Methodological problems intrinsic to suicide prediction are discussed. The research on suicide predictive signs derived from attributes of the suicidal person, his behavior, and his surroundings, and on predictive instruments is reviewed. The results of these researches are equivocal, in that neither single signs, standard psychological tests, specially devised tests, clinical judgments, nor scales are found to be able to predict suicide at useful levels. Scales are shown to offer the best predictive potential, but to date their construction has not been systematic. It is proposed that scales might be better constructed and prediction improved if investigators were to consider the focus, duration, inclusion‐exclusion, interactions and subgroupings, and generalizability of suicidal indices.