Implications of modular song learning for species isolation

Understanding how reproductive barriers between species arise and are maintained is paramount to the study of speciation and evolutionary biology. The contribution of avian vocal learning to the process of speciation has been the subject of considerable debate, with some suggesting that vocal learning has strongly promoted rapid speciation in the speciesrich oscine passerines. Furthermore, vocal learning has been characterized through the focused study of a small and closely-related group of taxa and so the common understanding of vocal learning in birds may reflect this narrow taxonomic coverage. While the evolutionary origin of this complex trait remains mysterious, I address the complexities of the evolution and evolutionary implications of vocal learning in songbirds through three lines of research. In a meta-analysis of isolate studies, through comparison of isolate to normal song for 16 species, I find that the learned components of song vary significantly among species, in line with literature that finds different degrees of genetic guidance in the development of song production modules. These findings form the basis for the theory of modular vocal learning. In a Setophaga hybrid zone, I investigate the potential for differential evolution in response to secondary contact for song traits according to their degree of genetic guidance by pairing a production-based acoustic analysis with a geographic cline analysis. I find that song rhythm shows a cline that resembles that of a quantitative trait under moderate iv selection while other song traits do not. In the same system, female response to song differs according to song rhythm in a way that reflects the cline shape for song rhythm uncovered previously; this suggests that this song trait, which is presumably under stricter genetically guided development than other aspects of song production, is important in mate attraction. In contrast, males do not respond differently to song based on any measured song trait. Together, my research results show that vocal learning contributes variably to the acoustic features of song across diverse species and that this contribution may not necessarily be informative during mating decisions. In contrast, components of song more likely to be under genetic control appear to function in mate attraction and evolve accordingly. From my results, vocal learning does not appear to function in accelerating speciation rates, and so the details of the origin and maintenance of vocal learning in songbirds remain open for investigation.

IMPLICATIONS OF MODULAR SONG LEARNING FOR SPECIES ISOLATION by Jay William Love A dissertation submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Biology School of Biological Sciences The University of Utah August 2019 Copyright © Jay William Love 2019 All Rights Reserved The University of Utah Graduate School STATEMENT OF DISSERTATION APPROVAL The dissertation of Jay William Love has been approved by the following supervisory committee members: Franz Goller , Chair 6/4/2019 Date Approved Frederick R. Adler , Member 5/22/2019 Date Approved David R. Carrier , Member 6/11/2019 Date Approved Karen Kapheim , Member 5/22/2019 Date Approved Ayako Yamaguchi , Member 5/22/2019 Date Approved and by M. Denise Dearing the Department/College/School of , Chair/Dean of Biological Sciences and by David B. Kieda, Dean of The Graduate School. ABSTRACT Understanding how reproductive barriers between species arise and are maintained is paramount to the study of speciation and evolutionary biology. The contribution of avian vocal learning to the process of speciation has been the subject of considerable debate, with some suggesting that vocal learning has strongly promoted rapid speciation in the speciesrich oscine passerines. Furthermore, vocal learning has been characterized through the focused study of a small and closely-related group of taxa and so the common understanding of vocal learning in birds may reflect this narrow taxonomic coverage. While the evolutionary origin of this complex trait remains mysterious, I address the complexities of the evolution and evolutionary implications of vocal learning in songbirds through three lines of research. In a meta-analysis of isolate studies, through comparison of isolate to normal song for 16 species, I find that the learned components of song vary significantly among species, in line with literature that finds different degrees of genetic guidance in the development of song production modules. These findings form the basis for the theory of modular vocal learning. In a Setophaga hybrid zone, I investigate the potential for differential evolution in response to secondary contact for song traits according to their degree of genetic guidance by pairing a production-based acoustic analysis with a geographic cline analysis. I find that song rhythm shows a cline that resembles that of a quantitative trait under moderate selection while other song traits do not. In the same system, female response to song differs according to song rhythm in a way that reflects the cline shape for song rhythm uncovered previously; this suggests that this song trait, which is presumably under stricter genetically guided development than other aspects of song production, is important in mate attraction. In contrast, males do not respond differently to song based on any measured song trait. Together, my research results show that vocal learning contributes variably to the acoustic features of song across diverse species and that this contribution may not necessarily be informative during mating decisions. In contrast, components of song more likely to be under genetic control appear to function in mate attraction and evolve accordingly. From my results, vocal learning does not appear to function in accelerating speciation rates, and so the details of the origin and maintenance of vocal learning in songbirds remain open for investigation. iv For the birds. What does wild do that tame does not? - Sarah Audsley TABLE OF CONTENTS ABSTRACT....................................................................................................................... iii LIST OF FIGURES............................................................................................................. x LIST OF TABLES..............................................................................................................xi ACKNOWLEDGEMENTS...............................................................................................xii Chapters 1. INTRODUCTION........................................................................................................... 1 1.1 References............................................................................................................................... 9 2. SONG FEATURE SPECIFIC ANALYSIS OF ISOLATE SONG REVEALS INTERSPECIFIC VARIATION IN LEARNED COMPONENTS.................................. 14 2.1 Abstract........................................................................................................... 15 2.2 Introduction..................................................................................................... 15 2.3 Methods........................................................................................................... 16 2.4 Results............................................................................................................. 24 2.5 Discussion....................................................................................................... 29 2.6 Literature Cited................................................................................................32 3. DIVERGENT EFFECTS OF SELECTION ON SONG MODULES IN A SETOPHAGA HYBRID ZONE......................................................................................... 35 3.1 Abstract................................................................................................................................. 35 3.2 Introduction.......................................................................................................................... 36 3.3 Methods................................................................................................................................. 40 3.4 Results....................................................................................................................................44 3.5 Discussion............................................................................................................................ 49 3.6 References............................................................................................................................ 57 4. EVIDENCE FOR DIFFERENTIAL SEXUAL SELECTION ON VARIOUS SONG COMPONENTS IN A SETOPHAGA HYBRID ZONE....................................................74 4.1 Abstract..................................................................................................................................74 4.2 Introduction.......................................................................................................................... 75 4.3 Methods................................................................................................................................. 79 4.4 Results....................................................................................................................................87 4.5 Discussion.............................................................................................................................91 4.6 References.............................................................................................................................94 5. CONCLUSION............................................................................................................103 5.1 Modular Song Learning................................................................................................. 103 5.2 The Evolution of Oscine Vocal Learning................................................................. 106 5.3 Learning and Speciation.................................................................................................110 5.4 References.......................................................................................................................... 111 ix LIST OF FIGURES Figures 2.1 Proof of concept for method of creating sound file from spectrograms.................... 18 2.2 Interspecific variation in isolate song disparity......................................................... 24 2.3 Syllable acoustic morphology.................................................................................... 25 2.4 Example syllable comparisons with scores................................................................25 2.5 Results of the rhythm analysis................................................................................... 26 2.6 Rhythms of trilled song elements.............................................................................. 28 2.7 Syntax........................................................................................................................ 28 3.1 Density distributions of rhythm (a,b) and syllable morphology (c,d) LD1 scores in allopatry (top) and sympatry (bottom)...................................................................... 63 3.2 Geospatial distributions of samples (dots) and kernel smoothing representations of plumage ID (top left) and rhythmic song features (all others) ............................. 64 3.3 Geographic cline analysis ......................................................................................... 65 3.4 Geographic distribution of common song types ....................................................... 66 4.1 Playback songs varied according to the number of notes in the introductory syllables .....................................................................................................................99 4.2 Results of acoustic analysis of playback songs ....................................................... 100 4.3 Response to song rhythm ........................................................................................ 101 4.4 Map of female strong responses to playback .......................................................... 102 LIST OF TABLES Tables 2.1 Results of the literature search and spectrograms used in the analysis....................... 20 2.2 Species details............................................................................................................. 23 2.3 Rhythm principal component analysis details............................................................. 27 3.1 Coefficients of rhythm linear discriminant analysis....................................................67 3.2 Coefficients of syllable morphology linear discriminant analysis.............................. 68 3.3 Song type summary statistics...................................................................................... 69 3.4 Results from Tukey post-hoc “Honest Significant Difference” test of day of year as a function of song type, showing significant results (p<0.05) only............................ 73 ACKNOWLEDGEMENTS There are many people to thank who have assisted in my ability to do this research; I will start at the beginning. Thanks to my mom, who gave me summer projects as a kid, wherein I drew birds with colored pencils and looked up facts in the encyclopedia. Thanks to my dad, who taught me about muscadine vines and poison ivy and kudzu. Thanks to my sister, who gave me the impetus to go out in the woods on my own. Mark McCool, former Hoh subdistrict ranger, hired me as a backcountry ranger for no good reason, and he first taught me the magic of Olympic. Thanks, Mark. Thanks to Mandy Holmgren. She took a risk hiring a kid from Georgia to do point counts in Washington, and she worked really hard training me on how to identify birds by ear. If I hadn’t succeeded that first summer, I would not be here, and I owe much of that modest success to Mandy. Mandy continues to run the monitoring project for IBP, and I’ve really enjoyed hanging out with the crews over the years – thanks for including me. Thanks to Scott Gremel. Scott is a role-model naturalist who answered my bird and plant questions for years. He gave me advice about academia, my career, and life. He also let me crash at his house when I was beat from fieldwork. Thinking of Port Angeles connections, I must also mention Bill Baccus, who provides a safe haven for people like me. Thanks for everything you do, Bill. At Olympic National Park, additional thanks go out to Roger Hoffman, Patti Happe, Matt Dubeau, and everyone else who did paperwork so that I could study birds there. I really appreciate it! The same goes for Joel Nowack et al. at Olympic National Forest. My fellow graduate students at the University of Utah School of Biological Sciences are a group of bright, talented, and courageous humans. They have supported me in many ways over the last five years – thanks, everyone. A special thanks goes out to my labmates: Amanda Hoepfner, Sarah Garcia, Lindsey Reader, Manon Peltier, and Natasha Verzhbitskiy. The lab atmosphere cultivated discussions, both intellectual and otherwise, that were a great benefit to my time in school. Also, thanks to my committee members, Fred Adler, Dave Carrier, Karen Kapheim, and Ayako Yamaguchi, who have been invaluable in my progress, sacrificing their time in order to help me learn. Despite your busy schedules, none of you ever turned down a request for assistance, and for that I am deeply grateful. Franz Goller has shown me what it means to be a scientist. I continue to be astounded by the breadth and depth of his understanding of the natural world and by his drive to push the boundaries of knowledge. In addition, Franz is a great person to be around and a fantastic advisor. Franz pushed me when I needed pushing and gave me space when I needed space. He supported my sometimes-questionable endeavor to study the system that I wanted to study, despite there being plenty of other options that would have provided guaranteed results. I find it very likely that I would have been unsuccessful in graduate school with any other advisor. Thank you, Franz, for everything. The final person that I will thank is Molly Powers. Molly has been my unfailing source of support and encouragement. Thanks for moving to Utah and for putting up with my absences both physical, due to field work, and mental, when I’m still thinking about xiii birds during our dinners together. Thanks for encouraging me to take the GRE and for helping me study. Thanks for taking care of me when I got sick. Thanks for understanding and supporting my passions. I absolutely could not have made it through this without you; in fact, I probably never would have started. I’m inspired by you every day, and I only hope that I can repay you for all that you’ve done for me. xiv CHAPTER 1 INTRODUCTION A major goal of biology has been to explain biodiversity. Humans have long contemplated the diversity of life forms and have made attempts to describe the natural processes that could have led to the array of organisms that we see around us today. Charles Darwin famously conjured thoughts of the physical laws of matter in the last paragraph of On the Origin of Species, relating his “view of life” to the “fixed law of gravity” (Darwin 1859). While Darwin made enormous strides towards this goal, the search for the guiding principles of life that might explain the astonishing diversity of living things continues today, driving us onward towards an ever-refined view of the world around us. For many, the simplest way to observe biodiversity is to take a walk outside. It is very likely that one of the first organisms one recognizes is a bird. Boasting a global distribution, birds have been particularly evolutionarily successful; currently, there are over 10,000 recognized bird species. Over half of these are within Passeriformes, the perching birds, and a strong majority of that portion are within the oscine clade, the songbirds, which has undergone a rapid diversification (Ericson et al. 2006). This radiation provides an excellent opportunity to investigate the driving forces behind evolutionary diversification and speciation, but the exact relationship between clades is subject to continued debate (e.g., Barker et al. 2004; Hackett et al. 2008; Lovette et al. 2010; DuBay and Witt 2012; 2 McCormack et al. 2013; Jarvis 2014; Prum et al. 2015). In addition to the numerous ongoing discussions about specific taxonomic relationships between species that arise due to differing methodology (e.g., in birds, above), there is also a healthy debate about how to define species (Mallet 1995; De Queiroz 2007), and, indeed, if “species” truly exist at all (Hey et al. 2003; reviewed in Coyne and Orr 2004). Still, a commonly used qualifier for a species is the existence of reproductive isolation, which is the criterion for the biological species concept. Whether or not this concept is appropriate for the application to living things – whether or not it constitutes a reflection of the principles guiding all of life – is a question beyond the scope of this dissertation, but suffice it to say that partial or complete reproductive isolation does exist between groups of organisms, and that such isolation is often used to define the groups as distinct species, and that furthermore the evolutionary fate of a group of organisms is determined in part by the degree to which they are isolated from outside gene flow. Therefore, a keystone to the understanding of life is understanding the evolution of limitations on gene flow between populations. A seahorse and a raven cannot interbreed for a number of reasons, including physical barriers that result in geographic isolation – the two would never meet! However, why can a raven and a robin not interbreed? Again, there are a number of reasons, which are called reproductive barriers. There are two primary forms of barriers to reproduction, defined by their timing relative to fertilization. The first category is that of the postzygotic reproductive barriers, which can be characterized by an associated reduction in the fitness of offspring, even to the point of zero fitness. Included in this category are hybrid inviability and hybrid sterility. The second category is that of prezygotic reproductive 3 barriers. Within this category there are several types, including barriers that prevent two species from coming into contact (temporal, geographic, or habitat isolation) and barriers that prevent two species that do come into contact from interbreeding. Within this last group is the type of reproductive barrier that results from differences in behavior and which contributes to behavioral reproductive isolation between two species (Coyne and Orr 2004). This last type is the form of reproductive isolation that is of a primary concern to this dissertation. There has been debate about how these reproductive barriers form. Most assuredly, many come about through arbitrary processes; they are, at least in part, the byproduct of evolutionary adaptation to unrelated selection or of neutral processes involving random genetic drift (i.e., Kimura 1983; reviewed in Uy et al. 2018). When a population splits and the two new populations that result experience different selective pressures, each population will adapt to its given circumstances and, in due course, develop reproductive barriers as a symptom of that adaptation and of the random genetic divergence accrued while out of contact from each other. If the two populations then meet and if enough divergence has occurred for them to be reproductively isolated, then they are separate species. This description forms the basis for the allopatric model of speciation, which is thought to be vastly more common than sympatric speciation (Coyne and Orr 2004; Uy et al. 2018). However, an interesting situation occurs if, upon secondary contact between two previously isolated populations, sufficient divergence has not occurred for them to be completely reproductively isolated. In particular, if premating isolation is not complete and there is at least partial postmating reproductive isolation (e.g., reduced hybrid fitness), then 4 interbreeding between the two populations is expected to be selected against (Dobzhansky 1937). The natural selection pressure to reduce hybridization in such a situation is known as reinforcing selection or simply reinforcement. Reinforcing selection is expected to result in the formation of strong prezygotic reproductive isolation through various mechanisms, leaving behind the hallmark of reinforcement: divergent reproductive character displacement and stronger prezygotic isolation in sympatry than in allopatry (Dobzhansky 1937; Coyne and Orr 2004). The importance of reinforcement to the development of premating reproductive barriers and to the speciation process has been the subject of intense debate (reviewed in Coyne and Orr 2004). Some have contended that reinforcement is not simply coincident with but necessary for complete speciation (Lewontin 1974), though that position is currently less popular (Coyne and Orr 2004). While theoretical support for the reinforcement model of speciation is strong (Pfennig 2016; Butlin and Smadja 2018; Yeh et al. 2018), empirical evidence continues to lag behind (but see Pfennig 2003; Kozak et al. 2015; and a review in Coyne and Orr 2004). One notable exception is an analysis of a large dataset in Drosophila, which has provided evidence for strong reinforcement (Yukilevich 2012). One major difficulty in identifying reinforcement in nature is that most studies rely on detecting patterns in traits that are predicted by reinforcement theory; it is difficult to reject the possibility that such patterns could have been created by other natural forces instead (Coyne and Orr 2004). Irrespective of the importance of reinforcement in their evolution, behavioral mechanisms of prezygotic reproductive isolation are common. One of the best-studied examples comes from birds, class Aves, where complex mating behavior often involves 5 multimodal signaling. Plumage coloration, complex courtship rituals, and acoustic communication have been shown to be highly divergent among closely related species (Andersson 1994; Price 2008). The interspecific diversity and integration of these modes of behavior provides an excellent opportunity for the study of reproductive isolation and its importance to speciation. Within acoustic communication in birds, song of the oscine passerines is a heavily studied topic from many disciplines and represents perhaps the best-studied example of complex mating behavior. The oscine clade contains over 5,000 species and boasts a global distribution. In this clade, song is an acoustic signal that functions in dual roles of mate attraction and territory defense (Collins 2004). While song is known to be used in several contexts (e.g., duetting, Hall 2009) and by both sexes (Odom et al. 2014), most study has focused on species where predominantly the males sing to attract females and to negotiate territorial disputes with neighboring males. Song is often highly divergent between closely related species and this divergence is thought to play an important role in establishing species limitations (reviewed in Uy et al. 2018). Importantly, oscines exhibit vocal learning of song. Vocal learning refers to the ability of an animal to alter its vocal ontogeny and vocal signals in response to auditory information (Nottebohm 1972) and is found in birds and mammals (Janik and Slater 1997; Bolhuis et al. 2010). Song is the primary vocalization that is learned in oscines, though other vocal behaviors have also been shown to be learned in some species (e.g., Zann 1990; Hughes et al. 1998). Through cultural transmission, learned song can diverge rapidly within and between populations (Mason et al. 2017), potentially leading to reproductive isolation. This feature of learned song – rapid trait 6 evolution that is relevant to mating behavior – is thought to have increased the rate of speciation in the oscines and may have contributed to the success of the oscine clade (Lachlan and Servedio 2004). In part due to limited supporting empirical evidence, this hypothesis has been subject to continued discussion (Baptista and Trail 1992; Slabbekoorn and Smith 2002; Seddon and Tobias 2007; Olofsson et al. 2011; Verzijden et al. 2012; Yeh 2018; Yeh 2019). As a result, it is currently unclear what role song learning plays in the establishment of reproductive isolation. One potential explanation for the disparity between theoretical predictions of and empirical evidence for the impact of vocal learning on speciation may be found in recent evidence that suggests song learning is a modular process wherein certain aspects of song rely to differing degrees on learning for normal development. Laboratory studies investigating the neuromotor underpinnings of song learning and production in birds indicate that song, rather than being learned in full from one of a number of identified vocal learning processes (Beecher and Brenowitz 2005; Brenowitz and Beecher 2005), is learned in distinct pieces, each of which reflect separate neuromotor pathways that vary in their specificity of genetic control (literature reviewed in Chapter 2). A further complicating matter is that very little is known about the perceptual relevance of learned aspects of song in reproductive contexts (but see discussion in Chapter 3). Given these two points, theoretical models that predict evolutionary outcomes based on an unrestricted view of song learning and that make assumptions about the relative salience of learned song may find patterns different from those that occur in natural populations. On the other hand, validation of theoretical predictions may be evidenced through focused study of natural populations, particularly those currently experiencing secondary contact where evidence 7 of reinforcing selection on mating signals could be observed. There currently exist many areas of secondary contact between oscine species pairs. Secondary contact and subsequent hybridization, once thought to be rare, is in fact relatively common, especially in birds (Mallet 2005; Taylor and Larson 2019). In the oscine family Parulidae, for example, over half of the species found in North America are known to hybridize, many of them in maintained hybrid zones (Willis et al. 2013). One of these zones is a narrow contact zone between the hermit warbler (Setophaga occidentalis) and its closest relative, the Townsend’s warbler (S. townsendi). Study of this system in the late 1980s through the 2000s made interesting discoveries related to plumage traits, aggression, genetic introgression, and hybrid zone dynamics, but song was not thoroughly studied (Pearson 1997; Pearson and Rohwer 1998; Rohwer and Wood 1998; Pearson 2000; Pearson and Manuwal 2000; Pearson and Rohwer 2000; Smith and Rohwer 2000; Rohwer et al. 2001; Owen-Ashley and Butler 2004; Rohwer et al. 2007; Krosby and Rohwer 2010). The baseline of research on this system makes it an attractive venue for the investigation of the role of learned song in the establishment and maintenance of reproductive isolation. In particular, evidence for postzygotic reproductive barriers in the hermit warbler / Townsend’s warbler hybrid zone was found (Pearson 1997; Pearson and Rohwer 1998), and so one could expect to find reproductive character displacement that reflects reinforcing selection. With this in mind, we designed our research to explore the evolutionary response of song to secondary contact between hermit and Townsend’s warblers. Furthermore, we attempted to draw conclusions about mechanisms underlying any observed trait change and tried to identify any implications of song learning for the speciation process uncovered in the course of our study. 