REVERSAL DESIGN

Reversal Design (A-B-A/A-B-A-B Design)

Primary Disciplinary Field(s): Applied Behavior Analysis; Experimental Psychology; Single-Case Research Methodology

1. Core Definition and Methodology

The Reversal Design is a powerful type of single-case experimental design used primarily within experimental psychology and applied behavioral analysis to demonstrate a causal, or functional, relationship between an independent variable (treatment or intervention) and a dependent variable (target behavior). This methodology achieves experimental control by the systematic introduction and subsequent withdrawal of the intervention condition, fluctuating between baseline limitations and treatment conditions across phases, thereby functioning as an internal check on causality. It stands as a vital exploratory arrangement specifically designed to contravene confusing impressions of continuity, classification, and analysis that might arise in less rigorous sequential designs or complex arrangements like Latin squares by requiring the behavioral effect to repeatedly turn on and off with the introduction and removal of the treatment.

The most basic iteration of this methodology is the A-B-A design, where ‘A’ represents the baseline phase (measurement of behavior without the intervention) and ‘B’ represents the intervention phase. The critical element is the return to baseline (the second ‘A’), or the reversal phase. If the target behavior changes reliably when the intervention is introduced (A to B) and reliably returns toward baseline levels when the intervention is withdrawn (B to A), a strong argument for experimental control and a functional relationship can be established. This sequence ensures that the observed effect is due to the manipulation of the independent variable and not to confounding variables such as historical events or maturation.

Due to ethical concerns regarding the final withdrawal of a demonstrably effective treatment, the A-B-A-B variation is far more prevalent, particularly in clinical and applied settings. The A-B-A-B sequence provides an extra opportunity to verify the functional relationship through a second introduction of the treatment (B), while also concluding the experiment with the participant benefiting from the effective intervention. This design extension addresses the primary ethical weakness of the simple A-B-A structure.

Ultimately, the design aims to yield counterbalanced estimations of one condition versus another, allowing researchers to observe and quantify the specific magnitude of change attributable solely to the intervention. The systematic, cyclical nature of condition presentation acts as a built-in replication mechanism, strengthening the visual and analytical evidence of cause and effect within the individual subject.

2. Underlying Principles and Rationale

The fundamental rationale underlying the Reversal Design lies in its commitment to strong internal validity. By utilizing the participant as their own control, the design effectively controls for a wide range of subject-specific threats to validity that plague group designs, such as individual differences in aptitude, learning history, or environmental exposure. The core principle dictates that if the intervention (B) is truly responsible for the change, then the behavior should predictably change (improve or decline) every time the condition shifts from A to B or B to A, respectively.

The design provides robust evidence for a functional relationship, which is defined as a relationship where the dependent variable (behavior) changes systematically in the presence of, or contingent upon, changes in the independent variable (intervention). For a functional relationship to be established via a reversal design, the change must be immediate, substantial, and repeatable across phase changes. The demonstration of this contingency across multiple reversals serves as the primary mechanism for rejecting the hypothesis that the behavioral change was coincidental or caused by an uncontrolled third variable.

Reversal designs are uniquely suited for examining idiosyncratic effects, focusing on the intensive study of a single subject’s behavioral processes. Unlike nomothetic research, which seeks general laws across populations, single-case research seeks idiographic understanding. The detailed, time-series data collected during the reversal design captures moment-to-moment variability and responsiveness that would be lost in the averaged data produced by traditional group comparisons. This sensitivity allows practitioners to tailor interventions based on an individual’s specific reaction profile to the treatment variable.

3. Phases of the Reversal Design (Detailed Breakdown)

• Phase A1 (Initial Baseline): This foundational phase involves systematically measuring the frequency, duration, or magnitude of the target behavior prior to the introduction of any experimental intervention. Data collection continues until a stable data path is achieved. Stability is critical, meaning that there is minimal variability and no significant trend (either ascending or descending), ensuring that the subsequent effects of the intervention can be clearly differentiated from pre-existing natural variation.
• Phase B1 (Intervention): Once stability is confirmed in A1, the intervention (the independent variable) is introduced. Measurement of the target behavior continues under the new conditions. For experimental control to be demonstrated, the data path must show an immediate and significant change in level, trend, or both. The magnitude and direction of this change provide the first piece of evidence that the intervention is having an effect.
• Phase A2 (Reversal/Withdrawal): This is the pivotal phase. The intervention is completely removed, and conditions are returned exactly to those of A1. If the behavior returns to or near the levels observed in A1, the hypothesis that B caused the change is strongly supported. If the behavior maintains the improvements seen in B1, the intervention is deemed to have produced a therapeutic effect (a desirable outcome for the client) but the design fails to demonstrate experimental control because the effect did not reverse.
• Phase B2 (Reinstatement): In the A-B-A-B variant, the intervention is reintroduced following the verified reversal (A2). The purpose of B2 is twofold: ethically, it provides the participant with the effective treatment they need; experimentally, it serves as a final, confirming replication of the functional relationship, expecting the behavior to once again change in the desired direction and return to B1 levels.

4. Establishing Experimental Control and Internal Validity

The Reversal Design is arguably the strongest single-case methodology for establishing internal validity because it demands three distinct, highly correlated behavioral shifts that must correspond precisely to the planned manipulations of the independent variable. Each complete cycle (A-B and B-A) acts as its own internal replication, meaning the researcher does not need to rely on statistical generalization across subjects but rather on repeated demonstrations of the effect within the same subject over time. This high level of control allows for precise statements of causality regarding the individual participant’s behavior.

