Corn yield trials comprising 5 to 6 replications were planted in the years 1945 to 1953 in 4 to 13 locations in order to evaluate the y eld variability of several cultivars, single, double cross and three way hybrids in the years and in the locations. The locations are representative of the areas where corn is grown in the S ate of S. Paulo. Estimates of the variance components were calculated for the cultivars (s²v) and for the cultivar x locations (s²vl), cultivar x year (s²va), cultivar x locations x year (s²vla) and error (s²E) variances. The cultivar x locations is of higher magnitude than cultivar x year interaction, which indicates that the environmental conditions at the locations considered had a larger effect on the yield variability. Based on the data obtained the average genetic advance was calculated for several combinations varying the number of locations, years and replications, in order to find out the most efficient ones to be adopted for the conditions prevailing in the State of S. Paulo.