Why and When Leader’s Affective States Influence Employee Upward Voice
Department of Management and Marketing
Faculty of Business
Hong Kong Polytechnic University
Hung Hom, Kowloon, Hong Kong
Phone: (852) 3400-3954
NUS Business School
National University of Singapore
1 Business Link, BIZ 1 Building
Phone: (65) 6516-5739
NUS Business School
National University of Singapore
1 Business Link, BIZ 2 Building
Phone: (65) 6516-5739
NUS Business School
National University of Singapore
1 Business Link, BIZ 2 Building
Phone: (65) 6516-1344
We thank Yuxia Zhang for her help in data collection, and Cynthia Lee, Jian Liang, Xu
Huang, Subrahmaniam Tangirala, and Jason Shaw for their helpful comments on earlier
drafts of this paper. We also thank Brent Scott and three anonymous reviewers for their
helpful guidance and suggestions in the review process. This project was partially supported
by the Singapore Ministry of Education ACRF grant (R-317-000-110-112). An early version
of the paper was presented at the Academy of Management Annual Meeting, Orlando,
Florida, United States of America, 2013.
Why and When Leader’s Affective States Influence Employee Upward Voice
Although researchers have argued that employees often carefully examine social
contexts before speaking up to leaders, the role of leaders’ affective states has received little
attention. The current research addresses this important issue from an emotion-as-social-
information perspective by exploring whether, why, and when leaders’ affect influences
employees’ voice behavior. By collecting data of 640 daily interactions from both sides of 85
leader-employee dyads using the experience sampling method (ESM) through mobile surveys,
we found that leaders’ positive affect was positively related to employees’ voice behavior.
Furthermore, such a relationship could be accounted for through employees’ psychological
safety directly via emotional contagion mechanism (through employees’ own positive affect)
but not directly via signaling mechanism (through employees’ assessment of leaders’ positive
affect), and the effects of both employees’ own positive affect and their assessments of
leaders’ positive affect on psychological safety were stronger when the leader-member
exchange relationship was weak. Interestingly, we also found that leaders’ negative affect
was positively related to employees’ voice, but neither emotional contagion nor signaling
mechanisms explained this effect. These findings highlight the important role of leaders’
affect in the voice process and also provide insights for when employees would choose to
speak up to their leaders.
Keywords: Employee voice, affect, leadership, LMX
In many organizations, employees are uniquely positioned to identify emerging
problems and opportunities that can critically influence the effectiveness of work processes
and outcomes (Edmondson, 2003). In this context, upward voice, or employees’ expression
of constructive work-related ideas to organizational leaders (Detert & Burris, 2007; Tangirala
& Ramanujam, 2012), plays a critical role in linking employees’ private knowledge and
insights with leaders’ organizational influence (Ryan & Oestreich, 1998). To understand what
may promote or discourage employees from speaking up to their leaders, most of the existing
voice research has taken a between-individual approach to explore the antecedents of voice
(Morrison, 2011). These studies have identified employee characteristics, such as self-esteem
(LePine & Van Dyne, 1998) and dispositional factors (e.g., LePine & Van Dyne, 2001), and
leader characteristics, such as openness to change (e.g., Detert & Burris, 2007; Liu, Zhu, &
Yang, 2010) and leadership styles (e.g., Fast, Burris & Bartel, 2014; Tangirala & Ramanujam,
2012; Walumbwa & Schaubroeck, 2009), as important predictors of voice.
Although this line of research has provided valuable insights into the relatively stable
antecedents of voice, the phenomenon that both employees and leaders can behave differently
in the moment has almost been neglected. In reality, even when facing the same leader, an
employee may be more likely to speak up to the leader in some interaction episodes than in
other episodes (Detert & Edmondson, 2011; Detert & Treviño, 2010). Some prior discussions,
for example, have suggested that employees “read the wind” to discern whether a particular
situation is favorable to sharing their suggestions, opinions, or concerns with leaders (Dutton,
Ashford, Wierba, O'Neill, & Hayes, 1997; Milliken, Morrison, & Hewlin, 2003). However,
little attention has been paid to the possible fluctuation of employee voice from one episode
to another. As a consequence, we still know little about whether more dynamic, fluctuating
leader-relevant factors, such as leaders’ affect, influence employee upward voice (Morrison,
Addressing this fluctuation in employee voice as dependent on leader-relevant factors
is important for both practical and theoretical reasons. Practically, voice contributes to
organizational effectiveness (e.g., Detert, Burris, Harrison, & Martin, 2013), and leaders’
affective states have been argued to play a critical role in shaping employee behaviors (e.g.,
Van Kleef, Homan, Beersma, van Knippenberg, van Knippenberg, & Damen, 2009).
Theoretically, recent studies in social psychology have suggested that individuals, especially
those with less power, pay particular attention to the affect of others in order to behave
appropriately in social interactions (e.g., Melwani & Barsade, 2011; Van Kleef, De Dreu, &
Manstead, 2004, 2010). We expect that during interactions between employees and leaders,
leaders’ affect may importantly influence employee voice (Gooty, Connelly, Griffith, &
Gupta, 2010). Therefore, the first purpose of this research is to examine whether leaders’
affect influences voice by taking a within-individual or episodic approach. Accordingly, we
conceptualize voice as an episodic, social-interactional process between leaders and
employees, in which employees share constructive suggestions, ideas, and concerns with
leaders (c.f. Morrison, 2011).
We draw on the emotion-as-social-information (EASI) model (Van Kleef et al., 2009;
Van Kleef et al., 2010), which contends that individuals’ emotions influence others via two
distinct mechanisms. One mechanism is the emotional contagion process, by which leaders’
affect implicitly evokes employees’ affect and then influences employees’ consequent
attitudes and behaviors (Hatfield, Cacioppo, & Rapson, 1994), while the other is the
signaling process, by which leaders’ affect is cognitively assessed by employees and then
influences employees’ attitudes and behaviors (Van Kleef et al., 2009). More recently,
scholars have also argued that affect may influence psychological safety, defined in this
context as employees’ belief that they can show and express themselves to leaders without
fear of negative consequences during interactions with leaders (Kahn, 1990; Kish-Gephart,
Detert, Treviño, & Edmondson, 2009; Liang, Farh, & Farh, 2012) and thus voice behavior
(Detert & Burris, 2007; Liang et al., 2012). Therefore, the second purpose of this research is
to contribute to the voice and leadership literature by integrating the EASI model and
psychological safety studies to explain why leaders’ affect influences voice. More
specifically, we examine the indirect effect of leaders’ affect on employee voice through
psychological safety via both emotional contagion and signaling processes.
Another critical issue is identifying when leaders’ affect is more likely to influence
employees. Gooty and colleagues (2010), in a recent review of emotion research in the
leadership literature, suggest that we still know little about the contextual contingencies for
the effects of leaders’ affect. Answering this call, we draw on the EASI perspective (Van
Kleef et al., 2010) to examine when leaders’ affect has stronger or weaker influences on
voice. Specifically, this perspective argues that the social functions of affect vary contingent
upon the relationships between interaction parties (Van Kleef et al., 2004, 2009, 2010).
Applying this tenet, we propose that the strengths of both emotional contagion and signaling
mechanisms depend on the quality of the leader-member exchange (LMX)—the quality of
the social exchange relationship between leaders and employees (Graen & Uhl-Bien, 1995),
such that the paths would be stronger when LMX quality is weaker.
In summary, the current study presents a multilevel framework to investigate whether,
why, and when leaders’ affect influences employees’ voice behavior. Our study extends the
current voice and leadership literature in four unique ways. First, most previous voice
research has taken a between-individual approach by focusing on the stable characteristics of
leaders or employees, but it cannot explain all of the variance of voice (Morrison, 2011). By
taking a within-individual approach to investigate voice at the episodic level, our study not
only captures the hitherto missing within-individual variance of employee voice, but also
advances prior voice research by examining new antecedents of voice at the episodic level.
Second, we examine both the positive and negative affect of leaders as critical
antecedents of voice. This investigation not only extends the emerging conceptual discussions
that focus primarily on how employees’ own affect may lead to voice (Kish-Gephart et al.,
2009; Harvey, Martinko, & Douglas, 2009), but also enlarges the scope of recent research on
positive mood and voice (e.g., Liu, Tangirala, Lam, Chen, Jia, & Huang, 2015) by explicitly
scrutinizing the effects of both the positive and negative affect of leaders. Our study thus
offers a timely response to the recent call for exploring the connection between affect and
voice (Morrison, 2011).
Third, by integrating the EASI model (Van Kleef et al., 2010) with research on
psychological safety (Liang et al., 2012; c.f. Edmondson, 1999) and LMX (Graen & Uhl-
Bien, 1995), we help explain why and when leaders’ affect influences voice. Our study
extends existing voice research (e.g., Detert & Burris, 2007; Liang et al., 2012) by identifying
leaders’ affect as a driving force of employee psychological safety and also contributes to the
existing emotion research (e.g., Van Kleef et al., 2009) by identifying psychological safety as
a consequence of leaders’ affect via both emotional contagion and signaling mechanisms.
Moreover, our study advances prior voice research, which has addressed only the main effect
of LMX on voice (Burris, Detert, & Chiaburu, 2008; Van Dyne, Kamdar, & Joireman, 2008),
by theorizing that LMX, in conjunction with leaders’ affect, influences upward voice.
Finally, to unpack the interpersonal dynamics between leaders and employees in the
voice process, we employed the event-contingent version of the experience sampling method
(ESM, Wheeler & Reis, 1991) to collect field data on immediate interactional episodes from
both sides of leader-employee dyads in real work settings. Our research design sheds light on
how to resolve concerns about external validity that many experimental studies in the
emotion research have encountered (Chatman & Flynn, 2005), as well as how to minimize
the common method bias that most ESM studies have faced (Bolger, Davis, & Rafaeli, 2003;
Podsakoff, MacKenzie, Lee, & Podsakoff, 2003).
Figure 1 summarizes the hypothesized relationships proposed in this study. We tested
these hypotheses in a field study using multiple-source, experience sampling data from
leader-member interactions for two weeks in five information technology (IT) companies.
Theoretical Development and Hypotheses
Voice and the Emotion-as-Social-Information Model
Voice aims to challenge the status quo in organizations and is thus risky (Van Dyne,
Cummings, & McLean Parks, 1995). Leaders often perceive voice as threats and thus respond
negatively to employees who speak up (Burris, 2012; Milliken et al., 2003; Morrison &
Milliken, 2000). Given the risky nature of voice, employees would carefully evaluate the
social contexts before speaking up (Ashford et al., 1998; Liu et al., 2015). Therefore, it is
particularly meaningful to examine how leaders’ affective states influence voice, because
voice in essence is a form of social interaction between employees and leaders (Dutton &
Ashford, 1993; Morrison & Milliken, 2000) and leaders’ affective states provide important
social information that influences employee behaviors during leader-member interactions
(Gooty et al., 2010; Van Kleef et al., 2009).
Affect plays an important role in everyday life, and it is not only often evoked by
social interactions, but also serves as a form of communication that influences the behavior of
others in social interactions (Fridlund, 1994; Frijda, 1986; Parkinson, 1996; Van Kleef, De
Dreu, & Manstead, 2010). An individual's display of positive affect, for example, signals
security, openness, or an intention of affiliation to others; by contrast, an individual's display
of negative affect signals a threat, a fixed mind, or an intention of distance to others (Forgas,
1985; Forgas & George, 2001). When others unconsciously capture or consciously make
inferences from a person's displays of affect, it can influence their attitudes or behaviors (Van
Kleef et al., 2009).
Synthesizing and extending this emerging literature on the social functions of affect,
Van Kleef and colleagues (Van Kleef, 2009; Van Kleef et al., 2009, 2010) propose the EASI
model. One premise of the EASI model is that affect serves critical interpersonal functions in
social interactions (Keltner & Haidt, 1999; Manstead, 1991; Oatley & Johnson-Laird, 1987;
Van Kleef, 2009), because affect conveys information to actors about the target’s current
feelings, social intentions, and orientation toward the relationship (Ames & Johar, 2009; Van
Kleef et al., 2004). Moreover, according to this model, an individual’s affect influences
others in two distinct ways: emotional contagion and signaling processes (Van Kleef et al.,
In the following, we apply the EASI model to theorize the mechanisms by which
leaders’ affect influences employees’ voice. We investigate both the positive and negative
affect of leaders. Positive and negative affect is quite different in their characteristics, social
meanings, and functions (e.g., Fredrickson, 1998; Forgas & George, 2001; Van Kleef et al.,
2010). In addition, the states comprising negative affect are more differentiated than the
states comprising positive affect (e.g., de Rivera et al., 1989; Ellsworth & Smith, 1988;
Fredrickson, 1998). Therefore, in the following we first propose hypotheses on leaders’
positive affect and then raise several research questions related to leaders’ negative affect.
