A study is described in which the spectral similarities in the voices of monozygotic, dizygotic and non-twin controls is compared. The results are based on a statistical comparison of the differences between pairs of individuals at a number of spectral points. The data are interpreted as showing that monozygotic twins are more similar in the higher but not the lower spectral ranges. It is suggested that the lower frequencies are more sensitive to transitory factors while the higher frequencies are more likely to reflect anatomic similarity.