Professional Documents
Culture Documents
Liden 1998
Liden 1998
Multidimensionafity of
Leader-Member Exchange:
An Empirical Assessment through
Scale Development
Robert C. L i d e n
University of Illinois at Chicago
John M. M a s l y n
Vanderbilt University
Leader-Member Exchange (LMX) theory suggests that leaders do not use the
same style in dealing with all subordinates, but rather develop a different type of
relationship or exchange with each subordinate (Dansereau, Graen, & Haga,
1975; Graen & Cashman, 1975; Liden & Graen, 1980; Graen, Novak, &
Sommerkamp, 1982; Graen & Scandura, 1987). These relationships range from
those that are based strictly on employment contracts (i.e., low LMX, or "out-
group") to those that are characterized by mutual trust, respect, liking and recipro-
cal influence (i.e., high LMX or "in-group;" Dansereau, et al., 1975).
Research on LMX has shown significant associations with many important
work outcomes. For example, LMX is negatively related to turnover (Graen,
Liden & Hoel, 1982) and turnover intentions (Vecchio & Gobdel, 1984), and
positively related to organizational commitment (Duchon, Green, & Taber, 1986),
satisfaction with supervision (Schriesheim & Gardiner, 1992), supervisory ratings
of job performance (Duarte, Goodson, & Klich, 1993, 1994; Graen, Novak, &
Sommerkamp, 1982), autonomy (Scandura, Graen, & Novak, 1986), satisfaction
with work (Vecchio & Gobdel, 1984), and frequency of promotions (Wakaba-
Direct all correspondence to: Robert C. Liden, Department of Managerial Studies MC 243, University of Illinois
at Chicago, 601 S. Morgan, Chicago, IL 60607-7123, e-mail <bobliden@uic.edu>.
43
44 ROBERT C. LIDEN AND JOHN M. MASLYN
yashi, Graen, Graen, & Graen, 1988). Although this research has been appealing,
there have been criticisms and suggestions for improvement (e.g., Vecchio &
Gobdel, 1984). One suggestion is that LMX is a multidimensional, not a unidi-
mensional construct (Dienesch & Liden, 1986). Also, none of the published LMX
articles have presented formal psychometric support for the most frequently used
LMX measure. The main purpose of the current research was to assess the propo-
sition that LMX is a multidimensional construct. This was accomplished by
attempting to develop and validate a psychometrically sound, multidimensional
measure of LMX.
Mulfidimensionality of LMX
Theoretical support for the multidimensionality of LMX may be derived
from role theory and social exchange theory. The theoretical foundation of LMX
has been built using role theory (Graen, 1976). According to this approach, lead-
ers test subordinates with various work assignments in a series of role making
episodes. The degree to which subordinates comply with task demands and
demonstrate a worthiness to be trusted dictates the type of LMX relationship that
forms. In turn, the type of LMX determines the extent to which the leader recipro-
cates with work-related resources such as information, challenging task assign-
ments, and autonomy (Graen & Scandura, 1987). The provision of resources by
the leader in return for subordinate task behaviors represents an exchange.
According to Graen and his colleagues, these exchanges are limited to work-
related commodities. In this way, LMXs are unidimensional, based upon the work
behaviors of leaders and subordinates (Graen, 1976; Graen & Scandura, 1987;
Graen & Uhl-Bien, 1995). However, role theory, which has provided the theoreti-
cal foundation for LMX research (e.g., Graen, 1976), stresses that roles are multi-
dimensional (Katz & Kahn, 1978; Jacobs, 1971). For example, some subordinates
may focus on their tasks, while neglecting social interaction, some may stress
social interaction and not tasks, and others may be weak or strong on both dimen-
sions (Bales, 1958). Leaders' roles are also comprised of multiple factors, includ-
ing such activities as supervising, allocating resources, and serving as a liaison
(Kim & Yukl, 1995; Tsui, 1984).
As implied by the exchange in leader-member exchange, social exchange
theory is highly relevant to LMX research (Sparrowe & Liden, 1997). Consider-
able attention, beginning primarily in the 1950s and 1960s and extending to the
present, has been paid to exchange processes (Blau, 1964; Gouldner, 1960;
Homans, 1958; Sahlins, 1972; Thibaut & Kelley, 1959). Curiously, this research
was overlooked by LMX theorists (e.g., Dansereau, et a1.,1975; Dienesch &
Liden, 1986; Graen, 1976; Graen & Cashman, 1975; Graen & Scandura, 1987)
until recently (Keller & Dansereau, 1995; Liden, Sparrowe, & Wayne, 1997;
Settoon, Bennett, & Liden, 1996; Wayne & Green, 1993; Wayne, Shore, & Liden,
1997). Social exchange theorists have identified numerous material and non-
material goods that may be exchanged (Gouldner, 1960). Advice, workflow, and
friendship are some of the exchange types that have been recognized (Brass &
Burkhardt, 1992; Krackhardt, 1990; Sparrowe & Liden, 1997), suggesting that
do so to benefit the organization, not just the leader. In this way, contribution may
be related to organizational commitment (Shore & Wayne, 1993). As a result of
differential task assignments, loyal members may report higher levels of auton-
omy in carrying out their jobs and a resulting high level of work satisfaction
compared to less loyal members who receive less task autonomy from the leader.
Supervisor ratings of subordinate performance have been shown to be affected by
liking (Tsui & Barry, 1986; Wayne & Ferris, 1990), and thus should be related to
the affect dimension. In addition, demonstrated loyalty and willingness to contrib-
ute above and beyond job requirements may be recognized by leaders as superior
performance.
Other dimensions. Dienesch and Liden (1986) did not claim that contribu-
tion, loyalty, and affect are the only LMX dimensions. They acknowledged the
possibility that dimensions other than these three might be involved in developing
and established LMXs. Other possible dimensions are suggested by social
exchange theory (e.g., Hollander, 1980; Homans, 1958; Jacobs, 1971). For exam-
ple, other dimensions of LMX might include trust, respect, openness, and honesty
(Graen & Scandura, 1987). Another purpose of the current research was to empir-
ically explore the existence of LMX dimensions in addition to those proposed by
Dienesch and Liden.
Methods
Overview
Our approach in evaluating the dimensionality of the LMX construct was to
develop a scale designed to assess different aspects of LMX relationships. We
followed an accepted approach to scale development outlined by DeVellis (1991)
and Hinkin (1995). Items were generated from an understanding of the LMX liter-
ature as well as from reports of experiences of individuals from the work setting,
subjected to content validation, administered to a large and diverse sample of
working students, and again administered several weeks later to generate re-test
data. Items surviving these analyses were then administered to employees repre-
judges were retained. A total of 38 items survived the second content validation.
Because of the lack of previous theoretical delineation of the professional respect
dimension, we wrote four additional items for this dimension, to ensure an
adequate chance of being represented after further validation procedures.
