Reciprocal recurrent selection (RRS) was proposed for simultaneously improving two populations and their cross. A modification of the classical full-sib RRS (FS-RRS) was proposed in which the performance of full-sibs and S(2) families is combined in a selection index (FS-S(2)-RRS). The Mediterranean corn borer (MCB) is the main corn borer species in the Mediterranean and adjacent areas and produces important yield losses. We started two RRS programs (FS-RRS and FS-S(2)-RRS) from the same maize population in which the selection criterion was grain yield under artificial infestation with MCB eggs. Original populations, two cycles of selection derived from them by both RRS methods, and population crosses were evaluated under MCB attack and under insecticide treatment in three different environments. The objective was to compare the efficiency of the FS-RRS and the FS-S(2)-RRS methods for improving grain yield. We found that the FS-S(2)-RRS method was successful for improving the yield of the population cross under optimum conditions (the regression coefficient over cycles was b = 0.87** Mg ha(-1) cycle(-1)) without losing yield under high pressure of MCB attack (b = 0.07). On the contrary, FS-RRS failed to improve the yield of the population cross under optimum conditions (b = 0.65) and tended to decrease the yield under high levels of MCB attack (b = -0.26). We conclude that for developing high yielding and stable varieties, FS-S(2)-RRS is more efficient than the classical FS-RRS method.