APPEAL COURT, HIGH COURT OF JUSTICIARY
	[2013] HCJAC 145
Lord Eassie Lord Menzies Lord Bracadale	Appeal No: XC896/07   OPINION OF THE COURT   delivered by LORD MENZIES   in   the Reference from the Scottish Criminal Cases Review Commission   by   THOMAS ROSS YOUNG   Appellant;   against   HER MAJESTY'S ADVOCATE   Respondent:   _______

 

Appellant: Kerrigan, QC; Latif; Gilchrist; Drummond Miller LLP (for George Mathers & Co, Aberdeen)

Respondent: Ogg, QC, AD; Crown Agent

 

15 November 2013

 

Introduction

[1] The appellant was convicted at the High Court in Glasgow on 25 October 1977 of eight charges on an indictment, including a charge of murder, two charges of attempted murder, two charges of rape, a charge of assault and robbery, a charge of assault and a charge of theft. Each of these offences was committed between May and July 1977, in Glasgow or the surrounding area. All but one of the offences involved violence towards women, and many involved a significant sexual element. The most serious charge was in the following terms:

"(11) Between 10 and 27 June 1977, both dates inclusive, at the service road leading to Inchneuk Farm, Glenboig, or elsewhere in Scotland to the prosecutor unknown, you did assault Frances Barker, 289 Maryhill Road, Glasgow and did compress her throat, tie a ligature around her neck, push a pair of pants into her mouth and did murder her".

 

[2] On 2 November 1977 the appellant lodged a note of appeal against conviction, but on 3 January 1978 this appeal was abandoned.

[3] By letter dated 12 September 2005 Strathclyde Police wrote to the appellant's solicitors explaining that Strathclyde Police and Lothian and Borders Police were currently undertaking "Operation Trinity", a joint investigation into the murders of six women in 1977, including the murder of Frances Barker, and stating that the police, at the request of the procurator fiscal, wished to interview the appellant in the presence of his solicitor about the murder of Frances Barker. The police interviewed the appellant, in the presence of his solicitor, on 20 September 2005. As a result of this interview, the appellant formed the view that the police no longer believed that he murdered Frances Barker. If the safety of the murder conviction was open to question, this called into question the safety of the other convictions, because the appellant was convicted on the basis of the Moorov doctrine. Accordingly, the appellant made an application to the Scottish Criminal Cases Review Commission for review of his conviction.

[4] In the course of the Commission's consideration of the appellant's application, on 12 March 2007 Crown Office forwarded to the Commission a copy of the report Strathclyde Police had sent to Crown Office in respect of the Operation Trinity investigation, in which it is stated that the police had re-investigated the murders of six young women, including Frances Barker, between 11 June and 2 December 1977, all the women having gone missing following social nights out and whose bodies were subsequently discovered dumped in the open air a substantial distance from where they had been last seen alive. Reference was made in the report to what were described as "unique similarities" between the murders, and the report expressed the conclusion that the same person(s) were responsible for all murders within a short time frame. Appended to the Operation Trinity report were three further reports, namely a report by John Clark and Marjorie Black, forensic pathologists of the University of Glasgow dated 16 December 2004 ("the forensic pathology report"); a report by the Behavioural Analysis Unit at the Federal Bureau of Investigation, Quantico, Virginia, USA dated 23 January 2006 ("the BAU report"); and a report in which data relating to all homicides of female victims in Scotland between 1968 and 2004 have been collated, which is undated and unsigned ("the SCOTHOM report").

[5] The hypothesis which underlay the conclusion of the Operation Trinity investigation was that it may be possible to link crimes to form a series on the basis of perpetrator behavioural similarity, to enable a conclusion to be reached as to the likelihood of the crimes being perpetrated by the same person or persons. Although in the course of these proceedings this practice was referred to by several names, we propose to refer to it throughout as "Case Linkage Analysis" ("CLA").

[6] Strathclyde Police apparently took the view that CLA might be of assistance to them in the conduct of their Operation Trinity investigations. The Commission observes in its reference to this court (at paragraphs 117-119) in dealing with the BAU Report:

"117. On what is page 1 of the Commission's copy of the report (the report is unpaginated), it is stated that Strathclyde Police requested an analysis and an opinion from the FBI on whether 'linkage' between six homicides could be established. The purpose of the report is set out more fully at page 3, where it is stated:

'The purpose of this analysis was to behaviourally assess the overall condition of each crime scene and the dynamics of the offender(s) interaction with that scene (both contact and deposition sites) and the victim in order to determine if a linkage could be made between any of the six homicides'.

 

118. Strathclyde Police provided the FBI with aerial photographs of the crime scene and contact locations; detailed maps of the crime scenes and deposition sites; overall maps identifying the contact and deposition sites for each victim and the relationship of each site to the other sites; photographs of the last seen locations; photographs of the crime scenes, including the bodies; autopsy photographs; detailed photographs of the bindings, the ligatures and the gags used on the victims; photographs of the victims; a map highlighting the dimensions of Scotland in relation to the US and other countries in the UK; a map demonstrating the geographical relationship between Glasgow and Edinburgh; a map highlighting central Scotland and the contact and deposition sites; virtual representations of the bindings, the gags and the ligatures from the front and from the rear; and a photographic summary demonstrating the similarities in the crime scenes, the ligatures used to gag the victims, the ligatures used to bind their wrists and ankles, and the body positions.

