With the proliferation of suicide prevention centers in the United States, the task of rapidly and effectively assessing an individual caller's suicide potential has become an important research problem. However, the social science researcher is often confronted with an ethical problem when the results of his predictive equations can be used to label people. It may be necessary to switch from predicting negative features to those which are positive and insulating. Predictaion instruments are best made for defined classes of persons. Longer lengths for the assessment instrument are not an advantage since research shows no relationship between predictive accuracy and amount of information available. Predictive instruments are relative to the persons involved, the criterion behavior to be predicted, and the social system in which the person and behavior are found. While predictive instruments can distinguish between different kinds of risks, estimate future risks, and form a baseline against which the outcomes of treatments can be judged, the challenge of predicting suicidal behavior remains. (JAC)