Pages

Friday, September 5, 2008

Optimal Paper Length

Before I came to CU and joined the research group of Dr. Juri Toomre I asked a lot of questions about the group - what's the work load like? what's the average time to thesis? Do grad students get to travel? What's the funding situation? One question I didn't even think to ask at the time was "How long are the group's papers on average?" The answer: really long. For example, the senior grad student in our group, Ben Brown, just published his first first-author paper which came in at a whopping 19 pages. Basically, in our group, each grad student publishes only 2-3 first author papers but each one is ~20 pages in length.

This has spurred a discussion of whether it is better to publish more shorter papers or fewer larger papers. Here are the pros and cons of both approaches:
Long Papers
Pros
  • Longer papers tend to have more impact and get more citations
  • Fewer longer papers leads to less hassle with referees and editors
Cons
  • Longer papers require more time before the first paper is published
  • Longer papers suffer from the "deal with everything in one shot" problem where people think that you should address the entire universe in your paper (since you already have so much in there, what's a little more?).
Short Papers
Pros
  • More short papers increase your name recognition and give the feeling that your group is a highly productive research group
  • Longer lists of papers on CV always look good
Cons
  • Many short papers can give the impression that you are trying to cheat the system, especially if all of your papers cite all of your other papers
  • Short papers are often less thorough in addressing all potential problems with their conclusions, this leading to "paper wars" when misunderstandings arise
This debate has been interesting, but it was often based on purely anecdotal evidence. However today on arXiv.org, a paper by Krzysztof Zbigniew Stanek (a name with an incredible number of consonants if you ask me) provided some data to the discussion. You can find his paper here.

Stanek took all of the journal articles published in the 4 leading astronomy/astrophysics journals from 2000 through 2004 and found all of their lengths and citations. He has various plots in his paper, but the one that I found most useful is shown below.
Since this is about astrophysical papers, it's only fitting to use a log-log plot with the number of citations per page on the y-axis and the number of pages on the x-axis. If we ignore the bump that comes in at around 3 to 4 pages (which corresponds to Astrophysical Journal Letters - the equivalent of Physical Review Letters), then there is a range between 6 and 40 pages where the data show that, on average, it does not matter if you publish more shorter papers or fewer longer papers in terms of citations. And if citations are a measure of impact in your field (as most job searches assume they are) then going short or long doesn't have any impact. Of course if you can get a ApJ or PhysRev Letter go for it, but otherwise the citations are a wash.

So the important point is to publish - long, short, or medium length doesn't really matter.

2 comments:

  1. So from what I gather from the study is that at the beginning of your career it does not matter how long the papers are (unless they are ridiculously short) but what matters most is the number of papers published. So I think it would be best to publish two 10 page papers rather than one 20 page paper (unless you have something really good that will be cited a lot through out your career).

    But at some point during your career you need to publish one or more papers that will be cited fairly heavily (the sooner the better) so that you can establish some recognition. So at some point in your career, the number of papers published is no longer relevant, but how much of an impact they have. There is one professor here that has published several papers (+70), and is still publishing, but the problem is he is not publishing anything especially citeworthy (that is more than normal). So he is not known for anything in particular, just for publishing papers that no one bothers to read.

    So while the number of published papers may be useful when you are fist starting, at some point it becomes more important to publish very influential papers rather than a lot of them.

    ReplyDelete
  2. This is a good question but I don't have a definite answer.

    I think the length depends on what the paper is on. If the idea can be explained in a short paper the paper should be short. If it takes a lot of pages. then make it long.

    Personally I don't think "how long" a paper should be has crossed my mind. I was just going to write it up and how long it is is how long it is.

    But, maybe it should be something I am concerned with. I mean, if I am writing a PRL there are constraints and if I find my paper is becoming a dissertation I probably need to break it up, but assuming I have one important idea or result, I'll type it up and whatever it comes out to be I'll go with it.

    As to citations, I think the highest cited papers will be those written at the best length to properly go over the idea.

    ReplyDelete

To add a link to text:
<a href="URL">Text</a>