A Multilevel Field Investigation of Emotional Labor, Affect, Work Withdrawal, and Gender

Abstract

Using experience-sampling methodology, we examined within-individual relationships
among emotional labor, negative and positive affective states, and work withdrawal,
as well as the moderating role of gender. Fifty-eight bus drivers completed two
daily surveys over a two-week period, producing 415 matched surveys. Results of
hierarchical linear models revealed that affective states worsened when employees
engaged in surface acting but improved when they engaged in deep acting. Surface
acting was positively associated with work withdrawal, and state negative affect
mediated this relationship. Results also revealed moderating effects of gender: the
within-individual relationships were stronger for females than for males.

Full text

A MULTILEVEL FIELD INVESTIGATION OF EMOTIONAL
LABOR, AFFECT, WORK WITHDRAWAL, AND GENDER
BRENT A. SCOTT
Michigan State University
CHRISTOPHER M. BARNES
United States Military Academy at West Point
Using experience-sampling methodology, we examined within-individual relation-
ships among emotional labor, negative and positive affective states, and work with-
drawal, as well as the moderating role of gender. Fifty-eight bus drivers completed two
daily surveys over a two-week period, producing 415 matched surveys. Results of
hierarchical linear models revealed that affective states worsened when employees
engaged in surface acting but improved when they engaged in deep acting. Surface
acting was positively associated with work withdrawal, and state negative affect
mediated this relationship. Results also revealed moderating effects of gender: the
within-individual relationships were stronger for females than for males.
As countries such as the United States have in-
creasingly shifted to service-oriented economies
(Wharton, 1993), the emotions and moods (or more
generally, the affect) displayed by members of or-
ganizations has become increasingly important
(Ashforth & Humphrey, 1993; Grandey, 2000, 2003;
Hochschild, 1979, 1983; Morris & Feldman, 1996).
Generally speaking, customer service employees
are expected to conform to integrative display rules
(see Ekman & Friesen, 1969), which stipulate either
explicitly or implicitly that they should express
positive emotions and suppress negative emotions
(e.g., Brotheridge & Grandey, 2002; Hochschild,
1983; VanMaanen & Kunda, 1989). However, em-
ployees naturally experience a variety of emotions
throughout their workdays. As a result, they often
must manage their affective displays in order to
conform to display rules.
Hochschild (1979, 1983) referred to the self-man-
aging of affective displays as emotional labor and
distinguished between two primary forms: surface
acting and deep acting (see also Grandey, 2000).
Surface acting involves attempting to modify affec-
tive displays without changing underlying feelings.
In contrast, deep acting involves attempting to
modify actual feelings to match required displays.
Thus, with surface acting, an employee manipu-
lates or “fakes” the appropriate emotional display;
with deep acting, the employee tries to genuinely
feel the desired emotion.
Cote (2005) noted that a common assumption is
that emotional labor, whether it occurs via surface
acting or deep acting, is good for organizations yet
bad for employees. With the accumulation of re-
search on emotional labor, this assumption has been
challenged, as differences between surface acting and
deep acting, particularly with regards to their effects
on organizational and employee criteria, have be-
come apparent. For example, research has shown that
surface acting is negatively associated, but deep act-
ing is positively associated, with effective emotional
displays during customer interactions (Grandey,
2003). In addition, research has revealed that surface
acting, but not deep acting, is associated with per-
sonal costs such as emotional exhaustion, physical
complaints, and burnout (for a meta-analysis, see
Bono and Vey [2005]). Thus, the effects of emotional
labor appear to depend not only on the type of labor
involved (i.e., surface or deep acting), but also on the
criteria to which it is related.
Although knowledge about emotional labor is
accruing, much remains unknown. For instance,
despite the dynamic interplay between felt and dis-
played emotions in theories of emotional labor
(e.g., Grandey, 2000; Gross, 1998), to our knowl-
edge, no study has examined how affective states
actually experienced by employees change as a re-
sult of engaging in surface acting or deep acting. In
addition, little is known about the effects of emo-
tional labor on employees beyond outcomes such
as strain and burnout, even though theoretical
models specify links between emotional labor and
work withdrawal (Grandey, 2000), which refers to
such behavior as taking longer breaks than permit-
ted, spending work time on personal matters, or
putting less effort into one’s job (e.g., Lehman &
Simpson, 1992). Little is also known about the role
of individual differences in the context of emo-
tional labor, even though Grandey (2000) noted that
individual differences need to be taken into ac-
娀Academy of Management Journal
2011, Vol. 54, No. 1, 116–136.
116
Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s express
written permission. Users may print, download or email articles for individual use only.

count to more fully understand emotional labor.
Indeed, Bono and Vey (2005) called specifically for
future research to examine individual differences
as moderators of the relationships between emo-
tional labor and work outcomes.
Finally, the vast majority of existing research has
been cross-sectional, focusing on static differences
in surface acting and deep acting between individ-
uals at a single point in time. Yet, as Beal, Trouga-
kos, Weiss, and Green (2006) noted, emotions, and,
consequently, emotional labor, are dynamic, and
thus research should examine emotional labor
within individuals over time. This is an important
point, because, typically, in cross-sectional studies,
within-individual variation is implicitly treated as
error variance. These fluctuations, however, may
be meaningful and related to important work crite-
ria. To date, a few studies have taken a within-
individual approach to the study of emotional la-
bor, linking daily engagement in surface acting to
lower well-being in the form of emotional exhaus-
tion, negative affect, and job dissatisfaction (Judge,
Woolf, & Hurst, 2009; Totterdell & Holman, 2003)
and deep acting to the display of positive emotions
and self-rated job performance (Totterdell & Hol-
man, 2003). Indeed, Judge et al. (2009) found that
39.3 and 31.7 percent of the variance in surface
acting and deep acting, respectively, was within
individuals. These noteworthy studies suggest that
within-individual variation in emotional labor is
systematic and important to consider.
According to Kozlowski and Klein (2000), a mean-
ingful understanding of workplace phenomena re-
quires integrative approaches that traverse multiple
levels. Thus, if researchers are to more fully under-
stand emotional labor, theory and research must be
extended not only to take into account emotional
labor processes occurring within individuals, but also
to consider the ways in which between-individual
differences influence these within-person processes.
With this in mind, we aimed in the current study to
extend theory and research on emotional labor by
taking a multilevel approach and examining (1) the
effects of emotional labor on affective states and work
withdrawal within individuals over time and (2) the
moderating role of gender on these within-individual
relationships. Using Grandey’s (2000) model of emo-
tional labor as an overarching theoretical framework,
we hypothesized that when employees engage in sur-
face acting, they experience an increase in negative
affect and a decrease in positive affect and, conse-
quently, are more likely to report withdrawing from
work. In contrast, we hypothesized that when em-
ployees engage in deep acting, they experience a de-
crease in negative affect and an increase in positive
affect and, consequently, are less likely to report
withdrawing from work. Moreover, in accordance
with Grandey’s (2000) theorizing that emotional labor
relationships may vary by gender, we hypothesized
that these within-individual relationships are stron-
ger for females than males. Figure 1 presents a model
of our hypothesized relationships.
FIGURE 1
Hypothesized Multilevel Model of the Relationships among Emotional Labor, Affect,
Work Withdrawal, and Gender
2011 117Scott and Barnes

Overall, our multilevel approach responds si-
multaneously to calls for research on emotional
labor processes within individuals (Bono & Vey,
2005; Gosserand & Diefendorff, 2005) and calls for
research on the extent to which individual differ-
ences influence those processes (Bono & Vey,
2005). In doing so, we advance theory by demon-
strating the importance of taking into account the
dynamic, within-individual aspect of emotional
labor and by explicating how individual differ-
ences emphasized in extant theories of emotional
labor, such as gender (Grandey, 2000), may, in a
top-down manner, influence the strength of with-
in-individual emotional labor relationships. The
examination of gender in concert with intraindi-
vidual emotional labor processes links between-
and within-individual levels of analysis and
therefore extends the range of existing theory. In
addition, we advance practice by providing em-
ployees and managers with guidance on how to
facilitate the cultivation of desired emotional and
work outcomes on day-to-day basis and by illumi-
nating how men and women may best benefit from
emotional labor. Theories of affect have tended to
focus on external events as elicitors of emotion
(e.g., Weiss & Cropanzano, 1996). Here, we con-
sider a more internal mechanism (i.e., employee
choice of emotional labor strategy) whereby em-
ployees may alter their own experienced states and
subsequent perceptions about their actions at work.
Having provided an overview of our model, we
present the theoretical logic for each hypothesis in
the sections below.
HYPOTHESES
Emotional Labor and Affect
Despite being recruited, selected, socialized, and
rewarded for conforming to display rules (Rafaeli &
Sutton, 1987), employees may not actually experi-
ence the affect that display rules prescribe. Emo-
tion-eliciting events may occur at work (Weiss &
Cropanzano, 1996), or they may occur at home and
emotions may spill over into work (Judge & Ilies,
2004), or individuals may experience diffuse mood
states that are unconnected to specific events (Wat-
son, 2000). Regardless of what precipitates affect,
there are days and times when employees’ feelings
are at odds with the display requirements of their
organization. As noted above, employees can at-
tempt to close the gap between what they feel and
what they are supposed to feel either by faking
appropriate affective displays (surface acting) or by
actively modifying underlying affective states to
match display requirements (deep acting) (Hoch-
schild, 1979, 1983).
Integrating surface acting and deep acting with
Gross’s (1998) taxonomy of emotion regulation,
Grandey’s (2000) model of emotional labor posi-
tions surface acting as a form of response-focused
emotion regulation, whereby affective states are
manipulated after they have commenced. In con-
trast, deep acting is a form of antecedent-focused
emotion regulation, whereby affective states are
manipulated before they have commenced. Ac-
cording to Grandey (2000), with surface acting, in-
dividuals may attempt to suppress unwanted feel-
ings by simply faking appropriate displays. With
deep acting, individuals may deploy their attention
elsewhere by conjuring thoughts to elicit desired
affective states, or they may change their cognitive
perspective by reappraising their situation. The
end result of such efforts is that surface acting does
nothing to change underlying emotions, whereas
deep acting actually creates emotions that are com-
mensurate with display rules. Consequently, emo-
tional dissonance, which is a sense of tension that
occurs when experienced and displayed affect di-
verge (Hochschild, 1983), is exacerbated by surface
acting but diminished by deep acting (Grandey,
2003).
From an affective display standpoint, both sur-
face acting and deep acting lead to the same result:
namely, the expression of affective states that con-
form to display rules, which are typically “integra-
tive” in customer service settings, meaning that
positive emotions such as enthusiasm, interest, and
cheer are encouraged, and negative emotions such
as anger, hostility, and distress are discouraged
(Beal et al., 2006; Hochschild, 1983). However, the
affective states actually experienced in the two
forms of emotional labor differ. When individuals
surface act, they do nothing to change their under-
lying affective states, and indeed, those affective
states may actually worsen as a result of this emo-
tional labor strategy. Research has shown that at-
tempting to suppress negative emotions may have
the ironic effect of causing individuals to think
about the eliciting situation even more, which ul-
timately intensifies those negative feelings
(Wegner, 1994). In contrast, deep acting by defini-
tion produces a change in felt affect. It follows that,
to the extent that display rules emphasize positive
affective displays, surface acting throughout a
given day should increase negative affect and de-
crease positive affect. In contrast, deep acting
throughout a given day should decrease negative
affect and increase positive affect.
118 FebruaryAcademy of Management Journal