 

119. At pages 10 and 11 of the report, under the heading 'Crime Analysis', it is stated:

 

'Violent crimes scenes always tell a story - a story written by the offender, the victim and the unique circumstances of their interaction. It is this interaction that makes crime scenes dynamic events that vary in complexity. One of the fundamental tenants [sic] of crime scene analysis is to identify the multiple forensic and behavioural variables presented at the scene in order to interpret how and why these six women became the victim of homicide....While conducting an analysis of the behaviour manifested at violent crime scenes, it is essential to avoid becoming too focused on any one aspect and ascribing singular importance to it. It is important to remember that no attempt was made to reconstruct the exact sequence of events. Instead, it is the totality of the circumstances that is most important in assessing not only what happened but why and how it happened".

 

[7] The Commission considered the various reports which were placed before it, and the appellant's statement to the Commission. It addressed the test for additional evidence, and concluded that the terms of the Operation Trinity report were likely to have had a material bearing on, or a material part to play in, the jury's determination of the critical question which was put to them by the trial judge, namely "the question which you have to decide is whether you are satisfied beyond reasonable doubt that the accused is the person who murdered her" [i.e. Frances Barker]. The Commission concluded that there may have been a miscarriage of justice and that it was in the interests of justice that the case should be referred to this court. Accordingly the Commission made a reference to this court in November 2007.

 

Procedure

[8] The appellant lodged grounds of appeal on 6 June 2008. Despite the best efforts of the court to make progress, there then followed a very protracted procedural history, involving several changes of senior counsel and prolonged investigation. It is not necessary for us to explain here the reasons for this delay. In February 2011 the court appointed the appeal to a full hearing of 8 days duration to be held no earlier than September 2011 and appointed both parties to lodge written notes of argument before that hearing. In November 2011 amended grounds of appeal were lodged on behalf of the appellant.

[9] The ground of appeal lodged in June 2008 was in the following terms:

"That there has been a miscarriage of justice in this case in respect that new evidence not heard at the trial of the appellant is now available which would have had a material bearing on the jury's determination on charge 11 in the indictment. The new evidence is to the effect that the crime in charge 11 on the indictment facing the appellant may have been committed not by the appellant, but by X and/or Y".

 

[10] The evidence relied upon in support of this ground of appeal in the original grounds comprised (1) the forensic pathology report, (2) the BAU report and (3) the SCOTHOM report, together with two other evidential matters. The amended grounds of appeal deleted these two other evidential matters and substituted four other evidential matters, the last of which was in the following terms:

"(vii) The evidence recovered and findings made by the Lothian and Borders and Strathclyde Police in the investigation Operation Trinity, as contained in the Operation Trinity report, that

(a) could reasonably infer that the same person was responsible for the six murders under review; and

 

(b) could reasonably infer a circumstantial link between the target of the Report, X and the six murders under review".

 

[11] CLA has accordingly not only been an important element in the Operation Trinity report and the Commission's conclusion, but has formed a very substantial part of the grounds of appeal since 2008. Throughout the appeal procedure since the reference by the Commission, the Crown has adopted the position that CLA is not evidence that could competently have been admitted at the original proceedings nor is it now admissible in the appeal. At a preliminary hearing in January 2012 the court appointed the appeal to a hearing to determine the admissibility of the expert evidence to proceed, preferably in May 2012, and noted that such hearing would hear evidence from two witnesses only, these being Dr Jessica Woodhams and Professor David Canter, followed by legal submissions thereon. Further procedural difficulties then occurred, and the hearing on the admissibility of CLA evidence did not take place until the summer of 2013.

 

The evidence

[12] On 29 April 2013 parties lodged a substantial joint minute extending to some 68 pages, the purpose of which was to agree the facts and circumstances surrounding the murders of Frances Barker and six other women as far as possible from the information led at any trial or ingathered by the Crown. At the hearing, evidence was led on behalf of the appellant from Professor David Canter, who spoke to his opinion dated 22 March 2013, and on behalf of the respondent from Dr Jessica Woodhams, who spoke to her report dated 7 May 2013.

[13] Professor Canter was aged 69 and Professor of Psychology at the University of Huddersfield, where he directs the International Research Centre for Investigative Psychology. Before moving to the University of Huddersfield in 2009 he had held a similar post at the University of Liverpool, where he is also an emeritus professor. He is an honorary fellow of the British Psychological Society, a fellow of the Royal Society of Medicine, a fellow of both the South African and American psychological associations, and is a chartered forensic psychologist and a member of the Forensic Science Society. He has published over 30 books and written numerous articles in peer reviewed journals, as well as having edited several journals. Over the past 25 years he has been involved as a consultant to over 100 police investigations, and has given evidence in court and to government inquiries concerning a variety of aspects of human behaviour in criminal circumstances. One of his particular areas of study has been patterns of criminal actions. He has published on the patterns of behaviour in sexual assault, rape and homicide, advised police investigations in these matters and given presentations on this work at academic and professional conferences around the world.

[14] Professor Canter stated that he was first asked to look into aspects of behavioural science which might be of use in police investigations in the late 1980s. One aspect of this was whether a series of crimes might have been committed by the same person. This aspect of behavioural science has really developed over the last few years, and CLA is beginning to develop a body of literature and scientific activities which is recognisable, and which has exciting possibilities. He was asked on what basis he described CLA as a science; he replied that the components of any proposals can be clearly and objectively stated, and open to challenge against empirical evidence. If behaviour is completely random and displays no consistency at all, CLA will not work, but there tends to be a level of consistency in offenders' behaviour. It is necessary to select the appropriate, salient behaviour for comparison, and to focus on some behaviours that are less frequent than others. Relevant factors include geographical location, proximity and temporal factors. Training is required in order to collect the appropriate information, make sense of it, and then carry out statistical tests - there may be more than one way of measuring a particular similarity.

