Association for Behavior Analysis International

Differential reinforcement of other behavior (DRO) is a strategy commonly used to address challenging behavior but can be difficult to implement with high treatment integrity. While it is ideal that behavioral treatments are implemented with perfect treatment integrity, previous research has demonstrated that treatment effects can be observed even with occasional treatment integrity errors. However, the impact of various degrees of treatment integrity on outcomes of DRO is still unclear. The purpose of the current study was to conduct a parametric analysis of treatment integrity level (100%, 80%, 60%, 40%, 20%) in DRO in a human operant arrangement using a multiple treatments reversal design. Participants engaged with a computer program in which mouse-clicks on a moving button on the screen served as the target response and a proxy for challenging behavior. Rate of target responding was compared across baseline, DRO with 100% treatment integrity, and during reduced-integrity DRO (i.e., 80%., 60%, 40%, and 20% treatment integrity). Results indicate that DRO may be robust despite some level of treatment integrity error. These findings will inform future evaluations of treatment integrity errors in DRO and may inform mediator training practices.

Differential Reinforcement of Alternative Behavior (DRA) effectively reduces challenging behavior by reinforcing an alternative behavior. However, its efficacy can be compromised if implemented with errors. With other procedures (e.g., RIRD, DRO), interspersing high- and low-fidelity sessions has reduced negative effects of fidelity errors, but this has not yet been demonstrated with DRA. Therefore, the purpose of the present study was to identify how experimental design (phase-wide errors or interspersed high- and low-fidelity) impacted responding when fidelity was reduced. To address this question, we used a within-subject translational model. Undergraduate students engaged in arbitrary responses on a computer-based task that were analogous to challenging and appropriate behavior. Participants experienced three conditions: baseline, DRA 100%, and DRA 50%. During the phase-wide errors, participants experienced an extended period of DRA 100% followed by an extended period of DRA 50%. During the interspersed high- and low-fidelity phase, participants rapidly alternated between the DRA 100% and DRA 50% conditions. Fidelity errors had differential effects based on the experimental design. In some cases, DRA was still effective when implemented with interspersed errors.

A limitation of the parametric literature regarding procedural errors is that individual stimulus sets or tasks are typically implemented with one level of fidelity throughout a study. Currently, no studies have investigated how implementing procedures with varying levels of fidelity affects learning. In addition, a small body of research indicates that a history of errors may impact future learning (e.g., Bergmann et al., 2021; Hirst & DiGennaro Reed, 2015). In Experiment 1, we varied the levels of procedural fidelity across one of three stimulus groups during a matching-to-sample task. Subsequently, we implemented a 100% fidelity phase for the varying group of stimuli to evaluate how initially exposing learners to varying levels of fidelity affects ongoing performance once errors are corrected. In Experiment 2, we systematically replicated Experiment 1 and included an incentive component to further investigate how including a cumulative point system and the opportunity to obtain additional reinforcers would affect learning. Overall, the results showed that the varying levels with which procedural errors are introduced can be detrimental to learning. However, when errors are corrected, individuals can master the task.