Finally, these 42 items were revisited by the authors for theoretical content
adequacy prior to submitting them to empirical analysis. Content adequacy evalu-
ation is different from an evaluation by judges of whether items reflect defined
dimensions, and involved an evaluation of the "theoretical... correspondence
between a measure's items and a construct's delineated content domain"
(Schriesheim, Powers, Scandura, Gardiner, & Lankau, 1993, p. 389). This was
done to help ensure that items retained for empirical analysis were cleanly reflect-
ing LMX theory and the underlying theoretical dimensions of LMX. Each item
was reviewed for an indication of exchange in general, for consistency with the
notion of a "currency of exchange" versus a probable antecedent to an exchange
relationship or an outcome of an exchange I , or simply for potential measurement
problems. Eleven items were dropped from further scale consideration through
this process: three were reflective of perceived similarity between supervisor and
subordinate (an antecedent: Liden, et al., 1993) (e.g., "I share interests in leisure
activities (e.g., sports, movies) with my supervisor"), four were associated with
likely outcomes (e.g., "my supervisor has invited me to his/her home for a dinner,
party, or informal evening"), two required speculations regarding the opinions of
others (e.g., "my supervisor's skills and abilities are respected by his/her peers or
managers"), one was inconsistent with the on-going relationship (e.g., "I would
want to keep in touch with my supervisor after we no longer work together, even
if we were not working in the same organization"), and one contained potentially
gender-bound language. The item by dimension distribution of the resulting 31
items was: contribution, 5 items; loyalty, 9 items; affect, 6 items; and professional
respect, 11 items. All LMX item responses were scaled from Strongly Disagree =
1 to Strongly Agree = 7.
Participants
Data were collected from 302 working students and 251 organizational
employees. Initial item analysis was conducted with the students samples, while
convergent, discriminant, and criterion-related validity was assessed with the
organizational samples. For all respondents, participation was voluntary and
confidentiality of responses was assured. All students who were working or had
recent work experience were asked to participate and all did so, for a response rate
of 100% (including students who were absent from class when data were collected
produces a response rate of approximately 80%). Student participants were
located at one of three universities (two in the Midwest and one in the Southeast);
either current or recent work experience was required. Approximately one-third
were enrolled in graduate-level business programs, and two-thirds were under-
graduate students. Based on the findings of Liden et al. (1993) that LMXs develop
very early (typically within the first 2 weeks) in the life of leader-member dyads,
long-term tenure with the leader was not a criterion for selection into the sample.
The demographic breakdown of the 302 working students was: 62% male, 38%
female; 71% white, 8% black, 10% Asian, 7% Hispanic, 4% other; mean age of
26 years (SD = 6.4); and a mean of 7.8 years of work experience (SD = 6.1). One-
hundred and seventy-one students received questionnaires that included the 31
LMX items, demographic items, the LMX-7 items, as well as the Acquiescence
Response Set Scale and the Crowne-Marlowe Social Desirability Scale. Question-
naires that included only the LMX items and the demographic items were
received by 131 students. Of the 302 students, 128 students received a follow-up
questionnaire 8-10 weeks after the first questionnaire which included the same
LMX items. This questionnaire was administered for the purpose of assessing
test-retest reliability.
The organizational employees sample consisted of full-time employees from
two industry segments and were surveyed as part of larger projects (Liden,
Wayne, Bradway, & Sparrowe, 1994; Sparrowe, 1994). One hundred eighty-three
were employed in the hospitality industry (response rate of 75%), and 68
(response rate of 60%) were from a large heavy equipment manufacturer. In addi-
tion, 19 managers (response rate of 94.7%) from the heavy manufacturing organi-
zation rated the performance of the 68 employees in the sample. The 183
hospitality industry employees represented multiple work groups from multiple
properties. Most of these employees worked at the lower levels of the organiza-
tions, such as front desk clerks, housekeepers, cooks, and food servers. Partici-
pants from the heavy equipment manufacturer included staff and managers from
the purchasing, travel, and product distribution departments. The demographic
breakdown of the full-time working sample was: 53% female, 41% male, and 6%
not responding; 71% white, 10% black, 5% Asian, 5% Hispanic, 2% other, and
7% preferring not to respond. The mean age of this sample was approximately
33.6 years. The average length of tenure with the organization was 7 years, 7
months.
Measures
All students responded to the 31 LMX items that survived the initial phases
of scale development. In addition, acquiescence response set (Couch & Keniston,
1960) and social desirable response bias (Crowne & Marlowe, 1960) scales were
assessed from a subset of the student participants. All organizational employees
also responded to the LMX items. In addition, employees of the heavy manufac-
turing organization responded to the LMX-7 measure, used previously to assess
LMX, and other measures employed to examine the convergent and discriminant
validity of the resulting new scale. These included satisfaction with the supervi-
sor, co-workers, and work; autonomy, and turnover intentions. Managers of the
heavy manufacturing organization also rated the job performance of the partici-
pants. Organizational commitment was assessed with the hospitality industry
employees, but not in the heavy manufacturing organization. Thus, sample sizes
varied across analyses involving the validation variables and LMX-7.
LMX.7(previous LMXseale). Seven items adapted from Scandura and
Graen (1984) were used to reflect an established measure of LMX. Response cate-
gories were slightly modified to permit scaling from Strongly Disagree = 1 to
Strongly Agree = 7 so as to more closely approximate interval scaling than the
original LMX-7 scale. This scaling format has been used in previous LMX
research (e.g., Liden, et al., 1993). The items from each of the scales were
summed to create the scale (t~ = .89 for student samples; ct = .91 for organization
samples). This scale was included for purposes of establishing convergent valid-
ity, as well as to determine incremental validity of the new scale relative to this
existing measure of LMX.
Satisfaction with supervisor. Eighteen items from the revised version of
the Job Descriptive Index (JDI; Smith, Balzer, Brannick, Eggleston, Gibson, Iron-
son, Josephson, Paul, Reilly, & Whalen, 1987) were used to measure participant
satisfaction with their supervisor as related to the supervisor's demonstration of
interest in and concern about the employee. Total scores were established by
summing the item responses such that higher values reflected greater satisfaction
with the supervisor (o~ = .89).
Satisfaction with co-workers. Eighteen items from the revised version of
the JDI (Smith, et al., 1987) were used to measure participant satisfaction with
their co-workers. This scale was included for assessment of discriminant validity
of the new LMX scales. Specifically, a higher correlation was expected between
the LMX scales and supervisor satisfaction than between the LMX scales and co-
worker satisfaction. Total scores were established by summing the item responses
such that higher values reflected greater satisfaction with co-workers (oc = .87).
Satisfaction with work. Eighteen items from the revised version of the JDI
(Smith, et al., 1987) were used to measure satisfaction with the work itself. Total
scores were established by summing the item responses such that higher values
reflected greater satisfaction with work (or = .79).
Autonomy. Three items from the Job Diagnostic Survey (Hackman &
Oldham, 1975), revised according to recommendations of Idaszak and Drasgow
(1987), were used to assess member perceptions of control in deciding when and
how to carry out tasks. Responses were summed to form a composite (o~ = .72).
Job performance. Managers rated subordinates on seven items assessing
cooperation, communications, planning and organizing, technical competence,
quality of work, quantity of work, and overall job performance. This scale repre-
sents an adaptation of this organization's performance rating scale which was
based on job analyses. Each item was scaled from 1 = minimally acceptable to 5 =
superior. All seven items were summed to form a scale (o~ = .87).
