You are on page 1of 11

Perceived Task Characteristics and Employee Performance: A Literature Review

Author(s): Ricky W. Griffin, Ann Welsh and Gregory Moorhead


Source: The Academy of Management Review, Vol. 6, No. 4 (Oct., 1981), pp. 655-664
Published by: Academy of Management
Stable URL: http://www.jstor.org/stable/257645
Accessed: 07-03-2015 10:39 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

Academy of Management is collaborating with JSTOR to digitize, preserve and extend access to The Academy of Management
Review.

http://www.jstor.org

This content downloaded from 121.52.157.189 on Sat, 07 Mar 2015 10:39:04 UTC
All use subject to JSTOR Terms and Conditions
Academy of Management Review 1981, Vol. 6, No. 4, 655-664

Perceived Task and


Characteristics
Employee Performance:A Literature Reviewl'2
RICKYW. GRIFFIN
Texas A&M University
ANN WELSH
University of Missouri - Columbia
GREGORYMOORHEAD
Arizona State University
We review 13 studiesdealing with empiricalrelationshipsbetweenperceivedtask
scopeand employeeperformance.Resultsfrom thesestudiesare contradictoryand
inconclusive.Most studies use less than adequatemeasuresof employeeperfor-
mance.Moreover,althougha causalrelationshipis assumed,reciprocalor reverse
causality may exist. There is a clear needfor further theoreticalexplicationand
improvedlaboratoryandfield researchaimedat enhancingbothconstructvalidity
and substantiveconsiderations.

Recent reviews of the literature have concluded Hackman and Lawler [1971], research in the area of
that the task design/performance relationship has task design has mushroomed. For the most part,
not been as consistently demonstrated as have task these research efforts have focused on one or more
design/affective response relationships. Our pur- of the following points: (1) employee perceptions of
pose in this review is to examine task/performance their tasks, (2) relationships among these percep-
relationships more closely, with the objectives of tions and organizationally relevant outcome vari-
determining why reported results have been so ables, or (3) possible mediating variables between
inconclusive and suggesting future research direc- employee task perceptions and the outcome vari-
tions. Task design refers to the overall set of work- ables. This formulation of task design interrelation-
related activities performed by an employee; task ships is perhaps best represented by the job charac-
and job characteristicsrefer to specific attributes or teristics theory [Hackman & Oldham, 1976, 1980].
dimensions (e.g., variety, autonomy, feedback, One particularaspect of the formulation that war-
identity, and task significance) that can be used to rants critical attention is the predicted relationship
describe different tasks; and task scopeis the extent between perceived task scope and employee perfor-
to which a job has these various characteristics. mance. The rationale for such a relationship has
been well argued by Richard Hackman and his
Overview of Task Design Research associates. A paper by Hackman and Lawler [1971]
laid the foundation by developing a task design
Since the pioneering research of Turner and
framework from expectancy theory [Vroom, 1964].
Lawrence [1964], Hulin and Blood [1968], and
This initial framework, which suggested that
'We gratefully acknowledge the assistance of Diana Perl in the employees work harder to the extent that their
preparation of this manuscript. individual needs and organizational goals are con-
2Portions of this article were presented at the Annual Meeting of gruent, was more clearly explicated in 1976 by
the American Institute for Decision Sciences, Las Vegas, 1980. Hackman and Oldham. In their job characteristics
? 1981
by the Academy of Management 0363-7425 theory, they stated that all outcome variables, in-
655

