The increased incidence of nosocomial Legionnaires' disease in two hospitals prompted investigation of possible environmental sources. In the search for an effective DNA-typing technique for use in hospital epidemiology, the performance and convenience of three methods-SfiI macrorestriction analysis (MRA), amplified fragment length polymorphism (AFLP), and arbitrarily primed PCR (AP-PCR)-were compared. Twenty-nine outbreak-associated and eight nonassociated strains of Legionella pneumophila with 13 MRA types and subtypes were investigated. These strains comprised isolates from bronchoalveolar lavages, from environmental, patient-related sources, and type strains. All three typing methods detected one predominant genotype associated with the outbreaks in both hospitals. All of them correctly assigned epidemiologically associated, environmental isolates to their respective patient specimens. AP-PCR was the least discriminating and least reproducible technique. In contrast, AFLP was demonstrated as being the method with the best interassay reproducibility (90%) and concordance (94%) in comparison to the genotyping standard of MRA and the epidemiological data. Analysis of AFLP fragments revealed 12 different types and subtypes. Because of its simplicity and reproducibility, AFLP proved to be the most effective technique in outbreak investigation.