Gold-Standard Science Requires More Than Crisis Rhetoric

A response to NIH Director Dr. Jay Bhattacharya’s “Launching a Second Scientific Revolution”

Today I want to do something a little different.

About a year ago I made a video (We’re About to Lose a Generation of Scientists – And It’s Not an Accident) that stepped outside the usual lane I stay in. In that video I talked about how I’d spent decades training scientists, and I told you why I was worried about what was happening to funding at the National Science Foundation. That video came out of a conviction I’ve held for a long time: this blog and my YT channel exists to defend good science wherever it’s under threat, not only when young-earth organizations happen to be the ones doing the threatening.

I want to come back to that conviction today, because I think we’re at another moment where it matters.

This time it isn’t the NSF. It’s the National Institutes of Health, and specifically an essay written by its current director, Dr. Jay Bhattacharya, titled “Launching a Second Scientific Revolution.” It was adapted from a speech he gave this spring at Hillsdale College, and it has been circulating widely, including among people in my own circles who don’t normally follow science policy closely.

I should tell you upfront where I’m coming from. I am not a disinterested observer. I went to graduate school on NSF-funded research. I would not have the PhD I have, and I would not have had the career I’ve had, without the kind of public investment in basic science that this essay, in effect, calls into question. You should know that going in, and weigh what follows accordingly.

Reform is not the same thing as demolition

Dr. Jay Bhattacharya’s Imprimis essay, adapted from an April 2026 Hillsdale College speech, presents a deliberately ambitious claim: the National Institutes of Health should help launch a “second scientific revolution.” His argument begins with an image that will resonate with many scientists: science should not be governed by a priesthood. Scientific truth is not established by agency directors, journal editors, professional prestige, or institutional authority. It is tested by evidence, criticism, replication, and the capacity of alternative explanations to survive empirical scrutiny (Bhattacharya, 2026).

That premise is sound. It is also incomplete. The danger in the essay is not that Bhattacharya values reproducibility, innovation, or broader participation in biomedical research. Those are worthy goals. The danger is that he uses those goals to support an overdrawn narrative: that a replication crisis explains public confusion about science, that the NIH has failed because U.S. life expectancy has not improved enough, that existing peer review is structurally hostile to innovation, and that funding concentration can be repaired by severing the link between research grants and the infrastructure that makes research possible.

The full article makes the critique easier to sharpen. Bhattacharya explicitly identifies three problems: the replication crisis, scientific stagnation, and funding concentration. He then presents three corresponding remedies: fund replication and mark replicated work in PubMed; replace payline-based funding with a Unified Funding Strategy that can favor riskier ideas; and separate direct research funding from indirect facilities support so more institutions can compete for infrastructure resources (Bhattacharya, 2026).

A fair response should not treat all of this as wrong. NIH should fund rigorous replication. It should not confuse journal prestige with truth. It should support early-career scientists and fund riskier work. It should watch for self-reinforcing institutional concentration. But a reform agenda loses credibility when it turns real problems into a generalized indictment of biomedical research, especially at the same moment that researchers report grant terminations, payment freezes, staff losses, and funding decisions that appear to track political priorities rather than scientific merit.

What Bhattacharya gets right

1. Peer review is not a declaration of truth

Bhattacharya is correct that peer review is not an audit, a replication, or a certification of truth. Reviewers generally evaluate whether a manuscript or proposal is plausible, methodologically adequate, significant, and properly contextualized; they do not rerun experiments or reanalyze every raw dataset. This distinction matters. Published science is a provisional contribution to a continuing evidentiary process, not an oracle.

This point should be uncontroversial among practicing scientists. John Ioannidis’s influential 2005 paper argued that the probability that a published claim is true depends on study power, bias, the number of tested relationships, and the prior probability of the hypothesis being tested. The paper is often invoked carelessly, but its central message is not anti-science. It is a statistical argument for better design, adequate power, transparent analysis, and humility about isolated findings (Ioannidis, 2005).

2. Replication deserves more direct support

Bhattacharya’s proposal to fund replication is a genuinely constructive part of the essay. Replication studies often remain professionally undervalued because they are less novel, harder to publish, and less likely to generate conventional academic prestige. A funding agency with NIH’s scale can help correct that incentive problem.

