The study analyzed the personal characteristics, previous record, and subsequent convictions of 1,000 drunk drivers convicted in New South Wales in 1972 and followed up for 3 years from their conviction date or prison release date. Statistical controls, in the form of linear models analysis adjusted for the fact that offenders receiving heavier penalties are generally worse risks than those given lighter penalties. Offenders are classified as high, medium, or low entitlement for punishment, and as having received a high, medium, or low severity penalty. Neither heavy fines nor long disqualification periods were found more effective than low fines or short disqualification periods in reducing recidivism rates for drunk driving. Furthermore, imprisonment was no more effective than any other penalty for any group of offenders, and there is strong evidence that longer imprisonment, specifically beyond 6 months, encouraged reoffending, especially drunk driving. The likelihood of recidivism was not related to age. High-risk groups in the general motoring population may be impervious to countermeasures such as publicity campaigns and random breath tests. Sixty-six tables, 12 figures, and 51 tables are included.