472 MUTATION AND PLANT BREEDING 



the three crosses in Figure 1. Here the data are presented as per- 

 centage of the mean of all treatments in each cross, i.e., as genotypic 

 coefficients of variability. The center line labeled "cross mean" is 

 equal to 100 per cent of all treatments in the cross. The bases of 

 the bar graphs, shown slightly extended, represent the means of 

 the Fi generation families. The s G arising from differences among 

 progenies in each family is indicated by the height of the bar for 

 each treatment shown at the bottom of the chart. Since interest 

 is primarily in the variation in excess of the means, only the s G 

 above the mean is presented. The deviations of the family means 

 from the cross mean are shown on the same scale as the genotypic 

 standard deviations from the family means. 



The mean family yield of PX was less than that of any P in 18 

 out of the 30 comparisons made in these two treatments. Of the 12 

 remaining comparisons s G was nonsignificant in five PX families. 

 This is to say that in the P vs. PX comparisons 7 PX families out of 

 30 had equal or higher means in addition to higher variances than 

 any of their P standards. 



In the case of the F 2 vs. PX, 17 out of 30 family means in the 

 F 2 were higher than any PX family mean. In eight of these 17, s G 

 was small in magnitude, failing to attain significance. This is to 

 say that in 9 out of 30 families of F 2 the Fi family means were higher 

 than any PX and, in addition, possessed a significant variance. In 

 reverse, 13 of the 30 PX were lower than any F 2 and none higher. 



In the case of the F 2 vs. F 2 X 2 , 9 of 30 family means of the F 2 X 2 

 were lower than any family mean of F 2 . In reverse, seven F 2 were 

 higher than any F 2 X 2 , while only two F 2 X 2 were higher than any F 2 . 

 (This could happen since the comparisons were made cross by 

 cross.) 



It is obvious from the data that if outcrossing Avere too costly 

 either in time or loss of collateral characters, progress from selec- 

 tion in PX could be expected. If the character here measured were 

 the only one under consideration, it is likewise obvious that in two 

 of the three crosses greater progress from selection would be expected 

 in the F 2 , while in one of the three greater progress would be 

 expected in the F 2 X 2 . When this prognosis was tested (two loca- 

 tions, six replications, 100 plants per plot) with bulked F 2 fam- 

 ily seed in the F 4 generation, the results confirmed the prediction 



