language as a lightning rod: power contests, emotion ... · geographically dispersed team)...

Language as a lightning rod: Power contests,emotion regulation, and subgroup dynamics inglobal teams

Pamela J Hinds1,Tsedal B Neeley2 andCatherine Durnell Cramton3

1Center for Work, Technology and Organization,Department of Management Science andEngineering, Stanford University, Stanford, USA;2Harvard Business School, Boston, USA; 3Schoolof Management, George Mason University,Fairfax, USA

Correspondence:Pamela Hinds, Center for Work, Technologyand Organization, Department ofManagement Science and Engineering,Stanford University, Stanford, CA 94305-4026, USA.Tel: +1 650 723 3843;Fax: +1 650 723 2826

Received: 19 December 2012Revised: 10 October 2013Accepted: 16 October 2013Online publication date: 12 December 2013

AbstractBased on an ethnographic study comprising interviews with and observationsof 96 globally distributed members of six software development teams, wepropose a model that captures how asymmetries in language fluency contributeto an us vs them dynamic common in global teams. Faultlines, formedalong the dimensions of asymmetries in lingua franca fluency, location, andnationality of teammembers, were associated with subgrouping in some but notall of the teams. Our findings suggest that divisive subgroup dynamics occurredonly in teams that also suffered from power contests, suggesting that powercontests activate otherwise dormant faultlines. Our findings extend theory onsubgroup dynamics in global teams by adding language as a potential faultlinedimension, showing how power struggles activated faultlines and were, inturn, reinforced by them and documenting the emotion-regulation processestriggered by subgrouping and enacted through language-related choices andbehaviors.Journal of International Business Studies (2014) 45, 536–561. doi:10.1057/jibs.2013.62

Keywords: ethnography; qualitative research; language (language design, silent lan-guage, translation); teams and teamwork; geographic distance; intercultural workrelationship

INTRODUCTIONMultinationals are increasingly mandating English as the linguafranca, or common language, to facilitate collaboration acrossnational and linguistic boundaries (e.g., Bono & Vey, 2005; Feely &Harzing, 2003). Despite the central role of a lingua franca in globalwork, a gap remains in our understanding of how language affectswork and workers in international organizations (Harzing & Feely,2008). The burgeoning empirical studies examining the mandateof English in the global workplace shed light on the politics oflanguage choice (e.g., Marschan-Piekkari, Welch, & Welch, 2005;Vaara, Tienari, Piekkari, & Säntti, 2005), language fluency as asource of power and status (e.g., Kingston, 1996; Neeley, 2013), theinterplay between language and cultural diversity (Henderson,2005), and the role of language in social identity and categorization(e.g., Harzing & Feely, 2008). This body of research has identifieddisparities in co-worker lingua franca fluency and has highlightedtheir importance (Fixman, 1989; Knapp, 2003). At the same time,

Journal of International Business Studies (2014) 45, 536–561© 2014 Academy of International Business All rights reserved 0047-2506

www.jibs.net

http://dx.doi.org/10.1057/jibs.2013.62

http://www.jibs.net

the last decade has produced a significant body ofresearch on the dynamics of teams with memberslocated in multiple countries who must collabo-rate across cultural and language divides. A near-universal claim in this research is that globallydistributed teams tend to fracture into subgroupsthat are characterized by an us vs them mentality(e.g., Cramton & Hinds, 2005; Metiu, 2006;Polzer, Crisp, Jarvenpaa, & Kim, 2006). Researchsuggests that team dynamics are brittle in globalteams, in part because of geographic distance,missing information, power imbalances, and thelocal identifications that strengthen allegiancesat a given site. Despite the applicability of thesetwo parallel streams to global teams, they haveyet to be integrated, and little is known abouthow or whether asymmetries in language fluencycontribute to subgrouping or, more generally,how subgroup dynamics play out in globalteams.Over a decade ago, Lau and Murnighan (1998)

introduced the idea of faultlines that reflect align-ment across demographic characteristics, thus creat-ing a chasm between subgroups with the potential toerupt into conflict. Scholars quickly recognized theapplicability of the faultline model to global teams,suggesting that subgroups, aligned along demo-graphic faultlines and by location, are a naturalconsequence of global work, and interfere with astrong team identity (Cramton & Hinds, 2005;Polzer et al., 2006). O’Leary and Mortensen (2010),for example, report that geographically based sub-groups (e.g., subgroups aligned with location in ageographically dispersed team) experience weakerteam identification and more conflict. Subgroupscan also experience destructive power dynamics,such as when status differentials deepen the dividebetween team members across different locations(e.g., Metiu, 2006). Theory suggests that social cate-gorization explains this phenomenon because globalteam members identify more strongly with thosewho are collocated and, by default, are more similarto themselves (O’Leary & Mortensen, 2010; Polzeret al., 2006). Research on global teams documentsthe challenges of working across cultures and geo-graphic barriers, and highlights the role of potentialfissures between distant locations that can lead tounhealthy subgroup dynamics. Work to date, how-ever, remains largely silent on how these dynamicsunfold, on how the particular attributes of globallydistributed teams contribute to these potential fis-sures and their activation, and on the role thatlanguage plays in these dynamics.

We examine the role of language asymmetries insubgroup dynamics in globally distributed teamsthrough a qualitative study including interviewswith and observations of six software developmentteams in one company in the high-technology indus-try (GlobalTech, a pseudonym). Language asymme-tries, for the purposes of our study, refer to differinglevels of language competence in the lingua franca(English) across team members. We present a theore-tical model that captures the role of language asym-metries and documents the pivotal role of power insubgroup dynamics. In doing so, we provide insightinto why and when global teams tend to fracture. Ourtheoretical contribution extends faultline theory toarticulate how subgroups become salient and arereinforced in global teams and, particularly, howlanguage can act as a lightning rod for these dynamicsby attracting and conducting the tensions and then,in turn, reinforcing the emotions that further fortifythe dynamic. We observed strong location-based sub-groups in some but not all teams. We extend theoryby identifying power dynamics as an activator offaultlines, and by describing how power dynamicstriggered and reinforced language-based subgroupingbehavior.

LANGUAGE AND SUBGROUP DYNAMICS INGLOBAL TEAMS

Recent research has begun to explore languagedynamics among members of multinational teams.Henderson (2005), for example, in a review ofresearch on language diversity in international man-agement teams, reported that language diversityinterferes with trust and team building. Harzing andFeely (2008) similarly concluded in their study ofGerman and Japanese MNCs that the failure to com-municate effectively promoted faulty attributions,conflict, and distortion in management teams thatdid not share the same mother tongue. Neeley (2013)further documented status loss experienced by nativeFrench speakers in the presence of a newly mandatedEnglish lingua franca. In sum, early studies suggestthat uneven language fluency interferes with infor-mation sharing and the building of rapport on globalteams. Evidence also suggests that language is intri-cately tied with self-identity (e.g., Harzing & Feely,2008; also Giles & Johnson, 1981) and that socialidentities are forged and reinforced through interac-tion (see Bucholtz & Hall, 2004, 2005). Thus evidencesuggests that language asymmetries, exacerbatedthrough day-to-day interactions, could promote thesubgrouping so commonly observed in global teams.

Language as a lightning rod Pamela J Hinds et al537

Journal of International Business Studies

In a related line of inquiry, Lau and Murnighan(1998) introduced the idea of group faultlinesto reconcile inconsistent findings in research ongroup diversity and provide insight into subgroupdynamics. Faultlines are dividing lines formed bythe alignment of demographic characteristics acrossgroup members (Lau &Murnighan, 1998) and repre-sent a potential fracture that can lie dormant untilactivated. The faultline model has since been appliedto advance our understanding of subgroup dyna-mics in global teams. One of the earliest studiesexamining faultlines in global teams confirmedthe hypothesis that highly heterogeneous teamswere more successful at creating a shared identitythan were the teams that had multiple membersfrom each country and could splinter along countrylines (Earley & Mosakowski, 2000). Li and Hambrick(2005) later examined faultlines in Sino-Westernjoint venture management teams and reported thatdemographic faultlines (alignment of age, tenure,gender, and ethnicity) engendered emotional con-flict, which resulted in behavioral disintegrationacross subgroups. More recently, Polzer et al. (2006)studied teams of graduate business students whowere globally distributed. Subgrouping was mostpronounced when there were two location-basedsubgroups, as compared with teams dispersed acrossmore locations and when location-based subgroupswere more homogeneous in nationality. This nas-cent research affirms that, when aligned with otherdefining features, geographic location can generatea faultline between physically separated subgroupsin a team and can lead, as suggested by Lau andMurnighan, to subgrouping and, in turn, behavioraldisintegration and conflict in global teams.One gap in the extant research on subgroup

dynamics in global teams is a more precise specifica-tion of the characteristics likely to align by location.So far, scholars have argued that team members at agiven location will be more similar demographically:thus geographic dispersion will, by default, be asso-ciated with characteristics that align with location,such as nationality (e.g., Cramton & Hinds, 2005;Polzer et al., 2006). Unfortunately, most empiricalresearch to date has measured only location, and hasnot examined directly alignment between locationand other meaningful characteristics. As a result, wehave little understanding of what, if any, demo-graphic or other characteristics are influential whenaligned with geographic location to create faultlinesin global teams. One exception to this is Polzer et al.’s(2006) examination of homogeneity of nationalityby location, which found that alignment between

location and nationality exacerbated subgrouping.Language, we speculate, is also likely to align withglobal dispersion and contribute to faultlines, espe-cially in identity-based subgroups (Carton &Cummings, 2012), although the role of language hasso far been overlooked in examinations of subgroupdynamics on global teams. As organizations mandateEnglish as the lingua franca, however, there arecompelling reasons to anticipate that language asym-metries will contribute to, and perhaps be a powerfulsource of, geographic subgrouping.Although we set out in this study to shed light on

subgroup dynamics in global teams, especially therole of language, our inductive approach yieldedseveral unanticipated discoveries. We were firststruck by how emotional team members were whentalking about language asymmetries, the strategiesused for dealing with these asymmetries, and thediscomfort the strategies frequently engendered.Language asymmetries seemed to channel powerfulsubgroup dynamics, acting as a lightning rod or aconduit for intense emotion that was then reflectedin behaviors and further exacerbated tensions. Ouranalysis enables us to clarify and extend theoryrelated to language and subgroup dynamics in sev-eral ways. First, by virtue of having rich ethno-graphic data, we are able to identify the potent roleof language asymmetries as a faultline dimension,one that was persistently available and concentratedteam members’ emotional energy, leaving othermore taboo issues (such as power contests) obscured.Second, we capture the process by which subgroupsare likely to emerge from faultlines, and how theyare then held in place, in particular, by powerdynamics and the emotions that are ignited aroundlanguage asymmetries.

METHODA five-person research team conducted ethnographicinterviews and observations at a large high-techmultinational company, GlobalTech (a pseudo-nym), headquartered in Germany. GlobalTech offi-cially adopted English as its lingua franca 2 yearsprior to this study. Operationally, the rule requiredthat official verbal and written communicationoccur in English, so that employees from variouslanguage backgrounds could interact and collabo-rate. Our team traveled to two cities in Germany, onein India, and three cities in the United States toconduct interviews with 96 informants involved insix projects within GlobalTech. Most informantswere engineers, with the exception of 18 who servedas liaisons between engineers and the sales division.



Informants in India were 22–27 years of age and had1–5 years of tenure at GlobalTech. Informants inGermany and the United States were 30–40 years ofage and had 5–15 years of tenure at GlobalTech.Interviews were followed by observations of four ofthe six project teams.To facilitate understanding of cultural nuances,

our research team was composed of multilingualmembers who represented the three primary coun-tries involved in the study: Germany, India, and theUnited States. It is important to note, however, thatthe informants represented many more countries.The informants in the German office of GlobalTech,for instance, included people from several regions ofGermany, as well as from places as distant as theUnited States, Tunisia, and India. The Indian loca-tion included informants from many differentparts of India. Finally, the US office of GlobalTechincluded informants from China, India, Germany,Pakistan, and Australia.Our strategy was to collect data from teams that

were split between only two countries and betweenonly two locations. This focus served to minimize thevariations and complexities that arise from configura-tional differences involving more than two countriesor locations, which have been shown to significantlyaffect team dynamics (O’Leary & Mortensen, 2010).Table 1 presents a breakdown of the data collected on

each team and the team characteristics. Of the sixteams included in this study, three were split betweenGermany and the United States, and three were splitbetween Germany and India.

