Journal of Organizational Behavior J. Organiz. Behav.

21, 163183 (2000)

Antecedents of workplace emotional labor dimensions and moderators of their eects on physical symptoms
JOHN SCHAUBROECK*1 AND JAMES R. JONES2
1

Department of Management, LeBow College of Business, Drexel University, 33rd & Arch Streets, Philadelphia, PA 19104, U.S.A. 2 Department of Marketing/Management, College of Business Administration, University of Nebraska at Omaha, Omaha, NE, U.S.A.

Summary

The present study distinguished between two modal emotional display rules, demands to express positive eerence and demands to suppress negative eerence, that partially constitute the work roles of many employees. Perceived demands to express positive emotion were positively related to health symptoms primarily among those reporting: (1) lower identication with the organization; (2) lower job involvement; and (3) lower emotional adaptability. The eects of various personality traits and situational variables on perceived emotional labor diered depending on the nature of the emotional labor. The ndings are discussed in terms of implications of emotional labor for health and practices through which organizations might intervene to minimize its unhealthful consequences among employees. We also attempt to reconcile the ndings with some of the related research in psychology suggesting that some forms of required eerence may have salutary physiological consequences. Copyright # 2000 John Wiley & Sons, Ltd.

Introduction
Much research investigating the relationship between job demands and health has emphasized the key psychosocial nature of the demands. Pressures are placed on individuals in a social context and their behavioral responses to these pressures have socially derived implications for them. Being required to modulate the expression of one's own emotions in particular ways is a signicant component of the work role for many persons. This type of role requirement has been called `emotional labor' (cf. Wharton and Erickson, 1993, p. 458). The concept of emotional display rules was introduced by Paul Ekman (e.g., Ekman, 1972). Emotional display rules refer to norms about appropriate emotional expression for specic situations. Persons who have much customer or client contact (e.g., salespersons, nurses) are seen to be subject to stronger emotional display rules (Sutton, 1991; Sutton and Rafaeli, 1988). Several authors have observed that such display rules may be expected to compromise the psychological and/or physical health of workers
* Correspondence to: John Schaubroeck, Department of Management, LeBow College of Business, Drexel University, 33rd & Arch Streets, Philadelphia, PA 19104, U.S.A. E-mail: [email protected]

CCC 08869383/2000/02016321$17.50 Copyright # 2000 John Wiley & Sons, Ltd.

164

J. SCHAUBROECK AND J. R. JONES

because they often lead to a disturbing disequilibrium (or `dissonance') between felt emotions and the emotions one must exhibit (Ashforth and Humphrey, 1993; Morris and Feldman, 1996; Rafaeli and Sutton, 1987; Wharton and Erickson, 1993). Indeed there is considerable evidence from outside the work sphere that chronically experiencing conict among one's emotions has negative health consequences (see reviews by Friedman, 1989; King and Emmons, 1990). Emotional labor is a construct of increasing interest among theorists in the management literature (cf. Ashforth and Humphrey, 1993; Morris and Feldman, 1996; Rafaeli and Sutton, 1987; Wharton and Erickson, 1993). From an empirical point of view, however, `the expression of socially appropriate emotion constitutes a neglected form of role demand' (Ashforth and Humphrey, 1993, p. 110). As noted by Wharton and Erickson (1993), `virtually all previous studies [of emotional labor] have been case studies' ( p. 462). To our knowledge, only two quantitative studies of emotional labor in the workplace have been published in the literature. After sampling workers in two companies, Wharton (1993) constructed a dummy variable of emotional labor by using a scheme developed by Hochschild (1983) to classify some of the jobs as high in emotional labor. Wharton (1993) observed no signicant correlation between this variable and emotional exhaustion, and a signicant negative correlation with job satisfaction. Morris and Feldman (1997) developed a scale measure of emotional labor dierentiated along three dimensions: duration of contact with others, frequency of contact with others, and emotional dissonance inspired by such contact. Their study supported hypotheses linking emotional labor to characteristics of the job as well as emotional exhaustion. Despite the claims of Hochschild (1983) and others that emotional labor is related to physical symptoms, there is no research relating emotional labor to physical symptoms. As we review below, laboratory studies from psychology are suggestive of the possibility that certain types of emotional labor may in fact have salutary inuences on individual's physiological well-being whereas other forms may be toxic to health. We distinguish in this study between demands to suppress negative emotion, which may generally be expected to be unhealthful, and demands to express positive emotion. Second, we identify key individual and situational moderators of this emotional laborsymptoms relationship. Third, it is not clear whether or to what extent emotional labor reects dispositional tendencies and/or objective situational exposures. Accordingly, we examine how individual dierences and situational variables determined two dimensions of emotional labor and also how these variables moderated relationships between emotional labor and health symptoms. Figure 1 depicts the framework for this study. Objective characteristics of the role (i.e., requirements for interpersonal interaction) and personal traits (i.e., personality traits and gender) are posited to determine perceptions of emotional labor. Perceptions of emotional labor inuence physical health. The relationship between perceived emotional labor and health is moderated by emotional adaptability and work perceptions (i.e., job involvement and organizational identication).

Psychological research on emotional expression

Dierent aspects of emotional labor have been explicated and many propositions about their eects on well-being and other outcomes have been proposed. However, the prevailing concept of emotional display rules remains unidimensional in character despite the range of plausible types of display rules. For the present study we instead chose to focus on the character of the emotions being managed, such as whether a particular type of eerence (i.e., positive or negative) is
Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

WORKPLACE EMOTIONAL LABOR DIMENSIONS

165

Figure 1. Overview of research model

required to be expressed or suppressed. Perhaps because the concept emerged from a dierent eld (sociology), research and theory about emotional labor has generally overlooked relevant research from the eld of psychology. However, the psychological literature may usefully broaden perspectives about emotional labor. In addition to the emotional conict research conducted in eld settings (as reviewed by King and Emmons, 1990), a good deal of laboratory research has investigated the relationship between emotional eerence and physiological outcomes. The psychological studies of emotion have tended to distinguish between two types of emotion manipulation: the suppression of felt negative emotion and the required display of positive emotion. While there are many possible types of display rules, we believe these two general categories, the display of positive emotional eerence and the suppression of negative emotional eerence, capture many of the rules operating in organizational contexts. When one is explicitly or implicitly required to display emotional eerence in the organization, typically one is required to display cheerfulness, earnestness, and other emotions that are culturally dened as `positive'. Exceptions to this may include psychological therapists, morticians, and other jobholders that are often required to `deep act' (Rafaeli and Sutton, 1987) sadness, concern, and other less `positive' emotions (cf. the bill collectors studied by Sutton, 1991). In addition, interpersonal role relationships such as one has with coworkers and customers often require one to suppress negative emotions such as irritation and anger. Whereas these two types of role requirements may not capture all forms of emotional labor, we believe they represent emotional labor demands that are relevant to a wide range of occupations. Occupational categorization of emotional labor The present study takes a psychological, rather than sociological, approach to emotional labor. Thus in contrast to previous studies we do not equate emotional labor with the characteristics of an occupational setting (cf. Hochschild, 1983) or the frequency and duration of interaction with other persons (Morris and Feldman, 1997). Rather we operationalize requirements to express or suppress emotion in terms of the experience of workers in having to suppress negative emotions
Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