Hypothesis 1a. Within individuals, daily sur-
face acting is associated with an increase in
negative affect.
Hypothesis 1b. Within individuals, daily sur-
face acting is associated with a decrease in
positive affect.
Hypothesis 2a. Within individuals, daily deep
acting is associated with a decrease in negative
affect.
Hypothesis 2b. Within individuals, daily
deep acting is associated with an increase in
positive affect.
Emotional Labor, Affect, and Work Withdrawal
According to Grandey’s (2000) model, work
withdrawal is an outcome of emotional labor.
Drawing on our above arguments regarding the re-
lationships between emotional labor and experi-
enced affective states, we propose that employees
are more likely to withdraw from work when they
surface act throughout a day but are less likely to
withdraw from work when they deep act through-
out the day. To the extent that surface acting pro-
duces a change in affect for the worse (i.e., in-
creased negative affect and decreased positive
affect), surface acting should be positively associ-
ated with work withdrawal. In contrast, to the ex-
tent that deep acting produces a change in affect for
the better (i.e., decreased negative affect and in-
creased positive affect), deep acting should be neg-
atively associated with work withdrawal. Below,
we draw from the emotions literature to support
these propositions.
A fundamental tenet of theories of emotion is
that individuals’ behaviors are influenced by how
they feel. Specifically, affective experiences are ac-
companied by action tendencies, which are impul-
sive, automatic urges to achieve a particular goal
(Frijda, 1994, 2007). By taking “control prece-
dence” relative to other concerns (Frijda, 1994,
2007), affective states energize and prioritize be-
haviors (Elfenbein, 2007).
As Elfenbein (2007) noted, the action tendencies
of negative affective states address current prob-
lems to improve the situation, which is accom-
plished through avoidance (as opposed to ap-
proach) in either the short or long term (Cacioppo,
Gardner, & Bernston, 1999; Fitness, 2000). Applied
to a work context, the notion of action tendencies
implies that employees should report higher levels
of work withdrawal on days in which they experi-
ence negative affective states. Research on related
outcomes such as counterproductive work behav-
iors supports this idea, showing that such out-
comes are positively associated with negative affect
(for a review, see Elfenbein [2007]). Although un-
desired by their organization, work withdrawal
may be perceived as constructive by employees
insofar as it serves as a form of coping, a way to
calm negative affective states to return to a desired
baseline state (Grandey & Brauburger, 2002).
In contrast, positive affective states prompt ap-
proach rather than avoidance behavior by stimulat-
ing creativity and search as well as by expanding
individuals’ action repertoires (Cacioppo et al.,
1999; Fredrickson, 2001). Given that individuals
experiencing positive affective states engage their
environments, it follows that, in work contexts,
employees should report lower levels of work with-
drawal on days when they experience positive af-
fective states. Here again, research indicating that
related outcomes such as counterproductive work
behaviors are negatively associated with positive
affect provides some support for this idea (Elfen-
bein, 2007).
In view of the above, we propose that employees
will report higher levels of work withdrawal on
days when they experience high levels of negative
affect and lower levels of work withdrawal on days
when they experience high levels of positive affect.
Furthermore, we propose that “state negative” and
“state positive” affect transmit the effects of daily
surface acting and deep acting on reports of work
withdrawal.
1
According to Frijda (1994, 2007), af-
fective states are proximal drivers of action. Thus,
daily surface acting should be positively associated
with reports of work withdrawal because this emo-
tional labor strategy is associated with worsening
affect. In contrast, daily deep acting should be neg-
atively associated with reports of work withdrawal
because this emotional labor strategy is associated
with improved affect. Some indirect empirical ev-
idence supports this notion, as Judge et al. (2009)
found that state negative affect mediated the
within-individual relationships between surface
acting and both emotional exhaustion and job
dissatisfaction.
We expect this mediation to be partial rather than
full because reasons in addition to affect may ex-
plain associations between emotional labor and
work withdrawal. For example, surface acting re-
quires high levels of self-regulation and thus de-
1
Here, our concern is with affective “states,” which
refer to more transient, temporary feelings, as opposed to
affective “traits,” which refer to more enduring individ-
ual differences in the proclivity to experience particular
states (e.g., Elfenbein, 2007).
2011 119Scott and Barnes

pletes self-regulatory resources (Baumeister, Mu-
raven, & Tice, 2000; Beal et al., 2006; Richards &
Gross, 1999). Consequently, employees may utilize
work withdrawal as a form of rest, allowing them to
return to self-regulatory demands with renewed
strength (Muraven & Baumeister, 2000).
Hypothesis 3. Within individuals, daily surface
acting is positively related to perceptions of
work withdrawal.
Hypothesis 4. Within individuals, daily deep
acting is negatively related to perceptions of
work withdrawal.
Hypothesis 5a. Within individuals, negative af-
fect partially mediates the positive relation-
ship between daily surface acting and percep-
tions of work withdrawal.
Hypothesis 5b. Within individuals, positive
affect partially mediates the positive relation-
ship between daily surface acting and percep-
tions of work withdrawal.
Hypothesis 6a. Within individuals, negative
affect partially mediates the negative relation-
ship between daily deep acting and percep-
tions of work withdrawal.
Hypothesis 6b. Within individuals, positive
affect partially mediates the negative relation-
ship between daily deep acting and percep-
tions of work withdrawal.
Gender as a Boundary Condition on Emotional
Labor Effects
A large body of work has been devoted to under-
standing gender differences in the experience and
expression of emotion (for a review, see Brody and
Hall [2008]), making gender a natural individual
difference to examine in concert with emotional
labor (Grandey, 2000; Morris & Feldman, 1996).
Drawing from this literature, we propose that gen-
der moderates the within-individual relationships
among emotional labor, affect, and work with-
drawal in such a way that the relationships are
stronger for females than males.
A relatively consistent finding in the emotions
literature is that women are both expected to and
do show greater emotional intensity and emotional
expressiveness than men, and such differences
hold for both positive and negative emotions
(Brody & Hall, 2008). The root of such differences
may lie in role development, whereby females are
socialized to be more emotionally expressive and
men are socialized to be more emotionally re-
strained (Eagly, 1987; Grossman & Wood, 1993;
Kring & Gordon, 1998). Socialization pressures also
influence the types of emotions that females and
males are expected to express. Positive, relation-
ship-facilitating emotions such as warmth and
cheer are considered more role-appropriate for
women than men, whereas negative, distancing
emotions such as anger and hostility are viewed as
more role-appropriate for men than women (Brody,
1999; Brody & Hall, 2008; Simpson & Stroh, 2004).
According to poststructuralist feminist theorists
(e.g., Weedon, 1987; see also Mumby & Putnam,
1992), these expectations carry over into work-
places and are perpetuated by organizational norms
and practices that construct different emotional
roles for women and men.
These gender differences should have implica-
tions for the within-individual relationships hy-
pothesized above. Beginning with surface acting,
when women attempt to mask or fake an emotion,
they should experience greater emotional disso-
nance because their actions are at odds with their
tendency to display and express what they are ac-
tually feeling. Men, however, should experience
less emotional dissonance, because they are more
accustomed to hiding emotions from others and
faking affective states (Ashmore, Del Boca, &
Wohlers, 1986; Fabes & Martin, 1991). Indeed,
Kruml and Geddes (1998) found that women were
more likely to experience emotional dissonance
than men. Moreover, when display requirements
call for the expression of positive emotion, the use
of surface acting should be especially detrimental
to women, because the persistence of an underlying
negative affective state conflicts with gender role
requirements to display positive emotion. Taken
together, these findings suggest that the adverse
effects of daily surface acting on both experienced
affective states (i.e., increased negative affect, de-
creased positive affect) and reports of work with-
drawal should be greater for women than men. On
this point, in a between-individual study, Johnson
and Spector (2007) found that surface acting was
more strongly associated with decreased well-being
(e.g., emotional exhaustion, job dissatisfaction) for
females than males.
In contrast to surface acting, deep acting should
be more beneficial for females than males. In a
customer service context, the cultivation of posi-
tive affective states via deep acting should not only
be more role-appropriate for women than for men,
but also, to the extent that women experience such
states more intensely (Brody & Hall, 2008), deep
acting should yield a greater gain for women than
for men. In addition, research has shown that
women are more adept than men at producing “au-
thentic” smiles (Merton, 1997), implying that
120 FebruaryAcademy of Management Journal

