| Stimulus space in the olfactory modality is intrinsically high-dimensional. In contrast to retinotopy or cochlear frequency tuning, for example, the similarity of odor stimuli cannot be reliably represented in the brain by the physical proximity of responsive neurons. Hence, nearest-neighbor lateral inhibition, an established mechanism for similarity-dependent computations in other sensory systems, is incapable of mediating such transformations (such as the regulation of receptive field breadth) in olfaction. Instead, the olfactory bulb exhibits a novel winner-take-most neural circuit architecture utilizing local feed-forward inhibition and global feedback inhibition to generate the necessary competitive interactions among glomeruli. This “non-topographical contrast enhancement” or “self-surround” mechanism is independent of the physical proximity of olfactory bulb glomeruli, and innately distributes inhibition among second-order olfactory principal neurons (mitral cells) according to the similarities in their receptive fields. Consequently, it is robust to changes in the statistical structure of the odor environment over time, and greatly simplifies circuit models of olfactory learning. Both olfactory learning and neuromodulatory inputs to the olfactory bulb regulate behavioral odor generalization and the receptive fields of higher-order olfactory neurons, providing a reduced system for the study of complex learning in neural circuits. I will introduce the neural circuitry of the olfactory bulb and illustrate this sensory processing mechanism using studies of olfactory learning and behavior, electrophysiology, and computational modeling. |