Leaders’ Positive Affect and the Emotional Contagion Process
A person may unintentionally and automatically “catch” others’ emotions, and this
process is referred to as emotional contagion (Hatfield et al., 1994). Emotional contagion
takes place when a person unconsciously mimics another individual’s emotions and assumes
that an individual’s emotions are a consequence of facial, vocal, or gestural communications
(Anderson, Keltner, & John, 2003; Kelly & Barsade, 2001). Both laboratory (e.g., Sy, Côté,
& Saavedra, 2005) and field (e.g., Barger & Grandey, 2006; Song, Foo, & Uy, 2008;
Totterdell, Kellet, Teuchmann, & Briner, 1998) studies have demonstrated the existence of an
emotional contagion process. Moreover, emotions are more likely to be transferred from
high-power individuals to low-power ones, rather than the other way around (Anderson et al.,
2003). That is, low-power individuals, who are more dependent on high-power individuals
than the reverse, are more attentive to and are more likely to mimic the emotions of high-
power individuals (Anderson & Thompson, 2004; Van Kleef et al., 2004). In the context of
leader-member interactions, for example, Sy et al. (2005) found that when leaders were in a
positive mood (rather than a negative mood), their group members had more positive
experiences. Based on the above discussions, we predict that leaders’ positive affect transfers
to employees during leader-employee interactions and that employees experience positive
affect as a result of automatic mimicking and unconscious learning of leaders’ positive tones,
gestures, or facial expressions (Kelly & Barsade, 2001; Sy et al., 2005).
We further propose that leaders’ positive affect influences employee psychological
safety through employees’ own positive affect. That is, when a leader’s positive affect evokes
an employee’s positive affect, the employee, in turn, will be more likely to feel safe when
interacting with the leader. Psychological safety reflects the extent to which employees
believe that they can safely express themselves to leaders without fear of negative
consequences (Kahn, 1990; Kish-Gephart, Detert, Treviño, & Edmondson, 2009; Liang, Farh,
& Farh, 2012). An individual’s own affect provides him or her with information about
situations, and such information further influences cognitive processes and behavior
(Schwarz & Clore, 1983, 2003). Research has shown that an individual often attunes his or
her thought processes and behaviors to the information provided by his or her own affect in
order to function and adapt to an environment effectively (Schwarz, 2002). Positive affect
basically informs people that an environment is safe and things are going well (Clore, Gaspar,
& Garvin, 2001; Schwarz & Clore, 1983, 2003). Supporting this argument, research has
shown that when people are having positive feeling states, they perceive relatively neutral
consumer products more positively (Isen, Shalker, Clark, & Karp, 1978), perceive others
more positively (Forgas & Bower, 1987), and believe favorable outcomes are more likely
(Erez & Isen, 2002; Rottenstreich & Hsee, 2001) than people who are not experiencing
positive affective states. Based on these discussions, we propose that employees’ own
positive affect, evoked through contagion by leaders’ positive affect, help employees feel
psychologically safe during interactions with leaders.
Hypothesis 1: In an interactional episode, an employee’s positive affect mediates the
positive relationship between the leader’s positive affect and the employee’s
Psychological safety is a core psychological mechanism that drives employees to
speak out (Edmondson, 1999; Morrison & Milliken, 2000). As Milliken et al.’s (2003)
qualitative study documented, employees usually are afraid to convey negative, challenging,
or unpopular information to colleagues, because they expect negative consequences
associated with voice. When they feel safe enough, they are more likely to share their
opinions, suggestions, and concerns freely (Detert & Burris, 2007; Liang et al., 2012;
Nembhard & Edmondson, 2006). When interacting with leaders, employees who experience
psychological safety thus are more likely to express their opinions, suggestions, and concerns.
Hypothesis 2: In an interactional episode, an employee’s positive affect and
psychological safety sequentially mediate the positive relationship between the
leader’s positive affect and the employee’s upward voice (i.e., leader’s positive affect
employee’s positive affect
Leader’s Positive Affect and the Signaling Process
Another mechanism, according to the EASI model, is a signaling process. Affect
conveys meaningful information to a social interaction partner about an individual’s current
feelings, social intentions, and orientation toward the relationship (Ames & Johar, 2009; Van
Kleef et al., 2004). In response, the partner consciously makes judgments or takes follow-up
actions based on the information inferred from the individual’s emotions (Filipowicz,
Barsade, & Melwani, 2011; Miron-Spektor, Efrat-Treister, Rafaeli, & Schwarz-Cohen, 2011;
Van Kleef et al., 2009). Previous research has provided evidence that signaling process is
distinctive from the emotional contagion process (Eberly & Fong, 2013; Van Kleef et al.,
2009).For example, in a lab setting, Van Kleef and colleagues (2009) manipulated leader’s
emotions and showed that teams with high epistemic motivation were more influenced by the
signaling process whereas those with low epistemic motivation were more influenced by the
emotional contagion process.
In the context of voice, to avoid leaders’ misunderstanding or confusion, employees
often closely monitor leaders’ affective states to assess whether the context is favorable for
voice (c.f. Ashford, Rothbard, Piderit, & Dutton, 1998; Dutton et al., 1997). The initial step
of cognitively processing the meaning of leaders’ affect consists of an employee recognizing
such affect (Elfenbein, 2007). In other words, employees need to be consciously aware of
leaders’ affective states before they can make sense of them (Cropanzano, Weiss, Hale, &
Reb, 2003; Elfenbein, 2007; Lazarus, 1991). In interactions between a leader and an
employee, the employee may assess the leader’s affect by vocal tone, facial expressions, or
gestures (Ambady & Rosenthal, 1992; Gooty et al., 2010). When a leader displays a positive
affect, an employee is likely to infer that the leader is happy, excited, or pleased, and such an
inference would further influence this employee’s subsequent attitudes and behaviors (Van
Kleef et al., 2009).
Therefore, we argue that in addition to the emotional contagion pathway, leaders’
positive affect also influences employees’ psychological safety through a signaling pathway.
That is, when an employee assesses a leader’s positive affect, the employee, in turn, is more
likely to experience psychological safety. Previous research has suggested that people tend to
pay selective attention to mood-consistent details (Bower, 1981; Forgas & Bower, 1987;
Forgas & George, 2001). For example, given that they are observing actors engaging in
identical behaviors, people in a positive mood selectively look for lenient and optimistic
explanations, while those in a less positive mood tend to make more critical attributions
(Forgas, 1998; Forgas, Bower, & Moylan, 1990). From the employee’s perspective, the
employee would expect the leader’s affect to largely influence how the leader will respond to
voice. The leader would be expected to pay attention to the positive aspects of voice (e.g.,
that the employee is making constructive suggestions) when the leader is experiencing
positive affect, but more to the negative aspects (e.g., that the employee is showing off or
trying to embarrass the leader) when the leader is in a less positive affective state.
In addition, positive affect enables individuals to be flexible and open to new ideas
(Fredrickson, 2001). The employee would expect the leader to be more likely to accept and
take actions on voice when the leader is in a more positive affective state. Indeed, Ames and
Johar (2009) found that compared with targets displaying negative emotions, targets
displaying positive emotions are more likely to be ascribed to have prosocial intentions. In
addition, Gino and Schweitzer (2008) reported that compared with people who felt angry,
people who felt gratitude were more receptive to advice. Another recent research by Liu et al.
(2015) further suggests that a target member’s positive mood was positively related to a focal
member’s psychological safety with this target member. Based on the discussions above
about the signaling pathway and the effect of psychological safety on voice, we propose the
Hypothesis 3: In an interactional episode, an employee’s assessment of the positive
affect of the leader mediates the positive relationship between the leader’s positive
affect and the employee’s psychological safety.
Hypothesis 4: In an interactional episode, an employee’s assessment of the positive
affect of the leader and the employee’s psychological safety sequentially mediates the
positive relationship between leader’s positive affect and the employee’s upward
voice (i.e., leader’s positive affect
employee’s assessment of leader’s positive affect
Moderating Role of LMX Quality
According to the EASI model, the interpersonal effects of emotions depend on the
relations between the actor and the partner, because the nature of relations fundamentally
determines the meaning and social consequences of emotional expressions (Van Kleef et al.,
2009, 2010). A smile, for example, is likely to signify warmth to a friend, but disdain to an
enemy. Therefore, we further propose that although both emotional contagion and signaling
processes are important mechanisms by which leaders’ positive affect influences employees’
psychological safety and thereby upward voice, the strength of each path may vary across
different leader-employee dyads, contingent upon LMX quality.
Leaders may develop relationships with a variety of employees with different
qualities (Dansereau, Graen, & Haga, 1975; Gerstner & Day, 1997). Low-quality LMX
relationships are characterized by economic exchange and feature low levels of trust, support,
commitment, and loyalty (Cropanzano & Mitchell, 2005; Uhl-Bien & Maslyn, 2003). By
contrast, high-quality LMX relationships increasingly engender feelings of mutual obligation
and reciprocity (Liden, Sparrowe, & Wayne, 1997). Such high-quality relationships result in
increased affective attachments between leaders and followers, with such key features as trust,
support, commitment, and loyalty (Dulebohn, Bommer, Liden, Brouer, & Ferris, 2012). We
argue that LMX quality would respectively moderate the indirect effect of employees’ own
positive affect (the emotional contagion pathway) and the indirect effect of their assessment
of leaders’ positive affect (the signaling pathway) on voice via psychological safety.
For the emotional contagion pathway, the EASI theory suggests that the extent to
which an actor’s emotional reactions would further influence his or her following attitudes or
behaviors depends on the relational context (Van Kleef et al., 2010). In relationships where
informational cues are already stored and available for judgment, during interactions, actors
use a direct processing strategy without giving much consideration to their own affective
states in making evaluations; but in relationships involving complicated, unusual targets that
mandate more elaborate processing, actors rely more upon their affective states to make
judgments (e.g., Dunn & Schweitzer, 2005; Forgas, 1995; Van Kleef et al., 2009). Having a
prototype, for example, gives people structuralized and simplified information about others,
and therefore, when actors evaluate others consistent with a prototype, they are less likely to
use their own affect to make judgments; but when they encounter those not consistent with a
prototype, they engage in substantive processing by utilizing their own affect (Forgas, 1992).
Similarly, Dunn and Schweitzer (2005), in a serial of experimental studies, found that when a
truster had little history with a trustee, the truster’s trust judgments were heavily influenced
by the truster’s own affective states. By contrast, when the truster was familiar with the
trustee, the truster’s affective states had little influence on his or her trust judgments.
We argue that the quality of LMX influences the strategies by which employees use
their own affect in judging psychological safety before speaking up to leaders. Specifically,
when LMX quality is high, employees already regard their interactions with leaders as
mutually beneficial, trustful, and safe (e.g., Dulebohn et al., 2012; Liden et al., 1997). As a
consequence, employees naturally feel psychologically safe in speaking up to their leaders
without needing to consider their own affect in making safety evaluations. That is, employees
would use a direct strategy in this situation, and therefore, the effects of employees’ own
positive affect on psychological safety and, thus, on voice would be weak.
By contrast, when LMX quality is low, the work relationships between employees and
leaders feature low levels of affective attachment, trust, and support (Cropanzano & Mitchell,
2005; Uhl-Bien & Maslyn, 2003). Employees encounter relatively risky and unpredictable
situations when they interact with low-relationship-quality leaders, especially when judging
whether the situation is safe for speaking up, and thus these employees are motivated to use
auxiliary cues, such as their own affective states, to evaluate whether it is safe to voice. On
the basis of these discussions, we propose a “first-stage” moderated mediation (Edwards &
Lambert, 2007): The indirect effects of employees’ positive affect on voice via their
psychological safety are stronger when LMX quality is low rather than high.
Hypothesis 5: LMX quality moderates the indirect relationship between an
employee’s positive affect and upward voice (via the employee’s psychological safety),
such that this indirect relationship is stronger when LMX quality is low rather than
The EASI theory also proposes that people are more likely to monitor others’
affective states to make social judgments when they have a low level of trust in others (Van
Kleef et al., 2004, 2009). Under such a condition, an individual is motivated to expend effort
to systematically process social information, such as emotions displayed by the interaction
target, in order to make appropriate decisions, judgments, and behavioral strategies (Van
Kleef et al., 2010). Recent research has suggested that a negotiator deliberately analyzes the
negotiation partner’s affect in order to develop effective strategies, especially when the two
negotiators are competing rather than collaborating with each other (e.g., Van Kleef et al.,
We argue that in the workplace, the quality of LMX influences the extent to which
employees use leaders’ affect when judging psychological safety. Specifically, when LMX
quality is low, the interpersonal risks associated with voice are expected to be high (Burris et
al., 2008; Van Dyne et al., 2008). That is, employees would tend to have greater concerns that
their voice, however constructive, might be more negatively construed as a veiled criticism or
complaint by a partner with whom they do not get along than by a partner with whom they
have a positive relationship. Hence, when interacting with a leader with low LMX quality,
employees might pay more attention to the leader’s affect to discern whether the situation is
favorable for speaking up. As a consequence, employees’ assessments of leaders’ positive
affect would have a strong effect on psychological safety and, thus, on voice.
