Going beyond deep and surface acting: a bottom-up taxonomy of strategies used in response to emotional display rules

ABSTRACT Research on emotional labour has primarily focused on two emotional labour strategies that employees may use when confronted with organizational display rules: deep acting and surface acting. Initial evidence suggests that these two strategies do not fully cover the range of strategies that employees engage in when responding to display rules in interactions with customers. Yet, a systematic overview of the full range of such strategies is missing. Across two studies, we used a bottom-up approach to develop a taxonomy of strategies used in response to display rules. In Study 1, we conducted interviews in the Netherlands and in Turkey to collect a comprehensive list of employee-generated strategy statements. We found that deep and surface acting only partially accounted for the strategy statements (Netherlands: 47.6%; Turkey: 53.3%). In Study 2, we asked a new sample of participants to sort the strategy statements into categories based on their similarity. Hierarchical cluster analysis showed that employees engage in deep and surface acting in response to display rule requirements. However, employees also engage in additional strategies: changing customers’ cognitions or affect, solution-oriented, waiting, and avoidance strategies. These results reveal that employees rely on a wide range of strategies to deal with display rules.

The service sector is a key component of modern economies.In member countries of the Organization for Economic Cooperation and Development (OECD), more than half of the employees are working in service jobs (Central Intelligence Agency, 2020;The World Bank, 2020).Such jobs often involve display rules specifying which emotions employees should express when interacting with clients because employees' expressions are critical to enhancing service experience, establishing lasting relationships with clients, and ultimately promoting the organization's success (Diefendorff & Greguras, 2009;Grandey, 2000;Wang et al., 2017).For example, a waiter is expected to greet a customer with a friendly smile and positive attitude, or a flight attendant is expected to approach a fearful passenger with a reassuring smile and empathetic demeanour.
In the context of organizations, emotional display rules are often characterized as job requirements that serve to regulate and standardize employees' emotional expressions to achieve the organization's goals (Diefendorff & Greguras, 2009;Diefendorff et al., 2010;Grandey, 2000;Rafaeli & Sutton, 1987).They can either be explicit, conveyed through formal practices (e.g., training), or implicit, conveyed through informal practices and norms (e.g., work climate; Grandey et al., 2020).In client interactions, organizations often expect their employees to show positive emotions or hide their true emotions (Diefendorff & Greguras, 2009).Employees may adopt a wide range of strategies to deal with these rules.Prior research has predominantly focused on two classes of strategies: deep acting and surface acting (Grandey, 2000).These strategies are studied in the domain of emotional labour (Grandey, 2000) and it has been shown that both strategies are fundamentally different.When engaging in deep acting, employees modify their own inner emotions such that their emotional expression is aligned with their underlying feeling state.When engaging in surface acting, employees modify their emotional expression but without altering their feeling state, resulting in a mismatch between emotional experience and expression.More recently, automatic regulation has been proposed as a third emotional labour category (B. A. Scott et al., 2020;Martinez-Iñigo et al., 2007).Automatic regulation involves a naturally occurring regulation process resulting spontaneously in authentic and appropriate emotional expressions.However, while a number of studies have been conducted on automatic regulation (e.g., Hülsheger et al., 2015;Martinez-Iñigo et al., 2007), emotional labour research has largely focused on deep and surface acting.
Many studies have demonstrated that display rules predict usage of these emotional labour strategies, which are, in turn, related to key individual and organizational outcomes, including emotional exhaustion, burnout, job satisfaction, turnover intentions, and performance (for reviews and meta-analyses see Grandey & Gabriel, 2015;Grandey & Melloy, 2017;Hülsheger & Schewe, 2011;Kammeyer-Mueller et al., 2013).These studies converged on the conclusion that compared to surface acting, deep acting is overall a more adaptive strategy when considering employees' health and performance (Hülsheger & Schewe, 2011;Kammeyer-Mueller et al., 2013).
There is no doubt that studying these emotional labour strategies has furthered our understanding of how employees respond to display rules during (challenging) service encounters.However, the focus on deep and surface acting might be too narrow to do justice to the full scope of responses to display rules.While it is undoubtedly true that employees respond to display rules with emotional labour strategies of deep and surface acting, they may also respond with a range of other cognitive or behavioural strategies.For example, a customer service representative may transfer a difficult call to a colleague or calm down an angry customer.
While not necessarily directly targeting the emotions felt and expressed by the employee, such additional cognitive and behavioural strategies may be often used in response to organizational display rules.However, due to the current narrow focus on emotional labour strategies as responses to display rules, such potential additional strategies have not been systematically studied in employee-client interactions.It thus remains unknown what additional strategies are possibly at employees' disposal and to what extent they make use of them.
To date, no study has attempted to systematically map the full range of strategies employees use in response to organizational or occupational display rules.This is troublesome for several reasons.From a theoretical perspective, an incomplete understanding of employees' responses to display rules in organizations and occupations may lead to an underestimation of employees' strategy repertoire.These alternative strategies may be used as frequently as the emotional labour strategies of deep and surface acting.Similar to deep and surface acting, these cognitive and behavioural strategies may also function as mechanisms driving the effects of organizational display rules on employee performance and well-being.Uncovering and understanding these additional pathways is critical for a comprehensive understanding of strategies employees use in response to display rules.From an applied perspective, and considering the benefits and costs associated with following display rules, an understanding of the full range of strategies employees use when confronted with display rules is of vital importance.When recruiting employees, candidates could not only be selected based on their capacity to engage in traditional emotional labour strategies, but also on their ability to engage in other strategies that may even be more predictive of success or failure in jobs requiring interactions with customers.However, for this to happen, we first have to capture the full range of strategies.Similarly, without an overview of all possible strategies, major evaluation and training opportunities for employees may be overlooked.
The goal of the current study is therefore to develop a bottom-up taxonomy of strategies employees use in response to organizational and occupational emotional display rules.For this purpose, we asked employees to describe recent encounters during which they needed to comply with emotional display rules and we asked them to describe the strategies they used.To ensure that our results would not only pertain to a limited number of jobs in a particular country, we recruited employees from a variety of jobs involving interactions with the public in both the Netherlands and Turkey.This resulted in an extensive list of employee-generated strategy statements.We subsequently used this list for two related purposes.First, we examined to which extent the two dominant emotional labour strategies of deep acting and surface acting captured the list of strategy statements relying on expert-coders.The number of statements that could not be assigned to deep and surface acting reflects how much there is to gain by going beyond these two types of traditional emotional labour strategies.Second, we created a bottom-up taxonomy of strategies used in response to display rules.We examined the nature of these categories and to what extent they go beyond the two traditional emotional labour strategies of deep and surface acting.This allows us to examine to what degree the emotional labour domain (Grandey, 2000) captures all strategies employees report engaging in when facing display rules during interactions with customers, and whether certain strategies may have been overlooked.
The present study makes important contributions to the literature.First, we extend the current understanding of how employees respond to display rules in their interactions with clients.Recently, several scholars stressed the need to go beyond surface and deep acting, arguing that employees are likely to use a much broader variety of strategies in response to display rules (Grandey & Gabriel, 2015;Grandey & Melloy, 2017;Troth, Lawrence, Jordan, & Ashkanasy).To date, however, these calls have not been sufficiently addressed, because doing so "may require new methods of assessment" (Grandey & Melloy, 2017, p. 4).A bottom-up approach is such a new method of assessment as it may uncover new strategies that have, to date, been overlooked in top-down theoretical approaches.Our bottom-up approach allows us to test whether surface and deep acting also emerge organically when asking service workers how they deal with display rules in client interactions but also to identify additional strategies that employees may resort to.The bottom-up identification of novel strategies may stimulate subsequent theoretical work on the identified novel strategies.In doing so, our work lays the foundation for future research to explore previously overlooked ways of responding to display rule requirements and how this impacts individual and organizational outcomes.
In the next sections, we first provide a brief theoretical overview of display rules and research on the two emotional labour strategies that have most often been studied in this context: deep acting and surface acting.Next, we describe prior research and theory suggesting the existence of additional strategies employees resort to besides the two traditional emotional labour strategies.Finally, we elaborate on the advantages of a bottom-up approach to generate an extensive taxonomy of display rules strategies before articulating the hypotheses of the present study.
Two key strategies employees use to deal with display rules: the emotional labour strategies of deep and surface acting In Hochschild's (1983) original view, emotional labour jobs entail (a) regular contact with the public, (b) eliciting desired emotions in clients, and (c) monitoring and training employees' emotional displays (Gardner et al., 2009).These emotional displays are assumed to positively influence customers' emotions, service satisfaction and behaviours (Ashforth & Humphrey, 1993).
The dominant view in the literature argues that employees often act to comply with display rules, and their acting strategy is consequential for their well-being and performance (Grandey, 2000;Hochschild, 1983;Hülsheger & Schewe, 2011).According to Grandey (2000), such acting can be considered a specific type of emotion regulation to cope with display rules.She connected research on emotional labour to Gross' process model of emotion regulation (Gross, 1998) by relating deep and surface acting to antecedent-and response-focused strategies, respectively.The main difference between antecedent-and response-focused strategies is when the strategy affects the emotional response, either before the emotion is fully developed (antecedent-focused) or after the emotion is in full swing (response-focused, Gross, 1998).
When engaging in deep acting, employees align their internal feeling state with the required emotional expression by engaging in one of two antecedent-focused strategies: cognitive change or attentional deployment (Grandey, 2000).Cognitive change involves a reinterpretation of the situation.For example, when talking with an angry passenger, a flight attendant may try to understand the passenger's perspective (perspective-taking; Alabak et al., 2020), see the situation as a learning opportunity (positive reappraisal; Grandey, 2000) or accept the current situation being a part of her job (acceptance; Mikolajczak et al., 2009;Webb et al., 2012).Attentional deployment involves focusing attention on non-emotional aspects of the situation or shifting attention away from the situation altogether.For example, a flight attendant may think about an upcoming holiday to facilitate a positive emotional expression during an interaction with an angry passenger.In both cases (i.e., attentional deployment or cognitive change), the key principle is that employees attempt to change their own inner feelings such that they match the desired emotional expression.In contrast, when engaging in surface acting, employees only alter their emotional expression regardless of what they feel (Grandey, 2000).For example, flight attendants can simply hide their irritation or fake positive emotions without actually feeling them when dealing with an arrogant customer.
More recently, the categories of deep acting and surface acting have been complemented by the strategy of automatic regulation.This strategy pertains to the effortless expression of organizationally-desired emotions (Diefendorff & Gosserand, 2003;Martinez-Iñigo et al., 2007).Automatic regulation may be an especially adaptive strategy.For example, in positive display contexts (i.e., show positive emotions), automatic regulation has been shown to be more beneficial for employees' well-being than either deep or surface acting, and it was associated with less work withdrawal and more customer satisfaction (B. A. Scott et al., 2020;Hülsheger et al., 2015).
As such, emotional labour research mainly focused on deep acting, surface acting, and automatic regulation as the main strategies to deal with display rule requirements.Consistently, most research in the field of emotional labour has focused on the prevalence, antecedents, and consequences of these strategy classes.However, these emotional labour strategies may not provide a full picture of the strategies employees may resort to in response to display rules.This becomes particularly clear when considering adjacent literature in the fields of emotion regulation and coping.In the next sections, we elaborate on these adjacent streams of research suggesting alternative strategies to cope with display rules.