women are better at deep acting than men (see also
Johnson & Spector, 2007). Overall, this discussion
suggests that the beneficial effects of daily deep
acting on both experienced affective states (i.e.,
decreased negative affect, increased positive affect)
and reports of work withdrawal should be greater
for women than men.
Finally, as noted above, by taking control prece-
dence relative to other matters, affective states
stimulate action. This is especially so when those
affective states are high in intensity (Frijda, 2007).
Although a variety of factors contribute to the in-
tensity of affective states, including bodily arousal
and felt action tendencies, rumination over affect-
eliciting events appears to be a primary driver of
the intensity of both positive and negative affect
(Frijda, 2007). Interestingly, there is evidence that
women engage in rumination more than men (No-
len-Hoeksema & Jackson, 2001), suggesting a reason
underlying women’s greater intensity in both pos-
itive and negative affective states (Brody & Hall,
2008). To the extent that action tendencies are
stronger under conditions of higher emotional in-
tensity, it follows that women, compared to men,
should be more likely to report withdrawing from
work on days in which they experience negative
affective states and less likely to report withdraw-
ing from work on days when they experience pos-
itive affective states. Overall, given the above, we
hypothesize the following cross-level moderating
effects of gender:
Hypothesis 7a. The within-individual relation-
ship between daily surface acting and percep-
tions of work withdrawal is stronger for fe-
males than for males.
Hypothesis 7b. The within-individual relation-
ship between daily deep acting and percep-
tions of work withdrawal is stronger for fe-
males than for males.
Hypothesis 8a. The within-individual relation-
ship between daily surface acting and negative
affect is stronger for females than for males.
Hypothesis 8b. The within-individual relation-
ship between daily deep acting and negative
affect is stronger for females than for males.
Hypothesis 9a. The within-individual relation-
ship between daily surface acting and positive
affect is stronger for females than for males.
Hypothesis 9b. The within-individual relation-
ship between daily deep acting and positive
affect is stronger for females than for males.
Hypothesis 10a. The within-individual rela-
tionship between negative affect and percep-
tions of work withdrawal is stronger for fe-
males than for males.
Hypothesis 10b. The within-individual rela-
tionship between positive affect and percep-
tions of work withdrawal is stronger for fe-
males than for males.
Considered together, the above hypothesized pat-
tern of moderation implies moderated mediation,
whereby a mediated effect depends on the level of
a third variable (see Edwards & Lambert, 2007).
Specifically, to the extent that (1) the relationships
between surface acting, deep acting, and reports of
work withdrawal are stronger for females, (2) the
relationships between surface acting, deep acting,
and affective states are stronger for females, and (3)
the relationships between affective states and re-
ports of work withdrawal are stronger for females,
the mediated effects (i.e., indirect effects) of surface
acting and deep acting on work withdrawal
through state negative and positive affect may de-
pend on gender, so that the indirect effects are
stronger for females than males. Thus, we
hypothesize:
Hypothesis 11a. The mediated (indirect) ef-
fects of daily surface acting and deep acting on
perceptions of work withdrawal through state
negative affect are stronger for females than
males.
Hypothesis 11b. The mediated (indirect) ef-
fects of daily surface acting and deep acting on
perceptions of work withdrawal through state
positive affect are stronger for females than
males.
METHODS
Sample
Participants were 68 bus drivers working for the
same transportation company in the northwestern
United States. The sample was comprised of 25
females and 43 males; the average age was 48.4
years (s.d. ⫽8.6 years), and reported ethnicities
were as follows: African American (16 partici-
pants), Asian/Pacific Islander (2 participants), His-
panic/Latino (1 participant), white/Caucasian (44
participants); 3 participants chose an “other” cate-
gory, and 2 participants did not report ethnicity.
Bus service was provided all hours of the day
except 1:00 am to 2:45 am. Thus, bus drivers
worked various versions of early morning, after-
noon, and evening shifts. Drivers spent the over-
2011 121Scott and Barnes

whelming majority of their work time on buses and
had personal contact with every customer who en-
tered their bus. At a minimum, this contact con-
sisted of taking fare and offering a transfer slip.
Drivers also answered questions about schedules
and stops, gave personal greetings to riders upon
their entering and exiting, and had conversations
with riders while driving the bus. Because of this
frequent customer contact, this sample provided a
good setting in which to study the effects of emo-
tional labor.
Procedures
Participants were recruited through an organiza-
tional contact. Individuals interested in participat-
ing were directed to a web site containing an in-
formed consent form assuring confidentiality as
well as instructions on how to participate. We col-
lected all data using online surveys. Participants
were first asked to complete a baseline survey as-
sessing demographic characteristics. To capture
drivers’ emotional labor, affective states, and work
withdrawal daily, we employed interval-contin-
gent experience-sampling methodology (see Alliger
& Williams, 1993; Wheeler & Reis, 1991). We chose
a two-week period for the experience-sampling
phase, as Reis and Wheeler (1991) suggested that
two weeks represents a generalizable sample of in-
dividuals’ lives. During the two weeks (workdays
only), drivers were asked to complete two surveys
each day. The first survey, completed upon drivers’
arriving to work, before they had begun their
routes, assessed state negative and state positive
affect. The second survey, completed before drivers
left work, after they had completed their routes,
assessed surface acting and deep acting, state neg-
ative and state positive affect, and work withdrawal
occurring during work that day. Although not of
course establishing causality, by assessing affect
upon drivers’ arrival at work each day, we were
able to assess whether surface acting and deep act-
ing were associated with changes in their daily
affective states—that is, whether surface acting and
deep acting were associated with state negative and
state positive affect, with these affective states at
the beginning of the workday controlled for. Thus,
we included daily data for a given driver only if he
or she completed both the beginning- and end-of-
work surveys. Reminders placed in the drivers’
mailboxes each day of the study encouraged com-
pletion of the daily surveys, and drivers were en-
tered into a random drawing for five $100 prizes
that were distributed at the end of the study. Of the
68 drivers who volunteered for the study and com-
pleted the initial baseline survey, 58 took part in
the experience-sampling phase. An independent
samples t-test comparing mean levels of gender
between participants who took part in the experi-
ence-sampling phase and those who did not re-
vealed no significant differences. We obtained 425
matched beginning- and end-of-work surveys out of
a possible 580, an average of 7 matched surveys per
driver, which converts to a response rate of 73.3
percent. We used “timestamps” to assess whether
drivers adhered to the study instructions. Inspec-
tion of these timestamps revealed 10 instances in
which a driver completed the end-of-work survey
immediately after the beginning-of-work survey.
We thus excluded these 10 observations from the
analyses. The remaining 415 matched surveys were
completed as instructed, and the average time
elapsed between completion of the beginning- and
end-of-work surveys was 9 hours and 47 minutes
(s.d. ⫽2 hours, 47 minutes).
Measures
Surface acting and deep acting. To assess the
degree to which drivers engaged in surface and
deep acting each day, we used the scales developed
by Grandey (2003). At the end of each workday,
drivers were asked to indicate how often that day
they had engaged in the actions listed using a scale
ranging from 1, “almost never,” to 5, “very often.”
Examples from the five-item surface acting scale
included “Put on an act in order to deal with cus-
tomers in an appropriate way” and “Faked a good
mood.” Examples from the three-item deep acting
scale included “Tried to actually experience the
emotions I must show” and “Made an effort to
actually feel the emotions that I needed to display
toward others.” Coefficient alpha for each scale,
averaged over the days of data collection, was .97
for surface acting and .94 for deep acting.
State negative and state positive affect. Daily
affective states experienced by the drivers were
assessed in both the beginning- and end-of-work
surveys using the Positive and Negative Affect
Scale (PANAS; Watson, Clark, & Tellegen, 1988).
At the beginning of work, drivers were asked to
indicate the extent to which they were experienc-
ing each state using a scale ranging from 1, “very
slightly or not at all,” to 5, “very much,” thus cre-
ating a baseline measure of affect before their shift.
Using the same scale at the end of work, drivers
were asked to indicate the extent to which they
experienced each state at work that day. Examples
from the ten-item negative affect scale included
“irritable,” “distressed,” and “upset.” Examples
from the ten-item positive affect scale included
“enthusiastic,” “excited,” and “attentive.” Coeffi-
122 FebruaryAcademy of Management Journal

cient alpha for each scale, averaged over the days of
data collection, was: beginning-of-work survey, .72
for negative affect and .95 for positive affect; end-
of-work survey, .70 for negative affect and .95 for
positive affect.
Work withdrawal. Perceptions of daily work
withdrawal were assessed with the psychological
withdrawal scale developed by Lehman and Simp-
son (1992). To keep the survey brief, we a priori
eliminated four items that appeared less relevant to
our sample (i.e., bus drivers). Examples of excluded
items were “Let others do your work” and “Spent
work time on personal matters,” as it seemed un-
likely that bus drivers’ daily withdrawal could en-
compass such actions. The four items chosen for
inclusion were “Thought about being absent,” “Put
less effort into the job than you should have,”
“Thought about leaving current job,” and “Day-
dreamed.” At the end of each workday, drivers
were asked to indicate how often that day they
engaged in the items listed using a scale ranging
from 1, “almost never,” to 5, “very often” (
␣
⫽.78).
Analyses
Given the hierarchical nature of the data set (i.e.,
days were nested within individuals), we used hi-
erarchical linear modeling (HLM; see Raudenbush
& Bryk, 2002) to test our hypotheses. The data
consisted of two levels. The lower level (level 1)
comprised the repeated, daily assessments of driv-
ers’ state negative and positive affect, surface acting
and deep acting, and work withdrawal. The upper
level (level 2) comprised drivers’ gender. Thus,
the level 1 data could vary within individuals, and
the level 2 data could vary between individuals.
To test the hypothesized within-individual rela-
tionships among emotional labor (i.e., surface act-
ing and deep acting), state affect, and work with-
drawal (Hypotheses 1a through 6b), we regressed
each outcome variable on the predictor variables,
with all level 1 predictors centered at individuals’
means (i.e., group-mean-centered), following the
recommendation of Hoffmann, Griffin, and Gavin
(2000). To test the hypothesized cross-level moder-
ating effects of gender on the within-individual
relationships (Hypotheses 7a through 10b), we
added gender to the above equations as a level 2
predictor of the intercept and slope of each level 1
relationship of interest (e.g., the within-individual
relationship between surface acting and work with-
drawal). To test the moderated mediation hypoth-
eses (Hypotheses 11a and 11b), we compared the
indirect effects for women and men, following Ed-
wards and Lambert (2007).
As noted above, the drivers comprising our sam-
ple worked different shifts and thus did not begin
and end their routes at the same time every day.
Research has revealed significant diurnal variation
in affect, especially positive affect, which rises
sharply from morning until noon, remains steady,
and then falls in the evening (Clark, Watson, &
Leeka, 1989; Watson, 2000). As a result, any ob-
served differences in drivers’ daily affective states
could be attributed to their natural circadian
rhythms rather than to our hypothesized predictors
(i.e., surface acting and deep acting). To eliminate
this possibility, in all analyses we controlled for the
time of day that drivers began by including the time
that the beginning-of-work survey was completed
each day as a level 1 predictor (“start time”). In
addition, in all analyses we controlled for both age
and ethnicity at level 2.
RESULTS
Descriptive Statistics and Correlations
Means, standard deviations, and correlations are
shown in Table 1. Within-individual correlations
are above the diagonal, between-individual corre-
lations are below, and level 1 variables are aggre-
gated over the two-week period into single scores
for each driver. Although females were more likely
than males to experience state negative affect (r⫽
.27, p⬍.05), there were no significant gender dif-
ferences in the average daily levels of the remaining
variables (e.g., surface acting and deep acting).
Partitioning of Variance within and
between Individuals
Before testing our hypotheses, we inspected the
results of null models in HLM (regressions with no
level 1 or 2 predictors) for each endogenous level 1
variable to determine whether there was within-
individual variance to explain. Null models sepa-
rate the variance in a given level 1 variable within
and between individuals, and the intercept repre-
sents the average level (i.e., mean) of the variable
for the days of data collection. A lack of within-
individual variance in the outcome variables and
the presence of only between-individual variance
would have indicated that HLM was inappropriate
because there was only one level of variance (be-
tween-individual) to explain.
Table 2 shows the results for each null model,
indicating a significant amount of between-individ-
ual variance in each outcome. However, the out-
comes also varied within individuals, as 14.3 per-
cent of the variance in surface acting, 23.0 percent
2011 123Scott and Barnes

