The study analyzed the court records of 14,311 drunk-driving cases involving convictions in New South Wales in 1976. Occupational status was measured using Congalton's (1969) four-point scale. This scale classifies occupations according to their prestige as perceived by the Australian public. Penalties were analyzed simultaneously along the dimensions of the amount of fine, the period of license disqualification, and the outcome category. Sixteen variables, including the penalty dimensions, were incorporated into a series of multivariate linear models. About one-third of the total explanatory power of the model used was contributed by previous drunk-driving convictions, while one-fourth was due to magistrate variations. The model allowed an identification of the major dimensions along which magistrates varied. A general toughness-leniency dimension was evident, with most magistrates clustered at the lenient end of the spectrum. There was a general trend for older offenders and offenders of higher status to be dealt with more leniently than young offenders or offenders of low status. The policy of penalizing young and low-status offenders more severely on the grounds that the community is thereby being protected is without scientific or moral justification. Sentencing guidelines should be pursued based on a just deserts model with the dimensions of offense seriousness and the blameworthiness of the offender as measured by previous drunk-driving convictions and traffic offenses. The ranges of penalties specified for each cell in the grid should be sufficiently broad to allow for mitigating factors.