matching and nonmatching responses in the other 2 participants when 2 identical or nonidentical objects were presented. Transfer to identical objects presented through touch (i.e., tactile discrimination) was additionally observed for both participants. Procedures in the study were adapted from the PEAK Relational Training System to aid in clinical replication, and the translational results have implications for language training with individuals with autism.

Stimulus equivalence theory (Sidman, 1971; Sidman & Tailby, 1982) and contemporary extensions (e.g., relational frame theory, bidirectional naming; Hayes, Barnes-Holmes, & Roche, 2001; Horne & Lowe, 1996) have significantly progressed behavior science by describing the emergent nature of language in the absence of direct-acting contingencies (Barnes-Holmes & Barnes-Holmes, 2000; Hayes et al., 2001). Stimulus equivalence requires the demonstration of four events: reflexivity, symmetry, transitivity, and the transfer of stimulus function (see Sidman & Tailby, 1982, for an in-depth description of each of these behavioral phenomena). Reflexivity occurs when a participant can match identical stimuli (A = A), symmetry occurs when a participant is taught to match a stimulus A with a stimulus B (A = B) and can match the stimulus B with the stimulus A (B = A) without direct reinforcement, and transitivity occurs when a participant can match two stimuli based on a mutual relationship with a third stimulus (e.g., taught A = B and B = C, derive A = C and C = A). Symmetrical and transitive responding are considered arbitrarily applicable relational responses (Hayes et al., 2001), as formal similarity between class members is not required. Reflexivity, which emerges prior to symmetry and transitivity, encompasses a large class of relational responses that are nonarbitrary, in that class members are related in terms of identical stimulus properties, or sameness. Demonstrating reflexive relational responding, therefore, involves responding in terms of the identical properties of two stimuli presented at different locations and/or at different points in time.

One theory is that relational responding emerges as a set of higher order generalized operant response classes (Barnes-Holmes & Barnes-Holmes, 2000; Healy, Barnes-Holmes, & Smeets, 2000; Luciano, Becerra, & Valverde, 2007). When multiple-exemplar relations are directly reinforced, novel relations can be demonstrated in the absence of reinforcement. The generalized nature of relational responding may be most readily apparent in terms of reflexivity as generalized identity matching, in that once matching some identical stimuli in the environment is reinforced, matching virtually any identical stimulus may emerge. As noted by Urcuioli (2011), demonstrating the emergence of generalized identity matching with humans is unlikely given a preexperimental ability to match any two identical stimuli. To develop an experimental model of reflexive responding, research by Urcuioli (2011; Sweeney & Urcuioli, 2010) and others (e.g., Barros, Galvão, & McIlvane, 2002) has demonstrated the emergence of generalized identity matching in nonhuman animals (e.g., pigeons, monkeys). Research with children with autism may provide an alternative avenue for promoting the emergence of untrained, generalized identity matching in humans, as identity matching may not develop naturally (i.e., preexperimentally) with this population. Additionally, given the likely role of reflexivity in equivalence, and the relationship between relating equivalent arbitrary events and core learning deficits in autism (Belisle, Dixon, & Stanley, 2018; Dixon, Belisle, & Stanley, 2018b; Stewart, McElwee, & Ming, 2013), evaluating the efficacy of procedures that promote generalized identity matching is a socially valid application of prior research that may apply to broader habilitative efforts consistent with stimulus equivalence theory.

Several studies have now utilized a match-to-sample (MTS) arrangement to establish reflexive relational responding in individuals with autism (e.g., Dube, Iennaco, & McIlvane, 1993; Slocum, Miller, & Tiger, 2012; Toussaint, Scheithauer, Tiger, & Saunders, 2017). In a typical MTS arrangement, a sample stimulus (A) and a given number of comparison stimuli (A, B, and C) are presented, where one comparison stimulus is identical to the sample (A = A) and the other comparison stimuli are nonidentical to the sample (A ≠ B and A ≠ C). Reflexive relational responding is demonstrated when the participant selects the identical comparison. Once a set of reflexive relations is mastered, novel topographies are tested without reinforcement in a transfer test phase to determine if the reflexive response is a generalized operant. Slocum et al. (2012) implemented an MTS procedure to evaluate the efficacy of a blocking procedure in promoting visual reflexive responding in terms of colors, shapes, and symbols. Toussaint et al. (2017) extended these results by incorporating the tactile sensory system (i.e., touch), requiring early braille learners to match identical braille letters. In both studies, reflexive responding transferred to novel stimuli contained in the same sense mode, supporting the position that reflexivity may be topographically boundless following direct reinforcement of multiple exemplars. Additionally, the transfer of stimulus function is not restricted to the visual sense mode that is most frequently employed in studies on relational responding (Rehfeldt & Dixon, 2005); rather, this is likely a multimodal phenomenon occurring across each of the five sense modes (i.e., visual, auditory, olfactory, tactile, and gustatory).