[15] CLA begins with the process of identifying constituents of the circumstances of a set of crimes of any particular class; there may be as many as 50 or 100 actions, referred to as "variables", from many crimes committed by many different people. A procedure is then carried out to ensure that each of these variables is clearly defined in such a way that two independent experts would consistently assign the same actions to the same variables - a process known as establishing "inter-rater reliability (IRR)". This results in a "content dictionary" that defines all the variables relevant to any class of crime. The content dictionary may also provide information on the frequency with which each of the variables occurs across the total sample of cases on which it is based. This frequency is known as the "base rate". This is of central importance to CLA. Professor Canter suggested that it was of assistance to the triers of fact to have the detailed knowledge and expertise available from CLA, in order to form a judgment as to which aspects of the character and circumstances of crimes should be used to connect them.

[16] Another aspect of CLA is the identification of dominant themes in a particular series of crimes. To take the example of homicide, actions that are to do with sexual activity can be distinguished from those relating to dismemberment of the victim's body, the form the actual killing takes, and aspects of theft surrounding the crime. These components provide the constituents of any behavioural profile of a killing. Professor Canter maintained that to identify the components underlying any set of variables requires the sort of scientific study that goes beyond the knowledge and experience available to those outside this area of expertise. As with intelligence tests, it is based on developments in sophisticated statistical procedures.

[17] A variety of procedures are available for measuring the degree of linkage between two or more crimes. Professor Canter emphasised that these measures rarely give a definitive, very precise answer of the form that may be expected from DNA comparisons. Rather they give an indication that can assist the court in its decision making. There are two broad approaches to measuring linkage. The first is the nomothetic approach such as the index developed by Jaccard in 1901, which measures the frequency of co-occurrence of all occurrences. There is no agreed criterion for when a Jaccard's index can be taken to demonstrate that crimes are linked. Instead it has been used for background research to measure the co-occurrence of each of the components of crimes known to be linked. This has produced indications of which components may be most helpful in indicating that crimes are linked. The behaviours that have been found useful in this regard vary from study to study and crime type to crime type. The lack of very precise findings may be due to the unreliability or imprecision of the original information on which the analyses were conducted, or because of contingency influences on a criminal's actions that make absolute consistency and differentiation very unlikely. However, the results are clear enough to be of value to the court, if their context is clearly understood.

[18] The second approach to measuring linkage is an ideographic procedure which is shaped by features special to those individuals or their circumstances that are under study. It draws on particular aspects of the cases in question and examines their similarities and differences, and can take account of what is known about alleged offenders. This approach is described by some as "decision support" because it is the procedure for contributing to ongoing investigations.

[19] Professor Canter went on to explain, with the aid of a diagram, how a Receiver Operator Characteristic ("ROC") is compiled. This is a mathematical procedure developed by engineers to indicate the effectiveness of a decision system. The vertical axis measures sensitivity (or "frequency of hits"), and the horizontal axis measures specificity (or "frequency of false alarms"). A number of occurrences displaying similarities and differences are placed on this table, and a curved line of thresholds may then be seen. The area under the curve ("AUC") is then used to quantify the degree to which it is possible to link crimes. The shape of the curve is of importance - the larger the AUC, the more efficient is the system. However, Professor Canter accepted that no system of CLA was completely reliable or efficient, and the problems identified in the literature on the subject (eg Home Office Research Study 215 published in 2001, which observed that "the problems inherent in the data used in this study...cannot be overstated") remained and must be borne in mind. Nonetheless, he considered that CLA was of assistance in linking crimes for the purpose of police investigations. He was asked if this science was sufficiently developed that it can be relied on, and he replied that if it is carried out under the proper conditions it can give guidance, but it is "decision support" - it does not answer the final question. In order to qualify as an expert in this field one would require at least a Masters degree in investigative psychology or general forensic psychology which would touch on this area. An expert could assist an appeal court by drawing to the court's attention the salient facts, and any similarities in patterns which were revealed, and by offering an opinion as to the likelihood that these components indicated that the same offender was involved. This evidence might also assist a jury at a trial, and although Professor Canter anticipated that on the first occasion that this evidence was presented to a jury the court would require to establish the credentials of the science, thereafter he considered that the major issues could be quickly explained.

[20] Professor Canter regarded himself as an expert in this area, and also regarded Dr Woodhams as an expert in this area. He was familiar with her writings, which he described as more theoretical and academic, rather than addressing the implications and challenges for the real world. He was asked to comment on her report dated 7 May 2013, and her opinion that, at present, case linkage evidence is unsuitable for admission into legal proceedings; he considered that there was enough accuracy, if used appropriately and in the right context, to be of assistance to the court. He could think of some cases where case linkage was unsuitable for admission into legal proceedings, and some cases in which it would be suitable. The views which Dr Woodhams expressed at chapter 5 of her report, on Professor Canter's conclusions, suggested that she was not aware of all the studies which he had carried out. He described her as a very serious and focused researcher, but she is not explaining the implications of what she is doing in a way which reveals the power of her own research.

