Femur 2-4 Tibia 2 Fibula 1 Patella Pretibial bone Tarsals Metatarsal I 8 Phalanx 1 9 Phalanx 2 9 Metatarsal II 4 Phalanx 1 8 Phalanx 2 9 Phalanx 3 9 Metatarsal III 2 Phalanx 1 8 Phalanx 2 9 Phalanx 3 9 Phalanx 4 9 Metatarsal IV 2 Phalanx 1 8 Phalanx 2 9 Phalanx 3 12 Phalanx 4 11 Phalanx 5 9 If an element is unnumbered, it was unossified in all specimens examined. If two numbers are given, these represent the range of ranks over which a variable element can ossify.

282 Evolution of Avian Ossification Sequences

Erin E. Maxwell

Department of Biology

McGill University, Montreal

April 2008

A thesis submitted to McGill University in partial fulfillment of the requirements of the degree of Doctor of Philosophy

© Erin Maxwell 2008 Library and Archives Bibliotheque et 1*1 Canada Archives Canada Published Heritage Direction du Branch Patrimoine de ('edition

395 Wellington Street 395, rue Wellington OttawaONK1A0N4 Ottawa ON K1A0N4 Canada Canada

Your file Votre reference ISBN: 978-0-494-53317-8 Our file Notre reference ISBN: 978-0-494-53317-8

NOTICE: AVIS:

The author has granted a non­ L'auteur a accorde une licence non exclusive exclusive license allowing Library and permettant a la Bibliotheque et Archives Archives Canada to reproduce, Canada de reproduire, publier, archiver, publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public communicate to the public by par telecommunication ou par I'lntemet, prefer, telecommunication or on the Internet, distribuer et vendre des theses partout dans le loan, distribute and sell theses monde, a des fins commerciales ou autres, sur worldwide, for commercial or non­ support microforme, papier, electronique et/ou commercial purposes, in microform, autres formats. paper, electronic and/or any other formats.

The author retains copyright L'auteur conserve la propriete du droit d'auteur ownership and moral rights in this et des droits moraux qui protege cette these. Ni thesis. Neither the thesis nor la these ni des extraits substantias de celle-ci substantial extracts from it may be ne doivent etre imprimes ou autrement printed or otherwise reproduced reproduits sans son autorisation. without the author's permission.

In compliance with the Canadian Conformement a la loi canadienne sur la Privacy Act some supporting forms protection de la vie privee, quelques may have been removed from this formulaires secondaires ont ete enleves de thesis. cette these.

While these forms may be included Bien que ces formulaires aient inclus dans in the document page count, their la pagination, il n'y aura aucun contenu removal does not represent any loss manquant. of content from the thesis.

1*1 Canada ACKNOWLEDGMENTS

Discussion and Support I would like to thank my supervisor Hans Larsson, my supervisory committee (Ehab Abouheif and Bob Carroll) and the members of the Larsson- Carroll lab for discussion. Special thanks to Audrey Heppleston, Maria de Boef and Sylvie Tissandier for assisting with the lab work, and to Luke Harrison and Nadia Frobisch for commenting on early drafts of Chapter 1. Luke Harrison also provided essential assistance with programming issues. Matt Vavrek and Maria de Boef helped with proofreading, and Julie Rousseau translated the abstract. The comments of anonymous reviewers improved Chapters 3, 4 and 5.

Collections Access I appreciate the efforts of Virginie Millien and Anthony Howell for accessioning the Redpath Museum specimens for me. I would also like to thank Kristof Zyskowski, Richard Prum and the other Peabody Museum vertebrate zoology staff for their hospitality and allowing me access to the collections to collect the data used in Chapter 2. Thanks to Parker Cane for access to the cleared and stained Common Tern specimens, as well as Paul Sweet, Joel Cracraft and other members of the AMNH Department of Ornithology for allowing me access to the collections to assemble the data presented in Chapter 5. Thanks to David Evans for allowing me access to the ROM collections for comparative purposes.

Specimen Collection I would like to thank Dr. Katherine Mehl and the members of the Ducks Unlimited NL/Labrador field team (2004/2005) for collecting Common Eider embryos for me. Thanks also to the Presqu'ile ProvincialPark staff (2005) for collecting Double-crested Cormorant eggs on my behalf. Thanks to the Biodome de Montreal for providing me with access to altricial embryos, used for comparative purposes.

i Funding Funding for this project was provided by a Tomlinson Fellowship, a Wylie Memorial Fellowship, FQRNT, and NSERC CGS-M and CGS-D fellowships to myself, as well as an NSERC Discovery Grant and Canada Research Chair to H. Larsson which covered logistical expenses. A collections study grant helped cover the cost of the collections visit to the AMNH.

11 CONTRIBUTIONS OF AUTHORS

Chapter 5, "Ossification Sequence of the Common Tern (Sterna hirundo) and Its Implications for the Interrelationships of the Lari (Aves, Charadriiformes)" (In Press, Journal of Morphology) is a manuscript I co-authored with Luke Harrison. I examined the embryos and reconstructed the ossification sequence. I wrote all portions of the manuscript not directly pertaining to PGi analysis, and made figures 2-5.

111 TABLE OF CONTENTS

Acknowledgments i

Contributions of authors Hi

ABSTRACT/RESUME 1

INTRODUCTION 4

CHAPTER 1: Analysis of developmental sequence data: a review Introduction 8 Phenetic Methods 10 Phylogenetic Methods 16 Mapping sequences on an existing tree topology 16 The problem of simultaneity 17 The problem of polymorphism 18 The use of developmental sequence data for phylogenetic reconstruction 23 Conclusions 26 CHAPTER 2: Skeletal development in palaeognathous birds Introduction 28 Materials and Methods 30 Results Skull Dromaius novaehollandiae 31 Rhea americana 35 Struthio camelus 36 Eudromia elegans 39

Postcranial axial skeleton Dromaius novaehollandiae 42 Rhea americana 45 Struthio camelus 46 Eudromia elegans 47

Forelimb Dromaius novaehollandiae 48 Rhea americana 50 Struthio camelus 51 Eudromia elegans 52

Hind limb Dromaius novaehollandiae 55 Rhea americana 57 Struthio camelus 58

IV Eudromia elegans 60

Discussion Skull 63 Paedomorphosis and ratite cranial evolution 67 Postcranial axial skeleton 69 Forelimb 71 Hind limb 75 Incubation period, adult size and ossification 77 Mechanisms of skeletal reduction 78 Morphology and ossification sequence 79

CHAPTER 3: Comparative embryonic development of the skeleton of the Domestic Turkey (Meleagris gallopavo) and other galliform birds Introduction 87 Materials and Methods 88 Results Sequence variability 89 Skull 89 Postcranial axial skeleton 94 Forelimb 95 Hind limb 98 Discussion Skull 100 Postcranial axial skeleton 101 Forelimb 102 Hind limb 103 Factors affecting ossification sequence 105

CHAPTER 4: Ossification Sequence of the Avian Order Anseriformes, with Comparison to other Precocial Birds Introduction 115 Materials and Methods 117 Results Sequence variability 118 Skull 119 Postcranial axial skeleton 125 Forelimb 128 Hind limb 132 Discussion Sequence variability 137 Comparative ossification sequence Skull 137 Postcranial axial skeleton 140 Forelimb 141 Hind limb 142

v Life history and ossifcation sequence 143

CHAPTER 5: Ossification Sequence of the Common Tern {Sterna hirundo) and Its Implications for the Interrelationships of the Lari (Aves, Charadriiformes) Introduction 156 Materials and Methods 160 Results Skeletal development of Sterna hirundo Skull 163 Post-crartial axial skeleton 169 Forelimb 171 Hind limb 175 Intraspecific variability 178 Phylogenetic analysis Parsimony analysis of event-pairs 178 PGi analysis - 179 Discussion Skull 180 Post-cranial axial skeleton 181 Forelimb 182 Hind limb 183 Potential factors affecting ossification sequence 184 Intraspecific variability 186 Phylogenetic analysis 186

CHAPTER 6: Phylogenetic analysis of avian ossification sequence data Introduction 193 Materials and Methods 195 Results Topology reconstruction 197 Character evolution 198 Discussion 202

CONCLUSIONS AND FUTURE RESEARCH 250

BIBLIOGRAPHY 253

APPENDICES Appendix 1: List of specimens 274 Appendix 2: Ossification sequence of the Double-crested Cormorant 281 Appendix 3: Supporting documents 283

vi ABSTRACT

The relative timing and sequence of events during embryonic development plays an important role in adult shape and thus in evolution. The sequence in which bones form in the developing embryo should therefore contain a component capable of revealing evolutionary history, however processes relating to ossification sequence and the sequences themselves are poorly known and rarely discussed. In this thesis, I describe the embryonic skeletal development of Meleagris gallopavo, Sterna hirundo, Somateria mollissima, Anas platyrhynchos, Cairina moschata, Dromaius novaehollandiae, Rhea americana and Struthio camelus for the first time in the scientific literature, focusing on ossification. All species exhibited intraspecific variation in ossification sequence, but the level of polymorphism present was generally quite low. Specimens collected from the wild did not show more variability in ossification sequence than those incubated under constant conditions in the lab. Dermal bones did not always ossify before endochondral bones, nor did neural-crest derived elements always form before elements derived from the paraxial mesoderm. All of this suggests that the factors controlling ossification sequence are complex, and more than one variable may play a role. In order to examine sequences in a more explicit phylogenetic context, I converted them into a form that is easily analyzed (event-pairs) and used these as characters for phylogenetic reconstruction. While this technique is plagued with problems involving logical and biological non-independence, it is an efficient tool for surveying conservation and divergence in ossification sequences at different levels of phylogenetic relatedness. I also reconstructed shifts on an accepted topology. The analysis indicates that ossification sequences are influenced by relative evolutionary reduction or expansion in element size. Reduced elements ossify late in sequence, and also temporally later, as measured by stage. Enlarged elements ossify early in sequence and in stage. This pattern is seen in cranial and post-cranial elements and in dermal and endochondral elements. It is therefore not surprising that ossification sequences converted to event-pairs can correctly recover clades in a phylogenetic analysis if the morphological divergence between groups is large enough.

1 RESUME

Le temps de formation et la sequence d'evenements du developpement embryonnaire jouent un role important dans la forme adulte et dans revolution. La sequence selon laquelle les os se forment dans l'embryon devrait done contenir des informations capables de reveler l'histoire evolutionnaire. Cependant, les facteurs qui influencent la sequence d'ossification et les sequences elles-memes sont mal compris et rarement etudies. Dans cette these, je decris le developpement squelettique embryonnaire chez Meleagris gallopavo, Sterna hirundo, Somateria mollissima, Anasplatyrhynchos, Cairina moschata, Dromaius novaehollandiae, Rhea americana et Struthio camelus pour la premiere fois dans la litterature scientifique, en me concentrant sur l'ossification. Une variabilite intraspecifique entre les sequences d'ossification a ete observee chez toutes especes, mais le niveau de polymorphisme etait generalement bas. Les specimens d'especes sauvages n'ont pas montre plus de variabilite dans la sequence d'ossification que ceux incubes dans les conditions constantes du laboratoire. Les os membraneux n'ossifient pas toujours avant les os de cartilage, et les os derives de la crete neurale ne se forment pas toujours avant les elements derives du mesoderme paraxial. Ceci suggere que les facteurs qui controlent la sequence d'ossification sont complexes et que plus qu'une facteur peuvent y jouer un role. Afin d'examiner les sequences dans un contexte phylogenetique, je les ai convertis en une forme facile a analyser ('paires d'evenements') et ai utilise les caracteres de sequence d'ossification pour la reconstruction phylogenetique. Bien que cette technique presente des problemes lies a un manque d'independence logique et biologique, e'est un outil efficace pour examiner la conservation et la divergence des sequences d'ossification a differents niveaux de relation phylogenetique. J'ai aussi reconstruit les changements sur une topologie admise. L'analyse montre que les sequences d'ossification sont influencees par les reductions ou les expansions evolutionnaires dans la taille relative des os. Les elements reduits s'ossifient plus tard dans la sequence et dans l'etape du developpement. Les elements agrandis s'ossifient tot dans la sequence et dans l'etape. Les elements craniens et poste- craniens et les os membraneux et de cartilage suivent ce modele. 11 n'est done pas

2 surprenant que l'analyse ait demontree que ces sequences d'ossification converties en paires d'evenements peuvent retracer correctement les groupes evolutionnaires dans une analyse phylogenetique si la divergence morphologique entre les groupes est assez grande.

3 INTRODUCTION

Vertebrate embryos begin life as a single cell, in which all of the genetic information exists to determine the shape of the adult organism. Development is the process by which the information encoded in the genes becomes translated into adult morphology. Genes and morphologies can evolve, changing over time, but in addition to this development itself can evolve in response to selection acting on embryonic or larval stages. This greatly increases the degree of difficulty in assessing how differences in development between taxa affect the morphological endpoint. Development can generate novel morphologies in several ways. These can be simplified as evolutionary changes in the timing of onset or offset of a process or event (heterochrony) or changes in the location of this process or event (heterotopy). Heterochrony has two aspects: growth (how fast a process occurs) and sequence (the time of onset or offset of an event relative to the rest of the organism). Historically, heterochrony has been studied as a function of changes in growth between taxa resulting in changes in shape and size. The foundations for this approach were laid by Gould and Alberch (reviewed by Smith, 2001). The benefit of this approach is that the morphological outcomes of changes in growth are intuitive and easily visualized. More recently, however, the focus has shifted to changes in timing of onset of events during development. These events are often represented as a ranked sequence, as this avoids assumptions about the way in which embryos measure time (Smith, 2001). This change in emphasis opens up many new questions involving the way in which development evolves and generates morphological change; however the resulting morphological outcomes are less clear (Bininda-Emonds et al., 2003), as will be discussed throughout this thesis. The aspect of sequence heterochrony I am examining is the timing of onset of ossification in the various bones of the vertebrate skeleton, specifically in Class Aves. Ossification sequence is a relatively new and poorly understood way of looking at evolution and diversity. It has gained popularity due to its ease of application in studying the development of non-model organisms, but is especially

4 useful when looking at the development of fossil taxa where gene expression and molecular studies are not possible (Schoch, 2002; Balanoff and Rowe, 2007; Frobisch et al., 2007). Previous studies have questioned the utility of ossification sequence data in phylogenetic reconstruction, focusing instead on its use as a tool for understanding adaptation and modularity (Sanchez-Villagra, 2002; Schoch, 2006). Ossification sequences are known from a relatively small number of extant taxa, and the degree of intraspecific variation and variation within closely related groups may have a large effect on such studies. While it is assumed that changing the timing of onset of ossification of a give element has implications for both the embryo and for the size and shape of the resulting element in the adult, the downstream effects of small changes may be minimal. Ossification sequences are influenced by a number of unknown factors acting in combination. Several hypotheses have been advanced regarding which factors are most important, and it is likely that gene expression alone is not the determinant of ossification sequence. Physical factors, such as muscle development and embryonic movements (Adriaens and Vermes, 1998; Wagemans and Vandewalle, 2001), as well as constraints such as modularity, the sequence of chondrification and the source of osteogenic cells (Maxwell, 2008; Frobisch, In Press) have all been proposed to play a role. Sexual dimorphism affects sequence (Garn et al., 1966), as do shape variables such as adult morphology (Haluska and Alberch, 1983; Adams, 1992; Rieppel, 1993b). Additionally, phylogeny (Maisano, 2002b; Poe, 2006; Hofmann et al., 2007), adaptations to post- embryonic life (Adriaens and Verraes, 1998; Mabee et al., 2000; Sanchez- Villagra, 2002; Prochel, 2006), life-history (Prochel 2006), and other untested ecological variables like temperature, humidity and geography result in variation both between and within species (Sheil and Greenbaum, 2005). Some or all of these factors may affect the observed sequence, and can be difficult to isolate as they often interact and influence one another. In the past, hatchling (neonate) adaptation was believed to be the most important factor in determining ossification sequence (Adriaens and Verraes, 1998; Mabee et al., 2000; Sanchez- Villagra, 2002; Prochel, 2006). This is unlikely, as exceptions to some of the classic examples have been described, for instance the ability of some marsupial

5 neonates to climb to the pouch without ossified forelimbs (Gemmell et al., 1988), and feeding in cannibalistic fish larvae without complete jaw and palatal ossification (Vandewalle et al., 2005). A number of methodologies are available for the evolutionary analysis of ossification sequence data. Sequence data are problematic because they are inherently non-independent, both in the logical and in the biological sense. Many methodologies have been published attempting to circumvent these problems but often tradeoffs between philosophical rigour and computational intensity are involved (Chapter 1). For this reason, a descriptive approach to understanding ossification sequences was also employed. The descriptive approach was useful because avian ossification sequences and patterns of skeletal development are especially underreported in the literature, which is somewhat surprising since several bird species are of agricultural importance. Birds represent an ideal system for examining factors underlying ossification sequences, as well as the potential phylogenetic utility of these sequences. Many bird species are of agricultural importance and are farmed commercially in Canada, making it easy to obtain eggs and embryos. Bird eggs are relatively easy to incubate in a laboratory environment; development is postponed until the onset of incubation and it is not necessary to sacrifice the mother in order to collect embryos. The Domestic Chicken {Gallus gallus) and the Japanese Quail (Coturnix coturnix) are model systems in developmental biology, and gene expression patterns during development and the sources of osteogenic cells are well characterized (ie. Evans and Noden, 2006). Previous studies suggest that the degree of altriciality and precociality do not exert a profound effect on ossification sequences in birds (Rogulska, 1962; Starck, 1993). In this thesis, I characterize the ossification sequences of several of the more basal orders of extant birds: Struthioniformes (Chapter 2), Galliformes (Chapter 3) and Anseriformes (Chapter 4) with the objective of increasing the utility of this study for paleontologists interested in fossil archosaurs. I also describe skeletal development in the order Charadriiformes (Neoaves) (Chapter 5). I sampled several species per order in order to examine sequence conservation and divergence at the ordinal level. If more than one sequence was available for a

6 given species, I did not pool the results in order to examine intraspecific sequence conservation and divergence. I then analyzed the sequences described in Chapters 2 through 5 in an explicitly evolutionary context (Chapter 6). I conducted a phylogenetic analysis using ossification sequence data in order to evaluate its merits for phylogenetic reconstruction when intraspecific variability was taken into account and taxon and sequence length were increased. I then looked at sequence changes on an accepted topology in order to quantify the shifts previously described qualitatively. Throughout this thesis, I attempt to address the question of which variables are most important in driving ossification sequence shifts. By understanding the factors responsible for changes in sequence, the limitations and potential applications of this type of data can be better understood, and potential pitfalls avoided.

7 CHAPTER 1

Analysis of developmental sequence data: a review

Levels of morphological diversity and disparity change over evolutionary time, and the mechanistic basis of these morphological changes can be understood through the study of development in a comparative context. Heterochrony, or changes in developmental timing resulting in evolutionary changes, is considered one of several important mechanisms in understanding these processes. The study of heterochrony has been broken down into two main components: growth heterochrony, or differential growth rates which result in changes in size and shape, and sequence heterochrony, or changes in timing of developmental events (Smith, 2001). Within the past ten years the study of sequence heterochrony has undergone a rebirth, mostly due to efforts to improve analytical techniques and computing power. Historically, sequence heterochrony was dealt with by eyeballing descriptive data, and in many cases due to methodological limitations and restrictive taxonomic sampling, this trend continues to the present day. The role of sequence heterochrony in generating morphological changes in the adult has been questioned (Bininda-Emonds et al., 2003), and many authors have instead focused on its role in embryological and larval adaptations, and even considered it a potential source of characters for phylogenetic analysis. According to some authors, there appears to be phylogenetic signal present in developmental sequence data (Strauss, 1990; Velhagen Jr., 1997; Jeffery et al., 2002b). This is by no means universally accepted, and may be data-set specific (Velhagen Jr., 1997; Chipman et al., 2000; Schoch, 2006). In this review, I will discuss the strengths and weaknesses of the methodologies that have been proposed to quantitatively examine developmental sequence data, and studies that have attempted to implement them in an evolutionary context. These methods can be divided into two broad categories, phenetic comparisons between terminal taxa that strive to uncover integrations within the developmental sequences and putative shared sequence heterochronies, and phylogeny-based methods that derive ancestor-

8 descendent sequence heterochronies and establish statements of sequence evolution.

Ranked developmental sequence data constitute a series of developmental events that occur one after the other in a constrained order during the ontogeny of an individual. In practice, a ranked sequence of events is usually constructed using cross-sectional data; in other words based on many individuals. These ranked developmental events are isolated from their temporal context, and therefore can be examined independent of absolute time, size, and developmental stage, all of which have proven problematic for interspecific comparisons (Smith, 2001; Bininda-Emonds et al., 2002). Ranking developmental events also removes the timing and rate effects caused by environmental perturbations, making it a useful method for analyzing intraspecific data as well (Smith, 2003). Smith (2001) argues that regarding developmental events as a ranked sequence makes biological sense based on the sequential way embryos measure time. Presenting developmental events as a ranked sequence also allows for the inclusion of multiple different kinds of events in the same analysis (Smith, 2003). For instance, ossification events can be included in an analysis that also incorporates sequential gene expression patterns, as long as the two sets of events can be merged into a single ranked series. In spite of these useful features, ranked developmental data has several major detractors. Firstly, ranking events leads to a loss of information regarding the absolute timing of events. Ranking events standardizes the time units between each event, and so relatively condensed and relatively spread out parts of the sequence appear identical (King, 2004). There is also the problem of resolution and specimen number, a flaw characteristic of most developmental studies using cross-sectional data, which examines many specimens of different ages, rather than longitudinal data, which follows a single individual through its ontogeny (King, 2004). Many specimens are needed to obtain a perfectly resolved series with cross-sectional data. In the best-case scenario, as many specimens as there are events are needed. In reality, many more are needed, especially to resolve highly condensed parts of the sequence (Mabee et al., 2000), and even more to

.9 resolve any variability within a population. Developmental sequence data is influenced by two types of biological dependence: collective dependence, in which the events are subject to the same developmental constraints, and linear dependence in which an event can only occur after the occurrence of a prior event (Koenemann and Schram, 2002). The lack of event independence creates problems when attempting to analyse the data phylogenetically as discrete events. This differs from logical dependence, to be discussed later, which arises as a methodological artifact of event-pairing.

There are currently two major classes of methods used to analyze developmental sequence data. The first class, phenetic methods, includes both graphical and statistical approaches, and deals strictly with terminal taxa. Graphical analyses are intuitive and easy to interpret but do not work well with a large number of taxa or events. They can be placed in a statistical framework to provide a means to explicitly test widely held assumptions about the evolution of development, but the results are difficult to interpret and cannot easily be placed in an evolutionary context. The second type, phylogenetic methods, are equally difficult to interpret but clearly place developmental sequence data in an evolutionary context, providing nodal estimates of ancestral sequences, and can theoretically deal with much larger data sets than graphical methods.

1) Phenetic methods This class of methods can be divided into two subclasses, graphical analysis, followed by statistical analysis. These are grouped together based on their shared philosophical approach: both rely only on terminal taxa and both are unable to estimate ancestral sequences or place results in a phylogenetic context. Graphical studies are intuitive and easy to implement, but may be more appropriate for preliminary analyses (Poe, 2004; Poe and Wake, 2004), and for analyses involving few taxa and events (Smith, 2001). Smith (2001) formalized a bivariate graphical method to identify rank changes in the ontogeny of multiple species. Developmental rank is plotted against developmental event. One reference taxon is selected to be the standard,

10 and its sequence determines the arrangement of events on the x-axis. The developmental sequences of additional taxa are compared to it, and sequence changes appear as deviations from the reference line. This technique was recently used to compare the ossification sequence of the limbs of a fossil tetrapod to various extant species and draw conclusions regarding the evolution of divergent patterns of limb formation (Frobisch et al., 2007). The most complicated graphical method developed thus far can be attributed to Schlosser (2001; 2003). In this case, 'heterochrony plots' are used. These permit a direct comparison of the relative time of developmental events between two species. The x-axis is scaled according to a measure of ontogenetic time for species 1, and the y-axis is scaled using a similar measure of ontogenetic time for species 2. The developmental events under consideration are plotted by time of appearance. Error bars can be added after the fact to indicate where gaps in sampling occur. Changes in the slope of the line and deviations from the line indicate alterations in the rate and the sequence of development between the two taxa. Although this technique was originally designed to measure the relative timing of events in species 1 compared to species 2, criticisms involving uneven sampling and developmental rates can be alleviated by using ranked data (Bininda-Emonds et al., 2002). Schlosser applied his method to an examination of dissociated coevolution in direct developing versus indirect developing frogs (Schlosser and Roth, 1997; Schlosser, 2001, 2003). In this case, the method performs quite well and clearly detects codissociation of limbs and their innervation, the pharyngeal arches and the nervous system. Using Schlosser's method, it is necessary to compare every ingroup taxon independently to the outgroup in order to place the results in an evolutionary context and determine character polarity. There is also no procedure in place to infer internal nodes, making phylogenetic information difficult to include. The selection of regression lines is subjective, which is problematic because it has a strong influence on the results. For instance, it may be difficult to determine whether a series of points falls along a single line, or is better described as two lines, suggesting they belong to two separate modules (Bininda-Emonds et al.,

11 2002). This method is limited by the current absence of significance tests (Poe and Wake, 2004).

A statistical approach can also be applied to these inherently phenetic graphical approaches. These statistics are designed to test overall sequence conservation or divergence between two groups, as well as to test hypotheses regarding the movement of specific events. These methods provide only measures of overall similarity, and are difficult to place in a phylogenetic context. It is therefore difficult to tell where in the phylogeny changes occurred, and rank correlation measures cannot accommodate missing data (Bininda-Emonds et al., 2002). One of the first statistical methods was developed by Nunn and Smith (1998). They used Kendall's coefficient of concordance (W), a nonparametric measure of correlation, to examine the overall divergence of ranked data from multiple species. This statistical measure can allow the comparison of more than two groups. Values range between 0 and 1, with W=\ indicating that the two sequences are identical. A simulation is needed to determine critical values of W because Kendall's coefficient of concordance was designed to test the null hypothesis that two sequences have no similarity. This approach was applied to sequence divergence between marsupial and placental mammals. Results supported overall sequence conservation in mammals, with higher levels of similarity found within a group than between groups (Nunn and Smith, 1998). This method was also applied to examine the relative divergence between turtle cranial ossification sequences and those of other reptiles (Sheil, 2003b). While divergence wasn't extensive, higher divergence was seen between groups than within groups. The problem with this type of analysis is to determine the level of similarity that is biologically relevant (Smith, 2001). Kendall's coefficient of concordance, as well as the similar Spearman's rank correlation, have also been used to evaluate sequence divergence at lower taxonomic levels, such as between species and between experimental treatments (Mabee et al., 2000). An ANOVA can be used to test hypotheses that individual events shifted significantly during development in marsupials relative to placental mammals

12 (Nunn and Smith, 1998). Computer simulation was used to generate revised values for the F statistic, taking into account the effect of phylogeny on sequence similarity, and found support for significant rank differences in the development of the central nervous system and somatic elements of the craniofacial region between marsupial and placental mammals (Nunn and Smith, 1998). Unfortunately, critical values generated using this technique are highly sensitive to the model of evolution used (Jeffery et al., 2002a). AN OVA is a parametric statistic, and so data must meet these assumptions (Nunn and Smith, 1998). It can be misled by strong signal, and cannot determine in which lineage the heterochronic change occurred (Jeffery et al., 2002a). For both Kendall's coefficient of concordance and AN OVA, a single most parsimonious developmental sequence for every species is required, and simultaneous data caused by a poorly resolved sequence reduces the ability of these techniques to detect between-group differences. In order to examine the relative timing of chondrogenesis in the fore- and hind limbs of tetrapods, Bininda-Emonds et al. (2007) created an event-pair metric to measure relative timing (see "phylogenetic methods" section for details on the generation of event-pairs). The sum of all non-redundant event-pairs relating to the sequence of chondrogenesis of fore- and hind limb elements divided by the number of event-pairs yielded a value which, if less then 1, indicated an acceleration in the development of the hind limb, if approximately equal to one, the two limbs developed synchronously, and if greater than 1, the forelimb was accelerated. They termed this value the "average event-pair score" (EPS). The values were then analyzed using an AN OVA combined with Fisher's PLSD in order to identify trends in major taxonomic groupings, showing taxon- specific heterochrony in timing of limb development. The effect of simultaneity on the results is unclear, but should generally result in a convergence on synchronous limb formation. A single most parsimonious developmental sequence is required for each species (Bininda-Emonds et al., 2007). The number of secondary ossification centers in different limb regions of the neonates of three species of primates was analyzed using a Kruskal-Wallis test for multiple comparisons, followed by a Dunn's post-hoc test (Hofmann et al.,

13 2007). The purpose was to discriminate between taxa and draw phylogenetic conclusions. The forelimb, hind limb, and hand and carpal regions in isolation differed significantly between taxa. The post-hoc test revealed that the tamarin was significantly different from both Goeldi's monkey and the marmoset for all anatomical regions, but the latter two taxa were not significantly different from each other. The authors emphasized the need of increased taxon sampling to lend support to the phylogenetic aspects of their results. In a novel approach to phenetic similarity, which allows a systematic perspective to be included, King (2004) used discriminant function analysis to examine developmental sequences. This method has several key advantages, firstly that fewer specimens are needed to resolve the relative timing of ontogenetic events, secondly that intraspecific variation can be easily incorporated, and lastly that ontogenetic differences can be detected in taxa sharing identical developmental sequences. A series of developmental events is classified using a four-stage scoring system, which details the completeness of the event (for instance, 1= unossified, 2=35% ossified, 3=70% ossified, 4= completely ossified). The sum of the scores for a developmental event across all individuals of a species represents that event's relative timing in ontogeny, and the scores for all events can then be used to construct the developmental sequence of that species. Discriminant function analysis uses the raw scores of each individual to assign individuals to higher taxonomic units. Events that correlate most strongly with the discriminant function are the ones that differ between two taxa, and the sign of the function, when combined with the location of the group centroid, provides information on the direction of the heterochronic shift. When applied to 72 primate post-natal skeletal developmental events, 99% of all individuals of all ages were correctly classified to suborder. When a single species was analyzed, males and females could be separated. Although the data are not normally distributed, this is not problematic for the use of discriminant function analysis as a classification tool, but prevents the implementation of measures of statistical significance. Like other phenetic measures, it is difficult to place in an explicitly evolutionary context (King, 2004).

14 Ranked developmental sequence data can be extended past taxon-specifc cases to test general models of developmental evolution. For instance, Poe and Wake (2004) used Spearman's signed ranks test to address whether small (one- step) changes in rank were more likely than large shifts in an amniote organogenesis data set (Jeffery et al., 2002b). They also tested whether evolutionary lability was equal over the developmental period. Later, this test was expanded to include multiple datasets encompassing more aspects of development. In this case, Kendall's tau was used, with Bonferroni correction for multiple comparisons. The results were similar across datasets, with sequence heterochronies being equally likely to occur across all ranks, and small rank changes being more likely than large ones (Poe, 2006). If an a priori hypothesis of modularity is developed, it can be tested statistically using ranked data (Poe, 2004). In this case, Kendall's tau was used to test the hypothesis that a given set of events shows a higher degree of correlation than random, as determined through simulation. This technique recovered modularity in the vertebrate limb (Schlosser, 2001), but failed to detect modularity in skeletal ossification sequence (Nunn and Smith, 1998). The power of the test is influenced by the number of developmental events (Poe, 2004). Goswami (2007) applied this methodology to an expanded dataset of ossification sequences of marsupial and placental mammals to test if evolutionary modules based on adult stages and defined using morphometric data coincided with developmental modules, represented by ossification sequence data. Her results were not significant, indicating either that ossification sequence changes within a module can occur, that the morphometrically defined modules are not equivalent to ossification centers, or that there is no modularity in ossification sequences. A method that attempted to integrate statistical and phylogenetic approaches was pioneered by Larsson (1998). In order to test hypotheses of character integration, he created a bivariate plot of the ranked order of appearance of osteological characters in Alligator ontogeny against their ranked order of appearance in a linearized crocodilian phylogeny. Areas of the plot with stronger positive correlation between ontogenetic rank and phylogenetic rank were interpreted as being integrated over evolution and development. A Spearman rank

15 correlation was implemented to attach significance values to the graphical correlation. This approach is different from the other phenetic methods in that it sets up an explicit phylogenetic context, involving fossil taxa, in which hypotheses of character integration are tested over both evolutionary and developmental time.

2) Phylogenetic methods Many researchers attempt to draw explicit evolutionary conclusions from ranked developmental sequence data. This is done using two mutually exclusive but similar methodological approaches. The first is to map ranked developmental sequences on an existing phylogenetic tree and reconstruct the ancestral states and subsequent sequence shifts that must have occurred. The second attempts to create phylogenetic hypotheses using developmental sequence data. While these two approaches are similar in many fundamental ways, their outcomes are quite different.

a) Mapping sequences on an existing tree topology The most intuitive and common way sequence changes are mapped on to an independently derived phylogenetic tree is by comparing the order of occurrence of every developmental event to every other event using an all-pairs comparison. This converts ranked data to a series of 'event-pairs'. This technique was developed independently by several authors (Mabee and Trendler, 1996; Smith, 1996, 1997; Velhagen Jr., 1997). In essence, a matrix of TV by TV developmental events is constructed. The rank of column events and the rank of row events are compared. If the column event occurs earlier in the developmental sequence than the row event, that event-pair is assigned the character state of 2. If the reverse is true, the event-pair is given the character state of 0. If the two events appear to occur simultaneously due to poor sequence resolution, the event-pair is assigned character state 1 and the character is treated as ordered (Smith, 1996). Alternatively, simultaneous events can be coded as missing data, resulting in binary characters rather than multi-state characters (Velhagen Jr., 1997).

16 The problem of simultaneity in event-pairing. Simultaneity can either be coded as a separate state (Smith, 1996; Bininda-Emonds et al., 2002; Schoch, 2006), or as missing data (Velhagen Jr., 1997; Bininda-Emonds et al., 2003) when event-pairing is used. If it is coded as a separate state, the characters are frequently considered as ordered. This implies that a simultaneous state is necessarily a transitional event during a heterochronic shift. Yeh (2002) ran separate event-pairing analyses for both types of simultaneity coding. She found that the coding method chosen did not significantly affect the results and this is most likely why both types of coding have persisted in the literature. However, Bininda-Emonds et al. (2003) found that the number of changes detected on a branch was significantly correlated to the amount of simultaneity in the data set when simultaneity was coded as a separate state. This was improved by coding simultaneous events as missing data. The causes of simultaneity are not thoroughly understood. Theoretically, the number of embryos that must be sampled in order to generate a perfectly resolved sequence is equal to the number of developmental events examined. In practice, however, many more embryos are necessary because of variably condensed regions of the sequence, and intraspecific sequence variation (Mabee et al., 2000). There is also a debate regarding the nature of simultaneity. While there may be no truly simultaneous events in the ontogeny of an individual (Nunn and Smith, 1998), it is likely that even with infinitely close sampling intervals, it is impossible to eliminate simultaneity at the level of species simply due to individual variation (Harrison and Larsson, In Press). Although there are typically more instances of simultaneity with data-poor species, in some well-sampled data sets the amount of simultaneity does not decrease with more intensive sampling. This is partially because the amount of intraspecific variation detected increases as the number of embryos sampled increases. It has been suggested that there may be an optimal number of events that can be sequentially resolved during the developmental time span of a species (Bininda-Emonds et al., 2003). There is therefore an intersection between the problems of ideal sample size, polymorphism and simultaneity.

17 The problem of polymorphism in event-pairing. There are several options available for dealing with intraspecific variation in the analysis of developmental sequences. One of the more popular options is to select a single representative sequence for a taxon, and effectively ignore sequence variation. This risks biasing the results, depending on which sequence position is randomly or subjectively assigned to a variable element. For example, OSA (ontogenetic sequence analysis) uses parsimony followed by the construction of reticulating networks to identify a modal or average developmental sequence. This method can also be used to identify alternative sequences, as well as their predicted and observed frequencies in a population. While this technique attempts to address intraspecific variability in developmental sequence data, it is difficult to apply to data sets consisting of many events (Colbert and Rowe, 2008). A second approach is to use only conserved events in the construction of the sequence (Smith, 2001). This is perhaps not the ideal approach, as it has been demonstrated using other morphological character sets that polymorphic characters do contain phylogenetic information (Wiens, 2000). Lowering rank-order resolution in order to avoid variation also discards information. Some authors advocate analyzing alternate sequences independently (Bininda-Emonds et al., 2002), although this has never been applied in practice. Using the event-pairing method, variable event-pairs can either be coded as missing or as simultaneous data. Polymorphic coding is another alternative that can be implemented if simultaneity is coded as a separate state. The issue of sequence polymorphism is difficult to address using any of the methods, but is important and should be considered further.

Regardless of how polymorphism and simultaneity are addressed, the result is a matrix ofN(N-\)/2 event-pairs, representing a ranked sequence as a series of discrete binary characters. The resulting matrix can then be vectorized, and the event-pairs compared between taxa. This overcomes differences in sequence resolution and sample size. Event-pairs can then be optimized onto an existing phylogeny with standard phylogenetic software, usually using a parsimony algorithm, and the evolution of the developmental sequence under accelerated and delayed models of character transformation can be examined.

18 Event-pairs are non-independent characters in several ways, and this creates problems during optimization. In addition to any biological non- independence, this coding system creates further logical dependence (Bininda- Emonds etal., 2002; Schulmeister and Wheeler, 2004). This can lead to illogical ancestral character optimizations at internal nodes (for instance, event A precedes B, B precedes C but A follows C) (Schulmeister and Wheeler, 2004), and the problem is exacerbated when inapplicable events are included in the matrix (Jeffery et al., 2002b). Event-pairing in its simplest form has been widely used to analyze sequence data, generally quite successfully. Smith (1996, 1997) examined the development of 28 central nervous system and somatic events in nine marsupial and placental mammals, and found that a major shift in the relative timing of these events had occurred that was correlated with reproductive differences between the two groups. Simple probability statistics indicated that this distribution of character states was unlikely to have occurred by chance alone. A third study of mammalian developmental sequences examined postcranial events and also found a large number of autapomorphies in marsupials (Sanchez-Villagra, 2002). Velhagen (1997) analyzed a much smaller data set of 5 cranial ossification events in 6 species of thamnophiine snakes. The results were somewhat inconclusive. Chipman et al. (2000) examined the sequence of appearance of 14 histological features of the notocord, eye, ear, heart, kidney and epiphysis, which are used to define stages in early anuran development. They found a lack of any phylogenetic or ecological pattern in the observed interspecific variation, and large heterochronies in the relative timing of eye and ear development. In a second study of sequence heterochrony in early anuran development, Mitsgutch et al. (2008) used 25 developmental events, about half of which pertained to the development of cranial neural crest streams, to infer variability in early development between two closely related species. In a study of anuran skeletal development, a majority-rule consensus sequence of the ossification of 16 skull bones for a large number (70) of vertebrate species found delayed jaw ossification in anurans, but homoplasious reversals in pipoids emphasizing the importance of taxonomic sampling when examining developmental sequences (Yeh, 2002).

19 Jeffery et al. (2002b) assembled a large data set of 41 developmental events spanning organogenesis in 14 vertebrate taxa from all major classes. Among other things, they found that an acceleration in heart development evolved prior to the evolution of homeothermy in birds and mammals. These studies give an idea of the type and number of events and the number of taxa that have been examined using event-pairing. As in any other type of phylogenetic analysis, the quality of the results is strongly influenced by the phylogenetic distribution of the taxa sampled, as well as by the number of characters and taxa examined.

The event-pairing method does not provide information on the direction or the magnitude of the shifts occurring along a given branch. To address this problem, the method was expanded and termed 'event-pair cracking' (Jeffery et al., 2002a). In this approach, a heuristic search determines the smallest set of event movements at a node that accounts for the observed change, in other words the most parsimonious solution. The shifts that moving events have undergone are summed (+1 value for an acceleration, -1 for a delay) and the sums are ranked. The researcher then selects the magnitude of shifts thought to be significant. Of this subset of moving events, rankings are recalculated not counting movements between events in the subset. After this recalculation, events with scores above 0 are assumed to have moved, all other events only appear to have shifted. This technique turns out to be too conservative for use with most real data sets; the results do not end up explaining all of the changes that have occurred along a given branch (Jeffery et al., 2005). The internal nodal reconstructions suffer from the same problem of illogical reconstruction as described for event-pairing. The inclusion of error bars was suggested by the authors, although these would be difficult to implement (Jeffery et al., 2002a). The technique was used to reanalyse Smith's (1996) data set, and obtained similar results to the original analysis (Jeffery et al., 2002a). Event-pair cracking was also used to examine 116 events spanning organogenesis in 13 species of mammals and two outgroups (Bininda- Emonds et al., 2003). The procedure was made even more conservative by only examining events that were moving under both accelerated and delayed character optimization. The number of changes per branch was significantly correlated with

20 the number of simultaneous events, so simultaneous event-pairs were coded as missing data. The analysis found high levels of homoplasy on terminal branches, and much lower levels on internal branches suggesting rapid sequence evolution. This pattern might be a methodological artifact caused by lack of logical constraint in ancestral reconstructions (Harrison and Larsson, In Press). Schoch (2006) examined 16 cranial ossification events in 13 vertebrate taxa. He compared the sequences of the terminal taxa directly to the outgroup to infer shifts, rather than relying on nodal reconstructions. Total changes were therefore calculated relative to the basal sequence. While this approach is not identical to event-pair cracking, it might be more useful for looking at major shifts when dealing with poorly resolved topologies. In an attempt to address the flaws of event-pair cracking, Jeffery et al. (2005) developed the Parsimov method. All possible solutions for internal nodal reconstructions are investigated, and the most parsimonious solution is selected. If more than one most parsimonious solution exists, as is often the case, a consensus sequence is calculated. This is more computationally intensive than event-pair cracking, but heuristic search strategies can be implemented. Unlike event- pairing, the Parsimov method is unable to construct topologies. Because it still relies on an event-pairing framework, problems involving illogical ancestral reconstructions persist. This method was implemented in a study of 14 morphological and physiological events in a clade of freshwater snails (Smirthwaite et al., 2007). This study is unique because it incorporates physiological as well as morphological sequence data, and also because it examines invertebrates, a group that has been largely ignored in the developmental sequence literature. A less conservative approach was used in order to account for all the observed changes, with events found to have shifted under ACCTRAN or DELTRAN character optimization being examined. A large amount of homoplasy was present in the data set, but clear developmental differences existed at the interfamilial level. The authors suggested that these developmental traits may be playing a role in species divergence in this group, which would explain the high levels of homoplasy.

21 The use of event-paired data is only one of the possible ways to reconstruct the evolution of developmental sequences. Many authors would prefer to use the developmental sequence itself as a single character, and thus avoid the coding dependence implicit in the event-pairing approach (Mabee and Trendler, 1996). One way that has been proposed to achieve this is to use a step-matrix approach (Mabee and Humphries, 1993). This approach was initially designed for transformation series entailing strict ontogenetic dependence (Alberch, 1985), which are represented more accurately as ordered multi-state characters (Mabee and Humphries, 1993). Its use has been extended to all types of developmental sequence data, not just those showing ontogenetic dependence (Schulmeister and Wheeler, 2004). The PGi method (Harrison and Larsson, In Press) examines the ranked series of developmental events as a single complex character. The Parsimov algorithm (Jeffery et al., 2005) was selected as the edit-cost function, because it minimizes the number of shifts necessary to explain the observed sequences, but does not discriminate based on the magnitude of those shifts. Because PGi uses a genetic heuristic, rather than an exhaustive search (as in Schulmeister and Wheeler, 2004) it is much less computationally intensive than the latter method. Since it does not atomize the ranked developmental sequence into event-pairs, it avoids problems of illogical nodal reconstructions and overweighting of observed shifts. This method was applied to a dataset of shorebird cranial ossification (Chapter 5). While PGi could not differentiate between two of the three possible topologies, one was found to be less optimal. These results were congruent with an event-pairing analysis on an expanded ossification dataset involving the whole skeleton (Maxwell and Harrison, In Press). A combination of the step matrix and event-pairing approaches was used in the analysis of the evolution of amphibian metamorphosis (Reiss, 2002). Event- pairing was used to indicate the timing of transformation events relative to metamorphosis, while a transformation series represented the morphological changes each element underwent. This allowed for the reconstruction of both shape and temporal data. The redundancy inherent in this type of coding does not permit this technique to be used for phylogenetic reconstruction.

22 b) The use of developmental sequence data for phylogenetic reconstruction It is very tempting to. use ontogenetic data to reconstruct the evolutionary history of a group. Development can yield a large number of characters, and since it results in a stereotyped adult morphology known to contain phylogenetic information, development must also be constrained by evolution. An initial attempt to phylogenetically analyze ossification sequence data did not use formal event-pairing methodology (Strauss, 1990). Instead, the time of appearance of 17 cranial elements was used as characters, and a multivariate measure of size at which a given element ossified was used as the character state. This resulted in a loosely ranked series of events. The characters were then analyzed using parsimony and maximum likelihood. The topology generated by both analytical techniques was consistent with an accepted morphological phylogeny. This method is difficult to apply to more distantly related taxa, due to its emphasis on body size rather than ranked data, but it represents one of the first examples of the use of ontogenetic data in a phylogenetic context. Event-pairing has frequently been used as a way to generate phylogenetic hypotheses, although it violates the assumption of character independence required for phylogenetic analysis. Most authors are interested in using this technique as a way. to explore the data; very few place high levels of confidence in the resulting topology. Velhagen (1997) attempted to use both his snake cranial ossification data set, and a poeciliid fish data set (Strauss, 1990) to construct phylogenies. The former yielded a completely unresolved topology, but the latter produced results consistent with other data sets. When the snake cranial ossification data set was reanalyzed using a step-matrix approach, the resulting phylogeny was also unresolved (Schulmeister and Wheeler, 2004). Jeffery et al. (2002b) recovered some clades (amniotes, mammals, and artiodactyls), but failed to recover others using their data set of developmental sequence of organogenesis. An analysis of cranial ossification over all tetrapods failed to recover the monophyly of any major clades; only the sister-group relationship between Gallus and Alligator was

23 supported (Schoch, 2006). An analysis of chondrification and ossification events in the anuran postcranial axial skeleton yielded a poorly resolved consensus tree inconsistent with all phylogenies proposed using more conventional data sets (Blanco and Sanchiz, 2000). It should be noted that in this study, there were very few events that varied between the taxa sampled. An analysis of postcranial ossification events in mammals also failed to produce a tree congruent with accepted hypotheses; only the marsupials were recovered as a monophyletic group (Sanchez-Villagra, 2002). An analysis of post-embryonic ossification . centers in squamates reliably recovered crown clades, but was uninformative for more basal divergences (Maisano, 2002b). In a much larger analysis, 67 taxa in the higher taxonomic grouping Spiralia, which includes among other invertebrates polychaetes and molluscs, were coded for 53 ordered characters relating to the timing of cell lineage events relative to the total number of cells in the animal. Although there were some discrepancies and basal polytomies, there was general congruence with existing phylogenies in both the large and small scale patterns recovered (Guralnick and Lindberg, 2001). A simulation-based approach examined the reliability of parsimony as an optimality criterion for developmental sequence data, using a subset of characters from the amniote organogenesis data set (Jeffery et al., 2002b; Koenemann and Schram, 2002). The results of parsimony, neighbour joining and UPGMA were tested for both event-paired data and directly on ranked event data. A topology consistent with the accepted phylogeny was recovered by both the parsimony and neighbour-joining analyses when event-pair data was used. The ranked data failed to recover a consistent topology using any algorithm. In spite of the higher success rate of the event-paired data, the authors argued that ranked data was superior because of its ability to discriminate between dependent and independent events, as well as its ability account for constraints (Koenemann and Schram, 2002). Schulmeister and Wheeler (2004) developed a method to deal with developmental sequence data as a single complex character. Their method uses search-based optimization, with an edit-cost function (for instance, event-pairing) to compare among possible states, whether present in the data set or not. Missing

24 data can be dealt with by omitting it from possible sequences. This method cannot deal with intraspecific variation at the present time, and polymorphisms cannot be retained at internal nodes. The great benefit of this method is that because no recoding occurs, no impossible sequences are created at internal nodes. This method can be used not only to create topologies, but also to optimize characters on an existing phylogeny. It is computationally intensive, which restricts its use to all but the smallest data sets. To apply this method, the authors reanalyzed the snake cranial ossification data set (Velhagen Jr., 1997), and recovered a polytomy (Schulmeister and Wheeler, 2004) - the same result that was recovered using event-pairing (Velhagen Jr., 1997). Treating an ontogenetic sequence as a single complex character also entails assumptions about the nature of developmental sequence data, namely that all events are ontogenetically constrained to a single variable (a character) with alternative states (alternative sequences), similar to but less obviously than a transformation series. Developmental biology is complex, and the assumption that a single character is responsible for the myriad of events recorded in a developmental sequence may be an oversimplification. This assumption needs further scrutiny, but may be operationally valid depending on the data set involved (Maisano, 2002b; Poe, 2006). An alternative approach to dealing with developmental data is to code them as a series of discrete, independent characters similar to those used in other morphological analyses. The sequence itself cannot be analyzed in this way, but other developmental patterns can be included in a phylogenetic context. Maisano (2002b) found that these discrete characters, such as the number and location of cartilages and the pattern of fusion of elements could resolve deep divergences not resolved by ossification sequence data alone. A slightly different technique was used to examine anomuran phylogeny (McLaughlin et al., 2004). In this analysis, three life-history stages of 16 crabs were each coded for the same 27 discrete characters and a maximum parsimony search was performed. The resulting tree was unrooted, due to the lack of an appropriate outgroup, but major divisions supported by ontogenetic characters were revealed. A step-matrix approach may be preferable for this type of transformation data, with each of the 27 characters stepped through the three life history stages.

25 CONCLUSIONS

Phenetic and phylogenetic methodologies address fundamentally different questions regarding developmental sequence evolution. Phylogenetic methods examine the evolution of the sequence itself, calculating theoretical ancestral sequences and mapping the changes that have occurred along the branches of a phylogeny or estimating the phylogeny based on the sequence data. In contrast, phenetic methods provide a statistical estimate of sequence similarity between two groups, and can also be used to test hypotheses dealing with theoretical aspects of sequence evolution, such as modularity and character integration. All phylogenetic methods for the analysis of developmental sequence data retain flaws but the field is rapidly evolving and positive modifications to existing methods, as well as new methods are constantly being introduced. Developmental sequence evolution remains an intriguing field of study, and will gain prominence as the development of non-model organisms becomes increasingly accessible as a source of data.

26 Bridging text 1. The obvious tradeoff between computational intensity and philosophical rigour in the evolutionary analysis of developmental sequence data suggests that there is no single best way to accomplish this type of analysis. Reducing development, a continuous process, to a series of discrete presence / absence ossification sequence events is inherently a reductionist approach. It is important to view skeletal development as a dynamic process involving shape and timing variables as well as the rank sequence in which events occur. This approach may permit the identification of factors affecting the distribution, morphology and time of formation of ossification centers in a way that a rank sequence cannot. For this reason, I examine every order of birds separately, looking at the degree of developmental conservation and divergence within each. I begin with Palaeognathae, generally acknowledged to be the most basal group within extant birds (Gibb et al., 2007; Livezey and Zusi, 2007; Slack et al., 2007), and also one of the most morphologically diverse avian orders. These birds are not primitive, however, having undergone a high degree of morphological divergence related to loss of flight. This makes them a good system in which to understand the influence of skeletal reduction on ossification sequence. These birds are also widely variable in terms of body size, ranging from the chicken- sized kiwi to the largest extant bird, the Ostrich. Palaeognath skeletal development is relatively unknown, in spite of their important phylogenetic position. Here, I describe the skeletal development of the Emu {Dromaius novaehollandiae), Ostrich {Struthio camelus), Greater Rhea (Rhea americana), and Elegant Crested-Tinamou (Eudromia elegans), focusing on ossification. All of these taxa are characterized by element loss in the appendicular skeleton, but there are several developmental mechanisms through which this loss occurs including failure to chondrify, failure to ossify and fusion of cartilages prior to ossification. Evidence is presented here to support the reduction in size of skeletal elements resulting in a delay in the timing of ossification. This study provides an important first look at the timing and sequence of ossification in palaeognathous birds, and discusses the influence of changes in the pattern of skeletal development on morphological evolution.

27 CHAPTER 2

Skeletal development in palaeognathous birds (A portion of this chapter is from Maxwell, E.E. Comparative ossification and development of the skull in palaeognathous birds (Aves: Palaeognathae). Zoological Journal of the Linnean Society. In Press)

Ossification occurs as a species-specific pattern of bone formation, which spans the later 2/3 of embryogenesis and continues after hatching in birds (Hogg, 1980; Starck, 1989). Timing of bone formation is influenced by different factors depending on the taxon and element being considered, resulting in a pattern of incredible complexity, as will be discussed throughout this thesis. The size of the embryo, sequence of chondrification and source of osteogenic cells appear to be of minimal importance in determining ossification sequence within an order (Maxwell, 2008). Altriciality and precociality also do not seem to exert a large influence on ossification sequence in birds (Starck, 1993). Ratites are large flightless birds, now restricted to a Gondwanan distribution. They are generally assumed to be monophyletic and are positioned basally within crown-group Aves (Gibb et al., 2007; Livezey and Zusi, 2007; Slack et al., 2007). Ratites are characterized by a suite of morphological features thought to be correlated with gigantism and loss of flight, as well as by a palaeognathous palatal morphology (Cracraft, 1974). Whether flight has been lost once or multiple times in the clade is intensely debated (Cracraft, 2001; Briggs, 2003). Tinamous are the sister group of the ratites (Livezey and Zusi, 2007; Slack et al., 2007), and although they share a palaeognathous palatal morphology, they are weakly flighted and correspondingly much smaller in size. Paedomorphosis has been suggested to be one factor underlying many of the morphological peculiarities of ratite birds (Livezey, 1995). As many of the supposedly juvenile features involve characteristics of the skull and palate, the timing and sequence of developmental events relative to neognaths is especially relevant. Ratites also provide an excellent system in which to study the relative influence of adult morphology on ossification sequence, as they are so

28 morphologically diverse that they have been classified in separate orders rather than separate families by some authors (Wetmore, 1960). The skeletal development of ratite birds is not well known: much of the research has focused on the formation of discrete anatomical regions during very early stages of embryogenesis (typically pre-ossification). The early skeletal development of the Emu {Dromaius novaehollandiae) was examined by Lutz (1942); sampling was limited. The early development of the appendicular skeleton of the Ostrich {Struthio camelus) has been examined multiple times (Broom, 1906; Feduccia and Nowicki, 2002; Kundrat et al., 2002). The development of the skull of the Ostrich has been studied in detail (Parker, 1866; Webb, 1957); those of the Greater Rhea {Rhea americand) and Emu have also been examined (Kesteven, 1942; Muller, 1963). Glutz von Blotzheim (1958) described the development of the pectoral and pelvic girdles in 5". camelus, R. americana and D. novaehollandiae; sample sizes were small. The genesis of the entire skeleton of kiwis has been studied. Although the description was based on a composite of several species, it provides comparative information for the development of the genus Apteryx, which is now critically endangered and not amenable to embryological work (Parker, 1891, 1892). A partial skeleton of a late-stage embryo of the extinct Elephant Bird (Aepyornis sp.) has also been described (Balanoff and Rowe, 2007). The skeletal development of ratites remains understudied, but is especially important in addressing questions of homology and polarity in the ontogeny of extant birds. Here, I describe the ossification and skeletal development of the Emu (Dromaius novaehollandiae) and the Ostrich (Struthio camelus) from a well- sampled series of pre-hatching embryos. I supplement these observations with data from the Greater Rhea (Rhea americana) and the Elegant Crested-Tinamou (Eudromia elegans), which are derived from more poorly sampled series. This represents the most thorough description of skeletal development in palaeognathous birds to date, and focuses on the entire skeleton rather than on restricted anatomical modules in order to discuss the effects of localized skeletal reductions or expansions on ossification sequence.

29 MATERIALS AND METHODS

Eggs of the Emu {Dromaius novaehollandiae (Latham, 1790)) were purchased from Hunter Farms (Oshawa, Canada), and eggs of the Greater Rhea {Rhea americana L.) were purchased from Pfeffer Rhea Farm (St. Thomas, Canada). Both species were artificially incubated under conditions normal to these taxa (36.5°C and 40% humidity). This resulted in a sample size of n=35 viable embryos (Emu) and n=10 viable embryos (R. americana). Embryos were fixed in 10% neutral buffered formalin and staged using a normal table designed for Gallus gallus (Hamburger and Hamilton, 1951). As the last four stages (40-44) are based on the length of the beak and the third toe of G. gallus and are therefore not applicable to other species, these stages are referred to as "40+" in this paper. Specimens were then cleared and double stained for bone and cartilage, following the procedure outlined by Dingerkus and Uhler (1977). Late stage embryos were skinned, eviscerated, and soaked in a mixture of the histological clearing agent Citrisolve and ethanol in order to remove subcutaneous and intermuscular fat deposits prior to staining. Ossification sequences presented for D. novaehollandiae and R. americana are based on personal observation of specimens housed in the Redpath Museum (RM; Appendix 1). Cleared and stained Ostrich (Struthio camelus L.), Elegant Crested- Tinamou (Eudromia elegans),a second sample of Emu embryos, several late- stage Chilean Tinamou embryos (Nothoprocta perdicarid) and one late-stage Greater Rhea were prepared by C. Marshall (See Marshall, 2000 for details) and are housed in the Peabody Museum (YPM; Appendix 1). Embryos were not staged prior to clearing and staining so ossification sequence is described by day of incubation for these species. Ossification is described in the text by anatomical region, then by species, and by stage (Table 2.1) or day; Table 2.2 reflects the sequence of ossification of the entire embryo. Using ranked ossification sequence data allows for a direct comparison between the Ostrich and the Emu, independent of differences in incubation period. Because little is known about population-level variation in ossification sequences, the two samples of Emu are reported separately (Table

30 2.2). Such variation has previously been reported for turtles (Sheil and Greenbaum, 2005). The description in the text is based on the Redpath Museum specimens unless otherwise noted, and anatomical nomenclature follows Baumel and Witmer (4993).

RESULTS

Skull Dromaius novaehollandiae Stage 31. No ossification has occurred. The trabecula communis extends under the eyes, but not far rostral to them resulting in a very short prenasal process. The process is slightly flexed ventrally. Meckel's cartilage is in articulation with the quadrate cartilage, but is short as the beak is only weakly developed. The quadrate cartilage is triradiate. The infrapolar process is small, and is located dorsal to the quadrate cartilage. The pars canaliculi of the auditory capsule are round in shape and fully chondrified. The ceratohyal and hypohyal are fused to form a curved element directly posterior to the quadrate cartilage.

Stage 32. There is no ossification present in the skull. The prenasal process exceeds Meckel's cartilage in length, and remains triangular in lateral view. It remains slightly flexed ventrally. The parietotectal cartilage is growing posteriorly over the nasal capsule, and has flattened dorsoventrally. The postorbital cartilage has formed behind the orbit. The pars canaliculi of the auditory capsule have elongated dorsally, becoming more ovoid than in the previous stage. They are in close contact with the quadrate cartilage anteroventrally. The columella is situated posterior to the quadrate cartilage. Meckel's cartilage is considerably more elongate than in the previous stage, and has developed a retroarticular process. The basibranchial portion of the hyoid apparatus is approximately equal to the retroarticular cartilage in posterior extent.

31 Stage 33. The skull remains unossified. The prenasal process has straightened and forms a distinct angle with the posterior nasal septum, although outgrowth of the beak has been limited. Meckel's cartilage remains considerably shorter than the prenasal process. The infrapolar process has increased in prominence. The foramen for the ophthalmic artery is oval, with its long axis parallel to the long axis of the skull.

Stage 34. The prenasal process is slightly more elongate than in the previous stage, but the greatest amount of growth has been in the lower jaw, which is now almost equal to the prenasal process in anterior extent. The two rami of Meckel's cartilage share a broad contact. The area between the parietotectal cartilage and the prenasal process is not well chondrified. The external narial opening cuts a trough in the underlying trabecula, but does not form a discrete perforation in the nasal septum due to weak chondrification. The angular is ossifying. Later in this stage (Fig. 2.1 A), the squamosal ossifies around the quadrate articulation. The palatine, pterygoid, jugal and quadratojugal are also ossifying. In the lower jaw, the dentary and supra-angular ossify. The ossification of the jugal is variable, as it is absent in some stage 35 individuals (RM 8021, RM 8053).