Going beyond emotional labour strategies when dealing with display rules
First, fundamental research on emotion regulation suggests that individuals may engage in a wide range of strategies in response to emotional display rules (for a review and metaanalyses, Gross, 2015;Webb et al., 2012).Among these strategies, situation selection and situation modification may be strategically adopted to guarantee optimal client service.Situation selection involves selecting the emotion-eliciting situations one is involved in, while situation modification involves directly modifying aspects of an emotion-eliciting situation in order to change one's emotions (Gross, 1998(Gross, , 2015)).Such strategies may have been largely overlooked in the emotional labour literature due to the long-held assumption that service employees lack the opportunity to rely on situation-oriented strategies due to the nature of their job (Grandey, 2000).However, growing evidence suggests that this assumption might not hold true.For example, it has been found that when teachers anticipate students to misbehave during a class activity, they may adapt the nature of the class activity to prevent the misbehaviour to take place (M.Chang & Taxer, 2020;Taxer & Gross, 2018).Similarly, Diefendorff et al. (2008) asked employees to rate the extent to which they applied a range of predefined emotion regulation strategies in their interactions with clients.It was found that employees did not only use deep and surface acting but also engaged in situation selection and modification.
Second, research within the field of coping at work revealed that employees often rely on coping strategies when facing job demands (Zhang et al., 2019).As organizational display rules may be considered a job demand (Zapf et al., 2021), the coping literature may hint at what additional strategies employees may use in response to display rules.Coping strategies have been broadly categorized as emotion-focused or problem-focused (Lazarus & Folkman, 1984;Zhang et al., 2019).Emotion-focused strategies are primarily aimed at managing one's emotions (Carver et al., 1989) whereas problem-focused strategies are primarily aimed at solving the problem at hand (e.g., developing solutions) but problemfocused strategies may ultimately also impact one's emotions (Carver et al., 1989;Lazarus & Folkman, 1984).Prior work on emotional labour has focused on a subset of emotion-focused strategies (i.e., emotional labour strategies) while largely overlooking problem-focused coping.However, first evidence shows that problem-focused coping strategies are also used to deal with display rules.For example, M. L. Chang (2013) found that teachers frequently employ problem-focused strategies in response to disruptive student behaviours, which helps them to regulate their emotions and adhere to display rules.
Third, rather than managing their own emotions and emotional expressions, employees may also try to directly change the customer's feelings.For example, when dealing with an angry customer, waiters may try to improve their customers' mood.In the fundamental emotion regulation literature, this is referred to as inter-personal emotion regulation, defined as achieving or maintaining emotional goals using the presence of others (Dixon-Gordon et al., 2015;English & Eldesouky, 2020;Zaki & Williams, 2013).In response to display rules, employees may thus use strategies that also target the emotions of their interaction partner rather than only their own.Such strategies may be adopted to primarily change the customer's feelings (interpersonal goal) or to manage their own emotions via changing the customer's feelings (intrapersonal goal).However, it should be noted that service settings are inherently an interpersonal context and a strict separation between intrapersonal and interpersonal strategies might not be feasible.Nevertheless, strategies may differ in the extent to which they primarily target the emotions of the employee or the customer.
Finally, employees may at times deviate from display rules.Such emotional deviance involves the genuine display of inner feelings disregarding organizational display rules and it might be a legitimate response in some situations (Holman et al., 2008;Rafaeli & Sutton, 1987;Von Gilsa et al., 2014).While deviance behaviour might at times be unacceptable for customers or organizations (e.g., spitting on the meal of arrogant customers before serving their food), there might also be less extreme and perhaps even beneficial deviance behaviours.For example, employees may encounter situations where emotional labour strategies do not work (e.g., extremely rude customers; Dahling, 2017;Von Gilsa et al., 2014) and in such situations deviance can be chosen to prevent continued inappropriate behaviour by letting customers know that their behaviour is offensive and unacceptable.Little is known, however, about the specific deviance strategies that employees may use in response to display rules.

