Matthew ~ o o d e *
SOME OBSERVATIONS ON EVIDENCE OF DNA FREQUENCY . . .[Elvidence presented to a jury with the convincing rituals and trappings of scientific evidence, should be good science. This means that the assumptions on which a statistical analysis proceeds should be wellgrounded in empirical data, and if statistical inferences are made, the inferences should be fair ones that accurately capture the weight of the evidence as it pertains to questions the jury must answer. Critics acknowledge that DNA evidence should be admitted, and most agree that statistics can help juries understand how to weigh DNA evidence, but, the critics hold, the evidence must be presented to juries in such a way that its uncertainties are not obhscated nor its implications oversold.' man known as Popeye has been kicked to death in a park. There is a bloodstained right boot by his body. It is not Popeye's boot. He died with his boots on. The extra boot is spattered with Popeye's blood and tissue. The boot also bears substantial traces of blood which came from someone else. Bluto is arrested two days later after he injured another man in a j g h t at a homeless man's shelter. A DNA sample from Bluto is taken on his arrest. It is matched to the unknown blood sample taken from the boot found near Popeye's body. There is expert evidence that the chance of the boot DNA sample not being that of Bluto is one in one billion. Consider the following variations and complications: A single left boot, of the same size, make, style and state of wear as the one in the park, is found in Bluto's room in the shelter; No trace of DNA matching Popeye's DNA is found on the left boot;
1
LLB (Hons), LLM; Managing Solicitor, Policy and Legislation Section, South Australian Attorney-General's Department, Adjunct Associate Professor of Law, University of Adelaide. The views in this article should be taken to represent those of the author only and not those of the Attorney-General's Department or, indeed, anyone else. R Lempert, 'After the DNA Wars: A Mopping Up Operation' (1 997) 3 1 Israel Law Review 536, 538.
46
GOODE - SOME OBSERVATIONS ON EVIDENCE OF DNA FREQUENCY
Bluto is an Aborigine. The individuals on the statistical database did not include Aborigines and the frequency of Bluto 's DNA confguration in the Aboriginal subpopulation is not known.2
In R v itche ell,^ the accused was charged with attacking a young Korean with intent to commit rape. In the course of the attack, the victim suffered wounds which caused bleeding. The accused was apprehended near the scene of the attack. He, too, had a wound which was bleeding. Forensic examination by DNA analysis was made of the blood of both the accused and the victim and compared to the blood on the clothing of each. Evidence was led that blood consistent with that of the accused and inconsistent with that of the victim was found on the clothing of the victim. Evidence was also led that that blood consistent with that of the victim and inconsistent with that of the accused was found on the clothing of the accused. Two experts gave evidence. The first was to the effect that the blood on the clothing of the accused was that of the victim with a probability of one in 54 million. The second was that the chance of a random match was 99.9 per cent or between one in 18 to one in 500 million. The trial judge directed the jury as follows (his summary): [Tlhe DNA evidence enabled a conclusion to be drawn that the subject blood samples found on the clothing in question were consistent with blood deposited by the accused and the victim respectively. It would follow, of course, that such evidence would support a conclusion that there had been contact between the accused and the victim while each was shedding blood, at or near the crime scene. I directed the jury that whilst the DNA evidence was consistent with that hypothesis, the statistical evidence should not be used by them so as to conclude that there was merely a statistically minute chance of a random match of the order suggested by [the experts]. In effect, the weight to be
2
3
4
I am indebted to my colleague, Ian Leader-Elliott, for this scenario. He is far more ingenious and witty in this task than I shall ever be. I do not intend to analyse this hypothetical. It is designed to outline some basic issues that must be confronted. It is very often the case that a particular decision on the admissibility and reception of DNA evidence is greatly affected by its age. The technology of DNA analysis, its presentation in court and the scientific status of the technology has evolved rapidly in recent years. Cases in this field are not alike. (1997) 98 A Crim R 32.
(2002) 23 Adelaide Law Review 45-77
47
attached was for them on such other basis as they might consider appropriate.5
The first expert's opinion was based on a database from Victoria of samples tested on six loci compared to a database of people tested for the purpose of paternity testing. The size of the database varied from 344 to 523. In order to obtain the probability figures, the expert simply multiplied each of the individual percentages of matching at each site. The second expert gave evidence in applied mathematics about the same calculation. In R v ~ u r n ~ h rthe e ~accused ,~ was charged with murder. What appeared to be blood on the boot and jeans of the accused was analysed for DNA comparison purposes. A forensic scientist7 gave evidence that the accused was excluded as a source of the DNA but the deceased was not excluded. She also expressed the opinion that the statistical likelihood of the DNA profile found in the sample being repeated in the general population was less than one in ten thousand, or, put another way, that 99.99 per cent of the population would be excluded in relation to the boot stain and 99.98 per cent of the population in relation to the stain on the jeans. The defence sought a ruling that the statistical evidence be excluded unless it was proved that the database was reliable and unless it was proved that there was an absence of dependence between the alleles from which the frequency evidence was derived. In ruling on the defence submission, Bleby J held that it is not for the trial judge to determine which population group the data base should represent because in order to do so, one must make an assumption about the limits of the population from which the unknown samples came and, in any event, the question is one for the jury and to so decide would be to usurp the function of the jury. In addition, he held that the admissibility of the frequency evidence did not depend on evidence of the absence of dependence between the alleles from which the frequency evidence was derived because that too was a matter of the weight of the evidence for the jury to decide. However, Bleby J also ruled: Whilst I have ruled that the frequency evidence is admissible, if an appropriate foundation is not laid as to the validity of the databases for the purpose of calculating that frequency, and where that validity remains 5 6
7
Ibid 34. (1999) 103 A Crim R 434. Two of the forensic experts who gave evidence at this trial appear to be the same as those who gave evidence in R v Mitchell (1997) 98 A Crim R 32.
48
GOODE - SOME OBSERVATIONS ON EVIDENCE OF DNA FREQUENCY in dispute, the evidence of frequency may well carry little or no weight because the assertion of the expert is incapable of being tested. . . . when such evidence can and no doubt does have such a powerhl influence on identity, it can reach a point where, if the justification for the opinion is not led, the probative value of the evidence is substantially outweighed by its prejudicial effect, and it becomes inadmissible on discretionary grounds. . .8
In this case, the statistical database used for comparison purposes was South Australian and consisted of profiles at three loci ranging in number from 1002 to 538. The source was members of the staff at the Forensic Science Centre, paternity testing (excluding children) and complainant and defendants in criminal matters. Then, after an immensely long voire dire that took place over eight months in 2000-2001, Mullighan J of the Supreme Court of South Australia, delivered reasons for rulings in the test case of R v ~ a r ~ eThe r . accused ~ was charged with a murder that took place in the course of a house break. Two very small blood stains were found at the scene of the crime. Expert evidence was offered that these stains yielded a DNA profile that matched that of the accused, but the rest of the evidence against him was circumstantial. The DNA profiling was done on various samples with two systems - one a quadruplex and the other a system known as 'Profiler Plus'. The quadruplex system inspects DNA at four loci and 'Profiler Plus' at 10 loci. The prosecution intended to call expert evidence to the effect that there was a match probability of one in a little over 90 billion; or, more specifically, that it was about 90 billion times more likely that the accused was the source of the crime stain than if a random person was the source. The defence mounted a challenge on the voire dire to every aspect of the evidence and all of the components that led to it. In a comprehensive judgment, Mulligan J rejected every aspect of the challenge. The judgment is in itself a thorough guide to current forensic DNA technology. In so doing, he ruled that the 'Profiler Plus' based system was a generally accepted and validated scientific technique even though the manufacturer and marketer of the kit had declined to disclose the primer sequences they had created for the amplification of DNA." His Honour ruled that the statistical evidence sought to be led by the Crown could be led. 8
9
10
(1999) 103 A Crim R 434, 443. The general proposition is, of course, a routine evidence rule. His Honour cited R v Duke (1979) 22 SASR 46 for it. [2001] SASC 64. The decision is unlikely to be reported in fill at any time because of its extreme length. His Honour referred to a small number of US authorities in which admissibility was denied on this ground but refised to follow them on the ground that the test to be applied is general acceptability in scientific circles and the latter had been established. At para [230]:
(2002) 23 Adelaide Law Review 45-77
49
This paper will not include the customary outline of what DNA is and the basics of its significance. It assumes that the reader either knows that general material or can find it elsewhere. Nor is this paper about the law relating to the collection of DNA evidence - that is, the law of criminal investigation. That is a subject for another time. Instead, it is intended to concentrate on recent developments in an emergent and rapidly developing field of expert evidence and law.
