The aim of this work is to explore the representation of speech in the anteroventral cochlear nucleus by computational models based on neuroanatomical and neurophysiological data. A computational model of the anteroventral cochlear nucleus stellate cells with chop-S type response properties is described. Input that is excitatory and inhibitory to the model is in the form of simulated auditory-nerve spikes. The model is capable of generating a wide range of realistic chop-S unit responses. We investigate the speech information processing performed by our system that includes a peripheral model (cochlea, hair cells, auditory nerve spikes generation and chop-S units).