There's also a debate brewing in legal circles about the science of fingerprinting. This article in the Washington Post covers the issue, and is worth the trouble of free registration.



I recall learning about this in Michael Sak's "The Law, Litigation and Science" class back in my second year of law school. After reviewing basic scientific principles (correlation doesn't equal causality, results must be replicable, etc.), he pointed out: 1) There's actually no way to test the theory that all fingerprints must be distinct; 2) We rely on "points" of similarity - if a certain number of key characteristics are similar, then the powers-that-be declare the print to be a match. These things don't exactly jibe with the scientific principles.



I ran into this on a burglary case I had my first year out, where there were footprints being matched to the defendant's shoes. The DCI had found something like 9 points of similarity and were calling the shoes a match. Prior to depositions, I asked the agent what the standards were for matching - how many points did one need? I was told that there were no set standards, it would depend on how unique the characteristic is and the number of matches found, as well as other factors. That would have concerned me, had there not been sufficient corroborating evidence to convince me of the defendant's guilt, including a co-conspirator confession.



So what does one do with a partial print and eight points of similarity? Run the statistics on how likely the eight points are in that amount of a print? But fingerprint evidence is rarely put forth in terms of probability, like DNA evidence is. Maybe it should be.