This content downloaded from 121.52.157.189 on Sat, 07 Mar 2015 10:39:04 UTC
All use subject to JSTOR Terms and Conditions
cluding quality of work performance, "are expected increased study of the determinants of individual
to be more positive for jobs with high motivating employee performance is important to society in
potential" [p. 259], i.e., tasks with high scope. This general.
assertion is also found in Hackman [1977]. A secondjustification for continued interest in
The most recent presentation of the job charac- performance is more pragmatic. For obvious
teristics model [Hackman & Oldham, 1980] also reasons, many managers are interested in improving
includes performance or work effectivenessas a spe- employee productivity within their organizations.
cific outcome variable. Work effectiveness, in turn, Hence, a contribution that organizational re-
consists of both quality and quantity of output. The searchers could make to practicing managers would
rationale for assuming a quality relationship is that be increased understanding of the causes and con-
"when a job is high in motivating potential, people sequences of employee performance. Further re-
who work on that job tend to experience positive search efforts are needed to discover if we can, in
affect when they perform well"[p. 91]. The quantity fact, make a contribution.
prediction is based on three factors: (1) high-scope
tasks are generally not routine and boring; hence,
Task Design and Performance
employees are not as likely to search for ways of
avoiding work, (2) as a result of task changes aimed Two recent reviews have examined in general
at increasing scope, hidden inefficiences in the terms the relationship between task design variables
work system may be corrected, and (3) such and employee performance postulated by Hackman
changes may simplify and refine the total work and Oldham. Pierce and Dunham summarized 10
system. Hence, the theoretical and conceptual argu- studies involving main effects between task design
ments for predicting a task scope/performance re- and performance and 5 additional studies dealing
lationship would seem to be both reasonable and with interactions among task design and individual
logical. differences and performance. Also included in the
review were studies dealing with other outcome
variables, such as satisfaction and motivation.
Importance of the Performance Variable
Pierce and Dunham concluded that:
Performance is generally discussed within the
contexts of motivation, leader behavior, task The evidence emerging for the main effects in-
design, goal setting, and most other primary areas vestigations suggests that task designs are more
of organizational research. Yet, empirical tests frequently associatedwith positive affective, be-
havioral,andmotivationalresponsesthanarenar-
involving performance and the various areas of rowlydefinedtasks.... Affectiveandmotivational
study are generally disappointing or inconclusive. responses appearto be more strongly related to
The performance variable, however, should main- task design than are behavioralresponses. Satis-
tain its position of primacy in organizational factionwith work is more strongly relatedto task
research for at least two reasons. designthanareotheraffective,behavioral,or moti-
vationalvariables.... Inallbut one of the investiga-
First, the aggregate productivity growth rate in tions reviewed, evidence suggests improvements
the United States has fallen considerably behind in work-relatedbehaviorassociatedwith expanded
that of other industrialized countries such as Japan task design. [1976, p. 87]
and West Germany. For example, in the United
States, the value of goods produced by each worker In a more recent but less thorough review, Mow-
has increased by 62 percent from 1950 to 1977; the day [1978] argues that despite moderate empirical
corresponding increases in Japan and West Ger- support for a task design/performance relationship,
many for the same time period have been 531 and critical questions remain unanswered. He also
256 percent, respectively ["Productivity," 1979]. notes that even where predicted relationships are
Productivity growth is an important factor in sta- found, the results are often so weak as to be of little
bilizing our economy through improved living practicalvalue, or subject to alternative explanation.
standards, higher wages, an increase in goods In summary, the evidence regarding a task
available for consumption, and so forth. Hence, design/performance linkage is still open to debate.
656

