This research aimed to verify the performance of soybean (Glycine max (L.) Merrill) progenies over three environments, to evaluate the genotype x environment (G x E) interaction, and to estimate the correlation among characters. The estimates of these parameters can to facilitate the future selection steps, and the release of new food type, large-seeded genotypes for three kinds of use: vegetable soybean (edamame, imature, large-poded, without limitations on seed-coat and hilum-color); sweetmeat soybean (kuromame, mature seeds with black coat and hilum); salad soybean (mature seeds with light coat and hilum). The three environments were: a) Piracicaba-summer season (PV), sowing on November 15, 1996; b) Piracicaba-autumn season (PU), sowing on March 07, 1997; c) Anhembi-summer season (AV), sowing on November 17, 1996. The localities Piracicaba and Anhembi have as main contrast the physical-chemical properties of soils: Piracicaba has alfisol type; Anhembi presents a sandy soil, commonly found in the Brazilian savannas (cerrados) and with the acidity neutralized by using lime. There were evaluated 24 topcrossed progenies in F 9 : 4 , derived from 15 different topcrosses; the following two topcrosses has contributed with more than one progeny: Late Giant x Doko and Aliança Preta x Doko. The three experiments were designed in randomized complete blocks, each one with two replications stratified in experimental sets with common checks. The replications were stratified in two sets, each one having 12 progenies and four checks (food type, large-seeded soybeans cultivars). The experimental plot was represented by ten hills of individual plant, spaced l.OOm x O.SOm. Between plots the space was l.OOm. The following characters were evaluated: number of days to maturity (NDM), plant height at maturity (APM), lodging (AC), agronomic value (V A), visual width pod (L VV), one hundred seed weight (PCS), and seed yield (PG). The results permitted the following conclusions: a) Various of the superior progenies in F 9 : 4 had already exceeded in the previous generations (F 2 , F3:2 or F 5 : 4). b) It was obtained early progenies having less than 110 days of cycle, despite they were originated from topcrosses with the late parental 'Doko'. c) In the selection for increase plant height at maturity (APM) it is obligatory to consider the lodging (AC) together. d) The favorable environment (PV) permitted the best differentiation among progenies for the characters APM, L VV, and PG. The stressing environment (PU) permitted the differentiation among progenies for ali characters, except AC. The Anhembi-summer environment was relatively intermediate in discriminating the progenies. e) The correlation coefficients among characters were much altered by the environmental variations. f) The "Box-plot" evidenced significative environmental effects on the characters: V A, NDM, PG and APM. However the seed size, evaluated by the one hundred seed weight (PCS), was not much influenced by the environmental variations. g) The ecovalence method was mor...