As noted by Debert, Huziwara, Faggiani, Mathis, and McIlvane (2009), any procedure that establishes ordered pairs of related events may allow for responding relationally, including reflexive relational responding. An alternative to MTS arrangements may be successive matching training (SMT; Urcuioli, 2011; Wasserman, 1976), where identical or nonidentical stimuli are presented at different points in time, and a discriminated response is required given the stimuli that are sequentially presented. Matching occurs when one response is made more frequently in the presence of any combination of identical components (e.g., Response X given A–A or B–B). Conversely, nonmatching discrimination occurs when another response is made more frequently in the presence of any combination of nonidentical components (e.g., Response Y given A–B or B–A). SMT is similar to a go/no-go arrangement in that a compound stimulus is presented; however, in SMT there is a delay between the presentation of the sample stimulus and the identical or nonidentical comparison stimulus, and the trials are discrete rather than continuous. SMT has been used in several animal models of generalized reflexive responding (e.g., Urcuioli, 2011) and with typically developing adult participants to establish equivalence classes consisting of exclusively auditory stimuli with some experimental modifications (Dube, Green, & Serna, 1993). A successful demonstration of this technique with typically developing adult participants was provided by (Lantaya, Miguel, Howland, LaFrance, & Page, 2018). Across four experiments, the researchers conducted SMT by presenting a visual stimulus on the screen followed by a delay and a comparison visual stimulus. If the pairs were related, the participant touched the comparison stimulus. Results suggested that this was effective in establishing baseline relations, as well as derived relations, in this study.

In the studies by (Dube et al., 1993) and (Lantaya et al., 2018), these equivalence arrangements are likely too complex for children with autism; however, there may be advantages to adapting SMT procedures for evaluating emergent reflexive responding with this population. First, SMT arrangements have been used in research on memory and delayed recall with both human and animal subjects (Simmonds, Pekar, & Mostofsky, 2008; Winocur et al., 2005); this is possible when a short-term or long-term delay is imposed between the presentation of the sample and comparison stimuli. This design, therefore, affords similar advantages to a delayed MTS (DMTS) procedure in which a sample stimulus is presented and then removed for a set delay, followed by the presentation of a selection array (e.g., Ratkos, Frieder, & Poling, 2016). By imposing a delay, researchers can attempt to extend the delay through shaping to improve delayed recall (Arntzen, 2006). Other procedures such as verbal mediational training can even further extend the delay but may not be needed when the delayed presentation is 30 s or less (Ratkos et al., 2016). Second, reflexive relational responding can be evaluated in terms of each of the five sense modes that could not otherwise be evaluated using an MTS or DMTS procedure. For example, edible stimuli cannot be presented at the same time in a selection array; however, edible stimuli can be presented sequentially, when a sample stimulus is consumed followed by a second identical or nonidentical comparison edible stimulus. The SMT procedure could also allow for an analysis of the cross-model transfer of generalized identity matching, where reinforcing identity matching generalizes not only to untrained stimuli within the same sense mode (e.g., visual) but also to untrained stimuli in another, untrained sense mode (e.g., tactile). Third, SMT procedures extend the potential complexity of reflexive responding, where reflexivity may not exclusively refer to matching events that are both present, but also matching events that occur at different times (i.e., temporal locus). Therefore, it is not simply the case that symmetry is more complex than reflexivity, and transitivity is more complex than symmetry; rather, varying degrees of complexity exist within each of these relational behaviors. For example, identity matching in an SMT arrangement following a 10-s delay between the presentation of the stimulus and the sample represents a likely more complex response than identity matching in an MTS arrangement. Also, responding correctly in an SMT arrangement using tactile, olfactory, or gustatory stimuli may be more complex than with more typically used visual or auditory stimuli. Development of all of the previously described complexities of reflexive responding can be interpreted as generalized operants that are related but distinct and potentially build upon one another. Finally, correct responding in an SMT arrangement can require abstracting the concepts of “yes” and “no” through multiple exemplars, which is a common deficit experienced by individuals with autism. For example, a participant must abstract “yes” and “no” when presented sequentially with two identical or nonidentical stimuli and asked “Were those the same?” Although correct responding in an SMT arrangement may be expected in typically developing populations, procedures are needed to establish delayed reflexive responding in individuals with autism. Requiring the abstracted tacts “yes” and “no” was a feature of the present study to improve the social utility of the training arrangement.

