Single-Subject Experimental Design
- Introduced by
- Murray Sidman (foundational tactics); B. F. Skinner (applied behavior analysis lineage)
- Year
- 1960s (Sidman 1960; formal applied codification by Kazdin and Baer in 1970s–1980s)
- Type
- Experimental research design
- Family
- Experimental design
What does it do?
Single-subject experimental design (SSED) establishes experimental control by repeatedly measuring one individual (or a small number of individuals) across baseline and intervention phases, using the participant as their own control.
Instead of comparing groups, it compares the participant's own behavior across conditions over time. Widely used in applied behavior analysis, special education, rehabilitation, and clinical psychology, SSED allows causal inference from small or unique samples where group designs are impractical.
The intuition behind the method
Imagine you want to know whether a new reading intervention helps a specific child with dyslexia. You cannot recruit hundreds of children, but you can measure that one child's reading accuracy every day for weeks — first without the intervention (baseline), then with it (treatment), and perhaps back again. If the behavior changes each time the condition changes, the evidence that the intervention caused the change accumulates through replication within the same person. SSED makes the individual the unit of analysis rather than the group, treating variability within the person as the primary source of experimental evidence.
History
The intellectual foundations of SSED lie in the operant conditioning tradition pioneered by B. F. Skinner, who studied individual organisms under controlled schedules of reinforcement. Murray Sidman's 1960 book Tactics of Scientific Research articulated the logic of intrasubject replication as the basis of experimental inference in psychology. Donald Baer, Montrose Wolf, and Todd Risley's landmark 1968 article in the Journal of Applied Behavior Analysis established applied behavior analysis as a discipline and codified single-subject methodology as its primary research tool. Alan Kazdin's 1982 handbook systematized the design options, measurement requirements, and analytic conventions that remain standard today.
When to use it
Use SSED when you need experimental evidence about the effect of an intervention on a specific individual or a very small number of individuals — for example, a student with a rare disability, a patient with a unique clinical profile, or a participant pool too small for group designs. It is appropriate when ethical or logistical constraints prevent withholding treatment from a control group, and when frequent repeated measurement of a well-defined behavior over time is feasible. Do NOT use SSED when the research question concerns population-level prevalence or average effects, when the target behavior cannot be measured repeatedly and objectively, when carry-over or irreversible treatment effects would prevent a meaningful return to baseline, or when the goal is norm-referenced comparison rather than individual-level functional analysis.
Limitations and considerations
- Findings are specific to the individual(s) studied; statistical generalization to a population is not warranted from a single study.
- Requires frequent, repeated, objective measurement of a clearly operationalized target behavior — feasible for overt behaviors, harder for internal states.
- Reversal designs are inappropriate when the intervention teaches a skill that cannot be unlearned, making the return-to-baseline phase illogical or unethical.
- Visual analysis is inherently subjective; interrater agreement on phase-change conclusions is not always high, and quantitative effect-size supplements are underused.
- Carryover effects between rapidly alternating conditions (in alternating-treatments designs) can confound results.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Allows causal inference from very small or unique samples where group designs are impossible or unethical.
- The participant serves as their own control, eliminating between-subject variability and making individual change transparent.
- Continuously collected data reveal the temporal dynamics of behavior change — not just pre- and post-snapshots.
- Flexible: multiple design variants (AB, ABA, ABAB, multiple baseline, alternating treatments) accommodate diverse research questions and ethical constraints.
- Replication across participants, settings, and behaviors builds a cumulative evidence base analogous to meta-analytic convergence.
Applications
- Evaluating behavioral interventions for individuals with autism spectrum disorder, intellectual disabilities, or ADHD in special education settings.
- Assessing rehabilitation protocols for patients with acquired brain injury, stroke, or motor impairments in clinical settings.
- Testing instructional procedures or classroom management strategies for individual students whose needs differ markedly from the group.
- Investigating pharmacological or behavioral treatments for rare conditions where population-level RCTs are impractical.
- Evaluating technology-based or app-driven self-management programs at the individual level before scaling.
- Building the evidence base for novel interventions through systematic replication across individuals and settings.
How it works
How it works
Before introducing any intervention, the researcher collects repeated measurements of the target behavior under natural or unmanipulated conditions. This baseline phase (typically labeled A) establishes a stable, predictable pattern — the participant's performance without the treatment. A baseline should be stable (low variability) or trending in the direction opposite to the expected treatment effect; an unstable baseline undermines causal interpretation.
