"Well, it's more complicated than that"
The student has one job on this cross-examination: establish that the witness couldn't see the intersection clearly from where she was standing.
"You were standing behind the bus shelter, weren't you?"
"Well, I was sort of beside it, and I could see through the glass, and honestly I had a pretty good view because—"
He lets her finish. Then: "But the shelter was between you and the intersection?"
"Not really. I mean, partly. But I was looking the whole time, and I saw the red car go through—"
By the fourth question she has repeated her direct testimony twice, the jury has heard it three times, and his point — which was a good one — hasn't landed at all.
Control is question design
Students often believe that controlling a witness on cross is about force: a sharp voice, a firm manner, the confidence to cut someone off. Practitioners generally teach the opposite. A witness escapes control through the gaps in the question. Close the gaps and control follows.
Cross-examination training has to address three features of the runaway witness.
Every open door gets used. A question that allows explanation will receive one. Witnesses — especially those allied with the other side — want to repeat their story, and any question broad enough to let them will be treated as an invitation.
Compound questions invite escape. A question with two facts gives the witness a way to agree with one and fight the other — and to explain why.
And the jury is watching the contest. Arguing with the witness, visibly losing patience or repeating the question angrily all make the lawyer, not the witness, look unreasonable.
The design rules
One fact per question. "You were at the bus shelter." "The shelter has a metal frame." "The frame runs across the glass at eye level." "You were standing behind the shelter." Each question contains one fact the witness can only accept or deny.
Leading, short and statement-shaped. Leading questions are generally permitted on cross. The strongest form is a short declarative statement with the question implied: "You were behind the shelter."
Build to the point; don't ask it. The conclusion — she couldn't see clearly — is for closing. Cross assembles the facts that make it inevitable. Asking the conclusion ("So you couldn't really see, could you?") is the "one question too many," and it gives her the chance to say she could.
Use the witness's own words. Facts she has already committed to — in direct, a statement, a deposition — are the hardest to deny. "On direct you said you were 'next to the shelter.'"
Never ask why. "Why" invites a narrative. On cross, the narrative is hers, not yours.
Plan in chapters, not in questions
Control on cross starts before the lawyer stands up. Experienced examiners plan in short chapters — each one a single point, such as where she was standing or what was between her and the intersection — built from a handful of single-fact questions and anchored, wherever possible, in something she has already said.
Chapters do three things. They keep each point small enough to control. They let the lawyer abandon a chapter that isn't working without losing the whole examination. And they make it obvious, in rehearsal, which facts the lawyer can actually prove from the file and which ones depend on the witness cooperating — which is the difference between a cross that works and one that merely hopes.
When she runs anyway
Even well-designed questions meet witnesses who won't be controlled. The recovery techniques are well established, and the key is using them calmly.
Repeat the question, exactly. "My question was: you were standing behind the shelter?" Calm and identical. Juries notice when a witness won't answer a simple question twice.
Shorten it further. If the witness evades a statement, break it down: "You were at the shelter?" "Yes." "Behind it?"
Use her non-answer. After a long explanation: "So the answer to my question is yes?" Then stop.
Ask the court for help when appropriate. Courts can direct a non-responsive witness to answer the question, and the procedure — including whether to move to strike — varies. Know how your court handles it and use it sparingly; over-reliance looks like the lawyer can't manage the witness.
And don't argue. The moment cross becomes a debate, the witness has won — because she knows the facts and the lawyer only knows the questions.
Four ways it goes wrong
The open-door questioner, who asks questions broad enough to let the witness repeat her story.
The compounder, who packs two facts into one question and gets half an answer plus an explanation.
The one-question-too-many, who asks the conclusion instead of leaving it for closing.
The arguer, who responds to evasion with frustration and hands the jury a reason to side with the witness.
Why this isn't trained well
Students learn rules, not rhythm. They know leading questions are allowed on cross. They haven't built the habit of breaking every point into single-fact steps.
Practice witnesses cooperate. Classmates playing hostile witnesses rarely evade persistently. The witness who explains everything is the common real case and the rare practice case.
The one-question-too-many is invisible in rehearsal. A classmate answering the conclusion usually answers it the way the student hoped. A real witness doesn't.
And feedback comes as impressions. Faculty report that the cross "lost control." Without a transcript showing which question opened the door, students can't see where it happened.
What cross-examination simulation can rehearse
A ten-minute simulation can put the student opposite a witness who uses every open door, fights compound questions and responds only to single-fact, statement-shaped leading questions — so the student learns question design by experiencing what each form produces. The AI agent in Foretell AI plays the witness consistently for every student; the case file, prior statements and grading standards stay with the program.
Four versions to build:
- The explainer, who answers every question with a narrative.
- The half-agreer, who accepts one part of a compound question and fights the other.
- The conclusion-denier, who waits for the student to ask the ultimate question and then denies it.
- The prior-statement witness, where the student must use earlier words to hold her to a fact.
Design caution. Rules on leading questions, the scope of cross-examination, handling non-responsive witnesses and use of prior statements vary between federal and state courts and across jurisdictions. Modules should use the program's own evidence rules and case file; nothing here is legal advice.
Designing the module
Ten minutes, scored against a trial advocacy rubric focused on question design and control.
Pass one — design. What proportion of questions contained a single fact? How many were open, compound or "why" questions?
Pass two — control. When the witness evaded, did the student repeat, shorten, or use the non-answer — calmly?
Pass three — restraint. Did the student stop before asking the conclusion?
Rubric on observable behavior: Proportion of single-fact leading questions. Number of compound questions. Number of open or "why" questions. Number of times the witness repeated direct testimony. Was the conclusion asked? Did the student argue with the witness?
Single-fact proportion is the measure. It's countable, it predicts control almost perfectly, and students see immediately which questions let the witness escape.
For law schools and trial advocacy programs
It makes control teachable. "Maintain control" is advice. Single-fact question design is a technique, and it can be measured.
It pairs with direct. Running the same witness on direct and then on cross shows students the difference between building testimony and testing it.
It's consistent. Every student faces the same evasions, which makes feedback precise and grading fair.
And the practice payoff is direct. Junior litigators, prosecutors and defenders cross-examine constantly. The habit of one fact per question is learned slowly in court — and much faster in practice.
Frequently asked questions
How do you control a witness on cross-examination? Through question design: short, leading, statement-shaped questions containing one fact each, built step by step toward a point you save for closing.
What is the "one question too many" in cross-examination? Asking the witness to agree with your conclusion rather than leaving it for closing. It gives the witness the chance to deny it and undoes the facts you built.
What should you do when a witness won't answer yes or no? Repeat the question exactly and calmly, shorten it further, use the non-answer, and, if necessary and appropriate, ask the court for direction. Never argue.
Why shouldn't you ask "why" on cross-examination? Because it invites the witness to explain, and on cross the explanation belongs to them, not you.
The short version
She kept explaining because his questions let her. That's not a stubborn witness; that's an open door.
One fact per question. Short, leading, statement-shaped. Build the facts; save the conclusion for closing. When she runs, repeat the question calmly — then shorten it. Never ask why. Never argue.
Control doesn't come from being forceful. It comes from leaving her nothing to say except yes.
Foretell AI lets law schools and advocacy programs build cross-examination simulations — including evasive witnesses, question-design practice and control recovery like the one above — with configurable AI witnesses, recordings and rubric-based evaluation. If your students first meet a witness who won't say yes or no in front of a jury, we're happy to walk through how other programs have structured it.