Incorporating translation in qualitative studies: Two case studies in education
Authors’ names:
Agustian Sutrisno, Faculty of Education, Queensland University of Technology, Brisbane, Australia
Nga Thanh Nguyen, Faculty of Education, Queensland University of Technology, Brisbane, Australia
Donna Tangen, Faculty of Education, Queensland University of Technology, Brisbane, Australia
Correspondence datails: , Room 354 B Block, Kelvin Grove campus, Faculty of Education, Queensland University of Technology, Brisbane, Australia
Abstract:
Cross-language qualitative research in education continues to increase. However, there has been inadequate discussion in the literature concerning the translation process that ensures research trustworthiness applicable for bilingual researchers. Informed by Squires’ (2009) evaluation criteria for qualitative data translation, this paper compares two different procedures for incorporating translation in education qualitative research to provide a clear depiction of the complexities involved in translating qualitative data and the strengths and weaknesses of each procedure. To maintain the trustworthiness of the qualitative research, it is necessary to minimise translation errors, provide detailed accounts of the translation process, involve more than one translator, and remain open for scrutiny from those seeking to access the translation process. Taking into account the resource constraints often faced by novice bilingual researchers, this paper provides some strategies that can be employed in similar contexts.
Introduction
As the world becomes more globalised, research often crosses national and linguistic borders (Guest, Namey, & Mitchell, 2013). There are a growing number of researchers with research locus outside their country of origin. Included in this category are a significant number of international research students studying in countries where English is the dominant language and culture (UNESO Institute for Statistics, 2011). In these situations, there are research areas that demand the use of more than one language. For example, the research proposal may be written in English but the data were gathered in another country such as Indonesia or Vietnam using the local language. Translation needs to be employed in such research but the question then becomes: how efficiently is the translation of data done so as not to compromise the research?
At the onset of this paper, it is necessary to introduce some key terms used. First, cross-language research refers to research involving a translator at any stage during the research process (Temple, 2002). This cross-language research may be done by mono-lingual or bilingual researchers. Mono-lingual researchers do not have working knowledge regarding one of the languages used in cross-language research; whereas bilingual researchers are proficient in two languages used in the research (Liamputtong, 2010). Another key term, linguistic equivalence refers to the similarity between linguistic expressions in one language and their translation in another (Loos et al., 2004).
In general, satisfactory discussion about translation procedures employed in social science research is quite limited (Douglas & Craig, 2007; Liamputtong, 2010). In particular, cross-language qualitative studies are often not accompanied by sufficient explanation of the translation procedure employed to demonstrate the trustworthiness of the research findings (Fryer et al., 2012; Larkin, de Casterlé, & Schotsman 2007; Wong & Poon, 2010). Lincoln and Guba (1985) view trustworthiness as the degree of rigour in qualitative studies. In cross-language qualitative studies, trustworthiness does not only concern the research process and findings but also the translation procedures and the translation results upon which the final research findings are based. Some studies utilise complex translation processes and the involvement of numerous translators and translation reviewers, which are generally beyond the resources available for research students (Chen & Boore, 2009; Regmi, Naidoo, & Pilkington, 2010). It could be suggested that the reliance on such a complex and multi-partied translation process, as described in the literature, is necessary because many of these studies were written by mono-lingual researchers who depend on third-party translators to understand data gathered in a language and cultural context foreign to them (Squires, 2009; Croot, Lees, & Grant, 2011). In contrast, there is limited literature that delves into the issues faced by bilingual qualitative researchers.
In light of the complex issues mentioned above, this paper attempts to compare specific translation procedures employed in two case studies and how the resource limitations in each study have been addressed from the perspective of bilingual researchers. Lessons learned from these case studies may be valuable for other bilingual researchers facing similar circumstances in clarifying their translation procedures and demonstrating trustworthiness in cross-language research. The two case studies described in this paper involve the translation from: 1) Indonesian to English, and 2) Vietnamese to English and the focus is on written translation, rather than oral interpretation.
