Reading Between the Spikes: Real-Time Signal Processing in Neural Systems

This thesis discusses biological strategies for real-time signal processing in neural systems. Nearly all creatures encode information about the world as patterns of identically shaped action potentials, or "spikes". As a result, all the animal's knowledge of the world is contained in the occurrence times of these discrete events. Traditional approaches to the study of neural coding emphasize the encoding process, resulting in predictions of average neural responses to a limited class of stimuli. However, these studies fail to address the relevant biological question: What can the organism "learn" about the outside world from real-time observations of its own spike trains? Therefore, this thesis approaches neural coding from the point of view of the organism itself: We learn to decode neural spike trains to obtain real-time estimates of sensory stimuli. In particular, this ability to extract continuous signals from spiking cells, together with the definition of an equivalent spectral noise level for a spiking neuron allows characterization of the information contained in patterns of neural response as well as forming the basis for the prediction of optimal neural computation strategies with spike trains. These methods are applied to the design and analysis of experiments on a single wide field, movement -sensitive neuron (H1) in the visual system of the blowfly Calliphora erythrocephela and to the filiform hair receptors of the wind-sensing system of the cricket Acheta domestica. This thesis also discusses the generalization of these strategies to collections of neurons and the applications to future work in the context of neural computation in the retina.