8 Chapter 2 of this dissertation details an analysis of song development in 16 species. For each of these species, we compare song produced by normally-reared birds to song of birds with no exposure to conspecific song, and thus no opportunity to develop song through imitative vocal learning. We leverage vocal production mechanisms to develop a new approach to acoustic analysis and introduce a technique to reconstruct sound from published spectrogram images. We find no evidence of a phylogenetic signal in the degree of learning required for normal development in each of three production modules: syllable morphology, song rhythm, and syntax. We do find that song rhythm shows a lower degree of learning required for normal development than the other two modules, suggesting that this trait is under more strict genetic control. Furthermore, our results provide comparative evidence in support of the modularity of song learning in oscines. In Chapter 3, we use a production-based acoustic analysis to assess song trait change across the hermit / Townsend’s warbler hybrid zone. We construct geographic clines and assess differential song trait evolution in response to secondary contact. Our results further support the predominantly genetic control for song rhythm in these species and suggest that syllable morphology is comparatively more susceptible to change via cultural, rather than genetic, transmission. Intriguing patterns of bidirectional convergence for these two traits are discussed in detail along with potential explanations. The difference in cline shape between song rhythm and plumage traits suggests the possibility of positive directional selection for Townsend’s-like song rhythm within hermit warblers. In Chapter 4, we use a playback experiment to directly test predictions made in Chapter 3. Our methods involve establishing a panel of playback songs that differ systematically in key traits and then quantifying male and female responses to song in field 9 sites distributed across allopatric and sympatric locales. We find that males respond relatively equally to all songs regardless of acoustic characteristics, but that females respond more strongly to songs that exhibit Townsend’s-like rhythm. In addition, allopatric Townsend’s warbler males are more aggressive than allopatric hermit warblers, but this difference declines in sympatry, suggesting that hybridization coincides with reduced territorial aggression. 1.1 References Andersson MB (1994) Sexual selection. Princeton University Press Baptista LF, Trail PW (1992) The role of song in the evolution of passerine diversity. Syst Biol 41(2):242–247 Barker FK, Cibois A, Schikler P, Feinstein J, Cracraft J (2004) Phylogeny and diversification of the largest avian radiation. Proc Natl Acad Sci USA 101(30):11040– 11045. doi:10.1073/pnas.0401892101 Beecher MD, Brenowitz E (2005) Functional aspects of song learning in songbirds. Trends Ecol Evol 20(3):143–9 doi:10.1016/j.tree.2005.01.004 Bolhuis JJ, Okanoya K, Scharff C (2010) Twitter evolution: Converging mechanisms in birdsong and human speech. Nat Rev Neurosci 11(11):747–759. doi:10.1038/nrn2931 Brenowitz EA, Beecher MD (2005) Song learning in birds: Diversity and plasticity, opportunities and challenges. Trends Neurosci 28(3):127–132 Butlin RK, Smadja CM (2018) Coupling, reinforcement, and speciation. Am Nat 191(2):155–172 doi:10.1086/695136 Collins S (2004) Vocal fighting and flirting: The functions of birdsong. In: Nature’s music. Elsevier pp 39–79 Coyne JA, Orr HA (2004) Speciation. Sunderland, MA: Sinauer Associates, Inc Darwin C (1859) On the origin of species Dobzhansky T (1937) Genetics and the origin of species. New York: Colombia Univ. Press DuBay SG, Witt CC (2012) An improved phylogeny of the Andean tit-tyrants (Aves, Tyrannidae): More characters trump sophisticated analyses. Mol Phylogenet Evol 10 64(2):285–296. doi:10.1016/j.ympev.2012.04.002 Ericson PGP, Anderson CL, Britton T, Elzanowski A, Johansson US, Källersjö M, Ohlson JI, Parsons TJ, Zuccon D, Mayr G (2006) Diversification of Neoaves: Integration of molecular sequence data and fossils. Biol Lett 2(4):543–547. doi:10.1098/rsbl.2006.0523 Hackett SJ, Kimball RT, Reddy S, Bowie RCK, Braun EL, Braun MJ, Chojnowski JL, Cox WA, Han K-L, Harshman J (2008) A phylogenomic study of birds reveals their evolutionary history. Science 320(5884):1763–1768 Hall ML (2009) A review of vocal duetting in birds. Adv Study Behav 40(09):67–121 doi:10.1016/S0065-3454(09)40003-2 Hey J, Waples RS, Arnold ML, Butlin RK, Harrison RG (2003) Understanding and confronting species uncertainty in biology and conservation. Trends Ecol Evol 18(11):597–603. doi:10.1016/j.tree.2003.08.014 Hughes M, Nowicki S, Lohr B (1998) Call learning in black-capped chickadees (Parus atricapillus): The role of experience in the development of “chick-a-dee” calls. Ethology 104:232–249 doi:10.1111/j.1439-0310.1998.tb00065.x Janik VM, Slater PJB (1997) Vocal learning in mammals. Adv Study Behav 26:59–100 Jarvis ED et al. (2014) Whole-genome analyses resolve early branches in the tree of life of modern birds. Science 346(6215):1126–1138 Kimura M (1983) The neutral theory of molecular evolution. Cambridge University Press Kozak GM, Roland G, Rankhorn C, Falater A, Berdan EL, Fuller RC. 2015. Behavioral isolation due to cascade reinforcement in lucania killifish. Am Nat 185(4):491–506 doi:10.1086/680023 Krosby M, Rohwer S (2010) Ongoing movement of the hermit warbler x Townsend’s warbler hybrid zone. Fenton B, editor. PLoS One 5(11):e14164 doi:10.1371/journal.pone.0014164 Lachlan RF, Servedio MR (2004) Song learning accelerates allopatric speciation. Evolution 58(9):2049–2063 Lovette IJ, Pérez-Emán JL, Sullivan JP, Banks RC, Fiorentino I, Córdoba-Córdoba S, Echeverry-Galvis M, Barker FK, Burns KJ, Klicka J, et al. (2010) A comprehensive multilocus phylogeny for the wood-warblers and a revised classification of the Parulidae (Aves). Mol Phylogenet Evol 57(2):753–70 doi:10.1016/j.ympev.2010.07.018 Mallet J (1995) A species definition for the modern synthesis. Trends Ecol Evol 10(7):294– 299 doi:10.1016/0169-5347(95)90031-4 Mallet J (2005) Hybridization as an invasion of the genome. Trends Ecol Evol 20(5):229– 11 237 doi:10.1016/j.tree.2005.02.010 Mason NA, Burns KJ, Tobias JA, Claramunt S, Seddon N, Derryberry EP (2017) Song evolution, speciation, and vocal learning in passerine birds. Evolution 71(3):786–796 doi:10.1111/evo.13159 McCormack JE, Harvey MG, Faircloth BC, Crawford NG, Glenn TC, Brumfield RT (2013) A phylogeny of birds based on over 1,500 loci collected by target enrichment and high-throughput sequencing. PLoS One 8(1):e54848 doi:10.1371/journal.pone.0054848 Nottebohm F (1972) The origins of vocal learning. Am Nat 106(947):116–140 Odom KJ, Hall ML, Riebel K, Omland KE, Langmore NE (2014) Female song is widespread and ancestral in songbirds Nat Commun. 5:1–6 doi:10.1038/ncomms4379 Olofsson H, Frame AM, Servedio MR (2011) Can reinforcement occur with a learned trait? Evolution 65(7):1992–2003 doi:10.1111/j.1558-5646.2011.01286.x Owen-Ashley NT, Butler LK (2004) Androgens, interspecific competition and species replacement in hybridizing warblers. Proc R Soc B Biol Sci 271(SUPPL.):S498–S500 doi:10.1098/rsbl.2004.0230 Pearson S (1997) Behavioral and ecological tests of four models explaining narrow hybrid zones between hermit and Townsend’s warblers (PhD dissertation). University of Washington Pearson S (2000) Behavioral asymmetries in a moving hybrid zone. Behav Ecol 11(1):84– 92 doi:10.1093/beheco/11.1.84 Pearson S, Manuwal DA (2000) Influence of niche overlap and territoriality on hybridization between hermit warblers and Townsend’s warblers. Auk 117(1):175–183 Pearson S, Rohwer S (1998) Influence of breeding phenology and clutch size on hybridization between hermit and Townsend’s warblers. Auk 115(3):739–745 Pearson S, Rohwer S (2000) Asymmetries in male aggression across an avian hybrid zone. Behav Ecol 11(1):93–101 doi:10.1093/beheco/11.1.93 Pfennig KS (2003) A test of alternative hypotheses for the evolution of reproductive isolation between spadefoot toads: Support for the reinforcement hypothesis. Evolution 57(12):2842–2851 doi:10.1111/j.0014-3820.2003.tb01525.x Pfennig KS (2016) Reinforcement as an initiator of population divergence and speciation. Curr Zool 62(2):145–154 doi:10.1093/cz/zow033 Price T (2008) Speciation in birds. Greenwood Villiage, CO: Roberts and Co. Prum RO, Berv JS, Dornburg A, Field DJ, Townsend JP, Lemmon EM, Lemmon AR 12 (2015) A comprehensive phylogeny of birds (Aves) using targeted next-generation DNA sequencing. Nature 526(7574):569-573 doi:10.1038/nature15697 De Queiroz K (2007) Species concepts and species delimitation. Syst Biol 56(6):879–886 doi:10.1080/10635150701701083 Rohwer S, Bermingham E, Wood C (2001) Plumage and mitochondrial DNA haplotype variation across amoving hybrid zone. Evolution 55(2):405–422 Rohwer S, Martin PR (2007) Time since contact and gene flow may explain variation in hybrid frequencies among three Dendroica townsendi × D. occidentalis (parulidae) hybrid zones. Auk 124(4):1347-1358 Rohwer S, Wood C (1998) Three hybrid zones between hermit and Townsend’s warblers in Washington and Oregon. Auk 115(2):284–310 Seddon N, Tobias JA (2007) Song divergence at the edge of Amazonia: An empirical test of the peripatric speciation model. Biol J Linn Soc 90(1):173–188 doi:10.1111/j.10958312.2007.00753.x Slabbekoorn H, Smith TB (2002) Bird song, ecology and speciation. Philos Trans R Soc Lond B Biol Sci 357(1420):493–503 doi:10.1098/rstb.2001.1056 Smith CE, Rohwer S (2000) A phenotypic test of Haldane’s rule in an avian hybrid zone. Auk 117(3):578–585 Taylor SA, Larson EL (2019) Insights from genomes into the evolutionary importance and prevalence of hybridization in nature. Nat Ecol Evol 3(2):170–177 doi:10.1038/s41559018-0777-y Uy JAC, Irwin DE, Webster MS (2018) Behavioral isolation and incipient speciation in birds. Annu Rev Ecol Evol Syst 49:1–24 doi:10.1146/annurev-ecolsys-110617-062646 Verzijden MN, ten Cate C, Servedio MR, Kozak GM, Boughman JW, Svensson EI (2012) The impact of learning on sexual selection and speciation. Trends Ecol Evol 27(9):511– 519 doi:10.1016/j.tree.2012.05.007 Willis PM, Symula RE, Lovette IJ (2013) Ecology, song similarity and phylogeny predict natural hybridization in an avian family. Evol Ecol 28(2):299–322 doi:10.1007/s10682013-9677-4 Yeh DJ (2018) The interaction between learning and speciation (PhD dissertation). University of North Carolina Yeh DJ (2019) Assortative mating by an obliquely transmitted local cultural trait promotes genetic divergence: A model. Am Nat 193(1):81–92 doi:10.1086/700958 Yeh DJ, Boughman JW, Sætre GP, Servedio MR (2018) The evolution of sexual imprinting 13 through reinforcement. Evolution 72(7):1336–1349 doi:10.1111/evo.13500 Yukilevich R (2012) Asymmetrical patterns of speciation uniquely support reinforcement in drosophila. Evolution 66(5):1430–1446 doi:10.1111/j.1558-5646.2011.01534.x Zann R (1990) Song and call learning in wild zebra finches in south-east Australia. Anim Behav 40(5):811–828 CHAPTER 2 SONG FEATURE SPECIFIC ANALYSIS OF ISOLATE SONG REVEALS INTERSPECIFIC VARIATION IN LEARNED COMPONENTS Printed with permission from: Love J, Hoepfner A, Goller F (2019) Song feature specific analysis of isolate song reveals interspecific variation in learned components. Developmental Neurobiology 79(4):350369 doi:10.1002/dneu.22682 15 Song Feature Specific Analysis of Isolate Song Reveals Interspecific Variation in Learned Components Jay Love ,1 Amanda Hoepfner,1 Franz Goller 1,2 1 School of Biological Sciences, University of Utah, 257 South 1400 East, Salt Lake City, Utah 84112 2 Institute for Zoophysiology, University of Muenster, Muenster, Germany Received 27 November 2018; revised 14 April 2019; accepted 15 April 2019 ABSTRACT: Studies of avian vocal development without exposure to conspecific song have been conducted in many passerine species, and the resultant isolate song is often interpreted to represent an expression of the genetic code for conspecific song. There is wide recognition that vocal learning exists in oscine songbirds, but vocal learning has only been thoroughly investigated in a few model species, resulting in a narrow view of birdsong learning. By extracting acoustic signals from published spectrograms, we have reexamined the findings of isolate studies with a universally applicable semi-automated quantitative analysis regimen. When song features were analyzed in light of three different production aspects (respiratory, syringeal, and central programming of INTRODUCTION Vocal learning is a complex behavior that has evolved only in a few taxa. It entails the postnatal acquisition of at least some aspects of the vocal repertoire, which are then reproduced through vocal imitation (Brenowitz and Beecher, 2005). The song of oscine passerine birds, which functions mainly in mate Correspondence to: J. Love ([e-mail]). Contract grant sponsor: George R. Riser Research Award Contract grant sponsor: NIH; contract grant number: DC 06876. Additional Supporting Information may be found in the online version of this article. © 2019 Wiley Periodicals, Inc. Published online 00 Month 2019 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/dneu.22682 sequence), all three show marked interspecific variability in how close isolate song features are to normal. This implies that song learning mechanisms are more variable than is commonly recognized. Our results suggest that the interspecific variation shows no readily observable pattern reflecting phylogeny, which has implications for understanding the mechanisms behind the evolution of avian vocal communication. We emphasize that song learning in passerines provides an excellent opportunity to investigate the evolution of a complex, plastic trait from a phylogenetic perspective. © 2019 Wiley Periodicals, Inc. Develop Neurobiol 00: 1–20, 2019 Keywords: vocal learning; acoustic analysis; isolate song; sexual selection; vocal production; oscine attraction and territorial behavior, is a learned vocal behavior and is highly variable across the more than 5,000 species. In light of the variability in song characteristics, it is interesting to understand which song features must be acquired after hatching and to what degree species differ in these requirements. Although various pieces of evidence can be found in the long history of investigation of birdsong, this question has not been addressed in a comprehensive analysis. Modern understanding of birdsong learning has been mostly shaped by evidence from a few model systems, which gave rise to the template theory for vocal learning (Konishi, 1965; Marler, 1970; Marler and Sherman, 1983; Marler, 1997). This theory posits that a genetically encoded model (i.e., “innate template”) guides song acquisition during the sensory phase of vocal development, such that young birds will acquire conspecific rather than heterospecific song templates. 1 16 2 Love et al. During normal development, this genetically guided acquisition of a song model is followed by vocal production during the sensorimotor phase of vocal development, in which a bird practices and refines the imitation of the acquired song template. During this phase the young bird trains its own vocalizations to match the features of the model songs, which are recalled from memory. At the end of ontogeny (or seasonally at the end of each plastic period) song features become highly stereotyped (crystallization). More recent behavioral and neurobiological evidence suggests that social interactions also shape how vocal production proceeds during the sensorimotor period (West and King, 1988; Fehér et al., 2009; Caruso-Peck and Goldstein, 2019), thus expanding the template model for vocal ontogeny (Soha, 2017). Young oscines sing irrespective of whether they have been exposed to model song. Untutored birds practice singing and, although their plastic period seems to be extended relative to tutored birds (Marler, 1970), eventually arrive at a stereotyped song (“isolate song”). Isolate song displays aberrant features compared to song from birds with exposure to conspecific song, and studies of isolates have therefore been a key tool in avian vocal learning research. If imitative vocal learning is utilized for normal song development in a species, individuals of that species raised in acoustic isolation, and thus without the opportunity to acquire song models or to receive social input, will develop songs that are guided only by the genetically encoded information about song. Thus, song features that develop normally in isolates can be understood to be genetically encoded, while features that are missing or aberrant must require input from acquired song models or social feedback. Given the diversity of oscines, one can expect considerable diversity of song learning programs, but this topic has received only modest attention (Beecher and Brenowitz, 2005; Brenowitz and Beecher, 2005; Beecher, 2017), and the discussions have focused on different vocal learning strategies (e.g., imitation, improvisation and invention; closed-ended versus open-ended learning). However, even a cursory look at isolate songs from different oscine species reveals a range of deficits compared to normal song (e.g., Marler et al., 1962; Marler, 1970; Kroodsma et al., 1997; Leitner et al., 2002). It is therefore tempting to conclude that the interspecific variation in how much isolate song differs from normal song can be attributed to continuous variation in the degree to which the genetically encoded template guides the emergence of specific features of song. To what degree different species vary in their need for learning of specific features is largely unknown. Developmental Neurobiology A problem with a comparative analysis of divergent song features is how to define and quantify them. Here, we approach this problem by using song production mechanisms to identify song features, which involve, at least in part, distinct brain circuitry for controlling the different underlying motor systems. These different circuitries enable us to meaningfully categorize song features into modules which can be acted on independently by selective forces. Sound is produced by respiratory air pulses which drive oscillations of the labia. The nature of these vibrations (i.e., frequency, pulse shape, etc.) can be controlled by syrinx muscles. Syringeal motor control therefore shapes the acoustic morphology of individual song elements somewhat independently of the respiratory activity. In addition, upper vocal tract movements modify the harmonic content of the generated sound. We define this central control mechanism as Module 1. Respiratory activity involves expiratory and inspiratory muscles, which generate the air pulses for vibration and refill the air supply, respectively. This alternation of sound and silence contributes the coarse rhythm of song. We consider the respiratory pace making circuitry underlying song patterns as Module 2 (e.g., Suthers et al., 1999; Riede and Goller, 2014; Beckers, 2013). As a distinct, third neural programming task, we consider the sequential arrangement of song elements into a syntactical pattern. This third module is likely controlled by the elaborate circuit network for production of learned vocal behavior (e.g., Bolhuis et al., 2010). Learning of song, then, must involve the coordination of these modules of production into an integrated control system (Veit et al., 2011; Nick, 2015). The interspecific variation in the degree to which species rely on imitation for normal song development (Beecher and Brenowitz, 2005) might be attributed to differences in how neural control in these song production modules develops. For example, different species might rely more or less on learning for normal production of song rhythm, acoustic morphology of syllables, or song syntax. If this is the case, we expect to see interspecific variation in isolate song distributed across the ranges on these three axes, reflecting differential reliance between species on acquired information across modules. Here, we investigate the interplay between genetic control and learning-mediated features of song phenotypes across species. METHODS A literature search was conducted for studies on individual birds raised in acoustically and socially controlled environments and that included illustrations of 17 Variation in Learned Components of Song spectrograms of isolate song. Spectrographic images from .pdf files of each isolate and normal song were extracted as separate .png files, using image editing software (Adobe Photoshop CC 2015.5). The only editing consisted of setting the background of the spectrogram to white, if the page had become discolored when scanned, and of rotating the image to ensure that the spectrogram was oriented horizontally. Time stamp and frequency axes were included in the .png images. The .png file was then opened in an imageto-sound editing and synthesizing software package (Photosounder 1.9.3, photosounder.com). This software enables the conversion of images to sound and uses the standard format for spectrograms: frequency on the y axis, time on the x axis, and the depth of color relating to the amplitude. It was important to use the invert function to reverse the colors of the image, since the Photosounder software interprets white as sound and black as silence. Using the linear frequency scale setting, the frequency and time scales in the Photosounder window were manually aligned with the frequency and time scales in the .png image. This was completed by adjusting the minimum and maximum frequency sliders and the time resolution slider. The time and frequency stamps were then erased from the image using the editing tools in the Photosounder package. The image was converted to a 44.1 kHz .wav sound file by saving in the Photosounder software. To aid in the automated portion of the analysis, the name of each file included the publication source (author and year), figure number, species identity, and isolate/ normal identity. The process of converting spectrographic images to sound as detailed above consistently reconstructs the sounds that the spectrographic images represent. Visual examination of the spectrographic representation of sounds before and after the conversion process indicates a strong match. Song samples are provided in the supplementary materials and spectrograms and power spectra of an original sound accessed from xeno-canto.org and its spectrogram-converted sound are provided in Figure 1, along with original and converted syllables with similarity scores. Adjusting the Gamma control in Photosounder can make this conversion extremely accurate if an original recording is available for reference, but we kept the Gamma control set to default for all conversions in order to avoid potential artifacts. This process works consistently, but it should be noted that the sound produced through conversion is dependent upon the settings and conditions of the original recordings and spectrogram settings used to make the spectrogram image. Fortunately, this potential source 3 of artifacts, which we expect to be minimal, is effectively randomized for our purposes and should not favor one species over another. Additionally, there is no reason to assume that spectrograms created for the same species employed drastically different settings. Finally, our analyses (which use cross-correlational and other measures of similarity rather than being characterized by comparisons of exact frequency measurements) are fairly robust to minor differences in the distribution of frequencies of sounds. Song samples of isolate zebra finch, Taeniopygia guttata, brown-headed cowbird, Molothrus ater, and yellow-headed blackbirds, Xanthocephalus xanthocephalus, were obtained from individuals raised in a laboratory environment from a young age in acoustic isolation from adult, conspecific song (Jarrell, 2009; Hoepfner and Goller, 2013; Goller, unpublished data). Other species for which song samples were acquired via the above spectrogram conversion process differed in some aspects of the raising regime, but all were kept in acoustic isolation from adult, conspecific song from an early age (Table 1). Detailed information about exact sample sizes of songs and syllables for each species, as well information about the raising regime used in each publication from which song spectrograms were pulled, are detailed in Table 2. Acoustic Analysis: Syllable Morphology Each.wav file was opened in Sound Analysis Pro software (SAP) (Tchernichovski et al., 2000). A syllable table (spreadsheet) was created for each species by opening each song for that species, outlining the full song in the display window, selecting automated segmentation based on amplitude, and adding records to the table. Proper segmentation was confirmed using the red segmentation bars in the visual display window. For most songs, a threshold of 20.0 dB was satisfactory to ensure proper segmentation, but occasionally the amplitude threshold for segmentation was increased so that only syllables were highlighted or decreased so that definition between syllables was evident. The syllable tables were exported to .xlsx, then converted to .csv files for use in R (R Core Team, 2017). The syllable tables contained the following measures for each syllable: file name, syllable duration, syllable start, means and variances of amplitude, pitch, FM (frequency modulation), AM2 (squared amplitude modulation), entropy, pitch goodness, and mean frequency. Each species, then, had an associated table with data from all syllables from all songs in the dataset, including songs of both isolate- and normally reared birds. Developmental Neurobiology 18 4 Love et al. With the goal of quantifying overall acoustic similarity of syllables, custom code was constructed in R to identify the best possible normal match for every isolate syllable, the best possible normal match for each normal syllable, the best possible isolate match for each (a) normal syllable, and the best possible isolate match for each isolate syllable. While we used a cross-correlation approach to measure syllable similarity (see below), conducting cross-correlation for every possible combination of syllables and then selecting the best match (b) XC371330 Original 8 8 Frequency (kHz) 10 Frequency (kHz) 10 6 4 2 XC371330 Converted 6 4 2 1 Time (s) 0.5 0 (c) 0 2 1.24 Time (s) 0.5 (d) 2.48 Sound pressure level (dB/Hz) Sound pressure level (dB/Hz) 40 20 0 0 3103.64 6201.84 (e) Original 1 12 20 0 22050 0 10 8 8 6 6 4 4 2 2 0.1 10 10 8 8 6 4 4 2 2 0 0.1 Developmental Neurobiology Time (s) 0.1 (h) Converted 2 0.90 12 6 Frequency (Hz) (f) Converted 1 0 (g) Original 2 12 3083.13 6125.07 0.87 12 10 0 Frequency (kHz) Frequency (Hz) 40 0 0.1 22050 19 Variation in Learned Components of Song 5 Figure 1 Proof of concept for method of creating a sound file from spectrograms. (A) spectrogram of field recorded Vesper Sparrow (xeno-canto.org, XC371330) created with Praat software. No filtering was conducted. This spectrogram image was then converted to sound using Photosounder software, following the procedure in the methods section of the text. (B) spectrogram of the image-converted sound created with Praat software with identical settings to those used in (A). Note that, while sound depicted in (B) may at first appear temporally compressed compared to (a), silence has been added before and after the sound in the conversion process and that the time stamp and axis is therefore changed compared to (A). (C) and (D) are power spectra, taken from the same point in the song (the exact midpoint). These plots show the distribution of frequencies at a given time point. In the conversion process, background noise and low-amplitude sounds are sometimes lost or distorted, but the primary frequency peaks are quite accurate. The overall accuracy of the conversion process is highly reliant upon the original recording quality and the quality of the published spectrogram. (E) and (G) show high-quality recordings of brown-headed cowbird, Molothrus ater, syllables. Recordings of this quality are typical of those typically acquired and published in the isolate studies that are included in our analysis. (F) and (H) show spectrograms of the same syllables after going through the spectrogram image conversion process. The similarity scores between original and image-converted sounds are shown between panels: the 0.87 and 0.90 maximum correlation values signify only minor degradation of the signal that should not impact the conclusions of our study. would be unnecessarily computationally expensive. So, we found the best-matching syllable pairs using a pipeline approach, where key acoustic features of each syllable were used to narrow down the pool of possible matches. A syllable was determined to be a match if it came within 40% of the target syllable’s pitch, FM, and AM2. Of the pool of syllables that met these requirements, the member which most closely matched the target syllable’s duration was identified as the best match. If no syllable met these requirements, then the target syllable was defined as having no potential matches from the matching pool. Short .wav files were then created for each member of each pair of matching syllables by extracting the portion of the song .wav file from which the syllable originated. This was automated in R using custom code that relied heavily on the “Extract part…” functionality from the PraatR package (Albin, 2014). Syllable-length .wav files were visually examined using spectrogram visualization in AvisoftSASLab Pro (v. 5.2) to ensure that the appropriate portion of the song-length .wav file was extracted by the automated process. Finally, each pair of syllables—the target syllable and its best match— were scored for similarity using cross-correlation of the amplitude envelopes of the syllable-only .wav files and recording the maximum correlation (corenv function from the seewave package in R) (Sueur et al., 2008). The maximum correlation was used to quantify similarity in syllable morphology, after being adapted as in the following methods. When members of syllable pairs did not match in duration, a silent period of a duration equal to the difference in duration between the two syllables was added to the end of the shorter syllable in order to allow for cross-correlation to proceed (addsilw function from the seewave package). Those syllables that were not found to have a match, and thus for which cross-correlation was not conducted, were given a score of zero in place of the maximum correlation. To verify that our pipeline approach gave us similar answers to those that would be provided by conducting cross-correlation on all possible pairs of syllables, we completed this alternate analysis on one species, the wood thrush (Hylocichla mustelina). We found that the two methods differed in the best match for a given syllable ~ 90% of the time, but that this discrepancy did not significantly affect the average similarity score for the species (means of 0.16 and 0.14, t-test P-value = 0.7). We therefore are confident that both approaches provide reliable data to address our question. Since it was found that difference in duration of compared syllables had a significant effect on maximum correlation reported by the above method (simple linear regression: adjusted R-squared = 0.044, P < 0.0001), the reported values were corrected to account for this effect by utilizing the residuals of the data from the linear regression of maximum correlation score to difference in duration. Thus, the residuals (i.e., the magnitudes of the differences between the reported maximum correlations between syllable pairs and the maximum correlations predicted by the linear regression between correlation values and differences in syllable pair duration) become our basis for summary statistics of similarity. For each species, the mean of isolate-to-normal residuals was subtracted from the mean of the normal-to-normal residuals to create an overall summary statistic of similarity for each species. A positive score indicates that isolate syllables are, on average, different from normal syllables, whereas a negative score indicates that isolate syllables are found to match normal syllables better than normal syllables match other normal syllables. A Developmental Neurobiology Developmental Neurobiology Icteridae Emberizidae Yellow-headed Xanthocephalus YHBL blackbird xanthocephalus Zonotrichia leucophrys Zonotrichia leucophrys Zonotrichia leucophrys Melospiza melodia SOSP Melospiza melodia SOSP Melospiza melodia SOSP Whitecrowned Sparrow Whitecrowned Sparrow Whitecrowned Sparrow Song Sparrow Song Sparrow Song Sparrow Junco hyemalis Melospiza georgiana Dark-eyed (Oregon) junco Swamp Sparrow SWSP DEJU WCSP WCSP WCSP Icteridae RWBL Agelaius phoeniceus Red-winged blackbird Emberizidae Emberizidae Emberizidae Emberizidae Emberizidae Emberizidae Emberizidae Icteridae BHCO Brown-headed Molothrus ater cowbird Icteridae Family BHCO Abbreviation Brown-headed Molothrus ater cowbird Scientific name Marler and Peters, 1977 MarlerKreith and Tamura, 1962 Marler and Sherman, 1985 Marler and Sherman, 1983 Kroodsma, 1977 Plamondon et al, 2010 Marler, 1970 Baptista and Petrinovich, 1984 Jarrell, 2009 Marler et al., 1972 Hoepfner and Goller, 2013 West King EastzerStaddon, 1979; Reference ? 