The idea of making replication status more visible in PubMed is also worth serious discussion. A well-designed “replication” or “evidence context” function could help readers identify confirmatory studies, failed replications, systematic reviews, retractions, expressions of concern, and related methodological critiques. The difficulty is implementation: such a system would need careful curation, domain-specific standards, and protection against being reduced to a simplistic badge of “true” or “false.” Still, the underlying goal is appropriate.

3. NIH should fund high-risk science and younger investigators

Bhattacharya is also right that risk aversion can distort grantmaking. When paylines are low, reviewers and applicants often become conservative. Applicants are incentivized to present polished, already-feasible projects rather than genuinely uncertain ideas. Reviewers may penalize insufficient preliminary data even when the purpose of a public research portfolio should be to support work that private capital is unlikely to fund.

NIH has long recognized this problem. The NIH Common Fund’s High-Risk, High-Reward Research program supports exceptionally creative scientists pursuing innovative projects with potential for broad impact, and some of those programs do not require preliminary data (NIH Common Fund, n.d.). Bhattacharya’s call for risk is therefore not a novel discovery so much as a legitimate reminder that such mechanisms should be protected and perhaps expanded.

4. Funding concentration is a real problem

The article’s concern about concentration also has merit. If a small set of institutions receives a large share of NIH resources, the system may reinforce accumulated advantage: major universities have experienced grant offices, core facilities, reputational advantages, and established networks that make future awards easier to obtain. A healthy national biomedical enterprise should cultivate strong research capacity across a wider range of institutions and regions.

This concern is especially important for early-career investigators and for scientists at institutions that serve many students but lack elite research infrastructure. A reform agenda that expands access to instrumentation, core facilities, mentoring networks, and administrative grant support could make American biomedical science broader and more resilient.

Where the argument overreaches

1. Life expectancy is the wrong single scoreboard for NIH performance

Bhattacharya argues that NIH’s mission is to fund research that improves American health and longevity, then points to stagnant U.S. life expectancy since 2010 as evidence that NIH has not done its job well. That is rhetorically powerful, but analytically weak. Life expectancy is a population-level endpoint shaped by many variables outside NIH’s control: income inequality, access to care, insurance design, diet, physical activity, firearms, opioids, alcohol, traffic deaths, environmental exposures, public health capacity, and state policy choices.

The problem is not that longevity is irrelevant to NIH. It is that life expectancy cannot be treated as a clean response variable for NIH grant strategy. A drug discovery pipeline can improve cancer survival while overdose deaths, COVID-19, suicide, or cardiovascular risk factors depress national life expectancy. Conversely, changes in overdose mortality can raise life expectancy without proving that biomedical research suddenly became more rigorous.

Recent CDC mortality data illustrate the problem. CDC reported that U.S. life expectancy reached a record high in 2024 and that overdose death rates fell 26.2% from 2023 to 2024 (CDC/NCHS, 2026). That does not prove NIH strategy suddenly became optimal. It demonstrates that national life expectancy is too confounded to serve as a simple referendum on NIH’s scientific portfolio.

2. The egg example is not a clean example of a replication crisis

The weakest scientific example in the essay is the claim that changing advice about eggs illustrates the replication crisis. Bhattacharya writes that eggs were once treated as nearly poisonous, then only egg whites were treated as acceptable, and now eggs are sometimes described as a “superfood.” He concludes that “at the root of the confusion is the replication crisis” (Bhattacharya, 2026).

That framing is misleading. The egg story is not best understood as a simple case in which scientists failed to replicate an experiment. It is a case in which nutritional epidemiology, cardiovascular physiology, dietary guidelines, and public communication evolved as researchers learned more about a complex exposure embedded in complex diets.

Eggs are not a single-variable intervention. They contain dietary cholesterol, but they also contain protein, choline, carotenoids, vitamins, phospholipids, and other nutrients. Their health effects depend on the background diet, substitution effects, saturated fat intake, total caloric intake, diabetes status, LDL responsiveness, food preparation, and the foods they replace. A breakfast of eggs with vegetables is not biologically equivalent to a breakfast of eggs with bacon, buttered toast, and processed meat.

