Despite the proliferation of work on the ‘punitive turn’, issues concerning its definition and measurement remain largely under-examined in the mainstream literature. This article seeks to refocus attention on the best ways in which to measure punitiveness and to argue that a more accurate characterization of what we mean by the concept forms an important part of advancing our understanding in this area. To illustrate this point, punitiveness in three countries is measured according to a unidimensional measure, a broader test proposed by Tonry and a fully multidimensional test. It is contended that the very different results produced by these tests suggest the need for greater social-scientific attention to the measurement of the punitiveness concept itself.