In the present study, we sought to extend the now considerable work on generalized identity matching in children with autism in two ways. The first experiment was conducted with participants with severe forms of autism to determine if direct reinforcement and delay fading would lead to generalized identity matching of visual stimuli in an SMT arrangement. The second experiment was conducted with participants with milder forms of autism to determine if reinforcement of delayed matching and nonmatching would lead to an untrained cross-modal transfer to the same objects presented in another sense mode (i.e., tactile/touch). Cross-modal reflexive transfers of stimulus function have never been demonstrated in prior literature; thus this study potentially expands the scope of what is understood to comprise reflexive responding as a generalized operant. That is, matching in an SMT arrangement is a generalized operant that may generalize not only to other topographically dissimilar stimuli but also to stimuli contacted through an entirely different sense mode. Procedures in both experiments were adapted from the Promoting the Emergence of Advanced Knowledge Relational Training System (PEAK) (Dixon, 2015), a comprehensive language training protocol designed for use with individuals with autism. We elected to use standardized procedures to aid in the clinical replication of the procedures of the study.

Experiment 1: Picture–Picture Reflexive Responding and Novel Picture–Picture Transfer

Method

Participants and Setting

Two children with a medical diagnosis of autism participated in Experiment 1 of the study. Emily was a 4-year-old female, and Chris was a 4-year-old male. Both participants were diagnosed with severity Level 3, requiring very substantial support (American Psychiatric Association, 2013), and were receiving early intensive behavior intervention (EIBI) therapy using applied behavior analysis (ABA) for 20–30 hr per week in a clinical setting. Both participants engaged in similar topographies of challenging behavior: tantrums, self-injurious behavior, and physical aggression toward staff and peers. Additionally, both participants were not able to demonstrate vocal responding as identified on the PEAK Direct Training Assessment (Dixon, 2014), which was completed by the children’s parents at the onset of the study. No vocal responding was observed by the researchers throughout the study.

Participants’ language skills at the time of the study were evaluated by obtaining Verbal Behavior Milestones Assessment and Placement Program (VB-MAPP; Sundberg, 2008) results from client records and by conducting the PEAK Equivalence Preassessment (PEAK-E-PA; Dixon, 2015). The assessments were conducted within 1 month of the onset of the study. The VB-MAPP is an assessment instrument that identifies skills for language-based training, and the assessments were completed using a combination of naturalistic observation and direct testing. The intended typically developing age range for the VB-MAPP is up to 4 years of age. Emily scored a 41.5 out of a possible 170, which placed her developmental skill level comparable to a typically developing child aged 0–18 months. Chris scored a 60.5 out of a possible 170, which placed his developmental skill level comparable to a typically developing child aged 18–30 months. The PEAK-E-PA is a directly implemented standardized assessment instrument that provides an estimate of a participant’s ability to derive reflexive, symmetrical, transitive, and equivalence relations (Dixon, 2015). Both participants achieved a score of 2 points out of a possible 48, demonstrating the ability to match identical pictures in a standard MTS arrangement. Neither participant demonstrated reflexive responding in an SMT arrangement in the assessment using auditory, gustatory, and tactile stimuli. In this assessment, the assessor presents the first stimulus, then immediately presents a second stimulus and asks, “Were those the same?” representing a similar arrangement to that used in the present study. Visual reflexive SMT performance is not tested in this assessment. Rather, visual SMT performance was tested in the baseline phase of the current study. Neither participant could demonstrate symmetry, transitivity, or equivalence at the time of the study in any sense mode.

The study took place in the ABA clinic specialized for children with autism where the participants received EIBI therapy. All sessions were conducted in a therapy room isolated from other clients. The therapy room contained a table, two chairs, preferred stimuli, and materials needed to conduct the procedures in the study (described next). Sessions were conducted by the authors of the study, and occasionally a second observer was present in the therapy room.

Materials

The materials used in the study were those described in the PEAK Equivalence Module program “2A–Reflexivity: Pictures.” The program is one of 184 programs contained in the PEAK Equivalence Module and provides instructions for caregivers for how to conduct the procedures. Materials included two sets of three 6 cm × 6 cm picture cards. Set 1 pictures included a lemur (A1), a raisin (A2), and a speaker (A3), and Set 2 pictures included a snowflake (A4), a tree (A5), and a rainbow (A6). The pictures are shown in Fig. 1. A green text card and a red text card were placed on opposite sides of the table adjacent to the participant, containing the text “YES” and “NO,” respectively. The text cards were provided because neither participant could demonstrate vocal responding at the time of the study. Various preferred items were present in the room and included an iPad, wooden fruit set, train set, putty, books, music box, small figurines, and toy cars. The preferred items were identified as potentially preferred items by the participants’ parents.

Fig. 1.

Picture stimuli for Set 1 (left) and Set 2 (right) used in Experiment 1 of the study

Procedure

A multiple-baseline design across participants was used in Experiment 1 in an A-B-C-D design. The A phase was a baseline phase, the B phase involved training of Set 1 stimulus relations using differential reinforcement (DR) and delay fading, the C phase involved continued training of Set 1 without delay fading, and the D phase was a transfer phase where reflexive relational responding was evaluated using untrained Set 2 stimuli.

