Over the past two decades, neuroethologists have been unravelling the neural circuitry underlying some of the specialized capabilities animals use to obtain and process sensory information. Explicitly comparative studies of these systems can now be carried out to ask questions about the processes of evolutionary change in neural systems. In some cases, it can be seen how relatively minor modifications of neural networks may significantly expand the efficacy of sensory processing. Other comparative studies might ask whether expanding the number of steps involved in neural computation, or the area devoted to the representation of a particular sub-modality, constrains neural architecture in predictable ways.