A total of 2, 415 blood samples from 8 populations of Japanese native horses, 18 populations of Asian native horses and 8 breeds of European horses (race-horses, draft-horses and ponies) collected since 1971 were electrophoretically examined at 33 biochemical genetic loci. Gene diversity within populations or breeds measured by the proportion of polymorphic loci (Ppoly) and the expected average heterozygosity per individual (H) was observed to be higher in the Mongolian native horses than in other populations or breeds, and in the Tokara horse of Japan was on the lowest level. Gene diversity between populations or breeds was evaluated by the principal component analysis (PC). It was also measured by computing the Nei's standard genetic distances from the allele frequency data, and the dendrograms were drawn from the genetic distance matrix by the unweighted-pair-group (UWPG) method and the neighbor-joining (NJ) method for estimating the phylogenetic relationships among the horse populations or breeds. The results revealed that the horse populations or breeds were roughly classified into four clusters, that is, Asian native horses, European race-horse breeds, European draft-horse breeds and European pony breeds. The cluster of Asian native horses included some subclusters of Mongolian native horses, south and southeast Asian native horses, Indonesian native horses, and Japanese native horses, although the Batak horse of Indonesia was observed to diverge from the subcluster of Indonesian native horses and the Tokara and Noma horses of Japan diverged remarkably from the subcluster of Japanese native horses in gene constitution. Such diversifications are postulated to be on account of the effect of random genetic drift and/or a bottle-neck effect resulting from their geographic isolation and small reproductive population size. The Mori-Hayashida's two-wave immigration hypothesis for formation of eight local populations of the Japanese native horses can not be supported from our genetic data.