EVOLUTIONARY AND FUNCTIONAL NEUROBIOLOGY OF BIRDSONG
A Dissertation Presented to the Faculty of the Graduate School
of Cornell University In Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy
by Jordan Matthew Moore
February 2010

© 2010 Jordan Matthew Moore

EVOLUTIONARY AND FUNCTIONAL NEUROBIOLOGY OF BIRDSONG
Jordan Matthew Moore, Ph. D. Cornell University 2010
Birdsong is a complex, learned vocal communication signal that functions in reproduction. Many aspects of the neural mechanisms underlying this behavior remain unknown, with respect to both song production and perception. The work described here investigates these issues from different levels of analysis. Nearly every aspect of song is characterized by immense interspecific diversity. In particular, species range from those that learn a single syllable to others that produce hundreds or thousands, but neither the ultimate nor proximate causes for this variability have been identified. Chapter 2 uses comparative analyses across a wide phylogeny of songbirds to address these questions. Syllable repertoire size is positively and strongly correlated with the degree of convergence along the descending motor pathway, suggesting that the level of top-down control rather than motor circuit size per se is more closely associated with the complexity of behavioral output. Repertoire size is not related to the sizes of other brain regions, however. Chapter 3 describes general evolutionary patterns of functional neural circuits and tests predictions made by two models of brain evolution. Interestingly, support was found for both. Nuclei that develop late in ontogeny possess larger allometric slopes than those that develop early, which is consistent with a developmental conservation model in which a general stretching of neurogenetic schedules underlies increases in brain size. Functional circuits evolve in concert and do so independently of others, however, supporting a mosaic pattern. Finally, chapter 4 examines the function of the caudomedial

nidopallium (NCM), a region of the auditory forebrain, in adult females. Numerous correlative studies suggest that NCM is important for song learning. Consistent with this, inactivation of NCM eliminates female preferences for familiar songs but has much more modest effects on their preferences for high quality songs. Together, these experiments are the first to relate evolutionary changes in behavioral capacities to those in its underlying circuit, to document the coordinated evolution of functional circuits, and to demonstrate a causative role for the auditory forebrain in song perception.

BIOGRAPHICAL SKETCH Jordan Moore hails from the small town of Waterloo, Wisconsin, where the cows outnumber the people. It was here that Jordan developed a deep appreciation for the finer things in life, including the Green Bay Packers, bratwursts, and deep-fried cheese curds. Jordan graduated from Waterloo High School in 1998 and attended the University of Wisconsin-Madison, where he majored in biology and zoology. Unofficially, he also majored in making plays. Jordan had many great experiences at the University of Wisconsin. Among them, he gained his first research experience in the lab of Prof. Tom Yin working with Dr. Dan Tollin. These two were terrific mentors and this positive experience influenced Jordan to pursue a career in neuroscience. Before beginning graduate school, however, Jordan wanted to experience life outside the Midwest and moved to Salamanca, Spain for a year. He then came to Ithaca, a small, dreary town in New York. While not particularly fond of his new surroundings, he nevertheless met some great friends and good times were had. He fondly recalls Jon’s matchbox car, Damian playing sheepshead, golfing with Dave, the Rio trio, the lost week with Chuck crashing on the couch, Matt going balls to the wall, and of course Cakes’ Friday Night Lights. Six long years later, the curtain falls.
iii

To mom and dad iv

ACKNOWLEDGEMENTS I am indebted to many people for their generosity and support throughout my graduate career. First and foremost, I thank my advisor, Prof. Timothy DeVoogd, for kindly offering me the opportunity to work in his laboratory and for his enduring guidance, support, and patience. He provided me with excellent graduate training for which I am extremely grateful. I also thank my committee members, Profs. Elizabeth Adkins-Regan, Andrew Bass, and Sandra Vehrencamp, for their many helpful suggestions and encouragement throughout. I sincerely appreciate the assistance I received from Dr. Bruce Land, who helped design and build the song preference testing apparatus and Mr. Tim Van Deusen, who was an essential asset for all six years and selflessly devoted a great deal of time and effort to construct many devices for my research. Thank you to the many members of the DeVoogd lab who helped along the way, including Marc Weiskopf, Khalid Alkhalaifi, Emily Chen, Lisa Deuel, Katie McCrary, Xem Bui, Young Suh, Melanie Tai, and Setareh Omran. I also thank the NBB administrative staff for their willingness to help in any and every situation, especially Ms. Terri Natoli and Ms. Stacey Coil. I am grateful to Dawn and Geoff Potter for being wonderful friends and for providing many delicious meals and good times. I could not have completed this journey without the unending support and encouragement from my dear friends: Matt Arnegard, Bruce Carlson, Damian Elias, Steve Finckbeiner, Kristin Goble, Jon King, Dave McLean, Shane Peace, and Dustin Rubinstein. Last but not least, I thank my family and my wonderful fiancée, Janine, for their unending encouragement. Funding was provided by a Graduate Research Fellowship from the National Science Foundation, International Research Travel Grants from the Mario Einaudi Center, the Neurobiology and Behavior Chapter of the Animal Behavior Society, The Cornell Chapter of Sigma Xi, the American Ornithologists’ Union, and a Frank M.
v

Chapman Memorial Fund Award from the American Museum of Natural History. I thank them for their support.
vi

Biographical Sketch Dedication Acknowledgements Table of Contents List of Figures List of Tables Chapter 1 Chapter 2 Chapter 3 Chapter 4

TABLE OF CONTENTS

iii iv v vii viii ix 1 8 61 101

vii

Figure 2.1 Figure 2.2 Figure 2.3

LIST OF FIGURES Diagram of the song system Phylogeny Repertoire-song system convergence correlations

10 16 45

Figure 3.1 Figure 3.2 Figure 3.3

Pathways of four functional systems Allometry of telencephalic vs. thalamic and brainstem nuclei Song nuclei are more variable than other systems

66 77 80

Figure 4.1 Figure 4.2 Figure 4.3 Figure 4.4 Figure 4.5 Figure 4.6 Figure 4.7 Figure 4.8 Figure 4.9

Diagram of the ascending auditory system Diagram of phonotaxis cage and experimental timeline Spectrograms of tutored and isolate songs Hearing test results Familiar song preferences following NCM inactivation Call latencies to familiar songs following NCM inactivation High-quality song preferences following NCM inactivation Familiar song preferences following CM inactivation Cannulation site verification

104 107 111 114 116 117 119 120 122

viii

Table 2.1 Table 2.2 Table 2.3 Table 2.4 Table 2.5 Table 2.6 Table 2.7 Table 2.8 Table 2.9 Table 2.10

LIST OF TABLES Syllable repertoires and sources Phylogenetic signal of song nuclei Allometric equations of song nuclei Bivariate song nucleus correlations Multivariate song nucleus correlations HVC, RA, and nXIIts neuron number correlations Bivariate repertoire-song system correlations Multivariate repertoire-song system correlations Repertoire-song nucleus neuron number correlations Repertoire-song system convergence correlations

Table 3.1 Table 3.2 Table 3.3 Table 3.4 Table 3.5 Table 3.6 Table 3.7 Table 3.8 Table 3.9 Table 3.10

Phylogenetic signal of 23 nuclei Allometric equations Factor analysis of all nuclei Correlations between song nucleus volumes Correlations between limbic structure volumes Correlations between visual nucleus volumes Correlations between auditory nucleus volumes Correlations between nucleus volumes and neuron numbers Correlations between song nucleus neuron numbers Correlations between visual nucleus neuron numbers

20 24 27 29 30 34 35 36 42 43
73 76 78 82 83 84 85 86 87 88

ix

CHAPTER 1
INTRODUCTION Songbirds have long been admired their varied and complex songs, and a great deal of research has sought to fully describe this behavior with respect to each of Tingbergen’s (1963) four questions: its function, evolution, development, and underlying neural mechanisms. Birdsong is a learned vocal communication signal that functions in reproduction and is generally thought to have evolved in response to sexual selection. Typically, males sing to attract females as mates and/or defend territories against competing males. The song features under selection can differ between taxonomic groups, but large syllable repertoires are a favored phenotype in several. Females of many species prefer to mate with males that possess larger repertoires (Byers & Kroodsma 2009), and repertoires must somehow be costly to senders to have evolved as honest signals of male quality (Zahavi 1975a; Bradbury & Vehrencamp 1998). Within species, repertoire size reflects developmental history: males that experience nutritional or immunological stresses early in life develop smaller repertoires (Nowicki et al. 1998; Nowicki & Searcy 2004), and this association can persist throughout life as adult male song sparrows (Melospiza melodia) with larger repertoires have superior immune systems and are in better physiological condition (Pfaff et al. 2007). Repertoires may also signal general properties of the brain. Song learning and production are controlled by a discrete, evolutionarily conserved neural circuit dedicated specifically to this purpose. Within species, the size of the premotor nucleus HVC is heritable, is positively related to song complexity, and is correlated with overall brain size (Airey et al. 2000; Garamszegi & Eens 2004). The research described here further characterizes some of these relationships from different levels of analysis, focusing on the coordinated evolution
1

between song complexity and its underlying neural circuit, the evolution of the song system with respect to other neural circuits, and the neural pathways underlying song perception in females.
Songbirds are remarkable among the animal kingdom because they learn a large portion of their vocal repertoire. While they are not unique in this regard, they are by far the largest group of animals possessing this capacity with each of the approximately 4000 species learning its own characteristic song. Species can vary in nearly every aspect of the learning process, including the accuracy with which they copy tutor songs, their ability for improvisation, the types of sounds they attempt to copy, the time of life during which memorization and vocal modification occurs, and the social interactions necessary and/or sufficient to produce normal learning. This immense diversity is perhaps best exemplified by interspecific differences in the capacity to learn multiple song components, or syllables, with species ranging from those that learn a single syllable to others than produce hundreds or thousands. Moreover, song is controlled by a conserved network of interconnected nuclei dedicated specifically to this purpose. Together, these attributes make songbirds ideally suited for comparative analyses to help discover principles of motor circuit organization because differences in behavior can be directly related to those in the structure of its underlying pathway.
Previous studies have identified positive relationships between repertoire size and the volume of nucleus HVC (e.g., DeVoogd et al. 1993; Garamszegi & Eens 2004). These anatomical correlates are consistent with physiological data showing that HVC neurons sparsely encode song, whereby each individual cell only fires during a single, brief segment of song (Hahnloser et al. 2002). Thus, increases in repertoire appear to require the addition of HVC neurons to encode this increase in vocal versatility. However, these correlations are often quite weak and HVC is not
2

singly responsible for song production. Chapter 2 explores this brain-behavior relationship in greater depth by relating syllable repertoire size to the structure of the entire song circuit. Surprisingly, the degree of convergence along this motor pathway is more closely related to repertoire size than its overall size, suggesting that the level of top-down control could be a primary determinant of behavioral complexity. This chapter also seeks to identify potential functions of repertoires by searching for evolutionary correlations between repertoire size and the relative volumes of structures not directly involved in song production. No such relationships are evident, however, suggesting that repertoires have not evolved as reliable signals for general cognitive capacities.
A related issue concerns the evolution of the brain in general. Two models have been proposed to describe changes in brain composition as its overall size changes. The first, referred to as the model of developmental conservation, suggests that neurogenetic schedules are largely conserved but stretched or shortened in duration and differences in composition between large and small brains are largely predictable from their developmental sequence (Finlay & Darlington 1995). This model specifically predicts that late-developing structures will become disproportionately large as overall size increases owing to exponential increases in the number of neuronal precursors. The second model, referred to as mosaic evolution, focuses more on the relationships between areas and predicts that those sharing functional connections will evolve in concert but will be independent of structures underlying other functions (Barton & Harvey 2000). While not mutually exclusive, these ideas are often portrayed as such and several reports have sought to falsify one model by providing support for the other (e.g., de Winter & Oxnard 2001). Furthermore, nearly all work in this area has focused on mammals and all analyses have been limited to the sizes of major brain subdivisions that contain areas belonging
3

to many distinct functional systems. Chapter 3 addresses some of these limitations by documenting the evolution of discrete neural circuits in songbirds, in which the pallia are organized into nuclei rather than layers which permits the direct and accurate measurement of areas belonging to various functional systems. Interestingly, elements of both models are supported. Telencephalic nuclei incorporate neurons late in ontogeny and exhibit greater allometric slopes than areas in the thalamus and brainstem, which cease neuronal recruitment at earlier times. The relative volumes of nuclei within one system are also strongly related to each other, regardless of brain subdivision, and are largely independent of structures belonging to a different system. Finally, the song system develops late compared to surrounding tissue and is much more variable than nuclei within other systems, perhaps facilitating the extreme variability of the song system compared to other systems and the immense behavioral diversity between species.
Lastly, female song preferences drive many evolutionary changes in male song and thus in underlying brain structure, but much remains unknown about the nature of these preferences. In particular, the neural substrate underlying song perception has only recently been identified and the specific functions of different regions are unclear. One area that has received special attention is the caudomedial nidopallium (NCM) because this area responds selectively to song and its physiological and genomic responses habituate to repeated stimulus presentations, suggesting that it may play an important role in song memory formation and/or storage (Mello et al. 1992, 1995; Chew et al. 1995). Yet, despite the many studies that have characterized responses in this area following presentation of various auditory stimuli, none have demonstrated its causal role in song perception. Chapter 4 assesses the role of NCM in song perception directly in female zebra finches (Taeniopygia guttata). Using a phonotaxis task, the role of NCM in song perception is measured for two distinct types
4

of song preferences shown by adult females. The first is based on song familiarity; females prefer the songs of their mate or their father over unfamiliar songs (Miller 1979a,b). The second is based on song quality and is independent of previous song exposure; females prefer longer songs composed of more syllables and also prefer normal, learned songs to improvised songs that have abnormally simple phonology (Neubauer 1999b; Lauay et al. 2004). Inactivation of NCM eliminated female preferences for their mate’s song but had much more modest effects on those for high quality songs, indicating that NCM is necessary for preferences based on song familiarity but not those based on song quality.
In conclusion, the work presented here contributes to new areas of birdsong research from different levels of analysis. It is the first to relate evolutionary changes in behavioral capacities to the structure of its underlying neural circuit. It is the first to document the evolution of discrete neural circuits. It is also the first to demonstrate that an area of the auditory forebrain is required for a specific type of song preference. These findings provide a foundation for future work, such as investigations into the evolution of neuronal morphologies within song nuclei, developmental mechanisms that underlie coordinated changes between nuclei, and the neural pathways that produce the expression of distinct song preferences by female songbirds.
5

REFERENCES Airey, D. C., Castillo-Juarez, H., Casella, G., Pollak, E. J. & DeVoogd, T. J. 2000
Variation in the volume of zebra finch song control nuclei is heritable: developmental and evolutionary implications. Proc. R. Soc. Lond. B 267, 2099-2104. Barton, R. A. & Harvey, P. H. 2000 Mosaic evolution of brain structure in mammals. Nature 405, 1055-1058. Bradbury, J. W. & Vehrencamp, S. L. 1998 Principles of animal communication. Sunderland, MA: Sinauer Associates. Byers, B. E. & Kroodsma, D. E. 2009 Female mate choice and songbird song repertoires. Anim. Behav. 77, 13-22. de Winter, W. & Oxnard, C. E. 2001 Evolutionary radiations and convergences in the structural organization of mammalian brains. Nature 409, 710-714. DeVoogd, T. J., Krebs, J. R., Healy, S. D. & Purvis, A. 1993 Relations between song repertoire size and the volume of brain nuclei related to song: comparative evolutionary analyses amongst oscine birds. Proc. R. Soc. Lond. B 254, 75-82. Finlay, B. L. & Darlington, R. B. 1995 Linked regularities in the development and evolution of mammalian brains. Science 268, 1578-1584. Garamszegi, L. Z. & Eens, M. 2004 Brain space for a learned task: strong intraspecific evidence for neural correlates of singing behavior in songbirds. Brain Res. Rev. 44, 187-193. Hahnloser, R. H. R., Kozhevnikov, A. A. & Fee, M. S. 2002 An ultra-sparse code underlies the generation of neural sequences in a songbird. Nature 419, 65-70. Lauay, C., Gerlach, N. M., Adkins-Regan, E. & DeVoogd, T. J. 2004 Female zebra finches require early song exposure to prefer high-quality song as adults. Anim. Behav. 68, 1249-1255. Miller, D. B. 1979a Long-term recognition of father's song by female zebra finches. Nature 280, 389-391. Miller, D. B. 1979b The acoustic basis of mate recognition by female zebra finches (Taeniopygia guttata). Anim. Behav. 27, 376-380. Neubauer, R. L. 1999 Super-normal length song preferences of female zebra finches (Taeniopygia guttata) and a theory of the evolution of bird song. Evol. Ecol. 13, 365-380. Nowicki, S., Peters, S. & Podos, J. 1998 Song learning, early nutrition and sexual selection in songbirds. Am. Zool. 38, 179-190.
6

Nowicki, S. & Searcy, W. A. 2004 Song function and the evolution of female preferences: why birds sing, why brains matter. Ann. N.Y. Acad. Sci. 1016, 704-723.
Pfaff, J. A., Zanette, L., MacDougall-Shackleton, S. A. & MacDougall-Shackleton, E. A. 2007 Song repertoire size varies with HVC volume and is indicative of male quality in song sparrows (Melospiza melodia). Proc. R. Soc. B 274, 20352040.
Tinbergen, N. 1963 On aims and methods of ethology. Z. Tierpsychol. 20, 410-433. Zahavi, A. 1975 Mate selection--a selection for a handicap. J. Theor. Biol. 53, 205-
214.
7

CHAPTER 2
DEGREE OF CONVERGENCE ALONG A MOTOR PATHWAY IS STRONGLY RELATED TO BEHAVIORAL COMPLEXITY ACROSS A WIDE PHYLOGENY
OF SONGBIRDS Abstract Songbirds have long been renowned for their varied and complex songs, which are learned vocal communication signals that have been shaped by sexual selection. Species especially differ in their capacity for learning multiple song components, but much remains unknown about the proximate and ultimate bases for such variety. This comparative study describes evolutionary correlations between syllable repertoires and the structure of the song system across a wide phylogeny of mostly unstudied songbird species. The relative volumes of song nuclei evolved in concert but can differ markedly in their variability about allometric expectations. In particular, the premotor nucleus HVC and one of its primary afferents, the medial magnocellular nucleus of the anterior nidopallium (MMAN), are much more variable than other areas. Both relative HVC and MMAN volumes are positively correlated with repertoire size. Relationships were substantially improved, however, when nucleus volumes were considered relative to their efferent target(s) rather than overall size. These findings suggest that evolutionary changes in repertoire size have been more closely associated with changes in the degree of convergence along this motor pathway than relative circuit size per se. Introduction
Birdsong is a complex, learned vocal communication signal that functions in reproduction. It is an incredibly diverse behavior, with each of the approximately 4000 oscine species producing characteristic song(s) that can differ in nearly every
8

respect. This variability is perhaps best exemplified by repertoires; species range from those that learn only a single song component, or syllable (e.g., chipping sparrow, Spizella passerina; Liu & Kroodsma 1999) to others that produce hundreds of distinct sounds (e.g., brown thrasher, Toxostoma rufum; Kroodsma & Parker 1977). Such immense disparities in behavioral capacities are seldom seen within so large a group of closely related species, yet it remains unknown how or why such large differences exist.
Song learning and production are primarily controlled by the song system, a discrete and evolutionarily conserved neural circuit dedicated specifically to this purpose (Figure 2.1; Nottebohm et al. 1976; 1982). This network is largely comprised of two parallel pathways arising from sensorimotor nucleus HVC: a caudal motor pathway underlies song production (Nottebohm et al. 1976; Vu et al. 1994), and a rostral, basal ganglia pathway generates motor variability and is required for song learning but not for stable song production in adults (Bottjer et al. 1984; Scharff & Nottebohm 1991; Kao et al. 2005; Ölveczky et al. 2005). The discrete nature and functional specificity of this circuit, coupled with its quantifiable output, make songbirds well suited for comparative analyses that can address the evolution, function, and proximate causes of a complex motor behavior.
The neural bases underlying the capacity for complex songs have not been unequivocally identified, but three lines of evidence suggest that HVC size is a primary determinant of this ability. First, repertoire size is positively correlated with HVC volume within species (Nottebohm et al. 1981; Garamszegi & Eens 2004), between species (DeVoogd et al. 1993; Székely et al. 1996), and between sexes (Brenowitz et al. 1985; MacDougall-Shackleton & Ball 1999). HVC volume is largely independent of early auditory experience, therefore increases in HVC neuron number appear to be permissive for more learning rather than a consequence of it
9

Figure 2.1. Diagram of the song system. The caudal motor pathway is shown in black, the anterior forebrain pathway in gray, and feedback projections with dashed lines and open circles. DMP: dorsomedial nucleus of the posterior thalamus; DLM: dorsolateral nucleus of the lateral anterior thalamus; VRG: ventral respiratory group.
10

(Brenowitz et al. 1995; Leitner et al. 2002). Second, song complexity and the size of HVC covary after various experimental treatments, such as nutritional or immunological stresses during development (Spencer et al. 2005; Pfaff et al. 2007) or administration of endocrine-disrupting chemicals (Markman et al. 2008). Third, intricate physiological studies have shown that HVC encodes song with a sparse code, whereby individual neurons that project to the robust nucleus of the arcopallium (RA) fire a temporally precise burst of action potentials that corresponds to one specific syllable segment during each song rendition (Hahnloser et al. 2002). If such specificity is general across songbirds, and it appears to be (Mooney et al. 2001; Prather et al. 2008), then an ability to produce a larger repertoire would seem to necessitate more HVC neurons.
Nevertheless, the relationship between HVC and repertoire size remains contentious because several inconsistencies among studies appear to contradict notions of a strict correspondence between the two. The majority of intraspecific studies have not detected significant correlations (e.g., Kirn et al. 1989; Brenowitz et al. 1991); there are sizeable outliers from the parallel sex differences in song behavior and anatomical dimorphisms (Gahr et al. 1998, 2008); age-related increases in repertoire are not necessarily accompanied by expansions of HVC (Nottebohm et al. 1986); and seasonal fluctuations in HVC volume are not accompanied by changes in repertoire (Kirn et al. 1989; Brenowitz et al. 1991; Smith et al. 1995). These observations call attention to the seemingly unreliable nature of the HVC and song complexity relationship.
Comparative analyses can also provide clues about potential behavioral function(s). Birdsong has evolved in the context of sexual selection, where males typically sing to attract mates and/or defend territories and females assess song to gauge male quality. Females of many species prefer large repertoires, which can
11

convey direct and/or indirect benefits to be gained (Buchanan & Catchpole 2000; Nowicki & Searcy 2004). Regarding the latter, intraspecific studies have shown that male repertoires can honestly signal genetic diversity, developmental history, physiological condition, immunocompetence, parasite load, and overall viability (e.g., Hasselquist et al. 1996; Nowicki et al. 1998; Marshall et al. 2003; Buchanan et al. 2004; Pfaff et al. 2007). By virtue of their relation to HVC, repertoires may also indicate features about the brain in general. Such a function was suggested by the findings that HVC volume is heritable and is related to the sizes of other song nuclei and the telencephalon (Burek et al. 1991; Airey et al. 2000).
The goals of this study were threefold: (1) to re-examine the evolutionary relationship between syllable repertoire size and song system anatomy using a new data set of mostly unstudied species; (2) to describe general evolutionary patterns of the song circuit, and (3) to test for correlations between repertoire and brain regions not directly involved in song. We find that consideration of multiple song nuclei rather than HVC alone greatly enhances the correlation with behavior. Methods (a) Specimen collection and preparation
All procedures were approved by the Cornell University Institutional Animal Care and Use Committee. One to four adult male songbirds of 58 temperate zone species were wild-caught under permit with mist nets. Collections were restricted to spring months (April – June) when birds were reproductively active to minimize seasonal variation in song system anatomy. Most were collected throughout Hungary from 1993-1995 or in Tompkins County, New York in 2004. Exceptions are the white-throated sparrows (Zonotrichia albicollis), Ontario, Canada in 1991; European pied flycatchers (Ficedula hypoleuca), central Norway in 1995; yellow-throated buntings (Emberiza elegans), China in 1991; and northern mockingbirds (Mimus
12

polyglottos), North Carolina in 2003. Birds were deeply anesthetized with barbiturate anesthetic at the time of capture and transcardially perfused with 0.8% saline followed by 10% formalin in saline. Brains were extracted, post-fixed for at least 24 hours, cryoprotected with 30% sucrose/10% formalin, embedded in gelatin, and sectioned at 40 μm in the coronal plane with a freezing, sliding microtome. Sections were then mounted onto gel-coated slides, Nissl-stained with cresyl violet, and coverslipped using Permount. (b) Brain measurements
Alternate sections were viewed under 40× or, for larger regions [mesopallium (MP) and hippocampus (Hp)], every fourth section was viewed under 20× magnification and nucleus boundaries were traced using a camera lucida. Unmagnified digital images of every fourth section were used to measure the telencephalon. The song nuclei measured were HVC, RA, Area X of the striatum, the lateral and medial magnocellular nuclei of the anterior nidopallium (LMAN and MMAN, respectively), uvaeformis (Uva), the dorsomedial nucleus of the intercollicular complex (DM), and the tracheosyringeal portion of the hypoglossal nucleus (nXIIts). Other brain regions included MP, three limbic structures [Hp, septum (Sep), and nucleus taeniae of the amygdala (TnA)], and an auditory nucleus [dorsal part of the lateral mesencephalic nucleus (MLd)]. Measurements were made in one side of the brain, typically the left except in cases where torn tissue or poor staining prevented measurements. Cross-sectional areas of scanned traces and telencephalon images were measured using NIH ImageJ software (Rasband 2007) and final volumes were computed by summing the areas and multiplying by the sampling interval (0.08 or 0.16 mm).
The boundaries of most song nuclei were unambiguous. Song control nucleus HVC, however, is complex and comprised of at least three distinct subdivisions
13

