Reciprocal recurrent selection (RRS) and a modified reciprocal recurrent selection procedure (RRSI), conducted in BS21 and BS22 maize (Zea mays L.) synthetics, were evaluated for changes in agronomic traits after three cycles of selection. BS21(R)CN and BS22(R)CN were used reciprocally as testers for each other in RRS. In RRSI, A632, an inbred line related to BS22, was used to test S[subscript]0 plants of BS21(A632HI)CN, while the tester for S[subscript]0 plants of BS22(H99HI)CN was H99, an inbred line related to BS21;BS21(R)CN x BS22(R)CN had highly significant (P < 0.01) improvements for grain yield (0.31 Mg ha[superscript]-1 cycle[superscript]-1), root lodging (-2.91% cycle[superscript]-1), and stalk lodging (-2.61% cycle[superscript]-1) from RRS. Yield improved in BS21(A632HI)CN x A632 after RRSI at a highly significant rate of 0.18 Mg ha[superscript]-1 cycle[superscript]-1, but the procedure was ineffective for improving yield in BS22(H99HI)CN x H99 or in the interpopulation cross of BS21(A632HI)CN x BS22(H99HI)CN. Yield improvement was highly significant for all testcrosses when averaged over all inbred and noninbred testers, and the trends were similar for both procedures (RRS and RRSI). However, the significance of the rate of yield response with specific testers varied in all testcrosses. RRSI resulted in no significant changes for root and stalk lodging in BS21(A632HI)CN x A632 and BS22(H99HI)CN x H99. Neither procedure was effective for improving yield of the populations per se. In general, all populations per se and population crosses had reductions for days to flowering, plant height, and ear height after three selection cycles;RRS was more effective than RRSI for improving the interpopulation crosses between BS21 and BS22. Masking effects of the elite inbred testers were evident for most traits. However, both procedures were similar in their effectiveness for improving yield in the crosses between the cycle populations and the testers, when averaged over all testers. If these changes were associated with selection, rather than random fluctuations, then RRSI may show significant progress in later cycles. Additional cycles of selection, currently in progress, may establish more definite trends.