tl;dr-ELT

I’m sure we’ve all made use of dictation at one time or another: as a spelling focus or as a means to encourage learners to notice connected speech. It’s easy to think of dictation as something that belongs squarely in the classroom. However, it’s also common in the criminal justice system. Before electronic recording became standard, a suspect’s words were often typed up “verbatim” on a typewriter & treated as an accurate record of what was said.

Helen Fraser’s new open-access article in the Australian Journal of Linguistics takes us deep into that world. It’s not an ELT study, instead it’s an unsettling case study of a 1988 police interview that became the sole basis for a life sentence. It’s a powerful reminder of how much hinges on our assumptions about transcription, accuracy & what it means to “capture” spoken language & shows just how far the consequences of those assumptions can reach.

The study

Fraser examines the case of Stephen “Shorty” Jamieson, convicted in 1990 for a murder he has always denied. The only substantial evidence was a typed “verbatim” Record of Interview (ROI) produced by detectives on a mechanical typewriter. No audio. No shorthand. No witness until the final minutes.

Her method involved:

• reconstructing the interview timeline using the ROI’s timestamps
• analysing word counts, typing speeds & speech rates
• comparing the ROI’s structure with known features of spoken interaction
• reviewing trial transcripts, appeal documents & judicial commentary
• drawing on research into entextualisation, epistemic injustice & forensic transcription

Rather than comparing the ROI to samples of Jamieson’s speech (as earlier experts attempted), Fraser asks a simpler question: Could this transcript have been produced in the way police claimed?

The findings

Several findings are hard to ignore.

• The ROI could not have been typed verbatim.
  Part 1 alone (832 words) would require typing at around 55 wpm on a mechanical typewriter. Other sections would require typing far slower than natural speech, meaning the typist would have been constantly behind.
• Key events in the transcript could not have happened as written.
  Detectives testified that the first two pages were faxed during a 10-minute break, yet those pages contain dialogue that would have had to be typed after the lead detective left the room.
• Crucial spoken interactions are missing.
  Nearly an hour of breaks contain no dialogue. Routine “business talk” (e.g. removing pages, clarifying errors) is absent. Yet one trivial exchange — “Would you like something to eat?” — appears immediately after a long confessional monologue.
• A major typographical error was never corrected.
  The word “not” is missing from the legal caution (“you do have to answer…”). The [1]Justice of the Peace later claimed he noticed & re-read it correctly, but the ROI shows no correction & the timing leaves no space for such an intervention.
• The ROI is uncorroborated until 6:07pm — after the confession was fully typed.
  There is no independent evidence of what was said before that point.
• The legal assumption that “even if not verbatim, it captures the gist” is untenable.
  Without corroboration, a typed ROI cannot reliably represent any spoken event.

Context

Fraser situates the case within wider research on:

• entextualisation & the interpretive nature of transcription (Park & Bucholtz)
• epistemic injustice (Fricker)
• false confessions & interrogation practices (Kassin)
• the history of “verballing” in Australian policing. This term refers to the once‑common Australian police practice of fabricating or embellishing a suspect’s “dictated” statement in typed form.
• the shift to mandatory electronic recording in the 1990s

Fraser’s argument is clear: forensic transcription is not clerical work. It is expert linguistic work, & the discipline needs theoretical tools that treat transcripts as artefacts shaped by power, procedure & human limitation.

Imagine typing a verbatim transcript of a lively pair-work exchange in real time. Even with a laptop, you’d miss turns, tidy up grammar, paraphrase, omit hesitations & unconsciously “smooth” the interaction. Now imagine that transcript being treated as legally equivalent to the original speech event. The gap between spoken reality & written representation becomes a matter of justice, not convenience.

Fraser chose this particular case because it provides an unusually clear, well-documented example of how legal misconceptions about transcription can produce ongoing injustice: Stephen “Shorty” Jamieson remains in prison more than 35 years after the conviction, with the typed record of interview still treated as the central evidence against him.

Check out this post for a broader look at Forensic Linguistics.

Teacher takeaways?

• Transcription is interpretive. Even in classroom research, our choices shape the story the data tells.
• Spoken language is fast, messy & multimodal. Any written version is a reconstruction.
• Language ideologies matter. The belief that “anyone can transcribe” mirrors the belief that “anyone can judge language”, often to the detriment of marginalised speakers.

How do you use dictation in class?

---

[1] a layperson authorised to witness documents & administer oaths. In the 1980s Australian system, a JP was often brought in at the end of a police interview to certify that the typed ROI had been read aloud to the suspect & that the suspect agreed it was accurate.