An epidemiological and microbiological investigation of a cluster of eight cases of Legionnaires' disease in Los Angeles County in November 1997 yielded conflicting results. The epidemiological part of the investigation implicated one of several mobile cooling towers used by a film studio in the centre of the outbreak area. However, water sampled from these cooling towers contained L. pneumophila serogroup 1 of another subtype than the strain that was recovered from case-patients in the outbreak. Samples from two cooling towers located downwind from all of the case-patients contained a Legionella strain that was indistinguishable from the outbreak strain by four subtyping techniques (AP-PCR, PFGE, MAb, and MLEE). It is unlikely that these cooling towers were the source of infection for all the case-patients, and they were not associated with risk of disease in the case-control study. The outbreak strain also was not distinguishable, by three subtyping techniques (AP-PCR, PFGE, and MAb), from a L. pneumophila strain that had caused an outbreak in Providence, RI, in 1993. Laboratory cross-contamination was unlikely because the initial subtyping was done in different laboratories. In this investigation, microbiology was helpful for distinguishing the outbreak cluster from unrelated cases of Legionnaires' disease occurring elsewhere. However, multiple subtyping techniques failed to distinguish environmental sources that were probably not associated with the outbreak. Persons investigating Legionnaires' disease outbreaks should be aware that microbiological subtyping does not always identify a source with absolute certainty.