Carryover effect

Carryover changes what a later phase means

Single-case designs compare repeated measurements across conditions. Interpretation depends on knowing which condition plausibly controls the observed pattern. When an earlier condition leaves a residual influence, a later phase is no longer a clean test of its own procedure.

Suppose a learner acquires a communication skill during teaching. Returning to baseline conditions will rarely erase that skill. High responding after withdrawal can reflect durable learning rather than failure to implement the withdrawal. The later data still matter, though the original prediction of reversal no longer fits.

Other carryover can be temporary. Fatigue from one condition may suppress responding in the next session. A dense reinforcement schedule may affect responding after the schedule changes. Repeated exposure to test materials may create practice effects.

Carryover, maintenance, and sequence effects differ

Maintenance describes continued performance after intervention changes or ends. It is often a desired outcome. Carryover effect describes how an earlier condition influences interpretation of a later one. The same persistent skill can be desirable clinically and problematic for a reversal comparison.

A sequence effect is the broader influence of condition order. Carryover is one mechanism that can produce a sequence effect. Contrast, fatigue, practice, sensitization, and changing preferences can also make A followed by B differ from B followed by A.

Use exact language in a report. “The skill maintained after teaching” states an outcome. “Prior teaching may have carried into the withdrawal phase” states a design limitation.

Look for evidence in timing and response form

Possible signals include:

• abrupt changes that follow a prior phase rather than the current procedure
• a gradual washout pattern after a condition ends
• different results when conditions occur in another order
• persistent prompted or practiced response forms
• phase differences that shrink across repeated comparisons
• data changes after medication, staffing, setting, or schedule exposure

These patterns support hypotheses. They do not identify a cause alone. Check procedural integrity, measurement reliability, opportunity counts, session spacing, health and contextual events, and the exact transition between conditions.

Choose designs with behavior change in mind

Reversal logic works best when the dependent variable can reasonably change with introduction and withdrawal of the independent variable. Learned skills, irreversible environmental changes, and procedures with lasting effects may call for multiple-baseline, changing-criterion, or other designs.

In multielement or alternating-treatment arrangements, rapid switching can create interaction across conditions. Distinct stimuli, counterbalanced order, adequate spacing, limited session exposure, and separate materials may reduce some risks. Each control should fit the actual behavior and procedure.

The BACB BCBA Test Content Outline covers internal and external validity, single-case design features, interpretation, and comparative, component, and parametric analyses. It is exam content rather than a design protocol.

A fictional teaching sequence

Maya is learning to request a break using speech or AAC. During four baseline probes with communication access, she independently requests in 1 of 12 opportunities. During eight teaching sessions, she requests in 20 of 24 opportunities.

The team then conducts five planned probes without teaching prompts while continuing to honor clear communication. Maya requests in 12 of 15 opportunities. That persistence is useful for Maya. It also means the post-teaching phase cannot recreate the original baseline history.

The report keeps all three denominators and states that the later level is consistent with learning carried from teaching. It avoids claiming that the withdrawal phase independently verified a functional relation. The team uses replication across other targets and settings to answer later questions.

Report the limitation rather than hiding it

Record condition order, start and end dates, transition rules, spacing, materials, implementers, prompts, unexpected exposures, and procedural changes. Graph phase boundaries clearly. If a planned comparison becomes uninterpretable, preserve the data and revise the question prospectively.

The WWC Single-Case Design Technical Documentation describes repeated measurement, internal-validity standards, visual analysis, and evidence of a relation. Its 2010 framework helps explain why phase history and replication matter. It is technical review guidance rather than individual clinical authorization.

Ethical care remains primary. A team should never remove necessary communication, safety, health, or dignity supports to create a visually cleaner reversal.

Questions for design review

Before interpreting a phase comparison, ask:

1. Could the prior condition plausibly produce durable learning or residual effects?
2. Was the current phase long enough to show its own pattern?
3. Were opportunities and measurement comparable?
4. Did order, fatigue, practice, health, or setting change?
5. Which replication can answer the question safely?