In August and September 1977 a discrete cluster of 27 serologically or pathologically confirmed cases of Legionnaires' disease, plus six highly presumptive cases were identified in the area of Kingsport, Tennessee. Three patients died. Most patients manifested severe pneumonia and fever; no mild or asymptomatic disease forms were recognized despite intensive case-finding efforts. Illness was epidemiologically associated with residing, visiting, or working in one geographic area of Kingsport, residence there being the factor most strongly associated. Although the attack rate for area residents was 0.64%, the randomly determined prevalence of serologic reactors was 5.2%, which is not significantly different from that in a nonimplicated control neighborhood. The epidemic did not correlate temporally with any identified environmental or demographic event. No source of the bacterium was found either by a detailed case-control study of area associations or by bacterial isolation from sentinel guinea pigs or environmental specimens. There was no evidence of person-to-person spread.