Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Journal of Applied Psychology1989, Vol. 74, No. 5,739-751
Copyright 1989 by the American Psychological Association, Inc0021-9010/89/$00.75
Integrating Work Environment Perceptions:Explorations into the Measurement of Meaning
Lois A. James and Lawrence R. JamesGeorgia Institute of Technology
Many of the perceptual variables used in industrial/organizational psychology assess the meaningthat work environment attributes have for individuals (e.g., the ambiguity of role prescriptions).This study represents an initial attempt to test the hypothesis that a unifying theme exists for inte-grating diverse measures of meaning. The unifying theme is based on a hierarchical cognitive model
wherein each assessment of meaning reflects a general appraisal of the degree to which the overallwork environment is personally beneficial versus personally detrimental to the organizational well-
being of the individual. Results of confirmatory factor analyses on multiple samples supported ahierarchical cognitive model with a single, general factor underlying measures of meaning. Theseresults are used to explain the substantive impact of work environment perceptions on individual
outcomes.
Industrial/organizational (I/O) psychologists use many vari-
ables to assess perceptions of work environments. Examples in-
clude perceptual indicators of job attributes (e.g., job challenge,
job autonomy), characteristics of leaders and leadership pro-
cesses (e.g., leader consideration and support, leader work facili-
tation), workgroup characteristics and processes (workgroup
cooperation, workgroup esprit), and interfaces between individ-
uals and subsystems or organizations (e.g., role ambiguity, fair-
ness and equity of reward system). The following two principles
have guided many applied psychologists' efforts to measure
work environment perceptions: (a) Individuals respond to envi-
ronments in terms of how they perceive them and (b) the most
important component of perception is the meaning or mean-
ings imputed to the environment by the individual (Ekeham-
mer, 1974; Endler & Magnusson, 1976; Lewin, 1938, 1951;
Mischel, 1968).
The Concept of Meaning
Our intention is to explore the concept of meaning in the
context of work environments. We begin with basics by noting
that the attribution of meaning, or a meaning analysis, refers to
the use of stored mental representations or schemas (i.e., beliefs
that are products of learning and experience) to interpret (i.e.,
to make sense of) stimuli; in this case, work environment attri-
butes (e.g., events, objects, processes, structures; cf. Jones &
Gerard, 1967; Mandler, 1982; Stotland & Canon, 1972). The
objective of a meaning analysis is often to describe what is "out
there" in the environment (Mandler, 1982, p. 8). This form of
cognition focuses on perceptions of the presence or absence of
features and structures of environmental attributes. The em-
phasis on description is reflected in the use of designations such
Correspondence concerning this article should be addressed to LoisA. James, School of Psychology, Georgia Institute of Technology, At-lanta, Georgia 30332.
as descriptive cognition and descriptive meaning (Mandler,
1982), cold cognition (Zajonc, 1980), descriptive schemas
(Stotland & Canon, 1972), and denotative meaning (Osgood,
Suci, & Tannenbaum, 1957). Examples of variables from the
I/O literature that reflect primarily descriptive meaning include
technological complexity, span of control, centralization of de-
cision making, functional specialization, physical space charac-
teristics (temperature, lighting, sound), formal communication
networks, and formal rules, regulations, and reward structures.
Information processing may proceed beyond describing what
is "out there" to a valuation of environmental attributes
(Mandler, 1982). Valuation refers to cognitive appraisals of en-
vironmental attributes in terms of schemas derived from values
such as equity, freedom from threat, and opportunity for gain
(Mandler, 1982). In comparison with descriptive meaning, val-
uation is more internally oriented and requires additional infor-
mation processing to judge how much of a value is represented
in or by (perceived) environmental attributes (e.g., how much
equity is represented by a pay raise, how much challenge is pres-
ent in a set of job tasks, how much friendliness is represented in
interactions with coworkers; cf. Jones & Gerard, 1967; Mandler,
1982; Stotland & Canon, 1972). The subjective, value-based
meanings furnished by the valuation process are variously re-
ferred to as evaluative meaning (Mandler, 1982), affective (con-
notative) meaning (Osgood, May, & Miron, 1975; Osgood et
al., 1957), emotionally relevant cognitions or simply emotional
cognitions (Reisenzein, 1983; Schacter & Singer, 1962), and
cognitive appraisals (Lazarus & Folkman, 1984). Each of these
designations points to the central role played by valuation in
causal models of emotion and well-being.
To review this role briefly, Lazarus and his colleagues (Laza-
rus, 1982, 1984; Lazarus & Folkman, 1984) viewed emotional
responses to environmental attributes as functions of cognitive
appraisals of the significance of these attributes for one's well-
being. Valuation appears to be the key to such cognitive apprais-
als inasmuch as (a) values serve as standards for assessing wel-
fare (Locke, 1976), where welfare is defined in terms of a sense
739
740 LOIS A. JAMES AND LAWRENCE R. JAMES
of well-being (Mirriam-Webster, 1975); and (b) valuation pro-
vides appraisals of the degree to which these standards are rep-
resented in environmental attributes. The cognitive appraisals
thus described are viewed as "emotional cognitions" (Reisen-
zein, 1983; Schacter & Singer, 1962) because they reflect the
subjective meanings that, in combination with perceived physi-
ological arousal, help to label the emotion and to determine the
direction and intensity of the experienced emotion. Finally, and
most importantly, a single latent component is believed to be
shared by all emotionally relevant cognitions. It is this single
latent component, or g-factor, that most directly furnishes the
emotional cognitions with the facility to assess significance for
well-being. As proposed by Lazarus (1982, 1984), this g-factor
comprises a higher order schema for appraising the degree to
which the environment is personally beneficial versus person-
ally detrimental (damaging or painful) to the self and therefore
to one's well-being (Lazarus, 1982, 1984).
It is noteworthy that perceptions may vary between those
conveying primarily descriptive meaning and those conveying
primarily valuations (Mandler, 1982). We are interested in the
latter because we wish to understand the psychological pro-
cesses linking cognitions of work environments to affect and,
ultimately, to behavior. In pursuit of this objective, we shall now
apply the theoretical perspective to measures reflecting valua-
tions of work environment attributes, with special attention
given to the hypothesis that a general factor underlies emotional
cognitions.
A Hierarchical Model of Emotional Cognitions
in Work Environments
Examples of emotional cognitions are easily found in I/O
psychology. These are constructs that provide cognitive apprais-
als of work environment attributes in terms of schemas engen-
dered by work-relevant values. Based on Locke (1976), a work-
relevant value is denned as that which one desires or seeks to
attain because it is that which one regards as conducive to one's
sense of organizational well-being. Sense of organizational well-
being is defined as the degree to which positive affect exceeds
negative affect in regard to one's overall work experiences (this
is a focused definition based on overall well-being; see Brad-
burn, 1969; Diener, 1984). Values serve as standards for assess-
ing organizational well-being and may be expressed in quite spe-
cific terms, such as one desires a $ 10,000 raise. However, a po-
tentially large number of specific expressions are believed to be
manifestations of a much smaller number of latent or psycho-
logical values (Locke, 1976). A nonexhaustive but important
set of latent values appears to include desires for (a) clarity, har-
mony, and justice; (b) challenge, independence, and responsibil-
ity; (c) work facilitation, support, and recognition; and (d)
warm and friendly social relations (Locke, 1976, p. 1329).
