Publishing for Impact

Choices made in the process of doing research can lead to a 5-10x difference in number of publications for exactly the same underlying work. For example, do I publish my idea in a workshop, then a conference, and then a journal? On the other hand I can just do the conference version. Do I take the trouble to write up a paper about an MS thesis where the student has somehow managed to escape without giving me a paper? (The common tactic for accomplishing this is to get a great job offer with a short fuse, and then utterly disappear.) Do I incrementalize my work into a collection of neat little papers each based on a single idea, or do I glom it all together into monster papers that contain a half-dozen or more research contributions?

Today’s question is:

If my goal is simply to maximize my own research impact over my career, should I aim for the low end (1-3 papers per year) or the high end (ten times more) of this spectrum?

I’ve puzzled through this and have concluded that as long as we keep the goal simple — maximize impact — there are few arguments for publishing more.

The first problem with publishing more is that writing a paper requires significant effort over and above that of doing the research in the first place. I usually estimate the cost of writing a paper at one month, but it could be considerably more or less depending on a number of factors. Since teaching uses up about three months of my effort each year, grant proposals cost 1-3 months, and service and administration another three months, the loss of a month of research time is a serious matter. In a nutshell: since my time is finite, writing more papers means I get less research done. Over a 40-year career, the loss of productivity due to writing 5-10 extra papers per year would be huge.

The other big problem with publishing more is that my target audience consists of extremely busy people. Having impact means that people who matter need to read and understand the material that I produce. If I produce 20 papers per year, how many will be read carefully? Probably not that many. My own reading time is highly constrained, and therefore I have become strongly prejudiced against reading papers by authors who seem like paper machines, because all too often any individual paper is weak and/or it contains a lot of similar thinking to something I’ve already read by that author. On the other hand, there are other researchers who produce less material where I’ll drop everything to read their latest piece of work because it reliably teaches me something new. In a nutshell: producing watered-down content creates a disincentive for people to read my work.

The thing that surprised me, when I sat down and thought through these issues, was how one-sided the arguments against prolific publication seem to be. Of course I’m not unaware of the arguments in the other direction. First, a larger number of papers impresses people who primarily count. Second, publishing more benefits students by giving them practice writing and by building up their vitae. Third, prolific publishing in a wide variety of venues probably exposes a larger number of people to one’s work — though I’m not sure this factor is very important given the social, searchable web as it exists today.


6 responses to “Publishing for Impact”

  1. > First, a larger number of papers impresses people who primarily count.

    Who are you talking about? I doubt your research peers can’t recognize when you’re publishing the same result three times or publishing different interesting papers. Maybe your administration is impressed, but I thought that they used “impact factor” and all that these days, so the payoff of writing to a small workshop (except when you’re actually interested in the workshop and have an interesting but not-quite-done thing to tell) is probably not so important.

    > Third, prolific publishing in a wide variety of venues probably exposes a larger number of people to one’s work.

    If you’re looking for quantity, I’m not sure publishing a lot is the most efficient strategy. I’d say that you should try to publish vulgarization articles in non-research journals and magazine. This can actually be very interesting both for you and the readers. On a personal note, having read your blog for a few weeks, I think your research area would translate relatively well into a vulgarization piece (I’m thinking Scientific American, this range of publications); wouldn’t be too easy (if the audience has no idea of what a compiler is), but doable (you’re not doing category theory), practical but with abstraction possibilities… I don’t know if you have already done it, but you should certainly try.

  2. Daniel, definitely. Also there’s a nontrivial amount of satisfaction in producing a really good paper — which takes an awful lot of time.

  3. Gasche, I’m talking about administrators: chairs, deans, college-level committees, etc. The chairs and deans make decisions about raises and also in many cases have effective veto power over hiring and promotion decisions.

  4. Gasche, I have written one non-academic article, a number of years ago:

    My take-away message from the experience was that I’d be better off spending the time on blog entries, which is what I now do. Of course, something like Scientific American is at a different level — I have no idea if they’d be interested in the compiler bugs. But this is interesting, thanks for the suggestion.