8 from four populations, two years 5 2 2 isolate, 2 group housed (multi-species) 6 2 individual isolates, 9 “social-group” isolates 5 6 5 5 4 Sample size I I, G I, G/heteroG G I, G G G 3–10 6–10 G G, heteroG Egg, from canary heteroG at 10 3–10, deafened at G 17–23 egg 5–9 5 6–13 5–7 Taken 6, isolated I 39–46 <10 ? Age at isolation Isolation (DPH) treatment Normal Normal, Isolate Isolate Normal Normal, Isolate Normal, Isolate Normal, Isolate Normal, Isolate Normal, Isolate Normal, Isolate Normal, Isolate Normal, Isolate Utilized spectrograms? W W – W W W W T T W T W Normal:tutored/ wild? (Continues) 6 Species Table 1 Results of the Literature Search and Spectrograms Used in the Analysis 20 Love et al. Turdidae Mimidae Northern Mimus polyglottos NOMO mockingbird Turdus merula Turdidae Cardinalidae Cardinalidae Cardinalidae Fringillidae Estrildidae Emberizidae Emberizidae Family EUBL Eurasian blackbird WOTH Wood thrush Hylocichla mustelina INBU Indigo bunting Passerina cyanea NOCA INBU Cardinalis cardinalis Northern Cardinal CHAF SWSP Indigo bunting Passerina cyanea Fringilla coelebs Taeniopygia guttata ZEFI Zebra finch Chaffinch Melospiza georgiana Swamp Sparrow SWSP Melospiza georgiana Swamp Sparrow Abbreviation Scientific name (Continued) Species Table 1 6 3 isolates, 5 normal 5 6 Sample size Hatch, 1967 Thielcke-Poltz and Thielcke, 1960; Lanyon, 1979 Payne, 1981 Rice and Thompson, 1968 5–8 ~5 0–1 I, G Pair, I I Egg, from canary G then I at 10, group until ~ 30dph, then individually isolated with white noise 3–10, deafened at D 17–23 1 2? 2 2–3 ? egg Normal, Isolate Isolate Normal, Isolate Normal, Isolate Normal, Isolate Isolate Normal Utilized spectrograms? I I Normal, Isolate Normal, Isolate Pair since egg, Normal, white noise at Isolate 3 weeks heteroG until ~ 60 Taken “a few days G, heteroG dph, 4 ph” individuall isolated, 3 group reared to 9 months, then isolated 3 isolated, 2 could 3–7, isolated from I, heteroG hear other each other at species, but not singing onset each other Dittus and Lemon, 1969 3 isolates, more group isolates Thorpe, 1958 Goller, unpublished data; Marler and Sherman, 1985 Marler and Sherman, 1983 Reference Age at isolation Isolation (DPH) treatment T W W W, T – T W, T Wild* – W Normal:tutored/ wild? (Continues) * = aviary raised 21 Variation in Learned Components of Song 7 Developmental Neurobiology 22 Love et al. score close to zero indicates that isolates produce syllables that were not found to be different from those produced by normally reared birds. Since syllables that were not found to have a best match did not have a reported difference in duration, and thus could not be given a residual in the linear regression, above, these syllables were excluded from the syllable morphology analysis. Developmental Neurobiology W Egg I Normal, Isolate Acoustic Analysis: Rhythm Note: Sample size refers to number of isolate individuals raised and reported. 2 Thamnophilidae Touchton et al, 2014 Hylophylax naevioides Spotted antbird SPAN T^ G until ~ 35dph, Normal, then I Isolate 6–8 8 Chaiken and Bohner, 2007 Sturnidae Sturnus vulgaris European starling EUST Scientific name Species Table 1 (Continued) Abbreviation Family Reference Sample size Age at isolation Isolation (DPH) treatment Utilized spectrograms? Normal:tutored/ wild? ^ = some tutored birds were group housed, others individually housed 8 Using the syllable start time and syllable duration from the SAP software output, summary features of each song were computed in R. The rhythmic features that were recorded are: number of syllables, mean syllable length, mean silent period length, song length, syllable rate (number of syllables/song length), syllable regularity, and silent period regularity. Syllable regularity is a normalized measure of consistency of syllable duration. This measure was computed within a single song by first finding the proportion of total sound that each syllable represents (syllable length divided by the sum of all syllable lengths). Then, the proportions were normalized by dividing each proportion by the maximum proportion. Finally, each of these normalized proportions was divided by the total number of syllables in the song. The resulting value yielded the syllable regularity. The same procedure was utilized to produce the silent period regularity, substituting silent period lengths for syllable lengths. For each species, two-tailed t-tests were conducted between normal and isolate songs on each rhythmic feature. The absolute value of the t-statistic served as the metric for similarity, with a higher value signifying greater disparity between isolate and normal songs for a given feature. Principal component analysis (PCA) was conducted (prcomp function in stats package in R) on the t-statistic data to identify major contributory rhythmic features. The first three principal components together explained 90% of the variance in the dataset. To increase interpretability, the t-scores for the three variables that contributed most to the first three principal components (identified as syllable rate, syllable regularity, and silent period regularity using an eigenvalue cutoff of 0.70, see results) were averaged to create a single index for isolate dissimilarity for each species, with a higher value signifying that isolate song rhythm differs from normal in a species, and a lower value signifying that isolate song rhythms are similar to those found in normal songs of the species. Note that multiple testing is not an issue here, as tests were conducted not to search for significant relationships, but simply to serve as a convenient way 23 Variation in Learned Components of Song 9 Table 2 Species Details Species abbreviaton Scientific name Common name BHCO Molothrus ater Brown-headed cowbird CHAF Fringilla coelebs Common chaffinch Proportion of missing syllables Number of Isolate songs Number of normal songs Number of isolate syllables Number of normal syllables 0.314 44 44 181 185 0 19 4 237 39 DEJU Junco hyemalis Dark-eyed junco 0.005 21 12 194 195 EUBL Turdus merula Common blackbird 0.215 12 6 109 93 EUST Sturnus vulgaris European starling 0.303 3 7 12 109 INBU Passerina cyanea Indigo bunting 0.013 13 10 144 160 NOCA Cardinalis cardinalis Northern cardinal 0.125 5 10 39 64 NOMO Mimuspoly glottos Northern mockingbird 0.039 2 3 28 51 RWBL Agelaius phoeniceus Red-winged blackbird 0.132 15 17 72 53 SOSP Melospiza melodia Song sparrow 0 31 11 402 193 SPAN Hylophylax naevioides Spotted antbird 0.038 3 2 43 52 SWSP Melospiza georgiana Swamp sparrow 0 8 3 175 79 WCSP Zonotrichia leucophrys White-crowned sparrow 0.041 11 6 75 49 WOTH Hylocichla mustelina Wood thrush 0.148 2 5 11 27 YHBL Xanthocephalus xanthocephalus Yellow-headed blackbird 0.057 24 20 83 106 ZEFI Taeniopygia guttata Zebra finch 0.058 15 19 83 121 Notes: Proportion of missing syllables is the proportion of normal syllables for which no match was found to be produced by isolates. Sample sizes for all analyses are shown in subsequent columns. to quantify differences between two sets of data and compare those differences between species. Acoustic Analysis: Syntax Ten pairs of songs were randomly chosen from each species, 5 of which contained one isolate song and one normal song, the other 5 of which contained two normal songs. Resampling was kept to a minimum by running the code used to produce the random song pairs several times and using the set of pairs from the run that included the fewest resampling events. After the identity of each song was hidden, a single expert judge (JL) scored syntactical similarity for each pair according to the following metric: if both songs used the same or morphologically similar syllables for the nth syntactical position, one point was awarded for that position. Morphological similarity was determined by the judge and reflected rough visual similarity of the frequency–time trace on the spectrogram, in line with previous studies (e.g., Plamondon et al., 2010). If the nth syllable of one song matched the n + 1th syllable of the other, a penalty zero was given, and scoring continued with an offset. If the nth syntactical position used different syllables in each song, then a point was not awarded for that position. This procedure progressed sequentially through the syntactical positions until the final syllable of the song that contained the higher number of syllables was reached. If the songs had unequal numbers of syllables, a score of zero was given for each surplus syntactical position. The sum of the points was then divided by the total number of Developmental Neurobiology 24 10 Love et al. possible scores, including any penalty zeros, yielding a score between 0 and 1, with higher scores indicating greater syntactical similarity and lower scores indicating greater syntactical divergence. The difference in the means of the normal–normal pairs and the isolate–normal pairs was recorded for each species to use as the summary statistic for syntax, with a higher score indicating that isolates produce syntax different from that of normally reared birds and a lower score indicating that isolates produce similar syntax to that produced by normally reared birds. Tests for Phylogenetic Signal To test for phylogenetic signal, we first built a consensus tree containing all 16 species built from 100 phylogenetic trees created using an online resource (birdtree. org) and using the full Ericson backbone. Then, we tested for phylogenetic signal for each production module by calculating Blomberg’s K (phylosig function from phytools package v0.6-44 in R; Revell, 2012). RESULTS In the literature, we found 15 species for which data were sufficient for complete analysis. This dataset was expanded or supplemented with our own recordings of three species: brown-headed cowbird (Hoepfner and Goller, 2013), zebra finch (FG unpublished data), and yellow-headed blackbird (Jarrell, 2009), resulting in a sample of 16 species, including 15 oscine species and 1 suboscine species (detailed information, including sample sizes used in analyses, in Table 2). As detailed below for each production module, species show substantial variation in how and the degree to which isolate song differs from normal song (Fig. 2). None of the three modules showed evidence of phylogenetic signal, as detailed below. These results suggest the potential for major differences among species in the degree of genetic coding of song. Syllable Morphology Comparisons of corrected maximum correlation reported by cross-correlation of the amplitude envelopes of isolate–normal syllable pairs to normal–normal syllable pairs show interspecific variation, with four species showing significant positive differences: Song sparrow, Melospiza melodia, (two-tailed t-tests; t = 7.25, df = 345.56, P = 2.79e-12), dark-eyed junco, Junco hyemalis, (t = 7.26, df = 341.47, P = 2.63e12), zebra finch (t = 2.0044, df = 146.45, P = 0.047), and brown-headed cowbird (t = 3.21, df = 300.5, P = 0.0015) (Fig. 3; see Table 2 for sample sizes of syllables and songs for each species; statistical independence in our sample varies among species). Species other than the four listed above showed low or negative mean differences. Blomberg’s K for syllable morphology was 0.16, suggesting a lack of phylogenetic signal (P = 0.96). Representative examples of syllable matches and quantitative comparison used in the analysis are shown in Figure 4. Plots of raw acoustic (a) (b) Legend Icteridae Fringillidae Emberizidae Turdidae Sturnidae Cardinalidae Mimidae Suboscine Estrilldidae −1.56 standardized value 3.24 ax nt Sy hy th m SD units R S M ylla or b ph le ol og y Hylophylax naevioides Turdus merula Hylocichla mustelina Mimus polyglottos Sturnus vulgaris Taeniopygia guttata Fringilla coelebs Passerina cyanea Cardinalis cardinalis Melospiza georgiana Melospiza melodia Zonotrichia leucophrys Junco hyemalis Molothrus ater Agelaius phoeniceus Xanthocephalus xanthocephalus Figure 2 Interspecific variation in isolate song disparity. (A) Phylogeny with character heatmap. Data were scaled and centered by column. Darker shading indicates greater difference between isolate and normal songs for a given feature and species. Abundant interspecific variation with no phylogenetic signal was observed for each of three production modules. (B) Color key for families. Developmental Neurobiology 25 Variation in Learned Components of Song 11 0.15 mean difference 0.10 0.05 0.00 −0.05 −0.10 Agelaius phoeniceus Xanthocephalus xanthocephalus Hylocichla mustelina Hylophylax naevioides Fringilla coelebs Melospiza georgiana Passerina cyanea Cardinalis cardinalis Turdus merula Zonotrichia leucophrys Sturnus vulgaris Mimus polyglottos Molothrus ater Taeniopygia guttata Junco hyemalis Melospiza melodia −0.15 Figure 3 Syllable Acoustic Morphology. Mean differences between normal-to-normal and isolate-to-normal syllable similarities with 95% CI. A positive score indicates that isolate syllables are, on average, different from normal syllables, whereas a negative score indicates that isolate syllables are found to match normal syllables better than normal syllables match other normal syllables. A score close to zero indicates that isolates produce syllables that were not found to be different from those produced by normally reared birds. Colors as in Figure 2b. Spotted Antbird 12 10 10 (a) 8 (b) 6 4 2 Frequency (kHz) Frequency (kHz) Indigo Bunting 12 8 6 (c) (d) 4 2 0 0 0.5 0 Time (s) 0 0.1 Time (s) Figure 4 Example syllable comparisons with scores. (A) an isolate Indigo bunting syllable (B) a normal Indigo bunting syllable. Maximum cross-correlation score between a and b is 0.31. Duration difference-corrected residual score is −0.092 (C) an isolate spotted antbird syllable (D) a normal spotted antbird syllable. Maximum cross-correlation between c and d is 0.72. Duration difference-corrected residual score is −0.173. The syllables pictured here include silent periods before and after sound for display only. Actual analysis was conducted using exact start and end of syllables. These four syllables are included in one .wav file in the ESM in a, b, c, d order. Developmental Neurobiology 26 12 Love et al. measurement data for isolate versus normally reared birds for all species are included in the supplemental information. Syllable morphology scores can be positive or negative. While positive scores indicate clearly that isolate syllable morphology is, on average, different from normal, the interpretation of negative scores is less intuitive. A brief description of how these scores may be interpreted follows: One possible outcome of vocal learning is the opportunity to increase individual identity by differentiating one individual’s syllables from those of other individuals (Nelson and Poesel, 2007). Since isolates do not have this opportunity, it is possible that, in species where this process occurs, isolate syllables would be expressed as a “default” syllable, with few individually identifying features, and would thus share some features with most normal syllables, while normal syllables would show higher diversification among themselves, which would produce the pattern of negative syllable morphology scores observed in some species in this study. Three species showed significant negative mean differences: Redwinged blackbird (t = −2.1, df = 101.45, P = 0.038), swamp sparrow (t = −2.20, df = 202.63, P = 0.029), and yellow-headed blackbird (t = −2.35, df = 156.69, P = 0.020) (Fig. 3; see Table 2). Rhythm Quantitative measures of rhythmic difference between songs of isolation-reared and normally reared birds also show interspecific variation. In a PCA, syllable rate, silence regularity, and sound regularity emerged as the major variables explaining variation between species (first three principal components together explain 90% of the variation between species, top contributory variables were identified with a 0.70 eigenvalue cutoff; see Methods). The first principal component primarily distinguished between song sparrow, Melospiza melodia, and all other species, while PC2 and PC3 explained variation between all species (Fig. 5b and Table 3). To increase Syllable Rate Silence Regularity Sound Regularity Rhythm Score 1.5 2.0 8 PC3 4 0.0 0.5 1.0 6 Melospiza melodia −2.0 −1.5 −1.0 −0.5 Rhythmic Difference Score 10 We modified the analysis framework to find the proportion of normal syllables for which no match was found to be produced by isolates. The results showed that the proportion of missing syllables in the pool of isolate syllables did not have a significant impact on overall syllable morphology scores (linear regression adjusted R-squared = −0.07, P = 0.91). In addition, proportion of missing syllables does not show evidence of a phylogenetic signal (Blomberg’s K = 0.47, P = 0.21). Incidence of missing syllables is reported in Table 2. 2 Mimus polyglottos Turdus merula Cardinalis cardinalis Agelaius phoeniceus Hylophylax naevioides Junco hyemalis Hylocichla mustelina Zonotrichia leucophrys Passerina cyanea Xanthocephalus xanthocephalus Taeniopygia guttata Melospiza georgiana Molothrus ater Fringilla coelebs Sturnus vulgaris Melospiza melodia 0 −10 −12 −4 −3 −2 −1 0 1 2 −8 −6 3 PC2 Figure 5 Results of the Rhythm analysis. (A) t-statistics from syllable rate, syllable regularity, and silent period regularity were averaged together to produce the overall rhythmic difference score, presented by the bold dots, with 95% CI. A higher score indicates a greater difference between isolate- and normal-produced song rhythms. (B) xyz plot showing the first three principal components of the rhythm PCA. PC1 primarily differentiates between song sparrow (Melospiza melodia) and all other species, while PC2 and 3 explain variation among all species. PCA details in Table 3. Colors as in Figure 2b. Developmental Neurobiology −4 −2 0 1 PC 2 4 27 Variation in Learned Components of Song 13 Table 3 Rhythm Principal Component Analysis Details PC1 PC2 PC3 PC4 PC5 PC6 Mean syllable length −0.51 0.23 −0.06 −0.15 0.40 0.71 Mean silence −0.32 −0.41 0.01 0.56 −0.55 0.34 Song length −0.31 −0.34 0.57 −0.63 −0.24 −0.06 Syllable rate −0.72 0.34 −0.08 0.16 −0.03 −0.58 Rhythmic silence regularity −0.17 −0.74 −0.29 0.03 0.55 −0.20 Rhythmic sound regularity −0.05 −0.03 −0.76 −0.49 −0.42 0.04 Proportion of variance 0.69 0.14 0.07 0.06 0.03 0.01 Cumulative proportion 0.69 0.83 0.90 0.96 0.99 1.00 Notes: PCA was conducted on t-values from t-tests between normal and isolate songs for each rhythmic feature. Explained cumulative variance reached 90% once three PCs were included, and a cutoff of 0.70 was used for the individual eigenvalues for each PC. The first PC primarily differentiates between song sparrow, Melospiza melodia, and all other species. interpretability relative to the use of PCA scores, the absolute value of the t-scores for these three components for each species were averaged, creating a single index score that relates to the difference in rhythm between isolate and normal song for each species (Fig. 5a). Blomberg’s K for rhythm was 0.16, suggesting a lack of phylogenetic signal (P = 0.91). In addition, no phylogenetic signal was detected for the individual measures of rhythm (syllable rate, K = 0.12, P = 0.95; silence regularity, K = 0.20, P = 0.89; sound regularity, K = 0.30, P = 0.63). The spotted antbird, Hylophylax naevioides, showed strong differences in rhythm between songs produced by isolate-reared and the two examples of normally reared birds. This result was surprising, since this species is not a vocal learner and since nonvocal learning species are not expected to have variable song. Therefore, we wanted to be sure that the results in this case were not due to an artifact. Because the small sample of normal songs (n = 2) might not represent the naturally found variation in components of the rhythm, we used two additional recordings from a public database (XC363516, XC377884; xeno-canto. org) to supplement the data on normally reared song. After including these additional songs, the previously observed difference was no longer present, suggesting that low sample size generated an artifactual difference. The results of the rhythm analysis, but not the syllable morphology or syntax analyses, shown here were conducted with this expanded sample of normal songs for the spotted antbird. Low sample size could affect the results in a similar way for two other species: Northern mockingbird and swamp sparrow, both of which have a sample of three normal songs. However, we have no a priori reason to doubt our findings from these species—song features are expected to be variable in species that have learned song. In the interest of comparing the components of rhythm in detail for species that have similar song organization, we chose to conduct additional comparative analyses on the rhythms of swamp sparrows and dark-eyed juncos. While the two species had similar rhythm scores, they arrived at those scores through different trajectories. In dark-eyed junco, scores for sound and silence regularity were low, but the score for syllable rate was high, while in swamp sparrow scores for syllable rate were moderately low, but scores for sound and silence regularity were high (Fig. 5a). Also, isolate swamp sparrows utilize a rhythmic consistency not statistically different from that of isolate dark-eyed juncos (Student’s t-tests: P = 0.324 and P = 0.793 for sound and silence, respectively), while normal swamp sparrow’s rhythms are statistically different from normal dark-eyed juncos (Student’s t-tests: P = 0.002 and P = 0.004 for sound and silence regularity, respectively) (Fig. 6). Syntax Comparisons of syntactical patterning show interspecific variation, with isolates of some species producing syntax similar to that found in the songs of normally reared birds and isolates of other species producing syntax different from that found in the songs of normally reared birds (Fig. 7). Scores are distributed fairly continuously across the observed range (Fig. 7). Of the three components analyzed (syllable morphology, rhythm, and syntax), syntax showed the highest level of deviation in isolate song when compared to normal song. The major exception is the European starling, Sturnus vulgaris, where normally reared birds sing songs with complex syntax which only becomes apparent over a long timescale, but did not emerge in the current analysis (Eens et al., Developmental Neurobiology 28 14 Love et al. (a) (b) DEJU DEJU SWSP SWSP 1.0 rhythmic silence regularity rhythmic sound regularity 0.8 0.8 0.6 0.6 0.4 0.4 I N N I I N I N Figure 6 Rhythms of trilled song elements. Isolate dark-eyed juncos and swamp sparrows use similar rhythms (Student's t-tests: P = 0.324 and P = 0.793 for sound and silence, respectively), while normal individuals of the two species use different rhythms (Student's t-tests: P = 0.002 and P = 0.004 for sound and silence regularity, respectively). (A) Syllable length regularity. (B) Silent period length regularity. Mean Difference 1.0 0.5 0.0 Sturnus vulgar is Turdus merula Hylophylax naevioides Fringilla coelebs Mimus polyglottos Cardinalis cardinalis Melospiza georgiana Junco hyemalis Melospiza melodia Xanthocephalus xanthocephalus Passerina cyanea Hylocichla mustelina Taeniopygia guttata Agelaius phoeniceus Molothrus ater Zonotrichia leucophrys −0.5 Figure 7 Syntax. Difference in the means between syntactical similarity scores for isolate–normal and normal–normal song pairs with 95% CI. Colors as in Figure 2b. 1988). Our comparison shows positive values for the difference in the mean between normal–normal song pairs and isolate–normal song pairs. Thus, in general Developmental Neurobiology it was unlikely that our analysis revealed strong differences in species with variable or complex syntax (e.g., European starling, Northern cardinal Cardinalis 29 Variation in Learned Components of Song cardinalis, Northern mockingbird Mimuspoly glottos), as the analysis regime was unlikely to capture higher order or variable syntax that may be evident on longer timescales than were available for inclusion in this analysis (see Berwick et al., 2011). Still, it is notable that species which received lower scores in the analysis had either very large or very small syllable repertoires, while species which received higher scores mostly sing stereotyped songs composed of a brief sequence of well-defined syllables. Blomberg’s K for syntax was 0.60, again suggesting a lack of phylogenetic signal (P = 0.081). Effects of Rearing Condition Since non-adult social interaction (West and King, 1988; Fehér et al., 2009) and pre-hatching auditory experience (Katsis et al., 2018) have been identified to play a role in avian vocal development, the effects of rearing condition on scores were analyzed. No significant effect of isolate rearing condition was found for syllable morphology (LM, highest f = 2.9 for heterospecific group), rhythm (highest f = 2.5 again for heterospecific group), or syntax (highest f = 1.2 for pair in masking noise). DISCUSSION Our data reveal differences in the degree to which different oscine species rely on acquired information for developing species-specific song. This may come to many as no surprise, since qualitative comparison of spectrographic evidence from isolate and normally raised birds suggests this differential reliance on learned features and that this has been noted since early isolate studies (Thorpe, 1964). Despite this, authors of studies that discuss or even directly investigate oscine vocal learning may be limited by accepting the common narrow view of vocal learning in this clade, which can confuse conclusions about observed patterns in the evolution and use of vocal learning and song in birds. Our analysis is, to our knowledge, the first attempt at quantifying the differences in learning of production modules of song across species, and should serve as an impetus to refine the investigation of vocal learning in birds going forward. The results show that species vary markedly in how much information about song features needs to be learned, and it is not apparent that related species share the degree to which they have to acquire features. No phylogenetic signal was found for any of the three modules. The members of individual families did not cluster, and scores for individual species were distributed across 15 the range found in all species. While it is possible that closely related species may cluster if a larger dataset was available, we find no current evidence to suggest that this would be the case. More importantly, we assessed the respective roles of genetically encoded and learned information for the development of distinct modules of vocal production, which correspond to different song features. It is possible that in each production module distinct interactions exist between learned and genetically encoded features, and different species may vary in the degree that information must be acquired for each module. The rationale for the distinction of production modules is derived from peripheral and central song production mechanisms. The main acoustic features such as sound frequency, amplitude, and harmonic content are determined by our first production module, syringeal fine control (e.g., Sitt et al., 2008; Riede and Goller, 2014; Srivastava et al., 2015; Mencio et al., 2017), which originates in the forebrain motor production circuitry, which projects to the motor neuron pool in the hypoglossal nucleus, nXII. The main song rhythm is predominately determined by the second module, respiratory patterns of song (e.g., Suthers and Goller, 1997; Suthers and Zollinger, 2008), and the instructions from the forebrain motor circuitry (HVC, proper name, and robust arcopallial nucleus, RA) must be integrated into the pacemaking circuitry for respiration (RAM/PAM; e.g., Schmidt and Goller, 2016) and use its potential for generating various rhythmic patterns (e.g., Trevisan et al., 2006; Alonso et al., 2009) as well as accommodate limitations from limited available volume and gas exchange needs (e.g., Franz and Goller, 2003; Riede and Goller, 2014). Finally, our third module, the sequence of syllables, is mainly determined by central motor programming, which may take place in HVC and other connected areas in the song circuitry. The anterior forebrain pathway (area X, the thalamic dorsal lateral nucleus of the medial thalamus, DLM, and lateral magnocellular nucleus of anterior nidopallium, LMAN) plays an important role in vocal plasticity