The use of self-organizing feature maps (SOFM) in models of cognitive development has frequently been associated with explanations of "critical" or "sensitive periods". By contrast, error-driven connectionist models of development have been linked with "catastrophic interference" between new knowledge and old knowledge. We introduce a set of simulations that systematically evaluate the conditions under which SOFMs demonstrate critical/sensitive periods in development versus those under which they display interference effects. We explored the relative contribution of network parameters (for example, whether learning rate and neighbourhood reduce across training), the representational resources available to the network, and the similarity between old and new knowledge in determining the functional plasticity of the maps. The SOFMs that achieved the best discrimination and topographic organization also exhibited sensitive periods in development while showing lower plasticity and hence limited interference. However, fast developing, coarser SOFMs also produced topologically organized representations, while permanently retaining their plasticity. We argue that the impact of map organization on behaviour must be interpreted in terms of the cognitive processes that the map is driving.