Flattened code-switches can turn uncertainty into certainty by removing the pauses, qualifiers, and shifts in language that show how a participant is weighing an answer. Before publishing, replay the original audio and keep the Twi-English phrasing beside any English rendering that carries a key finding.
The sentence that sounded more certain on paper
At 6:40 p.m., Akosua, a composite researcher with a cold cup of sobolo beside her laptop, was preparing the final findings section of an interview study. Her participant had spoken about choosing where to seek care. The English transcript read: “I prefer the clinic because it is more reliable.”
It was clean. It was quotable. It also supported a conclusion Akosua had already begun to write.
Then she replayed the recording.
The participant had said, “Me deɛ, maybe clinic no, because wɔ hɔ deɛ, at least, sometimes, wotumi nya obi ka wo ho asɛm.” The voice slowed around “maybe.” There was a small laugh before “at least.” The participant had moved between Twi and English while searching for a careful way to describe an uneven experience.
The finding had become too firm. “I prefer the clinic” suggested a settled choice. What the participant had actually conveyed was conditional hope: the clinic could be helpful, sometimes, if someone was available to explain what was happening.
Akosua had one evening left before sending the draft to her co-authors. If she kept the flattened sentence, the published report could overstate confidence that was never there. That stronger wording might shape recommendations, citations, and future decisions built on a participant’s voice.
She paused the document and started again.
Code-switching can carry the meaning of hesitation
A switch between Twi and English does not automatically need interpretation. Sometimes a word is simply the easiest word available. But in an interview, a switch can also mark distance, uncertainty, emphasis, or a change in how safe a person feels making a claim.
“Maybe,” “at least,” “I think,” and their Twi equivalents can look disposable when a transcript is edited for smooth English. They are often the difference between a finding about certainty and a finding about possibility.
The risk grows when a researcher hears natural speech but works later from a polished transcript alone. Spoken language contains timing, repetition, self-correction, and changes of register. A participant may begin confidently in English, move into Twi while explaining a family concern, then return to English for a phrase they know carries institutional weight. Those movements belong to the data.
This does not mean every transcript should preserve every “um,” restart, or repeated word. Research still needs readable material. The job is to decide deliberately what can be cleaned up and what changes the claim.
For a key quote, ask: What level of certainty did the person express? Did the English version add confidence? Did a qualifier disappear? Did the participant use one language for the fact and another for their feeling about it?
That last question can change the finding.
Keep the original voice close to the interpretation
Akosua created a small review table for the quotes that carried her central argument. One column held the original Twi-English excerpt. One held a close English rendering. The last column recorded what she could safely infer, including uncertainty where it remained.
Her revised wording became: “Participants described the clinic as a possible source of clearer explanations, although access to staff support appeared inconsistent.” It was less dramatic than the first version. It was also truer to what had been said.
A natural Twi and Ghanaian English conversation tool can help at this stage when researchers need to talk through wording without forcing each turn into a single language. They can type or speak an excerpt as it was said, then examine whether an English rendering preserves its caution and context. The tool should support the researcher’s judgment, not replace it.
Sensitive interview material deserves the same care as the interpretation. Before entering an excerpt into any cloud-connected tool, remove identifying details and check exactly what will be sent. Nkomo gives people explicit choices for cloud consent: never, ask each time, or this session. Its on-device history can be turned off and purged immediately, and people can export or delete their account data in one tap.
That clarity matters when a recording includes family details, health concerns, workplace conflict, or a participant’s private uncertainty. See also What Happens When the Key Detail Arrives in Twi?.
Build a replay step into the publication process
The useful habit is simple: replay the original audio for every quote that anchors a conclusion. Do it before the final draft circulates, when changing one sentence can still change the interpretation without creating a scramble.
Listen once for words. Listen again for certainty, hesitation, and the point where the speaker changes language. Keep the original excerpt available to a bilingual reviewer when the finding carries real weight. If a translation has to choose between smoothness and caution, keep the caution.
Later that night, Akosua listened again to the participant’s final sentence. The clinic was not a clear preference. It was a place the participant hoped might offer help on a difficult day. In the revised findings, that hope remained visible.
Comments
No comments yet.