This article presents two common accounts of the question of inference of source in forensic science. The first, the classic view, leads to direct opinions about source propositions, either categorically or in terms of graded conclusions. The second account focuses on assessing the value of the findings with respect to competing source propositions. It is based on probability theory and represents a preliminary to the most recent, decision-theoretic conceptualization of individualization. This account allows one to critically expose and resolve shortcomings, limitations and scientific drawbacks of the classic view, and to clarify the distinct duties and roles of experts and fact finders.