ÁñÁ«ÊÓƵ

Did Keats get his inspiration from smoking opium? Very likely. Did Van Gogh cut his ear off because he was being driven mad by Meniere's disease? Maybe. Was Napoleon poisoned with arsenic? Possibly - but probably not fatally so. Solving historical puzzles is a fascinating pastime, especially if we can call on modern forensic techniques. The analysis of Keats's and Napoleon's hair shows that the former was using drugs and the latter absorbing more arsenic than normal. But before we can arrive at a verdict based on such information, we need to know how much value to place on such evidence. The same should apply in a court of law.

Here are two books that complement each other very well: the first, Science and the Detective, tells how the forensic evidence is collected and what its limitations are; the second, Interpreting Evidence, shows how it should be evaluated in court, at least in theory.

Science and the Detective is by Brian Kaye of Laurentian University, Canada, and is designed for law students, whom he believes should be exposed to a course on forensic science, the better to appreciate its value and to dispel their fears that they lack the scientific training to question it. Here is a book in which they will discover what expert witnesses are really talking about, and I am sure it will find its way on to every budding lawyer's bookshelf. I can also recommend it to chemists, biologists, medics, sociologists and even historians. The text is clearly written, copiously illustrated with photographs and scientific charts, and there is a good index. This is an ideal source book for anecdotes to enliven lectures; its science is sound and the material has been well researched and referenced.

Kaye leans over backwards to help the non-scientist by explaining every unusual word he uses, and this is where the book's weakness lies. He admits his hobby is writing dictionaries, and he is fascinated by etymology. When he uses a scientific term he highlights it in bold, as in a dictionary, and explains it in the body of the text. Footnotes or a glossary would have been better.

The book has chapters on fingerprints, footprints, dust, bullets, bombs, body armour, speech analysis, scents, alcohol, poisons, forgery, fraud, dead bodies and genetic fingerprinting. This is the right approach. Kaye is a scientist and might easily have succumbed to a format based on the various forensic methods of analysis, citing interesting cases along the way. By focusing on the evidence itself, he has widened the book's appeal to the main audience: the legal profession.

I must confess that I most enjoyed the tasty titbits, such as the witches of Salem, the Turin shroud and Piltdown man. Most purchasers will seek out this book for its technical data, but many will find it worth buying for its entertaining stories, especially about forgeries. The Vinland map, for example, was reputed to show that the Vikings discovered America, and when it came to light in 1957 it appeared authentic because it was bound together with genuine medieval documents. That it was a forgery was glaringly obvious from the start, because it shows Greenland as an island, which it is, but this could not have be known to the Vikings because of the polar ice sheet. Scientific analysis of its ink finally proved it must have been drawn this century, but not before 10,000 reproductions had been sold to the American public at $15 each.

Interpreting Evidence, by Bernard Robertson of Massey University and Tony Vignaux of Victoria University, New Zealand, also has its case histories, but these take second place to theory. This book is seeking to revolutionise the way forensic evidence is evaluated. It is not the probability that a single piece of evidence is true, which concerns the authors, but how one piece of evidence reinforces others. To put it more simply: how do we decide when items of circumstantial evidence add up to near certainty?

The authors show it is possible to arrive at such a conclusion by using Bayes's Rule to determine a "likelihood ratio" for each piece of evidence and then combining these to calculate an overall probability. The approach is based on the theory of an 18th century cleric, the Reverend Thomas Bayes, who showed that dividing the statistical probability of something being true by the probability of its not being true gives a useful number, the likelihood ratio. In the case of forensic evidence the higher this is, the stronger the evidence in support of an hypothesis.

Each piece of evidence so assessed can then be used to up-grade the probability of the existing evidence. The authors refer to the process as converting "prior odds" to "posterior odds". Unless the numerical odds are significantly increased by this process then the value of the extra evidence is debatable. But juries, and lawyers, do not like thinking in terms of numbers, so Robertson and Vignaux have produced a range of likelihood ratios with corresponding describers. Thus evidence with a likelihood ratio in the range of 1 to 33 is described as weak, 33 to 100 as fair, 100 to 330 as good, 330 to 1,000 as strong and over 1,000 as very strong.

Some kinds of evidence, such as fingerprints and DNA, have likelihood ratios that run into millions and seem almost impossible to deny, although even here Robertson and Vignaux show how these one-in-a-million chances tend to be considerably reduced when combined with prior odds. On the other hand, evidence with relatively small likelihood ratios, such as matching glass, fibres or paint fragments, can mount up to produce "strong" evidence.

All those whose job it is to administer and uphold the rule of law - forensic scientists, lawyers, barrister, judges, crime reporters and the police - should study this book. They need not fear that a degree in formal logic or probability statistics is a necessary prerequisite. The basic propositions are explained simply and clearly and the maths is kept to a bare minimum. The book makes compulsive reading and draws upon a wide range of cases to illustrate its points. The authors even include reference to the early stages of the O. J. Simpson proceedings. I suspect that in a second edition, which this book will surely run to, this famous case will feature in more detail as a classic for them to analyse and prove their approach really would have worked better. It might then become a bestseller.

Both books are superb in their different ways. Perhaps in the next millennium we will see forensic textbooks that combine the science of Kaye with the logic of Robertson and Vignaux. Lawyers taught how to understand and evaluate forensic evidence correctly is surely a promising way in which to reverse the declining faith in our legal system.

John Emsley is the science writer in residence, Imperial College, London, and the winner of the 1995 Rhone Poulenc Science Book Prize for his Consumer's Good Chemical Guide.