Professional Documents
Culture Documents
Chester A. Schriesheim”
University of Miami
Stephanie L. Castro
Louisiana State University
Claudia C. Cogliser
Oregon State University
Research conducted since the construct of leader-member exchange (LMX) was first investigated
in 1972 is reviewed with respect to the theoretical, measurement, and analytic adequacy of LMX
studies. It is shown that conceptual definitions of LMX and its subdimensions have evolved over
time, often with little reason or rationale given for changes. Likewise, the measures employed
to assessLMX have varied widely and have included an almost bewildering array of diverse item
content. Finally, LMX research has rarely examined the level of analysis at which its findings
hold. All of these shortcomings lead to the conclusion that we may know less than we should about
fundamental leader-member exchange processes and that future research must be conducted with
greater attention devoted to the key issues outlined in this review.
* Direct all correspondence to: Chester A. Schriesheim, Department of Management, School of Business
Administration, University of Miami, 414 Jenkins Building, Coral Gables, FL 33124-9145; e-mail: chet@
miamiedu.
two very different lines of development. The first branch of development from the
early VDL approach appears to be most commonly called the Leader-Member
Exchange (LMX) model (e.g., Graen, Novak, & Sommerkamp, 1982b), although
it has sometimes been given other labels as well (e.g., the “Leadership-Making”
model: cf., Graen & Uhl-Bien, 1991). The second branch of VDL development
has been the recent “Individualized Leadership” (IL) model of Dansereau and
colleagues (1995b), which is quite different from the LMX approach and is briefly
discussed later in this review.
A recent meta-analysis (Gerstner & Day, 1997) has shown that interest in the
first branch of the VDL approach, the LMX model, has been increasing substantially
over the years, as evidenced by a dramatic increase in the number of scholarly
papers recently produced in this domain (see the first line in Table 3 for a summary
of LMX studies over time). The review by Gerstner and Day (1997) also shows
that LMX research has been quite fruitful, as LMX has been a significant correlate
of such variables as increased subordinate satisfaction (e.g., Graen et al., 1982b),
increased subordinate performance (e.g., Dansereau, Alutto, Markham, & Dumas,
1982) enhanced subordinate career outcomes (e.g., Wakabayashi & Graen, 1984).
and decreased propensity to quit (e.g., Vecchio, 1982).
However, as we demonstrate in this review, there are some fundamental problems
related to the validity of the LMX construct, its measurement, and the data analytic
procedures which have been used in the majority of LMX investigations to date.
In particular, the construct has some basic definitional problems, the measures used
to assess LMX have varied substantially without explanation, and the analytic
procedures utilized have generally not been aligned with the theory being proposed
and tested. These issues draw into question the usefulness of any substantively
oriented synthesis of extant LMX literature. However, our intent is not to criticize
early or more recent reviews, but rather to highlight the fundamental issues which
are raised regarding the LMX construct and its associated research.
BACKGROUND
Summarizing the evolution of LMX theory over nearly 30 years, Graen and Uhl-
Bien (1995) recently suggested that LMX theory has passed through four stages, with
each stage building on the stages preceding it (in terms of theoretical clarification of
the LMX process). Stage One research found that leaders developed differentiated
relationships with their subordinates, a departure from the prevailing approach to
leadership which assumed that leaders displayed consistent behavior toward all
subordinates in their work units (the so-called “Average Leadership Style” or ALS
model). The second stage focused on these different relationships the leader had
within the work unit and began the explication of the nomological network sur-
rounding the LMX construct; the majority of LMX research has been conducted
with a Stage Two focus (Graen & Uhl-Bien, 1995).
The Leadership Making model (Graen & Uhl-Bien, 1991) shifted LMX research
into Stage Three, and moved the emphasis from the leader’s differentiation of
subordinates to “how they may work with each person on a one-on-one basis to
develop a partnership with each of them” (Graen & Uhl-Bien, 1995, p. 229). The
LMX Theory, Measurement, and Analysis 6.5
final stage broadens the scope from the dyad to larger collectives, exploring how
dyadic relationships are organized within and beyond the organizational system.
Despite the apparent high level of scholarly interest in LMX theory and the
description of the theory’s evolution offered by Graen and Uhl-Bien (1995), as
mentioned earlier, serious concerns about this approach still remain. In particular,
a number of scholars have expressed reservations regarding the adequacy of LMX
theory (e.g., Dansereau, Yammarino, & Markham, 1995a; Dienesch & Liden, 1986).
the adequacy of LMX measures which have been employed in LMX studies (e.g.,
Barge & Schlueter, 1991, Yukl, 1994) and the appropriateness of the methodologies
which have been used for data analysis in LMX research (e.g., Keller & Dansereau,
1995; Schriesheim, Cogliser, & Neider, 1995).
Much of this concern has arisen from the fact that the theoretical conceptualiza-
tion and operational measurement of the LMX construct have evolved since its
inception (Yukl, 1994). Additionally, we believe that it is not unreasonable to take
issue with the assertion that development of the LMX approach has followed as
orderly and chronological a progression as Graen and Uhl-Bien (1995) portray.
The four-stage model presents an overview of the progression of LMX theory
through various stages of development; we maintain that the development of the
theory within each of its stages of development is equally if not more important.
Additionally, while Graen and Uhl-Bien have categorized various LMX studies, it
is difficult to follow the development of the construct (i.e., the ideas) from their
presentation. For example, Graen and Uhl-Bien’s discussion of Stages One, Two,
and Three all contain references to papers within the same time frame (from 1984
to 1987). This chronological “mixing” of studies thus appears to contradict the
assertion that the theoretical development of the LMX construct has been chrono-
logically progressive and based on previous LMX theory: there does not appear to
have been a clear development and refinement of ideas over time.
The above concerns notwithstanding, however, the evolution of LMX theory,
measurement, and analytic methodology should be seen positively in general, partic-
ularly as contrasted with “frozen” or “static” theory (cf., Schriesheim & Kerr, 1977a,
1977b), and we trace this evolution in some detail below. While the theory has
been the subject of previous reviews and critiques (e.g., Dienesch & Liden, 1986;
Vecchio & Gobdel, 1984) we extend and update these prior reviews by systemati-
cally evaluating the LMX approach from its birth to the present in a detailed,
comprehensive manner. Before beginning our review, we should note that our
purpose is not to be critical but to take stock of the past.
We thus review the state of the LMX approach with an eye toward advancing
specific recommendations for future research. These are offered with the hope that
they will serve to further stimulate increased interest in the study of leader-member
exchange and, at the same time, help improve the quality of future work in this
domain. (Parenthetically, we should also note that we rely heavily on tabular presen-
tations to efficiently and concisely summarize a large amount of information in the
review which follows.)
In reviewing LMX theory and research, we hope to focus on the development
of concepts related to LMX. Additionally, the various perspectives in the LMX
literature on the issue of levels-of-analysis will be reviewed. The scales used to
66 LEADERSHIP QUARTERLY Vol. t 0 No. 1 1999
measure LMX will also be examined, The multiple changes which have been made
in LMX scales will be addressed, as will problems associated with these changes.
The revisions made in the measure of LMX which Graen and Uhl-Bien (1995)
recommend indicate that the measurement problems which have existed since LMX
(or VDL) was first introduced into the literature have not been corrected or ad-
dressed. Finally, the extent to which studies have examined levels-of-analysis issues
will be reviewed.
An offer will not be made and accepted without (1) mutual respect for the
capabilities of the other, (2) the anticipation of deepening reciprocal trust
78 LEADERSHIP QUARTERLY Vol. 10 No. 1 19W
with the other, and (3) the expectation that interacting obligation will grow
over time as career-oriented social exchanges blossom into a partnership
(Graen & Uhl-Bien. 1995. p. 237).
needed with respect to the appropriate level(s) of analysis for conducting future
LMX research.
Dansereau and colleagues have also distanced themselves from the LMX ap-
proach, as they have recently proposed an alternative conceptualization of leader-
ship relationships (Dansereau, 1995; Dansereau et al., 1995b). Dansereau’s ap-
proach, called “Individualized Leadership” (IL), focuses on dyadic relationships
outside of any collective context, and is different from both the ALS and VDLi
LMX perspectives on leadership. Dansereau (1995) proffers that the VDL model
differs substantially from the LMX model (as conceptualized by Graen and col-
leagues) in that the VDL model asserts that leaders form in-groups and out-groups
(within their work units), and that the VDL can co-exist with the ALS approach
(i.e., both ALS and VDL effects may operate simultaneously). He argues that the
LMX model assumes that “different relationships must occur within supervisory
work groups,” thus replacing the ALS approach as opposed to complementing it
(Dansereau, 1995, p. 482). Dansereau’s IL model proposes that both individuals
involved in the dyad are considered distinct from their respective others (i.e., each
follower is independent of other followers, and each leader is viewed as unique).
However, a linkage still exists between each leader and follower (while the dyad
remains independent of other dyads).
Summary
In summary, and as documented in Table 1, LMX was initially conceptualized
quite broadly, as being comprised of the amount of interpersonal attraction between
a leader and a member and the degree of loyalty that existed between a leader and
a member (Dansereau et al., 1975; Graen & Cashman, 1975) as well as attention,
support, and sensitivity (Dansereau et al., 1975). However, Cummings (1975), in
commenting on Graen and Cashman’s (1975) explication of the approach, con-
tended that the theory’s constructs were vague and ambiguous. Perhaps in response
to this criticism, multiple iterations of LMX definitions have since followed. These
have been confusing and sometimes appear to be contradictory but, more impor-
tantly, they have appeared with little or no discussion provided as to why the theory
has evolved and why particular changes in the LMX construct have been offered.
