Due to the low evolutionary rate and the limited horizontal transmission of human T-lymphotrophic virus type I (HTLV-I), its phylogenetic analysis reveals the movements and contacts of ancient populations. Since simian strains cannot be distinguished from human strains by phylogenetic criteria, this virus has appropriately been called primate T-lymphotropic virus type I (PTLV-I). We sequenced the LTR of six PTLV-I strains: three HTLV-I strains from African patients with tropical spastic paraparesis (TSP) (Equateur, Zaire), two laboratory HTLV-I strains of Japanese origin, MT-2 and MT-4, and one STLV-I from a baboon of the primate center in Sukhumi, Georgia. We applied four phylogenetic inference methods: neighbor-joining (NJ), unweighted pair group method using arithmetic averages (UPGMA), Fitch and Wagner parsimony (pars), and maximum likelihood (ML), to these 6 LTR sequences and 18 published LTR sequences (cosmopolitan, African, and Melanesian HTLV-I strains and African and Asian STLV-I strains). Three major HTLV-I subtypes can be identified with all four methods: the cosmopolitan HTLV-Ia, the central African HTLV-Ib, clearly descendant from a STLV-I CH-like African ancestral simian strain, and the Melanesian HTLV-Ic, probably descendant from an Asian STLV-I strain. We observe a segregation of PTLV-I sequences according to their geographical origin and not according to host species. The Zairean strains form a cluster closely related to an STLV-I strain isolated from a chimpanzee (STLV-I CH) and distinct from western African strains, which belong to the cosmopolitan subtype of HTLV-I. The Sukhumi STLV-I strain found in a captive-born baboon was of Asian descent. We experienced rooting problems with UPGMA when using HTLV-II as an outgroup. Concordant results with all four methods were obtained by eliminating HTLV-II LTR sequence fragments with bad alignment to HTLV-I. This resulted in a HTLV-II root node on the African STLV-I TAN90 terminal branch (with bootstrap values above 92% for the NJ and pars methods) and not on the Asian STLV-I PtM3 branch, as has been derived by others based on their use of UPGMA. The results of the analyses also support a higher evolutionary rate of PTLV-I in Asia, implying that the trees obtained with the NJ and ML methods have a higher reliability. These results are more compatible with an ancient African origin of PTLV-I than with an Asian origin.