An extensive investigation was carried in the years 1952/53 to study the importance of rainfall distribution on coffee yields. Fourteen coffee counties were selected to represent the different soil and climatic conditions prevalente in the State of São Paulo, for the purpose of correlating official estimates of coffee yields and monthly precipitation for the period 1945 up to 1953, known to have had quite dry spells. Linear correlation was shown ta decrease with the average coffee yields of the counties, losing its significance for the counties with lowest yields. Closer linear correlation was found for the drier period of April-to-September, as well as for July-to-September, this period appearing to be the most important as there is the coincidence of coffee blooming with the driest period in the year. Linear correlation studies were also made with soil water defficiences determined by using Thornthwaite's method. The results obtained paralleled those of the rainfall-yield studies previously mentioned. The data best suited for the linear correlation studies were found to be progressive two years averages of the years t n and t n + 1 for yield data, and t n-1 and t n for rainfall dato, which appeared to minimized the biennial bearing effect on coffee yield. The high degree of correlation obtained might suggest the possibility of improving the method for the purpose of yield estimates. However, the law values obtained for the regression coefficients show that the importance of rainfall distribution is minimized by the biennial bearing effect not completely eliminated and other factors not considered in this study, as shown by the results with counties of lowest yields. Higher values of the regressions coefficients for counties of highest yields may indicate the method will prove better when applied to coffee of high yielding capacity.