Organizational eommitment. Three items from O'Reilly & Chatman
(1986) were used to measure the affective component of organizational commit-
ment. Response categories ranged from Strongly Disagree = 1 to Strongly Agree =
7, with higher values indicating greater commitment. These items were summed
to form a composite (o~ = .80). Data on this variable were available only for the
hospitality employee sample.
Turnover intentions. Employee intentions to quit their jobs were assessed
with 3 items from the Michigan Organizational Assessment Questionnaire
(Seashore, Lawler, Mirvis, & Cammann, 1982). For example, "How likely is it
that you will actively look for a new job in the next year?", with anchors, "Not at
all likely = 1 to Extremely likely = 7." These items were summed to form a
composite (o~ = .74).
Data analysis overview. Data analysis was conducted in five phases. Given
the size and diversity of the sample, we performed item selection analyses using
the sample of working students, while the organizational employees sample was
utilized for confirmation and validation of the resulting scale. First, identification
of items that had little or no variance or were significantly correlated with either
acquiescence or social desirability, were eliminated. Second, exploratory factor
analysis was used to guide the selection of a reduced set of items. Third, confirma-
tory factor analysis was used with the organizational samples to assess the good-
ness of fit of the selected scale items with the proposed dimensions and with
competing models. Fourth, comparisons were made between the resulting dimen-
sion scales and the validation measures. And finally, to further examine the multi-
dimensionality of the new scale, differences among the four dimensions with
respect to relations with outcome variables was assessed.
Results
All of the initial item analyses, as well as exploratory factor analyses were
conducted using the working student samples.
Tests of Item Variance, Acquiescence and Social Desirability Response Sets
The first step in analyzing the data was to calculate the variance on each of
the 31 items. Items showing little variability would not be of much value and thus
would be removed from use in scale development. Although there is not an estab-
lished criterion for "adequate variability," a standard deviation of 1.0 was chosen
as representing an adequate amount of variability for usefulness as an item (L.R.
James, personal communication, November 4, 1993). All 31 items had standard
deviations exceeding 1.0 with a range from 1.19 to 2.08, and thus no items were
removed for lack of variability.
Next, all items were correlated with the Acquiescence Response Set scale
(N = 162) and with the Crowne-Marlowe Social Desirability scale (N = 161). The
2 items that correlated significantly with the Acquiescence Response Set scale and
the 3 items that correlated significantly with the Crowne-Marlowe Social Desirabil-
ity scale were removed• These 5 significant correlations ranged in size from. 15 to
•18 ( all p < .05). At this point, 26 items remained for consideration in scale devel-
opment.
Factor Analyses
Using data collected from the sample of 302 working students, we conducted
an exploratory factor analysis using principal components with the number of factors
not specified• The magnitude and scree plot of the eigenvalues indicated four factors.
In the next factor analysis, we set the number of factors to four and interpreted factor
loadings based on the pattern matrix which resulted from oblique rotation (Tabach-
nick & Fidell, 1989)• Oblique rotation was appropriate because of the anticipated
relation among the factors (cf. Dienesch & Liden, 1986). Analysis of the 26 items
resulted in four factors that explained 69.1% of the variance. Based on the oblique
factor pattern, each factor clearly reflected one of the four a priori dimensions•
Subsequent iterations were performed following deletion of cross-loaded items or
items that were theoretically inconsistent with their factor. The resulting solution
consisted of 11 items explaining 79.4% of the variance. The breakdown of these
items was affect (3 items), loyalty (3 items), contribution (2 items), and professional
respect (3 items). The rotated factor loadings for these 11 items appear in Table 2.
Although the exploratory analysis resulted in groups of items reflective of
the four a priori dimensions, the extent to which multidimensionality was truly
captured needed to be verified. To further assess multidimensionality, conf'trma-
tory factor analysis was performed. Differing from exploratory factor analysis,
confirmatory factor analysis allows for better integration of theory and measure-
ment (Hughes, Price, & Marrs, 1986; James, Mulaik, & Brett, 1982). This is
accomplished through specific constraints imposed by the researcher based on
relevant theory (Long, 1983; McDonald & Marsh, 1990). As noted by Church and
Burke, "Whereas exploratory methods are advantageous when no a priori struc-
ture can be hypothesized, confirmatory methods provide significance tests and
goodness-of-fit indices for hypothesized models" (1994, p. 93).
Differences Between Organizations
Prior to merging the two organizational samples for the confirmatory analy-
ses, we tested for differences between the two organizations on the four LMX
Professional
Item Affect Loyalty Contribution Respect
1 like my supervisor very much as a person. .91 .02 .10 --.03
My supervisor is the kind of person one would like to
have as a friend. .80 .09 -.01 .11
My supervisor is a lot of fun to work with. .72 .l I - . l0 .22
My supervisor defends my work actions to a superior,
even without complete knowledge of the issue in -.08 .91 -.05 -.07
question.
My supervisor would come to my defense if I were .13 .74 .06 .05
"attacked" by others.
My supervisor would defend me to others in the organi- .15 .70 .06 .07
zation if I made an honest mistake.
| do work for my supervisor that goes beyond what is
.21 -.06 .86 - . 13
specified in my job description.
I am willing to apply extra efforts, beyond those nor-
mally required, to further the interests of my work - . 19 .l0 .81 .18
group.
I am impressed with my supervisor's knowledge of his/ - . 0 3 -.03 .00 .97
her job.
I respect my supervisor's knowledge of and competence .05 --.01 --.01 .91
on the job.
I admire my supervisor's professional skills. .13 .06 .05 .79
factors that emerged from the exploratory factor analyses. Means were signifi-
cantly higher in the hospitality sample than the manufacturing sample for affect
(t = 4.48, p < .01), loyalty (t = 2.17, p < .05), and professional respect (t = 2.91, p
< .01), and approached significance for contribution (t = 1.78, p < .08). However,
of greater importance than mean differences in scale development efforts, no
significant differences were found in the intercorrelations between the dimensions
as shown in Table 3. For example, of all correlations reported in Table 3, the
correlation between affect and contribution differed most between the manufac-
turing (r = .49) and the hospitality (r = .30) samples. Applying the formula for
testing the difference between correlations in independent samples (Hays~ 1973),
showed this difference in correlations to be nonsignificant (z = 1.56, ns). Given
these results, we combined the two samples for all further analyses.
Confirmatory Factor Analysis
Item selection was executed with the student samples, but all confirmatory
analyses were conducted exclusively with the data collected from the 251 organi-
zational participants (less 2 due to missing data, for a total N = 249). As in previ-
ous uses of confirmatory factor analysis in assessing construct dimensionality
(e.g., Hackett, Bycio, & Hausdorf, 1994; Meyer, Allen, & Smith, 1993; Rahim &
Magner, 1995), the models were tested with correlated factors and uncorrelated
error. The 4-factor, or hypothesized model, reflected the four dimensions of
affect, loyalty, contribution, and professional respect, as defined by the four origi-
nal a priori categories for the 11 chosen scale items.
