A neural model suggesting how speech spectra may be represented by the nervous system

A computer model of nerve cell somata was made using Rall-type cylindrical compartments and Hodgkin-Huxley-like channels for sodium, potassium, and calcium ions. Seven of these modeled neurons, representing three spherical bushy cells in the auditory system, were networked as an input layer, with each connected to the dendrites of three neurons in a middle layer by excitatory synapses. These all synapsed with one common output neuron. The input layer neurons were each stimulated at slightly different frequencies to represent differences in sound intensity in a sound spectrum. Each middle- and output-layer neuron fired at two rates: at the fastest input frequency and with bursts at regular intervals matched to the difference frequency between the input signals. This would then allow a neuron to represent the sound intensity at a tonotopically localized frequency and the intensity of a neighboring frequency as seen in a sound spectrum