This content downloaded from 121.52.157.189 on Sat, 07 Mar 2015 10:39:04 UTC
All use subject to JSTOR Terms and Conditions
The reviews by Pierce and Dunham and Mowday progress or in press that would meet our criteria.
essentially raise more questions than they answer. This effort yielded one unpublished manuscript by
Furthermore, characteristics of each review limit Pierce [1979]. In addition, one of us had completed a
their applicability to the task design/performance study, not yet published, that also met the criteria
question. The Pierce and Dunham paper was a [Griffin, 1981b]. Hence, a total of 13 studies are
review of the total body of task design literature. included in the review.
Hence, specific issues were, of necessity, summar-
ized in brief form. The Mowday paper was prepared Results of the Review
for a symposium at a professional meeting and was,
therefore, superficial in its coverage, owing to The primary characteristics of the 13 studies are
severe space and time constraints. It should also be summarized in Table 1. In the discussion of the
noted that a number of relevant studies have results, studies will be identified as offering sup-
appeared since the last review. Hence, there is a port, no support, or mixed support for a task charac-
clear need for a thorough analysis of the literature teristics/performance relationship. A study in
pertaining to task design and performance. which most main effects relationships were positive
and significant will be classified as supportive, one
in which a few positive and significant main effects
Scope of the Review relationships or significant mediating effects were
We limited the literature review to empirical stu- found will be classified as offering mixed support,
dies relating perceived task characteristics and em- and one in which no main or interactive effects
ployee performance because (1) many studies do were present, as nonsupportive.
not measure task design variables or are otherwise
only minimally acceptable in terms of appropriate Measures of Task Characteristics
research design [Pierce & Dunham, 1976], and (2)
As noted by Pierce and Dunham [1976], the most
the perceived task design framework has become
the primary research model in the area, and it is widely used measure of perceived task characteris-
tics has been the Job Diagnostic Survey (JDS)devel-
therefore useful to approach the task design/per-
formance issue from this particular perspective. oped by Hackman and Oldham [1975]. Of the 7
studies using the JDS, support was found in 3,
Hence, to be included in this review, a study had to
mixed support in 1, and no support in 3. A second
(1) include measures of employee perceptions of
measure often used is the Job Characteristic Inven-
tasks and (2) empirically test the degree of associa-
tion between these task perceptions and some tory (JCI) [Sims, Szilagyi, & Keller, 1976]. Three
studies in Table 1 report use of the JCI. Two of
primary facet of employee performance. these studies provide support for a task character-
Studies to be included were located by two means.
First, all major social science journals published istics/performance relationship; the other offers no
since 1971 were surveyed (we considered the 1971 support. Finally, of the three studies using the Yale
Hackman and Lawler study a logical starting point). Job Inventory [Hackman & Lawler, 1971], 2 pro-
vide mixed support and the other provides no sup-
This approach identified 67 papers. The majority of
these articles were eventually eliminated from the port. Interestingly, no study to date has included
more than one measure of perceived task charac-
review, however, because (1) the article was general-
teristics, even though such research is clearly
ly descriptive as opposed to empirical [e.g., Schwab warranted.
& Cummings, 1976], (2) the study did not measure
task perceptions [e.g., Robey, 1974], or (3) the Measures of Performance
study related task perceptions to affective rather
than behavioral outcomes [e.g., Stone & Porter, Because the criterion variable in question-
1975]. A total of 11 studies met our two criteria. employee performance-is of such importance,
A second search procedure entailed contacting careful analysis of performance measures used in
researchers known to be working on task design the task design studies is essential. These measures
issues and asking them for copies of articles in are summarized in Table 2. Even a casual examina-

657

This content downloaded from 121.52.157.189 on Sat, 07 Mar 2015 10:39:04 UTC
All use subject to JSTOR Terms and Conditions
Table 1
Summary of Studies Relating Task Characteristics to Performance
Measure of
Task Measure of Mediator(s) Type of Sample Nature of
Study Characteristicsa Performancec Testedd Studye Size Samplef

1. Hackman & Lawler [1971] YJI SE GNS FS 208 HE

2. Wanous [19741 YJI SE GNS FS 80 HO


PWE
RU

3. Hackman & Oldham 11976] JDS SE GNS FS 658 HE

4. Oldham, Haickman, & Pearce [1976] JDS SE GNS FS 201 HO

CS

GNS x CS

5. Sims & Szilaagyi [1976] JCI SE GNS FS 766 HE

LC

6. Umstot, Bell, & Mitchell [1976] JDSb DI GNS LE 42 HO

7. Steers & Spencer [1977] YJI SE NA FS 115 HO

8. Evans, Kiggundu, & House [1979] JDS SR GNS FS 343 HO


NA

9. Orpen [1979] JDSb SE GNS FE 72 HO


DI CS

10. White & Mitchell [1979] JDSb DI None LE 41 HO

11. Griffin [1981a] JCI DI GNS LFS 109 HO


160 HO

12. Griffin [1981b] JCI DI GNS FS 80 HO

13. Pierce [1979] JDS SE None FS 397 HO

aJCI= Job Characteristic Inventory, JDS = Job Diagnostic Survey, YJI= Yale Job Inventory
bUsed as a manipulation check
CSE= Supervisory evaluation, SR = Self-rated, DI = Direct measure of quantity and/or quality
dGNS = Growth need strength, CS = Contextual satisfaction, NA = Need for Achievement, PWE = Protestant Work Ethic, RU = Rural/Urban background,
LC = Locus of Control

658

This content downloaded from 121.52.157.189 on Sat, 07 Mar 2015 10:39:04 UTC
All use subject to JSTOR Terms and Conditions
Main Interactive
Effectsg Effectsh

