It was shown in a recent paper that the occurrence of empty locules in coffee fruits of the variety Mundo Novo (Coffea arabica L.) is due to an arrest in the endosperm development that takes place at a definite stage, leading to the formation of a small disc ; this rudimentary type of endosperm has been called "discoid endosperm". A suggestion was then made that this process might be controlled by genetic factors. Under this assumption, normal plants were DD; plants in which the defect occurred, Dd, and the homozygous, recessive form was lethal. The high correlation between discoid endosperm and empty locules permitted a clear-cut separation of Mundo Novo coffee populations into two classes : a) plants in which the discoid endosperm is not formed (normal plants) ; and b) plants in which the disc occurs (hight percentage of empty locules). A survey was made of three Mundo Novo plantings representing different generations of an once-existing small group of plants. Under the assumption that the original plants (maximum of eight plants) were Dd (F1), the first plantings made with their seeds represented F2 populations and should have DD and Dd individuals in a ratio of 1:2. A second planting has been made with seed collected from 400 plants that represented another group of the supposed F2 generation; assuming that this F2 had 1DD:2Dd, and that 10% of cross pollination occurred, this F3 population should consist of DD and Dd individuals in the ratio of 59:41. A third group of plants that was examined had been planted with seed collected from 15 plants selected from the first planting of the F2 generation. Sampling of these 15 parent plants indicated that according to the presence of discoid endosperm they were 4DD:11Dd; assuming that 10% of cross pollination occurred and taking into account that an equal number (20) of seedlings was selected from each progeny, it was expected that the ratio in this F3 lot would be 51:49. Samples of 100,100, and 262 plants that were examined in the one lot that represented an F2 generation and in the two others that represented F3 generations, respectively, confirmed the above-mentioned expectations. Individual study of the plants belonging to the 15 progenies showed that the 197 plants derived from the 11 Dd plants were in the ratio of 1DD:2Dd; among the 65 plants derived from the 4 DD individuals, only four were Dd, which corresponds closely to a possible 10% of cross pollination. Emasculated flowers of Dd plants that were pollinated with DD pollen produced only normal seeds, thus confirming the genetic origin of empty locules. The frequency of empty locules only seldom reaches the expected 25%. To explain this slight discrepancy, the suggestion is made that some ddd endosperm may attain normal size and shape, although their embryos may not be viable since no dd plants has ever been found. Data on germination obtained in 1950 and 1951 have indicated that apparently normal seed from plants that had a high percentage of empty locules (probably Dd) gave a lower germination than those from plants with a low percentage of empty locules (probably DD).