Translanguaging Practices in Local Employees’ Negotiation to create Linguistic Space in a Multilingual Workplace Círculo de Lingüística Aplicada a la Comunicación

. This study explores the language practices and beliefs of local employees at a Shanghai-based subsidiary of a German multinational company. We conducted a seven-month ethnographic study and collected data from the company’s publicly accessible documents, meeting transcripts, semi-structured interviews with five employees, and ethnographic notes. Qualitative data analysis revealed that local employees frequently utilized translanguaging practices despite the company’s implicit assumption that English would be used as the common corporate language. Four major translanguaging practices were identified: key terms in English, bilingual label quest, cross-language recapping , and cross-language alternation . In addition, local employees perceived language as both a resource and an obstacle, often engaging in translanguaging practices to establish their own linguistic and communicative spaces, indicating that translanguaging is a complex multilingual practice influenced by internal and external factors, subject to social milieu, personal language competence, and beliefs. Ultimately, this study extends the notion of translanguaging and probes its analytical benefits for understanding fluid and discursive activities in multilingual workplaces and the sustainability of linguistic ecology and knowledge dissemination. fluid and discursive communicative activities in the business context, and strengthen sustainable linguistic ecologies for knowledge sharing in the multilingual translanguaging practices and beliefs, particularly in a non-Anglophone Asian context, in order to fully understand the complex linguistic ecologies of multilingual workplaces and the tension that multilingualism brings to these contexts. Specifically, this study explores the characteristics and environmental contexts of local employees’ translanguaging practices and their associated beliefs, as contextualized in a Chinese subsidiary of a German MNC. Two research questions were formulated to guide the inquiry:


Introduction
The increasing pace of globalization has encouraged most multinational companies (MNCs) to adopt a common corporate language to overcome language barriers encountered while attracting linguistically diverse employees (Sanden, 2020). Against this backdrop, English "seems to become the common corporate language by default" in MNCs (Piekkari et al., 2005: 333). However, the adoption of English as the working language frequently poses significant challenges to linguistically heterogeneous employees who may experience communicative difficulties using English. For example, employees complain that the mandatory use of English as a corporate language lowers their communicative efficiency, resulting in misinterpretation of messages and working delays, to name a few common difficulties (e.g., Fredriksson et al., 2006;Heikkilä & Smale, 2011;Sanden, 2020). It has also been acknowledged that MNCs are multilingual by nature, with English and other languages in use at the same time (Angouri, 2013(Angouri, , 2014Fredriksson et al., 2006;Kingsley, 2009Kingsley, , 2013. For example, despite the existence of an official working language in an MNC, employees and managers typically take a "what works" approach regarding language practice (Angouri, 2013). In this sense, employees in the multilingual workplace rely on their entire linguistic repertoires in their workplace communication, which is often characterized by complex and fluid linguistic practices (Angouri, 2013(Angouri, , 2014Kingsley, 2009Kingsley, , 2013Fredriksson et al., 2006).
Despite increasing research attention directed to language practices in multilingual workplaces such as MNCs, most studies are limited to European contexts (e.g., Barner-Rasmussen et al., 2014;Ehrenreich, 2010;Lauring, 2008), and little research has investigated how employees negotiate language practices in Asian contexts (with the exception of Fairbrother, 2018). In addition, most studies have focused on the challenges brought by corporate language policies, such as English as the common corporate language, but few have looked into how local employees cope with communicative challenges (Ehrenreich, 2010;Neeley, 2013;Sanden & Lønsmann, 2018). Therefore, we contextualized our study at a Chinese branch of a German MNC, focusing on the language practices and beliefs of local employees during business communication. This study aims to extend the concept of translanguaging (García & Li, 2014;Li, 2018) to the exploration of the fluid language practices in the multilingual workplace, and to probe the local employees' negotiation of linguistic space. The findings of the present study may shed light on fluid and discursive communicative activities in the business context, and strengthen sustainable linguistic ecologies for knowledge sharing in the multilingual workplace.