The design effectively manages several common threats to internal validity. For instance, the threat of maturation (the subject naturally changing over time) is largely ruled out because maturation effects typically occur slowly and linearly; they would not produce the sharp, immediate, and reversible changes necessary to validate the design. Similarly, while history (external events) remains a threat, the likelihood of an external event precisely coinciding with the introduction, withdrawal, and reintroduction of the treatment across three or four phases is statistically minimal.

A key component of establishing control is the concept of withdrawal versus reversal. While the terms are often used interchangeably, true experimental control is strongest when the withdrawal of the treatment causes an actual *reversal* of the behavior toward baseline levels. If the intervention’s effect is maintained after removal—such as the subject continuing to exhibit a newly learned skill—the design fails to prove that the intervention is necessary for the maintenance of the behavior, thus weakening the demonstration of control, even if the outcome is clinically successful.

Furthermore, the integrity of the design relies on the ability of the behavior to be sensitive to the environmental conditions. If the behavior is heavily influenced by factors that cannot be completely controlled or reversed (e.g., physiological changes or long-term drug effects), the design’s power to demonstrate functional control is diminished. Therefore, researchers must meticulously ensure that all relevant variables are successfully reversed during the A2 phase to isolate the effect of the B condition.

5. Variations of the Design

While the A-B-A-B structure is standard, researchers employ various modifications to suit different research questions, ethical constraints, and behavioral contexts. One prominent variation is the B-A-B Design, which is employed when immediate intervention is deemed necessary, such as when dealing with severe problem behaviors, and delaying treatment for a stable baseline is inappropriate. This design starts immediately with the intervention (B), reverses to baseline (A), and concludes with treatment (B). While ethically advantageous in high-risk scenarios, its demonstration of control is slightly weaker than the A-B-A-B because it lacks the initial baseline comparison necessary to rule out potential reactive effects of the initial measurement process.

Another variation is the use of multiple intervention comparisons, such as the A-B-C-B Design, where ‘C’ represents a second distinct intervention, often used to compare the efficacy of two different treatments against the same baseline measure. This complex design allows the researcher to determine which intervention is superior, provided both B and C are functionally linked to the behavior and that the behavior reverses when B or C is withdrawn (or when switching between them). Similarly, alternating treatment designs sometimes incorporate a reversal phase (e.g., A-B-A-C-A) to confirm that the differences observed between B and C are indeed due to the interventions and not environmental drifts.

For research requiring extremely high statistical confidence or dealing with highly variable baseline data, researchers may implement multiple replications, resulting in sequences like A-B-A-B-A-B. While increasing the certainty of the functional relationship, such extensive designs significantly increase the overall duration of the study and heighten the ethical exposure of the participant to repeated reversals, making them less common in standard clinical practice unless absolutely essential for establishing scientific rigor.

6. Analytical Techniques and Data Interpretation

Data analysis in Reversal Designs relies predominantly on visual inspection, a hallmark of single-case research methodology. Researchers meticulously graph the dependent variable across sessions and phase lines, observing for three critical visual indicators of an experimental effect: changes in level, trend, and variability. A strong functional relationship is demonstrated when there is an immediate and substantial change in the data’s level or trend upon phase transition (e.g., A to B), and the behavior reliably returns to the previous phase’s characteristics upon reversal (B to A).

Criteria for establishing a causal link through visual analysis are strictly defined. Firstly, there must be a clear and non-overlapping difference between the data points in the A phase and the B phase. Secondly, the trend of the data must change direction or slope when the intervention is introduced, and this trend must reverse when the intervention is withdrawn. Thirdly, variability should ideally remain low within each phase, ensuring that the changes observed are true effects of the intervention rather than random fluctuation. High variability obscures the visual determination of the functional relationship.

While visual analysis is the primary standard, researchers sometimes utilize statistical adjuncts to supplement interpretation, particularly when visual inspection is ambiguous due to moderate variability or marginal effects. These methods include calculating non-parametric statistics, determining percentage of non-overlapping data (PND), or employing advanced time-series analysis techniques. However, in the realm of Applied Behavior Analysis, the philosophical preference remains rooted in visual evidence that is sufficiently clear to influence clinical practice directly.

7. Ethical and Practical Limitations

The most significant limitation of the Reversal Design is its inherent ethical concern. The requirement to withdraw an effective treatment (A2 phase) means that if the intervention successfully reduced a dangerous or problematic behavior, the researcher is obligated to temporarily restore the conditions that allow the problematic behavior to return. In cases involving self-injurious behavior, aggression, or other severe behaviors, the reversal phase is often ethically indefensible, leading many practitioners to choose alternative designs, such as the Multiple Baseline Design, which does not require withdrawal of the intervention.

A critical practical limitation arises when dealing with irreversible effects. If the target behavior involves the acquisition of a skill (e.g., academic mastery, learning a complex motor skill), the learning gained during the B phase may be permanent. If the behavior does not revert to baseline upon withdrawal of the intervention, the design cannot demonstrate experimental control, even if the learning is beneficial. In such scenarios, the Reversal Design is inappropriate for establishing causality, and other single-case methods must be employed.

Furthermore, the design may suffer from sequence effects, whereby the experience in Phase B affects the subject’s behavior in Phase A2, even after the intervention is withdrawn. While the design is intended to mitigate sequence effects better than simpler designs, it cannot eliminate them entirely. If the subject anticipates the return of the intervention or if the B condition creates permanent changes in the environment or the organism, the required reversal may be partial or absent. Finally, the repeated introduction and withdrawal of the treatment can sometimes negatively impact social validity, creating frustration or confusion for the participant, who may struggle with the instability of their behavioral contingencies.

8. Further Reading

• Reversal Design (Wikipedia)
• Applied Behavior Analysis (ABA)
• Single-Case Experimental Designs
• Internal Validity