Computer simulation is a tool increasingly used by archaeologists to build theories about past human activity; however, simulation has had a limited role theorising about the relationship between past behaviours and the formation of observed patterning in the material record. This paper visits the argument for using simulation as a means of addressing the gap that exists between archaeological interpretations of past behaviours and their physical residues. It is argued that simulation is used for much the same reason that archaeologists use ethnographic or experimental studies, and that computational models can help to address some of the practical limitations of these approaches to record formation. A case study from arid Australia, examining the effects of episodic surface erosion on the visibility of the record, shows how simple, generative simulations, grounded in formational logic, can be used to compare different explanatory mechanisms and suggest tests of the archaeological record itself.