The multiple baseline design is one of the most widely used designs in applied behavior analysis, and one of the most tested on the BCBA exam. It lets you show that your intervention caused a change in behavior without ever taking an effective treatment away. This practical guide explains how a multiple baseline works, the three types and when to use each, how to set one up and read the graph, and how to answer multiple baseline questions on the BCBA exam.
Key takeaways
- What it is: baseline data are collected on two or more tiers at once, and the intervention is introduced to one tier at a time, at staggered points.
- Why it works: if each tier changes only when the intervention reaches it, while the others stay flat, that is strong evidence of a functional relation.
- Three types: across behaviors, across settings, and across participants.
- When to use it: when reversing the intervention would be unethical or the behavior will not reverse, such as a newly learned skill.
- Design rule of thumb: plan at least three tiers so you can show at least three demonstrations of effect.
In this guide
- What is a multiple baseline design?
- How a multiple baseline shows experimental control
- The three types of multiple baseline design
- Concurrent vs. nonconcurrent multiple baseline
- When to use a multiple baseline design
- How to set up a multiple baseline: step by step
- How to read a multiple baseline graph
- Advantages and limitations
- Multiple baseline on the BCBA exam
- FAQ
What is a multiple baseline design?
A multiple baseline design is a single-case experimental design in which you measure the same target in two or more tiers (behaviors, settings, or participants) at the same time, then introduce the intervention to each tier one at a time, at different points. Each tier is its own A-B comparison: baseline (A), then intervention (B).
Baer, Wolf, and Risley described the design in their 1968 paper defining applied behavior analysis, and it remains the most common design in the field. Reviews have found multiple baseline designs in roughly 69% to 80% of single-case studies.
How a multiple baseline shows experimental control
A single A-B comparison cannot rule out coincidence. Something else might have changed at the same moment you started the intervention. A multiple baseline solves this with baseline logic, which has three parts:
- Prediction: stable baseline data in every tier predict what would happen if nothing changed.
- Verification: the tiers still in baseline keep showing the same level while the first tier changes. This verifies that the change was not caused by something affecting everyone.
- Replication: when the intervention reaches the next tier, that tier changes too. Each staggered replication makes coincidence less and less likely.
In short: behavior should change when, and only when, the intervention is introduced to each tier.
The three types of multiple baseline design
| Type | What stays the same | What varies across tiers | Example |
|---|---|---|---|
| Across behaviors | One participant, one intervention | Two or more different behaviors | Teaching a learner to request, label, and answer questions using the same prompting procedure, starting each skill at a different time |
| Across settings | One participant, one behavior, one intervention | Two or more settings or contexts | Introducing a token system for on-task behavior in math, then reading, then science class |
| Across participants | One behavior, one intervention | Two or more people | Starting behavioral skills training for three new staff members one week apart |
Choosing the right type
- Across behaviors works when one person has several target behaviors that will respond to the same intervention but are independent of each other.
- Across settings works when one behavior occurs in several places and you can introduce the intervention in one place without it spreading to the others.
- Across participants works when several people share the same target and are not likely to influence each other, for example by watching each other receive the intervention.
Concurrent vs. nonconcurrent multiple baseline
In a concurrent multiple baseline, all tiers are measured during the same stretch of calendar time. Session 1 in every tier happens before session 2 in any tier. In a nonconcurrent multiple baseline, the tiers are run at different times, for example one participant in March and another in June.
The concurrent version gives stronger control over history effects, because an outside event that happens on a given day would show up in every tier at once. Nonconcurrent designs are useful when you cannot recruit all participants at the same time, but they rely more heavily on replication to rule out coincidence.
When to use a multiple baseline design
| Use a multiple baseline when… | Consider another design when… |
|---|---|
| Withdrawing an effective intervention would be unethical (e.g., self-injury, aggression) | The tiers are likely to influence each other (behaviors are related, settings overlap) |
| The behavior is unlikely to reverse (a newly learned skill) | You cannot wait: later tiers must stay in baseline for a long time |
| You have several behaviors, settings, or people with a shared target | You only have one behavior, setting, and participant |
| You want to show generality across behaviors, settings, or people | You need to compare two or more interventions directly (consider alternating treatments) |
How to set up a multiple baseline: step by step
- Choose at least three tiers. Pick behaviors, settings, or participants that share the same target and intervention but are independent of each other.
- Write operational definitions and pick a measure. Use the same measurement procedure in every tier, and plan interobserver agreement checks.
- Start baseline in all tiers at the same time. Collect data until each tier is stable, with no clear trend toward improvement.
- Introduce the intervention to the first tier only. Keep every other tier in baseline.
- Wait for a clear change before moving on. Introduce the intervention to the next tier only after the first tier changes and the remaining tiers stay stable.
- Repeat for each tier. Keep measuring procedural integrity so you know the intervention was delivered as planned.