By contrast, when LMX quality is high, we expect employees to pay less attention to
transient cues, such as the partner’s affect, because their relationship is already based on trust
and psychological safety (e.g., Dulebohn et al., 2012; Liden et al., 1997). Under such
conditions, employees should care less about “reading the wind” when speaking up and, thus,
are less influenced by their assessments of leaders’ positive affect. Applying a similar logic,
Liu, Tangirala, Lam, and colleagues (2015) proposed and found that in the context of teams,
a target member’s positive mood was positively related to a focal member’s psychological
safety with this target member, especially when the relationship quality between the two
members was low rather than high. Based on the above discussions, we propose a “first-
stage” moderated mediation model, where the effect of the assessment of leaders’ positive
affect on voice (via psychological safety perceptions) is stronger when LMX quality is low
rather than high.
Hypothesis 6: LMX quality moderates the indirect relationship between an
employee’s assessment of the leader’s positive affect and upward voice (via the
employee’s psychological safety), such that this indirect relationship is stronger when
LMX quality is low rather than high.
Role of Negative Affect
The above discussions have highlighted the role of positive affect during leader-
member interactions, but prior research have also suggested that negative affect may play an
important role in the context of voice (e.g., Edwards, Ashkanasy, & Gardner, 2009; Harvey et
al., 2009; Kish-Gephart et al., 2009; Milliken et al., 2003). For example, in their qualitative
interviews with employees, Milliken et al. (2003) reported that employees usually fear
speaking up to their managers. Kish-Gephart et al. (2009) further theoretically elaborated the
origin of fears associated with voice and suggested that such fears arise from deeply rooted
emotions that are evolutionary based and further reinforced by socialization and habituation.
Harvey et al. (2009) and Edwards et al. (2009), from the perspective of the observers of
wrong-doing, theorized that experiences of anger and resentment may drive employees to
blow the whistle in organizations. Therefore, it is meaningful to discuss how leaders’
negative affect may influence employee voice.
First, we propose that leaders’ negative affect may be contagious to employees; that is,
employees tend to experience negative affect when their leaders display negative affect.
Previous research has found that negative affective states, such as stress and burnout, can be
transferred among friends, couples, and colleagues (e.g., Gump & Kulik, 1997; Van Kleef et
al., 2009; Westman, Vinokur, Hamilton, & Roziner, 2004). However, the effect of
employees’ negative affect on psychological safety or voice may be not as clear as that of
positive affect. One important reason is that the meanings of negative affect are more
diversified and differentiated than those of positive affect (de Rivera et al., 1989; Ellsworth &
Smith, 1988; Fredrickson, 1998). An employee who is experiencing fear, for example, may
not dare to speak up (Kish-Gephart et al., 2009; Milliken et al., 2003), while an employee
who is experiencing anger, another type of negative affect, may take the risk to stick out and
speak up (e.g., Edwards et al., 2009; Harvey et al., 2009).
Second, we also contend that when leaders display negative affect, employees are
likely to recognize and assess the negative affect of leaders. However, the effect of
employees’ assessments of leaders’ negative affect on psychological safety or voice may be
not as clear as that of the assessments of leaders’ positive affect. When an employee
perceives that a leader is angry, for example, the leader’s anger may trigger unsafe feelings
within the employee, thus preventing voice (Milliken et al., 2003); but such anger may also
signal the leader’s dissatisfaction with the status quo, thus prompting the employee to speak
up (Van Kleef et al., 2009).
In summary, the states comprising negative affect are more differentiated than those
comprising positive affect. In addition, some negative affect may have opposite effects on
psychological safety and voice, and we are lack of concrete research or theory in making
precise predictions. As a result, in this study we chose to explore how leaders’ negative affect
may influence employee voice in an open manner, rather than proposing specific hypotheses.
We are interested in the following research questions:
Research question 1: In an interactional episode, do leader’s negative affect states
influence employee voice?
Research question 2: If the answer to RQ 1 is yes, then is such an effect explained by
psychological safety via the emotional contagion mechanism (i.e., via the employee’s
Research questions 3: If the answer to RQ 1 is yes, then is such an effect explained by
psychological safety via the signaling mechanism (i.e., via the employee’s assessment
of leader’s negative affect)?
We collected data from five small- and medium-sized enterprises in the IT industry in
China. We contacted a total of 45 middle-level managers and 135 employees to participate in
the current study. To qualify for the study, leaders had to have at least three employees
reporting directly to them. If those leaders directly supervised more than three subordinates,
we randomly selected three of their employees to participate.
All participants were invited to briefing sessions, in which the purpose, content, and
procedures of the study were communicated. Participation was completely voluntary and
confidentiality was assured. During the briefing, every participant was trained on a one-to-
one basis on how to use the mobile survey system through which we collected interaction
data (described in the next paragraph); toward the end of the briefings, they finished a
baseline survey, which collected demographic information and information on control
We used an event-contingent version of the experience sampling method (Wheeler &
Reis, 1991) to collect interaction data from leaders and employees. That is, both a leader and
an employee were asked to fill out mobile surveys only when an interaction that met our pre-
established standard (i.e., an event) occurred (Wheeler & Reis, 1991). We employed the
mobile survey technique (MST, Li & Townsend, 2008; Song et al., 2008) to trace leaders’
and employees’ immediate interaction experiences in real work settings. A mobile survey
refers to survey research using electronic questionnaires based on a mobile platform. We used
J2ME and WAP as two alternative ways to collect data (Li & Townsend, 2008). Specifically,
J2ME provides a robust, flexible environment for applications running on nearly all types of
mobile devices (including low-end cell phones), such as electronic questionnaires, while
WAP is a standardized protocol that enables mobile devices (smart or PDA phones) to access
web-based information. Combining these two methods, most mobile devices on the market
can be “equipped” as data-collection tools. To conduct the mobile survey, we programmed
electronic questionnaires based on J2ME and WAP and guided participants on how to
complete these questionnaires through their mobile phones during briefing sessions. One day
before the formal data collection, we also ran a simulation session to make sure that
participants understood the protocol and could correctly submit their mobile surveys through
For a period of 10 working days over two weeks’ time (including extra working
hours), participants were required to respond to the mobile survey within one hour after each
interaction with their leaders or employees (c.f., Bolger et al., 2003; Laurenceau, Feldman, &
Pietromonaco, 1998). We specifically defined “interaction” in our study as a “face-to-face
conversation between leaders and their immediate employees”
that lasts for more than two
minutes. If participants were answering a J2ME-based questionnaire, their responses were
sent back to researchers via short message service (SMS); if participants were answering a
WAP-based questionnaire, their responses were submitted through mobile network (e.g.,
GPRS and 3G) to an online database. J2ME- and WAP-based questionnaires had similar
formats, which allowed us to combine the data later for analysis. Responses were time-
stamped, allowing for accurate recording of the time that the responses were received. We
then matched responses from leaders and employees regarding the same interaction event.
To facilitate data collection and increase the response rate, we sent two types of SMS
reminders to participants. The first type was a “general reminder,” which was sent out to
every participant at 9:00 a.m. (the normal beginning of working hours in the morning) and
1:30 p.m. (the normal beginning of working hours in the afternoon) on each working day. A
sample message is, “Good morning (afternoon), please do not forget to answer the survey
The interaction, as defined in this study, excludes non-face-to-face communication via phones, e-mails,
teleconferences, or others. It also excludes interactions that members had with leaders other than their direct
after interacting with your supervisor (employee). Thank you and have a pleasant day!” The
second type was a “conditional reminder,” which was triggered when a leader (or an
employee) submitted a mobile survey but the employee (leader) did not. Two research
assistants monitored the system from 8:30 a.m. to 9:30 p.m. on each working day. They
checked the system every 30 minutes and sent “conditional reminders” to corresponding
participants once responses from either a leader or an employee showed up in the system. A
sample message is, “Hi, please do not forget to submit your response for the interaction you
just had with your supervisor (employee).” In addition to SMS reminders, participants also
were encouraged to contact researchers via e-mail and telephone for instructions and help.
Since participants were the ones who initiated the mobile surveys, we provided cash
and lotteries as incentives to motivate them to report each real interaction. In particular, for
each pair of valid mobile survey responses, a leader and an employee each received 10 RMB
(approximately 1.58 U.S. dollars). No upper limit was placed on the number of mobile survey
responses. At the end of the study, leaders and employees were entered into a random
drawing in which they had a chance to win an iPod-touch player as a reward. Participants
also completed a short reflection survey at the end of the whole study.
Among the 45 leaders and 135 employees, 9 leaders (20%) decided to drop out in the
middle of data collection for reasons such as fatigue or busy work; accordingly, 27
employees who reported to these 9 leaders were dropped due to unmatched data. The
remaining 36 leaders and 109 employees submitted a total of 1,849 mobile surveys, and
1,468 (79%) from 36 leaders and 96 employees were successfully matched (i.e., 734 episodes
with paired surveys). We further cleaned the paired mobile survey data by deleting (a) 87
episodes with surveys submitted more than one hour after the interaction took place in order
to reduce retrospection bias (e.g., Ilies et al., 2007; Laurenceau et al., 1998), and (b) 7
episodes with missing data on core variables. After the data cleaning, 640 paired responses
from 36 leaders (80%) and 85 employees (63%) remained, resulting in an average of 7.53
paired responses per dyad.
In the reflection survey, we asked leaders and employees to provide their estimation of
how many interactions they had every day over the past two weeks. Then, we estimated the
response rate as the actual number of interactions of each dyad received divided by the
average of leaders’ and employees’ self-reported numbers of interactions. Overall, the mobile
survey captured about 30% of the total number of interactions that had taken place during the
In the final sample, 22.8% of the leaders were female; their ages ranged from 23 to 47,
with an average of 32.5; 88.6% had received college education or above; and their average
organizational tenure was 49 months. For the employees, 51.8% were female; their ages
ranged from 19 to 38, with an average of 28.2; 81.5% had received college education or
above; and their average organizational tenure was 25.4 months. On average, leaders and
employees submitted the mobile survey 32 minutes after the interaction, and an employee (or
leader) responded to the mobile survey 22 minutes after the other party submitted the
response. In addition, 15% of participants responded through the J2ME questionnaire and
85% responded through the WAP questionnaire. T-tests showed that there were no significant
differences between the data collected by J2ME- and by WAP-based questionnaires.
Positive affect. Leaders reported their positive affect in the mobile surveys, rating the
extent to which they displayed specific types of affective states during their interactions with
T-tests confirmed that employees who were eliminated did not differ significantly from those in the final
sample along demographic dimensions or LMX (t = 1.72, p > .05). Moreover, although the eliminated paired
mobile surveys had significantly lower scores on leader positive affect (t = 7.61, p < .001), employee positive
affect (t = 8.52, p < .001), employee perceived leader’s positive affect (t = 8.15, p < .001), psychological safety
(t = 4.75, p < .001), and voice (t = 3.73, p < .001) than did those in the final sample, including these data in the
regressions did not substantially change results. To ensure data quality, we decided not to include them in the
employees on a five-point scale (from 1 = “to a small extent” to 5 = “to a large extent”). As
participants needed to report every interaction they had, a lengthy survey would have been
demanding. For this reason, we used four items (delighted, excited, happy, and joyful) that
reflected both positive valence and high activation from the positive affect scale of Tellegen,
Watson, and Clark (1999) to represent leaders’ positive affect (α = .93).
Employees reported their positive affect in mobile surveys, rating the extent to which
they experienced specific types of affective states during each interaction with leaders on the
same four-item five-point scale (α = .95). Employees also reported their assessments of their
leaders’ positive affect in the mobile surveys, rating the extent to which they perceived
leaders displaying specific types of affect during the interaction on the same four-item five-
point scale (α = .95).
Negative affect. Leaders reported their negative affective states in the mobile surveys,
rating the extent to which they displayed specific types of states (distressed, angry, sad, and
afraid)—four items from Tellegen et al.’s negative affect scale (1999, α = .78), during each
interaction with employees on a five-point scale (from 1 = “to a small extent” to 5 = “to a
Employees reported their negative affect in the mobile surveys, rating the extent to
which they experienced specific types of affect during each interaction with leaders on the
same four-item five-point scale (α = .83). Employees also reported the extent to which they
perceived leaders’ negative affect during the interaction (1 = “to a small extent” to 5 = “to a
large extent”), using the same scale (α = .87).
Employees’ psychological safety. Employees reported the extent of their
psychological safety during each interaction with leaders in the mobile survey using three
positively described items adapted from the scale proposed by Liang et al. (2012). A sample
item was: “In the interaction with the leader just now, I feel that expressing my true opinions
is welcomed by this leader” (from 1 = “strongly disagree” to 5 = “strongly agree,” α = .90).
Employees’ upward voice. Leaders reported employees’ upward voice during the
interaction in the mobile survey. Following Morrison’s (2011) suggestion, we selected three
items from Van Dyne and LePine (1998) and from Liang et al.’s (2012) voice scale by
focusing on suggestion, opinion, and concern, respectively. Sample items were: “In the
interaction with me just now, this employee (1) gave me constructive suggestions regarding
work-related issues, (2) expressed his/her opinions to me, which are different from mine, and,
(3) pointed out problems in our work or company” (from 1 = “strongly disagree” to 5 =
“strongly agree,” α = .79).