6 of 12 correlations between task characteristics and No support for total sample; modest support only for
performance positive and significant high-scope tasks

Main effects not reported No support for any mediating variable; most correla-
tions non-significant'

4 of 5 task characteristics and MPS scores signifi- Correlations for high and low GNS groups different in
cantly correlated with performance predicted direction but differences not significant

MPS and performance significantly correlated Correlations for high and low GNS groups different in
predicted direction but differences not.significant

Correlations for high and low CS groups different in


predicted direction; 1 of 5 pairs significantly different

Correlations for high GNS/high CS and low GNS/low CS


groups all significantly different and in predicted direction

6 of 6 correlations between task characteristics and 3 of 6 pairs of correlations for high and low GNS groups
performance positive and significant different in predicted direction (1 pair significantly
different); remaining 3 pairs of correlations significantly
different in opposite direction

No mediating effects detected

No main effects between task characteristicslMPS and 4 of 6 pairs of correlations for high and low GNS groups
performance detected different in predicted direction but differences not
significant; autonomy significantly correlated with
performance for high GNS group
No significant correlations between task characteristics 5 of 6 pairs of correlations for high and low Nach groups
and performance significantly different in predicted direction
5 of 7 correlations between task characteristics/MPS No mediating effects detected
and performance positive and significant

No significant correlations between task characteristics No mediating effects detected


and performance

No significant relationships found Not applicable

No significant correlations between task characteristics No mediating effects detected


and performance

3 of 4 task characteristics significantly correlated with No mediating effects detected


performance

Significant R obtained with 3 task characteristics regressed Not applicable


on performance

eFS = Field survey, FE = Field experiment, LE = Lab experiment, LFS = Longitudinal field survey
fHE = Heterogeneous, HO = Homogeneous
gLinear relationship between perceived task characteristics and performance
hExtent to which predicted mediator effects were found
'Specific results not presented in article; interpretation based on author's discussion of results

659

This content downloaded from 121.52.157.189 on Sat, 07 Mar 2015 10:39:04 UTC
All use subject to JSTOR Terms and Conditions
tion of Table 2 reveals that performance has been ations of employee performance are well docu-
measured and defined in a variety of ways. Super- mented [Cummings & Schwab, 1973; Heneman,
visory ratings have been used in 8 studies, objective 1974; Klimoski & London, 1974]. Specifically, such
measures in 5, and self-ratings in 1 (the number of ratings often suffer from problems of leniency,
studies discussed in this section will add to 14, restriction of range, or halo effects. These factors
rather than 13, because Orpen [1979] used 2 represent error variance in the performance evalua-
measures). tions. Because measures of perceived task charac-
The problems associated with supervisory evalu- teristics are also characterized by a certain amount

Table 2
Summary of Performance Measures Used in Task Design Research

Study Performance Labela Method of Measurement

1. Hackman & Lawler [1971] Performance Quantity and quality of work produced measured by 7-point
rating scales completed by supervisor(s); overall performance
effectiveness measured by summing supervisory ratings
across seven separate performance scales (quantity and
quality plus five unspecified variables); all 3 measures used in
data analysis

2. Wanous [1974] Performance Supervisory ratings of employee's quality and quantity of per-
formance were summed and called job performance

3. Hackman & Oldham [1976] Work Performance Supervisory ratings (7-point scales) of effort expended on the
job, work quality, and work quantity were summed and called
work effectiveness

4. Oldham, Hackman, & Pearce Performance Supervisory ratings (5-point scale) of performance effective-
[1976] Effectiveness ness; salary

5. Sims & Szilagyi [1976] Performance Supervisory ratings (5-point scales) of nine performance
dimensions; results either summed or averaged and called job
performance

6. Umstot, Bell, & Mitchell [1976] Job Performance Average number of units produced per hour for one day;
adjusted for set-up and called productivity

7. Steers & Spencer [1977] Performance Supervisory ratings (5-point scale) of overall performance rela-
tive to performance of others

