Reciprocal recurrent selection in maize (:Zea mays L.) has been ve1y successful in improving yields of crosses of two populations wherein each population has been used as a tester for the other. A modification of the procedure was suggested that uses two inbred lines as testers instead of the populations. The purpose of our research was to compare the two procedures for progress achieved after three cycles of recurrent selection. The populations were BS21 and BS22 Synthetics used in a reciprocal recurrent selection program (BS21R x BS22R). For the modified procedure, inbred A632 was the tester for BS21, [BS21(A632Hl); and inbred H99, for BS22, [BS22(H99HI)]. Progress was evaluated in six crosses: BS2l(R) x BS22(R), BS21(A632HI) x A632, BS22(H99Hl) x H99, BS21(R) x A632, BS22(R) x H99, and BS21(A632H[) x BS22(H99HI). Grain yield showed highly significant P < 0.01) linear gains for BS2l(R) x BS22(R) (4.9% cycle), BS21 (A632H]) x A632 (3.6 cycle), and BS21(R) x A632 4.7%/cycle); gains for the other crosses were positive, hut not significant. Evidently, after three cycles of selection, the modified procedure was not successful in improving yield for crosses of the populations. The results indicated that choice of inbred testers may he very critical in the modified procedure. Grain moisture had highly significant linear decreases for BS21(A632HI) x A632, BS22(R) x H99, and BS2l(A632HI) x BS22(H99Hl) Highly significant changes were observed for lodging as follows: root lodging decreased for BS21(A632H[) x A632 and BS21(R) x A632; stalk lodging decreased for BS21{R) x BS22(R) and BS21(R) x A632, but increased for BS21(A632HI) x BS22(H99Hl).