Leader-member exchange quality. Employees reported LMX quality in the baseline
survey using the seven-item leader-member exchange scale suggested by Graen and Uhl-Bien
(1995). A sample question was: “How would you characterize your working relationship with
your leader?” (1 = “extremely ineffective” to 5 = “extremely effective,” α = .85).
Control variables. To exclude alternative explanations, we controlled variables that
could be related to voice and affective experience. First, interaction quality may influence
both leaders’ and employees’ affect, so we created a 3-item scale and had leaders report the
interaction quality of each episode. A sample item was: “This interaction was effective”
(from 1 = “strongly disagree” to 5 = “strongly agree,” α = .92). Second, as employees may be
more likely to engage proactive behaviors when they show initiative (Frese & Fay, 2001), we
controlled who initiated the interaction in each episode (the leader, the employee, or a third
party). We used two dummy variables to code these three choices (“initiated by the leader,” 1
= “yes,” 0 = “no”; “initiated by the employee,” 1 = “yes,” 0 = “no”).
Third, we controlled for dyadic tenure (month) between leaders and employees and
also the estimated the interaction time with the leader during the survey period, which
employees reported at the end of our study, because we wanted to exclude the possibility that
employees spoke up to their leaders merely due to having had more opportunities to approach
their leaders during the survey period. Fourth, employees’ proactive personality (e.g., Detert
& Burris, 2007) and employees’ positive affectivity and negative affectivity (e.g., Grant,
Parker, & Collins, 2009; Tangirala & Ramanujam, 2008) have been identified as personality
factors associated with voice and affective states, so we used Seibert, Crant, and Kraimer’s
(1999) 10-item proactive personality scale to measure proactive personality (from 1 =
“strongly disagree” to 5 = “strongly agree,” α = .79) and Watson, Clark, and Tellegen’s (1988)
20-item affectivity scale (from 1 = “to a small extent” to 5 = “to a large extent”) to measure
positive (α = .86) and negative (α = .87) affectivity. Further, research has shown that
individual differences in susceptibility to emotional contagion influence affective transfer in
the workplace (Ilies, Wagner, & Morgeson, 2007), so we measured this variable using
Doherty’s (1997) emotional contagion scale (from 1 = “strongly disagree” to 5 = “strongly
agree,” α = .78). In addition, individuals differ in their ability to recognize others’ emotions
(cf., Mayer, Roberts, & Barsade, 2008). Hence, we controlled for employees’ emotional
appraisal ability using Wong and Law’s (2002) sub-scale of emotional intelligence.
Finally, as transformational leadership style may influence employees’ emotional
experience (Bono, Foldes, Vinson, & Muros, 2007) and voice (Detert & Burris, 2007), we
asked each employee to report the extent of their leader’s transformational leadership using
the scale from MLQ 5X (Avolio, Bass, & Jung, 1999, α = .96). Statistical tests revealed that it
was appropriate to aggregate this scale to the leader level (the median of Rwg = .95, ICC
= .20, ICC = .42, F (1, 35) = 1.73, p < .05), so we conducted aggregation and used it as a
leader-level variable in our analyses. Moreover, as leaders’ positive and negative affectivity
may also influence leaders’ affective states and employees’ perceptions (Rubin, Munz, &
Bommer, 2005), we asked leaders to report positive and negative affectivity using Watson et
al.’s (1988) positive (α = .70) and negative (α = .87) affectivity scale.
We first conducted confirmatory factor analyses (CFA) to confirm the discriminant
validity of our measures. Next, we checked variances of episode-level variables (e.g., voice
and psychological safety) with HLM 6.02 (Raudenbush, Bryk, Cheong, & Congdon, 2004) to
confirm that hierarchical linear models would be appropriate to analyze our data. Then to
partition the variance at the episode, employee, and leader levels in hypothesis testing, we
used HLM 6.02 to test our hypotheses. We centered episode-level predictors with the group-
mean technique due to our research interests, as well as to separate the cross-level
interactions from the between-group interactions when testing the cross-level interactive
effects (Aguinis, Gottfredson, & Culpepper, 2013; Hofmann & Gavin, 1998). We centered
employee- and leader-level predictors with the grand-mean technique to reduce potential
collinearity between level-2 intercept and slope terms and to model the potential influences of
both within- and between-team variances (Hofmann & Gavin, 1998; Mathieu & Taylor,
2007). When testing the hypothesized multilevel mediated relationships, we used the Monte
Carlo method recommended by Selig and Preacher (2008) and Preacher, Zyphur, and Zhang
(2010) to estimate confidence intervals for determining their significance, with the help of an
open-source software R-based simulator (which can be found at http://www.quantpsy.org).
Descriptive Statistics and Confirmatory Factor Analyses
Table 1 shows the means, standard deviations, and correlations of the variables. We
conducted CFAs on eight focal variables (leader’s positive affect, employee’s positive affect,
employee’s assessment of leader’s positive affect, leader’s negative affect, employee’s
negative affect, employee’s assessment of leader’s negative affect, psychological safety, and
voice) and one critical control variable (interaction quality) at the episode level. The nine-
factor model fit the data well (χ2 = 1490.28, χ2/df = 3.25, RMSEA = .06, non-normed fit
index (NNFI) = .93, comparative fit index (CFI) = .94). This model fit the data better than
alternative models when the following variables were combined: (a) leader’s positive affect
and employee’s positive affect (∆χ2 ∆(8) = 2133.85, p < .01); (b) leader’s positive affect and
employee’s assessment of leader’s positive affect (∆χ2 ∆(8) = 2147.72, p < .01); (c)
employee’s positive affect and employee’s assessment of leader’s positive affect (∆χ2 ∆(8) =
748.07, p < .01); (d) leader’s negative affect and employee’s negative affect (∆χ2 ∆(8) =
833.83, p < .01); (e) leader’s negative affect and employee’s assessment of leader’s negative
affect (∆χ2 ∆(8) = 824.37, p < .01); (f) employee’s negative affect and employee’s
assessment of leader’s negative affect (∆χ2 ∆(8) = 801.42, p < .01); (g) psychological safety
and voice (∆χ2 ∆(8) = 727.05, p < .01); (h) leader’s positive affect , voice, and interaction
quality (∆χ2 ∆(15) = 1783.82, p < .01); and (i) all nine variables as a single factor (∆χ2 ∆(36)
= 9639.12, p < .01). The results indicated discriminant validity for these variables.
Partitioning of Variance
To check if the theoretical reason for using HLM (i.e., variance at episode and
employee levels) was justified empirically, we inspected the results of null models in HLM
(regressions without any predictors) for the eight core episode-level variables. Null models
separated the variance in these variables into episode, employee, and leader levels, and the
intercept represents the mean of the variable. The three-level HLM is justified only when
variances in the outcome variables are present at different levels. Table 2 shows the results
for each null model. First, these variables all had significant episode-level variances, ranging
from 31.3% to 69.9% (leader’s positive affect, 46.7%; employee’s positive affect, 31.3%;
employee’s assessment of leader’s positive affect, 42.4%; leader’s negative affect, 53.2%;
employee’s negative affect, 48.2%; employee’s assessment of leader’s negative affect, 69.9%;
psychological safety, 50.6%; and upward voice, 61.6%). In addition, except for leader’s
negative affect, all variables had significant employee-level variances, ranging from 7.8% to
67.8% (leader’s positive affect, 7.8%; employee’s positive affect, 67.8%; employee’s
assessment of leader’s positive emotion affect, 51.2%; employee’s negative affect, 51.7%;
employee’s assessment of leader’s negative affect, 30.1%; psychological safety, 37.2%; and
upward voice, 7.8%). Finally, four variables also had significant leader-level variance
(leader’s positive affect, 45.4%; leader’s negative affect, 46.8%; psychological safety, 12.3%;
and upward voice, 30.6%). Therefore, these results indicate that HLM was a more
appropriate analytic technique than standard OLS (LeBreton & Senter, 2008).
Table 3 presents the results of our HLM analysis. Hypothesis 1 predicted that
employee’s positive affect would mediate the positive relationship between leader’s positive
affect and employee psychological safety. As shown by the results in Table 3, Model 1,
leader’s positive affect was positively related to employee’s positive affect (γ = .26, p < .01).
In addition, the employee’s positive affect was positively related to employee psychological
safety even when the leader’s positive affect was controlled (γ = .23, p < .01; Table 3, Model
6). To further confirm this mediation, we used a Monte Carlo-based simulation methodology
(20,000 repetitions), which is similar to parametric bootstrapping, as suggested by Selig and
Preacher (2008). Results indicated that the indirect path from leader’s positive affect to
employee psychological safety via employee’s positive affect was significant (.06; 95% CI
[03, .09]). Hence, Hypothesis 1 was supported.
Hypothesis 2 predicts that an employee’s positive affect and psychological safety
would sequentially mediate the positive relationship between leader’s positive affect and
employee’s upward voice. As shown by the results in Table 3, Model 9, employee
psychological safety was positively related to employee’s upward voice (γ = .11, p < .01).
Based on this coefficient and the results in testing Hypothesis 1, we used the same Monte
Carlo-based simulation methodology (20,000 repetitions) and found that the indirect path for
leader’s positive affect employee’s positive affect psychological safety voice was
significant and positive (.0077; 95% CI [.00040, .0151]). Therefore, Hypothesis 2 was
supported. Although the point estimates (.0077) for this indirect effect look small, Preacher
and Kelley (2011) suggested that the estimates of indirect effects are determined by the range
of possible values of each link in the mediation process and are very likely to differ from the
population parameters. Therefore, the indirect effect still provides meaningful and important
support to our hypothesis about the mediation relationships between variables.
Hypothesis 3 predicted that the employee’s assessment of the leader’s positive affect
would mediate the positive relationship between the leader’s positive affect and the
employee’s psychological safety. As shown by the results in Table 3, Model 2, the leader’s
positive affect was positively related to the employee’s assessment of the leader’s positive
affect (γ = .31, p < .01). The employee’s assessment of the leader’s positive affect, however,
was not significantly related to employee psychological safety (γ = .07, n.s., Table 3, Model
6). Hence, Hypothesis 3 was not supported. As a result, we did not further test Hypothesis 4,
which predicted that an employee’s assessment of the leader’s positive affect and employee
psychological safety would sequentially mediate the positive relationship between the
leader’s positive affect and the employee’s upward voice.
Hypothesis 5 predicted that LMX would moderate the relationship between the
employee’s positive affect and his or her psychological safety. As shown in Table 3, Model 7,
the interaction term was significant (γ = -.18, p < .05). With the comparison of a raw random-
slope model without any slope predictors, LMX explained 7% of the variance of the slope.
Following Aiken and West (1991), we present this interaction graphically at two levels of
LMX (i.e., +1 SD and –1 SD) in Figure 2a. A simple slopes test indicated that employee’s
positive affect was positively related to employee psychological safety at lower levels of
LMX (γ = .28, t = 3.78, p < .01), but not significantly related to it at higher levels of LMX (γ
= .07, t = .95, n.s.); and the two simple slopes were significantly different from each other (t
= 2.02, p < .05). We also examined indirect paths using the Monte Carlo-based simulation
methodology (20,000 repetitions). When LMX was low, the indirect path from employee’s
positive affect to upward voice via psychological safety was significant and positive (.036;
95% CI [.0013, .070]), which was significantly stronger than the indirect path when LMX
was high (.011, 95% CI [-.0083, .031], t = 1.98, p < .05). Therefore, Hypothesis 5 was
Hypothesis 6 predicted that LMX moderates the relationship between the employee’s
assessment of the leader’s positive affect and employee psychological safety. As shown in
Table 3, Model 6, the interaction term was significant (γ = -.20, p < .05). With the
comparison of a raw random-slope model without any slope predictors, LMX explained 15%
of the variance of the slope. This interaction at two levels of LMX (i.e., +1 SD and –1 SD;
Aiken & West, 1991) is presented graphically in Figure 2b. A simple slopes test indicated
that the perceived positive affect of leaders was positively related to employee psychological
safety at lower levels of LMX (γ = .20, t = 2.36, p < .05), but it was not significantly related
to it at higher levels of LMX (γ = -.04, t = -.44, n.s.), and the two simple slopes were
significantly different from each other (t = 2.12, p < .05). We also examined indirect paths
using the Monte Carlo-based simulation methodology (20,000 repetitions). When LMX was
low, the indirect path from the employee’s assessment of the leader’s positive affect to
upward voice via psychological safety was significant and positive (.029; 95% CI
[.0002, .0571]), which was significantly stronger than the indirect path when LMX was high
(-.0054, 95% CI [-.0264, .0156], t = 2.01, p < .05). Therefore, Hypothesis 6 was supported.
We were also interested in examining the role of negative affect in the context of
voice. To address our first research question regarding whether a leader’s negative affect
influences employee voice, Model 9 in Table 3 showed that the leader’s negative affect was
positively and significantly related to voice (γ = .30, t = 3.61, p < .01). Our second and third
research questions concerned the mechanisms by which a leader’s negative affect influences
voice. Model 3 in Table 3 showed that the leader’s negative affect was positively and
significantly related to the employee’s negative affect (γ = .13, t = 3.24, p < .01), which,
however, was not significantly related to psychological safety (Model 6 in Table 3, γ = -.01,
n.s.) or voice (Model 9 in Table 3, γ = .02, n.s.). In addition, Model 4 in Table 3 showed that a
leader’s negative affect was positively and significantly related to the employee’s assessment
of the leader’s negative affect (γ = .19, t = 3.53, p < .01), which, however, was not
significantly related to psychological safety (Model 6 in Table 3, γ = .03, n.s.) or voice
(Model 9 in Table 3, γ = .05, n.s.).
