Leader-member exchange, differentiation, and task ...

25
See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/211390177 Leader-member exchange, differentiation, and task interdependence: Implications for individual and group performance Article in Journal of Organizational Behavior · September 2006 Impact Factor: 3.85 · DOI: 10.1002/job.409 CITATIONS 145 READS 1,444 4 authors, including: Robert C Liden University of Illinois at Chicago 113 PUBLICATIONS 12,751 CITATIONS SEE PROFILE Berrin Erdogan Portland State University 52 PUBLICATIONS 1,982 CITATIONS SEE PROFILE All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately. Available from: Berrin Erdogan Retrieved on: 10 May 2016

Transcript of Leader-member exchange, differentiation, and task ...

Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/211390177

Leader-memberexchange,differentiation,andtaskinterdependence:Implicationsforindividualandgroupperformance

ArticleinJournalofOrganizationalBehavior·September2006

ImpactFactor:3.85·DOI:10.1002/job.409

CITATIONS

145

READS

1,444

4authors,including:

RobertCLiden

UniversityofIllinoisatChicago

113PUBLICATIONS12,751CITATIONS

SEEPROFILE

BerrinErdogan

PortlandStateUniversity

52PUBLICATIONS1,982CITATIONS

SEEPROFILE

Allin-textreferencesunderlinedinbluearelinkedtopublicationsonResearchGate,

lettingyouaccessandreadthemimmediately.

Availablefrom:BerrinErdogan

Retrievedon:10May2016

Journal of Organizational Behavior

J. Organiz. Behav. 27, 723–746 (2006)

Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/job.409

*Correspondence to:Morgan Street, Chica

Contract/grant sponsoManagement (SHRM

Copyright # 2006

Leader-member exchange, differentiation,and task interdependence: Implicationsfor individual and group performance

ROBERT C. LIDEN1*, BERRIN ERDOGAN2, SANDY J. WAYNE1

AND RAYMOND T. SPARROWE3

1University of Illinois at Chicago, Department of Managerial Studies (M/C 243), IL, Chicago, U.S.A.2School of Business, Portland State University, Oregon, U.S.A.3John M. Olin School of Business, Washington University, St. Louis, Missouri, U.S.A.

Summary We investigated the effects of leader-member exchange (LMX) differentiation on individualand group performance with a sample of 120 work groups consisting of 834 employees whorepresented six different organizations. LMX differentiation was defined as the degree ofvariability in the quality of LMX relationships formed within work groups. HierarchicalLinear Modeling (HLM) results did not indicate support for a main effect of LMX differ-entiation on individual performance. Rather, the results demonstrated that LMXmoderated therelation between LMX differentiation and individual performance, such that increases in LMXdifferentiation were accompanied by increases in individual performance for low LMXmembers, but no change in individual performance for high LMX members. At the grouplevel, there was not a main effect for LMX differentiation on group performance. However, thehierarchical regression results revealed that the relation between LMX differentiation andgroup performance was moderated by task interdependence, such that for groups high in taskinterdependence, the greater the differentiation among group members, the higher theperformance of the group. Conversely, for groups with relatively lower levels of taskinterdependence, differentiation among subordinates was not related to group performance.Finally, LMX differentiation was positively related to group performance in groups with a lowLMX median, but was not related to performance in groups with a high LMX median.Copyright # 2006 John Wiley & Sons, Ltd.

Introduction

Leader-member exchange (LMX) theory, which was developed as an extension of the vertical-dyad

linkage model, is based on the differential types of relationships that form between leaders and group

members. The quality of leader-member relationships, in turn, is assumed to reflect the extent to which

the leader and subordinate exchange resources and support beyond what is expected based on the

R. C. Liden, University of Illinois at Chicago, Department of Managerial Studies (M/C 243), 601 Southgo, IL 60607-7123, USA. E-mail: [email protected]

rs: University of Illinois, Center for Human Resource Management (CHRM); Society for Human Resource) Foundation.

John Wiley & Sons, Ltd.

Received 4 June 2004Revised 17 April 2006Accepted 7 June 2006

724 R. C. LIDEN ET AL.

formal employment contract (Dansereau, Graen, & Haga, 1975). LMX quality has been found to play

an important role for individual performance (Gerstner & Day, 1997). Despite extensive research on

LMX and individual performance, and other individual level outcomes such as employee attitudes

(Dansereau et al., 1975; Duchon, Green, & Taber, 1986), substantially less attention has been paid to

the implications of LMX differentiation on individual and group performance (Liden, Sparrowe, &

Wayne, 1997; Scandura, 1999). Gerstner and Day (1997) emphasized the need for research exploring

the relationship between patterns of LMX and group outcomes. Despite the importance of group

effectiveness, understanding the concomitant relations between LMX and individual and group

effectiveness remains a largely unexplored research avenue (cf. Dotan, Goldstein, Nishii, Mayer, &

Schneider, 2004; McClane, 1991).

An important research question confronting LMX researchers is the extent to which LMX

differentiation is related to performance. LMX theory is based on the premise that leaders differentiate

among their subordinates. The level of differentiation tends to differ across groups such that, in some

groups there is a high level of differentiation, and the leader forms high quality exchanges with some,

but not with all members. On the other hand, in other groups, although relatively less common (Liden&

Graen, 1980), leaders form similar quality exchanges with all members. It is interesting that even

though LMX theory is based on leader differentiation between group members, the implications of

differentiation for individuals and groups remain largely unknown.

In the current study, we used a multi-level approach to examine the relations between LMX

differentiation and individual performance, as well as between LMX differentiation and group

performance. Furthermore, we examined the potential complexity of the relations of LMX

differentiation on both individual and group performance. Specifically, we explored the possibilities

that the effect of LMX differentiation on individual performance is influenced by one’s own LMX, and

that the effect of differentiation on group performance is moderated by task interdependence and

median LMX within the group.

Previous research has demonstrated that LMX is positively related to support, and creates

obligations in individuals (Maslyn & Uhl-Bien, 2001; Settoon, Bennett, & Liden, 1996), who then

reciprocate through higher levels of performance (Masterson, Lewis, Goldman, & Taylor, 2000;

Wayne, Shore, & Liden, 1997). Indeed, at the individual level, a positive relation between LMX and

individual performance has been demonstrated (Gerstner & Day, 1997). But predicting the effect of

LMX differentiation on individual performance is more complex, such that there are theoretically

compelling explanations for both positive and negative associations between LMX differentiation and

individual performance. To relieve the tension created by the existence of viable competing

hypotheses, we contend that one’s LMX status filters members’ reactions to leader differentiation.

Specifically, we offer a moderating hypothesis that proposes that the relation between LMX

differentiation and individual performance varies as a function of focal members’ LMX status.

Similarly, at the group level, we offer competing explanations of the relation between LMX

differentiation and group performance. However, we attempt to settle the apparent conflict between these

explanations by offering the hypothesis that this relation is moderated by task interdependence, a critical

structural variable identified in research on groups (Goodman, Ravlin, & Schminke, 1987). We also

contend that median within group LMX moderates the relation between LMX differentiation and group

performance. In summary, the present study contributes to the literature on leadership and effectiveness by

examining the influence of LMX differentiation on both individual and group performance.

LMX Differentiation and individual performance

Considerable research has examined the relation between LMX and individual performance (Erdogan

& Liden, 2002; Liden et al., 1997), with a meta-analysis showing a positive relation between LMX and

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

LMX DIFFERENTIATION 725

performance ratings (Gerstner & Day, 1997). This effect has been explained as occurring due to a

combination of valid differences in the performance of high versus low LMX subordinates as well as by

supervisor rating bias. Valid differences are thought to occur due to the value of resources provided to

high LMX subordinates from their leaders (Feldman, 1986). For example, subordinates enjoying high

LMX relationships are more likely than those involved in lower LMX relationships to receive

challenging task assignments, training opportunities, resources, information, and support (Liden,

Wayne, & Sparrowe, 2000; Scandura, Graen, & Novak, 1986). Although receipt of greater support

appears to result in actual performance increments for high LMX subordinates, leaders may also be

guilty of leniency biases in rating the job performance of high LMX subordinates (Duarte, Goodson, &

Klich, 1994; Heneman, Greenberger, & Anonyuo, 1989).

Moving from an individual- to a multi-level examination of individual performance we ask, ‘How

does LMX differentiation relate to individual performance?’ One point of view is that individuals could

respond negatively to LMX differentiation due to perceptions of unfairness that result when leaders

provide different levels of resources and support across members of their work groups (Uhl-Bien,

Graen, & Scandura, 2000). Such differential treatment may challenge expectations of future fairness,

undermining overall levels of trust in the leader’s intentions. Characteristic of the main approaches to

justice, including relative deprivation theory (Crosby, 1976), equity theory (Adams, 1965), and

procedural justice theory (Thibaut &Walker, 1975), is a theme that individuals are not only influenced

by their perceptions of the fairness with which they are treated, but also by their perceptions of the way

others are treated when evaluating their own standing in a group. If employees perceive that wide

differences exist in the quality of relationships between the leader and members that are not justified,

they may respond by withholding effort and performing at lower levels.

Hypothesis 1a. LMX differentiation and individual performance are negatively related.

