Development of precise neural control during vocal learning in the juvenile songbird Restricted; Files Only

Abstract

Motor skills that we learn early in life (e.g. talking or reaching) are essential to everyday function. Skilled behavior relies on precisely timed signals from the brain to coordinate the activation of muscles. Through repeated practice, the brain reorganizes these signals through sensorimotor feedback to refine behavior until consistent high levels of performance is achieved. While the anatomical, molecular, and neural circuitry changes that accompany skill learning have been the focus of much prior work, how neural activity is transformed to refine behavior has been less examined. This is particularly understudied during development, when changing activity patterns occur with the maturation of a developing animal’s muscles and biomechanics –potentially altering the impact of patterns on motor output. My dissertation, therefore, focuses on two questions: how are neural patterns changed across developmental learning (“neural vocabulary”), and with what level of temporal precision do these patterns transduce changes in behavior (“motor code”)? To answer these questions, I leveraged songbirds’ unique experimental tractability: Juvenile songbirds refine their song through sensorimotor learning across weeks. Using prior knowledge of their well-characterized song circuitry, I longitudinally recorded spiking activity from neurons in the vocal motor cortex (RA) across learning. I then related the rendition-by-rendition variations in spiking patterns to acoustic behavior, applying information-theoretic methods sensitive to the precise spike timing at different temporal resolutions from millisecond-scale timing to spike count. This work contributes three main insights: 1) learning constitutes large turnovers in spiking repertoire, suggested by gross changes in temporal structure; 2) motor coding is on the millisecond-scale from even the earliest periods of sensorimotor learning and persists through adulthood; 3) song acquisition is a process of reducing repertoire diversity and selecting patterns that likely tune descending signals to the mechanics of the vocal apparatus. Altogether, this work is the first to demonstrate in any species millisecond-scale motor coding during skill acquisition. It proposes a mechanism for developmental skill learning by which behavioral sensitivity to millisecond-precise neural variations is leveraged to achieve reliable, precise motor control. Broadly, these results will provide a framework for investigating how transformations in spike patterns drive learning across different species. 

Table of Contents

CHAPTER 1: INTRODUCTION 1

1.1 How does the nervous system organize activity to enable skill learning? 1

1.2 The neuroarchitecture of skill learning and the role of motor cortex 3

1.2.1 Descending control and distributed regional contributions 3

1.2.2 Motor cortical contributions to skill learning 5

1.3 Temporal precision of motor control during learning 8

1.3.1 From rate-based views to spike timing-sensitive control 8

1.3.2 Unexplored question on the timescale of control during learning 10

1.4 Birdsong: a tractable system to study skill learning 11

1.4.1 Birdsong learning 11

1.4.2 Circuits that parallel vertebrate motor systems 11

1.4.3 RA as vocal motor cortex: timescale considerations of acoustic control 13

1.5 Goals of this Dissertation 15

CHAPTER 2: BACKGROUND AND RATIONALE 17

2.1 Neural control of behaviors during developmental skill acquisition: Gaps in knowledge 17

2.1.1 Central question framed by a theory of motor learning 17

2.1.2 Gaps in knowledge 19

2.2 Skilled motor control in non-bird vertebrate systems 20

2.2.1 Motor coding during adult skill learning: timescale and trial-averaging considerations 21

2.2.2 Subcortical Loops as Tutors for Motor Cortex 24

2.2.3 Developmental perspectives on neural activity 28

2.3 Skilled motor control in songbirds 31

2.3.1 Songbird and mammalian motor analogies 31

2.3.2 Timescale considerations in songbird RA control of syllable acoustics 32

2.3.3 Behavioral selection and circuitry 35

2.3.4 Developmental changes in neural activity 37

2.4 Approaches to Examining the Motor Code 38

2.4.1 Central Analytical Framework 38

2.4.2 Gross statistical measures of spiking 39

2.4.3 Classical statistical approaches to variability (rates and dispersion) 39

2.4.4 Advantages of information-theoretic methods 40

2.4.5 Spike pattern-level analysis at millisecond-scale resolutions. 41

2.4.6 Applying the information-theoretic framework for examining motor codes across developmental learning. 44

CHAPTER 3: MILLISECOND-SCALE MOTOR CODING PRECEDES SENSORIMOTOR LEARNING IN SONGBIRDS 47

Abstract 47

3.1 Introduction 48

3.2 Materials and Methods 51

3.2.1 Subjects 51

3.2.2 Song Acoustic Analysis 52

3.2.3 Electrophysiological Recordings 53

3.2.4 Spiketrain Analysis 55

3.3 Results 62

3.3.1 Gross changes in RA spiking statistics across sensorimotor learning in juvenile Bengalese finch 62

3.3.2 Millisecond-scale motor coding is present across development 67

3.3.3 Evolution of RA spike pattern vocabularies across learning 70

3.4 Discussion 75

3.5 Supplementary Materials 82

3.5.1 Emergence of bursting activity during song acquisition in Bengalese finches 82

3.5.2 Spike rate variability and acoustic variability decrease simultaneously during song learning 84

3.5.3 Millisecond-scale motor coding of vocal acoustics throughout song learning 86

3.5.4 Entropy of activity patterns 87

3.5.5 Examining Potential Sources of Bias for Entropy Calculation 91

CHAPTER 4: EXTENDED DISCUSSION 98

4.1 Brain and body considerations on the role of RA motor coding 98

4.2 A causal perspective on the motor code temporal precision and changes in motor cortical spiking features 100

4.2.1 Result 1: Millisecond-scale precision is behaviorally relevant and constant across learning. 100

4.2.2 Result 2: Decreased variability in RA spiking - intrinsic properties supporting more reliable and selective use of spike patterns 104

4.2.3 Result 3: RA spike pattern reorganization deviates from Poisson-like statistics 106

4.3 Network‑level theoretical hypotheses: learning rules stabilizing spike and neural recruitment patterns 108

4.3.1 Reorganization and selection of spike patterns through recurrent population dynamics 108

4.3.2 Spike pattern selection via reinforcement-modulated spike-timing dependent plasticity rules 110

4.3.3 Effect of population size on neural ensemble recruitment and pattern selection 111

4.4 Future Directions 113

4.4.1 What is the content of spike patterns across learning, and how do patterns map onto behavior? 114

4.4.2 What features of spike patterns are favored by central and peripheral mechanisms of learning? 122

4.4.3 How do ensembles change? Timing across neurons, recruitment size, and downstream impact 124

4.4.4 How do shared dictionaries diverge into variant-specific motifs across learning? 125

BIBLIOGRAPHY 127

About this Dissertation

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Laney Graduate School
Department	Biological and Biomedical Sciences
Subfield / Discipline	Neuroscience
Degree	Ph.D.
Submission	Dissertation
Language	English
Research Field	Biology, Physiology Psychology, Behavioral Biology, Neuroscience
Keyword	Information Theory Skill Learning Motor Code Motor Control Motor Cortex Neural Code Motor Learning
Committee Chair / Thesis Advisor	Samuel J. Sober, Emory University
Committee Members	Gordon J. Berman, Emory University Astrid A. Prinz, Emory University Ilya Nemenman, Emory University

Last modified

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	File download under embargo until 12 July 2026	2025-12-15 15:41:28 -0500	File download under embargo until 12 July 2026

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions