Single-Subject Experimental Designs: Reversal, Multiple Baseline, Alternating Treatments & Changing Criterion

Reversal design in ABA is a single-case experimental design that measures behavior during baseline, introduces an intervention, and withdraws or reverses it to test whether behavior changes back. ABA and ABAB patterns let the participant serve as their own comparison.

It can demonstrate experimental control when behavior should change with treatment and safely return when treatment is withdrawn. It is not appropriate for every case; consider risk, carryover, ethics, feasibility, and safer designs.

What Is Reversal Design in ABA?

In a reversal design, the analyst changes the independent variable across phases and observes the dependent variable. A baseline A shows behavior before treatment; B introduces treatment; a reversal returns to the earlier condition; a second B may replicate the effect.

Reversal describes the condition change, not an exact return to the original level. Partial return, delay, carryover, learning, or maturation may occur. Strength comes from repeated, time-linked changes consistent with planned conditions.

Element	Meaning	Example
A phase	Baseline or comparison condition.	Data are collected with the current routine and no programmed token intervention.
B phase	Intervention condition.	A token system is implemented according to the written protocol.
Reversal	Return to the earlier condition or withdrawal of the intervention.	The token system is removed while other conditions are held as constant as possible.
Replication	Reintroduction or repeated test of the intervention effect.	The token system is reinstated to see whether behavior changes again.

The BACB BCBA Test Content Outline, 6th edition includes reversal, multiple-baseline, multielement, and changing-criterion designs. On an exam, identify the phase sequence, changing variable, and repeated effect.

How ABA, BAB, ABAB, and ABABA Patterns Work

The letters describe conditions in sequence. A is often baseline and B treatment, but the study defines the labels. ABA is baseline, treatment, return; BAB starts with treatment, withdraws it, and returns; ABAB adds a second treatment phase for replication.

ABABA adds another withdrawal and reintroduction. Do not memorize abbreviations in isolation: recognize repeated changes to the same independent variable and corresponding behavior changes.

Sequence	Phase logic	What the analyst is testing
ABA	Baseline, intervention, return to baseline.	Whether behavior changes with treatment and moves back when treatment is removed.
BAB	Intervention, withdrawal or comparison, intervention.	Whether the behavior changes when the intervention is absent and returns when it is reinstated.
ABAB	Baseline, intervention, baseline, intervention.	Whether the effect is replicated across two treatment introductions.
ABABA	Two reversals and three treatment introductions.	Whether repeated alternation produces a consistent pattern, when ethically and practically justified.

A reversal need not return to a risky natural baseline. The comparison may be a defined safe control, and withdrawal may stop early when risk is unacceptable. Tie phase labels to the protocol, not the graph alone.

What Makes a Reversal Design Experimental?

A reversal design is experimental because the analyst systematically manipulates an independent variable and evaluates a measured dependent variable. Planned timing, repeated data, and withdrawal or change test whether behavior shifts with the condition.

Control is stronger when the dependent variable changes soon after phase shifts, remains stable within phases, and replicates on reintroduction. High variability, poor timing, unknown fidelity, or coincident events weaken it.

Feature	Strong pattern	Interpretive caution
Manipulation	The active intervention is clearly introduced and withdrawn.	Several components change together, so the active variable is unclear.
Measurement	The same operational definition is used across phases.	Observers or measurement rules change between conditions.
Phase stability	Data show a reasonably interpretable level, trend, and variability before the next phase.	The phase changes before the baseline or treatment pattern can be evaluated.
Replication	Behavior changes again when the intervention returns.	Only one treatment introduction is observed.
Alternative explanations	Concurrent events are identified and controlled as much as possible.	Staff, schedule, setting, or reinforcement changes at the same time.

Repeated measurement is essential; pre- and post-treatment scores cannot show a phase-by-phase relation. Define phase duration, stability criteria, and when safety overrides experimental goals in advance.

How to Read an ABAB Graph

At each phase-change line, compare data before and after the condition changes. Examine level, trend, variability, immediacy, overlap, and consistency. A convincing effect shifts with B, back when B is withdrawn, and again when B returns.

Do not reduce the graph to “up is bad” or “down is good.” Direction depends on the target and goal: lower may be desirable for reduction, higher for skill acquisition. Read the variable, dimension, and goal together.

