Empty fruit locules in Coffea spp., have been reported several times. Although an explanation for the occurrence of this phenomenon in the C. arabica species has apparently never been advanced, it seems that empty fruit locules result from an arrest in the endosperm development, which probably occurs at the time when it begins to fill out the already developed ovary locule. When this process occurs, development of the fruit locule proceeds normally, but a normal seed is not formed. Usually a residue of what seems to be perisperm tissue is present in the locule of the fully developed fruit, accompanied or not, by a small amount of endosperm which may occasionally contain a small embryo. Since 1950 samples of 100 ripened fruits from individual plants of several progenies of Mundo Novo coffee (Cojfea arabica L. var. bourbon (B. Rodr.) Choussy), growing in experiment stations located at Campinas, Pindorama, Mococa and Jau, have been examined for the presence of empty fruit locules. In some of the progenies examined from open pollinated flowers most plants produced only very low quantities of fruits with empty locules, regardless of the location where they were grown. Other progenies presented plants which produced fruits with small amounts of empty fruit locules and plants in which this amount was higher. The ratio between these types of plants was rather constant for the same progenies at different locations. The number of empty fruit locules for a given plant remained nearly the same from year to year and did not show any apparent correlation with yield, vigor, or plant type. Among the 1096 plants studied in four locations, it was observed that 456 plants (42%) presented high quantities of empty fruit locules. If this sample is considered representative of the common Mundo Novo coffee plantations, which were established without selection to eliminate this defect, it may be assumed that 39 to 45% of such plants will present high quantity of empty fruit locules. It has been noted that, as a rule, fruits of Mundo Novo coffee with at least one empty locule will float when put in water. Approximately the same results are obtained by counting the number of floating fruits in a sample of 100 as when the 100 fruits are cross-sectioned and examined for the presence of empty locules. The method of counting the number of fruits that float in water is accurate enough in field work, when quick results are needed such as when mother trees free of this defect are to be selected. The observation already made indicate that the tendency to produce fruits with empty locules is a genetical character. Although the inheritance of this trait is not yet fully understood, it must be taken into consideration in the improvement of coffee varieties.