Modified contemporary sire evaluation procedures, using first lactation records only, were used to compute Predicted Differences for all Holstein bulls having 10 or more daughters calving in any calendar year between January 1964 and June 1973. Significant heterogeneity of within sire daughter milk yield variance was found although only 18% of the differences in sire variances could be accounted for by factors examined. Repeatability of sire daughter variances suggested that 23 to 32% of the differences in sire variances could be attributed to sire differences. Heritability estimates of sire daughter variance were low to moderate and ranged from -.14 to .29. The standard deviations of mean milk yield, adjusted for number of contemporaries, of paternal half-sibs in the same herd-year-season ranged from 999 to 522 kg for groups of 2 to 16 daughters. Repeatabilities of contemporary comparison sire summaries were examined to determine their impact on the accuracy of sire selection. A set of 455 bulls have final Repeatabilities exceeding 90% was used to examine the effects of initial Predicted Difference and Repeatability on final Predicted Difference. The effect of initial Predicted Difference on final proof was highly significant (P .01) and was the most important predictor of final proof. However, the linear effect of Repeatability on final proof was also significant (P< .05), indicating an increase of 1.53 kg in final proof per 1% increase in initial Repeatability. Relationships between semen cost and Repeatability indicated that low Repeatability bulls frequently represented most economical semen purchases. / Master of Science