Data from retrospective studies of endophthalmitis vary widely with respect to incidence and to the pathogens implicated. To see whether we could provide more accurate data, we have done a prospective multicentre national survey of endophthalmitis over one year in France. Records of 36,241 operations and 1148 cases of ocular trauma from 64 centres specialising in eye surgery were analysed. There were 167 cases of endophthalmitis; incidence of postoperative endophthalmitis was 0.31 per 100 operations, and the risk after penetrating ocular trauma was 2.8%. In contrast with most previous studies, Staphylococcus epidermidis was the most frequently isolated organism, with gram-negative organisms accounting for only a small proportion of cases. Patients infected with streptococci had the least favourable outcome. The survey confirms data from retrospective studies showing that the incidence of postoperative or post-injury endophthalmitis is low. The low frequency means that large numbers of patients would be required for a trial of antibiotic efficacy, but such a trial is worthwhile because there are now antibiotics with good ocular bioavailability that are effective against most of the bacteria that cause endophthalmitis.