Even when simultaneously performing a task unrelated to sounds, the human auditory cortex can precisely model the invariances of the acoustic environment. Data acquired in a mismatch negativity (MMN) paradigm have shown that temporally and spectrally complex sounds as well as their relations are automatically represented in the human auditory cortex. Furthermore, MMN data indicate that these neural sound representations are spatially distinct from phonetic and musical sounds within and between the cerebral hemispheres. Most MMN studies were conducted in pitch dimension, but temporal aspects of sound processing are also under increasing experimentation. To some extent, musical expertise is also reflected in sound representation accuracy as indexed by the MMN paradigm.