Upon studying 178 cotton fertilizer experiments, conducted on soils cf the State of São Paulo, from 1957 up to 1965, a correlation was found between cotton yield responses to application of 60 kg/ha of K(2)0 and soil test results. The effect of potash was found to be associated with the Ca+ +/K + relation and the amount of K + , exchangeables in soils. A multiple regression with R = 0.804++ could explain, thereby, 65% of variation in the responses. With regard to linear simple correlations, the variable Ca+ +/K+ explained 59% of the fertilizer effects, but the amount of exchangeable K+ could not explain, by itself, more than 23%. No additional gain in accuracy was obtained by considering a quadratic form for these relationships.