Biology Department
The City College of CUNY
10/1/03
Young zebra finches can be trained to imitate a song in a controlled environment, where the entire vocal development can be recorded and analyzes. I will present methods for representing the learning process by tracking trajectories of vocal changes. The entire learning process is then presented as dynamic maps (movie clips) showing how features of sounds form clusters (or types). This approach allows us to uncover specific methods the bird uses to learn to perform complex syllables and match their syntax to the model. I will present some technical issues of analyzing large databases of sounds and then discuss the following phenomena:
1. Vocal learning has strong circadian oscillations, with sharp
deterioration in the performance of newly learned sounds after
night-sleep.
2. Trajectory of vocal changes are not independent and non-monotonic
(that is, learning trajectories are indirect and subject to
constraints)
3. The details of transition from subsong to mature song have some
similarities with transition from babbling to speech.
4. The effect of age on vocal learning is not generic: some
performances improve with age and other deteriorates with age.
Here too, there are some interesting similarities with speech
development.