While early VDL theory can be said to have evolved along two separate branches
of thinking (LMX and IL), differences among the three approaches are often not
clearly represented by persons doing research in this domain. However, we believe
that these three approaches may be differentiated largely as follows. The VDL
approach has employed negotiating latitude as its key variable and has focused on
differentiated dyads in groups as its level of analysis. The LMX approach has used
measures of leader-member exchange as its central variable and left the level of
analysis open or unspecified (despite using the terms “dyad” and “dyadic,” LMX
theory and research has typically been unclear as to whether dyads in groups, dyads
independent of groups, or some other level of analysis is involved). Finally, the 1L
model has employed self-worth, satisfaction, and performance as its main focus and
has used whole dyads (independent of groups) as its level of analysis (Dansereau,
1995; Dansereau, Alutto, & Yammarino, 1984; Dansereau et al., 1995a, 1995b).
Although some LMX researchers have attempted to respond to criticisms raised
80 LEADERSHIP QUARTERLY Vol. IO No. 1 1999
about the theory, future theoretical treatments need to include better explanations
and justifications for changes which are proposed. Additionally, it would be quite
useful if future work linked current conceptualizations to the previous ones from
which they derived. [As an illustration, Dansereau’s (1995) treatment of the evolu-
tion of the IL approach did an exemplary job of this.] Finally, levels-of-analysis
issues must be specifically addressed in future theoretical treatments. The level(s)-
of-analysis at which LMX phenomena are expected to hold must be clearly specified
and theoretical justification for the proposed level(s) must be offered. Otherwise,
criticisms that the theory is vague (Cummings, 1975) and that it suffers from “. . . a
lack of specificity” and “. . . needs substantial clarification” (Dansereau et al., 1995a,
p. 108) will continue to detract from the value and contribution of this approach
to the study of leadership phenomena.
11. Graen & Ginsburgh (1977) Leader attention to needs, Leader’s divulging job information, Lead- Raw Scores (dichotomized): MANOVA.
er’s support (content of these 3 is similar to Supervisory Treatment
scale [Haga, Graen, & Dansereau, 19741) Leader’s personal sensi-
tivity, Leader’s allowance for self determination (number of items
and source of scales not reported); Leader acceptance (2 items)
(new scale; similar to Negotiating Latitude scale of Dansereau,
Graen, & Haga [1975])
12. Graen, Cashman, Ginsburgh, Leader-boss Linking Pin Quality (4 items) (source of scale not rc- Raw Scores (dichotomized); ANOVA
& Schiemann (1977) ported); Leader- and member-reported Latitude and Support (num-
ber of items and source of scales not reported)
13. Schiemann (1977) Leader Behavior Index (Dansereau, Graen, & Haga, 1975); Dyadic Raw Scores: Correlations: Mult. Regression:
Exchange scale (Graen & Cashman, 1975) MANOVAs (dyadic exchange scores tri-
chotomized).
14. Graen & Schiemann (1978) LMX (4 items) (new scale: referenced Dansereau. Graen. & Haga Raw Scores; Profile Similarity: Pattern
119751 and Graen & Cashman [1975]) Agreement.
15. James, Gent, Hater, & Coray 3 items based on discussion in House & Mitchell (1974) and Dansereau, Raw Scores: Standard Dev.; Correlations:
(1979) Graen, & Haga (1975) Subgroup Moderator Analysis.
16. Liden & Graen (1980) Vertical Exchange~egotiating Latitude (4 items) (same items as in Raw Scores (trichotomized); MANOVA.
Graen & Schiemann 119781)
17. Schriesheim (1980) Initiating Structure and Consideration (10 items each) modified from Raw Scores: Subgroup Moderator
the LBDQ-XII (Stogdill, 1963) Analysis: Mult. Regression.
18. James. Hater, & Jones (1981) Influence Opportunity (3 items) and Control (5 items) based on Raw Scores: Parallelism of Regression:
Dansereau, Graen, & Haga (1975) Planned Comparisons.
19. Katerberg & Horn (1981) LBDQ (Consideration and Initiating Structure subscales) (Stogdill, LMX scores partitioned into within and
1963) between variance using regression: raw
scores used for other variables.
20. Wakabayashi. Minami, Vertical Exchange (12 items) (new scale; referenced Graen Raw Scores (trichotomized); ANOVA.
Hashimoto, Sane, Graen. & Cashman [1975])
& Novak (1981)
Table 2. (Continued)
stwiy Reported Measrrrr Use& Amlviis Method
-
21. Dansereau. Alutto. Markham, Leadership Attention (11 items) (Dansereau, Graen. & Haga, 1975) Within and Between Analysis (WABA).
& Dumas (1982)
22. Graen. Liden, & Hoel (lYX2a) LMX (5 items) (4 items similar to Graen Ot Schiemann [lY78]; Deviation Scores correlated with turnover.
1 new item)
23. Graen, Novak, & Sommer- LMX-7 (new scale; referenced Graen & Cashman (19751 and Liden Raw Scores; ANOVA.
kamp (1982b) & Graen [1980]; items not provided)
24. Kim & Organ (1982) Non-Contractual Social Exchange (15 items) (new scale) Raw Scores; ANOVA.
25. Green, Blank, & Liden (1983) LMX (10 items), some of which were adapted from Dansereau, Raw Scores: Correlations; Canonical
Graen. & Haga (1975); Dyadic Contribution (4 items) (new scale) Correlations; Part-canonical analysis.
26. Nachman, Dansereau, Negotiating Latitude (2 items) (Dansereau, Graen, & Haga. 1975) Raw Scores and between and within
& Naughton (1983) scores for Negotiating Latitude: WABA.
27. Rosse & Kraut (1983) Negotiating Latitude (4 items) (new scale) Raw Scores: Correlations.
28. Chassie (1984) LMX (4 items) (Graen & Schiemann. 1978) Raw Scores (trichotomized); Correlations:
Logistic Regression; t-tests; Path Analysis.
29. Fukami & Larson (1984) LMX (3 items) adapted from Dansereau. Graen. & Haga (1975) Raw Scores: Correlations; Mult.
Regression.
30. Novak (1984) LMX-7 (Scandura & Graen, 1984) Raw Scores; Mult. Regression.
31. Scandura 81 Graen (1984) LMX-7 (new scale, 7-item scale also reported to be used in Graen. Raw Scores (trichotomized): MANOVA.
Novak. & Sommerkamp [lY%b])
32. Seers & Graen (19X4) I,eadership Exchange measures reportedly from Dansereau. Graen. Raw Scores: Mutt. Regression.
& Haga (1975). Gracn & Cashman (1975). and Graen (1976) (num-
ber of items not reported)
33. Snyder, Williams, & Cashman VDL (4 items) (Cashman. 1976) Raw Scores: Correlations: Muh.
(19X4) Regression.
34. Vecchio & Gobdel (1984) Negotiating Latitude (4 items) (source of scale not reportcdk In/out Raw Scores (trichotomized); Also variance
status (1 item) (new scale) partitioned as in Katerberg & Horn
(1981).
35. Wakahayashi & Graen (19X4) Vertical Exchange (12 items) (new scale based on Graen CuC’ashman Raw Scores: Mult. Regression.
jiY75] and Cashman. Dansereau. Graen & Haga j197hJ)
Table 2. (C~~~~~~e~)
stunv Reported Measure Use& Analytic Method
36. Ferris (1985) LMX (5 items) (Graen, Liden, & Noel. 1982~1) Deviation Scores for LMX: Raw Scores for
other variables: WABA.
37. Liden (1985) LMX and Leadership Interpersonal Sensitivity (7 items) adapted from Raw Scores; t-tests.
Graen & Cashman (1975), and Liden & Graen (1980)
38. Snyder & Bruning (1985) VDL (4 items) (Cashman, 1976) Raw Scores: Mutt. Regression.
39. Vecchio (1985) Negotiating Latitude (4 items) (source of scale not reported) Average and Deviation Scores used for
LMX; Raw Scores for other variables:
Correlations: Hit Rate Analysis.
41. Duchon, Green. & Taber 5-item scale adapted from 4-item Negotiating Latitude scale (used in Raw Scores; Correlations.
(19886) Graen & Cashman [1975]): also used nominations of best and
worst working relationships
42. Scandura, Graen, & Novak LMX-7 (Graen, Novak, & Sommerkamp, 1982b; Scandura Raw Scores; Mult. Regression.
(1986) 62 Graen, 1984)
43. Vecchio, Griffeth, & Nom LMX (5 items) (Graen. Liden, &, Hoe], 1982a) Raw Scores (trichotomized): Partial Corrr-
(1986) lations.
44. Dienesch (1987) 4-item LMX scale (described as typical of previous LMX measures); Raw Scores: Mult. Regression; Hierarchical
Attribution/Expectation (15 items)(Dienesch, 1986); Behavioral Regression (aggregated values for
Incident scale (9 items)(Dienesch, 1986); a single item to assess the step 1, individual values for step 2).
relationship between self and other dyad member; 3 behaviorai indi-
cators of LMX
45. Fairhurst, Rogers, & Saar 7 items adapted from Negotiating Latitude scale (Graen & Cashman. Raw Scores; Mult. Regression.