The goodness-of-fit of a 4-factor model was tested in comparison to compet-
ing models through use of the LISREL 7 program in SPSS (J6reskog & S6rbom,
1989). Mulaik, James, Van Alstine, Bennett, Lind, & Stilwell (1989) noted that
even good fitting models can suffer from misspecification, suggesting that alter-
native models be considered when using structural equation modeling. The
competing models were: 1) the null model in which none of the variables are
related; 2) a single factor model, in which all items represent a single dimension.
Support for this model over all others would confirm the proposition of Graen and
his colleagues that LMX is a unidimensional construct (Graen & Scandura, 1987;
Graen & Uhl-Bien, 1995); 3) a 2-factor model composed of one factor containing
the work-related scales, contribution and professional respect, with the second
factor including the affect and loyalty items. Support for this model would suggest
that respondents are not able to distinguish between four dimensions, but only
between more objective aspects of the relationship that involve the work itself and
subjective aspects involving liking and loyalty; and 4) a 3-factor model in which
one factor contains the same work-related scale (combination of the contribution
and professional respect items) as the 2-factor model, but affect and loyalty are
retained as separate factors. In other words, the 3-factor model is the same as the
hypothesized model except that the work-related scales, contribution and profes-
sional respect are combined into 6ne factor. This model approximates that
described by Dienesch and Liden (1986) and parallels the three-dimensional LMX
scale developed by Schriesheim et al. (1992).
Several statistics were used to assess the relative fit between the obtained
data and each of the competing models. Although the X2 for the 4-factor model
was significant, signaling a difference between the hypothesized and observed
structures, this is often the case with large samples (Bentler & Bonett, 1980).
Thus, we used other recommended measures to determine fit (Medsker, Williams,
& Holahan, 1994). Goodness-of-fit statistics appearing in Table 4 indicate that.the
hypothesized 4-factor model provides a good fit (CFI = .986; GFI = .960; AGFI =
.930), and a better fit than the alternative models tested. Of the fit statistics
presented, the comparative fit index (CFI), which is a modified version of the
normed fit index (NFI), is the index that best accounts for parsimony without
over-penalizing models that hypothesize more paths (Bentler, 1990; Williams &
Holahan, 1994). The goodness-of-fit (GFI) and adjusted goodness-of-fit (AGFI)
indices are presented for purposes of comparison with previous research.
As noted, alternative models may also adequately fit the observed relation-
ships. The fit of the hypothesized model relative to the competing models was
also assessed by examining the significance of the AX2 for adjacent models based
on their degrees of freedom. That is, differences between the: null and the 1-factor
models; 1-factor and 2-factor models; 2-factor and 3-factor models; and the 3-
factor and hypothesized or 4-factor models (see Anderson & Gerbing, 1988). A
significant difference in g2 between models indicates that the specified relations
among the variables in one model are a significantly better representation of the
data under analysis. As shown in Table 4, each of the AX2 tests was significant,
demonstrating the superiority of the 4-factor model over all competing models,
Providing further support for the hypothesized model are the completely stan-
dardized factor loadings (lambdas), ranging from .551 to .875 (shown in Figure 1).
With completely standardized estimates, both indicators and latent variables are
standardized. All of the factor loadings were statistically significant with t's rang-
ing from 7.10 to 16.96. Especially important for demonstrating multidimensional-
ity are the factor correlations (phi coefficients), because support for the 4-factor
model simply indicates that the factor correlations are less than 1.0. The factor
correlations from thephi matrix, also reported in Figure 1, range from .456 to .763.
Although moderately high, they do not reflect redundancy between the four dimen-
sions. Latent variable reliabilities were .90, .78, .59, .89, respectively, for affect,
loyalty, contribution, and professional respect. Completely standardized theta
deltas are also presented in Figure 1. Completely standardized theta deltas provide
an estimate of the error present in each item such that 1-theta delta = reliability of
the item. Theta deltas can also be derived by subtracting the squared completely
standardized lambdas from one.
Creation of Scales
Scales were created by summing the raw scores (i.e., unit weighting) for
selected items loading on each of the 4 factors, and dividing by the total number
of items representing that factor. The new scale was named LMX-MDM for
multidimensional measure of leader-member exchange. Table 5 shows the inter-
correlations between the four scales measuring LMX dimensions for both the
student and employee samples. The four new LMX scales were moderately corre-
lated, with coefficients ranging from .26 to .63 in the student samples and .32 to
.67 in the organizational employee samples.
Reliability
Internal consistency reliabilities were acceptable for the affect, loyalty, and
professional respect scales but low for the contribution scale. Coefficient alphas
were .90, .78, .60 and .92, respectively, for affect, loyalty, contribution, and
professional respect in the student samples, and .90, .74, .57, and .89, respectively,
for affect, loyalty, contribution, and professional respect for the organizational
employee samples. Test-retest correlations for the subsample of 126 students were
.83, .66, .56, and .79, respectively, for affect, loyalty, contribution, and profes-
sional respect.
Validity
Validity of the the LMX-MDM may be derived from support for the 4-factor
model using exploratory factor analysis and confirmation using CFA with inde-
pendent samples (Bagozzi, Yi, & Phillips, 1991; Rahim & Magner, 1995).
JOURNAL OF MANAGEMENT, VOL. 24, NO. 1, 1998
60 ROBERT C. LIDEN AND JOHN M. MASLYN
63 .234 ~l .875
~12.763
84 .693
63 400 ~ l ~ l ~ %2.775
t.oyalty
86 .267 2.856
~3.563
~4.740
57 .455
~3.738
58 .697 ~3.551
~4.456 /
~..285 ~ _..._.~ ///
til1.248
Figure 1. ConfirmatoryFactorAnalysisResults:
CompletelyStandardizedThetaDeltas,Lambdas,and
Phi Coefficientsfor the Hypothesized4-FactorModel
JOURNAL OF MANAGEMENT, VOL. 24, NO. 1, 1998
MULT1DIMENSIONALITY OF LMX 61
Note: a. Mean values can range from 1 to 7. Values left and below the diagonal represent the development
sample of working students; N ranges from 301-302. Values right and above the diagonal represent the
validation sample of organizational employees, N = 249. All correlations are significant at p < .01.
However, consistent with Schmitt and Klimoski's (1991) argument that validity is
best assessed using multiple approaches, we examined our scale with respect to
response bias susceptibility, convergent validity, discriminant validity, and crite-
rion-related validity.
Response bias susceptibility. Validity of a scale may be questioned if it is
prone to response biases. Although none of the retained items correlated signifi-
cantly with acquiescence or social desirability response sets, it is possible that the
cumulative effects of combining items to form scales might render significant
correlations with response bias tendencies. Correlations between the four LMX
dimensions and acquiescence response set scores were not significant: -.01, .04,
-. 10, and -.01 for affect, loyalty, contribution, and professional respect, respec-
tively. Correlations between social desirability and affect (.04), loyalty (.08), and
professional respect (.01) were not significant. The correlation between social
desirability and contribution was significant (.16, p < .05), but not of practical
significance based on explained variance (.162 = .026). In sum, these results
suggest that the LMX-MDM scales are not susceptible to two common response
biases (cf. Moorman & Podsakoff, 1992).