11 THEPROSECUTOR'S FALLACY Before delving into more complex matters, it is necessary to mention a matter which is commonly known as the 'prosecutor's fallacy'." It is not intended to be comprehensive about it in these observations. Rather, it is intended to provide an outline and record the problem. It appears that, in general terms, the 'prosecutor's
11
I accepted ... that there have been modifications to the primer sequences by Perkin Elmer during the developmental stage which have not been published. However, the failure to do so has not been an obstacle to general acceptance by the forensic science community of Profiler Plus with Genotyper as an accurate and reliable system of DNA analysis for the purpose of human identification. The same reasoning applied to the non-publication of chromosomal locations of the loci inspected by Profiler Plus: see para [465]. The reasoning and result in R v Karger, particularly as it relates to Profiler Plus, were duplicated in NSW in, for example, R v Gallagher [2001] NSWSC 462 (4 May 2001) and R v McIntyre [2001] NSWSC 31 1 (1 1 April 2001). The explanation given for non-publication has been that scientific journals will not publish data which do not represent new concepts. However, since November 1999, the Journal of Forensic Sciences has decided to publish this kind of data. See, most recently, Hocking et al, 'DNA, Human Rights and the Criminal Justice System' (1997) 3 Australian Journal ofHuman Rights 208. The debate was known in the United States as the 'DNA Wars' and, technically, the debate was about the reliability of random DNA inheritance. If DNA inheritance is random, then a statistician can validly use the product rule when calculating frequency estimates see, for example, Note: 'The Dark Side of DNA Profiling: Unreliable Scientific Evidence Meets the Criminal Defendant' (1990) 42 StanfordLaw Review 465; W C Thompson, 'Evaluating the Admissibility of New Genetic Identification Tests: Lessons From the "DNA War"' (1993) 84 Journal of Criminal Law & Criminology 22. There was an unsuccessful attempt to end the DNA War in 1992 by the (USA) National Research Council (NRC), DNA Technology in Forensic Science (1992). In 1996, the NRC withdrew the effect of the 1992 report: National Research Council Report 11, The Evaluation ofForensic DNA Evidence (1996). The second report came down strongly in favour of the proponents of DNA evidence. See for a general and thorough account, R Lempert, above n 1.
50
GOODE - SOME OBSERVATIONS ON EVIDENCE OF DNA FREQUENCY
fallacy' has been properly handled by the courts, but one needs to be carehl about it - and it persists in popular imagination. DNA statistical evidence can be startling indeed. For example, as early as R v Welsh,I2 the expert evidence was that the odds of another person having the same ' ~ Cox, an blood was between 1/88 million and 1/99 million. In R v ~ l e t c h e r , Mr expert witness from the John Tonge Laboratory in Brisbane, was apparently content to testify that the random occurrence ratio was 1 in 6.4 billion. In R v Karger, the evidence was about 1/90 billion. The reason for the increasing numbers has to do with the increase in the number of loci examined by the increasingly sophisticated DNA comparison process. Early cases involved technology which examined few loci - such as three or four (the triplex and the quadruplex respectively) - whereas Profiler Plus examines 10. Increasingly sophisticated systems may well examine more loci and so the numbers will become more extreme. l 4 So what is wrong with the following statement? 'The chance that Roy Whiting was not Sarah Payne's killer was one in a billion, a jury was toldyesterday. ,I5 The answer is that the statistical calculation does not transfer the likelihood of matched samples to guilt. The journalist got it wrong - and the prosecution almost certainly did not say that. The 'prosecutor's fallacy' is that the statistics of the match16 necessarily translates into the equivalent chance of the accused being guilty. 12
13 14
15
16
[I9921 SLT 193. [I9981 2 QdR 437. It is predicted that, in the foreseeable future, examination will not move beyond the 13 core short tandem repeats (STR) loci commonly employed in the United States. See the (US) Department of Justice, National Institute of Justice, The Future of Forensic DNA Testing: Predictions of the Research and Development Working Group; A Report From the National Commission on the Future of DNA Evidence (2000). That report states that using the standard US 13 loci, the probability of a match between the profiles of unrelated persons in a randomly mating population of ~ one in 575 trillion (at 19). On these extreme Caucasian Americans is 1.74 x 1 0 - l or numbers, see further discussion below. The Times, Saturday, November 19, 200 1. There is a great deal of terminological debate in this area about what the right words are. Different sources use different terms for 'match probability', and so on. This paper does not intend to enter into this debate. Nothing turns on the words so long as the idea is clear. But, on the words, see I W Evett, et al, 'DNA profiling: A Discussion of Issues Relating to The Reporting of Very Small Match Probabilities' [2000] Criminal Law Review 34 1.
(2002) 23 Adelaide Law Review 45-77
51
Another less statistically stellar example is R v Milat,'' in which the Crown sought to lead evidence that the probability ratio for the DNA 'match' made in that case was 111 18,000. The court allowed the evidence to be given, in an articulate and informed judgment which repays close study in its reference to the scientific evidence. The court noted the 'prosecutor's fallacy' in the following way: Care must be taken not to transpose that proposition to one that the blood is 118,000 times more likely to have come from the deceased rather than from someone else. This transposition has been called the 'prosecutor's fallacy', and the distinction is explained in an article by Evett, IW, 'Avoiding the transposed conditional' (1995) Science and Justice 127. Its subtlety is such that I could not do it justice by describing it here. The distinction may, however, be demonstrated by adopting one of the examples (at 130) to read: the chance that a person other than the deceased would leave (a) blood of this type is one in 118,000 (which is correct); and the chance that a person other than the deceased did leave (b) blood of this type is one in 118,000 (which is incorrect).18
A major starting point in consideration of the issues ought to be the decision in R v Doheney & Adams.l9 In that case, the prosecution sought to give evidence of a DNA match and the expert witness for the prosecution gave evidence in the form of a likelihood ratio. In the case of Doheney, the evidence was of a DNA probability matched by a multi-locus probe technique based on a blood-semen match of one in 40 million.20In the case of Adams, the evidence was based on single-locus probe tests and resulted in evidence citing a chance of one in 27 m i l l i ~ n . ~ ' The defence then produced an expert witness who explained to the jury how to combine the DNA evidence with other evidence in the case consistent with the innocence of the accused. This evidence assumed the correctness of Bayes' Theorem. Bayes' Theorem is a mathematical way of stating that an estimate of the probability of (in this case) the DNA match being a guilty match depends logically 17 18 19
20
21
(1996) 87 A Crim R 446. Ibid 45 1, n 22. [I9971 1 Cr App R 369. After all of the evidence was in, this turned out to be the 'real' statistical (as opposed to evidential) chance: 'If one excludes from the calculation the result of the multi-locus probe test, in a population base of 800,000 suggested as constituting those within a reasonable proximity of the scene of the crime, there may well have been 20 or so individuals with the same DNA profile as the appellant and the crime stain'. Ibid 380. In this case, the DNA calculation was not examined closely because of other corroborating evidence.