166

J. SCHAUBROECK AND J. R. JONES

and express positive emotions in order to meet the demands of their jobs. The `emotional dissonance' measure used by Morris and Feldman (1997) comes closest to our approach in measuring emotional labor (e.g., `Most of the time, the way I act and speak with patients matches how I feel anyway' (reverse-scored; p. 263)). This type of measure appears on its face to be more consistent with existing denitions of emotional labor than is mere interpersonal interaction. Whether being required to interact frequently with others creates emotional labor may be expected to depend on some interaction between the context of the interactions and the characteristics of the individual. Indeed, Morris and Feldman (1997) observed very little correlation between the frequency and duration of interactions with others and their index of emotional dissonance. As a sociologist, however, Hochschild (1983) was more interested in measuring the distribution of objective occupational dierences and exploring how incumbents of these occupations had to manage their emotions. It should also be noted that Hochschild's scheme for identifying occupations that require much emotional labor did take the expected modal content of interactions into account, not just the frequency and duration of these. However, her coding scheme did not distinguish this content in any way, and like all broad occupation variables hers provides a crude estimate of the psychological exposures experienced by dierent individuals in the same occupation. As we examine further on in this paper, it may well be that individual dierences inuence the perception and experience of emotional labor. Thus one goal of this study was to ascertain the extent to which emotional labor, as measured directly at the individual level, corresponds with the occupational scheme developed by Hochschild that is based primarily upon the volume and intensity of interpersonal interaction on the job. Hypothesis 1. The extent to which individuals' occupations require interpersonal interaction will be positively related to their reports of role expectations to suppress negative emotional eerence and to display positive emotional eerence.

The role of emotion-related traits in emotional labor In addition to emotional labor inherent within a particular job or occupation, the present study also seeks to examine its dispositional sources. In the case of emotional labor, the key resource domain for coping may be related to self-perceptions of the emotional resources and capabilities that are specic to emotional experience. Two separate literatures, one in sociology and the other in psychology, suggest a critical individual dierence variable. A sociological construct of `institutional' versus `impulsive' orientation of emotions refers to the extent to which `people may nd evidence of their real, deeper selves either in institutional behavior and feelings, or else in impulsive acts and feelings' (Gordon, 1989, p. 117). Persons whose emotions are institutionallyoriented `focus [their] eorts on achieving and maintaining institutional standards for feeling and expression, thereby upholding widely shared values', whereas `a perspective centered on impulse [italics his] emphasizes spontaneous, uninhibited emotion, unregulated by institutions`. Ashforth and Humphrey (1993) suggested that persons having an `institutional' orientation to their emotions may be expected to experience less emotional dissonance and higher well-being in conforming to display rules than do persons having an `impulsive' orientation ( p. 100). Within the eld of psychology, Laird and Crosby (1974) examined trait dierences in the extent to which persons used `self-produced cues' versus `situationally produced cues' in making selfattributions about their own emotions. Participants were required to either frown or smile while watching various cartoons. Those who reported that their emotions were typically selfproduced demonstrated a signicantly stronger relationship between the emotional expression
Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

WORKPLACE EMOTIONAL LABOR DIMENSIONS

167

manipulation and the moods they experienced just after the cartoons. There are evident parallels between the `institutional orientation' described by Gordon (1989) and this proclivity for one's emotional expression to follow situation-produced cues, just as there are parallels between Gordon's `impulsive orientation' and the tendency to use `self-produced cues' as stimuli for emotional expression. People who maintain an other-oriented disposition toward the expression of emotion (i.e., `institutional' or `situation-produced') are essentially more willing and able to adapt their emotional expression to the social environment. Hochschild (1983) referred to this as an ability to develop a `healthy false self' in which `inauthentic' emotions are readily expressed in service of the role ( p. 195). Hereafter we refer to this trait as `emotional adaptability'. These very similar lines of inquiry suggest that adaptability to emotional demands is an individual dierence variable. Although emotional adaptability may be construed in numerous ways, for the present purposes we dene persons who naturally adapt their emotions to the situational context as being `emotionally adaptable'. Persons who are more emotionally adaptable in this respect may tend to report fewer requirements to express or suppress emotion. This is because one's perception of the normative requirements to express or suppress emotion is at least partially an attribution that is sensitive to self-perception eects. Specically, when one readily meets demands for emotional expression, he or she does not require explicit cues from other coworkers or the supervisor to emote appropriately. Thus expressions of positive emotion are not easily attributed to role obligations. On the other hand, persons who must put more eort into appropriate emotional expression or who nd this aspect of their role dicult will be made more cognizant of social cues about norms for appropriate emotional expression. Moreover, the very challenge of this aspect of the role will make appropriate emotional expression more easily recognized as a role requirement for these persons than it is for emotionally adaptable persons. Thus we predict that whereas more emotionally adaptive persons may or may not emote dierently compared to less emotionally adaptive persons, the former will not experience role appropriate emotional expression as emotional labor to the same extent as the latter. Hypothesis 2. Emotional adaptability will be negatively related to reports of requirements to express or suppress emotion. When researchers speak of a dissonance created by demands to display emotions that one does not actually feel, then emotional labor must be seen as an interaction between dispositions of the individual and his or her job context. Persons are more likely to nd their spontaneous emotions to be at odds with emotional display rules when they have particular aect tendencies. Research has identied two basic, essentially orthogonal trait continua pertaining to aect. Trait positive aect (PA) refers to a tendency for one to be frequently energetic, active, alert, and enthusiastic. Persons scoring high on trait negative aect (NA) tend, even in normal situational contexts, to more frequently experience negative emotions such as anxiety, disgust, guilt, and fear. Trait PA is associated with positive evaluations of social environments, whereas trait NA is associated with correspondingly negative evaluations (Watson and Clark, 1984; Watson et al., 1988). A person's aect tendencies may lead directly to higher perceptions of requirements to express or suppress emotion because particular aective styles will be more or less congruent with particular emotional display rules. A person who naturally experiences positive emotions on a regular basis, for example, may hardly be cognizant of others' expectations to `act cheerful'. Such behavior is not likely to be perceived as role-oriented for such a person, but rather authentic. Likewise, a high NA person may have diculty suppressing the expression of negative emotions since these are felt more frequently and deeply than for a low NA person. Thus any social cues that he or she receives concerning the unacceptability of negative eerence (cues to which `cheerful' people are
Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

168

J. SCHAUBROECK AND J. R. JONES

seldom exposed) may be perceived to dene emotional display rules for such persons. This suggests the following hypotheses: Hypothesis 3. Trait positive aect will be negatively associated with perceptions of role expectations to display positive emotional eerence and negatively associated with role expectations to suppress negative emotional eerence. Hypothesis 4. Trait negative aect will be positively associated with perceptions of role expectations to suppress negative emotional eerence and positively associated with role expectations to express positive emotional eerence. Trait aect reects one's modal aective states and thus represents states experienced on the job as well as o the job. For some persons modal state aect may dier between work and other domains, and their aect on the job may more strongly inuence perceived emotional labor. Therefore, in conjunction with examining the eects of trait aect on perceived emotional labor, we also examined the eects of state aect while the individual is at work.

