A central theme of the Knowledge and Regional Economic Development Conference is that of the geographical dimensions of knowledge and innovation. In this paper we are going to investigate the importance of geographical proximity for partnership formation in innovation systems, in which there is an assumed tacit aspect to knowledge flows. Once the theoretical framework has been defined, we shall present our rationale for an agent-based modelling approach to understand these phenomena, and then describe the model we are developing. We report on an initial series of simulation experiments we have conducted with this model, showing that when the geographical bound on the ability to use knowledge in joint innovations is relaxed, the number of partnerships increases approximately twofold. However, geographical aspects of the simulation did not change the surprising result that firms showed a large discrepancy in the number of innovations they attained.