52
GOODE - SOME OBSERVATIONS ON EVIDENCE OF DNA FREQUENCY
on an estimate, whether mathematically expressed or otherwise, of the prior evidence of the likelihood of the guilt of the accused. When the likelihood ratio derived from the DNA match is combined with the 'prior odds' one obtains a result which is usually called the 'posterior odds '. This is, in short, a statement of logic. A simple example is a case in which a witness gave evidence that a blood test showed that the odds were, say, 1/72 that the person concerned was the father of a baby rather than a randomly selected person. That does not mean that the odds are 1/72 that the person is the father. If one took 720 people, one would expect that 10 would produce a match. The odds in that circumstance would be 1011 against the specified person being the father.22In R v Adams, the DNA match was with a blood stain left at the scene of the crime. The likelihood that the DNA matched the accused was inevitably affected by the likelihood that he left the stain. The problem with this is that the 'prior odds' are not scientifically statistically available, particularly when they are more complex in nature. Therefore scientists do not like them and want to leave them out of the equation. Courts have gone along with this point of view. The Court in R v Adams explained the intersection of the prosecutor's fallacy and Bayes' Theorem as follows: If one person in a million has a DNA profile which matches that obtained from the crime stain, then the suspect will be one of perhaps 26 men in who ' ~ share that characteristic. If no fact is known the United ~ i n ~ d o m about the defendant, other than that he was in the United Kingdom at the time of the crime the DNA evidence tells us no more than that there is a statistical probability that he was the criminal of 1 in 26. The significance of the DNA evidence will depend critically upon what else is known about the suspect. If he has a convincing alibi at the other end of England at the time of the crime, it will appear highly improbable that he can have been responsible for the crime, despite his matching DNA profile. . . . The reality is that, provided there is no reason to doubt either the matching data or the statistical conclusion based upon it, the random occurrence ratio deduced from DNA evidence, when combined with sufficient additional evidence to give it significance, is highly probative. As the art of analysis progresses, it is likely to become more so, and the stage may be reached when a match will be so comprehensive that it will be possible to construct a DNA profile that is unique and which proves the guilt of
22 23
I W Evett, et a1 above n 16, 345 call this evidence of 'relative frequency'. One might add that this calculation is optimistic in that it takes no account of uncounted illegal immigrants, visitors, tourists and the like.
(2002) 23 Adelaide Law Review 45-77 the defendant without any other evidence. So far as we are aware that stage has not yet been reached.24
The question in R v Adams was the admissibility of this kind of evidence about proper reasoning dressed up, as it were, as expert evidence about mathematical probability theory. This is not really a new legal question. If one thinks about it, the reasoning involved is always part of the assessment of chance and in particular the probabilitylcertainty of the guilt of the accused or of facts in evidence.25It becomes acute in cases involving DNA evidence because of the way in which that evidence may be expressed. It is quite clear that the court and the players in the court have to be aware of the 'prosecutor's fallacy' where, in particular, DNA evidence is given and probability ratios are discussed in evidence. It is obviously true that it is easy to fall into the reasoning error exposed in this area. In a very interesting and short piece on this and other like cases, Robertson and Vignaux stated: [Bayes' Theorem]. .. may indeed plunge the jury into unnecessary realms of theory and complexity, but it can hardly be described as a deflection from their proper task. Bayes' Theorem constitutes not a deflection from this task, but a formalisation of it, a formalisation which, we would agree may not always be necessary or helphl to a juror, but which is the only appropriate yardstick for those, such as Appeal Courts and academics, who judge Judges. There is arguably a much simpler and more compelling reason why experts cannot be allowed to give evidence on how jurors should combine non-scientific evidence using Bayes' Theorem. If such evidence were admissible in Adams it would be admissible in every single criminal case. While this might be good for those qualified to give the evidence, this would not do anything for the cost-effectiveness of the criminal justice system.26
One cannot with equanimity contemplate the instruction of a jury in Bayes' Theorem, the more so if it is contemplated, as is currently the case, that DNA 24 25
26
[I9971 1 Cr App R 369,373. The question in R v Adams is not new, albeit that the use and form of DNA evidence has given it a recent piquancy. The relationship between logic and mathematics on the one hand and the law of evidence on the other has been the subject of a great deal of writing and commentary. Sir Richard Eggleston wrote extensively on it, his most famous work being Evidence, Proof and Probability (Second Ed, 1983). Glanville Williams has produced his usual coherent and understandable version in 'The Mathematics of Proof [I9791 Criminal Law Review 297. B Robertson and T Vignaux, 'DNA on Appeal - I' [I9971 New Zealand Law Journal 210,212.
54
GOODE - SOME OBSERVATIONS ON EVIDENCE OF DNA FREQUENCY
evidence will become very much more common in criminal courts than is and has been the situation h i t h e r t ~ . ~The ' English Court of Criminal Appeal has said: To introduce Bayes Theorem, or any similar method, into a criminal trial, plunges the jury into inappropriate and unnecessary realms of theory and complexity deflecting them from their proper task.28
In R v Karger, Mullighan J adopted the reasoning in R v Doheney & Adams. The defence had then argued that, if Bayes' Theorem cannot be used, nor can the likelihood ratio because it is part of the theorem. Mullighan J rejected that argument. He did so on the basis that the statistical evidence sought to be led did not imply the use of Bayes' Theorem. In general, the answer to this particular problem lies in experts giving their evidence properly and judges instructing juries about how to use it appropriately. The former appears to be a question about professional ethics and o b j e c t i v i ~ m . ~ ~ Expert evidence, not only in relation to DNA evidence, but also generally, should move from the cult of the expert to transparent objectivity. The latter is a question of law - which is about social ethics in the sense of a right to a fair trial. The 27
28
29
It seems evident at the time of writing that this will be the case. The Commonwealth Government has set up a national DNA Database as part of its CrimTrac programme and the Model Criminal Code Officers' Committee has reported to the Standing Committee of Attorneys General on the appropriate legislative base for such a scheme: see MCCOC, Report, Model Forensic Procedures Bill and the Proposed National DNA Database, May 1999; MCCOC Final Draft: Model Forensic Procedures Bill and the Proposed National DNA Database (February 2000). These papers may be accessed at .Others have been less formal. See, for example, the rhetorical poems of praise to common DNA testing for even minor crimes made by visiting 'expert' Detective Superintendent Robin Napper, Sydney Morning Herald, 1112199: 'The more people on a DNA database the better. Those who commit highvolume crimes, such as simple burglary, should be included on the database, not just the serious offenders, as recommended by the model legislation ... All big crooks start out as little crooks. If you can interrupt this early enough, you can reduce crime.' It should be noted that this statement is illogical and, in terms of criminological analysis, unverifiable. R v Adams [I9961 2 Cr App R 467,482 quoted in R v Doheny and Adams [I9971 1 Cr App R 369, 375. However, I W Evett, et al, above n 16, 353 disagree. They think that it is strange that the only logical approach to the evidence should be thought to be inappropriate. At 354 they state that 'it is the triumph of reaction against progress'. For an alarming account in the US context, see P C Gianelli, 'Impact of PostConviction DNA Testing on Forensic Science' (2001) 35 New England Law Review 627.
