R< sat fdi Bulletin No. 2 
had a mean total stalk lengtb of only 40.29 decimeters while that 
of 1146 was 79.88 decimeters — almost twice as great. Evidently 
in actual practice one must select a number of promising F 2 
plants to stand a fair chance of getting an F 3 type to his liking. 
There is very little doubt that the striking results secured in 
the well known selection experiments with corn carried out at 
the Illinois Experiment Station (Smith 1908) are to be ex- 
plained on the same basis as that used in this discussion to show 
how types can be isolated by the proper use of small numbers 
of plants in F 2 , F 3 , F 4 . or later generations — types that could 
not be expected to appear in F 2 unless excessively large numbers 
were grown. If oil and protein content in corn are inherited in 
the same way that other quantitative characters are, we should 
not expect to find very high or very low oil or protein content 
in any 100 or 200 ears taken at random from a lot of open-pol- 
linated, unselected plants, most of which are almost certainly 
heterozygous in oil and protein factors — just as they are known 
to be for other characters — and have, therefore, an intermediate 
development of these characters. It is not necessary to resort to 
the somewhat mystic idea that selection (the mere act of choos- 
ing) has profoundly changed the oil and protein characters of 
this corn nor to call into service the possibility that mutations 
occurred repeatedly during the progress of the experiment. There 
is as much reason for assuming that the height-of-plant character 
as exhibited in the F 2 ranges (Table 25) was changed into a 
much smaller height character in F 3 by selection or that a 
mutation occurred opportunely. Between the isolation of a 
short-stalked type in F 3 from a largely heterozygous F 2 and the 
isolation of a low or high oil type from an open-fertilized lot of 
corn — also largely heterozygous — there is apparently no founda- 
mental difference. 
A word should be said of the practical possibilities of pro- 
ducing sizes greater or less than the parent sizes. A noticeable 
example of this from our own experiments is the production of 
a type with a total length of stalks more than twice that of the 
larger parent, I See Table 38.) This was accomplished by com- 
bining the many-stalked condition of one parent with the tall- 
stalked condition of the other parent. Visibly different parent 
characters were combined to form a new character complex. The 
production of a type with internodes much longer than those of 
the parents (Table 33) differs from the above only in that here 
the parent types were apparently very similar and it could not 
be foreseen that the similar internode lengths of the two varieties 
were due to different factors that could be combined into a new 
character complex by crossing. Tt is of course unlikely that 
