PSYCHOLOGICAL TESTS 
the point where they refuse to continue, whereas those who 
are able to complete these difficult problems find the sailing 
smoother with the problems which follow. The resulting 
differences in score exaggerate the real differences in ability. 
If the items are of equal difficulty, speed is measured. If 
they are in order of increasing difficulty, then not only speed 
but level of ability is measured. 
LENGTH OF TEST 
The test should provide a sufficient sampling of the ability 
to yield a reliable measure. The law of diminishing returns 
applies to the length of a test. The first few items are the 
most valuable. Additional items increase the reliability of 
the test, but in diminishing proportion. It is better, how- 
ever, to err on the side of having too many items than too 
few. The validity of each item may eventually be deter- 
mined and the test reduced to include only those items which 
show the greatest vocational predictive value. 
The test may also sometimes be shortened after correlat- 
ing scores made in a part of it with scores made in the 
whole test. Any part of the test which correlates more than 
.95 with the whole test may be used in place of it. 
TIME LIMITS 
A time limit is not essential to a test, since it may be 
scored on accuracy alone. If accuracy is the variable which 
the test aims to measure, the time limit may be set so 
liberally that all but a very few (about 5% or 10%) are able 
to complete the task. But there is no point in allowing a 
person to fumble over a task indefinitely when it is obvious 
that he will not be able to complete it satisfactorily. A 
time limit which is set for convenience aims to meet this 
situation, and not to measure speed. 
On the other hand, if it is necessary to measure speed, the 
IX