Graph feature	Question to ask	Why it matters
Level	Did the typical value shift after the phase change?	Shows the size and direction of a possible effect.
Trend	Was the behavior already moving before treatment?	A preexisting trend can mimic or obscure an intervention effect.
Variability	Are the points tightly grouped or widely scattered?	High variability makes phase comparisons less certain.
Immediacy	How quickly did the data change after the condition changed?	A rapid, repeated shift supports a functional relation.
Overlap	Do data ranges overlap substantially across phases?	Less overlap can make the phase distinction easier to see, but it is not the only criterion.
Replication	Does the pattern repeat when B returns?	Repeated demonstrations strengthen the experimental interpretation.

See the ABAB graph and Excel guide for graphing practice. Repeated changes after introducing and withdrawing the same intervention point to reversal design.

Reversal Design Versus Other Single-Case Designs

Reversal is one of several single-case designs. Match it to whether withdrawal is possible, multiple participants or settings are available, behavior changes quickly, and effects are reversible. The phase pattern alone is not enough.

Design	Defining pattern	When it is often useful
Reversal	One or more baselines alternate with intervention phases.	The effect is expected to reverse and withdrawal is acceptable.
Multiple baseline	The same intervention is staggered across behaviors, settings, or participants.	Withdrawal would be unsafe, unethical, or likely to produce a lasting skill change.
Multielement or alternating treatments	Two or more conditions are rapidly alternated.	Conditions can be discriminated and quick comparisons are feasible.
Changing criterion	The performance criterion changes step by step.	Behavior should move gradually toward successive target levels.

Multiple baseline does not withdraw treatment from an already treated baseline. Multielement rapidly alternates conditions. Changing criterion moves the target in planned steps. See the changing-criterion guide when the graph shows successive standards.

When Reversal Design Is a Poor Choice

Withdrawal may be poor when behavior is dangerous, treatment protects health or dignity, or learning is expected to last. Removing an effective communication or safety skill may be unacceptable; multiple baseline or changing criterion can provide control without withdrawal.

Carryover may prevent a newly taught skill from returning to baseline even when treatment worked. Extinction history, delayed learning, or context changes can also complicate the return phase. Lack of reversal does not prove lack of effect.

Concern	Why it weakens a reversal design	Possible design response
Safety risk	Withdrawal could expose the learner or others to serious harm.	Consider multiple baseline, a safe comparison condition, or an early-stop rule.
Irreversible learning	New skills may persist after treatment is withdrawn.	Use staggered introduction or another nonwithdrawal design.
Slow behavior change	The response may not change quickly enough for phase comparisons.	Plan longer phases or select a design suited to delayed effects.
Carryover between conditions	The prior condition continues influencing the next phase.	Assess carryover, increase separation, or use alternating treatments carefully.
High context change	Setting or staff changes make phase differences ambiguous.	Standardize procedures and document unavoidable events.

Ethics are part of design selection. Ask whether the question can be answered with less risk, whether assent and communication are respected, and whether monitoring and stopping criteria exist. The best design demonstrates control while protecting the participant.

How to Plan and Implement a Reversal Design

Define the dependent variable in observable terms and select frequency, rate, duration, latency, percentage, or another suitable dimension. Define each condition so a trained observer can identify A or B without vague labels.

Collect enough baseline to understand the pattern, then introduce treatment with fidelity checks. Change only the planned independent variable when practical, and document schedule, medication, staff, illness, or consequence changes that could explain the result.

Planning step	Decision	Documentation
Define the question	What causal relation should the design test?	Independent variable, dependent variable, and prediction.
Define conditions	What exactly happens in A and B?	Written phase protocols and examples.
Set phase rules	What indicates enough stability to change phase?	Duration, stability, safety, and early-stop criteria.
Plan measurement	How will data be collected consistently?	Recording form, observer instructions, and reliability plan.
Plan ethics	What risks arise during withdrawal?	Consent or assent process, safeguards, monitoring, and review.
Plan replication	How will the effect be demonstrated again?	Return condition, predicted pattern, and decision rule.

