University of Southern Denmark

Peripheral eco-morphology predicts restricted lineage diversification and endemism among corvoid

Kennedy, Jonathan D.; Marki, Petter Z.; Fjeldså, Jon; Rahbek, Carsten

Published in: Global Ecology and Biogeography

DOI: 10.1111/geb.13194

Publication date: 2021

Document version: Final published version

Document license: CC BY

Citation for pulished version (APA): Kennedy, J. D., Marki, P. Z., Fjeldså, J., & Rahbek, C. (2021). Peripheral eco-morphology predicts restricted lineage diversification and endemism among corvoid passerine birds. Global Ecology and Biogeography, 30, 79- 98. https://doi.org/10.1111/geb.13194

Go to publication entry in University of Southern Denmark's Research Portal

Terms of use This work is brought to you by the University of Southern Denmark. Unless otherwise specified it has been shared according to the terms for self-archiving. If no other license is stated, these terms apply:

• You may download this work for personal use only. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying this open access version If you believe that this document breaches copyright please contact us providing details and we will investigate your claim. Please direct all enquiries to [email protected]

Download date: 27. Sep. 2021

Received: 13 January 2020 | Revised: 19 June 2020 | Accepted: 28 July 2020 DOI: 10.1111/geb.13194

RESEARCH PAPER

Peripheral eco-morphology predicts restricted lineage diversification and endemism among corvoid passerine birds

Jonathan D. Kennedy1 | Petter Z. Marki2,3 | Jon Fjeldså2,4 | Carsten Rahbek2,5,6

1Department of and Plant Sciences, University of Sheffield, Sheffield, UK Abstract 2Center for Macroecology, Evolution and Aim: Across a variety of taxonomic scales, species diversity is unevenly distributed Climate, GLOBE Institute, University of among its constituent units, and clades with few species are more common than ex- Copenhagen, Copenhagen, Denmark 3Natural History Museum, University of pected assuming homogeneous rates of speciation and among lineages. In Oslo, Oslo, Norway order to explain the prevalence of species-poor families among a global and species- 4 Natural History Museum of Denmark, rich radiation of passerine birds, we test whether these groups share common eco- University of Copenhagen, Copenhagen, Denmark morphological, geographical and phylogenetic attributes. 5Department of Life Sciences, Imperial Location: Global. College London, Ascot, UK Time period: Late Oligocene to the present day. 6Danish Institute for Advanced Study, University of Southern Denmark, Odense, Major taxa studied: The Corvides (c. 790 species). Denmark Methods: We obtained 10 linear measurements of external morphology for 782 spe-

Correspondence cies of corvoid . Using these measurements as a proxy for species ecology, Jonathan D. Kennedy, Department of Animal we assessed the positioning of corvoid families in eco-morphological trait space and and Plant Sciences, University of Sheffield, Sheffield S10 2TN, UK. how these factors were associated with their species richness and rates of lineage Email: [email protected] diversification. Subsequently, we compared these same characteristics (species rich-

Funding information ness, morphological positioning and rates of lineage diversification) between families European Commission, Grant/Award that are currently endemic to the Australasian ancestral area of the Corvides with Number: Marie Sklodowska-Curie actions (MSCA-792534); Danmarks those that have dispersed and diversified throughout other continental and insular Grundforskningsfond, Grant/Award landmasses. Number: DNRF96; Carlsbergfondet, Grant/ Award Number: CF17-0239 Results: Families with low species richness and rates of diversification tend to oc- cupy the most peripheral positions in eco-morphological trait space, with almost all Editor: Adam Algar of these groups being endemic to Australasia. The peripheral eco-morphological po- sitioning of the Australasian groupings is generally greater than expected upon ac- counting for differences in phylogenetic isolation and heterogeneity in rates of trait evolution, implying that species-poor corvoid families repeatedly evolved towards marginal areas of morphospace. Main conclusions: The over-representation of species-poor clades across diverse sets of organismal groups is consistent with their evolution towards, and the mainte- nance in, marginal areas of ecological niche space. The evolution of peripheral eco- morphological characters represents a potentially significant limit to rates of range expansion and lineage diversification.

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2020 The Authors. Global Ecology and Biogeography published by John Wiley & Sons Ltd

 wileyonlinelibrary.com/journal/geb | 79 Global Ecol. Biogeogr. 2021;30:79–98. 80 | KENNEDY et al.

KEYWORDS diversification, ecology, endemism, morphology, phylogenetic distinctiveness, phylogenetic diversity, range expansion, taxon cycle

1 | INTRODUCTION (Grandcolas & Trewick, 2016). Ricklefs (2005) hypothesized that spe- cies-poor lineages consistently exploit peripheral or rare resources, The numbers of species in clades of the same age and taxonomic and that this directly inhibits opportunities for lineage diversifica- rank vary extensively. At a given taxonomic scale, it is common tion. These conclusions were formed from studying the trends of that a small number of lineages contain the bulk of species di- morphological positioning among a paraphyletic assemblage of pas- versity, whereas the vast majority are species poor (Bennett & serine species, but in the absence of information about the phyloge- Owens, 2002; Dial & Marzluff, 1989; Owens et al., 1999; Scotland netic relatedness among taxa. Given the significant improvement in & Sanderson, 2004). Compared with null expectations about the the and phylogeny of passerine birds in the 15 years since distribution of clade richness assuming homogeneous rates of spe- the initial study by Ricklefs (2005), we re-evaluate this hypothesis ciation and extinction among lineages, it is apparent that groups through the analysis of a comparative data set representing a large with low numbers of species are far more common than predicted monophyletic passerine clade. (Bennett & Owens, 2002; Dial & Marzluff, 1989; Owens et al., 1999; Adaptation to broadly distributed and generalized sets of re- Ricklefs, 2003). In attempting to understand the causes of the di- sources commonly leads to the evolution of life strategies that fa- versification rate heterogeneity that produces variation in species cilitate geographical expansion and, subsequently, the formation richness, phylogenetic comparative studies have predominantly of superspecies complexes in which the constituent species are focused on the factors underlying major upturns in diversification largely allopatric, having undergone limited ecological differentia- rates, and thus the processes that have generated the exception- tion (Kennedy et al., 2018; Mayr, 1947). In contrast, lineages with ally high richness of a relatively small subset of clades (e.g., Alfaro more unusual or specialized ecologies might become limited in their et al., 2009; Phillimore et al., 2006). However, any comprehensive distributional extent because the resources upon which they are explanation for the uneven distribution of species diversity across adapted are patchy or less widely distributed in geographical space the tree of life must also account for the prevalence of species-poor (Ricklefs, 2005). Ecological specialization might also result in the clades and the reasons underlying their exceptionally low rates of evolution of eco-morphological traits that further inhibit the poten- lineage diversification (Bennett & Owens, 2002; Owen et al., 1999; tial for successful dispersal and range expansion. The distributions Ricklefs, 2003, 2005). of taxa that exploit specialized and rare resources are therefore From horseshoe crabs to the duck-billed platypus, taxa that are expected to be restricted to climatically stable and geographically both species poor and phylogenetically isolated represent emblem- isolated areas, which harbour unusual habitats or environments atic examples of lineages with deep and unique evolutionary histo- (Fjeldså et al., 2012; Ricklefs, 2003, 2005). In light of the combined ries. The commonality of lineages that meet these criteria among predictions of these hypotheses, here we test the association be- diverse sets of organismal groups implies that a common mecha- tween eco-morphological attributes, geographical distribution and nism might underlie their formation and maintenance through time lineage diversification.

