Reciprocal recurrent selection (RRS) has proven to be a successful method to improve the performance of a cross population and to increase the heterosis between populations. However, RRS has not been widely adopted by the commercial breeders because RRS is not as efficient to recover inbred lines as other methods of inbred development. Use of two lines as testers instead of the opposite population as reciprocal tester in an RRS scheme could overcome this limitation;A breeding research program was initiated in 1974 at Iowa State University to evaluate the modified RRS (MRRS) procedure and RRS in BS21 and BS22 maize populations. The modification used inbred line A632 as tester for BS21 and inbred line H99 as tester for BS22;After six cycles of selection were completed in BS21 and BS22 using MRRS and RRS, an experiment was conducted to evaluate the response to selection. The populations per se, testcrosses to inbred testers, and crosses between cycle populations of RRS and MRRS were evaluated in replicated yield trials;There were significant increases in grain yield in all six cross populations as a consequence of selection. The rate of direct response was greater for the RRS procedure than for the MRRS (4.4, 1.6, and 2.8%cycle-1 for BS21(R)xBS22(R), H99xBS22(HI), and A632xBS21(HI), respectively). RRS was as effective as MRRS to improve the grain yield of the populations in crosses with the inbred lines, but MRRS was not as effective as RRS to improve the cross population BS21xBS22, with a significantly lower rate of increase in yield of 1.6%cycle-1. Realized heritability and response to selection for yield were 25 to 50% of their predicted values;In this study there was no evidence that the genetic variance among testcrosses for yield was greater when using inbred lines as testers than when using populations as tester. The traditional RRS procedure was more effective than MRRS to improve grain yield in the cross population BS21xBS22.