The contemporary literature remains nuanced rather than triumphalist. A pooled analysis of six U.S. cohorts in JAMA found that higher dietary cholesterol or egg consumption was associated with higher risk of incident cardiovascular disease and all-cause mortality, though the egg association was no longer significant after adjustment for dietary cholesterol (Zhong et al., 2019). By contrast, a BMJ analysis of three large U.S. cohorts and an updated meta-analysis found no overall association between moderate egg consumption and cardiovascular disease risk (Drouin-Chartier et al., 2020). The American Heart Association’s science advisory recommended considering dietary patterns rather than isolated cholesterol targets, and it allowed that healthy individuals can include up to one whole egg per day within a heart-healthy diet while advising caution in some groups (Carson et al., 2020).

That pattern is not a simple failure to replicate. It is the normal difficulty of causal inference in diet and chronic disease. Observational studies have measurement error, confounding, and substitution problems. Randomized trials can measure lipids but are often too short or too small to measure clinical endpoints. Dietary guidelines must translate heterogeneous evidence into practical advice for populations with different baseline risks. The lesson is not that “science keeps changing, therefore scientists cannot be trusted.” The lesson is that complex biological systems often require a move from simple rules to conditional, mechanistic, and population-specific recommendations.

This matters because Bhattacharya uses the egg example to justify a larger policy narrative. If common cases of scientific refinement are rhetorically reclassified as replication failure, then uncertainty itself becomes a weapon against expertise. That is bad science policy. Replication reform should improve inference, not encourage a public habit of treating every revised recommendation as evidence of corruption or incompetence.

3. Reproducibility reform should not be used to delegitimize whole research areas

A serious reproducibility agenda would be procedural: preregistration where appropriate, better statistical training, open methods, data sharing when ethical and legal, replication funding, robust metadata, image-integrity screening, independent validation, and stronger incentives for negative results. Bhattacharya gestures toward some of these reforms, especially replication funding and PubMed discoverability.

The concern is that his essay is situated within a broader funding environment in which scientists have reported abrupt grant cancellations, payment freezes, and political screening of research topics. The Bethesda Declaration, published by NIH staff and supporters, charged that NIH had halted high-quality, peer-reviewed grants and contracts through indiscriminate terminations, payment freezes, and blanket holds regardless of scientific quality, progress, or impact (Stand Up for Science, 2025). Nature reported in January 2026 that 5,843 NIH grants had been terminated or frozen during 2025 and that NIH issued 24% fewer grants than the previous ten-year average (Kozlov et al., 2026).

These reports do not prove that every cancelled grant was scientifically excellent or that no reform was warranted. They do show why scientists are skeptical when “gold standard science” language is paired with sweeping disruption. If replication is the stated rationale, then terminations should be tied to transparent evidentiary criteria: flawed methods, unverifiable data, misconduct, lack of progress, or changed scientific priorities that are articulated in advance. If terminations instead track politically disfavored topics, the result is not a second scientific revolution. It is viewpoint governance.

4. “Bleeding-edge” science is not the same as good science

Bhattacharya’s second major claim is that biomedical science has stagnated because NIH funds too much work based on older ideas and too little work based on new ideas. He argues that the Unified Funding Strategy will empower reviewers and program officers to support risky projects that may fail but could produce fundamental advances (Bhattacharya, 2026). The underlying concern is reasonable: a portfolio composed entirely of safe incremental work would be suboptimal.

But “newness” is not a sufficient criterion for public funding. Many important biomedical advances emerge from mature research programs that require sustained refinement: structural biology, model organism genetics, immunology, epidemiologic cohorts, longitudinal clinical studies, rare disease registries, comparative genomics, and method development. A good portfolio needs novelty, replication, refinement, infrastructure, long-term datasets, and basic discovery whose applications are not yet visible.

