Library / research-methodology-guide
Understanding reproducibility
Audience: patients-public, students-learners ยท Review state: published
Two words, two tests
Reproducibility and replicability are often used loosely, but a major United States National Academies report gives them distinct meanings. Reproducibility means getting consistent results when the same data are reanalysed with the same methods. Replicability means getting consistent results when a new study collects new data to answer the same question.
Both matter: the first checks the analysis, the second checks whether the finding survives contact with the world again.
Why results fail to repeat
Common causes are mundane rather than scandalous: small samples, flexible analysis choices made after seeing the data, incomplete method descriptions, unshared code and data, and publication bias that leaves failed replications invisible.
Deliberate misconduct exists but explains only a minority of failures. The larger problem is a system that rewards novel positive results more than careful, checkable ones.
What makes work reproducible
The practical toolkit is well established: pre-registered hypotheses and analysis plans, shared data and code where ethics allow, exact method reporting, version-controlled workflows, and computing environments that others can rebuild.
When another team can rerun the analysis end to end and reach the same numbers, the result is reproducible; when independent teams gather fresh evidence and find the same effect, it is replicated. Confidence grows with each successful round.
Reading claims with this lens
Ask whether the study's data and code are available, whether its analyses were pre-registered, and whether independent replications exist. A single un-replicated finding is a hypothesis with evidence, not settled knowledge.
Large collaborative replication projects in psychology and cancer biology have shown that many celebrated results do repeat, but a substantial share do not, which is exactly why the checking matters.
Limitations
Educational overview of reproducibility concepts, not an assessment of any specific study or field.