Police interviews of witnesses as well as suspects should routinely be tape recorded to ensure that there is an accurate account of the conversation, according to barrister Anthony Heaton-Armstrong.'This would do away at a stroke with unjustifiable allegations of prompting, pressurising, tailoring, police canteen collusion sessions and many other forms of witness manipulation which have long been suspected but are impossible to prove or disprove,' he told a seminar on eye witness testimony, organised by the British Academy of Forensic Sciences, the Criminal Bar Association and the London Criminal Law Solicitors Association.Maxwell McLean, an inspector with West Yorkshire Police, agreed that tape recording is necessary.

'Witnesses are more vulnerable interviewees than suspects,' he said.

Suspects will have been cautioned about the seriousness of their answers, will generally have legal assistance and may be guarded for fear of self-incrimination.Witnesses, however, have no such protection.

Mr McLean said: 'Witnesses will wish to appear co-operative and competent respondents and will provide answers which it is assumed the questioner wants to hear.''A witness's suggestibility, combined with an unethical interviewer is a recipe for conversational chaos,' he warned.

'Whatever the interviewer says, goes.'Mr McLean compared 16 tape recorded conversations between police and witnesses with the subsequent written statement in each case.

He found there were an average of 14 pieces of information given by the witness which were not included in the written statement.The number of 'edited items' ranged from four to 38 per interview.

In some cases the information omitted was key to the investigation, such as: 'I saw a sticker in the back of the car', and 'I explained what had happened to the person in the office.'In four cases, Mr McLean found that the police officer wrote down a fact that was directly contrary to what the witness had actually said.

'So not only can information of real relevance be left out of written statements, but where necessary for the sake of plausibility, contrary information may be included,' he said.Despite the omissions and contradictions, all 16 statements were signed by the witnesses as a true record.Mr McLean was also critical of police interviewing techniques, 'which were worse for witnesses than for suspects,' he said.

The ideal split of talking time for an effective interview is 20-30% for the questioner and 70-80% for the interviewee.

Yet in the 16 witness interviews he studied, on average, talking time was split almost evenly between the two.

'In one example, a witness was only allowed to speak for one minute 19 seconds - and he was describing a brawl in a pub,' said Mr McLean.The types of questions being asked reflected 'inappropriate interviewer control'.

65% of all questions were either leading or invited only a yes/no answer.

This is nearly double the amount of counterproductive questions put to suspects, as determined by a similar study.

'It would appear that the witness is in much greater danger of being led than the suspect,' he concludes.