Synthesizing the Multiple-Probe Experimental Design With the PEAK Relational Training System in Applied Settings

Abstract

The scientist-practitioner model necessitates embedding experimental designs within applied practice. This technical report describes a procedure for embedding a multiple-probe experimental design within the PEAK Relational Training System across all four PEAK modules. Baseline probes provide a direct test of target skills negatively endorsed within the PEAK assessment battery and can provide an estimate of skill acquisition in the absence of direct training. Temporal staggering of the probes maintains the fidelity of the experimental design and allows for the design to evolve along with learner skill acquisition. Achievement of mastery criteria demonstrates the efficacy of programming, and failure to achieve mastery can be remedied through programming adjustments that can be captured within the design. We additionally conducted a field test of the design with a child with disabilities, supporting the viability of this procedure within applied settings.

The experimental analysis of behavior is foundational to applied behavior analysis. Whereas distinctions may be drawn in other fields between those who conduct research and those who practice, the same cannot be said of applied behavior analysis, which generally follows a scientist-practitioner model (Barlow et al., 1984; Dixon et al., 2015; see also Critchfield & Reed, 2004). The scientist-practitioner model emphasizes the ongoing experimental analysis of events that are functionally related to the challenges experienced by people. The most extensively researched technology within applied behavior analysis exemplifies this approach—the experimental functional analysis (EFA; Iwata et al., 1994). An EFA typically utilizes a multielement design to ascertain the function of challenging behavior. Practitioners may then embed function-based treatment approaches within a number of designs to evaluate the efficacy of interventions developed from the EFA. For example, in a reversal design, the practitioner will temporarily withdraw the treatment to determine if the treatment is responsible for behavior change. When withdrawal is undesirable, practitioners may instead attempt a multiple-baseline design, where the treatment is introduced at another point in time with another client or targets another behavior with the same client. In both cases, addressing challenging behavior and experimental design logic are interwoven directly within the functional assessment and treatment of challenging behavior.

Research within many applied behavior-analytic journals is increasingly evaluating the efficacy of behavior-analytic technologies that address the language-learning deficits experienced by children with autism (Belisle, Paliliunas, et al., 2020; O’Connor et al., 2017). One technology that has emerged from this research, and has more recently contributed heavily to this research, is the Promoting the Emergence of Advanced Knowledge (PEAK) Relational Training System (Dixon et al., 2017). PEAK is an assessment and curriculum designed to promote the development of generative language skills in children with autism or related disabilities. PEAK is composed of four modules: The Direct Training and Generalization modules (PEAK-DT and PEAK-G, respectively) were developed from Skinner’s verbal operant theory (Dixon, 2014a, 2014b; Skinner, 1957) and emphasize the development of verbal operants through contingency learning and generalization. The Equivalence and Transformation modules (PEAK-E and PEAK-T, respectively) extend upon this approach by incorporating advances in stimulus equivalence theory (Dixon, 2015; Sidman & Tailby, 1982) and relational frame theory (Dixon, 2016; Hayes et al., 2001). These latter two modules emphasize the development of derived relational responding as a generalized operant, as well as changes in behavior resultant from transformations of stimulus function. Research on PEAK has begun to support the reliability and validity of the assessments, as well as the efficacy of the curriculum (Dixon et al., 2017), exceeding considerably the research supporting other tools that serve a similar function (Ackley et al., 2019).

