20 April, 2012

Guest lecture from Bayes and Law network member Dr Geoffrey Stewart Morrison at Queen Mary, University of London

Dr Geoffrey Stewart Morrison is Director of the Forensic Voice Comparison Laboratory, School of Electrical Engineering & Telecommunications, University of New South Wales. For his website click here.

Time: 11am-12pm

Date: Friday 20 April, 2012

Place: Rm 3.02, Bancroft Teaching Rooms, QMUL, Mile End Road, London, E1 4NS

Title: What did Bain really say? A forensic analysis of the disputed utterance based on data, acoustic analysis, statistical models, calculation of likelihood ratios, and testing of validity

The David Bain legal case is (in)famous in New Zealand. In 1995 Bain was convicted of murdering his family. He maintained his innocence and a retrial was eventually held in 2009. He was found not guilty. Prior to the retrial, a police officer thought he heard the words “I shot the prick” in the telephone call that Bain made to the emergency services shortly after the murders. The defence contended that these were not the words spoken. An alternative hypothesis which emerged was that Bain had said “I can’t breathe”. 

A number of forensic experts offered their opinion, but all except one focussed on what they heard. Rose, in contrast, pointed out that what anyone hears is irrelevant, what matters is what Bain said, and the latter should be assessed via acoustic analysis not auditory perception. Further, the proper way for a forensic scientist to evaluate the strength of evidence is via a likelihood ratio: What is the probability of getting the acoustic properties of the disputed utterance if Bain had said “I shot the prick” versus what is the probability of getting the acoustic properties of the disputed utterance if he had said “I can’t breathe”. 

The key segment of the recording which Rose focussed on was the part immediately following “I”. Were the acoustics of this segment more likely if what had been said was a postalveolar fricative (a realisation of the first phoneme in “shot”) or were they more likely if what had been said was a palatal fricative (a realisation of the first phoneme in “can’t”)? Rose concluded that the likelihood of the latter was greater than that of the former, but at the time he did not go as far as building a database, and performing acoustic and statistical analyses to calculate a numeric likelihood ratio, and to test the validity of such procedures. 

I have now conducted a preliminary version of such an analysis and in this presentation provide a step-by-step exposition of how to calculate and test the validity of a numeric likelihood ratio for the disputed utterance in the Bain case.


Dr Geoffrey Stewart Morrison received his PhD from the University of Alberta in 2006. His dissertation focussed on statistical modelling of speech data. He began working on forensic voice comparison in 2007 as a Research Associate with Dr Philip Rose at the Australian National University. Since 2010 he has been a Senior Researcher and the Director of the Forensic Voice Comparison Laboratory, School of Electrical Engineering & Telecommunications, University of New South Wales. He is also an Invited Lecturer in the Judicial Phonetics Specialisation, Master in Phonetics and Phonology Programme, Spanish National Research Council / Menéndez Pelayo International University, an Adjunct Associate Professor in the Department of Linguistics, University of Alberta, and is Chair of the Forensic Acoustics Subcommittee of the Acoustical Society of America. In collaboration with the Australian Federal Police, New South Wales Police, Queensland Police, National Institute of Forensic Sciences, Australasian Speech Science and Technology Association, and the Guardia Civil, he is the lead researcher on an Australian Research Council Linkage Project on making demonstrably valid and reliable forensic voice comparison a practical everyday reality in Australia. He has published on forensic voice comparison and on the evaluation of forensic evidence in leading journals in acoustics, speech processing, forensic science, and law. (Any opinions expressed by Dr Morrison are his own and do not necessarily represent the policies or opinions of any of the above-mentioned organisations or individuals.)

21 March 2012

Research Seminar on the Bayes and Law network at the Jill Dando Institute for the Forensic Sciences

Professor Norman Fenton will be giving a seminar on the work of the Bayes and Law Seminar  at UCL’s Jill Dando Institute for The Forensic Sciences. For more information or to register to attend, click here

Title: The Bayes and Law Network

Time: 4pm, 21 March, 2012

Place: Jill Dando Institute, UCL, 35 Tavistock Square, London, WC1H 9EZ

29 Feb 2012

Distinguished Lecturer Seminar: Professor David Balding

As part of the department of Electronic Engineering and Computer Science's Distinguished Lecturer Seminar series, network member Professor David Balding will speak on DNA profiling at Queen Mary, University of London on 29 February, 2012

