influence of the non-uniform distortions of the film. The use of a larger-than-four 
number of collimating marks would produce a similar improvement in connection 
with the use of cameras that have no register glass. There is no difficulty in the 
analytical treatment of photographs carrying the reseau and, as Prof. Thomp- 
son ©) has shown, no additional labour is involved. But what use can be made of the 
information supplied by the reseau of collimating marks when using a plotter if 
one is not prepared to do a large amount of calculations? 
These remarks are not intended to imply that the plotting machine would 
not give sufficiently accurate results for many purposes when circumstances are 
favourable. They only suggest that the use of plotting machines in air triangulation 
cannot obtain the best results from any quality of photography, except with such 
an amount of calculations that would offset the very advantage of using the 
plotter. But this argument in praise of analytical methods can hardly convince 
the practical air surveyors to use them because of the practical shortcomings of 
their application. It does, however, show the need for developing the analytical 
approach, and has encouraged investigations (^: ^9 into the sources of difficulty 
and the ways in which they can be overcome. 
2. The Problem of Computations. 
It is well to realize at the outset that the difficulties of the analytical method 
cannot be overcome by reducing the number of observations or by applying 
approximate solutions. There is in fact ample evidence (^ that it is necessary to 
observe more points on each overlap than the usual six points. The logical con- 
clusion to these findings is that the only promising approach to the practical 
problem of calculation is to mechanize as much as possible of its processes. 
2.1. The Use of Electronic Computing Machines. 
The use of electronic calculating machines is a fascinating proposition. It may 
however be worth while considering its limitations. 
Electronic engineers have been able to produce machines which execute any 
sequence of arithmetical operations on a given set of numbers in a minute fraction 
of the time that a human being would take. But the machine does this if, and only 
if, the following conditions are satisfied: (1) That the numbers on which the 
operations are required must be suitably coded (e.g. on punch cards, (2) That the 
instructions as to what the machine is required to do with these numbers must be 
worked out in advance and must be suitably coded. And, (3) that these instructions 
must be logical. 
It is important to realize that the machine does not think for itself, and 
would certainly fail if the instructions given to it are inadequate. If for example 
a zero appears as denominator of a fraction during the operation of the machine, 
this would certainly try hard to give the right answer, namely infinity, unless 
it had been given explicit instructions to stop when such a situation arises. 
The accuracy of the results depends not only on the precision of the observati- 
onal data which are put into the machine, but also on the efficiency of the criteria 
which are coded in the programme of operation and on which the machine shall 
2