The big idea
Knowing the designs (D.7) is half the job. Applying them means choosing the one that fits the question, the behavior, the setting and the client’s welfare, and then running it well: measuring reliably, changing conditions at the right time and replicating the effect.
In practice, the design serves the client. A behavior analyst does not withdraw an effective treatment for dangerous behavior just to get a cleaner graph. But when a decision depends on knowing whether an intervention is responsible for a change (keeping a restrictive procedure, recommending an expensive service, sharing data with a physician about a medication), an experimental element such as a brief reversal or staggered starts gives a far better answer than an A-B comparison.
Breaking it down
Start with the question
- Does this intervention work for this person? Reversal or multiple-baseline design.
- Which of two or more procedures works better? Multielement design, often followed by a phase with the best treatment alone.
- What maintains the behavior? A functional analysis in a multielement design (F.6).
- Can behavior change step by step toward a goal? Changing-criterion design.
- Which parts, or how much, of a procedure matter? Component or parametric analyses (D.8).
Then check the behavior and the setting
- Is the behavior dangerous? Avoid a long return to baseline. Use a multiple baseline, or a brief reversal only if the risk is acceptable and the team and family agree.
- Will the behavior reverse? A learned skill, or behavior now maintained by natural reinforcers, often will not. Cooper et al. (2020) call this irreversibility. Use a multiple-baseline or multiple-probe design.
- Are there several similar behaviors, settings or clients? That makes a multiple baseline possible, as long as the tiers are independent of each other.
- Can staff run conditions that switch quickly? If not, a multielement design may suffer from poor integrity.
- Is the behavior already in the repertoire, and is gradual change wanted? Then a changing criterion fits.
Run it well
- Define the target behavior and measure it reliably, with interobserver agreement checks (C.1, C.8).
- Collect baseline until it is steady or trending away from the expected effect (D.4).
- Introduce the IV, changing one variable at a time, and measure treatment integrity (C.12).
- Make phase changes based on the data, not the calendar.
- Plan for at least three demonstrations of the effect.
- Graph and inspect the data often (D.6), and change course if the intervention is not working.
Ethics
Evaluation in service delivery must put the client first. Behavior analysts collect and graph data to make decisions about services (Standard 2.17, Collecting and Using Data) and keep evaluating the intervention, changing it when it is not working (Standard 2.18, Continual Evaluation of the Behavior-Change Intervention). When the evaluation is also research, client services and welfare still take priority (Standard 6.03, Research in Service Delivery), and research consent is required (Standard 6.04, Informed Consent in Research).
Seeing it in practice
| Situation | Good fit | Why |
|---|---|---|
| Kenji, age 9, hits peers at school; staff want to know whether a new plan works. | Multiple baseline across settings (recess, lunch, classroom) | No need to remove a working plan for aggression |
| A clinic teaches Sofia three sets of signed requests. | Multiple probe across sets | The skills will not reverse, and constant testing of untaught sets would frustrate her |
| A teacher asks whether peer praise or teacher praise increases Dev’s participation more. | Multielement, then the better condition alone | Fast comparison without long phases |
| Ms. Al-Mansour, 72, who lives in a retirement community, wants to walk farther each day. | Changing criterion | She already walks; the goal is gradual increases |
| A token system for on-task behavior in a resource room, where the behavior is not dangerous | Reversal (A-B-A-B) with a brief withdrawal | Clear evidence, low risk, and the behavior is likely to reverse |
Common mix-ups
- An A-B design is a monitoring tool, not an experiment. It tracks progress but cannot show that the intervention caused the change.
- A reversal is not always unethical. A brief withdrawal for behavior that is not dangerous can be reasonable and gives strong evidence.
- A multiple baseline needs independent tiers. If treating one behavior changes the others, verification is lost.
- Plan the design before treatment starts, but stay flexible. Adjusting phases in response to the data is a strength of single-case designs, not a failure of planning.
Exam tip: In “which design should the analyst use” questions, eliminate first. Dangerous behavior or a skill that will not reverse rules out a reversal. One behavior, in one setting, with one client rules out a multiple baseline. Behavior not yet in the repertoire rules out a changing criterion. Then pick the design that answers the question asked.
Quick recap
- Match the design to the question: effect, comparison, function, stepwise change, or parts and amounts.
- Avoid reversals for dangerous or irreversible behavior; use multiple-baseline or multiple-probe designs instead.
- Multielement designs compare quickly but need conditions that are easy to tell apart and good integrity.
- Changing criterion fits behavior already in the repertoire that should change in steps.
- Measure reliably, change phases on data, replicate, and keep the client’s welfare first.
Further reading
- Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). Applied behavior analysis (3rd ed.), chapters 8 to 10.
- Kazdin, A. E. (2011). Single-case research designs: Methods for clinical and applied settings (2nd ed.). Oxford University Press.
Key terms
- Reversal design (A-B-A-B design, Withdrawal design)
- A single-case design in which baseline and intervention conditions are introduced, withdrawn and reintroduced, typically in an A-B-A-B sequence. The return to baseline provides verification, and the reinstated intervention provides replication.
- Multiple-baseline design
- A single-case design in which baselines are measured at the same time for two or more behaviors, settings or participants, and the intervention is introduced to them one at a time in a staggered sequence. Untreated tiers provide verification, and each tier that changes when treated provides replication.
- Multiple-probe design
- A variation of the multiple-baseline design in which baseline data are gathered through intermittent probes instead of continuous measurement, with a short series of consecutive probes just before each tier receives the intervention. It suits skills taught in sequence and cases where continuous baseline testing would be impractical or aversive.
- Multielement design (Alternating treatments design, ATD)
- A single-case design in which two or more conditions alternate rapidly, often from session to session, each paired with a distinct stimulus and plotted as its own data path. Differences between the data paths show the effect of the conditions.
- Changing-criterion design
- A single-case design in which, after baseline, an intervention is applied in a series of steps, each with a new criterion for performance. Experimental control is shown when behavior changes to match each new criterion.
- Irreversibility
- A situation in which behavior does not return to its earlier level when the conditions of an earlier phase are reinstated. It makes a reversal design unsuitable, as with learned skills or behavior now maintained by natural reinforcers.
- Baseline
- A condition in which the independent variable is absent, used to measure behavior before or without the intervention. A steady baseline predicts what behavior would do if nothing changed, giving later phases something to be compared against.
Check yourself
5 questions on D.9. Choose an answer to see whether you're right and why.
Show the answer
Answer: D. Staggering the intervention across the three settings can show experimental control without withdrawing treatment, which the team has ruled out. A reversal design depends on the very return of self-injury the team wants to avoid.
Show the answer
Answer: C. A reversal design needs behavior to return toward baseline when the intervention is withdrawn. Learned academic skills usually persist, which is irreversibility. A multiple-baseline or multiple-probe design across skill sets or students would fit better.
Show the answer
Answer: A. A multielement (alternating treatments) design compares conditions quickly, and a final phase with only the better method puts the result to use. Month-long phases would answer the question far more slowly, and combining the methods could not show which one is responsible.
Show the answer
Answer: A. Untreated tiers are supposed to stay at baseline so they can verify the prediction. When they change along with the treated tier, the change could come from something other than the intervention, or the behaviors may be linked. Replication requires each tier to change only when it is treated.
Show the answer
Answer: D. The behavior is already in the repertoire, and the goal is stepwise change, so a changing-criterion design fits: control is shown when exercise tracks each new criterion. A multiple-probe design suits new skills taught in sequence.
Adds a step to your study record. Saved in this browser only.