Flynn & Anor, R v (Rev 1)

On 7 December 2006 at Reading Crown Court the appellants, Kris Flynn and Joe Phillip St John, were convicted of conspiracy to commit robbery. On 8 December 2006 each was sentenced to 13 years’ imprisonment less the appropriate days spent on remand. There were two other co-accused. Mark Bannister pleaded guilty and was sentenced to 12 years’ imprisonment. On 4 December the full court quashed his sentence and substituted for it a sentence of 10 years’ imprisonment. Frank Sines was acquitted of conspiracy and discharged. The appellants appeal with leave of the full court.

These appeals raise issues relating to identification by voice recognition. The evidence was as follows. On 29 April 2006 at about 7.50 am, Samia Taourit, an employee of Elonex Computers plc, arrived for work at the company premises situated at Cricklewood, London. The company acted as a distribution warehouse for IBM computers and stored on its premises expensive computer equipment. The premises had in the past been a target for criminal activity as a result of which a strong room had been placed on the premises as a security measure.

Mr Taourit was the first employee to arrive at the premises on that day. He unlocked the main gates to the premises and drove his car into the car park. As he was walking into the main building he noticed a white Mercedes Sprinter van which had driven through the open gates. Believing the driver to be lost, Mr Taourit approached the van and spoke to the passenger. A number of hooded masked men then jumped out of the rear of the van and pushed a double-barrelled shotgun into Mr Taourit’s stomach. He was pushed into the premises and told to de-activate the alarm. This he did in order to prevent the robbers harming him. All of the men were wearing masks or balaclavas and some had crowbars with them. The robbers took personal documents from Mr Taourit before taking him to a reception room where his feet and hands were bound with red parcel tape and cable ties.

At some point during the course of the robbery, the robbers panicked on hearing a noise outside. They left the premises, having been unable to penetrate the steel walls of the strong room. In fact the strong room contained approximately £10,000 worth of computer chips. After the robbers had left in the white van Mr Taourit was found by colleagues in a very distressed condition. It was not in dispute that between 19 March 2006 and 29 April 2006 there existed a criminal conspiracy to rob Elonex Computers plc. The issue for the jury was whether or not these two appellants were parties to the conspiracy.

In support of its case against each of these two appellants the prosecution relied on evidence of a number of circumstantial facts, together with some forensic evidence. In addition, the prosecution relied upon evidence of police officers who purported to recognise the voice of each appellant taken from recordings of conversations covertly obtained from a microphone in the Sprinter van just before it entered the premises.

The van used in the attempted robbery had been stolen on 20 March 2006. On 29 March 2006 it was seen parked at an address in Slough bearing registration plates with the number VN04WFW. These registration plates were false, since that registration number belonged to another Sprinter van owned by a different company which was at the time locked up in the owner’s compound. Whilst the van was being watched on 29 March the appellant, Flynn, and the co-accused, Sines, arrived in a silver Subaru which belonged to Flynn. They went into the premises where the van was. In the following days before 28 April 2006 covert surveillance cameras tracked the Sprinter van in different locations and recorded visits to it by Flynn and other men. On 28 April 2006, the night before the attempted robbery, the Sprinter van was moved by Bannister and Flynn to Heron Close in Ascot, close to Flynn’s father’s home. At some stage before 29 April 2006 police officers fitted a secret listening and transmitting device in the van. On 29 April at 05.10am, Flynn and Sines appeared together and were seen to spend three or four minutes with the bonnet of the van open. They were apparently using jump leads to start it. Flynn was wearing gloves; Sines was not. The van then was then driven out of view. The van was next seen by Mr Towritt when it entered the premises of Elonex Computers plc.

Following the aborted robbery the van was seen by police officers to leave the premises. It and an accompanying Volkwagen Golf motor vehicle, with two people in the front, both wearing balaclava helmets, were followed by police vehicles. Although police were unable to follow the Sprinter van and Volkswagen Golf to their ultimate destinations, the Sprinter van was subsequently located and removed for forensic examination. This examination revealed fingerprints attributable to both of the appellants.

