03 · What You Need to Know
How to Distinguish Decisions That Must Be Fixed From Those That Can Remain Flexible
What Does Prespecification Mean in a Research Protocol?
Prespecification means establishing a research decision or decision rule before the information that could improperly influence that decision becomes available.
For example, identifying the primary outcome before examining outcome data helps distinguish the intended confirmatory analysis from later exploratory comparisons.
Similarly, specifying participant eligibility criteria before recruitment can reduce the risk of selecting participants according to emerging findings.
However, not every operational decision serves the same inferential purpose.
Choosing which approved meeting room to use may be inconsequential in one study, while the environment of an assessment room may be central to another.
The importance of prespecification therefore depends on what the decision could change.
Why Is Prespecification Important?
Prospective decisions can reduce opportunities for selectively choosing methods or analyses after observing results.
For example, a researcher who examines several outcomes and subsequently labels the most favorable one as primary may create a misleading impression of confirmatory evidence.
Prespecification can also improve operational consistency by preventing personnel from independently resolving consequential questions in incompatible ways.
Nevertheless, prespecification does not guarantee validity. A poorly justified analysis remains unsuitable even when planned in advance.
Its principal value is to clarify commitments, support transparency, and distinguish planned decisions from subsequent adaptations.
Which Decisions Usually Deserve Strong Prespecification?
Decisions should generally be established prospectively when they affect the study's central scientific comparisons or participant protection.
| Decision area |
Typical prospective commitment |
Why it matters |
| Research design |
Study structure, comparison conditions, and allocation procedures. |
Defines the intended inference. |
| Eligibility |
Participant inclusion and exclusion criteria. |
Protects the definition of the study population. |
| Primary outcomes |
Outcome definition, measurement, and timing. |
Distinguishes planned confirmatory evidence. |
| Sample size |
Target size, justification, and applicable stopping rules. |
Limits inappropriate data-dependent decisions. |
| Intervention |
Essential components and comparison condition. |
Protects the meaning of the intervention contrast. |
| Analysis |
Principal estimand, models, exclusions, and relevant decision rules. |
Supports transparent interpretation. |
| Participant protection |
Consent, confidentiality, and critical safety requirements. |
Establishes ethical boundaries. |
The required level of detail varies across designs. For example, qualitative inquiry may appropriately specify sampling principles rather than a fixed final sample size.
Which Operational Decisions Can Often Remain Flexible?
Routine decisions may remain open when their alternatives do not materially affect the approved methodology or participant protection.
Examples may include assigning qualified personnel to particular sessions, selecting among approved rooms, arranging administrative meetings, or using equivalent authorized equipment.
However, the same decision may be consequential in another study.
For example, room selection could matter substantially in research examining the influence of classroom noise on attention.
Researchers should therefore avoid categorizing decisions as trivial solely because they appear administrative.
What Is the Difference Between a Fixed Decision and a Prespecified Decision Rule?
A fixed decision identifies the action that will be taken. A decision rule establishes how an action will be selected when specified circumstances arise.
Fixed decision
The procedure or analytical choice is established before implementation.
Prespecified decision rule
The conditions and criteria governing a future choice are established prospectively.
Permitted operational discretion
Personnel may choose among authorized alternatives within defined boundaries.
Unplanned methodological change
A consequential decision is revised outside the prospective plan and requires appropriate justification and disclosure.
For example, a protocol may specify a fixed assessment duration while allowing a repeat session under predefined conditions following a technical failure.
The repeat session is not necessarily an unplanned methodological decision when its conditions were established beforehand.
Should Recruitment Decisions Be Fixed?
Eligibility criteria and the intended recruitment population should generally be established before recruitment begins.
However, operational recruitment arrangements may need adjustment when participation is slower than anticipated.
For example, a study may permit additional approved recruitment announcements without changing who is eligible.
Expanding recruitment to a substantially different population could have more serious methodological and ethical implications.
The protocol should distinguish permissible recruitment adjustments from changes requiring amendment.
Should Sample Size and Stopping Rules Be Prespecified?
For confirmatory quantitative research, sample-size decisions and stopping rules are often important prospective commitments.
Unplanned decisions to stop or extend collection after examining outcome patterns can affect statistical inference.
However, some designs legitimately use adaptive sample-size procedures when supported by appropriate statistical methods and prospective decision rules.
Qualitative studies may instead use iterative sampling principles informed by emerging analysis.
Researchers should distinguish these methodological arrangements rather than impose one universal stopping rule.
The principles for preregistering sample size and stopping rules provide further guidance.
How Should Measurement Decisions Be Prespecified?
Measurement decisions that define primary outcomes or important comparisons should generally be established before relevant outcome information is examined.
These may include the instrument, scoring procedure, measurement occasions, and treatment of incomplete assessments.
For example, changing a primary achievement assessment after observing unfavorable results could compromise the distinction between confirmatory and exploratory evidence.
By contrast, replacing malfunctioning equipment with a previously approved equivalent device may be an operational adjustment.
Researchers should evaluate whether the replacement preserves the intended measurement rather than assume equivalence.
Can Qualitative Research Prespecify Decisions Without Becoming Rigid?
Yes. Qualitative research can specify methodological commitments and decision principles while preserving appropriate iterative development.
For example, a grounded theory study may establish the logic of theoretical sampling without identifying every eventual participant.
