Abstract: CASPUR allows many academic Italian institutions located in the CentreSouth of Italy to access more than 7 million of articles through a digital library platform. We analyzed the behaviour of its users by considering their "traces" stored into the web server log file. Using several Web Mining and Data Mining techniques we discovered that there is a gradual and dynamic change in the way how articles are accessed; in particular there is evidence of a Journal browsing increase in comparison to the searching mode. We interpreted such phenomenon by considering that browsing better meets the need of users when they want to keep abreast about the latest advances in their scientific field, in comparison to a more generic searching inside the digital library.