In addition to the above evidence, the prosecution relied on voice recognition evidence given by four police officers. Their evidence was to the effect that they recognised the voices of Flynn and St John on the covert recording of voices in the sixty minutes before the Sprinter van entered the premises at Cricklewood. The evidence of the police officers was that, having spoken to the two appellants during and after their arrest, they were able to recognise their voices on the recording captured by the covert recording device. The officers had also obtained covertly a sample of each of the appellants’ voices in conversations at the police station. Further, the jury had before them two transcripts of the recording taken from the van. One was produced by one of the police officers and the other by Dr Frederika Holmes, an independent forensic consultant. The latter had been instructed on behalf of the prosecution. At trial, counsel on behalf of both appellants objected to the introduction of the voice recognition evidence and the transcripts. The judge ruled both the evidence of the police officers and the transcripts admissible.

Shortly after the discovery of the Sprinter van Flynn was arrested. At interview he replied “No comment” to all questions asked of him. St. John left the United Kingdom on 11 May to get married abroad and returned on 26 May 2006. On that date he was arrested. He was interviewed twice on the same day but answered “No comment” to all questions put to him.

Each of the appellants gave evidence. Both denied taking part in the conspiracy. Flynn said that he had been helping Bannister to repair a Sprinter van because it had starting problems. He did not dispute the forensic evidence connecting him with the van. He denied that his voice was on the covert recording taken from the van.

St John said that he had spent the weekend during which the robbery occurred with his family. Both appellants relied on the evidence of first Dr Frederika Holmes and second, Professor John French, a speech recognition expert, in support of their case that it was not possible to recognise the voice of either of them on the covert listening recording. Much of Professor French’s evidence was distilled into written admissions which were placed before the jury.

There are two grounds of appeal. They are common to both appellants. First, the judge’s decision to admit the voice recognition evidence of the four police officers in evidence is challenged. Secondly, it is submitted that the judge failed to give an appropriate direction in relation to this evidence. Before we turn to these grounds of appeal we must make some general comments about the background to voice recognition evidence.

There can be no doubt that the admission of voice recognition evidence is controversial, perhaps highly controversial. In this appeal we are concerned with two categories of such evidence. They are the agreed evidence given by two experts in voice identification and the evidence of the four police officers.

The first category, expert evidence, in this field can be of two types: firstly, auditory analysis and secondly acoustic/spectrographic analysis. The second category of evidence falls into a group described by the experts in this case as lay listener evidence. The latter requires that the witness possesses some special knowledge of the suspect that enables him or her to recognise the suspect’s voice. The most common example of such evidence is the knowledge of a close relative or friend. However, there are other persons who may acquire sufficient special knowledge by their familiarity with the suspect’s voice. For example, a person may acquire such familiarity by the frequency of his or her contact with the suspect. A yet further group may comprise those who acquire specialist knowledge by listening to a sample of the speech of a known person and comparing it with a recording of a disputed voice. The latter are referred to in some judgments and academic papers on this topic as ad hoc experts. In our opinion, this description is unhelpful, since those who acquire the specialist knowledge cannot in our view properly be referred to as experts.

In this case there was before the judge and jury some general expert evidence and rather more before us. At the outset of the appeal we heard unsworn evidence from Dr Frederika Holmes. Strictly, this evidence was not fresh evidence since it was available to both sides at trial. Dr Holmes had been instructed by the prosecution. She did not give oral evidence at the trial but, as we have already said, her evidence, in the form of witness statements, was read to the jury. Her evidence at the hearing of the appeal was taken, without objection from the prosecution, and heard by us de bene esse. Dr Holmes expanded and clarified her written evidence which was before the judge and jury. We also have, as had the jury, expert evidence in the form of a report and the written admissions from Professor French.

