402 G. F. SPRAGUE 



series can be readily accounted for by the assumption of epistasis, though no 

 claim is made that this is the only or even the correct explanation. Where the 

 correlations involve some character of the inbred parent and the same char- 

 acter in their Fi crosses, epistatic effects would be expected to be at a maxi- 

 mum. When a character of the parent is correlated with the mean of all cross- 

 bred progeny opportunity would be provided for a considerable degree of can- 

 cellation of the epistatic effects. 



The results reported by Hayes and Johnson are more directly comparable 

 with Jenkins' group 2. Various characters of the inbred parent were corre- 

 lated with the yield of their topcross progeny. The correlations for individual 

 characters ranged from .19 to .54, and the multiple r for 12 characters of the 

 inbred parent and yield of the topcross progeny was .67. 



As a result of these studies some investigators have decided that the cor- 

 relations were too low to provide a wholly satisfactory basis for prediction, 

 and the only safe measure of the worth of an inbred line was to evaluate it in 

 hybrid combinations. 



EARLY TESTING 



Since the characteristics of the inbred lines did not provide an adequate 

 index as to the value of a line, and since this value must be determined by 

 crossbred progeny tests, it seemed advisable to determine whether crossbred 

 performance could be evaluated at an earlier stage of inbreeding. Several 

 lines of reasoning suggested that this might be feasible and desirable. First 

 the ear-to-row tests with all of their limitations suggested that there were 

 marked differences in yielding ability between individual carefully selected 

 open-pollinated ears. The genotype of such high yielding ears was modified 

 or diluted in ear-to-row testing procedure, but the identity of these individual 

 ears could readily be maintained by self-pollination. Second, it appeared logi- 

 cal to assume that a potential ceiling was established for any derived line at 

 the time of the selfing of the So or F2 parent plant. This ceiling is established 

 by the genotype of the parent plant and the most desirable combination of 

 genes which can be isolated from this gene sample. 



The small population commonly grown from each selfed ear, the hindrance 

 of linkage in preventing random recombination of genes, and the limited ef- 

 ficiency of visual selection would all operate to render the probability of 

 isolating this most desirable gene combination very unlikely. The effort ex- 

 pended in growing and continued inbreeding and selection of strains having 

 the less desirable genotype might represent a considerable waste. Third, if 

 facilities were limited, as they always are, greater progress might be achieved 

 by the early discarding of the less desirable genotypes and the growing of 

 larger progenies of the more desirable genotypes in the early generations of 

 selfing when variability and the efficiency of visual selection would be ex- 

 pected to be at a maximum. 



Before these ideas could be put to a test, data were presented by Jenkins 