Semi-Structured InterviewsFollowing procedures for conducting ethnographicinterviews (Spradley, 1979), we structured interviewsaround a common set of open-ended questionsaimed at learning about informants’ experienceswith collaboration across sites and the nature of theirinterpersonal relationships with co-workers.We askedinformants to provide in-depth project explanationsas we posed descriptive questions to learn about theirdaily tasks, related communication demands, andteam dynamics (Spradley, 1979). Sample promptsincluded “Could you tell us about your project?” and“How do you interact with your colleagues at othersites?” We also posed grand tour questions (Spradley,1979) such as “Describe what you did yesterday fromthe time you walked into your office.” Grand tourquestions allowed us to hone in with mini-tourquestions (Spradley, 1979) to get details about specificevents and participants’ experience of them. This lineof inquiry was particularly useful in probing delicatesubjects such as the emotions and behaviors asso-ciated with a mandated lingua franca and the subse-quent impact of those behaviors on distant team

Table 1 Summary of task descriptions and data collected by team

Name Locations Interviewsa Observationdays

Description

HH1 US 6 0 Product management team supporting strategic deals and special projectimplementation.Germany 12 0

HH2 US 7 5 US-based team members tasked with creating the UI, while German-basedmembers provide the back-end functionality software and the databaseapplication.

Germany 9 5

MM1 India 14 0 Provided support for retail software solution. Ownership over softwaresupport transferred almost completely from Germany to India.Germany 4 0

ML1 India 6 5 Quality management in India with development support across sites andproduct management primarily in Germany.Germany 5 5

LL1 India 16 5 Responsible for ensuring that data formats are compatible between thecompany’s technology and the client’s systems. India-based team workedmostly on client side while German team members worked on server side.

Germany 8 5

LL2 US 2 5 In charge of software development for personnel-related functionality. USteam members worked on two aspects of personnel management, whileGerman-based members worked on one.

Germany 7 5

Totalsb 96 40 —

aInterview numbers include interviews with the managers on these teams.bFollow-up team meetings a year later are not included in the total.



members. While our questions were aimed at cover-ing topics connected to work experiences, the inqui-ries were open-ended, so that the most pressing issuesfor individual informants guided the conversations.We conducted interviews primarily in English, but

when it was evident that a German informant wasanxious about being interviewed in English, weoffered to interview him or her in German. With theexception of two informants, all agreed to be inter-viewed in English. Each interview averaged about 1 h,and most were conducted in conference rooms, pri-vate offices, and – on some rare occasions – cafeteriaswhere informants and interviewers were seated inisolated areas. Interviews were tape-recorded, trans-lated, and transcribed. Occasionally we collected dia-grams, email correspondences, and other artifacts thatinformants offered to supplement interviews.

Concurrent ObservationsBecause we were interested in exploring subgroupdynamics in teams spread across two locations, wecarried out concurrent observations of four of theproject teams (two split between Germany and theUnited States and two split between Germany andIndia). We use the term concurrent observations torefer to the process of simultaneously observingteam members located at multiple locations. Forexample, for a team split between Germany andIndia, one researcher conducted observations at theGerman site, while another researcher observed inIndia during the same week. This format allowed usto record social interaction and team dynamicsas they occurred and provided rich data on howindividuals at each location experienced team inter-actions, how people experienced cross-site and localmeetings, and how events were interpreted similarlyor differently across sites.Observations of each team lasted from 8 to 12 h

per day for 1 week. Extensive field notes document-ing our observations were typed and distributed toeveryone on the research team at the end of eachday. During observations, we regularly exchangedemails to notify each other about important activ-ities. For instance, one researcher observing inIndia sent her counterpart in Germany an emailabout a teleconference scheduled by someone inIndia that involved individuals on the team inGermany. These regular communications enabledus to observe the same events from our respectivelocations. During observations, we occasionallyasked informants questions to clarify issues andget explanations and interpretations of activities.We paid close attention to the interactions,

attitudes, and responses of individuals as theycommunicated with both collocated and distribu-ted colleagues. In addition to sitting with infor-mants to observe them while working, we attendedmeetings, observed conference calls, had lunchwith informants, and went to after-work socialgatherings whenever possible.Although there were only a few cross-site meetings

during a given week, we captured some examples oflanguage dynamics (e.g., code switching) in themeetings that occurred. Observations were also use-ful in detecting team dynamics and, in particular,cross-location subgroup dynamics and power con-tests. Further, because the concurrent observationswere conducted between 3 and 7 months after ourinitial interviews, we were able to capture how thedynamics between team members and perceptionsof coordination challenges across sites evolved overthis period.Approximately 1 year after the initial set of inter-

views and observations, two of the authors returnedto the GlobalTech offices in Germany, India, and theUnited States to discuss, validate, and better under-stand some of the observations from our data analy-sis. These meetings were important in helping us tohone our understanding of the language, subgroup,and power dynamics at play.

Data Analysis: Individual LevelFor our data analysis, we submitted all of our inter-view transcripts and field notes to a software pro-gram that we used to assign codes to passages of ourdata, as is common practice when working withethnographic data. Our analysis followed empiricalgrounded-theory procedures (Strauss & Corbin,1998). We began with a process of open coding, inwhich two of the authors read all of the data andassigned codes based on what emerged from it. Onerecurring theme from open coding revolved aroundcommunication anxieties and frustrations that arosefrom the use of English as the lingua franca. Withinthe first week of interviews, in fact, we discoveredthat English language communication was a deeplyfelt concern for many of the people with whom wespoke. This was further supported during opencoding, when we found that nearly 70% of infor-mants talked about their experiences with lan-guage asymmetries at GlobalTech. At the sametime, we began to cycle back and forth betweendata analysis and reviewing the literature in inter-national management, organizational behavior,sociolinguistics, communications, and social psy-chology to make sense of our findings, as well as to



refine our coding. In this process, the link betweenlanguage asymmetries, emotions, and emotion reg-ulation emerged.In subsequent rounds of coding, we classified

individuals into three broad levels based on theirlevel of fluency in English and German. Informantswere either native English speakers (e.g., those whowere born or raised within an English-speakingcountry, or who reported having learned Englishas a first language), or bilingual speakers (those wholearned English as a second language, but demon-strated speaking proficiency at the level of an edu-cated native speaker), or professional-level speakers(e.g., those demonstrating the ability to use Englishaccurately and participate in conversations, but whosometimes made errors, interspersed non-Englishwords, and/or frequently asked for questions orcomments to be repeated). We made use of theInteragency Language Roundtable (ILR, 2012) skill-level descriptors for competence in interculturalcommunication, designed for use in governmentsettings such as the US Foreign Service. The ILR treatsnative and bilingual fluency as the same, but wedifferentiated these as described above, because thedynamics between native and non-native Englishspeakers has been shown to be important in globalteams (Neeley, 2013). We also combined the ILRLevels 3 and 4 (professional competence and advan-ced professional competence) to create a singlecategory of professional competence, because wewere unable to distinguish these two levels ade-quately in our data. We had no need for Levels 1and 2 (elementary and limited working competence)because our informants were all at least profession-ally competent. For coding German language com-petence, we used only two levels of competence:those who spoke German fluently and those whodid not. There were several informants who hadattempted to learn some German and had mastereda few words or phrases, but they were not ableto interact in German or participate in German-language meetings. All other informants were eithernative or bilingual German speakers, or spoke noGerman.In analyzing the data, we were struck by the

strategies that team members used to avoid experi-encing anticipated negative emotions or to alleviatetheir experience of them. We then moved to theintermediary step of axial coding (Strauss & Corbin,1998), in which we captured the host of strategiesused by our informants to deal with uneven lan-guage fluency, the emotions or anticipated emotionsthat were associated with these strategies, and the

effects of these behaviors on team members. In thelast stage of analysis, selective coding (Strauss &Corbin, 1998), we integrated and developed ouroverall model by iterating with theory and literatureon emotion regulation and language asymmetries ininternational organizations, subgroup dynamics, andpower dynamics in global teams. We noticed, forexample, that informants were describing emo-tion-regulation strategies to deal with the negativeemotions associated with language asymmetries. Theemotion-regulation process includes stimulus regula-tion, reappraisal, experience regulation, and displayregulation (see Table 2, also Elfenbein, 2007). Wetherefore applied these codes to our data to under-stand how emotion-regulation processes were used,and to uncover their effect on global collaboration.Finally, as noted, we scrutinized the validity of ourfindings by sharing themwith a subset of GlobalTechstudy participants. Doing so yielded additionaldetails that sharpened our interpretations.

Data Analysis: Team LevelAfter completing our individual-level analysis of therelationship between language asymmetries andemotion regulation, we turned to team-level analysesof the six teams in our study. We first read the data byteam and, in doing so, noticed that strong subgroupdynamics coincided with emotionally charged lan-guage issues, and that power dynamics were inter-twined with subgrouping and language challenges. Inthe next round of coding, we therefore calculatedfaultlines and focused on informants’ response tolanguage asymmetries (as described above), subgroupidentities and subgrouping, and power dynamics.Our coding for subgroup identities and subgroup-

ing was derived from previous theory related tolanguage and identity (Bucholtz & Hall, 2004, 2005;Miller, 2000), focusing particularly on informants’social positioning of self vs other (see Bucholtz &Hall, 2005). We coded mentions of differencesbetween other social groups (including cross-site,cross-discipline, etc.), overt categories or labels suchas themselves being “other,” evaluative commentsregarding the other location, and mentions ofingroup/outgroup dynamics (us vs them). We alsocaptured mentions of similarity across sites andassociations with a team-level identity. To code forpower dynamics, we examined the sources of poweras described by team members (including access toinformation, decision makers, customers, key mar-kets, and growth opportunities), and the commentsor behaviors that demonstrated power struggles. Incharacterizing the power dynamics in each team,



Table 2 Emotion-regulation process definitions and examples of strategies enacted by German and non-German speakers

Process Definition Strategies enacted by German speakers Strategies enacted by non-German speakers

Situation selection Prior to experiencing an event thatstimulates an emotional response,people use their anticipated emotionsto make choices about exposingthemselves to an experience that islikely to give rise to a particularemotion, whether it be desirable orundesirable (Gross & Thompson,2007).

“If we are going to extend the meeting to a largerforum, if we have to talk in English, then I say no!No, I don’t want to do this.”A German informant said that he almost alwaysrefrained from calling his co-worker in India to avoidthe stress of conversing in English. He said that hewould either find another way to address hisconcern or search for collocated colleagues whowould be able to help.“If I needed perhaps to write documentation, itwould have to be in English. It’s too hard to write inEnglish, so I don’t do it and there’s nodocumentation at all because it is too difficult towrite it in English.”

“I don’t actually stay because I feel it’s a waste ofmy time and perhaps of theirs because I will notunderstand it. So, it’s really no use. If you misstwo words, you miss the whole idea of what theywere talking about.”

Situationmodification

People change a situation to alter itsemotional impact (Elfenbein, 2007;Gross & Thompson, 2007). In cases ofsituation modification, the individualstill exposes himself or herself to thesituation, but attempts to modify theconditions that evoke the desirable orundesirable emotion.

One of our informants said, “Some people just don’tspeak up at all [in meetings] because they are tooshy to show their [poor] language skills.”“Sometimes people (engineers) do not speak up indesign reviews because they don’t feel comfortableexpressing themselves in English.” She told us thatshe knew this because these engineers wouldsometimes speak with her in German after meetingsabout issues that they were otherwise reluctant toraise.A German informant said that they attempted tospeak in English most of the time “but sometimessomeone asks a question in German, and theneverybody starts talking in German.”

A US informant of Indian origin told us how shedealt with a constant flood of emails in German.She first tried to alter the situation by askingthem to send an English version of the email andthen, after getting no response, reminded themof GlobalTech’s lingua franca policy. She saidwith annoyance:“We had a colleague last year who constantlykept sending emails in German, right? And I keptsending back emails. I mean it was through adistribution list, and they very convenientlyassumed that everybody on the DL is German. SoI replied back to them saying … you know, I’mreally sorry that I don’t know German. I am tryingto learn, but until then, could you please sendmean English version? Right? And ignore … I thinkwe actually had an email from our CMR head,like the guy who leads the whole division sayingthat every communication, if it is Germany,should follow in an English version. I actually hadto send him a copy of that.”A developer in the United States explained howshe dealt with code switching. She said, “I justtell them, like … guys, I’m sorry, you can’t speakGerman … or something like that.”