55
(2002) 23 Adelaide Law Review 45-77
appropriate direction on the use that can and should b e made of seemingly conclusive and unanswerable evidence is, it is submitted, crucial. As the (US) Department of Justice has said: 'The human psyche seems to have an overwhelming proclivity to misinterpret this [DNA statistical The necessity for an appropriate direction is further emphasised b y the following scientific opinion:
evidence^'.^^
There is an infinite range of ways of carrying out the calculation that underlies the figure given. The method chosen in the individual case must be seen to be as much a matter of opinion as one given in other areas of forensic science. The match probability is personal '. It is based on what the scientist considers to be the most appropriate calculation given the circumstances of the case.31 If anything calls for a clarifying jury direction, this does. T w o further matters arise from this observation. First, it should be noted in this context that it has been argued that the match ratio given in evidence should never be expressed b y figures which are higher than the probability that there has been an error in the chain of evidence or in the laboratory.32 It is surely common sense that the error rate in police and laboratory procedures is bound to b e higher than the
30
31 32
(US) Department of Justice, National Institute of Justice, The Future of Forensic DNA Testing: Predictions of the Research and Development Working Group; A Report From the National Commission on the Future of DNA Evidence (2000), 14n5 available at . I W Evett, et al, above n 16, 346 (emphasis added). See, for example, JJ Koehler, 'Error and Exaggeration in the Presentation of DNA Evidence at Trial' (1993) 34 Jurimetrics 21, discussed in G K Chambers, et al, 'Forensic DNA Profiling: The Importance of Giving Accurate Answers to the Right Question' (1997) 8 Criminal Law Forum 445, 4 5 3 4 and R Lempert, above n 1, 545. It can be argued with some force that, given the orders of magnitude of the likelihood ratios sometimes expressed, the possibility of planted evidence or police error is far greater. It has also been suggested in the context of the OJ Simpson evidence that the likelihood ratio which truly expressed the evidentiary position was that which expressed the likelihood that the police lied: see W C Thompson, 'DNA Evidence in the OJ Simpson Trial' (1996) 67 Colorado Law Review 287. That was clearly a sensational and unusual case. But police and scientists, like any other human beings, acquire a vested interest in maintaining the notion that their work is error free. There is nothing sinister in that. Expert evidence just happens not to be true all of the time. The jury has the function of judging this on the facts of any given case, but cannot be expected to guess about such a decision and should be given the instruction not to do so.
56
GOODE - SOME OBSERVATIONS ON EVIDENCE OF DNA FREQUENCY
often fantastic likelihood match ratios that are presented to courts.33Indeed, some of the literature supports the suggestion that the newer method of DNA analysis (the PCR method) is by reason of its sensitivity more prone to handling error and accidental ont tam in at ion.^^ However, while it is true that laboratory or other handling errors can lead to a false inclusion and that it is very difficult for such errors to be discovered, even with the best of wills and the best of crossexamination, the problem is that there is no defensible and cost-effective standard for measuring laboratory error.35An attempt to take this ground was rejected in the unreported Canadian case of Love (No 2).36 It was also rejected in Karger. Mullighan J said in response to that argument: There is no support for this opinion to be found in the literature as disclosed in the evidence. Dr Buckleton said that an error should not be taken into account and that view is expressed in NRC I1 and other publications. His view is that the possibility of error is a separate matter. Mr Pearman said that there is much evidence to show that an error rate cannot be calculated. He could not see any difference between a frequency or likelihood ratio with regards to an error rate. I accepted their evidence. If the tribunal of fact concludes that there is a reasonable possibility of an error causing an incorrect match, be it in the handling of samples, or in the laboratory or in any other way, there could not be a finding of a match between the DNA of the accused and that in the crime scene sample to the required degree of proof, with the consequence that the DNA part of the case would collapse and there could be no need to consider the statistical e ~ i d e n c e . ~ '
In short, while the statistical calculation does not take into account the laboratory error rate, it cannot do so because that is statistically incalculable. Impliedly, it is up to the accused to show the possibility of laboratory error. The question of possible error is a matter for the jury. The fact that there is always the possibility of an error does not make the frequency ratio evidence inadmissible. This holding is consistent with the way in which the courts have dealt with the question of Bayes' 33
34
35
36
37
See generally, W C Thompson, 'Accepting Lower Standards: The National Research Council's Second Report on Forensic DNA Evidence' (1997) 37 Jurimetrics 405. K Inman and N Rudin, An Introduction to Forensic DNA Analysis (1997) 103; A L Friedman, 'DNA Profiling in the 2 1st Century' (1999) 43 International Journal of Offender Therapy and Comparative Criminology 168, 173. In theory, it would be possible to do a double blind test for any given laboratory, but even that test would be subject to interpretation given the subtlety of testing procedures. See the account in T R McDonald, 'Genetic Justice: DNA Evidence and the Criminal Law in Canada' (1998) 26 Manitoba Law Journal 1, 16-7. [2001] SASC 64, paras [665]-[666].
(2002) 23 Adelaide Law Review 45-77
57
Theorem. It is, however, troubling, because it has the effect of placing a factual onus on the accused, it gives an artificial ring of confidence to startling statistical scientific opinion evidence and there is no evident requirement that the jury receive a direction about the problem (even if, in the end, it is a matter for them in the ordinary way). There is an additional consideration. It may be that the larger the database, the greater the possibility of error or mismatch. In a case known as the Easton case, the UK database operating on a profile of seven loci using STR technology matched a suspect with a stated probability of one in 37 million. The suspect was arrested and charged despite the fact that there was no other evidence against him and there was some evidence suggesting that he was unlikely to have committed the offence. On retesting with 11 loci, the accused was eliminated as a suspect. The Forensic Science Service (FSS) admitted that, at the time, there were 700,000 suspectloffender profiles on their database and 70,000 crime scene samples, so that testing by 6 loci would produce several hundred matches just by chance. With 1 1 loci, it is alleged by the FSS that chance matches would be reduced to less than one.38 However, the (US) Department of Justice, Report From the National Commission on the Future of DNA Evidence states: [I]n traditional forensic analysis one computes the probability, P, of finding in an unrelated person a particular profile matching that of the evidence, E. But if several DNA samples are examined instead of a single suspect, the probability of finding the matching profile is increased correspondingly. To deal with this problem, the first National Academy of Sciences Committee (NRC 1992) recommended that the information employed in the database search be used for identification of a suspect, but notfor evidence in court. For this purpose use of a separate set of loci was recommended. The observations on the second set obviously do not depend on the manner in which the suspect is found and cannot be biased by it.39
There is at least one safeguard in this regard in the requirement contained in the Model Criminal Code Officers Committee Draft Bill for forensic procedures that the sample taken from the accused for DNA purposes be split into two and one
38
39
An account of the story is to be found in USA Today, 8/2/00. It is not to be found in any press release on the FSS website. See the (US) Department of Justice, National Institute of Justice, The Future of Forensic DNA Testing: Predictions of the Research and Development Working Group; A Report From the National Commission on the Future of DNA Evidence (2000), 7, available at . (Emphasis added.)