of the variance in deep acting, 52.3 percent of the
variance in state negative affect, 17.2 percent of the
variance in state positive affect, and 15.1 percent of
the variance in perceptions of work withdrawal
was within individual. Overall, although the
amount of within-individual variance was smaller
than the amount of between-individual variance for
some of the outcomes, the above results suggest
that HLM was appropriate and that there was with-
in-individual variance to be explained.
Tests of Hypotheses
Within-individual hypotheses. Hypothesis 1a
predicts that daily surface acting is associated with
increased negative affect within individuals, and
Hypothesis 1b predicts that daily surface acting is
associated with decreased positive affect within
individuals. Hypothesis 2a predicts that daily deep
acting is associated with decreased negative affect
within individuals, and Hypothesis 2b predicts
that daily deep acting is associated with increased
positive affect within individuals. Table 3 provides
the results of the HLM regressions testing these
hypotheses. Surface acting was positively associ-
ated with state negative affect (b
40
⫽0.08, p⬍.05)
but was not significantly associated with state pos-
itive affect (b
40
⫽⫺0.00), supporting Hypothesis 1a
but not Hypothesis 1b. Deep acting was negatively
associated with state negative affect (b
50
⫽⫺0.06,
p⬍.05) and positively associated with state posi-
tive affect (b
50
⫽0.12, p⬍.05), supporting Hypoth-
eses 2a and 2b. Given that we controlled for state
negative affect and state positive affect at the start
of the drivers’ workdays, these results can be inter-
preted as showing change in their affective states.
Overall, drivers’ negative affect worsened on days
when they engaged in surface acting. However, on
days when they engaged in deep acting, their affec-
tive states improved, and they experienced less
negative affect and more positive affect.
Hypothesis 3 predicts that, within individuals,
daily surface acting is positively associated with
perceptions of work withdrawal. Hypothesis 4 pre-
TABLE 2
Parameter Estimates and Variance Components of Null Models for Level 1 Endogenous Variables
a
Variable
Intercept
b
00
Within-Individual
Variance (e
2
)
Between-Individual
Variance (r
2
)
Percentage of
Within-Individual
Variance
Surface acting 1.85* .14 0.84* 14.3%
Deep acting 2.66* .38 1.25* 23.0%
State negative affect 1.23* .08 0.07* 52.3%
State positive affect 2.80* .17 0.81* 17.2%
Work withdrawal 1.77* .09 0.48* 15.1%
a
n⫽415. b
00
is the pooled intercept representing average level of variable across individuals; e
2
is the within-individual variance in
a variable; and r
2
is the between-individual variance in the variable. The percentage of variance within-individuals was computed as
e
2
/(e
2
⫹r
2
).
*p⬍.05
TABLE 1
Descriptive Statistics and Correlations
a
Variable Mean s.d. 1 2 3 4 5 6 7
1. State negative affect, start of work 1.17 0.24 ⫺.10 .27* .00 .16 ⫺.02 .10
2. State positive affect, start of work 2.95 0.88 ⫺.15 ⫺.07 ⫺.01 ⫺.10 .32* ⫺.10
3. Surface acting 1.85 0.93 .23 ⫺.63* .07 .25* .01 .37*
4. Deep acting 2.66 1.15 ⫺.01 ⫺.05 .19 ⫺.22* .21* ⫺.08
5. State negative affect 1.24 0.34 .80* ⫺.15 .27* ⫺.09 .02 .17*
6. State positive affect 2.80 0.91 ⫺.13 .91* ⫺.62* ⫺.03 ⫺.18 ⫺.05
7. Work withdrawal 1.77 0.70 .22 ⫺.57* .71* .10 .31* ⫺.55*
8. Gender
b
0.34 0.48 .22 ⫺.10 .19 ⫺.23 .27* ⫺.14 .21
a
Variables 1 through 7 are within-individual (level 1) variables. Variable 8 is a between-individual (level 2) variable. Within-individual
correlations are shown above the diagonal (n⫽415); between-individual correlations are shown below the diagonal, with within-
individual variables aggregated to the between-individual level (n⫽58). Means and standard deviations are based on between-individual
scores.
b
“Female” ⫽1, “male” ⫽0.
*p⬍.05
124 FebruaryAcademy of Management Journal

dicts that, within individuals, daily deep acting is
negatively associated with perceptions of work
withdrawal. Table 4 provides results of the HLM
regressions testing these predictions, showing that
surface acting was positively associated with work
withdrawal (b
40
⫽0.20, p⬍.05), supporting Hy-
pothesis 3. Although the coefficient for deep acting
was negative, it was not significant (b
50
⫽⫺0.04),
failing to support Hypothesis 4. Thus, drivers re-
ported higher levels of work withdrawal on days
when they engaged in surface acting, but their deep
acting was not associated with work withdrawal.
Hypothesis 5a predicts that state negative affect
partially mediates the within-individual relation-
ship between daily surface acting and perceptions
of work withdrawal, and Hypothesis 5b predicts
that state positive affect partially mediates the
within-individual relationship between daily sur-
face acting and perceptions of work withdrawal.
Hypothesis 6a predicts that state negative affect
partially mediates the within-individual relation-
ship between daily deep acting and perceptions of
work withdrawal, and Hypothesis 6b predicts that
state positive affect partially mediates the within-
individual relationship between daily deep acting
and perceptions of work withdrawal. As shown in
Table 4, state negative affect (b
60
⫽0.20, p⬍.05),
but not state positive affect (b
70
⫽0.02), was sig-
nificantly associated with drivers’ reports of work
withdrawal, and the coefficients for surface acting
and deep acting decreased slightly when these af-
fective states were controlled. To determine specif-
ically which relationships were mediated, we ex-
amined the significance of the indirect effects of
TABLE 3
HLM Results Predicting State Negative Affect and State Positive Affect
a
Predictor
Outcome: State Negative Affect Outcome: State Positive Affect
bs.e. tbs.e. t
Intercept (b
00
)1.12 0.22 5.15* 1.54 0.63 2.44*
Gender (b
01
)0.09 0.08 1.12 ⫺0.11 0.19 ⫺0.57
Age (b
02
)0.00 0.00 0.46 0.03 0.01 2.44*
Ethnicity (b
03
)0.01 0.06 0.20 0.11 0.25 0.44
Start time (b
10
)0.01 0.00 2.95* ⫺0.03 0.01 ⫺4.89*
State negative affect, start of work (b
20
)0.20 0.15 1.33 0.19 0.14 1.38
State positive affect, start of work (b
30
)⫺0.02 0.04 ⫺0.38 0.33 0.06 5.81*
Surface acting (b
40
)0.08 0.03 2.39* ⫺0.00 0.06 ⫺0.07
Deep acting (b
50
)⫺0.06 0.02 ⫺3.25* 0.12 0.04 3.16*
a
All level 1 predictors were centered at individuals’ means; n⫽415. Gender was coded 0, “male,” 1, “female.” Ethnicity was coded
0, “Caucasian,” 1, “other.”
*p⬍.05
TABLE 4
HLM Results Predicting Work Withdrawal
a
Predictor
Direct Effects Model Mediated Model
bs.e. tbs.e. t
Intercept (b
00
)1.47 0.39 3.80* 1.64 0.38 4.34*
Gender (b
01
)0.40 0.18 2.21* 0.35 0.17 2.04*
Age (b
02
)⫺0.00 0.01 ⫺0.26 ⫺0.01 0.01 ⫺0.78
Ethnicity (b
03
)⫺0.36 0.17 ⫺2.09* ⫺0.40 0.17 2.39*
Start time (b
10
)⫺0.01 0.01 ⫺1.51 ⫺0.01 0.01 ⫺1.97
State negative affect, start of work (b
20
)0.04 0.06 0.55 ⫺0.05 0.08 ⫺0.66
State positive affect, start of work (b
30
)⫺0.04 0.04 ⫺0.84 ⫺0.05 0.04 ⫺1.15
Surface acting (b
40
)0.20 0.05 4.26* 0.17 0.04 4.35*
Deep acting (b
50
)⫺0.04 0.03 ⫺1.24 ⫺0.01 0.02 ⫺0.53
State negative affect (b
60
)0.20 0.04 5.58*
State positive affect (b
70
)0.02 0.04 0.45
a
All level 1 predictors were centered at individuals’ means; n⫽415. Gender was coded 0, “male,” 1, “female.” Ethnicity was coded
0, “Caucasian,” 1, “other.”
*p⬍.05
2011 125Scott and Barnes