Within the scientist-practitioner model, however, behavior analysts cannot rely exclusively on this research to support their use of PEAK or any other technologies with a given client. Peer-reviewed research should be used to guide practitioners toward potentially efficacious technologies, but to be analytic, practitioners must use experimental design strategies to evaluate whether the treatment as conducted is achieving the targeted socially significant behavior change and, where success is not achieved, to continually adjust intervention components or attempt new interventions altogether. The experimental logic of the multiple-probe design (MPD; Horner & Baer, 1978) may be uniquely situated for use by practitioners implementing PEAK, allowing for the ongoing analysis of interventions’ efficacy within daily practice. The MPD represents a synthesis of the multiple-baseline design and probe designs and allows for (a) an initial baseline test of performance, (b) baseline tests after a criterion is reached within other steps, (c) another baseline test prior to the introduction of the independent variable, and (d) ongoing testing throughout the intervention. The design is capable of affirming that, without the intervention, improvements in performance would not have occurred (i.e., failure to reach the mastery criterion in baseline); that the intervention was efficacious in achieving the mastery criterion; and that mastery of one step does not necessarily lead to mastery of other steps within the MPD. These design characteristics of the MPD are uniquely situated to allow for an ongoing analysis of the efficacy of PEAK instruction.

Each PEAK module contains a criterion-referenced list of 184 target skills (736 total items). Selection of appropriate targets should be based on direct preassessment results and direct or indirect testing of each of the target skills (Dixon, 2016). Belisle, McDonald, et al. (2020) described a bias-informed selection algorithm as one way to combine these sources of data in a systematic way to identify potential programs for individualized programming. Regardless of the selection method, direct testing of potential programs may be preferable to exclusive indirect testing (Witts, 2018). When conducting an MPD across skills, the initial probe can serve as a direct test of the target skill. For example, if the target skill is tacting animals, the probe can evaluate the percentage of correct tacts given pictures of common animals. If the learner achieves a mastery criterion (e.g., 90% correct responding) during this initial probe, practitioners can progress to testing more complex programs. If the learner fails to achieve this criterion, a second probe may be conducted at a later time immediately preceding the introduction of PEAK training for that target skill. This may occur once the learner has mastered a different, earlier skill. One advantage of this approach is that skill acquisition of the target program may result from the mastery of earlier skills. Given research suggesting verbal operant behavior may be interdependent (see Fryling, 2017), such that learning a tact can result in the untrained emergence of a mand, mastering one skill as a function of learning other skills may be likely throughout instruction. The baseline probe structure of the MPD can save practitioners considerable time that may have been spent training skills that were either earlier or later in training. The PEAK-G, PEAK-E, and PEAK-T modules contain not only skills that are directly reinforced but also skills that are tested throughout instruction to evaluate generalization (PEAK-G), derived relational responding (PEAK-E and PEAK-T), and transformations of stimulus function (PEAK-T). Therefore, the probe design additionally allows practitioners to embed test probes (a) within baseline testing and (b) following mastery of the trained targets. When mastery of either trained or tested targets is not achieved, practitioners may adjust programming systematically to ensure mastery of all target skills, and the design logic further allows for adjustments to remain constant across all subsequent steps of the MPD.

In this technical report, we provide a detailed overview of how practitioners can embed an MPD across-skills experimental design in the context of PEAK instruction. This is implemented following the selection of programs. The design is intended to be highly flexible to fit within any applied context, allowing clinical decision making to affect the design, rather than the design to constrain clinical decision making. The number of steps or target skills within the MPD is theoretically infinite, as new steps are added as additional programs are introduced within the curriculum for a given client. Another feature of the design is that it allows for baseline testing of new programs throughout the implementation of the curriculum. In this way, target skills that are endorsed as absent within direct and indirect assessments are directly tested within trial blocks at two points in time before implementing PEAK training. After describing the general features of this design, we provide field-test data of the MPD within a special education program with a child participant. These data provide a test and demonstration of the design in a context similar to that in which this design is likely to be implemented.

Implementing the Multiple Probe

The progression of the MPD is shown in Figure 1. For the sake of description, we assume that practitioners are implementing eight programs concurrently (i.e., two programs from each module); however, this number can be adjusted based on learner characteristics and contextual considerations. For example, if a learner demonstrated a high score on the PEAK Comprehensive Assessment (PEAK-CA) and several hours are devoted to language and cognitive training, programming could include 10 or 15 concurrent programs. Conversely, if a learner demonstrated a low score on the PEAK-CA and fewer hours are devoted to language and cognitive training, programming may only include two or four concurrent programs. Another factor may be the inclusion of all modules or the exclusion of some modules, as may occur given a variety of factors beyond the scope of the present article. In the present example, we assume that practitioners are implementing all four modules concurrently, as recommended within PEAK (Dixon et al., 2016). Regardless of the idiosyncrasies that are necessary for individualizing instruction guided by PEAK, the MPD should be sufficiently flexible to allow for such individualization while maintaining experimental design logic.

