Fingerprints

For most of the century since it made its courtroom debut, fingerprinting has enjoyed an impeccable reputation for identifying criminals. What jury would acquit a suspect if his prints matched those found at the scene of a crime? It was thus understandable that when a speaker at a recent meeting on science and law held in San Diego by America's Justice Department hinted that the technique might not deserve its aura of infallibility, an FBI agent in the audience was later overhead calling him an unprintable name. Understandable, but not, says the speaker, justified. For he is one of a small group of people that has started looking at the technique which, above all others, gave forensic science its scientific status. And, surprisingly, he has found it is scientifically and statistically wanting.

This is not to say that the world's prisons are full of innocent victims of dodgy evidence. But the fact is according to Dr Cole, who researched the subject at Cornell University, that fingerprints has never been subjected to scientific scrutiny required in a modern courtroom. An he thinks it should be. Modern fingerprinting goes back to Francis Galton, a 19th century British scientist who, ironically, helped to pioneer the use of statistics. In 1892 Galton looked at the pattern of whorls, arches and loops that make up fingerprints, and estimated that the chance of two prints matching at random was about one in 64 billion. That estimate, however, has never been backed up by any data. Besides, Galton was not really comparing whole prints. Instead, he identified places where the ridges of which fingerprints are composed, either end or split. These are now known as "points of similarity", or "Galton details" and if two prints have enough points in common they are deemed to be identical.

Galton's estimate relied on using every available point, there are generally between 30 and 50. Current practice, which varies widely from one place to another, has been to declare a match if there are somewhere between 8 and 16 points of similarity linking a fingerprint found at a crime scene and one taken from suspect. Unfortunately, the validity of this process, and the number of points of similarity needed to make it statistically secure, have not been scientifically investigated. Nor has the alternative technique, introduced in England of relaying on an examiner's overall impression of a match, without any attempt at quantification. That puts fingerprinting on shaky theoretical ground.