90 
Research Bulletin No. 2 
varieties, the frequency distributions for 1909 and 1910 are per- 
haps more useful than those of 1911. While the 1911 stocks of 
these varieties were doubtless more nearly homozygous than the 
stocks of previous years, they probably did not possess just the 
factors carried by the two germ cells whose union resulted event- 
ually in the 1911 F 3 families. Approaching homozygosis of other 
factors as well as of size factors rendered the 1911 parent stocks 
less vigorous than they were in previous years and thereby 
lessened the tendency to tiller, somewhat as poor or dry soil is 
known to do. The very few tillers produced by the 1911 stock 
of Tom Thumb, No. 1126, was due in part perhaps to this cause, 
in part to the very unfavorable weather conditions while it was 
developing, and doubtless in part also to the isolation of a type 
with comparatively few stalks. 
Whether any of the F 3 lots of this cross were homozygous for 
all the factors that have to do with number of stalks cannot be 
determined without further breeding. It seems probable that 
some of them, like 1145 for instance, that had a small range of 
variation will be found to breed practically true, while from 
others it will perhaps be possible to isolate several types.* 
Among the F 3 families of Tom Thumb X Black Mexican there 
was a noticeable difference in number of stalks per plant. Two 
families scarcely ever tillered, while one family averaged four 
tillers per plant. 
* While some intimation of what the future behavior of any F 3 family 
will be may perhaps be obtained from an inspection of the range of vari- 
ation and from the statistical constants, particularly the standard devi- 
ation, too much dependence cannot be put upon these things. It may be, 
as was suggested earlier for number of rows per ear, that types of plants 
with an inherently large number of stalks naturally fluctuate more than 
types with an inherently small number of stalks because each inherent 
stalk may furnish a basis for fluctuation. From this point of view, the 
coefficient of variation seems a better measure of relative fluctuation than 
does the standard deviation, because it expresses the deviation as a per- 
centage of the thing that deviates. From an inspection of the arrays 
(Table 36) and even more from an inspection of the plants themselves, it 
would, however, be difficult to believe that family 1146 is less variable than 
family 1145, even tho the coefficients of variation are 29.86 ± 1.60 per cent 
for the former and 48.31 ± 3.08 per cent for the latter. The standard devi- 
ations of these families are respectively 1.27 ± .06 and 0.80 ± .04 stalks. 
Only further breeding experiments, rather than statistical constants, can 
show how much of this variation in either case is mere fluctuation and 
how much due to genetic differences. 
