Deciding to submit a Registered Report is one decision; actually writing a Stage 1 protocol that survives review is a different problem, with its own craft. This guide covers that second problem: how to draft the hypotheses, methods, and analysis plan of a Stage 1 manuscript so that it holds up under review conducted before any data exist. For the mechanics of the format itself — what in-principle acceptance guarantees, how Stage 2 review differs, which journals offer the track — see CASRAI’s companion guide, Registered Reports: How Preregistered Study Plans Get Accepted Before Data Collection, which this page assumes as background rather than repeats.
The practical guidance below draws substantially on Henderson & Chambers’ “Ten Simple Rules for Writing a Registered Report” (PLOS Computational Biology, 2022), the most widely cited practical writing guide for the format, together with the general structure of Stage 1 review described in CASRAI’s format guide above.
Before you start drafting
Three things need to be settled before you write a word of the Stage 1 manuscript itself:
- Pick your review route. Decide between a journal’s own in-house Registered Reports track (for example Nature Human Behaviour) or a journal-independent service like PCI Registered Reports, which reviews Stage 1 and Stage 2 on preprints outside any single journal’s editorial office. Each has its own author guidelines and, often, its own required Stage 1 template — read it before drafting, not after.
- Decide when you’ll seek ethics approval. Some journals require IRB/ethics-committee approval already in hand at Stage 1 submission; others accept a Stage 1 submission with approval pending and require it before data collection begins. This affects your drafting timeline, since ethics committees may themselves want to see a near-final protocol.
- Understand what cannot change later. Once a Stage 1 manuscript is in principle accepted, its stated hypotheses, methods, and confirmatory analysis plan are locked. Anything not specified at Stage 1 can still be reported at Stage 2, but only as clearly labelled exploratory work that carries less evidentiary weight. Write with that constraint in mind from the first draft, not as an afterthought.
Step 1: Write hypotheses that are explicit, numbered, and testable
Vague or implicit hypotheses are a common source of Stage 1 revision requests. State each hypothesis as a discrete, numbered prediction (H1, H2, H3…) rather than folding several predictions into a paragraph of prose. A workable hypothesis specifies:
- The direction or pattern predicted, not just that “an effect” will occur.
- The specific comparison or relationship the data need to show to count as support.
- What result would count as disconfirming it — a hypothesis a reviewer cannot imagine failing is not doing any work.
Where the predicted pattern depends on conditions, write it as an explicit if-then statement (for example: “if condition A produces a larger effect than condition B, this supports H1; if the effect sizes are statistically indistinguishable, this supports H2 instead”) rather than leaving the decision rule implicit for the results section to reveal. This is the single habit that most directly forecloses HARKing risk, because it removes the discretion a results-first framing would otherwise leave for reinterpreting the hypothesis after the fact — see CASRAI’s HARKing entry for the underlying problem this addresses.
Step 2: Motivate the question without leaning on the eventual result
A Stage 1 Introduction has to justify why the question is worth asking using only what is already known — prior theory, prior findings, an identified gap or contradiction in the literature — because no result yet exists to justify it retroactively. Two habits that read poorly to Stage 1 reviewers, precisely because they smuggle in outcome-dependent framing:
- Writing as though a particular finding is expected and building the argument around that expectation, rather than building the argument around why the question needs answering regardless of which way it comes out.
- Deferring the significance of the work to “results will clarify X” instead of stating plainly, before data exist, why either a positive or a null result would be informative.
A useful test while drafting: could this Introduction be published unchanged if the eventual result came out the opposite way? If not, the framing is doing work that belongs in the Discussion, after Stage 2, not in the Stage 1 Introduction.
Step 3: Specify methods precisely enough for someone else to run the study
Stage 1 methods sections are held to a higher bar of exhaustiveness than a conventional manuscript’s, because reviewers are evaluating whether the design can actually answer the question — the only thing left to evaluate, since there are no results yet. Henderson & Chambers’ guidance is to write with enough precision that an independent researcher could carry out the study without needing to contact the authors for clarification. In practice that means specifying, in advance and in order:
- The full procedure and materials, including exact wording of any manipulation, instructions, or stimuli where feasible.
- Inclusion and exclusion criteria, and critically, the order in which exclusion rules are applied when more than one could apply to the same case.
- Randomization and blinding procedures, stated explicitly rather than assumed.
- Any planned manipulation check or “outcome-neutral” check — a check that confirms the study procedure is working as intended (for example, that a manipulation was noticed, or equipment is calibrated) without revealing anything about the substantive hypothesis. Include these deliberately; their absence is a recurring point of Stage 1 reviewer feedback, because without them a null result is ambiguous between “the effect isn’t there” and “the manipulation didn’t work.”
Step 4: Justify the sample size with an a priori power analysis
An undefended sample size is one of the most common reasons a Stage 1 submission comes back for revision. A defensible justification needs three things stated explicitly, not just a target N:
- The statistical test or model the power calculation corresponds to, matching what the analysis plan actually specifies.
- The effect size the calculation is built on, sourced from prior literature or theory rather than assumed — and, per Henderson & Chambers, anchored to a conservative estimate rather than the largest published one, since published effect sizes are systematically inflated by publication bias toward significant findings.
- What the study can and cannot detect at the resulting sample size, including how a null result at that sample size would still be informative (a sensitivity or precision analysis alongside the power analysis is common practice here, and directly answers a question reviewers will otherwise ask themselves).
