![\"Proc\u00e8s](\"https:\/\/www.forenseek.fr\/wp-content\/uploads\/2026\/10\/Proces-de-1024x578-1.png\")
<\/figure>By Monica Casey, WRAL Durham reporter<\/em><\/p> A DNA match, even one accompanied by an extremely high likelihood ratio, does not necessarily show how the DNA came to be on a surface. The evaluation may address the biological origin of a trace, which forensic science refers to as the source level [6]. It may also seek to discriminate between different activities that could explain its presence, such as direct contact or indirect transfer. This second level, the activity level, requires additional information and, where the scientific data allow, a specific evaluation [7]. A likelihood ratio calculated for the source question therefore cannot be carried over as it stands to the activity question [6].<\/p> DNA can persist after contact and be transferred indirectly through a person or an object [7]. Its presence alone therefore does not allow the deposit to be dated precisely or the action that produced it to be established. The nature of the biological material does, however, bear on this evaluation. In the Hedgepeth case, the detection of spermatozoa in intimate swabs considerably narrows the range of plausible activities, without establishing when the deposit occurred or its connection with the killing [3]. The reports presented in court thus provide information on the possible contributors to the DNA mixtures, but they should not automatically be read as demonstrating the circumstances of the murder.<\/p> This distinction is central to the guidance of the European Network of Forensic Science Institutes (ENFSI), which recommends that scientific findings be evaluated against clearly defined propositions suited to the issues in the case [5].<\/p> One of the best-known errors in the interpretation of scientific evidence is the prosecutor’s fallacy, a term coined by Thompson and Schumann in 1987 [8]. Despite its name, it is not confined to the prosecution and can be committed by anyone involved in a trial. It consists in confusing the probability of observing a result if a proposition is true with the probability that the proposition is true given the result. An everyday example shows that the two are far from equivalent. The probability that a judge in France speaks French is close to 1, whereas the probability that a French speaker is a judge is minute. Nobody would make this confusion here, but it becomes much harder to detect when it involves numerical values derived from a DNA analysis. Reading the 170 septillion as a probability of guilt, an interpretation rejected above, would be one example.<\/p> Bayesian reasoning places the contribution of the expert evidence in its proper context. The likelihood ratio measures how strongly the findings support one proposition rather than the other, without indicating on its own the probability that either is true. It acts as a multiplier applied to the view that could be taken of these propositions before the DNA analysis, on the basis of the other evidence in the case, such as the whereabouts of the person concerned or the witness statements obtained. That assessment is a matter for the fact-finder, not the expert [9]. The same likelihood ratio can therefore lead to very different conclusions depending on whether the person was identified through a body of corroborating evidence or solely through a hit on a DNA database such as the FNAEG, the French national DNA database. When the ratio reaches values comparable to those in the Hedgepeth trial, this prior assessment carries much less weight on the source question, and the debate shifts to the choice of propositions, the risk of error and the circumstances of the deposit.<\/p> The Sally Clark case in England illustrates the consequences of this confusion. Tried for the murder of her two sons, who died when only a few weeks old, she was convicted in November 1999 after the paediatrician Roy Meadow told the jury that the probability of two sudden infant deaths in a family like hers was 1 in 73 million [10]. The figure had been obtained by squaring an estimate of 1 in 8,543, as if the two deaths were independent, although genetic or environmental factors shared within a family make that assumption highly questionable [11]. Above all, it was understood, including by part of the press, as the probability that the deaths were accidental, in other words as the probability that the mother was innocent [12]. Yet the murder of two infants by their mother is also an extremely rare event, and the only relevant question was the relative probability of the two explanations in the light of the findings [11]. The convictions were quashed in January 2003, mainly because of microbiological results that had not been disclosed to the defence [10]. As early as October 2001, the Royal Statistical Society had publicly condemned the misuse of statistics in the case [11].<\/p> The growth of probabilistic genotyping makes this vigilance all the more necessary. Software such as STRmix or EuroForMix can interpret complex mixtures and low-template traces by modelling the phenomena that affect profiles, in particular allele drop-out and drop-in [13]. These systems frequently produce likelihood ratios in the billions, and sometimes values of the same order as those presented in Durham. The value obtained nevertheless depends on the propositions compared, the number of contributors assumed, the model parameters and the population data used, so two analyses of the same profile may yield different values without either necessarily being wrong.<\/p> Nor does a likelihood ratio of 170 septillion mean that the risk of error is 1 in 170 septillion. The calculation assumes that the profile was correctly generated and attributed to the right sample. The risks of contamination or sample mix-up, even if small, are generally far greater than such values and, where they are not negligible, can limit the probative value of the DNA evidence much more than the rarity of the profile [14].