The patterns that appear in exchanges between researchers, scientific journal publications and the demand for scientific articles often intersect, but the logic behind each type of activity is not necessarily the same. Analyses of requests for scientific articles from document suppliers may help to interpret current developments in electronic publishing. This study of article requests to the Institut de l'information scientifique et technique (INIST) shows that, in France, document supply customers fall into three main categories : business, academic libraries and public research organisations, in descending order. Demand focuses mainly on medicine, pharmacology, biology and chemistry, and the distribution of articles is entirely in accordance with the laws of bibliometrics. A further comparative analysis shows a high reciprocal correlation (except in the physical sciences) between the 50 journals most frequently requested from INIST, and the 50 most frequently cited journals according to ISI (Institute for Scientific Information). The titles which did not appear in either one list or the other show that the most frequently cited physics journals are not necessarily requested from the document supplier, and that, conversely, some frequently requested journals are not often cited. It may therefore be assumed that a trade in electronic articles is likely to develop quite rapidly in disciplines which are common to both lists, although this would focus on reputed titles only, but that a different pattern of electronic document exchange would emerge for scientific literature in other disciplines.