Some preliminary evidence points to the relative difficulty of
interpreting tactical photographs, especially those of a type found in
the Korean War. In the try-out administration of five performance mea-
sures using Korean War photographs, the average &ccuracy of & class of
l7 enlisted men about to graduate from the PI course was 31% and the
average completeness was 29%. Of the 65 different targets located by
at least one member of the class, the average examinee located approxi-
mately 19 targets correctly. On the other hand, the average number of
targets wrongly identified was approximately 43. One examinee made over
90 misidentifications. Most photointerpreter misidentifications were
the result of wrongly calling natural formations targets of significance.
This was especially evident for tactical scenes in which targets were
spread out over a large area.
The performance of individual photointerpreters in the school
sample varied considerably, as was also the case for the sample of ex-
perienced interpreters. The number of correct identifications ranged
from a low of 6 to a high of 29 for the sample of 17 interpreters. The
number of misidentifications ranged from 22 to 91. This wide variation
in performance points to the possible fruitfulness of employing effec-
tive selection measures to obtain men who will not only provide many
correct identifications but who will also provide fewer misidentifica-
tions. A small group of officer graduates of the PI course (N = 10),
on whom the performance measures were also tried out, did not differ
significantly from the enlisted men in number of correct identifications,
but made on the average about half as many misidentifications as were
made by the enlisted men.
These data also bring out the necessity for human factor research
in improving work methods and techniques. The possibility of achieving
marked gains in performance by the simple procedure of pooling independ-
ent photointerpreter responses is indicated by the results obtained with
three of the l7 enlisted men who had performed relatively well on the
measures. When only those responses which either two or all three of
the interpreters had independently agreed upon were considered, the ob-
tained accuracy jumped to 85%, and there was a slight increase in com-
pleteness as well. These results, which are similar to those obtained
for the modal responses of the sample of experienced interpreters,
point to the idiosyncratic nature of wrong responses. If one photoin-
terpreter sees something which is really not there, the probability is
apparently very high that no one else will make the same 'inventive'
misidentification. However, there is & good chance that if the target
is really there, other photointerpreters will agree with the identifi-
cation.
The importance of obtaining the correct procedures for integrating
the responses of & team of PI's is indicated by some other preliminary
results. In & try-out administration to & second class of enlisted men,
- 18 -
TIN Ll ft Spd UT el fub ^mi as
Archives 6