Yield data for barley, oats, hard red spring and durum wheats from 10 years of testing m western Canada were analyzed. Variety, variety × location interaction, and error components of variance were estimated and the relative values of the latter two were found to differ from those already reported in the literature. The relative values of the variety × location interaction and error variances are important factors in determining the optimum allocation of test sites within years, and in designing experiments to test different aspects of quantitative genetic theory. The year-to-year variability of these relative values as demonstrated in this study will have to be considered in certain applications of estimates of variance components.