( 1987) 1975 and Liden & Graen, 1980); Relational coding scheme (Rogers
& Farace, 1975)
46. Gast (1987) LMX assessed with 5 scales from Novak & Graen’s (1982) Manager- Raw Scores; Correlations; Structural Equa-
Employee Questionnaire: 8-item LMX (Novak. 1985); Trust (3 tions Modeling (SEM).
items; Roberts & Q’Reilly [3974] and Novak [1982]); Availability
(Novak, 1982); Assistance with uncertainty (Novak, 1982); Parlicipa-
lion (Novak, 1982)
47. Lagace (1987) 7 items developed by Graen, Novak, and Sommerkamp (1982b) Mult. Regression.
Raw Scores: Correlations: ---~.
(L~ilrrLMreil,
Table 2. (Continued)
Study Reported Measure Used” Amafytic Mehod
49. Novak & Graen (1987) LMX-7 (Graen, Novak, 6;r Sommerkamp, 1982b: Scandura Raw Scores; Mult. Regression.
& Graen, 1984)
50. Vecchio (1987) LMX (4 items) (Liden & Graen, 1980) Raw Scores; Correlations; Mult. Regres-
sion; ANOVA.
51. Blau (1988) Quality of Relationship (3 items) (new scale; based on Dansereau, Raw Scores; Correlations; Mult.
Graen, & Haga [1975]) Regression.
52. K’Obonyo (1988) 15 items from the 17-item LMX scale of Graen (1985) Raw Scores; ANOVA; Hierarchical
Regression.
53. Leana (1988) LMX (9 items} based on discussion in Liden & Graen (1980) Raw Scores; Correlations; Mult.
Regression.
54. Peck (1988) 7-item LMX scale (Graen cited as author on questionnaire in appendix; LMX scores trichotomized for visual
items are identical to those reported in Scandura & Graen [1984], inspection; Raw Scores used for Correla-
with minor word changes) tions and ANOVAs.
55. Scandura (1988) LMX-7 (Graen, Novak, & Sommerkamp, 1982b) Raw scores; Simple and Hierarchical
Regression.
56. Sidhu (1988) f7-item scale (referenced Graen and associates) Raw scores; Hierarchical Regression;
Correlations; Canonical Correlation:
SEM.
57. Steiner (1988) LMX-7 (Scandura & Graen, 1984) Raw Scores; ANOVA.
58. Wakabayashi, Graen, Vertical Exchange (12 items) (new scale based on Graen & Cashman Raw Scores: Correlations; Mult.
Graen, & Graen (1988) 119751 and Cashman, Dansereau, Graen, & Haga 119767; item con- Regression.
tent appears similar to Wakabayashi & Graen 119841)
59. Fairhurst & Chandler (1989) 7 items adapted from Negotiating Latitude scale (Graen & Cashman, Raw Scores (trichotomized); Conversa-
1975 and Liden & Graen, 1980); Conversational analysis of conflict tion analysis.
situation
60. Heneman, Greenberger, & Nominations of best and worst relationships Raw Scores; ANOVA
Anonyuo (1989)
Table 2. (Continued)
Study Reported Measure Used” Analytic Method
61. Kozlowski & Doherty (1989) 2 measures of negotiating latitude: 7-item LMX scale (Scandura Correlations; Mult. Regression; Dichoto-
& Graen, 1984) and Measure of Information Exchange (13 items mized LMX and Box M test for homoge-
developed for study) neity of variance: MANQVA and
ANOVA.
62. Seers (1989) Team Member Exchange (10 items) (new scale); some items adapted Raw Scores; Correlations: Mu&.
from Seers & Graen (1984); LMX (7 items) (Graen, Novak, Regression; ANCGVA.
& So~erkamp, 1982b)
63. Wakabayashi & Graen (1989) Vertical Exchange (number of items not reported in study 1; 14 items Raw Scores: Correlations; Mult.
used in study 2) (new scale; based on Dansereau, Graen, & Haga Regression.
[1976]; content areas similar to Wakabayashi & Graen [1984])
64. Weitzel & Graen (1989) Quality of the Working Relationship (25 items from Graen & Scan- Raw Scores; Mutt. Regression.
dura [1985], including LMX-7 items [Graen & Scandura, 19841)
65. Zalesny & Kirsch (1989) LMX (6 items) developed by Dansereau, Graen, & Haga (1975) and Raw Scores; Correlations; Similarity
Graen & Schiemann (1978) Indices.
66. Dobbins, Cardy, & Negotiating Latitude (2 items) (Dansereau. Graen, & Haga, 1975) Raw Scores; Correlations; Mult.
Platz-Vieno (1990) Regression.
67. Dockery & Steiner (1990) LMX (16 items) adapted from 17-item LMX scale (Graen & Scandura, Deviation scores for LMX; Raw scores for
1985) for supervisors; LMX (12 items) adapted from 17-item other variables; Correlations; Mult.
LMX scale (Graen & Scandura. 1985) for subordinates Regression.
68. Graen, Wakabayashi, LMX (12 items) based on Graen & Cashman (1975) and Cashman. Raw Scores; Muft. Regression.
Graen, & Graen (1990) Dansereau, Graen, & Haga (1976)
69. Lagace (1990) LMX-7 (Scandura 6%Graen, 1984) Raw Scores (dichotomized); MANGVA
70. Nystrom (19~) Vertical Exchange Quality (5 items) (Graen, Liden, & Hoe], 1982a) Raw Scores; Mult. Regression.
71. Tanner & Castleberry (1990) LMX (17 items) (Graen, 1985): LMX (4 items) (Graen & Schiemann. Raw Scores (dichotomized in study 1 and
1978); LMX (4 items) (Rosse & Kraut, 1983); LMX (1 item) trichotom~ed in study 2); Correlations:
(Vecchio & Gobdel, 1984) ANOVA.
72. Turban, Jones. & Rozelle LMX (4 items) adapted from LMX-7 (Scandura & Graen. 1984) Raw Scores: MANOVA.
(1990)
73. Whi-Bien, Tierney, Graen. LMX (14 items) (Graen & Scandura 119871 referenced as source) Raw Scores; Correlations; Mult.
& Wakabayashi (1990) Regression.
Table 2. (Continued)
Study Reported Measure Used“ Analytic Method iT
.f
74. Wakabayashi, Graen, & Vertical Exchange (14 items) (Wakabayashi, Graen, Graen, Raw Scores; Correlations; Mult. Regres- ;‘;
Uhl-Bien (1990) & Graen, 1988) sion; Path Analysis. 3
75. Wayne & Ferris (1990) LMX-7 (Scandura & Graen, 1984) Raw Scores: SEM. 3
76. Yammarino & Dubiisky Supervisor Attention (9 dim.); Latitude (4 items) (Dansereau, WABA. E
5
ww Graen, & Haga, 1975) 3
77. Basu (1991) LMX (5 items) modified from Graen, Novak, & Sommerkamp (1982b) Raw Scores; Mult. Regression; Logistic 3
.-
and Seers & Graen (1984) Regression.
78. Deluga & Perry (1991) 6 items of 17-item LMX scale (Graen & Scandura, 1985) Raw Scores; Mult. Regression. sa
80. Krone (1991) LMX (5 items) (Graen, Liden, & Hoel, 1982a) Raw Scores (dichotomized); ANOVA. >
81. McClane (1991a) Negotiating Latitude (4 items) adapted fron Liden & Graen (1980) Raw Scores”; Correlations; Mult. 2
Regression. -5-.
82. McClane (1991b) Negotiating Latitude (4 items) adapted from Liden & Graen (1980) Raw Score@; Mult. Regression. lm
83. Salzmann & Grasha (1991) LMX scale (5 items) (Graen, Liden, & Hoel, 1982a) Raw Scores; Mult. Regression.
84. Stepina, Pemewe, Hassell, LMX (4 items) (Graen & Cashman, 1975) Raw Scores; Mult. Regression.
Harris, & Mayfield (1991)
85. Uhl-Bien (1991) LMX (14 items) (Graen, Novak, & Sommerkamp, 1982b and Raw Scores; Correlations; MANOVA.
Wakabayashi, Graen, Graen, & Graen, 1988)
86. Waldron (1991) LMX (5 items) (Graen, Liden, & Hoel, 1982a) Raw Scores (trichotomized): Correlations;
MANOVA.
87. Baugh (1992) LMX (14 items) (Dansereau, Graen, & Haga, 1975; Graen & Cashman, Raw Scores; Correlations; Mult. Regres-
1975; Graen, Novak, & Sommerkamp, 1982b) sion; WABA.
88. Carnevale & Wechsler (1992) LMX (5 items) adapted from Dansereau, Cashman, & Graen (1973) Raw Scores; Factor Analysis; Correlations:
Mult. Regression.
89. Day & Cram (1992) LMX-7 for subordinates (Scandura & Graen, 1984); LMX (3 items} Raw Scores; Mult. Regression.
for supervisors (new scale)
90. Deluga (1992) LMX (17 items) (Graen & Scandura, 1985) Raw Scores; Mult. Regression.
91. Dunegan, Duchon, & S-item scale adapted from 4-item Negotiating Latitude (used in Raw Scores; Mult. Regression.
Uhl-Bien (1992) Graen & Cashman [1975J): also used nominations of best and worst
working relationships
Table 2. (Continued)
Study Reported Measure Used” Analytic Method
92. Dunegan, Tierney, & Duchon h-item scale KIansereau. Gracn. & Haaa. 197.5: Duchon, Green, & Raw Scores: Correlations; MANOVA:
(1992) Taber, 1986; Graen & Cashman, 19%; &andura & Graen, 19&1) Hierarchical Regression.