5 (a) 14 Australasian (b) (c) Widespread 0.30 12 4

0.25 10 chness) 3 8

er My ) 0.20 amily species ri Frequenc y 6 2

Mean DR (P 0.15 Log (F

4

1 0.10

2

0.05 0 0

0102030405060708090 100 110 120 130 2 3456 2 3456

Family species richness Distance from the centroid of PC space Distance from the centroid of PC space

FIGURE 1 (a) Frequency histogram showing the species richness of the 30 corvoid passerine families defined by the temporal banding approach proposed by Jønsson et al. (2016). (b,c) Scatterplots showing the relationship between (b) the log10-transformed family species richness and (c) rates of lineage diversification [mean diversification rate (DR) values], as a factor of the mean distance that each family occurred from the centroid of principal component (PC) space (all PC axes are standardized to have equal variance). Black lines represent the least squares regression fit. Families that are endemic to Australasia are shown in blue, whereas those that have dispersed and diversified on other continental or insular landmasses are shown in red [Colour figure can be viewed at wileyonlinelibrary.com] KENNEDY et al. | 81

In common with many animal and plant taxa (Bennett & downstream analyses on the species scores of the individual princi- Owens, 2002; Dial & Marzluff, 1989; Owen et al., 1999), the distri- pal component (PC) axes. Given that the tail measurements gener- bution of species diversity among families of the global radiation of ally exhibit weak correlations with major aspects of corvoid ecology corvoid passerines (c. 790 species overall) is heavily right skewed, (Kennedy et al., 2019), we also repeated the PCA and subsequent with the majority of families containing < 10 species (Figure 1a). analyses after the exclusion of these measurements. Loadings of

Independent biogeographical analyses provide strong support that the original log10-transformed morphological measurements upon Australasia (Australia, New Zealand and New Guinea) represents the individual PC axes can be found in the Supporting Information the ancestral area of this radiation (Jønsson et al., 2011; Kennedy (Table S2). Given that a significant aim of this study was to assess the et al., 2017; Moyle et al., 2016; Oliveros et al., 2019), and many cor- morphological positioning of clades containing different numbers of void families remain endemic to this region. In this study, initially we species, it is important to consider the influence that species-rich assess whether species-poor families are over-represented among clades (containing many closely related species) have in determin- the Corvides, assuming homogeneous rates of lineage diversifica- ing the relative trait loadings on the different PC axes, in contrast tion among lineages. After establishing this to be the case, using to lineages that are phylogenetically isolated and species poor. We morphology as a proxy for ecology we test whether morphological therefore compared our findings using the original PC scores with positioning predicts variation in species richness and rates of lineage those attempting to correct for the phylogenetic non-independence diversification among corvoid families. We ask whether endemism of the species-level data in their computation (Revell, 2009). We re- to the Australasian ancestral area is associated with taxa that have covered highly consistent results upon analysis of the phylogenetic limited species diversity, depressed diversification and marginal PC (pPC) scores, and therefore, we present the results from analyses eco-morphology. Finally, we determine whether the trait combi- of the original uncorrected PC scores in the main text of the article nations that determine morphological peripherality are consistent and those using pPCs in the Appendix. among clades. Estimates of the phylogenetic relationships among the Corvides were obtained from Kennedy et al. (2017). This phylogeny built upon the analysis of Jønsson et al. (2016) that sampled ≤ 10 mito- 2 | METHODS chondrial and nuclear loci for c. 85% of all corvoid species. The re- maining unsampled species were added as polytomies by Kennedy 2.1 | Morphological and phylogenetic data et al. (2017), with their phylogenetic position determined from tax- onomic information alone. These species were generally placed in Detailed descriptions of the morphological and phylogenetic data terminal positions within well-resolved and species-rich clades of used in this study can be found in the papers by Kennedy et al. (2017, the Corvides (Supporting Information Table S3), with the branch 2018, 2019). Briefly, 10 linear measurements of external morphol- lengths subtending these species generated using the polytomy ogy were quantified from museum specimens to capture the vari- resolver method (Kuhn et al., 2011). The maximum clade credibility ation in different aspects of the bills, tarsi, wings, tails and feet. tree generated from the pseudo-posterior distribution of phyloge- These measurements have been shown to be associated with eco- nies that was analysed in this study can be downloaded from Dryad logical variation among passerine birds (Miles & Ricklefs, 1984; Pigot (https://datad​ryad.org/resou​rce/doi:10.5061/dryad.80n42). For et al., 2016, 2020; Ricklefs, 2005; Ricklefs & Miles, 1994; Ricklefs downstream analyses performed at the species level, congruent re- & Travis, 1980), and specifically within the Corvides (Kennedy sults were recovered both on the complete phylogeny and after the et al., 2019). Measurements of 4,092 specimens were obtained that removal of the taxonomically placed species (Appendix). In the con- together represent 782 species, with a mean of 5.56 ± 1.22 indi- text of the main aims of our study, it is important to note that 95% viduals measured per species. We used the mean trait values of each of the species that are members of clades endemic to Australasia species in our analyses because the vast majority of the variance in were sampled in the phylogeny using molecular data (Supporting the individual trait measurements is between rather than within spe- Information Table S3). cies (Kennedy et al., 2019). The mean morphological measurements We defined corvoid families following the temporal banding of all species can be found in the Supporting Information (Table S1), approach proposed by Jønsson et al. (2016). This method involves whereas the morphological data for all measured individuals can slicing the phylogeny at the point in time (21.6 Ma) that generates a be downloaded from Dryad (https://doi.org/10.5061/dryad.fbg79​ selection of monophyletic clades that are as congruent as possible cnr6). with the original taxonomic units [here, the IOC family classifica- Species of the Corvides range in mass from < 10 to > 1,000 g tions (Gill & Donsker, 2010) used in previous comparative analy- (Kennedy et al., 2012), but the overall size distribution is significantly ses of the Corvides; Kennedy et al., 2017, 2018], while also being right skewed, and the raw morphological measurements demon- temporally consistent in their definition. Species-level member- strate strong positive correlations with one another (Kennedy ship of the respective family units was extremely similar under the et al., 2019). To produce uncorrelated axes of morphological varia- two alternative taxonomic schemes, and consequently, the results tion, we log10-transformed the original morphological measurements of all downstream analyses were in accordance with one another and performed a principal components analysis (PCA), executing our (Appendix). To perform comparative analyses at the family level, we 82 | KENNEDY et al. pruned the phylogeny of Jønsson et al. (2016) to contain a single that was equivalent to the empirical corvoid phylogeny (root representative of each family. We also repeated the family-level age = 30.1 Ma). Subsequently, we sliced the tree at 21.6 Ma (the analyses using a more recent phylogeny generated by Oliveros time slice used to derive the families under the temporal banding et al. (2019). Although the phylogeny of Oliveros et al. (2019) was approach), summed the overall number and richness of the result- produced using a far greater amount of DNA sequence data than ing clades, and compared the frequency distribution of these values that of Jønsson et al. (2016), because some of our analysed metrics with those of the empirical corvoid families. were derived from the latter species-level tree (stem/crown group ages and family diversification rates), we focus on the results derived from this phylogeny in the main text of this article. Results produced 2.3 | Phylogenetic comparative analyses upon analysis of the Oliveros et al. (2019) phylogeny remained fully in agreement with these findings (Appendix). Using phylogenetic least squares regression (PGLS) implemented We estimated current rates of lineage diversification using the in the R package caper (Orme et al., 2012; R Core Team, 2018), we