We also investigated whether LMX quality might moderate the effects of an
employee’s negative affect and of an employee’s assessment of leader’s negative affect on
psychological safety. The slope-as-random HLM models, however, showed that there was not
significant variance for either the relationship between employee’s negative affect and
psychological safety or the relationship between employee’s assessment of leader’s negative
affect and psychological safety. Therefore, we did not further explore.
In this study, we have highlighted the important role of leaders’ affect in the process
of employees’ upward voice during leader-member interactions. Using the emotion-as-social-
We also conducted supplementary analyses by separately examining each negative affect item (i.e., distressed,
angry, sad, and afraid). We found that (1) except for afraid, leader’s single negative state was positively related
to employee’s single negative state (contagion effect), (2) except for distressed, leader’s single negative state
was positively related to employee’s assessment to leader’s single negative state (signaling effect); (3) we did
find that employees who were afraid were less likely to engage in voice, though this relationship was marginal
(p < .10); and (4) employee’s assessment of leader’s distressed was positively but marginally (p < .10) related to
information model (Van Kleef et al., 2010) as our theoretical lens, we conceptualized and
examined voice as a dynamic interaction between leaders and employees. Our findings
suggest that leaders’ positive affect influences voice through psychological safety directly via
employees’ own positive affect (emotional contagion pathway) but not directly via
employees’ assessments of leaders’ positive affect (signaling pathway). Moreover, we found
that employees’ own positive affect and their assessments of leaders’ positive affect were
related to employees’ upward voice via employees’ psychological safety only when LMX
was low rather than high. Interestingly, leaders’ negative affect was also positively related
with voice; although leaders’ negative affect was related to employees’ negative affect
(emotional contagion pathway) and employees’ assessments of leaders’ negative affect
(signaling pathway), none of which were significantly related to psychological safety or voice.
The findings of our study generate some interesting implications for theory and practice.
Our findings contribute to the voice, affect, and leadership literatures in several
important ways. First, this study unveils upward voice as a dynamic behavior with episodic
variance. Most previous studies have investigated voice at the individual or group level and
thus only focused on relatively stable personal, relational, or situational predictors of voice
(Morrison, 2011, 2014). Although some scholars have discussed the episodic characteristics
of voice (e.g., Detert & Edmondson, 2011; Detert & Treviño, 2010), little effort has been
made to investigate voice empirically at the episode level. Our data showed that 61.6% of the
variance of upward voice occurred at the episode level, which indicates that it is meaningful
and important to examine voice as an episodic behavior in organizations. Our finding,
together with recent research suggesting substantial within-individual variance in work
behaviors (e.g., Dalal, Lam, Weiss, Welch, & Hulin, 2009; Ilies, Scott, & Judge, 2006),
demonstrates that taking an episodic approach to examine organizational behavior can be
fruitful and beneficial (Beal, Weiss, Barros, & MacDermid, 2005).
Second, this research enhances our understanding of the role of leaders’ affective
states in affecting employee voice. Past qualitative research and conceptual discussions have
focused mainly how an individual’s own affect is associated with his or her voice (Edwards,
Ashkanasy, & Gardner, 2009; Harvey et al., 2009; Kish-Gephart et al., 2009; Milliken et al.,
2003). However, the effects of others’ emotions, such as leaders’ positive affect, have
remained unknown. We have addressed this important research question by conducting a
multilevel, experience sampling field study to provide empirical evidence of the effects of
leaders’ affect on employee voice. Our study not only provides empirical evidence of
“reading the wind” (Dutton et al., 1997), but also responds to scholars’ calls for a fine-tuned
framework to explore the connection between affect and voice (Grant & Ashford, 2008;
Morrison, 2011). Using three-level hierarchical data, we have demonstrated that leaders’
positive and negative affective states have unique, independent effects on employee voice.
Hence, our research also adds to the leadership literature showing that leaders’ affect plays a
crucial role in influencing employees’ attitudes and behaviors (Gooty et al., 2010). This
contribution is important because the extant literature has focused primarily on attitudinal
outcomes or in-role performance as consequences of leaders’ affect.
A third theoretical contribution of this research is that it integrates emotional
contagion and signaling mechanisms with psychological safety to explain why leaders’
positive affect influences voice. Although recent years have witnessed increasing interest in
exploring the effects of leaders’ affect, studies of the explanatory mechanisms have been very
limited (Madera & Smith, 2009). Introducing the EASI model to voice research, we have
investigated whether the emotional contagion or signaling mechanisms can explain the
effects of leaders’ positive affect on employees’ psychological safety and in turn, employees’
voice. Employees’ positive affect was found to account for the positive effects of leaders’
positive affect on psychological safety and, thus, on voice. This finding supports emotional
contagion theory (Hatfield et al., 1994) and also sheds light on the EASI literature (Van Kleef
et al., 2010), in that psychological safety may provide another mechanism explaining why
leaders’ affect influences employees. Interestingly, by contrast, employees’ assessments of
leaders’ positive affect was not significantly related to psychological safety or voice. This
finding suggests that the signaling mechanism is probably more complicated than the
emotional contagion process (Van Kleef et al., 2010) because whether the assessments of
leaders’ positive affect lead to psychological safety or voice depends on contextual factors,
such as LMX.
Fourth, our study extends the EASI model and the LMX literature by theorizing the
moderating role of LMX in attenuating the effects of leaders’ affect. With a primary focus on
conflict resolution contexts, prior EASI research has found that personal traits, such as
agreeableness (Van Kleef, Homan, Beersma, & van Knippenberg, 2010) and need for closure
(Van Kleef et al., 2009), and relational characteristics, such as power difference (Van Kleef et
al., 2004), are the boundary conditions for the social functions of emotions. Applying the
EASI model to leader-member interaction contexts, our study has indicated that LMX quality
moderates both emotional contagion and the signaling process. That is, employees who have
low rather than high LMX quality are more likely to be influenced by leaders’ affect —both
being more susceptible to leaders’ affect and being more motivated to cognitively analyze
leaders’ affect in formatting psychological safety in interactions with leaders. Therefore, our
findings suggest that in the workplace, LMX quality plays an important role in influencing
people’s reactions to affect as social information during interactions (Dunn & Schweitzer,
2005), an outcome that extends EASI research.
This research also adds to the existing LMX literature (Dulebohn et al., 2012), which
has not addressed the potential role of LMX in the effects of leaders’ affect. This is partially
because past LMX research has rarely taken a within-individual approach to examine leader-
member interaction dynamics, thus neglecting LMX as an important contingent contextual
factor that influences interaction episodes. Directly addressing this research gap, we
combined experience sampling data (mobile survey) and LMX data (baseline survey) in our
study. Besides the moderating role of LMX, we also observed that LMX did not have a
significant effect on voice (Model 9, Table 3, γ = .01, n.s.), a finding contradictory to
previous voice research at the individual level (e.g., Burris et al., 2008; Liu, Tangirala, &
Ramanujam, 2013; Van Dyne et al., 2008). We surmise that this is probably because we
collected data for only a relatively short period (10 working days), such that momentary
characteristics, such as leaders’ affective states, are more salient predictors than stable
predictors, such as LMX.
Finally, with parallel data of both positive affect (leader’s positive affect, employee’s
positive affect, and employee’s assessment of leader’s positive affect) and negative affect
(leader’s negative affect, employee’s negative affect, and employee’s assessment of leader’s
negative affect), our study showed some similar as well as different patterns for these two
types of affective states. For example, interestingly, like leader’s positive affect, leader’s
negative affect was positively related with employee voice. Moreover, we found that similar
to leaders’ positive affect, leaders’ negative affect was transferred to employees, who could
also assess them cognitively. These findings suggest that like positive affect, negative affect
also serves important social functions (e.g., Lelieveld, Van Dijk, Van Beest, Steinel, & Van
Kleef, 2011; Madera & Smith, 2009) and that emotional contagion and signaling are the two
possible influential processes (e.g., Eberly & Fong, 2013; Van Kleef et al., 2009).
Unlike positive affect, however, neither employee’s overall negative affect nor
employee’s assessment of leader’s overall negative affect had any significant effects on
psychological safety or voice. Nevertheless, it is too early to conclude that negative affect do
not predict voice. There are several possible explanations for our non-findings. From a
methodological perspective, these non-findings might be due to the relatively few negative
interactions in our sample. For example, compared with employee’s positive affect and
employee’s assessment of leader’s positive affect (Ms = 3.66/3.66, SD = 1.01/.99),
employee’s negative affect and employee’s assessment of leader’s negative affect (Ms =
1.23/1.23, SD = .48/.53) were more restricted in range. From a theoretical perspective, these
non-findings suggest that the social functions of negative affect may be more differentiated
than those of positive affect. Previous research has shown that compared with positive
emotions, negative emotions have more dimensions and are richer in their meanings (e.g., de
Rivera et al., 1989; Ellsworth & Smith, 1988; Fredrickson, 1998). Moreover, some negative
emotions, such as fear, deactivate one’s willingness to speak up (Kish-Gephart et al., 2009)
because these emotions represent warning, threat, or punishment (Elfenbein, 2007; Larsen &
Ketelaar, 1991). In our supplementary analyses, we did find that employees’ feeling of being
afraid was negatively (but marginally) related to voice. By contrast, some other negative
emotions, such as angry and distressed, suggest something is wrong and change is needed
(e.g., e.g., Edwards et al., 2009; Harvey et al., 2009). In our supplementary analyses, we did
find that employee’s assessment of leader’s distress was positively (but marginally) related to
voice. Of course, these initial findings were limited because they were based on the single-
item measure of each type of negative affect. Therefore, it is critical to investigate discrete
negative emotions, such as anger, disappointment, and sadness, with more sophisticated
methods in the future (e.g., Lelieveld et al., 2011; Van Kleef et al., 2010).
Our study has significant implications for managerial practices. We have shown that
leaders’ positive affect promotes employee psychological safety and, thus, voice. This result
can serve as advice to organizations that managers should display positive affect when
interacting with their subordinates. Managers’ affect, something they can control but often
neglect in their interactions with employees, can be an effective management tool influencing
employees (Huy, 2002). In organizations where employees’ suggestions, opinions, and
concerns have critical implications for organizational functions, managers should pay more
attention to their display of positive affect in everyday contacts with employees, because
showing positive affect to employees can increase employees’ psychological safety in freely
expressing their ideas (Edmondson, 2003). In addition, organizations should implement
training programs to teach management how to express positive affect to employees. Our
results showed that both leaders’ positive affect and employees’ assessments of leaders’
positive affect fluctuated with a large magnitude. Hence, organizations should invest in
training programs to help management develop more positive attitudes toward work, as well
as better interpersonal skills. Furthermore, to encourage employees to speak up, managers
should consider exercising transformational leadership, which has been suggested to promote
employees’ positive affect (Bono et al., 2007) and voice (Detert & Burris, 2007). Moreover,
our finding about the emotional contagion route also suggests that managers should consider
having more face-to-face interactions with employees. Individuals are likely to experience
similar emotions when they are exposed to emotionally laden faces, bodies, and voices
(Hatfield et al., 1993). Therefore, having more positive physical (rather than virtual)
interactions would be more likely to encourage employees to speak up.
Another implication of our research relates directly to LMX in leader-member
interaction dynamics. Our findings showed that leaders’ positive affect play a salient role in
eliciting psychological safety and voice, especially for employees who have low LMX with
leaders. This suggests that employees with low LMX are more likely to be the wind-readers
(Ashford et al., 1998) and that subtle cues signaled by management, such as positive affect,
significantly shape their attitudes and behaviors. Accordingly, when interacting with
employees in whom LMX is low, managers should be more careful about their display of
affective states. Organizations should share this finding with newly appointed leaders who
have not yet established high LMX with organization members (Bauer & Green, 1996), who
especially need employees to speak up in order to collect constructive ideas and bring
changes to the new environment (Morrison & Milliken, 2000; Sauer, 2011).
Limitations and Directions for Future Research
Our study has several limitations that point to directions for future research. First, our
research focused on leaders’ high-activated positive affect (i.e., delighted, excited, happy, and
joyful) and high-activated negative affect (distressed, angry, sad, and afraid). A valuable
extension to our research would be to examine discrete emotions (e.g., excitement) rather
than high-activated positive affect in general. Van Kleef and colleagues (2010) have
emphasized that discrete emotions can give more accurate meanings to partners in social
interactions. A less activated positive emotion, for example, might signal the target’s
satisfaction with the status quo (Bindl, Parker, Totterdell, & Hagger-Johnson, 2012), and
therefore its positive effect on the actor’s change-oriented behaviors, including voice, might
not be as prominent. Existing research, however, has suggested that compared with negative
emotions, positive emotions are fewer in number and more diffuse (de Rivera et al., 1989;
Ellsworth & Smith, 1988; Fredrickson, 1998). This may alleviate, to a certain degree, the
concern that the findings obtained in our study cannot be generalized to other positive
emotions. Future studies, especially those with a focus on negative emotions, should explore
the effects of discrete emotions. An even more interesting and aggressive agenda would be to
examine and compare different emotions in the emotion circumplex (Russell, 1980) in the
context of voice.
