In this paper are reported the results obtained in 10 experiments located at seven different sites comprising the main soil types of the State of São Paulo. Such experiments were carried out from 1949-50 to 1953-54 and, with exception of one, were conducted on the same plots for two or three consecutive years, giving a total of 22 annual results. The rates of application were 60 kilograms of P2O5 per hectare and, according to the experiment, 20 or 30 kilograms of N and 30 or 40 kilograms of K2O per hectare, these nutrients being applied alone or in several combinations. Nitrogen, phosphorus and potassium were used respectively as Chilean nitrate, superphosphate and potassium chloride. Phosphorus was always applied at planting time, whereas nitrogen and potassium were applied at three different times: at planting, 30-40 and 60-80 days after the emergence of the plants. The fertilizers applied at planting time were, as usually, added to the furrows opened for the seed and slightly mixed with the soil just before planting; the applications after the emergence were done as top-dressing. The response.to nitrogen was generally much greater when it was top-dressed 30-40 days after emergence than when the application was done at planting time; the former also gave better results than the delayed top-dressing (60-80 days after emergence). The division of the dose of nitrogen in two equal portions - studied in four experiments, comprising eight annual results - gave still better results than a single application 30-40 days after emergence. When the two portions were top-dressed the response was fairly greater than in the combinations in which one portion was applied at planting time in contact with the seed. A retrospective study based on the present experiments and in others reported in a previous paper, involving 30 annual results, confirmed entirely the above conclusions. In the experiments permitting the comparison of the effect of potassium applied at different times, top-dressing showed to be inferior to the application in the planting furrows. In this case, however, the results were satisfactory only when it rained during the period immediately after planting. Otherwise, germination and stand were impaired. Apparently the top-dressed potassium did not move into the root zone quick enough to benefit the cotton plants when they needed it most; this explains the poor results of that method of application. As the risk of immediate leaching of potassium is generally low, it should be applied before or at planting time in a position easily accessible to the roots, but not in contact with the seed.