R v Humphrey

R v HUMPHREY
[1999] SASC 67

Criminal (Reasons for Ruling)
Bleby J

The accused was charged with the murder of Eugene Charles McLean at Lyrup Flats, in the South Australian Riverland, on 23 October 1997. There were no eye witnesses to the murder, and one of the issues in the case was whether the Crown could prove that it was the accused who killed McLean. What appeared to be blood stains or spots, one on the upper toe area of the left boot being worn by the accused at the time of his arrest (marked PWH4.A), and one from his jeans (marked PWH6.D), were submitted for DNA analysis by Ms Williams at the South Australian Forensic Science Centre. In the declaration submitted by Ms Williams she expressed the opinion that in respect of both spots the accused was excluded as the source of the DNA and the deceased was not.
In respect of the stain on the boot she further expressed the opinion that greater than 99.99% of the population would be excluded and that the statistical likelihood of that profile being repeated in the general population was less than one in ten thousand. In respect of the sample from the jeans, approximately 99.8% of the population would be excluded as the source of the blood stain.
Before the empanelling of a jury the defendant sought an order that that “frequency” evidence of Ms Williams was inadmissible unless it were proved that the data base from which the frequency evidence was derived was an appropriate one and unless it were proved that there was an absence of dependence between the alleles from which the frequency evidence was derived.
In order to resolve the question of admissibility I heard evidence on the voir dire from Ms Williams, a forensic scientist whose qualifications and experience include molecular biology, and Associate Professor Chaseling, an Associate Professor in Applied Statistics at Griffith University, both called by the prosecution. I also heard evidence from Dr Atchison, Manager, Molecular Biology, Victorian Institute of Forensic Medicine, part of the Department of Forensic Medicine at Monash University.
After hearing argument as to the admissibility of the frequency evidence I ruled that it was admissible provided that (if such matters were not otherwise agreed) evidence was also led from which the jury could determine the appropriateness or otherwise of the data base or bases used by Ms Williams, in respect of such matters as size, representation of the population, lack of dependency between and within particular loci and any matters which might affect the validity or weight of the frequency evidence. I indicated at the time that I would publish reasons for that ruling, and these are those reasons.
Deoxyribonucleic acid, or DNA, is contained in the chromosomes in the nucleus of most cells in the human body, and contains the code for each of a person’s characteristics or that person’s genetic makeup. The testing in this case was by means of the polymerase chain reaction (PCR) technique which has been described in many other cases. I am content to rely on the description of the process which appears in R v Jarrett (1994) 62 SASR 443, and I will not repeat it. It is sufficient for present purposes to say that on each of the many loci on each strand of DNA there are two alleles - one contributed by the male parent and one by the female. At a number of these loci the alleles will vary in size and can be measured by the PCR technique. Other than in the case of identical twins, no two persons will contain the same DNA profile. However, techniques do not yet exist for conveniently measuring and recording the alleles at every locus, of which there are many millions, and thus it cannot be said with absolute certainty that the DNA taken from any two samples is identical where the origin of one of those samples is unknown. However, a number of loci have been identified where variations between individuals are known to exist, although a number of individuals may bear the same two alleles at a given locus.
By means of statistical sampling of the population it is possible to calculate the mathematical probability of particular combinations of alleles at given loci being repeated in the population represented by the sample. Hence, if the alleles are measured at each locus in a sample of unknown origin, an estimate can be made of the frequency of that same profile occurring in the population represented by the sample contained in the data base. This is sometimes expressed as a certain percentage of the population being excluded as the donor of the unknown sample, or as the frequency of occurrence being one in a certain number or range of numbers according to a 95%, 99% or some other percentage probability factor, depending on the size of the sample.
In respect of the boot stain (PWH4.A) the sample was able to be typed at 3 loci, namely those known as D1S80, VWA and THO1. In respect of the stain on the jeans (PWH6.D) the sample was successfully typed at 2 loci, namely VWA and THO1. The method of analysis, testing and recording of the results of these samples against blood samples taken from the deceased was not challenged. The only challenge was to the adequacy of the data base from which the relevant frequencies were calculated.
The data base used by Ms Williams was the general data base held at the Forensic Science Centre, South Australia. That was composed of profiles collected at the Centre from 1993 relating to members of staff of the Centre, persons (excluding children) tested for paternity and complainants and defendants in criminal matters. No profile was excluded from entry into the data base on the basis of race, although the racial composition of the data base had not been recorded. The number of individuals in each data base was as follows:

