Merging the NSF and NIH
Two speeds
In the fall of 2014, Ebola was loose in West Africa and, for a few frightening weeks, in Dallas. WHO had first reported cases in Guinea on March 23. By October, the virus had reached an American hospital, and the National Science Foundation moved.
On October 16, the NSF Director issued a Dear Colleague Letter inviting proposals on how Ebola spreads and how people behave in an epidemic. The mechanism was RAPID: internal review only, up to $200,000, one year. NSF staff at the time put the turnaround at one to two weeks from proposal to money.
What could an NIH-funded investigator do in the same weeks? The answer is: not much, quickly. NIH had no standing agency-wide mechanism for urgent extramural awards. Its fastest disaster mechanism, run out of NIMH, took about eight weeks from submission to award. NIH didn’t adopt an agency-wide urgent award policy until June 2018, after Ebola and Zika, and even that version is open only to existing grantees, needs an ad hoc review panel, and still goes to Council.
I want to be fair here. NIAID was fast in its own way. Its intramural vaccine trial started at the NIH Clinical Center the first week of September 2014, built on candidates that had been in human testing since 2003. That’s real speed. It’s also the speed of an in-house lab with a decade’s head start, and it tells you nothing about how fast the government can put money in the hands of a university scientist with a new idea.
Two agencies. Two clocks. Same virus.
The seam in ordinary times
Ebola shows what the split between NSF and NIH costs in a crisis. The BRAIN Initiative shows what it costs on an ordinary Tuesday.
I’ve written about BRAIN before (”The Coordination Problem“), so I’ll be brief. When I led NSF’s part, NIH, NSF, DARPA, IARPA, and later others funded the brain, each with its own review culture, vocabulary, and sense of what counted as progress. NIH had most of the money. NSF had most of the physicists, engineers, and theorists the field needed. Getting the two to point in the same direction took a remarkable amount of senior time, many meetings, and a fair number of compromises that made diplomatic sense and less scientific sense.
So here’s the question. Why do we fund the study of the brain’s molecules in one agency and the mathematics of its networks in another? Nobody designed it that way. It just happened.
The foundation Bush proposed
In July 1945, Vannevar Bush sent Science, the Endless Frontier to President Truman. It proposed a single National Research Foundation, with one division for medical research. Bush wanted physicists, chemists, and physicians under one roof, governed by scientists, insulated from day-to-day politics, and funded for the long haul.
Then came five years of fighting. Senator Kilgore had his own plan, with more political control and more emphasis on applied work. Congress finally passed a Bush-style bill in 1947, and Truman vetoed it. His objection was governance: the bill put the foundation’s director under a part-time board of private citizens instead of under the President. Truman wanted a science agency that answered to him. Bush wanted one that didn’t.
While Washington argued, biomedicine found a home. In 1946 the medical research contracts left over from the wartime Office of Scientific Research and Development went to the Public Health Service, and the NIH extramural grants program grew up around them. By the time the National Science Foundation Act finally passed in 1950, NIH had four years’ head start and a growing constituency in Congress. Mary Lasker and her allies were about to show the country what disease-named institutes could do for an appropriation.
That’s the history in one sentence: the split between NSF and NIH came out of a delay. Nobody ever sat down and decided that basic biology and biomedicine belonged in different agencies with different review systems and different cultures. We’ve lived with an accident for 76 years and gotten used to calling it architecture.
The consolidation we’re getting
I know how this argument sounds in October 2026.
The administration proposed collapsing NIH’s 27 institutes and centers into eight. Congress said no in the FY2026 funding bill. The restructuring has gone ahead anyway, administratively, including changes to how NIH runs peer review, made by directive rather than statute. On the NSF side, the FY2027 request eliminates the Social, Behavioral and Economic Sciences directorate outright. Consolidation is already happening. It’s just happening by fiat, and in the direction Truman wanted: more presidential control, less insulation.
So let me be explicit about where I stand. I’m not arguing for that. What’s underway at NIH is the opposite of Bush’s idea, wearing some of its clothes. Bush’s foundation was built to be hard for any president to bend. The current project is built to make science easy for this current administration to bend.
It’s worth remembering that Congress protected both agencies this year. NIH came through FY2026 at about $48.7 billion, roughly flat. NSF ended up near $8.75 billion, up from a requested $3.9 billion. The current structure survived its hardest year yet. Why would anyone want to rebuild it now?
Because surviving isn’t the same as being well designed, and because the next reorganization is coming whether the science community writes it or someone else does.
Bush’s terms
Here’s what I’m proposing. One National Research Foundation, combining NSF with NIH’s extramural research programs (but not the NIH intramural program in Bethesda), built on Bush’s terms.
Governance first. A board of distinguished scientists and citizens with staggered, fixed terms. A director with a fixed term, removable only for cause. Readers of “The Board That Couldn’t Be Fired“ will know why I care about that clause. An agency the President can reshape by memo is a science foundation that will be reshaped by memo.
