HYBRID VIGOR IN DROSOPHILA 491 



General and specific combining abilities are strictly applicable to the par- 

 ticular experiment from which they are calculated. The values for the dif- 

 ferent lines vary with the group of lines from which they are calculated. This 

 is a serious defect, for the results have no significance unless they may be used 

 for the prediction of the future performance of the particular line or line com- 

 bination. As the interest is in the operational use of these parameters in guid- 

 ing breeding work, the theoretical objections to comparing successive values 

 for general or specific combining abilities are outranked by the practical con- 

 sideration. This study measures the repeatability of the estimates of general 

 and specific combining ability for particular inbred lines when the crosses are 

 grown in different locations, different years, and in different combinations. 



Sixty-two inbred lines were the parents of the Fi crosses. The data include 

 451 determinations of general and 2033 estimates of specific combining abil- 

 ity. As pointed out above, these determinations are not of equal weight be- 

 cause of differences in numbers and lines in the different Fi hybrid tests. 

 However, for the purposes of this comparison they are regarded as of equal 

 weight, since it is on this basis that the data will be used for guiding future 

 breeding operations. 



The intraclass correlation between the repeated tests of the general com- 

 bining abilities was 0.29; that for the specific combining abilities of the re- 

 peated crosses of the pairs of lines was 0.27. These correlations are definitely 

 higher than those observed for the Drosophila data. They are high enough to 

 be of reasonable importance in practice. The data for the general combining 

 abilities become of somewhat greater value when the determinations are re- 

 stricted to particular regions of the state, the over-all correlation becoming 

 0.31. When arranged within years but allowing free range over the 10 differ- 

 ent geographical regions of the state, the over-all correlation becomes 0.53. 

 The specific combining abilities do not show an equal improvement in predic- 

 tive values when subdivided by these categories. Specific combining abilities 

 drop when the data are subdivided by geographical regions of the state, the 

 over-all correlation becoming 0.18. When the subdivision is made by years, 

 the over-all correlation becomes 0.34. 



These results reemphasize the effects of the environmental-genotypic in- 

 teractions on performance as discussed earlier. The corn hybrids are fitted to 

 the geographical regions of the state by selection of the place of planting for 

 season of maturity. Little or no selection is possible for fitting the plantings 

 to the vagaries of the different years. The effects are noted in the intraclass 

 correlations. Double selection for genotypic environmental correlation when 

 the data are subdivided by years leads to definitely increased correlations for 

 the general combining abilities of the particular lines and to slightly increased 

 correlations for the specific combining abilities of these same lines. Where the 

 years X genotypic effects are allowed to express themselves, the correlations 

 are no greater than those of the whole or are reduced. 



