150 THE CANADIAN SURVEYOR May, 1960 
It seems that the cause of this fluctuation is in part physiological. There is, 
however, another factor that may play a more important role in photogrammetric 
operations than is generally recognized. This is the arbitrary decision regarding the 
apparent location of the point. This decision is made each time the measuring mark 
is set on the point. Once the decision is made the operator seems to adhere to it during 
a limited period of time. As a result, consecutive pointing proved to be more consistent 
than an “independent” pointing after a certain lapse of time when a new decision as 
to the location of the point must be made. This can be proved by a simple experiment. 
First the accuracy in consecutive pointing at a certain number of points is determined 
by the usual procedure—performing ten consecutive measurements at each point. Then 
the procedure is altered so that all points are read one after another in ten rounds. It 
can be shown that the pointing accuracy computed from the second determination is 
worse than that determined from the first observations. 
The practical significance of this phenomenon is obvious. For the absolute orien- 
tation of a model only a limited number of ground control points is available. Each 
point is presumably read a few times with the utmost care, but as we have seen, 
fluctuations in the elevation readings cannot be avoided. This uncertainty in pointing 
produces a tilting of the optical model and a wrong shifting of the reference plane, 
and both these factors contribute to the existence of systematic errors, positive or 
negative, depending on the sign of committed errors. 
This situation is somehow aggravated by the psychological factors involved. As 
is commonly known, the greatest importance is unconsciously attached to the first 
observation or impression. In consequence, the photogrammetric operator often 
adjusts the consecutive readings to the first one. Therefore, a series of consecutive 
readings of one point is seldom composed of really independent observations. Worse, 
it is extremely difficult to divorce the real value from the desired values known to the 
operator. After completing the first approximative rotation of the stereomodel the 
operator has already noted signs of discrepancies on particular ground control points, 
and therefore the consecutive setting of the measuring mark on the same points is 
affected by his desire to obtain good agreement between photogrammetric and ground 
values. In our experimental work, for instance, we discovered that the readings of two 
operators at control point 48 differed consistently by about 1.6 m. After comparing 
the results with ground values it was found that the error lay with only one operator. 
He was told what the ground value was and from that moment all his determinations 
at point 48 were within the usual tolerance. 
In order to assess the personal errors better, a stereomodel with 39 check elevations 
were set on the Wild Autograph A-7 and six operators read all the check points and 
the five ground control points for the absolute orientation. The photographs used 
were RC-5 Aviogon photographs, size 18 X 18 cm, f = 115 mm. The scale of the 
photographs was 1:23,000. The temperature in the laboratory was kept constant 
within 2°C, and the operators knew neither the ground values of points nor the results 
of previous operators. Each result was then reduced separately to a reference plane 
established from five control points used for absolute orientation according to the 
readings by a particular operator. The results are tabulated in diminishing order of 
accuracy in Table VI. The interesting part of it is that whereas the result produced 
by Operator 1 is free from systematic errors (shift of —0.04 m is negligible) an analysis 
of the results of Operator 6 seems to indicate a curvilinear distortion of the model 
which, however, is not visible from the readings of the same operator at the five points 
used in absolute orientation. On the other hand, Operator 4 produced a very random 
distribution of errors, with the exception of recordings of the five absolute orientation 
points which are notably shifted. If there were no systematic shift of —0.26 m, the 
result would be better by 209% and the mean square elevation error would amount to 
+ 0.43 m only.