In general terms the expert evidence before us demonstrates the following:

The admissibility of voice recognition evidence is treated differently in different jurisdictions (see “Sounding Out Expert Voice Identification” by Professor David Ormerod CLR October 2002). We propose to refer only to authorities in our jurisdiction and one decision of the CACD in Northern Ireland. They are R v Robb [1991] 93 Cr.App.R 161; R v Clare and Peach [1995] Cr.App.R 333; R v O’Doherty [2003] 1 Cr.App.R 5 (NI) and A-G’s Reference [No 2 of 2002] [2003] 1 CAR 21.

In Robb, the court held that voice recognition evidence given by a phonetician was admissible as expert evidence; and that evidence of police officers who listened to disputed tapes and recognised the voice of the person speaking was admissible as factual evidence. Bingham LJ (as he then was), giving the judgment of the court, accepted that the phonetician was sufficiently qualified to give expert evidence on voice recognition. He said of the expert, Dr Baldwin:

“He was entitled to be regarded as a phonetician well qualified by academic training and practical experience to express an opinion on voice identification. We do not doubt that his judgment, based on close attention to voice quality, voice pitch and the pronunciation of vowels and consonants, would have a value significantly greater than that of the ordinary untutored laymen, as the judgment of a hand-writing expert is superior to that of the man in the street.”

Of the police officers’ evidence, the court said that they were entitled to give evidence as witnesses of fact based on their familiarity with the suspect’s voice. Subject to consideration of objections under s.78 of the Police and Criminal Evidence Act 1984 (PACE) the evidence was admissible provided the police officers had acquired the relevant familiarity sufficiently to recognise the suspect’s voice.

Mr Moore, counsel for the respondent, relies on this decision to support his submission that the evidence of the police officers in the case before us, was admissible as evidence of fact.

Clare and Peach is a case on visual recognition. In that case, the court held that a police officer, who had acquired knowledge by many hours spent studying video films of a violent disorder, was entitled to give evidence identifying the appellants in the video films. Giving the judgment of the court, Lord Taylor of Gosforth CJ said:

“The phrase ‘expert ad hoc’ seeks to put witnesses like Detective Parsons and PC Fitzpatrick into the traditional category of those qualified to give expert opinion. Whether or not the tag is appropriate, we are clearly of the view that PC Fitzpatrick had ‘special knowledge that the Court did not possess’ … PC Fitzpatrick had acquired the knowledge by lengthy and studious application to material which was itself admissible evidence. To afford the jury the time and facilities to conduct the same research would be utterly impracticable. Accordingly, it was in our judgment legitimate to allow the officer to assist the jury by pointing to what he asserted was happening in the crowded scenes on the film. He was open to cross-examination, and the jury, after proper direction and warnings, were free either to accept or reject his assertions.”

The evidence in the instant case is not quite the same. We are told that although a control sample of the appellants’ voices was made, the police officers did not use that for the purpose of identification. Their familiarity with the voices of the appellants was acquired by contact with each of them post arrest.

In O’Doherty the court ruled that evidence of an expert using auditory techniques was inadmissible. The court went so far as to rule that save for three exceptions no prosecution should be brought in Northern Ireland based on voice identification given by an expert which was solely confined to auditory analysis. The court ruled that to be admissible such evidence must be supported by acoustic analysis. The three exceptions are first, where the voices of a known group were being listened to and the issue was which voice had spoken which words; secondly, where there were rare characteristics which rendered a speaker identifiable; or thirdly, where the issue related to the accent or dialect of the speaker.

The court went on to deal with evidence of voice recognition given by a police officer. One of the experts who had given evidence to the court had said that identifications made by police officers were, in his opinion, “highly problematic”. He continued, “auditory identification of speakers known to untrained listeners, contrary to popular belief, yields high error rates even under ideal listening conditions”. Dr Holmes echoed this observation in her evidence to us.