Translanguaging Practices and Language Beliefs in the Multilingual Workplace
As MNCs are constituted by heterogeneous groups of employees, they form a typical multilingual setting featuring linguistic diversity (Nekvapil & Nekula, 2006;Sanden, 2020). In a multilingual workplace, "communicative events are considerably more complex than the label of English as a lingua franca would suggest" (Nickerson, 2005: 371), and a multilingual workplace does not become monolingual simply because one language (often English) is taken as the working language (Fredriksson et al., 2006). Fredriksson et al. further showed that internal communication in multinational companies involves "crossing language boundaries and operating at the interface between several languages" (2006: 407), and "one strategy for managing language diversity may be non-management in the form of conscious ambiguity" (2006: 419). Fairbrother's (2018) study of the use of English in the Japanese branches of European MNCs also showed that some local Japanese employees regarded using English as face-threatening due to the unfamiliar communication norms. Piekkari, Oxelheim, and Randøy (2015) found that in a linguistically diverse environment, the introduction of English as a working language in Nordic boards caused some board members to become silent. Having reviewed all the findings regarding corporate language policies, Sanden (2020) argued that in some cases a common corporate language may create more problems than it solves.
As speakers of multiple languages (despite differences in language proficiencies), employees may form their own integrated and complex language systems in workplace communication (Angouri, 2013), and may thus be engaged in translanguaging practices (García & Li, 2014;Li, 2018). For example, Barner-Rasmussen and Aarnio's (2011) study of 61 MNCs in Finland showed that, even though the official business language was clearly English, employees freely switched between seven different languages. Steyaert, Ostendorp, and Gaibrois' (2011) study of two MNCs in Switzerland showed that employees changed their language according to their current space and situation, forming their own lingual landscape. Employees' language practices appear to be flexible-that is, employees and managers typically take a "what works" approach (Angouri, 2013) when they make decisions about language choice. Sanden and Lønsmann (2018) found that in three Scandinavian MNCs where English was held as the corporate common language, employees tended to use their discretion to choose when to diverge from using English when the corporate language policy was incompatible with everyday demands of their job. Some researchers have proposed utilizing translanguaging as an emerging paradigm to analyze language practices in multilingual communication in resistance to monolingual beliefs (e.g., Conteh & Meier, 2014;Paquet-Gauthier & Beaulieu, 2016), and we believe that this argument may also apply to the investigation of language practices in multilingual workplaces.
The term 'translanguaging' originated in language teaching (Williams, 1996), where it is used to explain a pedagogy characterized by converting language to realize teaching objectives in the input and output of bilingual education (Lewis, Jones, & Baker, 2012). More recently, the concept has been developed to refer to "the act performed by bilinguals of accessing different linguistic features or various modes of what are described as autonomous languages, in order to maximize communicative potential" (García, 2009: 140). It refers to the language practice whereby bilinguals and multilinguals spontaneously or consciously use multiple languages to complete communication (Cenoz & Gorter, 2017Creese & Blackledge, 2010;Hornberger & Link, 2012;. Overall, translanguaging has been proposed as a practical theory of language that offers a new lens to look at how bilingual or multilingual speakers draw on the entirety of their linguistic repertoire to make meaning in their own social context (Li, 2018).
Previous studies have classified translanguaging practices in detail, mainly according to four categories: (1) bilingual label quest-"the use of corresponding labels in one language to introduce a term or fixed expression in another language" (Martin, 2005: 83); (2) cross-language alternation-"explaining the meaning in both languages together" (Creese & Blackledge, 2010: 108); (3) cross-language recapping-repeated interpretation of the same meaning in two languages (Wang & Curdt-Christiansen, 2019); and (4) dual language substantiation-the concrete interpretation of a meaning with the concept of localization in another language (Wang & Curdt-Christiansen, 2019). Although these categories originate from the fields of classroom observation and language teaching, they are believed to be suitable to describe the language practices observed in the multilingual workplace.
Language beliefs, also known as language ideology, are people's deeply-held attitudes and assumptions about what constitutes appropriate language choices or practices in a community or communicative context (Spolsky, 2004). Language beliefs are arranged along multiple dimensions and can be very complex (Collins & Slembrouck, 2005;King, 2003;Kroskrity, 2000). In particular, language beliefs derive from and motivate language practices, and language practices can reflect an individual's ideas, evaluations, and judgments (Spolsky, 2004). In the multilingual workplace, employees' language choice is constructed according to their language beliefs towards their own and others' perceived competence in language with reference to native-speaker competence (Millar & Jensen, 2009). On the other hand, employees and managers in Angouri's (2013) study generally reported plurilingualism as an asset and an opportunity in daily business communication, which is associated with flexible language practices in the workplace. These findings suggest a link between language practices and language beliefs in the workplace, and it seems that employees' beliefs regarding translanguaging practices, if there are any, can then also be seen as manifestations of their attitudes towards multilingualism in the workplace.
Given that translanguaging provides a new perspective to investigate language practices in bilingual and multilingual settings, the present study chose to use it as the theoretical lens to investigate MNCs' employees' translanguaging practices and beliefs, particularly in a non-Anglophone Asian context, in order to fully understand the complex linguistic ecologies of multilingual workplaces and the tension that multilingualism brings to these contexts. Specifically, this study explores the characteristics and environmental contexts of local employees' translanguaging practices and their associated beliefs, as contextualized in a Chinese subsidiary of a German MNC. Two research questions were formulated to guide the inquiry: (1) What translanguaging practices do local employees conduct in a Chinese branch of a German multinational company in Shanghai, China?

