INTRODUCTION Language has been described as the essence of human life (Gadamer, 2004): it produces rather than just transmits meaning. In the social sciences, greater awareness of the constitutive role played by lan- guage in society (Alvesson & Kärreman, 2000; Deetz, 2003; Iedema & Wodak, 1999) has seemingly not prompted researchers to engage in greater methodological reflection about how to approach cross- language studies (Temple, 2005); that is, studies in which linguistic boundaries are being crossed by the researcher and/or participants. Inattention to the methodological implications of cross-language research was observed early on by Brislin (1970), whose review of 80 articles on cross-cultural psychology found that translation issues were either not mentioned or were under-reported. There is evidence that this silence about cross-language methodo- logical decisions is not just confined to Brislin’s discipline of psychology, and that it persists even in more recent times. For example, Bradby (2002) points to ” [sociology’s lack of interest in language”. Temple and Young (2004: 163) assert that research on minority ethnic communities “is written without any reference to This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:53:43 UTC All use subject to https://about.jstor.org/terms 563 language issues”. Even anthropology, with its focus on in-depth fieldwork, has been accused of treating foreign language interviewing as “a taken for granted issue” (Winchatz, 2006: 84). In this paper, we exam- ine how scholars in international business (IB) have approached cross-language research. Our key research question is: “How do IB authors account for the translation decisions they make in their research?” This paper draws inspiration from Davis’s (1971) seminal insight that challenging assumptions is what makes academic study valuable. We adopt Alvesson and Sandberg’s (2011) “problematization methodology” to develop new theoretical insights into translation. This theoretical approach involves identification and critique of the taken-for-granted assumptions underlying a particular theory or domain; and then development of an “alternative assumption ground”. Our paper is structured in accordance with these steps. We commence by critically reviewing what, following Pym (2007), we term the “equivalence paradigm”, which has domi- nated existing literature on cross-language research methodology in IB. According to this approach, translation is the quest for conveying identical meanings. We then shift from the methodological discussion about equivalence to empirical practice, analysing the approach to translation taken in cross- language studies published over a 10-year period in four IB journals: International Business Review ( IBR ), Journal of International Business Studies (JIBS), Journal of World Business (JWB) and Management Interna- tional Review (MIR). Our evidence highlights that the technicist assumptions of the equivalence para- digm also underlie empirical studies in IB. Having uncovered the prevailing assumptions in methodological and empirical studies in IB, we then turn to an alternative approach that has emerged in translation studies. While we agree with the need for transparency and rigour when it comes to translation, we argue that this does not require a mechanical process of arriving at the most identical target-language text as possible. Instead, we propose the shift to a more contextualised approach based on theoretical developments in translation studies. We conclude our paper by outlining the implications of such a shift for conducting and reporting on cross- language studies. We argue that adopting a contex- tualised approach to translation would enable researchers to take greater account of the cross- border and cross-cultural differences that are, after all, central to the rationale for IB forming a separate field of enquiry. THE EQUIVALENCE PARADIGM: THE VI ^ Translation In cross-language international business research Agnieszka Chidlow et al 564 international marketing, back-translation “is still the primary method used to check translation accu- racy”. Their analysis of cross-language studies pub- lished in the Journal of International Marketing from 1993 to 2005 finds that 75% of papers report they used back-translation (compared with 62% in Schaffer and Riordan’s review of a different set of journals). The popularity of back-translation is per- haps not a surprise, given it is grounded in etic assumptions of a symmetrical correspondence bet- ween languages (Douglas & Craig, 2007). Yet the assumption that back-translation estab- lishes equivalence is questioned by the very author whose seminal paper on back-translation, Brislin (1970), is heavily cited in IB. In this paper, he warns that “a researcher cannot depend solely on the back- translation technique” (Brislin, 1970: 213) and advo- cates that it should be combined with other meth- ods. He outlines what he regards as the ideal of a multiple-method procedure, which comprises seven steps – only one of which is back-translation.2 In particular, his study finds that a pre-test is “neces- sary” (Brislin, 1970: 212) to eliminate translation errors, even after several rounds of back-translation. Yet Brislin’s advocacy of multi-method procedures has been largely overlooked by