Woodpeckers and Allies)

Home , Blythipicus, Campephilus, Campethera, Chrysocolaptes, Chrysophlegma, Colaptes

Coexistence, Ecomorphology, and Diversification in the Avian Family Picidae (Woodpeckers and Allies)

A Dissertation SUBMITTED TO THE FACULTY OF UNIVERSITY OF MINNESOTA BY

Matthew Dufort

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

F. Keith Barker and Kenneth Kozak

October 2015

Acknowledgements

I thank the many people, named and unnamed, who helped to make this possible. Keith Barker and Ken Kozak provided guidance throughout this process, engaged in innumerable conversations during the development and execution of this project, and provided invaluable feedback on this dissertation.

My committee members, Jeannine Cavender-Bares and George Weiblen, provided helpful input on my project and feedback on this dissertation. I thank the

Barker, Kozak, Jansa, and Zink labs and the Systematics Discussion Group for stimulating discussions that helped to shape the ideas presented here, and for insight on data collection and analytical approaches. Hernán Vázquez-Miranda was a constant source of information on lab techniques and phylogenetic methods, shared unpublished PCR primers and DNA extracts, and shared my enthusiasm for woodpeckers. Laura Garbe assisted with DNA sequencing.

A number of organizations provided financial or logistical support without which this dissertation would not have been possible. I received fellowships from the National Science Foundation Graduate Research Fellowship Program and the Graduate School Fellowship of the University of Minnesota. Research funding was provided by the Dayton Fund of the Bell Museum of Natural History, the

Chapman Fund of the American Museum of Natural History, the Field Museum of

Natural History, and the University of Minnesota Council of Graduate Students.

My work would have been impossible without the extensive collections of specimens and tissue samples housed in various natural history museums, and i the collectors, preparators, and curators who generate and maintain those collections. Tissue samples were generously provided by Janet Hinshaw at the

University of Michigan Museum of Zoology, Mark Robbins at the University of

Kansas Natural History Museum, and Paul Sweet at the American Museum of

Natural History. For access to specimens, I thank curators and collections managers at the National Museum of National History, the Burke Museum of

Natural History and Culture at the University of Washington, the University of

Kansas Natural History Museum, the University of Michigan Museum of Zoology, the Slater Museum of Natural History at the University of Puget Sound, the

Florida Museum of Natural History, and the US Fish and Wildlife Forensics

Laboratory. I thank the woodpecker phylogeneticists who came before me for making their data available in public repositories. The computationally intensive analyses presented here were made possible by the computing resources of the

CIPRES Science Gateway and the Minnesota Supercomputing Institute.

Finally, I thank Amber for always being there for me, and for believing in me even when I doubted myself. I thank Owen and Theo for making life more fun every day. And I thank all of my family and friends for their constant support.

Dedication

I dedicate this dissertation to Amber, Owen, and Theo, who remind me every day what’s most important.

iii

Abstract

Interspecific competition has well-documented effects on evolution in simple systems over short timescales. However, the effects in more complex communities, and over the timescales of continental radiations, are less clear. In this dissertation, I addressed the relationships between coexistence, morphological evolution, and diversification at multiple spatial and temporal scales, using birds in the family Picidae, which includes woodpeckers and the related piculets and wrynecks. Morphology of these birds is correlated with diet and foraging mode, allowing similarity in morphological measurements to be used as a proxy for ecological similarity. Members of Picidae occur throughout the Americas, Africa, and Eurasia, and local diversity ranges from a single species to up to 13. I first generated a phylogenetic hypothesis of Picidae, and then tested predictions from community ecology and macroevolutionary theory at the community level and across the entire global radiation of Picidae.

In Chapter 1, I inferred phylogenetic relationships among ~75% of extant species of Picidae, the most comprehensive molecular phylogeny of the family to date, using publicly available sequence data augmented by targeted collection of new data. While most results matched previous findings, a few species with new molecular data were placed in unexpected regions of the tree, and several genera as currently delineated appear to be paraphyletic. Relationships within most genera and previously described tribes were well resolved, but relationships among most major clades remained unclear. A number of tightly spaced iv branching events near the base of the family were not resolved with available data.

In Chapter 2, I evaluated the roles of ecological assortment of species and in situ trait evolution in driving trait distributions in communities of North American woodpeckers. I used multiple null models and metrics of community trait distributions to test for deviations from randomness in six communities including a total of 10 species. I recovered a signal of divergent displacement across all populations and all communities, suggesting local evolution away from morphologically most similar species. However, trait distributions in most individual communities did not differ from random expectations. In average size and overall morphology, one community showed evidence of species sorting for dissimilar traits. In the size-scaled shape data, I found evidence of divergent and convergent local evolution in one community each. These results suggest that trait differences are related to both species sorting and local evolution, but that other processes or a lack of statistical power prevent detection of effects in many communities.

In Chapter 3, I tested for relationships between coexistence and rates and modes of diversification and trait evolution across all Picidae. I used phylogenetic comparative methods to evaluate correlations among subclades of Picidae in relevant variables. I found strong and consistent positive correlations between geographic range overlap, rates of diversification, and rates of shape evolution— but not body size or overall morphological evolution. In addition, time-dependent v models of morphological evolution and diversification fit better to subclades with greater range overlap.

Taken together, these findings suggest that coexistence with similar species does affect evolution in Picidae. Shape evolution shows clear connections with coexistence in both the community-level and family-wide comparative analyses. The community analyses suggest that coexistence is the cause, rather than the consequence, of this trait evolution, as the trait changes are more spatially restricted and temporally recent than apparent coexistence among many of the species. Variation in diversification rates may be driven by other factors that covary with local diversity and evolution in body shape. Future work in this group should focus on solving the remaining puzzles in phylogenetic relationships, determining the contribution of body size and shape to competitive interactions, and understanding possible relationships of other variables with diversification, morphological evolution, and coexistence.

One online supplementary file (OSF) accompanies this dissertation: a spreadsheet containing the GenBank accession numbers or other sources for the

DNA sequence data used in Chapter 1.

Table of Contents

List of Tables ...... viii

List of Figures...... ix

Chapter 1: An augmented supermatrix phylogeny of the avian family Picidae reveals uncertainty deep in the family tree ...... 1 Introduction...... 2 Materials and Methods ...... 5 Results...... 16 Discussion ...... 24 Conclusions ...... 39

Chapter 2: Both species sorting and trait evolution explain non-random trait distributions in North American woodpecker communities ...... 48 Introduction...... 49 Materials and Methods ...... 55 Results...... 73 Discussion ...... 77 Conclusions ...... 88

Chapter 3: Diversification and trait evolution are correlated with coexistence in the avian family Picidae...... 103 Introduction...... 104 Materials and Methods ...... 108 Results...... 122 Discussion ...... 129 Conclusions ...... 141

Bibliography...... 155

Appendices...... 174 Appendix 1. Specimens measured by sex for each taxon for comparative analyses...... 174 Appendix 2. Results of PGLS regression fits to subclade variables for all models tested...... 178 Appendix 3. Map of global Picidae species richness...... 186

vii

List of Tables

Table 1.1. Specimens used for DNA sequencing ...... 40

Table 1.2. Genetic loci and primers used for sequencing...... 41

Table 1.3. Calibration points for divergence time estimation ...... 42

Table 1.4. Genetic loci used in phylogenetic analyses ...... 43

Table 2.1. Site localities and number of individuals measured by site...... 89

Table 2.2. Community metric significance values for overall morphology data .. 90

Table 2.3. Community metric significance values for size data ...... 91

Table 2.4. Community metric significance values for shape data...... 92

Table 2.5. Community metric significance values using the within-species population permutation (WSPP) ...... 93

Table 2.6. Comparison of community metrics from the within-species population permutation (WSPP) and the general permutation (GP) ...... 94

Table 2.7. Displacement values and results of statistical tests...... 95

Table 3.1. Fits of diversification and morphological evolution models to subclades ...... 143

Table 3.2. Results from PGLS regression fits to subclade variables for select models ...... 144

viii

List of Figures

Figure 1.1. DNA sequence coverage by taxon...... 44

Figure 1.2. Phylogenetic relationships of Picidae inferred from maximum likelihood analysis of the concatenated alignment of all loci in RAxML .... 45

Figure 1.3. Phylogenetic relationships of Picidae inferred from STAR analysis of RAxML nuclear gene trees and concatenated mtDNA analysis ...... 46

Figure 1.4. Divergence time estimation from BEAST ...... 47

Figure 2.1. Morphological measurements taken from woodpecker skeletons.... 97

Figure 2.2. Schematic of the general permutation (GP) model ...... 98

Figure 2.3. Schematic of the within-species population permutation (WSPP) model...... 99

Figure 2.4. Hypothetical example of displacement of a population in trait space ...... 100

Figure 2.5. Distribution of populations on the first two PCA axes of overall morphological data ...... 101

Figure 2.6. Distribution of populations on the first two PCA axes of shape data...... 102

Figure 3.1. Phylomorphospace plots of all taxa...... 146

Figure 3.2. Histogram of summed overlap values by taxon ...... 147

Figure 3.3. Results from BAMM diversification analysis for all Picidae ...... 148

Figure 3.4. Results from BAMM analyses of morphological evolution for all Picidae...... 149

Figure 3.5. Rates of morphological evolution versus geographic overlap by subclade ...... 150

Figure 3.6. Boxplots of parameter values from models fit to rates of morphological evolution and geographic overlap in Picinae...... 151

Figure 3.7. Diversification rate versus average summed range overlap and rate of shape evolution by subclade ...... 152

Figure 3.8. Boxplots of parameter values from models fit to diversification rates and geographic overlap in Picinae...... 153

Figure 3.9. Boxplots of parameter values from models fit to diversification rates and rates of morphological evolution in Picinae ...... 154

Chapter 1

An augmented supermatrix phylogeny of the avian family Picidae reveals

uncertainty deep in the family tree

Abstract

The accumulation of DNA sequence data in public repositories allows for phylogenetic inference on unprecedented taxonomic scales using supermatrix approaches. Careful analysis of available data allows strategic augmentation with new sequences in order to maximize taxonomic sampling and coverage of informative loci. I inferred relationships among 179 species (76%) in the avian family Picidae (woodpeckers, piculets, and wrynecks), using publicly available sequence data supplemented with targeted sequencing to increase species-level and locus-level sampling and maximize resolution. Results of these analyses generally corroborate previous molecular studies, with consensus on the membership of most genera and tribes. However, several newly placed taxa, including Dryocopus galeatus, Dendropicos obsoletus, and Guyana populations of Colaptes rubiginosus, show surprising affinities. As currently delineated, the genera Campethera, Dryocopus, and Melanerpes appear to be paraphyletic.

Relationships among major clades of Picidae remain poorly resolved, particularly among the three lineages of piculets, the unusual woodpecker genus Hemicircus, and the remaining woodpeckers, and among the major groups of true 1 woodpeckers (Picinae). If these deep relationships are to be resolved, phylogenomic approaches may be necessary.

1. Introduction

The avian family Picidae includes the charismatic woodpeckers (subfamily

Picinae), as well as the piculets (Picumninae and Nesoctitinae) and wrynecks

(Jynginae). The family is nearly global in distribution, and includes species with varying degrees of specialization in foraging modes (Winkler and Christie, 2002).

Woodpeckers are well known as ecosystem engineers, excavating tree cavities that are subsequently used by numerous other species of animals (Aitken and

Martin, 2007; Cockle et al., 2011; Martin and Eadie, 1999). The group also includes striking examples of visual mimicry and convergent evolution (Cody,

1969; Prum, 2014; Prum and Samuelson, 2012; Weibel and Moore, 2005). A comprehensive species-level phylogeny of the family is needed in order to more fully understand their biogeography, convergence, morphological evolution and the evolution of foraging modes and cavity-excavating behavior.

Early classification schemes based on anatomical characters (Goodge, 1972;

Swierczewski and Raikow, 1981) and on varied behavioral, ecological, and plumage characters (Goodwin, 1968; Short, 1982) generated two markedly different taxonomies of Picidae; the discrepancies between schemes were largely based on differing opinions about the frequency of convergence in plumage and 2 ecological characters. More recently, a series of studies based on DNA sequence data have clarified the major evolutionary groups within Picidae, as well as species-level relationships within select subclades (Benz et al., 2006; Fuchs et al., 2007; Fuchs et al., 2013; Moore and DeFilippis, 1997; Prychitko and Moore,

1997; Webb and Moore, 2005; Winkler et al., 2014). One major finding has been that convergence, particularly in plumage, has occurred quite frequently; results from molecular phylogenetics generally support the historical classifications that assumed convergence common. The consensus from these studies includes 5 major clades generally supported as monophyletic: Jynginae (the wrynecks),

Picumninae (the piculets excluding Nesoctites), Malarpicini, Dendropicini, and

Chrysocolaptes+Blythipicus, and three genera whose placement is not well resolved: Nesoctites, Hemicircus, and Campephilus. However, relationships among these groups, and among many of the subclades within them, remain poorly resolved.

To date, all DNA-based studies of Picidae have either sampled broadly across genera (Benz et al., 2006; DeFilippis and Moore, 2000; Fuchs et al., 2007; Fuchs et al., 2013; Moore and DeFilippis, 1997; Prychitko and Moore, 1997, 2000;

Webb and Moore, 2005; Winkler et al., 2014) or more intensively within particular portions of the family (Azevedo et al., 2013; Benz and Robbins, 2011; Fleischer et al., 2006; Fuchs et al., 2006; Fuchs et al., 2008; García-Trejo et al., 2009;

Honey-Escandón et al., 2008; Moore et al., 2011; Moore et al., 2006; Overton 3 and Rhoads, 2006; Pons et al., 2011; Weibel and Moore, 2002a, b; Winkler et al.,

2005). However, no molecular study focused on Picidae has included extensive species-level sampling of all genera: the most inclusive study to date analyzed just 65 of the ~235 currently recognized species (Winkler et al., 2014).

Sufficient molecular data are now available to attempt a species-level analysis of

Picidae using supermatrix methods. Accumulation of DNA sequence data in public archives has opened avenues to research combining and synthesizing data sets collected by many research groups for varied purposes. Supermatrix phylogenetic analyses use these accumulated data to infer evolutionary relationships among large numbers of taxa, often with sparse sampling of loci and only partial overlap among taxa (de Queiroz and Gatesy, 2007; Driskell et al., 2004; Thomson and Shaffer, 2010). While there is some disagreement about the effects of missing data in such sparse matrices on inference of phylogenetic relationships (e.g. Lemmon et al., 2009), numerous simulation and rarefaction studies indicate that with appropriate analytical techniques the benefits of adding data generally outweigh problems caused by missing data at poorly sampled loci, especially if at least one locus is sampled for most taxa (Jiang et al., 2014; Roure et al., 2013; Wiens, 2006; Wiens and Moen, 2008; Wiens and Morrill, 2011). One additional benefit of supermatrix approaches is that they allow strategic supplementation of available data with additional sequence data, in order to maximize sampling of taxa or informative loci. 4

The current accumulation of abundant DNA sequence data for Picidae, with broad sampling across the clade, and denser sampling within many subclades, is ideal for the application of supermatrix approaches. Here, I infer relationships among nearly three-quarters of the currently recognized species of Picidae, using public sequence data, supplemented with newly collected data. I present the most complete molecular phylogeny of Picidae to date, and place several enigmatic taxa for the first time. This phylogeny generally recovers similar relationships to those from previous studies, but raises new hypotheses and questions about deep relationships within the family. It also brings to light limitations of the currently available data and highlights areas requiring further study. This phylogeny will be invaluable for analyses of diversification, trait evolution, and biogeography within this diverse and ecologically important group.

2. Materials and Methods

2.1. Taxon Sampling and Public Sequence Data

The taxonomy of Picidae is currently in flux, as genera are being modified to account for molecular phylogenetic results, and many lineages previously considered subspecies are being recognized as full species after additional study. In general, I follow the classification of Winkler et al. (2014), which includes revisions based on recent work (Benz and Robbins, 2011; Collar, 2011;

García-Trejo et al., 2009; Moore et al., 2011; Pons et al., 2011) I retain the genus 5 name Veniliornis (subsumed in Dryobates by Winkler et al.), and split

Leuconotopicus from Dryobates. I also include several taxa not given species rank by Winkler et al., for which sequence data or tissue samples are available from phenotypically distinctive populations. A recent taxonomic revision elevated a number of forms to species status based on an analysis of external characters

(del Hoyo et al., 2014; Tobias et al., 2010). These taxa were not included because their validity is unclear, but future molecular studies should address their status.

I downloaded all DNA sequence data from GenBank

(http://www.ncbi.nlm.nih.gov/genbank/) for Picidae and outgroups on July 31,

2014. I used clusters of sequences identified by PhyLoTa as a starting point for locus identification; PhyLoTa uses BLAST searches and clustering algorithms to identify groups of similar sequences (Sanderson et al., 2008). I checked definitions of sequences in these clusters for possible anomalous sequences, such as nuclear mitochondrial pseudogenes and GenBank definition discrepancies. I augmented these clusters with data added to GenBank after the most recent PhyLoTa update (GenBank Release 194), and identified additional clusters based on GenBank sequence definitions. I also obtained sequences from previous studies where the sequence data is not yet accessioned in

GenBank, by personal communication (García-Trejo et al., 2009) or by transcribing sequences from published manuscripts (Cicero and Johnson, 1995). 6

I excluded loci with data for fewer than 4 ingroup taxa, as these are unlikely to substantially increase resolution of relationships. I used members of the families

Indicatoridae (honeyguides) and Ramphastidae (toucans) as outgroups.

Indicatoridae and the clade containing Ramphastidae have consistently been recovered as successive sister groups to Picidae (Ericson et al., 2006; Hackett et al., 2008; Lanyon and Zink, 1987). In order to maximize outgroup sequence coverage, data were included from multiple species of Indicator for Indicatoridae, and multiple species of Pteroglossus for Ramphastidae.

One of the primary challenges of supermatrix analyses is reconciling variation in taxonomy across data sets (Thomson and Shaffer, 2010). Authors may use different names to refer to the same taxon, and changes in species limits over time may render taxon names ambiguous with regard to current taxonomy. To account for these possible ambiguities, I used multiple sources to generate a list of taxa for which names and/or species limits have changed in the past 30 years

(Winkler and Christie, 2002; Winkler et al., 1995; http://avibase.bsc-eoc.org/); sequences for these taxa might correspond to one of several currently recognized species. For all sequences attributed to taxa on this list, I used additional data where possible to assign the source specimens unambiguously to a single taxon. Most such ambiguities were resolvable using subspecies identity, locality information, or additional specimen data. Sequences that could not be unambiguously assigned were excluded from analyses. 7

Another potential pitfall of supermatrix analyses is the possible variation in data quality in public repositories. Several sequences were excluded based on judgments made by previous authors, or on preliminary analyses. Benz and

Robbins (2011) concluded that Celeus elegans contained mitochondrial DNA introgression from Celeus lugubris; I used C. elegans sequences only from individuals outside the area of inferred introgression. Fuchs et al. (2007) discovered that a CYTB sequence attributed to Geocolaptes olivaceus was actually from Dryobates minor; this sequence has since been removed from

GenBank, and was not included in this study.

2.2. Collection of Additional Sequence Data

In order to increase taxonomic coverage and improve the resolution of phylogenetic inference, I sequenced 21 additional taxa for a subset of six loci. I selected taxa based on three criteria: (1) those identified in preliminary analyses as rogue taxa by RogueNaRok (Aberer et al., 2013), (2) taxa spanning poorly- resolved nodes with little or no sequence data, and (3) taxa of uncertain relationships for which tissue samples but no DNA sequence data was available.

A list of specimens utilized, with locality and voucher information, is in Table 1.1.

I extracted DNA from tissue samples using DNEasy extraction kits (Qiagen). I selected loci to amplify and sequence based on the extent of taxonomic coverage 8 for each locus in the existing public data. Fragments were amplified using polymerase chain reaction (PCR), in 25 µL reaction volumes. Sequenced loci and corresponding PCR primers are listed in Table 1.2. Primer pairs were tailed with

M13 forward and reverse primer sequences to facilitate cycle sequencing.

Reactions included approximately 50-200 ng of target DNA, 10 pmol of each primer, and 12.5 µL of GoTaq Green Master Mix (Promega). All reactions used cycles of 30s denaturation at 95°C, 30s annealing, and 60s extension at 72°C, followed by 7 min extension at 72°C. Annealing temperatures followed a touchdown procedure, with 5 cycles each at 58°C, 56°C, and 54°C, followed by

25 cycles at 52°C.

Purification and sequencing were performed by Beckman Coulter Genomics in

Danvers, MA. Amplicons were purified with Solid Phase Reversible

Immobilization (Beckman Coulter), sequenced with ABI BigDye Terminator v. 3.1

(Applied BioSystems) using standard M13 primers, cleaned with Agencourt

CleanSEQ (Beckman Coulter), and analyzed on an ABI Prism 3730xl (Applied

Biosystems). All amplicons were sequenced in both directions, and sequencing reads were assembled and edited in Sequencher v. 5.2 (Gene Codes). Putative heterozygous sites and ambiguous base calls were assigned the appropriate

IUPAC ambiguity codes.

A number of nuclear intron sequences contained heterozygous indel regions.

These sequences were resolved using Indelligent v. 1.2 (Dmitriev and Rakitov,

2008, 2008-2014), and the allele lacking autapomorphies relative to other Picidae was used in analyses. All mitochondrial protein coding sequences were tested for the presence of open reading frames and absence of internal stop codons using the ExPASy Translate tool (Artimo et al., 2012); this helps to ensure that sequences are target mitochondrial loci, and not nuclear pseudogenes. Biases in base composition can mislead phylogenetic inference, as taxa with similar base composition are grouped together by “compositional attraction” (Foster, 2004;

Mooers and Holmes, 2000; Weisburg et al., 1989). I tested each sequence for significant deviations from average locus-level base composition following Gruber et al. (2007). I calculated the empirical average base composition for each locus using Paup* v. 4.0 (Swofford, 2003), and conducted χ2 goodness-of-fit tests for each sequence against this average, with sequential Bonferroni correction for multiple tests.

2.3. Sequence Selection and Alignment

After compiling all existing and new sequence data, I selected at most one sequence per taxon-locus combination to use for analyses. Where multiple sequences were available for a single taxon and locus, I included the sequence with greatest coverage as identified by BLAST alignment to a reference sequence for each locus. Ties were broken randomly, or, where possible, to 10 retain multiple loci from the same individual specimen. In a few cases, multiple non-overlapping sequences from the same taxon covered different portions of the locus; these were combined to form a chimeric sequence.

Sequences were aligned by locus using PASTA, which aligns sequences by iteratively inferring a guide tree, breaking that tree into subsets, aligning sequences within subsets, and merging the subset alignments (Mirarab et al.,

2014). Alignments were examined for obvious errors and ambiguously aligned regions in Seaview v. 4 (Gouy et al., 2010). Ambiguous alignment regions were manually trimmed prior to downstream analyses.

All alignments were further trimmed to exclude regions with fewer than four taxa.

Many mitochondrial DNA sequences included adjacent loci, and locus-by-locus alignment could therefore duplicate these regions in analyses. To prevent such duplication, I trimmed all mtDNA alignments to a reference sequence based on the complete mitochondrial genome of Dryocopus pileatus (Gibb et al., 2007).

Trimmed alignments for each locus were concatenated to generate full alignments for all loci, all mtDNA loci, and all nuDNA loci, with all missing sequences coded as missing data. Full alignments were submitted to TreeBASE

(http://purl.org/phylo/treebase/phylows/study/TB2:S18109). Trimming and concatenation was performed in R (R Core Team, 2015), using code written by the author (available at https://github.com/mjdufort). 11

2.4. Phylogenetic Inference

I inferred relationships among taxa using maximum likelihood (ML), Bayesian inference, and species-tree estimation from individual gene trees.

I estimated phylogenetic relationships using maximum likelihood in RAxML v. 8.0

(Stamatakis, 2014). I first analyzed data for each locus independently, in order to identify any strongly conflicting signal among loci, which may indicate anomalous gene histories or data errors. I used the GTR+Γ model of DNA substitution

(GTRGAMMA in RAxML), and evaluated support for the topology of the highest- scoring ML tree with 100 rapid bootstrap replicates per locus. These gene trees were also used in the species tree analyses described below.

I analyzed concatenated datasets in RAxML as described above, with alignments partitioned by locus. I also ran analyses with protein-coding sequences partitioned by codon position. In order to more broadly search tree space, I performed 100 independent runs with random-order stepwise addition maximum parsimony starting trees. Support for the topology of the highest-scoring ML tree was evaluated with 1000 rapid bootstrap replicates.

I inferred phylogenetic relationships in a Bayesian framework in MrBayes v. 3.2

(Ronquist et al., 2012). I partitioned alignments by locus, and estimated per-locus 12 models of sequence substitution using jModelTest v. 2.3 (Darriba et al., 2012;

Guindon and Gascuel, 2003), with best-fit models chosen by Bayesian

Information Criterion (BIC). Analyses were also run with protein-coding sequences partitioned by codon position. The tree was constrained to be ultrametric, with a relaxed molecular clock and inverse-gamma-distributed rate variation. Analyses consisted of 4 runs of 4 chains each for 40,000,000 generations, sampled every 10,000 generations, with the first 10,000,000 generations discarded as burn-in. Runs were evaluated for stationarity and convergence in Tracer (Rambaut et al., 2013) and AWTY (Nylander et al., 2008).

Analyses of concatenated alignments in RAxML and MrBayes were run on resources of the CIPRES Science Gateway (Miller et al., 2010).

Results of concatenated analyses may be misleading in the presence of conflict among gene trees due to coalescent variation (Edwards, 2009; Kubatko and

Degnan, 2007; Pamilo and Nei, 1988). To account for this possibility, I inferred species trees in STAR, which uses the ranks of coalescent events across gene tree to estimate a species history (Liu et al., 2009). Input for STAR analyses was individual gene trees from RAxML analyses of nuclear loci, plus the maximum likelihood tree from RAxML analysis of the concatenated mitochondrial DNA.

Mitochondrial DNA is typically inherited as a single linkage block, and should therefore be included only once in species tree analyses. In addition to using the single maximum-likelihood tree for each locus from RAxML, I analyzed the 13

RAxML bootstrap replicates for each locus with STAR’s multilocus bootstrapping method. STAR analyses were run on the STRAW web server (Shaw et al., 2013).

Fuchs et al. (2013) found strongly supported conflicts between FGB-I7 and mitochondrial and other nuclear loci. They attributed this conflict to an inferred hybridization event between the Campephilus lineage and the ancestor of

Melanerpes and Sphyrapicus. To address possible conflicting histories among these loci, I ran all multilocus analyses with and without FGB-I7.

One potential issue in phylogenetic inference is the unstable placement of so- called “rogue taxa” (Sanderson and Shaffer, 2002; Wilkinson, 1996); this may be amplified in supermatrix analyses due to the extensive quantity of missing data and varied overlap in characters among taxa (Thomson and Shaffer, 2010). I used RogueNaRok (Aberer et al., 2013) to identify rogue taxa using bootstrap replicates (for ML analyses) or samples from the post-burnin posterior distribution

(for Bayesian analyses), and re-ran analyses after removal of rogues.

2.5. Time Calibration

Supermatrix analyses may be subject to branch-length mis-estimation due to the varying quantities of missing data across taxa, and the ways that missing data is handled in phylogenetic inference. I used multiple approaches to calibrate trees to relative and absolute time. 1 4

First, I calibrated the maximum-likelihood tree from the analysis of concatenated data in RAxML using non-parametric rate smoothing in r8s v. 1.8 (Sanderson,

2003). Each run included 10 random starting points to minimize the probability of recovering local optima. The tree was calibrated to both relative time (root age set to 1), and to absolute time using calibrations described below.

Second, I analyzed the concatenated alignment of all loci in BEAST v. 1.7.5

(Drummond et al., 2012), using the RAxML maximum-likelihood tree as a topological constraint. This uses the sequence data to infer branch lengths, while maintaining the topology inferred via maximum likelihood. The alignment was partitioned by locus, with a single topology, but all other parameters free to vary among partitions. Each partition was assigned the DNA substitution model inferred in jModelTest, as described above for MrBayes analyses. As with the r8s analyses, I calibrated the tree both to relative time (root age set to 1) and to absolute time using the calibration points described below. I conducted 2 independent runs of 40,000,000 generations each, sampling every 10,000 generations, with the first 20,000,000 generations discarded as burnin.

The fossil record of Picidae is limited. However, a small number of fossils have been placed phylogenetically within the family based on morphological synapomorphies. I used these fossils, node age estimates from previous studies, 15 and one biogeographic split as calibration points in divergence time estimates.

Table 1.3 shows the calibration points and how they were used in r8s and

BEAST. In general, fossil calibrations were assigned to the stem age of the most inclusive clade to which they could be assigned. r8s analyses used fossils as minimum-age constraints; in BEAST analyses they were used to set informative lognormal priors on node ages. I also used the biogeographic calibration of the split between Sasia abnormis and S. ochracea, estimated by Fuchs et al. (2007) to be between 4.5 and 5.5 Ma based on the formation of large seaways in the

Isthmus of Kra. Finally, I used estimates of the Picidae-Indicatoridae and Picidae-

Indicatoridae-Ramphastidae divergences from several sources (Moore and

Miglia, 2009) to set minima and maxima (r8s analyses) or informative normal priors (BEAST analyses) on the corresponding node ages.

3. Results

3.1. Sequence data

Clustering of GenBank and other public data resulted in a total of 25 loci for 172

Picidae taxa and 2 outgroups. These loci and their characteristics are listed in

Table 1.4. New DNA sequence data were collected for 7 species with no publicly available data. Data were also collected for 14 species to supplement existing data in public repositories. GenBank accession numbers for newly collected sequences are KT455620 - KT455706 (see online supplementary file).

No evidence of nuclear pseudogenes was found in the mitochondrial loci. The

COI sequence of Jynx torquilla showed significant divergence from average base composition (p = 3·10-5); divergence in a single taxon at a single locus is not expected to substantially impact phylogenetic results. Several sequences were excluded from analysis based on unusual behavior in single locus gene trees.

The sequences of Melanerpes striatus and Xiphidiopicus percussus obtained by

Overton and Rhoads (2006) showed extraordinarily long terminal branches in the cytochrome b (CYTB) gene tree. Inspection of these sequences revealed an unusual number of nucleotide and amino acid autapomorphies, suggesting that sequencing errors may be present, or that these sequences may represent nuclear mitochondrial pseudogenes (though the sequences do not contain stop codons or base-compositional discrepancies). These sequences were therefore excluded from analyses. ND2 sequences from Campephilus robustus and

Campethera abingoni appeared in unexpected positions in the single-locus gene tree. Further inspection revealed that these sequences were identical to those from Dryocopus javensis and Dryocopus pulverulentus, respectively, and likely due to contamination; they were excluded and analyses re-run. These problem data highlight the importance of inferring gene trees based on single loci, even in large supermatrix analyses, as such erroneous data could lead to incorrect inference of relationships.

A total of 1438 sequences were selected for analysis. GenBank accession numbers, museum voucher numbers, or alternate sources for all sequences used are available in the accompanying online supplementary file. Figure 1.1 shows coverage by locus across all taxa and loci. The alignment of all 25 loci included

68% missing data, while the mtDNA alignment had 57%, and the nuDNA alignment 66%. The mean number of loci per taxon was 8.0 (SD=6.2); mean number of taxa per locus was 57.5 (SD=40.3)

3.2. Maximum-likelihood inference

RAxML analyses of concatenated data recovered a generally well-resolved tree.

Analyses of mitochondrial loci, nuclear loci, and all loci combined were generally concordant, with conflicts only at weakly supported nodes. I therefore focus on results from the combined analysis (Figure 1.2). A number of major clades were recovered with strong support. Jynginae were found sister to the remaining

Picidae. Verreauxia+Sasia and Picumnus were both strongly supported as monophyletic, though the monophyly of Picumninae received only moderate bootstrap support. Hemicircus and Nesoctites were both placed basal to all other woodpeckers, but relationships among those three lineages were poorly resolved. Within Picinae, a number of major clades were recovered with strong support. These include Chrysocolaptes+Blythipicus, Campephilus, the

Malarpicini, and the Dendropicini (sensu Webb and Moore, 2005). Relationships among these major clades were poorly resolved, with only moderate support for 18 a sister relationship of Campephilus and Chrysocolaptes+Blythipicus. The

Malarpicini were further divided into a (Dryocopus, (Celeus, (Colaptes, Piculus))) clade, a ((Micropternus, Meiglyptes), (Gecinulus, Dinopium)) clade, and a clade containing Picus, Campethera, and Geocolaptes. The Dendropicini were divided into Melanerpes+Sphyrapicus and a clade containing Picoides, Leiopicus,

Dendropicos, Dendrocopos, Dryobates, Leuconotopicus, and Veniliornis.

Several taxa were placed for the first time in uncontroversial positions on the tree: Jynx ruficollis was sister to J. torquilla, Campephilus magellanicus and C. robustus clustered with other Campephilus, Dryocopus pulverulentus and D. fuliginosus both clustered with other Dryocopus species formerly in the genus

Mulleripicus. D. fuliginosus and D. funebris were previously considered conspecific; my analyses suggest they are not each other’s closest relatives.

Dryocopus javensis was placed with other “core” Eurasian Dryocopus, and

Chrysocolaptes lucidus was placed with other members of that genus. Guyana specimens of Colaptes rubiginosus were placed sister to Colaptes atricollis with strong support, rather than with other lineages of the C. rubiginosus superspecies. Chrysophlegma miniaceum clustered with other Chrysophlegma species, and not with the members of Picus, in which the Chrysophlegma species were formerly included. The enigmatic Dendropicos obsoletus, which has at times been placed in Dendrocopos, was found to be well within the genus

Dendropicos, sister to several species more characteristic of that genus. 19

Most genera recognized by Winkler et al. (2014) received strong support for monophyly; however, several genera were recovered as paraphyletic. Dryocopus galeatus clustered with Celeus with strong support, rather than with other

Dryocopus. Campethera nubica was placed with other members of Campethera, but this clade also included Geocolaptes olivaceus; Geocolaptes has previously been placed sister to a more limited sampling of Campethera (Benz et al., 2006;

Fuchs et al., 2007; Fuchs et al., 2013; Fuchs et al., 2008; Webb and Moore,

2005). Finally, Melanerpes striatus was sister to the remaining Melanerpes and all members of Sphyrapicus, though this placement received only weak bootstrap support.

3.3. Bayesian Inference

Assessment of MrBayes runs in Tracer and AWTY indicated that they had reached stationarity and convergence after 10,000,000 generations of burnin. As with RAxML trees, conflict among the mitochondrial DNA, nuclear DNA, and full concatenated data set was generally limited to weakly supported nodes. I therefore present results from the full analysis. The 50% majority rule consensus tree was well resolved, with the majority of nodes recovered with posterior probability of 1 (posterior probabilities shown on RAxML tree topology in Figure

1.2). The tree topology was generally concordant with the ML analysis, though many nodes that showed strong support in the Bayesian analyses were 20 unresolved or only weakly supported in the ML analyses. The monophyly of

Picumninae was strongly supported, with Picumnus sister to Verreauxia+Sasia.

Hemicircus and Nesoctites were found sister to each other, though with a posterior probability of only 0.75; regardless of their relationship to each other, these two species were recovered as the sister taxa of the remaining Picinae.

Campephilus was placed sister to Chrysocolaptes and Blythipicus in a monophyletic Megapicini, which was sister to a clade composed of Dendropicini and Malarpicini.

