Nosocomial infection is a major problem in intensive therapy units (ITUs) and a significant cause of mortality. Selective decontamination of the digestive tract (SDD) has been advocated as a means to reduce ITU morbidity and mortality. Ninety-one patients in a general ITU underwent SDD, consisting of topical polymyxin E, tobramycin and amphotericin B administered throughout the unit stay together with parenteral ceftazidime for the first three days, and were compared with 84 historical and 92 contemporaneous control patients who were treated conventionally. Twenty-seven historical and 32 contemporaneous control patients developed unit-acquired infections, in comparison with only three patients in the SDD group (P less than 0.01). Mortality in the SDD group (36%) was not significantly different from that in the other two groups (historical control 40%, contemporaneous control 43%). Screening specimens revealed a significantly higher rate of colonization with resistant Acinetobacter spp. in the contemporaneous control than in the other two groups of patients; infection caused by resistant bacteria did not occur. SDD did not lead to a significant reduction in the use of systemically-administered antibiotics when compared with either control group. SDD may be used selectively in an ITU without ill effects on those patients not receiving SDD; nevertheless, microbiological monitoring is needed to detect emergence of resistant bacteria in the unit.