Source: (1997) Ph.D. dissertation, Graduate Faculty of Criminal Justice, Claremont Graduate School, California.Federal sentencing laws were reformed in 1984. The result has been a limitation of sentencing discretion. Federal courts now employ sentencing guidelines calculated by probation officers and reviewed by the courts. Application of these guidelines to a given defendant results in a sentencing range and a determination of available sentencing options. These options are limited to probation, supervised release, community confinement, and imprisonment. Since the sentencing guidelines came into effect the United States Sentencing Commission has reported annual statistics of the sentencing practices of the various federal district courts. Among these is the rate of compliance for any given district. According to the sample of sentencing decisions collected by the author, the United States Sentencing Commission has overestimated compliance in the Central District of California. The sample of sentencing decisions shows that the sentencing judge, the number of prior arrests and convictions, and a guilty plea by the defendant, were significantly related to non-compliant sentencing outcomes. One reason that compliance may be less than reported by official statistics is because some judges sentence outside of the presumptive guideline range without making formal findings that their decisions are in fact departure from the guidelines. It is suggested that the reasons for this practice involve the personal evaluation of offenders by judges and that these evaluations are based on factors which are not accounted for by the sentencing guidelines, but are none-the-less defensible in terms of certain principles of sentencing. It is also suggested that judges may hold implicit models of criminal behavior not accounted for by the sentencing guidelines. One such model is based on the sentencing principle of social justice. A second model is based on the principle of restorative justice. Author’s abstract.