The use of selective decontamination of the digestive tract (SDD) as prophylaxis against nosocomial respiratory tract infection remains controversial, largely because of concerns that, in the long term, it may promote the emergence of antibiotic-resistant strains. This report describes the results of surveillance cultures and susceptibility testing undertaken during the course of a 2-year, double-blind study of the efficacy of SDD which was conducted in a respiratory intensive care unit (ICU). Surveillance specimens from the alimentary tract and trachea were obtained from each patient on admission and then twice weekly until 48 h after discharge from the unit. Specimens were cultured semiquantitatively and organisms from morphologically distinct colonies were identified by standard methods; the susceptibilities of these isolates were determined by the disc diffusion method. Five thousand, nine hundred and sixty surveillance samples from 239 patients were processed in this way. Compared with the placebo group, SDD caused a significant reduction in the incidence of colonization of the alimentary tract with aerobic Gram-negative bacilli (AGNB), and Candida spp. were almost totally eliminated. The incidence of colonization with enterococci increased in both groups, while the incidence of both colonization of the alimentary tract with strains of coagulase-negative staphylococci and methicillin-resistant Staphylococcus aureus and infection caused by these organisms increased in the SDD group. Acinetobacter spp. were the most common bacteria associated with unit-acquired colonization and lower respiratory tract infection in both groups. The acquisition of strains of Pseudomonas aeruginosa and cefotaxime- and/or tobramycin-resistant Enterobacteriaceae was significantly greater in the placebo group than in the SDD group, although tobramycin-resistant strains of Proteus, Morganella and Providencia spp. were isolated from three of 114 patients receiving SDD. The use of SDD did not lead to an overall increase in antibiotic resistance amongst the AGNB usually associated with ICU-acquired infection. However, colonization with strains which were either resistant to one or more of the antibiotic components of the regimen or which were not inhibited by the regimen was observed and may subsequently lead to infection.