8. Evans, Kiggundu, & House [1979] Performance Self-ratings (5-point scales) of relative performance were
averaged

9. Orpen [1979] Job Performance Supervisory ratings (8-point scale) of general competence

Productivity Average daily output (i.e., quantity) over 6 months

10. White & Mitchell [1979] Productivity Number of units produced divided by total time worked
(i.e., items per minute)

11. Griffin [1981a] Productivity Average daily output (quantity adjusted for set-up time, down
time, and scrappage) for a 10-day period

12. Griffin [1981b] Productivity Average daily output (quantity adjusted for set-up time, down
time, and scrappage) for a 12-month period

13. Pierce [1979] Performance Supervisory ratings (3-point scale) of overall job performance

aRefers to the term used by the author(s) to reflect employee behavior on the job
660

This content downloaded from 121.52.157.189 on Sat, 07 Mar 2015 10:39:04 UTC
All use subject to JSTOR Terms and Conditions
of error variance, analysis of these data involve high and low task-scope groups were 4.37 and 4.36,
correlating two sets of measures, each of which respectively. One explanation for such similarity
contains an unknown amount of variance not might be that the manner in which the performance
attributable to the phenomena at issue. To illus- measure was constructed resulted in outcomes that
trate the point rather crudely, let us assume that were independent of task characteristics effects. In-
each measure is characterized by 80 percent true deed, the results of the study support this possible
variance (i.e., variance attributable to construct explanation. The artificial nature of the lab setting
validity) and 20 percent error variance. It follows may have confounded experimental effects to the
that the computed correlation may be somewhat extent that the measure of performance lacked
higher or lower than the "real"correlation between validity.
the variables by some proportional amount of error Apparent weaknesses associated with the per-
variance. formance measure are found in a study by Griffin
This point becomes critical when the studies sum- [1981a]. This study used a quasi-experimental
marized in Table 2 are considered. Of the 8 studies design (cross-lagged panel correlations) in a field
using supervisory evaluations of performance, 3 setting. The organization that hosted the study
offer support for a task design/performance rela- generated a daily index of productivity (quantity of
tionship, 3 offer mixed support, and 2 offer no sup- output adjusted for scrappage, down-time, and set-
port. Yet, most of the correlations reported in all 8 up time). Griffin defined performance as a 10-day
studies cluster around the levels needed to attain average of these daily productivity indices. No sup-
significance. Correlations tend to be significant at port was found for a task design/performance rela-
the .05 level, but seldom at or beyond the .01 level. tionship. A potential explanation for this finding is
On the other hand, many of the nonsignificant that a 10-day sampling was not sufficient to reflect
correlations actually approach a level of minimal long-term employee performance.
significance. Hence, if it were somehow possible to The other two studies using direct measures of
extract the error variance from either or both of the performance offer frustratingly contradictory evi-
measures, the results might be more conclusive. dence. In probably the best-designed task study to
This problem of error variance is partially over- date, Orpen [1979] measured performance by both
come in those studies using direct indices of perfor- supervisory evaluations and a direct index of out-
mance. Although a direct measure of performance put. He found no relationships at all between task
may also be deficient in some ways [Cummings & variables and performance. Because of the nature of
Schwab, 1973], it would presumably have less error this study, considerable confidence in the findings
variance than a supervisory rating. Of the 5 studies may be warranted.
using a direct measure of performance, 1 offers sup- However, a second study by Griffin [1981b] also
port and 4 offer no support for a task designlperfor- appears to have a reasonable amount of validity. In a
mance linkage. Closer scrutiny, however, indicates different plant within the same organization where
that the direct index used in 3 of the studies may not the other Griffin study was conducted, a follow-up
adequately represent true employee performance. was completed approximately one year later. In the
Umstot, Bell, and Mitchell operationalize perfor- follow-up, employee performance was defined as
mance as the average hourly output for 4 hours per the daily productivity index averaged for an entire
day for 2 days. Mean values of productivity, how- year. This time, task perceptions and performance
ever, increased for all groups in the experiment were positively and significantly correlated. Hence,
[1976, p. 391], suggesting that improvements may this finding reinforces the previous contention that
have been at least partially attributable to increased the original 10-day sampling was not representative.
proficiency resulting from the previous day's ex- In summary, then, results appear to be contradic-
perience. Therefore, the validity of the performance tory. One important point that should be noted,
index used by Umstot et al. may be suspect. however, is that employee performance has not
In another lab experiment, White and Mitchell been measured in exactly the same way in even two
[1979] also used a direct measure of performance: studies. Consequently, even if results were clear-
the number of items completed per minute during a cut, we still would be hard pressed to make valid
90-minute period. Mean values of productivity for generalizations.
661