Convergent. Perhaps the most common method for establishing conver-
gent validity is to compare factor loadings from exploratory factor analysis and/or
CFA with loadings that would be expected based on theory. The exploratory and
confirmatory factor analysis results strongly supported our four a priori dimen-
sions, providing solid evidence of convergent validity. Convergent validity can
also be tested by examining correlations with constructs that should, based on
theory or past empirical findings, be significantly correlated (Campbell & Fiske,
1959). Certainly, a "new" scale of a construct should correlate with an existing
measure of that same construct (DeVellis, 1991). Thus, we expected that the
dimensions of the LMX-MDM would be significantly correlated with the LMX-7
measure. This expectation was confirmed by correlations of .71, .71, .55, and .70
(and .64, .53, .33, and .42 in the student samples) between LMX-7 and affect,
loyalty, contribution, and professional respect (respectively).
Discriminant. Support for the 4-factor model using exploratory factor
analysis and confirmation using CFA with an independent sample provides
support for discriminant validity (Rahim & Magner, 1995). Discriminant validity
can also be tested by examining correlations with constructs that should, based on
theory or past empirical findings, not be significantly correlated (Campbell &
Fiske, 1959). Because LMX focuses on the relationship between a supervisor and
subordinate, comparisons of the LMX scales with satisfaction with co-workers
would provide evidence of discriminant validity. Correlations of. 12, .25, .00, and
.00 between satisfaction with co-workers and affect, loyalty, contribution, and
professional respect, respectively, provide support for the discriminant validity of
the LMX-MDM.
Criterion-related. Another way to assess the validity of our new measure is
to determine whether the four dimensions are differentially related to various
outcomes consistent with theory and research. Expected relations between the
three Dienesch and Liden (1986) dimensions were presented above, and the same
may be done for the new dimension, professional respect, which was identified in
the critical incidents interviews. A global outcome such as satisfaction with super-
vision, which encompasses multiple beliefs about the leader, is expected to be
positively related to professional respect as well as to affect, loyalty, and contribu-
tion. We would also expect that professional respect would be positively related to
performance ratings. The more a member believes that the leader commands
respect as a professional in the field, the more that member would be expected to
contribute to the work unit and be rated higher on performance.
To test ability of the LMX scales to discriminate among various outcomes,
the outcome variables were separately regressed on the four scales using simulta-
neous entry. This method indicates whether a dimension of LMX helped to
explain variance in the dependent variable after accounting for the effects of the
other dimensions. As shown in Table 6, for four of the six outcome variables, the
LMX-MDM scales as a group were significant. More importantly, the results also
provide support for the multidimensionality of the LMX construct. For each
outcome, only a subset of the dimensions were significant, and each of the dimen-
sions was significant for at least one of the outcomes. However, caution should be
used in interpreting the results for the contribution dimension given its low (.57)
internal consistency reliability.
The LMX dimensions with a significant beta weight varied depending on the
outcome variable and were generally consistent with the expectations noted
above. For example, beta weights for affect, loyalty, and professional respect, but
not contribution were significant for satisfaction with supervision. Beta weights
for affect and loyalty toward the supervisor were not significant with respect to
organizational commitment, but the beta weights for contribution and professional
respect were significant. Although the regression model for turnover intentions
was significant, none of the beta weights were significant (the beta weight for
loyalty approached statistical significance at p < .08). Of all outcomes measured,
performance ratings is of greatest interest because it was assessed by the subordi-
nates' managers, thus reducing the effect of common method variance biases.
Variance in subordinate performance ratings was explained by contribution and
loyalty, but not by affect. Professional respect appeared to play a suppressor role
as it was not correlated with performance (r = .05) but had a marginally signifi-
cant (p < .07) beta weight.
Table 6. R e g r e s s i o n A n a l y s e s :
O u t c o m e Variables R e g r e s s e d on the F o u r L M X D i m e n s i o n s
sample (N = 68), they suggest that LMX-MDM may explain incremental variance
in some outcome variables beyond that explained by LMX-7.
In some studies, such as those in which LMX is not a key variable, LMX
dimensionality may not be a concern and a unidimensional measure may suffice.
Researchers who desire a measure of global LMX could combine all 11 items into
a composite. In our organizational sample, the reliability for the 11-item compos-
ite was .89 and it correlated ,84 with LMX-7. Alternatively, when using structural
equations modeling, researchers could use each of the four dimensions as indica-
tors of global LMX. Providing support for this approach, results of a higher-order
CFA showed that each of the four dimensions contributes to global LMX. 3
Discussion
Through the validation of a multidimensional measure of LMX (labeled
LMX-MDM), support was found for the proposition that LMX is composed of
multiple dimensions as originally suggested by Dienesch and Liden (1986). Four
factors, consistent with a priori categories based on theory and critical incident
interviews, were found using exploratory factor analysis. Confirmatory factor
analyses with independent samples of organizational employees showed a signifi-
cantly better fit for the proposed four-dimensional model than the null model and
the one-, two-, or three-factor models. The affect, loyalty, and professional respect
scales showed high reliability assessed both with internal consistency and test-
retest methods. Further scale development is needed for the contribution dimen-
sion to add an additional item and to increase reliability. With respect to validity,
all four scales were shown to be unrelated to acquiescence bias. Affect, loyalty,
and professional respect were also found to be unrelated to social desirability, and
although contribution was significantly related to social desirability, the effect
was small in magnitude (r 2 = .026), Enhancing convergent validity beyond that
shown by the exploratory and confirmatory factor analyses, the LMX-MDM
scales were shown to be correlated LMX-7. Similarly, discriminant validity
beyond that shown by the factor analysis results was demonstrated by the small
correlations with less theoretically related constructs such as satisfaction with co-
workers. Further support with LMX as a multidimensional construct was provided
by regression results showing that each dimension contributed differentially in the
explanation of variance in each of the outcome variables.
In summary, support for LMX as a multidimensional construct was provided
by a consistent set of results: 1) factor loadings from exploratory factor analysis
provided support for four separate factors; 2) the CFA results showed the four-
dimensional model to be superior to competing models, including the unidimen-
sional model advocated by Graen and colleagues (Graen & Scandura, 1987;
Graen & Uhl-Bien, 1995); 3) the four dimensions correlated with theoretically
similar variables such as satisfaction with supervision and showed small or zero
correlations with theoretically dissimilar variables such as satisfaction with co-
workers and 4) regression results indicated that different LMX dimensions were
significant in the explanation of variance in outcome variables.
One weakness of the study was the size of the sample used for the critical
incident interviews, which were designed to determine the relevance of the affect,
loyalty, and contribution dimensions of LMX proposed by Dienesch and Liden
(1986) and to identify other possible dimensions of LMX. Although a sample
larger than 24 would have been preferred, this step in item generation is typically
skipped in scale development efforts. Another weakness is that both organizations
involved in the validation segment of the study were for-profit, private sector
organizations. To extend generalizability, the new scale needs to be used in a wide
variety of public and private sector organizations. Also, although data were
collected from two sources (in the manufacturing sample), no data were collected
from sources independent of the dyad. For example, performance data are suspect
because they are from leader reports. Performance data from an independent
source or an objective measure would be desirable (cf. Duarte et al., 1994).
