Suicide seems to be increasing in young people in various countries and causes the greatest loss of years of life under the age of 65 in the Swedish population. Data from a national survey of 50,465 conscripts in Sweden were used in a prospective follow up study to assess personality and behavioural predictors of suicide in young men. Altogether 247 completed suicides occurred in the cohort during 13 years' follow up. Baseline data on social conditions, psychological assessments, and psychiatric diagnoses of the conscripts were entered into a Cox regression model with suicide as the outcome variable. Several early indicators of antisocial personality (poor emotional control, contact with a child welfare authority or the police, and lack of friends) were strongly predictive of suicide. None of the few conscripts who had a diagnosis of schizophrenia or affective psychosis committed suicide. A diagnosis of neurosis was associated with a twofold increase in the suicide rate and personality disorder with a threefold increase. Although the risk of suicide is difficult to assess in an unselected population owing to the low base rate of suicide, the predictors identified in the study may help to identify those at high risk in units where people with deviant behaviour and personality disorders cluster.