Figure 1.

Process flow diagram of the multiple probe experimental design across skills within a single PEAK module (e.g., Direct Training). Note. (a) Develop the curriculum structure and conduct the assessment battery. (b) Test initial programs and continuously test upcoming programs throughout the training of target programs. The multiple-probe design creates a loop where new programs are continuously being tested while the learner progresses throughout the curriculum

For an overview of program selection, we refer readers to the introduction chapters of the PEAK modules (Dixon, 2014a, 2014b, 2015, 2016). Programs are selected using a combination of direct and indirect assessment methods that identify items or skills that are negatively endorsed (i.e., skills that the learner has not yet mastered), which provides targets for curricular programming. We also refer readers to the bias-informed selection algorithm developed by Belisle, McDonald, et al. (2020), which describes one selection method that may allow for more precise endorsement for the PEAK-DT and PEAK-G modules. We provide a brief overview of this method in the sections that follow; we used this method to select PEAK-DT and PEAK-G programs for field-testing of the MPD. Because the MPD allows for the direct testing of all items or skills that were negatively endorsed within the PEAK assessment battery, we encourage readers to utilize program selection methods that, when errors occur, are most likely to produce false negatives over false positives. A false negative occurs when a skill is endorsed as absent when the learner in fact can demonstrate the skill. A false positive occurs when a skill is present when the learner in fact cannot demonstrate the skill. If false negatives occur, these will be identified within the baseline probes when the learner achieves the mastery criterion (e.g., 90% independent correct responding or greater) within the trial block prior to training. However, false positives will not be directly assessed and could serve as prerequisite skills for later programs, which would inhibit learning. Whereas false negatives may delay the training of some items, false positives could adversely impact the efficacy of PEAK training if prerequisite skills are never targeted. The following steps are completed after initial skills for targeted instruction have been identified.

Initial Baseline Testing

Each program within the MPD is tested twice within the baseline phase to determine whether the target skill is already mastered prior to training. The first test serves as the initial direct test of the program, and the second test is used to determine if the participant mastered the program either due to maturation or due to the participant mastering earlier programs. We expect this to occur given that one purpose of the PEAK curriculum is to promote emergent learning, or the acquisition of new skills in the absence of direct contingency training (Dixon et al., 2016). The first step is to determine the number of programs from each module that will be trained simultaneously (n, e.g., n = 2 within each module; n × 4 modules = 8 total programs). In Figure 1, this is part of developing the curricular structure for a given learner, and the number of programs will vary across learners and contexts. The number of programs that should be tested initially in baseline within each module should be equal to n + 1, where n is the number of training programs. For our example, we are targeting two programs from each module simultaneously. Therefore, the number of programs that should be initially tested is equal to 3 for each module, or 12 if all four modules are being implemented simultaneously (total test = 2 + 1 = 3; 3 × 4 modules = 12 total tested programs). If the learner demonstrates mastery of any of these initial programs, the program should be scored as mastered within the 184-item assessment, and the next program should be directly tested. By following this method, the result will be the identification of three initial programs that are not yet mastered, thus providing initial skill targets.

