This paper deals with the statistical analysis of data referring to raw coffee processed by cooperatives in the State of São Paulo, Brasil. Six cooperatives were studied, with a total 289 batches of coffee for all of them. For each batch the weighted average of defects was determined, before and after processing. The square roots of the data thus obtained were submitted to the technique of analysis of variance. Significant improvement in coffee type was proved in all cases, with only one exeption. The exception ocurred for the "Cooperativa da Média Sorocabana", where observed improvement was not statistically significant, this being probably due to tre following reasons: a) data of only 7 batches were available; b) the cooperative is new, and had not yet all the equipment needed. Excellent results were obtained, for example, in the "Cooperativa dos Cafeicultores da Alta Morgiana", were a difference significant at the 1% level of probability was observed, with the following means of square roots of defects and standard errors of the mean: Average number of defects before processing = 7.81 ±0.21. Average number of defects after processing = 3.53 ± 0.21. This mans that the average number of defects was lowered from aproximately 61, before processing, to 12, after processing.