Cortical codes for sound localization

Abstract

Abstract:It is well known that there is a point-to-point map of auditory space in the midbrain: each neuron is tuned to a particular sound-source location, and neurons' preferred locations are topographically represented in a neural structure. In the auditory cortex, however, researchers have consistently failed to demonstrate evidence for such an auditory space map, despite the well-known necessity of the auditory cortex for normal sound localization. Cortical neurons show generally broad spatial tuning, and the preferred locations are not systematically organized on the cortex in a topographical fashion. An alternative hypothesis is presented here: Individual single neurons represent auditory space panoramically by space-specific characteristics of their spike patterns. Information about any particular sound-source location is distributed across a large population of neurons, and we predict accurate localization judgement by combining information across those neurons. In our analyses of experimental data using an artificial neural network algorithm, we were able to recognize spike patterns of single neurons to identify sound-source locations throughout 360 • of space. The amount of information carried by a moderate size of neural ensemble appeared sufficient to account for the accuracy of location judgements by behaving animals.