Fifty-eight sugarcane virus isolates were obtained from leaves showing mosaic symptoms, and collected in Cameroon (26 isolates), Congo (20 isolates), Egypt (1 isolate), South Africa (3 isolates) and the U.S.A. (8 isolates). All these isolates belonged to Sugarcane mosaic virus (SCMV) based on the amplification product obtained by RT-PCR with SCMV-specific primers. The amplicons (0.9 kb) from the coat protein (CP) coding region were cloned, sequenced and compared to each other as well as to the sequences (GenBank accessions) of 16 SCMV isolates from sugarcane (Australia, South Africa and U.S.A.) and 12 SCMV isolates from maize (Australia, Germany and China). Maximum likelihood and maximum parsimony analyses robustly supported two major monophyletic groups that were correlated with the host of origin: the SCE or sugarcane group that included all isolates from sugarcane and the MZ or maize group that contained all isolates from maize. The 86 virus isolates were distributed in 13 minor phylogenetic groups, four (I-IV) restricted to maize and nine (V-XIII) to sugarcane. A strong correlation was observed between the sugarcane groups and the geographical origin of the SCMV isolates. Each SCMV type strain from sugarcane (A, B, D, E and SC) was distributed in a different phylogenetic group or subgroup. The 26 isolates from Cameroon constituted a relatively homogenous group (group V) whereas the 20 isolates from Congo belonged to two other groups (VI and VII). All the isolates from Cameroon and Congo were different from the SCMV type strains and other strains or isolates studied so far. It appears, therefore, that the population of SCMV from sugarcane in Africa contains virus genotypes that have not yet been described.