A bottom-up taxonomy of strategies used in response to display rules
The overall goal of the present studies is to create a taxonomy of strategies employees use in response to display rule requirements.To do so, we used a bottom-up approach and directly asked service workers how they navigate display rules in their jobs.This approach offers many advantages.A bottom-up approach informs on what strategies organically emerge when using service workers as a source of information and allows examining the extent to which bottom-up derived strategies map on prior top-down derived strategies.Specifically, it allows examining to what degree the traditional deep actingsurface acting framework captures the full range of strategies employees themselves report adopting in service settings involving display rules.Moreover, a bottom-up approach helps uncovering strategies that have been overlooked in prior work on how to deal with display rule requirements.Finally, it should be noted that a bottom-up approach has proven its value in the field of fundamental emotion regulation (Livingstone & Srivastava, 2012;Niven et al., 2009;Parkinson & Totterdell, 1999) and coping research (Ayers et al., 1996;Walker et al., 1997).By following a procedure that is similar to the one adopted in these other bottom-up approaches, we will provide a systematic and comprehensive account of the many ways of dealing with display rules.

Research overview
The overall goal of the present studies is to create a complete bottom-up taxonomy of how employees respond to display rules.Using semi-structured interviews with Dutch and Turkish employees, we created a list of strategy statements (Study 1, phase 1).Next, expert-coders allocated the Dutch and Turkish statements to previously derived emotional labour categories (Study 1, phase 2), allowing us to quantify the degree to which these categories cover employees' selfreported strategies to deal with display rules.Finally, we asked non-expert participants to assess the similarity between strategy statements both for the Dutch and Turkish data (Study 2, phase 1) and ran a cluster analysis on these similarity ratings to obtain a bottom-up classification of strategies in dealing with display rules (Study 2, phase 2).
Even though the present studies were inevitably largely exploratory, a number of hypotheses could be formulated.First, we expect (Study 1) that a significant number of Dutch and Turkish strategy statements do not fall within any of the main emotional labour categories currently distinguished in the field (i.e., deep acting, surface acting, automatic regulation, and deviance).Second, we expect (Study 2) that both the Dutch and Turkish bottom-up classification will include deep acting and surface acting.Third, we expect (Study 2) that the Dutch and Turkish bottom-up taxonomy will include situation selection, situation modification, and strategies pertaining to problem solving.

Study 1
The overall aim of Study 1 was to collect bottom-up data on strategies employees use in response to display rules and to assess to what extent the current dominant taxonomy of emotional labour describes this set of strategies.

Phase 1: bottom-up harvesting of strategies used in response to display rules
The aim of Phase 1 of Study 1 is to collect a list of strategies for dealing with display rules.To ensure that our results do not only pertain to a limited number of jobs in one particular country we recruited employees from a variety of jobs in the Netherlands and in Turkey.Both countries have a service-based economy (Central Intelligence Agency, 2020) but compared to the Netherlands, Turkey is a more collectivistic and hierarchical society (Hofstede et al., 2010).

Participants
To take part in the study, participants needed to work in a job that required direct interaction with customers broadly defined (Grandey et al., 2005) for at least 20 hours per week.We relied on convenience sampling in recruiting participants.Participants were initially recruited with personal and professional contacts of the authors and the research assistant.Our participants connected us with their colleagues, which helped us reach out to more participants (the snowball technique).No participants have been excluded from our data.
Dutch sample.The Dutch sample included 77 Dutch employees (29 male; 48 female) from a variety of ages who voluntarily participated.More than half of the participants (55.8%) were employed in service jobs demanding welcoming emotional expressions, such as waitresses, receptionists, salespeople, flight attendants, hairdressers, or taxi drivers (Humphrey, Pollack, & Hawver, 2008).A major part of the other employees (29.9%) were employed in caring occupations where employees are expected to be sympathetic, such as teachers, nurses, or doctors (Humphrey et al., 2008).The remainder of the employees (14.3%) worked in social control jobs that require employees to stay calm and patient in emotionally intense situations, such as firefighters (C.Scott & Myers, 2005).
Turkish sample.The Turkish sample included 103 Turkish employees (54 male; 49 female) from a variety of ages who voluntarily participated in the study.The majority of participants (88.5%) were employed in service jobs and the others were employed in caring jobs (Humphrey et al., 2008).

Procedure
Semi-structured, face-to-face interviews on strategies were conducted in Dutch or Turkish.The exact time and place were planned at the convenience of the participant.We designed the interview protocol to prompt participants to think about strategies they use in response to display rules.The term "customer" was replaced by the term "student" when interviewing teachers, and by the term "patient" when interviewing nurses and doctors.Participants were first asked to describe their job and to explain whether they have customer interactions as a part of their job.Next, the interviewer explained that many jobs require certain emotional displays and provided some examples of common display rules (e.g., a waiter is expected to behave in a friendly way when interacting with customers).Subsequently, the interviewer asked participants to describe a recent challenging situation involving display rules: Can you think of any recent challenging situations where you were supposed to express certain emotions towards a customer/student/ patient without actually feeling those emotions, or situations where you were supposed to stay calm and neutral while actually feeling very emotional?"Then, participants were asked to report the strategies they adopted to display the required emotions: I understand that you were angry/upset . . .but you were expected to show a positive/negative/neutral expression.

