Geomorphologists are currently using a wide array of techniques and approaches, ranging from computer simulation of drainage networks to detailed study of the physical character and behaviour of soil materials, to investigate landforms and land-forming processes. We have suggested (Mosley and Zimpfer, 1976) that the many approaches in use are complementary, in that each provides information not provided by another; nevertheless, it is desirable occasionally to 'take stock', to evaluate the relative merits and contributions of the available techniques, and identify areas of neglect and potential progress (Chorley, 1967, 60).There has been much interest recently in the explicit use of models in most branches of geography and the earth sciences, although model-building has long been a fundamental technique of science and the scientific method. Several writers (Chorley and Haggett, 1967; Harvey, r969;Minshull, 1975) have reviewed the justification for, types of, and uses of models in geographical research. In general, the use of models is associated with, and required by, the application to geographical problems of the techniques of scientific method-the generation and testing of hypotheses, collection and analysis of data, and formulation of general laws and theories. It would be inappropriate here to present a detailed discussion; suffice it to say that, of the many types of models identified (symbolic, analogue, mathematical, etc.-see Minshull (1975) for an overview), we are here concerned with only the hardware model.Discussions to date on the use and appropriateness of various types of models have been highly abstract. While it may not be possible to make general assertions regarding the appropriateness of a given type of model, because this must be decided on a case by case basis, it should be possible to evaluate the model's contribution to and role in scientific progress.We and a number of colleagues have been involved in an extensive research programme based primarily on hardware modelling techniques, which has considered a wide range of aspects of landform development. It is our present purpose to assess the potential contribution of hardware modelling to an understanding of landforms and landforming processes. In keeping with our topic, the approach will be 'empirical', and use the experience gained during the research programme at Colorado State University to formulate general conclusions.