and is therefore a major part of vocal learning processes (e.g., Brainard, 2004; Aronov et al., 2008). Our data indicate strongly that the genetically encoded information for the three production modules provides young birds of different species with differently detailed instructions for the development of species-specific song. It remains unclear whether song development follows distinct separate pathways as outlined with these production modules, but some experimental evidence suggests that this may be so. Developmental Neurobiology 30 16 Love et al. For example, the genetically encoded syntax rules of white-crowned sparrows (i.e., the whistle is the first syllable; (Soha and Marler, 2001) can be overwritten by tutoring information, while the acoustic features of syllables are not entirely dependent on full syntax information (Rose et al., 2004; Plamondon et al., 2008). In this species, different parts of the song may provide different information, such as species, regional, and individual identity (Nelson and Poesel, 2007; Nelson, 2017), which is also consistent with the notion of the proposed modules. Similarly, research in zebra finches and in song and swamp sparrows suggests that syntax and syllable morphology are learned separately from one another and prioritized differently (Marler and Peters, 1977; Lipkind et al., 2017). Other recent research on the neural control of song rhythm strongly suggests that this aspect of song is controlled by a distinct production module that requires different neural input than that required for the control of other aspects of vocal production (Ali et al., 2013) and that rhythm is strongly innately controlled (Araki et al, 2016). Our study uses a comparative approach to arrive at similar conclusions (see Fig. 5). Whereas the observed differences across the three different production modules clearly illustrate that the degree to which genetically encoded information guides song development varies markedly, the specific values must be interpreted with caution in light of the different song organization characteristics. The species in the dataset encompass substantial variation in how song is organized, ranging from species with a single multisyllable song to species with very large repertoires. Additionally, other species sing songs composed of trills (individual syllables are repeated multiple times at often high rates) with different repertoire sizes of trill types. The dataset is too small to allow the exploration of whether song organization may explain different learning strategies, but comparison of species pairs with similar song organization suggests complex combinations of how genetically encoded and acquired information may guide song development. For example, the dark-eyed junco showed a moderately low rhythm index score (= 1.35), for its songs composed of fairly stereotyped trills, which is consistent with a respiratory rhythm that emerges easily without much input from song models. In comparison, the closely related swamp sparrow also sings songs composed of relatively simple trills but showed a higher difference in rhythm between isolate and normal songs (rhythm index score = 1.83). Interestingly, these two closely related species arrive at comparable rhythm index scores by different trajectories (see Results). The two species appear to rely on imitation Developmental Neurobiology for the normal development of different aspects of song rhythm. These two species are in the same family but different genera, and there are many species with different song organization (i.e., songs not composed of trills) that are more closely related to each of these two species (Carson and Spicer, 2003). Still, given the limited taxonomic coverage of isolate studies, this remains an intriguing result that supports investigation of phylogenetic trends in innate song-associated respiratory patterns. This finding supports the possibility that the innate, genetically programmed respiratory pattern for song production is highly regular in some species and may be expressed as such in adult song of those species but is nonetheless subject to modification through learning. These findings also inform our views on how the evolution of vocal learning may have affected speciation. Vocal learning has been identified as a mechanism through which reproductive isolation may be achieved earlier in the process of speciation, potentially encouraging the high diversification rates observed in the oscines (Lachlan and Servedio, 2004), though there has been some discussion about this purported role (Baptista and Trail, 1992). Our results highlight the importance of genetically encoded guidance of vocal development in the oscines. It has been recognized that a system of purely cultural evolution of song is difficult to rectify with the patterns of song observed. An isolate-founded population of zebra finches, relying on genetically guided song development, nevertheless displayed song features approaching those of normally reared birds within a few generations (Fehér et al., 2009). While supporting the idea that song arises naturally through ingrained processes, this study indicates that genetically guided social influences during song development play an important role in shaping the final crystallized song, and that imitative vocal learning as it has classically been depicted (e.g., Konishi, 1965; Marler, 1970; Marler and Sherman, 1983; Marler, 1997) is not necessarily the primary means of effecting change in song traits over time in all oscine species. This influence needs to be considered in our comparative data set as well, because species in our dataset are likely to vary in the importance of social influences, and the rearing conditions also varied in different isolation studies (see Table 1 and results). Our study compliments this and other previous work by highlighting the importance of genetically guided evolution of song in vocal-learning species. From this conceptual vantage point, we can begin to more clearly investigate natural phenomena that may shed light on the evolutionary emergence of vocal learning. The sister group to the vocal-learning 31 Variation in Learned Components of Song oscines is the suboscine clade. Although vocal learning may be present in the Cotingidae (Saranathan et al., 2007; Kroodsma et al., 2013), most families of the suboscines are not considered to be vocal learners. A true isolation study has been conducted in only one suboscine species: the spotted antbird (Touchton et al., 2014). This species does not show signs of vocal learning, similar to other suboscines that have been studied using alternate techniques (Kroodsma, 1984; Kroodsma and Konishi, 1991; Liu et al., 2013). The antbird data therefore set a lower limit against which scores of oscines can be compared. The degree to which scores of oscine species differ from those of the antbird indicates reliance on acquired information. Clearly, we need more studies of suboscines to be able to perform better comparative analyses and, thus, gain better insight into the evolution of vocal learning and its role in the speciation process. If the evolution of song occurred primarily via cultural mechanisms, we would expect to find evidence of closely related species sharing major aspects of the vocal learning program (i.e., that closely related species require similar levels of learning in each of the three components analyzed in this study). To the contrary, we show here, even among closely related species, markedly different specific features of the vocal learning program. Indeed, the possibility that continuous variation exists in the genetic control of production-based song features suggests that a model of disparate vocal learning strategies (e.g., Beecher and Brenowitz, 2005; Brenowitz and Beecher, 2005) may not be supported. Instead, future investigation may provide evidence that most species in fact can be seen to utilize a mixture of these strategies to develop their song: primarily improvising syllables but producing them with an innate rhythm and external model imitated syntax, for example. A productive research program would employ a taxonomically diverse, production-based approach to focus on identifying genetically encoded rules for song development and on characterizing the interplay between the genetic basis for song and the mechanisms of vocal learning. By embarking on such a research program, one would be investigating the evolution of behavioral and phenotypic plasticity for a complex trait. Song is subject to sexual and natural selective pressures that have seemingly alternately favored more or less plasticity according to yet-unknown variables. Recent studies focusing on the evolutionary context of learned and unlearned avian vocal behavior show that progress in understanding the evolutionary forces that control plasticity in vocal behavior is imminent (Seddon, 2005; Weir and Wheatcroft, 2011; Medina-García 17 et al., 2015; Mason et al., 2017), and our study joins these in promoting a broad, comparative approach. In addition, future studies that involve oscine song may benefit by taking into account the evidence presented here. We show that some song traits are more strongly genetically guided than others and that this phenomenon varies in its expression by species, thus it may be possible to conduct more direct inquiries regarding such various topics as song trait change across hybrid zones, species/subspecies/population delineation based on song traits, and neural control of song production mechanisms. It may also be worthwhile to revisit older studies, as we do here, and “mine” them using a production-based acoustic analysis for potentially elucidative data. The novel method of spectrogram-to-sound conversion presented here may be of some utility to such endeavors. This work would not be possible without the hard work of the many authors of the studies used in this analysis, seen in table 1. Members of R Club and Theory Lunch at the University of Utah provided important and thoughtful discussion that benefitted this study. We also thank Kent Livezey and Bruce Lagerquist for contributing recordings to xeno-canto.org, licensed under CC-BY-NC-SA 4.0. Two anonymous reviewers provided comments that greatly improved this manuscript. JL was supported by a George R Riser Research Award. AUTHORS’ CONTRIBUTIONS JL and FG conceived and designed the study. JL conducted the literature search and the analysis. AH and FG provided recordings. JL, AH, and FG drafted and edited the manuscript. ETHICS Research on live animals directed by the authors was carried out under the approval of the University of Utah IACUC. DATA AVAILABILITY STATEMENT The data that support the findings of this study are available from the corresponding author upon reasonable request. CONFLICT OF INTEREST The authors report no conflict of interest. Developmental Neurobiology 32 18 Love et al. LITERATURE CITED Albin, A.L. (2014) PraatR: an architecture for controlling the phonetics software “Praat” with the R programming language. Journal of the Acoustical Society of America, 135(4), 2198–2199. https://doi.org/10.1121/1. 4877175. Ali, F., Otchy, T.M., Pehlevan, C., Fantana, A.L., Burak, Y. and Ölveczky, B.P. (2013) The basal ganglia is necessary for learning spectral, but not temporal, features of birdsong. Neuron, 80, 494–506. Alonso, L.M., Alliende, J.A., Goller, F. and Mindlin, G.B. (2009) Low-dimensional dynamical model for the diversity of pressure patterns used in canary song. Physical Review E, 79(4), 1–8. https://doi.org/10.1103/PhysR evE.79.041929. Araki, M., Bandi, M.M. and Yazaki-Sugiyama, Y. (2016) Mind the gap: neural coding of species identity in birdsong prosody. Science, 354(6317), 1282–1287. Aronov, D., Andalman, A.S. and Fee, M.S. (2008) A specialized forebrain circuit for vocal babbling in the juvenile songbird. Science, 320(5876), 630–634. https://doi. org/10.1126/science.1155140. Baptista, L.F. and Petrinovich, L. (1984) Social interaction, sensitive phases and the song template hypothesis in the whitecrowned sparrow. Animal Behaviour, 32(1), 172–181. Baptista, L.F. and Trail, P.W. (1992) The role of song in the evolution of passerine diversity. Systematic Biology, 41(2), 242–247. Beckers, G.J. (2013) Peripheral mechanisms of vocalization in birds: a comparison with human speech. In: Bolhuis, J.J. and Everaert, M. (Eds.) Birdsong, Speech, and Language: Exploring the Evolution of Mind and Brain. MIT Press, pp. 399–422. Beecher, M.D. (2017) Birdsong learning as a social process. Animal Behavior, 124, 233–246. https://doi.org/10.1016/j. anbehav.2016.09.001. Beecher, M.D. and Brenowitz, E.A. (2005) Functional aspects of song learning in songbirds. Trends in Ecology & Evolution, 20(3), 143–9. https://doi.org/10.1016/j. tree.2005.01.004. Berwick, R.C., Okanoya, K., Beckers, G.J. and Bolhuis, J.J. (2011) Songs to syntax: the linguistics of birdsong. Trends in Cognitive Sciences, 15(3), 113–121. https://doi. org/10.1016/j.tics.2011.01.002. Bolhuis, J.J., Okanoya, K. and Scharff, C. (2010) Twitter evolution: converging mechanisms in birdsong and human speech. Nature Reviews Neuroscience, 11(11), 747–759. https://doi.org/10.1038/nrn2931. Brainard, M.S. (2004) Contributions of the anterior forebrain pathway to vocal plasticity. Annals of the New York Academy of Sciences, 1016, 377–394. https://doi. org/10.1196/annals.1298.042. Brenowitz, E.A. and Beecher, M.D. (2005) Song learning in birds: diversity and plasticity, opportunities and challenges. Trends in Neurosciences, 28(3), 127–132. https:// doi.org/10.1016/j.tins.2005.01.004. Developmental Neurobiology Carson, R.J. and Spicer, G.S. (2003) A phylogenetic analysis of the emberizid sparrows based on three mitochondrial genes. Molecular Phylogenetics and Evolution, 29(1), 43–57. https://doi.org/10.1016/ S1055-7903(03)00110-6. Caruso-Peck, S. and Goldstein, M.H. (2019) Female social feedback reveals non-imitative mechanisms of vocal learning in zebra finches. Current Biology, 29, 1–6. Chaiken, M.L. and Böhner, J. (2007) Song learning after isolation in the open-ended learner the European starling: dissociation of imitation and syntactic development. The Condor, 109(4), 968. Dittus, W.P.J. and Lemon, R.E. (1969) Effects of song tutoring and acoustic isolation on the song repertoires of cardinals. Animal Behaviour, 17523–533. Eens, M., Pinxten, R. and Verheyen, R.F. (1988) Temporal and sequential organisation of song bouts in the starling. Ardea, 77(August), 75–86. Fehér, O., Wang, H., Saar, S., Mitra, P.P. and Tchernichovski, O. (2009) De novo establishment of wild-type song culture in the zebra finch. Nature, 459(7246), 564–568. https:// doi.org/10.1038/nature07994. Franz, M. and Goller, F. (2003) Respiratory patterns and oxygen consumption in singing zebra finches. Journal of Experimental Biology, 206(6), 967–978. https://doi. org/10.1242/jeb.00196. Hatch, J.J. (1967) Diversity of the song of mockingbirds (Mimus Polyglottos) reared in different auditory environments. PhD dissertation, Duke University. Hoepfner, A.R. and Goller, F. (2013) Atypical song reveals spontaneously developing coordination between multimodal signals in brown-headed cowbirds (Molothrus ater). PLoS One, 8, 1–7. https://doi.org/10.1371/journ al.pone.0065525. Jarrell, E. (2009) Development of multimodal song displays and use of song in the yellow-headed blackbird (Xanthocephalus xanthocephalus). Master's thesis, University of Utah. Katsis, A.C., Davies, M.H., Buchanan, K.L., Kleindorfer, S., Hauber, M.E. and Mariette, M.M. (2018) Prenatal exposure to incubation calls affects song learning in the zebra finch. Scientific Reports, 8, 15232. https://doi.org/10.1038/ s41598-018-33301-5. Konishi, M. (1965) Effects of deafening on song development in American robins and black-headed grosbeaks. Ethology, 22(5), 584–599. Kroodsma, D.E. (1977) Correlates of song organization among North American wrens. The American Naturalist, 111(981), 995–1008. Kroodsma, D.E. (1984) Songs of the alder flycatcher (Empidonax alnorum) and willow flycatcher (Empidonax traillii) are innate. Auk, 101(1), 13–24. Kroodsma, D.E. and Konishi, M. (1991) A suboscine bird (eastern phoebe, Sayornis phoebe) develops normal song without auditory feedback. Animal Behavior, 42, 477–487. https://doi.org/10.1016/S0003-3472(05)80047-8. Kroodsma, D.E., Houlihan, P.W., Fallon, P. and Wells, J. (1997) Song development by grey catbirds. Animal 33 Variation in Learned Components of Song Behavior, 54(2), 457–64. https://doi.org/10.1006/ anbe.1996.0387. Kroodsma, D., Hamilton, D., Sánchez, J.E., Byers, B.E., Fandiño-Mariño, H., Stemple, D.W., et al. (2013) Behavioral evidence for song learning in the suboscine bellbirds (Procnias spp.; Cotingidae). Wilson Journal of Ornithology, 125(1), 1–14. https://doi. org/10.1676/12-033.1. Lachlan, R.F. and Servedio, M.R. (2004) Song learning accelerates allopatric speciation. Evolution, 58(9), 2049–2063. Lanyon, W. (1979) Development of song in the wood thrush (Hylocichla mustelina), with notes on a technique for hand-rearing passerines from the egg. American Museum of Natural History. Leitner, S., Nicholson, J., Leisler, B., DeVoogd, T.J. and Catchpole, C.K. (2002) Song and the song control pathway in the brain can develop independently of exposure to song in the sedge warbler. Proceedings of the Royal Society B-Biological Sciences, 269(1509), 2519–2524. https://doi.org/10.1098/rspb.2002.2172. Lipkind, D., Zai, A.T., Hanuschkin, A., Marcus, G.F., Tchernichovski, O. and Hahnloser, R.H.R. (2017) Songbirds work around computational complexity by learning song vocabulary independently of sequence. Nature Communications, 8(1), 1247. https://doi. org/10.1038/s41467-017-01436-0. Liu, W., Wada, K., Jarvis, E.D. and Nottebohm, F. (2013) Rudimentary substrates for vocal learning in a suboscine. Nature Communications, 4(May), 2082. https://doi. org/10.1038/ncomms3082. Marler, P. (1970) A comparative approach to vocal learning: song development in white-crowned sparrows. Journal of Comparative and Physiological Psychology, 71(2), 1. Marler, P. (1997) Three models of song learning: evidence from behavior. Journal of Neurobiology, 33(5), 501–516. Marler, P., Kreith, M. and Tamura, M. (1962) Song development in hand-raised oregon juncos. The Auk, 79(1), 12–30. Marler, P. and Peters, S. (1977) Selective vocal learning in a sparrow. Science, 198(4316), 519–521. Marler, P. and Sherman, V. (1983) Song structure without auditory feedback: emendations of the auditory template hypothesis. Journal of Neuroscience, 3(3), 517–531. Marler, P. and Sherman, V. (1985) Innate differences in singing behaviour of sparrows reared in isolation from adult conspecific song. Animal Behaviour, 33(1), 57–71. Marler, P., Mundinger, P., Waser, M.S. and Lutjen, A. (1972) Effects of acoustical stimulation and deprivation on song development in red-winged blackbirds (Agelaius phoeniceus). Animal Behaviour, 20(3), 586–606. Mason, N.A., Burns, K.J., Tobias, J.A., Claramunt, S., Seddon, N. and Derryberry, E.P. (2017) Song evolution, speciation, and vocal learning in passerine birds. Evolution, 71(3), 786–796. https://doi.org/10.1111/evo.13159. Medina-García, A., Araya-Salas, M. and Wright, T.F. (2015) Does vocal learning accelerate acoustic diversification? Evolution of contact calls in Neotropical parrots. Journal 19 of Evolutionary Biology, 28(10), 1782–1792. https://doi. org/10.1111/jeb.12694. Mencio, C., Kuberan, B. and Goller, F. (2017) Contributions of rapid neuromuscular transmission to the fine control of acoustic parameters of birdsong. Journal of Neurophysiology, 117(2), 637–645. https://doi. org/10.1152/jn.00843.2015. Nelson, D.A. (2017) Geographical variation in song phrases differs with their function in white-crowned sparrow song. Animal Behavior, 124, 263–271. https://doi.org/10.1016/j. anbehav.2016.05.016. Nelson, D.A. and Poesel, A. (2007) Segregation of information in a complex acoustic signal: individual and dialect identity in white-crowned sparrow song. Animal Behavior, 74(4), 1073–1084. https://doi.org/10.1016/j. anbehav.2007.01.018. Nick, T. (2015) Models of vocal learning in the songbird: historical frameworks and the stabilizing critic. Developmental Neurobiology, 75(10), 1091–1113.https:// doi.org/10.1002/dneu.22189. Payne, R.B. (1981) Song learning and social interaction in indigo buntings. Animal Behaviour, 29(3), 688–697. Plamondon, S.L., Goller, F. and Rose, G.J. (2008) Tutor model syntax influences the syntactical and phonological structure of crystallized songs of white-crowned sparrows. Animal Behavior, 76(6), 1815–1827. https://doi. org/10.1016/j.anbehav.2008.07.029. Plamondon, S.L., Rose, G.J. and Goller, F. (2010) Roles of syntax information in directing song development in white-crowned sparrows (Zonotrichia leucophrys). Journal of Comparative Psychology, 124(2), 117–132. https://doi.org/10.1037/a0017229. R Core Team. (2017) R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Available at: https://www.R-proje ct.org/. Revell, L.J. (2012) phytools: An R package for phylogenetic comparative biology (and other things). Methods in Ecology and Evolution, 3, 217–223. Riede, T. and Goller, F. (2014) Morphological basis for the evolution of acoustic diversity in oscine songbirds. Proceedings of the Royal Society B, 281 (1779), 20132306. Rice, J.O. and Thompson, W.L. (1968) Song development in the indigo bunting. Animal Behaviour, 16(4), 462–469. Rose, G.J., Goller, F., Gritton, H.J., Plamondon, S.L., Baugh, A.T. and Cooper, B.G. (2004) Species-typical songs in white-crowned sparrows tutored with only phrase pairs. Nature, 432(December), 750–753. https://doi. org/10.1038/nature03073.1. Saranathan, V., Hamilton, D., Powell, G.V.N., Kroodsma, D.E. and Prum, R.O. (2007) Genetic evidence supports song learning in the three-wattled bellbird Procnias tricarunculata (Cotingidae). Molecular Ecology, 16(17), 3689–702. https://doi.org/10.1111/j.1365-294X.2007.03415.x. Schmidt, M.F. and Goller, F. (2016) Breathtaking songs: coordinating the neural circuits for breathing and singing. Physiology, 31(6), 442–451. https://doi.org/10.1152/physi ol.00004.2016. Developmental Neurobiology 34 20 Love et al. Seddon, N. (2005) Ecological adaptation and species recognition drives vocal evolution in neotropical suboscine birds. Evolution, 59(1), 200–215. https://doi. org/10.1111/j.0014-3820.2005.tb00906.x. Sitt, J.D., Amador, A., Goller, F. and Mindlin, G.B. (2008) Dynamical origin of spectrally rich vocalizations in birdsong. Physical Review E, 78(1), 011905. https://doi. org/10.1103/PhysRevE.78.011905. Soha, J. (2017) The auditory template hypothesis: a review and comparative perspective. Animal Behavior, 124, 247– 254. https://doi.org/10.1016/j.anbehav.2016.09.016. Soha, J. and Marler, P. (2001) Vocal syntax development in the white-crowned sparrow (Zonotrichia leucophrys). Journal of Comparative Psychology, 115(2), 172–180. https://doi.org/10.1037//0735-7036.115.2.172. Srivastava, K.H., Elemans, C.P.H. and Sober, S.J. (2015) Multifunctional and context-dependent control of vocal acoustics by individual muscles. Journal of Neuroscience, 35(42), 14183–14194. https://doi.org/10.1523/JNEUR OSCI.3610-14.2015. Sueur, J., Aubin, T. and Simonis, C. (2008) Seewave: a free modular tool for sound analysis and synthesis. Bioacoustics, 18, 213–226. Suthers, R.A. and Goller, F. (1997) Motor correlates of vocal diversity in songbirds. In: Nolan, V., Ketterson, E.D.and Thompson, C.F. (Eds.) Current Ornithology. Boston, MA: Springer US, pp. 235–288. Suthers, R.A. and Zollinger, S.A. (2008) From brain to song: the vocal organ and vocal tract. Neurosci, birdsong.: 78–98. Suthers, R., Goller, F. and Pytte, C. (1999) The neuromuscular control of birdsong. Philosophical Transactions of the Royal Society B: Biological Sciences, 354(1385), 927– 939. https://doi.org/10.1098/rstb.1999.0444. Tchernichovski, O., Nottebohm, F., Ho, C.E., Pesaran, B. and Mitra, PP. (2000) A procedure for an automated Developmental Neurobiology measurement of song similarity. Animal Behavior, 59(6), 1167–1176. https://doi.org/10.1006/anbe.1999.1416. Thielcke-Poltz, H. and Thielcke, G. (1960) Akustisches Lernen verschieden alter schallisolierter Amseln (Turdus merula L.) und die Entwicklung erlernter Motive ohne und mit künstlichem Einfluß von Testosteron. Zeitschrift für Tierpsychologie, 17(2), 211–244. Thorpe, W.H. (1958) The learning of song patterns by birds, with especial reference to the song of the chaffinch Fringilla coelebs. Ibis, 100(4), 535–570. Thorpe, W.H. (1964) The isolate song of two species of Emberiza. Ibis (Lond 1859), 106(1):115–118. Touchton, J.M., Seddon, N. and Tobias, J.A. (2014) Captive rearing experiments confirm song development without learning in a tracheophone suboscine bird. PLoS One, 9(4), e95746. https://doi.org/10.1371/journal.pone.0095746. Trevisan, M.A., Mindlin, G.B. and Goller, F. (2006) Nonlinear model predicts diverse respiratory patterns of birdsong. Physical Review Letters, 96(5), 1–4. https://doi. org/10.1103/PhysRevLett.96.058103. Veit, L., Aronov, D. and Fee, M.S. (2011) Learning to breathe and sing: development of respiratory-vocal coordination in young songbirds. Journal of Neurophysiology, 106(4), 1747–1765. https://doi.org/10.1152/jn.00247.2011. Weir, J.T. and Wheatcroft, D. (2011) A latitudinal gradient in rates of evolution of avian syllable diversity and song length. Proceedings of the Royal Society B: Biological Sciences, 278(1712), 1713–20. https://doi.org/10.1098/ rspb.2010.2037. West, M.J. and King, A.P. (1988) Female visual displays affect the development of male song in the cowbird. Nature, 334, 244–246. https://doi.org/10.1038/332141a0. West, M.J., King, A.P., Eastzer, D.H. and Staddon, J.E. (1979) A bioassay of isolate cowbird song. Journal of Comparative and Physiological Psychology, 93(1), 124–133. CHAPTER 3 DIVERGENT EFFECTS OF SELECTION ON SONG MODULES IN A SETOPHAGA HYBRID ZONE 3.1 Abstract In cases of secondary contact between two species where hybrids are less fit than parents, theoretical and empirical evidence suggests that mating signals should diverge to decrease fitness loss due to hybridization, while aggressive signals should converge to limit the negative effects of aggressive interactions. Bird song is used both as a mating signal and as an aggressive signal, thus this trait is uniquely positioned to experience oppositional selective pressures in cases of secondary contact. The evolutionary trajectory of song in such a system will depend on the dynamics between these two competing forces, complicated by the interaction between genetic and cultural inheritance of song. In a hybrid zone between two Setophaga warblers, we observe a pattern of “divergent convergence” of song. While song rhythm shows a clinal convergence towards that typical of one species, acoustic morphology of syllables shows a nonclinal convergence towards that which is typical of the other species. These two song features, song rhythm and syllable morphology, are generated by different neuromuscular production modules, reflecting primarily respiratory and syringeal fine motor control, respectively. Because these modules 36 differ in the degree to which genetically encoded and learned aspects contribute to their proper development, the traits controlled by different song production modules can respond in different ways to the same selection pressures. The observed trend highlights the importance of trait evolution in response to secondary contact in generating cultural and genetically heritable variation. 