Dependent Variables and Interobserver Agreement

The dependent variable was the percentage of correct responding within five-trial blocks in each phase. The percentage of correct responding was calculated by summing the number of correct independent responses per trial block, dividing by the number of trials in each block, and multiplying by 100. All trials that required prompting were considered incorrect. Interobserver agreement (IOA) was assessed for 50% of trials and was calculated by summing the number of trials in which both observers reported the same score, dividing by the total number of trials, and multiplying by 100. IOA was 100%.

Baseline Phase

In the baseline phase, reflexivity was tested for both stimulus sets. (For Chris, Set 2 stimuli were tested on two occasions to evaluate stability in this phase.) For each trial, the researcher presented the sample picture (A1, A2, or A3), removed the sample picture, and, following a 2-s delay, presented the second picture (A1, A2, or A3). The experimenter then asked, “Were those the same?” All trials were presented in a tabletop format with the pictures clearly visible. If the pictures were identical (i.e., A1–A1, A2–A2, or A3–A3), then a correct response occurred when the participant touched the “YES” text card. If the pictures were not identical (e.g., A1–A3, A2–A1), then a correct response occurred when the participant touched the “NO” text card. The participants were given 5 s to respond before the next trial was presented. The presentation of picture cards was randomized in each trial block. Each trial block contained at least two identical stimulus presentations and two nonidentical stimulus presentations. In all phases, the implementers were instructed to avoid differentially gazing toward the correct response card to avoid cuing the correct response.

DR, Prompting, and Delay Fading

Prior to running each five-trial block in the DR + delay-fading phase, the researcher conducted a brief free-operant preference assessment to establish the preferred item that would be provided to the participant following each correct response. To conduct the preference assessment, all available items described in the Materials section were placed on the floor, and the item that the participant engaged with the most within a 30-s period was provided by the experimenter to the participant as a reinforcer in this phase. Only Set 1 stimuli were used in this phase, and the stimulus presentation was similar to that of the baseline phase. For each trial, the experimenter sequentially presented the two picture cards and asked, “Were those the same?” Trials varied systematically in this phase both in terms of the delay between the presentation of the sample and comparison stimuli and in terms of the use of error correction and prompt strategies used to evoke the correct response. First, the experimenter presented the picture cards simultaneously (i.e., no delay; A–A), and if the participant demonstrated an incorrect response or did not demonstrate a selection response within 5 s, the experimenter physically prompted the correct response. Once the participant achieved three consecutive trial blocks with a mean percentage correct responding of at least 40%, a 0.5-s (A-0.5-A) delay was imposed between the presentation of the sample and comparison stimuli, and a verbal/gestural error correction procedure was added to the sequence. The verbal/gestural correction involved the experimenter saying, “No, try again,” and pointing to the correct text card following an incorrect selection response. If the error correction procedure failed to evoke the correct response, or if the participant failed to demonstrate a response within 5 s of the presentation of the comparison stimulus, a physical prompt was used. Again, once the participant achieved three consecutive trial blocks with a mean percentage correct responding of at least 40%, the delay was increased to 1 s (A-1-A), and only the verbal component of the error correction procedure was used when the participant demonstrated the incorrect response. That is, the experimenter no longer modeled the correct response along with the verbal correction, and if a correct response was not evoked by the correction, the experimenter simply progressed to the next trial after 5 s. Finally, if the participant achieved 40% correct responding for three consecutive trial blocks, a 1.5-s delay (A-1.5-A) was imposed, and only the verbal correction was delivered contingent on an incorrect response, as in the prior sequence. Throughout all trials in this phase, participants were provided with 30 s of access to the preferred item contingent on independent correct responding. A mastery criterion for this phase was set at a mean of at least 60% correct responding across five consecutive trial blocks and an increasing trend across those five trial blocks. This criterion was adopted to ensure that the participant could demonstrate a “yes” or “no” response consistently and respond to verbal prompts. That is, although the response might be incorrect at this stage, a response was consistently occurring, allowing for the contingencies to be effective in later stages. The mastery criterion in this phase was selected to establish the “yes” and “no” selection responses by fading the prompts, as well as to build toward the terminal 2-s delay in the next phase. The DR alone phase and the transfer phase were then introduced to build toward the terminal mastery criterion in the present study.

DR, Prompting, and Transfer Phases

The same preference assessment method as was used in the DR + delay-fading phase was used throughout the DR alone phase. Again, only Set 1 stimuli were presented during this phase, with a 2-s delay (A-2-A) between the presentation of the sample and comparison stimuli. If participants demonstrated the correct response, then they were provided 30 s of access to their preferred item, and if participants demonstrated the incorrect response, the experimenter delivered the verbal prompt. In this phase, a mastery criterion for the training stimuli was set at a mean of 80% correct responding or greater across at least five consecutive trial blocks before progressing to the transfer test phase. During the transfer phase, the experimenter exclusively presented Set 2 stimuli to determine if reflexive relational responding transferred to the untrained identical and nonidentical stimulus pairs. Reinforcement and prompts were not delivered during this phase, and the stimuli were presented with a 2-s delay (A-2-A).