surface acting and deep acting on work withdrawal
through state negative and state positive affect. In
their discussion of mediation in multilevel models,
Krull and MacKinnon (1999) recommended the use
of the first-order Taylor series expansion (Sobel,
1982; see also MacKinnon, Lockwood, Hoffman,
West, & Sheets, 2002) to estimate the standard error
of the indirect effect in cases in which the level 2
sample size is 50. Given that our level 2 sample size
was 58, we followed their recommendation. Re-
sults of our analyses revealed significant indirect
effects of both surface acting (z⫽2.20, p⬍.05) and
deep acting (z⫽2.81, p⬍.05) on perceptions of
work withdrawal through state negative affect. The
indirect effects of surface and deep acting on per-
ceptions of work withdrawal through state positive
affect were not significant, which was to be ex-
pected given that state positive affect was not sig-
nificantly associated with drivers’ reports of work
withdrawal. Thus, Hypotheses 5a and 6a, but not
5b and 6b, were supported.
We should note that although the indirect effect
of deep acting on perceptions of withdrawal
through state negative affect was significant, the
total effect of deep acting on perceptions of with-
drawal was not (see above). Judd and Kenny (1981)
acknowledged the possibility of mediation without
a significant total effect (see also MacKinnon, Krull,
& Lockwood, 2000). In the case of the current re-
sults, this finding is likely the result of two indirect
effects of opposite sign (i.e., the indirect effect of
deep acting on perceptions of work withdrawal
through state negative affect was negative in sign,
while the indirect effect of deep acting on percep-
tions of work withdrawal through state positive
affect was positive in sign), which resulted in a
nonsignificant total effect. Thus, our mediation re-
sults for deep acting should be interpreted with this
condition in mind.
Cross-level hypotheses. Hypothesis 7a predicts
that the within-individual relationship between
daily surface acting and perceptions of work with-
drawal is stronger for females than males, and Hy-
pothesis 7b predicts that the within-individual re-
lationship between daily deep acting and
perceptions of work withdrawal is stronger for fe-
males than males. Results of slopes-as-outcomes
regressions revealed that gender was a significant
predictor of the within-individual slopes between
both daily surface acting and perceptions of work
withdrawal (b
41
⫽0.25, p⬍.05) and daily deep
acting and perceptions of work withdrawal (b
51
⫽
⫺0.14, p⬍.05). Figure 2 presents plots of these
interactions, which show that the relationships be-
tween surface acting, deep acting, and perceptions
of work withdrawal were stronger for females than
males. Simple slopes analyses (see Preacher, Cur-
ran, & Bauer, 2006) revealed that the relationship
between surface acting and perceptions of work
withdrawal was significant for both females (b
41
⫽
0.35, z⫽5.07, p⬍.05) and males (b
41
⫽0.10, z⫽
2.07, p⬍.05). The relationship between deep act-
ing and perceptions of work withdrawal was sig-
nificant for females (b
51
⫽⫺0.14, z⫽2.54, p⬍.05)
but not males (b
51
⫽0.00, z⫽.03). Overall, Hypoth-
eses 7a and 7b were fully supported.
As an anonymous reviewer pointed out, for
women, higher levels of work withdrawal were ob-
served at the lowest levels of deep acting versus the
lowest levels of surface acting. A closer inspection
of the data revealed that, at the lowest levels of
deep acting (i.e., 1 on a 5-point scale), women’s
average daily level of surface acting was higher
than men’s (mean ⫽1.40 versus 1.13). In contrast,
at the lowest levels of surface acting (i.e., 1 on a
5-point scale), women’s average daily level of deep
acting was lower than men’s (mean ⫽1.65 versus
2.76). In other words, on days when women did not
deep act, they engaged in surface acting more than
men; but on days when women did not surface act,
they engaged in deep acting less than men. Thus,
women’s reports of work withdrawal tended to be
higher on days when they did not deep act than on
days when they did not surface act, because on
days when they did not deep act, they tended to
surface act, and as our results showed, surface act-
ing was rather strongly associated with work
withdrawal.
Hypothesis 8a predicts that the within-individ-
ual relationship between daily surface acting and
state negative affect is stronger for females than
males, and Hypothesis 8b predicts that the within-
individual relationship between daily deep acting
and state negative affect is stronger for females than
males. This hypothesis was fully supported, as gen-
der was a significant predictor of the within-indi-
vidual slopes between both daily surface acting and
state negative affect (b
41
⫽0.14, p⬍.05) and daily
deep acting and state negative affect (b
51
⫽⫺0.11,
p⬍.05). The plots of these interactions in Figure 3
show that the relationships between surface acting,
deep acting, and state negative affect were stronger
for females than males. Simple slopes analyses re-
vealed that the relationship between surface acting
and state negative affect was significant for females
(b
41
⫽0.16, z⫽3.25, p⬍.05) but not males (b
41
⫽
0.02, z⫽.54). Similarly, the relationship between
deep acting and state negative affect was significant
for females (b
51
⫽⫺0.14, z⫽4.17, p⬍.05) but not
males (b
51
⫽⫺0.03, z⫽⫺1.29).
Hypothesis 9a predicts that the within-individ-
ual relationship between daily surface acting and
126 FebruaryAcademy of Management Journal

state positive affect is stronger for females than
males, and Hypothesis 9b predicts that the within-
individual relationship between daily deep acting
and state positive affect is stronger for females than
males. These predictions were not supported, as
gender did not moderate the within-individual re-
lationships between daily surface acting and state
positive affect (b
41
⫽⫺0.01) or daily deep acting
and state positive affect (b
51
⫽0.02).
Hypothesis 10a predicts that the within-individ-
ual relationship between state negative affect and
perceptions of work withdrawal is stronger for fe-
males than males, and Hypothesis 10b predicts that
the within-individual relationship between state
FIGURE 2
Cross-Level Moderating Effects of Gender on Within-Individual Relationships
1.50
1.90
2.29
2.69
3.09
Work
Withdrawal
−0.88 0.12 1.12 2.12 3.12
Surface Acting
Males
Females
1.60
1.75
1.91
2.06
2.22
Work
Withdrawal
−1.65 −0.65 0.35 1.35 2.35
Dee
p
Acting
(a) Relationship between Surface Acting and Work Withdrawal
(b) Relationship between Deep Acting and Work Withdrawal
2011 127Scott and Barnes

positive affect and perceptions of work withdrawal
is stronger for females than males. These hypothe-
ses were also not supported. Although the effect of
gender on the within-individual slope between
state negative affect and perceptions of work with-
drawal approached significance (b
61
⫽0.15, p⫽
.053), the effect of gender on the within-individual
slopes between state positive affect and percep-
tions of work withdrawal was not significant (b
71
⫽
⫺0.06). Given that the probability value for the
coefficient of gender on the within-individual rela-
tionship between state negative affect and percep-
tions of work withdrawal was close to .05, we pro-
ceeded to plot the interaction to assess whether the
FIGURE 3
Cross-Level Moderating Effects of Gender on Within-Individual Relationships
1.00
1.21
1.43
1.64
1.85
State Negative
Affect
−0.88 0.12 1.12 2.12 3.12
Surface Acting
Males
Females
1.00
1.15
1.29
1.44
1.58
State Negative
Affect
−1.65 −0.65 0.35 1.35 2.35
Dee
p
Acting
(a) Relationship between Surface Acting and State Negative Affect
(b) Relationship between Deep Acting and State Negative Affect
128 FebruaryAcademy of Management Journal

influence of gender was in the hypothesized direc-
tion. Figure 4 shows that this was the case, as the
relationship between state negative affect and per-
ceptions of work withdrawal was slightly stronger
for females than males. A simple slopes analysis
revealed that the relationship between state nega-
tive affect and perceptions of work withdrawal was
significant for both females (b
61
⫽0.28, z⫽3.97, p
⬍.05) and males (b
61
⫽0.13, z⫽3.41, p⬍.05).
2
Hypothesis 11a predicts that the indirect effects
of daily surface acting and daily deep acting on
perceptions of work withdrawal through state neg-
ative affect are stronger for females than males, and
Hypothesis 11b, that these indirect effects through
state positive affect are stronger for females than
males. Given that the indirect effects of emotional
labor on perceptions of work withdrawal through
state positive affect were not significant (see
above), we examined moderated mediation for re-
lationships involving state negative affect only. To
test Hypothesis 11a, we examined the indirect ef-
fects of surface acting and deep acting on percep-
tions of work withdrawal through state negative
affect for both females and males. For females, re-
sults revealed significant indirect effects of surface
acting (z⫽2.07, p⬍.05) and deep acting (z⫽2.31,
p⬍.05) on perceptions of work withdrawal via
state negative affect; for males however, the indi-
rect effects were not significant for either surface
acting (z⫽1.27) or deep acting (z⫽1.43). Despite
these differences, the confidence intervals for each
indirect effect overlapped, thus precluding the con-
clusion that the mediated effects differed signifi-
cantly by gender.
Additional Analyses
Although our primary focus at the within-indi-
vidual level concerned the effects of emotional la-
bor on affective states and work withdrawal, our
data also allowed us to examine whether employ-
ees’ affective states felt at the beginning of work
were associated with engagement in surface acting
or deep acting that day, potentially creating a larger
initial discrepancy between their feelings and the
display rules of their job. Results of HLM regres-
sions revealed that although the within-individual
relationship between state negative affect at the
start of work and surface acting was positive, it
only approached significance (b
20
⫽0.17, p⫽
.087). State negative affect at the start of work was
unrelated to deep acting (b
20
⫽⫺0.14), and state
positive affect at the start of work was unrelated to
2
On an exploratory basis, we examined the potential
moderating influence of age and ethnicity. Results of
these analyses revealed that neither age nor ethnicity
moderated any of the hypothesized relationships, with
one exception: the within-individual relationship be-
tween state negative affect and withdrawal was stronger
for Caucasians (b
63
⫽.27, p⬍.05). Importantly, the
moderation results for gender held.
FIGURE 4
Cross-Level Moderating Effect of Gender on the Within-Individual Relationship between
State Negative Affect and Work Withdrawal
1.60
1.91
2.22
2.52
2.83
Work
Withdrawal
−0.20 0.60 1.40 2.20 3.00
State Negative Affect
Males
Females
2011 129Scott and Barnes