Research Context
This study was situated in a local subsidiary located in Shanghai, China, anonymized as GIT, affiliated with a German MNC anonymized as HQ-GIT (Headquarters of GIT). HQ-GIT is a market leader in the business software industry with nearly 10,000 employees in 83 offices around the world, including Germany (the headquarters), the United States, China, India, South Korea, and Japan. According to GIT's official website, the company has over 1,500 employees. Over 75% of the staff are Chinese, with the rest mainly coming from Germany, the US and other Asian countries such as India, South Korea, Singapore, and Japan, using a variety of languages including Standard English, Indian English, Chinese, German, and other languages.
The present study focused on a small team anonymized as GIT-SF. With around 10 employees, GIT-SF serviced software development under a department with around 100 employees. All the members of GIT-SF are native Chinese speakers. They often work with teams from other departments to develop common projects, and also communicate with clients from around the world. Therefore, the language practices of GIT-SF are considered typical of the language practices of local employees at GIT.

Data Collection
To better address the complex linguistic ecologies of multilingual workplaces, an ethnographic approach is believed to be potentially fruitful to gain hands-on knowledge about language and communication practices (Angouri, 2013;Sanden, 2020). Following the principles of ethnographic research (Johnson, 2011), the first author immersed herself at GIT as a short-term intern and conducted a seven-month ethnographic study from July 2018 to January 2019. The employees and work she could reach were relatively junior, mostly entry-level local employees, and she could not access sensitive corporate information. The data she collected include documents, meeting transcripts, semi-structured interviews, and ethnographic notes.
(1) Documents This set of data included GIT's corporate management documents released to the public, such as recruitment language requirements and relevant regulations on the company's document translation.
(2) Meeting Transcripts To capture the actual translanguaging practices generated by the local employees at GIT, the first author recorded 98 meetings with a total time of 82 hours and 28 minutes, with an average of about 50 minutes per meeting. The main language used in the meetings was English (77.5%); the majority of speakers were Chinese (39.7%), while the audience was multilingual (67.3%). Based on meeting transcripts provided by GIT, we conducted a detailed analysis of the translanguaging practices during these meetings.
(3) Semi-structured Interviews To obtain more in-depth information about language beliefs towards translanguaging, we also conducted semistructured interviews with five colleagues on the team. Relevant information about these participants is as follows (all names are pseudonyms): Sarah and Daniel were fresh graduates with intermediate English proficiency . GIT was their first employer. Cathy was also a recent graduate, but she had worked as an intern at GIT for half a year before beginning her formal work, so she was more familiar with the company's language environment. Arianna had been at GIT for three years, before which she had worked for another top foreign software company in China. Tom was the most senior employee with the longest working experience among all the interviewees. He had been working for GIT for five years. Before that, he had worked for another smaller private enterprise.
(4) Ethnographic Notes Ethnographic notes include the first author's observation of GIT's linguistic environment in meetings, daily work, and other contexts, such as the nationality of the participants in a meeting, their language, English fluency, etc. While these notes were fragmentary and unsystematic, they represent an initial understanding of GIT's overall language choices, and also provided a preliminary foundation of understanding for the semi-structured interviews and further analysis.

Data Analysis
The study adopted a qualitative approach to data analysis. The authors first transcribed all the interviews in English, and then confirmed the interview transcripts with the interviewees to ensure their trustworthiness. For data analysis, we used QDA Miner 4.0 to code the transcripts of our documentary data and meeting transcripts following a thematic analysis procedure (Heller, 2011).
To address RQ1, we used the meeting transcripts to provide a description of the translanguaging practices typical in this multilingual workplace. The four types of translanguaging practices previously identified in the literature review were used to provide the coding scheme: key terms in English, bilingual label quest, cross-language recapping, and cross-language alternation. The first and second authors coded the meeting transcripts together to provide an overall description of the local employees' translanguaging practices. However, it should be noted that translanguaging practices were difficult to quantify in the data, as the meeting participants freely chose different languages to serve their communicative purposes. Therefore, we grouped the practices under the four categories qualitatively. After the meeting transcripts were coded, we sifted through the interview transcripts to examine the factors that shaped the employees' translanguaging practices. Documentary data were also consulted for interpretation of the interview data. Two types of factors-social milieu as an external factor, language beliefs and competence as internal factors-emerged as the major themes.

Research Ethics
This study strictly followed relevant requirements of ethnographic research ethics (Hult & Johnson, 2015: 34). First, the first author ensured that her identity and research purpose were open and transparent upon entry into the research context. All the materials collected were public data and materials released from GIT's public website. We strictly adhered to the principles of confidentiality and anonymity when dealing with internal or private information. Second, the researcher did not actively intervene, change, or guide the participants' language behaviors, but merely observed the language practices of her colleagues. Third, the researchers double-checked the data analysis results with the participants to ensure authenticity and transparency.

Different Types of Translanguaging Practices
This section aims to provide an overall description of the translanguaging practices used in this multilingual workplace. In the discussion below we group the practices under four categories. To better understand these practices, we also present some interview data to illustrate the employees' own rationalization of such linguistic practices.

Key Terms in English
For the purposes of the present study, key terms include technical terms, concepts, specific name expressions, etc. According to our analysis, employees would often use professional and technical terms without further explanation in Chinese. Notably, this type of translanguaging practice has not been explicitly identified in previous studies. The following excerpts serve as telling examples.
[Note: All excerpts are presented in their original language in italics. Translation is provided directly under the original text. The number in front of each utterance indicates speaking order. We refer to these numbers as Line 1, Line 2 (L1, L2), etc.] 1 现在讲一下什么是 machine learning，这个 slide 展示的是一个韦恩图。 Now, let's talk about machine learning. This slide shows a Wayne's diagram. 2 还有给大家开发的各种 big data 存储的中心应用，比如 spark, kafuka，来帮大家来处理些数据。从方 法来说，我们有了deep learning。 There are various big data storage center applications developed for us, such as spark and kafuka, to help us process these data. In terms of methods, we have deep learning.
In the above excerpts, the key terms machine learning, slides, big data, spark, and kafuka were expressed in English. Participants' direct borrowing of key terms from English was coded under the category of "Key terms in English" because these terms are all short words mostly originating from English-speaking countries. The following interview excerpts may help to illustrate the rationale behind this translanguaging behavior.
"The terms are all in English. I'm used to them. No one can translate them into Chinese. I don't know how to translate them." (Interview, Arianna, 05/08/2019) "If they [key terms] are expressed in English, everyone will understand them. They are professional words used for computers. Everyone in this field can understand them, right?" (Interview, Tom, 06/08/2019) The participants speculated that English was the source language of many key terms used in the industry. As "no one can translate them into Chinese," these technical terms embody key concepts that have yet to enter the Chinese linguistic repertoire. This echoes the widely-held sociolinguistic concern of "domain loss" due to the dominant position of English in business communication and high-tech industries (Hultgren, 2013). On the other hand, the translanguaging practice of expressing key terms in English may also imply that the English key terms have constructed the participants' integrated linguistic system (Cenoz & Gorter, 2017.
Interestingly, this translanguaging practice seems both spontaneous and unconscious. When we interviewed the five employees about this specific translanguaging practice, we found that they either could not recall exhibiting such behavior, or else they could not explain the reasons underlying their language use: Interviewer: When we have a meeting together, we often speak English, and suddenly people change to Chinese, or switch back and forth between Chinese and English. Have you ever met such a situation? Interviewee: No, I can't remember. Interviewer: For example, when talking about "Alipay" in English, it will be "zhifubao" rather than "Alipay." (zhifubao is the Chinese name of Alipay) Interviewee: Emm… Well, I don't know. I'm used to it. That's what everyone else says. Maybe sometimes I can't remember how to say it in another language. (Interview, Cathy, 24/07/2019) In our ethnographic observation, even though employees themselves seemed to have little awareness of using this particular type of translanguaging practice, it was the most frequently identified among all the translanguaging practices. Employees were unable to explain their language behavior, let alone recall the specific behavior, suggesting that their use of this translanguaging practice was most likely unconscious and unintentional. On one hand, this finding suggests that insiders in this workplace community have constructed a taken-for-granted language practice (Angouri, 2013), while on the other hand, it may also suggest that these key terms have been integrated into their daily communication and become part of their holistic linguistic repertoire. Employees simply said the words in the language that immediately came to mind. The "flexibility" observed in the employees' language practices (Angouri, 2013) further suggests that multilinguals care less about linguistic boundaries (Conteh & Meier, 2014; Paquet-Gauthier & Beaulieu, 2016) and may have developed higher multilingual awareness (Zheng, Lu, & Ren, 2020).

