2                  Professor A. Leopold              6-15-44 
The statistical errors due to small numbers, themselves, 
can be treated and evaluated in a rather straight-forward manner. 
In any considerable body of data, the statistical sources of error 
are relatively small, and the biological sources relatively large. 
In almost every case where a biological source of error exists, a 
special variation of the method can be devised which gives, at the same 
time, a corrected estimate of the population and a quantitative 
expression of the biological phenomenon that causes the error when 
the simple general method was used. 
As I have said before, this method is an idealized base 
of reference based on the assumption that each individual is 
equally likely to be caught in a trap. 
At first blush, biologists regard the method as hazy and 
lacking in the concreteness that they feel when they use what might 
be called the "strong arm" methods of censusing. They seem to 
forget that our knowledge of physics and chemistry is based on 
such probabilities. In the physieal sciences they never attempt 
to make precise counts of the number of atoms, molecules, ate. 
Their conclusions are based entirely on deductive methods such as 
I talk about here. 
Also, biologists seem to be less impressed by the sim- 
plicity of this method than by the loss of particularity inherent 
in its generalities and abstractness. I think that this objection 
can be overcome when they realize that it is an idealized base of 
reference which can be used to measure the deviations brought about 
by certain sets of circumstances. 
Now for some imaginary examples -- Suppose that in trap- 
ping the pheasants from the Arboretum in winter, something inter- 
feres before you are satisfied that you have handled every bird, 
thus making a system of estimating of some sort necessary. You 
estimate your population and, then, use certain tests to determine 
its validity. For example:. 
i.        Repeaters.--It may appear that certain birds seem to be 
recaptured  oo o  en; i.e., they have become "trap happy".  In
any 
series of catches, once-taken birds, 2-, 5-, 4-, 5-time, etc. 
repeats should occur in certain proportions if the method is non- 
selective. If the distribution is reasonably close, all right. If 
one or a few birds are 'way out of line, they will lower your popu- 
lation estimates. The best way is to omit them from your calcula- 
tions, and add their number to your final estimate. Moreover, you 
can say that the probability that these birds are "plus-trap" con-

ditioned is e.g. 100 to 1.