Understanding the temporal dynamics underlying cortical processing of auditory categories is complicated by difficulties in equating temporal and spectral features across stimulus classes. In the present magnetoencephalography (MEG) study, female voices and cat sounds were filtered so as to match in most of their acoustic properties, and the respective auditory evoked responses were investigated with a paradigm that allowed us to examine auditory cortical processing of two natural sound categories beyond the physical make-up of the stimuli. Three cat or human voice sounds were first presented to establish a categorical context. Subsequently, a probe sound that was congruent, incongruent, or ambiguous to this context, was presented. As an index of a categorical mismatch, MEG responses to incongruent sounds were stronger than the responses to congruent sounds at ~250ms in the right temporoparietal cortex, regardless of the sound category. Furthermore, probe sounds that could not be unambiguously attributed to any of the two categories ("cat" or "voice") evoked stronger responses after the voice than cat context at 200-250ms, suggesting a stronger contextual effect for human voices. Our results suggest that categorical templates for human and animal vocalizations are established at ~250ms in the right temporoparietal cortex, likely reflecting continuous on-line analysis of spectral stimulus features during auditory categorizing task.