Pattern analysis in forensic identification

A study assesses the reliability of a forensic identification technique. Photographic pattern analysis, in which features on a perpetrator’s face, hands, or clothing, obtained from crime scene photographs, are matched to those of a suspect, has been widely used in criminal cases over the past 50 years. However, the reliability and reproducibility of such analyses have not been rigorously tested. Sophie Nightingale and Hany Farid analyzed a technique for matching purportedly distinct patterns of wear and tear on denim jeans. The authors quantified the differences in such patterns between different pairs of jeans, and between different images of the same pair of jeans taken under varying conditions. From the distributions of these differences, the authors calculated the probability of correctly matching the same seams–true positive rate–and of incorrectly matching different seams–false positive rate. Differences between multiple images of the same pair were comparable to differences between different pairs. As a result, even at a relatively high false positive rate of 1 in 1,000, true positive rates were less than 40%. The results suggest that identification based on denim jean patterns should be used with extreme caution and that other forensic photographic pattern analyses should be validated using similarly rigorous studies, according to the authors.

###

Article #19-17222: “Assessing the reliability of a clothing-based forensic identification,” by Sophie Nightingale and Hany Farid.

MEDIA CONTACT: Hany Farid, University of California, Berkeley, CA; tel: 510-664-4438; e-mail:

[email protected]