An alternative method of reciprocal recurrent selection (RRS) in which populations A and B are each evaluated in a different environment is proposed. This method is called dual-environment reciprocal recurrent selection (DERRS). Two genetic models are considered in the theoretical study. A comparison of selection methods shows that genetic gain is larger in DERRS than in RRS for the two models. The difference grows greater as the dominance effects operating in the two environments are more divergent and as the number of selection cycles increases. A greater gain is obtained when the genetic covariances between crosses in the two chosen environments are lower.