"This can mean the most complete, accurate and reliable evidence is not elicited from the child," it said.
Auckland University of Technology researchers said most children who testify in jury trials allege they have been sexually abused. AUT research last year found cases involving children took an average of 15 months to reach trial, making it hard to remember details and preventing them from moving on.
They found lawyers often used closed and leading questions, difficult words, accused witnesses of lying and changed topics abruptly - a practice used to put liars off balance and test whether children have memorised testimony.
"It seems possible that such a tactic could just as easily confuse the child who is trying to answer questions honestly," they say.
The experiments used an adult to play an 8-year-old girl who had been sexually assaulted. Questions were asked by an experienced forensic interviewer for the first experiment and a speech language therapist for the second, using three questioning options.
No one liked the first two options, where the intermediary asked all the questions in one block after being briefed by both lawyers, or translated the questions one by one. A third option, where questions were asked in topic chunks, with lawyers able to enter follow-up questions after each chunk, was preferred.
But defence lawyer Richard Earwaker said yesterday that this too was "problematic" because it took control from the lawyers.
"I have spent 25 years developing my skills in cross-examination," he said. "If someone else is asking the questions for you, particularly in a way that wouldn't be how you would ask them, is that effective in terms of getting your defence across?"
The report recommends a working group develop an intermediary model. It also supports pre-recording children's testimony, including cross-examination, as soon as possible after a crime is alleged.