Psychological climate (PC) furnishes perhaps the most
readily identifiable set of variables in I/O psychology for ap-
praising work environments in terms of schemas based on these
latent values (see PC variables in Table 1). This is not a chance
occurrence; the PC variables were designed to assess work envi-
ronments as they are "cognitively represented in terms of their
psychological meaning and significance to the individual"
(James, 1982, p. 219). Indeed, items and scales for PC variables
were developed, using interviews, observations, and literature
reviews, to obtain empirical indicators of what we now refer to
as valuations or emotional cognitions (see James & Sells, 1981,
and Jones & James, 1979, for reviews of the development of the
PC measures). It follows that we would expect a congruence
between the four latent values described previously and the em-
pirically derived factors underlying the PC variables in Table 1.
Such a congruence is evident. Four PC factors that have demon-
strated factorial invariance over diverse work environments are
(a) Role Stress and Lack of Harmony, (b) Job Challenge and
Autonomy, (c) Leadership Facilitation and Support, and (d)
Workgroup Cooperation, Friendliness, and Warmth (see Table
2 and James & Sells, 1981).
The four PC factors above were derived by exploratory factor
analyses, which included orthogonal rotations. As shown later
in Table 2, the PC variables tended to cluster within a priori
domains denned by an environmental referent (e.g., all leader-
ship PC variables loaded on a single factor). These results sug-
gest the somewhat obvious notion that specific environmental
attributes (e.g., workgroup members) are related most directly
to specific values (e.g., friendliness and warmth). At a more psy-
chological level of explanation, the results also suggest that indi-
viduals tend to separate emotional cognitions pertaining to
jobs, leaders, workgroups, and individual/organizational inter-
faces into separate internal compartments.
However, compartmentalization need not imply the lack of
relations among the factors indicated by orthogonal rotation.
The preceding discussion furnishes two related bases for pro-
posing that the PC factors are related. First, each PC factor, and
thus each PC variable, has in common the judgmental process
of a cognitive appraisal of an environmental attribute or attri-
butes in terms of that attribute's significance to the well-being
of the individual. To be specific, each PC variable furnishes in-
formation pertaining to the efficacy of the perceived environ-
ment to provide for the achievement of a desired standard or
standards and thereby to promote or to detract from the welfare
of the individual in the work environment. By efficacywe mean
facility to produce an intended effect (Mirriam-Webster, 1975).
For example, perceptions that a job is challenging are believed
to be efficacious to personal welfare because increasing levels of
perceived job challenge have been shown to produce higher lev-
els of job satisfaction (James & Jones, 1980; James & Tetrick,
1986). Second, judgments of significance to well-being—that
is, efficacy to promote or detract from personal welfare—are
hypothesized to be based substantively on a single, higher order
schema or general factor, which is the degree to which the indi-
vidual believes that membership in this work environment is
personally beneficial versus personally detrimental to his or her
organizational well-being. This hypothesized general factor not
only implies that the four PC factors are correlated but also
provides a basis for integrating the PC factors and explaining,
at least in part, the substantive impact of perceived work envi-
ronments on individuals.
Our objective is to test the hypothesis that a general factor
underlies PC perceptions. We propose that each of the four PC
factors reflects a cognitive appraisal of the degree to which the
overall work environment is believed to be personally beneficial
versus personally detrimental to the organizational well-being
of the individual. This prediction presupposes a hierarchical,
MEASUREMENT OF MEANING 741
workgroup Cooperation,Warmth and Friendliness
Figure 1. A hierarchical model of meaning.
integrative model of PC perceptions wherein "personal benefit
versus personal detriment to organizational well-being" serves
as a single, higher order, general factor. Such a model is graphi-
cally displayed in Figure 1, where the general factor is desig-
nated PCg (i.e., general factor of psychological climate). The
causal arrows extending from PQ to each of the PC factors de-
note that PC8 is a latent, psychological common denominator
for the factors. The substantive interpretation of these causal
arrows is that perceptions reflecting valuations of individual/
organizational interfaces (Role Stress and Lack of Harmony),
job characteristics (Job Challenge and Autonomy), leaders
(Leadership Facilitation and Support), and workgroups (Work-
group Cooperation and Friendliness) have in common the la-
tent, cognitive appraisal of the degree to which the overall work
environment is personally beneficial versus personally detri-
mental to the organizational well-being of the individual.
Empirical Tests of the Model
Confirmatory factor analysis was used to test the goodness of
fit of the model in Figure 1 on data from four different work
environments. The key assumptions subjected to empirical
tests were as follows:
1. The 4 first-order PC factors are correlated.
2. The correlations among the 4 first-order PC factors may
be explained by a single general factor, which we have desig-
nated PCg.
3. PCB explains reasonable amounts of variance in the initial,
manifest variables used to measure PC (i.e., the PC variables in
Table 1).
Issues pertaining to covariation between PCg and affect, and
potential contamination due to systematic bias in the method
of measurement, are also addressed.
Method
Sample
Sampling was designed to include individuals and work environments
representative of varying levels of technology and of civilian versus mili-
tary settings. The samples were drawn from the data sets used by James
and Jones (1980) and James, Hater, and Jones (1981). The four samples
of nonsupervisory personnel were as follows: (a) aircraft maintenance
personnel from two Navy Air Training Commands located in the south-
ern portion of the United States (N = 422). Jobs varied across a fairly
wide range of technologies, from the very routine (e.g., fueling aircraft)
to the very complex (e.g., repairing sophisticated equipment). Average
age was 22.2 years, average tenure was 3.1 years, and average education
was 12 years (12 years of education represents completion of high
school), (b) Systems analysts and programmers from an information-
systems department of a large, western, private health care program.
Average age was 36.5 years, average tenure was 1.5 years, and average
742 LOIS A. JAMES AND LAWRENCE R. JAMES
education was 15 years (N = 128). (c) Front-line firefighters from a large
metropolitan fire department. Average age was 37.5 years, average ten-
ure was 7 years, and average education was 12.6 years (N = 288). (d)
Production-line personnel from four plants of a paper-product manu-
facturing organization. Average age was 34.1 years, average tenure was
3.5 years, and average education was 11.2 years (Ar = 208).
Instruments
All data included in this study were collected by means of question-
naires administered to the nonsupervisory personnel on a confidential
basis. All items were measured on 5-point Likert scales or Likert-type
scales. The data sets involved measures of the perceived work environ-
ment and variables used for additional construct validation analyses. A
brief description of each data set follows.
Psychological climate. The latest version of the psychological climate
(PC) inventory developed by James, Jones, and colleagues was used to
measure perceived work environment variables (cf. James & Sells,
1981). Seventeen PC variables and accompanying items were common
to the four studies. One additional PC variable, esprit de corps, was
included for Navy personnel. Each of these variables represents a com-
posite of 3-11 items designed to measure a perceptual construct. Co-
efficient alpha for each PC variable in each of the four samples is re-
ported in Table 1. Scores on the variables are means taken over items.
In general, the coefficient alphas are acceptable for research purposes.
It is more important, as we shall see, that the modest values in some
cases do not appear to influence overall results of the analyses.
Variables used in additional construct validation analyses. As ex-
plained in the next section of this article, the variables added to the
analysis were overall job satisfaction and personality indicators of
achievement motivation, rigidity, and self-esteem. The overall job satis-
faction variable was based on the Minnesota Satisfaction Questionnaire
(Weiss, Dawis, England, & Lofquist, 1967), with minor revisions to in-
sure that (a) the items were applicable for all samples, (b) all a priori
factors of PC were represented in the item domains (i.e., satisfaction
with aspects of jobs, individual/organizational interfaces, groups, and
leaders). The same 20 items were used in all studies. An assessment of
general satisfaction was used as a manifest indicator of overall organiza-
tional well-being. This variable was based on a composite of the 20
items, for which coefficient alpha was .90 for Navy personnel, .91 for
systems analysts, .91 for production-line personnel, and .89 for fire-
fighters.
