Leader-Member Exchange Theory: Graen, Uhl-Bien, In-Groups, Trust, and LMX
Updated: 2 days ago
Author: Ukrainian Psychological Hub · Published: September 25, 2026 · Editorial Policy
Leader-member exchange theory (LMX theory) is a relationship-based theory of leadership. Its defining claim is that leadership inside a work unit is not adequately described by one average “leadership style” applied equally to everyone. A leader develops a separate working relationship with each member, and the quality, reciprocity, trust, respect, obligation, role negotiation, and resource exchange within those dyads can differ. The approach grew out of vertical dyad linkage research in the 1970s and was later developed into a broader theory of relationship-based leadership by George B. Graen, Mary Uhl-Bien, and other scholars (Dansereau, Graen, & Haga, 1975; Graen & Uhl-Bien, 1995).
The empirical record is substantial, but the construct needs careful handling. Meta-analyses associate higher-quality LMX with better task performance, more organizational citizenship behavior, less counterproductive work behavior, higher satisfaction and commitment, and lower turnover intentions. Yet these are mostly probabilistic associations, many studies rely on perceptual surveys, leader and member ratings do not always agree, and major critiques question whether commonly used scales fully capture the exchange process the theory claims to explain (Dulebohn et al., 2012; Martin et al., 2016; Gottfredson, Wright, & Heaphy, 2020; Yuan et al., 2023).
LMX is therefore best understood as a major relational leadership framework with a large evidence base and important unresolved construct-validity questions. It is not a personality test, not a diagnosis of a “good” or “bad” boss, not a synonym for trust, not a leadership style such as transformational or transactional leadership, and not a claim that favoritism is desirable. For the wider map of styles, traits, theories, power, followership, and effectiveness, see Leadership Psychology: Styles, Traits, Power, Followership, and Leadership Effectiveness.
What Is Leader-Member Exchange Theory?
Leader-member exchange theory focuses on the dyadic relationship between a leader and one particular member. “Dyadic” matters. If a manager supervises eight people, LMX does not begin by assuming there is one psychologically uniform leader–team relationship. It treats the unit as containing eight leader–member dyads, each with its own history of interaction, expectations, negotiated roles, access to information, discretion, support, and reciprocity. The theory asks how these relationships develop, why their quality differs, and what those differences predict.
That focus separates LMX from many classic leadership theories. Trait approaches ask what characteristics leaders tend to possess. Behavioral and style approaches ask what leaders do. Contingency theories ask when particular leader characteristics or behaviors fit a situation. LMX shifts the unit of analysis toward the relationship itself. This is why it belongs in a map of leadership theories rather than being treated as one more style label.
The word “exchange” also has a specific theoretical purpose. The relationship develops through repeated contributions and responses: information, effort, discretion, support, feedback, access, loyalty, opportunities, problem solving, and other resources can be exchanged over time. Later LMX scholarship has often interpreted these processes through social exchange theory. At the same time, critics have argued that many LMX measures assess evaluations of relationship quality rather than exchange behavior itself, creating a gap between theory and measurement (Bernerth et al., 2007; Sheer, 2015; Gottfredson et al., 2020).
The Core Idea: Leadership Happens in Relationships, Not Only in Leader Traits or Styles
LMX theory challenged an “average leadership style” assumption that was common in earlier research. Two employees reporting to the same manager can experience the manager differently because the dyads can differ in negotiated latitude, mutual influence, communication, reliability, and access to resources. Early longitudinal work by Dansereau, Graen, and Haga found that leaders used richer leadership exchanges with some members while relying more heavily on formal supervisory authority with others; role-negotiation latitude predicted later behavior in the dyad (Dansereau et al., 1975).
This does not mean a leader has no stable behavior or personality. It means those leader-level characteristics do not exhaust the phenomenon. A conscientious, extraverted, charismatic, directive, or supportive leader can still develop different relationship qualities with different members. Conversely, a high-quality dyad should not be inferred simply because the leader possesses a trait or displays a behavior often associated with effective leadership.
The distinction also protects against a common interpretive error: leader emergence and leadership effectiveness are separate questions. LMX primarily concerns relationships after leader and member roles exist. It does not by itself explain who will emerge as a leader in a leaderless group, and a favorable LMX score is not a complete measure of a leader’s effectiveness across team performance, ethics, strategic judgment, coordination, or organizational outcomes.
