66 
SYMPOSIUM PHOTO INTERPRETATION, DELFT 1962 
Accuracy and completeness in general do not seem to be dependent on 
whether or not an interpreter views a positive transparency, negative trans 
parency or print, although a further study needs to be made to check on the 
finding for 30-minute completeness at low photo quality. 
Good photo quality does result in better accuracy and completeness than 
does poor photo quality. However, this finding applies only to the materials 
at hand. Our “good” photo quality is actually not fixed. We have no way of 
measuring the quality of our “high” quality photos and we downgraded these 
photos to a barely acceptable level. Since we were not able to improve the 
photo quality of our “good” photos, we are unable to say that an increment 
in photo quality beyond that of our good quality would likewise yield an 
increment in accuracy and completeness. Therefore the finding does not argue 
for improved photo quality. What should be done first is to develop a measure 
of photo quality and then systematically for different quality levels determine 
the accuracy and completeness of interpretations to establish the point of 
diminishing returns, which would then be the goal for the desired level of 
photo quality. 
Photo interpreters generally tend to have more confidence in their right 
responses than in their wrong ones. This means that the probability of accurate 
identification will be higher for responses associated with high confidence than 
for those associated with low confidence. However, the use of this knowledge 
must await the determination as to whether or not one mode is favored over 
another with respect to confidence differential between right and wrong 
responses. Should there be such a mode, this mode would be preferred, every 
thing else being equal, since it would make possible the identification of 
correct responses with a higher degree of probability than any other mode. 
Mode differences most likely don’t have much of an effect on accuracy as 
a function of time, however quality does. The accuracy curves do not taper 
off in the same way for the time period of concern. There is a continuing very 
evident drop in accuracy for high photo quality, whereas accuracy for low 
quality remains fairly constant until almost the very end of the period where 
there are beginnings of a loss in accuracy. The use of specified work time to 
establish desired accuracy levels therefore requires the establishment of ac 
curacy-over-time-functions for different levels of quality.