456 FRED H. HULL 



gous lines. Or we may consider the more efBcient technic of recurrent testing 

 of controlled testcrosses of So plants with the parent variety and recombin- 

 ing the better ones into an improved variety. We know this will not work, al- 

 though it has not been fairly tried. Finally, in modern corn breeding the same 

 technic with S4 and higher lines has been extensively advanced through at 

 least two cycles. Most corn breeders will admit that a general recombination 

 or synthetic blend of parent lines of present elite-yield hybrids would hardly 

 yield more than a random blend of parent varieties of today or of 50 years 

 ago. 



A few recombinations of lines selected wholly for general combinability 

 have been reported with significantly higher yields than improved varieties. 

 This result I will attempt to show later is a different matter, fully consistent 

 with overdominance theory. 



It seems likely that improvement of general combinability, accumulation 

 of dominant favorable genes with respect to grain yield, in the field corn of 

 our central Corn Belt in the past fifty years has been hardly significant except 

 for that depending on disease resistance, resistance to lodging, to ear drop- 

 ping, etc. Almost any one of the common breeding technics is quite efifective 

 with general improvement of morphology of the corn plant, or with oil and 

 protein of the grain. Genetics of vigor would appear to differ in some impor- 

 tant respect from genetics of the other traits. 



Overdominance has seemed the more likely, but I have never meant to in- 

 sist that the existence of every other alternative had been disproven. Refrac- 

 tory repulsion linkage has seemed insufficient alone to explain the apparent 

 degree of overdominance in corn (Hull, 1945a). 



The main point now is accumulation of general combinability with recur- 

 rent selection. It is axiomatic with most of us, including the corn breeders, 

 that general combinability is the first consideration, despite the evidence 

 cited here. This kind of evidence has been largely ignored and almost taboo. 



Comstock et al. (1949) have proposed Reciprocal Recurrent Selection to 

 obtain maximum utilization of general and specific combinability together. 

 In this they have accepted that specific combinability might be accumulated 

 in successive cycles, and that the inbreeding interphase could be abandoned 

 entirely. This variation of the general plan was compared on theoretical con- 

 siderations with selection in a crossbred for combinability with a homozygous 

 tester. Now, since a homozygous tester is clearly impracticable in many cases 

 and heterozygosity would impair efficiency of a tester except for reciprocal 

 selection, there is an advantage in the reciprocal plan which the authors did 

 not record. 



It has never been my intent, however, to attempt to rule out judicious 

 reciprocal selection. We have crossed each of the two tester lines of corn to a 

 goodly number of unrelated strains, and have backcrossed in bulk to each 

 tester line separately. The two lots are being held in separate crossbreeding 