The reliability for the contribution dimension was not acceptable. Despite
writing 30 items to capture contribution, only two items survived all stages of
scale development. Many of the items dealing with member perceptions of the
leader's behaviors that reflected contribution to work did not survive content vali-
dation. Others tended to cross-load on other dimensions, such as professional
respect. The items asking members to report their own work-related contributions~
of which two items survived, appeared to suffer from leniency bias similar to that
uncovered in research on self-performance ratings (Farh & Dobbins, 1989; John
& Robins, 1994; Levy, 1993). An additional issue related to reliability is that it
would be preferable for the contribution scale to consist of three items as is true of
the other three dimensions. This is especially important for uses of the scale
involving structural equations modeling. It has been recommended that with
multiple indicator models, each latent variable should have at least three indica-
tors (Harris & Schaubroeck, 1990).
One of the main strengths of this investigation was the thorough process used
in creating the item pool. Many of the items were based on interviews designed
expressly for this study, and it was through the interview process that the profes-
sional respect dimension was identified. Previous LMX theory had not recognized
the importance of this dimension. Also improving upon previous LMX scale
development efforts (Dienesch, 1985; Phillips, et al., 1993; Schriesheim, et al.,
1992) was the rigorous content validation procedure involving 14 faculty and
doctoral students, and the evaluation of all items and scales for social desirability
and acquiescence response sets. Finding four factors that matched our a priori
dimensions using the conservative approach of exploratory factor analysis with
the number of factors not specified provided strong support for the hypothesized
multidimensionality of LMX (DeVellis, 1991). In addition, the scales developed
from a sample of working undergraduates and MBA students were validated using
CFA with data collected from an independent sample of employees representing
two diverse organizations.
Implications and Future Research
The value of identifying multiple dimensions of LMX lies in understanding
when and how these dimensions relate to issues of development and maintenance
JOURNAL OF MANAGEMENT. VOL. 24, NO. 1, 1998
MULTIDIMENSIONALITY OF LMX 67
Addendum
Since the current manuscript was accepted for publication, further psycho-
metric research was conducted with the purpose of improving measurement of the
"Contribution" dimension. This dimension now contains three items, the same as
the other dimensions. The first item below remains unchanged. The second item
has been modified from the version that appears in Table 2. The third item below
is new. The anchors for these items remain Strongly Disagree = 1 to Strongly
Agree = 7. The three Contribution items are:
1. I do work for my supervisor that goes beyond what is specified in my job
description.
2. I am willing to apply extra efforts, beyond those normally required, to
meet my supervisor's work goals.
3. I do not mind working my hardest for my supervisor.
Notes
JOURNAL OF MANAGEMENT, VOL. 24, NO. 1, 1998
MULTIDIMENS1ONALITY OF LMX 69
References
Anderson, J. C., & Gerbing, D. W. (I 988). Structural equation modeling in practice: A review and recommended
two-step approach. Psychological Bulletin, 103:411--423•
Bagozzi, R. P., Yi, Y., & Phillips, L. W. (1991)• Assessing construct validity in organizational research. Admin-
istrative Science Quarterly, 36: 421-458•
Bales, R. F. (1958). Task roles and social roles in problem-solving groups• In E. E. Maccoby, T. M. Newcomb &
E. L. Hartley (Eds.), Readings in social psychology, (pp. 437--447). New York: Holt and Company.
Barnard, C. 1. (1938). The functions of the executive• Cambridge, MA: Harvard University Press.
Bass. B. M. ( 1981 ). Stogdill's handbook of leadership: A survey of theory and research, (2nd edition). New York:
The Free Press.
• (1990). Bass & Stogdill's handbook of leadership: Theory. research, and managerial applications (3rd
edition). New York: The Free Press.
Bauer, T. N., & Green, S. G. (1996). The development of leader-member exchange: A longitudinal test. Academy
of Management Journal, 39: 1538--1567.
Bentler, P. M. (1990). Comparative fit indexes in structural models• Psychological Bulletin, 107: 238--246.
Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness of fit in the analysis of covariance struc-
tures. Psychological Bulletin, 88: 588--606.
Blau, P. M. (1964). Exchange andpower in social life. New York: Wiley•
Bollen, K. A., & Hoyle, R. H. (1990). Perceived cohesion: A conceptual and empirical examination. Social
Forces, 69: 479--504.
Brass, D. J., & Burkhardt, M. E. (1992). Centrality and power in organizations. In R. G. Eccles & N. Nohria (Eds.),
Networks and organizations: Structure, form, and action, (pp. 191--215). Boston: Harvard Business
School Press.
Bridge, K., & Baxter, L. A. (1992). Blended relationships: Friends as work associates. Western Journal of
Communications, 56: 201)--225.
Butler, J. K. (1991 ). Toward understanding and measuring conditions of trust: Evolution of a conditions of trust
inventory. Journal of Management, 17: 643--663.
Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminate validation by the multitrait-multimethod
matrix• Psychological Bulletin, 56:81-- 105.
Church, A. T., & Burke, P. J. (1994). Exploratory and confirmatory tests of the big five and Tellegen's three-and
four-dimensional models. Journal of Personality and Social Psychology, 66: 93--114.
Couch, A., & Keniston, K. (1960). Yeasayers and naysayers: Agreeing response set as a personality variable•
Journal of Abnormal and Social Psychology, 60:151-- 174.
Crowne, D. P., & Marlowe, D. A. (1960)• A new scale of social desirability independent of psychopathology.
Journal ~f Consulting Psychology, 24: 349--354.
Dansereau, F., Graen, G., & Haga, W. (1975). A vertical dyad approach to leadership within formal organizations.
Organizational Behavior and Human Performance. 13: 46--78.
DeVellis, R. F. ( 1991 ). Scale development: Theory and applications. Newbury Park, CA: Sage.
Dienesch, R. M. (1985, August)• A three dimensional model of leader-member exchange: An empirical test.
Paper presented at the Academy of Management meeting, San Diego, CA.
Dienesch, R. M., & Liden, R. C. (1986). Leader-member exchange model of leadership: A critique and further
development. Academy of Management Review. 11: 618--634•
Dockery, T. M., & Steiner, D. D. (1990)• The role of the initial interaction in leader-member exchange. Group
and Organization Studies, 15: 395--413•
Duarte. N. T., Goodson, J. R., & Klich, N. R. (1993). How do I like thee? Let me appraise the ways. Journal of
Organizational Behavior, 14: 239"-249.
. ( 1994)• Effects of dyadic quality and duration on performance appraisal. Academy of Management Jour-
nal, 37: 499--521.
Duchon, D., Green, S. G., & Taber, T. D. (1986). Vertical dyad linkage: A longitudinal assessment of antecedents,
measures, and consequences. Journal of Applied Psychology, 71: 56--60•
Dunegan. K. J., Duchon, D., & Uhl-Bien, M. (1992). Examining the link between leader-member exchange and
subordinate performance: The role of task analyzability and variety as moderators. Journal of Manage-
ment, 18: 59--76.
Fahr, J. L., & Dobbins, G. H. (1989). Effects of self-esteem on leniency bias in self-reports of performance: A
structural equation model analysis. Personnel Psychology, 42: 835--850.
Fairhurst, G. T., & Chandler, T. A. (1989). Social structure in leader-member interaction. Communication Mono-
graphs, 56: 215--239.
