Q Academy of Management Review

2017, Vol. 42, No. 2, 233–258.

[Link]

AFFECTIVE EVENTS AND THE DEVELOPMENT OF

LEADER-MEMBER EXCHANGE
RUSSELL CROPANZANO
University of Colorado at Boulder

MARIE T. DASBOROUGH
University of Miami

HOWARD M. WEISS
Georgia Institute of Technology

We propose that consideration of affective events theory can enrich our understanding
of leader-member exchange (LMX) development. Drawing from previous research, we
argue that high-quality LMX relationships progress through three stages: role taking,
role making, and role routinization. Affective events theory indicates that emotions are
relevant at each of these three stages, although their influence is manifested in different
ways and at different levels of analysis. During the initial role-taking stage, leaders’
affective expressions serve as affective events influencing member emotions through
the processes of emotional contagion and affective empathy, which determine the
progress of further relationship development. Next, during the role-making stage,
leaders and members are both sources of affective events, and they may gradually
become affectively entrained such that their affective states tend to fluctuate in a com-
mon rhythm. This pattern of dyadic-level affect helps to build high-quality LMX re-
lationships over time. Finally, during the role routinization stage, an LMX relationship
has been formed but, we argue, could subsequently change based on member emotional
responses to the distribution of LMX relationships within a workgroup (LMX
differentiation).

Over the past few decades, leader-member ex- it becomes incumbent on scholars to clearly
change (LMX) theory has emerged as among the articulate how high-quality LMX relationships
most successful approaches for studying orga- can be nurtured and developed. While promising
nizational leadership (e.g., Dinh et al., 2014; work exists (see especially Dienesch & Liden,
Schriesheim, Castro, & Cogliser, 1999). Much of 1986, and Graen & Scandura, 1987), prevailing
the success of the theory is attributed to the research is somewhat limited, and there is much
documented benefits found to result from positive we do not yet know.
relationships between leaders and followers We argue that currently available research on
(Graen & Uhl-Bien, 1995). These include such im- LMX development, with a few exceptions, has
portant outcomes as more favorable employee job yet to pay sufficient attention to the role of affect
attitudes, less conflict, better performance, more in relationship formation and maintenance.
frequent organizational citizenship behaviors Rather, in work on LMX formation, scholars have
(OCBs), higher creativity, and lower turnover (for tended to explore such variables as delegation
quantitative reviews see Dulebohn, Bommer, (e.g., Schriesheim, Neider, & Scandura, 1998;
Liden, Brouer, & Ferris, 2012; Gerstner & Day, Schyns, Paul, Mohr, & Blank, 2005) and leader
1997; and Ilies, Nahrgang, & Morgeson, 2007; for personality (Nahrgang, Morgeson, & Ilies, 2009).
narrative reviews see Erdogan & Bauer, 2014, and These are important factors, to be sure, but as
Liden, Sparrowe, & Wayne, 1997). Given the ad- Ferris et al. state, “There is little about a re-
vantages of strong leader-member relationships, lationship that can be understood without un-
derstanding its affective tone, and the emotions
and feelings the partners experience in their
The first two authors contributed equally but are presented
in alphabetical order. We thank the special issue editor and
association with each other” (2009: 1384). Con-
four anonymous reviewers for their helpful suggestions and sistent with Ferris and colleagues’ view, affec-
insightful ideas. tive tone is the LMX factor on which leaders and
233
Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright
holder’s express written permission. Users may print, download, or email articles for individual use only.
234 Academy of Management Review April

followers most agree (Sin, Nahrgang, & Morgeson, that leaders and members will not always agree
2009). Moreover, leader-member similarity in pos- (see Sin et al., 2009), in our proposed model we
itive affectivity predicts LMX quality (Bauer & emphasize both perspectives and shared emo-
Green, 1996), and affect may even have stronger tional states. Of course, focusing on the dyadic
effects on LMX than follower job performance relationship is not to say that the each party will
(Liden, Wayne, & Stilwell, 1993; for related evi- be equally active at every developmental stage.
dence see Newcombe & Ashkanasy, 2002, and During the early role-taking stage, when the
Sears & Hackett, 2011). member lacks formal power, the leader is often
To address this issue, we propose a new theo- likely to take initiative in relationship develop-
retical model, integrating affective events theory ment (e.g., Dienesch & Liden, 1986; Graen & Uhl-
(AET; Weiss, 2002; Weiss & Cropanzano, 1996) with Bien, 1995). Here the leader offers a member the
existing work on LMX relationship development. opportunity for a higher-quality relationship
Summarized in Figure 1, our model proceeds (Nahrgang et al., 2009). In this initial stage affec-
in three stages, previously identified by LMX tive expressions by the leader serve as anchoring
researchers (Bauer & Green, 1996; Graen & affective events (Ballinger & Rockmann, 2010),
Scandura, 1987; Nahrgang et al., 2009; Sin et al., influencing follower emotions via emotional
2009): role taking, role making, and role routini- contagion and laying the foundation for future
zation. Underappreciated, as we shall argue, is LMX development.
the critical role played by the affective experi- After this opening encounter, the leader and the
ences of leaders and members in the develop- member enter the role-making stage, during
ment of LMX relationships. AET, with its strong which the two individuals undergo a series of
focus on events, affective changes, discrete transactions. These are referred to as “role epi-
emotions, and so forth, is a strong framework for sodes” (e.g., by Sin et al., 2009: 1049) or “interacts”
articulating the role of affect in these three (e.g., by Bauer & Green, 1996: 1544). As the role-
stages (cf. Weiss & Beal, 2005). For each stage we making stage unfolds, the affective experience of
draw on AET to propose important affective pro- the leader and member may become emotionally
cesses that can shape LMX quality: contagion in entrained (e.g., Totterdell, 2000; Totterdell, Kellet,
Stage 1, entrainment in Stage 2, and elicitation of Teuchmann, & Briner, 1998)—that is, in high-
moral emotions in Stage 3 (see Figure 1). We also quality LMX relationships, the leader’s and
discuss specific emotions—those suggested or member’s feeling states begin to synchronize (cf.
implied by AET and LMX theorists—that are in- Bauer & Green, 1996). During this role-making
volved in each of these processes, although we stage, the member and the leader are both active
remind the reader that our list of emotions is not participants in the relationship development
intended to be exhaustive. process.
Consistent with prior research, we treat LMX as Finally, in the role routinization stage, the LMX
a dyadic relationship (Graen & Scandura, 1987; relationship is often stable. Nevertheless, we ar-
Graen & Uhl-Bien, 1995). While we acknowledge gue that subsequent changes in LMX quality are

FIGURE 1
Leader-Member AET

Role taking Role making Role routinization

LMX development: LMX development: LMX development:

Leader initiates Series of dyadic Stable relationship
possibility of LMX affective events disrupted by LMX
relationship shapes LMX quality differentiation

Key affective Key affective Key affective

process: process: process:
Leader emotional Entrainment of Relative LMX elicits
expressions are leader/member emotions
affective events affect

Level: Individual Level: Dyadic Level: Group

2017 Cropanzano, Dasborough, and Weiss 235

possible and their likelihood may have been Before the formulation of VDL/LMX, most existing
previously understated. We suggest that there theories emphasized a leader’s overall treatment
are circumstances in which the members’ re- of the group as a whole (Dansereau et al., 1995;
sponses to their LMX relationship, relative to Vecchio, Griffeth, & Hom, 1986). This average
others in the workgroup, will change the nature leadership style (ALS) model (Schriesheim et al.,
of their current relationship. According to studies 1998: 298) implicitly assumed that the leader’s
of LMX differentiation, leaders sometimes sort style was roughly stable among all members
members into ingroups—members with higher- (Dienesch & Liden, 1986). In contrast, LMX theory
quality LMX relationships—and outgroups— maintained that leaders treat members differ-
members with lower-quality LMX relationships ently, depending on whether they are in the
(cf. Henderson, Liden, Glibkowski, & Chaudhry, leaders’ ingroup or outgroup (Graen & Scandura,
2009; Schriesheim, Castro, & Yammarino, 2000; 1987; Graen & Uhl-Bien, 1995). That is, LMX theory
Tse, Ashkanasy, & Dasborough, 2012; Tse, is concerned with the quality of relationships
Dasborough, & Ashkanasy, 2008). If these di- within leader-member dyads (Erdogan & Bauer,
visions are stark, then individuals who benefit are 2014). Each dyadic relationship can possess its
likely to experience the other-praising emotion of own quality, depending on the nature of ex-
gratitude (Emmons, 2007; Emmons & Crumpler, changes between the participants. Leader re-
2000; Fredrickson, 2004; Haidt, 2003). However, lationships with some members will rate higher
those who feel they are unfairly disadvantaged in terms of trust, respect, and liking. Relationships
could experience other-condemning moral emo- with other members will be lower on these attri-
tions, such as anger, disgust, or contempt (cf. butes (Liden et al., 1997).
Hooper & Martin, 2008; Scandura, 1999), toward the
leader. The other-praising positive emotion of
The Role of Affect and LMX Development
gratitude tends to further strengthen the LMX re-
lationship, whereas the other-condemning nega- Although research to date remains limited,
tive emotions could further weaken the LMX there is tantalizing evidence that LMX relation-
relationship. At this final stage, in line with LMX ships are likely to develop out of affectively tinged
differentiation research, we focus on the members’ interactions between leaders and members.
perceptions and emotions in response to their From our present perspective, this work suggests
relationships. that these interactions serve as affective events
There are certainly many aspects of emotion that (e.g., Weiss, 2002), which will help shape the
could potentially impact interpersonal relation- quality of future leader-member exchanges. For
ships. In the interest of theoretical parsimony, example, Sears and Hackett (2011) examined
we emphasize three: contagion, entrainment, and both role clarity and affect as predictors of LMX.
moral emotions. We pay special attention to these They found that member perceptions of LMX
three emotion concepts because a review of the were largely explained by affect toward leaders,
AET literature and leadership literature provided more so than by perceptions of role clarity.
strong conceptual rationales and solid empirical Likewise, in a longitudinal study Liden and
evidence, both of which connect them to the phe- colleagues (1993) investigated LMX develop-
nomena in question—LMX relationship develop- ment over a six-month period. They found that,
ment over time. In the following pages we elaborate over time, performance became less important
on these concepts, paying special attention to the and affect took over as the main predictor of
literature on LMX development and also leader/ leader LMX ratings. Consequently, it appears
follower affect. Before doing so, we offer brief re- that affect plays an important role in the devel-
views of LMX and AET so as to provide a conceptual opment of LMX, although its role is not yet well
grounding for our proposed framework. understood. To begin providing a more complete
theoretical grounding, we now turn our attention
to AET.
THE ROLE OF AFFECTIVE EVENTS IN LMX
Originally named the vertical dyad linkage
AET
(VDL) model (Dansereau, 1995; Dansereau, Graen,
& Haga, 1975), the LMX approach was a consider- Originally formulated as a theory of workplace
able departure from prior leadership theories. emotion (e.g., Weiss, 2002; Weiss & Cropanzano,
236 Academy of Management Review April