Languag

eas

alig

htn

ingro

dPam

elaJH

indset

al542

JournalofInternationalBusinessStudies

we noticed that two groups had power contests(e.g., expressions of competition, disagreementsabout who should get more work), two groups wereexperiencing a shift of power between locations(e.g., mentions by either location that one groupwas losing and another gaining resources anddecision-making authority), and two groups hadan imbalance of power that was not contested byeither location (e.g., mentions that the other loca-tion had significantly more resources, but no objec-tions to this state or evidence of competition overresources).In our team-level analysis, we coded each team,

one location at a time. With our first Germany–USteam, for example, we coded the US informants firstand then the German informants. In doing so, wegot a strong sense of the team by location and thecross-location dynamics. After coding the data foreach team, we created team summaries that capturedwithin-team and across-location dynamics for thatteam. Our summaries enabled cross-team compari-son to better identify the patterns and under-stand the similarities and differences between thedynamics within the six teams in our study. Usingthe team summaries and the coded data, we classi-fied each team based on the strength of subgroupdynamics, the extent to which language issues weredealt with empathically or were negatively charged,and the nature of the power dynamics. Our classifi-cation process is described in the relevant sections ofthe findings. Finally, to classify the strength offaultlines, we captured each informant’s location,nationality, English language fluency (native, bilin-gual, and professional), and German fluency (yes,no).1 Based on this, we assessed alignment betweenlocation and nationality, location and English lan-guage fluency, and location and Germany fluency.For example, if location, nationality, and languagefluency were aligned, strong location-based faultlineswere indicated. If some but not all were aligned and/or nationality and language fluency were cross-cut(e.g., German speakers at both locations) faultlineswere classified as moderate.2

FINDINGSOur research was grounded in the study of languagechallenges in global teams to help explicate howasymmetries in the lingua franca contributed tofracturing and subgroup dynamics in these groups,but our inductive investigation also revealed that,when subgrouping occurred, language asymme-tries stimulated emotion-regulation strategies thatreinforced subgroup dynamics. Further, we foundRe

appraisal

Aproc

essby

which

workersco

gnitively

refram

etheeven

tto

chan

getheway

that

they

aredirectingtheirattention

(e.g.,igno

ringit)

orch

ange

their

assessmen

tof

theeven

titself(Gross,

1998

;Elfe

nbein,

2007

).

Aninform

anttold

uswha

the

thou

ghtitmight

belikeforno

n-German

spea

kers,a

ndho

wim

portan

tit

was

tospea

kEn

glish:

“[If]y

ousitin

agrou

pof

German

san

dthey

talk

German

toea

chothe

rand

youdo

n’tk

now

ifthey’re

talkingab

outyo

uor

not,an

d…

Iwou

ldfeel

unco

mfortab

lebe

ingin

China

andthem

justtalking

Chine

sean

dmeno

tbe

ingab

leto

unde

rstand

anything

.So,

beingsensitive,I

guesswetryto

talk

Englishas

muc

has

wecanan

dtryto

invo

lve

peop

le.”

Inspea

king

abou

thisGerman

team

mem

bers,a

USinform

antsym

pathized

:“So

metim

esthey

feel

they

cann

otco

mmun

icatesometech

nicalissue

usingEn

glish,

they’vego

tto

goba

ckto

their

mothe

rtong

ue.A

ndthat

oneIcan

unde

rstand

.”Ano

ther

nativ

eEn

glish-spea

king

inform

ant

expressedap

preciatio

nfortheeffortsof

their

German

colleag

ues,saying

:“Ithink

even

thou

ghitisreallytoug

hforthem

,the

ycrea

teallthe

docu

men

tsin

English.

That

way

they

aredo

inga

reallygo

odjob.”

Expe

rienc

ean

ddisplayregu

latio

nDeliberatech

ange

sin

one’sem

otiona

lstate,

essentially

supp

ressingor

deny

ingan

emotionthat

existsrather

than

alterin

gthesituationor

tran

sformingtheem

otionby

reap

praising

thestim

ulus

(Gross

&Th

ompson

,200

7).

Noex

amples

inou

rda

ta.

AnIndian

team

lead

explaine

dho

wthey

wou

ldha

vemee

tings

that

wou

ldslip

into

German

,“bu

tweshou

ldsm

ile,y

eah,

wearesm

iling

,”he

said,

despite

thefact

that

they

couldno

tun

derstand

wha

twas

beingsaid.



that power contests played a pivotal role, bothactivating otherwise dormant faultlines and beingreinforced through the language strategies invokedby workers.We start by describing the interactive situation

that set the context for communication amongglobally distributed workers at GlobalTech, and thendescribe how team members responded to and dealtwith language dynamics, focusing particularly onthe emotion-regulation strategies employed and onco-workers’ responses to these strategies. We startwith responses to language asymmetries because thiswas most salient for our informants and provides afoundation for the team-level dynamics. We thenpresent the team-level analysis, in which we describethe subgroup dynamics that characterized the teamsin our study. Finally, we describe the power struggleswe observed and how the balance (or not) of poweracross locations played a crucial role in determiningthe extent of subgrouping, and was likewise rein-forced by the emotion-regulation strategies used byteam members.

The Interactive Situation at GlobalTechAs mentioned earlier, GlobalTech established Eng-lish as its business language 2 years prior to ourstudy. Reaction to the language policy among itsworkers was mixed. US-based and India-based per-sonnel perceived the introduction of English as theformal business language positively. US-based infor-mants were mostly native English speakers, despitetheir wide range of ethnic origins, although somelearned English as their second language. Most ofthe informants we interviewed in India were bilin-gual in English, although a few of them displayedprofessional fluency, demonstrating some struggleswith understanding and articulating themselveswithout errors in English. For the most part, how-ever, our Indian informants learned English at ayoung age, received their technical education inEnglish, spoke English extensively every day at work,and expressed confidence in their English fluency.In contrast, although many of our German infor-mants were bilingual, speaking and writing inEnglish was often difficult. Many told us that,although professionally fluent, they received theirtechnical education in German, and were for-mally trained to think and articulate ideas in theirnative language. In particular, most received theirtechnical education in German. Their confidencein English was mixed and nearly all of themexpressed feeling some anxiety about having accessto appropriate words, particularly in situations that

were emotional or highly technical, or when theywere fatigued. The German team members, unlesscollaborating with non-German speakers, spokeprimarily German at and outside work. With fewexceptions, team members in the United States andIndia spoke negligible German. It should be notedthat asymmetries in fluency existed for both verbaland written communication.All of the teams that we studied were interdepen-

dent across location and depended on team mem-bers at the distant location for project success.As a result, they interacted regularly in the Englishlingua franca through a combination of face-to-facemeetings (site visits), videoconferences, telecon-ferences, phone calls, and emails. These were theprimary occasions for anxieties associated with lan-guage asymmetries. German informants, for exam-ple, referred to the need to speak and write emailsand documents in English as a “bottleneck” or“handicap,” or as what “holds them back” fromeffective communication. In describing his experi-ence, one German informant said that “the Englishexpression is not really making the point thatyou are trying to make, and you are maybe an inchaway from it.” German informants explained thatthe experience of speaking English was more timeconsuming and, consistent with research on secondlanguage acquisition (see Mettler, 1984), more stress-ful. Attempting to communicate technical or busi-ness topics precisely in English was especiallyfrustrating. One manager told us that, even if peoplepursue English language instruction, “they don’treally teach engineering jargon there.” In additionto feeling hindered by the official lingua franca atGlobalTech, many German informants said theyfelt awkward speaking in English in front of otherGermans, particularly when they lacked confidencein the language. Likewise, many Germans preferredto write emails and documents in German if theyexpected the only readers to be their countrymen.One informant told us that “it always feels stupidto have two German colleagues talking to each otherin English.”Non-German speakers felt similarly anxious about

the dynamics associated with the lingua franca.They reported frustration, anger, guilt, and feelingsof exclusion as a result of strategies invoked by theirGerman-speaking collaborators to deal with lan-guage asymmetries. Greg, a developer based in theUnited States, explained how he felt when his Ger-man colleagues would speak German in his pre-sence. In his words, “It didn’t feel good at all. It feltthat the other person didn’t really respect you.” In



sum, the lingua franca dynamics at GlobalTechtriggered emotional responses from the native, bilin-gual, and professionally fluent English speakers.

Emotion-Regulation Strategies of GermanSpeakersEmotion regulation refers to a set of processes bywhich emotions, negative or positive, are regulatedwith or without conscious awareness (Gross &Thompson, 2007). Regulation can occur throughoutthe process of an emotional experience, from stimu-lus to expression, including situation selection,situation modification, reappraisal, experience regu-lation, and display regulation (see Elfenbein, 2007;Gross & Thompson, 2007). We provide definitionsfor each of these processes with additional illustra-tions from our data in Table 2 (see also Neeley,Hinds, & Cramton, 2012).German speakers at GlobalTech who had profes-

sional fluency in English used most of these strate-gies. Informants told us, for example, about how theyavoided language situations that provoked anxietyand shame by not attending meetings in which theyanticipated having to speak in English and, whenthey had control over the invitation list, inviting onlyGerman-speaking members of the team (situationselection). In addition, informants said that decisionsabout who to include in communications were some-times made on the basis of whether the involvementof that person would require that English be spoken,resulting in a more difficult exchange. Avoidance ofthe lingua franca was not isolated to spoken English.Workers also resisted pressures to create documenta-tion, including emails, in English. Excluding nativeEnglish speakers or avoiding lingua franca situationsseemed to be particularly prevalent when informantswho did not feel proficient in English felt heightenedapprehension due to time pressure.Even more prevalent in our data were situation

modification strategies designed to alter the anxiety-producing language situation. German speakers wholacked confidence in the lingua franca attendedmeetings, but remained silent, switched to Germanfor some of the meeting, and, when faced withhaving to produce emails and documents, gener-ated those in German as well. Our informantsrepeatedly told us that when they were mentallyor physically fatigued, for example, at the end ofthe workday or under time pressure, they simplyremained silent at English-language meetingsbecause the additional effort required for them tospeak English became overwhelming, and heigh-tened anxiety. Another prevalent regulation

strategy of the German speakers was to switch toGerman during English-language meetings. Codeswitching is a phenomenon in which a bilingualspeaker shifts from one language to another in thecourse of a conversation (Auer, 2000; see alsoHarzing, Köster, & Magner, 2011). In our data, codeswitching was employed as a means of altering asituation that instilled anxiety, shame, and frustra-tion in speakers. When apprehensive about theirlanguage skills, workers at GlobalTech simplyreturned – at least temporarily – to their nativelanguage. During our observations, we documentedoccasions in which German colleagues code-switchedduring the course of teleconferences that involvedtheir distant co-workers. Many Indian and US infor-mants also discussed code switching during our inter-views with them. In extreme cases, we were told,meetings turned entirely into German-language ses-sions. German informants were aware that they attimes reverted to their native language in the courseof communications involving non-German speakers.They perceived this as a minor incident that occurredonly for short periods of time. They said theyattempted to speak in English most of the time “butsometimes someone asks a question in German andthen everybody starts talking in German.”The final strategy we observed among the German

speakers was reappraisal, which refers to emotion-regulation strategies in which people reframe thestimulus, thus changing their emotional responseto it. A German informant, for example, explainedthe importance of getting accustomed to speakingEnglish so that Indian colleagues would have accessto the information. He said: “Perhaps if I do it inEnglish, there’s a chance an Indian colleague canfollow what I did.” In reappraising the situation,reframing it by being empathic about their distantcolleagues’ experience, non-native English speakersfound a way to avoid the unpleasant emotionsassociated with articulating their ideas in English bychanging their interpretation of the meaning theevent.

Non-German Speakers’ ResponseWhen German team members regulated their ownemotional responses through situation selection andsituation modification, this often triggered negativeemotional responses in their non-German-speakingcollaborators. Native, bilingual, and professionallyfluent English speakers who spoke no German toldus of their anger, frustration, and discouragementwhen excluded from meetings, unable to readdocuments and emails, and confronted with code



switching during visits and video or teleconferences.While the company’s lingua franca policy stipulatedthat work-related communication be conducted inEnglish, the policy was not always followed. Infor-mants who did not speak any German told tales ofbeing forwarded email threads that had begun inGerman and then crossed country and linguisticlines. Sometimes a brief English summary wasattached, which was intended to capture the gist ofthe trail of communication exchanges, and whichcontained a request for a response. Informants, how-ever, said that the summaries typically did not conveyto their satisfaction the previous communication, therequest put to them, or an indication of how torespond. The central source of discontent seemed tobe the inability to derive meaning from a commu-niqué where tacit, contextual, and jointly constitutedknowledge was obscured. An Indian informant,Geet,3 recounted what he said was a frequentoccurrence:

Emails get written in German where two, three emailsbetween people go back and forth on some topic, and thenit’s forwarded to me with some comments in English…Mostof the time, somebody who is handling some problem wouldhave analyzed the problem in German before it is sent to me.If that piece is in German, it’s difficult for me to understandwhat’s going on. That is the main problem. They need tostick to the company language.