Second, in our study we only asked leaders to report employee voice, an observable
behavior by others; we did not explore a related behavior, silence or information withholding
(Morrison & Milliken, 2000; Tangirala & Ramanujam, 2008). As negative emotions may be
more tied to silence rather to voice (Liu et al., 2015), future research should consider
exploring such a direction. In addition, as only the employees themselves are probably aware
of silence, future research should examine both voice and silence by taking into account
multiple perspectives from the actor and the target and by objectively measuring these
Third, although we drew upon the EASI model (Van Kleef et al., 2010) to theorize
mediators (e.g., employees’ positive affect [EPA], assessment of leaders’ positive affect
[ALPA], and psychological safety) and a contingent factor (LMX) for the relationship
between leaders’ positive emotions and voice in our research, future research should further
investigate other possible mediating mechanisms and moderators. Interestingly, our data
showed that leaders’ positive affect s (LPA), the distal predictor in our model, was
significantly associated with voice, yet none of the more proximal employee predictors (EPA
and ALPA) were significant (Model 9, Table 3). This finding may result from common
method bias, as leaders reported their own positive affect as well as voice, but it may also
suggest that besides emotional contagion (via EPA) and signaling (via ALPA) and
psychological safety, there are other potential mediating mechanisms linking LPA and voice.
LPA, for example, may energize employee to have a sense of power, which in turn may lead
to voice (Tangirala & Ramanujam, 2012). Future research should explore other mechanisms.
Another cautious point is that we did not directly measure employee’s strategic
inferences of leaders’ affective statesfor the signaling pathway, although we believe that
employees’ assessment of leaders’ affect should be the precondition for making strategic
inferences (e.g., Elfenbein, 2007). In other words, an employee must first perceive and assess
a leader’s affect before making an inference about it. As such, future research should measure
strategic inferences directly, as Van Kleef et al. (2010) suggested, to discern the functions of
signaling mechanism in social interactions.
Fourth, given that participants reported all of our key variables after each interaction,
our study could not firmly establish causality for the hypothesized relationships. Prior
research and supplemental analyses, however, may help overcome this limitation. Existing
emotion research, for example, has shown that the emotions of an individual with high power
are more likely to (a) exert influence on, and (b) be transmitted to individuals with low power,
rather than vice versa (Anderson et al., 2003; Van Kleef et al., 2004). Hence, in our study,
leaders’ affect is more likely to be the predictors of employees’ positive affect and their
assessment of leaders’ positive affect. Moreover, to establish the causal relationship between
leaders’ positive emotions affect and voice, a core research interest in this study, we
conducted supplemental lagged analyses with a subsample of participants who reported more
than two episodes within a day. In support of our argument, the results showed that (a)
leaders’ positive affect led to employee voice but not vice versa; (b) leader’s positive affect
led to employee’s assessment of leader’s positive affect but not vice versa; and (c)
employee’s assessment of leader’s positive affect led to psychological safety but not vice
Despite these theoretical arguments and analytic efforts, we still cannot firmly
We thank the Editor for providing this suggestion. Specifically, we selected paired mobile surveys submitted
within the same day (n = 322) to conduct lagged analyses and obtained some meaningful findings. First, we
found that employee voice in the previous episode did not lead to leaders’ positive affect in the current episode
(γ = .05, n.s.), when controlling leaders’ positive affect in the previous episode. By contrast, leaders’ positive
affect in the previous episode did lead to employee voice in the current episode (γ = .15, p < .05), when
controlling voice in the previous episode. Second, leader’s positive affect in the previous episode was positively
related with employee’s assessment of leader’s positive affect in the current episode (γ = .20, p < .05), even
when controlling employee’s assessment of leader’s positive affect in the previous episode, but employee’s
assessment of leader’s positive affect in the previous episode was not significantly related with leader’s positive
affect in the current episode (γ = .03, n.s.) when controlling leader’s positive affect in the previous episode.
Third, employee’s assessment of leader’s positive affect in the previous episode was positively related with
establish the causal links proposed in our study without an experimental design or pre- and
post-interaction measures of affect using ESM. Future research should employ a more
microscopic approach to fill in this gap.
Fifth, most of the measures in our study (i.e., EPA, ALPA, and psychological safety)
were all from employees, which may raise questions about common method bias (i.e.,
inflation of relationships among study variables). The intra-individual level correlations,
however, were not substantial (the correlations ranged from .05 to .59). In addition, as we
used intra-individual analyses by centering the variable scores at the individuals’ means, we
thus sufficiently eliminated the potential response tendencies that stem from personal
characteristics and experiences. Another possible source of common variance that causes
concern in between-individual analyses, LMX quality, was conceptualized and analyzed as a
moderator in this study and thus alleviated the concern about common method bias.
Nevertheless, future research that overcomes the methodological limitations associated with
this study (by using observations, for example) could provide more accurate assessments of
the relationships of interest.
Sixth, although the mobile-survey method has the advantage of making it possible to
conduct a survey anytime and anywhere (Li & Townsend, 2008), we were able to record only
30% of the interactions that took place. There were several possible reasons for this relatively
low response rate. First, consistent with guidance for conducting an interactional study
(Bolger et al., 2003), to avoid placing too much burden on participants, we asked our
participants to only report interactions that last for more than two minutes. It is thus likely
that our study neglected some short conversations or small talk. Second, participants might
have been too busy with their work, and thus had no time to record their conversations in
psychological safety in the current episode (γ = .09, p < .05), even when controlling psychological safety in the
previous episode, but psychological safety in the previous episode was not significantly related with employee’s
assessment of leader’s positive affect in the current episode (γ = -.05, n.s.) when controlling employee’s
assessment of leader’s positive affect in the previous episode.
time. Third, the survey could not be completed when participants did not bring their phones
with them, when their phones had no power, or when they were in places with poor mobile
network coverage. Lastly, our study asked participants to record only their face-to-face
interactions, thus excluding interactions through electronic media, such as phone
conversations, email, and SMS. Future studies are encouraged to examine hypothesized
relationships in the above situations and to apply innovative methods to capture more dyadic
Another limitation of our study is that we did not measure leaders’ behavior at the
episode level, which may inflate the effects of leaders’ affect. Although we did control
transformational leadership at the leader level, we could not exclude the possibility that voice
results from leaders’ behavior rather than leaders’ affect. Future research should measure
leaders’ episodic behavior to exclude such an alternative explanation.
Finally, our data were from China, a culture that features a highly collectivistic
orientation and high power distance (House, Hanges, Javidan, Dorfman, & Gupta, 2004). So
it is questionable whether our results could be extended to other cultures. As existing voice
behavior research has involved mainly Western cultures (e.g., Detert & Burris, 2007; Van
Dyne & LePine, 1998; Tangirala & Ramanujam, 2008), however, our research, supported by
data from Mainland China, may bring fresh perspectives. Of course, future research should
explore whether our findings can be replicated in other cultures.
Our study has highlighted the importance of leaders’ affect as a critical factor that
influences both employees’ psychological safety and upward voice behavior. Our results also
indicate that leaders’ positive affect are more likely to influence employees who are low in
LMX, through both employees’ own positive affect and their assessments of leaders’ positive
affect. These findings indicate that leaders’ emotions affect matter to upward voice and
suggest the importance of taking a dynamic, within-individual approach to study the
connection between affect and voice.
Aguinis, H., Gottfredson, R. K., & Culpepper, S. A. 2013. Best-practice recommendations for
estimating cross-level interaction effects using multilevel modeling. Journal of
Management, 39: 1-39.
Aiken, L. S., & West, S. G. 1991. Multiple Regression: Testing and interpreting
interactions. Newbury Park, CA: Sage.
Ambady, N., & Rosenthal, R. 1992. Thin slices of expressive behavior as predictors of
interpersonal consequences: A meta-analysis. Psychological Bulletin, 111: 256-274.
Ames, D. R., & Johar, G. V. 2009. I'll know what you're like when I see how you feel: How
and when affective displays influence behavior-based impressions. Psychological
Science, 20: 586-593.
Anderson, C., Keltner, D., & John, O.P. 2003. Emotional convergence between people over
time. Journal of Personality and Social Psychology, 84: 1054-1068.
Anderson, C., & Thompson, L.L. 2004. Affect from the top down: How powerful individuals'
positive affect shapes negotiation. Organizational Behavior and Human Decision
Processes, 95: 125-139.
Ashford, S.J., Rothbard, N.P., Piderit, S.K., & Dutton, J.E. 1998. Out on a limb: The role of
context and impression management in selling gender-equity issues. Administrative
Science Quarterly, 43: 23-57.
Avolio, B. J., Bass, B. M., & Jung, D. I. 1999. Re-examining the components of
transformational and transactional leadership using the multifactor leadership
questionnaire. Journal of Occupational and Organizational Psychology, 72: 441-462.
Barger, P.B., & Grandey, A.A. 2006. Service with a smile and encounter satisfaction:
Emotional contagion and appraisal mechanisms. Academy of Management Journal, 40:
Bauer, T.N., & Green, S.G. 1996. Development of leader-member exchange: A longitudinal
test. Academy of Management Journal, 39: 1538-1567.
Beal, D. J., Weiss, H. M., Barros, E., & MacDermid, S. M. 2005. An episodic process model
of affective influences on performance. Journal of Applied Psychology, 90: 1054-1068.
Bindl, U.K., Parker, S.K., Totterdell, P., & Hagger-Johnson, G. 2012. Fuel of the self-starter:
How mood relates to proactive goal regulation. Journal of Applied Psychology, 97: 134-
Blau, P. M. 1964. Exchange and power in social life. New York: John Wiley.
Bolger, N., Davis, A., & Rafaeli, E. 2003. Diary methods: Capturing life as it is lived.
Annual Review of Psychology, 54: 579-616.
Bono, J.E., Foldes, H.J., Vinson, G., & Muros, J.P. 2007. Workplace emotions: The role of
supervision and leadership. Journal of Applied Psychology, 92: 1357-1367.
Bower, G. H. 1981. Mood and memory. American Psychologist, 36: 129-148.
Burris, E.R. 2012. The risks and rewards of speaking up: Managerial responses to employee
voice. Academy of Management Journal, 55: 851-875.
Burris, E.R., Detert, J.R., & Chiaburu, D.S. 2008. Quitting before leaving: The mediating
effect of psychological attachment and detachment on voice. Journal of Applied
Psychology, 93: 912-922.
Chatman, J.A., & Flynn, F.J. 2005. Full-cycle organizational psychology research.
Organization Science,16: 434-447.
Clore, G. L., Gaspar, K., & Garvin, E. 2001. Affect as information. In J. P. Forgas (Ed.),
Handbook of affect and social cognition (pp. 121-144). Mahwah, NJ: Erlbaum.
Costa, P. T. Jr., & McCrae, R. R. 1989. The NEO-PI/NEO-FFI manual supplement. Odessa,
FL: Psychological Assessment Resources.
Cropanzano, R., & Mitchell, M. S. 2005. Social exchange theory: An interdisciplinary review.
Journal of Management, 31: 874-900.
Cropanzano, R., Weiss, H., Hale, J., & Reb, J. 2003. The structure of affect: Reconsidering
the relationship between negative and positive affectivity. Journal of Management, 29:
Dalal, R.S., Lam, H., Weiss, H.M., Welch, E.R., & Hulin, C.L. 2009. A within-person
approach to work behavior and performance: Concurrent and lagged citizenship-
counterproductivity associations, and dynamic relationships with affect and overall job
performance. Academy of Management Journal, 52: 1051-1066.
Dansereau, F., Jr., Graen, G., & Haga, W. J. 1975. A vertical dyad linkage approach to
leadership within formal organizations—A longitudinal investigation of the role making
process. Organizational Behavior and Human Performance, 13: 46-78.
De Rivera, J., Possel, L., Verette, J. A., & Weiner, B. 1989. Distinguishing elation, gladness,
and joy. Journal of Personality and Social Psychology, 57: 1015-1023.
Detert, J.R., & Burris, E.R. 2007. Leadership behavior and employee voice: Is the door really
open? Academy of Management Journal, 50: 869-884.
Detert, J.R., Burris, E.R., Harrison, D.A., & Martin, S. 2013. Voice flows to and around
leaders: understanding when units are helped or hurt by employee voice. Administrative
Science Quarterly, 58: 624-668.
Detert, J.R., & Edmondson, A.C. 2011. Implicit voice theories: Taken-for-granted rules of
self-censorship at work. Academy of Management Journal, 54: 461-488.
Detert, J.R., & Treviño, L.K. 2010. Speaking up to higher ups: How supervisor and skip-level
leaders influence employee voice. Organization Science, 21: 241-270.
