43,99 €
Mehr erfahren.
- Herausgeber: John Wiley & Sons
- Kategorie: Wissenschaft und neue Technologien
- Sprache: Englisch
This book explains the correct logical approach to analysis of forensic scientific evidence. The focus is on general methods of analysis applicable to all forms of evidence. It starts by explaining the general principles and then applies them to issues in DNA and other important forms of scientific evidence as examples. Like the first edition, the book analyses real legal cases and judgments rather than hypothetical examples and shows how the problems perceived in those cases would have been solved by a correct logical approach. The book is written to be understood both by forensic scientists preparing their evidence and by lawyers and judges who have to deal with it. The analysis is tied back both to basic scientific principles and to the principles of the law of evidence. This book will also be essential reading for law students taking evidence or forensic science papers and science students studying the application of their scientific specialisation to forensic questions.
Sie lesen das E-Book in den Legimi-Apps auf:
Seitenzahl: 474
Veröffentlichungsjahr: 2016
Ähnliche
Table of Contents
Cover
Title Page
Copyright
Preface to the First Edition
Who Is This Book Aimed At?
Preface to the Second Edition
Chapter 1: Introduction
1.1 Three ‘principles’
1.2 Dreyfus, Bertillon, and Poincaré
1.3 Requirements for Forensic Scientific Evidence
1.4 What We Will Cover
Chapter 2: Interpreting Scientific Evidence
2.1 Relevance and Probative Value
2.2 The Likelihood Ratio and Bayes' Theorem
2.3 Admissibility and Relevance
2.4 Case Studies
2.5 Summary
Chapter 3: The Alternative Hypothesis
3.1 Some Symbols
3.2 Which Alternative Hypothesis?
3.3 Exclusive, Exhaustive, and Multiple Hypotheses
3.4 Immigration and Paternity Cases
3.5 ‘It Was My Brother’
3.6 Traces at the Scene and Traces on the Suspect
3.7 Hypothetical Questions
3.8 Pre-Trial Conferences and Defence Notice
3.9 Case Studies
3.10 Summary
Chapter 4: What Questions Can the Expert Deal With?
4.1 The Hierarchy of Propositions
4.2 The Ultimate Issue Rule
4.3 Summary
Chapter 5: Explaining the Strength of Evidence
5.1 Explaining the Likelihood Ratio
5.2 The Weight of Evidence
5.3 Words Instead of Numbers?
5.4 Dealing with Wrongly Expressed Evidence
5.5 Case Studies
5.6 Summary
Chapter 6: The Case as a Whole
6.1 Combining Evidence
6.2 Can Combined Weak Evidence Be Stronger Than Its Components?
6.3 The Standard of Proof and the Cost of Errors
6.4 Assessing Prior Odds
6.5 The Defence Hypothesis and the Prior Odds
6.6 Case Studies
6.7 Summary
Chapter 7: Forensic Science Methodology
7.1 A General Methodology for Comparative Analysis
7.2 Assessing the Performance of an Expert or a Comparison System
7.3 System Performance Characteristics
7.4 Case Assessment and Interpretation (CAI)
7.5 Context Bias
7.6 Summary
Chapter 8: Assigning Likelihood Ratios
8.1 DNA
8.2 Glass Refractive Index
8.3 Colour Comparison
8.4 Fingerprints
8.5 Signatures
8.6 Psychological Evidence
8.7 Summary
Chapter 9: Errors of Thinking
9.1 A Brace of Lawyers' Fallacies
9.2 Double-Counting Evidence?
9.3 The Accuracy and Reliability of Scientific Evidence
9.4 Case Studies
9.5 Summary
Chapter 10: Frequentist Statistics and Database Matching
10.1 The Frequentist Statistical Approach
10.2 Databases
10.3 The Right Questions and the Wrong Questions
10.4 Summary
Chapter 11: Implications for the Legal System
11.1 What Is Expert Evidence?
11.2 Who Is an Expert?
11.3 Insanity and the Ultimate Issue Rule
11.4 Novel Forms of Scientific Evidence
11.5 Knowledge of Context
11.6 Court-Appointed Experts
11.7 Summary
Chapter 12: Conclusion
12.1 Forensic Science as a Science
12.2 Conclusions
12.3 The Fundamental Questions
Appendix
A.1 Probability, Odds, Bayes' Rule and the Weight of Evidence
A.2 Laws of Probability
A.3 The Weight of Evidence
Index
End User License Agreement
Pages
xi
xii
xiii
xv
1
2
3
4
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
29
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
147
148
149
150
151
152
153
154
155
156
157
158
159
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
179
180
181
182
183
184
185
186
187
188
189
190
191
193
194
195
196
197
Guide
Cover
Table of Contents
Preface to the Second Edition
Begin Reading
List of Illustrations
Chapter 7: Forensic Science Methodology
Figure 7.1 Graph (a) shows a distribution of probability over the outcomes of a roll of a loaded die. There are six possible outcomes and the total of the probabilities of those outcomes is therefore 1. Graph (b) shows a distribution of probability over the expected amount of rain for tomorrow. The amount of rain is a continuous variable, and the probability of any exact amount (e.g. 3.00 mm) of rain falling is zero. The curve gives the probability density, and the probability of the amount of rain being in a certain range (here: 2.8–3.2 mm) is given by the grey area under the curve (here: 0.07 or 7%). The total area under the curve is the total probability, 1.