(Fortune & Margoliash 1995). Identified boundaries were based on two that are visible in the coronal plane: a central region that contains many large, darkly-stained cells and the caudomedial region or paraHVC, which is comprised of small, denselypacked neurons. In most cases, MMAN and LMAN were not contiguous and were easily distinguished. When the two were juxtaposed, MMAN was delineated on the basis of its slightly smaller somata and generally higher cell density than LMAN. The nucleus interface of the nidopallium (NIf) could not be reliably identified in many specimens and the medial portion of the dorsolateral nucleus of the anterior thalamus (DLM) lacks clear boundaries, therefore neither was measured.
Rostral MP was identified by differences in staining intensity compared to adjacent regions; in more central and caudal sections, it was clearly distinguished by the superior frontal and mesopallial laminae. The lateral extent of Hp was identified by its lower cell density compared to adjacent areas (Székely & Krebs 1996). Caudal Hp does not have a clear boundary with the surrounding parahippocampal area, therefore measurements were arbitrarily stopped at the section in which the cerebellum reached the dorsal-most extent of the telencephalon. Sep boundaries were identified by differences in staining intensity and largely coincided with those described by chemoarchitecture (Goodson et al. 2004); the estimates reported here include most of the four major subdivisions but exclude portions of the nucleus of the diagonal band ventral to the septopallio-mesencephalic tract. Both TnA and MLd were delineated on the basis of their staining intensity and/or cell density.
Neuronal densities in HVC, RA, and nXIIts were estimated from one brain of each species. Presumptive neurons were discriminated from glia on the basis of their soma size, uniformly stained cytoplasm, and a single, darkly stained nucleolus. Nucleolus counts were tallied within sampling windows that were evenly distributed throughout each plane of each structure, and a neuron was only counted if its
14

nucleolus was within the grid. Grid dimensions were 80 × 80 μm (600×, HVC and nXIIts) or 120 × 120 μm (400×, RA). On average, 13 (range: 7-26) counts were made in each nucleus. (c) Syllable repertoires
The behavioral unit of interest was a song syllable. A syllable is generally defined as a continuous trace on a spectrogram or a stereotyped sequence of notes separated by less than 25 msec, and a repertoire is the number of unique syllables produced by an individual bird. Species-typical syllable repertoires were obtained for 49 of the 58 species, most from published sources. When multiple sources were found for a species, one was chosen on the basis of sample size, clarity of the syllable definition and conformance to that above, proximity of the recording site to the site of specimen collection, and the season during which recordings were made. Repertoires for seven species were estimated from recordings held by the Macaulay Library at Cornell University (mean: 3 birds per species, range: 1-5) with the Syrinx sound analysis program (Burt 2006). All recorded syllables known to originate from one individual (mean: 199 syllables per bird, range: 14-528) were used to construct plots of new syllables encountered versus total syllables sampled. The function y = a(1 – e-bx) was fit to the data and the asymptote (a) was used as the estimated repertoire. (d) Phylogeny construction
A fully resolved, composite phylogeny was constructed from published molecular phylogenies (Figure 2.2). Species spanned 18 families, 16 of which are within the Passerida parvorder. All inter-family relationships were inferred from the analysis of two nuclear genes (RAG-1 and RAG-2). Most intra-family relationships were based on the mitochondrial gene(s) cytochrome b and/or ND2, occasionally in
15

Figure 2.2. Composite phylogeny of 58 species used in this study. Branch lengths were scaled using the method of Grafen (1989) and the rho transform (ρ = 0.3). Sources were: inter-family (Barker et al. 2004); Sylviidae (Blondel et al. 1996; Leisler et al. 1997; Helbig & Seibold 1999; Drovetski et al. 2004); Hirundinidae (Sheldon et al. 2005); Fringillidae (Arnaiz-Villena et al. 2001; Klicka et al. 2003; Zamora et al. 2006); Motacillidae (Arctander et al. 1996); Emberizidae (Klicka et al. 2007; Alström et al. 2008); Cardinalidae (Yuri & Mindell 2002); Muscicapidae (Wink et al. 2002; Voelker & Spellman 2004); Turdidae (Klicka et al. 2005); Mimidae (Cibois & Cracraft 2004).
16

17

addition to a nuclear gene; two exceptions were the Mimidae family (nuclear RAG-1) and the Sylvia genus (nuclear DNA hybridization). (e) Data analysis
Specimens representing 13 of the 58 species in this data set were the subjects of previous studies (DeVoogd et al. 1993; 1995; Székely et al. 1996). All areas were re-measured, and exclusion of these data did not alter the significance of principal findings reported here. Mesquite (Maddison & Maddison 2008) was used to manage data and trees, and the PDAP:PDTree v1.14 module was used for all analyses of independent contrasts (Garland et al. 1999; Garland & Ives 2000). Phylogenetic signal was computed with the Matlab program PHYSIGLL.m (Blomberg et al. 2003) and phylogenetic generalized least squares (PGLS) models were constructed with REGRESSIONv2.m (Lavin et al. 2008). For these analyses, trees were first converted to variance-covariance matrices with the PDDIST module of Phenotypic Diversity Analysis Programs (Garland & Ives 2000).
PGLS models using log10-transformed data were used for most analyses. Arbitrary branch lengths computed with the method of Grafen (1989) and scaled with the rho (ρ = 0.3) transform were used because of this tree’s consistent performance in diagnostic tests for all traits (Garland et al. 1992; Blomberg et al. 2003). For comparison, data were also analyzed using a star phylogeny (a single basal polytomy with equal branch lengths, as assumed by conventional statistics). Models were constructed to determine relationships between song nuclei and between syllable repertoire and various combinations of nuclei; in each case, a ‘size’ parameter was included as a covariate. Values used for the ‘size’ reference depended on the trait(s) in question and/or the particular analysis. For song nuclei, limbic structures, and MP it was the difference between the telencephalon and respective total circuit volume (i.e. sum of song nuclei or limbic structures) or subdivision (i.e. MP) to control for
18

part-whole artifacts. For MLd, it was telencephalon volume and for the telencephalon it was species-typical body mass (Dunning 2008). Residual analyses were also performed in some cases, in which relative trait sizes were first computed and then regressed with another trait (Darlington & Smulders 2001). Here, allometric slopes of log10-transformed song nucleus volumes were estimated using either the same ‘size’ reference listed above (e.g., HVC versus telencephalon) or another nucleus (e.g., HVC versus RA). These were then used to compute relative trait volumes with the formula log10[trait/(sizeb)], where b was the allometric exponent, and ‘trait’ and ‘size’ were original data values. Data from all 58 species were used for brain-only analyses, but only the 49 for which repertoires were obtained were used in the brain-behavior correlations. Prior to the latter analyses, the phylogenetic tree was pruned, branch lengths re-scaled and diagnostic tests re-run. Nested and non-nested models were compared using ln maximum likelihood ratio tests (LRTs) and the Akaike Information Criterion (AIC; smaller is better), respectively, and partial F-tests were used to test for significance of independent variables in multivariate models. Note that results from PGLS models and independent contrasts are identical (Garland & Ives 2000), and R2 values from PGLS models are not comparable to those from traditional ordinary least squares models. Results
Species-typical repertoires and their respective sources are listed in Table 2.1. The use of different branch lengths produced qualitatively consistent results, therefore only those based on the hierarchical tree are presented here. (a) Allometry of song nuclei
All absolute and most relative nucleus volumes exhibited significant phylogenetic signal, indicating that they were better described by the hierarchical than star phylogeny under a Brownian motion model of evolution (Table 2.2). The
19

Table 2.1. Syllable repertoires and sources used in this study. Common names follow Gill & Wright (2006).
20

21

Common Name
American Goldfinch American Robin Barn Swallow Barred Warbler Black Redstart Black-capped Chickadee Blue Jay Blue-winged Warbler Brown-headed Cowbird Chipping Sparrow Common Blackbird Common Chaffinch Common Grasshopper Warbler Common House Martin Common Linnet Common Nightingale Common Starling Common Whitethroat Common Yellowthroat Corn Bunting Crested Lark Eurasian Reed Warbler Eurasian Skylark Eurasian Tree Sparrow European Goldfinch

Scientific Name
Carduelis tristis Turdus migratorius Hirundo rustica Sylvia nisoria Phoenicurus ochruros Poecile atricapillus Cyanocitta cristata Vermivora pinus Molothrus ater Spizella passerina Turdus merula Fringilla coelebs Locustella naevia Delichon urbicum Carduelis cannabina Luscinia megarhynchos Sturnus vulgaris Sylvia communis Geothlypis trichas Emberiza calandra Galerida cristata Acrocephalus scirpaceus Alauda arvensis Passer montanus Carduelis carduelis

Syl Rep
33 38 20 74 5 2

Source
Macaulay Library (n = 5) Dobson & Lemon (1979) Garamszegi et al. (2005) Székely et al. (1996) Macaulay Library (n = 1) Kroodsma et al. (1999)

4 Gill & Murray (1972) 15 Allan & Suthers (1994) 1 Albrecht & Oring (1995) 108 Garamszegi et al. (2005) 18 Garamszegi et al. (2005) 1 Wahlström (1966)

1160 44 119 5 9 55 80 341 10 60

Hultsch (1980) Garamszegi et al. (2005) Garamszegi et al. (2005) Lemon et al. (1983) Macaulay Library (n = 2) Macaulay Library (n = 1) Catchpole (1980) Briefer et al. (2008) Lang & Barlow (1997) Güttinger & Clauss (1982)

22

Table 2.1 (Continued).
Common Name
European Pied Flycatcher European Stonechat Great Reed Warbler Grey Catbird Hawfinch House Sparrow Indigo Bunting Lesser Grey Shrike Lesser Whitethroat Long-tailed Bushtit Marsh Warbler Moustached Warbler Northern Cardinal Northern Mockingbird Northern Wheatear Red-backed Shrike River Warbler Rose-breasted Grosbeak Sand Martin Savi's Warbler Sedge Warbler Song Sparrow Spotted Flycatcher

Scientific Name
Ficedula hypoleuca Saxicola rubicola Acrocephalus arundinaceus Dumetella carolinensis Coccothraustes coccothraustes Passer domesticus Passerina cyanea Lanius minor Sylvia curruca Aegithalos caudatus Acrocephalus palustris Acrocephalus melanopogon Cardinalis cardinalis Mimus polyglottos Oenanthe oenanthe Lanius collurio Locustella fluviatilis Pheucticus ludovicianus Riparia riparia Locustella luscinioides Acrocephalus schoenobaenus Melospiza melodia Muscicapa striata

Syl Rep
27 93 30 328

Source
Espmark & Lampe (1993) Schwager & Güttinger (1984) Forstmeier et al. (2006) Kroodsma et al. (1997)

6 Nivison (1978) 8 Payne (1981)

68 Klit (1999)

412 Bell et al. (2004) 134 Feßl & Hoi (2000) 14 Anderson & Conner (1985) 216 Wildenthal (1965) 46 Macaulay Library (n = 5)

2 Wahlström (1966) 19 Lemon & Chatfield (1973) 3 Macaulay Library (n = 3) 1 Wahlström (1966) 68 Buchanan & Catchpole (2000) 38 Borror (1965)

23

Table 2.1 (Continued).
Common Name
Tawny Pipit Tree Pipit Veery Western Yellow Wagtail Whinchat White Wagtail White-throated Sparrow Wood Thrush Yellowhammer Yellow-throated Bunting

Scientific Name
Anthus campestris Anthus trivialis Catharus fuscescens Motacilla flava Saxicola rubetra Motacilla alba Zonotrichia albicollis Hylocichla mustelina Emberiza citrinella Emberiza elegans

Syl Rep
1 11 12 2 130

Source
Wallschläger (1964) Wallschläger (1964) Dobson & Lemon (1979) Ödeen & Björklund (2003) Schwager & Güttinger (1984)

5 Falls & Kopachena (1994) 16 Dobson & Lemon (1979) 5 Garmaszegi et al. (2005) 109 Zeng et al. (2005)

Table 2.2. Phylogenetic signal of log10-transformed trait values. Tree branch lengths were scaled using the method of Grafen (1989) and the rho transform (ρ = 0.3). Refer to Blomberg et al. (2003) and Revell et al. (2008) for discussions of the K-statistic and phylogenetic signal.
24

Trait log(Mass)

df

MSE0

MSEstar

MSE

obs. exp. MSE0/MSE MSE0/MSE

K

p ln lklhd ln lklhdstar

56 0.078 0.075 0.058

1.351

1.518 0.890 <0.001 0.801

-6.847

log(Telen)

56 0.057 0.053 0.041

1.373

1.518 0.904 <0.001 10.711

3.190

log(HVC)

56 0.118 0.117 0.087

1.360

1.518 0.896 <0.001 -10.832

-19.599

log(RA)

56 0.066 0.065 0.060

1.090

1.518 0.718 0.004 -0.379

-2.416

log(Area X)

56 0.082 0.079 0.064

1.288

1.518 0.848 <0.001 -1.978

-8.265

log(LMAN)

56 0.069 0.068 0.065

1.057

1.518 0.696 0.001 -2.644

-3.949

log(MMAN) 56 0.099 0.098 0.071

1.394

1.518 0.918 <0.001 -5.074

-14.509

log(Uva)

56 0.029 0.028 0.024

1.242

1.518 0.818 <0.001 26.834

21.483

log(DM)

56 0.024 0.023 0.022

1.082

1.518 0.713 0.002 28.916

27.685

25

log(nXIIts)

56 0.045 0.044 0.044

1.011

1.518 0.666 0.010 8.643

8.789

log(MP)

56 0.067 0.063 0.050

1.338

1.518 0.881 <0.001 4.940

-1.767

log(Hp)

56 0.052 0.047 0.040

1.313

1.518 0.865 <0.001 11.635

6.886

log(Sep)

56 0.039 0.035 0.030

1.292

1.518 0.851 <0.001 19.711

15.162

log(TnA)

56 0.046 0.043 0.028

1.636

1.518 1.078 <0.001 21.877

9.291

log(MLd)

56 0.026 0.025 0.024

1.091

1.518 0.718 0.001 26.683

25.109

log(syl rep)

47 0.592 0.592 0.445

1.331

1.436 0.927 <0.001 -49.196

-56.165

rel_HVC

47 0.084 0.084 0.058

1.445

1.436 1.006 <0.001 0.734

-8.283

rel_RA

47 0.029 0.029 0.027

1.075

1.436 0.748 <0.001 19.835

18.130

rel_nXIIts

47 0.022 0.022 0.025

0.881

1.436 0.613 0.117 21.253

24.455

HVC_rel_RA 47 0.044 0.044 0.032

1.367

1.436 0.952 <0.001 15.062

7.435

RA_rel_nXIIts 47 0.020 0.020 0.020

0.997

1.436 0.694 0.013 27.265

27.339

allometric slopes of song nuclei were related to the brain subdivision within which they are located, with telencephalic nuclei having consistently larger slopes (0.77 – 0.85) than thalamic and brainstem structures (0.63 – 0.67) (Mann-Whitney rank sum test, p = 0.04; Table 2.3). There were striking differences in the variability of some nuclei. HVC and MMAN had comparatively weak allometric relationships (0.31 ≤ r2 ≤ 0.34), while RA, Area X, LMAN, and nXIIts varied more directly with telencephalon size (0.40 ≤ r2 ≤ 0.50) and Uva and DM tightly corresponded to it (0.72 ≤ r2 ≤ 0.76). These relationships were homoscedastic (all p > 0.12). The extreme variability of HVC and MMAN is further underscored by comparisons between similarly-sized species. For example, the spotted flycatcher (Muscicapa striata) has a slightly smaller telencephalon (84%) than the common yellowthroat (Geothlypis trichas) yet has 15- and 9-fold larger absolute HVC and MMAN volumes, respectively. By comparison, its RA, Area X, and LMAN volumes are 3-4 times as large. (b) Correlations between song nuclei
The volumes of most song nuclei were positively and strongly correlated after accounting for overall telencephalon size (Table 2.4). The main exception to this was DM, which was not correlated with any nuclei but had nearly significant relationships with HVC, RA, and Uva (0.08 ≤ p ≤ 0.14). Associations were usually strongest between monosynaptically-connected nuclei, and these generally remained significant in multivariate GLS models that explained variation in one nucleus as a function of all others (Table 2.5). For example, the only two significant partial coefficients in a model explaining HVC variation were attributable to a primary afferent source (MMAN, p = 4.3 × 10-5) and efferent target (RA, p = 5.0 × 10-5). Most significant relationships in these models were positive; the two exceptions included Uva, which
26

Table 2.3. Allometric data for traits used in the present analysis. Tree branch lengths were scaled using the Grafen (1989) and the rho transform (ρ = 0.3). In all cases, df = 56. SS: sum of song nuclei, LimbS: sum of limbic structures.
27

28

Brain Subdivision
Telencephalon
Diencephalon Mesencephalon Rhombencephalon
Telencephalon
Mesencephalon

X
log(Body Mass) log(Telen-SS) log(Telen-SS) log(Telen-SS) log(Telen-SS) log(Telen-SS) log(Telen-SS) log(Telen-SS) log(Telen-SS) log(Telen-MP) log(Telen-LimbS) log(Telen-LimbS) log(Telen-LimbS) log(Telen)

Y
log(Telen) log(HVC) log(RA) log(Area X) log(LMAN) log(MMAN) log(Uva) log(DM) log(nXIIts) log(MP) log(Hp) log(Sep) log(TnA) log(MLd)

r
0.908 0.560 0.705 0.685 0.631 0.586 0.850 0.869 0.647 0.945 0.916 0.914 0.834 0.759

p
<1.00E-16 4.94E-06 6.68E-10 3.08E-09 1.12E-07 1.33E-06
<1.00E-16 <1.00E-16
4.19E-08 <1.00E-16 <1.00E-16 <1.00E-16
4.44E-16 5.07E-12

Slope
0.766 0.810 0.852 0.851 0.793 0.768 0.642 0.634 0.669 1.052 0.898 0.779 0.685 0.576

Table 2.4. Matrix of relationships between nuclei with log(T-SS) as a confounding variable in each regression. Statistics refer to the partial coefficients of the variables indicated. Shaded cells indicate directly connected nuclei. In all cases, df = 55.

29

log(RA) log(Area X) log(LMAN) log(MMAN) log(Uva) log(DM) log(nXIIts)

r p r p r p r p r p r p r p

log(HVC) 0.702
1.19E-09 0.678
6.65E-09 0.497
8.42E-05 0.741
4.26E-11 0.325 0.014 0.222 0.097 0.329 0.013

log(RA) log(Area X) log(LMAN) log(MMAN)

0.698 1.58E-09
0.699 1.44E-09
0.461 3.09E-04
0.446 4.99E-04
0.197 0.142
0.630 1.48E-07

0.717 3.46E-10
0.569 3.93E-06
0.461
3.10E-04 0.190 0.157
0.504 6.32E-05

0.398
0.002
0.432
0.001 0.101 0.456
0.509
5.29E-05

0.047 0.730 0.141 0.297 0.176 0.191

log(Uva)
0.237 0.076 0.122 0.364

log(DM)
0.163 0.224

Table 2.5. Parameters of multivariate GLS models describing relationships between song nuclei. Each column represents a separate model with the nucleus in the top row as the response variable and all others as explanatory variables. Shaded cells indicate directly connected nuclei. In all cases, df = 50.
30

log(HVC) log(RA) log(Area X) log(LMAN) log(MMAN) log(Uva) log(DM) log(nXIIts) intercept

coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p R2

log(HVC) ----
0.687
13.881
4.96E-04 0.290 2.735 0.104 -0.178 1.527 0.222
0.497
20.099
4.31E-05 -0.029 0.012 0.912 -0.084 0.136 0.713 -0.199 1.607 0.211 0.055 0.023 0.880 0.814

log(RA)
0.316
13.881
4.96E-04 ----0.021 0.031 0.862
0.226
5.759
0.020 -0.020 0.051 0.822
0.348
4.352
0.042 -0.043 0.079 0.780
0.382
16.653
1.61E-04 0.620 7.314 0.009 0.877

log(Area X) 0.179 2.735 0.104 -0.029 0.031 0.862
----
0.310
8.518
0.005 0.185 3.460 0.069
0.428
5.007
0.030 -0.101 0.328 0.570 0.228 3.574 0.064 1.381 45.026 1.71E-08 0.844

log(LMAN) -0.166 1.527 0.222
0.457
5.759
0.020
0.469
8.518
0.005 ----
0.094 0.559 0.458 0.259 1.126 0.294 -0.239 1.227 0.273 0.095 0.385 0.538 -0.675 4.045 0.050 0.770

31

Table 2.5 (Continued).

log(MMAN)

coef.

0.577

log(HVC)

F

20.099

p 4.31E-05

coef.

-0.051

log(RA)

F

0.051

p 0.822

coef.

0.349

log(Area X) F

3.460

p 0.069

coef.

0.117

log(LMAN) F

0.559

p 0.458

coef.

--

log(MMAN) F

--

p --

coef.

-0.536

log(Uva)

F

4.092

p 0.048

coef.

0.512

log(DM)

F

4.832

p 0.033

coef.

-0.190

log(nXIIts) F

1.257

p 0.268

coef.

