A model is developed to account for recent molecular observations. It postulates four alleles: normal (N), small rather stable insert (S), larger, unstable insert (Z), and large insert (L). The last-named allele causes the fragile-X phenotype, inactivation of the FMR1 locus by methylation, and mental impairment; the FMR1 locus (for fragile-X mental retardation locus 1) resides in the FRAXA region. When this model is fit to pre-molecular data, the Z allele appears to be no more frequent than L, while the S allele is polymorphic. Predictions of the model are in reasonable agreement with observation and suggest much more powerful tests of molecular data, including the Laird hypothesis that conversion of Z to L does not occur in active X chromosomes.