Figure 7.2 An illustrative probability density for any particular comparison score, , (i.e. distance) for items from the same source (solid curve) or from different sources (dashed curve) and the corresponding log LRs. The vertical dashed line in each graph demonstrates how an LR can be assigned for distance in a particular case. The ratio of the two probability densities gives an LR of about 10 (i.e. ) at this point. The LR decreases as the distance increases and becomes less than 1 to the right of the density curves' intersection. In the right-hand graph, log LR is plotted against score. At the intersection, the evidence is neutral (, ).
Figure 7.3 Partly overlapping LR distributions, with (a) bad calibration and (b) improved calibration. LRs that are equally likely under both hypotheses (where the graphs intersect) should be assigned the value 1.
Figure 7.4 (a) The distributions of a system's log LRs. (b) The corresponding Tippett plot. The rates of misleading evidence can be read from the Tippett plot at from the size of the gaps at the top and bottom.
Figure 7.5 Entropy as an information theoretical metric for uncertainty. (a) As a function of the probability of and (b) as a function of log odds.
Figure 7.6 Graphs (a) and (b) show examples of empirical cross entropy (ECE) curves. ECE curves are shown for LRs equal to 1 as a reference (dotted curve), for the system's reported LRs (solid curve), and for the calibrated LRs (dashed curve). Graph (b) shows ECE curves for a system whose calibration is so bad that the ECE of the uncalibrated LRs exceeds that of the reference system for higher log odds.
Chapter 8: Assigning Likelihood Ratios
Figure 8.1 Probability densities for the refractive index of any glass fragments recovered in casework (between-source measurements, thin line) and fragments from a particular broken window
12
(within-source measurements peaking at 1.5200, thick line). The bottom graph is a very much expanded part of the central range of refractive index and the vertical axis is on a logarithmic scale for both LR and probability density. The dashed line shows the source-level likelihood ratios for different refractive indices determined from dividing the within-source and between-source probability densities. When these are equal, the , of course.
Figure 8.2 (a) In this RGB cube, the colours present in an image of two inks on paper are plotted. The empty paper background colour is represented by the large spherical blob at
P
. Ink 1 (a blue ink) and ink 2 (a black ink) form elongated clusters of blobs. (b) Spherical angles
x
and
y
capture the direction along which an ink colour varies and are ideal as a feature for ink colour comparison.
Figure 8.3 Feature vectors for 262 different blue ballpoint inks (open dots) and for 100 samples of the same blue ballpoint ink (solid dots).
Figure 8.4 (a) Probability density functions for the same-source (narrow distribution on the left) and the different-source colour differences (wide distribution). The curve models the same-source probability density as what is known as a Rayleigh distribution; (b) The log LR follows from dividing the two probability densities and taking the logarithm.
Figure 8.5 Tippett plot showing the proportion of log LRs above a certain value for same-source and different-source comparisons.
Figure 8.6 The most basic types of minutiae: ridge end and line split (or bifurcation). More types of minutiae can be defined in terms of these two. Apart from the type, several features are linked to the minutiae, such as their location and direction.
Figure 8.7 LR distributions for same-source and different-source fingerprint comparisons, for 5- and 10-minutiae configurations.
Figure 8.8 Tippett plots for 5- and 10-minutiae configurations.
Figure 8.9 ECE plots for forensic fingermark comparisons for marks with 5–12 minutiae. The remaining uncertainty indicated by the dashed line is clearly reduced when more minutiae are compared.
Figure 8.10 Comparison schemes for the generation of LRs based on comparison scores for same-source and different-source comparisons. (a) An approach where the distributions of comparison results are giving probability density functions which can be divided to give an LR. (b) An approach where an LR is obtained by calibrating the comparison scores.
Figure 8.11 Score distributions of several competing systems before and after Calibrating. Different-source comparison results are shown by the dashed line and same-source comparison results are shown by the solid line.
List of Tables
Chapter 7: Forensic Science Methodology
Table 7.1 Pre-assessment of evidence E
Table 7.2 Pre-assessment of the GSR evidence
Interpreting Evidence
Evaluating Forensic Science in the Courtroom
Second Edition
by
Bernard Robertson
Barrister, Wellington, New Zealand
G.A. Vignaux
Emeritus ProfessorVictoria University, Wellington, New Zealand
Charles E.H. Berger
Principal Scientist, Netherlands Forensic InstituteProfessor of Criminalistics, Leiden University,The Netherlands
This edition first published 2016 © 2016 by John Wiley & Sons, Ltd
Registered office: John Wiley & Sons, Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, UK
Editorial offices: 9600 Garsington Road, Oxford, OX4 2DQ, UK
The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, UK
For details of our global editorial offices, for customer services and for information about how to apply for permission to reuse the copyright material in this book please see our website at www.wiley.com/wiley-blackwell.
The right of the authors to be identified as the authors of this work has been asserted in accordance with the UK Copyright, Designs and Patents Act 1988.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by the UK Copyright, Designs and Patents Act 1988, without the prior permission of the publisher.
Designations used by companies to distinguish their products are often claimed as trademarks. All brand names and product names used in this book are trade names, service marks, trademarks or registered trademarks of their respective owners. The publisher is not associated with any product or vendor mentioned in this book.
Limit of Liability/Disclaimer of Warranty: While the publisher and author(s) have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. It is sold on the understanding that the publisher is not engaged in rendering professional services and neither the publisher nor the author shall be liable for damages arising herefrom. If professional advice or other expert assistance is required, the services of a competent professional should be sought.
Library of Congress Cataloging-in-Publication Data applied for.