Results

The results for Emily and Chris in Experiment 1 of the present study are summarized in Fig. 2. Baseline results across both participants show that neither participant could correctly demonstrate the matching or discriminated nonmatching response in baseline across both stimulus sets. Although not displayed in the graphed data, both participants failed to touch either the “YES” or “NO” options when asked, “Were those the same?” by the experimenter.

Fig. 2.

Percentage correct independent responses for Emily and Chris during the SCD arrangement. Transfer probes were conducted with novel stimuli (Set 2)

DR + delay fading was conducted with Emily following four baseline test blocks using Set 1 stimuli. Across the first six trial blocks, Emily demonstrated 0% correct responding, requiring a physical prompt to demonstrate the correct response. On the sixth trial block, Emily demonstrated a correct response (20%), and correct responding increased throughout the DR + delay-fading phase. In the final five trial blocks, Emily’s mean percentage correct responding was 84%. Emily also met the mastery criterion of five consecutive trial blocks with an increasing trend and a mean percentage correct responding above 60%. In the DR alone phase, Emily’s correct responding initially decreased given a 2-s delay between the presentation of the sample and comparison stimuli. Throughout the phase, Emily’s correct responding increased, resulting in a mean percentage correct responding of 80% in the final five trial blocks, thus achieving the mastery criterion in this phase. Stimuli from Set 2 were then tested in the transfer phase, and the results suggest an increase in correct responding relative to baseline, resulting in a mean of 93%.

DR + delay fading was then conducted with Chris following 10 trial blocks, including two test probes of Set 2 stimuli. Chris demonstrated 0% correct responding across the first five trial blocks, requiring a physical prompt to demonstrate the correct response. His rate of correct responding increased throughout the phase, resulting in a mean percentage correct responding of 64%, thus meeting the mastery criterion to progress to the next phase. In the DR alone phase, where the delay was increased to 2 s, Chris initially showed greater variability in correct responding relative to the end of the previous phase; however, his correct responding increased throughout the phase, resulting in a mean percentage correct responding of 84% in the final five trial blocks, thus meeting the mastery criterion. Although Chris technically met the mastery criterion on the 58th trial block, a decreasing trend was observed within those five trials, so the phase was continued to ensure that his performance was consistent before progressing to the transfer phase. Stimuli from Set 2 were then tested in the transfer phase, and Chris demonstrated an increase from baseline, achieving a mean percentage correct responding of 88%.

Experiment 2: Object–Object Reflexive Responding and Tactile–Tactile Transfer

Method

Participants and Setting

Two children with a medical diagnosis of autism with severity Level 1, requiring support (American Psychiatric Association, 2013), participated in the second experiment. Joe was a 6-year-old male and was enrolled in a regular first-grade classroom in a public school. Ben was a 7-year-old male and was enrolled in a specialized school for children with autism. Both participants received 4 hr of ABA therapy guided by the PEAK Relational Training System either at a clinic (Joe) or at the specialized school (Ben). No challenging behaviors were reported to the experimenters for either of the participants. For inclusion in this experiment, both participants were required to demonstrate consistent vocal “yes” and “no” responses. Their ability to respond vocally with “yes” and “no” was endorsed by both participants’ parents on the PEAK Direct Training Assessment, which was completed indirectly at the onset of the study. This item was also assessed in a 10-trial block by the experimenters by asking common yes/no questions, where both participants demonstrated 100% correct responding at the onset of the study.

As in Experiment 1, the participants’ language skills at the time of the study were evaluated by obtaining VB-MAPP results from client records and by conducting the PEAK-E-PA directly with the participants. Both participants achieved the maximum possible score of 170 on the VB-MAPP, indicating that the participants had language skills exceeding those expected of a typically developing 4-year-old. Joe’s PEAK-E-PA score was 17 out of 48, suggesting that Joe could demonstrate reflexivity in an MTS arrangement with visual stimuli, as well as reflexive responding in an SMT arrangement with auditory stimuli. Joe was not able to demonstrate reflexive responding in an SMT arrangement with gustatory or tactile stimuli, but he was able to demonstrate basic and intermediate symmetrical relations. Ben’s PEAK-E-PA score was 4 out of 48, suggesting that Ben could demonstrate matching in an MTS arrangement but not in an SMT arrangement with auditory, gustatory, or tactile stimuli.

