The two main criteria that the National Science Foundation uses to evaluate grant proposals are Intellectual Merit and Broader Impacts. Every proposal submitted to NSF is judged against both, and a proposal that's strong on one but weak on the other is vulnerable no matter how good the science is. The current definitions come from the NSF Proposal & Award Policies & Procedures Guide (PAPPG), and reviewers score each proposal against them separately.

Understanding what each criterion actually asks for is the difference between a fundable proposal and one that gets returned without review or sinks in panel. The two criteria are weighted equally in principle, but they're evaluated through different evidence, and reviewers apply them with different instincts. Here's how each works, what reviewers look for, and where proposals most often fail.

How the Two Review Criteria Are Defined in the PAPPG

The PAPPG defines Intellectual Merit as the potential to advance knowledge, and it asks reviewers to consider five specific things: the importance of the idea, the creativity and originality of the approach, whether the plan is well-reasoned and organized, whether the proposers have the needed expertise and resources, and whether there's adequate access to resources. That's a deliberately broad list, and it covers both the idea and the team's capacity to execute it.

Broader Impacts is defined as the potential to benefit society and contribute to the achievement of specific, desired societal outcomes. The PAPPG lists the kinds of outcomes NSF has in mind: full participation of underrepresented groups, integration of research and education, broadening of participation, enhancement of infrastructure for research and education, broad dissemination to enhance scientific understanding, and benefits to society. Those aren't a checklist you must hit all of, but they show what reviewers are primed to recognize.

Both criteria get evaluated through two standard review elements: what the proposal aims to do and how well it's planned. Reviewers are told to assess the intellectual merit and broader impacts of the proposed activity, not the merit of the people involved in the abstract. A famous researcher with a thin plan and a junior investigator with a rigorous one are supposed to be judged on the proposal.

The definitions are stable across recent PAPPG editions, but the exact phrasing and the review questions have shifted modestly over time. The version that governs your submission is the one in effect for the deadline you're targeting, so check the current PAPPG on NSF's site rather than relying on a colleague's older copy.

What Intellectual Merit Actually Requires Reviewers to Judge

Intellectual merit is where most scientists feel at home, and it's also where overconfidence costs proposals. Reviewers are asked whether the work advances knowledge in the field, whether the idea is creative and original, whether the plan is logically organized and well-reasoned, whether the team has the expertise and resources, and whether the proposal has adequate access to needed resources. Five questions, and a proposal can stumble on any one of them.

The innovation question is the one people misread. Creativity and originality don't require a brand-new field. A well-designed test of an existing hypothesis, a methodological improvement, or a careful replication can be original if the proposal explains why it matters. The reviewer's question is what knowledge changes if this succeeds, and a proposal that can't answer that in a sentence is in trouble regardless of its technical sophistication.

The plan question is where experienced reviewers spend the most time. A proposal can have a brilliant idea and a plan that reads like a wish list. Reviewers look for a logic that connects the hypothesis to the methods to the expected outcomes, with milestones that make the work checkable. A proposal that describes what will be done but not how anyone will know it worked tends to get middling scores even when the science is sound.

Expertise is judged as a fit between the team and the work, not as a prestige ranking. A new investigator with the right skills and a credible plan can score highly. A senior team proposing work outside its demonstrated competence can score poorly. The proposal has to show the team can do this specific thing, not just that the team is accomplished.

What Broader Impacts Requires and Why It Sinks So Many Proposals

Broader impacts is the criterion that most often separates funded from unfunded proposals at the margin, and it's the one most often treated as an afterthought. NSF asks reviewers to judge the potential to benefit society and contribute to specific, desired societal outcomes. The PAPPG lists examples, and reviewers are trained to look for a genuine, planned activity rather than a sentence about how the work might someday help people.

The most common failure is the vague promise. "This work will train the next generation of scientists" without saying who, how many, through what mechanism, or how it'll be assessed reads as filler. Reviewers see hundreds of these. A specific, modest activity with a clear plan and a way to measure it scores better than an ambitious, vague one. A proposal that partners with a named community organization to run a defined workshop series, with attendance and evaluation plans, is stronger than one that gestures at outreach.

NSF doesn't require every proposal to hit every example on the list. It requires a credible, specific plan for societal benefit that's appropriate to the work and the investigators' capacity. A small team can propose a focused activity it can actually deliver. Reviewers penalize overpromising as much as underpromising, because a plan that can't be executed doesn't benefit anyone.

