We used a prospective longitudinal study to examine the utility of Moffitt's (1993) trajectories of antisocial behaviour. Data on registered criminality in three time periods - before age 15 (childhood), from 15 to 20 (adolescence) and from 21 to 35 (adulthood) - were used to construct life-course trajectories of offending for males. Life-course-persistent and adolescence-limited groups were found. The life-course-persistent males had the most problematic upbringing conditions, school problems and adjustment difficulties in adolescence, and the highest social and mental health problems in middle age. Adolescence-limited offenders did not differ much from non-offenders. In these respects, Moffitt's typology was confirmed. However, there was an equally large childhood-onset desister group. They had many of the same problems as the life-course-persistent males up to age 15, but did not differ much from the non-registered males in mid-adolescence or at the middle-age follow-up. These males are not predicted from Moffitt's model, but cannot be ignored. There was also a group of males who started to offend in adolescence and continued in adulthood, who had about the same problematic upbringing conditions, mid-adolescent maladjustment, and middle-age social and mental health problems as the life-course-persistent group.