Statute second. Here’s a fact about NSF that deserves more attention than it gets. Apart from global change research, nearly everything that made NSF NSF (research computing, the major facilities, its role as the place other agencies came to for the basic science under their missions) rested on custom. Custom works fine until someone decides to move it. A merger would require a new organic act, and a new organic act is a chance that comes along about once every 75 years to write the franchise into law. I’d take it.
Culture third. The story of Ebola in 2014 should give you a big clue as to which agency culture I’d prefer for the merged successor. The ability to invest quickly, with program officers having a high degree of discretion, informed by merit review when possible, strikes me as very important.
Names last. Keep the disease-named institutes as divisions inside the foundation. The National Cancer Institute should still exist, with that name, and the people who march for it every spring should still have something to march for. Bush’s Medical Research division can contain multitudes.
What do you get for all this? One director defending one top line for American research. One review system with one set of standards. No seam for BRAIN to fall into, and no second clock when the next virus shows up.
The best objection
The strongest argument against what I’m proposing comes from NSF’s own appropriations history, and I’d rather meet it head-on.
Until 2005, the VA-HUD appropriations subcommittee funded NSF. That year both chambers abolished VA-HUD and moved NSF and NASA into the new Commerce-Justice-Science subcommittee. The community was pleased. A science bill! In hindsight, the move was never priced. Relationships reset: Kit Bond, NSF’s most reliable Senate champion, stayed behind. The competition changed from veterans’ medical care to the FBI, NASA, the Census, and federal prisons. And NSF lost the best position an agency can have: a small, sympathetic line in a big bill. In CJS it became a big account in a medium bill.
Apply that to my foundation. A combined agency of something like $58 billion would be the biggest discretionary item in whatever bill it landed in. By the small-in-a-big-bill rule, that’s about the worst place to be.
My answer is NIH. NIH is already the biggest item in Labor-HHS, and it does fine. It doubled between 1998 and 2003, on schedule, the only completed doubling in modern American science funding. Big in a big bill hasn’t hurt it at all. Why? Disease. Every member of Congress has constituents with cancer, Alzheimer’s, or diabetes, and every one of those diseases has an institute with its name on the door. That rule describes agencies without constituencies. That’s NSF’s problem, and it’s exactly the problem a merger is meant to solve.
NSF has never lacked for friendly authorizers. Its authorization doubled on paper in 2002, again in the COMPETES Acts, again in CHIPS and Science, and the appropriators never followed through. Basic science outside medicine has never found an appropriations constituency. The only proven machine for building one belongs to NIH. A merger bets that the machine can carry physics, ecology, and mathematics along with it.
It’s a bet, and it’s somewhat out there (easy for a nearly retired professor). The coattails could run the other way, with disease advocates turning the whole foundation into a translational shop. That’s why the culture and the governance clause matter more than the org chart.
What stands in the way
Any proposal this big owes the reader a bill. Here’s mine.
The culture could go the wrong way. NIH’s extramural budget is roughly five times NSF’s. Put them together without a deliberate design, and the bigger culture wins. The study section absorbs the program officer, and the next Ebola response gets slower. Every agency merger I’ve watched has followed the money. This one would too unless the statute says otherwise.
Congress would have to willingly redraw its own committee map. NIH’s appropriation currently lives in Labor-HHS, NSF in CJS, with different authorizers on each side. Wherever the successor foundation ends up, it changes who it competes with for allocations, and it resets relationships on both sides at once. The 2005 move shows Congress can rehouse a science agency when it wants to. It doesn’t show Congress will.
One foundation is one target. In an upcoming commissioned piece, I’ll argue that the American research system’s resilience comes partly from having many funders, so that a hit to one doesn’t take down the whole. A single foundation is the biggest hub you could build. The governance protections above are there because of this risk, and I won’t pretend they eliminate it. A future administration with a grudge against science would have one director to fire instead of two.
“Single” is an overstatement. DOE’s Office of Science, DOD, NASA, USDA, and ARPA-H stay where they are. So do NIH’s intramural program and the Clinical Center, in my version, because their mission-driven, in-house work belongs with HHS. What I’m proposing is a single foundation for investigator-initiated research. That’s still most of the federal basic research enterprise, and it’s where the return on investment would be the highest.
The argument Bush lost
Bush lost in 1947 to a President who wanted control over science. Seventy-nine years later, the same argument is being had again, in the opposite order. This time, consolidation comes first, and whoever moves fastest settles who controls it.
I’d rather the science community moved first. If American research is going to be reorganized, and it is, the reorganization should be written in statute, governed by a board that can’t be fired on a whim, and run on the culture that could fund Ebola research in two weeks. That’s Bush’s foundation. We should have built it in 1950.
We can still build it. Somebody is going to.
