16 BULLETIN 1354, U. S. DEPARTMENT OF AGRICULTURE 
varietal cross, WhatleyxSt. Charles White, therefore seems en- 
tirely justified. 
Whether particular crosses can be relied upon to give larger yields 
year after year remains to be determined. The higher yields of 
crosses involving 10-3- in 1921 and 1923, the consistency with which 
crosses involving 7-4-1-6-6-2 were superior, and all of the other 
evidences of individuality among the self-fertilized lines and of 
crosses between them indicate that the large yields of certain crosses 
were not due to chance. The lines had been self-fertilized six gener- 
ations and were reasonably uniform when the crosses compared in 
1923 were made. The experiments of Jones (3) suggest that these 
lines will remain fairly constant from now on, except as germinal 
changes of one kind or another may occur. Even such changes 
should interfere little with maintaining the lines sufficiently constant 
for practical purposes. 
The data presented offer a number of points of interest besides 
the question of the possibility of obtaining larger yields from crosses 
between self -fertilized lines. The erratic yields of the lines in the 
comparison of successive generations give an excellent idea of what 
to expect during the earlier generations of self-fertilization in such an 
experiment. Lines are isolated occasionally that apparently breed 
true almost from the beginning. The 12-1-1- and 12-2-2- lines 
afford a good example of this condition. Segregating following the 
second selfed generation, these lines have remained consistently 
different through what is now the eighth generation of self-fertiliza- 
tion. Other lines become constant more slowly, offering greater 
opportunity for selection. 
There is abundant evidence in the data of Table 3 that selection 
can be effective in obtaining lines more productive than the average 
of the generation ; in fact, when the small number of lines involved in 
these experiments is considered, there seems to be every reason to 
expect that self-fertilized lines can be obtained that will be productive 
enough to obviate the need for utilizing double crosses for commercial 
planting. An even better example of the effects of selection is 
afforded by the 10-3- strain. Little difference was apparent between 
the rows grown from 10-3-1-1-, from 10-3-1-2-, and from 10-3-2-1- 
in the breeding plat in 1921. Selection since then has resulted in 
isolating reasonably fair lines of 10-3-1-2- one of which is shown in 
Plate IV. Lines of the 10-3-1-1- strain have been obtained that 
are uniformly as good as the best plants shown in this illustration. 
The 10-3-2- lines, on the other hand, became steadily poorer until 
they can be propagated now only with great difficulty (Pi. V). 
The data from the comparison of crosses made following different 
numbers of generations of self-fertilization indicate no general ad- 
vantage for crosses made following five generations over analogous 
crosses made after three generations of self-fertilization. This sug- 
gests that there is little inherent relation between the yield of a cross 
and the number of generations that its parent lines had been self- 
fertilized before crossing. In other words, self-fertilization appears 
to be a means of obtaining definite entities from which specific high- 
yielding combinations can be selected rather than a direct cause of 
these high yields. 
The lack of any definite correlation between the yields of the parent 
lines and their crosses indicates that selection for crossing must be 
