Sex differences of physical traits can be regarded as population characteristics, and the analysis of their interpopulation vari ation is relevant to phylogenetic studies as well as to the evaluation of factors operative on sex differences. Since interpopulation variations of sex differences tend to be small, attention should be paid to distinguish them from sampling errors. The author proposed a multivariate statistical method for the significance test and description of the interpopulation variation of sex differences. The method is essentially a factorial discriminant analysis, but allows the inequality of sample sizes and the use of the basic statistics commonly reported in anthropological works. By this method, the sex differences of the total of 22 measurements of humerus, radius, ulna, femur, and tibia were compared among the Modern Japanese consisting of five regional populations, the prehistoric Yayoi (Doigahama), Earliest/Early Jomon, Late/Latest Jomon, the modern Sakhalin Ainu, Koreans, and Indians with reference to "size" and "shape".<br>The interpopulation variation of sex differences was statistically highly significant in multivariate comparison while it was non-significant for most of the variables in univariate comparison. And the most part of the multivariate interpopulation variation of sex differences was attributed to "shape".<br>No statistically significant difference was detected between the sex difference of Indians and that of Modern Japanese. This raises the possibility of cross-racial application of the common discriminant function coefficients to the sex determination from long bone measurements. But its use in the field of forensic anthropology requires some restrictions since the sex difference of Modern Japanese exhibited a statistically significant geographic variation.<br>The total interpopulation variation is expressed by the sum of two terms; the sexcommon term or the interpopulation variation common for both sexes and the sex difference term or the interpopulation variation of the male mean minus female mean. A Parallelism was found for the interpopulation relationships between these two kinds of variations. The humeral robustness was greater in Late/Latest Jomon than in Earliest/Early Jomon for both terms of interpopulation variation, in other words, the humeri of the later Jomon periods were more robust than those of the earlier periods, and the difference was greater for males. This may be interpreted by some ecological change in the subsistence of the Jomon people. Sakhalin Ainu shared some traits with Jomon populations, but showed a resemblance to Koreans and Modern Japanese. A possible interpretation for this is to assume the existence of common NeoMongoloid genes for these populations introduced independently through two different routes from the Asian continent. The Doigahama population of the Yayoi period showed a close similarity to Koreans for both terms of interpopulation variation. This seems to reflect the genetic affinity between them.