Transcription as a Key Phase of Data Analysis in Qualitative Research: Experience from KwaZulu-Natal,South Africa

Introduction

Transcription has been described as a data management mechanical task of putting spoken words into textual format (Davidson 2009; Loubere 2017). In recent years, there has been a change in thinking about transcription, and many scholars have emphasized the need for researchers to report on transcription as an independent component of a study rather than viewing it as a less important part of data management (Bhattacherjee 2012; Davidson 2009; Halcomb and Davidson 2006; Loubere 2017; Ochs 1979; Oliver et al. 2005; Wellard and McKenna 2001). It has been argued that rather than being a chore, transcription is a key stage of data analysis in which interpretations and meanings are generated from interview transcripts (Agar 1996; Halcomb and Davidson 2006; Silverman 2013). We share our experiences of how transcription facilitated an interpretative act in a study using qualitative data collection methods.

Approaches to Transcription as an Interpretative Act

Our study, known as mHealth Africa (mAfrica) focused on the co-development (with potential users) of a mobile phone-connected technology for HIV diagnostic tests and online clinical care pathways needed for optimal delivery of decentralized HIV testing (including HIV self-testing), prevention and care in low- and middle-income countries (Adeagbo et al. 2019). The study was approved by the local institutional review board, and ethical approval included the provision for participants to agree to the use of their data, including quotes, in reports and publications. The study was conducted in two communities (semi-urban and rural communities) in uMkhanyakude district in KwaZulu-Natal, South Africa. Fifty-four semi-structured in-depth interviews and nine focus group discussions were conducted in the local language (IsiZulu) and English, with community members and service providers by four local social science research assistants, under the mentorship of a senior social scientist and two senior team members experienced in qualitative research. Two common transcription paradigms—naturalism (complete transcription of all words and sounds [such as stutters and coughs etc.]) and denaturalism (where sounds are removed) described by Bucholtz—were adopted during transcription, initial data coding, and analysis (Bucholtz 2000). In the following section, we discuss three key steps to transcription as an interpretative act in our study.

Step 1: Ensuring Quality: Fieldworker Training and the Transcription Process

The fieldworkers responsible for the data collection were IsiZulu native speakers. They were also involved in the transcription of interviews, translation, and analysis. As described by Bucholtz (2000:1441), transcription is “representational insofar as they offer a version of events and a portrait of the participants in those events.” Thus, a good quality transcript would be detailed with the transcriber considering both verbal and nonverbal communications and reflecting the originality in the audio-files (Silverman 2013).

To address the issue of trustworthiness, quality, and transcription representativity in our study, the fieldworkers received training on effective interviewing and good quality audio-recorders were also provided during fieldwork. Then fieldworkers were trained to do transcription and translation, and this gave them guidelines on how and what to transcribe. Fieldworkers were able to transcribe and interpret not only the spoken words, but also nonverbal communications recorded during interviews. During the fieldwork, they observed and noted participants’ emotions and reactions to the topics discussed. This corresponds to Lucas’s view that researchers should endeavor to pay close attention to the conversations as well as nonverbal expressions during transcription (Lucas 2010). This example shows both verbal and nonverbal expressions of a young man responding to a question about how he was previously affected by HIV/AIDS:

(Sighs worriedly)…What can I say, like my brother, the older one. He passed on, it started off as TB and he took treatment but stopped taking it and continued to drink alcohol…(takes a deep breath). Then after some time there was HIV (shakes head), he was not taking the pills. He was just throwing it in the toilet. That’s when I saw that HIV is actually killing people.

Finally, all transcripts were first reviewed by the fieldworkers and subsequently by their supervisors and independent senior colleagues fluent in both IsiZulu and English. Fieldworkers’ supervisors randomly selected (across the study sites and by gender) and listened to some interview audio-files while reading the transcripts to check if what was documented was a true representation of what was said by participants. A similar process was repeated by colleagues who served as independent reviewers. This simplified the identification of transcription errors like words that were skipped or typing errors (Poland 1995). Reviewing the transcripts has proven to be an essential practice that enhances the quality of transcripts in qualitative research (Davidson 2009; McLellan et al. 2003).

Step 2: IsiZulu to English: Navigating the Translation Quandary

Noting the “unconscious interference” of translators as described by Shread (2007), we sought to reflect and inflect the English interpretation of our IsiZulu interview transcripts rather than presenting a “perfectly equivalent translation” (p. 226). Coupled with relevant training, the fieldworkers are native speakers of the language in which the interviews were conducted, and this contributed to transcription accuracy and limited representation errors during translation. We had several debriefing sessions on translation (including back-translation) to improve transcript quality and appropriate representation of IsiZulu texts in English, given that some Zulu words have no direct English synonyms. After the initial review of the transcripts (both IsiZulu and English versions), fieldworkers’ supervisors and independent colleagues outside of the study reviewed and evaluated the transcripts, a recognized process of ensuring quality and accuracy during translation (Clark et al. 2017). The translated transcripts were coded and analyzed once we were satisfied with the quality of the transcripts. These quotes from a young man describing sexual behaviors among males illustrate the use of representativeness:

Direct translation: “There is nothing else, it is saving yourself. If you get tempted, you must get tempted with the hats that are worn”

Representative text: “There is nothing else besides saving yourself. If it happens that you get tempted, use the condom.”

Step 3: Data Selectivity and Analysis

It is impossible for transcripts to capture all dialogue and interaction during an interview (Duranti 2006). All transcripts are selective in some way (Davidson 2009; Loubere 2017; Ochs 1979), and the analysis process adds another layer of selectivity. We adopted selectivity especially during data analysis, as the data features that did not form part of the analysis were omitted. We focused mainly on selected information from the transcripts during data analysis to make the texts readable and presentable. Finally, we tried to keep the meaning and flow of the original texts during translation and data analysis.

Conclusion

As shown in the above steps, transcription was a key phase of analysis as it promotes transcripts knowledge and immersion, and adoption of reflexivity throughout the transcription process could prove to be a valuable for the research as well as the representation of participants’ worldviews. Thus, qualitative researchers should endeavor to constantly reflect on interview transcripts and improve their transcription skills. Also, involving interviewers in the transcription process is empowering and impacts positively on the outcome of the research since some of the local idioms and meanings enunciated would be captured.