Where LMX Theory Came From: Vertical Dyad Linkage to Relationship-Based Leadership
The historical starting point was vertical dyad linkage (VDL), associated with George Graen and colleagues. Instead of averaging follower descriptions of the leader, VDL researchers examined vertical pairs: one superior and one subordinate. The 1975 longitudinal study by Fred Dansereau, George Graen, and William Haga treated leadership as an exchange relationship developing during role-making and showed that members under the same superior could occupy meaningfully different relational positions (Dansereau et al., 1975).
The terminology and theory changed over time. Graen and Uhl-Bien’s influential 1995 review explicitly emphasized that contemporary LMX theory was “far different” from early VDL work. They traced four evolutionary stages in the research program: discovery of differentiated dyads; study of LMX characteristics and outcomes; leadership-making or partnership building; and expansion beyond isolated dyads toward networks and larger systems of interdependent relationships (Graen & Uhl-Bien, 1995).
Those four evolutionary stages describe the history and scope of the theory. They should not be confused with the three phases often used to describe development of one leader–member relationship. That distinction is easy to lose in simplified summaries and is important for reading the original literature accurately.
In-Groups and Out-Groups: What the Early LMX Language Means
The in-group/out-group distinction comes from early differentiated-dyad research. In a comparatively high-quality exchange, a member may receive more negotiating latitude, information, support, discretion, and opportunities while also accepting broader responsibilities and contributing beyond the minimum formal role. A comparatively low-quality exchange may remain more contractual, role-bound, and dependent on formal authority. Early VDL research described this pattern; it did not establish that leaders should deliberately create permanent favored and disfavored camps (Dansereau et al., 1975; Graen & Uhl-Bien, 1995).
Modern LMX research is usually better represented as variation in relationship quality than as a binary sorting system. Real dyads can fall anywhere along a continuum, can change over time, and can be evaluated differently by leader and member. The labels “in-group” and “out-group” remain useful historically, but they can become misleading when they imply fixed social categories or when managers use them to rationalize favoritism.
Graen and Uhl-Bien’s leadership-making formulation moved toward developing effective partnerships broadly rather than reserving high-quality exchange for a small elite. This normative move matters because descriptive differentiation and prescriptive favoritism are different claims. LMX research documents that leaders differentiate among relationships; it does not make inequity, exclusion, or arbitrary preferential treatment psychologically harmless.
How LMX Relationships Develop
Relationship development in LMX is often summarized with a three-phase leadership-making sequence: stranger, acquaintance, and mature partnership. The phases describe increasing reciprocity, negotiated influence, and relational maturity. They are a theoretical developmental model rather than a guaranteed ladder every dyad follows (Graen & Uhl-Bien, 1995).
1. Stranger Phase
In the stranger phase, interaction is comparatively formal and role bound. Leader and member rely heavily on job descriptions, rules, contractual expectations, and one-way authority. Each side has limited evidence about the other’s reliability, judgment, motives, and willingness to reciprocate. The relationship can function adequately at this level; “low LMX” should not automatically be read as hostility or abuse.
2. Acquaintance Phase
In the acquaintance phase, one or both parties test whether the relationship can expand beyond strict contractual exchange. A leader may offer greater discretion, information, responsibility, or influence; a member may accept broader responsibility, provide extra effort, share expertise, or take initiative. Repeated interactions create evidence about dependability and reciprocity. This is a testing process, not simply a leader choosing favorites.
3. Mature Partnership
In a mature partnership, roles can become more negotiated and influence more reciprocal. Both parties have a history that supports expectations of mutual respect, reliability, support, and longer-horizon reciprocity. The word “partnership” should not erase hierarchy: the leader may still possess formal authority, control resources, evaluate performance, or make final decisions. LMX describes a relational quality within that structure rather than abolishing the structure.
A related stream describes LMX development through role taking, role making, and role routinization. Cropanzano, Dasborough, and Weiss used that sequence in a later theoretical account of affective events and LMX development, emphasizing that emotions and repeated events can shape the dyad as roles are sampled, negotiated, and stabilized (Cropanzano, Dasborough, & Weiss, 2017). These labels overlap conceptually with leadership-making, but they should not be treated as a single official staging system without attribution.
What Does High-Quality LMX Mean?
