1153 RESEARCH S oybean [Glycine max (L.) merr.] seed yield has been increasing between 0.5 and 1% per year during the last 20 years in the United States (Specht et al., 1999). Approximately 50% of this increase is due to genetic improvement while the other 50% is attributed to better management practices (Duvick, 2005;Egli, 2008;Specht et al., 1999). Future yield increases must continue to come, at least partially, from genetic improvement.A more rapid rate of genetic gain could be achieved through increased interaction between crop physiologists and plant breeders (Cooper and Hammer, 2005;Duvick, 2005;Hammer et al., 2006). Traditional breeding of autogamous species is an empirical ABSTRACT Crop production increase needed to satisfy a growing world population depends, at least partially, on increasing current genetic gain in yield. Theory proposes that increased genetic gain can be attained using diverse high-yielding parents. Physiological traits, compared to molecular or morphological markers, are hypothesized to better estimate parental diversity. A trait-based hybridization approach will require assessing diversity for physiological traits. Here, we phenotyped several traits using two physiological frameworks to assess diversity in 25 and 65 elite cultivars from Argentina (ARG) and the United States (USA), respectively. First, we identified the highest-yielding cultivar clusters across two environments in each country. These cultivars had the highest N uptake at both ARG and USA. Therefore, there was no genotypic diversity for total N uptake within each cluster. For other traits, the highest yielding clusters did not show the highest values. There was residual diversity within ARG and USA highest yielding clusters in the temporal pattern of N uptake, N use efficiency, and N harvest index. Stacking these traits in one cultivar could potentially increase yield by 13%. The possibility of such stacking, however, depends on the nature of the phenotypic correlation among traits. We demonstrated that several trade-off correlations between phenotyped traits, thought to hinder stacking, are actually not biologically based.