Reciprocal recurrent selection (RRS) has not been widely adopted by maize (Zea mays L.) breeders because pedigree selection methods were effective in developing improved inbred lines. RRS as been used successfully, but modified reciprocal recurrent selection (MRRS) uses elite inbred lines as testers, which may be more useful in applied breeding programs. A study was conducted to compare responses to selection via RRS and MRRS in the BS21 and BS22 maize populations. After six cycles of RRS and MRRS, an experiment was conducted to determine the direct and the indirect responses to selection. The populations themselves, testcrosses to inbred testers, and crosses between BS21 and BS22 were evaluated at four locations for 2 yr. Grain yield increased significantly in all population crosses. Direct response to selection was greater for the RRS method than for the MRRS method: 4.4% cycle−1 for BS21(R) × BS22(R); 2.8% cycle−1 for BS21(HI) × A632; and 1.6% cycle for BS22(HI) × H99. RRS was as effective as MRRS for improving grain yield of BS21(R) and BS22(R) in crosses with A632 and H99, but MRRS was not as effective as RRS in the improvement of the BS21(HI) × BS22(HI) cross populations: 1.6%−1 cycle for BS21(HI) × BS22(HI) vs. 4.4% cycle−1 for BS21(R) × BS22(R).There was no evidence that the genetic variation among testcrosses for grain yield was greater with use of inbred lines as testers compared with use of populations as testers.