<\/p> Erin Ermish’s cross-examination focused on the analytical methods and on the possibility that samples stored for several years had degraded [3]. The defence also questioned the sampling choices and the results obtained from different parts of the rum bottle [4]. These exchanges show how important it is to make scientific conclusions understandable without altering their scope. A striking figure can give an impression of absolute certainty, whereas its interpretation remains limited to a specific question.<\/p> ENFSI recommends in particular that expert reports set out the propositions considered and the limitations of the evaluation [5]. The aim is to enable those who receive the report to understand the strength of the findings without confusing this scientific assessment with the judicial decision. This requirement applies directly to experts, who must explain their results in accessible language without losing the necessary nuance. It also applies to legal practitioners and investigators, whose questions can help determine whether a conclusion genuinely addresses the propositions at issue in the case.<\/p> In the Hedgepeth trial, understanding the ratio of 170 septillion therefore requires knowing which samples it relates to and which propositions were used to calculate it. The results obtained from the other exhibits must be examined for what each contributes in its own right, without being overshadowed by the most striking figures.<\/p> The scientific testimony heard in Durham is a reminder that a DNA analysis does not necessarily answer every question raised by an investigation. It can provide extremely strong support for a proposition about the origin of a trace without establishing the exact circumstances of its deposit. A likelihood ratio replaces neither the examination of the other evidence in the case nor the assessment of propositions about the facts, which is a matter for the court. For judges and lawyers, the reported value only makes sense in the light of the conditions under which it was calculated. Knowing which sample was analysed, which propositions were compared and whether the evaluation addresses the source of the DNA or the activity that led to its deposit makes it possible to determine what the expert evidence actually establishes. The clarity with which the expert sets out these elements, both in the report and in court, is an integral part of the quality of the work [5].<\/p> R\u00e9f\u00e9rences<\/strong><\/p> [1] WRAL. 170 septillion times more likely, Expert testifies about DNA in UNC student murder trial<\/em><\/a>. 15 octobre 2026. A likelihood ratio of 170 septillion was presented to the jury in Durham, North Carolina, at the trial of Miguel Salguero-Olivares, who is accused of murdering Faith Hedgepeth in 2012 [1]. Derived from the analysis of a DNA mixture recovered from the victim’s body, this figure provides extremely strong statistical support for the proposition that … Continue reading The Hedgepeth trial, when DNA statistics put the interpretation of evidence to the test<\/span> Identifying a source does not explain a contact<\/h2>
The risk of transposing the conditional<\/h2>
Expert evidence that must be understood and open to challenge<\/h2>
Interpretation and the true value of the evidence<\/h2>
[2] ABC11. Court documents reveal how investigators linked suspect in Faith Hedgepeth’s murder<\/em><\/a>. 24 janvier 2022.
[3] 9th Street Journal. Day 14 of the Faith Hedgepeth trial, Scientist testifies on DNA evidence<\/em><\/a>. 15 octobre 2026.
[4] The Daily Tar Heel. Forensic scientist connects defendant’s DNA to crime scene during Faith Hedgepeth murder trial<\/em><\/a>. Octobre 2026.
[5] European Network of Forensic Science Institutes (ENFSI). ENFSI Guideline for Evaluative Reporting in Forensic Science<\/em>. Version 3.0, 2015.
[6] Gill P., Hicks T., Butler J.M. et al. (2018). DNA commission of the International society for forensic genetics, assessing the value of forensic biological evidence, guidelines highlighting the importance of propositions, Part I, evaluation of DNA profiling comparisons given (sub-) source propositions<\/em>. Forensic Science International, Genetics, 36, 189-202.
[7] Gill P., Hicks T., Butler J.M. et al. (2020). DNA commission of the International society for forensic genetics, assessing the value of forensic biological evidence, guidelines highlighting the importance of propositions, Part II, evaluation of biological traces considering activity level propositions<\/em>. Forensic Science International, Genetics, 44, 102186.
[8] Thompson W.C., Schumann E.L. (1987). Interpretation of statistical evidence in criminal trials, the prosecutor’s fallacy and the defense attorney’s fallacy<\/em>. Law and Human Behavior, 11(3), 167-187.
[9] Aitken C., Roberts P., Jackson G. (2010). Fundamentals of probability and statistical evidence in criminal proceedings, guidance for judges, lawyers, forensic scientists and expert witnesses<\/em>. Royal Statistical Society, Practitioner Guide No 1.
[10] Court of Appeal (England and Wales). R v Clark<\/em> [2003] EWCA Crim 1020, 11 avril 2003.
[11] Royal Statistical Society (2001). Royal Statistical Society concerned by issues raised in Sally Clark case<\/em>. Communiqu\u00e9 de presse, 23 octobre 2001.
[12] Royal Statistical Society (2002). Lettre au Lord Chancellor relative \u00e0 la pr\u00e9sentation des preuves statistiques dans les proc\u00e8s p\u00e9naux, janvier 2002.
[13] Gill P., Gusm\u00e3o L., Haned H. et al. (2012). DNA commission of the International Society of Forensic Genetics, recommendations on the evaluation of STR typing results that may include drop-out and\/or drop-in using probabilistic methods<\/em>. Forensic Science International, Genetics, 6(6), 679-688.
[14] Thompson W.C., Taroni F., Aitken C.G.G. (2003). How the probability of a false positive affects the value of DNA evidence<\/em>. Journal of Forensic Sciences, 48(1), 47-54.<\/p>","protected":false},"excerpt":{"rendered":"