The authors study the influence of Rio coffee in mixture with Brasilian coffee of Soft taste. Increasing percentages of Rio coffee were tried: 0,0; 0,5; 1,0; 1,5; 2,0; 2,5; 3,0; 3,5; 4,0; 4,5; 5,0; 7,5; 10,0; 12,5; 15,0; 20,0;25,0; 30,0; 35,0; 40,0; 50,0. Two experiments were carried out, in a balanced incomplete block design, with t = 21 treatments (those just mentioned), k = 3 plots per block; r = 10 replications; b = 70 blocks, L = 1. Each plot had 3 standard cups, and was tasted by 3 tasters, who gave only one opinion for each plot. Data thus collected were, therefore, 630 for each experiment (210 plots x 3 tasters). But each trial had indeed only 210 plots, to every plot corresponding the average of 3 opinions. The experiments gave results in good agreement, which led to the following conclusions: a) Transformation of the data is necessary, since variances corresponding to different treatments are rather different. b) The square root transformation gave satisfactory results. c) Rio coffee injures a lot Soft coffee taste, for percentages from 2,0% upwards. d) For percentages of 4,5% and over the mixture has taste Rioy or Rio. e) The regression obtained is not strictly linear, but the straight line gives a reasonable approximation. f) Taken in account only percentages 0,0 through 10,0%, the regression equation is: T = 1,7045 - 0,1278 X, where X is the percentage of Rio coffee, and T gives the square root of the score corresponding to tasting categories. g) For treatments 0,0 through 10,0% the regression equation for the original data is: Y = 3,0997 - 0,3281 X, that is, there is a decrease of 0,3281 in the score corresponding to coffee taste for every unity of percentage of Rio coffee.