3.2 Introduction Divergence in mating signals is expected to occur during the speciation process, but the evolutionary timing of such divergence is subject to debate and likely varies significantly according to various factors (reviewed in Coyne and Orr 2004; Wilkins et al. 2013; Grether et al. 2017). It is expected that such divergence contributes to reproductive isolation between two populations and thus increases speciation rates relative to the absence of divergence in mating signals (Pfennig and Pfennig 2009). Aggressive signals are found widely in animals and aid in securing access to resources, including food or mates (interference competition; Grether et al. 2017). While often more species-general than mating signals, aggressive signals are also expected to diverge during the speciation process via adaptive (Morton 1975; Endler 1992) or passive (Irwin et al. 2008) processes. However, unlike mating signals, aggressive signals are expected to converge upon secondary contact if genetic or ecological divergence between the two species is incomplete, a condition which encourages interspecific competition or territoriality (convergent agonistic character displacement; Grether et al. 2017). This phenomenon presumably arises out of selection for effective aggressive communication across species boundaries while reducing the potential costs of a direct physical encounter 37 (Cody 1969; Tobias and Seddon 2009; Grether et al. 2017). Convergence in aggressive signals should decrease loss of fitness due to hybridization (Grether et al. 2017), since mates and mating territories will be defended against competing members of both species. Birdsong is used as a mating signal and as an aggressive signal (Collins 2004). Thus, in cases of secondary contact, we expect oppositional selective pressures to act on song. Selection for the facilitation of aggressive signaling should exert pressure for song to converge, while selection for effective mate signaling should exert pressure for song to diverge between populations. The patterns in bird song that occur in response to secondary contact between two closely-related species can be viewed as an opportunity to dissect these competing selective pressures. It has been hypothesized that in cases of dual-purpose signals, the mating function should be the primary driver of evolutionary change (Okamoto and Grether 2013). This suggests that, in secondary contact where hybrids are less fit than parents, bird song of two species should diverge. However, empirical studies have shown mixed results, with some finding song convergence (de Kort et al. 2002; Secondi et al. 2003; Haavie et al. 2004; Kenyon et al. 2011; Laiolo 2012) and others divergence (Haavie et al. 2004; Kirschel et al. 2009) or stasis (Halfwerk et al. 2016) in cases of secondary contact. It is likely that the disparate patterns of song trait evolution in response to secondary contact observed by these studies are, in part, the result of the action of the competing selective pressures described above. Complicating the situation of potentially opposing trajectories of song in response to secondary contact is the fact that for all investigated oscine species, some aspects of species-specific song need to be learned by vocal imitation. At present, it remains difficult 38 to interpret how learning influences the observed evolutionary patterns of song, especially in light of the fact that the features of song subject to learning appear to vary across species (Love, Hoepfner and Goller 2019). Disentangling culturally transmitted from genetically inherited song traits would allow a more accurate assessment of song trait evolution, but, to date, little headway has been made in that regard. Still, a case can be made for the distinction between three main production “modules” that show evidence of differential reliance on genetically encoded information and of being reliant on distinct mechanisms for production: song rhythm, syllable morphology, and song syntax (Love, Hoepfner and Goller 2019). Learning and production of song is likely to involve the coordination of these distinct modules into an integrated system of control. Since laboratory studies suggest that the rhythm of song – the coarse patterns of sound and silence which are produced primarily by the respiratory pattern driving song activity – is likely to be under stricter genetic control and less subject to modification through learning than other production modules (syllable acoustic morphology and syntax; Marler and Sherman 1983; Ali et al. 2013; Araki et al. 2016; also see Lipkind et al. 2017), analysis of song across contact zones that distinguishes between production modules may provide an additional level of insight into the forces that guide signal evolution in response to secondary contact. For example, if one module shows convergence while another shows divergence, we can begin to draw conclusions about the targets and strengths of different selective forces. By gaining insight into the conditions which favor rapid signal evolution and the physiological mechanisms that drive or facilitate rapid signal evolution, a more detailed understanding of the conditions under which vocal learning evolved may emerge. Hermit warblers (Setophaga occidentalis) and Townsend’s warblers (Setophaga 39 townsendi) are sister species that are estimated to have diverged in glacial refugia during the middle or late Pleistocene and reached secondary contact approximately 5,000 years ago (Lovette and Bermingham 1999; Rohwer et al. 2001). Individuals of both species are interspecifically territorial and hybridize where their ranges overlap in the mountains of the northwestern United States, forming three narrow hybrid zones (Rohwer and Wood 1998). Hybrids between the two species have been reported to produce fewer eggs per clutch than either parent species, a sign of decreased hybrid fitness (Pearson 1997; Pearson and Rohwer 1998). Therefore, we should expect reinforcing selection to act on mating signals and decrease fitness loss due to hybridization. The results of an aggression study suggest that Townsend’s males are more aggressive than hermit males (Pearson 2000; Pearson and Rohwer 2000). Along with asymmetrical clutch sizes (Townsend’s females produce slightly larger clutches than hermit females at the same latitude), this aggressive asymmetry has been used as an explanation of the supposed movement of the hybrid zone from north to south, with Townsend’s replacing hermit warblers (Pearson and Rohwer 1998; Pearson 2000; Krosby and Rohwer 2010). While previous researchers have stated that species “sing the same songs” in the hybrid zone (Pearson and Rohwer 2000), a formal acoustic analysis has, to date, not been conducted, nor has the role of song in interspecific interactions been assessed. The Townsend’s warbler – hermit warbler hybrid zone system provides an excellent opportunity to examine the selective pressures acting on a dual-use, sexually selected signal during secondary contact of two closely related species. Here, we report how different acoustic features, based on the production modules, of the songs of hermit and Townsend’s warblers vary across the hybrid zone located in the Olympic mountains of Washington, 40 USA. We tested two possibilities of how song might be affected by the opposing selective forces. First, we hypothesized that songs of the two species will show distinct patterns of convergence or divergence in different song traits, according to the interaction of selective pressures for effective territorial aggression and for effective mate attraction, respectively, which could indicate division of functions within song. As an alternative, we also tested whether differentiation of function between different qualitatively assigned song types showed patterns of differential change across the hybrid zone. The results provide strong support for the first evolutionary scenario. 3.3 Methods From May to July of 2015-2018, the Olympic hybrid zone (described in Rohwer and Wood 1998) and surrounding allopatric regions were surveyed for singing Hermit, Townsend’s, and hybrid warbler males. A roving approach was used, wherein roads, trails, and off-trail routes were followed in vehicles or on foot until a singing individual was heard; that individual was then located, and high-quality recordings were made (Sennheiser directional microphone and Marantz PMD661 digital recorder). The individual was then attracted using a playback speaker (Apple iPhone 4, JBL Flip 2 speaker, connected with a miniplug cable) with the observer located over 10m away. Visual identification of the individual was made using binoculars and digital photography (Canon SL1 camera body with either a Canon 75-300 mm lens or a Canon 70-300 mm L lens in .jpg or .RAW format), and each individual was scored by plumage following guidelines established in Rohwer and Wood 1998 and as employed in the field by Pearson 1997 and Pearson 2000. There was one exception: face coloration was included in the overall score because recent study 41 on a closely related species suggests that this aspect of plumage (auricular) is likely to be a multilocus trait (Brelsford et al. 2017), not a single-locus trait as has been assumed by Rohwer and Wood (1998) and subsequent studies (but see S. Wang, International Ornithological Congress, Vancouver, B.C., August 2018). Song recordings made prior to playback attraction were used in all analyses. For each recording, the single song with the greatest acoustic quality was selected for analysis (high signal-to-noise ratio). This song was extracted from the original .wav file and saved as a new .wav file using Praat software. Each short file was band-pass filtered between 2.59.5 kHz, values that were selected to reduce background noise while maintaining full acoustic integrity across the frequency range of each song. These files were then segmented using RavenPro v1.5 by highlighting each syllable in a joint spectrogram and waveform view and compiling a “selection table” that includes start and end times for each selection. In addition to spectrogram and waveform visualization, power spectra were used to identify more precise start and end points of vocalizations. Syllables were defined as a single continuous sound separated from other sounds by a silent period longer than 0.015 seconds or as a group of sounds separated from each other by less than 0.015 seconds and separated from other sounds by more than 0.015 seconds. This definition was used to identify vocalizations made during a single expiratory event, in contrast to those made in a single pulse during pulsatile vocal production. 0.015 seconds is a value that approximates the shortest recorded inspiratory event (“mini-breath”) of any bird species (e.g., Hartley and Suthers 1989; Goller and Daley 2001). Two separate acoustic analyses were conducted: syllable morphology and rhythm. The acoustic morphology of syllables was analyzed by taking automated 42 measurements with the “specan” function from the warbleR package in R (Araya-Salas and Smith-Vidaurre 2017; R Core Team 2017). The measurements used were: mean frequency, standard deviation of frequency, 1st quartile frequency, 3rd quartile frequency, inter-quartile range of frequency, skewness, and kurtosis. In order to include a measure of frequency modulation unaffected by the duration of a syllable, an additional custom measurement, “modulation”, was created by dividing warbleR’s “modulation index” by the duration of the syllable. These eight measurements were then included in a linear discriminant analysis (“lda” function in R). The lda was trained on the data from allopatric populations of hermit and Townsend’s warblers, and then applied to the full dataset, including samples from both allopatric and sympatric populations. Rhythm was analyzed using custom code in R that assessed components of songs relating to the pattern of sound and silence; no frequency measures were included. Using the syllable start time and syllable duration from the RavenPro selection table, summary features of each song were computed in R. The rhythmic features that were recorded are: mean syllable length, mean silent period length, syllable rate (number of syllables/song length), “syllable regularity,” and “silent period regularity.” Syllable regularity is a normalized measure of consistency of syllable duration. This measure was computed within a single song by first finding the proportion of total sound that each syllable represents (syllable length divided by the sum of all syllable lengths). Then, the proportions were normalized by dividing each proportion by the maximum proportion. Finally, each of these normalized proportions was divided by the total number of syllables in the song. The resulting value yielded the syllable regularity. The same procedure was utilized to produce the silent period regularity, substituting silent period lengths for syllable lengths. A linear 43 discriminant function was then applied, following the same procedure that was used for syllable morphology. Independent one-dimensional geographic cline analyses were conducted on plumage index scores, rhythm LD1 scores, and syllable morphology LD1 scores. We used the hzar package in R, following a protocol outlined in (Derryberry et al. 2014) for quantitative traits and using default settings, no parameter constraints, and a single cycle of fit-requests (two sequential model fittings per analysis) to reach a maximum likelihood cline. Since these analyses relied on grouped sampling events, we binned our samples by every kilometer following a North-South axis. We used a direct North-South axis as our single geographical dimension, which fits the species and plumage score distribution across our study area as well as any other possible straight-line transect (see results). Assessment of song type analysis was conducted by qualitative visual inspection of spectrograms, categorizing songs by type using broad-scale similarity in the syntactical pattern of acoustically differentiated syllables. A single spectrogram of each song was created using the “catalog” function in the warbleR package. The spectrograms were visually analyzed sequentially, and each time that it was determined that a song did not match a pre-existing song type, a new type was created. These species have high interindividual variation in song, even within a song “type.” For example, the number of repetitions of the introductory phrase syllables varies among bouts of song given by a single individual within a short period of time, from the same perch, and (to the best of our knowledge) under the same social conditions (JL – personal observation). In our methodology, the songs with different numbers of introductory phrase syllables would not be categorized as different song types. 44 3.4 Results In the hermit warbler – Townsend’s warbler hybrid system, songs show considerable variation within and between species for most features. In both species, song is generally characterized by two phrases. The first phrase is composed of a repeated syllable or sequence of syllables, which consist either of a single note or several tightly clustered notes. The second phrase is variable and sometimes not present, but it often includes notes of longer duration than those found in the first phrase. The second phrase is usually shorter in duration than the first phrase. Occasionally, songs were found that do not adhere to this two-phrased structure. Frequencies fall in the range of approximately 3-9 kHz with high tonality, i.e., little upper harmonic content. In terms of the basic acoustic characterization of song in these species, our findings are roughly in line with the descriptions of previous studies (Morrison and Hardy 1983; Janes and Ryker 2006; Janes and Ryker 2016), but our analytical methods and large dataset lead us to reach new and different conclusions for a number of aspects. 3.4.1 Rhythm analysis Coefficients of the first linear discriminant (LD1) in the linear discriminant analysis are shown in Table 3.1. Songs that receive more positive rhythm LD1 scores have fewer syllables, longer silent periods between syllables, and a lower syllable repetition rate with less consistent syllable length. In allopatry, rhythm LD1 scores readily distinguished between hermit and Townsend’s warbler song, with Townsend’s warblers primarily singing songs with rhythm LD1 scores greater than zero (peak estimated at LD1=0.97) and hermit warblers primarily singing songs with rhythm LD1 scores less than zero (peak 45 estimated at LD1= -1.56) (Figure 3.1a). In sympatry, rhythm scores for hermit warbler song showed a positive shift (Figure 3.1b; red; peak at LD1= 0.89). Townsend’s warbler songs also showed a positive, albeit minor shift (Figure 3.1b; blue; peak at LD1 = 1.40). LD1 rhythm scores for hybrid songs were more broadly distributed with a positive LD1 peak (Figure 3.1b; violet; peak at LD1 = 1.41). The resulting pattern is that most songs from sympatric populations have rhythms that are similar to those produced in allopatric Townsend’s warbler populations. Rhythm LD1 scores showed a strong positive relationship with plumage ID scores (linear regression; Adjusted R-squared: 0.25, F-statistic: 91.9 on 1 and 276 DF, p-value: < 2.2e-16). A weaker but still significant relationship was found when excluding allopatric populations (linear regression; Adjusted R-squared: 0.032, F-statistic: 7.67 on 1 and 198 DF, p-value: 0.0061). Geographic distributions of rhythm features are illustrated in Figure 3.2. With the exception of song length, features roughly follow the clinal transition of plumage. However, many rhythmic features show the greatest level of divergence near the southern end of the hybrid zone, at the edge of the hermit allopatric population. This trend is only weakly evident, possibly due to the low population density in that geographic region coincident with decreased available suitable habitat. 3.4.2 Syllable morphology analysis Coefficients of the first linear discriminant (LD1) in the linear discriminant analysis are shown in Table 3.2. Syllables with lower LD1 scores have higher mean frequency and decrease in frequency over time. In allopatry, syllable morphology LD1 scores for 46 Townsend’s warblers show a distribution that has a strong peak in the positive range (estimated LD1 value = 0.48; Figure 3.1c; blue). The distribution for hermit warblers shows a noticeably bimodal distribution, with a strong peak close to that for Townsend’s warblers (LD1= 0.31), and a secondary peak in the negative range (Figure 3.1c; red; LD1= -1.52). In sympatry, a striking negative shift results in almost all songs, regardless of plumage phenotype or species identity, scoring in the negative range for LD1 (Figure 3.1d). Notably, the peak for each phenotype falls almost directly in line with the secondary peak in the allopatric hermit warbler syllable morphology (LD1 peak for hermit: -1.43; Townsend’s: 1.35; hybrid: -1.40). Syllable morphology LD1 shows a weak negative relationship with plumage ID score (linear regression; Adjusted R-squared: 0.024 F-statistic: 48.18 on 1 and 1953 DF, p-value: 5.27e-12). When excluding allopatric populations, this relationship is lost (linear regression; Adjusted R-squared: -0.00014, F-statistic: 0.81 on 1 and 1322 DF, p-value: 0.37). 3.4.3 Geographic cline analysis One-dimensional geographic cline analysis of plumage identity scores showed a cline with center at 5270800m N (UTM Zone 10N) and a width of 82.6 km. The cline created for rhythm LD1 scores had a center 46 km to the south at 5224637m N and a greater width of 193.4 km. The cline created for syllable morphology LD1 scores had a center at 5044372m N, 226 km to the south of the plumage score cline center, and a width of 938.4 km, an order of magnitude wider than the width of the plumage score cline. The model predicts the cline center for syllable morphology south of our study area, and the 47 distribution of syllable morphology LD1 scores relative to Northing did not show a logistic pattern that would be expected for quantitative traits across a hybrid zone (Figure 3.3). 3.4.4 Song type analysis We identified 24 song “types” used across the study area, based on qualitatively common acoustic morphology and syntax. While hermit and Townsend’s warblers showed some distinction in song types used, there is also a high degree of overlap. Song types showed more evidence of local geographic clustering across all phenotypes than of being more closely associated with one species, but it is difficult to disentangle the two factors, since some song types were found only in one allopatric region and are therefore de facto associated with one parental phenotype (Figure 3.4; Table 3.3). The most common song type was found across the full range of the study area and was used by all phenotypes. Interestingly, despite similar acoustic morphology and syntactical patterning, the rhythms of these “type 2” songs showed trends in rhythm that closely matched those found in the full song dataset. Hermit songs had lower rhythm LD1 scores than those of Townsend’s songs in allopatry, but rhythm score density distributions for both species shifted toward more positive values in sympatry and hybrid songs scored primarily in the positive range. Furthermore, comparison of hermit and Townsend’s song within each song type show the same trends; regardless of type, songs sung by Townsend’s warblers showed positive rhythm scores, while songs sung by hermit warblers showed negative rhythm scores in allopatry, but scored more positively in sympatry (Table 3.3). One exception is song type 9, all instances of which had negative rhythm scores. We did not find strong support for the two song type classification that has been 48 found in other Parulid species (Spector 1992) and which has been suggested may exist in these species (Morrison and Hardy 1983; Janes and Ryker 2006; Janes and Ryker 2013; Janes and Ryker 2016). Change from one song type to another has been used to indicate change in pairing status in these species (Pearson 1997), though no thorough analysis has, to our knowledge, been conducted on intraseasonal song change at the level of the individual (Janes and Ryker used songs qualitatively classified as “type I” but that were in part sung later in the season). Our song type 3 shares acoustic features with what has been described as “Type II” songs in Townsend’s warblers (Janes and Ryker 2016), but song type 3 was found relatively rarely in our study population (n=15, or roughly 5% of the total). However, it is notable that we did not find hermit or hybrid warblers singing this song type; only Townsend’s warblers sang song type 3 (Table 3.3). We used a Tukey posthoc test to assess differences in Julian date between song types, and the results showed that, out of 276 comparisons, 9 pairs of song types differed significantly in Julian occurrence dates. Song type 7 was found later in the season than 5 other song types, song type 12 was found earlier in the season than 2 other song types, song type 19 was found earlier in the season than two other song types, and song type 1 was found later in the season than two other song types (Table 3.4). With our current dataset, we cannot rule out the possibility that, as has been suggested, each individual male has a song type that is used pre-dawn before pairing and a song type used pre-dawn after pairing (i.e., “performanceencoded”; Byers 1995), though for our species we find that this suggestion has not been confirmed by sufficient evidence (but see Janes and Ryker 2016). It is clear that song type use in this system is more complex than has been found for some other parulid species that have two distinct song type classes (e.g., accented and 49 unaccented ending songs or type I and type II; reviewed in Spector 1992), though it is possible that our dataset may include songs of two main classes plus those of more rare forms (e.g., Byers 2017). The two most common song types (our type 1 and type 2) did not show a difference in seasonal timing from each other in the Tukey post-hoc test (adjusted P value=0.36). Our dataset includes samples from a broad altitudinal range, which could impact comparisons of seasonal timing of song use. For a more direct assessment of song type use, a future study would benefit by tracking song use by known individuals across the breeding season. Notably and as discussed in the results, the patterns observed for rhythm scores are maintained in analyses of songs within song types. 3.5 Discussion Our results highlight the complexities of evolutionary response to secondary contact for a trait that is subjected to opposing evolutionary pressures while its ontogeny is directed by both genetic control and cultural transmission. In syllable acoustic morphology, which is less likely to be genetically encoded (Love, Hoepfner and Goller 2019; Ali et al. 2013; Marler and Sherman 1983; Marler and Sherman 1985), we found an abrupt trait shift in the hybrid zone towards acoustic features found in allopatry exclusively among hermit warblers. In song rhythm, which is more likely to be genetically encoded (Marler and Sherman 1983; Marler and Sherman 1985; Love, Hoepfner and Goller 2019; Araki et al. 2017), we found an abrupt shift in the hybrid zone towards rhythmic features most commonly present in allopatry among Townsend’s warblers. The observed pattern for both acoustic features is consistent with that expected under agonistic character convergence of aggressive signals. However, the observation that the two features, acoustic 50 morphology and rhythm, converge toward different parental species gives important new insight. Cline analyses further clarifies these trends in song features. Again, song rhythm and acoustic features showed different trends. Song rhythm exhibited asymmetrical introgression of Townsend’s warbler rhythm into hermit warbler populations, while syllable acoustic morphology shows a nonclinal pattern of convergence towards hermit warbler syllable morphology in the hybrid zone. 3.5.1 Song rhythm: a genetically controlled trait under moderate selection Two possible explanations for the observed asymmetrical introgression of song rhythm are positive selection and a difference in parent species abundance. In areas of secondary contact, if a trait expressed by a single species is adaptive, it is expected to introgress into the population of the other parent species further than is expected of neutral traits (Barton 1979; Barton and Hewitt 2001). It is possible that a Townsend’s warbler-like rhythm could be adaptive in allopatric hermit populations if it provides an advantage in mate attraction or territory defense, but there is no current published evidence that this is the case. A playback study (Chapter 4 of this dissertation) suggests that there may be an adaptive benefit to singing songs with a Townsend’s warbler-like rhythm. However, it is also possible that differences in parent species abundance drive the observed asymmetrical introgression of song rhythm. In hybrid zones with major differences in the abundance of both parent species, neutral traits of the more abundant species are expected to introgress into the population of the less abundant species to a 51 greater degree than minority species traits introgress into the majority species population (Burgess et al. 2005; Currat et al. 2008). In the hybrid zone studied here, Townsend’s warblers vastly outnumber hermit warblers (see Figure 3.4), and this asymmetry in abundance may be underlying the observed asymmetrical introgression of rhythm. In addition to the difference in center location between the plumage index and the rhythm LD1 score clines, the cline widths also differed, with the plumage cline being more narrow than the rhythm cline. The notion that cline width is expected to decrease with increasing selective pressure (Barton and Hewitt 1985) suggests that song rhythm is under lower selective pressure than plumage. Thus, the more likely explanation for the observed pattern of rhythm introgression is a difference in parent species abundance, rather than positive selection. Still, we expect that both forces are likely to contribute (see Chapter 4). 3.5.2 Syllable morphology: culturally transmitted The change in syllable morphology across this hybrid zone is surprising in that it does not transition - neither sharply nor gradually - from one condition (hermit-like) to the other (Townsend’s-like). Rather, allopatric populations of hermit and Townsend’s warblers regularly used syllables with similar acoustic morphology, though hermits sometimes use syllable forms that are distinct from those found in Townsend’s song. Near the center of the hybrid zone, all individuals, regardless of plumage phenotype, used syllables with similar acoustic morphology to those found more rarely in the allopatric hermit population. There are a number of possible explanations for this pattern of convergence. First, the pattern could be the result of the vocal learning process. In most oscine species that have been studied, development of typical syllable morphology and syntax 52 requires exposure to conspecific syllables early in life (Love, Hoepfner and Goller 2019; Marler 1970). Typically, young birds learn from the social father or neighboring males. In some species, initial overproduction and post-dispersal selective attrition of learned song components has been shown to occur (Nelson and Marler 1994; Marler 1997; Nelson 2000; Beecher 2017). By these processes, song learning facilitates formation of local and regional dialects and geospatial clustering of song-type use. If hermit warblers, Townsend’s warblers, and hybrids share most aspects of the vocal learning process, from the innate song template to the timing of the sensitive and sensorimotor phases, then we expect that learned song traits should be culturally transmitted across species boundaries relatively uninhibited. By this process, dialects of learned song features and geospatial clustering of song-type use would proceed irrespective of species boundaries. Several studies of crosstutored individuals suggest that sharing of features of the song learning process occurs between some closely-related species pairs, sufficient to allow heterospecific song learning (Thorpe 1958; Immelmann 1969; Eriksen et al. 2009). Our results support the possibility that hermit warblers and Townsend’s warblers are capable of learning syllable morphology from each other’s songs. Second, the pattern may result from imbalanced selective pressures. If the fitness cost of hybridizing outweighs the fitness benefit of using song to negotiate territorial disputes, then we should see song divergence. In contrast, if the fitness benefit of using song to negotiate territorial disputes outweighs the fitness cost of hybridizing, then we should see song convergence. In the case of this hybrid zone, it appears that the former applies to hermit warblers, while the latter applies to Townsend’s warblers. This finding is in accordance with reinforcement theory, which predicts that the rarer species will 53 experience stronger selection pressure to evolve premating reproductive isolation (Noor 1995; Yukilevich 2012). Here, reinforcing selection causes hermit warblers in the contact zone to use syllables with distinct acoustic morphology. However, simultaneous selection pressure to share aggressive signals causes Townsend’s warblers in sympatric populations to use syllables that are characteristic of hermit warbler song. Together, these two processes could have formed the observed pattern of syllable morphology across the hybrid zone. If this is the case, we expect continued trait evolution in this direction, representing an adaptive scenario reminiscent of that occurring in host-parasite evolution (the “red queen hypothesis”) (Van Valen 1973). Constant pressure for hermit warblers to diverge in syllable morphology relative to Townsend’s warblers is counteracted by constant pressure for Townsend’s warblers to converge in syllable morphology relative to hermit warblers; the trait will evolve rapidly in the same direction for both species, while the level of reproductive isolation, and thus the rate of hybridization, remains stable. This interaction of forces may explain similarly complex patterns in song observed in other avian hybrid zones (e.g., Haavie et al. 2004). 