High-quality LMX usually refers to a relationship characterized by stronger mutual respect, trust, obligation, communication, support, and reciprocal influence than a low-quality exchange. In practice, it can involve greater discretion, richer information exchange, more frequent developmental feedback, more willingness to help, and more confidence that contributions will be reciprocated. These are tendencies, not a checklist that converts a workplace relationship into a clinical category.
Low-quality LMX usually refers to a more formal, limited, and contractual relationship. Communication may focus on required tasks; discretion and mutual influence may be smaller; exchanges may remain close to explicitly specified obligations. Low-quality LMX is not equivalent to abusive supervision, toxic leadership, incompetence, conflict, or a hostile relationship. A relationship can be distant and rule-bound without being abusive, and abusive behavior requires its own constructs and evidence.
Likewise, high LMX is not automatically ethical or effective. A close leader–member relationship can coexist with poor strategic decisions, favoritism, exclusion of others, or unethical collusion. Relationship quality is one construct among many. Leadership effectiveness requires outcomes and criteria beyond interpersonal closeness.
Trust and LMX: Closely Related, Not Identical
Trust is central to many accounts of high-quality LMX, but trust and LMX should not be collapsed into one variable. Trust is an expectation or willingness to accept vulnerability based on beliefs about another party; LMX is a broader relational construct involving exchange, role negotiation, mutual influence, support, obligations, and relationship quality. Brower, Schoorman, and Tan proposed an integrated model precisely because the two literatures overlap while remaining conceptually distinguishable (Brower, Schoorman, & Tan, 2000).
Meta-analytic evidence also suggests that trust can function as a mechanism connecting LMX to performance. Martin and colleagues found that trust in the leader, motivation, empowerment, and job satisfaction mediated associations between LMX and task and citizenship performance, with trust showing the largest mediating effect among those variables in their analysis (Martin et al., 2016). A mediator is not the same thing as the construct it helps explain.
How Is Leader-Member Exchange Measured?
LMX is usually measured with questionnaires completed by followers, leaders, or both. This sounds straightforward, but measurement is one of the field’s most persistent problems. Different scales embody different definitions of the construct, and a relationship is dyadic even when the research design asks only one person to rate it.
LMX-7
LMX-7 is the best-known short measure in the literature. It asks respondents to evaluate aspects of their working relationship with the leader, such as understanding, recognition, support, effectiveness of the relationship, and willingness to defend or help one another. Gerstner and Day’s 1997 meta-analysis concluded that, among the measures available in the literature they reviewed, LMX-7 had the soundest psychometric properties (Gerstner & Day, 1997).
That historical finding should not be converted into the claim that LMX-7 perfectly operationalizes the theory. Later work has raised content-validity and construct-validity concerns. Bernerth and colleagues argued that widely used LMX measures were not sufficiently representative of social exchange and developed a leader–member social exchange measure; Gottfredson and colleagues later argued that common LMX measures are misaligned with parts of LMX’s theoretical foundations (Bernerth et al., 2007; Gottfredson et al., 2020).
LMX-MDM
The multidimensional LMX measure, commonly called LMX-MDM, was developed by Robert Liden and John Maslyn. Their validation work supported four dimensions: affect, loyalty, contribution, and professional respect (Liden & Maslyn, 1998). This approach captures the possibility that a leader–member relationship has several distinguishable facets rather than one undifferentiated quality.
A multidimensional scale solves one problem while creating another interpretive obligation: researchers must specify whether they are theorizing about overall LMX or particular dimensions. A study using affect or professional respect is not necessarily testing the same proposition as a study treating LMX as a single global factor.
Leader Ratings, Member Ratings, and Agreement
A particularly important issue is that leaders and members can disagree about the same dyad. Gerstner and Day (1997) found only moderate correspondence between leader and member perceptions in the earlier literature. A later meta-analysis of 64 independent samples involving 10,884 dyads estimated overall leader–member agreement at about ρ = .37 (Sin, Nahrgang, & Morgeson, 2009), again indicating meaningful but far from perfect convergence. More recently, Yuan and colleagues showed meta-analytically that greater LMX agreement strengthened relationships between LMX and follower task performance and organizational citizenship behavior (Yuan et al., 2023).
For assessment, this means one person’s LMX score is a perspective on a relationship, not an objective photograph of the dyad. In organizational research or development, multi-source measurement is often more informative when the question genuinely concerns the relationship rather than one party’s experience of it.