NIH’s own statements about the Unified Funding Strategy are more measured than the Imprimis essay. NIH says the framework, effective for the January 2026 Council round, is intended to support meritorious research, address health priorities, sustain the biomedical workforce, consider career stage, and promote broader distribution and geographic balance (NIH, 2025a). NIH institute leaders also emphasize that peer review remains foundational and that funding decisions will consider scientific merit, programmatic relevance, portfolio balance, and financial stewardship rather than relying solely on paylines (NIH, 2026).

That is defensible in principle. The risk is opacity. Paylines are imperfect, but they are legible. A system that moves beyond paylines can become better if it adds transparent portfolio criteria, public rationales, conflict-of-interest safeguards, and post hoc reporting. It can become worse if it gives political appointees more discretion to label preferred topics “innovative” and disfavored topics “ideological.” The difference depends not on slogans but on governance.

5. Basic science is undervalued when “health outcomes” are interpreted too narrowly

The most important scientific omission in Bhattacharya’s framework is the role of basic science. The essay praises NIH’s historical contributions to heart disease, cancer, rheumatoid arthritis, cystic fibrosis, and sickle cell anemia. Yet many such advances depended on decades of basic work that could not have been justified by immediate population-health deliverables at the time it was funded.

NIH itself has stated that the path from basic research to health benefits is often long and difficult to trace, but that fundamental knowledge frequently produces unanticipated breakthroughs. NIH’s basic research materials note that basic science accounts for roughly half of the NIH budget and cite an analysis finding that NIH contributed to published research associated with every one of 210 FDA-approved new drugs from 2010 to 2016, with 90% of that research being basic research on biological targets rather than the drugs themselves (NIH, 2022).

This is why “make Americans healthier” cannot be reduced to short-term translational metrics. A country that funds only immediately legible health outcomes will eventually deplete the conceptual and technical base from which future therapies arise. CRISPR, mRNA vaccines, monoclonal antibodies, cancer immunotherapy, MRI, cryo-EM, and molecular genetics all depended on basic science that initially looked remote from the clinic. A funding strategy that rewards only immediate clinical payoff will cannibalize the future.

6. Indirect costs are not merely institutional rent

Bhattacharya’s third proposed reform is to “sever the link” between direct research funding and indirect facilities support. He argues that the current system rewards institutions with expensive facilities and entrenches concentration. The concern about concentration is legitimate. The proposed remedy is underdeveloped.

Indirect costs are not a slush fund. They pay for the shared conditions that make research possible: animal care facilities, biosafety systems, radiation safety, hazardous waste disposal, core instrumentation, data security, compliance systems, grant administration, building operations, utilities, and the maintenance of laboratories that must meet regulatory and technical standards. Those costs are real whether the grant is held by a famous university or a regional institution.

The recent legal fight over NIH’s attempt to cap indirect costs at 15% shows why the issue is not merely administrative. The U.S. Court of Appeals for the First Circuit upheld a ruling that NIH could not impose an across-the-board cap on research overhead reimbursement, concluding that the agency violated statutory law and its own regulatory procedures. Reporting on the case noted that indirect costs support information technology, utilities, administrative support, and the operation of laboratories (Higher Ed Dive, 2026).

A better policy would ask how infrastructure can be broadened without destabilizing ongoing work. That could include regional core facilities, shared instrumentation grants, transparent facilities competitions, capacity-building awards for emerging research institutions, bridge support for compliance infrastructure, and mechanisms that allow smaller institutions to buy access to high-quality cores. But simply separating facilities from research support risks creating unfunded mandates: investigators receive awards, but institutions lack the infrastructure to execute them safely and reproducibly.

The current funding context changes how the essay should be read

If Bhattacharya’s article had appeared in isolation, it could be read as a conventional metascience reform essay. In the current NIH context, it reads as a policy rationale for a broader reorientation of federal biomedical science. That context matters.

Scientific organizations have warned that proposed cuts would damage the research enterprise. FASEB stated that the administration’s FY 2026 budget proposal included unprecedented cuts of roughly 40% to NIH and nearly 60% to NSF and would severely erode biomedical and biological research (FASEB, 2025a). FASEB later urged Congress to continue funding grants for researchers and recommended $51.3 billion for NIH and $16.7 billion for NSF (FASEB, 2025b).