3.5.3 Synthesis and relevance We hypothesized that the main different functions of song might be reflected in song features arising from different production aspects (“modules”) and thus show differential responses to diverging selective pressures. As an alternative, we suggested that song types may diverge in function in a similar way. We found support for the first hypothesis: song rhythm showed a markedly different pattern of evolution in response to secondary contact from that observed for syllable morphology. However, we have not 54 directly assessed song function in the current analysis. Chapter 4 of this dissertation addresses this issue more directly. Why, in our study system, syllable morphology responds to secondary contact in such a dramatically different way from that observed for song rhythm remains an open and intriguing question. The idea that differences in receivers’ perception of different song modules could drive our observed diverging patterns of convergence towards either species’ song for different song traits is addressed in Chapter 4 of this dissertation. It is possible that aspects of song convey different information and that the traits underlying this information experience a given selection pressure in accordingly different ways. As a result, unexpected patterns of trait evolution may occur. In addition, the disparity between evolutionary response of different song modules likely relates to differences in the strength of genetic control for these traits. In this study, we interpret song rhythm as a predominantly genetically-controlled trait. Support for this position is found in both the clinal transition of song rhythm across the hybrid zone and in the significant relationship between song rhythm LD1 score and plumage score, a genetically controlled, quantitative trait. It has long been recognized that in some species temporal structure, or rhythm, of song appears to be more strongly genetically heritable than other aspects of song (Thorpe 1958; Marler and Sherman 1983), and more recent studies confirm this for other species (Ali et al. 2013; Araki et al. 2016). Rhythm perception has been found to be remarkably similar in distantly related vocal learning species (ten Cate and Spierings 2018). Additionally, a comparative approach has been applied to those species that have been raised in isolation, which unveiled a common trend: rhythm is more strongly innately controlled than other aspects of song for most oscine species that have 55 been thoroughly studied (Love, Hoepfner and Goller 2019). The results of our study, presented here, agree strongly with this body of work; we found that rhythm transitioned smoothly from one species to the other across the hybrid zone, just as one would expect from a quantitative, genetically-controlled trait experiencing moderate selection (Barton and Hewitt 1985), while another aspect of song, acoustic structure, showed a strikingly different pattern, which is presumably the result of cultural transmission. To what degree vocal learning and genetically encoded aspects of song differentially drive phenotypic changes needs to be investigated further for this system, and our suggested mechanisms present testable hypotheses. Interestingly, studies of other avian hybrid zones find similar, complex patterns of trait evolution characterized by a clinal convergent transition in song rhythm paired with a nonclinal pattern of syllable morphology. A 2003 study by Secondi et al. found a similar pattern for song of melodious, Hippolais polyglotta, and icterine warblers, H. icterina, with rhythmic features (in their study, syllable duration and silent period duration) showing a clinal transition while syllable morphology (“frequency” song features in their study) did not. The authors note that “… the spatial variation of song suggests the presence of a vocal cline as usually observed for allelic frequencies or genetically determined traits in hybrid zones” (Secondi et al. 2003, p. 515). Similarly, in a recent study of a subspecies hybrid zone (white-crowned sparrow, Zonotricha leucophrys) Lipshutz et al. found that rhythmic features (whistle length and average syllable length) showed a clinal transition pattern, and they acknowledge the likely innate nature of those song features (Lipshutz et al. 2017). It may be the case that differences in genetic control between song production modules are observable in all bird song hybrid zones, however, since many studies use a multi- 56 dimensional scaling approach to song analysis, the distinction between learned and genetically determined song traits may go unnoticed or the patterns uncovered in the process of analysis may be difficult to interpret. We therefore propose that distinguishing between rhythm, syllable acoustic morphology, and syntax in analyses can contribute an additional level of distinction and thus the nature of questions that can be answered by the study of bird song in natural populations. Reanalyzing data from past studies with this approach may be highly elucidative. Furthermore, future advances in our understanding of the neurobiology of bird song should prove to be of great assistance when incorporated into study of birdsong in natural populations. The results presented here can be seen to highlight the contribution of secondary contact to increasing culturally and genetically heritable variation on which selection can act. Secondary contact and hybridization, once thought to be evolutionarily rare, is now acknowledged to be a regular occurrence on evolutionary time-scales (Mallet 2005). Furthermore, hybridization has been acknowledged as a potential major contributor to rapid speciation (Seehausen 2004). In our study, we have a present-day example of secondary contact and subsequent hybridization encouraging increased trait variation, upon which selective pressures act to produce rapid phenotypic change. Our evidence for potential Red Queen selection for rapid change in a learned aspect of vocal production (syllable morphology) in response to secondary contact, as described above, may have played a major role in the evolution of vocal learning in birds, which is characterized by the propensity for rapid evolutionary change in comparison to the assumed slower pace for equivalent genetically-controlled vocal production. 57 Our findings suggest that, contrary to theoretical predictions (Okamoto and Grether 2013), selection acts to support the territory defense function of song more strongly than the mate attraction function of song. It is possible that the historical focus on male song as a primary mating signal may be somewhat misleading. Perhaps, song serves only a basic mate locating function for many species. Other signals such as for example soft song (e.g., Titus 1998; Dabelsteen et al. 1998) may provide more important information leading to the actual mate choice by females than territorial song does. Nevertheless, the respective signal value may vary between species, explaining evidence that song is, at least under certain conditions, used by females to assess mates (e.g., Nowicki et al. 2002; Hauber et al. 2013). Whether or not our observation that implies a greater role of song in intrasexual than intersexual communication applies broadly to all similar systems with dual-use signals is, for now, unknown, but the evolutionary trajectory of such signals is likely to vary based on the relative fitness costs and benefits associated with hybridization and territorial defense, as has been previously noted (ten Cate 2004), coupled with the relative functional importance of the signal. Repeating our methods in hybrid systems where parent species pairs vary in their genetic, ecological, and behavioral divergence would allow for a more systematic comparison of selection pressure strengths. The numerous existing avian hybrid zones provide an excellent opportunity to conduct such a study. 3.6 References Ali F, Otchy TM, Pehlevan C, Fantana AL, Burak Y, Ölveczky BP (2013) The basal ganglia is necessary for learning spectral, but not temporal, features of birdsong. Neuron 80(2):494–506 doi:10.1016/j.neuron.2013.07.049 Araki M, Bandi MM, Yazaki-Sugiyama Y (2016) Mind the gap: Neural coding of species identity in birdsong prosody. Science 354(6317):1282–1287 doi:10.1126/science.aah6799 58 Araya-Salas M, Smith-Vidaurre G (2017) warbleR: An r package to streamline analysis of animal acoustic signals. Methods Ecol Evol 8(2):184–191 doi:10.1111/2041-210X.12624 Barton NH (1979) Gene doi:10.1038/hdy.1979.86 flow past a cline. Heredity 43(3):333–339 Barton NH, Hewitt GM (1985) Analysis of hybrid zones. Annu Rev Ecol Syst 16:113–148 doi:10.1146/annurev.es.16.110185.000553 Barton NH, Hewitt GM (2001) The role of hybridization in evolution. Mol Ecol 10:551– 568 Beecher MD (2017) Birdsong learning as a social process. Anim Behav 124:233–246 doi:10.1016/j.anbehav.2016.09.001 Brelsford A, Toews DPL, Irwin DE (2017) Admixture mapping in a hybrid zone reveals loci associated with avian feather coloration Proc Roy Soc B 284(1866) 20171106 doi:10.1098/RSPB.2017.1106 Burgess KS, Morgan M, Deverno L, Husband BC (2005) Asymmetrical introgression between two Morus species (M. alba, M. rubra) that differ in abundance. Mol Ecol 14(11):3471–3483 doi:10.1111/j.1365-294X.2005.02670.x Byers BE (1995) Song types, repertoires and song variability in a population of chestnutsided warblers. Condor 97(2):390–401 Byers BE (2017) Chestnut-sided warblers use rare song types in extreme aggressive contexts. Anim Behav 125:33–39 doi:10.1016/j.anbehav.2017.01.007 ten Cate C (2004) Birdsong and evolution. In: Nature’s Music. San Diego, CA: Elsevier. pp. 296–317 ten Cate C, Spierings M (2018) Rules, rhythm and grouping: Auditory pattern perception by birds. Anim Behav 151:259-257 doi:10.1016/J.ANBEHAV.2018.11.010 Cody L (1969) Convergent characteristics in sympatric species: A possible relation to interspecific competition and aggression. Condor 71:222–239 doi:10.2307/1366300 Collins S (2004) Vocal fighting and flirting: The functions of birdsong. In: Nature’s Music. San Diego, CA: Elsevier. pp 39–79 Coyne JA, Orr HA (2004) Speciation. Sunderland, MA: Sinauer Associates, Inc. Currat M, Ruedi M, Petit RJ, Excoffier L (2008) The hidden side of invasions: Massive introgression by local genes. Evolution 62(8):1908–1920 doi:10.1111/j.15585646.2008.00413.x Derryberry EP, Derryberry GE, Maley JM, Brumfield RT (2014) Hzar: Hybrid zone 59 analysis using an R software package. Mol Ecol Resour 14(3):652–663 doi:10.1111/17550998.12209 Endler JA (1992) Signals, signal conditions, and the direction of evolution. Am Nat 139:S125–S153 doi:10.1086/285308 Eriksen A, Lampe HM, Slagsvold T (2009) Interspecific cross-fostering affects song acquisition but not mate choice in pied flycatchers, Ficedula hypoleuca. Anim Behav 78(4):857–863 doi:10.1016/j.anbehav.2009.07.005 Goller F, Daley MA (2001) Novel motor gestures for phonation during inspiration enhance the acoustic complexity of birdsong. Proc R Soc B Biol Sci 268(1483):2301–2305 doi:10.1098/rspb.2001.1805 Grether GF, Peiman KS, Tobias JA, Robinson BW (2017) Causes and consequences of behavioral interference between species. Trends Ecol Evol 32(10):760–772 doi:10.1016/j.tree.2017.07.004 Haavie J, Borge T, Bures S, Garamszegi LZ, Lampe HM, Moreno J, Qvarnström A, Török J, Sætre GP (2004) Flycatcher song in allopatry and sympatry - Convergence, divergence and reinforcement. J Evol Biol 17(2):227–237 doi:10.1111/j.1420-9101.2003.00682.x Halfwerk W, Dingle C, Brinkhuizen DM, Poelstra JW, Komdeur J, Slabbekoorn H (2016) Sharp acoustic boundaries across an altitudinal avian hybrid zone despite asymmetric introgression. J Evol Biol 29(7):1356–1367 doi:10.1111/jeb.12876 Hartley RS, Suthers RA (1989) Airflow and pressure during canary song: Direct evidence for mini-breaths. J Comp Physiol A 165(1):15–26 doi:10.1007/BF00613795 Hauber ME, Woolley SMN, Cassey P, Theunissen FE (2013) Experience dependence of neural responses to different classes of male songs in the primary auditory forebrain of female songbirds. Behav Brain Res 243:184–90 doi:10.1016/j.bbr.2013.01.007 Immelmann K (1969) Song development in the zebra finch and other estrildid finches. In: Bird Vocalizations. Cambridge, UK: Cambridge University Press. pp 61–74 Irwin DE, Thimgan MP, Irwin JH (2008) Call divergence is correlated with geographic and genetic distance in greenish warblers (Phylloscopus trochiloides): A strong role for stochasticity in signal evolution? J Evol Biol 21(2):435–448 doi:10.1111/j.14209101.2007.01499.x Janes SW, Ryker L (2006) Singing of hermit warblers: Dialects of type I songs. Condor 108(2):336–347 Janes SW, Ryker L (2013) Rapid change in a type I song dialect of hermit warblers (Setophaga occidentalis). Auk 130(1):30–35 doi:10.1525/auk.2012.11273 Janes SW, Ryker L (2016) Type I and II songs of Townsend’s warblers in Oregon and 60 Washington. West Birds 47:67–73 Kenyon HL, Toews DPL, Irwin DE (2011) Can song discriminate between Macgillivray’s and mourning warblers in a narrow hybrid zone? Condor 113(3):655–663 doi:10.1525/cond.2011.100182 Kirschel ANG, Blumstein DT, Smith TB (2009) Character displacement of song and morphology in African tinkerbirds. Proc Natl Acad Sci USA 106(20):8256–6. doi:10.1073/pnas.0810124106 de Kort SR, den Hartog PM, ten Cate C (2002) Diverge or merge? The effect of sympatric occurrence on the territorial vocalizations of the vinaceous dove Streptopelia vinacea and the ring-necked dove S. capicola. J Avian Biol 33(2):150–158 doi:10.1034/j.1600048X.2002.330205.x Krosby M, Rohwer S (2010) Ongoing movement of the hermit warbler X Townsend’s warbler hybrid zone. Fenton B, editor. PLoS One 5(11):e14164. doi:10.1371/journal.pone.0014164 Laiolo P (2012) Interspecific interactions drive cultural co-evolution and acoustic convergence in syntopic species. J Anim Ecol 81(3):594–604 doi:10.1111/j.13652656.2011.01946.x Lipkind D, Zai AT, Hanuschkin A, Marcus GF, Tchernichovski O, Hahnloser RHR (2017) Songbirds work around computational complexity by learning song vocabulary independently of sequence. Nat Commun. 8(1) doi:10.1038/s41467-017-01436-0 Lipshutz SE, Overcast IA, Hickerson MJ, Brumfield RT, Derryberry EP (2017) Behavioural response to song and genetic divergence in two subspecies of white-crowned sparrows (Zonotrichia leucophrys). Mol Ecol 26(11):3011–3027 doi:10.1111/mec.14002 Lovette IJ, Bermingham E (1999) Explosive speciation in the New World Dendroica warblers. Proc R Soc B 266:1629–1636 Mallet J (2005) Hybridization as an invasion of the genome. Trends Ecol Evol 20(5):229– 237 doi:10.1016/j.tree.2005.02.010 Marler P (1970) A comparative approach to vocal learning: Song development in whitecrowned sparrows. J Comp Physiol Psychol 71(2) Marler P (1997) Three models of song learning: Evidence from behavior. J Neurobiol 33(5):501–16 Marler P, Sherman V (1983) Song structure without auditory feedback: Emendations of the auditory template hypothesis. J Neurosci 3(3):517–531 Marler P, Sherman V (1985) Innate differences in singing behavior of sparrows reared in isolation from adult conspecific song. Anim Behav 33(1):57–71 61 Morrison ML, Hardy JW (1983) Hybridization between hermit and Townsend’s warblers. The Murrelet 64(3):65–72 Morton ES (1975) Ecological sources of selection on avian sounds. Am Nat 109(965):17– 34. doi:10.1086/282971 Nelson D (2000) Song overproduction, selective attrition and song dialects in the whitecrowned sparrow. Anim Behav 60(6):887–898 doi:10.1006/anbe.2000.1560 Nelson D, Marler P (1994) Selection-based learning in bird song development. Proc Natl Acad Sci USA 91(22):10498–10501 doi:10.1073/pnas.91.22.10498 Noor MA (1995) Speciation driven by natural selection in Drosophila. Nature 375(6533):674–675 Nowicki S, Searcy W, Peters S (2002) Quality of song learning affects female response to male bird song. Proc Biol Sci 269(1503):1949–54 doi:10.1098/rspb.2002.2124 Okamoto KW, Grether GF (2013) The evolution of species recognition in competitive and mating contexts: The relative efficacy of alternative mechanisms of character displacement. Ecol Lett 16(5):670–678 doi:10.1111/ele.12100 Pearson S (1997) Behavioral and ecological tests of four models explaining narrow hybrid zones between hermit and Townsend’s warblers (PhD dissertation). University of Washington Pearson S (2000) Behavioral asymmetries in a moving hybrid zone. Behav Ecol 11(1):84– 92 doi:10.1093/beheco/11.1.84 Pearson S, Rohwer S (1998) Influence of breeding phenology and clutch size on hybridization between hermit and Townsend’s warblers. Auk 115(3):739–745 Pearson S, Rohwer S (2000) Asymmetries in male aggression across an avian hybrid zone. Behav Ecol 11(1):93–101 doi:10.1093/beheco/11.1.93 Pfennig KS, Pfennig DW (2009) Character displacement: Ecological and reproductive responses to a common evolutionary problem. Q Rev Biol 84(3):253–276. doi:10.1086/605079 R Core Team (2017) R: A language and environment for statistical computing. Rohwer S, Bermingham E, Wood C (2001) Plumage and mitochondrial DNA haplotype variation across amoving hybrid zone. Evolution 55(2):405–422 Rohwer S, Wood C (1998) Three hybrid zones between hermit and Townsend’s warblers in Washington and Oregon. Auk 115(2):284–310 Secondi J, Bretagnolle V, Compagnon C, Faivre B (2003) Species-specific song 62 convergence in a moving hybrid zone between two passerines. Biol J Linn Soc 80(3):507– 517 doi:10.1046/j.1095-8312.2003.00248.x Seehausen O (2004) Hybridization and adaptive radiation. Trends Ecol Evol 19(4):198– 207 doi:10.1016/j.tree.2004.01.003 Spector D (1992) Wood-warbler song systems. Curr Ornithol 9:199–238 Thorpe WH (1958) The learning of song patterns by birds, with especial reference to the song of the chaffinch fringilla coelebs Ibis 100(4):535–570 Titus RC (1998) Short-range and long-range songs: Use of two acoustically distinct song classes by dark-eyed juncos Auk 115(2):386–393. doi:10.2307/4089197 Tobias JA, Seddon N (2009) Signal design and perception in hypocnemis antbirds: Evidence for convergent evolution via social selection. Evolution 63(12):3168–3189 doi:10.1111/j.1558-5646.2009.00795.x Van Valen L (1973) A new evolutionary law. Evol Theory 1:1–30. doi:10.1038/344864a0 Wilkins MR, Seddon N, Safran RJ (2013) Evolutionary divergence in acoustic signals: Causes and consequences. Trends Ecol Evol 28(3):156–166 doi:10.1016/j.tree.2012.10.002 Yukilevich R (2012) Asymmetrical patterns of speciation uniquely support reinforcement in Drosophila. Evolution 66(5):1430–1446 doi:10.1111/j.1558-5646.2011.01534.x 63 Figure 3.1: Density distributions of rhythm (a,b) and syllable morphology (c,d) LD1 scores in allopatry (top) and sympatry (bottom). 64 Figure 3.2: Geospatial distributions of samples (dots) and kernel smoothing representations of plumage ID (top left) and rhythmic song features (all others). The scales for number of syllables, song length, and syllable rate were inverted for consistent representation: cooler colors indicate hermit-like trait values and warmer colors indicate Townsend’s-like traits values for the features indicated in plot titles. Plots produced with spatstat package (Baddeley; www.spatstat.org) in R. We used a bandwidth of 1300 for the kernel smoothing. 65 Figure 3.3: Geographic cline analysis. Samples were averaged every 1 kilometer. Black line represents the maximum-likelihood cline. Gray area indicates 95% credible cline region. Dashed red lines indicate estimated plumage ID cline center and edges, dashed blue lines indicate estimated Rhythm LD1 cline center and edges. Plots produced with HZAR package (Derryberry et al. 2014) in R. 66 Figure 3.4: Geographic distribution of common song types. Polygons indicate the geographic range found for 6 of the most common song types. Representative spectrograms are shown on right. 67 Table 3.1: Coefficients of rhythm linear discriminant analysis. Number of syllables Mean syllable duration Mean silent period duration Song length Syllable rate Rhythmic silence regularity Rhythmic sound regularity LD1 -1.09 -0.18 0.36 0.0052 -0.4 0.032 -0.22 68 Table 3.2: Coefficients of syllable morphology linear discriminant analysis. Mean frequency Standard deviation of frequency First quartile frequency Third quartile frequency Interquartile range frequency Skewness Kurtosis Modulation LD1 2.05 0.89 -1.51 -1.51 -0.41 -0.84 0.99 -0.21 69 Table 3.3: Song type summary statistics. Sample size, mean, median, and standard deviation of rhythm LD1 for each song type, sung by each species, in allopatry and in sympatry. Shading groups rows by song type (left column). song type 1 1 1 1 1 2 2 2 2 2 3 3 3 3 3 4 4 4 4 4 5 5 5 5 5 6 6 6 6 6 7 7 species TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA allopatry 1 1 0 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 0 1 1 n 0 0 16 11 5 10 4 45 18 4 4 0 11 0 0 3 0 15 1 0 4 1 3 1 0 10 0 6 8 0 1 0 mean NA NA 0.33 0.50 0.51 1.50 -0.01 1.47 1.67 0.69 1.58 NA 2.11 NA NA 0.93 NA 0.71 2.01 NA 1.91 -0.34 1.98 1.16 NA 1.15 NA 0.12 0.09 NA 1.69 NA median NA NA 0.36 0.36 0.79 1.48 -0.16 1.55 1.77 0.83 1.57 NA 2.01 NA NA 1.09 NA 0.88 2.01 NA 1.69 -0.34 1.87 1.16 NA 0.87 NA 0.29 -0.30 NA 1.69 NA sd NA NA 0.98 0.90 1.34 0.97 1.12 0.82 0.96 0.45 0.29 NA 1.08 NA NA 0.33 NA 0.78 NA NA 0.98 NA 0.52 NA NA 0.83 NA 1.33 1.18 NA NA NA 70 Table 3.3 continued song type 7 7 7 8 8 8 8 8 9 9 9 9 9 10 10 10 10 10 11 11 11 11 11 12 12 12 12 12 13 13 13 13 13 14 14 14 14 14 species TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA allopatry 0 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 0 n 11 2 3 0 0 1 0 0 0 6 2 2 1 0 1 2 0 0 0 1 0 0 0 0 15 2 0 0 0 1 0 0 0 0 0 5 1 0 mean 1.57 1.41 0.94 NA NA 1.34 NA NA NA -2.19 -3.53 -3.21 -2.69 NA -0.33 -0.89 NA NA NA -2.39 NA NA NA NA -1.86 1.66 NA NA NA 0.32 NA NA NA NA NA 1.71 -0.31 NA median 1.51 1.41 1.35 NA NA 1.34 NA NA NA -1.99 -3.53 -3.21 -2.69 NA -0.33 -0.89 NA NA NA -2.39 NA NA NA NA -1.87 1.66 NA NA NA 0.32 NA NA NA NA NA 1.85 -0.31 NA sd 0.58 0.14 1.80 NA NA NA NA NA NA 0.84 0.38 0.61 NA NA NA 1.34 NA NA NA NA NA NA NA NA 0.71 0.51 NA NA NA NA NA NA NA NA NA 0.71 NA NA 71 Table 3.3 continued song type 15 15 15 15 15 16 16 16 16 16 17 17 17 17 17 18 18 18 18 18 19 19 19 19 19 20 20 20 20 20 21 21 21 21 21 22 22 22 species TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA allopatry 1 1 0 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 n 2 0 0 0 0 1 0 3 0 0 0 0 2 0 0 0 0 2 0 0 0 7 8 3 2 0 1 1 2 0 2 0 1 0 0 2 0 0 mean 0.11 NA NA NA NA 0.29 NA 0.76 NA NA NA NA -0.77 NA NA NA NA 0.91 NA NA NA -0.88 0.30 0.15 -0.25 NA -0.04 2.31 -1.03 NA 1.46 NA 2.46 NA NA -0.26 NA NA median 0.11 NA NA NA NA 0.29 NA 1.08 NA NA NA NA -0.77 NA NA NA NA 0.91 NA NA NA -1.03 0.57 -0.29 -0.25 NA -0.04 2.31 -1.03 NA 1.46 NA 2.46 NA NA -0.26 NA NA sd 1.01 NA NA NA NA NA NA 0.67 NA NA NA NA 1.74 NA NA NA NA 0.14 NA NA NA 0.80 0.92 1.19 0.71 NA NA NA 2.32 NA 1.09 NA NA NA NA 1.24 NA NA 72 Table 3.3 continued song type 22 22 23 23 23 23 23 24 24 24 24 24 species HTHW HEWA TOWA HEWA TOWA HTHW HEWA TOWA HEWA TOWA HTHW HEWA allopatry 0 0 1 1 0 0 0 1 1 0 0 0 n 0 0 1 0 0 0 0 1 0 0 0 0 mean NA NA NA NA NA NA NA NA NA NA median NA NA 0.88 0.88 NA NA NA NA 1.35 1.35 NA NA NA NA sd NA NA NA NA NA NA NA NA NA NA NA NA 73 Table 3.4: Results from Tukey post-hoc “Honest Significant Difference” test of day of year as a function of song type, showing significant results (p<0.05) only. Columns are song types, difference between the means, lower end and upper end of the 95% family-wise CI, and the p-value after adjustment for multiple comparisons. types 21-1 19-1 7-4 7-6 9-7 12-7 19-7 17-12 19-17 diff -16.75 -15.73 13.93 13.80 -16.52 -20.35 -19.33 31.47 -30.45 lwr -28.54 -26.93 0.82 1.34 -31.72 -33.83 -32.29 2.10 -59.59 upr -4.96 -4.53 27.05 26.25 -1.31 -6.88 -6.37 60.84 -1.31 p-adj 9.43E-05 1.25E-04 2.33E-02 1.28E-02 1.70E-02 1.86E-05 2.63E-05 2.07E-02 2.90E-02 CHAPTER 4 EVIDENCE FOR DIFFERENTIAL SEXUAL SELECTION ON VARIOUS SONG COMPONENTS IN A SETOPHAGA HYBRID ZONE 4.1 Abstract The song of oscine passerine birds is a learned vocal signal that is thought to be subjected to strong sexual and natural selection (Nowicki and Searcy 2014). The rapid emergence of song culture through vocal learning has been viewed as a mechanism for increasing speciation rates by effectively decreasing the time for premating isolating mechanisms to evolve (Lachlan and Servedio 2004), though empirical evidence is rare. In a hybrid zone between two Setophaga species, we find evidence that sheds doubt on this hypothesis. Male and female responses to song do not differ according to divergent learned song features. On the contrary, female responses systematically differ according to song rhythm, a feature of song that is less likely to be learned. Females show greater response to playback songs that exhibit rhythms found in allopatric populations of Townsend’s warblers, Setophaga townsendii, than to songs with hermit warbler, Setophaga occidentalis, rhythms. This finding suggests a role for intersexual selection in the evolution of genetically inherited song traits in response to secondary contact. Together, these findings indicate that song learning alone does not necessarily constitute an efficient 75 mechanism for increasing speciation rates. Instead, the aspects of song that are more strongly genetically-controlled appear to serve a greater role in mate attraction. Furthermore, hybridization may coincide with reduced territoriality in this Setophaga hybrid zone. 4.2 Introduction Vocal learning refers to the ability of an animal to alter its vocal ontogeny and vocal signals in response to auditory information (Nottebohm 1972) and is found in birds and mammals (Janik and Slater 1997). In birds, vocal learning is thought to have played a major role in the rapid speciation of oscine passerines (Passeri), which comprise nearly half of all known species (Lachlan and Servedio 2004). Through cultural transmission, learned song can diverge rapidly within and between populations (Mason et al. 2017) and thus contribute to their isolation from each other. Song is a multiuse signal in passerines that is commonly cited as serving dual roles in mate attraction and territory defense and is used at long and short range (Collins 2004). While song is known to be used in several contexts (e.g., duetting, Hall 2009) and by both sexes (Odom et al. 2014), most study has focused on species where predominantly the males sing to attract females and to negotiate territorial disputes with neighboring males. Because song is viewed as an important mating signal under strong sexual selection, it has been hypothesized that the culturally mediated increase in the rate of song divergence in the oscines may be associated with the rapid diversification of this group (Lachlan and Servedio 2004). In part due to limited supporting empirical evidence, this hypothesis has been subject to continued discussion (Baptista and Trail 1992; Slabbekoorn and Smith 76 2002; Seddon and Tobias 2007; Olofsson et al. 2011; Verzijden et al. 2012; Yeh 2018; Yeh 2019). A hybrid zone between two Setophaga warblers, which are presumed to learn their songs, provides a valuable system to test some of the key assumptions underlying this hypothesis. Hermit warblers (Setophaga occidentalis) and Townsend’s warblers (S. townsendi) are sister taxa that are interspecifically territorial and interbreed to produce viable hybrid offspring where they meet in three narrow hybrid zones in Washington and Oregon, USA (Rohwer and Wood 1998). Hybrid pairing is apparently directional, as Townsend’s males have been shown to readily pair with heterospecific mates, while hermit males appear to successfully pair with conspecific females exclusively (Pearson 2000). Pearson (2000) suggests that female preference for Townsend's-like traits may be driving hybridization since, despite asymmetries in territorial aggression with Townsend’s being more aggressive than hermit warblers, no difference has been found in territory quality between the two species or their hybrids within the hybrid zones (Pearson 2000). Notably, playback of song was used with mounts to investigate territorial aggression in males of the species pair, but based on preliminary analysis of song the assumption was made that song does not vary with species identity in the hybrid zone (Pearson and Rohwer 2000). As described in Chapter 3 of this dissertation, details of song use in the hybrid zone are somewhat more complex than was suggested by previous researchers. Song of the two species does converge in the hybrid zone, but in two distinct ways. The syllable morphology of song converges in a nonclinal way towards hermit-like syllables, a pattern which likely reflects song learning across species boundaries. In contrast, song rhythm, which is more likely to be under stricter genetic control, converges on Towsend’s-typical 77 rhythm in a clinal pattern, reminiscent of a pattern shown by a quantitative trait under moderate selection. In an additional analysis, we expanded our dataset by combining our own recordings with song recordings from public databases (xeno-canto.org and Macaulay Library, Cornell Lab of Ornithology) and performed a qualitative analysis of song features from allopatric and sympatric populations as well as hybrids. Again, aspects of song that were divergent in allopatry converged in sympatry, opposite of the expectation of reproductive character displacement predicted by reinforcing selection acting to reduce hybridization events. In particular, the syllable morphology of the “introductory” notes was observed to be clearly divergent between species when comparing allopatric populations of hermit warblers to interior allopatric populations of Townsend’s warblers (see Figure 4.1 (a,c): multinote hermit and Figure 4.1 (f,h): single-note Townsend’s song). Some parent-species males from the hybrid zone sang songs that used introductory syllables found in allopatric song of heterospecifics rather than those found in song from allopatric areas of their own species (see Figure 4.1 (b,d,e,g)). See methods for details on species identification and acoustic analyses, which differ slightly from those used in the analysis in Chapter 3. This results in an additional pattern of song character convergence upon secondary contact, which is evident over a broader geographic range than that presented in Chapter 3 of this dissertation. Previous researchers have found similar patterns of uni- or bidirectional song trait convergence in hybrid zones of other species, and have suggested that heterospecific song learning may contribute to the production of such patterns (Haavie et al. 2004; Qvarnström et al. 2006; Kenyon et al. 2011). Whether or not learned aspects of song observed to be 78 convergent in hybrid zones are salient to the signal receiver is an important question that has seen little study (but see Qvarnström et al. 2006). Less emphasis has been placed on the role of genetically-guided aspects of song in facilitating inter- and intraspecific interactions in vocal learning species (but see studies in nonvocal learning taxa: Seddon and Tobias 2007, Tobias and Seddon 2009, Seddon and Tobias 2010; and studies of strictly genetically controlled song features guiding song development: Soha and Marler 2000, Nelson 2017). Though direct studies of avian vocal learning have long recognized a strong innate basis for some aspects of song production (Thorpe 1958; Marler and Sherman 1983; Regulus songs: reviewed in Packert et al. 2003), it is common in studies of bird song in natural populations to assume that any measurable aspect of song is subject to development through the learning process. However, as we demonstrate in Chapters 2 and 3 of this dissertation, a strong case can be made for the distinction between three main production “modules” that show evidence of differential reliance on genetically encoded information and of being reliant on distinct mechanisms for production: song rhythm, syllable morphology, and song syntax (Love, Hoepfner and Goller 2019). Learning and production of song is likely to involve the coordination of these distinct modules into an integrated system of control. Since laboratory studies suggest that the rhythm of song – the coarse patterns of sound and silence which are produced primarily by the respiratory pattern driving song activity – is likely to be under stricter genetic control and less subject to modification through learning than other production modules (syllable acoustic morphology and syntax) (Marler and Sherman 1983; Ali et al. 2013; Araki et al. 2016; Lipkind et al. 2017), analysis 79 of song across contact zones that distinguishes between production modules may provide an additional level of insight into the forces that guide signal evolution in response to secondary contact. We executed such an analysis in Chapter 3 of this dissertation. In the playback study that is the focus of the current chapter, we put to test the hypothesis that sexual selection has driven the observed patterns of trait evolution in response to secondary contact. If Townsend’s and hermit warblers use divergent, learned song traits to assess competitors for territories or potential mates, then we expect that they will respond differently to songs that vary along this line. Conversely, if these learned song traits are relatively unimportant for the assessment of potential competitors or mates, then individuals should not respond differently to songs that differed in this trait. The same holds true for genetically controlled aspects of song; if they are used to assess territorial competitors or mates, we should observe different responses to songs that vary in that aspect. To address this issue, we devised a playback experiment that measured response to a suite of songs that differed in both learned (syllable morphology and syntax) and nonlearned (rhythm) features and were sung by members of both species (Figure 4.1). Males and females of each species were tested in both allopatry and sympatry. 