In regard to the personality variables, the item-composites used to
measure achievement motivation (13 items, alpha = .77 on the total
sample) and rigidity (12 items, alpha = .74 on the total sample) were
the same as those described in James, Gent, Hater, and Coray (1979).
The achievement motivation item-composite was designed to assess ori-
entation toward success and included items that measured need for
achievement, aspiration level, persistence, and preference for achieve-
ment-related activities. The rigidity item-composite was based on a
need for certainty and included items such as "I don't like things to be
uncertain or unpredictable" and "I like to have everything organized
before I start a task." The measure for self-esteem was a general manifes-
tation of self-esteem and general self-confidence in the work setting (11
items, alpha = .68 on the total sample) and included items such as "I
take a positive attitude toward myself and "My ability gives me an
advantage in my current job" (cf. James & Jones, 1980).
Analytic Procedure
The goodness of fit of the proposed hierarchical model in Figure 1
was tested by the confirmatory factor-analytic program for higher order
latent variables contained in LISREL vi (Joreskog & Sorbom, 1986). The
17 PC variables (18 PC variables for Navy personnel) served as manifest
Table 1
Classification of the Item Composites by the Four
Psychological Climate (PC) Domains and Their Respective
Coefficient Alphas for Each Sample
Coefficient alphas
PC variables NP SA PL FF
Role stress and lack of harmony
Role ambiguityRole conflictRole overloadSubunit conflict
.70 .83 .68 .76
.81 .63 .66 .71
.80 .36 .64 .52
.60 .82 .66 .63.71Lack of organizational identification .79 .87 .81
Lack of management concern andawareness .76 .82 .80 .82
Job challenge and autonomy
Challenge and varietyAutonomyJob importance
.83 .82 .67 .72
.71 .73 .60 .69
.73 .66 .62 .65
Leadership facilitation and support
Leader trust and supportLeader goal facilitationLeader interaction facilitationPsychological influenceHierarchial influence
.85
.86
.80
.82
.84
.85
.88
.77
.76
.70
.87
.81
.61
.75
.63
.85
.89
.73
.73
.61
Workgroup cooperation, friendliness, and warmth
Workgroup cooperationWorkgroup friendliness and warmthReputation for effectivenessEsprit de corps
.75
.73
.59
.61
.76
.82
.64
.74
.67
.41
.74
.77
.61
Note. NP = Navy personnel, SA = systems analysts, PL = productionline, FF = firefighters.
indicators, the four PC factors from Table 1 were treated as first-order
latent variables or factors, and PCg was the predicted higher order latent
variable or factor. In the first-order analysis—that is, the test of relations
between the 4 first-order factors (FOFs) and the 17 (or 18) manifest PC
variables—each manifest variable was allowed to be an indicator of only
one FOF (i.e., a free parameter to be estimated, although one such pa-
rameter, randomly selected, had to assume a fixed value of 1.0 for each
FOF; Joreskog & Sorbom, 1986). Relations between each such indica-
tor and the other three factors were fixed at zero.
On the basis of prior exploratory factor-analytic work by Jones and
James (1979), we expected that the 17 (or 18) manifest PC variables
would factor into the domains indicated in Figure 1. Two differences
between this study and the prior research by Jones and James were that
(a) the PC questionnaire was revised and shortened by collapsing the
role and subsystem/organizational domains to obtain the individual/
organizational interface domain (a product being a single factor to rep-
resent the interface) and (b) the first-order factors were left free to be
correlated. The second-order analysis ascertained whether a single
higher order factor could account for the covariation among the four
FOFs. With four FOFs, this model is overidentified and thus the model
is disconfirmable by tests of fit (cf. James, Mulaik, & Brett, 1982). All
coefficients in the vector relating PCg to the FOFs were free to be esti-
mated, and the variance of PCg was fixed at unity.
It is noteworthy that, in each of the four samples, the determinant of
the observed variance-covariance matrix for the 17 (or 18) PC manifest
MEASUREMENT OF MEANING 743
variables was very small relative to the magnitude of the diagonal ele-ments (e.g., .000006 for the Navy sample). These results indicated ahigh degree of collinearity. In such conditions, Joreskog and Sorbom(1986) recommended the use of unweighted least squares (ULS) to esti-
mate the values of the free parameters.The statistical distribution for ULS is unknown. One ramification is
that chi-square (x2) cannot be used to test the goodness of fit of themodel. Even though LISREL does provide other indices to assess the over-
all fit of the model to observed data, we did attempt to obtain a maxi-mum likelihood (ML) solution and included a chi-square test to assessthe fit of the estimated models. The models were therefore tested byusing both ULS and ML. Comparisons are provided between the ULS
and ML (a) parameter estimates; (b) correlations among the first-orderfactors; (c) the goodness-of-fit index (GFI), a measure of the relativeamount of variance and covariance jointly accounted for by the model;(d) the root-mean-square residual (RMSR), the average residual vari-ance and covariance, respectively; and (e) the squared multiple correla-tion (J?2), a measure of how well the indicators jointly accounted for thelatent variables. Standardized data were used in these analyses to dealwith the scale dependency associated with ULS estimation (cf. Long,
1983).The observed covariance matrix was then used to determine a chi-
square for the first-order factor model (x2) and the second-order factor
model (x!), based on the ML solution. The x? was used as a stand-aloneindex (cf. Marsh, Balla, & McDonald, 1988) to assess the fit of the first-order model. The fit of the hierarchical model was assessed with (a) adifference chi-square test (\l - x? with dfc- df, degrees of freedom), a
measure of the adequacy of PC, to explain the covariation among thefirst-order factors; and (b) the Tucker-Lewis (1973) index (TLI) [(XJ/4/i — Xi/dK)/(xtJdfn - 1.0), where n represents the null model], a mea-sure of the covariance accounted for by the hierarchical model relativeto the total covariance among the PC composites.
Results
Results of the confirmatory factory analyses (CFAs) are re-
ported in an order that corresponds to the empirically testable
assumptions presented earlier. Additional construct validation
efforts are then addressed.
Tests for PCg
The ULS standardized solution provided by the first-order
CFA on all four samples (i.e., Navy personnel, systems analysts,
production line, and firefighters) are presented in Table 2. The
data in Table 2 are consistent with prior research (cf. James &
Sells, 1981) regarding relations between manifest (PC) variables
and the four FOFs (see Figure 1). The goodness-of-fit indices
based on the ULS solution for all four samples generally indi-
cated that the fit of the first-order model to the data was good
(e.g., the GFI was greater than .95 for all samples). Nonetheless,
inspection of the normalized residual matrix revealed a large
number of normalized residuals with absolute values greater
than two. This finding is indicative of a possible specification
error in the first-order model (cf. Hayduk, 1987; Joreskog &
Sorbom, 1986).