Most of the genus-level findings listed for the ML analyses were found in the

Bayesian analyses, with much stronger support in many cases. Dryocopus galeatus was placed within Celeus, with posterior probability of 1, though its specific affinities were unclear. Geocolaptes olivaceus was nested within

Camepthera. Melanerpes striatus, Sphyrapicus, and the remaining Melanerpes formed a three-way polytomy. One surprise was the recovery of Leiopicus mahrattensis sister to Dendropicos, with posterior probability 0.94. This species was placed with other Leiopicus in the ML analyses, though with only weak bootstrap support (60%).

3.4. Species tree analyses

Two species that were represented only in 1-2 nuclear gene trees had problematic placement in the STAR analyses, leading to negative branch lengths 21 and poor resolution of relationships among the remaining taxa. These species,

Campephilus robustus and Colaptes chrysoides, were removed, and analyses re-run. Following their removal, the optimal tree inferred by STAR had a topology very similar to those in the ML and Bayesian analyses (Figure 1.3). However, many nodes received only weak multilocus bootstrap support. Few clades had

100% bootstrap support; those that did included Jynx,

Picumninae+Nesoctites+Picinae, Sasia+Verreauxia, Picinae excluding

Hemicircus, and the genera Sphyrapicus, Chrysophlegma, and Campephilus.

The generic placements of newly analyzed species were entirely in agreement with concatenated analyses, though with much lower support values.

3.5. Analytical variations

RogueNaRok identified 13 and 8 species as rogues in the ML and Bayesian analyses, respectively. Rogue taxa were generally consistent across ML and

Bayesian analyses; in several cases one of a pair of closely related taxa were identified as rogues in each analysis. Rogues included Campephilus robustus;

Chrysocolaptes strictus; several species of Melanerpes , including M. striatus, M. herminieri, M. hypopolius, and M. superciliaris; Dendrocopos major (Bayesian) or

D. himalayensis and D. noguchii (ML); Veniliornis kirkii (Bayesian) or V. cassini

(ML); Picoides temminckii (ML only); and Picumnus lafresnayi (Bayesian only).

Trees inferred with and without rogues had only one topological conflict: the weakly supported placement of Leuconotopicus borealis, either sister to other 22

Leuconotopicus or to Veniliornis. Support values generally increased for relationships in the region of the tree close to the rogues, but were otherwise unchanged. Because of these strong similarities, I present trees including the rogue taxa, with the low support values appropriately indicating uncertainty about their placement.

Analyses with protein coding loci partitioned by codon position were nearly identical to those partitioned only by locus, and the few conflicts were at nodes within genera that were weakly supported both with and without partitioning by codon position. Only analyses partitioned by locus are presented. Similarly, analyses with and without FGB-I7 were nearly identical; only a few conflicts were detected, all were at weakly supported nodes, and most were within genera. In

ML analyses, removing FGB-I7 reverses the placement of Nesoctites and

Hemicircus, but this placement is weakly supported in all cases. Removing FGB-

I7 did not affect the placement of Campephilus; it was sister to

Chrysocolaptes+Blythipicus in all analyses, though with only moderate support.

3.6. Divergence time estimation

Time calibration of the ML tree in BEAST yielded estimates of divergences consistent with previous work (Figure 1.4; Fuchs et al., 2013). Parameter values converged within and between runs, and ESS values for most parameters exceeded 200. Node ages from analyses with and without internal calibrations 23 were highly concordant (Pearson’s r for node ages = 0.997), indicating that the calibrations do not conflict strongly with relative timing of splits based solely on sequence divergence, though posterior density intervals were much wider without internal calibrations.

Calibration to absolute time in r8s yielded results very similar to those from

BEAST (Pearson’s r of node ages = 0.989). Runs from multiple random starting points converged on similar likelihood scores and parameter values. Runs with and without internal calibrations yielded similar results (Pearson’s r for node ages

= 0.987).

4. Discussion

This study provides the most complete estimate of woodpecker phylogenetic relationships to date. Phylogenetic relationships were broadly consistent across different analytical techniques and subsets of the data. I recovered relationships that generally agree with previous studies, though relationships of several newly placed species were surprising, and multiple genera appear to be paraphyletic despite recent taxonomic revisions. The most interesting result may be the previously underexplored uncertainty about relationships deep in the tree, calling into question the monophyly of several major groups for the first time. This uncertainty is exposed primarily by the more extensive taxon sampling in this

24 study, but in several cases appears to have been overlooked by previous authors despite their results showing low support or conflict with historical groupings.

4.1. Comparisons across data sets and analytical methods

Relationships recovered from analyses of different data sets were broadly concordant. Gene trees inferred from single loci had variable relationships, as expected from varied taxon sampling, coalescent variation, and lack of strong signal in many of the loci. Results from the maximum likelihood, Bayesian, and species-tree analyses were strikingly similar, with highly concordant topologies across the three analyses. Most exceptions to this consistency were at poorly supported nodes. In addition, while the absolute support values varied widely across the three analyses, from very low in the STAR tree, to intermediate in the

RAxML tree, to very high for Bayesian posterior probabilities, the relative ranks of support for nodes were remarkably similar. Pairwise Pearson correlation coefficients of support values for shared nodes ranged from 0.34 for the ML and

STAR trees to 0.68 for the ML and Bayesian trees (all p<0.0001). The variation in absolute support values is likely due to the different ways in which the analyses estimate uncertainty in the data; Bayesian posterior probabilities are very often higher than bootstrap support values (Alfaro and Holder, 2006; Rannala and

Yang, 1996), whereas concatenated analyses overestimate support from a species tree perspective (Kubatko and Degnan, 2007). However, the robustness of results to the different analytical methods suggests that, at a minimum, the 25 clades strongly supported in all three analyses are not attributable to method- specific artifacts.

Missing data did not appear to cause major problems – despite the much larger quantity of missing data in this study vs. the original studies many of these sequences came from, most of the relationships recovered were similar to those found in original studies (where taxon sampling allows comparison), frequently with concordant support values.

4.2. Implications for woodpecker relationships

The results of this and previous studies make it clear that numerous historical groupings within Picidae were or are paraphyletic (Benz et al., 2006; Fuchs et al.,

2008; Moore and DeFilippis, 1997; Moore et al., 2011; Moore et al., 2006; Webb and Moore, 2005; Weibel and Moore, 2002a; Winkler et al., 2005). A recent taxonomic revision (Winkler et al., 2014) proposed genus names based on results from recent molecular phylogenetic studies; the monophyly of nearly all of these genera is supported by my analyses (in contrast to many of the previously named genera), indicating that our understanding of woodpecker relationships is stabilizing. However, uncertainty persists regarding the relationships among many genera and higher-level groupings; I address these in detail below.

Jynginae was consistently recovered as sister to all other Picidae. This was strongly supported in all analyses. This study includes the first molecular placement of Jynx ruficollis; its relationship to J. torquilla, which was expected based on the high morphological similarity between the two species, was confirmed.

My results cast doubt on relationships among the Picinae and piculets. The discovery that piculets are paraphyletic, with Nesoctites resolved as the sister lineage to Picinae (Benz et al., 2006), was particularly notable, though it was suggested by previous authors based on anatomical characters (Goodge, 1972).

Shortly thereafter, Fuchs et al. (2007) found Hemicircus to be the sister group to all other woodpeckers, and Nesoctites sister to all Picinae, though this was weakly supported. Hemicircus and Nesoctites have since been assumed to be successive sister taxa to the Picinae, with the remaining piculets sister to Picinae

+ Nesoctites (e.g. Winkler et al., 2014). The analyses reported here suggest that much remains to be resolved about these basal relationships within Picidae.

First, Picumnus and Sasia+Verreauxia are not strongly supported as monophyletic; this grouping has posterior probability of 1 in the Bayesian analysis, but is recovered in only 75% (RAxML) and 61% (STAR) of bootstrap replicates. Alternate topologies generally place Picumnus sister to Nesoctites +

Picinae. Winkler et al.’s (2014) tree shows Sasia as the sister group to Nesoctites

+ Picinae, with 99% bootstrap support. Surprisingly, the authors note this result 27 without comment; if an accurate representation of relationships, this would render their tribe Picumninae paraphyletic. Moving to Nesoctites, its exact placement is similarly unclear. In my analyses, both STAR and Bayesian analyses placed

Nesoctites sister to Hemicircus. This may be an artifact of the sparse supermatrix approach. Only two loci, ND2 and FGB-I7, are available for both Nesoctites and

Hemicircus; both gene trees show weak support for Hemicircus sister to remaining Picinae, with Nesoctites sister to all Picinae. The maximum-likelihood tree placed Nesoctites sister to Picinae, with Hemicircus the sister group to all other Picinae, but this received only 21% bootstrap support, and a

Nesoctites+Hemicircus clade was recovered in 65% of bootstrap replicates. The monophyly of Picinae (excluding Hemicircus) is clearer, with 93% (ML) and 100%

(STAR) bootstrap support and Bayesian posterior probability of 0.97. A clade comprising Nesoctites and Hemicircus is unexpected in light of their plumage and structural differences, the apparent similarities between Hemicircus and other woodpecker species, and claimed synapomorphies of all Picinae inclusive of

Hemicircus (Benz et al., 2006; Fuchs et al., 2007; Webb and Moore, 2005).

However, convergence in plumage and ecological specialization are widespread in Picidae (Benz et al., 2006; Moore et al., 2006; Weibel and Moore, 2005). Given the uncertainty regarding their relationships, the internodes separating

Nesoctites, Hemicircus, and Picinae are probably quite short, and convergence may explain the phenotypic similarity of these taxa. These short branches, fairly deep in the tree, may require phylogenomic approaches to resolve. 28

Three main groupings within Picinae have been named as tribes: Megapicini,

Malarpicini, and Dendropicini (Webb and Moore, 2005); Winkler et al. (2014) renamed these same groupings Campephilini, Picini, and Melanerpini to correspond to named genera. All of my analyses of the full data set place the genus Campephilus within Megapicini, though with varying support (ML: 72%;

STAR: 65%; MrBayes: 1.0). Given the moderate bootstrap support values, this relationship may best be viewed as probable but not yet certain. Consequently,

Campephilus should probably be considered a separate grouping, as it has alternatively been recovered as sister to Blythipicus and Chrysocolaptes, as sister to all Picinae excluding Hemicircus, or as sister to

Melanerpes+Sphyrapicus (Fuchs et al., 2013).

Fuchs et al. (2013) contended that placement of Campephilus has been confused by a historical hybridization event with melanerpine woodpeckers.

While one locus (FGB-I7) strongly supports a

Campephilus+(Melanerpes+Sphyrapicus) clade in both their analyses and those reported here (ML bootstrap 94%), only one other locus (BRM-I15) weakly supports this relationship, and mtDNA and several nuclear loci (ACO1-I9, MUSK-

I4, PER-I9) strongly contradict it. Fuchs et al. used simulations to estimate the probability of this conflict arising through coalescent variation at 0.026. However, by selecting the most strongly supported conflict in their analyses, then running 29 the test on that conflict, they are effectively conducting multiple tests, and therefore underestimating the probability of finding such a strongly supported conflict. In addition, Fuchs et al.’s own analyses show that the purported hybridization must have occurred in a very short time window, and that the introgressed alleles would have to have reached fixation throughout the ancestor of Campephilus in a similarly short time period, prior to the basal split within the genus. In short, the ambiguous placement of Campephilus may be more easily explained by coalescent variation than by an inferred hybridization event.

Excepting Campephilus, memberships of the tribes Megapicini, Malarpicini, and

Dendropicini are consistently recovered and strongly supported by my analyses; this corroborates previous findings (Benz et al., 2006; Fuchs et al., 2007; Fuchs et al., 2013; Webb and Moore, 2005; Winkler et al., 2014). By contrast, relationships among the tribes remain uncertain. ML and Bayesian analyses recover (Megapicini, (Malarpicini, Dendropicini)), with varying support (ML: 53%,

MrBayes: 0.95), while STAR recovers (Malarpicini, (Megapicini, Dendropicini), but with only 30% bootstrap support. My own divergence time estimates and previous studies (Fuchs et al., 2007) suggest that the splits among these three tribes (and Campephilus, whatever its placement) happened in a relatively short period of time; this is another case where phylogenomic data may be needed to resolve relationships, assuming they are resolvable.

Relationships within Megapicini are fairly clear. Blythipicus is found sister to

Chrysocolaptes (including the former genus Reinwardtipicus), with strong support except in the STAR tree (73% bootstrap support). The monophyly of these two genera is strongly supported across all analyses, and there is no reason to question it.

Relationships within Malarpicini are more complicated. Several subclades are consistently recovered and strongly supported. These include the genus Picus, the genus Chrysophlegma, a clade containing Campethera and Geocolaptes, the clade ((Dinopium, Gecinulus), (Meiglyptes, Micropternus)), and the clade

(Dryocopus, (Celeus, (Colaptes, Piculus))). Relationships among these subclades vary among analyses. While both ML and Bayesian concatenated analyses find Picus sister to Campethera+Geocolaptes with strong support,

STAR places Campethera+Geocolaptes sister to the clade containing Dinopium,

Gecinulus, Meiglyptes, and Micropternus. Fuchs et al. (2007) treat the Malarpicini in detail; the rapid radiation at the base of this tribe remains another challenging phylogenetic problem.

Most of the genera within Malarpicini were strongly supported as monophyletic across all analyses. Two exceptions deserve particular mention. The South

American species Dryocopus galeatus, listed as Vulnerable on the IUCN Red

List, was recovered within Celeus in all my analyses. This placement is not 31 completely unanticipated; the species is unusual for Dryocopus, is often described as having Celeus-like characteristics (Short, 1982; Winkler and

Christie, 2002), and an independent analysis placed it within Celeus (B. Benz, pers. comm.). This relationship is based only on a single ND2 sequence, and needs to be confirmed with additional data. To address the possibility that this result was an artifact, I re-sequenced ND2 from Dryocopus galeatus from a separate PCR reaction; the two sequences were identical. In addition, it is unlikely that this sequence is due to laboratory contamination, as no work on any species of Celeus has been conducted in our lab. Confusion of specimens is also unlikely to be the cause: two species of Celeus, C. lugubris and C. flavescens, were collected on the same expedition, but both are placed with strong support in a different subclade of Celeus than this D. galeatus sample. An extensive treatment of relationships within Celeus (Benz and Robbins, 2011), included populations of all Celeus species found in the region where this D. galeatus specimen was collected.

Moore et al. (2011) found Colaptes rubiginosus to be paraphyletic, with individuals from Veracruz, Mexico sister to C. auricularis and those from Peru sister to C. atricollis. My analyses placed Guyana specimens of C. rubiginosus sister to C. atricollis to the exclusion of Peruvian C. rubiginosus, with strong support. South American C. rubiginosus are therefore paraphyletic, and this complex may include additional independent lineages. Appropriate names for 32 these lineages depend on the subspecies they contain, which is currently unclear. At present, I refer to the Peru lineage as C. canipileus, the Guyana lineage as C. rubiginosus, and the Mexico lineage as C. yucatanensis (del Hoyo and Collar (2014) name it C. aeruginosus, but it probably includes yucatanensis, which has taxonomic priority). This complex requires a well-sampled phylogeographic study, including closely related species of Colaptes, to clarify species limits and relationships.

Geocolaptes olivaceus was consistently placed nested within the radiation of

Campethera, though its placement was only strongly supported in the Bayesian analysis. All previous studies have found Geocolaptes sister to Campethera, but sampling within Campethera was restricted to one or two representatives (Benz et al., 2006; Fuchs et al., 2007; Fuchs et al., 2013; Fuchs et al., 2008; Webb and

Moore, 2005). While definitive placement of Geocolaptes awaits more complete sampling of Campethera, if it is nested within Campethera, Geocolaptes is a strikingly divergent member of this clade. Unlike nearly all other woodpeckers, it occupies open habitats devoid of trees, nesting in burrows (Winkler and Christie,

2002). Its assumed generic status has been based in part on these unique characteristics. It would also present an interesting taxonomic quandary, as the ecologically specific name Geocolaptes Swainson, 1832 predates Campethera

Gray, 1841.

Relationships within Dendropicini were some of the first to receive in-depth treatment with DNA sequence data (Moore et al., 2006; Weibel and Moore,

2002a, b), and are some of the best resolved. Melanerpes and Sphyrapicus consistently form a monophyletic group, which is sister to a clade of pied and brown woodpeckers including Picoides, Leiopicus, Dendropicos, Dendrocopos,

Dryobates, Leuconotopicus, and Veniliornis (though three of these genera have often been subsumed in Picoides and/or Dendrocopos).

Contrary to Overton and Rhoads (2006), I find no evidence that Melanerpes is monophyletic; in my analyses, Melanerpes striatus either clusters with

Sphyrapicus, forms the sister group to the Melanerpes+Sphyrapicus clade, or forms a polytomy with Sphyrapicus and other Melanerpes. In none of these cases is the placement of M. striatus within the clade well supported.

Xiphidiopicus probably also belongs in this clade, but given the poor quality of the only sequence available for it (see section 3.1; Overton and Rhoads, 2006), its placement requires re-analysis with new data.

The other major clade of Dendropicini has a consistent topology across analyses

(see Figures 1.2-1.4), with generally strong support in ML and Bayesian analyses, though some nodes had fairly low bootstrap support in the STAR analyses. The STAR tree also has a weakly supported conflict with the ML and

Bayesian trees in the placement of Dendrocopos and Dendropicos+Leiopicus 34 relative to Dryobates+Leuconotopicus+Veniliornis. The monophyly of most genera was strongly supported, though STAR analyses again show low support values. One notable exception to monophyletic genera is that Leiopicus mahrattensis was found sister to Dendropicos in the Bayesian analysis, with 0.94 posterior probability. The ML and species tree analyses recovered it as sister to other Leiopicus, but with low bootstrap support (ML: 60%; STAR: 32%). Finally,

Dendropicos obsoletus, which has alternately been placed in the genus

Dendrocopos or the monotypic genus Ipophilus based on morphology, was placed within Dendropicos, with strong support in all three analyses.

Several groups within Picidae await in-depth treatment. The most prominent of these are the diverse Neotropical piculet genus Picumnus, and the African genera Campethera and Dendropicos. Other groups needing more complete treatment are the Asian Dendrocopos, which recent evidence suggests may include species actually related to Dryobates or other genera (Winkler et al.,

2014), and the clade including Melanerpes, Sphyrapicus, and Xiphidiopicus, as

Melanerpes may be paraphyletic, and the only sequences placing Xiphidiopicus appear problematic. Another enigmatic taxon requiring molecular work is

Dendrocopos dorae, which Winkler et al. (2014) place in Dendropicos. This species is probably a member of the pied/brown clade of Dendropicini, but it has no obvious affinities with any of those genera.

In conclusion, while these and previous analyses resolve many of the major phylogenetic relationships within Picidae, numerous outstanding questions remain, including fundamental questions about relationships at the base of the family. Further work sampling additional species and loci will be required if these problems are to be resolved.

4.3. Comparison with supertree and supermatrix analyses

Three recent studies inferred relationships among all birds, using supertree

(Davis and Page, 2014), supermatrix (Burleigh et al., 2015), or hybrid supertree- supermatrix (Jetz et al., 2012) approaches. While these studies included taxon sampling far beyond that treated here, and did not focus specifically on Picidae, each included extensive sampling of Picidae.

Davis and Page (2014) included 112 species of Picidae, with relationships estimated by supertree construction using matrix representation with parsimony

(MRP; Baum, 1992; Ragan, 1992). Notably, this approach does not yield meaningful support values or branch lengths. Davis and Page recovered the general structure of the Picidae tree presented herein, with a few discrepancies among the major groupings. They found Jynx sister to all other Picidae, Sasia sister to Picumnus+Nesoctites+Picinae, and Picumnus sister to

Nesoctites+Picinae. Their tree placed Hemicircus and Nesoctites in a clade sister to the remaining Picinae. Within Picinae, they found (Megapicini, Malarpicini), 36

(Campephilus, Dendropicini)), with composition of each of these groups similar to that recovered here. However, their tree also included a number of anomalous findings, apparently due to the varied overlap in taxon sampling. For example, they placed one member of Meiglyptes within Chrysophlegma, and Celeus was split, with some members forming a clade sister to Dryocopus, and others forming a clade sister to Colaptes+Piculus.

Jetz et al. (2012) conducted supermatrix analyses of bird subclades, including

Picidae before grafting these subclades onto a backbone tree. They also placed species for which no sequence data were available, but I ignore these taxonomy- based analyses. Burleigh et al. (2015) inferred relationships of all birds using supermatrix methods. The two studies use similar data: Jetz et al. included 7 loci for 137 picid taxa, while Burleigh et al. included 9 loci for 142 taxa. Unsurprisingly for such large-scale analysis, their data matrices included numerous problematic sequence data and taxonomic mis-assignments. For example, both studies included the suspect CYTB sequences from Melanerpes striatus and

Xiphidiopicus percussus excluded here, as well as a Picus awokera CYTB sequence deemed a pseudogene by Fuchs et al. (2008). Jetz et al. included at least 28 sequences for 10 different species, and Burleigh et al. 14 sequences for

5 species, which have been or could possibly be assigned to other species in their taxonomy; these errors are generally due to historical taxon names in

GenBank not corresponding to current species limits. Several of the taxa in each 37 study include no data that can be unambiguously assigned to that taxon. Their approaches also yielded unnecessarily sparse data matrices; at least 22 of the

~500 sequences in each study had higher-coverage alternatives available to them at the time.

Despite these shortcomings from the perspective of picid sampling, both Jetz et al. and Burleigh et al. recovered trees quite similar to those presented here. They found Jynx sister to remaining Picidae, Sasia+Picumnus sister to

Nesoctites+Picinae, and Nesoctites+Hemicircus sister to remaining Picinae. They recovered Malarpicini and Dendropicini with the same members as presented here, but Campephilus was found nested within Megapicini by Jetz et al.

Relationships among genera within tribes were also similar to those presented here, though both found a monophyletic Picus+Chrysophlegma, which is strongly contradicted by the Bayesian and ML analyses discussed above. Other points of conflict with my results include the placement of Celeus sister to Dryocopus in

Jetz et al. (vs. sister to Colaptes+Piculus), and the odd placement of

Xiphidiopicus percussus sister to Picoides stricklandi in Burleigh et al. With the exception of the problematic Melanerpes+Sphyrapicus+Xiphidiopicus group, relationships within Dendropicini were nearly identical to those found herein; the topology of this portion of the tree has been consistent across numerous studies.

One notable exception is the placement of Dendrocopos maculatus sister to D. moluccensis in Jetz et al., with a branch length close to 0. This is due to absence 38 of overlapping characters in the data matrix; in my analyses, these two species are not sister taxa.

5. Conclusions

I present the most complete phylogeny to date of the avian family Picidae.

Findings are generally consistent across analytical techniques, and in agreement with previous studies. More thorough taxon sampling allows placement of a number of species for the first time, and highlights regions of the tree where uncertainty remains. Future analyses may require much more extensive locus sampling in order to resolve the short internodes deep within the tree.

Table 1.1 Specimens used for DNA sequencing, with collection locality and voucher information.

Species Locality Collection Voucher # Jynx ruficollis Dem. Rep. Congo UMMZ 235397 Campephilus magellanicus Prov. Río Negro, Argentina AMNH DOT-13532 Campephilus magellanicus Prov. Río Negro, Argentina AMNH DOT-13541 Campephilus robustus Itapúa, Paraguay UMMZ 200730 Campethera nubica Kenya AMNH DOT-13018 Chrysocolaptes lucidus Leyte, Philippines KU 110763 Chrysophlegma miniaceum Singapore AMNH DOT-17251 Colaptes auratus Minnesota, USA JFBM 48408 Colaptes cafer Oregon, USA JFBM 47195 Colaptes cafer Oregon, USA JFBM 47194 Colaptes chrysoides Baja California Sur, Mexico MZFC BCHVM-176 Dendropicos obsoletus Northern Province, Sierra Leone KU 113226 Dinopium benghalense unknown AMNH DOT-11104 Dinopium javanense Singapore AMNH DOT-17259 Dryobates pubescens Minnesota, USA JFBM 47204 Dryocopus fuliginosus Mindanao, Philippines KU 112437 Dryocopus galeatus Canindeyú, Paraguay UMMZ 202091 Dryocopus javensis Samar, Philippines KU 110756 Dryocopus pileatus Oregon, USA JFBM 47199 Dryocopus pulverulentus Palawan, Philippines KU 99094 Sphyrapicus ruber Oregon, USA JFBM 47193 Sphyrapicus ruber Oregon, USA JFBM 47221 Sphyrapicus thyroideus Oregon, USA JFBM 47232 Sphyrapicus thyroideus Oregon, USA JFBM 47229 Sphyrapicus varius Minnesota, USA JFBM 47202 Abbreviations: AMNH = American Museum of Natural History; KU = University of Kansas Natural History Museum; JFBM = Bell Museum at the University of Minnesota; MZFC = Museo de Zoología "Alfonso L. Herrera," Universidad Nacional Autónoma de México; UMMZ = University of Michigan Museum of Zoology.

Table 1.2 Genetic loci and primers used for PCR amplification and sequencing. All primer pairs were tailed with standard M13 forward and reverse primers; M13 primers were used for sequencing.

Locus Chromosome Primer Sequence (5'-3') Source ND2 mitochondrial L5215U TATCGGGCCCATACCCCGAAWAT 2 H1056U RTYTAAGGCTTTGAAGGCCTTYGG 2 CYTB mitochondrial L14863-Pic ACATTCGCCCTRTCCGTYCTYAT 2 H16040-Pic AGACCAATGTTTTMGATAAACTATTAGAG 2 H16060 CTTCAATYTTTGGYTTACAAGACCAATG 1 ACO1-I9 z-linked WP_ACO1_F CTCCTTCTGGCTCCTTGCTTTACA 2 WP_ACO1_R CTTTTTGGTCCACTACAGCAAGGC 2 FGB-I7 autosomal WP_FIB7_F GACAATGATGGATGGTATGTGCTTGC 2 WP_FIB7_R CTTGGATCTGAAGTTAACCTGATGAAGG 2 MYO-I2 autosomal Myo2Pi-F CCTGTCAAATATCTGGAGGTATG 3 Myo3F GCAAGGACCTTGATAATGACTT 4,5 TGFB-I5 autosomal TGFBwp.F CTAGATGCTGCCTATTGTTTTAGG 2 TGFBwp.R TATCCTGCACGTTCCTGGGAA 2 Sources: (1) This study; (2) Vazquez-Miranda, 2014; (3) Fuchs et al., 2006; (4) Heslewood et al., 1998; (5) Fuchs et al., 2013.

Table 1.3 Calibration points for divergence time estimation in r8s and BEAST. Precise dates are unavailable for most fossils listed; ages are based on boundaries of stratigraphic ranges from source materials.

Node (MRCA of taxa listed) r8s calibration BEAST calibration Sources Pici (Picidae plus not used normal, µ=51.6 Ma, σ=12.0 Ma 1,2,3,4 Ramphastidae and barbets) Picidae and Indicatoridae min. 20 Ma, normal, μ=39.5 Ma, σ=10.0 Ma 1,2,3,4 max. 40 Ma Picidae min. 22 Ma lognormal, offset=22.0 Ma, 5 μ=30.0 Ma, log(σ)=4.165 Picumninae and Nesoctitinae min. 23.0 Ma lognormal, offset=23.0 Ma, 6 μ=28.0 Ma, log(σ)=4.216 Sasia abnormis / Sasia min. 4.5 Ma, normal, μ=5.0 Ma, σ=0.25 Ma 7 ochracea max. 5.5 Ma Campephilus min. 3.6 Ma lognormal, offset=3.6 Ma, μ=7.2 8,9 Ma, log(σ)=3.4 Colaptes, Piculus, min. 11.6 Ma lognormal, offset=11.6 Ma, 9,10,11 Dryocopus, and Celeus μ=13.8 Ma, log(σ)=4.0 Colaptes, Piculus, and min. 3.6 Ma not used 9,12 Celeus Sources: (1) Ericson et al., 2006; (2) Brown et al., 2007; (3) Brown et al., 2008; (4) Moore and Miglia, 2009; (5) De Pietri et al., 2011; (6) Laybourne et al., 1994; (7) Fuchs et al., 2006; (8) Brodkorb, 1971; (9) Manegold and Louchart, 2012; (10) Wetmore, 1931; (11) Short, 1965; (12) Becker, 1986.

Table 1.4 Genetic loci included in phylogenetic analyses. Abbreviations listed are used throughout the text. Sequence lengths are the number of base pairs of the longest sequence in the analysis, and the single-locus alignment after trimming.

# Locus Abbreviation Chromosome Taxa Length Length Substitution (max.) (aligned) Model 1 12S ribosomal RNA 12S mitochondrial 78 973 876 GTR+I+Γ 2 16S ribosomal RNA 16S mitochondrial 7 1603 436 GTR+I 3 ATP synthase subunit 6 ATP6 mitochondrial 71 684 684 GTR+I+Γ 4 ATP synthase subunit 8 ATP8 mitochondrial 18 168 168 GTR+I 5 cytochrome c oxidase subunit 1 COI mitochondrial 129 1551 1551 GTR+I+Γ 6 cytochrome c oxidase subunit 3 COIII mitochondrial 28 784 252 HKY+Γ 7 control region (D-loop) CR mitochondrial 25 1215 934 HKY+I+Γ 8 cytochrome b CYTB mitochondrial 154 1143 1143 GTR+I+Γ 9 NADH dehydrogenase 2 ND2 mitochondrial 129 1041 1041 GTR+I+Γ 10 NADH dehydrogenase 3 ND3 mitochondrial 68 352 353 GTR+I+Γ 11 oocyte maturation factor CMOS autosomal 40 605 605 HKY+I+Γ 12 fibrinogen beta chain, intron 5 FGB-I5 autosomal 41 612 588 HKY+Γ 13 fibrinogen beta chain, intron 7 FGB-I7 autosomal 121 902 961 GTR+Γ 14 glyceraldehyde-3-phosphate dehydrogenase, GAPDH-I11 autosomal 62 425 427 HKY+Γ intron 11 15 non-histone chromosomal protein HMG-17, exon 2 HMGN2 autosomal 15 691 686 HKY 16 inter-photoreceptor retinoid-binding protein IRBP autosomal 43 801 774 K80+Γ 17 lactate dehydrogenase, exons 1-2 LDH autosomal 10 1395 1000 HKY+Γ 18 myoglobin, intron 2 MYO-I2 autosomal 83 715 694 K80+Γ 19 phosphoenolpyruvate carboxykinase, intron 9 PEPCK-I9 autosomal 40 697 696 HKY+Γ 20 period protein, intron 9 PER-I9 autosomal 41 783 650 HKY+Γ 21 recombination activating gene 1 RAG1 autosomal 12 1914 1469 K80+I 22 transforming growth factor beta 2, intron 5 TGFB-I5 autosomal 73 588 565 K80+Γ 23 aconitase 1, intron 9 ACO1-I9 z-linked 51 1085 1024 HKY+Γ 24 brahma protein, intron 15 BRM-I15 z-linked 58 373 364 HKY+Γ 25 muscle skeletal receptor tyrosine kinase, intron 4 MUSK-I4 z-linked 41 580 581 HKY+Γ 43 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Figure 1.1. DNA sequence coverage by taxon. Included data are represented by black pixels; missing data is represented by white pixels. Each horizontal line corresponds to a single species, listed in the order in which they appear in the RAxML tree in Figure 1.2. The time-calibrated RAxML tree is included to aid in identifying blocks of taxa. Each pixel represents ~20bp of sequence. Tick marks on the x-axis indicate boundaries of individual loci; numbers correspond to locus numbers in Table 1.4.

