AI-generated summaries. Not legal advice. Always verify against the official judgment on NSW Caselaw.
← All decisions
22
Court of Criminal Appeal 16 October 2001

REGINA v GK

[2001] NSWCCA 413

Also reported as (2001) 53 NSWLR 317 (2001) 125 A Crim R 315
Sexual offences
EVIDENCE S137 EVIDENCE ACT 1995 PATERNITY INDEX RELATIVE CHANCE OF PATERNITY STATISTICS BAYES' THEOREM "PROSECUTOR'S FALLACY". (D)

Citation: Regina v GK [2001] NSWCCA 413
Court: New South Wales Court of Criminal Appeal
Date: 16 October 2001
Judge(s): Mason P, Sully J, Dowd J

Background

The accused (GK) faced charges of sexual assault against his step-daughter. Central to the Crown case was DNA evidence tendered to establish that GK had fathered the complainant's child. At the first trial in September 1999, Ford DCJ ruled on a voir dire (a hearing conducted in the absence of the jury to determine admissibility) that the DNA evidence could only be admitted to show that GK was not excluded as the father. The judge refused to allow numerical probability figures to be placed before the jury.

At a retrial in February 2000 before Moore DCJ, the Crown sought to lead evidence from further DNA testing conducted after the first trial. Moore DCJ concluded he was bound to follow Ford DCJ's earlier ruling and again excluded all arithmetical figures. A third trial proceeded on the same basis, ending in verdicts of not guilty on all eight counts.

The Director of Public Prosecutions then referred two questions of law to the Court of Criminal Appeal under s 5A(2) of the Criminal Appeal Act 1912 (NSW). The referral was for the purpose of obtaining authoritative answers, not to disturb the acquittals.

  • Whether a trial judge at a retrial (following a hung jury) is bound to follow an admissibility ruling made by the judge at the earlier trial.
  • Whether Moore DCJ erred in excluding numerical probability statistics derived from DNA testing, specifically the Paternity Index figures and the Relative Chance of Paternity percentage figures, under s 137 of the Evidence Act 1995 (NSW).

Decision

Question 1: Binding effect of prior admissibility rulings

The Court answered this question unanimously in the negative. A ruling under s 137 of the Evidence Act does not involve the exercise of a true discretion in the conventional sense. Rather, it requires the judge to make connected findings of fact and law upon which the section then operates. An evidentiary ruling made during a trial is final only for the duration of that trial. If the verdict is set aside on appeal, or if the jury fails to agree, the judge presiding at any subsequent trial is not bound by the earlier ruling.

Question 2: Admissibility of DNA probability statistics

The Court drew a clear distinction between two types of DNA-derived statistics. The Paternity Index, which expresses the relative probability that the accused rather than an unrelated man fathered the child, should have been admitted accompanied by careful directions to the jury warning against the "prosecutor's fallacy." That fallacy involves misreading a DNA match statistic as directly expressing the probability of guilt, rather than as one piece of evidence to be weighed with all others. The Court found that excluding the Paternity Index altogether distorted the experts' evidence without improving trial fairness.

By contrast, the Court upheld the exclusion of the Relative Chance of Paternity percentage figures. Those figures expressed the probability of paternity as a percentage so close to 100 percent that the Court considered there was a real and unacceptable risk that even properly directed jurors would treat the residual difference as negligible and effectively treat the question of paternity as conclusively resolved. Because the Paternity Index evidence already gave the Crown a fair opportunity to advance its case on paternity, the exclusion of the raw percentage figures was justified under s 137.

The Court also addressed, in general terms, the use of Bayes' Theorem in criminal proceedings. It expressed the view that inviting a jury to determine guilt by applying mathematical formulae to aggregate different items of evidence is inappropriate, endorsing observations made in English and other Australian authorities to the same effect.

Orders Made

  • Question 1: A trial judge at a subsequent trial, following a failure by the first jury to agree, is not bound to follow a prior judge's admissibility ruling. Answer: No.
  • Question 2: Moore DCJ erred in excluding the Paternity Index statistics. Answer: Yes as to Paternity Index statistics. Moore DCJ did not err in excluding the Relative Chance of Paternity statistics. Answer: No as to Relative Chance of Paternity statistics.

Key Takeaways

  • A ruling under s 137 of the Evidence Act 1995 (NSW) is not a discretionary ruling in the traditional sense. It involves findings of fact and law, not a residual judicial choice, and a judge at a retrial is therefore free to reconsider it afresh.

  • Where a hung jury ends a trial, the judge at any subsequent trial is not bound by admissibility rulings from the previous proceeding, consistent with the principle that those rulings expire with the verdict or non-verdict in which they operate.

  • Under s 137, the operative concept is "unfair" prejudice. Evidence is not rendered unfair simply because it points strongly toward guilt. Highly probative DNA statistics, properly explained through qualified expert evidence, cannot be excluded on the basis of probative weight alone.

  • The Court of Criminal Appeal confirmed that the "prosecutor's fallacy" represents a real and identifiable risk in DNA evidence cases. This fallacy arises when a jury conflates the rarity of a DNA match with the probability that the accused is guilty, treating a statistical likelihood as proof of identity. Appropriate jury directions can address this risk without requiring wholesale exclusion of the figures.

  • Percentage probability-of-paternity statistics that approach 100 percent so closely that jurors might treat them as conclusive may, in appropriate circumstances, be excluded under s 137 even where the underlying Paternity Index figures are admitted. The distinction turns on the assessed risk that the figures will be misused, rather than on any general rule about the admissibility of DNA statistics.

Legislation and Cases Referenced

Legislation
- Evidence Act 1995 (NSW), ss 135, 137
- Criminal Appeal Act 1912 (NSW), s 5A(2)

Key Australian Cases
- Rogers v The Queen (1994) 181 CLR 251
- Papakosmas v The Queen (1999) 196 CLR 297
- R v BD (1997) 94 A Crim R 131
- R v Edwards (1997) 94 A Crim R 204
- Pantoja (1996) 88 A Crim R 554
- Latcha (1998) 104 A Crim R 390
- Milat (1996) 87 A Crim R 446
- Toki (2000) 116 A Crim R 536
- Humphrey (1999) 103 A Crim R 434
- Mitchell (1997) 98 A Crim R 32
- Ordukaya v Hicks [2000] NSWCA 180
- R v Karger [2001] SASC 64
- Gibson v R [2001] TASSC 59
- R v Jarrett (1994) 62 SASR 443

Key English and Other Cases
- Doheny and Adams [1997] 1 Cr App R 369
- R v Denis Adams (No 2) [1998] 1 Cr App R 377