What Is Preregistration, and What Problem Is It Supposed to Solve?

What Does It Actually Mean to Preregister a Study?

You may encounter papers that describe a study as “preregistered,” journals that award badges for preregistration, or research methods courses that recommend registering a study before collecting data. That can make preregistration sound like another administrative requirement attached to research.

It is more useful to understand the problem it is trying to address.

Researchers make many decisions while designing, conducting, analyzing, and reporting a study. Some are made before the results are known. Others emerge after researchers have seen patterns in the data. Without a record of those earlier decisions, readers may have difficulty telling the difference.

Preregistration creates that record. It does not prevent researchers from learning from their data. Rather, it can make the chronology of their decisions more visible.

Preregistration Records the Plan Before the Relevant Results Are Known

Why Preregistration Exists in the First Place

Research Decisions Are Often More Flexible Than They Look in the Final Paper

A published study usually presents a fairly orderly account. There is a research question, a method, an analysis, and a conclusion. The actual research process may have involved considerably more uncertainty.

Researchers might need to decide which outcome will receive the most attention, when recruitment should stop, which observations should be excluded, how variables should be constructed, whether covariates should be included, which statistical model should be used, or which subgroup analyses are worth pursuing.

Many of these choices have legitimate alternatives. That is not inherently a problem. The difficulty arises when researchers can make consequential choices after seeing the data and the final report does not reveal that those choices were informed by the observed results.

This is sometimes discussed in terms of researcher degrees of freedom: the collection of defensible choices available during the research process. Flexibility is often necessary, but undisclosed flexibility can make evidence harder to evaluate because readers do not know how many alternative analytical paths might have been available.

Preregistration Creates a Record That Exists Before the Outcome

The Open Science Framework describes preregistration as posting a time-stamped version of a study plan before beginning the relevant data collection or analysis. The timing is crucial. A plan written after the results are known cannot establish which decisions were made independently of those results.

A preregistration might specify research questions, hypotheses, sampling procedures, variables, exclusion rules, stopping criteria, outcome measures, and planned analyses. Exactly what should be preregistered depends on the design and on which decisions matter for the claims the researchers intend to make.

The resulting record provides a reference point. When the study is eventually reported, researchers and readers can compare what was planned with what actually happened.

One Problem Is Presenting a Discovery as Though It Were a Prediction

Suppose a researcher measures several variables and finds an unexpected association between two of them. There is nothing improper about investigating that association. Unexpected observations are often where interesting research begins.

But imagine that the eventual paper presents the association as though the researcher had predicted it before examining the data. The reader is now evaluating a different evidential story from the one that actually occurred.

This practice is related to what has been called hypothesizing after the results are known, or HARKing. The broader issue is not simply whether a hypothesis was written before or after analysis. It is whether readers can tell which claims represent prospective tests and which arose through exploration.

Preregistration can provide evidence about that chronology because the earlier plan has already been time-stamped.

Another Problem Is Undisclosed Analytical Flexibility

Consider a dataset that could reasonably be analyzed in several ways. Researchers might use different exclusion thresholds, alternative outcome definitions, multiple covariate specifications, transformations, subgroup definitions, or statistical models.

If researchers try several approaches and report only the one that produces the most favorable result, readers may see a single apparently straightforward analysis without knowing about the alternatives.

Advance specification can reduce this ambiguity. Researchers can identify the primary analysis before observing the relevant results and then distinguish that analysis from additional analyses conducted later.

This does not mean every analytical contingency can or should be predicted. Effective preregistration often requires specifying the decisions that are consequential for the primary claims and, where feasible, stating contingency rules for foreseeable problems.

Preregistration Can Also Make Selective Reporting More Visible

Selective reporting can occur when researchers collect multiple outcomes, test multiple hypotheses, or conduct numerous analyses but emphasize only a subset in the eventual publication.

A preregistration does not physically prevent selective reporting. What it can do is leave evidence that other prespecified outcomes or analyses existed. If a registered primary outcome disappears from the paper, for example, a reader who consults the preregistration may notice the omission.

This distinction is important. Preregistration is primarily a transparency mechanism. Its existence does not guarantee compliance with the plan.

The Deeper Distinction Is Between Prediction and Exploration

Nosek and colleagues have argued that preregistration helps distinguish prediction from postdiction. A prediction exposes an expectation to the possibility of being wrong before the outcome is known. An explanation constructed after seeing the outcome has access to information that the original prediction did not.

Both activities matter. Science needs researchers to test expectations, and it also needs them to notice surprising patterns and develop new explanations.

The difficulty comes when those modes of inquiry become indistinguishable in the final report. Preregistration can help preserve the distinction between exploratory and confirmatory research without requiring researchers to choose one and abandon the other.

Preregistration Is a Plan, Not a Promise That Nothing Will Change

A preregistration records intentions at a particular point in time. It does not guarantee that those intentions will remain methodologically appropriate.

Perhaps recruitment becomes impossible under the original sampling plan. A measurement problem emerges. An analytical assumption is violated. Researchers discover an error in the preregistration itself. Better methodological information becomes available.

In such circumstances, blindly following the original plan can be worse than changing it.

The relevant principle is transparency. Researchers can preserve the original registration, explain what changed and why, and distinguish the revised analysis from the one originally planned. This is why a preregistered study can still change.

Preregistration Is Not Necessarily Public Immediately

Researchers sometimes hesitate to preregister because they do not want to disclose a new idea before conducting the study. Public visibility and time-stamped registration are related but distinct issues.

Some registries allow an embargo. For example, the Open Science Framework currently permits researchers to submit a registration while keeping it private for a defined embargo period. The record can therefore establish when the plan was submitted without immediately revealing its substantive content publicly.

Repository rules can change, so researchers considering an embargo should verify the current options and conditions directly with the registry they intend to use.

Preregistration Is Not the Same Thing as a Registered Report

Preregistration does not ordinarily involve a journal evaluating the plan before the research proceeds. Researchers create the advance record, conduct the study, and later decide where to submit the resulting manuscript.

A Registered Report adds prospective peer review. The proposed study is submitted to a participating journal before the main results are known, and successful Stage 1 review may lead to in-principle acceptance.

The two practices therefore overlap but should not be conflated. The difference between preregistration and a Registered Report becomes especially important when deciding whether you want only an advance record or also methodological review before conducting the research.

Preregistration Does Not Solve Every Threat to Research Quality

A precise record of a weak design is still a weak design.

Preregistration does not ensure valid measurement, adequate sampling, appropriate statistical reasoning, ethical research conduct, accurate data collection, sensible interpretation, reproducibility, or theoretical importance. It cannot make an underpowered or conceptually confused study rigorous simply by documenting it in advance.

Nor does preregistration completely eliminate researcher discretion. Plans differ considerably in specificity, and unforeseen decisions may still arise during the study.

The appropriate claim is therefore narrower: preregistration can make some important research decisions more transparent and can help readers distinguish planned work from later exploration. Whether that ultimately improves a particular study depends on how well the practice is implemented and how the resulting evidence is interpreted.

How the Same Finding Can Mean Something Different With an Advance Record

Misconceptions About What Preregistration Does

Think of Preregistration as a Record of Decision Timing

If you are considering preregistration, ask which decisions in your study would be easier for another researcher to interpret if their timing were known.

The practical objective is not to demonstrate obedience to an earlier version of yourself. It is to leave enough information for readers to reconstruct which important decisions preceded the evidence and which decisions were informed by it.

That is a more modest goal than eliminating bias, but also a more defensible one.

Before You Call a Study Preregistered

Frequently Asked Questions About Preregistration

Preregistration Makes the History Behind the Results More Visible

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.