93. Gerras (1992) 24 items developed, designed to measure Dienesch & Liden’s Raw Scores; Mult. Regression; Hierarch-
(1986) dimensions ical Regression.
94. Gessner (1992) 7-item LMX (Graen, Novak, & Sommerkamp, 1982b; Scandura Raw Scores; Correlations; Regression;
& Graen, 1984) Dichotomized LMX scores, r-tests.
96. Markham, P/furry, & Scott 4-item scale adapted from D-item Leadership Attention scale WABA; Raw score ANOVA.
ww (Dansereau, Alutto, & Yammarino, 1984)
97. Schriesheim, Neider, Scan- LMX (6 items) (new scale) Raw Scores; SEM: Correlations.
dura, & Tepper (1992a)
98 Schriesheim, Scandura, LMX (6 items) (S~h~esheim, Neider, Scandura. & Tepper, 1992af Raw Scores; SEM; Correlations.
Eisenbach, & Neider (1992b)
99. Tierney (1992) LMX (14 items) (Graen, Novak, & Sommerkamp, 1982b) Deviation Scores for LMX; Raw Scores for
other variables; Correlations: Muit.
Regression.
101. Yammarino & Dubinsky Leadership Attention, Job Latitude (Dansereau, Graen, & Haga, 1975) WABA.
WQ) (number of items not reported}; Satisfaction with Performance
(1 item) (Dansereau, Alutto, Markham, & Dumas, 1982); Job Con-
gruence (1 item) (Dansereau, Alutto, Markham, & Dumas, 1982)
102. Baugh, Graen, & Page (1993) Centroid item from 14-item LMX scale (discussed in Graen Raw Scores: MANCOVA; ANCOVA.
& Cashman, 1975)
103. Duarte, Goodson, & Klich LMX-7 (Scandura & Graen. 1984) Raw Scores; Mult. Regression.
(1993)
104. Fairhurst (1993) 7 items adapted from Negotiating Latitude scale (Graen Bc Cashman, Raw LMX scores used to classify subjects
1975 and Liden & Graen, 1980) as in-, middle-, or out-group; fonversa-
tional Analysis.
105. Jones, Glaman, & Johnson LMX (8 items) adapted from Scandura & Graen (1984) Raw Scores; Correlations.
(1993)
106. Judge & Ferris (1993) LMX (5 items) adapted from Dansereau. Graen, & Haga (1975) and Raw Scores; SEM.
Graen & Schiemann (1978)
Table 2. (Co~~i~~ed)
Study Reported Measure Used” Analytic Method
107. Liden, Wayne, & Stilwell LMX-7 (Scandura & Graen, 1984) Raw Scores: Mult. Regression.
(1993)
108. Murry (1993) LMX (5 items) adapted from Scandura & Graen (1984) and Graen, WABA.
Novak, & Sommerkamp (1982b) 2
109. Phillips, Duran, & Howe11 18-item Attribution/Expectation scale (Dienesch, 1985) Exploratory and Confirmatory Factor 3
(1993) Analysis (CFA). 2
110. Scott (1993) LMX (14 items) (reported as developed by Graen, Novak, & Sommer- Raw Scores; Correlations: Mult. Regres- %
kamp [ 1982b]) sion; SEM. 2
111. Tanner, Dunn, & Chonko Exchange Relationship Quality (6 items) (new scale) Raw Scores (trichotomized): MANOVA. s
(1993) 2
112. Tansky (1993) LMX (7 items) (Graen & Casbman, 1975; this is a 4-item scale) Raw Scores; t-tests: Correlations; Mult. -5-.
Regression. v,
113. Vansudevan (1993) LMX (modified version of 14-item scale) (source of scale not reported) Raw Scores; Cluster Analysis; Correlations;
Mult. Regression; SEM.
114. Vecchio (1993) LMX (4 items) (Liden & Graen, 1980) Raw Scores for LMX (trichotomized);
Intraclass Correlation Coefficient for one
variable; Correlations; ANOVA; Sub-
group Analysis.
115. Wayne & Green (1993) LMX-7 (Scandura & Graen, 1984) Raw Scores; Correlations; ANOVA.
116. Wilhelm, Herd, & Steiner 7-item LMX scale adapted from Dansereau, Graen, & Haga (1975) Raw Scores; Mult. Regression.
(1993) and Graen & Cashman (1975)
117. Ashkanasy & O’Connor ( 1994) Negotiating Latitude (4 items) (Liden & Graen. 1980) Raw Scores (both continuous and categori-
cal); Mult. Regression; Multidimensional
Scaling.
118. Bauer & Green (1994) LMX-7 (Scandura & Graen, 1984) Raw Scores; Mult. Regression.
119. Borchgrevink & Boster (1994) 14-item scale (obtained from Graen, personal communication, 1990) Raw Scores; CFA.
120. Deluga & Perry (1994) LMX (17 items) (Graen & Scandura, 1985) Raw Scores; Mult. Regression.
121. Duarte, Goodson, & Khch LMX-7 (Scandura & Graen, 1984) Raw Scores; Mutt. Regression.
(1994)
Table 2. (Continued)
Study Reported Measure Used#
122. Kinicki & Vecchio (1994) LMX-7 (Scandura & Graen, 1984) Raw Scores: Mu&. Regression.
123. Phillips & Bedeian (1994) LMX-7 (Scandura & Graen. 1984) Deviation Scores for L‘MX correlated with
Raw Scores of other variables.
124. Scandura & Schriesheim LMX-7 (Scandura & Graen, 1984) Raw Scores; Mult. Regression: SEM.
(1994)
125. Scott & Bruce (1994) LMX (14 items) (reported as developed by Graen, Novak, & Sommer- Raw Scores; Correlations; Mult. Regres-
kamp [1982b]) sion; SEM.
126. Basu & Green (1995) I(-item scale adapted from Graen, Novak & Sommerkamp (1982b) Raw Scores; Mult. Regression.
and Seers & Graen (1984).
129. Keller & Dansereau (1995) LMX (5 items) (Graen, Liden, & Hoel, 1982a) for supervisors: WABA.
Modified version of same scale for subordinates.
130. Kramer (1995) 3-level nominal LMX scale (1 item) (new) Raw Scores; Correlations; ANOVA:
MANOVA.
131. Major, Kozlowski, Chao. LMX-7 (Scandura & Graen, 1984) Raw Scores; Mult. Regression.
& Gardner (1995) I;;
132. Sias & Jablin (1995) 7 items (Graen & Cashman, 1975) Raw Scores (dichotomized); t-tests: G
Chi-square. 5;:
133. Bauer & Green (1996) LMX (8 items) (Scandura & Graen, 1984) Raw Scores: Mult. Regression. ;
134. Bhal & Ansari (1996) Sample 1: 24 new items, and LMX (5 items) (Graen, Liden, & Raw Scores; Correlations: Factor Analysis.
Hoel, 1982a) x31
Sample 2: 10 items (reduced from Sample 1 results), and items of Raw Scores; Correlations; E and F ratios z
Attention and Latitude (Dansereau, Alutto, & Yammarino, 1984) of WABA I. r:
13.5. Maslyn, Farmer, & Fedor LMX-7 (Scandura & Graen, 1984) Raw Scores: Correlations: SEM. 5G
wf4 7
136. Green, Anderson, & Shivers 7 items (Graen, Novak, & Sommerkamp, 1982b) Raw Scores; SEM.
o<
0.996) .-
138. Settoon, Bennett, & Liden LMX-MDM (Liden & Maslyn, 1998) Raw Scores: SEM. ;;
(1996)
139. Thibodeaux & Lowe (1996) LMX-7 (Scandura & Graen, 1984) Raw Scores (dichotomized): Correlations: ;
Chi-square; MANOVA.
Study Reported Measure Used0 Analytic Method 2
.i
140. Williams, Gavin, & Williams LMX (8 items) (Graen, Novak, & Sommerkamp, 1982b) Raw Scores; Partial Correlations; Variance 5
(1996) Reduction Rate; SEM.
141. Basu & Green (1997) 5 items (Graen. Novak, & Sommerkamp, t982b; Seers & Graen, 1984) Raw Scores; Regression; MANOVA. ;
142. Engle & Lord (1997) LMX-7 (Scandura & Schriesheim, 1994) Raw Scores: Correlations; Hierarchical z
Regression. sVI
144. Wayne, Shore & Liden (1997) LMX-7 (Scandura & Graen, 1984) Raw Scores; CFA; SEM.
145. Klein & Kim (1998) LMX (7 items) (Graen, Novak, & Sommerkamp, 1982b) Raw Scores; ANOVA; Mult. Regression.
146. Liden & Maslyn (1998) LMX-MDM (13 items) (new scale) Raw Scores; Correlations; Mult.
Regression; SEM.
147. Schriesheim, Neider, LMX-6 (Schriesheim, Neider, Scandura. & ‘pepper, 1992a) Raw Scores; WABA; Mutt. Regression.
& Scandura (1998)
Nore. Modified versions of previously-used instruments are labeled “new” above if items were changed, added, or deleted.
” References for scale source(s) reported in this column are those made by the author(s) of the respective articles. The items used in studies in some instances did not correspond
with the citation given by the author(s) as the source of the measure.
“The LMX-7 scafe refers to the 7-item LMX scale reported in Scandura and Graen (1984) and developed by Graen, Novak, and Sommerkamp (1982b).