System No of Individuals

D1S80 1002

VWA 538

THO1 540

The reason for the lower numbers in the VWA and THO1 systems was because the recording for those loci did not commence until 1995.
Before elaborating on the argument of Mr Barrett QC for the defendant, I need to say a little more about the circumstances, at least as gleaned from the declarations submitted by the prosecution. It was alleged that the murder took place at what might be described as a “humpy” or camp on the banks of the River Murray near Lyrup. It appears that there were a number of such humpies in the vicinity. The deceased was killed at the one where he was living at the time. The accused had been staying at another, approximately 150 to 200 metres away. The Crown case alleged that the accused, until approximately a week before the murder, had been living in Adelaide for at least some months. Between the time of leaving Adelaide and his arrest, less than 48 hours after the murder, besides living at the camp, he had been in Berri, Loxton (twice), Lyrup and places in between. There was no evidence as to when the blood samples were placed either on his boot or on his jeans, other than that they were there at the time of his arrest. The donors of those samples were quite unknown. Some of the people with whom he had been associating on the river bank were Aboriginal or of Aboriginal descent. There was no suggestion in the material before me that the deceased was other than caucasian.
Mr Barrett’s first argument was that there was a series of cases which established that the evidence of frequency could not be admitted without the Court being satisfied that the data base was truly representative of the population from which the samples could have come. He did not argue that this was a discretionary exclusion based on fairness to the accused in his trial. His argument was that unless that condition could be satisfied, the evidence was inadmissible. In order to be satisfied that the data base was truly representative of the population Mr Barrett submitted that the Court first had to determine the population from whom the samples could have come. In this case that could have been from the population of the banks of the River Murray, the population in the Berri area (including Lyrup), the population of the South Australian Riverland or the population of South Australia. Dr Atchison, in his evidence, had initially suggested the first of these. However, when it was pointed out that the accused had also been in Adelaide and other places, he conceded that some other criterion might have to be devised. His opinion was, however, that unless the appropriate population was identified, a data base relating to any other section of the population was of no use. Mr Barrett therefore argued that it was for the Court to determine the relevant population from which the sample came, and if there were no data base truly representative of that population, the evidence should be excluded.
Mr Barrett went further, however, and submitted that once the relevant population was defined, if there existed a data base relevant to that population, the evidence should also be excluded unless the Court could be satisfied that the data base was an accurate statistical sample of that relevant population. The submission was that if, for example, I was satisfied that the appropriate data base was one which was to reflect the population of South Australia, the data base used by Ms Williams could not be shown to be a proper sample of that population, and any evidence of frequency relying on that data base should be excluded for that reason as well.
Without descending into detail Professor Chaseling expressed the view that the SA general data base was sufficiently representative of the population for the purpose. Dr Atchison disputed that view. For the purposes of my decision on this argument it is not necessary to resolve that conflict, although Mr Barrett’s argument was that I should, and should do so in favour of Dr Atchison’s view. But whatever view I took would resolve the question of admissibility. It was not a discretionary exclusion.
The first part of the argument can be dealt with quite simply. The purpose of the evidence in question is to establish the statistical probability that the donor of each of the samples found on the accused’s clothing might not have been the victim. For the Court to determine before the trial begins what is the appropriate description of the population from which the samples might have come is to make a judgment on evidence which it has not heard, and which judgment may or may not turn out to be correct when the evidence unfolds before the jury. In any event such determination will properly be open to immediate criticism by the defence because it was too restricted or too generous or failed to include some particular racial or other sub‑group, thereby adding significantly to the likelihood of an unsafe or wrong verdict.
