The neural basis of auditory perception, including speech perception, has typically been studied using standard neuropsychological paradigms (e.g. deficit-lesion correlation). The development of the functional imaging methods (PET, fMRI, MEG) has made it possible to study such cognitive processes in vivo. Although these methods all provide important physiological data, magnetoencephalography (MEG) is particularly suited to study auditory processing. MEG has the appropriate (ms) temporal resolving power and is especially sensitive to neuronal activity in areas such as the supratemporal plane. Magnetic source imaging (MSI) amounts to coregistering MEG data with high-resolution magnetic resonance images to determine the location of auditory-evoked electromagnetic activity. We use pure tone and band-pass noise stimuli as well as synthesized speech stimuli and record the auditory evoked neuromagnetic responses with a large-array biomagnetometer. One illustrative result based on pure tone presentation is discussed: The dominant auditory evoked neuromagnetic field (M100) generated by tone stimuli, which has been used in modeling spatial tonotopy, is frequency-dependent. That is, the characteristic evoked field of the recruited neuronal population reaches a peak at a latency that depends on the frequency of the stimulus. The latency varies parabolically with frequency, with low (100-500Hz) and high (3-5kHz) frequencies yielding longer latencies. This suggests that, in addition to spatial tonotopy, there may be a frequency coding mechanism that relies on temporal response properties.