ISBN: 9781118492482
A catalogue record for this book is available from the British Library.
Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books.
Cover image: © Dmitrii Kotin/gettyimages
Preface to the First Edition
This book started as part of a wider project, the examination of the applicability of logical and probabilistic reasoning to evidence generally. This has been the subject of vigorous discussion in the legal literature and is one of the main threads of the ‘New Evidence Scholarship’.
Forensic science suggested itself as a case study as there seemed to be some degree of consensus that forensic scientific evidence should be thought about in probabilistic terms, but when we surveyed the field it appeared to be a mess.
Some expert witnesses, such as fingerprint officers, make categorical statements that two impressions are from the same finger.
Some experts, such as glass experts, would only say that a sample could have come from a particular source and then gave some straightforward sounding statistics about the frequency of glass of that type.
Some types of evidence, such as DNA, seemed to involve statistical arguments of impenetrable complexity.
The law seemed in equal confusion.
There was a rule preventing giving an opinion on the ultimate issue, yet courts regularly heard witnesses talk about the probability of paternity.
A court would reject evidence in one case because it usurped the role of the jury and in another because it was not definitive and conclusive.
Courts sometimes pointed out problems with evidence that the forensic science profession did little about and sometimes ruled evidence out for reasons that had little to do with its probative value.
It also seemed to us that courts and textbook writers were keener to bandy words such as ‘reliability’ and ‘regard the evidence with caution’ than to explain what ideas lay behind these phrases.
The time had clearly come for some fundamental re-evaluation of forensic science. As we studied the matter, we realised that the application of a few basic logical principles solved the problems of description and prescription with which we were faced. That is not to say that solutions came easily; the application of these principles requires hard thinking and we cannot pretend to offer answers to all the questions. The results lead to advice about how to think about evidence of much more practical value than an admonition to ‘regard the evidence with caution’.
While preparing this book we found some forensic scientists who had been thinking along the same lines and had published papers in the scientific literature. The most prolific current writer is Dr Ian Evett of the British Home Office Forensic Science Service. Gradually, and despite opposition from within the scientific and legal fraternities, these ideas have begun to appear in legal literature and to influence the giving of evidence.
The result is that while the insights in this book will seem to some readers as revelatory as they first did to us, this book is, in fact, part of a movement throughout the forensic scientific world to re-evaluate scientific evidence and, at the same time, to encourage a greater sense of unity and professionalism amongst forensic scientists. So far as we know, however, this book is the first to be written as a single book-length work on the subject.
Who Is This Book Aimed At?
The task of convincing forensic scientists that they must rethink their interpretation of scientific evidence is one for scientists writing in scientific journals. At some point, however, the scientist has to communicate with a lawyer and perhaps with a jury. Likewise, the lawyer who understands the law and is an expert at communicating with juries has to be able to understand the scientist. It is evident that in the past there has been a sad failure of communication.
This book attempts to remedy that. It is designed to be read by both lawyers and forensic scientists so that each will better understand the other and they will be better equipped to work together to explain the evidence to the court.
We intend that the book will also be of value to academics and students. The basic logical principles we apply provide the intellectual tool-kit for re-evaluating the law relating to expert evidence and indeed to evidence generally. We believe that this is a classic example of deep theoretical thinking appropriate to university courses providing far more practical solutions to practitioners' problems than the ad hoc reasoning which has been applied to expert evidence in the past.
In completing this task we have been helped and encouraged enormously by academic colleagues and forensic scientists including, through the wonders of electronic mail, those from the United States and the United Kingdom. Particular mention must be made of Dr Evett, who has not only been of invaluable technical assistance but who chivvied us vigorously when we were slacking on the job. Valuable comments on drafts were provided in the later stages by Richard Friedman, David Kaye and Dennis Lindley and by David Wilson of John Wiley and Sons Ltd who supported us enthusiastically. We have also benefited from discussion at many conference and staff seminar presentations at our own and other universities, and from a presentational point of view we have even benefited from the outright hostility we have met on occasions. We have conducted thoroughly enjoyable (to us at any rate) Masters and Honours courses in which a number of enthusiastic students have contributed ideas and sharpened up our presentation. Some are mentioned by name at appropriate points in the book.
We have been generously supported by research grants from the Victoria University of Wellington Internal Grants Committee, which have enabled us to employ several research assistants as the project ground through its various phases. Isobel Egerton, Andrew Fairfax, Victoria Heine, Michael Sleigh and Victoria Wicks-Brown have all contributed during vacations and term time.
Certain passages are adapted versions of papers which we have published elsewhere. More than one passage is extracted from our paper ‘Expert evidence: law, practice and probability’ (1992) 12 Oxford Journal of Legal Studies 392; the passage on stylometry is adapted from ‘Stylometric Evidence’ [1994] Crim L R 645, of which Isobel Egerton was co-author; much of Chapter 7 is to be found, differently arranged, in ‘DNA Evidence: Wrong Answers or Wrong Questions’ (1995) 96 Genetica 145; the section on fingerprints is adapted from ‘The Interpretation of Fingerprints’ (1994) 3 Expert Evidence 3. The assistance we have had from the editors and publishers of those journals is also gratefully acknowledged.
This book is based on a logical argument and the state of the law in any particular jurisdiction is not important for its thesis. Nonetheless, we have endeavoured to state the law correctly where we give examples and citations and hope that the law is correct as of 1 January 1995.