An opposing view of the association between LMX differentiation and individual performance is

that especially in individualistic societies, such as the settings for the current research, where a

premium is placed on equity across individuals, group members desire leaders who differentiate

between subordinates. Because followers of a leader inevitably differ in the contributions that they are

able to make to the workgroup, group members may be disappointed in leaders who developed the

same type of exchange relationship with each subordinate (Sias & Jablin, 1995). In addition, leaders

who differentiate may be in a better position to cultivate high individual performance. For example,

leaders who assign more routine tasks to subordinates who are less capable of handling more

challenging tasks, and assign relatively complex tasks to more capable subordinates would be more

likely to cultivate high individual performance than leaders who do not differentiate between

subordinates.

Hypothesis 1b. LMX differentiation and individual performance are positively related.

Although we expect a main effect for the association between LMX differentiation and individual

performance, we argue that one’s own LMX serves as a filter through which perceived LMX

differentiation is interpreted. That is, we propose that LMX moderates the relation between LMX

differentiation and individual performance. We contend that for several reasons, high LMX members

are motivated to perform at high levels regardless of the degree of LMX differentiation within their

workgroups. First, it has been empirically demonstrated that LMX is positively related to interactional

justice with the leader (Erdogan & Liden, 2006; Masterson et al., 2000; Murphy, Wayne, Liden, &

Erdogan, 2003; Tekleab, Takeuchi, & Taylor, 2005), implying that high LMX members view their

quality LMX relationship as being deserved (cf. Saavedra & Kwun, 1993). Given that high LMX

members feel that the supervisor is fair, and that they deserve their high LMX status, they should be

motivated to continue the performance level that they feel qualified them for high LMX status. Second,

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

726 R. C. LIDEN ET AL.

high LMXmembers have a desire to reciprocate for the support provided by the leader in a high quality

exchange relationship. One key way to reciprocate is to put forth high levels of individual performance

(Graen & Scandura, 1987). Because high LMXmembers feel that their quality relationships, complete

with high levels of support, are justified, they should be relatively unaffected by the degree to which the

leader differentiates between members.

On the one hand, low LMX members may respond differently to low versus high levels of LMX

differentiation within their workgroups. We argue that low LMX members under conditions of low

LMX differentiation may feel that there is no incentive to increase performance levels, as there is little

chance of improving their LMX status. On the other hand, high LMX differentiation may be perceived

by low LMX members as a sign of hope for eventually developing a high LMX relationship with the

leader, as the leader is not forming low LMXs with everyone. For these individuals, high LMX

differentiation suggests that the leader is not uniformly developing low quality relations, and that the

leader is capable of, and interested in forming at least some high quality relationships. Such hopeful

individuals may increase their individual performance levels as a way of proactively increasing the

chances of improving their LMX quality.

Hypothesis 2: The relation between LMX differentiation and individual performance is moderated

by LMX, such that LMX differentiation is positively related to individual performance at low levels

of LMX but not related to individual performance at relatively higher levels of LMX.

The above hypothesis is consistent with the common finding of positive relations between LMX and

individual performance ratings. We expect that high LMX members tend to perform better than low

LMX members as shown in previous research (Gerstner & Day, 1997).

LMX Differentiation and group performance

Group performance is perhaps the most critical group-level variable in the leadership literature (Bass,

1990; Yukl, 2002, p. 11). For example, studies on Fiedler’s (1967) contingency model and the

transformational model of leadership (Bass, 1990; House & Aditya, 1997) have focused on the effects

of leadership on group performance. Given that group performance is arguably the key criterion for

evaluating leader effectiveness, it is surprising that group performance has been ignored in LMX

research. Although there is evidence from other streams of leadership research that leader behavior has

an impact on group performance, prior studies have not examined how the differentiation process, as

described in the LMX literature, is related to group performance. To the extent that the LMX

perspective on leadership is unable to offer an account of how leaders affect the performance of their

groups, it will be seen as an incomplete theory of leadership.

Although the LMX literature is silent with respect to the effects of differentiation on group

performance, role theory suggests that group performance should be enhanced by role differentiation.

A critical manifestation of LMX differentiation is a division of labor that occurs as the result of role

differentiation during the formation of LMX relationships. Specifically, the type of LMX relationship

that emerges from differentiation can be interpreted as an indicator of a member’s status within a group.

Whereas low LMX members have been portrayed in the LMX literature as being of lower status with

terms such as ‘hired hands’ used to describe them, high LMXmembers are thought to be of high status,

as indicated by terms such as ‘cadre’ to characterize them (Dansereau et al., 1975). Thus, there is a

direct correspondence between status and LMX. Because group tasks typically range from the

mundane to those requiring special technical or social skills, leaders rely on low status members to

perform routine work and high status ‘special assistants’ to help them with the more challenging tasks

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

LMX DIFFERENTIATION 727

(Graen, 1976, p. 1241). Without such assistance, leaders either fail, or are forced to attempt handling all

of these tasks themselves, eventually succumbing to burnout (Graen, 1976).

Stogdill (1959) developed a theoretical framework in which role structure/differentiation mediates

the relation between member behaviors and group achievement, defined primarily as group

performance. Stogdill (1959, p. 44) contended that ‘the subdivision of the group task makes it possible

to apply individual member skills in places where they can be used most effectively for the

accomplishment of the group purpose.’ Stogdill (1959, p. 133) further argued that ‘a differentiated role

structure has utility in enabling the members to operate upon a complicated and continuing task, each in

accord with his [sic] specialized skills and abilities.’ Experimental research backs this assertion, as the

recognition of each group members’ expertise has been found to be positively related to group

performance (Littlepage, Schmidt, Whisler, & Frost, 1995; Moreland, Argote, & Krishnan, 1998). In

essence, the clearer and more accepted one’s own role and the roles of others, the better the team works

together and the higher the subsequent group performance (Stogdill, 1959).

In sum, we propose that leaders should benefit from selectively investing in their relationships with

subordinates, resulting in high and low-quality exchanges (i.e., high degree of LMX differentiation)

represented in the work group. Role differentiation also serves to clarify goals for all members, which

further facilitates group performance (Slater, 1965; Stogdill, 1959).

Hypothesis 3a: The greater the degree of LMX differentiation, the higher the group performance.

An alternative to Hypothesis 3a would be that greater LMX differentiation is negatively related to

group performance. Rather than fostering cooperation as members accept their roles, differentiation

might instead leadmembers to reduce or withhold effort in group tasks. The quality of the relationship a

member has with the leader is an important signal of the extent to which his or her contribution to the

group is valued relative to other members. Differentiation thus establishes a hierarchy within the group

whereby high LMX members enjoy greater standing relative to others (Sparrowe, Soetjipto, &

Kraimer, in press). Because relative standing is an antecedent of group cooperation (Tyler & Blader,

2000), in highly differentiated groups low LMX members are likely to withhold effort to the detriment

of the performance of the group as a whole. Empirical evidence of a similar relationship is found in

research on pay dispersion. Invoking a compensation analogy, Bloom (1999) demonstrated with major

league baseball statistics that teams with greater pay dispersion tended to have lower team performance

relative to teams with greater pay compression. In some respects, LMX status within a group may be

viewed as a reward. To the extent that there is overlap between LMX differentiation and pay dispersion,

a negative association between LMX differentiation and group performance may be expected.

Hypothesis 3b: The greater the degree of LMX differentiation, the lower the group performance.

Moderating effects of task interdependence

Stogdill’s thesis implies a positive relation between LMX differentiation and team performance as

members accept their roles, but if differentiation creates disparities in relative standing and group

members withhold effort then this relationship may instead be negative. However, the tension created

by these opposing hypotheses may be resolved by considering the context as a moderator of the relation

between LMX differentiation and group performance. Specifically, we propose that the nature of this

relation varies based on the degree of task interdependence. Task interdependence refers to the degree

to which group members need to work closely with others, share material, information, and expertise in

order to complete their tasks (Cummings, 1978). Although task interdependence is portrayed by

organizational theorists as predicated by technology, the social psychological approach we adopt views

interdependence as being based on ‘cooperation requirements’ related to ‘group-level goals and

feedback, areas not dependent on technology’ (Stewart & Barrick, 2000, p. 138). Task interdependence

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

728 R. C. LIDEN ET AL.

is a group-level phenomenon (Schnake & Dumler, 2003), and by definition, as task interdependence

increases, so does each group member’s amount of dependence on others in the team. The increased

interdependence creates the need to coordinate in a more effectivemanner, and assist group members in

the completion of tasks.

When task interdependence is high, greater communication, cooperation, and coordinated action

among group members are necessary for goal accomplishment (Shea & Guzzo, 1987; Wageman &

Baker, 1997). We argue that leaders are more likely to have an impact on group performance when task

interdependence is high. That is, in groups characterized by high task interdependence, there is a

greater need for the leader’s assistance in coordinating interaction among group members (Zalesny,

Salas, & Prince, 1995). Differentiating among members in the group and developing varying exchange

relationships may be one mechanism by which leaders coordinate the contributions of group members.

Because LMX is associated with the assignment of tasks, it is likely that leaders assign the most critical

tasks to individuals who have a high-quality exchange and less important tasks to individuals who have

a low-quality exchange (Graen, 1976). Thus, when there is greater LMX differentiation in the group,

the leader plays a more active role in distributing assignments among work group members. Through

the process of strategically allocating tasks among subordinates, the leader increases the likelihood of

high group performance. Finally, in highly task interdependent groups, LMX differentiation may be

perceived as more equitable than it is under conditions of low task interdependence.Whenmembers are

highly interdependent, low performers are more likely to be noticed by peers (Lepine & Van Dyne,

2001) enabling leaders to engage in more effective role differentiation.

