Covariances among first-lactation milk records of cousins.

Covariances among first-lactation milk records expressed as deviations from herdmate averages of Holstein cows related as cousins of varying degree, as daughterdan1, as full and maternal sibs, and as aunt-niece of varying degree were examined to dete~nine whether genetic maternal effects are important. The multiple regression of the expected coefficients of additive direct, additive maternal, dominance direct, and dominance maternal genetic variance components; and additive direct by additive maternal and dominance direct by dominance maternal genetic covariances on the 24 estimated covariances between pairs of relatives gave a multiple correlation coefficient of .95. The regression of actual covariance on the coefficient o£ additive direct genetic variance gave a correlation of .92. These results suggest for these data that only additive genetic effects are important for this trait. Additive genetic variance accounted for 38% of total variance. The intercept of the regression line indicated that the average environmental eovariance among pairs of relatives accounts for 2 to 3% of the total variance. Bradford and Van Vleck (1) reported a difference in heritability estimates for first-lactation milk production expressed as deviations from herd-mate averages from daughter-dam regression (.44) and paternal half-sib correlation (.25). Several analyses of more of the same data have not completely deter~nined the reasons for the difference (3, 5-7). Additive by additive genetic variance is apparent ly not impor tant (5). Enviromnental correlations between daughter and dam records apparent ly account for .01 to .02 of the total variance (4). The increase in variance with change in time and production may also bias the daughter-dam regression upwards by about 10% (3). Genetic maternal effects a r e a possible cause of the remaining difference between the two methods of estimation. Van Vleck and Har t (7) found evidence to cast doubt on the importance of maternal effects. Willham (8) has suggested that the best sets of relatives to use to estimate Received for publication August 14, ]965. the importance of genetic maternal effects are sets of cousins. This study attempted to estimate genetic and genetic maternal components of variance from the covariances between records of pairs of cousins with additive relationships of 1/]6, 1/8, 5/16, and 3/8. Materials and Methods The data were first.lactation (305-day, 2 × , M.E.) milk production records expressed as deviations from herd-mate averages (2) of registered Holstein cows. These records were taken from the files of the New York Dairy Records Processing Laboratory. Covariances between records of pairs Of grand daughters (cousins) with relationships as described in Table i for the four mating patterns were computed. Records of the dams were also available. Thus, covariances and variances were computed among the set of four records for each mating pa t te rn-Danghter One, Daughter Two, Dam One, and Dam Two. The expected coefficients of an arbi t rary six components of genetic and genetic maternal variance included in the various covariances are given in Table 2 for the four analyses. (See Willham (8) for the derivation of these coefficients.) Results and Discussion The expectations shown in Table 2 are all based on the assumption of no selection on any of the records. The means and variances shown in Table 3, however, suggest that, at least, the records of dams were selected. Consequently, the eovarianee between any daughter (niece) and any dam (aunt) was expressed as the regression of the daughter on the dam (aunt) . The covarianee between cousins was expressed as the correlation between them, and similarly the covarianee between the dams (maternal or full sibs) was expressed as the correlation. These standardized estimates appear in the right-hand column of Table 2. Note that in each analysis there were two estimates from daughter-dam regression and two estimates from niece-aunt regression. An attempt was made to equate the estimates to their expectations for each analysis by limiting the number of components to four or less. In all cases the equations were not consistent, in that components such as Am ', which must be