To test the hypothesis that the frequencies of normal alleles (ANs) with a relatively large number of CAG repeats (large ANs) are related to the prevalences of the dominant spinocerebellar ataxias (SCAs)-SCA types 1, 2, 3 (Machado-Joseph disease), 6, and dentatorubral-pallidoluysian atrophy (DRPLA)-we investigated the relative prevalences of these diseases in 202 Japanese and 177 Caucasian families and distributions of the number of CAG repeats of ANs at these disease loci in normal individuals in each population. The relative prevalences of SCA1 and SCA2 were significantly higher in Caucasian pedigrees (15% and 14%, respectively) than in Japanese pedigrees (3% and 5%, respectively), corresponding to the observation that the frequencies of large ANs of SCA1 (alleles >30 repeats) and of SCA2 (alleles >22 repeats) were significantly higher in Caucasians than in Japanese. The relative prevalences of MJD/SCA3, SCA6, and DRPLA were significantly higher in Japanese pedigrees (43%, 11%, and 20%, respectively) than in Caucasian pedigrees (30%, 5%, and 0%, respectively), corresponding to the observation that the frequencies of large ANs of MJD/SCA3 (>27 repeats), SCA6 (>13 repeats), and DRPLA (>17 repeats) were significantly higher in Japanese than in Caucasians. The close correlations of the relative prevalences of the dominant SCAs with the distributions of large ANs strongly support the assumption that large ANs contribute to generation of expanded alleles (AEs) and the relative prevalences of the dominant SCAs.