Forensic identification science involves two fundamental steps. The first step is to compare a questioned item of evidence to an exemplar from a known source and judge whether they appear so alike that they can be said to match. The second step is to assess the meaning of that reported match: What is the probability that the questioned and the known originated from the same source?
Different risks of error are present at each step. The risk of error in the first step is that a reported match between a questioned and a known sample might not really match. Even if the method used to compare questioned and known samples were flawless, an error could occur if, for example, one of the samples had been mislabeled or mixed up with a different sample. The risk of error associated with the second step is that, while accurate, the reported match may have arisen through coincidence and not because the samples share a common source. The risks of error at both steps affect the ultimate inferences that can be drawn about the identification evidence in a case.
Both risks are subjects of far too little research. As to the first step, existing standards and procedures do not provide sufficient protection from erroneous conclusions that two marks are indistinguishably alike-that is, that they "match" when in fact they differ. Few, if any, criminalistics subfields have objective standards for deciding whether two patterns match.
Michael J. Saks and Jonathan J. Koehler,
The Individualization Fallacy in Forensic Science Evidence,
61 Vanderbilt Law Review
Available at: https://scholarship.law.vanderbilt.edu/vlr/vol61/iss1/4