-1.159

intercept

F

10.757

p 0.002

R2 0.736

log(Uva) -0.009 0.012 0.912
0.230
4.352
0.042
0.213
5.007
0.030 0.085 1.126 0.294
-0.141
4.092
0.048 ----
0.563
33.938
4.08E-07
-0.188
5.035
0.029 -0.751 19.668 5.06E-05 0.791

log(DM) -0.033 0.136 0.713 -0.037 0.079 0.780 -0.064 0.328 0.570 -0.100 1.227 0.273
0.172
4.832
0.033
0.718
33.938
4.08E-07 ----
0.221
5.495
0.023 0.041 0.033 0.858 0.714

log(nXIIts) -0.157 1.607 0.211
0.654
16.653
1.61E-04 0.293 3.574 0.064 0.081 0.385 0.538 -0.129 1.257 0.268
-0.486
5.035
0.029
0.448
5.495
0.023 ----1.201 19.387 5.62E-05
0.712

32

was inversely related to MMAN (p = 0.048) and nXIIts (p = 0.029) after accounting for the rest of the song circuit.
The cellular basis for some of these relationships was assessed by estimating neuron densities and numbers in HVC, RA, and nXIIts (Table 2.6). Neuron densities in each area were inversely related to telencephalon volume (all p < 8.1 × 10-9). After accounting for overall size, the volumes of all three nuclei were positively related to neuron number (0.88 ≤ r ≤ 0.98; all p < 1.3 × 10-19), but only those of RA (r = -0.46, p = 2.7 × 10-4) and nXIIts (r = -0.56, p = 6.1 × 10-6) were also inversely related to neuron density. Relative neuron numbers in each of the three nuclei were positively correlated (0.31 ≤ r ≤ 0.58; all p < 0.02), and those in HVC and RA were inversely related to neuron density in RA and nXIIts, respectively (-0.46 ≤ r ≤ -0.42; both p < 0.001). (c) Syllable repertoire size and song system neuroanatomy
Initial analyses sought to explain variation in syllable repertoire size as a function of individual song nucleus volumes with size (T-SS) as a confounding variable (Table 2.7). Here, log10(repertoire) was positively correlated with relative HVC (r2 = 0.46; p = 1.3 × 10-7) and MMAN (r2 = 0.26; p = 2.2 × 10-4) volumes but not with other nuclei (all p > 0.08). Numerous multivariate models were then compared (Table 2.8). A full model that included all song nuclei performed significantly better than either of the previous correlations (R2 = 0.72; ln ML = -17.7; LRT, both p < 2.9 × 10-5), and only HVC retained a significant coefficient (3.0; p = 1.4 × 10-7). Reduced models that contained nuclei along the caudal motor pathway were comparable to the full model, including those with all three nuclei (HVC, RA, nXIIts; R2 = 0.71; LRT, p = 0.80), HVC and RA (R2 = 0.67; p = 0.19), and HVC and nXIIts (R2 = 0.67; p = 0.20).
33

Table 2.6. Correlations between the volumes, neuron numbers, and neuron densities of HVC, RA, and nXIIts. The first three are bivariate correlations between the size factor and neuronal densities of the three nuclei. For each, df = 56. The remaining regressions are two-factor models with size [log(Telen-SS)] as a confounding variable. The partial r and corresponding p-values refer to each trait after accounting for size. For these models, df = 55.

34

Size log(Telen-SS)
log(Telen-SS) log(Telen-SS) log(Telen-SS) log(Telen-SS) log(Telen-SS)

X Trait --
log(HVC volume) log(RA volume) log(nXIIts volume) log(HVC neuron #) log(RA neuron #)

Y
log(HVC density) log(RA density) log(nXIIts density) log(HVC neuron #) log(HVC density) log(RA neuron #) log(RA density) log(nXIIts neuron #) log(nXIIts density) log(RA neuron #) log(RA density) log(nXIIts neuron #) log(nXIIts density)

Trait

rp

-0.815 7.20E-15

-0.671 8.08E-09

-0.708 5.14E-10

0.983 -0.070

2.84E-42 0.604

0.915 2.54E-23

-0.465 2.71E-04

0.882 1.32E-19

-0.559 6.06E-06

0.583 1.96E-06

-0.464 2.75E-04

0.524 2.91E-05

-0.424

0.001

Table 2.7. Two-factor models explaining the corresponding size reference are explanatory

variation in log(repertoire). The volume of variables, and the partial r, r2, and p-values

an individual structure refer to each trait after

and its accounting

for

size.

The ln ML and AIC values describe the entire model. In all cases, df = 47.

35

X

Size
log(Telen-SS) log(Telen-SS) log(Telen-SS) log(Telen-SS) log(Telen-SS) log(Telen-SS) log(Telen-SS) log(Telen-SS)

Trait
log(HVC) log(RA) log(Area X) log(LMAN)
log(MMAN) log(Uva) log(DM) log(nXIIts)

log(Body Mass) log(T-MP) log(T-LimbS) log(T-LimbS) log(T-LimbS) log(Telen)

log(Telen) log(MP) log(Hp) log(Sep) log(TnA) log(MLd)

Y
log(syl rep) log(syl rep) log(syl rep) log(syl rep)
log(syl rep) log(syl rep) log(syl rep) log(syl rep) log(syl rep) log(syl rep) log(syl rep) log(syl rep) log(syl rep) log(syl rep)

r
0.677 0.116 0.243 0.040
0.510 0.150 0.061 -0.251 0.183 -0.080 -0.113 -0.058 -0.117 0.021

Trait r2
0.458 0.013 0.059 0.002
0.260 0.022 0.004 0.063 0.033 0.006 0.013 0.003 0.014 4.60E-04

p
1.30E-07 0.434 0.096 0.788
2.15E-04 0.310 0.681 0.085 0.214 0.591 0.446 0.697 0.429 0.885

ln ML
-33.066 -47.728 -46.565 -48.019
-40.687 -47.502 -47.967 -46.465

AIC
74.131 103.457 101.130 104.038
89.375 103.003 103.933 100.929

Table 2.8. Parameters of some multivariate GLS models with log(repertoire) as the response variable. Each column represents a separate model with various combinations of log-transformed nucleus volumes as explanatory variables and log(Telen-SS) as a confounding variable. Likelihood ratio tests are in comparison to the full model (left column).
36

log(Telen-SS) log(HVC) log(RA) log(Area X) log(LMAN) log(MMAN) log(Uva) log(DM) log(nXIIts) intercept

coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p df R2
ln ML AIC LRT p

log(rep) 0.619 0.622 0.435
2.999
41.087
1.40E-07 -1.129 2.262 0.141 -0.551 0.887 0.352 0.016 0.001 0.975 -0.178 0.148 0.702 -0.358 0.134 0.716 0.337 0.165 0.687
-1.162
4.396
0.043 -1.852 0.300 0.587
39 0.723 -17.708 57.415 --

log(rep) 0.553 1.575 0.216
2.680
72.395
7.60E-11
-1.440
5.898
0.019 ----------------
-1.159
6.038
0.018 -1.784 1.766 0.191
44 0.710 -18.889 49.779
0.797

log(rep) -0.277 0.411 0.525
2.803
60.063
9.24E-10 ----0.910 2.429 0.126
-1.000
5.051
0.030 ------------1.627 1.947 0.170
44 0.638 -24.284 60.567 0.022

log(rep) 0.400 0.756 0.389
2.904
83.403
8.24E-12
-2.393
25.576
7.60E-06 -------------------0.299 0.056 0.814 45 0.670
-22.040 54.080
0.193

log(rep) -0.305 0.458 0.502
2.806
55.227
2.35E-09 ----
-1.717
12.768
0.001 ---------------2.894 7.376 0.009
45 0.597 -26.946 63.892 0.005

37

Table 2.8 (Continued).

log(Telen-SS) log(HVC) log(RA) log(Area X) log(LMAN) log(MMAN) log(Uva) log(DM) log(nXIIts) intercept

coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p df R2
ln ML AIC LRT p

log(rep) -0.483 1.332 0.255
2.483
67.361
1.72E-10 -------
-1.426
16.031
2.30E-04 ------------1.491 1.593 0.213 45 0.597
-26.946 63.892
0.015

log(rep)
-0.969
4.153
0.047
1.826
16.424
1.98E-04 ---------0.020 0.002 0.967 ---------4.150 9.940 0.003 45 0.618
-25.600 61.200 2.87E-05

log(rep) -0.421 0.365 0.549
1.964
38.070
1.75E-07 -------------0.965 1.083 0.304 ------1.698 0.430 0.515 45 0.482
-33.065 76.129 4.79E-05

log(rep) -0.694 0.843 0.363
1.862
38.052
1.75E-07 ----------------0.454 0.207 0.651 ---2.949 1.095 0.301 45 0.494
-32.483 74.966 3.17E-05

log(rep) 0.182 0.174 0.679
2.169
77.290
2.50E-11 -------------------
-1.907
-5.077
7.12E-06 -1.220 0.768 0.386 45 0.485
-32.953 75.906
0.202

38

Table 2.8 (Continued).

log(Telen-SS) log(HVC) log(RA) log(Area X) log(LMAN) log(MMAN) log(Uva) log(DM) log(nXIIts) intercept

coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p df R2
ln ML AIC LRT p

log(rep) -0.127 0.036 0.851 ----0.522 0.839 0.365 -------
1.660
16.100
2.24E-04 ---------3.136 2.733 0.105 45 0.306
-40.235 90.470 4.56E-08

log(rep) 0.609 0.738 0.395 ----
1.933
7.425
0.009 ----------------
-2.318
10.200
0.003 -2.173 1.015 0.319
45 0.232 -42.723 95.446 4.64E-09

log(rep) 0.009 0.000 0.989 -------
1.656
5.131
0.028 -1.013 2.243 0.141 ------------0.102 0.003 0.954
45 0.144 -45.373 100.747 3.97E-10

log(rep) -0.303 0.259 0.614 -------0.433 0.586 0.448 ----
1.726
12.946
0.001 ---------3.964 6.529 0.014
45 0.302 -40.370 90.741 4.03E-08

log(rep) 0.768 1.373 0.248 -------
1.609
9.835
0.003 -------------
-1.947
10.060
0.003 -3.285 2.594 0.114
45 0.266 -41.622 93.244 1.28E-08

39

Table 2.8 (Continued).

log(Telen-SS) log(HVC) log(RA) log(Area X) log(LMAN) log(MMAN) log(Uva) log(DM) log(nXIIts) intercept

coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p df R2
ln ML AIC LRT p

log(rep) -0.202 0.132 0.719 ----------0.704 2.535 0.118
1.774
19.116
7.19E-05 ---------2.938 3.040 0.088 45 0.331
-39.345 88.690 1.03E-07

log(rep)
1.383
4.453
0.040 ---------0.986 3.049 0.088 ----------
-1.818
6.196
0.017 -3.327 2.316 0.135
45 0.162 -44.859 99.717 6.41E-10

log(rep) -1.113 1.691 0.200 -------------
1.473
15.524
2.81E-04 0.918 0.829 0.367 ------6.698 4.474 0.040 45 0.306
-40.240 90.481 4.54E-08

log(rep) -0.467 0.276 0.602 -------------
1.499
15.572
2.76E-04 ----0.062 0.003 0.958 ---4.012 1.423 0.239 45 0.293
-40.686 91.372 3.02E-08

log(rep) 0.420 0.485 0.490 -------------
1.631
21.469
3.08E-05 -------
-1.360
7.506
0.009 0.196 0.009 0.924
45 0.394 -36.908 83.816 9.38E-07

40

A surprising pattern among partial coefficients emerged in the reduced models: several nuclei whose relative volumes alone were not related to repertoire (RA, Area X, LMAN, nXIIts) became significant explanatory variables after the variation associated with other area(s) had been factored out. Moreover, the sign of these partial slopes differed such that the most afferent nucleus relative to the syrinx was positive while those of efferent area(s) were usually negative. This pattern was evident in multiple combinations of nuclei and robust to the use of different trees. The only exception was HVC, which always had a positive coefficient, even when paired with MMAN or Uva (both p < 2.0 × 10-4). MMAN had a positive partial slope in every case except when paired with HVC, in which case it was not significant (p = 0.97, all other p < 0.001). Conversely, nXIIts had a negative coefficient when paired with every telencephalic nucleus (all p < 0.02) and nearly significant negative slopes with Uva and DM (both p < 0.08). The sign of coefficients for other nuclei depended on the nucleus with which it was paired; RA (-2.4), Area X (-1.7), and LMAN (-1.4; all p < 0.001) all had negative partial slopes when paired with HVC, but RA had a positive slope when paired with nXIIts (1.9, p = 0.009), as did Area X when paired with LMAN (1.7, p = 0.03) or nXIIts (1.6, p = 0.003). The same pattern was present in models that used HVC, RA, and nXIIts neuron numbers as explanatory variables (Table 2.9). Here, all models performed slightly less well than their volumetric counterparts, though the full model nevertheless explained a large portion of the observed behavioral variation (R2 = 0.64; AIC = 59.9).
These opposing relationships suggested that the degree of convergence along the motor pathway rather than its overall relative volume is more closely associated with syllable repertoire size. This was assessed with residual analyses whereby the size of one nucleus relative to another or to the telencephalon was first calculated and then regressed with repertoire size (Table 2.10). The volume of HVC relative to RA
41

Table 2.9. Parameters of some multivariate GLS models with log(repertoire) as the response variable. Each column represents a separate model with a size reference and various combinations of log-transformed nucleus neuron numbers as explanatory variables. Likelihood ratio tests are in comparison to the full model (left column).

42

log(Telen-SS) log(HVC n#) log(RA n#) log(nXIIts n#) intercept

coef. F p
coef. F p
coef. F p
coef. F p
coef. F p df R2
ln ML AIC LRT p

log(rep) 0.412 0.943 0.337
2.331
62.417
5.61E-10 -1.107 3.972 0.052
-1.369
6.056
0.018 -2.660 3.452 0.070
44 0.643 -23.957 59.915 0.012

log(rep) 0.349 0.608 0.440
2.410
60.671
7.00E-10
-1.812
13.043
0.001 ----4.169 9.339 0.004
45 0.594 -27.117 65.629 0.040

log(rep) 0.054 0.019 0.892
2.047
58.866
1.04E-09 ----
-1.940
15.564
2.77E-04 -3.111 4.544 0.039 45 0.611
-26.075 62.150 4.75E-11

log(rep) 0.719 1.221 0.275 ---1.010 1.822 0.184
-1.844
4.705
0.035 1.163 0.315 0.577
45 0.137 -45.595 101.190 8.31E-05

log(rep) -0.510 1.438 0.237
1.850
37.836
1.72E-07 -------6.582 24.277
1.12E-05 46
0.476 -33.352 74.704 3.51E-11

log(rep) 0.646 0.915 0.344 ---0.146 0.049 0.826 ----0.719 0.135 0.715 46 0.046
-48.032 104.063 1.52E-10

log(rep)
1.189
4.603
0.037 -------1.233 2.884 0.096 2.252 1.369 0.248
46 0.102 -46.568 101.135 0.012

Table 2.10. Parameters of some multivariate GLS models with log(repertoire) as the response variable. Each column represents a separate model with various combinations of relative nucleus volumes as explanatory variables. Volumes are relative to another nucleus (e.g., HVC_RA is HVC relative to RA) or to Telen-SS.

43

HVC_RA RA_nXIIts HVC_T-SS RA_T-SS nXIIts_T-SS intercept

coef. F p
coef. F p
coef. F p
coef. F p
coef. F p
coef. F p df R2
ln ML AIC

log(rep)
2.632
71.522
6.33E-11
1.442
13.037
7.52E-04 ----------0.592 3.169 0.082 46 0.694
-20.221 48.441

log(rep)
2.891
72.499
4.39E-11 ------------0.302 1.473 0.231 47 0.607
-26.334 59.201

log(rep) ----
2.218
13.030
7.42E-04 ----------0.216 0.172 0.680 47 0.217
-43.201 92.403

log(rep) -------
1.879
40.044
8.57E-08 ------5.384 59.954
6.18E-10 47
0.460 -34.098 74.195

log(rep) ---------0.592 1.004 0.322 ---2.857 3.207 0.080 47 0.021
-48.678 103.357

log(rep) -------------0.925 2.383 0.129 -1.491 0.658 0.421 47 0.048
-47.985 101.969

was strongly and positively related to repertoire (r2 = 0.61; AIC = 59.2; p = 4.4 × 1011) and was a significantly better predictor than HVC relative to telencephalon (r2 = 0.46; AIC = 74.2; Figure 2.3). The size of RA relative to nXIIts was also positively correlated with repertoire (r2 = 0.22; AIC = 92.4; p = 7.4 × 10-4) despite the fact that neither RA nor nXIIts volumes relative to telencephalon were related to it (both p > 0.13). Interestingly, the HVC-to-RA and RA-to-nXIIts values were not correlated (r = 0.23, p = 0.11) and both retained significant and positive partial coefficients in a twofactor model explaining the variation in repertoire (HVC-to-RA: 2.6, p = 6.3 × 10-11; RA-to-nXIIts: 1.4, p = 7.5 × 10-4). This model was one of the best as judged by the AIC (R2 = 0.69; AIC = 48.4); it was comparable to that containing HVC, RA, and nXIIts with T-SS as a confounding size variable (R2 = 0.71; AIC = 49.8). (d) Song and other brain structures
Syllable repertoire size was not correlated with the volumes of the telencephalon (r = 0.18, p = 0.21), MP (r = -0.08, p = 0.59), Hp (r = -0.11, p = 0.45), Sep (r = -0.06, p = 0.70), TnA (r = -0.12, p = 0.43), or MLd (r = 0.02, p = 0.88) after accounting for overall size (Table 2.7). Discussion
Behavioral specializations are often associated with hypertrophy of their underlying neural substrate. Examples are widespread and include brain structures related to auditory specializations (Kubke et al. 2004; Iwaniuk et al. 2006), echo- and electrolocation (Clark et al. 2001), binocular vision (Barton 2004; Iwaniuk et al. 2008), visual motor coordination (Iwaniuk & Wylie 2007), and even more abstract capabilities such as spatial memory (Krebs et al. 1989), feeding innovations (Timmermans et al. 2000), and complex social interactions (Whiten & Byrne 1988; Dunbar 1995; Burish et al. 2004). The nature of these brain-behavior relationships is complex and many of their intricacies remain unclear, but the enhanced behavioral
44

45

Figure 2.3. Correlations between standardized independent contrasts of log10(syllable repertoire) and (a) HVC volume relative to the telencephalon, (b) HVC volume relative to RA, and (c) RA volume relative to nXIIts. Black and open symbols in (a) to (b) highlight the reduction of the largest positive and negative deviations, respectively, when repertoire is correlated with HVC relative to RA rather than the telencephalon.

capacities seem likely to result from a larger region’s increased processing power and/or its expanded influence over other areas (Striedter 2005). These possibilities are difficult to dissociate because they are not mutually exclusive, are almost certainly linked, and their relative roles are likely to vary with the behavior produced. Moreover, the entire pathways relevant to a particular behavior are often unknown or their complete measurement is impractical. As a result, the evolutionary relationships between behaviors of interest and the structure of their underlying neural circuits have gone largely unexplored.
The neurobiology of birdsong provides distinct advantages in this regard. First, song is a conspicuous and quantifiable behavior that circumvents the need for categorical classifications of ‘specialists’ or ‘generalists’. Second, the song system is well characterized and the functional roles of several areas are known in detail. The premotor nucleus HVC encodes song sequence (Hahnloser et al. 2002), and its volume is positively related to the repertoire size both within and across species (DeVoogd et al. 1993; Garamszegi & Eens 2004). The present study replicates these findings across a wide phylogeny of mostly unstudied songbirds. However, some species deviate substantially from this regression line and relative HVC size cannot be said to accurately predict their capacity (or lack thereof) for complex songs. Inclusion of downstream nuclei in the present analyses accounts for these discrepancies remarkably well, and it does so in a surprising way because these relative volumes are inversely correlated with the residuals from the HVC-repertoire regression. Thus, species that have fewer HVC neurons than expected from their repertoire also have few RA and nXIIts neurons, while those with many more HVC neurons than expected from their repertoire have correspondingly high neuron numbers in RA and nXIIts. This indicates that repertoire size may not be regulated by the size of its underlying neural
46

circuit per se and instead highlights the potential importance of top-down control in the encoding of this complex motor behavior.
Syllables are an operationally convenient unit for behavioral quantification, but they can also obscure details about the nature of the observed vocal diversity. For example, one strategy for building a larger repertoire might be to produce a more diverse array of acoustically distinct sounds, while another might be to modify the temporal properties or sequence of sounds already mastered. Both would effectively produce new, distinct syllables, but they seem likely to involve different physiological mechanisms. Evolutionary changes in convergence from HVC to RA and RA to nXIIts are independent and may represent two means by which these behavioral differences arise. In the zebra finch, 40% of all HVC neurons project to RA and this proportion is remarkably stable across individuals (Ward et al. 2001). These cells sparsely encode song whereby each neuron produces a short (~6 ms), temporally precise burst of action potentials that corresponds to one specific syllable segment, and the sequential activity of these neurons ultimately specifies motor output (Hahnloser et al. 2002; Long & Fee 2008; Prather et al. 2008). This correspondence between single neurons and brief sounds highlights a mechanism by which increases in HVC neuron number could enhance syntactical capacities, even if many share similar axonal projections onto a given number of RA neurons. More HVC neurons will create more activity chain possibilities, which will increase the number of ways that short sounds can be combined. Conversely, decreases in the number of HVC neurons that can excite a particular RA neuron will decrease the number of ways that neuron can be incorporated into a particular song sequence.
Individual RA neurons burst multiple times throughout each song rendition, but population-level activity patterns are specific to short syllable segments and are equally temporally precise as activity in HVC (Yu & Margoliash 1996; Leonardo &
47

Fee 2005). RA also presents the first myotopic map of syringeal and respiratory muscles along this pathway (Vicario 1991; Wild 1993), and syringeal labial adduction or abduction and the airflow past them largely shape song spectral features (Goller & Suthers 1996). It was recently suggested that the strength of each muscular contraction is determined by the linear sum of convergent RA inputs onto nXIIts neurons rather than the specific identity of active RA ensembles (Fee et al. 2004; Leonardo & Fee 2005). If true, evolutionary increases in RA-to-nXIIts convergence could expand the range of attainable nXIIts firing rates (and thus the contractile strength of each motor unit) and/or enhance contractile precision if accompanied by decreased synaptic weights and/or axonal arborization. Such changes would increase the level of control over syringeal shape and could permit the production of more spectrally varied sounds. These ideas are admittedly speculative because many interspecific differences in circuit structure could affect the proposed relationships, including the number of syringeal muscle fibers, motor unit size, synaptic weights, axonal branching, convergence onto single neurons, and physiological activity patterns. Nevertheless, evolutionary changes in repertoire size are closely related to the degree of convergence along the caudal motor pathway, and current hypotheses on song encoding provide clues as to why this might be the case.
Much less is known about the functions of areas afferent to HVC and, incidentally, the mechanisms underlying song organization and song type switching. In the zebra finch, sequential activity of HVC neurons appears to be determined by intrinsic connectivity within the nucleus itself (Long & Fee 2008). Yet, most other species show far more behavioral variability and many can organize syllables into distinct song types or produce continuous strings of numerous syllables (e.g., Beecher et al. 2000; Briefer et al. 2008). It is not known how these feats are accomplished, but they seem likely to involve areas afferent to HVC given its role in syllable encoding.
48

The thalamic nucleus Uva supplies feedback to HVC that appears critical for interhemispheric coordination (Coleman & Vu 2005). The relative volume of Uva is positively correlated with that of HVC across species, but its deviations from allometric expectations are quite small. Insofar as the extensive interspecific variability in song organization can be expected to accompany large anatomical differences, Uva appears unlikely to serve such a role. The telencephalic nucleus MMAN also projects to HVC but its function is less clear. In the zebra finch, MMAN axons ramify throughout HVC and lesions in juveniles both severely disrupt song learning and prevent song stabilization (Foster et al. 1997; 2001). The effects of lesions made in adults are much more subtle, but they do increase syntactical variability and particularly affect the first syllable in a song. In light of its strong correlation with HVC and considerable variability across species, MMAN appears well positioned to function in syllable and/or song type organization.
We also tested for patterns of correlated evolution between repertoires and other neural traits. In the zebra finch, repertoire size is correlated with HVC volume, which is heritable and related to overall telencephalon size (Airey et al. 2000; Airey & DeVoogd 2000). Across species, however, repertoires are not related to relative telencephalon size (Garamszegi & Møller 2004) or to the relative volumes of structures involved in other behaviors, including MP, a telencephalic subdivision correlated with feeding innovation rate (Timmermans et al. 2000); Hp, which is involved in spatial memory (Krebs et al. 1989); Sep, which is involved in sociality and territoriality (Goodson et al. 1999; 2006); TnA, which underlies aspects of sexual behavior and pair bonding (Thompson et al. 1998; Svec et al. 2009); or MLd, which encodes general sound properties (Woolley & Casseday 2005). Thus, elaborate syllable repertoires do not appear to have evolved in coordination with brain areas underlying other behaviours and cognitive tasks.
49