What Predicts LMX Quality?
LMX quality is not created by a single personality variable. The most comprehensive antecedent meta-analysis by Dulebohn and colleagues synthesized 247 studies, 290 samples, 21 antecedents, and 16 consequences. Leader behaviors and perceptions, follower characteristics, interpersonal relationship variables, and contextual variables all showed relationships with LMX; leader variables explained the largest share of variance among the antecedent groups examined (Dulebohn et al., 2012).
This multicausal picture matters practically. Similarity, competence perceptions, affect, expectations, leader behavior, follower behavior, relationship history, work design, and context can all shape exchange quality. It is therefore scientifically weak to infer a single cause from a low LMX score—for example, “the employee is disengaged” or “the manager lacks empathy”—without independent evidence.
The dyadic logic also means causation can run through feedback loops. A leader may grant autonomy because a member has performed well; the member may then perform better because autonomy and trust increased; the leader may respond with still greater responsibility. Cross-sectional correlations compress that sequence into one snapshot and cannot by themselves determine the direction of causation.
What Does the Evidence Show About LMX Outcomes?
Task Performance
The strongest direct synthesis of LMX and performance is Martin and colleagues’ meta-analysis. Across 146 samples, LMX correlated positively with task performance at ρ = .30. The association remained positive for objective task performance across 20 samples at ρ = .24. These are meaningful average relationships, but they are not large enough to imply that relationship quality determines performance on its own (Martin et al., 2016).
The same meta-analysis found that observed relationships were weaker when LMX and performance came from different sources or methods, an important warning about common-method inflation. If the same manager rates both the relationship and the employee’s performance, shared rater perceptions can increase the observed association.
Citizenship and Counterproductive Work Behavior
Martin and colleagues reported a positive meta-analytic association between LMX and organizational citizenship performance across 97 samples (ρ = .34) and a negative association with counterproductive work behavior across 19 samples (ρ = −.24) (Martin et al., 2016). These patterns fit exchange-based accounts: employees who experience supportive, reciprocal relationships may be more willing to contribute beyond formal requirements. The evidence still describes average relationships, not a deterministic rule for any individual employee.
Satisfaction, Commitment, Role Clarity, and Turnover Intentions
Earlier meta-analytic work found significant associations between LMX and job performance, satisfaction with supervision, overall job satisfaction, organizational commitment, role conflict, role clarity, perceived member competence, and turnover intentions. Actual turnover was not significantly related in that 1997 synthesis, illustrating why intentions and realized behavior should not be treated as the same outcome (Gerstner & Day, 1997).
Dulebohn and colleagues’ larger later meta-analysis reinforced the view that LMX is embedded in a broad network of employee attitudes and behaviors, while also showing that antecedents and consequences span leader, follower, dyadic, and contextual domains (Dulebohn et al., 2012).
Culture
LMX findings are not culturally invariant in every respect. Rockstuhl and colleagues meta-analyzed 282 independent samples involving 68,587 participants across 23 countries. National culture moderated LMX relationships with organizational citizenship behavior, justice perceptions, job satisfaction, turnover intentions, and trust in the leader, while relationships with task performance, organizational commitment, and transformational leadership were not significantly moderated in the same way (Rockstuhl et al., 2012).
This is a useful check against treating one organizational norm as universal. Hierarchy, individualism, collectivism, and expectations about leader–follower relationships can change how the same relational pattern is interpreted and how strongly it connects to particular outcomes.
Correlation, Longitudinal Evidence, Mediation, and Causality
The evidence base supports prediction more strongly than universal causal claims. Much LMX research is observational. A cross-sectional association between LMX and performance does not establish whether relationship quality improved performance, strong performance improved the relationship, a third variable influenced both, or several processes operated together.
Martin and colleagues went further than a simple correlation summary. Their synthesis found mediation through trust, motivation, empowerment, and job satisfaction, and their temporal analyses supported a path from LMX to task performance rather than a reverse or reciprocal pattern in the available longitudinal evidence (Martin et al., 2016). That strengthens the case for temporal ordering in that outcome domain, but it does not turn the entire LMX literature into experimental causal evidence.
Gottfredson and colleagues highlighted endogeneity as a major challenge: unmeasured causes, reciprocal processes, and construct overlap can make estimated LMX effects look more causal or distinct than the design warrants (Gottfredson et al., 2020). Strong practice recommendations should therefore be based on mechanisms and evidence, not on the slogan that “high LMX causes performance.”
