Knowledge base Updated 2 November 2026 1 views

Preregistration

Specifying a study's hypotheses, design and analysis plan in a time-stamped, read-only record before the data are collected or examined, so that confirmatory claims can be distinguished from exploratory ones.

Definition

A preregistration records, before data are seen: the hypotheses, the sampling plan and stopping rule, the variables and how they will be computed, the exclusion criteria, the analysis including how each hypothesis maps to a specific test, and the handling of missing data.

It does not forbid exploratory analysis. It requires that exploration be labelled as exploration.

A Registered Report goes further: the introduction and method are peer reviewed and accepted in principle before data collection, and publication does not depend on how the results come out.

History

Clinical trial registration was mandated first, in medicine, after selective reporting of trial outcomes was documented repeatedly. The ICMJE required registration for publication from 2005.

Psychology adopted the practice much later, driven by the replication crisis. The Open Science Framework, AsPredicted and the Registered Reports format all date from the 2010s, and uptake has grown quickly from a very low base.

Theory

The mechanism is straightforward: it removes researcher degrees of freedom that would otherwise be exercised after seeing the data, and it makes HARKing — presenting a post hoc hypothesis as a prior one — detectable.

The deeper point concerns what a p-value means. Error rates are only valid for a test specified in advance. A test chosen after looking at the data does not have the properties the test statistic claims, whatever the arithmetic says. Preregistration is what makes the inferential machinery honest, not merely what makes the researcher honest.

Registered Reports additionally remove publication bias at source, because the accept decision precedes the result.

Research

The strongest evidence comes from comparing outcomes across formats.

Scheel and colleagues compared Registered Reports with conventionally published psychology papers and found positive results in about 44 percent of Registered Reports against about 96 percent in the standard literature. That gap is the size of the filter that was operating.

In medicine, trials registered after mandatory registration show a marked drop in positive findings — the well-known analysis of NHLBI cardiovascular trials found large null-result increases after 2000.

Audits also find that a substantial share of preregistrations are incomplete or are deviated from without disclosure, so the practice is not self-enforcing.

Applications

Any confirmatory study. Preregistration is cheap, and for experimental work it costs little more than writing the method section early.

For observational and secondary-data work, preregistration is harder but possible; blinded analysis and held-out data serve similar purposes.

As a reader: check that a registration exists, then check that the reported analysis is the registered one. The second step is the one that gets skipped and it is the step that matters.

Criticism

Deviation without disclosure is common, which undermines the guarantee.
Vague registrations provide little constraint and can give false assurance to readers who do not check.
It can be misread as a ban on exploration, which would be a real loss; discovery matters and does not need to be disguised as confirmation.
Poorly suited to some designs, particularly qualitative work, long-running cohort analyses, and computational modelling where the analysis develops with the model.
It does not fix bad theory or bad measurement. A precisely preregistered test of an incoherent hypothesis is still worthless.

References

Nosek, B. A., et al. (2018). The preregistration revolution. Proceedings of the National Academy of Sciences, 115(11), 2600-2606.
Chambers, C. D. (2013). Registered Reports: A new publishing initiative at Cortex. Cortex, 49(3), 609-610.
Scheel, A. M., Schijen, M. R. M. J., & Lakens, D. (2021). An excess of positive results. Advances in Methods and Practices in Psychological Science, 4(2).
Kaplan, R. M., & Irvin, V. L. (2015). Likelihood of null effects of large NHLBI clinical trials has increased over time. PLoS ONE, 10(8), e0132382.
Claesen, A., et al. (2021). Comparing dream to reality: An assessment of adherence of the first generation of preregistered studies. Royal Society Open Science, 8(10).

Written for PsychologyOrbit and reviewed before publishing. It is a summary for professionals and students, not clinical guidance, and it is not a substitute for the primary sources listed above.