On cochlear encoding: Potentialities and limitations of the reverse-correlation technique

Abstract

This paper presents a description of the interrelation between two major properties of the responses recordable from auditory nerve fibers:f r e q u e n c y s e l e c t i v i t y and p a r t i a l s y n c h r o n y between stimulus and response. In the course of this work the influence of nonlinearity on the cochlear encoding process can be assessed. The theory of the r e v e r s e‐c o r r e l a t i o n t e c h n i q u e is derived in a most general way. It is based on a model in which a filter—assumed to be linear—is followed by a stochastic pulse generator—the probability of producing an output pulse being an instantaneous but nonlinear function of its input signal. Insofar as such a model represents stimulus transformations in a primary auditory neuron, the technique can be applied to the responses recorded from an auditory nerve fiber. Several illustrative examples of experimental reverse‐correlation functions−abbreviated: r e v c o r f u n c t i o n s—are presented and discussed. These functions have the general character of impulse responses of sharp bandpass filters. They show very little phase modulation. For noise stimuli of up to 70 dB per third octave the revcor functions are almost invariant. Above that level some (but not all) of the revcor functions show a loss of frequency selectivity. If a nerve fiber can be contacted for a sufficiently long time, it is possible to compare the response with that of a model filter, in which the revcor function of that fiber is substituted as its impulse response. The output signal of the model filter is shown to be a very good predictor of the firing probability of the fiber under study. This property is demonstrated for noise as well as for tone stimuli. There is surprisingly little evidence of nonlinear filtering in these results. This so‐called simulation method can also be applied when the stimulus is switched on and off. The results show, apart from effects due to filtering, clear manifestations of fast adaptation. Again, the filtering appears to be independent of the latter effect. It is concluded that for wide‐band noise and single‐tone signals the firing probability is predominantly controlled by a linearly filtered version of the acoustical stimulus; this constitutes the principle of s p e c i f i c c o d i n g. The conspicuous absence of nonlinear effects in the results can partly be explained in terms of the response properties of a class of networks in which sharp filtering occurs after the generation of nonlinear distortion products. It can then be predicted that this property will hold only for wide‐band and tonal stimuli. That our results show so little evidence of cochlear distortion appears to be a property of signal transformations and is not due to linearization tendencies of the experimental method.