Artificial Neural Networks simulation of learning of auditory equivalence classes for vowels

Abstract

In a series of behavioral experiments rats were trained to discriminate between synthetic vowels characterized by an increase in fundamental frequency correlated with an upward shift in formant frequencies. The results demonstrate that rats are able to generalize the discrimination to new instances of the same vowels and that the performance depended on the relation between fundamental and formant frequencies that they had previously been exposed to. Simulation results using artificial neural networks could reproduce most of the behavioral results and suggest that equivalence classes for vowels are associated with an experience-driven process based on general properties of peripheral auditory coding mixed with elementary learning mechanisms.