The big idea
A single-case experimental design compares one person’s behavior under different conditions, measured many times, to show whether an independent variable changes that person’s behavior. Single-case refers to the unit of analysis, not the number of participants. Most studies include several people, but each is analyzed separately and serves as their own control.
What makes these designs experimental is baseline logic. A steady baseline predicts what behavior would do without intervention, and the design then verifies that prediction and replicates the effect (Sidman, 1960; Cooper et al., 2020). Without those steps, a graph that drops after treatment starts is only a hopeful story.
Breaking it down
The individual serves as their own control
Each participant’s behavior during baseline is compared with the same participant’s behavior during intervention. There is no separate control group. Differences between people cannot explain the result, and the analyst can see whether this person, specifically, was helped.
Repeated measurement
Behavior is measured many times within each condition, usually every session. Repeated measurement shows the level, trend and variability of the data (see D.6), lets the analyst see changes as they happen, and gives each phase enough data points to judge.
Baseline and steady state
A baseline is a condition in which the independent variable is absent. Its job is to show what behavior looks like without the intervention, so there is something to compare against.
The analyst usually waits for steady-state responding before changing conditions: a pattern with little variation in level and trend across a series of sessions. A steady state makes the prediction trustworthy. A baseline trending in the direction opposite the expected effect also works, because improvement will stand out. A baseline already trending toward improvement is a problem, since further improvement could simply continue that trend.
Prediction, verification and replication
- Prediction: steady baseline data predict that, if nothing changes, behavior will continue at about the same level. The intervention is introduced, and the data depart from the prediction.
- Verification: the design shows that the baseline prediction would have held without the IV. A reversal design does this by withdrawing the IV and seeing behavior return toward baseline. A multiple-baseline design does it by showing that untreated tiers stay at baseline while one tier is treated.
- Replication: the effect is produced again, by reintroducing the IV in a reversal design or by introducing it to the next tier of a multiple baseline. Each replication makes coincidence a less likely explanation.
A widely used standard asks for at least three demonstrations of the effect, at three different points in time, before a functional relation is claimed (Kratochwill et al., 2013).
Other features
- Visual analysis of graphed data is the main method of analysis (see D.6).
- Flexibility: the analyst can extend a phase, adjust a procedure or add a condition in response to the data.
- One variable at a time: only one variable changes when moving between conditions.
Seeing it in practice
A behavior analyst works with Jun, age 10, who leaves his seat during independent work.
| Phase | What happens | Baseline logic |
|---|---|---|
| A (baseline) | Out-of-seat counts are 12, 14, 13, 12 and 13 per session. | Prediction: without change, counts will stay near 13. |
| B (token system) | Counts fall to 3, 2, 2, 1 and 2. | The data depart from the prediction. |
| A (tokens withdrawn) | Counts rise to 10, 12 and 13. | Verification: the baseline level returns without the IV. |
| B (tokens reinstated) | Counts fall to 2, 1, 2 and 1. | Replication: the effect occurs again. |
Now picture only the first two phases. Jun’s improvement could be due to the token system, a new seat or a calmer week. An A-B design supports prediction only, so it cannot demonstrate a functional relation.
Common mix-ups
- Single-case does not mean one participant. It means each participant is analyzed individually and compared with their own baseline.
- A case study is not a single-case experiment. An anecdotal account of one client’s progress has no baseline logic and no experimental control.
- Prediction here is narrower than in A.1. In baseline logic, prediction is the expected course of behavior if conditions stay the same. In A.1, prediction is a level of understanding based on covariation.
- Verification and replication are different steps. Verification confirms what would have happened without the IV. Replication repeats the effect of the IV.
- A perfectly flat baseline is not required. A steady level, or a trend opposite the expected effect, is enough to support a prediction.
Exam tip: In an A-B-A-B design, the return to baseline (the second A) verifies, and the second B phase replicates. In a multiple baseline, the untreated tiers verify and each later tier replicates.
Quick recap
- Single-case designs compare each participant with their own baseline, using repeated measurement.
- Change conditions at steady state, or when baseline trends away from the expected effect.
- Prediction: baseline forecasts future behavior. Verification: shows the forecast would have held. Replication: repeats the effect.
- An A-B design offers prediction only and cannot show a functional relation.
Further reading
- Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). Applied behavior analysis (3rd ed.), chapter 7.
- Sidman, M. (1960). Tactics of scientific research: Evaluating experimental data in psychology. Basic Books.
- Kratochwill, T. R., Hitchcock, J. H., Horner, R. H., Levin, J. R., Odom, S. L., Rindskopf, D. M., & Shadish, W. R. (2013). Single-case intervention research design standards. Remedial and Special Education, 34(1), 26-38.
Key terms
- Single-case experimental design (Single-subject design, Single-case design)
- A research design in which each participant's behavior is measured repeatedly under different conditions and compared with their own baseline, so each person serves as their own control. Experimental control is shown through prediction, verification and replication.
- 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.
- Steady-state responding (Steady state)
- A pattern of responding that shows relatively little variation in level and trend over a series of measurements. Changing conditions at steady state makes it easier to see whether a change in behavior is due to the new condition.
- Baseline logic
- The reasoning that makes single-case designs experimental, in three steps: prediction (steady baseline data forecast future behavior if conditions stay the same), verification (the design shows that the forecast would have held without the intervention) and replication (the effect of the intervention is produced again).
- Verification
- In baseline logic, evidence that behavior would have continued at its baseline level if the independent variable had not been introduced. It comes from withdrawing the intervention in a reversal design or from untreated tiers staying unchanged in a multiple-baseline design.
- Replication
- Repeating an experiment, or a condition within an experiment, to see whether the same result occurs again. Within a study it strengthens the case for a functional relation; across studies it shows how reliable and general a finding is.
Check yourself
5 questions on D.4. Choose an answer to see whether you're right and why.
Show the answer
Answer: D. Returning to baseline conditions and seeing behavior return toward its baseline level shows that the original prediction would have held without the intervention. That is verification. Replication comes next, when the intervention is reintroduced and the effect appears again.
Show the answer
Answer: A. A baseline already trending toward improvement makes it impossible to tell whether later improvement comes from the intervention or from the existing trend. Waiting for stability gives a usable prediction. Starting now would leave any effect uninterpretable.
Show the answer
Answer: D. Single-case designs compare each person's behavior across conditions, so each participant serves as their own control. The name refers to the unit of analysis, not the number of participants; most studies include several people.
Show the answer
Answer: C. An A-B design supports prediction only. Without verification and replication, something else that changed when intervention began could explain the improvement. The size of a change does not substitute for experimental control.
Show the answer
Answer: C. When the intervention reaches the cafeteria and behavior changes there too, the classroom effect has been repeated: replication. Earlier, while the cafeteria was still in baseline, its unchanged data served as verification. It is not generalization, because behavior changed only after the intervention began there.
Adds a step to your study record. Saved in this browser only.