The study took place in the same location in which the participants regularly received ABA therapy. For Joe, sessions were conducted in a therapy room containing two chairs, a table, and the materials needed to complete the study (described next). For Ben, sessions were conducted in a therapy room at the school that was separate from the other students. The room contained a desk, three chairs, and the materials needed to complete the study (described next). Sessions were conducted by the authors of the study, and occasionally a second observer was present in the therapy room. No preferred items were present, as verbal praise was used as a reinforcer for both participants.

Materials

The materials used in the study were those described in the PEAK Equivalence Module programs “2B–Reflexivity: Objects” and “3C–Reflexivity: Tactile Matching.” The first program was used to train tactile–tactile reflexive responding, and the second program was used to evaluate the transfer of reflexive responding when stimuli were presented through the tactile sense mode. Stimuli were identical across both programs. For Joe, stimuli included a small rubber ball, a rag, and a toy tractor. For Ben, stimuli included a small rubber ball, putty, and a toy truck. The stimuli were selected to ensure that they were discriminable by touch. Other objects included a paper screen and a tactile sensory box. The tactile sensory box had a hole on either end such that the experimenter could insert the object and the participants could insert their hands to feel the object.

Procedure

A multiple-baseline design across participants was used in Experiment 2, with embedded transfer test probes throughout the baseline (A) and training (B) phases.

Dependent Variable and Interobserver Agreement

The dependent variable was the percentage of correct responding within six-trial blocks in each phase. Six-trial blocks were conducted to ensure that each stimulus was presented twice within a block. The percentage correct responding was determined by dividing the number of independent correct responses by the number of trials in a block and multiplying by 100. IOA was evaluated for 34% of the trials and was calculated by summing the number of trials in which two observers reported the same score, dividing by the total number of trials, and multiplying by 100. IOA was 96% for Joe (range 92%–100%) and 95% for Ben (range 90%–100%).

Baseline Phase

In the baseline phase, visual–visual and tactile–tactile reflexive responding were both tested. (For Ben, tactile–tactile reflexive responses were probed on two occasions to assess for stability in this phase.) To test for visual–visual responses, the experimenter held two objects (either identical or nonidentical) behind a screen. The experimenter then presented a sample stimulus (A1, A2, or A3), removed the stimulus, then presented the comparison stimulus (A1, A2, or A3) following a 3-s delay (A-3-A). Following the removal of the comparison stimulus, the experimenter asked, “Were those the same?” All trials were presented in a tabletop format with the pictures clearly visible. A correct response occurred if the stimuli were identical (A1–A1, A2–A2, A3–A3) and the participant responded by saying “yes,” or if the stimuli were nonidentical (e.g., A1–A3, A3–A2) and the participant responded by saying “no.” To test for tactile–tactile responses, the experimenter placed a sample object (A1, A2, or A3) inside the tactile sensory box and said, “Feel this item.” Once the participant touched the object for 5 s, the experimenter removed the item, placed the comparison object (A1, A2, or A3) in the box following a 3-s delay, and said, “Now, feel this item.” Once the participant touched the object for 5 s, the experimenter removed the sensory box and asked, “Were those the same?” Again, a correct response occurred if the stimuli were identical (A1–A1, A2–A2, A3–A3) and the participant responded by saying “yes,” or if the stimuli were nonidentical (e.g., A1–A3, A3–A2) and the participant responded by saying “no.”

Visual–Visual Training and Transfer Test Probes

The stimulus presentation for visual–visual and tactile–tactile responses was identical to the baseline phase in the training phase and during the transfer probes. In this phase, transfer probes were conducted following every three visual–visual training trial blocks. For visual–visual training blocks, participants were provided verbal praise, such as “Good job” or “That’s right,” when they demonstrated the correct response. When the participants demonstrated the incorrect response, the experimenter prompted the correct response verbally by saying, “No, try again.” No additional prompts were provided other than the verbal prompt after an error. Transfer test probes were identical to the baseline phase in that prompts and reinforcement were never provided contingent on participants’ responses to tactile–tactile stimulus presentations. The mastery criterion in this phase was at least five consecutive trial blocks with 100% correct responding.

Results

The results for Joe and Ben in Experiment 2 of the present study are summarized in Fig. 3. Baseline results for Joe demonstrate levels of correct responding for visual–visual probes that approximate results that would be expected by chance alone (i.e., 50%), with a mean percentage correct responding of 49.75%. The tactile–tactile test probe resulted in 66% correct responding. In the visual–visual training phase, Joe continued to demonstrate chance levels of correct responding for visual–visual stimuli for eight trial blocks, and test probes for tactile–tactile stimuli remained low, resulting in a mean percentage correct responding for the first two blocks of 58%. Throughout the remainder of the training phase, Joe’s correct responding for visual–visual stimuli increased, resulting in five consecutive trial blocks with 100% correct responding, thus achieving the mastery criterion. Additionally, Joe’s percentage correct responding for tactile–tactile stimuli increased to 100% on the third trial block and remained stable at 100% throughout the remainder of the training phase.

Fig. 3.