44 Jynx torquilla Jynginae 100/1.00 Jynx ruficollis Verreauxia africana

Sasia abnormis Picumninae 100/1.00 100/1.00 Sasia ochracea 75/1.00 Picumnus innominatus Picumnus nebulosus 100/1.00 56/0.44 Picumnus rufiventris Picumnus exilis 100/1.00 100/1.00 Picumnus aurifrons 94/0.93 77/0.21 Picumnus lafresnayi 100/1.00 Picumnus cirratus Picumnus pygmaeus 99/1.00 62/0.84 Picumnus temminckii 34/0.63 Picumnus spilogaster Nesoctites micromegas Nesoctitinae, Hemicircus

100/1.00 Hemicircus canente Megapicini Blythipicus rubiginosus 98/1.00 100/1.00 Blythipicus pyrrhotis Chrysocolaptes validus Chrysocolaptes guttacristatus 99/1.00 Chrysocolaptes strictus 72/0.99 99/1.00 Chrysocolaptes xanthocephalus 43/0.84 Chrysocolaptes haematribon 72/0.21 87/0.76 Chrysocolaptes lucidus Campephilus haematogaster Campephilus principalis Campephilus 99/1.00 Campephilus imperialis 100/1.00 100/0.86 52/0.21 Campephilus bairdii 85/0.93 Campephilus magellanicus Campephilus rubricollis 59/0.70 63/0.58 Campephilus leucopogon Campephilus robustus 64/0.67 Campephilus pollens 6/0.26 Campephilus melanoleucos 93/0.78 Campephilus gayaquilensis 96/0.74 42/0.89 Campephilus guatemalensis 21/0.17 Micropternus brachyurus Meiglyptes tukki 100/1.00 100/1.00 Meiglyptes tristis 100/1.00 Gecinulus grantia Dinopium benghalense 98/1.00 Dinopium shorii 100/1.00 41/0.77 100/1.00 Dinopium javanense Chrysophlegma miniaceum Chrysophlegma mentale 100/1.00 100/1.00 Chrysophlegma flavinucha Campethera caroli 47/1.00 100/1.00 Campethera nivosa 100/1.00 Geocolaptes olivaceus Campethera cailliautii 66/1.00 97/1.00 83/1.00 Campethera nubica Picus puniceus 100/1.00 Picus chlorolophus Picus awokera 100/1.00 100/1.00 Picus erythropygius Picus canus 100/1.00 Picus vaillantii 93/0.97 62/0.84 Picus viridis 97/1.00 Picus sharpei 91/1.00 100/1.00 99/1.00 Picus squamatus 100/1.00 Picus xanthopygaeus 100/1.00 Picus rabieri Picus viridanus 53/0.75 100/1.00 Picus vittatus Dryocopus lineatus Dryocopus pileatus 100/1.00 Malarpicini Dryocopus martius Dryocopus hodgei 100/1.00 100/1.00 98/0.89 Dryocopus javensis 98/1.00 Dryocopus fuliginosus Dryocopus funebris 100/1.00 Dryocopus fulvus 93/0.83 93/0.83 Dryocopus pulverulentus Celeus torquatus 100/1.00 Celeus loricatus 99/1.00 Celeus ochraceus Celeus flavescens 100/1.00 100/1.00 Celeus elegans 76/1.00 38/0.94 Celeus lugubris 88/0.99 Dryocopus galeatus Celeus flavus Celeus spectabilis 77/0.64 99/1.00 100/1.00 Celeus obrieni 89/1.00 80/1.00 Celeus castaneus Celeus grammicus 100/1.00 100/1.00 Celeus undatus Piculus chrysochloros Piculus flavigula 100/1.00 97/0.98 Piculus callopterus Colaptes fernandinae 99/1.00 Colaptes cafer 53/0.95 99/1.00 Colaptes auratus 100/1.00 Colaptes chrysoides

69/0.62 Picinae Colaptes rupicola 68/0.99 99/1.00 Colaptes pitius 96/0.99 Colaptes melanochloros 58/1.00 Colaptes rivolii 100/1.00 Colaptes campestris 99/1.00 Colaptes punctigula 100/1.00 Colaptes auricularis 65/0.82 Colaptes yucatanensis 100/1.00 Colaptes canipileus 100/1.00 Colaptes rubiginosus 99/1.00 Colaptes atricollis Melanerpes striatus Sphyrapicus thyroideus Sphyrapicus varius 100/1.00 Sphyrapicus nuchalis 100/1.00 100/1.00 98/0.99 Sphyrapicus ruber 71/0.92 Melanerpes erythrocephalus 67/0.95 Melanerpes lewis 41/0.25 Melanerpes formicivorus Melanerpes candidus 74/1.00 33/0.22 Melanerpes cactorum 41/0.54 Melanerpes herminieri 92/1.00 Melanerpes hypopolius 50/0.46 Melanerpes pucherani Melanerpes cruentatus 82/0.99 88/0.64 Melanerpes flavifrons 55/0.47 Melanerpes pygmaeus Melanerpes chrysogenys 90/0.96 Melanerpes hoffmannii 73/0.72 Melanerpes uropygialis 96/0.97 48/0.52 Melanerpes superciliaris 71/0.99 Melanerpes santacruzi 41/0.41 Melanerpes carolinus 84/0.99 100/1.00 Melanerpes aurifrons Picoides arcticus Picoides tridactylus 96/0.97 100/1.00 Picoides dorsalis 81/0.94 Picoides kizuki 32/0.13 Picoides temminckii 75/0.82 Picoides maculatus 99/1.00 Picoides canicapillus 96/0.68 Picoides moluccensis 60/0.06 Leiopicus mahrattensis Leiopicus auriceps Dendropicini 100/1.00 Leiopicus medius 96/0.97 87/1.00 Dendropicos pyrrhogaster Dendropicos fuscescens 98/1.00 98/1.00 Dendropicos elliotii 78/0.99 Dendropicos obsoletus Dendropicos griseocephalus 87/1.00 98/1.00 Dendropicos goertae Dendrocopos hyperythrus Dendrocopos atratus 85/0.99 Dendrocopos macei 99/1.00 77/0.82 64/0.98 Dendrocopos analis Dendrocopos leucotos 98/0.91 39/0.03 Dendrocopos noguchii Dendrocopos darjellensis 94/1.00 Dendrocopos syriacus 61/1.00 17/0.14 Dendrocopos major 30/0.24 Dendrocopos himalayensis 98/0.99 Dryobates minor 87/0.98 Dryobates cathpharius 98/1.00 Dryobates pubescens Dryobates nuttallii 94/1.00 100/1.00 Dryobates scalaris 59/0.88 Leuconotopicus borealis 97/0.99 Leuconotopicus fumigatus Leuconotopicus villosus 99/1.00 100/1.00 Leuconotopicus arizonae 97/1.00 Leuconotopicus albolarvatus 94/1.00 Veniliornis chocoensis 40/0.32 Veniliornis cassini 42/0.70 Veniliornis kirkii Veniliornis spilogaster 98/1.00 Veniliornis mixtus 97/1.00 100/1.00 Veniliornis lignarius 79/0.99 97/1.00 Veniliornis sanguineus Veniliornis frontalis 100/1.00 Veniliornis passerinus Veniliornis dignus 98/1.00 65/0.98 Veniliornis callonotus 96/1.00 Veniliornis affinis 0.02 69/0.92 Veniliornis nigriceps

Figure 1.2. Phylogenetic relationships of Picidae inferred from maximum likelihood analysis of the concatenated alignment of all loci in RAxML. Numbers at nodes indicate percentage of bootstrap replicates and posterior probabilities for the descendant clade. Bars and labels at right indicate named subfamilies, tribes, and genera that do not fit into existing groupings. 45 Jynx torquilla 100 Jynx ruficollis Sasia ochracea Verreauxia africana 100 67 Sasia abnormis 61 Picumnus innominatus Picumnus nebulosus 98 Picumnus rufiventris Picumnus exilis 78 77 Picumnus lafresnayi 84 81 Picumnus aurifrons 100 Picumnus cirratus Picumnus temminckii 95 Picumnus pygmaeus 83 44 36 Picumnus spilogaster Nesoctites micromegas 45 Hemicircus canente Campethera nivosa Campethera caroli 93 Campethera cailliautii Geocolaptes olivaceus 33 46 Campethera nubica 99 Gecinulus grantia Dinopium benghalense Dinopium shorii 89 85 100 86 Dinopium javanense 74 Micropternus brachyurus Meiglyptes tukki 28 89 99 Meiglyptes tristis Chrysophlegma mentale Chrysophlegma miniaceum 100 57 Chrysophlegma flavinucha Picus puniceus 22 99 Picus chlorolophus Picus squamatus 66 Picus xanthopygaeus 95 61 Picus rabieri Picus vittatus 39 80 Picus viridanus 70 Picus awokera Picus erythropygius 39 Picus canus 38 64 Picus vaillantii 49 Picus viridis 99 94 Picus sharpei 77 Dryocopus pileatus 99 Dryocopus lineatus Dryocopus martius Dryocopus hodgei 92 83 73 Dryocopus javensis 86 Dryocopus fuliginosus Dryocopus funebris 84 Dryocopus fulvus 64 70 Dryocopus pulverulentus Celeus torquatus 99 Celeus loricatus 97 Celeus ochraceus Celeus flavescens 96 73 Celeus lugubris 52 33 Celeus elegans 58 Dryocopus galeatus Celeus castaneus Celeus undatus 63 44 99 Celeus grammicus 94 81 Celeus flavus Celeus spectabilis 58 90 Celeus obrieni Piculus chrysochloros Piculus callopterus 94 88 Piculus flavigula Colaptes fernandinae 97 Colaptes cafer 72 Colaptes auratus 95 Colaptes rivolii 100 Colaptes melanochloros 66 16 Colaptes pitius 48 89 Colaptes rupicola 94 Colaptes punctigula 92 Colaptes campestris Colaptes yucatanensis 39 98 Colaptes auricularis 85 Colaptes canipileus Colaptes rubiginosus 59 80 Colaptes atricollis Blythipicus rubiginosus 99 Blythipicus pyrrhotis Chrysocolaptes validus 73 Chrysocolaptes guttacristatus 99 Chrysocolaptes strictus 99 Chrysocolaptes xanthocephalus 56 Chrysocolaptes lucidus 65 81 84 Chrysocolaptes haematribon Campephilus haematogaster Campephilus principalis Campephilus imperialis 100 99 64 Campephilus bairdii 90 Campephilus magellanicus Campephilus rubricollis 53 73 Campephilus leucopogon Campephilus pollens 66 Campephilus guatemalensis 92 Campephilus melanoleucos 46 62 Campephilus gayaquilensis Melanerpes striatus Sphyrapicus thyroideus 37 Sphyrapicus varius 100 Sphyrapicus ruber 30 97 88 Sphyrapicus nuchalis Melanerpes lewis 99 58 Melanerpes erythrocephalus Melanerpes formicivorus 59 Melanerpes candidus 70 Melanerpes herminieri 38 51 Melanerpes cactorum 85 Melanerpes hypopolius Melanerpes pucherani Melanerpes flavifrons 87 46 83 Melanerpes cruentatus 61 Melanerpes pygmaeus Melanerpes chrysogenys 91 Melanerpes hoffmannii 55 Melanerpes uropygialis 64 Melanerpes superciliaris 97 69 24 Melanerpes santacruzi 35 Melanerpes carolinus 99 Melanerpes aurifrons Picoides arcticus Picoides tridactylus 96 99 Picoides dorsalis 84 Picoides temminckii Picoides kizuki 77 Picoides maculatus 50 Picoides moluccensis 89 92 Picoides canicapillus Dendrocopos hyperythrus Dendrocopos atratus Dendrocopos macei 97 72 95 70 Dendrocopos analis 82 Dendrocopos leucotos Dendrocopos noguchii 61 Dendrocopos major 37 Dendrocopos syriacus 22 Dendrocopos darjellensis 42 39 Dendrocopos himalayensis Leiopicus mahrattensis 84 Leiopicus medius 32 50 Leiopicus auriceps 58 Dendropicos pyrrhogaster Dendropicos fuscescens 63 76 Dendropicos elliotii 75 Dendropicos obsoletus Dendropicos griseocephalus 89 97 Dendropicos goertae 66 Dryobates minor 54 Dryobates cathpharius 66 Dryobates pubescens Dryobates nuttallii 81 31 Dryobates scalaris Leuconotopicus fumigatus 95 Leuconotopicus borealis 66 Leuconotopicus villosus 47 Leuconotopicus arizonae 83 84 Leuconotopicus albolarvatus 66 Veniliornis chocoensis Veniliornis cassini Veniliornis kirkii 96 Veniliornis spilogaster 26 Veniliornis mixtus 77 75 94 Veniliornis lignarius Veniliornis sanguineus 37 Veniliornis passerinus 95 100 Veniliornis frontalis 95 Veniliornis affinis 45 Veniliornis nigriceps 89 Veniliornis dignus 56 Veniliornis callonotus Figure 1.3. Phylogenetic relationships of Picidae inferred from STAR analysis of RAxML nuclear gene trees and concatenated mtDNA analysis. Numbers at nodes are support values from multilocus bootstrapping in STAR. 46 Ramphastidae spp Indicatoridae spp Jynx torquilla Jynx ruficollis Verreauxia africana Sasia abnormis Sasia ochracea Picumnus innominatus Picumnus nebulosus Picumnus rufiventris Picumnus exilis Picumnus aurifrons Picumnus lafresnayi Picumnus cirratus Picumnus pygmaeus Picumnus temminckii Picumnus spilogaster Nesoctites micromegas Hemicircus canente Blythipicus rubiginosus Blythipicus pyrrhotis Chrysocolaptes validus Chrysocolaptes guttacristatus Chrysocolaptes strictus Chrysocolaptes xanthocephalus Chrysocolaptes haematribon Chrysocolaptes lucidus Campephilus haematogaster Campephilus principalis Campephilus imperialis Campephilus bairdii Campephilus magellanicus Campephilus rubricollis Campephilus leucopogon Campephilus robustus Campephilus pollens Campephilus melanoleucos Campephilus gayaquilensis Campephilus guatemalensis Micropternus brachyurus Meiglyptes tukki Meiglyptes tristis Gecinulus grantia Dinopium benghalense Dinopium shorii Dinopium javanense Chrysophlegma miniaceum Chrysophlegma mentale Chrysophlegma flavinucha Campethera caroli Campethera nivosa Geocolaptes olivaceus Campethera cailliautii Campethera nubica Picus puniceus Picus chlorolophus Picus awokera Picus erythropygius Picus canus Picus vaillantii Picus viridis Picus sharpei Picus squamatus Picus xanthopygaeus Picus rabieri Picus viridanus Picus vittatus Dryocopus lineatus Dryocopus pileatus Dryocopus martius Dryocopus hodgei Dryocopus javensis Dryocopus fuliginosus Dryocopus funebris Dryocopus fulvus Dryocopus pulverulentus Celeus torquatus Celeus loricatus Celeus ochraceus Celeus flavescens Celeus elegans Celeus lugubris Dryocopus galeatus Celeus flavus Celeus spectabilis Celeus obrieni Celeus castaneus Celeus grammicus Celeus undatus Piculus chrysochloros Piculus flavigula Piculus callopterus Colaptes fernandinae Colaptes cafer Colaptes auratus Colaptes chrysoides Colaptes rupicola Colaptes pitius Colaptes melanochloros Colaptes rivolii Colaptes campestris Colaptes punctigula Colaptes auricularis Colaptes yucatanensis Colaptes canipileus Colaptes rubiginosus Colaptes atricollis Melanerpes striatus Sphyrapicus thyroideus Sphyrapicus varius Sphyrapicus nuchalis Sphyrapicus ruber Melanerpes erythrocephalus Melanerpes lewis Melanerpes formicivorus Melanerpes candidus Melanerpes cactorum Melanerpes herminieri Melanerpes hypopolius Melanerpes pucherani Melanerpes cruentatus Melanerpes flavifrons Melanerpes pygmaeus Melanerpes chrysogenys Melanerpes hoffmannii Melanerpes uropygialis Melanerpes superciliaris Melanerpes santacruzi Melanerpes carolinus Melanerpes aurifrons Picoides arcticus Picoides tridactylus Picoides dorsalis Picoides kizuki Picoides temminckii Picoides maculatus Picoides canicapillus Picoides moluccensis Leiopicus mahrattensis Leiopicus auriceps Leiopicus medius Dendropicos pyrrhogaster Dendropicos fuscescens Dendropicos elliotii Dendropicos obsoletus Dendropicos griseocephalus Dendropicos goertae Dendrocopos hyperythrus Dendrocopos atratus Dendrocopos macei Dendrocopos analis Dendrocopos leucotos Dendrocopos noguchii Dendrocopos darjellensis Dendrocopos syriacus Dendrocopos major Dendrocopos himalayensis Dryobates minor Dryobates cathpharius Dryobates pubescens Dryobates nuttallii Dryobates scalaris Leuconotopicus borealis Leuconotopicus fumigatus Leuconotopicus villosus Leuconotopicus arizonae Leuconotopicus albolarvatus Veniliornis chocoensis Veniliornis cassini Veniliornis kirkii Veniliornis spilogaster Veniliornis mixtus Veniliornis lignarius Veniliornis sanguineus Veniliornis frontalis Veniliornis passerinus Veniliornis dignus Veniliornis callonotus Veniliornis affinis Veniliornis nigriceps

60 50 40 30 20 10 0 Figure 1.4. Divergence time estimation from BEAST, based on branch length estimation with fossil and biogeographic calibrations. Gray bars indicate 95% highest posterior density intervals for node ages. Nodes with calibrations are marked with arrows. Scale bar reflects time before present, in millions of years (Ma). 47

Chapter 2

Both species sorting and trait evolution explain non-random trait

distributions in North American woodpecker communities

Abstract

Patterns of trait variation among coexisting species are often used to infer the processes driving community structure. In many cases, the same patterns could be driven either by ecological assortment processes or by evolutionary character displacement within communities. I used multiple metrics and null models to test for non-random trait distributions within North American woodpecker communities. I analyzed variation within species across communities to determine whether observed community-level patterns were best explained by ecological sorting of species or in situ trait evolution. Data from ten species at six sites showed an overall signal of displacement, suggesting local evolution. By contrast, most individual communities did not differ from random expectations, though two had members either evenly spread or clustered in morphological trait space. Some of these patterns are best explained by species sorting, while others appear to be driven by local evolution. I conclude that character displacement may be operating at the population level, but in most cases its effects are not strong enough to generate detectable community-level patterns.

This may be due to non-equilibrium processes or lack of power to detect patterns in communities with small numbers of species.

1. Introduction

Patterns in ecological communities may provide a window into processes operating over timescales that cannot be directly observed. Extensive research, including numerous controlled experiments, has demonstrated the effect that ecological assortment processes such as habitat filtering and competitive exclusion can have on community patterns (Vellend, 2010; Weiher et al., 2011).

Such assortment processes, also known as species sorting, determine the composition of communities as a limited subset of members of a broader species pool. More recently has come the recognition that evolutionary processes may also play a role in generating such patterns (Brooks and McLennan, 1991;

Cavender-Bares et al., 2009; Johnson and Stinchcombe, 2007; Losos, 1995). For example, populations may evolve in situ in response to interactions with other community members or abiotic conditions. Alternatively, broader scale diversification, trait evolution, and biogeography may shape the species pools from which community members are drawn. Both of these processes can operate on similar timescales (Hubbell, 2006; Urban et al., 2008), and can collectively contribute to community structure (Cavender-Bares et al., 2009; Johnson and

Stinchcombe, 2007). However, we lack a general understanding of the relative roles of species sorting and evolution in determining community structure. 49

The search for a general model of ecological community assembly dates back at least to the 1950s (Diamond, 1975; Hutchinson, 1959; Weiher and Keddy, 1999).

Despite early difficulties (e.g. Connor and Simberloff, 1979; Simberloff and

Boecklen, 1981), extensive work has yielded a nuanced understanding of the interplay of processes such as habitat filtering, competitive exclusion, and dispersal, the underlying mechanisms driving those processes, and how their effects vary across habitats, taxa, and spatial and temporal scales

(HilleRisLambers et al., 2012; Vellend, 2010; Weiher et al., 2011). However, much of this understanding rests on the assumption that species are static entities, and that the presence or absence of patterns in communities is due to the particular set of those entities able to persist in a given community.

But species are not static entities. They change over space and time through drift and response to various selective processes (Darwin, 1859; Wright, 1931). The potential importance of such evolution in communities has long been recognized

(MacArthur and Levins, 1967), but has been largely ignored in work on community assembly. More recently, the recognition that community assembly and evolutionary processes can operate on similar timescales has refocused attention on the role of evolution in generating community-level patterns

(Cavender-Bares et al., 2009; Johnson and Stinchcombe, 2007; Ricklefs, 2008).

It is well established that interactions among species can influence evolution. The 50 most obvious cases involve tightly linked interactions, such as host-parasite

(Clayton and Moore, 1997), predator-prey (Abrams, 2000), and mutualism

(Boucher, 1988). Competition among ecologically similar species can also drive evolution (Grant and Grant, 2006; Schluter, 1994).

There is a growing body of research on patterns of diversification and trait evolution among species, and how that diversity may be driven by or contribute to community-level processes (e.g. Graham et al., 2012; Kozak et al., 2009;

Losos, 1990). Given the spatial and temporal scales at which such processes operate, we may also expect to see effects among populations within species.

Ecological character displacement is one area in which such intraspecific variation has been explored for possible interplay with ecology (Brown and

Wilson, 1956; Grant, 1972). While initially formulated for interactions among species pairs, character displacement has been extended to community-wide processes and patterns (Brown, 1975; Dayan and Simberloff, 2005; Strong et al.,

1979). In general, patterns used to test for community-wide character displacement cannot differentiate ecological sorting processes from local evolution, as they use only characteristics within communities, which do not provide information on evolutionary changes. To discriminate between these two possible causes, we need additional data to quantify evolution, such as variation among populations of the component species. We can then test for community-

51 level patterns, and for population-level variation that may contribute to those patterns.

In addition, most empirical work on competitive exclusion and character displacement has assumed that competitive interactions are most likely to result in divergence or exclusion of similar species. However, a number of theoretical studies have identified conditions under which the expected result is convergence or coexistence of similar species (Abrams, 1986; Mayfield and

Levine, 2010; Slatkin, 1980; Taper and Case, 1985), yielding communities of species that are clustered or clumped in trait space (Scheffer and van Nes,

2006). Testing only for patterns expected under divergence may thus miss any signature of convergence and underestimate the prevalence of character displacement or competitive exclusion. Therefore, in order to clarify the roles of ecological and evolutionary processes in generating community patterns, we need to test for patterns expected from convergence and divergence (or clumped and evenly distributed species), and to examine intraspecific variation to determine possible evolutionary explanations for such patterns.

Finally, many of the metrics used to quantify variation in community trait patterns have low power in communities with low species richness (Mouchet et al., 2010).

The failure to detect community-level patterns may be due to this lack of power, rather than to an absence of character displacement. Testing for expected trait 52 shifts across many populations could allow detection of such a signal, even if its effects are not strong enough to be detectable at the community level.

North American woodpeckers (Aves: Picidae) are particularly well-suited for such analyses. Most woodpecker species share a common resource base: arthropods on or within a substrate, often woody plant tissue (though some species seasonally utilize other resources; Winkler and Christie, 2002). Previous research has identified morphological variation that corresponds to differences in diet and foraging mode, and is therefore a suitable proxy for ecological similarity (Burt,

1930; Leonard and Heath, 2010; Short, 1978; Spring, 1965). Specifically, species vary primarily in size and along an excavator-to-gleaner axis, with robust, short- limbed excavators more capable of delivering strong blows, and gracile, long- limbed gleaners able to move more rapidly over substrates (Burt, 1930; Spring,

1965). These differences relate to trade-offs in efficiency of acquiring different resources, and therefore represent niche differences (sensu Mayfield and Levine,

2010) or stabilizing mechanisms (sensu Chesson, 2000) that facilitate coexistence by decreasing interspecific competition.

In order for community patterns to be driven by either ecological or evolutionary processes, species must interact over timescales relevant to both ecology and evolution. A number of factors suggest that interactions among woodpecker species fit this requirement. Most species are generally widespread, habitat 53 generalists, and non-migratory (Winkler and Christie, 2002), increasing both the ecological and temporal extent of interactions. Experimental evidence has demonstrated local competitive exclusion (Peters and Grubb, 1983; Williams and

Batzli, 1979), and coexisting woodpeckers are typically strongly territorial within and among species (Cody, 1969; Short, 1971b; Winkler and Christie, 2002), or partition resources among species (Short, 1971b, 1978; Török, 1990; Ugalde-

Lezama et al., 2011). If competitive interactions have effects on communities through ecological or evolutionary processes, we might reasonably expect to detect those effects in woodpeckers.

In this study, I examined the relative influence of ecological sorting processes and trait evolution on community trait distributions. I tested for non-random distributions of traits in six North American woodpecker communities comprising

10 species, and used intraspecific variation across communities to assess possible ecological and evolutionary explanations for those patterns. I found only a few cases of communities deviating from random expectations, and attribute those cases to a mix of both ecological and evolutionary processes. I explore possible reasons for the variation among communities, and suggest future research directions to improve our understanding of the interplay of community ecology and trait evolution.

2. Materials and Methods

2.1. Study sites, taxon sampling, and individual sampling

Sampling was designed to capture broad-scale ecological sorting and possible evolutionary responses to coexistence. These processes are expected to depend upon long-term interactions among species. A growing body of evidence suggests that local community assemblages may not persist over long periods of time (Leibold et al., 2004; Ricklefs, 2008; Stralberg et al., 2009). Regional coexistence may be a better predictor of long-term interactions among species than co-occurrence in a patch at a single point in time, especially for highly mobile organisms such as birds (Holt, 1993; Ricklefs, 2008; Wiens, 1989). In addition, my research assumes that there is some degree of constancy of community membership at the timescales that are relevant to evolution, which is more reasonable for regional communities (but see Davis, 1981). Consequently, for each site, all specimens included were collected within a 50-km radius; this distance incorporates substantial habitat heterogeneity and multiple generations of typical dispersal (Bull and Jackson, 1995; Kesler et al., 2010; Walters et al.,

2002). Study locations (Table 2.1) were distributed throughout the continental

United States, and were determined based on three factors: 1) sites must be sufficiently distant from each other for the reasonable assumption of independent community assortment; 2) sites must have sufficient existing museum specimens for all common woodpecker species; 3) sites should vary as much as possible in species composition. 55

Within each site, I measured up to five males and five females of each species. I excluded juveniles and specimens of migratory populations collected outside the breeding season. In many cases, sufficient samples were not available for rare species; however, these species are not expected to contribute as substantially to interspecific interactions as common species. Numbers of specimens measured for each sex of each species is listed in Table 2.1. Analyses were run with and without the available samples of rare species.

2.2. Measurements

For each specimen, I measured skeletal characters using digital calipers and photographic images (Figure 2.1). Differences in these measurements correlate with variation among woodpecker species in diet and foraging mode (Burt, 1930;

Leonard and Heath, 2010; Short, 1978; Spring, 1965). The following measurements were taken from each specimen using digital calipers; anatomical terminology follows Baumel et al. (1993). (1) greatest length of the cranium, measured from the most caudal point of the prominentia cerebellaris to the most rostral point of the maxilla, (2) length of the cranium from the occipital condyle, measured from the most caudal point of the condylus occipitalis to the most rostral point of the maxilla, (3) length of the mandible, measured from the most caudal point of the processus retroarticularis to the most rostral point of the mandible, (4) length of the humerus, measured from most proximal point of the 56 caput humeri to the most distal point of the condylus ventralis, (5) length of the ulna, measured from the most proximal point of the olecranon to the most distal point of the condylus dorsalis, (6) length of the femur, measured from most proximal point of the trochanter femoris to the most distal point of the condylus lateralis, (7) length of the tibiotarsus, measured from the most proximal point of the crista patellaris to the most distal point of the condylus medialis, (8) length of the tarsometatarsus, measured from the most proximal point of the eminentia intercotylaris to the most distal point of the trochlea accessoria, (9) length of the pygostyle, measured from the most ventral point of the discus pygostyli to the most dorsal point of the apex pygostyle, (10) width of the pygostyle, measured between the most lateral points on the left and right side of the discus pygostyli.

For the mandible and all limb measurements, both the left and right side were measured when available.

In order to ensure data consistency and account for day-to-day variation in the measurement apparatus, I measured two reference specimens on each day that measurements were taken. Data from these reference specimens showed that caliper measurements varied predictably over time due to instrument drift, but were otherwise consistent. I fit a linear model with date-specific slopes to the among-date variation in these reference measurements, and used this model to adjust all caliper-based measurements before analysis. This model explained

69% of the among-date variation in measurements of the reference specimens. 57

Images of the cranium in profile were taken with a digital SLR camera equipped with macro lens. To ensure consistent positioning across specimens, multiple cranial landmarks were aligned prior to imaging. A scale bar was placed in the plane of each image to allow calibration. Measurements were taken from images using ImageJ (Rasband, 1997-2015), with each image calibrated separately using the scale bar. I measured (11) length of the maxilla, from the nasofrontal hinge to the most distal point of the maxilla, and (12) depth of the maxilla at its base, from the dorsal surface to the nearest point on the cutting edge of the maxilla or the ventral surface of the arcus jugalis.

Many individual specimens had damage that prevented measurement of all characters. I conducted all analyses both with and without these incomplete measurement sets.

2.3. Processing of measurement data

Data processing was conducted using a larger data set of 964 individuals collected from a global sample of 151 species of woodpeckers; this larger quantity of data allowed for more robust assessment of sources of error. The data set was then trimmed following processing to include only individuals from the study sites described above. Unless otherwise noted, all data processing and

58 analysis was conducted in R (R Core Team, 2015), using scripts written by the author (available at https://github.com/mjdufort).

All measurements were first ln-transformed. Means of left and right sides were then taken for all bilateral measurements for each measurement set, and all measurements were adjusted for date variation as described above. If multiple sets of measurements were available for a single individual, I took the mean value of these measurements. Many species of woodpeckers are sexually dimorphic in size and/or shape, and averaging across samples with varying numbers of each sex may bias estimates of species-level morphology. I therefore calculated the mean across individuals for each sex of each population, and used the mean of the male and female mean values as the population mean.

Due to concerns about the repeatability of certain measurements across specimens, I tested the consistency of measurements within species by fitting

ANOVA models to the individual-level log-transformed date-adjusted data for each measurement. I treated species as a random effect, with sex nested within species. Measurements with little variance explained by species are either measured with low precision, or show large individual variation; in either case, they are not good indicators of species-level ecomorphology. As this analysis was conducted on the larger data set of 151 species, the contribution of variation among populations measured for analyses of community patterns is expected to 59 be small; results were similar using a trimmed data set excluding the species analyzed herein. For depth of the maxilla, species explained much less of the variance (87%) than for all other measurements (98-99%); I therefore discarded this measurement.

Bird skeleton preparations vary in the presence or absence of the ramphotheca, the keratin sheath that covers the maxilla and mandible on live birds. Using specimens only with or without the ramphotheca would have dramatically reduced the sample size for many populations, but measurements with and without the ramphotheca are not directly comparable. I used regression models built from the preliminary species means to impute individual measurements without the ramphotheca via stochastic regression imputation (Enders, 2010).

While this approach is not widely used in ecological research, it has been extensively validated and is commonly used to deal with missing data in other fields (Gold and Bentler, 2000; Little and Rubin, 2002). I fit phylogenetic generalized least squares (PGLS) models to the species means of all cranial measurements with and without the ramphotheca. For individuals measured with the ramphotheca, I then used predictions from these models to generate estimated measurements without the ramphotheca. I preserved variance by adding a random variable to each observation, using a random draw from a normal distribution with variance equal to residual variance in the model. The species-level PGLS models are strongly predictive, with pseudo-R2 ranging from 60

0.92 to 0.96. However, preliminary analyses showed that the imputed data inappropriately biased some populations of some species toward the regression line. To allow for species-level variation in the relationship between measurements with and without ramphotheca, I adjusted the imputed individual measurements using species-level residuals from the model fit to species means.

I then recalculated means of cranial measurements for each sex and each population; the final population means used in analyses were the means of the male and female mean value for each population of each species.

These population mean values were then used to generate three sets of size and shape variables: (1) unscaled variables or overall morphology (which includes size and shape), (2) average size, and (3) size-scaled shape variables. I calculated the average size of each population as the mean of the log of all measurements (equivalent to the log of the geometric mean of the untransformed measurements). I then generated size-scaled shape variables by taking the residuals from a PGLS regression of each variable on the average size (Revell,

2009). Rotation of data using principal components analysis (PCA) does not affect the multivariate community metrics described below, as the metrics are based on Euclidean distances, which are unaffected by PCA if all axes are retained. However, using orthogonal variables (as generated by PCA) simplifies the phylogenetic simulations used to generate null expectations. I conducted 61 phylogenetic PCA on both the overall morphological variables and the size- scaled shape variables using functions in the R package phytools (Revell, 2009,

2012).

2.4. Phylogenetic Information

For all analyses requiring phylogenetic relationships (e.g. PGLS, phylogenetic

PCA, Brownian motion models of trait evolution), I used a phylogeny of 178 picid species described in Chapter 1. The phylogeny used here was inferred using

RAxML (Stamatakis, 2014), with branch lengths scaled in BEAST using sequence data and fossil and biogeographic calibrations (Drummond et al.,

2012). Several methods described below require a population-level tree to account for evolutionary covariance within the data. I generated a population- level tree by introducing polytomies at ~200,000 years ago on the terminal branch leading to each species with multiple populations; this assumes star-like relationships among populations within species. 200,000 years is slightly more recent than the most recent interspecific split among the study species. Variation of the precise dating of population divergence as 50,000 years or 25-50% of the age since the most recent interspecific split did not affect results (not shown).

The tree was then trimmed as needed to include only the species or populations in a given analysis.

2.5. Patterns within Communities 62

2.5.1. Community metrics

A number of metrics have been used to quantify variation in community trait distributions (Kraft et al., 2008; Mouchet et al., 2010; Stubbs and Wilson, 2004). I used metrics from the literature on functional diversity and community assembly, as methodological development has proceeded somewhat independently in these two related areas. Because many of these metrics may be redundant, I conducted a PCA of their scores in simulated communities generated using permutation and simulated trait evolution (described below), and selected a subset based on variation in their coefficients on PCA axes. All of the metrics used are based on Euclidean distances among species in trait space, and are appropriate for univariate and multivariate analyses. While some were designed to incorporate relative abundance, they are also suitable for the presence- absence data of this study.

Four measures of functional diversity capture different aspects of trait variation among species within communities and have been validated through simulation

(Mouchet et al., 2010; Villeger et al., 2008). Functional richness (FRic) quantifies the volume of the minimum convex hull containing all community members (Cornwell et al., 2006; Villeger et al., 2008). Functional evenness

(FEve) uses the sum of the branch lengths in the minimum spanning tree connecting all community members (Villeger et al., 2008). Functional divergence (FDiv) quantifies variation in the distances of community members 63 from the centroids (Villeger et al., 2008). Functional dispersion (FDis) is the average distance of species trait values from the centroids (Laliberté and

Legendre, 2010). I calculated FRic, FEve, FDiv, and FDis using the function dbFD in the R package FD (Laliberté and Legendre, 2010).

Three metrics previously used in research on community assembly captured different types of variation than the functional diversity metrics described above.

The minimum distance (MinDist) between two species in trait space suggests a minimum degree of similarity at which two taxa can coexist (Stubbs and Wilson,

2004). The mean of nearest neighbor distances (MNN) in trait space and the ratio of the minimum to maximum link in the minimum spanning tree

(RatMST) both provide measures of evenness (large values) or clumping (small values; Stubbs and Wilson, 2004).

For all of these metrics, large values indicate community members are broadly or evenly spread in trait space, while small values indicate community members are narrowly spread or clumped in trait space. High values are therefore consistent with divergence or sorting favoring reduced similarity, while low values are consistent with convergence or sorting for similar species.

All metrics were calculated for the three data sets: overall structure, size, and size-scaled shape. 64

2.5.2. Null models and significance tests for community distributions

Debates over appropriate null models for community patterns have persisted for

30+ years (Gotelli and Graves, 1996). In theory, the goal of null models is to incorporate all processes that could influence a pattern except those being tested for (Tokeshi, 1986); in practice, this can be difficult to achieve, as separating the contributions of various processes is not always straightforward. I used multiple null models that incorporate different assumptions about the processes generating variation in community membership and trait values.

I generated expected distributions of community metrics under two different null models: (1) random permutations of population means from all measured populations of all species (Strong et al., 1979) and (2) simulation of Brownian motion (BM) trait evolution on a phylogeny (Davies et al., 2012). Permutation null models generate an expectation based on the assumption that communities are assembled by random draw from a pool of possible populations or species. Null models using simulated trait evolution assume that the shared evolutionary history as captured by the tree of community members can shape the expected community trait distribution; for example, species that diverged recently are expected to have more similar trait values, and this affects the distribution of community members in trait space. BM models provided a better fit to the

65 woodpecker trait data, as quantified by AICc, than more parameter-rich models such as Ornstein-Uhlenbeck and early-burst models (not shown).

For each model, I generated samples specific to each community, containing the same number of taxa as were measured from the empirical community.

Permutation null models were generated by randomly sampling species, without replacement, from the set of all species included in this study (Figure 2.2). For each sampled species, I randomly selected a single population from all measured populations, and used the population mean trait values from that population.

Each iteration of the model is thus a set of populations equal in number to the number of species in the empirical community, with at most one population from each sampled species. I refer to this model as the general permutation (GP) to distinguish it from a more restricted permutation model described below.

Several species complexes included in this study either hybridize extensively or do not co-occur where their ranges come into contact: Sphyrapicus nuchalis, ruber, and varius; Colaptes auratus and cafer (Johnson and Johnson, 1985;

Short, 1965a). Though considered separate species under some species concepts, each group appears to act like a single species in ecological interactions. As such, including more than one of them in a null community would be analogous to sampling the same species more than once, and would bias null model communities towards including species with extremely similar trait values. 66

I therefore filtered random communities to allow only a single member of each of these species groups.

For the BM simulation model, trait evolution was simulated on a tree trimmed to include only measured species from the empirical community. BM models were fit and simulated using data rotated with phylogenetic PCA, as the orthogonal axes of PCA greatly simplify simulations of trait evolution. I estimated σ2 by fitting a BM model to the empirical population means from the observed community, using the phylogeny trimmed to include only community members. Models were fit with the fitContinuous function in the R package geiger (Harmon et al., 2008). I then simulated BM evolution on the selected tree with the fastBM function in the

R package phytools (Revell, 2012), to generate expected trait values.

Null distributions were generated by conducting 1000 iterations of each model simulation for each community, then calculating all metrics described above for each iteration. Empirical values for each community were then compared to these expected distributions. If divergent character displacement and/or species sorting for dissimilarity are operating, I expect empirical communities to have higher values of the community metrics described above than a large proportion of the randomly generated communities. Conversely, if convergent character displacement and/or sorting for similarity are operating, I expect lower values of community metrics in empirical communities. 67

Finally, for any communities that showed deviations from randomness in their traits, I used an additional null model to test the roles of species sorting and local evolution. This model, termed the within-species population permutation (WSPP), samples a single population at random from each species present in the given observed community (Figure 2.3). This generates a set of simulated communities that sample the variation among populations within the component species. I calculated all metrics for each of these simulated communities to generate a distribution of expected values. I then compared distributions of metrics from the

WSPP to values from the observed communities and distributions from the general permutation model.

If in situ evolution is the primary driver of trait distributions, community-level patterns should be due to characteristics of the particular populations within that community. In that case, I expect values from the empirical community to differ significantly from the WSPP model distribution, and for distributions generated under the WSPP model to be similar to those generated under the general permutation (GP) model. If values from the observed community exceed the values from a very large proportion of WSPP iterations, this suggests divergent local evolution; if values from the observed community are small than values from a large proportion of WSPP iterations, this suggests convergent local evolution.