DR statistic (Jetz et al., 2012), computing these values on the com- tested the correlations between DR/log10-transformed species rich- plete phylogeny of the Corvides (Kennedy et al., 2017). DR provides ness as a factor of distance from the centroid of PC space among an estimate of diversification rate at the species level, by consider- corvoid families. To determine whether species richness, patterns ing the number of nodes separating a species from the root of the of lineage diversification, differences in the stem/crown group ra- phylogeny, weighted by how close those nodes occur to the present. diations and morphological positioning were associated with wide- This statistic has been shown to be correlated strongly with spe- spread/restricted geographical distribution, we also compared these ciation rate estimates in birth–death simulations (Jetz et al., 2012). factors between families of corvoid passerines that have remained We chose DR as our measure of diversification because these values endemic to the Australasian ancestral area (herein, “Australasian” could be analysed at both the family and species levels. For the fam- families; Jønsson et al., 2011; Kennedy et al., 2017; Moyle ilies, we calculated the mean value of DR among their constituent et al., 2016; Oliveros et al., 2019) with those that have dispersed species, in addition to the difference between the stem and crown and diversified on other continental or insular landmasses (herein, group radiations on the species-level tree. “widespread” families; Supporting Information Table S3). These We considered positioning in morphospace as the mean distance analyses were performed using phylogenetic ANOVA (Garland et al., that the members of each family occurred from the centroid of over- 1993), implemented in the R package phytools (Revell, 2012). Given all corvoid PC space. Before calculating the distance to the PC cen- that a single Australasian family has accumulated substantial species troid, we standardized all PC axes to have equal variance. Given that diversity relative to the remaining families in this region [the birds- the majority of the variance of the morphological measurements is of-paradise (BOP), Paradisaeidae; c. 41 species], probably owing to explained by principal component 1 (PC1; a strong proxy for overall significant sexual and ecological selection pressures, we repeated size; Supporting Information Table S2), standardizing the PC axes our analysis with the exclusion of this clade from the grouping of increases the emphasis of morphological positioning on the lower Australasian families. Overall, the 15 Australasian families are repre- PC axes, which have been shown to retain important ecological sented by a combined 78 species (37 species upon the exclusion of information despite accounting for limited variation in the original BOP), in comparison to the 15 widespread families, which together traits (Kennedy et al., 2019; Pigot et al., 2016, 2020; Ricklefs, 2005; total 704 species. Analyses comparing rates of lineage diversifica- Ricklefs & Travis, 1980). tion and distance from the centroid of PC space were also repeated among these family groupings at the species level. For the analyses of morphological positioning among the 2.2 | Variation in species richness among corvoid Australasian and widespread families, we investigated how differ- families assuming equal rates of lineage diversification ences in the distance from the centroid depended on the number of PC axes combined and, consequently, the amount of variance that Previous studies have reported that species-poor clades are more the respective axes explained in the original morphological measure- prevalent than expected assuming equal rates of diversification ments. To do this, we removed the PC axes that explained the least among lineages (Bennett & Owens, 2002; Dial & Marzluff, 1989; variance in the data (e.g., analysing all PC axes, PC9–1, PC8–1 etc.) Owens et al., 1999; Ricklefs, 2003). To assess this possibility among incrementally until we were left with only PC1, before recalculat- the Corvides, we simulated phylogenetic trees to contain a number ing the distance from the centroid and repeating the phylogenetic of tips equal to the overall species diversity of the group (789 extant ANOVAs. species), assuming the same constant rate of diversification through To evaluate the diagnosable morphological characteristics of time among all lineages. Five sets of trees were simulated assuming each family, we performed a linear discriminant analysis (LDA) using a constant rate of speciation and a differing fraction of extinction (0, the original log10-transformed measurements. For families that 0.2, 0.4, 0.6 and 0.8), in the R package TreeSim (Stadler, 2017). We possess marginal morphology, this analysis allowed us to quantify repeated each set of simulations 1,000 times and then scaled the re- the most diagnosable morphological traits for these clades. For the sulting phylogenies to correspond to the time-scale of diversification analysis of each individual family, we placed all of its members into KENNEDY et al. | 83 a single class, with the remaining corvoid species considered as a For each individual BM or VR simulation, we generated a set second class. of standardized PC axes, then calculated the distance that each species, or family, occurred from the centroid of PC space. We then assessed whether the following empirical relationships could 2.4 | Null models of trait evolution be reproduced using the simulated trait data derived from these null models: (a) the trait divergence between the Australasian and We assessed whether divergences in PC values, or a negative cor- widespread families on the individual PC axes; (b) the distance from relation between species richness/DR as a factor of distance from the centroid of PC space, again compared between Australasian the centroid of PC space, could arise primarily owing to differences and widespread families; (c) the correlation between family spe- in the species richness and phylogenetic isolation among families. cies richness and the distance from the centroid of PC space; and Species-poor families might be more likely to occur on the periphery (d) the correlation between diversification rate (DR) and the dis- of PC space in the absence of any direct underlying ecological or tance from the centroid of PC space. Finally, to assess the extent evolutionary processes: (a) because they contribute a limited num- to which the BM and VR models capture the variation in our empir- ber of data points to the PCA (and thus generally account for a small ical morphological data, and thus accurately describe the patterns amount of the explained variance on the PC axes); and (b) as a re- of trait evolution across the individual PC axes, we performed the sult of the long time they have been separated from their closest model adequacy tests proposed by Pennell et al. (2015) in the R ancestors, plus random drift in the evolution of their trait values. package arbutus (Pennell et al., 2014). Therefore, we assessed morphological divergence and trait correla- tions when the data were simulated under two different null mod- els, namely Brownian motion (BM) and Variable rates (VR) (Venditti 3 | RESULTS et al., 2011). Brownian motion models the scenario in which continuous traits 3.1 | Are species-poor families over-represented can evolve through time in any direction, with the expected differ- among the Corvides? ences accrued among taxa being proportional solely to the time since divergence from a common ancestor. Variance in trait values Regardless of the relative rate of extinction, we were unable 2 is determined by the overall evolutionary rate (σ ). For each of the to produce a similar frequency distribution of clades (in terms 2 empirical PC axes, we calculated σ in the R package geiger (Harmon of either the number of clades or their respective species rich- et al., 2008) before simulating sets of traits under BM using the em- ness) in any of our simulated data sets, when compared against 2 pirical σ values on phylogenies transformed by the maximum like- the empirical corvoid families (Supporting Information Figure lihood value of Pagel's λ (Pagel, 1999). We repeated the simulations S2). Consistently across all of our simulations, we tended to re- 1,000 times for each PC axis. cover fewer clades (Supporting Information Figure S2) when slic- In contrast to BM, VR models the possibility that each branch ing the tree at 21.6 Ma, with those clades having much higher in the phylogeny has its own independent rate of evolution, which species richness on average (Supporting Information Figure S2). we estimated in BayesTraits v.2 (available from http://www.evolu​ Phylogenetically isolated species-poor lineages are comparatively tion.rdg.ac.uk/). The VR model uses reversible-jump Markov chain uncommon in our simulated phylogenetic trees. Therefore, assum- Monte Carlo (rjMCMC) algorithms and two scaling mechanisms to ing constant rates of diversification among lineages, it is apparent identify changes in rate along single branches and for whole clades that the Corvides contain a far greater number of species-poor across the phylogeny (Venditti et al., 2011). We used default priors families than expected. for the phylogenetic mean (α) and Brownian variance (σ) param- eters, running a single rjMCMC chain upon the empirical spe- cies-level scores of each PC axis for 50 million generations. Chains 3.2 | Does morphological positioning predict were sampled every 5,000th generation. We assessed mixing and variation in species richness and rates of lineage convergence of the chains, before the first five million generations diversification? were removed as burn-in. BayesTraits outputs a posterior distribu- tion of trees, in which the branch lengths are scaled by the rate The distance from the centroid of morphological space, species of evolutionary change that best explains the distribution of the diversity and rates of lineage diversification vary extensively trait values at the tips, assuming BM evolution. We summarized among families of the Corvides (Figures 1 and 2). In general, the the posterior distribution of the rate-scaled trees for each PC greater the distance that each family occurs from the centroid axis as a maximum clade credibility (MCC) tree in TreeAnnotator of PC space, the lower its species richness (all families: slope = 2 2 (Drummond & Rambaut, 2007). We then simulated 1,000 null −0.62, R = 0.15, p = .04; excluding BOP: slope = −0.8, R = 0.23, trait data sets under BM for each axis (again using the empirical p = .009; Figure 1b; Supporting Information Figure S3) and rate 2 2 σ values for each PC axis) on the scaled MCC trees (Supporting of lineage diversification (all families: slope = −0.03, R = 0.11, 2 Information Figure S1). p = .07; excluding BOP: slope = −0.03, R = 0.18, p = .02; Figure 1c; 84 | KENNEDY et al.

Mohouidae (3) (a) Eulacestomidae (1) 5 30 F = 19.62 Paramythiidae (2) F = 58.55 (c) (b) p = 0.001 Pteruthidae (9) 4 p = 0.001 25 Vireonidae (54) 20 Oreoicidae (3) 3 wn age (My) ichness (5) 15 Falcunculidae (1) 2 Cinclosomatidae (9) 10 Oriolidae (35) 1 Pachycephalidae (51) 5 Neosittidae (3) Log species r 0 Campephagidae (92) Stem age − Cro 0 Rhagologidae (1) Machaerirhynchidae (2) Widespread Australasian Widespread Australasian Artamidae (23) (N = 15) (N = 15) (N = 15) (N = 15) Aegithinidae (4) Malaconotidae (51) Platysteiridae (30) 0.35 6 F = 38.63 F = 10.33 Vangidae (39) (d) (e) 0.30 p = 0.001 p = 0.005 Dicruridae (24) Lamprolidae (2) 0.25 5

Rhipiduridae (44) er My ) 0.20 Melampittidae (2) 4 Corcoracidae (2) 0.15 Paradisaeidae (41) Ifritidae (1) 0.10 3 Monarchidae (94) Mean DR (P 0.05 Laniidae (34) (127) 0.00 2 Widespread Australasian

Widespread Australasian Distance from the centroid of PC space 0.03 0.33 0 5.7 (N = 15) (N = 15) (N = 15) (N = 15) Mean Diversification 30 25 20 15 10 50 Distance from the Rate (DR) per centroid of PC space million years

FIGURE 2 (a) Phylogenetic relationships among the 30 families of corvoid passerines proposed by Jønsson et al. (2016). Tip colours reflect the mean diversification rate (DR) of each family, with the black squares indicating the crown group ages. The species richness of each family is shown in parentheses, with the mean distance each family occurs from the centroid of principal component (PC) space (all PC axes are standardized to have equal variance) shown next to the tip labels. Family values of distance from the centroid shown in blue represent those present in the Australasian region, whereas those highlighted in red indicate families that have dispersed and diversified on other insular or continental landmasses (widespread). Violin plots indicate the differences in: (b) log10-transformed species richness; (c) the stem to crown group radiations; (d) rates of lineage diversification (mean DR values); and (e) distance from the centroid of PC space between the Australasian and widespread families. The median (white dot), interquartile range (thick black lines) and the upper/lower adjacent values (thin black lines) are shown on all violin plots [Colour figure can be viewed at wileyonlinelibrary.com]

