This paper addresses issues that may influence the evaluation of evidence when crime material from different cases is compared and no common source of the evidence is known.
Generally, decisions about a potential common source of evidence or case linkage involve risky arguments that require sound rules of reasoning. This paper discusses probability theory as a means of dealing with such uncertainty. In addition to linking items of evidence on the basis of similarities in compared characteristics, forensic scientists may also be called on to assist in the evaluation of the nature and the strength of inferred links. The latter is the focus of this paper. The authors discuss ways scientists can assist in developing arguments for a common source or case linkage. Within this process, several intermediate propositions are usually encountered, together with sources of uncertainty. Probability theory is an appropriate tool for guaranteeing coherence of the decisionmaker's inferences under such circumstances; however, even when only a few variables are involved, formal probabilistic calculus may become intractable. Such difficulties can be addressed by formalisms such as Bayesian networks. Their application to decisionmaking about common sources of evidence and case linkage is discussed. In order to illustrate the proposed decisionmaking scheme, the authors show how it applies in deciding whether two stains with no reputed common source are linked. The Bayesian networks presented in this paper allow for a distinction between the observations on a crime stain and the characteristics of its source. This enables scientists to account for uncertainties scientists may have when drawing inferences about properties of a source based on observations of a possibly nonrepresentative and/or degraded sample. 7 tables, 9 figures, and 20 references