Doherty, R.W. 1997. The emotional contagion scale: A measure of individual differences.
Journal of Nonverbal Behavior, 21: 131-154.
Dulebohn, J.H., Bommer, W.H., Liden, R.C., Brouer, R.L., & Ferris, G.R. 2012. A meta-
analysis of antecedents and consequences of leader-member exchange: Integrating the
past with an eye toward the future. Journal of Management, 38: 1715-1759.
Dunn, J.R., & Schweitzer, M.E. 2005. Feeling and believing: The influence of emotion on
trust. Journal of Personality and Social Psychology, 88: 736-748.
Dutton, J.E., & Ashford, S.J. 1993. Selling issues to top management. Academy of
Management Review, 18: 397-428.
Dutton, J. E., Ashford, S. J., Wierba, E.E., O'Neill, R., & Hayes, E. 1997. Reading the wind:
How middle managers assess the context for issue selling to top managers. Strategic
Management Journal, 15: 407-425.
Eberly, M.B., & Fong, C.T. 2013. Leading via the heart and mind: The roles of leader and
follower emotions, attributions and interdependence. Leadership Quarterly, 24: 696-711.
Eckel, C.C., & Wilson, R.K. 2003. The human face of game theory: Trust and reciprocity in
sequential games. In: Ostrom, E., Walker, J., (Eds). Trust and Reciprocity:
Interdisciplinary Lessons from Experimental Research. New York: Russel Sage
Foundation, pp. 245-274.
Edmondson, A.C. 1999. Psychological safety and learning behavior in work teams.
Administrative Science Quarterly, 44: 350-383.
Edmondson, A. C. 2003. Speaking up in the operating room: How team leaders promote
learning in interdisciplinary action teams. Journal of Management Studies, 40: 1419-
Edwards, M.S., Ashkanasy, N.M., & Gardner, J. 2009. Deciding to speak up or remain silent
following observed wrongdoing: The role of discrete emotions and climate of silence. In J.
Greenberg &M. Edwards (Eds.), Voice and silence in organizations (pp. 83-110).
Bingley, England: Emerald.
Edwards, J. R., & Lambert, L. S. 2007. Methods for integrating moderation and mediation:
A general analytical framework using moderated path analysis. Psychological Methods,
Elfenbein, H. A. 2007. Chapter 7: Emotion in organizations. Academy of Management
Annals, 1: 315-386.
Ellsworth, P. C, & Smith, C. A. 1988. Shades of joy: Patterns of appraisal differentiating
pleasant emotions. Cognition and Emotion, 2: 301-331.
Erez, A., & Isen, A.M. 2002. The influence of positive affect on the components of
expectancy motivation. Journal of Applied Psychology, 87: 1055-1067.
Fast, N., Burris, E., & Bartel, C. 2014. Managing to stay in the dark: Managerial self-efficacy,
ego defensiveness, the aversion to employee voice. Academy of Management Journal,
Filipowicz, A., Barsade, S., & Melwani, S. 2011. Understanding emotional transitions: The
interpersonal consequences of changing emotions in negotiations. Journal of Personality
and Social Psychology, 101: 541-556.
Forgas, J. P. 1992. On bad mood and peculiar people: Affect and person typicality in
impression formation. Journal of Personality and Social Psychology, 62: 863-875.
Forgas, J. P. 1995. Mood and judgment: The affect infusion model (AIM). Psychological
Bulletin, 117: 39-66.
Forgas, J. P. 1998. Happy and mistaken? Mood effects on the fundamental attribution error.
Journal of Personality and Social Psychology, 75: 318–331.
Forgas, J. P., & Bower, G. H. 1987. Mood effects on person–perception judgments. Journal
of Personality and Social Psychology, 53: 53-60.
Forgas, J. P., Bower, G. H., & Moylan, S. J. 1990. Praise or blame? Affective influences on
attributions for achievement. Journal of Personality and Social Psychology, 59: 809–818.
Forgas, J.P., & George, J.M. 2001. Affective influences on judgments and behavior in
organizations: An information processing perspective. Organizational Behavior and
Human Decision Processes, 86: 3-34.
Fredrickson, B. L. 1998. What good are positive emotions? Review of General Psychology, 2:
Fredrickson, B. L. 2001. The role of positive emotions in positive psychology: The broaden-
and-build theory of positive emotions. American Psychologist, 56: 218-226.
Frese, M., & Fay, D. 2001. Personal initiative (PI): An active performance concept for work
in the 21st century. In B.M. Staw & R.M. Sutton (Eds.), Research in Organizational
Behavior (Vol. 23, pp. 133-187). Amsterdam: Elsevier Science.
Fridlund, A. J. 1994. Human facial expression: An evolutionary view. San Diego, CA:
Frijda, N. H. 1986. The emotions. Cambridge, U.K.: Cambridge University Press.
George, J. M., & Brief, A. P. 1992. Feeling good—doing good: A conceptual analysis of the
mood at work–organizational spontaneity relationship. Psychological Bulletin, 112: 310–
Gerstner, C. R., & Day, D. V. 1997. Meta-analytic review of leader-member exchange theory:
Correlates and construct issues. Journal of Applied Psychology, 82: 827-844.
Gino, F., & Schweitzer, M.E. 2008. Blinded by anger or feeling the love: How emotions
influence advice taking. Journal of Applied Psychology, 93: 1165-1173.
Gooty, J., Connelly, S., Griffith, J., & Gupta, A. 2010. Leadership, affect and emotions: A
state of the science review. Leadership Quarterly, 21: 979-1004.
Graen, G. B., & Uhl-Bien, M. 1995. Relationship-based approach to leadership: Development
of leader-member exchange (LMX) theory of leadership over 25 years: Applying a multi-
level multi-domain perspective. Leadership Quarterly, 6: 219-247.
Grant, A.M. 2013. Rocking the boat but keeping it steady: the role of emotion regulation in
employee voice. Academy of Management Journal, 56: 1703-1723.
Grant, A. M., & Ashford, S. J. 2008. The dynamics of proactivity at work. In B. M. Staw & A.
Brief (Eds.), Research in Organizational Behavior, vol. 28, 3-34.Greenwich, CT: JAI.
Grant, A.M., Parker, S., & Collins, C. 2009. Getting credit for proactive behavior: Supervisor
reactions depend on what you value and how you feel. Personnel Psychology, 62: 31-55.
Gump, B.B., & Kulik, J.A. 1997. Stress, affiliation, and emotional contagion. Journal of
Personality and Social Psychology, 72: 305-319.
Harvey, P., Martinko, M.J., & Douglas, S.C. (2009). Causal perceptions and the decision to
speak up or pipe down. In J. Greenberg & M. Edwards (Eds.), Voice and silence in
organizations (pp. 63-82). Bingley, England: Emerald University Press, New York, NY.
Hatfield, E., Cacioppo, J.T., & Rapson, R.L. 1994. Emotional Contagion, Cambridge
Hofmann, D. A., & Gavin, M. B. 1998. Centering decisions in hierarchical linear models:
Implications for research in organizations. Journal of Management, 24: 623-641.
House, R.J., Hanges, P.J., Javidan, M., Dorfman, P.W., & Gupta, V. 2004. Culture,
leadership, and organizations: The GLOBE study of 62 societies. Thousand Oaks, CA:
Huy, Q.N. 2002. Emotional balancing of organizational continuity and radical change: The
contribution of middle managers. Administrative Science Quarterly, 47: 31-69.
Ilies, R., Scott, B.A., & Judge, T.A. 2006. The interactive effects of personal traits and
experienced states on intraindividual patterns of citizenship behavior. Academy of
Mangement Journal, 49: 561-575.
Ilies, R., Wagner, D.T., & Morgeson, F.P. 2007. Explaining affetive linkage in teams:
Individual differences in susceptibility to contagion and individualism-collectivism.
Journal of Applied Psychology, 92: 1140-1148.
Isen, A. M., Shalker, T. E., Clark, M., & Karp, L. 1978. Affect, accessibility of material in
memory, and behavior: A cognitive loop? Journal of Personality and Social Psychology,
Kahn, W. A. 1990. Psychological conditions of personal engagement and disengagement at
work. Academy of Management Journal, 33: 692-724.
Kelly, J. R., & Barsade, S. G. 2001. Moods and emotions in small groups and work teams.
Organizational Behavior and Human Decision Processes, 86: 99-130.
Keltner, D., & Haidt, J. 1999. Social functions of emotions at four levels of analysis.
Cognition and Emotion, 13: 505-521.
Kish-Gephart, J.J., Detert, J.R., Treviño, L.K., & Edmondson, A.C. 2009. Silenced by fear:
The nature, sources and consequences of fear at work. Research in Organizational
Behavior, 29: 163-193.
Kruglanski, A. W. 1989. Lay epistemics and human knowledge: Cognitive and motivational
bases. New York: Plenum.
Larsen, R. J., & Ketelaar, T. 1991. Personality and susceptibility to positive and negative
emotional states. Journal of Personality and Social Psychology, 61: 132-140.
Laurenceau, J.P., Feldman, B. L., & Pietromonaco, P.R. 1998. Intimacy as an interpersonal
process: The importance of self-disclosure, partner disclosure, and perceived partner
responsiveness in interpersonal exchanges. Journal of Personality and Social
Psychology, 74: 1238-1251.
Lazarus, R. S. 1991. Cognition and motivation in emotion. American Psychologist, 46: 352-
LeBreton, J. M., & Senter, J. L. 2008. Answers to 20 questions about interrater reliability and
interrater agreement. Organizational Research Methods, 11: 815-852.
Lelieveld, G.-J., Van Dijk, E., Van Beest, I., Steinel, W., & Van Kleef, G. A. 2011.
Disappointed in you, angry about your offer: Distinct negative emotions induce
concessions via different mechanisms. Journal of Experimental Social Psychology, 47:
LePine, J.A., & Van Dyne, L. 1998. Predicting voice behavior in work groups. Journal of
Applied Psychology, 83: 853-868.
LePine, J.A., & Van Dyne, L. 2001. Voice and cooperative behavior as contrasting forms of
contextual performance: Evidence of differential relationships with big five personality
characteristics and cognitive ability. Journal of Applied Psychology, 86: 326-336.
Li, H., & Townsend, L. 2008. Mobile research in marketing: Design and implementation
issues. International Journal of Mobile Marketing, 3: 32-40.
Liang, J., Fahr, C., & Fahr, L. 2012. A two-wave examination of the psychological
antecedents of voice behavior. Academy of Management Journal, 55: 71-92.
Liden, R. C., Sparrowe, R. T., & Wayne, S. J. 1997. Leader-member exchange theory: The
past and potential for the future. In G. R. Ferris (Ed.), Research in personnel and human
resources management (Vol. 15): 47-119. Greenwich, CT: JAI.
Liu, W., Tangirala, S., Lam, W., Chen, Z., Jia, R., & Huang, X. 2015. When and why peers’
positive mood influences team members’ voice behavior. Journal of Applied Psychology,
Liu, W., Tangirala, S., & Ramanujam, R. 2013. The relational antecedents of voice targeted
at different leaders. Journal of Applied Psychology, 98: 841-851.
Liu, W., Zhu, R.H., & Yang, Y.K. 2010. I warn you because I like you: Voice behavior,
employee identifications, and transformational leadership. Leadership Quarterly, 21:
Luthans, F. 2002. The need for and meaning of positive organizational behavior. Journal of
Organizational Behavior, 23: 695-706.
Luthans, F., & Youssef, C. M. 2007. Emerging positive organizational behavior. Journal of
Management, 33: 321-349.
Madera, J. M., & Smith, D. B. 2009. The effects of leader negative emotions on evaluations
of leadership in a crisis situation: The role of anger and sadness. Leadership Quarterly,
Manstead, A. S. R. 1991. Emotion in social life. In A. S. R. Manstead (Ed.), Emotion in
social life (pp. 353-362). London: Erlbaum.
Mathieu, J. E., & Taylor, S. R. 2007. A framework for testing meso-mediational relationships
in organizational behavior. Journal of Organizational Behavior, 28: 141-172.
Mayer, J. D., Roberts, R. D., & Barsade, S. G. 2008. Human abilities: Emotional intelligence.
Annual Review of Psychology, 59: 507-536.
Melwani, S., & Barsade, S.G. 2011. Held in contempt: The psychological, interpersonal, and
performance consequences of contempt in a Work context. Journal of Personality and
Social Psychology, 101: 503–520.
Milliken, F.J., Morrison, E.W., & Hewlin, P.F. 2003. An exploratory study of employee
silence: Issues that employees don’t communicate upward and why. Journal of
Management Studies, 40: 1453-1476.
Miron-Spektor, E., Efrat-Treister, D., Rafaeli, A., & Schwarz-Cohen, O. 2011. Others’ anger
makes people work harder not smarter: The effect of observing anger and sarcasm on
creative and analytic thinking. Journal of Applied Psychology, 96: 1065-1075.
Morrison, E.W. 2011. Employee voice behavior: Integration and directions for future
research. Academy of Management Annals, 5: 373-412.
