During a two month period doctors' motives for prescribing antimicrobial chemotherapy were studied. Of the 1775 patients admitted, 374 (21%) received antimicrobial agents for 384 infective episodes. In 44 (11.5%) no infection could be subsequently proven. Of the remaining 340, drug selection was based on a susceptibility report in 120 (35%), on the advice of a medical microbiologist with the aid of a Gram stain in 66 (19%), and without this assistance in 51 (15%). In 103 (30%) episodes the clinician selected the agent without advice and laboratory tests. Infection was proven in 237 of 238 episodes in which therapy was started following a known susceptibility result or the advice of the microbiologist. Four (1.7%) of these later required a change of therapy, compared with 22 (21%) of the 103 courses started without such help (P less than 0.0005). This study has shown that the use of the laboratory and consultation with a medical microbiologist improves the rational prescribing of antimicrobial chemotherapy.