[21] In cross-examination Professor Canter acknowledged that he had never given evidence on CLA in any court. He was asked if anyone had ever given evidence of CLA in any court in the world, and he observed that some of the research he was considering had only just been published, and that we were only just coming to the point where it could be placed before a court. He had searched his sources and contacted colleagues around the world, and concluded that it was very likely that CLA evidence has never yet been used in any court anywhere in the world. He observed that CLA certainly has more obvious value in contributing to police investigations, and it has been shaped by and grew out of these investigations, but he was now discussing whether it was robust enough to be used in court. He accepted that the error rate was not quite so important if it was only being used as an investigative tool rather than as evidence before a court. He also accepted that the best AUC performance that he had seen is 75%; this means that, using the best methodology, 25% of cases which are shown as being linked are not in fact linked. There was no way in which a jury can decide, in a particular case, whether it falls into the category of being correctly linked with other cases or incorrectly linked with other cases - but the CLA evidence would require to be placed in the context of all the other evidence, and no CLA expert would give a clear decision as to whether one offence was or was not linked to another. He accepted that there was a large error rate in CLA research at present, and that this was quite different from other areas of evidence such as DNA or fingerprint evidence. He accepted that there would be a substantial number of cases where the CLA evidence "got it wrong".

[22] It was put to him that the methodology involved working with known, closed cases in which the offender had been convicted on the basis of evidence other than CLA; he agreed that this was the usual process, and that this was a weakness in the whole area of criminal research, which was only dealing with solved crimes. He accepted that there was a potential for bias in the methodology - there may have been a very stable set of behaviours by an offender which led the police to him. He understood why Dr Woodhams expressed the view that it is a step too far to use this evidence as probative, in the real world. However, he had two concerns about awaiting further research and development and improvement of the methodology, and tidying up CLA as a proper professional discipline - (1) there are cases where one can be confident that the error rate will be low, and (2) he was aware that at present many wrong assumptions are made about behaviour and similarities between offences.

[23] Professor Canter agreed that it was correct that some of the linked crime pairs in the research have been shown to be as inconsistent or dissimilar as the unlinked crimes, and that some unlinked crime pairs have a high degree of similarity when using CLA methodology. On occasions there are crimes committed by two different people which bear a striking number of similarities.

[24] Dr Jessica Woodhams was aged 35 and a senior lecturer in forensic psychology at the University of Birmingham. She holds a Bachelor degree in psychology, a Masters in forensic psychology and a doctorate in child sexual offending. She is a registered forensic psychologist, a chartered psychologist and has published works in forensic psychology. Her primary area of research is CLA. Her first job was as a crime analyst with the Metropolitan Police, in which she carried out CLA in about 30 cases, the majority of which were sex offences against strangers. Much of her research since then has been in case linkage of sex offences. She spoke to her report dated 7 May 2013.

[25] Dr Woodhams expressed the conclusion that the two theories underpinning the practice of case linkage, the theories of behavioural consistency and behavioural distinctiveness, have received empirical scrutiny from academic researchers across the globe. She was of the opinion that these studies suggest some support for the theories of behavioural consistency and distinctiveness; however, they also indicate a degree of error in making predictions about whether crimes are the work of the same offender. In samples of solved and unsolved serial crime, some linked crime pairs (two crimes by the same offender) are characterised by behavioural inconsistency meaning that with such offenders case linkage would not be accurate in practice. Further the behavioural similarity shared by some unlinked crime pairs (two crimes by two different offenders) is so high that it could lead a practitioner erroneously to consider them the work of the same serial offender. She expressed the opinion that, at present, case linkage evidence is unsuitable for admission into legal proceedings.

[26] Dr Woodhams explained that CLA requires the compilation of a coding dictionary or checklist of behavioural features, against which each offence may be checked for points of similarity or dissimilarity. The person conducting the coding is known as the "rater". Different raters may code behavioural features differently, hence it is important to establish a system of inter-rater reliability. For example, the Violent Crime Linkage Analysis System ("VICLAS") has been developed by the Serious Organised Crime Agency. Rating (or coding) under this system is performed by assistant crime analysts who have received thorough training, and their rating is checked by another member of staff. However, rating is normally based on information obtained by the police from a victim; if this is not possible, it may be based on a statement from the first witness on the scene, from CCTV footage, photographs etc. However reliable the rating is, if it is based on poor information the result will be poor.

[27] The assessment of similarity of offences is usually performed using Jaccard's coefficient. If there is a perfect match, this would result in a coefficient of 1, and if there was no match at all, this would result in a coefficient of zero. However, experience has shown that even if a coefficient of 1 is achieved, it does not follow that the crimes were committed by the same person - there have been cases where the highest possible score was achieved, yet as a matter of fact the two crimes had been committed by different people. Equally, there were instances of crimes which were not linked because they contained so few similarities, but which had in fact been committed by the same person - Dr Woodhams had seen Jaccard scores of zero, suggesting no linkage, when it was known that the crimes had been committed by the same person. She regarded this as one of the weaknesses of the system, and it was one of her main concerns about such evidence being admitted in court.

[28] Dr Woodhams also had concerns about the ecological validity of CLA. A study is ecologically valid if it approximates the real world that is being examined. If it does, the findings can be applied to practice. CLA can be applied to large populations of crimes, but it is less reliable with regard to specific cases in real life. There has been no research on the efficacy of CLA predictions in real life. The checklist/dictionary varies from one researcher to another, and from one country to another. An offender's behaviour may be influenced by his victim's behaviour or reaction. She gave an example of a single male offender who engaged in a series of four rapes, pointing a gun at each victim to overcome their resistance. If three of the victims' resistance was overcome by this means, but the fourth victim continued to resist and was battered to death, the fourth crime would appear sufficiently different that, using CLA methodology, it would not be linked with the other three crimes. CLA is, Dr Woodhams said, still so undeveloped that it cannot be confidently asserted that the theory works in practice, and even when it appears to be achieving successful results, there is a high and inexplicable error rate.