REFERENCES
Airey, D. C., Castillo-Juarez, H., Casella, G., Pollak, E. J. & DeVoogd, T. J. 2000 Variation in the volume of zebra finch song control nuclei is heritable: developmental and evolutionary implications. Proc. R. Soc. Lond. B 267, 2099-2104.
Airey, D. C. & DeVoogd, T. J. 2000 Greater song complexity is associated with augmented song system anatomy in zebra finches. Neuroreport 11, 2339-2344.
Albrecht, D. J. & Oring, L. W. 1995 Song in chipping sparrows, Spizella passerina: structure and function. Anim. Behav. 50, 1233-1241.
Allan, S. E. & Suthers, R. A. 1994 Lateralization and motor stereotypy of song production in the brown-headed cowbird. J. Neurobiol. 25, 1154-1166.
Alström, P., Olsson, U., Lei, F., Wang, H. t., Gao, W. & Sundberg, P. 2008 Phylogeny and classification of the old world Emberizini (Aves, Passeriformes). Mol. Phylogenet. Evol. 47, 960-973.
Anderson, M. E. & Conner, R. N. 1985 Northern cardinal song in three forest habitats in eastern Texas. Wilson Bull. 97, 436-449.
Arctander, P., Folmer, O. & Fjeldså, J. 1996 The phylogenetic relationships of Berthelot's pipit Anthus berthelotii illustrated by DNA sequence data, with remarks on the genetic distance between rock and water pipits Anthus spinoletta. Ibis 138, 263-272.
Arnaiz-Villena, A., Guillén, J., Ruiz-del-valle, V., Lowy, E., Zamora, J., Varela, P., Stefani, D. & Allende, L. M. 2001 Phylogeography of crossbills, bullfinches, grosbeaks, and rosefinches. Cell. Mol. Life Sci. 58, 1159-1166.
Barker, F. K., Cibois, A., Schikler, P., Feinstein, J. & Cracraft, J. 2004 Phylogeny and diversification of the largest avian radiation. Proc. Natl. Acad. Sci. USA 101, 11040-11045.
Barton, R. A. 2004 Binocularity and brain evolution in primates. Proc. Natl. Acad. Sci. USA 101, 10113-10115.
Beecher, M. D., Campbell, S. E., Burt, J. M., Hill, C. E. & Nordby, J. C. 2000 Songtype matching between neighbouring song sparrows. Anim. Behav. 59, 21-27.
Bell, B. D., Borowiec, M., Lontkowski, J. & Pledger, S. 2004 Short records of marsh warbler (Acrocephalus palustris) song provide indices that correlate with nesting success. J. Ornithol. 145, 8-15.
50

Blomberg, S. P., Garland, T., Jr. & Ives, A. R. 2003 Testing for phylogenetic signal in comparative data: behavioral traits are more labile. Evolution 57, 717-745.
Blondel, J., Catzeflis, F. & Perret, P. 1996 Molecular phylogeny and the historical biogeography of the warblers of the genus Sylvia (Aves). J. Evol. Biol. 9, 871891.
Borror, D. J. 1965 Song variation in Maine song sparrows. Wilson Bull. 77, 5-37.
Bottjer, S. W., Miesner, E. A. & Arnold, A. P. 1984 Forebrain lesions disrupt development but not maintenance of song in passerine birds. Science 224, 901903.
Brenowitz, E. A., Arnold, A. P. & Levin, R. N. 1985 Neural correlates of female song in tropical duetting birds. Brain Res. 343, 104-112.
Brenowitz, E. A., Lent, K. & Kroodsma, D. E. 1995 Brain space for learned song in birds develops independently of song learning. J. Neurosci. 15, 6281-6286.
Brenowitz, E. A., Nalls, B., Wingfield, J. C. & Kroodsma, D. E. 1991 Seasonal changes in avian song nuclei without seasonal changes in song repertoire. J. Neurosci. 11, 1367-1374.
Briefer, E., Aubin, T., Lehongre, K. & Rybak, F. 2008 How to identify dear enemies: the group signature in the complex song of the skylark Alauda arvensis. J. Exp. Biol. 211, 317-326.
Buchanan, K. L. & Catchpole, C. K. 2000 Song as an indicator of male parental effort in the sedge warbler. Proc. R. Soc. Lond. B 267, 321-326.
Buchanan, K. L., Leitner, S., Spencer, K. A., Goldsmith, A. R. & Catchpole, C. K. 2004 Developmental stress selectively affects the song control nucleus HVC in the zebra finch. Proc. R. Soc. Lond. B 271, 2381-2386.
Burek, M. J., Nordeen, K. W. & Nordeen, E. J. 1991 Neuron loss and addition in developing zebra finch song nuclei are independent of auditory experience during song learning. J. Neurobiol. 22, 215-223.
Burish, M. J., Kueh, H. Y. & Wang, S. S. H. 2004 Brain architecture and social complexity in modern and ancient birds. Brain Behav. Evol. 63, 107-124.
Burt, J. M. 2006 Syrinx, ver. 2.6h. http://www.syrinxpc.com/.
Catchpole, C. K. 1980 Sexual selection and the evolution of complex songs among European warblers of the genus Acrocephalus. Behaviour 74, 149-165.
51

Cibois, A. & Cracraft, J. 2004 Assessing the passerine "tapestry": phylogenetic relationships of the Muscicapoidea inferred from nuclear DNA sequences. Mol. Phylogenet. Evol. 32, 264-273.
Clark, D. A., Mitra, P. P. & Wang, S. S. H. 2001 Scalable architecture in mammalian brains. Nature 411, 189-193.
Coleman, M. J. & Vu, E. T. 2005 Recovery of impaired songs following unilateral but not bilateral lesions of nucleus uvaeformis of adult zebra finches. J. Neurobiol. 63, 70-89.
Darlington, R. B. & Smulders, T. V. 2001 Problems with residual analysis. Anim. Behav. 62, 599-602.
DeVoogd, T. J., Houtman, A. M. & Falls, J. B. 1995 White-throated sparrow morphs that differ in song production rate also differ in the anatomy of some songrelated brain areas. J. Neurobiol. 28, 202-213.
DeVoogd, T. J., Krebs, J. R., Healy, S. D. & Purvis, A. 1993 Relations between song repertoire size and the volume of brain nuclei related to song: comparative evolutionary analyses amongst oscine birds. Proc. R. Soc. Lond. B 254, 75-82.
Dobson, C. W. & Lemon, R. E. 1979 Markov sequences in songs of American thrushes. Behaviour 68, 86-105.
Drovetski, S. V., Zink, R. M., Fadeev, I. V., Nesterov, E. V., Koblik, E. A., Red'kin, Y. A. & Rohwer, S. 2004 Mitochondrial phylogeny of Locustella and related genera. J. Avian Biol. 35, 105-110.
Dunbar, R. I. M. 1995 Neocortex size and group size in primates: a test of the hypothesis. J Human Evol 28, 287-296.
Dunning, J. B. 2008 Handbook of avian body masses, 2nd ed. Boca Raton, FL: CRC Press.
Espmark, Y. O. & Lampe, H. M. 1993 Variations in the song of the pied flycatcher within and between breeding seasons. Bioacoustics 5, 33-65.
Falls, J. B. & Kopachena, J. G. 1994 White-throated sparrow (Zonotrichia albicollis). In Birds of North America Online (ed. A. Poole). Ithaca: Cornell Lab of Ornithology.
Fee, M. S., Kozhevnikov, A. A. & Hahnloser, R. H. R. 2004 Neural mechanisms of vocal sequence generation in the songbird. Ann. N.Y. Acad. Sci. 1016, 153-170.
52

Feßl, B. & Hoi, H. 2000 Song complexity and song structure in the moustached warbler Acrocephalus melanopogon. J. Avian Biol. 31, 144-150.
Forstmeier, W., Hasselquist, D., Bensch, S. & Leisler, B. 2006 Does song reflect age and viability? A comparison between two populations of the great reed warbler Acrocephalus arundinaceus. Behav. Ecol. Sociobiol. 59, 634-643.
Fortune, E. S. & Margoliash, D. 1995 Parallel pathways and convergence onto HVc and adjacent neostriatum of adult zebra finches (Taeniopygia guttata). J. Comp. Neurol. 360, 413-441.
Foster, E. F. & Bottjer, S. W. 2001 Lesions of a telencephalic nucleus in male zebra finches: influences on vocal behavior in juveniles and adults. J. Neurobiol. 46, 142-165.
Foster, E. F., Mehta, R. P. & Bottjer, S. W. 1997 Axonal connections of the medial magnocellular nucleus of the anterior neostriatum in zebra finches. J. Comp. Neurol. 382, 364-381.
Gahr, M., Metzdorf, R., Schmidl, D. & Wickler, W. 2008 Bi-directional sexual dimorphisms of the song control nucleus HVC in a songbird with unison song. PLoS One 3, e3073.
Gahr, M., Sonnenschein, E. & Wickler, W. 1998 Sex difference in the size of the neural song control regions in a dueting songbird with similar song repertoire size of males and females. J. Neurosci. 18, 1124-1131.
Garamszegi, L. Z., Balsby, T. J. S., Bell, B. D., Borowiec, M., Byers, B. E., Draganoiu, T., Eens, M., Forstmeier, W., Galeotti, P., Gil, D., Gorissen, L., Hansen, P., Lampe, H. M., Leitner, S., Lontkowski, J., Nagle, L., Nemeth, E., Pinxten, R., Rossi, J.-M., Saino, N., Tanvez, A., Titus, R., Török, J., Duyse, E. V. & Møller, A. P. 2005 Estimating the complexity of bird song by using capture-recapture approaches from community ecology. Behav. Ecol. Sociobiol. 57, 305-317.
Garamszegi, L. Z. & Eens, M. 2004 Brain space for a learned task: strong intraspecific evidence for neural correlates of singing behavior in songbirds. Brain Res. Rev. 44, 187-193.
Garamszegi, L. Z. & Møller, A. P. 2004 Extrapair paternity and the evolution of bird song. Behav. Ecol. 15, 508-519.
Garland, T., Jr., Harvey, P. H. & Ives, A. R. 1992 Procedures for the analysis of comparative data using phylogenetically independent contrasts. Syst. Biol. 41, 18-32.
53

Garland, T., Jr. & Ives, A. R. 2000 Using the past to predict the present: confidence intervals for regression equations in phylogenetic comparative methods. Am. Nat. 155, 346-364.
Garland, T., Jr., Midford, P. E. & Ives, A. R. 1999 An introduction to phylogenetically based statistical methods, with a new method for confidence intervals on ancestral values. Am. Zool. 39, 374-388.
Gill, F. & Wright, M. 2006 Birds of the World: Recommended English Names. Princeton, NJ: Princeton University Press.
Gill, F. B. & Murray, B. G. 1972 Song variation in sympatric blue-winged and goldenwinged warblers. Auk 89, 625-643.
Goller, F. & Suthers, R. A. 1996 Role of syringeal muscles in controlling the phonology of bird song. J. Neurophysiol. 76, 287-300.
Goodson, J. L., Eibach, R., Sakata, J. & Adkins-Regan, E. 1999 Effect of septal lesions on male song and aggression in the colonial zebra finch (Taeniopygia guttata) and the territorial field sparrow (Spizella pusilla). Behav. Brain Res. 98, 167-180.
Goodson, J. L., Evans, A. K. & Lindberg, L. 2004 Chemoarchitectonic subdivisions of the songbird septum and a comparative overview of septum chemical anatomy in jawed vertebrates. J. Comp. Neurol. 473, 293-314.
Goodson, J. L., Evans, A. K. & Wang, Y. 2006 Neuropeptide binding reflects convergent and divergent evolution in species-typical group sizes. Horm. Behav. 50, 223-236.
Grafen, A. 1989 The phylogenetic regression. Phil. Trans. R. Soc. Lond. B 326, 119157.
Güttinger, H. R. & Clauss, G. 1982 Der gesangsaufbau von stieglitz-kanarienbastarden (Carduelis carduelis × Serinus canaria) im vergleich zu den elternarten. J. Ornithol. 123, 269-286.
Hahnloser, R. H. R., Kozhevnikov, A. A. & Fee, M. S. 2002 An ultra-sparse code underlies the generation of neural sequences in a songbird. Nature 419, 65-70.
Hasselquist, D., Bensch, S. & von Schantz, T. 1996 Correlation between male song repertoire, extra-pair paternity and offspring survival in the great reed warbler. Nature 381, 229-232.
54

Helbig, A. J. & Seibold, I. 1999 Molecular phylogeny of palearctic-African Acrocephalus and Hippolais warblers (Aves: Sylviidae). Mol. Phylogenet. Evol. 11, 246-260.
Hultsch, H. 1980 Beziehungen zwischen struktur, zeitlicher variabilitaet und sozialem einsatz des gesangs der nachtigall (Luscinia megarhynchos). Ph.D. Dissertation, Free University of Berlin.
Iwaniuk, A. N., Clayton, D. H. & Wylie, D. R. W. 2006 Echolocation, vocal learning, auditory localization and the relative size of the avian auditory midbrain nucleus (MLd). Behav. Brain Res. 167, 305-317.
Iwaniuk, A. N., Heesy, C. P., Hall, M. I. & Wylie, D. R. W. 2008 Relative Wulst volume is correlated with orbit orientation and binocular visual field in birds. J. Comp. Physiol. A 194, 267-282.
Iwaniuk, A. N. & Wylie, D. R. W. 2007 Neural specialization for hovering in hummingbirds: hypertrophy of the pretectal nucleus lentiformis mesencephali. J. Comp. Neurol. 500, 211-221.
Kao, M. H., Doupe, A. J. & Brainard, M. S. 2005 Contributions of an avian basal ganglia-forebrain circuit to real-time modulation of song. Nature 433, 638-643.
Kirn, J. R., Clower, R. P., Kroodsma, D. E. & DeVoogd, T. J. 1989 Song-related brain regions in the red-winged blackbird are affected by sex and season but not repertoire size. J. Neurobiol. 20, 139-163.
Klicka, J., Burns, K. & Spellman, G. M. 2007 Defining a monophyletic Cardinalini: a molecular perspective. Mol. Phylogenet. Evol. 45, 1014-1032.
Klicka, J., Voelker, G. & Spellman, G. M. 2005 A molecular phylogenetic analysis of the "true thrushes" (Aves: Turdinae). Mol. Phylogenet. Evol. 34, 486-500.
Klicka, J., Zink, R. M. & Winker, K. 2003 Longspurs and snow buntings: phylogeny and biogeography of a high-latitude clade (Calcarius). Mol. Phylogenet. Evol. 26, 165-175.
Klit, I. 1999 The function of song forms in the lesser whitethroat Sylvia curruca. Bioacoustics 10, 31-45.
Krebs, J. R., Sherry, D. F., Healy, S. D., Perry, V. H. & Vaccarino, A. L. 1989 Hippocampal specialization of food-storing birds. Proc. Natl. Acad. Sci. USA 86, 1388-1392.
55

Kroodsma, D. E., Byers, B. E., Halkin, S. L., Hill, C., Minis, D., Bolsinger, J. R., Dawson, J.-A., Donelan, E., Farrington, J., Gill, F. B., Houlihan, P., Innes, D., Keller, G., Macaulay, L., Marantz, C. A., Ortiz, J., Stoddard, P. K. & Wilda, K. 1999 Geographic variation in black-capped chickadee songs and singing behavior. Auk 116, 387-402.
Kroodsma, D. E., Houlihan, P. W., Fallon, P. A. & Wells, J. A. 1997 Song development by grey catbirds. Anim. Behav. 54, 457-464.
Kroodsma, D. E. & Parker, L. D. 1977 Vocal virtuosity in the brown thrasher. Auk 94, 783-785.
Kubke, M. F., Massoglia, D. P. & Carr, C. E. 2004 Bigger brains or bigger nuclei? Regulating the size of auditory structures in birds. Brain Behav. Evol. 63, 169180.
Lang, A. L. & Barlow, J. C. 1997 Cultural evolution in the Eurasian tree sparrow: divergence between introduced and ancestral populations. Condor 99, 413-423.
Lavin, S. R., Karasov, W. H., Ives, A. R., Middleton, K. M. & Garland, T., Jr. 2008 Morphometrics of the avian small intestine compared with that of nonflying mammals: a phylogenetic approach. Physiol. Biochem. Zool. 81, 526-550.
Leisler, B., Heidrich, P., Schulze-Hagen, K. & Wink, M. 1997 Taxonomy and phylogeny of reed warblers (genus Acrocephalus) based on mtDNA sequences and morphology. J. Ornithol. 138, 469-496.
Leitner, S., Nicholson, J., Leisler, B., DeVoogd, T. J. & Catchpole, C. K. 2002 Song and the song control pathway in the brain can develop independently of exposure to song in the sedge warbler. Proc. R. Soc. Lond. B 269, 2519-2524.
Lemon, R. E. & Chatfield, C. 1973 Organization of song of rose-breasted grosbeaks. Anim. Behav. 21, 28-44.
Lemon, R. E., Struger, J. & Lechowicz, M. J. 1983 Song features as species discriminants in American warblers (Parulidae). Condor 85, 308-322.
Leonardo, A. & Fee, M. S. 2005 Ensemble coding of vocal control in birdsong. J. Neurosci. 25, 652-661.
Liu, W. C. & Kroodsma, D. E. 1999 Song development by chipping sparrows and field sparrows. Anim. Behav. 57, 1275-1286.
Long, M. A. & Fee, M. S. 2008 Using temperature to analyse temporal dynamics in the songbird motor pathway. Nature 456, 189-194.
56

MacDougall-Shackleton, S. A. & Ball, G. F. 1999 Comparative studies of sex differences in the song-control system of songbirds. Trends Neurosci. 22, 432436.
Maddison, W. P. & Maddison, D. R. 2008 Mesquite: a modular system for evolutionary analysis, ver. 2.5. http://mesquiteproject.org/.
Markman, S., Leitner, S., Catchpole, C., Barnsley, S., Müller, C. T., Pascoe, D. & Buchanan, K. L. 2008 Pollutants increase song complexity and the volume of the brain area HVC in a songbird. PLoS One 3, e1674.
Marshall, R. C., Buchanan, K. L. & Catchpole, C. K. 2003 Sexual selection and individual genetic diversity in a songbird. Proc. R. Soc. Lond. B (Suppl.) 270, S248-S250.
Mooney, R., Hoese, W. & Nowicki, S. 2001 Auditory representation of the vocal repertoire in a songbird with multiple song types. Proc. Natl. Acad. Sci. USA 98, 12778-12783.
Nivison, J. J. 1978 The vocal behavior and displays of the house sparrow, Passer domesticus L., in the United States. Ph.D. Dissertation, Wayne State University.
Nottebohm, F., Kasparian, S. & Pandazis, C. 1981 Brain space for a learned task. Brain Res. 213, 99-109.
Nottebohm, F., Kelley, D. B. & Paton, J. A. 1982 Connections of vocal control nuclei in the canary telencephalon. J. Comp. Neurol. 207, 344-357.
Nottebohm, F., Nottebohm, M. E. & Crane, L. 1986 Developmental and seasonal changes in canary song and their relation to changes in the anatomy of songcontrol nuclei. Behav. Neural Biol. 46, 445-471.
Nottebohm, F., Stokes, T. M. & Leonard, C. M. 1976 Central control of song in the canary, Serinus canarius. J. Comp. Neurol. 165, 457-486.
Nowicki, S., Peters, S. & Podos, J. 1998 Song learning, early nutrition and sexual selection in songbirds. Am. Zool. 38, 179-190.
Nowicki, S. & Searcy, W. A. 2004 Song function and the evolution of female preferences: why birds sing, why brains matter. Ann. N.Y. Acad. Sci. 1016, 704-723.
Ödeen, A. & Bjröklund, M. 2003 Dynamics in the evolution of sexual traits: losses and gains, radiation and convergence in yellow wagtails (Motacilla flava). Mol. Ecol. 12, 2113-2130.
57

Ölveczky, B. P., Andalman, A. S. & Fee, M. S. 2005 Vocal experimentation in the juvenile songbird requires a basal ganglia circuit. PLoS Biol. 3, e153.
Payne, R. B. 1981 Song learning and social interaction in indigo buntings. Anim. Behav. 29, 688-697.
Pfaff, J. A., Zanette, L., MacDougall-Shackleton, S. A. & MacDougall-Shackleton, E. A. 2007 Song repertoire size varies with HVC volume and is indicative of male quality in song sparrows (Melospiza melodia). Proc. R. Soc. B 274, 20352040.
Prather, J. F., Peters, S., Nowicki, S. & Mooney, R. 2008 Precise auditory-vocal mirroring in neurons for learned vocal communication. Nature 451, 305-310.
Rasband, W. S. 2007 ImageJ, ver. 1.38x. Bethesda, MD: US National Institutes of Health.
Revell, L. J., Harmon, L. J. & Collar, D. C. 2008 Phylogenetic signal, evolutionary process, and rate. Syst. Biol. 57, 591-601.
Scharff, C. & Nottebohm, F. 1991 A comparative study of the behavioral deficits following lesions of various parts of the zebra finch song system: implications for vocal learning. J. Neurosci. 11, 2896-2913.
Schwager, G. & Güttinger, H. R. 1984 Der gesangsaufbau von braunkehlchen (Saxicola rubetra) und schwarzkehlchen (S. torquata) im vergleich. J. Ornithol. 125, 261-278.
Sheldon, F. H., Whittingham, L. A., Moyle, R. G., Slikas, B. & Winkler, D. W. 2005 Phylogeny of swallows (Aves: Hirundinidae) estimated from nuclear and mitochondrial DNA sequences. Mol. Phylogenet. Evol. 35, 254-270.
Smith, G. T., Brenowitz, E. A., Wingfield, J. C. & Baptista, L. F. 1995 Seasonal changes in song nuclei and song behavior in Gambel's white-crowned sparrows. J. Neurobiol. 28, 114-125.
Spencer, K. A., Buchanan, K. L., Leitner, S., Goldsmith, A. R. & Catchpole, C. K. 2005 Parasites affect song complexity and neural development in a songbird. Proc. R. Soc. B 272, 2037-2043.
Striedter, G. F. 2005 Principles of Brain Evolution. Sunderland, MA: Sinauer Associates, Inc.
Svec, L. A., Licht, K. M. & Wade, J. 2009 Pair bonding in the female zebra finch: a potential role for the nucleus taeniae. Neuroscience 160, 275-283.
58

Székely, A. D. & Krebs, J. R. 1996 Efferent connectivity of the hippocampal formation of the zebra finch (Taenopygia guttata): an anterograde pathway tracing study using Phaseolus vulgaris leucoagglutinin. J. Comp. Neurol. 368, 198-214.
Székely, T., Catchpole, C. K., DeVoogd, A., Marchl, Z. & DeVoogd, T. J. 1996 Evolutionary changes in a song control area of the brain (HVC) are associated with evolutionary changes in song repertoire among European warblers (Sylviidae). Proc. R. Soc. Lond. B 263, 607-610.
Thompson, R. R., Goodson, J. L., Ruscio, M. G. & Adkins-Regan, E. 1998 Role of the archistriatal nucleus taeniae in the sexual behavior of male Japanese quail (Coturnix japonica): A comparison of function with the medial nucleus of the amygdala in mammals. Brain Behav. Evol. 51, 215-229.
Timmermans, S., Lefebvre, L., Boire, D. & Basu, P. 2000 Relative size of the hyperstriatum ventrale is the best predictor of feeding innovation rate in birds. Brain Behav. Evol. 56, 196-203.
Vicario, D. S. 1991 Organization of the zebra finch song control system: II. Functional organization of outputs from nucleus robustus archistriatalis. J. Comp. Neurol. 309, 486-494.
Voelker, G. & Spellman, G. M. 2004 Nuclear and mitochondrial DNA evidence of polyphyly in the avian superfamily Muscicapoidea. Mol. Phylogenet. Evol. 30, 386-394.
Vu, E. T., Mazurek, M. E. & Kuo, Y. C. 1994 Identification of a forebrain motor programming network for the learned song of zebra finches. J. Neurosci. 14, 6924-6934.
Wahlström, S. 1966 En akustisk jämförelse mellan sången hos tre olika Locustellaarter. Vår Fågelvärld 25, 161-166.
Wallschläger, D. 1964 A bioacoustical contribution to the systematics of th palearctic Motacillidae. I. Songs and call-notes of the genus Anthus. Mitt. Zool. Mus. Berl. (Suppl.) 60, 37-56.
Ward, B. C., Nordeen, E. J. & Nordeen, K. W. 2001 Anatomical and ontogenetic factors producing variation in HVc neuron number in zebra finches. Brain Res. 904, 318-326.
59

Whiten, A. & Byrne, R. W. 1988 The Machiavellian intellect hypothesis. In Machiavellian Intelligence (ed. R. W. Byrne & A. Whiten), pp. 1-9. Oxford: Oxford University Press.
Wild, J. M. 1993 Descending projections of the songbird nucleus robustus archistriatalis. J. Comp. Neurol. 338, 225-241.
Wildenthal, J. L. 1965 Structure in primary song of the mockingbird (Mimus polyglottos). Auk 82, 161-189.
Wink, M., Sauer-Gürth, H. & Gwinner, E. 2002 Evolutionary relationships of stonechats and related species inferred from mitochondrial-DNA sequences and genomic fingerprinting. Brit. Birds 95, 349-355.
Woolley, S. M. N. & Casseday, J. H. 2005 Processing of modulated sounds in the zebra finch auditory midbrain: responses to noise, frequency sweeps, and sinusoidal amplitude modulations. J. Neurophysiol. 94, 1143-1157.
Yu, A. C. & Margoliash, D. 1996 Temporal hierarchical control of singing in birds. Science 273, 1871-1875.
Yuri, T. & Mindell, D. P. 2002 Molecular phylogenetic analysis of Fringillidae, "new world nine-primaried oscines" (Aves: Passeriformes). Mol. Phylogenet. Evol. 23, 229-243.
Zamora, J., Lowy, E., Ruiz-del-Valle, V., Moscoso, J., Serrano-Vela, J. I., Rivero-deAguilar, J. & Arnaiz-Villena, A. 2006 Rhodopechys obsoleta (desert finch): a pale ancestor of greenfinches (Carduelis spp.) according to molecular phylogeny. J. Ornithol. 147, 448-456.
Zeng, S. J., Lu, K., Hua, F. Y., Zhao, W. L., Zhang, X. W. & Luo, M. X. 2005 Song complexity is associated with the volume of brain song-control nuclei in ten oscine species. Acta Zool. Sinica 51, 68-75.
60

CHAPTER 3
EVOLUTION OF FUNCTIONAL NEURAL CIRCUITS IN THE OSCINE BRAIN Abstract Two models of brain evolution have been proposed to explain the source of changes in brain composition. The model of developmental conservation asserts that neurogenetic schedules are conserved between species but are altered in duration. As a result, late-developing areas become disproportionately large as overall brain size increases because the number of neuronal precursors increases exponentially. The model of mosaic evolution focuses on coordinated changes between regions, and specifically predicts that functionally connected areas change in concert but independently of other areas. The present study tests predictions of the two models at the level of functional neural circuits across a wide phylogeny of songbirds. Interestingly, the data support elements of both. Telencephalic nuclei develop later in ontogeny and possess greater allometric slopes than nuclei in the thalamus and brainstem. Functionally connected nuclei evolve in concert as a result of coordinated changes in relative neuron number. Finally, the song system is much more variable than other systems, which may be related to its late development relative to surrounding tissues. Introduction
The vertebrate brain comes in many shapes and sizes, and little is currently known about the evolutionary processes that give rise to these various forms. In particular, considerable debate has centered on differences in brain composition between large and small animals and how differences may relate to ecological adaptations and/or phylogenetic relationships. This research has focused on two main arguments. The ‘developmental conservation’ hypothesis asserts that brain structure is
61

largely determined by its sequence of development, which supports graceful scaling of brain components with brain and body size (Finlay & Darlington 1995; Finlay et al. 2001). According to this view, neurogenesis schedules for different brain regions are conserved across species but altered in duration, therefore compositional differences between large and small brains are predictable from their developmental program (Clancy et al. 2001). Behavioral evolution is therefore thought to occur within the context of these conserved developmental mechanisms, which promote computationally efficient structural changes by means of graceful scaling and the allocation of new functions to favorable locations. By contrast, a ‘mosaic’ model argues that selective forces drive specific changes to individual systems and/or brain regions and that these forces are strong compared to those that might favor conserved developmental programs (Barton & Harvey 2000; de Winter & Oxnard 2001). This hypothesis contends that grade shifts, relatively large and abrupt changes in brain composition, occur regularly and can explain the majority of evolutionary changes in brain structure. While competing, these hypotheses are not exclusive. They instead differ in their assertions about the degree to which brain evolution is concerted or piecemeal and about the frequency with which large grade shifts occur (Striedter 2005).
Nearly this entire debate stems from measures of gross brain subdivisions in mammals. The case for concerted brain evolution is based on attempts to explain the covariation among these different components in terms of the total observed variation (Finlay & Darlington 1995; Reep et al. 2007). Not surprisingly, the vast majority of variation is explained by overall brain size. The proportional sizes of different regions are not constant across brains, however, and these differences are closely predicted by neurogenetic schedules. Areas that continue to develop late in ontogeny (e.g., telencephalon) possess larger allometric slopes and thus constitute a greater proportion
62

of larger brains. The olfactory bulb and limbic structures (e.g., subicular cortices, hippocampus, septum) deviate from this general rule but covary with each other, an effect that could arise from a common developmental source such as the expansion or contraction of prosomeric boundaries (Puelles & Rubenstein 2003; Reep et al. 2007). The remaining variance left unexplained by these two factors (~1%) might be associated with behavioral specializations in individual species, which permits up to ~2.5-fold differences in species of similar size. Thus, the model of concerted brain evolution seeks to explain the total variation in brain composition and identifies developmental events, such as the timing of neurogenesis and/or the size of neuronal precursor pools, as its primary determinants. It further contends that graceful scaling of brain regions is a generally more efficient means to effect behavioral change than the independent restructuring of individual brain components.
The second approach has focused on differential changes in specific brain areas or brain composition and their relationships to behavioral or ecological specializations (e.g., de Winter & Oxnard 2001). The literature is filled with examples of brain-behavior relationships that span sensory (Welker et al. 1964; Kubke et al. 2004), motor (DeVoogd et al. 1993; Yopak et al. 2007), and even fairly abstract cognitive capacities (Krebs et al. 1989; Emery & Clayton 2004; Lefebvre et al. 2004). Moreover, evolutionary changes in brain structure reflect its connectivity, with interconnected regions changing in parallel but seemingly independently of other areas (Barton & Harvey 2000; Whiting & Barton 2003). Of central importance, then, is the magnitude of these behavior-related differences and the frequency with which largescale changes occur. Clear examples of mosaic changes exist, including large isocortical volume differences between primates and insectivores (Stephan et al. 1981) and the nearly 10-fold difference in superior colliculus volumes between the ground squirrel and rat (Woolsey et al. 1971). Such large differences have usually been
63

limited to distantly related species, however, as brain composition appears quite stable within closely related taxa (Clark et al. 2001). Analyses in other groups of animals can shed light on the frequency with which these large changes occur and the applicability of either model across vertebrates.
Recent studies on the avian brain have highlighted several instances of mosaic evolution. Brain composition varies widely between species and often reflects both phylogeny and behavioral and/or ecological specializations (Boire & Baron 1994; Burish et al. 2004; Iwaniuk & Hurd 2005). These large differences may arise through any of several changes to developmental programs, including delayed and/or prolonged periods of neurogenesis, increased progenitor cell cycle rates, and expanded proliferative cell regions (Charvet & Striedter 2008, 2009a; Striedter & Charvet 2008). Interestingly, the strategies employed reflect developmental mode, with altricial but not precocial species extending brain development (Charvet & Striedter 2009b). Thus, large grade shifts are evident between avian orders and multiple differences in developmental programs appear to underlie these changes. Evolutionary changes within orders have not been characterized, however, therefore it remains unknown if gradual scaling exists in birds as in mammals.
To the best of our knowledge, no study has actually measured the evolutionary patterns of discrete functional circuits with respect to these questions. The present study does so in songbirds, whose nuclear (rather than laminar) pallial organization facilitates the measurement of telencephalic areas with known connectivity and function. It also provides an opportunity to ascertain the cellular basis of correlations between connected brain regions (Barton & Harvey 2000; Whiting & Barton 2003; Iwaniuk et al. 2004). These relationships could result from any of three causes, including (1) changes in cell density of one region due to more or fewer axons from afferent sources, (2) trophic-induced cell growth resulting from increased afferent
64

activity, or (3) coordinated changes in neuron number (Finlay et al. 2001). While the third is commonly assumed to be the primary cause, this has not been explicitly shown. Finally, songbirds are capable of many diverse and complex behaviors that include learned songs (Marler & Slabbekoorn 2004), intricate social interactions (Burish et al. 2004), food caching and tool use (Emery & Clayton 2004), and various innovations to cope with variable seasonal and environmental demands (Lefebvre et al. 2004). This study examines whether functional pathways in the songbird brain exhibit similar amounts of diversity. Methods (a) Specimen collection and preparation
A portion of the data presented here were the subject of a previous study (Moore et al. 2009), and new measurements were obtained from the same brains. Specimens include 58 temperate zone species that were wild-caught under permit with mist nets during spring months (April – June) when in reproductive condition. At the time of capture, birds were deeply anesthetized with a barbiturate and transcardially perfused with 0.8% saline followed by 10% formalin in saline. Brains were extracted, post-fixed for at least 24 hours, cryoprotected with 30% sucrose/10% formalin, embedded in gelatin, and sectioned at 40 μm in the coronal plane with a freezing, sliding microtome. Sections were then mounted onto gel-coated slides and Nisslstained with cresyl violet. All procedures were approved by the Cornell University Institutional Animal Care and Use Committee. (b) Brain measurements
In total, 23 discrete nuclei from 4 functional systems (song, limbic, tectofugal visual, ascending auditory) and spanning 4 major brain subdivisions (telencephalon, diencephalon, midbrain, hindbrain) were measured (Figure 3.1). Song nuclei included HVC (abbreviation used as proper name) in the nidopallium, robust nucleus of the
65

Figure 3.1. Diagram of major connections between nuclei in the (a) song system, (b), limbic system, (c) tectofugal visual system, and (d) ascending auditory system. Nuclei that were measured are shown in black while those that were not are in gray.
66

67

arcopallium (RA), Area X of the striatum, the lateral and medial magnocellular nuclei of the nidopallium (LMAN and MMAN, respectively), uvaeformis of the thalamus (Uva), the dorsomedial nucleus of the intercollicular complex (DM) in the midbrain, and the tracheosyringeal portion of the hypoglossal nucleus (nXIIts) in the hindbrain. Major projections of the song system are well established and nuclear boundaries were unambiguous (Nottebohm et al. 1976; 1982; Moore et al. 2009). Nuclei shown in Figure 3.1 but not measured include the nucleus interface of the nidopallium (NIf), the dorsomedial nucleus of the posterior thalamus (DMP), the medial portion of the dorsolateral nucleus of the anterior thalamus (DLM), and nuclei within the ventral respiratory group (VRG).
Limbic structures included the hippocampus (Hp), septum (Sep), and nucleus taeniae of the amygdala (TnA), all of which are in the telencephalon and interconnected (Székely & Krebs 1996; Cheng et al. 1999; Atoji & Wild 2004). The lateral extent of Hp was identified by its lower cell density compared to the adjacent parahippocampal area. Caudal Hp does not have clear boundaries, therefore measurements were arbitrarily stopped at the section in which the cerebellum reached the dorsal-most extent of the telencephalon. Sep boundaries were identified by differences in staining intensity and largely coincided with those described by chemoarchitecture (Goodson et al. 2004); the estimates reported here include most of the four major subdivisions but exclude portions of the nucleus of the diagonal band ventral to the septopallio-mesencephalic tract. TnA was delineated from surrounding arcopallium on the basis of its high cell density. Nuclei shown in Figure 3.1 but not measured are the medial preoptic area (POM), ventral tegmental area (VTA), and locus coeruleus (LoC).
Tectofugal visual nuclei were MVL (abbreviation used as a proper name) of the mesopallium, entopallium (E; including the perientopallial belt, Ep) in the
68

nidopallium, nuclei rotundus (Rt), pretectalis (Pt), and spiriformis lateralis (SpL) of the thalamus, and subpretectalis (SP) and nucleus isthmi pars semilunaris (SLu) of the midbrain. Connectivity between these areas has been described in detail (Hellmann et al. 2001; Theiss et al. 2003; Krützfeldt & Wild 2004). MVL was clearly distinguishable from surrounding mesopallium based on its small somata and high cell density, and E boundaries were determined from the relatively high staining intensity of the Ep belt. Boundaries of thalamic and mesencephalic nuclei were unambiguous. Areas shown in Figure 3.1 but not measured include the lateral striatum (LSt), nucleus isthmi pars parvocellularis (Ipc), nucleus isthmi pars magnocellularis (Imc), and the tectum (Tec).
Finally, nuclei along the ascending auditory pathway included nucleus ovoidalis (Ov) of the thalamus, mesencephalicus lateralis, dorsalis (MLd) of the midbrain, and the superior olivary (SON), laminar (NL), and magnocellular (NM) nuclei of the hindbrain (Carr & Konishi 1990; Vates et al. 1996; Burger et al. 2005). Telencephalic auditory areas [e.g., Field L, caudomedial nidopallium (NCM) or caudal mesopallium (CM)] do not have clear boundaries and could not be measured. MLd was distinguished from the surrounding intercollicular complex on the basis of cell density and staining intensity, all other areas were unambiguous. Nuclei shown in Figure 3.1 but not measured were Field L of the nidopallium, the lateral lemniscus (LL), and nucleus angularis (NA).
Alternate sections were viewed under 40× or every fourth section under 20× (Hp and E) magnification and nuclei were traced using a camera lucida. Unmagnified digital images of every fourth section were used to measure the telencephalon and entire brain. All measurements were made in one side of the brain, typically the left except in cases where torn tissue or poor staining precluded their measurement, in which case the entire structure was measured in the right. Cross-sectional areas of
69

scanned traces and brain images were measured using NIH ImageJ software (Rasband 2007) and final volumes were computed by summing the areas and multiplying by the sampling interval (0.08 or 0.16 mm).
Neuron densities were estimated in 11 nuclei from three systems. Presumptive neurons were discriminated from glia on the basis of their soma size, uniformly stained cytoplasm, and a single, darkly stained nucleolus. Nucleolus counts were tallied within sampling windows that were evenly distributed throughout each plane of each structure, and a cell was only counted if its nucleolus was within the grid. For nuclei RA and Rt, sampling grid dimensions were 120 × 120 μm (400×); for all other nuclei (HVC, LMAN, MMAN, nXIIts, Hp, Sep, E, SpL, SP), grid dimensions were 80 × 80 μm (600×). On average, 12 (range: 7-26) samples were collected from each region, and overall neuron density estimates were calculated by averaging the neuron counts and dividing by the volume sampled. One specimen from each species was sampled for three song nuclei (HVC, RA, nXIIts) whereas one specimen from approximately half of the species was sampled for all others. Species were chosen to maximize differences in overall brain size and relative nucleus volumes. (c) Data analysis
Phylogenetically-based statistical analyses were used in all cases. A fully resolved composite phylogeny was constructed from published molecular studies and was the same as that used previously (Moore et al. 2009). For analyses of log10transformed data, the phylogenetic tree was scaled using the arbitrary branch lengths method of Grafen (1989) and the rho transform (ρ = 0.3) based on this tree’s superior performance in diagnostic tests (Garland et al. 1992; Blomberg et al. 2003). Allometric slopes and confidence intervals were estimated using a generalized least squares (GLS) approach, which is equivalent to analyses of independent contrasts when performed with the same tree (Garland & Ives 2000). To measure the allometric
70

slope for the telencephalon, the independent variable was brain volume minus telencephalon; for each nucleus it was brain volume minus the sum of all nuclei.
A factor analysis was performed on standardized independent contrasts of log10-transformed data to ascribe portions of variance to common components. Because factor analysis centers variables at their mean and contrasts must be centered at 0, contrasts for each variable were entered twice and the second entry was multiplied by -1, thus giving each a mean = 0 (and n = 2*57 = 114, therefore no significance tests were performed). For the subsequent model construction, axes were rotated using an oblique Oblimin rotation.
Finally, correlations between nuclei were calculated in two ways. Relationships between two nuclei were measured by constructing two-factor GLS models that explained the variation in one nucleus as a function of the other with size (i.e. brain – nuclei) as a confounding variable. The same form of model was used for analyses of neuron numbers and densities, and trees were pruned and re-scaled prior to those for which not every species was sampled. Multivariate models were also constructed whereby the variation in the volume of one nucleus was explained as a function of all others within that functional circuit. Finally, phylogenetic signal was measured in both log-transformed and relative trait volumes. The latter were computed with the formula log10[trait/(sizeb)], where b was the allometric exponent, and ‘trait’ and ‘size’ were original data values of a nucleus and size reference, respectively.
Mesquite v2.5 (Maddison & Maddison 2008) was used to manage data and trees, and the PDAP:PDTree v1.14 module was used for all analyses of independent contrasts (Garland et al. 1999; Garland & Ives 2000). Phylogenetic signal was computed with the Matlab program PHYSIGLL.m (Blomberg et al. 2003) and phylogenetic generalized least squares (PGLS) models were constructed with
71

REGRESSIONv2.m (Lavin et al. 2008). For these analyses, trees were first converted to variance-covariance matrices with the PDDIST module of Phenotypic Diversity Analysis Programs (Garland & Ives 2000). Results (a) Allometry
All log-transformed and most relative nucleus volumes exhibited significant phylogenetic signal, the latter of which are listed in Table 3.1, indicating that the data were better described by the hierarchical than star phylogeny. The telencephalon displayed hyperallometry relative to the rest of the brain (slope = 1.08; SE = 0.06; Table 3.2). This was also evident at the level of individual nuclei, where those in the telencephalon had significantly greater slopes than nuclei within the thalamus or brainstem (Figure 3.2; t-test, p < 0.001). (b) System variability
A factor analysis was performed on independent contrasts of log10-transformed nucleus volumes, and two factors explained 81% of the total variance (Table 3.3a). Most nuclei loaded highly on the first, which explained 70.7% of the total variance and primarily reflects changes in overall brain size (Table 3.3b). However, the majority of song nuclei loaded more strongly on a second factor that accounted for an additional 10.5% of the total variance. The two factors were correlated (r = 0.59), suggesting that a ‘size’ component was present in the ‘song system’ factor but that much of the variability associated with this component was not attributable to overall brain size. This disparity between the song system and the three others is also evident from correlations between system volumes, where the value for each was calculated by summing the raw values of their respective nuclei. Correlations between the song system and each system (0.55 ≤ r ≤ 0.64) were notably weaker than those between the other systems (0.78 ≤ r ≤ 0.88).
72

Table 3.1. Phylogenetic signal of relative trait volumes. Tree branch lengths were scaled using the method of Grafen (1989) and the rho transform (ρ = 0.3). In all cases, df = 56. For additional discussion of phylogenetic signal, see Blomberg et al. (2003) and Revell et al. (2008).
73

74

Trait
rel_Telen
rel_HVC rel_RA rel_Area X rel_LMAN rel_MMAN rel_Uva rel_DM rel_nXIIts
rel_Hp rel_Sep rel_TnA
rel_MVL rel_E rel_Rt rel_Pt rel_SpL rel_SP rel_SLu

MSE0
0.0063
0.0831 0.0296 0.0351 0.0373 0.0726 0.0051 0.0070 0.0210
0.0084 0.0052 0.0103
0.0130 0.0095 0.0099 0.0149 0.0110 0.0126 0.0082

MSEstar
0.0061
0.0828 0.0294 0.0351 0.0370 0.0724 0.0051 0.0069 0.0209
0.0077 0.0048 0.0101
0.0130 0.0095 0.0099 0.0149 0.0110 0.0126 0.0082

MSE
0.0063
0.0603 0.0292 0.0334 0.0381 0.0469 0.0064 0.0055 0.0253
0.0066 0.0044 0.0081
0.0120 0.0076 0.0079 0.0108 0.0116 0.0109 0.0085

obs.
MSE0/MSE 0.9911
1.3772 1.0168 1.0519 0.9774 1.5477 0.7984 1.2745 0.8308
1.2593 1.1780 1.2699
1.0872 1.2523 1.2514 1.3782 0.9478 1.1622 0.9688

exp.
MSE0/MSE 1.5183
1.5183 1.5183 1.5183 1.5183 1.5183 1.5183 1.5183 1.5183
1.5183 1.5183 1.5183
1.5183 1.5183 1.5183 1.5183 1.5183 1.5183 1.5183

K
0.6528
0.9071 0.6697 0.6928 0.6438 1.0194 0.5259 0.8395 0.5472
0.8294 0.7759 0.8364
0.7161 0.8248 0.8242 0.9078 0.6242 0.7655 0.6381

p
0.020 <0.001
0.007 0.002 0.015 <0.001 0.389 <0.001 0.280
<0.001 0.003
<0.001 0.001
<0.001 <0.001 <0.001
0.020 <0.001
0.019

ln lklhd
65.0298
-0.3677 20.7243 16.7779 12.9325
6.9353 64.5757 69.2888 24.8271
63.6107 75.6655 57.7920
46.5562 59.8981 58.5218 49.5505 47.4413 49.3009 56.5504

ln
lklhdstar 65.9427
-9.5339 20.4429 15.3626 13.8317 -5.6416 71.2047 62.3579 30.3284
59.1881 72.8376 51.3859
44.1330 53.3788 52.0188 40.2648 49.0100 45.0030 57.6091

75

Table 3.1 (Continued).

Trait
rel_Ov rel_MLd rel_SON rel_NL rel_NM

MSE0
0.0055 0.0088 0.0041 0.0063 0.0045

MSEstar
0.0055 0.0088 0.0039 0.0060 0.0044

MSE
0.0064 0.0094 0.0038 0.0048 0.0050

obs.
MSE0/MSE 0.8620 0.9372 1.0841 1.3201 0.8870

exp.
MSE0/MSE 1.5183 1.5183 1.5183 1.5183 1.5183

K
0.5678 0.6173 0.7140 0.8695 0.5842

p
0.149 0.024 0.003 <0.001 0.096

ln lklhd
64.6996 53.5522 79.6769 73.0195 71.6187

ln
lklhdstar 69.2641 55.4625 78.8400 66.5477 75.4874

Table 3.2. Parameters and statistics of allometric equations from GLS regressions of the telencephalon and each nucleus. The size reference for the telencephalon was log(Brain – Telen) and for each nucleus was log(Brain – nuclei). In all cases, df = 56.

Nucleus
log(Telen) log(HVC) log(RA) log(Area X) log(LMAN) log(MMAN) log(Uva) log(DM) log(nXIIts) log(Hp) log(Sep) log(TnA) log(MVL) log(E) log(Rt) log(Pt) log(SpL) log(SP) log(SLu) log(Ov) log(MLd) log(SON) log(NL) log(NM)

r2 coef.

0.846
0.303 0.517 0.476 0.416 0.339 0.728 0.751 0.428
0.833 0.855 0.710
0.744 0.803 0.749 0.691 0.656 0.620 0.751
0.790 0.604 0.793 0.731 0.744

1.083
0.852 0.930 0.917 0.867 0.816 0.690 0.676 0.724
0.960 0.846 0.742
0.981 0.922 0.810 0.818 0.782 0.702 0.840
0.817 0.630 0.636 0.602 0.638

SE
0.062 0.173 0.120 0.129 0.137 0.152 0.056 0.052 0.112 0.057 0.047 0.063 0.077 0.061 0.063 0.073 0.076 0.073 0.065 0.056 0.068 0.043 0.049 0.050

95% Confidence
Interval
lower upper 0.981 1.212
0.486 1.140 0.688 1.170 0.668 1.146 0.599 1.138 0.540 1.122 0.584 0.802 0.575 0.779 0.503 0.914
0.850 1.069 0.761 0.933 0.624 0.868
0.825 1.121 0.806 1.049 0.686 0.929 0.663 0.947 0.631 0.928 0.555 0.837 0.704 0.960
0.716 0.931 0.506 0.757 0.554 0.725 0.505 0.695 0.544 0.730

int.
0.094 -2.447 -3.003 -2.282 -3.372 -3.155 -2.980 -2.913 -3.291 -1.656 -1.952 -2.078 -3.218 -1.799 -2.043 -3.405 -2.793 -2.744 -3.345 -3.205 -2.064 -2.954 -2.795 -2.806

SEE
0.080 0.248 0.172 0.184 0.197 0.219 0.081 0.075 0.161 0.082 0.067 0.091 0.110 0.088 0.090 0.105 0.109 0.105 0.093 0.081 0.098 0.062 0.070 0.072

76

Figure 3.2. Nuclei within the telencephalon (black) have greater allometric slopes than those in the thalamus or brainstem (gray). Note that slopes from GLS models are identical to those from regressions of independent contrasts when the latter are forced through the origin.
77

Table 3.3. (a) First five eigenvalues and percent total variance explained from a factor analysis of independent contrasts of log10-transformed data. (b) Factor loadings for independent contrasts of each nucleus in a two-factor model.

