I was going to explain this in the Strassmann blog discussion, but since even G-rated comments about people writing like they are building a medieval labyrinth seem to have a way of disappearing over there, I'll post it here. Ya' know, because we have to keep this discussion civil.
If you have been following the discussion over at the Strassmann blog regarding changes to NSF, you'll notice a few familiar themes. One of the most popular is comparison to the Canadian system and a call for smaller grants to more people. This debate was a particular hobby horse of the now tumbleweeded NSF is Broken forum. Suffice to say it's not a good solution without overhead reform, another re-occurring topic. Unfortunately, NSF's hands are tied on that front as well unless NIH wants to play along. So what to do?
IMO, the calls for a hybrid system make the most sense and we need not even leave the DC metro area to find a model for how to make it work. Regular readers will know that I have some limited experience applying to NIH. Whereas NIH has it's issues with success rates, one of the things they have right is the variety of mechanisms for funding. If we think of the traditional NSF grants as an R01-like mechanisms (the big individual PI grants), then I think NSF would be serving it's constituents well by instituting an R15 and R21-like mechanism of limited budget funding.
For those unfamiliar with NIH, the R15 awards are limited budget ($300K/3yrs in direct costs) awards aimed specifically at institutions that are not in the top XX% of NIH money getters. This mechanism would not only give NSF PIs a limited budget option, but also deal with the "the top schools take all the money and leave none for the little guy" concern, because the top schools would not be eligible. This has been a very popular program at NIH.
The R21s are also limited in their budget ($275K/
3yrs2yrs), but are for exploratory research. Another major (and valid) criticism of the NSF process is that the bar for "preliminary" data is so high that you need 50% of the work done already. The R21 mechanism would solve this issue and provide an alternative to NSF's seemingly arbitrary EAGAR program by providing a source of funding to get a line of research off the ground.
If you want to get crazy, we could throw in an R03-like mechanism, but I feel less strongly about this.
But obviously the money for these programs can't come out of thin air. Instituting the change would require cutting some money from the regular programs, so how would that work? I don't know the numbers, meaning I'm just tossing out ideas here, but I would support something like a 5% cut to all existing grants (yes, mine included) and a slight reduction of the pot for the next round. While people will howl about further reductions to the pot given the current funding rates, I think the potential success of these programs will make up for it. In addition to that, I think panels should be empowered to tell investigators that if they want to resubmit a certain "regular" project as it is, it should come back as another (smaller) grant type.
So far, though, we have not dealt with the review load problem. For that issue I would support both a shorter grant application (8-10 pages) and a deadline set up like MCB has gone to - two deadlines a year, but an 8 month cycle of review with a limit of two proposals per PI per year. I think the 4 page preproposal is relatively useless, particularly when there are different people reviewing the preproposals and the full proposals.
Perhaps this doesn't fix all the problems instantly, but I think it goes a long way towards a good compromise. There may be good reasons why NSF is uninterested in the smaller grant mechanisms, but the argument that one can always submit a smaller budget doesn't fly because the science is judged first. There needs to be a way to separate out these smaller proposals so they can be judged against one another. I could see budgets for each of the smaller mechanisms limited to something in the $100K-$150k range in direct costs over a three year proposal. We could even shorten the length of the proposal for these grants to 6 pages if you want.
There's my suggestion, feel free to make your own.