[29] Dr Woodhams was of the opinion that CLA was presently useful in guiding police investigations, particularly investigations which crossed police boundaries, in that it could lead the police to real evidence. However, she was unhappy about the use of CLA as evidence in court, at least in its present state of development. There is at present no certification or charter mark in CLA, and there are no international standards. There is no agreed procedure for CLA within the UK, nor in other countries.

[30] Dr Woodhams explained that she had been awarded a grant to research examples of unsolved rape cases which are linked by DNA across seven different countries. She has gathered a team of leading academics from Canada, America, Finland, the UK, Belgium and the Netherlands, as well as several leading analysts working for police services around the world. She explained that she had not selected Professor Canter to join this group, because he has not published a research paper on CLA since 2002. She regarded him as a much respected figure, but not a current researcher.

[31] She explained that predictions as to linkage are made by computer software - there was no human input after the statistical work had been completed. However, there is no standard method of assessing frequency - there are different methods, and different error rates. There is not yet the level of knowledge for a standardised, agreed system.

[32] In cross-examination Dr Woodhams confirmed that she would not be happy about guilt being determined in court on the basis of CLA on its own, because there would still be questions about its reliability. Even with other evidence she has concerns about its use in court because of the potential for mistakes. She was asked if CLA would return a figure greater than chance, and she replied that in all her reports, the figures are better than chance, but chance is simply 50%, and she considered that one ought to have accuracy rates much higher than that. It was put to her that despite the shortcomings of CLA, it might still be an aid, an assistance to the court, but she disagreed. The research now available shows that the assumptions in CLA do not always hold, and indicates that it will not be reliable in all cases. She regarded it as a useful investigative tool, but even in that context there are issues about assumptions and conclusions; there is a potential for mistakes, and for the police to be led down the wrong avenue. For CLA to be a useful tool in evidence before a court, it would have to be proved to be reliable - and, moreover, reliable not in the context of a research study but in a real life analysis. Her concern was that in the present state of knowledge it would not be useful because it might mislead the court.

 

Submissions

[33] Both parties helpfully provided the court with written outline submissions, which we have taken into account in reaching our decision. The position adopted by each of the parties, in their written and oral submissions to the court, may be summarised as follows:-

 

Submissions for the respondent

[34] The advocate depute submitted that CLA, as a technique in its infancy, cannot at present be categorised as a recognised branch of knowledge sufficiently reliable to be admitted for court purposes. At best it is an investigative tool which might be usefully deployed at the investigative stage of a criminal case in an effort to determine whether individual cases might be linked and with the ultimate aim of identifying "real" evidence upon which a prosecution of those linked cases can be brought.

[35] CLA is a theoretical approach to analyse the likelihood of any series of crimes being committed by the same person. The Crown submits that the theory is unproven and only recently tested. It relies on two assumptions, namely behavioural consistency and behavioural distinctiveness. In themselves these assumptions are not very controversial, and it is accepted that there is some evidence to support them. However, the issue for the court is whether such assumptions can be carried over in a checklist which, after weighting, can be reliably used by statistical software to predict correctly whether two or more crimes are linked. The theory suggests that this can be done, although with an error rate. This rate of error is not yet explained, nor can it be improved upon at present.

[36] Both witnesses accepted that (1) there is an error rate, (2) it is high, (3) it cannot be explained by the theory, and (4) it allows for any case to be a false positive, even if checklists show identical scores and results. It follows that it is not possible to attribute a percentage degree of certainty to the figures, nor to indicate a percentage likelihood that they may be correct, because CLA includes false positives even for the best performance. There is a known high error rate in every application of the theory. This was one of the reasons for Dr Woodhams' conclusion that CLA is unreliable at present for probative (and even investigative) purposes. The theory is unproved, and the underlying assumptions are still unproved in a criminal context. More research and testing may result in the theory being disproved and the assumptions destroyed. Dr Woodhams' conclusion was that any such evidence would be unreliable, and should not be admitted as it may mislead a jury (or an appellate court). Evidence which is inherently unreliable should never be placed before a jury or fact finder. To admit such evidence would be to invite a jury to speculate, rather than to deduce. Any evidence with a high error value which cannot be explained or weighted cannot be admissible in our law.

[37] CLA research carried out to date has been conducted in an artificial environment dealing with closed, solved cases; we do not know whether the techniques can be transferred to the real world. Dr Woodhams has real life experience of CLA in The Metropolitan Police, and gave evidence of the way that raters can make errors or can rate or code offences differently from other raters. There is therefore evidence of inter-rater unreliability.

[38] The best that can be achieved by CLA is to say that a set of crimes may be linked or may not be linked. In the present state of knowledge it would be impossible to lead CLA evidence before a jury that the appellant could not have committed the index offence because it was committed by another, or even that there was a real doubt on this point. Even if the best researchers adopted the best practices in the best conditions, CLA evidence would still be unreliable as the results might be a false positive. Both witnesses agreed that there was no such thing as a unique signature or a unique group of behaviours.

