Estimates of genotype by environment interaction variances were obtained from the western Canada Co-operative fall rye tests grown in 1963–1967. All first-order interactions and the second-order interactions were significantly greater than zero at the 0.05 level of probability. Although the estimate of cultivar by year interaction variance was relatively small, it accounted for 40% of the variance of a cultivar mean when only three years of testing were considered. However, testing in 20 locations for three years with four replicates could detect yield differences of approximately 10% of the mean of the highest yielding cultivar. If further reduction of the yield difference detectable is desired, more locations should be included in the test.