A host of potential misspecifications exist, some of which are
indigenous to the hierarchical mode) and are not substantively
damaging to the theoretical fit of the model. For example, when
using item-composites as manifest indicators, it is plausible that
the items from different composites within the same FOF co-
vary beyond the covariation accounted for by that FOF (Hert-
zog, 1989). Closer inspection of the normalized residual matrix
suggested that this was a viable explanation for the large number
of inflated residuals. This was because the majority of the large
residuals were associated with PC variables that loaded on a
common FOF. Second, it was possible that some of the PC com-
posites loaded on more than one factor. We tested both of these
assertions with a ML solution as follows.
Covariation among composites unaccounted for by a com-
mon FOF was assessed by allowing several of the unique vari-
ances for the PC composites (i.e., the off-diagonal elements of
the theta epsilon [TE] matrix) to covary. Specifically, the off-
diagonal TE within each factor with the highest modification
index was set free to be estimated. We then allowed a PC com-
posite to load on more than one factor if it was theoretically
reasonable and if the modification index indicated that the freed
parameter would result in a substantial gain in the fit of the
model. Overall, these analyses were conducted following the
guidelines recommended by MacCallum (1986) for follow-up
analyses.
The ML standardized solutions provided by the first-order
CFAs for all four samples are presented in Table 3. These results
are highly similar to the ULS solution. Differences between the
solutions included the correlated TEs and the additional factor
loadings obtained by composites loading on more than one
FOF. There were 3-5 correlated TEs over samples, all but one
of which were within a common factor. The one exception oc-
curred in the production sample, in which it was indicated that
identification with an organization was conditional on the repu-
tation of the individual's workgroup. This relationship could
not be accounted for by the four FOFs, even when the PC com-
posites for identification and group reputation were allowed to
cross-load on the factors. In regard to PC composites loading
on more than one factor, this occurred rarely and primarily for
the psychological influence composite. In general, however, re-
sults reported in Tables 2 and 3 are highly consistent.
The chi-squares provided by the ML analysis for the four
samples were as follows: Navy, x2(123, AT = 422) = 446.51; sys-
tems analysts, x2(109, jV = 128) = 196.28; production line,
X2(107,JV= 208) = 248.23; and firefighters, x2(109,JV = 288) =
323.35. The chi-squares appear to be acceptable given the sam-
ple sizes and the possible violation of multivariate normality.
Although lack of multivariate normality does not affect param-
eter estimates, it might inflate the chi-square (Bentler & Chou,
1987). We thus checked the joint normality assumption for all
four samples. Specifically, squared generalized distances were
computed for each observation in each sample (cf. Johnson &
Wichern, 1988, p. 154). Because the squared generalized dis-
tances behave like chi-square random variables with sample
sizes greater than 30, a chi-square test of significance could be
used to test the multivariate normality assumption. All four chi-
square tests were significant (p < .001), which indicates that the
data sets from which the four samples were drawn were not
distributed according to multivariate normal distributions.
Additional support for the model was provided by the TLI,
which was greater than .90 for all samples. More importantly,
for each of the four samples, there were only five normalized
residuals slightly greater than 2 (the range was 2.1 to 2.7), all of
which were associated with role overload. These results suggest
744 LOIS A. JAMES AND LAWRENCE R. JAMES
Table 2
Standardized Estimates of Relations of Manifest Psychological Climate (PC) Variables on 4
First-Order Factors, Usingan Unweighted Least Squares Solution
Navy personnel Systems analysts Production line Firefighters
PC variables 1 1
1. Role ambiguity .86"2. Role conflict .763. Role overload .484. Subunit conflict .485. Lack of organization identification .696. Lack of management concern and
awareness .68
Role stress and lack of harmony
.87* .76"
.65 .65
.38 .53
.62 .67
.90 .82
.80 .88
.80"
.67
.39
.62
.68
.82
Job challenge and autonomy
7. Challenge and variety8. Autonomy9. Job importance
.74
.79"
.72
.76
.78"
.67
.61
.77"
.54
.69
.66"
.39
Leadership facilitation and support
10. Leader trust and support1 1 . Leader goal facilitation1 2. Leader interaction facilitation13. Psychological influence14. Hierarchical influence
.93"
.88
.84
.83
.87
.84'
.85
.75
.72
.75
.88"
.75
.63
.83
.78
.90"
.85
.83
.79
.78
Workgroup cooperation, friendliness, and warmth
15. Workgroup cooperation16. Workgroup friendliness and warmth17. Reputation for effectiveness18. Esprit de corps
.71"
.69
.67
.85
.76"
.77
.90
.76"
.71
.76
.91"
.85
.76
" Fixed parameter (not equal to 1.0 because results are presented in terms of a standardized solution such that the Psi matrix has 1 .Os in diagonal).
the need for additional psychometric work on this PC com-
posite.
Of particular interest are the results shown in Table 4, which
demonstrate high correlations among the four FOFs for all four
samples. The similarity of the correlations among the FOFs for
ULS (below the diagonal) and ML solutions (above the diago-
nal) are noteworthy.
In sum, the key parameter estimates—factor loadings and
correlations among the first-order factors—were essentially the
same for the ULS solution and the ML solution. The ML solu-
tion was more complex in that parameters were added to the
model to address correlated errors and factorial complexity. Al-
though this added complexity enhanced the statistical fit of the
ML model, it had no meaningful effect on the estimates of key
parameters. We thus proceeded to use the simple structure ULS
first-order factor model as well as the more complex ML first-
order factor model to test for a single, higher order factor. Spe-
cifically, a second-order factor structure was created in which
each first-order factor was accounted for by a common compo-
nent due to PC, and a residual first-order factor.
Results of analyses designed to test the goodness of fit of a
single, higher-order factor model are reported in Tables 5
through 7. Tests of overall fit, presented in Table 5, generally
indicate that the single, higher-order factor model was con-
firmed in all samples. The difference chi-square was nonsig-
nificant in three of the samples. (Even though the difference chi-
square was significant in the production sample, the difference
was small.) The TLI was greater than .90 in all samples. The
ULS and ML goodness-of-fit indices were also generally compa-
rable. The RMSR values were uniformly low, and the multiple
correlations were high across all samples. The GFI values were
all .97 or .98 for the ULS solution but dropped to the middle to
upper .80s for the ML solution. This indication of a possible
discrepancy in the ULS and ML solutions may reflect the effects
of skewed distributions or collinearity, or both, on the ML esti-
mates. However, these effects do not appear to have been sub-
stantial given that the GFI values based on the ML solution are
still reasonably high and not seriously inconsistent with the GFI
values based on ULS estimates.
Estimated loadings for the FOFs on PC, are presented in Ta-
ble 6. The generally high magnitude of these estimates suggests
that PCg explained substantial portions of variance in each of
the FOFs. The estimated loadings between the ULS and ML
solutions were again similar. The mean loading for absolute val-
ues for both ULS and ML was .82, which denotes that an aver-
age of 67% of the variance in the FOFs was attributable to PCg.
These results, combined with those of the goodness-of-fit tests,
furnish strong support for the prediction that the FOFs share a
common element of meaning.
