We have described earlier how the multiscale spectro-temporal representations of the auditory cortex provide for the analysis and extraction of perceptually significant features, and how these might in turn prove useful in a variety of applications involving detection and recognition of acoustic signals ( Thrust areas I(A), II, V(B)). The physiological basis for this representation in the primary auditory cortex (AI) is by now well established, especially the two most important response properties: linearity and separability [.Shamma rip2 1995, Shamma mov1 1996.]. What remains unknown, however, is: (1) to what extent can linearity and separability be used to account for the representation of all dynamic spectra such as speech? (2) How does this organization arise in earlier auditory structures such as the inferior colliculus (IC), and how is it further developed in secondary cortical areas such as the anterior auditory field (AAF)? (3) And why does the system behave in a substantially linear manner despite the numerous and obvious response nonlinearities?
These questions are addressed in the following proposed experiments which employ a unique methodology founded on the observation that the auditory responses are substantially linear with respect to the stimulus spectral envelope and its temporal dynamics [.shamma rip2 1995, shamma mov2 1996.]. Specifically, the experiments seek to accomplish three objectives: (a) Predict single-unit responses in AI to speech and other complex spectra. This will be the culmination of a series of experiments that build upon our previous experimental and theoretical findings with ripple stimuli in AI [.shamma rip2 1995, shamma mov2 1996.]. (b) Extend the applicability of the ripple stimuli and their data analysis to the responses of the AAF and ICC. Our goal is to describe the functional organization of these structures based on the spectral and temporal parameters of their responses to ripple stimuli. (c) Delineate the origin and extent of nonlinearities in responses to narrowband vs. broadband spectra (e.g., tonal vs. ripple stimuli). These nonlinearities include threshold, half-wave rectification, saturation, and adaptation.
The results of these investigations will enhance our understanding of the encoding of complex acoustic spectra such as speech and music in the auditory system. More generally, however, they will place auditory cortical processing within the larger framework of visual and other sensory processing in the brain.