Nature’s 2026 analysis described thousands of federal grant cancellations and suspensions, including 5,843 NIH grants, and reported that roughly 2,600 cancelled or frozen grants had not been reinstated or unfrozen, amounting to $1.4 billion in unspent funding. The same analysis reported that NIH issued 24% fewer grants in 2025 than the previous ten-year average (Kozlov et al., 2026). A PNAS analysis of NIH terminations reported that from February to August 2025 NIH terminated 2,291 active grants, rescinding $2.45 billion from a $5.08 billion investment (Oliveira et al., 2026).

A policy environment like that makes rhetoric about “gold standard science” especially consequential. Scientists are generally willing to be evaluated. They are not generally willing to have years of peer-reviewed work interrupted by opaque administrative decisions, sudden payment freezes, or politically coded judgments about what research is acceptable. Reform requires rules. Disruption without transparent rules is not reform.

A better reform agenda

Bhattacharya is right that NIH can improve. But the right reform agenda should be more disciplined than the one implied by the essay.

First, NIH should fund replication without turning replication into a punitive weapon. Dedicated replication grants, registered reports, adversarial collaborations, and confirmatory studies should be normalized. Replication should be targeted strategically: influential findings, clinical claims with high public impact, studies with small samples or fragile statistical support, and domains where policy depends on uncertain evidence.

Second, NIH should improve evidence context in PubMed, but not with a simplistic truth badge. A useful evidence-context tool should distinguish exact replication, conceptual replication, systematic review, meta-analysis, correction, retraction, expression of concern, and independent validation. It should also make uncertainty more interpretable without pretending that every paper can be categorized as simply replicated or not replicated.

Third, NIH should preserve transparent peer review while adding explicit portfolio logic. Moving beyond paylines can be beneficial, but only if the criteria are public, consistently applied, and auditable. “Innovation,” “health priority,” “programmatic relevance,” and “financial stewardship” must not become containers for political preference.

Fourth, NIH should expand high-risk and early-career funding through mechanisms designed for that purpose. If early-career investigators are disadvantaged by preliminary-data expectations and institutional prestige, NIH should expand New Innovator-style awards, starter R01 equivalents, bridge funding, and review criteria calibrated to career stage.

Fifth, NIH should broaden institutional participation by building infrastructure rather than pretending infrastructure is optional. Regional cores, shared instrumentation, capacity-building grants, and transparent facilities support would address concentration more directly than starving indirect costs.

Sixth, NIH should defend basic science explicitly. The history Bhattacharya praises is largely a history of basic discoveries becoming clinically meaningful later. If current leaders want future cures, they must protect the curiosity-driven and mechanism-driven work that makes those cures possible.

Conclusion: humility cuts both ways

Bhattacharya repeatedly appeals to humility about what science can claim to know. Scientists should accept that admonition. Peer review is fallible. Published studies can be wrong. Prestige can distort judgment. Replication deserves more funding. NIH should support creative younger scientists and a wider set of institutions.

But humility about what science can claim to know also applies to reformers. It should make NIH leaders cautious about overinterpreting life expectancy trends, cautious about turning the egg-and-cholesterol story into a caricature of scientific failure, cautious about treating “new ideas” as inherently superior to mature research programs, cautious about weakening the infrastructure that makes research reproducible, and cautious about using crisis language to justify opaque funding disruption.

The central problem with the Imprimis essay is that it identifies several real weaknesses in science but repeatedly draws conclusions stronger than the evidence warrants. The result is a reform rhetoric that could support good policy if disciplined by transparency, pluralism, and respect for basic science, but could also rationalize political selection of research under the label of “gold standard science.”

A second scientific revolution is not launched by declaring that science has too much authority and then concentrating more discretionary authority in agency leadership. It is launched, if at all, by strengthening the practices that make scientific communities self-correcting: open evidence, independent replication, methodological rigor, durable infrastructure, transparent review, and long-term support for the basic science whose value often becomes visible only after the fact.

