英语论文网

are unique or used in only one study.
One reason for the JDI's wide use is the care with which it was developed.
Its development is described in detail in Smith et al. (1969) and in a
recent paper by Smith and Sandman (1979). Another reason is its applicability
across a wide variety of demographic groups (Golembiewski &
Yeager, 1978).
The 72-item instrument was designed to measure 5 theoretically and
practically useful dimensions of job satisfaction—satisfaction with the
work itself, supervision, co-workers, promotion opportunities, and pay.
However, a recent study has questioned this five-dimensional structure.
Smith et al. (1974) report seven rather than five dimensions because a scree
test (Cattell, 1966) indicated that there were seven non-trivial factors in
their data. These factors are discussed later in this paper. Smith et al.
(1974) did not focus on the dimensionality of the JDI, nor did they examine
the utility of more than five factors. The present study examines the dimensionality
of the JDI, the utility of a larger number of factors, and their
implications.
Smith et al. (1974) reported results from analysis of 3 different samples
containing 212 white and 107 black government employees, and 110 white
bank employees, respectively. The supervision items split into performance
and interpersonal dimensions in all three of their samples. The
work scale split into two dimensions in only one (white civil servants) of
their samples. The items for co-workers, pay, and promotion opportunities
loaded consistently on the a priori factors. Given the results of Smith
et al., a similar split of the supervision items was expected in the present
study. Because the co-worker items specifically refer either to peer performance
on the job or to interpersonal relations with peers, it was expected
that they might form two factors. This would parallel the split along similar
lines reported for supervisory items by Smith et al. (1974). The pay and
promotion items were expected to load entirely on the a priori factors for
several reasons. First, these factors remained intact in all three samples in
Smith et al.'s study. Examination of these items further reinforces this expectation.
Divisions or subtopics are not clearly apparent as they are
among the supervision and co-worker items. Third, there are fewer items
in these scales, 9 instead of 18, and this lessens the likelihood, given the
comparative homogeneity of items, of a split occurring. Expectations
about the pattern of the factor loadings for the work itself items were
more problematic given Smith et al.'s mixed results and their small sample
1981 Yeager 207
sizes. The items in the scale split in only one of their three samples. This
could be an artifact unique to that sample. Its small size increases the likelihood
that the sample is unique in some way that would cause unique results.
Differences in results between the Smith et al. (1974) study and the present
study also are expected for several methodological reasons. The characteristics
of the subjects in the sample used in this study are different
from their three samples. For example, the type of work they perform and
the communities they work in are different. As a result, there is variation
in each sample that is unique to it, which can influence factor analytic results.
Second, when the nu