both surface acting (b
30
⫽⫺0.05) and deep acting
(b
30
⫽0.06).
3
DISCUSSION
In their quantitative review of the emotional la-
bor literature, Bono and Vey (2005) called for re-
searchers to go beyond cross-sectional studies ex-
amining the correlates of emotional labor to
experience-sampling studies that track emotional
labor and its effects over time. Those authors also
called for researchers to examine individual differ-
ences as moderators of emotional labor relation-
ships. Toward those ends, we drew from the theo-
retical work of Grandey (2000) and utilized
experience-sampling methodology in a field setting
to examine the influence of emotional labor (i.e.,
surface acting and deep acting) on affective states
and perceptions of work withdrawal on a within-
individual basis, as well as the moderating role of
gender on these within-individual processes. Al-
though we are not the first to examine within-
individual variation in surface acting and deep act-
ing (see Judge et al., 2009; Totterdell & Holman,
2003), our multilevel investigation contributes ele-
ments absent in these noteworthy studies in several
respects and consequently extends theory and re-
search on emotion labor.
First, our results show that emotional labor strat-
egies vary within individuals as well as between
individuals: a given employee may use surface act-
ing at one time, but deep acting at another, to reg-
ulate affective states. The vast majority of research
to date has focused on static differences in surface
acting and deep acting between individuals. Our
findings reinforce the notion that emotional labor is
also dynamic, and thus for a more comprehensive
understanding of it, both perspectives (between-
and within-individual) are necessary. Indeed, the
importance of a within-individual approach is es-
pecially apparent in light of our findings regarding
the impact of emotional labor on employees’ affec-
tive states and reports of work withdrawal.
We found that surface acting is associated with a
change in affect for the worse, in that negative
affect increases on days in which individuals sur-
face act. This finding is consistent with research on
ironic effects (Wegner, 1994), which occur when
negative feelings intensify as a result of trying to
suppress them. In contrast, we found that deep
acting is associated with a change in affect for the
better, in that negative affect decreases and positive
affect increases. Previous research has yet to exam-
ine the changes in felt affect associated with engag-
ing in surface and deep acting—a surprising omis-
sion, given the integral role of experienced emotion
in theories of emotional labor. In doing so, our
investigation extends theory to incorporate within-
individual variation by demonstrating how a given
individual’s choice of an emotional labor strategy
may alter daily negative and positive affective
states. Although Judge et al. (2009) also positioned
affect as an outcome of emotional labor within in-
dividuals, their study did not control for affect at
the start of workdays. We acknowledge, of course,
that although controlling for affect at the beginning
of work removes some of the ambiguity regarding
casual direction, it does not by itself establish
causality.
In addition, our results add to existing knowl-
edge by demonstrating that although a given em-
ployee may be able to conform to organizational
display rules via either surface acting or deep act-
ing, the choice of surface acting over deep acting is
likely to have deleterious effects on organizations
via its effects on work withdrawal. Importantly, we
were able to explain part of the process by which
emotional labor is associated with work with-
drawal. Specifically, we found that state negative
affect mediated the within-individual relationships
between surface acting, deep acting, and work
withdrawal. In contrast, state positive affect did not
mediate these within-individual relationships, as it
was not significantly associated with work with-
drawal. The presence of a relationship between
state negative affect and work withdrawal, and the
lack of a relationship between state positive affect
and work withdrawal, is consistent with recent
meta-analytic findings at the trait level, which
show that trait negative affect, but not trait positive
affect, is associated with such behavior (Kaplan,
Bradley, Luchman, & Haynes, 2009). Taken to-
gether, these findings extend the range of Grandey’s
(2000) theory not only by demonstrating the rele-
vance of emotional labor to work withdrawal at
the within-individual level, but also by illuminat-
ing a mechanism fitting to that level (i.e., affect)
through which emotional labor influences work
withdrawal.
Finally, at the between-individuals level, our re-
sults for gender tell a consistent story. Adding to
findings by Judge et al. (2009) that emotional labor
is generally more detrimental to introverts than to
3
Following suggestions by anonymous reviewers,
we also examined whether the lengths of drivers’ shifts
and the day of the week were associated with choice of
emotional labor strategy. Results of HLM regressions
revealed that both shift length and day of week were
unrelated to both surface acting and deep acting. More-
over, controlling for shift length and day of week did
not alter the significance of our findings.
130 FebruaryAcademy of Management Journal

extraverts, we found that, relative to males, women
were more likely to experience negative affect and
to report withdrawing from work on days when
they engaged in surface acting, but when they en-
gaged in deep acting, women were less likely than
men to experience negative affect and to report
withdrawing from work. Moreover, when women
experienced negative affect, they were more likely
to report withdrawing from work, although this
effect was marginal. Findings at the between-indi-
vidual level on interactions between gender and
emotional labor have been rather equivocal, with
some research supporting gender as a moderator
(Johnson & Spector, 2007), other research finding
no support (Erickson & Ritter, 2001), and still other
research finding results opposite to predictions
(Simpson & Stroh, 2004). It may be that the moder-
ating influence of gender is most pronounced for
within-individual relationships. If true, this would
further reinforce the importance of taking a multi-
level approach to the study of emotional labor, and
theory on emotional labor could be revised to in-
corporate the cross-level influences of gender on
emotional labor processes. Overall then, our find-
ings for gender illustrate how a more complex and
nuanced understanding of emotional labor can be
achieved via the integration of between- and with-
in-individual levels of analysis.
Limitations and Strengths
Although our study possesses a number of
strengths (e.g., the use of experience-sampling
methodology in a field context, temporal separa-
tion between daily measures of affect), there are of
course some limitations. First, although our posi-
tioning of affective states and work withdrawal as
outcomes of emotional labor was based on existing
theory (e.g., Grandey, 2000), alternative causal or-
derings may be possible, and our data cannot es-
tablish causality. However, controlling for affective
states experienced at the beginning of work did
allow us to show whether surface acting and deep
acting were associated with changes in positive and
negative affect, relationships that have not been
tested in previous research.
A second limitation concerns the scope of our
model. Although the experience-sampling method-
ology utilized in the current study has a number of
strengths, a drawback of this methodology is con-
straint on the number of variables that can be as-
sessed, particularly when employees complete sur-
veys at work each day (Scollon, Kim-Prieto, &
Diener, 2003). Consequently, some variables likely
relevant to emotional labor were excluded. For ex-
ample, in addition to work withdrawal, perfor-
mance is positioned in Grandey’s (2000) model as a
work outcome of emotional labor. In addition, al-
though we investigated state negative and positive
affect in concert with emotional labor, other dimen-
sions of affective experience may be relevant, such
as pleasure and arousal (Russell, 1980), as well as
the low ends of the negative and positive affect
dimensions, which the PANAS does not capture
(Watson & Vaidya, 2003). Future research adopting
a within-individual approach to the study of emo-
tional labor may benefit by addressing performance
and a greater diversity of affective states.
A third limitation is that all of our measures
relied on self-reports, which raises the possibility
of common method variance. However, several fac-
tors lessen this possibility. Specifically, the mea-
sures of affect were separated temporally each day,
and because we centered the daily measures rela-
tive to participants’ means, we avoided several
sources of common method variance, such as re-
sponse tendencies and affectivity. Indeed, negative
and positive affect were substantive variables in
our analyses and thus were valid sources of vari-
ability (see Podsakoff, MacKenzie, Lee, & Podsa-
koff, 2003). Finally, the cross-level interactions in-
volving gender are unlikely to be explained by
common method variance.
Related to the above, our measure of work with-
drawal, strictly speaking, assessed perceptions of
behavior rather than actual behavior. Although our
approach is consistent with existing experience-
sampling studies that have relied on self-reports of
behaviors such as task performance (Fisher & No-
ble, 2004), organizational citizenship (Ilies, Scott, &
Judge, 2006), and workplace deviance (Judge, Scott,
& Ilies, 2006), collecting objective data on actual
withdrawal behaviors (e.g., absenteeism, tardiness)
would be worthwhile, particularly in studies exam-
ining lagged effects of daily emotional labor on
work behavior. Thus, future research could attempt
to replicate our findings using more objective data.
In addition, for most of our variables, the majority
of the variance was at the between-individual level.
Despite the lower amount of within-individual
variance relative to between-individual variance,
we were able to detect significant relationships at
the within-individual level.
Finally, although the interval-contingent experi-
ence-sampling methodology employed in the cur-
rent investigation allowed us to capture employees’
emotional labor, affective states, and perceptions of
work withdrawal daily, we were unable to separate
multiple emotional labor episodes that may have
occurred during drivers’ workdays. Future studies
obtaining multiple observations during the work-
day and/or using event-contingent experience-sam-
2011 131Scott and Barnes

