114 THE AMERICAN NATURALIST. [VoL. XXXVI. 
III. METHODS AND PRECAUTIONS. 
In order that the personal equation should modify results as 
little as possible, all of the heads of each individual were col- 
lected, and none were discarded on the ground of abmodality. 
The only material rejected was such as was eaten by insects, 
or had been blighted to such extent as to make a correct count 
impossible. It is evident that by the discarding of healthy 
material because of its great departure from the usual condition, 
or, as Strong (1901, p. 295) says, the choice of ‘individuals 
which appear on inspection to be typical the statistical 
method may be made to give any result for which the investi- 
gator may be looking. 
There is one element of error in the choice of material which 
must be mentioned. In those cases in which it seemed desir- 
able to get curves and * constants," representing the conditions 
in the capitula of single individuals, there was necessarily a 
choice of those individuals which had the largest number of 
blooming heads. In this respect, therefore, the determinations 
would represent the conditions in the more robust specimens 
rather than in the general population. To counteract this 
tendency, there was an occasional collection of material from 
a sufficiently large group of smaller individuals to give equally 
valid results. As these collections from a number of smaller 
individuals did not show a marked difference from those of 
single larger individuals, this selection has probably not greatly 
modified the results. 
In making counts the liability of error is not so great as in 
the taking of measurements. However, even in counting there 
are sources of error, which I found it necessary to eliminate as . 
far as possible. To free my work as far as possible from these 
sources of error, the capitula were carefully dissected and the 
parts kept separate until the count was completed. When 
there was the least doubt as to the correctness of the result 
they were recounted. Also in cases of great abmodality the 
. results were verified by a recount. 
In calculating the various constants I have used the formula 
given by Davenport (99). All the mathematical processes 
