WORKING GROUP 1
BIRNBAUM
65
Fig. 1 shows the cumulative number of right and wrong responses for five-
minute intervals of viewing time, plotted separately for high and low photo
qualities. Notice first the far larger number of wrong than right responses
at all time intervals and regardless of photo quality. When we look at quality
separately we find for high photo quality that the number of new identifications
scored right tapers off much sooner than the number of those scored wrong.
This is not as evident for the low photo quality. Moreover, and very significantly
when we compare the performance for high and low photo quality, we find
invariably for all time intervals that the high quality photographs provide
more right and fewer wrong responses than do the low quality ones.
0 5 10 15 20 25 30
Time (minutes)
Fig. 1. Cumulative Right and Wrong Fig. 2. Accuracy as a Function of Time
Responses as a Function of Time by Photo by Photo Quality
Quality
Fig. 2 shows accuracy which relates right to right plus wrong responses,
plotted as a function of time for the same performance measures. For high
photo quality there is a continuous decrease in accuracy as time increases,
a finding which conforms to previous ones. For low photo quality, on the
other hand, accuracy remains relatively constant at a lower level at all time
intervals as time increases, although towards the end it begins to decrease.
Actually, if additional time had been provided to the interpreters working on
the low quality photographs, a continuing drop in accuracy would not be
unexpected, since any responses that go beyond the limits of what there is in
the photograph would have to be wrong. Moreover as the interpreter addresses
himself to objects that are increasingly more difficult to identify we would
expect the ratio of wrong to right responses to increase and hence for accuracy
to decrease.
Implications of photo-mode study findings
To the extent to which the test materials and conditions and test subjects
are representative, the findings of these studies are generalizable. If they are
the following conclusions can be made.