If species sorting is driving trait distributions, non-random distributions should be recovered for any set of populations of the species present in the community. In that case, I expect values from the empirical community to be similar to the

WSPP model, and for distributions generated under the WSPP model to differ significantly from those generated under the general null model. If values from the WSPP model exceed the values from a very large proportion of general permutation model iterations, this suggests sorting for differences; if values from the WSPP model are smaller than the values from a large proportion of general permutation model iterations, this suggests sorting for similarity.

2.5.3. Community metrics, null models, and power to discriminate patterns

The combination of metrics and null models described above specifies a large number of tests, which could yield a small number of Type I errors. In assessing the results, I focus on general patterns across metrics and null models, as such

Type I errors are not expected to concentrate in particular trait sets or communities. In addition, the various community metrics and null models address different expectations and deviations from those expectations. Community metrics were selected to capture different types of variation in trait distributions, and communities may differ in some but not all of these metrics. In addition, the ideal null model is unknown, and testing against multiple null models accounts for different effects of the underlying assumptions. Finally, the tests used herein have relatively low power to discriminate patterns in communities with low 69 species richness (Mouchet et al., 2010). Accounting for multiple tests (e.g. via

Bonferroni correction) would further increase the Type II error rate, resulting in very low probability of recovering any true patterns.

2.6. Intraspecific Variation Across Communities

2.6.1. Displacement of focal populations from the species mean

It is possible that local populations could show evolutionary responses that are insufficient to generate community-level patterns. Some species may be more affected by character displacement than others, and such effects could be masked from detection by variation in other species. Testing for character displacement across all populations of each species could reveal such effects.

Alternatively, failure to detect community-level patterns could be due to lack of power in discriminating differences in trait distributions due to small sample sizes.

Testing for a common signal of character displacement across all populations could reveal shifts not detectable at the community level due to insufficient power in the community metrics.

To determine if populations appear to be consistently convergent with or divergent from the other members of their community, I quantified the displacement of populations in ecomorphospace relative to other species in the community (Figure 2.4). For each species in each community, I defined the community nearest neighbor as the population within the community that is 70 closest in morphospace to the species mean phenotype of the focal species. I then calculated the displacement towards (positive) or away from (negative) the community nearest neighbor as the difference between the distance from the community nearest neighbor to the species mean phenotype of the focal species and the distance from the community nearest neighbor to the focal population of the focal species. Positive values of this metric suggest convergence; negative values suggest divergence. I excluded species for which measurements were available for only a single population. I calculated average displacement values for each species across all communities, for all species in each community, and for all populations across all communities. Consistently positive values of displacement would indicate convergence at the species, community, or study scale, while negative values would indicate divergence.

2.6.2. Significance tests for displacement

I tested patterns of displacement against the null hypothesis that intraspecific variation is random with respect to other community members. First, I generated null distributions using permutation of populations similar to those described for the WSPP above. For each community, I retained the same set of species measured from the empirical community. For each of those species, I randomly selected a single population from all measured populations. I then calculated the displacement values for each community member. If local evolution as measured by displacement is consistently convergent with or divergent from nearest 71 neighbors in trait space, observed displacement values should be consistently greater than (convergence) or less than (divergence) values from simulated communities. Second, I tested for deviations from expected statistical distributions. If population variation is random with respect to other community members, we expect values of displacement to have a symmetric distribution about 0. I tested for deviations from a mean displacement of 0 using t-tests, assuming normally-distributed displacement values. I tested for symmetry of the distribution about 0 by comparing the counts of positive and negative values to a binomial distribution with probability of 0.5 (equal probability of negative and positive values). Large numbers of positive values would suggest convergence, while large numbers of negative values suggests divergence.

The permutation- and distribution-based tests described above were conducted for the overall morphology data, the average size data, and the size-scaled shape data. In order to detect possible signals of displacement at multiple levels,

I conducted tests on data aggregated across all populations in each community, all populations of each species, and all populations of all species in all communities.

3. Results

3.1. Measurement data and PCA axes

I measured a total of 271 individuals from 6 sites; these represent 34 populations of 10 species (Table 2.1). Figures 2.5 and 2.6 show the distribution of populations on the first two principal component axes (PC1 and PC2) for the overall morphology data and shape data. In the overall morphology data, PC1 captured 87.6% of the variation among species. All variables had negative coefficients; this axis represents size and allometric scaling relationships. PC2 had negative coefficients for skull measurements and positive coefficients for pygostyle measurements. The factors associated with this axis of variation have not been noted previously as being critical to community assembly, and require further investigation; in this sample of taxa, the axis primarily divides Melanerpes and Colaptes from other genera. PC3 approximates the ecological excavator- gleaner axis, with positive scores representing more robust excavators with large heads and pygostyles and short limbs, and negative scores representing gleaners with long limbs and small heads and pygostyles. For the data scaled by

PGLS regression on size, PC1 and PC2 show variable coefficients very similar to those for PC2 and PC3 of the overall morphology data.

3.2. Community metrics and results from null models

Tables 2.2-2.4 show significance values from null model tests of community metrics for the overall structure, size, and shape data. Due to the large number of 73 null models and metrics, I focus on general patterns across metrics and null models to identify communities that show consistent deviations from random expectations.

Choice of null model did not substantially alter results. I conducted ANOVA of the arcsin-transformed significance values (the proportion of simulations for which the empirical value of the metric exceeded that for the simulated community) to quantify variation in values explained by the null model and community. In nearly all cases, the communities explained the large majority of the variation. However, two metrics yielded problematic results for some or all null models, and in several cases the null model explained most of the variation in significance values. FRic showed consistent problems with the phylogenetic null models, as these models sometimes generated communities with species much more widely distributed in trait space than empirical communities; this appears to be due to simulated populations occupying regions of trait space that are not represented in actual species. FDiv showed anomalous results for some communities, with significance values markedly different than those for any other metrics.

Most communities did not show any deviations from random expectations, for any of the null models, for any of the metrics. However, a few communities did show consistent deviations from expectations in certain data sets. The

Washington community showed the strongest deviations from randomness in the 74 overall morphology data and average size data, with species more evenly distributed than expected. This was particularly pronounced for the average size data: of the 7 metrics in the 5% right tail of their respective null distributions

(indicating evenness or overdispersion) across all communities, metrics, and null models, all were for Washington. With the overall morphology data, 7 metrics were in the 5% right tail of their respective null distributions (again indicating overdispersion) across all communities, metrics, and null models; 4 of these were for Washington. This community also showed non-random distributions in the shape data, but in the opposite direction: excluding the problematic numbers from FRic, all 3 metrics in the 5% left tail of their respective null distributions were for Washington, suggesting that species were clumped or clustered in shape space. The Florida community appeared more evenly distributed than expected in the shape data only; 5 of the 9 metrics in the 5% right tail of their null distributions for were for Florida (the other 4 all corresponded to the problematic

FDiv metric).

No communities showed strong evidence of clumping in the overall morphology or average size data. Kansas and Michigan showed some evidence of clumping in the overall morphology data, and of evenness in the shape data; however, this was indicated by only one metric, FDiv. Given the idiosyncratic interactions among metrics, community trait patterns, and null models, recovering a result for a single metric does not provide strong evidence of a non-random distribution. 75

The even distributions in the overall morphology and size data in the Washington community appear to be driven by the set of species present in the community, as expected under species sorting, rather than by the particular populations of those species. Metrics from the empirical community did not differ from expected distributions based on the WSPP model (Table 2.5). Similarly, the WSPP model for the Washington community showed trait distributions that consistently differed from the general null models (Table 2.6). Thus, any community made up of any populations of that set of species would have been more evenly distributed in overall morphological trait space or in body size than expected at random.

In the shape data, the clumped distribution for the Washington community and the even distribution for the Florida community both appear to be due to particulars of those sets of populations, as expected if differential evolution is driving the patterns, rather than to characteristics of those species as a whole. In both cases, several metrics show the empirical communities to be outliers relative to the WSPP model (Table 2.5), and communities generated using the

WSPP model do not differ strongly from the communities simulated using the more general null models (Table 2.6).

3.3. Displacement

In general, individual communities and species did not show displacements biased toward positive or negative values (Table 2.7). Only two communities show statistical support for biased displacements, for the shape data only. Both

Florida and Michigan have mean displacements less than 0 using either t-tests or simulations (Florida only), and more negative displacement values than expected from a binomial distribution. Washington, which had non-random community-level trait distributions in all data sets, did not have biased displacement relative to nearest neighbors.

Across all communities and species, for the overall morphology data, size data, and shape data, many more of the displacements are negative than positive, as expected if divergent character displacement is occurring. This is marginally statistically significant for all three data sets. None of the mean displacements for all populations differ markedly from 0.

4. Discussion

4.1. Previous research on character displacement in communities

The current state of research on character displacement has been summarized in several reviews over the last decade (Dayan and Simberloff, 2005; Pfennig and

Pfennig, 2010; Schluter, 2000a). In brief, these reviewers concluded that (1) while many other processes also affect trait evolution, there is widespread observational evidence for divergent character displacement due to resource 77 competition in nature; (2) there is stronger evidence for character displacement at the species level than for non-random trait distributions within communities; and

(3) tests using body size or size of trophic morphology have been much more extensive than tests using shape. Theoretical work has repeatedly found that convergent character displacement is sometimes to be expected, especially when communities are highly diverse or intraspecific and interspecific competition are of similar intensity (Abrams, 1986; Scheffer and van Nes, 2006; Slatkin,

1980; Taper and Case, 1985), Despite this, few empirical studies have even tested for convergence (Dayan and Simberloff, 2005; but see Knouft, 2003;

Sidorovich et al., 1999).

While studies of character displacement in species pairs frequently attempt to exclude species sorting as a possible cause of observed patterns, almost all studies of community-wide character displacement have focused solely on patterns within communities, which do not allow discrimination of these two processes (Dayan and Simberloff, 2005). At least one prior study tested for patterns both within communities and within species across communities. Knouft

(2003) studied body size ratios in Etheostoma darters. He found evidence of convergent evolution in sympatry for two pairs of species. He also recovered an unusual pattern where the similarity of two species varied with the number of other Etheostoma species present. I did not test for context-dependent variation

78 such as that found in Etheostoma, as it is unclear which sets of species should be considered for such tests.

This study is unique in quantifying variation (1) among species within communities and among populations within species, (2) in body size and shape variables, and (3) relative to predictions of divergence and convergence in trait space. This allows discrimination of species sorting from trait evolution, assessment of the importance of body size versus shape characteristics in these processes, and evaluation of the rarely-tested predictions of convergent character displacement or and assortment for greater similarity.

4.2. Community Patterns

Results from analyses of community patterns indicate that trait distributions of most measured communities of North American woodpeckers do not deviate from random expectations. Interactions among species have not had strong effects on the overall patterns of trait variation in communities. The exceptions to this are the Washington community, which had species evenly distributed in overall morphology and size data, and clustered in shape data, and the Florida community, which had species evenly spaced in shape data.

Of the non-random trait distributions that were present, some are explained solely by the species found within the community. For overall morphology and 79 size data in the Washington community, any community made up of any of the measured populations of those species would show a similar non-random distribution. This suggests that local evolution is not responsible for the recovered pattern; rather, ecological sorting at some level is the more likely cause.

By contrast, all non-random distributions in the shape data are due to the characteristics of the particular populations found in those communities. This suggests that local evolution may be responsible. The two communities that show such patterns appear to have evolved in opposite directions; it is notable that both of these cases involve size-scaled shape data. Washington shows convergent evolution across species, with populations more clustered than expected. Florida shows divergent evolution, with populations more spread out in trait space than expected. This mixture of convergence and divergence is not totally unexpected, as modeling has shown that each can occur under certain conditions (Abrams, 1986; Scheffer and van Nes, 2006; Slatkin, 1980; Taper and

Case, 1985).

4.3. Displacement

While there is not strong evidence for character displacement in most communities and species, combining all populations across all sites yields a surprising surplus of displacements away from nearest neighbors in trait space, for overall morphology, size, and shape data. This suggests that, despite the 80 absence of significantly non-random trait distributions in most communities, divergent character displacement has occurred in these communities. One possible explanation for this is a lack of power in the tests of community metrics.

The ability to discriminate patterns of trait distributions from null models depends strongly on the number of community members (Kraft and Ackerly, 2010;

Mouchet et al., 2010). The recovery of displacement across all populations suggests that community-level effects are present, but do not differ strongly enough from null models to be detectable using the community metrics.

In addition, the Michigan and Florida communities show strong evidence of character displacement in shape space; displacements for all species are away from their nearest neighbors. This is particularly notable for Florida, given that community’s non-random distribution in shape space. Together, these tests provide evidence that Florida populations have evolved in situ to be more divergent in their ecomorphology. This may be attributable to more stable interactions over relevant timescales in the Florida community (see below).

This pattern of divergent character displacement warrants further study at finer spatial scales, targeted at populations spanning the range boundaries of nearest neighbors in trait space.

4.4. Size and overall structure versus shape 81

Patterns of overall structure and body size differ markedly from those for size- scaled shape. None of the communities have non-random trait distributions that are concordant between overall/size and shape; the only community that shows distinct patterns in both has species evenly spaced in overall/size and clustered in shape. This suggests that different processes are operating on the evolution and ecology of woodpeckers in size and shape.

It is somewhat surprising to find patterns in shape data but not in the overall or size data, as is the case in Florida. Most of the variation among woodpecker species and populations is in size, rather than in shape; a single PC axis, primarily representing body size, captured nearly 90% of the variation in the overall morphology data. In addition, body size is strongly connected to a wide range of ecological and evolutionary processes (Brown et al., 2004; Calder,

1984; LaBarbera, 1989; Peters, 1983; Trebilcol et al., 2013). One possibility is that the communities are consistently structured by size, but that this effect is incorporated into the null models through constraints on the sampling approach, and therefore not detectable in the observed communities in comparison to those models (Diamond and Gilpin, 1982). However, if that were the case, we should see still be able to detect patterns in size data using the null model of BM trait evolution. The BM model is not subject to sampling constraints, and is not expected to reflect any size structuring in the regional species pool. The results from the permutation and phylogenetic models are highly concordant (see Table 82

2.3), indicating that such model constraints are not preventing detection of true patterns.

Size and shape may correspond to different aspects of interspecific interaction in woodpeckers. Shape is correlated with diet and foraging modes in this group

(Burt, 1930; Leonard and Heath, 2010; Selander, 1966; Short, 1970, 1978;

Spring, 1965). However, this has not been explicitly compared to size and its connections to diet. It is notable that woodpeckers also compete for nesting cavities (Martin et al., 2004), the supply of suitable trees for cavity excavating may be population-limiting for some species (Lorenz et al., 2014; Raphael and

White, 1984), and preferred nest tree size varies with species size (Adkins Giese and Cuthbert, 2005). Similarity in size may thus be a better predictor of competition for nest sites than similarity in shape.

Further research is required to determine intensity of competition among woodpecker species based on similarity in size, in shape, or both. It may be that shape corresponds to niche differences, while size represents differences in both niche and competitive ability (Chesson, 2000; Mayfield and Levine, 2010).

4.5. Variation among communities

There are a number of possible explanations for why patterns would appear in only some communities. Communities may differ in age or stability, as species 83 respond differently to climatic fluctuations and habitat shifts. Such differences would change the length and constancy of interactions among species, and the expected outcomes of such interactions. While there are many published studies and maps of vegetation distribution over the past 100,000-200,000 years, no study has explicitly quantified temporal variation in vegetation across North

America over that time period. Such a study would be complicated by varying inferences from pollen records and climate-based vegetation models (Adams,

1997), and by idiosyncratic differences between sites. The Minnesota and

Michigan sites sampled here were mostly or entirely covered by glaciers at the last glacial maximum (LGM; Dyke and Prest, 1987), and subsequently probably covered by taiga and open boreal forest (Adams, 1997; Delcourt and Delcourt,

1981; Overpeck et al., 1992). Inferences of vegetation in Florida since the LGM vary widely, with some authors concluding that it was consistently forested, and others that it was at times covered by desert or scrub (Adams, 1997; Delcourt and Delcourt, 1981; Jackson et al., 2000; Overpeck et al., 1992; Watts and

Stuiver, 1980; Williams, 2002). The Kansas site may have experienced great variation since the LGM, as it sits at the edge of grassland and varied forest ecosystems (Adams, 1997; Overpeck et al., 1992; Williams, 2002). Variation at the Oregon and Washington sites is more difficult to summarize, as estimates of vegetation differ widely among studies and with local topography (Adams, 1997;

Davis, 1995; Ray and Adams, 2001; Thompson et al., 1993). Thus, despite abundant data and modeling, the only simple conclusions that can be drawn 84 regarding community-level patterns and estimated variation in vegetation since the LGM are the extreme changes in the Michigan and Minnesota sites following glacial retreat.

Despite this uncertainty, it is likely that most communities in this study would have had to shift or re-form following the last glacial maximum. Such shifts could be tested explicitly by inferring ranges of woodpeckers under historical climatic conditions, to determine if their ranges were expected to overlap throughout their recent evolutionary history, or by testing for concordant phylogeographic structure. It is important to note that communities need not remain in the same location in order to be fairly constant over evolutionary time; species may move in synchrony in response to climatic changes (Martínez-Meyer et al., 2004; Tingley et al., 2009). However, evidence from tree pollen records shows that not all species respond in the same way or at the same rate, resulting in historical community compositions without contemporary analogs (Davis, 1981; Jackson and Overpeck, 2000; Williams and Jackson, 2007). As birds do not have an abundant fossil record, it may be impossible to conclusively determine their historical interactions and variation among sites.

A phylogeographic study of one species (Leuconotopicus villosus, which occurs at all sites in this study), found deep population structure in western North

America, and a signature of population expansion in eastern North America 85

(Klicka et al., 2011). The same study found that the predicted distribution at the

LGM included Oregon and Washington but not Kansas, Michigan, or Minnesota.

These two results, if they are mirrored by other species, would strongly suggest greater community stability at the Washington and Oregon sites than the eastern sites (excepting Florida). This could explain the lack of community-level patterns in Kansas, Michigan, and Minnesota, but does not explain why the Oregon community shows no non-random trait distributions.

Alternatively, variation in topographical and habitat complexity may affect the frequency and intensity of interactions among species. Though woodpeckers are habitat generalists, more variable topography and habitats could facilitate spatial partitioning through habitat specialization, reducing interactions and any evolutionary effects of those interactions. However, of the two sites with the most variable topography and vegetation, Washington and Oregon, one shows numerous deviations from random expectations, while the other shows none.

Variation in resource abundance could also influence both the location of adaptive peaks in trait space and the intensity of interspecific competition

(Tilman, 1982), and may explain the variation in patterns among sites. However, assessing such effects would require quantifying resources at each site, which is beyond the scope of this study. The presence of predators can affect the intensity of competition among prey species (Paine, 1966; Sih et al., 1985). The primary 86 predators of woodpeckers are largely unknown, as direct observations of predation are rare (Winkler and Christie, 2002). However, there is no reason to suspect that their abundance would strongly vary among the sites studied here.

4.6. Is this character displacement?

Schluter and McPhail (1992) presented six criteria for character displacement.

While they focused on divergent effects of character displacement, the same criteria can be similarly applied to convergent character displacement. In order for differences to be considered character displacement, they should (1) not be explainable by chance, (2) have a genetic basis, (3) be the outcome of evolution rather than failure to coexist, (4) reflect differences in resource use (or similarity, for convergence), (5) not be due to among-site differences in resources, and (6) correspond to reduced competition for resources. The results presented herein satisfy criteria 1, 4, and probably 6. All patterns were tested against a variety of null models. The studied traits have documented connections to resource utilization, and there is some evidence that species compete for the corresponding resources. In addition, the patterns in shape data in the Florida and Washington community are not explained by species sorting, and therefore satisfy criterion 3. While it is possible that these patterns are due to plasticity, this is a problem common to most studies of character displacement. While some work has been done on the heritability of body size in wild birds (e.g. Gebhardt-

Henrich and van Noordwijk, 1991; Merilä, 1997), little work has been done on 87 other morphological variation (but see James, 1983), and heritability can vary from among populations and environments (Hoffmann and Merilä, 1999). Finally,

I do not have data on distributions of resources across sites – this should be addressed in future work. In conclusion, these differences may be character displacement under Schluter and McPhail’s criteria, but a number of assumptions remain to be tested.

5. Conclusions

I present evidence for both ecological sorting and local evolution driving non- random distributions of traits in communities. Species sorting was found to influence overall structure and body size distributions in one community.

Convergent and divergent evolution were each found in one community in shape space. Populations of all ten species across six sites showed an overall pattern of displacement in both size and shape. These varied results raise questions about the importance of size and shape in interspecific competition among woodpeckers, and when trait evolution is expected in response to such competition. In order to more accurately determine effects of interactions on evolution, we require better estimates of those interactions over time, through inference of historical ranges and vegetation shifts. In addition, a number of assumptions about the genetic and ecological nature of the studied traits still need to be tested.

Table 2.1 Site localities and number of individuals measured by site. Latitude and longitude are the centers of 50km radius circles.

Individuals Location of site measured by Site center sex # of Lat. Long. Species Species F M Florida 29.80 N 82.49 W 7 Colaptes auratus 1 1 Dryobates pubescens 5 6 Dryocopus pileatus 3 8 Leuconotopicus borealis 5 6 Leuconotopicus villosus 3 2 Melanerpes carolinus 6 5 Melanerpes erythrocephalus 4 3 Kansas 39.23 N 95.18 W 5 Colaptes auratus 3 2 Dryobates pubescens 6 5 Leuconotopicus villosus 7 5 Melanerpes carolinus 6 6 Melanerpes erythrocephalus 2 2 Michigan 42.40 N 83.30 W 5 Colaptes auratus 4 4 Dryobates pubescens 5 5 Leuconotopicus villosus 3 4 Melanerpes carolinus 4 5 Melanerpes erythrocephalus 4 4 Minnesota 44.97 N 93.09 W 6 Colaptes auratus 0 3 Dryobates pubescens 7 7 Dryocopus pileatus 2 4 Leuconotopicus villosus 2 4 Melanerpes carolinus 1 3 Melanerpes erythrocephalus 1 1 Oregon 42.48 N 122.22 W 6 Colaptes auratus 2 5 Dryocopus pileatus 3 1 Leuconotopicus villosus 5 7 Melanerpes formicivorus 5 3 Sphyrapicus ruber 4 5 Sphyrapicus thyroideus 1 6 Washington 47.44 N 121.93 W 5 Colaptes auratus 6 4 Dryobates pubescens 0 4 Dryocopus pileatus 6 7 Leuconotopicus villosus 5 4 Sphyrapicus ruber 4 5

Table 2.2 Community metric significance values for overall morphology data. Cell values indicate the proportion of 1000 simulations for which the observed value of the metric exceeded the simulated value for the given null model. Values close to 1 indicate evenness or overdispersion; values close to 0 indicate clustering. Proportions >= 0.95 and <= 0.05 are shown in bold.

Community Null model FRic FEve FDiv FDis MinDist MNN RatMST Florida permutation 0.880 0.878 0.335 0.723 0.702 0.272 0.701 phylogenetic 0.943 0.547 0.137 0.757 0.599 0.195 0.437 Kansas permutation 0.460 0.362 0.069 0.389 0.564 0.274 0.397 phylogenetic 0.790 0.292 0.041 0.598 0.616 0.304 0.277 Michigan permutation 0.413 0.417 0.126 0.366 0.594 0.290 0.441 phylogenetic 0.657 0.287 0.047 0.602 0.580 0.285 0.254 Minnesota permutation 0.925 0.525 0.137 0.848 0.960 0.511 0.705 phylogenetic 0.951 0.402 0.062 0.673 0.822 0.405 0.523 Oregon permutation 0.785 0.495 0.892 0.683 0.402 0.143 0.498 phylogenetic 0.968 0.295 0.615 0.899 0.381 0.190 0.263 Washington permutation 0.681 0.967 0.611 0.943 0.945 0.584 0.947 phylogenetic 0.966 0.970 0.449 0.790 0.969 0.619 0.945

Table 2.3 Community metric significance values for average size data. Cell values indicate the proportion of 1000 simulations for which the observed value of the metric exceeded the simulated value for the given null model. Values close to 1 indicate evenness or overdispersion; values close to 0 indicate clustering. Proportions >= 0.95 and <= 0.05 are shown in bold.

Community Null model FRic FEve FDis MinDist MNN RatMST Florida permutation 0.690 0.818 0.648 0.512 0.165 0.568 phylogenetic 0.910 0.659 0.709 0.404 0.125 0.359 Kansas permutation 0.447 0.325 0.349 0.479 0.210 0.483 phylogenetic 0.796 0.285 0.511 0.613 0.290 0.436 Michigan permutation 0.414 0.267 0.342 0.428 0.208 0.411 phylogenetic 0.779 0.235 0.491 0.487 0.207 0.341 Minnesota permutation 0.939 0.486 0.842 0.497 0.191 0.443 phylogenetic 0.881 0.386 0.692 0.291 0.111 0.231 Oregon permutation 0.495 0.649 0.758 0.759 0.325 0.757 phylogenetic 0.877 0.527 0.901 0.801 0.435 0.708 Washington permutation 0.838 0.974 0.975 0.956 0.625 0.961 phylogenetic 0.828 0.977 0.811 0.969 0.625 0.964

Table 2.4 Community metric significance values for shape data (measurement data scaled by average size). Cell values indicate the proportion of 1000 simulations for which the observed value of the metric exceeded the simulated value for the given null model. Values close to 1 indicate evenness or overdispersion; values close to 0 indicate clustering. Proportions >= 0.95 and <= 0.05 are shown in bold.

Community Null model FRic FEve FDiv FDis MinDist MNN RatMST Florida permutation 0.930 0.993 0.442 0.611 0.996 0.904 0.990 phylogenetic 0.969 0.934 0.814 0.833 0.985 0.872 0.911 Kansas permutation 0.582 0.461 0.970 0.521 0.823 0.479 0.697 phylogenetic 0.865 0.153 0.962 0.929 0.897 0.625 0.371 Michigan permutation 0.531 0.489 0.987 0.774 0.907 0.561 0.699 phylogenetic 0.780 0.222 0.979 0.939 0.918 0.597 0.388 Minnesota permutation 0.627 0.833 0.580 0.297 0.876 0.847 0.939 phylogenetic 0.757 0.505 0.679 0.800 0.793 0.863 0.698 Oregon permutation 0.804 0.455 0.215 0.307 0.759 0.327 0.674 phylogenetic 0.946 0.259 0.562 0.707 0.835 0.595 0.417 Washington permutation 0.345 0.739 0.017 0.037 0.566 0.555 0.788 phylogenetic 0.869 0.382 0.045 0.715 0.781 0.706 0.454

Table 2.5 Community metric significance values using the within-species population permutation (WSPP). Values shown are for data variants and communities with substantial deviations from the more general null models. Cell values indicate the proportion of 1000 simulations for which the observed value of the metric exceeded the simulated value for the given null model. Values close to 1 indicate evenness or overdispersion; values close to 0 indicate clustering. Proportions >= 0.95 and <= 0.05 are shown in bold.

Data variant Community FRic FEve FDiv FDis MinDist MNN RatMST Overall Morphology Washington 0.031 0.867 0.695 0.473 0.565 0.215 0.811 Average Size Washington 0.323 0.893 NA 0.517 0.665 0.258 0.873 Shape Washington 0.012 0.453 0.051 0.012 0.123 0.477 0.415 Shape Florida 0.938 0.965 0.368 0.763 0.983 0.930 0.953

Table 2.6 Comparison of community metrics from the within-species population permutation (WSPP) and the general permutation (GP). Significance values were determined by calculating metrics from the WSPP simulations, and calculating the proportion of times this metric exceeded the metric from the general permutation model simulations. Each cell contains the median significance value from 1000 simulations. Values close to 1 indicate distributions from WSPP are more even or overdispersed than those from general null models; values close to 0 indicate greater clustering in the WSPP relative to general null models. Proportions >= 0.95 and <= 0.05 are shown in bold.

Data variant Community Null Model FRic FEve FDiv FDis MinDist MNN RatMST Overall Washington permutation 0.865 0.926 0.530 0.946 0.945 0.793 0.892 Morphology phylogenetic 0.996 0.924 0.400 0.790 0.967 0.781 0.900 Average Size Washington permutation 0.872 0.954 NA 0.973 0.947 0.808 0.936 phylogenetic 0.844 0.929 NA 0.810 0.909 0.797 0.893 Shape Washington permutation 0.697 0.761 0.105 0.145 0.774 0.559 0.813 phylogenetic 0.982 0.420 0.147 0.881 0.926 0.715 0.489 Shape Florida permutation 0.735 0.806 0.506 0.404 0.787 0.612 0.832 phylogenetic 0.876 0.546 0.850 0.777 0.627 0.563 0.509

Table 2.7 Displacement values and results of statistical tests, across communities and species. See text for details of statistical tests. Probability values less than 0.05 are shown in bold.

Mean Binomial Data t-test Simulation Positive Negative variant Community Mean (SE) probability probability Count Probability Count Probability Overall Florida -0.0047 (0.0525) 0.932 0.548 2 0.891 4 0.344 Morphology Kansas -0.0039 (0.0131) 0.780 0.523 2 0.812 3 0.500 Michigan 0.0076 (0.0188) 0.706 0.669 3 0.500 2 0.812 Minnesota -0.0337 (0.0239) 0.218 0.216 2 0.891 4 0.344 Oregon -0.0461 (0.0241) 0.151 0.207 0 1.000 4 0.062 Washington -0.0249 (0.0515) 0.654 0.281 2 0.812 3 0.500 Total -0.0168 (0.0139) 0.237 NA 11 0.965 20 0.075 Average Florida -0.0044 (0.0176) 0.814 0.403 2 0.891 4 0.344 Size Kansas -0.0048 (0.0046) 0.357 0.380 1 0.969 4 0.188 Michigan 0.0042 (0.0062) 0.538 0.731 3 0.500 2 0.812 Minnesota -0.0079 (0.0075) 0.345 0.279 2 0.896 4 0.344 Oregon -0.0115 (0.0097) 0.320 0.231 1 0.938 3 0.312 Washington -0.0099 (0.0166) 0.585 0.217 2 0.812 3 0.500 Total -0.0055 (0.0046) 0.237 NA 11 0.965 20 0.075 Shape Florida -0.0264 (0.0079) 0.021 0.010 0 1.000 6 0.016 Kansas 0.0041 (0.0064) 0.558 0.919 3 0.500 2 0.812 Michigan -0.0096 (0.0019) 0.007 0.439 0 1.000 5 0.031 Minnesota -0.0089 (0.0067) 0.244 0.461 1 0.984 5 0.109 Oregon 0.0115 (0.0111) 0.375 0.974 3 0.312 1 0.938 Washington 0.0032 (0.0071) 0.676 0.910 3 0.500 2 0.812 Total -0.0057 (0.0035) 0.113 NA 10 0.985 21 0.035

Table 2.7 (continued)

Mean Binomial Data t-test Simulation Positive Negative variant Species Mean (SE) probability probability Count Probability Count Probability Overall Colaptes auratus -0.0499 (0.0416) 0.284 NA 3 0.656 3 0.656 Morphology Dryobates pubescens -0.0022 (0.0388) 0.957 NA 2 0.812 3 0.500 Dryocopus pileatus -0.0009 (0.0758) 0.992 NA 1 0.938 3 0.312 Leuconotopicus villosus -0.0288 (0.0158) 0.128 NA 1 0.984 5 0.109 Melanerpes carolinus -0.0090 (0.0171) 0.634 NA 1 0.938 3 0.312 Melanerpes erythrocephalus -0.0025 (0.0084) 0.781 NA 2 0.688 2 0.688 Sphyrapicus ruber 0.0061 (0.0559) 0.931 NA 1 0.750 1 0.750 Average Colaptes auratus -0.0152 (0.0131) 0.298 NA 3 0.656 3 0.656 Size Dryobates pubescens 0.0000 (0.0116) 1.000 NA 2 0.812 3 0.500 Dryocopus pileatus 0.0000 (0.0225) 1.000 NA 1 0.938 3 0.312 Leuconotopicus villosus -0.0132 (0.0086) 0.185 NA 2 0.891 4 0.344 Melanerpes carolinus 0.0000 (0.0100) 1.000 NA 1 0.938 3 0.312 Melanerpes erythrocephalus -0.0002 (0.0026) 0.942 NA 1 0.938 3 0.312 Sphyrapicus ruber 0.0000 (0.0180) 1.000 NA 1 0.750 1 0.750 Shape Colaptes auratus -0.0076 (0.0063) 0.279 NA 1 0.984 5 0.109 Dryobates pubescens -0.0169 (0.0112) 0.205 NA 1 0.969 4 0.188 Dryocopus pileatus -0.0073 (0.0110) 0.557 NA 2 0.688 2 0.688 Leuconotopicus villosus 0.0055 (0.0086) 0.549 NA 3 0.656 3 0.656 Melanerpes carolinus -0.0039 (0.0110) 0.746 NA 1 0.938 3 0.312 Melanerpes erythrocephalus -0.0096 (0.0071) 0.269 NA 1 0.938 3 0.312 Sphyrapicus ruber 0.0015 (0.0143) 0.931 NA 1 0.750 1 0.750

96 5 12

11 4 1 9,10

3 7

Figure 2.1. Morphological measurements taken from woodpecker skeletons. Red lines indicate measurements from cranium and post-cranial skeleton. Numbers correspond to those used in the list of detailed mea-

97 among speciesacrossthestudyarea. then repeatedtogenerateadistributionofsimulatedcommunitiesthatcapturevariation iteration, nopopulationswereselectedfromSpecies A orSpeciesE. This samplingprocessis populations areshowninblack,whilenotselectedgray. Inthis in theobservedcommunity, withatmostonepopulationperspecies.Intheexample,selected the model,arandomsetofpopulationsisselectedequalinnumbertomeasuredspecies with (a)thefullpoolofsamplepopulationsfromallcommunities.For(b)asingleiteration population mean;populationsofeachspeciessharethesameplotsymbol. The modelstarts Figure 2.2.Schematicofthegeneralpermutation(GP)model.Eachpointrepresentsasingle Trait Axis 2 (a) Population Pool Trait Axis1 Species E Species D Species C Species B Species A 98 (b) Single IterationofGP (3species) Trait Axis1 Species E Species D Species C Species B Species A (a) Full Population Pool (b) Observed Community Trait Axis 2

Species A Species A Species B Species B Species C Species C Species D Species D Species E Species E

Trait Axis 1 Trait Axis 1

Figure 2.3. Schematic of the within-species population permutation (WSPP) model. Each point represents a single population mean; populations of each species share the same plot symbol. The model starts with (a) the full pool of sampled populations from all communities. This pool is trimmed to include only populations from species in (b) the observed community, yielding (c) the reduced population pool. In the example, the observed community does not include Species A, so populations of Species A are excluded from the reduced population pool. For (d) a single iteration of the WSPP model, one population is selected at random from each species in the reduced population pool. In the example, selected populations are shown in black, while populations not selected are shown in gray. This sampling process is then repeated to generate a distribution of simulated communities that capture among-population variation in the species comprising the given community.

99 Species A

Species B Species C Species D

Species E Species G

Trait Axis 2 Species H (species mean)

Species F Species H (population mean)

Focal population Species mean Community nearest neighbor Other community members

Trait Axis 1

Figure 2.4. Hypothetical example of displacement of a population in trait space. The arrow shows the displacement in trait space of the focal population from the species mean.

