A model of song syntax generation in the Bengalese finch

To better understand the song generation processes taking place in the bird’s brain, they developed a spiking neuronal network model. The point of note is that the individual short segments of a song, the syllables, are combined stochastically, i.e. the process obeys clear probabilistic rules.

The model, developed by Alexander Hanuschkin and Abigail Morrison, encodes each syllable as a feed-forward subnet within a larger network. Each feed-forward subnet reliably propagates spiking activity when it is stimulated. Priming signals from an auditory network enable the activity to be switched only to syllables that are permitted by the syntax.

The model not only produces biologically realistic syllable sequences, but also accounts for a wide range of experimental findings, such as the gradual loss of syntax when a bird is deafened, and restoration of song syntax when hearing is restored. The study paves the way for further investigations into the learning and production of other complex action sequences on the basis of structured neuronal activity.

Alexander Hanuschkin has since moved to Zürich to take up a postdoc position in the laboratory of Richard Hahnloser, a world expert on learning and vocal production in songbirds.

Full article (open access)

Hanuschkin, A.; Diesmann, M. & Morrison, A. (2011) A reafferent and feed-forward model of song syntax generation in the Bengalese finch. J Comput Neurosci 31(3)