1996), AET has also been a useful conceptual tool during the early stages of LMX relationship de-
for understanding leadership (e.g., Dasborough, velopment (Ballinger & Rockmann, 2010).
2006; Gaddis, Connelly, & Mumford, 2004; Johnson, AET also recognizes that affective states can
2008, 2009; Pirola-Merlo, Härtel, Mann, & Hirst, show regular patterns. These may be the result of
2002; Volmer, 2012). At its core AET is a model regular patterns of work events (we tend to eat
describing within-person changes in affective lunch at the same time every day, for example) or
states, their root in events of both a stochastic and underlying biological rhythms. Yet a key objec-
regular nature, and their influences on concurrent tive of AET was to encourage researchers to look
changes in performance-related behaviors (Weiss at the within-person changes in affective states
& Beal, 2005). At the time AET was originally pre- and concurrent behaviors influenced by those
sented, interest in emotional experiences at work changing states, and to think about the causes of
was growing alongside an expanding interest in those states as arising from the events people
emotions within psychology generally. AET was experience (Weiss & Cropanzano, 1996). Weiss
offered as a framework for organizing existing lit- and Cropanzano (1996) described the episodic
erature on work related to emotional experiences, structure of affective states, borrowing from
as well as for guiding future research (Weiss & earlier work by Frijda (1993). Beal, Weiss, Barros,
Beal, 2005). A few key conceptual ideas form the and MacDermid (2005) introduced the concept
core of the AET framework. Because these concepts of episodic performance as a way of better un-
have relevance to the problem of studying emo- derstanding how changes in affective states
tional experiences within leader-member dyads, influence concurrent performance. Hence, we
they are worth discussing. are interested in not only single affective events
To begin, AET makes clear that emotions are but, rather, numerous affective events over time.
states, and as states they vary within persons over Once again, the tendency of feeling states to
time (Weiss, 2002)—for example, “I’m angry now, change together can inform research on LMX. As
we discuss in this article, the affective states of
but I won’t be later,” or “This morning I was in
leaders and members can fluctuate together
a terrible mood.” This point is obvious, of course,
(e.g., Johnson, 2008, 2009; Johnson & Connelly,
but from it AET develops certain other points that
2014), becoming mutually entrained (cf. Gooty,
were less obvious at the time AET was first pre-
Connelly, Griffith, & Gupta, 2010; Mansfield,
sented. Of particular importance for this article is
Hood, & Henderson, 1989). Emotional entrain-
the idea that events are the proximal causes of
ment refers to the process where two interacting
emotions. Things happen to people at work, and
oscillating systems assume the same emotional
often people react emotionally as a result (Weiss
pattern over time (Chartrand & Bargh, 1999;
& Cropanzano, 1996). Most organizational theories Lakin, Jefferies, Cheng, & Chartrand, 2003). This
look at the relationship between features of the pattern relies on emotional contagion (when in-
work environment (such as reward structure or dividuals catch each other’s emotions), and we
role clarity) and work outcomes. AET argues that argue that entrainment provides the theoretical
using stable environmental features of work set- explanation for the formation of the empathic
tings to explain and predict changeable affective bond between a leader and his or her followers
states represents a mismatch of construct type. (see Kellett, Humphrey, & Sleeth, 2002, 2006).1 As
Changing states need causal variables that are we shall see, emotional entrainment clearly has
themselves changing. To be sure, features may implications for developing high-quality LMX
make certain types of events more frequent, but in relationships.
the final analysis events are the proximal causes A third important feature of the AET framework
of affect states. As such, the core of AET is the is the difference between affect and other evalu-
relationship between workplace events and ative judgments, such as attitudes and ratings of
changes in affective states (Weiss & Cropanzano, leader effectiveness (Weiss, 2002). For instance,
1996). Throughout this article we attend to this prior to the presentation of AET, job satisfaction
issue, arguing that doing so will inform our had been generally defined as both an attitude
knowledge of LMX. Toward this end, we pay spe- and an emotional response one has toward one’s
cial attention to leader emotional expressions as
affect-inducing events for members (see also 1
We thank special issue editor Ronald Humphrey for
Humphrey, 2002), which particularly have impact pointing this out to us.
2017 Cropanzano, Dasborough, and Weiss 237

work. Most researchers saw no conflict between include all such discrete emotions in a single
the satisfaction as attitude and satisfaction as theoretical model. Nor would it be appropriate to
emotion definitions, believing attitude and emo- do so, since many discrete emotions have not re-
tion to be equivalent constructs (see Fisher, 2000, ceived adequate research attention in workplace
for an investigation). AET, on the other hand, settings. Therefore, we emphasize only those emo-
views an emotion as an affective state one expe- tions that are theoretically relevant and discuss
riences and satisfaction as the evaluative judg- them when their importance is most salient.
ment (attitude) one makes (Weiss & Cropanzano, Emotions from AET and LMX theories. AET
1996). The theory recognizes that emotional ex- places emphasis on the basic emotions (Ekman,
periences can influence attitudes but argues, 1992a,b), which have a characteristic physiology,
consistent with current attitude theory, that they particular set of antecedents, and recognizable
are separate constructs. An essential implication facial expressions (Izard, 1977, 2007; Plutchik,
of this observation is the differentiation between 1994). Basic emotions are also useful because
affect-driven behaviors—behaviors that are di- they show a degree of cross-cultural consistency
rectly influenced by immediate emotional states (e.g., Shaver, Wu, & Schwartz, 1992). We follow
and show significant within-person variability AET by listing six basic emotions that are recog-
as a result—and judgment-driven behaviors— nized by most researchers: joy (or happiness),
behaviors directly influenced by work attitudes sadness, fear, anger, love, and surprise (Shaver
(Weiss & Cropanzano, 1996). This is an important et al., 1987).
assumption for understanding LMX because it While this list is based on the emotions presented
suggests that a member’s judgment of the leader in AET (Weiss & Cropanzano, 1996: 21), we also
(or attitude) is likely to be influenced by affective wished to avoid making grandiose claims here that
events. were not merited by available research. We excluded
Finally, AET also called attention to the sub- surprise because research suggests that it may be
jective, phenomenal structure of affect. At the time better considered as an aspect of fear (see Jack,
that AET was presented, most research on work- Garrod, & Schyns, 2014), since these two emotions
related affect focused on mood (positive or nega- have not always been reliably discriminated in
tive state affect). Weiss and Cropanzano (1996) studies of basic emotions (Ellsworth, 2014). We ex-
called for more research on the causes and con- cluded love because there is only limited research on
sequences of discrete emotions, since these emo- the effects of this emotion in work settings, and re-
tions have distinct structures, have a specific cent work on companionate love has been presented
target, are the result of different events, and pro- at the unit level of analysis (Barsade & O’Neill, 2014).
duce different outcomes. Thus, AET discusses the However, the findings for companionate love, al-
subjective structure of affect states (moods and though tentative, are promising for future research
discrete emotions), a point we consider at length. on workplace relationships, and we will return to this
For both positive and negative feeling states, we emotion in the discussion section.
emphasize the distinction between generalized We also discuss emotions that are implied by
affect and discrete emotions. Consistent with AET, the LMX literature. According to Hooper and
we argue that increased theoretical precision can Martin (2008) and Scandura (1999), workers may
result from consideration of the latter. feel upset if they are excluded from their leader’s
ingroup. That is, if members believe their relative
standing is low compared to that of coworkers,
Which Emotions and When They Matter
they will sometimes feel deprived and unfairly
A major contribution of AET to the management treated (Bolino & Turnley, 2009; Tse et al., 2012).
literature was the call for organizational scholars This emphasis on unfairness suggests a moral
to focus on discrete emotions rather than general judgment, and there is a set of emotions pertinent
feeling states. However, a theoretical challenge to these sorts of evaluations. For this reason,
was deciding which discrete emotions we should Stage 3 of our model considers the other-praising
include in our model. Shaver, Schwartz, Kirson, and other-condemning moral emotions identified
and O’Connor (1987) identified over 200 words that by Haidt (2003): gratitude, anger, disgust, and
pertained to different emotions, and this was after contempt. LMX could become stronger in the case
shortening their (much longer) original list. As of gratitude but weaker in the case of the other-
explained by Coté (2005), it would be impossible to condemning moral emotions.
238 Academy of Management Review April

