A Revision of the Job Diagnostic Survey: Elimination of a Measurement Artifact

Abstract

The dimensionality of the original Job Diagnostic Survey (JDS) and a revision were investigated. Factor analyses of two data sets identified six dimensions underlying the original JDS. Five of the factors correspond to the pattern expected for the JDS items; the sixth was identified as a measurement artifact. Five of the JDS items were subsequently rewritten to eliminate the artifact. The revised survey was administered to employees of a printing company (N = 134) and the a priori five-factor solution was obtained with no artifact factor. Scale–factor correlations were also computed. The resulting coefficients suggest that the revised JDS scales are measuring their underlying constructs with reasonable accuracy. As a result of the measurement artifact in the original JDS, it is recommended that the revised JDS should be used in future research concerned with task characteristics. (18 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)

Full text

lounul of ApplMd Pncholoty
1917.\W,72.No
l,<»-74
Co|lyn|ht
J
987 by the American Psycholoxical Aaaociation, Inc.
O021-9O1O/87/S00.75
A Revision of the Job Diagnostic Survey: Elimination
of a Measurement Artifact
Jacqueline R. Idaszak and Fritz Drasgow
University of Illinois, Urbana-Champaign
The dimensionality of the original Job Diagnostic Survey (JDS) and a revision were investigated.
Factor analyses of two data sets identified six dimensions underlying the original JDS. Five of the
factors correspond to the pattern expected for the JDS items; the sixth was identified as a measure-
ment artifact. Five of the JDS items were subsequently rewritten to eliminate the artifact. The revised
survey was administered to employees of a printing company (N = 134) and the a priori five-factor
solution was obtained with no artifact factor. Scale-factor correlations were also computed. The
resulting coefficients suggest that the revised JDS scales are measuring their underlying constructs
with reasonable accuracy. As a result of the measurement artifact in the original JDS, it is recom-
mended that the revised JDS should be used in future research concerned with task characteristics.
Characteristics of jobs play a central role in organizational
theory. They can be viewed as "technology's most direct conse-
quences" (Hulin & Roznowski, 1985, p. 71). Job enlargement
and job enrichment programs typically treat job characteristics
as the independent variables that should be matiipulated. More
generally, a variety of organizational theories hypothesize that
job characteristics are precursors of job-related affect, produc-
tivity, and withdrawal (Hackman & Oldham, 1974,1975; Mow-
day, Porter, & Steers, 1982; Turner & Lawrence, 1965).
At present, the most popular perceptual measure of job char-
acteristics seems to be Hackman and Oldham's (1974, 1975)
Job Diagnostic Stirvey (JDS). Its popularity, however, is more a
consequence of Hackman and Oldham's theory of job charac-
teristics (upon which the JDS is based) than the psychometric
properties of the instrument
itself.
Specifically, several ques-
tions remain unanswered with respect to the latent structure
of the JDS. Until recently, evidence that the JDS measures the
hypothesized dimensions was weak (E>unham, 1976; Dunham,
Aldag, &
Brief,
1977; Fried & Ferris, 1986; Green, Armenakis,
Marbert, & Bedeian, 1979; Pierce & Dunham, 1978; Pokomey,
Gilmore, & Beehr, 1980).
Dunham and his colleagues (Dunham, 1976; Dunham et al.,
1977; Pierce & Dunham, 1978) and Pokomey et al. (1980)
looked at the factor solutions for a wide variety of samples and
found very few that resembled the a priori five-factor structure.
Dunham (1976), for example, advocated a single-factor solu-
tion representing job complexity. Other studies have accepted
two-,
three-, and four-factor solutions in addition to the rarely
encountered five-factor structures (Dunham et al., 1977; Fried
We extend special thanks to Charles L. Hulin for his helpful com-
ments on earlier versions of this article and to Greg R. Oldham and
J. Richard Hackman for their help in revising the Job Diagnostic Sur-
vey items.
Correspondence concerning this article should be addressed to Jac-
queline R. Idaszak, Department of Psychology, University of Illinois,
Psychology Building, 603 East Daniel Street, Champaign, Illinois
61820.
& Ferris, 1986; Green et al., 1979; Pierce & Dunham, 1978;
Pokomey et al., 1980). From this set of studies it would seem
appropriate to conclude that the JDS should be empirically ex-
amined in each new subpopulation.
