The purpose of this paper is to survey several statistical procedures for testing exogeneity. The definitions of exogeneity and noncausality that are of relevance for this problem are first given. The classical instrumental variable procedure is then considered, and it is shown that an alternative formulation can be given in terms of a "completed" model. This alternative formulation indicates that most of the interference problems in the context of instrumental variables can be solved within a framework analogous to the limited information maximum likelihood procedure. This is notably the case for the exogeneity test.