We combine experimental and computational methods to study pattern generation and homestasis in small neural networks. Our model system is the stomatogastric ganglion in crustaceans, one of the best understood neural networks. We want to understand how cellular and synaptic properties shape the output of a neural network and how neural circuits manage to function reliably in spite of ongoing molecular turn-over and developmental or environmental changes. The tools we use to study these questions include standard electrophysiological recording techniques, hybrid network methods that couple biological and model neurons, and computational brute-force explorations of high-dimensional parameter spaces.