Hypothesis 4: Task interdependence moderates the relation between LMX differentiation within

groups and group performance, such that differentiation enhances group performance when task

interdependence is high.

Moderating effects of Median LMX

We argue that another key contextual variable surrounding the LMX differentiation process is the

nature of LMX relationships across leaders’ work groups. Specifically, LMX differentiation may have

differing effects on group performance depending on the median level of LMX within a group. LMX

median is a measure of central tendency of LMXs in the group, thus serving as a group level indicator of

LMX. Because LMX is not a variable that has been studied at the group level in the past, there is not a

precedent for the appropriate measure of central tendency. We chose the median because LMX is a

variable for which group level agreement is not expected. Kozlowski and Klein (2000, p. 34) termed

these variables as having ‘configural unit properties.’ Kozlowski and Klein argued that variables such as

teammember abilities, or demographic characteristics can be aggregated to the group level to represent

the overall tendency in the group, but agreement in these characteristics is not expected. They also

contended that the method of aggregation of such characteristics depends on the theory in question, and

may include minimum, maximum, or measures of dispersion. Even though Kozlowski and Klein

(2000) did not comment on the appropriateness of the use of median, they did caution researchers

against the use of the mean. They contended that without a measure of agreement, the mean may not be

an appropriate summary statistic because of its ambiguity and equifinality (p. 34).

Graen and Uhl-Bien (1995) argued that when leaders form high quality relationships with most of

their members, the satisfaction of members with their relationships and with the support received from

their leaders leads to high levels of performance regardless of the level of differentiation. In a review of

40 studies, Brockner and Wiesenfeld (1996) concluded that individuals are more strongly affected by

fairness of decision-making procedures when they receive unfavorable outcomes. Thus, we argue that

when most members have high LMXs, they do not pay much attention to how the resources are

distributed. Additionally, in groups with a high LMXmedian, LMX differentiation may not necessarily

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

LMX DIFFERENTIATION 729

create a clear role and status differentiation, given that there are manymembers whomay be considered

as the trusted assistants of the leader. Therefore, we argue that LMX differentiation is not related to

group performance when LMX median is high.

When the LMX median is low, that is, when leaders are unable to form high LMXs with many

members, LMX differentiation will be more strongly related to group performance. Even when many

LMXs are of low quality, leaders of low median LMX groups may achieve high levels of group

performance through differentiation. In the absence of satisfaction and support derived from having

many high LMX relationships in the group, the strategic distribution of resources has a salient influence

on group performance. Additionally, high differentiation in groups with a low LMX median may lead

to a clear role and status differentiation in the group, by granting one or more members the role of

trusted assistants. Therefore, leaders may achieve high group performance by differentiating in groups

with a low LMX median.

Significantly lower group performance should be characteristic only of groups in which LMX

median and differentiation are low. This could occur because the leader does not differentiate and

forms uniformly low LMX relationships with all subordinates. Group performance would be low

because of the disenchantment among subordinates resulting from the failure of the leader to provide

support to members, and the lack of equity due to the homogeneity of the relationships. An explanation

for this low performance may be derived from early formulations of LMX theory (Jacobs, 1970; Graen,

1976) suggesting that leaders and their subordinates engage in reciprocal exchanges. Just as

subordinates are dependent on their leaders for support and challenging assignments, leaders are also

dependent on their subordinates to effectively fulfill group performance objectives (Hollander &

Offermann, 1990). Should all subordinates withhold effort and contribution in their exchanges with

leaders, group performance would suffer.

Our arguments pertaining to the effects of LMX differentiation on group performance are consistent

with the role theory argument that differentiation is positively related to group performance, because it

acknowledges that the perceived legitimacy of role differentiation and role acceptance is more likely

when some members attain high LMX status (cf. Slater, 1965; Stogdill, 1959). That is, if no workgroup

members have high LMX relationships with the leader, members are likely to reject the role

expectations sent by the leader, and thus group performance will suffer.

Hypothesis 5: The median level of LMX within a group moderates the relation between LMX

differentiation and group performance, such that greater differentiation will be positively related to

group performance when LMX median is low, but will not be related to group performance when

LMX median is high.

Organizational Context

Description of Organizations and Employees

Hypotheses were tested with data collected from employees representing six organizations located

in the Midwestern portion of the U.S. Organization 1 was a Fortune 100 firm operating within the

manufacturing sector. Our sample from this organization included non-unionized, mid-level

employees from the distribution and transportation areas, production employees such as foremen

and their supervisors, and low-level workers and engineers. Organization 2 was a Fortune 500

insurance company, and the respondents represent various departments of the company such as

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

730 R. C. LIDEN ET AL.

claims, sales, and business insurance. In this organization, mid-level employees constituted the

sample. Prior to data collection, Organization 1 and 2 had implemented an empowerment initiative

covering all participants of our study. Organization 3 was a telecommunications company and the

sample consisted of low-level customer service employees working in a call center. Organization 4

was a public Midwestern university, and the respondents were mid-level, non-academic,

administrative employees. Organization 5 was a large firm engaged in product distribution and mid-

level and managerial employees constituted the sample. Organization 6 was a manufacturing

organization, and the sample was made up of non-unionized assembly line workers engaged in the

packaging of consumer products.

Time Frame

The data were collected between 1993 and 1996.

Method

Sample

We collected data from 931 employees, representing 190 groups. Due to reasons such as employee

absence on the date of data collection, most groups had a response rate of less than 100 per cent.

Because of the importance of the group level in this study, a high within-group response rate was

essential. Therefore, we set 60 per cent as the minimum within-group response rate we used to include

groups in our analyses. This rate enabled us to eliminate groups in which the majority of respondents

did not respond, and include groups in which 2 out of 3 members responded. We also excluded five

‘groups’ with fewer than three members. As a result of these criteria, the number of groups that were

eligible for inclusion in the study was 120, consisting of 834 employees. Eighteen groups had within-

group response rates between 60 per cent and 69 per cent, 14 groups had response rates between 70 and

79 per cent, and the remaining 88 groups had response rates over 80 per cent. Organization 2 provided

the largest number of groups (41 groups), followed by Organization 1 (35 groups), and Organization 6

(21 groups). The rest of the organizations each supplied fewer than 10 groups to the sample

(Organization 3: seven groups; Organization 4: eight groups; Organization 5: eight groups). Actual

group sizes ranged between 3 and 24 with a mean of 8.45.

The resulting sample was diverse in terms of demographic characteristics: 51 per cent had a high school

education, professional training, or less; 14.3 per cent held an associate degree; 28.1 per cent held a college

degree; 6.4 per cent had a graduate degree; and the remaining employees declined to report their level of

education. 56.1 per cent of respondents were male, 43 per cent were female, and 0.8 per cent did not report

their gender. Caucasians constituted the majority in the group member sample (74.0 per cent). African

Americans comprised 12.8 per cent of the sample, Hispanics 5.5 per cent; Asians 3.6 per cent; and the

remainder of the participants did not report their race. Group members had an average organizational

tenure of 9.37 years and position tenure of 3.83 years. Among group leaders, 19.2 per cent had a high

school degree, professional degree, or less; 14.2 per cent held an associate degree; 55 per cent held a

college degree; and 11.7 per cent held an advanced degree. 65.8 per cent of leaders were male, 33.3 per

cent were female. Caucasians constituted 84.2 per cent of the leader sample, followed by African

Americans (12.5 per cent), Hispanics (1.7 per cent) and other (0.8 per cent). Leaders’ organizational tenure

per cent averaged 13 years and their average position tenure was 3.34 years.

Procedure

In all organizations, researchers (including the authors and five assistants) collected the data on site

during work hours. All members of the selected groups, as well as the formal leaders of the groups were

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

LMX DIFFERENTIATION 731

invited to participate. Members completed surveys in their groups on company time, and leaders

completed surveys individually. Group members (subordinates) supplied data on LMX and task

interdependence, and group leaders provided the information on individual performance, group

performance, and group size.

Measures

Individual performance : In all organizations, leaders evaluated member performance, using

measures designed to fit the tasks performed in the studied organizations. In Organizations 1, 2, and 3,

leaders rated individual performance in terms of technical competence, quantity of work, quality of

work, and overall job performance. In the first site of Organization 1, and in Organizations 2, and 3, the

leaders used a 5-point scale (1¼Minimally acceptable, 5¼ Superior-exceptional performance) and in

the second site of Organization 1, a 7-point scale (1¼Very Poor to 7¼Outstanding) to evaluate

performance. These scales had Cronbach alphas of 0.84 and 0.91 in Organization 1, and 0.92 and 0.89

in Organizations 2 and 3. Finally, in Organizations 4, 5, and 6, individual performance was measured

with a 7-item in-role behavior scale developed by Williams and Anderson (1991). Leaders were asked

to indicate their level of agreement with the items, using 7-point scales (1¼ Strongly Disagree to

7¼ Strongly Agree). In these organizations, reliabilities of individual performance scales were 0.87,

0.90, and 0.89. The averages of the individual performance items were taken to form a composite score

for each individual. Individual performance scores were standardized for each organization in order to

account for differences in the nature of the jobs that were performed.

