Major funders require applicants to address scientific rigor explicitly: how the study design controls for bias, why the sample size is adequate, how key resources will be authenticated, and how results will be reported. Each funder publishes its own criteria, and an application is assessed against that funder's rules, not a general standard.
Why did funders start asking about rigor?
Funders added explicit rigor requirements because a large share of published findings could not be reproduced, and the failures traced back to design decisions made long before publication. More than 70% of 1,576 researchers surveyed said they had tried and failed to reproduce another scientist's experiments, and more than half had failed to reproduce their own, in a survey published by Nature.
That result reframed reproducibility as a funding problem rather than a publishing one. A study that is underpowered, unblinded, or built on unauthenticated cell lines will produce an unreliable result no matter how carefully the paper is written afterwards. Those flaws are cheapest to fix before the money is committed, which puts the burden on the application.
The requirements that followed share a common shape across funders even though the wording differs. Applicants are asked to describe the scientific premise, including a critical assessment of the prior work the proposal rests on. They are asked to describe design features that reduce bias, typically randomisation, blinding, and predefined analysis. They are asked to justify sample size rather than assert it. They are asked to state how key biological and chemical resources will be authenticated. And they are asked to address whether relevant biological variables, including sex, have been considered.
None of that is a formatting exercise. Reviewers read those sections as evidence about whether the applicant will produce a result anyone can build on.
How do requirements differ between funders?
Each funder maintains its own criteria, and the differences are substantive enough that a proposal written for one cannot be resubmitted to another without rework. The practical point for a research office is that "meeting funder requirements" is never a single checklist.
The major funders whose requirement sets an institution typically deals with include:
- NIH, whose R01 and R21 mechanisms differ from each other in scope and in what the reviewer is asked to weigh, with the exploratory mechanism judged on feasibility rather than on preliminary data.
- ERC, whose Starting, Consolidator, and Advanced schemes are assessed primarily on excellence, with the applicant's track record weighted differently at each career stage.
- Horizon Europe, which layers programme-level requirements, including open science obligations, on top of the scientific case.
- Wellcome, the Gates Foundation, HHMI, and ISF, each with their own published criteria, formats, and emphases.
Two structural differences matter most when an office is reviewing a draft. The first is what the scheme is actually judged on: a mechanism that rewards feasibility punishes an application padded with preliminary data, and a mechanism that rewards excellence punishes one that reads as incremental. The second is the open science and data-sharing obligation, which varies from a brief statement to a detailed management plan with compliance consequences after the award.
Because the criteria are public, they can be checked systematically rather than from memory. That is the difference between an office that reviews applications against the rules they will actually face and one that reviews them against general good practice.
Where do rigor sections go wrong?
Rigor sections fail in predictable ways, and most failures are omissions rather than errors.
The most common is asserting rather than justifying. "The sample size is adequate to detect a meaningful effect" states a conclusion where a reviewer expects the reasoning: the effect size assumed, where that assumption came from, and the resulting power calculation.
The second is treating the scientific premise as a literature summary. Funders ask for a critical assessment of the prior work, including its weaknesses. An applicant who describes the foundational studies without noting that they were small or never independently replicated has not answered the question, and a reviewer who knows the field will notice.
The third is silence on authentication. Cell line misidentification and antibody variability are long-standing sources of irreproducible results, and a proposal that does not say how key resources will be validated leaves an obvious gap.
The fourth is reporting. Funders and journals both expect study designs to align with the relevant reporting guideline, and for health research the EQUATOR Network maintains the searchable library of those guidelines across the major study designs. An application that commits to a specific standard signals that the eventual paper will be complete.
The fifth is more subtle: describing rigor practices in the methods but not connecting them to the specific threats to validity in that study. Blinding matters because a particular readout is subjective. Randomisation matters because a particular confounder is plausible. Naming the threat makes the safeguard legible.
What can a research office check before sign-off?
A research office reviewing a draft before submission is checking the same things a reviewer will, with the advantage of time to fix them.
| Check | What good looks like |
|---|---|
| Scientific premise | Prior work cited and critically assessed, including its limitations |
| Design safeguards | Randomisation, blinding, and controls named against specific threats |
| Sample size | A calculation with stated assumptions, not an assertion |
| Resource authentication | Named method for validating cell lines, antibodies, and reagents |
| Relevant variables | Sex and other biological variables addressed or justified as not applicable |
| Reporting standard | The applicable guideline identified and committed to |
| Citations | Every reference resolves to the work it claims, none retracted |
The last row is worth its own attention. Applications carry substantial bibliographies, they are often assembled under deadline pressure, and a citation to a retracted paper in the premise section undermines exactly the part of the proposal that is meant to establish credibility. The odds of that happening keep rising: more than 10,000 research papers were retracted in 2023, a record annual figure, according to Nature. Checking a bibliography against the retraction record is now practical because Crossref acquired the Retraction Watch database in September 2023 and made it openly and freely available.
Fabricated citations deserve the same scrutiny, for a newer reason. A peer-reviewed test published in Scientific Reports found that a majority of the bibliographic citations generated by GPT-3.5 were fabricated, that GPT-4 still produced a substantial share of fabricated or erroneous citations, and that even citations to real works frequently contained substantive errors. Any part of an application drafted with a language model needs its references resolved individually before the office signs off.
An office running these checks consistently across departments is doing quality assurance on the institution's output, not policing individual investigators. The institutions that get the most from it treat the review as a service to the applicant, delivered early enough to change the draft. Offices that support this across labs and departments tend to run the same review on manuscripts later, which makes rigor a continuous expectation rather than a grant-season scramble. The evidence behind these requirements is worth reading directly: why so many experiments fail to reproduce.
How Octym helps
Octym reviews a grant application against the criteria of the funder it is going to, alongside checks on references, statistics, and methods completeness. Findings arrive as signals with their location and evidence attached, so the applicant and the research office can see what a reviewer would see while there is still time to act. Nothing is scored or certified, and every decision stays with the people submitting. You can see how a review is configured to a specific funder's requirements on the tailored page.