Fingerprints Founds At Crime Scenes

The first difficulty that makes the situation worse in practice is that fingerprints found at crime scenes tend to be incomplete. What are being compared are thus not whole prints, but mere fragments. Nothing, not even Galton's original analysis, has anything to say about the likehood of fragments of prints coinciding in different individuals.
The Second difficulty is that most fringerprint evidence found at the scene of crime is latent. In other words it requires treatment with chemicals, or illumination with ultraviolet light, In order to make it visible enough to work with and, even then it is often indistinct. How valid it is to compare such filtered evidence with the clean crisp prints obtained from suspects in controlled conditions is another unexplored question.

The upshot is that at least by comparison with the techniques used to process DNA evidence which are often, in tribute to the awe in which the older technique is held, referred to as DNA FINGERPRINTING, fingerprints look technically flawed. And lawyers(that set standards for the admission of scientific evidence in court) are starting to notice.

The turning point was a case of Byron who allegedly drove the getaway car in a robbery carried out in Pennsylvania in 1991. In 1998 Mr Mitchell appealed against his conviction. The case turned on two latent prints, one found on the getaway car's steering wheel and the other on its gear lever second that were said to link him to the crime. The detail of the case are tortuous; Mr Mitchell's conviction was upheld, but his lawyer another doubter of the value of fingerprints, is still trying to have it overturn. During the course of the trial, however the FBI did something that had never been done before. It carried out a rough and ready experiment to test the reliability of the fingerprints.
It did this sending the latent prints, plus inked prints of Mr Mitchell's fingers, to the laboratories of 53 state law enforcement agencies. Eight of the 35 agencies that responded were unable to find match for one of the latent prints, and six failed to match the other - an average failure rate of 20%. That is was a shocking result so the use of a technique that may have an error rate as high as 20% raises a lot of legal questions.