Fernandez, R. M. (1991). Structural bases of leadership in intraorganizational networks. Social Psychology Quar-
terly, 54: 36-53.
Ferris, G. R. (1985). Role of leadership in the employee withdrawal process: A constructive replication. Journal
of Applied Psychology, 70: 777--781.
French, J. R. P., & Raven, B. H. (1959). The bases of social power. In D. Cartwright (Ed.), Studies in socialpower,
(pp. 150--167). Ann Arbor: University of Michigan Press.
Gergen, K. J., Greenberg, M. S., & Willis, R. H. (1980). Social exchange: Advances in theory and research. New
York: Plenum.
Gerstner, C. R., & Day, D. V. (1997). Meta-analytic review of leader-member exchange theory: Correlates and
construct issues. Journal of Applied Psychology, 82: 827--844.
Gouldner, A. W. (1960). The norm of reciprocity: A preliminary statement. American Sociological Review, 25:
161--177.
Graen, G. (1976). Role-making processes within complex organizations. In M.D. Dunnette (Ed.), Handbook of
industrial and organizational psychology, (pp. 1201-- 1245). Chicago: Rand McNally.
Graen, G., & Cashman, J. F. (1975). A role making model in formal organizations: A developmental approach. In
J. G. Hunt & L. L. Larson (Eds.), Leadership Frontiers, (pp. 143--165). Kent, OH: Kent State Press.
Graen, G., Liden, R. C., & Hoel, W. (1982). Role of leadership in the employee withdrawal process. Journal of
Applied Psychology, 67: 868--872.
Graen, G., Novak, M. A., & Sommerkamp, P. (1982). The effects of leader-member exchange and job design on
productivity and satisfaction: Testing a dual attachment model. Organizational Behavior and Human
Performance, 30: 109--131.
Graen, G. B., & Scandura, T. A. (1987). Toward a psychology of dyadic organizing. Research in Organizational
Behavior, 9: 175--208.
Graen, G. B., & Uhl-Bien, M. (1995). Development of leader-member exchange (LMX) theory of leadership over
25 years: Applying a multi-level mult-domain perspective. Leadership Quarterly, 6: 219-247.
Hackett, R. D., Bycio, P., & Hausdorf, P. A. (1994). Further assessments of Meyer and Allen's (1991) three-
component model of organizational commitment. Journal of Applied Psychology, 79:15--23.
Hackman, R. J., & Oldham, G. R. (1975). Development of the job diagnostic survey. Journal of Applied Psychol-
ogy, 60: 159--170.
Harris, M. M., & Schaubroeck, J. (1990). Confirmatory modeling in organizational behavior/human resource
management: Issues and applications. Journal of Management, 16: 337--360.
Hays, W. L. (1973). Statistics for the social sciences. New York: Holt, Rinehart & Winston.
Hinkin, T. R. (1995). A review of scale development practices in the study of organizations. Journal of Manage-
ment, 21: 967--988.
Hollander, E. P. (1980). Leadership and social exchange processes. In K. J. Gergen, M. S. Greenberg, & R. H.
Willis (Eds.), Social exchange: Advances in theory and research, (pp. 103--118). New York: Plenum.
Homans, G.C. (1958). Social behavior as exchange. American Journal of Sociology, 63: 597--606.
Hughes, M.A., Price, R.L., & Marrs, D.W. (1986). Linking theory construction and theory testing: Models with
multiple indicators of latent variables. Academy of Management Review, 11: 128--144.
lbarra, H. (1993). Network centrality, power, and innovation involvement: Determinants of technical and admin-
istrative roles. Academy of Management Journal, 36:471--501.
Idaszak, J. R., & Drasgow, F. (1987). A revision of the job diagnostic survey: Elimination of a measurement arti-
fact. Journal of Applied Psychology, 72: 69"-74.
Jacobs, T.O. ( 1971 ). Leadership and exchange informal organizations. Alexandria, VA: HumRRO.
James, L.R., Mulaik, S.A., & Brett. J.B. (1982). Causal analysis: Assumptions, models, and data. Beverly Hills,
CA: Sage Publications.
John, O. P., & Robins, R. W. (1994). Accuracy and bias in self-perception: Individual differences in self-enhance-
ment and the role of narcissism. Journal of Personality and Social Psychology, 66: 206--219.
Judge, T. A., & Ferris, G. R. (1993). Social context of performance evaluation decisions. Academy of Management
Journal, 36:80--105.
J6reskog, K.G., & S6rbom, D (1989). Lisrel 7: A guide to the program and applications. Chicago: SPSS, Inc.
Katerberg, R., &Hom, P.W. (1981). Effects of within-group and between groups variation in leadership. Journal
of Applied Psychology, 66:218--223.
Katz, D., & Kahn, R.L. (1978). The socialpsychology of organizations. (2nd ed.). New York: Wiley.
Keller, T., & Dansereau, F. (1995). Leadership and empowerment: A social exchange perspective. Human Rela-
tions, 48: 127--145.
Kim, H., & Yukl, G. (1995). Relationships of managerial effectiveness and advancement to self-reported and
subordinate- reported leadership behaviors from the multiple linkage model. Leadership Quarterly, 6:
361-377.
Kozlowski, S.W., & Doherty, M.L. (1989). Integration of climate and leadership: Examination of a neglected
issue. Journal of Applied Psychology, 74: 546--553.
Krackhardt, D. (1990). Assessing the political landscape: Structure, cognition, and power in organizations.
Administrative Science Quarterly, 35: 342-369.
Kramer, M. W. (1995). A longitudinal study of superior-subordinate communication during job transfers. Human
Communication Research, 22: 39--64.
Kumar, K., & Beyerlein, M. (1991). Construction and validation of an instrument for measuring ingratiatory
behaviors in organizational settings. Journal of Applied Psychology, 76: 619--627.
Levy, P. E. (1993). Self-appraisal and attributions: A test of a model. Journal of Management, 19:51--62.
Liden, R. C., & Graen, G. (1980). Generalizability of the vertical dyad linkage model of leadership. Academy of
Management Journal, 23:451--465.
Liden, R. C., Sparrowe, R. T., & Wayne, S. J. (1997). Leader-member exchange theory: The past and potential
for the future. Research in Personnel and Human Resources Management, 15:47--119.
Liden, R. C., Wayne, S. J., Bradway, L. K., & Sparrowe, R. T. (1994, April). Leading empowered work groups.
Paper presented as part of a symposium at the annual meetings of the Society of Industrial and Organiza-
tional Psychologists, Nashville, TN.
Liden, R. C., Wayne, S. J., & Stilwell, D. (1993). A longitudinal study on the early development of leader-member
exchanges. Journal of Applied Psychology, 78: 662--674.
Long, J.S. (1983). Confirmatory factor analysis. Beverly Hills, CA: Sage Publications.
McDonald, R. P., & Marsh, H. W. (1990). Choosing a multivariate model: Noncentrality and goodness of fit.
Psychological Bulletin, 107: 247--255.
Medsker, G. J., Williams, L. J., & Holahan, P.J. (1994). A review of current practices for evaluating causal models
in organizational behavior and human resources management research. Journal of Management, 20: 439--
464.