58
GOODE - SOME OBSERVATIONS ON EVIDENCE OF DNA FREQUENCY
given to the accused for independent testing.40 While current legislation does not deal with crime scenes and analysis, it might be a wise precaution41 for that procedure to be followed in relation to the samples from the time of retrieval from a crime scene, witness or victim.42 The second observation is about the mathematical models used to produce the startling statistics. Again, the thoroughness of the observations of Mullighan J in Karger provides a useful guide.43According to the evidence given to Mullighan J, there are two acceptable methods of producing the relevant statistic. The first is the 'product rule'. The product rule is simply the result of taking chances of a match at any given locus and multiplying the result by the number of loci at which the match occurs. Clearly, the more loci tested and matched, the higher the statistical certainty. Hence, as testing has moved from three or four loci to ten (or maybe more), the statistical confidence has risen to the very high rates noted above. Currently, if this method is used, the statistical chances of a random match become so improbable that it appears that one might as well say one in 10 billion and state that such a figure is conservative. According to Mullighan J: The product rule method requires that relevant states of independence, or equilibrium, exist in the databases. There must be independence between alleles at each locus, which is known as Hardy-Weinberg equilibrium and there must be independence between each locus and other loci, which is known as linkage equilibrium. -
40
41
42
43
MCCOC, Final Draft: Model Forensic Procedures Bill and the Proposed DNA Database, (2000) cl 45. See, for example, the South Australian Criminal Law (Forensic Procedures) Act, 1998, s 39 available at . See R v Gordon [I9951 1 Cr App R 290. In that case, the prosecution took the precaution of having the sample analysed in three separate laboratories. Although each pronounced a match, they did so on different criteria. The conviction was overturned as unsafe and unsatisfactory on the basis that the statistical evidence could not have been the same. There is at least one lesson in this. The vexed question of whether the statistical calculation goes to admissibility or weight and the tendency to treat it as going to weight may be resolved in the end by the application of the unsafe and unsatisfactory rule on appeal. This occurred in the OJ Simpson case, and forestalled defence attacks on the evidence on this point. See R Lempert, see above n 1, 544-5. See also (US) Department of Justice, National Institute of Justice, The Future of Forensic DNA Testing: Predictions of the Research and Development Working Group; A Report From the National Commission on the Future of DNA Evidence (2000) at 24: '[Tlhe best protection for a suspect who may have been wrongly accused because of an error is the opportunity for an independent retest.' [2001] SASC 64, paras [643]-[652].
(2002) 23 Adelaide Law Review 45-77
59
The Hardy-Weinberg Equilibrium is, generally stated, the idea that gene frequencies will remain constant from generation to generation so long as mating remains random. Of course, it is a perfectly reasonable assumption for the purposes of some scientific experiments, but is simply not true enough to ensure proof beyond reasonable doubt in all cases.44 In Karger itself, these pre-conditions were not met. The expert evidence could not exclude a mild disequilibrium between alleles at each locus and between loci. Therefore the alternative method was used, called Equation 4.10. Said to be commonly used and accepted, the Equation requires a confidence which in this case was set at 95 per cent plus what is known as an Fst value. The Fst value is a measure of the co-ancestry or relatedness of alleles and reflects the probability that alleles within a population have a common ancestor. Clearly, that value will differ according to the derivation and nature of any given statistical index employed as a comparison. In this instance, the Fst value used was 3 per cent. That was said to be conservative, given that ancestral relatedness among Caucasians is accepted to be less than one per cent. The result of the Equation in Karger was the figure of one in 90.558 billion. But even so, it appears that, on the view of some experts not cited by Mullighan J, evidence of this kind of extreme number should not have been given. After discussion the validation of statistical calculation by the product rule using a statistical index, Evett, Foreman, Jackson and Lambert state: However, probabilities of the order one in trillions, that follow from calculations based on 10-locus profiles, require independence assumptions that 44 45
[2001] SASC 64, para [644]. Mullighan J described the effect of the evidence on this factor in Karger in the following way: If 200 persons are profiled and allele frequencies are determined on that basis and the process is again undertaken with 200 different persons, slightly different values would be obtained for allele frequencies simply because different people within the same population are being sampled. That sampling variation is taken into account by use of a confidence level which defines, with some certainty, that the true value lies between a common value and a rare value. This means that it is certain that 95 per cent of the time the true allele frequency will fall between the upper limit and the lower limit. In the present case the upper confidence limit was one in 90 billion and the lower confidence limit was one in 272 billion and so the figure adopted is the conservative and lower figure. [2001] SASC 64, para [649].
60
GOODE - SOME OBSERVATIONS ON EVIDENCE OF DNA FREQUENCY cannot be evaluated by statistical experiment, given the size of existing databases. As more and more loci are employed, so the standard calculation produces numbers that are further and further beyond the realms that can be investigated by statistical methods. The tiny numbers, although not necessarily 'wrong', are without any real meaning. ... For these reasons the FSS has adopted a policy of not calculating case-specific match probabilities for full profiles. ... The policy of the FSS ... will be to report a general figure of 1 in one billion for all 10 locus match cases.46
There is, therefore, some scientific opinion for the proposition that the product rule should not be employed to produce such very tiny match probability statistics in court as were proposed to be led in Karger. It is not intended to explore these complications further in this series of observation^.^' Again, it suffices to note that the complications are there and that the complications serve to distinguish, clearly, DNA profiling from
46
47
I W Evett, et al, above n 16, 347-8. The authors go on to defend the general validity of the one in one billion approximation. Ibid. The authors do so in detail. Given the information above, though, the decision of the Victorian Court of Appeal in R v No11 [I9991 3 VR 704 may be correct but must be unwise in the more general sense. In that case, there was an attack on the expertise of an expert witness to give evidence as to the statistical probabilities. The evidence given was that the match, based on a four loci comparison, was one in about 180,000. The basis of the evidence was the product rule. The four loci were each 113, 118, 11360 and 1/23. The product of these chances is 11198,720. The expert was expert in biochemistry. He had no formal qualification in statistics. It does not appear from the report of the case that any thought was given to the Hardy-Weinberg equilibrium. Callaway JA simply said (at 709): 'It is uncontroversial that independent frequencies may be multiplied.' No confidence limit was expressed, it being said that one was unnecessary with a database of over 200. This appears to be wrong. Callaway JA thought that the confidence limit should have been about 95% (without explaining what that meant in the context of the evidence). In the end, Callaway JA concluded (at 7 10): At the end of the day, however, the jury would have appreciated that the blood on the [item] and the applicant's blood had been tested in four different respects, that they were said to be the same in those respects and that that would be true of only about 25 people in Victoria. Whether or not that would be sufficient in other cases, it was sufficient here having regard to the way in which the trial was conducted. For defence counsel, in the course of his final address, said that it was no longer part of his case to contest that it was the applicant's blood on the [item]. Given what has been said in the text, it is submitted that this decision must be confined to the way in which the trial in this case was conducted.
(2002) 23 Adelaide Law Review 45-77
61
fingerprinting.48 In addition, the highly subjective nature of the mathematical processes remains concealed behind the apparent certainty of a bald statistic. It may also be that the larger the number of loci compared, the higher the statistic, the more need there is for an appropriate direction to the jury about what the really impressive statistic really means.
I11 FORMULATING GUIDELINES FORTHEPRESENTATION OF DNA FREQUENCY ANALYSIS A fascinating aspect of the report of the decision of the English Court of Appeal in R v Adams is that the headnote purports to set out a sequence of rules about this kind of evidence and its disputation. It is an interesting compilation of various statements made in the judgment and is worthy of reproduction here. They are a mixture of evidential propositions, procedural propositions and guidance for judges, and so, for convenience, they have been separated into those three groups. First, the evidentiary propositions: 1.
