Last modified: 2014-04-01
Abstract
Background:
Decision-making models are commonly based on reinforcement learning: reward promotes action, punishment inhibits it. This fails to explain diversity of rhythmic, innate behaviours. One part of innate courtship in the fruit fly Drosophila Melanogaster is courtship song. Song consists of two modes: pulse and sine song. Switches between song modes seem random and cannot be explained by reinforcement.
We are the first to develop and present a decision-making model of fly song.
Description of model:
Our model includes two coupled oscillators. They are driven by one common source. These assumptions conform with physiological findings. Sine and pulse are generated by different sets of muscle though activation of one neuron activates both sine and pulse generation. We assume the decision on which song is sung to take place in the periphery, rather than centrally in the brain. The model consists of two levels: the level of decision which song to sing, and the level of song generation. Song is generated by fast wing movements – a pendulum motion. The sound wave corresponds to the velocity of the wing.
Results:
Accurate simulation of fly song is achieved twofold: the diverse structure of fly song is mimicked and the distinct shape of song bouts is rendered. Analysis of stability indicates song production to be sensitive to peripheral modulation of muscle contractibility.
Conclusion:
The proposed fly song model represents a simple model of decision-making for innate, stereotyped behaviours. Potentially, parameters can be linked to physiological measures. This will be tested with targeted excitation and depression of song neurons.