A grant proposal sent back after revision tends to score better than the same idea sent in cold. At the US National Institutes of Health, resubmissions have cleared review at roughly 30 to 40 percent in recent years, against 15 to 20 percent for first submissions. The gap is real and persistent. What causes it is far less settled.

The Resubmission Advantage

The pattern shows up wherever funders publish outcome data by submission type. Revised applications win more often than new ones, and the margin is large enough to matter for anyone planning a career around soft money. It holds across institutes with very different pay lines and across years when overall budgets rose or fell.

The effect is modest in any single panel but stubborn in aggregate. A proposal that fails on the first pass and returns with revised aims, new preliminary data, and a point-by-point response sits somewhere in the range of twice as likely to fund. That is not a rounding error. It shapes how labs sequence their submissions and how long a project waits before it starts.

There is an uncomfortable reading of this. If revision helps that much, the first review may be measuring familiarity as much as quality. Panels reward proposals they have already argued about once, which is a strange thing for a system that claims to reward the best new ideas.

The mechanics of the NIH process matter here. Most applications that miss the funding line on the first pass are allowed a single resubmission, and the agency treats that second look as a distinct category in its own reporting. That administrative choice is what makes the comparison visible at all. Agencies that fold resubmissions into a single pool, or that allow unlimited revision cycles, give researchers no clean way to see whether a second attempt pays off. The transparency of the NIH numbers is a side effect of how the paperwork is structured, not a deliberate experiment.

What the Data Shows

The clearest numbers come from NIH RePORTER, the agency's public database of applications and awards. Because submission type is recorded, anyone can pull success rates for new and resubmitted applications by institute, year, and study section. Analyses of these records consistently find resubmissions outperforming first submissions by a wide margin.

The size of the gap varies. Some institutes report resubmission success near 40 percent while first submissions land in the mid-teens. Others show a narrower spread. The direction almost never reverses, which is the part worth taking seriously: this is not one institute's quirk.

Pay lines complicate the picture. A funder that sets a hard percentile cutoff will fund a fixed fraction of scored applications regardless of type, so the raw rates partly reflect how many resubmissions reach the scoring range at all. Selection starts before the panel meets, and the published rates carry that history inside them.

There is a second arithmetic trap. The overall success rate for any cycle is a weighted blend of new and revised applications, and the weights shift year to year as budgets change. A rising headline rate can mean the agency is funding more first submissions, or it can mean a larger share of the applicant pool is resubmitting after a prior miss. The two stories imply opposite things about how hard it is to enter the system, and the aggregate figure cannot tell them apart. Only the split-by-type table can.

Reviewer Psychology and Score Drift

A resubmission arrives with baggage the panel can see. Reviewers get the revised aims, the introduction explaining what changed, and often the prior critiques. That context does two things at once. It gives reviewers a scaffold for reading the proposal, and it gives them a story about a team that listened.

Familiarity probably softens initial skepticism. A reviewer encountering an unfamiliar model system on a first submission has to build confidence from scratch; on a resubmission, the same reviewer has already worked through the objections and can check whether they were addressed. The second read is easier, and easier reads tend to score better.

Anchoring may compound this. If reviewers recall a previous score, even vaguely, the new score can drift toward it. Panels rarely see prior scores formally, but study section members rotate slowly and memory travels. No published experiment has isolated how much of the resubmission advantage comes from anchoring rather than genuine improvement.

The cognitive work of a first read is heavier than it looks from outside. A reviewer facing an unfamiliar proposal has to reconstruct the applicant's reasoning, judge whether the methods fit the question, and estimate whether the team can deliver, all inside a few hours spread across a stack of a dozen or more applications. On a second pass, much of that reconstruction is already done. The revised proposal gets read against a model the reviewer built months earlier, and models are easier to update than to build. That asymmetry is a feature of how panels work, not a flaw anyone introduced on purpose, which is exactly why it is hard to legislate away.

Novelty Penalty for First Timers

Unfamiliar proposals get harder scrutiny. A reviewer who does not recognize the approach, the assay, or the investigator has to be persuaded on the page, and persuasion takes space the applicant may not have. Proposals that look like work the panel already trusts move through more smoothly.

New investigators carry the cost. Someone submitting a first application has no prior review to revise from, no established relationship with the study section, and often a thinner preliminary dataset. The resubmission advantage is, in practice, an advantage for people already inside the system long enough to have failed once and come back.

Survey work on reviewer behavior has documented a persistent tension between innovation and feasibility. Reviewers say they want bold work; their scoring tends to reward work that looks likely to finish. A first submission has to clear both bars without the benefit of a second hearing.

The early-career arithmetic is worth spelling out. A lab that has never held a major award has no prior summary statement to learn from, and the first critique it receives is also the first real signal about how the panel reads its work. Established labs enter each cycle with a mental map of which study section to target, which framing has landed before, and which reviewers are likely to be sympathetic. None of that is corrupt. It is accumulated information, and it compounds. The cost falls hardest on exactly the applicants a funder most wants to reach.

This site has followed adjacent questions about how funding structures shape what gets finished, including a piece on model organism findings publishing faster than the ecology grants supporting them. The resubmission gap is a different mechanism but the same underlying issue: the review process leaves marks on the science that follows it.

Limits of the Evidence

Every number above is observational. Nobody has randomized applications to first-submission or resubmission tracks and compared outcomes, and it is hard to imagine a funder agreeing to try. Without that, the comparison is between two groups that differ in ways the data do not capture.

Resubmitters are self-selected. An applicant who revises has decided the idea is worth another year, has usually collected more data, and has had months to sharpen the argument. The revised proposal may simply be better, in which case the higher success rate reflects improvement rather than bias. Separating those explanations requires information the public records do not contain.

There is also a survivorship problem. Proposals that scored badly enough on the first pass are often never resubmitted, so the resubmission pool is already filtered for ideas that came close. Comparing that pool to all first submissions overstates the advantage of revising.

A cleaner test would compare revised proposals against new proposals on closely matched topics, with similar preliminary data and comparable teams. Nothing in the public record does this at scale. The closest available substitute is to look within a single study section and a single cycle, where the pay line and the reviewer pool are held roughly constant. Even there, the applicants who chose to resubmit differ from those who walked away, and the difference is unmeasured.

A related piece on this site argued that code repositories outlive the grants that funded their upkeep, which is a reminder that funding decisions have consequences stretching well past the review cycle. The resubmission literature rarely tracks those downstream effects.

What Funders and Applicants Can Do

Funders can publish success rates split by submission type for every study section, not just agency-wide. The data already exist in RePORTER; the useful step is making the comparison routine and visible to the people who sit on panels.

Panels can test blinded review of prior scores, where reviewers score revised aims before seeing the earlier critique. Some institutes have experimented with variants of this. The results are not widely reported, which is itself a problem.

Applicants can treat a first review as a diagnostic rather than a verdict. The specific objections in a summary statement are the most valuable document a lab receives about a project, and the response section is where the resubmission advantage is earned or lost.

There is a trade-off funders should weigh openly. Every allowance for revision consumes review capacity, and a system that encourages resubmission also encourages applicants to hold a project back for a cycle rather than send it early. Whether that delay buys better science or simply longer timelines is an empirical question nobody has answered. A funder could, in principle, compare matched cohorts of projects that submitted early against those that waited, and track output years later.

Researchers studying the research system can push for actual experiments: randomized submission tracks, matched comparisons of revised and new proposals on the same topic, follow-up on whether funded resubmissions deliver more or less than funded first submissions. The last question has barely been asked, and it is the one that would tell us whether the advantage is wisdom or inertia.