Researchers could garner more citations simply by making their papers longer, a study seems to imply.

In an analysis of 30,027 peer-reviewed papers published between 2000 and 2004 in top astronomy journals, astronomer Krzysztof Stanek of Ohio State University in Columbus found that the median number of citations increases with the length of the paper — from just 6 for papers of 2–3 pages to about 50 for 50-page papers1.

There is, however, a limit to the benefits of size: citations start to tail off when papers reach lengths of 80 pages or so, perhaps because fewer people have the stamina to read them.

It is unexpected, says astronomer Jörg Dietrich of the European Southern Observatory headquarters in Germany, who recently conducted a similar analysis and found the same results but didn't publish them. "I expected that shorter papers would be cited more than longer ones," he says. "I assumed that people don't have the time to read long papers."

Papers of about 4 pages — the length of Letters in Astrophysical Journal and Astronomy and Astrophysics, which report brief summaries of work that is usually published in more detail later — fare better than papers 5–10 pages long. But brevity offers no such benefit for papers in the other two journals considered, Astronomical Journal and Monthly Notices of the Royal Astronomical Society, which do not have Letters.

There is definitely too much obsession with citations and other indices. ,

Stanek says he can't explain the length effect. Dietrich thinks that longer papers are more useful because on average they contain more information. But he suspects that such papers are often not read in their entirety. A study of the propagation of citation errors has revealed that citations are often simply copied without being read2.

Stanek does not intend to submit his study to a journal (although he encourages readers to "feel free to cite it as often as possible"). But it follows other recent investigations of citation statistics in astronomy3, 4, which looked at the effects of placing papers on the preprint server On average this doubles the number of citations received3, and Dietrich found that the timing of posting, and thus the paper's position in the mailing list, also had an important effect4.

This apparent potential to rig citations simply by adjusting a paper's size, place and timing of exposure sounds worrying at a time when citation statistics are increasingly being used, both formally and informally, as indicators of performance. Stanek points out that the 'h index'5, a measure of the cumulative impact of a researcher's published output, has become a fetish among some scientists. "You will become obsessed with it," he advises young scientists in his paper, and will start "telling other astronomers that your h is bigger than their h ".

"There is definitely too much obsession with citations and other indices," Stanek says. This is partly because they are easy to calculate regardless of whether they actually mean very much. But he confesses to using them himself.

Many of the greatest past discoveries, such as the structure of DNA, were reported extremely concisely. ,

Stanek treats his results semi-humorously, using them with earlier studies to draw up a set of guidelines for how graduate students might manage their publications and citations to greatest advantage as their careers progress. "Make sure you submit your papers to [Arxiv] just after 4 p.m. US Eastern time on Wednesday," he suggests, for example.

But his study highlights some important questions.One is whether — in the face of new dissemination channels such as preprint servers and an increased sensitivity to citation indices — it is realistic to regard citations as an accurate measure of achievement.

Another question is how long a paper ought to be. If length really does matter, will that encourage researchers simply to inflate their results unnecessarily? Many of the greatest past discoveries, such as the structure of DNA, were reported extremely concisely. "Most astronomical publications in the nineteenth century were very short observation reports, a few paragraphs at most," says Dietrich. But Stanek has no problem with length, saying that he prefers papers to "be as self-contained as possible — and I have seen a lot of short papers that are not".

Yet Dietrich thinks the risk of encouraging people to inflate their papers with waffle is small. "Writing a bloated paper is considerably more work, and incurs the risk of diluting an interesting work to the point that readers don't find the interesting aspects. Also, referees and editors are usually very critical of bloat. The best tactic is still to write a good paper, and a bloated paper is not a good paper."