Broader impacts also gets scrutinized for fit with the proposed research. A plan that's bolted on and unrelated to the science reads as transactional. The strongest proposals connect the societal benefit to the work itself β€” the data, the methods, the partners, the communities affected. That connection is what makes the plan credible to a reviewer who's seen a lot of generic outreach paragraphs.

How Reviewers Score and Weigh the Two Criteria

NSF reviews typically use a scoring system where reviewers rate intellectual merit and broader impacts separately, often on a scale from excellent to poor, plus an overall rating. The exact scale varies by program and by the review system in use, but the separation matters: reviewers are asked to justify their rating on each criterion independently. A proposal with excellent intellectual merit and poor broader impacts gets a mixed review, and panels discuss those cases at length.

Panels don't simply average the two scores. A proposal that's excellent on one and poor on the other is a harder case than one that's good on both, because the criteria are meant to be complementary. Panel discussions often turn on whether the weaker criterion is fixable or disqualifying. Proposals that are strong on both tend to move forward. Proposals that are excellent on one and weak on the other are the ones that generate the most debate.

The weighting is equal in principle, but in practice the panel's discussion and the program officer's judgment shape outcomes. Program officers consider the panel's ratings, the program's priorities, and the portfolio they're building. A proposal that's strong on both criteria but unexciting on either can still lose to one that's excellent on one and solid on the other, depending on the program's needs.

Reviewers are also asked to consider both criteria in the context of the proposal's aims and the field's norms. A theoretical physics proposal and a community-engaged public health proposal will look different on both criteria, and reviewers are supposed to judge each on its own terms. That's why proposals that explain their field's context and norms do better than ones that assume the reviewer shares them.

The Most Common Intellectual Merit Weaknesses Reviewers Flag

The recurring intellectual merit weakness is a proposal that describes activity rather than a question. Reviewers want to know what hypothesis is being tested, what would count as evidence for or against it, and what changes in the field if it's answered. A proposal that says it will "study" or "investigate" something without framing a testable question reads as unfocused. The fix is usually a sentence or two early in the project description stating the question and the expected contribution.

A second weakness is a plan that doesn't match the question. Methods that can't answer the stated question, or a timeline that's unrealistic for the methods, get flagged. Reviewers with expertise in the methods will notice if the sample size is too small for the claim, or if the analysis plan is underspecified. A proposal that says it'll use "standard statistical methods" without naming them leaves the reviewer to guess.

Preliminary data is powerful but not required. A proposal with preliminary results that supports the feasibility of the approach scores better than one without, but a proposal without preliminary data can still score highly if the plan is strong and the team is credible. Reviewers penalize proposals that overclaim based on thin preliminary data more than they penalize the absence of preliminary data.

The fourth weakness is a team that doesn't match the work. A proposal that adds a co-investigator for credibility without demonstrating their role in the plan gets flagged. Reviewers look for evidence that each person has a defined, necessary contribution. A proposal with a clear division of labor and demonstrated relevant expertise scores better than one with a long list of senior personnel and no explanation of what they'll do.

Broader Impacts Failure Cases and Edge Scenarios

The classic broader impacts failure is the proposal that promises a big outreach program with no budget, no partner, and no evaluation plan. Reviewers read that as a promise the team can't keep. A proposal that budgets for the activity, names the partner, and describes how it'll be evaluated reads as real. The budget and the evaluation plan are the evidence that the plan is credible, not the ambition of the description.

Edge cases matter here. A proposal that plans to work with a community organization needs to show the partnership is real, not just named. Reviewers have seen enough proposals that claim a partnership that doesn't exist to look for evidence: a letter of collaboration, a description of prior work with the partner, or a clear explanation of the partner's role. A proposal that names a partner without evidence reads as boilerplate.

Another edge case is a proposal whose broader impacts are inseparable from the research. A proposal that trains graduate students as part of the research plan is doing broader impacts through the work, but it still needs to explain how that training is designed and assessed. The PAPPG's list includes integration of research and education, and reviewers look for a plan, not just an assumption that training happens.

A third edge case is a proposal whose societal benefit is in the future and speculative. Reviewers distinguish between a plan that will produce a benefit and a hope that the work might. A proposal that says the results could someday inform policy is weaker than one that says it will produce a specific product β€” a dataset, a tool, a curriculum β€” for a specific audience. The specificity is what makes the benefit credible.