This content downloaded from 121.52.157.189 on Sat, 07 Mar 2015 10:39:04 UTC
All use subject to JSTOR Terms and Conditions
Mediators potentially invalid and meaningless. Second, indi-
The most commonly tested mediating variable, vidual difference variables have not been shown to
growth need strength (GNS), has been investigated be a major factor in the task design/performance
in 10 studies. Contextual satisfaction and need for relationship. Perhaps one primary explanation for
achievement have been investigated twice, and sev- problems of interpretability is that measurement
eral other variables only once each. Overall, results deficiencies are of such magnitude that they consid-
are not particularly supportive of the individual dif- erably confound the results reported. Hence, a
ference perspective. Although some studies found clear-cut outline of future research needs is impor-
differences in the predicted directions, few differen- tant if the issue is to ever to be resolved.
ces were statistically significant. Therefore, little
Future ResearchNeeds and Directions
support exists for an interactive relationship be-
tween task and individual variables and employee First, a more precise formulation of task design
performance. (Another issue pertaining to mediat- and performance interrelationships is needed. The
ing variables relates to statistical problems in data performance variable itself needs to be more fully
analysis. Fordetails, see Peters & Champoux [1979].) explicated in terms of its various components
(quantity, quality, effort, etc.) and appropriate
Type of Study techniques for measuring these components de-
scribed. Each component should then be tied to
Ten of the 13 studies in Table 1 are field surveys.
Of these surveys, 5 offer support, 3 mixed support, specific task characteristics from a theoretical point
of view. For example, feedback seems likely to have
and 2 no support for a task design/performance
a positive effect on work quality, whereas the
relationship. The remaining 3 studies are experimen- effects of identity on quantity might be less pro-
tal in nature, and all offer no support. In summary,
nounced. (These examples are pure conjecture and
then, reasonable support is obtained from field sur- are offered only to illustrate the point.) What is
veys, but not from experimental studies. Implica- needed is a theoretical framework linking task
tions of this observation will be discussed later.
scope and performance, with each variable viewed
from a multidimensional perspective.
Sample Characteristics Second, causal priorities among variables need
Sample sizes range from 41 to 766, and are char- both theoretical and empirical clarification. In gen-
acterized by extremes of both heterogeneity and eral, the presumption is made that task design is an
homogeneity. No discernible pattern of results can independent variable and that performance is a
be detected when studies are compared along these dependent variable. However, an alternate expla-
dimensions. The only observation of potential im- nation might be that the linkages are reversed. For
portance is that 10 of the 13 samples are essentially example, high-performing employees may be given
homogeneous. Of the 3 studies based on hetero- more autonomy by their supervisors. The em-
geneous samples, all provide at least mixed support ployees, in turn, may not be able to differentiate
for a task design/performance relationship. How- between sources of autonomy and, therefore, may
ever, since heterogeneous data bases might be char- perceive more task autonomy. Similarly, by focus-
acterized as having more error variance, these ing more attention on the task itself as a necessary
results must be interpreted with caution. antecedent for high performance, employees may
develop a better feel for how the task relates to
Summary others and may perceive more identity as a result.
One of the stated objectives of this review was to This possibility of reverse causality is supported by
determine why research on the task design/perfor- the fact that many of the cross-sectional field sur-
mance linkage has been so inconclusive. Based on veys summarized in Table 1 provide evidence for
this examination of the research literature, two correlational relationships, whereas the experi-
generalizations appear to be warranted. First, the mental studies do not.
available studies, with few exceptions, are charac- Third, organization context variables [Oldham &
terized by performance measures that are at best Hackman, 1980] must be integrated with the study
only moderately valid and meaningful and at worst of task design variables if meaningful proportions
662