100 (a) (b) 0.2 0.1 0.0 PC2 (6.53%) −0.1 Colaptes auratus Dryobates pubescens Dryocopus pileatus Leuconotopicus borealis Leuconotopicus villosus Florida Melanerpes carolinus Kansas

−0.2 Melanerpes erythrocephalus Michigan Melanerpes formicivorus Minnesota Sphyrapicus ruber Oregon Sphyrapicus thyroideus Washington

−1.5 −1.0 −0.5 0.0 0.5 1.0 1.5 −1.5 −1.0 −0.5 0.0 0.5 1.0 1.5 PC1 (87.6%) PC1 (87.6%)

Figure 2.5. Distribution of populations on the first two PCA axes of overall morphological data, with (a) points identified by species and (b) points identified by community. Note that the scale of the axes differs by approximately one order of magnitude; distances along the x-axis contribute much more variation than distances along the y-axis.

101 (a) (b)

Colaptes auratus Florida Dryobates pubescens Kansas Dryocopus pileatus Michigan Leuconotopicus borealis Minnesota Leuconotopicus villosus Oregon Melanerpes carolinus Washington Melanerpes erythrocephalus Melanerpes formicivorus

0.10 Sphyrapicus ruber Sphyrapicus thyroideus 0.05 PC2 (24.83%) 0.00 −0.05

−0.2 −0.1 0.0 0.1 0.2 −0.2 −0.1 0.0 0.1 0.2 PC1 (52.6%) PC1 (52.6%)

Figure 2.6. Distribution of populations on the first two PCA axes of shape data, with (a) points identified by species and (b) points identified by community.

102

Chapter 3

Diversification and trait evolution are correlated with coexistence in the

avian family Picidae

Abstract

Macroevolutionary theory predicts that rates of lineage diversification and morphological evolution should vary with species coexistence. Explicit tests of these predictions have been largely limited to island taxa and species pairs. I used phylogenetic comparative analyses of the global diversity of woodpeckers and allies (Aves: Picidae) to test for predicted relationships among clades between geographic range overlap and rates and modes of diversification and morphological evolution. I found positive correlations between rates of diversification, rates of shape evolution but not size evolution, and extent of range overlap with morphologically similar species. I also found better fits for time-dependent models of diversification and morphological evolution in clades with greater range overlap. Collectively, the results suggest that increases in clade diversity and local coexistence occur more rapidly with extensive differentiation in shape space. Picidae morphology is associated with diet and foraging mode, suggesting that ecological differentiation promotes both coexistence and clade diversification. The correspondence of this finding with results from analyses of community structure demonstrates that the probable 103 drivers of morphological evolution can extend from micro- to macro-evolutionary scales. Future research should explore the proximate mechanisms of interspecific interactions in Picidae and the possible contribution of latitudinal gradients in diversity to the patterns described here.

1. Introduction

The origin of lineage and phenotypic diversity is a major question in evolutionary research. Ecological interactions have clear effects on evolutionary processes over short timescales (Grant and Grant, 2006; Schluter, 1994; Schoener, 2011); however, it has been difficult to connect microevolutionary processes with macroevolutionary patterns. Various authors have attributed this to a lack of suitable data, inappropriate analyses, bounded evolution, or the effect of processes varying across timescales (Estes and Arnold, 2007; Gingerich, 1983,

2001; Gould, 2002; Uyeda et al., 2011).

A number of theories make specific predictions about how evolution should relate to coexistence and ecological interactions. The ecological theory of adaptive radiation lays out expectations for how lineages diversify and differentiate in trait space over time (Schluter, 2000b). Specifically, both diversification and trait evolution are expected to decline as diversity accumulates and ecological opportunity is reduced (Schenk et al., 2013; Schluter, 2000b; Yoder et al., 2010).

Character displacement theory predicts that rates of evolution may increase in 104 sympatry, as coexisting species repeatedly shift relative to each other in trait space (Abrams, 1986; Brown and Wilson, 1956; Carlson et al., 2009; Losos,

1995). However, such shifts may be limited by available trait space; the concept of ecological locking proposes that cumulative interspecific interactions may drastically reduce evolution in trait space, as competitive interactions prevent species from evolving to new optima (Morris et al., 1995). More recently, equilibrium states and ecological limits to diversification have been invoked to explain declining diversification through time (Cornell, 2013; Rabosky, 2009a, b), though actual ecological data are often absent from empirical studies of these processes.

Many studies of these processes have focused on taxa confined to islands or island-like habitats, and/or on pairs of closely related species (e.g. Chiba, 2004;

Gillespie, 2004; Grant and Grant, 2008; Losos, 2009; Radtkey, 1996; Robichaux et al., 1990; Schluter and McPhail, 1992). This allows for more straightforward reconstruction of historical processes, but results from these systems may not extend to complex continental radiations. Thus, additional tests are required using diverse clades with distribution patterns more representative of global diversity.

Studies that have tackled these ideas in continental systems have found varied results. Carlson et al. (2009) found higher rates of morphological evolution in 105 sympatry in Percina darters. Tobias et al. (2014) found no relationship existed between coexistence and divergence in traits related to competition, though their results suggest the opposite (see Discussion). Price et al. (2014) concluded that niche filling through morphological evolution early in a radiation led to declining rates of speciation over time. These findings call for additional research to clarify whether a general pattern exists, or whether each lineage is characterized by its own unique history.

The work described above used variation in rates among clades or divergence among sister taxa to quantify patterns of diversification and trait evolution.

However, the same theoretical processes are also expected to have effects on changes in rate over time, and in how such rate change varies across the tree

(Harmon et al., 2010; Rabosky, 2009b; Schluter, 2000b; Yoder et al., 2010).

Time-dependent rates have been examined in this context, primarily by fitting models to entire radiations (e.g. Rabosky and Lovette, 2008; but see Rabosky and Glor, 2010). Predicted patterns could also be revealed by testing for relationships between coexistence and the fit of constant-rate or variable-rate models of diversification and trait evolution to subclades of such radiations.

The avian family Picidae includes woodpeckers (subfamily Picinae) and the related piculets and wrynecks. Its approximately 237 species are distributed on all continents but Antarctica and Australia, and local diversity ranges from a 106 single species up to 13 or more (Short, 1978; Winkler and Christie, 2002). Most species share a common resource base: arthropods found on or within a substrate, especially wood (Winkler and Christie, 2002). Variation in specific morphological characters has documented relationships with differences in diet and foraging mode (Burt, 1930; Leonard and Heath, 2010; Selander, 1966; Short,

1970, 1978; Spring, 1965). Many species have high interspecific aggression

(Winkler and Christie, 2002), and experimental work shows that competition can affect diet and foraging behavior (Peters and Grubb, 1983). The Picidae are therefore an ideal system for studying relationships between coexistence, diversification, and morphological evolution.

In the previous chapter, I documented non-random patterns in woodpecker communities attributable to local body shape evolution. In the present study, I used a time-calibrated phylogeny, morphological measurements, and range data from a global sample of the avian family Picidae to test for macroevolutionary patterns predicted by theory and results from community-level analyses. I found strong relationships between shape evolution, diversification, and range overlap at macroevolutionary scales, and between the fit of time-dependent evolutionary models and degree of coexistence across clades. These results fit expectations from the theory of adaptive radiation, and reveal that microevolutionary processes may scale up to influence macroevolutionary patterns in this group. I

107 discuss implications of these findings, and propose future work to fill gaps in understanding of the effects of coexistence on the accumulation of diversity.

2. Materials and Methods

2.1. Taxon and Individual Sampling

The taxonomy used in this study is identical to that described in Ch. 1. In general, it follows the taxonomy proposed by Winkler et al. (2014), which incorporated revisions based on recent findings (Benz and Robbins, 2011; Benz et al., 2006;

Collar, 2011; García-Trejo et al., 2009; Moore et al., 2011; Pons et al., 2011).

To include a taxon in all analyses, I required both morphological measurements and phylogenetic information. Taxa with morphological data but no phylogenetic information were included in some, but not all, portions of analyses. Sampling was therefore designed to maximize both diversity of measured taxa and intersection with available phylogenetic information. Because skeletal preparations make up only a small percentage of ornithological collections

(Banks et al., 1973), many taxa had few or no measurable skeletons, and could not be included.

For each available taxon, I measured up to ten individuals (mean=6.15, range=1 to 61), including five males and five females if available. Several taxa had larger sample sizes due to the more extensive sampling required for Ch. 2. Juveniles 108 were excluded from analyses, as were obviously aberrant individuals, which may represent pathology or misidentification.

The family Picidae is composed of several sub-families, which differ substantially in morphology and ecology. Because these differences may reflect underlying differences in the historical evolutionary processes studied here, I conducted analyses including all Picidae, and including only genera unambiguously assigned to the subfamily Picinae (woodpeckers). The latter grouping excludes the genera Jynx, Picumnus, Sasia, and Verreauxia, as well as the ambiguously placed Nesoctites and Hemicircus. In addition, sampling of taxa outside of

Picinae was less extensive, as many members of the diverse South American genus Picumnus are poorly known.

2.2. Phylogenetic Information

All analyses incorporating phylogenetic information used the phylogenetic hypotheses of 178 species of Picidae presented in Ch. 1, based on analysis of a supermatrix DNA sequence dataset. Phylogenies of only Picinae contained 161 taxa. Phylogenies were trimmed as needed to include only taxa with morphological data. In four cases, one taxon had phylogenetic information but no morphological data, while a presumed closely related taxon had morphological data but no phylogenetic information: Gecinulus grantia and G. viridis,

Dendropicos pyrrhogaster and D. namaquus, Melanerpes herminieri and M. 109 portoricensis, and Piculus callopterus and P. simplex. In these cases, I replaced the former taxon with the latter in the trees, in order to include one member of these taxon complexes.

All analyses were conducted using the Maximum Likelihood tree hypothesis from

RAxML v. 8.0 (Stamatakis, 2014), with divergence times estimated in BEAST v.

1.7.5 (Drummond et al., 2012) using fossil and biogeographic calibrations.

Analyses were also conducted using the same tree topology with branch lengths scaled by r8s v. 1.8 (Sanderson, 2003); as results of these analyses were highly concordant, they are not presented here.

2.3. Morphological Measurements

For all sampled individuals, I measured a set of skeletal characters with documented correlations with foraging mode and diet (Burt, 1930; Leonard and

Heath, 2010; Short, 1978; Spring, 1965). Based on this previous work, variation in these characters is used here as a proxy for ecological similarity.

Measurement procedures were identical to those described in Ch. 2, and are described more fully therein. With digital calipers, I measured (1) greatest length of the cranium, (2) length of the cranium from the occipital condyle, (3) length of the mandible, (4) length of the humerus, (5) length of the ulna, (6) length of the femur, (7) length of the tibiotarsus, (8) length of the tarsometatarsus, (9) length of 110 the pygostyle, (10) width of the pygostyle. I used digital images and ImageJ

(Rasband, 1997-2015) to measure (1) length of the maxilla and (2) depth of the maxilla.

I adjusted for day-to-day variation in caliper measurements using data from reference specimens, as described in Ch. 2. All analyses were conducted with and without partial sets of skeletal measurements.

2.4. Processing of Measurement Data

Unless otherwise noted, all data processing and analysis was conducted in R (R

Core Team, 2015), using scripts written by the author (available at https://github.com/mjdufort).

Morphological data were cleaned and processed as described in Ch. 2.

Measurements were first ln-transformed. ANOVA of the individual-level data revealed that depth of the maxilla had high variance within species relative to among species, due either to measurement error or to intraspecific variation; I therefore excluded this measurement from analyses. Means were then taken by individual, sex, and taxon. If a taxon did not have at least one measured individual of each sex, I took the mean of all individuals regardless of sex. Taxa with no data for any single character were excluded. Taxon-level means were used in all downstream procedures. 111

For specimens with the ramphotheca intact, I imputed measurements of cranial characters without the ramphotheca by stochastic regression imputation (Enders,

2010), using predictions and variances from phylogenetic generalized least squares (PGLS) regression of taxon-level measurements without the ramphotheca on measurements with the ramphotheca (see Ch. 2 for additional details). I then recalculated sex and taxon means, including these imputed data.

I generated three sets of morphological data for use in later analyses: (1) unscaled variables or overall morphology, (2) average size, and (3) size-scaled shape variables. I calculated average size of each taxon as the mean across all variables of the log-transformed values. I generated size-corrected shape data by taking the residuals from PGLS regression of each variable on average size

(Revell, 2009). Analyses of morphological evolution are greatly simplified by orthogonal trait axes. I therefore rotated the overall morphology data and the size-scaled shape data using phylogenetic principal components analysis

(phylogenetic PCA; Revell, 2009), as implemented in the R package phytools

(Revell, 2012).

PGLS size correction and phylogenetic PCA exclude the few taxa with morphological data but no phylogenetic information. In order to include these taxa in calculations of similarity-scaled range overlap (see below), I applied the 112 transformations from PGLS and phylogenetic PCA to these data. For size-scaled shape data, I calculated the predicted value of each measurement from the

PGLS models, then calculated the difference between the actual value and that prediction (a pseudo-residual). To project the overall morphology data and the size-scaled shape data into the same coordinate system as the PCA-rotated data, I subtracted the phylogenetic mean from all values to center them, then multiplied the trait matrix by the PCA eigenvector matrix. This approach uses the phylogeny-aware regression models and PCA rotation to project morphological data from all species into a common coordinate system. These data from species without phylogenetic information were used only in calculating similarity-scaled range overlap values; they were not used for estimating rates or fitting models of morphological evolution.

2.5. Range Data and Overlap

I obtained digital range maps for all species of Picidae from BirdLife International and NatureServe (BirdLife International and NatureServe, 2014). These maps were updated in 2014 based on a thorough survey of published and unpublished data, and represent the most complete and up-to-date distribution data available for the family as a whole.

I modified the maps to reconcile the taxonomy used by BirdLife International and

NatureServe with the taxonomy used herein. Where necessary, I lumped or split 113 taxon ranges manually in QGIS v. 2.8 (QGIS Development Team, 2015). Most cases involved splitting or combining discontinuous ranges. In the few cases where I split contiguous ranges, I used published sources (Winkler and Christie,

2002; Winkler et al., 1995) to determine the appropriate distribution boundary between the split taxa. I further split ranges of migratory taxa into separate geometries for breeding and nonbreeding range, using the season labels specified for each polygon in the digital files. I visually checked all ranges in

QGIS to ensure that anomalous regions were not included.

I calculated range sizes and overlaps using the R package rgeos (Bivand and

Rundel, 2015). All geometries were reprojected from their original WGS84 coordinate system to the NSIDC equal-area grid (EPSG:3410). Invalid geometries were corrected by buffering with a width of 0. Geometries were flattened with the function gUnaryUnion to ensure that overlapping polygons would not duplicate portions of the range.

I calculated range sizes in m2 using the function gArea. For migratory species, I used the arithmetic mean of the breeding and nonbreeding ranges. I calculated pairwise overlaps for all pairs of species using the function gIntersection, then calculated the area of the resulting geometry. Overlaps for migratory species used the arithmetic mean of the breeding and nonbreeding range overlaps. For overlaps between two migratory species, I calculated the arithmetic mean of the 114 overlap of the two breeding ranges and the two nonbreeding ranges; as all migratory Picidae are boreal migrants, this ensures that seasonal overlap is represented as accurately as possible. Finally, I scaled all range overlaps by the range size of the focal taxon, so that overlap values represent the proportion of the focal taxon’s range that is also occupied by each other taxon, and summed these scaled overlaps for each focal taxon with all other taxa. These summed values can also be viewed as the average number of other Picidae (or Picinae) taxa present in the focal species’s range, averaged across the range. Using this quantity assumes that the strength of any relationship between range overlap and other variables is directly proportional to the fraction of a species range occupied by a possibly interacting species.

2.6. Similarity-scaled range overlap

Analyses were conducted with raw overlap values as described above, and with overlap scaled by similarity in trait space. I scaled the raw overlap values using the overall morphological data, average size, and size-scaled shape data. For each morphological data set, I generated a matrix of pairwise Euclidean distances among all taxa in trait space. For taxa lacking morphological data, I used close relatives with morphological data as proxies. This may overestimate the similarity-scaled overlap with close relatives, as proxies are treated as morphologically identical. I then divided the pairwise overlap values by the

115 corresponding Euclidean distance in trait space, and summed the scaled overlaps for each focal taxon with all other taxa.

This scaling of overlap values assumes that the strength of any relationship between range overlap and other variables is proportional both to the extent of overlap in distribution, and to ecomorphological similarity as measured by

Euclidean distance in trait space.

2.7. BAMM analyses of diversification and morphological evolution

I tested for large-scale variation in diversification and morphological evolution across the tree using Bayesian Analysis of Macroevolutionary Mixtures (BAMM;

Rabosky, 2014; Rabosky et al., 2014a). This approach uses reversible-jump

Markov Chain Monte Carlo to fit models with one or more diversification or morphological evolution regimes to subsets of the tree. Analyses were run on both the BEAST- and r8s-scaled trees. Diversification analyses used trees taxa without morphological data.

For diversification analyses, runs included four chains of 100,000,000 generations each, sampled every 50,000 generations, with the first 10,000,000 generations discarded as burnin. I incorporated non-random incomplete sampling by estimating sampling fractions for each taxon and clade. I assigned each unsampled taxon to the least inclusive group to which it could be reliably 116 attributed, based primarily on taxonomy (Winkler and Christie, 2002; Winkler et al., 1995; Winkler et al., 2014). I then calculated sampling fractions based on the quantity of sampled and unsampled taxa at each node in the tree.

For morphological evolution analyses, runs included four chains of 50,000,000 generations each, sampled every 25,000 generations, with the first 5,000,000 generations discarded as burnin. Morphological analyses were run separately for average size and for each of the first three PC axes of the overall morphology and size-scaled shape data.

Priors for each analysis were set with the setBAMMpriors function in the R package BAMMtools (Rabosky et al., 2014b). Default values were used for all other parameters. Assessment of convergence and post hoc analyses were conducted with BAMMtools.

2.8. Subclades and subclade metrics

I tested for relationships among clade-level variables using subclade regression, following the general approach of Adams et al. (2009) and Kozak and Wiens

(2010). I extracted all suitable subclades from the trimmed trees of taxa with morphological data. I included only subclades with at least 6 taxa, as this is the minimum number required to calculate AICc for some models of diversification

117 and morphological evolution. For each subclade, I calculated a number of subclade-level variables.

Many diversification metrics utilize both the tree structure and the total number of taxa in the clade, including unsampled taxa. To quantify unsampled taxa for each subclade, I mapped each unsampled taxon to one or more phylogenetic proxies, based primarily on taxonomy (Winkler and Christie, 2002; Winkler et al., 1995). A given unsampled taxon was then included in a subclade if all phylogenetic proxies were included in that subclade. For example, if a taxon can be reasonably placed within a genus, but not with any subset of that genus, then it was included in a subclade only if all sampled genus members were included.

Most anomalous Picidae taxa have been placed using molecular phylogenetics, as previous studies have intentionally sampled taxa with unclear placement.

Therefore, most the remaining taxa are apparent close relatives of sampled taxa, and can be placed with reasonable confidence. I believe the increased accuracy of subclade diversity outweighs the effect of any errors due to incorrect placement.

The relationship between clade diversity and clade age can help to determine if clade diversity is at or near equilibrium; in that case, total diversity may better explain variation in diversity and be more appropriate for analyses than diversification rates (Rabosky, 2009a). I calculated total diversification as the 118 natural logarithm of the total number of taxa, both sampled and unsampled, in the subclade. I calculated crown age as the depth in Ma of the node uniting all subclade members.

Diversification model fits and estimation of diversification rates were conducted with the full untrimmed tree, including taxa without morphological data. To match clades across these trees, I selected the least inclusive subclade from the full tree that contained all members of the subclade from the trimmed tree. This allows for a more accurate estimate of rates and model fits, but leads to differences in the set of taxa used to calculate morphological rates and models versus diversification rates and models.

If interspecific interactions have important effects on diversification, we may expect that diversification either increases or decreases with time or with picid diversity; in that case, models with time-dependent or diversity-dependent rate variation should have better fits to the data. To test this prediction and to estimate diversification rates, I fit constant-rate, time-dependent, and diversity- dependent diversification models to each subclade using functions in the R package DDD (Etienne and Haegeman, 2015; Etienne et al., 2012; Nee et al.,

1994). For all models, I calculated net diversification rate as estimated lambda – mu. I retained parameter estimates and sample-size corrected Akaike

Information Criterion (AICc) scores. In addition, I calculated the gamma statistic 119

(Pybus and Harvey, 2000), which quantifies the position of internodes relative to the tips and the root, and can provide a general indication of increasing vs. decreasing diversification. I also extracted average subclade rates of net diversification from BAMM runs described above, using the BAMMtools function getCladeRates.

We may also expect rates of morphological evolution to vary with time as diversity accumulates; if this is the case, early-burst models should fit the data better than constant-rate models such as Brownian motion. To test this prediction and estimate rates of morphological evolution, I fit models of continuous character evolution to the morphological data using the function fitContinuous in the R package geiger (Harmon et al., 2008). I fit Brownian motion (BM;

Felsenstein, 1973), Ornstein-Uhlenbeck (OU; Butler and King, 2004), and early-burst (EB; Harmon et al., 2010) models to each subclade. Models were fit independently to data for overall morphology, size, and size-scaled shape. I retained parameter estimates and AICc scores. In addition, I extracted average rates of morphological evolution from BAMM runs described above. Rates for overall morphology and size-scaled shape data were calculated as the sum of the rates from the first three PCA axes.

120

Finally, I calculated average overlap and average similarity-scaled overlap values by taking the arithmetic means across all subclade members, including those without phylogenetic or morphological data.

2.9. Random subclade regression

There are numerous ways to divide a phylogenetic tree into reciprocally monophyletic subclades, and any single division may result in biased estimates of relationships among subclade variables. To account for variation due to subclade selection, I randomly selected combinations of subclades, and fit regression models to subclade variables using each of these random combinations.

I generated random subclade combinations by selecting a single subclade at random, then selecting additional subclades at random until no suitable subclades remained, with the constraint that each additional subclade was reciprocally monophyletic with all subclades already selected for that combination. I retained only combinations with 5 or more suitable subclades.

For each random subclade combination, I trimmed the full tree to a backbone tree with the subclades as tips, and used the backbone tree in PGLS regression of the subclade variables. I truncated the terminal branches of these backbone trees at the depth of the most rootward node in the subclade, so that the branch 121 terminates at the crown of the subclade. I then fit PGLS regression models to the subclade metrics for each random combination of subclades. I tested for relationships between rates and models of diversification, rates and models of morphological evolution, and range overlap metrics (see Appendix 2 for a complete list of models fit). The goal of fitting these models was to test for relationships among subclades between rates and modes of diversification and morphological evolution, and geographic overlap with family members and ecomorphologically similar species. For each model, I summarized the distribution across all 100 random combinations of pseudo-R2, p-value for the model fit, and the slope of the relationship.

3. Results

3.1. Measurement data and PCA axes

I measured a total of 964 individuals from 151 out of 237 total taxa in Picidae.

The reduced data set of only Picinae contained 916 individuals, representing 134 of 203 total taxa. Appendix 1 shows the quantity of specimens measured by sex for each taxon.

Figure 3.1 shows the distribution of all taxa on the first two PCA axes (PC1 and

PC2) for the overall morphology and size-scaled shape data. PC1 of the overall morphology data captured 91.0% of all among-species morphological variation for Picidae and 93.4% for Picinae. All variables had high negative coefficients; 122 this axis primarily captures overall size and allometric scaling. PC2 of the overall morphology data captured 5.0% and 2.7% of variation for Picidae and Picinae, respectively. It had negative coefficients for skull measurements, positive coefficients for the pygostyle, and negative coefficients in Picidae for leg measurements. In this data set, it primarily differentiated the piculets (genera

Picumnus, Sasia, Verreauxia, and Nesoctites) from the remaining taxa, with no clear pattern within Picinae. PC3 of the overall morphology data corresponded to the previously documented excavator-gleaner axis of variation (Burt, 1930;

Leonard and Heath, 2010; Short, 1978; Spring, 1965), with opposing coefficients for skull and limb measurements.

For the size-scaled shape data, PC1 captured 55.7% and 40.5% and PC2 captured 19.6% and 27.2% of the variation for Picidae and Picinae, respectively.

Variable coefficients for PC1 and PC2 of the shape data correspond closely to those of PC2 and PC3 in the overall morphology data. PC3 of the shape data has opposing coefficients for length of the wing bones and leg bones, possibly indicating differences between more volant and more scansorial or terrestrial species.

3.2. Range data and combined data

Range sizes, raw overlaps, and similarity-scaled overlaps were calculated for all

237 Picidae and 203 Picinae taxa. Summed overlap scaled by focal species 123 range shows high variability among species (Figure 3.2), and is strongly correlated with summed overlaps scaled by morphological similarity (R2 = 0.96,

0.89, and 0.95 for overlaps scaled by average size, overall morphology, and size- scaled shape, respectively).

Morphological data, range data, and phylogenetic information were all available for 134 Picidae and 121 Picinae taxa (Appendix 1). These sets of taxa were used in analyses of morphological evolution.

3.3. BAMM analyses

Assessment of MCMC chains from BAMM analyses indicated that all analyses converged and reached stationarity within 5,000,000 generations. Effective sample sizes for the log-likelihood and number of shifts were >500 for all analyses.

Diversification analyses of all Picidae recovered two as the number of shifts with highest posterior probability (0.42); however, the single best configuration had zero shifts, and the 95% credible shift set also contained configurations with zero and one shift with high posterior probabilities (Figure 3.3). For models with shifts, a single shift was consistently placed near the base of Picinae, with an increase in diversification gradually declining to the background rate. When a second shift was included, it indicated an increase in diversification near the base of the 124 piculet genus Picumnus. Consistent with the homogeneity inferred for Picinae in the all-Picidae analyses, diversification analyses of Picinae only recovered zero shifts with high posterior probability (0.96). The best shift configuration and the only configuration in the 95% credible set included had a single regime across the tree. This discrepancy in diversification dynamics between Picinae and non-

Picinae reinforces the importance of conducting separate analyses with Picinae only.

Analyses of morphological evolution in both Picidae and Picinae consistently recovered zero shifts as both the highest posterior probability and single best configuration for average size and PC1 of the overall morphology data. Other

PCA axes of the overall morphology and size-scaled shape data recovered one or two shifts with high posterior probability; in nearly all cases, these inferred shifts indicated increases in rate either at the base of Picinae, or within the tribe

Malarpicini. Figure 3.4 shows average rates by branch for the first three PCA axes of the overall morphology data. Results for average size were nearly identical to those for overall morphology PC1, and size-scaled PC1 and PC2 are very similar to PC2 and PC3 from the overall morphology data, respectively.

3.4. Subclades and subclade metrics

The fits of models of diversification and morphological evolution to all subclades extracted from the phylogenies of Picidae and Picinae are shown in Table 3.1. 125

Diversity-dependent diversification models fit the data marginally better than constant-rate models and substantially better than time-dependent models.

However, for randomly selected sets of reciprocally monophyletic subclades, rates from constant-rate models were strongly positively correlated with rates from diversity-dependent models (median pseudo-R2 = 0.38, p = 0.03). In addition, rates from constant-rate models are simpler to use in the subclade regression analyses due to the difficulty in incorporating the K parameter of diversity-dependent models. Average net diversification rates from BAMM were strongly negatively correlated with diversification rates inferred using these other methods (with constant-rate models, median pseudo-R2 = 0.35, p = 0.0006). This appears to be due to BAMM’s inference of a declining rate of diversification over time, such that subclades closer to the tips have lower diversification rates; the opposite relationship is recovered by other methods. This issue does not extend to diversity-dependent diversification models fit by subclade because those models are fit independently to each subclade, rather than fitting a single model or small number of models across the entire tree. BAMM diversification analyses allowing only constant-rate diversification regimes resulted in models with largely homogeneous rates across the tree. I therefore focused on rates from the maximum-likelihood fit of a constant-rate model.

BM models provided better fits to the average size data, PC1 of the overall morphology data, and PC1 of the size-scaled shape data than did OU, EB, or 126 trend models (Table 3.1). Rates of morphological evolution from BAMM were strongly positively correlated with BM rates for all data sets. I used BM rate estimates in downstream analyses of rates of morphological evolution.

3.5. Random subclade combinations and subclade regression models

Random combinations of reciprocally monophyletic subclades from Picidae included an average of 10.2 subclades, containing on average 141.0 of the 178 total taxa in the full tree and 106.7 of the 134 total taxa with all data. Random combinations of reciprocally monophyletic subclades from Picinae included an average of 9.2 subclades, with 130.2 of the 161 total taxa in the full tree and 98.4 of the 121 total taxa with all data. The median number of combinations in which a subclade was included was 18.5 for Picidae and 18 for Picinae, but three subclades were included in 60 or more combinations for both Picidae and

Picinae: the genus Dryocopus, the genus Picoides, and a subset of Melanerpes.

Subclades comprising the genus Celeus showed unusually large values of range overlap (see Figure 3.5 and 3.7); results from analyses with and without these high-overlap subclades were not qualitatively different.

Table 3.2 shows details of PGLS models fit to random combinations of reciprocally monophyletic subclades, for a selection of variable combinations

(details on all models tested are in Appendix 2). Subclade diversity was

127 consistently positively correlated with crown age, supporting the use of diversification rates rather than total diversification in analyses (Rabosky, 2009a).

Across Picinae and all Picidae, the rate of shape evolution was strongly positively correlated with average range overlap, whether raw or scaled by similarity in shape space (Table 3.2; Figures 3.5 and 3.6). Rates of size and overall morphological evolution showed a trend towards negative correlations with average range overlap, but the model fits were generally not significant, and explanatory power of these models was low (pseudo-R2 between 0.02 and 0.04).

In Picinae, diversification rate was positively correlated with both the rate of shape evolution and average range overlap, whether unscaled or scaled by similarity in any of the morphological data sets (Table 3.2; Figures 3.7-3.9).

These relationships were only marginally significant in Picidae. Diversification rate was not correlated with rate of size or overall morphological evolution.

Variation in rates of diversification and trait evolution over time was related to overlap in some cases (Table 3.1 and Appendix 2). Gamma showed no significant correlation with average overlap, regardless of scaling (p = 0.18 to

0.48). In Picinae only, time-dependent and diversity-dependent diversification models provided better fits to subclades with greater overlap. Similarly, early- burst models fit better to subclades with greater overlap for all morphological data 128 sets. These relationships were not recovered in the broader Picidae. Significant correlations were not detected between parameter values from these models and overlap values in either group.

4. Discussion

The results presented here provide strong evidence for correlations between geographic range overlap, diversification rates, and rates of shape evolution across Picidae. In addition, declining-rate models of diversification and trait evolution provided better fits to subclades with high geographic overlap.

However, some predicted relationships were absent or weak, e.g. with rates of size and overall morphological evolution. Below, I discuss details of these findings and more general implications for how interspecific interactions relate to diversification and trait evolution.

4.1. BAMM analyses of diversification and morphological evolution

BAMM results indicate that the mode of diversification is consistent throughout

Picinae, but differs between Picinae and other groups within Picidae. Specifically,

BAMM placed an increase in diversification at the base of Picinae, followed by a decline to the background rate, and a lasting increase in diversification within the genus Picumnus. Previous studies have inferred a number of morphological changes related to woodpeckers’ scansorial habits occurring near the base of

Picinae (Manegold and Töpfer, 2013). Given the approximate concordance 129 between these traits and the inferred increase in diversification, it is tempting to point to these changes as key innovations that enabled Picinae to diversify more rapidly. However, given the unreplicated nature of these events, it is difficult to determine if morphological change enabled a burst of diversification, or if this is only a chance association (Maddison, 2006; Maddison and FitzJohn, 2015).

BAMM models of morphological evolution were uniform across all Picidae for average size and the first two PCA axes of the overall morphology data. BAMM recovered an increase in rate of morphological evolution in PC3 of the unscaled data at the base of the tribe Malarpicini. This PC axis reflects differences between excavators and gleaners, with excavators having larger heads and shorter limbs, and gleaners showing the reverse. Malarpicini contains several independent origins of highly terrestrial habits (Colaptes and Geocolaptes), as well as numerous more typically arboreal species (Novaes, 2013; Short, 1971a;

Winkler and Christie, 2002). Results from BAMM indicate that this ecological differentiation has been accompanied by corresponding higher rates of morphological evolution in related characters.

4.2. Random subclade regression

Analyses of subclade variables recovered positive correlations among diversification rates, shape evolution, and range overlap. Given this three-way correlation, causality is difficult to determine from the present analyses. None of 130 these three measures can be assumed as only a cause or only an outcome of the other two. Range overlap may drive diversification or morphological evolution.

Alternatively, both diversification and morphological evolution may be required in order for coexistence to increase. Similarly, morphological evolution may be either driven by or contributing to diversification. I consider these varied possibilities in more detail below.

The relationship between shape evolution and range overlap was the strongest detected, with both the highest explanatory power and lowest p-values. This is particularly interesting in light of results from Ch. 2, where coexisting populations showed greater differentiation in shape than expected at random. I attributed this differentiation to microevolution rather than ecological assortment, as it was based on morphology specific to the coexisting populations, rather than to the species. The pattern found here is what we would expect if such microevolution scales up to affect macroevolutionary patterns.

Carlson et al. (2009), in a study of Percina darters, also found higher rates of shape evolution with greater coexistence with close relatives. As with the present study, they could not conclusively determine causal relationships. Tobias et al.

(2014) found a conflicting result: trait divergence did not correlate with coexistence, after adjusting for the age of lineage splits. However, this conclusion is based on a model in which differences among species pairs in the relevant trait 131 are largely explained by an interaction between coexistence and the age of the split, rather than by the main effect of coexistence (see their Figure 4b); Tobias et al. incorrectly interpret this as a lack of relationship between coexistence and trait divergence, when in fact the opposite is true.

As noted above, shape differentiation may be either driven by or a prerequisite for coexistence. Determining the directionality of this relationship requires additional information, such as precise inference of the relative timing of coexistence and morphological changes. However, the coexistence of species without detectable differentiation in shape, as found for a number of communities in Ch. 2, points to overlap as the cause, rather than the effect. The connection between microevolutionary processes and macroevolutionary patterns has been a major topic in evolutionary research in recent years, as both fossil and phylogenetic analyses have recovered differences (Gould, 2002; Uyeda et al.,

2011). The correspondence between coexistence and shape evolution across timescales in woodpeckers provides additional evidence that micro- and macroevolution can in some cases reflect common underlying processes.

A positive correlation between diversification rate and range overlap was found in

Picinae, but was only marginally significant in Picidae. As with evolution in shape, diversification may be either the cause or the result of increased coexistence.

Some authors have proposed that local diversity can increase speciation 132

(Emerson and Kolm, 2005) (but see Cadena et al., 2005); however, a broad body of theoretical and empirical work indicates that diversification rates decline as diversity increases (Etienne et al., 2012; Rabosky and Lovette, 2008; Schluter,

2000b; Weir and Mursleen, 2013; Yoder et al., 2010). Clades occupying more diverse regions have more opportunity to coexist with other species. Given the body of previous work, I find it more likely that diversification has increased range overlap, rather than the converse. In general, we should expect the pattern recovered here unless variation in diversification is completely decoupled from geography, or species are strongly excluded from each other’s ranges.

A positive correlation between rates of shape evolution and diversification is predicted by both the ecological theory of adaptive radiation (Schluter, 2000b) and the theory of punctuated equilibrium (Eldredge and Gould, 1972; Gould and

Eldredge, 1977). A number of studies have examined corollaries of this prediction, such as the effect of key innovations on diversification (e.g. Dornburg et al., 2011), the relationship between species diversity and trait disparity

(Ricklefs, 2004), or the modes of accumulation of trait disparity and species diversity (e.g. Derryberry et al., 2011; Harmon et al., 2003). However, few have explicitly tested for correlations between rates of trait evolution and diversification, and recent phylogenetic tests have yielded conflicting results.