When are emotions manifested? Based on Overview of Stage 1 Mechanisms

prior literature, we discuss emotions that have
Emotional contagion. Based on AET, the leader
been shown to be important for each of these
expresses some sort of emotion at work (e.g.,
processes, such as basic emotions in Stage 1
Humphrey, 2002; Walter & Bruch, 2009). Acting
and moral emotions in Stage 3. Of course, in-
as a type of affective event (e.g., Dasborough, 2006;
dividuals can experience these emotions at
Pirola-Merlo et al., 2002), this emotion spreads
other times and outside the context of LMX re-
to other individuals through a process of conta-
lationships. Hence, we do not list specific dis-
gion (Barsade, 2002; Johnson, 2009). Hatfield,
crete emotions in Figure 1. Further, to conserve
Cacioppo, and Rapson refer to contagion as “the
space, we only review an emotion when the
tendency to automatically mimic and synchronize
emotion is especially relevant to a particular
theoretical mechanism. facial expressions, vocalizations, postures, and
movements with those of another person and,
consequently, to converge emotionally” (1992:
153–154). Accordingly, contagion produces in-
STAGE 1: ROLE TAKING AND LEADER terpersonal exchanges of feelings, such as those
AFFECTIVE EXPRESSIONS between a leader and a member (for similar def-
Role taking is a process that begins very quickly initional approaches, see George, 2002; Hsee,
(Bauer & Green, 1996) and can last from a few Hatfield, & Chemtob, 1992; Pugh, 2001).
hours to a number of weeks (Graen & Scandura, During the process of emotional contagion,
1987). During this stage, which is sometimes emotions are transferred among team members
called the “initial interaction” (e.g., by Dienesch & (Hatfield, Carpenter, & Rapson, 2014). Members
Liden, 1986: 626), the leader may choose to offer the tend to “catch” or “pick up” their leader’s affect
member a high-quality relationship (Graen & Uhl- (Bono & Ilies, 2006; Dasborough, Ashkanasy, Tee,
Bien, 1995). LMX theory posits that the leader ac- & Tse, 2009), so the leader’s emotion serves as an
tively initiates this process (Nahrgang et al., 2009). affective event that can initiate a process of con-
Commensurate with this observation, various tagion throughout a work team (Johnson, 2008,
emotion scholars have argued that emotional 2009). We have more to say about collective emo-
expressions of leaders are of particular impor- tions when we discuss mutual entrainment,
tance to leadership effectiveness (e.g., Humphrey, which occurs during Stage 2 (role making) of our
2002; Humphrey, Pollack, & Hawver, 2008; Rimé, model. Suffice it to say that this initial process of
2009). In this way emotional expressions act as contagion in Stage 1 sets the stage for subsequent
a type of affective event that provides information and more dyadic affective sharing, which occurs
to the observer (follower) about the individual as the LMX relationship matures.
(leader) that can inform future relationship de- Affective empathy. All members will not de-
velopment (see Coté, 2005). Moreover, as Sy, Choi, velop the same relationships with their leaders,
and Johnson (2013) have observed, leader ex- even if they happen to observe the same emotions
pression of affect is likely to be the initiating in them. This occurs because people are differ-
mechanism for linking leaders and followers entially responsive to affective expressions from
affectively. others. Affective empathy refers to the ability to
Consequently, leader affective expressions share the emotional experiences of others (Batson,
serve as an anchoring event influencing the 2009), and it tends to be relatively automatic and
course of subsequent LMX development. Ballinger nonconscious (Zaki & Ochsner, 2012). That is, we can
and Rockmann (2010) define anchoring events as experience it without reflecting or exerting effortful
those events in an exchange relationship that in- thinking (Keysers & Gazzola, 2007). Significantly,
volve high-intensity emotions and that later shape individuals differ in their capacity for affective
the way successive events unfold. These re- empathy, although aspects of the situation can also
searchers explain that once the rules for the re- make other people’s emotions salient (Patient &
lationship have been impacted by such events, the Skarlicki, 2010). As one might expect, when af-
relationship becomes resistant to later changes. fective empathy is high, emotional contagion is
Hence, an anchoring event during Stage 1 of LMX more likely (e.g., Barsade, 2002; Nummenmaa,
development will set the affective tone for what Hirvonen, Parkkola, & Hietanen, 2008; Omdahl &
may follow as the relationship progresses. O’Donnell, 2001).
2017 Cropanzano, Dasborough, and Weiss 239

The process of emotional contagion is en- mood in MBA student leaders who then gave
hanced when individuals are high in affective speeches. Those leaders who experienced a posi-
empathy, since this encourages perspective tak- tive mood induction subsequently expressed
ing and placing oneself in another’s shoes (Kellett more positive affect and were viewed by observers
et al., 2002). As such, empathic individuals are as more charismatic. Those leaders experiencing
more likely to pay attention to and to experience a negative mood induction subsequently expressed
the emotions they witness in other people more negative affect and were judged by observers
(Humphrey, 2002), including negative emotions to be less charismatic. Although these studies did
such as fear (Olsson et al., 2016). Empathic leaders not measure LMX per se, they did establish the
have also been found to develop higher-quality connection between leader positive affect and fol-
LMX relationships than those who are low in lower perceptions of leadership. We argue that this
empathy (see Mahsud, Yukl, & Prussia, 2010; relationship occurs because of positive emotional
Kellett et al., 2002, 2006). We build on this earlier contagion.
work by suggesting that this LMX quality is Happiness/joy. Consistent with AET, there is
somewhat due to enhanced emotional contagion. reason to believe that discrete emotions, not
Thus, in Stage 1, if members are high in affective simply global affective states, influence LMX
empathy, they are more likely to experience development (cf. Connelly & Ruark, 2010). In this
emotional contagion after exposure to their regard, there is reason to believe that expres-
leaders’ emotions. sions of happiness or joy tend to promote high-
Later, in Stage 2, the empathy of both the leader quality LMX relationships. We argue that this
and member are relevant as they become emo- occurs when the target of the leader’s happiness/
tionally entrained. This is what Kellett et al. (2006) joy is within and/or outside of the leader-member
refer to as “interactive empathy.” In situations dyad. Some direct support comes from an ex-
where leaders and members do not become perimental study by Visser, van Knippenberg,
emotionally entrained, we argue that their LMX van Kleef, and Wisse (2013), who studied discrete
relationship will not stabilize (Stage 3, role rou- emotions and found that leaders are perceived
tinization). Unfortunately, previous research on as more effective when displaying happiness
leadership, including but not limited to work rather than sadness.
pertaining to LMX, has not taken into account the
Proposition 1: During the role-taking stage
role of affective empathy in contagion. Given its
(Stage 1), leader expressions of happiness/
importance in establishing the empathic bond
joy—regardless of the target—will act as
between leaders and members, we do so in the
affective events, which will tend to pro-
research propositions that follow.
mote happiness/joy in members. This, in
turn, will promote the development of
Positive Emotional Expressions As a higher-quality LMX relationship. These
Affective Events effects are stronger when members are
high in affective empathy.
Leader displays of positive affect tend to im-
prove members’ moods and also their opinion of Of course, we caution the reader that there
their leader (e.g., Johnson, 2008; Michel, Pichler, & could be specific instances where a leader’s ex-
Newness, 2014; Walter & Bruch, 2009). While evi- pression of joy could serve as an affective event
dence directly testing LMX relationships is lim- that generates negative emotion. For instance, if
ited, theoretical articles and supportive research the leader appears to be degrading the employee
can be found in relation to transformational lead- by taking pleasure at their misfortune or if the
ership (e.g., Ashkanasy & Tse, 2000; Dasborough & leader appears to be disregarding a stressful
Ashkanasy, 2002) and the closely aligned construct work situation, the employee is unlikely to
of leader charisma (e.g., Walter & Bruch, 2009). respond as positively as we suggest in our
In two field studies Bono and Ilies (2006: Studies Proposition 1. Nevertheless, in most settings
1 and 2) found that leaders who expressed positive expressions of happiness or joy tend to be re-
emotions were rated by others as more charis- ciprocated with positive feelings (Visser et al.,
matic; they replicated these findings in a sub- 2013). This effect occurs when the discrete
sequent experiment (Study 4). In a later study positive emotion is targeting either a member of
Johnson (2009) induced either positive or negative the dyad itself or an outside object.
240 Academy of Management Review April