A semi-structured interview study may specify core domains while permitting responsive probing.
In reflexive thematic analysis, researchers may identify their theoretical orientation and intended analytical practices without fixing all codes and themes in advance.
Prespecification should support methodological transparency rather than impose procedures inconsistent with the research approach.
The distinction is particularly important when managing research procedures that cannot be fully standardized.
What About Mixed-Methods Research?
Mixed-methods studies may combine fixed quantitative commitments with more flexible qualitative procedures.
For example, a sequential explanatory design may specify that qualitative interviews will follow the quantitative analysis to investigate particular patterns.
The precise interview questions or participant selection may legitimately depend on the quantitative results.
However, the protocol should establish the purpose of that dependence and the principles governing subsequent decisions.
Such an arrangement differs from retrospectively inventing a qualitative component solely to support a preferred conclusion.
Can Operational Decisions Be Delegated to Research Personnel?
Yes, when responsibilities and decision boundaries are clear.
For example, a site coordinator may be authorized to reschedule an assessment within an approved window.
However, the coordinator may not have authority to change the primary outcome definition or eligibility criteria.
Delegation should distinguish routine implementation from consequential methodological changes.
Where a decision falls outside delegated authority, personnel should seek appropriate review rather than improvise a permanent solution.
How Can a Protocol Specify Flexibility?
A protocol can identify the circumstances under which flexibility is permitted, the acceptable alternatives, and the authority responsible for selecting among them.
For example, it may specify that assessment sessions can be rescheduled within a defined window when technical disruptions prevent completion.
It should also establish how the rescheduling will be documented and whether additional conditions apply.
When an alternative could affect measurement comparability, the protocol may require further review.
These arrangements should be based on the study's scientific and ethical needs rather than arbitrary administrative convenience.
Should Every Possible Contingency Be Anticipated?
No. Researchers cannot reasonably predict every operational event.
However, foreseeable consequential problems should be addressed through contingency plans or decision principles.
For example, equipment failure, recruitment shortfalls, missing primary outcomes, and participant protection concerns may justify advance planning.
The protocol should also identify who may authorize responses to unexpected situations.
The broader question of including contingency plans in a research protocol is relevant when unexpected problems could alter implementation.
What Happens When an Unplanned Change Becomes Necessary?
Unplanned changes should be evaluated according to their scientific, ethical, and operational implications.
For example, replacing an assessment instrument may affect measurement comparability and require a protocol amendment.
Changing a routine administrative arrangement may be permissible under existing procedures.
Researchers should document the reason for consequential changes, the information available when the decision was made, and the relevant approval process.
Where participant protection is immediately at risk, urgent protective action may be necessary under applicable governance arrangements.
Subsequent reporting or review may still be required.
How Does Prespecification Relate to Preregistration?
Preregistration provides a time-stamped record of selected research plans before relevant data or results are examined.
A research protocol may contain additional ethical and operational information that is not included in a preregistration.
Conversely, a preregistration may specify analytical decisions in greater detail than the main institutional protocol.
Researchers should ensure consistency between documents where they address the same commitments.
Consequential changes should be disclosed rather than silently incorporated into later descriptions of the original plan.
How Does Prespecification Differ From Registered Reports?
Registered Reports involve prospective peer review of research questions, design, and analysis plans, typically before the principal results are known.
After in-principle acceptance, consequential methodological changes may require editorial consultation under the journal's policies.
However, routine operational adjustments do not necessarily alter the accepted scientific plan.
Researchers should distinguish changes to essential methodological commitments from ordinary implementation decisions.
Where changes affect the accepted design, the principles for managing design changes after Stage 1 acceptance may apply.
Can Prespecification Become Counterproductive?
Yes. Excessive prespecification can impose unnecessary rigidity when the research requires contextual responsiveness or when operational alternatives are scientifically equivalent.
For example, fixing every qualitative interview probe before data collection may undermine the purpose of semi-structured inquiry.
Similarly, requiring a formal protocol amendment for every change of approved room may create administrative burden without improving the evidence.
However, flexibility should not be used to justify selective changes after examining results.
The appropriate balance depends on the decision's scientific consequences and the methodology's assumptions.
How Should Researchers Document Decisions Made During Implementation?
Consequential decisions should be traceable through appropriate records.
These may include protocol amendments, deviation logs, implementation records, reflexive memos, sampling decisions, or approved SOP revisions.
Documentation should establish what changed, why, when, and under whose authority.
For confirmatory studies, it may also be important to establish whether relevant outcome information was available when the decision was made.
The level of documentation should be proportionate to the decision's implications.
How Do You Decide Whether a Decision Must Be Fixed?
Consider whether the decision could alter the research question, participant population, measurement, intervention, analysis, or participant protection.
Also consider whether the decision could be influenced by emerging results in ways that affect the credibility of confirmatory claims.
If either concern is substantial, stronger prespecification or a prospective decision rule may be appropriate.
If several alternatives are scientifically equivalent and fall within approved boundaries, operational discretion may be sufficient.
The protocol should establish who is authorized to make the decision and how its consequences will be evaluated.
Watch Out
Do not describe a methodological decision as prespecified merely because it was documented before the final manuscript was written. The relevant timing concerns when the decision was made relative to access to information that could influence it. Decisions made after examining relevant results should be disclosed appropriately, even when scientifically justified.
