A c.:omputational model of song learning in the song sparrow (M elospiza melodia) learns to categorize the different syllables of a song sparrow song and uses this categorization to train itself to reproduce song. The model fills a crucial gap in the computational explanation of birdsong learning by exploring the organization of perception in songbirds. It shows how competitive learning may lead to the organization of a specific nucleus in the bird brain, replicates the song production results of a previous model (Doya and Sejnowski, 1995), and demonstrates how perceptual learning can guide production through reinforcement learning.

The passeriformes or songbirds make up more than half of all bird species and are divided into two groups: the os cines which learn their songs and sub-oscines which do not. Oscines raised in isolation sing degraded species typical songs similar to wild song. Deafened oscines sing completely degraded songs (Konishi, 1965), while deafened sub-oscines develop normal songs (Kroodsma and Konishi, 1991) indicating that auditory feedback is crucial in oscine song learning. Innate structures in the bird brain regulate song learning.