Fifty four experiments designed to test corn varieties and hybrids were carried out in nine experiment stations of the Instituto Agronômico de Campinas from 1943 to 1954. Two types of analysis were made with data obtained from these experiments : first, they were analysed as from randomized block experiments ; second, as from completely randomized designs. Randomized blocks were 30 per cent more effective than completely randomized designs. This result indicates that a randomized block experiment with six replications offers as much precision as that obtained with eight replications in the latter design. A mean coefficient of variation of 16.1 was found. The experiments carried out at the experiment stations of Ribeirão Prêto, Tietê, Ipanema, and Capão Bonito had the highest coefficients of variation.