It is no strange fact that the requirements of accuracy and economy should 
be conflicting. The difficulties of compromising is aggravated in photogrammetry 
by the large number of factors involved. 
These are clearly problems for experimental research. 
Many experiments have been made. Many more are needed. The point, 
however, is that the results of some would not pass the critical test of experience. 
Recurrence of this failure can only lead to distrust experimentation, and to 
strengthen the belief that trial and error is the only path of wisdom. 
It is in the nature of things that no limited amount of experimentation can 
lead to 100 per cent correct results. Modern statistics has, however, done much 
to show how to improve their reliability. It tells us how to plan experiments to 
obtain the maximum amount of information from the number of observations 
which we can afford to make; and how to determine the reliability of inference 
from them. 
Scrutiny of some publications in photogrammetry has shown that consider- 
able improvement in their yielding could be achieved had statistical literature 
been consulted. Few publications seem to deviate so much from the well esta- 
blished lines of experimental research, in their design or presentation, or in both, 
that hardly any use can be made of them. 
Much has been written in the subject of experimentation in various fields 
of research, and I gratefully record my indebtness to the writers from whom ! 
borrowed the statistical approach expounded in the following pages. 
In the following pages, the argument is illustrated by discussions of parts of 
a few of the outstanding publications in air survey and photogrammetry; but no 
attempt was made toward a comprehensive discussion of these publications, 
since the idea was simply to provide examples to elucidate certain points in 
this paper. 
2. 
A discussion of parts of some well known experiments on the dimensional 
stability of air survey films makes an appropriate starting point, by illustrating 
a few rather common shortcomings. 
2.1. The first remark which one might make on some photogrammetric 
publications is the inadequacy of the given information. 
For instance, the long term shrinkage of four types of films was investiga- 
ted (1) by storing samples for periods up to 12 months at temperatures of 70°, 
90°, 120° F; then measuring the distortions under identical conditions. The 
published account omits two necessary pieces of information: 
(a). The exact number of film pieces in each sample was not given. It is however 
mentioned that „two to five strips (35 millimetres X 15 inches) or grids were 
tested from each sample of film..." That this information is not adequate can 
be shown by calculating the uncertainty about the actual numbers when only 
two films are involved. Assume that the two films demonstrate equal, sampling 
variability, i.e. equally erratic. Let s be the sampling variance (or the square of 
the standard deviation, or the mean square deviation); and let us suppose that 
the two samples consisted of two and five pieces respectively. The variances of 
9