The mammalian brain is assembled from thousands of neuronal cell types that are organized into distinct circuits to perform behaviourally relevant computations. To gain mechanistic insights about brain function and to treat specific diseases of the nervous system it is crucial to understand what these local circuits are computing and how they achieve these computations. By examining the structure and function of a few genetically identified and experimentally accessible neural circuits we plan to address fundamental questions about the functional architecture of neural circuits. First, are cell types assigned to a unique functional circuit with a well-defined function or do they participate in multiple circuits (“multitasking cell types”), adjusting their role depending on the state of these circuits? Second, does a neural circuit perform “a single computation” or depending on the information content of its inputs can it carry out radically different functions? Third, how, among the large number of other cell types, do the cells belonging to the same functional circuit connect together during development? We use the mouse retina as a model system to address these questions. Finally, we will study the structure and function of a specialised neural circuit in the human fovea that enables humans to read. We predict that our insights into the mechanism of multitasking, network switches and the development of selective connectivity will be instructive to study similar phenomena in other brain circuits. Knowledge of the structure and function of the human fovea will open up new opportunities to correlate human retinal function with human visual behaviour and our genetic technologies to study human foveal function will allow us and others to design better strategies for restoring vision for the blind.