Gender and emotional labor Hochschild (1983) described the close relationship between emotional labor and stereotypical gender roles in American society. `In general, they [women] are thought to manage expression and feeling not only better but more often than men do . . . the evidence seems clear that women do more emotion managing than men' (both italics hers; p. 164). Indeed, much of the literature on emotional labor focuses on women's roles, both at work and at home. Hochschild (1983) did not distinguish conceptually between the two dimensions of requirements to express or suppress emotion proered in this study. However, in her studies of female ight attendants, Hochschild referred to both facets in referencing the higher prevalence of emotional labor among women: `Women . . . do extra emotion work especially emotion work that arms, enhances, and celebrates the well-being and status of others' [italics hers]. She also suggested that `Women are more likely to be presented with the task of mastering anger and aggression in the service of ``being nice'' ' (1983, p. 163). These suppositions allude to the distinction we have made between demands to express positive emotion and demands to suppress negative eerence. Whereas Wharton (1993) found a higher prevalence of women in jobs that tend to require high amounts of emotional labor (using Hochschild's occupational coding scheme), that relationship was explained by other demographic variables and income level. But Hochschild's discourse about women and emotional labor goes beyond the sociological task of describing pre-market factors which dierentially assign the sexes to occupations. Rather, her discussion of the social psychology of women in all manner of occupations, including traditionally male roles, suggests that women will experience more emotional labor even after the sociological status factors are taken into account. Women were suggested to take more initiative in undertaking emotional labor at work, and co-workers were suggested to have higher expectations of emotion work from women: `The world turns to women for mothering, and this fact silently attaches itself to many a job description' (Hochschild, 1983; p. 170). This suggests the following hypothesis: Hypothesis 5. After controlling for pre-market factors that may dierentiate women and men generally, women will still be found to report higher levels of requirements to express or suppress emotion than men.
Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

WORKPLACE EMOTIONAL LABOR DIMENSIONS

169

Direct and interactive eects on physical symptoms

Laboratory research suggests that requirements to display emotions often lead to physical and emotional reactions that are consistent with the instructions. As stated by Notarious et al. (1982), `Facial expressions are subject to strong personal monitoring and voluntary control and thus likely vary as a function of the display rules operating in a given context' ( p. 401). One related line of laboratory research in psychology concerns the `facial feedback hypothesis' (Izard, 1971). As reviewed by Adelmann and Zajonc (1989), facial feedback studies typically require participants to manipulate their facial expressions in ways that contradict the emotional context of their immediate environment. Negative emotional eerence (e.g., frowning) is found to promote adverse emotional experience and physiological indicators, just as positive emotional eerence (e.g., smiling) promotes positive physical and psychological reactions. This research suggests that positive emotional display rules (e.g., `always smile when one is on the job') might have salutary inuences on health. Hypothesis 6. Requirements to express positive emotional eerence on the job will be negatively associated with physical symptoms. Laboratory ndings indicate that eorts to suppress `negative' emotions (i.e., anger, irritability, grief, sadness) often lead to patterns of physiological response that presage somatic illness. When one is prevented from expressing emotions in a psychologically stimulating context, generally a `discharge model' of emotion is supported, wherein high levels of physiological reactivity and negative moods are experienced (e.g., Notarious and Levenson, 1979; Notarious et al., 1982). Likewise, the psychological characteristics of denial and suppression are consistently linked to lower immune levels and susceptibility to viral infection (see the review by S. Cohen and T. B. Herbert, 1996) as well as cardiovascular illness (see the review by Friedman, 1989). Thus this type of emotional labor is expected to be positively associated with physical symptoms. Hypothesis 7. Requirements to suppress negative emotional eerence on the job will be positively associated with physical symptoms.

Role identication moderators Rationalization processes may buer individuals from the health eects of requirements to express or suppress emotion. Specically, individuals who identify strongly with their organizations and/or their jobs may more fully subscribe to the belief that they must often behave in an emotionally `inauthentic' fashion to serve the purposes of their roles. Thus they will experience little cognitive dissonance following such `inauthentic' acts. Varied research indicates that, provided that they are fully identied with a higher organizational purpose, people have little trouble behaving in ways they would normally nd unnatural or even abhorrent. When an individual lacks that sense of purpose, however, considerable emotional dissonance may be created by his or her inauthentic emotional behavior. Hypothesis 8. Perceived requirements to express or suppress emotion will interact with organizational identication and job involvement to predict physical symptoms. As organizational identication or job involvement increase, there will be signicantly weaker relationships between requirements to express or suppress emotion dimensions and physical symptoms.
Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

170

J. SCHAUBROECK AND J. R. JONES

Individual dierence moderators Individual dierences in personality may not only inuence the extent to which people perceive role requirements for emotional expression as noted above, they may also inuence the strength of the relationship between these perceptions and physical symptoms. As noted above, the less emotionally adaptable the individual is, the more he nds it dicult and/or onerous to modulate his emotions in ways that are socially appropriate. Thus the emotionally adaptive individual may be better able to cope with perceived requirements for emotional expression than is the less emotionally adaptive person. Hypothesis 9. As emotional adaptability increases, requirements to express or suppress emotion will be less strongly related to physical symptoms.

Methods
Sample and procedure
Questionnaires were collected from the full-time headquarters employees of a major survey research organization in the mid-west United States. Because the nature of much of the work in this organization is rather complex and requires internal and external interactions with people, the site provided a fairly high incidence of emotional labor. A total of 227 persons returned a questionnaire, of which 217 included complete data for the analyses. Nearly all occupations within the organization were represented in the sample. The average respondent's age was 36.84 years, and 62 per cent of participants were women. Respondents averaged 7.31 years in the organization and 4.29 years in their current position. The mean level of educational attainment was quite high, averaging 0.58 years of postgraduate college education. Respondents did not dier signicantly from the overall organization on any of these demographic data.