LMX. LMX was measured from the group member’s (subordinate’s) perspective. Across research

sites, two different measures of LMX were used, but for both measures of LMX, 7-point (1¼ Strongly

Disagree to 7¼ Strongly Agree) response scales were used. Example items are: ‘I admire my

manager’s professional skills,’ ‘My manager is the kind of person one would like to have as a friend,’ ‘I

can trust my manager to look out for my best interest,’ ‘My manager is very loyal to me,’ and ‘‘My

manager would come to my defense if I were ‘attacked’ by others.’’ In the first site of Organization 1, in

Organization 2, and in Organization 3, LMX was measured with the LMX-13 scale (Settoon et al.,

1996). The resulting reliabilities were 0.95, 0.96, and 0.96, respectively. In the second site of

Organization 1, and in Organizations 4, 5 and 6, 11 items from LMX-MDM scale (Liden and Maslyn,

1998) were used. Coefficient alphas in these sites were 0.91, 0.91, 0.88, and 0.91, respectively. The

twelfth item of the LMX-MDM measure, which was reported in the addendum of Liden and Maslyn

(1998), was not used because it was developed after the collection of the data reported in this study.

Both LMX-13 and LMX-MDM are multidimensional measures of LMX, and the four dimensions have

been previously shown to fall under a second-order factor (Settoon et al., 1996; Liden &Maslyn, 1998).

Becausewewere interested in capturing overall LMX quality rather than specific dimensions, the items

were averaged to form a composite for both LMX measures.

In order to ensure that the two LMX scales used in this study both captured the same construct, we

collected data from a sample of 179 working undergraduate students representing two Midwestern

universities (100 per cent response rate for students attending class on the day of survey

administration). Students responded to 13 items from the LMX-13 scale (Settoon et al., 1996), 11 items

from the LMX-MDM scale (Liden &Maslyn, 1998), and 7 items from the LMX-7 scale (Graen & Uhl-

Bien, 1995) with reference to their current or recent supervisors. Our results showed that LMX-13 and

LMX-MDM were significantly correlated (r¼ 0.95, p< 0.01). In addition, the LMX-13 and LMX-

MDM scales correlated with LMX-7 at 0.83 (p< 0.01) and 0.83 (p< 0.01), respectively. These results

showed that the different LMXmeasures were strongly correlated, and provided support for their use in

the current investigation.

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

732 R. C. LIDEN ET AL.

LMX Differentiation : Within-group variance in individual-level LMX scores was used to

operationalize LMX differentiation for each group.

LMXMedian : The within-group median of individual-level LMX scores was used to operationalize

LMX median for each group.

Task interdependence : Consistent with prior research (Van Der Vegt, Emans, & Van de Vliert,

2001), we measured task interdependence at the individual level. Employees reported the extent to

which they depended on others for completion of their work using three items adapted from Pearce and

Gregersen (1991). Respondents reported their agreement with items such as ‘I frequently must

coordinate my efforts with others in my work group’ using a 7-point scale (1¼ Strongly Disagree,

7¼ Strongly Agree). The reliability of this measure in the overall sample was 0.74. We used the

average individual task interdependence to operationalize within-group task interdependence. Before

aggregating the individual responses to the group level, we calculated interrater agreement using three

indices. Median rwg (James, Demaree, &Wolf, 1993) was 0.86, suggesting high levels of within-group

agreement. ICC (1) was 0.13 and ICC (2) was 0.52, suggesting moderate levels of agreement and

between-group differences (Bliese, 2000). These indices support the aggregation of individual ratings

to create group scores for task interdependence.

Group performance : Group performance was measured with five items. During interviews with

researchers, leaders were asked to rate their work group as a whole on the following performance

dimensions: quality of work, quantity of work, overall group performance, completing work on time,

and responding quickly to problems using 7-point scales (1¼Very poor to 7¼Outstanding). For all six

organizations, the items for the group performance measure were averaged to form a composite.

Cronbach alphas were: 0.78 (Site 1, Organization 1), 0.91 (Site 2, Organization 1), 0.88 (Organization

2), 0.91 (Organization 3), 0.87 (Organization 4), 0.83 (Organization 5), and 0.94 (Organization 6).

Control Variables : We controlled for Employee organizational tenure because organizational

experience may relate to understanding norms for performance and therefore may be related to

individual and group performance. Moreover, employees with greater organizational tenure may find it

easier to form high quality relationships with their leaders. Group leaders’ organizational tenure was

also controlled for, as this could affect the ability of leaders to differentiate effectively. We measured

organizational tenure of both employees and leaders in months. We controlled for Organization by

creating a dummy code for five of the six organizations that were surveyed in the study. Group sizewas

reported by group leaders, and was operationalized as the number of employees in the work group.

Finally, average individual performance was the mean individual performance ratings reported by

group leaders, and was added as a control variable in group level analyses.

Results

Preliminary analyses

Before merging the data from different organizations, we checked whether there were any differences

across organizations on the variables of interest by performing one-way ANOVAs. At the individual

level, ANOVA results showed that organization was related to individual LMX (F (5,814)¼ 4.41,

p< 0.01) but was not related to individual performance (F (5, 814)¼ 0.19, p> 0.05). At the group

level, ANOVA results suggested that organization was related to team performance (F (5, 114)¼ 5.20,

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

LMX DIFFERENTIATION 733

p< 0.01) and task interdependence F (5, 114)¼ 4.43, p< 0.01), but was not related to average

individual performance (F (5, 114)¼ 0.67, p> 0.05), LMX differentiation (F (5, 114)¼ 0.44, p> 0.05

or LMX median (F (5, 114)¼ 1.49, p> 0.05). Because of the between-organization differences in

some of the variables of interest, we controlled for organization in all analyses.

Tests of Hypotheses

The means, standard deviations, and correlations among the variables appear in Table 1. Hypotheses 1

and 2 examined the relationship between LMX differentiation and individual performance. Because of

their multi-level nature, we tested Hypotheses 1 and 2 using Hierarchical Linear Modeling (HLM).

HLM is the recommended technique for studying relationships at cross-levels and when the data are

hierarchically nested (Bryk & Raudenbusch, 1992; Hofmann, 1997). Hypothesis 3, which predicted a

relationship between LMX differentiation and group performance was tested using hierarchical

regression analysis. Finally, Hypotheses 4 and 5 examined the moderating effects of task

interdependence and LMX median on the relationship between LMX differentiation and group

performance, and were tested using hierarchical moderated regression.

Hypotheses 1a and 1b stated that there would be a relationship between LMX differentiation and

individual performance. In order to test Hypotheses 1a and 1b, we first examined the level of within and

between group variance (Hofmann, 1997; Snijders & Bosker, 1999). Thus, we performed the one-way

analysis of variance procedure in HLM and calculated the ICC coefficient. Between-group variance in

individual performance was significant (t00¼ 0.25, x2 (120)¼ 378.47, p< 0.01) and the ICC was 0.24.

These results suggested that there was substantial within and between-group variation in individual

performance. Then, we performed an intercepts-as-outcomes model in HLM. All variables were

entered after centering them by the grand mean. We entered employee LMX and organizational tenure

as individual-level control variables, and LMXmedian, organization, leader’s organization tenure, and

group size as group-level control variables. As presented in Table 2, LMX differentiation was not

significantly related to individual performance, failing to provide support for Hypothesis 1a or 1b. Only

individual LMXwas positively related to individual performance, whereas neither LMX differentiation

nor LMXmedian were related to individual performance controlling for individual LMX. The residual

variance remaining in the intercept was significant (t00¼ 0.23, x2 (93)¼ 179.39, p< 0.01) and a

comparison of this parameter with the residual variance in the random regression model suggested that

the level-2 variables did not explain any variance in the intercept.

Hypothesis 2 suggested that an individual’s own LMX would moderate the relationship between

LMX differentiation and individual performance. We estimated a slopes-as-outcomes model in HLM

to assess the moderating effect of individual LMX on the relation between LMX differentiation and

individual performance. In this approach we treated a level-1 variable (LMX) as the moderator, and the

level-2 variable (LMX differentiation) as the component ‘main effect,’ rather than the more common

approach whereby the level-2 variable is conceptualized as the moderator. However, the mathematical

computations are identical in either case, just as in tests for moderation using ordinary least squares and

a single level of analysis (Cohen, Cohen, West, & Aiken, 2003, p. 555). We entered LMX

differentiation as a predictor of the intercept and LMX—individual performance slope, while we

controlled for employee organizational tenure, leader’s tenure, group size, organization, and LMX

median. We centered the level-1 predictor (LMX) by the group mean, as recommended by Hofmann

and Gavin (1998). The remainder of the variables were centered by the grand mean. The results of this

analysis are presented in Table 2. The results showed that there was a significant interaction between

LMX differentiation and individual LMX. Comparing the residuals of this model with the residual

of the intercepts as outcomes model (Hofmann, 1997), we found the pseudo R2 for the interaction to be

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

Table

1.Descriptivestatistics

andcorrelationsa

Variable

Mean

SD

12

34

56

78

910

11

12

Individual

level

variables

1.Individual

perform

ance

0.02

1.00

2.Individual

LMX

5.14

1.16

0.16

3.Employee

organizational

tenure

112.54111.01

0.05

0.02

Grouplevel

variables

1.LMX

differentiation

1.03

0.81

2.LMX

median

5.28

0.74�0.28��

3.Taskinterdependence

5.35

0.56�0.09

0.23��

4.Groupperform

ance

4.88

0.88

0.07

0.20�

0.05

5.Averageem

ployee

tenure

116.95

84.78

0.07

0.13

�0.15

0.09

6.Groupleader

organizational

tenure

156.05118.12

.02

0.17

�0.16

0.39��

0.33�

7.Organization1

0.29

0.45�0.04

�0.08

0.08

�0.06

�0.04

0.23�

8.Organization2

0.34

0.47

0.04

0.20�

�0.18�

0.28��

0.13

0.13

�0.46��

9.Organization3

0.05

0.23

0.10

�0.14

0.06

0.10

0.07

0.04

�0.16

�0.17

10.Organization5

0.06

0.25�0.08

�0.08

�0.28��

�0.06

�0.10

�0.00

�0.17

�0.19�

�0.06

11.Organization6

0.17

0.38�0.00

0.01

0.25��

�0.36��

�0.03

�0.48��

�0.29��

�0.33��

�0.11�0.12

12.Groupsize

8.45

4.58�0.00

�0.04

�0.14

�0.08

�0.14

�0.03

0.36��

�0.09

�0.03

0.22��0.31��

13.Averageindividual

perform

ance

0.07

0.65�0.00

0.23��

0.16

0.56��

0.11

0.17

�0.03

0.09

0.11�0.01

�0.09

�0.15

aForindividual

level

correlations,n¼834.Forgrouplevel

correlations,n¼120.Organizational

tenure

was

measuredin

months.

