In the two previous instalments to the Researching Multilingually series, I asked a number of hypothetical questions that researchers in multilingual settings might face, and discussed some challenges involved in obtaining informed consent. This post moves the discussion forward, by looking into questions of data generation. Once again, I will approach this topic by drawing on the experience of my own PhD research, a case study set in a language school in Greece, where English was used as a working language.

When thinking of how to elicit data from the teachers and students in the language school, I originally planned to use English as my default language. This was in line with the school’s monolingual policy, which encouraged the use of English in the premises at all times, for pedagogical reasons (and, I suppose, for reasons of prestige as well). I tacitly acknowledged that some students might struggle with expressing complex concepts in English, and I expected that they would perhaps resort to using Greek. However, my overall attitude towards Greek was one of tolerance rather than encouragement. I certainly had not imagined any problems with using English when interviewing the teachers. I was wrong.

Dealing with non-standard forms

The first problem with my monolingual data generation strategy was the frequency of non-standard data in the interviews. While transcribing the interviews, I was surprised to notice that the participants, who were non-native speakers of English tended to use non-standard forms, false-starts, repetitions and other language infelicities rather frequently. Here’s an extract from one of the interviews, where a young, rather inexperienced and very nervous teacher described how she taught grammar:

…and then ask my students to give me examples. For example (2 sec) I when I teached the passive voice, I gave them the rule, and I told them ‘try to give me an example’. But this didn’t work with everybody because they feel the pressure that ‘I have to use the foreign language’ and they couldn’t do this.

While occasional slips such as *teached are to be expected in oral discourse, especially in when a speaker is stressed, I had not realised how frequent they might be. Moreover, they created two serious ethical problems:

• Firstly, I felt that when I eventually shared the transcripts with the participants, such deviations from ‘correct’ English would likely undermine their confidence in the ability to use the language. Many people report that the first time they hear their recorded voice, they feel uncomfortable with how they sound. I was very apprehensive that the same thing would happen when I confronted teachers with a record of their oral discourse, especially since their professional identity depended on them using English ‘correctly’.
• Furthermore, I was concerned that if the transcripts were to be made more broadly available, that could result in unfair judgements being made about the teachers’ competence, or the quality of second language education provided in the language school. It wasn’t unimaginable that a professional competitor might use my data to disparage the language school, and it was also quite plausible that if an unfavourable reputation was created, however unjustly, the teachers would find find it difficult to find employment elsewhere in the future.

I eventually solved that problem by creative improvisation: I began using several versions of interim transcripts, with different degrees of detail and editing, for purposes such as validation, analysis and dissemination. But I decided that in the next rounds of interviews, I would have to be more flexible with language use.

Dealing with nervousness

Another problem associated with the monolingual data generation strategy was that it tended to make interactions too awkward. By repeatedly listening to the recordings, I began to notice that a number of participants sounded distinctly nervous. In part, I think that this was due to a perception of power differentials between me (an ‘expert’) and the teacher participants, who may have felt that their professional practice was under scrutiny. However, it seems very plausible that these feelings were exacerbated by the fact that we were using English, as I was thought to be a more fluent speaker than they were.

To confirm this tentative hypothesis, I asked some teachers for feedback on the experience of being interviewed in English. Perhaps unsurprisingly, none admitted having experienced any discomfort. However, a number of participants suggested that the opportunity to practice English in the context of a sophisticated conversation was excellent practice for them, which (I think) suggests that the interviews did take them outside their comfort zone.

A revised data generation strategy

To mitigate these problems in subsequent interviews, I began to actively encourage the use of Modern Greek. In practical terms this meant that I would begin a pre-interview phase in Greek, and before starting the main interview I would ask the participating teacher if she would like us to speak in Greek or in English. Here’s an extract, showing how language choice was negotiated:

Unexpected findings

These interviews provided me with a number of unexpected insights. Firstly, I noticed that the teachers tended to code-switch very frequently in the Greek interviews, especially when using professional terminology (Figure 1).

Although some sources on Greek ELT will argue otherwise, such metalanguage is available in Modern Greek, so the difficulty the participants experienced in accessing it hints at the Anglo-centric orientation of their teacher education, and the disconnect between the ELT teacher-training programmes and mainstream Greek pedagogy. It also offers clues about the ways in which the teachers at the language school constructed a professional identity, by displaying their mastery of the profession’s metalanguage.

Another finding that surprised me even more was that I sounded considerably less confident in the Modern Greek interviews. The number of repetitions, and false starts In my discourse was considerably higher than what it had been in the English ones, even though one would have expected my my interviewing skills to increase over time. My provisional explanation is that I seem to have been subconsciously privileging the use of English, in which I was more proficient than other participants, in order to compensate for deficiencies in my interpersonal skills. I was using English, in other words, as an instrument of power, and I wasn’t even aware of it.

~

In summary, when faced with the task of interviewing teachers, I had the option of using either English or Greek. My language of choice was English, and I justified this choice by drawing on the school’s monolingual policy. Conducting the interviews in English, however, reinforced power imbalances between myself and the other participants, and also generated a number of thorny issues which I was unsure how to confront. A more flexible data generation policy, which gave the participants language choice proved to be better option.

In the next two posts in this series, I will address various considerations relating to how bilingual data might be presented, and I shall conclude this series by presenting a cline of representational options. Till next week!

Image Credit: The LEAF Project @ Flickr | CC BY-SA 2.0