Principal Investigator Michale Fee
Sensitivity to temporal sequence is a striking and nearly universal aspect of brain function -- not only at a sensory level, but also at a motor and cognitive level. The ability of the brain to step rapidly through a learned sequence of states underlies not only the performance of complex motor tasks such as speech, but perhaps our ability to think and plan as well. Despite the fundamental significance of temporal ordering in animal behavior, little is known about the biophysical and circuit mechanisms underlying the generation, learning and detection of complex sequences in the brain. Animal vocalizations provide a marvelous example of these phenomena, and we are using the songbird as an experimental system to explore detailed models of neural sequence generation. Most songbirds, such as the zebra finch, produce a stereotyped pattern of acoustic signals with structure and modulation over a wide range of time-scales, from milliseconds to several seconds. Another remarkable aspect of this behavior is that the specific acoustic pattern produced by a songbird is learned, rather than being innately controlled: Vocalizations are learned from the parents through a series of well-defined stages. Moreover, avian brain areas involved in song learning are closely homologous to mammalian brain areas involved in motor learning. Thus, the song control system may have a great deal to teach us about general principles of sequence generation and learning in the vertebrate brain.
Recent Projects in the Songbird -- For the past six years, our laboratory has focused on studying the cellular, circuit, and mechanical underpinnings of songbird vocalizations. In one current project, we are studying nucleus RA, an area that projects directly to motor neurons of the vocal organ. During song, RA neurons each generate a distinctive and reproducible sequence of brief bursts of spikes. Using a new miniature motorized microdrive developed in this lab (see below), we have been able to record from large populations of RA neurons (~50) in the singing bird to understand how premotor activity maps to vocal output.
We are also investigating how the complex sequence of bursts in RA is generated. The primary motor-related input to RA comes from nucleus HVC; previous studies in singing birds found that HVC neurons generate relatively stochastic, dense patterns of spikes. Furthermore, the relationship of these firing patterns to the song suggested that the temporal control of song is organized hierarchically, such that HVC codes for motor patterns on the timescale of song syllables (50-100 ms), and RA codes for motor and acoustic patterns on a much briefer timescale (~10 ms). Thus, the prevailing view in the songbird field has been that burst patterns in RA are largely generated by circuitry intrinsic to RA.
One difficulty with these earlier experiments is that HVC is known to contain at least three classes of neurons: neurons that project to RA, neurons that project to area X (a brain area involved in vocal learning), and local interneurons. We were the first to characterize the firing patterns of antidromically identified HVC neurons in the singing bird, and have shown that RA-projecting HVC neurons are extremely sparsely active - each generating at most a single brief burst of spikes precisely at one time in the song motif. We also found that these neurons burst sequentially, with each neuron bursting at a different time in the song.
What is the causal relationship between the sparse bursts of HVC(RA) neurons and the bursts in downstream nucleus RA? Do bursts in HVC(RA) neurons drive bursts in RA, and if so, is every burst in RA driven from HVC? We are examining these questions in a new head-fixed sleeping bird preparation that permits sophisticated electrophysiological experiments that would be difficult or impossible to do in the freely behaving bird. Recent observations have shown that during sleep, RA neurons generate spontaneous high-frequency bursts, and that the sequence of sleep burst produced by an individual RA neuron can closely resemble the burst sequence of that neuron during singing. We have shown that HVC neurons similarly replay brief snippets of highly sparse song-like sequences during sleep. Furthermore, we have shown that during sleep, nearly every RA burst is driven directly by a small population of HVC(RA). The similarity of song- and sleep- related burst patterns suggest that song-related bursts in RA are likewise directly driven from HVC.