Stage 35. The jugal process of the premaxilla is present. The angle between the prenasal process and the nasal capsule has decreased relative to stage 34. The base of the interorbital cartilage curves ventrally to a point just rostral to the external nares before it flattens out to form the prenasal process. The external narial opening is slit-like, with its long axis parallel to the prenasal process. The perforation in the underlying cartilage corresponding to the narial opening is complete! Later in this stage, the parasphenoid rostrum, palatal and frontal processes of the premaxilla, the vomer, and lacrimal are ossifying. The jugal and frontal processes of the premaxilla ossify from separate centers, as do the dorsal and ventral portions of the orbital process of the lacrimal. The dorsal ossification center of the orbital process of the lacrimal forms slightly before the ventral center. The orbital

32 process forms in contact with a cartilaginous lateral extension of the ectethmoid. The vomer is ossifying from paired centers. The lacrimal, parasphenoid rostrum and frontal process of the premaxilla ossify variably, as they are occasionally unossified in early stage 36 individuals (RM 8022).

Stage 36. The beak has outgrown to the point where the angle between the cartilaginous prenasal process and the nasal septum has been eliminated. The splenial is ossifying along the medial margin of the lower jaw. The jugal and palatal processes of the maxilla are ossifying. The jugal and frontal processes of the premaxilla retain separate ossification centers. This is followed by the ossification of the nasal from a single center lying along the roof of the nasal capsule. The prearticular ossifies along the posterior medial margin of the lower jaw. The squamosal is long and thin, forming a spur paralleling the otic process of the quadrate cartilage and also forming an arch around the external auditory meatus. The dentary is ossifying from multiple centers along the anterior portion of the lower jaw. The hyoid apparatus remains short and stout. Late in this stage, the basisphenoid ossifies. The two ossification centers of the lacrimal remain unfused, as do the two ossification centers of the premaxilla.

Stage 37 (Fig. 2.2E). The frontal is ossifying along the dorsal margin of the orbit. The body of the quadrate is ossified. The ossification centers of the lacrimal have fused, as have the premaxillary ossification centers. Later in this stage, the parietal and the parasphenoid alae ossify. The parasphenoid alae form a plate of bone ventral to the postorbital cartilage; this cartilage remains unossified.

Stage 38 (Fig. 2.IB). The premaxilla has completely surrounded the cartilaginous prenasal process. The frontal has expanded its ossified area ventrally into the orbit. The ceratobranchials are ossifying. The laterosphenoid is also ossifying from its ventrolateral corner. This is followed by the ossification of the exoccipitals and parasphenoid lamina. The parasphenoid lamina ossifies from

33 right and left paired ossification centers located posterior to the parasphenoid rostrum and basisphenoid. Otoliths are calcified. The laterosphenoid is variable in its timing of ossification, as it is cartilaginous in some stage 40+ individuals (RM 8047).

Stage 39. The lacrimal has a large foramen on the anterior surface of the descending process. The supraoccipital is ossifying from a single center; this appears to be variable as two centers on either side of the cranial midline were observed in some individuals. The laterosphenoid is ossified along its entire ventral edge. The mesethmoid is ossifying from two centers - one on the anteroventral edge of the interorbital septum, the second located in the lamina dorsalis representing the ossification of the parietotectal cartilage. The lamina dorsalis has extensive dorsal exposure in the Emu, first as cartilage and later as bone. It is bordered rostrally by the frontal process of the premaxilla (which does not contact the frontal), laterally by the nasals and frontals and in later stages, posteriorly by the frontals. The supraoccipital and mesethmoid are variable in their timing of ossification, remaining unossified in some stage 40+ individuals (RM 8047).

Stage 40+. Day 36. The basioccipital is ossifying as a single linear ossification center on the cranial midline. The ossified area of the laterosphenoid has expanded to include the entire lateral edge. The epiotic is beginning to ossify from the medial margin of the supraoccipital. It is not yet exposed on the lateral surface of the skull. The ossified portion of the parietal has expanded to reach the posterior wall of the orbit. There is an extra medial ossification center of the parasphenoid lamina, located between the two principle wings of that element. Day 38. The basioccipital is elongate and diamond-shaped, broadening posteriorly. The laterosphenoid is entirely ossified. The opisthotic ossifies separately from the exoccipital, and is exposed along the lateral wall of the braincase. The ossified portion of the lamina dorsalis of the mesethmoid contacts the frontals posteriorly, but is not overlapped by any elements. The interorbital

34 septum is ossified over the anterior third of the element. The articular is ossified, beginning from the center of the jaw joint. There also appears to be some bony ossicles in the jaw joint that are separate from the articular, and may represent sesamoid elements termed ossicula articularia (Jollie, 1957). Day 43 (Fig. 2.1C). The opisthotic contacts the squamosal, forming the posterior border of the external auditory meatus. Day 45. The prootic ossifies medial to the squamosal. It is not visible in lateral view.

Rhea americana Stage 34 (Fig. 2.11). No ossification is present at this stage. The prenasal process is slightly swollen at its tip, and is only slightly longer than Meckel's cartilage. The roof of the nasal tectum has grown posteriorly, dorsal to the orbits. The pars canaliculi of the auditory capsule is prominent, encompassing the entire lateral posterior margin of the skull. The contact between the two rami of Meckel's cartilage is relatively narrow.

Stage 35. The contact between the two rami of Meckel's cartilage becomes broader in this stage. The squamosal, palatine, vomer, pterygoid, jugal and quadratojugal are ossifying. The lacrimal is ossifying, beginning from its orbital process. The frontal process of the premaxilla is ossifying, as is the jugal process of the maxilla. The supra-angular and angular are ossified in the lower jaw. This is followed by the ossification of the parasphenoid rostrum.

Stage 37 (Fig. 2.2C). The nasal is ossifying, as are the splenial and prearticular and dentary. The jugal process of the premaxilla is ossified, but is not fused with the frontal process. The maxilla is triradiate. There are two independent, parallel ossification centers posterior to the parasphenoid rostrum; these represent the initiation of ossification of the parasphenoid lamina. The squamosal forms a process that parallels the otic process of the quadrate.

35 Stage 38 (late). The parietal and frontal are ossifying, as are the quadrate and the ceratobranchials.

Stage 40+ Day 22. The supraoccipital is ossifying from a single center. The parasphenoid alae are ossified, as is the basisphenoid. The lacrimal is triradiate. Although the nasal lacks a descending process, the ascending process of the maxilla in Rhea americana is much better developed than in the Emu. There is a hole located in the middle of the squamosal, perhaps due to osteological restructuring caused by muscle development, as hypothesized for Meleagris gallopavo (Maxwell, 2008). Day 26 (Fig. 2.1 J). The basioccipital, exoccipital, laterosphenoid, prootic, opisthotic, epiotic and mesethmoid are ossifying. The lamina dorsalis of the mesethmoid is ossifying from paired ossification centers, rather than from a single median center. The prearticular has developed a second ossification center, located along the posterior edge of the articular cartilage. Day 28. The facial region of the skull has become extremely elongate. The articular is ossifying, and the dermal bones of the skull roof are in contact. Day 30. The opisthotic and epiotic have developed large lateral exposures (Fig. 2.2D).

Struthio camelus Day 12. This embryo is poorly prepared, however some general observations are possible. The beak is broad, and the prenasal process does not extend far rostral to the eyes. The distal tip of the prenasal process is directed ventrally. The lower jaw extends only as far as the anterior margin of the orbit. The pars canaliculi of the auditory capsule are ovoid, with the long axes oriented dorsoventrally. They are not in contact posteriorly. Day 15 (Fig. 2.IF). There is no ossification in the skull. The prenasal process is straight and narrow. The roof of the nasal capsule is chondrified. It widens posteriorly, and contacts the orbits laterally. There is an angle formed between the nasal capsule and the prenasal process. The contact between these

36 two structures occurs near the proximal end of the prenasal process. The cartilages of both the upper and lower jaw are elongate, the upper jaw extends further anteriorly than the lower jaw. The lower jaw is y-shaped, with the two rami contacting each other medially along a broad contact. The retroarticular process of Meckel's cartilage extends posterior to the quadrate articulation. The pars canaliculi retain their ovoid morphology, and do not contact each other posteriorly. There is a ventrally directed chondrification on the anterior margin of the pars canaliculi, which is situated ventral to the external auditory meatus. This represents the metotic cartilage, although its morphology is different than in other birds. The definitive occipital arch is ventral and medial to the pars canaliculi, and is distinct from this element. The quadrate cartilage and stapes are present. Day 16. The beak is slightly longer than in the younger embryo. The prenasal process is slightly hooked at the tip, and is also wider. The pterygoid is ossified. The dentary is also ossifying from the ventral surface of the mandibular symphysis. The timing of onset of ossification is variable, with some day 17 embryos lacking ossified skull elements (YPM 112440). Day 17. The maxilla is weakly ossified near the base of the ascending process. The squamosal is ossifying around the quadrate articulation. The quadratojugal, supra-angular and angular are also ossifying. Day 19. The parasphenoid rostrum, parietal, lacrimal, nasal, premaxilla, palatine, vomer, jugal, and splenial are ossifying. The lacrimal, parietal, nasal and vomer are variable in their timing of ossification, being absent in some day 21 embryos (YPM 112444). Day 21. The hyoid apparatus is more elongate, extending posterior to the external auditory meatus. There is a process of the metotic cartilage overlying the quadrate articulation. The squamosal is anterior to this. The prenasal process is not completely enveloped by the premaxilla. Its anterior tip has a spatulate morphology. The frontal is ossifying, as is the ceratobranchial. The ceratobranchials are variable in its timing of ossification, remaining cartilaginous in some day 22 embryos (YPM 112446, YPM 112447). Day 22. The basisphenoid and prearticular are ossified. The premaxilla completely covers the prenasal process. The basisphenoid is variable in its timing

37 of ossification, being absent from some day 23 to day 25 embryos (YPM 112448, YPM 112450). Day 23 (Fig. 2.1G). The quadrate is ossifying (Fig. 2.2A); this is variable as it is cartilaginous in some day 24 embryos (YPM 112449). The maxilla lacks a distinct ascending process, although it is triangular at its midpoint. The nasal lacks a descending process. Day 26. The basioccipital is ossifying from paired linear ossification centers along the cranial midline. The parasphenoid alae and lamina are now ossified. Day 28. The two ossification centers of the basioccipital have fused into a single oblong element (Fig. 2.2B). The supraoccipital is ossifying, as are the laterosphenoid, exoccipital and prootic. Both the lamina dorsalis and interorbital septum of the mesethmoid are ossified. The prearticular extends along the medial surface of the lower jaw, and posteriorly along the anterior portion of the medial process. The nasal has an ossified descending process. The ossification of the prootic and the lamina dorsalis of the mesethmoid are variable in timing of occurrence, and remain cartilaginous in some day 30 - 32 embryos (YPM 112454, YPM 112455, YPM 112458). Day 30. There is a large patch of reduced ossification in the center of the squamosal, corresponding to changes in bone architecture. This is similar to what was seen in stage 40+ (day 22) Rhea americana, and stage 39 Meleagris gallopavo (Maxwell, 2008). Day 31. The stapes is ossified. This is followed by the ossification of the opisthotic and epiotic. The epiotic ossifies from the anterolateral margin of the supraoccipital. The calvarium is very heavily ossified, with all dermal roofing elements contacting each other (Fig. 2.1H). Day 36. The nasal trabeculae are ossifying, but this event is very variable in timing and they remain cartilaginous in some day 37-38 individuals (YPM 112462, YPM 112464, YPM 112465). Day 38. The articular is ossifying from the dorsal surface of the medial process.

38 Eudromia elegans Day 9. There is no cranial ossification. The two rami of Meckel's cartilage have a broad contact, as in ratites. Day 10. The prenasal process is curved ventrally, as is the area of the lower jaw where the rami of Meckel's cartilage are in contact. Dorsal outgrowth of the trabecula communis in the area of the nasal capsule initiates the formation of the parietotectal cartilage and the preoptic root of the orbital cartilage. The jugal and frontal processes of the premaxilla are ossified, as is the quadratojugal. Day 11 (Fig. 2.ID). The prenasal process is straighter than in younger embryos, and only a slight terminal hook remains. The squamosal, premaxillary process of the nasal, maxilla, palatine, vomer, and jugal are ossifying, as are the dentary, supra-angular, angular and splenial. This is followed by the ossification of the parasphenoid rostrum and lamina, the basisphenoid, parietal, frontal, lacrimal, pterygoid, quadrate, prearticular and ceratobranchials. The ossification of the quadrate is variable, as it is does not always ossify before day 12 (YPM 112522). The premaxilla completely covers the prenasal process. The descending process of the nasal does not contact the ascending process of the maxilla. The beak forms a rigid ossified framework, but the calvarium is not well ossified. Day 12. The basioccipital is ossifying Day 14. The exoccipital, laterosphenoid, prootic, opisthotic, epiotic and mesethmoid are ossifying (Fig. 2.IE, 2.2F). The supraoccipital is ossifying from a single center. The mesethmoid lacks a broad dorsal exposure.

39 Figure 2.1. Lateral view of the skull of palaeognath embryos. A-C: Dromaius novaehollandiae. A: Stage 34 (day 22 of incubation, RM 8020). B: Stage 38 (day

32 of incubation, RM 8030). C: Stage 40+ (day 43 of incubation, RM 8039). D-E:

Eudromia elegans. D: Day 11 of incubation (YPM 112520). E: Day 15 of incubation (YPM 112525). F-H: Struthio camelus. F: Day 15 of incubation

(YPM 112437). G: Day 23 of incubation (YPM 112448). H: Day 35 of incubation (YPM 112465). I-J: Rhea americana. I: Stage 34 (day 14 of incubation, RM 7217). J: Stage 40+ (day 26 of incubation, RM 7223). Grey shaded regions represent cartilage; black regions represent ossified tissue. The density of stippling reflects the relative degree of ossification. Scale bar equals

5mm.

40 * • •. ( j i\

^ :^35P-L^&:

^ ^ ^

<#**• ^ 3^=^

- •-- ^L-' '< '-if Figure 2.2. Palatal view of selected palaeognath embryos. A, B: Struthio camelus

(modified from Parker, 1866). C: Rhea americana, stage 37 (day 17 of incubation, RM 7219). D: Rhea americana (modified from Muller, 1963). E:

Dromaius novaehollandiae, stage 37 (day 28 of incubation, RM 8026). F:

Eudromia elegans, day 14 of incubation (YPM 112524). Scale bar equals 5mm.

41

Postcranial axial skeleton Dromaius novaehollandiae Stage 31. The cervical and dorsal ribs are chondrified, as have the anterior three sternal rib segments. The dorsal ribs have not separated from the vertebral bodies. The sternal plates are absent. There are approximately 18 post-sacral (24 post-acetabular) vertebrae chondrified.

Stage 32. The postcranial axial skeleton remains entirely cartilaginous during this stage. The dorsal ribs are more elongate than in the previous stage, and four now have sternal segments. These segments articulate both with the dorsal ribs and with the sternal plates. The sternal plates are faintly chondrified, and are unfused. The tail is shorter than in the previous stage, with approximately 13 post- sacral centra (21 post-acetabular centra).

Stage 33. There is no ossification in the postcranial axial skeleton. The sternal plates extend from the ventral edge of the coracoid to the most posterior dorsal rib, and are hot fused.

Stage 34. The cartilaginous cervical ribs progressively increase in length posteriorly. The dorsal ribs are more robust than in previous stages, and the sternal plates are partially fused beginning from the anterior end. The number of post-sacral vertebrae has decreased: 11 are now present. Later in this stage, the second-most anterior dorsal rib begins to ossify. The uncinate processes on the fifth and sixth dorsal ribs are chondrified. Nine cartilaginous post-sacral vertebrae are present. The ossification of the dorsal ribs, as well as the chondrification of the uncinate processes is variable; these events have not occurred in some stage 35 and early stage 36 individuals (RM 8045, RM 8022, RM 8053).

Stage 35. In some individuals, an extra dorsal rib is present and in this case this element is partially covered by the ilium. Late in this stage, the number of cartilaginous post-sacral centra is reduced to eight and the sternal plates fuse along their entire length.

42 Stage 36. In some individuals of this stage only a single uncinate process is present. The remainder of the dorsal ribs ossify from anterior to posterior. The number of free caudal centra is reduced to seven. Late in this stage, the number of free caudal centra is reduced to the adult count of six.

Stage 37. Neither of the most posterior two dorsal ribs has a sternal segment, and both are relatively short. In some individuals at this stage, three uncinate processes are chondrified. Uncinate processes appear to be variable both in absolute number and in timing of appearance in the Emu embryo.

Stage 38. Cervical centra 2-4 are ossifying. The thoracic centra are also ossifying, beginning at T3. The fourth and fifth sternal ribs are ossified; these are the most posterior sternal ribs. Later in this stage, the atlantal arch begins ossifying, as do the most posterior two (and largest two) cervical ribs. The synsacral centra located under the medially contacting anterior alae of the ilia are ossified. ' There is variability in the timing of ossification of the cervical arches and cervical ribs; in some stage 40+ individuals these are not yet ossified (RM 8047).

Stage 39. By this stage all cervical centra are ossifying, with the exception of the atlas. The atlantal arch is strongly ossified, the axial arch less so. No other cervical arches are ossified. All of the thoracic centra are ossifying; all of the synsacral centra are as well although the most posterior centra retain right and left paired ossification centers. Later in this stage, the third most posterior cervical rib ossifies, as do the second and third sternal ribs. The two most anterior free caudal centra are ossifying; they have right and left paired ossification centers as do the most posterior three synsacral centra.

43 There is a great deal of variability surrounding the ossification of the free caudal centra, with these elements remaining cartilaginous in many stage 40+ individuals (RM 8034, RM 8035, RM 8036, RM 8047, RM 8048).

Stage 40+ Day 36. All cervical arches are weakly ossified around the base of the transverse processes; only the atlantal arch is strongly ossified. The thoracic arches are also all ossified around the base of the transverse processes. Day 38. The cervical ribs are all ossified. All sternal ribs are also heavily ossified. Day 42. Five free caudal centra are ossifying from dorsal and ventral ossification centers. Day 44. The body of the sternum is ossifying from a center located lateral to the sulcus articularis coracoideus and posterior to the processus craniolateralis. The dorsal ribs in the Emu have ossified a laterally compressed region immediately distal to the capitulum. This was initially a cartilaginous flange off the posterior margin of the rib. Day 45. The cervical arches are now heavily ossified; there is an ovoid gap in the ossified region around the base of the neural spine, which remains cartilaginous. Day 46. All free caudal centra are ossifying, followed by the anterior vertebral component and linear notocordal component of the pygostyle ossifying as two separate centers. The posterior six post-acetabular synsacral transverse processes are ossifying. Day 47. The body of the sternum is extensively ossified ventral to the articular facets for the sternal ribs. The number of ribs in articulation with the sternum is intraspecifically variable in the Emu, and ranges between four and five. Day 48. All synsacral transverse processes are ossifying. The pygostyle retains two ossification centers. In the most ossified individual examined (YPM Emu 25) the two large sternal ossifications remain unfused medially. All axial other elements are ossified.

44 Rhea americana Stage 34. The postcranial axial skeleton remains entirely cartilaginous. The sternal plates are fused anteriorly, but remain unfused posteriorly. There are three sternal ribs and two uncinate processes chondrified. 15 post-sacral centra vertebrae are present.

Stage 35. 10 post-sacral vertebrae are chondrified.

Stage 37. Thoracic centra T3-T6 are ossifying from right and left paired ventral ossification centers. The dorsal ribs are cartilaginous. The sternal plates are now fused along their entire length. Four uncinate processes are chondrified.

Stage 38 (late). The cervical centra are ossifying from ventral ossification centers. All of the dorsal ribs are ossified. The number of post-sacral centra remains at 10; this higher count is due to the reduced length of the posterior ilium in Rhea americana.

Stage 40+ Day 22. The cervical centra have developed dorsal ossification centers. The thoracic centra are completely ossified. The laterocranial process of the sternum is also ossifying, although it remains cartilaginous in some day 30 embryos. Day 24. The most posterior cervical ribs have ossified, as have the synsacral centra. Day 26. The cervical arches have begun to ossify. The atlantal arch is not significantly more ossified than the other cervical arches. The more anterior cervical ribs are now ossifying. Day 28. The thoracic arches are ossifying beginning from the transverse processes. The sternal ribs are also ossified. Day 30. The body of the sternum is ossifying from its lateral margin, immediately posterior to the last sternal rib. The synsacral transverse processes are ossified.

45 YPM Rhea 1. The posterior two uncinate processes are ossifying. The ossification centers of the sternum have expanded, but do not yet meet along the midline. The laterocranial processes of the sternum are heavily ossified from separate ossification centers.

Struthio camelus Day 12. There is no ossification in the postcranial axial skeleton. The dorsal ribs are chondrified, but are not separated from the vertebral cartilages. Sternal plates are present, but unfused. Day 15. The postcranial axial skeleton is entirely cartilaginous. The sternal plates remain unfused. Both the dorsal ribs and sternal segments are distinct and separate from each other. Cervical ribs remain small, if present. 16 post-sacral centra are chondrified. Day 16. There is still no ossification in the postcranial axial skeleton. The sternal plates are fused over the anterior 2/3 of their length. The number of post- sacral centra has decreased to 11. Day 17. No uncinate processes are chondrified. 10 cartilaginous post- sacral vertebrae are present. Day 20. The dorsal ribs are ossified, but this event occurs variably between days 20 and 23. Cartilaginous cervical ribs are short but present. Day 21. Cervical ribs are present and normal. A single cartilaginous uncinate process is present on the fourth dorsal rib, of which there are a total of nine. This is followed by the appearance of a second process on the third dorsal rib. Nine post-sacral centra are present. Day 23. Three uncinate processes have chondrified, on dorsal ribs 3-5. The thoracic and pre-acetabular synsacral centra are ossifying, although they remain cartilaginous in some day 24 embryos. The thoracic centra are ossifying from dorsal and ventral ossification centers. The sternal cartilages are fused along their entire length. Day 24. The number of post-sacral centra has been reduced to seven. Day 25. The cervical centra are now ossifying from dorsal and ventral ossification centers.

46 Day 28. The cervical centra are completely ossified, except for the atlas and axis, which have only the ventral ossification center. The cervical arches are ossifying, with the exception of CI and C2. The cervical ribs are ossified. The thoracic transverse processes are also ossifying, with the degree of ossification decreasing posteriorly: Ossification of the transverse processes begins at Tl. The synsacral centra posterior to the acetabulum retain dorsal and ventral ossification centers. Day 30. Only the ventral ossification center is present in the atlas. The axis now has a dorsal ossification center as well, but the two centers are not fused. The free caudal centra are ossifying from a dorsal ossification center, followed by the appearance of a yentral ossification center. Day 31. The cervical arches are completely ossified except for an ovoid gap surrounding and including the base of the cartilaginous neural spine. All six free caudal centra are ossifying from dorsal and ventral centers. Day 34. The thoracic neural spines remain cartilaginous. The synsacral transverse processes are ossified from the acetabulum to the widest point of the sacrum. The free caudal transverse processes and pygostyle are also ossifying. The transverse processes of the free caudal centra and the pygostyle remain cartilaginous in some day 35 to 38 embryos (YPM 112460, YPM 112464, YPM 112465). Day 35. Seven free caudal centra are ossifying (YPM 112460), implying intraspecific variability in the number of free caudal centra in the Ostrich Day 36. The synsacral arches are ossifying, although they can remain cartilaginous in some day 38 embryos (YPM 112464, YPM 112465). Day 37. The pygostyle has a linear ossification center that connects the separate ossification centers of the vertebrae making up this element. Cartilaginous uncinate processes persist. The second most posterior sternal rib (number 5 of 6) is ossifying, although it is not ossified prior to hatching in all embryos (YPM 112466).

Eudromia elegans

41 Day 10. The postcranial axial skeleton is entirely cartilaginous. Seven post-sacral centra are present. Day 11. The dorsal ribs are ossifying. The cartilaginous cervical centra have a large ventral foramen located between the two halves of the vertebral body. The two halves of the centrum are unfused. The posterior thoracic centra and anterior synsacral centra are ossifying from right and left paired ossification centers. Day 14. No uncinate processes are present. The cervical centra and cervical arches are ossifying. The atlantal arch is ossified, as are the arches C4 and posterior. All thoracic centra and synsacral centra are ossifying, retaining right and left paired ossification centers posterior to the acetabulum. Day 15. All cervical arches are ossifying; the atlantal arch is more ossified than the arches of C2 or C3. The cervical ribs are also ossified, as are thoracic transverse processes T1-T6.

Forelimb Dromaius novaehollandiae Stage 31. The humerus, radius, ulna and metacarpal IV are chondrified. A large chondrification proximal to the humerus is the precursor of the scapulocoracoid.

Stage 32. The forelimb is entirely cartilaginous. The scapula and coracoid are chondrified, but not as separate elements. The glenoid opens laterally. No carpals are visible, and there is no flexure at the wrist. Three manual digits are chondrified, with the middle digit being the longest. The three digits completely fill the autopodial region of the limb, which is not significantly wider than the zeugopodium. No phalanges are visible.

Stage 33. The scapula and coracoid are separate chondrifications in some individuals of this stage (RM 8050, RM 8051). Although there is a large cartilaginous mass in the carpal region, no distinct carpal elements are present, nor is there any flexion of the wrist. The handplate is more splayed than in stage

48 32. The middle digit is the longest of the three chondrified digits. One phalanx is present per digit.

Stage 34. The scapula and coracoid appear to be variably fused, as both states are observed during this stage. The humerus has a faint collar of ossified tissue around the midpoint of the shaft. A semi-lunate shaped carpal mass caps metacarpal III. The hand plate has further elongated. Three cartilaginous digits are present. Digit II is relatively long, thin and only faintly chondrified. Digit IV is robust, and has two phalanges, the most distal of which is quite faint. Digit III also has two phalanges. The handplate is splayed posteriorly, with digit II being parallel to digit III.

Stage 35 (Fig. 2.3A). Although the zeugopodium remains cartilaginous, metacarpal III is ossified. The radiale is distinct, flattened between metacarpal II and the radius. The semilunate carpal mass capping metacarpal III is also present, and the ulnare is not visible. Metacarpal II is a long thin splint of cartilage. Metacarpal IV is robust, and extends distal to the ossified area of metacarpal III. The phalanx is not in articulation with the metacarpal. The ossification of metacarpal III is variable; it remains cartilaginous in some stage 36 embryos (RM 8022).

Stage 36. Early in this stage, the scapula and furcula ossify. The furcula ossifies from two rami; these do not contact each other at any point in development. The furcula does not contact the coracoid, being located anterior and medial to it. The position of the furcula is variable, since in some individuals of this stage the rami do not extend medially past the medial edge of the coracoid. The radiale remains distinct from the rest of the cartilaginous mass in the carpal region. Metacarpal II is in contact with the carpometacarpus, and is quite robust. Later in this stage, the radius and ulna ossify (Fig. 2.3B). The hand plate is long and narrow, and all elements are parallel. Metacarpal II is a thin cartilaginous splint. Metacarpal III has three phalanges. Metacarpal IV is short and has one phalanx, which is not in articulation. In some embryos, metacarpal IV may have

49 two phalanges (RM 8046); in this case the most distal is fused to the carpometacarpus. An alternative morphology of metacarpal II is a rounded remnant of cartilage not in articulation with the carpometacarpus (RM 8025).

Stage 38. The coracoid ossifies from a separate ossification center than the scapula. The semilunate carpal mass has fused to the carpometacarpus. Metacarpal IV is elongate, over 50% of the length of metacarpal III, but this is a variable morphology. The coracoid is variable in its time of ossification, remaining cartilaginous in some stage 40+ individuals (RM 8047).

Stage 40+. The radiale is distinct and separate, but during this stage it partially fuses to the carpometacarpus. The portion of the distal phalanx (3) of metacarpal III that is covered by the keratinous sheath ossifies (Fig. 2.3C). This is followed by the ossification of metacarpal IV in embryos having the elongate morphology of this element. It fails to ossify prior to hatching in embryos with the short morphology. Lastly, the proximal phalanx of metacarpal III ossifies.

Rhea americana Stage 34 (Fig. 2.31). There is no ossification in the wing. The scapula and coracoid are fused. There is a prominent procoracoid process, although it is questionable whether this derives from the scapula or the coracoid. It does not grow further ventrally, as seen in the Ostrich. The radiale is prominent in the carpal region. The pisiforme is triangular, and is located on the posterior ulna. The distal carpals form a semilunate mass, but aren't fused to the metacarpals. There are two phalanges on digits II and III, and one phalanx on digit IV.

Stage 35. The furcula is not present in Rhea. The humerus, radius and ulna are ossified. The semilunate mass is in the process of fusing to the metacarpals, and there are two phalanges present on all manual digits.

50 Stage 37. Metacarpals III and IV are ossifying.

Stage 38 (late). The scapula and coracoid are ossifying from independent ossification centers. The ulnare is an extremely prominent element in the carpal region, approaching the radiale in size. The proximal phalanx of digit II is ossifying, although it remains cartilaginous in some stage 40+ embryos. The two phalanges of digit IV have fused to form a single element.

Stage 40+ (Fig. 2.3J). ' Day 30. The distal phalanx of digit II is ossified, as is the proximal phalanx of digit III and the only phalanx of digit IV. YPM Rhea 1. The scapular and coracoidal ossification centers have not yet fused, and metacarpal II remains cartilaginous. All manual phalanges are ossified.

Struthio camelus Day 12. The wing bud is considerably smaller than the hind limb bud, even at this early stage. Day 13. The forelimb is unossified. The humerus, radius and ulna are chondrified. One digit is chondrified; it presumably corresponds to metacarpal IV. Day 15 (Fig. 2.3D). The forelimb remains entirely cartilaginous. The scapula and coracoid are not separate cartilages. The scapula is very short relative to the length of the wing, spanning only two ribs. The procoracoid process of the coracoid is absent. In the carpal region, the radiale is prominent. The semi-lunate carpal mass caps the metacarpals. The ulnare is not visible. Three digits are chondrified in the manus. Digits III and IV are of equal length, and digit II is splayed rather than lying parallel to the other two. All manual digits have one cartilaginous phalanx. Day 16. The furcula is not ossified, nor does it ossify prior to hatching in Struthio camelus. The scapula and coracoid are separate cartilages. A projection from the anteromedial margin of the coracoid represents the initiation of chondrification of the procoracoid process. The humerus, radius and ulna are

51 ossifying. The radiale and semi-lunate carpal mass are visible; an ulnare is not observed. The phalangeal count is 1-2-1. Digit IV is separated from digit III distally, with its phalanx forming a distinct projection on the posterior margin of the hand plate. Day 17. The scapula has elongated posteriorly, and now spans four ribs. The procoracoid process is complete and extends from the anterior coracoid to the sternum, however it remains unfused to the sternal edge of the coracoid. Day 19. The semilunate carpal mass is large, and is not fused to the metacarpals. Metacarpals III and IV are ossifying, but remain cartilaginous in some day 22 embryos (YPM 112446). The manual phalangeal count has increased to 2-3-2. Day 21 (Fig. 2.3E). The cartilages of the scapula and coracoid are fused. Manual phalanges 111:1,2 are ossifying, although 111:1 remains cartilaginous in some day 22-23 individuals (YPM 112446, YPM 112448), and 111:2 remains cartilaginous in some day 23-24 individuals (YPM 112448, YPM 112449). Day 22. The scapula is ossifying. Day 24. Phalanx 11:1 is ossifying. Day 28. The main shaft of the coracoid is ossified; the procoracoid process remains cartilaginous until after hatching. Phalanx IV: 1 is ossifying. Day 30. Phalanx 11:2 is ossifying, although it is not always ossified prior to hatching. Day 34.111:3 is ossifying, although it is not always ossified prior to hatching. Day 35. Metacarpal II is ossifying as a perichondral ossification around the shaft, beginning from the posterior surface of the element. Day 36 (Fig. 2.3F). Phalanx IV:2 is ossifying, although it is not always ossified prior to hatching.

Eudromia elegans Day 9. The humerus, radius and ulna are ossifying. The phalangeal count is 1-2-1.

52 Day 10 (Fig. 2.3G). The area where the furcula would have been is damaged, so it is unknown how early it ossifies. The scapula and metacarpal III are ossifying. The radiale and semi-lunate carpal are visible, although the latter is an oblong mass proximal to metacarpal III rather than a semi-lunate chondrification capping the metacarpal. The radiale has an anterior projection at its proximal end, rather than having a simple rectangular morphology. The ulnare is not visible. Day 11. In the carpal region, the semilunate carpal mass has fused to metacarpal II. In addition, there are now two posterior elements visible: a distal carpal just proximal to metacarpal IV, as well as a posteriorly offset proximal element (the pisiform). The furcula and metacarpal IV are now ossifying. This is followed by the ossification of manual phalanges 11:1,111:1,2. The semilunate mass has fused to the rest of the carpometacarpus. The distal carpal has fused to the pisiform to form the ulnare complex. Day 14. The coracoid is ossifying (Fig. 2.3H).

53 Figure 2.3. Lateral view of the forelimbs of palaeognath embryos. A-C:

Dromaius novaehollandiae. A: Stage 35 (RM 8053). B: Stage 36 (day 25 of incubation, RM 8023). C: Stage 40+ (day 43 of incubation, RM 8039). D-F:

Struthio camelus.D: Day 15 of incubation (YPM 112437). E: Day 21 of incubation (YPM 112444): F: Day 36 of incubation (YPM 112461). G-H:

Eudromia elegans.G: Day 10 of incubation (YPM 112519). H: Day 15 of incubation (YPM 112525). I-J: Rhea americana. I: Stage 34 (day 14 of incubation, RM 7217). J: Stage 40+ (day 26 of incubation, RM 7223). Grey shaded regions represent cartilage; black regions represent ossified tissue. The density of stippling reflects the relative degree of ossification. Scale bar equals

2mm for part A, B, D, E, G, I; 5mm for parts C, F, H and J.

54

Hind limb Dromaius novaehollandiae Stage 31. The ilium is the only element of the pelvic girdle that is chondrified. It is elongate and curved in shape. The femur, fibula and two digits (III and IV) have chondrified. Some diffuse staining is present in the anterior portion of the autopodial region, but distinct digits are not visible.

Stage 32 (Fig. 2.4A). There is no ossification present in the hind limb. The ilium is much wider than in the previous stage, and has expanded both anteriorly and posteriorly. The ischium and pubis are still not visible. The fibula remains in contact with the tarsal region. A large fibulare is present. There are three digits chondrified in the pes; the middle one is longest. Two phalanges are present per digit.

Stage 33. The hind limb remains entirely cartilaginous. The ilium is similar to the previous stage. The ischium lies ventral to it, and roughly parallel with a slight downward curvature. The pubis is directed ventrally. The fibula is no longer connected to the tarsal region, and the proximal tarsals have fused to each other and to the tibia. The putative ascending process of the astragalus and the fibulare can be distinguished. The distal tarsals have fused to each other and to the metatarsals. Digit I is a splint of cartilage on the lateral surface of the tarsometatarsus. The metatarsals are not yet in contact with each other. Two phalanges are present per digit.

Stage 34. The pelvic girdle is cartilaginous. The ilium is broad, with its posterior wing curving ventrally. The ischium is ventral to it, but is straight. The pubis is rotating posteriorly. The femur, tibia, fibula and metatarsals II-IV are ossifying. The ascending process of the astragalus is cartilaginous, and is quite large. Metatarsal I is clearly visible as a thin splint located just distal to the ossified portion of the tarsometatarsus. Three phalanges are well-chondrified on digit III; the fourth is present but has not acquired a distinct morphology. The

55 phalanges of digit IV are not yet noticeably cuboid. Later in the stage, the pubis becomes-parallel to the ischium, and contacts the posterior portion of this element.

Stage 35. The pelvic girdle remains cartilaginous. The anterior ilium reaches the most posterior dorsal rib, but does not overlap it. The ilium thins posteriorly, and curls ventrally, following the curvature of the tail. It contacts the ischium posteriorly. The pubis is directed posteroventrally. Metatarsal I is present, approximately lA way down the tarsometatarsus and distal to the ossified region. It does not lie exactly parallel to the other metatarsals. All of the phalanges have chondrified; those of digit IV do not yet have the cuboid morphology.

Stage 36. The ischium is ossifying, as are pedal phalanges 11:1,3; 111:1,4; and IV:5. Metatarsal I is no longer clearly visible (Fig. 2.4B). This is followed by the ossification of the anterior wing of the ilium from its ventral margin; the posterior wing ossifies slightly later also from its ventral margin. The pubis is ossifying, as are pedal phalanges 111:2 and IV: 1. The ungual phalanges and 111:2 are variable in their timing of ossification, remaining cartilaginous in some stage 37 embryos (RM 8026).

Stage 37. The ilium has expanded anteriorly to cover the two most posterior dorsal ribs.

Stage 38. The entire anterior wing of the ilium is ossified, but the most posterior portion of the posterior wing remains cartilaginous. The ascending process of the astragalus is ossifying; the ossified portion does not extend distal to the most distal ossification of the tibia. Phalanges IV:2-4 are very short and broad. They have lost the constriction at mid-shaft that characterized them during previous stages.

Stage 40+. Day 36 (Fig. 2.4C). The ossified ascending process of the astragalus extends distal to the most distal tibial ossification. At its distal end, it broadens,

56 representing the ossification of the astragalus (tibiale) itself, but there is not a separate ossification center. Day 44. Pedal phalanges 111:3 and IV:2 are ossifying. ' Day 47. A distal tarsal is ossifying. YPM 112503 (Day 41). Pedal phalanx IV:3 is ossifying. YPM Emu 25. This represents the most ossified individual. All of the pedal phalanges are ossified. The crista cnemialis is ossifying from a separate center located within the cartilaginous proximal end of the tibia. It has a grainy texture and is ossifying from the medial surface of the proximal tibiotarsus. Two tarsals are ossified, a single plate-like distal tarsal capping the proximal tarsometatarsus and a fused tibiale and fibulare capping the distal tibiotarsus.

Rhea americana Stage 34 (Fig. 2.4D). The pelvic girdle is cartilaginous. The ilium extends anteriorly over the most posterior dorsal rib. It does not contact the ischium and pubis, but the latter two elements are in contact. The femur, tibia, fibula and metatarsals II-IV are ossified. The fibula is 2/3 the length of the tibiotarsus. The tarsals have fused together to form the tarsometatarsus and tibiotarsus. There is no trace of metatarsal I. All pedal phalanges are present. Those phalanges with a short, square morphology in the adult already exhibit this morphology.

Stage 35. The cnemial crest is prominent, and the ascending process of the astragalus is large.

Stage 37. Both the anterior and posterior wings of the ilium are ossifying from their ventral margins. The ilium is much shorter than the ischium and the pubis in posterior extent. It is not in contact with the ischium or the pubis, both of which are also ossifying. Pedal phalanges 11:1,3,111:1,4 and IV:5 are ossified.

Stage 38 (late). Pedal phalanges 111:2 and IV: 1 are ossified.

Stage 40+

57 Day 26 (Fig. 2.4H). The ascending process of the astragalus is ossifying, , Day 28. Pedal phalanges 111:3 and IV:2 are ossified. Day 30. The ossified portion of the ascending process of the astragalus now extends into the tarsal region. The distal tarsals are ossifying. Phalanges 11:2, and IV:3 are ossified. YPM Rhea 1. Phalanx IV :4 is ossified.

Struthio camelus. Day 12. Three pedal digits are distinct. Digits III and IV are of equal length. Poor preparation makes it impossible to assess chondrification Day 13. The ilium is chondrified, as are the femur, tibia and fibula. Four pedal digits are chondrified. Day 15 (Fig. 2.4E). The hind limb remains cartilaginous. The ilium is thickest anteriorly, and thins posteriorly. It is not closely associated with the axial skeleton posterior to the acetabulum. The ischium and pubis are both thin, and are directed posteriorly. There is no contact between the posterior ends the elements of the pelvic girdle. The tibia is ossifying. The fibula has lost contact with the tarsal region, but remains quite long. It is only slightly shorter than the tibia. The proximal tarsals have fused to each other, and are very closely associated with the distal tibia. The distal tarsals form a flat plate capping the metatarsals, but they have not fused to the latter. The metatarsals are separate, the only contact occurs at their proximal ends. Digit II has a poorly chondrified phalanx. There are three phalanges on digits III and IV. Digit III has a cartilaginous disc intercalated between the proximal phalanx and the metatarsal. Day 16. The posterior ends of the ilium, ischium and pubis are not in contact. The posterior pubis exhibits strong medial curvature. The femur, fibula and metatarsals II-IV are ossifying. The proximal and distal tarsals are fused to the tibia and metatarsals to form the tibiotarsus and tarsometatarsus, respectively. Metatarsal I is a small cartilaginous remnant located near the distal edge of the ossified shaft of the tarsometatarsus. Digit II has one phalanx, a small cartilaginous remnant located distal to the metatarsal. Digit III is much longer than digit IV, and both have four phalanges.

58 Day 17. There is a broad gap between the posterior ilium and ischium, but the posterior ischium is approaching the pubis. The fibula is relatively shorter, being slightly more than half of the length of the tibia. Metatarsal I is no longer visible. Day 19. The ischium contacts the pubis posteriorly, but the broad gap between the ilium and the ischium persists. The pubis is ossifying, as are the proximal phalanges of digits III and IV. The pubis is variable in its timing of ossification, remaining cartilaginous in some day 21 embryos (YPM 112444). Day 20. The ischium is ossifying, as is phalanx 111:2. This phalanx is variable in its timing of ossification, remaining cartilaginous in some day 21 embryos (YPM 112444). Day 21 (Fig. 2.4F). The cartilaginous ilium extends anteriorly over the two most posterior dorsal ribs. Metatarsal II retains a cartilaginous phalanx. Day 23. The ilium overlaps the three most posterior dorsal ribs. Pedal phalanx 111:4 is ossified. Day 24. Both the anterior and posterior wings of the ilium are ossifying, with ossification beginning dorsal to the acetabulum. Day 25. The ascending process of the astragalus is ossifying, beginning at its proximal end. The anterior wing of the ilium is now also ossifying from its ventral margin. None of the elements of the pelvic girdle are in contact posteriorly. Day 28. Phalanx 111:3 is ossifying from a dorsal and a ventral center. IV:2 is ossifying from the dorsal center only. Ossification has expanded over both the anterior and posterior wings of the ilium to include almost the entire element. Day 30. The pectineal process is ossifying from a discrete ossification center. This process is part of the pubis in ratites (Beddard, 1898). Day 31. Phalanx IV:2 has now acquired a separate ventral ossification center. The ascending process of the astragalus extends beyond the ossified end of the tibia into the tarsal region. The dorsal and ventral ossification centers of phalanx 111:3 have merged. Day 34 (Fig. 2.4G). Phalanx IV:3 is ossifying from a dorsal ossification center. Phalanx IV:5 is also ossified, although it is variably cartilaginous in day 38

59 embryos. Two distinct morphologies exist for phalanx IV:5 (tiny or only slightly smaller than the ungual of digit III), and timing of ossification depends on size, with the smaller morphology failing to ossify prior to hatching. Day 36. All pedal phalanges are ossified, although IV:4 is not always ossified in day 38 embryos (YPM 112464, 112465). IV:3,4 retain paired dorsal and ventral ossification centers.

Eudromia elegans Day 9. The femur, tibia and fibula are ossifying. There are three proximal tarsals, a large tibiale and two tarsal elements in the region to be occupied by the fibulare. The distal tarsals form a single cartilaginous mass. The metatarsals are widely separated from each other and remain cartilaginous. Metatarsal I is not present. Each digit has two phalanges. Day 10 (Fig. 2.41). The pelvic girdle remains cartilaginous. The ischium and pubis are directed posteriorly. The two lateral chondrifications in the proximal tarsal region have merged to become the fibulare. The tibiale and distal tarsals capping the metatarsals remain distinct. The metatarsals are ossifying. The proximal phalanges are ossified, as is phalanx 2 digits II and III. Day 11. The ischium is ossifying, as is pedal phalanx 111:3. The tarsals have not fused to the tibiotarsus or to the tarsometatarsus, but are closely associated with those elements. This is followed by the ossification of the anterior and posterior wings of the ilium, dorsal to the acetabular region. All pedal phalanges are ossifying. Day 13. The pubis is ossifying, as is the ascending process of the astragalus (Fig. 2.4J).

60 Figure 2.4. Lateral view of the hind limb and pelvic girdle of palaeognath embryos. A-C: Dromaius novaehollandiae. A: Stage 32 (RM 8052). B: Stage 36

(day 25 of incubation, RM 8023). C: Stage 40+ (day 36 of incubation, RM 8034).

E-G: Struthio camelus. E: Day 15 of incubation (YPM 112437). F: Day 21 of incubation (YPM 112444). G: Day 34 of incubation (YPM 112459). I-J:

Eudromia elegans. I: Day 10 of incubation (YPM 112520). J: Day 15 of incubation (YPM 112525). D,H: Rhea americana. D: Stage 34 (day 14 of incubation, RM 7217). H: Stage 40+ (day 26 of incubation, RM 7223). Grey shaded regions represent cartilage; black regions represent ossified tissue. The density of stippling reflects the relative degree of ossification. Scale bar equals

5mm.

61

Figure 2.5. Evolution of embryonic and adult palatal morphology of the ratites

(phylogeny sensu Bledsoe, 1988, palaeognathous characters 1-3 sensu Zusi and

Livezey, 2006). Top row: adult morphology (Parker, 1869; Parker, 1891;

Beddard, 1898; Simonetta, 1960; Zusi and Livezey, 2006; Silveira and Hofling,

2007). Bottom row: embryonic morphology. Gallus (Jollie, 1957), Rhea, Apteryx

(Parker, 1891) and Dromaius are all HH stage 37; Struthio (Parker, 1866) is slightly older; Aepyornis (Balanoff and Rowe, 2007) and Eudromia (Tinamidae) are late-stage individuals. The palatine is shaded to facilitate comparison.

62 K W A« A /H /lt\

<^#& /I

• Ratitae

Palaeognathae [ - pterygopalatine arc, pterygovomeral arc sutured to the maxilla - lack of a flexible zone in either of the above arcs - bony palate supported by a vomer - parasphenoid and pterygoid - basipterygoid articulation -. pterygoid forms medial to palatine during embryonic development] DISCUSSION

Skull As in most birds, the dermal bones are the first cranial elements to ossify. The ossification of the skull is delayed relative to the long bones, as in Anseriformes, Charadriiformes, and some Galliformes (Chapter 4; Schumacher and Wolff, 1966a; Nakane and Tsudzuki, 1999; Maxwell and Harrison, In Press). In the palatal region, the vomer and parasphenoid rostrum ossify earlier in sequence and in stage in ratites (HH 35 compared to late 36) than in neognath embryos (Anseriformes, Galliformes) (Chapter 4; Maxwell, 2008). There are some confounding factors potentially influencing this relationship: the palate of Anseriformes is considered desmognathous (Beddard, 1898), which is characterized by the possession of a reduced vomer. Galliformes have reduced the vomer even further, and it is entirely absent from some pre-hatching embryos (Beddard, 1898; Maxwell, 2008). As has been hypothesized elsewhere, adult morphology, specifically evolutionary expansion or reduction of elements, exerts an influence on ossification sequence (Haluska and Alberch, 1983; Rieppel, 1993b; Maxwell, 2007; Maxwell and Harrison, In Press). This might make the timing of vomeral ossification in Galloanseres atypical for neognathous embryos, as the vomer in this clade represents a reduced condition. One of the features of the palaeognathous palate is a dramatically expanded vomer (Cracraft, 1974), and thus the opposing evolutionary reduction and expansion in the groups being compared might be increasing the magnitude of the shift observed. The premaxilla ossifies from two distinct centers in both the Emu and Rhea americana, corresponding to the frontal and maxillary processes. The number of ossification centers was unresolved in the Ostrich and Eudromia elegans. The lacrimal also ossifies from two centers in the Emu, a dorsal and a ventral center. Kiwis lack the supraorbital process of the lacrimal, and the orbital process ossifies from a single center (Parker, 1891, 1892). Several skull elements have been reported as ossifying from separate ossification centers in the avian cranium (frontal (Erdmann, 1940), nasal (Goldschmid, 1972; Maxwell and Harrison, In Press), and premaxilla (Maillard, 1948; Starck, 1989)) with various

63 developmental and functional implications proposed. It is clearly not unusual for a single dermal element to have multiple ossification centers. It is unlikely that the two centers observed in either the premaxilla or the lacrimal in ratites are due to differences in the origin of the osteogenic cells between the two centers, as both of these elements are neural-crest derived in their entirety in the Domestic Chicken (Evans and Noden, 2006) and no other source of osteogenic cells has been suggested. A functional explanation as proposed to explain multiple centers of ossification in the snake (Elaphe) maxilla (Richman et al., 2006) and charadriiform nasal (Maxwell and Harrison, In Press) is improbable, as both of the premaxilla and lacrimal are robust and relatively inflexible in the adult as well as in late-stage embryos. Differences in the timing of muscle development might induce bone deposition at multiple sites within a single element, as external forces are known to affect ossification (Glucksmann, 1942; Herring, 1993). In the case of the Emu lacrimal, interaction with the underlying cartilaginous ectethmoid or the developing nasolacrimal duct might stimulate bone formation at a more ventral location of the element, and result in the formation of an upper and lower ossification center. The nasolacrimal duct is entirely surrounded by the lacrimal in the Emu (Witmer, 1995). In spite of approximate similarities in topographical location, it is unlikely that the dorsal and ventral ossification centers are homologous to the lacrimal and prefrontal of other archosaurs as suggested by Erdmann (1940). Theoretically, both the putative lacrimal and prefrontal osteogenic domains may have been conserved, even though the prefrontal bone was reduced and then lost in the archosaur lineage leading to birds. These two domains may be the origins of the two ossification centers present in the Emu, and may represent the ancestral lacrimal and prefrontal contributions to the orbital margin. This possibility is more likely than the acquisition of a de novo ossification center if ossification centers are more easily lost than gained, as has been proposed (Sidor, 2001). Stage 40+ individuals of both the Ostrich and Rhea americana exhibited a change in the structure of the squamosal, resulting in a large patch of reduced ossification in the center of this element. This is similar to what was observed in Meleagris gallopavo, which was attributed to osteological restructuring following

64 increased strain imposed by the developing M. adductor mandibulae externus complex (Maxwell, 2008). Reorganization of bone following tension and pressure changes have been widely reported (Scott, 1957; Herring, 1993). The ossification of the lamina dorsalis and interorbital septum of the mesethmoid from separate centers has previously been reported for the Ostrich (Parker, 1866), and Anseriformes (Maxwell, In Press), and was also observed in the Bobwhite Quail {Colinus virgianus; pers.obs). This pattern appears to be broadly conserved among the ratites, although it could not be confirmed for Aepyornis (Balanoff and Rowe, 2007). The basioccipital ossifies from a single center in the Emu, as previously reported for both Emu and Rhea americana (Kesteven, 1942; Miiller, 1963). It is therefore interesting to note that the basioccipital of the Ostrich clearly ossifies from paired linear centers (also previously reported (Webb, 1957)), as observed in some Galliformes, Anseriformes and Charadriiformes (Maxwell, 2008; Maxwell and Harrison, In Press). It is possible that sampling gaps in the Emu and R. americana series missed the paired ossification center stage, as they fuse to form a single ossification over a relatively short time span. Alternatively, the basioccipital might form either from paired or unpaired centers depending on the species in question, similar to the ontogeny of the supraoccipital. A single center might be a derived condition within ratites. The supraoccipital in the Emu ossifies from paired centers in some individuals, and a single median ossification center in others. This provides additional evidence for the flexibility of this character; intraordinal differences have already been reported in neognaths (Maxwell, 2008). In the Emu, both ossification centers are located in the main body of the element near the midline rather than widely separated on the ventral processes as in Anseriformes (Maxwell, In Press). The ossification of the prootic is greatly delayed in the Emu, relative to neognaths as well as to other ratites. This delay was observed in both Emu populations, with the prootic being the last skull element to ossify, after the opisthotic, epiotic, articular and mesethmoid. Histological studies did not report this delay, and identified the prootic as ossifying prior to the epiotic (Kesteven,

65 1942). This conflict between histological findings and cleared and stained specimens was unexpected, as previous studies reported differences in timing but not in sequence when the two methodologies were compared (Clark and Smith, 1993). Either there is a delay in the mineralization of the prootic, a difference between authors as to the identity of the prootic, or intrapopulational variation in the Emu. Kesteven (1942) also reported that the laterosphenoid was the last element in the braincase to begin ossification, rather than ossifying prior to the otic series as reported here. Major differences between this study and that of Kesteven in the sequence and pattern of ossification of the Emu cranium suggest either that differences are present between the Emu populations examined, or that there are differences in element identification and homology between authors. The skull is very heavily ossified in all of the ratites examined here, as well as in kiwis (Parker, 1891) and Aepyornis (Balanoff and Rowe, 2007). All of the dermal roofing bones contact each other prior to hatching. Both species of tinamous, on the other hand, had relatively weakly ossified skull roofs, with a large portion of the dorsal skull remaining unossified. There are some caveats that must be considered, namely that the oldest individual of Eudromia elegans examined was only 15 days old in a 20-21 day incubation period, and so further dermal skull ossification likely occurs. The embryos of Nothoprocta perdicaria are of uncertain age, although their heavily ossified postcranial skeletons suggest that they are close to hatching. In all palaeognaths examined, the two rami of Meckel's cartilage meet over a broad medial contact (Fig. 2.1 A,I). This was not observed in any neognaths, including anseriforms whose bill shape most closely approximates that of most ratites. It was surprising to see this morphology in a tinamou {Eudromia elegans) as well as in ratites, since bill morphology is dissimilar between the two groups. The ossification of the dentary from multiple centers, as found in Dromaius novaehollandiae, has been reported as the norm for birds (Jollie, 1957). It is possible this pattern is observed simply due to the length of the element relative to its thickness: the islands of ossification are ragged and spread out, rather than appearing as discrete, well-formed centers.

66 The ceratobranchials ossify later in ratites than in the Galliformes and Anseriformes examined (stage 38 compared to stage 36) (Chapter 4; Maxwell, 2008), and later in sequence than in Charadriiformes (simultaneously with the frontal and after the parietal; (Maxwell and Harrison, In Press)). The hyoid apparatus is relatively smaller in ratites, due to obligate inertial feeding in the group, rather than lingual feeding as in neognaths (Tomlinson, 2000). The functional importance of the tongue during feeding and drinking in palaeognaths is also reduced compared to neognaths (Gussekloo and Bout, 2005). This evolutionary reduction in size may be correlated with the delay in ossification of the ceratobranchials, as previously discussed for the vomer. The functional reduction and reduction in associated musculature may also play a role in the delay in ossification.

Paedomorphosis and ratite cranial evolution. Ratites have been interpreted as being paedomorphic (Livezey, 1995). This idea has gained support from endocrinological studies indicating that typical features of ratites, such as proportionately longer legs, a short wide bill, protuberant eyes, delayed fusion of skull sutures and adult feathers with smooth, symmetrical barbules could have arisen from the neognath condition via paedomorphosis associated with hypothyroidism (Dawson et al., 1994; Dawson et al., 1996). Preliminary examination of the embryological development of ratites suggests that paedomorphosis is not a likely mechanism for the generation of ratite-specific morphological features. Ratite feather development is initiated in a manner similar to neognaths, and at a similar stage (pers. obs.). The elements roofing the cranial vault are in closer contact in the ratites than in other precocial embryos approaching hatching, rather than being delayed. It has been clarified that skull sutures do fuse in adult ratites, similar to what is observed in neognaths (Elzanowski, 1986; Marshall, 2000). Ossification of most of the skeleton occurs at approximately the same stage in both neognaths and ratites. This includes the ossification of the palate, some elements of which are actually accelerated in ratites. At no point in development does the pterygoid-palatine complex in neognaths resemble that of adult palaeognaths, as would be expected given

67 paedomorphosis (Gussekloo and Bout, 2002; Zusi and Livezey, 2006). The development of the ratite palate is distinct from the neognathous palate early in embryonic development based on several features, including accelerated ossification of the vomer. The early appearance and relatively large size of this element excludes the palatine from contact with the parasphenoid rostrum. In ratites, the palatines begin ossification lateral to the pterygoids, and relatively close to the braincase (Figs 2.2, 2.5). This results in the formation of the pterygovomeral arch (Zusi and Livezey, 2006) due to a difference in orientation of the pterygoid relative to the palatine. This is in contrast to the condition in a taxonomically diverse sample of neognaths, where the palatine ossifies further rostrally and approximately parallel to the pterygoid (Schinz and Zangerl, 1937; Maillard, 1948; Tokita, 2003), leading to the early formation of the pterygopalatine arch (Zusi and Livezey, 2006) and preventing the formation of the pterygovomeral arch. This pattern may be caused either by differential growth that separates the earliest ossification centers in neognaths, or by a shift in the location of the ossification centers. A more complete series of tinamou embryos will clarify whether they follow the ratite pattern or the neognath pattern of palatal ossification, but the ratite pattern is predicted based on the possession of a palaeognathous palate. This early difference in palatal formation implies that the palate in ratites is not a developmentally truncated version of the palate in neognaths, and the neognathous palate is morphologically distinct from the palate of ratites prior to the onset of ossification. The absence of a pterygovomeral arch has been hypothesized to be an apomorphy of neognaths (Zusi and Livezey, 2006). This leaves the absence of a postnarial bar as one of the only ratite features that can be derived from the condition in neognaths and tinamous via a truncation of development (Elzanowski, 1986). These observations support the conclusion that ratites are highly apomorphic, both in their adult morphology and in their skeletal development. With one exception the ratite condition cannot be derived by truncating neognath development. Likewise, the neognathous condition does not involve a recapitulation of the ratite condition during early development.

68 Postcranial axial skeleton The vertebral column in birds and other amniotes chondrifies in an anterior to posterior direction, and many authors assume ossification follows a similar pattern (Starck, 1996). While examples demontrating that exeptions occur do exist (Strong, 1925; Rieppel, 1993b; Sheil, 2003a; Blom and Lilja, 2004; Maxwell, 2008), more posterior regions have not previously been reported as ossifying prior to more anterior regions in other avian taxa (Starck, 1993). This is clearly not the case in palaeognaths, where the thoracic centra begin ossification prior to the cervical centra in the Ostrich, Rhea americana and Eudromia elegans, and demonstrates that there can be dramatic dissociation between the sequence of chondrification and ossification, even at a general, regionalized level. The Ostrich and the Emu exhibit very different patterns in the ossification of the atlantal arch. In the Emu, ossification of this arch is accelerated relative to the other cervical arches. The atlantal centrum, however, is delayed in ossification. The same pattern is observed in kiwis, (Parker, 1891). In the Ostrich, the atlantal arch is one of the last cervical arches to ossify. The atlas itself is also delayed in ossification relative to more posterior cervical centra. In earlier stages of Rhea americana, the atlantal arch is not more strongly ossified than the other arches, and in Eudromia elegans it ossifies slightly earlier than C2-C3. Changes in timing of the ossification of the atlantal arch may involve development of the cervical musculature or early movements of the head. This has been suggested to play a role in the early ossification of the exoccipitals in marsupials (Smith, 1996). The accelerated ossification of the atlantal arch in the Emu coincides with an acceleration of the exoccipitals relative to the supraoccipital, basioccipital and the atlantal centrum (relative to the Ostrich, and to a lesser degree R. americana). This may be caused by predominantly lateral head movements in ovo in embryonic Emus, and dorsal and ventral movement in other species. More detail studies of embryonic behaviour and muscle development are needed to test this hypothesis. The pygostyle in palaeognaths ossifies prior to hatching in the Emu, Ostrich and Nothoprocta perdicaria. The most ossified Rhea americana examined lacked any ossified caudal centra as well as an ossified pygostyle, and kiwis also

69 fail to ossify the most posterior portion of the axial skeleton prior to hatching, (Parker, 1891). In other precocial birds, the pygostyle ossifies prior to hatching in Anseriformes, and in the Domestic Turkey {Meleagris gallopavo), but does not consistently ossify prior to hatching in the Domestic Chicken or Japanese Quail {Coturnix coturnix) (Maxwell, 2008, In Press). Although the uncinate processes are reduced in the Emu and are absent in all developmental stages of the Elegant Crested-Tinamou {Eudromia elegans), they are present and well-developed in the Ostrich, Rhea americana, and Chilean Tinamou (Nothoprocta perdicaria). In all species where they are present, the uncinate processes develop at around stage 34, at approximately the same time as the dorsal ribs begin ossifying and prior to the fusion of the sternal cartilages. The uncinate processes serve as attachment for muscles involved in breathing in birds (Codd et al., 2005), and have been reported as absent only in the adult Emus, screamers (Anhimidae) and megapodes (Baumel and Witmer, 1993; Codd et al., 2005). Apparently, the loss of uncinate processes in the Emu takes place after their initial chondrification in the embryo. However, in E. elegans, the uncinate processes never chondrify. Uncinate processes have not been reported as variably present in tinamous, and this demonstrates that skeletal reduction by failure to chondrify and failure to ossify both operate in the axial skeleton (Rieppel, 1992). The posterior projection of the proximal dorsal ribs in the Emu may take over the role of muscle attachment for Mm. appendicocostalis and obliquus externus abdominis from the uncinate processes, and this needs to be confirmed through dissection and myological studies. There is a great deal of variation in the timing of ossification of the thorax in palaeognaths. In other orders of birds, this is conserved at the ordinal level: for instance, in Galliformes, the sternal ribs, uncinate processes, laterocranial and laterocaudal sternal processes all ossify relatively early in embryonic development, whereas in ducks these elements typically do not ossify prior to hatching (Maxwell, 2008, In Press). In the Emu, the sternal ribs and sternum are heavily ossified prior to hatching, and a sternum ossified prior to hatching was also been observed in the Elephant Bird {Aepyornis sp.) (Balanoff and Rowe, 2007). In Rhea americana, ossification has progressed even further, with

70 ossification beginning in the uncinate processes as well as strongly ossified sternal ribs and laterocranial processes of the sternum. In the Ostrich and in kiwis, however, the pattern observed is more similar to the pattern seen in Anseriformes, where the sternal ribs, uncinate processes and sternum remain entirely cartilaginous in some birds at hatching (Parker,, 1891). This pattern appears to be independently derived in Apteryx and Struthio (Fig. 2.5; Bledsoe, 1988). In both tinamou species (Eudromia elegans and Nothoprocta perdicaria), the sternal ribs ossify well prior to hatching. The laterocranial and laterocaudal sternal processes, as well as the body of the sternum and uncinate processes, remain entirely cartilaginous at hatching. In this sense, tinamous demonstrate a different pattern than that exhibited by the Emu, R. americana and Galliformes, but are more ossified in the thoracic region than the Ostrich, kiwis and Anseriformes. The cause of these very different patterns is unknown, and given similarities in both cursoriality and precociality between ratite and galliform hatchlings, it is unlikely such similarities arose as post-hatching adaptations.