* One variable, Rote Differentiation. was obtained by summing the absolute value of differences between each member’s Negotiating Latitude score and the mean Negotiating
Latitude of the work unit.
d Member scores on Negotiating Latitude were adjusted to the group’s average.
92 LEADERSHIP QUARTERLY Vol. 70 No. 1 1999
Structure items from the Ohio State studies’ Leader Behavior Description Quesdon-
naire (LBDQ), with one study “augmenting” these with 20 additional items (Graen,
Dansereau, Minami, & Cashman, 1973a).’ Schriesheim (1980) and Katerberg and
Horn (1981) also used Consideration and Initiating Structure items from the revised
LBDQ (Schriesheim, 1980, used a modified version of the revised LBDQ [Schries-
heim, 19791).
Additionally, the Role Orientation Index (RQI) was used along with the LBDQ
by Graen et al. (1972b, 1973a) as a supplementary leadership measure. The Graen
et al. (1972s) study used a third measure, Leader-Member Influence, which was
reported to be similar to one employed by Tannenbaum (1968).
Leader-member exchange was later measured by the Supervisory Attention
scale, also referred to by several other labels (Supervisor Treatment, Leadership
Attention, Leadership Support, Leader Attention, and Support). However, as only
Haga et al. (1974) give a complete list of items and response categories (caned
Supervisor Treatment in that article), it is difficult to determine whether the four
studies reporting use of this measure actually used the same instrument. However,
the several that did report the content areas they measured presented somewhat
different lists (Graen, Orris, & Johnson, 1973b; Haga et al., 1974; Dansereau et al.,
1975; Craen gi Ginsburgh, 1977).
A two-item measure, Negotiating Latitude, was first used by Dansereau et al.
(1975) to measure LMX in conjunction with measures of Supervisor Attention
(Leadership Attention and Leadership Support). In the same year, Graen and
Cashman (1975) published a study that used a four-item measure of Negotiating
Latitude: information was not provided regarding this measure. However, reference
was made to Haga et al. (1974), Dansereau et al. (1975), Graen (1976), and Graen,
Dansereau, Haga, and Cashman (1975). Graen and Ginsburgh (1977) later provided
information on two of the four items used in the Graen and Cashman (1975)
study, and the two appear similar to the two used by Dansereau et al (1975). No
information was given by the authors regarding the two new items’ development
or why they were included.
The two-item measure of Negotiating Latitude was also used in two additional
studies. Cashman et al. (1976) used a two-item measure called Superior Negotiating
Latitude; the two items were similar to Dansereau et al’s (1975) with only minor
wording changes. Cashman et al. (1976) also used a three-item measure, Superior
Vertical Exchange, but no information was presented regarding the items nor were
any sources cited as references for the measure. Graen and Ginsburgh’s (1977)
measure, Leader Acceptance, was composed of two items that appear similar to
those of the Negotiating Latitude measure of Dansereau et al. (1975).
However, the authors differentiate Leader Acceptance from Negotiating Lati-
tude; Leader Acceptance is based on the leader’s response to the two items regarding
the member, whereas Negotiating Latitude involves the member’s response concern-
ing the leader. This study also looked at Leadership Treatment (as a dependent
variable), composed of five scales: Leader Attention, Info~ation, Support, Personal
Sensitivity, and Allowance for Self-determination; some content areas appear simi-
lar to those in the Supervisor Treatment scale of Haga et al, (1974).
The Leader-Boss Linking-Pin Quality scale, a four-item measure, was used to
LMX Theory, Measurement, and Analysis 93
assess LMX by Graen et al. (1977). One of the items was similar to the first item
of Negotiating Latitude (Dansereau et al., 1975), but the measure included three
new items. The authors refer to Graen and Cashman (1975) for details on data
collection, and state that validity evidence for the measure is presented in Graen
and Cashman (197.5) and Cashman and Graen (1977). However, the Graen and
Cashman (1975) study does not present the items used, and one of the items Graen
and Ginsburgh (1977) report Graen and Cashman (1975) as using was not used by
Graen et al. (1977). Graen et al. (1977) used a Leader’s Latitude and Support scale,
in addition to the Leader-Boss Linking-Pin Quality scale, as a measure of leadership.
No information was provided regarding the items or the response categories of this
measure, nor were any supportive or informative citations given.
Schiemann’s (1977) dissertation did not clearly distinguish between measures
used to capture the exchange relationship and those that were just related to it.
However, it appears that the Leader Behavior Index [for which Dansereau (1975)
is referenced on p. 45 but Dansereau et al. (1975) was apparently meant] and the
Dyadic Exchange scale [Graen & Cashman (1975) are cited] were intended as
measures of the exchange relationship. The Leader Behavior Index consisted of
three subscales: (1) Information, for which some item content appears similar to
the Supervisory Attention scale (Graen et al., 1973b) and the Supervisory Treatment
scale (Haga et al., 1974); (2) Assurance and Attention (consisting of four items, all
corresponding to the Supervisory Attention scale; Graen et al., 1973b); and (3)
Support, appearing to correspond to the Leadership Support scale as described by
Graen and Ginsburgh (1977). The second instrument Schiemann used to measure
the exchange relationship, the Dyadic Exchange scale, is composed of four items
which appear to be a combination of those used in the Negotiating Latitude scales
of Graen et al. (1977) and Graen and Schiemann (1978).
A four-item measure of LMX was again used in 1978 by Graen and Schiemann,
with two new items and two items drawn from the Negotiating Latitude scale of
Dansereau et al. (1975) ( minor wording changes were made). The authors refer-
enced Graen and Cashman (1975) as having extended and validated the Dansereau
et al. (1975) measure, so it is possible that the two “new” items were actually the
same as those used by Graen and Cashman (1975); however, it is not possible to
determine whether this is the case, since Graen and Cashman (1975) did not report
their items.
The 1980s did not bring any clearer consensus in LMX measurement, although
the development of a seven-item scale used in Graen et al. (1982b) and reported
in Scandura and Graen (1984) (LMX-7) has become the most commonly-used
measure for LMX operationalization (Gerstner & Day, 1997). The majority of
LMX research throughout the 1980s continued to use many different LMX scales,
without providing clear rationale for the use of a particular measure and without
much (if any) psychometric support for the soundness of the measures employed.
Graen and his colleagues used 5-,7-, 12-, 14-, and 17-item measures of LMX during
this period, as well as a four-item Negotiating Latitude scale (see Table 2).
The rest of the research in the 1980s employed 16 different measures, reported
to assess Negotiating Latitude (two- and four-item scales), Influence Opportunity,
Vertical Exchange, Leadership Attention, LMX (three-, four-, five-, six-, seven-,
94 LEADERSHIP QUARTERLY Vol. 10 No. 1 1499
nine-, lo-, and 25item scales), Non-Contractual Social Exchange, and VDL, as well
as nominations of best and worst working relationships. As mentioned previously.
most studies (1) did not indicate the rationale for choosing a particular measure;
(2) did not provide scale item content (even if the measure was new or modified);
or (3) did not indicate the source or origin of the measure (even if an existing scale
was used, sources would typically be listed that do not correspond to the measure
employed).
As shown in Table 2, LMX measurement in the 1990s is still characterized by
researchers using different LMX measures, some of which are developed on an ad-
hoc basis or modified from existing measures without adequate psychometric testing.
Of the 75 empirical works conducted during this decade and reported in Table 2,
numerous different LMX scales were employed. Because of a great deal of variation
in reporting practices employed in the studies reviewed, it is difficult to determine
whether the scales used were identical to those previously employed or to what
extent they were altered, For example, as can be seen in Table 2, measures consisting
of the same number of items referenced different studies as the source of the
measure, It does appear, at a minimum, that at least 12 different scales were used
to measure LMX in the 1990s (since the number of items varied); various studies
reported using one-, four-, five-, six-, seven-, eight-, 12-, 13-, 14-, 16-, 17-, and 24-
item scales, In addition, Negotiating Latitude (both two- and four-item versions~,
Supervisory/Leadership Attention (four- and nine-item scales), Vertical Exchange
Quality (five and 14 items), Exchange Relationship Quality (six items), nominations
of best and worst working relationships, an Attribution/Expectation scale (18 items),
and a three-level “nominal LMX scale” .were used as operationalizations of the
construct. The bulk of studies employed either LMX-7 or a five-item LMX measure
(it does not appear that the studies used the same five-item scale, however).
Somewhat noteworthy is the recent recommendation by Graen and Uhl-Bien
(1995) that researchers use a revised seven-item LMX measure. This new scale is
quite similar to the LMX-7, with the major difference being a change in response
categories (from four to five categories and revised scale anchors). However, it is
unclear as to why this new measure is recommended and, unfortunately, Graen
and Uhl-Bien (1995) provide no evidence of adequate psychometric testing of the
newly revised LMX-7 scale. [As noted by Schriesheim et al. (1993) and others,
making even minor modifications to a scale may change its psychometric properties,
so that testing revised measures is essential to the development of con~den~e in
their psychometric adequacy.]
Several similar but slightly different versions of LMX measures have also evolved
from the LMX research stream. Seers (1989) developed a lo-item Team-Member
Exchange (TMX) scale that was adapted from an LMX scale used in Seers and
Graen (1984). Uhl-Bien and Graen (1992) used a Project LMX (PLMX) scale and
a Team-Member Exchange (TMX) scale. The PLMX measure consisted of one
item with a 5-point response scale (the item and the end anchors were reported).
