I had been reading articles about how "incontrovertible" eyewitness identification and/or fingerprint identification led to the conviction of indicted persons, only to have the conviction legally overturned when DNA evidence showed scientifically that the person so identified was not the person who committed the crime. I began to investigate the use of fingerprints, looking for the basic science involved in saying that a fingerprint found at the scene of a crime matched that of a suspect, and I was surprised and shocked at what passed for "scientific evidence". As you will see below, any fingerprint identification is purely a judgment call, and the only basic science is that no two individuals have been shown to have the same ten fingerprints, much as no two individuals have exactly the same face (and both statements do apply to identical twins). The only distinction between the two is that some people are born without fingerprints and never develop them, while everyone has a face.
The fingerprint found at the scene of a crime is called a latent fingerprint. It typically is only a portion of a fingerprint, and is usually smudged and distorted. The job of the fingerprint expert (with or without the help of a computer) is to make the final decision as to whether or not this partial fingerprint can be reliably and validly said to have come from a person whose fingerprints are on file. The decision is a judgment call, based on the expert's experience and visual intuition, and the fact that no two people have the same set of fingerprints is not the same as saying that no two people could leave the same latent fingerprint.
The examiner looks for points of agreement, called Galton points, between the partial ridges of the latent fingerprint and the fingerprints on file, and no one knows the error rate for fingerprint identification. How many points of agreement are needed to call it a match (assuming the determination of a point of agreement is made without error)? New York, California, and London all have different numbers. In Italy, 16 points of agreement are needed to declare a match, while France and Australia require 12, and our own FBI has no minimum number of matching points to declare an ID. In America the number of points of agreement required varies from state to state, and even within a state.
No pair of latent and rolled (recorded) fingerprints are ever 100% identical, so the question is: how much agreement is enough? There is no available evidence to answer the question as to how much correspondence between two fingerprints is needed to say that they were made by the same person. The examiner decides how much is enough, and there are no universally agreed upon standards as to what constitutes a match, nor are there any published studies of the ability of experts to match one fingerprint to another.
I could find mention of only two tests given nationwide to fingerprint experts. In one case, the FBI sent two fingerprints from a robbery (where the identity was known) to 50 state fingerprint laboratories, and 10 failed to identify them correctly. In 1995 the Collaborative Testing Service, in a test okayed by the International Association for Identification sent 156 experts 4 suspect cards with all 10 prints along with 7 latents for identification. Only 66 of the 156 correctly identified all 7 latents, a success rate of less than 50%, and there were 44 incorrect identifications.
Lest my readers think that this is all theoretical, and I am exaggerating the room for error, let me close by citing two internationally known cases of fingerprint error, one by the FBI and one by Scotland Yard, which led to the arrest of innocent people (one of them a policewoman), and then payment of monetary damages ($2M to the American) to both individuals because of the harm done to them.
In 1998 the fingerprint of Policewoman McKie was found in the room of a murdered man in Scotland. The police had arrested a man for the murder, and asked her to account for the presence of her fingerprint. She swore that she had never been in the room, but even her father, a former policeman, believed the evidence. On the basis of the one fingerprint, she was indicted and put on trial for perjury. Her expert (who was later hounded out of his job at Scotland Yard and shunned by his co-workers) testified that not only wasn't it her fingerprint, but they police had claimed that the partial had matched her left thumb, while he clearly demonstrated that the latent was of a right forefinger. She was acquitted but left the force.
In 2004 a bombing in Madrid, Spain killed over 100 people. The Spaniards found one latent print, and circulated it around the world. The FBI claimed that it was the print of an Arab in the U.S. named Mayfield, and arrested him as a material witness. They said they had 15 points of agreement, and that this was verified by three different FBI fingerprint experts. The Spanish police said that there were only 7 points of agreement, and that he couldn't have been the man (plus he had never left the country), and that the FBI had misidentified him. The Spaniards later found a man whose fingerprint they said was a match, and they then connected him with the presence of explosives, arrested him and convicted him.
Until I began my reading, I did not realize how subjective the "science" of fingerprint-matching really was. There are no hard and fast standards, and the match is really in the eye of the beholder, much as is facial identification. The fact that fingerprints are almost certainly unique does not in and of itself make fingerprint ID valid and correct, any more than the uniqueness of our face makes witness ID 100% accurate.