Geet’s plea that people “stick to the companylanguage” captured the ongoing frustration that hefelt at being excluded from communications that wereimportant to his work. Code switching created parti-cularly strong emotional responses. A number ofnon-German-speaking informants described feeling“lonely” when they were excluded from meetingsor when their German team members conversed inGerman in their presence. English speakers, at times,concluded that the switch was intended to excludethem. Said one US informant:

The classic fear of all Americans, unilingual inhabitants ofthis continent, is that when somebody’s talking in Germanand they know you don’t speak German, the natural thoughtis – they’re talking about you.

In India and the United States, engineers won-dered about how, in the face of important docu-ments and emails being only in German, a languagethey could not speak or read, they could really beconsidered valuable members of the team.To avoid these real or anticipated negative emo-

tions, non-German speakers also relied on emotion-regulation strategies. We saw evidence that non-German speakers, albeit infrequently, used

situation selection strategies to avoid or changesituations that they anticipated would yield anxi-ety, discomfort, or frustration. In some cases wheretheir German-speaking colleagues would switch toGerman, non-German-speaking team memberswould leave the meeting to avoid “wasting theirtime” and feeling excluded. More common, how-ever, were attempts at situation modification. Non-German speakers attempted to alter the situationwhen they confronted emails or documents thatthey could not understand, or code switching onthe part of their German colleagues. They asked forexplanations or translations, declared that theircolleagues’ behavior was rude, asked that meetings,documents, and emails be in English, invoked thecompany rule stipulating that English be spoken inthe presence of non-German speakers, and eventried to learn German. A common response to codeswitching was to draw attention to the fact thatnon-German speakers were present and to askthat English be spoken. A developer in the UnitedStates, for example, explained how she dealt withcode switching. She said, “I just tell them, like …

guys, I’m sorry, you can’t speak German … orsomething like that.”Non-German speakers also frequently described a

reappraisal strategy in which they recognized thelingua franca dilemma as “just part of the job,” orthought about the language issue as one thatplagued everyone, thus not taking the coping strate-gies personally, and even empathizing with theirGerman colleagues. Empathy is an emotional pro-cess in which people feel compassion and concernfor others (Barsade, 2002; Davis, 1983). When co-workers took the perspective of others, empathizedwith their struggles, and altered their action accord-ingly, a break in the negative emotional cycleoccurred. Empathic responses led to sensitivity andbehavioral adaptations that seemed to amelioratesome of the tensions associated with language asym-metries. When co-workers empathized with theplight of their distant colleagues, they often tookaction that was less likely to be interpreted asthreatening or devaluing. Informants told us, forexample, about listening more carefully, workinghard to involve others, and being painstakinglycareful in their communication to avoid causingoffense. Although empathy was not the primaryresponse to asymmetries in lingua franca fluency,this alternative path eased the emotional burdenof the lingua franca mandate.We also saw ample evidence of experience and

display regulation among non-German speakers,



particularly when confronted with emails and docu-ments in German and code switching in meetings.An American informant, for example, said thatinitially she would be “infuriated” when her collea-gues switched languages in her presence. Later, shesaid that her anger turned into resignation and shejust sat silently and “let them continue to speak ontheir own.” They would also regulate their display ofemotions by pretending they understood and smil-ing at jokes, regardless of how they were feeling.Overall, language asymmetries triggered intense

emotional responses, which were then regulated toreduce discomfort. German speakers relied heavilyon situation selection andmodification, and some ofthese strategies, particularly code switching, evokedpowerful emotional responses in their non-German-speaking team members. Non-German speakers inIndia and the United States then responded, primar-ily with strategies grounded in frustration and anger,but occasionally with empathy.

Language Asymmetries, Subgroup Dynamics, andEmotion RegulationOur data show that the team dynamics, especiallysubgrouping, were intertwined with the emotion-regulation strategies described above. In particular,the teams that experienced less subgrouping acrosslocations also relied on reappraisal that was moreempathic and reduced the emotional charge asso-ciated with language asymmetries, whereas theteams that experienced more severe subgroupingused more situation selection and modification(e.g., excluding team members from or leaving ameeting, code switching) that ignited emotionalreactions in distant colleagues.For our analysis, we categorized each of the teams

with regard to the strength of subgrouping acrosslocations based on the codes described in themethod section. So, for example, teams whose mem-bers at one or both locations consistently referred tothe differences between locations, a sense of being“other,” and a strong us vs them dynamic were ratedas high, and teams that talked primarily about simi-larities between locations and referred to their teamsas integrated across the two locations, were rated aslow. The teams we rated as medium had some us vsthem dynamics, but they were not felt by the entireteam, and/or the discussion of differences was coun-terbalanced with discussion of similarities by multi-ple members of the team. We also categorized eachteam as high, medium, or low based on their affec-tive responses to language asymmetries, where highindicates that the team, on the whole, experienced a

highly charged emotional environment around lan-guage asymmetries. On these teams, there was highnegative affectivity and/or low positive affectivityacross most team members (or most team membersat one location), reflecting unpleasant emotionsaround language asymmetries in these teams. Lowratings indicate a weak emotional charge, with gen-erally empathic or neutral comments around lan-guage asymmetries in the team, and medium wasused to capture teams that had some frustration, butalso some empathic responses to language asymme-tries. To illustrate our findings, we describe thecontext, structure, and dynamics of four of theseteams, two with strong subgrouping across loca-tions and highly charged emotional responses tolanguage asymmetries (high/high (HH)), one withmoderate subgrouping and some negative affectaround language (medium/medium (MM)), and onewith weak subgrouping and virtually no emotionalcharge around language asymmetries (low/low (LL)).Table 3 captures the faultline strength and distance,the location-based subgroup dynamics, and the lan-guage dynamics and emotion regulation associatedwith language asymmetries in the six teams in ourstudy.

High/high 1The HH1 team was split between the United Statesand Germany. Of the 18 members, 12 were inGermany and 5, plus the German VP, were locatedin the United States. This was a technical productmanagement group that sat between marketing anddevelopment. They gathered customer requirementsand created demos for potential customers to gener-ate interest in product offerings. The faultline onHH1was moderate. Team members in Germany werehomogeneous with regard to nationality, but bothlocations had mixed language proficiency, includingthree German speakers on the US-based team.This team experienced strong divisive subgroup

dynamics between locations, using the language“our team and their team” frequently to describethe two locations, despite their stated dependenceon one another. The team members in Germanyperceived the US teammembers as being quite differ-ent from themselves. In interviews, the German teammembers pointed out their different work styles,“mentalities,” communication styles, and capabil-ities, suggesting that these differences were a sourceof “tension.” One German team member, for exam-ple, described how “we’re trying here to sometimesfix the most tiny little screw and might not beworking … whereas in the US, they would rather



Table 3 Faultline, subgroup dynamics, emotion regulation around language asymmetries, and coordination challenges by team

Team Locations Team composition by location Faultline strength anddistance

Emotion regulation around languageasymmetries

Subgroup identification andsubgroup dynamics

Powerdynamics

HH1 US (6)Germany (12)

US-based team members described asCaucasian American (one with Germanheritage), Indian, and ChineseAmerican. There were two Germanteam members in the United States,including the team lead. Three female/three male. Mixed English fluency,ranging from native to bilingual. ThreeGerman speakers.German-based team members allappeared to be from Germany. Fourfemale/eight male. Mixed Englishfluency from bilingual professionalfluency. All fluent German speakers.

Moderate faultline withnationality homogeneityin Germany, but cross-cutting language.

The division between the United Statesand Germany was strongly felt on thisteam, with numerous attributions tolocational and cultural differences.Competition was mentioned on thisteam. In particular, team members inGermany worried about getting theattention of their boss, who hadrelocated to the United States.

This team had acutelanguage issues and strongnegative emotions associatedwith the regulation strategies(e.g., code switching,exclusion). Team membersfelt disrespected, infuriated,and resigned to the situation.There was little evidence ofreappraisal as a strategy.

Powercontested.

HH2 US (7)Germany (9)

US-based team members described asCaucasian American, Asian-American,or Indian. Some team members wereborn and raised in Asia. One female/sixmale. Mixed English fluency, rangingfrom native to professional fluency.One fluent German speaker.German-based team members allappeared to be from Germany. Ninemale. Mixed English fluency, frombilingual to professional fluency.

Moderate faultline withnationality and genderhomogeneity inGermany, but cross-cutting language acrosslocations.

There were strong tensions betweensubgroups by location, but this wasmostly expressed by the US-basedteam members.Competition and fear were mentionedfrequently on this team, and teammembers in Germany questionedwhether the US location was needed.

This team had a highlyemotionally chargedsituation around languageasymmetries. Teammembers, particularly in theUnited States, felt ostracized,angry, left out, and resigned.They rarely used reappraisalas a strategy.

Powercontested.

MM1 India (14)Germany (4)

India-based team members from India.Two female/twelve male. MixedEnglish fluency, from bilingual toprofessional fluency. No Germanspeakers.German-based team members allappear to be from Germany. Fourmale. Mixed English fluency, frombilingual to professional fluency. Allfluent German speakers.

Strong faultline, withnationality homogeneityin Germany and India andno cross-cutting Germanlanguage fluency.However, there are alsono fluent Englishspeakers.

Although the German team membershad negative impressions of Indianteam members, and there was a clearfaultline because the work was beingmoved, the group still talked aboutbeing a single team, and there was notmuch evidence of subgroup tensions.

This team had somelanguage issues andmentioned exclusion andcode switching astroublesome, but they alsoreported constructiveresponses, and emotionsaround language were nothighly charged.

Powershifting.

Languag

eas

alig

htn

ingro

dPam

elaJH

indset

al548


ML1 India (6)Germany (5)

India-based team members from avariety of Indian states. Two female/four male. Mixed English fluency, frombilingual to professional fluency. NoGerman speakers.German-based team members weremixed, with one from Yugoslavia andone from India at the time of theinterviews. Two female/three male.Mixed English fluency, from fluent toprofessional fluency. Two fluentGerman speakers (three non-Germanspeakers).

Moderate faultline withnationality homogeneityin India and no cross-cutting German languagefluency, but withcross-cutting Indiannationality (e.g., Indiansin Germany).

Overall, good team integration. Teammembers had a lot of autonomy, butspoke of themselves as a team. Teammembers at both locations spokepositively about the other location,although there was some mention ofcultural differences (such as the Indianteam members not saying “no”) asculturally different.

This team had somelanguage issues, anddiscussed feeling neglectedand excluded, but often hadadaptive and empathicresponses. Indian teammembers reported usingeasier words and speakingmore slowly to compensatefor their team members’ lackof English proficiency.

Powershifting.

LL1 India (16)Germany (8)

Indian-based team members fromvarious locations in north and southIndia. Two female/thirteen male.Mixed English fluency, from bilingualto professional fluency. No Germanspeakers.German-based team memberspredominantly from Germany, withone from Tunisia with recent Germancitizenship. Three female/five male.Mixed English fluency, from bilingualto professional fluency. All fluentGerman speakers.

Strong faultline withnationality homogeneityin Germany and India,and no cross-cuttingGerman languagefluency. Also, no fluentEnglish speakers.

Generally positive and integrated.Team members feel comfortablecalling one another, and respect oneanother’s work and work ethic.Communication is described as openand positive.

Despite languageasymmetries, this team rarelymentioned language as aproblem. They describedadaptive and empathicresponses to what they sawas a challenge for allinvolved.

Acceptanceof powerimbalance.

LL2 US (2)Germany (7)

US-based team members included oneChinese American and one whodescribed himself as Turkish Australian.One female/one male. Mixed Englishfluency, from fluent to bilingual. NoGerman speakers.German-based team members weremixed, with one from Austria, one fromChina, and one from Spain. Onefemale/six male. Mixed English fluency,from bilingual to professional fluency.Mixed German fluency, from fluent tobilingual.

Moderate faultline withmixed-nationality teamsat both locations, but nocross-cutting Germanlanguage fluency bylocation.

Team members were generally wellintegrated, and respected one another.

Despite languageasymmetries, this team rarelymentioned language as aproblem. They describedadaptive and empathicresponses to what they sawas a challenge for allinvolved.

Acceptanceof powerimbalance.

