BACKGROUND AND OBJECTIVES: The development of electronic editions of scientific journals and the rapid spread of scientific information might modify the pattern the bibliographic citations, and thus the impact factor and quality of journals. We assessed changes in the impact factor over years of a number of journals and whether the presence of an electronic version of the journal was associated with the impact factor score. DESIGN AND METHODS: This is a retrospective longitudinal study. The availability of journals (table of contents (TOC), abstracts, full text and free full text) on Internet, in years 1995-2000, was assessed between December 2000 and January 2001. The first 20 top-journals from 8 subject categories were included. Changes in impact factor over time and association with Internet availability were modeled. RESULTS: Overall, 118/139 journals (85%) had their TOC on the Internet, of these 107 (77%) had abstracts, 97 (70%) had full text and 33 (24%) free full text. The median impact factor for all journals was 1.65, 2.08, 2.10, 2.21 and 2.35 for the years from 1995 to 1999, respectively. This increase was statistically significant, with differences among subject categories. The presence of TOC, abstracts and full text on the Internet was also significantly associated with higher impact factor, after accounting for time and subject category. INTERPRETATION AND CONCLUSIONS: The impact factor has been used for assessing the quality of journals. We identified a new limitation of this indicator: the impact factor seems to be related to the amount of circulation of information through Internet. This could be a temporary limitation, associated with diffusion of journals on, and spread of Internet.