Skip to content
Back to search results

ABA Design — ABA Reversal Design

Introduced by
Montrose Wolf, Donald Baer, Todd Risley (applied behavior analysis tradition)
Year
1968
Type
Single-subject experimental design
Family
Experimental design
Font size

What does it do?

The ABA design is a single-subject experimental design that demonstrates experimental control through three sequential phases: a baseline phase (A1), an intervention phase (B), and a return-to-baseline withdrawal phase (A2).

By removing the intervention in the final phase and observing whether behavior reverts toward baseline levels, researchers establish a functional relationship between the treatment and the target behavior for an individual participant.

The intuition behind the method

Imagine you want to show that a classroom reward system genuinely causes a student to stay on-task — not just that on-task behavior and the reward happened to co-occur. The ABA design provides that evidence by removing the reward after it appears to be working. If on-task behavior drops when the reward is withdrawn, you have compelling evidence that the reward, and not some coincidental factor, was responsible for the improvement. This reversal logic is the core of the design's power — and its ethical challenge.

History

The conceptual roots of reversal logic in single-subject research trace to B. F. Skinner's operant conditioning laboratory work in the 1950s and 1960s. The ABA (and ABAB) reversal design was formalized as a methodological standard with the founding of the Journal of Applied Behavior Analysis in 1968, anchored by Baer, Wolf, and Risley's seminal paper defining the dimensions of applied behavior analysis. Through the 1970s and 1980s, the design became the cornerstone methodology for evaluating behavioral interventions in special education, developmental disabilities, and clinical psychology.

When to use it

Use the ABA design when you need to demonstrate a functional relationship between a specific intervention and a target behavior in an individual participant, and when the behavior is expected to be reversible after the intervention is withdrawn. It is appropriate in applied behavior analysis, special education, rehabilitation, and clinical psychology when group designs are impractical (rare conditions, small N, individualized interventions). Do not use the ABA design when withdrawal of the intervention poses ethical risk — for example, when the target behavior is self-injury, academic skill acquisition (learned skills may not reverse), or any situation where withholding a beneficial treatment during A2 would harm the participant. In those cases, prefer the ABAB design or a multiple baseline design.

Limitations and considerations

  • Findings are specific to the individual participant; statistical generalization to a population requires systematic replication across multiple participants and settings.
  • The design is unsuitable when the target behavior is irreversible (e.g., acquired academic skills, motor learning) — a non-reversal in A2 confounds interpretation.
  • Ends in a withdrawal phase, meaning the participant concludes the study without the intervention active; this raises ethical concerns for beneficial treatments.
  • Visual analysis of time-series data requires training and can be subject to analyst bias; inter-rater reliability checks are necessary.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Provides clear, within-subject demonstration of a causal relationship without a separate control group.
  • Appropriate for studying rare behaviors or small populations where group experiments are infeasible.
  • Replication logic within a single participant controls for many maturation and history confounds.
  • Continuous measurement yields rich temporal data on behavioral dynamics, not just pre-post snapshots.
  • Low participant burden compared with multi-group randomized trials; each person serves as their own control.

Applications

  • Evaluating the effect of token economy systems on classroom on-task behavior in students with ADHD or autism spectrum disorder.
  • Assessing whether differential reinforcement procedures reduce stereotypic or self-stimulatory behavior in clinical settings.
  • Testing the impact of ergonomic workplace modifications on repetitive-strain injury behaviors in occupational health research.
  • Examining the functional effect of pharmacological treatments on behavior when dosing can be ethically suspended.
  • Demonstrating functional relationships in animal behavior research and comparative psychology laboratories.

How it works

How it works

1

The first A phase establishes a stable pre-intervention baseline. The target behavior is observed and recorded repeatedly across sessions (typically at least three to five data points) until performance is stable or trending in a direction opposite to the predicted intervention effect. This baseline characterizes the participant's natural rate of behavior and serves as the counterfactual reference.

2

The B phase introduces the independent variable — the treatment, intervention, or environmental manipulation under study. Behavioral data collection continues at the same measurement frequency. Researchers look for a clear change in level, trend, or variability in the target behavior relative to the A1 baseline, providing initial evidence of an effect.

3

The second A phase withdraws the intervention and returns to baseline conditions. If the target behavior reverts toward the A1 level, this replication within the same individual strengthens the inference that the behavior change was caused by the intervention rather than by extraneous factors. The degree of reversal is assessed visually and, where applicable, through effect-size statistics.

4

Data across all three phases are evaluated using visual analysis: changes in level (immediate shift in mean performance), trend (slope of the data path), and variability (scatter of data points around the trend line). The overall pattern — improvement in B, return toward baseline in A2 — constitutes the logical demonstration of a functional relationship. Effect sizes such as Tau-U or PND (percentage of non-overlapping data) may supplement visual inspection.

Common pitfalls

  • Withdrawing an effective intervention that addresses harmful behavior (e.g., self-injurious behavior) is ethically indefensible; always conduct an ethical review before committing to the A2 withdrawal.
  • Collecting too few data points per phase — fewer than three stable data points per phase makes it impossible to determine whether performance has truly stabilized, undermining the experimental logic.
  • Ignoring trend direction in baseline — if the A1 baseline already trends in the direction predicted for intervention, any apparent B-phase improvement may simply be a continuation of the pre-existing trend rather than an intervention effect.
  • Confusing ABA with ABAB — the ABA design ends after withdrawal and does not re-introduce the intervention; researchers who wish to end with the intervention active and strengthen replication evidence should use the ABAB design instead.

Frequently asked questions

What is the difference between the ABA and ABAB designs?

Both are reversal designs. The ABA design has three phases and ends with withdrawal of the intervention. The ABAB design adds a second B (intervention) phase, ending with the treatment active. ABAB provides stronger causal evidence through a second replication of the effect and is ethically preferable when the intervention is beneficial, because the participant ends the study receiving the treatment.

How do I know when the baseline is stable enough to proceed?

Common criteria require at least three consecutive data points with no systematic trend in the direction predicted for the intervention, and with variability (the range of scores) that is small enough not to obscure an anticipated effect. If baseline trends upward and you predict an upward intervention effect, wait for the trend to flatten or reverse before introducing B.

Can I analyze ABA data statistically instead of using visual analysis?

Yes, visual analysis is standard and required, but effect-size statistics can supplement it. Widely used metrics include Tau-U (which accounts for baseline trend), the percentage of non-overlapping data (PND), and the improvement rate difference (IRD). These metrics help quantify the magnitude of the effect and support comparisons across studies in meta-analyses.

Is the ABA design the same as an ABAB or a withdrawal design?

The terms reversal design and withdrawal design are often used interchangeably for ABA-family designs. Technically, a reversal involves actively reinstating baseline conditions (removing the reinforcer), while a withdrawal simply discontinues the intervention. In practice, most researchers treat ABA and withdrawal design as synonyms. An ABAB design is a specific extension that re-introduces the intervention after the second baseline.

Can I generalize the findings from an ABA study to other people?

Not directly. Single-subject designs establish a functional relationship for the individual studied. Generalization to other individuals and settings is achieved through systematic replication — repeating the study with different participants, therapists, or settings. A pattern of consistent effects across replications, often reported in literature reviews or meta-analyses, supports broader conclusions.

Related methods

Sources

  1. Baer, D. M., Wolf, M. M., & Risley, T. R. (1968). Some current dimensions of applied behavior analysis. Journal of Applied Behavior Analysis, 1(1), 91–97.

    Open DOI reference
    Copied.
  2. Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). Applied Behavior Analysis (3rd ed.). Pearson.

    ISBN: 978-0134752556
    Copied.