7.
Planning an experiment of the type just discussed asks no more than can be
secured within the same establishment. More comprehensive experimentation
requires the cooperation of external organizations, and is therefore more difficult
to arrange. It is believed, however, that an appreciation of the advantages of
large scale experimentation in the multi-factor style, and the limited usefulness of
disconnected individual pieces of work make the necessary cooperation easier to
obtain.
8.
The comparison of independently conducted experiments meets with certain
difficulties arising primarily from the subjectiveness of our judgment of the
quality of photography, the experience of observers, the state of adjustment of
the instruments, and so on.
It is clear that such comparison becomes less unreliable when the experiments
are designed to include more than one level of the subjective factor. It is never-
theless necessary to seek for objective counterparts to the subjective criteria, in
order to derive the largest amount of useful information from our experiments.
To fix ideas, let us consider the quality of photography which ranks first in
the list of factors influencing the results of any air survey work from every point
of view.
The quality of photography is usually described by such attributes as good,
poor, normal, usual, etc. The subjectivity of these attributes is self evident. The
important point is that they do not correspond to any universally accepted scale,
primarily because they are not related to an efficient objective criterion.
I shall venture to suggest it is both necessary and sufficient to grade photo-
graphs according to their detail rendering, e.g. according to the percentage of
detail which they depict identifiably, since we are not interested in their artistic
value.
It is clear that the resolving power of the lens, or the lens-emulsion combi-
nation cannot be expected to offer an adequate criterion for grading, especially
when the results of work carried out under varying conditions of photography are
concerned. I am not of course suggesting that the resolving power as determined
in the laboratory is of no use. But resolving power is one of several factors, and is
therefore not sufficient by itself.
To derive a more efficient estimator, the conditions of photography must be
taken into account. These include the type of ground, flying height, atmospheric
conditions, ground movement, vibrations, statics, etc... It is therefore first neces-
sary to sort out by experimentation, the relevant among all conceivable factors:
and, second, to find out a function (or a small number of functions) of the relevant
factors which discriminate most effectively between the detail rendering of the
photographs. The quality of photography used in any experimental investigation
can then be assessed by means of this function, using data which may have to be
specially recorded during photography and later stages.
Experimentation of this type demands more elaborate methods of statistical
treatment of data than any referred to in earlier Sections of this paper. But the
general idea of multi-factor experimentation is the same.
26