Data of 109.200 lactations were used to verify the effects of heterogeneity of variance on genetic evaluation of cows and bulls of Holstein breed raised in Brazil. Total milk production adjusted for adult age was used to classify the herds among three classes of phenotypic standard deviation: Low (<1427 kg), medium (between 1427 kg and 1625 kg) and high (>1625 kg) classes. Records from the first lactation were analyzed for total milk production adjusted for adult age (TOTAJU) and for adult age and 305-day lactation length (TAJU305). Milk yield means, and genetic, residual and phenotypic variance components increased from the low to the high standard deviation classes. Genetic correlation's among milk production were .97, .89, .91 for TOTAJU and .97, .92, .96 for TAJU305, respectively, between low and medium, low and high and medium and high classes. Correlation's among genetic values for the low, medium and high standard deviation classes obtained from multitrait analyses (considering the expression of the trait in the three classes as different) and those obtained from the general univariate analysis (all classes as a unique trait), were all close to one. However, the bulls presented high genetic values in herds from high standard deviation classes. In genetic evaluation programs, it is important to consider differences in variability among herds, otherwise, under selection, the most variables herds would contribute with the majority of the animals, and the genetic evaluation of the animals could be more a function not only from its genetic potential, but also from its environment where the progenies expressed their traits.
dairy cattle; genetic parameter; variance component