The scientist should adduce the evidence of the DNA comparisons between the crime stain and the defendant's sample together with his calculations of the random occurrence ratio.
.... 5.
6.
7.
48
In giving evidence the expert will explain to the jury the nature of the matching DNA characteristics between the DNA in the crime stain and the DNA in the defendant's sample. The expert will, on the basis of empirical statistical data, give the jury the random occurrence ratio - the frequency with which the matching DNA characteristics are likely to be found in the population at large. Provided that the expert has the necessary data, it may then be appropriate for him to indicate how many people with the matching characteristics are likely to be found in the
There is another distinction worth noting. In the United States, some prosecutors have been able to obtain 'John Doe' warrants based solely on the DNA profile. There is no evidence that this could be done on the basis of fingerprints alone. See, for example, A C Bernasconi, 'Beyond Fingerprinting: Indicting DNA Threatens Criminal Defendants' Constitutional and Statutory Rights' (2001) 50 American Universities Law Review 979. and the authorities cited there.
62
GOODE - SOME OBSERVATIONS ON EVIDENCE OF DNA FREQUENCY
8.
9.
10.
United Kingdom or a more limited relevant sub-group, for instance, the Caucasian, sexually active males in the Manchester area. It is for the jury to decide, having regard to all the relevant evidence, whether they are sure that it was the defendant who left the crime stain, or whether it is possible that it was left by someone else with the same matching DNA characteristics. The expert should not be asked his opinion on the likelihood that it was the defendant who left the crime stain, nor when giving evidence should he use terminology which may lead the jury to believe that he is expressing such an opinion.49 It is inappropriate for an expert to expound a statistical approach to evaluating the likelihood that the defendant left the crime stain, since unnecessary theory and complexity deflect the jury from their proper task.50
Second, the procedural propositions:
. . .. 2.
3.
4.
Whenever DNA evidence is to be adduced, the Crown should serve on the defence details as to how the calculations have been carried out which are sufficient to enable the defence to scrutinise the basis of the calculations. The Forensic Science Service should make available to a defence expert, if requested, the databases upon which the calculations have been based, Any issue of expert evidence should be identified and, if possible, resolved before trial. This area should be explored by the court in the pre-trial review.51
Lastly, the judicial instructions:
49
50 51
Some members of the (UK) Forensic Science Service (FSS) appear to object to this principle on the ground that the same rule does not apply to fingerprints. See I W Evett, et al, above n 16, 342. The distinction between DNA and fingerprint evidence is discussed briefly elsewhere in this paper. They are simply not comparable. The authors later (at 344) agree with that proposition. The fingerprint rule, insofar as it is a rule, is plainly wrong. R v Doheney andAdams [I9971 1 Cr App R 369,369-70. Ibid 369.
(2002) 23 Adelaide Law Review 45-77
11.
12.
13.
63
In the summing up careful directions are required in respect of any issues of expert evidence and guidance should be given to avoid confusion caused by areas of expert evidence where no real issue exists. The judge should explain to the jury the relevance of the random occurrence ratio in arriving at their verdict and draw attention to the extraneous evidence which provides the context which gives the ratio its significance, and to that which conflicts with the conclusion that the defendant was responsible for the crime stain. In relation to the random occurrence ratio, a direction along the following lines may be appropriate, tailored to the facts of the particular case: 'Members of the jury, if you accept the scientific evidence called by the Crown this indicates that there are probably only four or five white males in the United Kingdom from whom that semen stain can have come. The defendant is one of them. If that is the position, the decision you have to reach, on all the evidence, is whether you are sure that it was the defendant who left that stain or whether it is possible that it was one of that other small group of men who share the same DNA
characteristic^.'^ The Australian equivalent is ~ a t c h a , ' ~ in which the accused was charged with having sexual intercourse with a woman without her consent. The victim's underwear had human semen on them, and a forensic biologist gave evidence that there was a match at three loci between the semen and blood taken from the accused so that (a) with 97.5 per cent confidence, the relative frequency of occurrence of such a match was rarer than one person in 8,000, (b) with 95 per cent confidence, between one in 8,0000 and 20,000 and (c) since 1996, when the comparison was done, the 95 per cent confidence area was between one in 11,000 and 20,000. The case for the defence was that no sexual intercourse took place at all. It is worthy of note that the accused was an Afro-Carribbean from Guyana. The trial judge instructed the jury to ignore the statistical evidence except insofar as it did not exclude the accused as the person responsible for the seminal stain. The jury convicted. On appeal, the Court of Criminal Appeal ordered a retrial, holding (a) that the DNA evidence as presented should have been wholly excluded and (b) setting out general guidelines for the admission of such evidence in the future.
52
53
Ibid 370. (1998) 127 NTR 1.
GOODE - SOME OBSERVATIONS ON EVIDENCE OF DNA FREQUENCY
The principal reason why the DNA evidence should have been rejected was the basis on which it was calculated - in particular, the composition of the database with which it was compared. The expert gave evidence that the database used was based upon the 'general population' of the Northern Territory, excluding full-blood Aboriginals. The court held that this was insufficient: In a case such as this, where the prosecutrix's evidence has identified the accused as the perpetrator, the relevant question is: 'What is the probability of obtaining a matching analysis of the crime scene if someone else left it?' In order to answer that question, the evidence as to probability must be based on the whole population, and not on a limited part of it. ... It therefore could not be accepted in this case that the likelihood ratio was statistically valid, bearing in mind that Aboriginals,, who were excluded from the database from which the ratio was derived, comprise a significant proportion of the Northern Territory's population. On this basis, the statistical evidence should have been rejected.. .54
The court then prescribed a set of rules for the use of DNA evidence which clearly have been adapted from those set out in R v Doheney and Adams. However, the rules are by no means the same. There is a greater procedural emphasis. The rules are set out below, with the matching R v Doheney and Adams rule indicated where appropriate: 1.
54
Whenever DNA evidence and statistical evidence based thereon is to be adduced, the Crown should serve on the defence prior to the committal hearing a statement or statements from the expert or experts the Crown intends to call, which provide details of DNA
Ibid 6-7 citing Pantoja (1996) 88 A Crim R 554; M Redmayne, 'Doubts and Burdens: DNA Evidence, Probability and the Courts' [I9951 Criminal Law Review 464; and B Robertson and T Vignaux, 'DNA on Appeal - 11' [I9971 New Zealand Law Journal 247. The importance of raciallcultural subgroups has been explained in the following way: 'The answer lies in the way that forensic scientists calculate match probabilities. Their calculations are based on mathematical models obtained from population genetics theory. Such models generally assume that individuals in populations mate at random with respect to genotype. This assumption may be significantly violated if populations are genetically structured, that is, if they contain genetically ethnic groups or races, and especially if individuals mate "assortively" or preferentially with respect to ethnic group or race. If so, it may not be safe to calculate how often pairs of different types of any particular gene (its "alleles") may occur in combination (their genotype frequency) by simply multiplying together the frequencies at which each individual allele is found in the population as a whole. Used in the courtroom, this method would be potentially biased against defendants.' G K Chambers, et al, above n 32, 448-9. The interested reader should read further from this quotation.
(2002) 23 Adelaide Law Review 45-77 testing carried out, the nature of the matching DNA characteristics between the DNA in the crime sample and the DNA obtained from the defendant, and details as to how the calculations of the likelihood ratios have been carried out which are sufficient for the defence to scrutinise the basis of the calculations. [R v Doheney and Adams 21 2.