Studies of Plethodontid salamanders recovered a lack of correlation between rates of diversification and morphological evolution (Adams et al., 2009), but a 133 positive correlation between species richness and rate of morphological evolution

(Rabosky and Adams, 2012). Thacker (2014) found a negative correlation between diversification and shape evolution in gobies. Positive relationships between rates of diversification and morphological evolution have been recovered in ray-finned fishes (Rabosky et al., 2013) and the Madagascan vangas (Jønsson et al., 2012).

In the analyses presented here, Picinae showed a positive correlation between the rates of shape evolution and diversification; Picidae showed a marginally significant but consistently positive correlation. Picinae thus presents another empirical example supporting the importance of trait divergence in diversification.

Additional tests on other taxa are necessary to determine why some groups conform to these predictions while others do not.

If interspecific interactions have effects on diversification or trait evolution, we may expect these effects in the form of rate changes over time, rather than in consistent differences in rates over the entire history of a lineage. This is expected as adaptive radiation proceeds and available niche space is progressively filled (Schluter, 2000b). This is particularly true if interactions have increased to the present day, as these increased interactions should have a stronger effect on diversification and trait evolution in recent time. Such effects should be detectable by better fits of time-variable models of evolution in 134 subclades with greater overlap. These predicted relationships between model fits and overlap were recovered in Picinae but not in Picidae. Diversity-dependent and time-dependent models of diversification showed better fits (lower ΔAICc values) in subclades with higher range overlap. Early-burst models of shape evolution fit better with greater range overlap. Surprisingly, this was also true for size and overall morphological evolution. In all cases, these time-variable models consistently showed declining rates of diversification or morphological evolution over time.

Despite clades with greater range overlap showing higher rates of diversification and shape evolution, they also show more clearly declining rates over time. This suggests that these clades may be approaching saturation both in lineage diversity and in occupation of ecomorphospace.

4.3. Drivers of diversification, morphological evolution, and coexistence

In broad-scale studies of diversity, latitude is nearly universally correlated with variation in diversity. A huge range of organisms shows greater diversity in the tropics than in temperate or polar zones. While the causal relationships between latitude and diversity continue to be debated, the pattern itself is quite clear

(Hillebrand, 2004). This pattern of higher diversity in the tropics holds for Picidae

(Appendix 3). In addition, alpha diversity, which corresponds to coexistence as quantified in this study, also tends to be higher in the tropics (Swenson et al., 135

2012; Wright, 2002), and this is similarly true in Picidae (Short, 1978; Winkler and

Christie, 2002). Latitudinal variation in trait evolution is less clear, as evidenced in part by the proposed use of the neutral theory to explain tropical diversity.

It is possible that the observed correlations between diversification, shape evolution, and range overlap in Picinae and Picidae are all explained by underlying processes that covary with latitude. Higher diversification rates in the tropics may lead to greater range overlap in those same areas. Increased morphological evolution in tropical regions may be enabled by the availability of additional ecological roles; indeed, tropical woodpeckers appear to show greater variation in diet and microhabitat use than temperate species (Short, 1978;

Winkler and Christie, 2002), though this has not been explicitly tested.

The better fit of time-dependent models to clades with greater overlap is less easily explained by latitudinal variation. However, it is possible that older clades are more likely to show this time-dependent signature. Such older clades would also be more likely to occur in the tropics (Mittelbach et al., 2007; Wiens and

Donoghue, 2004), and therefore more likely to overlap with a greater diversity of other Picidae.

In addition to latitudinal gradients in diversity, research has also identified altitudinal variation, with greater diversity often found at intermediate elevations 136

(Rahbek, 1995, 2005), or in regions with greater topographic complexity

(Badgley, 2010; Cracraft, 1985; Kozak and Wiens, 2006). While the expected effects of topography on diversification are relatively clear, the causes of the latitudinal mid-elevation hump in diversity are not well resolved (McCain, 2009).

Diversity of Picidae does not have obvious relationships with altitude or topographic variation, but analyses of available data could reveal patterns that are not immediately apparent. Future research in this group should address latitudinal and altitudinal variation in diversity, morphology, and coexistence; a thorough analysis of these variables is beyond the scope of this study.

Finally, even if the correlations described here are effects of geographic variation in other processes, that does not diminish their importance in influencing the accumulation of diversity. In this group, elevated diversification and greater coexistence appears to require increased shape evolution. This is an interesting result regardless of the ultimate causes of those changes. There does not appear to be accumulation of neutral diversity in Picidae. The specific processes underlying this remain to be determined.

4.4. Shape vs. size in woodpeckers

Body size in animals has strong, well-documented connections to a wide range of biological processes (Brown et al., 2004; Calder, 1984; LaBarbera, 1989; Peters,

1983; Trebilcol et al., 2013). In many studies, it is used as the single 137 morphological proxy for ecological similarity (e.g. Davies et al., 2012; Dossena et al., 2012; Kozak et al., 2009). Given this ubiquity of size as an important ecological variable, it is somewhat surprising that significant correlations were found with body shape evolution, but not with body size evolution, in the present data. Many woodpecker communities show apparent structuring based on size

(Short, 1978; Winkler and Christie, 2002), and at least some of these patterns are statistically non-random (e.g. Washington community in Ch. 2). I attributed this pattern to species assortment at large geographic scales; if this process operates over greater spatial and temporal scales, we would expect higher rates of size evolution to correspond with increased range overlap. This was not the case in the present analyses; the relationship between size evolution and range overlap, though generally not significant, tended toward a negative correlation.

One possible explanation is that communities of Picidae may be more saturated along the size axis, or have approached saturation along that axis earlier, than along shape axes. This would result in species becoming ecologically “locked” along the size axis (Morris et al., 1995), as an evolutionary shift to a new optimum would require passing through ecomorphological space already occupied by other species. Differentiation along shape axes could free lineages from this constraint, allowing increased morphological evolution, diversification, and sympatry with other species. Alternatively, evolution of size and shape may reflect different underlying genetic and developmental variation that shapes or 138 constrains the range of possible variation in these characteristics (Maynard Smith et al., 1985).

Past research on woodpecker ecology identified a number of morphological characters that correlate with diet and foraging mode, including many of the measurements used in this study. The results here and in the previous chapter suggest that variation in these characters may be more important for long-term interspecific interactions than body size per se. The proximate determinants of woodpecker coexistence merit additional research.

4.5. Difficulties identifying characters associated with diversification rate variation

Many recent analyses of trait-dependent diversification used models from the state-dependent speciation-extinction (SSE) modeling framework (e.g. BiSSE,

QuaSSE; FitzJohn, 2010; Maddison et al., 2007). A recent study using simulated traits on simulated and empirical phylogenies found alarmingly high false positive rates with BiSSE, and these problems are expected to apply to other models in the SSE framework (Rabosky and Goldberg, 2015). Other studies have highlighted similar challenges with discriminating true associations from spurious ones, especially for characters with low transition rates (Maddison and FitzJohn,

2015).

139

Whether the problems discussed in those studies apply to the methods used here remains to be determined, and should be addressed using simulation. One of the primary problems of SSE and similar models is the attribution of unreplicated shifts to modeled processes based on the proportion of the tree affected by those shifts. Treating subclades as the data points in PGLS analyses may alleviate some of these issues. Specifically, analyzing the subclade data in a

PGLS framework should address the problem of unreplicated changes, as transitions among subclades should show strong phylogenetic signal that can be accounted for at least in part by the tree-based correlation structure. Utilizing random combinations of subclades helps to avoid acquisition bias and anomalous results based on the specific set of subclades utilized. Simulation studies on subclade regression would be invaluable in evaluating the tendency towards Type I errors found with other similar methods.

4.6. Current versus historical distributions

Perhaps the biggest challenge of inferring evolutionary processes is the reliance on present data to understand the past. This is particularly true for cases where we have little ability to test our assumptions about connections between past and present. I used present distributions to approximate historical interactions among species. Especially in the temperate zone, current distributions may have shifted in recent time, and not reflect important past interactions. However, I contend that they are a better approximation than ignoring geography entirely, particularly for 140 tropical species that make up the majority of picid diversity. Range overlaps as quantified here show substantial phylogenetic signal (Blomberg et al.’s (2003) K

= 0.20, p=0.002; Pagel’s (1999) λ = 0.60, p < 0.0001), supporting the assumption that interactions deeper in the tree are reflected by current overlap.

Future work should focus on models that incorporate inference of historical interactions (e.g. Quintero et al., 2015), perhaps using biogeographical analyses to determine the time at which each clade colonized major land masses (Ree and

Smith, 2008; Ronquist, 1997). Projection of current ranges to past climatic conditions could provide an indication of the stability of interactions over time, though such projections rely on a number of questionable assumptions (Kearney,

2006; Nogués-Bravo, 2009; Soberón and Nakamura, 2009).

5. Conclusions

I found strong associations between diversification, shape evolution, and geographic overlap in woodpeckers and related species. These results indicate that morphological ecology and geography interact to determine the species and trait diversity in this group. In addition, fit of time-varying models to diversification and morphological evolution show that rates decline over time in clades with more extensive ecological interactions. This aligns with predictions from the ecological theory of adaptive radiation, and demonstrates that this group

141 warrants further study to determine the specific mechanisms generating these patterns.

142

Table 3.1 Fits of diversification and morphological evolution models to subclades extracted from phylogenies of Picinae and Picidae. Values shown are the median AICc and ΔAICc values from models fit to each subclade. ΔAICc values are relative to baseline AICc from constant-rate diversification models or Brownian motion models of morphological evolution.

Model AICc Model ΔAICc Taxon Set Data set Constant-rate Time-dependent Diversity-dependent Picidae Diversification 80.58 3.55 -1.59 Picinae Diversification 64.58 3.55 -1.59

Model AICc Model ΔAICc BM OU EB trend Picidae Average size -13.06 3.18 3.47 2.44 Overall morphology PC1 16.73 3.18 3.52 2.44 Size-scaled shape PC1 -18.57 2.70 3.47 2.70 Picinae Average size -13.06 3.19 3.52 2.71 Overall morphology PC1 15.98 3.26 3.58 2.77 Size-scaled shape PC1 -18.66 2.74 3.67 2.77

143

Table 3.2 Results from PGLS regression fits to subclade variables for select models. Values shown are the median from models fit to 100 random selections of reciprocally monophyletic subclades. Pseudo-R2 was calculated as the correlation between observed values and fitted values from the models.

Taxon sign of Set Variable 1 Variable 2 pseudo-R2 p-value slope Picidae Total diversification Crown age 0.30 0.05 + Diversification rate Rate of size evolution 0.02 0.73 + Diversification rate Rate of overall morphological evolution 0.03 0.71 + Diversification rate Rate of shape evolution 0.26 0.06 + Diversification rate Overlap 0.03 0.12 + Diversification rate Overlap scaled by size 0.23 0.06 + Diversification rate Overlap scaled by overall morphology 0.30 0.06 + Diversification rate Overlap scaled by shape 0.25 0.10 + Rate of size evolution Overlap 0.02 0.07 - Rate of size evolution Overlap scaled by size 0.04 0.04 - Rate of overall morphological evolution Overlap 0.02 0.14 - Rate of overall morphological evolution Overlap scaled by overall morphology 0.02 0.10 - Rate of shape evolution Overlap 0.50 0.008 + Rate of shape evolution Overlap scaled by shape 0.59 0.004 + ΔAICc for time-dependent diversification model Overlap 0.14 0.20 - ΔAICc for diversity-dependent diversification model Overlap 0.11 0.22 - ΔAICc for EB model of size evolution Overlap 0.11 0.27 - ΔAICc for EB model of overall morphological evolution Overlap 0.11 0.27 - ΔAICc for EB model of shape evolution Overlap 0.13 0.21 -

144

Table 3.2 (continued)

Taxon sign of Set Variable 1 Variable 2 pseudo-R2 p-value slope

Picinae Total diversification Crown age 0.24 0.11 + Diversification rate Rate of size evolution 0.02 0.45 + Diversification rate Rate of overall morphological evolution 0.04 0.34 + Diversification rate Rate of shape evolution 0.36 0.04 + Diversification rate Overlap 0.33 0.04 + Diversification rate Overlap scaled by size 0.37 0.03 + Diversification rate Overlap scaled by overall morphology 0.35 0.03 + Diversification rate Overlap scaled by shape 0.32 0.04 + Rate of size evolution Overlap 0.04 0.15 - Rate of size evolution Overlap scaled by size 0.04 0.13 - Rate of overall morphological evolution Overlap 0.02 0.24 - Rate of overall morphological evolution Overlap scaled by overall morphology 0.02 0.10 - Rate of shape evolution Overlap 0.64 0.002 + Rate of shape evolution Overlap scaled by shape 0.66 0.002 + ΔAICc for time-dependent diversification model Overlap 0.35 0.03 - ΔAICc for diversity-dependent diversification model Overlap 0.30 0.05 - ΔAICc for EB model of size evolution Overlap 0.31 0.05 - ΔAICc for EB model of overall morphological evolution Overlap 0.31 0.05 - ΔAICc for EB model of shape evolution Overlap 0.34 0.02 -

145 (a) Overall Morphological Data (b) Size-scaled Shape Data 0.2 0.5 0.1 0.0 0.0 PC2 (5.0%) PC2 (19.6%) −0.1 −0.5

Picinae −0.2 Picinae non−Picinae non−Picinae −1.0

−4 −3 −2 −1 0 1 2 −1.0 −0.5 0.0 0.5

PC1 (91.0%) PC1 (55.7%)

Figure 3.1. Phylomorphospace plots of all taxa. Axes represent the first two principal components for (a) overall morphology or unscaled data and (b) size-scaled shape data. Taxa included in Picinae are shown as black circles; non-Picinae are shown as gray circles. Line segments indicate branches in the phylogeny, and are included to indicate the phylogenetic structure of the data.

146 30 25 20 15 Frequency 10 5 0

0 5 10 15 Overlap scaled by focal taxon range size

Figure 3.2. Histogram of summed overlap values by taxon. Overlap values reflect the quantity of the focal taxon’s range occupied by each other taxon, summed across all taxa.

147 0.32

0.18

0.035

Figure 3.3. Results from BAMM diversification analysis for all Picidae. Colors show average rate of net diversification across all samples from the posterior distribution. The model with highest posterior probability included a single diversification regime across the entire tree. Other models with high posterior probability included shifts in diversification regime near the base of Picinae and the base of Picumnus (approximate location indicated by arrows). 148 (a) (b) (c)

0.13 0.01 0.0035

0.081 0.0059 0.0021

0.03 0.0014 0.00064

Figure 3.4. Results from BAMM analyses of morphological evolution for all Picidae, for PCA axes 1-3 of the overall morphology data. Colors show average rate of trait evolution across all samples from the posterior distribution. Black circles indicate locations of shifts in the model with the highest posterior probability for each data set. 149 Size Evolution Overall Morphological Evolution Size−scaled Shape Evolution

2 2 median R = 0.04 0.012

0.07 median R = 0.02 0.06 0.010 0.005 ) 2 0.05 σ 0.008 (

0.04 Rate 0.006 0.003 0.03 0.004 0.02 median R2 = 0.64 0.002 0.001 0.01

5 6 7 8 9 5 6 7 8 9 5 6 7 8 9

Average Range Overlap Average Range Overlap Average Range Overlap

Figure 3.5. Rates of morphological evolution versus average summed range overlap by subclade, for (a) average size, (b) overall morphological data, and (c) size-scaled shape data, for all Picinae subclades with all data for at least 6 taxa. Rates shown are σ2 from BM models fit to the data for each subclade. Lines reflect the median slope and intercept from PGLS models fit to 100 random selections of reciprocally monophyletic subclades. R2 values are pseudo-R2 calculated as the correlation between the observed and fitted values from the PGLS models.

150 distributions ofvalues(a)pseudo-R Figure 3.6.Boxplotsofparametervaluesfrommodelsfittorates ofmorphologicalevolutionandgeographicoverlapinPicinae.Boxplotsshow evolution. The dottedlinesin(b) and (c)areatp=0.05slope0,respectively. monophyletic subclades.Eachboxplotincludesvalues frommodelsfittoratesofoverallmorphologicalevolution,sizeandshape 0.0 0.2 0.4 0.6 0.8 1.0 a) Morphology Ov er all pseudo-R Siz e 2 Shape 2 , (b)p-values,and(c)slopesfrommodelsfittorandomly selectedcombinationsofreciprocally

0.0 0.2 0.4 0.6 0.8 1.0 b) Morphology Ov er all p-value 151 Siz e Shape

0.010 0.005 0.000 0.005 c) Morphology Ov er all slope Siz e Shape 0.30 0.30 0.25 0.25 0.20 0.20 ersification Rate Di v 0.15 0.15 median R2 = 0.33 median R2 = 0.36 0.10 0.10

5 6 7 8 9 0.002 0.004 0.006 0.008 0.010 0.012

Average Range Overlap Rate of Size−scaled Shape Evolution

Figure 3.7. Diversification rate versus (a) average summed range overlap and (b) rate of shape evolution by subclade, for all Picinae subclades with all data for at least 6 taxa. Diversification rates are net rates (lambda-mu) from constant-rate maximum likelihood models fit to the each subclade. Lines reflect the median slope and intercept from PGLS models fit to 100 random selections of reciprocally monophyletic subclades. R2 values are pseudo-R2 calculated as the correlation between the observed and fitted values from the PGLS models.

152 2 a) pseudo-R b) p-value c) slope 1.0 1.0 0.10 0.8 0.8 0.08 0.6 0.6 0.06 0.04 0.4 0.4 0.02 0.2 0.2 0.00 0.0 0.0

Unscaled Overall Size Shape Unscaled Overall Size Shape Unscaled Overall Size Shape Morphology Morphology Morphology

Figure 3.8. Boxplots of parameter values from models fit to diversification rates and geographic overlap in Picinae. Boxplots show distributions of values of (a) pseudo-R2, (b) p-values, and (c) slopes from models fit to randomly selected combinations of reciprocally monophyletic subclades. Each boxplot includes values from models to unscaled geographic overlaps and overlaps scaled by differences in overall morphology, size, and shape. The dotted lines in (b) and (c) are at p = 0.05 and slope = 0, respectively.

153 2 a) pseudo-R b) p-value c) slope 1.0 1.0 30 0.8 0.8 20 10 0.6 0.6 0 0.4 0.4 10 0.2 0.2 20 30 0.0 0.0

Overall Size Shape Overall Size Shape Overall Size Shape Morphology Morphology Morphology

Figure 3.9. Boxplots of parameter values from models fit to diversification rates and rates of morphological evolution in Picinae. Boxplots show distributions of values of (a) pseudo-R2, (b) p-values, and (c) slopes from models fit to randomly selected combinations of reciprocally monophyletic subclades. Each boxplot includes values from models fit to rates of morphological evolution, size evolution, and shape evolution. The dotted lines in (b) and (c) are at p = 0.05 and slope = 0, respectively.

154

Bibliography

Aberer, A.J., Krompass, D., Stamatakis, A., 2013. Pruning rogue taxa improves phylogenetic accuracy: An efficient algorithm and webservice. Syst. Biol. 62, 162-166. Abrams, P.A., 1986. Character displacement and niche shift analyzed using consumer-resource models of competition Theoretical Population Biology 29, 107-160. Abrams, P.A., 2000. The evolution of predator-prey interactions: Theory and practice. Ann. Rev. Ecol. Syst. 31, 79-105. Adams, D.C., Berns, C.M., Kozak, K.H., Wiens, J.J., 2009. Are rates of species diversification correlated with rates of morphological evolution? Proc. R. Soc. Lond. B 276, 2729-2738. Adams, J.M., 1997. Global land environments since the last interglacial. Oak Ridge National Laboratory, Oak Ridge, Tennessee, USA. Adkins Giese, C.L., Cuthbert, F.J., 2005. Woodpecker nest tree characteristics in Upper Midwestern oak forests. Can. Field-Nat. 119, 367-376. Aitken, K.E.H., Martin, K., 2007. The importance of excavators in hole-nesting communities: availability and use of natural tree holes in old mixed forests of western Canada. J. Ornithol. 148, S425-S434. Alfaro, M.E., Holder, M.T., 2006. The posterior and the prior in Bayesian phylogenetics. Ann. Rev. Ecol. Evol. Syst. 37, 19-42. Artimo, P., Jonnalagedda, M., Arnold, K., Baratin, D., Csardi, G., de Castro, E., Duvaud, S., Flegel, V., Fortier, A., Gasteiger, E., Grosdidier, A., Hernandez, C., Ioannidis, V., Kuznetsov, D., Liechti, R., Moretti, S., Mostaguir, K., Redaschi, N., Rossier, G., Xenarios, I., Stockinger, H., 2012. ExPASy: SIB bioinformatics resource portal. Nucleic Acids Res. 40, W597-W603. Azevedo, L.d.S., Aleixo, A., Santos, M.P.D., Sampaio, I., Schneider, H., Vallinoto, M., Rêgo, P.S.d., 2013. New molecular evidence supports the species status of Kaempfer's Woodpecker (Aves, Picidae). Genet. Mol. Biol. 36, 192-200. Badgley, C., 2010. Tectonics, topography, and mammalian diversity. Ecography 33, 220-231. Banks, R.C., Clench, M.H., Barlow, J.C., 1973. Bird collections in the United States and Canada. Auk 90, 136-170. Baum, B.R., 1992. Combining trees as a way of combining data sets for phylogenetic inference, and the desirability of combining gene trees. Taxon 41, 3-10. Baumel, J.J., King, A.S., Breazile, J.E., Evans, H.E., Vanden Berge, J.C. (Eds.), 1993. Handbook of avian anatomy: Nomina anatomica avium. Nuttall Ornithological Club, Cambridge, Massachusetts, USA. Becker, J.J., 1986. Fossil birds of the Oreana Local Fauna (Blancan), Owyhee County, Idaho. Great Basin Naturalist 46, 765-768.

155

Benz, B.W., Robbins, M.B., 2011. Molecular phylogenetics, vocalizations, and species limits in Celeus woodpeckers (Aves: Picidae). Mol. Phylogen. Evol. 61, 29-44. Benz, B.W., Robbins, M.B., Peterson, A.T., 2006. Evolutionary history of woodpeckers and allies (Aves: Picidae): placing key taxa on the phylogenetic tree. Mol. Phylogen. Evol. 40, 389-399. BirdLife International, NatureServe, 2014. Bird species distribution maps of the world. Cambridge, UK and Arlington, VA, USA. Bivand, R., Rundel, C., 2015. rgeos: Interface to Geometry Engine - Open Source (GEOS). Blomberg, S.P., Garland, T., Jr., Ives, A.R., 2003. Testing for phylogenetic signal in comparative data: behavioral traits are more labile. Evolution 57, 717-745. Boucher, D.H. (Ed.), 1988. The biology of mutualism: Ecology and evolution. Oxford University Press, New York, New York, USA. Brodkorb, P., 1971. Catalogue of fossil birds: part 4 (Columbiformes through Piciformes). Bull. Fla. State. Mus. Biol. Sci. 15, 163-266. Brooks, D.R., McLennan, D.A., 1991. Phylogeny, ecology, and behavior: A research program in comparative biology. University of Chicago Press, Chicago, IL. Brown, J.H., 1975. Geographical ecology of desert rodents. In: Cody, M.L., Diamond, J.M. (Eds.), Ecology and evolution of communities. Belknap Press, Cambride, MA, USA, pp. 315-341. Brown, J.H., Gillooly, J.F., Allen, A.P., Savage, V.M., West, G.B., 2004. Toward a metabolic theory of ecology. Ecology 85, 1771-1789. Brown, J.W., Payne, R.B., Mindell, D.P., 2007. Nuclear DNA does not reconcile ‘rocks’ and ‘clocks’ in Neoaves: A comment on Ericson et al. Biol. Lett. 3, 257- 259. Brown, J.W., Rest, J.S., García-Moreno, J., Sorenson, M.D., Mindell, D.P., 2008. Strong mitochondrial DNA support for a Cretaceous origin of modern avian lineages. BMC Biol. 6, 6. Brown, W.L., Wilson, E.O., 1956. Character displacement. Syst. Zool. 5, 49-64. Bull, E.L., Jackson, J.A., 1995. Pileated woodpecker (Dryocopus pileatus). In: Poole, A. (Ed.), The Birds of North America Online. Cornell Lab of Ornithology, Ithaca, NY. Burleigh, J.G., Kimball, R.T., Braun, E.L., 2015. Building the avian tree of life using a large-scale, sparse supermatrix. Mol. Phylogen. Evol. 84, 53-63. Burt, W.H., 1930. Adaptive modifications in the woodpeckers. Univ. Cal. Pub. Zool. 32, 455-524. Butler, M.A., King, A.A., 2004. Phylogenetic comparative analysis: a modeling approach for adaptive evolution. Am. Nat. 164, 683-695. Cadena, C.D., Ricklefs, R.E., Jiménez, I., Bermingham, E., 2005. Ecology: Is speciation driven by species diversity? Nature 438, E1-E2.

156

Calder, W.A., 1984. Size, function, and life history. Harvard University Press, Cambridge, Massachusetts, USA. Carlson, R.L., Wainwright, P.C., Near, T.J., 2009. Relationship between species co-occurrence and rate of morphological change in Percina darters (Percidae: Etheostomatinae). Evolution 63, 767-778. Cavender-Bares, J., Kozak, K.H., Fine, P.V.A., Kembel, S.W., 2009. The merging of community ecology and phylogenetic biology. Ecol. Lett. 12, 693-715. Chesson, P., 2000. Mechanisms of maintenance of species diversity. Ann. Rev. Ecol. Syst. 31, 343-366. Chiba, S., 2004. Ecological and morphological patterns in communities of land snails of the genus Mandarina from the Bonin Islands. Journal of Evolutionary Biology 17, 131-143. Cicero, C., Johnson, N.K., 1995. Speciation in sapsuckers (Sphyrapicus): III. Mitochondrial-DNA sequence divergence at the cytochrome-B locus. Auk 112, 547-563. Clayton, D.H., Moore, J. (Eds.), 1997. Host-parasite evolution: General principles and avian models. Oxford University Press, Oxford, United Kingdom. Cockle, K.L., Martin, K., Wesolowski, T., 2011. Woodpeckers, decay, and the future of cavity-nesting vertebrate communities worldwide. Front. Ecol. Environ. 9, 377-382. Cody, M.L., 1969. Convergent characteristics in sympatric species: A possible relation to interspecific competition and aggression. Condor 71, 222-239. Collar, N.J., 2011. Species limits in some Philippine birds including the Greater Flameback Chrysocolaptes lucidus. Forktail 27, 29-38. Connor, E.F., Simberloff, D.S., 1979. The assembly of species communities: Chance or competition? Ecology 60, 1132-1140. Cornell, H.V., 2013. Is regional species diversity bounded or unbounded? Biological Reviews 88, 140-165. Cornwell, W.K., Schwilk, D.W., Ackerly, D.D., 2006. A trait-based test for habitat filtering: convex hull volume. Ecology 87, 1465-1471. Cracraft, J., 1985. Biological diversification and its causes. Annals of the Missouri Botanical Garden 72, 794-822. Darriba, D., Taboada, G.L., Doallo, R., Posada, D., 2012. jModelTest 2: more models, new heuristics and parallel computing. Nat. Methods 9, 772-772. Darwin, C., 1859. On the origin of species by natural selection. Murray, London, United Kingdom. Davies, T.J., Cooper, N., Diniz-Filho, J.A.F., Thomas, G.H., Meiri, S., 2012. Using phylogenetic trees to test for character displacement: a model and an example from a desert mammal community. Ecology 93, S44-S51. Davis, K.E., Page, R.D.M., 2014. Reweaving the tapestry: A supertree of birds. PLoS Curr. Tree Life.

157

Davis, M.B., 1981. Quaternary history and the stability of forest communities. In: West, D.C., Shugart, H.H., Botkin, D.B. (Eds.), Forest Succession: Concepts and Application. Springer-Verlag, New York, NY, pp. 132-177. Davis, O.K., 1995. Climate and vegetation patterns in surface samples from arid western U.S.A.: Application to Holocene climatic reconstructions. Palynology 19, 95-117. Dayan, T., Simberloff, D., 2005. Ecological and community-wide character displacement: the next generation. Ecol. Lett. 8, 875-894. De Pietri, V.L., Manegold, A., Costeur, L., Mayr, G., 2011. A new species of woodpecker (Aves; Picidae) from the early Miocene of Saulcet (Allier, France). Swiss J Palaeontol 130, 307-314. de Queiroz, A., Gatesy, J., 2007. The supermatrix approach to systematics. Trends Ecol. Evol. 22, 34-41. DeFilippis, V.R., Moore, W.S., 2000. Resolution of phylogenetic relationships among recently evolved species as a function of amount of DNA sequence: an empirical study based on woodpeckers (Aves: Picidae). Mol. Phylogen. Evol. 16, 143-160. del Hoyo, J., Collar, N.J., Christie, D.A., Elliott, A., Fishpool, L.D.C., 2014. Illustrated checklist of the birds of the world. Lynx Edicions, Barcelona, Spain. Delcourt, P.A., Delcourt, H.R., 1981. Vegetation maps for eastern North America: 40,000 yr BP to the present. In: Romans, R.C. (Ed.), Geobotany II. Plenum Press, New York, NY, USA, pp. 123-166. Derryberry, E.P., Claramunt, S., Derryberry, G., Chesser, R.T., Cracraft, J., Aleixo, A., Pérez-Emán, J., Remsen, J.J.V., Brumfield, R.T., 2011. Lineage diversification and morphological evolution in a large-scale continental radiation: The Neotropical ovenbirds and woodcreepers (Aves: Furnariidae). Evolution 65, 2973-2986. Diamond, J.M., 1975. Assembly of species communities. In: Cody, M.L., Diamond, J.M. (Eds.), Ecology and evolution of communities. Belknap Press of Harvard University Press, Cambridge, MA, pp. 342-444. Diamond, J.M., Gilpin, M.E., 1982. Examination of the "null" model of Connor and Simberloff for species co-occurrences on islands. Oecologia 52, 64-74. Dmitriev, D.A., Rakitov, R.A., 2008. Decoding of superimposed traces produced by direct sequencing of heterozygous indels. PLoS Comp. Biol. 4, e1000113. Dmitriev, D.A., Rakitov, R.A., 2008-2014. Indelligent v. 1.2. http://dmitriev.speciesfile.org/indel.asp. Dornburg, A., Sidlauskas, B., Santini, F., Sorenson, L., Near, T.J., Alfaro, M.E., 2011. The influence of an innovative locomotor strategy on the phenotypic diversification of triggerfish (Family: Balistidae). Evolution 65, 1912-1926. Dossena, M., Yvon-Durocher, G., Grey, J., Montoya, J.M., Perkins, D.M., Trimmer, M., Woodward, G., 2012. Warming alters community size structure and ecosystem functioning. Proceedings of the Royal Society B-Biological Sciences 279, 3011-3019. 158

Driskell, A.C., Ané, C., Burleigh, J.G., McMahon, M.M., O'Meara, B.C., Sanderson, M.J., 2004. Prospects for building the tree of life from large sequence databases. Science 306, 1172-1174. Drummond, A.J., Suchard, M.A., Xie, D., Rambaut, A., 2012. Bayesian phylogenetics with BEAUti and the BEAST 1.7. Mol. Biol. Evol. 29, 1969- 1973. Dyke, A.S., Prest, V.K., 1987. Late Wisconsinan and Holocene history of the Laurentide ice sheet. Géographie Physique et Quaternaire 41, 237-263. Edwards, S.V., 2009. Is a new and general theory of molecular systematics emerging? Evolution 61, 1-19. Eldredge, N., Gould, S.J., 1972. Punctuated equilibria: An alternative to phyletic gradualism. In: Schopf, T.J.M. (Ed.), Models in paleobiology. Freeman, Cooper and Co., San Francisco, CA, pp. 82-115. Emerson, B.C., Kolm, N., 2005. Species diversity can drive speciation. Nature 434, 1015-1017. Enders, C.K., 2010. Applied missing data analysis. Guilford, New York, New York, USA. Ericson, P.G.P., Anderson, C.L., Britton, T., Elzanowski, A., Johansson, U.S., Källersjö, M., Ohlson, J.I., Parsons, T.J., Zuccon, D., Mayr, G., 2006. Diversification of Neoaves: integration of molecular sequence data and fossils. Biol. Lett. 2, 543-547. Estes, S., Arnold, S.J., 2007. Resolving the paradox of stasis: models with stabilizing selection explain evolutionary divergence on all timescales. Am. Nat. 169, 227-244. Etienne, R.S., Haegeman, B., 2015. DDD: Diversity-dependent diversification. Etienne, R.S., Haegeman, B., Stadler, T., Aze, T., Pearson, P.N., Purvis, A., Phillimore, A.B., 2012. Diversity-dependence brings molecular phylogenies closer to agreement with the fossil record. Proceedings of the Royal Society B-Biological Sciences 279, 1300-1309. Felsenstein, J., 1973. Maximum-likelihood estimation of evolutionary trees from continuous characters. Am. J. Hum. Genet. 25, 471-492. FitzJohn, R.G., 2010. Quantitative traits and diversification. Syst. Biol. 59, 619- 633. Fleischer, R.C., Kirchman, J.J., Dumbacher, J.P., Bevier, L., Dove, C., Rotzel, N.C., Edwards, S.V., Lammertink, M., Miglia, K.J., Moore, W.S., 2006. Mid- Pleistocene divergence of Cuban and North American ivory-billed woodpeckers. Biol. Lett. 2, 466-469. Foster, P.G., 2004. Modeling compositional heterogeneity. Syst. Biol. 53, 485- 495. Fuchs, J., Ohlson, J.I., Ericson, P.G.P., Pasquet, E., 2006. Molecular phylogeny and biogeographic history of the piculets (Piciformes: Picumninae). J. Avian Biol. 37, 487-496.

159

Fuchs, J., Ohlson, J.I., Ericson, P.G.P., Pasquet, E., 2007. Synchronous intercontinental splits between assemblages of woodpeckers suggested by molecular data. Zool. Scr. 36, 11-25. Fuchs, J., Pons, J.-M., Liu, L., Ericson, P.G.P., Couloux, A., Pasquet, E., 2013. A multi-locus phylogeny suggests an ancient hybridization event between Campephilus and melanerpine woodpeckers (Aves: Picidae). Mol. Phylogen. Evol. 67, 578-588. Fuchs, J., Pons, J.M., Ericson, P.G.P., Bonillo, C., Couloux, A., Pasquet, E., 2008. Molecular support for a rapid cladogenesis of the woodpecker clade Malarpicini, with further insights into the genus Picus (Piciformes: Picinae). Mol. Phylogen. Evol. 48, 34-46. García-Trejo, E.A., Espinosa de los Monteros, A., Arizmendi, M.C., Navarro- Sigüenza, A.G., 2009. Molecular systematics of the red-bellied and golden- fronted woodpeckers. Condor 111, 442-452. Gebhardt-Henrich, S.G., van Noordwijk, A.J., 1991. Nestling growth in the great tit I. Heritability estimates under different environmental conditions. Journal of Evolutionary Biology 3, 341-362. Gibb, G.C., Kardailsky, O., Kimball, R.T., Braun, E.L., Penny, D., 2007. Mitochondrial genomes and avian phylogeny: complex characters and resolvability without explosive radiations. Mol. Biol. Evol. 24, 269-280. Gillespie, R.G., 2004. Community assembly through adaptive radiation in Hawaiian spiders. Science 303, 356-359. Gingerich, P.D., 1983. Rates of evolution: Effects of time and temporal scale. Science 222, 159-161. Gingerich, P.D., 2001. Rates of evolution on the time scale of the evolutionary process. Genetica 112-113, 127-144. Gold, M.S., Bentler, P.M., 2000. Treatments of missing data: a Monte Carlo comparison of RBHDI, iterative stochastic regression imputation, and expectation-maximization. Structural Equation Modeling 7, 319-355. Goodge, W.R., 1972. Anatomical evidence for phylogenetic relationships among woodpeckers. Auk 89, 65-85. Goodwin, D., 1968. Notes on woodpeckers (Picidae). Bull. Br. Mus. (Nat. Hist.) Zool. 17, 1-44. Gotelli, N.J., Graves, G.R., 1996. Null models in ecology. Smithsonian Institution Press, Washington, DC. Gould, S.J., 2002. The structure of evolutionary theory. Belknap Press, Cambridge, Massachusetts, USA. Gould, S.J., Eldredge, N., 1977. Punctuated equilibria: The tempo and mode of evolution reconsidered. Paleobiology 3, 115-151. Gouy, M., Guindon, S., Gascuel, O., 2010. SeaView version 4: a multiplatform graphical user interface for sequence alignment and phylogenetic tree building. Mol. Biol. Evol. 27, 221-224.