Forelimb The scapula and coracoid originate as a single chondrification, although it appears that hypertrophy of cartilage cells begins from the scapular and coracoidal regions rather than from the midpoint of the element (Lutz, 1942). The two elements later separate in some individuals of the Emu and Ostrich, only to fuse again as cartilages. They invariably ossify from separate centers, as in all avian species examined to date as well as in Alligator mississippiensis (Rieppel, 1993a). The scapulocoracoid is formed after hatching by the fusion of the two ossification centers, and the two elements are never separate in adults (Beddard, 1898). The chondrification of the procoracoid process in the Ostrich is delayed relative to the main body of the coracoid. Its ossification is also delayed until after hatching. In Rhea americana, the procoracoid is much smaller than in the Ostrich, but also chondrifies later than the main body of the coracoid. In several individuals, it appears to be more closely derived from the scapula (Fig. 2.3J). Both a coracoidal and a scapular origin of the avian procoracoid process have been argued (reviewed by Vickaryous and Hall, 2006). In kiwis, Parker (1891)

71 noted that the procoracoid is originally cartilaginous and large, as in the Ostrich, but during embryonic development the cartilage degenerates and the structure is almost entirely ligamentous at hatching. The Emu has a furcula in a topographically similar location to the procoracoid of the other species (Fig. 2.3B,C), and lacks a well-developed procoracoid process at all points in development. The morphology of the manus of the Emu is highly variable, both in the adult (Maxwell and Larsson, 2007) as well as during embryonic development. These differences include timing of events, such as the number of digits that ossify prior to hatching, and the morphology of the digits. The location and morphology of the furcula is also variable both developmentally and in the adult (Maxwell and Larsson, 2007). The carpals fuse very early in development in the Emu. There are some indications that the radiale ossifies after hatching from a separate ossification center following fusion of the cartilages early in development (Maxwell and Larsson, 2007), but there is no trace of an ulnare either late in development or in the adult. Lutz reported five digital condensations in the manus of the Emu, and also identified six carpals (Lutz, 1942), but based on the width of the handplate, this appears unlikely. Only three digits were reported as chondrifying (Lutz, 1942). The number of phalanges that chondrify on digits II and IV is variable, ranging between none and two, but by the end of development the count has stabilized at 0-3-1. The major metacarpal has three phalanges throughout development, although not all ossify prior to hatching. The manus of kiwis is also incredibly variable in adults, and this variation is also reflected in the morphologies seen in the developing embryos. Whether different morphologies result in changes in the timing of ossification was not reported (Parker, 1891). Ratites demonstrate several different methods for reducing the manual phalangeal count. In the Emu the terminal phalanges on digits II and IV appear to be resorbed after initially chondrifying. In Rhea americana, the phalangeal count is 2-2-2 but becomes reduced to 2-2-1 by fusion of the proximal and distal cartilaginous phalanges of digit IV. In the Ostrich, the count remains stable at 2-3- 2 prior to hatching. In Eudromia elegans, only 1-2-1 ever form. This demonstrates skeletal reduction by failure of chondrification (fewer phalanges chondrify in

72 some taxa); and also by divergence in the number of chondrifications and ossifications (Rieppel, 1992) - not all elements that chondrify ossify, either by failure of ossification but persistence of these elements as cartilages, or late-acting apoptotic mechanisms leading to degeneration of elements (Kundrat et al., 2002). It is clear that both of these mechanisms are active in the avian manus. There is further dissociation between the sequence of chondrification and ossification in the avian forelimb observed in the Emu. While it is apparently not uncommon for some birds, for instance the Common Tern {Sterna hirundo) and the Domestic Pigeon {Columba livid), to ossify the zeugopodium before the ossification of the stylopodium (Schinz and Zangerl, 1937; Maxwell and Harrison, In Press), in no bird is the metapodium been reported as ossifying prior to the zeugopodium. The Emu appears to be unique in this regard, as kiwis, the Ostrich and Rhea americana follow a strictly proximal to distal pattern of wing ossification (Parker, 1891). According to a study of allometry in ratites, the ratio, of body length to wing length remains constant throughout the incubation period in the Emu, Ostrich, Domestic Chicken and Eudromia elegans (Marshall, 2000). That is, the wing of the Emu is relatively small in early ontogeny, and remains small throughout development. The wings of Gallus gallus and E. elegans were largest relative to body size. If small size delays the onset of ossification, it is therefore not surprising that ossification in the Emu wing is delayed compared to the state observed in flying birds. The humerus begins ossifying during stage 34 (Fig. 2.3A), which is similar in timing to what is observed in Galliformes, and only slightly delayed relative to the state in the Common Eider (Maxwell, 2008, In Press). The radius and ulna, on the other hand, are delayed and do not ossify prior to stage 36 whereas in Galliformes, they ossify around stage 34 and in the Common Eider, these elements ossify during stage 33. In Rhea americana, a ratite with proportionately larger wings than the Emu but still smaller in proportion to body length than thoes of neognaths, these elements are ossified by stage 35. The major metacarpal in the Emu is also delayed, ossifying variably between stages 35 and 36. In the Muscovy Duck, this element ossifies by stage 34 (Maxwell, In Press). The first manual phalanx to ossify in the Emu does not do so prior to stage

73 40+. This represents a very large delay - manual phalanges begin ossification in Galliformes during stage 36, and during stage 37 in Anseriformes (Maxwell, 2008, In Press). Ossification of the manual phalanges in R. americana has usually occurred by stage 38. It is poorly understood whether the delayed ossification seen in the wings of ratites is due to reduced size or delayed maturity in the cartilaginous precursors, reduced motion when in the egg due to poorly developed musculature, or reduced need for function in the region during post-hatching life. The Ostrich ossifies metacarpal II very early compared to both flying birds and other ratites. The only other order of birds in which metacarpal II was found to ossify prior to hatching was Anseriformes (Maxwell, In Press). In neither Rhea americana, nor in either of the tinamou species studied does this element ossify prior to hatching. In ducks, metacarpal II appears to begin ossification as an endochondral center located between metacarpal II and III (Chapter 4). In the Ostrich, this element ossifies in a manner more typical of other long bones. Ossification does begin on the posterior shaft, but rapidly spreads around the element as a collar of bone. Both the Ostrich and R. americana ossify the proximal phalanx of metacarpal IV prior to hatching, and the Ostrich also ossifies the distal phalanx of metacarpal IV, an element which is absent in R. americana. In tinamous, both metacarpal II and manual phalanx IV: 1 are cartilaginous. In the three duck species examined, only the Common Eider ossified phalanx IV :1 prior to hatching (Maxwell, In Press). This phalanx is also not uncommon as an ossified element in late-stage galliforms (Maxwell, 2008). It is interesting that although ratites have relatively reduced wings in which the ossification of some bones is delayed, the ossification other elements in the manus is accelerated. This type of mosaic sequence change suggests functional and developmental modules do not constrain ossification sequence. Although all embryos examined here were stained with alcian blue for the presence of cartilage, the two youngest Ostrich embryos were very poorly prepared, which interfered with detailed observations. The youngest well- prepared embryo was 15 days old. This is one possible explanation for the failure to detect a chondrified digit I in this study. If this structure conforms to the pattern exhibited by other reduced structures in the Ostrich manus and pes, it is

74 undoubtedly transitory, as it was observed only in day 14 embryos of both studies which reported its presence (Feduccia and Nowicki, 2002; Kundrat et al., 2002). An earlier study examined embryos that were several days younger, and failed to locate a cartilaginous digit I (Broom, 1906) and digit I was not seen in any of the other palaeognaths examined.

Hind limb There is great morphological diversity in the ratite pelvic girdle. In the embryonic Emu, for instance, all three elements contact each other as in neognaths; however in the adult none of the elements are in contact (Beddard, 1898; Cracraft, 1974). This character may show intraspecific variability, as the ilium and ischium fuse in some individuals (Glutz von Blotzheim, 1958). The ischium is the first element to ossify (Fig. 2.4B). The ischium is also the first element to ossify in the pelvic girdle of Eudromia elegans, but in this species the ilium does not contact the ischium and pubis. In Rhea americana, the ischium and pubis are in contact, but at no point in development does the ilium contact either element {contra Glutz von Blotzheim, 1958). The ilium contacts the ischium in the adult (Beddard, 1898). A third state is exemplified by the Ostrich, where the ischium and pubis come into contact and then separate later in embryonic development. They remain separate in the adult (Beddard, 1898). Glutz von Blotzheim (1958) found that they remained connected throughout embryonic development, as well as being fused in the adult. The ischium ossifies first, variably relative to the pubis. The pectineal process ossifies from its own ossification center, although it is technically part of the pubis (Beddard, 1898). A separate ossification center was not reported in kiwis (Parker, 1891).

In contrast to the delayed wing ossification observed in the Emu, the hind limb ossifies at approximately the same stage as in neognaths (stage 34) (Maxwell, 2008, In Press). In the Ostrich, there is dissociation between the sequence of chondrification and that of ossification, with the tibia ossifying prior to the femur. This is not uncommon in birds: the tibia is the first element of the hind limb to ossify in the Common Tern {Sterna hirundo), as well as in the

75 Domestic Chicken {Gallus gallus) (Schinz and Zangerl, 1937; Maxwell and Harrison, In Press). The tarsal region of ratites is variably developed prior to hatching. The putative ascending process of the astragalus is typically prominent as cartilage. There is no evidence that the astragalus ossifies from a separate ossification center, as is the case in neognaths, where a separate tarsal ossification center is present (McGowan, 1984, 1985; Maxwell, In Press). The pretibial bone in neognaths loses its association with the astragalus (tibiale) while still cartilaginous, and becomes associated with the calcaneum (fibulare) (McGowan, 1985). Neognaths appear to have acquired a de novo ossification center rather than a new element. The Emu, Rhea americana and Chilean Tinamou {Nothoprocta perdicarid) ossify a distal tarsal prior to hatching, but the Ostrich fails to do so. In N. perdicaria, the tarsal associated with the ascending process ossifies first, followed by a second proximal tarsal associated with the medial condyle of the tibiotarsus. Neither tarsal element is associated with the lateral condyle of the tibiotarsus. This pattern of tarsal ossification was also reported for the Red-winged Tinamou {Rhynchotus rufescens) (McGowan, 1985). Although early-stage embryos of the Chilean Tinamou were not available, only a single tibiale cartilage was present in Eudromia elegans. In contrast, two lateral cartilages were present that later fused to form the fibulare. Histological sectioning of the tarsal region of both species will be required to effectively resolve the issue. In the Emu embryo, Lutz (1942) reported two proximal tarsal cartilages on the medial side of the tarsus, but only a single tarsal on the lateral side. All of the palaeognaths examined have undergone some degree of digital reduction. The Emu, Rhea americana and Eudromia elegans have lost the hallux in the adult, and the Ostrich has partially lost digit II as well (the metatarsal is retained). In some cases, however, the missing elements form in cartilage during development (Fig. 2.4E). Their presence appears to be variable and transitory. For instance, metatarsal I is variably present in the Emu through stages 33-35. Lutz (1942) described this element as being associated with the tarsal region rather than as being located more distally on the tarsometatarsus; this is shown not be

76 the case here. Metatarsal I is even more transitory in the Ostrich, having only been observed around day 16 of incubation; the day 11 and 14 samples in which it was observed appear to be of a similar stage (Broom, 1906; Feduccia and Nowicki, 2002). The proximal phalanx of digit II is also transitory in nature, having been observed in the Ostrich from days 15 and 21 as well as in the above mentioned samples (days 11 and 14) (Broom, 1906; Feduccia and Nowicki, 2002). Reduction appears to be mediated by apoptosis, as in the Ostrich manus (Kundrat et al., 2002): the cartilaginous elements fail to ossify, but also do not persist as cartilages. This mechanism is different from that observed in Hemiergis, a genus of squamates with variably reduced limbs. In Hemiergis, all elements that chondrify eventually ossify, and so failure of elements to ossify is not a mechanism regulating limb reduction (Shapiro, 2002). As in the Emu manus, intraspecific variability in size and shape mediates the timing and sequence of ossification in the Ostrich pes. In the Emu manus, if metacarpal IV has an elongate morphology it ossifies prior to hatching. If it is short and does not articulate with the rest of the carpometacarpus, it will not ossify. The same applies to pedal phalanx IV:5 in the Ostrich pes. If it has a ' typical ungual morphology, it ossifies around day 34. If it has a small, nail-like morphology, it fails to ossify prior to hatching. In none of the ratites examined did ossification of the pedal phalanges proceed in a proximal to distal sequence. In general, the shorter, more cuboid phalanges ossified last, especially in the Emu and Rhea americana where the pedal phalanges of digit IV remained cartilaginous almost until hatching. The phalanges in the Ostrich show a pattern of ossification previously unreported in avian long bones. They have separate dorsal and ventral perichondral ossification centers in the later-ossifying phalanges, which persist as separate centers for a long time prior to fusion. The cause of this pattern is unknown, since the pedal phalanges of the Emu are relatively shorter and more cuboid, but ossify normally.

Incubation period, adult size and degree of ossification The Ostrich is the largest adult bird and has the largest egg of the species examined (Table 2.3). The Emu lags behind the Ostrich in egg size, but has the

77 longest incubation period of the three ratites. The Emu is most ossified at hatching, excluding the wing and the uncinate processes, which are reduced or absent: it has ossified the synsacral arches, caudal centra and pygostyle, as well as the sternum, sternal ribs, crista cnemialis and distal tarsals. This is consistent with the theory that a longer incubation period results in a more ossified hatchling (Starck and Sutter, 2000), although the difference in degree of ossification between the Emu and other ratites is minimal. Rhea americana has delayed ossification of the postcranial axial skeleton, lacking ossification of the caudal centra, synsacral arches, pygostyle, and crista cnemialis but has ossified all of the other elements present in the Emu. Additionally, it has ossified all manual phalanges, the uncinate processes and laterocranial processes of the sternum. The Ostrich lags behind both of the other species in the ossification of the thoracic region: the sternal ribs are largely cartilaginous and the sternum and uncinate processes are entirely cartilaginous. The distal tarsals and crista cnemialis have not yet ossified. The rest of the axial skeleton is as described for the Emu. The Ostrich has, however, a more ossified wing than either of the other species, having ossified all manual phalanges and metacarpal II. Both the Ostrich and R. americana, while delaying ossification regionally, share a similar degree of ossification upon hatching.

Mechanisms of skeletal reduction Reduction mediated by initial development followed by degeneration has been argued to be a common mechanism of limb reduction (Lande, 1978; Galis et al., 2001, 2002). This theory was advanced in reference to the development of mesenchymal condensations, followed by failure of these condensations to chondrify, a pattern which has been observed in the avian manus (Larsson and Wagner, 2002). Chondrification followed by degeneration is a second way digital reduction is accomplished in birds, both in the manus and the pes; this represents reduction by a failure to ossify (Rieppel, 1992). There is more opportunity for ontogenetic change mediated by increasing or decreasing the number of ossification centers forming within a given cartilage. Fusion of cartilages does not necessarily affect the number of ossification centers.

78 This is demonstrated in several ways in the appendicular and axial skeleton of ratites: the pygostyle ossifies from several distinct vertebral ossification centers, the scapulocoracoid ossifies from two centers, and the tarsals ossify from multiple centers. This additional layer of complexity in development is also exhibited by the mesopodium of Hemiergis, where carpal and tarsal elements fuse as cartilages, but retain separate ossification centers (Shapiro, 2002). Alternatively, and also seen in ratites, an element that begins as two separate cartilages may ossify from only a single ossification center. This is the case in the manus of Rhea americana. This case appears to be much less common than the former situation. Lastly, elements that form from a single cartilage may have multiple ossification centers, as in the pubis of the Ostrich.

Morphology and ossification sequence There is mounting evidence that morphology influences ossification sequence. This evidence comes both from intraspecific variability, as well as variability between lineages. Even within an individual, the most posterior cervical ribs are the most robust and these ossify first. Metacarpal IV of the Emu has either an elongate or a short morphology. The short morphology fails to ossify prior to hatching, whereas the elongate morphology ossifies. Pedal phalanx IV:5 of the Ostrich presents a similar case. In either the short or long metacarpal morphology, the state in the Emu is much reduced relative to its state in both neognaths and ratites, and is considerably delayed in sequence. The same applies with regard to the entire wing of the Emu. Between lineages, the vomer of palaeognaths is greatly enlarged relative to its state in the Galloanseres. It is consequently accelerated in timing (stage) in the lineage in which it is enlarged. The ceratobranchials are reduced in ratites, and these are delayed relative to both sequence and stage. This raises the question of whether delayed ossification is the cause of morphological reduction, or vice versa. The former is unlikely to be the case; for instance although the wing in ratites is delayed in ossification relative to the hind limb, the limb bud itself is smaller than the hind limb throughout ontogeny. The wing bud is also proportionately smaller than in neognaths. Delayed ossification

79 is probably a consequence rather than a cause of morphological reduction in endochondral bones. Examination of the membranous precursors of dermal elements will clarify whether this phenomenon is specific to endochondral bones, or can be broadly applied across the entire skeleton.

CONCLUSIONS

Ossification sequence is shown to be variable in palaeognathous birds, but not unusually so when the morphological diversity of the group is accounted for. The developmental patterns resulting in the reduction of the appendicular skeleton in ratite birds are various, and elements may be lost at all steps (condensation, chondrification and ossification). The location of ossification centers does not always correspond to cartilaginous precursors. Some sequence shifts occur within ratites and between ratites and neognaths, but the general pattern of ossification in palaeognaths is similar to other birds. A comparison with neognaths demonstrates that relative expansion or reduction of a structure in the adult results in a relative acceleration or delay in the ossification of that structure in the embryo, respectively, with regard to both stage and sequence. This applies to dermal bones, such as the vomer, as well as endochondral bones, for instance the ceratobranchials and wing elements. The ossification sequence of the palatal elements is not consistent with a paedomorphic origin of the palaeognathous palate; the ossification of several elements is in fact accelerated in ratites. A detailed analysis of character evolution in this region is needed to address hypotheses of heterochrony involving shape variables rather than relative timing variables. This work provides an important overview of ossification in the most basal order of extant birds, and is critical if explicit hypotheses regarding the developmental evolution of the skeleton are to be formulated for Aves.

80 TABLE 2.1: Summary of elements ossified by stage in the Greater Rhea and Emu.

Stage Rhea americcma Dromaius novaehollandiae (RPM) 31 No ossification 32 No ossification 33" No ossification 34 Day 14 Day 22 Femur, tibia, fibula, metatarsals Squamosal*, palatine*, pterygoid*, jugal*, II-IV quadratojugal Dentary, supra-angular, angular Dorsal ribs* Humerus Femur, tibia, fibula, metatarsals II-IV

35 Day 13-15 Day 21-23 Squamosal, lacrimal, premaxilla, Squamosal*, palatine*, pterygoid*, maxilla, parasphenoid rostrum, premaxilla, jugal*, parasphenoid rostrum*, palatine, pterygoid, vomer, jugal, lacrimal*, vomer, quadratojugal Dorsal ribs* Supra-angular, angular Metacarpal III* Humerus, radius, ulna

36 Day 24-27 Parasphenoid rostrum, basisphenoid, lacrimal*, maxilla, nasal Splenial, prearticular Dorsal ribs* Scapula, furcula, radius, ulna, metacarpal III* Ilium, ischium, pubis, pedal phalanges 11:1,3*; III:1,2*,4*;IV:1,5*

37 Day 17 Day 28-29 Nasal, parasphenoid lamina Parasphenoid alae, frontal, parietal, Dentary, splenial, prearticular quadrate Thoracic centra Pedal phalanx 11:3*, I1I:2*,4*, IV:5* Metacarpals III and IV Ilium, ishium, pubis, pedal phalanges 11:1,3; 111:1,4; IV:5

38 Day 20 Day 30-33 Parietal, frontal, quadrate Exoccipital, parasphenoid lamina, Ceratobranchial laterosphenoid* Cervical centra, dorsal ribs Ceratobranchial Scapula, coracoid, manual Cervical centra, thoracic centra, synsacral phalanx 11:1* centra, cervical arches*, cervical ribs*, Pedal phalanges 111:2, IV: 1 sternal ribs Coracoid* Ascending process of the astragalus

39 Day 34-35

81 Supraoccipital*, mesethmoid Free caudal centra*

40+ Day 22 - hatching Day 34 - hatching Basioccipital, supraoccipital, Basioccipital, supraoccipital*, prootic, laterosphenoid, prootic, opisthotic, epiotic, laterosphenoid*, opisthotic, epiotic, parasphenoid mesethmoid* alae, basisphenoid, mesethmoid Articular Articular Free caudal centra*, pygostyle, cervical Synsacral centra, cervical arches, arches*, cervical ribs*, thoracic transverse thoracic transverse processes, processes, synsacral transverse processes, synsacral arches, synsacral sternum transverse processes, cervical Coracoid*, metacarpal IV, manual phalanx ribs, sternal ribs, sternum, 111:1,3 laterocranial process of the Distal tarsals, pedal phalanx 111:3, IV:2 sternum, uncinate processes Manual phalanx 11:1*, 2; 111:1,2; IV:1 Ascending process of the astragalus, distal tarsals, pedal phalanges 11:2,111:3, IV:2,3,4 . Asterisk denotes elements that ossify variably with respect to stage.

82 TABLE 2.2. Rank order of element ossification

Element Dromaius D. novae- Rhea Struthio Eudromia novae- hollandiae americana camelus elegans hollandiae (YPM) (RM)

Skull Basioccipital 14-21 8 8 14-15 5 Exoccipital 14 7 8 16 7 Supraoccipital 18 7 6 14 7 Parasphenoid rostrum 6 3 3 5 4 Parasphenoid ala 12 7 6 16 7 Parasphenoid lamina 14 4 4 16 4 Basisphenoid 10-12 4-7 6 8-10 4-6 Laterosphenoid 13-18 10 8 16 7 Prootic 25 19 8 17 7 Opisthotic 22 12 8 18 7 Epiotic 21 13 8 18 8 Squamosal 3 2-4 2 4 3 Parietal 12 6 6 7. 4 Frontal 11 6 6 8 4 Lacrimal 6 4 2 6 4 Mesethmoid 18 8 8 14 7 Trabeculae 22-24 23 Nasal 8 3 4 6 3 Premaxilla 4 2 2 5 2 Maxilla 7 3 2 4 3 Palatine 3 3 2 5 3 Pterygoid 3 3 2 3-5 3-4 Vomer 5 3 2 6 3 Jugal 3-5 3 2 5 3 Quadratojugal 2 1-3 2 4 2 Quadrate 11 6 5 11 4 Ectethmoid Dentary ' 2 1 4 3 3 Supra-angular 2 1 2 4 3 Angular 1 3 2 4 3 Splenial 7 4 4 5 3 Prearticular 9 5 4 10 4-6 Articular 23 13-18 9 19 Mandibular Entoglossal Basihyal Urohyal Ceratobranchial 13 6 5. 9-11 4 Epibranchial

83 Postcranial axial skeleton Cervical centra 15 7 5 13 7 Thoracic centra 15 7 4 11 4-6 Synsacral centra 16 7 7 11 4-6 Caudal centra 19-24 11-13 16-18 Pygostyle 27-29 21 20 Cervical neural arch 17 9 8 14 7 Thoracic neural arch 20 9-11 9 14 8 Synsacral transverse processes 27 15-16 10 19 Caudal transverse processes 17 20-22 Synsacral arch 19-21 11 21 Cervical ribs 17 7 7 14 8 Dorsal ribs 3-7 4 5 7 4 Sternal ribs 15 7 9 22-23 Uncinate 11 processes Sternum (body) 25 15 10 Laterocranial process 6-11 Laterocaudal process

Forelimb Scapula 5-7 5-7 5 10 . 2 Coracoid 15-18 7 5 14 5 Furcula 5-7 4 3-4 Humerus 1 1 2 2 1 Radius 7 1-3 2 2 1 Ulna 7 1-3 2 2 1 Radiale Ulnare Metacarpal II 19 Phalanx 1 5-9 12 4 Phalanx 2 10 16-23 Metacarpal III 4-7 4 4 5-8 2 Phalanx 1 21-29 12 10 9-14 4 Phalanx 2 11 9-14 4 Phalanx 3 20-25 13 20-23 Metacarpal IV 25-30 11-24 4 5-8 3 Phalanx 1 10 14 Phalanx II 21-23

Hind limb Ilium 9 6 4 12 4

84 Ischium 7-9 4 4 5-7 5 Pubis 8 4 4 5-7 5 Femur 1 1 1 2 1 Tibia 1 1 1 1 1 Fibula 1 1 1 2 1 Patella 23 Ascending process of the 8 6 astragalus 14 7 13 Distal tarsals 28 16 10 Metatarsal I Phalanx 1 Phalanx 2 Metatarsal II 1 1 1 2 2 Phalanx 1 7 6 4 2 Phalanx 2 18 10 2 Phalanx 3 7 6 4 4 Metatarsal III 1 1 1 2 2 Phalanx 1 7 6 4 5 2 Phalanx 2 8-13 6 5 7 2 Phalanx 3 25 13-18 9 14 3 Phalanx 4 7 6 4 11 4 Metatarsal IV 1 1 1 2 2 Phalanx 1 7 6 5 6 •2 Phalanx 2 26-29 15-18 9 16 4 Phalanx 3 20 10 19 4 Phalanx 4 20-22 11 21 4 Phalanx 5 7 7 4 16-23 4 If an element is unnumbered, it was unossified in all specimens examined. If two numbers are given, these represent the range of ranks over which a variable element can ossify.

Table 2.3: Palaeognath size and incubation period (from Marshall 2000 and references therein, unless otherwise noted) Dromaius Struthio Rhea Eudromia novaehollandiae camelus americana elegans Incubation . 50-57 39-42 38-42 (2) 20-21 period (days) Egg(g) 586.17±78.06g 1618 ± 575.2+/-43.4 32-52 (4) 14.05 (5) (2) Adult body Male: 31.5kg 63-104kg 20-40 kg (3) Male: 680g mass Female: 36.9kg (3) Female: 709g (2) (Prinzinger et al., 1997) (3) (Davies, 2002) (4) (Dzialowski and Sotherland, 2004) (5) (Sahan et al., 2003)

85 Bridging text 2. What is known about the processes and mechanisms of skeletal development in neognaths (all extant birds to the exclusion of palaeognaths, which were discussed in the previous chapter) is largely restricted to a single order of birds, Galliformes. This literature bias exists because both the chicken {Gallus gallus) and the Japanese Quail {Coturnix coturnix) are considered vertebrate model systems and are of agricultural importance. Similar to the ratites discussed previously, galliforms have precocial young and vary widely in body size and incubation period, but unlike the ratites they are osteplogically conservative. This permits a more in-depth discussion of the influence of life- history variables and egg size on ossification sequences, while minimizing the confounding influence of divergent morphology. Ossification sequences can differ even between birds in the same order having similar life histories, and so it is relevant to discuss these differences as they occur in Galliformes in spite of the perceived morphological uniformity of the group. Galliformes is a good group in which to observe the effects of source of osteogenic cells and differences in bone type on. ossification sequence, as their development has been well studied experimentally. In this chapter, I describe the formation of the skeleton in the Domestic Turkey (Meleagris gallopavo), and compare it to the Domestic Chicken {Gallus gallus) and the Japanese Quail {Coturnix coturnix). Ossification sequences in this order are not affected by egg size or incubation period. They also appear to be independent of the embryological source of the osteogenic cells, both in terms of the spatial location as well as in which tissue these cells originated.

86 CHAPTER 3

Comparative embryonic development of the skeleton of the Domestic Turkey {Meleagris gallopavo) and other galliform birds

(Reference: Maxwell, E.E. 2008. Comparative embryonic development of the skeleton of the domestic turkey (Meleagris gallopavo) and other galliform birds. Zoology 111:242- 257).

Galliforms, especially phasianids (the family to which the chicken, turkey and quail belong), are generally considered to be morphologically uniform birds (Verheyen, 1956), although this historical idea has received criticism (Dyke et al., 2003). Relationships between the phasianid genera to be examined in this study are not agreed upon (Crowe et al., 2006; Kaiser et al., 2007). Phasianidae is an appropriate family to sample in order to control for the possible effects of adult morphology on intraordinal differences in ossification sequence: although morphological characters do exist to separate the genera, general shape and the number of skeletal elements remain invariant between taxa. All three species are precocial (Starck and Ricklefs, 1998), and differ mainly in incubation period and body size (Table 3.1). For instance, the incubation period of the turkey is almost double that of the quail, and its egg mass is eight times greater (Tazawa et al., 2001). Previous studies providing detailed descriptions of skeletogenesis and ossification sequence for the chick are legion (Parker, 1869, 1888; Sonies, 1907; Schinz and Zangerl, 1937; Erdmann, 1940; Jollie, 1957; Rogulska, 1962; Schumacher and Wolff, 1966a, b; Hogg, 1980). The ossification sequence of the Japanese Quail was described more recently, and remains poorly resolved (Starck, 1989; Nakane and Tsudzuki, 1999). Although contradictions between authors are inevitable due to differences in the manner of reporting ossification sequences, variation in the completeness of the embryonic series examined, and intraspecific variability, there is general consensus regarding the order of appearance of most elements. This allows the existing literature to supplement personal observation in the determination of ossification sequences. The relative sequence of ossification

87 is the same whether enzymatic clearing and staining or histological sections are used, and so studies using different methodologies are also comparable (Clark and Smith, 1993). The ossification sequence of the Domestic Turkey {Meleagris gallopavo) has not previously been described, and may reveal the relative importance of egg size and incubation period to ossification sequence when compared to other Galliformes.

MATERIALS AND METHODS

Eggs of the Domestic Chicken {Gallus gallus), the Domestic Turkey {Meleagris gallopavo) and the Japanese Quail {Coturnix coturnix) were obtained from a local hatchery and incubated at species-specific temperatures in the laboratory (37.5°C and 40% humidity). Embryos were sampled at daily intervals and fixed in 10 % phosphate buffered formalin. Embryos were staged using the normal table presented by Hamburger and Hamilton (1951) for the Domestic Chicken, supplemented by the results of Mun and Kosin (1960) for the turkey. This scheme defines the last four stages (40^4) based on the length of the beak and the third toe. These measurements are not applicable to species other than G. gallus, and so these stages are referred to here as "40+". Embryos were skinned and eviscerated, soaked in a mixture of Citrisolve (a histological clearing agent) and ethanol to remove subcutaneous and intermuscular fat, then cleared and stained for the presence of bone and cartilage following a standard protocol (Dingefkus and Uhler, 1977). The uptake of alizarin red corresponds to the presence of calcium and not always to the initiation of bone formation (Rieppel, 1993a). In this study, the first occurrence of ossification was determined based on the presence of osteoclast spicules that appear as white to cream-colored fibrous columns of bone matrix that do not yet have enough calcium to be stained by alizarin red. All elements were dealt with in the same way to ensure results are comparable to other taxa and to the literature. Ossification sequences presented for Gallus gallus, Coturnix coturnix and Meleagris gallopavo are based on personal observation of specimens housed in

88 the Redpath Museum (RM; Appendix 1). Ossification of M. gallopavo is described in the text by anatomical region, then by stage (Table 3.2); Table 3.3 reflects the sequence of ossification of the entire embryo with a comparison to G. gallus and C. coturnix. Although the sample sizes used in this study are too small to make a detailed analysis of the variability of ossification sequence data in birds, previous studies have shown intraspecific variability to be within the realm of other morphological character sets (Mabee et al., 2000; Maisano, 2002b). Ossification sequences can be compared between taxa by breaking the ranked order of ossification events down to event-pairs (Smith, 1997; Velhagen Jr., 1997). This allows elements that have been accelerated or delayed in the sequence to be identified. Table 3.4 lists sets of events having a shared sequence in two of the three taxa examined. For instance, if element A ossifies following element B in the quail and the turkey, but A precedes B in the chicken, the delayed ossification of A relative to B will be listed as a shared feature of the quail and the turkey. Events having a constant position in all taxa are not listed; nor are events whose placement in any of the taxa is uncertain due to lack of sequence resolution or variability. Intraspecific variability in the turkey is calculated by dividing the number of variable event-pairs (v) by the total number of event-pairs (2v/n(n-l), where n is the number of elements ossified in the most ossified individual examined). Anatomical nomenclature follows Baumel and Witmer (1993). The digits of the manus are numbered II, III and IV.

RESULTS

Sequence variability The oldest turkey examined had 79 ossified elements, which can be described as 3081 event-pairs. Of these, 288 were variable (in some individuals, B ossified before A, while A ossified before B in others). Therefore 288/3081, or 9.3% of all events, showed intraspecific variability. This is within the range of other variability estimates for ossification sequence data (Maisano, 2002b).

Skull

89 Stage 33 (Fig. 3.1 A). The eyeball is very large in proportion to the head, even when compared to other Galliformes, with a diameter equaling approximately 60% of the length of the head. The prenasal process is short and is strongly flexed ventrally. The chondrocranium appears to be close to its maximum development. The quadrate and Meckel's cartilages are in close articulation and the postorbital cartilage and the canalicular portion of the auditory capsule are solid and well-developed. The infrapolar process is relatively prominent, and lies much further anterior to the postorbital process than the state described for ducks (de Beer and Barrington, 1934). The epibranchial cartilage extends posterior to the mandible. The cartilaginous retroarticular process is not fully developed, and does not extend far past the quadrate cartilage.

Stage 34. The prenasal process remains strongly directed ventrally, although it is considerably more elongate than in the previous stage. The interorbital septum has become proportionately shorter relative to the skull, and the posterior portion of the orbital cartilage is beginning to extend dorsad. The foramen for the ophthalmic artery is especially prominent, and is completely surrounded by cartilage. The quadratojugal has ossified, beginning from its posterior end. By the end of this stage, the posterior portion of the orbital cartilage has developed into a solid element posterior to the orbit. The ventral flexure of the prenasal process has decreased considerably, and although it continues to be directed ventrally the cartilage is no longer bowed. The angular is ossified along the ventral edge of the lower jaw, followed by the supra-angular, which ossifies from its posterior end.

Stage 36. Early in this stage, the prenasal process is thickened along its dorsal surface, forming a faint bump near its distal tip. It remains slightly bowed in one individual. The foramen for the ophthalmic artery has also opened up dorsally in this individual. The squamosal is ossified, but the supra-angular is not. In a second individual of the same stage, the prenasal process is completely

90 straight and lacks any anterior thickening. The palatine and pterygoid have ossified, but the supra-angular is absent. Later in this stage (Fig. 3.IB), the prenasal process is straight, but the tip is thickened and bent downwards in a hook. The frontal process of the premaxilla has begun to ossify. The orbital process of the lacrimal is ossified parallel to the nasal, which has begun to ossify from a single center posterodorsal to the external nares. The maxilla is ossified and triradiate. The parasphenoid rostrum and the jugal are ossified. In the mandible, the dentary and splenial bones are visible, and the supra-angular is consistently present. The ceratobranchial portion of the hyoid apparatus is ossified. This is followed by the ossification of the quadrate, then the frontal. Large vacuities in the cartilaginous interorbital septum are present.

Stage 37. The bones of the palate and beak have expanded their ossified areas to form a framework for the skull. The supraorbital process of the lacrimal extends laterally rather than lying appressed to the frontal. The squamosal has ossified around the otic process of the quadrate, forming an embayment between two slender bony processes. The basisphenoid and the parietal are ossified.

Stage 38. The exoccipitals are ossifying, followed by the parasphenoid lamina. The premaxilla now completely covers the prenasal process, giving the beak its definitive shape. Otoliths are present.

Stage 39. Early in this stage, the posterior process of the squamosal exceeds the zygomatic process in distal extent. The structure of the squamosal is also changing, with an area of dense bone extending posterodistally from the zygomatic process, and a pneumatic area beneath (Fig. 3.1C). The supraoccipital ossifies from a single median ossification center located dorsal to the foramen magnum. This is followed by the ossification of the parasphenoid alae. Later, the basioccipital ossifies from elongate paired ossification centers. This is followed by the ossification of the ventrolateral corner of the laterosphenoid.

91 Stage 40+. The prootic has begun to ossify, and the basioccipital ossifications have fused into a single median element. This is followed by the ossification of the opisthotic from the posterior surface of the exoccipital. The laterosphenoid remains unossified in some individuals at this stage (day 18 or 19 of incubation). By day 20, the epiotic has ossified, beginning from the lateral surface of the supraoccipital. The mesethmoid is ossifying anterior to the orbit. At this point, the frontal, parietal and squamosal contact each other. The supraoccipital and exoccipitals are also largely ossified by day 20, resulting in a unified ossification of the posterior skull. Ossification of the columella auris begins shortly after this (day 21). Ossification of the laterosphenoid proceeds from a secondary ossification center located on the dorsomedial corner of the cartilage. The prearticular ossifies as a sleeve around the medial process of the mandible by day 23 (Fig. 3. ID).

92 Figure 3.1. Lateral view of the developing skull of selected Galliformes. Grey shaded regions represent cartilage, black regions represent ossified tissue. The density of stippling reflects the relative degree of ossification. Only a few drawings have been labeled in order to maintain clarity, since skeletal elements are equivalent across stages and across taxa. A: Meleagris gallopavo stage 33. B: M. gallopavo stage 36. C: M. gallopavo stage 39 (early). D: M. gallopavo stage 40+. E: Gallus gallus stage 37 (modified from Erdmann 1940). F: Coturnix coturnix stage 35. G: C. coturnix stage 36. H: C. coturnix stage 40+. I: G. gallus stage 43. Scale bar = 2 mm. a, angular; ac, auditory capsule; cb, ceratobranchial; d, dentary; ee, ectethmoid; exo/oo, fused opisthotic and exoccipital; fr, frontal; ip, infrapolar process; j, jugal; lac, lacrimal; Is, laterosphenoid; M, Meckel's cartilage; me, mesethmoid; mx, maxilla; n, nasal; pa, parietal; pi, palatine; pmx, premaxilla; pnp, prenasal process; pop, postorbital process; psa, parasphenoid ala; psr, parasphenoid rostrum; pt, pterygoid; q, quadrate; qj, quadratojugal; sa, supra- angular; so, supraoccipital; sq, squamosal.

93 * *

-/ 1 • r" • W^-!-P°P

P —

9 r 6t» ' ~' Postcranial axial skeleton Stage 33. The cervical ribs have not yet chondrified, nor have the uncinate processes. Dorsal and sternal ribs are both present as cartilage. There are 16 cartilaginous post-acetabular vertebrae.

Stage 34. The cartilaginous dorsal and sternal ribs have segmented. At the beginning of this stage, no uncinate processes have formed, and no haemal arches are present. There are 20 post-acetabular vertebrae. Late in this stage, the cervical ribs and uncinate processes are both chondrified. The notarium is beginning to form.

Stage 36. The uncinate processes have elongated. Approximately 17 post- acetabular or 12 post-sacral vertebrae are present. Later in this stage, the dorsal ribs ossify.

Stage 37. There are no major differences in the postcranial axial skeleton from stage 36.

Stage 38. The most anterior cervical centra (those located at or anterior to the cervical flexure) are ossifying from a single ventral ossification center. The number of post-acetabular vertebrae has been reduced to 15, with the remaining vertebrae incorporated into the pygostyle. Although these posterior-most vertebral elements have fused into the pygostyle, ridges along their dorsal and ventral edges are still visible.

Stage 39. Some individuals still lack ossified centra at the beginning of this stage. In others, dorsal and ventral ossification centers have appeared in the cervical centra, although dorsal centers are not present in the most posterior vertebrae. The sternal ribs have also begun to ossify in this stage with the ossification beginning in the mid-region of the series and extending anteriorly and posteriorly.

94 Stage 40+. Some individuals early in this stage lack ossified sternal ribs, suggesting that ossification events in the axial skeleton do not track as closely with stage as in othqr regions of the body. Dorsal and ventral ossification centers are present for all cervical vertebrae by day 18. The dorsal and ventral centers have merged in CI and C2. The cervical neural arches are ossifying, as are the cervical ribs (those close to C4). The thoracic neural arches anterior to the notarium are ossifying. Nine free caudal centra are present. By day 19, the dorsal and ventral ossification centers of the cervical centra have fused. The anterior five synsacral centra are ossifying from right and left paired ossification centers. By day 20, all thoracic neural arches have begun ossification. All cervical ribs have ossified. The anterior 13 synsacral centra are ossified, but retain paired centers. By day 21, the synsacral centra are not yet completely ossified, with the most posterior three vertebrae lacking all traces of ossification centers. The ossification of the inner laterocaudal process of the sternum occurs around this time. By day 22, both laterocaudal processes of the sternum have ossified. The most anterior transverse process of the synsacrum has ossified by day 23, as have the free caudal centra and arches. The notochordal portion of the pygostyle is ossified as well. The laterocranial processes of the sternum ossify prior to hatching, at around day 24.

Pectoral appendage Stage 33. One phalanx is chondrified for all digits. The scapular and coracoidal cartilages have separated.

Stage 34 (Fig. 3.2A). Both manual phalanges of the major digit are now chondrified. A second, very small cartilaginous phalanx is variably present on the alular digit (digit II). It does not ossify prior to hatching in any individuals. During this stage, the humerus ossifies, followed by the radius and ulna. In some specimens, the furcula ossifies at the same time as the humerus; in others it is

95 delayed until after the zeugopodial elements ossify. The hypocleideum is wedge- shaped, and the furcula extends approximately parallel to the coracoids. The radiale complex and the ulnare are chondrified. The cartilaginous semilunate distal carpal mass caps the major metacarpal.

Stage 36. Metacarpal II has become closely associated with the carpometacarpus, and its posterior end is continuous with metacarpal III. The scapula, and metacarpals III and IV have ossified. The last element to ossify during this stage is the proximal phalanx of digit III.

Stage 37. The coracoid ossifies, followed by the proximal phalanx of digit II.

Stage 38. The distal phalanx of digit III is ossified.

Stage 39 (Fig. 3.2B). There are no qualitative changes to the wing skeleton during this stage.

Stage 40+. Late in this stage (around day 21), the proximal phalanx of digit IV ossifies.

96 Figure 3.2. Lateral view of the forelimb of selected Galliformes. Grey shaded regions represent cartilage, black regions represent ossified tissue. The density of stippling reflects the relative degree of ossification. Only one drawing has been labeled in order to maintain clarity, since skeletal elements are equivalent across stages and across taxa. A: Meleagris gallopavo stage 34. B: M. gallopavo stage 39 (early). C: Gallus gallus stage 34. D: G. gallus stage 43. E: Coturnix coturnix stage 35 (late). F: C. coturnix stage 40+. Scale bar = 2 mm. car, central carpals; H, humerus; mcll-IV, metacarpals; R, radius; ra, radiale; scap, scapula; U, ulna; ul, ulnare complex.

97 ^eiSC^Sfc,

scap Pelvic limb and girdle Stage 33. All pelvic limb and girdle elements present at this stage are cartilaginous (Fig. 3.3). The ilium, ischium and pubis are all in contact and are directed posteriorly. The fibula is detached from the tarsal region. The hallux has shifted distally towards the other digits, but is not yet fully reversed. Two phalanges are present on each of the pedal digits.

Stage 34 (Fig. 3.3A). The femur and tibia are ossified, followed by the fibula and tarsometatarsus later in this stage. The tarsals are not yet fused to the tibia. All of the phalanges are segmented, and metatarsal V is clearly visible as cartilage. The cartilaginous hallux is fully reversed.

Stage 36 (Fig. 3.3B). The cartilaginous tarsals capping the tibia have fused to this element to form the condyles. Late in this stage, the pubis ossifies.

Stage 37. In this stage, pedal phalanges 11:1,11:3, and 111:1,2,4 are ossified, followed by 11:2 and 111:3. The ventral margin of the anterior wing of the ilium, phalanges 1:2, IV:1 and IV:5 are variably ossified.

Stage 38. The ischium is ossified, as is phalanx 1:1. This is followed by the ossification of phalanx IV:2.

Stage 39. The posterior wing of the ilium is now ossifying by extension of the anterior ossification center over the acetabular region. Metatarsal I is ossified at its proximal tip. In a second individual, phalanges 1:2,11:3 and IV:5 are not ossified, indicating that extensive intraspecific variability exists in the sequence of phalangeal ossification in this species.

Stage 40+ (Fig. 3.3C). The proximal end of the pretibial bone ossifies, followed by phalanx IV:4. Phalanx IV:2 is still not consistently ossified in all individuals. Phalanx IV:3 always ossifies last.

98 Figure 3.3. Lateral view of the hind limb of selected Gahiformes. Grey shaded regions represent cartilage, black regions represent ossified tissue. The density of stippling reflects the relative degree of ossification. Only one drawing has been labeled in order to maintain clarity, since skeletal elements are equivalent across stages and across taxa. A: Meleagris gallopavo stage 34. B: M. gallopavo stage 36. C: M. gallopavo stage 40+. D: Coturnix coturnix stage 35. E: C. coturnix stage 37. F: C. coturnix stage 40+. G: Gallus gallus stage 34. H: G. gallus stage 36,1: G. gallus stage 43. Scale bar = 2 mm. F, femur; Fib, fibula; il, ilium; isch, ischium; ptb, pretibial bone; pub, pubis; T, tibia; tmt, tarsometatarsus.

99

DISCUSSION

Skull The early ossification of the quadratojugal, squamosal and angular is found in all three galliform genera (Fig. 3.1B,E-G). The vomer is present in the chicken prior to hatching, however it was not located in any of the turkeys examined. It has been described as present in adult Meleagris gallopavo (Tomlinson, 2000) so presumably it ossifies after hatching, or is only variably present in adult birds. The mesethmoid ossifies well before hatching in M. gallopavo, similar to what is observed in Coturnix coturnix where it ossifies prior to any ossification of the caudal centra (Nakane and Tsudzuki, 1999). It also ossifies prior to the ossification of the laterocranial processes of the sternum, caudal neural arches and pygostyle of Gallus gallus (Fig. 3.11, Table 3.4). The squamosal of Meleagris gallopavo undergoes osteological restructuring during stage 39. This may correspond to the development of M. adductor mandibulae externus (Zusi and Livezey, 2000), resulting in reorientation of bony trabeculae due to strain caused by muscle development (Biewener et al., 1996). Further research into embryonic behaviour and cranial muscle development in this species is necessary to test this hypothesis. The basioccipital is reported as being an unpaired midline element in the chicken (Jollie, 1957), however my observations suggest that it forms from paired centers. The parallel ossification centers are reminiscent of the ossification of the cervical vertebrae, providing support for long-term conservation of ossification pattern (the occipital region is thought to be derived from vertebrae that become transferred to the skull during development (Romanoff, I960)). In Meleagris gallopavo, the supraoccipital ossifies as an unpaired median element from an ossification center located dorsal to the foramen magnum. This is in contrast to the chicken, which has paired centers (Erdmann, 1940; Jollie, 1957). Species in many other orders have been reported as also forming this element from a medial unpaired ossification center (reviewed by Jollie, 1957). In all galliforms examined, the ossification of the basisphenoid is not accelerated relative to the

100 ossification of the parietal, exoccipital and quadrate. The supraoccipital and basioccipital ossify after these elements (Jollie, 1957). In Meleagris gallopavo, the ossification of the prearticular is greatly delayed, ossifying after most cranial elements, including the opisthotic and epiotic. This is not so in Gallus gallus or Coturnix coturnix, where this element ossifies around the same time or slightly before the occipital series (Jollie, 1957; Starck, 1989; Nakane and Tsudzuki, 1999). The articular was also not ossified in the oldest individual examined, although in other Galliformes, it ossifies at the same time as the laterocranial process of the sternum, the first phalanx of digit IV and the caudal centra (Schumacher and Wolff, 1966a; Nakane and Tsudzuki, 1999), all of which were ossified in the oldest M. gallopavo embryos (Table 3.4). In all Galliformes, the nasal ossifies from a single ossification center located posterior to the external narial opening. This is in contrast to the situation in Charadriiformes (specifically in the common tern, the Great Skua and the Herring Gull) (Maillard, 1948; Maxwell and Harrison, In Press), possibly due to schizorhinal narial morphology in the latter clade. Erdmann (1940) reported two centers of ossification for the frontal of Gallus gallus. These have been interpreted as representing the separation between portions of the frontal that are derived from paraxial mesoderm and those portions that are of neural crest origin (Evans and Noden, 2006). Other authors have failed to observe two centers for this element (Jollie, 1957), and multiple centers were not observed in the embryos examined here.

Postcranial axial skeleton In the chicken, the cervical vertebrae ossify roughly in an anterior to posterior sequence, although the most anterior centra may be slightly delayed (Schinz and Zangerl, 1937). The most posterior cervical centrum ossifies ahead of the preceding cervical centra, and ossification spreads posteriorly into the thoracic series. In the quail, ossification spreads both anteriorly and posteriorly from the middle of the cervical series. Ossification also begins in the middle of the thoracic series and spreads both anteriorly and posteriorly (Nakane and Tsudzuki, 1999). However, a strict anterior to posterior sequence of vertebral ossification was

101 reported for the quail by Blom and Lilja (2004). In the turkey, ossification appears to proceed in a strictly anterior to posterior fashion, but this part of the sequence is not very well resolved. In the turkey the notochordal portion of the pygostyle ossifies well before hatching. This differs from Anseriformes (specifically the Muscovy Duck, Pekin Duck and Common Eider), in which the ossification of the pygostyle begins with the most anterior vertebral components (Maxwell, In Press). According to some authors (Schinz and Zangerl, 1937; Hogg, 1980), the pygostyle ossifies prior to hatching in the chicken, but this was not observed in any of the specimens examined here. The pygostyle and most posterior caudal centra of the quail were reported as unossified (Starck, 1996). In the turkey, the laterocranial and laterocaudal processes of the sternum ossify prior to hatching. The laterocaudal processes ossify first, the medial process followed by the lateral process. This is identical to the pattern observed in the chicken. The ossification of the laterocaudal processes preceding the ossification of the laterocranial process was also observed in the quail (Nakane and Tsudzuki, 1999). These processes appear to be derived from sternal ribs in Galliformes, perhaps accounting for their accelerated ossification relative to other avian orders (Starck, 1993). Ossification of the uncinate processes was not observed in the turkey, although it is well documented in the chicken and the quail (Schinz and Zangerl, 1937; Hogg, 1980; Nakane and Tsudzuki, 1999). Ossification of the sternal ribs prior to hatching, beginning with the middle ones and spreading anterior and posterior, has been reported for all phasianids (Schinz and Zangerl, 1937; Nakane and Tsudzuki, 1999), but was reported absent in more basal Galliformes (Starck, 1996).

Forelimb In the forelimb of the chicken (Fig. 3.2C and D), the humerus, radius and furcula ossify soon after the initial formation of dermal skull elements, followed by the ulna. Schinz and Zangerl (1937) recorded the radius as being delayed relative to the ulna, but ossifying simultaneously with the furcula. This pattern is

102 roughly the same as that recorded for the turkey, although apparent simultaneity and polymorphism prevent a definitive comparison. In the quail (Fig. 3.2E and F), the long bones are recorded as ossifying before the furcula (Nakane and Tsudzuki, 1999), although personal observation suggests otherwise. The delay in furcular ossification is also reported for the chicken (Schumacher and Wolff, 1966a). The rapid ossification of all of these elements is reflected in their apparent simultaneity in many taxa. The flexibility of the ossification sequences relative to each other also suggests that these elements ossify independently of each other, though their formation may be invariant relative to other morphological, functional, or developmental modules. They also create a strong argument for the case of biologically real simultaneity. The simultaneous ossification of elements was previously dismissed as sampling artifact (Nunn and Smith, 1998), but extensive sequence variability may mean that assuming simultaneity is a valid methodological approach, especially in well-sampled series. This leads to the counterintuitive conclusion that simultaneous ossification of elements might not exist within an individual, but can exist within a species because of polymorphisms. In both the chicken and the quail, the proximal phalanx of digit II ossifies prior to the proximal phalanx of digit III. In the turkey, this sequence is reversed. The proximal phalanx of digit IV ossifies prior to hatching in all three taxa (Rogulska, 1962; Schumacher and Wolff, 1966b; Hogg, 1980; Nakane and Tsudzuki, 1999), but it ossifies immediately prior to hatching in the chicken and . the quail, whereas in the turkey it ossifies one week prior to hatching, representing a significant acceleration (Table 3.4). In none of the galliforms studied did metacarpal II ossify prior to hatching, nor was it previously reported (Hogg, 1980) although its presence as an ossified element in late-stage Anseriformes is not unusual: it was found to ossify in both the Pekin Duck and the Common Eider during the embryonic period (Maxwell, In Press).

Hind limb In the pelvic girdle of Meleagris gallopavo, the pubis ossifies first, followed by the ilium and then the ischium. This is highly divergent from the

103 sequence reported for the quail (Fig. 3.E) by some authors (Nakane and Tsudzuki, 1999), but identical to that described by others (Starck, 1989, pers. obs.). Both sequences have also been observed in the chicken (Table 3.3) (Schumacher and Wolff, 1966a, pers. obs.). In the forelimb of the turkey, ossification is slightly advanced over the hind limb. This is similar to the chicken (pers. obs.), although the opposite has also been reported (Schinz and Zangerl, 1937). In the quail, the hind limb appears to precede the forelimb in ossification (Table 3.3). The stylopodium (femur) ossifies before the zeugopodium, and the tibia ossifies before the fibula in the turkey. Again, differences have been reported for the chicken where the tibia ossifies before the femur (Schinz and Zangerl, 1937), followed by the fibula. Therefore, although the proximal-distal sequence of long bone ossification may not be present in the limbs of all Galliformes, the ossification of the fibula after the tibia may be a more stable sequence as it is consistently observed in all three genera examined. Ossification in both the fore- and hind limbs of the turkey begins with the stylopodium, and it is unresolved whether ossification of the ulna follows that of the radius, as would be expected if the pattern observed in the hind limb was identical to the forelimb. The pattern in the wing of the chicken is variable, but Schinz and Zangerl (1937) reported that in the forelimb ossification of the radius is delayed, whereas in the hind limb ossification of the fibula is . delayed. This reversal of ossification of anterior and posterior zeugopodial elements suggests there may be dissociation between the ossification sequences of the fore- and hind limbs. The metatarsals always ossify after the stylo- and zeugopodial elements. Although it is impossible to deduce the sequence of ossification of the tarsometatarsus in the turkey or the quail, the ossification of metatarsal IV is delayed in Gallus gallus (Schumacher and Wolff, 1966a). Tarsals were not found to ossify in the latest-stage turkey examined, and ossify immediately prior to hatching in the chicken (Schinz and Zangerl, 1937; Schumacher and Wolff, 1966b; Hogg, 1980) and at hatching in the quail (Starck, 1996). The phalanges in the turkey are variable in their ossification sequence, and do not follow a strictly proximal to distal pattern.

104 Factors affecting ossification sequences Variability in ossification sequences is well known both within and between populations of amniotes (Garn et al., 1966; Maisano, 2002b; Sheil and Greenbaum, 2005). Large differences in the amount of sequence variation observed have been reported for vertebrates with feeding larval stages as well (reviewed by Mabee and Trendler, 1996). Variation is not constrained by the function of the element, or lack thereof (Mabee et al., 2000), and so amniotes are not expected to demonstrate much greater levels of ossification sequence variability than other vertebrate classes. Ossification sequences should remain generally consistent, at least at the species level, and therefore reflect some aspect of the biology of the organism. Although the turkey has a considerably longer incubation period than either the chicken or the quail, it does not demonstrate a higher number of ossified elements. In fact, several skull elements are delayed in their formation. In megapodes (Galliformes), a high degree of ossification at hatching has been attributed to the extended incubation period (Starck and Sutter, 2000). Having a larger egg size also does not seem to influence the ossification sequence: the turkey's egg is much closer in size to the chicken's than to the quail's (Table 3.1), however, the number of sequence changes is similar (Table 3.4). The delay in the ossification of some skeletal elements is also difficult to explain based on body size. Avian species with higher growth rates tend to have a higher proportion of cartilage in their skeletons (Starck, 1996; Blom and Lilja, 2004). A high growth rate may also lead to a delay in the ossification of some skeletal elements (Arendt and Wilson, 2000). Larger galliforms such as turkeys have been found, however, to have lower post-hatching growth rates than quail (Dietz and Ricklefs, 1997). The ossification sequence in Galliformes appears to be largely independent of several variables historically believed to constrain it. For instance, the ossification sequences of the fore- and hind limbs are independent, not identical as one might expect for serial homologues. This type of dissociation has also been noted in squamates and turtles (Rieppel, 1994; Sheil, 2005).

105 Ossification of the skull does not proceed from anterior to posterior or vice versa, even when the source of the osteogenic cells is considered rather than the absolute position of the element in the adult (Evans and Noden, 2006). For instance, the neural crest-derived portion of the squamosal ossifies prior to the neural crest-derived maxilla in all taxa, even though the precursor cells of the maxilla arise further anteriorly. The splenial ossifies following the more anterior dentary, further demonstrating the independence of osteogenic cell source. Within the elements derived from the paraxial mesoderm, the ossification of the parietal precedes that of the more anteriorly located laterosphenoid, as well as that of the more posteriorly located supraoccipital. Bones derived from the paraxial mesoderm do not always ossify after elements of neural-crest origin. For instance, ossification of the parietal precedes the neural crest-derived mesethmoid. Elements within the paraxial mesoderm and neuroectoderm- derived portions of the avian skull have previously been shown to evolve in a correlated way (Marugan-Lobon and Buscalioni, 2006). It is necessary to evaluate a larger number of taxa in order to test whether these modules truly constrain the potential for changes in ossification sequence between avian taxa with divergent skull morphologies. A recent study on the mammalian skull indicates that such evolutionary modules do not necessarily correspond to developmental modules (Goswami, 2007). The sequence of ossification has also become dissociated from the sequence of chondrification. Within the limb, chondrification proceeds in a proximal to distal direction (Shubin and Alberch, 1986). Differences in the pattern of ossification in amniotes have previously been noted (Sheil, 2005; Frobisch, In Press), mostly caused by alteration of the sequence of ossification within digits. Birds appear to alter the sequence of ossification from that of chondrification not just within digits but also between the stylopodial arid zeugopodial elements. In the axial skeleton, the sequence of chondrification proceeds from anterior to posterior (Romanoff, 1960). The sequence of ossification does not follow this strict progression, nor is there always a single initial site from which ossification proceeds (Rieppel, 1993b; Sheil, 2003a; Blom and Lilja, 2004).

106 The clear independence of some aspects of ossification from chondrification may not pertain to recently evolved morphological features. For instance, the early ossification of the laterocranial and laterocaudal processes of the sternum may be caused by developmental constraint: the deep embayments of the phasianid sternum are unique among avian taxa examined thus far. If the sternal processes are homologous with sternal ribs (Starck, 1993), the ossification of which are already accelerated relative to other birds, this might be the cause of accelerated ossification of these components in the sternum as well.

