Design guide
Which design fits your question?
Answer a few questions to find a design that fits your behavior, your setting, and your ethics. Each design page then shows what experimental control looks like on a graph, and what it doesn't.
Decision guide
Do you want to compare two or more treatments or conditions, such as in a functional analysis?
Are you comparing ways to teach a new skill, one that stays learned once it's acquired?
For example, comparing two prompting procedures for teaching sight words. A word learned under one procedure can't be unlearned before the other procedure is tested on it.
Can the conditions switch from session to session without one carrying over into the next?
Carryover means a condition's effect lingers into the next session. Conditions need to be easy to tell apart, for example by a different therapist, room, or color cue.
Do you plan to change the behavior gradually, in steps toward a final goal?
For example, raising daily exercise from 10 to 30 minutes, or reducing cigarettes per day, with reinforcement for meeting each step.
Would the behavior likely return to baseline if you removed the intervention, and is removing it safe and acceptable?
Learned skills such as reading usually don't reverse. Removing treatment for dangerous behavior may not be acceptable.
Do you have at least three participants, behaviors, or settings that are similar but independent of each other?
Similar means each tier should respond to the same intervention. Independent means treating one tier shouldn't change the others. Two behaviors that the same skill would improve together, or two students who sit together and could copy each other, are not independent.
Can you measure every tier in every session throughout a long baseline?
Repeated testing can be impractical, tiring, or reactive, for example when testing itself teaches the skill.
Recommended
Alternating treatments or multielement design
Alternate the conditions rapidly and compare their data paths. A functional analysis is a multielement design.
Recommended
Adapted alternating treatments design
Make separate sets of targets that are equally difficult and don't overlap, and assign one set to each procedure. Alternate the procedures across sessions, and compare how quickly each set reaches mastery. Keep a control set that is tested but never taught, and replicate the comparison across participants or with new sets of targets.
Recommended
Reversal design with more than one treatment
Give each treatment its own reversal, for example A-B-A-B-A-C-A-C, so each one is compared with baseline three times. To compare the treatments with each other directly, put them in adjacent phases, for example A-B-A-B-C-B-C. The earlier treatment can change how behavior responds to the later one (multiple-treatment interference), so counterbalance the order across participants when you can.
This works only if behavior returns toward baseline when a treatment is withdrawn, and withdrawing it is safe. If it isn't, test each treatment in its own multiple baseline across independent participants, behaviors, or settings. That shows whether each treatment works, but comparing results across separate designs is weaker than comparing the treatments directly.
Recommended
Changing criterion design
Change the criterion in at least three steps, and look for behavior that tracks each step.
Recommended
Reversal (ABAB) design
Introduce, withdraw, and reintroduce the intervention. Each change is a chance to show the effect again.
Recommended
Multiple baseline design
Start intervention at different times across three or more tiers, while measuring them all continuously.
Recommended
Multiple probe design
A multiple baseline that uses intermittent probes in baseline, with several consecutive probes just before each tier starts intervention.
Consider carefully
An AB design can describe change, but it can't show control
With one participant and a behavior that won't reverse, an AB design can track progress, but it can't rule out other explanations. To show that the intervention caused the change, add at least two more participants and start intervention after a different number of baseline sessions for each: a multiple baseline across participants. Run their baselines over the same period if you can. If participants can only enroll at different times (a nonconcurrent multiple baseline), the design still helps, but it is weaker at ruling out outside events, and some reviewers and standards don't accept it as showing a functional relation.
The designs at a glance
Every design gets its evidence the same way: by showing the effect at least three times. They differ in how they create those three chances.
| Design | How it shows the effect three times | Needs | Minimum to meet WWC pilot standards |
|---|---|---|---|
| AB | It doesn't. There is only one comparison. | Nothing beyond a baseline | Does not meet standards |
| Reversal (ABAB) | Introducing, withdrawing, and reintroducing the intervention | A reversible behavior; a withdrawal that is ethical | 4 phases, 5 points per phase (3 with reservations) |
| Multiple baseline | Staggered starts across tiers | 3 or more independent participants, behaviors, or settings | 6 phases (3 tiers), 5 points per phase (3 with reservations) |
| Multiple probe | Staggered starts, with intermittent baseline probes | The same as multiple baseline | As for multiple baseline, plus three probe rules from a later WWC handbook: every tier probed in each of the first 3 sessions, waiting tiers probed when another tier starts intervention, and 3 consecutive probes just before each tier starts |
| Alternating treatments / multielement | Repeated, rapid alternation of conditions | Discriminable conditions; little carryover | 5 repetitions of the alternating sequence (4 with reservations) |
| Changing criterion | Behavior shifting with each new criterion | A target that can change in steps | At least 3 different criteria |
Minimums are from the What Works Clearinghouse single-case design pilot standards (Kratochwill et al., 2010), which are widely used to judge single-case research. The 2010 standards don't cover multiple probe designs; the probe rules come from the WWC Standards Handbook, version 4.0 (What Works Clearinghouse, 2017, pp. A-5 to A-6). Later WWC handbooks revised how studies are rated, so check the current version for formal reviews. See sources.
Each design, with examples
Every design page shows a graph with experimental control and graphs without it, explains the difference, and ends with live practice.
AB design
Why a clear change still isn't experimental control.
Open →Reversal (ABAB)
Prediction, verification, and replication in one participant.
Open →Multiple baseline and probe
Staggered starts across participants, behaviors, or settings.
Open →Alternating treatments and multielement
Comparing conditions quickly, including functional analysis.
Open →Changing criterion
Behavior that steps up or down with each criterion.
Open →Visual analysis lessons
Level, trend, variability, immediacy, overlap, and consistency.
Open →