Bilingual Label Quest
Bilingual label quest refers to "expressing professional terms, concepts, specific name expression, etc. in one language and eliciting corresponding labels from another" (Martin, 2005: 83). This translanguaging practice appeared in a majority of meetings. Some typical examples are presented below: 1 AI可以做一些在自己狭小的领域内专注能做好的事情，比如说computer vision, computer vision就是视 觉模式识别。 AI can do some things that can be done in its own narrow field, such as computer vision, which is visual pattern recognition. 2 从方法来说，我们有了deep learning, deep learning说难也不难，深度学习，他就说把很多结构，做 多层次的重塑.
In terms of methods, we have deep learning, which is not hard to use, and deep learning, which means that we should reshape many structures and levels. (Conference 1.2 Overview of Artificial Intelligence Applications -11_30_2018) In the above excerpts, concepts such as computer vision and deep learning were first introduced in English and then explained in Chinese to ensure mutual understanding. This translanguaging practice was limited to simple definitions of terms. When it came to complex definitions, employees used the translanguaging practice of crosslanguage recapping, which will be discussed below.
The use of "bilingual label quest" may be due to the fact that these terms were novel to the meeting attendees, so the speakers found it crucial to ensure that everyone had an accurate understanding of them. In contrast to the first practice of key terms in English, this practice does not appear to be spontaneous. To some extent, the speakers purposefully emphasized the concept in both English and Chinese. They easily crossed the language barrier, utilizing their multilingual resources to achieve mutual understanding. English and Chinese (the local language) were both used in the specific interactional context, and for the speakers, monolingual language use, such as to use English alone, would have been unacceptable as it might have caused ambiguity (Cummins, 2007).

Cross-language Recapping
Cross-language recapping refers to first mentioning certain content in one language and then repeating that same content in another language (Wang & Curdt-Christiansen, 2019). It should be noted that GIT's implicit language preference required employees to prepare all slides, videos, and official work material in English. However, in practice employees generally found it difficult to prepare all materials in English in advance. As a result, if most of the participants in a meeting were Chinese, speakers tended to introduce content in English and switched to Chinese for detailed explanations. Speakers considered this practice convenient for both their audience and themselves, as illustrated in the following excerpt: 1 (Slide) Figure 1. Screenshot of a slide (Speech) 我在网上找了几个对于AI的定义，这个叫Hinton的人呢说，AI是人去思考自己的大脑是如何 运行之后所作出的一种人工物，AI的运行不需要明显进行编程，就是你不需要告诉电脑怎么运行而 电脑可以自己想出来。 (Slide: Model AI is modeled after ideas about how the brain works […] it learns to produce the right answers without you over programming. -Geoffrey Hinton) (Speech) I found several definitions of AI on the internet. The person named Hinton said that AI is an artifact made after people think about how their brain works. AI doesn't need to be programmed obviously, that is, you don't need to tell the computer how to run, but the computer can figure it out by itself. (Conference 1.2 Overview of Artificial Intelligence Applications -11_30_2018) In this excerpt, we can see that while the slide was created in English, the speaker recast the content in Chinese. Similar to bilingual label quests, speakers used this practice deliberately. In this excerpt, the speaker moved flexibly between languages to achieve mutual understanding, revealing that they may have consciously adopted the translanguaging practice.
Follow-up interviews shed some light on the motives behind such a practice: "Slides can be prepared in advance but speaking Chinese on site is easier for me. I just need to translate the key words to Chinese. Speaking all in English is difficult for me." (Arianna, 05/08/2019) As this interviewee further explained, polishing their English required excessive time and energy, and they believed that using a large quantity of professional English expressions would pose challenges to the on-site Chinese audience. However, insufficient linguistic proficiency was also a reason that drove the participants into this translanguaging practice.

Cross-language Alternation
The fourth type of translanguaging practice is cross-language alternation, or simultaneous code-switching, which occurs when speakers simultaneously adopt different languages in the meaning-making process. Speakers use each language to express distinct content; only by integrating the different languages can listeners understand the complete meaning (Creese & Blackledge, 2010: 108).

Good afternoon, ladies and gentlemen, on behalf of our team our presentation topic is financial robot. (向控 制台打手势) 下一页，okay.
Good afternoon ladies and gentlemen, on behalf of our team our presentation topic is financial robot. (gestures to the console) Next page, okay. (Conference 1.2 Overview of Artificial Intelligence Applications -11_30_2018) 2 网络层数深了参数数量就非常恐怖的，我们普通所用的比如说support machine, run the forest, 参数数 目大概几十个几百个, 或者是根据你的feature数目自己figure out出来的数目。 When the network layer number is deep, the number of parameters is terrible. For example, support machine, run the forest, the number of parameters is about dozens and hundreds, or figure out the number according to your feature number.

(Conference 57 Labs China Internal Incubation Program Graduation Ceremony -financial robot-11_12_2018)
In the first excerpt, the speaker used English to present and changed to Chinese for content irrelevant to the speech. This was a typical language practice during meetings, whereby employees discussed informal matters in Chinese and switched to English for important work-related topics. In the second excerpt the speaker frequently alternated between Chinese and English. As revealed in the interviews, participants might simply be speaking the language that comes to mind first: "I can't recall [the translanguaging practice]. I think I didn't change my language on purpose. I'm just uttering the words that come to my mind. For example, I might say '我们team在localization上做得蛮好的' (Our team is doing great in localization). I don't mean to use English words on purpose. I'm just used to it. That's how everyone talks." (Interview, Daniel, 04/08/2019).
Drawing on our previous analysis, we find that this cross-language conversion might indicate participants' high proficiency in both languages, enabling them to flexibly mobilize their linguistic resources to enhance mutual understanding. Alternatively, it may be a decision forced by insufficient linguistic proficiency. In addition, speakers use English for topics related to meetings and Chinese for unrelated content, demonstrating that these multilingual speakers rely on their own linguistic systems to flexibly convey relevant and less relevant messages (Angouri, 2013). Multilingual language use is inevitable in multinational corporations (Luo & Shenkar, 2006).
In summary, local employees as multilingual speakers are gradually building their own linguistic systems within GIT. This appropriation can be regarded as a struggle for "linguistic space" in workplaces where English is treated as the lingua franca for business (Johnson, 2009), subject to their language beliefs associated with translanguaging. In line with previous findings, we found a link between employees' language practices in the workplace and their language beliefs (Angouri, 2013;Millar & Jensen, 2009). In the next section, we will illustrate how these factors shape the employees' translanguaging practices as a process of negotiation to create a linguistic space.

External and Internal Factors Associated with Translanguaging
To address the second research question, the findings suggested that employees' translanguaging practices were shaped by two main types of factors: employees' social milieu as an external factor; and employees' language beliefs and language competency as internal factors. In this section, we will illustrate how employees' translanguaging practices develop in the linguistic social milieu, and how language beliefs and language competency shape their translanguaging practices.

