In order to breed efficiently, it is necessary to identify individual quantitative trait loci (QTLs) as well as interactions between these loci and to determine which QTLs produce phenotypes that are environment specific. This can be done by linking QTLs to molecular markers. The objective of this research was to carry out such an analysis for yield, one of the most complex agronomic traits. To do this, recombinant inbred lines of soybean [Glycine max (L.) Merrill] were characterized for molecular genetic markers and analyzed for yield in different environments. Interactions between QTLs were identified by subdividing the segregants into four sub-populations defined by molecular alleles at pairs of unlinked loci. Differences in the mean yields of these sub-populations defined interactions between QTLs. Measurements of yield in genotyped, recombinant inbred populations derived from crosses of 'Minsoy' with 'Archer' (MA population) and 'Noir 1' with Archer (NA population) have identified a pair of interacting yield QTLs whose effect was independent of environment as well as a pair of loci whose interaction was environment specific. Each example of epistasis, involved an allele specific interaction between the two QTLs. In the NA population, a pair of QTLs was identified in which Noir 1 alleles interact to specify a significant increase in yield that is not environment specific. These loci, located on linkage groups (LG) U3 and U9, do not affect either height or maturity. In all environments, the interaction between the QTLs was significant. In the MA population, a pair of QTLs was identified in which the Minsoy alleles interact to specify a significant increase in yield. However, this significant interaction is environment specific. One of the loci (on LG U14) is also associated with effects on height, seed weight, and maturity that are found in other environments, but these latter effects do not appear to involve any interactions with other loci. The data from the MA population support the concept that interactions between QTLs also can result in location-specific effects on quantitative traits.