4.3 Methods Individuals were located in the field by song. Song was recorded and then the individual was attracted using a playback speaker for visual identification. Species identity for each individual was scored by plumage as in Rohwer and Wood (1998), with one exception: “face coloration” was included in the overall score because recent study on a closely related species suggests that this aspect of plumage (“auricular”) is likely to be a 80 multilocus trait (Brelsford et al. 2017), not a single-locus trait as has been assumed by Rohwer and Wood (1998) and subsequent studies (but see S. Wang, International Ornithological Congress, Vancouver, B.C., August 2018). For the purposes of the present study, the inclusion of face patch in the scoring metric does not impact the conclusions, since we are primarily concerned with “pure” parental phenotypes, all of which have the species-typical face patch. The initial scoring was conducted by eye with binoculars and telephoto photography in the field and then confirmed later by closer inspection of photographs. Audio recordings were taken in the field with a Sennheiser microphone and Marantz PMD661 digital recorder with a sample rate of 44.1 kHz. Photographs were taken with either a Fujifilm FinePix S700 or a Canon SL1 camera body with either a Canon 75300 mm lens or a Canon 70-300 mm L lens in .jpg or .RAW format. Playback was conducted with an Apple iPhone 4 and a JBL Flip 2 portable speaker connected with a miniplug cable. Volume level was left consistent for all playbacks. If an individual could not be scored by plumage, the audio recordings for that individual were left out of all analyses. 4.3.1 Broad-level acoustic analysis In addition to the methods described in Chapter 3 of this dissertation, our recordings from the Olympic hybrid zone (described in Rohwer and Wood 1998) taken in 2015 and 2016 were compared to song recordings from public databases (xeno-canto.org and the Macaulay Library). All songs from these databases that met standards of time of year (after spring and before fall migration: defined as May 15th to August 8th), acoustic quality, and validity were included, resulting in a sample of 141 Townsend’s warbler and 46 hermit 81 warbler songs. JL visually inspected spectrograms of each song and categorized them as either “multiple-note introductory syllable” or as “single note introductory syllable” songs. We briefly report the results of our qualitative categorical song study in the introduction to provide context for our playback study. The detailed results of the qualitative song study are as follows: Of the 43 Townsend’s warbler samples from the Macaulay Library, 2 songs, or 5% of the total, from allopatric populations used multinote introductory syllables: one near Glacier Bay in Alaska and one on Vancouver Island, British Columbia. There were 5 other samples in the Macaulay Library that used multinote introductory syllables, but these samples did not meet standards for inclusion. Additionally, of the 98 Townsend’s warbler song samples in the online citizen-science nature sound repository xeno-canto.org, 21 used multi-note introductory syllables. Of those 21, 14 were from Vancouver Island, immediately to the north of the Olympic hybrid zone, 5 were from Gwaii Haanas Island in extreme western British Columbia, 1 was from Montana, and 1 was from Central B.C. Multi-note introductory syllables are not common in allopatric populations of Townsend’s warblers that are removed from hybrid zones. Overall, 4% of published recordings of the inland populations (East of the Coast Range) and ~45% of recordings from coastal populations (West of the Coast Range) of Townsend’s warblers use multi-note introductory syllables. The bulk of the coastal multi-note syllables are from Vancouver Island, again, immediately to the North of the Olympic hybrid zone. It is possible that this pattern results from dispersal from the hybrid zone to Vancouver Island. Hermit warblers also appear to show an interesting trend in the morphology of their introductory syllables. In an allopatric population of hermit warblers in northern Oregon, 82 all encountered individuals sang songs with multinote introductory syllables (n=12 individuals recorded by JL in 2016). Out of 46 recordings of Hermit Warbler song from the Macaulay library, 42 (91.3%) use multinote introductory syllables. The 4 songs (8.7%) that use single-note introductory syllables are from the vicinity of Mount Hood in Oregon, the location of another hybrid zone (Rohwer and Wood 1998). 4.3.2 Playback experiment: male response The playback experiment was conducted in May-July of 2016, 2017 and 2018. Locations were determined through a roving approach of the entire study area. Individuals were located from a distance by their song and then tracked to their perch. After an individual was located and its song recorded as above, playback was initiated. The playback song used was randomly selected from a suite of songs recorded previously which were selected for features described in the following text. Prior to use in the experiment, these songs were edited in PRAAT software to remove any background noise with a bandpass filter and spectral subtraction, using a window length of 0.025 seconds and a smoothing bandwidth of 1.0 Hz across the entire frequency range that was not filtered out with the band-pass filter. This was done to preserve acoustic integrity as much as possible while removing the potential effects of background noise, thus ensuring a high signal-tonoise ratio. After filtering, mean intensity over the duration of the song was scaled to 75 db. Peak intensity was not normalized in order to preserve any potentially informative differences between songs. Eight songs were used for playback (see Figure 4.1): four of which were produced by individuals from the hybrid zone scored as Townsend’s warblers, four of which were produced by individuals from the hybrid zone scored as hermit 83 warblers. Of each set of four, half were designated as having multinote introductory syllables, while the other half were designated as having single-note introductory syllables. Songs to include in this suite were first selected by their meeting the qualifications above and then by recording quality. Two control playback songs of locally common birds, one of a dark-eyed junco (Junco hyemalis) and one of an American robin (Turdus migratorius), were included in the song suite to test for the possible effect of response by the focal species relying in part on a general disturbance response. At no time did the controls elicit a measurable or noticeable response from the target bird, so these controls will not be included in analyses or discussed further. These species live in the canopy of trees. Playback was initiated under the perched bird, immediately adjacent to the perch tree, in an exact spot selected for the availability of perches at continuously decreasing distances from the playback speaker, which was placed on the ground. Next to the speaker was placed a clay mount, one side being painted with the plumage pattern of a Townsend’s warbler, the other side being painted with the plumage pattern of a hermit warbler. This was done to provide a “target” for the responding bird while effectively randomizing any potential effect of plumage type on response. To make sure that this mount did not unduly influence the results, we conducted a trial wherein the same individuals were tested twice, once with each half of the mount covered. This test did not reveal a clear effect of mount type on the response measures used in this study. Care was taken to avoid placing the mount so that the most prevalent perches would favor display of one side over the other. The use of a mount focused the responding bird’s attention to one point (the mount was even physically attacked several times) but avoided the potential of a specific coloration, rather than the song features, affecting the response. 84 In trials where we did not use a clay mount, playback also elicited strong responses, but the responding bird often flew around the forest floor, searching a wide area for the intruder, making it difficult to quantify the response accurately. In previous research that used real mounted skins, researchers found that in allopatry, hermit males were more aggressive to hermit mounts than to Townsend’s mounts, but Townsend’s males did not differ in aggression towards mounts of either plumage phenotype (Pearson and Rohwer 2000). This difference is likely explained by our use of a painted clay mount rather than a feathered skin, which was purposeful. In order to specifically address our question of the role of song in eliciting aggressive interactions, we did not include measures of direct physical aggression in our response PCA (i.e., pecks, hits, and wing flicks, which contributed heavily to Pearson and Rohwer’s response PC1) and are behaviors likely to be more strongly induced by close-range signals such as plumage coloration than by longrange signals such as song. Ten meter, 3 meter, and 1 meter distances from the speaker were noted by the experimenter. Naturally available objects were used as markers, since a preliminary trial showed that birds targeted brightly-colored or foreign objects (like flagging). Notably, these distances directly above the speaker were measured with a pre-measured light weight pole, which was removed before playback began. With the experimenter standing over 10 meters away in a location that allowed for observation of the full study area, playback was commenced. A song was played at a rate of once every 3.5 seconds for 10 minutes. During playback, response distance category (>10m, <10m, <3m, <1m) from the playback speaker was recorded continuously, as was duration of time spent <1m from the playback speaker. After the playback period was 85 complete, song recordings and photographs were taken. Many individuals were tested a second time with a different playback song a minimum of 30 minutes after the end of the first playback period, an amount of time which trials showed was sufficient for the focal bird to resume normal behavior. Order of playback response was recorded (i.e., 1st or 2nd playback response) in the field and was subsequently included as a random effect in analyses. The summary statistic used to represent overall response in subsequent analyses was obtained by using the first principal component values after conducting a PCA (prcomp function in R) that included the nearest response distance (PC1 loading: -0.57), duration of time spent <1m from the playback speaker (0.54), and the duration of time between the start of playback and the closest response (-0.62). PC1 explained 66% of the variance. 4.3.3 Playback experiment: female response Since they are more difficult to detect and locate in the field, we could not reliably target females in the same way that we did males. Still, whenever we had direct confirmation that a female was in the immediate vicinity of the male-directed playbacks described above, we recorded their behavior as well. We categorized their response more broadly as close (less than 3 meters from the playback speaker), far (over 3 meters away, but clearly altering their behavior by approaching the playback speaker), and ignore (did not clearly alter their behavior by approaching the playback speaker). Sample sizes were much lower than for males (n=32 total female responses recorded; 10 hermit, 6 hybrid, and 16 Townsend’s). 86 4.3.4 Acoustic analysis Analysis of the acoustic morphology of syllables was performed using the warbleR (Araya-Salas and Smith Vidaurre 2017) package in R (R Core Team 2017). Syllables were selected using the “manualoc” and “seltailor” functions, then the “specan” function was used to measure acoustic features of syllables and all measurements were exported in a .csv file. The “trackfreqs” function was used to visualize dominant frequency tracking, and all measured values were inspected and spot-checked to ensure accuracy. A principal component analysis (PCA) was run in R (“prcomp” function from “stats” package) and included the following measured values, using the naming scheme according to warbleR: duration, meanfreq, freq.median, freq.Q25, freq.Q75, freq.IQR, time.median, time.Q25, time.Q75, time.IQR, skew, kurt, sp.ent, time.ent, entropy, sfm, meandom, mindom, maxdom, dfrange, modindx, dfslope, meanpeakf. PC1 explained 98.5% of the variance and PC2 explained 0.8% of the variance in the data. The top 5 PC1 loadings, in descending order of loading values, were: kurtosis (kurt, 9.98e-01), skewness (skew, 5.04e-02), modulation index (modindx, -3.42e-02), slope of the change in the dominant frequency (dfslope, 1.18e-02), and dominant frequency range (dfrange, -6.68e-03) (Figure 4.2). Analysis of the rhythm of songs was performed using a custom protocol. Using the syllable start time and syllable duration from the warbleR output table, summary features of each song were computed in R. The rhythmic features that were recorded are: mean syllable length, mean silent period length, syllable rate (number of syllables/song length), syllable regularity, and silent period regularity. Syllable regularity is a normalized measure of consistency of syllable duration. This measure was computed by summing the concatenation of the proportion of total sound of each syllable (syllable length divided by 87 the sum of all syllable lengths), normalized to the maximum syllable length. The resulting value was then divided by the number of syllables to yield the syllable regularity. The same procedure was utilized to produce the silent period regularity, substituting silent period lengths for syllable lengths. Then, a PCA was performed, as above. PC1 explained 98% of the variance and PC2 explained 1.98% of the variance in the data. The top PC1 loadings were syllable rate (0.966), silence regularity (-0.18), and syllable regularity (-0.17) (Figure 4.2). In a separate analysis, we also used the song rhythm and syllable morphology scoring from Chapter 3 of this dissertation to assess differential response to songs. This allowed us to gauge the response to song features relevant to the variation observed across the entire study area. As discussed in the results below, by chance, syllable morphology LD1 scores of the playback songs were not representative of that observed across the full range of the study area, so we were unable to conduct our full analysis using this feature. In contrast, rhythm LD1 scores fell approximately continuously across the range of scores found in the full population of songs, so we conducted our full analysis using rhythm LD1 scores. Given the density distribution of song features observed in Chapter 3, we would not necessarily expect a linear relationship between response and LD1 scores. Therefore, we used t-tests comparing response between positive and negative LD1 scores instead of linear regressions. 4.4 Results Acoustic analysis shows that the playback songs clustered better by introductory syllable morphology category than they did by plumage-based species identity category 88 (Figure 4.2). If individuals of these species use introductory syllable morphology as a cue in territorial disputes, then we expect them to respond to the playback songs differently according to this feature. Likewise, if song rhythm is used, then we expect to see differential response according to these features. We found that categorical parameters (single/multiple introductory syllable morphology and species identity of the singer) did not significantly affect response, but song rhythm did. 4.4.1 Statistical analysis: categorical parameters A linear mixed-effects model was constructed to assess the relationship between male response to song and the following explanatory variables: species identity of the responding bird, allopatry/sympatry status, introductory syllable morphology category of the playback song, and species identity of the playback song singer (“lmer”, lme4 package, p-values from Satterthwaite's method). Only parent species male responses were considered in these analyses. Order of playback was included as a random effect. Species identity of the responding bird showed a significant relationship (df=169.24, t=3.18, P=0.0018), as did allopatry status (df=169.28, t=3.80, P=0.00020). Interestingly, neither syllable morphology category nor species identity of the singer of the playback song had a significant effect on response (df=169.26, t=-0.96, P=0.34 and df=169.06, t=1.56, P=0.12 respectively). To further characterize these relationships, additional statistical analyses were conducted. Response to playback song was lower in the hybrid zone compared to that found in allopatric populations, and this trend appears to be heavily driven by a reduced response to song by Townsend’s warblers in the sympatric population. When accounting for the 89 additional random effects of allopatry status and both playback-song-associated variables, species identity showed a significant relationship with response such that hermit warblers respond less overall than Townsend’s warblers do (lmer, df=165.27, t=3.061, P=0.0026). When analyzed independently, hermit warbler response did not appear to be impacted by allopatry or playback song variables (lmer, species ID: df=46.64, t=1.56, P=0.13; syllable morphology: df=47.00, t=1.378, P=0.175; allopatry: df=46.13, t=1.31, P=0.20), while Townsend’s warbler response showed a significant relationship with allopatry status (df=119.57, t=3.81, P=0.00023) and a weakly significant effect of introductory syllable morphology (df=119.01, t=-2.00, P=0.048), but this significance level decreased when including allopatry status as a random effect (lmer, df=120.14, t=-1.87, P=0.065). Hybrid males did not respond significantly differently to hermit versus Townsend’s warbler song (lmer, hybrid data only with fixed effects for both playback song variables and order as a random effect, df=34.18, t=0.74, P = 0.46), and, like the parent species males, showed no significant effect of syllable morphology on response (same lmer, df=34.69, t=-1.71, P=0.097). Female response to song did not differ significantly by introductory syllable morphology category (pooled data for females of all species, Welch two sample t-tests: t=1.27, df=29.97, P=0.21) or by species identity of the playback singer (t = -1.74, df = 28.87, p-value = 0.09). Sample sizes were too low to reliably compare female response in sympatric versus allopatric regions. 90 4.4.2 Statistical analysis: learned versus innate song components, male and female response Our playback songs did not differ in average syllable morphology PC1 score in a way that would allow us to assess response to syllable morphologies that span the full range found in allopatric populations; all playback songs had negative average syllable morphology PC1 scores. This is a byproduct of the selection criteria for our suite of playback songs, which were established to test categorical variables, as above. By chance, we did not select songs that were representative of the full range of syllable morphologies in songs across the hybrid zone. This does, however, allow us to assess other aspects of song without potential artifacts induced by differences in average syllable morphology. Our playback songs ranged in rhythm PC1 scores somewhat continuously between -1.74 and 1.83 (-1.74, -0.91, -0.76, -0.011, 0.35, 1.00, 1.49, 1.83). This allowed us to assess response according to playback song rhythm. In allopatric populations, parent species males did not respond significantly differently to songs with positive versus negative rhythm PC1 scores (Welch two-sample t-tests; t=1.33, df=77.60, P=0.19). In sympatry, we also found no significant difference in parent species male response to songs with negative versus positive rhythm PC1 scores (t=1.20, df=89.31, P= 0.23, Figure 4.3a). In contrast, female response showed a different trend: higher response to positive rhythm PC1 scores and lower response to negative rhythm PC1 scores (Welch two-sample t-test: t=2.30, df=26.98, P=0.030; linear regression: adjusted R-squared=0.15, df=30, P=0.017). Tests showed this trend was apparent in Townsend’s females (t=-3, df=9, P=0.015), but not hermit females (t=0, df=8, P=1), though sample sizes were very low. The highest female response magnitude was found for playback song with rhythm LD1 score 91 of 1.83 (Figure 4.3b). Hermit, Townsend’s, and hybrid warbler females responded to song playback. Seasonal timing also affected female response: response was higher later in the summer season (multiple regression; rhythm LD1: t=2.29, df=30, P=0.030; Julian day: t=3.12, P=0.0040), though it should be noted that no females were detected after day 168 (approx. June 16th). However, the low sample size of female responses requires cautious interpretation; local singing or social conditions may play a role in female response (Figure 4.4). 4.5 Discussion Within the hybrid zone, females responded more strongly to Townsend’s-like song rhythms (positive LD1) than they did to hermit-like song rhythms (negative LD1). This finding suggests that sexual selection has driven the introgression of Townsend’s-like song rhythm into the hermit warbler population, which was revealed in Chapter 3. In contrast, no significant difference was detected in the response by individuals to aspects of the acoustic morphology of song introductory syllables that are divergent in allopatric populations. Pearson (2000) hypothesized that female preference for Townsend’s warbler traits was driving hybridization. While low sample size places a limit on the achievable level of confidence, our results provide support for this hypothesis and should encourage further research. The level of sexual selection evident in our experiment is in line with that which would be expected by the cline width for song rhythm. Rhythm score had a significant but overall moderate effect on female response. This can be interpreted as reflecting a moderate level of selection on song rhythm, as predicted by cline shape in Chapter 3. Likewise, the 92 cline width for syllable morphology from Chapter 3 was exceedingly large, suggesting very low rates of selection for syllable morphology, just as we see in our playback response experiment here. Sexual selection is thought to be a major driver of phenotypic change and important to the process of speciation (Andersson 1994, Coyne and Orr 2004, Servedio and Boughman 2017). In birds, sexual selection has been implicated in the evolution of vocal signals, including song (Searcy and Andersson 1986), and song is said to be important for mating decisions in females (Searcy and Yasukawa 1996, Collins 2004). While it is important to take into account the evidence that female preference may be experience dependent (Riebel 2000, Lauay et al. 2004, Hauber et al. 2013), there is also ample evidence that songs convey consistent information to potential mates. In relation to our study, it is important to note that several laboratory tests have shown that females prefer males who exhibit signs of good song learning ability (e.g., Nowicki et al. 2002, Spencer et al. 2005) and that frequency-related song components are important in female choice (e.g., Tomaszycki and Adkins-Regan 2005). However, song rhythm also seems to be important; one study found that rhythm-related components of song were more important than frequency-related components in guiding female response (Searcy et al. 2010), and another found that rhythm-related components were predictive of female preference (Holkveck and Riebel 2007). Still, despite experiments investigating female preference related to frequency bandwidth/trill rate ratio (e.g., Draganoiu et al. 2002, Ballentine et al. 2004), there have been few studies that directly investigate female preferences for song rhythm. Our research adds to this body of literature by attempting to discern between the 93 forces that act to induce trait change via cultural versus genetic mechanisms. In our hybrid system, the feature of song most obviously divergent between the two parent species (syllable morphology, which is probably learned) does not show evidence of being under strong sexual selection – at least by our metric. Rather, a song feature that is comparatively subtle to the human observer (song rhythm, which is unlikely to be learned) shows direct evidence of being under sexual selection via female mate choice, which has apparently led to introgression and a dramatic phenotypic shift for one species. The convergence in syllable morphology in this system could be driven by the hypotheses laid out in Chapter 3. These results demonstrate that the most apparent acoustic convergence of song in the hybrid area, which is more likely to be due to heterospecific song learning than to genetic change, does not appear to be of perceptual relevance in regard to the perceived territorial threat or mating suitability. Thus, cases of convergence or divergence of acoustic features of signals between closely related species in sympatry and allopatry may lead to the false conclusion that these vocal features also play a role in determining the nature of interspecific interactions during secondary contact. Instead, selection may be acting more strongly on more subtle aspects of song that are under stricter genetic control; these song features may show alternate trends that reflect interspecific interactions and population dynamics during periods of secondary contact. Pearson (2000) found asymmetrical aggression between hermit and Townsend’s warblers. In our study, we also found that Townsend’s warblers were more aggressive than hermit warblers, but that this difference was only evident in allopatric populations. In the center of the hybrid zone, Townsend’s warblers were significantly less aggressive than in 94 allopatry. Notably, hermit warblers in the hybrid zone also exhibited lower levels of aggressive response to song than their allopatric counterparts, but this difference was not significant. The resulting pattern is that the two species did not differ in their aggressive response to song playback in sympatric populations, where hybridization occurs. While the reduced territorial aggression observed by one parent species in the hybrid zone compared to that found in allopatric populations suggests that hybridization coincides with lower territorial aggression in this system, causality cannot be determined from our data. It is possible that, as discussed in Chapter 3 of this dissertation, convergence in song represents selection against hybridization through an increase in the efficiency of territorial defense by males, which could result in lower rates of physically aggressive encounters. Sill, a more direct test of this hypothesis is necessary. Indeed, we found that song that lacked convergent, Townsend’s-like rhythm elicited lower levels of aggression than that elicited by Townsend’s-like rhythm (Figure 4.3a), opposite of the expectation of the convergent agonistic character displacement hypothesis (Grether et al. 2017). Overall, this study casts serious doubt on the notion that vocal learning accelerates speciation through rapid cultural changes in song. In recently diverged species, learned aspects of song may not function any more as a barrier than do the obvious morphological differences in plumage coloration between these taxa. Why hybridization occurs despite these differences therefore remains puzzling. 