Cronbach. Studies of Acquiescense.
Cronbach. Studies of Acquiescense.
Cronbach. Studies of Acquiescense.
EDUCATIONAL PSYCHOLOGY
Volume XXXIII September, 1942 Number 6
Standard
Number of items M jan Score P^ AI nit i f\ n
Reliability
LJCv 1UtIVli
"*" Critical Correlation
1 OS I
No of
Field Directions ratio* Trues X Fnlsea
No. cases Keyed Keyed Trues Falses Trues Falsea Trues Falses Total
Total
true false
(1) (2) 3) (4) (5) (O (7) (C )-(5) (8)
recognize. After all students have marked those items on which they
think they are correct, one would expect to find the acquiescent
student to have marked more false items " t r u e " than vice versa.
Then, when forced to balance his responses, on the items he does not
know, he will be forced to make further errors, marking true state-
ments "false" to compensate. In effect, these guessed responses
are determined for him by the responses he considered certain, so
that for the acquiescent student the number of degrees of freedom
is reduced. For the non-acquiescent student, who responds " t r u e "
and "false" equally often, the number of degrees of freedom equals
the total number of items in the test, as it does under normal directions.
The result of this factor in the experimental directions is to shorten
the test for the overacquiescent or overcritical student, thus reducing
reliability and validity. A similar factor may explain the failure of
the novel directions proposed by Dunlap, De Mello, and Cureton
to improve the accuracy of testing.
DISCUSSION OF IMPLICATIONS
Findings.—The data gathered in these studies, taken as a whole,
reinforce the belief that the tendency to respond "true" more often
than "false" affects scores of students on true-false tests. Evidence
supports, but does not prove, the viewpoint that this tendency is a
trait present in different amounts in different students. False items
are on the whole superior to true items, as the reliability and validity
of scores based on false items are usually greater than those based on
true items. Discovery that, within a test of sixty-seven items having
a reliability of 0.616, the thirty-four false items scored alone would
give a reliability of 0.723, while the thirty-three true items give only a
reliability of 0.111, is certainly worth the attention of those using
true-false tests. True items and false items do not measure the same
behaviors, which means that one or both is in part invalid. Further
research will be needed to clarify many questions raised by these
explorations. At present, it is significant to translate these findings
into suggestions for improving the true-false test.
Suggestions for Improving the Validity of Testing.—Directions
intended to eliminate the effect of acquiescence proved useless, raising
the question whether any other approach can control this factor. It
is doubtful whether control through other novel directions will solve
the problem, but some other more or less useful procedures suggest
themselves. First, one may use more false items and fewer true items
412 The Journal of Educational Psychology