CONCLUSIONS

In spite of their differences in size and incubation period, the three galliforms examined here demonstrated remarkable consistency in their degree of ossification at hatching, as well as the identity of the elements that were ossified. For instance, accelerated ossification of the sternal ribs, sternal processes and uncinate processes appears to be unique to Galliformes, even when other precocial taxa are considered (Maxwell, In Press). Various developmental parameters, such as embryological origin of the elements and sequence of chondrification, do not seem to be solely responsible for the pattern observed. In order to estimate the impact of phylogeny, function and adult morphology on ossification sequences, increased taxonomic sampling is necessary.

107 Table 3.1. Mean egg mass (± SD) and incubation period of selected Galliformes (Tazawa et al. 2001) Species Incubation period (days) Fresh egg mass (g) Coturnix coturnix 17 10.7 ±0.7 Gallus gallus 21 64.9 ±2.5 Meleagris gallopavo 28 82.9 ± 2.6

108 Table 3.2: Summary of elements ossified by stage for the Domestic Turkey, Meleagris gallopavo. Stage Day of incubation Elements ossified 33 10 34 11 Quadratojugal Angular, supra-angular* Furcula*, humerus, radius, ulna Femur, tibia, fibula, metatarsals II-IV

36 11.5-13.5 Squamosal, palatine, pterygoid, parasphenoid rostrum, premaxilla, maxilla, jugal, quadrate, lacrimal, nasal, frontal Dentary, supra-angular*, splenial Ceratobranchial Dorsal ribs Scapula, furcula*, metacarpals III-IV, manual phalanx 111:1 Pubis

37 15 Basisphenoid, parietal Coracoid, manual phalanx 11:1 Ilium, pedal phalanges 1:2*, 11:1,2,3*; 111:1,2,3,4; IV:1,5*

38 16 Exoccipital Cervical centra* Manual phalanx 111:2 Ischium, pedal phalanges 1:1,2*; 11:3*, IV:2*,5*

39 16.5-18 Supraoccipjtal, parasphenoid alae and laminae, , basioccipital, laterosphenoid* Cervical centra*, sternal ribs* Pretibial bone*, metatarsal I, pedal phalanges 1:2*, 11:3*, IV:2*,4*,5*

40+ 19-24 Laterosphenoid*, prootic, opisthotic, epiotic, mesethmoid Prearticular Cervical neural arches, cervical ribs, thoracic centra, thoracic neural arches, synsacral centra, synsacral transverse processes, caudal centra, caudal neural arches, pygostyle, sternal ribs*, laterocranial and laterocaudal sternal processes Manual phalanx IV: 1 Pretibial bone, pedal phalanges IV:2*3,4* Asterisk denotes elements that ossify variably with respect to stage.

109 Table 3. Rank order of element ossification for three galliform genera.

Element Meleagris Gallus Coturnix Coturnix gallopavo gallus coturnix coturnix (Nakane (pers obs) and Tsudzuki 1999) Skull Basioccipital 17 17 4 Exoccipital 12 13 4 10 Supraoccipital 13-16 17 6 Parasphenoid rostrum 6 8 4 5 Parasphenoid ala 16 10 Parasphenoid lamina 15 10 Basisphenoid 9-11 17 4 10 Laterosphenoid 18-21 22-24 6 Prootic 19 19 6 Opisthotic 20 22-24 6 Epiotic 22 22-24 6 Squamosal 4 2 2 1 Parietal 9 13 4 8 Frontal 8 9 3 6 Lacrimal 7 7-9 3 5 Ectethmoid • Mesethmoid 22 29 7 Trabeculae Nasal 7 5 2 5 Premaxilla 6 9 3 5 Maxilla 6 4 3 5 Palatine 5 4 3 5 Pterygoid 5 4 2 5 Vomer <20 5 Jugal 6 4 2 2 Quadratojugal 1 1 2 1 Quadrate 6-8 10 4 7 Dentary 6 4 2 5 Supra-angular 2-6 4 2 5 Angular 2 2 2 1-4 Splenial 6 9 3 Prearticular 24 11 4 Articular 9 Mandibular Entoglossal Basihyal Urohyal Ceratobranchial 6 8-9 3 3 Epibranchial

Postcranial axial skeleton

110 Cervical centra 14 15-17 4 10 Thoracic centra 19 19 5 Synsacral centra 20 21 6 Caudal centra 24-25 28 9 Pygostyle 24 Cervical arch 19 24 5 Thoracic transverse processes 19 24 6 Synsacral transverse processes 24 28 8 Caudal neural arches 24 31 9 Synscral arch 30 8 Cervical ribs 19-24 20 6 Dorsal ribs 8 9 3 7 Sternal ribs 18 16 6 Uncinate processes 28-31 9 Sternum (body) Laterocranial processes 25 31 9 Laterocaudal processes 23 27 8

Fore limb Scapula 6 9 2 5 Coracoid 9-11 9 3 7 Furcula 1-6 3 2 1 Humerus 1 3 1 - 1-4 Radius 2 3 1 1-3 Ulna 2 4 1 1-3 Radiale Ulnare Metacarpal 11 Phalanx 1 9-11 11 3 7 Phalanx 2 11-31 5 9-11 Metacarpal III 6 7 2 1-3 Phalanx 1 7 14 4 7 Phalanx 2 11 11 4 7 Metacarpal IV 6 8 2 1-3 Phalanx 1 23 10

Hind limb Ilium 9-11 11 3 7 Ischium 12 17 3 8 Pubis 8 9 4 6 Femur 1 4 1 1 Tibia 2 4 1 1 Fibula 3 4 1 1-3 Patella Pretibial bone 17-24 26 Tarsals Metatarsal I 16 18 5 Phalanx 1 11 12 4 5-9 Phalanx 2 9-17 11 5 5-9

111 Metatarsal II 3 6 2 1-4 Phalanx 1 9 12 4 5-7 Phalanx 2 10 15 4 7 Phalanx 3 9-17 11 5 7-10 Metatarsal III 3 6 2 1-4 Phalanx 1 9 11 3 5-7 Phalanx 2 9 15 3 5-7 Phalanx 3 10 15 4 7 Phalanx 4 9 11 5 7-10 Metatarsal IV 3 6 2 1-4 Phalanx 1 9 11 3 5-7 Phalanx 2 11-21 23 3 7-10 Phalanx 3 21 " 25 5 7-10 Phalanx 4 19-21 18 5 7-10 Phalanx 5 9-11 11 5 7-10 If an element is unnumbered, it was unossified in all specimens examined. If two numbers are given, these represent the range of ranks over which a variable element can ossify.

112 Table 3.4. Shared sequence changes in the ossification sequences of selected galliforms. Coturnix + Gallus Coturnix + Meleagris Meleagris + Gallus soDbo bo A str so A (tc, ca) pra A (so, Is, po, oo, eo, pa A (exo, psa, psl, pra) psr A (131,132,141) tc, sc, ca, tt, str, lea, w41,mtl) pa D w32 psaD(w21, w31,w32, po A 122 pub, 111,121,122,131,133, 134,141) w21 D (ft, dr, cor, w21, pslD(w32, pub, 141) 143 D (so, oo, tc, sc, ca, tt, il) str, mtl) n A (pmx, psr) bs D pub 142 D (pa, q, sp, w31,121, 122,133,134) jApl eo D 143 pi A sp w41 D (ar, st, en, ssa, dr D (lac, pmx, cb) pt A (j, mc3, mc4) up, lcr) cac A 112 n D (mc3, mc4, mt4) qA(w21,131,132,141) ssa A ct pt D mt4 sp A (cor, w21,131,132, 141) mtl D 144 pra D(cc,w32,111,121, cc D (134,145) 122,132,133,141,142) sc D ca str A (tc, ca) W32D141 cac A ssa 11 ID 132 134A(w32,111,122,133) w31 A 132 pub A 141 Only sets of events showing variation between the three genera are listed. Shifts for Gallus and Coturnix are based on consensus sequences between personal observation, Schumacher and Wolff (1966) and Nakane and Tsudzuki (1999) respectively. See materials and methods section for more details. A= relative acceleration, D= relative delay ar, articular; bo, basioccipital; bs, basisphenoid; ca, cervical arch; cac, caudal centra; cb, ceratobranchial; cc, cervical centra; en, caudal neural arches; cor, coracoid; dr, dorsal ribs; eo, epiotic; exo, exoccipital; fr, frontal; il, ilium; jjugal; 111-145, pedal phalanges; lac, lacrimal; lea, laterocaudal process; lcr, laterocranial process; Is, laterosphenoid; mc2-4, metacarpals; mtl-4, metatarsals; n, nasal; oo, opisthotic; pa, parietal; pi, palatine; pmx, premaxilla; po, prootic; pra, prearticular; psa, parasphenoid alae; psl, parasphenoid lamina; psr, parasphenoid rostrum; pt, pterygoid; pub, pubis; q, quadrate; sc, synsacral centra; so, supraoccipital; sp, splenial; ssa, synsacral arches; st, synsacral transverse processes; str, sternal ribs; tc, thoracic centra; tt, thoracic transverse processes; up, uncinate processes; w21 -w41, manual phalanges.

113 Bridging text 3. Anseriformes (waterfowl) is the sister-group to Galliformes. While the two orders do share several osteological similarities in palatal structure, they have very different beak and foot morphologies, and ecologies. While galliforms are primarily terrestrial, with conical bills, anseriforms are generally aquatic, with webbed feet and dorsoventrally compressed bills. Many anseriforms are also accomplished migrants, flying long distances to take advantage of seasonal habitats. Diversity in patterns of skeletal development has not been described in anseriforms, although much of the historical work on early skeletal development in birds was based on the Domestic Duck (Sieglbaur, 1911; de Beer and Barrington, 1934). The expression patterns and the phenotypic effects of upregulation and downregulation of many genes are well characterized in the anseriform and galliform craniofacial region. This provides a unique opportunity to examine whether those altered expression patterns have downstream effects on ossification sequences. In this chapter, I describe the ossification sequences of three ducks, the Common Eider (Somateria mollissima), the Pekin Duck {Anas platyrhynchos) and the Muscovy Duck (Cairina moschata). Sequence differences exist both within and among these species, but are generally minor. The Common Eider has the most ossified skeleton prior to hatching, contrary to what is expected in a subarctic migrant species. This can be attributed to a tradeoff between growth rate and locomotory performance. Growth rate is higher in hatchlings with more cartilaginous skeletons, but this may compromise locomotion. No major ossification sequence differences were observed in the craniofacial skeleton when compared to Galliformes, which suggests that the influence of adult morphology on ossification sequence might be relatively minor in many taxa. Galliformes and Anseriformes, while both highly ossified at hatch, differ in the identity of their late-stage ossifications. In Anseriformes, these are most often located in the appendicular skeleton, while in Galliformes they are in the thoracic region and form the ventilatory apparatus.

114 CHAPTER 4

Ossification Sequence of the Avian Order Anseriformes, with Comparison to other Precocial Birds

(Reference: Maxwell, E.E. Ossification sequence of the avian order Anseriformes, with comparison to other precocial birds. Journal of Morphology In Press, doi 10.1002/jmor.l0644).

Ossification sequence is influenced by numerous developmental factors, including muscle development and embryonic movements (Adriaens and Verraes, 1998; Wagemans and Vandewalle, 2001), as well as constraints such as modularity, the sequence of chondrification and the source of osteogenic cells (Maxwell, 2008; Frobisch, In Press). Shape variables also may influence the observed sequence; these include sexual dimorphism (Garn et al., 1966) and adult morphology (Haluska and Alberch, 1983; Adams, 1992; Rieppel, 1993b). Additionally, phylogeny (Maisano, 2002b; Poe, 2006; Hofmann et al., 2007), adaptations to post-embryonic life (Adriaens and Verraes, 1998; Mabee et al., 2000; Sanchez-Villagra, 2002; Prochel, 2006), life-history (Prochel 2006), and other untested ecological variables such as temperature, humidity and geography result in variation both between and within species (Sheil and Greenbaum, 2005). Some or all of these factors may affect the observed sequence, and can be difficult to isolate as they often interact and influence one another. In addition, the degree to which these variables affect ossification sequence is not consistent across all taxa. For instance, in organisms with a feeding larval stage such as many fish and amphibians, it is important for maxillary and palatal elements to be ossified for efficient feeding and to prevent damage to the brain caused by large food items (Adriaens and Verraes, 1998). In birds, however, the palatal, maxillary and mandibular elements ossify well before any feeding activity, and so an adaptationist explanation cannot explain their early formation. Rather, developmental constraint or phylogenetic retention of these traits from an ancestral condition in which they were selectively advantageous may better explain the pattern observed.

115 Low levels of polymorphism do not discount the potential phylogenetie utility of ossification sequence data. The amount of ossification sequence variation detected is not significantly greater than in more traditional morphological data sets (Berger, 1956; Mabee et al., 2000; Maisano, 2002b; Maxwell and Larsson, 2007; Maxwell, 2008). Consistent intraspecific variation in some taxa suggests that ecological factors may also play a role (Sheil and Greenbaum, 2005), but these remain poorly defined. Anseriformes (waterfowl) form a monophyletic sister group to Galliformes (game birds) (Cracraft, 1988; Mayr and Clarke, 2003; Sorenson et al., 2003; Chubb, 2004; Fain and Houde, 2004; Gibb et al., 2007; Livezey and Zusi, 2007; Slack et al., 2007). The monophyly of Anseriformes has never been seriously questioned and classification at the family level is stable (Livezey, 1997; Sorenson et al., 1999; Donne-Gousse et al., 2002), however little is known about their skeletal development. Here I describe the late stages of skeletal formation for three ducks, the Muscovy Duck Cairina moschata, the Pekin Duck Anas platyrhynchos, and the Common Eider Somateria mollissima dresseri. These three ducks are thought to be closely related, but the nature of their relationship is debated (Livezey, 1997; Sorenson et al., 1999; Donne-Gousse et al., 2002). Starck (1989) described the ossification sequence of the Muscovy Duck. That sequence and the one presented in this study are largely congruent, but the sequence described here is better resolved. The development of the chondrocranium in the Domestic Duck {Anas platyrhynchos) was described by de Beer and Barrington (1934), allowing for a comparison to be made with the chondrocranial development of the Common Eider and Muscovy Duck, even though no Pekin Ducks of the appropriate stage were sampled in this study. The early development of the limbs of the A. platyrhynchos has been previously described (Sieglbauer, 1911). An understanding of skeletal formation in ducks is important in order to gain an appreciation of the variability and conservation of ossification sequences in birds as a group, and especially to provide a more robust estimate of character polarity for evolutionary interpretations of ossification sequences.

116 MATERIALS AND METHODS

Eggs of the Pekin Duck (Anas platyrhynchos L.) and Muscovy Duck (Cairina moschata (L.)) were purchased from Couvoir'Simetin (Montreal, Canada) and were incubated in the lab at 37.5°C and a relative humidity of 50%. Eggs were turned by hand twice daily. This resulted in a sample of 30 and 28 viable embryos, respectively. Common Eider embryos (Somateria mollissima dresseri (L.)) were collected in the field from Table Bay and Grey Islands, NL Canada (24 embryos; permits SC2323, ST2392). Embryos were fixed in 10% neutral buffered formalin and staged using a normal table designed for Gallus gallus (Hamburger and Hamilton, 1951). Early stages are defined based on limb and branchial arch development, but the development of epidermal structures (feathers, scales, claws, eyelids, beak) is used to define later stages. As the last four stages (40 to 44) are based on the length of the beak and the third toe of G. gallus and are therefore not applicable to other species, these stages are referred to as "40+" in this paper. Specimens were then cleared and double-stained for bone and cartilage, following the procedure outlined by Dingerkus and Uhler (1977). Late-stage embryos were skinned, eviscerated, and soaked in a mixture of the histological clearing agent Citrisolve and ethanol in order to remove subcutaneous and intermuscular adipose deposits prior to staining. Ossification sequences presented for Anas platyrhynchos, Cairina moschata and Somateria mollissima are based on personal observation of specimens housed in the Redpath Museum (RM; Appendix 1). Ossification is described in the text by anatomical region, followed by stage, and lastly by species (Table 4.1); Table 4.2 reflects the sequence of ossification of the entire embryo. The sample sizes used are too small to make a detailed analysis of the variability of ossification sequence data within each species, however previous studies have shown intraspecific variability for this type of data to be within the range of that of other morphological data sets (Mabee et al., 2000; Maisano, 2002b; Maxwell, 2008).

117 Ossification sequences can be compared between taxa by describing the ranked order of ossification events as event-pairs (Smith, 1997; Velhagen Jr., 1997). This allows elements that have been accelerated or delayed in the sequence to be identified. Table 4.3 lists sets of events having a shared sequence in two of the three taxa examined. For instance, if Element A ossifies before Element B in the Muscovy Duck and the Pekin Duck, but A follows B in the Common Eider, the earlier occurrence of A relative to B will be listed as a shared feature of the Muscovy Duck and the Pekin Duck. Table 4.3 summarizes shared sequence changes between species pairs. The element that has the largest number of pairwise timing changes attributed to it is the one that is assumed to have shifted, recorded as either a sequence acceleration or a delay relative to other elements. Events having a constant position in all taxa are not listed, nor are events whose placement in any of the taxa is uncertain due to lack of sequence resolution or variability. Intraspecific variability in the ossification sequence was calculated by dividing the number of variable event-pairs by the total number of event-pairs (2V/N(N -1), where V is the number of variable event-pairs and N is the number of elements ossified in the most ossified individual examined). Anatomical nomenclature follows Baumel and Witmer (1993). The digits of the manus are numbered II, III and IV.

RESULTS

Sequence variability The oldest Pekin Duck examined had 80 ossified elements, which can be described by 3160 event-pairs. Of these, only 36 were variable (i.e., in some individuals, B ossified before A, whereas A ossified before B in others). Therefore 36/3160, or 1.1% of all events showed intraspecific variability. The oldest Muscovy Duck examined had 81 ossified elements, resulting in 3240 event-pairs of which 224 were variable; therefore in this species, 6.9% of ossification event-pairs were intraspecifically variable. The oldest Common Eider examined had 87 ossified elements, resulting in 3741 event-pairs, of which 208 were variable. This leads to a variability estimate of 5.6%. This measure is biased

118 by the absolute number of embryos sampled for each species, as well as by the number of embryos sampled at each stage of incubation.

Duck skeletal development ' A stage-by-stage description of the skeletal development of the three species is presented. The stage at which the onset of onset of ossification occurs is summarized in Table 4.1, and the ossification sequence is summarized in Table 4.2. Results are organized by anatomical region for comparative purposes, both in the text and in the tables.

Skull (Fig. 4.1) Stage 31 Cairina moschata (Fig. 4.ID): An abrupt angle is present between the prenasal process and the interorbital septum, as the parietotectal cartilage of the nasal capsule has not started growing posteriorly, nor has the beak elongated anteriorly. The dorsoventral flattening of the prenasal process of the nasal septum, typical of ducks, has not occurred. The planun antorbitale is clearly present, and the parietotectal cartilage forms a bridge over the nasal capsule. The postorbital cartilage forms the posterior border of the orbit. The quadrate cartilage is in direct contact with Meckel's cartilage, and the processus retroarticularis is elongate, extending posterior to the jaw articulation. The stylohyal cartilage and the columella are visible in the inner ear region. The posterior occipital arch is continuous with the canalicular portion of the otic capsule. The metotic cartilage is rounded and is located at the junction of the occipital arch and the otic capsule.

Stage 33 Somateria mollissima (Fig. 4.1G): The prenasal process is broader in dorsal view, and is elongate relative to the stage 31 Muscovy Duck (Fig. 4.ID). The angle between the prenasal process and the antorbital cartilage has increased. The quadrate cartilage is almost contacting the articular portion of Meckel's cartilage. The retroarticular cartilage is directed downwards; it is not yet extended posteriorly. The hyoid apparatus is short, and does not extend past the jaw

119 articulation. The canalicular portion of the auditory capsule is present, and the two halves do not meet posteriorly.

Stage 34 Cairina moschata: An abrupt angle is still observed between the prenasal process of the nasal septum and the interorbital septum. This angle is smaller than the angle described for the stage 33 Somateria mollissima, although the overall degree of chondrification of the skull is greater. The prenasal process is spatulate anteriorly, as is typical of ducks. The metotic cartilage has elongated dorsally, and now forms the posterior edge of the tympanic cavity. Posteriorly, the supraoccipital is chondrifying. The quadrate cartilage is in articulation with the otic capsule. The cartilaginous epibranchials have lengthened, and now extend further posteriorly than the retroarticular process. In a slightly older specimen, the angle between the beak and the interorbital cartilage has increased due to the anterior extension of the beak and the posterior growth of the parietotectal cartilage of the nasal capsule. The prenasal process remains spatulate. The otic capsule is enlarged, forming the posteroventral braincase. The supraoccipital is fully chondrified, and bridges the pars canaliculi of the otic capsules. Along the ventral margin of the lower jaw, the angular is ossified. Starck (1989) identified the pterygoid as ossifying at approximately this stage, although this was not observed here. Late in this stage, the squamosal, lacrimal, nasal, maxilla and quadratojugal ossify. Ossification of the squamosal is initiated near the otic process of the quadrate. The lacrimal ossifies anterior to the orbit, beginning from its orbital process, and ossification of the maxilla begins from its jugal process. Initiation of the ossification of the quadratojugal occurs at the quadrate - quadratojugal articulation. The nasal is ossifying from a single center posterior to the external narial opening. The beak has greatly elongated.

Stage 35 Anasplatyrhynchos (Fig.4.IB): Chondrification is well-advanced, and ossification has already begun in the first specimen sampled. The squamosal is

120 ossifying, as is the quadratojugal, supra-angular and angular. The supra-angular ossifies from the dorsal margin of the lower jaw. Later in this stage, the lacrimal ossifies, beginning from the orbital process. Simultaneously, the nasal is ossifying from a single center, as in the Cairina moschata. The jugal process of the maxilla, the palatine and pterygoid are also ossifying. The pterygoid and palatine ossification centers are widely separated at the time of their initiation.

Stage 36 Anas platyrhynchos: The frontal process of the premaxilla, the jugal and the ceratobranchial are ossifying. This is followed by the ossification of the nasal process of the maxilla. The dentary is present, beginning ossification from its anteroventral edge. Later in this stage, the otic process of the quadrate and parietal ossify, the latter ossifying from its ventral edge. Cairina moschata (Fig. 4. IE): The frontal process of the premaxilla, the palatine, pterygoid, and jugal are ossified. In the lower jaw, the dentary and supra-angular are ossifying, as is the ceratobranchial. This is followed by the frontal and the splenial. The maxilla is triradiate. Late in this stage, the parietal, parasphenoid rostrum and vomer are present. The parasphenoid rostrum ossifies beginning from its posterior end. Differences in the degree of ossification between Anas platyrhynchos and Cairina moschata are more likely due to gaps in sampling than to actual differences in the timing of events between the two genera at this stage.

Stage 37 Anas platyrhynchos: The frontal is ossifying posterior and dorsal to the orbit. This is followed by the vomer, and later by the parasphenoid rostrum and the splenial. Cairina moschata: Early in this stage, the basisphenoid and quadrate ossify. Somateria mollissima: By this stage, the parasphenoid rostrum, basisphenoid, squamosal, parietal, frontal, lacrimal, nasal, premaxilla, maxilla, palatine, vomer, pterygoid, jugal, quadratojugal, the otic process of the quadrate,

121 dentary, supra-angular, angular, splenial and ceratobranchial are ossifying. The prenasal process is spatulate, but has not yet been surrounded by the premaxilla.

Stage 38 Anas platyrhynchos: The supraoccipital is ossifying from two centers located on its ventral corners. The exoccipitals, parasphenoid lamina and basisphenoid are also now ossifying. Late in this stage, the prearticular ossifies from the anterior surface of the medial mandibular process. Somateria mollissima (Fig. 4.1H): The exoccipitals, and the supraoccipital are ossifying. The latter ossifies from paired ventral centers, as in A. platyrhynchos. The vomer now contacts the palatine. The lacrimal is thickening, with bone filling the region between the anterior process and the supraorbital process. This is followed by the ossification of the basioccipital from paired linear centers. The parasphenoid rostrum and lamina are now ossified. Otoliths are present.

Stage 39 Anas platyrhynchos: The basioccipital is ossifying. Cairina moschata: Early in this stage, the basioccipital is ossifying from paired elongate centers. The parasphenoid alae and laminae, exoccipital, supraoccipital and prearticular are ossifying simultaneously with the basioccipital. This is followed by the ossification of the prootic. Later, the opisthotic ossifies on the medial margin of the exoccipital. The lacrimal remains clearly triradiate, but is beginning to become more robust. This is followed by the ossification of the articular, which ossifies variably in either stage 39 or 40+. Somateria mollissima: The laterosphenoid is ossifying from its ventral edge. The paired ossification centers of the basioccipital have merged to form a single ossification. The frontal and parietal are fused, and the orbital process of the quadrate is ossified along 50% of its length. The prearticular is ossifying.

Stage 40+

122 Anasplatyrhynchos (Fig. 4.1C): The prootic and opisthotic are ossifying, followed by the articular. Later, the laterosphenoid ossifies from its dorsomedial corner, then the epiotic ossifies from the anterior surface of the most ventral process of the supraoccipital. In the oldest individuals, an unpaired ossification situated medial to the ectethmoid and anterior to the interorbital septum develops. This initial center of ossification spreads dorsally to ossify the lamina dorsalis of the mesethmoid Cairina moschata (Fig. 4.IF): The laterosphenoid is ossifying from its ventral edge, followed by the ossification of the epiotic as described for A platyrhynchos. Later, the unpaired lamina dorsalis of the mesethmoid ossifies, also as in A. platyrhynchos. Somateria mollissima: The parietal, frontal and squamosal are in contact. The prootic and articular are ossifying. A large portion of the laterosphenoid remains cartilaginous, but later in this stage it ossifies completely (Fig. 4.11). The opisthotic has appeared as a small ossification center medial to the exoccipital. The recessus conicalis is obvious on the posterior mandible, and there is one ossification center located on either side of the recess. This is followed by the variable ossification of the interorbital portion of the mesethmoid and nasal trabeculae, which may be delayed in many embryos, and the epiotic as in both Anas platyrhynchos and Cairina moschata. The lamina dorsalis of the mesethmoid is consistently ossified late in this stage. Closer to hatching, a triangular ossicle is present at the dorsolateral corner of the palatal process of the maxilla. In a second individual, an ossicle is found dorsal to the palatal process of the maxilla, lying parallel to the process. The identity of these elements is questionable.

123 Figure 4.1. Lateral view of the developing skull of selected Anseriformes. Grey shaded regions represent cartilage; stippled regions represent ossified tissue. The density of stippling reflects the relative degree of ossification. A: Anas platyrhynchos day 9.5 (modified from de Beer and Barrington, 1934). B: A. platyrhynchos stage 35. C: A. platyrhynchos stage 40+ (day 26). D: Cairina moschata stage 31 (late). E: C. moschata stage 36. F: C. moschata stage 40+ (day 27). G: Somateria mollissima stage 33. H: S. mollissima stage 38 (early). I: S. mollissima stage 40+ (day 15). Scale bar equals 2 mm for parts D and G; 5 mm for B, C, E, F, H, I. a, angular; cb, ceratobranchial; d, dentary; eb, epibranchial; f, frontal; foa, foramen for the ophthalmic artery; is, interorbital septum; j, jugal; lac, lacrimal; me, Meckel's cartilage; mx, maxilla; n, nasal; pa, parietal; pal, palatine; pao, planum antorbitale; pea, pars canaliculi of the otic capsule; pmx, premaxilla; pnp, prenasal process; poc, postorbital cartilage; psr, parasphenoid rostrum; pt, pterygoid; q, quadrate; qj, quadratojugal; rp, retroarticular process of Meckel's cartilage; sa, supra-angular; sop, supraoccipital process; sq, squamosal; v, vomer.

124

Postcranial axial skeleton Stage 31 Cairina moschata: The cervical ribs are cartilaginous, small and indistinct. Nine dorsal ribs are chondrified, eight of which have a sternal component. The cartilaginous ilium overlaps the most posterior rib. Approximately 18 post-sacral vertebrae are chondrified, and six haemal arches are present. Somateria mollissima: The dorsal ribs are faintly chondrified, as are their sternal components; the cervical ribs have not yet chondrified. The tail remains long (20+ cartilaginous post-acetabular vertebrae), and the most posterior vertebrae are not yet fully formed.

Stage 33 Somateria mollissima: The cervical ribs are beginning to chondrify. Nine cartilaginous dorsal ribs are present; the uncinate processes have not yet formed. The sternal ribs have also chondrified. Approximately 20 chondrified post- acetabular vertebrae are present.

Stage 34 Cairina moschata: Cervical ribs have become much more visible. Nine cartilaginous dorsal ribs are present; uncinate processes have chondrified on four. 15 to 16 post-acetabular vertebrae are chondrified, with five cartilaginous haemal arches. Late in this stage, the number of free haemal arches is reduced to three; however this character exhibits some intraspecific variability.

Stage 35 Anasplatyrhynchos: Four uncinate processes have chondrified, but there is no sign of ossification in the postcranial axial skeleton. There are two free haemal arches present.

Stage 36 Anas platyrhynchos: The dorsal ribs are ossifying.

125 Cairina moschata: Five uncinate processes have chondrified; this count increases to six during stage 36. Four free haemal arches are present. Towards the end of this stage, or early in stage 37, the dorsal ribs ossify.

Stage 37 Anas platyrhynchos: The cervical centra are ossifying from paired ventral centers. Cairina moschata: Only four uncinate processes are chondrified in one of the individuals examined at this stage, suggesting variability in either the timing of chondrification or in the total number of processes. Somateria mollissima: The dorsal ribs are ossified, the middle six of which have chondrified uncinate processes.

Stage 38 Anas platyrhynchos: Late in this stage, the cervical ribs ossify. Somateria mollissima: The cervical centra are ossifying in an anterior to posterior direction. In the most posterior centra, paired ventral ossification centers are visible; otherwise unpaired dorsal and ventral centers are present within a centrum. The anterior thoracic centra possess a higher degree of ossification than the posterior cervical centra, suggesting two different sites of initiation of ossification in the vertebral column. The first 10 synsacral centra are ossified from paired ventrolateral ossification centers. Five cartilaginous uncinate processes are present on the dorsal ribs, indicating that this character is variable, as in Cairina moschata. Later, the cervical ribs ossify, beginning with the fourth cervical rib. Ossification proceeds bidirectionally.

Stage 39 Anas platyrhynchos: The cervical centra still have only the ventral ossification center present. The thoracic centra have begun to ossify, and the synsacral centra have right and left paired ossification centers. This is followed by ossification around the base of the transverse processes of the cervical arches.

126 Cairina moschata: In the vertebral column, the cervical centra are ossified, as are the thoracic and synsacral centra. This is followed by the ossification of the most anterior cervical ribs. Late in this stage, the first to third free caudal centra ossify from right and left paired centers. Somateria mollissima: The cervical arches and the two most anterior thoracic arches are ossifying. All of the synsacral centra are ossified; only the most posterior eight retain right and left paired ossification centers.

Stage 40+ Anas platyrhynchos: The cervical arches are ossifying by this stage, and the dorsal and ventral ossification centers of the cervical centra have fused. Only half of the synsacral centra are ossified. This is followed by the ossification of the arches of the two most anterior thoracic vertebrae. Later, the free caudal centra ossify from dorsal and ventral ossification centers. In at least one individual, the dorsal and ventral centers of the cervical centra have not merged, suggesting that there is sequence variability present in the ossification of the vertebral column. This is followed by the ossification of all of the synsacral centra, and the fusion of their paired ossification centers. Next, all thoracic arches ossify. Eight free caudal centra are ossifying, as is one of the most anterior vertebral components of the pygostyle. Lastly, the transverse processes of the synsacral vertebrae, then the synsacral arches and the caudal neural arches ossify. Cairina moschata: The first six cervical ribs are ossified, and the cervical arches are also ossifying but are not closed. The anterior seven free caudal centra are ossifying. Next, the first thoracic arch begins ossification. After this, the vertebral portion of the pygostyle ossifies as a linear ossification center. Later, the caudal neural arches begin ossifying, followed by the synsacral arches and lastly the synsacral transverse processes. In the case of the former, only the narrowest arches in the posterior region of the synsacrum are ossified. Somateria mollissima: Ossification of the thoracic arches proceeds in an anterior to posterior direction. The free caudal centra are also ossifying. This is followed by the ossification of the transverse processes of the synsacrum, then the synsacral arches. The arches are more ossified in the posterior half of the

127 synsacrum, as in C. moschata. Later, the free caudal neural arches ossify, followed by the vertebral components of the pygostyle. Vertebrae located anterior to the pygostyle ossify from single dorsal and ventral centers. Prior to hatching, the sternal keel ossifies, as do the sternal ribs. The latter ossify in an anterior to posterior direction.

Forelimb (Fig. 4.2) A detailed description of the early phases of wing development in A platyrhynchos can be found in Sieglbauer (1911). Stage 31 Cairina moschata (Fig. 4.2C): All elements in the shoulder girdle are cartilaginous. The scapula and coracoid are separate. The sternal plates are chondrified, but their medial edges are unfused. The radiale is fused with the intermedium to form the radiale complex, distal to the radius. A faint chondrification distal to this is the semilunate mass of distal carpals that caps the metacarpals. The ulnare is located directly distal to the ulna, and is larger and located anterior to its position in Anas platyrhynchos (Sieglbauer, 1911). Two elements are present posterodistal to the ulna, the pisiform and metacarpal V. Only the proximal phalanx is present for all manual digits. Somateria mollissima: All elements of the forelimb are cartilaginous. The coracoid and scapula are separate. The sternal cartilages are unfused, and articulate with the coracoid. The radiale complex can be distinguished distal to the radius; the ulnare is located distal to the ulna and the pisiform is located posterodistal to the ulna as in C. moschata. The semilunate mass of distal carpals - capping metacarpal III is visible. A proximal phalanx is present on each of the manual digits.

Stage 33 Somateria mollissima (Fig. 4.2F): The sternal plates are cartilaginous and unfused. The humerus, radius and ulna each ossify from single ossification centers. The radiale complex, the ulnare and the distal carpal complex capping metacarpal III are visible. The Ulnare is relatively smaller and located more

128 posteriorly than in stage 31. Two phalanges each are present on manual digits II and III.

Stage 34 Cairina moschata: The cartilaginous sternal plates have almost entirely fused, beginning from the anterior margin. The sternal keel begins chondrifying from its anterior edge. The crista bicipitalis is clearly present on the humerus. Two phalanges each have formed on the manual digits. The humerus, radius and ulna are ossifying, followed by the furcula. The latter ossifies from two rami that later fuse medially. All manual phalanges are chondrified. Late in this stage, metacarpal III ossifies.

Stage 35 Anas platyrhynchos (Fig. 4.2A): The furcula, humerus, radius and ulna are ossifying. All phalanges are chondrified. This is followed by the ossification of metacarpals III and IV. The radiale, metacarpal V and the semilunate carpal mass capping metacarpal III are cartilaginous.

Stage 36 Anas platyrhynchos: The scapula is ossifying. Cairina moschata (Fig. 4.2D): The scapula and the metacarpal IV are ossifying. In the carpal region, the distal carpals and metacarpal II are fused to each other, but remain separate from the rest of the carpometacarpus. The radiale complex is triangular in shape.

Stage 37 Anas platyrhynchos: The coracoid is ossifying, as is the proximal phalanx of digit II. This is followed by the ossification of the two most proximal phalanges of digit III. Cairina moschata: The coracoid is ossifying, as are manual phalanges 11:1,2 and 111:1,2,3; however in some more ossified individuals manual phalanges

129 11:2 and 111:3 are not ossified indicating variability in the timing of ossification in the terminal phalanges. Somateria mollissima: The scapula, coracoid, furcula, both phalanges of digit II, metacarpals HI and IV are ossifying. The ossification of the coracoid may be delayed until after phalanx 111:2, and the ossification of phalanx 11:2 may be delayed until after stage 38. Later, manual phalanx 111:1 ossifies, followed by 111:2. The radiale complex and pisiform are distinct; the semilunate mass is continuous with metacarpal II but is not fused to the rest of the carpometacarpus. The furcula has an ossified extension contacting the scapula.

Stage 38 Anasplatyrhynchos: Manual phalanx 11:2 may ossify during this stage, although as in Somateria mollissima and Cairina moschata, intraspecific variability is present.

Stage 39 No significant changes in the manus are noted for this stage. All genera show polymorphism in the ossification of the terminal phalanges of digits II and III, with these elements variably ossifying during this stage.

Stage 40+ Anas platyrhynchos (Fig. 4.2B): Metacarpal II ossifies prior to hatching. Somateria mollissima (Fig. 4.2H): Manual phalanx IV:1 is ossifying, followed by metacarpal II. The latter element ossifies from a center located between it and metacarpal III. This center is distinct from that of metacarpal III, as the ossified region of metacarpal III has not extended that far posteriorly.

130 Figure 4.2. Lateral view of the forelimb of selected Anseriformes. Grey shaded regions represent cartilage; black regions represent ossified tissue. The density of stippling reflects the relative degree of ossification. A: Anas platyrhynchos stage 35. B: A. platyrhynchos stage 40+ (day 23). C: Cairina moschata stage 31 (late). D: C. moschata stage 36. E: C. moschata stage 40+ (day 27). F: Somateria mollissima stage 33. G: S. mollissima stage 38 (early). H: S. mollissima stage 40+ (day 19). Scale bar equals 2 mm (parts A, B, C, D, F, G) and 5 mm (parts E, H). dec, distal carpal complex; furc, furcula; H, humerus; mcll - V, metacarpals; R, radius; ra, radiale; scap, scapula; U, ulna; ul, ulnare.

131 scap- Pelvic girdle and hind limb (Fig. 4.3) Early stages in the development of the hind limb of Anas platyrhynchos are described by Sieglbauer (1911). Stage 31 Cairina moschata (Fig. 4.3C): All hind limb elements are cartilaginous for the duration of this stage. The ischium and pubis are directed posteroventrally, as is the posterior ilium. The three elements are not in contact posteriorly. The hallux is positioned halfway along the tarsometatarsus and has not rotated posteriorly. The fibula is detached from the tarsal region. The central tarsals are fused to form a single element, but the fibulare is separate. The distal tarsals are also fused to form a single narrow element capping the metatarsals. Two phalanges are present on all pedal digits. Somateria mollissima: All hind limb elements remain cartilaginous for the duration of this stage. The ilium, ischium and pubis are chondrified. The pubis is rotating posteriorly but has not yet assumed its adult orientation. The ilium has not expanded anteriorly or posteriorly, and does not contact the ischium. The fibula remains connected to the tarsal region; the fibulare, tibiale and central tarsals are distinct. Metatarsal I remains close to the tarsal region, rather than at the distal end of the tarsometatarsus. The metatarsals are separate and parallel; they have not yet fused. One to two pedal phalanges are present on all digits. This embryo is slightly younger than the Muscovy Duck embryo described previously.

Stage 33 Somateria mollissima (Fig. 4.3F): The ilium does not yet overlap the most posterior dorsal rib. The ischium and pubis are directed posteriorly, but do not contact each other or the posterior ilium. The femur, tibia, fibula and metatarsals II, III and IV are ossifying from single centers; ossification, however, is less advanced than in the forelimb. In the tarsal region, a large fibulare, an equally large distal tibial centrale and a smaller, medially located proximal tibial centrale are visible. The hallux is detached from the tarsal region, but has not completely rotated posteriorly. The metatarsals are approaching each other proximally, and

132 metatarsal III has slipped behind metatarsals II and IV. Two cartilaginous phalanges are present on each of the pedal digits.

Stage 34 Cairina moschata: The anterior wing of the ilium overlaps the most posterior one to two ribs. The ilium, ischium and pubis are in contact posteriorly. The hallux is located further distally, and three phalanges are visible on pedal digits II-IV. The femur, tibia, fibula and metatarsals II, III and IV are ossifying. By the end of this stage, all pedal phalanges have chondrified.

Stage 35 Anas platyrhynchos (Fig. 4.3 A): The ilium extends anteriorly over the most posterior two ribs. The femur, tibia, fibula, and metatarsals II, III and IV are ossifying. All of the pedal phalanges have chondrified.

Stage 36 Anas platyrhynchos: The ilium has extended anteriorly over the three most posterior dorsal ribs. Pedal phalanges 111:1 and IV: 1 are ossifying. This is followed by the ossification of pedal phalanx 11:1. Cairina moschata: The ilium extends anteriorly over the three most posterior ribs. The ischium is the first element of the pelvic girdle to ossify. Later in this stage, the pubis ossifies, followed by pedal phalanges 11:1,2,111:1, and IV: 1,4.

Stage 37 Anas platyrhynchos: The ilium extends over the four most posterior ribs. Ossification is beginning from the posterior wing of the ilium, and also from a separate center in the ischium. Pedal phalanx 11:2 is ossifying, followed by pedal phalanges 11:3, and 111:4. Later, the pubis ossifies, then pedal phalanges 111:2 and IV:5. Cairina moschata (Fig. 4.3D): Pedal phalanges 11:3,111:3,4, and IV: 1,5 are ossifying, followed by pedal phalanges 1:1,2,11:1,2,3,111:1,3,4, and IV:1,4,5. Late

133 in this stage, the anterior and posterior wing of the ilium ossify. In some individuals, pedal phalanges 1:2,11:3,111:4 or IV:5 are not ossified by this point, indicating sequence variation; however by the end of this stage, all pedal phalanges except 1V:3 are consistently ossified. Somateria mollissima: The ischium, pubis, and pedal phalanges 11:1,3, 111:1,4, and IV:1,5 are ossified, followed by pedal phalanges 1:1,11:2,111:2,3 and variably by 1:2. In the pelvic girdle, the ilium has extended anteriorly over the last four dorsal ribs. Late in this stage, it ossifies from two separate centers of ossification, one in the anterior wing and one in the posterior wing dorsal to the acetabulum.

Stage 38 Anas platyrhynchos: The anterior wing of the ilium is ossifying. All pedal phalanges except IV:3 are ossified by this stage. Somateria mollissima (Fig. 4.3G): All pedal phalanges except IV:3 are ossified, and by the end of this stage this phalanx is also ossified.

Stage 39 Anas platyrhynchos: Metatarsal I ossifies, followed by pedal phalanx IV:3. Cairina moschata: The pretibial bone is beginning to ossify; only an initial ossification center is present. Although its timing of appearance is variable relative to other skeletal elements, it is always present by the end of this stage.

Stage 40+ Anasplatyrhynchos: The pretibial bone is ossifying, although the portion extending into the tarsal region is not present until much later in this stage (Fig. 4.3B). The ossification of metatarsal I may be delayed until this stage in some individuals. Cairina moschata: Metatarsal I is ossifying from its proximal end, followed by pedal phalanx IV:3 (Fig. 4.3E). Somateria mollissima: Metatarsal I and the pretibial bone ossify early in this stage (Fig. 4.3H). This is followed by the ossification of two proximal tarsals,

134 the fibulare, then the tibiale. Towards the end of this stage, the hypotarsus ossifies from a separate center located on its distal medial corner.

135 Figure 4.3. Lateral view of the hind limb of selected Anseriformes. Grey shaded regions represent cartilage; black regions represent ossified tissue. The density of stippling reflects the relative degree of ossification. A: Anas platyrhynchos stage 35. B: A. platyrhynchos stage 40+ (day 23). C: Anterior view of the leg of Cairina moschata stage 31 (late). D: Lateral view of C. moschata stage 35. E: C.moschata stage 40+ (day 28). F: Somateria mollissima stage 33. G: S. mollissima stage 38 (early). H: S. mollissima stage 40+ (day 19). Scale bar equals 5 mm. F, femur; fi, fibula; il, ilium; isch, ischium; ptb, pretibial bone; pub, pubis; tib, tibia; tmt, tarsometatarsus.

136

DISCUSSION

Variability in ossification sequence Variability estimates for the three duck species examined ranged from 1.1% to 6.9%. These numbers are within the range of other variability estimates for ossification sequence data (Maisano, 2002b), and are lower than reported values for other birds (Maxwell, 2008). The eggs of Somateria mollissima were collected directly from the nests, but did not show higher ossification sequence variability than domestic species whose eggs were artificially incubated under constant conditions in the lab. This suggests that incubation conditions may not have a strong effect on ossification sequence in birds, although more experiments are necessary to confirm this. The higher genetic diversity in the wild population of eiders (compared to the farmed species, which have a relatively small effective population size) also did not appear to lead to higher sequence variability.

Comparative ossification sequence Skull The majority of dermal skull elements in Anseriformes ossify from single ossification centers. For instance, the nasal ossifies from an ossification center posterior to the external narial opening, similar to the pattern observed in Galliformes and the Double-crested Cormorant (Phalacrocorax auritus) (pers. obs.), but differing from the two centers observed in Charadriiformes and Coliiformes (Goldschmid, 1972; Maxwell and Harrison, In Press). The premaxilla also forms from a single ossification center in Anseriformes, unlike the state reported for some Charadriiformes (Maillard, 1948). The supraoccipital in Anseriformes forms from two ossification centers. This differs from the single median center reported for Meleagris gallopavo (Maxwell, 2008), but two centers were reported for Gallus gallus (Jollie, 1957). The amount of variation observed within Galliformes suggests that this feature cannot be interpreted as an anseriform synapomorphy. In anseriforms, the angular is the first skull bone to ossify (Table 4.2). In galliforms, the first element to ossify is the quadratojugal (Maxwell, 2008). This

137 is a relatively minor difference, as both elements are part of the earliest phase of cranial ossification in both orders. The early phase of cranial ossification is approximately simultaneous with the onset of ossification in the stylopodial and zeugopodial elements. The vomer ossifies prior to the basisphenoid, occipital series (basioccipital, supraoccipital, and exoccipital), and prearticular in Anseriformes (Table 4.2). This represents a considerable acceleration relative to its sequence position in Galliformes, where it ossifies after many of these elements (Jollie, 1957; Schumacher and Wolff, 1966b). This might be due to the relatively greater contribution of the vomer to the palate in ducks; it is greatly reduced in Galliformes (Beddard, 1898). In all anseriforms examined, ossification of the basisphenoid is not accelerated relative to the quadrate and parietal (Table 4.2). This is a characteristic shared with Galliformes (Maxwell, 2008), but not with Charadriiformes, in which the basisphenoid ossifies well before the latter elements (Maillard, 1948; Maxwell and Harrison, In Press). Whether this shift will be demonstrated to be of phylogenetic importance awaits increased taxon sampling within Neoaves. Unlike Gallus gallus, the anseriform species examined begin to ossify the mesethmoid from a point anterior to the orbit, rather than within the orbit. Ossification then proceeds dorsally, to include the lamina dorsalis of the mesethmoid prior to hatching. This characteristic is shared with ratites (pers. obs.). In Anseriformes, however, the lamina dorsalis is completely covered by the frontal process of the premaxilla prior to ossification and therefore lacks dorsal exposure. It fails to ossify from a separate center of ossification. In Anasplatyrhynchos the metotic cartilage arises as an independent. chondrification located at the posterior end of the metotic fissure and ventral to the pars canaliculi of the otic capsule. It grows anteriorly and dorsally, forming the posterior border of the tympanic cavity and almost reaches the otic process of the quadrate cartilage (Fig. 4.1 A), as in Cairina moschata (Fig. 4.ID). The prenasal process of the nasal septum also follows a similar ontogenetic trajectory, beginning as a straight outgrowth and later assuming a distal swelling at the tip (de Beer and Barrington, 1934). The general ontogenetic trajectory is similar

138 between A. platyrhynchos and C. moschata, suggesting general conservation of chondrocranial development within ducks. Variability has been reported between different orders of birds (reviewed by Romanoff, 1960). Although general features are conserved, variation in the relative timing of chondrocranial developmental events is present between species of ducks. For instance, the prenasal process in Somateria mollissima flattens before the retroarticular process extends posteriorly and Meckel's cartilage and the quadrate cartilage become tightly associated with each other (Fig. 4.1G). This is contrary to what is observed in either late stage 31 Cairina moschata embryos, or day 9 A. platyrhynchos embryos (Fig. 4.1 A) (de Beer and Barrington, 1934). There appears to be a high degree of ossification sequence conservation in the early formation of the skull in Anas platyrhynchos and Cairina moschata. An exception to this is the supra-angular, which is accelerated relative to the jugal, pterygoid and palatine in the ossification sequence of A. platyrhynchos. Major sequence differences between the three genera examined here became evident in stage 37. The ossification of the parietal and the quadrate are accelerated, and the basisphenoid is delayed in A. platyrhynchos relative to C. moschata (Table 4.1). Ossification of the laterosphenoid is greatly accelerated in Somateria mollissima relative to both Cairina moschata and Anas platyrhynchos, occurring by stage 39 (Table 4.1). Its ossification precedes that of the prootic, opisthotic and articular (Table 4.2). In Meleagris gallopavo, the laterosphenoid ossifies variably during either stage 39 or 40+; it also is variable in sequence relative to the prootic and opisthotic. Increased sample size is necessary to determine whether this variability is also found in S. mollissima. Unlike the situation observed in galliforms (Maxwell, 2008), the prenasal process of the nasal septum acquires a dorsoventrally flattened, rounded anterior lobe in anseriforms. The premaxilla, the ossified element that makes up the beak in the adult bird, forms around this cartilaginous process leading to a dorsoventrally flattened and more elongate bill. The developmental processes leading to beak formation are well understood in birds, and much of the work has been done using Anas platyrhynchos as a model (Schneider and Helms, 2003; Wu et al., 2004; Wu et al., 2006). The cranial neural crest is responsible for generating

139 both the underlying cartilage and the premaxilla (Schneider and Helms, 2003; Evans and Noden, 2006). The Bmp4 pathway controls beak size and shape (Wu et al., 2004; Wu et al., 2006), specifically the size and shape of the underlying prenasal cartilage (Abzhanov et al., 2004). Whether it affects the shape of the premaxilla directly, or indirectly by changing the shape of the cartilaginous template (proposed by Richman et al., 2006) has not yet been demonstrated. The paired premaxillae ossify around the prenasal process posterior to its anterior tip (Fig. 4. IE, H), and only much later in development encases the anterior tip of the process. This allows for evolutionary and developmental flexibility in beak outgrowth, and demonstrates that even in well-studied instances, the shape and pattern of bone formation is not well understood and may not be under strict genetic control (Hall, 2001). In spite of large morphological differences in the shape of the craniofacial skeleton between Galliformes and Anseriformes, no major ossification sequence changes are observed between the two orders.

Postcranial axial skeleton The cervical ribs in Somateria mollissima do not ossify in an anterior to posterior direction. Although this is the first time this dissociation in ossification sequence from chondrification sequence has been noted for cervical ribs in birds, this is not exceptional when the entire skeleton is considered. Loss of an anterior to posterior gradient in the Ossification of the axial skeleton is not uncommon among amniotes (Strong, 1925; Rieppel, 1993b; Nakane and Tsudzuki, 1999; Sheil, 2003a; Blom and Lilja, 2004; Maxwell, 2008). Unlike the situation in Galliformes (Maxwell, 2008), the uncinate processes fail to ossify prior to hatching in any of the ducks examined. Sternal ribs ossify relatively late in Anseriformes when compared to Galliformes, being among the last skeletal elements to ossify in Somateria mollissima. They remain unossified in the oldest Cairina moschata and Anas platyrhynchos embryos examined, even in those embryos showing ossification of the pygostyle (Table 4.2). This differs from Galliformes, where the ossification of the sternal ribs occurs at approximately the same time as the vertebral arches (Maxwell, 2008).

140 Also unlike Galliformes, Anseriformes lack ossified laterocranial and laterocaudal sternal processes at hatching (Maxwell, 2008). The ossification of the pygostyle in birds reflects its dual origin from fused caudal centra and the posterior notochord (H. Larsson, pers comm). In Anseriformes, the pygostyle ossifies as individual vertebral centra prior to hatching. These vertebrae fused as cartilages to form the pygostyle, but retain separate ossification centers. The notochordal portion of the pygostyle ossifies last. This is somewhat different from Meleagris gallopavo, where the notochordal portion of the pygostyle ossifies first and then ossification spreads anteriorly (Maxwell, 2008).

Appendicular skeleton: forelimb Somateria mollissima initiates ossification in the wing prior to the chondrification of all manual phalanges (Fig. 4.2F). This is unlike the state described for Anas platyrhynchos, where the final phalangeal count is present prior to the onset of ossification (Sieglbauer, 1911). S. mollissima also demonstrates an unusually high degree of ossification in the pectoral skeleton prior to hatching. For instance, the proximal phalanx of metacarpal IV ossifies before the mesethmoid (Table 4.2). Although this phalanx ossifies prior to hatching in the Domestic Turkey (Meleagris gallopavo), the Japanese Quail (Coturnix coturnix) (Nakane and Tsudzuki, 1999; Maxwell, 2008) and variably so in Gallus gallus (Hogg, 1980), it ossifies quite late in galliforms and is always preceded by the mesethmoid. Both S. mollissima and A. platyrhynchos also ossify metacarpal II prior to hatching, which has not been reported for birds. A. platyrhynchos is not a wing-propelled diving duck, so the early ossification of this element cannot be considered an adaptation for increased locomotory performance. There is a great deal of intraspecific variability in the timing of ossification of the manual phalanges in ducks; however they always follow a proximal to distal sequence of ossification, unlike the pedal phalanges. This dissociation between ossification sequence in the manus and pes is not uncommon in amniotes (Rieppel, 1994; Sheil, 2005; Maxwell, 2008), and may be expected in birds based on the amount of morphological divergence between the fore- and

141 hind limbs. The three genera examined all ossified two phalanges on digit II and three phalanges on digit III prior to hatching.

Appendicular skeleton: hind limb The early chondrification of the hind limb appears to be quite conserved within ducks. There were no major differences observed between the stage 31 embryos of Somateria mollissima and Cairina moschata observed here and the day 7 to 8 embryos of Anas platyrhynchos described by Sieglbauer (1911). However, the onset of ossification of the metatarsals occurs earlier relative to phalangeal chondrification in S. mollissima than in A, platyrhynchos (Sieglbauer, 1911). This acceleration of ossification relative to chondrification in S. mollissima was also observed in the forelimb. Both Somateria mollissima and Anas platyrhynchos share the initial ossification of the ischium and pubis, followed by the ilium (Table 4.2). In Cairina moschata, the pubis is the last element of the pelvic girdle to ossify, although Starck (1989) reported that the ilium was the last element to ossify in this species as well. Sequence variability for the three elements of the pelvic girdle is also widespread in Galliformes (reviewed by Maxwell, 2008). The pedal phalanges do not ossify in a proximal to distal sequence, and although the sequence is variable, the third phalanx of digit IV is consistently last to ossify in all genera (Table 4.2). This is also true of Galliformes (Schinz and Zangerl, 1937; Maxwell, 2008), but not for pigeons (Schinz and Zangerl, 1937) terns (Maxwell and Harrison, In Press), or ratites (Chapter 2) and may be specific to Galloanseres. The dissociation of the sequence of ossification from the sequence of chondrification in the limbs is even more widely reported than in the axial skeleton, having been found in a taxonomically diverse sample of tetrapods (Sheil, 2005; Frobisch, In Press). The pretibial bone, two tarsals and the hypotarsus ossify in Somateria mollissima prior to hatching. Although the pretibial bone ossifies in many birds prior to hatching (McGowan, 1984, 1985), the presence of tarsals prior to the ossification of the pygostyle and sternal ribs is unusual. Tarsals have been reported as ossified in the hatchling chicken (Hogg, 1980; McGowan, 1985), but

142 were absent in all embryonic specimens examined in a second study (Maxwell, 2008). Aside from this, ossified tarsals have been observed in embryonic buttonquails (Turnicidae), megapodes (Galliformes) (Starck, 1993) and ratites (McGowan, 1985). Megapodes are considered unusual for their high degree of precociality at hatching (Starck and Ricklefs, 1998), but this has been attributed to their extremely long incubation period (Starck and Sutter, 2000). S. mollissima has an incubation period that is shorter than the other ducks examined here (26 days versus 28 days for Anasplatyrhynchos, and 35 days for the Cairina moschata), and so a different explanatory mechanism is required.

Life history and ossification sequence

Somateria mollissima is considerably more ossified prior to hatching than either Cairina moschata or Anas platyrhynchos, even though its incubation period is shorter and all three species are equally precocial. There are several potential explanations for this phenomenon. It is possible that S. mollissima has a slower post-hatching growth rate than either of the two domestic species. A high growth rate can be achieved by retaining a high ratio of cartilage to ossified tissue in the skeleton (Starck, 1996). Populations of the pumpkinseed (Lepomis gibbosus) with faster growth rates are found to have fewer ossification centers when compared to slower-growing populations of the same age (Arendt and Wilson, 2000). However, hatchlings of migratory species are under strong selection for fast fledging times (Meiri and Yom-Tov, 2004), and this should be reflected in relative growth rates. For instance, the Horned Grebe (Podiceps auritus) is a medium-distance migrant that breeds at high latitudes and has precocial young (Stedman, 2000) and its hatchlings are inferred to exhibit less ossification as they lack an ossified prootic, most of the synsacral and caudal centra, pygostyle and tarsals (Schinz and Zangerl, 1937; Starck, 1998). S. mollissima in southern Labrador have a short time available for development due to the high latitudes at which they nest, and are short-distance migrants (Goudie et al., 2000). The Pekin Duck (A. platyrhynchos) is descended from the Mallard (also A. platyrhynchos), a facultative migrant. Both S. mollissima and A. platyrhynchos should therefore be

143 less ossified than C. moschata, a non-migrant tropical species. This explanantion is not consistent with what is observed, suggesting that a potential tradeoff exists between growth rate and ossification in ducks. In order to rule out the effects of domestication and selection for rapid growth rate leading to a more cartilaginous skeleton at hatching, the skeletal development of the Mallard must be compared to that of the Pekin Duck.

Alternatively, Somateria mollissima may show a higher degree of ossification due to the high locomotor performance required for foot- and wing- propelled diving. A high degree of ossification has been suggested to improve locomotor performance, as well as the structural integrity of the skeleton (Kirkwood et al., 1989; Arendt and Wilson, 2000). S. mollissima ducklings can dive and procure their own food within 24 hours of hatching. The young are frequently grouped together in large creches; there is little parental brooding (Goudie et al., 2000). Anas platyrhynchos and Cairina moschata do not dive, and so locomotory performance may not be as important. Although the young of the Horned Grebe (Podiceps auritus) are capable of diving immediately after hatching, they are fed and transported by the parent up to 10 days after hatching (Stedman, 2000) and so locomotor performance is also not as critical. If selection for increased locomotory performance is driving the high degree of ossification observed in Somateria mollissima, is selection acting on a higher degree of ossification, or on the precocial development of muscles or muscle flexion that induces mechanical strain leading to bone formation? This mechanism has been suggested to influence the time of appearance and shape of ossification centers in fishes (Adriaens and Vermes, 1998), as well as the ossification of the exoccipital in marsupials (Smith, 1996). However, in much of the marsupial skeleton, muscles and tendons develop significantly before the ossification of their attachment points (Smith, 2006). The timing of muscle development is unknown in Somateria mollissima, and so the cause and effect relationship cannot be determined at the present time. Whereas Galliformes ossify the uncinate processes, sternal ribs, laterocranial and laterocaudal sternal processes relatively early, these ossifications

144 are all either greatly delayed or absent in the ducks examined. These elements all contribute to ventilatory capacity and rigidity of the thorax. The late-stage ossifications found in ducks, on the other hand, are generally found in the appendicular skeleton; for instance, metacarpal II and the proximal phalanx of the minor metacarpal in the wing, and the tarsals and hypotarsus in the hind limb. Whereas the increased resistance encountered while moving through water when compared to movement in air might result it increased appendicular ossification in Anseriformes, the reason for the increased number of thoracic ossifications in Galliformes is less clear. It is possible that increased thoracic rigidity is selectively advantageous during bipedal running.

CONCLUSION

f The large morphological changes to a generalized beak resulting in a duck bill, as well as selection for faster post-hatching growth rates in migratory species are not reflected in the ossification sequence of Anseriformes. Differences in the ossification of migratory altricial species compared to resident altricial species need to be examined, because in these birds the influence of locomotory function in hatchlings is minimal, allowing for the effective separation of these variables. The impact of artificial selection for higher growth rate on ossification sequence and degree of ossification also needs to be confirmed. Anseriformes are highly ossified prior to hatching, but the identity of these late-stage ossifications is different than in their sister group, Galliformes. This implies that highly ossified hatchlings evolved independently and for different reasons in these two avian orders, in spite of the fact that precociality is primitive for birds.

