a
Archives 6
mail SINGES RE FE rte
Intra-individual differences are also demonstrated by an examina-
tion of the component raw scores (number right or wrong) by target.
Each individual's performance varied so evidently from target to target
that it was not necessary to transform these scores into standard scores
to demonstrate the differences. For example, Examinee A had the lowest
component right score among all 16 interpreters (well below average per-
formance) on Target 9, and his right score on Target l was the highest
(well above average).
Contribution of Individual Differences to Improved Output. Such
extensive inter- and intra-individual differences point to improvements
in output possible through utilizing individuals in terms of their best
abilities. Table 5 provides for each target the highest, average, and
lowest completeness scores achieved by the 16 photointerpreters. The
average of scores for the highest target was 65% compared with 39% for
the average target score and 10% for the average of the lowest target
scores. Thirty-nine percent is the average score (output) to be ex-
pected if photointerpreters were assigned at random to interpret the
nine targets. On the other hand, if the interpreters had been pre-
viously assigned to interpret their speciality targets, the average
completeness would have increased to 65%. The effective use of classi-
fication and assignment measures is clearly indicated by the difference
between 65% and 39% output.
Accuracy and Completeness of Modal Responses. The analysis up to
this point has been concerned with the output of the individual photo-
interpreter. This section deals with the quality of information pro-
vided by the group as a whole and by sub-groups of interpreters.
Although individual interpreter performance has been shown to be
far from perfect, the outlook as to the type of product that can be ex-
pected is quite encouraging. It must be remembered that results to this
point deal with base performance of individual interpreters working in-
dependently. An analysis of the &ccuracy and completeness of modal re-
Sponses was made for the experimental group of experienced Pl's,
Responses by individuals were compared in terms of completeness with
(1) modal responses by the total group of 16 interpreters, (2) modal
responses by any three interpreters selected at random, and (3) modal
responses for the best two interpreters for a given target.
The group modal response for component scoring was found to be 93%
accurate as compared to 75% for the average individual; for the two best
on each target it was 94%, and for three random interpreters, 96%. For
component-modifier scoring, the respective accuracies were TTh, 50%, 91%,
and 87%.
- 15 a