Representation of auditory space in human cortex

Auditory space poses a difficult computational challenge to the nervous system. The localization of a sound source is based on the extraction of cues embedded in a neural representation organized according to sound frequency. Single-neuron studies on the neural representation of space and the computations leading to it have been performed on animals. This has given rise to two alternative models of auditory spatial representation: a place code consisting of narrow spatial receptive fields and a hemifield code formed by neurons tuned widely to the left or to the right. The aim of this thesis was to reveal which of these codes explains the representation of auditory space in human cortex. Predictions based on the place and the hemifield code were tested in a series of magnetoencephalography (MEG) experiments utilizing a stimulus-specific adaptation paradigm. The pattern of location-specific adaptation of brain responses found for realistic spatial sound stimuli closely followed that predicted by the hemifield code. Further, results consistent with the hemifield code were found also with sound containing only the interaural time difference cue for which place coding has long been assumed to apply. The right hemisphere contained more neurons tuned to the left than to the right hemifield whereas such asymmetries did not occur in the left hemisphere. Cortical activity was found in parietal and frontal areas but only after the presentation of a target stimulus requiring an active response. The implications of wide neural tuning for sound discrimination were explored in a neural network model. The best discrimination power of neurons was found to be related to the slopes of the tuning curves which in the hemifield code coincide with frontal sound source directions that are optimally localized by human listeners. In conclusion, the results support a hemifield code representation of sound source location in human cortex formed by two populations of neurons: one tuned to the left and the other to the right hemifield. Further, the present studies provide an encouraging example on how theories originating from studies of single-neuron tuning properties can be tested with methods available for the study of human brain function at the mass-action level