single-subject-design

name	single-subject-design
description	Use when selecting and implementing single-subject experimental designs including reversal, multiple baseline, alternating treatments, and changing criterion for behavior analytic research and clinical evaluation.

Single-Subject Experimental Design

Single-subject designs (also called single-case experimental designs) are the primary research methodology in applied behavior analysis. Each participant serves as their own control, and experimental control is demonstrated through within-subject replication of effects.

Baseline Logic

All single-subject designs rest on three elements of baseline logic (Sidman, 1960):

1. Prediction: Baseline data predict what behavior would look like without intervention. A stable baseline allows a confident prediction.
2. Verification: Demonstrate that the predicted pattern would have continued in the absence of the IV. Achieved through reversal to baseline conditions or continued baseline in untreated tiers.
3. Replication: Repeat the effect at different points in time to rule out coincidence. A minimum of three demonstrations of effect is the standard for establishing a functional relationship.

Reversal/Withdrawal Designs

A-B-A-B Design

The standard reversal design.

• A1 (Baseline): Collect baseline until stable.
• B1 (Intervention): Introduce the IV. Continue until effect is clear.
• A2 (Withdrawal): Remove the IV. Behavior should return toward baseline levels (verification).
• B2 (Reintervention): Reintroduce the IV. Behavior should change again (replication).

Variations

• B-A-B: Begin with treatment when withholding treatment is unethical (e.g., SIB). Weaker design—lacks the initial prediction from baseline.
• A-B-A-B-C-B-C: Compare two treatments (B and C) against baseline. Useful for treatment comparison.
• A-B-A-B with follow-up: Add maintenance probes after the final intervention phase.

When to Use

• The behavior is expected to be reversible (behavior will return to baseline when the IV is withdrawn).
• Ethical to withdraw treatment temporarily (no safety concerns during withdrawal).
• Only one primary target behavior to evaluate.

When NOT to Use

• Behavior is not reversible (learned skills that cannot be unlearned).
• Withdrawal poses safety risks (severe SIB, dangerous aggression).
• Carryover effects are expected (skill acquisition, awareness changes).

Multiple Baseline Designs

Stagger the introduction of the IV across two or more tiers while collecting concurrent baseline data.

Across Participants

Same behavior and intervention applied to different participants at staggered times.

Across Behaviors

Same participant, different behaviors treated with the same intervention at staggered times.

Across Settings

Same participant and behavior, intervention introduced in different settings at staggered times.

Design Requirements

• Minimum of 3 tiers (preferably 4–5 for stronger evidence).
• Begin baseline simultaneously across all tiers.
• Introduce the IV to the first tier only after baseline is stable.
• Monitor all remaining tiers during baseline—their data must remain stable while untreated.
• Introduce the IV to successive tiers only after the previous tier demonstrates a clear effect.

Strengths

• Does not require withdrawal of treatment (no return to baseline).
• Appropriate for irreversible behaviors (skill acquisition).
• Can demonstrate generality across participants, behaviors, or settings.

Threats to Validity

• Covariation: If untreated tiers change when the IV is introduced to the first tier, experimental control is compromised. This may indicate the tiers are not independent.
• Insufficient stagger: If the IV is introduced to all tiers at similar times, the design approximates an A-B design without replication.

Alternating Treatments Design (Multi-Element)

Rapidly alternate two or more conditions within a single phase.

Procedure

1. Define the conditions to compare (e.g., Treatment A vs. Treatment B, or Treatment vs. Control).
2. Randomly assign or counterbalance conditions across sessions (e.g., alternating days, morning vs. afternoon).
3. Plot data for each condition on the same graph using different data paths.
4. Compare the data paths: clear separation (differentiation) between conditions indicates differential effects.

Counterbalancing

Control for sequence and time-of-day effects by ensuring each condition is equally represented across all positions.

When to Use

• Comparing two or more treatments for the same behavior.
• Need to evaluate treatment effects quickly (without extended baselines).
• Behavior must be capable of rapid change between conditions (no long carryover).

Evidence of Effect

• Clear and consistent separation between data paths.
• Minimal overlap between conditions.
• Stable data paths within each condition.

Changing Criterion Design

Demonstrate experimental control by showing that behavior changes incrementally to match a stepwise criterion.

Procedure

1. Collect baseline.
2. Set an initial criterion slightly above (or below) baseline level.
3. Deliver reinforcement when the behavior meets the criterion.
4. Once behavior stabilizes at the criterion (meets it consistently), increase the criterion.
5. Repeat through multiple criterion changes.

Requirements for Experimental Control

• Behavior should closely track each criterion change.
• Include at least 3–4 criterion changes.
• Vary the size of criterion changes (some small, some large) to rule out maturation.
• Consider including a reversal to a previous criterion to strengthen the demonstration.

When to Use

• Behaviors that are shaped gradually (increasing exercise, decreasing cigarettes, building fluency).
• When reversal is impractical and multiple baselines are not available.

Combined and Hybrid Designs

• Multiple baseline with embedded reversal: Strengthen evidence by reversing the IV in one tier.
• Multiple probe: Variant of multiple baseline where baseline probes are taken periodically rather than continuously. Reduces assessment burden.
• Nonconcurrent multiple baseline: Baselines are collected at different times rather than simultaneously. Weaker experimental control but practical when participants are recruited sequentially.

Selecting a Design

Factor	Recommended Design
Reversible behavior, one target	Reversal (A-B-A-B)
Irreversible behavior (skill acquisition)	Multiple baseline
Comparing treatments	Alternating treatments
Gradual, shaped behavior	Changing criterion
Safety concerns with withdrawal	Multiple baseline
Multiple participants available	Multiple baseline across participants
Multiple settings to address	Multiple baseline across settings
Rapid treatment comparison needed	Alternating treatments

Threats to Internal Validity

Threat	Design Most Vulnerable	Mitigation
Maturation	A-B (no replication)	Use designs with replication
History	All designs with concurrent events	Multiple baseline across settings
Testing/reactivity	All designs	Collect unobtrusive measures
Instrumentation	All designs	Maintain IOA and observer training
Covariation	Multiple baseline	Select independent tiers
Carryover	Alternating treatments, reversal	Counterbalance, use extended phases
Sequence effects	Alternating treatments	Randomize condition order

Key References

• Sidman, M. (1960). Tactics of Scientific Research. Basic Books.
• Kazdin, A. E. (2010). Single-Case Research Designs (2nd ed.). Oxford University Press.
• Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). Applied Behavior Analysis (3rd ed.). Pearson.
• Kratochwill, T. R., et al. (2010). Single-case designs technical documentation. What Works Clearinghouse.