LMX Differentiation: Why One Good Dyad Is Not the Whole Team
LMX differentiation is the degree to which a leader has relationships of different quality with different members of the same group. It is a team-level or within-group pattern, not the same variable as one employee’s LMX score. Henderson and colleagues reviewed differentiation as a multilevel phenomenon with antecedents and consequences at individual, group, and organizational levels (Henderson et al., 2009).
Differentiation can create both functional and social consequences. Some role differentiation can reflect legitimate differences in expertise, responsibility, interdependence, tenure, or task demands. But visible differences in access, information, support, or influence can also trigger social comparison, justice concerns, subgroup formation, and weakened cohesion.
Yu, Matta, and Cornfield’s meta-analysis covered 4,114 workgroups and 21,745 individuals. LMX differentiation was negatively related to collective emergent states and group processes. Its relationship with group performance was more complex: the authors found a positive direct relationship alongside negative indirect effects through poorer group processes, showing a tradeoff rather than a simple “differentiation is good” or “differentiation is bad” conclusion (Yu, Matta, & Cornfield, 2018).
For managers, the implication is not that every relationship must be identical. Equality of treatment, equity of resources, role requirements, and individualized support are different principles. The psychologically important questions include whether differences are explainable, fair, task-relevant, revisable, and compatible with genuine access to development.
LMX and Transformational Leadership: Related but Not the Same Construct
Transformational leadership and LMX are frequently correlated because transformational behaviors can help create high-quality relationships and because high-quality LMX includes perceptions—such as support, respect, trust, and development—that overlap with how transformational leadership is experienced. But the theories point to different focal units. Transformational leadership is primarily a pattern of leader behavior and influence; LMX is primarily a dyadic relationship construct.
A 2026 critical umbrella review by Mortaza Zare and Carol Flinchbaugh identified 36 meta-analytic studies and reported corrected meta-analytic correlations between transformational leadership and LMX ranging from .69 to .77. The authors concluded that the constructs overlap substantially but also show different patterns of relationships with outcomes, so the evidence does not justify simply treating them as redundant (Zare & Flinchbaugh, 2026).
This overlap is exactly why construct discipline matters. A study showing that transformational leadership predicts LMX does not mean the two measures are interchangeable. For the behavioral theory, historical origins, Four I’s, evidence, and criticisms, see Transformational Leadership: Burns, Bass, the Four I's, Evidence, and Criticism.
LMX Is a Leadership Theory, Not a Leadership Style
A leadership style is a recurring pattern of leader behavior, such as directive, participative, transformational, transactional, or servant-oriented behavior. LMX is a theory about the quality and development of a relationship between a leader and a member. A leader can use the same broad style with two employees yet have different LMX quality with them, or use different behaviors within the same high-quality dyad as tasks change.
This distinction is especially important in online leadership content, where “LMX leadership style” is sometimes used casually. That phrase can be convenient, but it obscures the unit of analysis. LMX does not prescribe a fixed behavioral style analogous to a style inventory.
LMX Is Not a Personality Trait
LMX scores do not identify a stable personality type in the leader or follower. Personality can be an antecedent or correlate of relationship development, but the construct itself is relational. A person can have high-quality LMX with one supervisor and lower-quality LMX with another; a manager can have different exchange relationships with members of the same team.
For the same reason, LMX should not be used to infer narcissism, psychopathy, attachment style, emotional intelligence, or any clinical diagnosis. Those are separate constructs with different measures and evidentiary requirements.
LMX Is Not Leader Emergence or a Complete Measure of Leadership Effectiveness
Leader emergence asks who becomes influential or is perceived as leader-like, often before formal leadership is assigned. LMX assumes an identifiable leader–member relationship and studies the dyad. A person can emerge as a leader because of status, expertise, dominance, prestige, social identity, or network position without LMX being the mechanism that explains emergence.
Leadership effectiveness is also broader than LMX. High-quality exchanges are associated with useful outcomes, but effectiveness can include team performance, adaptation, decision quality, safety, ethics, innovation, retention, coordination, and long-term organizational results. LMX can contribute to effectiveness without defining it.
