01 · The Question
Does Preregistration Mean You Cannot Change Your Research Plan?
You preregister a study, begin collecting data, and discover that an important assumption was wrong. Perhaps recruitment is slower than expected, a measurement instrument proves unreliable, or the planned statistical model is inappropriate for the observed data structure. Can you change the plan without undermining the purpose of preregistration?
The confusion comes from treating preregistration as an irreversible commitment to every methodological detail. Its central purpose is different: to establish what researchers intended before obtaining information that could influence their decisions. Understanding which commitments require particular protection, and how subsequent changes should be reported, is essential to interpreting the resulting evidence.
02 · The Short Answer
What Must Stay Fixed, and What Can Be Revised?
In Brief
The original, time-stamped preregistration must remain an accurate historical record, but the research plan itself may change when there is a defensible reason. Decisions defining confirmatory hypotheses, primary outcomes, eligibility criteria, sample size, stopping rules, and planned analyses require particular transparency when revised.
Changes should be documented with their rationale, timing, and implications for interpretation. A revision made before relevant results are known is different from one selected after examining those results. Neither should be concealed or presented as though it formed part of the original plan.
03 · What You Need to Know
How Fixed Commitments and Legitimate Flexibility Work
The Original Registration Is Fixed, Not Necessarily Every Research Decision
Preregistration creates a time-stamped account of planned research decisions. Its value lies partly in allowing readers to compare those intentions with what researchers eventually conducted and reported.
On platforms such as the Open Science Framework (OSF), a submitted registration preserves the original record, while subsequent updates can document changes. AsPredicted likewise protects submitted registrations against ordinary editing, although its procedures for recording subsequent revisions differ.
This distinction matters. Changing a study procedure does not require pretending the original plan never existed. Instead, the research record should preserve both the original commitment and the reason for departing from it.
Fixed historical record
What was specified at the original registration date remains identifiable and attributable to that date.
Revisable research plan
Study procedures and analyses may evolve, provided consequential changes are disclosed and interpreted appropriately.
Which Decisions Require the Strongest Protection?
Decisions are especially consequential when changing them could alter the evidence used to support a hypothesis, increase the probability of obtaining a favorable result, or change the population to which the findings apply.
Researchers should therefore pay particular attention to prespecified design and analysis decisions , including the following.
Decision
Why consistency matters
Possible justified change
Confirmatory hypotheses
Establishes which predictions preceded the results.
Clarifying a hypothesis before relevant evidence is examined.
Primary outcome
Limits selective emphasis among measured outcomes.
Replacing an unusable measure for a documented technical reason.
Eligibility criteria
Defines the intended study population.
Correcting an operationally unworkable criterion before recruitment decisions are affected.
Sample size and stopping rules
Protects against outcome-dependent recruitment decisions.
Revising recruitment targets because of documented feasibility constraints.
Exclusion and missing-data rules
Limits selective inclusion of observations.
Addressing an unforeseen data-quality problem through a documented procedure.
Primary analysis
Defines how the principal hypothesis will be evaluated.
Using a justified alternative when planned model assumptions are untenable.
These are not universally immutable categories. Their importance depends on the study design, inferential goals, and the information available when the change is proposed.
What Can Usually Change With Less Risk to Confirmatory Interpretation?
Some decisions are primarily operational and may have little bearing on the study's inferential claims. Examples include adjusting interview appointments, replacing a malfunctioning recording device with an equivalent one, modifying internal project-management arrangements, or correcting typographical errors in working documentation.
Nevertheless, even an apparently minor operational change can become methodologically important. Switching an online questionnaire to an interviewer-administered format, for example, might alter how participants respond. Changing data collection sites could affect the composition of the sample.
The relevant question is not simply whether a decision is operational. It is whether the change could influence measurement, selection, exposure, analysis, or interpretation.
Timing Matters: Before Data, During Collection, or After Results?
The evidential status of an amendment depends partly on what researchers already knew when making it.
Timing
Interpretive concern
Recommended treatment
Before data collection and access to relevant outcomes
Usually lower risk of outcome-driven selection.
Record a dated amendment or new linked registration.
During collection, before relevant outcome information is examined
Potential influence from recruitment or operational information.
Document what information was available and why the change was needed.
After inspecting relevant outcome patterns
Greater risk of selecting decisions based on observed results.
Disclose the departure and distinguish revised or exploratory analyses from original confirmatory tests.
Access to data is not necessarily equivalent to knowledge of the relevant results. A blinded data-quality assessment may reveal missing values or implausible measurements without revealing treatment differences. Even then, researchers should explain the blinding arrangements and the information used to justify the revision.
Does Every Deviation Turn an Analysis Into an Exploratory Analysis?
No. The implications depend on the nature of the deviation and whether the revised decision could have been influenced by the results it is intended to evaluate.
