There is generally a positive correlation between antibiotic consumption and incidence of resistance to antibiotics used either for prophylaxis or therapy in human infections. This was not the case for two surgical wards in our hospital. A 15-year study showed that the incidence of methicillin-resistant Staphylococcus aureus (MRSA) was unrelated to cloxacillin consumption, and in fact fell after introduction into the two wards of an antibiotic policy based on cloxacillin. The two wards, a 90-bed orthopaedic unit and a 60-bed trauma unit, had an incidence of MRSA that has remained below the hospital average (23% in 1989, 32% in 1992). Before introduction of the policy the incidence of MRSA in 1977–1979 in the orthopaedic ward was 31%, and in the trauma ward 33%. In 1989 an investigation revealed no MRSA carriers in staff of either ward. In contrast, seven MRSA carriers were found among staff and patients of three other surgical units selected, because the percentage of MRSA isolated was above the average in our hospital. However a different type of patient is found in these units, the treatment techniques differ and broader-spectrum antibiotics are used. In addition to the usual precautions regarding nursing care and isolation techniques, the best means of reducing MRSA epidemics is to reduce the reservoir of carriers. The fall in the MRSA infection rate in orthopaedic and traumatology wards can be explained by antibiotic policy but also by other infection control measures.