Measures
Emotional labor Our approach to measuring emotional labor was to index the character of the emotions that are perceived to be encouraged on the job. The items were scaled using a ve-point Likert format. They focused on norms that must be complied with for: (1) eective job performance; or (2) `To make a good impression on others (e.g., bosses, co-workers, customers, etc.) in my job'. We formed a composite of demands for positive eerence (a 0.87) and suppression of negative eerence (a 0.96) using the products of duration and frequency items. This type of composite measure therefore reects the degree of emotional labor by combining the frequency and the overall amount of attention devoted to these demands. This was done because it is unlikely, for example, that someone who believes it is a norm to smile at passers-by when walking across an oce space experiences this as demanding. Likewise, a worker who has only one or two emotional display rule-laden interactions per day may nd this demanding if these encounters are lengthy. All items were subjected to a principal factor analysis with oblique rotation. All item-products loaded distinctly with coecients exceeding 0.40 on one of two factors, both with eigenvalues
Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

WORKPLACE EMOTIONAL LABOR DIMENSIONS

171

exceeding 1.0 (6.87 and 3.35, respectively). The rst factor comprised the positive eerence items (e.g. `To be eective in my job, I must try to share in the enthusiasm or liveliness of another person.') and explained 42.9 per cent of the overall item variance. The item-products loading on the second factor measured suppression of negative eerence (e.g., `To be eective in my job, I must try to suppress how upset or distressed I may feel.'). These item-products explained the next 21.0 per cent of the variance. (The complete list of emotional labor items is available from the rst author upon request.) Interpersonal interactionoccupation In order to further guarantee condentiality, rather than asking for the respondent's specic job title the questionnaire inquired about his or her job category from a list of 35 occupational units present in the organization. Most all of these categories were readily related to the broad occupational dimensions which Hochschild (1983) identied as having high levels of emotional labor. We also asked two managers to identify which categories on the list contained high needs for intensive interaction with others. A few job categories were not in Hochschild's list of occupations but clearly required high levels of intensely personal interaction with other persons (i.e., Executive Interviewer, Research Consultant, Interviewer, Computing Servicesoperational support). Given the service nature of the organization, it is not surprising that a signicant portion of the respondents who identied their occupational category (n 195) were included in the `high emotional labor' group (72.3 per cent). The remaining occupational categories were not on Hochschild's list. The managers indicated that these latter occupations required little interpersonal interaction or those interactions that were required were not qualitatively demanding in interpersonal interaction. Examples of jobs that were not coded as being high in emotional labor were building services, research analyst, data coding/entry, mailing/scanning/proong, and accountant. High emotional labor job categories were coded `2' and the other categories were coded `1' to form an interpersonal interaction variable. Physical symptoms Physical symptoms were measured using Caplan et al.'s (1975) 20-item somatic complaints index (a 0.86). Respondents are asked `How often have you experienced any of the following events during the past month? Rarely (1) to Very often (5)', (e.g., `You were in ill health that aected your work'; `You had trouble sleeping at night.`). Individual dierence variables Age, gender, trait negative aect, and education were measured on the questionnaire. These were included as control variables in tests of models that examined physical symptoms as a dependent variable. We also included quantitative measures of smoking and physical exercise, but these variables did not correlate signicantly with physical symptoms. State negative and positive aect were controlled in analyses examining trait aect and its correlations with requirements to express or suppress emotion. Trait negative aect has been found to be a confound variable in examining relationships between stressors and health outcomes when these are both measured by selfreports (Brief et al., 1988). For the analyses examining the relationship between gender and perceived requirements to express or suppress emotion, we also controlled for self-reported salary and job complexity. Job complexity was measured using a shortened six-item version of the instrument used by Caplan et al. (1975) and House (1980). These items (e.g., `How often does your job require that you think
Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

172

J. SCHAUBROECK AND J. R. JONES

through complex problems or situations?') provided an index of the extent to which the job was a traditionally `male-gendered' role carrying higher social status, and thus together with the other controls (e.g., education) it assisted in controlling for social status. Salary data were obtained by a separate survey designed by the organization and conducted one month earlier. Data on salary and the other variables combined were thus available for just over one-half (n 123) of the respondents. The respondent was asked to indicate which of six salary range categories included his or her own salary. We used the Positive and Negative Aect Scale (PANAS; Watson et al., 1988) to measure both forms of aect. There are 20 adjectives, 10 reecting negative emotional states (e.g., `irritable`) and 10 reecting positive emotional states (e.g., `excited'). The PANAS respondent is asked to record the extent ( from `very slightly or not at all' [1] to `extremely' [5]) to which he or she has experienced the emotion during a specied interval. Watson et al. (1988) note that this interval can be dened to represent state aect ranging from momentary states, single days, weeks, to years, with `generally feel that way' representing trait aect. Both the positive state aect (PA) and negative state aect (NA) scales evidenced sound internal consistency reliability (a 0.91 and 0.83, respectively). Watson et al. (1988) observed similarly strong internal consistencies for `momentary' PANAS scales as well as reasonably high testretest reliabilities over an eight week period (r 0.54 (PA) and r 0.45 (NA)). For our state aect indexes, `during the past hour' was the reference period. The same PANAS questionnaire modied to measure trait aect (`. . . indicate the extent you generally feel that way [both at work and away from work], that is, on average.') was included near the end of the questionnaire. The trait PA and the trait NA items provided coecient alpha reliabilities of 0.89 and 0.87, respectively. 'Emotional adaptability' represents the extent to which one's expressed emotions are adapted easily to the institutional context. This instrument was adapted from three items developed by Laird and Crosby (1974); (a 0.68; e.g., `Frequently one moves from a situation on the job in which one is feeling one way to a situation in which it is appropriate to feel some other way, as, for instance, when you have an argument with someone and then immediately must be with someone else. How rapidly do you change from the rst kind of feeling to the second? 1 very slowly to 4 very rapidly'). Other dispositional variables We adapted the ve items developed by Mael (1988unpublished doctoral dissertation, Wayne State University, Detroit) to measure organizational identication. The items (a 0.92) were adapted by making specic reference to the name of the organization (e.g., `When someone criticizes [name of organization], it feels like a personal insult.'). The three-item job involvement instrument from the Michigan Organizational Assessment Questionnaire (MOAQ; Camman et al. (1969)) used a 17 `very inaccurate' to `very accurate' scale (a 0.84; `I am very much involved personally in my job'; `I live, eat, and breathe my job'; `The most important things which happen to me involve my job').