Percentage correct independent responses for Ben and Joe during the SCD arrangement with embedded probes of tactile–tactile discrimination

Baseline results for Ben also demonstrate levels of correct responding for visual–visual stimuli that would be expected by chance alone, resulting in a mean percentage correct responding of 39% in this phase. In addition, two test probes were conducted to evaluate the stability of Ben’s correct responding for tactile–tactile stimuli, resulting in percentage correct values of 33% and 50%. Although an increasing trend was observed, correct responding did not exceed results expected by chance alone. In the visual–visual training phase, Ben’s correct responding increased to 100% for visual–visual stimuli in the second trial block and remained stable at 100% for the remainder of the training phase, thus achieving the mastery criterion. Although Ben achieved the mastery criterion early in the phase, we continued the training phase to allow for at least two test probes to ensure stability in correct responding. Similarly, Ben’s percentage correct for tactile–tactile stimuli increased to 100% on the second trial block and remained stable across the two trials presented after the mastery criterion was achieved.

General Discussion

The results of the study suggest that an SMT arrangement can establish reflexive responding as a more complex generalized operant with children with severe and mild forms of autism. In both experiments, identity matching of a subset of visual–visual stimuli was directly reinforced, and the response transferred to untrained visual–visual stimuli. In the second experiment, the participants demonstrated a transfer not only to novel stimuli presented in the same sense mode (i.e., visual–visual) but also to the same stimuli contacted through touch (i.e., train visual–visual, test tactile–tactile). Prior research has shown that SMT procedures can effectively establish identity matching in animal models (e.g., Urcuioli, 2011), as well as equivalence class formation in typically developing adult participants (e.g., Dube, Green, & Serna, 1993; Lantaya et al., 2018). Anecdotally, establishing sequential reflexive responding within a larger equivalence-based curriculum is challenging for practitioners, and the data from Experiment 1 describe a procedure that can be effective with individuals with more severe forms of autism. Additionally, to our knowledge, this study is the first to demonstrate a cross-modal transfer of identity matching within an SMT arrangement and provides preliminary evidence that generalized reflexive responding can occur between sense modes in children with milder forms of autism. Although the transfer was only observed in terms of tactile–tactile conditional discriminations, there is no reason to assume that the same outcome could not be obtained in terms of gustatory–gustatory, olfactory–olfactory, or auditory–auditory reflexive relations. Prior research has demonstrated that cross-modal relations between arbitrarily applicable stimuli can emerge without direct reinforcement based on prior directly reinforced relationships in an equivalence paradigm (e.g., Dixon, Belisle, Stanley, Munoz, & Speelman, 2017), and the current demonstration of cross-modal transfers in the absence of any common relationship adds to a growing body of literature on the development of complex, multimodal reflexive responding.

The results reported here may additionally have several implications for language training with individuals with autism. Reflexive relational responding is likely the first higher order relational operant to develop, requiring participants to respond in terms of the identical properties of stimuli present at another location and/or at another time. This simple relation is needed to demonstrate stimulus equivalence, which has been put forward by several authors as the foundation for language development (Hayes et al., 2001; Horne & Lowe, 1996) that is delayed in individuals with autism (Rehfeldt & Barnes-Holmes, 2009). Prior research has demonstrated that reflexive responding can be established with children with autism in MTS arrangements (Slocum et al., 2012; Toussaint et al., 2017); however, advantages afforded by the use of an SMT arrangement can expand the complexity of untrained identity-matching responses demonstrated by children with autism.

Although the present study is inherently translational, occurring in a contrived experimental arrangement, potential applications of this approach could be used to teach children with autism to identify the same caregivers present at different times or to identify differences between objects when the objects are not both present in the environment. Delayed reflexive responding could also occur in several problem-solving tasks, such as when fixing an object, which requires making the object resemble itself before it was broken. Another example involves recognizing a person as the same person at two different points in time. Both of these potential extensions and others undoubtedly require further empirical evaluation, especially as to whether language training using an SMT approach while successively building complexity results in improvements in other socially valid skills. The transfer tasks across the two experiments additionally provide support for interpreting relational responding as a higher order operant set of response classes and for demonstrating its potential utility for participants with autism in this capacity (Barnes-Holmes & Barnes-Holmes, 2000; Luciano et al., 2007).

Finally, the procedures in the present study were adapted from the PEAK Relational Training System (Dixon, 2015) to aid in clinical replication. Translational protocols such as PEAK have only started to emerge that aim to develop language skills more broadly in individuals with autism, and require extensive empirical evaluation. Prior research on PEAK with this population has supported the technology in teaching a myriad of socially valid skills, including foundational perspective taking (Belisle, Dixon, Stanley, Munoz, & Daar, 2016), geometry (Dixon, Belisle, Stanley, Daar, & Williams, 2016a), metaphorical responding (Dixon, Belisle, Munoz, Stanley, & Rowsey, 2018a), and categorization (Dixon, Belisle, Stanley, Speelman, et al., 2016b). The present study is the first to evaluate the SMT procedure utilized in several PEAK programs and therefore extends the empirical support for this technology with individuals with autism.

