Kant is widely regarded as holding a retributivist view of punishment. I think this common opinion is obviously correct, supported explicitly by many texts. But Kant also devised an entire practical philosophy, a theory about the foundations of right and ethics, and a theory of justice and ethical duties and a theory of justice based upon it. If we are to do Kantian ethics properly, we must constantly ask how Kant's own moral convictions relate to his practical philosophy as a whole – for instance, how, or even whether, these convictions can be supported by his theory. It cannot be a foregone conclusion that everything Kant says follows validly from his fundamental principles or is even consistent with them.

It is a sound hermeneutical principle that in studying any philosopher we should at the start provisionally assume that the philosopher's thought constitutes a coherent unity. Thus if Kant emphatically asserted a retributivist theory of punishment, then we should begin with the assumption that this retributivism can be supported by, or somehow integrated into, his larger theories of right and morality. Our first task should be to look for a way that Kant's retributivism can be seen to fit into his practical philosophy. Yet it is an equally valid hermeneutical principle that this assumption of unity and coherence should be only provisional or tentative.