Electricité de France has conducted a probabilistic safety study on one of its 1300-MW(electric) reactors since the beginning of 1986. Practically all the reliability data used in this study are derived from the experience feedback from EdF’s pressurized water reactor units. The data concerning common-cause failures (CCFs) are therefore calculated on the basis of the information in EdF’s national files or on-site investigations. For most of the components (in particular, pumps and valves), the SRDF (system for gathering reliability data) and the event file of EdF’s Nuclear and Fossil Generation Division were used and >1200 data sheets were analyzed. Approximately 100 data sheets concerning CCFs were identified and CCF rates were deduced using the binomial failure rate method. In addition, a number of exhaustive investigations were carried out at the site chosen for the probabilistic study and samples of CCFs were analyzed in qualitative terms. The results were taken into account in the studies, and corrective actions were implemented on the site. Several observations were made on the basis of this work. The CCFs seem inevitable as they are of a complex and variable nature: problems of design or operation, meteorological conditions, human error, etc. As a result, it is very difficult to find a simple remedy and the number of solutions is often equal to the number of cases observed. Experience feedback is of fundamental importance to the rapid identification and correction of CCFs, especially for utilities that, like EdF, operate many standardized units. Finally, it appears that quantification of the problem is possible to a certain extent and that the order of magnitude of the rates of order 2 (ß2) CCFs is relatively well known, as is that of most components of order 0.1. The determination of the β rates at higher orders (β3, β4, etc.) still seems to be full of uncertainties, however.