145 TABLE 4.1: Summary of elements ossified by stage and by day of incubation for selected anseriform species.

Skull Anas Cairina moschata Somateria mollissima Stage platyrhynchos 31 Day 12 Day 6 No ossification No ossification

33 Day 9

34 Day 13-16 Lacrimal, maxilla, nasal, quadratojugal, squamosal Angular

35 Day 11-11.5 Quadratojugal, squamosal, lacrimal, maxilla, nasal, palatine, pterygoid Angular, supra- angular

36 Day 12-13 Day 15-17 Jugal, premaxilla, Palatine, parietal, quadrate parasphenoid Dentary rostrum, pterygoid, Ceratobranchial vomer, jugal, premaxilla, frontal, parietal Dentary, supra- angular, splenial Ceratobranchial

37 Day 14-15 Day 16-21 Day 10-13 Frontal, vomer, Basisphenoid, Basisphenoid, frontal, parasphenoid quadrate jugal, lacrimal, rostrum maxilla, nasal, Splenial palatine, parasphenoid rostrum, parietal, premaxilla, pterygoid, quadrate, quadratojugal, squamosal, vomer Angular, dentary, splenial, supra- angular Ceratobranchial

146 38 Day 16 Day 10-16 Basisphenoid, Exoccipital, exoccipital, supraoccipital, parasphenoid alae, basioccipital, parasphenoid parasphenoid alae, lamina, parasphenoid lamina, supraoccipital Prearticular

39 Day 17-18 Day 22-24 Day 13-16 Basioccipital Parasphenoid Laterosphenoid lamina, Prearticular parasphenoid alae, basioccipital, exoccipital, supraoccipital, prootic, opisthotic Prearticular, articular*

40+ Day 18-26 Day 25-30 Day 13-24 Prootic, opisthotic, Laterosphenoid, Prootic, opisthotic, laterosphenoid, epiotic, mesethmoid epiotic, mesethmoid, mesethmoid, Articular* nasal trabeculae epiotic Articular Articular

Postcranial Anas Cairina moschata Somateria mollissima axial platyrhynchos skeleton Stage 31 Day 12 Day 6 No ossification No ossification

33 Day 9 No ossification

34 Day 13-16 No ossification

35 Day 11-11.5 No ossification

36 Day 12-13 Day 15-17 Dorsal ribs Dorsal ribs*

37 Day 14-15 Day 16-21 Day 10-13 Cervical centra Dorsal ribs* Dorsal ribs

38 Day 16 Day 10-16

147 Cervical ribs Cervical centra, thoracic centra, synsacral centra, cervical ribs

39 Day 17-18 Day 22-24 Day 13-16 Thoracic centra, Cervical centra, Cervical neural synsacral centra thoracic centra, arches, thoracic synsacral centra, neural arches cervical ribs, caudal centra

40+ Day 18-26 Day 25-30 Day 13-24 Cervical neural Cervical neural Caudal centra, arch, thoracic arches, thoracic synsacral neural neural arch, caudal neural arches, arches, synsacral centra, pygostyle, pygostyle, caudal transverse processes, synsacral neural arches, caudal neural arches, transverse synsacral neural pygostyle, sternal processes, arches, synsacral ribs, sternal keel synsacral arches, transverse processes caudal neural arches

Forelimb Anas Cairina moschata Somateria mollissima Stage platyrhynchos 31 Day 12 Day 6 No ossification No ossification

33 Day 9 Humerus, radius, ulna

34 Day 13-16 Humerus, radius, ulna, furcula, metacarpal III

35 Day 11-11.5 Humerus, furcula, metacarpals 111, IV, radius, ulna

36 Day 12-13 Day 15-17 Scapula Metacarpal IV, scapula,

37 Day 14-15 Day 16-21 Day 10-13 Coracoid, manual Coracoid, manual Coracoid, furcula, phalanx 11:1, phalanges 11:1,2; metacarpals 1II-1V, 111:1,2 111:1,2,3* manual phalanges 11:1,2*, scapula,

148 manual phalanges III:1*,2*

38 Day 16 Day 10-16 Manual phalanx Manual phalanx 11:2 11:2*, 111:1 *,2*

39 Day 17-18 Day 22-24 Day 13-16 Manual phalanx Manual phalanx 11:2* 111:3*

40+ Day 18-26 Day 25.-30 Day 13-24 Metacarpal II, Manual phalanx Metacarpal II, manual manual phalanges 111:3* phalanges 11:2*, 111:3, 11:2*, 111:3 IV: 1

Hind limb Anas Cairina moschata Somateria mollissima Stage platyrhynchos 31 Day 12 Day 6 No ossification No ossification

33 Day 9 Femur, fibula, metatarsals II-IV, tibia

34 Day 13-16 Femur, fibula, metatarsals II-IV, tibia

35 Day 11-11.5 Femur, fibula, metatarsals II-IV, tibia

36 Day 12-13 Day 15-17 Pedal phalanges Ischium, pubis, 111:1, IV:1* 11:1 pedal phalanges II:1*,2*; 111:1*, IV:1*,4*

37 Day 14-15 Day 16-21 Day 10-13 Ilium, ischium, Pedal phalanges Ischium, pubis, pedal pedal phalanx 11:2, II:1*,2*, 111:1*, phalanges 11:1, pubis, pedal IV:1*,4*, 11:3, 111:1,4, IV:1,5, ilium, phalanges 11:3, 111:3,4,1V:5,1:1,2, pedal phalanges 111:4,2,1V:1*,5 111:2, ilium, pedal II:2*,3, III:2*,3* phalanx IV :2

38 Day 16 Day 10-16

149 Pedal phalanges Pedal phalanges 1:2, 1:1,2; 111:3; IV:2,4 11:2*; I1I:2*,3*, 1:1, IV: 2,4,3

39 Day 17-18 Day 22-24 Day 13-16 Metatarsal I, pedal Pretibial bone phalanx IV:3

40+ Day 18-26 Day 25-30 Day 13-24 Metatarsal I*, Metatarsal I, pedal Pretibial bone, pedal phalanx phalanx IV:3 metatarsal I, tibiale, IV:3*, pretibial fibulare, hypotarsus bone Asterisk denotes elements that ossify variably with respect to stage.

150 TABLE 4.2. Rank order of element ossification for three anseriform genera. Anas Cairina Somateria Element platyrhynchos moschata mollissima

Skull Basioccipital 12 12 6 Basisphenoid 10 8 2 Ectethmoid C C C Epiotic 21 19 12 Exoccipital 10 12 5 Frontal 6 5 2 Jugal 3 4 2 Lacrimal 2 3 2 Laterosphenoid 20 17 7 Maxilla 2 3 2 Mesethmoid 20 21 11-17 Nasal 2 3 2 Opisthotic 15 14 9 Palatine 2 3-4 2 Parasphenoid ala 10 3-6 6 Parasphenoid lamina 10 3-6 6 Parasphenoid rostrum 8 3-6 2 Parietal 5 6 2 Premaxilla 3 4 2 Prootic 15 13 8 Pterygoid 2 3-4 2 Quadrate 5 8 2 Quadratojugal 1 3 2 Squamosal 1 3 2 Supraoccipital 10 12 5 Trabeculae c C 11-17 Vomer 7 3-6 2.

Angular 1 2 2 Articular 19 15 8 Dentary 3 4 2 Prearticular 11 12 7 Splenial 8 4-5 2 Supra-angular 1 4 2

Basihyal C C C Ceratobranchial 3 4 2 Entoglossal C C C Epibranchial C C C Urohyal c c c Postcranial axial skeleton Caudal centra 18 16 8

151 Caudal neural arches 24 21 12 Cervical centra 9 12 5 Cervical neural arch 14 17 7 Cervical ribs 11 14 6 Dorsal ribs 4 7 2 Laterocaudal process C C C Laterocranial process C C C Pygostyle 21 20 15 Sternal ribs C C 16 Sternum (body) C C 16 Synsacral arch 23 22 11 Synsacral centra 12 12 5 Synsacral transverse processes 22 23 10-12 Thoracic centra 12 12 5 Thoracic neural arch 16 18 7 Uncinate processes C C C

Forelimb Scapula 3 4 2 Coracoid 6 8 2-4 Furcula 1 2 2 Humerus 1 1 1 Radius 1 1 1 Ulna 1 1 1 Radiale C C C Ulnare C C C Metacarpal 11 24 c 12-15 Phalanx 1 6 8 2 Phalanx 2 10-25 7-12 2-11 Metacarpal III 2 3 2 Phalanx 1 8 8 3 Phalanx 2 8 8 4 Phalanx 3 24-25 7-19 10 Metacarpal IV 2 4 2 Phalanx 1 C c 11-15 Hind limb Ilium 6 10 3-5 Ischium 6 5 2 Pubis 7 6 2 Femur 1 1 1 Tibia 1 1 1 Fibula 1 1 1 . Patella C c C Pretibial bone 17 12-14 8 Tarsals C C 13 Metatarsal I 13-17 17-18 8 Phalanx 1 10 9 5 Phalanx 2 10 9-11 2

152 Metatarsal II 1 1 1 Phalanx 1 5 . 7 2 Phalanx 2 6 7 3 Phalanx 3 7 8-11 4 Metatarsal III 1 1 1 Phalanx 1 4 7 2 Phalanx 2 8 10 4 Phalanx 3 10 8 4 Phalanx 4 7 8-11 2 Metatarsal IV 1 1 1 Phalanx 1 4 7 2 Phalanx 2 10 11 5 Phalanx 3 14 19 6 Phalanx 4 10 7-8 5 Phalanx 5 8 8-11 2 If an element is labeled "C", it was unossified in all of the specimens examined. If two numbers are given, these represent the range of ranks over which a variable element can ossify.

153 TABLE 4.3. Shared timing of developmental events in the ossification sequences of selected anseriforms. Anas + Cairing Anas + Somateria Somateria + Cairina psl A [tc, sc] psl D [q, cc, dr, cor, w21, psr A 123 w31, il, 121,122,123,131, 132,134,141,145] 122 A [bs, 134,145] 143 A [Is, po, oo, ar, cac, ta, il D [psr, bs, v, spl, pub] ptb] Is D [po, oo, ar, cac, ptb] ca A [po, oo, ar, cac, ptb] cc D [bs, 133] ta D [po, oo] . cr A po bs A 132 oo A [ar, cac] q A 122 122 D [psr, v, pub] pra A 143 ta A [ar, cac, ptb] spl D mt3 pyg A [sstp, caa, ssa, tar] caa D ssa cr D [tc, sc] 132 A [111,144] caaDw22 123 A 144 Only sets of events showing variation between the three genera are listed. See materials and methods section for more details. This represents one of several equally parsimonious solutions regarding the identity of moving elements and the directionality of their shifts. A, relative acceleration, D, relative delay ar, articular; bs, basisphenoid; ca, cervical neural arch; caa, caudal neural arch; cac, caudal centra; cc, cervical centra; cor, coracoid; cr, cervical ribs; dr, dorsal ribs; il, ilium; 111-145, pedal phalanges, where the first number represents the digit and the second number represents the phalanx, with 1 being the most proximal; Is, laterosphenoid; mtl-4, metatarsals; oo, opisthotic; po, prootic; pra, prearticular; psl, parasphenoid lamina; psr, parasphenoid rostrum; ptb, pretibial bone; pub, pubis; pyg, pygostyle; q, quadrate; sc, synsacral centra; spl, splenial; ssa, synsacral neural arches; sstp, synsacral transverse processes; ta, thoracic neural arches; tar, tarsals; tc, thoracic centra; v, vomer; w21-w41, manual phalanges, where the first number represents the digit and the second number represents the phalanx, with 1 being the most proximal.

154 Bridging text 4. Palaeognaths (Chapter 2), Galliformes (Chapter 3) and Anseriformes (Chapter 4) constitute the three most basal lineages of birds. The vast majority of extant bird species comprise a large clade referred to as Neoaves. Phylogenetic relationships within this group are intensely debated. Within Neoaves, I was able to obtain ossification sequences for several species in the order Charadriiformes (shorebirds). While there is disagreement between morphological and molecular data regarding the interrelationships of Charadriiformes, there is an increasing degree of consensus among molecular studies. This makes Charadriiformes an excellent order on which to test the power of ossification sequence data to resolve phylogenetic relationships at the family level. Here, I use two techniques (event- pairs and PGi) to examine the potential of ossification sequence data to address phylogenetic problems when small numbers of recently diverged taxa are considered. Developmental sequence data is thought to contain phylogenetic information, but has never been applied to the problem of avian systematics. In this chapter, I describe the ossification sequence of the Common Tern (Sterna hirundo), and compare the pattern observed to published descriptions of other Charadriiformes, specifically the Great Skua (Stercorariidae) and various species of gulls (Laridae). Changes in ossification sequence are then used to elucidate the relationship between these three taxa, using both qualitative and systematic approaches. The first analysis of the ossification sequence data does not support a close relationship between Stercorariidae and Laridae, as has been proposed in some morphological analyses; however it was unable to differentiate between a Laridae - Sternidae sister-group relationship or a Sternidae - Stercorariidae sister- group relationship. The second analysis was unable to differentiate between any topology, including a polytomy, for these taxa. These results highlight the potential for use of ossification sequence data in an evolutionary context but caution that analyses are highly dependent on sequence resolution and the taxonomic level of the data set.

155 CHAPTER 5

Ossification Sequence of the Common Tern (Sterna hirundo) and Its Implications for the Interrelationships of the Lari (Aves, Charadriiformes)

(Reference: Maxwell, E.E. and L.B. Harrison. Ossification sequence of the Common Tern (ISterna hirundo) and its implications for the interrelationships of the Lari (Aves: Charadriiformes). Journal of Morphology In Press).

Charadriiformes (shorebirds) is an order of birds deeply nested within Neoaves. It is generally thought to be closely related'to Gruiformes (rails and allies) (Sibley and Ahlquist, 1990; Paton et al., 2003; Livezey and Zusi, 2007), but various authors place Charadriiformes as a sister group to either Columbiformes (pigeons) or to a large and diverse assemblage of other neoavian orders (Livezey and Zusi, 2001; Mayr and Clarke, 2003). Relationships within Charadriiformes are also contentious. Three principal charadriiform clades are recognized - Scolopaci (sandpipers), Lari (gulls), and Charadrii (plovers) (Sibley and Ahlquist, 1990; Christian et al., 1992; Chu, 1995; Ericson et al., 2003; Paton et al., 2003; Thomas et al., 2004; Paton and Baker, 2006) - but the arrangement of these clades, as well as that to which Alcidae (auks) belongs, differs depending on whether morphological or molecular data are examined. The relationships among families within the clades are also debated, although we will focus only on the relationships of three families in this paper: Laridae (gulls), Sternidae (terns) and Stercorariidae (skuas). Morphological analyses find either a sister-group relationship between Stercorariidae and Sternidae (Fig. 5.IB) (Mickevich and Parenti, 1980; Chu, 1998), or a sister group relationship between Stercorariidae and Laridae (Fig. 5.1C) (Chu, 1995). Molecular analyses recover a sister-group relationship between Laridae and Sternidae (Fig. 5.1 A) (Sibley and Ahlquist, 1990; Ericson et al., 2003; Paton et al., 2003; Paton and Baker, 2006; Fain and Houde, 2007). The conflict within the morphological data set, when compared to the relative consensus achieved using molecular data, suggests that new characters are needed to yield improved resolution of the morphological tree. This is especially true since two of the three

156 morphological studies cited rely on reanalysis of the same character set (Strauch, 1978). Using ontogenetic data to answer evolutionary questions provides a powerful tool for ornithologists. Ontogenetic information has been successfully used to address problems of skeletal homology in birds (McGowan, 1984; Zusi and Livezey, 2000). Juvenile features, such as the coloration pattern of natal down, are frequently used as characters in phylogenetic analyses (Livezey, 1991, 1996; Chu, 1998; Livezey, 1998). Similarities in post-hatching growth trajectories have also been used to support particular clades (Cane, 1994), and the number and timing of postnatal ossification centers in squamates has been used to reconstruct phylogeny (Maisano, 2002b). Structures present in pre-adult stages are dependent on the function of the structures in the adult form, but are also subject to unique selective pressures imposed on them during development (Cane, 1993; Badyaev and Martin, 2000; Minelli, 2003). Developmental pathways can evolve, but are not infinitely flexible: similar patterns of covariation of traits within a stage and between adjacent stages indicate that constraints exist limiting potential changes to an ontogenetic trajectory (Badyaev and Martin, 2000). These constraints are useful because by limiting the potential for change, phylogenetic information may be conserved in the pathway. By looking at the relationship between the ontogenetic sequence of a single species to the sequence in other species, the pattern of development itself can become a single phylogenetic character (Mabee, 2000), or can be divided into a set of relatively independent characters (Maisano, 2002b). By comparing the trajectory of change in a single species over developmental time, patterns may be found that provide evidence for the trajectory of change over evolutionary time. In spite of this, ontogenetic data have been very difficult to incorporate into phylogenetic analyses due to their inherently non-independent nature. There is an increasing amount of evidence supporting extensive dissociation of the sequence of ossification from the sequence of chondrification in amniotes (Strong, 1925; Maillard, 1948; Maisano, 2002a; Maxwell, 20Q8; Frobisch, In Press). Reports for birds suggest that the sequence of ossification,

157 while dissociated from the sequence of chondrification, is relatively invariant and is not linked to the mode of development (Starck, 1998). This implies that ossification sequences have the potential to vary, but do not vary infinitely. Detailed studies of ossification in birds are rare, and most have focused on the relationship between skeletal development and precociality (Rogulska, 1962; Starck, 1989, 1993,1996). Comparative avian skeletal development has never been examined in a phylogenetic context. Previous work on charadriiform skeletal development is limited to a description of the chondrogenesis of the carpal region of Sterna wilsonii (Leighton, 1894), two studies examining the ossification sequence of the Black-headed Gull (Larus ridibundus) (Rogulska, 1962; Schumacher and Wolff, 1966a), the ossification sequence of the Mew Gull (L. canus) (Schumacher and Wolff, 1966a), and a very thorough description of the ossification of the Great Skua {Stercorarius skua) (Maillard, 1948). Here, we describe the ossification sequence of the Common Tern {Sterna hirundo), and compare it to other Charadriiformes as well and to the Domestic Chicken {Gallus gallus). We attempt to find phylogenetic markers that can be used to resolve relationships within the Lari.

158 Figure 5.1. Ossification sequence characters mapped on to the phylogenetic hypotheses for the relationships between the charadriiform families Laridae, Sternidae and Stercorariidae. The event-pair characters are from the whole dataset, and the PGi shifts are from the skull dataset. An upwards pointing arrow represents an accelerated event, and a downwards pointing arrow represents a delayed event. A: Molecular hypothesis (Paton and Baker, 2006). B: Morphological hypothesis (Chu, 1998). C: Morphological hypothesis (Chu, 1995). ang, angular; bo, basioccipital; bs, basisphenoid; cb, ceratobranchial; cor, coracoid; d, dentary; dr, dorsal ribs; exo, exoccipital; f, frontal; il, ilium; isch, ischium; 111 to 145, pedal phalanges; lac, lacrimal; mc, metacarpals III and IV; mx, maxilla; n, nasal; pa, parietal; pal, palatine; pmx, premaxilla; pt, pterygoid; pub, pubis; q, quadrate; qj, quadratojugal; sa, supra-angular; sc, synsacral centra; so, supraoccipital; sp, splenial; tc, thoracic centra; verts, vertebral column; w21 to w41, manual phalanges.

159 #• Gallus Event-Pair PGi

Stercorariidae 142,143 -f bo,so,verts,q cb 4 f.lacsp. mc-f n,pt,qj,sa exo 4- bo,q Sternidae pub>Jrl41 q>|rw21 Laridae dr'f lac,sp,cb

## Gallus Event-Pair PGi bo^w21,l23,l4S bc>4-exo,q Laridae bs-f-ischjl pa 4 bo,exo,q cor^pub exo-^ lac Sternidae pa 4-123,134,145 Stercorariidae i44Srtc,sc ang'f mx,qj,pt,pal

Gallus #•*# Event-Pair PGi Sternidae exo4q,l23J34,l45 exo4-bo,q,pa,f sp4Pmx'd,cb Stercorariidae • sp4-cb w21^122 Laridae MATERIALS AND METHODS

Cleared and stained embryos of the Common Tern (Sterna hirundoL.) examined ranged from day 5 to day 20 of incubation in a 22 to 27 day incubation period. A smaller series of Herring Gull embryos (Larus argentatus Pontoppidan, 1763) ranged from day 8 to day 13 of incubation with one older individual of an unknown age. Both series are housed at the American Museum of Natural History (AMNH), and were prepared by W.P. Cane (WPC). For information on collection and preparation, refer to Cane (1994). Embryos were not staged prior to clearing, and so day of incubation is used as a measure of age (Appendix 1). This is only an approximation, however, as length of the incubation period depends on the egg's rank in the clutch: for instance, in L. argentatus, the incubation period of the egg laid last is 27 days, whereas it is 30 days for the first- laid egg (Pierotti and Good, 1994). A similar pattern is observed in S. hirundo (Nisbet, 2002). Embryos of the Domestic Chicken, (Gallus gallus) referred to in this paper were cleared and stained for the presence of bone and cartilage by the first author following a standard protocol (Dingerkus and Uhler, 1977), and are housed in the Redpath Museum. This sequence was supplemented by previously published results (Jollie, 1957; Schumacher and Wolff, 1966b). Ossification sequences were coded for 99 skeletal features, including entire elements and distinct parts of elements if separate ossification centers had previously been reported in other species (Table 5.1). The data were subjected to two analyses: a large-scale event-pair based analysis (Smith, 1997; Velhagen Jr., 1997) and a more restricted, computationally intensive evolutionary analysis of selected skull elements (Harrison and Larsson, In Press). In the first analysis, ossification sequences were entered into the PGi program to convert the sequence to event-pairs (Harrison and Larsson, In Press). This transformed the numerical sequence to a series of comparisons between all pairs of events, reflecting relative timing. Simultaneous events were coded as missing data, rather than as a separate character state as is sometimes the case (Schulmeister and Wheeler, 2004; Jeffery et al., 2005). This was to prevent biasing the data set in favor of taxa with poorly

160 resolved sequences. Polymorphic events were coded as missing with regard to the parts of the sequence affected by the polymorphism (for instance, if the femur ossified either at event 2 or event 4, it was coded as unknown relative to other elements ossifying at events 2, 3 and 4). This approach allowed variable elements to be maintained in the data set, which is important since the timing of ossification appears to be constrained relative to most, but not all, other elements. An estimate of intraspecific variability for the Common Tern was calculated by dividing the number of variable event-pairs by the total number of event-pairs (2v/n(n-l), where v is the number of variable event-pairs and n is the number of elements ossified in the most ossified individual examined). The sequence for Gallus gallus reported by Maxwell (2008) was amalgamated with the sequence published by Schumacher and Wolff (1966b), and event-pairs that differed between the two were coded as unknown. Ossification sequences of LOTUS canus, L. ridibundus, and Stercorarius skua were obtained from the literature (Maillard, 1948; Schumacher and Wolff, 1966a). The three species of the genus Larus were treated as a single taxonomic unit. Variable event-pairs were coded according to the majority (the state shared by two of the three taxa). Although the genus Larus is commonly believed to be polyphyletic, there is no evidence that the family Laridae is not a valid taxonomic unit (Chu, 1998). The event-pairs were assessed via a parsimony algorithm using PAUP* (Swofford, 2002). Gallus was used as an outgroup, and the three charadriiform families formed the ingroup. This yielded a topology that was biased by both the biological (developmental) and logical non-independence of the data. To correct for the latter, variable event-pairs were grouped into compound characters. For example, if the angular was accelerated in sequence relative to the squamosal, maxilla, palatine and pterygoid, traditional parsimony analysis would count this as four characters supporting the node, when in reality only the angular shifted. Biological non-independence (modularity) was more difficult to correct for. Events that were developmentally constrained did not shift relative to each other, and so were not informative in a parsimony analysis. In addition, preliminary tests suggest that ossification sequences do not behave in a modular way (Poe, 2004;

161 Goswami, 2007). The approach taken here is similar to that employed by Maisano (2002b), and yielded a character list that was reanalyzed using parsimony. Noting the above methodological complications, a subset of the data was subjected to another more computationally intensive analysis of sequence heterochrony {sensu Smith, 2001). The PGi method treats the developmental sequence as a single complex character, and identifies sequence heterochrony on a known phylogenetic topology without the need for event-pair coding (Harrison and Larsson, In Press). This sidesteps the logical dependence created with event- pairs, and directly examines evolutionary change in developmental sequences at the level of the movements of events (sequence heterochronies). The method is conceptually similar to that proposed by Schulmeister and Wheeler (2004) and optimizes the ancestral states of the developmental sequence onto the phylogeny by searching within all possible ancestral developmental sequences. PGi uses the Parsimov method (Jeffery et al., 2005) as an edit-cost function and uses heuristics to enable execution on larger data sets. Nonetheless, the size of this data set necessitates a limited analysis. In this case, PGi was used to analyze the cranial ossification of these taxa. The selected data set consisted of the following elements: basioccipital, exoccipital, supraoccipital, basisphenoid, squamosal, parietal, frontal, lacrimal, nasal, premaxilla, maxilla, palatine, pterygoid, vomer, jugal, quadratojugal, quadrate, dentary, supra-angular, angular, splenial and ceratobranchial. While this method was designed to examine the evolution of developmental sequences on a single phylogenetic topology, it can also be used to determine the comparative length of multiple phylogenetic topologies. While computational restrictions precluded a tree search, the three possible topologies (Fig. 5.1), as well as a topology in which the ingroup was collapsed into a polytomy, were examined and the length (number of shifting events / sequence heterochronies) of each cladogram was computed. Unlike the first method, simultaneous data were considered a separate state in this analysis in order to maintain a uniform level of simultaneity in the ancestral reconstructions and the extant sequences. In the PGi analysis, sequence changes of a single rank-order involving an event entering or leaving a block of simultaneous events are treated

162 as putatively artifactual and removed from the described heterochronies. They remain included in the tree length value for comparative purposes. These may constitute legitimate sequence heterochronies, but without further taxonomic sampling this is impossible to determine. The reported heterochronies were identified using a majority rule consensus based on a pseudoreplicate dataset created during computation (for more details, see Harrison and Larsson, In Press).

RESULTS

Skeletal development of Sterna hirundo Skull Day 5 (Fig. 5.2A): The major cartilages in the head are present, and the skull is unossified. The parietotectal cartilage of the nasal capsule extends into the rostral area from the trabecula communis (the elongate cartilage connecting the posterior portion of the orbital cartilage to the nasal capsule). The anterior portion of the orbital cartilage forms as a posteriorly directed spur of the trabecula communis. It is connected to the trabecula anteriorly by a broad pedestal. The posterior portion of the orbital cartilage is present, extending along the posterior margin of the orbit and widening dorsally. It appears connected to the suprapolar cartilage posterodorsal to the latter by a faint chondrification. The identity of this cartilage as the suprapolar cartilage is probable due to its close spatial association with the posterior portion of the orbital cartilage, as in the Domestic Duck (Romanoff, 1960), although due to gaps in sampling, its association with the pila antotica could not be confirmed. Posterior and ventral to the suprapolar cartilage and ventral to the posterior portion of the orbital cartilage, the trabecula communis widens to form the polar cartilage. Ventral to the polar cartilage is the quadrate cartilage, which arises from an oval chondrification and forms a right angle with Meckel's cartilage. Immediately posterior to the quadrate cartilage is the pars canaliculi of the auditory capsule. At this stage, the paired capsules are still widely separated, being located more laterally than posteriorly on the skull. The occipital arch is visible between the auditory capsules and the atlas. The two halves of the arch are also widely separated dorsally.

163 Day 6: The skull is unossified. The right and left auditory capsules and the occipital arches almost touch medially; they have also rotated posteriorly. The hyoid is chondrified. The stylohyal cartilage is elongate, and is beginning to segment into the columella. The two sides of Meckel's cartilage are still not in contact anteriorly. Day 7: The skull remains unossified. The parietotectal cartilage is dorsoventrally flattened and forms the roof of the nasal capsule. The prenasal process is a distinct rostrum, although it does not considerably exceed the parietotectal cartilage in anterior extent. The interorbital septum is chondrifying anterior to posterior. The quadrate cartilage has assumed its triradiate shape. The occipital arch and the pars canaliculi of the auditory capsule are now very closely associated. The right and left sides of Meckel's cartilage contact each other anteromedially. The paraglossal and ceratobranchial cartilages are distinct, and the stylohyal and columellar cartilages have also separated. Day 8: The interorbital septum is fully chondrified. The paraglossum, basi- and urohyal, cerato- and epibranchial are chondrified, although the elements have not yet completely separated from each other. The pars canaliculi are located on the posterior surface of the braincase, but the right and left sides do not contact each other. Day 9 (Fig. 5.2B): Both the posterior orbital cartilage and the canalicular portion of the auditory capsule form broad, plate-like cartilaginous elements at the back of the skull. The planum antorbitale is also especially noticeable. The squamosal is ossifying dorsal to the cotyla quadratica squamosi. Day 10: Ossification has advanced to include the posterior portion of the pterygoid and the anterior portion of the pars choanalis of the palatine. The central portion of the maxilla is ossified. Day 11 (Fig. 5.3A): The premaxilla is ossifying from two centers corresponding to the frontal and maxillary processes. The jugal, premaxillary, and maxillopalatine processes of the maxilla are ossified, but not from separate ossification centers. The nasal is ossifying from two splints corresponding to the maxillary and premaxillary processes; the frontal process ossifies slightly later but does not have a separate center. The orbital process of the lacrimal is ossified,

164 and is tri-radiate in shape with an anteriorly directed process halfway along its length. The jugal bar, including both the jugal and quadratojugal, is ossifying. The frontal is ossifying as a narrow band along the dorsal margin of the orbit, and reaches as far anteriorly as the frontal process of the nasal. This event occurs at the same time as the ossification of the frontal process of the nasal. The ossified portion of the squamosal is triangular in shape and completely surrounds the quadrate articulation. The choanal, maxillary and rostral processes of the palatine are ossified, as is the pars palatina and the body of the pterygoid. The pars palatina is still closely associated with the pterygoid proper and does not approach the palatine. The vomer is present, and the parasphenoid rostrum appears slightly later than the rest of the palate. In the lower jaw, the dentary is forked posteriorly. The supra-angular is already well developed, and completely surrounds the fenestra mandibula caudalis. The angular is ossifying as a thin splint along the ventral margin of the mandible, and the splenial is also present. The ceratobranchial is ossified. Day 12: The supraorbital process of the lacrimal is ossified. The parietal is ossifying as a separate center overlying the auditory capsule. The pterygoid process of the palatine is present, and lies lateral to the pars palatina of the pterygoid. The basisphenoid ossifies as two small, paired elements immediately posterior to the mesethmoid cartilage. The parasphenoid lamina is also beginning to ossify as two dorsoventrally flattened wings on either side of the basisphenoid. Day 13, 14 (Fig. 5.3D): The ossiculum lacrimosuborbitale is visible. It is a triangular element, not in articulation with the ectethmoid. The parasphenoid alae have begun to ossify. They are located dorsal to the parasphenoid lamina and are anteroposteriorly flattened. The basioccipital is ossifying as two parallel linear centers, oriented anteroposteriorly. There is no change in the number of ossified elements of the skull between days 13 and 14. Day 15 (Fig. 5.3B): The otic process of the quadrate is ossified. Day 16, early day 17: The posterior frontal contacts the squamosal, and the anterior parietal is also in contact with the squamosal. The supraoccipital is ossifying from a single medial center. The exoccipitals ossify on either side of the basioccipital; the paired centers of the latter have fused. The prearticular is present

165 by this point, although it is possible that it was present earlier and was not noted due to its medial position in the lower jaw. Day 17: The lagenar otoliths are calcified, and lie just posterior to the ossified portion of the parasphenoid lamina. The prootic is ossifying as two parallel sheets of bone oriented perpendicular to the long axis of the skull in dorsal view. Day 18 (Fig. 5.3C): Ossification is advancing up the orbital process of the quadrate. There is extensive overlap between the palatine process of the pterygoid and the palatine, with the former lying medial and parallel to the latter. The laterosphenoid is ossifying from its ventromedial corner. A second otolith is present, posterior to the double laminated ossification center of the prootic. By day 20, a second center of ossification has developed medial to the first; this may represent the opisthotic. Day 19, 20 (Fig. 5.3E): The anterior mesethmoid is ossifying. The laterosphenoid has developed a second center of ossification on its ventrolateral corner.

166 Figure 5.2. A: Lateral view of a. Sterna hirundo embryo on the fifth day of incubation (WPC 255). Scale bar equals 2mm. B: Lateral view of the chondrocranium of a S. hirundo embryo on the ninth day of incubation (WPC 252); squamosal ossified but not shown. Scale bar equals 5mm. Grey shaded regions represent cartilage, ac, auditory capsule; aoc, anterior portion of the orbital cartilage; ent, entoglossal cartilage; io, interorbital septum; M, Meckel's cartilage; pao, pila antotica; pla, planum antorbitale; pnp, prenasal process; poc, posterior part of the orbital cartilage; pit, processus tectalis; ptc, parietotectal cartilage; q, quadrate; spc, suprapolar cartilage; tr, trabecula communis.

167

Figure 5.3. Lateral view of the skull of Sterna hirundo embryos. A: Day 11 of incubation (WPC 251). B: Day 15.5 of incubation (WPC 275). C: Day 18 of incubation (WPC 301). D: Dorsal view of the braincase, day 14 of incubation (WPC 276). E: Dorsal view of the braincase, day 20 of incubation (WPC 310). Grey shaded regions represent cartilage; black regions represent ossified tissue. The density of stippling reflects the relative degree of ossification. Scale bar equals 5 mm for parts A-C, E; scale bar equals 3 mm for part D. a, angular; at, atlas; bo, basioccipital; bs, basisphenoid; d, dentary; exo, exoccipital; f, frontal; j, jugal; lac; lacrimal; Is, ossiculum lacrimosuborbitale; mx, maxilla; n, nasal; oto, otoliths; pa, parietal; pi, palatine; pmx, premaxilla; po, prootic; psl, parasphenoid lamina; psr, parasphenoid rostrum; pt, pterygoid; q, quadrate; qj, quadratojugal; sa, supra-angular; so, supraoccipital; sq, squamosal.

168

Postcranial axial skeleton Day 5 (Fig^5.2A): The postcranial axial skeleton is entirely cartilaginous. Approximately 14 cervical, 13 dorsal, and 17-18 post-acetabular vertebrae are present (sacral vertebrae, except for the one immediately above the hind limb, are counted as either part of the thoracic or caudal series as the pelvic girdle has not chondrified). Eight of the thoracic vertebrae bear dorsal ribs; ribs 4 to 7 have associated sternal segments. Day 6: The postcranial axial skeleton remains cartilaginous. Some individuals possess nine dorsal ribs, although eight is more usual. The second to seventh ribs are associated with sternal segments. There are 17 to 18 post- acetabular vertebrae, 13 of which are post-sacral. Six to nine of these are associated with haemal arches; the number is also intraspecifically variable. Day 8: Cervical ribs are distinct. Day 11: Cartilaginous uncinate processes are present on the first five true ribs. There are 10 free caudal vertebrae, five of which have free haemal arches, and the pygostyle. In slightly older embryos, the number of free caudal vertebrae decreases to eight. Day 12, 13: All dorsal ribs are ossifying. The degree of ossification decreases posteriorly, suggesting an anterior to posterior sequence of ossification. Day 14, 15: There are four cartilaginous haemal arches, all of which are now fused to the caudal vertebrae. A sixth cartilaginous uncinate process is variably present. From this point on, chronological age no longer closely corresponds with the degree of ossification of the axial skeleton. For that reason, the sequence of ossification is described without reference to day of incubation, but rather in reference to individual specimens. All cervical centra remain cartilaginous, but the thoracic centra are ossifying from the first complete rib posteriorly (WPC 248). The number of ossifying synsacral centra is variable, with the sixth ossifying after the cervical centra in some individuals (WPC 275). The centra begin ossification from a pair of ventral centers, one on either side of the midline. The size of these centers decreases from anterior to posterior in the thoracic series.

169 The cervical centra all have ossification centers, as do the thoracic centra (WPC 275). Following this, the paired ventral ossification centers in the cervical, thoracic, and first synsacral centra merge. A dorsal ossification center is also present. The remaining six ossifying synsacral centra still retain the paired ventral centers (WPC 294). Nine synsacral centra are ossifying; the first five have a dorsal center, while the remaining four have only the paired ventral centers (WPC 256). The first four cervical centra are completely ossified, but the more posterior ones retain dorsal and ventral ossification centers. The cervical arches are ossifying; the arches of the 4th and 5th cervical vertebrae are most complete, and the degree of completeness decreases both anterior and posterior to this point. Ossification is greatest at the anterior edge of the arch. Cervical ribs 4-9 are ossified, although 4-5 may be delayed in some individuals until after complete ossification of the synsacral centra (WPC 302, 310). All thoracic centra are ossified. The first and second thoracic vertebrae have ossification centers in their transverse processes. Ossification of the transverse processes may also be delayed relative to the cervical centra, cervical arches, and free caudal centra (WPC 310, 313). Twelve synsacral centra are ossified; the first two are complete and the 10 more posterior have paired ventral ossification centers (WPC 306). The cervical and thoracic centra are completely ossified, although the posterior cervical centra retain dorsal and ventral ossification centers until after the beginning of arch closure in some individuals. Cervical arches five and six are the most ossified. The first three to four synsacral centra are completely ossified, and the rest have paired ventral ossification centers. The first free caudal centrum is also ossified via paired ventral ossification centers (WPC 310, 313). The cervical arches near the middle of the series are almost completely closed; the ones anterior and posterior to them less so. Cervical ribs 3-9 are ossified. The first two to four caudal centra have paired ventral ossification centers (WPC 302). The synsacral centra are completely ossified. Cervical arches 5-7 are the most complete, but all exhibit some degree of ossification. The thoracic transverse

170 processes are ossifying along their posterior surfaces from the first to the seventh vertebrae (WPC 301, 302). Cervical arches 5-12 are complete, as are all cervical ribs, although cervical ribs one to three are occasionally delayed. Seven thoracic vertebrae have extensively ossified transverse processes and arches; the eighth has only a small center of ossification on the posterior surface of its transverse process. The transverse processes of the first five synsacral vertebrae are ossifying, as is the transverse process of the first caudal vertebra (WPC 254). The latter may be delayed (WPC 257). Most cervical arches are completely ossified; the dorsal edges of the arch are not medially joined only in the atlas, axis, and the most posterior two cervical vertebrae. The transverse processes of all the thoracic vertebrae except the most posterior are ossifying. The transverse processes of the first eight synsacral vertebrae are ossifying (WPC 257). The arches of the first two caudal vertebrae are beginning to ossify, and the first free caudal transverse process is also ossified. The cervical arches are all completely ossified except for the atlas. The thoracic arches are also all complete except for the most posterior one. The synsacral arches all have ossification centers around the proximal end of the transverse processes. The costal process of the acetabular vertebra is ossified (WPC 247).

Appendicular skeleton: forelimb The development of the carpal region in Sterna hirundo is not described in detail here, as it is not significantly different from the description presented by Leighton(1894). Day 5 (Fig. 5.2A): There is no ossification present in the forelimb. The coracoid and scapula are chondrifying, but only the anteriormost part of the latter is present. The humerus, radius, ulna and digit IV are chondrifying. Digit IV appears continuous with the ulna due to diffuse chondrification in the carpal region and the small size of the embryo. Observations of sectioned material clarify that the ulna, carpals and metacarpals do initiate chondrification from

171 independent centers (Leighton, 1894). The metacarpal cannot be differentiated from its phalanges. Day 6: Ossification has not begun in the forelimb. The sternal plates are chondrified. Digits III and IV are present, and are distinct from the zeugopodium. The ulnare is located between the ulna and the fourth digit. Day 7: The sternal plates are cartilaginous but unfused. The ulnare and digit IV have continued to shift posteriorly, and are no longer in line with the ulna. Digit V is a small cartilage visible medially. There is a large cartilaginous mass attached to the distal end of the radius; this is interpreted as the radiale complex. Digit II has also chondrified, and is separated from the radiale. The metacarpals are now separated from the phalangeal precursors. Following this, the radius and ulna begin ossifying from paired centers, which form parallel to each other mid-shaft. Those of the radius are slightly less advanced. The manual digits have the full complement of phalanges. The carpal region consists of the chondrified radiale complex, pisiform, ulnare and carpals distal to the ulnare, with metacarpal V located posteriorly. Day 8 (Fig. 5.4A): The furcula is ossified, and the humerus has also developed a faint ossification center. There is a cartilaginous central carpal mass continuous with the ulna and extending to the base of metacarpal II. The ulnare is fused to the pisiform element on its ventral surface, and metacarpal V is located distal to the pisiform. Metacarpal V is a only visible in medial view and is extremely closely associated with metacarpal IV. There is some extra diffuse alcian staining distal to digit IV, suggesting the possible presence of a second phalanx. Day 9, JO: Metacarpals III and IV are ossifying, with metacarpal IV lagging behind metacarpal III. The sternal plates are not completely fused; the apex of the carina is in the process of chondrifying. Day 11: The dorsal supracondylar process forms as an independent chondrification near the distal end of the anterolateral humerus. Metacarpal II is now fused to distal carpal II and the central carpal mass. This is followed by the ossification of the proximal phalanx of digit III, and later by the ossification of the distal phalanx of digit III, the proximal phalanx of digit II, and the scapula.

172 Day 12: The central carpal complex (including metacarpal II), the radiale and the ulnare complexes are the only free cartilaginous elements in the mesopodium. Metacarpal V is still visible. The proximal phalanx of digit III has developed a semicircular cartilaginous flange along its posterior edge. Day 13,14: The coracoid is ossifying. The dorsal supracondylar process has developed a cartilaginous connection to the dorsal condyle of the humerus. Day 15 (Fig. 5.4B): The anterior edge of the phalanx 1 digit III flange is beginning to ossify. Day 18: The ossification center at the anterior edge of the phalanx 1 digit III flange has extended via a thin pillar and expanded over the posterior edge, so that there is a gap between the two ossified regions of the phalanx. Later, the two regions fuse (Fig. 5.4C). The dorsal supracondylar process is fused to the humerus.

173 Figure 5.4. Lateral view of the forelimbs of Sterna hirundo embryos. A: Day 8.5 (WPC 291). B: Day 15 (WPC 249). C: Day 20 (WPC 247). Grey shaded regions represent cartilage; black regions represent ossified tissue. The density of stippling reflects the relative degree of ossification. Scale bar equals 2 mm for part A; 5 mm for parts B and C. c2+3, carpals; dsp, dorsal supracondylar process; H, humerus; R, radius; re, radiale complex; U, ulna; ue, ulnare.

174 - I

A Appendicular skeleton: hind limb Day 5 (Fig. 5.2A): The femur, tibia, fibula, and digits III-V are chondrified in the pes. The girdle elements are not visible. The fibulare is a large chondrification between digit IV and the fibula. Day 6: There is no ossification present in the hind limb. The posterior wing of the ilium is chondrifying, as is the ischium. The ilium lies parallel to the vertebral column, while the ischium points posteroventrally at a 45° angle. The femur, tibia and fibula are separated, with joints forming between the elements. The fibula is equal in length to the tibia and has not yet detached from the tarsal region. The fibulare is well chondrified relative to the rest of the tarsal region. Pedal digits II-V are chondrified, but their phalanges are not yet separate. Digits III and IV are the most chondrified; digit I is not clearly formed. Day 7: The pubis is chondrified, and is oriented ventrally. The anterior wing of the ilium is still not well formed. The tibia and the fibula remain equal in length, and the latter retains its association with the tarsal region. The intermedium is situated near the lateral condyle of the tibia, medial to the fibulare. The distal tibial centrale is situated on the medial condyle of the tibia. Distal tarsals IV and V appear continuous with metatarsal V; distal tarsals II and III are proximal to metatarsal III. All five pedal digits are now distinct, with digit I located halfway down the tarsometatarsus. The proximal phalanges are separated from the metatarsals. Day 8 (Fig. 5.5A): The anterior wing of the ilium is now chondrified. The dorsomedial region closest to the vertebral column is the most advanced. The ischium and pubis remain cartilaginous, and retain their ventral orientation. The tibia is ossifying, followed by the fibula and later the femur. The ossification of the latter may be variably delayed until the ninth day of incubation. The fibula has become detached from the tarsal region. The intermedium is roughly triangular in shape and is associated with the tibia, a morphology and position consistent with the ascending process of the astragalus. The fibulare as well as the proximal and distal tibial centrales are still distinct cartilages. The distal tarsals are fused. The hallux is halfway down the tarsometatarsus, and has partially rotated posteriorly.

175 Metatarsals II to IV have elongated, and the two most proximal phalanges are distinct. Day 9: The cartilaginous ischium and pubis have rotated posteriorly and have almost reached their adult orientation. The pubis has thinned considerably. Metatarsal III is showing faint signs of cartilage degeneration, and has shifted posteriorly relative to metatarsals II and IV. All phalanges are chondrified. The fibulare and distal tibial centrale are distinct cartilages, and the proximal central tarsal has merged with the intermedium so that only a single medial triangular element is present. Day 10: Metatarsals II to IV are ossifying. The hallux is fully reversed. Day 11, 12: The proximal phalanges of digits II-IV is ossified, as is phalanx 2 digits III (variable) and IV (variable) (Fig. 5.5B). Later, the pubis, phalanx 2 digits II-IV and phalanx 4 digit III ossify. Day 13, 14: The anterior wing of the ilium ossifies from two centers, one on the dorsal margin, the other directly below it on the ventral margin. The posterior wing remains cartilaginous. The ischium is also ossifying. Either pedal phalanx 3 digit IV, or phalanx 4 digit III is ossified. This indicates polymorphism in the order of ossification of pedal phalanges. Later, all phalanges except those of the hallux and phalanx 4 digit IV are ossified. Day 15: Digit 1 shows no signs of ossification, but all other phalanges are ossified. Day 16: Phalanx 2 digit I is ossified, followed by phalanx 1 and the pretibial bone. Day 17: Metatarsal I is ossifying, beginning from its proximal end. Day 18: The posterior wing of the ilium is ossifying via the posterior extension of the mineralized area. The anterior wing of the ilium is completely ossified (Fig. 5.5C).

176 Figure 5.5. Lateral view of the hind limb and pelvic girdle of Sterna hirundo embryos. A: Day 8.5 (WPC 291). B: Day 11 (WPC 251). C: Day 20 (WPC 247). Grey shaded regions represent cartilage; black regions represent ossified tissue. The density of stippling reflects the relative degree of ossification. Scale bar equals 2 mm for parts A and B, and 5 mm for part C. dt, distal tarsals; F, femur; fe, fibulare; Fib, fibula; il, ilium; isch, ischium; ptb, pretibial bone; pub, pubis; te, tibiale; Tib, tibia.

177 te. fe .™ ^^i—isch Intraspecific variability in ossification sequence data The oldest Common Tern examined had 74 ossified elements, resulting in a total of 2701 event-pairs. One hundred fifteen of these event-pairs were intraspecifically variable. This leads to a variability estimate of 4.3%.

Phylogenetic analyses Parsimony analysis of event-pairs. Forty of the 4851 event-pairs evaluated were unambiguous and phylogenetically informative for the taxa in question. Unambiguously optimized event-pairs were amalgamated into compound characters. These characters support two of three possible topologies. Sterna hirundo and Larus spp. share (Fig. 5.1 A): 1) a delay in dorsal rib ossification relative to the lacrimal, splenial and ceratobranchial, 2) an acceleration of metacarpals III and IV relative to the nasal, pterygoid, quadratojugal and supra-angular, 3) a delay in the ossification of the pubis relative to pedal phalanx 1 digit IV, 4) a delay in the ossification of the quadrate relative to manual phalanx 1 digit II, 5) a delay in the ossification of the basioccipital, supraoccipital, vertebral column and quadrate with regard to pedal phalanges 2 and 3 digit IV. Sterna hirundo and Stercorarius skua share (Fig. 5.1 B): 1) a delay in the ossification of the angular relative to the maxilla, quadratojugal, pterygoid and palatine, 2) an early ossification of the basioccipital relative to manual phalanx 1 digit II and pedal phalanges 3 digit II and 5 digit IV, 3) an acceleration in the ossification of the basisphenoid relative to the ilium and ischium, 4) an acceleration in the thoracic and sacral centra relative to pedal phalanx 4 digit IV, 5) an acceleration of the coracoid relative to the pubis, 6) an acceleration of the parietal relative to the ungual phalanges of digits II, III and IV. Stercorarius skua and Larus spp. share (Fig. 5.1C):

178 1) an acceleration in the ossification of the exoccipital relative to the quadrate as well as to the ungual phalanges of digits II, III and IV, 2) an acceleration in the ossification of the splenial relative to the ceratobranchial, 3) a delay in the ossification of manual phalanx 1 digit II relative to pedal phalanx 2 digit II. Based on this method of character optimization, the most parsimonious topology is the sister-group relationship between Sternidae and Stercorariidae, to the exclusion of Laridae. The sister-group relationship between Sternidae and Laridae is supported by one less shift. The similar levels of support received by these two hypotheses suggest that ossification sequence data cannot distinguish between them. PGi Analysis. The PGi analysis returned cladograms based on the topologies in Figure 5.1 with the following lengths, respectively: 34, 34, 37 and a length of 33 for a polytomous ingroup. Given the heuristic nature of PGi, such small differences in cladogram length are probably not significant, with the exception of the third topology. Thus, the PGi analysis does not support any one topology better than the unresolved topology. While the sequences support no topology better than an unresolved relationship, the Stercorarius-Larus topology is particularly poorly supported. Nevertheless, in order to compare with the event- pair based method, the following majority-rule consensus sequence heterochronies were identified to support the three possible informative topologies (support levels not shown). Sterna hirundo and Larus spp. share (Fig. 5.1 A): 1) an acceleration of the ceratobranchial relative to the frontal, lacrimal and splenial, 2) a deceleration of the exoccipital relative to the basioccipital and quadrate. Sterna hirundo and Stercorarius skua share (Fig. 5.IB): 1) an acceleration of the basioccipital relative to the exoccipital and quadrate, 2) an acceleration of the parietal relative to the basioccipital, quadrate, and exoccipital, 3) a deceleration of the exoccipital relative to the lacrimal. Stercorarius skua and Larus spp. share (Fig. 5.1C):

179 1) an acceleration of the exoccipital relative to the basioccipital, parietal, frontal and quadrate, 2) an acceleration of the splenial relative to the premaxilla, dentary and ceratobranchial.

DISCUSSION

Skull The ossification of the skull in Sterna hirundo is generally similar to that of LOTUS argentatus, with a few exceptions. For example, while the squamosal is the first element in the skull of S. hirundo to ossify, and among the first to ossify in Stercorarius skua (Maillard, 1948), it is preceded by the angular in L. argentatus. The angular and various palatal elements (maxilla, palatine, pterygoid, quadratojugal, and supra-angular) ossify prior to the squamosal in L. canus and L. ridibundus. In S. hirundo, the angular is delayed relative to the maxilla, pterygoid and palatine. The nasal ossifies from two separate centers corresponding to the maxillary and premaxillary processes in Sterna hirundo, Larus argentatus and Stercorarius skua (Maillard, 1948). This is in contrast to the chicken, as well as to the Double-crested Cormorant (Phalacrocorax auritus), where the nasal ossifies from a single ossification center at the junction of the three processes (pers. obs.). Two nasal ossification centers were reported in the Red-faced Mousebird (Colius indicus) (Goldschmid, 1972), which has holorhinal nares. These centers correspond to the frontal and maxillary processes, unlike the state in Charadriiformes. The unique pattern of ossification of the nasal observed in shorebirds might be attributable to the schizorhinal narial morphology shared by many charadriiform birds, including the Lari. This is also associated with an extremely kinetic skull (Zusi, 1984). It is difficult to say which factor is responsible for the observed pattern of ossification: the development of a rhynchokinetic skull, adult morphology or by evolutionary inheritance of this ontogenetic feature.

180 There is a delay in the ossification of the lacrimal in Larus argentatus relative to the state in Sterna, with its ossification lagging behind the frontal. This delay was not observed in the other two species of Larus, Stercorarius, or Gallus gallus (Maillard, 1948; Jollie, 1957; Schumacher and Wolff, 1966a), and so might be unique to L. argentatus. The ossification of the basioccipital is delayed relative to the exoccipitals and quadrate in Larus argentatus. This is most similar to the state in Stercorarius skua, L. canus, L. ridibundus, and Gallus gallus, where the ossification of the basioccipital follows the ossification of the exoccipitals (Maillard, 1948; Jollie, 1957; Schumacher and Wolff, 1966a, b). In Sterna hirundo, the basioccipital ossifies prior to both of these elements. The relationship of the quadrate to the exoccipital is highly variable, including reports of intraspecific sequence variability in G. gallus (Jollie, 1957; Schumacher and Wolff, 1966b). These findings suggest that the sequence of ossification of these three elements is highly variable in birds, specifically the relationship of the exoccipital to the quadrate, and may contain little phylogenetic information relevant to more ancient divergences. Only a single, unpaired ossification center was detected in the supraoccipital of Sterna hirundo and Stercorarius skua, compared with the two observed in the chicken (Jollie, 1957). In most taxa surveyed, only a single ossification center has been reported for this element (Jollie, 1957). Although a separate ossification center of the laterosphenoid has been described as the orbitosphenoid, the two centers described for. the Common Tern do not appear to correspond to two separate elements. The laterosphenoid has been described as ossifying from the dorsomedial edge of the cartilage, rather than from the ventrolateral edge as observed here (Baumel and Witmer, 1993). Also, as previously discussed with regard to the nasal, the occurrence of several ossification centers resulting in a single element is not unheard of.

Postcranial axial skeleton Comparative data on the ossification of the postcranial axial skeleton is limited for Charadriiformes, as the Larus argentatus embryos examined in this

181 study were too young to show vertebral ossification. It is not unique to Sterna hirundo that the ossification of the cervical vertebrae is delayed relative to the thoracic vertebrae; Schumacher and Wolff (1966a) also identified this pattern in L. canus. The two series begin ossification independently in Gallus gallus, as well as in Stercorarius skua (Maillard, 1948). In the chicken, the cervical centra begin ossification before the thoracic centra. The initial thoracic vertebral ossification center in the chicken is located in the most posterior cervical centrum, rather than at the first thoracic vertebra (Maxwell, 2008). Although the sternal ribs do not ossify prior to hatching in Lari, they ossify immediately prior to hatching in the Barred Buttonquail, Turnix suscitator (Starck, 1989). This does not reflect the degree of precociality, as Gallus gallus ossifies its sternal ribs even before the ossification of the thoracic vertebrae, and other precocial species such as Anseriformes do not ossify the sternal ribs prior to hatching (Chapter 4; Starck, 1998). The laterocaudal processes of the sternum have been demonstrated to take their origin from sternal ribs in at least one galliform (Starck, 1993), and so it is not surprising that their ossification is accelerated relative to the rest of the sternum in G. gallus, in keeping with their embryological identity as sternal ribs. Although the uncinate processes in G. gallus ossify prior to hatching, they do not do so in T. suscitator even though it is also considered to be precocial (Starck, 1989). A similar late ossification of uncinate processes is found in Anseriformes (Chapter 4). Thus, the cartilaginous sternal ribs and uncinate processes may be phylogenetically relevant, and may not reflect the semi-precocial classification of Charadriiformes.

Forelimb The radius and ulna are not reported as ossifying before the humerus in most birds (Starck, 1993), although this pattern was reported in the pigeon (Schinz and Zangerl, 1937). This does not necessarily reflect a close relationship between Columbiformes and Charadriiformes, however, as the three elements ossify simultaneously in several avian species (Maillard, 1948; Rogulska, 1962; Nakane and Tsudzuki, 1999). Increased sampling intensity may show this pattern to be more common in birds than suspected. Alternatively, the reversal of this

182 sequence may be autapomorphic for Sterna hirundo. Ossification of the radius and the ulna from paired centers has also not been noted for other species, and may also be unique to the Common Tern. In Sterna hirundo, the ossification of the scapula follows that of metacarpals III and IV, as well as some of the manual phalanges. This differs greatly from what is observed in Stercorarius skua, where the ossification of the scapula precedes all of the wing elements distal to the zeugopodium (Maillard, 1948). In Larus spp. and Turnix suscitator, resolution was insufficient to place the timing of scapular ossification relative to metacarpal ossification, but the scapula ossified prior to any manual phalanges (Rogulska, 1962; Schumacher and Wolff, 1966a; Starck, 1989). This might be caused by different growth trajectories in the two groups, as gulls and skuas group together on the basis of wing length, to the exclusion of terns (Schnell, 1970). Delayed ossification has been linked to more rapid growth rate at the scale of the whole organism (Arendt and Wilson, 2000); this may also apply at a modular scale. The dorsal supracondylar process has been suggested to be a synapomorphy of Charadriiformes, with the exception of Jacanidae (Strauch, 1978). The development of this process as a chondrification distinct from the humerus, observed in both Sterna hirundo and Larus argentatus suggests that the dorsal supracondylar process is an intratendinous ossification of M. extensor carpi radialis. Its development is clearly distinct ontogenetically from the dorsal epicondylar process, and so this process may well form a good character uniting Charadriiformes (contra Rotthowe and Starck, 1998). The dorsal supracondylar process has arisen independently several times within birds (Baumel and Witmer, 1993) and so a definitive interpretation of its phylogenetic significance cannot be made without a functional study of M. extensor carpi radialis. In all charadriiform species examined, the manual phalanges ossify in a proximal to distal sequence (Maillard, 1948; Schumacher and Wolff, 1966a).

Hind limb The pubis ossifies before the ilium and the ischium in both Sterna hirundo, Larus argentatus, and Stercorarius skua (Maillard, 1948); this differs from L.

183 ridibundus and L. canus where the ilium and ischium both ossify prior to the pubis (Schumacher and Wolff, 1966a). As in the forelimb, the zeugopodium ossifies before the stylopodium in Sterna hirundo. This was also reported for Stercorarius skua, Larus canus and L. ridibundus (Maillard, 1948; Schumacher and Wolff, 1966a). Metatarsal II ossifies noticeably later than metatarsals III and IV in L. argentatus and S. skua, although these events could not be differentiated in S. hirundo. The pedal phalanges follow a roughly proximal to distal sequence of ossification in S. hirundo, but some intraspecific variability is observed, indicating that ossification is not constrained to this pattern. The exception to the proximal-to-distal pattern is phalanx 3 digit IV, which ossifies after the terminal phalanx of digit IV, and in this way S. hirundo resembles L. ridibundus (Schumacher and Wolff, 1966a). L. canus does not adhere to such a strict proximal to distal sequence, with the third phalanx of digit III ossifying after the terminal phalanx of that digit, and phalanges 3 and 4 ossifying after the terminal phalanx of digit IV. Ossification of digit I is considerably delayed; this includes the ossification of metatarsal I. The order of ossification of pedal digit I in Sterna hirundo, with the distal phalanx ossifying first followed by the proximal phalanx, is shared with Larus canus and Stercorarius skua; the proximal phalanx in the latter taxon does not ossify prior to hatching. L. ridibundus ossifies digit I in a proximal to distal sequence. Metatarsal I in 5". skua ossifies prior to its proximal phalanx (Maillard, 1948), but the sequence is reversed in S. hirundo, L. ridibundus and L. canus. This delay in the ossification of digit I is not seen in Gallus gallus, and its cause is unknown.

Potential factors affecting ossification sequence Previous work on bird skeletal development has suggested that there are no differences in the timing of appearance of most ossification centers in relation to the position of the taxon on the altricial - precocial spectrum (Starck, 1993). Even if small, as yet undetected differences existed, all of the ingroup taxa examined in this study are semi-precocial (Starck and Ricklefs, 1998; Nisbet, 2002). One would therefore not expect any one family to be pulled to a basal

184 position due to characters it shares with the outgroup that are associated with precociality. The number and location of nasal ossification centers raises questions regarding the influence of function and adult morphology on skeletal development. For instance, the action of muscles on ossifying elements has been shown to be significant (Hall, 2001). Shape changes in the human skull result in extra ossifications, and the number and location of those ossifications also depends on shape (White, 1996), suggesting that ossification pattern is not independent of adult morphology. It is therefore very difficult to determine whether the nasal has two ossification centers in charadriiforms because of a difference in morphology (schizorhinal nares), because of the development of muscles, influencing movement around the nasal - frontal hinge, or because of a historical pattern of ossification. A similar situation is encountered in the ossification of the maxilla in the colubrid snake Elaphe obsoleta. Multiple ossification centers are visible, and this has been postulated to be due to the posterior extension of the maxilla in snakes (Haluska and Alberch, 1983). The presence of multiple ossification centers has also been attributed to the formation of a hinged joint in the colubrid cranium (Richman et al., 2006): again, the effects of adult morphology and kinesis cannot be positively disentangled. The presence of multiple ossification centers in the nasal of mousebirds suggests that in some cases, a functional explanation does not make sense when taxon sampling is increased. This makes it difficult to interpret changes in number and position of ossification centers. An examination of the morphological synapomorphies used to unite Stercorarius skua and Sterna hirundo, and to define Larus (Chu, 1998) could not be used to explain the pattern observed; in fact changes in adult morphology could not explain any of the changes in ossification sequence or number of centers observed within the Lari. While it is popular to explain some developmental changes in anamniotes, as well as in mammals in terms of their potential adaptive value (Adriaens and Verraes, 1998; Mabee et al., 2000; Sanchez-Villagra, 2002; Prochel, 2006), this has not been tested in birds and there is little evidence here to support it.