I am interested in what you did or thought to achieve to display a positive/negative/ neutral expression during the interaction. Can you tell me what you did or thought?
Once the participant shared an episode, the interviewer asked them whether they would like to share another episode.If the participant answered yes, the interviewer asked again the same questions described above.If the participant answered no, the interview ended.All interviews were audio-recorded with the permission of the participants.The duration of the interviews ranged from 5 to 15 minutes.The study was approved by the local ethical review board.
Data processing.The Dutch interviews were transcribed and translated to English by the Dutch interviewer who speaks both Dutch and English fluently.Similarly, the Turkish interviews were transcribed and translated to English by the original Turkish interviewer who speaks both Turkish and English fluently.Subsequently, we extracted strategy statements from participants' responses.To this end, the first author reviewed the transcripts for specific words and phrases that indicated the use of a particular strategy, such as "I took a deep breath", or "I tried to understand what the customer wanted from me".Responses including multiple strategies were broken down into substatements.Next, to ensure a consistent format, each single statement was rephrased in the simple present tense.Finally, we looked for duplicates with close to identical wording (e.g., breathe deeply, take a deep breath) and only kept one of them in the list.This approach resulted in 147 distinct statements from the Dutch sample and 165 statements from the Turkish sample.
As an example, a waiter from our sample shared the following episode.
Interview excerpt: "There was a family.The father complained about our slow service.He was loud and angry.Other customers heard his complaints.He argued with me by saying 'you (referring to all waiters) are terrible at what you do, you are carelessly doing your job, you do not give full attention to customers'.We often come across these kinds of guys who think they are always right and powerful." The employee further said: "I said: if we continue to argue, I cannot serve other customers.As you see, they are waiting to be served, and we are wasting their time.We disrespect other customers.But the customer was obstinate.I mentioned A, he understood as B. I wanted to show him this discussion was really pointless.We could not find a middle ground.I gave up convincing him.I thought 'never mind, I am wasting my time'.Their meals were served by other waiters while we argued.We also have very nice customers who say everything was so delicious or thank you for your excellent service." We extracted the following statements: Show a customer that the current argumentation is pointless by saying ", I cannot serve other customers.As you see, they are waiting to be served, and we are wasting their time.We disrespect other customers." Think that "Never mind, I am wasting my time".

Phase 2: matching bottom-up data with theory-based categories
The aim of phase 2 was to match the bottom-up generated strategy statements in phase 1 with the dominant taxonomy of emotional labour strategies.The number of statements that could not be assigned to any category distinguished in the dominant taxonomy of emotional labour reflects the extent to which that taxonomy does not fully reflect employees' strategies in response to display rules.

Coders
Three persons coded the Dutch and Turkish statements.The coders were experts in the domain of emotional labour and coded the statements independently using the coding scheme described below.

Coding scheme
We created a coding scheme based on the emotional labour strategies described in the literature and their associated substrategies.As can be seen from Table 1, the first three categories pertained to deep acting and included cognitive change, attentional deployment and an overall unspecified deep acting category (Grandey, 2000;Gross, 1998).The next three categories pertained to surface acting and included faking, hiding and an overall unspecified surface acting category (Brotheridge & Lee, 2003;Grandey, 2000).The seventh category captured automatic regulation (Martinez-Iñigo et al., 2007, p. 284).The eighth category covered emotional deviance, showing the genuine display of inner feelings disregarding organizational display rules (Rafaeli & Sutton, 1987;Von Gilsa et al., 2014), to identify instances during which employees deviated from organizational display rules.Finally, the ninth category covered strategies that could not be captured by any of the categories mentioned above (1-8).
To test the scheme's utility, the coders initially coded ten randomly selected strategy statements and then discussed how distinguishable the statements were using the categories.This phase helped us ensure that our categories allowed us to code the data.Finally, the coders coded all statements in the Dutch and Turkish data set and achieved substantial agreement with Fleiss' Kappa (Fleiss, 1971) being .78for the Dutch data and .80 for the Turkish data.Statements that were initially classified differently by the coders were ultimately also assigned to a particular category following a discussion between the coders.

Results
Table 1 (left panel) contains the coding results for the Dutch data.Deep acting was used often by employees, with 27.9% of the statements being assigned to categories one to three.Interestingly, in the majority of those cases, employees relied on cognitive change rather than attentional deployment to change their inner feelings.Employees also engaged in other subtypes of deep acting strategies (11.6%) to regulate their emotions which were coded in category 3.For instance, participants reported adjusting their breathing and to slow down their speech to feel calm, or reminding themselves not to treat others in a way they do not want to be treated.
Surface acting was also well-represented in the data, with 19.7% of the statements being assigned to categories four to six.When engaging in surface acting, people relied more often on faking than on hiding.Interestingly, faking and hiding provided a complete description of the surface acting category.The category surface acting -other/unspecified remained empty.Automatic regulation (5.4%), and emotional deviance (7.5%) were also represented in the list of statements, but less frequently than deep acting and surface acting.
Critically, the largest category was category nine (39.5%), consisting of strategy statements that could not be assigned to any of the categories recognized in the current dominant theoretical framework on emotional labour.

Brief discussion of study 1
Taken together, these results suggest that the dominant taxonomy of emotional labour strategies (focusing on deep acting and surface acting), only partially captures the full range of strategies used in response to display rules.Deep acting and surface acting only partially accounted for the strategy statements in both countries (Netherlands: 47.6%; Turkey: 53.3%).This demonstrates the need to create a taxonomy of strategies used in response to display rules that goes beyond deep and surface acting (Grandey & Melloy, 2017).Although this is an important finding in and of itself, Study 1 did not provide any insights into the nature of possible additional strategies.To address this shortcoming, we conducted Study 2.

Study 2
The overall aim of this study was to expand on prior work on emotional labour by establishing a broader taxonomy capturing all bottom-up derived strategies employees use in response to display rules in our Dutch and Turkish sample.Rather than taking a top-down approach in identifying new strategies, we followed procedures used in adjacent fields of research (e.g., Niven et al., 2009) and adopted a bottom-up approach in establishing a taxonomy by asking laypeople to assess the similarity between strategy statements.Doing so allowed us to investigate which categories organically emerge.

Phase 1: rating the similarity between strategies
In the first phase of this study, we assessed the similarity between the strategy statements derived in Study 1.As the eventual taxonomy of strategies might be different in both countries, we assessed the similarity between the collected strategy statements separately for the Dutch and Turkish dataset.Specifically, naïve participants, who were not familiar with the emotional labour literature, assessed the similarity between the strategy statements derived in Study 1 using a card sorting task.

