Mapping the structure of science through usage

Science has traditionally been mapped on the basis of authorship and citation data. Due to publication and citation delays such data represents the structure of science as it existed in the past. We propose to map science by proxy of journal relationships derived from usage data to determine research trends as they presently occur. This mapping is performed by applying a principal components analysis superimposed with a k-means cluster analysis on networks of journal relationships derived from a large set of article usage data collected for the Los Alamos National Laboratory research community. Results indicate that meaningful maps of the interests of a local scientific community can be derived from usage data. Subject groupings in the mappings corresponds to Thomson ISI subject categories. A comparison to maps resulting from the analysis of 2003 Thomson ISI Journal Citation Report data reveals interesting differences between the features of local usage and global citation data.