185 Neoteny has been proposed as a potential mechanism to explain large body size and 'juvenile'-type plumage in Stercorarius skua (Anderssbn, 1999). Only weak evidence was found to support this during the embryonic period. For instance, the proximal phalanx of pedal digit I fails to ossify prior to hatching, although it does so in both Larus spp. and Sterna hirundo. In order to use ossification sequence data to examine heterochronies in development, it is necessary to stage the embryos using a normal table. This is because heterochronic shifts involving the entire animal, such as neoteny, will not result in any ossification sequence changes, only relative changes in timing. Staging permits the direct comparison of the timing of developmental events. Even taking this into account, a paedomorphic growth trajectory in the Japanese Quail did not result in any major differences in the number of ossifications at hatching when compared to other Galliformes (Starck and Sutter, 2000; Maxwell, 2008).

Intraspecific variability Intraspecific variability in the ossification sequence of the Common Tern represented approximately 4.3% of event-pairs. This number may be influenced by sample size (Bininda-Emonds et al., 2003), but intraspecific variability estimates of 10% or less are not unusual for either traditional morphological data sets or ossification sequence data sets (Maisano, 2002b; Maxwell and Larsson, 2007; Maxwell, 2008). The presence of variability in the ossification sequence of Sterna hirundo affects the two phylogenetic analyses in different ways. The event-pairing analysis used only unambiguous characters, and since variable event-pairs were coded as unknown, their presence decreased the number of phylogenetically informative characters. The PGi method cannot accommodate sequence polymorphism, and so any sequence variability decreases the accuracy of this method in reconstructing sequence changes. In our analysis of selected skull bones, none of the elements chosen exhibited sequence variability, and so polymorphism had no effect on the results.

Phylogenetic analysis

186 Although the phylogenetic content of ossification sequences is far from certain, preliminary studies such as this one are necessary in order to assess their potential utility for systematic purposes. Previous work on fish (Mabee et al., 2000) and on postnatal ossification in squamates (Maisano, 2002b) have suggested that phylogenetic signal is present, although the direct analysis of sequence data remains problematic. The first approach used in this paper relies on very few taxa and characters in order to be able to visualize the resulting shifts in sequence and manually correct for logically non-independent data;, this precludes the use of ossification sequence data in large-scale analyses at present. This analysis was plagued by poor resolution, with the ossification sequence data supporting two of the three possible topologies. The support for the Sternidae - Stercorariidae relationship was slightly, though probably not significantly, higher. The only topology that was ruled out was that of a sister-group relationship between Laridae and Stercorariidae. Causes of ossification sequence change are poorly understood, especially in birds where an adaptive explanation does not immediately present itself. Thus the observed shifts may be the result of a variety of factors, including neutral sequence evolution. The second approach was more methodologically robust, but computationally intensive. While completely correcting for logical dependence by avoiding the use of event-pair coding, PGi could not be applied to the full-size dataset because of the computational burden required. When compared to the results from the first analysis, PGi is more sensitive and identifies relatively more sequence heterochronies in the subset of the total ossification sequence data set. This is likely because the first method is based solely upon the unambiguous optimizations of event-pair characters where the identified shifts are large and easily detected. While PGi identifies relatively more sequence heterochronies than the first analysis, like the first analysis it is unable to differentiate between the proposed topologies other than to rule out a Stercorarius-Larus grouping (Fig. 5.1C). From this, we conclude that the cranial ossification patterns of these taxa, at least at this taxonomic level, contain limited phylogenetic signal. The addition of further taxa, particularly the skimmers (Rynchopidae), which are thought to be closely related to both Sterna and Larus (Fain and Houde, 2007), might better

187 resolve the phylogenetic tree. This conjecture is supported by the lack of any identified sequence heterochronies in the first analysis involving cranial elements shifting relative to each other, save two. PGi recovered one of the same sequence heterochronies as the first analysis; the acceleration of ossification of the splenial (Fig. 5.1C). The second v^as not recovered due to differences in both ancestral sequence optimization and the treatment of simultaneous data (Fig. 5. IB). A thorough investigation of multiple maximally-computable sized subsets from this dataset may be required to establish if any phylogenetic signal is present in the complete sequences. Additional difficulties for both methods of analysis originate from weighting certain types of shifts. For instance, should shifts in ossification sequence within a "module" (term used loosely), in this case the membrane bones of the cranium, be weighted more heavily than those occurring between relatively disconnected elements, for instance, a shift in the basioccipital relative to the pelvic girdle? Based on an absence of a priori information, all changes were weighted equally in this analysis. Both may be informative, the first because a shift within a module represents a fundamental change in the way that structure develops, the second because it reflects a change in developmental timing of one region of the body relative to another. By restricting the analysis to developmentally integrated subsets of events it is possible to sidestep this problem, but as seen above, this limits the temporal resolution of the analysis. The use of ossification patterns can be valuable' for understanding the biology and morphology of a species, and also in examining the evolutionary history of that species. Many problems remain, first and foremost the limited availability of analytical techniques that are biologically realistic and able to analyze the large data sets needed for good temporal resolution. Increased taxonomic sampling will reveal the relative influence of external environmental factors, function and morphology on ossification sequences.

188 TABLE 5.1. Rank order of element ossification for selected charadriiforms.

Sterna Stercorarius Larus Z. L. canus hirundo skua argentatus ridibundus Schumacher Maillard Schumacher and Wolff (1948) and Wolff (1966b) Element (1966b)

Skull Basioccipital 10 9 16 14 Exoccipital 13 8 1 13 11 Supraoccipital 14 11 18 14 Parasphenoid rostrum 6 4 9 6 Parasphenoid ala 8 Parasphenoid lamina 7 8 17 12 Basisphenoid 7 5 7 10 10 Laterosphenoid 20-23 21 16 Prootic 18 17 14 Opisthotic 19 Epiotic Squamosal 3 3 3 5. 4 Parietal 8 12 7 18 14 Frontal 6 6 7 11 8 Lacrimal 5 6 6 7 8 Mesethmoid 22 22 18 Trabeculae Nasal 5 5 4 6 7 Premaxilla 5 5 4 6 6 Maxilla 4 5 3 4 3 Palatine 4 5 3 4 4 Pterygoid 4 5 3 4 3 Vomer ? 6 4 9 6 Jugal 5 6 4 6 6 Quadratojugal 4 3 4 4 3 Quadrate 12 10 7 17 14 Ectethmoid Dentary 5 6 4 5 5 Supra-angular 5 5 3 4 3 Angular 5 6 2 • 4 3 Splenial 6 6 4 5 4 Prearticular <13 7 Articular 16 Mandibular 5 5 Entoglossal Basihyal Urohyal Ceratobranchial 5 8 4 5 6 Epibranchial

189 Postcranial axial skeleton Cervical centra 13 8 11 15 Thoracic centra 13 8 11 14 Synsacral centra 13 9 Caudal centra 19 14 Pygostyle Cervical neural arch 18 12 13 Thoracic neural arch 17-19 Synsacral transverse processes 21 Caudal transverse processes 20-22 Synsacral arch . 23 Cervical ribs 18 14 Dorsal ribs 8-10 5 8 8 Sternal ribs Uncinate processes Sternum (body) Laterocranial process Laterocaudal process

Forelimb Scapula 9 5 4 4 3 Coracoid 10 8 5-7 11 11 Furcula 2 3 1 3 2 Humerus 1 1 1 1 1 Radius 1 1 1 1 1 Ulna 1 1 1 1 1 Radiale Ulnare Metacarpal 11 Phalanx 1 8 11 7 12 13 Phalanx 2 20 Metacarpal 111 3 6 2 4 3 Phalanx 1 6-8 8 7 13 8 Phalanx 2 9 9 7 13 11 Metacarpal IV 4 6 2 4 3 Phalanx 1 11 21

Hind limb Ilium 10 6 7 8 8 Ischium 10 6 7 7 5 Pubis 8 5 4 15 13

190 Femur 2-4 2 1 2 2 Tibia 1 1 1 1 1 Fibula 1 1 1 1 1 Patella Pretibial bone 20 Tarsals Metatarsal I 12-19 15 20 Phalanx 1 12-16 12 17 Phalanx 2 12-15 13 19 14 Metatarsal 11 4 6 , 2 4 3 Phalanx 1 5 7 4 12 9 Phalanx 2 9 8 7 12 11 Phalanx 3 12 • 13 7 17 11 Metatarsal III 4 3 1 4 3 Phalanx 1 5 7 4 12 8 Phalanx 2 5 8 7 12 11 Phalanx 3 12 9 7 13 13 Phalanx 4 9-11 13 7 14 11 Metatarsal IV 4 4 1 4 3 Phalanx 1 5 7 4 13 9 Phalanx 2 6-9 11 7 16 11 Phalanx 3 10-12 13 7 18 13 Phalanx 4 14 13 7 18 13 Phalanx 5 12 13 7 16 11 If an element is unnumbered, it was either unossified in all specimens examined or unreported by other authors. If two numbers are given, these represent the range of ranks over which a variable element can ossify.

191 Bridging text 5. As discussed in the previous chapter with regard to the interrelationships of the Lari, ossification sequence data lack a strong phylogenetic signal when a small taxonomic sample of recently diverged species is considered. This is not necessarily informative with regard to the capacity of ossification sequences to reveal more ancient evolutionary splits, or for their potential to resolve relationships when more taxa are considered. This possibility remains not unlikely, given levels of simultaneity within sequence data. In this chapter, I pool all of the ossification sequence data collected (Chapters 2-5, Appendix 2). I examine the capability of ossification sequence data to reconstruct phylogeny, and I also examine patterns of character evolution on an accepted topology. Both of these approaches provide objective information for the role phylogeny plays in ossification sequence, and also how ecological and morphological convergence affect ossification sequences. An analysis such as this one, in which all skeletal elements are considered together as a single data set, has never been implemented before using ossification sequence data. While it is possible, even likely, that there is no biological significance in heterochronies involving cranial events moving relative to postcranial events in terms of constraint and modularity, pooling all of the skeletal elements in a single data set allows for increased power to detect heterochronies in the strict sense (ie. as changes in timing). This analysis reveals that relative changes in size of homologous elements in different lineages results in sequence changes, and that many of the largest observed heterochronies in the avian dataset can be explained in this way.

192 CHAPTER 6

Phylogenetic analysis of avian ossification sequence data

Morphology changes over evolutionary time, and the mechanistic causes of this change can be understood by studying development in a comparative context. Heterochrony, or changes in developmental timing resulting in evolutionary changes, is considered to be an important mechanism in morphological change. The study of heterochrony has been broken down into two main components: growth heterochrony, or differential growth rates which result in changes in size and shape, and sequence heterochrony, or changes in timing of developmental events (Smith, 2001). Ossification sequences are species-specific patterns of bone formation, which on a qualitative level appear to exhibit some degree of evolutionary conservation. The sequence is not invariant, however, and heterochronies do occur (Chapters 2, 3, 4, 5). Other researchers have not found ossification sequences to exhibit strong phylogenetic signal when tested in an explicit context using objective methodologies (Sanchez-Villagra, 2002; Schoch, 2006). In these studies, however, the objective was not to characterize ossification sequence changes between closely related taxa. The number of elements in the series was also relatively low and restricted to localized parts of the skeleton. Other studies have used statistical methods to examine divergence in ossification sequences both within and between groups, but failed to tie the results to a specific hypothesis of character evolution (Sheil, 2003b). The sequence of formation of epiphyseal ossification centers has also been examined in a phylogenetic context, but these are unlikely to be influenced by the same factors that drive primary centers of ossification (Maisano, 2002b; Hofmann et al., 2007) due to differences in the mode of development between the two. Many factors are believed to influence ossification sequences. Adaptationist hypotheses predominate in the literature, as ossification sequences are poorly understood and this encourages speculation in order to explain the patterns observed (Adriaens and Verraes, 1998; Mabee et al., 2000; Sanchez-

193 Villagra, 2002; Prochel, 2006). Even some of the most concrete examples of larval adaptation and ossification sequence have been questioned: for instance, in the cannibalistic fish Brycon moorei, teeth are present prior to the ossification of the jaw elements and carnivory begins prior to the completion of jaw and palatal ossification (Vandewalle et al., 2005), something that was previously believed to be necessary for function (Adriaens and Verraes, 1998). In some marsupials, the neonate crawls to the pouch prior to the ossification of the forelimbs, providing a second example of maintenance of function in the absence of ossification (Gemmell et al., 1988). Previous studies have also implicated growth rate (Arendt and Wilson, 2000) and muscle action (Adriaens and Verraes, 1998; Wagemans and Vandewalle, 2001; Smith, 2006) as important determinants of ossification sequence. Few studies have found that adult morphology affects ossification sequence, and those that do do not attempt to extend their observations past the ontogeny of skeletal adaptations (Adams, 1992; Rieppel, 1993b). Extant birds provide a good system on which to study the evolution of ossification sequences. While the relationships within orders may not be well- resolved (Livezey, 1997; Donne-Gousse et al., 2002; Crowe et al., 2006; Kaiser et al., 2007), there is consensus as to the topology of the most basal divergences (Gibb et al., 2007; Livezey and Zusi, 2007; Slack et al., 2007). This creates a simple benchmark with which to test the accuracy of any topology derived using developmental data. The majority of primary ossification centers arise in ovo, and so larval adaptation is not a factor influencing avian ossification sequences. Avian morphology is limited by the constraints imposed by flight, and most birds share a basic body plan making homology determination straightforward. However, a degree of morphological and ecological differentiation has occurred, and the way in which this affects skeletal development can be observed. In this study, I use an objective approach to examine ossification sequence evolution in the avian dataset I have compiled (Chapters 2, 3, 4, 5, Appendix 2). This approach relies on a series of pairwise comparisons to describe the sequence (Smith, 1997; Velhagen Jr., 1997) and to determine which events have shifted. This is the best method available for topology construction. A second analysis is needed to examine character movements on an existing topology, and for this

194 purpose the Parsimov method presents a good compromise (Jeffery et al., 2005). It examines all possible solutions for internal nodal reconstructions, and the most parsimonious solution is selected. If more than one equally parsimonious solution exists, as is often the case, a consensus sequence is calculated. A heuristic search diminishes computational intensity. Unlike event-pairing, the Parsimov method is unable to construct topologies. Because it still relies on event-pairs, however, problems involving illogical ancestral reconstructions persist, but the reasonable computation time more than outweighs this flaw. In this analysis, I test whether ossification sequence data contains phylogenetic signal in birds at the ordinal level, and whether morphology affects ossification sequence when the data are objectively analyzed, as was hypothesized based on comparative observations (Chapter 2, 4).

MATERIALS AND METHODS

The order of ossification of 99 skeletal elements (see Legend, Table 6.2) was coded for 14 avian species (Anasplatyrhynchos, Cairina moschata, Coturnix coturnix (two samples), Dromaius novaehollandiae (two samples), Gallus gallus (two samples), Larus argentatus, L. canus, L. ridibundus, Meleagris gallopavo, Phalacrocorax auritus, Somateria mollissima, Stercorarius skua, Sterna hirundo, Struthio camelus) based on the examination of cleared and stained specimens, or from previously published sequences (Maillard, 1948; Schumacher and Wolff, 1966a; Nakane and Tsudzuki, 1999). Of these, only the ossification sequence of the Double-crested Cormorant (Phalacrocorax auritus) has not been described in the literature or in this thesis (Appendix 2). Only species for which the ossification sequence was resolved to more than 10 events were used in order to avoid excessive simultaneity and therefore missing data. When the ossification sequences of two populations of a single species were available, these were kept as separate taxonomic units in the analysis. The factors affecting ossification sequences are poorly understood, and at the present time there is little justification for assuming that there is no population-level divergence in ossification sequence (Sheil and Greenbaum, 2005). Ossification sequences were transformed into

195 event-pairs (Smith, 1997; Velhagen Jr., 1997) using a subroutine of a sequence heterochrony analysis program developed by Harrison and Larsson (In Press). 99 events yielded a matrix of 4851 event-pairs. Intraspecifically variable event-pairs were coded as missing data ('?'), as were simultaneously occurring events. There is a precedent for both of these coding approaches in the literature (Chapter 1). The event-pairs were used as independent characters to reconstruct a phylogeny using parsimony implemented with PAUP* 4.0M0 (Swofford, 2002). Although this approach has little philosophical justification (Chapter 1), it is currently the only method available for topology reconstruction based on large developmental data sets. All characters were treated as unweighted and unordered. A heuristic search was performed using 10 000 random starting trees and a TBR branch-swapping algorithm. A strict consensus of the most parsimonious trees is reported. Character evolution was reconstructed on the consensus tree using the Parsimov algorithm (Jeffery et al., 2005). This method examines all sequence changes needed to explain the observed shifts on a given branch, but produces a consensus output that includes only those changes found in every equally parsimonious solution. While the consensus sequence does not contain all observed changes, the large number of changes in the data makes this approach feasible. 'Ambiguous' optimization was allowed to increase the chance of detecting large shifts in the vicinity of the root. Only shifts common to both ACCTRAN and DELTRAN optimizations are discussed (Tables 6.1, 6.2). The same approach was used to reconstruct ossification sequence changes on an accepted topology (Dyke and van Tuinen, 2004; Thomas et al., 2004). Orders where interrelationships remain unresolved (Galliformes, Anseriformes) were treated as polytomies. Both trees were rooted using Palaeognathae (Struthio and both populations of Dromaius) as an outgroup (Figures 6.1, 6.2). Palaeognaths are recognized as being basal within extant birds, when either morphological or molecular data are analyzed (Gibb et al., 2007; Livezey and Zusi, 2007). In the case of the tree constructed with developmental sequence data, the addition of a root did not affect the resulting topology.

196 RESULTS

Topology reconstruction A phylogenetic analysis using event-pairs as characters produced six most parsimonious trees of length 2161. 829 of the event-pairs were parsimony- informative. The resulting trees were identical, except for the variable placement of two taxa, Somateria mollissima and Coturnix coturnix (Nakane and Tsudzuki, 1999). The strict consensus is shown in Figure 6.1. An Adams consensus, preferred for data sets in which a lot of data is coded as missing, gives the same result. The observed topology is not consistent with what is known of the •interrelationships of extant birds, but several clades were correctly recovered. 1) Palaeognathae and Neognathae: Struthio and Dromaius were recovered as sister taxa prior to the addition of a root. Supporting shifts include a change in timing of the ossification of the frontal, nasal, scapula, metacarpal IV, dentary, prearticular and pedal phalanges IV:2 and 4. The directionality of these shifts is problematic (see Discussion). 2) The two samples of Dromaius formed a monophyletic group. Many shifts characterize this node, including delays in the ossification of the prootic, manual phalanges 111:2 and IV: 1, metacarpal IV and pedal phalanx IV:2, and early ossification of the prearticular, sternal ribs, sternum, and tarsals. 3) Galliformes + Anas: Anas is correctly placed as sister taxon to the Galliformes, although a monophyletic Anseriformes was not recovered. This node is supported by a sequence shift between the pretibial bone and opisthotic, and a delay in the ossification of the vomer. 4) Galliformes: The order Galliformes was recovered as a monophyletic group, although specific relationships within Galliformes were not recovered (specifically Coturnix and Gallus). The parietal, vomer and

197 pedal phalanx 11:1 are delayed in sequence. The ossification of the sternal ribs and laterocaudal processes of the sternum are accelerated. 5) Neoaves (to the exclusion of Stercorarius): This node is supported by the early ossification of pedal phalanges 111:2 and IV:2, and the late ossification of the quadrate.

Character evolution The accepted phylogeny had a length of 2280 (Fig. 6.2). Only shifts defining more inclusive clades will be discussed. 1) Palaeognathae and Neognathae were separated by shifts involving the frontal, nasal, dentary, prearticular, ceratobranchials, scapula, metacarpal IV, pretibial bone, and pedal phalanges IV:2, 3, and 4. The uncertainty in polarity is addressed in the discussion. 2) Dromaius: The prootic, manual phalanx 111:2, metacarpal IV and pedal phalanx IV :2 are delayed in sequence. The prearticular, sternal ribs, sternum and tarsals ossify early. 3) Galloanseres: This clade is defined by a sequence change between the parasphenoid alae and the prearticular. The supraoccipital ossifies early, and the basisphenoid, vomer and pedal phalanx 111:2 are delayed. 4) Galliformes: The vomer, ischium and pedal phalanx 11:1 ossify late, whereas the sternal ribs, laterocranial and laterocaudal processes of the sternum and manual phalanx III: 1 ossify early. 5) Gallus: Both samples of Gallus gallus share a late ossification of the parietal, a delay in the ossification of pedal phalanx IV:2, and a further acceleration in the ossification of the sternal ribs. 6) Coturnix: Both samples of Coturnix coturnix share a late ossification of the palatine, and an early ossification of pedal phalanx IV:3. 7) Anseriformes: The ducks all share an early ossification of the lacrimal and articular, and a late ossification of metatarsal I. 8) Neoaves: This clade is characterized by an early ossification of the basisphenoid and a late ossification of the quadrate.

198 9) Charadriiformes: The ossification of the parietal is delayed, as are pedal phalanges 1:1 and 1:2. 10) Larus + Sterna: The sister-group relationship between Larus and Sterna is supported by the early ossification of both the major and minor metacarpals and the late ossification of pedal phalanx 111:3. 11) Larus: All three species in the genus Larus share a late ossification of the squamosal and parietal, and an early ossification of the splenial and scapula.

199 Figure 6.1. Strict consensus of the six most parsimonious trees generated when event-pairs were used as characters. Tree length = 2161, CI = 0.621. Node numbers correspond to the list of changes in Table 6.1.

200 " Somateria mollissima

" Sterna hirundo

• Lams argentatus CO • Phalacrocorax auritus

• Lams ridibundus

• Larus canus

• Anas ptatyrhynchos

• Gallus gallus (S&W)

z " Gallus gallus CD O (O • Meleagris gallopavo o en " Coturnix coturnix 3" 3 CD • Coturnix coturnix (N&T)

' Cairina moschata

• Stercorarius skua

• Dromaius novaehollandiae (RM)

" Dromaius novaehollandiae (YPM)

"0 SL • Struthio camelus

o

0) Figure 6.2. Accepted phylogenetic relationships of the avian species studies. Tree length = 2280, CI = 0.588. A complete list of the shifts supporting nodes are listed in Table 6.2.

201 !~

Anseriformes . 'Galliformes Lacrimal, articular early Sternal ribs, Metatarsal I iate laterocranial anc laterocauda! processes, manual phalanx 1 Galloanseres. 111:1 early •Charadriiformes Palaeognathae Supraoccipita! eariy Vomer, ischium, Parietal, pedal Basisphenoid, vomer, pedal phalanx phalanges 1:1,2 pedal phalanx l!l:2 late 11:1 late late

1 Neoaves Basisphenoid early Quadrate late

Frontal, nasal, ceratobranchials, scapula, metacarpal IV. pedal •Neognathae phalanges IV:2,3,4 early Prearticular, dentary, pre-tibial bone late

• Neomithes DISCUSSION

Stercorarius skua is pulled to a basal position within Neognathae in the topology constructed using event-pairs due in part to late ossification of the ceratobranchials, a feature shared with palaeognaths. Stercorarius skua also exhibited delayed ossification of pedal phalanges IV:2 and 4 as reported for ratites, and an early ossification of manual phalanx IV: 1 as in Struthio camelus (Table 6.1). The ossification sequence of Stercorarius skua was taken from the literature (Maillard, 1948), rather than being based on personal observation and so the accuracy of these observations could not be confirmed. Stercorarius skua is the most displaced taxon in this analysis. Aside from the displacement of Stercorarius, the greatest failure of the event-pair topology was the inability to recover the monophyly of Anseriformes. Even when the accepted topology was considered, this order is united by only four shifts, two involving cranial bones and one involving a postcranial element (Fig. 6.2). This lack of support at the ordinal level is also present for Charadriiformes. These results could be caused by several factors, including methodology. Better taxonomic sampling, especially within Neoaves, might improve resolution at the ordinal level. It has been suggested that event-pairing creates a bias forcing shifts on to terminal branches, leaving internal branches character-impoverished (Harrison and Larsson, In Press). Alternatively, poor support at the ordinal level might be a feature of ossification sequence data itself. Intraspecific variation, both within and between populations, does exist and it has not been demonstrated that the rate of evolution of ossification sequences is appropriate for resolving relationships at the ordinal level. As has been suggested based on description (Chapter 2, 4), evolutionary reduction or enlargement of elements has a strong effect on their relative timing of ossification. This analysis provides more quantitative evidence to support the qualitative examples discussed in previous chapters. For instance, the vomer is reduced in both Galliformes and Anseriformes relative to the state seen in ratites (Beddard, 1898), and is further reduced in Galliformes to the point where it did not ossify prior to hatching in Meleagris gallopavo. This is reflected in two

202 sequence changes involving the vomer: the first shared by Galloanseres, in which the vomer is delayed relative to the parietal and frontal, and the second shared by Galliformes in which the vomer is delayed relative to 15 additional elements. These delays are enough to maintain the monophyly of Galliformes in the event- pair topology, and also to maintain Anas as the sister group of Galliformes. The ceratobranchials, reduced in palaeognaths due to obligate inertial feeding in ratites (Tomlinson, 2000), were found to ossify much earlier in neognaths. This difference amounted to a shift relative to 10 other elements. The delayed ossification of the ceratobranchials in Stercorarius cannot be explained using this model and reinforces the hypothesis emerging from this thesis that ossification sequences are determined by multiple factors. Metacarpal IV ossifies later in sequence in ratites than in neognaths. Ratites are widely cited as having reduced wings, with the wings being more reduced in Dromaius than in Struthio. Metacarpal IV was further delayed in both populations of Dromaius, ossifying late relative to 13 elements. The scapula of ratites is also proportionally smaller than its homologue in flying birds (Parker, 1881), and a delay in the ossification of the scapula supports the separation of palaeognaths from neognaths in the tree constructed using event-pairs as well as in the accepted topology. A delay in the ossification of the elements of the hallux characterizes the order Charadriiformes (in this case the clade Lari). The Lari are described as having a reduced hallux (Gill, 1995). This delay is probably larger than is optimized, as an acceleration of these elements was found to occur on the branch leading to Phalacrocorax. The large delay in Charadriiformes was likely averaged between the node Charadriiformes and Neoaves during optimization. In a putative example of this phenomenon operating in the dermal skeleton, the ossification of the lacrimal in Anseriformes is accelerated relative to Galliformes. The lacrimal in Anseriformes is very robust, while it is reduced in Galliformes (Cracraft, 1968). This acceleration was also observed in a swan (Cygnus sp.). Increased sampling within Galliformes is necessary in order to confirm that this sequence is not an artifact of intraspecific polymorphism. Intraspecific differences in the size of an element resulting in changes in the sequence and timing of ossification are found in the appendicular skeleton of ratites and have already been discussed (Chapter 2).

203 Relative changes in element size between species leading to changes in ossification sequence has also been found in other amniotes (Rieppel, 1993b), implying that evolutionary changes in the size of a structure may alter sequence position in a predictable way in most vertebrate taxa. The directionality of shifts around the root (Fig. 6.2) is somewhat arbitrary and does not provide accurate information on the primitive state for birds. For instance, given the data, the ceratobranchials can be interpreted as ossifying early in ratites and late in neognaths, or late in ratites, or early in neognaths. Reconstructing the ancestral sequence for Aves requires additional information, and this process may be aided by the observation that morphological reduction or enlargement affects the sequence position of an element. Using this information, the delays in the ossification of the scapula and metacarpal IV are most likely autapomorphic to ratites, which have secondarily reduced forelimbs, rather than being primitive to birds. Conversely, an early ossification of the pretibial bone (ascending process of the astragalus) is most likely primitive for birds, as the ontogeny and morphology of this element is modified in neognaths, but not in ratites and tinamous (McGowan, 1985). This creates a powerful tool for reconstructing ancestral ossification sequences, and helps identify potential apomorphies in the ontogeny of a group where developmental sequences are unknown or unknowable. The underlying cause of relative size changes resulting in ossification sequence changes remains unknown, as does the extent of the phenomenon and its consequences and evolutionary implications. Alteration in the width of a bone compared to changes in bone length might affect sequence position differently, but at the scale this analysis was conducted, these effects cannot be distinguished. Likewise, it is uncertain whether complex allometries would change the results observed (for instance, if in species A element X was relatively large in the embryo, but there was no relative size change in the adult). In all of the examples presented here, the affected elements have undergone changes in size, proportionately similar in the embryo and the adult. A possible way to test this pattern is to examine sequence changes in organisms with more complex ontogenies. The forelimbs of marsupial mammals, for instance, are relatively

204 larger in early ontogeny compared with their homologues in placental mammals, and the hind limbs are relatively smaller. This relationship is not maintained in adulthood, however, as bipedal locomotion has evolved in several marsupial groups. In the bandicoot adult, locomotion is facultatively bipedal (Bennett and Garden, 2004). In the neonate, however, forelimbs are larger than hind limbs and ossified earlier in sequence (Gemmell et ah, 1988), suggesting that embryonic morphology and not adult morphology is controlling timing of ossification. It has been shown that the ossification of the hind limb ossification is delayed in marsupials relative to placental mammals (Harrison and Larsson, In Press). Hind limb formation is delayed relative to the forelimbs at all points in development, including limb bud formation and chondrification (Bininda-Emonds et ah, 2007). Therefore hind limb ossification may occur later due to immaturity of the tissues, not because hind limbs are smaller in size in the neonate. Delayed tissue maturity in smaller structures may also be playing a role in the avian system, although it is not as obvious as in marsupials. The evolutionary effects of changes in ossification sequence are unknown. Ossification may be independent of morphological evolution if sequence heterochrony is a by-product of morphological change, or sequence heterochronies may facilitate morphological change, with delays in ossification leading to small size or, in the extreme case, the loss of an element. Previous studies imply that the last elements to ossify are the first to be lost during the evolution of a lineage (Miiller and Alberch, 1990). Precociality and altriciality are phylogenetically correlated in this sample, with semialtricial and semiprecocial taxa being restricted to the clade Neoaves. No fully altricial species were sampled. The degree of precociality is not thought to influence ossification sequence (Rogulska, 1962; Starck, 1993, 1998), but ossification sequence in birds has not previously been examined using a metric as sensitive as ranked data. The failure of the semiprecocial Stercorarius skua to group with the other semiprecocial and semialtricial Neoaves supports the previously held assertion that the degree of precociality at hatching does not affect ossification sequence in birds - at least not in an easily recognizable way such as a

205 large delay in the ossification of specific elements and an acceleration in the ossification of other bones (Fig. 6.1). An important factor influencing the validity of these results was homology determination. For instance, the neognath pretibial bone was considered to be homologous to the ascending process of the astragalus in ratites, in spite of different ontogenies of this element at the cartilaginous stage (McGowan, 1984, 1985). There was a shift observed between ratites and neognaths in the timing of ossification of this element, with neognaths ossifying the pretibial bone later than ratites ossified the ascending process. The significance of this shift is unknown at the present time. The implication of the phalangeal count to the ossification sequence of the phalanges was also unknown. For instance, Dromaius delays ossification of manual phalanx 111:2. However, the phalanges do not ossify in a proximal to distal sequence, and the ossification of phalanx 111:3 occurs much earlier in sequence than phalanx 111:2. In a bird with only two phalanges, such as the chicken, the distal phalanx (111:2) may correspond to the distal phalanx of Dromaius (111:3), or to 111:2. For this analysis, the terminal element is assumed to be absent in species with phalangeal count 111:2. Lastly, some 'elements' were really multiple ossification centers within a single bone - the parasphenoid rostrum, alae and lamina or the laterocranial and laterocaudal processes and body of the sternum for instance. In some birds, like Anseriformes, separate ossification centers were not seen for the parasphenoid alae and lamina, and so they were assumed to ossify simultaneously. Whether or not this was a valid approach is open to debate. Likewise, the mentomandibular and dentary (Schumacher and Wolff, 1966a) and prearticular begin ossification in variable locations and may not be homologous across taxa. Lastly, event-pairing has frequently been criticized in the literature as it may lead to illogical reconstructions at internal nodes (Schulmeister and Wheeler, 2004; Harrison and Larsson, In Press). The Parsimov algorithm for character reconstruction combined with event-pairing was selected for this study for two reasons: it allows for the incorporation of polymorphic data, and it has a reasonable runtime for a large data set. The PGi method (Harrison and Larsson, In Press) will be used in the future; however the runtime for this dataset exceeded

206 three months and so it is not presented here. There is some evidence that event- pairs were resulting in illogical reconstructions at internal nodes in this analysis. When comparing ACCTRAN and DELTRAN optimizations, some elements were reconstructed as ossifying early in one optimization, while in the other optimization the same character was noted as being late to ossify at the same node (Tables 6.1, 6.2). This discrepancy raises questions regarding the accuracy of the results obtained. The best method for the analysis of developmental sequence data remains an open question.

Biological implications of ossification sequence heterochrony Ossification sequence changes can be thought of as a combination of events undergoing heterochronic changes (changes in developmental timing) and 'placeholder' events, or those that allow changes in timing to be identified as changes in sequence. Larger changes in timing are more likely to be identified because they will shift relative to more placeholder events, although the magnitude of a sequence heterochrony is largely dependent on how condensed the region of the sequence is in which it occurs. This means that in principle ossification sequence heterochronies are dependent on growth and maturity of other tissues, not on other ossification events (collective rather than linear dependence; Koenemann and Schram, 2002). Changes in ossification may therefore be governed by the same factors determining the timing of other ontogenetic events in the presumptive region, specifically growth and tissue differentiation. An example of this concept can be seen in the delayed ossification of reduced elements. Relative reduction of developmental fields will result in a series of effects that may act individually, or more likely, in a more complex combinatorial fashion. The most obvious effect of a reduced developmental field is an overall reduction in cell number. For instance, reducing Shh expression in the developing limb bud results in reduced cell proliferation, and affects digit number and identity (Towers et al., 2008). This experimental evidence supports theoretical studies finding reduction of cell number and absolute dimensions of a developmental field result in different stable morphogenetic patterns that, in turn,

207 give rise to different numbers of structural elements (Newman and Frisch, 1979). Reduced field sizes may also result in reduced muscular development, which in turn, may diminish embryonic muscle activity and forces and limit the mechanical stimulation of ossification (Glucksmann, 1942; Herring, 1993). These factors, probably acting in a combinatorial mode, result in delays or absence of ossification. Conversely, a high proportion of cartilage in the hatchling skeleton results in higher post-hatching growth rates (Starck, 1996). Ossification sequence is invariant, however, rather the degree of ossification is variable (Starck, 1989, 1993, 1996). Thus the initiation of ossification is decoupled from overall structural and functional maturity of an element. This is as expected based on the observation of delayed ossification in reduced elements - if delaying ossification resulted in increased growth, evolutionarily enlarged structures would experience delays in their formation, not vice versa as is actually observed. Aside from these factors, many ossification sequence heterochronies may result from small changes in very condensed areas of the ossification sequence. Although these changes do not reflect large timing shifts and may not be biologically relevant, they appear as equal to real, larger shifts in less condensed areas of the sequence. This means that care must be taken a priori in selecting placeholder events, as well as events hypothesized to have shifted in order to ensure an even distribution of events in the developmental time frame of interest. Incorporating some measure of developmental time may also be desirable, as it is a more precise way of ensuring that observed heterochronies are actually heterochronies and not artifactual.

CONCLUSION

Ossification sequence data were found to contain a stronger phylogenetic signal than was recovered in previous studies (Sanchez-Villagra, 2002; Schoch, 2006). This might have been due to the sheer volume of elements and species examined in this study, as this is by far the largest data set assembled for this purpose. In spite of the volume of data collected, however, the results did not

208 provide any justification for considering ossification sequence data a solution to difficult systematic questions. In fact, it seems as if ossification sequences may be more closely linked to morphology, being influenced by the evolutionary enlargement or reduction of structures. This suggests that this type of data may be just as susceptible to convergence as other morphological data, and for the same reasons. However, because of the hypothesis that developmental timing and embryonic and adult morphologies are related, albeit in a complicated way, the use of developmental sequences may offer unique insight into questions of developmental evolution such as modularity and constraint without being particularly useful for reconstructing phylogenetic topologies.

209 Table 6.1: ACCTRAN, DELTRAN and consensus reconstructions of hypothesized sequence changes on an avian phylogeny constructed using event- pairs. Node DELTRAN ACCTRAN Consensus Node 38 (Root) Twins (50, 3) (76, Twins (30, 25) (52, Char 14 moved E -> Node 35 61) (80, 1) 48) relative to 86, 90, Char 14 moved E Char 1 moved E 95 relative to 86, 90,95 relative to 3,42 Char 18 moved E Char 18 moved E Char 9 moved E relative to 23, 24, relative to 23,24, 31, relative to 43, 46 31,90 90 Char 14 moved E Char 28 moved L Char 28 moved L relative to 86,90, 95 relative to 20, 21, relative to 20, 21,22 Char 16 moved L 22 Char 57 moved E relative to 8, 10, 11 Char 57 moved E relative to 74, 90, 95 Char 18 moved E relative to 74, 90, Char 62 moved E relative to 23, 24, 95 relative to 89, 94 31,90 Char 71 moved E Char 71 moved E Char 19 moved E relative to 86, 90, relative to 86, 90, 95 relative to 18 95 Char 28 moved L relative to 20, 21, 22 Char 32 moved L relative to 2, 5,26 Char 57 moved E relative to 74, 75, 90,95 Char 71 moved E relative to 74, 86, 90,95 Char 80 moved L relative to 50, 69 Char 87 moved L relative to 88 Char 88 moved L relative to 5, 13, 26, 69, 73 Char 91 moved L relative to 69, 73, 87 Char 93 moved L relative to 69, 87 Char 96 moved L relative to 87 Char 99 moved L relative to 87 Node 35 -> Twins (97, 83) Char 38 moved E Char 38 moved E Node 34 Char 16 moved L relative to 7, 14, 19, relative to 7, 14, relative to 8, 10 23,31,51,86,90, 86, 90, 95 Char 32 moved L 95 Char 66 moved E relative to 5, 6, 26 Char 66 moved E relative to 42 Char 38 moved E relative to 42, 69, Char 68 moved E relative to 7, 14,86, 70, 92 relative to 19, 22

210 90, 95 Char 68 moved E Char 72 moved L Char 66 moved E relative to 19,20, relative to 33, 43, relative to 41, 42 21,22,29 67 Char 68 moved E Char 71 moved E Char 91 moved L relative to 19,22 relative to 19, 21, relative to 26 Char 70 moved E 22,51 Char 96 moved E relative to 40, 41 Char 72 moved L relative to 1,42 Char 72 moved L relative to 33, 43, Char 98 moved E relative to 33,43, 67 44, 48, 67 relative to 1, 3, Char 74 moved L Char 73 moved L 42,45 relative to 75 relative to 13,26,87 Char 75 moved L Char 83 moved E relative to 14, 57 relative to 96, 97 Char 80 moved L Char 84 moved E relative to 58, 69 relative to 83 Char 91 moved L Char 91 moved L relative to 26, 58, 69 relative to 13, 26 Char 92 moved E Char 96 moved E relative to 2,40, 41 relative to 1, 42 Char 96 moved E Char 98 moved E relative to 1,41,42 relative to 1, 3, 42, Char 98 moved E 45,96 relative to 1, 3,41, 42,45 Node 34 --> Twins (43, 82) (48, Twins (43, 82) Twins (43, 82) Node 33 49) Char 5 moved L Char 6 moved L Char 6 moved L relative to 2, 87, 92, relative to 87, 88, relative to 83, 87, 88, 96 92,99 92, 99 Char 6 moved L Char 71 moved E Char 38 moved E relative to 2, 26, 66, relative to 18, 20 relative to 19, 23 84, 87, 88, 92, 93, Char 75 moved L Char 66 moved E 96, 99 relative to 86, 90 relative to 69, 70, 92 Char 49 moved E Char 80 moved L Char 68 moved E relative to 44, 48 relative to 45,46 relative to 20, 21 Char 67 moved L Char 97 moved E Char 71 moved E relative to 9, 10 relative to 8, 9, relative to 15, 18, 19, Char 71 moved E 10,45,46 20,21 relative to 18, 20 Char 75 moved L Char 74 moved L relative to 86, 90 relative to 90, 95 Char 80 moved L Char 75 moved L relative to 3,45, 46, relative to 58, 66, 50 86, 87, 90, 93, 95 Char 97 moved E Char 80 moved L relative to 8, 9, 10, relative to 45, 46 45, 46 Char 87 moved L relative to 26, 66 Char 97 moved E relative to 8, 9, 10, 32,45,46 Node 33 --> Twins (79, 65) (82, Twins (59, 94) (79, Twins (79, 65)

211 Somateria 45) 65)(82, 45) (82, 45) mollissima Char 5 moved L Char 1 moved L Char 5 moved L relative to 2, 3, 41, relative to 3, 41,42 relative to 3, 41, 42, 83, 84, 87, 92, Char 5 moved L 42, 83, 84, 98 96,98 relative to 3, 41, 42, Char 6 moved L Char 6 moved L 83, 84, 88, 98 relative to 3,41, relative to 3,41, 42, Char 6 moved L 42, 98 84,96,98 relative to 3, 41, 42, Char 8 moved E Char 8 moved E 98 relative to 43, 80 relative to 43, 80 Char 8 moved E Char 9 moved E Char 9 moved E relative to 43, 80 relative to 8 relative to 8 Char 9 moved E Char 32 moved L Char 32 moved L relative to 8 relative to 1,41, relative to 1, 3, 41, Char 32 moved L 42 42, 97 relative to 1,41,42 Char 99 moved E Char 73 moved L Char 46 moved E relative to 69, 70 relative to 26, 66, 87, relative to 9 93 Char 81 moved E Char 87 moved L relative to 44, 72 relative to 26, 66, 93 Char 93 moved E Char 91 moved L relative to 69, 73 relative to 13, 87 Char 99 moved E Char 99 moved E relative to 69, 70, relative to 69, 70 73,91 Node 33 --> Twins (40, 41) (46, Twins (49, 44) (62, Twins (80, 10) Node 28 9) (80, 10) 76)(67,48)(68, 29) Char 23 moved L Char 23 moved L (78, 61)(80, 10) relative to 13, 14, relative to 13, 14, 15, Char 3 moved E 15,86,87,90,95 86, 87, 90, 95 relative to 41 Char 42 moved L Char 4 moved L relative to 3, 50 relative to 14, 15 Char 7 moved L relative to 58, 66, 69, 70, 87, 91 Char 18 moved E relative to 20, 21 Char 23 moved L relative to 5, 13, 14, 15,58,66,86,87, 90, 93, 95 Char 24 moved E relative to 19,28 Char 26 moved E relative to 58, 75, 86 Char 40 moved E relative to 84, 98 Char 41 moved E relative to 40 Char 51 moved E relative to 31, 74 Char 55 moved E

212 relative to 17, 54 Char 56 moved E relative to 17,54,79 Char 73 moved E relative to 31, 66, 75,87 Char 74 moved E relative to 73 Char 88 moved E relative to 70,91, 99 Char 96 moved L relative to 5,6, 84 Char 97 moved L relative to 32, 82 Node 28 --> Char 3 moved E Twins (11, 16) Char 4 moved L Anas relative to 32, 41 Char 4 moved L relative to 58, 66, platyrhynchos Char 4 moved L relative to 58, 66, 86, 90, 95 relative to 14, 58, 66, 86, 90,95 Char 7 moved L 86, 90, 95 Char 7 moved L relative to 93, 99 Char 7 moved L relative to 88,93, 99 Char 13 moved L relative to 66, 69, 87, Char 13 moved L relative to 14 91,93,99 relative to 14 Char 14 moved L Char 13 moved L Char 14 moved L relative to 86, 90, relative to 14 relative to 86, 90,95 95 Char 14 moved L Char 15 moved E Char 15 moved E relative to 86, 90, 95 relative to 19,24, relative to 19,24, Char 15 moved E 28, 38, 57 28,57 relative to 19,24,28, Char 31 moved L Char 31 moved L 57 relative to 23, 58, relative to 23, 58, Char 26 moved E 66, 75, 93,95 66,75,93,95 relative to 58, 73 Char 40 moved E Char 40 moved E Char 31 moved L relative to 2, 83, 92 relative to 2, 83 relative to 23, 58, 66, Char 44 moved E Char 44 moved E 75, 93, 95 relative to 47, 48 relative to 47,48 Char 40 moved E Char 69 moved L Char 74 moved L relative to 2, 83, 84 relative to 88, 93 relative to 13, 86 Char 44 moved E Char 74 moved L Char 75 moved L relative to 47, 48, 72 relative to 13,26,86 relative to 13 Char 50 moved E Char 75 moved L relative to 1, 41 relative to 13 Char 74 moved L relative to 13,86,90, 95 Char 75 moved L relative to 13,87,95 Char 88 moved E relative to 69, 99 Char 91 moved L relative to 13, 87 Node 28 -> Twins (15, 38) Twins (19, 31) Char 13 moved L Node 27 Char 13 moved L Char 13 moved L relative to 70

213 relative to 70, 73 relative to 58, 66, Char 23 moved L Char 23 moved L 70, 83, 84, 87, 88, relative to 1,2,4, relative to 1, 2, 4, 5, 93,99 6, 7, 70, 83, 84, 6, 7, 32, 70, 83, 84, Char 20 moved L 98, 99 88,91,92,93,98,99 relative to 24, 28, Char 52 moved E Char 24 moved E 38,57 relative to 8, 9, relative to 19,20,21, Char 21 moved L 16 Char 56 28 relative to 24, 38, 57 moved E Char 52 moved E Char 23 moved L Char 86 moved L relative to 8, 9, 16, relative to 1, 2, 4, 6, 72 7, 40, 70, 83, 84, 98, Char 53 moved E 99 relative to 72, 81 Char 32 moved E Char 54 moved E relative to 3, 84 relative to 53 Char 33 moved L Char 55 moved E relative to 11, 16,47 relative to 54, 72, 81 Char 42 moved L Char 56 moved E relative to 9, 45, 97 relative to 54, 72, 79, Char 43 moved L 81 relative to 11, 16 Char 58 moved E Char 50 moved L relative to 90, 95 relative to 9, 97 Char 86 moved L Char 52 moved E relative to 26, 73 relative to 3, 8,9, 10, 11,16,46,47, 48,49 Char 56 moved E relative to 47, 48,49 Char 65 moved L relative to 53, 55 Char 82 moved E relative to 3, 41 Char 86 moved L relative to 58, 66 Char 90 moved L relative to 26, 58 Char 95 moved L relative to 26, 58, 66,75 Node 27 --> Twins (16, 56) Twins (14, 73) (16, Twins (16, 56) Gallus gallus Char 3 moved L 56) Char 5 moved E (Schumacher relative to 1, 52 Char 2 moved L relative to 66, 69, &Wolff) Char 5 moved E relative to 32, 98 91,93,99 relative to 66, 69, 88, Char 5 moved E Char 6 moved E 91, 93, 99 relative to 66, 69, relative to 5 Char 6 moved E 86,87,91,92,93, Char 12 moved L relative to 5 99 relative to 22, 29, Char 12 moved L Char 6 moved E 68,71 relative to 22, 29, 68, relative to 5, 84 Char 13 moved L 71 Char 7 moved E relative to 1, 2, Char 13 moved L relative to 66, 69, 98

214 relative to 1, 2, 83, 86, 87,91 Char 51 moved E 87, 88, 92, 93, 99 Char 12 moved L relative to 15, 24, Char 51 moved E relative to 22, 29, 38 relative to 15,24,38 68,71 Char 70 moved L Char 70 moved L Char 13 moved L relative to 83, 99 relative to 83, 99 relative to 1, 2, 32, Char 74 moved E Char 74 moved E 98 relative to 15, 19 relative to 15, 19 Char 51 moved E Char 75 moved L Char 75 moved L relative to 15,24,38 relative to 83, 88, relative to 83, 87, 88, Char 70 moved L 92, 99 91,92,93,99 relative to 32, 83, Char 95 moved L Char 95 moved L 98, 99 relative to 83, 86, relative to 66, 83, 86, Char 74 moved E 87, 91 87,91 relative to 15, 19 Char 96 moved L Char 96 moved L Char 75 moved L relative to 1, 3, 9, relative to 1, 3, 9, 52, relative to 83, 88, 52 83 92, 99 Char 97 moved L Char 97 moved L Char 95 moved L relative to 9, 52 relative to 9, 52 relative to 32, 83, 86,87,91 Char 96 moved L relative to 1, 3, 9, 32,52 Char 97 moved L relative to 9, 52 Node 27 --> Char 1 moved E Char 5 moved L Char 7 moved L Node 26 relative to 41, 42 relative to 83, 84 Char 7 moved L Char 7 moved L relative to 66, 91 relative to 83, 92 Char 23 moved L Char 24 moved E relative to 40, 96 relative to 18, 57, Char 52 moved E 68,71 relative to 47, 49 Char 28 moved E Char 82 moved E relative to 18, 57 relative to 42, 50 Char 58 moved E relative to 26, 73 Char 74 moved L relative to 66, 91 Char 75 moved E relative to 26, 86, 87, 93 Char 85 moved L relative to 25, 30, 59 Char 89 moved L relative to 25, 30, 59 Char 98 moved L relative to 1, 84 Node 26 -> Twins (23, 97) Twins (33, 72) Char 4 moved E Node 25 Char 4 moved E Char 3 moved E relative to 51 relative to 51, 73 relative to 1, 82 Char 74 moved L Char 5 moved L Char 4 moved E relative to 69, 86,

215 relative to 83, 87, 92 relative to 14, 15, 51 87,92 Char 7 moved L Char 18 moved L Char 75 moved E relative to 69, 70, 83, relative to 20, 21, relative to 58, 66 87, 92, 93 31,38 Char 24 moved E Char 25 moved E relative to 18,57 relative to 61, 62, Char 74 moved L 77, 78 relative to 14,26,69, Char 30 moved E 86,87,90,91,92,95 relative to 61, 62, 78 Char 75 moved E Char 57 moved L relative to 58, 66, 73 relative to 20, 21,38 Char 67 moved L relative to 16, 46 Char 74 moved L relative to 13, 69, 70, 83, 86, 87, 92 Char 75 moved E relative to 51, 58, 66, 90 Char 84 moved E relative to 2, 40, 83 Node 25 --> Twins (18, 28) Twins (85, 29) Char 97 moved L Node 24 Char 25 moved E Char 4 moved E relative to 8, 10, relative to 30, 61, 62, relative to 31 45,46 77, 85, 89,94 Char 13 moved E Char 98 moved L Char 26 moved E relative to 87, 91,92 relative to 3 relative to 90, 95 Char 31 moved E Char 30 moved E relative to 51, 73, 74 relative to 85, 89, 94 Char 96 moved L Char 41 moved L relative to 1, 3, 9, relative to 3, 82 41,42,50,52,82 Char 52 moved E Char 97 moved L relative to 10, 11,44, relative to 3, 8, 9, 45, 46, 50 10,42,45,46,50, Char 56 moved E 52 relative to 44, 47, 49 Char 98 moved L Char 74 moved L relative to 3, 52, 82 relative to 13,83,84, 99 Char 75 moved E relative to 86, 90 Char 97 moved L relative to 8, 10, 45, 46,82 Char 98 moved L relative to 1,3, 84 Node 24 --> Twins (44,49) (97, Twins (1,52) (44, Char 12 moved E Gallus gallus 11) 49)(45, 42) relative to 61, 62, (Chapter 3) Char 7 moved L Char 12 moved E 89,94 relative to 88, 99 relative to 61, 62, Char 16 moved L Char 9 moved E 89,94 relative to 47, 56

216 relative to 45, 50 Char 16 moved L Char 32 moved E Char 12 moved E relative to 43, 47, 56 relative to 2,40, relative to 59, 61,62, Char 18 moved E 83, 86, 98 89,94 relative to 31, 38, Char 69 moved L Char 13 moved L 68,71 relative to 13, 84, relative to 83, 84, 88, Char 32 moved E 88, 93, 99 93,99 relative to 2, 7, 13, Char 85 moved L Char 16 moved L .40, 83, 86, 96, 98 relative to 20, 21, relative to 47, 56 Char 69 moved L 22 Char 25 moved E relative to 13,84, relative to 59 86, 88, 93, 99 Char 30 moved E Char 85 moved L relative to 59, 61, 62 relative to 20, 21, 22 Char 32 moved E relative to 2, 40, 83, 86, 98 Char 52 moved E relative to 1, 3 Char 59 moved E relative to 89, 94 Char 68 moved L relative to 18,20,24 Char 69 moved L relative to 2, 13,70, 83,84,88,93,99 Char 71 moved L relative to 18,20,24 Char 73 moved L relative to 58, 74 Char 74 moved L relative to 58, 66, 70 Char 85 moved L relative to 18, 20, 21, 22,29 Char 87 moved L relative to 2, 88, 99 Char 91 moved L relative to 13, 70, 83, 88, 99 Char 92 moved L relative to 2, 70, 83 Char 96 moved L relative to 1, 3, 41, 42, 50, 82, 83, 84, 98 Node 24 --> Twins (5,2) (24, 21) Twins (61, 76) Twins (61, 76) Meleagris (61, 76) (70, 13) Char 13 moved E Char 18 moved L gallopavo Char 18 moved L relative to 70, 83 relative to 4, 19, relative to 4, 19,38, Char 18 moved L 57 57 relative to 4, 19, 57 Char 32 moved L Char 32 moved L Char 32 moved L relative to 1, 8, 9, relative to 1,3, 8, 9, relative to 1, 8, 9, 41,42

217 41,42,97 16,41,42 Char 69 moved E Char 67 moved L Char 69 moved E relative to 66, 73, relative to 10, 46, 48 relative to 51, 66, 75,90,95 Char 69 moved E 73, 75,90, 95 Char 72 moved E relative to 66, 73, 75, Char 72 moved E relative to 47, 49, 86,90,95 relative to 43,47, 53,55 Char 72 moved E 49, 53, 55, 67 relative to 33,47, 49, Char 74 moved E 53, 55 relative to 66, 70, 73 Char 82 moved E relativetol,98 Node 25 --> Char 13 moved L Char 24 moved L Char 24 moved L Coturnix relative to 87, 92 relative to 31 relative to 31 coturnix Char 18 moved L Char 28 moved L Char 28 moved L (Chapter 3) relative to 31 relative to 31,38 relative to 31 Char 19 moved L Char 29 moved L Char 31 moved L relative to 31 relative to 68, 71 relative to 58, 66, Char 24 moved L Char 31 moved L 75,86,90,91, relative to 31 relative to 58, 66, 93, 95 Char 28 moved L 75,86,90,91,93, Char 97 moved E relative to 31 95 relative to 41 Char 31 moved L Char 97 moved E relative to 58, 66, 75, relative to 41, 82 86,90,91,93,95 Char 38 moved E relative to 20,21,57 Char 51 moved L relative to 14, 75 Char 73 moved L relative to 74 Char 74 moved L relative to 58, 66, 70 Char 84 moved E relative to 5, 6 Char 97 moved E relative to 1,3,32, 41,42,50 Node 26 --> Char 3 moved L Twins (21, 28) Char 4 moved L Coturnix relative to 1, 82 Char 4 moved L relative to 58, 90 coturnix Char 4 moved L relative to 58, 66, 90 Char 84 moved L (Nakane & relative to 14, 15, 58, Char 26 moved L relative to 1 Tsudzuki) 90 relative to 66, 90 Char 88 moved L Char 20 moved L Char 67 moved E relative to 1, 2, relative to 18, 28, 57 relative to 9, 10,43, 40, 83, 92 Char 21 moved L 48 Char 92 moved L relative to 18, 28, 57 Char 84 moved L relative to 93 Char 83 moved L relative to 1, 13 Char 93 moved L relative to 84 Char 88 moved L relative to 1, 2, Char 84 moved L relative to 1,2, 13, 40, 70, 83 relative to 1,32, 40 40, 83, 92 Char 96 moved E Char 88 moved L Char 92 moved L relative to 2, 13,

218 relative to 1,2,32, relative to 93 40, 70, 87 40, 70, 83, 92 Char 93 moved L Char 99 moved L Char 92 moved L relative to 1,2, 13, relative to 1,2, relative to 93 40, 70, 83 40, 83, 92 Char 93 moved L Char 96 moved E relative to 1,2, 32, relative to 2, 13,40, 40, 70, 83 70, 83, 87 Char 96 moved E Char 99 moved L relative to 2, 13,40, relative to 1,2, 13, 70, 87 40, 83, 92 Char 98 moved L relative td 1,32 Char 99 moved L relative to 1,2,32, 40, 83, 92 Node 33 --> Twins (12, 68) (74, Twins (18, 28) (95, Char 26 moved L Node 32 31)(76, 62)(85, 59) 86) relative to 66, 70, Char 26 moved L Char 17 moved E 88, 92, 93, 99 relative to 66, 70, 88, relative to 65, 72, 81 Char 91 moved E 92, 93, 99 Char 26 moved L relative to 93 Char 38 moved E relative to 66, 69, Char 96 moved E relative to 15, 24 70,73,88,92,93, relative to 88 Char 91 moved E 99 relative to 66, 93 Char 33 moved L Char 96 moved E relative to 16, 72 relative to 6, 88 Char 40 moved L relative to 5, 6 Char 59 moved E relative to 85, 89 Char 68 moved E relative to 12, 25 Char 71 moved E relative to 12, 29 Char 75 moved L relative to 23, 92 Char 82 moved E relative to 41, 42, 50 Char 91 moved E relative to 69, 87,93 Char 96 moved E relative to 83, 88, 92, 98, 99 Char 97 moved E relative to 1, 3, 41, 42,50 Node 32 --> Twins (26, 13) Twins (25, 12) (26, Twins (26, 13) Node 30 Char 57 moved L 13) Char 57 moved L relative to 4, 19,28,' Char 32 moved E - relative to 4, 19, 31,38 relative to 40, 83, 84 28,31,38 Char 74 moved L Char 57 moved L Char 74 moved L relative to 4, 7, 14 relative to 4, 19,24, relative to 4, 7

219 Char 87 moved E 28,31,38 relative to 58, 73 Char 58 moved L Char 91 moved E relative to 66, 87, relative to 58, 69, 73 91,93 Char 96 moved E Char 73 moved L relative to 2, 58, 73 relative to 70, 91, 93 Char 97 moved E Char 74 moved L relative to 1,2, 42, relative to 4, 7, 23, 50 86 Node 30 --> Char 14 moved L Twins (29, 20) Char 51 moved L Node 29 relative to 86, 90,95 Char 51 moved L relative to 7 Char 25 moved L relative to 7, 13 Char 74 moved L relative to 12, 68 Char 69 moved E Char 29 moved L relative to 58, 73 relative to 68, 71 Char 74 moved L Char 51 moved L relative to 13, 70, 87 relative to 7, 14, 86, Char 75 moved E 90,91,95 relative to 57, 92,93 Char 74 moved L Char 83 moved L relative to 75, 86, 90, relative to 1, 2, 97 95 Char 86 moved E relative to 4, 14 Char 90 moved E relative to 4, 14, 15, 57 Char 91 moved E relative to 7, 14 Char 95 moved E relative to 4, 14, 57 Node 29 --> Twins (16, 47) (44, Char 1 moved E Char 1 moved E Sterna hirundo 49) relative to 26, 66, relative to 26, 66, Char 1 moved E 88,92 88 relative to 26, 41,42, Char 5 moved E Char 5 moved E 66, 83, 84, 88 relative to 2, 26, 58, relative to 26, 58, Char 4 moved L 70, 87,92, 93 70, 93 relative to 15,86,90, Char 12 moved E Char 12 moved E 95 relative to 71, 85, relative to 85, 89, Char 5 moved E 89,94 94 relative to 26,40, 58, Char 30 moved L Char 30 moved L 70,88,93 relative to 20, 21, 22 relative to 21, 22 Char 12 moved E Char 57 moved L Char 57 moved L relative to 85, 89, 94 relative to 7, 13, 14 relative to 7, 13, Char 17 moved E Char 98 moved L 14 relative to 65, 72, 81 relative to 40, 41,42 Char 98 moved L Char 30 moved L relative to 40, 41, relative to 21, 22, 25 42 Char 31 moved L relative to 15,38,86, 90,95 Char 57 moved L relative to 7, 13, 14,