Participants
Participants were psychology students at a European university who had not yet taken any courses on the topic of emotional labour.They assessed the similarity between the strategy statements derived from the Dutch data in one session and from the Turkish data in another session.The two sessions were separated by one week in a counter-balanced order (i.e., they assessed either the similarity between the Dutch strategy statements or between the Turkish strategy statements first).
Participants were compensated with course credit.The study was approved by the local ethical review board.Similarity of Dutch strategy statements.A total of 122 participants assessed the similarity between the Dutch strategy statements.The final sample consisted of 95 participants (24 male; 71 female) after excluding 22 students who had prior knowledge on emotional labour (i.e., scored 3 or higher on the emotional labour knowledge question) and 5 students who failed an attention check (control card) during the similarity rating (card sorting) task.The average age of participants was 21.6 (SD = 3.00).More than half of the participants (58.9%) worked in a customer service job.
Similarity of Turkish strategy statements.A total of 125 participants assessed the similarity between the Turkish strategies.The final sample included 101 participants (19 male; 82 female) after excluding 22 students who had prior knowledge on emotional labour and 2 students who failed the attention check.The participants' average age was 21.50 (SD = 2.96).A high percentage of participants (63.4%) worked in a customer service job.

Procedure
Participants first provided demographic information (gender, age, nationality and work experience) and reported their knowledge of emotional labour (i.e., 1 = no prior knowledge at all, 4 = I already know a lot about emotional labour).Next, participants took part in a card-sorting task online (Cardsorting.net; Blanchard et al., 2017).
During the card sorting task, all Dutch or Turkish strategy statements were presented in random order.Participants were instructed to group strategy statements (cards) based on their similarity by adding similar strategies to the same pile, creating as many piles as they thought fit.Moreover, participants were asked to name each pile.To check whether participants performed the task carefully, we inserted a control card, which included text asking participants to sort it in a separate category.To make sure that participants understood the procedure, they were given a video demonstration of a card-sorting task, and were explained that each card represented a strategy statement.

Phase 2: clustering the similarity data
The aim of phase 2 of this study was to create a bottom-up taxonomy of strategies in response to display rules by running cluster analyses on the similarity data obtained in Phase 1 of Study 2.

Analytical procedure
In order to create a taxonomy of strategies, we employed hierarchical cluster analysis using the "cluster" package in R (Maechler et al., 2017).In hierarchical clustering, the goal of the algorithm is to form groups by sequentially merging similar clusters (Aldenderfer & Blashfield, 1984;Kassambara, 2017).It is a suitable approach to analyse the data in a bottom-up manner as it does not rely on a pre-defined number of clusters (Aldenderfer & Blashfield, 1984;Kassambara, 2017).Among the hierarchical cluster algorithms, we opted for Ward's (1963) hierarchical agglomerative cluster method using Euclidean distance because it outperforms other agglomerative cluster methods (e.g., complete linkage, average linkage) in generating homogeneous groups (Aldenderfer & Blashfield, 1984;Milligan & Sokol, 1980).Euclidean distance was used as a dissimilarity measure.
To determine the optimal number of clusters, we followed several procedures recommended in the clustering literature (Kassambara, 2017).Specifically, we combined the elbow method (i.e., choosing the cluster solution such that increasing the number of clusters does not markedly improve the total intra-cluster variation; Thorndike, 1953), average silhouette method (i.e., choosing the cluster solution with maximal average silhouette value, indicating that the strategies are well-matched to their clusters; Rousseeuw, 1987), and the gap statistic (i.e., choosing the cluster solution with maximal gap statistic, indicating that the clustering structure of the data is the strongest compared to a reference distribution with no clustering structure; Tibshirani et al., 2001).In addition to these statistics, we also took the interpretability of the cluster solution into account.
In addition to correctly estimating the appropriate number of clusters, it is important to evaluate and then improve the quality of the final cluster solution (Kaufman & Rousseeuw, 2009;Rousseeuw, 1987).An important challenge of any hierarchical cluster algorithm is that poor cluster assignments are not improved throughout the analysis because objects are grouped sequentially and irrevocably (Everitt et al., 2011;Ketchen & Shook, 1996).As in previous research (e.g., Solymosi, 2019), we followed the recommended procedure to check and improve poor placements in clusters, and in turn reach the optimum cluster solution in the data.Specifically, we improved cluster solutions based on individual silhouette values of strategy statements (i.e., re-assigning strategy statements with poor silhouette index to the closest neighbouring cluster).
For potential cluster solutions, we obtained silhouette values for the current and closest neighbouring clusters of each strategy (Rousseeuw, 1987).Silhouette values allowed us to assess to what extent each strategy fits the cluster to which it was assigned.The silhouette value stands between −1 and 1 (Rousseeuw, 1987).Values closer to 1 indicate a good fit while values closer to −1 indicate misfit (Rousseeuw, 1987).Following Kaufman and Rousseeuw (2009), we reassigned poorly fitting strategies (i.e., strategies with negative silhouette values) to their closest neighbouring cluster if they were a better match for the final cluster solutions both statistically and theoretically.It should be noted that we reassigned poorly fitting strategies only if they theoretically fit the newly assigned cluster.

Number of clusters in Dutch data
The elbow method suggested seven clusters, while the average silhouette method and the gap statistic both suggested a ninecluster solution.The nine-cluster solution was therefore retained.However, cluster 2 had a negative average silhouette index, reflecting the presence of poorly assigned strategies.To further optimize the nine-cluster solution, as suggested by Kaufman and Rousseeuw (2009), we reassigned poorly fitting strategy statements with a negative silhouette index (n = 14) to their closest neighbouring cluster, but only when this was theoretically meaningful.After these reassignments, the average silhouette index of cluster 2 became positive, increasing from −0.0061 to 0.006.

Number of clusters in Turkish data
The elbow method proposed a five-cluster solution, the average silhouette method proposed that any solution between five and ten clusters could be retained, and the gap statistic method proposed ten as the ideal number of clusters.We therefore next carefully examined cluster solutions with 5 to 10 clusters.The ten-cluster solution was the most interpretable and was therefore retained.As all clusters had a positive average silhouette index, we did not reassign any strategies.

Interpreting the cluster solutions
The full cluster solution containing all individual strategies can be found in supplementary materials.We first discuss clusters that directly match the categories of the current dominant taxonomy of emotional labour strategies (see top of Table 2): deep acting (attentional deployment and cognitive change) and surface acting (faking and hiding).Subsequently, we discuss the clusters containing strategies that have been overlooked by the current dominant taxonomy of emotional labour strategies 1 (see bottom of Table 2).It should be noted that our cluster labels reflect the majority of the cluster member strategy statements.As non-experts tend to create groups with high within-group variability (Fincher & Tenenberg, 2005), not each single individual cluster member may be equally well represented by the cluster label.