Results
Correlations among all variables are shown in Table 1. Emotional adaptability was positively associated with state and trait positive aect and negatively correlated with state and trait
Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

Table 1. Correlations among study variables and descriptive statistics

Variable name 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. Mean S.D. 1 0.06 0.22 0.44 0.19 0.08 0.12 0.27 0.16 0.21 0.11 0.18 0.21 0.11 0.90 0.12 0.27 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Age (years) 36.84 12.3 Gender (1 male; 2 female) 1.62 0.49 Education (years) 16.58 13.1 Tenure ( job) 4.29 4.91 Job complexity 3.80 0.79 Demands to express positive emotions 11.33 5.83 Demands to suppress negative emotions 5.53 4.39 Physical symptoms 1.85 0.53 Organization identication 5.89 1.32 Job involvement 4.96 1.45 Emotional adaptability 2.74 0.83 State positive aect 3.33 0.86 Trait positive aect 3.93 0.62 State negative aect 1.55 0.55 Trait negative aect 1.76 0.60 Emotional labor occupation (Hochschild, 1983) 1.72 0.45 Salary 3.49 1.69

0.23 0.13 0.19 0.03 0.29 0.17 0.22 0.07 0.03 0.10 0.11 0.03 0.11 0.04 0.30 0.08 0.21 0.07 0.16 0.03 0.00 0.23 0.01 0.21 0.09 0.41 0.12 0.16 0.01 0.15 0.11 0.22 0.05 0.27 0.09 0.20 0.06 0.36 0.09 0.09 0.09 0.02 0.12 0.05 0.09 0.16 0.06 0.02 0.12 0.16 0.21 0.39 0.30 0.33

0.35 0.28 0.14 0.13 0.07 0.08 0.09 0.13 0.05 0.17 0.17

WORKPLACE EMOTIONAL LABOR DIMENSIONS

0.43 0.23 0.18 0.04 0.21 0.31 0.37 0.48 0.03 0.07

0.03 0.02 0.67 0.04 0.25 0.17 0.14 0.36 0.44 0.20 0.39 0.44 0.42 0.17 0.13 0.58 0.25 0.17 0.04 0.01 0.01 0.05 0.01 0.24

0.24 0.23 0.58 0.07 0.18 0.18 0.17 0.21 0.34 0.52 0.03 0.04 0.05 0.08 0.09 0.02 0.23 0.13 0.04 0.08

0.06

Variables 115: pairwise n's 218222 (critical value of r ( p 5 0.05) 0.13; p 5 0.01 0.18; p 5 0.001 0.23). Variable 16: pairwise n's 191195 (critical value of r ( p 5 0.05) 0.15; p 5 0.01 0.19). Variable 17: pairwise n's 121123 (critical value of r ( p 5 0.05) 0.19; p 5 0.01 0.24; p 5 0.001 0.30).

173

174

J. SCHAUBROECK AND J. R. JONES

negative aect. The two measures of emotional labor, demands to suppress negative eerence and demands to express positive eerence, correlated signicantly (r 0.35, p 5 0.0001).

Predictors of emotional demands

Consistent with Hypothesis 1, Hochschild's (1983) coding of the occupational units in this sample for `objective' expectation of emotional labor (based largely on expected quantity of interpersonal interaction) was positively correlated with demands to express positive emotion (r 0.17, p 5 0.01). However, the hypothesis was only partially supported as this variable did not correlate with demands to suppress negative eerence (r 0.03). The second hypothesis predicted that emotional adaptability would be negatively associated with reports of requirements to express or suppress emotion. After controlling for trait negative aect, age, education, and gender, emotional adaptability was not associated with perceived demands to suppress negative emotion (DR2 0.00, F(1, 206) 0.72, p 5 0.40), nor was this trait associated with perceived demands to express positive emotion on the job (DR2 0.01, F(1, 206) 2.31, p 5 0.13). Thus Hypothesis 2 was not supported. To examine the incremental contributions of the dispositional aect variables on emotional labor perceptions, we rst regressed each emotional labor index on the demographic controls (age, sex, and education) and then we entered both trait aect variables. These variables (trait NA and PA) together explained a good deal of the variance in demands for suppression of negative eerence (DR2 0.25, F(4, 203) 18.09, p 5 0.0001), but very little of the variance in demands for positive emotional expression (DR2 0.01, F(4, 203) 0.56, p 5 0.67). Thus Hypothesis 3 ( pertaining to trait PA) and Hypothesis 4 ( pertaining to trait NA) were supported for the suppression of the negative eerence variable but not the positive emotional expression variable. The state aect variables (state PA and NA) next entered the equation. After controlling for trait aect, the block of state aect variables was marginally related to demands for suppression of negative eerence ((DR2 0.01, F(2, 203) 2.85, p 5 0.06) but it was unrelated to demands for positive expression (see Table 2). Gender ( female 2, male 1) correlated positively with perceived demands to express positive emotion (r 0.22, p 5 0.001). There was no relationship between gender and Hochschild's (1983) coding of occupations that are seen as high in emotional labor (`interpersonal interactionoccupation'; r 0.06, n.s.). To test Hypothesis 5, positive eerence demands was rst regressed on age, education, job complexity, position tenure, trait negative aect, trait positive aect, and salary level (all employees were salaried). As noted above, salary data were available for only half of the respondents. Together, these variables explained 12 per cent of the variance in positive eerence demands (F(7, 111) 2.13, p 5 0.05). Gender was entered at the next step, and its strong incremental eect on positive eerence demands (DR2 0.06, F(1, 110) 8.18, p 5 0.005) supported the hypothesis (see Table 2). Gender was not signicantly related to perceived demands to suppress negative eerence. Hypothesis 6 predicted a negative relationship between reported demands for positive eerence and physical symptoms. This hypothesis was not supported as there was a positive eect of these kinds of demands on symptoms, after controlling for trait negative aect, age, education, and gender (see Table 3). The eect of demands to suppress negative eerence (Hypothesis 7) was not statistically signicant ( p 5 0.10). Whereas neither hypothesis was supported, together the two dimensions of emotional labor explained signicant variance in symptoms (DR2 0.04, F(2, 200) 6.91, p 5 0.001).
Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

Table 2. Dimensions of emotional labor predicted by trait aect variables and gender (regression coecients are standardized) Variable name Trait aect hypotheses Demands to express positive eerence Step 1 Age Gender Education DR2 0.07; F(3, 207) 5.52} Step 2 Trait negative aect Trait positive aect DR2 0.01; F(2, 205) 0.56 Step 3 State negative aect State positive aect DR2 0.01; F(2, 203) 1.63
* p 5 0.10; { p 5 0.05; { p 5 0.01; } p 5 0.001

Variable name

Gender hypothesis Demands to express positive eerence Demands to suppress negative eerence 0.06 0.06 0.47} 0.16 0.06 0.01 0.14 DR2 0.32; F(7, 111) 7.36} 0.08 DR2 0.01; F(1, 110) 0.74

