A high proportion of plants of the Mundo Novo coffee (a strain of Coffea arabica L. var. bourbon (B. Rodr.) Choussy) produce an abnormally large number of fruits with empty locules. Since this character reduces the rate of coffee beans per weight of fruit, selection was started aiming at reducing or eliminating plants having this defect, and thereby increasing the yeild of coffee beans. Preliminary investigations carried out in 1950 on the occurrence of empty locules in selfed progenies and hybrids of the bourbon plant No. 44, which presents a high percentage of this defect, indicated that this characteristic was heritable. It was also determined that only normal plants were found in progenies of normal individuals, even when these belonged to progenies in which some of the sibs produced large number of fruits with empty locules. Based on this information, seeds from normal, high yielding plants of Mundo Novo coffee were released to coffee growers in 1950. In one of the first farms to use selected seeds of Mundo Novo coffee, approximately 7,000 trees (one tree is really a group of four plants in the same hill) were planted with seeds produced by the best plants of five of the most outstanding progenies of this variety. A small crop was available in 1953. Due to the increasing demand for selected seeds of Mundo Novo coffee, it was then suggested that the State Extension Service buy the seeds produced in this farm and distribute them to coffee growers that were willing to plant them for seed production. Before this suggestion was put into effect, the whole planting was surveyed for the presence of plants producing large numbers of fruits with empty locules. By means of a sampling technique, it was determined that only from 2 to 7 per cent of abnormal plants were found in the plantation. When these numbers were contrasted with the 50 per cent index of undesirable trees normally found in the original Mundo Novo plantings, it became evident that selection to eliminate plants producing large number of fruit with empty locules had been quite successful. If must be noted that such a low percentage of undesirable plants having the same defective character is normally found in plantings of the current commercial varieties such as bourbon and others. Details of the sampling results and the influence of a high rate of empty locuels on the yield of coffee beans are discussed. The possibility of selecting exceptionally high-vielding individuals, although having this defect, is considered. The elimination of trees which gave unusually large quantity of fruit with empty locules allowed the establishment of pilot plantings. It mg/y be said that an increase in productivity was attained by this method of selection.