Experiments designed to compare three levels of nitrogen, five of phosphorus, and four of potash were planted for three consecutive years (1949 through 1951) at Campinas in the "terra roxa misturada" type of soil. Treatment 1-1-1 received 25 kg/ha of each, N, P2O5, and K2O, applied respectively as Chilean nitrate, superphosphate, and potassium chloride. The fertilizers were placed in the furrow at planting time and mixed with the soil direct under the seed. Injury to the germinating seed resulting from salt concentration was noted only in 1949, when germination was substantially affected, especially in plots that received high nitrogen and potash. The number of plants per plot after thinning was, however, comparable for all treatments. Phosphorus promoted a considerable increase in yield. A study of the adjusted means showed that a single dose of phosphorus increased the yield in 521 kg/ha when compared with plots receiving only NK. Double rates of phosphorus increased the yield in 806 kg/ha (44%), but higher rates of this element did not promote any further increase. No gain in yield due to the application of nitrogen or potash was noticed in these experiments.