Bernard RobertsonPalmerston North
G. A. (Tony) VignauxWellington1995
Preface to the Second Edition
It has been 20 years since the first edition of Interpreting Evidence. It was written in such a way that neither changes in the law nor advances in technology would invalidate the discussion. Since then, however, there have been substantial advances in the application of the principles discussed to new areas of forensic science. At the same time, there has been some confused reaction in the courts and little sign of great increase in understanding in the legal profession or academia.
The original authors had been asked by several scholars to prepare a new edition both to update the book so that it remained as comprehensive as possible and also because there was a need to get the book back into circulation. One of these, Charles Berger, a forensic scientist, not only urged the writing of a new edition but offered to participate and provide much-needed insight into recent advances. The authors were therefore delighted to recruit him to the team.
While the principles and Chapters 2 and 3 remain largely the same, a number of improvements have been made:
We have removed reference to obsolete methods such as blood-grouping, now replaced by DNA testing and to methods such as stylometry, which has been effectively dismissed as being of any value;
Chapters have been reordered, so that the whole logical method is set out before we discuss problems caused by the use of other methods;
There has been a general rewriting to improve style and presentation and to take into account various detailed criticisms we have received; and
Chapters 7 and 8 are largely new and, in particular, take account of advances in the application of Bayesian analysis to new areas of evidence.
We have benefited from feedback about the first edition from forensic scientists and lawyers around the world. We are especially grateful for comments and help while preparing this edition from Colin Aitken, Niko Brümmer, John Buckleton, Christophe Champod, Ian Evett, Tacha Hicks Champod, Daniel Ramos, and Marianne Vignaux, none of whom, of course, are responsible for the views expressed or any errors made.
Bernard RobertsonWellington
G. A. (Tony) VignauxWellington
Charles E. H. BergerThe Hague1 June 2016
Chapter 1Introduction
Forensic scientific evidence can help us to establish:
that a particular person was at a given place at a given time;
that a particular person carried out an activity, such as signing a cheque or breaking a window;
that something was done with a particular instrument, for example, a door was forced with a particular tool, a shot fired from a particular weapon, or a call made from a particular telephone;
a relationship between two people, for example, in paternity disputes and incest or immigration cases.
There is a whole range of techniques used for forensic purposes, and new methods are continually being added to the arsenal of the forensic scientist. Our purpose is not to discuss the technical details of these methods, which rapidly become dated. We propose to concentrate on how such evidence should be interpreted and incorporated into the court process.1
1.1 Three ‘principles’
Traditionally, several ideas have been proposed as principles for forensic science:
Locard's ‘Principle’: A perpetrator will either leave marks or traces on the crime scene, or carry traces from the crime scene. This is often misquoted as ‘every contact leaves a trace’ but Locard never actually claimed this.
Edmond Locard (1877–1966) was a French forensic scientist. He proposed that we should always consider whether traces of the victim or crime scene can be found on the accused and whether traces of the accused can be found on the crime scene or victim. After an assault, for example, we might find skin and blood under a deceased's fingernails and infer that they come from the attacker. We might arrest a suspect on the basis of other evidence and find, on him or his clothing, fibres which might come from the deceased's clothes, blood which might come from the deceased or soil and plant material which might come from the scene.
‘Principle’ of individuality: Two objects may be indistinguishable but no two objects are identical.
2
The combination of these two ideas together might seem to have enormous potential value to the forensic scientist. If every contact provides ample opportunity for the transfer of traces, and every trace is different that seems to be cause for optimism. However, if no two objects are identical, then, for example, no two fingerprint impressions will be identical even if they are taken from the same finger; no two samples of handwriting by the same author will be identical. The question is whether two marks have the same source, and how much our observations help us in answering that question.
We describe these two statements as proposed principles rather than laws because neither meets the standard definition of a law of science. The philosopher Karl R. Popper (1902–1994) said that for a law to be regarded as scientific it must be potentially falsifiable, that is, it must be possible, at least in theory, to design an experiment which would disprove it.3
It seems to be impossible to design an experiment to refute the first of these principles. If an experiment fails to find an impression after two objects have been in contact, it may be that all that is revealed is the limitations of the detection process. The proposed principle that no two objects are identical does not require proof, since two objects that would be identical in every way would – by definition – be one object. Unfortunately, it does not follow from the uniqueness of every object that we can correctly point out its unique source.
Individualisation ‘Principle’: If enough similarities are seen between two objects to exclude the possibility of coincidence, then those objects must have come from the same source.
This ‘principle’ has a long history in forensic science, as can be seen from the following quotes that span the 20th century:
The principles which underlie all proof by comparison of handwritings are very simple, and, when distinctly enunciated, appear to be self-evident. To prove that two documents were written by the same hand, coincidences must be shown to exist in them which cannot be accidental.4
When any two items have characteristics in common of such number and significance as to preclude their simultaneous occurrence by chance, and there are no inexplicable differences, then it may be concluded that they are the same, or from the same source.5
…we look for unique characteristics in the items under examination. If we find a sufficient number of characteristics to preclude the possibility or probability of their having occurred by coincidence in two different objects, we are able to form a conclusion of individualization. It's as simple as that.6
This popular so-called principle, while simple, is fraught with problems. The possibility of a coincidence can never be completely excluded, which precludes categorical statements of individualisation. There is no general criterion possible for the number of coincidences needed to decide an individualisation; whatever level is chosen is purely arbitrary. How certain we would want to be for a decision would depend on the gravity of the crime involved (e.g. capital murder versus shoplifting). How certain we could be would also depend on other evidence and information in the case. Clearly, such issues and decisions are not up to the forensic scientist but rather the trier of fact. The role of the forensic scientist is not to decide the issue, but to describe what the evidence is worth. This ‘principle’ should therefore not be used.