This content downloaded from 121.52.157.189 on Sat, 07 Mar 2015 10:39:04 UTC
All use subject to JSTOR Terms and Conditions
of performance variance are to be explained. It is specify the nomological network [Schwab, 1980] of
unlikely that task design, or any other narrowly each construct.
defined set of organizational variables, ever ac- Once these two steps have been taken, attention
counts for more than a comparatively small can then appropriately be refocused on substantive
amount of performance variance. More likely, per- research issues. Specifically, the extent of covaria-
formance is a function of a complex set of individual tion and direction of causality among the two nomo-
variables (e.g., motivation, experience, ability), logically described constructs can then be investi-
group variables (e.g., performance norms, cohe- gated with a more precise scientific foundation and
sion), and organizational variables (e.g., task a clearer understanding.
design, structure, technology). This call for inte- From a research point of view, future directions
gration is not new; others have been arguing the then become much more obvious. Validation work
same point for years. Pierce and Dunham [19761 on both the task design and performance constructs
identified integrative research as an important would involve field and laboratory research aimed
need. We concur. at (1) identifying and describing relevant attributes
These three issues are all related to construct of each, (2) explicating interactions among the
validity. As noted recently by Schwab [1980], a use- attributes, and (3) developing better devices and
ful distinction can be made between substantive techniques for measuring each construct. Con-
research and construct validation research. Schwab comitantly, better understanding of the nomologi-
also accurately notes that "substantive relative to cal networks could be gained by: (1) theoretical
construct validation research has been overempha- analysis and (2) laboratory and cross-sectional field
sized in organizational behavior" [p. 4]. The task studies aimed at determining other covariates.
design literature is no exception. All of the studies The return to substantive research could progress
reviewed earlier fall into the substantive category; through three phases. Phase 1 would involve rigor-
few construct validation studies are available. Yet, ous, tightly controlled laboratory experimentation
construct validation must clearly precede substan- aimed at (1) determining directions of causality and,
tive research if knowledge is to be advanced. (2) determining interconstruct attribute combina-
Construct validation of the task design/perfor- tions. Phase 2 would focus on cross-sectional field
mance issue should focus on theoretical explication, surveys in order to establish the external validity of
clarification, and measurement of multidimensional laboratory findings. Phase 3 would include experi-
conceptualizations of tasks and of essential em- mentation in the field to determine the practicaland
ployee task-related behaviors. Next, each construct scientific significance of the task design construct in
should be framed within its appropriate organiza- influencing employee performance. If organiza-
tional context. That is, relationships among task tional behavior as a field of research is to mature,
design and performance constructs and other rele- there must be movement toward research pro-
vant organizational variables should be assessed grams rather than isolated studies, and professional
(theoretically and then empirically). In the nomen- and organizational support for the integration of
clature of construct validity, it is imperative to research efforts.

REFERENCES
Cummings, L.L.; & Schwab, D.P. Performancein organizations: Griffin, R.W. Perceived task characteristicsand employee pro-
Determinants and appraisal. Glenview, Ill.: Scott, Foresman, ductivity and satisfaction. Paper presented at the annual meet-
1973. ing of the Academy of Management, San Diego, 1981. (b)

Evans, M.G.; Kiggundu, M.N. & House, R.J. A partial test and Hackman, J.R. Work design. In J.R. Hackman & J.L. Suttle
extension of the job characteristic model of motivation. Or- (Eds.), Improving life at work: Behavioral science approachesto
ganizational Behavior& Human Performance,1979, 24, 354-381. organizationalchange. Santa Monica, Calif.: Goodyear, 1977.
Hackman, J.R.;& Lawler, E. Employee reactions to job charac-
Griffin, R.W. A longitudinal investigation of task characteris- teristics. Journal of Applied Psychology,1971, 55, 259-286.
tics relationships. Academy of Management Journal, 1981, 24,
99-113. (a) Hackman, J.R.;& Oldham, G.R. Development of the Job Diag-
663

This content downloaded from 121.52.157.189 on Sat, 07 Mar 2015 10:39:04 UTC
All use subject to JSTOR Terms and Conditions
nostic Survey. Journalof Applied Psychology,1975, 60, 159-170. 1978, 2 1, 123-128.

Hackman, J.R.; & Oldham, G.R. Motivation through the de- Pierce, J.L. Technologyandjobdesign:An exploratoryexamination
sign of work: Test of a theory. Organizational Behavior & of employeeabsenteeism,performance,and work satisfaction. Un-
Human Performance,1976, 16, 250-279. published manuscript, University of Minnesota - Duluth, 1979.

