Cognitive agents are continuously faced with new problems. To facilitate adaptation, emerging theories of neural reuse propose that evolution might often favor re-purposing existing brain structures for new functions. This paper presents a novel approach to the study of neural reuse based on the evolution of simulated agents in an object-categorization task. We artificially evolve populations of dynamic neural networks to perform two variants of a categorization task that alternate over evolutionary time. We find that populations become increasingly adaptive over repeated exposures to the tasks. Analysis of evolved networks reveals two types of equally-fit solutions: one that is specialized to a given task variant and does not adapt to changes easily; and another that is more general, in that it can adapt to the other task with minimal change to its structure. Interestingly, we find that populations exposed to alternating tasks spontaneously locate the latter type of structures.