A model for the genetic evaluation of dairy bulls and cows for subjectively scored conformation traits using random contemporary groups is proposed. Variance parameters for the random factors of the model were estimated for 14 traits in Jersey cattle. Three alternatives for reducing the influence of parent averages on estimated breeding values were compared using the new model. One alternative was to modify the construction of the inverse of the additive genetic relationship matrix, and the other alternatives involved the application of two artificially high heritability values. Comparisons of estimated breeding values of bulls and cows from the three alternatives and usual BLUP with the new model were made to determine the effect on rankings of animals. Alternatives with higher heritabilities were similar to usual BLUP, but the modified inverse matrix method caused more rerankings especially for cows. The question is whether procedures for calculating estimated breeding values should be modified to reduce the effect of inbreeding in subsequent selections or whether genetic evaluation procedures should be left unaltered and strategies for optimum mating decisions to avoid or minimize inbreeding should be adopted.