pling methodology (whereby individuals immedi-
ately respond following the occurrence of a
predesignated event) would better capture discrete
emotional labor episodes and allow researchers to
examine the interplay between surface acting and
deep acting within a given day—for instance,
whether individuals use surface acting and deep
acting simultaneously, or whether they use them in
some sequence. Such methods were not feasible in
our sample of bus drivers, who would have had to
temporarily discontinue their routes to complete
measures delivered in these ways.
Implications and Suggestions for
Future Research
Our findings have several implications for prac-
tice as well as future research. Regarding practice,
our results add to existing research showing the
detrimental effects of surface acting. The recom-
mendation against using surface acting is not en-
tirely new (Cote, 2005; Grandey, 2000; Hochschild,
1983). However, our findings supplement this rec-
ommendation by revealing that surface acting is
harmful to both employees and organizations via
its association with work withdrawal. That is, al-
though surface acting may result in the display of
positive affect, which is desirable from an organi-
zational standpoint, it simultaneously increases
felt negative affect and the likelihood of work with-
drawal, which is undesirable from an organization-
al standpoint. In contrast, deep acting may actually
change experienced affect for the better. Conse-
quently, managers may wish to discourage workers
from engaging in surface acting and instead encour-
age them to engage in deep acting via training in
techniques such as cognitive reappraisal (Gross,
1998).
This recommendation may be true particularly
for women, who, relative to men, appear to be
harmed more by surface acting but helped more by
deep acting. The encouragement of deep acting,
especially for females, not only would increase the
likelihood that the positive affective displays of
employees are authentic, resulting in more favor-
able customer reactions (Cote, 2005; Grandey,
2003), but also would decrease the likelihood of
work withdrawal following engagement in emo-
tional labor. Moreover, to the extent that deep act-
ing increases positive affect, organizations may
benefit from resultant emotional contagion effects,
which have been shown to positively influence
customers’ evaluations of service quality (Pugh,
2001) as well as individual performance and coop-
eration in groups (Barsade, 2002). Indeed, such
contagion effects may be more pronounced in fe-
males, given their tendency to experience and ex-
press emotions more intensely than males (Brody &
Hall, 2008). Of course, it is likely that surface acting
cannot be eliminated entirely, and thus organiza-
tions should have practices in place to manage the
negative outcomes of this emotional labor strategy,
such as supervisor monitoring of employees to de-
tect withdrawal. Such monitoring could provide
opportunities for supervisors to intervene with ei-
ther a short break, temporary job rotation, or
prompting to refocus attention on the job.
With regards to future research, we see several
directions in which our findings could be ex-
tended. First, although we found some support for
affect as a mediator of the relationship between
emotional labor and work withdrawal, the media-
tion effects were not particularly strong, and thus
future research should test additional mediators of
this relationship. As stated above, it may be that
surface acting is associated with work withdrawal
via its effects on resource depletion (e.g., Baumeis-
ter et al., 2000). Other mediators might include
physical exhaustion and fatigue (Barnes & Van
Dyne, 2009). Additionally, although we focused on
work withdrawal as our work outcome of interest,
other work outcomes may be relevant. For example,
to the extent that surface acting increases negative
affect, deviant workplace behaviors may result
(Judge et al., 2006). In contrast, to the extent that
deep acting increases positive affect, organizational
citizenship behaviors may result (Ilies, Scott, &
Judge, 2006). Thus, future research could extend
our model to other important work criteria.
Future research could ascertain whether our re-
sults hold in different cultures and contexts. Norms
for emotional expression vary across cultures (e.g.,
Ekman & Friesen, 1971), with some cultures valu-
ing the regulation of emotions in adherence to in-
stitutional roles and standards, and other cultures
valuing the impulsive expression of unregulated
emotions (Gordon, 1989). It may be that detrimen-
tal effects of surface acting on affective states and
work withdrawal are stronger in more institution-
ally oriented cultures. Indeed, in a between-indi-
vidual study, Grandey, Fisk, and Steiner (2005)
found some evidence that the negative relationship
between surface acting and job satisfaction was
stronger in U.S. employees than French employees;
the former are more institutionally oriented.
In terms of contexts, it would be interesting to
examine the within-individual dynamics of emo-
tional labor in different types of organizations, in-
cluding “feminist” organizations and organizations
characterized by “bounded emotionality.” In such
organizations, the expression of a wider range of
emotions by employees is encouraged (e.g., Kark &
132 FebruaryAcademy of Management Journal

Medler-Liraz, 2007; Martin, Knopoff, & Beckman,
1998; Mumby & Putnam, 1992). It may be that al-
lowing employees to “be themselves” at work leads
to greater within-individual variation in affect,
which may positively influence day-to-day feelings
of personal well-being (Mumby & Putnam, 1992).
However, if employees who prefer emotional re-
straint feel pressured to openly display feelings,
negative outcomes may result (Martin et al., 1998).
Indeed, in such organizations, men, who tend to be
more emotionally restrained (e.g., Eagly, 1987),
may need to engage in emotional labor more fre-
quently to conform to bounded emotionality
norms.
Finally, although our two-week investigation re-
vealed important short-term effects of emotional
labor, it would be informative if future studies
could examine emotional labor over longer periods
of time. It may be that surface acting has more
detrimental consequences in the short term, but
deep acting has more detrimental consequences in
the long term. Continuously devoting effort toward
creating desired affective states may eventually
harm employees’ feelings of personal authenticity
(Ashforth & Humphrey, 1993). Such long-term in-
vestigations not only could better illuminate the
causal relationships between emotional labor and
work outcomes, but also could reveal whether emo-
tional labor processes are cyclical (see, for example,
Amabile, Barsade, Mueller, and Staw [2005]). Such
research, in combination with the current study
and existing studies examining emotional labor at
multiple points in time (Judge et al., 2009; Totter-
dell & Holman, 2003), would culminate in a greater
understanding of the temporal dynamics of emo-
tional labor.
REFERENCES
Alliger, G. M., & Williams, K. J. 1993. Using signal-con-
tingent experience sampling methodology to study
work in the field: A discussion and illustration ex-
amining task perceptions and mood. Personnel Psy-
chology, 46: 525–549.
Amabile, T. M., Barsade, S. G., Mueller, J. S., & Staw,
B. M. 2005. Affect and creativity at work. Adminis-
trative Science Quarterly, 50: 367–403.
Ashforth, B. E., & Humphrey, R. H. 1993. Emotional labor
in service roles—The influence of identity. Acad-
emy of Management Review, 18: 88–115.
Ashmore, R. D., Del Boca, F. K., & Wohlers, A. J. 1986.
The social psychology of female-male relations: A
critical analysis of central concepts. Orlando, FL:
Academic Press.
Barnes, C. M., & Van Dyne, L. 2009. “I’m tired:” Differ-
ential effects of physical and emotional fatigue on
workload management strategies. Human Relations,
62: 57–90.
Barsade, S. G. 2002. The ripple effect: Emotional conta-
gion and its influence on group behavior. Adminis-
trative Science Quarterly, 47: 644– 675.
Baumeister, R. F., Muraven, M., & Tice, D. M. 2000. Ego
depletion: A resource model of volition, self-regula-
tion, and controlled processing. Social Cognition,
18: 130–150.
Beal, D. J., Trougakos, J. P., Weiss, H. M., & Green, S. G.
2006. Episodic processes in emotional labor: Percep-
tions of affective delivery and regulation strategies.
Journal of Applied Psychology, 91: 1053–1065.
Bono, J. E., & Vey, M. A. 2005. Toward understanding
emotional management at work: A quantitative re-
view of emotional labor research. In C. E. Ha¨rtel,
W. J. Zerbe, & N. M. Ashkanasy (Eds.), Emotions in
organizational behavior: 213–233. Mahwah, NJ:
Erlbaum.
Brody, L. R., 1999. Gender, emotion, and the family.
Cambridge, MA: Harvard University Press.
Brody, L. R., & Hall, J. A. 2008. Gender and emotion in
context. In M. Lewis, J. M. Haviland-Jones, & L. F.
Barrett (Eds.), Handbook of emotions: 338–349.
New York: Guilford.
Brotheridge, C. M., & Grandey, A. A. 2002. Emotional
labor and burnout: Comparing two perspectives of
“people work.” Journal of Vocational Behavior, 60:
17–39.
Cacioppo, J. T., Gardner, W. L., & Bernston, G. G. 1999.
The affect system has parallel and integrative pro-
cessing components: Form follows function. Journal
of Personality and Social Psychology, 76: 839– 855.
Clark, L. A., Watson, D., & Leeka, J. 1989. Diurnal varia-
tion in the positive affects. Motivation and Emotion,
13: 205–234.
Cote, S. 2005. A social interaction model of the effects of
emotion regulation on work strain. Academy of
Management Review, 30: 509–530.
Eagly, A. H. 1987. Sex differences in social behavior: A
social-role interpretation. Hillsdale, NJ: Erlbaum.
Edwards, J. R., & Lambert, L. S. 2007. Methods for inte-
grating moderation and mediation: A general ana-
lytic framework using moderated path analysis. Psy-
chological Methods, 12: 1–22.
Ekman, P., & Friesen, W. V. 1969. The repertoire of non-
verbal behavior: Categories, origins, usage, and cod-
ing. Semiotica, 1: 49–98.
Ekman, P., & Friesen, W. V. 1971. Constants across cul-
tures in the face and emotion: Journal of Personality
and Social Psychology, 17: 124–129.
Elfenbein, H. A. 2008. Emotions in organizations. In J. P.
Walsh & A. P. Brief (Eds.), Academy of Manage-
ment annals, vol. 1: 315–386. New York: Erlbaum.
Erickson, R. J., & Ritter, C. 2001. Emotional labor, burn-
2011 133Scott and Barnes

