Reconvictions for DWI and other offenses were determined for 3 years from conviction or release from prison. The penalties involved were fines, license suspension, and prison/bond. The relationship between penalties and reconviction rates was controlled by 25 offender and offense characteristics. Out of the 1,000 in the sample, 378 were reconvicted for some offense. A total of 149 were reconvicted for a DWI offense. The effects of penalties were found to depend on offender characteristics, i.e., whether they were 'good risks' or 'high risks.' For good risk offenders, long periods of license suspension (up to 18 months) were most effective in preventing further traffic offenses other than DWI, and this did not appear to increase the offense of driving under a license suspension. For most high risk offenders, no penalty appeared effective in preventing DWI reconvictions. There was no justification for using imprisonment as a deterrent, and the relatively harsher penalties currently imposed on young offenders in New South Wales were not effective in reducing recidivism. Ignition interlock devices fitted to the cars of high risk offenders so that the ignition will operate only after successful completion of a psychomotor or breathalyzer test appear urgently needed. Tabular and graphic data, 11 notes, and 26 references are provided. (Author summary modified)