Genotype x environment interactions were estimated in a 3-year study in South Africa in which cotton cultivars were evaluated in 54 trials under two systems: (1) at a single fixed location and (2) at several different farm sites in the Loskop production area. For six parameters measured, first and secondorder interactions were of similar magnitude for both systems, and relatively small, compared to main effects. Results indicate that the two evaluation systems are equally effective establishing differences in performance potential.