Abstract
The breeding values of 3975 horses, included in the data set, were assessed simultaneously for the 10 traits by a Best Linear Unbiased Prediction (BLUP) procedure. The correlation among the traits were considered and the additive genetic relationship among the horses were accounted for. A description is given of the computational procedure. Canonical transformation of the variates substantially reduced the computing effort entailed and made it possible to solve the BLUP equations separately for each trait. The method is currently being applied to routine evaluation of breeding horses in the Icelandic toelter horse population.