A web of scholarship

Citation statistics are widely used to assess the impact of scholarly work. The simplest statistic is the number of times that a particular article, a particular author, or a particular journal has been cited. Thomson Scientific's impact factor¹ is probably the oldest and best known, but there are a variety of alternatives. One feature all citations statistics share is that they are derived from the list of citations found in the literature cited section of published articles. Clearly cited works have had some impact on the authors of the paper, but some of those works are likely to have been more influential than others, so other ways of assessing that influence are needed.

Johan Bollen and his colleagues present such an assessment in PLoS One. Instead of compiling statistics from works that are cited, they compile statistics from works that are used. Specifically, they use log data from scholarly web portals to determine which journals are being used and to visualize the connections among them. The first thing that caught my eye was the table to the left. The usage data on which Bollen and his colleagues based their analysis shows a much closer match to the number of degrees awarded in a large, diverse university system than the number of citations in Thomson's Journal Citation Reports. Those data in and of themselves should give pause to any administrators who think that citation statistics alone provide a good measure of scholary impact and influence.

What's really cool, though, is the image to the left (click on the image for a full-size version). It one way of visualizing the connections among the 2,307 journals with at least 170 connections to other journals. Again, the pattern is very striking. The authors caution that the exact geometric positions of the various clusters depends on the visualization algorithm used, but even so, there appear to be some patterns likely to be preserved in any visualization. First, there are three relatively large clusters (physical sciences, life sciences, social sciences and humanities) and a couple of smaller ones (nursing, clinical pharmacology). Second, the humanities and social sciences are at least as connected to other clusters as any of the others. If anything,² it's the physical sciences that are the most isolated from other areas of scholarly work. They are mostly connected indirectly through connections in the life sciences. That's not too surprising, since quarks and molecules are a long way from law and archaeology, but it's interesting to see the patterns displayed visually. Third, and most important it illustrates for those who might need convincing that the social sciences and humanities are large, active fields of scholarly work, fields that are intimately connected to many others.

Here's hoping that as leaders at the University of Connecticut seek ways to bridge the large gap in its budget, they will remember that the social sciences and humanities are as central its future as the life sciences and the physical sciences.³