Supporting Information Figure S3). Although it is important ac- 3.3 | Is endemism to the Australasian knowledge that similar negative correlations could potentially ancestral area associated with limited species be expected when the trait data were simulated for all families diversity, diversification and marginal eco- under the BM [species richness: mean null slope −0.46 (95% CI, morphology? −0.42, −0.5); DR: mean slope, −0.028, (95% CI, −0.026, −0.03); Table 1; Figure 3], shallow negative [species richness: mean slope, Families of corvoid passerines that are endemic to Australasia contain −0.05 (95% CI, −0.02, −0.08); Table 1; Figure 3] or positive rela- fewer species (Table 2; Figures 1 and 2b), have longer waiting times tionships [DR: mean slope, 0.002 (95% CI, 0.004, 0.001); Table 2; between the stem and crown group radiations (Table 2; Figure 2c) Figure 3] were expected for the VR model. Furthermore, the and are currently diversifying at lower rates (Table 2; Figure 2d) ranges of expected relationships generated from the simulated compared with families that have dispersed and diversified on other data were extremely broad, and the explanatory power of these continental or insular landmasses. On average, Australasian families models (particularly in the case of VR) was generally low (Table 2; possess morphologies that are distributed further from the centroid Figure 3). Few of our simulated data sets produced models with of PC space than those that are more widespread (Table 2; Figure 2e). both a similar or more negative slope and equivalent or greater The greater distance that the Australasian families occur from the explanatory power, compared with those generated from the centroid of morphospace was far larger than expected if the trait empirical data (Table 1; Figure 3). Importantly, model adequacy data were simulated under the BM or VR null models (Supporting tests suggest that BM is a particularly poor approximation of the Information Figure S5). patterns of trait evolution among the Corvides on many PC axes At the species level, we also found that members of the (Supporting Information Figure S4), being limited in its ability Australasian families occupy increasingly peripheral areas of to account for heterogeneity in rates of evolution among line- PC space regardless of whether the BOP are included in the ages (CVAR; Supporting Information Table S4) and the bursts of Australasian grouping (all measurements: F = 174, t = 13.2, p = .007; phenotypic evolution on individual lineages (DCDF; Supporting excluding tail measurements: F = 94.1, t = 9.7, p = .01; Supporting Information Table S4). Consequently, the VR model should be Information Figure S6) or not (all measurements: F = 25, t = 5, considered the most appropriate null expectation from which to p = .02; excluding tail measurements: F = 32.1, t = 5.7, p = .01; compare the empirical patterns of trait evolution and the associ- Supporting Information Figure S6). In accordance with results at ated correlations among the Corvides. the family level, the increased peripherality of the Australasian KENNEDY et al. | 85

5 (a) 70 (b) 300 (c) 60 250 4 ichness ) 50 200 3 40 150 30 2 Frequency Frequency 100 20 amily species r 1 10 50 Log (F 0 0 0

23456 −2 −1 012 0.00.1 0.20.3 0.40.5 Distance from the centroid of PC space PGLS Model slope (BM null) PGLS R2

0.25 (d) 150 (e) 300 (f) 250 0.20

100 200 er My) 0.15 150 Frequency Frequency 0.10 50 100 Mean DR (P 50 0.05 0 0

23456 −0.2 −0.10.0 0.10.2 0.00.1 0.20.3 0.40.5 Distance from the centroid of PC space PGLS Model slope (BM null) PGLS R2

5 (g) 120 (h) 350 (i) 100 300 4 chness)

ri 250 80 3 200 60 150 2 Frequenc y 40 Frequenc y 100 amily species 1 20 50 Log (F 0 0 0

23456 −2 −1 012 0.00 0.05 0.10 0.15 0.20 Distance from the centroid of PC space PGLS Model slope (VR null) PGLS R2

0.25 350 (j) 100 (k) (l) 300 0.20 80 250 er My ) 0.15 60 200

150

Frequenc y 40 Frequenc y 0.10 100

Mean DR (P 20 50 0.05 0 0

23456 −0.10 −0.05 0.00 0.05 0.10 0.00 0.05 0.10 0.15 0.20 Distance from the centroid of PC space PGLS Model slope (VR null) PGLS R2

FIGURE 3 The relationships between log10-transformed family species richness (a–c,g–i) and rates of lineage diversification (d–f,j–l), as a factor of the mean distance that each corvoid family occurred from the centroid of principal component (PC) space (all PC axes are standardized to have equal variance). In these analyses, the distance from the centroid of PC space was calculated on sets of PC axes simulated under a Brownian motion (a–f) or Variable rates (g–l) model of continuous trait evolution. The phylogenetic least squares regression (PGLS) fits generated across the 1,000 simulations are plotted in a,d,g,j. Regressions with p < .05 are shown as red lines, and those with p > .05 are shown in grey. The mean regression line across all simulations is shown in black. Histograms show the slopes (b,e,h,k) and R2 values (c,f,i,l) of the PGLS models assessing the correlations. Grey bars represent models that showed statistically significant correlations between the variables (p < .05). Red vertical lines represent the values extracted from the PGLS models estimated on the empirical data [Colour figure can be viewed at wileyonlinelibrary.com] families was far more extensive than when trait values were sim- Hallux and hind claw length were the only measurements found ulated under BM or VR (Supporting Information Figure S7). Lower to be differentiated significantly between the Australasian and rates of lineage diversification were the general property of spe- widespread families (Supporting Information Figure S9; Table S5). cies that are members of the Australasian families, with these dif- However, these differences became significantly weaker upon the ferences becoming more evident upon the exclusion of the BOP exclusion of the BOP from the Australasian category (Supporting from the Australasian grouping (including the BOP: F = 61, t = −7.8 Information Figure S10; Table S5). The apparent discrepancy be- p = .04; and excluding the BOP: F = 139.7, t = −11.8, p = .001; tween these sets of results reflects the strong correlation of the Supporting Information Figure S8). values of the original measurements with overall body size (Kennedy 86 | KENNEDY et al. et al., 2019) and that many BOP species are both large bodied and Supporting Information Table S6). We were unable to replicate the possess relatively long hallux and hind claws (Figure 4b,c). Repeating extent of the morphological divergences on PC3 when the trait data the comparisons of morphological positioning using the species were simulated under either the BM or VR null models (Supporting scores on the individual PC axes, we found that the Australasian and Information Figures S12 and S13). These findings are in accordance widespread species had distinct values on PC3, but not on any fur- with the results of the analyses in which PC axes explaining the ther axes (Supporting Information Figure S11; Table S6). The signifi- smallest amounts of variation in the original measurements were cant differences on PC3 remained consistent whether the BOP were removed progressively until only the axes that explained the largest grouped with the Australasian families (F = 316, t = −17.8, p = .001; amount of variance remained (Supporting Information Figure S14; Supporting Information Table S6) or not (F = 85, t = −9.3, p = .001; Table S7).

TABLE 1 Probabilities of observing phylogenetic least squares regression models with (a) an equal or more negative slope, (b) 3.4 | Are the trait combinations that reflect equal or greater R2 value, or (c) both an equal or more negative morphological peripherality consistent among 2 slope and an equal or greater R value, when assessing the families? correlation between log10-transformed species richness or DR, as a factor of family distance from the centroid of PC space Tarsus, hallux, hind claw, bill depth, bill width and outer rectrix were Brownian motion Variable rates the individual measurements that had the greatest loadings on PC3 (Supporting Information Table S2). Considered across different fami- Log (species Log (species richness) DR richness) DR lies, the relative importance of the individual traits in contributing to the patterns of divergence on PC3 between the species-poor and (a) Slope 0.41 0.506 0.08 0.055 species-rich groups varied greatly (Figure 4; Supporting Information (b) R2 0.106 0.169 0.015 0.028 Table S8). The importance of different morphological traits in de- (c) Slope 0.1 0.163 0.013 0.013 2 termining peripherality can be illustrated directly by comparing the and R trait loadings upon the linear discriminants of the individual fami- Note.: In these analyses, the morphological data were simulated under lies (Supporting Information Table S8) and by visualizing the original Brownian motion or Variable rates models of trait evolution 1,000 log -transformed traits plotted against PC1 (a proxy for size), as- times, before the correlations were assessed. DR = diversification rate; 10 PC = principal component. sessing the taxa that were outliers in these relationships (Figure 4).

(a) 3.0 (b) (c) 4.0 3.5 ) aw 2.5 rsus) 3.5 Ta 3.0 Log (Hallux ) Log ( Log (Hind cl 2.0 3.0 2.5

2.5 1.5 −2 −1 0 1234 −2 −1 0 1234 −2 −1 0 1234

PC1 (87%) PC1 (87%) PC1 (87%)

3.5 3.0 (d) (e) 5.0 (f)

3.0 x) 2.5 4.5 2.5 Aegithinidae Corcoracidae Eulacestomidae 4.0 Ifritidae 2.0 2.0 Machaerirhynchidae

Log (Bill width) Melampittidae Log (Bill depth)

Log (Outer rect ri Mohouidae Neosittidae 1.5 3.5 Oreoicidae 1.5 Paramythiidae Psophodidae Rhagologidae Lamprolidae 1.0 3.0 Falcunculidae

−2 −1 0 1234 −2 −1 0 1234 −2 −1 0 1234

PC1 (87%) PC1 (87%) PC1 (87%)

FIGURE 4 Scatterplots showing the relationship between (a) tarsus, (b) hallux, (c) hind claw, (d) bill width, (e) bill depth, and (f) outer rectrix length, as a factor of principal component 1 (PC1; a strong proxy for overall body size). Dashed lines represent the least squares regression fit. Species highlighted by a plus sign are members of the birds-of-paradise, whereas all other highlighted taxa are members of families that contain five or fewer species. The remaining corvoid species are highlighted by an open grey circle . Illustrations by J.F. correspond to the following species, with their data points highlighted by red borders: (a) Melampitta lugubris (Melampittidae); (b) Daphoenositta chrysoptera (Neosittidae); (c) Ifrita kowaldi (Ifritidae); (d) Machaerirhynchus nigripectus (Machaerirhynchidae); (e) Eulacestoma nigropectus (Eulacestomidae); and (f) Machaerirhynchus flaviventer (Machaerirhynchidae) [Colour figure can be viewed at wileyonlinelibrary. com] KENNEDY et al. | 87

TABLE 2 Values of F, t and p derived All families Excluding Paradisaeidae from phylogenetic ANOVAs investigating F t p F t p the differences in species diversity, lineage diversification, waiting times Log (species richness) 58.6 −7.7 < .001 94.3 −9.7 < .001 between the stem and crown group Stem–crown group age 19.6 4.4 < .001 24.7 5 < .001 radiations and distance from the centroid Diversification rate 38.6 −6.2 < .001 53.8 −7.3 < .001 of principal component space among families that are endemic to Australasia Distance from the 10.3 3.2 .003 12 2.9 .006 against those that have dispersed and PC centroid (all diversified on other continental and measurements) insular landmasses Distance from the 11.9 3.45 .003 10.4 3.2 .003 PC centroid (no tail measurements)

PC = principal component.