1.2 Dreyfus, Bertillon, and Poincaré
In 1894, Alfred Dreyfus (1859–1935), an officer in the French army, was charged with treason in what was to become one of the most famous criminal trials in history. The charges were espionage and passing information to Germany. The espionage had definitely taken place and one of the central items of evidence was the comparison of the handwriting in an incriminating note with Dreyfus's own handwriting. A prominent witness for the prosecution was Alphonse Bertillon (1853–1914).
Bertillon was a Paris police officer who rose to found a police laboratory for the identification of criminals. He was well known for proposing a system of anthropometry, which became known as Bertillonage. Anthropometry simply means the measurement of humans. Bertillonage required taking a photograph and recording a series of measurements of bone features which were known not to change after adolescence. Later, fingerprints were added to the features recorded. The basis of the system was that it would be unlikely that any two people would have the same measurements over the whole range of features.
Bertillonage suffered from a number of problems. The method was slow and expensive and was far from error free. The officers taking the measurements had to be specially trained; this involved more expense, and even then, at the levels of accuracy called for, no two would take the same measurements from the same series of features. Nor could the system be applied to juveniles.
The purpose of the system was to determine whether or not a person had the same measurements as a person who had earlier been arrested. This can be very useful, for example, when someone is arrested on suspicion of failing to attend court or when a person being sentenced denies that previous convictions relate to him. However, Bertillonage could not help investigators by providing evidence that a particular person had been, for example, at the scene of a crime.
Although fingerprints were later taken as one of the Bertillonage measurements and Bertillon himself solved a crime using fingerprints in 1902, there was no formal classification system for them. Once such systems were developed (by Galton and Henry in England and India, and Vucetich in Argentina) it was possible to quickly exclude the majority of the fingerprint collection (i.e. the other classes) on each search. Fingerprints became a far quicker and simpler method of identification than anthropometry. In the first full year of operation by the London Metropolitan Police, fingerprints identified 3 times as many persons as anthropometry and, 2 years later, 10 times as many. Not only were fingerprints far simpler and cheaper to obtain and record but they could also help investigators identify the perpetrators of crimes. Bertillonage was dropped.
Bertillon gave evidence in the Dreyfus case as a handwriting expert and claimed that Dreyfus had written the incriminating document. His evidence referred to certain similarities and multiplied together the probabilities of each of the similarities occurring by chance to arrive at a very low probability of them occurring together by chance. His evidence was subjected to devastating critique by a number of people including Poincaré, an eminent mathematician.7 Poincaré made three important points about Bertillon's evidence. The first was that Bertillon had simply erred in that the figure he produced was the probability of getting the four similarities amongst four examined characteristics. There were far more characteristics examined, and so the chances of finding four similarities were actually much greater than Bertillon's figure. The second point Poincaré made was that events that have actually occurred might be seen beforehand as highly improbable. The example he gave was the drawing of a particular number or set of numbers in a lottery. The probability that any particular set of numbers will be drawn is extremely low. Once it has been drawn, however, that low probability does not mean that the draw has been dishonest.
Most importantly of all, Poincaré discussed what is called the inverse probability problem, the difference between calculating in advance the probability of an effect and calculating after the event the most probable cause of an effect:
As an example of probability of effects, we usually choose an urn containing 90 white balls and 10 black balls; if we randomly draw a ball from this urn, what is the probability for this ball to be black; it is evidently 1/10.
The problems of probability of causes are far more complicated, but far more interesting.
Let us suppose for example two urns of identical exterior; we know that the first contains 90 white balls and 10 black balls, and the second contains 90 black balls and 10 white balls. We draw arbitrarily a ball from one of the urns, without knowing from which, and we observe that it is white. What is the probability that it came from the first urn?
In this new problem, the effect is known, we observed that the ball drawn was white; but the cause is unknown, we do not know from which urn we made the draw.
The problem that we are concerned with here is of the same nature: the effect is known, the indicated coincidences on the document, and it is the cause (forgery or natural writing) that is to be determined.8
Poincaré identifies a crucial point for forensic science and, indeed, all reasoning about evidence in court. This is a central theme of this book and will be explained in the following chapters. Courts are not concerned with the probability that some observation would be made. They are concerned with what can be inferred from the fact that the observation has been made. The question for the court then is what inferences can be drawn as to the guilt of the accused.
Poincaré went on to make the point that single items of evidence enable us to alter our assessment of the probability of an event but they cannot determine the probability of an event on their own9
:
To be able to calculate, from an observed event, the probability of a cause, we need several data:
we need to know what was à priori, before the event, the probability of this cause.
we then need to know for each possible cause, the probability of the observed event.
1.3 Requirements for Forensic Scientific Evidence
Photographs are still used to help identify criminals and are recorded with the details of their convictions. They have a number of advantages: they can be transmitted and reproduced easily and can enable people to be recognised at a distance. In most cases, a photograph will settle a question of identity. Where this is seriously challenged, however, a photograph is of questionable value, particularly if much time has passed since it was taken.10 Similarly, physical descriptions can be broadcast on police radios and even the most rudimentary description will eliminate a large proportion of the population. However, when identity is seriously challenged, descriptions and even eyewitness identification are of questionable value, perhaps because the question has become whether the perpetrator was the accused or someone else of similar appearance.