In A-G’s Ref (No.2 of 2002) the court considered again the issue of admissibility of identification evidence given by a police officer who had spent much time studying video films before identifying the defendant. The court outlined four circumstances in which such evidence might be admissible. Giving the judgment of the court, Rose LJ said:

“In our judgment, on the authorities, there are, as it seems to us at least four circumstances in which, subject to the judicial discretion to exclude, evidence is admissible to show and, subject to appropriate directions in the summing up, a jury can be invited to conclude that the defendant committed the offence on the basis of photographic image from the scene of the crime:

(i)

where the photographic image is sufficiently clear, the jury can compare it with the defendants sitting in the dock …;

(ii)

where a witness knows the defendant sufficiently well to recognise him as the offender depicted in the photographic image, he can give evidence of this …;

(iii)

where a witness who does not know the defendant spends substantial time viewing and analysing photographic images from the scene, thereby acquiring special knowledge which the jury does not have, he can give evidence of identification based on a comparison between those images and a reasonably contemporary photograph of the defendant, provided that the images and the photograph are available for the jury …;

(iv)

a suitably qualified expert with facial mapping skills can give opinion evidence of identification based on comparison between images from the scene, (whether expertly enhanced or not) and a reasonably contemporary photograph of the defendant, provided the images and the photograph are available to the jury …”

Mr Moore submits that this approach should be adopted by the court when dealing with identification by voice recognition. He submits that the evidence of the police officers in this case was admissible under the second circumstance above, and possibly the third as well.

Although A-G’s Ref (No.2 of 2002) was concerned with visual identification, in our judgment it has some application to the issues in relation to voice identification. We accept Mr Moore’s submission that the evidence of police officers in the circumstances of the instant case is evidence of fact. It would seem from O’Doherty that in Northern Ireland voice recognition evidence by a lay listener in the circumstances existing in this case would almost certainly be ruled inadmissible.

There were three issues which the judge had to determine in respect of the voices recorded on the covert device in the Sprinter van. The first two issues are not the subject of any ground of appeal. The third issue was the question of whether or not the voice recognition evidence of the four police officers was admissible.

The judge had before him the witness statements and reports of Dr Holmes and Professor French. He also heard evidence from the four police officers on a voir dire. That evidence has been helpfully put before us in the form of agreed facts. We summarise this evidence in respect of each appellant separately.

So far as Flynn is concerned, he was seen by DC Gittings on 29 April 2006 on two occasions. First he was seen in the cells following his arrest when his clothing was seized and swabs taken from his hands. This operation lasted 30 minutes. Later on the same morning he was interviewed by DC Gittings, during the course of which he answered “no comment” to all questions asked of him. The interview lasted approximately ll minutes. On the following day Flynn was interviewed a second time by DC Gittings and DC Fleck. The interview lasted 55 minutes. Again, he answered “no comment” to all questions asked of him. The same day he was seen by DC Gittings and DC Fleck, either together or separately, on four short visits.

DC Fleck listened to the covert recording from the Sprinter van on 9 May 2006. It was on that occasion that he claimed to recognise Flynn’s voice. He said that on that occasion he had in his possession the transcript prepared by DC Gittings. DC Gittings listened to the covert recording on 18 May 2006. His evidence was that it was on that occasion that he made the transcript which DC Fleck said he had in his possession on 9 May. DC Gittings said on that occasion he annotated that transcript with the name Flynn against the passages which he identified as being Flynn’s voice. His evidence was that this was the only transcript made by him. It is an agreed fact that there was only one transcript made and that was the transcript prepared by DC Gittings. DC Flynn said that the transcript which he had in his possession, when he listened to the covert recording, had no annotations on it. On 8 June each of these two officers made witness statements.

