Every single-case design answers the same question: did the independent variable, and not something else, change the behavior? The four designs in this lesson get there by different routes. Knowing them helps you pick a design and read published graphs.
Why it matters
As a BCBA you will decide whether an intervention is working and whether it is worth keeping. A poorly matched design wastes sessions, misleads you, or forces you to remove an effective treatment for dangerous behavior.
Practice items on this task are usually short vignettes: name the design used, choose the one that fits a case, or point to where a graph shows prediction, verification, or replication.
Key ideas
- Reversal (A-B-A-B) design. You collect baseline (A), introduce the intervention (B), withdraw it to return to baseline conditions (A), then reinstate it (B). Many authors call this version a withdrawal design. The first A phase predicts what would happen without treatment. The return to A verifies that prediction. The second B phase replicates the effect. It is a strong demonstration of control, but it only works when behavior is likely to reverse, and withdrawing treatment can be unethical for harmful behavior.
- Multiple-baseline design. You take baselines at the same time on two or more tiers (behaviors, settings, or participants), then introduce the intervention to one tier at a time in a staggered sequence. Verification comes from tiers still in baseline staying unchanged while another tier gets treatment. Replication comes from each tier changing only when the intervention reaches it. Nothing has to be withdrawn, so it suits skills that will not "unlearn" and behavior you would not want to bring back. The tiers should be functionally similar but independent; if treating one tier changes the others, control is unclear.
- Multielement design (also called the alternating treatments design). You rapidly alternate two or more conditions, often session by session in a counterbalanced or random order, and pair each condition with a distinct stimulus so the learner can tell them apart. Control is shown by differentiation: the data paths separate and stay separated. It can compare treatments quickly and does not require a baseline phase or withdrawal. Functional analyses typically use this logic. Its main weakness is multiple-treatment interference.
- Changing-criterion design. After baseline, you set a performance criterion and change it in a series of steps (for example, 10, then 15, then 20 minutes of exercise). Control is shown when behavior shifts to match each new criterion and stays there. You strengthen the case by varying how long each phase lasts and how big each step is, or by briefly returning to an earlier criterion. It fits behavior the learner can already do that you want to increase or decrease gradually.
Definitions
- Reversal design: a design in which the intervention is introduced, withdrawn, and reintroduced, so behavior is compared across alternating baseline and treatment phases.
- Multiple-baseline design: a design in which baselines run at the same time on two or more tiers and the intervention is introduced to the tiers one at a time.
- Multielement design: a design in which two or more conditions are rapidly alternated, each linked to a distinct stimulus, and their effects are compared by how far apart their data paths are.
- Changing-criterion design: a design in which a single behavior is changed in graded steps by setting a new performance criterion in each treatment phase.
- Multiple-treatment interference: a change in behavior during one condition that is partly caused by exposure to another condition in the same study, through sequence, carryover, or alternation effects.
Examples from a session
Matching the design to the question
- Reversal: At home, Devon, 6, gets a token for each minute of independent play. Play rises, drops when tokens stop, and rises again when they return.
- Multiple baseline across settings: A high school uses a self-monitoring card for Priya, 15, in math, then science, then English, starting each class a week apart. Her on-task behavior improves in each class only after the card is introduced there.
- Multielement: In a clinic, a BCBA alternates two prompting procedures across sessions for Mr. Alvarez, 42, using a blue folder for one and a green folder for the other. Each procedure teaches its own set of equally difficult grocery words. The two data paths separate within a few sessions.
- Changing criterion: Grace, 28, earns a preferred activity when she drinks a set amount of water each day. The goal rises from 2 to 3 to 4 cups, and her intake steps up to each level.
- Non-example (changing criterion): A coach raises Leo's daily reading goal each week, but his minutes climb steadily whatever the goal is. Behavior does not track each criterion, so this shows no control.
When you meet a vignette, ask three questions in order: Is anything withdrawn? Is anything staggered across tiers? Are conditions alternating quickly, or is a goal changing in steps? The answers usually name the design.
Common mistakes / exam traps
- Treating the multiple baseline as one A-B design. A single tier is just an A-B comparison. Control comes from the staggered starts and the untreated tiers staying stable.
- Thinking a multielement design needs a baseline phase. It can start with conditions alternating from the first session. A baseline condition can be one of the alternated conditions, though.
Exam trap
Reversal vs. multiple baseline. The discriminating feature is how the prediction is verified: a reversal withdraws treatment and expects behavior to return to baseline; a multiple baseline keeps untreated tiers in baseline and expects them to stay flat. If behavior will not reverse (a learned skill) or should not (injurious behavior), choose the multiple baseline.
Exam trap
Effect under alternation vs. effect alone. In a multielement design, a treatment's effect while alternating with another may not match its effect when used by itself. That is multiple-treatment interference. Counterbalancing reduces sequence effects but cannot remove every alternation or carryover effect.
Quick check
Ines, a BCBA, wants to show that a new reinforcement plan reduces head hitting for a client in a clinic. The plan has already cut head hitting sharply, and the family does not want it removed. Which design consideration matters most here?