[39] The advocate depute pointed out that Professor Canter has not been involved in research in recent years, whereas Dr Woodhams has been, and continues to be, involved in research, and collaborates with other researchers and experienced crime analysis practitioners. What, the advocate depute asked, would a jury make of Professor Canter's evidence, and how could they become the masters of it? Professor Canter cannot give a weight to whether crimes are linked or not, and if he cannot, a jury cannot on the basis of his evidence. As Professor Canter accepted, CLA results are not remotely equivalent to DNA results in which a jury may be told that the odds of a sample originating from someone other than an identified person are (for example) one billion to one - each CLA result has an equal value to every other CLA result, and computer software will always make the same prediction on the codings provided.

[40] There has been no research carried out as to whether the results of CLA are better than chance, and Dr Woodhams (and other researchers to whom she referred in evidence) criticised the unreliability of inter-rater systems. Moreover, there has been limited CLA research in relation to serial homicides (a total of about six studies, compared with eleven or twelve studies on rapes). Dr Woodhams expressed the view that this was a failing or weakness. Another weakness was the lack of victimology in the CLA checklists - ie the impact of the behaviour of victims on the behaviour of offenders, which may cause widely differing scores. Moreover, no research has been carried out as to how "copycat" crimes and the staging of crime scenes are to be fitted into the theory of CLA. There are other difficulties - foreign research with no certification or adequate quality controls, the removal of extreme behaviours from other cultures and the method of dealing with "one off" cases and how this affects the results. Moreover, there is no crosscheck at present against different crimes - for example, a lorry driver with a wide geographical range who commits a rape on one occasion, and on another occasions abducts his victim, on another occasion attempts to murder and on another occasion murders. The advocate depute observed that under present CLA research, attempted murders are not included in murder researches, despite the fact that they have the same intention.

[41] Before CLA could be safely regarded as admissible evidence in court, there would need to be an agreed set of principles in the scientific community, with generally recognised credentials, published protocols and procedures and confirmation that agreed procedural steps had been taken. None of these exist at present.

[42] In addition to all the foregoing concerns, the advocate depute submitted that in any event CLA evidence is not admissible applying the general rules of Scots law against the admission of evidence about collateral issues. In this regard he relied on A v B (1895) 22R 402 (per Lord President Robertson at 404), Moorov v HM Advocate 1930 JC 68 (per Lord Sands at 87), Brady v HM Advocate 1986 JC 68 (per Lord Justice Clerk Ross at 73) and CJM (No 2) v HM Advocate [2013] HCJAC 22, 2013 SLT 380. This approach was consistent with the approach taken in England - R v McAllister [2008] EWCA Crim 1544 and O'Dowd v R [2009] EWCA Crim 905. The advocate depute observed that it would be difficult to imagine how a trial could be conducted if there was a retrial and CLA evidence was admissible - it would be necessary for evidence to be led about the individual circumstances of each crime in the alleged series, witnesses would be required to give evidence about each crime, and there would then require to be an analysis of the similarities and dissimilarities between each crime. Weeks would be added to the length of the trial, and there would be lengthy and detailed evidence about cases not referred to on the indictment. All witness statements in each of the investigations would have to be assessed, and the jury would be swamped by collateral evidence and the trial protracted unnaturally. In the present case, the appellant had been convicted of this charge of murder by a unanimous verdict of the jury, on the basis of real evidence. It was impossible to think of a better example to show the rationale for the rule against the admission of collateral evidence.

[43] The advocate depute concluded that, in the present state of CLA research, neither a jury nor an appellate court would be assisted by the admission of CLA evidence. No witness speaking to CLA evidence could put a value on it, and the evidence might be quite wrong. The court would not be in any better position having heard this evidence than it would without it - cf Hainey v HM Advocate [2013] HCJA 47, at paragraph [49].

[44] The advocate depute considered the use of CLA in comparative jurisdictions, in particular in the United States of America, and referred the court to Daubert v Merrell Dow Pharmaceuticals Inc 113 S.Ct 2786 (1992), and (amongst others) an article by D Bernstein "Junk science in the United States and the Commonwealth" (1996) 21 Yale Journal of International Law 123. He submitted that whilst some courts permit the use of CLA in the United States, some do not and no authority has been found by the respondent or by the appellant where the defence has sought to use this technique as a method to delink an accused person from "real" evidence or where CLA evidence has been used to establish the identification of an offender in the absence of direct evidence pointing towards that offender's guilt.

[45] Professor Canter referred in his report to the Moorov doctrine. However, the advocate depute observed that one significant limitation of the Moorov doctrine is that it may only operate where there is a positive identification of the accused in respect of each offence (in a typical case, there being evidence of each complainer positively identifying the accused).

[46] The advocate depute concluded that the research studies for CLA are small and reveal it to be a technique which is prone to error. Its theories are not valid for all offenders, all the time. These limitations suggest the potential for unreliability in case linkage predictions. CLA evidence is not supportive of the appellant's exculpation or undermining of the Crown case and is unsuitable for admission in legal proceedings. In any event to admit evidence of this type, which is not "instantly verifiable" but which instead involves the jury having to consider a number of satellite trials and conflicting testimony from experts in a field which on any view is in a constant state of discovery and refinement would overburden the jury and prolong unduly the trial process. The evidence sought to be led is inadmissible and irrelevant, would cause practical difficulties to the administration of justice and would serve only to distract the jury from the compelling "real" evidence upon which the jury in 1977 unanimously convicted the appellant of the murder of Frances Barker.

