SYSTEMATIC IMPROVEMENT OF PLANTS — 201 
that is, we are unable to discriminate between the pure J)? and 
the 2 Dr with its unnoticeable recessive, 7. How, now, shall 
these be separated ? 
It is a long and difficult process. The only procedure is to 
plant the seeds, separately if possible. Those that are pure 
dominants will of course produce only dominants, while those 
that are mixed will produce both kinds; that is, among these 
no recessive will appear. In self-fertilizing species we can 
quickly separate the pure dominant strains, but when it is neces- 
sary to resort to cross fertilization, either natural or artificial, it 
is evident that the work is still more difficult. Under such circum- 
stances the only way is to proceed at random until a strain 
appears that produces no recessives, when it may be confidently 
assumed that the parents were both pure dominants and that 
the separation has been effected. 
When more than two characters are involved. It is sufficiently 
difficult to separate two characters, one of which is dominant. 
Manifestly, it is still more difficult to effect separations when 
three or more characters are involved, especially if we are con- 
cerned with all of them. 
Of course, in practical improvement we neglect all characters 
that do not concern us, whether they are dominant or recessive ; 
but, on the other hand, it is seldom that we are concerned with 
so simple a problem as the separation of a single character from 
its recessive or dominant associations. When our problem is 
to separate two or three such characters from their hereditary 
entanglements, the job becomes akin to hunting for the tradi- 
tional ‘‘needle in a haystack,” because the combinations are 
exceedingly complicated ; for we remember that the individuals 
which are recessive as to one character may be dominant or 
mixed as to others.!. The only way, however, is to run it down 
1 This is why, if Jerseys and Holstein-Friesians should be crossed, some of 
the offspring would be rich in certain Jersey characters and others in other 
characters, either Jersey or Holstein-Friesian ; but under the law of chance not 
once in a million times, or more perhaps, would a single animal be pure Jersey 
with reference to a// the Jersey characters. Besides this, it is more than likely 