Social Milieu as an External Factor
Although GIT did not have clear written rules on which language to use in any given situation, English was the most widely used language and became the de facto common corporate language. According to our interview data, employees were fully aware that the company expected its employees to use English for official communication.
"The slides put on the internet are also in English, perhaps because they are the company's property. In the future, people may need to review them, and only English can spread more widely." (Interview, Daniel, 04/08/2019) As shown above, slides and documents were related to the GIT's linguistic branding of their corporate image, and the notion that using English for external communication may allow the documents to be accessed and understood by a wider global audience. English was also a preferred language choice for internal communication. Among the total of 1441 email messages received by the first author during her time at GIT, 90% were written in English. It seems that the default choice of English as the common corporate language applies to both external and internal communication, conforming to previous literature on corporate English-only language policies (Sanden, 2020). The presence of the common corporate language therefore constituted a salient linguistic aspect of the social milieu in which the local employees were embedded.
Employees mostly complied with the corporate choice of English as the common language. For example, they all agreed that emails should be in English, as demonstrated by the following quote: "Emails are better in English, although there is no rule. They are usually CC'ed to many people, maybe even including foreigners." (Interview, Cathy, 24/07/2019) This finding indicates the participants' acquiescence to using English as part of the corporate language management, although there was no written rule in this regard. This is in line with previous studies that investigated the role of English as a shared language for information exchange in international business, thereby reducing potential misunderstandings or ineffective communication (Neeley, 2013;Marschan-Piekkari, Welch, & Welch 1999). Furthermore, it is likely that the very discourse of English as the default choice may itself construct the reality for the employees, to the extent that they perceived the language choice as "commonsensical" (Angouri, 2013: 569-570;Millar & Jensen, 2009). However, the translanguaging practices identified in the employees' daily communication did not always align with the English-monolingual assumption.

Language Beliefs and Language Competence as Internal Factors
For those employees who frequently utilized translanguaging strategies, language was regarded as a resource and translanguaging as an efficient method to express themselves completely and communicate smoothly. It is particularly noteworthy that the employees were tolerant of non-standard use in pronunciation, words, grammar, and so on. For example, "industrialization" was frequently misspelled as "industrization" in meeting transcripts, and one of the employees commented: "These problems don't matter, as long as we make the core issues clear." (Interview, Adrianna, 05/08/2019) It seems that translanguaging practices shift the employees' attention from adherence to linguistic standards and correctness in language use to meaning making (Wang & Curdt-Christiansen, 2019). In addition, the interview responses indicate a low level of language awareness. "I often mix Chinese and English together. I'm afraid that sometimes the expression is not complete. I don't think communication is a problem. Just talk more." (Interview, Tom, 06/08/2019) The above attitudes towards language are relatively positive, which might be connected to the employees' personal backgrounds. For example, because Tom had the longest working experience among the interviewees, having worked at another multinational enterprise before entering GIT, he was more comfortable in a multilingual environment and his English level was higher than that of his colleagues. As multilinguals draw on all the linguistic resources at their disposal to facilitate smooth communication (Hornberger & Link, 2012), translanguaging allows language users to draw on the entirety of their linguistic repertoire (Li, 2018) to create a linguistic space where different languages are functionally integrated to facilitate effective communication.
However, it was also found that employees who experienced communicative difficulties using English held different beliefs, as demonstrated in the following excerpts regarding these individuals' avoidance and fear of using English: "I know the meaning of the words in Chinese, but I can't remember how to say them in English. It's too hard. I'd better speak Chinese. I'm afraid I may make mistakes again. It's embarrassing." (Interview, Arianna, 05/08/2019) "Sometimes I just don't know how to express my ideas in English." (Interview, Sarah, 23/07/2019) For Sarah and Arianna, language seems to be a barrier and using English only could be a frustrating experience, leading to a loss of professional status and reduced self-esteem (see also Neeley, 2013). In this scenario, frequent translanguaging practices might be a strategy to avoid revealing their English incompetency.

Discussion
In this study we investigated local employees' language practices and beliefs in a multilingual workplace through the lens of translanguaging. Although the MNC in the study adopts English as its common corporate language, the local employees tended to mix languages to establish a linguistic space to accommodate their working needs. These employees exerted individual agency to draw on the entirety of their linguistic resources for effective communication (Cenoz & Gorter, 2020;. Sometimes even unconsciously, they were building their own language system and using their multilingual communicative repertoire to achieve mutual understanding (Creese & Blackledge, 2015;Hornberger & Link, 2012;Wang & Curdt-Christiansen, 2019). However, local employees' translanguaging could also be an involuntary choice to compensate for insufficient English skills.
Our findings suggest that, despite the presence of English as the common corporate language which constitutes the linguistic front of the social milieu, the employees demonstrated a variety of translanguaging practices. They crossed language boundaries and operated at the interface of different languages (Fredriksson et al., 2006), exercised their discretionary power (Sanden & Lønsmann, 2018), and engaged in flexible language use to meet their situated communicative needs (Angouri, 2013). The findings also demonstrate the employees' language beliefs compatible with translanguaging: boundaries between languages are blurred (Cenoz & Gorter, 2020), language is used as a fluid resource at one's disposal to make sense of the multilingual world (García & Li, 2014), and non-standard usage is tolerated. Translanguaging, then, can serve as a useful approach to gaining additional resources for communication in the multilingual workplace.
However, our findings also showed that translanguaging could also be a strategy for multilingual users to negotiate their linguistic space within the social milieu of English as the common corporate language (Angouri, 2013;Johnson, 2009). Translanguaging emerges as an alternative strategy to the English monolingual expectation for employees of relatively lower language competence, in order for them to establish a linguistic and communicative space (Johnson, 2009). In this regard, translanguaging can be seen as scaffolding to compensate for their defective communication and facilitate communication. From this perspective, the multilingual employees' language choice might not be as commonsensical as is normally assumed (Angouri, 2013;Millar & Jensen, 2009), but rather a constant negotiation process when the internal factors (i.e., language beliefs and language competence of the employees) interact with the external environment (i.e., English by default used as the common corporate language).
The findings of the present study extend our understanding of the spatiality and fluidity of translanguaging practices in the multilingual workplace. In a social milieu where a MNC's monolingual English assumption does not match the multilingual reality of the workplace, local employees' translanguaging practices construct a complex and fluid space to achieve communicative efficiency. In turn, diverse translanguaging practices expand the employees' linguistic space. On the other hand, local employees' translanguaging practices are also subject to complex language beliefs that simultaneously include language as resource and language as barrier. These complex and sometimes conflicting language beliefs in relation to translanguaging may affect the employees' language awareness with respect to their own translanguaging practices. Overall, the interaction of social environment with language proficiency and language beliefs may partially explain local employees' translanguaging practices.

Conclusion
In a seven-month ethnographic study situated in the context of a multilingual workplace, we identified four major translanguaging practices used by local employees and explored the rationales behind these translanguaging practices. Our results indicate that in a multilingual workplace, local employees' translanguaging practices are driven by the interaction between the external environment, where English is by default used as the common corporate language, and the internal factors of the employees' own language beliefs and language proficiencies. Our findings have also shown that many of the employees were not aware of their own translanguaging practices.
Echoing Wang and Curdt-Christiansen's (2019) argument in an educational context, an ideological reorientation towards flexible multilingualism may be a promising and favorable move in a multilingual workplace. In the context of MNCs like GIT where translanguaging frequently occurs, a multilingual ecology is conducive to the company's development by retaining the characteristics of local languages as well as the persistence of the unifying language. Policymakers in MNCs may find it beneficial to take full account of companies' historical and cultural backgrounds, local employees' language proficiencies, and their de facto language practices and language beliefs. They should also take a more proactive role in corporate language management so as to foster a sustainable multilingual ecology that stimulates employees' language awareness and self-esteem, which can ultimately promote a more sustainable working environment for knowledge dissemination as well as local employees' well-being.
However, it should be acknowledged that the present study was limited to a small team in a single multilingual workplace. Future studies should explore translanguaging in other business contexts and for other purposes; for example, it would be beneficial to investigate the multimodality and efficacy of translanguaging strategies to offer practical suggestions for employees who may encounter difficulties when practicing translanguaging.