This paper both presents the preliminaries of citation analysis as designed by Eugene Garfield of the Institute for scientific information and an application of this method to the evaluation of chemical journals. It points out that a restricted number of journals, or " core ", are cited in seventy-five per cent of the citations. The increase in the percentage of the items cited and in the subsequent cost of subscriptions seems to follow a geometrical ratio. This method emphasizes the necessity for homogeneity in collection selectivity and shows that thresholds or levels can be defined. It could become an evaluation tool valid for collection development policies and cooperation in documentation acquisitions.