LMX, Followership, and Why People Follow Leaders
LMX makes the follower an active party in leadership rather than a passive recipient of leader behavior. Members accept, reject, reinterpret, and reciprocate relationship offers; their performance, expertise, initiative, communication, and reliability can change the dyad. This relational emphasis overlaps with followership research but does not make LMX a general theory of why people follow.
People follow leaders for many reasons beyond a particular dyadic exchange: formal authority, identification, shared identity, expertise, perceived legitimacy, incentives, norms, charisma, dependence, and coordination needs can all matter. For that broader question, see Why Do People Follow Leaders? Charisma, Authority, Social Identity, and Followership.
LMX, Management, and Supervision
The historical roots of LMX make the difference between leadership and supervision unusually visible. Dansereau and colleagues reported that superiors used both leadership exchanges and authority-based supervision within the same units (Dansereau et al., 1975). This is an important conceptual clue: formal supervision creates a role relationship, while leadership exchange concerns the quality and reciprocal influence developing inside that relationship.
Management includes planning, staffing, budgeting, coordination, control, systems, and formal accountability. Supervision includes directing and evaluating work. Leadership includes influence around direction, meaning, coordination, motivation, and collective action. One person can perform all three functions, but the constructs remain analytically distinct. See Leadership vs Management: Psychology, Influence, Authority, and Decision-Making.
LMX, Power, Authority, Status, Dominance, and Prestige
LMX is also distinct from power. Power concerns asymmetric capacity to affect outcomes, control resources, shape behavior, or influence another person. Authority is legitimized or role-based right to direct or decide. Status is socially conferred rank or respect. Dominance and prestige are routes to hierarchical influence often studied in social hierarchy research. LMX concerns the quality and structure of a dyadic working relationship.
These constructs can interact. Formal authority can initiate a supervisor–employee dyad. Resource control can shape dependence. Status can affect whose advice is taken seriously. Prestige can facilitate voluntary deference. Dominance can affect compliance. None of these variables is identical to high-quality LMX, and a close dyad does not erase unequal formal power. For the cross-cluster distinction, see Power and Leadership: Influence, Status, Authority, and Followership.
Major Criticisms and Limitations of LMX Theory
1. The Construct Has Changed Over Time
A recurring criticism is definitional drift. Schriesheim, Castro, and Cogliser’s comprehensive review showed that definitions, measures, and analytic practices changed substantially across the literature, sometimes without sufficient theoretical explanation (Schriesheim, Castro, & Cogliser, 1999). When a field uses the same label for somewhat different constructs, apparent accumulation of evidence can overstate conceptual continuity.
2. Measurement Does Not Always Match the Theory
If LMX is fundamentally an exchange process, a scale that mainly asks whether the relationship feels effective, supportive, or respectful may capture relationship quality without directly measuring exchange behavior. This is the central concern behind critiques from Bernerth and colleagues, Sheer, and Gottfredson and colleagues (Bernerth et al., 2007; Sheer, 2015; Gottfredson et al., 2020).
3. Dyadic Theory Is Often Tested With One-Sided Data
A relationship belongs to two people, yet many studies use only follower-reported or leader-reported LMX. Moderate leader–member agreement means those perspectives cannot be assumed to be interchangeable. Yuan and colleagues’ 2023 meta-analysis further showed that agreement itself has substantive implications for LMX–outcome relationships (Yuan et al., 2023).
4. Common-Method Bias Can Inflate Relationships
When the same respondent rates LMX and an outcome such as satisfaction, commitment, or performance, shared response tendencies can increase correlations. Martin and colleagues found weaker LMX–performance associations when measures came from different sources or methods (Martin et al., 2016). This does not erase the relationship; it changes how confidently its magnitude should be interpreted.
5. Endogeneity and Reverse Causation Are Real Problems
Leaders may form better relationships with stronger performers, and better relationships may also improve performance. Similarity, expectations, liking, job design, organizational support, and unmeasured ability can influence both relationship quality and outcomes. Gottfredson and colleagues emphasized endogeneity as a central challenge for causal interpretation (Gottfredson et al., 2020).
6. In-Group/Out-Group Language Can Legitimize Favoritism if Misused
The early descriptive distinction is sometimes converted into managerial advice to create an inner circle. That move is not warranted. Differentiation can affect justice perceptions, cohesion, and group processes, and leadership-making explicitly shifted toward broader opportunities for high-quality relationships. Managers need defensible role differences and fair access to information and development rather than personality-based favoritism.