220 15,24,75,90,95 Char 59 moved E relative to 89, 94 Char 66 moved E relative to 58, 73, 87 Char 69 moved E relative to 26, 58, 73, 87,93 Char 70 moved E relative to 73, 88 Char 74 moved L relative to 13, 87 Char 91 moved E relative to 7, 14 Char 97 moved L relative to 98 Char 98 moved L relative to 40, 41,42 Char 99 moved L relative to 97 Node 29 --> Twins (74,23) Twins (25, 30) Char 5 moved L Larus argentatus Char 5 moved L Char 5 moved L relative to 98 relative to 87, 92, 96, relative to 88, 98 Char 75 moved E 98 Char 75 moved E relative to 7, 14 Char 32 moved E relative to 7, 14, 15, relative to 6, 40 91 Char 71 moved E relative to 12, 22 Char 75 moved E relative to 7, 14 Node 30 -> Twins (24, 31) Char 7 moved E Char 7 moved E Phalacrocorax Char 7 moved E relative to 86, 90, 95 relative to 86, 90, auritus relative to 86, 90, 95 Char 20 moved L 95 Char 20 moved L relative to 18,28 Char 20 moved L relative to 18,28 Char 50 moved E relative to 18,28 Char 50 moved E relative to 1, 3, 41, Char 50 moved E relative to 1, 3, 6, 41, 42 relative to 1, 3, 42 Char 51 moved E 41,42 Char 51 moved E relative to 15, 74 Char 51 moved E relative to 15, 74 Char 69 moved L relative to 15, 74 Char 58 moved L relative to 26, 70, 96 Char 69 moved L relative to 88, 93, 99 Char 75 moved L relative to 70, 96 Char 62 moved L relative to 13, 96 Char 75 moved L relative to 89, 94 Char 82 moved E relative to 13, 96 Char 66 moved L relative to 1, 13, 96, Char 82 moved E relative to 88, 92, 93, 97, 98 relative to 1, 13, 96, 99 Char 84 moved E 96, 97, 98 Char 69 moved L relative to 1,97,98 Char 84 moved E relative to 70, 88, 96, relative to 98 99 Char 70 moved L

221 relative to 96, 99 Char 73 moved L relative to 88, 92,93, 99 Char 75 moved L relative to 13,87,88, 92, 93,96,99 Char 82 moved E relative to 1,3, 13, 41,42,50,96,97,98 Char 83 moved E relative to 26, 82, 98 Char 84 moved E relative to 6, 98 Node 32 --> Char 12 moved L Char 2 moved E Char 12 moved L Node 31 relative to 20, 22,29, relative to 75, 88, relative to 20, 22, 57,71 92, 93, 99 29 Char 13 moved L Char 12 moved L Char 13 moved L relative to 6, 58, 66, relative to 20, 22, 29 relative to 6, 58, 70, 87, 88, 92, 93, Char 13 moved L 66,87,88,92, 96,99 relative to 1, 6, 58, 93,99 Char 18 moved L 66,87,88,92,93, • Char 18 moved L relative to 28, 38, 57 98,99 relative to 38 Char 31 moved E Char 18 moved L Char 31 moved E relative to 19,24 relative to 19,24,38 relative to 19 Char 32 moved L Char 26 moved L Char 32 moved L relative to 1,3, 8, 9 relative to 1, 6 relative to 1, 8 Char 59 moved E Char 31 moved E Char 73 moved E relative to 89, 94 relative to 19,28,38 relative to 7, 86, Char 73 moved E Char 32 moved L 95 relative to 7, 86, 95 relative to 1,8, 16, Char 75 moved L 41, 72, 82 relative to 23, 58, 70, Char 69 moved E 87,93 relative to 75, 95 Char 70 moved E relative to 75, 93 Char 73 moved E relative to 7, 14, 66, 75,86,87,90,95 Char 74 moved E relative to 15,90,95 Char 83 moved L relative to 40, 41 Char 84 moved L relative to 3, 6 Node 31 --> Char 1 moved E Char 1 moved E Char 1 moved E Larus relative to 6, 13, 26, relative to 6, 88, 97, relative to 6, 98 ridibundus 84,98 98 Char 9 moved E (Schumacher & Char 3 moved E Char 4 moved L relative to 97, 98 Wolff) relative to 84 relative to 15,51,73 Char 83 moved E Char 9 moved E Char 7 moved E relative to 13, 26,

222 relative to 82, 97, 98 relative to 14,86,90 70 Char 32 moved L Char 9 moved E Char 95 moved L relative to 41, 82 relative to 97, 98 relative to 87 Char 83 moved E Char 23 moved L relative to 13,26,70, relative to 15,51,73 96, 98 Char 58 moved E Char 84 moved E relative to 86, 90 relative to 83 Char 83 moved E Char 95 moved L relative to 13, 26, relative to 86, 87 69, 70, 75, 88, 92, 93,99 Char 92 moved E relative to 88, 99 Char 93 moved E relative to 92 Char 95 moved L relative to 87, 91 Node 31 --> Twins (13,98) (28, Char 18 moved L Char 18 moved L Larus canus 31) relative to 4, 23 relative to 4, 23 (Schumacher & Char 6 moved E Char 66 moved L Char 66 moved L Wolff) relative to 26, 40 relative to 2, 70, 88, relative to 2, 70, Char 18 moved L 93, 99 88, 93, 99 relative to 4, 19,23, Char 69 moved E Char 69 moved E 24 relative to 7, 58, 86, relative to 7, 58, Char 66 moved L 87,91 86,87 relative to 2, 70, 88, Char 74 moved E Char 74 moved E 93, 96, 99 relative to 19,38 relative to 19, 38 Char 69 moved E Char 75 moved L Char 75 moved L relative to 7, 58, 86, relative to 88, 96, 99 relative to 88, 96, 87,95 Char 83 moved L 99 Char 74 moved E relative to 1, 2, 3, Char 83 moved L relative to 15, 19, 38 84,97,98 relative to 1, 2, 3 Char 75 moved L relative to 88, 96, 99 Char 83 moved L relative to 1,2,3,40, 41 Char 92 moved L relative to 2, 70, 96 Node 34 --> Twins (16, 11) Twins (25, 30) (29, Char 2 moved L Cairina Char 2 moved L 20)(59, 94) relative to 84, 96, moschata relative to 84, 96, 98 Char 2 moved L 98 Char 5 moved E relative to 84, 96, 98 Char 5 moved E relative to 58, 66, 69, Char 5 moved E relative to 58, 66, 86, 88, 90, 93, 95, 99 relative to 40, 58, 69, 86, 90, 93, Char 6 moved E 66, 69, 86, 90, 93, 95,99 relative to 58, 66, 69, 95, 99 Char 6 moved E 86, 90, 93, 95 Char 6 moved E relative to 58, 69, Char 8 moved L relative to 40, 58, 86, 90, 95 relative to 10, 33 69, 86, 90, 95 Char 8 moved L Char 9 moved E Char 8 moved L relative to 10, 33

223 relative to 45, 50, 82 relative to 10, 33 Char 9 moved E Char 15 moved E Char 9 moved E relative to 45, 50, relative to 19, 24, 28, relative to 45, 50, 82 82 31,57 Char 15 moved E Char 15 moved E Char 29 moved L relative to 19,24, relative to 19,24, relative to 20, 68 28,31,57 28,31,57 Char 40 moved L Char 44 moved E Char 44 moved E relative to 96, 98 relative to 47, 48 relative to 47, 48 Char 44 moved E Char 46 moved L Char 46 moved L relative to 47,48, 72 relative to 8, 10 relative to 8, 10 Char 46 moved L Char 51 moved L Char 51 moved L relative to 8, 10 relative to 14, 75 relative to 14, 75 Char 51 moved L Char 67 moved E Char 67 moved E relative to 14, 75 relative to 43, 50 relative to 43, 50 Char 67 moved E Char 73 moved L Char 73 moved L relative to 43, 50 relative to 58, 69, relative to 58, 69, Char 73 moved L 70,83,92 70, 83, 92 relative to 26, 58, 66, Char 87 moved E Char 87 moved E 69, 70, 83, 92 relative to 7, 58 relative to 7, 58 Char 87 moved E Char 91 moved L Char 91 moved L relative to 7, 58 relative to 70, 83 relative to 70, 83 Char 91 moved L Char 97 moved L Char 97 moved L relative to 70, 83 relative to 33,43, 82 relative to 33, 43, Char 96 moved L 82 relative to 83, 98 Char 97 moved L relative to 33, 43, 82 Node 35 --> Twins (76, 62) Char 7 moved E Char 7 moved E Stercorarius Char 7 moved E relative to 14, 23, relative to 14, 23, skua relative to 14,23, 74, 74, 86, 90, 95 74, 86,90, 95 86, 90, 95 Char 13 moved L Char 13 moved L Char 13 moved L relative to 1, 2, 6, relative to 58, 66, relative to 58, 66, 69, 58, 66, 69, 70, 87, 69, 70, 87, 92, 96 70,87,92,96 92,96 Char 26 moved L Char 26 moved L Char 26 moved L relative to 69, 70, relative to 69, 70, 92 relative to 1, 2, 69, 92 Char 29 moved L 70,92 Char 29 moved L relative to 30 Char 29 moved L relative to 30 Char 30 moved L relative to 30 Char 30 moved L relative to 12, 19,20, Char 30 moved L relative to 12, 19, 21,22,25 relative to 12, 19, 20,21,22 Char 50 moved L 20,21,22 Char 50 moved L relative to 45, 97 Char 50 moved L relative to 45, 97 Char 51 moved E relative to 45, 97 Char 51 moved E relative to 15,23,24, Char 51 moved E relative to 15, 23, 28,31,74 relative to 15, 23, 24,28,31,74 Char 72 moved E 24,28,31,74 Char 72 moved E relative to 97, 98 Char 66 moved L relative to 97, 98 Char 73 moved E relative to 1, 2 Char 73 moved E relative to 86, 90, 95 Char 70 moved L relative to 86, 90, Char 75 moved E relative to 2, 6 95

224 relative to 15, 31 Char 72 moved E Char 75 moved E Char 82 moved L relative to 97, 98 relative to 15, 31 relative to 45, 83 Char 73 moved E Char 82 moved L Char 83 moved L relative to 86, 90, 95 relative to 45, 83 relative to 3,40,41 Char 75 moved E Char 83 moved L Char 84 moved L relative to 15, 31 relative to 3, 40, relative to 3,40 Char 82 moved L 41 Char 85 moved L relative to 45, 83 Char 84 moved L relative to 12,20,21, Char 83 moved L relative to 3, 40 22, 25, 29, 59 relative to 1, 2, 3, Char 85 moved L Char 88 moved L 40, 41 relative to 12, 20, relative to 40, 70, 92, Char 84 moved L 21,22,25,29,59 96 relative to 1,3,40 Char 88 moved L Char 92 moved L Char 85 moved L relative to 40, 92, relative to 93 relative to 12, 20, 96 Char 93 moved L 21,22,25,29,59 Char 92 moved L relative to 40,41,42, Char 88 moved L relative to 93 69,70,91,96 relative to 1,2,40, Char 93 moved L Char 94 moved L 92,96 relative to 40, 41, relative to 12,25,59 Char 92 moved L 42,70,91,96 Char 99 moved L relative to 93 Char 94 moved L relative to 3,40, 41, Char 93 moved L relative to 12, 25 42, 45, 92, 96 relative to 1,2,40, Char 99 moved L 41,42,70,91,96 relative to 3, 40, Char 94 moved L 41,42,45,92,96 relative to 12, 25 Char 99 moved L relative to 1, 2, 3, 40,41,42,45,92, 96 Char 28 moved E Char 4 moved E Char 32 moved E relative to 68, 71 relative to 18,31,74 relative to 40, 73 Char 32 moved E Char 16 moved E Char 57 moved L relative to 40, 73 relative to 43, 46 relative to 15, 19, Char 57 moved L Char 23 moved E 31,51 relative to 15,19,23, relative to 74, 75 Char 96 moved L 24,31,51 Char 32 moved E relative to 3, 45, Char 96 moved L relative to 40, 73, 50,80 relative to 3, 45, 46, 93,99 Char 98 moved L 50, 80 Char 57 moved L relative to 10, 33, Char 98 moved L relative to 7, 15, 19, 44, 48, 50, 80 relative to 8, 10,33, 31,51 44, 46, 48, 50, 80 Char 59 moved L relative to 12, 19, 21,22,25,29,30 Char 69 moved L relative to 3,40, 41, 42, 45, 50 Char 70 moved L relative to 1, 3, 42, 45,50

225 Char 71 moved L relative to 14, 31 Char 87 moved L relative to 1,2, 3, 10,40,41,42,43, 45, 47, 50,52, 54, 69,80,81 Char 88 moved E relative to 91, 99 Char 92 moved L relative to 1,3, 10, 42, 45, 50, 52, 80 Char 96 moved L relative to 3,43, 45, 50,80 Char 97 moved L relative to 33, 43, 47, 67, 98 Char 98 moved L relative to 10, 33, 43,44,47,48,50, 80 Node 37 --> Char 4 moved E Char 9 moved L Char 9 moved L Node 36 relative to 31, 74 relative to 10, 11,33 relative to 10, 11 Char 6 moved E Char 31 moved L Char 32 moved E relative to 40, 58 relative to 15, 23 relative to 13 Char 9 moved L Char 32 moved E Char 52 moved E relative to 10, 11 relative to 13, 91 relative to 1, 10, Char 23 moved E Char 52 moved E 11, 16,33,43, relative to 31, 74, 75 relative to 1, 3, 8, 44, 45, 47, 49 Char 32 moved E 10, 11,16,33,42, Char 54 moved E relative to 13, 14,38 43, 44, 45,47, 49, relative to 17, 44, Char 52 moved E 50, 97, 98 48,49 relative to 1, 10, 11, Char 54 moved E Char 70 moved L 16,33,43,44,45, relative to 17,44, relative to 9, 10, 46, 47, 49, 65 48,49,97,98 11,43,44,96 . Char 54 moved E Char 61 moved L Char 71 moved L relative to 17, 44, 48, relative to 12, 25, 30 relative to 1, 3, 7, 49, 65 Char 62 moved L 26,38,40,41, Char 59 moved L relative to 12, 25, 30 42, 45, 50, 57, relative to 19,21,22, Char 70 moved L 58, 80 25,28,29,30 relative to 9, 10, 11, Char 72 moved L Char 69 moved L 33, 43, 44,47, 96, relative to 43, 44 relative to 2, 3, 5, 40, 98 Char 81 moved E 41,42,45,46,50,99 Char 71 moved L relative to 48, 49 Char 70 moved L relative to 1, 3, 7, Char 96 moved L relative to 1, 2, 3, 5, 16,26,38,40,41, relative to 10, 11 9, 10, 11,42,43,44, 42, 45, 50, 57, 58, 45, 46, 50, 96, 99 80 Char 71 moved L Char 72 moved L relative to 1, 2, 3, 5, relative to 33, 43,

226 7,13,14,24,26,31, 44, 48 38,40,41,42,45, Char 81 moved E 50, 57, 58, 73, 80, 93 relative to 44, 48, Char 72 moved L 49, 97, 98 relative to 43, 44 Char 93 moved E Char 81 moved E relative to 13, 26, 73 relative to 48, 49 Char 96 moved L Char 87 moved L relative to 10, 11 relative to 1,2,3, 10, 11,26,40,41,42, 43, 45, 46, 50, 69, 80 Char 92 moved L relative to 1,3, 10, 42, 45, 46, 50, 80 Char 96 moved L relative to 10, 11,43 Node 36 --> Twins (20,24) (31, Twins (3, 45) (28, Twins (48, 44) Dromaius 15)(46, 8)(48, 44) 30) (48, 44) (96, 9) (96,9) novaehollandiae (96, 9) Char 57 moved E Char 75 moved L (RM) Char 5 moved E relative to 7, 18 relative to 86, 90, relative to 40, 58 Char 75 moved L 95 Char 6 moved E relative to 86, 90, 95 Char 99 moved E relative to 5 Char 99 moved E relative to 13, 26, Char 52 moved E relative to 13, 26, 73,91 relative to 42, 50 73,91 Char 75 moved L relative to 86, 90, 95 Char 88 moved E relative to 6, 91 Char 93 moved E relative to 13, 26, 73 Char 99 moved E relative to 6, 13, 26, 58,73,91 Node 36 -> Twins (1,50) (8, 52) Twins (1,50) Twins (1,50) Dromaius (44, 81) (57, 7) (70, Char 6 moved E Char 6 moved E novaehollandiae 98) relative to 13, 26, relative to 13, 26, (YPM) Char 6 moved E 32,73,86,90,91, 32, 73, 86, 90, relative to 13, 26, 32, 93, 95 91, 93, 95 73,86,90,91,93,95 Char 18 moved E Char 18 moved E Char 18 moved E relative to 31, 68, 90 relative to 31, 90 relative to 31, 90 Char 32 moved E Char 32 moved E Char 29 moved L relative to 86, 88, 95 relative to 86, 88, relative to 30 95 Char 30 moved L relative to 19,28 Char 32 moved E relative to 86, 88, 91, 93,95 Node 37 -> Twins (20, 21) (30, Twins (20, 21) (79, Twins (20, 21) Struthio 28)(79, 65)(81, 17) 65)(81,17) (79, 65)(81, 17)

227 camelus Char 2 moved L Char 2 moved L Char 2 moved L relative to 40, 50 relative to 40, 50 relative to 40, 50 Char 5 moved L Char 5 moved L Char 5 moved L relative to 3, 41, 42 relative to 3, 41, 42 relative to 3, 41, Char 6 moved L Char 6 moved L 42 relative to 1, 3, 41, relative to 1, 3, 41, Char 6 moved L 42,50 42,50 relative to 1, 3, Char 57 moved L Char 72 moved E 41,42,50 relative to 7, 91 relative to 47, 49, Char 72 moved E Char 67 moved L 97,98 relative to 47, 49, relative to 16, 46 Char 73 moved L 98 Char 72 moved E relative to 26, 38 Char 73 moved L relative to 47, 49, 98 Char 93 moved L relative to 26, 38 Char 73 moved L relative to 40, 41,42 Char 93 moved L relative to 26, 38 Char 99 moved L relative to 40, 41, Char 93 moved L relative to 1, 3,40, 42 relative to 32, 40,41, 41,42,45,97 Char 99 moved L 42,91 relative to 1,3, Char 97 moved L 40,41,42,45,97 relative to 67, 98 Char 99 moved L relative to 1, 3, 32, 40,41,42,45,97 E = early; L = late. See Legend for bone identities.

228 Table 6.2: ACCTRAN, DELTRAN and consensus reconstructions of hypothesized sequence changes on an accepted avian phylogeny.

Node ACCTRAN DELTRAN Consensus Root --> Twins (30,25) (52, Char 14 moved E Char 14 moved E Neognathae 48) (72, 67) (76, 61) relative to 74, 75, relative to 74, 75, Char 1 moved E 86,90,91,95 .86,90,91,95 relative to 3,42 Char 18 moved E Char 18 moved E Char 9 moved E relative to 23, 31,90 relative to 23, 31, relative to 43, 46 Char 28 moved L 90 Char 14 moved E relative to 21, 22 Char 28 moved L relative to 74, 75, 86, Char 32 moved L relative to 21, 22 90,95 relative to 2, 6, 26 Char 32 moved L Char 16 moved L Char 38 moved E relative to 2, 6 relative to 8, 10, 11 relative to 51, 74, Char 38 moved E Char 18 moved E 75, 86,90, 95 relative to 51, 74, relative to 23, 24,31, Char 57 moved E 75, 86, 90, 95 90 relative to 74, 75, Char 57 moved E Char 19 moved E 90, 95 relative to 74, 75, relative to 18 Char 71 moved E 90, 95 Char 28 moved L relative to 74, 75, Char 71 moved E relative to 21, 22 86,90,91,95 relative to 74, 75, Char 32 moved L Char 80 moved L 86, 90, 95 relative to 6, 26 relative to 1, 3, 45, Char 80 moved L Char 38 moved E 50, 69, 70 relative to 45, 50, relative to 51, 74, 75, 70 86, 90, 95 Char 57 moved E relative to 74, 75, 90, 95 Char 62 moved E relative to 89, 94 Char 71 moved E relative to 74, 75, 86, 90,95 Char 80 moved L relative to 10,45,46, 50, 70 Neognathae —> Twins (32, 5) (50, Twins (4, 57) (16, Twins (32, 5) Galloanseres 67) (72, 33) (78, 61) 10) (18, 19) (31, 38) Char 3 moved E Char 3 moved E (32, 5) relative to 41, 42 relative to 41, 42 Char 3 moved E Char 7 moved L Char 7 moved L relative to 41, 42, 50 relative to 87 relative to 69, 70, 87 Char 7 moved L Char 23 moved L Char 23 moved L relative to 58, 87 relative to 13, 14 relative to 13, 14,86, Char 23 moved L Char 91 moved L 95 relative to 13, 14, 15 relative to 69 Char 26 moved E Char 72 moved L relative to 88, 93 relative to 33, 67 Char 31 moved L Char 88 moved L relative to 38, 51 relative to 26, 58, 73

229 Char 66 moved E Char 91 moved L relative to 87, 92 relative to 26, 58, Char 74 moved L 69,73 relative to 86, 90,95 Char 93 moved L Char 75 moved E relative to 26, 58 relative to 86, 95 Char 91 moved L relative to 69, 99 Char 96 moved L relative to 83, 98 Galloanseres —> Twins (7,99) (22, Twins (1,98) (74, Char 15 moved E Anseriformes 68) (25, 30) (97, 67) 75) . relative to 19,38, Char 1 moved L Char 15 moved E 57 relative to 3,42, 98 relative to 19, 31, Char 33 moved E Char 2 moved L 38,57 ' relative to 11, 16 relative to 40, 98 Char 33 moved E Char 82 moved L Char 4 moved L relative to 11, 16 relative to 41, 42, relative to 14, 74 Char 52 moved L 50 Char 10 moved E relative to 10, 11,48 relative to 8, 46 Char 73 moved L Char 13 moved E relative to 26, 66 relative to 26, 75, 91 Char 82 moved L Char 15 moved E relative to 41, 42, 50 relative to 19,24,38, 57 Char 33 moved E relative to 8, 11, 16 Char 44 moved E relative to 47, 48 Char 70 moved E relative to 91, 92 Char 82 moved L relative to 3, 9, 41, 42, 45, 50 Char 87 moved E relative to 69, 91 Anseriformes —> Twins (79, 65) Twins (79, 65) (88, Twins (79, 65) Somateria Char 5 moved L 69) Char 5 moved L mollissima relative to 2, 3, 41, Char 5 moved L relative to 2, 3, 41, 42, 84, 88, 96, 98 relative to 2, 3,41, 42, 84, 96, 98 Char 6 moved L 42, 84, 96, 98 Char 6 moved L relative to 3,41,42, Char 6 moved L relative to 3, 41, 96, 98 relative to 2, 3, 41, 42, 96, 98 Char 7 moved E 42, 96, 98 Char 8 moved E relative to 69, 70 Char 8 moved E relative to 43, 80 Char 8 moved E relative to 43, 80 Char 9 moved E relative to 33, 43, 80 Char 9 moved E relative to 8 Char 9 moved E relative to 8 Char 32 moved L relative to 8 Char 32 moved L relative to 41, 42 Char 32 moved L relative to 1,41,42 Char 46 moved E relative to 41, 42 Char 46 moved E relative to 9

230 Char 44 moved L relative to 9 relative to 47,48,49, Char 73 moved L 81 relative to 7,93, 99 Char 46 moved E Char 87 moved L relative to 9 relative to 7, 66, 99 Char 97 moved E Char 91 moved L relative to 45, 82 relative to 7, 99 Anseriformes — > Twins (48,49) Twins (16, 11) Char 5 moved E Cairina moschata Char 5 moved E Char 5 moved E relative to 58, 66, relative to 26, 58, 66, relative to 40, 58, 69, 83, 86, 87, 92, 69, 83, 86, 87, 92, 66, 69, 83, 86, 87, 95,99 93,95,99 88, 92, 95, 99 Char 6 moved E Char 6 moved E Char 6 moved E relative to 58, 66, relative to 2, 26, 58, relative to 40, 58, 69, 83, 84, 86, 87, 66, 69, 83, 84, 86, 66, 69, 83, 84, 86, 88, 92, 95, 99 87, 88, 92,93, 95, 99 87,88,92,95,99 Char 51 moved L Char 23 moved E Char 8 moved L relative to 14 relative to 26, 58, 66, relative to 10, 33 Char 73 moved L 86,95 Char 46 moved L relative to 69, 70, Char 29 moved L relative to 8, 10 92 relative to 20, 68 Char 51 moved L Char 97 moved L Char 51 moved L relative to 14, 74 relative to 10, 33, relative to 13, 14,31 Char 73 moved L 46 Char 73 moved L relative to 7, 58, 69, relative to 69, 70, 92 70, 92 Char 74 moved E Char 75 moved E relative to 26, 86, 95 relative to 86, 95 Char 87 moved E Char 91 moved L relative to 26, 66 relative to 7, 70 Char 97 moved L Char 97 moved L relative to 10,33,46 relative to 8, 10,33, 46 Char 98 moved E relative to 2, 40 Anseriformes —> Twins (11, 16) Char 4 moved L Char 4 moved L Anas platyrhynchos Char 4 moved L relative to 14, 15, relative to 51, 58, relative to 51, 58, 66, 51,58,66,73,74, 66, 73, 86, 90, 95 73, 86,90, 95 86, 90, 95 Char 7 moved E Char 7 moved L Char 7 moved L Char 13 moved L relative to 73, 91 relative to 69, 99 relative to 14 Char 13 moved L Char 10 moved E Char 14 moved L relative to 14 relative to 46, 80 relative to 86, 90, Char 14 moved L Char 13 moved L 95 relative to 86, 90, 95 relative to 14 Char 23 moved L Char 23 moved L Char 14 moved L relative to 73 relative to 73, 87 relative to 86, 90, 95 Char 31 moved L Char 31 moved L Char 23 moved L relative to 58, 66, relative to 58, 66, 73, relative to 26, 58, 73, 75, 93, 95 75,93,95 66, 73, 86, 95 Char 40 moved E Char 40 moved E Char 31 moved L relative to 83, 84, relative to 83, 84, 92, relative to 23, 51, 96

231 96 58, 66, 73, 75, 93, Char 50 moved E Char 50 moved E 95 relative to 41, 42 relative to 41,42 Char 40 moved E Char 74 moved L Char 67 moved L relative to 2, 83, 84, relative to 13 relative to 10,48 96 Char 75 moved L Char 74 moved L Char 44 moved E relative to 26, 58, relative to 13 relative to 47,48 66,73 Char 75 moved L Char 50 moved E relative to 26, 58, 66, relative to 1,41,42 73, 86, 87,95 Char 74 moved L Char 88 moved E relative to 13, 26, relative to 69, 99 86, 90, 95 Char 75 moved L relative to 13, 26, 58, 66, 73 Galloanseres --> Twins (19, 31) Twins (78, 61) Char 23 moved L Galliformes Char 7 moved L Char 23 moved L relative to 2, 4, 5, relative to 73, 83, 91, relative to 2, 4, 5, 6, 6,7,70,73,84,87, 92 7, 26, 32, 70, 73, 84, 88,91,92,93,98, Char 13 moved L 86,87,88,91,92, 99 relative to 58, 66, 70 93, 95,98, 99 Char 52 moved E Char 20 moved L Char 24 moved E relative to 9,46, relative to 38, 57 relative to 20, 28 47, 49, 72 Char 23 moved L Char 52 moved E Char 55 moved E relative to 2, 4, 5, 6, relative to 9,46, 47, relative to 54, 81 7, 40, 70, 73, 83, 84, 49, 50, 72 Char 56 moved E 87,88,91,92,93, Char 55 moved E relative to 44, 47, 96, 97, 98,99 relative to 17,54, 81 48, 49, 54, 72, 81 Char 25 moved E Char 56 moved E Char 69 moved E relative to 59, 85, 89 relative to 17, 44, relative to 93, 99 Char 30 moved E 47, 48,49, 54, 72, Char 74 moved L relative to 85, 89 79,81 relative to 91 Char 42 moved L Char 69 moved E Char 86 moved L relative to 45, 50 relative to 7, 88, 93, relative to 26, 73 Char 43 moved L 99 relative to 11, 16 Char 74 moved L Char 52 moved E relative to 90,91, 95 relative to 8,9, 16, Char 86 moved L 45, 46,47, 48, 49, relative to 26, 73 65, 72 Char 53 moved E relative to 65, 72, 81 Char 54 moved E relative to 53 Char 55 moved E relative to 54, 65, 72, 81 Char 56 moved E relative to 44, 47, 48, 49,54,65,72,81 Char 69 moved E

232 relative to 93, 99 Char 73 moved L relative to 74 Char 74 moved L relative to 87, 91,92 Char 86 moved L relative to 26, 58, 66, 73 Char 87 moved L relative to 58, 93 Char 90 moved L relative to 26, 58, 73 Char 95 moved L relative to 26, 58, 66, 73 Char 97 moved L relative to 10,46 Char 98 moved L relative to 83, 84 Galliformes --> Twins (31, 4) (44, Twins (16, 56) Char 13 moved L Gallus spp. 49)(97, 11) Char 13 moved L relative to 84, 88, Char 5 moved E relative to 70, 83, 93, 99 relative to 66, 69, 70, 84, 88, 93, 99 Char 52 moved E 83,86,90,91,92, Char 52 moved E relative to 3 93, 95, 99 relative to 3, 8 Char 96 moved L Char 6 moved E Char 90 moved L relative to 1, 2, 3, relative to 2, 5, 70, relative to 26, 58 82, 84 84, 95 Char 95 moved L Char 13 moved L relative to 26, 58, 75 relative to 2, 84, 88, Char 96 moved L 93,99 relative to 1, 2, 3, Char 16 moved L 82, 83, 84, 98 relative to 43, 47, 56 Char 18 moved E relative to 68, 71 Char 32 moved E relative to 1,2,40, 70,83,84,86,91, 92, 95, 98 Char 45 moved L relative to 9, 42 Char 51 moved E relative to 15, 38 Char 52 moved E relative to 1, 3 Char 83 moved E relative to 91, 92 Char 87 moved L relative to 88, 99 Char 96 moved L relative to 1, 2, 3, 9,

233 40,41,42,50,82,84 Gallus sp. --> Char 7 moved E Char 4 moved L Char 7 moved E Gallus gallus relative to 66, 69, 70, relative to 74 relative to 66, 69, (Schumacher & 75, 83, 86, 87, 90, Char 5 moved E 87,93 Wolff) 91,92,93,95 relative to 66, 69, Char 12 moved L Char 12 moved L 87, 88, 92, 93, 99 relative to 22, 29, relative to 22, 29, 68, Char 6 moved E 68,71 71 relative to 5 Char 21 moved L Char 21 moved L Char 7 moved E relative to 68, 71 relative to 68, 71 relative to 66, 69, Char 23 moved E 87,93 relative to 96, 97 Char 12 moved L Char 26 moved E relative to 22, 29, relative to 75 68,71 Char 66 moved E Char 13 moved L relative to 75 relative to 2, 87, 92 Char 69 moved E Char 14 moved L relative to 75, 90, 95 relative to 74 Char 73 moved E Char 15 moved L relative to 14,26,31, relative to 74 58, 75 Char 19 moved L Char 74 moved E relative to 74 relative to 14, 15, 19, Char 21 moved L 26, 69, 86, 87, 90, relative to 24, 68, 71 91,92,95 Char 70 moved L Char 75 moved E relative to 32, 83 relative to 74 Char 74 moved L Char 83 moved E relative to 75 relative to 70, 75,95 Char 75 moved L Char 86 moved E relative to 26, 58, relative to 75, 95 66,83,87,88,91, Char 87 moved E 92, 93, 99 relative to 75, 95 Char 83 moved L Char 88 moved E relative to 84 relative to 75 Char 84 moved L Char 91 moved E relative to 2, 32 relative to 75, 95 Char 90 moved L Char 92 moved E relative to 74 relative to 75 Char 91 moved L Char 93 moved E relative to 74 relative to 75 Char 95 moved L Char 98 moved E relative to 32, 66, relative to 2, 70 74, 83, 86, 87, 91 Char 99 moved E relative to 75 Gallus spp. —> Twins (90, 75) (97, Twins (16, 47) (44, Char 7 moved L Gallus gallus 8) 49) relative to 88, 99 (Chapter 3) Char 7 moved L Char 4 moved E Char 12 moved E relative to 32, 88, 99 relative to 31, 57 relative to 59, 61, Char 12 moved E Char 7 moved L 62,94 relative to 59, 61,62, relative to 70, 83, Char 25 moved E

234 94 88, 92, 99 relative to 61, 62, Char 25 moved E Char 9 moved E 77,78 relative to 61, 62, 77, relative to 45, 50 Char 30 moved E 78 Char 12 moved E relative to 61, 62, Char 30 moved E relative to 59, 61, 78 relative to 61, 62, 78 62, 89,94 Char 69 moved L Char 69 moved L Char 23 moved L relative 83, 88, 93, relative to 32, 83, 86, relative to 96, 97 99 88, 93, 99 Char 25 moved E Char 85 moved L Char 74 moved L relative to 59, 61, relative to 18, 21, relative to 31, 66, 83, 62, 77, 78, 89 22, 29 93,99 Char 30 moved E Char 85 moved L relative to 61, 62, relative to 18,21,22, 78, 89 29 Char 32 moved E relative to 40, 83, 86 Char 52 moved E relative to 1,41,45 Char 59 moved E relative to 89, 94 Char 68 moved L relative to 18,20, 24,28 Char 69 moved L relative to 2, 13, 70, 83, 88,93, 99 Char 71 moved L relative to 20, 24, 28 Char 73 moved L relative to 26, 74 Char 74 moved L relative to 13, 26, 31,66,69,70,83, 86,87,92,93,99 Char 75 moved E relative to 86, 90 Char 85 moved L relative to 18, 20, 21,22,29 Char 87 moved L relative to 2, 88, 99 Char 91 moved L relative to 13, 70, 83,99 Char 92 moved L relative to 2, 70, 83 Char 96 moved L relative to 40, 41,42 Char 97 moved L relative to 8, 11,46 Galliformes —> Twins (5, 2) (24, 21) Char 2 moved E Char 18 moved L

235 Meleagris Char 1 moved L relative to 5, 6 relative to 19,38, gallopavo relative to 3, 82 Char 18 moved L 57 Char 18 moved L relative to 4, 19,31, Char 25 moved E relative to 19, 38, 57 38,57,71 relative to 62, 77 Char 25 moved E Char 25 moved E Char 32 moved L relative to 62, 77 relative to 61, 62, relative to 3, 9, 41, Char 32 moved L 77, 78, 85, 89 42,97 relative to 3,9, 16, Char 26 moved E Char 69 moved E 41,42,82,97 relative to 73, 90, 95 relative to 58, 66, Char 67 moved L Char 30 moved E 75,95 relative to 10, 16,43, relative to 78, 85, 89 Char 72 moved E 46, 48, 50 Char 32 moved L relative to 33, 47, Char 69 moved E relative to 1,3,8,9, 49 relative to 58, 66, 75, 41,42,97 Char 74 moved L 95 Char 52 moved E relative to 83, 84, Char 72 moved E relative to 41, 45 93,99 relative to 33,43,47, Char 67 moved L Char 98 moved L 49,53,55 relative to 10,43, relative to 3 Char 73 moved L 46,48 relative to 74 Char 69 moved E Char 74 moved L relative to 51, 58, relative to 83, 84, 93, 66, 73, 75, 86, 90, 99 95 Char 98 moved L Char 72 moved E relative to 3, 52, 82 relative to 33,47,49 Char 74 moved L relative to 13, 31, 83, 84, 86, 87, 92, 93, 99 Char 75 moved E relative to 86, 90, 95 Char 97 moved L relative to 8, 9, 10, 46 Char 98 moved L relative to 3, 84 Galliformes ~> Twins (51, 14) Char 21 moved L Char 21 moved L Coturnix sp. Char 13 moved L relative to 24, 68, 71 Char 97 moved E relative to 87, 92 Char 23 moved L relative to 3, 42, 50 Char 21 moved L relative to 40, 96 relative to 28, 38, 57 Char 97 moved E Char 41 moved E relative to 3, 42, 50 relative to 3, 52 Char 45 moved E relative to 3, 50, 52 Char 88 moved L relative to 2, 40, 70, 83,92 Char 92 moved L relative to 93 Char 93 moved L

236 relative to 40, 70, 83, 87 Char 96 moved E relative to 70, 83, 87, 98 Char 97 moved E relative to 3, 10,41, 42, 45, 46, 50, 52, 82 Char 99 moved L relative to 2,40, 70, 83, 92 Coturnix sp. --> Twins (7, 84) (74, Char 7 moved L Char 38 moved E Coturnix coturnix 70) relative to 70, 83,92 relative to 57 (Chapter 3) Char 31 moved E Char 13 moved L relative to 18,24, 28 relative to 70, 87,92 Char 3 8 moved E Char 18 moved L relative to 18, 28, 57 relative to 31 Char 75 moved E Char 19 moved L relative to 31, 51 relative to 31 Char 86 moved E Char 24 moved L relative to 31 relative to 31 Char 93 moved E Char 28 moved L relative to 31 relative to 31 Char 31 moved L relative to 51, 58, 66, 73, 75, 86, 90, 91,93,95 Char 38 moved E relative to 20, 21,57 Char 51 moved L relative to 14, 75 Char 73 moved L relative to 74 Char 74 moved L relative to 26, 58, 66, 69, 70, 86, 87, 92 Char 97 moved E relative to 23, 32, 41,45,82 Coturnix sp. --> Char 1 moved E Twins (43, 16) Char 4 moved L Coturnix coturnix relative to 88, 99 Char 3 moved L relative to 14, 51 (Nakane & Char 4 moved L relative to 1, 41 Char 84 moved L Tsudzuki) relative to 14, 51 Char 4 moved L relative to 2, 40 Char 32 moved E relative to 14, 15, Char 93 moved L relative to 88, 98, 99 51,58,66,73,74, relative to 2, 7 Char 58 moved E 90,95 Char 96 moved E relative to 75 Char 20 moved L Char 66 moved E relative to 18, 57 relative to 75 Char 21 moved L Char 73 moved E relative to 18, 28,57

237 relative to 75 Char 26 moved L Char 74 moved E relative to 58, 66 relative to 26, 69, 86, Char 45 moved E 87,92 relative to 42, 50 Char 75 moved E Char 67 moved E relative to 74 relative to 9, 50 Char 84 moved L Char 72 moved L relative to 2,40 relative to 53, 55 Char 93 moved L Char 74 moved L relative to 2, 7 relative to 75 Char 96 moved E Char 75 moved L relative to 75 relative to 58, 66, 73 Char 83 moved L relative to 84 Char 84 moved L relative to 1,2,40 Char 88 moved L relative to 1,2,40, 70, 83, 92 Char 92 moved L relative to 93 Char 93 moved L relative to 1, 2, 7, 40, 70, 83, 87 Char 96 moved E relative to 13, 70, 87 Char 99 moved L relative to 1,2,40, 70, 83, 92 Neognathae --> Twins (44, 49) (67, Twins (6, 2) (50, 97) Char 7 moved E Neoaves 43X76,62), Char 7 moved E relative to 66, 75, Char 7 moved E relative to 13, 66, 93 relative to 66, 75, 86, 73,75,91,93 Char 26 moved L 93 Char 26 moved L relative to 58, 70 Char 17 moved E relative to 58, 70, relative to 65, 72, 81 87,92 Char 24 moved L Char 59 moved E relative to 38, 57 relative to 30, 85 Char 26 moved L relative to 58, 66, 70, 99 Char 30 moved L relative to 12, 21, 22, 59 Char 32 moved L relative to 8, 9, 16 Char 85 moved L relative to 12, 59 Char 91 moved E relative to 75, 88, 93 Char 92 moved E

238 relative to 88, 99 Char 96 moved E relative to 5, 6, 66, 88,99 Char 97 moved E relative to 1, 3, 41, 42, 45, 50 Neoaves --> Twins (18, 19) Twins (28, 2.0) (67, Char 13 moved L Charadriiformes Char 1 moved E 43)(90,23) relative to 66 relative to 66, 88, 99 Char 13 moved L Char 83 moved L Char 2 moved E relative to 6, 66, 96 relative to 40, 41 relative to 88, 92, 93, Char 83 moved L Char 84 moved L 96,99 relative to 2,40, 41 Char 13 moved L Char 84 moved L relative to 58, 66 relative to 2, 6 Char 20 moved E Char 88 moved L relative to 29, 30 relative to 58, 70, Char 21 moved E 73,91,96 relative to 20 Char 93 moved L Char 69 moved E relative to 58, 70, relative to 58, 66, 93, 87,91 99 Char 99 moved L Char 70 moved E relative to 70, 96 relative to 88, 92,93, 99 Char 73 moved E relative to 86, 95 Char 82 moved L relative to 3, 9, 41, 42, 43,50,97 Char 83 moved L relative to 3, 40, 41, 97 Char 84 moved L relative to 3, 97 Char 97 moved L relative to 98 Char 98 moved L relative to 41, 42 Charadriiformes —> Twins (82, 83) Twins (75, 91) (89, Char 68 moved E Sterna + Larus Char 1 moved E 59) relative to 21, 22, relative to 40, 41 Char 26 moved L 25,29 Char 7 moved L relative to 66, 96 Char 71 moved E relative to 86, 95 Char 68 moved E relative to 29 Char 18 moved L relative to 21, 22, Char 92 moved L relative to 24, 38 25,29 relative to 66 Char 51 moved L Char 71 moved E relative to 14, 15,23, relative to 18, 29 74 Char 92 moved L Char 68 moved E relative to 66, 70 relative to 12, 21, 22, Char 97 moved E

239 25,29 relative to 3,41, 42, Char 71 moved E 45 relative to 21, 22, 29 Char 75 moved L relative to 23, 69 Char 92 moved L relative to 66, 96 Sterna + Larus --> Char 2 moved L Twins (16, 47) (44, Char 4 moved L Sterna hirundo relative to 26, 88, 92, 49)(52, 48) relative to 15, 86, 93,99 Char 1 moved E 90,95 Char 4 moved L relative to 41, 42, Char 5 moved E relative to 15,86,90, 66,88 relative to 70, 93 95 Char 4 moved L Char 12 moved E Char 5 moved E relative to 15, 86, relative to 25, 89 relative to 26, 70, 93 90, 95 Char 14 moved L Char 12 moved E Char 5 moved E relative to 86, 95 relative to 25, 89 relative to 40, 70, Char 31 moved L Char 13 moved E 83, 88, 93 relative to 15, 38, relative to 26, 70 Char 12 moved E 86, 90, 95 Char 14 moved L relative to 25, 85, Char 51 moved L relative to 86, 95 89,94 relative to 7, 86, 95 Char 31 moved L Char 14 moved L Char 57 moved L relative to 15,24,38, relative to 86, 90, 95 relative to 7, 14, 86, 90, 95 Char 17 moved E 15,31,38,75,90, Char 51 moved L relative to 65, 81 95 relative to 7, 86, 95 Char 30 moved L Char 73 moved L Char 57 moved L relative to 21, 22, 25 relative to 70, 87 relative to 7, 14, 15, Char 31 moved L Char 74 moved L 24,28,31,38,75, relative to 15, 38, relative to 70, 86, 90,95 86, 90, 95 87, 90, 95 Char 73 moved L Char 51 moved L relative to 69, 70, 86, relative to 7, 14, 86, 87,95 90,95 Char 74 moved L Char 57 moved L relative to 69, 70, 86, relative to 7, 14, 15, 87, 90, 95 31,38,75,90,95 Char 73 moved L relative to 70, 87 Char 74 moved L relative to 70, 86, 87, 90, 95 Sterna + Larus --> Char 1 moved L Twins (6, 96) (75, Char 12 moved L Larus spp. relative to 66, 88, 99 23) relative to 20, 22, Char 4 moved L Char 12 moved L 29 relative to 74 relative to 20, 22, Char 13 moved L Char 5 moved L 29,68 relative to 88, 99 relative to 2, 88 Char 13 moved L Char 31 moved E Char 7 moved L relative to 2, 58, 70, relative to 19 relative to 73, 74 87, 88, 99 Char 57 moved E Char 12 moved L Char 26 moved L relative to 19 relative to 20, 22, 29, relative to 2, 99

240 71,85,94 Char 31 moved E Char 13 moved L relative to 19,24 relative to 88, 93,99 Char 57 moved E Char 14 moved L relative to 4, 19,24, relative to 73, 74 28 Char 30 moved E Char 73 moved E relative to 20, 21,22, relative to 7, 91 25 Char 74 moved E Char 31 moved E relative to 7, 14,51 relative to 18, 19,28 Char 97 moved E Char 57 moved E relative to 32, 84 relative to 18, 19 Char 58 moved E relative to 87, 91 Char 74 moved L relative to 75 Char 75 moved L relative to 2, 70, 73, 87,93 ' Char 98 moved E relative to 40, 41,42 Larus spp. — > Char 6 moved L Twins (22, 71) (25, Char 6 moved L Larus argentatus relative to 97, 98 12) relative to 97, 98 Char 12 moved E Char 1 moved L Char 32 moved E relative to 25, 57 relative to 26, 97, 98 relative to 40, 83, Char 13 moved E Char 5 moved L 84 relative to 91, 92, 93 relative to 2, 96, 98 Char 51 moved L Char 32 moved E Char 6 moved L relative to 7, 69, 91 relative to 3, 8,9, 16, relative to 97, 98 Char 74 moved L 40, 83, 84 Char 14 moved L relative to 23 Char 51 moved L relative to 86, 95 relative to 7, 69, 91 Char 32 moved E Char 74 moved L relative to 40, 83, 84 relative to 4, 23 Char 51 moved L Char 75 moved E relative to 7, 14,23, relative to 2, 7, 58, 69,86,91,95 69,70,87,91,93 Char 74 moved L relative to 23, 86, 95 Larus spp. —> Char 1 moved E Twins (12, 57) (24, Char 1 moved E Larus ridibundus relative to 6, 13,88, 38) relative to 6, 13, 88 (Schumacher & 97, 98 Char 1 moved E Char 4 moved L Wolff) Char 4 moved L relative to 6, 13, 88 relative to 51 relative to 15, 51, 73 Char 2 moved E Char 9 moved E Char 9 moved E relative to 88, 93, relative to 97, 98 relative to 97, 98 96, 99 Char 32 moved L Char 14 moved L Char 4 moved L relative to 41, 72, relative to 7, 51 relative to 51, 74 82 Char 23 moved L Char 7 moved E Char 83 moved E relative to 15,51,73 relative to 14, 86,90 relative to 1, 2, 13, Char 32 moved L Char 9 moved E 26, 70, 88, 93, 97, relative to 41, 72, 82 relative to 82, 97, 98 98,99

241 Char 69 moved L Char 32 moved L Char 84 moved E relative to 58, 87 relative to 3, 8, 16, relative to 83 Char 83 moved E 41,72,82 Char 95 moved L relative to 1, 2, 3, 6, Char 58 moved E relative to 91 13,26,70,88,93, relative to 86, 87, 97, 98, 99 90,91 Char 84 moved E Char 73 moved E relative to 83 relative to 14, 86, 90 Char 86 moved L Char 74 moved L relative to 7, 58 relative to 75 Char 90 moved L Char 75 moved L relative to 7, 14,51, relative to 66, 69, 70 58,73 Char 83 moved E Char 95 moved L relative to 1, 2, 13, relative to 7, 58, 87, 26, 70, 88,92, 93, 91 96, 97,98,99 Char 84 moved E relative to 83 Char 92 moved E relative to 13, 88, 99 Char 93 moved E relative to 13, 92 Char 95 moved L relative to 66, 86, 91 Lams spp. --> Twins (21, 57) Twins (28,31) Char 18 moved L Larus canus Char 18 moved L Char 1 moved L relative to 4 (Schumacher & relative to 4, 23 relative to 97, 98 Char 66 moved L Wolff) Char 66 moved L Char 4 moved L relative to 2, 70, relative to 2, 6, 70, relative to 74 88, 93, 99 88, 93, 99 Char 6 moved E Char 69 moved E Char 69 moved E relative to 26, 40 relative to 7, 86, relative to 7, 86, 91, Char 13 moved L 91,95 95 relative to 92,93, 98 Char 75 moved L Char 74 moved E Char 15 moved L relative to 88, 96, relative to 15, 19, 38 relative to 74 99 Char 75 moved L Char 18 moved L Char 92 moved L relative to 6, 88, 96, relative to 4 Char 19 moved L 99 relative to 74 Char 92 moved L Char 32 moved L relative to 6, 88, 99 relative to 1,8,9, 16 Char 3 8 moved L relative to 74 Char 57 moved E relative to 12, 21 Char 66 moved L relative to 2, 70, 88, 93, 96, 99 Char 69 moved E relative to 7, 58, 86, 91,95

242 Char 73 moved E relative to 86, 95 Char 74 moved L relative to 75 Char 75 moved L relative to 2, 58, 70, 87, 88, 93,96, 99 Char 92 moved L relative to 2, 96 Charadriiformes --> Char 7 moved E Char 7 moved E Char 7 moved E Stercorarius skua relative to 14, 23 relative to 14, 23, relative to 14, 23 Char 38 moved L 38,86,90,95 Char 38 moved L relative to 14, 15,23, Char 13 moved L relative to 14, 15, 24,74,86,90,95 relative to 1, 2, 58, 23, 74, 86, 90, 95 Char 51 moved E 70, 87, 92 Char 51 moved E relative to 24, 31 Char 26 moved L relative to 31 Char 66 moved L relative to 2, 6 Char 66 moved L relative to 2, 6, 26, Char 29 moved L relative to 2, 6, 41, 41,42,70 relative to 30 42, 70 Char 70 moved L Char 30 moved L Char 70 moved L relative to 2, 6, 41 relative to 12, 21, relative to 2, 6,41 Char 72 moved E 22, 25 Char 72 moved E relative to 43,48, 67 Char 38 moved L relative to 43, 48 Char 75 moved E relative to 14, 15, Char 75 moved E relative to 14, 15, 31, 23,74,86,90,95 relative to 14, 15, 86,91,95 Char 51 moved E 31,86,95 Char 85 moved L relative to 15, 31 Char 85 moved L relativeto21,22, 29 Char 66 moved L relative to 21, 22, Char 88 moved L relative to 1, 2, 6, 29 relative to 6, 26 41,42,69,70 Char 88 moved L Char 92 moved L Char 70 moved L relative to 6, 26 relative to 6, 41,93 relative to 2, 6,41 Char 92 moved L Char 93 moved L Char 72 moved E relative to 6, 41,93 relative to 6,26, 41, relative to 43, 48, 97 Char 93 moved L 42,96 Char 73 moved E relative to 6, 26, Char 96 moved L relative to 86, 90, 41,42,96 relative to 6, 26, 41, 91,95 Char 98 moved L 42 Char 75 moved E relative to 3 Char 97 moved L relative to 14, 15, Char 99 moved L relative to 3, 41,42, 31,86,90,95 relative to 6, 41,42 45 Char 82 moved L Char 98 moved L relative to 41, 42, relative to 3, 45 43, 50, 83 Char 99 moved L Char 83 moved L relative to 3, 6, 26, relative to 3, 6 41,42,45 Char 85 moved L relative to 12, 21, 22, 25, 29 Char 88 moved L relative to 1, 2, 6, 26, 92

243 Char 92 moved L relative to 2, 6, 41, 93 Char 93 moved L relative to 1, 2, 6, 26,41,42,96 Char 94 moved L relative to 12, 25 Char 96 moved L relative to 1, 2, 41, 42 Char 98 moved L relative to 3,41,42 Char 99 moved L relative to 1, 2, 3, 6, 41,42,92 Neoaves —> Twins (20, 18) Twins (15, 51) Char 50 moved E Phalacrocorax Char 50 moved E Char 7 moved E relative to 1,2, 6, auritus relative to 1, 2, 3, 6, relative to 86, 90, 95 41,42 41,42 Char 26 moved L Char 66 moved L Char 57 moved L relative to 13, 66 relative to 88,93 relative to 31, 38 Char 50 moved E Char 69 moved L Char 58 moved L relative to 1, 2, 6, relative to 70, 87 relative to 88, 93,99 41,42 Char 73 moved L Char 62 moved L Char 66 moved L relative to 87, 92, relative to 89, 94 relative to 88, 93 93 Char 66 moved L Char 69 moved L Char 74 moved L relative to 88, 93, 99 relative to 70, 87 relative to 75 Char 69 moved L Char 73 moved L Char 75 moved L relative to 70, 87, 88 relative to 13,87, relative to 87, 88, Char 73 moved L 92,93 92,93 relative to 87, 88,91, Char 74 moved L Char 82 moved E 92, 93, 99 relative to 75 relative to 1, 13, Char 74 moved L Char 75 moved L 96,98 relative to 75 relative to 13, 87, Char 83 moved E Char 75 moved L 88,91,92,93 relative to 98 relative to 87, 88, 92, Char 82 moved E Char 84 moved E 93,99 relative to 1, 13, 96, relative to 98 Char 82 moved E 98 Char 97 moved E relative to 1, 13, 96, Char 83 moved E relative to 2, 6 98 relative to 70, 82, 98 Char 98 moved E Char 83 moved E Char 84 moved E relative to 1, 2, 6, relative to 98 relative to 98 40 Char 84 moved E Char 96 moved E relative to 98 relative to 6, 70 Char 97 moved E Char 97 moved E relative to 2, 6 relative to 1, 2, 3, 6, Char 98 moved E 40,41,42,45 relative to 1,2,6,40 Char 98 moved E relative to 1,2,6,40 Root --> Char 4 moved E Char 32 moved E Char 32 moved E

244 Palaeognathae relative to 18, 31 relative to 40, 58, 73 relative to 40, 58, Char 23 moved E Char 38 moved L 73 relative to 51, 74, 75 relative to 13, 14, Char 38 moved L Char 32 moved E 15, 19 relative to 13, 14, relative to 40, 58, 73, Char 57 moved L 15,19 93 relative to 15,31,51 Char 57 moved L Char 38 moved L Char 96 moved L relative to 15, 31, relative to 7, 13, 14, relative to 1,3,40, 51 15,19 41,42,45,46,50, Char 96 moved L Char 57 moved L 80 relative to 1,3,40, relative to 7, 15, 31, Char 97 moved L 41,42,45,50,80 51 relative to 8, 10, 11, Char 97 moved L Char 59 moved L 43, 46, 80 relative to 10, 11, relative to 12, 19, 21, Char 98 moved L 43, 80 22, 25, 29 relative to 3, 8, 10, Char 98 moved L Char 66 moved L 11,33,41,42,43, relative to 3, 10, relative to 41, 42 44, 45,46, 48, 50, 11,33,41,42,43, Char 68 moved L 80 44, 45, 48, 50, 80 relative to 19, 20, 24 Char 69 moved L relative to 3, 40, 41, 42, 45, 50 Char 70 moved L relative to 1,3,40, 41,42,45,50 Char 71 moved L relative to 13, 14, 15, 19,20,24,31,51 Char 87 moved L relative to 1, 2, 3, 5, 6,10, 11,40,41,42, 43, 45, 47, 50, 52, 54, 58, 80, 81 Char 92 moved L relative to 1, 2, 3, 5, 6,10,40,41,42,45, 50, 52, 80 Char 96 moved L relative to 1,2,3,40, 41,42,43,45,50,80 Char 97 moved L relative to 10, 11,33, 43, 47, 67, 80, 98 Char 98 moved L relative to 3, 10, 11, 33,41,42,43,44, 45, 47, 48, 50, 80 Palaeognathae —> Char 9 moved L Char 6 moved E Char 9 moved L Dromaius sp. relative to 10, 11,33 relative to 40, 58 relative to 10, 11 Char 32 moved E Char 9 moved L Char 32 moved E relative to 13, 14, 91 relative to 10, 11 relative to 13, 14

.245 Char 52 moved E Char 23 moved E Char 52 moved E relative to 1, 3, 8, 16, relative to 74, 75, 90 relative to 1, 16, 33, 42, 43, 45, 47, Char 32 moved E 33, 45,47, 49 49, 50, 97, 98 relative to 13, 14,38 Char 54 moved E Char 54 moved E Char 52 moved E relative to 17,44, relative to 17,44,48, relative to 1, 16, 33, 48,49 49, 97, 98 45, 46,47,49, 65 Char 70 moved L Char 61 moved L Char 54 moved E relative to 9, 10, relative to 12,25,30 relative to 17,44, 11,43,96 Char 62 moved L 48, 49, 65 Char 71 moved L relative to 12,25,30 Char 59 moved L relative to 1, 3,7, Char 70 moved L relative to 19, 21, 26,38,40,41,42, relative to 9, 10, 11, 22, 25, 28,29 45, 50, 57, 58, 80 33, 43, 44, 47, 96, 98 Char 69 moved L Char 81 moved E Char 71 moved L relative to 2, 3, 5, relative to 48, 49 relative to 1, 3, 7, 16, 40,41,42,45,46, Char 96 moved L 26,38,40,41,42, 50 relative to 10, 11 45, 50, 57, 58, 80 Char 70 moved L Char 72 moved L relative to 1, 2, 3, 5, relative to 33, 44 9,10,11,40,41,42, Char 73 moved E 43, 45,46, 50, 96 relative to 14, 38 Char 71 moved L Char 81 moved E relative to 1, 2, 3, 5, relative to 44,48, 49, 7,13,14,15, 19,24, 97, 98 26,31,38,40,41, Char 93 moved E 42,45,50,51,57, relative to 13, 73 58, 73, 80, 93 Char 96 moved L Char 81 moved E relative to 10, 11 relative to 48, 49 Char 87 moved L relative to 1, 2, 3, 5, 10,11,40,41,42, 43,45,46,50,58, 80 Char 92 moved L relative to 1, 2, 3, 5, 10,40,41,42,45, 46, 50, 80 Char 96 moved L relative to 2, 10, 11, 43 Dromaius sp. --> Twins (3, 45) (5, 6) Twins (20,24) (31, Twins (48, 44) (96, Dromaius (28, 30) (48, 44) (96, 15) (46, 8) (48, 44) 9) novaehollandiae 9) (96, 9) Char 99 moved E (RM) Char 57 moved E Char 5 moved E relative to 13, 26, relative to 7, 18 relative to 40, 58 73,91 Char 99 moved E Char 6 moved E relative to 13, 26, 73, relative to 5 91 Char 52 moved E relative to 42, 50 Char 93 moved E

246 relative to 13, 73 Char 99 moved E relative to 13, 26, 58, 73, 91 Dromaius sp. —> Twins (1,50) Twins (1,50) (8, 52) Twins (1,50) Dromaius Char 6 moved E (44, 81) (48, 49) (57, Char 6 moved E novaehollandiae relative to 13,26,32, 7) (70, 98) relative to 13, 26, (YPM) 73,86,88,90,91, Char 6 moved E 32,73,86,88,90, 93, 95, 99 relative to 13, 26, 91,93,95 Char 18 moved E 32, 73, 86, 88, 90, Char 18 moved E relative to 31, 68, 90 91,93,95 relative to 31, 90 Char 32 moved E Char 17 moved E Char 32 moved E relative to 86, 88, 95 relative to 65, 72 relative to 86, 88, Char 75 moved E Char 18 moved E 95 relative to 86,90, 95 relative to 31,90 Char 75 moved E Char 29 moved L relative to 86, 90, relative to 30 95 Char 30 moved L relative to 19,28 Char 32 moved E relative to 86, 88, 91,93,95 Char 68 moved L relative to 19,24 Char 75 moved E relative to 86, 90, 95 Palaeognathae --> Twins (20, 21) (23, Twins (20, 21) (23, Twins (20, 21) (23, Struthio camelus 4) (79, 65) (81, 17) 4)(30, 28)(79, 65) 4) (79, 65) (81,17) Char 2 moved L (81,17) Char 2 moved L relative to 40, 50 Char 2 moved L relative to 40, 50 Char 5 moved L relative to 40, 50 Char 5 moved L relative to 3,41, 42 Char 5 moved L relative to 3, 41, 42 Char 6 moved L relative to 3, 41, 42 Char 6 moved L relative to 1, 3, 41, Char 6 moved L relative to 1, 3, 41, 42,50 relative to 1, 3, 41, 42, 50 Char 52 moved L 42,50 Char 52 moved L relative to 10, 11,44 Char 52 moved L relative to 10, 11, Char 72 moved E relative to 10, 11,44 44 relative to 43, 47, 48, Char 57 moved L Char 72 moved E 49, 97, 98 relative to 7, 91 relative to 43, 47, Char 93 moved L Char 66 moved L 48, 49, 98 relative to 40, 41,42 relative to 41, 42 Char 93 moved L Char 99 moved L Char 67 moved L relative to 40, 41, relative to 1,3,40, relative to 16, 46 42 41,42,45,69,70,97 Char 72 moved E Char 99 moved L relative to 43, 47, relative to 1,3,40, 48, 49,98 41,42,45,69,70, Char 93 moved L 97 relative to 32, 40, 41,42,91 Char 97 moved L

247 relative to 67, 98 Char 99 moved L relative to 1,3,32, 40,41,42,45,69, 70,97 E = early; L = late. See Legend for bone identities.