IB scholars. Nevertheless, there are indications that in the past decade, a more critical assessment of back-translation can be detected among commentators on survey methodology (Harkness, Villar, & Edwards, 2010), although this trend is rarely reflected in the IB literature (for exceptions, see Douglas & Craig, 2007; Usunier, 2011). Concerns about back-translation are in fact not new, but rather are being rediscovered. Sechrest, Fay, and Zaidi (1972) warned early on about the “paradox” of equivalence: the more equivalent the translation, the less likely it will be that cultural differences will be found (see also Sekaran, 1983). Another common criticism is that back-translation encourages “a spurious lexical equivalence” (Deutscher, 1973: 167); in other words, it may estab- lish that two words refer to the same object, but this does not necessarily convey the intended meaning of the original text (Peng et al., 1991; see also McGorry, 2000 for an example in an empirical study). Werner and Campbell (1973) suggest that in order for equiva- lent meaning and not just lexical equivalence between source- and target-language versions to be achieved, both versions may need to be modified in the process of translation. While Werner and Camp- bell still operate within the equivalence paradigm, they capture a crucial insight: loyalty to the source version may result in a text that is not easily Journal of International Business Studies comprehensible in the target language. Their version of back-translation – decentering – avoids source- text dominance by involving several iterations of (back) translation, with the original text as well as the translated version being successively modified. Instead of back-translation, Harkness et al. (2010: 128) assert that another technique canvassed by Brislin (1970, 1976) should be regarded as the “cur- rently most favoured” approach in survey methodol- ogy (see also Douglas & Craig, 2007): namely, com- mittee or “team translations” involving input from a diversity of individuals, including monolinguals whose assessment of the quality of the translation may well differ from that of bilinguals. The committee approach allows a much more thorough discussion of alternatives and different perspectives, in contrast to back-translation, which Harkness and Schoua- Glusberg (1998: 112) liken to using a metal detector: “It cannot identify what it picks up and, neither, unfortunately, can the monolingual researcher”. Ulti- mately, back-translation involves a subjective judge- ment as to whether two versions of a text are equivalent or not (Sechrest et al., 1972). Given this dependence on judgement, Brislin (1970) and his contemporaries emphasise the importance of careful selection of the people to conduct the translation. In the end, this is what assures the quality of the transla- tion, not the application of specific techniques and mies (e.g., Werner & Campbell, 1973). In the literature on qualitative cross-language methodology (which is currently largely based in nursing and sociology, with little influence on man- agement research), the equivalence paradigm has also been influential. Squires (2009) lists multiple criteria for evaluating the quality of cross-language qualita- tive research, with conceptual equivalence playing a prominent role in ensuring the trustworthiness of a study: “Maintaining the conceptual equivalence of what a participant said during an interview is … the most important part of mediating the methodological issues that arise from using translators” [our empha- sis]. Among qualitative researchers, there has also been interest in back-translation as “[t]he most com- mon and highly recommended procedure for trans- lating” (Chen & Boore, 2009: 235). Language dif- ferences are viewed as a technical problem, as a “barrier” to be overcome or at least reduced through the application of rigorous techniques; thus Squires (2009) explicitly uses the phrases “language barrier” and “methodological challenge”. To conclude, while equivalence is widely accepted as the goal of translation, the equivalence paradigm has struggled with the notion of what equivalence This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:53:43 UTC All use subject to https://about.jstor.org/terms Translation In cross-language International business research Agnieszka Chidlow et al 565 actually means and how it can be achieved. At its worst, the drive for equivalence can result in a narrow focus on the lexical similarity of texts rather than their meaning, lack of clarity as to how to proceed if there is insufficient correspondence bet- ween two languages, over-zealous fidelity to the source text, inattention to the inescapable subjectiv- ities involved in judging equivalence, over-reliance on back-translation and an underlying positivism which treats language as the neutral transmission of messages. Limitations and assumptions that accom- pany the equivalence paradigm have been acknowl- edged in the literature, although less so in IB, but critics have typically suggested alternative routes to equivalence (e.g., decentering, use of team transla- tions) rather than questioning it as a goal. We now turn to our analysis of empirical cross-language research in IB, in which we find even greater adher- ence to equivalence as the overriding