Glossary
Pre-registration
What is Pre-registration?
Pre-registration is the practice of recording a study’s hypotheses, methods, and analysis plan before observing the relevant results. It creates a time-stamped record that helps distinguish planned tests from data-informed exploration.
How it works
A pre-registered study commits to specific hypotheses, sample size, outcome measures, exclusion criteria, and statistical tests before observing the relevant outcomes. With existing data, researchers must disclose what they already know.
Deviations should be identified and explained so readers can assess whether they were outcome-informed. This does not eliminate flexibility (researchers can still explore their data) but requires that exploration be labeled as such.
A dated original plan and a documented amendment history can make later changes detectable. Registrations can be embargoed, and amendment rules differ; registration does not make undisclosed changes impossible.
Applied example
In a hypothetical study, a researcher pre-registers that they will test whether a gratitude intervention reduces depression scores on the PHQ-9, with 100 participants, using an independent samples t-test. If the result is non-significant but an unplanned analysis shows an effect on a different measure, the non-significant primary result must still be reported.
Pre-registration, HARKing and p-hacking
Comparing the plan with the report can help reveal hypotheses presented as predictions only after results were known, often called HARKing, or undisclosed searches across analyses for favorable results, often called p-hacking. This depends on a sufficiently clear plan and transparent reporting; a registration alone cannot rule either practice out.
Why it matters
Nosek and colleagues’ paper explains the distinction between prediction and postdiction, including applications with existing data. A registered plan still needs sound design, measurement, analysis, and transparent reporting.
What makes a plan useful?
“We will test whether the program improves wellbeing” leaves most analytical choices unresolved. A useful preregistration specifies the primary outcome and time point, the comparison, the sample or stopping rule, exclusions, handling of missing values and the main analysis. It should be detailed enough for someone to distinguish the intended test from a later choice.
In the hypothetical gratitude study above, stating “PHQ-9 at four weeks, comparing assigned groups” would identify a different test from “whichever mood measure changes most at any follow-up.” A defensible sample-size rationale is still needed; merely registering 100 participants does not establish that this number is sufficient. If a planned measurement becomes impossible, document the change and when it was made, report the limitation and distinguish the replacement analysis.
The Center for Open Science's guidance separates data-independent plans from data-informed exploration and includes registration before analyzing existing data. Neither registration nor a badge establishes that a program works. For Registered Reports, the additional feature is peer review of the plan and a conditional publication decision before the main results are known.