Auditory scene analysis (ASA) is one of the most important, and least well-defined, tasks of audition. I will review a number of experimental paradigms that can be interpreted as evidence for ASA in the auditory system. Modeling ASA requires a combination of low-level and high-level processes. While we understand to a degree the low-level processes, and we have a suprisingly large arsenal of subcortical processing mechanisms that may underlie them, the high-level end of ASA is much more mysterious, and we have only a rough idea of what the end product of ASA, the auditory object, is. I will argue that part of the confusion in understanding ASA is the result of confounding the physical cues for ASA with the auditory object. Based on insufficient amount of evidence, I will argue that (i) auditory objects are predictive models of the incoming sound sequence, (ii) they are expressed in neuronal responses at the level of primary auditory cortex, and (iii) auditory objects are created before their properties are assigned to them. In this view of ASA, primary auditory cortex serves as a mid-level processor that creates auditory objects, while higher-order fields represents the properties of the resulting objects.
BACK