7. LMX Overlaps With Other Positive Leadership Constructs
High-quality LMX is strongly related to transformational leadership, trust, support, satisfaction with the leader, and other favorable relational perceptions. The 2026 umbrella review of transformational leadership and LMX found substantial meta-analytic overlap while also identifying distinct outcome patterns (Zare & Flinchbaugh, 2026). This is a construct-validity issue, not a reason to discard all LMX findings.
What Can Leaders and Organizations Do With LMX Research?
The most defensible practical use of LMX research is to treat relationship quality as something leaders and members jointly build, observe, and improve—not as a label for insiders and outsiders. The following practices follow from the evidence without turning LMX into a corporate personality quiz.
Make Access to Relationship-Building Opportunities Broad
Leaders can create recurring one-to-one contact, invite questions, explain decisions, and provide developmental assignments through transparent criteria. High-quality relationships cannot be forced, but access to the interactions from which they can develop can be distributed more fairly.
Separate Role Requirements From Personal Closeness
A manager can have warm rapport with one employee and still need objective standards for assignments, information, rewards, and performance evaluation. Friendship, liking, and LMX are related but not interchangeable. High-quality working exchange should not become a private channel for arbitrary advantage.
Use Negotiated Autonomy, Not Undefined Favor
Early LMX research emphasized negotiating latitude. In practical terms, leaders can make discretion explicit: which decisions may the member make, what information is needed, what risks require escalation, what outcomes define success, and how will authority expand as evidence of capability accumulates? This converts vague trust into observable working arrangements.
Check Whether Both Sides See the Relationship Similarly
Because leader–member agreement is imperfect, a leader should not assume that “we have a great relationship” is shared by the member. Regular discussion of expectations, support, autonomy, communication, workload, and obstacles can reveal mismatched assumptions before they harden into conflict. This practical implication is consistent with the agreement literature without pretending that an informal conversation is a psychometric assessment (Yuan et al., 2023).
Monitor Differentiation at the Team Level
Managers should ask not only “How good is my relationship with this person?” but also “What pattern am I creating across the team?” Differences in responsibility can be legitimate. Persistent differences in access to information, sponsorship, voice, or tolerance for mistakes can produce fairness problems and damage group processes even when one favored dyad performs well (Yu et al., 2018).
Do Not Use LMX Scores as Diagnoses
An LMX questionnaire can support research, reflection, or structured organizational assessment when used appropriately. It does not diagnose personality, mental disorder, abuse, toxicity, competence, or moral character. A low score should trigger questions about the relationship and context, not a label attached to a person.
A Worked Example: Same Manager, Three Different Dyads
Imagine a product manager with three direct reports. The first has worked with the manager for four years, regularly receives early strategic information, is trusted to make customer decisions without approval, and challenges the manager openly. The second is new, receives clear task instructions and scheduled feedback, and is gradually being given more discretion. The third has a long tenure but communicates only when required after repeated misunderstandings about priorities.
LMX theory would not require calling the first employee an “in-group personality” and the third an “out-group personality.” It would analyze three relationships at different levels and histories of exchange. The first dyad may approximate mature partnership. The second may be developing through testing and role negotiation. The third may be a low-quality or damaged exchange that still requires fair supervision and clear performance management.
A strong organizational response would examine both dyadic and team-level processes. Is the first employee’s autonomy based on demonstrated capability and transparent role demands, or personal favoritism? Does the second employee have genuine access to development? Can the third dyad repair expectations through specific feedback and negotiated responsibilities? Do other team members interpret the differences as fair? LMX becomes useful when it produces these questions rather than simplistic labels.
Frequently Asked Questions
Who developed leader-member exchange theory?
LMX developed through a research program rather than one isolated paper. George B. Graen and colleagues were central to early vertical dyad linkage research. Fred Dansereau, George Graen, and William Haga published a landmark longitudinal VDL study in 1975. George Graen and Mary Uhl-Bien’s 1995 review became a major statement of mature, relationship-based LMX theory (Dansereau et al., 1975; Graen & Uhl-Bien, 1995).
What is the simplest definition of LMX theory?
LMX theory proposes that leaders develop distinct one-to-one relationships with different members, and the quality and development of those dyadic relationships are associated with important attitudes and work outcomes.
