This article comments upon a previous article by Leonard (1977). Concern is raised about: a) the complexity of multivariate prediction and its lack of replicibility, and b) the combining in research of patients who attempt and threaten suicide. This article calls for clinical utility in suicide prediction indices.