In an attempt to explain these inconsistencies. Fried and
Ferris (1986) investigated possible moderators of the underlying
JDS factor structure. Based on their results they suggested that
age,
education level, and position level influence the factor
structure.
Harvey, Billings, and Nilan (1985) used a different approach
in an attempt to resolve the JDS dimensionality issue. They
used confirmatory factor analysis to evaluate the factor struc-
tures suggested in past research. Their results suggest that none
of the factor structures found in the past provide an adequate
fit. Instead, an oblique solution vvith Hackman and Oldham's
a priori dimensions plus one or two method factors (factors for
the reverse-scored items and the three-anchor scale items) pro-
vided the best fit. As a result of the existence of these measure-
ment artifacts, it was concluded that the JDS, in its present
state, is psychometrically "troublesome." Furthermore, a re-
vised version of the JDS should be developed that eliminates
the measurement artifact.
Although Harvey et al. (1985) provide a plausible explana-
tion for the inconsistencies found in past JDS research, it is still
possible, as Dunham et al. (1977) suggested, that the results of
JDS factor analyses are sample specific. Further evidence from
heterogeneous samples is needed to verify the measurement ar-
tifacts suggested by Harvey et al.
The research described in this article provides replications of
Harvey et al.'s (1985) finditigs with two separate samples. In
Study 1, the responses of a heterogeneous subsample from Old-
ham, Hackman, and Stepina's (1978) JDS data base were se-
lected and factor analyzed. Study 2 cross validated the factor
structure identified in Study 1, using an independent subsam-
ple.
In Sttidy 3, several JDS items were revised in an attempt to
correct the measurement artifact identified in Studies 1 and 2.
The revised items were administered to a new sample, and the
data set was factor analyzed. We obtained a five-factor solution
that closely reflected Hackman and Oldham's theory.
69

70JACQUELINE
R.
IDASZAK
AND
FRITZ DRASGOW
Table
1
Six-Factor Maximum Likelihood Factor Structure Using
the
First Sample From
the 1978 JDS
Data Base
Item
Factor
Autonomy
1
2
3
Task Identity
4
5
6
Skill Variety
7
8
9
Task Significance
10
II
12
Feedback
13
14
15
25(04)
18(05)
32(03)
74(02)
67(02)
63(05)
23(04)
-14(05)
50(03)
78(06)
70(03)
08(03)
44'(03)
39*(04)
37'(03)
42*
(03)
44*(04)
-08 (04)
-11(04)
47(04)
66(04)
40(04)
-13(04)
58(04)
31(04)
72(03)
-22(06)
-24(04)
-19(05)
83(04)
73(03)
46(04)
22(04)
30(08)
49(04)
55(04)
49
(03)
05
(05)
19(05)
50(04)
41(04)
—
04(07)
21(07)
23(06)
Factor correlation matrix
31(05)
53(04)50(03)
Note.
All
factor loadings fixed
at
zero are omitted; decimal points are omitted; standard errors
of
parameter estimates are
in
parentheses.
JDS
•-
Job Diagnostic Survey.
• This item is a reverse-scored item on the original JDS.
General Analysis
For each study
the
following analyses were conducted. First,
a reduced correlation matrix, using squared multiple corre-
lations
as
communality estimates, was obtained
for the 15 JDS
items.
Principal axes factor analyses were computed
for one
through
six
dimensions.
A
parallel analysis (Humphreys
&
Montanelli,
1975) was
performed
on the
eigenvalues. Then
each principal axes solution
was
rotated orthogonally
by
vari-
mai and obliquely
by the
direct artificial personal probabilities
factor rotation
(DAPPFR)
of Tucker and Finkbeiner (1981).
In
the
next step
of the
analysis,
the
DAPPFR
solutions were
tised
as
starting values
for
maximum likelihood factor analyses
computed
by the
LISREL IV
computer program.
A
minimally
identified solution
was
first obtained. Then small loadings were
fixed at zero according
to the
Kroonenberg
and
Lewis (1982)
procedure.
In
this procedure,
t
statistics provided
for the
load-
ings
in the
LISREL IV
output were examined
and all
loadings
associated vAth sttiall
/
values were set
to
zero. Chi-sqtiare statis-
tics were used
as
measures
of
the goodness
of fit of
the various
solutions.