� p<0.05;��p<0.01.

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

734 R. C. LIDEN ET AL.

Table 2. HLM results testing Hypotheses 1 and 2

Independent variables

Hypothesis 1 Hypothesis 2

CoefficientStandarderror df t-ratio Coefficient

Standarderror df t-ratio

Intercept1, b0Intercept, g00 0.06 0.05 110 1.18 0.08 0.05 110 1.44Group size, g01 �0.03 0.01 110 �2.61�� �0.03 0.01 110 �2.48�

Leader’s organizationaltenure, g02

�0.00 0.00 110 �0.06 �0.00 0.00 110 �0.10

Organization 1, g03 0.34 0.18 110 1.85 0.26 0.17 110 1.51Organization 2, g04 0.24 0.15 110 1.66 0.18 0.14 110 1.31Organization 3, g05 0.56 0.29 110 1.92 0.51 0.29 110 1.76Organization 5, g06 0.36 0.23 110 1.58 0.26 0.22 110 1.17Organization 6, g07 0.11 0.20 110 0.57 �0.01 0.19 110 �0.05LMX median, g08 0.06 0.07 110 0.88 0.16 0.06 110 2.44�

LMX differentiation, g09 0.09 0.06 110 1.49 0.02 0.06 110 0.43Individual LMX, b1Intercept, g10 0.12 0.03 119 3.62�� 0.16 0.03 118 4.35��

LMX differentiation, g11 — — — — �0.08 0.03 118 �2.38��

Employee organizationaltenure, b2Intercept, g20 0.00 0.00 119 0.68 0.00 0.00 119 0.75

�p< 0.05, one-tailed; ��p< .01, one-tailed.

LMX DIFFERENTIATION 735

24 per cent. Pseudo R2 refers to the percentage of the slope variance accounted for by LMX

differentiation. It is calculated by subtracting the residual slope variance of the current model from the

residual slope variance of the intercepts-as-outcomes model, and dividing this figure by the residual

slope variance of the intercepts-as-outcomes model.

After we entered LMX differentiation into the equation, the residual variance in the LMX—

performance slope was no longer significant (t11¼ 0.01, x2 (102)¼ 118.24, p> 0.05). We plotted the

significant interaction in Figure 1, at high and low levels of individual LMX and LMX differentiation

(one standard deviation above and below the mean). As presented in the figure, the relation between

LMX differentiation and individual performance was not significant for high LMX individuals,

whereas the relationship was positive for individuals low in LMX. These results provide support for

Hypothesis 2.

Hypotheses 3a and 3b predicted a relationship between LMX differentiation and group performance.

To test these hypotheses, we conducted a hierarchical regression analysis. We entered the control

variables in the first step (group size, leader and employee tenure, organization, LMX median, and

average individual performance), and entered LMX differentiation in the second step. The results are

presented in Table 3. Results from this analysis suggested that LMX differentiation was not

significantly related to group performance, failing to provide support for Hypothesis 3a and 3b.

Hypothesis 4 predicted a moderating effect of task interdependence on the LMX differentiation to

group performance relationship. We tested Hypothesis 4 using hierarchical moderated regression

analysis. We regressed group performance on the control variables in the first step. LMX differentiation

and task interdependence were entered in the second step after they were centered (Aiken & West,

1991). The interaction term was entered into the equation in the last step. The results of this analysis are

presented in Table 4. The change in R2 between the 2nd and 3rd steps was significant; indicating that

entering the interaction term explained additional variance in group performance (DR2¼ 0.02,

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

Figure 1. The relationship between LMX differentiation and individual performance for high and low levels ofLMX

Table 3. Hierarchical regression analyses with group performance as the dependent variable

Independent variable Step 1a Step 2

Group size 0.00 0.00Leader’s organizational tenure 0.24�� 0.23��

Organization 1 �0.25 �0.24Organization 2 �0.04 �0.05Organization 3 �0.04 �0.05Organization 5 �0.15 �0.14Organization 6 �0.31� �0.31�

Average employee tenure �0.07 �0.08Average individual performance 0.49�� 0.49��

LMX median 0.03 0.06LMX differentiation 0.08DR2 0.50 0.01DF 10.88�� 1.24Adj. R2 0.45 0.45Overall R2 0.50 0.51Overall F 10.88�� 10.02��

aStandardized regression coefficients are reported.�p< 0.05; ��p< 0.01.

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

736 R. C. LIDEN ET AL.

Table 4. Moderated hierarchical regression analyses with group performance as the dependent variable

Independent variable

Hypothesis 4 Hypothesis 5

Step 1a Step 2 Step 3 Step 1 Step 2 Step 3

Group size 0.00 0.00 0.00 0.01 0.00 0.01Leader’s organizational tenure 0.24�� 0.24�� 0.24�� 0.25�� 0.23�� 0.22�

Organization 1 �0.25 �0.24 �0.22 �0.25 �0.24 �0.22Organization 2 �0.04 �0.04 �0.00 �0.04 �0.05 �0.00Organization 3 �0.04 �0.05 �0.04 �0.05 �0.05 �0.03Organization 5 �0.15 �0.13 �0.11 �0.16 �0.14 �0.11Organization 6 �0.31� �0.32� �0.31� �0.30� �0.31� �0.28�

Average employee tenure �0.07 �0.07 �0.08 �0.07 �0.08 �0.05Average individual performance 0.49�� 0.48�� 0.49�� 0.50�� 0.49�� 0.48��

LMX median 0.03 0.04 0.05 0.06 0.06LMX differentiation 0.08 0.09 0.08 0.05Task interdependence 0.06 0.10 — — —LMX differentiation� task interdependence 0.14� — —LMX differentiation�LMX median �0.16�

DR2 0.50 0.01 0.02 0.50 0.01 0.02DF 10.88�� 0.94 3.96� 12.15�� 0.71 5.23�

Adj. R2 0.45 0.45 0.47 0.45 0.45 0.48Overall R2 0.50 0.51 0.53 0.50 0.51 0.53Overall F 10.88�� 9.21�� 9.04�� 12.15�� 10.02�� 9.98��

aStandardized regression coefficients are reported.�p< 0.05; ��p< 0.01.

LMX DIFFERENTIATION 737

DF¼ 3.96, p< 0.05). The overall regression equation was also significant (R2¼ 0.53, F (13,

106)¼ 9.04, p< 0.01). Figure 2 demonstrates the nature of the moderating relationship of task

interdependence. We plotted LMX differentiation and group performance at high and low levels of task

interdependence (Aiken & West, 1991). High and low were defined as one standard deviation above

and below the mean, respectively. For groups with low task interdependence, LMX differentiation was

Figure 2. The relationship between LMX differentiation and group performance for high and low levels of taskinterdependence

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

Figure 3. The relationship between LMX differentiation and group performance for high and low levels of LMXmedian

738 R. C. LIDEN ET AL.

not significantly related to group performance (b¼�0.09, t¼�0.82, p> 0.05). On the other hand, for

groups with high task interdependence, the relationship between LMX differentiation and group

performance was positive and significant (b¼ 0.27, t¼ 2.28, p< 0.05). These results provided support

for Hypothesis 4.

Finally, we tested Hypothesis 5 using hierarchical moderated regression. In this hypothesis, we

expected LMX median to moderate the relationship between LMX differentiation and group

performance. As presented in Table 4, there was a significant interaction between LMX differentiation

and LMX median, which explained 2 per cent of the variance in group performance. The interaction is

plotted in Figure 3. For groups with a low LMX median, LMX differentiation was positively related to

group performance (b¼ 0.23, t¼ 2.38, p< 0.05). On the other hand, for groups with a high LMX

median, the relationship between LMX differentiation and group performance was not significant

(b¼�0.11, t¼�1.05, p> 0.05). These results provided support for Hypothesis 5.