The assumption on which the Crown begins with any such DNA analysis is that it does not know who the donor was or from which particular section of the population it came. The purpose of the analysis is to attempt to show the frequency within the population at large with which such profile is likely to occur. It is not for the trial judge to make a factual determination as to the population group which the data base should represent. I say so for two reasons.
In the first place, to do so means that one must immediately make an assumption about the limits of the population from which the unknown samples came. Without evidence clearly delimiting that population group, which may happen in very rare cases, such an assumption cannot be made.
Secondly, to do so is to usurp the function of the jury. If a data base has been used which, for some reason, is not shown to be representative of the population from which the unknown sample may have come, or if it appears that the data base is not representative of the racial group from which the known sample comes and that the DNA profiles of that racial group do or may bear different characteristics from those revealed in the population represented in the data base, then that will no doubt be exploited before the jury. The jury will be able to give the evidence such weight as it deserves in the light of any criticisms that are advanced, taking into account any suggested weaknesses on which the opinion is based.
Mr Barrett argued, however, that there was an emerging line of authority since R v Jarrett (1994) 62 SASR 443 which indicated that such matters, including the representative nature of the sample within the appropriate data base, if such data base were shown to exist, went to admissibility, and were to be determined by the trial judge.
As I have said, the selection of the appropriate population for the data base will depend on all the evidence led at the trial, possibly including some expert evidence. Whether a particular data base is sufficiently representative of that population for statistical purposes will depend on expert evidence.
Mr Barrett conceded, in my view, properly, that if R v Jarrett were the only relevant authority he would have difficulty in persuading me that I should rule against the admissibility of the evidence. That case concerned the admissibility of DNA “frequency” evidence in respect of which there was conflicting expert testimony. Mullighan J said (at 451):
“The starting point is that all evidence is admissible if it is relevant to a fact in issue and may only be excluded if it falls within one of the well‑known policy grounds for exclusion or within one of the recognised grounds for the exclusion of evidence in the exercise of the trial judge’s discretion. Expert evidence is no exception. Once it is determined that expert evidence is relevant to a fact in issue, and there is no reason on policy or discretionary grounds for its exclusion, it should be admitted even though it is contested and there is apparently credible expert testimony to the contrary: Chamberlain v The Queen (No 2) (1984) 153 CLR 521, per Brennan J at 598.”
His Honour referred to a number of passages from the judgment in Chamberlain to the effect that it was inappropriate for an appellate court to decide which of two conflicting experts should be accepted, and that that was a matter for the primary finder of fact. His Honour continued (at 453):
“In my view, admissible expert testimony relevant to a fact in issue and of probative value may not be withheld from the jury. Once the expert testimony is found to be admissible, it should be admitted even though challenged and it is a matter for the jury, properly and adequately directed, to decide in the context of all of the evidence in the case, whether the evidence should be accepted ...”
After discussing a number of authorities His Honour concluded (at 456, 458):
“The resolution of conflicts in expert testimony and the assessment of the weight to be attached to it is the function of the jury. If there is no reason to exclude evidence in the exercise of discretion, it should be admitted and appropriate directions given to the jury as to how any conflicts in the evidence should be approached and what use can be made of the evidence which is accepted ...
To make a judgment before the trial commences that a jury could not understand evidence or resolve a dispute between experts is, in my view, to make a judgment at the wrong time. That judgment can only be made when all of the evidence in the case is before the jury and then it is a judgment, on a question of fact, for the jury to make, with proper directions.”
I respectfully agree with all that Mullighan J said. To the extent that there may be conflicting expert evidence as to the use that can be made of particular data bases, that is a matter for the jury, properly directed, to determine. If they are satisfied, having heard the evidence, that there are limitations on the data base used by Ms Williams, it will be for them to determine what weight (if any) they can put on her evidence. In my view, the other cases to which Mr Barrett referred do not qualify that approach.
