Courtroom Discourse: Examination, Cross-Examination, and Question Control

Discourse analysts describe the courtroom as an institutional speech event, meaning its conversational structure is governed by explicit rules rather than emerging spontaneously between participants. In ordinary talk, anyone can change the topic, interrupt, or ask a clarifying question. In court, the questioner controls all of those moves. The witness may not question the lawyer, may not introduce a new topic, and in many common-law systems may be stopped mid-sentence if the answer exceeds the scope of the question. These rules have a rationale: they prevent witnesses from making speeches and ensure both sides get to test the evidence. But they have a linguistic cost, and courts are increasingly aware of it.

John Gibbons's 2003 textbook Forensic Linguistics identifies four features that mark court language as different from everyday discourse: the asymmetrical distribution of questions, the formal register required of witnesses, the fixed turn-taking system, and the presence of a third-party audience (judge and jury) who are the real recipients of everything said. That last point matters. When a barrister asks a long question that the witness struggles to understand, the question has already performed its function for the jury before the answer arrives. The witness's confusion itself becomes data.

In examination-in-chief (called direct examination in American practice), the advocate who called the witness is generally prohibited from asking leading questions. The result is a preference for open-format questions: Tell me what happened next. What did you see? What did he say to you? These question types invite the witness to construct a narrative in their own terms, which is why examination-in-chief often produces the longest and most coherent witness accounts in a trial.

The preference for narrative in examination-in-chief has both a procedural and a psychological logic. Procedurally, an open answer commits the witness to a version of events that cross-examination can then probe. Psychologically, research on memory shows that free recall produces more complete and more accurate accounts than recognition or cued recall, which is what leading questions trigger. Susan Ehrlich and other discourse analysts have noted, however, that the preference for narrative also means that a skilled examiner can still shape the narrative indirectly, by choosing the sequence of questions and by what they choose not to ask.

Cross-examination is where the linguistic asymmetry is most pronounced. The cross-examining advocate is permitted to lead, and the most effective cross-examiners typically do so with short, syntactically simple questions that require a yes or no. 'You were in the building at ten o'clock.' 'You did not see the defendant.' Each question constrains the witness's response space to a single binary dimension. Volunteering an explanation requires the witness to actively resist the question's format, and many witnesses, especially inexperienced or vulnerable ones, do not know they are entitled to do that.

• Short confirmatory questions: one factual proposition per question, phrased as a statement awaiting confirmation. Maximally constraining.
• Tag questions: 'You were angry, weren't you?' The tag grammatically invites agreement and exploits the conversational norm of cooperating with a question's expected answer.
• Either/or questions: 'Did you run or walk?' frames a response space of two options when the true answer may be neither.
• Complex embedded questions: a long question with multiple clauses that the witness may not fully process before agreeing. Answering yes to the last clause accepts earlier ones by default.

A presupposition is a proposition that a question treats as already established, regardless of what the speaker or listener actually believes. The classic teaching example is 'Have you stopped beating your wife?'. Both yes and no accept that beating happened. In courtroom practice, presuppositions are rarely that crude, but they are pervasive. 'When did you decide to leave?' presupposes a decision was made. 'How many times did you argue?' presupposes arguments occurred. A witness who answers these questions is grammatically assenting to propositions they may never have explicitly agreed to.

Elizabeth Loftus's classic 1974 studies, which used questions about filmed car accidents, showed that changing one word inside a question changed what witnesses reported seeing. Witnesses asked 'How fast were the cars going when they smashed into each other?' gave higher speed estimates and were more likely to report seeing broken glass than witnesses asked the same question using 'hit' instead of 'smashed'. The implication for legal practice is direct. The form of a question does not just elicit a memory. It can partially construct one.

Sandra Harris's 1991 paper on evasion in political interviews drew on broadcast data but has been widely applied to courtroom testimony. Harris's core finding was that failure to answer a direct question is not a single act. It is a repertoire of moves, each with a different linguistic signature, and that repertoire is available to anyone in an institutional interview, including a witness in the box.

1. Challenging the question's relevance or legitimacy
   The respondent implies or states that the question ought not to be asked. In court: 'That question assumes facts not in evidence, I can't answer it that way.'
2. Treating a premise as false
   Rather than answering the main proposition, the respondent disputes an embedded claim. 'I wasn't angry, so your question doesn't apply.' This rejects the presupposition rather than the question itself.
3. Giving an incomplete answer
   Responding to part of the question, often the less damaging part, while leaving the core unaddressed.
4. Topic shift
   Moving to a related but different subject under the guise of answering. The new topic is usually less problematic for the speaker.

Harris's insight was that these moves are systematic enough to be classified and that the questioner's ability to force a direct answer is limited by the structural rules of the interaction. A cross-examiner who encounters none of these moves is facing either a fully cooperative witness or a thoroughly prepared one. The forensic linguist asked to comment on evasion in a transcript needs to distinguish the two cases, because cooperation and evasion produce different patterns at the level of turn structure and sentence type.

Research across Australia, the UK, the US, and South Africa converges on a troubling finding: standard cross-examination techniques produce false agreements from vulnerable witnesses at rates that are not explained by lying or evasion. The mechanism is straightforward. Leading questions exploit the conversational norm of responding cooperatively. Tag questions exploit acquiescence bias. Complex embedded questions exceed working memory capacity. These effects are present in all witnesses to some degree and amplified in children, people with intellectual disabilities, and people testifying in a second or third language.

In England and Wales, the Youth Justice and Criminal Evidence Act 1999 introduced 'special measures' including video-recorded pre-trial interviews, live video links, and court-appointed intermediaries for vulnerable witnesses. Intermediaries assess the witness's communication capacity, sit in the box, and can ask for questions to be rephrased if they are likely to be misunderstood. This is a direct institutional response to the forensic-linguistic finding that standard questioning harms testimony quality.

Australia, South Africa, and several Canadian provinces have adopted comparable intermediary or support-person schemes. Diana Eades's extensive work on Aboriginal witnesses in Queensland courts documented how the question formats used in cross-examination clashed systematically with the communication norms of Aboriginal English, producing apparent evasion where the witness was in fact observing their own cultural conventions about when and how to answer. The problem was not deception. It was a collision of discourse norms across two different speech communities.