Clarity Before Causality: Understanding and Articulating Approaches to SCD

• In Event: Analysis and Meta-Analysis of Single-Case Experimental Data: Methodological Developments

Abstract

As argued by Johnson and Cook (2020), there are two fundamentally distinct approaches to single-case design (SCD): inductive/dynamic and deductive/static. These two approaches represent completely different goals and methodologies in SCD. The inductive/dynamic approach is in many ways synonymous with applied behavior analysis (ABA), as SCD is leveraged for the potentially unnamed purpose of investigating those contingencies which control animal behavior and then to isolate their influence via systematic manipulation of the stimuli which interact with that behavior. Critically, if an effect is not observed, this is assumed to be due to the incorrect identification of those controlling variables, and as such, further manipulation is considered necessary and often undertaken as a continuation of the same experimental design to identify those variables which do in fact control the target behavior. In contrast, the deductive/static approach is focused on determining whether the implementation of an independent variable can be inferred to have caused change in a dependent variable. A study is designed which will allow for the testing of a causal claim (i.e., implementation of this token economy will decrease disruptive behavior), and then the results are used to inform whether such a causal relationship is present or not.
We contend that a major point of contention in SCD, and indeed a fundamental roadblock to progress in the methodology, is the unnamed nature of these assumptions. Members of those disparate fields which utilize a common methodology in SCD are not actively naming whether they are engaging in work to identify those contingencies which influence a target behavior, or if they are instead working to construct and then test a causal claim. Although seemingly self-evident, such unnamed assumptions have significant implications for what is a defensible methodological decision.
We would further argue that, although not named as such, assumptions like the necessity of a stable baseline and the value of response-guided implementation decisions are directly derived from an inductive/dynamic viewpoint which lacks much in common with a deductive/static approach. We further argue that when a deductive/static approach is utilized, complex causal claims of the type articulated by Shadish et al. (2007) can be utilized alongside the logic of the counterfactual to engage in single-case work which does not require a stable baseline nor response-guided intervention decisions. Indeed, we argue that such requirements fundamentally hinder the ability to engage in defensible causal inference. Additionally, we contend that research conducted using an inductive/dynamic approach should not be considered eligible for consideration in the identification of evidence-based practices, insofar as we identify evidence-based practices as discrete interventions and not overarching theoretical approaches (e.g., the effectiveness of a token economy, not the effectiveness of ABA).
In this poster presentation, we illustrate these issues through the selection and description of case studies in SCD. We demonstrate extant research which exemplifies each of the two major approaches to single-case design, unpack their implications for the production of evidence, and illustrate an alternative approach to conceptualizing and undertaking SCD which clearly states its underlying assumptions.

Authors

• Austin H. Johnson, University of California - Riverside

• Daniel Swan, University of Oregon

• Maggin Daniel, University of Illinois at Chicago