The limitations of Bertillonage prompt us to consider the features of an ideal scientific system for identifying people. These would include:
that it uses features that are highly variable between individuals;
that those features do not change or change little over time;
that those features are unambiguous so that two experts would describe the same feature the same way;
that those features can be transferred to traces at a crime scene; and
that it is reasonably simple and cheap to operate.
Inevitably, few systems will satisfy all these requirements and in particular there may be a trade-off between the last requirement and the others. Each of the systems that we examine later will satisfy some of these requirements but not all.
If we can establish features that help distinguish between individuals or groups, it becomes useful to maintain a database of observed features of known individuals. Large databases of DNA profiles have now been established, as happened with fingerprint collections over the last century. In investigations, such databases allow police to search for individuals that could have left a crime scene trace.
If a suspect has been identified and the observed features of this known person are, for example, similar to those of the traces from the crime scene, we need to evaluate what those observed similarities are worth. If the suspect had nothing to do with the crime, what would be the probability of finding those similarities? That probability can be assessed with the help of databases of features that are representative of some population. It does not require the contributors to the database to be known.
As we have seen, evidence should not be expected to give certainty. This does not make evidence ‘unreliable’ or inadmissible. Lawyers often tend to ignore evidence that does not claim to provide certainty, but by doing so, they lose relevant and probative evidence.11 Uncertainty is inherent to the limited amount of information that can be extracted from traces, which may be minute, old and contaminated. Poincaré did not tell us to simply discard such evidence, but to assess the probability of the observed effects for the possible causes.
It follows that a scientific witness will not, in principle, be able to say that two samples came from the same person. The evidence can only lead to an assessment of the probabilities that the evidence would be found if the prosecution case was true and if the defence case was true. The legal system has not been successful in dealing with this kind of evidence, and our purpose is to explain how such evidence should be given and integrated into the case.
1.3.1 Reliability
Rather than think rigorously about these problems, the legal system has been prone to ask questions such as ‘how reliable is this evidence?’. This question is difficult to answer since ‘reliable’ appears to have no fixed meaning. We discuss its different possible meanings and the consequences of each in Chapter 5.
1.4 What We Will Cover
We adopt a structure different from that of most other books on forensic scientific evidence. Those intended for scientists are usually built round the different techniques available. Those for lawyers are often structured round rules such as the Basis Rule, the Field of Expertise Rule, the Qualifications Rule and the Ultimate Issue Rule. That such a structure is unsatisfactory is shown by the extent to which these ‘rules’ are intertwined. Courts sometimes refer to one, sometimes to another. Cases that are decided on the basis of one rule are often explicable by reference to another. In this book, we
explain the fundamentals of logical reasoning about evidence and show how these principles apply to all forms of evidence in court cases. These principles explain how individual items of evidence should be thought about (
Chapters 2
and
3
);
consider what kinds of questions forensic scientific evidence can answer (
Chapter 4
);
discuss how the strength of evidence can be explained (
Chapter 5
);
show how to combine evidence with the case as a whole (
Chapter 6
);
look in more detail at how forensic scientists evaluate evidence and the methods they use (
Chapter 7
);
discuss the analysis of some specific types of scientific evidence to show how the principles apply to particular problems (
Chapter 8
);
discuss various misleading and fallacious styles of presentation of evidence, some of which are still in common use (
Chapters 9
and
10
);
examine some of the more traditional legal questions in the light of our analysis and make recommendations for reform (
Chapter 11
).
1
We use evidence here in the sense of observations (that are certain) that influence our degree of belief in the truth of things we cannot be certain about, such as those listed here. We do not limit evidence to information that has been designated as such by a court.
2
Wittgenstein: ‘Roughly speaking, to say of two things that they are identical is nonsense, and to say of one thing that it is identical with itself is to say nothing at all’ (
Tractatus
, 5.5303).
3
Popper KR,
Conjectures and Refutations: The Growth of Scientific Knowledge
, 5th ed (Routledge and Kegan Paul, London, 1989).
4
Osborn AS,
Questioned Documents
, (Rochester, New York, 1910), p. 211.
5
Huber RA, Expert witnesses, (1959), 2,
Criminal Law Quarterly
, 276–296.
6
Tuthill H,
Individualization: Principles and Procedures in Criminalistics
(Lightning Powder Company, Salem, Oregon, 1994) p. 27.
7
Taroni F, Champod C, and Margot P, Forerunners of Bayesianism in early forensic science, (1998), 38,
Jurimetrics
, 183–200.
8
Poincaré H, Darboux G, Appell P (1908) Rapport de MM. les experts Darboux, Appell et Poincaré, In
Affaire Dreyfus; La révision du procès de Rennes; Enquête de la chambre criminelle de la Cour De Cassation
vol. 3, p. 502. Paris: Ligue française pour la défense des droits de l'homme et du citoyen.
9
Poincaré pointed out that in the example he gave, before we draw the ball, we intuitively assess the probability that the urn chosen was the one with 90 white balls and 10 black balls as 0.5, or odds of 1 to 1. The problem would be changed if there were 11 urns to choose from and we knew that 10 of them had 90 white balls and only one had 90 black balls.
10
At the trial in Israel of the alleged Nazi concentration camp guard Demjanjuk various techniques were used to try to show that the defendant in 1989 was the person in a photograph on a 50-year-old identity card. Conversely, methods of altering photographs, either to implicate or exculpate a person, are now readily available.
