Peer Review and the Most Influential Publications

Thanks to Josh Greenberg, I’ve been mulling over this fascinating paper I missed from last winter about the relative impact of science articles published in three different ways in the Proceedings of the National Academy of Sciences (PNAS). It speaks to the question of how important traditional peer review is, and how we might introduce other modes of scholarly communication and review.

PNAS now allows for three very different modes of article submission:

The majority of papers published in PNAS are submitted directly to the journal and follow the standard peer review process. The editorial board appoints an editor for each Direct submission, who then solicits reviewers. During the review process the authors are blinded to the identities of both the editor and the referees. PNAS refers to this publication method as “Track II”. In addition to the direct submission track, members of the National Academy of Sciences (NAS) are allowed to “Communicate” up to two papers per year for other authors. Here, authors send their paper to the NAS member, who then procures reviews from at least two other researchers and submits the paper and reviews to the PNAS editorial board for approval. As with Direct submissions, authors of Communicated papers are at least in theory blinded to the identity of their reviewers, but not to the identity of the editor. PNAS refers to this publication method as “Track I”. Lastly, NAS members are allowed to “Contribute” as many of their own papers per year as they wish. Here, NAS members choose their own referees, collect at least two reviews, and submit their paper along with the reviews to the PNAS editorial board. Peer review is no longer blind, as the authoring NAS member selects his or her own reviewers. PNAS refers to this publication method as “Track III”… Examining papers published in PNAS provides an opportunity to evaluate how these differences in the submission and peer review process within the same journal affect the impact of the papers finally published. The possibility that impact varies systematically across track has received a great deal of recent attention, particularly in light of the decision by PNAS to discontinue Track I. The citation analysis we now present provides a quantitative treatment of the quality of papers published through each track, a discussion which as hitherto been largely anecdotal in nature.

Here’s the eye-opening conclusion:

The analysis presented here clearly demonstrates variation in impact among papers published using different review processes at PNAS. We find that overall, papers authored by NAS member and Contributed to PNAS are cited significantly less than papers which are Direct submissions. Strikingly, however, we find that the 10% most cited Contributed papers receive significantly more citations than the 10% most cited Direct submissions. Thus the Contributed track seems to yield less influential papers on average, but is more likely produce truly exceptional papers. [emphasis mine]

I suspect this will hold true for many new kinds of scholarly communication that are liberated from traditional peer review. Due to their more open and freewheeling nature, these genres, like blogging, will undoubtedly contain much dreck, and thus be negatively stereotyped by many in the professoriate, who (as I have noted in this space) are inordinately conservative when in comes to scholarly communication. But in that sea of nontraditionally reviewed material will be many of the most creative and influential publications. I’m willing to bet this pattern will be even more pronounced in the humanities, where traditional peer review is particularly adept at homogenizing scholarly work.

Just a thought for Open Access Week.


Citations does not equal influence. From what I’ve read from science blogs on NAS publications, the Contributed track is much more likely to include both mediocre and stupendously wrong material, the latter of which often gains considerable attention in the media and literature by way of refutation. If attention is your sole metric, then fine, but to use citations as a proxy for ‘truly exceptional’ is careless, at best.

Dan Cohen says:

@Jonathan: Do you have some links to those criticisms? In the follow-up to this article, including in commentary by the article’s authors, I don’t see that issue discussed.

Godefroy says:

I have not read the original paper so this comment may be wrong, but I wonder about another possible explanation for this finding. Maybe the most cited papers are cited because people who wrote them are usually very cited authors — be it because they get refuted (as Jonathan Dresner pointed out) or because there are numerous people that are ready to cite anything they contribute as they hold positions of power in Academia. So contribution would be an easy way for “academic power playrs” to get published, and cited… Of course, such a game is reserved to the “top power players”, hence the poor mean results in citations.

Leave a Reply