Meyer, J. P., Allen, N. J., & Smith, C. A. (1993). Commitment to organizations and occupations: Extension and
test of a three-component conceptualization. Journal of Applied Psychology, 78:538--551.
Moorman, R.H., & Podsakoff, P.M. (1992). A meta-analytic review and empirical test of the potential contbund-
ing effects of social desirability response sets in organizational behaviour research. Journal of Occupa-
tional & Organizational Psychology, 65:131--149.
Mulaik, S. A., James, L. R., Van AIstine, J., Bennett, N., Lind, S., & Stilwell, C. D. (1989). An evaluation of good-
ness of fit indices for structural equation models. Psychological Bulletin, 105: 430--445.
Novak, M.A. (1985). A study of leader resources as determinants of leader-member exchange. Doctoral disserta-
tion, Universi~' of Cincinnati.
O'Reilly, C. III, & Chatman, J. (1986). Organizational commitment and psychological attachment: The effects of
compliance, identification, and internalization on prosocial behavior. Journal of Applied Psychology, 71:
492--499.
Phillips, R.L., Duran, C.A., & Howell, R.D. (1993, November). An examination of the multidimensionality
hypothesis of leader member exchange, using both factor analytic and structural modeling techniques.
Paper presented at the meetings of the Southern Management Association, Atlanta, GA.
Rahim, M. A., & Magner, N. R. (1995). Confirmatory factor analyses of the styles of handling interpersonal
conflict: First-order factor model and its invariance across groups. Journal of Applied Psychology, 80:
122--132.
Rees, C. R., & Segal, M. W. (1984). Role differentiation in groups: The relationship between instrumental and
expressive leadership. Small Group Behavior, 15: 109--123.
Sahlins, M. (1972). Stone age economics. New York: Aldine De Gruyter.
Scandura, T. A., & Graen, G. B. (1984). Moderating effects of initial leader-member exchange status on the effects
of a leadership intervention. Journal of Applied Psychology, 69: 428--436.
Scandura, T. A., Graen, G. B., & Novak, M,A. (1986). When managers decide not to decide autocratically: An
investigation of leader-member exchange and decision influence. Journal of Applied Psychology, 71: 579--
584.
Schmitt, N. W., & Klimoski, R. J. (1991). Research methods in human resources management. Cincinnati, OH:
South- Western Publishing.
Schriesheim, C.A., & Gardiner, C.C. (1992). An exploration of the discriminant validity of the leader-member
exchange scale (LMX7) commonly used in organizational research. In M. Schnake (Ed.), Proceedings of
the Southern Management Association, (pp. 91--93). Valdosta, GA: Southern Management Association.
Schriesheim, C.A., Neider, L.L., Scandura, T.A., & Tepper, B.J. (1992). Development and preliminary validation
of a new scale (LMX-6) to measure leader-member exchange in organizations. Educational and Psycho-
logical Measurement, 52:135-- 147.
Schriesheim, C.A., Powers, K.J., Scandura, T.A., Gardiner, C.C., & Lankau, M.J. (1993). Improving construct
measurement in management research: Comments and a quantitative approach for assessing the theoretical
content adequacy of paper-and-pencil survey-type instruments. Journal of Management, 19:385--417.
Seashore, S.E., Lawler, E.E., Mirvis, P., & Cammann, C. (Eds.) (1982). Observing and measuring organizational
change: A guide to field practice. New York: Wiley.
Seers, A., & Graen, G. B. (1984). The dual attachment concept: A longitudinal investigation of the combination
of task characteristics and leader-member exchange. Organizational Behavior and Human Performance,
33: 283--306.
Settoon, R.P., Bennett, N., & Liden, R.C. (1996). Social exchange in organizations: The differential effects of
perceived organizational support and leader member exchange. Journal of Applied Psychology, 81: 219--
227.
Shore, L. M., & Wayne, S. J. (1993). Commitment and employee behavior: Comparison of affective commitment
and continuance commitment with perceived organizational support. Journal of Applied Psychology, 78:
774--780.
Smith, P. C., Balzer, W., Brannick, M., Eggleston, S., Gibson, W., Ironson, G., Josephson, H., Paul, K., Reilly,
C., & Whalen, M. (1987). Manual for the revised JDI and Job in General Scales. Bowling Green, OH:
Department of Psychology, Bowling Green State University.
Sparrowe, R. T. (1994). Empowerment in the hospitality industry: An exploration of antecedents and outcomes.
Hospitality Research Journal, 17:51--73.
Sparrowe, R. T., & Liden, R.C. (1997). Process and structure in leader-member exchange. Academy of Manage-
ment Review, 22: 522--552.
Tabachnick, B. G., & Fidell, L. S. (1989). Using multivariate statistics. New York: Harper-Collins.
Thibaut, J. W., & Kelley, H. H. (1959). The socialpsychology of groups. New York: John Wiley & Sons.
Tsui, A. S. (1984). A role set analysis of managerial reputation. Organizational Behavior and Human Perfor-
mance, 34: 64--96.
Tsui, A. & Barry, B. (1986). Interpersonal affect and rating errors. Academy of Management Journal, 29: 586-598.
Vecchio, R. P. (1985). Predicting employee turnover from leader-member exchange: A failure to replicate. Acad-
emy of Management Journal, 28: 478--485.
Vecchio, R. P., & Gobdel, B. C. (1984). The vertical dyad linkage model of leadership: Problems and prospects.
Organizational Behavior and Human Performance, 34: 5--20.
Wakabayashi, M., & Graen, G. (1984). The Japanese career progress study: A seven year follow-up. Journal of
Applied Psychology, 73: 217--227.
Wakabayashi, M., Graen, G., Graen, M., & Graen, M. (1988). Japanese management progress: Mobility into
middle management. Journal of Applied Psychology, 73:217--227.
Wakabayashi, M., Graen, G., & Uhl-Bien, M. (1990). The generalizability of the hidden investment process in
leading Japanese corporations. Human Relations, 43: 1099--1116.
Wayne, S. J., & Ferris, G. R. (1990). Influence tactics, affect, and exchange quality in supervisor-subordinate
interactions: A laboratory experiment and field study. Journal of Applied Psychology, 75: 487--499.
Wayne, S. J., & Green, S. A. (1993). The effects of leader- member exchange on employee citizenship and impres-
sion management behavior. Human Relations, 46: 1431--1440.
Wayne, S. J., Shore, L. M., & Liden, R. C. (1997). Perceived organizational support and leader-member exchange:
A social exchange perspective. Academy of Management Journal, 40:82--111.
Williams, L. J., & Holahan, P.J. (1994). Parsimony based fit indices for multiple indicator regression models: Do
they work? Structural Equation Modeling, 1: 161--189.
Williams, M. L., Podsakoff, P. M., & Huber, V. (1992). Effects of group-level and individual-level variation in
leader behaviours on subordinate attitudes and performance. Journal of Occupational and Organizational
Psychology, 65:115--129.
Yukl, G. (1989). Managerial leadership: A review of theory and research. Journal of Management, 15:251--289.
Zerbe, W. J., & Paulhus, D. L. (1987). Socially desirable responding in organizational behavior: A reconception.
Academy of Management Review, 12: 250--264.