What are the in-group and out-group in LMX theory?
They are historical labels for relatively high-quality and low-quality leader–member exchanges. High-quality dyads tend to involve greater reciprocal influence, discretion, support, information, and negotiated responsibility; low-quality dyads tend to remain more formal and contractual. Contemporary use should avoid treating these labels as fixed social castes or as a justification for favoritism.
What are the three stages of LMX?
In the leadership-making model, relationship development is commonly described as stranger, acquaintance, and mature partnership. A related role-development vocabulary uses role taking, role making, and role routinization. These are theoretical models of relationship development, not validated diagnostic stages that every dyad must pass through at a fixed rate.
Are the four stages of LMX the same as the three relationship stages?
No. Graen and Uhl-Bien’s four stages describe the historical evolution of LMX theorizing from differentiated dyads through broader relationship networks. The three phases—stranger, acquaintance, and mature partnership—describe development within a leader–member relationship. Mixing these two frameworks is a common source of confusion.
Is LMX a leadership style?
No. LMX is a relationship-based leadership theory or framework. A style describes recurring leader behavior; LMX describes the quality and development of a particular leader–member dyad.
Is high LMX the same as trust?
No. Trust is closely related to high-quality LMX and may mediate some LMX–performance relationships, but LMX includes a wider pattern of exchange, support, influence, obligations, role negotiation, and relationship quality (Brower et al., 2000; Martin et al., 2016).
Does high LMX improve performance?
Higher LMX is positively associated with task performance on average. Martin and colleagues reported a meta-analytic correlation of ρ = .30 across 146 samples and ρ = .24 for objective task performance across 20 samples (Martin et al., 2016). The relationship is meaningful but not deterministic, and causal interpretation requires more than a cross-sectional correlation.
Is LMX-7 a validated test?
LMX-7 is a widely studied research measure with a long psychometric literature. Gerstner and Day found it had the strongest psychometric properties among LMX instruments in their 1997 review, but later scholars raised content-validity and construct-validity concerns about what LMX-7 actually measures (Gerstner & Day, 1997; Bernerth et al., 2007; Gottfredson et al., 2020). It should be used as a research or organizational assessment tool with attention to purpose, scoring, source, and limitations—not as a diagnostic test.
Can leaders have high LMX with everyone?
Leadership-making theory moved toward offering high-quality partnership opportunities broadly, but equal relationship quality cannot be guaranteed. Roles, history, interdependence, performance, preferences, and context differ. The defensible goal is broad opportunity, fair processes, transparent role differences, and avoidance of arbitrary exclusion—not forced identical closeness.
Is LMX differentiation always harmful?
No. Team-level evidence shows tradeoffs. Yu and colleagues found differentiation was negatively associated with group emergent states and processes, while its association with group performance included a positive direct pathway and negative indirect pathways through group processes (Yu et al., 2018). The consequences depend on why differentiation exists, how it is interpreted, and what it does to coordination and justice.
How is LMX different from management?
Management includes formal planning, coordination, control, staffing, and accountability. LMX describes relationship quality within leader–member dyads. A formal manager can have high- or low-quality LMX with different employees, and supervisory authority can operate even when LMX remains low.
What is the biggest scientific criticism of LMX theory?
There is no single criticism, but the central cluster concerns construct definition and measurement: what exactly counts as “exchange,” whether popular scales capture that construct, whether one-sided ratings adequately represent a dyad, how much common-method bias inflates observed relationships, and whether causal claims survive endogeneity and construct overlap (Schriesheim et al., 1999; Sheer, 2015; Gottfredson et al., 2020).
The Bottom Line
Leader-member exchange theory changed leadership research by making a simple but consequential move: it treated leadership as a set of relationships rather than only as a property of the leader. That move explains why two people reporting to the same manager can experience different levels of trust, discretion, support, influence, and role negotiation—and why those differences are associated with performance, citizenship, satisfaction, commitment, and turnover intentions.
The mature scientific reading of LMX is equally important. High-quality LMX is associated with desirable outcomes, but it is not a universal causal switch. The field contains measurement ambiguity, leader–member disagreement, common-method problems, endogeneity, construct overlap, cultural moderation, and team-level costs of differentiation. LMX is most useful when it directs attention to how working relationships are built, reciprocated, compared, and embedded in a wider team—without turning “in-group” and “out-group” into labels for people.