Such basic conceptual tools can facilitate observations about a single system and can enable comparative analyses of numerous systems. The author posits certain working hypotheses about the design of sentencing systems in the context of the many observed possibilities for the parceling out of sentencing discretion. These propositions are not so much "rules" or "principles" as tentative observations and normative propositions that will be useful in the analysis of particular systems. The first hypothesis is that it is undesirable to concentrate all sentencing discretion at the systemic level and foreclose all meaningful sentencing discretion at the case-specific level. The second hypothesis is that no attempt to concentrate sentencing discretion at the systemic level can extinguish meaningful discretion at the case-specific level. The third hypothesis is that discretion exercised late in the sentencing process is likely to carry disproportionate influence over sentencing outcomes when compared with discretion exercised at early or intermediate stages. The fourth hypothesis is that the goals of a sentencing system are best furthered by visible, professional, and accountable actors who have primary responsibility to further systemic goals as opposed to the interests of particular parties; and the fifth hypothesis is that every sentencing system should be designed so that at least one visible, professional, neutral, and accountable public actor with responsibility to uphold systemic goals has power to exercise meaningful sentencing discretion at both the systemic and case-specific levels. The author applies these five hypotheses to three longstanding guideline jurisdictions in an attempt to generate useful raw and comparative observations: the Federal guidelines system, which began operation in late 1987; and the Pennsylvania and Minnesota systems which took effect in 1982 and 1980 respectively. 6 figures, 46 notes, and 70 references