Multiple Baseline Design — Single-Subject Experimental Design
- Introduced by
- Donald M. Baer, Montrose M. Wolf, Todd R. Risley
- Year
- 1968
- Type
- Single-subject experimental design
- Family
- Experimental design
What does it do?
The multiple baseline design is a single-subject experimental design that demonstrates functional control by introducing an intervention at staggered time points across two or more baselines — typically across different behaviors, individuals, or settings.
Because no withdrawal of treatment is required, it is especially suitable when the target behavior is irreversible or when removing an effective intervention would be unethical.
The intuition behind the method
Imagine teaching three students to read sight words. Instead of starting the intervention with all three simultaneously, you wait and collect baseline data on all three, then introduce the intervention first with Student 1, then — only after Student 1 shows improvement — with Student 2, and finally with Student 3. If improvement tracks each introduction of the intervention rather than happening spontaneously across all students, you can conclude the intervention, not some external event, caused the change. The staggered start is the logical engine of the design: coincidence could explain one improvement at one moment, but not three improvements at three different, deliberately chosen moments.
History
The multiple baseline design was formally described by Baer, Wolf, and Risley in their 1968 foundational paper in the Journal of Applied Behavior Analysis, which established the core methodological standards of applied behavior analysis. The design addressed a critical limitation of reversal designs: the inability to demonstrate control when learned behaviors cannot or should not be removed. Through the 1970s and 1980s, researchers in special education and clinical psychology adopted and refined the design, codifying visual analysis conventions and expanding applications. Today it remains one of the most commonly used single-case experimental designs in behavior-analytic and educational intervention research.
When to use it
Use a multiple baseline design when the target behavior is expected to be irreversible once learned (e.g., reading skills, social skills acquisition), making a withdrawal design unethical or impractical. It is well suited when you can identify at least two or three conceptually similar, functionally independent tiers (behaviors, participants, or settings) that share the same intervention logic. Minimum requirement: at least two tiers with stable baselines of at least three to five data points each. Do not use this design if the tiers are not independent — concurrent improvement across all tiers during any single baseline phase undermines the logic of the staggered demonstration. Also avoid it when the intervention cannot be withheld from some tiers for a meaningful period for ethical or practical reasons.
Limitations and considerations
- Requires at least two (preferably three or more) functionally independent tiers; a design with only two tiers offers weaker experimental control.
- Data collection must be continuous across all tiers simultaneously from the start, which is resource-intensive.
- Generalization of findings across participants or settings is limited by the small, often non-random samples inherent in single-subject research.
- If tiers are not truly independent, behavior may change in untreated tiers (multiple baseline covariation), threatening internal validity.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Demonstrates experimental control without requiring a return-to-baseline phase, making it ethical when skill acquisition is irreversible.
- Each tier serves as an independent replication, strengthening the case for a functional relation.
- Applicable across diverse contexts: different participants, behaviors, or settings can serve as tiers.
- Well-accepted in applied behavior analysis, special education, and clinical psychology literatures.
- Produces rich, continuous behavioral data that reveal the trajectory and immediacy of intervention effects.
Applications
- Evaluating skill acquisition interventions (e.g., reading, communication, daily living skills) in students with developmental disabilities.
- Testing behavioral interventions across multiple participants in applied behavior analysis and special education research.
- Assessing the generalization of a therapeutic technique across different settings (clinic, home, school).
- Examining the sequential introduction of workplace training protocols across different departments or job roles.
- Validating new clinical procedures in small-n health-sciences research where group designs are impractical.
- Demonstrating treatment fidelity and replication in evidence-based practice research.
How it works
How it works
Simultaneous, continuous baseline data are collected across all tiers — whether those tiers are multiple behaviors of the same individual, the same behavior in multiple individuals, or the same behavior in multiple settings. All tiers must show stable, predictable baseline performance before the intervention is introduced to any tier.
The intervention is introduced to the first tier while the remaining tiers continue in baseline. The length of each baseline is deliberately staggered so that each tier's baseline is longer than the preceding tier's, typically differing by at least three to five data points.
Once the first tier demonstrates a clear, stable behavior change following the intervention, the intervention is introduced to the second tier while subsequent tiers continue in baseline. The same criterion applies: a stable, replicable change should be evident before the next tier is treated.
The intervention is introduced to each remaining tier in turn, following the same logic. Each new tier provides an independent replication of the intervention effect. Ideally, behavior in tiers still in baseline should remain unchanged while each treated tier improves.
A functional relation between the independent variable (intervention) and the dependent variable (target behavior) is established when behavior changes predictably and only after the intervention is introduced to each successive tier. Visual analysis of graphed data — examining level, trend, variability, immediacy of effect, and overlap between phases — is the primary analytic method.
Common pitfalls
- Too few data points in baseline — a minimum of three to five stable data points per tier is needed; rushing to intervention without stable baselines makes it impossible to detect a true effect.
- Dependent tiers — if the same behavior across different settings improves together regardless of intervention timing (e.g., a child generalizes skills automatically), the staggered logic collapses; verify tier independence before beginning.
- Treating tiers too quickly — advancing to the next tier before the previous tier shows a clear, stable change conflates the effects and undermines replication logic.
- Ignoring trend in baseline — a consistently improving baseline trend means behavior may already be changing; introducing the intervention into an already-improving baseline makes the effect uninterpretable.
- Relying on statistical tests alone — visual analysis is the field-standard primary method; supplementary statistical approaches (e.g., effect size measures such as Tau-U) should complement, not replace, careful visual inspection.
Frequently asked questions
How many tiers do I need?
A minimum of two tiers is required, but three or more tiers are strongly recommended. With only two tiers the design provides a single replication, which is a relatively weak demonstration of experimental control. Most methodologists and journal reviewers consider three tiers the practical minimum for a credible functional relation claim.
What counts as a 'tier' — can I mix behaviors, participants, and settings?
Each study uses one type of tier for conceptual clarity. A multiple baseline across behaviors uses the same participant in the same setting but targets different behaviors. A multiple baseline across participants uses the same behavior and setting but different individuals. A multiple baseline across settings uses the same participant and behavior in different settings. Mixing tier types within a single study is non-standard and complicates interpretation.
How do I know when to advance to the next tier?
The standard criterion is that the tier currently receiving the intervention shows a clear, stable change in level and/or trend relative to its baseline — and that this change appears unlikely to reverse. Common conventions include three to five consecutive data points in the intervention phase that demonstrate the expected direction of change with low variability. There is no single universal rule; the criterion should be pre-specified and reported.
Is the multiple baseline design suitable for group data?
No. The multiple baseline design is a single-subject (single-case) design. Each tier is analyzed individually, and the logic of replication is based on within-subject comparison over time, not between-group comparison. If you have a sample large enough for group-level inference, a randomized controlled trial or quasi-experimental design is more appropriate.
Can I use statistical analysis instead of visual analysis?
Visual analysis is the conventional and primary method in single-case research. Effect size indices such as Tau-U, non-overlap of all pairs (NAP), or the percentage of non-overlapping data (PND) are widely used as supplements and for meta-analytic purposes, but they do not replace careful visual inspection of graphed data for phase-level interpretation.
Related methods
Sources
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.Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). Applied Behavior Analysis (3rd ed.). Pearson.
ISBN: 978-0134752556Copied.