Negative Emotional Expressions As shared. If a leader is expressing anger toward the

Affective Events member, their relationship will be harmed, even
if only temporarily, until the anger is revolved. Part
Research on negative affect suggests that when
of the reason is that, as predicted by research on
members experience unpleasant affect toward
emotional contagion, anger tends to be shared.
their leaders, they tend to rate them less favorably
That is, leader expressions of anger evoke re-
(Dasborough & Ashkanasy, 2002; Dasborough
ciprocal anger in members (see Johnson &
et al., 2009). That said, it is important not to over-
Connelly, 2014). When anger is expressed, the
simplify in this case. Negative affect does not al-
target of the anger reduces their liking for the
ways have unfavorable effects. Even if members
person expressing it (van Kleef, 2014). Once again,
dislike ill-tempered leaders, expressions of neg-
we emphasize the moderating effects of affective
ative mood can boost creativity (George & Zhou,
empathy.
2002) and member performance (Sy, Côté, &
Saavedra, 2005), at least in certain situations. In Proposition 2a: During the role-taking
addition to the environmental setting, AET sug- stage (Stage 1), leader expressions of
gests that is also important to take into account anger targeting objects within the dyad
the idiosyncratic nature of discrete negative will act as affective events, which will
emotions, rather than generalized negative affect tend to promote anger in members. This,
(Weiss, 2002). Although some emotions are he- in turn, will promote the development of
donically unpleasant, they do not necessarily a lower-quality LMX relationship. These
have the same directional impact on LMX. As we effects are stronger when members are
shall now see, there is evidence that discrete high in affective empathy.
negative emotions have distinct effects on leader- Where anger is concerned, there does appear to
member relationships. be an especially important exception to Proposi-
Anger. While it may be unpleasant to expe- tion 2a. Shared anger does not always destroy
rience anger, people who vent their ire are often a relationship. When it is directed at an external
perceived by others to be dominant, and this injustice (e.g., Leonard, Moons, Mackie, & Smith,
does benefit leaders in particular situations 2011) or a harmful individual (e.g., Rimé, 2009), it
(e.g., Knutson, 1996; Lewis, 2000; Lindebaum & can enhance interpersonal cooperation. In this
Fielden, 2011; Reed, DeScioli, & Pinker, 2014; way LMX relationships can actually be strength-
Sinaceur & Tidens, 2006). There is also evidence ened through shared anger (Jasper, 1998).
that managers occasionally display sham an-
ger in order to get their way (Fitness, 2000). Proposition 2b: During the role-taking
Anger can provide a helpful signal to em- stage (Stage 1), leader expressions
ployees about possible threats and can lead to of anger will act as affective events,
adaptation and learning (Ashkanasy & Härtel, which can improve LMX quality, if the
2014). Jackson (1998) showed that expressions of anger is targeted toward an external
anger at Intel, during confrontations, was ac- target and the member shares the same
tually helpful for boosting performance. Simi- feelings toward the target. These effects
larly, in an experiment utilizing student teams are stronger when members are high in
with leaders expressing anger, van Kleef et al. affective empathy.
(2009) found that when people respond to their Sadness. Although we do not often hear about
performance feedback (high epistemic moti- leaders expressing sadness, it can certainly
vation), anger improves their performance. happen, particularly in times of crisis (see Madera
Another recent study by Chi and Ho (2014) also & Smith, 2009). As anticipated by our model,
demonstrates the possible positive effects of sadness is a contagious emotion (Hancock, Gee,
leaders’ expressing anger. Ciaccio, & Lin, 2008). Besides the fact that in-
Although leader expressions of anger may lead dividuals do not usually enjoy feeling sad, there is
to improvements in member performance, our pri- an accompanying problem for unhappy leaders.
mary concern here is with the LMX relationship. Sadness can be perceived as a weakness. Ex-
When it comes to determining LMX relationships, pressions of sadness can lead to perceptions of
we argue that the most important factors to consider reduced dominance (Knutson, 1996) and needi-
are whom the anger is targeting and if the anger is ness (van Kleef & Côté, 2014). Leaders who express
2017 Cropanzano, Dasborough, and Weiss 241

sadness are evaluated lower in leadership effec- of a lower-quality LMX relationship.

tiveness than those with neutral expressions (van These effects are stronger when mem-
Kleef et al., 2009; Visser et al., 2013), and this is bers are high in affective empathy.
especially true for males (Lewis, 2000).
Similar to our propositions regarding anger,
Proposition 3: During the role-taking there is evidence to suggest that fear does not
stage (Stage 1), leader expressions of always destroy relationships. Individuals can
sadness—regardless of the target—will bond together over their shared fear of the same
act as affective events, which will tend external object (e.g., employees share the same
to promote sadness in members. This, fear of managers; Ashkanasy & Nicholson, 2003).
in turn, will promote the development Recently, in a qualitative study Vuori and Huy
of a lower-quality LMX relationship. (2016) found that both top and middle managers
These effects are stronger when mem- experienced fear, but about different targets. They
bers are high in affective empathy. found that top managers feared external com-
petitors, whereas middle managers feared top
We caution, however, that it is not our claim that
managers but not external competitors. Had these
expressing sadness is a universally bad thing in
managers feared the same external target (the
all situations. In some rare scenarios sadness can
external competitors), there may have been more
be beneficial. Madera and Smith (2009), for ex-
positive outcomes and their relationships may
ample, found that a leader expressing sadness
have been stronger.2
following a crisis was evaluated more favorably
than a leader expressing anger. Followers of the Proposition 4b: During the role-taking
leader perceived this sadness favorably, and stage (Stage 1), leader expressions of
sharing the emotion actually made them feel fear will act as affective events, which
connected to the leader. Expressions of sadness can improve LMX quality, if the fear is
can lead to acts of support and cooperation. targeted toward an external target and
Consequently, sadness can bring particular the member shares the same feelings
short-term benefits in specific situations, al- toward the target. These effects are
though expressing it repeatedly will likely harm stronger when members are high in
LMX relationships in the longer term. affective empathy.
Fear. Fear is another basic emotion that has not
been widely studied by leadership scholars. This
lack of research is unfortunate, since leaders of- STAGE 2: ROLE MAKING AND
ten experience fear (Ferris, 1998; Vuori & Huy, DYADIC ENTRAINMENT
2016), as do the rest of us (Dozier, 1998). In general,
fear is a hedonically unpleasant emotion, and The quality of the LMX relationship begins
individuals are likely to prefer leaders who do to crystallize during the role-making stage
not make them afraid (cf. Zoghbi Manrique (Nahrgang et al., 2009; Sin et al., 2009). The leader
de Lara, 2006). This conclusion is supported by instigates this stage by providing the member
with promising assignments, often through del-
evidence indicating that effective leaders tend
egation (Dienesch & Liden, 1986; Schyns et al.,
to be seen by their followers as courageous
2005). The member then responds with, among
( Johnson, 1994). Likewise, Knutson (1996) found
other things, higher or lower performance. If the
that leaders who expressed fear were viewed
leader views the member’s performance as
as less authoritative, although observers ap-
strong, the leader may provide the member with
peared to appreciate the willingness to be openly
additional delegation opportunities. Over time,
vulnerable. From this finding follows our next
and if the member successfully meets these
proposition.
challenges, a high-quality LMX relationship may
Proposition 4a: During the role-taking result (e.g., Schriesheim et al., 1998). We endorse
stage (Stage 1), leader expressions of this model but find it to be incomplete.
fear targeting objects within the dyad Research on LMX development indicates that
will act as affective events, which will affect is part of the picture (e.g., Liden et al., 1993;
tend to promote fear in members. This,
in turn, will promote the development 2
We thank a special issue editor for pointing this out to us.
242 Academy of Management Review April