Discussion

The current investigation empirically examined implications of the major premise of LMX theory, that

leaders differentiate between followers in terms of the quality of relationship formed with each. Rather

than directly testing the effects of differentiation, LMX research has proceeded for three decades with a

focus on identifying the antecedents and outcomes of follower LMX status. Although informative,

extant research has not addressed the bottom line question, ‘Does leader differentiation among

followers hurt or enhance performance?’ Results of the current investigation suggest that the answer to

this question is ‘It depends.’ Although there was not a main effect for LMX differentiation with respect

to individual performance, our results revealed that LMX differentiation and individual performance

were positively related for low LMX members, but not related for high LMX members. At the group

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

LMX DIFFERENTIATION 739

level, our results suggested that the relationship between LMX differentiation and group performance

could be best understood by considering moderators, the level of task interdependence and the median

level of LMX within the group. The association between LMX differentiation and group performance

was significant and positive for groups high in task interdependence. But for groups that were low in

task interdependence, the relation between LMX differentiation and group performance was not

significant. With respect to the LMX median, the association between LMX differentiation and group

performance was positive in groups with a low LMXmedian, but not significant for groups with a high

LMX median.

At the individual level, increasing LMX differentiation has opposite effects for high and low LMX

members, even when controlling for LMXmedian. Although our results show, consistent with previous

research (Gerstner & Day, 1997), that higher LMX members tend to be rated higher than low LMX

members on performance, the difference in rated performance between high and low LMX members

becomes increasingly smaller with increases in LMX differentiation. When low LMX employees

belong to a group with high LMX differentiation as opposed to low differentiation, they may feel that

with an increase in effort, they too might succeed in increasing their LMX status. On the other hand, the

degree of LMX differentiation appeared to have little effect on high LMX members. Regardless of the

degree of LMX differentiation, social exchange theory suggests that high LMX members feel the need

to reciprocate the preferential treatment by performing at high levels. Research is needed to explore the

causal mechanisms behind our observed interaction between LMX differentiation and LMX with

respect to individual performance.

The moderating effect of task interdependence on the relation between LMX differentiation and

group performance informs both the LMX and the role theory literatures. This finding adds value to the

LMX literature by demonstrating the relevance of studying LMX relationships at the group level, and

for providing an initial examination of the way in which LMX differentiation relates to group

performance. Because LMX differentiation is a specific type of role differentiation (Graen, 1976), the

results contribute to the literature on roles by identifying contextual factors that influence the effects of

creating different roles within groups and resulting group performance.

Rather than LMX differentiation having a direct positive (Dansereau et al., 1975; Dienesch & Liden,

1986; Liden et al., 1997) or negative (Uhl-Bien et al., 2000) effect on group performance, our results

suggest that the relationship between LMX differentiation and group performance depends on the level

of structural interdependence present in the work group, specifically task interdependence. Our results

also demonstrate that LMX differentiation is more highly related to group performance in groups with

low levels of median LMX than in groups with high median LMX. So, the advantages of strategically

differentiating between group members appear to materialize only when the nature of tasks require

group members to work closely together or when high LMX relationships are relatively rare within the

group.

Many of our hypotheses that were framed with social exchange/role theory explanations for

reactions to differentiation could be contrasted with rationale derived from compensation research on

reward distributions. Interestingly, research on reward dispersion has generally shown that group

performance suffers with wide pay dispersion and benefits from compression (Bloom, 1999; Pfeffer &

Langton, 1993). With respect to individual performance, Bloom’s argument that the impact of pay

dispersion on performance depends on one’s pay relative to others in the group is similar to our

hypothesis that the impact of LMX differentiation depends on one’s own LMX. Bloom’s results showed

that the relationship between pay dispersion and individual performance was more negative among

relatively low-paid team members. In contrast, our results supported the claim that the performance of

low LMXmembers increased with differentiation—a relationship we attributed to members’ efforts to

improve their LMX status and relative standing through improved performance. One possible

explanation for this difference lies in how easily members are able to affect the status and rewards they

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

740 R. C. LIDEN ET AL.

receive through higher levels of performance versus moving up the scale in terms of relative

compensation. In our sample, it is plausible that members enjoyed frequent opportunities to improve

their LMX status through performance-oriented efforts. However, in Bloom’s sample of major league

baseball players, compensation is negotiated contractually on a multi-year basis. Thus, with respect to

LMX, opportunities for redress of perceived unfairness and/or improving status were available to

members. Not so for pay dispersion; relatively low paid team members could only suffer any perceived

inequities and withhold effort.

With respect to group performance, Bloom (1999) found a negative relationship between pay

dispersion and team performance and attributed it to the context in which demands for coordination and

cooperation are high (e.g., baseball teams). Unlike Bloom’s sample, in which one might assume that all

baseball teams are equivalent in terms of task interdependence, our sample included groups that

performed a variety of different tasks. By measuring task interdependence among these groups we are

able to speak directly to the question of how the task context shapes the relationship between

differentiation and group performance. Our results demonstrate a moderating relationship that is the

opposite of what Bloom found with respect to pay dispersion; that is, LMX differentiation enhanced

performance where work required cooperation and coordination (i.e., high task interdependence).

One possible explanation why some forms of differentiation, such as pay dispersion, are detrimental

to team performance, but others, such as LMX differentiation, enhance group performance where the

requirements for coordination and cooperation are high, lies in their respective differences. Pay

dispersion—particularly in settings where salaries are contractually negotiated on amulti-year basis—

is largely exogenous to the work of the team. It is not a factor the leader (or, in Bloom’s [1999] sample,

the coach of the baseball team) can employ in coordinating the work on a day-to-day or moment by

moment basis. LMX differentiation, we suggested above, reflects the leader’s assignments of tasks and

coordination of the work. Where group performance involves relatively high levels of coordination and

cooperation, LMX differentiation is effective.

Limitations and Strengths

There are several limitations of this study. First, we were not able to pinpoint exactly what

characteristics, behaviors, and attitudes leaders relied on most when differentiating between

subordinates. It is plausible that employees respond to varied types of differentiation in different ways.

For example, members may respond more positively to differentiation based on work-related criteria

(Scandura, 1999; Sias & Jablin, 1995), as opposed to differentiation based on personality similarity or

demographic similarity. Because the literature revealed a large set of potential variables used for

differentiation as evidenced by the substantial literature on antecedents of LMX (Liden et al., 1997), we

were not able to include a comprehensive measure of the varied contributions members can make to the

group and organization. Similarly, it would have been ideal to have included a role perceptions

measure. For example, in future research on LMX differentiation, it might be useful to include

measures of group member work activities as were assessed by Dansereau et al., (1975) and Liden &

Graen (1980). These data would enable researchers to identify more precisely how leaders differentiate

between group members.

A second limitation was that because data from several studies were used to test the hypotheses, the

items comprising the variables were not exactly the same for all respondents in the sample. Although

these differences are minor, we might have been measuring slightly different domains of our constructs

in each organization. Also, the relatively low ICC values (Bliese, 2000) for task interdependence

indicate that results involving this variable may represent conservative estimates of population effects.

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

LMX DIFFERENTIATION 741

A third limitation was that participation in this study was voluntary and some members chose not to

participate. Response rates within groups are a critical issue for all group-level studies. The research

hypotheses in this study would have been most accurately tested with data collected from all members

of the groups. But in order to achieve adequate power, it was necessary to retain all groups with a

response rate of at least 60 per cent. If, for example, non-respondents were mostly low LMX

employees, failing to obtain their responses may have influenced the results. It was not possible to test

the assumption that non-respondents did not differ on such critical variables as LMX and individual

performance. Thus, the fact that we were unable to gather data from all members of the groups is one of

the limitations of our study.

A final limitation is the cross-sectional design employed, which does not allow us to infer causality

from the results. Our theory predicts that the direction of the relationship should flow from LMX to

performance. However, alternative explanations, such as performance leading to LMX (Bauer &

Green, 1996), are also possible.

Despite these limitations, there are also a number of strengths in the present study. One strength,

which serves to increase external validity, is that the data were collected in six organizations

representing diverse fields. Also, we gathered data from two independent sources, largely reducing

common source bias explanations for our results. Finally, we believe that our study explored an

important issue that offers several practical implications and may serve as an impetus for future

research.

Practical Implications

The results of our study suggest a number of implications regarding the role of LMX differentiation on

individual and group performance. Interestingly, LMX differentiation was positively related to

individual performance for low LMX members but was not related to performance for high LMX

members. It seems that low LMX members may be motivated to enhance their relationship with their

leader by increasing their performance because they may see first hand how high LMX members

benefit from high-quality exchanges with the leader. Thus, there are potential benefits of having low

LMX members observe coworkers’ high LMX relations with the leader. Rather than creating

resentment, low LMX members seem to be motivated to improve their standing and performance.

Thus, organizations should consider providing low LMX members with opportunities to observe

contrasting relationships (i.e., high quality) that coworkers have with their leader.

Our findings also suggest that the performance of task interdependent groups can be enhanced

through LMX differentiation. This result raises several intriguing practical implications. In team

environments, which are characterized by high levels of task interdependence, performance gains can

be achieved if leaders differentiate between subordinates. By providing additional resources and

support to some members, differentiation can ensure an effective utilization of the resources available

to the leader.When task interdependence is high, LMX differentiation may thus provide benefits to

groups. On the other hand, when task interdependence is low, or when individuals accomplish their

tasks in relative isolation from each other, LMX differentiation does not seem to have any effects on

group performance. Thus, provided that task interdependence is high, our results suggest that leaders

may enhance group performance through LMX differentiation. In order to elaborate on the above

results, for a subset of the organizations studied in the current investigation, job autonomy,

empowerment, and perceived control were measured. The significant correlations found between LMX

and autonomy (r¼ 0.29, p< 0.01 N¼ 480), empowerment (r¼ 0.29, p< 0.01, N¼ 429), and

perceived control (r¼ 0.21, p< 0.01, N¼ 612) are consistent with previous research and suggest that

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

742 R. C. LIDEN ET AL.

LMX differentiation involves the differential provision of challenge and responsibility in task

assignments.