ABAB labels do not replace a protocol. If B contains several components, the design may show the package works without identifying which part is necessary. Component analysis may follow when safe. Fidelity and observer agreement remain essential.

Data Collection and Treatment Integrity

Record the dependent variable often enough to reveal phase changes and trends. Collect fidelity data for B and confirm the planned comparison in A. If treatment is withdrawn, record its extent rather than trusting the label.

When possible, collect interobserver agreement or another reliability measure. Review data in real time for safety and quality, including generalization, new problems, and ethical justification for reversal.

Data stream	What to monitor	Interpretation benefit
Dependent variable	Repeated target measures using the same definition.	Shows level, trend, variability, and phase changes.
Phase fidelity	Whether A and B conditions were delivered as written.	Confirms the planned independent variable was manipulated.
Observer agreement	Agreement between trained observers when appropriate.	Supports confidence in the measurement system.
Context events	Staffing, schedule, setting, health, and unusual events.	Identifies plausible competing explanations.
Participant impact	Safety, distress, assent signals, and quality-of-life indicators.	Ensures experimental control does not override ethical care.

Interpret the entire data path, not only final B. Improvement in B that remains low in A may reflect carryover or another event. Delayed effects require attention to expected response latency; visual analysis and clinical judgment work together.

BCBA Exam Traps and Ethical Boundaries

“Reversal” is not simply any treatment stop. Look for a systematic phase sequence and repeated tests of the independent variable. BAB can be reversal without starting at baseline; ABAB adds replication when interpretable and ethical.

Exam wording	Likely concept	Reasoning move
“Treatment is introduced, withdrawn, and reintroduced.”	Reversal design	Map the phases and identify the repeated manipulation.
“The intervention begins at different times across settings.”	Multiple baseline	Look for staggered baselines without withdrawal.
“Two conditions alternate rapidly across sessions.”	Multielement or alternating treatments	Look for quick condition comparisons.
“The goal changes from 5 minutes to 10, then 15.”	Changing criterion	Identify successive performance standards.
“Withdrawal would expose the client to dangerous behavior.”	Ethical design-selection issue	Consider a nonwithdrawal design or safe comparison.
“The graph changes, but fidelity is low.”	Interpretive limitation	Do not overstate experimental control when implementation is uncertain.

Do not confuse ABA design notation with an ABA intervention acronym. In design notation A and B label conditions; in clinical writing the letters may mean antecedent, behavior, and consequence. Use the context and graph.

See the site’s ABA withdrawal-design guide, component-analysis guide, and free BCBA mock exam. Match the answer to the defining pattern and ethical facts.

Reversal Design Checklist and FAQ

Use this compact checklist before naming a design or interpreting an ABAB graph:

1. Define the dependent variable and measurement dimension.
2. Write the exact procedures for each condition.
3. Identify whether the same independent variable is repeatedly introduced and withdrawn.
4. Check level, trend, variability, immediacy, overlap, and replication.
5. Review treatment integrity, observer agreement, and contextual events.
6. Ask whether withdrawal is safe, ethical, and likely to be interpretable.
7. Choose a multiple-baseline, multielement, or changing-criterion design when reversal is not appropriate.
8. Separate phase labels from the actual procedures.
9. Check carryover before interpreting a weak return phase.
10. Document stopping rules and participant protections.

Question	Short answer
What is reversal design in ABA?	A single-case design that introduces, withdraws, and often reintroduces an intervention to test a functional relation.
Is ABA the only reversal sequence?	No. ABA, BAB, ABAB, and longer patterns can all use reversal logic when the conditions are systematically changed.
Does reversal require behavior to return exactly to baseline?	No. Return toward the comparison pattern is informative, but carryover and delayed effects can limit reversal.
When is multiple baseline preferable?	When withdrawal is unsafe, unethical, or likely to leave a lasting skill that will not reverse.
What is the biggest interpretation mistake?	Calling any treatment stop a reversal design without confirming systematic phase changes and repeated experimental control.

In one sentence: reversal tests whether behavior changes with an intervention, changes back when it is withdrawn, and changes again when it returns. Design selection must balance control, safety, dignity, and measurement quality.

Related BCBA Study Guides

• Comparative, Component, and Parametric Analysis in ABA: A BCBA Exam Guide