Factor
1 2 3 4 5
Nucleus
HVC RA Area X LMAN MMAN Uva DM nXIIts Hp Sep TnA MVL E Rt Pt SpL SP SLu Ov MLd SON NL NM

Eigenvalue
16.253 2.416 0.737 0.641 0.583
Factor 1
-0.099 0.119 0.066 0.052 -0.024 0.543 0.703 0.273 0.900 0.833 0.626 0.960 1.021 1.007 1.015 0.951 0.991 0.945 0.765 0.727 0.805 0.784 0.746

% Variance
70.667 10.503 3.205 2.785 2.535
Factor 2
0.952 0.864 0.900 0.847 0.849 0.448 0.265 0.616 0.022 0.114 0.268 -0.083 -0.095 -0.094 -0.168 -0.061 -0.154 -0.010 0.229 0.178 0.172 0.160 0.239

78

The pronounced variability of the song system is further underscored by comparisons between species of similar size (Figure 3.3). For example, the spotted flycatcher (Muscicapa striata) has an HVC volume that is 15-fold larger than that of the common yellowthroat (Geothlypis trichas); the common starling (Sturnus vulgaris) has an RA volume that is 5-fold larger than that of the blue jay (Cyanocitta cristata); and the sand martin (Riparia riparia) has an LMAN volume that is nearly 5fold larger than that of the black-capped chickadee (Poecile atricapillus) despite its substantially smaller brain volume. By comparison, maximum differences across other nuclei rarely exceeded a factor of 3. The relatively large Hp of the black-capped chickadee, for example, is only 2.8-times larger than that of the European goldfinch (Carduelis carduelis), which had one of the smallest relative Hp volumes measured. (c) Relationships between and within systems
Relationships between systems were assessed by first adding the volumes of all nuclei within one system and then testing for correlations between them with size [i.e. log(brain – nuclei)] included as a confounding variable. After accounting for size, the volume of the tectofugal visual system was positively correlated with the limbic structures (r = 0.30, p = 0.02) and was nearly significantly related to the auditory system (r = 0.25, p = 0.06). In contrast, the volume of the song system was inversely correlated with the limbic structures (r = -0.25, p = 0.04) and was nearly so with the visual system (r = -0.25, p = 0.06). These associations involving the song system were investigated in more detail. Significant relationships existed between Hp and RA (r = -0.43, p = 0.0008), LMAN (r = -0.28, p = 0.03), and nXIIts (r = -0.37, p = 0.005) and between Sep and Area X (r = -0.27, p = 0.04) and nXIIts (r = -0.37, p = 0.005). They were also present between MVL and LMAN (r = -0.30, p = 0.02) and E and Area X (r = -0.29, p = 0.03).
79

Figure 3.3. Variability in allometric relationships for several nuclei across systems. Note that the plots depict log-transformed data and not independent contrasts for illustrative purposes. Relative song nucleus volumes (a –c) were much more variable than those from the limbic (d), visual (e), or auditory (f) systems.
80

The volumes of nuclei within each system were positively and strongly correlated after accounting for size in most cases, and the strongest correlations were generally between directly connected areas (Tables 3.4-3.7). Significant associations existed between most song nuclei, including all telencephalic areas (all p ≤ 0.001) and between those areas and both Uva and nXIIts (all p ≤ 0.01). Nucleus DM was an exception and was not related to any areas but had nearly significant relationships with HVC and Uva (0.05 ≤ p ≤ 0.09). Among limbic structures, Hp and Sep were positively correlated with each other (r = 0.50, p = 6.0 × 10-5) but neither was related to TnA (both p > 0.65). All visual nuclei (all p < 0.03) and all but one pair of auditory nuclei (Ov and NM, p = 0.09; all other p < 0.04) were significantly correlated.
The cellular basis of these relationships was investigated by estimating neuron densities and numbers in 11 nuclei within the song, limbic, and visual systems. Densities in all 11 were inversely and strongly related to overall brain size (all p < 0.0001). For all areas except the Sep, relative neuron number was strongly and positively correlated with relative nucleus volume (Table 3.8; 0.54 ≤ r2 ≤ 0.98, all p < 0.0001) whereas neuron density was inversely related to relative volume for only 5 of these 10 (RA, MMAN, nXIIts, Hp, Rt; all p < 0.03). In each case, density explained a much smaller proportion of relative volume variation (0.13 ≤ r2 ≤ 0.36) than did neuron number. The opposite was true for relative Sep volume. Here, a smaller amount of the variation in relative volume was explained by relative neuron number (r2 = 0.12, p = 0.048) than neuron density (r2 = 0.27, p = 0.002). Finally, most of the correlations between relative nucleus volumes also existed beween their relative neuron numbers, though these relationships did tend to lose strength (Tables 3.9 and 3.10). Two relationships lost statistical significance; relative neuron numbers were not correlated between HVC and nXIIts (r = 0.21, p = 0.11) nor between Hp and Sep (r = 0.34, p = 0.051).
81

Table 3.4. Regression matrices between the log10-transformed song nucleus volumes. The size factor log(brain – nuclei) was a confounding variable in each regression and statistics refer to the partial correlations. Directly connected nuclei are indicated by shaded cells. In all cases, df = 55.

82

HVC

RA Area X LMAN MMAN

r 0.709

RA t 7.453

p 6.82E-10

r

0.684

0.690

Area X t

6.951

7.079

p 4.54E-09 2.81E-09

r

0.498

0.689

0.710

LMAN t

4.258

7.048

7.471

p 8.12E-05 3.15E-09 6.39E-10

r

0.744

0.459

0.570

0.394

MMAN t

8.269

3.832

5.144

3.178

p 3.18E-11 3.28E-04 3.70E-06

0.002

r

0.340

0.420

0.451

0.408

0.050

Uva

t

2.679

3.436

3.744

3.311

0.375

p

0.010

0.001 4.36E-04

0.002

0.709

r

0.241

0.168

0.181

0.073

0.151

DM

t

1.841

1.266

1.363

0.544

1.130

p

0.071

0.211

0.179

0.588

0.264

r

0.332

0.621

0.496

0.497

0.174

nXIIts

t

2.610

5.880

4.242

4.249

1.308

p

0.012 2.52E-07 8.57E-05 8.37E-05

0.196

Uva
0.233 1.778 0.081 0.103 0.770 0.444

DM
0.150 1.127 0.265

Table 3.5. Regression matrices between the log10-transformed volumes of limbic structures. The size factor log(brain – nuclei) was a confounding variable in each regression and statistics refer to the partial correlations. Directly connected nuclei are indicated by shaded cells. In all cases, df = 55.

83

Sep TnA

Hp r 0.508 t 4.370 p 5.57E-05 r 0.019 t 0.141 p 0.888

Sep
-0.045 -0.334 0.739

Table 3.6. Regression matrices between the log10-transformed tectofugal visual nucleus volumes. The size factor log(brain – nuclei) was a confounding variable in each regression and statistics refer to the partial correlations. Directly connected nuclei are indicated by shaded cells. In all cases, df = 55.

84

MVL

E

Rt

Pt SpL SP

r 0.642

E t 6.215

p 7.21E-08

r

0.580

0.825

Rt t 5.274 10.837

p 2.32E-06 2.90E-15

r

0.464

0.718

0.703

Pt

t

3.889

7.658

7.322

p 2.74E-04 3.15E-10 1.12E-09

r

0.306

0.514

0.483

0.557

SpL

t

2.381

4.445

4.088

4.975

p 0.021 4.30E-05 1.43E-04 6.76E-06

r

0.366

0.745

0.773

0.673

0.531

SP

t

2.912

8.289

9.026

6.740

4.650

p 0.005 2.96E-11 1.92E-12 1.00E-08 2.12E-05

r

0.304

0.523

0.621

0.513

0.558

0.555

SLu

t

2.365

4.553

5.869

4.435

4.989

4.946

p 0.022 2.97E-05 2.61E-07 4.46E-05 6.44E-06 7.49E-06

Table 3.7. Regression matrices between the log10-transformed auditory nucleus volumes. The size factor log(brain – nuclei) was a confounding variable in each regression and statistics refer to the partial correlations. Directly connected nuclei are indicated by shaded cells. In all cases, df = 55.

85

MLd SON NL NM

Ov

MLd

SON

NL

r 0.588

t 5.387

p 1.53E-06

r

0.321

0.413

t

2.513

3.368

p

0.015

0.001

r

0.273

0.493

0.518

t

2.106

4.202

4.489

p 0.040 9.79E-05 3.71E-05

r

0.232

0.438

0.385

0.600

t

1.771

3.613

3.093

5.556

p

0.082

0.001

0.003 8.29E-07

Table 3.8. Regressions between log10-transformed nucleus volumes and neuron number or density of that nucleus. The size factor log(brain – nuclei) was a confounding variable in each regression and statistics refer to the partial correlations.

Volume HVC RA LMAN MMAN nXIIts Hp Sep E Rt SpL SP

neuron #
r2 0.977 t 48.695 p <0.0001 r2 0.873 t 19.638 p <0.0001 r2 0.875 t 16.545 p <0.0001 r2 0.901 t 18.800 p <0.0001 r2 0.734 t 12.427 p <0.0001 r2 0.605 t 7.003 p <0.0001 r2 0.117 t 2.060 p 0.048 r2 0.594 t 6.161 p <0.0001 r2 0.544 t 5.573 p <0.0001 r2 0.663 t 7.152 p <0.0001 r2 0.713 t 8.040 p <0.0001

neuron density
0.037 -1.469 0.147
0.305
-4.956
<0.0001 0.059 -1.543 0.131
0.134
-2.457
0.019
0.257
-4.406
<0.0001
0.355
-4.131
3.00E-04
0.268
-3.422
0.002 0.049 -1.163 0.255
0.162
-2.238
0.034 0.022 0.761 0.454 0.020 -0.729 0.473

86

Table 3.9. Regression matrix of log10-transformed neuron numbers between various song nuclei. The size factor log(brain – nuclei) was a confounding variable in each regression and statistics refer to the partial correlations.

RA n# LMAN n# MMAN n# nXIIts n#

r t p r t p r t p r t p

HVC n#
0.612
5.796
<0.0001
0.488
3.494
0.001
0.785
7.925
<0.0001 0.214 1.638 0.107

RA n# LMAN n# MMAN n#

0.662 5.517 <0.0001 0.470 3.327 0.002 0.533 4.710 <0.0001

0.488 3.492 0.001 0.608 4.780 <0.0001

0.177 1.126 0.267

87

Table 3.10. Regression matrix of log10-transformed neuron numbers between various visual nuclei. The size factor log(brain – nuclei) was a confounding variable in each regression and statistics refer to the partial correlations.

Rt n# SpL n# SP n#

E n#

Rt n#

SpL n#

r 0.657

t 4.441

p 1.00E-04

r

0.604

0.449

t

3.867

2.560

p

0.001

0.017

r

0.569

0.563

0.573

t

3.529

3.473

3.562

p

0.002

0.002

0.001

88

Discussion The evolution of the avian brain has resembled that of mammals in several
respects, and the data presented here reveal patterns that support both concerted and mosaic models of evolution. First, graceful scaling of brain subdivisions is present across species. The telencephalon exhibits hyperallometry relative to the rest of the brain and telencephalic nuclei have greater allometric slopes than those in the thalamus and brainstem. This is consistent with scaling patterns in mammals, in which the evolution of large brains has been suggested to result from a gradual stretching of conserved neurogenetic schedules (Finlay & Darlington 1995). While changes in the timing of neurogenesis do not entirely account for the large telencephalic fractions of songbirds, parrots, and waterfowl compared to other taxa (Striedter & Charvet 2008; Charvet & Striedter 2009a,b), it may still underlie the gradual differences within oscines. In the songbird, the telencephalic fraction is roughly equivalent to the tectum and medulla early in embryogenesis but expands disproportionately thereafter because its neural progenitors divide for longer periods of time (Charvet & Striedter 2009a). This growth appears to occur in two waves, one in ovo when new neurons are born throughout the entire brain and another after hatching, when they are almost exclusively added to the telencephalon (AlvarezBuylla et al. 1994; Charvet & Striedter 2009a). Thus, gradually stretching developmental schedules as brain size increases, especially that which occurs after hatching, would disproportionately expand the number of telencephalic progenitors and could cause the gradual increase in telencephalic fraction observed here.
Second, the relative volumes of connected nuclei are strongly correlated and these associations are primarily due to coordinated changes in cell number. Similar relationships are present between the volumes of large brain regions in both mammals and birds (Barton & Harvey 2000; Whiting & Barton 2003; Iwaniuk et al. 2004). The
89

relative contributions of neuron number and density to relative nucleus volume generally correspond to the degree of convergent input each receives from its afferent source(s). For areas that are larger or comparable in size to their afferents, such as HVC, E, and SpL, relative volume accurately reflects neuron number and is unrelated to neuron density. By contrast, the relative volumes of nuclei that are densely innervated by a larger afferent source, such as RA, nXIIts, Sep, and Rt, also reflect differences in neuron density. This effect is presumably the result of increases in the number of axons and perhaps trophic-induced cell growth. As a consequence, interspecific differences in the relative volumes of nuclei that receive such dense innervation seem prone to overestimate the true differences in neuronal composition.
Coordinated changes in neuron number are perhaps to be expected if basic circuit organization and physiological mechanisms are conserved within orders. The developmental mechanisms that produce these correlations are less clear. At one extreme, brain development could be mosaic in that alterations in the regional boundaries, progenitor cell cycle rate, and/or timing of neurogenesis could be independently regulated in individual nuclei. Alternatively, cell proliferation and neurogenesis schedules may be broadly specified within larger subdivisions or regions, and new neurons could be differentially recruited into various nuclei as overall brain size changes. In either case, selection may then favor phenotypes in which nuclei comprising a functional pathway have changed together. It seems likely that activity-dependent mechanisms and/or epigenetic cascades also serve some role in this coordination given that correlations exist between widely distributed nuclei and that numerous molecular pathways are involved in neural tube regional specification (Garcia-Lopez et al. 2009). More descriptions of developmental differences between large and small species are needed to test these possibilities. It is clear, however, that correlations between connected areas cannot distinguish between them and that these
90

associations can exist within the context of scaling differences between subdivisions; therefore they do not, by themselves, provide clear evidence for mosaic brain evolution.
The most striking feature of this data set is the immense variability of the song system compared to other pathways. This is most obvious in nuclei HVC and MMAN, but even the relative volumes of RA, Area X, LMAN, and nXIIts vary by nearly twice as much as all other nuclei measured. To be sure, direct comparisons across nuclei are difficult because they differ in their connectivity, intrinsic circuitry, and physiology, all of which will influence their structure through evolution. Nevertheless, we are struck by the fact that nuclei within the three other pathways exhibit similar amounts of variation despite these differences and that they are all consistently and drastically less variable than the major song nuclei. This does not appear to reflect a lack of consequences when elaborated; hypertrophy of the hippocampus is associated with enhanced spatial memory (Krebs et al. 1989) and enlarged auditory brainstem nuclei are related to superior sound localization and perceptual abilities (Kubke et al. 2004). Interestingly, the interspecific variability in these three systems largely conforms to the ~2.5-fold differences allowed by the model of Finlay and Darlington (1995). Such uniformity clearly does not apply to the song system, however, which contains several nuclei that exhibit 5-fold or greater differences between species of similar size. Thus, the song system exhibits a degree of variation indicative of mosaic evolution, especially given that very large volumetric differences have emerged multiple times within the Passerida parvorder, in some cases quite rapidly (e.g., Sylviidae).
At first glance, this extreme variability of the song system would seem to invalidate the hypothesis that conserved developmental programs can be very influential in the evolution of the oscine brain. Yet, the development of the song
91

system is decidedly unusual. Neurons are incorporated into several song nuclei after surrounding tissues have been established, and this delay may effectively release these structures from otherwise powerful constraints and/or create opportunities for their selective alteration independently from other brain structures (Alvarez-Buylla et al. 1994). It is also interesting that the most variable nucleus measured, HVC, is one of the last to develop (e.g., Kirn & DeVoogd 1989; Nordeen et al. 1989). These observations are consistent with the notion that the sequence of brain development makes some changes more likely than others, and they suggest that later-developing nuclei may be more plastic. On the other hand, the late development and marked variability of the song system may be coincidental or related to other factors. The extreme variability of the song system may simply reflect more varied selection pressures on song, be they sexual or natural, than on other behaviors. It could also be a consequence of distinct physiological mechanisms between circuits, akin to a functional incompatibility between memory systems (Sherry & Schacter 1987). HVC neurons encode birdsong by means of a sparse code where individual neurons are specifically dedicated to the production of a single, brief sound (Hahnloser et al. 2002); incidentally, the elaboration of song repertoires is associated with dramatic increases in HVC neuron number (DeVoogd et al. 1993; Moore et al. 2009). The behavioral capacities associated with the other systems measured may not be encoded in such a way and can perhaps be effectively altered by more subtle changes, such as to intrinsic connectivity, synaptic weights, or internal chemistry (Goodson et al. 2006).
These data also highlight tendencies for ‘discrete’ systems to evolve in concert. For example, we have generally treated the visual and auditory systems as being functionally independent, though cross-modal integration certainly occurs and the degree of specialization in one sensory system may depend on the other. Many mammals, for instance, localize sounds for the purpose of visually fixating a source,
92

and sound localization acuity is related to best visual field width (Heffner & Heffner 1992). Similarly, enhanced perceptual systems that permit better visual pattern recognition (Watanabe et al. 2008) or auditory perception (Kubke et al. 2004) could aid in tasks that employ limbic structures, such as episodic memory (Hp; Clayton et al. 2003), territoriality or sociality (Sep; Goodson et al. 1999), and mating behaviors (TnA; Thompson et al. 1998; Svec et al. 2009). Hints at such relationships are reported here: limbic structures are positively correlated with visual nuclei and are nearly significantly related with auditory nuclei. Interestingly, the two negative trends observed both involved the song system, a circuit that underlies production of a reproductive advertisement signal that must somehow be costly to the senders (Zahavi 1975b; Bradbury & Vehrencamp 1998). At the level of individual nuclei, relative RA and nXIIts volumes are inversely correlated with the limbic system. Trade-offs will inevitably occur when comparing the relative sizes of various structures because a single brain cannot have all components that are relatively large or small. Yet, these would seem unlikely to be artifactual given that each occupies a small proportion of the brain. If these relationships genuinely exist, expansions of the song system could provide a means to advertise this neural trade-off. The exact location of arcopallial precursors in the neural tube is not known, but the dorsal ventricular ridge (meso-, nido-, and arcopallium) arises from lateral and ventral pallium, which are adjacent to the medial pallium that generates hippocampal neurons (Garcia-Lopez et al. 2009). Thus, a trade-off between the arcopallium and hippocampus may exist similar to that between neocortex and limbic structures in some orders of mammals (Reep et al. 2007).
In conclusion, this study documents the correlated evolution of functional neural circuits and provides evidence in support of both the concerted and mosaic models of evolution. The song system is dramatically more variable than others
93

measured, which may be related to its late development. Future work on interspecific differences in brain development will provide valuable insights into its role in the evolution of brain structure.
94

REFERENCES
Alvarez-Buylla, A., Ling, C. Y. & Yu, W. S. 1994 Contribution of neurons born during embryonic, juvenile, and adult life to the brain of adult canaries: regional specificity and delayed birth of neurons in the song-control nuclei. J. Comp. Neurol. 347, 233-248.
Atoji, Y. & Wild, J. M. 2004 Fiber connections of the hippocampal formation and septum and subdivisions of the hippocampal formation in the pigeon as revealed by tract tracing and kainic acid lesions. J. Comp. Neurol. 475, 426461.
Barton, R. A. & Harvey, P. H. 2000 Mosaic evolution of brain structure in mammals. Nature 405, 1055-1058.
Blomberg, S. P., Garland, T., Jr. & Ives, A. R. 2003 Testing for phylogenetic signal in comparative data: behavioral traits are more labile. Evolution 57, 717-745.
Boire, D. & Baron, G. 1994 Allometric comparison of brain and main brain subdivisions in birds. J. Brain Res. 35, 49-66.
Bradbury, J. W. & Vehrencamp, S. L. 1998 Principles of Animal Communication. Sunderland, MA: Sinauer Associates.
Burger, R. M., Cramer, K. S., Pfeiffer, J. D. & Rubel, E. W. 2005 Avian superior olivary nucleus provides divergent inhibitory input to parallel auditory pathways. J. Comp. Neurol. 481, 6-18.
Burish, M. J., Kueh, H. Y. & Wang, S. S. H. 2004 Brain architecture and social complexity in modern and ancient birds. Brain Behav. Evol. 63, 107-124.
Carr, C. E. & Konishi, M. 1990 A circuit for detection of interaural time differences in the brain stem of the barn owl. J. Neurosci. 10, 3227-3248.
Charvet, C. J. & Striedter, G. F. 2008 Developmental species differences in brain cell cycle rates between northern bobwhite quail (Colinus virginianus) and parakeets (Melopsittacus undulatus): implications for mosaic brain evolution. Brain Behav. Evol. 72, 295-306.
Charvet, C. J. & Striedter, G. F. 2009a Developmental origins of mosaic brain evolution: morphometric analysis of the developing zebra finch brain. J. Comp. Neurol. 514, 203-213.
95

Charvet, C. J. & Striedter, G. F. 2009b Developmental basis for telencephalon expansion in waterfowl: enlargement prior to neurogenesis. Proc. R. Soc. B 276, 3421-3427.
Cheng, M.-F., Chaiken, M., Zuo, M. & Miller, H. 1999 Nucleus taenia of the amygdala of birds: anatomical and functional studies in ring doves (Streptopelia risoria) and European starlings (Sturnus vulgaris). Brain Behav. Evol. 53, 243-270.
Clancy, B., Darlington, R. B. & Finlay, B. L. 2001 Translating developmental time across mammalian species. Neuroscience 105, 7-17.
Clark, D. A., Mitra, P. P. & Wang, S. S. H. 2001 Scalable architecture in mammalian brains. Nature 411, 189-193.
Clayton, N. S., Bussey, T. J. & Dickinson, A. 2003 Can animals recall the past and plan for the future? Nat. Rev. Neurosci. 4, 685-691.
de Winter, W. & Oxnard, C. E. 2001 Evolutionary radiations and convergences in the structural organization of mammalian brains. Nature 409, 710-714.
DeVoogd, T. J., Krebs, J. R., Healy, S. D. & Purvis, A. 1993 Relations between song repertoire size and the volume of brain nuclei related to song: comparative evolutionary analyses amongst oscine birds. Proc. R. Soc. Lond. B 254, 75-82.
Emery, N. J. & Clayton, N. S. 2004 The mentality of crows: convergent evolution of intelligence in corvids and apes. Science 306, 1903-1907.
Finlay, B. L. & Darlington, R. B. 1995 Linked regularities in the development and evolution of mammalian brains. Science 268, 1578-1584.
Finlay, B. L., Darlington, R. B. & Nicastro, N. 2001 Developmental structure in brain evolution. Behav. Brain Sci. 24, 263-308.
Garcia-Lopez, R., Pombero, A. & Martinez, S. 2009 Fate map of the chick embryo neural tube. Dev. Growth Differ. 51, 145-165.
Garland, T., Jr., Harvey, P. H. & Ives, A. R. 1992 Procedures for the analysis of comparative data using phylogenetically independent contrasts. Syst. Biol. 41, 18-32.
Garland, T., Jr. & Ives, A. R. 2000 Using the past to predict the present: confidence intervals for regression equations in phylogenetic comparative methods. Am. Nat. 155, 346-364.
96