In addition to task interdependence, our results suggest that LMX at the group level, assessed by

LMX median, is an important contextual factor in understanding the relation between LMX

differentiation and group performance. In groups where leaders have developed very few high LMX

relationships (i.e., LMX median is low), greater LMX differentiation is positively related to group

performance. Thus, if a leader has lower quality relations with many group members, it is advantageous

in terms of enhancing group performance for the leader to identify a selected few members and

establish higher-quality exchanges with those members. This situation may occur when a well-liked

leader is abruptly fired or replaced and many members are unreceptive to a new leader. The new leader

should focus on increasing LMX differentiation by developing high-quality exchanges with a few

selected members. In contrast, group performance benefits do not occur when a leader who has many

high-quality relations (high LMX median) develops a few additional high LMX relationships with

other group members.

Future Directions

In future studies, it is necessary to search for additional moderating variables of the LMX

differentiation-group performance relation. Perhaps the most important moderator not tested in the

current investigation is member perceived fairness of LMX differentiation. That is, to what extent do

subordinates feel that leaders’ formation of different quality relationships with each subordinate is

justified. When subordinates generally believe that the leader’s differentiation criteria are justified or

fair, they are more inclined to accept their level of status in the group (Sias & Jablin, 1995), and thus the

relation between LMX differentiation and group performance may be stronger. A potentially fruitful

area of future research would be to examine group member perceptions of LMX differentiation fairness

as a moderator of the relationship between LMX differentiation and performance. The relationship

between LMX differentiation and group performance may be positive only when employees perceive

LMX differentiation to be fair, whereas the relationship could be negative when LMX differentiation is

perceived as unfair. The first step in such as study is to examine what members view as fair

differentiation. Currently, it is not clear what norm of fairness (equity, equality, or need) would be

relevant for LMX differentiation. Furthermore, testing this hypothesis requires the presence of within-

group agreement in fairness perceptions of employees. The relational model of justice (Lind & Tyler,

1988) is particularly germane to understanding the responses of group members to leader

differentiation, because this approach to justice focuses on social bonds and relationships rather than

the value of resources received from exchange partners. ‘The basic assumption of the relational model

is that people are predisposed to belong to social groups and that they are very attentive to signs and

symbols that communicate information about their position within groups (Tyler, 1994, p. 851).’ The

symbols that appear to be most salient are the: (1) quality of one’s own relationship with the leader; (2)

degree to which individuals feel that leaders create a ‘level playing field’ in the treatment of group

members; and (3) extent to which group members feel that the leader treats members fairly.

Another moderator that might be explored in future research is the effects of leader time-pressures on

differentiation and subsequent group performance. Although Kinicki and Vecchio’s (1994) results

suggested that time-pressured leaders may invest in more, rather than fewer, high LMX relationships

(thus differentiating less), it may be necessary to conduct social network research to establish a clearer

picture of the effects of time-pressure. Social network researchers have argued that developing many

strong relationships is costly in terms of time and resources (Brass, 1995). However, this prescription

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

LMX DIFFERENTIATION 743

may not apply to a leader’s immediate work group, consistent with Krackhardt’s (1992) work on the

value of strong ties.

It is also important to examine the relations between LMX differentiation and group process

variables in addition to task interdependence. The literature in social psychology on group performance

has identified a plethora of explanatory variables (Guzzo & Dickson, 1996). Therefore, the population

effect size for the LMX differentiation—group performance relation might be small, and difficult to

uncover. However, LMX differentiation may have a stronger association with group process variables

such as group cohesiveness, within-group conflict, and group potency. The analysis of LMX

differentiation with respect to these and other group-level variables would make valuable additions to

both the LMX and group literatures.

Finally, this study examined LMX differentiation in relation to one outcome, in-role performance.

Given that LMX has been linked to contextual performance (e.g., Liden et al., 1997), future research

should explore whether our results extend to other outcomes including contextual performance.

Research on such issues would serve to further extend LMX theory to the group level beyond our initial

efforts in the current research.

Acknowledgments

This study was funded by a grant from the Center for Human Resource Management (CHRM) at the

University of Illinois. Additional funding was provided by the Society for Human Resource Manage-

ment (SHRM) Foundation. The interpretations, conclusions, and recommendations, however, are those

of the authors, and do not necessarily represent those of the CHRM or the SHRM Foundation. The

authors gratefully acknowledge the research assistance of Lisa Bradway, Monica Gavino, Isabel Graf,

Maria Kraimer, and Susan Murphy.

Author biographies

Robert C. Liden (Ph.D., University of Cincinnati) is Professor of Management and Director of the OB/

HR doctoral program at the University of Illinois at Chicago. His research focuses on interpersonal

processes within the context of such topics as leadership, groups, career progression, and employment

interviews.

Berrin Erdogan (Ph.D. University of Illinois at Chicago) is Assistant Professor of Management at

Portland State University. Her research focuses on attachment of employees to organizations as a result

of leader-subordinate relationships, organizational justice, and fit with organizational values.

Sandy J. Wayne (Ph.D., Texas A&M University) is Professor of Management and Director of the

Center for Human ResourceManagement at the University of Illinois at Chicago. Her research interests

include employee-supervisor relationships, employee-organization relationships, upward influence,

and perceived organizational support.

Raymond T. Sparrowe (Ph.D. University of Illinois at Chicago) is an Associate Professor of

Organizational Behavior in the John M. Olin School of Business at Washington University in

St. Louis. His research focuses on the social context of leadership.

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

744 R. C. LIDEN ET AL.

References

Adams, J. S. (1965). Inequity in social exchange. Advances in Experimental Social Psychology, 2, 267–299.Aiken, L. S., & West, S. G. (1991).Multiple regression: Testing and interpreting interactions. Newbury Park, CA:Sage Publications.

Bass, B. M. (1990). Bass & Stogdill’s handbook of leadership: Theory, research, and managerial applications (3rded.). New York: The Free Press.

Bauer, T. N., & Green, S. G. (1996). Development of leader-member exchange: A longitudinal test. Academy ofManagement Journal, 39, 1538–1567.

Bliese, P. D. (2000).Within-group agreement, non-independence, and reliability: Implications for data aggregationand analysis. In K. K. Klein, & S. W. J. Kozlowski (Eds.), Multilevel theory, research, and methods inorganizations (pp. 349–381). San Francisco: Jossey-Bass.

Bloom, M. (1999). The performance effects of pay dispersion on individuals and organizations. Academy ofManagement Journal, 42, 25–40.

Brass, D. J. (1995). A social network perspective on human resources management. Research in Personnel andHuman Resources Management, 13, 39–79.

Brockner, J., & Wiesenfeld, B. M. (1996). An integrative framework for explaining reactions to decisions:Interactive effects of outcomes and procedures. Psychological Bulletin, 120, 189–208.

Bryk, A. S., & Raudenbusch, S. W. (1992). Hierarchical linear models: Applications and data analysis methods.Newbury Park, CA: Sage Publications.

Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation analysis for thebehavioral sciences. Mahwah, NJ: Lawrence Erlbaum.

Crosby, F. (1976). A model of egoistical relative deprivation. Psychological Review, 83, 85–113.Cummings, T. G. (1978). Self-regulating work groups: A socio-technical analysis. Academy of ManagementReview, 3, 625–634.

Dansereau, F., Graen, G., & Haga, W. J. (1975). A vertical dyad linkage approach to leadership within formalorganizations: A longitudinal investigation of the role making process. Organizational Behavior and HumanPerformance, 13, 46–78.

Dienesch, R. M., & Liden, R. C. (1986). Leader-member exchange model of leadership: A critique and furtherdevelopment. Academy of Management Review, 11, 618–634.

Dotan, O., Goldstein, H., Nishii, L., Mayer, D. M., & Schneider, B. (April, 2004). Leader-member exchange,group-level processes, and group performance. Paper presented at the Annual Meeting of Society of Industrialand Organizational Psychology (SIOP).

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. G., & Taber, T. D. (1986). Vertical dyad linkage: A longitudinal assessment of antecedents,measures, and consequences. Journal of Applied Psychology, 71, 56–60.

Erdogan, B., & Liden, R. C. (2002). Social exchanges in the workplace: A review of recent developments andfuture research directions in leader-member exchange theory. In L. L. Neider, & C. A. Schriesheim (Eds.),Leadership (pp. 65–114). Greenwich, CT: Information Age Publishing.

Erdogan, B., & Liden, R. C. (2006). Collectivism as a moderator of responses to organizational justice:Implications for leader-member exchange and ingratiation. Journal of Organizational Behavior, 27, 1–17.

Feldman, J. M. (1986). A note on the statistical correction of halo error. Journal of Applied Psychology, 71, 173–176.

Fiedler, F. E. (1967). A theory of leadership effectiveness. New York: McGraw-Hill.Gerstner, C. R., & Day, D. V. (1997). Meta-analytic review of leader-member exchange theory: Correlates andconstruct issues. Journal of Applied Psychology, 82, 827–844.

Goodman, P. S., Ravlin, E., & Schminke, M. (1987). Understanding groups in organizations. Research inOrganizational Behavior, 9, 121–173.