4.6 References Ali F, Otchy TM, Pehlevan C, Fantana AL, Burak Y, Ölveczky BP (2013) The basal ganglia is necessary for learning spectral, but not temporal, features of birdsong. Neuron 80(2):494–506 doi:10.1016/j.neuron.2013.07.049 95 Araki M, Bandi MM, Yazaki-Sugiyama Y (2016) Mind the gap: Neural coding of species identity in birdsong prosody. Science 354(6317):1282–1287 doi:10.1126/science.aah6799 Andersson MB (1994) Sexual selection. Princeton University Press Araya-Salas M, Smith-Vidaurre G (2017) warbleR: An r package to streamline analysis of animal acoustic signals. Methods Ecol Evol 8:184–191 Ballentine B (2004) Vocal performance influences female response to male bird song: an experimental test. Behav Ecol 15(1):163–168 doi:10.1093/beheco/arg090 Baptista LF, Schuchmann KL (1990) Song learning in the Anna hummingbird (Calypte anna). Ethology 84:15–26 Baptista LF, Trail PW (1992) The role of song in the evolution of passerine diversity. Syst Biol 41:242–247 Bolhuis JJ, Okanoya K, Scharff C (2010) Twitter evolution: Converging mechanisms in birdsong and human speech. Nat Rev Neurosci 11:747–759 Brelsford A, Toews DPL, Irwin DE (2017) Admixture mapping in a hybrid zone reveals loci associated with avian feather coloration. Proc R Soc B 284: 20171106 doi:10.1098/RSPB.2017.1106 Clark CW, Marler P, Beeman K (1987) Quantitative analysis of animal vocal phonology. Ethology 76:101–115 Collins S (2004) Vocal fighting and flirting: The functions of birdsong. in Nature’s Music: the Science of Birdsong (Marler P, Slabbekoorn H, eds). Elsevier, Sand Diego, CA, pp 3979 Coyne JA, Orr HA (2004) Speciation. Sunderland, MA: Sinauer Associates, Inc. Drǎgǎnoiu TI, Nagle L, Kreutzer M (2002) Directional female preference for an exaggerated male trait in canary (Serinus canaria) song. Proc R Soc B Biol Sci. 269(1509):2525–2531 doi:10.1098/rspb.2002.2192 Emlen ST (1972) An experimental analysis of the parameters of bird song eliciting species recognition. Behavior 41:130–171 Grether GF, Peiman KS, Tobias JA, Robinson BW (2017) Causes and consequences of behavioral interference between species. Trends Ecol Evol 32(10):760–772 doi:10.1016/j.tree.2017.07.004 Haavie J, Borge T, Bures S, Garamszegi LZ, Lampe HM, Moreno J, Qvarnström A, Török J, Sætre GP (2004) Flycatcher song in allopatry and sympatry - Convergence, divergence and reinforcement. J Evol Biol 17(2):227–237 doi:10.1111/j.1420-9101.2003.00682.x 96 Hall ML (2009) A review of vocal duetting in birds. Adv Study Behav 40:67–121 Hauber ME, Woolley SMN, Cassey P, Theunissen FE (2013) Experience dependence of neural responses to different classes of male songs in the primary auditory forebrain of female songbirds. Behav Brain Res 243:184–90 doi:10.1016/j.bbr.2013.01.007 Holveck MJ, Riebel K (2007) Preferred songs predict preferred males: Consistency and repeatability of zebra finch females across three test contexts. Anim Behav 74(2):297–309 doi:10.1016/j.anbehav.2006.08.016 Janik VM, Slater PJB (1997) Vocal learning in mammals. Adv Study Behav 26:59–100 Kenyon HL, Toews DPL, Irwin, DE (2011) Can song discriminate between MacGillivray’s and mourning warblers in a narrow hybrid zone? Condor 113:655–663 Lachlan RF, Servedio MR (2004) Song learning accelerates allopatric speciation. Evolution 58:2049–2063 Lauay C, Gerlach NM, Adkins-Regan E, Devoogd TJ (2004) Female zebra finches require early song exposure to prefer high-quality song as adults. Anim Behav 68(6):1249–1255 doi:10.1016/j.anbehav.2003.12.025 Lipkind D, Zai AT, Hanuschkin A, Marcus GF, Tchernichovski O, Hahnloser RHR (2017) Songbirds work around computational complexity by learning song vocabulary independently of sequence. Nat Commun 8(1) doi:10.1038/s41467-017-01436-0 Marler P, Sherman V (1983) Song structure without auditory feedback: Emendations of the auditory template hypothesis. J Neurosci 3(3):517–531 Mason NA, Burns KJ, Tobias JA, Claramunt S, Seddon N, Derryberry EP (2017) Song evolution, speciation, and vocal learning in passerine birds. Evolution 71:786–796 Morrison ML, Hardy JW (1983) Hybridization between hermit and Townsend’s warblers. The Murrelet 64:65–72 Nelson DA (2017) Geographical variation in song phrases differs with their function in white-crowned sparrow song. Anim Behav 124:263–271 Nowicki S, Searcy W (2014) The evolution of vocal learning. Curr Opin Neurobiol 28:48– 53 Nowicki S, Searcy W, Peters S (2002) Quality of song learning affects female response to male bird song. Proc Biol Sci 269(1503):1949–54 doi:10.1098/rspb.2002.2124 Nottebohm, F (1972) The origins of vocal learning. Am Nat 106:116–140 Odom KJ, Hall ML, Riebel K, Omland KE, Langmore NE (2014) Female song is widespread and ancestral in songbirds. Nat Commun 5:3379 97 Olofsson H, Frame AM, Servedio MR (2011) Can reinforcement occur with a learned trait? Evolution 65:1992–2003 Päckert M, Martens J, Kosuch J, Nazarenko A, Veith M (2003) Phylogenetic signal in the song of crests and kinglets (Aves: Regulus). Evolution 57(3):616–629 doi:10.1554/00143820(2003)057 Pearson S (2000) Behavioral asymmetries in a moving hybrid zone. Behav Ecol 11:84–92 Pearson S, Rohwer S (2000) Asymmetries in male aggression across an avian hybrid zone. Behav Ecol 11:93–101 Qvarnström A, Haavie J, Sæther SA, Eriksson D, Pärt T (2006) Song similarity predicts hybridization in flycatchers. J Evol Biol 19:1202–1209 R Core Team (2017) R: A language and environment for statistical computing. Riebel K (2000) Early exposure leads to repeatable preferences for male song in female zebra finches. Proc R Soc B Biol Sci 267(1461):2553–2558 doi:10.1098/rspb.2000.1320 Rohwer S, Wood C (1998) Three hybrid zones between hermit and Townsend’s warblers in Washington and Oregon. Auk 115:284–310 Searcy WA, Andersson M (1986) Sexual selection and the evolution of song. Annu Rev Ecol Syst 17:507–533 Searcy WA, Peters S, Kipper S, Nowicki S (2010) Female response to song reflects male developmental history in swamp sparrows. Behav Ecol Sociobiol 64(8):1343–1349 doi:10.1007/s00265-010-0949-5 Searcy WA, Yasukawa K (1996) Song and female choice. in: Ecology and evolution of acoustic communication in birds (Kroodsma DE, Miller EH, eds). Cornell University Press, Ithaca, NY, 454-473 Secondi J, Bordas P, Hipsley CA, Bensch S (2011) Bilateral song convergence in a passerine hybrid zone: Genetics contribute in one species only. Evol Biol 38:441–452 Secondi J, Bretagnolle V, Compagnon C, Faivre B (2003) Species-specific song convergence in a moving hybrid zone between two passerines. Biol J Linn Soc 80:507– 517 Seddon N, Tobias JA (2007) Song divergence at the edge of Amazonia: An empirical test of the peripatric speciation model. Biol J Linn Soc 90(1):173–188 doi:10.1111/j.10958312.2007.00753.x Seddon N, Tobias JA (2010) Character displacement from the receiver’s perspective: Species and mate recognition despite convergent signals in suboscine birds. Proc Biol Sci 277(1693):2475–83 doi:10.1098/rspb.2010.0210 98 Servedio MR, Boughman JW (2017) The role of sexual selection in local adaptation and speciation. Annu Rev Ecol Evol Syst 48:85–109 doi:10.1146/annurev-ecolsys-110316 Slabbekoorn H, Smith T (2002) Bird song, ecology and speciation. Philos Trans R Soc Lond B Biol Sci 357:493–503 Soha J, Marler P (2000) A species-specific acoustic cue for selective song learning in the white-crowned sparrow. Anim Behav 60(3):297–306 doi:10.1006/anbe.2000.1499 Spector D (1992) Wood-warbler song systems. Curr Ornithol 9:199–238 Spencer KA, Buchanan KL, Leitner S, Goldsmith AR, Catchpole CK (2005) Parasites affect song complexity and neural development in a songbird. Proc R Soc B Biol Sci 272(1576):2037–2043 doi:10.1098/rspb.2005.3188 Thorpe WH (1958) The learning of song patterns by birds, with especial reference to the song of the chaffinch Fringilla coelebs. Ibis 100(4):535–570 Tobias JA, Seddon N (2009) Signal design and perception in Hypocnemis antbirds: Evidence for convergent evolution via social selection. Evolution 63(12):3168–3189 doi:10.1111/j.1558-5646.2009.00795.x Verzijden MN, ten Cate C, Servedio MR, Kozak GM, Boughman JW, Svensson EI (2012) The impact of learning on sexual selection and speciation. Trends Ecol Evol 27(9):511– 519 doi:10.1016/j.tree.2012.05.007 Yeh DJ (2018) The interaction between learning and speciation. (University of North Carolina). PhD thesis Yeh DJ (2019) Assortative mating by an obliquely transmitted local cultural trait promotes genetic divergence: A model. Am Nat 193(1):81–92 doi:10.1086/700958 99 Figure 4.1: Playback songs varied according to the number of notes in the introductory syllables. Songs with multinote introductory syllable on left (a,c,e,g), single-note introductory syllables on right (b,d,f,h). Upper four (a-d) are hermit warbler, lower four (e-h) are Townsend’s warbler. Graphics are representative, showing divergent plumage. Red box in (a) indicates multinote introductory syllables. Blue box in (b) indicates single-note introductory syllables. Photos by JL. 100 Figure 4.2: Results of acoustic analysis of playback songs. Visual assessment suggests that song features cluster better by intro-syllable category than they do by singer identity. a) Acoustic morphology cluster comparisons and b) song rhythm cluster comparisons. Color indicates clustering by k-means (a,b: left panels), or by categorization by morphology of the introductory syllables (a,b: center panels; red = single-note, black = multinote introductory syllables), or by species identity of the singer (a,b: right panels; red = Townsend’s warbler, black = hermit warbler). K means clustering (left) more closely matches introductory syllable morphology clustering (middle) than does species identity clustering (right). PCA loadings in methods. 101 Figure 4.3: Response to song rhythm. a) male overall response in allopatry and sympatry. Mean +/- 95% CI. b) female close responses. Mean +/- 95% CI. Black dots are negative LD1 (hermit-like), red dots are positive rhythm LD1 (Townsend’s-like). 102 Figure 4.4: Map of female strong responses to playback. Most strong female responses to song are female Townsend’s warblers responding to “Townsend’s-like” song rhythm in both allopatry and sympatry. Two strong responses were from female hermit warblers: one in allopatry which responded strongly to hermit-like song rhythm, and one in sympatry which responded to Townsend’s-like song rhythm. CHAPTER 5 CONCLUSION 5.1 Modular Song Learning Through the research presented in this dissertation, I find sufficient supporting evidence for differential reliance on imitative learning of different song features to propose the hypothesis of modular song learning. In Chapter 2, I conducted a comparative analysis of isolate studies using spectrograph-to-sound conversion and found interspecific variation in the degree to which song components ("modules") rely on imitation for normal development. Notably, isolates of most species produce normal song rhythm, which relates to the respiratory pattern associated with song activity. This suggests that development of song rhythm is, in large part, strongly genetically controlled. In Chapter 3, I analyzed song in the hermit-Townsend’s warbler hybrid system with the production-module approach developed in the previous chapter. Rhythm showed a striking clinal pattern of convergence across the hybrid zone, reminiscent of a quantitative trait under moderate selection, while other song features (i.e., syllable morphology) did not. Rather, syllable morphology appeared to be culturally transmitted within and across species limits in a pattern typical of song "neighborhoods" or dialects in a single, large population. This provides further evidence that, in these species, rhythm (i.e., song-associated respiratory pattern) is strongly genetically inherited, while syllable morphology is not. Furthermore, rhythm converged on 104 one species (Townsend's-like rhythm) and syllable morphology on the other species (hermit-like syllables). In Chapter 4, I conducted a playback study in the field. Males did not significantly differ in their response to any measured aspect of song. On the other hand, females responded significantly more strongly to songs with Townsend's-like rhythm than to hermit rhythms, while other aspects of song did not show a relationship with response. This suggests that intersexual selection has contributed to the evolution of song rhythm in the system and that syllable morphology and song rhythm experience different selection pressures, though future research should be conducted to identify female preferences across the hybrid zone with greater geographic coverage that I achieved here. Together, the findings presented here represent strong evidence in support of the distinction between modules in development and therefore production. The evidence that I provide that these modules may be developed through separate and somewhat independent processes is an important insight into the neuromotor systems that underlie avian vocal learning. Because birdsong is a model system for understanding imitative learning, including human speech development, this insight and approach may also be useful in other systems. In addition to studies that find support for the distinction between modules which have been discussed in the above chapters (Marler and Peters 1977; Ali et al. 2013; Araki et al. 2016; Lipkind et al. 2017), there is evidence from other laboratory studies that can be interpreted as supporting the separation between song modules. For example, through cooling the HVC in the brain of zebra finches and canaries, song rhythm can be readily manipulated with only minor changes in syllable acoustic morphology, such that songs have a lower syllable rate when the HVC is at a lower temperature (Long and Fee 2008), 105 up to a certain “breaking point” where syllable rate sharply increases (Goldin et al. 2013). Since in these studies the fine acoustic structure of the syllables was maintained across experimentally manipulated HVC temperatures, the neural input to the syrinx necessary for the production of syllables with consistent acoustic morphology would then originate elsewhere within the brain. Further research has been conducted that supports a neural distinction between production modules, which is discussed in Chapter 2. Still, the neural coding of song rhythm remains contentious, as does the integration between the proposed modules. To what degree the development of production modules is linked may vary between species. Several studies have found that fine control of syllable morphology seems to be affected by sequential organization of syllables within song. For example, in Bengalese finches, syllable morphology and syntax appear to be linked (Wohlgemuth et al. 2010). It is plausible to assume that fine motor control of acoustic structure imposes constraints on syllable sequencing and vice-versa. In addition, physiological and physical limits of sound production may impose a tradeoff between frequency bandwidth of syllables and the rate at which syllables are repeated. This phenomenon has been thoroughly discussed in the context of species whose songs are comprised of trilled elements (e.g., Podos 1996). Since the song-associated respiratory pattern must be integrated with gas exchange requirements (Schmidt and Goller 2016) and since, assuming a single sound source and cyclic gestures controlling driving respiratory pressure and syringeal labial tension (Mindlin and Latje 2006), the time required to change the vibrating frequency of structures within the vocal tract will increase with greater absolute difference in the magnitude of the intersyllable change in frequency, a negative relationship between syllable repetition rate and syllable frequency bandwidth is not 106 surprising. Nonetheless, through this relationship, syllable morphology and song rhythm can be interpreted as being linked in production and potentially in learning. In another example of a link between modules, there is a necessary relationship between rhythm, syntax, and syllable acoustic morphology in species that use syllables of differing durations; for example, the relative frequency of use of long versus short syllables will always affect song rhythm. Indeed, although we suggest that from a production perspective a separate view can be helpful, the proposed modules are, to some extent, inextricably intertwined. Future study should focus on defining the relationship between these proposed modules of avian vocal production and learning. In addition to laboratory-based studies employing a range of techniques, comparative analysis of natural populations also promises to be productive to this end. Using the modular approach to acoustic analysis presented here across a range of taxonomic groups would likely prove fruitful, giving some perspective on the relationship between song modules while at the same time shedding light on the evolution of learned vocal behavior in birds. Integrating findings from comparative studies of natural populations with those from laboratory investigations of the neuromotor mechanisms of song learning and production should allow for a general increase in the value of the scientific contribution of birdsong research. 5.2 The Evolution of Oscine Vocal Learning It has been previously stated that the evolution of vocal production learning can be described as the degradation of genetic control over vocal production (Nottebohm 1972). If oscine song learning is to some degree modular, then it is likely that the evolution of this 107 trait progressed by the degradation of genetic control over each module independently, rather than simultaneously or in concert. If this is the case, then we expect to find a phylogenetic pattern in the strength of genetic control over modules, with major clades showing different features of the song learning program. In Chapter 2, I attempted to identify such a phylogenetic signal, but did not recover one. It is possible that my results are due to extremely high evolutionary lability which would be predicted by gene-culture coevolution (i.e., Feldman and Laland 1996; Lachlan and Feldman 2003) and by a LandeKirkpatrick mechanism that could act on sexual signals (i.e., Prum 2010). However, it is likely that, if a phylogenetic trend does exist, I was unable to detect a phylogenetic signal due to the relatively sparse taxonomic coverage of my sample. A more complete taxonomic coverage would be necessary to elucidate any such pattern, but currently such data are not available. Early-branching oscine clades themselves show diversity in their vocal production, with some, for example the superb lyrebird (Menura novaehollandiae), showing notoriously high variability in all song modules while others, for example the rufous scrubbird (Atrichornis rufescens), show very little variation in acoustic structure of syllables. Even comparing the song of the rufous scrubbird, which consists of a repeated single-note syllable, to that of its congener the noisy scrubbird (Atrichornis clamosus), which exhibits a repertoire of syllable types composed in variable syntax and rhythm, is a testament to the abundance of obstacles present in an investigation of the evolution of song learning. Still, a broad comparative approach investigating variation in song modules using ancestral state reconstruction could prove informative. The most well-established explanation for the evolution of vocal learning in birds is that of sexual selection, where learned vocal behavior allows for increased variety of 108 vocal signals that females prefer (Nowicki and Searcy 2014). However, studies have found conflicting evidence for the role of song learning in mate attraction (literature reviewed in Chapter 4). In Chapter 4, we provide evidence that the components of song most likely to be learned (i.e., syllable morphology) do not appear to impact mating decisions by females. Rather, the component of song most likely to be genetically controlled (i.e., rhythm) appears to play a larger role in guiding female response to song. While it remains possible that our methods did not allow for a sensitive assessment of female mate choice for all song features, if our results are reliable then learned features of song may not be of highest importance for mate choice in these species. Accurate song learning ability has been found to be associated with reproductive success in some species (Nowicki et al. 2002) and learned song features have been proposed to be honest indicators of quality (i.e., the developmental stress hypothesis; Nowicki et al. 1998; Spencer et al. 2003). However, despite a strong publication bias, few systems provide supporting evidence (reviewed in Sewall et al. 2016). In contrast, the acoustic representation of the genetic basis for song is more likely to guide female mating decisions. In Chapter 2, I present evidence that shows that birds raised in isolation produce, for the most part, recognizable species-typical song with some relatively minor differences. We know very little about how important these differences are to the cultural function of song. When tested, conspecific females and males respond to isolate song (Searcy et al. 1985; Searcy and Marler 1987), in some cases to a greater degree than they do to normal song (King and West 1977; West et al. 1979; Kreutzer, Vallet, and Nagle 1996). We would not expect such results if learned song was key for guiding mating decisions. While there are undoubtedly many cases where females use learned components of song to guide 109 mating decisions, the importance of the genetic guidance of song development is likely to be universally greater in terms of mate attraction. Furthermore, the evidence in support of the aggressive function of learned song is at least as bountiful as the evidence in support of the mate choice function of learned song, especially beyond a basic level of mate location (e.g., Johnson and Searcy 1996). Due in part to the fact that singing behavior proceeds long after mating pairs have been established in many species (Collins 2004), in the course of a male songbird’s life, the intended receiver of the majority of songs will be males, not females, though I am unaware of any study that has quantified directed singing effort in this way. In addition, the role of song in aggressive and territorial interactions has been well described (Searcy and Beecher 2009). Notably, the phenomenon of song type matching, where neighboring males of some species negotiate territorial disputes through an organized system of song type use, provides a clear adaptive advantage of increasing the vocal repertoire through song learning (Akçay et al. 2013; Beecher 2017). If studies on other taxa confirm our findings that learned components of song are not the most important aspects for guiding mating decisions, then it is very possible that song learning did not evolve exclusively in a mate choice context, as is proposed in the sexual selection hypothesis (e.g., Nowicki and Searcy 2014). In light of this, I have presented in Chapter 3 a possible explanation for the striking pattern of convergence in syllable morphology within the hermit-Townsend’s warbler hybrid zone. My proposal involves the interaction of simultaneous selective pressures for one species to diverge (via intersexual selection or reinforcement) and the other species to converge (via intrasexual selection) in trait values, resulting in the antagonistic coevolution of the trait which is 110 reminiscent of an “evolutionary arms race.” This combination of pressures selects for very rapid trait evolution in both species involved. It is possible that the learned features of song in this system evolved in this context, and this explanation may extend to the origin of oscine song learning. The interaction between selective forces that bird song experiences have likely been important in the course of evolution of song learning in oscines. Still, there remains a host of other expressions of vocal learning within the oscines, about which very little is known. Whether song learning was coopted from the vocal learning process that originally evolved to facilitate vocal production learning of non-song vocalizations (i.e., Sewall et al. 2016) or vice-versa remains a question that deserves further study. Above all else, I provide in this dissertation evidence that song learning is relatively constrained in the species that have been well-studied and that such constraint is evident in natural populations experiencing selection for song trait change. Investigating other forms of vocal learning in taxonomically diverse passerine groups is likely to provide important evidence regarding the origin and evolutionary history of vocal learning in birds. 5.3 Learning and Speciation I did not find evidence that supports the hypothesis that learned song facilitates stronger or more rapidly evolving reproductive barriers compared to song that is not learned (i.e., Lachlan and Servedio 2004). While it remains possible that my assay was not sensitive enough to capture the effect of learned song features on female mate choice that could possibly be based on some unmeasured component, the results from my study do not provide support for the hypothesis that learned song traits feature in guiding mate choice 111 decisions. Rather, features of song that are unlikely to be learned appear to be more important in mate choice and thus divergence in such features is more likely to contribute to reproductive isolation between two incipient species. It is possible that, through the processes of gene-culture coevolution or genetic assimilation, learned song features could become genetically fixed in a population and thus learning could facilitate strong reproductive barriers in this manner (Waddington 1953; Feldman and Laland 1996; Lachlan and Feldman 2003). Recent theoretical work argues in support of this possibility, suggesting that, while reinforcement is unlikely to act directly on learned traits, statistical correlations between learned and genetically-controlled traits may allow for divergence in learned song to promote reproductive isolation between species (Yeh 2019). However, very little is currently known about this process and the timescale at which it might act in nature. Promising directions for future research are (1) to directly investigate the relative contribution of learned and nonlearned components of song to reproductive isolation in other systems and (2) to thoroughly explore the process of genetic assimilation of learned song traits using empirical data. When these lines of research have been conducted, our understanding of the contribution of song learning to the process of speciation will be more complete. 5.4 References Akçay Ç, Tom ME, Campbell SE, Beecher MD (2013) Song type matching is an honest early threat signal in a hierarchical animal communication system. Proc R Soc B 280(20122517) Ali F, Otchy TM, Pehlevan C, Fantana AL, Burak Y, Ölveczky BP (2013) The basal ganglia is necessary for learning spectral, but not temporal, features of birdsong. Neuron 80(2):494–506 doi:10.1016/j.neuron.2013.07.049 Araki M, Bandi MM, Yazaki-Sugiyama Y (2016) Mind the gap: Neural coding of species 112 identity in birdsong prosody. Science 354(6317):1282–1287 doi:10.1126/science.aah6799 Beecher MD (2017) Birdsong learning as a social process. Anim Behav 124:233–246 doi:10.1016/j.anbehav.2016.09.001 Collins S (2004) Vocal fighting and flirting: The functions of birdsong. In: Nature’s Music. San Diego, CA: Elsevier pp 39-79 Feldman MW, Laland KN (1996) Gene-culture coevolutionary theory. Trends Ecol Evol 11(11):453–457 doi:10.1016/j.cub.2008.02.055 Goldin MA, Alonso LM, Alliende JA, Goller F, Mindlin GB (2013) Temperature induced syllable breaking unveils nonlinearly interacting timescales in birdsong motor pathway. PLoS One 8(6) doi:10.1371/journal.pone.0067814 Johnson LS, Searcy WA (1996) Female attraction to male song in house wrens (Troglodytes aedon). Behaviour 133(5/6):357–366 King AP, West MJ (1977) Species identification in the North American cowbird: Appropriate responses to abnormal song. Science 195(4282):1002–1004 doi:10.1126/science.841321 Kreutzer M, Vallet E, Nagle L (1996) Female canaries display to songs of early isolated males. Experientia 52(3):277–280 Lachlan RF, Feldman MW (2003) Evolution of cultural communication systems: The coevolution of cultural signals and genes encoding learning preferences. J Evol Biol 16(6):1084–1095 doi:10.1046/j.1420-9101.2003.00624.x Lachlan RF, Servedio MR (2004) Song learning accelerates allopatric speciation. Evolution 58(9):2049–2063 Lipkind D, Zai AT, Hanuschkin A, Marcus GF, Tchernichovski O, Hahnloser RHR (2017) Songbirds work around computational complexity by learning song vocabulary independently of sequence. Nat Commun 8(1) doi:10.1038/s41467-017-01436-0 Long MA, Fee MS (2008) Using temperature to analyse temporal dynamics in the songbird motor pathway. Nature 456(7219):189–194 doi:10.1038/nature07448 Marler P, Peters S (1977) Selective vocal learning in a sparrow. Science 198(4316):519– 521 Mindlin GB, Laje R (2006) The physics of birdsong. Verlag Berlin Heidelberg: Springer Nottebohm F (1972) The origins of vocal learning. Am Nat 106(947):116–140 Nowicki S, Peters S, Podos J (1998) Song learning, early nutrition and sexual selection in songbirds. Am Zool 38(1):179–190 doi:10.1093/icb/38.1.179 113 Nowicki S, Searcy W (2014) The evolution of vocal learning. Curr Opin Neurobiol 28:48– 53 doi:10.1016/j.conb.2014.06.007 Nowicki S, Searcy W, Peters S (2002) Quality of song learning affects female response to male bird song. Proc Biol Sci 269(1503):1949–54 doi:10.1098/rspb.2002.2124 Podos J (1996) Motor constraints on vocal development in a songbird. Anim Behav 51:1061–1070 doi:10.1006/anbe.1996.0107 Prum RO (2010) The Lande-Kirkpatrick mechanism is the null model of evolution by intersexual selection: Implications for meaning, honesty, and design in intersexual signals. Evolution 64(11):3085–3100 doi:10.1111/j.1558-5646.2010.01054.x Schmidt MF, Goller F (2016) Breathtaking songs: Coordinating the neural circuits for breathing and singing. Physiology 31(6):442–451 doi:10.1152/physiol.00004.2016 Searcy W, Marler P, Peters S (1985) Songs of isolation-reared sparrows function in communication, but are significantly less effective than learned songs. Behav Ecol Sociobiol 17(3):223–229 doi:10.1007/BF00300140 Searcy WA, Beecher MD (2009) Song as an aggressive signal in songbirds. Anim Behav 78(6):1281–1292 doi:10.1016/j.anbehav.2009.08.011 Searcy WA, Marler P (1987) Response of sparrows to songs of deaf and isolation-reared males: Further evidence for innate auditory templates. Dev Psychobiol 20(5):509–519 Sewall KB, Young AM, Wright TF (2016) Social calls provide novel insights into the evolution of vocal learning. Anim Behav 120:163–172 doi:10.1016/j.anbehav.2016.07.031 Spencer KA, Buchanan KL, Goldsmith AR, Catchpole CK (2003) Song as an honest signal of developmental stress in the zebra finch (Taeniopygia guttata). Horm Behav 44(2):132– 139 doi:10.1016/S0018-506X(03)00124-7 Veit L, Aronov D, Fee MS (2011) Learning to breathe and sing: Development of respiratory-vocal coordination in young songbirds. J Neurophysiol 106(4):1747–1765 doi:10.1152/jn.00247.2011 Waddington CH (1953) Genetic assimilation of an acquired character. Evolution 7(2):118– 126 West MJ, King AP, Eastzer DH, Staddon JE (1979) A bioassay of isolate cowbird song. J Comp Physiol Psychol 93(1):124–133 doi:10.1037/h0077577 Wohlgemuth MJ, Sober SJ, Brainard MS (2010) Linked control of syllable sequence and phonology in birdsong. J Neurosci 30(39):12936–12949 doi:10.1523/JNEUROSCI.269010.2010 114 Yeh DJ (2019) Assortative mating by an obliquely transmitted local cultural trait promotes genetic divergence: A model. Am Nat 193(1):81–92 doi:10.1086/700958