Turning to St John, DC Nicoll and DC Seymour arrested him at Gatwick Airport on 26 May 2006 and transported him to Loddon Valley Police Station. They were in his company for two hours and 20 minutes. At ll.30am on the same day, both officers together listened to the covert recording from the Sprinter Van. They listened only to a section of the recording which lasted 59 seconds. St John’s voice was said to be a voice speaking over a walkie-talkie radio to the men in the van. The time they spent listening to the recording was not recorded but it cannot have been very long because the next recorded time on which St John was seen by them was at 12.13pm on the same day.

In his ruling, the judge records that both DC Gittings and DC Fleck listened to the covert recording on 9 May. This must be an error, because the agreed facts placed before us show, as we have already stated, that DC Fleck said he listened to the recording on that date but DC Gittings said he listened to the recording on 18 May.

In his ruling, the judge rejected allegations made on behalf of each appellant that the police officers had acted in bad faith. He went on to deal with the submission that in the face of Dr Holmes’ evidence the police officers’ voice recognition was too unreliable to be admissible. He said in a passage, which seems to be determinative of his ruling, the following:

“I agree with Mr Moore that there is no direct contradiction between Dr Holmes’ evidence and that of the officer. On the basis of the material submitted to her, Dr Holmes said it was not possible reliably to attribute anything in the recording to particular speakers, but that was as far as she was able to go. She did not, for example, say that the quality of the tape was so poor that not even someone who had spent some time in conversation with a particular individual would be unable reliably to recognise his voice on the recording.”

In the final paragraph of the ruling the judge said that the criticisms raised of this evidence were perfectly proper points to be pursued in evidence and submissions to the jury but none justified him in holding under s.78 of PACE that the evidence should be excluded.

The first submission on this ground of appeal made by Mr Stephen Batten QC (not trial counsel) is that the judge, in the passage cited above in his ruling misinterpreted Dr Holmes’ evidence. Following the trial fresh counsel and solicitors obtained a further statement from Dr Holmes which formed the basis of the application for leave to appeal which was granted by the full court. The full court further directed Dr Holmes to attend the hearing of the appeal to give evidence.

We must deal first with the basis upon which we received further evidence from Dr Holmes. As already noted, no objection to us hearing the evidence was taken by Mr Moore. As also noted, strictly speaking it is not fresh evidence. However, we take the view that we are able to receive this evidence as necessary and expedient in the interest of justice (see s.23 – (1)(c)).

We go back to the passage in the judge’s ruling referred to above (see paragraph 34). That passage is a reference to two sentences in Dr Holmes’ report dated 5 October 2006 which read:

“Although some of the speech was clearly intelligible much had lower intelligibility or could not be decoded with any degree of certainty. Because of the fragmented nature of the recorded speech and the poor technical quality, it was not possible reliably to attribute utterances to individual speakers.”

In her further statement dated 5 June 2007, prepared for the purpose of the application for leave to appeal, Dr Holmes said:

“A pre-requisite for my making a speaker-identification is that there is sample of adequate size from the disputed recording that can confidently be attributed to a single speaker. This sample forms the basis for the process of analysis. If it is not possible reliably to attribute utterances to individual speakers then the basis for an analysis is lacking. In the present case, on the basis of the analysis carried out by me in the course of producing the orthographic transcript of the disputed material, I made a judgment that I was not able reliably to attribute utterances to individual speakers, and consequently that it was not appropriate for me to carry out a speech comparison analysis.”

It is submitted by Mr Batten that this passage demonstrates that the judge in his ruling acted on a misconception of Dr Holmes’ evidence. Mr Moore, on the other hand, submits that there is no misconception. Relying on a passage in an e-mail from Dr Holmes to the appellant’s counsel shortly before this appeal (properly disclosed to the prosecution) he submits that Dr Holmes’ evidence remains as stated by the judge. Mr Moore relies on a passage in the e-mail which reads “There is no problem in principle with a police officer having an opinion regarding the identity of a speech he/she hears on a recording”.

