A sample of 173 cases of cytogentically diagnosed 21-trisomy mosaicism, 225 cases of complete 21 trisomy and 302 normal controls were used to form discriminant functions for mosaics versus normal controls, and also for complete 21 trisomics versus normal controls; for each of these discriminants optimal sets of ten dermatoglyphic characters were selected. The mean discriminants and discriminant coefficients were calculated on the basis of these characters, using Fisher's method. The relative position of mosaics in respect of the mean discriminant values, was found closer to the Down's syndrome group than to the normal controls. Consequently, reasonably, good separation of the mosaics from the normal controls has been achieved: the distance between the two mean discriminants is greater than two standard deviations. The distribution curves were then constructed of individual discriminants, calculated in a majority of cases from all the three groups. The distribution curve for mosaics has been found bimodal so that more detailed study of the individuals forming its first peak was undertaken; a possible relationship between the degree of dermatoglyphic distortion and cytogenetic findings is discussed and some limitations of both methods are indicated. The results concerning discriminant function indices are recommended for further use, particularly in genetic counselling. Their application in the studies of populations of parents of Down's syndrome cases is demonstrated in the group of 79 parents of two or more Down's cases and the group of 111 parents of one Down's case. On the basis of these data it is postulated that there may be some proportion of undetected mosaics amongst the parents of Down's cases, but larger numbers are necessary to support and further specify these preliminary results.