The
first
derivatives of loadings fixed
at
zero were also
used
as
measures
of fit. A
ratio criterion
of
increase
in chi-
square
to
increase
in
degrees of freedom
was
used
to
select
the
"best" restricted solution. Analyses were stopped when
a
large
ratio was obtained.
The final analysis performed provided
a
measure
of
the
ex-
tent
to
which scale scores correspond
to the
underlyitig
con-
structs.
The
correlations between scales
and
factors, termed ^-
delity coefficients (Drasgow & Millet; 1982), were computed
for
each scale derived from
the
principal axes factor analysis solu-
tions and the restricted maximum likelihood factor analysis so-
lutions.
Study
1
Purpose
The first study
was
performed
to
investigate whether
mea-
surement artifacts distort the factor structure of the JDS.
Method
Subjects.
DaU for
Study
1
consisted
of
subsamples from
the
Old-
ham et
al.
(1978)
JDS
data base randomly selected within job category.
This sample
(N=
1,672)
consisted
of
377
professionals, 380 managers,
416 clerical workers,
336
processing workers,
and 163
machine trade
workers. These categories were chosen because
of
their large sample
sizes
and the
heterogeneity
of
the group.
Of
the subjects, 53% were
women;
60%
were between the
ages
of 20
and
40
years;
49%
had a high
school degree or less
education;
and
47%
had some technical or college
education
or a
degree
from a technical school
or a
college.

JDS DIMENSIONALITY71
Table 2
JDS Scale-Factor Correlations From the Restricted
Maximum Likelihood Factor Analysis
Scale
Task Identity
Task Significance
Skill Variety
Feedback
Autonomy
Reverse scored
Task Identity
Task Significance
Skill Variety
Feedback
Autonomy
Revised items
1
m
.17
.13
.31
.35
.31
M
.08
.25
.26
.35
.53
2
Study 1
.10
Jl
.39
.37
.27
.32
Study 3
-.11
M
.19
.23
.13
.35
Factor
3
.25
.50
3^
.38
.58
.54
.21
.16
£1
.24
.34
.52
4
.29
.47
.40
.44
.47
.23
.33
.44
M
.39
.69
5
.37
.33
.40
.41
J!2
.40
.37
.15
.37
,42
M
.59
6
.14
.26
.41
.37
.38
Jl
Note. The diagonal entries are fidelity coefficients. Factors are presented
in the same order as the scales. JDS = Job Diagnostic Survey.
Instrument. The JDS includes 15 items that measure the five core
job characteristics of task identity, task significance, skiU variety, auton-
omy, and feedback. Each of these core dimensions was measured by
three 7-point Ukert-type items. The internal consistency reliabilities of
the scales, using the entire sample ofN= 6,930, ranged from .71 (skill
variety and feedback from the job itself) to .59 (task identity).
Results
Eigenvalues of the reduced carrelation matrix for one
through seven factors were 3.68, 0.87, 0.64, 0.42, 0.38. 0.20,
and 0.02, respectively. The corresponding parallel analysis ei-
genvalues were 0.17, 0.13, 0.10, 0.08, 0.06, 0.05, and 0.03.
These eigenvalues indicate that there are six factors underlying
the JDS items. The maximum likelihood goodness-of-fit mea-
sure also clearly indicated that more than five factors were re-
quired to model the correlations among the JDS items.
Table 1 presents the end result of the maximum likelihood
factor analyses. The chi-square goodness-of-fit measure for this
solution is 162.58 with 58 dji. The six-factor solution for the
principal axes factor analysis usiqg the
DAPPFR
rotation closely
resembles this solutian and therefore is not presented. In bath
af the salutions, five of the factors correspond to the JDS a pri-
ori pattern. Skill variety items load on Factor 1, task identity
items load on Factor 2, task significance items load on Factor
4,
autonomy items load on Factor 5, and feedback items load
on Factor 6. In addition, all items requiring reverse scoring load
on a single factor (Factor 3). The same reversed scoring factor
was obtained when we extracted five factors.