Having compared the passages in Dr Holmes’ various reports and the e-mail, and having heard her give evidence, we have no doubt that the passage in the judge’s ruling does not fully reflect Dr Holmes’ opinion. As Dr Holmes made clear in her evidence to us, because of the poor quality of the covert recording of the voices in the Sprinter van, she was not able to get to the stage where it was possible to start to analyse the voices by reference to individual speakers. In our judgment the key passage in her second statement is the sentence, “A pre-requisite for my making a speaker-identification is that there is sample of adequate size from the disputed recording that can confidently be attributed to a single speaker”. If she, using her sophisticated equipment, was unable to identify individual speakers it seems to us dangerous to assume that a lay listener could not only distinguish different speakers, but go on to identify accurately the voice of one individual speaker. Dr Holmes in her e-mail added to the passage already quoted from the e-mail the following:

“Further, there is no reason why a particular officer’s recognition of a voice should not be used as a basis for further investigation (no reference specifically to the present case; I am thinking more generally of e.g. cases involving hoax telephone calls made by someone already known to the police). The question here is rather, whether the opinion of a police officer, who is in this context a lay listener, should have any evidential weight in the absence of an expert opinion. The points raised in 1. above should give an indication of the potential issues raised.”

The issues raised above are factors to which we have already referred in the passage in this judgment summarising general comments made by the experts on lay listener voice recognition.

We approach the judge’s decision on the basis that he did misconstrue part of Dr Holmes’ report. We attach no blame to the judge for this. Neither the prosecution nor the defence sought to call Dr Holmes on the voir dire. This was in our view an error. Had Dr Holmes been called and her evidence explored, in our opinion, the judge would have been in a far better position to evaluate the probative value of the police officers’ evidence.

It is unnecessary for us to recite the general matters upon which Dr Holmes gave evidence. We have incorporated these matters into the general factors outlined above (see para 16). Specifically, in respect of the recognition by the police officers of the appellants’ voices on the covert tape, Dr Holmes said that the tape was sub-optimal and even by that standard poor. In her opinion the quality was such that it was not suitable for speech analysis. She said that before one can start to compare a disputed voice with a known voice one must be able to differentiate one voice from another on the disputed tape. In this case she was unable to distinguish the different voices on the covert tape and was not able reliably to distinguish patterns of speech. As to DC Gittings’ transcript, she said that although it was available to her when she made her transcript, she made her own before evaluating it against DC Gittings’ transcript.

Dr Holmes added that both she and Professor French used auditory and acoustic techniques. It was agreed that her equipment used to listen to the covert tape was more sophisticated than that used by the police officers. The latter was a standard issue Metropolitan Police laptop.

Dr Holmes said that recording by means of a telephone device, as had been done in this instance, further distorts the voice.

In his written report, Professor French stated that in his opinion it was not possible to work from a raised voice, as identified by the police officers, “backwards” to conclude, as the police officers had done, that the person speaking would have a quiet or high-pitched voice when speaking normally. This was a reference to the voice speaking over the walkie-talkie to the men in the van. In his view the most that it is safe to conclude is that the voice identified by police officers as that of St John, could be his voice. In his report he stated “it is an open possibility; I cannot eliminate him”.

In our judgment, central to the decision of whether the police officers’ evidence was admissible is the issue of whether it had any potential probative value. Mr Moore submits that it was relevant evidence and the judge was bound to rule it admissible. It was for the jury to evaluate its probative value in the light of the criticisms of it and the expert evidence. He further submits that when considering the application to exclude the evidence under s.78 of PACE the judge exercised his discretion in a way which was neither unreasonable nor flawed in the Wednesbury sense.

On the basis of the expert evidence in this case and the agreed evidence relating to the police officers’ experience of the voice of each appellant, our conclusion is that the evidence of voice recognition by the police officers should have been excluded by the judge.