4 | DISCUSSION negative associations between these variables are more likely under BM (Figure 3), this model is ultimately a poor approximation of the Species-poor and phylogenetically isolated lineages account for evolutionary trends observed in our empirical trait data (Supporting millions of years of unique evolutionary history and are over-rep- Information Table S4). Upon quantifying heterogeneity in rates of resented within higher taxonomic groups, assuming equal rates evolution among corvoid lineages (a process that appears common of speciation and extinction (Bennett & Owens, 2002; Dial & among a wide variety of organismal groups; Chira & Thomas, 2016; Marzluff, 1989; Owens et al., 1999; Scotland & Sanderson, 2004). Cooney et al., 2017; Venditti et al., 2011), it is unclear that lineages For the global radiation of corvoid passerines, we show that species- with the highest evolutionary rates should always be expected to be poor lineages are more common than expected given these assump- located at the periphery of morphospace (Supporting Information tions (Supporting Information Figure S2), and that these taxa tend to Figures S1, S4 and S15). Many species-rich families currently located occupy marginal areas of eco-morphological trait space (Figures 1 closer to the centroid of morphospace also recovered high rates of and 2). Marginality in morphospace is also associated with limited trait divergence on some axes (Supporting Information Figure S1, lineage diversification and restricted geographical distributions, S4 and S15). Therefore, the commonality of marginal morphological such that the majority of species-poor, morphologically peripheral characters among species-poor families (Figures 1, 2 and 4) implies corvoid families are endemic to the ancestral area of Australasia directional evolution in the morphological traits of these lineages to- (Figures 1 and 2). These patterns of diversification are in direct con- wards peripheral ecological positions. trast to the families that possess more generalized combinations of The relative proportions of the tarsus, hallux, hind claw, bill width, morphological traits, which have radiated to accumulate almost an bill depth and the outer rectrix are the morphological characters that order of magnitude more species (Figure 1a) and together are dis- predominantly lead to the marginality of the species-poor families tributed over a near-global extent (Jønsson et al., 2011; Kennedy endemic to Australasia (Figure 4; Supporting Information Table S7) et al., 2018). We argue that the evolution towards peripheral areas and reflect the relatively unusual ecologies of these lineages. For ex- of eco-morphological niche space represents a major limit to geo- ample, the comparatively longer hallux and hind claws of some spe- graphical expansion and lineage diversification. This hypothesis cies depauperate families (Figure 4b,c) are associated with gripping/ can potentially provide a general explanation for why species-poor perching on bark (e.g., Neosittidae, Mohouidae and Ifritidae) or bam- clades are so well represented across the tree of life. boo (e.g., Eulacestomidae) while foraging, and with scansorial habits The finding that species-poor corvoid families are distributed more generally (del Hoyo et al., 2007, 2009). In contrast, the elon- in marginal areas of morphological space (Figures 1 and 2e) is in gated tarsi of the families Melampittidae, Corcoracidae, Oreoicidae agreement with the findings of Ricklefs (2005; see also Chira et al., and Psophodidae (Figure 4a) enable increased locomotion and 2018), despite a significantly increased number of phylogenetically foraging on the ground (del Hoyo et al., 2006-2009). The dispro- isolated species-poor clades being recognized by molecular phylo- portionately wide or deep bills of other clades are features that rep- genetic analyses since that study (only five of 14 corvoid families resent morphological adaptations towards taking large insect prey with fewer than five species were considered by Ricklefs, 2005). in flight (e.g., Machaerirhynchidae; Figure 4d; del Hoyo et al., 2006) Although a negative correlation between species richness/DR and or digging/chipping bark to probe for insects, respectively (e.g., morphological peripherality can potentially be derived from trait Eulacestomidae, Falcunculidae and Mohouidae; Figure 2e; del Hoyo data that are simulated under the VR null (Figure 3), slopes that et al., 2007). Finally, although two species-poor Australasian families are equivalent or more negative than the empirical relationship are (Paramythiidae and Rhagologidae) were not exceptionally morpho- rare (Table 1), and the associated models tend to have extremely logically peripheral in the traits considered in the present analysis, low explanatory power (Figure 3). In comparison to VR, although we note that these characters are not an exhaustive representation 88 | KENNEDY et al. of corvoid eco-morphology (Kennedy et al., 2019), with both of from fossil data (Foote et al., 2007; Quental & Marshall, 2010), the these groups being somewhat atypical ecologically in the respect fossil record is notoriously poor among passerine birds (including that berries represent the major component of their diet (del Hoyo the Corvides), making it difficult to conclude definitively whether (1) et al., 2007, 2008). consistently low rates of net diversification, or (2) extinction leading The morphological traits analysed here have been shown to to a decline from a larger clade size, are the ultimate cause of the low represent an appropriate surrogate for broad-scale ecological dif- species richness of the Australasian families. Despite the potentially ferences among corvid passerines (Kennedy et al., 2019), support- differing formative processes, we argue that the marginal ecology of ing the idea that the evolution towards peripheral morphological the Australasian lineages represents the ultimate factor maintaining positions (Figure 2) is likely to reflect the occupation of marginal the low species diversity and endemism of these clades in the pres- areas of ecological niche space (Cooney et al., 2017; Ricklefs, 2005; ent day. Schluter, 2000; Simpson, 1944, 1953; Figure 2). Divergent selection Although some phylogenetically isolated, highly specialized and pressures that are the outcome of interspecific competition are a species-poor avian taxa have attained relatively widespread distri- likely driver of these trends (Darwin, 1876; Parsons, 2005). Given butions [e.g., the hoatzin (Opisthocomus hoazin) or members of the that Australasia is consistently recovered by biogeographical analy- Bombycillidae], many others tend to be restricted to large, isolated ses to represent the ancestral area of the Corvides, and oscine pas- islands [e.g., Philepittidae in Madagascar, and kiwi (Apteryx spp.) in serines more broadly (Boles, 1995; Jønsson et al., 2011; Kennedy New Zealand]. Endemism to large islands is also apparent among et al., 2017; Moyle et al., 2016; Oliveros et al., 2019), many closely the most species-poor corvoid lineages, generally being restricted related passerine lineages are likely to have been co-distributed in their distributions to New Guinea, New Zealand or Australia (only with one another throughout their evolutionary histories, mak- one of 14 Corvoid families with five or fewer species is distributed ing competitive interactions highly probable (Marki et al., 2019). outside of these areas; Aegithinidae). The range restriction and as- Interactions with competitively superior lineages are theorized to sociated endemism of these lineages imply that large islands located lead to the progressive evolution of ecological specialization and in- within tropical or subtropical latitudes have maintained marginal creasingly restricted geographical distributions (Pepke et al., 2019; ecological niche space over long periods of time, enabling these Ricklefs & Bermingham, 1999, 2002). In turn, the evolution of eco- lineages to persist towards the present. In direct contrast to these logical specialization and marginality limits competition with other distributional and phylogenetic patterns, corvoid families with more taxa, reducing extinction risk and enabling the persistence of these generalized morphology have expanded geographically and diversi- lineages over long periods of evolutionary time (Parsons, 2005; fied much more extensively, attaining a distribution that together Ricklefs, 2005). Consistent with these predictions, our results show encompasses the world's remaining major continental and insular that eco-morphological marginality and endemism to Australasia are landmasses (Jønsson et al., 2011; Kennedy et al., 2017, 2018). If spe- general properties of species-poor corvoid families that have long cies-poor lineages are adapted to specialized sets of resources that stem branches preceding the crown group radiation (Figure 2c,e). are geographically restricted in their distribution, this conceivably Many of the species-poor phylogenetically isolated Australasian lin- inhibits their ability to colonize new areas (Kennedy et al., 2017; eages might well represent examples of species in the later stages Ricklefs, 2005), expand their ranges and undergo allopatric spe- of the taxon cycle, with their marginal morphological forms hav- ciation (Kennedy et al., 2018; Price, 2008). Peripheral trait combi- ing evolved as a consequence of this process (Jønsson et al., 2017; nations might therefore directly inhibit dispersal and geographical Wilson, 1961). range expansion. The taxon cycle hypothesis proposes that a period of geographical The appreciation that many species-poor lineages have un- expansion and diversification precedes relictualization, implying that usual ecologies and associated morphological adaptations has a lineages that are currently species poor and phylogenetically isolated long history in the biological sciences that dates at least as far were once much more species rich (Ricklefs & Bermingham, 1999, back as Darwin (1876). Charismatic and diverse examples of such 2002; Wilson, 1961). Although more species-rich corvoid families lineages are abundant across different taxa, which, in addition to potentially show evidence for a variety of taxon cycle stages [e.g., those previously referred to, also include horsetails (Equisetum), Monarchidae (Fabre et al., 2014) and Pachycephalidae (Jønsson Gingko biloba (Ginkgoaceae), coelacanth fish (Coelacanthiformes), et al., 2014)], an alternative explanation is that corvoid lineages that tuataras (Sphenodontidae), mousebirds (Coliiformes) and serie- are currently species poor and phylogenetically isolated have had mas (Cariamidae). In line with the broad taxonomic span of these consistently low rates of net diversification and have maintained low examples, the uneven distribution of species diversity among species diversity throughout their histories (Ricklefs, 2005). Such a higher taxonomic units can be considered a ubiquitous biological process could also be an expected outcome of the evolution of mar- pattern (Bennett & Owens, 2002; Dial & Marzluff, 1989; Owens ginal ecology among these lineages, because ecological opportuni- et al., 1999; Ricklefs, 2003; Scotland & Sanderson, 2004). Given ties for further diversification would be limited in peripheral areas these observations, it is surprising that explaining the prevalence of the adaptive landscape. Consequently, the carrying capacity of of phylogenetically isolated species-poor lineages has received these clades would be expected to be low. Although the waxing and relatively limited attention from phylogenetic comparative studies waning of clade diversity is well evidenced among a variety of taxa to date. Part of the problem reflects that many small clades were KENNEDY et al. | 89 previously considered to be atypical members of larger families cnr6) and the phylogenetic hypotheses (https://datad​ryad.org/resou​ that have only been recognized as distinct and phylogenetically rce/doi:10.5061/dryad.80n42) can be downloaded from Dryad. isolated taxa in relatively recent times, following molecular phylo- genetic studies. A further issue is that many current methods for ORCID assessing diversification rate variation account for differences in Jonathan D. Kennedy https://orcid.org/0000-0002-2843-122X species richness among lineages through a small number of major rate shifts (Morlon, 2014; Morlon et al., 2010; Rabosky, 2014). REFERENCES Finer-scale rate variation, particularly among phylogenetically Alfaro, M. E., Santini, F., Brock, C., Alamillo, H., Dornburg, A., Rabosky, D. isolated lineages, has thus tended to be overlooked (Huang & L., Carnevale, G., & Harmon, L. J. (2009). Nine exceptional radiations plus high turnover explain species diversity in jawed vertebrates. Rabosky, 2014). The recent development of metrics (Jetz et al., Proceedings of the National Academy of Sciences USA, 106, 13410– 2012) and methods (Maliet et al., 2019) that quantify diversifica- 13414. https://doi.org/10.1073/pnas.08110​87106 tion rate variation at the lineage scale is likely to bring the ex- Bennett, P. M., & Owens, I. P. (2002). Evolutionary ecology of birds: Life tremely low rates of diversification of some lineages to wider histories, mating systems and extinction, Oxford University Press, attention, facilitating statistical analysis that can enable assess- Oxford. Boles, W. E. (1995). The world's oldest . Nature, 374(6517), 21– ment of their underlying causes. 22. https://doi.org/10.1038/374021b0 Phylogenetically isolated lineages represent millions of years Chira, A. M., Cooney, C. R., Bright, J. A., Capp, E. J. R., Hughes, E. C., of independent evolutionary history that tend to be maintained Moody, C. J. A., Nouri, L. O., Varley, Z. K., & Thomas, G. H. (2018). by relatively few extant species. The prevalence of such lineages Correlates of rate heterogeneity in avian ecomorphological traits. Ecology Letters, 21, 1505–1514. https://doi.org/10.1111/ele.13131 is pervasive across taxonomic groups, and explaining their abun- Chira, A. M., & Thomas, G. H. (2016). The impact of rate heterogeneity on in- dance and geographical distribution is necessary to comprehen- ference of phylogenetic models of trait evolution. Journal of Evolutionary sively determine the factors underlying the uneven distribution Biology, 29, 2502–2518. https://doi.org/10.1111/jeb.12979 of species diversity among clades. Species-poor isolated lineages Cooney, C. R., Bright, J. A., Capp, E. J., Chira, A. M., Hughes, E. C., Moody, C. J., Nouri, L. O., Varley, Z. K., & Thomas, G. H. (2017). Mega- have been suggested to be of limited future evolutionary poten- evolutionary dynamics of the adaptive radiation of birds. Nature, tial, yet evolutionary history provides multiple instances where 542(7641), 344–347. https://doi.org/10.1038/natur​e21074 extensive diversification has followed within lineages that were Darwin, C. R. (1876). The origin of species by means of natural selection, or of low species richness in previous time periods, as physical and the preservation of favoured races in the struggle for life. John Murray. environmental conditions changed (e.g., Eutherian mammals, del Hoyo, J., Elliot, A., & Christie, D. A. (2006–2009). Handbook of the birds of the world (Vols 11–14). Lynx Edicions. Neoaves). We conclude that the commonality of species-poor phy- Dial, K. P., & Marzluff, J. M. (1989). Nonrandom diversification within logenetically isolated lineages reflects the evolution of marginal taxonomic assemblages. Systematic Biology, 38, 26–37. https://doi. ecologies among their constituent taxa, which limits geographical org/10.1093/sysbi​o/38.1.26 range expansion, suppresses lineage diversification and increases Drummond, A. J., & Rambaut, A. (2007). BEAST: Bayesian evolutionary analysis by sampling trees. BMC Evolutionary Biology, 7, 214. endemism. Fabre, P. H., Moltensen, M., Fjeldså, J., Irestedt, M., Lessard, J. P., & Jønsson, K. A. (2014). Multiple waves of colonization by monarch ACKNOWLEDGMENTS flycatchers (Myiagra, Monarchidae) across the Indo-Pacific and their We thank the museum collections and associated staff that made implications for coexistence and speciation. Journal of Biogeography, possible collection of morphological data. In this regard, we 41, 274–286. Fjeldså, J., Bowie, R. C., & Rahbek, C. (2012). The role of mountain ranges would particularly like to thank Mark Adams, Hein van Grouw and in the diversification of birds. Annual Review of Ecology, Evolution, and Robert Prys-Jones at the British Museum of Natural History, Lydia Systematics, 43, 249–265. https://doi.org/10.1146/annur​ev-ecols​ys- Garetano, Joel Cracraft and Paul Sweet at the American Museum 10271​0-145113 of Natural History, and Pepijn Kamminga and Steven van der Mije at Foote, M., Crampton, J. S., Beu, A. G., Marshall, B. A., Cooper, R. A., Maxwell, P. A., & Matcham, I. (2007). Rise and fall of species occu- the Naturalis Biodiversity Center (The Netherlands). Chris Cooney pancy in Cenozoic fossil mollusks. Science, 318(5853), 1131–1134. and Gavin Thomas provided useful comments that helped improve https://doi.org/10.1126/scien​ce.1146303 the manuscript. J.D.K. was supported by an Internationalisation Garland Jr, T., Dickerman, A. W., Janis, C. M., & Jones, J. E. (1993). Fellowship (CF17-0239) from the Carlsberg Foundation and an Phylogenetic analysis of covariance by computer simulation. Individual Fellowship from Marie Sklodowska-Curie actions (MSCA- Systematic Biology, 42, 265–292. Gill, F., & Donsker, D. (2010). IOC world names (v.2.7). Retrieved from 792534). All authors wish to thank the Danish National Research http://www.world​birdn​ames.org Foundation for its support of the Center for Macroecology, Evolution Grandcolas, P., & Trewick, S. A. (2016). What is the meaning of extreme and Climate (DNRF96). phylogenetic diversity? The case of phylogenetic relict species. In: R. Pellens & P. Grandcolas, (Eds.), Biodiversity conservation and phyloge- netic systematics. Topics in biodiversity and conservation (Vol. 14, pp. DATA AVAILABILITY STATEMENT 99–115). Springer. The mean morphological measurements are provided in the Harmon, L. J., Weir, J. T., Brock, C. D., Glor, R. E., & Challenger, W. (2008). Supporting Information (Table S1). The morphological measure- GEIGER: Investigating evolutionary radiations. Bioinformatics, 24, ments of individual specimens (https://doi.org/10.5061/dryad.fbg79​ 129–131. https://doi.org/10.1093/bioin​forma​tics/btm538 90 | KENNEDY et al.