objective. OUR ANALYTICAL APPROACH In order to examine how IB authors account for their translation decisions, we conducted an interpretive content analysis3 of published journal articles. Unlike previous treatments of the topic (Huit et al., 2008; Schaffer & Riordan, 2003), our analysis was primarily qualitative, allowing for fine-grained coding, a more holistic interpretation of the meaning of the text, the consideration of context and the possibility of emer- gent insights. It also had a broader scope, covering not only cross-cultural studies, but IB research in general.4 In addition, it was not restricted to quantitative studies but also included qualitative papers, which were excluded from previous reviews. Content analysis is often regarded as largely a quantitative technique based on the categorisation of various textual features and frequency counts of the resulting categories (Ahuvia, 2001). Yet there has long been recognition that a purely quantitative approach to content analysis is not only restrictive, but can potentially even be misleading as it does not capture the full contextual meaning of texts (Kracauer, 1952). Traditional content analysis distinguishes between “manifest” (literal, surface-level, direct) con- tent and “latent” (implicit, underlying, connotative) content, with the former regarded as more objective and more amenable to quantification (Berelson, 1971). More recently, there has been increasing recog- nition that drawing distinctions between the two types of content is misleading, given that the recep- tion to any content is necessarily interpretive in nature (Ahuvia, 2001). In the current study, we took advan- tage of the interpretive strengths of a more qualitative approach to content analysis (see also Welch, Piekkari, Plakoyiannaki, & Paavilainen-Mäntymäki, 2011). However, being interpretive is compatible with being systematic (Schreier, 2012), and in the remainder of this section we will detail the multi-stage process we went through to build our interpretation. A key aspect of every content analysis is the selection of the texts to analyse. In our analysis we employed a purposeful sampling approach in order to increase the information richness and diversity of the investigated texts. Our sample comprises the leading IB-specific journals – IBR, JIBS, JWB and MIR – all of which impose English as the language of dissemination (to use the term suggested by Tietze and Dick, 2009). We chose these four journals because they represent the most highly ranked and specialised IB outlets, thus providing us with an insight into disciplinary practices (DuBois & Reeb, 2000; Piekkari, Welch, & Paavilainen, 2009). Chron- ologically, we followed up the study by Schaffer and Riordan (2003), which examined research in the 1990s. In the time period under investigation (2000-2009), a total of 1440 articles was published in the four journals (Table 1). Our analysis of the four journals concentrated on empirical quantitative and qualitative articles. As far as quantitative articles are concerned, we analysed only survey-based studies, excluding empirical research based on secondary data or experiments. The reason for this focus is that survey research is not only the most popular form of data collection in IB, but it is also featured in the relevant methodolo- gical literature on cross-language research (e.g., Douglas & Craig, 2007). We commenced the analy- sis by categorising every empirical article (omitting editorials, commentaries and conceptual papers) in the period under investigation based on the type of data collection used by the authors (qualitative and quantitative).5 We then scanned each of these arti- cles to select for further analysis those that were cross-language in nature. Table 1 shows the results of this process: 401 cross-language studies, of which 334 were quantitative (72% of the total number of quantitative survey-based articles) and 67 qualitative (69% of the total number of qualitative articles).6 Having assembled our data set of cross-language empirical papers, we then employed two “cycles” of analysis (Saldafta, 2009), with a different emphasis in each cycle. The first iteration could be labelled “summative” (Hsieh & Shannon, 2005): we recorded and classified all the occurrences in the texts related to translation decisions. We scanned the entire con- tent of each journal article, although the methods Journal of International Business Studies This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:53:43 UTC All use subject to https://about.jstor.org/terms *** Translation In cross-language International business research Agnieszka Chidlow et al 566 Table 1 Categorisation of journal articles 2000-2009 Type of articles journal Total JIBS IBR IWB MIR Quantitative survey articles Quantitative survey articles published(%) 122(26%) 131(28%) 96(2,%) 1 1 3(24%) 462(83%) Quantitative survey cross-language articleS(%) 111(33%) 89(27%) 59(18%) 75(22%) 334(72%) Qualitative articles Qualitative articles published(%) 24(25%) 19(20%) 35(36%) 1 9(20%> 97(17%) Qualitative cross-language artides(%) 12(i8%) H(i6%) 29(43%) 15(22%) 67(40%) Total quantitative survey and qualitative articles 146(26%) 150(27%) 131(23%) 132(24%) 559(39%) Published articles per journal 449(31%) 382(27%) 289(21%) 320(22%) 1440 Note: JIBS = Journal of International Business Studies; IBR = International Business Review, JWB = journal of World Business; MIR = Management International Review. section of published articles received par- ticular attention because language issues are commonly regarded as a methodological concern (Squires, 2009). Unlike in a quantitative analysis, interpretive content analysis is exploratory and flex- ible: it is both theory and data driven in that the preliminary list of concepts is extended or even challenged in light of new findings (Schreier, 2012). We commenced with, but then supplemented, a coding frame that was based on recommendations of the extant methodological literature on quantita- tive and qualitative cross-language research (e.g., Brislin, 1970; Schaffer & Riordan, 2003; Squires, 2009). 