6 BULLETIN 1092, U. S. DEPARTMENT OF AGRICULTURE. 
third generation, but they did not result in strains of value. This 
negative result may perhaps be due to the fact that only a limited 
number of crosses were made. The crosses between selections from * 
the Blue-Blossom Dutch and White-Blossom Dutch varieties resulted 
in strains that were intermediate in resistance to flax wilt. The 
flax under greenhouse conditions took four and one-half to five < 
months to mature and made an abnormal growth. Some of the 
plants reached a height of 170 centimeters (5J feet), so that it was 
found necessary to support them on wires. 
It was found impracticable to handle large numbers of selections 
with the centgener method, because of the time consumed in planting- 
seeds one at a time, and flax selections after 1914 were sown in drill 
rows. Uniformity of growth conditions similar to those of the cent- 
gener method was secured in 1918 by thinning out the rows to one 
plant to the inch. 
ELIMINATION OF POORER SELECTIONS. 
Very early in the work those selections that were most promising 
were sown on a larger scale than the others, so that in 1914 enough 
seed had been secured from the one then considered the best to sow 
it at frequent intervals throughout the experimental plats. From 
year to year additional selections were made, as in the year 1909, and 
wherever one of them growing next to this standard selection has 
been judged inferior it has been discarded. This standard selection 
has been called the "check," as it acts as a check on the soil condi- 
tions. The check serves the same purpose as a ruler placed alongside 
a plant and is in one respect better than a ruler, for it measures the 
soil conditions by growing tall where the soil is rich and short where 
the soil is poor. 
As more and more of the selections were discarded, those retained 
became more and more like each other, because all of them possessed 
in some degree each of the desired characters. (See the selections 
shown in Fig. 1.) Thus, it became necessary to study a larger num- 
ber of characters as a means of elimination of the poorer selections. 
In addition to length of stem and stem weight, the following char- 
acters were added : Strength of fiber per individual stem, the amount 
of basal branching, the vitality of seed, and resistance to disease. 
The check selection has furnished a ready means of comparison be- 
tween the different selections for all these characters. Data were 
rapidly accumulated on the check, as it was sown in many duplicate 
plats and measured extensively. 
Strength tests were begun in 1912, and during the years 1913 to 
1917 hundreds of strength tests were made under the direction of 
Mr. Frank C. Miles. In this way many inferior selections were 
eliminated. It was found that the middle portion of the stem 
