Obtaining information from the environment to guide behavior is one of the most fundamental functions of nervous systems. Since individual cues are rarely 100% reliable, many animals (including humans) combine cues across sensory modalities to improve the quality of their decisions and make their behavior more robust under variable circumstances. In mammals, this "multisensory integration" involves many billions of neurons across several densely interconnected brain areas, so to gain an understanding of integration processes at the level of individual neurons and their synaptic connections, simpler animals are a more promising starting point. I will present recent behavioral data on the hunting behavior of the medicinal leech that demonstrate that this animal uses a nonlinear combination of visual and mechanical cues extracted from water waves to locate its prey. It has previously been shown that as leeches mature their prey preferences change, and our data show that their hunting strategy shifts commensurately. Remarkably, this shift is implemented as a change in how cues are integrated across modalities rather than as a change in the responses to individual modalities. The leech is an ideal model for studying the neural basis of behaviors, since its nervous system is readily accessible and comprises a mere 10,000 neurons arranged in 21 nearly identical ganglia plus small head and tail brains. Toward the end of my talk I will discuss some technological advances that allow us to probe sensory processing in this nervous system, and show preliminary results on early visual processing.