7 In the first analytical cycle we constantly refined and expanded this initial coding frame to reflect the data set and resolved difficulties in its application to our textual material. At least two of us coded each article, independently at first and then jointly in order to discuss insights and multiple meanings associated with the coding process. During these multiple rounds of coding, we did not just record and count instances of keywords (e.g., back-translation, equivalence), but we also examined the entire textual segment in which the keywords appeared, allowing us to understand their meaning in context. This holistic treatment allowed us to look at what was absent and not just what was present in the texts, the way in which particular terms were used, the meanings that authors attached to them, the words associated with key terms (e.g., the way in which “ensure” was coupled with “equivalence”) and – critical to our problematisation approach – the assumptions underlying this word usage. The second cycle of analysis could be likened to what Gephart (1997) has termed “expansion analy- sis”. In this cycle, we linked the text segments that Journal of International Business Studies we had coded to the text’s broader linguistic, meth- odological and theoretical contexts. We linked the textual segments we had coded to the linguistic context of the study, as well as to the methodologi- cal literature that the authors cited. Having identi- fied the dominance of the equivalence paradigm in both the empirical and methodological literature, we began a theoretical journey to identify a contrasting paradigm, which took us to the area of translation studies. This stage required a high degree of theore- tical sensitivity (Ahuvia, 2001), as well as reflexive questioning of our own disciplinary assumptions. Consistent with our approach, we are not claiming our analysis is either objective or the only possible interpretation of our data. Instead, we are claiming it is a plausible interpretation that is based on a careful reading of the texts. Rather than inter-rater reliabil- ity, interpretive content analysis relies on collabora- tive coding (Ahuvia, 2001; Schreier, 2012): this allowed us to check each other’s coding for consis- tency, pose rival interpretations and develop an intersubjective understanding. We now turn to the key findings from this interpretive process. FINDINGS In this section, we present our findings about how authors report on their translation decisions (for the cross-journal comparison please see the Appendix). We start by discussing what emerged from our analy- sis as an important consideration when making translation decisions: the linguistic context of the study. We distinguish between four types of cross- language study, each of which presents a different context for translation. We then tum to the quanti- tative studies in our data set that mentioned transla- tion, in which equivalence was the leading concern. This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:53:43 UTC All use subject to https://about.jstor.org/terms 567 Authors typically frame equivalence as a straight- forward transfer of meaning that is attainable through the application of the correct technical procedures, foremost among them back-translation. We conclude the section by analysing the qualitative articles, find- ing that, while it was unusual to discuss translation at all, or refer to equivalence directly, authors regard language as a potential barrier and threat to accuracy. Translation Context: Four Types of Cross-Language Study The cross-language studies in our data set represent diverse linguistic (or translation) contexts. Not only do studies potentially involve multiple national lan- guages, but IB researchers are also investigating a variety of multilingual communities: the multina- tional enterprise (MNE) itself, expatriate communities and MBA classrooms. We differentiated four types of cross-language studies, based on their national scope (single or multiple-country) and the linguistic makeup of the population being studied (mono- or multi- lingual). Table 2 shows the frequency of each type: in descending order, they are Types 3, 1, 2 and 4. In addition, the table displays the extent to which articles in each type raised what we found to be the most common translation-related issues in quantita- tive and qualitative papers. During the course of coding, we also created a fifth “not clear” category for those articles that contained too little information on the linguistic