Lost in transcription: the case of police interviews
Photo by Van Tay Media on Unsplash
The study in a sentence
Perception experiments can examine how people make judgments based on language use. In this study, we investigated how the presentation of police interview evidence (either as an audio recording or as different types of transcripts) would affect perceptions of an interviewee.
Results showed that people made different judgments about the same interviewee, depending on whether they read a transcript or heard the audio version of the interview. We also found that judgments differed depending on how linguistic information was represented in the transcripts.
Photo by Michal Czyz on Unsplash
The question
We know that people make personality and social judgments about other people based on language use. But how consistent are these kinds of judgments between representations of speech in writing (a transcript) and an audio recording of the same speech?
This is really important in the context of police interviews, where transcripts of interviews are frequently used as evidence. Audio recordings and transcripts of police interviews may be used interchangeably, overlooking potentially important differences between speech and writing. This study explores the issues with equating transcripts and audio recordings in this legally relevant context.
Key concepts
Speech perception: How people process, identify and interpret information from speech.
Person perception: How different aspects of language use can, often baselessly, influence social evaluations of a speaker.
Research questions
Does the presentation of a police interview in audio or written format affect people’s perceptions of an interviewee?
Does the representation of linguistic features in a transcript influence people’s perceptions of an interviewee?
Methods
The authors use a Likert Scale judgment task to investigate this question – exploring people’s attitudes and evaluations towards audio recordings and transcripts of a section of a police interview. Participants were presented with a section of a police interview either in audio format or in one of a series of transcript formats, and asked to make a series of judgments about the interviewee.
We manipulated whether and how pauses were represented within the audio and transcript test items to assess how this would influence perceptions.
The answer
The results showed that the interviewee was judged differently depending on whether participants read a transcript of the interview, or heard an audio recording.
Participants also paid far more attention to use of pauses by an interviewee when pauses were marked in a transcript, compared to in an audio recording. Some people listed pauses as a factor leading to a more favourable judgement of the interviewee; others listed pauses as a factor contributing to a more negative perception.
Overall, the results highlight the dangers of automatically treating a transcript as a 'copy' of original spoken language, in high-stakes settings such as police interviews.
Classroom activities
In more detail
Workshop talk video (will be available shortly)
![](https://www.google.com/images/icons/product/drive-32.png)
Workshop talk (slides)
Meet the author
James Tompkinson is a Lecturer in Sociolinguistics, and teaches modules in Forensic Linguistics, Language Variation and Change, and Phonetics.