16 BULLETIN 869, IT. S , DEPARTMENT OF AGRICULTURE. 
appeared to give homozygous dense progeny. Of the entire 22 
plants, representing all types of F 2 densities, nine proved about as 
variable in F 3 as the F 2 generation. 
Seven F 3 selections which appeared to be breeding true, as deter- 
mined by the frequency distribution and coefficient of variability. 
were tested in the F 4 generation. This was done by selecting 10 heads 
of different densities and growing the progeny of each separately. 
Where all heads gave similar results, they are combined in the table 
and are given as the result of 10 plants. 
The F 3 line 325-5, of which only 26 plants were available for study. 
gave a mean of 3.15 ±0.02 mm. in 1917, with a low coefficient of 
variability. On testing this line in 1918, when data from 213 plants 
were available, a somewhat higher mean was obtained, or 3.43 ±0.01 
mm. Its coefficient of variability is also somewhat larger than in 
the homozygous parental forms. Selection 325-15 proved pure in 
F 4 with the exception of the progeny of one plant which gave as 
great a variability as the F 2 generation. Why one plant should 
behave so differently from the nine others is difficult to explain. 
The possibility of a natural cross must not be overlooked, although 
observations show that these are very infrequent. An occasional 
error is also a possibility, although precautions were taken to elimin- 
ate these as far as possible. 
The F 4 means for the seven lines which gave evidence in F 3 and F 4 
indicating that they were homozygous are as follows: 325-5 (10 
plants), 3.43 ±0.01*^ mm.; 325-13 (10 plants), 3.47 ±0.01 mm.; 
325-16 (9 plants), 3.74 ±0.01 mm.; 325-18 (10 plants), 2.24 ±0.01 
mm.; 325-20 (10 plants), 2.47 ±0.02 mm.; 325-21 (10 plants), 
3.95 ±0.01 mm.: 325-22 (10 plants), 3.72 ±0.01 mm. 
Of these, five have mean densities which are not very different 
from that of the Jet (lax) parent, while the means of the other two 
are similar to that of the P}Tamidatum parent. The most dense 
and the least dense of the five lax homozygous segregates have mean 
internode lengths of 3.43 ±0.01 mm. and 3.95 ±0.01 mm., respectively. 
As great a difference as this in any one season would not be expected 
in a sort homozygous for similar characters. It is not much greater, 
however, than seasonal variation in the means of several of the pure 
2-rowed forms, which seem more susceptible to such variability 
than the 6-rowed parents. Inheritance of such a reaction difference 
might possibly explain the results here represented. Whatever expla- 
nation may be given for these new means, here, as in the- Man- 
churia X Svanhals cross, no homozygous forms were produced 
which differed materially in density from the density of one or the 
other parent. 
