Mechanisms of Vocal Learning in Songbirds: how song syllable structure is learned

by Primoz Ravbar
Advisor: Professor Ofer Tchernichovski,
Co-advisors: Professor Lucas C. Parra and Professor David S. Vicario

When learning to perform continuous actions one needs to cope with conflicting motor requirements: while some parts of the action may require exploratory variability to find motor states that can efficiently produce a desired outcome, other parts may require consolidation if they are already close to the desired goal. A possible solution to this problem is partitioning the action into segments that could be controlled independently, but it is not known if exploratory variability can be locally regulated during the sensory-motor learning of a continuous action. This dissertation explores song learning in songbirds as a model of learning to perform continuous actions. In the first set of experiments we manipulated song learning in zebra finches (Taeniopygia guttata) to experimentally control different requirements for vocal exploration in different parts of the song. We first trained birds to perform a one-syllable song, and once they mastered it we added a new syllable to the song model. Remarkably, when practicing the modified song, birds rapidly alternated between high and low acoustic variability to confine vocal exploration to the newly added syllable. Analysis of exploratory variability within syllables revealed that acoustic variability changed independently across song elements that were only milliseconds apart. The variability of each song element decreased as it approached the target, correlating with momentary local error and less so with the global error (the mean of local errors across the syllable). We concluded that vocal error is computed locally in sub-syllabic time scales and that song elements can be learned and consolidated independently. However, our evidence for partitions singing behavior into segments that may be learned independently, also indicate  that those segments are not developmentally stable. For example, we observed that the appearance of distinct syllable types usually precedes the appearance of distinct sub-syllabic song elements, suggesting that the time scales at which syllables are learned might get shorter over development, perhaps in a hierarchical manner, from the learning of course temporal structure to the learning of fine temporal structure. To test this hypothesis we studied song learning at the articulatory level by measuring the control of respiratory pressure during song learning. We found that during early song learning, respiratory pressure patterns were coarse, and fine temporal structure was later added to that pre-existent coarse structure. This effect was not observed in socially isolated birds that developed their song without learning a tutored model. We therefore propose that the learning of continuous singing action is achieved by first partitioning the song into coarse segments, and then to finer units, where exploratory variability in each unit is dynamically gated until a local match to the song model is achieved.


Chapter 1: Introduction

Chapter 2: Vocal exploration is locally regulated during song learning

Chapter 3: Regulation of vocal exploration during continuous actions

Chapter 4: How the developing song is compared to the song model

Chapter 5: Song learning at the articulatory level: the development of respiratory pressure patterns

Chapter 6: Learning of some parts of continuous action may affect other parts

Chapter 7: Integration and Discussion