Many empirical studies have found both inter-industry and intra-industry externalities in the form of local knowledge spillovers in research. This paper makes some assumptions reflecting these empirical regularities in order to analyse their implications for the allocation of economic activities between two regions. The two main assumptions are that R&D guarantees a positive equilibrium rate of growth in the volume of output by increasing the marginal productivity of labour, and that it is characterised by geographically bounded intra-industry as well as inter-industry knowledge spillovers. The existence of an iceberg type cost in transporting consumption goods from one region to the other, together with increasing returns to scale in production, introduces a centripetal force; this is opposed by a centrifugal force associated with congestion costs: agents living in crowded areas suffer a reduction in their level of utility. In equilibrium, different locations of research and manufacturing firms can result. Where transport costs are higher (congestion costs lower), centripetal forces dominate and all economic activities end up concentrated in one region. As transport costs decrease (congestion costs increase), an equilibrium with activities in both regions becomes more likely.