160

Graham, C.H., Parra, J.L., Tinoco, B.A., Stiles, F.G., McGuire, J.A., 2012. Untangling the influence of ecological and evolutionary factors on trait variation across hummingbird assemblages. 93, S99-S111. Grant, P.R., 1972. Convergent and divergent character displacement. Biol. J. Linn. Soc. 4, 39-68. Grant, P.R., Grant, B.R., 2006. Evolution of character displacement in Darwin's finches. Science 313, 224-226. Grant, P.R., Grant, B.R., 2008. How and why species multiply: The radiation of Darwin's finches. Princeton University Press, Princeton, New Jersey, USA. Gruber, K.F., Voss, R.S., Jansa, S.A., 2007. Base-compositional heterogeneity in the RAG1 locus among didelphid marsupials: Implications for phylogenetic inference and the evolution of GC content. Syst. Biol. 56, 83-96. Guindon, S., Gascuel, O., 2003. A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood. Syst. Biol. 52, 696-704. Hackett, S.J., Kimball, R.T., Reddy, S., Bowie, R.C.K., Braun, E.L., Braun, M.J., Chojnowski, J.L., Cox, W.A., Han, K.-L., Harshman, J., Huddleston, C.J., Marks, B.D., Miglia, K.J., Moore, W.S., Sheldon, F.H., Steadman, D.W., Witt, C.C., Yuri, T., 2008. A phylogenomic study of birds reveals their evolutionary history. Science 320, 1763-1768. Harmon, L.J., Losos, J.B., Davies, T.J., Gillespie, R.G., Gittleman, J.L., Jennings, W.B., Kozak, K.H., McPeek, M.A., Moreno-Roark, F., Near, T.J., Purvis, A., Ricklefs, R.E., Schluter, D., Schulte, J.A., Seehausen, O., Sidlauskas, B.L., Torres-Carvajal, O., Weir, J.T., Mooers, A.Ø., 2010. Early bursts of body size and shape evolution are rare in comparative data. Evolution 64, 2385-2396. Harmon, L.J., Schulte, J.A., Larson, A., Losos, J.B., 2003. Tempo and mode of evolutionary radiation in iguanian lizards. Science 301, 961-964. Harmon, L.J., Weir, J.T., Brock, C.D., Glor, R.E., Challenger, W., 2008. GEIGER: Investigating evolutionary radiations. Bioinformatics 24, 129-131. Heslewood, M.M., Elphinstone, M.S., Tidemann, S.C., Baverstock, P.R., 1998. Myoglobin intron variation in the Gouldian finch Erythrura gouldiae assessed by temperature gradient gel electrophoresis. Electrophoresis 19, 142-151. Hillebrand, H., 2004. On the generality of the latitudinal diversity gradient. The American Naturalist 163, 192-211. HilleRisLambers, J., Adler, P.B., Harpole, W.S., Levine, J.M., Mayfield, M.M., 2012. Rethinking community assembly through the lens of coexistence theory. Ann. Rev. Ecol. Evol. Syst. 43, 227-248. Hoffmann, A.A., Merilä, J., 1999. Heritable variation and evolution under favourable and unfavourable conditions. Trends Ecol. Evol. 14, 96-101. Holt, R.D., 1993. Ecology at the mesoscale: The influence of regional processes on local communities. In: Ricklefs, R.E., Schluter, D. (Eds.), Species diversity in ecological communities. University of Chicago Press, Chicago, Illinois, USA, pp. 77-88.

161

Honey-Escandón, M., Hernández-Baños, B.E., Navarro-Sigüenza, A.G., Benítez- Díaz, H., Peterson, A.T., 2008. Phylogeographic patterns of differentiation in the acorn woodpecker. Wilson J. Ornith. 120, 478-493. Hubbell, S.P., 2006. Neutral theory and the evolution of ecological equivalence. Ecology 87, 1387-1398. Hutchinson, G.E., 1959. Homage to Santa Rosalia, or, why are there so many kinds of animals? Am. Nat. 93, 145-159. Jackson, S.T., Overpeck, J.T., 2000. Responses of plant populations and communities to environmental changes of the late Quaternary. Paleobiology 26, 194-220. Jackson, S.T., Webb, R.S., Anderson, K.H., Overpeck, J.T., Webb, T., Williams, J.W., Hansen, B.C.S., 2000. Vegetation and environment in eastern North America during the Last Glacial Maximum. Quaternary Science Reviews 19, 489-508. James, F.C., 1983. Environmental component of morphological differentiation in birds. Science 221, 184-186. Jetz, W., Thomas, G.H., Joy, J.B., Hartmann, K., Mooers, A.O., 2012. The global diversity of birds in space and time. Nature 491, 444-448. Jiang, W., Chen, S.-Y., Wang, H., Li, D.-Z., Wiens, J.J., 2014. Should genes with missing data be excluded from phylogenetic analyses? Mol. Phylogen. Evol. 80, 308-318. Johnson, M.T.J., Stinchcombe, J.R., 2007. An emerging synthesis between community ecology and evolutionary biology. Trends Ecol. Evol. 22, 250-257. Johnson, N.K., Johnson, C.B., 1985. Speciation in sapsuckers (Sphyrapicus): II. Sympatry, hybridization, and mate preference in S. ruber daggetti and S. nuchalis. Auk 102, 1-15. Jønsson, K.A., Fabre, P.-H., Fritz, S.A., Etienne, R.S., Ricklefs, R.E., Jørgensen, T.B., Fjeldså, J., Rahbek, C., Ericson, P.G.P., Woog, F., Pasquet, E., Irestedt, M., 2012. Ecological and evolutionary determinants for the adaptive radiation of the Madagascan vangas. Proc. Natl. Acad. Sci. USA 109, 6620-6625. Kearney, M., 2006. Habitat, environment and niche: what are we modelling? Oikos 115, 186-191. Kesler, D.C., Walters, J.R., Kappes, J.J., Jr., 2010. Social influences on dispersal and the fat-tailed dispersal distribution in red-cockaded woodpeckers. Behavioral Ecology 21, 1337-1343. Klicka, J., Spellman, G.M., Winker, K., Chua, V., Smith, B.T., 2011. A phylogeographic and population genetic analysis of a widespread, sedentary North American bird: The hairy woodpecker (Picoides villosus). Auk 128, 346- 362. Knouft, J.H., 2003. Convergence, divergence, and the effect of congeners on body size ratios in stream fishes. Evolution 57, 2374-2382.

162

Kozak, K.H., Mendyk, R.W., Wiens, J.J., 2009. Can parallel diversification occur in sympatry? Repeated patterns of body-size evolution in coexisting clades of North American salamanders. Evolution 63, 1769-1784. Kozak, K.H., Wiens, J.J., 2006. Does niche conservatism promote speciation? A case study in North American salamanders. Evolution 60, 2604-2621. Kozak, K.H., Wiens, J.J., 2010. Accelerated rates of climatic-niche evolution underlie rapid species diversification. Ecol. Lett. 13, 1378-1389. Kraft, N.J.B., Ackerly, D.D., 2010. Functional trait and phylogenetic tests of community assembly across spatial scales in an Amazonian forest. Ecological Monographs 80, 401-422. Kraft, N.J.B., Valencia, R., Ackerly, D.D., 2008. Functional traits and niche-based tree community assembly in an Amazonian forest. Science 322, 580-582. Kubatko, L.S., Degnan, J.H., 2007. Inconsistency of phylogenetic estimates from concatenated data under coalescence. Syst. Biol. 56, 17-24. LaBarbera, M., 1989. Analyzing body size as a factor in ecology and evolution. Ann. Rev. Ecol. Syst. 20, 97-117. Laliberté, E., Legendre, P., 2010. A distance-based framework for measuring functional diversity from multiple traits. Ecology 91, 299-305. Lanyon, S.M., Zink, R.M., 1987. Genetic variation in Piciform birds: monophyly and generic and familial relationships. Auk 104, 724-732. Laybourne, R.C., Deedrick, D.W., Hueber, F.M., 1994. Feather in amber is earliest New World fossil of Picidae. Wilson Bull. 106, 18-25. Leibold, M.A., Holyoak, M., Mouquet, N., Amarasekare, P., Chase, J.M., Hoopes, M.F., Holt, R.D., Shurin, J.B., Law, R., Tilman, D., Loreau, M., Gonzalez, A., 2004. The metacommunity concept: a framework for multi-scale community ecology. Ecol. Lett. 7, 601-613. Lemmon, A.R., Brown, J.M., Stanger-Hall, K., Lemmon, E.M., 2009. The effect of ambiguous data on phylogenetic estimates obtained by maximum likelihood and Bayesian inference. Syst. Biol. 58, 130-145. Leonard, D.L., Jr., Heath, J.A., 2010. Foraging strategies are related to skull morphology and life history traits of Melanerpes woodpeckers. J. Ornithol. 151, 771-777. Little, R.J.A., Rubin, D.B., 2002. Statistical analysis with missing data, 2nd Edition. Wiley & Sons, New York, New York, USA. Liu, L., Yu, L., Pearl, D.K., Edwards, S.V., 2009. Estimating species phylogenies using coalescence times among sequences. Syst. Biol. 58, 468-477. Lorenz, T.J., Vierling, K.T., Johnson, T.R., Fischer, P.C., 2014. The role of wood hardness in limiting nest site selection in avian cavity excavators. Ecological Applications 25, 1016-1033. Losos, J.B., 1990. A phylogenetic analysis of character displacement in Caribbean Anolis lizards. Evolution 44, 558-569.

163

Losos, J.B., 1995. Community evolution in Greater Antillean Anolis lizards: phylogenetic patterns and experimental tests. Philosophical Transactions of the Royal Society B: Biological Sciences 349, 69-75. Losos, J.B., 2009. Lizards in an evolutionary tree: Ecology and adaptive radiation of anoles. University of California Press, Berkeley, California, USA. MacArthur, R.H., Levins, R., 1967. The limiting similarity, convergence, and divergence of coexisting species. Am. Nat. 101, 377-385. Maddison, W.P., 2006. Confounding assymetries in evolutionary diversification and character change. Evolution 60, 1743-1746. Maddison, W.P., FitzJohn, R.G., 2015. The unsolved challenge to phylogenetic correlation tests for categorical characters. Syst. Biol. 64, 127-136. Maddison, W.P., Midford, P.E., Otto, S.P., 2007. Estimating a binary character's effect on speciation and extinction. Syst. Biol. 56, 701-710. Manegold, A., Louchart, A., 2012. Biogeographic and paleoenvironmental implications of a new woodpecker species (Aves, Picidae) from the Early Pliocene of South Africa. Journal of Vertebrate Paleontology 32, 926-938. Manegold, A., Töpfer, T., 2013. The systematic position of Hemicircus and the stepwise evolution of adaptations for drilling, tapping and climbing up in true woodpeckers (Picinae, Picidae). Journal of Zoological Systematics and Evolutionary Research 51, 72-82. Martin, K., Aitken, K.E.H., Wiebe, K.L., 2004. Nest sites and nest webs for cavity- nesting communities in interior British Columbia, Canada: Nest characteristics and niche partitioning. Condor 106, 5-19. Martin, K., Eadie, J.M., 1999. Nest webs: a community-wide approach to the management and conservation of cavity-nesting forest birds. For. Ecol. Manage. 115, 243-257. Martínez-Meyer, E., Peterson, A.T., Hargrove, W.W., 2004. Ecological niches as stable distributional constraints on mammal species, with implications for Pleistocene extinctions and climate change projections for biodiversity. Global Ecology and Biogeography 13, 305-314. Mayfield, M.M., Levine, J.M., 2010. Opposing effects of competitive exclusion on the phylogenetic structure of communities. Ecol. Lett. 13, 1085-1093. Maynard Smith, J., Burian, R., Kauffman, S., Alberch, P., Campbell, J., Goodwin, B., Lande, R., Raup, D., Wolpert, L., 1985. Developmental constraints and evolution: A perspective from the Mountain Lake conference on development and evolution. The Quarterly Review of Biology 60, 265-287. McCain, C.M., 2009. Global analysis of bird elevational diversity. Global Ecology and Biogeography 18, 346-360. Merilä, J., 1997. Expression of genetic variation in body size of the collared flycatcher under different environmental conditions. Evolution 51, 526-536. Miller, M.A., Pfeiffer, W., Schwartz, T., 2010. Creating the CIPRES Science Gateway for inference of large phylogenetic trees. Gateway Computing Environments Workshop (GCE), New Orleans, LA, pp. 1-8. 164

Mirarab, S., Nguyen, N., Warnow, T., 2014. PASTA: ultra-large multiple sequence alignment. In: Sharan, R. (Ed.), Research in Computation Molecular Biology: 18th Annual International Conference, RECOMB 2014, Pittsburgh, PA, USA, April 2-5, 2014, Proceedings. Springer. Mittelbach, G.G., Schemske, D.W., Cornell, H.V., Allen, A.P., Brown, J.M., Bush, M.B., Harrison, S.P., Hurlbert, A.H., Knowlton, N., Lessios, H.A., McCain, C.M., McCune, A.R., McDade, L.A., McPeek, M.A., Near, T.J., Price, T.D., Ricklefs, R.E., Roy, K., Sax, D.F., Schluter, D., Sobel, J.M., Turelli, M., 2007. Evolution and the latitudinal diversity gradient: speciation, extinction and biogeography. Ecol. Lett. 10, 315-331. Mooers, A.Ø., Holmes, E.C., 2000. The evolution of base composition and phylogenetic inference. Trends Ecol. Evol. 15, 365-369. Moore, W.S., DeFilippis, V.R., 1997. The window of taxonomic resolution for phylogenies based on mitochondrial cytochrome b. In: Mindell, D.P. (Ed.), Avian molecular evolution and systematics. Academic Press, San Diego, CA, pp. 83-119. Moore, W.S., Miglia, K.J., 2009. Woodpeckers, toucans, barbets, and allies (Piciformes). In: Hedges, S.B., Kumar, S. (Eds.), The Timetree of Life. Oxford University Press, pp. 445-450. Moore, W.S., Overton, L.C., Miglia, K.J., 2011. Mitochondrial DNA based phylogeny of the woodpecker genera Colaptes and Piculus, and implications for the history of woodpecker diversification in South America. Mol. Phylogen. Evol. 58, 76-84. Moore, W.S., Weibel, A.C., Agius, A., 2006. Mitochondrial DNA phylogeny of the woodpecker genus Veniliornis (Picidae, Picinae) and related genera implies convergent evolution of plumage patterns. Biol. J. Linn. Soc. 87, 611-624. Morris, P.J., Ivany, L.C., Schopf, K.M., Brett, C.E., 1995. The challenge of paleoecological stasis: Reassessing sources of evolutionary stability. Proc. Natl. Acad. Sci. USA 92, 11269-11273. Mouchet, M.A., Villéger, S., Mason, N.W.H., Mouillot, D., 2010. Functional diversity measures: An overview of their redundancy and their ability to discriminate community assembly rules. Funct. Ecol. 24, 867-876. Nee, S., May, R.M., Harvey, P.H., 1994. The reconstructed evolutionary process. Philosophical Transactions of the Royal Society B 344, 305-311. Nogués-Bravo, D., 2009. Predicting the past distribution of species climatic niches. Global Ecology and Biogeography 18, 521-531. Novaes, L.S., 2013. Osteological adaptation for terrestrial habit in woodpeckers (Aves: Picidae). Department of Zoology. University of São Paulo. Nylander, J.A.A., Wilgenbusch, J.C., Warren, D.L., Swofford, D.L., 2008. AWTY (are we there yet?): a system for graphical exploration of MCMC convergence in Bayesian phylogenetics. Bioinformatics 24, 581-583.

165

Overpeck, J.T., Webb, R.S., Webb, T., 1992. Mapping eastern North American vegetation change of the last 18 Ka: no-analogs and the future. Geology 20, 1071-1074. Overton, L.C., Rhoads, D.D., 2006. Molecular phylogenetic relationships of Xiphidiopicus percussus, Melanerpes, and Sphyrapicus (Aves: Picidae) based on cytochrome b sequence. Mol. Phylogen. Evol. 41, 288-294. Pagel, M., 1999. Inferring the historical patterns of biological evolution. Nature 401, 877-884. Paine, R.T., 1966. Food web complexity and species diversity. Am. Nat. 100, 65- 75. Pamilo, P., Nei, M., 1988. Relationships between gene trees and species trees. Mol. Biol. Evol. 5, 568-583. Peters, R.H., 1983. The ecological implications of body size. Cambridge University Press, New York, New York, USA. Peters, W.D., Grubb, T.C., Jr., 1983. An experimental analysis of sex-specific foraging in the downy woodpecker, Picoides pubescens. Ecology 64, 1437- 1443. Pfennig, D.W., Pfennig, K.S., 2010. Character displacement and the origins of diversity. Am. Nat. 176, S26-S44. Pons, J.-M., Olioso, G., Cruaud, C., Fuchs, J., 2011. Phylogeography of the Eurasian green woodpecker (Picus viridis). J. Biogeogr. 38, 311-325. Price, T.D., Hooper, D.M., Buchanan, C.D., Johansson, U.S., Tietze, D.T., Alstrom, P., Olsson, U., Ghosh-Harihar, M., Ishtiaq, F., Gupta, S.K., Martens, J., Harr, B., Singh, P., Mohan, D., 2014. Niche filling slows the diversification of Himalayan songbirds. Nature 509, 222-225. Prum, R.O., 2014. Interspecific social dominance mimicry in birds. Zool. J. Linn. Soc. Prum, R.O., Samuelson, L., 2012. The hairy–downy game: A model of interspecific social dominance mimicry. J. Theor. Biol. 313, 42-60. Prychitko, T.M., Moore, W.S., 1997. The utility of DNA sequences of an intron from the b-fibrinogen gene in phylogenetic analysis of woodpeckers (Aves: Picidae). Mol. Phylogen. Evol. 8, 193-204. Prychitko, T.M., Moore, W.S., 2000. Comparative evolution of the mitochondrial cytochrome b gene and nuclear β-fibrinogen intron 7 in woodpeckers. Mol. Biol. Evol. 17, 1101-1111. Pybus, O.G., Harvey, P.H., 2000. Testing macro-evolutionary models using incomplete molecular phylogenies. Proc. R. Soc. Lond. B 267, 2267-2272. QGIS Development Team, 2015. QGIS Geographic Information System. Open Source Geospatial Foundation Project. Quintero, I., Keil, P., Jetz, W., Crawford, F.W., 2015. Historical biogeography using species geographical ranges. Syst. Biol. R Core Team, 2015. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. 166

Rabosky, D.L., 2009a. Ecological limits and diversification rate: alternative paradigms to explain the variation in species richness among clades and regions. Ecol. Lett. 12, 735-743. Rabosky, D.L., 2009b. Ecological limits on clade diversification in higher taxa. Am. Nat. 173, 662-674. Rabosky, D.L., 2014. Automatic detection of key innovations, rate shifts, and diversity-dependence on phylogenetic trees. PLoS ONE 9, e89543. Rabosky, D.L., Adams, D.C., 2012. Rates of morphological evolution are correlated with species richness in salamanders. Evolution 66, 1807-1818. Rabosky, D.L., Donnellan, S.C., Grundler, M., Lovette, I.J., 2014a. Analysis and visualization of complex macroevolutionary dynamics: An example from Australian Scincid lizards. Syst. Biol. 63, 610-627. Rabosky, D.L., Glor, R.E., 2010. Equilibrium speciation dynamics in a model adaptive radiation of island lizards. Proc. Natl. Acad. Sci. USA 107, 22178- 22183. Rabosky, D.L., Goldberg, E.E., 2015. Model inadequacy and mistaken inferences of trait-dependent speciation. Syst. Biol. 64, 340-355. Rabosky, D.L., Grundler, M., Anderson, C., Title, P., Shi, J.J., Brown, J.W., Huang, H., Larson, J.G., 2014b. BAMMtools: an R package for the analysis of evolutionary dynamics on phylogenetic trees. Methods in Ecology and Evolution 5, 701-707. Rabosky, D.L., Lovette, I.J., 2008. Density-dependent diversification in North American wood warblers. Proc. R. Soc. Lond. B 275, 2363-2371. Rabosky, D.L., Santini, F., Eastman, J., Smith, S.A., Sidlauskas, B., Chang, J., Alfaro, M.E., 2013. Rates of speciation and morphological evolution are correlated across the largest vertebrate radiation. Nature Communications 4, 1958. Radtkey, R.R., 1996. Adaptive radiation of day-geckos (Phelsuma) in the Seychelles archipelago: A phylogenetic analysis. Evolution 50, 604-623. Ragan, M.A., 1992. Phylogenetic inference based on matrix representation of trees. Mol. Phylogen. Evol. 1, 53-58. Rahbek, C., 1995. The elevational gradient of species richness: A uniform pattern? Ecography 18, 200-205. Rahbek, C., 2005. The role of spatial scale and the perception of large-scale species-richness patterns. Ecol. Lett. 8, 224-239. Rambaut, A., Suchard, M.A., Xie, W., Drummond, A.J., 2013. Tracer: MCMC trace analysis tool. http://beast.bio.ed.ac.uk/. Rannala, B., Yang, Z.H., 1996. Probability distribution of molecular evolutionary trees: A new method of phylogenetic inference. J. Mol. Evol. 43, 304-311. Raphael, M.G., White, M., 1984. Use of snags by cavity-nesting birds in the Sierra Nevada. Wildlife Monographs 86, 1-86. Rasband, W.S., 1997-2015. ImageJ. U. S. National Institutes of Health, Bethesda, Maryland, USA. 167

Ray, N., Adams, J.M., 2001. A GIS-based vegetation map of the world at the Last Glacial Maximum. Internet Archaeology 11, 1-44. Ree, R.H., Smith, S.A., 2008. Maximum likelihood inference of geographic range evolution by dispersal, local extinction, and cladogenesis. Syst. Biol. 57, 4-14. Revell, L.J., 2009. Size-correction and principal components for interspecific comparative studies. Evolution 63, 3258-3268. Revell, L.J., 2012. phytools: An R package for phylogenetic comparative biology (and other things). Methods in Ecology and Evolution 3, 217-223. Ricklefs, R.E., 2004. Cladogenesis and morphological diversification in passerine birds. Nature 430, 338-341. Ricklefs, R.E., 2008. Disintegration of the ecological community. Am. Nat. 172, 741-750. Robichaux, R.H., Carr, G.D., Liebman, M., Pearcy, R.W., 1990. Adaptive radiation of the Hawaiian silversword alliance (Compositae-Madiinae): Ecological, morphological, and physiological diversity. Annals of the Missouri Botanical Garden 77, 64-72. Ronquist, F., 1997. Dispersal-vicariance analysis: A new approach to the quantification of historical biogeography. Syst. Biol. 46, 195-203. Ronquist, F., Teslenko, M., van der Mark, P., Ayres, D.L., Darling, A., Höhna, S., Larget, B., Liu, L., Suchard, M.A., Huelsenbeck, J.P., 2012. MrBayes 3.2: efficient Bayesian phylogenetic inference and model choice across a large model space. Syst. Biol. 61, 539-542. Roure, B., Baurain, D., Philippe, H., 2013. Impact of missing data on phylogenies inferred from empirical phylogenomic data sets. Mol. Biol. Evol. 30, 197-214. Sanderson, M.J., 2003. r8s: inferring absolute rates of molecular evolution and divergence times in the absence of a molecular clock. Bioinformatics 19, 301- 302. Sanderson, M.J., Boss, D., Chen, D., Cranscton, K.A., Wehe, A., 2008. The PhyLoTA browser: processing GenBank for molecular phylogenetics research. Syst. Biol. 57, 335-346. Sanderson, M.J., Shaffer, H.B., 2002. Troubleshooting molecular phylogenetic analyses. Ann. Rev. Ecol. Syst. 33, 49-72. Scheffer, M., van Nes, E.H., 2006. Self-organized similarity, the evolutionary emergence of groups of similar species. Proc. Natl. Acad. Sci. USA 103, 6230-6235. Schenk, J.J., Rowe, K.C., Steppan, S.J., 2013. Ecological opportunity and incumbency in the diversiﬁcation of repeated continental colonizations by muroid rodents. Syst. Biol. 62, 837-864. Schluter, D., 1994. Experimental evidence that competition promotes divergence in adaptive radiation. Science 266, 798-801. Schluter, D., 2000a. Ecological character displacement in adaptive radiation. Am. Nat. 156, S4-S16.

168

Schluter, D., 2000b. The ecology of adaptive radiation. Oxford University Press, Oxford, UK. Schluter, D., McPhail, J.D., 1992. Ecological character displacement and speciation in sticklebacks. Am. Nat. 140, 85-108. Schoener, T.W., 2011. The newest synthesis: Understanding the interplay of evolutionary and ecological dynamics. Science 331, 426-429. Selander, R.K., 1966. Sexual dimorphism and differential niche utilization in birds. Condor 68, 113-151. Shaw, T.I., Ruan, Z., Glenn, T.C., Liu, L., 2013. STRAW: Species TRee Analysis Web server. Nucleic Acids Res. 41, W238-W241. Short, L.L., 1965a. Hybridization in the flickers (Colaptes) of North America. Bulletin of the American Museum of Natural History 129, 307-428. Short, L.L., 1965b. Variation in West Indian flickers (Aves, Colaptes). Bull. Fla. State. Mus. Biol. Sci. 10, 1-42. Short, L.L., 1970. Reversed sexual dimorphism in tail length and foraging differences in woodpeckers. Bird-Banding 41, 85-92. Short, L.L., 1971a. The evolution of terrestrial woodpeckers. American Museum Novitates 2467, 1-23. Short, L.L., 1971b. Systematics and behavior of some North American woodpeckers, genus Picoides (Aves). Bulletin of the American Museum of Natural History 145, 1-118. Short, L.L., 1978. Sympatry in woodpeckers of lowland Malayan forest. Biotropica 10, 122-133. Short, L.L., 1982. Woodpeckers of the world. Delaware Museum of Natural History, Greenville, Delaware, USA. Sidorovich, V., Kruuk, H., MacDonald, D.W., 1999. Body size and interactions between European and American mink (Mustela lutreola and M. vison) in eastern Europe. Journal of Zoology 248, 521–527. Sih, A., Crowley, P., McPeek, M., Petranka, J., Strohmeier, K., 1985. Predation, competition, and prey communities: A review of field experiments. Ann. Rev. Ecol. Syst. 16, 269-311. Simberloff, D., Boecklen, W., 1981. Santa Rosalia reconsidered: Size ratios and competition. Evolution 35, 1206-1228. Slatkin, M., 1980. Ecological character displacement. Ecology 61, 163-177. Soberón, J., Nakamura, M., 2009. Niches and distributional areas: Concepts, methods, and assumptions. Proceedings of the National Academy of Sciences 106, 19644-19650. Spring, L.W., 1965. Climbing and pecking adaptations in some North American woodpeckers. Condor 67, 457-488. Stamatakis, A., 2014. RAxML version 8: A tool for phylogenetic analysis and post-analysis of large phylogenies. Bioinformatics 30, 1312-1313.

169

Stralberg, D., Jongsomjit, D., Howell, C.A., Snyder, M.A., Alexander, J.D., Wiens, J.A., Root, T.L., 2009. Re-shuffling of species with climate disruption: a no- analog future for California birds? PLoS ONE 4, e6825. Strong, D.R., Jr., Szyska, L.A., Simberloff, D.S., 1979. Tests of community-wide character displacement against null hypotheses. Evolution 33, 897-913. Stubbs, W.J., Wilson, J.B., 2004. Evidence for limiting similarity in a sand dune community. Journal of Ecology 92, 557-567. Swenson, N.G., Erickson, D.L., Mi, X., Bourg, N.A., Forero-Montaña, J., Ge, X., Howe, R., Lake, J.K., Liu, X., Ma, K., Pei, N., Thompson, J., Uriarte, M., Wolf, A., Wright, S.J., Ye, W., Zhang, J., Zimmerman, J.K., Kress, W.J., 2012. Phylogenetic and functional alpha and beta diversity in temperate and tropical tree communities. Ecology 93, S112-S125. Swierczewski, E.V., Raikow, R.J., 1981. Hind limb morphology, phylogeny, and classification of the Piciformes. Auk 98, 466-480. Swofford, D.L., 2003. PAUP*. Phylogenetic analysis using parsimony (*and other methods). Version 4. Sinauer Associates, Sunderland, Massachusetts, USA. Taper, M.L., Case, T.J., 1985. Quantitative genetic models for the coevolution of character displacement. Ecology 66, 355-371. Thacker, C.E., 2014. Species and shape diversification are inversely correlated among gobies and cardinalfishes (Teleostei: Gobiiformes). Org. Divers. Evol. 14, 419-436. Thompson, R.S., Whitlock, C., Bartlein, P.J., Harrison, S.P., Spaulding, W.G., 1993. Climatic changes in the western United States since 18,000 yr B.P. In: Wright, H.E., Jr., Kutzbach, J.E., Webb, T.I., Ruddiman, W.F., Street-Perrott, F.A., Bartlein, P.J. (Eds.), Global climates since the Last Glacial Maximum. University of Minnesota Press, Minneapolis, Minnesota, USA, pp. 468-513. Thomson, R.C., Shaffer, H.B., 2010. Sparse supermatrices for phylogenetic inference: taxonomy, alignment, rogue taxa, and the phylogeny of living turtles. Syst. Biol. 59, 42-58. Tilman, D., 1982. Resource competition and community structure. Princeton University Press, Princeton, New Jersey, USA. Tingley, M.W., Monahan, W.B., Beissinger, S.R., Moritz, C., 2009. Birds track their Grinnellian niche through a century of climate change. Proc. Natl. Acad. Sci. USA 106, 19637-19643. Tobias, J.A., Cornwallis, C.K., Derryberry, E.P., Claramunt, S., Brumfield, R.T., Seddon, N., 2014. Species coexistence and the dynamics of phenotypic evolution in adaptive radiation. Nature. Tobias, J.A., Seddon, N., Spottiswoode, C.N., Pilgrim, J.D., Fishpool, L.D.C., Collar, N.J., 2010. Quantitative criteria for species delimitation. Ibis 152, 724- 746. Tokeshi, M., 1986. Resource utilization, overlap and temporal community dynamics: A null model analysis of an epiphytic chironomid community. J. Anim. Ecol. 55, 491-506. 170

Török, J., 1990. Resource partitioning among three woodpecker species Dendrocopos spp. during the breeding season. Holarctic Ecology 13, 257- 264. Trebilcol, R., Baum, J.K., Salomon, A.K., Dulvy, N.K., 2013. Ecosystem ecology: Size-based constraints on the pyramids of life. Trends Ecol. Evol. 28, 423- 431. Ugalde-Lezama, S., Tarango-Árambula, L.A., Ramírez-Valverde, G., Equihua- Martínez, A., Valdez-Hernández, J.I., 2011. Tropic coexistence of woodpeckers (Picidae) in a Pinus cembroides (Zucc.) forest of the natural protected area known as Peña Alta, San Diego de la Union, Guanajuato. Revista Chapingo Serie Ciencias Forestales y del Ambiente 17, 361-377. Urban, M.C., Leibold, M.A., Amarasekare, P., De Meester, L., Gomulkiewicz, R., Hochberg, M.E., Klausmeier, C.A., Loeuille, N., de Mazancourt, C., Norberg, J., Pantel, J.H., Strauss, S.Y., Vellend, M., Wade, M.J., 2008. The evolutionary ecology of metacommunities. Trends Ecol. Evol. 23, 311-317. Uyeda, J.C., Hansen, T.F., Arnold, S.J., Pienaar, J., 2011. The million-year wait for macroevolutionary bursts. Proc. Natl. Acad. Sci. USA 108, 15908-15913. Valcu, M., Dale, J., Kempenaers, B., 2012. rangeMapper: a platform for the study of macroecology of life-history traits. Global Ecology and Biogeography 21, 945-951. Vazquez-Miranda, H., 2014. Multilocus assessment of population differentiation in Baja California birds: implications for community assembly and conservation. Ecology, Evolution, and Behavior. University of Minnesota, St. Paul, Minnesota, USA. Vellend, M., 2010. Conceptual synthesis in community ecology. Q. Rev. Biol. 85, 183-206. Villeger, S., Mason, N.W.H., Mouillot, D., 2008. New multidimensional functional diversity indices for a multifaceted framework in functional ecology. Ecology 89, 2290-2301. Walters, E.L., Miller, E.H., Lowther, P.E., 2002. Red-naped Sapsucker (Sphyrapicus nuchalis). In: Poole, A. (Ed.), The Birds of North America Online. Cornell Lab of Ornithology, Ithaca, NY. Watts, W.A., Stuiver, M., 1980. Late Wisconsin climate of northern Florida and the origin of species-rich deciduous forest. Science 210, 325-327. Webb, D.M., Moore, W.S., 2005. A phylogenetic analysis of woodpeckers and their allies using 12S, Cyt b, and COI nucleotide sequences (class Aves; order Piciformes). Mol. Phylogen. Evol. 36, 233-248. Weibel, A.C., Moore, W.S., 2002a. Molecular phylogeny of a cosmopolitan group of woodpeckers (genus Picoides) based on COI and cyt b mitochondrial gene sequences. Mol. Phylogen. Evol. 22, 65-75. Weibel, A.C., Moore, W.S., 2002b. A test of a mitochondrial gene-based phylogeny of woodpeckers (Genus Picoides) using an independent nuclear gene, β-fibrinogen intron 7. Mol. Phylogen. Evol. 22, 247-257. 171

Weibel, A.C., Moore, W.S., 2005. Plumage convergence in Picoides woodpeckers based on a molecular phylogeny, with emphasis on convergence in downy and hairy woodpeckers. Condor 107, 797-809. Weiher, E., Freund, D., Bunton, T., Stefanski, A., Lee, T., Bentivenga, S., 2011. Advances, challenges and a developing synthesis of ecological community assembly theory. Philosophical Transactions of the Royal Society B 366, 2403-2413. Weiher, E., Keddy, P. (Eds.), 1999. Ecological assembly rules: Perspectives, advances, retreats. Cambridge University Press, Cambridge, United Kingdom. Weir, J.T., Mursleen, S., 2013. Diversity-dependent cladogenesis and trait evolution in the adaptive radiation of the auks (Aves: Alcidae). Evolution 67, 403-416. Weisburg, W.G., Giovannoni, S.J., Woese, C.R., 1989. The Deinococcus- Thermus phylum and the effect of rRNA composition on phylogenetic tree construction. Syst. Appl. Microbiol. 11, 128-134. Wetmore, A., 1931. Record of an unknown woodpecker from the Lower Pliocene. Condor 33, 255-256. Wiens, J.A., 1989. Spatial scaling in ecology. Funct. Ecol. 3, 385-397. Wiens, J.J., 2006. Missing data and the design of phylogenetic analyses. J. Biomed. Inf. 39, 34-42. Wiens, J.J., Donoghue, M.J., 2004. Historical biogeography, ecology and species richness. Trends Ecol. Evol. 19, 639-644. Wiens, J.J., Moen, D.S., 2008. Missing data and the accuracy of Bayesian phylogenetics. J. Syst. Evol. 46, 307-314. Wiens, J.J., Morrill, M.C., 2011. Missing data in phylogenetic analysis: reconciling results from simulations and empirical data. Syst. Biol. 60, 719-731. Wilkinson, M., 1996. Majority-rule reduced consensus trees and their use in bootstrapping. Mol. Biol. Evol. 13, 437-444. Williams, J.B., Batzli, G.O., 1979. Competition among bark-foraging birds in central Illinois: Experimental evidence. Condor 81, 122-132. Williams, J.W., 2002. Variations in tree cover in North America since the last glacial maximum. Global Planet. Change 35, 1-23. Williams, J.W., Jackson, S.T., 2007. Novel climates, no-analog communities, and ecological surprises. Front. Ecol. Environ. 5, 475-482. Winkler, H., Christie, D.A., 2002. Family Picidae (woodpeckers). In: del Hoyo, J., Elliott, A., Sargatal, J. (Eds.), Handbook of the Birds of the World, Vol. 7: Jacamars to Woodpeckers. Lynx Edicions, Barcelona, Spain, pp. 296-555. Winkler, H., Christie, D.A., Nurney, D., 1995. Woodpeckers: a guide to the woodpeckers of the world. Houghton Mifflin, Boston, Massachusetts, USA. Winkler, H., Gamauf, A., Nittinger, F., Haring, E., 2014. Relationships of the Old World woodpeckers (Aves: Picidae): new insights and taxonomic implications. Ann. Naturhist. Mus. Wien (B Bot. Zool.) 116, 69-86. 172

Winkler, H., Kotaka, N., Gamauf, A., Nittinger, F., Haring, E., 2005. On the phylogenetic position of the Okinawa woodpecker (Sapheopipo noguchii). J. Ornithol. 146, 103-110. Wright, S., 1931. Evolution in Mendelian populations. Genetics 16: 97:159. Wright, S.J., 2002. Plant diversity in tropical forests: A review of mechanisms of species coexistence. Oecologia 130, 1-14. Yoder, J.B., Clancey, E., Des Roches, S., Eastman, J.M., Gentry, L., Godsoe, W., Hagey, T.J., Jochimsen, D., Oswald, B.P., Robertson, J., Sarver, B.A.J., Schenk, J.J., Spear, S.F., Harmon, L.J., 2010. Ecological opportunity and the origin of adaptive radiations. Journal of Evolutionary Biology 23, 1581-1596.

