Sunday, June 24, 2007

Fingerprint-matching pseudoscience

Fingerprint matching has been used as an investigative tool by law enforcement and as a key piece of evidence to convict criminals in courts for over a century, but its accuracy has not actually been scientifically tested until recently. It turns out that claims of its accuracy have been exaggerated, and fingerprint matching is often more art than science.

In 2004, the FBI claimed that a fingerprint found on a bag at the sign of a terrorist bombing in Madrid, Spain on March 11 was a match to the left index finger of Brandon Mayfield, an attorney in Beaverton, Oregon who converted to Islam and married to a Muslim woman from Egypt. Despite the fact that Spanish police disagreed, claiming that there was no match to Mayfield, the FBI insisted they had a "one hundred percent identification" with fifteen separate points of agreement between the latent print from Spain and Mayfield's fingerprint, validated by at least three FBI fingerprint examiners. Mayfield was arrested and detained on May 6, 2004. On May 20, Spanish police announced that they had matched the fingerprint to Ouhnane Daoud of Algeria, who--unlike Mayfield--had actually been in Spain. Mayfield was released and the FBI ended up apologizing.

This case has resulted in scientific scrutiny of fingerprint evidence that has been long overdue. A decade ago, Tucson printer and publisher of the anarchist periodical The Match!, Fred Woodworth, published "A Printer Looks at Fingerprints," in which he pointed out pseudoscientific reasoning in fingerprint matching methodology as described in fingerprint textbooks. Subsequently, Simon Cole authored the book Suspect Identities: A History of Fingerprinting and Criminal Identification (2002, Harvard University Press), and has just authored an article on the subject in the July/August 2007 issue of Skeptical Inquirer, titled "The Fingerprint Controversy."

In Cole's Skeptical Inquirer article, he states that "The very first study containing accuracy data was finally published just recently, finding very high accuracy rates in a class of trainees on latent prints of unknown difficulty; but the study contains some methodological flaws (Haber and Haver 2006). Moreover, the authors again argue strongly against inferring accuracy rates from their own data (Wertheim, Langenburg, and Moenssens 2006)."

No doubt scientific investigation will demonstrate that proper use of fingerprint analysis is a reliable method of identification, but more importantly, it will find its limits and weaknesses so that it does not continue to be pressed beyond its capabilities and result in false judgments of guilt in criminal cases. Unfortunately, law enforcement and prosecutors have a vested interest in the flexibility of techniques that can be used to produce the judgment they want, as demonstrated by the difficulty in getting police departments to modify their procedures of eyewitness identification of suspects to correct for well-known cognitive biases.

No comments: