The probability of a heterozygote being affected was estimated from the distribution of frequencies of early-replicating fragile X [fra(X)] chromosome in normal and mentally retarded heterozygotes, taking into account the prior probabilities of 0.35 for mental retardation and 0.65 for normality. The estimated probability of a heterozygote with 100% early-replicating fra(X) being mentally retarded was 78%, which coincides with the value of penetrance in males. Therefore, the manifestation of retardation in females seems to differ from that in males due solely to X inactivation. The frequencies of early-replicating fra(X) were significantly increased among the heterozygotes with the highest frequencies of fra(X) both in the normal group and in the mentally retarded. The mean frequencies of early-replicating fra(X) were 0.42 and 0.68 for normal and mentally retarded heterozygotes, respectively. Considering the overall frequency of retarded heterozygotes as 0.35, the mean frequency of early-replicating fra(X) obtained for all heterozygotes was 0.51, which is in accordance with the hypothesis of random X inactivation. Thus the fragile site appears to have equal chances of being detected when located either on the early- or on the late-replicating X. This leads to the conclusion that the frequency of the fragile site is a consequence of the proportion of cells with the active Martin-Bell syndrome (MBS) gene and not the result of a better visualization of the site on the early-replicating X.