 

Submissions for the appellant

[47] Senior counsel for the appellant adopted his written submissions (with the exception of paragraph 4). He submitted that evidence is relevant where it tends - or an inference which can be drawn from it tends - to render a fact in issue more or less probable. Assessment of relevancy is more a matter of logic than law. The ultimate issue in the present appeal is the guilt or innocence of the appellant. "Prima facie all evidence which is relevant to the question whether the accused is guilty or innocent is admissible." - DS v HM Advocate [2007] UKPC 36, per Lord Hope of Craighead at paragraph [26]. Any evidence which tends to undermine proof of the appellant as the perpetrator of the murder of Frances Barker, or which renders it less probable, must be relevant and admissible. Such evidence can include circumstantial evidence which yields a reasonable inference that the person who murdered Frances Barker murdered the other women in the series, thereby excluding the appellant.

[48] Most courts in common law jurisdictions "have been extremely cautious in restricting the power of the accused to call evidence in his or her defence, a reluctance founded in the fundamental tenet of our judicial system that an innocent person must not be convicted" - R v Seaboyer 1991 2 SCR 577 per McLachlin J at 611.

[49] Similar fact evidence is regularly admitted in criminal trials in Scotland provided that it is relevant and does not breach any exclusionary rule - eg HM Advocate v Ritchie and Morren (1841) 2 Swin 581; HM Advocate v Pritchard (1865) 5 Irv 88; HM Advocate v Joseph 1929 JC 55 and HM Advocate v Bickerstaff 1926 JC 65. The same principles that apply to the Crown leading such evidence must apply to the defence. The leading of similar fact evidence by the defence is permitted in other common law jurisdictions, eg Canada (R v Arcangioli [1994] 1 RCS 129), Australia (Cheney v R (1991) 99 ALR 360) and South Africa (S v Letsoko 1964 (4) SA 768).

[50] The relevance and probative value of evidence of similarity in respect of the circumstances of how offences were committed has been repeatedly recognised in Scotland - Moorov v HM Advocate 1930 JC 68; Stewart v HM Advocate 2007 JC 198; McMahon v HM Advocate 1996 SLT 1139; MR v HM Advocate [2013] HCJAC 8.

[51] Evidence of a collateral issue is relevant if it could rationally affect, directly or indirectly, the assessment of the probability of the existence of a fact in issue in the trial. For example, in the present case, circumstantial evidence which when taken together can yield the inference that someone else was responsible. To regard evidence of such similar crimes, supported by evidence of a connection to these by the suspect, as potentially undermining proof that the appellant was responsible, is a process of thought which a rational, objective and fair minded person might follow. If such a person would, or might, attach importance to evidence such as this, it would require good reasons to deny a judicial decision maker the opportunity to consider it - O'Brien v Chief Constable of South Wales [2005] 2 AC 534 (per Lord Bingham of Cornhill at paragraph 4).

[52] Senior counsel for the appellant submitted that CLA opinion evidence was sufficiently reliable to be admitted as it is predicated on sound principles, techniques and assumptions (see Daubert, supra). The underlying theories or assumptions upon which CLA is based - the theories of consistent and distinctive behaviour - are theoretically sound and are susceptible to and have been subject to scientific study. The underlying theories have been subject to review which has supported those theories. Review has shown that there remains a degree of error, and prediction is not perfect and varies according to the type of behaviour under analysis. However, it was submitted, predictive accuracy is statistically significantly better than chance and that accuracy can be good or excellent depending upon the type of serial crimes. Sexual offences and murders have been studied most, and it has been found that such crimes which have features indicative of planning and exerting control have a higher rate of accuracy. This means that case linkage based on such behaviours should be more accurate. CLA theories have been subject to peer review and publication.

[53] Senior counsel submitted that looking at the evidence of Professor Canter as a whole, including his reservations, CLA evidence met the test of admissibility. He acknowledged that Dr Woodhams had expressed reservations about the underlying assumptions and the practical implications arising from that; this was a matter of evaluation for the court to assess. Senior counsel's overall submission was that CLA, even with its imperfections, if properly explained, can be applicable to this case. Inevitably it would be explained to the court that there is an element of error in the CLA results. Dr Woodhams was seeking perfection. Quite responsibly, she was stating that she had grave reservations about CLA, but senior counsel submitted that it was a supplemental tool and would assist the court. There was sufficient merit in it to render it admissible. The court could reach a view without CLA, but it would assist. Despite the evidence about the high rate of error, this should properly be pointed out by any responsible scientist giving evidence. The error rate is to be weighed alongside other factual information; it is simply one of the matters to be taken into account. Even with the error rate, the court can be assisted by this process.

 

Discussion

[54] Evidence about relevant matters which are not within the knowledge of everyday life reasonably to be imputed to a jury or other finder of fact may be admissible if it is likely to assist the jury or finder of fact in the proper determination of the issue before it. The expert evidence must be relevant to that issue (and so not concerned solely with collateral issues), and it must be based on a recognised and developed academic discipline. It must proceed on theories which have been tested (both by academic review and in practice) and found to have a practical and measurable consequence in real life. It must follow a developed methodology which is explicable and open to possible challenge, and it must produce a result which is capable of being assessed and given more or less weight in light of all the evidence before the finder of fact. If the evidence does not meet these criteria, it will not assist the finder of fact in the proper determination of the issue; rather, it will risk confusing or distracting the finder of fact, or, worse still, cause the finder of fact to determine the crucial issue on the basis of unreliable or erroneous evidence. For this reason, the court will not admit evidence from a "man of skill" or an "expert" unless satisfied that the evidence is sufficiently reliable that it will assist the finder of fact in the proper determination of the issue before it. We agree with, and adopt, the general observations of the court with regard to evidence from a person claiming specialist knowledge and expertise which were made by the court in Hainey v HM Advocate [2013] HCJAC 47, particularly at paragraph [49].

