03 · What You Need to Know
When Single-Case Experimental Research Can Support Credible Causal Inference
What is a single-case experimental design?
A single-case experimental design, often abbreviated SCED, is a research approach that examines the effects of systematically manipulated interventions through repeated observation of individual cases or units.
The case may be a person, classroom, organization, or another clearly defined unit, depending on the design.
Researchers repeatedly measure a target outcome across baseline and intervention conditions and evaluate whether changes correspond systematically with planned intervention changes.
Unlike a conventional parallel-group experiment, the primary experimental contrasts often occur within cases over time.
However, many single-case experimental studies include several participants or units to support replication.
The term single-case therefore describes an experimental logic centered on individual cases rather than a requirement that exactly one participant be studied.
How is a single-case experiment different from a case study?
A case study is a broad research approach involving detailed investigation of a bounded case within its context.
Case studies may be descriptive, interpretive, explanatory, or otherwise organized around different research objectives.
A single-case experimental design specifically seeks to establish an intervention effect through systematic manipulation and repeated measurement.
| Feature |
Case study |
Single-case experimental design |
| Primary purpose |
Understand or explain a bounded case. |
Evaluate the effect of a manipulated intervention. |
| Intervention manipulation |
Not necessarily required. |
Central to experimental control. |
| Repeated outcome measurement |
May or may not be used. |
Fundamental to the design. |
| Experimental contrasts |
Not inherently required. |
Established through planned intervention conditions or timing. |
| Typical inference |
Contextual, descriptive, interpretive, or explanatory. |
Evidence about functional relationships between intervention and outcome. |
A richly documented case study can be scientifically valuable without being experimental. Conversely, a single-case experiment can provide evidence about intervention effects without attempting to explain every contextual aspect of the case.
The distinction is explored further in single-case studies versus single-case experimental designs.
When is a single-case experimental design especially useful?
These designs may be useful when relatively few eligible participants are available, interventions are individualized, and repeated outcomes can be collected reliably.
They are commonly associated with behavioral, educational, rehabilitation, and applied intervention research.
For example, researchers may investigate whether a structured prompting intervention increases independent task completion among students requiring individualized instructional support.
They may also study changes in communication behavior, classroom engagement, or other repeatedly observable outcomes.
However, small population size alone is not sufficient justification.
The intervention must be suitable for systematic manipulation, and the outcome must permit meaningful observation across experimental conditions.
Why is repeated measurement essential?
Repeated measurements allow researchers to examine the level, trend, variability, and temporal pattern of an outcome before and after intervention changes.
For example, measuring task completion once before and once after an intervention provides only two observations.
If performance improves, researchers cannot readily determine whether the change reflects the intervention, an existing upward trend, or ordinary fluctuation.
Repeated baseline measurements provide information about the outcome's behavior before intervention.
Repeated intervention measurements then allow examination of whether the outcome changes in a manner consistent with the experimental manipulation.
However, repeated measurement alone does not establish experimental control. The intervention must be manipulated according to a design capable of producing a credible contrast.
The role of measurement frequency is examined in repeated measurement in single-case experimental research.
What is experimental control in single-case research?
Experimental control concerns the ability to demonstrate that changes in the outcome are systematically associated with intervention manipulation rather than plausible alternative explanations.
For example, if several participants show little change during baseline but improve when an intervention is introduced at different prespecified times, the pattern may strengthen evidence of an intervention effect.
The staggered introduction helps distinguish the intervention from events occurring simultaneously across all participants.
However, the strength of the inference depends on the design, measurement quality, intervention fidelity, and observed data patterns.
Experimental control is not established merely by labeling study phases A and B.
Researchers must show that the planned design provides sufficient opportunities to demonstrate an effect.
What is an AB single-case design?
An AB design contains a baseline phase, labeled A, followed by an intervention phase, labeled B.
The researcher measures the outcome repeatedly during both phases.
This arrangement can describe changes associated with intervention introduction.
However, a simple AB design generally provides limited evidence of experimental control because it contains only one transition from baseline to intervention.
Other events occurring at the same time may explain the observed change.
Researchers should therefore distinguish an AB phase comparison from stronger single-case experimental arrangements incorporating replication or additional planned contrasts.
What are ABA and ABAB designs?
Withdrawal or reversal designs introduce and withdraw intervention conditions in a planned sequence.
An ABA design contains baseline, intervention, and a return to baseline or withdrawal condition.
An ABAB design adds a second intervention phase.
When an outcome changes systematically with intervention introduction and withdrawal, the repeated contrasts may strengthen evidence of a functional relationship.
However, these designs require careful consideration of reversibility.
For example, withdrawing a treatment may not reverse knowledge already learned or skills already acquired.
Withdrawal may also be ethically inappropriate when an intervention provides essential support.
The distinctions among AB, ABA, and ABAB designs should be evaluated before selecting a phase structure.
When is a multiple-baseline design preferable?
A multiple-baseline design introduces an intervention at different times across participants, behaviors, or settings while repeatedly measuring outcomes.
For example, three students may begin an intervention at staggered points, with baseline measurements continuing for students who have not yet received it.
If changes occur following intervention introduction rather than at the same calendar time, the pattern may provide evidence supporting a functional relationship.
Multiple-baseline designs can be useful when withdrawing the intervention is undesirable or when treatment effects are not readily reversible.
However, baseline lengths and intervention timing must be planned carefully.
The design also requires sufficiently independent tiers or an appropriate understanding of dependencies among them.
See when multiple-baseline designs are preferable for the detailed decision.
Can single-case experiments use randomization?
Yes. Randomization can be incorporated into selected single-case designs.
For example, researchers may randomly determine intervention start points from a prespecified set of permissible occasions.
Randomization may support design-based inference when the assignment mechanism is clearly specified and followed.