Training and Continuous Baseline Testing

Readers will notice that the number of target programs within each module exceeds the number of training targets by one. This is a critical feature of the MPD. The next step is to initiate training on the first n programs. In our example, this is the first two programs within each module. The distinction between training and baseline is that participant responses are contingently reinforced (correct responses) or prompted (incorrect responses). It is important to note that for the PEAK-G, PEAK-E, and PEAK-T programs, some targets are directly trained and others are tested throughout training. These include generalization targets (PEAK-G), derived relational responding targets (PEAK-E and PEAK-T), and transformation of stimulus function targets (PEAK-T). Implementers can choose to test targets intermittently within training or delay testing until the training targets are mastered. Both approaches are supportable, and empirical research may begin to vet which approach leads to faster acquisition of both training and testing trials. In either case, when improved performance is not observed for tested targets, implementers may attempt several strategies to promote stimulus or response generalization, derived relational responding, or transformations of stimulus functions. Strategies are described in the PEAK modules (e.g., Dixon, 2016), as well as in several studies with children with and without disabilities (see Belisle, Paliliunas, et al., 2020).

Any changes in programming to address failures to acquire trained or tested targets can be indicated with a phase change line. If the adjustment is made for a single program, then the phase change line should only appear within that program. If, however, changes are made to all programs at the same time, such as adjusting the reinforcement strategy, this can be indicated with a vertical phase change line that passes through all steps of the MPD. This strategy is shown in the following sections when we embedded a token-economy reinforcement system within all on-going PEAK programs with our pilot participant. This is indicated within PEAK as “levels,” where Level 1 represents initial targets, Level 2 represents additional targets, and subsequent levels can be added to ensure the participant has sufficiently mastered exemplars of the target skill. The number of levels is left to the discretion of the implementer. When new levels are added, this, too, can be denoted by adding a phase change line. Indicating new levels is important because performance will generally decrease initially with the addition of untrained targets. We encourage readers to consider conducting a baseline probe prior to the introduction of the next level to ensure that the participant did not master new targets as a function of learning earlier targets.

Once the participant masters a program within a module, the implementer conducts the second baseline test probe for the remaining program. In our example, this would be the second test of the third program. At the same time, the implementer conducts the first baseline for another program. In our example, this would be the first test of the fourth program. As in the initial baseline tests, if the first or second probe indicates that the skill is mastered, this should be indicated on the 184-item list and the next program should be tested. Implementers continue this procedure throughout PEAK instruction, where each time a target program is mastered, the second test probe is conducted for one program and the first test probe is conducted for another program. By following this method, implementers will achieve a stepwise progression of baseline testing that contains at least two probes. One potential limitation of this procedure is that an increasing trend may occur from the first baseline probe to the second baseline probe, where conventional baseline logic requires that baseline performance indicate stability or a decreasing trend for an accelerative target such as skill acquisition. Some implementers may choose to accept this limitation in order to more quickly progress to training. That is, although an increasing trend was observed, the skill has not yet been mastered and therefore could benefit from targeted training. Other implementers may consider running additional probes if an increasing trend is observed to ensure stability prior to progressing to the training phase. An intentional element of the MPD is that the design can be fit into practical contexts to allow implementers to show experimental control while maximizing the learner’s access to training guided by PEAK.

Field-Testing the MPD

We have described an adaptation of an MPD that could be conducted within a variety of applied contexts implementing PEAK and that could be sufficiently individualized to fit the needs of any learner. This field test provides a real-world example of how to conduct the MPD in a context that includes rapid testing of new skills, other educational demands, missed days, and many other factors that can impede implementers from embedding experimental designs within daily programming. We believe that our experimental designs should be adjustable to fit the demands of the real world, rather than adjust the real world to fit within our experimental design.

Participant and Setting

Colton was an 8-year-old male with medical diagnoses of attention-deficit/hyperactivity disorder and autism. An evaluation conducted 6 months prior to the study indicated he had average intelligence with weaker working-memory and processing-speed abilities. At the time of the study, Colton participated in applied behavior analysis programming in an intensive specialized program within a public school for 20 min per day, 5 days per week, and he participated in the regular education setting for the remainder of his day with support. Sessions were conducted in a separate classroom within the specialized program in a one-on-one format with only the therapist and Colton present. The number of programs implemented within a given day varied, and a classroom-wide token economy was introduced early in the school semester with all students throughout instruction. Both of these events can be seen in the MPD results in the Figs. 2, 3, and 4.