out, and inauthenticity: Does gender matter? Social
Psychology Quarterly, 64: 146–163.
Fabes, R. A., & Martin, C. L. 1991. Gender and age ste-
reotypes of emotionality. Personality and Social
Psychology Bulletin, 17: 532–540.
Fisher, C. D., & Noble, C. S. 2004. A within-person ex-
amination of correlates of performance and emotions
while working. Human Performance, 17: 145–168.
Fitness, J. 2000. Anger in the workplace: An emotion
script approach to anger episodes between workers
and their superiors, co-workers, and subordinates.
Journal of Organizational Behavior, 21: 147–162.
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.
Frijda, N. H. 1994. Varieties of affect: Emotions and epi-
sodes, moods, and sentiments. In P. Ekman & R. J.
Davidson (Eds.), The nature of emotion: 59– 67.
New York: Oxford University Press.
Frijda, N. H. 2007. The laws of emotion. Mahwah, NJ:
Erlbaum.
Gordon, S. L. 1989. Institutional and impulsive orienta-
tions in selective appropriating emotions to self. In
D. D. Franks & D. McCarthy (Eds.), The sociology of
emotions: Original essays and research papers:
115–136. Greenwich, CT: JAI Press.
Gosserand, R. H., & Diefendorff, J. M. 2005. Emotional
display rules and emotional labor: The moderating
role of commitment. Journal of Applied Psychol-
ogy, 90: 1256–1264.
Grandey, A. A. 2000. Emotional regulation in the work
place: A new way to conceptualize emotional la-
bor. Journal of Occupational Health Psychology,
5: 95–110.
Grandey, A. A. 2003. When “the show must go on”:
Surface and deep acting as determinants of emo-
tional exhaustion and peer-rated service delivery.
Academy of Management Journal, 46: 86–96.
Grandey, A. A., & Brauburger, A. L. 2002. The emotion
regulation behind the customer service smile. In
R. G. Lord, R. J., Klimoski, & R. Kanfer (Eds.), Emo-
tions in the workplace: 260–294. San Francisco:
Jossey-Bass.
Grandey, A. A., Fisk, G. M., & Steiner, D. D. 2005. Must
“service with a smile” be stressful? The moderating
role of personal control for American and French
employees. Journal of Applied Psychology, 90:
893–904.
Gross, J. 1998. Antecedent- and response-focused emo-
tion regulation: Divergent consequences for experi-
ence, expression, and physiology. Journal of Per-
sonality and Social Psychology, 74: 224–237.
Grossman, M., & Wood, W. 1993. Sex differences in in-
tensity of emotional experience: A social role inter-
pretation. Journal of Personality and Social Psy-
chology, 65: 1010–1022.
Hochschild, A. R. 1979. Emotion work, feeling rules, and
social-structure. American Journal of Sociology, 85:
551–575.
Hochschild, A. R. 1983. The managed heart: Commer-
cialization of human feeling. Berkeley: University
of California Press.
Hoffmann, S. A., Griffin, M. A., & Gavin, M. B. 2000. The
application of hierarchical linear modeling to organ-
izational research. In K. Klein & S. Kozlowski (Eds.),
Multilevel theory, research, and methods in organ-
izations: Foundations, extensions, and new direc-
tions: 75–170. San Francisco: Jossey-Bass.
Ilies, R., Scott, B. A., & Judge, T. A. 2006. The interactive
effects of personal traits and experienced states on
intraindividual patterns of organizational citizen-
ship behavior. Academy of Management Journal,
49: 561–575.
Johnson, H. A. M., & Spector, P. E. 2007. Service with a
smile: Do emotional intelligence, gender, and auton-
omy moderate the emotional labor process? Journal
of Occupational Health Psychology, 12: 319–333.
Judd, C. M., & Kenny, D. A. 1981. Estimating the effects
of social interventions. New York: Cambridge Uni-
versity Press.
Judge, T. A., & Ilies, R. 2004. Affect and job satisfaction:
A study of their relationship at work and at home.
Journal of Applied Psychology, 89: 661–673.
Judge, T. A., Scott, B. A., & Ilies, R. 2006. Hostility, job
attitudes, and workplace deviance: Test of a multi-
level model. Journal of Applied Psychology, 91:
126–138.
Judge, T. A., Woolf, E. F., & Hurst, C. 2009. Is emotional
labor more difficult for some than for others? A mul-
tilevel, experience-sampling study. Personnel Psy-
chology, 62: 57–88.
Kaplan, S., Bradley, J. C., Luchman, J. N., & Haynes, D.
2009. On the role of positive and negative affectivity
in job performance: A meta-analytic investigation.
Journal of Applied Psychology, 94: 162–176.
Kark, R., & Medler-Liraz, H. 2007. Leading with a smile:
The influence of managers’ leadership behavior on
the emotional experience of employees and custom-
ers. In C. E. J. Hartel, N. M. Ashkanasy, & W. J. Zerbe
(Eds.), Research on emotion in organizations:
Functionality, intentionality, and morality: 177–
198. Oxford, U.K.: JAI Press.
Kozlowski, S. W. J., & Klein, K. J. 2000. A multilevel
approach to theory and research in organizations. In
K. J. Klein & S. W. J. Kozlowski (Eds.), Multilevel
theory, research, and methods in organizations:
Foundations, extensions, and new directions: 3–90.
San Francisco: Jossey-Bass.
Kring, A. M., & Gordon, A. H. 1998. Sex differences in
emotion: Expression, experience, and physiology.
134 FebruaryAcademy of Management Journal

Journal of Personality and Social Psychology, 74:
686–703.
Krull, J. J., & MacKinnon, D. P. 1999. Multilevel media-
tion modeling in group-based intervention studies.
Evaluation Review, 23: 418– 444.
Kruml, S., & Geddes, D. 1998. Exploring the dimen-
sions of emotional labor: The heart of Hoch-
schild’s work. Paper presented at the First Confer-
ence of Emotions in Organizational Life, San
Diego.
Lehman, W. E., & Simpson, D. D. 1992. Employee sub-
stance use and on-the-job behaviors. Journal of Ap-
plied Psychology, 77: 309–321.
MacKinnon, D. P., Krull, J. L., & Lockwood, C. M. 2000.
Equivalence of the mediation, confounding and
suppression effect. Prevention Science, 1: 173–
181.
MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M.,
West, S. G., & Sheets, V. 2002. A comparison of
methods to test mediation and other intervening
variable effects. Psychological Methods, 7: 83–
104.
Martin, J., Knopoff, K., & Beckman, C. 1998. An alterna-
tive to bureaucratic impersonality and emotional la-
bor: Bounded emotionality at The Body Shop. Ad-
ministrative Science Quarterly, 43: 429– 469.
Merton, J. 1997. Facial-affective behavior, mutual gaze,
and emotional experience in dyadic interactions.
Journal of Nonverbal Behavior, 21: 179–201.
Morris, J. A., & Feldman, D. C. 1996. The dimen-
sions, antecedents, and consequences of emotional
labor. Academy of Management Review, 21: 986–
1010.
Mumby, D. K., & Putnam, L. L. 1992. The politics of
emotion: A feminist reading of bounded emotional-
ity. Academy of Management Review, 17:
465–486.
Muraven, M., & Baumeister, R. F. 2000. Self-regulation
and depletion of limited resources: Does self-control
resemble a muscle? Psychological Bulletin, 126:
247–259.
Nolen-Hoeksema, S., & Jackson, B. 2001. Mediators of the
gender difference in rumination. Psychology of
Women Quarterly, 25: 37–47.
Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsa-
koff, N. P. 2003. Common method biases in behav-
ioral research: A critical review of the literature and
recommended remedies. Journal of Applied Psy-
chology, 88: 879–903.
Preacher, K. J., Curran, P. J., & Bauer, D. J. 2006. Compu-
tational tools for probing interaction effects in mul-
tiple linear regression, multilevel modeling, and la-
tent curve analysis. Journal of Educational and
Behavioral Statistics, 31: 437–448.
Pugh, S. D. 2001. Service with a smile: Emotional conta-
gion in the service encounter. Academy of Manage-
ment Journal, 44: 1018–1027.
Rafaeli, A., & Sutton, R. I. 1987. Expression of emotion as
part of the work role. Academy of Management
Review, 12: 23–37.
Raudenbush, S. W., & Bryk, A. S. 2002. Hierarchical
linear models: Applications and data analysis
methods. Thousands Oaks, CA: Sage.
Reis, H. T., & Wheeler, L. 1991. Studying social interac-
tion with the Rochester Interaction Record. In M. P.
Zanna (Ed.), Advances in experimental social psy-
chology, vol. 24: 270–318. San Diego: Academic
Press.
Richards, J. M., & Gross, J. J. 1999. Composure at any
cost? The cognitive consequences of emotion sup-
pression. Personality & Social Psychology Bulletin,
25: 1033–1044.
Russell, J. A. 1980. A circumplex model of affect. Jour-
nal of Personality and Social Psychology, 39:
1161–1178.
Scollon, C. N., Kim-Prieto, C., & Diener, E. 2003.
Experience-sampling: Promises and pitfalls,
strengths and weaknesses. Journal of Happiness
Studies, 4: 5–34.
Simpson, P. A., & Stroh, L. K. 2004. Gender differ-
ences: Emotional expression and feelings of per-
sonal inauthenticity. Journal of Applied Psychol-
ogy, 89: 715–721.
Sobel, M. E. 1982. Asymptotic confidence intervals for
indirect effects in structural equation models. In
S. Leinhardt (Ed.), Sociological methodology:
290–312. Washington, DC: American Sociological
Association.
Totterdell, P., & Holman, D. 2003. Emotion regulation in
customer service roles: Testing a model of emotional
labor. Journal of Occupational Health Psychology,
8: 55–73.
VanMaanen, J., & Kunda, G. 1989. Real feelings: Emo-
tional expression and organizational culture. In L. L.
Cummings & B. M. Staw (Eds.), Research in organ-
izational behavior, vol. 11: 43–103. Greenwich, CT:
JAI Press.
Watson, D. 2000. Mood and temperament. New York:
Guilford.
Watson, D., & Vaidya, J. 2003. Mood measurement: Cur-
rent status and future directions. In I. B. Weiner
(series ed.) & J. A. Schinka & W. F. Velicer (vol. eds.),
Handbook of psychology: vol. 2, Research methods
in psychology: 351–375. New York: Wiley.
Watson, D., & Clark, L. A., & Tellegen, A. 1988. Devel-
opment and validation of brief measures of posi-
tive and negative affect: The PANAS scales. Jour-
nal of Personality and Social Psychology, 54:
1063–1070.
2011 135Scott and Barnes

Weedon, C. 1987. Feminist practice and poststructural-
ist theory. Oxford, U.K.: Blackwell.
Wegner, D. M. 1994. Ironic processes of mental control.
Psychological Review, 101: 34–52.
Weiss, H. M. & Cropanzano, R. 1996. Affective events
theory: A theoretical discussion of the structure,
causes and consequences of affective experiences at
work. In B. M. Staw & L. L. Cummings (Eds.),
Research in organizational behavior, vol. 18: 1–74.
Greenwich, CT: JAI Press.
Wharton, A. S. 1993. The affective consequences of ser-
vice work: Managing emotions on the job. Work &
Occupations, 20: 205–232.
Wheeler L., & Reis, H. T. 1991. Self-recording of everyday
life events: Origins, types, and uses. Journal of Per-
sonality, 59: 339–354.
Brent A. Scott (scott@bus.msu.edu) is an assistant pro-
fessor of management in the Eli Broad College of Busi-
ness at Michigan State University. He received his Ph.D.
from the University of Florida. His research interests
include mood and emotion in the workplace, organiza-
tional justice, and coworker relationships.
Christopher M. Barnes (christopher.barnes@usma.edu) is
an assistant professor of character development and
research in the Center for the Army Profession and
Ethic, United States Military Academy at West Point.
He received his Ph.D. from Michigan State University.
His primary research interests include fatigue in organ-
izations, team performance and decision making, and
behavioral ethics.
136 FebruaryAcademy of Management Journal