DNA Database Searches and the Legal Consumption of Scientific Evidence

Abstract

DNA evidence has transformed the proof of identity in criminal litigation, but it has also introduced daunting problems of statistical analysis into the process. In this Article, we analyze a problem related to DNA evidence that is likely to be of great and increasing significance in the near future. This is the problem of whether, and how, to present evidence that the suspect has been identified through a DNA database search. In our view, the two well-known reports on DNA evidence issued by the National Research Council ("NRC"), each of which has carried great authority with the American courts on various issues, have been badly mistaken in their analysis of this problem. Similar errors affect the analyses of scholars who have supported the NRC reports. We will also offer some reflections on the habits of mind, of both lawyers and statisticians, that may have led to this result. Finally, we will suggest an approach that legal decisionmakers might take in general with respect to scientific and statistically-based evidence to avoid this kind of difficulty. This approach is significantly different from that which pervades the Supreme Court's decisions in Daubert v. Merrell Dow Pharmaceuticals, Inc. 1 and its recent sequel, General Electric Co. v. Joiner, 2 as well as their precursor, Frye v. United States. 3 And the

The DN A Database Search Controversy

“…The viewpoint that I have sketched above is now in broad terms shared by most statisticians working in the field (Robertson and Vignaux, 1995, Chapter 7;Balding and Donnelly, 1996;Dawid and Mortera, 1996;Weir, 1996;Taroni and Aitken, 1997;Evett and Weir, 1998, Chapter 9;Donnelly and Friedman, 1999;Dawid, 2000). Had history followed a different course, the effect of database searches on identification evidence might now be regarded as merely an interesting statistical phenomenon suitable for textbook examples.…”

Selection effects and database screening in forensic science

“…These observations are based on contributions of many authors, including Robertson and Vignaux [13], Balding and Donnelly [17], Donnelly and Friedman [21], Dawid and Mortera [18], Evett and Weir [1] and Dawid [19].…”

Forensics

“…The confusion has arisen because of the observation that the expected number of matches increases with the size of the database, and there has been a recommendation (National Research Council, 1996) that a profile probability of P should be modified to NP when a suspect is identified by searching a database of size N . Balding and Donnelly (1996) and Donnelly and Friedman (1999) present careful analyses to show that the LR for the database search case cannot be greater than for the single suspect case. The evidence against the one person who goes to trial is essentially unaltered by the means by which he was identified.…”