Figure 2.

Multiple-Probe Design Across Skills 1. Note. The first phase included baseline probes of all train and test stimuli and relations. Filled data points indicate stimuli or relations that were directly reinforced in the training phases. Data paths indicate where training occurred. The gray phase line shows the introduction of a classroom-wide token-economy reinforcement strategy. Legends are provided for PEAK-G, PEAK-E, and PEAK-T programs. Cul = Cultural Relation, Arb = Arbitrary Relation, Com = Comparison Relation

Figure 3.

Multiple-Probe Design Across Skills 2. Note. The first phase included baseline probes of all train and test stimuli and relations. Filled data points indicate stimuli or relations that were directly reinforced in the training phases. Data paths indicate where training occurred. The gray phase line shows the introduction of a classroom-wide token-economy reinforcement strategy. Legends are provided for PEAK-G, PEAK-E, and PEAK-T programs. Cul = Cultural Relation, Arb = Arbitrary Relation, Com = Comparison Relation

Figure 4.

Multiple-Probe Design Across Skills 3. Note. The first phase included baseline probes of all train and test stimuli and relations. Filled data points indicate stimuli or relations that were directly reinforced in the training phases. Data paths indicate where training occurred. The gray phase line shows the introduction of a classroom-wide token-economy reinforcement strategy. Legends are provided for PEAK-G, PEAK-E, and PEAK-T programs. Cul = Cultural Relation, Arb = Arbitrary Relation, Com = Comparison Relation

Procedure

First, we selected the initial programs for baseline testing. To select programs, we first conducted the PEAK-CA (Dixon, 2019), which contains the preassessments for all four PEAK modules and instructions for program selection. Because of the potential interdependency of the verbal operants (REFs), we additionally had two independent raters indirectly complete the 184-item indirect PEAK Direct Training and PEAK Generalization assessments. The PEAK-CA contains 64 items for PEAK-DT (35% of all PEAK-DT assessment items) and 64 items for PEAK-G (35% of all PEAK-G assessment items). Results from the PEAK-CA, the PEAK-DT indirect assessments from both raters, and the PEAK-G indirect assessments from both raters were synthesized using the bias-informed selection algorithm described by a bias informed selection algorithm was used. The algorithm considers item-by-item agreement between both raters and the PEAK-CA and uses the estimated bias of raters to make decisions on items where there is no alignment between the PEAK-CA and the 184-item assessments (65% of all items). For the PEAK Equivalence and PEAK Transformation programs, we selected programs using the instructions contained within the PEAK-CA. Unlike the verbal operants, simpler forms of relational responding (e.g., symmetry, mutual entailment) are likely necessary for the development of more complex forms of relational responding (e.g., transitivity and equivalence, combinatorial entailment). Therefore, performance on the PEAK-CA may provide a sufficient measure of the relational operants measured within each factor to select programs.

We conducted the MPD as described in the previous section for the identified target skills. Our goal was to implement training for 2 programs concurrently from each module for a total of 12 programs. Therefore, we conducted initial baseline testing until three programs from each module were identified on which Colton achieved a score of 80% or below independent correct responding. For the PEAK-G, PEAK-E, and PEAK-T programs, this criterion was set both for skills that would be directly trained and skills that were tested for generalization, derivation, or transformation. Each PEAK-DT program contained 10 targets that were directly trained, each PEAK-G program contained 5 targets that were directly trained and 5 that were tested, each PEAK-E program contained 5 classes that included trained and tested relations, and each PEAK-T program contained 5 classes that included trained and tested relations and tests for transformations of stimulus function. Depending on the nature of the program and the timing within the semester, a second level of stimuli was introduced, which included the same number of new stimuli. Therefore, mastery of any PEAK-DT or PEAK-G program required mastery of 20 skills (10 trained and 10 tested for PEAK-G), and mastery of any PEAK-E or PEAK-T program required mastery of 20–40 relations (e.g., 10 trained and 10 tested).