Despite the results of the present study, they should be interpreted with caution due to several limitations. First, although DR was conducted across each of the four participants to establish matching and discriminated nonmatching responses, the delay-fading procedure was only implemented with two of the participants. Additionally, each of the assessed transfers of stimulus function was also only replicated across two of the participants. Although differentiating the procedures provides a degree of external validity in that procedures were individualized based on the autism severity of the participants, greater internal validity could be obtained by the inclusion of more participants in the multiple-baseline design in both experiments.

Second, although the procedures were effective in promoting the emergence of reflexive relational responding across the four participants, we do not know the degree to which conducting the experiments with the opposite participants would have been effective. For example, although DR + response fading was effective with Emily, and the transfer to novel picture stimuli was observed, we do not know if DR alone would have been effective, nor if the effect would transfer to tactile–tactile reflexive relations. Given the differences in the number of training sessions required in Experiment 1 and Experiment 2, it is unlikely that the two participants in Experiment 1 would have achieved the same outcome if they were exposed to Experiment 2’s instructional and experimental conditions. Future research may address this limitation by exposing participants to all experimental phases regardless of diagnosis and severity of language impairment.

Third, the results should be considered preliminary in Experiment 2 due to the baseline performance demonstrated by Joe and the rapid acquisition demonstrated by Ben. In Experiment 2, the number of baseline trial blocks was determined a priori; however, Joe demonstrated a score above chance levels on the final trial block. An additional trial block below chance levels would increase the certainty that this skill would not emerge due to maturation. Joe’s responding, however, immediately returned to chance levels at the beginning of the training phase, suggesting that this data point was likely due to chance, falling within the 10% range of what would be expected due to chance. Consistent responding above chance levels was not observed until the end of the training phase. For Ben, rapid rates of correct responding support the efficacy of this procedure; however, it is unclear what component parts of this complex skill were absent at baseline. Future research may attempt to ascertain how to assess reflexive SMT deficits in a way that contributes to rapid acquisition like that observed with Ben.

Fourth, procedural fidelity was not assessed to ensure that the independent variable was implemented as intended. Both experimenters had completed at least 4 hr of PEAK workshop training, as well as in situ behavioral skills training to conduct the procedures in the current study; however, evaluating procedural fidelity would increase the internal validity of the obtained results. We also did not evaluate the degree to which the instruction led to meaningful, socially valid changes in the participants’ behavior outside of the experimental arrangement. Therefore, the study should be considered strictly translational in nature.

Beyond addressing the limitations in the current study, future research may expand upon its results by further evaluating the emergence of reflexive responding and untrained transfers of stimulus function in an SMT arrangement, or by extending the translational scope of the procedures in terms of more socially valid behavioral phenomena. One potential extension could involve evaluating transfers of stimulus function in terms of the other sense modes. The SMT procedure, by imposing a delay between the presentation of the sample and comparison stimuli, may be especially amenable to gustatory and olfactory stimuli because the sensations associated with these sense modes (i.e., tastes and smells) tend to remain after the stimulus has been removed. Another potential avenue for future research could involve increasing the complexity of the conditional discrimination by presenting not only the discriminative stimulus “Were those the same?” but also the competing discriminative stimulus “Were those different?” This arrangement would require the participant to respond both in terms of matching or nonmatching properties of the sample and comparison stimuli and in terms of the contextual cues same and different, which would produce opposite answers given the same stimulus presentation. Finally, transfers of stimulus function that are more socially valid, such as identifying if a person is the same across contexts, may be evaluated to extend the utility of the SMT arrangement.

In summary, reflexive relational responding is required for the demonstration of stimulus equivalence, but prior research has almost exclusively evaluated reflexive responding using an MTS arrangement (Debert et al., 2009). An SMT arrangement provides an alternative to an MTS arrangement by sequentially presenting identical or nonidentical stimuli with a delay imposed between the sample stimulus and the comparison stimulus. The results of the present study suggest that DR with and without delay fading was effective in establishing consistent matching and discriminated nonmatching responding across four children with autism. In addition, two of the participants demonstrated a transfer of stimulus function to novel picture stimuli, and the other two participants demonstrated a transfer of function to the same stimuli presented through the tactile sense mode. Taken together, these results have implications for language training for children with autism who experience communication and relational learning deficits.

Data Availability

De-identified aggregate data will be made available upon request to the corresponding author.

Compliance with Ethical Standards

Conflict of Interest

The authors declare that they have no conflict of interest.

Ethical Approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed Consent

Informed consent was obtained from all individual participants included in the study.