The Team-Member Exchange (TMX) scale was a seven-item scale, and both Seers
(1989) and Uhl-Bien (1991) were referenced but the items and response categories
were not given. Baugh, Graen, and Page (1993) used several scales, all reported to
measure LMX. They used two “process measures,” both single-item scales; the first
LMX Theory, Measurement, and Analysis 95
was described as the “centroid item from the fourteen-item LMX scale (Graen and
Cashman, 1975)” (Baugh et al., 1993, p. 10) and the second was said to be the
“centroid item from the 7-item team member exchange scale” (Baugh et al., 1993,
p. 10) (no reference was given for this second measure). Although only one item
of each measure was used for data analysis, both of the scales from which the items
came were included in their entirety on the research questionnaire. The two process
measure items and their response categories were listed, and the response alterna-
tives listed for the two items which made up the process measures are identical to
the response categories listed by Graen and Uhl-Bien (1995) for a similar question
they present in their recommended LMX measure.
Paralleling the development of the LMX measure as described above, another
measure of exchange quality, Vertical Exchange, was developed by Wakabayashi,
Minami, Hashimoto, Sano, Graen, and Novak (1981), and consisted of 12 items.
The items and response categories were not provided, but the authors indicated
the five content areas from which the items were drawn, and the Graen and Cashman
(1975) study (which used a four-item measure) was referenced. A 12-item Vertical
Exchange scale was also used by Wakabayashi and Graen (1984) and by Wakabay-
ashi, Graen, Graen, and Graen (1988) both listing five content areas measured by
the scale and both referencing Graen and Cashman (1975) (a four-item measure)
and Cashman et al. (1976) ( a t wo-item Superior Negotiating Latitude scale and a
three-item Superior Vertical Exchange measure). Wakabayashi and Graen’s 1989
investigation involved two studies, each employing Vertical Exchange measures:
one for which four content areas were discussed, and a second which consisted of
14 items (the items and response categories were not presented for either). Both
referenced Dansereau et al.% (1975) study, which utilized a two-item measure.
Addressing the multidimensionality of the LMX construct, several different mea-
sures have been developed to assess the mutual affect, contribution, and loyalty
subdimensions that were suggested by Dienesch and Liden (1986) as being the key
subdimensions of LMX. Dienesch (1987) used two scales developed by Dienesch
(1986) to measure the three dimensions hypothesized by Dienesch and Liden (1986):
an Attribution/Expectation scale and a Behavioral Incident scale. Gerras (1992)
developed a 24-item scale designed to measure these three subdimensions. Phillips.
Duran, and Howell (1993) used 18 of 20 items from an Attribution/Expectation scale
(developed by Dienesch, 1986) to measure the three subdimensions. Schriesheim,
Neider, Scandura, and Tepper (1992a) developed and Schriesheim, Scandura, Eisen-
bath, and Neider (1992b) further validated a six-item measure with two items for
each subdimension. Liden and Maslyn (1998) developed a 13-item scale that includes
these three dimensions plus the added dimension of professional respect. Both the
LMX-6 of Schriesheim et al. (1992a) and the LMX-MDM of Liden and Maslyn
(1998) have undergone reasonable psychometric testing and have shown promising
evidence of satisfactory reliability and validity.
Summary
Given the history of LMX research which is chronicled above, it is not surprising
that confusion exists within the literature about the nature of the phenomenon or
that equivocal relationships have sometimes been found with expected outcomes
96 LEADERSHIP QUARTERLY Vol. 10 No. 1 19%
(Gerstner & Day, 1997). Additionally, mirroring other leadership domains, without
the use of a single construct definition and operationalization, it is very difficult to
make comparisons across studies or to make incremental progress in replicating
study findings. Finally, it should be noted that even the primary developer of LMX
theory himself has at times expressed uncertainty as to what some of the LMX
research instruments actually measure.
In Graen’s (1976) discussion of two previous studies, Dansereau et al. (1975)
and Haga et al. (1974), he questions, “What was ‘negotiating latitude’ measuring?
Obviously we had a handle on an important phenomena [sic], but what was its
nature?” (Graen, 1976, p. 1240). The measure of negotiating latitude (the variable
used to indicate development of a dyadic structure in initial LMX studies) was
purported to “assess the relative openness of a leader to individualized assistance
for a member” (Graen & Cashman, 1975, p. 145). Seven years later the same items,
now referred to as LMX, were characterized as assessing individualized leadership
(Graen et al., 1982a, p, 869) and the quality of LMX (Graen et al., 1982b, p. 118).
The same inconsistency and confusion appears to exist for the LMX-7 and revised
LMX-7 scales (Graen & Uhl-Bien, 1995). Simply put, this is highly problematic,
and we consider this issue further in the discussion and conclusion sections of this
article.
Table 3 thus presents a brief summary of the analyses employed in the 137
empirical LMX studies we uncovered and examined for this review (further detail
on the analyses used in these studies is provided in the right-hand column of Table
2). Although there are several approaches which might be employed to test for
different levels of analysis, the within- and between-entities analytical technique of
Dansereau, Alutto, and Yammarino (1984) appears to be the only fully appropriate
one used in the literature to date (hierarchical linear modeling and other potenti~ly
suitable approaches have apparently not been employed thus far). We therefore
briefly use this approach to illustrate why the testing of multiple levels is critical
for research in domains such as leader-member exchange.
their between-entity scores (i.e., the appropriate [xk] or [yk] for the entity within
which each part is situated). The within-entity correlation may be computed by
correlating the within-entity deviation scores (i.e., [” - x~] and [y” - yh]) for the n
parts, while the between-entity correlation may be computed by assigning each part
its appropriate between-entity scores([xk] and [yk]) and then correlating these across
the parts (i.e., between-entity correlations are simply correlations between the entity
means. weighted by the number of parts within each entity).
As can be seen from the fundamental WABA equation (Equation l), any raw
score correlation is just the sum of two separate components-a between-entity (cell)
component (3BxqBy’Bxy) and a within-entity (deviation) component (~W~~W~rW~~);
both are the products of multiplying their appropriate etas and component correla-
tions. Thus, for example, a raw score correlation of 0.00 may be highly misleading
if it is based upon a large negative within-entity component and a large positive
between-entity component (the conclusion drawn would be one of “no relationship,”
while the data actually support a negative within- and a positive between-entities
relationship). Consequently, raw score correlations (IT,,) cannot be unambiguously
interpreted-the explicit examination of levels of analysis is absolutely critical for
the drawing of sound conclusions.
LMX scale which was taken from Graen et al. (1982a), while Schriesheim, Neider,
and Scandura (1998) used Schriesheim et al.‘s (1992a, 1992b) six-item LMX scale.
Thus, even discounting Graen and Uhl-Bien’s (1995) recent contention that the
LMX-7 is the only measure which should be used, it seems clear that at least five
of the 10 studies that used an appropriate analytic technique also used measures
which are questionable operationalizations of the LMX construct (Dansereau et
al., 1982; Markham et al., 1992; Nachman et al., 1983; Yammarino & Dubinsky,
1990, 1992).
DISCUSSION
This review has examined and summarized 147 works which were completed since
the constructs of VDL and LMX were first investigated in 1972. Several important
difficulties have been highlighted and discussed. Some may view the numerous
definitions and subdimensions of LMX, the multiple measures, and the unspecified
level(s) and analysis to be part of the “richness” of LMX. However, we feel that
there is a good possibility of further developing the potential of the LMX approach
by providing greater clarity in the theoretical conceptualization of the construct
and by enhanced measurement and attention to level(s) of analysis concerns. Thus,
we make some brief recommendations with respect to needed theory, measurement,
and analysis in future LMX research.
Theory
As we discussed during our review of LMX theory, it has evolved substantially
over the past 25 years. This is good. One of the major criticisms which has been
advanced about other leadership theories is that there has been a lack of respon-
siveness to constructive criticisms provided by the field, as well as a failure to
incorporate needed changes as evidence has accumulated (cf. Schriesheim & Kerr,
1977a, 1977b). Neither of these criticisms can be leveled at LMX theory.
On the other hand, we believe that some of the changes which have been made
in LMX theory have not be accompanied by an adequate explanation of those
changes or the theoretical rationale which underlies them. For example, if the
theoretical definitions of LMX which are provided in Table 1 are examined, along
with the subaspects which various writers have specifically mentioned as being
included within the LMX construct, one cannot help but be struck by a lot of
similarities and a lot of differences. While it would be hoped that work which
elaborates or modifies the conceptualization of LMX would include theory, litera-
ture and discussion (rationale) which integrates the new treatment with previous
work (and which develops it in reasonable detail), this simply has not been done.
As a recent case in point, we believe that Graen and Uhl-Bien’s (1995) most
recent revision of the theory does not adequately justify why the theory now only
focuses on the relationship between a leader and a follower (and no longer considers
the leaders and followers themselves). Additionally, it is now becoming increasingly
accepted that “good” theory clearly states the level(s) of analysis at which its
hypothesized relationships hold (cf., Klein et al., 1994). Unfortunately, again, the
100 LEADERSHIP QUARTERLY Vol. 10 No. 1 1994
most current version of LMX theory (Graen & Uhl-Bien, 1995) now appears to
have been modified so that it is unacceptable in this regard. We should note that
we are not the only ones believing this-Dansereau et al. (1995a, p. 100) feel that
currently the “LMX approach provides little guidance as to how to view leaders or
followers from a levels-of-analysis perspective,” and that this is a critical deficiency.
Additionally, Dansereau (1995, p. 484) now argues that the newest “LMX approach
is simply unclear as to what level [of analysis] it applies.”