Results

The MPD for Colton is shown in Figures 2, 3 and 4, and the major components of the design are indicated in Figure 5. Over the course of approximately 3 months, we directly tested 39 items/programs. Of those skills, 19 were mastered within the initial probe, 5 were not mastered during the initial probe but were mastered during the second probe, and 15 were not mastered during either baseline probe. Those 15 programs that were not mastered during either probe are shown in the figures, and included 4 PEAK-DT programs, 3 PEAK-G programs, 4 PEAK-E programs, and 4 PEAK-T programs. Figure 5 shows all data stacked together with markers of the three major components of the design. The first and second probes are demarcated by the numbers 1 and 2. By conducting these probes in sequence, the practitioner can determine if mastering other skills was sufficient to allow for untrained mastery of a new skill, or if further instruction is still required. The individual skills are sequenced in the figure based on the date during which training was introduced (2). In this way, although the spacing of the initial probes (1) appears scattered, the stepwise sequence of the programs is ordered. As can be seen in the figure, the MPD was able to stagger the introduction of the programs, producing a stepwise introduction of training. This stepwise staggering is a necessary element of the baseline logic of the MPD, reviewed previously. Altogether, there were 31 individual stimulus sets or relational sets. PEAK-G programs contained train stimuli and test stimuli (i.e., two sets of stimuli); PEAK-E programs for Colton were symmetry programs that contained trained relations (A-B) and test relations (B-A; i.e., two sets); and PEAK-T programs contained train, test, and transformation relations (i.e., three to four sets). Recall that each set contained 5–10 individual targets. If we assume the median number of targets within each set is 7.5, we may estimate that the total number of targets directly tested within baseline was 232.5 (31 × 7.5). Of the 31 sets tested in baseline, 17 showed an increasing trend from the initial baseline test probe to the second baseline test probe. Again, implementers may elect to conduct additional probes to obtain stability; however, due to applying the MPD within a real-world applied context, we elected to commence training immediately if this second probe was below the 90% criterion. A total of 11 sets showed a decreasing trend, and 3 did not change. One potentially interesting source of data produced by the MPD is the proportion of sets where an increasing trend was observed. In this example, the greatest number of sets showed an increasing trend (i.e., 17, or 54.8%), and five programs were mastered during baseline between the first and second probes. This outcome may provide evidence that conducting the curriculum in general leads to improvements in items even when those items are not directly targeted, as would be expected within a curriculum that targets generative language and cognitive learning processes.

Figure 5.

Multiple-Probe Design Across Skills. Note. Numbers indicate the three major considerations of this design: 1 indicates the initial baseline probe, 2 indicates the second baseline probe, and 3 indicates the terminal training probe for Level 1 stimuli corresponding with the mastery criterion set for a given program. Cul = Cultural Relation, Arb = Arbitrary Relation, Com = Comparison Relation

The training phase of the MPD showed that PEAK training was efficacious in developing several target skills, such as several listener and speaker tact repertoires, generalized imitation and play, derived mands and translations, and initial variants of comparative relational responding like faster/slower and bigger/smaller. Achievement of the mastery criterion is shown in Figure 5 demarcated by the number 3. This is a critical element of the design because achieving mastery of a skill is the goal of any PEAK program and also serves to signal the need to test the next program in cue (i.e., a program that has been initially tested and requires the second test). We set a mastery criterion of three consecutive trial blocks with 90% or greater independent correct responding in the present study with some variation in individual programs as needed to meet the needs of the learner in context. Ideally, any practical research design should be flexible to meet the needs of an individualized programming decision, as was achieved in this field test. We see that time to master each individual program varied considerably. For example, counting fluency required almost 2 months of training to obtain the mastery criterion, whereas derived manding and derived math symbols were mastered within a number of days. There are several potential explanations for this variability that are outside the scope of the current article; however, one thing to note is that the temporal loci of trial blocks varied across programs. Again, we attempted to embed the MPD within a real-world applied setting, so this variability provides a demonstration of the ability of this design to accommodate differences in implementation that may emerge within applied settings. Visual analysis of the training data also suggests that, even though newer programs that were introduced later were more complex, the rate of mastery appeared to improve. Although direct empirical testing is needed with multiple participants and greater experimental control, this outcome does suggest the MPD may be capable of capturing “learning how to learn” by comparing rates of mastery throughout the progression of PEAK instruction over time.

