The epidemiology of hospital infections due to Pseudomonas aeruginosa was investigated by pyocin typing. The typing method, which determined the pyocin activity of clinical isolates of P. aeruginosa on 27 indicator strains, was 43.7% reproducible, but elimination of 9 indicator strains doubled the reproducibility and yielded more readable pyocin inhibition zones. Seventy-eight of 1,084 isolates (7.2%) were untypable. In the second part of the study, P. aeruginosa was isolated from 110 patients (5.4% of all admissions) in a 3-month period and typed with the revised method. Twenty pyocin types were identified, 10 of which were obtained from five or more patients. P. aeruginosa was isolated from 45 of 353 environmental samples, including water fountains, ice machines, bar soaps, and germicide solutions for toilet brushes. Twenty percent of the environmental samples were untypable but, among typable strains, the five most common environmental strains were the same as the strains most frequently isolated from patients. The organism was frequently isolated from noses (39%), throats (39%), and stools (29%) of patients with P. aeruginosa infections or colonizations in urine, sputum, surgical wounds, or skin lesions. Six of eight patients had P. aeruginosa in their tracheostomy wounds. Autoinfections by strains already acquired on carrier sites may be significant.