Skip to content

Figuring Out Fingerprints

​FOR YEARS, attorneys and law enforcement officers argued that latent fingerprint identification was essentially infallible. That perception has slowly given way to a more realistic view of the science’s shortcomings, spotlighted by the misidentification of a fingerprint from the 2004 Madrid terrorist train bombing that led police to the erroneous arrest of Oregon lawyer Brandon Mayfield.

The reliability of fingerprinting and other forensic science was further called into question two years ago by a National Research Council report. Since then, President Obama’s administration has convened a committee on forensic sciences to study the issue, and Senator Patrick Leahy (D-VT) has introduced S.132, which calls for forensic science reform, including research into the validity of the various sciences and accreditation standards for forensic laboratory workers.

Researchers are also beginning to explore the reliability of the forensic experts who analyze and draw conclusions about latent fingerprints. The research is necessary because, while television shows portray computers scanning databases for an electronic match, in reality, it’s often a human who makes the final call about whether a fingerprint collected at a crime scene matches one on file.

“Current computer algorithms cannot be relied upon to make the critical judgments, and humans are likely to remain in the loop for many years to come,” says Matthew B. Thompson, a Ph.D. researcher at the University of Queensland and National ICT Australia who recently coauthored a study on latent fingerprint examination. “The problem is that, even though fingerprints have been used in criminal courts for more than 100 years, no properly controlled experiments on fingerprint examiners’ accuracy in identifying perpetrators have been conducted,” he notes.
In Thompson’s study, the experts correctly matched just over 92 percent of the prints to the criminal. They mistakenly matched 0.68 percent of the prints to the subject.

In another recent study conducted by the FBI, there was a .10 percent false positive rate and a false negative rate of 7.5 percent. Additionally, researchers differed in opinion about whether the information gathered from the fingerprints was sufficient for reaching a conclusion.

Thompson admits that one could view the error rate as low, but given the context, it may still not be acceptable to the extent that it results in either innocent persons being convicted or guilty ones getting away. Still, Thompson says, he leaves “the broader interpretation of the significance of this evidence to legal scholars and the fingerprint profession itself.”

Other fingerprint experts interviewed note that certain benchmarks are lacking. For example, there are currently no guidelines regarding how much of a latent fingerprint must be recovered for the print to be acceptable as evidence. Investigators are often given only very small bits of a print to examine.

“If you have a latent fingerprint of a certain size, a certain fraction, and it is distorted to a certain degree, what’s the likelihood that when the examiner looks at this and runs it through the [automated fingerprint identification system], AFIS, that he or she will come up with a match to somebody who did not leave the fingerprint? That’s what we really care about,” says Thomas L. Bohan, director of MTC Forensics and former president of the American Academy of Forensic Sciences. He admits, however, that with the different types of distortions that can occur to prints, it may be difficult to create meaningful standards.

But what can happen, he says, is that fingerprint evidence can be presented more honestly in trials. He equates it to the evolution that has occurred with eyewitness testimony. “When there is eyewitness testimony, there is testimony from the defense as to the flaws in eyewitness testimony. So that’s what I would see as being the change that would occur once we get some handle on the reliability of fingerprints and judges take judicial notice that there is not a zero error rate,” says Bohan.