HETEROSIS IN A NEW POPULATION 445 



which may have some bearing on the relative value of Si lines versus more 

 homozygous lines in the formation of hybrids. 



Preliminary data on the relative combining ability between Si and the S3 

 lines selected from each Si are available from topcrosses to the same tester. 

 Each topcross with an Si line was tested for yield in the same experiment 

 with the corresponding topcrosses involving the lines obtained from that Si 

 after three generations of selling and selection for desirable agronomic char- 

 acters. The number of S3 lines in each Si family varied from one to sixteen, 

 some families having a larger number of desirable S3 lines with respect to 

 agronomic characters than others. 



A frequency distribution of the differences in topcross yields in per cent 

 between Si line topcrosses and the average of the S3 line topcrosses within 

 each family is given in Table 27.13. The differences are expressed as Si minus 



TABLE 27.13 



DISTRIBUTION OF DIFFERENCES IN TOPCROSS YIELDS BE- 

 TWEEN Si LINES AND THE AVERAGE OF THE S3 LINES 

 WITHIN EACH FAMILY (Si - AVERAGE OF Sa's) 



Number of observations = 138 Mean = +0.90 



the average of the S3 within the respective Si family. The class mid-points, 

 therefore, range from to 65 per cent positive and from to 75 per cent 

 negative, with class intervals of 5 per cent. A positive difference means that 

 the Si topcross yield exceeded the average of the S3 topcrosses. A negative 

 difference indicates that the average of the S3 topcrosses was higher than 

 the Si topcross within the same family. It is evident from Table 27.13 that 

 the distribution of the differences approaches very closely that of a normal 

 curve. That is, there were as many cases in which the Si exceeded the aver- 

 age of the S3 as there were cases in which the average of the S3 exceeded the 

 Si. The mean difference between the 138 pairs was +0.90 per cent. These 

 data indicate that visual selection in advanced selfed generation progeny 

 based on agronomic characters is largely at random with respect to com- 

 bining ability. If visual selection in successive generations of inbreeding had 

 been effective in increasing combining ability, then the above curve would 

 have been skewed in the direction of the negative differences. 



Upon further inbreeding of Si lines at random without selection for com- 