The solid gray line shows where a school-wide token system was introduced. This was introduced following staff training on effective reinforcement strategies in schools. Because the introduction of the token system could have affected rates of program mastery, we embedded the phase line vertically through all programs. We encourage implementers of the MPD to do this, as it could elucidate confounds that may explain acceleration or deceleration of learning rates. In theory, any such programming-strategy change could be denoted using this method, which is important given that changes may be frequent in applied settings, not all of which are directly linked to momentary level and trend changes in the data. We also introduced Level 2 stimulus sets for 9 of the 15 target skills (22 sets, or approximately 165 additional targets). We introduced the second level if the skill was mastered within the first 2.5 months. New levels were not introduced after this time due to the proximity to the end of the semester for the learner. Colton achieved the mastery criterion on five of the nine programs within the initial trial blocks, so additional training was not conducted. For the four that were not initially mastered, training was again efficacious in promoting the acquisition of the target skills. Altogether, again assuming 7.5 stimuli or relations within each set and 1–4 sets within each program, Colton mastered approximately 397.5 targets contained within 15 programs (232.5 + 165). Baseline testing additionally supported mastery of 34 programs; however, we do not know if these programs would have been mastered in the absence of PEAK instruction (i.e., false negatives on the assessment) or were mastered without direct training because of the acquisition of other skills (i.e., emergent learning). Additional design considerations may be needed to further vet this difference, but this is again outside of the scope of the present report. What Colton’s data do show, however, is that the PEAK assessment approach was able to identify appropriate targets (true negatives) for training and that the training strategies were efficacious in promoting skill development. Additionally, the data may provide some evidence of “learning how to learn” in terms of increasing trends evident within successive baseline probes and faster mastery rates throughout the intervention. The purpose of the design is not to discover new elements of derived relational responding, but rather to provide a simple way to embed experimental design logic within existing practice to support the efficacy of PEAK instruction or to refine instruction when acquisition is not immediately forthcoming.

Conclusion

In conclusion, although behavior analysts are adept at embedding experimental designs within treatment approaches to challenging behavior, strategies are needed to address language and cognitive challenges. PEAK is one assessment to treatment strategy that has obtained considerable empirical support (Dixon et al., 2017). Although empirical support exists for this curriculum, practitioners will still benefit from ensuring that programming remains efficacious at the level of the single subject. Reviewing existing research is important to ensure that any program is likely to work; however, experimental validation should be pursued once a program is implemented to ensure that the program is, in fact, working. As well, when a program is not working as desired, to allow for adjustments to programming consistent with existing research and knowledge of behavior change principles. The present design is a simple way to structure baseline testing to achieve the experimental logic of an MPD and may be especially useful for practitioners utilizing PEAK or similar curricular programming. This technical article overviewed how an MPD can be embedded within regular instruction guided by PEAK to demonstrate treatment effectiveness and to determine if modifications are needed to regular programming at the level of the single subject. We field-tested the design with a child with disabilities within a special education setting, where the MPD was capable of demonstrating training effectiveness along with some evidence of faster rates of skill acquisition with or without direct training. We hope that this report aids practitioners in using experimental design logic to support and improve services for individuals with disabilities.

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 (Missouri State University, FWA: 00004733) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Funding

This project was funded by research awards developed by Project Alpha and Pender Public School (B02743-132022-73004-011) as part of an effort to infuse applied behavior-analytic research within applied settings.