The intervention or independent variable is introduced (phase B), and the same target behavior continues to be measured at the same frequency. The shift in level (the mean), trend (direction of slope), and variability between the A and B phases constitutes the primary evidence of an effect. More complex designs add additional phases (return to baseline, second intervention) or concurrent baselines across participants, settings, or behaviors.
Data are plotted on a time-series graph, and the primary method of analysis is visual inspection. Analysts examine within-phase data characteristics (level, trend, variability) and between-phase data characteristics (change in level, change in trend, overlap between adjacent phases). Low data overlap between phases and immediate, consistent phase changes are taken as evidence of a functional relationship. Effect size indices such as nonoverlap of all pairs (NAP) or Tau-U supplement visual analysis.
Experimental control in SSED rests on replication. In a reversal design, the intervention is withdrawn and reintroduced; in a multiple-baseline design, the intervention is staggered across different baselines. Each demonstration of a behavior change coinciding with a phase change is a replication. Three or more replications of the same effect within a study constitute convincing evidence of a causal relationship, following the logic of intrasubject replication.
The researcher draws conclusions about whether a functional relationship exists between the independent variable (intervention) and the dependent variable (target behavior), and assesses the social or clinical significance of the magnitude of change. External validity is established not by statistical inference across a sample but by systematic replication across studies, settings, and participants in the literature.
Common pitfalls
- Too few data points in baseline: a baseline of fewer than three to five stable data points provides an inadequate basis for predicting the counterfactual; collect until the pattern is clear before introducing the intervention.
- Changing more than one variable at a time: introducing multiple components simultaneously makes it impossible to attribute change to any single element — alter one condition at a time.
- Ignoring trend in baseline: a baseline that is already trending in the direction of the expected treatment effect means any subsequent improvement is uninterpretable; stabilize or account for the trend before proceeding.
- Over-relying on visual analysis alone: train in standardized visual-analysis criteria and supplement with at least one validated nonoverlap statistic (NAP, Tau-U, PND) to improve objectivity.
- Misapplying a reversal design to irreversible outcomes: skills acquisition (e.g., learning to read) cannot be reversed; use a multiple-baseline design instead.
Frequently asked questions
Is single-subject design 'real' experimental research, or is it just a case study?
It is experimental. A case study describes what happened; a single-subject experimental design actively manipulates an independent variable and controls for confounds through replication of phase changes within the same participant. The logic of causal inference differs from group designs (it rests on intrasubject rather than intersubject replication) but is nonetheless rigorous when design conditions are met.
How many data points do I need in each phase?
The general guideline is a minimum of three to five data points per phase, but more are always preferable. The criterion is not a fixed number but stability: collect data until the pattern is predictable (stable level, clear trend, or acceptable variability) before changing phases. Rushing to introduce the intervention after only one or two baseline points undermines causal interpretation.
When should I use a multiple-baseline design instead of a reversal (ABA/ABAB) design?
Use a multiple-baseline design when the intervention effect is likely to be irreversible (e.g., skill acquisition), when withdrawing treatment would be unethical, or when a reversal would undo genuine therapeutic gains. A reversal design is most appropriate when the target behavior is expected to return toward baseline once the intervention is removed — typically operant behaviors maintained by environmental contingencies rather than learned skills.
Can I use statistics with single-subject data, or is visual analysis enough?
Visual analysis is the primary and traditional method, but quantitative supplements are encouraged. Nonoverlap statistics such as Nonoverlap of All Pairs (NAP), Tau-U, or Percentage of Nonoverlapping Data (PND) provide an objective effect-size index and aid communication with audiences unfamiliar with time-series graphs. Multilevel modeling and randomization tests are also applicable when the data and design warrant them.
How do I establish external validity from a study with only one or a few participants?
External validity in SSED is built through systematic replication across studies — your own replication across multiple participants within the same study (direct replication) and subsequent studies across different settings, populations, and researchers (systematic replication). A single study does not establish external validity; a body of replicated studies does. This is analogous to the role of meta-analysis for group-design literatures.
Related methods
Sources
Kazdin, A. E. (1982). Single-Case Research Designs: Methods for Clinical and Applied Settings. Oxford University Press.
ISBN: 978-0195030440Copied.Sidman, M. (1960). Tactics of Scientific Research: Evaluating Experimental Data in Psychology. Basic Books.
Open publication referenceCopied.