This paper reports the results of 10 experiments conducted on sandy soils of the State of São Paulo to study the effect, on the cotton crop, of green manure with velvet beans (Stizoiobium sp.) and mineral nitrogen, phosphorus, and potassium, these treatments being used single or combined. When the velvet beans were sown, the plots without green manure were planted with cotton, these crops receiving no fertilizer. In the following year both groups of plots were splitted for the application, or not, of the mineral fertilizers. To study the residual effect of green manure, in four of the trials the plots were fertilized and planted with cotton another year. In two of the experiments, located in areas that had been under forest not long before, none of the treatments increased appreciably the yields. Averaging the results of the other trials, conducted in areas repeatedly cropped, the response to phosphorus was fair and that to potassium very poor. The utilized soils were not low in potassium and some of them had received phosphorus in the previous crops. Even in the absence of green manure, the response to mineral nitrogen was not satisfactory, apparently because in most cases it was applied too late. On the cropped soils, the average effeet of green manure corresponded, in the first year after its application, to +355 kg/ha of seedcotton in the absence and to +280 kg/ha in the presence of mineral nitrogen. In the absence of the latter and of potassium, the application of phosphorus enhanced the response to green manure, and the treatment phosphorus + green manure, which was almost as effective as others much more expensive, increased the yield 51%. The residual effect of green manure, in the average of the four trials designed to study it, corresponded to only 14% of its immediate effect.