Provided that the expert has the necessary data, it may then be appropriate for it to be indicated how may people with the matching characteristics are likely to be found in Australia, or in a more limited relevant subgroup, for instance, the sexually active males in the Darwin area, depending on the circumstances of the case. [R v Doheney and Adams 71
3.
If the Crown intends to supplement or change the DNA evidence or the statistical evidence based thereon, after the committal hearing, it should serve such additional statements as are necessary to comply with guideline (1) in sufficient time prior to the trial for the defence to be able to meet that evidence. [This merely extends R v Doheney a n d Adams 21
4.
The forensic section of the Northern Territory Police Department should make available to a defence expert, if requested, the databases on which the calculations have been based (but not information which identifies particular individuals included in the databases). Any failure to do so in time for the defence expert to be available to assist the defence at the trial may lead to the exclusion of any statistical evidence at the trial. [R v Doheney and Adams 31
5.
Wherever possible, sufficient of the crime scene sample should be kept by the forensic section of the Northern Territory Police Department for re-testing, and made available to the defence for that purpose, upon request. [No equivalent]
6.
It is not necessary for the Crown to lead evidence from an expert in population genetics or from another scientific expert as to the statistical validity of the databases kept by the forensic section of the Northern Territory Police Department where the defence notifies the Crown that this is not in issue, or where objection is not taken at the trial. [No equivalent but a specific instance of R v Doheney and Adams 41
7.
A scientist other than a population geneticist or an expert in a statistical discipline may have sufficient qualifications derived from professional experience and personal familiarity with the data on the relevant database and published population statistics to be
66
GOODE - SOME OBSERVATIONS ON EVIDENCE OF DNA FREQUENCY permitted to give evidence of the likelihood ratios in the relevant population. If the Crown proposes to adduce evidence of this kind from such a scientist, the Crown should serve on the defence in accordance with guidelines (1) and (2) a statement of the scientist's qualifications and experience. [No equivalent] 8.
Disputes as to the admissibility at trial of DNA and statistical evidence, including the qualifications of witnesses, should be determined wherever possible by utilising the procedure provided for in s 26L of the Evidence Act. [R v Doheney and Adams 41
9.
Experts called to give statistical evidence should be led by the Crown as to any assumptions made in their calculations which, even though widely accepted, are not supported by empirical research, including: (a) Hardy-Weinberg equilibrium;55 (b) where the offender is of a racial group or sub-group for which there is no valid database and a general database has been used which does not take that fact into account, that fact. [No eq~ivalent]~~
10
Experts should not give evidence as to the likelihood that it was the defendants's DNA found at the crime scene or use terminology suggesting that he or she is expressing such an opinion. [R v Doheney and Adams 9 and
The Supreme Court of British Columbia has been even more concise in its instructions:
55
56
57
This matter has been discussed above. There is a fertile field of reported cases about the validity of calculations involving more or less exotic populations. Aside from those cases which can be accessed with ease in Australia, such as R v Pantoja (1996) 88 A Crim R 554 and cases involving Aboriginal defendants, see also accounts of cases involving Canadian indigenous people recounted in T R McDonald, above n 36, 1 3 4 . The issues here are well litigated and discussed and there is no profit in repeating it all. There are also detailed questions about the scientific qualifications of some expert witnesses to give opinion evidence in relation to highly confined racial subgroups. An example is R v Lafferty (1993) 80 CCC 3d 150, where the evidence was admitted. The same basic questions arise. Is this a question of admissibility or weight? The scientific complications of this factor are beyond the scope of this article. (1998) 127 NTR 1, 8-9, citing R v Adams [I9971 1 Cr App R 369; R v Pantoja (1996) 88 A Crim R 554; R v Vivona (unreported, V CCA, 12 September, 1994); R v Green (unreported, NSW CCA 26 March, 1993).
(2002) 23 Adelaide Law Review 45-77 [I]t can be made sufficiently clear to the jury that: 1) the estimates are not intended to be precise; 2) they are the products of mathematical and scientific theory, not concrete facts; 3) they do not purport to define the likelihood of guilt; 4) they should only be used to form a notion of the rarity of the genetic profile of the accused; and 5) the DNA evidence must be considered along with all the other evidence in the case relating to the issue of identifi~ation.'~
In general terms, the future seems clear. Law enforcement will use DNA evidence in an increasing number of cases because it regards DNA technology as the new law enforcement breakthrough of the next decade.59The technology is increasing in sophistication and an interested reader can find detailed summaries of scientific progress in this area in a sufficient number of places to not warrant inclusion in this note. However, two more unorthodox comments require emphasis. The first point to be made is that, as scientists revise, refine and perfect the science of DNA matching, the confidence levels of the evidence that they bring to the trial will become higher and the random occurrence ratios which they will be able to produce to the jury will become more confident and more extreme.60This points, of course, to the need to examine with the greatest care how the jury will be told to deal with such evidence - further comment will be made about that below. But it also points to the fact that the trend is quite likely to be that if there is error in the evidence, it is more likely (in the statistical sense) to be in the chain of evidence; such as the collection of the sample, the handling of the relevant samples, the procedures which safeguard the integrity of the process at the laboratory, and the computer processes which will govern the matching of the samples within the data bases.61 Put at the level of oversimplification, if the scientific evidence is to the effect that the chances of a DNA match or mismatch are in the order of millions or billions to one, the chances that something else went wrong, however incalculable 58 59
60 61
R v Singh (1996) 108 CCC (3d) 244. Some judges also feel the same way: 'DNA evidence.. . is the most dramatic forensic evidence ever discovered...': Kerans JA in R v Love (1995) 102 CCC 3d 393. See, for example, I W Evett, et al, above n 16, 342. See the (US) Department of Justice, National Institute of Justice, The Future of Forensic DNA Testing: Predictions of the Research and Development Working Group; A Report From the National Commission on the Future of DNA Evidence (2000) at 7: 'The scientific foundations of DNA are solid. Any weaknesses are not at the technical level, but are in possible human errors, breaks in the chain of evidence, and laboratory failures.'
68
GOODE - SOME OBSERVATIONS ON EVIDENCE OF DNA FREQUENCY
in a mathematical form, in other areas are very much more likely to be less than that. This first point, then, leads to three conclusions. The first is that issues such as the chain of evidence and the possibility of contamination or mismatch of samples are likely to come under closer examination at trial than has hitherto been the case, and necessarily so. The second is that, however devoted to their task they may be, scientists working in laboratories or computer operators running a large DNA database may not be used to guaranteeing the entire process by which the samples are dealt with beyond a reasonable doubt. The third is that, as a consequence of this, the relevant processes will come under increasing pressure to be documented to the point of standing up to the most rigorous of examinations in court. These issues all need to be addressed in a systematic way which not only inspires public (and judicial) confidence, but which also ensures a minimum of litigious confusion and contention. The latter two conclusions require some brief amplification. It is simply not the case that DNA profiling is analogous in any specific way to fingerprinting.62A fingerprint is just that and, in interpreting it, the expert exhausts the possibility of the print. DNA 'fingerprinting' examines a minute part of the whole DNA structure at artificially determined points. A large number of other loci could be examined but are not. In addition, the interpretative skill of the expert is much more critical in DNA profiling than in fingerprint analysis, simply because the science is more complicated by quite an order of magnitude. Scientific experience and interpretation is central to the process.63 The more common problems that can
62
63
See A C Bernasconi, above n 48. At 1009 (footnotes omitted): 'As a general rule, fingerprints are more conclusive and reliable in establishing a link between a sample and a criminal suspect. Researchers theoretically have the ability to obtain and analyze all of the information fingerprints provide. In direct contrast, however, scientists examining DNA samples possess the capability to analyze only one millionth of the three billion units of human DNA. Though current DNA analysis capabilities lead to conclusions about the source of the DNA sample, a great deal of disagreement and inconsistency remains over the scope of DNA analysis required to produce a result as conclusive as an examination of fingerprint samples.' 'The interpretation of results is a matter of professional judgment and expertise. Not every situation can or should be covered by a preset rule. It is important that the laboratory develops and implements written guidelines for interpretation of analytical results. ... The laboratory's interpretation guidelines should be based upon validation studies, data from the literature, instrumentation used, andlor casework experience'. Introduction to the published guidelines of the Scientific Working Group on DNA Analysis Methods (SWGDAM) quoted in R v Kavger [2001] SASC 64, para [143].