It is quite clear that the covert recording of voices in the Sprinter van was poor. It was so poor that Dr Holmes, even using her sophisticated equipment was unable to distinguish between different voices. Comparing her transcript with that of DC Gittings a number of the words which DC Gittings purported to recognise as Flynn’s, are not distinguished by Dr Holmes as words. Professor French described the recording as poor and, as already stated, he was not able to rule out the possibility that the voice attributable to St John was either his or not his.

So far as the evidence of DC Gittings and DC Fleck is concerned, their familiarity with Flynn’s voice was gained from comparatively short periods of time in his company on two days. It was eighteen days before DC Gittings listened to the covert recording and compiled the transcript. In DC Fleck’s case, accepting his evidence which is in complete conflict with the Gittings evidence as to when the transcript was made, it was at least nine days after he had last heard Flynn speak before he listened to the covert recording.

So far as St John is concerned, the police officers listened to the covert recording very soon after they had been with him for a period of two hours twenty minutes. To that extent they were in a better position than DC Gittings and DC Fleck to identify his voice on the covert recording. However, DC Nicoll and DC Seymour had a limited time to familiarise themselves with St John’s voice. Furthermore, the part of the covert recording alleged to be attributable to him lasted only 59 seconds of a poor tape and was of a voice itself said to be communicated to the van via the further distorting medium of a walkie-talkie.

There are other considerations which apply to both appellants. We have already noted the two most important factors, namely the limited opportunity for the officers to acquire familiarity with the appellants’ voices and secondly, the poor quality of the covert recording. To these we add the following. First, the police officers’ purported recognition of the appellants’ voices is in marked contrast to the evidence of the two experts who are unable to recognise their voices; in Dr Holmes’ case she was unable to identify individual voices. Secondly, nothing is known of the ability of any of the police officers to recognise voices. There is no evidence that any of them had any training in auditory analysis. Thirdly, the identification of the voices was carried out by listening to the covert tape on a standard laptop computer, as opposed to the sophisticated equipment used by the experts. Fourthly, in our judgment, it is significant that DC Gittings’ transcript contains words attributed to Flynn which Dr Holmes could not distinguish as words, let alone recognise as attributable to Flynn. Fifthly, the expert evidence shows that lay listeners with considerable familiarity of a voice and listening to a clear recording, can still make mistakes.

Taking all these factors into account we conclude that there are powerful factors militating against the admission of the evidence of the police officers in the case of each of the appellants. Such evidence was self-evidently very prejudicial. As the experts point out, there are difficulties in challenging the evidence of lay listeners for the reasons noted earlier in this judgment. In our judgment, the general uncertainties about the evidence of voice recognition by lay listeners are enhanced by the specific facts in the case of each of these appellants. Quite simply, in our opinion, the prejudicial effect of this evidence far outweighed its probative value. Accordingly we conclude that the judge was wrong to rule it admissible.

There are other reasons why, in our judgment, the judge ought also to have excluded the evidence under s.78 of PACE. First, in our opinion, when the process of obtaining such evidence is embarked on by police officers it is vital that the process is properly recorded by those officers. The amount of time spent in contact with the defendant will be very relevant to the issue of familiarity. Secondly, the date and time spent by the police officer compiling a transcript of a covert recording must be recorded. If the police officer annotates the transcript with his views as to which person is speaking, that must be noted. Thirdly, before attempting the voice recognition exercise the police officer should not be supplied with a copy of a transcript bearing another officer’s annotations of whom he believes is speaking. Any annotated transcript clearly compromises the ability of a subsequent listener to reach an independent opinion. Fourthly, for obvious reasons, it is highly desirable that such a voice recognition exercise should be carried out by someone other than an officer investigating the offence. It is all too easy for an investigating officer wittingly or unwittingly to be affected by knowledge already obtained in the course of the investigation.

