Whereas in vision a large amount of information may in theory be extracted from instantaneous images, sound exists only in its temporal extent, and most of its information is contained in the pattern of changes over time. The “echoic memory” is a pre-attentive auditory sensory store in which sounds are apparently retained in full temporal detail for a period of a few seconds. From the long-latency auditory evoked potentials to spectro-temporal modulation of complex harmonic tones, at least two automatic sound analysis processes can be identified whose time constants suggest participation of the echoic memory. When a steady tone changes its pitch or timbre, “change-type” CP1, CN1 and CP2 potentials are maximally recorded near the vertex. These potentials appear to reflect a process concerned with the distribution of sound energy across the frequency spectrum. When, on the other hand, changes occur in the temporal pattern of tones (in which individual pitch changes are occurring at a rate sufficiently rapid for the C-potentials to be refractory), a large mismatch negativity (or MN1) and following positivity (MP2) are generated. The amplitude of these potentials is influenced by the degree of regularity of the pattern, larger responses being generated to a “deviant” tone when the pitch and time of occurrence of the “standards” are fully specified by the preceding pattern. At the sudden cessation of changes, on resumption of a steady pitch, a mismatch response is generated whose latency is determined with high precision (in the order of a few milliseconds) by the anticipated time of the next change, which did not in fact occur. The mismatch process, therefore, functions as a spectro-temporal auditory pattern analyser, whose consequences are manifested each time the pattern changes. Since calibration of the passage of time is essential for all conscious and subconscious behaviour, is it possible that some states of unconsciousness may be directly due to disruption of internal “clocks”? Abnormal mismatch potentials may provide a manifestation of a disordered auditory time-sense, sometimes being abolished in comatose patients while the C-potentials and similar responses to the onset of tones are preserved. Both C- and M-potentials were usually found to be preserved, however, in patients who had emerged from coma and were capable of discriminating sounds. Substantially intact responses were also recorded from three patients who were functionally in a “vegetative” state. The C- and M-potentials were once again dissociated in a group of patients with multiple sclerosis, only the mismatch potentials being found to be significantly delayed. This subclinical impairment of a memory-based process responsible for the detection of change in temporal sound patterns may be related to defects in other memory domains such as working memory.