(2002) 23 Adelaide Law Review 45-77
69
occur in profiling are neatly summarised in R v ~ a r ~ eThey r . are ~ ~listed as: peak height imbalance, stutter, pull-up, non-specific artefacts, stochastic effect and the identification of mixtures.65 Such technological problems are quite alien to the simpler science of fingerprint comparison. So too the mathematics involved in the calculation of match probability, described above, which produce the startling figures so seemingly probative of guilt but which involve a set of complex hidden assumptions. Those assumptions may be commonly accepted in the scientific community, but far too much can be made of the infallibility of scientific process and results. The second point to be made begins also with the advance of technology. When the issue was examined first in the 1980s and then in the 1990s, it seemed clear that what was contemplated was the use of DNA technology in the most serious of cases - murder, other homicide, serial crimes (especially serial sexual crimes), rape, and so on. Those were the kinds of cases in which one would expect to find blood or semen samples and, with very few exceptions, those appear to be the cases which have reached the courts to this point. In the next decade, perhaps much sooner, DNA technology may be able to reach the point of typing microscopic skin cells ordinarily shed as a part of mere presence at any place all of the time and therefore left by any person at the scene of a crime at any time.66 The third point to be made concerns the chances of laboratory error. As has been seen in discussion above, the courts at present take the view, put forward by expert opinion evidence, that the chances of laboratory error are not susceptible to quantification and should therefore not form a part of the match probability evidence. Nevertheless, it has been suggested in the United States that the standard or proficiency of the laboratory concerned should be the subject of evidence and should even be a condition for admissibility of the e ~ i d e n c e . ~It' does not appear that any Australian court has formally taken this position. However, in R v Karger, Mullighan J laid considerable emphasis on the adherence of the laboratory to the Scientific Working Group on DNA Analysis Methods (SWGDAM) guidelines, the accreditation and testing of the laboratory by the National Association of Testing Agencies (NATA) and its Scientific Advisory 64 65
66
67
Ibid, paras[116]-[136]. This is not, of course, an exhaustive list. It is not intended to explore these problems hrther. It suffices for the purpose of the present discussion to note that they exist. For example, in R v Rees [2000] NSWSC 544 (16 June 2000) The amount of DNA reported was 0.3 nanograms per 20 microlitres. See, generally, the discussion of the US material on this in E J Imwinkelried and D H Kaye, 'DNA Typing: Emerging Or Neglected Issues' (2001) 76 Washington Law Review 413, 460ff.
70
GOODE - SOME OBSERVATIONS ON EVIDENCE OF DNA FREQUENCY
Group and the fact that the laboratory had a practice of requiring agreement on interpretation by two separate tests by two separate qualified experts - with recourse to a third should any difference of opinion emergee6*There is no indication that such matters will become conditions of admissibility in Australian courts - the indications are that the courts will regard them as matters of weight to be left to the jury - but there is in this question a note of caution for testing laboratories and a possibly productive field for cross-examination by defence counsel. In concert with this development, it is clear, as noted above, that law enforcement wants to be able to use this kind of technology as much as is feasible. In the most general of terms, law enforcement is of the opinion that the talung of a satisfactory DNA sample for the purpose of inclusion in the data base should be a regular part of the normal process during charging of a suspect with a serious offence, together with fingerprinting, photographing and the filling in of an identification card by the person charged in their own handwriting, all of which (except DNA sampling) routinely occur now. Taken together, these developments pose a very real question. How is DNA identification seen as a tool of law enforcement in, say, the next 20 years - a routine procedure for most crime of any consequence, or a specialist tool used for very serious cases? The United Kingdom has very clearly opted for the first position.69It is aiming for a database of 3 million of 'currently active criminal^'.'^ At the moment, no Australian jurisdiction has gone so far. The arguments are not simple, but, at their most general, it can be argued on the one hand that having a huge national DNA offenders databank including every common or garden assault in or near a drinking establishment uses a sledge hammer to crack a nut and is an invasion of privacy quite out of proportion to the offence committed. On the other hand, it can be argued with conviction that, if our privacy safeguards are sound enough, what is the harm? All that will happen is that it will be much easier to identify and, quite clearly, to eliminate, potential offenders. But are our safeguards good enough given that they remain untested? We simply do not have general privacy laws that deal with this kind of material. 68
69
70
See generally [2001] SASC 64, para [142] ff. The reforms were contained in amendments made to the Police and Criminal Evidence Act, 1984 by the Criminal Justice Act, 1994. See, generally, the account given in MCCOC, Report, Model Forensic Procedures Bill and the Proposed DNA Database, (1999). Forensic Science Service (UK), Press Release 14 November 2000, . 'Currently active criminals' is a curious phrase. If they are in prison, they are not currently active. If they are suspects, they are not yet criminals.
(2002) 23 Adelaide Law Review 45-77
71
All depends on this sort of subjective vision. But there are more pragmatic considerations as well. This technology comes at a price. Quite apart from police resources, it should be noted that, should the law enforcement vision prevail, the financial implications of this alteration in criminal investigation focus are potentially enormous. On relatively recent figures, the Victorian Forensic Science Centre is already the best funded such centre in Australia with 17 FTE DNA scientists. It has been estimated that the additional costs of a general DNA database regime covering the needs of law enforcement will require ongoing commitments of more than $5 million a year in forensic science requirements in Victoria alone. The wider the database range, the more conservative those estimates appear to be.71 There are, therefore, two questions yet to be answered in this country which the law enforcement vision has yet to ask. The first is the budgetary question: whether Governments are willing to allocate the necessary funds to allow the vision reality and, if not, whether law enforcement is willing to divert resources from other criminal investigation techniques to make up any shortfall.72This question will, in the end, have to be answered by more than the vision and a demand for the resources to do it. It will ultimately be based on value for money and the calculus of returns for outlay. The second question is the start of the value for money question. Is this society willing to invest millions of dollars in high technology science and scientists in order to apprehend or eliminate from investigation (for example) the breaker of a domestic violence order, the car thief or the minor shop breaker? It seems clear from a combination of popular fiction and high profile reality that the early 1990s calculation of public opinion was right that the investment was and remains worthwhile for murderers, rapists, robbers and serial 71
72
The UK database currently stands at 1,150,000 samples and employs 2,500 people. See Forensic Science Service (UK), Press Release 14 November 2000, . In November 2000, FSS announced additional Government funding of f202m over three years for expansion of the database. See Forensic Science Service (UK), Press Release 14 November 2000, . In the WoolfReport. See Lord Woolf, 'Medics, Lawyers and the Courts' (1997) 16 Civil Justice Quarterly 302. See the recommendation of what is colloquially known as the 'Martin Report': Standing Committee of Attorney's General, Working Group on Criminal Trial Procedure, Report (September, 1999) available at
