Being able to detect unusual, possibly dangerous events in the environment is a fundamental ability that helps ensure the survival of biological organisms. Novelty detection requires a memory system that models (builds neural representations of) events in the environment, so that changes are detected because they violate the predictions of the model. The earliest physiologically measurable brain response to novel auditory stimuli is the mismatch negativity, MMN, a component of the event-related potential. It is elicited when a predictable series of unvarying stimuli is unexpectedly followed by a deviating stimulus. As the occurrence of MMN is not usually affected by the direction of attention, MMN reflects the operation of automatic sensory (echoic) memory, the earliest memory system that builds traces of the acoustic environment against which new stimuli can be compared. The dependence of attentive novelty detection on earlier, pre-attentive processes, however, has remained elusive. Previous, related studies seem to suggest a relationship between MMN and attentive processes, although no conclusive evidence has so far been shown. Here we address novelty detection in humans both on a physiological and behavioural level, and show how attentive novelty detection is governed by a pre-attentive sensory memory mechanism.

Studies of human auditory and somatosensory modalities have shown that there is an oscillatory response in the gamma-band (at about 40 Hz) frequency which is elicited by either steady state or transient stimulation. The auditory 40-Hz response is generated at least partially in the auditory cortex as a result of thalamocortical interaction and may serve perceptual integration and conscious perception. A connection to selective attention has been implied in human and animal studies, although the evidence is inconclusive. Moreover, fundamental differences between the human and animal 40-Hz responses prohibit generalization. Furthermore, most experiments have used steady-state stimulation during which the brain does not regain its resting state between stimuli as it does when transient stimulation is used. Here we study the effect of selective attention on the auditory gamma-band (40-Hz) transient response using subjects listening to tone pips presented in one ear while ignoring a concurrent sequence of tone pips in the other ear. The 40-Hz response was larger when subjects paid attention to stimuli rather than ignored them. This attention effect was most pronounced over the frontal and central scalp areas. Our results demonstrate a physiological correlate of selective attention in the 40-Hz transient response in humans.

The central auditory system for event perception involves the integrating mechanism of sequential information addressed by the present study. The mismatch negativity (MMN) component of the event-related potentials (ERP) reflects the automatic detection of sound change. ERPs to occasionally omitted stimuli were measured when sequences with constant stimulus onset asynchronies (SOAs) were presented. In separate blocks, the SOA was from 100 to 350 ms. A clear MMN was elicited by a stimulus omission in a sequence of regularly spaced tone pips only when the SOA was shorter than 150 ms, yielding an estimate for the duration of the temporal window of integration used the perceptual segregation of auditory events.

The present study analyzed the neural correlates of acoustic stimulus representation in echoic sensory memory. The neural traces of auditory sensory memory were indirectly studied by using the mismatch negativity (MMN), an event-related potential component elicited by a change in a repetitive sound. The MMN is assumed to reflect change detection in a comparison process between the sensory input from a deviant stimulus and the neural representation of repetitive stimuli in echoic memory. The scalp topographies of the MMNs elicited by pure tones deviating from standard tones by either frequency, intensity, or duration varied according to the type of stimulus deviance, indicating that the MMNs for different attributes originate, at least in part, from distinct neural populations in the auditory cortex. This result was supported by dipole-model analysis. If the MMN generator process occurs where the stimulus information is stored, these findings strongly suggest that the frequency, intensity, and duration of acoustic stimuli have a separate neural representation in sensory memory.

No abstract