11
As the New Zealand Court of Appeal said in respect of fingerprints in
R
v
Buisson
[1990] 2 NZLR 542, 548.
Chapter 2Interpreting Scientific Evidence
Expert scientific evidence usually involves the forensic scientist making an observation on some aspect of the case and, based on knowledge and past experience, reporting inferences to the court. For example, the scientist may compare a DNA profile from blood found at the scene with that of the accused and find them to be the same. It is the observations made which constitute the evidence and not the material examined.1 Our task is to see what inferences can and cannot legitimately be drawn from such observations. There is a simple and logical solution to these questions that deals with many of the difficulties courts have perceived with expert evidence.
In later chapters we discuss how the expert should report such inferences and how the court should interpret them, what weight the court should give them, and how they should be combined with other evidence to help the court to decide the issues before it. In this chapter we consider how to evaluate a single item of evidence that is offered to support a party's case.
2.1 Relevance and Probative Value
The first requirement of any piece of evidence tendered in court is that it must be relevant. In order to be considered, an item of evidence must be one that might rationally affect the decision. If it cannot, then it is surely worthless. A typical definition of relevance which reflects that used in all common law systems is found in Rule 401 of the United States Federal Rules of Evidence:
Evidence is relevant if:
it has any tendency to make a fact more or less probable than it would be without the evidence; and
the fact is of consequence in determining the action.
2
Rather than the term ‘fact’ in this book we will use the words ‘proposition’ or ‘hypothesis’ for a fact that needs to be proved in either a civil or criminal case. If an item of evidence does not cause us to change our probability assignment for the hypothesis, then we would not normally describe it as evidence either for or against it. Thus, an item of evidence that is worth considering is one that might cause us to increase or decrease our probability for some proposition which is of consequence in determining the case. ‘Good evidence’ is evidence which has a substantial effect on our probability. What is it about an item of evidence which enables us to change our probability assignment? To answer this question, we will consider some extreme cases.
2.1.1 Ideal and Useless Evidence
An ideal piece of evidence would be something that always occurs when what we are trying to prove is true and never occurs otherwise. In real life, evidence this good is almost impossible to find. Suppose a blind person needed to determine whether it was cloudy. Rain is not ideal evidence because absence of rain does not imply absence of cloud. If it is raining we can be sure there are clouds about but there may also be clouds if it is not raining.
At the other end of the scale, some observations are certainly useless as evidence. Imagine we are interviewing a child it is suspected has been sexually abused. We seek factors which indicate abuse (or otherwise). If we looked at ‘all data’ without discrimination, we might note that the child is breathing at the time of the interview. After many such interviews we conclude that all children who allege abuse are breathing at the time of the interview. We know, however, that this is useless as evidence of abuse simply because all other children breathe as well. In other words, the child is equally likely to be breathing whether the child has been abused or not. Despite being a characteristic shared by all abused children, breathing is not any sort of evidence for abuse. It does not discriminate between abuse and non-abuse.
Likewise, a large proportion of the DNA in our cells is indistinguishable in all human beings. This is why we nearly all have two eyes, two legs, etc. The presence of this part of the DNA in traces taken from the scene of a crime and in DNA from a suspect is useless as evidence of identification. Since everyone shares such characteristics, the finding is equally likely whether or not it was the accused who left the trace.3 DNA gets its immense discriminating power from those tiny parts of it which differ from person to person.
2.1.2 Typical Evidence
Ideal evidence is seldom found. Even if the evidence always occurs when the hypothesis is true, it may also occur when it is not (e.g. clouds as evidence for rain). Alternatively, when the hypothesis is true, the evidence may not invariably occur (e.g. rain as evidence for clouds). Thus, in the real world, evidence is something that is more or less likely to occur when what we are trying to prove is true, than when it is not. Good or strong evidence would be something that is much more likely to occur when what we are trying to prove is true, than when it is not.
For example, during a career of interviewing, a doctor might observe that a high proportion of abused children display signs of stress such as nail-biting. This will be evidence for abuse if and only if abused children are more likely to bite their nails than non-abused children. If it turned out that abused and non-abused children are equally likely to bite their nails, then this observation is useless as evidence of abuse. If abused children are much more likely to bite their nails than non-abused, then we have strong evidence of abuse. Suppose 80% of abused children bite their nails but only 10% of other children do so. Nail-biting would then be eight times more likely in an abused child than in some non-abused child. If, on the other hand, 90% of non-abused children bite their nails, then nail-biting would be evidence pointing away from abuse.
There are two points to notice: first, the strength (or probative value) of the evidence depends not only on how many abused children bite their nails but also on how many non-abused children do so; secondly, and most importantly, all we know at this stage is the probability of the evidence in each case. We do not know how likely it is that the child has been abused.
The probative value of any evidence can be evaluated in the same way. A scientific test result is good evidence for a particular hypothesis if it is much more likely to occur if the hypothesis is true, than if it is false. We will know this only if we have seen the result of the test both on a number of occasions when the hypothesis is true, but also when its negation is true. Even when we have evaluated the probability of the result under both hypotheses, we still only know the strength of the evidence in favour of a hypothesis and not the probability that the hypothesis is true.
2.1.3 An Aside on Probability and Odds
This section breaks the flow, but the simple ideas of probability and odds are so fundamental to our argument that it is important that the reader is reminded of them. Even readers already familiar with probability should read this section as some widespread misconceptions exist. There is fuller coverage in the Appendix.
