gaul: induced mutants in seed-propagated species 233 



From Tables 7 and 8 it is further evident that, on the average, 

 more extreme earliness mutants have lower yields. The higher 

 yielding mutants are all ]/ 2 to 4 days earlier than Haisa II. 



Another possible indicator character under investigation in our 

 laboratory is seed size. As with earliness, kernel size is certainly con- 

 trolled by polymeric genes and the great heritability of 1 ,000-kernel 

 weight is Avell known. In one of these investigations winter barley, 

 variety Breustedts Atlas, was used. The procedure of selection is out- 

 lined in Table 9. We started with nearly 1,500 normal-appearing M 2 



Table 9. — Procedure of Recurrent Selection for 1,000-kernel Weight in X-rayed 

 Progenies of Winter Barley, Variety Breustedts Atlas. 



Number of 

 plants or 1,000-kernel weight 



Year Generation progenies Grown as determined with 



investigated 



1957 M 2 1,494 Single plants 200 and 300 kernels 



(drilled bulk) 



1958 M s 135 Hills 2 X 300 kernels (from 2 field 



replications) 



1959 M 4 31 Micro-drill test 2 X 1 ,000 kernels (from 2 field 



replications) 



1960 M 5 31 Micro-drill test 2 X 1.000 kernels (from 2 field 



replications) 



plants which were taken at random from an M 2 bulk. From these, 135 

 with the highest 1,000-kernel weight were grown further. By a two- 

 fold selection we ended with 31 M 2 progeny lines which were 

 investigated more carefully in micro-drill tests for 2 years. 



The results of the first drill trial in 1959 are shown graphically 

 in Figure 3. From that histogram it is evident that only 4 out of 31 

 lines had a lower 1,000-kernel weight than the mother variety. Most 

 of the differences are only in the range of 6 per cent as compared with 

 the mother variety, but they go up to around 16 per cent. Thus, it 

 appears that the relatively simple selection procedure has been sur- 

 prisingly effective. According to an analysis of variance, differences 

 between the lines are highly significant (F value lines/error: 17.80, F 

 value of the table at 0.1 per cent level: 2.97). Also in 1960, differences 

 between the lines were highly significant which is not demonstrated 

 here (simple lattice, F value lines/intra-block error: 6.73, F value of 

 the table at 0.1 per cent level: 3.58). 