Translation equivalence
Appreciating how linguists have understood the different conceptualisations of equivalence is informative for the two case studies in determining which equivalence conceptualisation to be pursued in the translation process. Equivalence has been supported and rejected by different factions in linguistics and translation studies (Kenny, 2009). Those who reject equivalence view that it is irrelevant and blocks the progress and creativity in translation as it is difficult to determine what constitutes equivalence between languages. Others view that without having some set criteria regarding equivalence, it is impossible to decide which text is a translation and which one is not (Brislin, 1970; Kenny, 2009). Baker (2011) contends that the concept of equivalence can be accepted for the sake of convenience in discussing translation. While some sort of equivalence can be obtained, there are so many language and culture-specific factors that obtaining absolute equivalence is unattainable (Baker, 2011; Kenny, 2009).
The difficulties in achieving absolute equivalence can be seen upon examining the building blocks of a language (Kirkpatrick & van Teijlingen, 2009; Larkin et al., 2007). Often, the most basic way of determining equivalence is by examining the correspondent words in the source language and the target language—lexical equivalence (Baker, 2011). For example, the word rice in English is equivalent to the word padi in Indonesian or cơm in Vietnamese. However, rice can also be translated as gabah, beras, and nasi or lúa, thóc, and gạo in Indonesian and Vietnamese respectively, depending on the stage of its development as a plant and the degree of its preparation as a dish. A word in the source language may have more than one equivalent in the target language and vice versa. There are also cases of no equivalence (Larkin et al., 2007; Squires, 2008). Following the previous example of nouns related to food, the Australian cake lamington does not have any proper equivalence in Indonesian and Vietnamese. In addition, words do not make sense without the grammar that binds them together into sentences. In this sense grammar is a ‘straitjacket’ that locks a language into a particular way of thinking and of expressing ideas that in many cases do not translate into another language (Baker, 2011). For instance, with tense-free languages such as Indonesian or Vietnamese, translating accurately into English is a challenge. Unless there are explicit temporal adverbs, it is impossible to tell whether an Indonesian or Vietnamese sentence is in the present, past or future tense. As such, aspiring to develop some set criteria to evaluate equivalence and establish trustworthiness of translation results based on the equivalence of words or sentences is a simplistic idea that does not correspond to the complexity of language systems as observed by linguists (Liamputtong, 2010; Temple & Young, 2004). For these and many other reasons, lexical equivalence has been abandoned and conceptual equivalence has acquired more support in cross-language research (Larkin et al., 2007; Squires, 2008).
Conceptual equivalence can be seen as the comparability of concepts or ideas between two languages, rather than the exact similarity of lexical meanings across languages (Birbilli, 2000; Neuman, 2011). For example, some nuances of meaning in the source language may be altered when conveyed in the target language (Baker, 2011). Although conceptual equivalence is striven for, it is still necessary to consider how information is structured in one particular language, how speakers of the language maintain coherence of ideas and cohesion of statements, and how the language users manipulate expressions and utterances to achieve humour, conceal feelings and assert opinions. Indeed, it is impossible to list all the variables involved in the process of translating a text (Baker, 2011). Through the notion of conceptual equivalence translators may subjectively determine the best translation result they want as there is no absolute equivalence to pursue. However, this approach may yield translation results that are unchecked and purely subjective, jeopardising the trustworthiness of the entire research (Liamputtong, 2010).
Rather than pursuing lexical and conceptual equivalences, some linguists propose dynamic equivalence as a more acceptable goal in translating a text (Constantinescu, 2010; Nida, 1969). The emphasis on reproducing the message from the source language to the target language in the most natural manner is called dynamic equivalence (Constantinescu, 2010; Nida, 1969). Arguably it has been the most prominent way of understanding equivalence among translation theorists (Munday, 2008). Accuracy of the translation is then evaluated by virtue of examining the target language users’ ease of understanding and acceptance of the translation text (Constantinescu, 2010; Nida, 1969). Dynamic equivalence seemingly puts many translation theorists in general agreement that the trustworthiness of translation is not seen as universal and totally neutral. According to Baker (2011), “It is in fact virtually impossible, except in extreme cases, to draw a line between what counts as a good translation and what counts as a bad one. Every translation has points of strengths and points of weakness, and every translation is open to improvement” (p. 6). Translation can be seen as a dialogue between the original texts in the source language and the translated texts in the target language, mediated by the translator, which results in a co-dependence between the two texts (Baker, 2011). However, adherence to dynamic equivalence may shift the focus from the particular nuances contained in the source language to what is deemed understandable in the target language (Croot, Lees, & Grant, 2011). Hence this approach may not be preferred by researchers who want to emphasise the perspective of the source language users.
