Background Affective temperaments are associated with suicidal risk, but their predictive value relative to diagnosis of major affective disorder is uncertain. Methods We compared diagnoses, affective-temperament ratings (TEMPS-A), and other potential risk factors in 956 psychiatric inpatients, using bivariate analyses and multivariable logistic regression modeling for associations with suicidal status. Results Lifetime suicide-attempt rates were high (43.9% overall), ranking by diagnosis: bipolar-II (58.4%), major depressive (50.0%), bipolar-I (44.6%), other (38.0%), and psychotic (33.9%) disorders. TEMPS-A scores for depressive (dep), cyclothymic (cyc), irritable (irr), and anxious (anx) temperaments and their sum were strongly associated with suicidal risk; hyperthymic (hyp) temperament scores were inversely associated; and a composite measure (dep+cyc+irr+anx – hyp), even more strongly associated. The composite score was highly, independently associated with suicidal behavior (p