In addition to needing increased attention to the provision of logic and linkages
between new and previous LMX theory, it seems clear to us that more theoretical
work is needed on LMX. The major theoretical papers number only a few, and
with few exceptions, have been written by Graen and associates. We believe that
broader-based, more frequent, and more systematic synthesis and integration of
the LMX literature is needed, along with more basic theorization and model devel-
opment. Certainly, future theoretical explications of LMX should clearly state the
level(s) of analysis at which the phenomena are expected to hold.
While some theorists may choose to go back to VDL’s propositions regarding
differentiation among subordinates (a within-unit phenomenon), or to adopt Dan-
sereau’s (1995) suggestion of independent dyads, it may also be fruitful to explore
alterative levels of analysis possibilities (e..g, Wiley, 1988). If some of this were to
be performed by a cross section of the field, perhaps an increased fertilization of
new ideas would occur. In any event, regardless of authorship, it seems critical to
us that more attention be devoted to theory development. And, at the same time,
it seems important that scholars who use the LMX approach in the future be more
careful to not modify or extrapolate extant theory in their writings without providing
a well-developed and sound theoretical basis for doing so.
Measurement
As evidenced by our review of LMX measures, and ~fortunately following a
good portion of management research practice (Podsakoff & Dalton, 1987; Schries-
heim et al., 1993), LMX scales seem to have been developed on an ad-hoc, evolution-
ary basis, without the presentation of any clear logic or theory justifying the changes
which were made. Also, items have been added to and subtracted from LMX
measures without adequately discussing or presenting evidence regarding the ef-
fect(s) of such changes on scale validity.
Perhaps even more troublesome, none of the scales used to measure LMX
appear to be “based on either systematic psychometric study or explicit construct
validation” (Dienesch & Liden, 1986, p. 623; see also Bhal & Ansari, 1996, and
Vecchio & Gobdel, 1984), althou~ validity assessments should be conducted on
all measures on an on-going basis (Guion & Schriesheim, 1997; Nunnally & Bern-
stein, 1994). Schriesheim et al. (1992b) have suggested that extant LMX measures
have not been content valid because, “‘LMX scales have typically not been developed
using an a priori theoretical definition of its content subdomains” (Schriesheim et
al., 1992b, p. 984).
This clearly violates one of the more important principles of scale development
for, as noted by Nunnally and Bernstein (1994, p. 102), “one should ensure content
LMX Theory, Measurement, and Analysis 101
validity (adequacy of sampling the material on which people are tested) in terms
of a well-formulated plan and procedure of test construction before the actual test
is developed rather than evaluate this after construction.” Additionally, it should
be emphasized that content validity is not a trivial psychometric property, since a
measure that is not content valid cannot. be construct valid (Schriesheim et al.,
1993) and because Cook, Campbell, and Perrachio (1990, p. 505) suggest that one
of the more serious threats to construct validity is “the inadequate preoperational
explication of constructs.”
Our review thus suggests that the LMX construct may not be what was actually
measured by at least some of the LMX scales which have been employed to date.
Because a systematic program of development and validation has not been con-
ducted, exactly what these scale are measuring is unknown and attempting to
substantively synthesize the existing literature may therefore not make much sense.
As such, we believe that either a reconceptualization of the LMX construct is
needed or further development and validation of a suitable LMX scale should be
undertaken. With respect to this second course of action, it is probably worthwhile
to note that current LMX theory lacks a clear description of the exchange process
between a leader and subordinate (Dansereau et al., 1995a; Dienesch & Liden,
1986) and that it “doesn’t specify what pattern of downward exchange relationships
with different subordinates is optimal for leadership effectiveness” (Yukl, 1994, p.
239). Furthermore, although LMX has been considered to involve an “exchange”
process, neither the supervisor nor the subordinate are asked in any of the LMX
scales (with the exception of the Negotiating Latitude scale used by Dansereau et
al., 1975) what must be given (exchanged) for the reward or benefit to be received
(Gerstner & Day, 1997).
However, some possible dimensions of the leader-member exchange have been
suggested (Graen & Scandura, 1987). Thus, future work may be able to build upon
this initial effort and a great opportunity appears to exist for psychometric work
to advance future research on LMX. To avoid the mistakes of the past, such
future psychometric work needs to carefully follow established practices for scale
development and validation (cf., Hinkin, 1995; Nunnally & Bernstein, 1994; Schries-
heim et al., 1993). Additionally, a challenge exists for future researchers to better
and more accurately describe their measures and how they relate to previous
measures, and why one measure is employed rather than another.
Analysis
The level of analysis at which LMX theory should generally be tested appears
to be either dyadic or within-group, and there appears to be basic agreement within
the field regarding this position (see Footnote 2, however). The determination of
whether LMX is a dyadic or within-group phenomenon and whether LMX is a
better predictor of leadership outcomes than is the ALS approach is an empirical
question which requires the use of analytic techniques which test for various levels
of analysis. A dyadic approach would involve looking at each supervisor-subordinate
dyad as a “whole,” the analytic focus being the deviation of each dyad member’s
score from this whole (or dyad average) score. Alternatively, a within-group analysis
LEADERSHIP QUARTERLY Vol. 10 No. 1 I990
would consider the entire unit or work group as the “whole.” with deviations of
individual members’ scores from this whole (or group average) as the focus.
Unfortunately, although methodologies have existed for some time to permit
such testing (e.g., Dansereau et al., 1984), apparently only 10 studies have employed
an appropriate methodology. Because it is highly inappropriate to draw conclusions
about the level of analysis at which a particular effect operates without first testing
for such an effect, it therefore appears that support for LMX theory as a dyadic
or within-group protrayal of leadership processes is substantially less than it should
be. As we noted earlier, 10 studies, even if all were supportive of LMX level-of-
analysis predictions, cannot be considered an adequate weight of scientific evidence
on as complex a phenomenon as leader-member exchange.
The recommendation which seems most reasonable to rectify this situation seems
obvious to us: (1) all future research on LMX should first specify the level(s) of
analysis at which effects are expected to occur, and (2) suitable analytic methods
should be employed to test for such levels and effects. We have briefly highlighted
within- and between-entities analysis (WABA) as one particularly suitable approach
but we should note that others exist as well. Thus, it does not matter whether one
is or is not favorably disposed toward WABA as a data-analytic system. What
does matter is that future research on leader-member relations exercise caution in
framing and then testing hypothesized relationships. Without direct and explicit
treatment of the level of analysis issue future research is not any more likely to be
pertinent to our understanding of leader-member exchange than is much of the
extant literature to date.
CONCLUSION
Although perhaps seeming obvious at first glance, this review clearly indicates the
need for improved theorization about LMX and its basic process, for improved
measurement practices, and for enhanced and more appropriate data-analytic tech-
niques. It is true that many scholars believe that the LMX approach has substantially
contributed to deepening our understanding of fundamental leadership phenomena.
However, greater advances in the future appear possible if the basic theoretical,
psychometric, and statistical recommendations presented in this review are em-
braced and enacted by LMX researchers.
NOTES
1. Parenthetically, several versions of the LBDQ exist, but none have 40 scored items. The
Halpin (1957) and the Halpin and Winer (1957) versions (which are slightly different)
present 40 items but score only 30. This, plus the fact that Graen and associatesincorrectly
cite Stogdill and Coons (1957) Fleishman (1957) and Pleishman, Harris, and Burn
(1955) as the source of the LBDQ, makes it impossible to be sure exactly what leadership
LMX Theory, Measurement, and Analysis 103
measuresthey employed in these studiesor how they were used (cf., Schriesheim&
Kerr, 1977b).
2. As mentioned earlier, Craen and colleagues(e.g., Graen & Uhl-Bien, 1995;Graen &
Wakabayashi, 1994) have recently placed theoretical emphasison the relationship or
linkage betweenthe superior and subordinateand excluded the leader and the follower
(as individuals) from the LMX model (Dansereauet al., 1995a).As Dansereauet al.
(1995a)have noted, this new emphasisresultsin ambiguitiesregarding the appropriate
level of analysisfor this phenomenon.It seemsclear to us, however, and to Dansereau
and colleagues(e.g., Dansereau,Alutto, & Yammarino, 1984;Nachman,Dansereau,&
Naughton, 1983;Yammarino & Dubinsky, 1992) that in mostinstancestestsof the LMX
model need to examinewhether resultshold at the dyadic, within-group, or other levels
of analysis.
REFERENCES
Achinstein, P. (1968). Concepts of science. Baltimore, MD: JohnsHopkins University Press.
Ashkanasy, N. M., & O’Connor, C. (1994). Value differences as a barrier in leader-member
exchange: A multidimensional scaling analysis. Paperpresentedat the annual meeting
of the Academy of Management,Dallas, TX.
Bacharach,S. B. (1989). Organizational theories: Somecriteria for evaluation. Academy of
Management Review, 14,496515.
Barge, J. K., & Schlueter,D. W. (1991). Leadershipasorganizing: A critique of leadership
instruments.Management Communication Quarterly, 4, 541-570.
Basu, R. (1991). An empirical examination of LMX and transformational leadershipas
predictors of innovative behavior. UMI Dissertation Services. Ann Arbor, MI: Univer-
sity Microfilms International, 9201299.
Basu, R., & Green, S. (1995). Subordinateperformance, leader-subordinatecompatibility,
and exchangequality in leader-memberdyads:A field study. Journal of Applied Social
Psychology, 25, 77-92.
