RECURRENT SELECTION AND OVERDOMINANCE 457 



reserves with nominal selection for agronomic type. If either tester line should 

 develop a serious fault, or if the present main selection for specific combinabil- 

 ity should seem to reach a ceiling, reversal of selection would seem almost 

 inevitable. A tester would be chosen from the current crossbred and the two 

 bulk backcrosses would furnish a reciprocal crossbred to reverse the process, 

 temporarily at least. 



An accessory operation with bulk backcrosses is hardly practicable with 

 livestock. But here the tester would be one inbred line which would need 

 to be 50 per cent inbred for equal efficiency with the single cross of homozy- 

 gous lines employed as the corn tester. The tester should be the male parent 

 of the improved hybrid in livestock to avoid any impairment of the female 

 function by inbreeding. 



Beginning with a partly inbred or non-inbred stud flock or herd as the 

 tester, and continuing mild inbreeding, it is inevitable that choices among 

 young males for herd sires of the stud herd would depend partly on their 

 testcross progeny. Sufficient vigor must be retained in this herd to provide 

 satisfactory sires of commercial hybrids. The problem is real and obvious 

 enough, but I have thought the details must await a demonstration that 

 specific combinability can be accumulated in important amounts by recur- 

 rent selection. For an early test the more homozygous tester is probably to be 

 preferred. If uniformity of the product is of some moment, the operator of 

 reciprocal selection may expend considerable effort for it. Such expenditure 

 might be avoided by partial inbreeding of one of the groups. 



The two breeding plans, selection in a crossbred to a homozygous tester 

 and reciprocal selection between two crossbreds, are the extremes of recur- 

 rent selection for specific combinability. Between these we may have any 

 practicable degree of inbreeding of one of the groups at the start, or subse- 

 quently. Inbreeding restricts reciprocal selection but, aside from that, the 

 reciprocal feature may be varied at will. I do not know what factors may 

 determine the more efficient plans except that general combinability with 

 respect to vigor is probably not an important one. Nor is it likely to be im- 

 portant to choose an inbred tester with above-average general combinability. 



PHYSIOLOGICAL NATURE OF OVERDOMINANCE 

 Overdominance has been defined (Hull, 1946a) as aA > AA, which is a 

 sufficient definition for present purposes. However, there may be some value 

 in considering what the underlying physiology may be. Heterozygosis as con- 

 sidered by Shull and his early contemporaries is entirely or very nearly the 

 same concept. Fisher (1918, 1932) has discussed this concept more gen- 

 erally as super-dominance. Some recent writers have employed heterotic al- 

 leles or heterotic interaction of alleles as a modern form of heterozygosis. But 

 since any degree of dominance of the more favored allele is essentially a 

 heterotic interaction, heterotic alleles does not necessarily imply a A > AA. 