Morrison, E.W., & Milliken, F.J. 2000. Organizational silence: A barrier to change and
development in a pluralistic world. Academy of Management Review, 25: 706-725.
Nembhard, I. M., & Edmondson, A. C. 2006. Making it safe: The effects of leader
inclusiveness and professional status on psychological safety and improvement efforts in
health care teams. Journal of Organizational Behavior, 27: 941-966.
Oatley, K., & Johnson-Laird, P. N. 1987. Towards a cognitive theory of emotions. Cognition
and Emotion, 1: 29-50.
Parkinson, B. 1996. Emotions are social. British Journal of Psychology, 87: 663-683.
Podsakoff, P.M., MacKenzie, S.B., Lee, J.Y., & Podsakoff, N.P. 2003. Common method
biases in behavioral research: A critical review of the literature and recommended
remedies. Journal of Applied Psychology, 88: 879-903.
Preacher, K. J., & Kelley, K. 2011. Effect size measures for mediation models: Quantitative
strategies for communicating indirect effects. Psychological Methods, 16: 93-115.
Preacher, K. J., Zyphur, M. J., & Zhang, Z. 2010. A general multilevel SEM framework for
assessing multilevel mediation. Psychological Methods, 15: 209-233.
Raudenbush, S. W., Bryk, A. S., Cheong, Y. F., & Congdon, R. T., Jr. 2004. HLM 6:
Hierarchical linear and nonlinear modeling. Chicago: Scientific Software International.
Rottenstreich, Y., & Hsee, C. K. 2001. Money, kisses, and electric shocks: On the affective
psychology of risk. Psychological Science, 12: 185-190.
Rubin, R. S., Munz, D. C., & Bommer, W. H. 2005. Leading from within: The effects of
emotion recognition and personality on transformational leadership behavior. Academy of
Management Journal, 48: 845-858.
Russell, J.A. 1980. A circumplex model of affect. Journal of Personality and Social
Psychology, 39: 1161-1178.
Ryan, K.D., & Oestreich, D.K. 1998. Driving fear out of the workplace: Creating the high
trust, high performance organization (2nd Eds). San Francisco: Jossey-Bass.
Sauer, S.J. 2011. Taking the reins: The effects of new leader status and leadership style on
team performance. Journal of Applied Psychology, 96: 574-587.
Schwarz, N. 2002. Situated cognition and the wisdom of feelings: Cognitive tuning. In L.
Feldman Barrett & P. Salovey (Eds.), The wisdom in feelings (pp. 144-166). New York:
Schwarz, N., & Clore, G. L. 1983. Mood, misattribution, and judgments of well-being:
Informative and directive functions of affective states. Journal of Personality and Social
Psychology, 45: 513-523.
Schwarz, N., & Clore, G. L. 2003. Mood as information. Psychological Inquiry, 14: 296-303.
Seibert, S.E., Crant, J.M., & Kraimer, M.L. 1999. Proactive personality and career success.
Journal of Applied Psychology, 84: 416-427.
Selig, J. P., & Preacher, K. J. 2008. Monte Carlo method for assessing mediation: An
interactive tool for creating confidence intervals for indirect effects [Computer software].
Available from http://quantpsy.org/.
Song, Z., Foo, M-D., & Uy, M.A. 2008. Mood spillover and crossover among dual-earner
couples: A cell phone event sampling study. Journal of Applied Psychology, 93: 43-452.
Spector, P. E., & Fox, S. 2002. An emotion-centered model of voluntary work behavior:
Some parallels between counterproductive work behavior and organizational citizenship
behavior. Human Resource Management Review, 12: 269-292.
Spreitzer, G. M., & Sonenshein, S. 2004. Toward the construct definition of positive deviance.
American Behavioral Scientist, 47: 828-847.
Sy, T., Côté, S., & Saavedra, R. 2005. The contagious leader: Impact of the leader’s mood on
the mood of group members, group affective tone, and group processes. Journal of
Applied Psychology, 90: 295-305.
Tangirala, S., & Ramanujam, R. 2008. Employee silence on critical work issues: The cross-
level effects of procedural justice climate. Personnel Psychology, 61: 37-68.
Tangirala, S., & Ramanujam, R. 2012. Ask and you shall hear (but not always): An
examination of the relationship between manager consultation and employee voice.
Personnel Psychology, 65: 251-282.
Tellegen, A., Watson, D., & Clark, L. A. 1999. Further support for a hierarchical model of
affect: Reply to Green and Salovey. Psychological Science, 10: 307-309.
Totterdell, P., Kellet, S., Teuchmann, K., & Briner, R. B. 1998. Evidence of mood linkage in
work groups. Journal of Personality and Social Psychology, 74: 1504-1515.
Uhl-Bien, M., & Maslyn, J. M. 2003. Reciprocity in manager-subordinate relationships:
Components, configurations, and outcomes. Journal of Management, 29: 511-532.
Vadera, A.K., Pratt, M.G., & Mishra, P. 2013. Constructive deviance in organizations:
Integrating and moving forward. Journal of Management. 39: 1221-1276.
Van Dyne, L., Cummings, L.L., & McLean Parks, J. 1995. Extra-role behaviors: In pursuit of
construct and definitional clarity. In L.L. Cummings & B.M. Staw (Eds), Research in
organizational behavior, vol.17: 215-285. Greenwich, CT: JAI Press.
Van Dyne, L., Kamdar, D., & Joireman, J. 2008. In-role perceptions buffer the negative
impact of low LMX on helping and enhance the positive impact of high LMX on voice.
Journal of Applied Psychology, 93: 1195-1207.
Van Dyne, L., & Lepine, J.A. 1998. Helping and voice extra-role behaviors: Evidence of
construct and predictive validity. Academy of Management Journal, 41: 108-119.
Van Kleef, G. A. 2009. How emotions regulate social life. Current Directions in
Psychological Science, 18: 184-188.
Van Kleef, G. A., De Dreu, C. K. W., & Manstead, A. S. R. 2004. The interpersonal effects of
emotions in negotiations: A motivated information processing approach. Journal of
Personality and Social Psychology, 87: 510-528.
Van Kleef, G. A., De Dreu, C. K. W., & Manstead, A. S. R. 2010. An interpersonal approach
to emotion in social decision making: The emotions as social information model.
Advances in Experimental Social Psychology, 42: 45-96.
Van Kleef, G.A., Homan, A.C., Beersma, B., & van Knippenberg, D. 2010. On angry leaders
and agreeable followers: How leaders’ emotions and followers’ personalities shape
motivation and team performance. Psychological Science, 21: 1827-1834.
Van Kleef, G. A., Homan, A. C., Beersma, B., van Knippenberg, D., van Knippenberg, B., &
Damen, F. 2009. Searing sentiment or cold calculation? The effects of leader emotional
displays on team performance depend on follower epistemic motivation. Academy of
Management Journal, 52: 562-580.
Walumbwa, F.O., & Schaubroeck, J. 2009. Leader personality traits and employee voice
behavior: Mediating roles of ethical leadership and work group psychological safety.
Journal of Applied Psychology, 94: 1275-1286.
Watson, D., Clark, L.A., & Tellegen, A. 1988. Development and validation of brief measures
of positive and negative affect: The PANAS scales. Journal of Personality and Social
Psychology, 54: 1063-1070.
Westman, M., Vinokur, A.D., Hamilton, V.L., & Roziner, I. 2004. Crossover of marital
dissatisfaction during military downsizing among Russian army officers and their spouses.
Journal of Applied Psychology, 89: 769–779.
Wheeler, L., & Reis, H.T. 1991. Self-recording of everyday life events: origins, types and
uses. Journal of Personality, 59: 339-354.
Wong, C. S., & Law, K. S. 2002. The effects of leader and follower emotional intelligence on
performance and attitude: An exploratory study. Leadership Quarterly, 13: 243-274.
Means, Standard Deviations, and Correlations
Level 1 Variables a
1. Initiation by Leader
2. Initiation by Employee
3. Interaction Quality
10. Psy Safety
a. Episode-level correlations obtained using HLM analyses (n = 640). Reliabilities are reported on the diagonal.
b. LPA refers to “Leader Positive Affect,” EPA refers to “Employee Positive Affect,” ALPA refers to “Assessment of Leader’s Positive Affect,” LNA
refers to “Leader Negative Affect,” ENA refers to “Employee Negative Affect,” ALNA refers to “Assessment of Leader’s Negative Affect,” and Psy
Safety refers to “Psychological Safety.” ** p < .01, * p < .05.
Level 2 and 3 Variables a
1. Dyadic Tenure (month)
2. Interaction Times
3. Employee Proactivity
4. Employee Susceptibility
5. Employee EI
6. Employee PA
7. Employee NA
9. Leader PA
10. Leader NA
a. Below the diagonal are employee-level correlations (n = 85), and above the diagonal are leader-level correlations (n = 36). Reliabilities are reported on
b. Interaction Times refers to “Interaction times during last two weeks,” Proactivity refers to “Proactive Personality,” Employee EI refers to “Employee
Emotional Intelligence,” PA refers to “Positive Affectivity,” NA refers to “Negative Affectivity,” and TL refers to “Transformational Leadership.” **
p < .01, * p < .05
Parameter Estimates and Variance Components of Null Models for Episode-Level Variables a
(e2) / Percentage
Variance (r2) /
(u2) / Percentage
Leader’s positive affect
.36** / 46.7%
.06** / 7.8%
.35** / 45.4%
Employee’s positive affect
.36** / 31.3%
.78** / 67.8%
.01 / .1%
Assessment of leader’s positive affect
.44** / 42.4%
.53** / 51.2%
.07 / 6.4%
Leader’s negative affect
.11** / 53.2%
.00 / 0%
.10** / 46.8%
Employee’s negative affect
.10** / 48.2%
.11** / 51.7%
.00 / .1%
Assessment of leader’s negative affect
.19** / 69.9%
.08** / 30.1%
.00 / .01%
.30** / 50.6%
.22** / 37.2%
.07** / 12.3%
.45** / 61.6%
.06** / 7.8%
.23** / 30.6%
a. n = 640. g00 is the pooled intercept representing the average level of variable across individuals; e2 is the episode-level variance in a variable;
r2 is the employee-level variance in the variable; and u2 is the leader-level variance in the variable. The percentage of the episode-level variance
was computed as e2/( e2 + r2 + u2); the percentage of the employee-level variance was computed as r2/( e2 + r2 + u2); and the percentage of the leader-
was computed as u2/( e2 + r2 + u2).
** p < .01, * p < .05
HLM Regressions a
Level 1: Episode Level
Initiated by Leader
Initiated by Employee
Leaders’ Positive Affect (LPA)
Employee Positive Affect (EPA)
Assessment of Leaders’ Positive Affect (ALPA)
Leader’s Negative Affect (LNA)
Employee Negative Affect (ENA)
Assessment of Leaders’ Negative Affect (ALNA)
Employee Psychological Safety
Pseudo R2 c
Level 2: Employee-Level Main Effects
Employee Proactive Personality
Employee Positive Affectivity
Employee Negative Affectivity
Leader-Member Exchange (LMX)
Level 2: Employee-Level Cross-Level Effects
EPA x LMX
ALPA x LMX
Pseudo R2 c
Level 3: Leader-Level Main Effects
Leader Positive Affectivity
Leader Negative Affectivity
Pseudo R2 c
** p < .01 * p < .05 Note: a n (level 1) = 640, n (level 2) = 85, n (level 3) = 36. Unstandardized coefficients are reported. b EPA refers to “Employee Positive
Affect,” ALPA refers to “Assessment of Leader’s Positive Affect,” ENA refers to “Employee Negative Affect,” and ALNA refers to “Assessment of Leader’s
Negative Affect,” c Pseudo R2 indicates the proportional reduction in the total variance of variables at each level of analysis.
(reported by employee)
(reported by employee)
(reported by leader)
Assessment of Leader’s affect
(reported by employee)
(reported by leader)
(reported by employee)
The Moderating Effect of LMX
on Two Paths Leading to Psychological Safety
1 s.d. above the mean of
Employee’s assessment of
leader’s positive affect
1 s.d. below the mean of
Wu Liu (firstname.lastname@example.org) is an associate professor in the Faculty of Business at the Hong
Kong Polytechnic University. His research interests include leader-member and team dynamics
on extra-role behaviors, emotion, and cross-cultural conflict resolution. He received his Ph.D. in
organization studies at Vanderbilt University.
Zhaoli Song (email@example.com) is an Associate Professor in the Department of Management
and Organization at the School of Business at the National University of Singapore. He received
his Ph.D. in Human Resources and Industrial Relations from the University of Minnesota. His
research interests include behavior genetics, momentary work experience, job search and
unemployment, work-family relationship, leadership, Chinese management, and research
Xian Li's (firstname.lastname@example.org) research interests include leadership, leader-member
interaction, and organizational justice. He received his PhD from the Department of Management
and Organization at the National University of Singapore. He is now working for China Huarong
Asset Management Co., Ltd.
Zhenyu Liao (email@example.com) is a doctoral candidate in the Department of Management
and Organization at the National University of Singapore. His research interests include
leadership, leader-member interaction, abusive leadership behavior, and newcomer socialization.