Sears & Hackett, 2011). In Stage 2 (role making) the FIGURE 2

follower is now a more active participant in the Stage 2 of LMX Development: Dyadic Emotional
process of LMX development. Here the leader and Entrainment
follower are both sources of affective events for
each other. Each transaction between the dyad Pattern of shared emotion over time
members can cause an emotional response in one
or both members. As indicated by Sin et al.’s (2009)
findings, what is important here is not the fre-
quency of interactions but, rather, what happens Mutual
Leader Member
within the interactions. Now we see the social emotional
emotion emotion
exchange of emotions within the dyad. The fol- entrainment
lower is not only observing the emotions of the
leader but is also initiating emotional responses
and sharing emotions as a member of the dyad.
The active roles of the leader and member are il-
lustrated in Figure 2. This active exchange of
feelings allows the LMX relationship to become
routinized (Stage 3). Next we outline the empirical
LMX
support for our theorizing. As we shall see, AET
relationship
provides additional reasons to believe that af-
quality
fective states, when they are synchronized be-
tween two individuals over the long term, can
impact LMX relationship quality. Explaining this
phenomenon will require us to examine entrain- These observations can be extended to the
ment within individuals before moving on to dy- dyadic level, with leaders and members often
adic entrainment. sharing common work events. However, our
model goes beyond this in emphasizing that each
individual influences the other.
Rhythms of Positive Affect
Mutual entrainment. The studies mentioned
Affective events theorists have observed that previously explored the entrainment of indi-
affective states, by definition, are not stable. They viduals (not dyads or groups) to various experi-
tend to fluctuate together (Beal et al., 2005; Weiss, ences in their environment, such as the calendar
Nicholas, & Daus, 1999), oscillating in regular (Larsen, 1987). Thus, the emotions of coworkers
patterns (Cropanzano & Dasborough, 2015; could come to exhibit common rhythms because
McFarlane, Martin, & Williams, 1988; Stone, 1981). these individuals are entrained to the same work
Entrainment is a term that refers to a dynamic schedule (Larsen & Kasimatis, 1990) or experience
process of emotional fluctuations across some the same work stressors (Trougakos et al., 2008).
interval of time (Larsen, Diener, & Emmons, 1986). These shared experiences, though important, are
For example, Watson, Wiese, Vaidya, and only part of the story. Perhaps more relevant to our
Tellegen (1999) found that positive affect could theory, individuals tend to impact one another’s
become entrained to a circadian (daily) sched- feelings directly, as we saw earlier when discus-
ule, improving during the morning but declining sing emotional contagion. When two or more
later in the evening. Individuals also become people interact in close proximity, their physical
entrained to the weekly work schedule. Addi- actions tend to be come synchronized or
tional studies have shown that moods fluctuate “entrained,” often unintentionally (Chartrand &
throughout the week; they are worse early but Bargh, 1999: Study 1; Cropanzano & Dasborough,
improve as the weekend approaches (Larsen, 2015).
1987; Larsen & Kasimatis, 1990). A large part of Although the idea of entrainment was origi-
this affective variability is due to the events that nally used to examine biological rhythms
people encounter in their everyday life. For ex- (Lakin et al., 2003), McGrath and Kelly (1986)
ample, Sonnentag (2001) and Trougakos, Beal, extended it to examine behavioral rhythms
Green, and Weiss (2008) identified mood cycles and social synchrony (i.e., social entrainment).
as individuals recovered from stressors at work. Social entrainment has been used to understand
2017 Cropanzano, Dasborough, and Weiss 243

phenomena in organizational settings (e.g., (Stage 1). However, once these individuals be-
Ancona & Chong, 1996). While social entrain- come entrained to their work settings, as well as to
ment includes physical actions, it is not limited to one another, they are likely to exhibit shared af-
them. Recall our earlier comments on emotional fective rhythms (Stage 2). Moreover, when en-
contagion, which we discussed in Stage 1. The trainment is experienced, it may improve LMX
affect expressed by the leader influences the relationship quality. Kelly and Barsade (2001)
subsequent feelings of the member (Dasborough explained that the entrainment of individuals
et al., 2009). Over time, these emotional re- leads to more positive interactions and boosts
sponses may tend to follow a particular pattern. positive affect. Two experiments conducted by
In this way affect, like physical actions, can be- Damen, van Knippenberg, and van Knippenberg
come coordinated as well (e.g., Levenson & (2008) are relevant here. These authors found that
Gottman, 1983; Mansfield et al., 1989; Siegman leaders who displayed positive emotion were
& Reynolds, 1982). able to motivate followers to higher performance,
Notice that emotional entrainment (in Stage 2) is but only when the followers were also experi-
a process that follows smoothly from the emo- encing positive affect. This result parallels the
tional contagion that occurs during the role- aforementioned findings of Bauer and Green
making stage (Stage 1). Within our AET/LMX (1996). Building on these findings, we suggest
model, people can become mutually entrained that when leaders and followers share the same
(i.e., exhibit common patterns of alternating af- affective pattern over time, the outcomes will
fective states) in large measure because conta- generally be positive.
gion has already made their emotional states Shared mood oscillation. When moods are
more similar. Therefore, we extend earlier re- mutually entrained, they are likely to share
search on mutual entrainment by arguing that it a similar average valance (e.g., be similarly pos-
tends to result from contagion. Nevertheless, itive or similarly negative). However, entrainment
the concept of entrainment builds on the concept goes beyond shared means, suggesting that
of contagion in two important ways. First, entrained moods also fluctuate in roughly the
entrainment explicitly pertains to a predictable same rhythm. That is, they wax and wane in
pattern of change that occurs over time (Larsen, a shared sequence. We know this shared se-
1987). Contagion refers to the interpersonal quence occurs in leader-member relationships,
transfer of affective states from one individual to with various researchers pointing out that mem-
another, generally with the effect of making the bers often share their leader’s emotions (Gooty
people in question more similar (Hatfield et al., et al., 2010).
2014). Second, the fluctuating mood states ob- We illustrate the distinction between average
served by entrainment researchers are often in level and shared oscillation in Figure 3. This
response to some environmental experience representation depicts two individuals: a leader
(e.g., the work schedule) or event (e.g., work and a member. The level of positive affect is
stress). Contagion scholars focus on the in- indicated by the vertical axis on the left. The
terpersonal transfer of emotions (Hatfield et al., within-person mean levels of affect for each
1992), explicitly contrasting this transfer with
situational experiences (e.g., George, 2002; Hsee
et al., 1992). In short, contagion and entrainment FIGURE 3
are important ideas that complement one an- Entrainment of Positive Affect: Different Means
other. Contagion emphasizes emotional changes with Similar Patterns of Fluctuation
across individuals, generally making them more
alike, whereas entrainment emphasizes emo-
Level of positive affect

tional change across time, generally exhibiting

predictable oscillations.
Of course, if entrainment did not admit in-
terpersonal influences, it could not be used in our
present model, which seeks to explain dyadic
LMX relationship development. As we will dis-
cuss in a moment, contagion creates an initial
similarity between the leader and the member Time
 244 Academy of Management Review April

individual are shown in the two bisecting dot- individuals have about the same average level of
ted lines. In this instance one individual has positive affect. However, this dyad is misaligned
a lower mean relative to the other. The two sets with respect to the pattern of fluctuation. While
of moods are positively correlated, however, the curves cross in more places than in, say,
because they covary in the same directions, but Figure 3, the individuals in Figure 4 tend to move
they do not always agree because one person is in different directions. These scores are nega-
usually higher than the other. These people tively correlated because, as one mood is im-
have responded to their environments in the proving, the other is becoming worse. These are
same way, although not at the same average two very different individuals. They are happiest
level. In other words, they are at least somewhat at different times, and one presumes that they
entrained. If one of these individuals is a leader are made so by different things. Despite the
and the other a member, then, other things be- overall level of agreement in their mean posi-
ing equal, we would expect them to have a high- tive affect, we predict that the quality of their
quality LMX relationship, although perhaps not LMX relationship will suffer because of their
as high quality as it would be if the two in- lack of entrainment. Mutual entrainment of
dividuals had the same means. positive emotions is required for developing high-
The evidence bearing on this possibility is quality relationships.
very limited, but there are a few studies that
Proposition 5: During the role-making
are broadly consistent. For one thing, people in
stage (Stage 2), mutual entrainment
dyads are able to read one another’s feelings of positive emotions in response to af-
(Totterdell et al., 1998: Study 2). In terms of emo-
fective events over time—regardless
tional convergence, there is supportive evidence
of the target—will strengthen LMX
from research on romantic relationships. Stone
relationships.
(1981) had twenty-six married partners complete
mood diaries about their spouse’s affect. Over To our knowledge, these entrainment effects
a two-week period, both spouses’ ratings tended have not been studied with respect to LMX de-
to be correlated. Later, Anderson, Keltner, and velopment. This lack of attention is unfortunate,
John (2003) found that individuals in romantic since these effects have been documented in
relationships became more emotionally similar peer relationships. Totterdell and colleagues
over time. Further, they found that the less pow- (1998) followed thirteen teams of nurses for
erful member of the romantic couple was more three weeks. These researchers reported that
likely to make the changes necessary for emo- during this period, teammates’ moods tended to
tional convergence to occur. This emotional be correlated, although more for nurses who
convergence is of particular interest to us, given were older than for those who were younger. In-
our interest in LMX relationships where clear terestingly, the moods of these nurse teams were
power differentials exist. not perfectly stable over time. Totterdell and his
As a contrary comparison, consider Figure 4. colleagues reported that workload hassles and
As presented here, these two hypothetical work regulations negatively affected feeling
states, reflecting that entrainment can result
from shared events. However, this was not the
FIGURE 4 whole story, since the association of the indi-
Positive Affect over Time: Similar Means with vidual mood with the team mood persisted even
Different Patterns of Fluctuation when daily stressors were controlled. We would
expect this persistence if workers’ affective ex-
periences were entrained.
Level of positive affect