Step 5: Write the analysis plan so it leaves no discretion at Stage 2
The analysis plan is what Stage 2 review checks the completed study against, so ambiguity written into it at Stage 1 becomes a problem at Stage 2, not before. Specify the exact statistical test or model, the software and version where relevant, the variables entering it, and any data transformation or coding decisions, in enough detail that two different analysts running your plan on the same dataset would reach the same output. Clearly separate this confirmatory plan from any analyses you already know you want to run but that don’t bear directly on the stated hypotheses — those can be flagged in the Stage 1 manuscript as planned exploratory analyses, which keeps them legitimate to report at Stage 2 without letting them count as confirmatory evidence they were never positioned to provide. This is the same discipline that keeps a fixed, pre-specified analysis plan from sliding into p-hacking after the fact.
Step 6: Decide contingencies now, not during analysis
Real data rarely come out exactly as planned — missing values, an ambiguous distribution, a manipulation check that partially fails. The Stage 1 protocol is the place to decide, in advance, how those situations will be handled, using explicit if-then contingency statements: if assumption X is violated, then analysis Y will be used instead; if attrition exceeds Z%, then the pre-registered exclusion rule applies as follows. Writing these into the Stage 1 manuscript, rather than leaving them to be decided once the data are visible, is what keeps a deviation from becoming a disclosure problem at Stage 2 — a contingency that was already approved isn’t a deviation at all.
Step 7: Map hypotheses to design, analysis, and interpretation explicitly
Reviewers at Stage 1 are checking that the logic connecting question, hypothesis, method, and analysis actually holds together — not just that each piece is individually well written. A short summary table that lines up each numbered hypothesis with the specific analysis that tests it and the specific pattern of results that would support or refute it makes that logic checkable at a glance, and is standard practice recommended in the Registered Reports writing literature. It is worth building even if your target journal doesn’t require it as a formal element, because assembling it is also how an author catches a hypothesis with no corresponding analysis, or an analysis with no corresponding hypothesis, before a reviewer does.
Common reasons Stage 1 protocols get sent back for revision
Across the practical writing guidance and journal author instructions for the format, the same handful of issues recur:
- Hypotheses stated too vaguely to be clearly supported or refuted by a specific result.
- A sample size target with no power analysis behind it, or a power analysis built on an optimistic rather than conservative effect-size estimate.
- Procedural detail thin enough that a reviewer cannot tell exactly what will be done, or in what order exclusion criteria will be applied.
- No outcome-neutral or manipulation check, leaving a potential null result uninterpretable.
- An analysis plan with genuine researcher degrees of freedom left in it — unspecified model choices, transformations, or covariates that could be decided after seeing the data.
- Confirmatory and exploratory analyses not clearly distinguished from one another.
None of these are usually fatal on their own — Stage 1 review typically proceeds through one or more revision rounds, the same as conventional peer review, per CASRAI’s format guide. But each one specifically targets the thing Stage 1 review exists to check, so they tend to draw a revision request rather than a minor comment.
After in-principle acceptance
Once a Stage 1 manuscript passes review, three things typically happen in sequence: the approved protocol is deposited on a public registry, usually the Open Science Framework (OSF), at which point it also functions as the study’s formal preregistration; data collection begins under the approved plan; and any unanticipated deviation from that plan needs to be documented and, where substantial, approved by the editor before it’s simply written into the Stage 2 submission. When it comes time to write Stage 2, the Introduction and Methods carry over from the approved Stage 1 manuscript largely unchanged (conventionally shifted from future to past tense), with Results and Discussion added — not rewritten as a fresh manuscript reacting to what was found.
Frequently asked questions
Can I change my hypotheses after Stage 1 acceptance if the literature moves on?
Not within that Registered Report. The hypotheses reviewed and accepted at Stage 1 are what Stage 2 review checks the study against; a genuinely new hypothesis belongs in a new study, reported as exploratory within the current one, or as the basis for a separate Registered Report.
Do I need a professional statistician to write the power analysis?
Not necessarily, but the analysis needs to correspond exactly to the test specified in the analysis plan and use a defensible, literature-sourced effect size — getting this wrong is one of the most common sources of Stage 1 revision requests, so it is worth the same scrutiny as any other part of the protocol, including a statistician’s review where the design is unfamiliar.
How long should a Stage 1 protocol be?
Journals vary and generally publish their own word-count guidance for Registered Reports specifically (often close to, but not identical to, their standard article limits) — check the target journal’s or PCI Registered Reports’ own author guidelines rather than assuming a generic manuscript length applies.
What happens if I can’t specify every detail in advance, for example because the study depends on real-time developments?
Write explicit if-then contingencies for the specific uncertainties you can anticipate, and disclose which decisions genuinely cannot be pre-specified. Some journals and PCI Registered Reports accommodate this for particular study types; it needs to be addressed openly in the Stage 1 manuscript rather than left implicit.
Related CASRAI resources
- Registered Reports: How Preregistered Study Plans Get Accepted Before Data Collection — the format explained: Stage 1/Stage 2 mechanics, in-principle acceptance, participating journals, and how it differs from plain preregistration.
- Registered Report — the dictionary definition.
- Preregistration of a Study Protocol: What It Is and How to Do It — registry-based preregistration mechanics.
- HARKing (Hypothesising After Results are Known)
- P-hacking
- Metascience