The factor carrelation matrix is alsa included in Table 1. As
anticipated, the JDS factars are maderately intercorrelated;
The average factor correlations between JDS facton is .42. Con-
ceptually, these relations may result from the existence of a
common latent construct that is directly linked to manifest
charaaeristics of the task (Hulin & Roznawski, 1985). Haw-
ever, cammon method variance cannot be overlooked as a possi-
ble cause of these correlations.
The average correlation between the artifact factor (or arti-
factor) and the JDS factors is. 12 after the
DAPPFR
rotation and
.17 in the maximum likelihood factor analysis solution. These
coefficients are expected to be small because the arti-factor does
nat measure a latent construct related to the technological as-
pects of
jobs.
In addition, the small arti-factor-JDS factor co-
efficients suggest that the average correlation of .42 among the
JDS factars may not be due entirely to method variance. How-
ever, as one reviewer of this article painted out, the factor inter-
carrelatians may also be caused, in part, by other artificial
sources.
Job Diagnostic Survey scale-factor correlations were com-
puted using the method presented by Drasgow and Miller
(1982).
The fidelity coefficients abtained from the restricted
maximum likelihood factor analyses solutian are presented in
the main diagonal of Table 2. The range of the coefficients ex-
tends from .75 to .85 for the JDS scales; it is .71 for the arti-
factor The fidelity coefficients for principal axes factor analysis
and DAPPFR rotation were very similar.
According to Drasgow and Miller (1982), scale-factor co-
efficients above .90 may be necessary for construct validation
research. This criterion is very difficult to attain, however, for
scales as short as the JDS scales. For some types of research,
coefficients in the .80s may be acceptable as long as the variance
in scales not accounted for by the factars they measure is due
ta random measurement error. In Study 1, some of the fidelity
coefficients fell below the .80 criterion and the scales contained
systematic errors caused by the measurement artifact. It is evi-
dent that some revision of items requiring reverse scoring is nec-
essary.
Study 2
Purpose
The second study was performed to replicate Study 1 with an
independent sample.
Method
Data for Study 2 consisted of subsamples from Oldham et al.'s (1978)
JDS data base. This sample (N = 565) consisted of 132 service workers.
Table 3
Revised Job Diagnostic Survey Items
ScaleItem
Autonomy The job gives me a chance to use my
personal initiative and judgment in
carrying out the work.
Task Identity The job is arranged so that I can do an entire
piece of work from beginning to end.
Skill Variety The job requires me to use a number of
complex or high-level skills.
Task Significance The job itself is very significant and
important in the broader scheme of
things.
Feedback After I finish a job, I know whether I
performed well.

72JACQUELINE R. IDASZAK AND FRITZ DRASGOW
Table 4
DAPPFR Rotation of the Six-Factor Principal Axes Factor
Analysis Solution for the Revised JDS
Study 3
Item
Autonomy
1
2
3
Task Identity
4
5
6
Skill Variety
7
8
9
Task Significance
10
11
12
Feedback
13
14
15
1
2
3
4
5
6
1
.60
.77
.74
-.11
.30
.22
.42
.10
2
.72
.68
.64
Factor
3
.35
.74
.49
.76
4
.23
.60
.64
.57
5
.52
.43
.66
.33
Factor correlation matrix
.28
.27
.15
.09
.32
.47
-.18.33
.25-.15
6
.09"
-.21
.07*
.22
.09*
.05*
.03*
—
Purpose
Note. All loadings less than .20 are omitted except for the loadings of
the rewritten items on Factor 6.
DAPPFR
= direct artificial personal
probabilities factor rotation; JDS = Job Diagnostic Survey.
• This item is rewritten so that reverse scoring is not necessary.
161 bench workers, and 272 workers classified as
other.
These job cate-
gories were selected so that they were mutually exclusive with the cate-
gories used in Study 1. Of the sample 71
%
were male; 68% of the workers
were between the ages of 20 and 40 years; 50% had a high school degree
or less education; and 35% had some technical or college education but
did not receive a degree.
Results
The first seven eigenvalues of the reduced correlation matrix
for the second sample are 3.32,0.73,0.67,0.64,0.39,0.23, and
0.13,
respectively. The corresponding parallel analysis eigen-
values are 0.32, 0.25, 0.20, 0.16, 0.13, 0.09, and 0.06. Once
again,
six factors seem to be plausible based on the parallel anal-
ysis results.
