Representation of Complex Stimuli in the Peripheral Auditory System

Eric Young, Barbara, Calhoun, Jane Yu, Israel Nelkent
unpublished
The neural representation of stimulus spectral shape depends on a tonotopic array of neural activity in which each neuron carries information about stimulus energy at frequencies near its best frequency (BF). In neurons like auditory nerve fibers and most cells in ventral cochlear nucleus, spectraI integration can be approximated by a bandpass filter centered on BF; however, for neurons in the dorsal cochlear nucleus, the spectral integration is strongly dependent on the stimulus spectrum, and
more » ... ulus spectrum, and is different for tones versus noise. Here, methods for characterizing the spectral sensitivity of neurons are discussed and a new method based on random spectral shapes is described. TONOTOPIC ORGANIZATION AND AUDITORY-NERVE TUNING One basis for understanding auditory perception and signal representation in the auditory nervous system is the idea that the stimulus spectrum is represented tonotopically. At the level of single neurons, this means thal each neuron has a best frequency (B~, the frequency to which it is most sensitive; each neuron is then assumed to provide information to the brain mainly about components of the stimu[us at frequencies near the neuron's BF. With this assumption, the neural representation in a population of neurons can be studied by plotting some measure of neural activation against BF. In the simplest model, the degree of neural activation at a particular BF should be monotonically related to the amplitude of the stimulus spectrum at that frequency (16, 18). The tonotopic representation has been studied most extensively in A-20 the auditory nerve (AN), where the assumptions that underlie it are best I realized. The frequency selectivity of an AN fiber can be approximated by its tuning curve (TC; 9), a plot of the sound level required to reach ~ 40-TC \ 7 threshold at different frequencies (Figs. 1A and lB). AN fiber TCS : resemble bandpass filters. A fiber's frequency selectivity can be :-60 +...-"'-'\"': approximated by inverting the TC and using it as a bandpass filter to = select the frequency components to which the fiber will respond. 5 Models of AN fiber responses based on extensions of this idea are ~ '80 : reasonably successful at predicting perceptual phenomena (5). The assumption that the TC is an adequate description of AN-1oo 1 frequency selectivity is challenged by several behaviors of AN fibers at 0.1 0.2 0.5 1 2 5 suprathreshold levels. If tuning is defined as the profile of discharge rate B in response to tones of fixed sound pressure level but varying frequency 100-(iso-level rate contour), then the bandwidth of the [uning broadens as sound level increases and the BF of the tuning may decrease (8), This behavior is consistent with the tuning of basilar membrane motion, ~ measured as response amplitude at fixed input level (e.g. 17), which ~ m 50 broadens and shows a shift of BF as sound level increases. u Responses to complex stimuli often have properties that are not well-explained by either the TC or the iso-level rate contour. An example is provided by responses to speech-like stimuli, in this case a :..V.: /F2 synthetic vowel. Fig. 1C shows spectra of the responses of an AN fiber 12 to the vowel /d. The plots show the frequencies to which the fiber was c phase-locked as peaks; all of these peaks occur at harmonics of 1W Hz, the fundamental frequency of the periodic vowel. The vowel's second
fatcat:7uh2fytcofcwhcmi7fkvifjtpm