How to read a multiple baseline graph
When you read a multiple baseline graph, analyze it in two directions:
- Within each tier (horizontal): look for a clear change in level, trend, or variability from baseline to intervention.
- Across tiers (vertical): at the moment the intervention starts in one tier, look straight down. The tiers still in baseline should not change at that point.
Red flags that weaken experimental control: a baseline that is already trending in the therapeutic direction, a change in an untreated tier when another tier begins intervention (a sign the tiers are not independent), a change that appears several sessions after the intervention starts, or fewer than three tiers showing a clear effect.
Advantages and limitations
| Advantages | Limitations |
|---|---|
| No need to withdraw an effective treatment | Tiers must be independent, which is not always possible |
| Works for irreversible behaviors and skills | Later tiers stay in baseline longer, delaying treatment |
| Easy to fit into clinical practice | Requires repeated measurement across several tiers |
| Shows generality across behaviors, settings, or people | A change in one tier can spread to others and weaken control |
A common variation, the multiple probe design, reduces the burden of long baselines by taking intermittent probe measures instead of continuous data in the tiers still waiting for intervention.
Multiple baseline on the BCBA exam
Multiple baseline designs fall under Domain D, Experimental Design, of the BCBA Test Content Outline (6th ed.), which has 13 scored questions. The most relevant tasks are D.3 (threats to internal validity), D.4 (defining features of single-case designs), D.7 (types of single-case designs), and D.9 (applying single-case designs). Expect scenario questions that ask you to pick the right design, or to judge whether a graph shows experimental control.
Practice question
A BCBA wants to evaluate a token economy for on-task behavior with three students in the same classroom. The teacher will not allow the token economy to be removed once it starts. Which design is most appropriate?
- Reversal (ABAB) design
- Multiple baseline design across participants
- Alternating treatments design
- Changing criterion design
Show the answer
B. Multiple baseline design across participants. The intervention cannot be withdrawn, which rules out a reversal design. There is only one intervention, so an alternating treatments design does not fit, and the goal is not to change behavior in steps toward a criterion. Three students with the same target behavior make a natural set of tiers, and introducing the token economy to one student at a time shows experimental control without removing it.
Master Domain D, one task at a time
The free BCBA Study Checklist lists every task in the 6th edition Test Content Outline, including all nine Experimental Design tasks, so you can track exactly what you have mastered.
Multiple baseline design FAQ
What is a multiple baseline design in ABA?
A multiple baseline design is a single-case experimental design in which baseline data are collected on two or more behaviors, settings, or participants at the same time, and the intervention is introduced to each one at a different, staggered point in time. If behavior changes only when, and only where, the intervention is introduced, that supports a functional relation.
What are the three types of multiple baseline design?
Multiple baseline across behaviors (one participant, two or more different behaviors), across settings (one participant and behavior, two or more settings), and across participants (the same behavior and intervention for two or more people).
Why use a multiple baseline instead of a reversal (ABAB) design?
Use a multiple baseline when withdrawing the intervention would be unethical or impractical, or when the behavior is unlikely to reverse, such as a newly learned skill. A multiple baseline shows experimental control without removing an effective treatment.
How many tiers does a multiple baseline design need?
Quality standards such as the What Works Clearinghouse guidelines call for at least three demonstrations of effect at three different points in time, so most multiple baseline designs use at least three tiers.
What is the difference between concurrent and nonconcurrent multiple baseline designs?
In a concurrent multiple baseline, all tiers are measured during the same calendar time, so session 1 in every tier happens before session 2 in any tier. In a nonconcurrent multiple baseline, tiers are run at different times, which weakens control for history effects.
What does the BCBA exam test about multiple baseline designs?
Domain D (Experimental Design) of the BCBA Test Content Outline (6th ed.) covers the types of single-case designs (D.7), applying them (D.9), and threats to internal validity (D.3). Expect scenario questions asking you to choose the right design or judge whether a graph shows experimental control.
Related reading: BCBA 6th Edition Task List Explained · How to Pass the BCBA Exam in 2027
References
- 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.
- Behavior Analyst Certification Board. (2022). BCBA test content outline (6th ed.).
- Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). Applied behavior analysis (3rd ed.). Pearson.
- Kratochwill, T. R., Hitchcock, J., Horner, R. H., Levin, J. R., Odom, S. L., Rindskopf, D. M., & Shadish, W. R. (2010). Single-case designs technical documentation. What Works Clearinghouse.
- Lanovaz, M. J., & Turgeon, S. (2020). How many tiers do we need? Type I errors and power in multiple baseline designs. Perspectives on Behavior Science, 43, 605–616.
- Slocum, T. A., Pinkelman, S. E., Joslyn, P. R., & Nichols, B. (2022). Threats to internal validity in multiple-baseline design variations. Perspectives on Behavior Science, 45, 619–638.