Mr Barrett referred to R v Green, an unreported decision of the Court of Criminal Appeal of New South Wales (Gleeson CJ, Cripps JA and Abadee J, 26 March 1993). In that case, which incidentally was decided before Jarrett, evidence of the DNA testing and of the similarity between the sample obtained and that of the accused was held to have been properly admitted at the trial. However, it was also held that evidence of frequency was properly excluded. The reasons for decision on this point are somewhat unsatisfactory. It appears that they were related, at least in part, to a very small sample of full‑blooded aboriginals from the Northern Territory. The justification for exclusion seems to have been the unduly prejudicial nature of the evidence which, in the circumstances, could not be tested by the accused. It therefore appears to have been the upholding of a discretionary exclusion by the trial judge. If I am correct in that, it does not assist Mr Barrett’s argument.
R v Pantoja (1996) 88 A Crim R 554 was a case in which, although the accused was charged with the murder of his wife, it was relevant to the prosecution to establish that he had, shortly beforehand, had sexual intercourse with his wife’s sister. It was sought to do that by means of (inter alia) a comparison of DNA from a vaginal swab taken from the sister‑in‑law with DNA from the accused. Probability evidence was admitted at the trial but the appeal was allowed on the footing that it should not have been.
The objection taken at the trial was based on the race of the accused and of the sister‑in‑law (South American Indians). There was evidence that blood groupings of that race would be different from those of Australians of European stock. The frequency evidence was based on data bases reflective of the Australian general population. There was no evidence that such data bases included DNA from South American Indians. The argument was that if in fact the DNA from the swab was donated by a South American Indian (be that the accused or some other member of that community), it could well reveal a lower frequency of occurrence against the Australian population than would be the case if the donor were European Australian. That argument was rejected by the majority of the Court of Appeal (Hunt CJ and Hidden J). They considered that because of the possibility that the person who had sexual intercourse with the appellant’s sister‑in‑law may have been of that race required a direction to the jury that the chance or coincidence of a match thrown up by the use of general data bases would have to be reduced in order to take that possibility into account (at 563). In other words, that would accord with the view I previously expressed that the matter is an appropriate one for the jury. Abadee J on the other hand expressed the view that if the suspect belonged to a particular racial group or sub‑group, the preparation of a specific data base rather than a general data base might well be warranted. However, in allowing the appeal he also agreed with Hunt CJ that the evidence was inadmissible because it was not shown that the data base used (which appeared to be very small) was statistically valid, and that in those circumstances the evidence “would have overawed the jury by the seemingly scientific garb in which it was presented, with the very real risk that they would have thought that it had greater weight than it may have been capable of bearing” (at 561‑562). In so saying Hunt CJ relied on R v Duke (1979) 22 SASR 46 at 41 and 48, an authority to which I shall return. Thus, the exclusion in that case was based on discretionary considerations and not on a requirement that the trial judge himself should be satisfied as to the validity of the data base. I do not consider that R v Pantoja supports Mr Barrett’s argument.
The next case to which Mr Barrett referred was R v Milat (1996) 87 A Crim R 446 in which there was an objection taken to the leading of DNA evidence relating to the identity of blood, said to be from one of the victims, on a piece of rope in the possession of the accused. Objection was taken to the statistical validity of the data bases concerned. They were taken from the same sources as those in the data bases used by Ms Williams in this case, although they related to New South Wales residents rather than South Australian. Whilst the expert concerned conceded that the material could not be said to have been truly random of the New South Wales population, they were able to be demonstrated to be “representative” by reference to comparison with other data bases - a method of testing similar to that adopted by Associate Professor Chaseling in this case in relation to the South Australian data bases. Hunt CJ on that basis accepted the statistical validity of the data bases used in that case and admitted the evidence subject to some further directions to the jury relating to confidence limits, and the bases of them, which were reflected in the estimate made.

