Perception of timbre dimensions: Psychophysics and electrophysiology in humans.

Abstract

Timbre is a perceptual attribute that characterizes the identity of a sound source. Psychoacoustical experiments have revealed that this perceptual attribute is multidimensional, with timbre dimensions (such as sharpness or brightness) corresponding to a variety of acoustical parameters: attack time, spectral centroid, spectrum fine structure, spectral flux, etc. The question arises whether timbral dimensions are processed holistically or separately in the auditory system. This issue was addressed in a series of behavioral and event-related potential (ERP) studies in humans [Caclin et al., J. Acoust. Soc. Am. (2005); J. Cogn Neurosci. (2006); J. Cogn Neurosci. (2008); Brain Res. (2007)]. Different timbres were created by manipulating attack time, spectral centroid, and spectrum fine structure. Dissimilarity ratings confirmed that these three parameters were indeed major determinants of timbre. Their representation in auditory sensory memory was explored using the mismatch negativity (MMN) component of the auditory ERP. MMN results indicated mainly separate sensory memory representations of these dimensions in auditory cortex. Using Garner interference paradigm, evidence was found for an interactive processing of timbre dimensions at other levels of analysis than sensory memory, including early perceptual processing stages. These results can be synthesized in a model of timbre perception postulating separate channels of processing for different timbre dimensions, with cross-talk between those channels.