Principal axes factor analysis with a
DAPPFR
rotation and re-
stricted maximum likelihood factor analysis were computed for
six dimensions. The results of the two analyses were very similar
to our findings in Study 1 and consequently are not presented.
A sixth factor again emerged that was defined by the five reverse-
scored items.
In Studies 1 and 2 it was determined that an arti-factor is
needed to explain statistically the intercorrelations of JDS
items.
Moreover, the arti-factor is clearly detrimental to the
measurement accuracy of the JDS scales as indicators of the a
priori factors because it systematically affects scale scores.
In Study 3 we attempted to improve the measurement prop-
erties of the JDS scales by revising the items requiring reverse
scoring.
We attempted to rewrite these items in ways that main-
tained their original meanings yet did not require reversed scor-
ing.
The revised JDS was administered to a third sample of
workers and the responses were factor analyzed.
Method
Subjects. Data for Study 3 were collected from 94 female and 40
male employees of a printing plant in central Illinois. All of the employ-
ees working in the customer information/service, documentation, mail-
ing service/records, distribution, order processing, printing, and bind-
ery departments were asked to participate. Of the employees who re-
sponded, 87% were white, 47% were married, 56% had up to a high
school education, and 31% had I to 3 years of college education. The
average age of the respondents was 31, the average tenure with the orga-
nization was between I and 3 years, and the respondents worked an
average of 38 hr per week.
Instrument.
A revised version of the JDS was developed after con-
sultation with the instrument's original authors. The format of the re-
vised survey was the same as the original JDS. Only items requiring
reverse scoring in their original form were rewritten. The five rewritten
items are presented in Table 3. Note that this listing does not reflect the
format of the JDS.
Administrative procedures. The written questionnaires were dis-
tributed to employees by the supervisor of each department. Employees
were informed that their participation was voluntary, their replies
would be kept anonymous, and employees of the company would only
see results based on aggregated data. Employees were given 1 week to
complete the questionnaire. The return rate across all departments was
approximately 65%.
Results
The first seven eigenvalues of the reduced correlation matrix
are 3.72,1.17,0.83,0.72,0.52,0.15, and 0.07, respectively. The
corresponding parallel analysis eigenvalues are 0.76,0.59,0.48,
0.39, 0.31, 0.24, and 0.14. These eigenvalues indicate that five
factors underlie the revised JDS. We, nonetheless, examined the
six-factor solution in order to verify that our revisions elimi-
nated the arti-factor.
The six-factor solution that resulted from the
DAPPFR
rota-
tion is presented in Table 4. It clearly indicates that we have
extracted too many factors. Notice that all loadings of the re-
written items on Factor 6 are nearly zero.
Given the lack of an arti-factor in the principal axes factor
solution,
the five a priori core job dimensions should now ap-
pear as distinct dimensions in the five-factor solution. Principal
axes factor analysis and both unrestrirted and restricted maxi-
mum likelihood factor analysis were used to obtain five-factor
structures.
The goodness-of-fit measures for the unrestricted

JDS DIMENSIONALITY73
Table 5
Five-Factor Maximum Likelihood Factor Structure for the Revised Job Diagnostic Survey
Item
Factor
Autonomy
1
2
3
Task Identity
4
5
6
Skill Variety
7
8
9
Task Significance
to
II
12
Feedback
13
14
15
62(09)
79(08)
80(08)
29(09)
81(10)
32(10)
66(10)
70(09)
71(09)
68(09)
29(09)
31(10)
29(10)
25(08)
61(09)
85(08)
56(09)
42(10)
37(10)
89(11)
-13(11)
25(11)
26(10)
42(10)
16(11)
26(1.1)
13(11)
Factor correlation matrix
24(12)
37(11)37(10)
Note.
All factor loadings
fixed
at zero are
omitted;
decimal points
are
omitted;
standard
errors
of parameter estimates are in parentheses.
and restricted solutions, resjjectively, are x^ (40) = 44.04, and
X^
(74) = 86.90. Both measures indicate a good fit.
The results for the restricted maximum likelihood factor
analysis are presented in Table 5. The
DAPPFR
solution was very
similar. As expected, the hypothesized JDS structure is clearly
evident in Table 5. Factors 1 through 5 can be identified as the
job dimensions of Task Identity, Task Significance, Skill Variety,
Feedback, and Autonomy, respectively. The factor intercorre-
latiotis ranged from -.
13
to .42 in the maximum likelihood fac-
tor analysis solution.
To check the measurement accuracy ofthe revised scales, fi-
delity coefficients were computed. Table 2 presents the Study 3
fidelity coefficients that resulted from the restricted maximum
likelihood factor analysis. The fidelity coefficients for the Task
Identity, Task Significance, Skill Variety, Autonomy, and Feed-
back factors are all above .80. In sum, it appears that the fidelity
coefficients for the revised JDS scales are reasonably high and
we can conclude that the JDS scales are measuring their under-
lying constructs with accuracies that are adequate for theoreti-
cal research.
EMscussion
Since the introduction ofthe Job Diagnostic Survey, numer-
ous studies have anempted, with limited success, to obtain em-
pirical support for the hypothesized five-factor structure. Rea-
sons for the lack of success in this area of research were not
apparent until the Harvey et al. (1985) study. They suggested
that reverse-scored items were a major source of the inconsis-
tencies. In the present study we provide confirmation that the
reversed scored items are indeed the source ofthe problem. It
is interesting that this arti-factor can also be identified in the
factor structures presented by Dunham (1976), Dunham et al.
(1977),
and to a lesser extent, Pokomey et al. (1980) and Fried
and Ferris (1986). Given the diversity ofthe samples used in
these factor analyses, it seems safe to conclude that the reverse-
scored items have caused the difficulties in factoring the JDS.
Ironically, Hackman and Oldham (1974,1975) deliberately in-
corporated reverse-scored items into the survey to minimize
response bias. Unfortunately, their effort backfired and seems
to have caused a substantial amount of
mischief.
Identification of this measurement artifact may have been de-
layed, until recently, for several reasons. First, few studies have
extracted more than five factors for the original JDS items. In
many cases, such as in the Dunham et al. (1977) four-factor
solution, the arti-factor seems to be stronger than some of the
JDS factors. When the arti-factor ap[)eared in a solution, it may
have misled researchers to believe that they had overfactored
and, consequently, they may not have examined solutions in
higher dimensionalities.
In addition, identification may have been delayed because the
salience or the "strength" ofthe arti-factor may vary as a result
of differences in reading comprehension (Green et al., 1979),

74JACQUELINE R. IDASZAK AND FRITZ DRASGOW
education and position level (Fried & Ferris, 1986), and atten-
tiveness to negatively worded items (Schmitt & Stults, 1985).
Fried and Ferris's (1986) results suggest that the arti-factor is
"weaker" for workers with more education or higher reading
comprehension ability. They obtained good approximations to
the a priori structure only in subsamples with more education
and higher position levels. Such workers may be more adept at
reversitig the negatively scored items and may therefore respond
using the same criteria that are used on other items. Workers
with lower reading abilities and less education may have more
difficulty in reversitig items mentally and therefore may respond
incorrectly to the items, thus creating the arti-factor. It is inter-
esting to note that Schmitt and Stults (1985) found that "nega-
tive" factors appear when as few as 10% of the subjects fail to
notice that some items are reverse scored.
Our approach to revising the JDS was to replace reverse-
scored items with new items that did not have to be reverse
scored. Some of the results presented by Harvey et al. (1985)
suggest that it may also be necessary to use a single response
format to eliminate all measurement artifact factors. However,
we found no evidence of an arti-factor due to response format
in Study 3. Of course, additional research with large samples
collected from diverse organizations is needed before we can
conclude unequivocally that the measurement artifacts have
been eliminated. At present it seems appropriate to conclude
that the new JDS items have substantially improved measure-
ment properties and the new scales should be used in future
research concemed with task characteristics.
Although the measurement artifact seems to have been elimi-
nated, investigations are still needed to better understand why
reverse phrasing of some items can lead to artifact factors. It is
possible that the arti-factor is simply a consequence of the lack
of salience of the five items written in a different, reverse-scored
format. A reviewer suggested that it may be possible to elimi-
nate the arti-faaor by explicitly pairitig items that are worded
in opposite directions. To study this hypothesis for the JDS,
items could be written and paired with current JDS items. The
relative salience argument would be supported if the arti-factor
did not appear in a factor analysis of the augmented scale.
