Little is known about the origin of Yellow Bourbon coffee (Coffea arabica cv. Bourbon Amarelo), one of the most promising coffee varieties. It probably originated in the neighbourhood of Pederneiras, S. P. (Brazil), either as a mutation of Bourbon coffee (red fruits), or as a recombinant from an eventual natural hybridization between this variety and Amarelo de Botucatu (yellow fruits when ripe), which is a natural mutation of typica (C. arabica L. var. typica Cramer). In order to develop new strains of this variety, 30 outstanding mother trees were chosen in 1945, in a private plantation located in Jaú, S. P. These trees were selected mainly on the basis of their vigor, plant type and apparent yield at selection time. To secure vegetative propagation branches were taken from each selected mother tree and then grafted in Campinas. Open-pollinated seeds from each tree were planted at five experiment stations of the State of São Paulo, namely, Campinas, Ribeirão Prêto, Pindorama, Mococa, Jaú. These locations show differences in soil type, and slight dissimilarities in climatic conditions. Campinas is located in the "Terra roxa-misturada" soil type, Ribeirão Prêto and Jaú on "Terra-roxa" soil, Mococa on "Massapé", and Pindorama on "Arenito Bauru" soil. At the beginning of the rainy season of 1946 the seedlings were transplanted to the field. In each location 20 plants from each mother tree were grown in lines, without replications, with a single individual tree per "hill". Individual records were taken for the first 8 years of crop (1949-56). During two or three years nearly all plants were sampled in order to determine the percentages of normal, "moca" (peaberry) and "concha" (elephant) seeds, grade of normal beans, incidence of empty fruit locules, and, in Campinas only, the ratio between weight of ripe fruits and the weight of dry seeds from these fruits (out-turn), and the weight of a thousand normal beans. From tables 1, 6, 9, 12, 15 and 18 which summarize the annual yields of each progeny in five locations, it appears that the most promising progenies are those derived from mother trees J 30, J 3, J 8, J 10, J 11 and J 24. High yielding progenies tend to show less variability from year to year than lower yielding-groups. Outstanding trees among the highest yielding progenies have already been included in seed production plots, and their seeds are being released to farmers for the establishment of new plantations. All progenies presented growing yields, up to the fourth crop, after which a clear biennial bearing cycle took place. Total yields per biennium increased as a rule from the first to the sixth crop, decreasing slightly in the las two years (1955-56). It was noticed that Yellow Bourbon progenies as a rule are not so consistent in keeping their initial yield ranking, from the first four to the sixth or eighth crops, as it was shown to occur in the closely related Red Bourbon variety. This indicates that if the lowest yielding progenies were discarded on the basis of th'e first two years of crop, some late high yielding progenies would have been lost for selection at a later period. Six or eight consecutive crops may already give an indication of the best progenies and of the individual trees which are probably among those to be selected for their high yields. Out of the 30-highest yielding trees in Campinas, 70 per cent belong to progenies ranked among the first ten on the basis of progeny mean yields. In Ribeirão Preto, Pindorama, Mococa and Jaú these figures were 90, 73, 60 and 80 per cent, respectively. This seems to indicate that under these circunstances selection should rather be based on the average yields of progenies. Considering altogether the five groups of progenies in the different experiment stations, an average was found of less than 2 per cent of individual trees presenting reduced total yields, that is, less than x - 2,57 s. The study of seed samples revealed that the highest percentage of "moca" seed is that from Ribeirão Prêto and the lowest one those from Campinas and Pindorama. These two locations, however, presented the highest percentages of "concha", while in Ribeirão Prêto the incidence of this abnormality was the lowest. Tables 20 and 21 summarize the characteristics of the distributions of these individual percentages of "moca" and "concha" seeds. All plants were examined for the presence of empty fruit locules. The incidence of this abnormality is low, as it is normally found in commercial varieties such as Red Bourbon and Caturra. Out of 2,930 plants belonging to 30 progenies grown in five' locations, 82 or 2.57c had fruits of a light red colour, being heterozygotes Xcxc. These plants resulted, as expected, from natural cross-pollination in the original plot where the seeds of mother trees were collected. If taken as a measure of the amount of cross-pollination this percentage may be considered as relatively high, because few plants with red fruits (XcXc) were noticed in the original plot of Yellow Bourbon. From these studies it may be concluded that due to their excellent agronomic characteristics the new strains of Yellow Bourbon coffee will have an outstanding place among the commercial varieties for the Brazilian coffee areas.