Video version of this blog post:

References

Bhattacharya, J. (May 2026). Launching a second scientific revolution. Imprimis. Adapted from an April 28, 2026 Hillsdale College National Leadership Seminar speech. https://imprimis.hillsdale.edu/launching-a-second-scientific-revolution/

Carson, J. A. S., Lichtenstein, A. H., Anderson, C. A. M., Appel, L. J., Kris-Etherton, P. M., Meyer, K. A., Petersen, K., Polonsky, T., & Van Horn, L. (2020). Dietary cholesterol and cardiovascular risk: A science advisory from the American Heart Association. Circulation, 141(3), e39-e53. https://doi.org/10.1161/CIR.0000000000000743

Centers for Disease Control and Prevention, National Center for Health Statistics. (2026, January 29). U.S. life expectancy hits record high as drug overdose deaths decline in 2024. https://www.cdc.gov/nchs/pressroom/releases/20260129.html

Drouin-Chartier, J.-P., Chen, S., Li, Y., Schwab, A. L., Stampfer, M. J., Sacks, F. M., Rosner, B., Willett, W. C., Hu, F. B., & Bhupathiraju, S. N. (2020). Egg consumption and risk of cardiovascular disease: Three large prospective US cohort studies, systematic review, and updated meta-analysis. BMJ, 368, m513. https://doi.org/10.1136/bmj.m513

FASEB. (2025a, May 6). FASEB statement in response to FY 2026 administration budget request. https://www.faseb.org/getmedia/008022d5-ab65-49b2-96e9-d17a3bdcdca3/FASEB-Statement-In-Response-to-FY-2026-Administration-Budget-Request-FINAL.pdf

FASEB. (2025b, July 10). FASEB implores Congress to reject proposed research funding cuts. https://www.faseb.org/journals-and-news/latest-news/faseb-implores-congress-to-reject-proposed-research-funding-cuts

Higher Ed Dive. (2026, January 6). NIH cap on indirect research costs struck down on appeal. https://www.highereddive.com/news/nih-indirect-cost-cap-appeal-struck-down/808925/

Ioannidis, J. P. A. (2005). Why most published research findings are false. PLOS Medicine, 2(8), e124. https://doi.org/10.1371/journal.pmed.0020124

Kozlov, M., Tollefson, J., & Garisto, D. (2026, January 20). US science after a year of Trump: What has been lost and what remains. Nature. https://www.nature.com/immersive/d41586-026-00088-9/index.html

National Institutes of Health. (2022). Basic research digital media kit. https://stagetestdomain3.nih.gov/news-events/basic-research-digital-media-kit

National Institutes of Health. (2025a, November 21). Implementing a unified NIH funding strategy to guide consistent and clearer award decisions. NIH Extramural Nexus. https://grants.nih.gov/news-events/nih-extramural-nexus-news/2025/11/implementing-a-unified-nih-funding-strategy-to-guide-consistent-and-clearer-award-decisions

National Institutes of Health. (2026, February 23). NIH institute and center director perspectives on implementing the NIH Unified Funding Strategy. NIH Extramural Nexus. https://grants.nih.gov/news-events/nih-extramural-nexus-news/2026/02/nih-institute-and-center-director-perspectives-on-implementing-the-nih-unified-funding-strategy

NIH Common Fund. (n.d.). High-Risk, High-Reward Research. https://commonfund.nih.gov/highrisk

Oliveira, D. F. M., et al. (2026). How the 2025 NIH grant terminations varied by researchers and projects. Proceedings of the National Academy of Sciences. https://www.pnas.org/doi/10.1073/pnas.2527755123

Stand Up for Science. (2025, June 9). Bethesda Declaration. https://www.standupforscience.net/bethesda-declaration

Zhong, V. W., Van Horn, L., Cornelis, M. C., Wilkins, J. T., Ning, H., Carnethon, M. R., Greenland, P., Mentz, R. J., Tucker, K. L., Zhao, L., Norwood, A. F., Lloyd-Jones, D. M., & Allen, N. B. (2019). Associations of dietary cholesterol or egg consumption with incident cardiovascular disease and mortality. JAMA, 321(11), 1081-1095. https://doi.org/10.1001/jama.2019.1572