Police Interview Discourse: Power, Questions, and Control

Drew and Heritage identified three features that mark institutional talk off from ordinary conversation. First, participants orient to a specific institutional goal: in a police interview that is eliciting an account that can be used as evidence. Second, the interaction follows constraints on what is appropriate: only certain kinds of utterances are expected from each party. Third, the institutional context itself infers special inferential frameworks: silence from a suspect after a caution can be commented on in court, while silence in normal conversation carries no such formal meaning.

For the forensic linguist these features are tools for analysis. When a transcript deviates from the expected question-answer structure, something worth investigating has happened. When an officer fails to follow up an ambiguous answer, that omission may be as telling as an explicit statement. When an interviewee uses a word the officer then reformulates with a stronger term, and the interviewee accepts the stronger term without protest, the transcript now reads as if the interviewee used that stronger term first.

Researchers studying police interviews classify questions on a scale from open to coercive. The position on that scale directly affects how much of the interviewee's own account can survive.

Best-practice guidance in England and Wales (the PEACE model), in Australia (where forces have increasingly adopted PEACE-derived approaches), and in Canada explicitly recommends starting with open questions and using closed questions only to clarify, never to establish contested facts. In practice, recordings of real interviews show that officers routinely use leading and presupposition questions, especially when they have a working theory and want to test it quickly. The linguistic analysis of those recordings is often the evidence that makes a confession or identification challengeable in court.

Reformulation is among the most studied features of police interview discourse because its effects on the record can be significant while passing unnoticed in the moment. When an officer paraphrases an interviewee's answer, the restatement serves as a comprehension check and simultaneously produces a new version of what was said. Whether and how much that version differs from the original is the forensic question.

Some changes are neutral: tidying up disfluency, replacing a pronoun with a name. Others are semantically significant. The interviewee says 'I might have bumped into him'. The officer reformulates as 'So you made contact with him'. Bumped-into is accidental and minor; made contact is neutral but implies deliberate proximity. If the interviewee confirms the reformulation with 'Yeah' or 'Right', the record now has them saying they made contact, which is subtly more useful to a prosecution than what they actually said.

Joanna Thornborrow's research and Alison Johnson's work on British police interviews from the 1990s onward documented this pattern systematically. The forensic implication is that a linguist examining an interview transcript cannot simply read it at face value. They need to compare what the interviewee originally said against what the officer's reformulation recorded, and check whether the interviewee had a genuine opportunity to correct any divergence.

Register is the variety of language appropriate to a social context. Police officers typically open interviews in formal procedural register: rights read from a card, charges stated in legal language, caution delivered verbatim. This register creates social distance and signals institutional authority. Some interviewers then shift deliberately into informal, colloquial, even slang-inflected language when they want to build rapport or lower a suspect's defences. This is code-switching: moving between registers within a single interaction.

From a linguistic standpoint the switch is a control move. The officer who switches into casual register is implying 'we're just talking'. The suspect who accepts that framing and relaxes into casual speech may produce less guarded utterances. Any admission obtained during the informal phase of an interview is still a formal statement on a recording, however informal the language felt in the moment. This mismatch between the felt register and the legal register of the interaction has featured in several English appeal cases where defendants argued their admissions were made in a context they did not understand was formally recorded.

Most legal systems give criminal suspects the right to silence and the right to legal representation before and during questioning. These rights are usually delivered in the caution at the start of the interview, and they are the most linguistically complex utterances in the whole encounter. In England and Wales the caution reads: 'You do not have to say anything. But it may harm your defence if you do not mention when questioned something which you later rely on in your court.' Research by Janet Cotterill and others has shown that large proportions of detainees who have just heard that caution cannot paraphrase what it means. The right to silence is formally granted and linguistically obscured almost simultaneously.

• Complex syntactic structure: the caution uses a double negative, a conditional clause, and a forward reference to court proceedings, all in one sentence.
• Reading pace and context: detainees hear the caution after arrest, in unfamiliar surroundings, under cognitive stress, read at the pace the officer is used to.
• Register mismatch: the caution is a written legal formula delivered orally; it does not match the conversational register of what follows and so often fails to register as something that requires active decision-making.
• Socioeconomic variation: comprehension rates are consistently lower among younger suspects, those with lower literacy, and those in whose first language is not English, amplifying existing inequalities in the interview room.

An interpreted police interview involves at least three voices: officer, interpreter, and interviewee. It also involves at least two different linguistic systems, and any mismatch in them can introduce distortion. Research by Sandra Hale on interpreted legal proceedings has identified recurring problems: interpreters omitting hedges and uncertainty markers ('I think', 'maybe', 'kind of'); interpreters upgrading vague denials into strong ones; interpreters shifting the pragmatic force of a question in ways that change what the interviewee is being asked.

The forensic linguist's role in an interpreted interview case is to compare the original-language recording with the interpreter's renditions, and then compare those with the English record. In UK practice this comparison was central in the appeal of Bhupinder Singh in 2006, where a Punjabi interpreter had significantly altered what the defendant said in terms that mattered to the charge. The question for the court was not whether the interpreter was dishonest but whether the interpreted record was a fair representation of the defendant's account.