The Downward Spiral of the NSF’s GRFP

In April 2025, we were stunned by the news that the National Science Foundation’s Graduate Research Fellowship Program was offering only 1,000 awards, rather than the 2,300 targeted by the solicitation. Disappointment reverberated through the STEM research community. GRFP awards, which provide three years of financial support for graduate study, have gone to talented and high-achieving U.S. students early in their scientific training ever since NSF first began making awards in 1952.

This announcement was a throwback to 2008, the last year in which there were so few awards. But this time around, NSF’s public announcement also included an unusually high number of honorable mentions—3,035, in fact. What was going on?

We, two former GRFP program officers, downloaded the public data from NSF’s website to take a closer look. As we describe in a preprint of our findings, we quickly noticed that, out of the top pool of students recognized by NSF as awardees or honorable mentions, the awards were oddly distributed by their fields of study.

Out of the top pool of 4,035 applications within an estimated 15,000 total applications, the largest number had selected the life sciences as their field of study. But applicants in the life sciences were offered only 214 awards, whereas the second-most selected field of study, engineering, accounted for significantly more awards (283). Historically, the numbers of awards that NSF offers to early-career students in each field of study have been, for the most part, determined by the application pressure in each field—that is, a bottom-up response to the STEM community that lets students vote with their feet, as opposed to a top-down directive.

Two months after this disappointing and biased result, NSF abruptly announced another 500 awards (on June 13, well after the Council of Graduate Schools’ target date for students to accept their offers of support from graduate programs). Our analysis of this new 500 that were promoted from honorable mention to award status revealed an even more disturbing pattern: The bias had been amplified. This second tranche of awards contained not a single additional award in the life sciences; rather, the new awards were mainly distributed to computer science, materials research, physics and engineering. We shared the data with Jeff Mervis for his report in Science.

Note that back in 2020, for the first time in its history, GRFP’s solicitation for fiscal year 2021 listed a small set of “priority areas”: artificial intelligence, quantum science and computationally intensive research. At that time, this announcement caused an outcry from the STEM community. At least two major petitions to NSF were signed by thousands, including many former GRFP applicants, fellows, academics and scientists. The idea of priority areas was widely seen as “contrary to the program’s goals and harmful” (as Jason Williams of Cold Spring Harbor Laboratory and coauthors reported in one of the two petitions), further exacerbating elitist and exclusionary trends reported by Jane Hu in Science in 2019. Very few students early in their training, outside of R-1 universities, have research opportunities available to them in such fields. After this negative reaction from the scientific community, the GRFP competition appeared to back away from an emphasis on priority areas for a few years.

But not in 2025. As we pulled additional threads to analyze the results by subfields in order to understand how the award decisions were made, we discovered that fully 82 percent of applicants in the top pool who had listed a subfield of quantum information science and 85 percent of those who listed artificial intelligence or machine learning, a commonly used synonym in computer science, had been promoted to award status in the second tranche of 500. But listing a priority subfield did not help within the top pool of life sciences applicants.

This top-down finger on the scale came directly from the White House. At the July meeting of the National Science Board, current NSB member and former NSF assistant director Joan Ferrini-Mundy confirmed that the focus of 2025’s additional 500 awards was on “critical and emerging technologies.” At that meeting, Brian Stone, acting director of NSF, stated that these awards had been made with the “direct intervention” of the White House Office of Science and Technology Policy’s Michael Kratsios and Lynne Parker.

The losers in the 2025 competition were applicants in three long-standing NSF-supported fields of study: life sciences, psychology and geosciences. Full disclosure: We, the authors, earned our Ph.D.s in these fields (in cognitive psychology and biology, respectively). And it didn’t escape us that, according to the most recent data in the 2024 NCSES Survey of Graduate Students and Postdoctorates in Science and Engineering, 60 percent of Ph.D. students in the life sciences are women, as are 77 percent of Ph.D. students in psychology (with even higher percentages of women among master’s students).

The overall success rate for the GRFP competition was only 10 percent in 2025 (1,500 awards out of about 15,000 applicants). However, applicants in computer science enjoyed a success rate of 20 percent, more than four times the success rate in the life sciences (where fewer than 5 percent of applicants were offered awards) and more than three times that in psychology (6.4 percent).

It has been NSF’s mission since 1952 for GRFP to fund the person, not the project. GRFP fellowships are awarded to students on the basis of their high potential, not for their research topic—in fact, many do not even know where they will be attending graduate school at the time they apply. GRFP fellows have more independence and freedom to choose their own research topics, compared to other graduate students who are funded through teaching assistantships or research grants to faculty. GRFP predicts the future; for example, at least 50 Nobelists were offered GRFP awards three or four decades before they received their Nobel Prizes.

The basic, curiosity-driven science fostered by GRFP is not limited to a narrow set of fields. There is no evidence whatsoever that individuals with the highest potential for innovation, achievement and leadership in STEM fields cluster early in their careers in research fields favored by federal priorities.

NSF’s mission has been to support basic research. GRFP’s mission has been to encourage early-career students to enter science and engineering fields. The move toward advantaging AI and quantum is incompatible with this mission, especially since NSF already emphasizes support for AI and quantum research in its grant portfolio.

To foster innovation in artificial intelligence, and to ensure that its impacts are fairly distributed and ethical, it’s essential to include a broad range of perspectives, including from the life- and human-centered sciences. Excluding these STEM fields from GRFP fellowship support runs counter to those arguments.

GRFP continues to slide downhill. The solicitation for the 2026 competition was published two months late, with eligibility criteria that were changed abruptly to exclude students in their second year of graduate school. Thousands of early-career graduate students were unexpectedly deemed ineligible (after having been assured from the previous 10 years of GRFP policy that they could apply in their second year). NSF was largely unresponsive to the community during the time leading up to the competition deadline, and GRFP staff held none of the usual live informational webinars for applicants (posting only a recorded video shortly before the deadlines).

Then, in February, an unusually large number of GRFP applications were returned without review (RWR) with no justification beyond “ineligible research,” despite being eligible according to the solicitation. As of this writing, all 65 RWR applicants who have reached out to Grant-Witness.us so far happen to be in the life sciences and psychology.

There are significant risks to using field of study to exclude meritorious applicants. Creativity in science is not siloed. After all, Nobel laureates include scientists trained in life sciences, social sciences and psychology who make transformative discoveries in fields such as economics and medicine and physiology. In the currently chaotic landscape for U.S. federal grant funding, decisions about how to recruit talented early-career individuals into science and engineering will have lasting consequences for the U.S. workforce and the nation’s prosperity. NSF’s GRFP should return to its original mission of supporting the most promising students across all STEM fields.