Koehler, J. J. (1995). Discussion: Inference in forensic identification. Journal of the Royal Statistical Society, Series A, 158, 47.

ABSTRACT

In this commentary on Balding and Donnelly (1994), I agree that the frequency of matching traits may be misinterpreted and overweighted by jurors even when the prosecution carefully avoids committing the "prosecutor's fallacy" (Thompson & Schumann, 1987). However, I suggest that it may be more difficult to convince people not to equate trait frequency with the probability that a matching suspect is not the source of recovered genetic trace material. Just as prior probability of guilt estimates should influence judgments about a suspect's guilt or innocence, prior probability of source estimates should influence suspect source probability judgments (Koehler, 1993a, 1993b). This point is lost on many forensic scientists--including the head of the FBI Laboratory Division--who insist that trait frequencies identify the probability that the suspect is not the source of a recovered trace (see e.g., Hicks, 1993).