The neural mechanism involved in the analysis of complex sounds was studied. By using a given cat's voice, the responses of the neurons at various levels of the auditory pathway to a cat's voice were analyzed with regard to the time patterns and the probability of firing by using a band-pass or band-rejection filter at particular center frequencies. In the cochlear nerve fibers, the responses to a cat's voice were a simple sum of the responses to the filtered components of the cat's voice. No mutual inhibitory interaction due to the coexisting frequency components in the vocalization was found at this level. However, the mutual inhibitory interaction was clearly found in the cochlear nucleus and became more evident at higher auditory centers reaching up to the medial geniculate body. The degree of this inhibitory interaction was different from neuron to neuron. Several collicular auditory neurons did not respond to a cat's voice, but did to filtered components. In an extraordinary case, a similar neuron was also found in the cochlear nucleus.
The neural process of the feature extraction from complex sounds is attributed to the synaptic mechanism of mutual inhibitory interactions. On the other hand, the majority of neurons in AI of the cortex showed an integrative action.