Basu, R., & Green, S. G. (1997).Leader-memberexchangeand transformationalleadership:
An empirical examination of innovative behaviorsin leader-memberdyads.Journal
of Applied Social Psychology, 27, 477-499.
Bauer, T., & Green, S.(1994).LMX: Relationships withperformance, expectations, experience,
and leader delegation at two points in time. Paper presentedat the annualmeeting of
the Academy of Management, Dallas, TX.
Bauer, T. N., & Green, S.G. (1996).Developmentof leader-memberexchange:A longitudinal
test. Academy of Management Journal, 39, 1538-1567.
Baugh, G. (1992).Interpersonalinfluence in a project organization: A role setanalysis.UMZ
Dissertation Services. Ann Arbor, MI: University Microfilms International, 9233218.
Baugh, G., Graen, G. B., & Page,D. (1993). Effects of team gender and racial composition
on perceptions of team performance in cross-functional teams. Paper presentedat the
annual meeting of the Southern ManagementAssociation, Atlanta, GA.
Bhal, K. T., & Ansari, M. A. (1996). Measuring quality of interaction between leadersand
members.Journal of Applied Social Psychology, 26, 945-972.
Blau, G. (1988).An investigation of the apprenticeshiporganizationalsocializationstrategy.
Journal of Vocational Behavior, 32, 176-195.
Borchgrevink, C. P., & Boster, F. J: (1994).Leader-memberexchange:A test of the measure-
ment model. Hospitality Research Journal, 17, 75-100.
Carnevale,D. G., & Wechsler,B. (1992).Trust in the public sector:Individual and organiza-
tional determinants.Administration & Society, 23, 471-494.
104 LEADERSHIP QUARTERLY Vol. 10 No. 1 1999
Cashman, J. F. (1976). The nature of vertical dvad linkages within managerial units. UMJ
Dissertation Services, Ann Arbor, MI: University Microfilms International. 7616101.
Cashman.J.. Dansereau.F.. Graen, G., & Haga, W. J. (1976).Organizational understructurc
and leadership:A longitudinal investigation of the managerialrole-making process.
Organizational Behavior and Human Performance, 15, 278-296.
Cashman,J. F., & Graen, G. (1977).The nature of leadershipin the vertical dyad: The team
building process.Cited in Graen et al. (1977,p. 504) asforthcoming in Organizational
Behavior and Human Performance (but not publishedin that journal nor in any we
could find).
Chassie,M. B. (1984).Vertical dyadic linkage formation: Predictorsand processes determin-
ing quality superior-subordinaterelationships.UMI Dissertation Services. Ann Arbor,
MI: University Microfilms International, 8506411.
Cook. J. D., Hepworth, S. J., Wall, T. D., & Warr, P. B. (1981). The experience of work.
London: Academic Press.
Cook, T. D., Campbell, D. T.. & Perrachio, L. (1990). Quasi experimentation. In M. D.
Dunnette & L. M. Hough (Eds.), Handbook of industrial and organizational psychol-
ogy: Vol. 1 (2nd ed., pp. 491-576). Palo Alto, CA: ConsultingPsychologistsPress.
Copi. I. M. (1954). Symbolic logic. New York, NY: Macmillan.
Cummings,L. L. (1975). Assessingthe Graen/Cashmanmodel and comparingit with other
approaches.In J. G. Hunt & L. L. Larson (Eds.), Leadership frontiers (pp. 181-185).
Carbondale,IL: Southern Illinois University Press.
Dansereau,F. (1995).A dyadic approachto leadership:Creating andnurturing this approach
under fire. Leadership Quarterly, 6, 479-490.
Dansereau,F., Alutto, J. A., Markham, S. E., & Dumas,M. (1982).Multiplexed supervision
and leadership:An application of within and between analysis.In J. G. Hunt, U.
Sekaran, & C. A. Schriesheim(Eds.), Leadership: Beyond establishment views (pp.
81-103). Carbondale,IL: Southern Illinois University Press.
Dansereau,F., Alutto, J. A., & Yammarino, F. J. (1984). Theory testing in organizational
behavior: The variant approach. Englewood Cliffs, NJ: Prentice Hall.
Dansereau,F.. Jr., Cashman.J., & Graen, G. B. (1973). Instrumentality theory and equity
theory ascomplementaryapproachesin predicting the relationship of leadershipand
turnover among managers.Organizational Behavior and Human Performance, IO,
184-220.
Dansereau.F.. Jr., Graen. G., & Haga, W. J. (1975). A vertical dyad linkage approachto
leadershipwithin formal organizations:A longitudinal investigationof the role-making
process.Organizational Behavior and Human Performance, 13, 46-78.
Dansereau,F., Yammarino, F. J., & Markham, S. E. (1995a).Leadership:The multiple level
approaches.Leadership Quarterly, 6, 97-109.
Dansereau,F., Yammarino, F. J., Markham, S. E., Alutto, J. A., Newman, J., Dumas, M.,
Nachman, S., Naughton, T., Kim, I., Al-Kelabi, S., Lee, S., & Keller, T. (1995b).
Individualized leadership:A new multiple level approach. Leadership Quarterl.y, 6,
413-450.
Day, D. D., & Crain, E. C. (1992).The role of affect and ability in initial exchangequality
perceptions.Group and Organization Management, 17. 38&397.
Deluga, R. J. (1992).Supervisortrust building, leader-memberexchange,and organizational
citizenship behavior. Journal of Occupational & Organizational Psychology, 67, 315-
326.
Deluga, R. J., & Perry, J. T. (1991). The relationship of subordinate upward influencing
behaviour,satisfaction,andperceivedsuperioreffectiveness.Journal of Occupational &
Organizational Psychology, 64, 239-252.
LMX Theory, Measurement, and Analysis 105
Deluga, R. J., & Perry, J. T. (1994). The role of subordinateperformance and ingratiation
in leader-memberexchanges.Group and Organization Management, 19,67-86.
Dienesch,R. M. (1986).A three dimensionalmodelof leader-memberexchange:An empirical
test. Paper presentedat the Academy of Managementmeeting, Chicago,IL.
Dienesch,R. M. (1987). An empirical investigation of the relationship between quality of
leader-memberexchangeand subordinateperformanceandsatisfaction.UMI Disserta-
tion Services. Ann Arbor, MI: University Microfilms International, 8800026.
Dienesch,R. M., & Liden, R. C. (1986). Leader-memberexchangemodel of leadership:A
critique and further development.Academy of Management Review, 11. 618-634.
Dobbins, G. H., Cardy, R. L., & Platz-Vieno, S. J. (1990). A contingency approach to
appraisal satisfaction: An initial investigation of the joint effects of organizational
variables and appraisalcharacteristics.Journal of Management, 16, 619-632.
Dockery, T. M., & Steiner, D. D. (1990).The role of the initial interaction in leader-member
exchange. Group and Organization Studies, 1.5, 395413.
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.
Duarte, N. T.. Goodson,J. R., & Klich, N. R. (1994).Effects of dyadic quality and duration
on performanceappraisal.Academy of Management Journal, 37, 499-521.
Duchon. D., Green, S., & Taber, T. (1986).Vertical dyad linkage: A longitudinal assessment
of attitudes, 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 exchangeand subordinate performance:The role of task analyzability and
variety asmoderators.Journal of Management, 18, 59-76.
Dunegan, K. J., Tierney, P., & Duchon, D. (1992). Perceptionsof an innovative climate:
Examining the role of divisionalaffiliation, work group interaction, and leader/subordi-
nate exchange.IEEE Transactions on Engineering Management, 39, 227-235.
Engle, E. M., & Lord, R. G. (1997). Implicit theories, self-schemas,and leader-member
exchange.Academy of Management Journal, 40,988-1010.
Fairhurst, G. T. (1993).The leader-memberexchangepatternsof womenleadersin industry:
A discourseanalysis.Communication Monographs, 60, 321-351.
Fairhurst, G. T., & Chandler, T. A. (1989). Social structure in leader-memberinteraction.
Communication Monographs, 56. 215-239.
Fairhurst, G. T., Rogers,L. E., & Saar,R. A. (1987).Manager-subordinatecontrol patterns
and judgmentsabout the relationship. Communication Yearbook, IO, 395-415.
Ferris, G. R. (1985). Role of leadershipin the employeewithdrawal process:A constructive
replication. Journal of Applied Psychology, 70, 777-781.
Fleishman,E. A. (1957). A leader behavior description for industry. In R. M. Stogdill &
A. E. Coons(Eds.), Leader behavior: Its description and measurement (pp. 103-119).
Columbus,OH: Bureau of BusinessResearch,Ohio State University.
Fleishman,E. A., Harris, E. F., & Burtt, H. E. (1955).Leadership and supervision in industry.
Columbus,OH: Bureau of Educational Research,Ohio State University.
Fukami. C. V., & Larson,E. W. (1984).Commitmentto companyand union: Parallel models.
Journal of Applied Psychology, 69, 367-371.
Gast, I. F. (1987). Leader cognitive complexity and its effects on the quality of exchange
relationshipswith subordinates.UMI Dissertation Services. Ann Arbor, MI: University
Microfilms International, 8708350.
Gerras, S. J. (1992). The effect of cognitive busynessand nonverbal behaviors on trait
inferencesand leader-memberexchangejudgments. UMI Dissertation Services. Ann
Arbor, MI: University Microfilms International, 9236826.
LEADERSHIP QUARTERLY Vol. 10 No. ? 19Y1