Date of this Version
Dairy cattle evaluation schemes routinely assume homogeneous variance with respect to environment. Increasing evidence suggests the presence of systematic changes in variance components associated with mean level of performance. Best linear unbiased prediction procedures that account for heterogeneity are reviewed. The consequences of incorrectly assuming homogeneity for evaluation are demonstrated for a progeny test and an artificial breeding program that screens dams of sires from heterogeneous populations. Selection assuming homogeneity can be very efficient when heritability, and therefore accuracy of selection, is greatest in the more variable environment. Conversely, appreciable reduction in response results when heritability is greater in the less variable environment.