NSF vs. NIH Grant Writing: The Structural Differences That Change Everything

A researcher who has built a successful NIH funding record and then applies for their first NSF CAREER award often makes a revealing mistake: they write a scientifically sound application that scores poorly. The science is legitimate, the investigator is accomplished, and the proposed work is intellectually coherent. What is missing is not the science but the writing — specifically, the recognition that NSF and NIH are evaluating research proposals against fundamentally different frameworks, and that an application written for one framework does not translate directly to the other.

This is not a minor adaptation problem. NSF's review criteria and NIH's review criteria reflect different theories about what federal research funding is for, which questions matter, and what "impact" means in the context of scientific investment. The writing habits that produce fundable NIH applications are, in several specific ways, the habits that produce unfundable NSF applications — and vice versa. Understanding why requires examining each framework on its own terms before trying to translate between them.

The Core Framework Difference: Five Criteria vs. Two

NIH reviewers score applications on five criteria: Significance, Investigator(s), Innovation, Approach, and Environment. The Overall Impact score synthesizes these into a holistic judgment, but the five-criteria structure shapes how applications are written and how they are evaluated. A strong NIH application typically leads with clinical or translational significance, establishes the investigator's track record through publications and preliminary data, and devotes most of its pages to the Approach — the specific experimental design that will test the stated hypotheses.

NSF reviews applications on exactly two criteria: Intellectual Merit and Broader Impacts. These criteria are weighted equally in the review process — a point that surprises many NIH-trained researchers who assume Intellectual Merit is the primary criterion and Broader Impacts is supplementary. NSF is explicit in its guidance: "Both criteria are to be given full consideration during the review and decision-making processes." An application with excellent Intellectual Merit and weak Broader Impacts is disadvantaged relative to one that performs well on both.

For NIH-trained researchers, the Broader Impacts criterion is the most common site of failure. They write Broader Impacts sections that are brief, generic, and clearly treated as an afterthought — because in NIH's framework, impact is addressed in the Significance section through the lens of health outcomes, not through a separate criterion that encompasses education, diversity, societal benefit, and infrastructure development. Writing a cursory Broader Impacts section on an NSF application signals to reviewers that the investigator does not understand what NSF is funding.

What Broader Impacts Actually Requires

NSF's definition of Broader Impacts encompasses five areas: advancing discovery and understanding while promoting teaching, training, and learning; broadening participation of underrepresented groups; enhancing infrastructure for research and education; broad dissemination; and benefits to society. An effective Broader Impacts section addresses multiple of these areas with specific, credible activities — not vague commitments to "disseminate findings widely" or "mentor graduate students."

Specific activities that score well in Broader Impacts sections are ones that are connected to the investigator's existing institutional infrastructure and track record. An investigator at an R1 institution who has previously mentored underrepresented undergraduates can credibly commit to specific mentorship activities with specific recruitment pipelines. One who has developed publicly available datasets or tools can credibly propose to expand or disseminate those resources. An investigator with no prior track record of the activities they propose in Broader Impacts faces a credibility problem that is structurally similar to an NIH investigator claiming they can run a clinical trial without any preliminary data demonstrating feasibility.

NSF program officers reviewing applications in competitive areas have described the most common Broader Impacts failure as the absence of specificity: statements about what the investigator will do without any evidence of how, where, with what infrastructure, or through what mechanism. "We will train graduate and undergraduate students" is not a Broader Impacts plan. "We will recruit two PhD students annually from our existing partnership with [HBCU partner], leveraging the REU supplement funding we have received in two of the past four years" is a Broader Impacts plan.

Intellectual Merit and the NSF Emphasis on Advancing Knowledge

NSF's Intellectual Merit criterion asks whether the proposed activity will advance knowledge and understanding within and across fields. This is similar to NIH's Significance criterion but carries a different emphasis: NSF is primarily interested in fundamental knowledge generation rather than translational or clinical application. An application proposing research with immediate clinical utility is not disadvantaged at NSF, but the criterion prioritizes the advancement of basic understanding over the production of usable outputs.

The practical writing implication is that NSF applications should lead with the fundamental question being addressed rather than with the problem it solves. An NIH Significance section leads with the clinical gap and the patient population affected. An NSF Intellectual Merit section leads with the conceptual gap and what its resolution will teach the field. The reversal of emphasis is small in practice but significant in scoring: reviewers at NSF are evaluating whether the proposed work will teach us something we genuinely do not know, not whether it will solve a problem we know exists.