It should be mentioned that the patterns of loadings varied
MEASUREMENT OF MEANING 745
Table3Standardized Estimates of Relations of Manifest Psychological Climate (PC) Variables on 4
First-Order Factors, Using a Maximum Likelihood Solution
Navy personnel Systems analysts Production line Firefighters
PC variables
1. Role ambiguity2. Role conflict3. Role overload4. Subunit conflict5. Lack of organization identification6. Lack of management concern and
Role stress and lack of harmony
.79* .87' .75"
.76 .63 .67
.48 .35 .57
.51 .70 .68
.62 .85 .79
.66 .82 .56 -.33
.87"
.72
.42
.48
.59
.84
Job challenge and autonomy
7. Challenge and variety8. Autonomy9. Job importance
.69
.82'
.61
.76
.78'
.60
.51
.74'
.45
.70
.66'
.40
Leadership facilitation and support
10. Leader trust and support11. Leader goal facilitation12. Leader interaction facilitation13. Psychological influence14. Hierarchical influence
.92'
.87
.87.37 .55
.88
.81*
.88
.77.56 .28
.76
.87*
.79
.65.37 .55
.78
.90'
.85
.83
.81
.76
Workgroup cooperation, friendliness, and warmth
15. Workgroup cooperation16. Workgroup friendliness and warmth17. Reputation for effectiveness18. Esprit de corps
.63"
.60
.69
.84
.68"
.65
.87
.69-
.63
.71
.93'
.87
.71
Theta epsilon
TEI6TE 5TEHTE 3TE 7
15 =6 =
12 =2 =9 =
.38
.25
.08
.25.25
TE16 15 =TE1013 =TE 7 9 =
.42
.17
.13
TE15TE 17TE 7TE 3
16592
====
.35
.17
.37
.17
TE 6TE13TEH
11112
= .18= -.11= .10
• Fixed parameter.
somewhat among samples. Nevertheless, the general pattern of
high loadings suggests a general factor in all samples.
Indirect Effects ofPCs
The indirect effects of PC, on the 17 (or 18) manifest vari-
ables are presented in Table 7. The indirect effects are a product
of the direct effects of PC, on the FOFs and the direct effects of
the FOFs on the manifest variables. For example, the total effect
of PCg on perceived role ambiguity equals the direct effect of
PCg on the first-order factor Role Stress and Lack of Harmony
times the direct effect of Role Stress and Lack of Harmony on
the PC composite ambiguity (i.e., for ULS, the solution [•>•] i =
-.77] [X,i = .86] = indirect effects of-.662).
The range was from .29 to .84 over samples, and the means
of the indirect effects were essentially identical over samples.
These results indicate that reasonable amounts of variance in
the initial measures of meaning were explained by PCg (an aver-
age of 37% of the variance was explained) and thus that PQ
was an important cause of the initial perceptions of meaning.
Consequently, the third and final assumption tested by CFA was
supported.
Additional Construct Validation Issues
The results reported previously furnish empirical support for
the hierarchical model of PCg and, within the context of the
CFAs, one may say that the model was confirmed. However,
confirmation of a model suggests that the model furnishes a use-
ful basis for explaining how and why events occurred. It does
not connote that the confirmed model is unique or has been
proven to be true (James et al., 1982), which is to say that alter-
native models may explain the data equally well. An alternative
model of particular concern here is based on systematic or non-
random measurement errors. Such a model predicts that the
higher order PCg factor is actually a measure of common re-
746 LOIS A. JAMES AND LAWRENCE R. JAMES
S I S3 8I '
S Rr r
—-I OO OO
oo «rj 8 i,' ' ' I
£ 8
3 i r-
m o oor- t- >o
3 i &
sponse biases such as acquiescence or social desirability. In cur-
rent jargon, one might propose that this general factor is a prod-
uct of a pervasive method factor and thus lacks the substantive
meaning that we have attributed to it (i.e., an appraisal of per-
sonal benefit vs. personal detriment).
To test the hypothesis that the general factor represents a per-
vasive method factor, we estimated the correlations between the
general factor and overall job satisfaction (OJS) and three per-
sonality variables (achievement motivation, self-esteem, and ri-
gidity) known to be predictors of PC perceptions or moderators,
or both, of relations between PC perceptions and individual
outcomes (James etal., 1979, 1981; James & Jones, 1980). The
rationale underlying this test is that if the general factor is inter-
preted as PC8, then PCg should be positively, and probably
highly, correlated with OJS. This prediction follows from the
theory that an appraisal of the degree to which one is personally
benefited versus personally hindered or harmed is the basis on
which an overall evaluation is made of one's organizational
well-being (as measured by OJS).
However, one might argue that a strong, positive correlation
between the general factor and OJS is a mere extension of the
pervasive method factor that engendered the general factor in
the first place. If this is a valid inference, then it would also
follow that the pervasive method factor would influence re-
sponses to other measures in the questionnaire used to assess
meaning. In particular, this method factor should also influence
responses to items measuring achievement motivation, self-es-
teem, and rigidity. These personality variables (a) were mea-
sured in precisely the same manner and by the same method as
the PC variables and OJS, (b) had psychometric characteristics
(e.g., reliabilities) that were similar to the PC variables and OJS,
and (c) could be assumed to be subject to the same types of
response biases as the PC variables and OJS. It follows that a
pervasive method factor that produces spuriously high corre-
lations between the general factor and OJS should also promote
spuriously high correlations between this same general factor
and the personality variables.
The squared correlations provided by ULS between the gen-
eral factor and OJS, and between the general factor and each of
the personality variables, are reported in Table 8. A differential
pattern of relations is clear; the general factor correlated much
more highly with OJS than with the personality variables. Even
though these results are, again, only indicative of construct va-
lidity, they question an alternative model that poses a pervasive
method factor as an alternative explanation of PCg.
Discussion
This study was based on the theory that factors underlying
emotionally relevant cognitions (valuations) of work environ-
ments reflect appraisals of the degree to which the overall work
environment is believed to be personally beneficial versus per-
sonally detrimental to the organizational well-being of the indi-
vidual. Personal benefit versus personal detriment to organiza-
tional well-being thus served as a proposed general factor, desig-
nated PCg, in a hierarchical model of value-based meanings for
work environments. Results of confirmatory factor analyses on
multiple and diverse samples supported most empirical predic-
tions suggested by this model. Specifically, (a) first-order factors
MEASUREMENT OF MEANING 747
Table 5
Goodness-of-Fit Indices for a Hierarchical Model of Meaning
Measures NP SA PL FF
Chi-square for hierarchical modelChi-square difference*Tucker-Lewis index
<VN
447.30.79.91
125422
200.193.91.92
111128
256.408.17*
.92109206
327.894.54
.91111288
Comparison of relevant goodness-of-nt indices
ULS ML ULS ML ULS ML ULS ML
GFIRMSRR2, first-order factorsJJ!, second-order factors
.98
.06
.99
.89
.89
.06
.99
.90
.98
.07
.98
.90
.85
.06
.99
.90
.97
.07
.98
.93
.86
.06
.97
.96
.97
.06
.99
.92
.88
.06
.99
.93
Note. NP = Navy personnel, SA = systems analysts, PL = production line, FF = firefighters. ULS = un-weighted least squares; ML = maximum likelihood; GFI = goodness-of-fit index; RMSR - root-mean-square residual.* Chi-square difference is between first-order model and higher order model with 2 df.