Demands to suppress negative eerence 0.14{ 0.12* 0.09 DR2 0.03; F(3, 207) 2.11 0.40} 0.19{ DR2 0.24; F(2, 205) 33.84} 0.16{ 0.02 DR2 0.02; F(2, 203) 2.85 Step 1 Age Education Trait negative aect Trait positive aect Tenure ( job) Job complexity Salary DR2 0.10; F(7, 111) 1.81*

0.09 0.25} 0.16{

0.03 0.08

0.19 0.10 0.17 0.17 0.11 0.09 0.17*

WORKPLACE EMOTIONAL LABOR DIMENSIONS

0.13 0.08

Step 2 Gender (1 male; 2 female) DR2 0.07; F(1, 110) 9.13{

0.29{

175

Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

176 J. SCHAUBROECK AND J. R. JONES

Table 3. Eects of emotional labor on physical symptoms (regression coecients are standardized) Physical symptoms Step 1 Age Gender Education Trait negative aect DR2 0.43; F(4, 202) 37.27} Step 2 Demands to express positive eerence (EPE) Demands to suppress negative eerence (SNE) DR2 0.04; F(2, 200) 6.91} Step 3 Emotional adaptability (EA) DR2 0.01; F(1, 199) 3.10* Step 4 EPE EA SNE EA DR2 0.02; F(2, 197) 2.68{ 0.20} 0.23} 0.03 0.53} Physical symptoms Physical symptoms

0.13{ 0.12

0.09*

Step 3 Organizational identication (OI) DR2 0.01; F(1, 198) 4.39{ Step 4 EPE OI DR2 0.02; F(1, 197) 6.89{

0.12{ Step 4 SNE OI DR2 0.01; F(1, 197) 3.16

0.45{ 0.17

0.15{

0.10

Note: Where Step 1, 2, or 3 is omitted, the statistics are identical to corresponding steps in previous statement. * p 5 0.10, { p 5 0.05, { p 5 0.01, } p 5 0.001.

WORKPLACE EMOTIONAL LABOR DIMENSIONS

177

Job involvement and organizational identication were substantially correlated (r 0.67) and thus it is not surprising that their respective patterns of relationships with emotional labor were essentially isomorphic. Hypothesis 8 predicted that the relationship between emotional labor and physical health would be stronger among persons reporting little identication with the organization or lower job involvement. In omnibus regressions that included all main eects (age, gender, education, trait NA, both emotional labor factors, and then either organization identication or job involvement), followed by the two product terms, the overall interaction block was signicant when organizational identication was the dispositional variable (DR2 0.02, F(2, 196) 2.94, p 5 0.05) but not when job involvement was the dispositional variable (DR2 0.01, F(2, 196) 1.65, p 5 0.20). Because these variables were signicantly correlated, they were examined in separate regression equations. Examined individually, the interaction between organizational identication and demands for positive eerence predicting physical symptoms was signicant (DR2 0.02, F(1, 199) 6.14, p 5 0.01), as was the case for job involvement (DR2 0.01, F(1, 198) 3.98, p 5 0.05). We plotted this interaction by inserting cut-point values of / one standard deviation from the mean of each variable into the regression equation (Cohen and Cohen, 1983). The plot reveals that whereas positive eerence demands were positively related to physical symptoms among persons with low organizational identication, there was no relationship between these variables among persons reporting high organizational identication (see Figure 2). A largely isomorphic pattern was

Figure 2. Interaction between perceived demands for positive emotional eerence and organizational identication predicting physical symptoms Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

178

J. SCHAUBROECK AND J. R. JONES

observed for job involvement, with only the less job-involved persons appearing to suer symptoms from perceived demands to express positive eerence. Demands to suppress negative emotion did not interact with organizational identication or job involvement in predicting health. Hypothesis 9 predicted that emotional adaptability would moderate the relationship between requirements to express or suppress emotion and physical symptoms. To test this hypothesis we rst entered the control variables (trait negative aect, age, gender, and education) and then the main eects of the emotional labor indexes. Emotional adaptability was entered at the next step. Its incremental main eect on physical symptoms approached statistical signicance, with more emotionally adaptive persons reporting fewer symptoms (DR2 0.01, F(1, 199) 3.10, p 5 0.08). The block of two interaction terms (the products of emotional adaptability and each emotional labor index) was signicant (DR2 0.02, F(2, 197) 2.68, p 5 0.05) and therefore this hypothesis was supported (see Table 3). Of the two, only the product term involving emotional adaptability and demands to express positive eerence was signicant. The plot of this interaction (see Figure 3) indicated that this aspect of emotional labor was strongly and positively related to physical symptoms among persons with weaker tendencies in managing their emotions. There was virtually no relationship among persons having greater ability to adapt their emotions to a given context.

Figure 3. Interaction between perceived demands for positive emotional eerence and emotional adaptability predicting physical symptoms Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

WORKPLACE EMOTIONAL LABOR DIMENSIONS

179

Discussion
There has been very little quantitative empirical research investigating emotional labor in the workplace. The present study expanded on the conceptual work and laboratory studies in psychology by distinguishing two separate dimensions of emotional labor (demands to express positive eerence and requirements to suppress negative eerence) and by testing predictions about situational and dispositional determinants and moderators.

Dispositional correlates We observed that perceived demands to express positive eerence were correlated with perceived demands to suppress negative emotion, as might be expected, but these variables represented separate factors of emotional labor that exhibited discriminant validity. These two dimensions were correlated with other variables in dierent ways. Demands for positive eerence was related to gender and an occupational `emotional labor' classication ( following Hochschild's (1983) scheme). Demands for suppression of negative eerence was related to trait aect (PA and NA). Thus both situational and dispositional variables appear to be plausible antecedents of perceived emotional labor. A basic pattern observed within this study was that the personality variables trait NA and trait PA predicted perceived demands to suppress negative emotion, whereas they were largely uncorrelated with perceived demands to express positive emotion. Conversely, as we summarize below, the situational variable of occupational status (independently assessed volume and intensity of interpersonal interaction on the job) was correlated with perceived demands to express positive emotion but not perceived demands to suppress negative emotion.

