Yield data of 21 soybean variety from trials conducted in 28 environments in the Rio Grande do Sul State were used for a stability analyses using the discontinuous bi-segmented model (STORCK, 1989) with corrections due to errors in the variables. The main conclusions were: 1) The method was adequate for phenotype stability studies of soybean varieties since there was a wide environmental variation and a good data adjustment to the model; 2) The group of varieties can be considered as undesirables (Hill, IAS-2, Paraná, Prata, Sel. Foscarin, Louisiana, Hardee, Bienville, Industrial, Santa Rosa, Bossier and Viçoja); b) Adaptable to both unfavorable and favorable environments, within the region studied (Pérola, Bragg and Jakson); c) Adaptable to favorable environments (Planalto, Hale-7, IAS-5, IAS-1 and IAS-4); d) Adaptable to average environments (Hood).
soybean; stability analysis; cultivar-environment interaction