Analysis of the stationarity of models of auditory-nerve fiber discharge patterns

The spontaneous activity and the responses to single, high-frequency tones of single auditory-nerve fibers are often described as a stationary point processes. Using a doubly stochastic Poisson process model as the analysis framework, it is shown that the instantaneous rate of discharge varies more than is consistent with a stationary model. Variability indices (which equal unity for a stationary Poisson process model) were found to be no smaller than 1.5 and at least as large as 6.9 for the data analyzed. Assuming this variability to be due to the presence of a noise process modulating the instantaneous rate, the spectra of the noise processes were measured. These spectra were low pass, having upper frequencies no greater than about 0.1 Hz. Some spectra may contain spectral lines, indicating the presence of nearly periodic components. The presumed noise process, whether containing periodic components or not, could not be readily related to ambient acoustic noise. Work supported by NINCDS.