A classification of suicidal behaviour was derived by cluster analysis from descriptive data on parasuicide patients. The grouping was meaningful in terms of clinical interpretation and had both therapeutic and prognostic implications. The sample was broken down into three groups, identified as repeated, serious and non-serious attempts. The repeated attempters had the worst prognosis a 1-year follow-up. An excessively high suicide rate was found in a small group of elderly attempters.