EMPLOYMENT PSYCHOLOGY
by criterion and test. A mathematical expression of the
validity of a test answers the question, “To what degree does
the test measure abilities which are required for success in
the occupation?”
The reliability of a test is the accuracy of the measurements
 it yields. It answers the questions “How well does
this test predict itself?”, or “How accurately does it measure
 the abilities it is supposed to measure?” This accuracy
can be determined either (1) by repeating the test and correlating
 the two series of measurements, or (2) by correlating
records obtained with two similar forms of the test (or the
odd items with the even, or the first half with the second
half). Either coefficient of correlation is called the coeficient
 of reliability. The first is the upper limit and the second
 the lower limit of the true reliability coefficient (86, p.
203). Formulas for use in computing these correlations are
found on pages 174 ff.
Factors which make for a low reliability coefficient of a
test are:
1. Variability in the ability tested; that is, susceptibility
to embarrassment, fatigue, and other temporary influences,
and increase of the ability through training.
2. Errors in measurement due to faults in the test or in its
administration.
3. Restriction in the range of ability measured.
If a test has high reliability, the inference may be drawn
that the abilities it measures are not subject to fluctuations
or errors of measurement. Low reliabilities, however, must
be due to one or more of the above factors.
The first factor is diminished when only a few minutes are
allowed between the first and second forms, or when the odd
items are correlated with the even items of one form. It is
important, however, to know what fluctuations actually exist
in the ability in question, for if the ability is subject to ups
and downs as shown in low self-correlation of the test, then it
is difficult to measure that ability for purposes of vocational

166