Wilkins' empirically-based descriptive sentencing guidelines are derived from the application of multiple regression to actual case data to identify, weight, and combine the major independent case factors related to sentences. One difficulty with this approach is that some assumptions underlying multiple regression are inconsistent with the complexities of judicial thought. The Minnesota Sentencing guidelines use a two dimensional grid based in offense seriousness and the offender's criminal history. This system is unsatisfactory for Australian courts because it does not address how information about cases ought to be combined. A sentencing model appropriate for Australian courts should be a faithful representation of the structure of judicial thought, incorporate most of the case factors encompassed in legal analyses of sentencing judgments, and implement the sentencing rules for combining case information promulgated in Australian appellate decisions. The latter principles are a detailed statistical description of current sentencing practice and assistance in moving toward uniformity of judicial approach in sentencing. 20 references.