Graen, G. (1976). Role-making processes within complex organizations. In M. D. Dunnette (Ed.), Handbook ofindustrial and organizational psychology (pp. 1201–1245). Chicago: Rand McNally.

Graen, G. B., & Scandura, T. A. (1987). Toward a psychology of dyadic organizing. Research in OrganizationalBehavior, 9, 175–208.

Graen, G. B., & Uhl-Bien, M. (1995). Development of leader-member exchange (LMX) theory of leadership over25 years: Applying a multi-level multi-domain perspective. Leadership Quarterly, 6, 219–247.

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

LMX DIFFERENTIATION 745

Guzzo, R. A., & Dickson, M. W. (1996). Teams in organizations: Recent research on performance andeffectiveness. Annual Review of Psychology, 47, 307–338.

Heneman, R. L., Greenberger, D. B., & Anonyuo, C. (1989). Attributions and exchanges: The effects ofinterpersonal factors on the diagnosis of employee performance. Academy of Management Journal, 32,466–476.

Hofmann, D. A. (1997). An overview of the logic and rationale of hierarchical linear models. Journal ofManagement, 23, 723–742.

Hofmann, D. A., & Gavin, M. B. (1998). Centering decisions in hierarchical linear models: Implications forresearch in organizations. Journal of Management, 24, 623–641.

Hollander, E. P., & Offermann, L. R. (1990). Power and leadership in organizations: Relationships in transition.American Psychologist, 45, 179–189.

House, R. J., & Aditya, R. T. (1997). The social scientific study of leadership: Quo Vadis? Journal of Management,23, 409–473.

Jacobs, T. O. (1970). Leadership and exchange in formal organizations. Alexandria, VA: HUMRROJames, L. R., Demaree, R. G., &Wolf, G. (1993). rwg: An assessment of within-group interrater agreement. Journalof Applied Psychology, 78, 306–309.

Kinicki, A. J., & Vecchio, R. P. (1994). Influences on the quality of supervisor-subordinate relations: The role oftime-pressure, organizational commitment, and locus of control. Journal of Organizational Behavior, 15, 75–82.

Kozlowski, S.W., &Klein, K. J. (2000). Amultilevel approach to theory and research in organizations: Contextual,temporal, and emergent processes. In K. J. Klein, & S. W. Kozlowski (Eds.), Multilevel theory, research, andmethods in organizations: Foundations, extensions, and new directions (pp. 3–90). San Francisco: Jossey-Bass.

Krackhardt, D. (1992). The strength of strong ties: The importance of philos in organizations. In R. Eccles, &N. Nohria (Eds.), Networks and organizations: Structure, form and action (pp. 216–239). Cambridge, MA:Harvard Business School.

Lepine, J. A., & Van Dyne, L. (2001). Peer responses to low performers: An attributional model of helping in thecontext of groups. Academy of Management Review, 26, 67–84.

Liden, R. C., & Graen, G. (1980). Generalizability of the vertical dyad linkage model of leadership. Academy ofManagement Journal, 23, 451–465.

Liden, R. C., & Maslyn, J. M. (1998). Multidimensionality of leader-member exchange: An empirical assessmentthrough scale development. Journal of Management, 24, 43–72.

Liden, R. C., Sparrowe, R. T., & Wayne, S. J. (1997). Leader-member exchange theory: The past and potential forthe future. Research in Personnel and Human Resources Management, 15, 47–119.

Liden, R. C., Wayne, S. J., & Sparrowe, R. T. (2000). An examination of the mediating role of psychologicalempowerment on the relations between the job, interpersonal relationships, and work outcomes. Journal ofApplied Psychology, 85, 407–416.

Lind, E. A., & Tyler, T. R. (1988). The social psychology of procedural justice. New York: Plenum.Littlepage, G. E., Schmidt, G. W., Whisler, E. W., & Frost, A. G. (1995). An input-process-output analysis ofinfluence and performance in problem solving groups. Journal of Personality and Social Psychology, 69, 877–889.

Maslyn, J. M., & Uhl-Bien, M. (2001). Leader-member exchange and its dimensions: Effects of self-effort andother’s effort on relationship quality. Journal of Applied Psychology, 86, 697–708.

Masterson, S. S., Lewis, K., Goldman, B. M., & Taylor, M. S. (2000). Integrating justice and social exchange: Thediffering effects of fair procedures and treatment on work relationships. Academy of Management Journal, 43,738–748.

McClane, W. E. (1991). Implications of member role differentiation: Analysis of a key concept in the LMXmodelof leadership. Group & Organization Management, 16, 102–114.

Moreland, R. L., Argote, L., &Krishnan, R. (1998). Training people towork in groups. In R. S. Tindale, & L. Heath(Eds.), Theory and research on small groups: Social psychological applications to social issues (Vol. 4, pp. 37–60). New York, NY, US: Plenum Press.

Murphy, S. K., Wayne, S. J., Liden, R. C., & Erdogan, B. (2003). Understanding social loafing: The role of justiceperceptions and exchange relationships. Human Relations, 56, 61–84.

Pearce, J. L., & Gregersen, H. B. (1991). Task interdependence and extrarole behavior: A test of the mediatingeffects of felt responsibility. Journal of Applied Psychology, 76, 838–844.

Pfeffer, J., & Langton, N. (1993). The effect of wage dispersion on satisfaction, productivity, and workingcollaboratively: Evidence from college and university faculty. Administrative Science Quarterly, 38, 382–407.

Saavedra, R., & Kwun, S. K. (1993). Peer evaluation in self-managing work groups. Journal of AppliedPsychology, 78, 450–462.

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)

746 R. C. LIDEN ET AL.

Scandura, T. A. (1999). Rethinking leader-member exchange: An organizational justice perspective. LeadershipQuarterly, 10, 25–40.

Scandura, T. A., Graen, G. B., & Novak, M. A. (1986). When managers decide not to decide autocratically: Aninvestigation of leader-member exchange and decision influence. Journal of Applied Psychology, 71, 579–584.

Schnake, M. E., & Dumler, M. P. (2003). Levels of measurement and analysis issues in organizational citizenshipbehaviour research. Journal of Occupational and Organizational Psychology, 76, 283–301.

Settoon, R. P., Bennett, N., & Liden, R. C. (1996). Social exchange in organizations: The differential effects ofperceived organizational support and leader-member exchange. Journal of Applied Psychology, 81, 219–227.

Shea, G. P., & Guzzo, R. A. (1987). Groups as human resources. Research in Personnel and Human ResourcesManagement, 5, 323–356.

Sias, P. M., & Jablin, F. M. (1995). Differential superior-subordinate relations, perceptions of fairness, andcoworker communication. Human Communication Research, 22, 5–38.

Slater, P. E. (1965). Role differentiation in small groups. In A. P. Hare, E. F. Borgatta, & R. F. Bales (Eds.), Smallgroups: Studies in social interaction (pp. 610–627) New York: Knopf.

Snijders, T. A. B., & Bosker, R. J. (1999). Multilevel analysis: An introduction to basic and advanced multilevelmodeling. London, UK: Sage.

Sparrowe, R. T., Soetjipto, B. W., & Kraimer, M. L. (in Press). Do leaders’ influence tactics relate to members’helping behavior? It depends on the quality of the relationship. Academy of Management Journal.

Stewart, G. L., & Barrick, M. R. (2000). Team structure and performance: Assessing the mediating role ofintrateam process and the moderating role of task type. Academy of Management Journal, 43, 135–148.

Stogdill, R. M. (1959). Individual behavior and group achievement. New York: Oxford University Press.Tekleab, A. G., Takeuchi, R., & Taylor, M. S. (2005). Extending the chain of relationships among organizationaljustice, social exchange, and employee reactions: The role of contract violations. Academy of ManagementJournal, 48, 146–157.

Thibaut, J., & Walker, L. (1975). Procedural justice: A psychological analysis. Hillsdale, NJ: Erlbaum.Tyler, T. R. (1994). Psychological models of the justice motive: Antecedents of distributive and procedural justice.Journal of Personality and Social Psychology, 67, 850–863.

Tyler, T. R., & Blader, S. L. (2000). Cooperation in groups: Procedural justice, social identity, and behavioralengagement. Philadelphia, PA: Psychology Press.

Uhl-Bien, M., Graen, G. B., & Scandura, T. A. (2000). Implications of leader-member exchange (LMX) forstrategic human resource management systems: Relationships as social capital for competitive advantage.Research in Personnel and Human Resources Management, 18, 137–185.

Van der Vegt, G., Emans, B. J. M., & Van De Vliert, E. (2001). Patterns of interdependence in work teams: A two-level investigation of the relations with job and team satisfaction. Personnel Psychology, 54, 51–69.

Wageman, R., & Baker, G. (1997). Incentives and cooperation: The joint effects of task and reward inter-dependence on group performance. Journal of Organizational Behavior, 18, 139–158.

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., & Anderson, S. E. (1991). Job satisfaction and organizational commitment as predictors oforganizational citizenship and in-role behaviors. Journal of Management, 17, 601–617.

Yukl, G. (2002). Leadership in organizations (5th ed.). Upper Saddle River, NJ: Prentice Hall.Zalesny, M. D., Salas, E., & Prince, C. (1995). Conceptual and measurement issues in coordination: Implicationsfor team behavior and performance. Research in Personnel and Human Resources Management, 13, 81–115.

Copyright # 2006 John Wiley & Sons, Ltd. J. Organiz. Behav. 27, 723–746 (2006)