Huang, H., & Rabosky, D. L. (2014). Sexual selection and diversification: Moyle, R. G., Oliveros, C. H., Andersen, M. J., Hosner, P. A., Benz, B. W., Reexamining the correlation between dichromatism and speciation Manthey, J. D., Travers, S. L., Brown, R. M., & Faircloth, B. C. (2016). rate in birds. The American Naturalist, 184, E101–E114. https://doi. Tectonic collision and uplift of Wallacea triggered the global songbird org/10.1086/678054 radiation. Nature Communications, 7, 12709. https://doi.org/10.1038/ Jetz, W., Thomas, G. H., Joy, J. B., Hartmann, K., & Mooers, A. O. (2012). ncomm​s12709 The global diversity of birds in space and time. Nature, 491, 444–448. Oliveros, C. H., Field, D. J., Ksepka, D. T., Barker, F. K., Aleixo, A., Andersen, https://doi.org/10.1038/natur​e11631 M. J., Alström, P., Benz, B. W., Braun, E. L., Braun, M. J., Bravo, G. A., Jønsson, K. A., Borregaard, M. K., Carstensen, D. W., Hansen, L. A., Brumfield, R. T., Chesser, R. T., Claramunt, S., Cracraft, J., Cuervo, A. Kennedy, J. D., Machac, A., Marki, P. Z., Fjeldså, J., & Rahbek, C. M., Derryberry, E. P., Glenn, T. C., Harvey, M. G., … Faircloth, B. C. (2017). Biogeography and biotic assembly of Indo-Pacific corvoid (2019). Earth history and the passerine superradiation. Proceedings passerine birds. Annual Review of Ecology, Evolution, and Systematics, of the National Academy of Sciences USA, 116, 7916–7925. https://doi. 48, 231–253. https://doi.org/10.1146/annur​ev-ecols​ys-11031​ org/10.1073/pnas.18132​06116 6-022813 Orme, D., Freckleton, R., Thomas, G., Petzoldt, T., Fritz, S., Isaac, N., & Jønsson, K. A., Fabre, P.-H., Kennedy, J. D., Holt, B. G., Borregaard, Pearse, W. (2012). Caper: Comparative analyses of phylogenetics and M. K., Rahbek, C., & Fjeldså, J. (2016). A supermatrix phy- evolution in R. Retrieved from https://CRAN.R-proje​ct.org/packa​ logeny of corvoid passerine birds (Aves: Corvides). Molecular ge=caper Phylogenetics and Evolution, 94, 87–94. https://doi.org/10.1016/j. Owens, I. P., Bennett, P. M., & Harvey, P. H. (1999). Species richness ympev.2015.08.020 among birds: Body size, life history, sexual selection or ecology? Jønsson, K. A., Fabre, P. H., Ricklefs, R. E., & Fjeldså, J. (2011). Major Proceedings of the Royal Society of London. Series B: Biological Sciences, global radiation of corvoid birds originated in the proto-Papuan ar- 266, 933–939. chipelago. Proceedings of the National Academy of Sciences USA, 108, Pagel, M. (1999). Inferring the historical patterns of biological evolution. 2328–2333. https://doi.org/10.1073/pnas.10189​56108 Nature, 401(6756), 877–884. https://doi.org/10.1038/44766 Jønsson, K. A., Irestedt, M., Christidis, L., Clegg, S. M., Holt, B. G., & Parsons, P. A. (2005). Environments and evolution: Interactions between Fjeldså, J. (2014). Evidence of taxon cycles in an Indo-Pacific pas- stress, resource inadequacy and energetic efficiency. Biological serine bird radiation (Aves: Pachycephala). Proceedings of the Royal Reviews, 80, 589–610. https://doi.org/10.1017/S1464​79310​ Society B: Biological Sciences, 281, 20131727. 5006822 Kennedy, J. D., Borregaard, M. K., Jønsson, K. A., Holt, B., Fjeldså, J., & Pennell, M. W., FitzJohn, R. G., Cornwell, W. K., & Harmon, L. J. (2014). Rahbek, C. (2017). Does the colonization of new biogeographic re- arbutus: Evaluate the adequacy of continuous trait models. R package gions influence the diversification and accumulation of clade rich- version 0.1. https://github.com/mwpen​nell/arbutus ness among the Corvides (Aves: Passeriformes)? Evolution, 71, 38– Pennell, M. W., FitzJohn, R. G., Cornwell, W. K., & Harmon, L. J. (2015). 50. https://doi.org/10.1111/evo.13080 Model adequacy and the macroevolution of angiosperm func- Kennedy, J. D., Borregaard, M. K., Marki, P. Z., Machac, A., Fjeldså, J., tional traits. The American Naturalist, 186, E33–E50. https://doi. & Rahbek, C. (2018). Expansion in geographical and morphological org/10.1086/682022 space drives continued lineage diversification in a global passerine Pepke, M. L., Irestedt, M., Fjeldså, J., Rahbek, C., & Jønsson, K. A. (2019). radiation. Proceedings of the Royal Society B: Biological Sciences, 285, Reconciling supertramps, great speciators and relict species with the 20182181. taxon cycle stages of a large island radiation (Aves: Campephagidae). Kennedy, J. D., Marki, P. Z., Fjeldså, J., & Rahbek, C. (2019). The asso- Journal of Biogeography, 46, 1214–1225. https://doi.org/10.1111/ ciation between morphological and ecological characters across a jbi.13577 global passerine radiation. Journal of Animal Ecology, 89, 1094–1108. Phillimore, A. B., Freckleton, R. P., Orme, C. D. L., & Owens, I. P. (2006). https://doi.org/10.1111/1365-2656.13169 Ecology predicts large-scale patterns of phylogenetic diversifica- Kennedy, J. D., Weir, J. T., Hooper, D. M., Tietze, D. T., Martens, J., tion in birds. The American Naturalist, 168, 220–229. https://doi. & Price, T. D. (2012). Ecological limits on diversification of the org/10.1086/505763 Himalayan core . Evolution, 66, 2599–2613. https://doi. Pigot, A. L., Sheard, C., Miller, E. T., Bregman, T. P., Freeman, B. G., Roll, org/10.1111/j.1558-5646.2012.01618.x U., Seddon, N., Trisos, C. H., Weeks, B. C., & Tobias, J. A. (2020). Kuhn, T. S., Mooers, A. Ø., & Thomas, G. H. (2011). A simple polytomy Macroevolutionary convergence connects morphological form to resolver for dated phylogenies. Methods in Ecology and Evolution, 2, ecological function in birds. Nature Ecology & Evolution, 4, 230–239. 427–436. https://doi.org/10.1111/j.2041-210X.2011.00103.x https://doi.org/10.1038/s4155​9-019-1070-4 Maliet, O., Hartig, F., & Morlon, H. (2019). A model with many small Pigot, A. L., Trisos, C. H., & Tobias, J. A. (2016). Functional traits reveal shifts for estimating species-specific diversification rates. Nature the expansion and packing of ecological niche space underlying an Ecology and Evolution, 3, 1086–1092. https://doi.org/10.1038/s4155​ elevational diversity gradient in passerine birds. Proceedings of the 9-019-0908-0 Royal Society B: Biological Sciences, 283, 20152013. Marki, P. Z. M., Kennedy, J. D., Cooney, C., Rahbek, C., & Fjeldså, J. Price, T. (2008). Speciation in birds. Roberts and Company. (2019). Adaptive radiation and the evolution of nectarivory in a large Quental, T. B., & Marshall, C. R. (2010). Diversity dynamics: Molecular songbird clade. Evolution, 73, 1226–1240. https://doi.org/10.1111/ phylogenies need the fossil record. Trends in Ecology and Evolution, evo.13734 25, 434–441. https://doi.org/10.1016/j.tree.2010.05.002 Mayr, E. (1947). Ecological factors in speciation. Evolution, 1, 263–288. R Core Team (2018). R: A language and environment for statistical comput- Miles, D. B., & Ricklefs, R. E. (1984). The correlation between ecology ing. R Foundation for Statistical Computing. Retrieved from https:// and morphology in deciduous forest passerine birds. Ecology, 65, www.R-proje​ct.org/ 1629–1640. https://doi.org/10.2307/1939141 Rabosky, D. L. (2014). Automatic detection of key innovations, rate Morlon, H. (2014). Phylogenetic approaches for studying diversification. shifts, and diversity-dependence on phylogenetic trees. PLoS One, 9, Ecology Letters, 17, 508–525. https://doi.org/10.1111/ele.12251 e89543. https://doi.org/10.1371/journ​al.pone.0089543 Morlon, H., Potts, M. D., & Plotkin, J. B. (2010). Inferring the dynamics Revell, L. J. (2009). Size-correction and principal components for inter- of diversification: A coalescent approach. PLoS Biology, 8, e1000493. specific comparative studies. Evolution, 63, 3258–3268. https://doi. https://doi.org/10.1371/journ​al.pbio.1000493 org/10.1111/j.1558-5646.2009.00804.x KENNEDY et al. | 91

