Bibliometrics involves the measurement of various aspects of bibliographic publications--the average number of publications scientists in a given discipline produce, the average number of citations an average publication will receive, the impact that a journal has due to the frequency of citations to its papers, and so on. There are well-accepted models that explain variations in productivity, and this paper suggests that measures of Internet use may improve the predictive power of such models.

By analyzing various internal Unix accounting logs, the authors were able to identify usage of various processes (including mail, news, finger, and ftp) by natural resources faculty at the University of California, Berkeley. They used a variety of statistical analyses to determine correlation and predictive improvement of these processes with publication productivity. They found that usage of common Internet processes was equal to about one-half the factor of age since the user received a Ph.D--a key predictor in most models. In other words, Internet users produced more publications than non-Internet users, and their productivity increased as their usage increased. Although this result is useful in strengthening bibliometric models, the authors raise a good question about how this information can be obtained if Internet use becomes more client-based so that general data collection from servers is no longer feasible. They also point out that they do not have data for other disciplines, so they cannot generalize their findings.