Safe Bets and Stalled Breakthroughs: How NIH's Grant Machinery Discourages the Science It Claims to Champion

The National Institutes of Health is the largest public funder of biomedical research on the planet. In fiscal year 2023, it disbursed approximately $35 billion to universities, medical centers, and research institutions across the United States. Its stated mission—to seek fundamental knowledge and apply it to enhance health and reduce illness—carries an implicit promise of discovery. Yet a growing body of evidence, and the candid testimony of researchers who navigate the system daily, suggests that the machinery designed to distribute that investment is quietly working against the very breakthroughs it purports to fund.

The central tension is not one of malice or negligence. It is structural. And understanding it requires looking carefully at how scientific merit is evaluated, who does the evaluating, and what incentives govern both sides of that transaction.

The Peer-Review Bottleneck and Its Hidden Conservatism

When a researcher submits an R01 application—the NIH's flagship investigator-initiated grant mechanism—the proposal is assigned to a study section composed of established scientists working in adjacent fields. These reviewers assess the application across several criteria, including significance, innovation, approach, investigator qualifications, and environment. Each dimension receives a score, and the aggregate determines whether the proposal advances toward funding.

In principle, this system is elegantly meritocratic. In practice, it contains a structural bias that researchers have long recognized but seldom discuss publicly for fear of alienating the very colleagues who may one day review their work. Reviewers are experts in what is already known. They are well positioned to evaluate whether a proposed methodology is sound, whether the preliminary data are convincing, and whether the research question fits within an established framework. What they are considerably less equipped to do—by virtue of their expertise itself—is assess the value of questions that challenge that framework.

A proposal arguing that a widely accepted mechanistic pathway is fundamentally misunderstood faces an immediate credibility problem: the reviewers most qualified to evaluate it are precisely those whose careers were built on the prevailing model. This is not a conspiracy. It is an epistemological constraint baked into the architecture of expert review.

Preliminary Data as a Paradox

The NIH review process places substantial weight on preliminary data—evidence that the proposed research direction is feasible and that the investigator has the technical capacity to execute it. The logic is defensible: taxpayer dollars should not fund pure speculation. But the downstream effect is a quiet mandate for researchers to have already partially answered the question they are asking for money to answer.

For incremental science, this requirement is manageable. A laboratory studying a well-characterized protein can generate preliminary findings within its existing infrastructure and present reviewers with a trajectory that is legible and low-risk. For a researcher pursuing a genuinely novel hypothesis—one that may require new tools, new model systems, or conceptual frameworks that have not yet been validated—generating conventional preliminary data is either impossible or self-defeating. If the work can be done without the grant, the grant may not be necessary. If it cannot, the application will likely be scored poorly for insufficient feasibility.

This paradox effectively prices high-risk, high-reward science out of the standard funding market. Researchers who recognize this dynamic adapt accordingly, designing proposals that are ambitious in rhetoric but conservative in scope—a practice sometimes referred to, with weary familiarity, as "grantsmanship."

What the Success Rate Numbers Reveal

The NIH's own data on application success rates tell part of this story in aggregate. Across most standard mechanisms, success rates have hovered between 18 and 22 percent in recent years—meaning that roughly four out of every five competitive applications are rejected. Under conditions of such scarcity, risk aversion becomes rational behavior. A principal investigator whose laboratory depends on continuous grant funding cannot afford to spend two years developing and submitting a speculative proposal with a low probability of success when a more conventional application in an established area stands a meaningfully better chance of keeping the lights on.

The cumulative effect across thousands of investigators is a gravitational pull toward the center of established knowledge. Individual researchers are not making irrational choices. They are responding logically to a system whose incentive structure, whatever its stated values, rewards predictability.

The NIH's Own Acknowledgment—and Its Limits

The NIH has not been oblivious to these criticisms. The High-Risk, High-Reward Research program, which includes mechanisms such as the Pioneer Award, the New Innovator Award, and the Transformative Research Award, was explicitly designed to fund science that conventional review processes might reject. These programs use modified review criteria and, in some cases, smaller panels intended to be more receptive to unconventional thinking.

The results have been meaningful in individual cases. Pioneer Award recipients have produced research that diverged substantially from mainstream trajectories. But the program remains a narrow channel. The Transformative Research Award, for instance, funds only a handful of projects per cycle. As a proportion of the NIH's total extramural budget, high-risk mechanisms represent a rounding error. The implicit message—that transformative science is a valued exception rather than a systemic priority—is not lost on the research community.

Alternative Models and What They Suggest

Outside the NIH ecosystem, several funding models have emerged that deliberately invert the conventional logic. The Howard Hughes Medical Institute's Investigator program funds scientists rather than projects, providing long-term support on the premise that exceptional researchers should be liberated from the quarterly anxieties of the grant cycle. The arc of several major discoveries—including foundational work in gene editing and structural biology—traces back to HHMI's willingness to tolerate extended periods without publishable output.

Philanthropic initiatives such as the Open Philanthropy Project and the Wellcome Trust's Leap program have experimented with milestone-free funding, lottery-based selection among qualified applicants, and explicit prioritization of proposals that established reviewers find implausible. Early evidence from these experiments is cautiously encouraging, though the sample sizes remain too small for definitive conclusions.

Within the federal system, ARPA-H—the Advanced Research Projects Agency for Health, modeled loosely on DARPA—represents a more institutionalized attempt to fund ambitious, mission-driven science outside the traditional peer-review framework. Its program managers hold significant discretionary authority to pursue unconventional directions, and its evaluation criteria explicitly de-emphasize the preliminary data requirements that constrain R01 applicants. Whether ARPA-H can sustain that culture as it matures and faces congressional scrutiny remains an open question.

The Cost of Caution

The stakes of this conversation extend well beyond academic career dynamics. The diseases that remain most resistant to therapeutic intervention—Alzheimer's, many cancers, treatment-resistant psychiatric conditions—are precisely those where incremental progress along established mechanistic pathways has delivered the least. The scientific breakthroughs most likely to change outcomes for patients are, by definition, the ones that will surprise us. A funding system calibrated to minimize surprise is, in a meaningful sense, calibrated to minimize discovery.

None of this suggests that rigorous peer review should be abandoned, or that scientific skepticism is the enemy of innovation. It suggests, rather, that the mechanisms through which skepticism is institutionalized matter enormously—and that a system designed by experts, for experts, will inevitably reflect expert consensus in ways that can be difficult to distinguish from genuine scientific judgment.

The inquiring mind, confronting this evidence, is left with a discomforting conclusion: the United States may be spending tens of billions of dollars per year to produce the science it already expects, rather than the science it most urgently needs.