The goals of this research were to evaluate the phenotypic and genotypic correlations between agronomic traits, to perform path analysis, having as main character grain yield, and to identify indirect selection criteria for grain yield. The experiment was carried out in an experimental area located at Capim Branco farm, which belongs to Federal University of Uberlândia, during the growing season of 2015/2016.Twenty-four soybean genotypes were evaluated under randomized complete block design with three repetitions, of which agronomic traits and grain yield were measured. There was genetic variability for all traits at 5% probability level through the F-test. Thirty significant phenotypic correlations were also observed with values oscillating from 0.42 to 0.87, which indicated a high level of association between some evaluated traits. Additionally, we verified that phenotypic and genotypic correlations were essential of the same direction, being the genotypic ones of superior magnitudes. Plants with superior vegetative cycle had longer life cycles; this fact could be explained by the significant phenotypic correlations between the number of days to the blooming and number of days to maturity (0.76). Significantly positive phenotypic and genotypic correlations for the total number of pods per plant and grain yield per plant (0.84) were observed. Through the path analysis, the trait that contributed the most over grain yield was the number of pods with three seeds as it showed the highest direct effect on grain yield per plant, as well as a strong indirect effect on the total number of pods. Therefore, the phenotypic and genotypic correlations suggested high correlations between grain yield and number of branched nodes, the number of pods with two and three seeds, and the total number of pods. Also, the path analysis determined the number of pods with three seeds as having the highest favorable effect on grain yield, and thus, being useful for indirect selection toward productive soybean genotypes.