Epidemiologic studies indicate that human T-cell lymphotropic virus type I (HTLV-I), the causative agent of most cases of adult T-cell leukemidlymphoma (ATLL) in Southeast Japan and the Caribbean islands and the probable cause of a progressive neurological disorder often referred t o as tropical spastic paraparesis, occurs with unusual geographic clustering. The current large-scale serosuney was undertaken to im- prove our understanding of HTLV-I revalence in different parts of the world. We analyzed 43,&5 serum samples collected from various geographic locales worldwide; 76% of these sera came from clinically healthy donors. Samples were initially screened by an enzyme-linked immunosorbent assay (ELISA) and 4,353 were further evaluated by means of competition assays. In this study, which did not include sera from endemic areas of Japan, a high prevalence of infection was observed in several countries in the Caribbean basin. A significant age-sex difference was observed between populations in the Caribbean and non-endemic regions of Japan. The reason for the male excess in non-endemic areas of Japan will require further study, while the female excess in the Caribbean basin is compatible with the previously described pattern for other HTLV-I-endemic areas. A newly recognized area of possible endemicity was southern Florida, where evidence of infection with HTLV-I or a related virus was found in a group of native Americans whose sera were collected in 1968. In certain parts of the world, particularly sublaharan Africa, important problems in determining specificity of reactivity occurred, probably because of cross-reacting antibodies. No pattern was detected that could explain the cross-reactivity solely on the basis of geographic areas, specific patterns of non-viral parasitic infection, or methods of handling the specimens. It is possible that these cross-reactivities are antibodies t o proteins from HTLV-I-related retroviruses yet to be discovered. 'Antibody determined by confirmation or U S A ratio 2 8 . Data presented only for individuals with no clinical disease in vo$phic locales with the largest numbers of available results.-2Female prevalence significantly greater than male (p< e prevalence significantly greater than female @=0.003).