03 · What You Need to Know
When Preregistration Adds Something Useful to the Research
Preregistration Is Particularly Useful When the Study Makes Confirmatory Claims
The clearest case is a study designed to test a hypothesis specified before the relevant results are known.
Imagine that several outcomes could be analyzed, different observations could reasonably be excluded, multiple model specifications are defensible, and several stopping decisions are possible. Once the researchers see the data, those choices can potentially be influenced by the observed results.
A preregistration can specify the primary hypothesis, outcome, sampling or stopping rule, exclusions, and intended analysis in advance. Readers can then distinguish those decisions from alternatives developed after the evidence became available.
This is why preregistration is closely associated with the distinction between exploratory and confirmatory research. Nosek and colleagues argue that preregistration can clarify planned and unplanned research by reducing unnoticed flexibility and helping researchers calibrate the certainty of their claims.
The Potential Benefit Increases When Researchers Have Many Consequential Choices
Some studies involve substantial researcher degrees of freedom: reasonable alternative choices in study design, data processing, analysis, and reporting.
Consider an analysis in which researchers can choose among several outcome definitions, exclusion thresholds, transformations, covariates, subgroup specifications, and statistical models. Each individual choice may be defensible. Collectively, however, they can create many possible analytical paths.
Preregistration can be useful when it identifies which path was intended before researchers knew where those paths would lead.
The relevant issue is not the sheer number of decisions. It is whether those decisions could materially alter the claims and whether their timing matters for interpretation.
Preregistration Can Help Even When Bias Reduction Is Not the Main Objective
Advance planning can have practical benefits beyond distinguishing predictions from post hoc explanations.
Writing a preregistration may expose decisions that have not yet been resolved. Researchers may discover that they have not clearly defined the primary outcome, established a stopping rule, decided how a construct will be scored, or connected each hypothesis to an analysis.
Survey research on researchers' experiences with preregistration has found study planning among the commonly perceived benefits, alongside transparency and trustworthiness. These are perceptions rather than experimental demonstrations that preregistration necessarily improves every study, but they suggest that researchers may find value in the planning process itself.
In this sense, preregistration can function partly as a methodological stress test. Writing “we will decide later” looks rather more consequential once one has to specify what “later” will involve.
Experimental Research Is Not the Only Candidate
Many familiar preregistration practices were developed around relatively conventional hypothesis-testing designs. That history can create the impression that preregistration belongs exclusively to randomized experiments with straightforward statistical analyses.
Research practice is considerably broader.
Method-specific preregistration approaches have been developed for areas including secondary-data analysis, qualitative research, cognitive modelling, systematic reviews, and other designs. Crüwell and Evans, for example, argue that standard preregistration tools developed around conventional experimental psychology may not adequately capture the researcher degrees of freedom involved in cognitive modelling. Their response is not to abandon advance specification altogether, but to adapt the preregistration to the methodology.
This illustrates a useful principle: ask what decisions matter in the particular research design before deciding what preregistration should look like.
Exploratory Research Can Be Preregistered Without Pretending to Be Confirmatory
Exploratory research presents an apparent paradox. If the purpose is to discover patterns, generate hypotheses, or allow questions to develop through interaction with the evidence, how can you specify the research in advance?
You do not need to predict what you will discover.
You may instead document the initial research questions, data source, sampling approach, variables available, broad analytical strategy, decision rules, or boundaries of the intended exploration. The usefulness depends on the project and on what information an advance record adds.
What would be counterproductive is inventing precise hypotheses merely to make exploratory work resemble confirmatory research. Whether and how exploratory research can be preregistered therefore requires a different approach from preregistering a conventional hypothesis test.
Qualitative Research Can Benefit, but the Purpose May Be Different
Qualitative research is another important case because flexibility, interpretation, and responsiveness to emerging data can be integral to the methodology rather than problems that need to be eliminated.
Haven and van Grootel argue that preregistration can nevertheless contribute to qualitative research by documenting the development of the study and making methodological flexibility more visible. Subsequent work involving qualitative researchers produced an agreement-based qualitative preregistration template while explicitly emphasizing that preregistration should not eliminate qualitative research's capacity to adapt and respond.
The relevant advance decisions may concern the research questions, sampling rationale, data sources, researcher positioning, data-collection procedures, analytical approach, and anticipated changes rather than a fixed hypothesis and statistical test.
Researchers therefore need to ask whether qualitative preregistration fits their methodological approach, not whether qualitative work can be squeezed into a quantitative preregistration form.
Secondary-Data Research Creates a Different Timing Problem
Preregistration is often described as occurring before data collection. That formulation does not fit studies using data that already exist.
The important issue becomes what researchers know about the dataset before specifying their questions and analyses. If they have already inspected the outcome patterns, a new registration cannot retrospectively make those decisions independent of the observed evidence.
Nevertheless, researchers working with existing data may still prospectively specify analyses they have not yet conducted, particularly when they have not examined the relevant relationships. The preregistration should state honestly what information about the dataset was already available.
Here, preregistration can still add transparency, but the chronology is more complicated than “registered before data collection.”
Highly Emergent Research May Gain Less From Detailed Advance Constraint
Some research is intentionally designed to evolve through engagement with data, participants, cases, texts, settings, or theory. In such work, fixing too many decisions in advance may conflict with the methodology rather than improve it.
This does not necessarily mean that no prospective documentation is useful. Researchers might still record starting questions, selection principles, methodological commitments, or anticipated decision processes.
But the benefit of preregistration may be modest if most consequential decisions cannot meaningfully be made until researchers have engaged with the evidence.
A registration that merely fills a template with vague promises provides little transparency. A registration that artificially freezes decisions that the methodology requires researchers to develop iteratively may be worse.
Some Studies May Have Little Important Unobserved Flexibility to Constrain
Imagine a straightforward descriptive project in which the dataset, variables, calculation, and reporting procedure are already externally defined and there are few consequential analytical alternatives.
Preregistration might still provide a record of intentions, but the incremental benefit may be small because there is little ambiguity about which analytical decisions preceded the results.
The administrative effort of preregistration should therefore be weighed against the information it adds. Preregistration takes time, and empirical surveys of researchers have identified time and workload among its commonly perceived disadvantages.
This is not an argument against preregistration. It is an argument against treating it as a ritual whose value is independent of what the registration actually accomplishes.
A Poorly Matched Template Can Make Preregistration Less Useful
A standard template can be helpful because it reminds researchers of decisions they might otherwise overlook. But templates embody assumptions about how research works.
A form centered on directional hypotheses, fixed samples, dependent variables, and predetermined statistical tests may fit one research design well and another badly.
The growing availability of specialized templates reflects this limitation. The qualitative preregistration template developed by Haven and colleagues, for example, was created specifically to preserve the adaptability of qualitative inquiry while providing a systematic starting point.
The appropriate question is therefore not simply “Did you preregister?” but “Was the preregistration appropriate to this study?”
Preregistration Should Not Become a Proxy for Research Quality
Even where preregistration is highly suitable, its presence does not establish that the research is rigorous.
A preregistered hypothesis may be theoretically weak. A preregistered sample may be inadequate. A preregistered measure may lack validity. A preregistered statistical analysis may be inappropriate.
Preregistration can make the sequence of decisions more transparent, but it does not validate those decisions. This is why preregistration should not automatically be treated as evidence of research rigor.
Watch Out
A blanket requirement to preregister every project in exactly the same way can turn a transparency practice into a compliance exercise. The form and level of advance specification should reflect the study's inferential goals, methodology, and genuine decision points.