Revell, L. J. (2012). phytools: An R package for phylogenetic compara- APPENDIX tive biology (and other things). Methods in ecology and evolution, 3, 217–223. Ricklefs, R. E. (2003). Global diversification rates of passerine birds. PERIPHERAL ECO-MORPHOLOGY PREDICTS Proceedings of the Royal Society B: Biological Sciences, 270, 2285– RESTRICTED LINEAGE DIVERSIFICATION AND 2291. https://doi.org/10.1098/rspb.2003.2489 ENDEMISM AMONG CORVOID PASSERINE BIRDS Ricklefs, R. E. (2005). Small clades at the periphery of passerine mor- phological space. The American Naturalist, 165, 651–659. https://doi. 1. org/10.1086/429676 Defining corvoid passerine families and accounting for phy- Ricklefs, R. E., & Bermingham, E. (1999). Taxon cycles in the Lesser logenetic uncertainty in their placement Antillean avifauna. Ostrich, 70, 49–59. https://doi.org/10.1080/00306​ 525.1999.9639749 A primary aim of this study was to determine the relationships Ricklefs, R. E., & Bermingham, E. (2002). The concept of the taxon cycle between morphology, species richness, lineage diversification and in biogeography. Global Ecology and Biogeography, 11, 353–361. https://doi.org/10.1046/j.1466-822x.2002.00300.x geographical distribution among families of corvoid passerines. For Ricklefs, R. E., & Miles, D. B. (1994). Ecological and evolutionary inferences the analyses presented in the main text of this article, we defined from morphology: An ecological perspective, University of Chicago 30 families in accordance with the temporal banding approach out- Press, Chicago, IL. lined by Jønsson et al. (2016). However, in previous family-level com- Ricklefs, R. E., & Travis, J. (1980). A morphological approach to the study of avian community organization. The Auk, 97, 321–338. parative studies of the Corvides (Kennedy et al., 2017, 2018), family Schluter, D. (2000). The ecology of adaptive radiation, Oxford University units that broadly followed the IOC classifications have also been Press, Oxford. analysed (Gill & Donsker, 2010). Therefore, we assessed whether re- Scotland, R. W., & Sanderson, M. J. (2004). The significance of few ver- peating our analysis following the IOC classifications had any impact sus many in the tree of life. Science, 303(5658), 643. https://doi. org/10.1126/scien​ce.1091483 on our findings and conclusions. Although the temporal banding ap- Simpson, G. G. (1944). Tempo and mode in evolution, Columbia University proach of Jønsson et al. (2016) aimed to define family units that were Press, New York, NY. as consistent as possible with the IOC classifications, there are minor Simpson, G. G. (1953). The major features of evolution. Columbia University differences with respect to the lineages that are considered families Press. Stadler, T. (2017). TreeSim: Simulating phylogenetic trees. Retrieved from and, consequently, the species that are members of the different fam- https://CRAN.R-proje​ct.org/packa​ge=TreeSim ily units. In the temporal banding classifications, a single IOC family, Venditti, C., Meade, A., & Pagel, M. (2011). Multiple routes to mamma- Pityriasidae (monotypic; Pityriasis gymnocephala) is classified in the lian diversity. Nature, 479, 393–396. https://doi.org/10.1038/natur​ Malaconotidae, whereas the families Prionopidae, Philentomidae e10516 and Tephrodornidae all become members of the Vangidae. In con- Wilson, E. O. (1961). The nature of the taxon cycle in the Melanesian ant fauna. The American Naturalist, 95, 169–193. https://doi. trast, Falcunculidae (monotypic; Falcunculus frontatus), Lamprolidae org/10.1086/282174 (Chaetorhynchus papuensis and Lamprolia victoriae) and Pteruthidae (genus Pteruthius) are all raised to family status, splitting from IOC BIOSKETCH families Cinclosomatidae, Rhipiduridae and Vireonidae, respec- tively. Aside from these changes, all other IOC families contained Jonathan D. Kennedy is a postdoctoral research fellow based at the same sets of species as in the temporal banding classifications. the University of Sheffield. The aim of his research is to under- In the analyses of the IOC families, we classified Pityriasidae with stand the distribution and diversification of passerine birds at the Australasian grouping, because although endemic to Borneo this the global scale, through comparative analysis of geographical, colonization from Australasia happened relatively early in the history genetic and eco-morphological trait data. of the Corvides (Jønsson et al., 2011; Kennedy et al., 2017; Oliveros et al., 2019), and none of its relatives are distributed nearby (classify- SUPPORTING INFORMATION ing this family with the widespread families did not change our re- Additional supporting information may be found online in the sults). The families Prionopidae, Philentomidae and Tephrodornidae Supporting Information section. were all classified in the widespread grouping. Recently, Oliveros et al. (2019) generated a phylogeny of 137 passerine families, including almost all of the corvoid families con- How to cite this article: Kennedy JD, Marki PZ, Fjeldså J, sidered in the present study (30 of 31 families following IOC, and Rahbek C. Peripheral eco-morphology predicts restricted 30 of 30 using the temporal banding definitions). This phylogeny lineage diversification and endemism among corvoid passerine was generated from analysis of > 4,000 ultraconserved elements, birds. Global Ecol. Biogeogr. 2021;30:79–98. https://doi. a far greater number of loci than sampled by Jønsson et al. (2016). org/10.1111/geb.13194 Although the inter-familial relationships among the Corvides 92 | KENNEDY et al.

FIGURE A1 Phylogenies of the family-level relationships among the Corvides as proposed by the studies of Jønsson et al. (2016), defined based on temporal banding (left), and Oliveros et al. (2019) (right). Blue curved lines show the links between the positions of the individual families in the topologies of the two phylogenetic studies [Colour figure can be viewed at wileyonlinelibrary.com] proposed by Oliveros et al. (2019) remain largely similar to those diversifying at the lowest rates also tend to be significantly more inferred by Jønsson et al. (2016), there are some slight differences morphologically peripheral (Table A1). Comparing the Australasian in the topological positioning and inter-relationships among a few families with those that have dispersed and diversified in other con- groups (Figure A1). Therefore, we explored the influence of these tinental and insular settings, we also continue to support the find- topological and branch length differences on our results by repeat- ings that Australasian families have: (a) lower numbers of species; ing the relevant family-level comparative analyses (for both the IOC (b) longer waiting times between the stem and crown group radia- and temporal banding taxonomic classifications), using the phylo- tions; (c) lower rates of diversification; and (d) and more marginal genetic relationships among corvoid families inferred by Oliveros morphologies (Table A2). et al. (2019) to account for the non-independence of the data points. Regardless of (a) the backbone phylogeny analysed; (b) whether 2. