Journal of Applied Psychology Copyright 1988 by the American Psychological Association, Inc.

1988, Vol. 73, No. 3,462-466 002I-9010/88/S00.75

Measurement of Job Characteristics: Comparison of the Original

and the Revised Job Diagnostic Survey

Carol T. Kulik Greg R. Oldham

Carnegie Mellon University University of Illinois at Urbana-Champaign

Paul H. Langner
Animal Health Services, Wyeth Laboratories
Philadelphia, Pennsylvania

This study contrasted the original version of the Job Diagnostic Survey (JDS; Hackman & Oldham,
1975) with the revised version recently proposed by Idaszak and Drasgow (1987). A total of 224
dairy workers completed both versions of the JDS. Results of a confirmatory factor analysis showed
that the revised JDS job characteristics items conformed more closely to the hypothesized five-factor
structure than did the original JDS items. However, results of LISREL analyses indicated that the
revised items did not improve the usefulness of the JDS in predicting several outcomes (e.g., satisfac-
tion, internal motivation, and productivity).

In its most general form, job characteristics theory (Hack-
man & Oldham, 1980) posits that five job characteristics (au-
tonomy, task identity, task significance, skill variety, and task
feedback) prompt a number of personal and organizational out-
comes. Nearly all of the research that has tested the theory has
used the Job Diagnostic Survey (JDS; Hackman & Oldham,
1975), an instrument designed to measure incumbents' percep-
tions of the job characteristics, their satisfaction, and internal
motivation.
Much of the research regarding the JDS has focused on the
dimensionality of the job characteristics measures (cf. Dun-
ham, 1976; Dunham, Aldag, & Brief, 1977). Although a few
studies have confirmed the five hypothesized job dimensions
(e.g., Katz, 1978; Lee & Klein, 1982), most have reported solu-
tions inconsistent with the a priori five-factor structure (e.g.,
Dunham, 1976; Pierce & Dunham, 1978).
Recent research has examined the possibility that the JDS
itself might be responsible for these factor structure inconsis-
tencies. Idaszak and Drasgow (1987) conducted factor analyses
of two samples of employees. Results revealed six-factor solu-
tions for both samples—five factors corresponding to the hy-
pothesized factor structure and the sixth representing the nega-
tively worded, reverse-scored JDS items. Harvey, Billings, and
Nilan (1985) used confirmatory factor analysis to evaluate the
factor structures suggested in past research. Their results sug-
gested that the best fitting solution included the five a priori
The authors thank Fritz Drasgow, Doug May, and Motohiro Mori-
shima for their help with the data analyses. Robert Billings, Jeanne
Brett, Fritz Drasgow, and Motohiro Morishima provided helpful com-
ments on earlier drafts of this article.
Correspondence concerning this article should be addressed to Carol
T. Kulik, Graduate School of Industrial Administration, Carnegie Mel-
lon University, Schenley Park, Pittsburgh, Pennsylvania 15213.
dimensions plus one or two method factors (factors for the re-
verse-scored items and the three-anchor scale items).
In response to these criticisms, Idaszak and Drasgow (1987)
revised the JDS by rewriting the reverse-scored job characteris-
tics items. The factor structure of the revised JDS was then
investigated for a sample of printing plant employees. The re-
sulting factor structure conformed very closely to the a priori
five-dimension structure. As a result, Idaszak and Drasgow con-
cluded that the new scales should be used in future research
concerned with job characteristics.
Two limitations of the research on the revised instrument
lead us to believe that this recommendation may be premature.
First, the Idaszak and Drasgow (1987) study did not involve a
comparison of the original JDS with its revised counterpart
within the same sample. As demonstrated by Dunham et al.
(1977), some samples display a five-factor structure when the
original JDS is used. Thus, it is possible that the five-factor
structure observed by Idaszak and Drasgow is a function of the
characteristics of the sample they used to assess the instrument.
Second, it is unknown what impact the JDS revisions might
have on the criterion-related validity of the JDS. If the JDS revi-
sions result in "purer" measures of the job characteristics, there
should be stronger associations between the revised measures
and the personal and organizational outcomes than between the
original measures and the outcomes.
The present research uses confirmatory factor analysis to ex-
amine whether the revised JDS items conform more closely to
the a priori factor structure than do the original JDS items
within the same sample. We also use LISREL analyses to assess
the relative effectiveness of the two instruments in predicting
several outcomes. LISREL is useful in examining the psychomet-
ric properties of the original and revised JDS because it allows
us to separate the measurement model from the structural
model. Specifically, we expect that the structural model would
not differ across the two versions of the JDS, but there should
be improvements in the measurement model.

462
 JOB CHARACTERISTICS
Method
Sample and Procedure
A survey measuring job characteristics and outcomes was mailed to
384 Illinois dairy employees. The farms employing these workers were
a random sample selected from Dairy Herd Improvement Association
(DHIA) records, which also provided productivity indicators for these
employees. In all, 58% of the employees (N = 224) returned their sur-
veys by mail. There were no observable differences between respondents
and nonrespondents in terms of farm characteristics (e.g., size, location,
and production). The mean age of the participants was 35.6 years;
modal education level was a high school degree; and 79% were men.
Measures
Job characteristics. The 15 original JDS items, followed by the 5
revised items, were included in the questionnaire.
Affective outcomes. Four original JDS items measuring growth satis-
faction and four JDS items measuring internal motivation were in-
cluded in the questionnaire after the job characteristics items.
Intention to quit. Employees' intentions to quit the dairy farm were
measured by three items suggested by Colarelli (1984).
Productivity. Two indicators of dairy workers' productivity were ob-
tained from DHIA records. Milk production measures the average
amount of milk (in pounds) produced per day per cow for the dairy herd
that the employee handles. Freshening interval measures the average
length of time in days for pregnant cows to freshen (i.e., to become ready
to give milk following calving). This interval reflects the health of the
cow in general and of the reproduction system in particular. Several
studies have demonstrated that such productivity measures are substan-
tially affected by the care given to the herd by dairy employees (Sea-
brook, 1972,1975,1978). Moreover, interviews with managers of dairy
farms indicated that these two measures were considered to be indica-
tors of employees' work effectiveness.
Results
Confirmatory Factor Analysis
A reduced correlation matrix, using squared multiple corre-
lations as communality estimates, was obtained for all 20 (15
original and 5 revised) JDS items. This matrix was then submit-
ted to a six-factor confirmatory factor analysis. In this analysis,
only the parameters consistent with Job Characteristics Theory
were set free. For example, the three positively worded auton-
omy items were allowed to load on the autonomy factor but
fixed to load zero on all other factors. The negatively worded
items were allowed to load on their respective content factor
and on the sixth factor, but constrained to zero on all others.
This analysis yielded x2( 150, N = 224) = 313.24, p < .001, indi-
cating a significant discrepancy between the factor model and
the observed data. An alternative indication of fit can be ob-
tained by examining the ratio of the chi-square to its degrees of
freedom. If the ratio is less than 2.0, a reasonably good fit is
indicated. In our analysis, a ratio of 2.09 was obtained, once
again indicating a statistically poor fit.
The modification indexes (Joreskog & Sorbom, 1985) associ-
ated with this factor analysis were then examined. A modifica-
tion index estimates how much the fit of the model can be im-
proved by setting the corresponding parameters free. All of the
parameters that were already free in the model have modifica-
463
tion indexes of zero. The distributions of the modification in-
dexes are approximately chi-square values with one degree of
freedom. Eight modification indexes had chi-square values
greater than 5.0. In the current analysis, values greater than 5.0
have probabilities of less than .025. Consequently, a second six-
factor confirmatory factor analysis was conducted, specifying
as free these eight parameters. This analysis yielded x2(142,
N = 224) = 256.96, p < .001. The ratio of the chi-square value
to its degrees of freedom was 1.81, indicating that this factor
model provided a reasonably good fit to the data.
Table 1 presents the estimated factor loading matrix and the
estimated factor correlation matrix resulting from the six-factor
maximum likelihood solution. In confirmatory maximum like-
lihood factor analysis, the factor correlations can be treated as
free parameters to be estimated from the data. It is not neces-
sary to rotate obliquely when the factor correlations are free
parameters. The results provide some support for the Idaszak
and Drasgow (1987) findings. Specifically, the five a priori di-
mensions do exist and there is a separate factor for the nega-
tively worded items from the original JDS. In addition, the re-
vised items load on the appropriate factors and show slightly
higher loadings on the skill variety, task significance, and task
identity dimensions than do the original JDS items. However,
the loadings for the revised and original items are nearly identi-
cal for autonomy, and the original item has a higher loading
than the revised item on the feedback construct. These results
suggest that the revised items provide purer measures of variety,
significance, and identity than do the original items, but do not
substantially improve the measurement of autonomy and feed-
back.
Correlational Analysis
We next examined relations between the job characteristics
and outcomes. The original three items from the JDS were aver-
aged to obtain a summary score for each of the five job charac-
teristics. A second set of summary scores for each of the five
characteristics was next obtained by replacing the negatively
worded item in each scale with the revised item. To obtain a
summary measure of the job characteristics, a motivating po-
tential score was formed for both the original and revised JDS
using the formula suggested by Hackman and Oldham (1975).
Summary scores for the growth satisfaction, internal motiva-
tion, and intention to quit measures were obtained by averaging
the items tapping these variables.
Correlations between the job characteristics measures and
outcome measures are shown in Table 2. The pattern of corre-
lations for the original and revised version of the JDS is very
similar. Contrasts between corresponding elements of the corre-
lation matrices for the original and revised JDS were performed
using Hotelling's t test of the significance of the difference be-
tween correlations for correlated samples. Of the 30 possible
contrasts, only 2 were significant. Growth satisfaction was more
strongly related to the revised task identity measure than to the
original, Z(221) = 3.86, p < .01; internal motivation was more
strongly related to the original autonomy measure, ((221) =
1.99,/x.OS.
464 C. KULIK, G. OLDHAM, AND P. LANGNER

Table 1
Six-Factor Maximum Likelihood Solution for the Original and Revised Job Diagnostic Survey Items

Factor/Item 1 2 3 4 5 6
Skill variety
1 .66 (.07)
2 .45 (.09) .52 (.15) -.18(.15)
3" .44 (.08) .15 (.09)
4" .79 (.06)
Task significance
5 .80 (.06)
6 .48 (.07)
7* .38 (.10) .40 (.10)
8" .67 (.07)
Task identity
9 -.16 (.16) .31 (.11) .09(.ll) .60(.19)
10 .75 (.06)
11" .61 (.08) .30 (.09)
12" .79 (.06)
Autonomy
13 .54 (.10) .27 (.12)
14 -.26 (.18) .98 (.17) -.11 (.24)
IS" .49 (.10) .40 (.10)
16» .59 (.07)
Feedback
17 .57 (.07)
18 .53 (.07)
19" .52 (.09) .31 (.10)
20b .46 (.07)

Factor Factor correlation matrix

1. Skill variety
2. Task significance .60— —
3. Task identity .46 .52
4. Autonomy .73 .63 .66—
5. Feedback .57 .81 .64 .68— —
6. Method artifact .04 .05 .04 .04 .19 —
Note. N = 224; standard errors of parameter estimates are within parentheses.
" These items are the original Job Diagnostic Survey items that required reverse scoring.
b
These items are rewritten so that reverse scoring is not necessary.

LISREL Analyses
The computer program LISREL (Version 5, Joreskog & Sor-
bom, 1985) was used to obtain maximum likelihood estimates.
It also provides a chi-square statistic and other goodness-of-fit
measures that represent the overall adequacy of the model. We
investigated two structural models. In both, the 15 JDS items
were taken as manifest indicators of a set oflatent independent
variables (the a priori job dimensions). The remaining items
were taken as manifest indicators oflatent dependent variables.
Two parallel LISREL analyses were performed. In both analy-
ses, a 28 X 28 matrix of correlations was input to LISREL (the
15 JDS items and the 13 dependent variable items). In the first
LISREL analysis, the specifications for the factor loading matrix
for the original JDS items were identical to the ones from the
previous confirmatory factor analysis. That is, any factor load-
ings corresponding to original JDS items that were free parame-
ters in the confirmatory factor analysis were free parameters in
the LISREL analysis. In the second analysis, specifications for the
factor loading matrix for the revised JDS items were identical
to those in the previous confirmatory factor analysis.
The 15 JDS items were used as observed measures of the in-
dependent latent constructs. A 5 X 5 submatrix, corresponding
to the correlations among the JDS construct factors, was ex-
tracted from the factor correlation matrix shown in Table 1.
The correlation matrix of the JDS latent variables was set equal
to this 5 X 5 matrix and was held fixed at these values to reduce
the standard errors of elements in the gamma matrix.
Because only single measures of milk production and fresh-
ening interval were available, the two productivity measures
were assumed to have perfect reliabilities, and the parameter
values for the paths from the measures to the productivity con-
structs were set equal to one. Because the other dependent vari-
ables had multiple measures, paths from the first items of each
construct to the construct factor were fixed at one. This sets the
scales of the latent factors equal to the scales of the first items
measuring the constructs (Pedhazur, 1982).
The factor loadings of the individual items on the JDS con-
structs obtained in the LISREL analyses were examined and
found to be virtually identical to those obtained in the confir-
matory factor analysis. Thus, the factor loadings are not re-
ported here, and are available on request from the authors. The
 JOB CHARACTERISTICS 465

Table 2
Correlations Between the Original and Revised Job Diagnostic Survey (JDS) Dimensions and the Outcomes

Outcome

Growth satisfaction Internal motivation Intention to quit

JDS dimension (a = .80) (a = .59) (a = .71) Milk production Freshening interval

Original version

Skill variety
(a = .58) .39** .26** -.27** -.04 .08
Task significance
(« = .53) .35** .35** -.18** -.12* .03
Task identity
(« = .73) .32** .19** -.19** .05 .17*
Autonomy
(a = .68) .53** .30** -.35** .05 .11
Feedback
(« = .60) .38** .32** -.22** .05 .10
Motivating potential score .53** .36** -.34** .05 .12*

Revised version

Skill variety
(a - .74) .42** .26** -.25** -.05 .08
Task significance
(a = .65) .39** .33** -.19** -.13* .06
Task identity
(a = .78) .43** .22** -.22** .04 .16*
Autonomy
(a = .70) .51** .24** -.30** .01 .12
Feedback
(a = .56) .43** .30** -.20** .02 .05
Motivating potential score .57** .33** -.29** .00 .12*

Note. N = 224.
*p<.05.**p<.01.

individual items comprising the dependent variable measures Discussion

all loaded on the appropriate factors. Again, these results are
available on request. This study contrasted the original version of the JDS with a
The chi-square values, associated degrees of freedom, and revised version proposed by Idaszak and Drasgow (1987). Re-
probability levels for the models corresponding to the two ver- sults of a confirmatory factor analysis indicated that the five a
sions of the JDS are presented in Table 3. Also reported in Table priori job dimensions did exist but that there was a sixth factor
3 are three measures of the overall adequacy of each model pro- consisting of the negatively worded items from the original JDS.
vided by LISREL: root mean square residual (RMSR), goodness- These results were generally consistent with those obtained by
of-fit index (GFI), and the adjusted goodness-of-fit index Harvey et al. (1985) and by Idaszak and Drasgow (1987) and
(AGFI). suggest that the negatively worded items may have been the
A comparison of the indexes reported in Table 3 indicates cause of the inconsistent factor analytic results obtained in ear-
that there is little difference between the original and revised lier research (e.g., Dunham, 1976; Dunham etal., 1977).
versions of the JDS in terms of overall fit. The results are infor- Results also indicated that the revised JDS items suggested
mative because they demonstrate that the purer job characteris- by Idaszak and Drasgow (1987) conformed more closely to the
tics measures provided in the revised version really do not im-
prove the JDS's usefulness in predicting several outcomes.
Table 4 shows the gamma matrices for the LISREL analyses of
Table 3
the original and revised JDS. The gamma matrix describes the
Results of LISREL Analyses
pattern of relations between the JDS constructs and the depen-
dent variable constructs. The pattern of path coefficients is sim- Version df X2 Xz/df RMSR GFI AGFI
ilar across the two versions of the JDS. The revised JDS demon- Original 319 464.86* 1.45 0.07 0.85 0.81
strates more significant paths between the skill variety and task Revised 319 503.86* 1.56 0.07 0.84 0.79
significance constructs and the outcome variables. However, the
Note. RMSR = root mean square residual; GFI = goodness-of-fit index;
original JDS resulted in more significant paths involving the AGFI = adjusted goodness-of-fit index.
task identity construct.
466 C. KULIK, G. OLDHAM, AND P. LANGNER

Table 4
Maximum Likelihood Estimates of Gamma Elements

Job diagnostic Growth Internal Intention Milk Freshening

survey dimension satisfaction motivation to quit production interval
Original version
Skill variety -.35 (.11)** -.20 (.12) .23 (.13) -.38 (.17)* -.26 (.16)
Task significance .21 (.13) .16(.14) -.20(.16) -.10 (.20) .11 (.20)
Task identity -.38 (.10)** -.29 (.10)** .29 (.12)* -.36 (.15)* .06 (.14)
Autonomy .91 (.13)** .48 (.13)** -.76 (.14)** .22 (.17) .33 (.16)*
Feedback -.01 (.16) .19 (.17) .18(.19) .49 (.24)* -.11 (.23)

Revised version
Skill variety -.42 (.10)** -.21 (.10)* .30 (.11)* -.41 (.14)** -.25 (.14)
Task significance .29 (.12)* .17(.13) -.28 (.15) .06 (.18) .09 (.18)
Task identity -.22 (.09)* -.20(.ll)* .20 (.09) -.08 (.13) .15 (.13)
Autonomy .85 (.13)** .45 (.12)** -.71 (.14)** .27 (.16) .32 (.16)*
Feedback -,06(.15) .16(.15) .21 (.18) .05 (.20) -.16 (.20)
Note. N= 189. Standard errors of parameter estimates are in parentheses.
*p<.05. **;><.01.

five-factor structure proposed by Job Characteristics Theory
than did the original JDS items. However, results of LISREL
analyses demonstrated that the revised items did not generally
improve the JDS's usefulness in predicting several outcomes.
Although Idaszak and Drasgow (1987) suggested that future
research substitute their items for the original JDS items, our
results suggest that such a substitution may be premature. Not
only are the revised items likely to have little impact on the
overall criterion-related validity of the job characteristics mea-
sures, but the revisions make difficult comparisons with earlier
research efforts using the original JDS.
Therefore, rather than using the revised items, it may be best
to continue using the original JDS until items can be developed
that improve the JDS factor structure and the effectiveness of
the job characteristics in predicting personal and organizational
outcomes. Although the revisions improved the measures of
skill variety, task significance, and task identity, the revised
items did not substantially improve the measurement of auton-
omy and feedback. Future research should focus on developing
alternative autonomy and feedback items and examining their
impact on the factor structure of the JDS and its criterion-re-
lated validity. In addition, future research might examine the
possibility that the JDS outcome measures are subject to the
same impurities as the measures of job characteristics. For ex-
ample, one of the four JDS internal motivation items is reverse
scored, potentially "muddying" the measurement of the con-
struct. Two reviewers suggested that improvements in the criter-
ion-related validity of the JDS may be observed only when both
the job characteristics and outcomes measures are "pure" indi-
cators of their respective constructs.
References
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.
Hackman, J. R., & Oldham, G. R. (1975). Development of the Job Di-
agnostic Survey. Journal of Applied Psychology, 60, 159-170.
Hackman, J. R., & Oldham, G. R. (1980). Work redesign. Reading,
MA: Addison-Wesley.
Harvey, R., Billings, R., & Milan, K. (1985). Confirmatory factor analy-
sis of the Job Diagnostic Survey: Good news and bad news. Journal
of Applied Psychology, 70,461-468.
Idaszak, J. R., & Drasgow, F. (1987). A revision of the Job Diagnostic
Survey: Elimination of a measurement artifact. Journal of Applied
Psychology, 72, 69-74.
Joreskog, K., & Sorbom, D. (1985). LISREL: Analysis of linear struc-
tural relationships by the method of maximum likelihood. Chicago:
National Educational Resources.
Katz, R. (1978). The influence of job longevity on employee reactions
to task characteristics. Human Relations, 8,703-726.
Lee, R., & Klein, A. R. (1982). Structure of the Job Diagnostic Survey
for public health occupations. Journal of Applied Psychology, 67,
515-519.
Pedhazur, E. J. (1982). Multiple regression in behavioral research (2nd
ed.). New York: Holt, Rinehart & Winston.
Pierce, J. L., & Dunham, R. B. (1978). The measurement of perceived
job characteristics: The Job Diagnostic Survey versus the Job Char-
acteristics Inventory. Academy of Management Journal, 21,123-128.
Seabrook, M. F. (19 72). A study to determine the influence of the herds-
man's personality on milk yield. Agriculture Labour Science, 1, 45-
49.
Seabrook, M. F. (1975). The relationship of livestock to human beings
on the farm. Agricultural Engineer, 30,115-119.
Seabrook, M. F. (1978). Managing and motivating the dairy worker.
Journal of the Irish Grasslands Animal Production Association, 13,
77-88.

Colarelli, S. M. (1984). Methods of communication and mediating pro- Received March 25, 1987
cesses in Realistic Job Previews. Journal of Applied Psychology, 69, Revision received February 18, 1988
633-642. Accepted February 2, 1988 •