Suggestions for writing an NSF proposal

Suggestions for writing NSF proposals
Paul Ronney
1. Have at least one really good, novel, clever idea. “Business as usual” proposals don’t
sell. Even if the idea isn’t that well formulated, or is very speculative, our panel gave a lot of credit
to proposals that showed something new and innovative. We tried to find a reason to like proposals
with a good idea, no matter what the other weaknesses might have been.
2. Pose specific, testable hypotheses. That is, don’t say “we will measure the effect of x on
y. Blah, blah, blah.” Instead say “based on the material presented in the introduction, it is our
hypothesis that y will decrease as x increases, until x reaches a critical value, after which y will
increase.” This shows that you have a specific goal in mind, i.e. that you have an idea and intend to
test it.
3. Avoid the “kitchen sink” mentality. Don’t say, “we will do very extensive measurements of
everything using every imaginable diagnostic tool and vary every possible experimental
parameter.” I really disliked seeing the word “extensive,” “thorough” or “comprehensive” in
proposals. It shows that you have no idea what’s important, so you’ll just measuring everything and
leave it to someone else to understand the problem. Instead say, “here is the minimum set of
measurements and conditions needed to test these hypotheses. If we have time and money left over
after this is completed, here is what else we will try to do.”
4. Explain your end game. That is, what you will do once you have the data? A lot of proposals
go on and on about what data they will collect, but then don’t say what they will do once they have
the data and how it can be used to test a hypothesis.
5. Don’t spend much page space on your past work. NSF proposals are limited to 15 pages,
and a lot of proposals didn’t get to the point of what they propose to do in the new effort until page
11. Then you need one page on “broader impacts” at the end, so in some cases there were only 4
pages of real “beef” in the proposal.
6. Don’t say “just trust me.” This doesn’t apply to young investigators, but a lot of senior people
wrote what we called “trust me” proposals. That is, they say, “I’m great, look at how much great
stuff I’ve done in the past, now I’ll do more of it. I have a forest of lasers and an army of
supercomputers. I don’t need to be specific since I’m so great. Just trust me to do great work
again. And because I’m so great, I deserve twice as much money as a new, untested PI.” In our
evaluations, we gave every benefit of the doubt to younger PIs because we know they are struggling
and if they are not funded and encouraged, they will drop out of academia altogether.
7. Don’t forget the “broader impacts.” NSF must consider the “broader impacts” which is
something beyond the technical merit of the proposal. For sure, 100%, you MUST have support
for graduate students as one of your budget items. Proposals that funded mostly summer faculty
salaries, postdocs, etc. went nowhere. Also it’s a good idea to have undergraduate support listed,
even if just a few K. Beyond student support, broader impacts can include things like how you will
incorporate women/minorities (leverage this with any available programs, such as the USC Viterbi
School’s Merit Research Program), where your graduates will be employed and why they will be
valuable to industry, how your research results might be used by a constituency other than the
combustion community, etc.
And above all else, as yourself, as though you were the reviewer rather than the author of the
proposal, “where’s the beef?”
So a good proposal should break down as follows, roughly:
# pages
2
3
1
1
4
2
2
Topic
Introduction - what your problem is and why it is important
Previous work - what has been done in this area and what is lacking in the knowledge
Objectives - very specifically, what will you do that is better?
Hypotheses - what you think will happen
Approach - how you will test these hypotheses - experimental apparatus, procedures,
etc.
Closure - what you will do with the data once you have it
Broader impact – applications of the research and educational merit