Herdmate comparison used currently by Holstein-Friesian Association of America to calculate predicted difference type was compared to two iterative procedures, adjusting for average PDT of herdmates' sires and average cow index of herdmates respectively, and a mixed model method. PDTs were calculated using the different methods for 6816 bulls with ten or more daughters. Mean PDT for the herdmate comparison was -.18 while means for the two respective iterative procedures were -.24 and -.19. Mixed model proofs sum to zero. Adjustments for average PDT of herdmates' sires increased variance of proofs compared to the other systems. Mixed model PDTs were least variable.

Correlations between herdmate average and two measures of genetic merit of herdmates, average PDT of herdmates' sires and average cow index of herdmates, were .48 and .45. Positive regression coefficients of final score and herdmate average on two measures of genetic merit of herdmates also indicated a positive association between herd average and level of competition.

The effects of failure to account for genetic merit of herdmates under present system was examined. Progeny were divided into low, middle, and high herdmate average groups and ·summaries were calculated for 778 sires with at least 20 daughters in each of three groups. Mean PDTs calculated by the herdmate comparison were .12, -.28, and -.45 for the low, middle, and high groups respectively. Eighty-three percent of the low herdmate comparison PDTs exceeded summaries in high group and 70% of low group evaluations were greater than respective middle group PDTs indicating biases in present sire summaries due to failure to account for genetic level of competition. Adjustment for average PDT of herdmates' sires and average cow index of herdmates reduced the advantage of PDTs calculated in low levels. The mixed model method appeared to be most effective in reducing biases in favor of bulls proven in below average herds. Summaries in low group were consistently higher than PDTs calculated in the other levels due to inability to completely remove differences in selection intensities among herds. PDTs calculated in high group were consistently less variable than those calculated in low herd levels. Means of final score, average PDT of herdmates' sires, and average cow index of herdmates were larger and less variable than corresponding means in low group demonstrating increased selection for type in high scoring herds. Accuracy in sire evaluation for type will be improved by adjusting for genetic merit of herdmates, however bulls proven primarily in low scoring herds will continue to be favored slightly.