2l8 
E. M. EAST 
and any departure noted in order that a correction for any constant 
error due to it may be calculated, if it be advisable. But, one might 
ask, would not any trained geneticist have taken these precautions 
anyway? What has been gained? 
The advantages are real. Unsuspected constant errors often come 
to light through such preliminary investigations. The good fortune 
that none appeared here certainly makes it no less satisfactory. It 
showed that control of conditions in such a manner that constant 
errors will be negligible in the end result is technically possible. It 
gave a definite idea of the magnitude of the error produced when 
various environmental factors do vary, and this is very necessary in 
determining the probable limits of error. 
There is a way of testing the conclusion that with the conditions 
controlled as suggested the constant error is negligible. If the same 
plants are measured during similar portions of successive periods of 
flowering activity, there is but one other obvious variable — total age 
of plant. If the latter has no measurable effect the two frequency 
distributions should duplicate. On this point I have no data, but 
Goodspeed and Clausen have corroborated the expectation in their 
conclusion number two. I do have some data on random samples 
of the same pure line grown in different years. This will be taken up 
later, however, as another point is involved. 
Now the question arises: If records are made in this uniform 
manner, how many records from each plant are needed to obtain a 
measure of that plant with the precision necessary for a genetic investi- 
gation? Goodspeed and Clausen say that twenty-five flowers is the 
minimum. At the beginning of my Nicotiana investigations (1908), 
I used the same number, curiously enough. But I soon found that 
this was ''accuracy with no significance," and the number was reduced 
to five. I now use but one measurement per plant. This is done 
because the precision is so nearly that of using twenty-five flowers, 
that it would be a waste of labor to try to attain the other. Further- 
more the precision obtained by measuring twenty-five flowers is only 
appreciably greater when it can be done in a short time, otherwise 
constant errors may become very much greater. 
The precision attained by measuring one flower per plant is all 
that is required for the use to which the data are to be put, and it is 
a rule of experimental physics not to strive for greater accuracy. 
This matter can and has been tested in two ways. The first is to 