This case provides an example of why these minimal safeguards should have been observed. No notes were made by any of the police officers of the dates and times when they attempted to recognise the voices on the covert recording from the Sprinter van. There is an important inconsistency in the evidence of DC Gittings and the evidence of DC Fleck. If the former is correct, he did not produce the transcript until 18 May 2006, nine days after DC Fleck said he had the transcript in his possession when he listened to the disputed tape. This is not a minor discrepancy. DC Fleck said the transcript he had on 9 May did not contain DC Gittings’ notes in the margin indicating which passages he attributed to Flynn. Yet, DC Gittings said he only produced one copy of the transcript and on it he placed his opinion as to who was speaking. It is agreed there was only one transcript produced and no copies were made.

It is not clear if the judge, when he gave his ruling, was aware of this discrepancy. It does not feature in his ruling. However, the discrepancy further hindered the ability of counsel for Flynn to challenge the accuracy of DC Fleck’s evidence. This discrepancy and the failure of the officers to record essential details of how and when the transcript was compiled, and the dates when each listened to the tape, coupled with the dangers to the use of voice recognition evidence by lay listeners, are all sufficient in our judgment to make it unfair for the evidence to be admitted. For these reasons we would also have excluded this evidence and in our judgment the judge was wrong not to exclude it under s.78 of PACE.

The appellant submits that the judge misdirected the jury by instructing them that they should not attempt to compare the voices heard on the covert recording with the voices of the appellants which they had heard when they gave evidence in the trial. Apart from the decision in R v Chenia [2003] 2 Cr.App.R. 6 there is no decision which supports the direction given by the judge. On the contrary, there are passages in other authorities, including those referred to above, which suggest that the jury should be permitted to make such a comparison providing the judge directs the jury to listen to the tapes guided by the evidence of the voice recognition witnesses, whether expert or lay listeners.

Mr Moore conceded that the direction given by the judge in this instance was wrong. We agree, but point out, in fairness to the judge, that he gave the direction at the instigation and agreement of all counsel.

Mr Moore’s alternative submission was that this misdirection does not render the jury’s verdicts unsafe. We consider this latter submission in conjunction with a similar submission made by Mr Moore in respect of ground 1.

Mr Moore urged on the court a submission that if we upheld the appellants’ submissions on either or both of the grounds of appeal we should nevertheless find that the verdicts of the jury were safe. He submits that the case against each of these appellants was a strong one.

We accept that the case against each appellant was strong, particularly the case against Flynn. However, we are unable to conclude that the verdicts are safe. As we have already stated, the prejudicial effect of the voice recognition evidence given by the police officers must have been considerable. Without this evidence the prosecution’s case against both appellants relied on circumstantial evidence. We find ourselves quite unable to say that if the voice recognition evidence had been excluded the jury would inevitably have reached the same verdicts.

For the reasons explained in this judgment we allow the appeals of both appellants and quash their convictions.

As appears from the above we have been dealing in these appeals with issues arising out of voice recognition evidence. Nothing in this judgment should be taken as casting doubt on the admissibility of evidence given by properly qualified experts in this field. On the material before use we think it neither possible nor desirable to go as far as the Northern Ireland Court of Criminal Appeal in O’Doherty which ruled that auditory analysis evidence given by experts in this field was inadmissible unless supported by expert evidence of acoustic analysis. So far as lay listener evidence is concerned, in our opinion, the key to admissibility is the degree of familiarity of the witness with the suspect’s voice. Even then the dangers of a mis-identification remain; the more so where the recording of the voice to be identified is poor.

The increasing use sought to be made of lay listener evidence from police officers must, in our opinion, be treated with great caution and great care. In our view where the prosecution seek to rely on such evidence it is desirable that an expert should be instructed to give an independent opinion on the validity of such evidence. In addition, as outlined above, great care should be taken by police officers to record the procedures taken by them which form the basis for their evidence. Whether the evidence is sufficiently probative to be admitted will depend very much on the facts of each case.

It goes without saying that in all cases in which the prosecution rely on voice recognition evidence, whether lay listener, or expert, or both, the judge must give a very careful direction to the jury warning it of the danger of mistakes in such cases.