It is true that Hunt CJ appears to have treated the validity of the data bases as being a pre‑requisite to admissibility because (at 448) he was “satisfied that the sizes and the nature of the data bases used in this case are within the generally accepted limits of what is statistically valid...”. In so doing he cited Frye v United States 293 Fed 1013 (1923); US v Baller 519 Fed 2d 463 at 466 (1975) and Gilmore [1977] 2 NSWLR 935 at 939‑941. However, as appears from Gilmore at 939, quoting with approval from United States v Baller:
“‘Absolute certainty of result or unanimity of scientific opinion is not required for admissibility. “Every useful new development must have its first day in court. And court records are full of the conflicting opinions of doctors, engineers, and accountants, to name just a few of the legions of expert witnesses.”...Unless an exaggerated popular opinion of the accuracy of a particular technique makes its use prejudicial or likely to mislead the jury, it is better to admit relevant scientific evidence in the same manner as other expert testimony and allow its weight to be attacked by cross-examination and refutation...’” (Emphasis added)
Once again, the question would appear to have turned on a discretionary exclusion if popular opinion as to the accuracy of a particular process makes the evidence prejudicial or misleading. I certainly do not take Hunt CJ to be saying that once the trial judge is satisfied as to the validity of the data base that question should not be challenged before the jury.
The final case referred to by Mr Barrett was R v Mitchell (1997) 130 ACTR 48. In that case Higgins J admitted DNA evidence, including frequency evidence. He directed the jury that the blood samples found on clothing were consistent with the blood deposited by the accused and the victim respectively and that the 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. He went on to direct the jury that “while 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 Dr Atchison and Professor Chaseling. In effect, the weight to be attended to the match was for them on such other basis as they might consider appropriate” (at 49). The case is reported as reasons which he gave for that direction. The offence had occurred in the ACT. DNA evidence was admitted where the probabilities were based on a Victorian data base of Dr Atchison. There was no evidence that the Victorian data base adequately reflected the ACT population at the date of the offence. There was also no evidence either that the sites or loci selected were independent or as to the validity of the size of the data base used. The evidence appears, nevertheless, to have been admitted, and it seems to have been because of the lack of validation that His Honour was moved to put the qualification he did as to the weight to be afforded to the evidence by the jury. I do not see that the case supports Mr Barrett’s argument.
I therefore consider that the frequency evidence is not rendered inadmissible by virtue of any alleged weakness in the validity of the data bases used, either as to the selection of the particular data base or as to whether it is sufficiently representative of the population group from which it comes. Any alleged weaknesses in such matters are properly to be considered by the jury, with appropriate directions from the trial judge, in order to assist them to give whatever weight they consider appropriate to the evidence.
In a similar vein Mr Barrett attacked the admissibility of the evidence because it could not be shown that there was a lack of dependence between or within the loci selected. Dependence occurs where it is shown statistically that particular combinations of alleles at one site tend to appear with particular combinations of alleles at another site, or alternatively that a particular combination of alleles occur together within one site. This could have its manifestation, by way of example, in the recognised combination of blue eyes and fair hair, or an observable dependency of one on the other. If it occurs, this could render invalid the calculation of the frequency prediction. That is done by multiplying together the frequencies of occurrence of the particular genotypes in the data base, known as the “product rule”. If dependencies occur this can result in unacceptable distortion of the frequency calculated.
Professor Chaseling carried out four recognised tests on the South Australian general data base, two of which tests Dr Atchison himself had used in respect of his own data base, and all of which he accepted as being recognised scientific tests. Professor Chaseling did this in respect of 206 individuals for whom information existed at all three loci tested.
To the extent that it is necessary to do so, I accept Professor Chaseling’s evidence. Dr Atchison’s criticism in the end seemed to be that not only was the data base itself not sufficiently representative of the South Australian population, but that the 206 samples used were not themselves random samples within the data base.
To the extent necessary I accepted Professor Chaseling’s evidence that the South Australian data base was sufficiently representative of the South Australian population. She had carried out similar testing against other data bases assembled in other parts of Australia. Except in respect of data bases said to be of full‑blood aboriginals only, she obtained what she, as a statistician, considered to be sufficiently similar results. Although none of the data bases had been able to be assembled by a purely random sampling, the South Australian data base was, she considered, sufficiently representative of the South Australian population. I accepted that evidence.
As to the selection of the 206 samples, Dr Atchison’s criticism was unfortunately not put to Professor Chaseling. However, in the end it was a matter more closely the province of an expert statistician than of a molecular biologist, and I accepted Professor Chaseling’s opinion, to the extent that it may have been necessary to do so.
However, once again, if there is criticism to be made relating to proof of lack of dependence within the data base, that is also a matter properly to be explored before the jury. If at the end of the day there is shown to be doubt as to lack of dependence within the data base, that will obviously affect the weight of the frequency evidence led.
