Suicide assessment and prediction are important functions performed by psychiatrists. Although the specificity and sensitivity of assessment instruments are poor, trainees are instructed to become proficient in their application. The occurrence of suicides during training confronts trainees with personal and professional limitations. The charts of patients who committed suicide while on the rolls of a large university-affiliated private psychiatric hospital, or within one month of discharge, were reviewed for relevant clinical characteristics and to assess the parameters related to training. Suicided patients were matched by age, sex, and unit of the hospital with a control selected from those admitted one or more months before the index patient. Over an 8.5 year period, 20 men and 15 women committed suicide, the mean age was 32.7 years and the vast majority were white, single, and of low socioeconomic status. There were no significant differences between suicides and controls on race, religion, marital status, or socioeconomic status; number of prior hospitalizations; prior discharges; or length of stay for index hospitalization. There were significant differences in suicidal tendency ratings, presence and number of suicide attempts, history of suicidal ideation and treatment with antidepressant medications, and recent feelings of hopelessness. There were trends for suicided patients to have an Axis I diagnosis of Affective Disorder only and to have had more than one therapist during their index treatment episode. (Author/NB)