AB Design — AB Single-Subject Experimental Design
- Introduced by
- Murray Sidman; Baer, Wolf & Risley
- Year
- 1960s
- Type
- Single-subject experimental design
- Family
- Experimental design
What does it do?
The AB design is the simplest single-subject experimental design, consisting of two sequential phases: a baseline phase (A) in which the target behavior is observed under natural conditions without intervention, followed by an intervention phase (B) in which the treatment or manipulation is introduced.
Changes in the behavior's level, trend, or variability between phases are used to infer the effect of the intervention on the individual participant.
The intuition behind the method
Think of the AB design as a before-and-after comparison for a single person. You first watch and record what a student, patient, or client does without any intervention — this is the baseline. Then you introduce your treatment and keep recording the same behavior. If the behavior changes noticeably when — and only when — you introduce the treatment, you have evidence that the treatment may be working. The key word is 'may': because there is no withdrawal or return-to-baseline phase, you cannot definitively rule out other explanations for the change. The AB design is often the entry point for single-subject research and the building block from which stronger designs (ABA, ABAB) are constructed.
History
Single-subject experimental methodology was systematized by Murray Sidman in his 1960 text Tactics of Scientific Research, which established within-subject replication as an alternative to group designs. The two-phase AB logic was formalized within applied behavior analysis following the landmark 1968 article by Baer, Wolf, and Risley in the inaugural issue of the Journal of Applied Behavior Analysis. The AB design remains the simplest expression of this tradition and the scaffold on which all more complex single-case designs — ABA, ABAB, multiple baseline — are built.
When to use it
Use the AB design when you need a practical, low-resource single-subject framework to document the association between an intervention and behavior change in an individual — common in applied behavior analysis, special education, clinical psychology, and rehabilitation. It is appropriate when a reversal (withdrawing the treatment) is ethically or practically impossible, or when the goal is preliminary documentation rather than strong experimental control. Do not use the AB design when you need causal evidence: without a return-to-baseline or replication phase, the design cannot rule out history, maturation, or regression to the mean. When internal validity is the priority, upgrade to an ABA or ABAB design.
Limitations and considerations
- Weak internal validity: the absence of a withdrawal or replication phase means coincidental events (history) cannot be ruled out as the cause of the observed change.
- Results apply only to the individual studied; generalization to other people, settings, or behaviors requires systematic replication across multiple cases.
- Dependent on a stable, predictable baseline — variable or trending baselines make post-intervention change uninterpretable.
- The design cannot detect delayed treatment effects if behavior change lags behind the phase transition by multiple sessions.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Maximally simple to implement — requires only two phases and can be conducted by a single practitioner with one participant.
- Ethically feasible when withdrawing or reversing a presumably beneficial treatment would be inappropriate.
- Provides an individualized picture of behavior change that group-design averages would obscure.
- Serves as the foundational building block from which stronger withdrawal and multiple-baseline designs are constructed.
- Visual analysis of time-series data makes findings interpretable to practitioners without advanced statistical training.
Applications
- Documenting the effect of a reading fluency intervention on an individual student's words read correctly per minute in special education.
- Monitoring the impact of a behavioral management strategy on frequency of self-injurious behavior in a clinical setting.
- Tracking pain intensity or functional mobility before and after a physical therapy protocol in rehabilitation research.
- Assessing the effect of a workplace accommodation on the productivity or error rate of an individual employee.
- Evaluating the preliminary impact of a pharmacological or dietary change on a measurable behavioral or physiological outcome in a single patient.
- Piloting a new classroom instructional technique with one student before scaling to a group intervention.
How it works
How it works
Operationally define the dependent variable — the specific, observable, and measurable behavior or outcome you will track (e.g., number of words read correctly per minute, frequency of disruptive behaviors per session). Clear operationalization ensures consistent measurement across both phases.
Collect repeated measurements of the target behavior over multiple sessions before any intervention is introduced (the A phase). Baseline data establish the pre-intervention trend and variability. A stable, predictable baseline — at least three to five data points showing a clear level and trend — is required before moving to the intervention phase; an upward-trending baseline in the desired direction makes it nearly impossible to attribute subsequent improvement to the treatment.
At a clearly defined session boundary, introduce the treatment or manipulation while keeping all other conditions constant. Document precisely when and how the intervention begins so that the phase transition is unambiguous in the data record. Avoid making any concurrent procedural changes that would confound the phase shift.
Continue measuring the same target behavior under the same conditions using the same data-collection method throughout the B phase. Consistency in measurement across phases is critical: any procedural deviation other than the intervention itself constitutes a confound that undermines interpretation.
Plot all data points on a time-series graph with a phase-change line marking the A-to-B transition. Evaluate changes in level (mean of data per phase), trend (slope direction), and variability (spread around the trend). A clear, immediate shift in level or trend coinciding with the phase change provides initial evidence of an intervention effect; formal statistical supplements such as Tau-U may also be applied.
Common pitfalls
- Proceeding with an unstable baseline: if baseline data are highly variable or already trending in the desired direction, any improvement during the B phase cannot be attributed to the intervention — wait for stability before introducing the treatment.
- Too few data points: a minimum of three to five data points per phase is the conventional floor; fewer points make trends indiscernible and visual analysis unreliable.
- Confounding the phase transition: changing measurement procedures, settings, or observers at the same time the intervention is introduced creates an uninterpretable confound.
- Overclaiming causality: the AB design yields correlational, not causal, evidence; conclusions should be framed as 'associated with change' rather than 'caused by the intervention'.
Frequently asked questions
Why is the AB design considered weak compared to ABA or ABAB designs?
The AB design shows that behavior changed after the intervention was introduced, but it cannot demonstrate that the intervention caused the change. Without withdrawing the treatment and observing whether behavior returns to baseline — as in an ABA design — or replicating that reversal — as in ABAB — alternative explanations such as coincidental events, maturation, or spontaneous recovery cannot be ruled out. Stronger designs provide the within-subject replication that establishes experimental control.
How many data points do I need in each phase?
The conventional minimum is three to five data points per phase, though more is better. The practical criterion is stability: collect data until the level and trend are predictable enough that you can project where data would go if nothing changed. For the intervention phase, continue until the effect or lack of effect is clear and the series has stabilized.
Can I use statistical tests with AB data?
Yes. Visual analysis is the primary and traditional approach, but several statistical supplements are used: Tau-U accounts for baseline trend and provides an effect-size estimate; piecewise regression (interrupted time-series analysis) models level and slope changes at the phase boundary. These complement visual analysis and require at least eight to ten data points per phase to be meaningful.
When is the AB design ethically preferable over a stronger reversal design?
When the intervention is expected to produce irreversible learning (e.g., acquiring a new academic skill) or when withdrawing a beneficial treatment would be harmful or unethical, the reversal required by ABA or ABAB designs is inappropriate. In these situations the AB design — or a multiple-baseline design, which avoids withdrawal entirely — is the ethical choice.
Is the AB design the same as a pre-post design?
No. A pre-post design typically involves a single measurement before and one after an intervention, often across a group. The AB design involves repeated, frequent measurements over time within a single participant, producing a time-series that reveals trends and variability — information a single pre and post measurement cannot provide.
Related methods
Sources
Sidman, M. (1960). Tactics of Scientific Research: Evaluating Experimental Data in Psychology. Basic Books.
Open publication referenceCopied.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 referenceCopied.