COftff BREEDING 43 
purple would produce only plants having ears with purple kernels, 
regardless of how these plants were crossed among themselves. 
If the object were to eliminate kernels with purple aleurone, the 
problem would be more difficult. Seed of each of the classes 5 to 9 
would breed true for colorless aleurone as long as the plants were 
self-pollinated or as long as no cross occurred bringing C and R to- 
gether. Crosses of either 5 or 6 on the one hand, with 8 or 9 on the 
other, would have all or some purple seeds. The lines that could be 
crossed without producing purple kernels could be determined only 
by experiment. Once determined, however, the selected lines could be 
combined into an F x cross year after year, with the knowledge that 
the cross would produce only nonpurple kernels. 
This illustrates well the difference between the certainty of success 
of selection within selfed lines in obtaining the desired condition for 
such a relatively simple character and the almost equal certainty of 
failure of selection in open-fertilized corn. Under selection within 
selfed lines the production of a variety of corn with purple or white 
kernels can be undertaken with the definite knowledge that it will be 
obtained. Under selection in open-fertilized material the only hope 
of obtaining a purple sort would lie in the chance of planting nothing 
but kernels of class 1 in some season. That this is a chance is evi- 
dent from the fact that the kernels in class 1 are indistinguishable 
practically from those in classes 2, 3, and 4. Success of selection 
for a variety with colorless aleurone in open-fertilized corn would 
equally be dependent upon chance. Only the accidental selection of 
all seed ears either in one or more of classes 5, 6, and 7, or in one or 
more of classes 7, 8, and 9, would insure against purple seeds in later 
generations. 
Except as the possibilities for obtaining specific characters were 
considered, the preceding discussion of selection has been from the 
point of view of obtaining desirable selfed lines. It obviously is 
necessary to have reasonably productive lines because of the practical 
considerations of seed production. How effective selection will be 
during the earlier generations is not certain. Some data show little 
relation between the characters of lines in the first and in the fifth 
generations (42). Other data show a definite tendency for lines to 
behave similarly in successive generations both with regard to yield 
(30, 64) and with regard to other characters studied (30). The 
effectiveness of selection in the earlier generations presumably would 
differ in different varieties. Close-bred strains should become fixed 
more quickly, whereas more heterozygous stocks would respond less 
quickly. 
Whether the more productive selfed lines will produce the higher 
yielding crosses is even less certain, some data indicating that they 
do (48) and other data showing no general relation (64). It is 
easy to see why such a relation should be obscure, or possibly not 
exist. Consider, for example, a line breeding true for dwarfness, 
owing to a simple recessive factor. Such a line might carry also an 
'unusually desirable complement of dominant favorable factors 
which, nevertheless, could not produce large plants or yields because 
of the dwarfing action of the single pair of factors. The expression 
of all the favorable dominant factors carried by the dwarf line would 
be permitted in crosses between it and any other lines carrying the 