(FOFs) underlying manifest perceptual variables were corre- an endogenous variable, it is likely that Person X Situation in-
lated, (b) a general factor (PC,) explained much of the covaria- teraction models will be required to explain PC, (cf. Endler &
tion among the FOFs, and (c) the general factor explained rea- Magnusson, 1976; James et al., 1 979, 1981; James, Hater, Gent,
sonable portions of variance in the manifest perceptual vari- & Bruni, 1978). Discussions of value development are also
ables. Additionally, a model that attempted to attribute results likely to be productive (cf. Locke, 1976), as is recent cognitive
to a pervasive method factor was disconflrmed, and PC, was research on top-down versus bottom-up processing models (cf.
highly correlated with an indicator of overall organizational Diener, 1984). As a causal variable, models may be developed
well-being. that relate PC, to affect (see later discussion), behavioral dispo-
Such empirical support for PCg denotes only that this is an sitions, and behaviors themselves. Finally, measurement issues
area potentially worthy of future research. It does not suggest require further attention. The rudimentary construct valida-
that the PC, model is true, unique, or proven. The tenuous na- tion effort reported here suggests only that a pervasive method
ture of this model is evident when one considers the large num- factor could not explain the data. More intense efforts that use
ber of assumptions on which the model was based but which multiple measures of each construct, combined with more so-
were not tested in the reported analyses. Future research is phisticated analyses (cf. Schmitt & Stults, 1986; Widaman,
needed to assess if indeed values, and perhaps other compo- 1985), are required before one may feel confident that models
nents of belief systems (e.g., self-concepts, self-regulatory sys- based on nonrandom measurement errors are clearly discon-
tents), engender the cognitive constructs used to impute mean- firmable.
ing to work environments. Research is also needed that ad- In sum, our effort represents a first step in what may prove
dresses other causes as well as effects of PC,. When viewed as to be a fruitful area of research. It has many limitations, but it
Table 6
Standardized Estimates of Relations of the 4 First-Order Latent Variables on a Single, Higher
Order Latent Variable for ULS and ML Solutions
PC, for each sample
NP SA PL FF
First-order factors ULS ML ULS ML ULS ML ULS ML
Role Stress and Lack of Harmony -.77 -.84 -.81 -.83 -.95 -.97 -.80 -.78Job Challenge and Autonomy .80 .79 .71 .69 .67 .70 .92 .90Leadership Facilitation and Support .80 .77 .90 .88 .87 .82 .93 .94Workgroup Cooperation, Friendliness,
and Warmth .87 .88 .78 .84 .75 .82 .72 .69
Note. ULS = unweighted least squares; ML = maximum likelihood. NP = Navy, SA = systems analysts,PL = production line, FF = firefighters.
748 LOIS A. JAMES AND LAWRENCE R. JAMES
Table 7
Indirect Effects ofPCs on Manifest Indicators for VLS and ML Solutions
NP SA PL FF
PC variables ULS ML ULS ML ULS ML ULS ML
Role ambiguityRole conflictRole overloadSubunit conflictLack of organization identificationLack of management concern and
awareness
Role stress and lack of harmony
-.66 -.66 -.70 -.72-.59-.37-.37-.53
-.64-.40-.43-.52
-.52-.31-.50-.73
-.52-.29-.58-.71
-.72 -.73 -.64 -.67-.62-.50-.64-.78
-.65-.56-.66-.77
-.54-.31-.50-.54
-.52 -.55 -.65
-.56-.33-.53-.46
-.82 -.66 -.65
Job challenge and autonomy
Challenge and varietyAutonomyJob importance
.69
.63
.58
.55
.65
.48
.54
.56
.48
.53
.54
.42
.41
.52
.36
.36
.52
.32
.63
.61
.36
.63
.59
.35
Leadership facilitation and support
Leader trust and supportLeader goal facilitationLeader interaction facilitationPsychological influenceHierarchial influence
.74
.70
.67
.66
.69
.71
.67
.67
.71
.68
.75
.77
.68
.65
.68
.71
.77
.68
.64
.67
.76
.65
.55
.72
.68
.71
.65
.53
.71
.65
.84
.79
.77
.73
.73
.85
.80
.78
.77
.72
Workgroup cooperation, friendliness, and warmth
Workgroup cooperationWorkgroup friendliness and warmthReputation for effectivenessEsprit de corps
.62 .56
.60 .53
.58 .61
.74 .75
.59 .57
.60 .55
.70 .73
.57 .57
.53 .52
.57 .59
.66 .64
.61 .60
.55 .49
Summary statistics
RangeMean1
ULSML
.37-.7S
.60 .61
.60 .61
.29-.77
.61 .62
.61 .61
.32-.S2 .33-.S4
Note. NA = Navy personnel, SA = systems analysts, PL = production line, FF = firefighters. ULS = un-weighted least squares; ML = maximum likelihood.a Mean was computed on absolute values.
also has some interesting implications. For instance, the PC8
mode] suggests that underlying cognitive structures pertaining
to emotionally relevant cognitions of work environments are
Table 8
Squared Correlations ofPCs with Overall Job Satisfaction and
Personality Variables Provided by Unweighted Least Sqttares
PC8
Variables NP SA PL FF
Overall job satisfactionAchievement motivationRigiditySelf-esteem
.79
.04
.00
.06
.88
.06
.10
.01
.75
.11
.04
.06
.73
.05
.00
.03
Note. NA = Navy personnel, SA - systems analysts, PL = productionline, FF = firefighters.
more integrated and parsimonious than has been realized. The
existing tendencies to compartmentalize environmental per-
ception research by environmental domains such as jobs and
leaders have resulted in a large, diffuse set of perceptual con-
structs that not only lacks a unifying theme but also lacks an
integrated theory regarding the substantive impact of the per-
ceived work environment on individuals. We propose that such
an integrated theory may exist and that it is possible, in part, to
explain the substantive impact of the perceived work environ-
ment on individuals. Stated simply, people respond to work en-
vironments in terms of how they perceive these environments,
and a key substantive concern in perception is the degree to
which individuals perceive themselves as being personally bene-
fited as opposed to being personally harmed (hindered) by their
environment.
This article concludes with discussions of two models that
furnish alternative explanations for the correlations among the
PC variables. These models do not exhaust potentially viable
MEASUREMENT OF MEANING 749
alternatives to PCg theory, but we consider them to be among
the more salient competitors of such theory.
Situational Process Model
The logic here is that perceptions on the PC variables repre-
sent, primarily, the influences of situational processes and thus
covariation among the PC variables reflects, primarily, covaria-
tion among workgroup and other contextual processes. Partial
support for a situational process model is furnished by prior
studies that have reported significant correlations between mea-
sures of situational variables and one or more of the PC vari-
ables (cf. James et al., 1979, 1981; James & Jones, 1980; James
& Sells, 1981; Jones & James, 1979). It is also true, however,
that many of these studies indicated that perceptions on PC
variables were also related uniquely to person variables (e.g.,
job involvement) or to Person X Situation interactions, or both.
Thus, it cannot be said that the PC variables are primarily func-
tions of workgroup and other contextual processes.
An additional point is that to confirm a situational process
model with our data would require that the correlations among
situational processes not only be quite high but also that re-
lations involving differing sets of processes be similar. To illus-
trate, consider the rather flat profile of correlations among the
first-order PC factors, reported in Table 4, that had an average
absolute value of approximately .66. One might build a case
that the meanings associated with jobs, leaders, workgroups,
and individual/organizational interfaces should be correlated
simply because the environmental attributes that are the ob-
jects of perception are correlated (e.g., open-systems theory).
However, it is rather unlikely that such a case would predict a
flat profile of correlations over diverse environmental attri-
butes, particularly for an average correlation as high as .66. A
more cogent explanation of these findings is that each first-order
factor was generated at least in part by a common, latent, cogni-
tive variable. Of course, a meaningful proportion of the vari-
ance in the first-order factors is unique and, by definition, reli-
able, so we must be careful not to overstate our position. In
addition, it is likely that covariation among situational pro-
cesses carried over into the meanings attributed to these pro-
cesses. Nevertheless, a flat profile of high correlations is pre-
cisely what one would expect if a strong, general, cognitive fac-
tor were present in the data. Thus, these findings are consistent
with a cognitive "g-factor theory" and suggest that PCg is useful
for explaining salient portions of variance in work environment
perceptions.