One Original Calculation: How Reviewer Time Shapes the Scores

Reviews are time-constrained, and that constraint shapes how the two criteria get read. A typical panel reviewer reads a proposal in a limited window and writes a review that has to cover both criteria. The practical consequence is that proposals which make the reviewer's job easy β€” clear question, clear plan, clear broader impacts β€” get more accurate, more favorable readings than proposals that require the reviewer to reconstruct the case.

You can see this in the structure of a review form. The reviewer has to answer specific questions about intellectual merit and broader impacts, and each answer needs evidence from the proposal. A proposal that places the evidence where the reviewer expects it β€” a clear statement of the question, a clear plan, a clear broader impacts section β€” gets the benefit of the reviewer's time. A proposal that buries the case in dense prose gets a thinner review, which usually means a lower score.

The practical implication is that formatting and signposting aren't cosmetic. They're part of how the criteria get evaluated. A proposal that uses clear headings, states its question and its broader impacts early, and makes the evidence easy to find is doing real work, not just polish. Reviewers respond to clarity because clarity lets them do the job they're asked to do.

That's also why the two criteria get equal billing in the review form even when one is more familiar to the proposer. The reviewer has to fill in both, and a proposal that's strong on one but vague on the other forces the reviewer to write a mixed review. Making both easy to evaluate is the proposal's job.

How the Two Criteria Have Shifted and What Remains Stable

The fundamental structure β€” two criteria, intellectual merit and broader impacts β€” has been stable for years. The broader impacts criterion was introduced in the late 1990s and has been refined since, but the two-criterion framework has held. The definitions in the PAPPG have been edited, and the review questions have been updated, but the core architecture hasn't changed.

What has shifted is the emphasis and the review guidance. NSF has periodically updated how it asks reviewers to consider broader impacts, and the examples in the PAPPG have been refined. The direction of travel has been toward more specificity and more evidence, which matches what reviewers were already doing. Proposals that comply with the current guidance are judged more favorably than ones relying on older habits.

The stable core is the expectation that both criteria get a credible, specific treatment. A proposal that treats broader impacts as a box to check and intellectual merit as the real work is out of step with how the criteria are defined and how reviewers are trained. The proposals that do well are the ones that treat both as substantive.

For a proposal being prepared now, the practical takeaway is to check the current PAPPG for the deadline in question, because the exact language and review questions matter. The framework is stable enough to plan around, but the details are worth verifying against the current document rather than a memory of an older one.

FAQ

Does a proposal need to be strong on both criteria to get funded?

Yes, in practice. NSF reviews rate each criterion separately, and a proposal that's excellent on one and poor on the other is a hard case for a panel. The strongest proposals are strong on both, and proposals that are excellent on one and weak on the other generate the most debate and often lose to proposals that are solid on both.

Can broader impacts be part of the research plan itself?

Yes. The PAPPG's examples include integration of research and education, and a proposal that trains students as part of the research is doing broader impacts through the work. The key is that the activity still needs a plan and an assessment, not just an assumption that training happens as a byproduct.

What's the biggest misconception about intellectual merit?

That it's about the prestige of the investigator or the novelty of the field. Reviewers are asked to judge the idea, the plan, the team's fit for the work, and access to resources. A new investigator with a rigorous plan and a clear question can score highly. An accomplished team proposing work outside its demonstrated competence can score poorly.

What happens if broader impacts are vague but the science is excellent?

The proposal usually gets a mixed review and is at risk in panel. Reviewers are asked to justify a rating on broader impacts separately, and a vague promise doesn't give them evidence to justify a strong rating. A specific, modest, credible plan scores better than an ambitious, vague one.

How much does preliminary data matter for intellectual merit?

It helps but isn't required. Preliminary data that supports feasibility scores better than none, but a strong plan and a credible team can carry a proposal without it. Reviewers penalize overclaiming from thin preliminary data more than they penalize the absence of preliminary data.

Do the definitions of the two criteria change between PAPPG editions?

The core framework has been stable, but the exact phrasing and review questions have been edited over time. The version that governs your submission is the one in effect for your deadline, so check the current PAPPG on NSF's site rather than relying on an older copy.

Can a proposal fail on intellectual merit because of the team?

Yes, if the team doesn't match the work. Reviewers look for evidence that each person has a defined, necessary contribution and relevant expertise. A long list of senior personnel with no explanation of roles gets flagged. A clear division of labor with demonstrated expertise scores better.