In the experiment reported, conducted for two years on poor soil derived from the Botucatu sandstone, superphosphate was compared, in the presence of NK with Olinda and Araxa phosphates (ground natural phosphates), and mixtures containing 1/3, 1/2 and 2/3 of the total dosis of P2O5 (65 kilograms per hectare) as superphosphate, the remaining being supplied by one of the natural phosphates. The average yield of the two years was only 196 kg/ha (kilograms per hectare) in the unfertilized plots and varied around 650 kg/ha in the better treatments. Although much higher than that of the unfertilized treatment, the latter yields were unsatisfactory, apparently due to the use of inadequate rates of application, deficiency of elements other than those included in the treatments, and adverse weather conditions in the second year. Used as the only phosphorus source, the natural phosphates were far inferior to superphosphate; when mixed, however, they proved about as effective. The responses to the mixtures decreased as the contributions of superphosphate were reduced, but the differences were small.