Soybean is the second major world oil seed and the leading source of high quality vegetable protein. Two long-term selection programs were developed to improve soybean seed yield: the UP2Y population was developed from seven unimproved plant introductions (PIs) selected on their high seed protein concentration and diverse geographical origin, and the UP3Y population was developed from eight elite cultivars and experimental lines adapted to North America with diverse pedigrees selected on their high grain yield and above-average seed protein concentration. The objective of this study was to evaluate effects of recurrent selection (RS) for increased grain yield on major agronomic traits of the soybean UP2Y and UP3Y populations. Response to selection was measured by evaluating the 10 highest-yielding lines selected at the end of each cycle. The seven founders of the UP2Y and seven of the eight founders of the UP3Y populations were evaluated along with 10 selected lines of each of three (Cycle 1, Cycle 2 and Cycle 3) and two (Cycle 1 and Cycle 3) cycles of the UP2Y and UP3Y populations, respectively, in three replications of a randomized complete block design in eight environments. After three cycles of RS in the UP2Y population, grain yield increased significantly from a mean of 3070 kg ha-1 for Cycle 0 (C0) to a mean of 5037 kg ha-1 for Cycle 3 (C3) with an average significant gain of 567 kg ha-1 cycle-1. Seed protein concentration decreased significantly an average of 13 g kg-1 cycle-1. Seed oil concentration increased significantly an average of 7 g kg-1 cycle-1. Results of the UP3Y evaluation showed that after three cycles of RS grain yield increased significantly from a mean of 4200 kg ha-1 for C0 to a mean of 4499 kg ha-1 for C3, averaging a significant gain of 101 kg ha-1 cycle-1. Seed protein concentration decreased an average of 2.8 g kg-1 cycle-1 while seed oil concentration increased an average of 0.8 g kg-1 cycle-1. The UP2Y population is of particular interest because it was developed from seven high-protein PIs unsuitable for commercial agriculture in North America. We present evidence suggesting that effective selection methods can bring together the appropriate genetic makeup for elite-level performance from allegedly primitive landraces of exotic geographical origins in a relatively short time. Our evidence further supports previous findings of favorable alleles being present in untapped germplasm, and it goes a step further by presenting high-yielding material ready for incorporation into the U.S. soybean-growing region.