April, '24] 
hyslop: statistical methods 
179 
upon accurate count of adequate samples beyond that point where the 
probable error is less than the results obtained. The ascertaining of 
what is an adequate sample is far less simple than it might seem to the 
casual observer. There is no established correlation between insect 
abundance and unit areas; a gregarious insect of sedentary habits 
might easily be conceived as always remaining near its place of origin or 
introduction. Samples taken at any point in this place bear no relation 
to the contiguous uninfested territory. It would be an evident fallacy 
to take the estimated number of gipsy moths in the infested part of the 
United States (Ma) and divide this by the total area to get the average 
infestation for the United States and then apply this figure to Georgia as 
the infestation in Georgia is 0. Many other cases will be extremely 
difficult: observation on the abundance of an insect that moves with 
the wind will bear no definite relation to the total area surveyed unless 
its point of origin in that area is located and the samples taken with 
respect to that point. These, and many other factors, which introduce 
biased conditions must be worked out by the investigator. I believe 
that we could profitably spend much time in the next few years in 
investigating intensively the dispersal of given pests on given crops as 
only in this way can we ascertain what constitutes an adequate sample 
This will involve selecting an area such as an orchard, a wheat field, a 
cabbage patch, etc., small enough to be completely examined in a time 
short enough to eliminate the factor of time distribution of the pest in¬ 
volved. This area will then be divided into units as small as is reason¬ 
ably measurable as 1 apple tree, 1 cabbage plant, a square foot of alfalfa, 
etc. 
Counts are then made of each unit in the area to be surveyed and 
mapped. This will bring out trends, such as a constant diminishing in 
number of insects from high to low side of field, with and against the 
wind, on different soils, etc., and will probably necessitate subdividing the 
original area. 
Now an analysis of the data should show the point where a frequency 
of observational units picked at random over the area will reduce the 
probable error to a negligible factor. By repeating this experiment a 
number of times this adequate sample should become fixed under similar 
conditions. 
These basic investigations, of course, will take years so our immediate 
problem as I see it is to adopt, as far as possible, uniform unit samples. 
For example; future investigation may indicate that 1 per cent of the 
trees in an orchard will give an adequate sample for codling moth if all 