In a subsequent investigation Totterdell (2000)

had members of two professional cricket teams
complete mood diaries. Once again, the moods of
individual team members were correlated with
the collective mood of the team. Notably, only
positive affect—not negative affect—showed this
convergence. Individual performance (batting
Time average) was positively associated with team
2017 Cropanzano, Dasborough, and Weiss 245

mood. As was the case in Totterdell et al. (1998: divisive (e.g., Dasborough & Ashkanasy, 2002;
Study 1), cricket players who shared positive Dasborough et al., 2009). Thus, alignment of neg-
moods reported more commitment to the team. ative emotions would be expected to produce
Shared experiences promoted entrainment as poor-quality LMX relationships in the case where
well. Totterdell found that the moods of the cricket the emotions are targeted toward members within
players were most positive when their match was the dyad.
going well, as opposed to when it was going
Proposition 6a: During the role-making
poorly. Thus, feeling states tended to fluctuate
stage (Stage 2), mutual entrainment of
based on the circumstances in which the team
negative emotions in response to af-
found itself. Given that individual affect, at least
fective events over time will result in
positive affect, was associated with the team’s
the development of higher-quality LMX
collective mood, this suggests some degree of
relationships, if the negative emotions
team entrainment. The work of Totterdell (2000)
are targeted toward objects outside of
and Totterdell et al. (1998) is based on peer-to-peer
the dyad.
entrainment and not leader-to-member entrain-
ment. For this reason it does not directly test Proposition 6b: During the role-making
Proposition 5. However, this research does show stage (Stage 2), mutual entrainment of
that emotional entrainment effects exist and are negative emotions in response to af-
important to collective performance and commit- fective events over time will result in
ment. In this way they are consistent with the the development of lower-quality LMX
theoretical claims we make here regarding LMX relationships, if the negative emotions
relationships. are targeted toward objects within the
dyad.
Lawful Rhythms of Negative Affect
STAGE 3: ROLE ROUTINIZATION AND
As one might expect, individuals do tend to ex-
EMOTIONAL IMPACT OF LMX
press negative affect in regular patterns (Larsen
DIFFERENTIATION
et al., 1986: Study 1), although perhaps not as con-
sistently as with positive affect (Watson et al., In the third stage, role routinization, the quality
1999). As we discussed earlier, individuals tend of the LMX relationship is reasonably well
to acquire the negative moods of their leaders established. For this reason, when researchers
(Dasborough et al., 2009; Sy et al., 2005). However, have explored LMX development, their inves-
this tendency may be more likely in the case of tigations have sometimes terminated without
high-arousal negative emotions (e.g., fury) than closely investigating the third stage (e.g., Bauer &
in the case of lower-arousal negative emotions Green, 1996; Nahrgang et al., 2009; Sin et al., 2009).
(e.g., depression; see also Walter & Bruch, 2008). This theoretical position is not unreasonable and,
From this finding it appears that negative emo- in fact, is generally consistent with the evidence.
tions can demonstrate mutual entrainment. While While there can certainly be some changes over
mutual entrainment of positive emotions is ben- time, mature LMX relationships tend to exhibit
eficial for LMX relationships, the case of mutual reasonable stability (Gerstner & Day, 1997; Kelley
entrainment of negative emotions is not so & Bisel, 2014).
straightforward. That said, AET suggests that the role routinization
On the one hand, if a leader and member have stage deserves closer scrutiny because it may be
mutually entrained negative emotions targeting less permanent than is sometimes implied. If un-
someone or something external to the dyad, this anticipated events cause new affective reactions,
can be as bonding for the dyad as shared positive and if affect influences the development of LMX,
emotions (Rimé, 2009), although it may feel less then the relationship is likely to be further altered.
pleasant. On the other hand, sharing negative AET, in other words, causes us to look for attributes
affect will be deleterious to the quality of leader- of the work environment that are inherently dy-
member relationships if the negative affect is namic or, at least, that produce dynamic affective
targeted toward a member of the dyad. Negative events. We are not only discussing extraneous
emotions, especially those that are active or hos- events, although such events are certainly impor-
tile (e.g., anger, contempt, resentment), tend to be tant. Rather, our point here is that the roots of
246 Academy of Management Review April

change are built into the nature of work settings quality (Graen & Scandura, 1987; Henderson
and, by extension, even into the nature of LMX. et al., 2009; Scandura, 1999). Considerable evidence
An AET approach suggests that despite the im- supports this contention (e.g., Ford & Seers, 2006;
plied steadiness of the role routinization stage, Schyns, 2006). In this way the overall arrangement
LMX relationships often contain the seeds of of LMX scores can be diverse (high differentiation
future affective events, which in AET’s reckon- or low consensus) or similar (low differentiation
ing could upset the constancy of any relation- or high consensus). This differentiation predicts
ship. Specifically, research suggests that the member responses such that low consensus (high
relative standing on LMX impacts worker atti- differentiation) is sometimes disadvantageous.
tudes and, perhaps, behavior (cf. Henderson, For example, while research suggests that LMX is
Wayne, Shore, Bommer, & Tetrick, 2008; Schyns, positively related to commitment (Gerstner & Day,
2006). These effects are stronger when LMX dif- 1997), this association becomes weaker when
ferentiation is high (i.e., when there is a wide differential LMX is perceived to be high (van
distinction between those with high-quality Breukelen, Konst, & van der List, 2002). Likewise,
relationships and those with low-quality Hooper and Martin found that perceived LMX
relationships). differentiation (these authors actually used the
If for no other reason, changes in relative equivalent term variability [2008: 20]) appeared to
standing are likely to occur because membership increase the number of relational conflicts within
in workgroups changes over time. Leaders are a team (for similar findings see Boies & Howell,
forming LMX relationships with multiple mem- 2006). These conflicts, in turn, reduced member
bers, and a certain amount of personnel churning well-being and job satisfaction (but for more
occurs in most work settings (i.e., people come and mixed results see Schyns, 2006). As we shall now
go or move up and down), so any given in- see, these ill effects are worse for some members
dividual’s relative position is ripe for periodic than for others.
reassessment. This reassessment can be
a positive or negative affective event, depend-
LMX Differentiation Among Higher-Status
ing generally on whether it is “upwards” or
Members: The Role of the Other-Praising Moral
“downwards,” when individuals compare their
Emotion of Gratitude
own LMX relationship to others in the work-
group. For instance, when a respected senior As one might expect, the thorny issues that are
employee retires, the employee’s departure sometimes associated with LMX differentiation
could cause other members to move upwards tend to be less troublesome among those with
one step in the LMX social hierarchy. We would high-quality LMX relationships. For example, Ma
expect this to improve LMX relationships be- and Qu (2010) found that even after controlling for
tween the leader and the remaining benefi- objective performance, leaders tended to give
ciaries. At other times, changes do not always higher performance ratings to ingroup members.
go well. For example, workplace politics might This inflation was more likely when LMX differ-
move someone out of the leader’s ingroup. To entiation was high. Moreover, Henderson and
the extent that this is viewed as unjust, it will colleagues (2008) observed that members’ relative
likely lead to declines in LMX quality for that standing on LMX was positively related to per-
particular employee. ceptions of psychological contract fulfillment, and
We develop these ideas below. As we shall see, this was especially so when differentiation was
these observations require that we step up to high. These researchers further noted that relative
a higher level of analysis and consider affective standing on LMX tended to increase job perfor-
events that result from LMX at the group level mance and sportsmanship (a dimension of OCB).
(cf. Ashkanasy & Humphrey, 2011; Yammarino & Taken together, these findings suggest that when
Markham, 1992; see also Figure 1). LMX differentiation is high, this is a positive af-
fective event for members with high-quality
relationships.
LMX Differentiation
When individuals receive a personal benefit
LMX theory suggests that leaders develop in- from another person, and especially when that
dividual relationships with members such that benefit has value, they are likely to experience
some are of higher quality and others are of lower the other-praising moral emotion of gratitude
2017 Cropanzano, Dasborough, and Weiss 247