Emotional laborhealth linkages The perception of demands to express positive emotion on the job was positively related to physical symptoms. The eect of role requirements to express positive emotion on symptoms was mainly among individuals who reported little identication with the organization or who were less involved with their jobs. The latter nding is consistent with the perspective that emotional labor is most unhealthful when one's emotional expressions on the job are not an authentic representation of one's personal beliefs. One set of factors that may dierentiate whether conformance to an emotional display rule is `authentic' is one's trait aect. Whereas these factors did not moderate the relationship between perceived emotional labor and physical symptoms, they were related to reports of emotional labor in the predicted directions. After accounting for trait negative and positive aect, the corresponding state aect variables had little relationship with perceived emotional labor. However, future research might protably examine modal state aect on the job in addition to trait aect and aective states around the time the responses are elicited. How one perceives emotional labor may be even more strongly inuenced by `job aect' than it is by trait aect. Emotional labor seems to be embedded in social and cultural contexts in complex ways. `Authenticity' in how people respond to their own emotion work seems to play an important role in its eects on health. Hochschild (1983) noted that the psychological eects of emotion work among the ight attendants she studied may have been linked to such authenticity perceptions. She suggested that when workers do not `own' their emotions and behavior, they suer more
Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

180

J. SCHAUBROECK AND J. R. JONES

psychological consequences from emotional labor. It is not the emotion labor itself that is distressing, but one's perceptions of the emotions as being inauthentic. Two of our ndings provide clues about the veracity of this supposition. Positive eerence demands were positively associated with illness among persons reporting low organizational identication or low job involvement whereas they were not at all associated with illness among persons reporting higher identication or involvement. Emotion work would seem to be more authentic ( following a characteristic avenue of self-expression) among persons who identify closely with their roles. Hochschild (1983) cautioned, however, that in extreme cases when the emotion worker `identies too wholeheartedly with the job,' he or she risks burnout. The trick is for the worker to recognize that when engaging in false emotions he or she is doing so on behalf of the organization. This `healthy estrangement' from the role (1983, p. 188) mitigates dissonant self-perceptions, but it is often dicult to achieve because `the company's purposes insinuate themselves into the way workers are asked to interpret their own feelings' (1983, p. 197). Taken together with the evidence from case studies and the Morris and Feldman (1997) quantitative study, emotional labor may have some negative health consequences. Psychological studies have observed that the emotions people express, even when they are not `authentic', can inuence how they feel. On that basis we predicted that requirements to express positive emotion on the job generally have a salutary inuence on health. This hypothesis was not supported in the present study. Perhaps conformity to such display rules does positively inuence mood on the job (it correlated positively with state positive aect), but any salutary emotional inuences on health are over-ridden by feelings of pressure to behave in inauthentic ways. The emotional adaptability variable also provided evidence concerning the role of authenticity in how one responds to emotion work. A person scoring low on this factor may be expected to have diculty separating his or her or `true self' from his or her `healthy false self' (Hochschild, 1983, p. 195). Persons scoring high on emotional adaptability are more self-aware emotionally, and their emotions are `more subject to command and manipulation, more amenable to various forms of management' as Hochschild described many of her participants (1983, p. 193). Based on our regression ndings, emotional labor (in the form of having to express positive eerence on the job) was not related to physical symptoms among persons who were emotionally adaptive in this way. Such persons may be less conicted in their emotions when confronted with particular display rules. They do not experience cognitive dissonance over a conict between their felt emotions and the emotions they must display. Together with our ndings pertaining to organizational identication and job involvement as moderator variables, these results suggest that self-construals of emotional authenticity may indeed be critical in determining the consequences of emotional labor, as suggested by Hochschild (1983). The context wherein demands for eerence occur may be a major factor in determining their physiological outcomes. The facial feedback research, which nds that salutary eects follow forced positive eerence, is conducted under laboratory conditions in which participants participate willingly for brief intervals. As stated by Notarious et al. (1982), `it is possible to question the ecological validity of [ facial feedback] paradigms because their interpersonal context is probably not characteristic of emotional situations confronting persons in the real world' ( p. 401). Under laboratory conditions, any conicts between one's cognitions and the emotions one is required to express are less likely to lead to negative self-perceptions. Gender and emotional labor As had been speculated by Hochschild (1983) to be a phenomenon in American society in general, this study found a higher prevalence of perceived emotional labor among women than
Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

WORKPLACE EMOTIONAL LABOR DIMENSIONS

181

among men. In the present study her occupational coding scheme of emotional labor was observed to converge only weakly (albeit to a level of statistical signicance) with perceived emotional labor. In the present study we examined the actual experience of emotional labor as perceived by the job incumbent. A signicant relationship between gender and perceived demands to express positive emotion persisted even after variables which may account substantially for dierences in occupational and social status (i.e., salary, education, age, job complexity) were included in the analysis. This may support Hochschild's (1983) view that the traditional prevalence of women in jobs that require more emotional labor reects more than just a `pre-market' bias whereby women self-select into emotionally-intensive occupations. It is also possible that within the same occupation the expectations of customers and co-workers dier toward men and women in a way that `genders' the work role in this respect. Finally, women may engage in more emotional labor on their own initiative. These explanations are not mutually exclusive. The eect of gender was limited to positive eerence demands, and thus this study supports Hochschild's (1983) suspicions that women, to a greater extent than men, undertake `emotion work that arms, enhances, and celebrates the well-being and status of others' [italics hers; p. 165]. However, Hochschild also suggested that `Women are more likely to be presented with the task of mastering anger and aggression in the service of ``being nice'' ' ( p. 163). In this sample, however, we observed only a marginal relationship between gender and demands for suppression of negative eerence. The role of occupation in emotional labor Emotional labor was related to various trait and situational variables in complex ways. The fairly low magnitude correlations of our indexes with Hochschild's (1983) occupational index is not particularly surprising given that such occupational dummy variables are known to be rather decient indices of psychological demands (cf. Ganster and Schaubroeck, 1991). A great deal of within-occupation variation in emotional labor is to be expected. This study indicates that what an individual experiences as emotional labor may reect a complex matrix of determining factors, both internal and external to the individual. If this is true, then simpler formulations about the role of emotional labor in health may not be protable. Study limitations The usual caveats about making causal inferences based upon cross-sectional, correlational data apply here. For example, relationships between perceived emotional labor and health complaints may partially reect a consistency artefact. One might be especially sceptical about correlations between a putative stressor that has been shown to be related to trait aect, as prior work has shown that relationships between perceived stressors and self-report symptoms are often explained by other variables, especially trait negative aect (Brief et al., 1988). However, we believe the present ndings are quite strongly suggestive of a reliable linkage between emotional labor dimensions and physical symptoms for two reasons. First, the tests of their eects on symptoms controlled for trait negative aect and demographic risk variables. Second, some of our ndings were interaction eects, and these interaction eects followed patterns of theoretical expectations that were of such a complexity as to be less plausibly explained by response set biases such as priming or consistency. Nevertheless, several of our hypothesis tests involved simple linear relationships and therefore common method variance is a possible threat to internal validity. In addition, we focused on only a single organization and a limited range of occupations.
Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

182

J. SCHAUBROECK AND J. R. JONES

We must also caution against generalizing from our study, which predicted somatic symptoms in an occupationally diverse service organization, to a broader range of health disorders and to particular occupations.

