The fingerprint controversy: astonishingly, one of the most trusted forms of forensic evidence, fingerprint identification, has yet to be validated. Lacking validation studies, proponents of fingerprint evidence have resorted to pseudoscientific arguments

A group of self-professed experts claims to have developed a new technique that is able to analyze objects found at a crime scene and determine whether a particular individual touched that object. The technique is so powerful that it can determine that a touch was made by a particular individual uniquely--to the exclusion of all other individuals in the world.

When this new technique is offered into evidence in a criminal trial, the defendant objects and demands scientific proof that this technique can, in fact, do what it claims to be able to do. A hearing is held to evaluate the validity of the experts' claims.

At the hearing, the experts note that training in their technique is accomplished through an apprentice system. A background in science is desirable but not necessary to practice the technique. The experts have conducted no validation studies measuring the accuracy of their determinations of identity. They have conducted no studies of the rarity of the attributes they analyze in the human population. The experts can articulate no standard that governs when a determination is made that an individual touched an object; the experts simply "know it when they see it." Asked the error rate of their technique, the experts reply that it is zero. "A person either touched the object or he did not." When instances of error are pointed out, they claim that, in those cases, the method was "applied improperly."

Such "experts" would probably be laughed out of court. The courts would tell them to come back when they had some scientific data to show. Unless, of course, they were fingerprint experts.

Over the past eight years, a battle has been waged in courtrooms, on the Internet, and in journal articles over the scientific validity and legal admissibility of fingerprint evidence. As the controversy developed, the views of the legal and scientific communities have become increasingly out of sync, with judges unanimously decreeing fingerprinting "reliable" while an increasing number of voices from the scientific community have agreed that there is no empirical basis for such a conclusion.

The debate initially took place in obscurity, in the pages of law journals and isolated courtrooms. Gradually, however, the debate has percolated into the scientific and mainstream media, fueled in part by two recent high-profile misattributions, one by the vaunted FBI Latent Print Unit (Fine 2006) and one that resulted in the dismantling of the Boston Police Department's latent-print unit (Cole 2005). The National Academies are taking an interest (Mnookin 2003); the controversy was the subject of a panel at a recent NAS Sackler Colloquium (see www.nasonline.org/site/PageServer?page name=sackler_forensic). The editor of Science wrote an editorial about it (Kennedy 2003). Today, we find ourselves in a peculiar situation in which the scientific and legal communities are increasingly at odds over this issue. It is becoming increasingly difficult to find a credentialed scientist who will argue that latent-print identification has been validated, while the number of scientists who agree that it has not been validated has grown (Siegel et al. 2006). But, at the same time, not a single court has restricted the admissibility of latent-print identification, despite legal standards that clearly require proof of reliability for expert evidence to be admissible.

There is, of course, one crucial difference between the technique I imagined above and the venerated forensic technique of fingerprint (or, more precisely, "latent-print") identification: fingerprinting is not new. Unable to point to any validation studies or error measurements, longstanding use has emerged as fingerprinting's last-gasp defense. Echoing the arguments of the Ptolemains, defenders of fingerprinting have argued that techniques that have been used in criminal proceedings for a century must be accurate.

To be sure, the wide use of fingerprint identification for nearly a century counts for something. If the technique were highly inaccurate, if its accuracy was little better than chance, we would probably know it. A finding that a suspect's "friction-ridge detail" is consistent with an impression found at the scene of a crime has definite probative value in a courtroom and law-enforcement value for solving crimes and apprehending offenders. But inferring that it isn't wildly inaccurate is not the same as knowing that it is highly accurate. And it is certainly a far cry from the claims of infallibility advanced on behalf of fingerprinting (McRoberts, Mills, and Possley 2004; Federal Bureau of Investigation 1985).

Can science and law be reconciled? What is the empirical basis for latent-print examiners' extraordinary claims? Is fingerprinting accurate, or is it "junk science"?

Junk Science?

The lay public often interprets the term junk science to refer to claims that are just plain false. This is not exactly what Peter Huber meant when he popularized the term in his book Galileo's Revenge (1991). In Huber's view, clinical ecology is not junk science because it is wrong. The clinical ecologists may eventually turn out to be right that something in the environment causes certain illnesses, but this would not alter Huber's classification of clinical ecology as junk science. Clinical ecology is junk science because it has prematurely made causal claims without sufficient proof.