Earlier this month Johan Bollen and colleagues from the Los Alamos National Laboratories unveiled a much-publicized ‘Connections Map’ that shows how researchers navigate online between science journals and those of other academic disciplines (1). With access to as many as 1 billion ‘user interactions’ from 35,000 journals in the natural sciences, social sciences and humanities, the study was unique in its sheer scale (1,2). Furthermore, unlike other such studies that have mined inter-article citations data (papers that cite each other) to map connections, the work carried out by Bollen and colleagues relied on up-to-date internet usage and navigation information supplied by reputable online publishers such as Elsevier and Thomson Scientific (2).
The work of Bollen’s group appeared to be in every sense revolutionary. In their paper to PLOS One, for example, they drew attention to the rather biased nature of studies that use inter-article citations data, concluding that, “existing citations databases over-represent the natural sciences” (2). Other factors, such as the lengthy time that it takes for papers to get published, lent support to the claim that internet navigation provided a more temporally-accurate picture of traffic between journals. In contrast to citations data that focuses only on published authors, internet navigation information also reflects the activity of ‘a larger community’ that includes practicing scientists who do not necessarily publish (2).
In order to maximize the accuracy of their study Bollen and colleagues selected only those user interactions that involved requests for article abstracts or fully-published articles (2). The overall distribution of the interactions that they mapped ranged from 47 and 41 percent in the social and natural sciences respectively to 8 percent in the humanities (2). Bollen and colleagues were able to access individual ‘click streams’- that is, temporal sequences that show how researchers navigated between journals. The resulting Connections Map classified journals into ‘course-grained disciplines’ such as cognitive science, architectural design, international studies, religion, music, geology and plant genetics to name but a few (2).
While the timing of interactions in such maps are accurate to the second, some still question whether internet navigation-based connections really provide valuable information on the future trends of cross-discipline navigation. Anthony van Raan, director of the Leiden Centre for Science and Technology Studies suggested that Bollen’s approach may do nothing more than supply snapshots of current navigation fashions (1). Nevertheless data on such fashions can in itself be valuable for tracking “contemporary trends in scientific activity” and monitoring how such trends vary over time (1).
Bollen and colleagues admit that there is much work that can still be done (2). Future projects might include comparing connections maps with inter-article citations data, deconvoluting the different navigation patterns through which researchers move between journals and identifying the most influential journals in given areas of research (2). Indeed, if used correctly there is no denying that data from connections mapping could help improve the way online journals are made available to the research community.
1. Declan, B. (2009) Web Usage Data Outline Map Of Knowledge. Nature News (accessed 3/23/3009).
2. Bollen, J. et al. (2009) Clickstream Data Yields High-Resolution Maps of Science. PLoS ONE 4, e4803.