(Fredrickson, 2004; Haidt, 2003). Gratitude is effects when it is viewed as unjust but not other-
a positive and important emotion (Emmons & wise. Erdogan and Bauer (2010) found that if work
Crumpler, 2000; Fredrickson, 2004). People who units had a high justice climate, then high levels
experience gratitude report fewer physical of differentiation were somewhat weakly related
symptoms related to stress (Froh, Yurkewicz, & to work attitudes and withdrawal behaviors.
Kashdan, 2009), sleep better (Wood, Joseph, Lloyd, However, when justice climate was low, high
& Atkins, 2009), and report a higher level of sub- levels of differentiation were strongly and nega-
jective well-being (Emmons & McCullough, 2003; tively related to work attitudes and withdrawal.
Emmons & Shelton, 2002; Froh, Sefick, & Emmons, As Hooper and Martin (2008) and Scandura (1999)
2008; McCullough, Kilpatrick, Emmons, & Larson, argued, fairness appears to be a critical variable
2001). Of particular relevance to our model, grati- for understanding the effects of LMX differen-
tude also helps people develop richer interper- tiation. When LMX differentiation is high, team
sonal relationships (Emmons, 2007; Froh et al., members tend to engage in sensemaking activities
2009). Through this mechanism of increased grati- (cf. Weick, 1995), discussing the distribution among
tude, we expect that high LMX differentiation will themselves in order to understand whether or not it
improve the quality of LMX relationships among is fair (Sais & Jablin, 1995). As anticipated, studies
those who fare well. by Pillai, Scandura, and Williams (1999) and
Vecchio et al. (1986) have shown that LMX differ-
Proposition 7: During the role routini-
entiation may sometimes be negatively related to
zation stage (Stage 3), when LMX dif-
fairness perceptions. When LMX differentiation is
ferentiation is perceived by a member
high and unfair, members with low-quality re-
to be high, this will act as a positive
lationships (outgroup members) will respond
affective event generating gratitude
badly and will reevaluate the quality of their LMX
toward the leader among those with
relationship.
relatively high LMX. As a result, the
Although anchoring events, which set the stage
quality of that member’s LMX relation-
for relationship quality, are less likely in late
ship will further improve.
stages of a relationship, they are more likely to
occur in reciprocal relationships that are negative
(Ballinger & Rockmann, 2010). Hence, we are more
LMX Differentiation Among Lower-Status
likely to see late-stage anchoring events affect
Members: The Role of Other-Condemning
LMX outgroup members than ingroup members.
Moral Emotions
Here, in Stage 3, the perception of unfair LMX
We have already reviewed evidence indicating differentiation is viewed as a reanchoring event,
that LMX differentiation is negatively related to since this occurs after the initial anchoring events
work outcomes, at least among those who are presented in Stage 1. This reanchoring event
relatively lower status (Henderson et al., 2008). arouses intense emotions and changes the nature
However, other research suggests that high dif- of the LMX relationship once again.
ferentiation is not always a negative affective Negative moral emotional responses to
event. In a fascinating study, Liden, Erdogan, unfairness. When individuals witness unfair status
Wayne, and Sparrowe (2006) found that LMX changes, they are likely to feel other-condemning
differentiation may boost individual job per- moral emotions (Folger & Cropanzano, 2010). These
formance for those with low-quality LMX re- emotions could lead to the precipitous deterioration
lationships but have no impact on those with of LMX quality. In a detailed review of the literature,
high-quality LMX relationships. These authors Haidt (2003) identified three other-condemning
suggested that high differentiation, far from moral emotions: anger, disgust, and contempt. If
causing a decline in motivation, encouraged the level of LMX differentiation violates a moral
those with low scores to exert extra effort. At norm by creating an injustice, then one or more of
first glance, these findings may appear coun- these emotions are likely to be elicited. To the extent
terintuitive, but they have a straightforward that this occurs, the LMX relationship will go into
explanation. further decline, although for different reasons with
According to Hooper and Martin (2008) and each discrete emotion. As we shall see, anger tends
Scandura (1999), perceived fairness is an impor- to cause assertive or aggressive actions, disgust
tant moderator. LMX differentiation produces ill causes one to back away from another person, and
248 Academy of Management Review April

contempt greatly reduces the amount of respect one relationship. These ill effects occur when a sense
person has for another (Jones & Fitness, 2008). of injustice triggers one or more of the powerful
Earlier we discussed anger as a basic emotion. other-condemning moral emotions—anger, dis-
However, anger can also be treated as a moral gust, and contempt. Anger tends to elicit confron-
emotion, when the eliciting event is a violation of tational, or at least passive-aggressive, actions.
a norm (e.g., an unfair event) rather than some Disgust tends to produce avoidance, and contempt
other cause (Jasper, 1998). Anger is a common creates a sense of disrespect. In each case the de-
response when an individual is treated unjustly stabilization of previously established LMX re-
(Krehbiel & Cropanzano, 2000; Weiss, Suckow, & lationships is likely to result.
Cropanzano, 1999). Angry individuals tend to
Proposition 8: During the role routiniza-
seek revenge (Aquino, Tripp, & Bies, 2001; Bies &
tion stage (Stage 3), when LMX differ-
Tripp, 2002; Tripp & Bies, 2009, 2015). When this
entiation is perceived by a member to be
possibility is blocked, their response to a perceived
high and unjust, this will act as a nega-
injustice is likely to be passive-aggressive be-
tive affective event generating negative
havior (i.e., withholding a helpful behavior; Walker
moral emotions, such as anger, disgust,
& Richardson, 1998) or simply social avoidance
and contempt toward the leader, among
(Aquino, Tripp, & Bies, 2006; Fitness, 2000). All of
those with relatively low LMX. As a re-
these responses, especially the tendency to seek
sult, the quality of that member’s LMX
revenge, will reduce LMX quality.
relationship will further decline.
A “disgusting” action is morally impure (Rozin,
Haidt, & McCauley, 2000; Rozin, Lowery, & Ebert,
1994; Rozin, Lowery, Imada, & Haidt, 1999), and
DISCUSSION
disgust appears to impact other moral judgments
as well (Schnall, Benton, & Harvey, 2008; Schnall, In this article we discussed the importance of
Haidt, Clore, & Jordan, 2009). As we do with dis- affect for the development of LMX relationships.
gusting food, we tend to avoid “disgusting” peo- Toward this end, we expanded the multistage
ple (Haidt, 2003). Consequently, members are apt model of LMX through the application of AET.
to “back away” from their leader. This reduced Using AET, our model integrates several bodies
interaction, which is also negatively tainted, can of literature into a single theoretical framework.
diminish LMX quality. These include work on basic emotions, emo-
People respond with contempt when someone tional contagion, emotional entrainment, LMX
behaves in such an extremely reprehensible differentiation, and moral emotions. Based on
fashion that others feel morally superior to the earlier LMX research (Dienesch & Liden, 1986;
transgressor (Haidt, 2003). Indeed, contempt la- see also Ferris et al., 2009), we suggested that
bels the wrongdoer as beneath us (Izard, 1977; LMX relationships develop out of a series of affec-
Miller, 1997). Interestingly, contempt does not tively charged interactions between leaders and
necessarily motivate people to retaliate, as is the members. These interactions move through
case with anger. However, contempt appears to three stages (Graen & Scandura, 1987; Nahrgang
be more destructive regarding interpersonal re- et al., 2009; Sin et al., 2009): role taking, role making,
lationships (Fredrickson, 2009). When we are and role routinization. Each stage highlights
contemptuous of another individual, we are less a different affective phenomenon, which we drew
kind and supportive, seeing the other individual from AET (Weiss, 2002; Weiss & Cropanzano, 1996)
as a clownish idiot who is not worthy of effort and other theories of workplace emotion and
on our behalf (Oatley & Johnson-Laird, 1995). leadership (Ashkanasy, 2003; Ashkanasy &
This sort of disrespect is likely to harm LMX Humphrey, 2011).
relationships. During the role-taking stage, LMX development
Closing thoughts. Even during the routinization begins when the leader initiates the opportunity
stage, the quality of the relationship can occa- for a high-quality relationship with the member
sionally be disrupted by group-level factors that (Graen & Scandura, 1987). In our model, as in most
are external to the dyad itself. In particular, the of the existing research on LMX, this is treated as
overall distribution of LMX quality among team an individual-level process whereby leader af-
members can have an independent impact on fective expressions are affective events for mem-
the long-term prognosis for the dyadic LMX bers (Humphrey, 2002; Humphrey et al., 2008),
2017 Cropanzano, Dasborough, and Weiss 249

influencing their individual emotional responses quality—could be shifted within a work team as
via emotional contagion (e.g., Bono & Ilies, 2006; new members come and others go. This could, in
Johnson, 2008, 2009). turn, trigger other-condemning moral emotions
Once the role-taking stage is successfully toward the leader and damage existing LMX
navigated, the leader and member enter the relationships.
second stage, role making. At this time the leader
and member may become affectively entrained
Avenues for Additional Conceptual Development
such that they share the similar types of affect at
similar times (Chartrand & Bargh, 1999). Apart Apart from providing an account of LMX devel-
from their shared situations, the leader and opment over time, the integration of the AET and
member influence one another directly. These the LMX literature goes further, opening doors for
dyadic-level affective experiences shared between future research inquiry and further conceptual
the leader and member create an empathic growth. In this regard we believe that our present
bond between them, leading to a higher-quality model is especially rich. Not only does it offer an
LMX relationship. account of LMX development in terms of affective
In the final stage, role routinization, the LMX events theory but, in doing so, it also raises
relationship has reached an initial level of sta- a number of questions. While future directions are
bility. As such, it begins to act as an antecedent to difficult to predict, there are five issues that we
further emotion on the part of members. At the think are especially interesting.
group level of analysis, leaders may form highly Reverse causality and feedback loops. To keep
differentiated LMX relationships with members. our presentation straightforward, we emphasized
Those with lower-quality LMX relationships rel- certain causal directions while deemphasizing
ative to teammates may feel a sense of injustice others. The antecedent paths that we discussed
(Scandura, 1999). This can produce such moral are consistent with the literature, as well as with
emotions as anger or contempt, which may lead to both AET and LMX research. We believe that fu-
a decline in future LMX quality. ture research is likely to find further support for
In addition to providing a fuller account of them. However, AET indicates that feedback
LMX development, our theoretical model also loops, so-called reverse causality, will also exist.
extends AET, which has traditionally been For example, in Stage 1 we focus on the leader as
examined at the individual level of analysis. the origin of affective events. While this aligns
However, LMX has not been so limited (e.g., nicely with LMX theory (Graen & Scandura, 1987;
Henderson et al., 2009; Schriesheim et al., 1999; Nahrgang et al., 2009), there are no doubt in-
Schriesheim et al., 2000; Schyns & Day, 2010). stances where the leader is impacted by the
Consequently, in order to make these two con- emotional expressions of followers (see Tee &
ceptual models commensurate with one another, Ashkanasy, 2008). Likewise, in Stage 3 we em-
we have added a new level to AET—the dyad—in phasized that high levels of LMX differentiation
discussing how emotions can become entrained could produce negative affect in followers. While
over time. In the third stage of our model, role this is also consistent with LMX research
routinization, we further consider how a group- (e.g., Erdogan & Bauer, 2010), the alternative path
level variable—LMX differentiation—could act could be possible as well. That is, followers who
as an affective event. We hope that future work chronically express negative emotions may be-
will continue to develop AET at different levels of come less popular and therefore be relegated to
analysis. an outgroup. With this in mind, future investiga-
Theoretical thinking on LMX has been ex- tors should explore the possibility of feedback
panded as well. Not only have we articulated an loops, for the causal arrow will not always go
important role for affective events, but in our in a single direction.
model LMX relationships may remain less stable Another important candidate for future in-
than researchers have previously emphasized. quiry concerns LMX quality. It appears likely
Part of the problem is that the LMX relationship that the quality of LMX relationships can impact
may be impacted by affective events that are the future emotional expressions of leaders
beyond the direct control of the leader and (Dasborough et al., 2009). For that reason we
member. As a result, something like LMX suspect that there could be a reciprocal path
differentiation—the distribution of relationship from LMX to leader emotional expression. This
250 Academy of Management Review April

