410 G. F. SPRAGUE 



then the variation associated with generations is correctly judged significant. 

 If, however, the experimental material is assumed to represent a random 

 sample of lines and therefore typical of lines in general, the appropriate test 

 indicates generations to be non-significant. Since no yield data were presented, 

 no test of significance can be calculated for the linear component of genera- 

 tions. Their results, as presented, have little bearing on either early testing 

 or the effectiveness of visual selection in modifying combining abiUty during 

 the course of inbreeding. 



Richey (1945, 1947) has presented a re-analysis of Jenkins' (1935) data on 

 combining ability after successive generations of inbreeding and reached 

 conclusions differing from those presented by Jenkins. He questions the 

 stability of combining ability and the effectiveness of early testing in provid- 

 ing a satisfactory criterion of combining ability when the lines approach 

 homozygosity. He also presents some information on tester parents and their 

 effectiveness in revealing segregation. This latter is a very important prob- 

 lem but will not be discussed here. 



We return to the first criticism raised by Richey, namely that lines do not 

 reach stability early in the course of inbreeding. To demonstrate his ideas, 

 Richey has combined the eight generations into pairs, thus providing four 

 groups. Then by selecting certain inbreds he has shown by graphs that, 

 visually, quite different slopes are obtained over the period under study. 

 Other groupings than those used by Richey may be selected with equal 

 validity. These different groupings show quite an array of slopes upon visual 

 inspection. However if one extends the original analysis of variance pre- 

 sented by Jenkins separating generations into a linear and remainder com- 

 ponent, the linear component is not significant. This, of course, does not 

 prove that trends are absent. It does indicate that such trends as may exist 

 are small in comparison with the random variation. 



Richey 's second criticism deals with the effectiveness of early testing as a 

 measure of combining ability as the lines approach homozygosity. He con- 

 cludes that early testing would have been quite ineffective. The real basis 

 for the evaluation of any breeding or testing system depends upon the lines 

 which are produced or revealed which have sufficient value for use in com- 

 mercial hybrids. Of the twenty-seven lines on which Jenkins presented data, 

 two lines of the Lancaster series have been of sufficient value to be used ex- 

 tensively. These are L289 and L317. These two ranked in the upper half of 

 the lines tested and would have saved under an early testing procedure. 



Two other lines have been used to a limited extent. One of these, 1224, 

 exhibited the highest yields in the lodent Si test cross series and would cer- 

 tainly have been saved. The other line L304A was in the upper 50 per cent of 

 the Si Lancaster series. If early testing had been used with this material, 

 saving the upper 50 per cent of each frequency distribution, no commercially 

 useful lines would have been discarded. The early discarding of the remain- 