The Investigator Criterion and Career Stage

NIH's Investigator criterion evaluates whether the investigator(s) are appropriately trained and suited to carry out the proposed work. For established investigators, this criterion is largely satisfied by publication record and prior funding. For early-stage investigators, NIH has explicit policies — Early Stage Investigator (ESI) protections, lower payline thresholds at some institutes — that are intended to counterbalance the track-record disadvantage. Knowing where you fall in this taxonomy matters for how you write the Investigator section.

NSF does not have a formal equivalent to the Investigator criterion, but career stage plays out differently through the CAREER award mechanism, which is specifically designed for untenured investigators within five years of starting their faculty position. The CAREER award integrates research and education plans in a way that has no NIH parallel: the education plan is a substantial component of the application, not a brief statement of mentorship intentions. Investigators who apply for CAREER awards without understanding that the education plan is evaluated with the same rigor as the research plan frequently produce underdeveloped education components that limit their scores.

Preliminary Data: Essential at NIH, Less So at NSF

NIH applications are expected to include substantial preliminary data demonstrating that the investigator has the technical capability to execute the proposed work and that the approach is feasible. Summary statements from NIH study sections routinely cite inadequate preliminary data as a primary reason for poor Approach scores. For many NIH application types, especially R01s, reviewers are implicitly asking: "Has this investigator already demonstrated that this will work?"

NSF's stance on preliminary data is more nuanced and varies by program and directorate. Many NSF programs value preliminary data as evidence of feasibility and investigator expertise, but they do not carry the same expectation that the work is already partially done. Some NSF programs explicitly discourage the kind of preliminary data sections that read as reports of already-completed research, on the grounds that it suggests the proposed work is more incremental than transformative. The intellectual risk that NSF values — the genuine uncertainty about whether a line of inquiry will succeed — is sometimes in tension with the comprehensive preliminary data package that NIH requires as evidence of feasibility.

This does not mean preliminary data is unimportant at NSF. It means that the framing of preliminary data differs. At NIH, preliminary data answers: "Why should reviewers believe this will work?" At NSF, preliminary data answers: "Why is this investigator uniquely positioned to pursue this question?" The emphasis shifts from feasibility demonstration to expertise establishment.

Citation Practices and Community Expectations

NIH research strategies typically have extensive bibliographies — sometimes 50 to 100 or more citations for a 12-page document — because the field norms around establishing significance and situating the proposed work in existing literature are dense. Reviewers in biomedical fields expect specific citation of the papers that establish the gap, demonstrate the limitations of current approaches, and support the investigator's preliminary data claims.

NSF programs, particularly in physical sciences, mathematics, and engineering, have different citation cultures. Applications in these fields tend to be more selective in citation, with bibliographies that are shorter and citations that are more highly filtered for direct relevance to the proposed work. An NIH-trained researcher who carries their citation density habits to an NSF application in a field with lighter citation norms produces a document that reads as unfamiliar with the community's conventions — a signal that can affect how reviewers perceive the investigator's fit with the program.

Page Limits and Density Norms

NIH R01 Research Strategy sections are limited to 12 pages. NSF standard proposals are limited to 15 pages for the Project Description. These limits produce different density norms: NIH applications in biomedical fields tend to be very dense, with narrow margins, small fonts (within allowable limits), and compressed prose. NSF applications in many programs benefit from somewhat less density — clearer visual organization, more white space, and more explicit structural signaling — because the reviewer audience includes scientists across a broader range of subspecialties than the targeted study section that reviews most NIH applications.

The NIH habit of maximum density in minimum space is not universally portable. NSF program officers have noted informally that applications that are clearly designed to cram as much information as possible into the page limit — using figure legends to add text that would otherwise exceed the limit, for example — can read as attempts to circumvent the constraints rather than as evidence of good judgment about what the reviewers need. NSF's page limits are more clearly enforced, and the culture around them is different.

Writing Habits That Are Funding-Agency-Specific

Several writing habits are specific to one agency and counterproductive at the other. Knowing which habits transfer and which must be abandoned is the practical core of switching between NIH and NSF writing.

Leading with clinical impact: Appropriate at NIH, less so at NSF basic science programs. An NSF application in fundamental chemistry that leads with potential pharmaceutical applications is prioritizing the wrong frame for that reviewer audience.

Extensive preliminary data sections: Expected at NIH, should be reframed at NSF as evidence of expertise rather than demonstration of completed feasibility work.

Detailed experimental protocols in the main text: Expected and valued in NIH Approach sections; in NSF applications, detailed protocols often belong in supplementary materials rather than the main Project Description, where space is better used for conceptual framing.

Brief education/training statements: Acceptable at NIH, where they appear in the Investigator section; fatal at NSF CAREER awards, where the education plan is a substantive scored component.

Generic Broader Impacts language: No equivalent at NIH; at NSF, it is one of two scored criteria and must be as specific and credible as the Intellectual Merit section.