Languag

eas

alig

htn

ingro

dPam

elaJH

indset

al549


concentrate on … the general impression that some-body is getting.” Some of the German teammembersreported that the US team members were not ascompetent and had to be monitored closely which,they said, was difficult at a distance. The US teammembers repeatedly described themselves as being“outsiders” on the team and pointed out that they“put a US perspective on things.” Subgrouping wassevere in this team. As mentioned by one US-basedteam member: “There hasn’t been like a guidingprinciple … that like you’re … you guys should worktogether and leverage your work and collaborate.”Language issues also were highly charged in this

team, with many German team members expressinga lack of confidence in their language skills, and as aresult relying on code switching to alleviate theirdiscomfort. Code switching was common, andthose who did not speak German (all located in theUnited States) felt excluded, became infuriated, andcharged their German-speaking team memberswith being disrespectful and rude. In the interviewswith these team members, they became tearful asthey talked about being excluded from discussions,or spoke angrily as they described the situation inwhich they found themselves. As recounted by oneof the US team members talking about code switch-ing: “It’s not the most pleasant experience, andI think that’s a very big problem. It’s an HR issueat GlobalTech.” Another US informant said: “Theyare certainly entitled to speak their language.It’s just sometimes infuriating because they’lljust like break into it in mid-meeting and you’relike … okay.” To some extent, these US teammembers railed against code switching by tryingto change the situation. For example, they wentto management asking that the rest of the teammembers be required to speak the lingua franca.In the end, though, these informants said thatthey had “given up” and either resigned themselvesto their situation or considered leaving the teamas a result of their powerlessness to change thesituation. As stated by one team member: “Thereare so many times when I’m like … oh, why amI working on this team?”Overall, this team was fraught with tension.

Team members were anxious about being able toget their work done, and attributed breakdowns totheir distant counterparts. Language asymmetrieswere frequently pointed to as a source of frustra-tion on the team and US-based team membersidentified it as important evidence reflecting howthey were outsiders and devaluated by their Germancolleagues.

High/high 2The 16 members of HH2 were split more or lessequally between Germany (9) and the United States(7). The team had spent nearly 2 years developing aprototype software application in a new applicationspace. The US team members were primarily respon-sible for defining the technology for the user interface(UI), while the German teammembers worked on theback-end technology. The faultline in this team wasalso moderate, again with homogeneity of national-ity in Germany, but not the United States, and a mixof English fluency at both locations. In addition,there was a German speaker at the US location.As with HH1, there were strong location-based

subgroups in HH2, although the experience of sub-grouping was felt most by the US team members.The Germany-based team members, in contrast,generally described an integrated team with a strongshared identity, although during our observationsthey suggested multiple times that perhaps the teamshould be separated. Despite the fact that all teammembers were part of the same team in the samecompany under the same leader, team members inthe United States felt like “outsiders” and describedtheir location as an “outpost” and “isolated.” As oneUS team member explained: “I usually just considermy team to be the team here.” One team membereven described the US location as “a subsidiary in [UScity], we’re not in [German city].” Several US-basedteam members felt that their work on the UI wasbeing misrepresented by their German counterparts,which further exacerbated the us vs them divide.A US team member confided, “They are talking withone another and reporting that the technology forthis new UI is unstable and unusable.”The frustration around code switching was also

acute on this team, primarily from the perspective ofthe US members. They perceived frequent codeswitching during meetings, which was corroboratedby our observations, and were angry that emails anddocumentation were often in German. As relayed byone of the US team members: “If there is documen-tation we need, it’s in German. We asked them,‘Hey, can you translate to English?’ They say, ‘Oh,we think in German first and then we translateto English,’ but sometimes they’ll take forever totranslate to English.” The interpretation of codeswitching was experienced as exclusionary and theresponse of many US team members was anger,frustration, and defensiveness. One US team memberdescribed a situation where she faced an email thatwas entirely in German. She said, “They ticked me offwhen it happened.” In contrast to the US-based team



members, the German team members either did notbelieve language-related difficulties existed, or spokeempathically about the challenge faced by their dis-tant team members. In some cases there was ambi-guity about which documents and artifacts were inGerman and which were in English. One Germaninformant, for example, said that language was not aproblem, because “The PowerPoints are in English,the specifications, the documentations, the docu-mentations with the coding, everything is in Eng-lish.” Some also perceived that they “always switch toEnglish” to accommodate the US teammembers. Notall German team members, however, held this per-ception. Many of the German teammembers claimedthat most documents and “all development” were inGerman. They recognized that language was a prob-lem and that “there can be more misunderstandingsbecause of the language.” They attempted to adjust tothe situation by repeating themselves and explaining“it again with an example” to make communicationmore clear, but these adjustments were not perceived,nor did they assuage the frustration of the US-basedteam members.In sum, the US team members on HH2 felt that

code switching and lack of access to importantdocuments in English were prevalent, which con-tributed to their feelings of being isolated and deva-lued. During our study, the team splintered furtherand performance suffered. At the time of our obser-vations, the two sites were fighting for their owntechnology solutions and mistrust was rampant.

Medium/mediumThe MM1 team was split between Germany andIndia. Of the 18 team members, 4 were located inGermany and the remaining 14 were in India. Therewas a manager in each location, but the projectleader was located in Germany. Many of the Indianteammembers were new to the organization, havingbeen hired within the previous year. This was a teamresponsible for a large software application for cus-tomers in the United States and Europe. The applica-tion was fairly mature, so it was primarily in themaintenance phase, with limited new development.During our initial interviews, the team was in theprocess of moving all of the maintenance work toIndia. This was the third attempt to move themaintenance of this system to India, although theteam members in India were new. Our data suggestthat the faultline in this group was relatively strong.Only German nationals were in Germany and onlyIndians were in the India-based team. None of the

India-based team members spoke German, but Eng-lish fluency was fairly comparable across locations.Despite strong faultlines on this team, subgroup-

ing was moderate. In general, the Indians werepositive about the collaboration and about their Ger-man colleagues, although they had a sense that theyhad to be prepared for meetings, because the Ger-mans would not provide additional information ifnot asked about it explicitly. The Germans expressedsome frustration that the Indian team members werenot as capable as they had hoped and were anxiousto have the Indians learn German work practices. Asdescribed by one German teammember: “Sometimesthey are just … faced with a problem, they only seethis small part of the problem and… sometimes theydon’t take care of problem with side effects … theirview is a bit narrow sometimes.” There was a clearsense from both sides of the Germans having moreexpertise, although some of the Indian team mem-bers felt that their German colleagues underestimatedtheir talents. In addition to recognizing differences inskill and expertise, members from both locationsrecognized that there were cultural differences inhow people collaborated and framed the problem.One of the German team members, for example,observed that the Indian team members “do notinitially cooperate with women like we do.” AnIndian team member also described differences incommunication approaches, saying, “They’re veryfrank … about things. I wouldn’t say all the negativethings as they would say so openly … .”With regard to language, the German team mem-

bers in MM1 expressed concerns about their ownEnglish proficiency, and worries about not beingable to express fully what they meant, but didnot report any major difficulties with using thelingua franca. The Indian team members similarlydid not have much to say about language asymme-tries, although they reported that the Germanssometimes communicated among themselves inGerman, and these messages would make their wayto India, but not be accessible to the Indian teammembers, which they found frustrating. They alsocomplained that, during visits to Germany, theirGerman team members would speak in Germandespite their inability to understand and felt “leftout” as a result. Also, initially, major parts of thecode and documentation were written in English,but had been translated by the time we began ourstudy. A close inspection of the interviews showsthat the regulation strategies by team members weremore constructive than in teams HH1 and HH2, andthat the issue of language asymmetries was less



emotionally charged. Indian team members tried tolearn technical terms in German, used online trans-lation services, and asked clarifying questions.German team members worked to improve theirEnglish, patiently repeated themselves in theiraccented English, and tried to find ways to commu-nicate more effectively, given the language barrier.One German team member, for example, declined ameeting, asking whether they could “just send meemails, and if we can discuss this problem throughemails because I’m not comfortable with speakingEnglish.” An informant in India explained: “Andthat was fine with us and that’s how we resolvedthe issue, just through emails.” In general, membersof this team recognized that there was a languagebarrier that was sometimes annoying, but treated itas a problem to resolve together.Overall, despite the strong faultline, this group

demonstrated moderate subgrouping, with sometensions between the locations around how thework was being done. There were also some frustra-tions around language issues, especially associatedwith code switching on the part of the Germans, butgenerally the regulation strategies used in this teamwere more constructive than those used in the HHteams, and the affective tone was more neutral.

Low/lowTeam LL1 was composed of 24 members, splitbetween Germany (8) and India (16). This team wasresponsible for technical solutions to ensure thatdata in different systems was compatible. The systemwas primarily in maintenance mode, with little newdevelopment, so much of the work entailed fixingbugs in the system. The team had been together for6–7 years, although it had undergone a massivereorganization 6 months prior to the start of ourstudy. LL1 had a relatively strong faultline, similar toMM1, with homogeneity in nationality at both theGerman and Indian locations. No German speakerswere located in India. Both locations included a mixof bilingual team members and those with pro-fessional fluency in English. Despite the strongfaultline, subgrouping in this team was minimal.The Indian team members described the Germanmembers as friendly and helpful. The German teammembers said that they were generally comfortablewith their Indian colleagues, reporting that theIndian team members were working long hours andweekends to make deadlines andmeet requirements.In the past there had been some tension as a resultof shoddy work by the Indians, described by aGerman informant as “a quick and dirty job,” but

these issues had been resolved with time. The Indianteam members said that their German team mem-bers were responsive, and they appreciated the struc-ture that they brought to the task and the fact thatthe Germans did not over-commit.Neither the German nor the Indian teammembers

thought that language asymmetries were a problem.The Indian team members acknowledged the asym-metries, but empathized with their German teammembers’ struggle to express themselves fully. Theysaid that, as long as it did not interfere with thework, they were comfortable with the Germansswitching to their native language. In fact, duringour observations, we noted that some of the com-munication among Indian team members was inTamil, since they were mostly from that region, andat other times they switched between English, Hindi,and Tamil. Although asymmetries in fluency werenot generally an issue in this team, the Indiansreferred to a past time when there were problemsbecause major parts of the code and documentationwere in German, but said that all documentationwas now in English, and the problems were resolved.Overall, this team mentioned language asymme-

tries, but members were sympathetic to their distantteam members’ experience and did not see languageas a problem for the team. Team members typicallytalked about themselves as an entire team acrosslocations and rarely highlighted location-based dif-ferences. Tensions were described as low on this team.In addition to the four teams described above,

there was one additional LL team (LL2) and oneteam rated as ML (ML1) (see Table 2 for details). LL2had patterns similar to LL1. ML1 had some languageissues, with the India-based team members some-times feeling neglected and excluded, but this teamalso reported being empathic, and had adapted theirresponses.

POWER CONTESTS AS AN ACTIVATORAs described in the examples above, strong faultlinessometimes resulted in subgrouping, but not always.Further analysis reveals that power dynamics mayhave operated as an activator, triggering sub-group dynamics in teams along their geographic-,nationality-, and language-based faultlines. We fol-low Magee and Galinsky (2008: 361) in defining“social power as asymmetric control over valuedresources in social relations.” The value of resourcesis subjectively determined, but if one person con-trols it and the other values it, more power accrues tothe individual who controls the resource. Power isdifferentiated from status in that “power is based in



resources, which belong to an actor, whereas statusexists entirely in the eyes of others” (Magee &Galinsky, 2008: 363–364). In the teams that westudied, we noticed that some had intense strugglesfor resources and power across locations, some werein transition because power was shifting from onelocation to the other, and yet others recognized andaccepted a power imbalance across locations thatthey anticipated continuing into the near future.Our analysis revealed a pattern in which teams thathad severe power contests also experienced the mostdivisive cross-location subgrouping, teams with shift-ing power reported less subgrouping, and teamsthat accepted the power imbalance reported virtuallyno subgrouping across locations, suggesting thatpower dynamics moderate the relationship betweenfaultlines and the presence of divisive subgroups. Weagain provide descriptions of four of the six teams,this time organized by type of power dynamic.

Power ContestsBoth teams HH1 and HH2 experienced power con-tests. The German and US teammembers recognizedthat those located at the German headquarters werein the location where the bulk of the resourcesexisted, where most decisions were made, and wheremost people and therefore most development activ-ity resided. As stated by a German in HH1, thislocated them with the “heart” and “brain” of thecompany, where most of the ideas were generated.He went on to say: “If you’re just kind of a fingersomewhere else, then it always takes time to under-stand what the brain is thinking and also what theheart is doing.” Members of HH1, both in Germanyand in the United States, acknowledged that the USlocation had its own source of power, includingbeing closer to the key market for software, to thecore marketing team, and to the industry analystswho were knowledgeable about the local high-technology industry. In addition, and in partbecause of the resources near the US site, the vicepresident for this group, a German, had relocated tothe United States, thus giving the US team membersmore access to him on a day-to-day basis. Membersat both locations in HH1 described the teams as“equal,” but also reported power contests. At bothlocations, team members mentioned a sense ofcompetition between locations. A team member inGermany, for example, described insecurities whena new topic was assigned to someone in the group.He said they would ask themselves, “Now a newtopic and people feel like … is it more importantthan my topic?” A US-based team member echoed

this sentiment, saying that there is always the“push” to put the work in Germany. Much of thecompetition revolved around being appreciated andreceiving recognition from the VP, and aroundquestions about whether or not they had his atten-tion. German team members, in particular, won-dered whether he understood and valued what theywere doing, given that he was spending more timewith the US team members. They also worried abouthis ability to represent them in critical activities inGermany, thus heightening their fear of losingground. US-based team members, however, also feltthat their position was threatened. They said that“ownership” was predominantly in Germany andthat Germany set the timeframe so that, for exam-ple, they would have to work holidays even thoughtheir German counterparts did not.HH2, in contrast, did not report feeling equal. US

team members considered themselves “outnum-bered” by the German team members (despite thefact that the numbers were fairly equal). As onereported, “the development team here is outnum-bered by the people, the colleagues in [German city]”and went on to say that their problems were rooted inthe “number issue” and the “language issue,” whichled to their “opinions or feedback” not being “equallytaken as the people in [German city].” The US-basedteam members described the dynamic as a “rivalry”with their Germany-based teammembers. During ourobservations, we saw high levels of frustration and apervasive sense of threat and competition at the US“outpost.” In meetings we observed, German teammembers spoke most and appeared to control theagenda. Although the German team members gener-ally expressed a value for the diversity in thought thatthe US team members brought to the team, and sawtheir contribution as necessary and distinctive, theyalso felt a sense of competition. One German devel-oper, for example, made the case that perhaps all ofthe work on this project should be moved to Ger-many, telling us: “One of the other guys from theother country is really good, there’s no questionabout that. You can learn from him. But there’s noquestion that similarly good guys are around here aswell.” The manager of the team who was based inGermany acknowledged the issues with power, tell-ing us during our observations that every time theUS team members developed something, it mightget rejected, because it did not fit with what theGerman team members had in mind. In one case,the German sub-team completed a task assignedto some of the US team members because theUnited States was seen as “too slow,” which further



contributed to worries of power loss experienced bythe US team members. At the same time, the man-ager worried that the German developers’ jobs wouldbe moved offshore, because they spent so much timeorganizing the projects and communicating withthe United States rather than coding.In general, the teams that were characterized by

power contests saw their situations as unstable, andwere constantly on guard against the loss of valuedresources to the other location.

Power ShiftsTwo of the teams in our study were experiencingshifts in power from the German to the Indian site.The MM1 team, for example, had recently movedmost of the work from Germany to India, althoughthe transition was not yet complete. There wasgeneral acknowledgment of the transition of workand of ownership for that work, including the place-ment of a team manager in India, although newdevelopment and product management remained inGermany. As stated by a German teammember, “It iscompletely changed. We give over the work and theresponsibility.” We still saw, however, some resis-tance in Germany and some impatience in India.The Germans were described as retaining seniority,experience, and product expertise on which theIndian team members were dependent. As a result, asrelayed by an Indian team member, “mostly theagenda is set by colleagues in [German city].” OneGerman team member spoke about how the Indianteam members could be overconfident, but afterspending timewith the Germans, came to understandthe gaps in their own knowledge. Another describedthe Indian team members as “competitive.” He saidthat, at the beginning, they had “a good ego …

actually they don’t have a clue at that point.” TheIndian teammembers countered by arguing that “theGermans are learning something from us … the waywe do analysis of a problem … the kind of creativepeople we have … and we are very adaptable.” At thepoint when we conducted our interviews, the deci-sion making resided in Germany, but it was alsoexpected to shift to India in a fewmonths. The Indianmanager said, “June is the deadline, where the solu-tion will shift and then another six months, comple-tely, it will shift here … so by next year end, we canexpect to make decisions rather than be a part of thedecision-making.”

Acceptance of Power ImbalanceTwo of the teams in our study clearly had a powerimbalance in which the primary base of power was

in Germany. In LL1, for example, team members inboth locations were clear that the standards, deci-sion making, directions, strategy, and most of theopportunities flowed from Germany to India. Asstated by one of the German team members: “Somost of the teams working in that area … are in[German city]… the decision-makers are in [Germancity] … and a lot of good interesting projects are in[German city]… and when someone decides to havea manager for a small … project lead, then it’s mostof the time in [German city].” Indian team membersagreed. They said, for example, that “the decisions[are]…more concentrated in [German city] and here[India] are the executors. That’s always the normalfunctionality.” Our observation was that althoughthe power imbalance was acknowledged, it was notcontested. In fact, the Indian team members seemedto rationalize the current situation as somethingnormal in the short term, while taking the long viewthat there would eventually be more growth oppor-tunities in India.

THEORETICAL MODELOur contribution resides, in particular, in articulat-ing how language asymmetries contribute to the usvs them subgroup dynamics often documented inglobal teams (e.g., Cramton & Hinds, 2005; Polzeret al., 2006). By providing evidence that languageasymmetries contribute to faultlines, that powerdynamics activate faultlines, and that some dimen-sions of faultlines, specifically language, can channeland embody intense subgroup power struggles, weextend theory on language asymmetries and sub-group dynamics in global teams. Our conceptualiza-tion is more dynamic than previous theory onsubgroup dynamics, suggesting a process that is self-reinforcing. Language asymmetries constituted anelement of a faultline (combined with location andnationality), but ultimately were reflected in actualbehaviors and clashes between locations. For theseteams, language asymmetries were not simply invi-sible dividing lines. Subgroup identification andlanguage asymmetries fueled one another in a rein-forcing process in which language was a visiblemarker that kept subgroup identifications alive,while divisive subgroup dynamics exacerbated theemotionally charged experience of language asym-metries. Our data suggest that the teams in our studywith more severe subgrouping also suffered from amore negative affective state that was intertwinedwith their response to language asymmetries. A teamaffective state arises from a combination of teamcomposition (e.g., individual affective tendencies)



and team context that is transferred through affectivetransfer processes (e.g., Barsade & Gibson, 2012). Thepatterns we observed suggest that language asymme-tries and subgrouping created a context that contrib-uted to and reinforced negative team affective states.We posit that language asymmetries become a con-duit for heightened emotions when subgroups arestrong and divisive, because constant interactionaggravates an already negative affective state. Unlikemore subtle cultural differences between locations,language is a day-to-day behavior that is constantlyaccessible and thus persistently salient.The theoretical model that emerged from our

analysis reflects the relationship between linguafranca asymmetries and subgroup dynamics andthe process that unfolded (Figure 1). Specifically,language asymmetries aligned with distance andnationality to create faultlines that were dormant insome teams and activated in others. Power contestswere associated with stronger and more divisivesubgrouping and a negative team affective state,thus suggesting an “activating condition” that trig-gered otherwise dormant faultlines. Teams that frac-tured into divisive subgroups were also the oneswhose dominant responses to language asymmetrieswere protective (e.g., situation selection and modifi-cation) or reactive (e.g., experience and delay regula-tion) rather than empathic (reappraisal). Finally,these emotional responses to language asymmetrieswere associated with behaviors, such as code switch-ing or exclusion, that further inflamed anxietiesabout power in those teams already burdened withpower contests. As depicted in Figure 1, this fueled aself-reinforcing cycle for the teams who had lan-guage-based faultlines and who were suffering frompower contests.

DISCUSSIONOur goal in this study was to understand the effect oflanguage asymmetries on subgroup dynamics inglobally distributed teams. We succeeded in describ-ing how language was an emotionally potent issuefor both German- and non-German-speaking teammembers at GlobalTech, and how language asym-metries contributed to a faultline that lay dormantin some teams, but was activated in others. With theincreasing prevalence of internationally distributedwork, it is critical to introduce theoretical andempirical perspectives that explicate the impact of alingua franca and bring it into the mainstream ofinternational business and organizational behaviorscholarship. Although intercultural sociolinguistics,intercultural pragmatics, and symbolic interaction-ists have long acknowledged language as a criticalcarrier of communication, little work has exploredhow differences in language backgrounds and lan-guage fluency affect team dynamics. Cossette (1998:1364) argues that language in organizations “mustnecessarily be examined in the light of the tonguespoken or written by the individuals concerned – inother words, the system of linguistic signs used bythem.”Our work helps to fill this gap and extend ourunderstanding of language in organizations, withparticular attention to its role in the us vs themdynamic so commonly observed in global teams(e.g., Cramton & Hinds, 2005; Polzer et al., 2006).Our findings suggest that language asymmetries area potent dimension of a faultline and play a role infueling us vs them subgroup dynamics. In our case,language aligned with geographic location andnationality to create powerful faultlines. AlthoughEarley and Mosakowski (2000) asserted that nation-ality is an umbrella trait to which managers must

Language asymmetriesaligned with ageographiclocation and nationality

Faultline Situation selection andmodification andexperience and displayregulation

vs

Empathic responses(e.g. reappraisal)

Subgroup dynamics

Dominant emotion regulation processes inresponse to language

asymmetries

Strong subgroup identification anddivisive subgroupdynamics

Activating condition:power contests

Negative teamaffective state

Figure 1 A process model of the relationship between language asymmetries and subgroup dynamics in global teams.



attend, we join others (e.g., Maloney & Zellmer-Bruhn, 2006) in challenging this assertion. Our datasuggest that language may be more salient andexplosive than nationality, which can go unnoticedwhen there is a strong occupational culture, or whendistance renders some of these differences less visible(Maloney & Zellmer-Bruhn, 2006).The link we propose between language asymme-

tries and subgroup dynamics extends previousresearch on language and identity. Bucholtz and Hall(2004: 382), for example, argue that language is “afundamental resource for identity production.” Iden-tity is an emergent product of linguistic practices andis, as a result, a social and cultural phenomenon thatemerges in interaction (Bucholtz & Hall, 2004, 2005).This approach to identity and language suggests thatteam identity, language, and context are intertwined.Language is one of the ways in which we categorizeothers and ourselves as belonging to particular groupsor subgroups (Harzing & Feely, 2008; also Giles &Johnson, 1981). In global teams, we conjecture,language asymmetries can enact and help to fortifylocal identities, particularly location-based subgroups.Depending on the context in which the interactiontakes place, identity work can be used to obscuredifferences to create a shared identity (as in LL1 andLL2), or can highlight differences, thus reinforcingan ingroup–outgroup dynamic between locations(e.g., Bucholtz & Hall, 2004; Harzing & Feely, 2008),as we saw in HH1 and HH2. In their meta-analysisof research on multicultural work groups, Stahl,Maznevski, Voigt, and Jonsen (2010) suggest thatsurface-level differences (e.g., overt demographiccharacteristics) vs deep-level differences (e.g., psycho-logical characteristics) are more salient, and thus maytrigger categorization. Language, as a surface-leveldifference, was a constantly accessible category, rein-forced during nearly every interaction, and so washighly salient, and may therefore have exacerbatedthe categorization process and interfered with com-munication effectiveness, as suggested by Stahl andcolleagues. We extend their work by providing morespecificity to which conditions heighten the salienceof (e.g., activate) these categories, since all of theteams in our study faced and spoke about languageasymmetries, but some were more dispassionate intheir response to it.We also contribute to theory about faultline acti-

vation. Despite general agreement about the impor-tance of an activation process to trigger subgroupdynamics, there exists a dearth of research on acti-vators and the role they play. A prominent feature ofthe original formulation of faultline theory was the

need for faultlines to be activated in order to result insubgrouping. As argued by Lau and Murnighan(1998: 333), when faultlines are not activated, theyare not only impotent, but “their strength maynaturally decrease over time” as teams in the mean-time develop a shared identity. Scholars buildingon the faultline model for global teams have recog-nized the importance of activating events and havemustered anecdotal evidence for their role. Earleyand Mosakowski (2000), for example, noticed intheir study that one of the teams that includedmembers from three separate national identities waswell integrated, but that another similarly composedteam perceived a cultural schism. They conjecturedthat the perception (or lack thereof) of heterogeneitywas pivotal to the level of integration. That is,national identity had to be salient for subgroupingto occur. Li and Hambrick (2005) also noted thatactivation needs to occur for faultlines to generatesubgrouping, but were not able to test it with thedata they collected. They did, however, postulatethat their participants “almost certainty attachedsome salience to their delegate status” (p. 808), andthat this salience triggered the faultline. Polzer et al.(2006) similarly were unable to test the role ofactivating events, but reported exploratory analysissuggesting that location differences became salientwhen teams divided the work by location. Our studydemonstrates that it is not necessarily the teamswith the strongest faultlines that fracture, but thatthe activator – in our case power contests – is crucialin determining whether or not they do.Specifically, our analysis shows that power dynam-

ics may occupy a central role in faultline activationand subgroup dynamics. Sivanathan, Pillutla, andMurnighan (2008) conceded that research on powerloss is rare. Loss of power, however, is a primaryreason cited for wars among nations and is known toincrease vigilance and attention to more powerfulparties (Sivanathan et al., 2008). Threats to power(e.g., anticipated power loss) are also known togenerate feelings of anxiety and shame (Keltner,Gruenfeld, & Anderson, 2003). Previous research onpower suggests that those who hold power enactstrategies to maintain the status quo (Magee &Galinsky, 2008). In our study, power contests andpower shifts contributed to heightened tensions bet-ween subgroups. Power dynamics were associatedwith discontent regarding control or perceived controlover valued resources such as access to informationand decision makers. We found that the relationshipbetween the activator (e.g., power contests) and sub-grouping was also more complex than previously



proposed. Power contests activated faultlines, butwere then also reinforced by the language-based sub-grouping behavior that ensued. Articulating this rein-forcing cycle makes a new contribution to theory onlanguage-based subgroup dynamics by explicatingone of the reasons why subgroups, once activated,may be difficult to dismantle.It is interesting to note that the power contests and

struggles in the teams we studied were somewhatreflected in issues and shifts in society at large, suchas the push to outsource work to lower-cost labormarkets. The external nature of these sources ofpower and power shifts made it especially hard forteam members to talk about and resolve power-related issues internally. In a provocative article,Chrobot-Mason, Ruderman, Weber, and Ernst (2009)suggest that triggers from society at large can activatefaultlines, such as when “a Palestinian boss gaveovertime pay to a Palestinian and denied it to aJordanian,” thus heightening subgroup identificationand tension. Brannen (2003) captures this idea ofexternal factors in the negotiation of culture in globalventures by differentiating field-based power (fromthe external environment) and arena-based power(from the organization). In our case, two of our teamssplit between the United States and Germany (HH1and HH2) were the ones with the most severe sub-grouping and most intense emotional reaction tolanguage asymmetries. It is likely that the powercontests between the United States and Germanyreflected field-based dynamics, such as the flat ordiminishing growth in jobs and opportunities inGermany and the United States, as well as arena-based dynamics, such as the location of headquartersin Germany, potentially contributing to the sense ofpower loss felt by these teams. Levine and Moreland(1998) argued that more attention to the settings thatgroups occupy was necessary. Our reflections suggestthat attending to the context in which teams areembedded is as important as ever for understandingsubgroup dynamics and is highly relevant to globalteams that span multiple contexts.As early as Lau and Murnighan’s (1998) introduc-

tion of the faultlines model, there was speculationthat power may play a role. Lau and Murnighan(1998: 335), for example, suggested that powermoderates the relationship between subgroupingand conflict, proposing that “Groups that split intosubgroups of comparable power are likely to experi-ence intense, overt conflict.” Surprisingly, theytalked only about power being associated with the sizeof the subgroups (e.g., equally sized subgroups), andignored other sources of power. Jehn and Bezrukova

(2010), in one of the few studies of faultline activators,allude to the role of power by examining entitlementpersonality. Finally, Carton and Cummings (2012)argue that asymmetries in power create “disparity-based” faultlines that result in resource-based sub-grouping, although it was outside the scope of theirwork to test this proposition. Evident from this sum-mary is agreement that power plays a role in subgroupdynamics, but the point at which this happens (e.g.,as a moderator between subgroups and conflict, as afaultline, or as an activator) remains ambiguous. Ourstudy is, as far as we know, one of the first directexaminations of the role of power in subgroupdynamics, and therefore helps to resolve these con-flicting propositions. We found that power asymme-tries are not, in fact, faultlines. In our teams that facedpower struggles, resources were not (perceived as)distributed asymmetrically. Instead, it was the sym-metry or move toward symmetry that seemed toignite power contests. We found that equally power-ful subgroups and subgroups experiencing powershifts competed for resources (Jehn & Bezrukova,2010; Magee & Galinsky, 2008), thus making salientan otherwise dormant faultline with language as adimension, and affirming power as an activator insubgroup dynamics.Although mostly unacknowledged, on rare occa-

sions informants linked power issues directly tolanguage, recognizing that the power differencesacross locations were intertwined with languageasymmetries. On HH2, for example, there was afeeling that resources and opportunities went toGerman speakers. One informant told us: “I’vealways felt that Germans do get maybe a little bettertreatment here … if you can speak German, thenyou’ll always have an advantage over someone whocan’t speak German.” When asked about thesebenefits, this team member said, “Well, if it came tolayoffs, I’ve always thought that they would keep theGerman person over the non-German-speaking per-son. And then I think a German has more opportu-nities to move up in this company.” As documentedin previous research, more power accrues to thoseworking in their native language (Harzing & Feely,2008), while those required to work in a secondlanguage can experience status loss, especially ifworking with native speakers of the lingua franca(Neeley, 2013). In our case, awareness of theseimbalances fueled frustrations, which were enactedin self-protective language-related behaviors.Although harder to support with evidence, we also

speculate that informants used language to reclaimpower. The strategy of talking to the boss about



violations to the English lingua franca policy, forexample, was a way that US teammembers reassertedtheir power in the situation. Similarly, code switchingmay have been used by German team members notonly to alleviate unwanted emotions, but also as away of regaining diminishing power. This is consis-tent with work in linguistics demonstrating that codeswitching can be used to exert power and createdesired social situations (e.g., Auer, 1984; Jan, 2003).Heller (1992) further argues that language choicereflects the relative value and symbolic distributionof resources across communities. Team members’ res-ponses to language asymmetries in polarized teamswere therefore complex, in that they reinforced nega-tive affect and subgrouping, but also invoked strate-gies, consciously or unconsciously, that kept powerdynamics alive.Our research also contributes to a better under-

standing of the emotional labor demanded of work-ers in globally distributed teams. Emotional laborhas been extensively studied in organizations(Elfenbein, 2007; Hochschild, 1979; Wharton,2009). The term “emotional labor” is commonlyused to refer to situations in which workers areexpected to regulate their emotions in service to anorganization as part of their jobs. To date, mostresearch on emotional labor has been conducted inservice contexts in which workers must regulatetheir emotional expression during service encoun-ters (e.g., Hochschild, 1979). Only a small number ofstudies consider the requirements for emotionallabor in professional jobs, and even fewer attend tothe emotional demands of interacting with colla-borators. Orzechowicz (2008: 143) describes these as“privileged emotion managers,” because they havemore autonomy, resources, and support than doservice workers. In her recent review, Wharton(2009: 161) calls for more research on emotion,emotional expression, and emotional labor in inter-active work, suggesting that “an investigation ofthese issues can shed light on a range of workplaceprocesses and dynamics.” She further challengesresearchers to use globalization as a stimulus toexamine “the intersections between local and globalinfluences on the construction and regulation ofemotion at work” (Wharton, 2009: 162). Answeringthis call, our study at GlobalTech suggests that lan-guage asymmetries and reliance on a lingua francacan demand emotional labor from globally distribu-ted workers. The informants we observed and talkedwith grappled daily with emotional issues aroundlanguage. In several of our interviews, informantsbecame tearful or cried when talking about their

feelings about language asymmetries. Some spokewith anger about their experience with the linguafranca at GlobalTech. We argue that this demandsemotional labor from these employees, and concludethat global work and language asymmetries exact anemotional toll on these professionals.Our team-level analysis suggests that Germany–US

teams were more likely to experience power con-tests, subgrouping, and frustration around languageasymmetries. There are several possible factors thatintensified these effects. First, the US informantsinhabited a relatively mono-lingual society as com-pared with India. US informants generally spoke asingle language, were more inclined to experiencefear of being talked about, and were less accustomedto situations in which code switching occurred. Onthe Indian teams, in contrast, we sawmore empathicresponses. Evidence suggests that bilingual speakersare perceived as being less threatening than nativeEnglish speakers in a mixed exchange (Neeley,2013), and that multilingual speakers have morecognitive empathy (Dewaele & Wei, 2012). The factthat the Indian location had no native Englishspeakers may have contributed to more empathicresponses and/or the perception on the part of theGermans that they would be more understanding,thus reducing anxieties around asymmetries. It isalso intriguing to note that both of the HH teamshad a German speaker in the non-German location(e.g., the United States). We posit that the presenceof a native German speaker may have made thelanguage asymmetries more salient. In some of ourobservations, for example, we noticed that Germanspeakers could be overheard talking on the phone inGerman with collaborators in Germany. This mayhave been a persistent reminder to the US-basedteam members that they did not speak the languageof headquarters, thus reinforcing their concernsabout power loss.A final consideration is the role of task and the

division of work between locations. Lau andMurnighan (1998) suggested that the task itselfcould behave as an activator of faultlines. In ourstudy, however, we see little evidence of this. InHH1, the task was not split between locations. Therewere multiple functions on the team, and teammembers in Germany and the United States workedtogether on all of them. In HH2, in contrast, the taskwas split, with the UI in the United States and back-end development work in Germany. In neither ofthese teams was the work itself location dependent(e.g., localizing for a specific market). The remainingteams were mixed, with the work somewhat



modularized between locations or, as with MM1, intransition. With regard to the division of work, weheard complaints that the work was too hard tocoordinate when it was not modularized, but alsocomplaints that when it was modularized, it made itdifficult to know what was going on, which inter-fered with later integration. Localization of theproduct did not seem to be salient for any of theseteams. Another consideration is the status associatedwith different types of tasks. In this high-technologycompany, development work was most prized.Maintenance work and product management, incontrast, afforded less status. HH1 may have feltmore insecure by virtue of being employed in ahigh-technology company but in a lower-status task,which could have exacerbated power contests. Thiswas not, however, the case with HH2, a team thatwas involved in high-status development work. Taskstatus is therefore not likely to be the primaryexplanation for the dynamics we observed.As with any study, ours has limitations. First,

ethnographies are well suited to build theory, butnot to test it. We have identified patterns andassociations, but cannot establish causality. Addi-tional research is therefore needed to validate therelationships we have identified. This will likelyrequire quantitative studies with significantly moreteams. Second, although we conducted some obser-vations, they were limited, and did not enablemicro-level analysis of conversations, particularlyaround code switching and responses to codeswitching. An important future research approachwould therefore be to implement sociolinguisticmethods of discourse and conversation analysis tobetter understand the actual patterns of code switch-ing that informants described in our study. Doing sowould enable the analysis of micro-level patterns inthe structure of the conversations that might con-tribute to interactants’ identity construction andmeaning-making (Auer, 2000; Bucholtz & Hall,2005; Lo, 1999; Sebba & Wootton, 1998). Moreover,this type of analysis would be required to revealdifferences in code-switching practices between sub-groups within a team. Third, all of our teams werespread across only two locations and included Ger-many. The role of language and power might lookdifferent in global teams with more locations, morelanguages, and more cultures, so additional researchwill be required to explore these dynamics. Fourth,we examined only globally distributed, not collocatedteams, and therefore geographic location was a keyfaultline dimension. We speculate that on collocated,multinational, multicultural teams, language may

align with other team characteristics, such as profes-sional training or organizational boundaries, to createsimilar dynamics. For example, if a collocated teamcomposed of Spanish-speaking Chilean and Spanishentrepreneurs from Company X and English-speak-ing American and British designers from Company Yis working together on a product, we might expectlanguage and company boundaries to align to createdynamics similar to those seen in the study we reporthere, particularly in the presence of power contestsbetween the two subgroups. Future research will beneeded to assess how subgroup dynamics play outwhen aligned with other faultline dimensions oncollocated teams with language asymmetries.Despite the limitations of this study, our findings

have important implications for managers chargedwith overseeing globally distributed projects. In parti-cular, managers need to understand the communica-tion challenges that people face when they have amix of native, bilingual, and professional competencein the lingua franca. Managers also need to under-stand the probable coping strategies that workers mayemploy, which may involve withdrawal, exclusion,and code switching for less confident English speak-ers, and leaving, asking for translations, or requestingthat English be used for more fluent and confidentEnglish speakers. Perhaps more importantly, identify-ing persistent activating forces, such as power con-tests, and helping to alleviate the sense of threat andfear that accompany these dynamics would go a longway to reduce subgroup dynamics and, in turn, createthe conditions for more empathic responses to lan-guage asymmetries.In sum, language asymmetries aligned with geo-

graphic distribution can contribute to a potent fault-line in global teams and, when activated by powercontests, draw emotions like a lightning rod, thusfueling a negative affective state that radiates tensionacross locations in these global teams, and triggers aself-reinforcing cycle of language-related behaviorsthat fuel an us vs them dynamic.