From the above discussion, conceptual and dynamic equivalence have relevance for the two case studies. For example, conceptual equivalence was deemed more acceptable for the first case study. Provided it is combined with appropriate translation procedures, the subjectivity issue can be reduced. Dynamic equivalence, used in the second case study can be useful when the research agenda is to produce translation results that can be fully grasped by the target language users. For example, as research students who need guidance for the data analysis process, there may be a need to provide the English mono-lingual supervisory team translated data that can be fully grasped by English speakers. Implementation of these two concepts in the case studies are explored further later in this paper. As the more theoretical issues in translation have been discussed, the focus now turns to the practical translation procedures.
Common translation procedures in qualitative studies
There are three common procedures of translation that can be employed in qualitative studies: single translation, back translation and parallel translation (Liamputtong, 2010; Neuman, 2011). These procedures are examined in light of the previous discussion on equivalence, to identify a procedure suitable for the two studies and the translation limitations faced by the authors of this paper.
Single translation is perhaps the most straightforward procedure where the data is translated from the source to the target language (Neuman, 2011). It is often the least complicated and the quickest procedure in finalising the result. Due to its straightforward nature, the single translation procedure is also perhaps the weakest in ensuring transparency and trustworthiness since single translation only relies on the skills and interpretation of one translator, without any comparison process and input from others (McGorry, 2000). To improve the trustworthiness of the single translation result, the dynamic equivalence approach perhaps can be adopted by bilingual researchers undertaking single translation procedure. This requires the bilingual researcher to corroborate the translation results with one or more native speakers of the target language to produce results easily understood by the target language users, hence reducing personal bias and increasing transparency in the translation process.
Back translation procedure is widely used in social science research because of its potential to minimise inaccuracies in the translation with a result that strives for equivalence across languages (Liamputtong, 2010; Lopez et al., 2008). In this procedure, two translators are employed. One translates from the source to the target language. The other translator translates back the data from the target to the source language without knowing the original source language version (the back translation) (Brislin, 1970). The back-translation version is then compared with the original version in the source language. The comparability between the back-translation version and the original version is considered to indicate translation accuracy (Douglas & Craig, 2007). Compared to the single translation procedure, back-translation provides more opportunities for filtering translation errors as there is more than one translator involved. Another advantage is the researcher does not have to be a bilingual to participate in examining the translation accuracy (Chen & Boore, 2009; Jones et al., 2011). Nevertheless, the use of back-translation for a whole data set tends to be avoided given its laborious and lengthy process (Chen & Boore, 2009; Guest, Namey, & Mitchell, 2013). In addition, as stated earlier, a word-to-word equivalence between two languages does not always exist and examining conceptual equivalence between the texts can be challenging. However, there are ways that back translation can be used by bilingual researchers and this is described below.
First, the bilingual researchers can assign the translation to two third-party translators. The bilingual researcher can then take the role as the evaluator of the back-translation results (Guest, Namey, & Mitchell, 2013). This may give the researchers a sense of involvement in the translation process and allow them to understand the complexities of the translation process. Evaluating the back-translation results may also improve the accuracy, providing more pairs of eyes to re-check and triangulate the results.
Alternatively, bilingual researchers can take part actively in the back-translation process as the first translator. The back-translation can then be undertaken by a third-party translator and the back-translation results can subsequently be compared with the original texts in the source language. If there are inconsistencies, the bilingual researchers can revise the translation into the target language. This scenario can be more financially viable compared to engaging two third-party translators.