Probability is a rational measure of one's degree of belief in the truth of a proposition based on information. The hypothesis, proposition, or premise is itself either true or false. For example, the proposition ‘The driver is over the drinking limit’ is either true or false but we may not be sure of whether it is true. Our assigned probability expresses our degree of belief about the truth of the proposition.
All probabilities depend on the assumptions and information used in assigning them. There are no ‘real probabilities’ that we are attempting to estimate. We would assign a different probability for the proposition ‘the driver is over the drinking limit’ if we had the result of a breath test or we had observed erratic driving, than we would without that information. All the information that is used to assign a probability is known as the condition for the probability. All probabilities are conditional on the evidence used and background knowledge.
Evidence is also described in the form of propositions, but in this case there is no uncertainty about these statements. Thus ‘the light showed red’ is evidence for the hypothesis that ‘the person is over the limit’. We would assign a higher probability that the person was over the limit than if the light showed green. That, again, would be different if we either had no breath test result or had observed erratic driving.
Probabilities take values between 0 and l.4 A probability of 0 means that (taking into account the evidence listed in the condition) the proposition cannot be true and we are completely convinced it is false. A probability of 1 means that, given the condition, the proposition must be true. Thus, my probability for the proposition ‘the sun will rise tomorrow’, given my knowledge of the working of the solar system, is 1.5
Most probabilities fall between these limits. A probability of 0.5 for a proposition means that we are equally sure (or equally unsure) that the proposition is true and that its negation is true.
Probabilities can be expressed as a percentage. A probability of 0.5 could be described as a probability of 50%, one of 0.3 as a probability of 30%. We will sometimes use percentages in this book.
We can also express probabilities in the form of odds. Many people are familiar with odds, if only from betting. They also recognise that they are a description of uncertainty, like probability. However, not everyone realises that they are only another way of representing probability and one can go from one form to the other quite easily.
To get the odds from the probability of a proposition, you calculate the ratio of its probability to the probability of its negation and simplify as much as possible. Thus, a probability of 0.3 has equivalent odds of:
This could also be written as odds of 3 to 7 (in favour of the proposition).
Odds corresponding to a probability of 0.5 are:
These odds could alternatively be described as 1 to 1 or evens.
Odds of less than evens are sometimes reversed and described as ‘odds against’ the proposition. Odds of 3 to 7 in favour of a proposition might, instead, be described as odds of 7 to 3 against.6
To return from odds to probability, one calculates the fraction:
Thus, odds of 3 to 7 (or 3/7) would be the same as:
Even odds (1 to 1) correspond to a probability of .
2.1.4 A Breath-Testing Device
Imagine a primitive breath-testing device to be used at the roadside for checking whether a driver is over or under the legal alcohol limit. It is supposed to show a red light if the driver is over the limit, a green light if he is under. Suppose we can adjust the setting of the device that determines above which alcohol concentration the red light will show. We must guard against two types of error: a false positive and a false negative. A false positive – a red light shows when the person is actually under the limit – leads to someone being wrongly arrested and inconvenienced by being required to be tested by the more accurate device at the police station. A false negative – a green light shows when the person is really over the limit – leads to a drunk driver remaining on the road.
Unfortunately, it is inevitable that reducing the rate of one of these errors by adjusting the settings of the device will increase the rate of the other. There is presumably a reason for each false reading and technical improvement would reduce the errors but, bearing in mind that we are trying to produce a cheap and robust device, we may not be able to afford to investigate all the causes. It may be impossible in practice to eliminate errors altogether but we have a choice of which errors to make. For example, if we decrease the probability of a false negative (‘miss’), we will automatically increase the probability of a false positive (‘false alarm’). Which error is the more serious is a question for society, but let us suppose some figures purely for the sake of example.
Before using it, we use the testing device with samples of air from people with a measured alcohol content. Many such samples are tested. Suppose, as a numerical example, we test 1000 samples from people with an alcohol concentration marginally below the legal limit and 1000 samples from people that are marginally above. We adjust the device so that, of the samples from people over the limit, 950 read red and 50 read green, and, of the samples from people below the limit, 995 read green and 5 read red.7
From the data from the calibration tests, we can see that:
if the sample is from someone marginally over the limit there is a 95% probability (950/1000) that the device will indicate red and a 5% probability (50/1000) that it will indicate green – the odds are 19 to 1 that it will indicate red if the person is over the limit.
8
if the sample is from someone marginally under the limit there is a 0.5% probability that the device will indicate red and a 99.5% probability that it will indicate green – the odds are 199 to 1 that it will indicate green if the person is under the limit.
9
We can see that a red light on the breath test is good evidence for the proposition that ‘the person is over the limit’. If a person is over the limit, there is a 95% probability of a red light; if a person is under the limit there is only a 0.5% probability of a red light. Thus, a red light is 190 times more likely to occur if the subject is over the limit than if under .
In contrast, a green light is good evidence against the proposition that ‘the person is over the limit’. If a person is over the limit, there is a 5% probability of a green light; if under the limit there is a 99.5% probability of a green light. A green light is about 19.9 times less likely to occur if the person is over the limit than if under . Therefore, depending on the light shown, the device can provide good evidence either for or against the proposition that ‘the person is over the limit’. It discriminates well between the two cases.
However, let us re-emphasise that this is telling us only the probative value of the evidence and not the probability that the person is over (or under) the limit. The breath-test result is a good piece of evidence, which means that it should cause us to change our assignment for the probability that the person is over the limit. But how exactly is this to be done?
