Single-Case Design in Education

Because the experimental logic comes from replication of the effect within the design, not from a separate control group. By repeatedly showing that the outcome changes only when the intervention is applied — reversing it (ABAB) or staggering it across behaviors, settings, or people (multiple baseline) — the design rules out coincidence and maturation as explanations. The What Works Clearinghouse and Horner et al. standards typically require at least three such demonstrations at different times to credit experimental control.

In a reversal design, the intervention is introduced, withdrawn, and reintroduced, demonstrating control by showing the behavior tracks the presence and absence of the intervention. This requires effects that can reverse. A multiple-baseline design instead introduces the intervention at staggered times across several behaviors, settings, or participants, demonstrating control by showing each one changes only when the intervention reaches it. Multiple baseline is preferred when reversal is undesirable or impossible, as with newly learned skills.

Traditionally, primarily by systematic visual analysis of graphed data, judging changes in level, trend, variability, immediacy of effect, and data overlap across phases, using documented decision rules. Increasingly this is supplemented by quantitative single-case effect sizes (such as Tau-U and other nonoverlap indices) and by multilevel or regression models for combining cases and meta-analysis. Best practice uses visual analysis as the primary judgment with effect sizes as a complement, rather than either alone.