Correcting a programming error to implement the originally specified statistical test is different from selecting a new test because it produces a smaller p-value. Likewise, substituting a robust estimator in response to an objectively documented assumption violation may be defensible, although the revision should still be reported.
Researchers should avoid treating confirmatory and exploratory analyses as mutually exclusive labels for an entire paper. A study may contain originally planned confirmatory tests, transparently amended analyses, and additional exploratory investigations. Each deserves an accurate description.
How Should You Document a Preregistration Amendment?
A useful amendment record should enable another researcher to reconstruct the decision without relying on an informal explanation written after the study is complete.
For each consequential change, record the original specification, revised procedure, reason for the revision, date of the decision, stage of data collection or analysis, information already examined, and anticipated implications for inference.
Where possible, preserve a linked, time-stamped amendment through the registration platform. OSF supports registration updates while preserving the original record. If a platform does not support comparable amendments, a separate dated document or linked registration may be appropriate.
Watch Out
A new time-stamped amendment does not retroactively establish that a decision was made before the original data were observed. Report when the revised decision was made and what relevant information was already available.
What About Ethics Approval, Clinical Trials, and Registered Reports?
Preregistration transparency does not replace other governance requirements. Changes involving participant safety, consent, recruitment, interventions, or sensitive data may require approval from an ethics committee or other responsible authority before implementation.
Clinical trial registrations may also be subject to registry-specific update requirements, regulatory obligations, and protocol amendment procedures. A public registration update should not be assumed to satisfy all of these obligations.
Registered Reports involve an additional commitment through journal peer review. Following in-principle acceptance, consequential changes to an approved protocol ordinarily require discussion with the journal. The relevant question is addressed separately by the procedures for design changes after Stage 1 acceptance .
04 · A Practical Example
Changing a Planned Analysis After Discovering a Measurement Problem
Hypothetical Example
A Preregistered Study of Online Learning
A research team preregisters a randomized study comparing two instructional approaches. Its primary outcome is a standardized posttest score, and its planned analysis is a regression model adjusting for baseline achievement.
During data collection, the team discovers that a software error caused one posttest item to be scored incorrectly for an identifiable subset of participants.
Original commitment
The preregistration specifies the posttest scoring procedure and the primary regression analysis.
Problem identified
The team establishes the scoring error through technical logs, before examining outcome differences between randomized groups.
Decision and documentation
Researchers correct the scoring algorithm, document affected records, retain the original scoring specification, and create a dated amendment describing the correction.
Reporting and interpretation
The publication identifies the original plan and correction, explains that treatment-group differences had not been inspected when the decision was made, and evaluates whether the correction materially changes the results.
The correction need not automatically invalidate confirmatory interpretation. It addresses an objectively identified measurement error rather than selecting an analysis based on which result is favorable.
Suppose, however, the researchers had compared several scoring methods after examining treatment effects and retained the one producing statistical significance. That decision would introduce a substantially different inferential concern. It should not be represented as an unchanged preregistered analysis.
06 · What This Means for You
How to Decide Whether and How to Change Your Plan
Before implementing a revision, assess its methodological necessity, timing, and possible consequences for the original research question. A decision should not be preserved merely because it was preregistered if doing so would produce invalid measurements or an inappropriate analysis.
A simple decision framework
If the change corrects an objective error before relevant outcomes are examined
Document the correction promptly and explain why the original procedure was unsuitable.
If the change affects the primary outcome, sample, stopping rule, or planned test
Assess its inferential implications, preserve the original commitment, and report the amendment prominently.
If the change is motivated by observed results
Identify the analysis as outcome-informed and avoid presenting it as an unchanged prospective test.
If the change affects ethics approval or an accepted Registered Report
Consult the responsible review body or journal before proceeding where approval is required.
It is often useful to retain the originally planned analysis alongside the revised analysis when both remain interpretable. This can help readers assess the sensitivity of conclusions to the departure, although it may not be feasible or scientifically meaningful in every situation.
For future projects, the strongest approach is to prespecify recruitment and stopping decisions and other consequential rules clearly enough that researchers can identify deviations without ambiguity. Appropriate contingencies can be specified in advance without attempting to predict every possible problem.
07 · A Quick Checklist
Before Changing a Preregistered Decision
Before implementing a consequential amendment, check:
Identify the exact original decision and locate its time-stamped registration.
Establish why the original procedure is no longer appropriate or feasible.
Record whether relevant outcomes or group differences have already been examined.
Assess effects on hypotheses, measurement, selection, statistical inference, and interpretation.
Check whether ethics, regulatory, or journal approval is required.
Create a dated amendment or linked record without concealing the original specification.
Consider reporting original and revised analyses when both are meaningful.
Explain all consequential departures in the final research report.
11 · Cite this Guide
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.
Recommended (Field Guide)
APA
MLA
Chicago
Copy Citation
