EARLY TESTING AND RECURRENT SELECTION 415 



variability, and therefore opportunity for selection, would be largely ex- 

 hausted after five generations of selfing. For reasons mentioned earlier, it is 

 assumed that a considerable degree of genetic variability remains in the re- 

 current series. The disparity between the two systems would therefore be 

 expected to increase with additional generations of selfing and cycles of selec- 

 tion. 



Recurrent selection has been practiced for oil percentage in two additional 

 populations. One series had its origin in an F2 population of the single cross 

 I198XHy. This population started with a much lower average oil percentage, 

 but the effectiveness of selection was essentially the same as in the 111. 

 High OilXwxOs420 series. 



In a third series a strain known as Stiff Stalk Synthetic served as parental 

 material. This material also has been divided into a selfing and a recurrent 

 series to supplement the material already presented. This experiment has not 

 yet been completed. The difference between the two series, in so far as data 

 are available, closely parallels the wxOs420Xlll. High Oil series already dis- 

 cussed. 



Data on the effectiveness of recurrent selection in modifying combining 

 ability are still quite limited. One such comparison is shown in Figure 26.5. 

 The original stock used was the Stiff Stalk Synthetic, and the double cross 

 Iowa 13 was used as the tester parent. The yields for the two years were not 

 greatly different, but to facilitate a direct comparison the lower frequency 

 distribution has been displaced to the right so that the yield of Iowa 13 for 

 the two years falls on the same ordinate. Stands were somewhat variable in 

 the test crosses comprising the first cycle. The effect of this variation was 

 minimized by adjusting all yields to an average stand by means of a covari- 

 ance analysis. This adjustment reduced the range in yields so that the con- 

 trast between the two frequency distributions does not necessarily present a 

 true picture of the relative variation in the two populations. 



RECIPROCAL RECURRENT SELECTION 



A modification of the recurrent selection scheme has been suggested by 

 Comstock et al. They have designated this procedure reciprocal recurrent 

 selection. Under this modification two diverse foundation sources, A and B, 

 are to be used. Individual selected plants in A are self-pollinated and out- 

 crossed to source 5 as a tester parent. Similarly selected plants from source 

 B are self-pollinated and outcrossed to source .1 as a tester. When test cross 

 data become available, a group of selfed ears from source A having the best 

 test cross performance are recombined to produce ^1^. AB^ population is 

 formed in a similar manner. A^ and B^ then serve as source material for a 

 new cycle of selfing and test cross evaluation followed by the intercrossing 

 of the most desirable plants. No data are yet available from either their 

 experiments or ours using this method. 