248 Legend (Tables 6.1, 6.2) 1 Basioccipital 51 Dorsal ribs 2 Exoccipital 52 Sternal ribs 3 Supraoccipital 53 Uncinate processes 4 Parasphenoid rostrum 54 Sternum (body) 5 Parasphenoid ala 55 Laterocranial sternal processes 6 Parasphenoid lamina 56 Laterocaudal sternal processes 7 Basisphenoid 57 Scapula 8 Laterosphenoid 58 Coracoid 9 Prootic 59 Furcula 10 Opisthotic 60 Humerus 11 Epiotic 61 Radius 12 Squamosal 62 Ulna 13 Parietal 63 Radiale 14 Frontal 64 Ulnare 15 Lacrimal 65 Metacarpal II 16 Mesethmoid 66 Phalanx 1 17 Trabeculae 67 Phalanx 2 18 Nasal 68 Metacarpal HI 19 Premaxilla 69 Phalanx 1 20 Maxilla 70 Phalanx 2 21 Palatine 71 Metacarpal IV 22 Pterygoid 72 Phalanx 1 23 Vomer 73 Ilium 24 Jugal 74 Ischium 25 Quadratojugal 75 Pubis 26 Quadrate 76 Femur 27 Ectethmoid 77 Tibia 28 Dentary 78 Fibula 29 Supra-angular 79 Patella 30 Angular 80 Pretibial bone 31 Splenial 81 Tarsals 32 Prearticular 82 Metatarsal I 33 Articular 83 Phalanx 1 34 Mentomandibular 84 Phalanx 2 35 Entoglossal 85 Metatarsal II 36 Basihyal 86 Phalanx 1 37 Urohyal 87 Phalanx 2 38 Ceratobranchial 88 Phalanx 3 39 Epibranchial 89 Metatarsal HI 40 Cervical centra 90 Phalanx 1 41 Thoracic centra 91 Phalanx 2 42 Synsacral centra 92 Phalanx 3 43 Caudal Centra 93 Phalanx 4 44 Pygostyle 94 Metatarsal IV 45 Cervical arch 95 Phalanx 1 46 Thoracic transverse processes 96 Phalanx 2 47 Synsacral transverse processes 97 Phalanx 3 48 Caudal transverse processes 98 Phalanx 4 49 Synsacral arch 99 Phalanx 5 50 Cervical ribs

249 CONCLUSIONS AND FUTURE RESEARCH

Ossification sequences are not dependent on any one factor, be it sequence of chondrification, source of osteogenic cells, morphology or phylogeny. Many factors appear to play a role, and there are exceptions to every rule. Comparative skeletal development remains a useful approach for understanding the role ontogeny plays in evolutionary change, but ossification sequences themselves do not appear to be the best solution for reconstructing phylogenies, even when a large sample of closely related taxa is used. Resolution of persistent methodological problems will most likely not improve the results obtained when ossification sequence data is used in phylogenetic reconstruction. The small number of elements shifting to support monophyly at the ordinal level will be lost amid the general high rate of change no matter which method is used. In some ways, my research suggests that overweighting large shifts (as event-pairing does) is the best way to capture phylogenetically relevant information from a highly variable data source. Small shifts are of dubious evolutionary significance. The mechanism underlying reduction in element size over evolutionary time resulting in delays in the timing of onset of ossification is at present unclear. Testing of this hypothesis on an extended ossification sequence data set provided support for this pattern in other orders of amniotes, extending these results outside of Aves and suggesting that this pattern might be broadly applicable in many vertebrate groups (L. Harrison, pers. comm.). A potential mechanism might be conservation of the rate of ossification in homologous elements. A larger element would have to begin ossifying earlier in development in order to maintain proportionately equal ossification to its relatively smaller homologue at birth or hatching. This provides a simple starting point for examining the mechanisms underlying this pattern. Reduced field size and reduced mechanical stimulus caused by reduced or delayed muscle development are also possible factors that need to be examined (see Chapter 6 for discussion). In spite of these uncertainties regarding the causes and therefore the potential significance of ossification sequence heterochronies, studying this type

250 of data can have more far-reaching implications regarding modularity and developmental constraint in vertebrate evolution. Bones and cartilage represent the fundamental building blocks of the vertebrate body plan, and as such how they are capable of changing is inherently interesting. Ossification sequence and qualitative observations about bone development are also the only information available when fossil embryos are considered. Future work necessary to advance this field of research can be divided into two broad categories. The first employs an approach similar to the one used in this thesis, examining ossification sequences in a comparative context. Sequence differences between altricial and precocial embryos have not previously been examined at this fine a scale using ranked (rather than stage) data. In order to make a more robust test of the influence of precociality on ossification sequence, available sequences for altricial birds must be considerably expanded. This is essentially a test for hatchling adaptation, but will be complicated by the influence of phylogeny: precociality is primitive for birds, and none of the basal orders discussed in this thesis contain altricial taxa. Studying the evolution of ossification sequences among extant squamates represents a productive avenue for future research, in order to extend the results of this study to other diapsids with very different ecologies. A well-resolved sequence is also not currently available for Alligator or other crocodilians. This is critical for polarizing the direction of ossification sequence changes within extant birds, as crocodilians are the only non-avian archosaurs alive today and thus amenable to this type of study. The second approach involves experimental and histological work on a few species to address the mechanistic basis of some of the factors appearing to exert the most influence on ossification sequences. Interspecific comparisons of the rate of ossification of homologous elements relative to intraspecific comparisons of the rate of ossification of non-homologous elements may provide a mechanism for earlier ossification of larger bones. Allometric studies of bone growth relative to the onset of ossification and instraspecific studies of ossification sequence in domestic species where extensive morphological change has accrued through selective breeding would also be informative.

251 This thesis contributes the first description of mid- to late-stage skeletal development in several avian species at the organismal scale. These species include the Domestic Turkey, Pekin Duck, Common Eider, Ostrich, Emu, Rhea, Elegant-Crested Tinamou, and Common Tern. Sequence divergence and conservation was discussed at both the specific and ordinal scales, allowing for a broad picture of the evolution of the timing and pattern of skeletal development in birds to emerge. A large data set was compiled from the ossification sequence component of these comparisons, one of the largest of its kind. This data set did not contain strong phylogenetic signal, but some of the most significant shifts were found to result from morphological changes involving the relative expansion or reduction of an element resulting in acceleration or delay of that same element in sequence. While this hypothesis has been casually proposed in the past by other authors, this is by far the most convincing support it has received, and as a result it has become one of the most plausible hypotheses to explain ossification sequence change. This result will influence the direction and design of future studies in the field, and deserves further consideration, especially with regard to the search for potential mechanisms driving this pattern.

•>

252 BIBLIOGRAPHY

Abzhanov, A., M. Protas, B. R. Grant, P. R. Grant, andC. J. Tabin. 2004. Bmp4 and morphological variation of beaks in Darwin's finches. Science 305:1462-1465. Adams, R. A. 1992. Stages of development and sequence of bone formation in the little brown bat, Myotis lucifugus. Journal of Mammalogy 73:160-167. Adriaens, D., and W. Verraes. 1998. Ontogeny of the osteocranium in the African catfish, Clarias gariepinus Burchell (1822) (Siluriformes: Clariidae): ossification sequence as a response to functional demands. Journal of Morphology 235:183-237. Alberch, P. 1985. Problems with the interpretation of developmental sequences. Systematic Zoology 34:45-58. Andersson, M. 1999. Phylogeny, behaviour, plumage evolution and neoteny in skuas Stercorariidae. Journal of Avian Biology 30:205-215. Arendt, J. D., and D. S. Wilson. 2000. Population differences in the onset of cranial ossification in pumpkinseed (Lepomis gibbosus), a potential cost of rapid growth. Canadian Journal of Fisheries and Aquatic Sciences 57:351- 356. Badyaev, A. V., and T. E. Martin. 2000. Individual variation in growth trajectories: phenotypic and genetic correlations in ontogeny of the House Finch {Carpodacus mexicanus). Journal of Evolutionary Biology 13:290- 301. Balanoff, A., and T. Rowe. 2007. Osteological description of an embryonic skeleton of the extinct Elephant Bird, Aepyornis (Palaeognathae: Ratitae). Society of Vertebrate Paleontology Memoir 9:1-53. Baumel, J. J., and L. M. Witmer. 1993. Osteologia; pp. 45-132 in J. J. Baumel, A. S. King, J. E. Breazile, H. E. Evans, and J. C. Vanden Berge (eds.), Handbook of avian anatomy. Nuttall Ornithological Club, Cambridge. Beddard, F. E. 1898. The structure and classification of birds. Longmans, Green and Co., London, 548 pp.

253 Bennett, M. B., and J. G. Garden. 2004. Locomotion and gaits of the Northern brown bandicoot, Isoodon macrourus, (Marsupialia: Peramelidae). Journal of Mammalogy 85:296-301. Berger, A. J. 1956. Anatomical variation and avian anatomy, Condor 58:433-441. Biewener, A. A., N. L. Fazzalari, D. D. Konieczynski, and R. V. Baudinette. 1996. Adaptive changes in trabecular architecture in relation to functional strain patterns and disuse. Bone 19:1-8. Bininda-Emonds, O. R. P., J. E. Jeffery, and M. K. Richardson. 2003. Is sequence heterochrony an important evolutionary mechanism in mammals? Journal of Mammalian Evolution 10:335-361. Bininda-Emonds, O. R. P., J. E. Jeffery, M. I. Coates, and M. K. Richardson. 2002. From Haeckel to event-pairing: the evolution of developmental sequences. Theory in Biosciences 121:297-320. Bininda-Emonds, O. R. P., J. E. Jeffery, M. R. Sanchez-Villagra, J. Hanken, M. W. Colbert, C. Pieau, L. Selwood, C. J. ten Cate, A. Raynaud, C. K. Osabutey, and M. K. Richardson. 2007. Forelimb-hind limb developmental timing differences across tetrapod phylogeny. BMC Evolutionary Biology 7:182. Blanco, M. J., and B. Sanchiz. 2000. Evolutionary mechanisms of rib loss in anurans: a comparative developmental approach. Journal of Morphology 244:57-67. Bledsoe, A. H. 1988. A phylogenetic analysis of postcranial skeletal characters of the ratite birds. Annals of the Carnegie Museum 57:73-90. Blom, J., and C. Lilja. 2004. A comparative study of growth, skeletal development and eggshell composition in some species of birds. Journal of Zoology, London 262:361-369. Briggs, J. 2003. Fishes and birds: Gondwana life rafts reconsidered. Systematic Biology 52:548-553. Broom, R. 1906. On the early development of the appendicular skeleton of the Ostrich, with remarks on the origin of birds. Transactions of the South African Philosophical Society 16:355-369.

254 Cane, W. P. 1993. The ontogeny of postcranial integration in the Common Tern, Sterna hirundo. Evolution 47:113 8-1151. Cane, W. P. 1994. Ontogenetic evidence for relationships within the Laridae. Auk 111:873-880. Chipman, A. D., A. Haas, E. Tchernov, and O. Khaner. 2000. Variation in anuran embryogenesis: differences in sequence and timing of early developmental events. Journal of Experimental Zoology (Molecular and Developmental Evolution) 288. Christian, P. D., L. Christidis, and R. Schodde. 1992. Biochemical systematics of the Charadriiformes (shorebirds): relationships between the Charadrii, Scolopaci and Lari. Australian Journal of Zoology 40:291-302. Chu, P. C. 1995. Phylogenetic reanalysis of Strauch's osteological data set for the Charadriiformes. Condor 97:174-196. Chu, P. C. 1998. A phylogeny of the gulls (Aves: Larinae) inferred from osteological and integumentary characters. Cladistics 14:1-43. Chubb, A. L. 2004. New nuclear evidence for the oldest divergence among neognath birds: the phylogenetic utility of ZENK (i). Molecular Phylogenetics and Evolution 30:140-151. Clark, C. T., and K. K. Smith. 1993. Cranial osteogenesis in Monodelphis domestica (Didelphidae) and Macropus eugenii (Macropodidae). Journal of Morphology 215:119-149. Codd, J. R., D. F. Boggs, S. F. Perry, and D. R. Carrier. 2005. Activity of three muscles associated with the uncinate processes of the giant Canada Goose Branta canadensis maximus. Journal of Experimental Biology 208:849- 857. Colbert, M. W., and T. Rowe. 2008. Ontogenetic sequence analysis: using parsimony to characterize developmental sequences and sequence polymorphism. Journal of Experimental Zoology (Molecular and Developmental Evolution) 310B: 1-19. Cracraft, J. 1968. The lacrimal-ectethmoid bone complex in birds: a single character analysis. American Midland Naturalist 80:316-359.

255 Cracraft, J. 1974. Phylogeny and evolution of the ratite birds. Ibis 116:494-521. Cracraft, J. 1988. The major clades of birds; pp. 339-361 in M. J. Benton (ed.), The phylogeny and classification of the tetrapods Volume 1: Amphibians, reptiles, birds. Clarendon Press, Oxford. Cracraft, J. 2001. Avian evolution, Gondwana biogeography and the Cretaceous- Tertiary mass extinction event. Proceedings of the Royal Society of London Series B 268:459-469. Crowe, T. M., R. C. K. Bowie, P. Bloomer, T. G. Mandiwana, T. A. J. Hedderson, E. Randi, S. L. Pereira, and J. Wakeling. 2006. Phylogenetics, biogeography and classification of, and character evolution in, gamebirds (Aves: Oalliformes): effects of character exclusion, data partitioning and missing data. Cladistics 22:495-532. Davies, S. J. J. F. 2002. Ratites and tinamous: Tinamidae, Rheidae, Dromaiidae, Casuariidae, Apterygidae, Struthionidae; in C. M. Perrins, W. J. Bock, and J. Kikkawa (eds.), Bird Families of the World. Oxford University Press, Oxford, 310 pp. Dawson, A., F. J. McNaughton, A. R. Goldsmith, and A. A. Degen. 1994. Ratite- like neoteny induced by neonatal thyroidectomy of European starlings, Sturnus vulgaris. Journal of Zoology, London 232:633-639. Dawson, A., D. C. Deeming, A. C. K. Dick, and P. J. Sharp. 1996. Plasma thyroxine concentrations in farmed Ostriches in relation to age, body weight, and growth hormone. General and Comparative Endocrinology 103:308-315. de Beer, G. R., and E. J. W. Barrington. 1934. The segmentation and chondrification of the skull of the duck. Philosophical Transactions of the Royal Society of London Series B 223:411-467. Dietz, M. W., and R. E. Ricklefs. 1997. Growth rate and maturation of skeletal muscles over a size range of galliform birds. Physiological Zoology 70:502-510.

256 Dingerkus, G., and L. D. Uhler. 1977. Enzyme clearing of alcian blue stained whole small vertebrates for demonstration of cartilage. Stain Technology 52:229-232. Donne-Gousse, C, V. Laudet, and C. Hanni. 2002. A molecular phylogeny of Anseriformes based on mitochondrial DNA analysis. Molecular Phylogenetics and Evolution 23:339-356. Dyke, G. J., and M. van Tuinen. 2004. The evolutionary radiation of modern birds (Neornithes): reconciling molecules, morphology and the fossil record. Zoological Journal of the Linnean Society 141:153-177. Dyke, G. J., B. E. Gulas, and T. M. Crowe. 2003. Suprageneric relationships of galliform birds (Aves, Galliformes): a cladistic analysis of morphological characters. Zoological Journal of the Linnean Society 137:227-244. Dzialowski, E. M., and P. R. Sotherland. 2004. Maternal effects of egg size on Emu Dromaius novaehollandiae egg composition and hatchling phenotype. Journal of Experimental Biology 207:597-606. Elzanowski, A. 1986: Ontogeny and evolution of the ratites. Acta XIX Congressus Internationalis Ornithologici, Ottawa, 1986. Erdmann, K. 1940. Zur Entwicklungsgeschichte der Knochen im Schadel des Huhnes bis zum Zeitpunkt des Ausschliipfens aus dem Ei. Zeitschrift fur Morphologie und Okologie der Tiere 36:316-400. Ericson, P. G. P., I. Envall, M. Irestedt, and J. A. Norman. 2003. Inter-familial relationships of the shorebirds (Aves: Charadriiformes) based on nuclear DNA sequence data. BMC Evolutionary Biology 3:1-14. Evans, D. J. R., and D. M. Noden. 2006. Spatial relations between avian craniofacial neural crest and paraxial mesoderm cells. Developmental Dynamics 235:1310-1325. Fain, M. G., and P. Houde. 2004. Parallel radiations in the primary clades of birds. Evolution 58:2558-2573. Fain, M. G., and P. Houde. 2007. Multilocus perspectives on the monophyly and phylogeny of the order Charadriiformes (Aves). BMC Evolutionary Biology 7:15 pp.

257 Feduccia, A., and J. Nowicki. 2002. The hand of birds revealed by early ostrich embryos. Naturwissenschaften 89:391-393. Frobisch, N. B. In Press. Ossification patterns in the tetrapod limb - conservation and divergence from morphogenetic events. Biological Reviews. Frobisch, N. B., R. L. Carroll, and R. R. Schoch. 2007. Limb ossification in the Paleozoic branchiosaurid Apateon (Temnospondyli) and the early evolution of preaxial dominance in tetrapod limb development. Evolution and Development 9:69-75. Galis, F., J. van Alphen, and J. Metz. 2001. Why five fingers? Evolutionary constraints on digit numbers. Trends in Ecology and Evolution 16:637- 646. Galis, F., J. van Alphen, and J. Metz. 2002. Digit reduction: via repatterning or developmental arrest? Evolution and Development 4:249-251. Garn, S. M., C. G. Rohmann, and T. Blumenthal. 1966. Ossification sequence polymorphism and sexual dimorphism in skeletal development. American Journal of Physical Anthropology 24:101-116. Gemmell, R. T., G. Johnston, and M. M. Bryden. 1988. Osteogenesis in two marsupial species, the bandicoot Isoodon macrourus and the possum Trichosurus vulpecula. Journal of Anatomy 159:155-164. Gibb, G. C, O. Kardailsky, R. T. Kimball, E. L. Braun, and D. Penny. 2007. Mitochondrial genomes and avian phylogeny: complex characters and resolvability without explosive radiations. Molecular Biology and Evolution 24:269-280. Gill, F. B. 1995. Ornithology. 2nd Edition. W.H. Freeman and Company, New York, 766 pp. Glucksmann, A. 1942. The role of mechanical stresses in bone formation in vitro. Journal of Anatomy 76:231-239. Glutz von Blotzheim, U. 1958. Zur Morphologie und Ontogenese von Schultergurtel, Sternum und Becken von Struthio, Rhea und Dromiceius. Revue Suisse de Zoologie 65:609-772.

258 Goldschmid, A. 1972. Die Entwicklung des Craniums der Mausvogel (Coliidae, Coliiformes, Aves). IV. Die Entwicklung des Osteocraniums. Gegenbaurs Morphologisches Jahrbuch 118:553-569. Goswami, A. 2007. Cranial modularity and sequence heterochrony in mammals. Evolution and Development 9:290-298. Goudie, R. L, G. J. Robertson, and A. Reed. 2000. Common Eider {Somateria mollissima). The birds of North America, Inc., Philadelphia. Guralnick, R. P., and D. R. Lindberg. 2001. Reconnecting cell and animal lineages: what do cell lineages tell us about the evolution and development of Spiralia? Evolution 55:1501-1519. Gussekloo, S. W. S., and R. G. Bout. 2002. Non-neotenous origin of the palaeognathous (Aves) pterygoid-palate complex. Canadian Journal of Zoology 80:1491-1497. Gussekloo, S. W. S., and R. G. Bout. 2005. The kinematics of feeding and drinking in palaeognathous birds in relation to cranial morphology. Journal of Experimental Biology 208:3395-3407. Hall, B. K. 2001. Development of the clavicles in birds and mammals. Journal of Experimental Zoology 289:153-161. Haluska, F., and P. Alberch. 1983. The cranial development of Elaphe obsoleta (Ophidia, Colubridae). Journal of Morphology 178:37-55. Hamburger, V., and H. L. Hamilton. 1951. A series of normal stages in the development of the chick embryo. Journal of Morphology 88:49-92. Harrison, L. B., and H. C. E. Larsson. In Press. Estimating evolution of temporal sequence changes: a practical approach to inferring ancestral developmental sequences and sequence heterochrony. Systematic Biology. Herring, S. W. 1993. Epigenetic and functional influences on skull growth; pp. 153-206 in J. Hanken and B. K. Hall (eds.), The Skull. University of Chicago Press, Chicago. Hofmann, M. I., C. Schradin, and T. Geissmann. 2007. Radiographic evaluation of neonatal skeletal development in Callimico goeldii reveals closer

259 similarity to Callithrixjacchus than to Saguinus oedipus. American Journal of Physical Anthropology 69:420-433. Hogg, D. A. 1980. A reinvestigation of the centers of ossification in the avian skeleton at and after hatching. Journal of Anatomy 130:725-743. Jeffery, J. E., M. K. Richardson, M. I. Coates, and O. R. P. Bininda-Emonds. 2002a. Analyzing developmental sequences within a phylogenetic framework. Systematic Biology 51:478-491. Jeffery, J. E., O. R. P. Bininda-Emonds, M. I. Coates, and M. K. Richardson. 2002b. Analyzing evolutionary patterns in amniote embryonic development. Evolution and Development 4:292-302. Jeffery, J. E., O. R. P. Bininda-Emonds, M. I. Coates, and M. K. Richardson. 2005. A new technique for identifying sequence heterochrony. Systematic Biology 54:230-240. Jollie, M. T. 1957. The head skeleton of the chicken and remarks on the anatomy of this region in other birds. Journal of Morphology 100:389-436. Kaiser, V. B., M. van Tuinen, and H. Ellegren. 2007. Insertion events of CR1 retrotransposable elements elucidate the phylogenetic branching order in galliform birds. Molecular Biology and Evolution 24:338-347. Kesteven, H. L. 1942. The ossification of the avian chondrocranium, with special reference to that of the Emu. Proceedings of the Linnean Society of New South Wales 67:213-237. King, S. J. 2004. Relative timing of ontogenetic events in primates. Journal of Zoology, London 264:267-280. Kirkwood, J. K., P. J. Duignan, N. F. Kember, P. M. Bennett, and D. J. Price. 1989. The growth rate of the tarsometatarsus bone in birds. Journal of Zoology, London 217:403-416. Koenemann, S., and F. R. Schram. 2002. The limitations of ontogenetic data in phylogenetic analyses. Contributions to Zoology 71. Kundrat, M., V. Seichert, A. P. Russell, and K. Smetana Jr. 2002. Pentadactyl pattern of the avian wing autopodium and pyramid reduction hypothesis.

260 Journal of Experimental Zoology (Molecular and Developmental Evolution) 294:152-159. Lande, R. L. 1978. Evolutionary mechanisms of limb loss in tetrapods. Evolution 32:73-92. Larsson, H. C. E. 1998. A new method for comparing ontogenetic and phylogenetic data and its application to the evolution of the crocodilian secondary palate. Neues Jahrbuch fur Geologie und Palaontologie - Abhandlungen 210:345-368. Larsson, H. C. E., and G. P. Wagner. 2002. Pentadactyl ground state of the avian wing. Journal of Experimental Zoology (Molecular and Developmental Evolution) 294:146-151. Latham, J. 1790. Index ornithologicus, sive systema ornithologiae; complectens avium divisionem in classes, ordines, genera, species, ipsarumque varietates: adjectis synonymis, locis, descriptionibus, & c., Volume .1. Leigh and Sotheby, London, 500 pp. Leighton, V. L. 1894. The development of the wing of Sterna wilsonii. American Naturalist 28:761-774. Livezey, B. C. 1991. A phylogenetic analysis and classification of recent dabbling ducks (Tribe Anatini) based on comparative morphology. Auk 108:471- 507. Livezey, B. C. 1995. Heterochrony and the evolution of avian flightlessness; pp. 169-193 in K. J. McNamara (ed.), Evolutionary change and heterochrony. John Wiley and Sons, New York. Livezey, B. C. 1996. A phylogenetic reassessment of the Tadornine - Anatine divergence (Aves: Anseriformes; Anatidae). Annals of the Carnegie Museum 65:27-88. Livezey, B. C. 1997. A phylogenetic classification of waterfowl (Aves: Anseriformes), including selected fossil species. Annals of the Carnegie Museum 66:457-496. Livezey, B. C. 1998. A phylogenetic analysis of the Gruiformes (Aves) based on morphological characters, with an emphasis on the Rails (Rallidae).

261 Philosophical Transactions of the Royal Society of London Series B 353:2077-2151. Livezey, B. C, and R. L. Zusi. 2001. Higher-order phylogenetics of modern Aves based on comparative anatomy. Netherlands Journal of Zoology 51:179- 205. Livezey, B. C, and R. L. Zusi. 2007. Higher-order phylogeny of modern birds (Theropoda, Aves: Neornithes) based on comparative anatomy. II. Analysis and discussion. Zoological Journal of the Linnean Society 149:1- 95. Lutz, H. 1942. Beitrag zur Stammesgeschichte der Ratiten Vergleich zwischen Emu-Embryo und entsprechendem Carinatenstadium. Revue Suisse de Zoologie 49:299-399. Mabee, P. M. 2000. The usefulness of ontogeny in interpreting morphological characters; pp. 84-114 in J. J. Wiens (ed.), Phylogenetic Analysis of Morphological Data. Smithsonian Institution Press, Washington. Mabee, P. M., and J. Humphries. 1993. Coding polymorphic data: examples from allozymes and ontogeny. Systematic Biology 42:166-181. Mabee, P. M., and T. A. Trendler. 1996. Development of the cranium and paired fins in Betta splendens (Teleostei: Percomorpha): intraspecific variation and interspecific comparisons. Journal of Morphology 227:249-287. Mabee, P. M., K. L. Olmstead, and C. C. Cubbage. 2000. An experimental study of intraspecific variation, developmental timing, and heterochrony in fishes. Evolution 54:2091-2106. Maillard, J. 1948. Recherches embryologiques sur Catharacta skua Brunn (pterylose et ossification). Revue Suisse de Zoologie 55:1-113. Maisano, J. A. 2002a. Postnatal skeletal ontogeny in five xantusiids (Squamata: Scleroglossa). Journal of Morphology 254:1-38. Maisano, J. A. 2002b. The potential utility of postnatal skeletal developmental patterns in squamate phylogenetics. Zoological Journal of the Linnean Society 136:277-313.

262 Marshall, C. L. 2000. Embryonic growth patterns in Ostrich and other ratite skeletons. PhD dissertation, Yale University, New Haven, 209 pp. Marugan-Lobon, J., and A. D. Buscalioni. 2006. Avian skull morphological evolution: exploring exo- and endocranial covariation with two-block partial least squares. Zoology 109:217-230. Maxwell, E. E. 2007. The phylogenetic significance of ossification sequence data in birds, and its relevance to fossil taxa. Journal of Vertebrate Paleontology 27:115A. Maxwell, E. E. 2008. Comparative embryonic development of the skeleton of the domestic turkey (Meleaghs gallopavo) and other galliform birds. Zoology 111:242-257. Maxwell, E. E. In Press. Ossification sequence of the avian order Anseriformes, with comparison to other precocial birds. Journal of Morphology, doi 10.1002/jmor. 10644 Maxwell, E. E., and H. C. E. Larsson. 2007. Osteology and myology of the wing of the Emu {Dromaius novaehollandiae), and its bearing on the evolution of vestigial structures. Journal of Morphology 268:423-441. Maxwell, E. E., and L. B. Harrison. In Press. Ossification sequence of the Common Tern {Sterna hirundo) and its implications for the interrelationships of the Lari (Aves, Charadriiformes). Journal of Morphology. Mayr, G., and J. Clarke. 2003. The deep divergences of neornithine birds: a phylogenetic analysis of morphological characters. Cladistics 19:527-553. McGowan, C. 1984. Evolutionary relationships of ratites and carinates: evidence from ontogeny of the tarsus. Nature 307:733-735. McGowan, C. 1985. Tarsal development in birds: evidence for homology with the theropod condition. Journal of Zoology London (A) 206:53-67. McLaughlin, P. A., R. Lemaitre, and C. C. Tudge. 2004. Carcinization in the Anomura - fact or fiction? II. Evidence from larval, megalopal and early juvenile morphology. Contributions to Zoology 73.

263 Meiri, S., and Y. Yom-Tov. 2004. Ontogeny of large birds: migrants do it faster. Condor 106:540-548. Mickevich, M. F., and L. R. Parenti. 1980. Review of the phylogeny of the Charadriiformes (Aves): a new estimate using the method of character compatibility analysis. Systematic Biology 29:108-113. Minelli, A. 2003. The development of animal form: ontogeny, morphology and evolution. Cambridge University Press, New York, 323 pp. Mitgutsch, C, N. Piekarski, L. Olsson, and A. Haas. 2008. Heterochronic shifts during early cranial neural crest cell migration in two ranid frogs. Acta Zoologica 89:69-78. Miiller, G. B., and P. Alberch. 1990. Ontogeny of the limb skeleton in Alligator mississippiensis: developmental invariance and change in the evolution of archosaur limbs. Journal of Morphology 203:151-164. Miiller, H. 1963. Die Morphologie und Entwicklung des Craniums von Rhea americana Linne II. Viszeralskelett, Mittelohr und Osteocranium. Zeitschrift fur Wissenschaftliche Zoologie 168:35-118. Mun, A. M., and I. L. Kosin. 1960. Developmental stages of the broad breasted bronze turkey embryo. Biological Bulletin 119:90-97. Nakane, Y., and M. Tsudzuki. 1999. Development of the skeleton in Japanese quail embryos. Development Growth and Differentiation 41:523-534. Newman, S.A., and H.L. Frisch. 1979. Dynamics of skeletal pattern formation in developing chick limbs. Science 205:662-668. Nisbet, I. C. T. 2002. Common Tern {Sterna hirundo) in A. Poole and F. Gill (eds.), The birds of North America. The Birds of North America Inc., Philadelphia. Nunn, C. L., and K. K. Smith. 1998. Statistical analyses of developmental sequences: the craniofacial region of marsupial and placental mammals. American Naturalist 152:82-101. Parker, T. J. 1881. On the skeleton oiNotornis mantelli. Transactions of the New Zealand Institute 14:245-258.

264 Parker, T. J. 1891. Observations on the anatomy and development of Apteryx. Philosophical Transactions of the Royal Society of London Series B 182:25-134. Parker, T. J. 1892. Additional observations on the development of Apteryx. Philosophical Transactions of the Royal Society of London Series B 183:73-84. Parker, W. K. 1866. On the structure and development of the skull in the ostrich tribe. Philosophical Transactions of the Royal Society of London 156:113- 183. Parker, W. K. 1869. On the structure and development of the skull of the common fowl {Gallus domesticus). Philosophical Transactions of the Royal Society of London 159:755-807. Parker, W. K. 1888. On the structure and development of the wing in the Common Fowl. Philosophical Transactions of the Royal Society of London, B 179:385-398. Paton, T. A., and A. J. Baker. 2006. Sequences from 14 mitochondrial genes provide a well-supported phylogeny of the charadriiform birds congruent with the nuclear RAG-1 tree. Molecular Phylogenetics and Evolution 39:657-667. Paton, T. A., A. J. Baker, J. G. Groth, and G. F. Barrowclough. 2003. RAG-1 sequences resolve phylogenetic relationships within Charadriiform birds. Molecular Phylogenetics and Evolution 29:268-278. Pierotti, R. J., and T. P. Good. 1994. Herring Gull {Larus argentatus); pp. 1-28 in A. Poole and F. Gill (eds.), The birds of North America. The Academy of Natural Sciences, Philadelphia. Poe, S. 2004. A test for patterns of modularity in sequences of developmental events. Evolution 58:1852-1855. Poe, S. 2006. Test of von Baer's law of the conservation of early development. Evolution 60:2239-2245. Poe, S., and M. H. Wake. 2004. Quantitative tests of general models for the evolution of development. American Naturalist 164:415-422.

265 Pontoppidan, E. 1763. Den Danske atlas eller Konge-Riget Dannemark, med dets naturlige egenskaber, elementer, indbyggere, vaerter, dyr og andre aff0dninger, dets gamle tildragelser og naervaerende omstasndingheder i alle provintzer, staeder, kirker, slotte og herre-gaarde. Forestillet ved en udforlig lands-beskrivelse, saa og oplyst med dertil forfasrdigede land-kort over enhver provintz, samt ziret med staedernes prospecter, grund-ridser, og andre merkvaerdige kaabber-stykker. Efter Soy-kongelig allernaadigst befalning., Volume 1. Godiche, Kiobenhavn, 723 pp. Prinzinger, R., V. Dietz, and B. Nagel. 1997. Respiratory quotient and embryological development of metabolic heat production in the Rhea (Rhea americand). Journal of Thermal Biology 22:223-226. Prochel, J. 2006. Early skeletal development in Talpa europaea, the common European mole. Zoological Science 23:427-434. Reiss, J. O. 2002. The phylogeny of amphibian metamorphosis. Zoology 105:85- 96. Richman, J. M., M. Buchtova, and J. C. Boughner. 2006. Comparative ontogeny and phylogeny of the upper jaw skeleton in amniotes. Developmental Dynamics 235:1230-1243. Rieppel, O. 1992. Studies on skeleton formation in reptiles. III. Patterns of ossification in the skeleton of Lacerta vivipara Jacquin (Reptilia, Squamata). Fieldiana (Zoology) 68:1-25. Rieppel, O. 1993a. Studies on skeleton formation in reptiles, v. Patterns of ossification in the skeleton of Alligator mississippiensis Daudin (Reptilia, Crocodylia). Zoological Journal of the Linnean Society 109:301-325. Rieppel, O. 1993b. Studies on skeleton formation in reptiles: patterns of ossification in the skeleton of Chelydra serpentina (Reptilia, Testudines). Journal of Zoology (London) 231:487-509. Rieppel, O. 1994. Studies on skeleton formation in reptiles. Patterns of ossification in the skeleton of Lacerta agilis exigua Eichwald (Reptilia, Squamata). Journal of Herpetology 28:145-153.

266 Rogulska, T. 1962. Differences in the process of ossification during the embryonic development of the chick {Gallus domesticus L.), Rook (Corvus frugilegus L.) and Black-Headed Gull {Larus ridibundus L.). Zoologica Poloniae 12:223-236. Romanoff, A. L. I960. The avian embryo: Structural and functional development. Macmillan Company, New York, 1305 pp. Rotthowe, K., and J. M. Starck. 1998. Evidence for a phylogenetic position of button quails (Turnicidae: Aves) among the Gruiformes. Journal of Zoological Systematics and Evolutionary Research 36:39-51. Sahan, U., O. Altan, A. Ipek, and B. Yilmaz. 2003. Effects of some egg characteristics on the mass loss and hatchability of Ostrich {Struthio camelus) eggs. British Poultry Science 44:380-385. Sanchez-Villagra, M. R. 2002. Comparative patterns of postcranial ontogeny in therian mammals: an analysis of relative timing of ossification events. Journal of Experimental Zoology Part B: Molecular and Developmental Evolution 294:264-273. Schinz, R., and R. Zangerl. 1937. Beitrage zur Osteogenese des Knochensystems beim Haushuhn, bei der und beim Haubensteissfuss. Denkschriften der SchweizerischenNaturforschenden Gesellschaft Abteilung 2 72:116-162. Schlosser, G. 2001. Using heterochrony plots to detect the dissociated coevolution of characters. Journal of Experimental Zoology (Molecular and Developmental Evolution) 291:282-304. Schlosser, G. 2003. Mosaic evolution of neural development in anurans: acceleration of spinal cord development in the direct developing frog Eleutherodactylus coqui. Anatomy and Embryology 206:215-227. Schlosser, G., and G. Roth. 1997. Evolution of nerve development in frogs. Brain, Behavior and Evolution 50:94-128. Schneider, R. A., and J. A. Helms. 2003. The cellular and molecular origins of beak morphology. Science 299:565-568. Schnell, G. D. 1970. A phenetic study of the suborder Lari (Aves) I. Methods and results of principal components analysis. Systematic Zoology 19:35-57.

267 Schoch, R. R. 2002. The early formation of the skull in extant and Paleozoic amphibians. Paleobiology 28:278-296. Schoch, R. R. 2006. Skull ontogeny: developmental patterns of fishes conserved across major tetrapod clades. Evolution and Development 8:524-536. Schulmeister, S., and W. C. Wheeler. 2004. Comparative and phylogenetic analysis of developmental sequences. Evolution and Development 6:50- 57. Schumacher, G. H., and E. Wolff. 1966a. Zur vergleichenden Osteogenese von Gallus domesticus L., Larus ridibundus L. und Larus canus L. II. Zeitliches Erscheinen der Ossifikationen bei Larus ridibundus L. und Larus canus L. Gegenbaurs Morphologisches Jahrbuch 110:620-635. Schumacher, G. H., and E. Wolff. 1966b. Zur vergleichenden Osteogenese von Gallus domesticus L., Larus ridibundus L. und Larus canus L. I. Zeitliches Erscheinen der Ossifikationen bei Gallus domesticus L. Gegenbaurs Morphologisches Jahrbuch 110:359-373. Scott, J. H. 1957. The mechanical basis of bone formation. Journal of Bone and Joint Surgery 39B:134-144. Shapiro, M. D. 2002. Developmental morphology of limb reduction in Hemiergis (Squamata: Scincidae): chondrogenesis, osteogenesis and heterochrony. Journal of Morphology 254:211 -231. Sheil, C. A. 2003a. Osteology and skeletal development of Apalone spinifera (Reptilia: Testudines: Trionychidae). Journal of Morphology 256:42-78. Sheil, C. A. 2003b. Skeletal development in turtles: patterns of ossification through ontogeny in Apalone spinifera, Chelydra serpentina, Macrochelys temminckii, and Eretmochelys imbricata (Reptilia: Chelonii). PhD dissertation, Ecology and Evolutionary Biology, University of Kansas, Lawrence, 490 pp. Sheil, C. A. 2005. Skeletal development of Macrochelys temminckii (Reptilia: Testudines: Chelydridae). Journal of Morphology 263:71-106.

268 Sheil, C. A., and E. Greenbaum. 2005. Reconsideration of skeletal development of Chelydra serpentina (Reptilia: Testudinata: Chelydridae): evidence for intraspecific variation. Journal of Zoology, London 265:235-267. Shubin, N. H., and Pi Alberch. 1986. A morphogenetic approach to the origin and basic organization of the tetrapod limb. Evolutionary Biology 20:319-387. Sibley, C. G., and J. E. Ahlquist. 1990. Phylogeny and classification of birds. Yale University Press, New Haven. Sidor, C. A. 2001. Simplification as a trend in synapsid cranial evolution. Evolution 55:1419-1442. Sieglbauer, F. 1911. Zur Entwicklung der Vogelextremitat. Zeitschrift fur Wissenschaftliche Zoologie 97:262-313. Silveira, L. F., and E. Hofling. 2007. Osteologia craniana dos Tinamidae (Aves: Tinamiformes), com implicacoes sistematicas. Boletim do Museu Paraense Emilio Goeldi. Ciencias Naturais 2:15-54. Simonetta, A. M. 1960. On the mechanical implications of the avian skull and their bearing on the evolution and classification of birds. Quarterly Review of Biology 35:206-220. Slack, K. E., F. Delsuc, P. A. Mclenachan, U. Arnason, and D. Penny. 2007. Resolving the root of the avian mitogenomic tree by breaking up long branches. Molecular Phylogenetics and Evolution 42:1 -13. Smirthwaite, J. J., S. D. Rundle, O. R. P. Bininda-Emonds, and J. I. Spicer. 2007. An integrative approach identifies developmental sequence heterochronies in freshwater basommatophoran snails. Evolution and Development 9:122-130. Smith, K. K. 1996. Integration of craniofacial structures during development in mammals. American Zoologist 36:70-79. Smith, K. K. 1997. Comparative patterns of craniofacial development in eutherian and metatherian mammals. Evolution 51:1663-1678. Smith, K. K. 2001. Heterochrony revisited: the evolution of developmental sequences. Biological Journal of the Linnean Society 73:169-186.

269 Smith, K. K. 2003. Time's arrow: heterochrony and the evolution of development. International Journal of Developmental Biology 47:613-621. Smith, K. K. 2006. Craniofacial development in marsupial mammals: developmental origins of evolutionary change. Developmental Dynamics 235:1181-1193. Sonies, F. 1907. Ueber die Entwicklung des Chondrocraniums und der Knorpeligen Wirbelsaule bei den Vogeln. Petrus Camper 4:395-486. Sorenson, M. D., E. Oneal, J. Garcia-Moreno, and D. P. Mindell. 2003. More taxa, more characters: the Hoatzin problem is still unresolved. Molecular . Biology and Evolution 20:1484-1499. Sorenson, M. D., A. Cooper, E. E. Paxinos, T. W. Quinn, H. F. James, S. L. Olson, and R. C. Fleischer. 1999. Relationships of the extinct moa-nalos, flightless Hawaiian waterfowl, based on ancient DNA. Proceedings of the Royal Society of London Series B 266:2187-2193. Starck, J. M. 1989. Zeitmuster der Ontogenesen bei nestfluchtenden und nesthockenden Vogeln, Courier Forschungsinstitut Senckenberg. E. Schweizerbart'sche Verlagsbuchhandlung, Stuttgart, 318 pp. Starck, J. M. 1993. Evolution of avian ontogenies. Current Ornithology 10:275- 366. Starck, J. M. 1996. Comparative morphology and cytokinetics of skeletal growth in hatchlings of altricial and precocial birds. Zoologischer Anzeiger 235:53-75. Starck, J. M. 1998. Structural variants and invariants in avian embryonic and postnatal development; pp. 59-88 in J. M. Starck and R. E. Ricklefs (eds.), Avian growth and development: evolution within the altricial-precocial spectrum. Oxford University Press, New York. Starck, J. M., and R. E. Ricklefs. 1998. Patterns of development: the altricial- precocial spectrum; pp. 3-30 in J. M. Starck and R. E. Ricklefs (eds.), Avian growth and development: evolution within the altricial-precocial spectrum. Oxford University Press, New York.

270 Starck, J. M., and E. Sutter. 2000. Patterns of growth and heterochrony in moundbuilders (Megapodiidae) and fowl (Phasianidae). Journal of Avian Biology 31:527-547. Stedman, S. J. 2000. Horned Grebe {Podiceps auritus). The Birds of North America Inc., Philadelphia. Strauch, J. G. J. 1978. The phylogeny of Charadriiformes (Aves): a new estimate using the method of character compatibility analysis. Transactions of the Zoological Society of London 34:263-345. Strauss, R. E. 1990. Heterochronic variation in the developmental timing of cranial ossifications in poeciliid fishes (Cyprinodontiformes). Evolution 44:1558-1567. Strong, R. M. 1925. The order, time, and rate of ossification of the albino rat {Mus norvegicus albinus) skeleton. American Journal of Anatomy 36:313-355. Swofford, D. L. 2002. PAUP*. Phylogenetic Analysis Using Parsimony (*and other methods). 4.0b 10. Sinauer Associates, Sunderland, Massachussetts. Tazawa, H., J. T. Pearson, T. Komoro, and A. Ar. 2001. Allometric relationships between embryonic heart rate and fresh egg mass in birds. Journal of Experimental Biology 204:165-174. Thomas, G. H., M. A. Wills, and T. Szekely. 2004. Phylogeny of shorebirds, gulls, and alcids (Aves: Charadrii) from the cytochrome-^ gene: parsimony, Bayesian inference, minimum evolution, and quartet puzzling. Molecular Phylogenetics and Evolution 30:516-526. Tokita, M. 2003. The skull development of parrots with special reference to the emergence of a morphologically unique cranio-facial hinge. Zoological Science 20:749-758. Tomlinson, C. A. B. 2000. Feeding in paleognathous birds; pp. 359-394 in K. Schwenk (ed.), Feeding: form, function and evolution in tetrapod vertebrates. Academic Press, San Diego. Towers, ML, R. Mahood, Y. Yin, and C. Tickle. 2008. Integration of growth and specification in chick wing digit-patterning. Nature 452:882-887.

271 Vandewalle, P., G. Germeau, P. Besancenet, E. Parmentier, and E. Baras. 2005. Early development of the head skeleton in Brycon moorei (Pisces, Ostariophysi, Characidae). Journal of Fish Biology 66:996-1024. Velhagen Jr., W. A. 1997. Analyzing developmental sequences using sequence units. Systematic Biology 46:204-210. Verheyen, R. 1956. Contribution a l'anatomie et a la systematique des Galliformes. Bulletin Institut Royal des Sciences Naturelles de Belgique 32:1-24. Vickaryous, M. K., and B. K. Hall. 2006. Homology of the reptilian coracoid and a reappraisal of the evolution and development of the amniote pectoral apparatus. Journal of Anatomy 208:263-285. Wagemans, F., and P. Vandewalle. 2001. Development of the bony skull in common sole: brief survey of morpho-functional aspects of ossification sequence. Journal of Fish Biology 59:1350-1369. Webb, M. 1957. The ontogeny of the cranial bones, cranial peripheral and cranial parasympathetic nerves, together with a study of the visceral muscles of Struthio. Acta Zoologica 38:81 -203. Wetmore, A. 1960. A classification for the birds of the world. Smithsonian Miscellaneous Collections 139:1-37. White, C. D. 1996. Sutural effects of fronto-occipital cranial modification. American Journal of Physical Anthropology 100:397-410. Wiens, J. J. 2000. Coding morphological variation within species and higher taxa for phylogenetic analysis; pp. 115-145 in J. J. Wiens (ed.), Phylogenetic analysis of morphological data. Smithsonian Institution Press, Washington. Witmer, L. M. 1995. Homology of facial structures in extant archosaurs (birds and crocodilians), with special reference to paranasal pneumaticity and nasal conchae. Journal of Morphology 225:269-327. Wu, P., T. X. Jiang, S. Suksaweang, R. B. Widelitz, and C. M. Chuong. 2004. Molecular shaping of the beak. Science 305:1465-1466.

272 Wu, P., T. X. Jiang, J. Y. Shen, R. B. Widelitz, and C. M. Chuong. 2006. Morphoregulation of avian beaks: comparative mapping of growth zone activities and morphological evolution. Developmental Dynamics 235:1400-1412. Yeh, J. 2002. The evolution of development: two portraits of skull ossification in pipoid frogs. Evolution 56:2484-2498. Zusi, R. L. 1984. A functional and evolutionary analysis of rhynchokinesis in birds. Smithsonian Contributions to Zoology 395:1-40. Zusi, R. L., and B. C. Livezey. 2000. Homology and phylogenetic implications of some enigmatic cranial features in galliform and anseriform birds. Annals of the Carnegie Museum 69:157-193. Zusi, R. L., and B. C. Livezey. 2006. Variation in the os palatinum and its structural relation to the palatum osseum of birds (Aves). Annals of the Carnegie Museum 75:137-180.

273 APPENDIX 1.

Specimens examined for this study

Day of Species Specimen # incubation LOTUS argentatus WPC 288 8 LOTUS argentatus WPC289 10 Larus argentatus WPC 290 12 Larus argentatus WPC 243 <13 Larus argentatus WPC 244 <13 Larus argentatus WPC 245 13 Larus argentatus WPC 260 ? Sterna hirundo WPC 255 5 Sterna hirundo WPC 258 6 Sterna hirundo WPC 259 7 Sterna hirundo WPC 269 8 Sterna hirundo WPC 291 8-9 Sterna hirundo WPC 252 9 Sterna hirundo WPC 311 9 Sterna hirundo WPC 271 10 , Sterna hirundo WPC 251 11 Sterna hirundo WPC 309 11 Sterna hirundo WPC 316 11 Sterna hirundo WPC 273 12 Sterna hirundo WPC 305 12 Sterna hirundo WPC 250 13 Sterna hirundo WPC 307 13 Sterna hirundo WPC 276 14 Sterna hirundo WPC 295 14 Sterna hirundo WPC 249 15 Sterna hirundo WPC 317 14-16 Sterna hirundo WPC 275 15.5-16 Sterna hirundo WPC 256 16 Sterna hirundo WPC 294 16 Sterna hirundo WPC 248 17 Sterna hirundo WPC 306 17 Sterna hirundo WPC 254 18 Sterna hirundo WPC 301 18 Sterna hirundo WPC 302 18 Sterna hirundo WPC 257 19 Sterna hirundo WPC 247 20 Sterna hirundo WPC 310 20 Sterna hirundo WPC 313 ?

Dromaius novaehollandiae YPM 112472 17 Dromaius novaehollandiae YPM 112474 18 Dromaius novaehollandiae YPM 112475 18 Dromaius novaehollandiae YPM 112476 19 Dromaius novaehollandiae YPM 112482 26

274 Dromaius novaehollandiae YPM 112485 27 Dromaius novaehollandiae YPM 112486 27 Dromaius novaehollandiae YPM 112489 30 Dromaius novaehollandiae YPM 112490 32 Dromaius novaehollandiae YPM 112493 33 Dromaius novaehollandiae YPM 112494 34 Dromaius novaehollandiae YPM 112496 36 Dromaius novaehollandiae YPM 112497 36 Dromaius novaehollandiae YPM 112499 38 Dromaius novaehollandiae YPM 112501 40 Dromaius novaehollandiae YPM 112502 41 Dromaius novaehollandiae YPM 112503 41 , Dromaius novaehollandiae YPM 112505 43 Dromaius novaehollandiae YPM 112506 44 Dromaius novaehollandiae YPM 112507 45 Dromaius novaehollandiae YPM 112509 47 Dromaius novaehollandiae YPM N/A N/A Dromaius novaehollandiae YPM N/A N/A Dromaius novaehollandiae YPM 112477 Dromaius novaehollandiae RM 8020 22 34 Dromaius novaehollandiae RM8021 23 35 (early) Dromaius novaehollandiae RM8022 24 36 (early) Dromaius novaehollandiae RM8023 25 36 Dromaius novaehollandiae RM 8024 26 36 Dromaius novaehollandiae RM8025 27 36 (late) Dromaius novaehollandiae RM8026 28 37 Dromaius novaehollandiae RM8027 29 37 Dromaius novaehollandiae RM 8028 30 38 Dromaius novaehollandiae RM8029 31 38 Dromaius novaehollandiae RM8030 32 38 Dromaius novaehollandiae RM8031 33 38 Dromaius novaehollandiae RM 8032 34 39 Dromaius novaehollandiae RM8033 35 39 Dromaius novaehollandiae RM 8034 36 40+ Dromaius novaehollandiae RM8035 38 40+ Dromaius novaehollandiae RM8036 39 40+ Dromaius novaehollandiae RM8037 40 40+ Dromaius novaehollandiae RM8038 42 40+ Dromaius novaehollandiae RM8039 43 40+ Dromaius novaehollandiae RM8040 44 40+ Dromaius novaehollandiae RM8041 45 40+ Dromaius novaehollandiae RM 8042 46 40+ Dromaius novaehollandiae RM 8043 47 40+ Dromaius novaehollandiae RM 8044 48 40+ Dromaius novaehollandiae RM 8045 21 35 Dromaius novaehollandiae RM8046 24 36 (late) Dromaius novaehollandiae RM8047 34 40+ Dromaius novaehollandiae RM8048 38 40+ Dromaius novaehollandiae RM8049 31 Dromaius novaehollandiae RM8050 33

275 Dromaius novaehollandiae RM8051 33 Dromaius novaehollandiae RM8052 32 Dromaius novaehollandiae RM8053 35 Dromaius novaehollandiae RM8054 34

Struthio camelus YPM 112435 12 . Struthio camelus YPM 112436 13 Struthio camelus YPM 112437 15 Struthio camelus YPM 112438 16 Struthio camelus YPM 112439 17 Struthio camelus YPM 112440 17 Struthio camelus YPM 112442 19 Struthio camelus YPM 112443 20 Struthio camelus YPM 112444 21 , Struthio camelus YPM 112445 21 Struthio camelus YPM 112446 22 Struthio camelus YPM 112447 22 Struthio camelus YPM 112448 23 Struthio camelus YPM 112449 24 Struthio camelus YPM 112450 25 Struthio camelus YPM 112451 26 Struthio camelus YPM 112453 28 Struthio camelus YPM 112454 30 Struthio camelus YPM 112455 30 Struthio camelus .YPM 112456 31 Struthio camelus YPM 112457 3.1 Struthio camelus YPM 112458 32 Struthio camelus YPM 112459 34 Struthio camelus YPM 112460 35 Struthio camelus YPM 112461 36 Struthio camelus YPM 112462 37 Struthio camelus YPM 112463 37 Struthio camelus YPM 112464 38 Struthio camelus YPM 112465 38 Struthio camelus YPM 112466 41

Eudromia elegans YPM 112518 9 Eudromia elegans YPM 112519 10 Eudromia elegans YPM 112520 11 Eudromia elegans YPM 112521 11 Eudromia elegans YPM 112522 12 Eudromia elegans YPM 112523 13 Eudromia elegans YPM 112524 14 Eudromia elegans YPM 112525 15

Nothoprocta perdicaria YPM 112511 Nothoprocta perdicaria YPM 112512 Nothoprocta perdicaria YPM 112513 Nothoprocta perdicaria YPM 112514

276 + + + + + in rr i/n t-~ J3 o o o o o ->> ^ >^ >-. en en en m -—• •t •* •>* •*)- ^r i-.v-.-ti oo en J*2s CO 00 + + + + J£ c- o« + + + ^ m ~ ^Q as as 00 r- +_ _ s « » \o ^> o 000 -—-mm v£. en OOO m en ^—^ g, v£, ~—- en m « en •^f rr rj- T ^t ^r u z> m m 2 en so vo « m ON m en r-- 00 m rn m en en en

en rr unr-ocN'3-voooo o — • CN en en t/n^Or-00ON(Nm^r vO ON ,_

< vor~-ooo\o — N n t /i (NCNCMCNCNlCNCNrNCNCN ^-CNen^rinvor~-<— ooo\0^-.eNen'^-in>or-ooCT\0 — CNm^un •q-unvot--ooONO — (N r~r»-r-~r-r--r--r--t--r»r» OOOOOOOr-00~ — — — — — — — — — 2 222 at at at at at at at at

a a a a a a Q a CS Q- <3 R R R R R R R R R R R 00000000000000000000000000 <3 Q Q Q. Q .a0 .O .a0 .0 .O . «u ^cJ «ci 00000000000000000000000000 s B B £ B B 5 B S S B B a a o a a a Q Q a Q a c a Q *•o* o a a a ?3 a <3 Q a a 53 <3 y y y y y y y y y y y y y y y y y y y y y y y y y y •So3 -iS -iS -iS -iS -iS -iS -iS -iS «j «u su «u *J -R -R -R% > -R -R -R -R 8v ft^Q^C en c-> m cn en

o 2 :=: Nni-

sor-ooo\o—>tNrn'^}-v-)vor-ooo\0'— (Nm-*ir)^or~-oooo — OOONO—•c^mTri^voc~--oooso — cNm-^-in^ot--ooo\o. "ntr)ir)iovo^ovosOvovovovO'0^ot-~r-r-~r~-t^t^t-~t^r~t-~-oooo OOOOOOOOOOOOOOQOOOOOOOOOOO 00 t*-t>t^t~-r~t-~r~r-r-r-r~-r^r~r-r-r~r~t~-.t-~r-r~r~r>-

a a a Q a a a a a a a <3 53 53 53 53 53 53 53 53 53 53 53 B B S 6 B B B S B B S S B S B B B B B B S B 53 53 53 53 53 53 53 5353535353535353 53 53 • m to to to to to to to *J *5 t«5 *5 «j to to "5 to to ** •<%« *£2 *CJ ^»4 '*N« **^4 •*>* 5J «J 5j "--» *-•-«* ^ *ct •^^ ^ "**» *-— '^* •**« *-^ ^* to II ki k> o o o *-•« ^** »^»* ^C»i *-^ *~«* *-»» *-* *-»* "— *-*. *^* »*O^ •*~O* ^-» •*«-* O '*«O* O ^^O* o o ••» •N-* •*•« •S .a •?« 53 53 53 53 ik fc» ***V» ik i- ik ik ••-Ik• ik Ik ik ik Ik 53 ft) ft) **» Cft) ft) ft) ft) ft) ft) ft) ft) £ft> ft) ft> •*ft)* ft) ft) ft) ft) .SB >•« •K* •^ •*-* •*•* *•- •** ••>»* •** •""•i ft) •B -5 -5 .16 •cw ^ •K^* •K* c c c •»•* c cw *c> c •»-i .8k V .ik .Ik .Ik 53 **53» *-C*S 53 53 **» *5^3 53 ^5*3 53 53 53 53 53 53 *^53» 53 *5^3 53 S 53. 53 a a a a cs 5 13 Q 13 53 6 S S B B B B B B B B B B S B B S B B B B B O U U O U O kJ r r r r U O U o o os O o o o o o o o- o o o O O o O o 53 o „o o o u u u o u CO co CO CO to CO CO CO CO CO co CO CO CO CO CO CO CO CO CO CO CO CO + + + + + + + + + + + + + + + + in in AO so so so r- r— r- c-». oo J£ os as as oooooooooooooo o oooo — M^^tin^-ininmsoJ^it^ mmmc^mmmmmrom c^cAenenenenenenenenenenen • en en en en so oo en

O O en en (N(Nr<-iro-^f-rJ-u-iinsosor-~r-ooooosOSO^-CN(Nmri-rr^-iO>OSOSO & — ol en en Tj-n-'ininsosot~~t^oo ^-,,-^,-^,_,,-H,_H,___,^,_„_,_„„(N1(N(N(N(N(N(N(N{NCN

r-~r~r--r--t--r--r--t--

5 5 OOOOOOOOO -R-R-R-R-R-R-R-R-R & b: b a" ooooouooo a- a" Is a *R "-R ass (3 <3 S3 53 - J? .^ J? Je" •§" Js" J§* ^ Jt* a s a ^^k^ik&s^^&s I g g S S £>£^£>^£^£^ S S3 53 O o 2 3 S SJ o 8 a a »S«S~2-2»2-S-2i5»S o o o g O g ,^» .^* .a,&,ft,Q,ci,Q,sa!=«a^^l>5 QQQQCaaQQ g 3 3 R.RRRRRRRR g g g <3 <3 U 0 r^^i'r^'^i'^iT^T^'^'^ -R -R -R -R -R -R •R -R -R -R -R -R -R -R ***&l ^ 1V ^"8l ^- i **"l *** a, a. a, a. a, a, a, a. Phalacrocorax auritus RM 7109 18 37 Phalacrocorax auritus RM 7110 19 37 Phalacrocorax auritus RM 7111 19 37 Phalacrocorax auritus RM 7112 20 39 Phalacrocorax auritus RM 7113 20 38-40 Phalacrocorax auritus RM 7114 21.5 38 Phalacrocorax auritus RM 7115 23.5 40+ Phalacrocorax auritus RM 7116 23.5 40+

280 APPENDIX 2.

Rank order of element ossification Epibranchial for the Double-Crested Cormorant Postcranial axial {Phalacrocorax auritus). skeleton Cervical centra Phalacrocorax Thoracic centra Element auritus Synsacral centra Caudal centra Skull Pygostyle Basioccipital Cervical neural arch Exoccipital 14 Thoracic neural arch Supraoccipital Synsacral transverse Parasphenoid rostrum 5 processes Parasphenoid ala Caudal transverse Parasphenoid lamina processes Basisphenoid Synsacral arch Orbitosphenoid Cervical ribs 13 Prootic Dorsal ribs 6 Opisthotic Sternal ribs Epiotic Uncinate processes Squamosal 4 Sternum (body) Parietal 9 Laterocranial process Frontal 6 Laterocaudal process Lacrimal 7 Mesethmoid Forelimb Trabeculae Scapula 6 Nasal 4 Coracoid 10 Premaxilla 5 Furcula 3 Maxilla 5 Humerus 1 Palatine 4 Radius 2 Pterygoid 4 Ulna 3 Vomer Radiale Jugal 6 Ulnare Quadratojugal 4 Metacarpal II Quadrate 11 Phalanx 1 10 Ectethmoid Phalanx 2 Dentary 4 Metacarpal HI 4 Supra-angular 4 Phalanx 1 12 Angular 4 Phalanx 2 10 Splenial 5 Metacarpal IV 4 Prearticular Phalanx 1 Articular Mandibular Hindlimb Entoglossal Ilium 10 Basihyal Ischium 8 Urohyal Pubis 10 Ceratobranchial

281