context of the study to be able to categorise them to a cross-language type. The first type is a single-country study, conducted in an environment in which English is not an official or widely spoken language. For such a study, the devel- opment of a research instrument typically involves the researcher translating measures from English (if doing quantitative research) or translating an inter- view guide and interview transcripts (if doing qualita- tive research). The choice of language for such a study is usually not stated explicitly, but can be inferred to be the native language of the respondents. Other translation-related issues are not widely addressed either: back-translation in 31% of articles, pilot studies in 31% and the use of reviewers in 29% (see Table 2). Authors undertaking this type of study are reliant on their own language skills to translate and over- come language differences, unless they employ translators. Pla-Barber (2001) falls into the former camp: a Spanish author surveying Spanish compa- nies in Spain, he was able to use his own language skills to translate questionnaire items from English and the results back into English. In contrast, Jiang and Li (2008) used a translator rather than translating their survey instrument into German themselves. A qualitative example of a Type 1 study can be found in Gamble (2006: 332), who notes that: The author’s previous experience and facility in Chinese permitted interviews to be conducted on a one-to-one basis without a translator and for them to be transcribed directly by the author during the interview – tape-recording interviews would have inhibited interviewees’ readiness to speak openly. A notable exception to the use of the native language is Nielsen (2007) who, despite the fact that he is a Danish researcher surveying compatriots in Denmark, ran the questionnaire in English. He pro- vides the justification that “language was not a significant barrier to target respondents” (Nielsen, 2007: 347). The unstated assumption is that if possi- ble, it is preferable to conduct the survey in English. The second type consists of a single-country, cross- language study, in which the researcher is interview- ing or surveying a multilingual population, such as expatriate managers, who are to be found within the boundaries of the same country. Given the linguistic diversity of this kind of sample, inevitably some research participants will be communicating in a non-native language. In order to reduce this num- ber, one option in this type of study is to send out the survey in multiple languages. For example, Shi (2001: 191) provides a brief explanation of the decision to send out a survey targeting foreign invested enterprises in China in Chinese as well as English: “Many managers representing foreign par- ties are local Chinese because of the localization of managerial personnel … in recent years”. Another option is to use only English to survey these multilingual communities, with authors assum- ing that respondents had sufficient language ability to participate in the survey. For example, English was chosen by Barner-Rasmussen (2003), despite Finns comprising the majority of the expatriates they were surveying. The authors are aware of the challenge that being surveyed in a non-native language might pose. They note that personally administering the survey provided them with the opportunity to over- come any potential barriers to comprehension caused by the choice of language: “any terms or concepts respondents perceived as unclear during interviewing were explained to them in the language they felt most comfortable with (Finnish, Swedish or English)” (Barner-Rasmussen, 2003: 50). Barner-Rasmussen is, however, a rare exception in explicitly discussing the complexities of surveying a multilingual popula- tion; overall, articles in this type do not emphasise translation-related issues (see Table 2). journal of International Business Studies This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:53:43 UTC All use subject to https://about.jstor.org/terms * Translation In cross-language international business research Agnieszka Chidlow et al 568 1 9 ON O o 2 o o o; D O) (O D O) c JS I Vi O O O a; o_ * £ .2 •S2 B u S (O -C U PM Ji -Q |S ^ J J J J J J J . M m rs o oo On ■ – . N NO ^ VO ^ LO ‘ – p>» . >r N *”> vr’ ^ ^ ^ LO ‘ 2 p>» ro >r ^ ‘N N ^ ¡o -2 r> vr’ ^ iP s °n Tf’ o v s O f ^ c Si o c o ■e oí o w o w J -n £ PM «» c ï * ^ – r1″ *” * s çl3^(oo<Û c – °m C s S2. II l^o2.c£?-2* h mm c ü u ^ ^ L !ň ri r- ü u c D !ň c a> a» _Q .e ā. £ g £ ■M U <ü D !□ Z ąj ‘•u B ” Q_ CL) «3 OÌ ^ «3 § i C <■ Si h ^ o 3 c _c O = oi ■- r II „(NOOt-ONO^-vO ix tj- moo m ix <■ > ^ ^ ~ H I ” C £ ř 3 ^ 8. f ÎÎ ._!_ c 2 S “O 3 o C ou £ C -O w * ^ Č? 3 J O C ? £ £ £ £ £ £ £ o P* ^ *> ^ t; I Ü- ^ C S ou 1 Ã § o- “rs «S « “cMCOr-Tt JI §> kSl5 § =0^3 « “cMCOr-Tt M <* fi £ O I oo ^ gS-si à (O u _ KJ y-N v> +■» Ü _ £ g KJ ■go o s O) c o. jjj o ^ J2 “is csflt ř ii S”? ř ř ? “i °^ï *” J ° ¡2 •f I § S ! S S> g ^ CL) ON ON 00r- 7 TJ I CC VJ S S> CL) w g1 g AllrovOfMrvroui^ ^ aj t & (O O > aj c t en ^ oo iL ^ -9 g- ^ .E u O ~ 5 •£ > -âi «s E ^3 o. •Ś S .s¿ .Q. -O ^ ° ^ ö ž > h ^ -O c •S £ ¡5 h y ^ i § “2 “2 “Ss “S à -e “S = 2” ‘I “g i “g J “g J S” =2 £ -o .1 ‘S ļj Jt* “g J I I 1i¡e.!¡-&i¡!|> l|.l¡.i |¡t§K§i 5 •S “i I -g 5 ^ O