Garland, T., Jr., Midford, P. E. & Ives, A. R. 1999 An introduction to phylogenetically based statistical methods, with a new method for confidence intervals on ancestral values. Am. Zool. 39, 374-388.
Goodson, J. L., Eibach, R., Sakata, J. & Adkins-Regan, E. 1999 Effect of septal lesions on male song and aggression in the colonial zebra finch (Taeniopygia guttata) and the territorial field sparrow (Spizella pusilla). Behav. Brain Res. 98, 167-180.
Goodson, J. L., Evans, A. K. & Lindberg, L. 2004 Chemoarchitectonic subdivisions of the songbird septum and a comparative overview of septum chemical anatomy in jawed vertebrates. J. Comp. Neurol. 473, 293-314.
Goodson, J. L., Evans, A. K. & Wang, Y. 2006 Neuropeptide binding reflects convergent and divergent evolution in species-typical group sizes. Horm. Behav. 50, 223-236.
Grafen, A. 1989 The phylogenetic regression. Phil. Trans. R. Soc. Lond. B 326, 119157.
Hahnloser, R. H. R., Kozhevnikov, A. A. & Fee, M. S. 2002 An ultra-sparse code underlies the generation of neural sequences in a songbird. Nature 419, 65-70.
Heffner, R. S. & Heffner, H. E. 1992 Visual factors in sound localization in mammals. J. Comp. Neurol. 317, 219-232.
Hellmann, B., Manns, M. & Güntürkün, O. 2001 Nucleus isthmi, pars semilunaris as a key component of the tectofugal visual system in pigeons. J. Comp. Neurol. 436, 153-166.
Iwaniuk, A. N., Dean, K. M. & Nelson, J. E. 2004 A mosaic pattern characterizes the evolution of the avian brain. Proc. R. Soc. Lond. B (Suppl.) 271, S148-S151.
Iwaniuk, A. N. & Hurd, P. L. 2005 The evolution of cerebrotypes in birds. Brain Behav. Evol. 65, 215-230.
Kirn, J. R. & DeVoogd, T. J. 1989 Genesis and death of vocal control neurons during sexual differentiation in the zebra finch. J. Neurosci. 9, 3176-3187.
Krebs, J. R., Sherry, D. F., Healy, S. D., Perry, V. H. & Vaccarino, A. L. 1989 Hippocampal specialization of food-storing birds. Proc. Natl. Acad. Sci. USA 86, 1388-1392.
Krützfeldt, N. O. E. & Wild, J. M. 2004 Definition and connections of the entopallium in the zebra finch (Taeniopygia guttata). J. Comp. Neurol., 452-465.
97

Kubke, M. F., Massoglia, D. P. & Carr, C. E. 2004 Bigger brains or bigger nuclei? Regulating the size of auditory structures in birds. Brain Behav Evol 63, 169180.
Lavin, S. R., Karasov, W. H., Ives, A. R., Middleton, K. M. & Garland, T., Jr. 2008 Morphometrics of the avian small intestine compared with that of nonflying mammals: a phylogenetic approach. Physiol. Biochem. Zool. 81, 526-550.
Lefebvre, L., Reader, S. M. & Sol, D. 2004 Brains, innovations and evolution in birds and primates. Brain Behav. Evol. 63, 233-246.
Maddison, W. P. & Maddison, D. R. 2008 Mesquite: a modular system for evolutionary analysis, ver. 2.5. http://mesquiteproject.org/.
Marler, P. & Slabbekoorn, H. (ed.) 2004 Nature's Music: The science of birdsong. San Diego, CA: Elsevier Academic Press.
Moore, J. M., Székely, T., Büki, Y. & DeVoogd, T. J. 2009 Degree of convergence along a motor pathway is strongly related to behavioral complexity across a wide phylogeny of songbirds. Ph.D. Dissertation, Cornell Univesity.
Nordeen, K. W., Marler, P. & Nordeen, E. J. 1989 Addition of song-related neurons in swamp sparrows coincides with memorization, not production, of learned songs. J. Neurobiol. 20, 651-661.
Nottebohm, F., Kelley, D. B. & Paton, J. A. 1982 Connections of vocal control nuclei in the canary telencephalon. J. Comp. Neurol. 207, 344-357.
Nottebohm, F., Stokes, T. M. & Leonard, C. M. 1976 Central control of song in the canary, Serinus canarius. J. Comp. Neurol. 165, 457-486.
Puelles, L. & Rubenstein, J. L. R. 2003 Forebrain gene expression domains and the evolving prosomeric model. Trends Neurosci. 26, 469-476.
Rasband, W. S. 2007 ImageJ, ver. 1.38x. Bethesda, MD: US National Institutes of Health.
Reep, R. L., Finlay, B. L. & Darlington, R. B. 2007 The limbic system in mammalian brain evolution. Brain Behav. Evol. 70, 57-70.
Revell, L. J., Harmon, L. J. & Collar, D. C. 2008 Phylogenetic signal, evolutionary process, and rate. Syst. Biol. 57, 591-601.
Sherry, D. F. & Schacter, D. L. 1987 The evolution of multiple memory systems. Psychol. Rev. 94, 439-454.
98

Stephan, H., Frahm, H. & Baron, G. 1981 New and revised data on volumes of brain structures in insectivores and primates. Folia Primatol. 35, 1-29.
Striedter, G. F. 2005 Principles of Brain Evolution. Sunderland, MA: Sinauer Associates, Inc.
Striedter, G. F. & Charvet, C. J. 2008 Developmental origins of species differences in telencephalon and tectum size: morphometric comparisons between a parakeet (Melopsittacus undulatus) and a quail (Colinus virgianus). J. Comp. Neurol. 507, 1663-1675.
Svec, L. A., Licht, K. M. & Wade, J. 2009 Pair bonding in the female zebra finch: a potential role for the nucleus taeniae. Neuroscience 160, 275-283.
Székely, A. D. & Krebs, J. R. 1996 Efferent connectivity of the hippocampal formation of the zebra finch (Taenopygia guttata): an anterograde pathway tracing study using Phaseolus vulgaris leucoagglutinin. J. Comp. Neurol. 368, 198-214.
Theiss, M. P. H., Hellmann, B. & Güntürkün, O. 2003 The architecture of an inhibitory sidepath within the avian tectofugal system. Neuroreport 14, 879882.
Thompson, R. R., Goodson, J. L., Ruscio, M. G. & Adkins-Regan, E. 1998 Role of the archistriatal nucleus taeniae in the sexual behavior of male Japanese quail (Coturnix japonica): A comparison of function with the medial nucleus of the amygdala in mammals. Brain Behav. Evol. 51, 215-229.
Vates, G. E., Broome, B. M., Mello, C. V. & Nottebohm, F. 1996 Auditory pathways of caudal telencephalon and their relation to the song system of adult male zebra finches (Taeniopygia guttata). J. Comp. Neurol. 366, 613-642.
Watanabe, S., Maier, U. & Bischof, H. J. 2008 Pattern discrimination is affected by entopallial but not by hippocampal lesions in zebra finches. Behav. Brain Res. 190, 201-205.
Welker, W. I., Johnson, J. I., Jr. & Pubols, B. H., Jr. 1964 Some morphological and physiological characteristics of the somatic sensory system in raccoons. Am. Zool. 4, 75-94.
Whiting, B. A. & Barton, R. A. 2003 The evolution of the cortico-cerebellar complex in primates: anatomical connections predict patterns of correlated evolution. J. Human Evol. 44, 3-10.
99

Woolsey, C. N., Carlton, T. G., Kaas, J. H. & Earls, F. J. 1971 Projection of visual field on superior colliculus of ground squirrel (Citellus tridecemlineatus). Vision Res. 11, 115-127.
Yopak, K. E., Lisney, T. J., Collin, S. P. & Montgomery, J. C. 2007 Variation in brain organization and cerebellar foliation in chondrichthyans: sharks and holocephalans. Brain Behav. Evol. 69, 280-300.
Zahavi, A. 1975 Mate selection: a selection for a handicap. J. Theor. Biol. 53, 205214.
100

CHAPTER 4
THE CAUDOMEDIAL NIDOPALLIUM (NCM) IS NECESSARY FOR MATE SONG PREFERENCES IN FEMALE ZEBRA FINCHES (TAENIOPYGIA GUTTATA)
Abstract Birdsong is a sexually selected communication signal that functions in reproduction. Females actively attend to male song and can exhibit at least two general types of song preferences, one for familiar songs and another for high quality songs. The neural bases of these preferences are poorly understood. The present study sought to determine whether the caudomedial nidopallium (NCM), a region analogous to mammalian auditory association cortex, is necessary for the expression of these two types of preferences in adult female zebra finches (Taeniopygia guttata). NCM was bilaterally and reversibly inactivated with tetrodotoxin and subjects’ preferences for their mate’s song over a novel song and for complex ‘tutored’ song over abnormally simple ‘isolate’ song were tested using a phonotaxis task. Inactivation of NCM eliminated the mate song preference of most subjects but had a much more modest effect on tutored song preferences. Inactivation of the caudal mesopallium (CM) had no effect on mate song preferences. These results demonstrate for the first time that NCM is required for the expression of song preferences by adult female songbirds. Potential pathways underlying this behavior and functional compartmentalization of the auditory forebrain are discussed. Introduction
Birdsong is a learned vocal communication signal that functions in reproduction. While a great deal is known about the neural mechanisms underlying male song learning and production, much less is understood about the neural bases of
101

song perception. In particular, the specific functions of auditory forebrain areas have not been unambiguously identified. Female songbirds provide opportunities to investigate these issues because they are often the intended audience of male song and they actively attend to it. They can make acute discriminations between songs (Miller 1979a; Searcy & Brenowitz 1988; Riebel & Smallegange 2003) and sophisticated judgments about male quality based on song (Kroodsma 1976; Hasselquist et al. 1996; Nowicki et al. 2002). The present study seeks to identify areas underlying these complex abilities.
Female zebra finches (Taeniopygia guttata) exhibit at least two general types of song preferences. One is based on song familiarity; they prefer both their father’s and mate’s songs over those of unfamiliar males and these persist over long time periods, indicating that they form stable memories of specific songs (Miller 1979a,b; Riebel 2000). The second is based on song quality, which may be a result of song content or performance-based aspects. In the case of the former, females prefer longer songs comprised of more syllables (Neubauer 1999a) and songs that were faithfully copied from an adult over those improvised by males raised in isolation (Lauay et al. 2004). For the latter, they prefer higher song rates (Houtman 1992) as well as femaledirected songs, which are slightly faster and more stereotyped than undirected songs (Woolley & Doupe 2008). The neural bases of these two types of preferences are unknown, but they seem likely to involve different brain regions given that the first is based on a specific auditory memory and is not necessarily dependent on one song’s perceived quality relative to another, while the second can be shown with songs the bird has never heard.
The caudomedial nidopallium (NCM) appears to be especially important for the encoding of song-related memories. It receives highly processed auditory information from Field L, the avian analogue of mammalian auditory cortex, is
102

densely interconnected with other regions in the auditory forebrain, and is specifically activated by playback of conspecific song (Figure 4.1; Mello et al. 1992; Vates et al. 1996). Both physiological and immediate early gene (IEG) responses habituate to repeated presentations of a stimulus (Chew et al. 1995; Mello et al. 1995); this effect is specific to individual songs and is long-lasting, suggesting that that NCM has the capacity for storing many distinct auditory memories (Chew et al. 1996; Stripling et al. 1997). Memory of the tutor song may be stored here, for instance, as both IEG levels and physiological habituation rates following tutor song playback are related to the strength of song learning in males (Bolhuis et al. 2000; Phan et al. 2006). Similarly, ZENK protein levels are habituated in females following playback of the mate’s song compared to a novel song (Woolley & Doupe 2008) and are elevated following the acquisition of new song discriminations (Gentner et al. 2004).
Whether NCM also serves a role in song quality assessment is less clear. ZENK levels in female zebra finches are equivalent following playback of directed or undirected songs (Woolley & Doupe 2008). Yet, they are elevated in female European starlings (Sturnus vulgaris) following playback of longer songs, which they find more attractive (Gentner & Hulse 2000; Gentner et al. 2001), and in female white-crowned sparrows (Zonotrichia leucophrys) following playback of a more attractive local versus foreign dialect (Maney et al. 2003). Each of these latter findings may reflect song salience, however, because they both depend on previous experience (Sockman et al. 2002). The goal of this study was to ascertain whether NCM is necessary for these two types of song preferences in female adult zebra finches.
103

Figure 4.1. Diagram of the ascending auditory pathway and major auditory forebrain areas. Cannulae targeted NCM and a central region of CM that was slightly lateral to CMM.
104

Methods (a) Subjects
Subjects were 16 adult female zebra finches (≥240 days old) bred from an inhouse colony. Prior to mate song preference testing, each female was housed with an adult male in an individual cage (45×40×40 cm) outfit with a nest box and nesting material for at least 3 months. All subjects had successfully reared offspring to adulthood before testing. Before assessing preferences for tutored versus isolate songs, subjects were housed individually without a nest box. Four birds with cannulae directed at NCM were tested in both the song familiarity and song quality preference tests. Cages occupied two rooms; opaque sheets visually isolated all cages within a room and birds housed in one room never heard the songs of males in the other except during behavioral tests (see below). Birds were kept on a 14h:10h light:dark schedule at all times and food and water were available ad libitum. (b) Cannulation Surgery
All surgical procedures were approved by the Cornell University Institutional Animal Care and Use Committee. Birds were anesthetized with an intramuscular injection of Equithesin (0.006 ml · g-1) and placed in a stereotaxic apparatus outfitted with a custom beak bar. Stainless steel guide cannulae (27 ga., Small Parts, Inc.) cut to 4 mm were bilaterally placed over NCM (0.6 mm rostral and 0.5 mm lateral of sagittal sinus bifurcation, 0.3 mm ventral) or CM (2.0 mm rostral, 1.0 mm lateral, 0.1 mm ventral) and affixed to the skull using dental cement. Following surgery, small wire plugs were placed in the cannulae to prevent clogging and birds were administered the analgesic butorphanol (2.5×10-4 mg · g-1). Birds were placed in a small recovery cage with a heating pad until they were able to perch normally and then returned to their home cage. They were allowed to recover for 1 week before behavioral testing.
105

(c) Song recordings Male songs were recorded (44.1 kHz, 16-bit) in a soundproof chamber with a
Marantz PMD670 digital audio recorder and Sennheiser ME62 omnidirectional microphone with parabolic reflector. Playback clips of female-directed songs were created using Syrinx (Burt 2006). Each clip contained a single song bout with 3-5 introductory notes and 4 motifs per bout. Bouts were normalized for peak amplitude but were otherwise unmanipulated. In cases where the song bout selected to create the playback clip contained more than 4 consecutive motifs, the first 4 were used. Stimuli were fed through a Pioneer SA-410 amplifier, and maximum song amplitude was 70 dB SPL on the near perch of the testing cage. (d) Drug Infusions & Behavioral Testing
The behavioral testing cage consisted of three connected cages (135×40×40 cm) to form two approach zones and a central, neutral zone that contained a nest box with nesting material and food and water (Figure 4.2a). Passageways were visually emphasized with a white border to promote free movement throughout the entire cage. Speakers (Morel MDT29) with flat frequency responses (0.5 – 10 kHz) were placed at opposite ends of the cage and male zebra finch models affixed to perches on the front of each speaker stand. A white semi-sheer fabric was draped over each end of the cage to obscure fine differences in physical appearance between the models and the subjects’ mates. Four symmetrically-positioned perches traversed the width of the cage: two within the neutral cage and one in each approach zone located 12 cm from the cage edge and 16 cm from the cage floor. Infrared emitters and sensors (DigiKey) were aimed over the end perches and read with a National Instruments USB6008 data acquisition card and custom Matlab software to enable the automated recording of bird position at these two locations.
106

Figure 4.2. (a) Diagram of the phonotaxis testing cage with a central, neutral zone and two approach zones. Cage ends were draped with a white semi-sheer fabric to obscure details of the male models. w: water, f: food. (b) Timeline for the infusion, re-acclimation period, hearing test, and song preference test.
107

A timeline of the experimental procedure is shown in Figure 4.2b. Prior to testing, subjects were isolated in the testing cage for at least 3 hours. To administer drug infusions, they were restrained in a custom cloth jacket and placed in a body mold with a head strap. Subjects were infused bilaterally with 0.15 µl of 0.9% saline for control trials and with 0.15 µl of 10 µM tetrodotoxin (TTX; Sigma), a potent Na+ channel blocker, dissolved in saline for experimental trials. Infusions were administered with a 0.5 µl Hamilton syringe and 32-gauge needle. Polyethylene tubing covered the entire needle except 5 mm from the tip to control the needle’s depth (i.e. 1 mm past the base of the guide cannula). Infusion rate was ≤0.03 µl/min to prevent permanent tissue damage and the needle was left in place for at least 3 additional minutes before removal to allow drug diffusion. In total, infusions lasted approximately 20 minutes.
Following infusions, subjects were allowed to re-acclimate to the testing cage for 30 minutes. A second female was placed in the cage for companionship during the first 15 minutes, which helped birds recover from any distress caused by handling, and the subject was again isolated for the last 15 minutes. Next, a hearing test was conducted because both NCM and CM are immediately adjacent to Field L, the avian analogue of primary auditory cortex, and we wished to rule out the possibility that drug-induced changes in behavior resulted from deafening the bird. Recorded aviary noises were played through both loudspeakers (53 dB SPL in the central cage) and the number of distance (or long) calls given by the subject was recorded. Distance calls are conspicuously long (~0.4 sec) and loud vocalizations frequently given by birds isolated from a flock (Zann 1984). This test began with 5 minutes of silence to obtain the bird’s baseline calling rate, and then a 30 sec clip of aviary sounds was followed by 30 sec of silence; the number of calls given during this one-minute interval was
108

recorded. Some subjects (n = 5) were also presented with a 15 sec clip of white noise and/or a 3 kHz tone to assess whether calling was general to all auditory stimuli.
Female song preferences were measured in two ways: trials began with a twochoice phonotaxis paradigm (passive phase, ~35 min) and were immediately followed by a two-choice operant paradigm (active phase, 30 min) in which song playback was the reward. The passive phase began with the left speaker broadcasting 4 song bouts with 9 sec separating the onset of each bout (4 motifs per bout, each bout lasted ~5-7 sec) and a 15-sec silent interval followed playback of the fourth bout. This routine was then initiated from the right speaker. The total cycle (left and right playbacks) was repeated 20 times. Speaker approach times were recorded as the amount of time the subject spent on the left and right end perches. During the active phase, the subject controlled song playback with its position. Perching on either end elicited playback of one bout from that speaker, and a 5-sec silent interval separated each bout if the bird remained on a single perch for extended periods.
Trials were only included in the analyses if the subjects met one (passive) or two (active) criteria. For the passive phase, subjects were required to spend at least 10% of the total time on either end perch. For the active phase, subjects were required to satisfy the passive phase criterion and also spend 10% of this trial on either end perch. This occurred in all but two cases (the same bird), therefore no active phase data are presented for that bird. Birds were tested until they met these criteria for a total of 6 preference trials for song familiarity (mate vs. stranger) or song quality (tutored vs. isolate). In each case, two control (saline) trials preceded two experimental (TTX) trials which were followed by two additional control trials. For each treatment, the speaker associated with a particular song was switched between the two trials to control for potential side preferences. Trials were separated by at least three days, during which time the subject was returned to its home cage.
109

The song familiarity preference test paired the song of the subject’s mate with that of a stranger matched for syllable repertoire, and each trial used a novel stranger song. Songs of males housed in the same room as the subject were never used as a stranger song; however, to control for the possibility that mate songs were inherently more attractive than stranger songs, those of males housed in the other room (and mated with other female subjects) were used as stranger songs whenever possible. For the song quality preference tests, one tutored song (i.e. of a male that faithfully copied the song of a normal adult) was paired with one ‘isolate’ song (i.e. of a male that improvised its song because it was raised in isolation from adults). Isolate songs usually contain fewer syllables, are less spectrotemporally complex, and occasionally contain abnormal features such as slow, upward frequency modulations (Figure 4.3). The same song pair was used in all six trials for a given subject but different subjects were presented with different song pairs. Here, songs were chosen to maximize the disparity in quality, so that tutored songs typically had more syllables and were slightly longer in duration and isolate songs were especially simple. Even still, not all females consistently preferred tutored songs in initial saline trials; only those that did were used in subsequent tests.
Female discriminatory abilities were also measured from recordings made during the passive phase of song preference trials. Both call latency relative to bout onset, which is related to song familiarity (Stripling et al. 2003), and call number during each bout were measured. Measurements did not differentiate between contact (short) and distance (long) calls. The same measurements were made from initial saline trials of song quality preference tests, but no differences in call number or latency to tutored or isolate songs existed, therefore recordings from subsequent trials were not analyzed.
110

111

Figure 4.3. Spectrograms of a single motif from a representative (a) normal, tutored song and (b) isolate song. Each playback clip was comprised of a bout containing 3-5 introductory notes (i) and 4 motifs. In the isolate song, note the lack of syllables with complex spectrotemporal modulations and the abnormal, slow upward frequency modulation of the last syllable.