Job Satisfaction Model
The average correlation (over samples) of .89 between PCe
and OJS (see Table 8) stimulated yet again the question, What is
the difference between climate and job satisfaction (JS; Guion,
1973; Johannesson, 1973)? Indeed, one reviewer suggested that
the second-order factor that we designated PCg may instead be
OJS. This same reviewer noted the need to conduct empirical
tests to ascertain whether second-order factors for PC and JS
are empirically distinguishable. We agree, but we could not
conduct such a test with these data because a brief, short-form
instrument was used to assess OJS. Such an instrument does
not provide the range of data needed to meaningfully examine
the factor structure of JS or to compare OJS with PCg. Thus, a
test to ascertain whether PCg and OJS are distinguishable awaits
future research.
As part of this future research we propose a test of an addi-
tional model. The theory underlying this model is that PCg and
an indicator of organizational well-being such as OJS should be
highly correlated because emotionally relevant cognitions and
the emotions (or feeling states) are components of reciprocally
interacting, interdependent, nonrecursive, fused processes (cf.
Ittelson, Proshansky, Rivlin, & Winkel, 1974; James & Jones,
1974, 1980; James &Tetrick, 1986; Lazarus, 1982, 1984; Park,
Sims, & Motowidlo, 1986). In regard to a PCe -»• OJS relation,
a cognitive appraisal of the degree to which an event, agent, or
object is personally beneficial versus detrimental to one's organ-
izational well-being is believed to determine the type and qual-
ity of emotion experienced as well as the direction and intensity
of the emotion (Lazarus, 1982, 1984). Moreover, the global ap-
praisal of whether the work environment as a whole is person-
ally beneficial versus detrimental is hypothesized to be the pri-
mary cause of organizational well-being, which may be charac-
terized by a general feeling state (Isen, 1984) reflecting
primarily positive affect (e.g., happiness, contentment, satisfac-
tion) or primarily negative affect (e.g., depression, melancholy,
dissatisfaction; cf. Diener, 1984). The reciprocal OJS -»• PQ
causal relation is based on the hypothesis that existing or de-
sired levels of overall organizational well-being may cause indi-
viduals to restructure or redefine cognitions to make them con-
sistent with the experienced (or desired) level of affect (cf. James
&Tetrick, 1986).
Although the reciprocal relationship between PQ and OJS
has not been tested, a possible implication of the reciprocal in-
fluence of OJS on PCg is noteworthy. That is, the reciprocal
OJS -»• PCB causal relation suggests an additional source of co-
variation among the first-order PC factors. This possibility is
predicated on the assumption that overall job satisfaction is, in
part, a function of individual dispositions such that individuals
have a generalized predisposition toward a given level or degree
of satisfaction (Pulakos & Schmitt, 1983). If such a predisposi-
tion exists, then an individual's global cognitive appraisal of
benefit versus harm engendered by the work environment may
be a function of that individual's predisposition to be satisfied
or dissatisfied. It follows that the covariation among the first-
order PC factors attributable to PC, may be, in part, a function
of the reciprocal influence of OJS on PQ. In short, a key (albeit
indirect) cause of covariation among the first-order PC factors
may be an affective predisposition.
To illustrate, negative affectivity (NA) is viewed as a mood-
dispositional variable associated with negative emotions, such
as dissatisfaction. Individuals that measure high on NA tend
to interpret situations in a manner that is consistent with their
experienced negative emotions (e.g., dissatisfaction). Such indi-
viduals accentuate the negative aspects of most situations (Wat-
son & Clark, 1984). If these same individuals tend to interpret
the work environment in a manner that is consistent with exist-
ing levels of negative affect, then a predisposition toward dissat-
isfaction may cause them to view their work environment as
being detrimental to their overall well-being. The implied low
score on PCe would in turn stimulate perceptions that jobs lack
750 LOIS A. JAMES AND LAWRENCE R. JAMES
challenge, leaders are unsupportive, roles are stressful, and so
on. In this manner, covariation among the perceptions of chal-
lenge, support, stress, and so forth, would reflect the indirect
influences of a predisposition toward NA.
Research pertaining to the possibility of a dispositional
source of job satisfaction has been sparse. Nonetheless, the few
studies that have been done have suggested that job satisfaction
may in part be a function of stable individual characteristics
(Arvey, Bouchard, Segal, & Abraham, 1989; Schmitt & Pu-
lakos, 1985; Staw, Bell, & Clausen, 1986; Staw & Ross, 1985).
Thus, the possibility that predispositions in affect influence
both PCg and the first-order PC factors is a viable question for
future research. On the other hand, this is but one of several
hypotheses, other hypotheses being that situational factors are
ultimately responsible for covariations among first-order PC
factors or between repeated measurements on OJS, or both (cf.
Gerhart, 1987) and that stable individual values are responsible
for the valuations underlying interpretations of environments.
Of course, it is possible, if not likely, that PC perceptions and
PCBare multiply determined, that is, a function not only of situ-
ations, both actual and perceived, and stable individual predis-
positions (e.g., affect, values), but also of potentially complex
interactions among these various sources (cf. James, James, &
Ashe, in press).
In sum, the presence of alternative models points to the need
(and direction) for future research. A key hypothesis to be
tested is that a general perceptual factor of personal benefit ver-
sus personal detriment furnishes a theme for unifying percep-
tions of work environments and an integrated theory for ex-
plaining, in part, the substantive impact of work environment
perceptions on individual outcomes. A series of hypotheses then
pertain to the causes of this general perceptual factor, which
includes stable individual characteristics (values) and predispo-
sitions (affect), situational factors, and various interactions
among these potential causes.
References
Arvey, R. D., Bouchard, T. J., Jr., Segal, N. L., & Abraham, L. M.(1989). Job satisfaction: Environmental and genetic components.
Journal of Applied Psychology 74, 187-192.Bentler, P. M., & Chou, C. P. (1987). Practical issues in structural mod-
eling. Sociological Methods and Research. 16, 78-117.Bradburn, N. M. (1969). The structure of psychological well-being. Chi-
cago: Aldine.Diener, E. (1984). Subjective well-being. Psychological Bulletin. 95.
542-575.Ekehammer, B. (1974). Interaclionism in personality from a historical
perspective. Psychological Bulletin, 81, 1026-1048.Endler, N. S., & Magnusson, D. (1976). Toward an interactional psy-
chology of personality. Psychological Bulletin, S3, 956-974.Gerhart, B. (198 7). How important are dispositional factors as determi-
nants of job satisfaction? Implications for job design and other per-
sonnel programs. Journal of Applied Psychology, 72, 366-373.Guion, R. M. (1973). A note on organizational climate. Organizational
Behavior and Human Performance, 9, 120-125.Hayduk, L. A. (1987). Structural equation modeling with LISREL. Balti-
more, MD: Johns Hopkins University Press.Hertzog, C. K. (1989). Using confirmatory factor analysis for scale de-
velopment and validation. In M. P. Lawton & A. R. Herzog (Eds.),Special research methods for gerontology (pp. 281-306). Amityville,
NY: Bayweod Press.
Isen, A. M. (1984). Toward understanding the role of affect in cognition.In R. S. Wyer, Jr. & T. K. Srull (Eds.), Handbook of social psychology
(pp. 179-236). Hillsdale, NJ: Erlbaum.
Ittelson, W. H., Proshansky, H. M., Rivlin, L. G., & Winkel, G. H.(1974). An introduction to environmental psychology. New %rk: Holt,Rinehart & Winston.
James, L. R. (1982). Aggregation bias in estimates of perceptual agree-
ment. Journal of Applied Psychology, 67,219-229.
James, L. R., Gent, M. J., Hater, J. J., & Coray, K. E. (1979). Correlates
of psychological influence: An illustration of the psychological cli-
mate approach to work environment perceptions. Personnel Psychol-ogy, 32, 563-588.
James, L. R., Hater, J. J., Gent, M. J., & Bruni, J. R. (1978). Psychologi-
cal climate: Implications from cognitive social learning theory andinteractional psychology. Personnel Psychology, 31, 783-813.
James, L. R., Hater, J. J., & Jones, A. (1981). Perceptions of psychologi-
cal influence: A cognitive information processing approach for ex-
plaining moderated relationships. Personnel Psychology, 34, 453-477.
James, L. R., James, L. A., & Ashe, D. (in press). The meaning of orga-nizations enacted, transacted, or imposed. In B. Schneider (Ed.),Frontier Series.
James, L. R., & Jones, A. P. (1974). Organizational climate: A reviewof theory and research. Psychological Bulletin, SI, 1096-1112.
James, L. R,, & Jones, A. P. (1980). Perceived job characteristics andjob satisfaction: An examination of reciprocal causation. PersonnelPsychology, 55,97-135.
James, L. R., Mulaik, S. A., & Brett, J. M. (1982). Causal analysis
assumptions, models and data. Beverly Hills, CA: Sage.
James, L. R., & Sells, S. B. (1981). Psychological climate: Theoretical
perspectives and empirical research. In D. Magnusson (Ed.), Towardapsychology of situations: An interactional perspective (pp. 275-295).Hillsdale, NJ: Erlbaum.
James, L. R., & Tetrick, L. E. (1986). Confirmatory analytic tests ofthree causal models relating job perceptions to job satisfaction. Jour-
nal of Applied Psychology, 71, 77-82.
Johannesson, R. E. (1973). Some problems in the measurement of or-ganizational climate. Organizational Behavior and Human Perfor-
mance, 10, 118-144.
Johnson, R. A., & Wichern, D. W. (1988). Applied multivariate statisti-
cal analysis (2nd ed.). Englewood Cliffs, NJ: Prentice-Hall.
Jones, E. E., AGerard, H. B. (1967). Foundations of social psychology.New York: Wiley.
Jones, A. P., & James, L. R. (1979). Psychological climate: Dimensions
and relationships of individual and aggregated work environment
perceptions. Organizational Behavior and Human Performance, 23,
201-250.
Joreskog, K. G., & Sorbom, D. (1986). USRKL vi analysis of linearstructural relationships by maximum likelihood, instrumental vari-
ables, and least squares methods. Mooresville, IN: Scientific Soft-
ware.
Lazarus, R. S. (1982). Thoughts on the relations between emotion and
cognition. American Psychologist, 37, 1019-1024.
Lazarus, R. S. (1984). On the primacy of cognition. American Psycholo-
gist, 39, 124-129.
Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal, and coping.
New York: Springer,
Lewin, K. (1938). The conceptual representation of the measurement of
psychological forces. Durham, NC: Duke University Press.
Lewin, K. (1951). Behavior and development as a function of the totalsituation. In D. Cartwright (Ed.), Field theory >" social science (pp.238-303). New York: Harper Torchbooks.
Locke, E. A. (1976). The nature and causes of job satisfaction. In
MEASUREMENT OF MEANING 751
M. D. Dunnette (Ed.), Handbook of industrial and organizational
psychology (pp. 1297-1349). Chicago: Rand McNally.
Long, J. S. (1983). Confirmatory factor analysis. Beverly Hills, CA:Sage.
MacCallum, R. (19S6). Specification searches in covariance structure
modeling. Psychological Bulletin, 100, 107-120.Mandler, G. (1982). The structure of value: Accounting for taste. In
M. S. Clark & S. T. Fiske (Eds.), Affect and cognition (pp. 3-36). Hills-
dale, NJ: Erlbaum.Marsh, H. W., Balla, J. R., & McDonald, R. P. (1988). Goodness-of-fit
indexes in confirmatory factor analysis: The effect of sample size.Psychological Bulletin, 103, 391-410.
Mirriam-WebsterDictionary(l915). Springfield, MA: Author.
Mischel, W. (1968). Personality and assessment. New Trbrk: Wiley.
Osgood, C. E., May, W. H., & Miron, M. S. (1975). Cross-cultural uni-versals of affective meaning. Urbana: University of Illinois Press.
Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The measure-
ments of meaning. Urbana: University of Illinois Press.Park, O. S., Sims, H. P., Jr., & Motowidlo, S. J. (1986). Affect in organi-
zations. In H. P. Sims, Jr. & D. A. Gioia and Associates (Eds.), The
thinking organization: Dynamics of organizational social cognition(pp. 215-237). San Francisco: Jossey-Bass,
Pulakos, E. D., & Schmitt, N. (1983). A longitudinal study of a valencemodel approach for the prediction of job satisfaction of new employ-
ees. Journal of Applied Psychology, 68, 307-312.
Reisenzein, R. (1983). The Schacter theory of emotion: Two decadeslater. Psychological Bulletin, 94, 239-264.
Schachter, S., & Singer, J. E. (1962). Cognitive, social, and physiological
determinants of emotional state. Psychological Review 69, 379-399.Schmitt, N., & Pulakos, E. D. (1985). Predicting job satisfaction from
life satisfaction: Is there a general satisfaction factor? InternationalJournal of Psychology, 20, 155-167.
Schmitt, N., & Stults, D. M. (1986). Methodology review: Analysis ofmultitrait-multimethod matrices. Applied Psychological Measure-ment, 10, 1-22.
Staw, B. M., Bell, N. E., & Clausen, J. A. (1986). The dispositional ap-proach to job attitudes: A lifetime longitudinal test. AdministrativeScience Quarterly, 31, 56-77.
Staw, B. M., & Ross, J. (1985). Stability in the midst of change: A dispo-sitional approach to job attitudes. Journal of Applied Psychology, 70,
469-480.Stotland, E., & Canon, L. K. (1972). Social psychology and cognitive
approach. Philadelphia, PA: Sanders.
Tucker, L. R., & Lewis, C. (1973). The reliability coefficient for maxi-mum likelihood factor analysis. Psychometrika, 38, 1-10.
Watson, D., & Clark, L. A. (1984). Negative affectivity: The dispositionto experience aversive emotional states. Psychological Bulletin, 96,
465-490.Weiss, D. J., Dawis, R. V., England, G. W., & Lofquist, L. H. (1967).
Manual for the Minnesota Satisfaction Questionnaire. Minneapolis:Industrial Relations Center, University of Minnesota.
Widaman, K. F. (1985). Hierarchically nested covariance structuremodels for multitrait-multimethod data. Applied Psychological Mea-surement, 9, 1-26.
Zajonc, R. B. (1980). Feeling and thinking: Preferences need no infer-ences. American Psychologist, 35, 151-175.
Received July 26,1988
Revision received January 30, 1989
Accepted February 1,1989 •