ABAB design

Each letter names a condition

The phases answer related questions:

1. A1, initial baseline or comparison: What pattern occurs before the independent variable is introduced?
2. B1, intervention: Does the outcome change when the defined intervention begins?
3. A2, return to baseline or withdrawal: Does the pattern move toward the comparison condition when the intervention is withdrawn?
4. B2, reintroduction: Does the intervention-related pattern appear again?

The letters describe conditions rather than guarantee their contents. “A” may be baseline, business as usual, or another defined comparison. “B” must specify the manipulable independent variable. The dependent variable needs an observable definition, measurement unit, opportunity rule, and time window.

The manifest starter, WWC Single-Case Design Technical Documentation, explains that valid inference in an ABAB design comes from structured phase repetition. An AB sequence has one comparison and cannot supply the same within-case replication.

Repeated phase changes support the causal test

A baseline pattern predicts what may continue without the intervention. A change after B1 challenges that prediction. Movement after the return to A and another change after B2 add verification and replication. In a simple ABAB sequence, the three adjacent transitions create three opportunities to demonstrate an effect at different times.

Phase labels alone establish no functional relation. History, maturation, measurement drift, procedural inconsistency, and coincident environmental changes can produce misleading patterns. The clinician should document exactly what changed at each transition and keep other conditions as consistent as feasible.

The current WWC Procedures and Standards Handbook is Version 5.0. For a simple reversal or withdrawal design, it requires at least two phases per condition to qualify for its higher research ratings. Its research-review thresholds include specific numbers of observations per phase. Those thresholds serve WWC evidence ratings, while clinical phase decisions also depend on stability, risk, benefit, feasibility, consent, and the planned decision rule.

Visual analysis examines the whole pattern

Graph every observation in time order, mark phase boundaries, and inspect:

• level: the typical magnitude within each phase
• trend: the direction and rate of change
• variability: how much observations fluctuate
• immediacy: how quickly the pattern changes after a transition
• overlap: how much adjacent-phase data occupy the same range
• consistency: whether similar A phases resemble each other and similar B phases resemble each other

Compare A1 with B1, B1 with A2, and A2 with B2, then compare A1 with A2 and B1 with B2. A systematic visual-analysis protocol paper describes those adjacent and same-condition comparisons for ABAB graphs. A summary percentage or phase mean can supplement the graph; it should preserve every observation and phase sequence.

Measurement and implementation need separate evidence

Define the outcome before collecting data. Train observers, calibrate on shared examples, and sample interobserver agreement across relevant phases when scores will drive decisions. Report agreement separately from the outcome. An ambiguous definition cannot be repaired by a high agreement percentage.

Define the intervention with enough precision to reproduce it. Measure procedural fidelity in A and B conditions, including whether the intervention was absent during A and present as planned during B. Record prompts, missing opportunities, environmental changes, treatment-integrity errors, and exclusions. An opportunity remains in its cohort when a scored implementation step was missed.

The BACB BCBA Test Content Outline, 6th edition covers single-case design features, interpretation, reversal and other designs, and application. It is examination content, not a clinical protocol or authority to practice.

Reversibility and ethics determine whether ABAB fits

ABAB is poorly suited to durable learning, irreversible outcomes, carryover effects, or any condition in which returning to A would create unacceptable risk or remove a necessary support. A taught communication skill may persist after teaching stops, so the second A phase may fail to reproduce the initial baseline even when teaching mattered.

Never withdraw AAC, food, water, bathroom access, mobility, prescribed care, pain care, emergency help, or an effective safety protection to complete a graph. Preserve ordinary supports across phases unless a qualified, lawful, and ethically reviewed plan identifies a safe comparison.

The current BACB Ethics Code applies to BCBA and BCaBA certificants and people who have completed an application. It addresses competence, client and stakeholder involvement, informed consent and assent when applicable, assessment, intervention, risk, data, and evaluation. A participant's withdrawal, distress, adverse effect, or meaningful benefit can require a phase change or termination even when the planned sequence remains incomplete.

When withdrawal lacks reversibility or acceptability, consider another design. Multiple-baseline, changing-criterion, or alternating-treatments arrangements answer different questions and carry their own requirements.

A fictional ABAB example

Riley is a fictional BCBA who voluntarily tests notification batching during a de-identified chart-audit simulation. No client records or clinical deadlines are involved. In A phases, ordinary work-chat alerts appear. In B phases, alerts are held until each ten-minute block ends. Riley may stop at any time. The outcome is checklist items completed correctly per block.

Across five A1 blocks, Riley completes 4, 5, 4, 5, and 4 items; the median is 4. In five B1 blocks, the values are 7, 8, 7, 8, and 7; the median is 7. A2 produces 5, 4, 5, 4, and 5, with a median of 5. B2 produces 8, 7, 8, 9, and 8, with a median of 8.

All three adjacent phase changes move in the predicted direction, and the two A phases and two B phases show similar patterns. That replicated pattern supports a functional relation in this simulation if measurement and fidelity evidence are adequate. It does not establish effects on real documentation, other workers, client outcomes, or longer periods. Riley's preference, errors, fatigue, and feasibility remain part of the decision.