[55] There are countless examples of evidence about such matters which are routinely regarded as based on sufficiently developed theories, which have sufficiently developed and certifiable methodologies, and produce results which have a practical effect and which may be weighed and assessed by a finder of fact that such evidence is admissible in court. So, scientific evidence about DNA comparisons, fingerprint evidence, evidence of medical practitioners or pathologists is evidence based on a sufficiently clear and reliable basis that it may assist the finder of fact, and will be admitted as evidence for the finder of fact to consider. It does not of course follow that the finder of fact will accept the evidence, in whole or in part - there may be conflicting evidence, or the finder of fact may not be satisfied by the evidence. But in order to be admissible, the evidence must have a sufficiently reliable foundation to be capable of assisting the finder of fact in the proper determination of the issue before it.

[56] Having considered the evidence of both Professor Canter and Dr Woodhams we have little difficulty in reaching the conclusion that CLA evidence, in its present state of development, does not possess the necessary qualities to render it admissible in court, either before a jury or in appellate proceedings. It is an area of academic research which is still in its infancy - it is an aspect of behavioural science which has only been actively pursued since the 1980s, and Dr Woodhams stated that there were only six research papers analysing potentially linked murders, and about twelve papers analysing potentially linked rapes. While the underlying theories of behavioural consistency and behavioural distinctiveness appear to have some foundation in general experience, the application of these theories in the context of criminal behaviour, and CLA in particular, is not yet tested.

[57] There are several aspects of the methodology of CLA at present which suggest that it is not yet reliable. Most studies do not take account of victimology, and the effect of a victim's reaction and behaviour on the behaviour of the offender. Research to date is concerned only with closed, or solved, crimes, and it is not apparent that it can safely be applied to predictions, or unsolved crimes. Inter-rater reliability remains a real issue, and there appears to be no agreed or uniform procedure (either within the UK or worldwide) to check and certify this. Indeed, there are no agreed international or national standards in the field of CLA.

[58] Perhaps the aspect of CLA which is of most concern when considering whether or not it is sufficiently reliable to be admissible evidence is its known high error rate, and the lack of any means whereby a fact finder may weigh and assess the evidence in a particular case. Professor Canter accepted in his evidence that the best AUC which has been achieved in CLA is 75%. This means that in a quarter of potentially linked crimes (and perhaps more), CLA will indicate that crimes are linked, when it is known that in fact they are not linked. There is no way in which a fact finder in a particular case can form a view as to whether the crime which is being considered is or is not correctly linked to another crime. No weighting can be given to the CLA evidence - it is either right or wrong, but no assistance can be given by a CLA "expert" as to which, in a particular case. Even in the best conditions, with the best researchers using the best practices, there is still a high number of false positives and false negatives, and no satisfactory explanation has yet been found for this.

[59] It may be for these reasons that CLA evidence does not appear to have been given in any court anywhere thus far. Considering all of the evidence led before us, it may be that CLA has developed to the stage that it may be of assistance to those charged with the investigation of crimes, as an aid in the search for other real evidence (although standing the doubts expressed by Dr Woodhams on this point, we express no view on this). However, we have reached the clear conclusion that CLA evidence in its present state of development, and as it has been explained to us by Professor Canter and Dr Woodhams, is not admissible in these proceedings.

[60] As appears from the summary of the evidence given above, there was considerable common ground between Professor Canter and Dr Woodhams. Professor Canter stated that he understood why Dr Woodhams expressed the view that it was a step too far to use CLA as a probative tool in court, in the real world; however, his own view was that although some cases were unsuitable for CLA, others were suitable, and that if it was used appropriately and in the right context it was accurate enough to be of some assistance to the court. To the extent that it is necessary for us to do so, we preferred the evidence of Dr Woodhams to that of Professor Canter. Dr Woodhams has considerable practical experience of CLA as a crime analyst with the Metropolitan Police; she has continued to be actively involved in academic research in CLA and has published several recent articles in peer review journals, whereas Professor Canter has not published since about 2002; and she is currently forming an international group of academics and police investigators to discuss minimum standards of practice in this area. Dr Woodhams gave very measured and persuasive evidence, and we are satisfied that her reservations about the use of CLA in court proceedings, in its present state of development, are well founded.

[61] That is sufficient to dispose of the subject matter of this preliminary hearing. Although both parties made submissions to us about the allegedly collateral nature of this evidence, this is not a matter which featured in the grounds of appeal, nor in the Commission's reference to the court. When the court was considering the appropriate procedure in this appeal on 4 November 2011 it "held that it would be desirable to deal with the Crown's submissions on the admissibility of the expert evidence, and the appellant's response thereto, at a preliminary hearing in advance of any final appeal hearing; such preliminary hearing to include the leading of any evidence on the extent of any acknowledgement of the evidence as part of a recognised science, and to be heard before the three judges assigned to determine the substantive appeal."

[62] At the time of that interlocutor, no question had arisen about the possibly collateral nature of the evidence. Since then, the question of the admissibility of collateral evidence has been the subject of review and authoritative re-statement by a bench of five judges in CJM (No.2) v HM Advocate 2013 SLT 380. This was not the subject of any submissions before us on behalf of the appellant, and standing the terms of the interlocutor of 4 November 2011 and the conclusions which we have expressed above, we find it unnecessary to address this issue here.