Hypothesis Testing in Law and Forensic Science: A Memorandum

This Forum Commentary series presents views on a memorandum from a group of lawyers and judges advising the Organization of Scientific Area Committees for Forensic Science (OSAC).¹ In response to calls for improving the practices of forensic science,² the National Institute of Standards and Technology (NIST) created the Scientific Area Committees in 2014 to promote and develop standards “that are fit-for-purpose and based on sound scientific principles.”³ The memorandum from the Legal Resource Committee (LRC)⁴ responds to a question from a scientist on OSAC’s governing board about whether the criminal law’s concern with avoiding false convictions at the expense of false acquittals should affect the choice of a “significance level” for deciding whether pieces of glass match in their chemical composition (and hence might have a common origin). Must a criminalist favor the hypothesis that similarities are coincidental over the hypothesis that the fragments have a common origin? The underlying issue applies to many forms of identification evidence, including fingerprints, fibers, paint chips, bullets, and biological fluids. Indeed, arguments over the choice of a significance level arise for statistical evidence of all sorts, from econometrics to epidemiology.⁵

This Introduction is a preamble to the memorandum. Part I describes the technical standard that prompted the memorandum. Part II sketches the statistical ideas in the memorandum by using glass comparisons to illustrate the three main statistical approaches to reasoning about the implications of evidence. This Introduction is followed by the memorandum itself and two commentaries.

* Associate Dean for Research and Distinguished Professor of Law, Penn State Law. This Introduction benefited from discussions with José Almirall, Karen Kafadar, and members of the Legal Resource Committee (LRC) of the Organization of Scientific Area Committees for Forensic Science (OSAC). The views expressed here are the author’s. They should not be attributed to NIST, OSAC, the LRC, or any other individual or organization.