ACKNOWLEDGEMENTSThis material is based upon work supported by theNational Science Foundation under Grant No. IIS-0220098 and Grant No. IIS-0219754 to the first andthird authors, respectively. Any opinions, findings, andconclusions or recommendations expressed in thismaterial are those of the authors, and do not necessarilyreflect the views of the National Science Foundation.We thank Aditya Johri and Tine Koehler for their helpwith data collection, and Steven Shafer and HeatherAltman for their assistance with data analysis. We



appreciate the generous assistance of the informantsand the organization that participated in the study. Wealso would like to thank Steve Barley and members ofthe Harvard Business School Organizational Behaviorworkshop, especially Patricia Satterstrom, Robin Ely,David Thomas, and Lakshmi Ramarajan, for theirinsightful comments on earlier drafts.

NOTES1Although functional area or discipline has also been

shown to be an important faultline dimension, the

teams in our study were fairly homogeneous infunctional/disciplinary area, so functional area did notconstitute a faultline dimension.

2Ideally, we would have employed a quantitativemeasure of faultlines, but were unable to, becausemost existing measures were not robust to groups aslarge as those in our study (e.g., Bezrukova et al.,2009), and those that were tended to generate toomany subgroups to be meaningful.

3This is a pseudonym, as are all other informantnames reported in this paper.

REFERENCESAuer, P. 1984. Bilingual conversation. Amsterdam: John Benjamins.Auer, P. 2000. A conversation analytic approach to code-switching and transfer. In L. Wei (Ed), The bilingualism reader:166–187. London and New York: Routledge.

Barsade, S. 2002. The ripple effect: Emotional contagion and itsinfluence on group behavior. Administrative Science Quarterly,47(4): 644–675.

Barsade, S., & Gibson, D. 2012. Group affect: Its influence onindividual and group outcomes. Current Directions in Psycholo-gical Science, 21(2): 119–123.

Bezrukova, K., Jehn, K. A., Zanutto, E. L., & Thatcher, S. M. 2009.Do workgroup faultlines help or hurt? A moderated model offaultlines, team identification, and group performance. Organi-zation Science, 20(1): 35–50.

Bono, J. E., & Vey, M. A. 2005. Toward understanding emotionalmanagement at work: A quantitative review of emotional laborresearch. In C. E. Härtel, W. J. Zerbe, & N. M. Ashkanasy (Eds),Emotions in organizational behavior: 213–233. Mahwah, NJ:Lawrence Erlbaum Associates.

Brannen, M. 2003. What is culture and why does it matter?Current conceptions of culture from anthropology. InN. A. Boyacigiller, R. A. Goodman, & M. E. Phillips (Eds),Crossing cultures: Insights from master teachers: 20–37. NewYork: Routledge.

Bucholtz, M., & Hall, K. 2004. Language and identity. InA. Duranti (Ed), A companion to linguistic anthropology, Vol. 1.369–394. Oxford: Blackwell.

Bucholtz, M., & Hall, K. 2005. Identity and interaction: A socio-cultural linguistic approach. Discourse Studies, 7(4–5): 585–614.

Carton, A., & Cummings, J. 2012. A theory of subgroups in workteams. Academy of Management Journal, 37(3): 441–470.

Chrobot-Mason, D., Ruderman, M, Weber, T., & Ernst, C. 2009.The challenge of leading on unstable ground: Triggers thatactivate social identity faultlines. Human Relations, 62(11):1763–1794.

Cossette, P. 1998. The study of language in organizations:A symbolic interactionist stance. Human Relations, 51(11):1355–1377.

Cramton, C. D., & Hinds, P. 2005. Subgroup dynamics in inter-nationally distributed teams: Ethnocentrism or cross-nationallearning? Research in Organizational Behavior, 26: 231–263.

Davis, M. H. 1983. Measuring individual differences in empathy:Evidence for a multidimensional approach. Journal of Personalityand Social Psychology, 44(1): 113–126.

Dewaele, J. M., & Wei, L. 2012. Multilingualism, empathy, andmulticompetence. International Journal of Multilingualism, 9(4):352–366.

Earley, P. C., & Mosakowski, E. 2000. Creating hybrid teamcultures: An empirical test of transnational team functioning.Academy of Management Journal, 43(1): 26–49.

Elfenbein, H. A. 2007. Emotion in organizations: A review andtheoretical integration. Academy of Management Annals, 1(1):371–457.

Feely, A., & Harzing, A. W. 2003. Language management inmultinational companies. Cross-Cultural Management, 10(2):37–52.

Fixman, C. 1989. The foreign language needs of US-basedcorporations. National Foreign Language Center (NFLC) Occa-sional Papers Presentation. (also published in Annals, 511:25–46).

Giles, H., & Johnson, P. 1981. The role of language in ethnicgroup relations. In J. C. Turner, & H. Giles (Eds), Intergroupbehavior: 99–243. Oxford: Blackwell.

Gross, J. J. 1998. The emerging field of emotion regulation:An integrative review. Review of General Psychology, 2(3):271–299.

Gross, J., & Thompson, R. 2007. Emotion regulation: Conceptualfoundations. In J. J. Gross (Ed), Handbook of emotion regulation:3–24. New York: Guilford Press.

Harzing, A. W., & Feely, A. J. 2008. The language barrier and itsimplications for HQ–subsidiary relationships. Cross CulturalManagement: An International Journal, 15(1): 49–60.

Harzing, A. W., Köster, K., & Magner, U. 2011. Babel in business:The language barrier and its solutions in the HQ-subsidiaryrelationship. Journal of World Business, 46(3): 279–287.

Heller, M. 1992. The politics of codeswitching and languagechoice. Journal of Multilingual and Multicultural Development,13(1–2): 123–142.

Henderson, J. K. 2005. Language diversity in international man-agement teams. International Studies of Management and Orga-nization, 35(1): 66–80.

Hochschild, A. R. 1979. Emotion work, feeling rules, and socialstructure. American Journal of Sociology, 85(3): 551–575.

Interagency Language Roundtable (ILR). 2012. Skill level descrip-tions for competence in intercultural communication, http://www.govtilr.org/skills/competence.htm, accessed 27 June2013.

Jan, J. M. 2003. Code-switching for power wielding: Inter-genderdiscourse at the workplace. Multilingua, 22(1): 41–57.

Jehn, K., & Bezrukova, K. 2010. The faultline activation processand the effects of activated faultlines on coalition formation,conflict, and group outcomes. Organizational Behavior andHuman Decision Processes, 112(1): 24–42.

Keltner, D., Gruenfeld, D., & Anderson, C. 2003. Power, approach,and inhibition. Psychological Review, 110(2): 265–284.

Kingston, P. 1996. Bridging the language gap through interna-tional networking. In M. Berger (Ed), Cross-cultural team build-ing: Guidelines for effective communication and negotiation:58–70. London: McGraw-Hill.

Knapp, K. 2003. Approaching lingua franca communication. InK. Knapp, & C. Meierkord (Eds), Lingua franca communication:217–245. Frankfurt: P. Lang.

Lau, D., & Murnighan, J. K. 1998. Demographic diversity andfaultlines: The compositional dynamics of organizationalgroups. Academy of Management Review, 23(2): 325–340.



http://www.govtilr.org/skills/competence.htm

http://www.govtilr.org/skills/competence.htm

Levine, J. M., & Moreland, R. L. 1998. Small groups. InD. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds), The handbook ofsocial psychology: 415–469. New York: McGraw-Hill.

Li, J., & Hambrick, D. 2005. Factional groups: A new vantage ondemographic faultlines, conflict, and disintegration in workteams. Academy of Management Journal, 48(5): 794–813.

Lo, A. 1999. Codeswitching, speech community membership,and the construction of ethnic identity. Journal of Sociolinguis-tics, 3(4): 461–479.

Magee, J., & Galinsky, A. 2008. Social hierarchy: The self-reinforcing nature of power and status. Academy of Manage-ment Annals, 2(1): 351–398.

Maloney, M., & Zellmer-Bruhn, M. 2006. Building bridges,windows and cultures: Mediating mechanisms between teamheterogeneity and performance in global teams. ManagementInternational Review, 46(6): 697–720.

Marschan-Piekkari, R., Welch, R., & Welch, C. 2005. Handbook ofqualitative research methods for international business. Chelten-ham: Edward Elgar.

Metiu, A. 2006. Owning the code: Status closure in distributedgroups. Organization Science, 17(4): 418–435.

Mettler, S. 1984. Acculturation, communication apprehensionand language acquisition. Paper presented at the papers andreports on Pidgin and Creole languages, Vol. 16, Fifth BiennialConference, Kingston, Jamaica.

Miller, J. M. 2000. Language use, identity, and social interaction:Migrant students in Australia. Research on Language and SocialInteraction, 33(1): 69–100.

Neeley, T. 2013. Language matters: Status loss and achievedstatus distinctions in global organizations. Organization Science,24(2): 476–497.

Neeley, T., Hinds, P., & Cramton, C. 2012. The (un)hiddenturmoil of language in global collaboration. OrganizationalDynamics, 41(3): 236–244.

O’Leary, M. B., & Mortensen, M. 2010. Go (con)figure: Sub-groups, imbalance, and isolates in geographically dispersedteams. Organization Science, 21(1): 115–131.

Orzechowicz, D. 2008. Privileged emotion managers: The case ofactors. Social Psychology Quarterly, 71(2): 143–156.

Polzer, J., Crisp, C. B., Jarvenpaa, S., & Kim, J. 2006. Extendingthe faultline model to geographically dispersed teams: Howcollocated subgroups can impair group functioning. Academyof Management Journal, 49(4): 679–692.

Sebba, M., & Wootton, T. 1998. Sequential versus identity-related explanation in code-switching. In P. Auer (Ed), Code-switching in conversation: Language, interaction, and identity:262–289. London: Routledge.

Sivanathan, N., Pillutla, M., & Murnighan, J. K. 2008. Powergained, power lost.Organizational Behavior and Human DecisionProcesses, 105(2): 135–146.

Spradley, J. P. 1979. The ethnographic interview. New York: Holt,Rinehart and Winston.

Stahl, G., Maznevski, M., Voigt, A., & Jonsen, K. 2010. Unravelingthe effects of cultural diversity in teams: A meta-analysis of

research on multicultural work groups. Journal of InternationalBusiness Studies, 41(4): 690–709.

Strauss, A., & Corbin, J. 1998. Basics of qualitative research:Grounded theory procedures and techniques. Newbury Park, CA:Sage Publications.

Vaara, E., Tienari, J., Piekkari, R., & Säntti, R. 2005. Language andthe circuits of power in a merging multinational corporation.Journal of Management Studies, 42(3): 595–623.

Wharton, A. S. 2009. The sociology of emotional labor. AnnualReview of Sociology, 35: 147–165.

ABOUT THE AUTHORSPamela J Hinds is an Associate Professor in theDepartment of Management Science and Engineer-ing at Stanford University. She received her PhDin organizational science and management fromCarnegie Mellon University. She studies teams andthe dynamics of collaboration. She has conductedextensive research on the dynamics of globally dis-tributed work teams. She also examines the relation-ship between national culture and work practices.

Tsedal B Neeley is an Associate Professor of Organi-zational Behavior at Harvard Business School.She received her PhD in organization studies fromStanford University’s Management Science andEngineering Department. Her research focuseson the challenges that collaborators face whenattempting to coordinate work across national andlinguistic boundaries, with special emphasis inthe impact of language and communication onsocial dynamics.

Catherine Durnell Cramton is Associate Professorof Management in the School of Management,George Mason University. She received her PhDfrom Yale University. Her research interests includecollaboration and cross-cultural adaptation in globalteams, information sharing and relationship devel-opment in geographically distributed teams, andculture-specific coordination practices.

Accepted by Mary Yoko Brannen, Deputy Editor, 16 October 2013. This paper has been with the authors for two revisions.



language as a lightning rod: power contests, emotion ... · geographically dispersed team)...

Documents