Replication

Editorial approval scope: The team checked current source fidelity, scope boundaries, dates, arithmetic, reader usefulness, practical workflow, and general-information limitations.

Repetition becomes evidence through design

Doing the same activity twice is not automatically a replication. A credible replication creates another opportunity to observe whether the predicted pattern appears while alternative explanations are controlled as well as the design allows.

Define what repeats: a phase contrast, effect across behaviors, participants, settings, implementers, or an independent study. State what counts as a successful result before looking at the data.

Single-case designs build replication into the study

Reversal designs may repeat a condition change and predicted response change. Multiple-baseline designs stagger intervention across participants, settings, or behaviors, creating repeated demonstrations at different times. Alternating-treatment designs compare repeated condition patterns.

The WWC Single-Case Design Technical Documentation discusses causal questions, threats to internal validity, design standards, visual analysis, and criteria for evidence of a relation. The exact replication logic depends on the design.

Within-study and across-study replication differ

Within-study replication strengthens the study’s internal demonstration. Across-study replication tests whether the finding appears with a new sample, team, setting, or research group.

A direct replication keeps key methods close to the original. A systematic replication changes selected features to test generality or theory. The boundary is often a matter of degree, so authors should list what stayed constant and what changed.

A fictional multiple-baseline example

Three fictional community classes teach the same accessible help-request routine. Baseline begins at the same time, while partner training starts in weeks 2, 4, and 6.

Class A rises from 1 of 6 partner responses to 5 of 6 after training. Class B stays near 1 of 6 until its own training, then reaches 6 of 7. Class C stays near 2 of 7 until its training, then reaches 5 of 7.

The staggered pattern offers three demonstrations aligned with intervention timing. A full analysis still examines level, trend, variability, immediacy, overlap, fidelity, concurrent events, and every data point. The counts alone cannot certify a functional relation.

Replication needs comparable implementation

Document participants, setting, materials, instructions, dose, implementer qualifications, prompts, measurement, and deviations. If the second team changed several active components, the result may test a different intervention.

Fidelity does not mean rigidly ignoring client needs. Record adaptations, consent, assent, access supports, and safety changes, then interpret which procedure was actually tested.

Failed replications are informative

A result that does not repeat can reveal a boundary condition, implementation difference, measurement problem, chance finding, or weakness in the original claim. Report it with the same care as a successful replication.

Avoid turning “failure” into blame. Check procedure clarity, competence, context, participant fit, health, communication, access, data quality, and whether the original effect was overstated.

Generality requires planned variation

Repeated success with one participant and one clinician supports a narrow conclusion. To examine generality, studies can vary age, communication, culture, diagnosis, setting, implementer, materials, response, or outcome while preserving the theoretical relation.

Representation matters. Ten studies with similarly selected participants may leave major groups and contexts untested. Report who was included, excluded, and missing.

Quantity cannot repair shared bias

Several replications can repeat the same confound, weak measure, selective outcome, or researcher allegiance. Independent teams, preregistered methods, open materials, transparent deviations, and complete reporting strengthen the evidence.

Publication bias can leave failed or smaller effects unseen. A literature containing only positive papers may look more replicated than the complete research record.

Apply replicated evidence carefully

Replication increases confidence within the studied conditions. It does not guarantee benefit for one person. Compare the person, goal, setting, supports, burden, risk, and desired outcome with the evidence.

Use individual assessment and ongoing data. A well-replicated procedure can still be a poor choice when the goal lacks client value, the method conflicts with assent, or the setting cannot deliver it safely.

Plan the replication before collecting data

Write the predicted pattern, design, participant and setting criteria, procedure, outcome, phase-change rules, fidelity checks, analysis, and success criteria in advance. Preserve the original materials and version every adaptation.

If the replication tests a published study, identify any unavailable information and contact authors when practical. Registering the plan and reporting all outcomes reduce the chance that a disappointing pattern disappears. Share enough data and materials for the next team to understand what was actually repeated.

Set a decision rule for inconclusive data. Extending a phase, changing a measure, or excluding a participant after results appear can turn ordinary uncertainty into a misleading success.

After completion, report deviations, missing data, implementation fidelity, adverse events, and every planned outcome, including results that do not reproduce the original pattern. This record helps the next reviewer distinguish a true difference from a changed method.