Clusters consisting of strategies recognized in the dominant taxonomy of emotional labour
A number of the clusters of our bottom-up taxonomy consisted of either surface acting or deep acting.The existence of these clusters provides support for the current dominant taxonomy of emotional labour strategies capturing a significant part of the strategies used in response to display rules.We discuss these clusters first.Surface acting by hiding or faking.Both the Dutch (cluster one) and Turkish (cluster eight) cluster solution contained a surface acting by hiding or faking category.For example, statements like try not to show my emotions (Dutch data) or do not show that I am annoyed (Turkish data) are typical examples of hiding, while statements like fake my emotions and smile (Dutch data) or wear my mask and try to be friendly (Turkish data) are typical examples of faking.If this category indeed reflects surface acting, we might expect that it would mainly consist of strategy statements coded as hiding or faking in Study 1.This was the case, with the large majority of strategy statements in this cluster (84.2% in Dutch data and 81.8% in Turkish data) being coded as hiding or faking in Study 1 by expert raters.
Deep acting -cognitive change by perspective taking.Both the Dutch (cluster seven) and Turkish (cluster four) cluster solution contained a cognitive change category consisting mainly of perspective taking statements.For example, statements like put myself in customer's shoes or stay friendly and think that this is the first time, and for a student it is all scary (Dutch data) and try to empathize with the customer or put myself in customers' shoes (Turkish data).Consistently, 84.6% of Dutch strategies and 86.9% of Turkish strategy statements in these clusters were coded as "cognitive change" by the experts in Study 1.
Deep acting -cognitive change by positive reappraisal.The Turkish cluster solution contained a cognitive change category consisting mainly of positive reappraisal statements (cluster five).For example, statements like think that I am not the only one who is responsible for the situation or think that I cannot make every customer satisfied.Consistently, a high number of statements in this cluster (87.5%) were coded as "cognitive change" by the experts in Study 1.
Deep acting -cognitive change by acceptance.The Turkish cluster solution contained a cognitive change category consisting mainly of acceptance statements (cluster six).For example, accept the situation as it is or see customer problems as part of my job.Consistently, a high number of strategies in this cluster (91.7%) were coded as "cognitive change" by the experts in Study 1.

Clusters consisting of strategies that are not part of the dominant taxonomy of emotional labour
The cluster analysis also revealed a number of categories that are not captured by the dominant taxonomy of emotional labour.An overview of these categories can be found in Table 2.
Changing customers' cognitions.Both the Dutch (cluster six) and the Turkish (cluster one) cluster solution contained a category of strategy statements that described attempts to induce cognitive change in the customers.As such, unlike deep acting-cognitive change, employees primarily target their customers' rather than their own appraisals of the situation.Examples include try to calmly explain both sides of a problem or explain to a student how his behaviour affects others (Dutch data) and ask the customer to empathize with me by asking questions "If I requested the same thing how would you react?; How would you feel if I made this request?"(Turkish).Employees thus stimulate perspective taking or reappraisal in their customers, which may indirectly calm down the customer and may make it easier for employees to display organizationally expected emotions.As this strategy does not belong to the current dominant taxonomy of emotional labour, one would expect this strategy to be coded as "other" by experts in Study 1.This was the case, with 89.2% Dutch and 95.7% Turkish strategy statements belonging to this category being assigned to the "other" category by experts in Study 1.
Changing customers' affect.Both the Dutch (cluster nine) and the Turkish (cluster three) cluster solution contained a category of statements describing attempts to change the feelings of the customers.Similar to the changing customers' cognitions strategies, these strategies primarily target the customer, but rather than changing the customers' perception of the situation, these strategies aim to target the customer's emotions directly.Examples include make jokes to keep things positive or try to make a customer feel loved and welcomed because she is in a pitiable situation (Dutch data), and offer a free meal, dessert or drink to decrease customer's negative emotions or focus on saving my relationship with the customer by making the customer feel he is important for us (Turkish data).These strategies may restore customers' positive emotions, and may thereby help employees to subsequently display organizationally expected emotions.As this strategy does not belong to the current dominant taxonomy of emotional labour, one would expect this strategy to be coded as "other" in Study 1.This was the case, with 100% Dutch and 100% Turkish statements belonging to this cluster being assigned to the "other" category by experts in Study 1.
Solution-oriented strategies.Both the Dutch (cluster five) and the Turkish (cluster two) cluster solution contained a category of strategy statements aimed at solving the problem at hand.Examples include try to come up with a solution as fast as possible or think about how to solve a customer's problem (Dutch), and propose a quick compromise to solve the problem (Turkish) or focus on possible solutions (Turkish).These strategies may diminish the negative impact of the situation on both the employee and the customer, making it easier for employees to display appropriate emotions during the interaction.This strategy is not part of the dominant taxonomy of emotional labour strategies and consistently 100% (Dutch data) and 100% (Turkish data) of statements belonging to the solution-oriented cluster were coded as "other" by the experts in Study 1.
Waiting strategies.Both the Dutch (cluster four) and the Turkish (cluster nine) cluster solution contained a category of statements that consisted of a wait-and-see approach.In contrast to solution-oriented strategies, strategies included in this cluster described behaviour where employees refrained from actively approaching the situation, but instead waited until it was all over.Examples include count till 10 or just let it go (Dutch data), and count to 20 or say a very short prayer in my head (Turkish data).This strategy may mainly contribute to not complicating the situation and waiting until it is all over or for a solution to present itself.As this approach is not explicitly recognized in the current dominant taxonomy of emotional labour strategies, one might expect that most of these statements were coded as "other" in Study 1.This was the case (57.9% in Dutch data and 50% in Turkish data).
Avoidance.The Turkish cluster solution contained a cluster (cluster ten) consisting of behaviours aimed at avoiding directly dealing with the situation.Examples include divert my attention away from the situation or walk away from the interaction and approach other customers.Attentional deployment is one way to disengage from the situation and a subset of the strategies belonging to this cluster (23.07%) were categorized as such by the experts in Study 1.However, most strategies were coded as "other" (46.15%) by the experts in Study 1. Several of these other strategies pertain to avoidance such as stay away from the situation or stay away from the situation and take 10 minutes break.
Deviance in bad faith.The Turkish cluster solution contained a cluster (cluster seven) consisting of behaviours that deviated from emotional display rules.For example, provide my service scornfully or shake a customer's coke just before serving it to take revenge.Consistently, 100% of these Turkish strategy statements were categorized in the deviance category by the experts in Study 1.
Deviance in good faith.The Dutch cluster solution contained a cluster (cluster three) consisting mainly of behaviours that deviated from emotional display rules but that can also be seen as an act of self-protection, assertiveness, or boundary management.For example, think that I do still have some selfrespect that I won't allow this misbehaviour or tell the patient that now you are crossing the line; we are not continuing with this.While strictly speaking these statements describe emotional deviance, some of them might be better captured by a term like boundary management.The majority of strategy statements in this group were categorized in the deviance (66.7%) category by the experts in Study 1.

Discussion
Display rules, prescribed requirements of how emotions should be displayed in client interactions, are crucial for today's service-oriented economy.Research on emotional labour focused on how display rules result in employees' emotion regulation attempts, typically in the form of surface and deep acting (Grandey & Gabriel, 2015;Grandey & Melloy, 2017).In the present paper, we argue that although the two emotional labour strategies of surface and deep acting are undoubtedly two important ways in which employees respond to display rule requirements, employees may use a much wider range of cognitive or behavioural strategies.The aim of the present studies was to provide a comprehensive account of the variety of responses to display rules employees engage in when interacting with customers.For this purpose, we adopted a bottomup approach.This ensured that we comprehensively captured the full range of strategies employees actually use in real life interactions with customers without restricting strategies to predefined theoretically derived categories.Specifically, we first gathered a wide range of strategy statements by conducting semi-structured interviews with employees.Next, we examined to what extent these strategies are captured by the dominant emotional labour strategies taxonomy.Finally, we derived a broader bottom-up taxonomy by asking lay people to assess the similarity between these strategy statements and running a cluster analysis on these similarity ratings.Given the global growth in the service sector (Kim & Wood, 2020;The World Bank, 2020), we did so across two culturally diverse settings (the Netherlands and Turkey) to maximally cover strategies employees adopt in response to display rules.
We found that employees use a broad array of strategies in response to display rules.These include the two established emotional labour strategies of deep acting and surface acting, which confirms the importance of these two strategies that are often studied in the emotional labour domain (Grandey, 2000).However, we also found that employees' strategies are broader and more nuanced than those strategies covered by the emotional labour literature.Specifically, we found that employees do not only make use of deep and surface acting but also adopt solution-oriented strategies, waiting strategies, avoidance strategies, and strategies that change customers' cognition or affect.Below, we discuss the key theoretical and practical contributions of our findings.

Theoretical contributions
Deep acting and surface acting are two key strategies used in response to display rules.Our findings confirmed that employees make extensive use of deep and surface acting to meet display rules, which confirms the importance of these two emotional labour strategies.Notably, while we did not find additional emotional labour strategies besides surface and deep acting, we identified a number of sub-strategies that fall under the umbrella of deep acting.Specifically, while we found a single surface acting cluster (subsuming hiding and faking emotions) in both countries, the bottom-up taxonomy did not contain a unitary deep acting cluster.Instead, three separate clusters containing different sub-strategies of deep acting emerged: perspective taking, positive reappraisal, and acceptance.This paints a more fine-grained picture of deep acting, and confirms earlier suggestions that deep acting is not a unitary construct, but rather consists of distinct strategies with the common goal of changing emotions felt by employees (Hülsheger & Schewe, 2011).This finding also aligns with a recent emotional labour study, which took a top-down approach and showed that cognitive reappraisal and perspective taking items loaded onto separate factors and differentially related to emotional labour outcomes (Alabak et al., 2020).Moreover, aside from perspective taking and positive reappraisal, acceptance was found as a third deep acting strategy (Mikolajczak et al., 2009) in our bottom-up taxonomy.Although it is often studied in the fundamental emotion regulation literature (Webb et al., 2012), acceptance has been largely overlooked as an emotional labour strategy.Yet, it is an important strategy to consider, given that it may not only benefit employee performance, but also well-being.This is suggested by a longitudinal study of Bond and Bunce (2003), which showed that customer service employees' habitual use of acceptance positively predicts their well-being and performance measured one year later.
Interestingly, no separate attentional deployment cluster emerged in our bottom-up taxonomy.Employees did report engaging in attentional deployment (e.g., divert my attention away from the situation) but these statements clustered together with overt behavioural avoidance (e.g., stay away from the situation and take a 10-minute break) to form an overall avoidance cluster.This is theoretically comprehensible as attentional deployment pertains to a form of internal avoidance and pooling all avoidance-related strategies together aligns with the higher order categorization of emotion regulation strategies in the fundamental emotion regulation literature (Naragon-Gainey et al., 2017).These strategies are typically subsumed under the broader category of disengagement (Naragon-Gainey et al., 2017).
Taken together, deep acting and surface acting were found to be two key classes of strategies to deal with display rules.Moreover, our bottom-up taxonomy aligns with recent recommendations to focus on specific rather than broad emotional labour strategies (e.g., Alabak et al., 2020;Grandey & Gabriel, 2015).Rather than using omnibus deep acting scales that capture employees' attempts or efforts to align felt and required emotions, it may be more fruitful to measure the actual use of perspective taking, positive reappraisal, and acceptance (Alabak et al., 2020;Grandey & Gabriel, 2015).
Beyond emotional labour strategies: Deep acting and surface acting are not the only two strategies to deal with display rules.We found that in an attempt to deal with organizational display rules, employees use additional strategies that do not qualify as emotional labour strategies as they do not target the management of employees' own emotions or emotional expressions directly.However, they are used in settings involving organizational display rules.
First, although emotional labour researchers have speculated that there may be little opportunity for situation selection and modification in service interactions (Grandey, 2000), employees did report engaging in such strategies in the present study.Specifically, using solution-oriented strategies, employees change the situation by solving the customer's problem.Such problem-solving strategies not only ensure the attainment of organizational expectations but may also have a regulatory function for the employee (Gross, 2015) and the customer (Little et al., 2013) indirectly as solving the problem at hand might increase the mood of the employee and customer.It is also notable that the occurrence of solution-oriented strategies in our bottom-up taxonomy is consistent with prior work suggesting that employees do in fact engage in situation modification and problem solving in a service context (Diefendorff et al., 2008).However, while these prior findings were obtained using a top-down approach, the present findings were obtained using a bottom-up approach, which further validates the importance of integrating situation modification in research on strategies to cope with display rules.
Second, our bottom-up taxonomy consisted of a cluster of avoidance strategies.In contrast to solution-oriented strategies, avoidance strategies aim to disengage from the customer interaction.For example, employees reported walking away from interactions, staying away from difficult situations or taking a break.This may elicit a feeling of relief (Aldao et al., 2010) but employees may miss out on an opportunity to improve the current situation for both parties (Endler & Parker, 1990;Van Bockstaele et al., 2019).Interestingly, as mentioned above, the avoidance cluster also entailed attentional deployment statements.This makes sense as distraction allows people to disengage from emotionally demanding situations as well.However, while attentional deployment has been often studied as a strategy in a display rule context (e.g., Alabak et al., 2020), there is hardly any evidence on behavioural avoidance in such contexts.
Third, employees also reported engaging in waiting strategies.The waiting cluster included passive attempts holding oneself back from the situation.While this approach may not seem particularly effective, waiting strategies may be motivated by different reasons.For example, a wait-and-see strategy may be a smart choice to suppress immediate undesirable behavioural responses.Similarly, waiting strategies may be effective to allow protecting or recharging one's resources when dealing with unpredictable clients (Carver et al., 1989).
These strategies indeed resemble so-called passive avoidance coping (e.g., I wait and see what will happen"; Andreassen et al., 2012, p. 284) that has been studied outside the context of display rules.
Fourth, our taxonomy contained two categories of strategies where the primary target is the customer rather than the employee.In contexts involving display rules, employees reported sometimes engaging in behaviours aimed at changing customers' cognitions or affect.In the latter case, the strategies are directly aimed at changing how the customer feels while in the former case strategies are targeted at changing how the customer thinks about the situation which may indirectly also change how the customer feels in the situation.These strategies resemble the cognitive engagement (i.e., influencing target's thoughts) and affective engagement (i.e., influencing target's emotions) strategies that Niven et al. (2009) described in their classification of interpersonal emotion regulation strategies.Given that display rules can be perceived as inter-personal job demands (e.g., promoting customers' happiness; Diefendorff & Richard, 2003;Wilk & Moynihan, 2005), it is not surprising that our bottom-up taxonomy contained strategies where the customer's emotions are the primary target.However, these customer oriented strategies may indirectly also impact how employees feel as increases in the customers' mood may be followed by corresponding increases in the employee.Therefore, they may not only help employees meet display rules, but also regulate their own emotions.
Two types of deviance: deviance in good faith vs. deviance in bad faith.Although participants were asked to report on strategies they had adopted to comply with display rule requirements, they also spontaneously reported deviance strategies.Interestingly, our bottom-up taxonomy provided a novel perspective on emotional deviance as two different types of deviance strategies were found.Drawing a parallel to Rafaeli and Sutton's (1987) distinction between surface acting as faking in good vs. bad faith, we found a deviance in good faith and a deviance in bad faith cluster.Emotional deviance has traditionally been conceptualized as deviance in bad faith: emotional displays deviating from display rules as a result of poor regulation skills or negligence and bad will (Mikolajczak et al., 2009).Statements in the Turkish cluster solution provide many examples of this type of deviance strategies (e.g., shaking a customer's coke before serving it).Deviance in bad faith refers to strategies that are outright deviant and reflect an inability or unwillingness to comply with organizational display rules (Rafaeli & Sutton, 1987).
In contrast, statements reported in the Dutch deviance cluster reflected deviance in good faith.Deviance in good faith strategies differ from deviance in bad faith in their interpersonal and instrumental nature.When engaging in deviance in good faith, employees seem to engage in boundary management, which allows to protect oneself and in some cases may even lead to a de-escalation of a difficult situation.From an instrumental standpoint, these self-protective responses may be employed to prevent further resource loss (Dahling, 2017).While these strategies may seem costly within an interaction when focusing on immediate consequences only (e.g., drop in immediate customer satisfaction), in the long-run, deviance in good faith may help employees to regulate their own emotions, be prepared for the next encounter and to ultimately meet emotional labour demands in the benefit of the organization.
While previous research has indicated differences in display rules' perception and regulation across cultures (e.g., Matsumoto et al., 2008;Matsumoto, 1990;McDuff et al., 2017), other studies have shown that these differences may be minimized in a service context (Grandey et al., 2010).Yet, our findings on deviance suggest a notable difference between two culturally different settings.In particular, deviance in bad faith strategies may be more culturally congruent in a collectivistic Turkish society because they are not easily noticeable by customers (e.g., secretly sabotaging a service encounter) and harmony in customer relationships is protected (Hofstede et al., 2010).In contrast, the use of deviance in good faith strategies may be more common in individualistic countries (e.g., The Netherlands), where employees may feel more comfortable to confront customers in unpleasant interactions (Hofstede et al., 2010).

Practical implications
The present study highlights the wide variety of strategies employees may use when dealing with display rule requirements.This has several important practical implications given that display rules are closely monitored and trained by organizations.First, our findings suggest that there is a need to extend training programs aimed at enhancing employees' communication skills during difficult encounters with customers.Employees may not only benefit from training in deep acting but also training in other strategies (e.g., solutionoriented strategies) may help employees in client interactions.Future research on the antecedents and consequences of these other strategies is needed to corroborate these claims.
Second, organizations may benefit from realizing that employees may occasionally engage in deviant behaviours to protect their own boundaries.While deviance in good faith may conflict with the motto of the customer being king, organizations should consider the trade-offs between the potential rewards (e.g., relieving tension, preserving the employees' self-esteem and well-being) and costs (e.g., possible reduction in customer satisfaction) of deviance in good faith behaviours for employees and organizations.One organizational strategy could consist of protecting employees against uncivil treatment with zero-tolerance towards abusive customers (Van Jaarsveld et al., 2010), or assuring employees that their well-being is more important than following display rules.These protective actions may in the long term decrease deviance that will harm organizations' reputation, as employees would be less likely to be involved in escalated unpleasant encounters in the first place.

Limitations and directions for future research
As with any study, the present study has a number of limitations.First, as recommended by Wilhelmy and Koehler (2021), we tried to ensure the transparency and consistency in reporting our qualitative data.However, memory bias may have impacted our data.Nevertheless, it should be noted that none of our participants experienced any difficulty recalling customer encounters and strategies to deal with display rules.Second, the majority of the interviewed employees had to display positive emotions to meet their job requirements.Possibly, employees who are required to display negative emotions (e.g., bill collectors) may use additional strategies not included in the present taxonomy.Future research is needed to complement and expand the present taxonomy by including this group of employees.Third, during interviews, we were only able to capture strategies that are consciously accessible.As a result, the present study could not provide much information about automatic regulation strategies used in response to display rules.Future research should consider alternative assessment methods to explore more implicit forms of strategies.Fourth, our research was descriptive, and future research is needed to explore the predictors and outcomes of the strategies identified in the present taxonomy.

Conclusion
Which strategies do employees use in response to display rules at work?The present study confirms that deep acting and surface acting are two key strategies that employees engage in.However, almost 50% of the distinctive strategies derived from our bottom-up study involved other strategies: changing customers' cognitions, changing customers' affect, solutionoriented strategies, waiting strategies, avoidance strategies, and deviance in good and bad faith.

Table 1 .
Emotional labour strategies across the Dutch sample and the Turkish sample.

Table 2 .
Bottom-up derived strategy categories.
Deviance in good faith Protecting one's boundaries with self-protective and assertive behaviours Tell a patient "Now, you are crossing the line.We are not continuing with this."Deviance in bad faith Deviating from display rules with an intention of bad willServe without smiling and perform my job at a minimum level.