The effectiveness of the TTX was periodically assessed using male zebra finches. Two males had cannulae bilaterally implanted targeting HVC, a song premotor nucleus located in the dorsal nidopallium, and were housed individually thereafter. Approximately every two to three weeks, each was infused with the same volume and concentration of TTX as administered to the females, allowed to recover in their home cage for ~30 minutes, and then presented with a female. The males reliably approached the female and attempted to sing (as judged by their posture and production of introductory notes), but they failed to do so following TTX infusions. New drug dilutions were made from a stock solution whenever this effect did not last >2h. (e) Histology
Following behavioral testing, cannula placement was verified by infusing 0.15 µl of 1 mM BODIPY-muscimol (Molecular Probes, MW=607.46) dissolved in saline and DMSO. Birds were infused as described above, then administered an overdose of barbiturate anesthetic and transcardially perfused with 0.9% saline followed by 10% formalin in saline. Brains were post-fixed for 24h, cryoprotected in 30% sucrose/10% formalin for 24h, embedded in gelatin and cryoprotected for at least 5 days. Brains were sectioned at 40 µm in the parasagittal plane on a freezing, sliding microtome; alternate sections were mounted on gel-coated slides and coverslipped immediately with ProLong Gold with DAPI (Molecular Probes) or allowed to dry and Nissl-stained with cresyl violet. Needle track locations were measured in Nissl-stained sections with aid of a camera lucida; approximate drug spread was estimated from digital images of the BODIPY-muscimol. The spread of BODIPY-muscimol likely underrepresented the spread of TTX due to its slightly larger size and greater hydrophobicity, but it nevertheless provided visual confirmation of successful infusions.
112

(f) Data Analysis Following previous studies, we reserved the term ‘preference’ for instances
when a subject spent at least twice as much time on one side compared to the other (e.g., Miller 1979a; Woolley & Doupe 2008). The passive and active phases of the phonotaxis task were analyzed separately. Data were represented as the time spent on the mate or tutored song side as a proportion of time spent on either end perch [e.g., mate/(mate+stranger)]; this was viewed as the most representative index of relative preference strength because it factored out time spent in the middle cage or on the cage floor. Each trial was given equal weight in final analyses by first computing the preference scores or call data and then averaging the two from a given treatment. These averages were analyzed with a two-way repeated measures ANOVA with ‘treatment’ and ‘mate speaker’ (i.e. left or right) as factors. The Holm-Sidák Pairwise test was used for post hoc comparisons. Results (a) General activity levels
All birds called during all hearing tests following both saline and TTX infusions (Figure 4.4). Within each treatment, the number of distance calls during and immediately after aviary noise playbacks was greater than that during the preceding five minutes of silence (all p < 0.01). There were no differences between saline and TTX treatments (p = 0.48). None of the five birds presented with white noise and/or 3-kHz tones called during these stimuli following any treatment, indicating that they retained the ability to discriminate complex, species-typical sounds from simpler stimuli.
There were no significant differences in the proportion of total trial time spent on either end perch across the three treatments (i.e. Saline 1, TTX, Saline 2). For mate vs. novel song tests following NCM infusions, birds spent 42±16% (mean ± sd; p =
113

Figure 4.4. Results from hearing tests conducted before both types of song preference tests. All birds called during each hearing test, and the average number of calls to aviary noise playback was significantly greater than those produced during the preceding 5 min of silence for each treatment (all p < 0.01). The average number of calls did not differ across treatments (p = 0.48).
114

0.22) of the passive phase and 37±21% (p = 0.86) of the active phase on the end perches. For the tutored vs. isolate song tests, they spent 51±22% (p = 0.91) of the passive phase and 46±27% (p = 0.06) of the active phase on the end perches. Finally, in the mate-novel tests following CM infusions the birds spent 45±22% (p = 0.50) of the passive phase and 28±19% (p = 0.17) of the active phase on the end perches. (b) NCM inactivation
All birds exhibited strong preferences for their mate’s song over novel songs in both the passive and active phases following saline infusions into NCM (Figure 4.5). During the passive phase, nearly all of the time that subjects spent on an end perch in Saline 1 trials was next to the mate song speaker (0.98±0.04, mean ± sd). This preference was reduced following TTX infusions (0.57±0.25, p < 0.001) and was abolished in 5 out of 8 birds, one of which reversed its preference and strongly preferred the novel song in both trials. These effects were reversible, as mate song preferences were universally restored during Saline 2 trials (0.88±0.09; p = 0.001). The reduced mate song preferences in TTX trials remained significant after exclusion of the bird that preferred the novel song (both p < 0.001). The effects of NCM inactivation were even stronger in the active phase. Again, all birds strongly preferred their mate’s song in both Saline 1 (0.99±0.01) and Saline 2 (0.91±0.07) trials, but this preference was eliminated in all birds following administration of TTX (0.31±0.23; both p < 0.001). One bird did not move to an end perch during the active phase of either experimental trial, thus n = 7 for this task. The birds did not exhibit a side preference across trials and treatments in either phase (both p > 0.55).
Mirroring the phonotaxis data, females called more quickly to playback of their mate’s song than to novel songs during both saline trials (Wilcoxon signed rank test, both p < 0.03) but not during the TTX trials (p = 0.11; Figure 4.6a). However, the differences in ratios between treatments were not significant (p = 0.10) because the
115

Figure 4.5. Phonotaxis results from (a) passive and (b) active phases showing relative preferences for the mate’s song versus a repertoire-matched novel song [mate/(mate+novel)] following infusions into NCM. (a) Mate song preferences were significantly reduced during experimental trials (0.57±0.25, p < 0.001) and eliminated in 5 out of 8 birds but were universally restored during Saline 2 trials (p = 0.001). (b) Mate song preferences were abolished by the experimental treatment in all subjects during the active phase (both p < 0.001).
116

Figure 4.6. Call latency data collected during the passive phase of the mate-novel phonotaxis test following infusions into NCM. (a) Subjects called more quickly in response to their mate’s song playback than to the novel song in both control trials (both p < 0.03) but not in the experimental trials (p = 0.11). (b) There was no significant difference between treatments when call latencies were expressed as a ratio (mate/novel; p = 0.10).
117

latencies for 3 of the 7 birds were unaffected during experimental trials (Figure 4.6b). There was not a tight correspondence between birds that retained the mate song preference in the passive phase and shorter call latency. Of the five birds that lost their mate song preferences in the phonotaxis task following TTX infusions, three had substantially increased call latency ratios. Conversely, of the three that maintained their mate song preference in the phonotaxis task, one also maintained shorter call latency, one had an increased ratio, and the third was not recorded. No consistent differences in the number of calls between stimuli or across treatments were observed.
Inactivation of NCM also affected female preferences for high quality songs, (Figure 4.7). In the passive phase, this preference (0.89±0.09) was significantly reduced following TTX infusions (0.65±0.27, p = 0.004) and restored in Saline 2 trials (0.94±0.07, p = 0.01). As before, this reduction remained significant even after removal of the most extreme case (both p < 0.05). This effect was not present in the active phase, however. All birds displayed a preference for tutored song in the Saline 1 (0.92±0.05) and Saline 2 (0.80±0.16) trials. This preference was robust to TTX infusions (0.69±0.28, p = 0.12), though preferences were abolished in 2 out of 8 birds. As before, there was no side preference effect in either phase (both p > 0.49). (c) CM inactivation
In contrast with NCM, inactivation of CM had no effect on mate song preferences (Figure 4.8). In the passive phase, the strong preferences in Saline 1 trials (0.92±0.08) were maintained in TTX (0.86±0.07) and Saline 2 trials (0.89±0.15) (p = 0.38). The same was true in the active phase, where mate song preferences in the three trials were 0.89±0.09, 0.78±0.09, and 0.86±0.14, respectively (p = 0.47). Notably, TTX infusions into CM did not abolish this preference in any of the subjects during the passive phase and barely did so in one subject during the active phase (preference = 0.65). No side effect preferences were evident (both p > 0.32).
118

Figure 4.7. Phonotaxis results from (a) passive and (b) active phases showing relative preferences for a tutored song versus an isolate song [tutored/(tutored+isolate)] following infusions into NCM. (a) Tutored song preferences were significantly reduced (p < 0.05) following TTX infusions and were eliminated in 4 of 8 subjects, and these preferences were universally restored during the second control trial (p < 0.01). (b) Tutored song preferences in the active phase were not significantly affected by experimental treatments (p = 0.12), though they were eliminated in 2 of 8 subjects.
119

Figure 4.8. Phonotaxis results from (a) passive and (b) active phases showing relative preferences for the mate’s song versus a repertoire-matched novel song [mate/(mate+novel)] following infusions into CM. Experimental treatments did not affect this preference in either phase (both p ≥ 0.38).
120

(d) Cannulae site verification Needle tracks within left and right NCM were located 0.42±0.07 and
0.39±0.09 mm from the midline, respectively. All were at least 0.5 mm caudal of the mesopallial lamina (LaM) and stopped approximately half way along the dorsalventral axis. Needle tracks within left and right CM were located 0.86±0.11 and 0.91±0.13 mm from the midline, respectively, and were at least 0.3 mm rostral of LaM. Infusions of BODIPY-muscimol were generally contained within NCM or CM and the average drug spread around the needle track was 0.21±0.04 mm3, though some fluorescence was always detectable along the ventricular borders as well (Figure 4.9). Discussion
Recent research has delineated pathways within the auditory forebrain of songbirds that are believed to underlie song perception (Vates et al. 1996; Mello et al. 1998). Special attention has been paid to NCM because it is specifically activated by presentation of conspecific song and several lines of evidence suggest that these neurons are involved in song-related memory formation (Mello et al. 1992; Pinaud & Terleph 2008). First, both genomic and physiological responses habituate to repeated presentations of a song stimulus, and these altered responses are specific to individual songs and are long-lasting (Chew et al. 1995, 1996; Mello et al. 1995; Stripling et al. 1997). Second, NCM appears to encode contextual information about specific songs and not simply their acoustic features. Activation levels of the IEG zenk are correlated with the strength of learning in associative tasks (Jarvis et al. 1995), and previously learned songs can induce a full zenk response when merely associated with a new context (Kruse et al. 2004). Third, both IEG levels (Bolhuis et al. 2000; Terpstra et al. 2004) and physiological habituation rates (Phan et al. 2006) following tutor song playback are correlated with the strength of song learning. Finally, inhibition of zenk induction in juveniles disrupts song acquisition (London & Clayton 2008), and lesions
121

122

Figure 4.9. Images taken following infusions of 0.15 μl of 1 mM BODIPY-muscimol. Note that the brains were sectioned at an angle to keep NCM intact with the rest of the brain, therefore the relative sizes of other regions are not indicative of needle location along the medial-lateral axis. (a) Needle penetrations into NCM were restricted to the dorsal half and were always ≥0.5 mm caudal of the mesopallial lamina (LaM). The section shown is 360 μm lateral from the most medial section containing NCM tissue. (b) Needle penetrations into CM were generally centrally located and were always ≥0.3 mm rostral of LaM. The section shown is 760 μm lateral from the most medial section containing CM tissue.

of NCM in adults diminish their tutor song preferences (Gobes & Bolhuis 2007). Together, these studies indicate that NCM is important for several aspects of song perception, especially the formation and storage of song-related memories.
Female songbirds provide a useful model system to study the neural mechanisms of song perception because they actively attend to song and exhibit several types of distinct song preferences. This study shows that NCM is necessary for the expression of mate song preferences in adult female zebra finches. This effect does not appear to result from generally lower activity levels or from general perceptual deficits because subjects retained the ability to discriminate species-specific aviary noises from white noise and tones and their preferences for tutored over isolate songs were only modestly affected. These data cannot distinguish between two potential underlying causes, however. One is that the treatment disrupted the neural substrate that encodes song-related memories, in which case the subjects could not retrieve learned songs but their general perception and motivation were unaffected. On the other hand, NCM inactivation may have decreased subjects’ fidelity such that they recognized their mate’s song but were not compelled to approach that speaker. We attempted to distinguish between these possibilities by analyzing female calls, but the results were ambiguous. Approximately half (3 of 5) of the birds that lost their mate song preference also had comparable call latencies to the two songs, consistent with (but not necessarily indicative of) the idea that these birds did not recognize their mate’s song. Yet, two birds that did not prefer their mate’s song retained much shorter call latencies, suggesting that they could still discriminate between the stimuli. At least to some extent, then, it appears that decreased pair-bond fidelity contributed to the phonotaxis results. Other potential explanations appear less likely. The reduced mate song preferences do not appear to result from general deficits in song perception or from a general behavioral effect related to the task itself because tutored song
123

preferences were less affected or even retained. Finally, because inactivation of CM did not affect mate song preferences, this deficit is not attributable to a general brain effect.
In conjunction with previous reports, these results suggest some level of functional compartmentalization within the songbird auditory forebrain. Specifically, NCM appears to underlie the formation and/or storage of song-related memories but not the assessment of song quality. ZENK levels in NCM of female canaries are not related to the proportion of complex syllables within a song (Leitner et al. 2005), nor do levels in female zebra finches differ following playback of directed or undirected song (Woolley & Doupe 2008). They are related to song bout length in European starlings, a feature preferred by females, but only if subjects are first primed with such songs. This may therefore reflect a difference in stimulus salience rather than quality per se (Gentner et al. 2001; Sockman et al. 2002). In contrast, neighboring CMM appears to be involved in both functions. Its physiological and IEG activity are elevated in response to familiar songs (Gentner & Margoliash 2003; Gentner et al. 2004; Terpstra et al. 2006), it exhibits greater ZENK activation following higher quality songs (Sockman et al. 2002; Leitner et al. 2005; Woolley & Doupe 2008), and lesions of CMM abolish female preferences for conspecific over heterospecific song (MacDougall-Shackleton et al. 1998). In the present study, inactivation of CM had no effect on mate song preferences, perhaps because the cannulae were directed lateral to CMM in an effort to limit drug spread into Field L or NCM. Infusions centered in more medial regions could perhaps also affect mate song preferences because this subregion is densely interconnected with NCM and its activity reflects song familiarity (Hernandez & MacDougall-Shackleton 2004; Terpstra et al. 2006). Thus, the present results cannot rule out a role for CMM in the expression of preferences for familiar songs.
124

The neural mechanisms by which NCM may help to encode song-related memories remain unclear. NCM is topographically organized and neuronal tuning bandwidths reflect those of species-typical sounds (Terleph et al. 2006, 2007). Some studies have reported generally lower physiological activity in response to familiar songs (e.g., Phan et al. 2006), but others have not (George et al. 2008). In the latter study, however, NCM neurons were most responsive to the motif class that is most variable between individuals, thus still suggesting an important role in individual recognition. Nevertheless, it remains difficult to speculate about NCM’s precise role in the formation of song-related memories because detailed descriptions its intrinsic connectivity and projections are lacking. It currently seems most likely that the projection to CMM, which ultimately projects to multiple auditory and motor areas including the song system (Vates et al. 1996; Bauer et al. 2008), is especially pertinent.
Neurons within the male song system respond selectively to the bird’s own song, constituting the most specific auditory response described (McCasland & Konishi 1981). This selectivity emerges gradually through inputs from CM to the interface nucleus of the nidopallium (NIf) to HVC (Janata & Margoliash 1999; Bauer et al. 2008), the last of which gives rise to song-selectivity within other song nuclei (Doupe & Konishi 1991; Vates et al. 1997). With the exception of NIf, this selectivity for single songs has not been observed in the auditory forebrain (Amin et al. 2004), and it also remains unknown if such specificity exists in the song nuclei of nonsinging females. Lesions of HVC in female zebra finches do not affect their ability to discriminate conspecific from heterospecific song (MacDougall-Shackleton et al. 1998). In female starlings (which sing), HVC lesions do not diminish their ability to discriminate between familiar songs but they do adversely affect their ability to make novel associations with those songs (Gentner et al. 2000). In female canaries (which
125

also sing), HVC activity is greater in response to songs with complex syllables (Del Negro et al. 2000) and HVC lesions also affect their discriminatory abilities between songs (Brenowitz 1991; Del Negro et al. 1998; Halle et al. 2002). Thus, the song system can also play an important role in female song perception. It receives highly processed information from the auditory forebrain, and together with it may underlie several forms of song preferences.
This study demonstrates for the first time that NCM is necessary for mate song preferences in adult female zebra finches. NCM likely forms part of the neural substrate that encodes specific song memories and may also underlie aspects of pair bond strength. Future work on its intrinsic circuitry and projections to other systems will help to discover they mechanisms by which song-related memories are formed, stored, and retrieved.
126

REFERENCES
Amin, N., Grace, J. A. & Theunissen, F. E. 2004 Neural response to bird's own song and tutor song. J. Comp. Physiol. A 190, 469-489.
Bauer, E. E., Coleman, M. J., Roberts, T. F., Roy, A., Prather, J. F. & Mooney, R. 2008 A synaptic basis for auditory-vocal integration in the songbird. J. Neurosci. 28, 1509-1522.
Bolhuis, J. J., Zijlstra, G. G. O., den Boer-Visser, A. M. & Van der Zee, E. A. 2000 Localized neuronal activation in the zebra finch brain is related to the strength of song learning. Proc. Natl. Acad. Sci. USA 97, 2282-2285.
Brenowitz, E. A. 1991 Altered perception of species-specific song by female birds after lesions of a forebrain nucleus. Science 251, 303-305.
Burt, J. M. 2006 Syrinx, ver. 2.6h. http://www.syrinxpc.com/.
Chew, S. J., Mello, C., Nottebohm, F., Jarvis, E. & Vicario, D. S. 1995 Decrements in auditory responses to a repeated conspecific song are long-lasting and require two periods of protein synthesis in the songbird forebrain. Proc. Natl. Acad. Sci. USA 92, 3406-3410.
Chew, S. J., Vicario, D. S. & Nottebohm, F. 1996 A large-capacity memory system that recognizes the calls and songs of individual birds. Proc. Natl. Acad. Sci. USA 93, 1950-1955.
Del Negro, C., Gahr, M., Leboucher, G. & Kreutzer, M. 1998 The selectivity of sexual responses to song displays: effects of partial chemical lesion of the HVC in female canaries. Behav. Brain Res. 96, 151-159.
Del Negro, C., Kreutzer, M. & Gahr, M. 2000 Sexually stimulating signals of canary (Serinus canaria) songs: evidence for a female-specific auditory representation in the HVc nucleus during the breeding season. Behav. Neurosci. 114, 526542.
Doupe, A. J. & Konishi, M. 1991 Song-selective auditory circuits in the vocal control system of the zebra finch. Proc. Natl. Acad. Sci. USA 88, 11339-11343.
Gentner, T. Q. & Hulse, S. H. 2000 Female European starling preference and choice for variation in conspecific male song. Anim. Behav. 59, 443-458.
Gentner, T. Q., Hulse, S. H. & Ball, G. F. 2004 Functional differences in forebrain auditory regions during learned vocal recognition in songbirds. J. Comp. Physiol. A 190, 1001-1010.
127

Gentner, T. Q., Hulse, S. H., Bentley, G. E. & Ball, G. F. 2000 Individual vocal recognition and the effect of partial lesions to HVc on discrimination, learning, and categorization of conspecific song in adult songbirds. J. Neurobiol. 42, 117-133.
Gentner, T. Q., Hulse, S. H., Duffy, D. & Ball, G. F. 2001 Response biases in auditory forebrain regions of female songbirds following exposure to sexually relevant variation in male song. J. Neurobiol. 46, 48-58.
Gentner, T. Q. & Margoliash, D. 2003 Neuronal populations and single cells representing learned auditory objects. Nature 424, 669-674.
George, I., Cousillas, H., Richard, J.-P. & Hausberger, M. 2008 A potential neural substrate for processing functional classes of complex acoustic signals. PLoS One 3, e2203.
Gobes, S. M. H. & Bolhuis, J. J. 2007 Birdsong memory: a neural dissociation between song recognition and production. Curr. Biol. 17, 789-793.
Halle, F., Gahr, M., Pieneman, A. W. & Kreutzer, M. 2002 Recovery of song preferences after excitotic HVC lesion in female canaries. J. Neurobiol. 52, 113.
Hasselquist, D., Bensch, S. & von Schantz, T. 1996 Correlation between male song repertoire, extra-pair paternity and offspring survival in the great reed warbler. Nature 381, 229-232.
Hernandez, A. M. & MacDougall-Shackleton, S. A. 2004 Effects of early song experience on song preferences and song control and auditory brain regions in female house finches (Carpodacus mexicanus). J. Neurobiol. 59, 247-258.
Houtman, A. M. 1992 Female zebra finches choose extra-pair copulations with genetically attractive males. Proc. R. Soc. Lond. B 249, 3-6.
Janata, P. & Margoliash, D. 1999 Gradual emergence of song selectivity in sensorimotor structures of the male zebra finch song system. J. Neurosci. 19, 5108-5118.
Jarvis, E. D., Mello, C. V. & Nottebohm, F. 1995 Associative learning and stimulus novelty influence the song-induced expression of an immediate-early gene in the canary forebrain. Learn. Mem. 2, 62-80.
Kroodsma, D. E. 1976 Reproductive development in a female songbird: differential stimulation by quality of male song. Science 192, 574-575.
128

Kruse, A. A., Stripling, R. & Clayton, D. F. 2004 Context-specific habituation of the zenk gene response to song in adult zebra finches. Neurobiol. Learn. Mem. 82, 99-108.
Lauay, C., Gerlach, N. M., Adkins-Regan, E. & DeVoogd, T. J. 2004 Female zebra finches require early song exposure to prefer high-quality song as adults. Anim. Behav. 68, 1249-1255.
Leitner, S., Voigt, C., Metzdorf, R. & Catchpole, C. K. 2005 Immediate early gene (ZENK, Arc) expression in the auditory forebrain of female canaries varies in response to male song quality. J. Neurobiol. 64, 275-284.
London, S. E. & Clayton, D. F. 2008 Functional identification of sensory mechanisms required for developmental song learning. Nat. Neurosci. 11, 579-586.
MacDougall-Shackleton, S. A., Hulse, S. H. & Ball, G. F. 1998 Neural bases of song preferences in female zebra finches (Taeniopygia guttata). Neuroreport 9, 3047-3052.
Maney, D. L., MacDougall-Shackleton, E. A., MacDougall-Shackleton, S. A., Ball, G. F. & Hahn, T. P. 2003 Immediate early gene response to hearing song correlates with receptive behavior and depends on dialect in a female songbird. J. Comp. Physiol. A 189, 667-674.
McCasland, J. S. & Konishi, M. 1981 Interaction between auditory and motor activities in an avian song control nucleus. Proc. Natl. Acad. Sci. USA 78, 7815-7819.
Mello, C., Nottebohm, F. & Clayton, D. 1995 Repeated exposure to one song leads to a rapid and persistent decline in an immediate early gene's response to that song in zebra finch telencephalon. J. Neurosci. 15, 6919-6925.
Mello, C. V., Vates, G. E., Okuhata, S. & Nottebohm, F. 1998 Descending auditory pathways in the adult male zebra finch (Taeniopygia guttata). J. Comp. Neurol. 395, 137-160.
Mello, C. V., Vicario, D. S. & Clayton, D. F. 1992 Song presentation induces gene expression in the songbird forebrain. Proc. Natl. Acad. Sci. USA 89, 68186822.
Miller, D. B. 1979a Long-term recognition of father's song by female zebra finches. Nature 280, 389-391.
Miller, D. B. 1979b The acoustic basis of mate recognition by female zebra finches (Taeniopygia guttata). Anim. Behav. 27, 376-380.
129

Neubauer, R. L. 1999 Super-normal length song preferences of female zebra finches (Taeniopygia guttata) and a theory of the evolution of bird song. Evol. Ecol. 13, 365-380.
Nowicki, S., Searcy, W. A. & Peters, S. 2002 Quality of song learning affects female response to male bird song. Proc. R. Soc. Lond. B 269, 1949-1954.
Phan, M. L., Pytte, C. L. & Vicario, D. S. 2006 Early auditory experience generates long-lasting memories that may subserve vocal learning in songbirds. Proc. Natl. Acad. Sci. USA 103, 1088-1093.
Pinaud, R. & Terleph, T. A. 2008 A songbird forebrain area potentially involved in auditory discrimination and memory formation. J. Biosci. 33, 145-155.
Riebel, K. 2000 Early exposure leads to repeatable preferences for male song in female zebra finches. Proc. R. Soc. Lond. B 267, 2553-2558.
Riebel, K. & Smallegange, I. M. 2003 Does zebra finch (Taeniopygia guttata) preference for the (familiar) father's song generalize to the songs of unfamiliar brothers? J. Comp. Psychol. 117, 61-66.
Searcy, W. A. & Brenowitz, E. A. 1988 Sexual differences in species recognition of avian song. Nature 332, 152-154.
Sockman, K. W., Gentner, T. Q. & Ball, G. F. 2002 Recent experience modulates forebrain gene-expression in response to mate-choice cues in European starlings. Proc. R. Soc. Lond. B 269, 2479-2485.
Stripling, R., Milewski, L., Kruse, A. A. & Clayton, D. F. 2003 Rapidly learned songdiscrimination without behavioral reinforcement in adult male zebra finches (Taeniopygia guttata). Neurobiol. Learn. Mem. 79, 41-50.
Stripling, R., Volman, S. F. & Clayton, D. F. 1997 Response modulation in the zebra finch neostriatum: relationship to nuclear gene regulation. J. Neurosci. 17, 3883-3893.
Terleph, T. A., Mello, C. V. & Vicario, D. S. 2006 Auditory topography and temporal response dynamics of canary caudal telencephalon. J. Neurobiol. 66, 281-292.
Terleph, T. A., Mello, C. V. & Vicario, D. S. 2007 Species differences in auditory processing dynamics in songbird auditory telencephalon. Dev. Neurobiol. 67, 1498-1510.
Terpstra, N. J., Bolhuis, J. J. & den Boer-Visser, A. M. 2004 An analysis of the neural representation of birdsong memory. J. Neurosci. 24, 4971-4977.
130

Terpstra, N. J., Bolhuis, J. J., Riebel, K., van der Burg, J. M. M. & den Boer-Visser, A. M. 2006 Localized brain activation specific to auditory memory in a female songbird. J. Comp. Neurol. 494, 784-791.
Vates, G. E., Broome, B. M., Mello, C. V. & Nottebohm, F. 1996 Auditory pathways of caudal telencephalon and their relation to the song system of adult male zebra finches (Taeniopygia guttata). J. Comp. Neurol. 366, 613-642.
Vates, G. E., Vicario, D. S. & Nottebohm, F. 1997 Reafferent thalamo-"cortical" loops in the song system of oscine songbirds. J. Comp. Neurol. 380, 275-290.
Woolley, S. C. & Doupe, A. J. 2008 Social context-induced song variation affects female behavior and gene expression. PLoS Biol. 6, e62.
Zann, R. 1984 Structural variation in the zebra finch distance call. Z. Tierpsychol. 66, 328-345.
131