should be an important priority for future in- powerful peer. Likewise, compassion is more apt
quiry. As demonstrated by Cwir, Carr, Walton, to occur when affective empathy is present than
and Spencer (2011), in social relationships the when empathy is absent (cf. Batson, 2009). Given
quality of the social connection can lead to in- these observations, researchers in future may
creased sharing of emotional states. The possi- wish to go beyond the contagion of corresponding
bility of feedback loops suggests that there may emotions and extend our model to the examina-
also be positive affective spirals (cf. Walter & tion of complementary emotions.
Bruch, 2008), where high-quality LMX relation- Individual differences. Individual differences
ships breed future positive dyadic affective have proven to be of considerable importance
events and positive affective climates. Over for understanding organizational behavior
time, the two processes begin to reinforce one (e.g., Judge, 1992; Judge, Klinger, Simon, & Yang,
another, making relationships steadily stron- 2008). Research into AET and LMX is consistent
ger. Negative spirals are possible as well with this. In their original statement of AET, Weiss
(Dasborough et al., 2009). Implications of this and Cropanzano (1996) discussed the importance
kind should receive empirical investigation. We of individual differences, including dispositions to
hope that our present model, grounded in AET, express positive affect (trait PA) and negative af-
will prove to be a reliable guide. fect (trait NA; for further reviews of positive and
A closer look at contagion. In our proposed negative affect, which are relevant to work set-
theoretical model, contagion plays an important tings, see Cropanzano, Weiss, Hale, & Reb, 2003,
role. As we have noted, members tend to “catch” and Judge & Larsen, 2001). Within the LMX litera-
the emotions of their leaders, especially when ture, there is suggestive evidence that positive af-
affective empathy is high (Propositions 1–4; see fect (Bauer & Green, 1996) and other personality
also Omdahl & O’Donnell, 2001). Later, at Stage 2, traits (Nahrgang et al., 2009) impact LMX relation-
these shared emotions are likely to fluctuate to- ship quality. Within our model, individual differ-
gether through a process of entrainment. As is ences are likely to appear as causes of leaders’
true of much of the literature on leadership and emotional expressions. That is, they are ante-
affect (e.g., Barsade, 2002; Bono & Ilies, 2006; cedent to the antecedents in our model. There is
Johnson, 2008, 2009), we have emphasized the good evidence for this. Nahrgang and colleagues
contagion of consistent emotions, such as when (2009) found that leaders who are agreeable and
a leader’s sadness engenders a corresponding extroverted (this latter trait is closely related to
sadness in an employee (Hancock et al., 2008; trait PA) are more likely to initiate higher-quality
Visser et al., 2013). While this important process LMX relationships with followers.
has been empirically demonstrated, there is also The Nahrgang et al. (2009) study is important for
evidence of another possibility. At times the an additional reason. Apart from the broad af-
emotional expressions of one individual can elicit fective traits of negative and positive affect, it
complementary emotions from another individual appears to be use useful to examine other dispo-
(Folger & Cropanzano, 2010; Folger, Cropanzano, sitions as well. In selecting theoretically relevant
& Goldman, 2005). For instance, when a person traits, AET provides a relevant starting point. The
expresses anger, another individual might expe- theory indicates that discrete emotions are im-
rience fear (Brosch, Sander, Pourtois, & Scherer, portant. Fortunately, a number of well-validated
2008). These complementary emotions could some- traits could be helpful in this regard. For exam-
times cross the boundary between positive and ple, leaders with high levels of trait anger
negative hedonic tones. A frightened person, for (Deffenbacher, 1992) and trait anxiety (Freeman
example, might elicit compassion in an observer & Freeman, 2012) may be more likely to initiate
(Konov et al., 2004). negative affective events impacting their
While in this article we have not considered members.
complementary emotions in detail, it seems likely LMX research cautions that the member, as well
that many of the concepts included in our model as the leader, is important. With this in mind, we
could be extended to assist with this analysis. To should also consider how the personality of the
illustrate, consider the power dynamics between member impacts their responses to their leader
a leader and a member in an LMX relationship. An (Bauer & Green, 1996). Speaking very generally,
angry leader could be more likely to elicit fear for example, Brockner (1988) has found that people
in a less powerful member than in an equally with low self-esteem are more reactive to their
2017 Cropanzano, Dasborough, and Weiss 251

environments than are those with high self- and for that reason we have paid close attention to
esteem. Other dispositions, such as emotional moderator variables. For example, we have ar-
intelligence, trait affect, and susceptibility to gued that affective empathy strengthens the re-
emotional contagion, should impact how mem- lationship between leader emotional expressions
bers respond to affective events (Antonakis, and member affect. We have also maintained
Ashkanasy, & Dasborough, 2009; Dasborough, that fairness moderates the impact of LMX dif-
2006; Dasborough, et al., 2009; Doherty, 1997). We ferentiation and that the target of negative
encourage researchers to use our AET/LMX the- emotions affects their relationship to LMX. While
ory as a guide to identifying relevant traits and these moderating effects are important, we cau-
understanding how they fit into the model. For tion that there are innumerable other variables
example, the emotion of gratitude is important that could modify the relationships identified in
at Stage 3, particularly when LMX differen- our model.
tiation is high. Interestingly, Wood, Maltby, Other emotions. As we observed earlier, we
Stewart, Linley, and Joseph (2008) found that highlighted well-researched emotions that
gratitude can be expressed as a relatively stable were useful for illustrating the affective processes
disposition—higher in some and lower in others. described in our model. We further emphasized
Our Proposition 7 would be most likely to hold emotions that were suggested by AET and LMX
among members high in this trait. research. Of course, this means that there are
Additional situational antecedents and other potentially important emotions not consid-
moderators. We have proposed an integrative ered in this article (cf. Shaver et al., 1987). For ex-
model of AET and LMX development. However, ample, in a recent article Barsade and O’Neill
LMX researchers have identified a number of (2014) investigated work cultures of companionate
antecedents to LMX that do not (or, at least, do not love and found evidence of higher life satisfac-
directly) involve emotion or affect (cf. Graen & tion, less absenteeism, and less stress when
Uhl-Bien, 1995). While these variables are not these cultures were present. While Barsade and
explicitly affective, they are worth considering in O’Neill’s work was at the unit level of analysis, we
future research. These include such things as see great potential for this emotion to explain the
contingent reward behavior (Dulebohn et al., development of dyadic LMX relationships. Fur-
2012), group size (Cogliser & Schriesheim, 2000), ther explorations of this kind into this and other
span of control (Schyns et al., 2005), and follower discrete emotions would likely be fruitful.
performance (Bauer & Green, 1996), among
others. Future research should take these situa-
Conclusion
tional features into account. Consistent with
AET, we believe that many of these antecedents In this article we have considered the three
will exert their effects on LMX relationships by stages of LMX development from the perspective
way of the affect they produce in the leader and/ of AET. In so doing we have integrated a number
or member. For example, Borchgrevink and of bodies of literature into a single integrative
Boster (1997) proposed that the use of leader co- framework. This framework has spanned three
ercive power can damage LMX relationships, levels of analysis: the individual (Stage 1), the
whereas reward power can promote them. Part of dyad (Stage 2), and the group (Stage 3). We found
these effects could be due to the different feeling that the sharing of discrete emotions over time,
states created by the use of these two power through such processes as emotional contagion
bases, with coercive power creating resentment and mutual entrainment, can produce stronger or
and reward power creating gratitude. By treating weaker relationships. We also observed that
the application of certain power bases, or other leaders and members should never be compla-
leader behaviors, as affective events, it is likely cent, since even late-stage LMX relationships
that some of this earlier work can be integrated can be disrupted by member emotions resulting
with our current model. Such an eclectic ap- from the perception of unfair LMX differentiation.
proach might allow future researchers to build There is additional work to be done, especially
more comprehensive models of LMX. on such topics as feedback loops and individual
This discussion of situational factors also rai- differences. Fortunately, the theory provided
ses the possibility that some will interact with here provides a framework for exploring these
various features of our model. This is not unlikely, phenomena. We encourage scholars to take