However, randomization is not required in every single-case experimental tradition, and it does not automatically resolve weak measurement or insufficient experimental contrasts.
Researchers should distinguish random assignment of intervention timing from random sampling of participants.
These procedures serve different inferential purposes.
How many participants are needed?
There is no universal participant count appropriate for every single-case experimental design.
Some designs investigate one participant through multiple intervention reversals or other experimental contrasts.
Others use several participants, behaviors, or settings to demonstrate replication.
What matters is whether the design offers sufficient opportunities to demonstrate the intervention effect and whether the resulting evidence supports the intended inference.
Standards such as those developed by the What Works Clearinghouse specify requirements for design features and demonstrations of effects in educational single-case research.
Researchers should consult the standards relevant to their intended evidence classification rather than assuming that any small number of participants is sufficient.
The question of how many replications establish convincing effects requires attention to both design standards and the observed evidence.
How stable should baseline measurements be?
Baseline observations should provide enough information to evaluate the outcome's existing pattern and make the intervention contrast interpretable.
However, a perfectly flat baseline is not always required.
Some outcomes fluctuate naturally, and trends may be present before intervention.
Researchers should examine baseline level, trend, variability, and the direction of any existing change.
For example, a rapidly improving baseline may make subsequent improvement difficult to attribute to the intervention.
Extending baseline measurement may help clarify the pattern, but it can also delay potentially beneficial treatment.
The decision should balance inferential needs, participant welfare, and the expected outcome process.
How are single-case experimental data analyzed?
Visual analysis has a central role in many single-case experimental traditions.
Researchers examine changes in level, trend, variability, immediacy, overlap, and consistency of patterns across comparable phases.
Graphs should represent observations in their actual temporal sequence and identify intervention changes clearly.
Statistical approaches may supplement visual analysis, including randomization tests, effect-size estimators, time-series methods, and other design-appropriate procedures.
However, no single statistical method is appropriate for every single-case design.
Autocorrelation, baseline trends, unequal phase lengths, and nonindependent observations may complicate estimation.
Researchers should select analyses consistent with the experimental structure and intended estimand.
Why is intervention fidelity important?
Intervention fidelity concerns whether the intervention was implemented as intended.
For example, a prompting strategy may be introduced inconsistently across sessions or delivered differently by instructors.
If outcomes do not change, the absence of improvement may reflect ineffective intervention content, inadequate implementation, or both.
Similarly, improvements during baseline may occur if participants inadvertently receive elements of the intervention before its formal introduction.
Researchers should define the intervention operationally and establish appropriate fidelity monitoring.
Fidelity measurements should be sufficiently independent of outcome assessment where feasible.
Strong experimental logic requires confidence that the planned condition changes actually occurred.
Can single-case experiments establish causation?
They can provide evidence of causal effects when experimental manipulation, repeated measurement, and replicated contrasts support a functional relationship.
However, the strength of the inference depends on the specific design and its implementation.
A simple improvement after treatment introduction is not sufficient by itself.
Researchers should consider plausible alternative explanations, including maturation, historical events, measurement changes, and coincident interventions.
Experimental control is strengthened when outcome changes correspond systematically with planned intervention changes across sufficiently informative opportunities.
Even a convincing within-case effect does not automatically establish the same effect for every other individual or setting.
Can single-case findings generalize beyond the participants studied?
Potentially, but generalization requires evidence beyond a single observed effect.
Replication across participants, behaviors, settings, and occasions may strengthen confidence that an intervention effect is not unique to one case.
However, the population represented by the studied cases should be considered.
For example, an intervention effective for three students with particular learning needs may not necessarily have the same effect for students with substantially different characteristics.
Researchers should distinguish demonstrating a functional relationship within the studied cases from estimating an average treatment effect in a broader population.
The scope of inference should reflect the replication evidence and the similarities between studied and target settings.
What ethical issues arise?
Single-case experimental designs may involve delaying, withdrawing, or alternating interventions.
These manipulations require careful consideration when participants depend on the intervention for essential support or when withdrawal could produce harm.
For example, withdrawing a successful behavioral support intervention may be inappropriate if doing so exposes a participant to substantial risk.
Multiple-baseline or other arrangements may offer alternatives when reversal is unethical or impractical.
However, delaying intervention introduction can also raise ethical concerns.
Researchers should justify the timing and manipulation of conditions, obtain appropriate consent and review, and establish safeguards for participant welfare.
Experimental control should not be pursued without regard to the consequences for participants.
When should researchers avoid single-case experimental designs?
A single-case experiment may be unsuitable when the outcome cannot be measured repeatedly, the intervention cannot be manipulated meaningfully, or the research question requires a population-level estimate that the design cannot support.
It may also be inappropriate when outcome changes occur over very long periods and repeated experimental contrasts are impractical.
For example, studying a rare event that can occur only once may not permit the repeated phase-based demonstrations used in many single-case designs.
Similarly, an intervention producing irreversible outcomes may prevent a valid withdrawal design.
Researchers should not choose single-case experimentation solely because conventional sample size requirements are difficult to meet.
The design must fit the outcome, intervention, and intended inference.
How should researchers determine whether the design is feasible?
Begin by defining the intervention, target behavior or outcome, and expected timing of effects.
Determine whether the outcome can be measured frequently and reliably without excessive burden or reactivity.
Evaluate whether intervention introduction, withdrawal, alternation, or staggered timing can produce credible experimental contrasts.
Then assess the number of cases, phases, measurement occasions, and replications needed under relevant methodological standards.
Researchers should also plan visual and statistical analysis, intervention fidelity assessment, missing-data procedures, and ethical safeguards.
Where these requirements cannot be satisfied, another design may provide stronger evidence despite the small eligible population.
