Mismatch negativity for item rather than serial-order information in a 150-ms tone series that is not repeated as a...

Speech, for example, consists of fast-paced (>5/s) sounds in specific spectrotemporal patterns. Humans are generally held to be able to represent not only such sounds themselves (item information) but also their serial order (order-information) as a repeated melody with ease, as suggested by studies on the mismatch negativity (MMN) of event-related potentials (ERPs). The present study tested whether this ability tolerates the absence of the support of melodic repetitiveness. ERPs were recorded from adult humans presented with rare 150-ms series of three 50-ms tones ('deviants') interspersed with frequently repeated ones ('standards'). The frequency of each tone was pseudorandomly one of four alternative frequencies. The series were of type 'AAB' (two tones of one frequency followed by a tone of another frequency), 'ABB' (two tones of one frequency preceded by a tone of another frequency), or 'AAA' (three tones of one frequency). The MMN was robustly elicited by AAA deviants against AAB standards. It was, however, less distinct for ABB deviants against AAB standards and even statistically non-significant for AAB deviants against ABB standards. MMN generation in the human brain thus seems to be based on item rather than serial-order information in a rapid spectro-temporal pattern of acoustic signals that is not repeated frequently in the short term.