Time: 3pm, 29 February, 2012

Place: Room BR 3.02, Dept of Electronic Engineering and Computer Science, QMUL


David Balding, Institute of Genetics, University College London

Recently, forensic DNA profiling has been used with far smaller volumes of DNA than was previously thought possible.  This "low copy number" profiling enables DNA to be recovered from the slightest traces left by touch or even merely breath, but brings with it serious interpretation problems that courts have not yet adequately solved.  These problems have contributed to important cases collapsing or convictions being overturned, for example in R v Hoey in Northern ireland, and the case of Knox and Sollecito in Italy.  The most important challenge to interpretation arises when either or both of "dropout" and "dropin" create discordances between the crime scene DNA profile and that expected under the prosecution allegation. Stochastic artefacts affecting the peak heights read from the electropherogram (epg) are also problematic, in addition to the effects of masking from the profile of a known contributor.  I will outline a framework for assessing such evidence, based on likelihoods that involve dropout and masking by stutter and other artefacts, and discuss possible options for modelling dropin.  I will apply it to casework examples and reveal serious deficiencies in some reported analyses.

David Balding was educated in his native Australia before coming to the UK to study for a PhD in mathematics at the University of Oxford. He then held a junior academic post at Oxford for a year before moving successively to Queen Mary London, the University of Reading, and Imperial College London. Since October 2009, he has been Professor of Statistical Genetics at the Institute of Genetics, University College London. David researches a broad range of mathematical and statistical problems in genetics – evolutionary, population and medical. He has also developed widely-adopted methods of analysis for the interpretation of forensic DNA profiles, summarized in his monograph Weight-of-Evidence for Forensic DNA Profiles (Wiley, 2005). On occasions, he acts as an expert witness for cases involving complex DNA profile evidence, and he has until recently been a member of the External Advisory Group of the UK Forensic Science Service.  He is a Fellow of both the Royal Statistical Society and the Society of Biology. 

Dec 2011

Where next for Bayes? Can we agree on the role of Bayesian reasoning in Criminal Justice?

Brief report on the first meeting of the Bayes and Law network, 2pm-5pm, Wednesday 7 December, 2011, Queen Mary, University of London.

Almost 40 mathematicians, computer scientists, legal scholars and practitioners, forensic scientists and civil servants attended the meeting at Queen Mary, University of London chaired by Professor Norman Fenton. Hosted by the School of Law's Criminal Justice Centre, the meeting was welcomed by Professor Fenton and Amber Marks of QMUL and addressed by the Forensic Science Regulator, Andrew Rennison, Strategic Policy Advisor to the Crown Prosecution Service, Karen Squibb-William, and barrister Richard Thomas (who was Counsel in the R v T case).

Click here to listen to Norman Fenton's welcome

Click here to listen to Amber Marks' welcome

Click here to listen to Andrew Rennison's talk

Click here to listen to Karen Squibb-Williams' talk

Click here to listen to Richard Thomas' talk

The second half of the meeting consisted of a discussion of the network’s research goals and issues arising from the speakers’ talks. Research ideas put forward included creating databases of different types of forensic evidence to provide a statistical basis for Bayesian calculations, a comparative law survey of the rules on how statistical evidence is analysed and presented in other countries and jurisdictions, and empirical experiments in which different experts would be asked to analyse the same case evidence, first on their own and then together, to compare and contrast their results.

Karen Squibb Williams answered many questions on the practicalities of different research ideas and where Bayesian reasoning might be most successfully deployed in the legal system. She saw great potential for its use at the case management stage, in which the CPS decides which cases to bring to court, and urged researchers to target their ideas there. It was further remarked by several speakers that an inquisitorial rather than adversarial approach between experts evaluating evidence at this pre-trial stage would be useful. 

Another key theme to come out of both the talks and the open discussion was the need for transparency by forensic scientists using Bayes to evaluate evidence. Rather than approach the question of when it is appropriate to use Bayesian reasoning by defining which types of forensic evidence were suitable for Bayesian analysis, the general consensus was that any approach to evidence evaluation could be used, as long as it was done transparently and explained clearly in court. There was some debate over the usefulness of verbal scales for presenting expert opinions in court, with some people finding them useful and others thinking they were less informative than a numerical scale and had too great a potential to be misinterpreted by the jury.

Discussing what could be achieved over the next 6 months, it was suggested that creating a couple of examples of how Bayesian analysis worked in practice, by retrospectively analysing the evidence in a couple of appropriate recent trials would be helpful. Work on how best to present Bayesian arguments to lay people was also put forward. Lastly, Norman Fenton said that his recent experience as an expert witness led him to believe that in practice there was a widespread perception among trial lawyers that R v T meant it was too risky to admit Bayesian evidence in court. He therefore suggested that the network draft and publish a public statement expressing its consensus that this was incorrect.

Rosie Wild,
6 Dec 2011, 03:45
Rosie Wild,
22 Dec 2011, 04:16