By analysis of variance the total of all squared deviations from the grand mean of some heterogeneous population may be divided into 2 variances, namely, variability about the mean of each of the more homogeneous elements in that population and the variability of these partial means about the grand mean of the entire population. In head measurements for 13 traits it is found that the mean squared variance within a family is. in all 13 traits, less than the variance between families, and that the ratio of one to the other is always significant and usually highly significant.