RECURRENT SELECTION AND OVERDOMINANCE 453 



same sources, abandoning recurrent selection entirely. Usually, successful 

 but mysterious processes are not modified on theoretical grounds alone. How- 

 ever, most of us, and myself most of all I suspect, chose the alternative course 

 without question. New lines for a second cycle of selection were isolated from 

 crosses of elite first-cycle lines. Since it was soon apparent that second-cycle 

 lines as a group were a vast improvement over first-cycle lines, it was clear 

 that we were on the right track. Recurrent selection for higher frequencies of 

 dominant favorable genes was fulfilling expectation admirably. That it had 

 failed in ear-to-row selection (progeny testing without inbreeding) meant 

 that "selection within and among inbred lines" was the key. Apparently the 

 protagonists of "early testing" have not fully appreciated this latter i:)oint. 



DISAPPOINTMENT WITH SECOND-CYCLE HYBRIDS 



My first susj)icion that all was not well was aroused by disappointing 

 yield performance of second-cycle hybrids in 1941. The first reaction then 

 was to conclude that heterosis might involve complex gene interactions to a 

 greater extent than I had supposed. Cytoplasmic-nuclear interactions could 

 not be ruled out entirely. But no thought of heterozygosis, of overdominance, 

 was entertained at all, so thoroughly had I been weaned from it. 



In 1942 w^e began the process of separating Florida inbred lines into two 

 permanently distinct groups on the basis of combining values with two single 

 cross testers which were thought to make a good double cross. Subsequent 

 breeding operations after the initial separation were to consist of isolating 

 new lines within each group from crosses of the older lines within the group. 

 New lines were to be stabilized by at least three self-pollinations with ac- 

 companying selection for vigor and type, and then tested for combinability 

 with the reciprocal group. This, of course, was reciprocal recurrent selection 

 without early testing. I still adhered firmly to the efficacy of "selection within 

 and among inbred lines." 



Segregation of the breeding mass into two permanently distinct reciprocal 

 groups, first of all, did not cost anything. A search for satisfactory substitutes 

 for each of the four master tester lines was well in order. It seemed that the 

 necessity of recovering specific combinability again as the last step of each 

 breeding cycle might be avoided to some extent. Possibly higher levels of 

 specific combinability might be accumulated. 



Two years later, after interviews with a number of other corn breeders, 

 it seemed that a still higher rating might be in order for specific combinabil- 

 ity. Second and third-cycle hybrids were not much superior to first-cycle 

 hybrids in yield of grain. Recurrent selection for general combinability was 

 not proving to be very effective. 



An early test of recurrent selection for specific combinability seemed de- 

 sirable. One way to intensify the process already in operation was to adopt a 

 more specific tester. This was done by abandoning the reciprocal feature of 