Conclusions
The results of this study suggest that it may be necessary for researchers investigating emotional labor to distinguish conceptually and empirically between perceptions of required eerence and imputed characteristics of the job or occupation. The extent to which individuals perceive that they are required to express or suppress certain types of emotional expression may depend as much on their emotional predispositions as it does on the objective characteristics of their organizational roles. Although the present study did not exhaust the list of all possible individual dierence antecedents, the relationships we observed between individual dierences and perceived emotional labor were not so strong as to suggest that emotional labor is some form of psychological epiphenomenon. It may be that the characteristics of one's role requirements actually dier substantially among incumbents of the same job title. Future research may protably investigate the extent to which dierences in the expectations held by others determine self-perceptions of emotional labor. This would require collecting data from peers, supervisors, perhaps other co-workers, and even customers or other types of clients. Our study also observed that the extent to which people perceive that they are required to express or suppress emotion on the job has some adverse eect on physical health. Moreover, to the extent that people identify closely with their roles and/or they have diculty in adjusting their emotional expressions to the social context, the eect of perceiving such requirements is more unhealthful. Increasingly organizations put pressure on workers to maintain a positive `face' to the public and others in the workforce. If workers are to thrive in such service-oriented environments they would benet from training and/or selection practices that focus on emotion management. As noted by Ashkanasy (1996), emotional displays have import not just to the actor himself or herself but also to those individuals who must attempt to gauge and act upon what they infer from others' emotional displays. Thus if individuals can be supported in their attempts to manage their emotions in healthy ways, it is likely that their emotional displays will be more authentic. This may enhance the overall healthfulness of the work environment by improving the quality of interpersonal relationships. Future research should therefore explore not only the individual consequences of emotional labor, such as health symptoms and satisfaction, but how perceived emotional display rules inuence the quality of the interpersonal environment by inuencing social perception processes.

References
Adelmann PK, Zajonc RB. 1989. Facial eerence and the experience of emotion. Annual Review of Psychology 40: 249280. Ashforth BE, Humphrey RH. 1993. Emotional labor in service roles: the inuence of identity. Academy of Management Review 18: 88115. Copyright # 2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 163183 (2000)

WORKPLACE EMOTIONAL LABOR DIMENSIONS

183

Ashkanasy NM. 1996. Distinguishing felt from displayed emotion in the workplace: implications for management research. Paper presented at the annual meeting of the Academy of Management, Cincinnati, USA. Brief AP, Burke MJ, George JM, Robinson BS, Webster J. 1988. Should negative aectivity remain an unmeasured variable in the study of job stress? Journal of Applied Psychology 73: 193198. Camman C, Fichman M, Jenkins GD Jr, Klesh JR. 1969. Assessing the attitudes and perceptions of organization members. In Assessing Organizational Change, Goodman PR (ed.); Wiley: New York. Caplan RD, Cobb S, French JRP, Harrison R, Pinneau SR. 1975. Job demands and worker health. (NIOSH publication no. 75-160). Washington, DC: Department of Health, Education, and Welfare. Cohen J, Cohen P. 1983. Applied Multiple Regression Analysis in the Behavioral Sciences. Erlbaum: Hillsdale, NJ. Cohen S, Herbert TB. 1996. Health psychology: psychological factors and physical disease from the perspective of human psychoneuroimmunology. Annual Review of Psychology 47: 113142. Ekman P. 1972. Universals and cultural dierences in facial expressions of emotion. In Nebraska Symposium on Motivation, Cole JK (ed.); University of Nebraska Press: Lincoln, NE; 207283. Friedman HS. 1989. The role of emotional expression in coronary heart disease. In In Search of Coronaryprone Behavior: Beyond Type A, Siegman AW, Dembroski TM (eds); Erlbaum: Hillsdale, NJ; 149168. Ganster DC, Schaubroeck J. 1991. Work stress and employee health. Journal of Management 17: 235271. Gordon SL. 1989. Institutional and impulsive orientations in selectively appropriating emotions to self. In The Sociology of Emotions: Original Essays and Research Papers, Franks DD, McCarthy ED (eds); JAI Press: Greenwich, CT; 115135. Hochschild AR. 1979. Emotion work, feeling rules, and social structure. American Journal of Sociology 85: 551575. Hochschild AR. 1983. The Managed Heart: Commercialization of Human Feeling. University of California Press: Berkeley, CA. House JS. 1980. Occupational Stress and the Physical and Mental Health of Factory Workers. Institute for Social Research: Ann Arbor, MI. Izard CE. 1971. The Face of Emotion. Appleton-Century-Crofts: New York. King LA, Emmons RA. 1990. Conict over emotional expression: psychological and physical correlates. Journal of Personality and Social Psychology 58: 864877. Laird J, Crosby M. 1974. Individual dierences in self-attribution of emotion. In Thought and Feeling: Cognitive Alteration of Feeling States, London H, Nisbett RE (eds); Aldine: Chicago; 4459. Morris JA, Feldman DC. 1996. The dimensions, antecedents, and consequences of emotional labor. Academy of Management Review 21: 9861010. Morris JA, Feldman DC. 1997. Managing emotions in the workplace. Journal of Managerial Issues IX: 257274. Notarious CI, Levenson RW. 1979. Expressive tendencies and physiological response to stress. Journal of Personality and Social Psychology 37: 12041210. Notarious CI, Wemple C, Ingraham LJ, Burns TJ, Kollar E. 1982. Multichannel responses to an interpersonal stressor: interrelationships among facial display, heart rate, self-report of emotion, and threat appraisal. Journal of Personality and Social Psychology 43: 400408. Rafaeli A, Sutton RI. 1987. Expression of emotion as part of the work role. Academy of Management Review 12: 2337. Sutton RJ. 1991. Maintaining norms about expressed emotions: the case of bill collectors. Administrative Science Quarterly 36: 245268. Sutton RI, Rafaeli A. 1988. Untangling the relationship between displayed emotions and organizational sales: the case of convenience stores. Academy of Management Journal 31: 461487. Watson D, Clark LA. 1984. Negative aectivity: the disposition to experience negative emotional states. Psychological Bulletin 96: 465490. Watson D, Clark LA, Tellegen A. 1988. Development and validation of brief measures of positive and negative aect: the PANAS scales. Journal of Personality and Social Psychology 54: 10631070. Wharton AS. 1993. The aective consequences of service work: managing emotions on the job. Work and Occupations 20: 205232. Wharton AS, Erickson RJ. 1993. Managing emotions on the job and at home: understanding the consequences of multiple emotional roles. Academy of Management Review 18: 457486.

Copyright # 2000 John Wiley & Sons, Ltd.

J. Organiz. Behav. 21, 163183 (2000)