For these reasons I considered that the frequency evidence was admissible. However, I added a proviso that evidence was also led from which the jury could determine the appropriateness or otherwise of the data bases used. I should indicate the reasons why I added that proviso.
The original declaration of Ms Williams provided no information as to the nature or source of the data bases used. It was only after the Rule 9 application was filed challenging the admissibility of the frequency evidence contained in her declaration that she then produced a further statement giving some information as to the composition of the data base, and that a statement from Professor Chaseling was filed directed towards the validity, in a number of respects, of the South Australian general data base. Those additional documents were filed and the evidence was led in response to the challenge to the admissibility of the evidence.
Whilst I have ruled that the frequency evidence is admissible, if an appropriate foundation is not laid as to the validity of the data bases for the purpose of calculating that frequency, and where that validity remains in dispute, the evidence of frequency may well carry little or no weight because the assertion of the expert is incapable of being tested.
By itself, DNA frequency evidence is powerful evidence going to the identity of the perpetrator of a crime or, in this case, to prove that the accused was in the presence of the victim at a time when the victim was shedding blood. There is a possibility that the jury may be overawed by the scientific garb in which the evidence is presented without explanation as to its foundation. There is, perhaps, a greater risk that they will be overawed by the opinion itself without the ability of the accused to challenge the foundations on which the opinion is expressed. A judge sitting alone may well afford no weight to the evidence once it is realised that the foundation for the opinion has not been or cannot be justified. However, when such evidence can and no doubt does have such a powerful 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, as appears to have been the view taken (for different reasons) in some of the cases to which I have referred.
In R v Duke (1979) 22 SASR 46 the trial judge had declined to exclude certain evidence going to the identity of an accused charged with rape. In the course of his judgment in upholding the admissibility of the evidence King CJ said (at 48):
“Generally speaking the fact that the weight to be attached to a piece of evidence is slight does not render its introduction to the jury dangerous. It must be assumed that the jury will give the piece of evidence the weight that it deserves. The danger arises only if there is something in the nature of the evidence or the manner of its presentation which would render it gravely prejudicial in the sense referred to above. Where does the suggested prejudice lie? There is nothing in the intrinsic nature of the evidence which could give rise to prejudice. In this the case differs from those cases in which the evidence discloses a criminal record, or conduct or circumstances which may show the accused in a bad light in relation to matters which are extraneous to the issues in the case, or features of the case which might inflame the minds of the jury and thus deflect them from a dispassionate consideration of the issues. Is the fact that the evidence includes scientific opinion as to the significance of the physical evidence capable of creating prejudice which could lead to the exclusion of the evidence? I would not wish to say anything which might be thought to limit a trial judge’s discretion, which is so necessary to the discharge of his overall duty to secure a fair trial. There may be unusual cases in which the judge has reason to fear that the jury will be overawed by the scientific garb in which the evidence is presented and will attach greater weight to it than it is capable of bearing. It is to be remembered, however, that under our system, whatever criticisms of it there might be, the assessment of the weight to be attached to expert evidence, as to other evidence, is the function of the jury. That being the system, a trial judge must assume, in my view, that the jury is capable of understanding that it is not bound to accept the expert evidence, that it is capable of resolving conflicts of opinion amongst the expert witnesses, and that it will not be overawed by the scientific garb in which the evidence is presented to it. It is to be remembered, moreover, that for the purpose of enabling the judge to decide as to the exercise of his discretion the question of weight is to be judged upon the evidence which the prosecution proposes to present. It is not the function of the trial judge to hear conflicting expert evidence as to the significance of the physical evidence in order to determine for himself the disputed scientific issues. To do so would be to usurp the function of the jury.”
That passage merely reinforces the view I have taken that ultimately the expert evidence is a matter for assessment by the jury. However, this is one of those cases where, in my opinion, merely to lead the frequency evidence without the evidence on which that opinion may be questioned, thus denying the accused the opportunity to question what would normally be seen by the jury as a very telling piece of evidence, would be to cause great prejudice to the accused. Without proof of its foundation, where that is in dispute, the jury might have to be told that the evidence should be given little or no weight. But in those circumstances its prejudicial value is obvious. As King CJ pointed out, the exercise of such discretion is to be judged upon the evidence which the prosecution proposes to present. It was because of some uncertainty arising out of the manner and circumstances in which the additional declarations were provided that I considered it appropriate to add the proviso, lest it be thought that such additional material went only to the argument on the admissibility of the evidence, and did not form part of the evidence which the prosecution proposed to present.

[1999] SASC 67

25 February 1999