Hackman, J.R.; & Oldham, G.R. Work redesign. Reading, Productivity: An elusive essential. Chicago Tribune, Sep-
Mass.: Addison-Wesley, 1980. tember 16, 1979, p. 18.

Heneman, H.G. Comparisons of self- and superior-ratings of Robey, D. Task design, work values, and worker response: An
managerial performance. Journal of Applied Psychology,1974, experimental test. Organizational Behavior & Human Perfor-
59, 638-642. mance, 1974, 12, 264-273.

Hulin, C.L.; & Blood, M.R. Job enlargement, individual differ- Schwab, D.P. Construct validity in organizational behavior.
ences, and worker responses. PsychologicalBulletin, 1968, 69, In B.M. Staw & L.L. Cummings (Eds.), Researchin organizational
41-55. behavior(Vol. 2). Greenwich, Conn.: JAI Press, 1980.

Klimoski, R.J.;& London, M. Role of the rater in performance Sims, H.P.; & Szilagyi, A.D. Job characteristic relationships:
appraisal. Journal of Applied Psychology,1974, 59, 445-451. Individual and structural moderators. Organizational Behavior
& Human Performance,1976, 17, 211-230.
Mowday, R.T. Task characteristics and behavioral outcomes.
Proceedingsof theAmericanInstituteforDecisionSciences,1978, 168. Sims, J.P.; Szilagyi, A.D.; & Keller, R.T. The measurement of
job characteristics. Academy of ManagementJournal, 1976, 19,
Oldham, G.R.; & Hackman, J.R. Work design in the organiza- 195-212.
tional context. In B.M. Staw & L.L.Cummings (Eds.), Research
in organizational behavior (Vol. 2). Greenwich, Conn.: JAI Steers, R.; & Spencer, D. The role of achievement motivation
Press, 1980. in job design. Journalof Applied Psychology,1977, 62,472-479.

Oldham, G.R.; Hackman, J.; & Pearce, J. Conditions under Stone, E.F.; & Porter, L.W. Job characteristics and job atti-
which employees respond positively to enriched work. Journal tudes: A multivariate study. Journal of Applied Psychology,
of Applied Psychology,1976, 61, 395-403. 1975, 60, 57-64.

Orpen, C. The effects of job enrichment on employee satisfac- Turner, A.; & Lawrence, P. Industrial jobs and the worker.
tion, motivation, involvement, and performance: A field experi- Cambridge: Harvard University Press, 1965.
ment. Human Relations, 1979, 32, 189-217.
Umstot, D.; Bell, C.; & Mitchell, T. Effects of job enrichment
Peters, W.S.; & Champoux, J.E. The role of moderator vari- and task goals on satisfaction and productivity: Implications for
ables in organizational research. In R.T. Mowday & R.M. job design. Journal of Applied Psychology,1976, 61, 379-394.
Steers (Eds.), Research in organizations:Issues and controversies.
Santa Monica, Calif.: Goodyear, 1979. Vroom, V.H. Work and motivation. New York: Wiley, 1964.

Pierce, J.L.;& Dunham, R.B. Task design: A literature review. Wanous, J. Individual differences and reactions to job charac-
Academy of Management Review, 1976, 1(4), 83-97. teristics. Journal of Applied Psychology,1974, 59, 616-622.

Pierce, J.L.; & Dunham, R.B. The measurement of perceived White, S.E.; & Mitchell, T.R. Job enrichment versus social
job characteristics: The Job Diagnostic Survey versus the Job cues: A comparison and competitive test. Journal of Applied
Characteristics Inventory. Academy of Management Journal, Psychology,1979, 64, 1-9.

Ricky W. Griffin is Assistant Professor of Management in the


College of Business Administration of Texas A&M University,
CollegeStation.
Ann Welsh is a doctoralcandidatein the Departmentof Manage-
ment, Collegeof Businessand Public Administration, University of
Missouri at Columbia.

Gregory Moorhead is Assistant Professorof Management in the


College of Business Administration of Arizona State University,
Tempe.
Received 4/7/80
664

This content downloaded from 121.52.157.189 on Sat, 07 Mar 2015 10:39:04 UTC
All use subject to JSTOR Terms and Conditions

You might also like