Michael Brainard, PhD

Sensorimotor Learning in the Song System

Research in my laboratory focuses primarily on the question of how experience, particularly during early life, shapes the functioning of the nervous system. To approach this question we are currently using a combination of behavioral and neurophysiological techniques to investigate the mechanisms underlying vocal learning in songbirds.

The study of song learning offers the advantages of a well described behavior that exhibits a variety of general features of learning, and that is subserved by a discrete and extensively investigated set of brain regions. Because many species of songbirds breed well in captivity and develop rapidly, they are well suited for studying processes of developmental plasticity. Song learning proceeds in two stages. First, during a period of sensory learning, young birds listen to and memorize the song of an adult 'tutor'. Then, during a period of sensorimotor learning, they use auditory feedback to gradually refine their own initially rambling vocalizations so that they progressively resemble the previously memorized tutor song. Normal song learning requires appropriate experience during a sensitive period in early development, although recent studies have shown that auditory feedback also contributes to the adult maintenance of precisely calibrated vocal output. These features make song learning a useful model for studying the mechanisms that contribute to vertebrate sensory and sensorimotor learning in general, and to certain components of human language learning in particular; speech acquisition exhibits strikingly similar requirements for memorization and vocal practice during early development and for maintained auditory feedback throughout life.

Current Projects

Contributions of auditory feedback to song learning and maintenance
Prior experiments combining behavioral analysis of song with targeted lesions of song system nuceli suggest that signals arising from the songbird basal ganglia (the 'anterior forebrain pathway' or AFP) are necessary throughout life for the feedback-based modification of song (Brainard and Doupe 2000). Current experiments using feedback alteration, neural recording and microstimulation are designed to further test the hypothesis that this basal-ganglia circuit provides an error signal that reflects the quality of match between a bird's own vocalizations and the memorized tutor song.

Feedback alteration. To achieve a controlled and reversible alteration of auditory feedback, we have developed a computerized system that detects song elements (as the bird is singing) and generates sounds that are superimposed on, and hence alter, the normal auditory feedback that the bird experiences. This manipulation can cause acute changes to ongoing song production (such as alterations to the timing or sequencing of song elements). This indicates that auditory feedback influences vocal production on a moment by moment basis, as it does in humans. Extended exposure to feedback altered in this manner can also lead to persistent changes in song, indicating that altered feedback elicits neural signals that are capable of driving plastic changes in the song motor program. Ongoing experiments are designed to further assess the capacity of the song system to adapt to systematic alterations of auditory feedback.

Chronic neural recording. To characterize the location and nature of neural signals generated by alteration of auditory feedback, we are using chronic extracellular recordings from awake singing birds during conditions of normal and altered feedback. Initial experiments are focussing on the AFP, although this technique will also enable tracing the transformation of feedback at earlier stages of auditory processing.

Microstimulation. Neural signals elicited by feedback alteration could in principle participate in instructing changes in the motor pathway. Artificial introduction of patterned activity to the nervous system, by microstimulation, will allow testing the sufficiency of such signals to drive changes in the motor program for song. We are currently assessing the influence on vocal production of song-triggered microstimulation of the AFP. Output from this pathway converges on motor neurons within the pathway for vocal production. Because song structure is normally very stereotyped, behavioral analysis will enable detection both of acute and lasting effects of altering neural activity on the motor program for song.

Developmental regulation of plasticity
Sensory learning of song occurs during a well characterized 'sensitive period' beyond which the nervous system is progressively less influenced by exposure to other song models. Similarly, in the final stages of sensorimotor learning there is a decline in the susceptibility of song production to disruption by altered experience. Evidence for this includes experiments that have shown that the degree to which song deteriorates following loss of auditory feedback decreases progressively with age (Brainard and Doupe 2001). As we explore the mechanisms underlying sensory and sensorimotor learning, a general question of interest is how these mechanisms are developmentally altered so that the nervous system becomes less susceptible to the influence of experience.

Reinforcement learning
In addition to the critical role of auditory experience during song learning, non-auditory factors, especially arising from social interactions, help shape what songs are memorized and produced. For example, male birds with a variable repertoire of songs will preferentially retain those songs that elicit courtship displays from females. This indicates that rather global reinforcement signals (in this case dependent on visual cues from the female) can shape subsequent vocal production. Similar reinforcement signals play a widespread role in learning, and indeed refinement of song based on feedback evaluation may depend in part on global reinforcement signals that indicate quality of match between actual vocal production and a desired perceptual target. In other vertebrates, reinforcement learning is mediated in part by dopaminergic pathways originating from the midbrain. In songbirds, these dopaminergic pathways project heavily to the AFP and other structures of the song system, and are thus well situated to provide signals that could modulate or guide song learning. We are preparing to explore how these pathways contribute to different stages of vocal motor learning using targeted neural recording and microstimulation.

Lab Members

Lab Website