173 Appendix 1 Specimens measured by sex for each taxon for comparative analyses. Specimens of unknown sex were used only where at least one individual of each sex was not available. The "Picinae" column indicates whether the taxon was included in analyses of Picinae. The "All data" column indicates whether taxa had sufﬁcient morphological and phylogenetic data to be included in analyses of morphological evolution.

Taxon Individuals measured by sex Picinae All data Female Male Unknown Blythipicus rubiginosus 2 0 0 Y Y Campephilus guatemalensis 3 1 1 Y Y Campephilus leucopogon 1 2 0 Y Y Campephilus magellanicus 2 2 0 Y Y Campephilus melanoleucos 3 1 0 Y Y Campephilus principalis 0 3 2 Y Y Campephilus robustus 2 2 0 Y Y Campephilus rubricollis 2 5 1 Y Y Campethera abingoni 3 1 2 Y N Campethera bennettii 0 2 0 Y N Campethera cailliautii 3 2 0 Y Y Campethera caroli 2 4 0 Y Y Campethera maculosa 1 0 0 Y N Campethera nivosa 7 5 0 Y Y Campethera nubica 2 3 0 Y Y Campethera punctuligera 0 1 0 Y N Campethera tullbergi 1 0 0 Y N Celeus castaneus 4 4 1 Y Y Celeus elegans 1 1 1 Y Y Celeus flavescens 2 2 0 Y Y Celeus flavus 1 4 0 Y Y Celeus grammicus 1 2 0 Y Y Celeus loricatus 1 2 0 Y Y Celeus lugubris 2 3 0 Y Y Celeus torquatus 1 3 0 Y Y Celeus undatus 4 1 0 Y Y Chrysocolaptes guttacristatus 1 1 0 Y Y Chrysocolaptes haematribon 1 4 1 Y Y Chrysocolaptes lucidus 0 3 0 Y Y Chrysocolaptes validus 0 1 0 Y Y Chrysophlegma flavinucha 3 2 0 Y Y Chrysophlegma mentale 0 1 1 Y Y Colaptes auratus 18 21 1 Y Y Colaptes auricularis 0 1 0 Y Y Colaptes cafer 10 13 1 Y Y Colaptes campestris 1 1 0 Y Y Colaptes canipileus 1 0 0 Y Y Colaptes chrysoides 2 1 0 Y Y Colaptes melanochloros 1 1 0 Y Y Colaptes pitius 2 1 0 Y Y Colaptes punctigula 4 2 0 Y Y Colaptes rivolii 1 2 0 Y Y Colaptes rupicola 0 1 0 Y Y

174 Taxon Individuals measured by sex Picinae All data Female Male Unknown Colaptes yucatanensis 1 1 0 Y Y Dendrocopos analis 0 2 0 Y Y Dendrocopos atratus 1 1 0 Y Y Dendrocopos himalayensis 1 1 0 Y Y Dendrocopos hyperythrus 1 0 1 Y Y Dendrocopos leucotos 1 4 1 Y Y Dendrocopos major 5 5 0 Y Y Dendrocopos syriacus 2 3 0 Y Y Dendropicos elliotii 1 1 0 Y Y Dendropicos fuscescens 7 5 0 Y Y Dendropicos gabonensis 1 1 0 Y N Dendropicos goertae 1 1 0 Y Y Dendropicos griseocephalus 1 3 0 Y Y Dendropicos namaquus 1 0 0 Y Y Dendropicos xantholophus 1 1 0 Y N Dinopium benghalense 5 1 0 Y Y Dinopium everetti 4 3 2 Y N Dinopium javanense 0 4 0 Y Y Dryobates minor 1 2 0 Y Y Dryobates nuttallii 4 2 0 Y Y Dryobates pubescens 28 30 1 Y Y Dryobates scalaris 1 1 0 Y Y Dryocopus fulvus 1 1 0 Y Y Dryocopus funebris 3 2 0 Y Y Dryocopus galeatus 1 0 0 Y Y Dryocopus javensis 2 4 0 Y Y Dryocopus lineatus 4 4 0 Y Y Dryocopus martius 4 7 0 Y Y Dryocopus pileatus 18 22 1 Y Y Dryocopus pulverulentus 3 0 0 Y Y Geocolaptes olivaceus 2 1 0 Y Y Jynx ruficollis 3 0 1 N Y Jynx torquilla 6 5 0 N Y Leiopicus medius 0 4 0 Y Y Leuconotopicus albolarvatus 3 6 0 Y Y Leuconotopicus arizonae 2 5 1 Y Y Leuconotopicus borealis 5 6 0 Y Y Leuconotopicus fumigatus 2 1 0 Y Y Leuconotopicus stricklandi 2 3 0 Y N Leuconotopicus villosus 32 29 2 Y Y Meiglyptes tristis 0 1 0 Y Y Meiglyptes tukki 3 3 0 Y Y Melanerpes aurifrons 2 1 0 Y Y Melanerpes cactorum 2 1 0 Y Y Melanerpes candidus 1 1 0 Y Y Melanerpes carolinus 17 21 2 Y Y Melanerpes chrysogenys 3 2 0 Y Y Melanerpes cruentatus 2 1 0 Y Y Melanerpes erythrocephalus 18 21 4 Y Y Melanerpes flavifrons 1 1 0 Y Y Melanerpes formicivorus 5 4 1 Y Y 175 Taxon Individuals measured by sex Picinae All data Female Male Unknown Melanerpes hoffmannii 1 1 0 Y Y Melanerpes hypopolius 0 2 0 Y Y Melanerpes lewis 2 3 0 Y Y Melanerpes portoricensis 2 3 0 Y Y Melanerpes pucherani 1 1 1 Y Y Melanerpes radiolatus 6 4 0 Y N Melanerpes rubricapillus 5 5 0 Y N Melanerpes santacruzi 3 5 0 Y Y Melanerpes striatus 5 6 0 Y Y Melanerpes superciliaris 3 4 0 Y Y Melanerpes uropygialis 12 12 1 Y Y Nesoctites micromegas 0 1 1 N Y Picoides arcticus 1 3 0 Y Y Picoides canicapillus 1 3 0 Y Y Picoides dorsalis 6 7 0 Y Y Picoides kizuki 1 1 0 Y Y Picoides maculatus 4 3 0 Y Y Picoides tridactylus 4 5 0 Y Y Piculus chrysochloros 0 1 0 Y Y Piculus flavigula 2 1 0 Y Y Piculus litae 0 1 0 Y N Piculus simplex 1 1 0 Y Y Picumnus aurifrons 1 0 0 N Y Picumnus cinnamomeus 0 1 0 N N Picumnus cirratus 1 1 0 N Y Picumnus exilis 1 1 0 N Y Picumnus innominatus 1 1 0 N Y Picumnus minutissimus 1 1 1 N N Picumnus olivaceus 1 1 0 N N Picumnus rufiventris 5 3 0 N Y Picumnus spilogaster 1 1 0 N Y Picumnus subtilis 0 1 0 N N Picumnus temminckii 1 2 0 N Y Picus awokera 1 0 0 Y Y Picus canus 6 6 0 Y Y Picus chlorolophus 1 1 0 Y Y Picus erythropygius 1 1 0 Y Y Picus sharpei 1 1 0 Y Y Picus squamatus 1 0 0 Y Y Picus viridis 6 1 0 Y Y Picus vittatus 1 2 1 Y Y Sasia abnormis 2 0 0 N Y Sasia ochracea 1 0 0 N Y Sphyrapicus nuchalis 3 3 0 Y Y Sphyrapicus ruber 10 10 4 Y Y Sphyrapicus thyroideus 1 6 0 Y Y Sphyrapicus varius 4 5 0 Y Y Veniliornis affinis 1 2 0 Y Y Veniliornis cassini 2 5 0 Y Y Veniliornis lignarius 1 1 0 Y Y Veniliornis mixtus 1 1 0 Y Y 176 Taxon Individuals measured by sex Picinae All data Female Male Unknown Veniliornis nigriceps 0 2 0 Y Y Veniliornis passerinus 3 4 0 Y Y Veniliornis sanguineus 4 0 0 Y Y Veniliornis spilogaster 1 1 1 Y Y Verreauxia africana 0 1 0 N Y Xiphidiopicus percussus 3 4 2 Y N

177 Appendix 2 Results from PGLS regression fits to subclade variables for all models tested. Values shown are the median from models fit to 100 random selections of reciprocally monophyletic subclades. Pseudo-R^2 was calculated as the correlation between observed values and fitted values from the models.

Taxon Set Variable 1 Variable 2 pseudo-R^2 p-value sign of slope Picidae Total diversification Crown age 0.30 0.05 + Diversification rate (Magallon & Sanderson, ε=0.1) Rate of size evolution (BM) 0.04 0.52 - Diversification rate (ML constant rate) Rate of size evolution (BM) 0.02 0.73 + Diversification rate (BAMM) Rate of size evolution (BM) 0.01 0.61 + Diversification rate (Magallon & Sanderson, ε=0.1) Rate of size evolution (BAMM) 0.31 0.05 - Diversification rate (ML constant rate) Rate of size evolution (BAMM) 0.53 0.01 - Diversification rate (BAMM) Rate of size evolution (BAMM) 0.56 0.10 + Diversification rate (Magallon & Sanderson, ε=0.1) Rate of overall morphological evolution (BM) 0.05 0.46 - Diversification rate (ML constant rate) Rate of overall morphological evolution (BM) 0.03 0.71 + Diversification rate (BAMM) Rate of overall morphological evolution (BM) 0.03 0.62 - Diversification rate (Magallon & Sanderson, ε=0.1) Rate of overall morphological evolution (BAMM) 0.22 0.04 - Diversification rate (ML constant rate) Rate of overall morphological evolution (BAMM) 0.42 0.03 - Diversification rate (BAMM) Rate of overall morphological evolution (BAMM) 0.52 0.19 + Diversification rate (Magallon & Sanderson, ε=0.1) Rate of shape evolution (BM) 0.17 0.13 + Diversification rate (ML constant rate) Rate of shape evolution (BM) 0.26 0.06 + Diversification rate (BAMM) Rate of shape evolution (BM) 0.04 0.11 - Diversification rate (Magallon & Sanderson, ε=0.1) Rate of shape evolution (BAMM) 0.03 0.61 + Diversification rate (ML constant rate) Rate of shape evolution (BAMM) 0.03 0.64 + Diversification rate (BAMM) Rate of shape evolution (BAMM) 0.19 0.24 - Diversification rate (Magallon & Sanderson, ε=0.1) Overlap 0.12 0.14 + Diversification rate (ML constant rate) Overlap 0.03 0.12 + Diversification rate (BAMM) Overlap 0.28 0.07 - Diversification rate (Magallon & Sanderson, ε=0.1) Overlap scaled by size 0.25 0.07 + Diversification rate (ML constant rate) Overlap scaled by size 0.23 0.06 + Diversification rate (BAMM) Overlap scaled by size 0.04 0.09 - Diversification rate (Magallon & Sanderson, ε=0.1) Overlap scaled by overall morphology 0.28 0.07 + Diversification rate (ML constant rate) Overlap scaled by overall morphology 0.30 0.06 + Diversification rate (BAMM) Overlap scaled by overall morphology 0.03 0.20 - Diversification rate (Magallon & Sanderson, ε=0.1) Overlap scaled by shape 0.24 0.10 +

178 Taxon Set Variable 1 Variable 2 pseudo-R^2 p-value sign of slope Diversification rate (ML constant rate) Overlap scaled by shape 0.25 0.10 + Diversification rate (BAMM) Overlap scaled by shape 0.01 0.13 - Rate of size evolution (BM) Overlap 0.02 0.07 - Rate of size evolution (BAMM) Overlap 0.05 0.05 - Rate of overall morphological evolution (BM) Overlap 0.02 0.14 - Rate of overall morphological evolution (BAMM) Overlap 0.13 0.11 - Rate of shape evolution (BM) Overlap 0.50 0.01 + Rate of shape evolution (BAMM) Overlap 0.43 0.16 + Rate of size evolution (BM) Overlap scaled by size 0.04 0.04 - Rate of size evolution (BAMM) Overlap scaled by size 0.04 0.06 - Rate of overall morphological evolution (BM) Overlap scaled by overall morphology 0.02 0.10 - Rate of overall morphological evolution (BAMM) Overlap scaled by overall morphology 0.05 0.08 - Rate of shape evolution (BM) Overlap scaled by shape 0.59 0.00 + Rate of shape evolution (BAMM) Overlap scaled by shape 0.33 0.20 + ΔAICc for time-dependent diversification model Overlap 0.14 0.20 - Rate change for time-dependent diversification model Overlap 0.02 0.62 - ΔAICc for diversity-dependent diversification model Overlap 0.11 0.22 - Gamma Overlap 0.24 0.18 + ΔAICc for time-dependent diversification model Overlap scaled by size 0.19 0.13 - Rate change for time-dependent diversification model Overlap scaled by size 0.02 0.61 + ΔAICc for diversity-dependent diversification model Overlap scaled by size 0.21 0.11 - Gamma Overlap scaled by size 0.06 0.48 + ΔAICc for time-dependent diversification model Overlap scaled by overall morphology 0.27 0.10 - Rate change for time-dependent diversification model Overlap scaled by overall morphology 0.02 0.58 + ΔAICc for diversity-dependent diversification model Overlap scaled by overall morphology 0.26 0.07 - Gamma Overlap scaled by overall morphology 0.01 0.66 + ΔAICc for time-dependent diversification model Overlap scaled by shape 0.19 0.15 - Rate change for time-dependent diversification model Overlap scaled by shape 0.03 0.56 + ΔAICc for diversity-dependent diversification model Overlap scaled by shape 0.17 0.15 - Gamma Overlap scaled by shape 0.05 0.46 + ΔAICc for trend model of size evolution Overlap 0.11 0.22 - Slope for trend model of size evolution Overlap 0.04 0.56 + ΔAICc for EB model of size evolution Overlap 0.11 0.27 - α for EB model of size evolution Overlap 0.01 0.10 + 179 Taxon Set Variable 1 Variable 2 pseudo-R^2 p-value sign of slope ΔAICc for trend model of size evolution Overlap scaled by size 0.13 0.22 - Slope for trend model of size evolution Overlap scaled by size 0.06 0.49 - ΔAICc for EB model of size evolution Overlap scaled by size 0.12 0.22 - α for EB model of size evolution Overlap scaled by size 0.01 0.12 + ΔAICc for trend model of overall morphological evolution Overlap 0.12 0.21 - Slope for trend model of overall morphological evolution Overlap 0.04 0.52 + ΔAICc for EB model of overall morphological evolution Overlap 0.11 0.27 - α for EB model of overall morphological evolution Overlap 0.01 0.10 + ΔAICc for trend model of overall morphological evolution Overlap scaled by overall morphology 0.12 0.23 - Slope for trend model of overall morphological evolution Overlap scaled by overall morphology 0.05 0.41 - ΔAICc for EB model of overall morphological evolution Overlap scaled by overall morphology 0.13 0.23 - α for EB model of overall morphological evolution Overlap scaled by overall morphology 0.01 0.26 + ΔAICc for trend model of shape evolution Overlap 0.14 0.09 - Slope for trend model of shape evolution Overlap 0.03 0.11 + ΔAICc for EB model of shape evolution Overlap 0.13 0.21 - α for EB model of shape evolution Overlap 0.03 0.20 + ΔAICc for trend model of shape evolution Overlap scaled by shape 0.30 0.06 - Slope for trend model of shape evolution Overlap scaled by shape 0.06 0.09 + ΔAICc for EB model of shape evolution Overlap scaled by shape 0.17 0.16 - α for EB model of shape evolution Overlap scaled by shape 0.04 0.26 + Diversification rate (ML diversity-dependent) Rate of size evolution (BM) 0.06 0.44 - Diversification rate (ML diversity-dependent) Rate of size evolution (BAMM) 0.11 0.25 - Diversification rate (ML diversity-dependent) Rate of overall morphological evolution (BM) 0.05 0.48 - Diversification rate (ML diversity-dependent) Rate of overall morphological evolution (BAMM) 0.11 0.22 - Diversification rate (ML diversity-dependent) Rate of shape evolution (BM) 0.02 0.54 + Diversification rate (ML diversity-dependent) Rate of shape evolution (BAMM) 0.02 0.62 - Diversification rate (ML diversity-dependent) Overlap 0.02 0.62 + Diversification rate (ML diversity-dependent) Overlap scaled by size 0.08 0.31 + Diversification rate (ML diversity-dependent) Overlap scaled by overall morphology 0.13 0.20 + Diversification rate (ML diversity-dependent) Overlap scaled by shape 0.06 0.42 + Diversification rate (BAMM) Diversification rate (ML constant rate) 0.35 0.00 - Diversification rate (BAMM) Diversification rate (ML time-dependent) 0.07 0.28 - Diversification rate (BAMM) Diversification rate (ML diversity-dependent) 0.04 0.10 - Diversification rate (BAMM) Diversification rate (Magallon & Sanderson, ε=0.1) 0.12 0.00 - 180 Taxon Set Variable 1 Variable 2 pseudo-R^2 p-value sign of slope Rate of size evolution (BAMM) Rate of size evolution (BM) 0.13 0.06 + Rate of overall morphological evolution (BAMM) Rate of overall morphological evolution (BM) 0.23 0.13 + Rate of shape evolution (BAMM) Rate of shape evolution (BM) 0.51 0.05 + Rate change for time-dependent diversification model Overlap 0.02 0.62 - Rate change for time-dependent diversification model Overlap scaled by size 0.02 0.61 + Rate change for time-dependent diversification model Overlap scaled by overall morphology 0.02 0.58 + Rate change for time-dependent diversification model Overlap scaled by shape 0.03 0.56 + Diversification rate (BAMM with constant rates) Diversification rate (ML constant rate) 0.44 0.15 + Diversification rate (BAMM with constant rates) Diversification rate (ML time-dependent) 0.14 0.36 + Diversification rate (BAMM with constant rates) Diversification rate (ML diversity-dependent) 0.07 0.37 + Diversification rate (BAMM with constant rates) Diversification rate (Magallon & Sanderson, ε=0.1) 0.12 0.20 + Diversification rate (BAMM with constant rates) Rate of size evolution (BM) 0.03 0.41 - Diversification rate (BAMM with constant rates) Rate of size evolution (BAMM) 0.79 0.04 - Diversification rate (BAMM with constant rates) Rate of overall morphological evolution (BM) 0.04 0.39 - Diversification rate (BAMM with constant rates) Rate of overall morphological evolution (BAMM) 0.73 0.02 - Diversification rate (BAMM with constant rates) Rate of shape evolution (BM) 0.00 0.47 - Diversification rate (BAMM with constant rates) Rate of shape evolution (BAMM) 0.12 0.39 - Diversification rate (BAMM with constant rates) Overlap 0.22 0.32 - Diversification rate (BAMM with constant rates) Overlap scaled by size 0.00 0.31 + Diversification rate (BAMM with constant rates) Overlap scaled by overall morphology 0.07 0.28 + Diversification rate (BAMM with constant rates) Overlap scaled by shape 0.01 0.14 + Diversification rate (ML diversity-dependent) Rate of size evolution (BM) 0.06 0.44 - Diversification rate (ML diversity-dependent) Rate of size evolution (BAMM) 0.11 0.25 - Diversification rate (ML diversity-dependent) Rate of overall morphological evolution (BM) 0.05 0.48 - Diversification rate (ML diversity-dependent) Rate of overall morphological evolution (BAMM) 0.11 0.22 - Diversification rate (ML diversity-dependent) Rate of shape evolution (BM) 0.02 0.54 + Diversification rate (ML diversity-dependent) Rate of shape evolution (BAMM) 0.02 0.62 - Diversification rate (ML diversity-dependent) Overlap 0.02 0.62 + Diversification rate (ML diversity-dependent) Overlap scaled by size 0.08 0.31 + Diversification rate (ML diversity-dependent) Overlap scaled by overall morphology 0.13 0.20 + Diversification rate (ML diversity-dependent) Overlap scaled by shape 0.06 0.42 + Diversification rate (ML constant rate) Diversification rate (ML time-dependent) 0.13 0.33 + Diversification rate (ML constant rate) Diversification rate (ML diversity-dependent) 0.38 0.03 + Diversification rate (ML constant rate) Diversification rate (Magallon & Sanderson, ε=0.1) 0.81 0.00 + 181 Taxon Set Variable 1 Variable 2 pseudo-R^2 p-value sign of slope Diversification rate (ML constant rate) Diversification rate (Magallon & Sanderson, ε=0.5) 0.81 0.00 + Diversification rate (ML constant rate) Diversification rate (Magallon & Sanderson, ε=0.9) 0.68 0.00 +

Picinae Total diversification Crown age 0.24 0.11 + Diversification rate (Magallon & Sanderson, ε=0.1) Rate of size evolution (BM) 0.06 0.47 + Diversification rate (ML constant rate) Rate of size evolution (BM) 0.02 0.45 + Diversification rate (BAMM) Rate of size evolution (BM) 0.04 0.57 + Diversification rate (Magallon & Sanderson, ε=0.1) Rate of size evolution (BAMM) 0.15 0.13 - Diversification rate (ML constant rate) Rate of size evolution (BAMM) 0.15 0.14 - Diversification rate (BAMM) Rate of size evolution (BAMM) 0.27 0.01 + Diversification rate (Magallon & Sanderson, ε=0.1) Rate of overall morphological evolution (BM) 0.05 0.48 + Diversification rate (ML constant rate) Rate of overall morphological evolution (BM) 0.04 0.34 + Diversification rate (BAMM) Rate of overall morphological evolution (BM) 0.04 0.61 + Diversification rate (Magallon & Sanderson, ε=0.1) Rate of overall morphological evolution (BAMM) 0.04 0.49 - Diversification rate (ML constant rate) Rate of overall morphological evolution (BAMM) 0.03 0.50 + Diversification rate (BAMM) Rate of overall morphological evolution (BAMM) 0.05 0.14 + Diversification rate (Magallon & Sanderson, ε=0.1) Rate of shape evolution (BM) 0.18 0.14 + Diversification rate (ML constant rate) Rate of shape evolution (BM) 0.36 0.04 + Diversification rate (BAMM) Rate of shape evolution (BM) 0.13 0.21 - Diversification rate (Magallon & Sanderson, ε=0.1) Rate of shape evolution (BAMM) 0.09 0.25 + Diversification rate (ML constant rate) Rate of shape evolution (BAMM) 0.20 0.08 + Diversification rate (BAMM) Rate of shape evolution (BAMM) 0.07 0.52 - Diversification rate (Magallon & Sanderson, ε=0.1) Overlap 0.25 0.06 + Diversification rate (ML constant rate) Overlap 0.33 0.04 + Diversification rate (BAMM) Overlap 0.20 0.11 - Diversification rate (Magallon & Sanderson, ε=0.1) Overlap scaled by size 0.28 0.05 + Diversification rate (ML constant rate) Overlap scaled by size 0.37 0.03 + Diversification rate (BAMM) Overlap scaled by size 0.20 0.13 - Diversification rate (Magallon & Sanderson, ε=0.1) Overlap scaled by overall morphology 0.30 0.05 + Diversification rate (ML constant rate) Overlap scaled by overall morphology 0.35 0.03 + Diversification rate (BAMM) Overlap scaled by overall morphology 0.19 0.14 - Diversification rate (Magallon & Sanderson, ε=0.1) Overlap scaled by shape 0.22 0.07 + Diversification rate (ML constant rate) Overlap scaled by shape 0.32 0.04 + Diversification rate (BAMM) Overlap scaled by shape 0.18 0.16 - 182 Taxon Set Variable 1 Variable 2 pseudo-R^2 p-value sign of slope Rate of size evolution (BM) Overlap 0.04 0.15 - Rate of size evolution (BAMM) Overlap 0.03 0.05 - Rate of overall morphological evolution (BM) Overlap 0.02 0.24 - Rate of overall morphological evolution (BAMM) Overlap 0.11 0.50 - Rate of shape evolution (BM) Overlap 0.64 0.00 + Rate of shape evolution (BAMM) Overlap 0.60 0.01 + Rate of size evolution (BM) Overlap scaled by size 0.04 0.13 - Rate of size evolution (BAMM) Overlap scaled by size 0.03 0.05 - Rate of overall morphological evolution (BM) Overlap scaled by overall morphology 0.02 0.10 - Rate of overall morphological evolution (BAMM) Overlap scaled by overall morphology 0.07 0.65 - Rate of shape evolution (BM) Overlap scaled by shape 0.66 0.00 + Rate of shape evolution (BAMM) Overlap scaled by shape 0.60 0.01 + ΔAICc for time-dependent diversification model Overlap 0.35 0.03 - Rate change for time-dependent diversification model Overlap 0.03 0.54 - ΔAICc for diversity-dependent diversification model Overlap 0.30 0.05 - Gamma Overlap 0.13 0.27 + ΔAICc for time-dependent diversification model Overlap scaled by size 0.35 0.03 - Rate change for time-dependent diversification model Overlap scaled by size 0.02 0.63 + ΔAICc for diversity-dependent diversification model Overlap scaled by size 0.33 0.02 - Gamma Overlap scaled by size 0.08 0.33 + ΔAICc for time-dependent diversification model Overlap scaled by overall morphology 0.33 0.03 - Rate change for time-dependent diversification model Overlap scaled by overall morphology 0.02 0.53 + ΔAICc for diversity-dependent diversification model Overlap scaled by overall morphology 0.31 0.03 - Gamma Overlap scaled by overall morphology 0.06 0.43 + ΔAICc for time-dependent diversification model Overlap scaled by shape 0.32 0.05 - Rate change for time-dependent diversification model Overlap scaled by shape 0.02 0.56 - ΔAICc for diversity-dependent diversification model Overlap scaled by shape 0.28 0.04 - Gamma Overlap scaled by shape 0.12 0.26 + ΔAICc for trend model of size evolution Overlap 0.28 0.05 - Slope for trend model of size evolution Overlap 0.05 0.49 - ΔAICc for EB model of size evolution Overlap 0.31 0.05 - α for EB model of size evolution Overlap 0.03 0.32 + ΔAICc for trend model of size evolution Overlap scaled by size 0.23 0.08 - Slope for trend model of size evolution Overlap scaled by size 0.05 0.51 - 183 Taxon Set Variable 1 Variable 2 pseudo-R^2 p-value sign of slope ΔAICc for EB model of size evolution Overlap scaled by size 0.23 0.09 - α for EB model of size evolution Overlap scaled by size 0.01 0.33 + ΔAICc for trend model of overall morphological evolution Overlap 0.28 0.04 - Slope for trend model of overall morphological evolution Overlap 0.05 0.48 - ΔAICc for EB model of overall morphological evolution Overlap 0.31 0.05 - α for EB model of overall morphological evolution Overlap 0.03 0.30 + ΔAICc for trend model of overall morphological evolution Overlap scaled by overall morphology 0.18 0.13 - Slope for trend model of overall morphological evolution Overlap scaled by overall morphology 0.04 0.51 + ΔAICc for EB model of overall morphological evolution Overlap scaled by overall morphology 0.17 0.15 - α for EB model of overall morphological evolution Overlap scaled by overall morphology 0.01 0.38 + ΔAICc for trend model of shape evolution Overlap 0.37 0.01 - Slope for trend model of shape evolution Overlap 0.04 0.52 + ΔAICc for EB model of shape evolution Overlap 0.34 0.02 - α for EB model of shape evolution Overlap 0.06 0.27 + ΔAICc for trend model of shape evolution Overlap scaled by shape 0.35 0.01 - Slope for trend model of shape evolution Overlap scaled by shape 0.04 0.52 + ΔAICc for EB model of shape evolution Overlap scaled by shape 0.31 0.02 - α for EB model of shape evolution Overlap scaled by shape 0.06 0.21 + Diversification rate (ML diversity-dependent) Rate of size evolution (BM) 0.08 0.38 - Diversification rate (ML diversity-dependent) Rate of size evolution (BAMM) 0.05 0.35 - Diversification rate (ML diversity-dependent) Rate of overall morphological evolution (BM) 0.06 0.44 - Diversification rate (ML diversity-dependent) Rate of overall morphological evolution (BAMM) 0.04 0.48 - Diversification rate (ML diversity-dependent) Rate of shape evolution (BM) 0.02 0.56 + Diversification rate (ML diversity-dependent) Rate of shape evolution (BAMM) 0.01 0.72 + Diversification rate (ML diversity-dependent) Overlap 0.07 0.44 + Diversification rate (ML diversity-dependent) Overlap scaled by size 0.09 0.31 + Diversification rate (ML diversity-dependent) Overlap scaled by overall morphology 0.13 0.19 + Diversification rate (ML diversity-dependent) Overlap scaled by shape 0.07 0.45 + Diversification rate (BAMM) Diversification rate (ML constant rate) 0.72 0.00 - Diversification rate (BAMM) Diversification rate (ML time-dependent) 0.24 0.11 - Diversification rate (BAMM) Diversification rate (ML diversity-dependent) 0.28 0.08 - Diversification rate (BAMM) Diversification rate (Magallon & Sanderson, ε=0.1) 0.80 0.00 - Rate of size evolution (BAMM) Rate of size evolution (BM) 0.41 0.01 + Rate of overall morphological evolution (BAMM) Rate of overall morphological evolution (BM) 0.52 0.05 + 184 Taxon Set Variable 1 Variable 2 pseudo-R^2 p-value sign of slope Rate of shape evolution (BAMM) Rate of shape evolution (BM) 0.94 0.00 + Rate change for time-dependent diversification model Overlap 0.03 0.54 - Rate change for time-dependent diversification model Overlap scaled by size 0.02 0.63 + Rate change for time-dependent diversification model Overlap scaled by overall morphology 0.02 0.53 + Rate change for time-dependent diversification model Overlap scaled by shape 0.02 0.56 - Diversification rate (BAMM with constant rates) Diversification rate (ML constant rate) 0.01 0.47 + Diversification rate (BAMM with constant rates) Diversification rate (ML time-dependent) 0.01 0.74 + Diversification rate (BAMM with constant rates) Diversification rate (ML diversity-dependent) 0.18 0.15 + Diversification rate (BAMM with constant rates) Diversification rate (Magallon & Sanderson, ε=0.1) 0.02 0.37 + Diversification rate (BAMM with constant rates) Rate of size evolution (BM) 0.16 0.33 - Diversification rate (BAMM with constant rates) Rate of size evolution (BAMM) 0.13 0.37 - Diversification rate (BAMM with constant rates) Rate of overall morphological evolution (BM) 0.21 0.19 - Diversification rate (BAMM with constant rates) Rate of overall morphological evolution (BAMM) 0.33 0.06 - Diversification rate (BAMM with constant rates) Rate of shape evolution (BM) 0.08 0.28 - Diversification rate (BAMM with constant rates) Rate of shape evolution (BAMM) 0.21 0.17 - Diversification rate (BAMM with constant rates) Overlap 0.03 0.58 - Diversification rate (BAMM with constant rates) Overlap scaled by size 0.02 0.62 + Diversification rate (BAMM with constant rates) Overlap scaled by overall morphology 0.03 0.68 + Diversification rate (BAMM with constant rates) Overlap scaled by shape 0.03 0.58 + Diversification rate (ML diversity-dependent) Rate of size evolution (BM) 0.08 0.38 - Diversification rate (ML diversity-dependent) Rate of size evolution (BAMM) 0.05 0.35 - Diversification rate (ML diversity-dependent) Rate of overall morphological evolution (BM) 0.06 0.44 - Diversification rate (ML diversity-dependent) Rate of overall morphological evolution (BAMM) 0.04 0.48 - Diversification rate (ML diversity-dependent) Rate of shape evolution (BM) 0.02 0.56 + Diversification rate (ML diversity-dependent) Rate of shape evolution (BAMM) 0.01 0.72 + Diversification rate (ML diversity-dependent) Overlap 0.07 0.44 + Diversification rate (ML diversity-dependent) Overlap scaled by size 0.09 0.31 + Diversification rate (ML diversity-dependent) Overlap scaled by overall morphology 0.13 0.19 + Diversification rate (ML diversity-dependent) Overlap scaled by shape 0.07 0.45 + Diversification rate (ML constant rate) Diversification rate (ML time-dependent) 0.05 0.49 + Diversification rate (ML constant rate) Diversification rate (ML diversity-dependent) 0.38 0.04 + Diversification rate (ML constant rate) Diversification rate (Magallon & Sanderson, ε=0.1) 0.87 0.00 + Diversification rate (ML constant rate) Diversification rate (Magallon & Sanderson, ε=0.5) 0.88 0.00 + Diversification rate (ML constant rate) Diversification rate (Magallon & Sanderson, ε=0.9) 0.82 0.00 + 185 25

Appendix 3 Map of global Picidae species richness. Species richness was calculated by overlaying range maps from BirdLife International using the R package rangeMapper (Valcu et al., 2012). White areas on the map represent regions with no Picidae species.

186