Genetic analysis was performed on 72 Italian A-T families ascertained through 91 probands. The frequency of the A-T gene was estimated through Dahlberg's formula, which uses the frequencies of first cousin marriages among the patients' parents and in the general population. Data on consanguinity in the general population were obtained from the collection of Catholic Church dispensations for the period 1910–1964. Accurate estimates of the disease frequency were made by subdividing data into 5-year periods and between provinces, in order to take into account temporal (decrease over time) and spatial (higher frequency in Southern Italy) heterogeneity. The estimate of the gene frequency for the whole sample was q = 0·012 ± 0·0065, corresponding to a heterozygote frequency of 2·34% and to a disease frequency of 1 in 7090. When considering q before and after 1970, there was an apparent increase of heterozygote frequency from 1·69 to 3·43%, perhaps due to a better ascertainment of the disease in the latter period. A segregation ratio of 0·249 ± 0·043 was obtained by analyzing the offspring of the 72 families under the model of incomplete multiple selection. This value is not significantly different from 0·25, as expected for an autosomal recessive disease.