In any case, a revised JDS now exists that appears to ade-
quately measure the five task dimensions exphcated by Hack-
man and Oldham (1974, 1975). We are now able to go beyond
considerations of the psychometric adequacy of the JDS and
study the conceptual role of task characteristics in organiza-
tional behavior.
References
Drasgow, F,
&
Miller, H.
E.
(1982). Psychometric
and
substantive issues
in scale construction and validation.
Journal
of Applied
Psychology,
67,
268-279.
Dunham, R. B. (1976). The measurement and dimensionality of job
characteristics.
Journal
of Applied
Psychology.
61, 404-409.
Dunham, R. B., Aldag, R. J., &
Brief,
A. P. (1977). Dimensionality of
task design as measured by the Job Diagnostic Survey. Academy of
Management
Journal.
20. 209-223.
Fried, Y,
&
Ferris, G. R. (1986).
The dimensionality
of job
characteris-
tics:
Some
neglected
issues.
Manuscript submitted for publication.
Green, S. B., Armenakis, A. A., Marbert, L. D., & Bedeian, A. G.
(1979).
An evaluation of
the
response format and scale structure of
the
Job
Diagnostic Survey.
Human
Relations.
iZ
181
-188.
Hackman, J. R.,
&
Oldham, G. R. (1974).
The Job Diagnostic
Survey:
An instrument for
the diagnosis
of jobs and
the evaluation
of job
rede-
sign
projects (Report No. 4). New Haven, CT: Yale University, De-
partment of Administration Science.
Hackman, J. R,
&
Oldham, G. R. (1975). Development of the Job Di-
agnostic
Survey.
Journal
of Applied
Psychology.
60. 159-170.
Harvey, R., Billings, R.,
&
Nilan, K. (1985). Confirmatory factor analy-
sis of the Job Diagnostic Survey: Good news and bad news.
Journal
of Applied
Psychology.
70.461-468.
Hulin, C. L., & Roznowski, M. (1985). Organizational technologies:
Effects on originators' characteristics and individual resfwnses. In
L.
L. Cummings & B. M. Staw (Eds.), Research in organizational
behavior
(Vol.
7, pp. 39-85) Greenwich,
CT:
JAI Press.
Humphreys, L. G.,
&
Montanelli, R. G.,
Jr.
(1975). An investigation of
the parallel
analysis criterion for determining
the
number of common
(acton.
Multivariate Behavioral
Research,
10. 193-205.
Kroonenberg, P. M., & Lewis, C. (1982). Methodological issues in the
search
for a
factor
model:
Exploration through confirmation.
Journal
of Educational Statistics.
7.69-89.
Mowday, R. T, Porter,
L.
W.,
&
Steers,
R. M. (1982). Employee-organi-
zation
linkages:
The psychology
of commitment, absenteeism, and
turnover
New
York:
Academic Press.
Oldham, G. R., Hackman,
J.
R.,
&
Stepina,
L.
P.
(1978). Norms for
the
Job Diagnostic Sumey (Tech. Rep. No. 16). New Haven, CT: Yale
University, School of Organization and Management.
Pierce, J. L.,
&
Dunham, R. B. (1978). The measurement of perceived
job characteristics: The Job Diagnostic Survey versus the Job Chai-
acteristidnventory. Academy of
Management
Journal,
21, 123-128.
Pokomey, J. J., Gilmore, D. C, & Beehr, T. A. (1980). Job Diagnostic
Survey dimensions: Moderating effect of growth needs and corre-
spondence with dimensions of the Job Rating Form.
Organizational
Behavior
and
Human
Performance.
26. 222-237.
Schmitt, N.,
&
Stults,
D. M. (1985).
Factors
defined by negatively keyed
items:
The result of
careless
lespoadeDtsT Applied Psychological Mea-
surement,
9, 367-373.
Tucker, L. R, & Finkbeiner, C. T. (1981).
Transformation
of factors by
artificiat personal
probability functions (Research Rep. No. RR-81-
58).
Princeton, NJ: Educational Testing Service.
Turner, A. N.,
&
Lawrence,
P.
R. (1965). Industrial jobs and
the
worker.
Boston, MA: Harvey Graduate School of Business Administration.
Received February 28,1986
Revision received May 19, 1986