PSYCHOMETRIKA--VOL.16, NO.

3
S~.FrF.MB ~ , 1951

COEFFICIENT ALPHA AND THE INTERNAL

S T R U C T U R E OF TESTS*

LF~ J. CRONBACH
UNIVERSITY OF ILLINOIS

A general formula (a) of which a special case is the Kuder-

Richardson coefficient of equivalence is shown to be the mean of all
split-half coefficients resulting from different splittings of a test.
is therefore an estimate of the correlation between two random sam-
ples of items from a universe of items like those in the test. ~ is
found to be an appropriate index of equivalence and, except for v e r y
short tests, of the first-factor concentration in the test. Tests di-
visible into distinct subtests should be so divided before using the
formula. The index ~ j , derived from a , is shown to be an index of
inter-item homogeneity. Comparison is made to the Guttmau and
Loevinger approaches. Parallel split coefficients are shown to be un-
necessary for tests of common types. In designing tests, maximum
interpretability of scores is obtained by increasing the firat-facter
concentration in any separately-scored subtest and avoiding sub-
stantial group-factor clusters within a subtest. Scalability is not a
requisite.

I. Historical Resum~
A n y research based on measurement must be concerned with the
accuracy or dependability or, as we usually call it, reliability of meas-
urement. A reliability coefficient demonstrates w h e t h e r the test de-
signer w a s correct in expecting a certain collection of items to yield
interpretable statements about individual differences (25).
E v e n those investigators who regard reliability as a pale shadow
of the more vital m a t t e r of validity cannot avoid considering the re-
liability of their measures. No validity coefficient and no factor analy-
sis can be interpreted without some appropriate estimate of the mag-
nitude of the e r r o r of measurement. The p r e f e r r e d w a y to find out
h o w accurate one's measures are is to make t w o independent measure-
ments and compare them. In practice, psychologists and educators
have often not had the opportunity to recapture their subjects f o r a
second test. Clinical tests, or those used for vocational guidance, a r e
generally worked into a crowded schedule, and there is always a de-
*The assistance of Dora Damrin and Willard Warrington is gratefully ac-
knowledged. Miss Damrin took major responsibility for the empirical studies re-
ported. This research was supported by the Bureau of Research and Service, Col-
lege of Education.
297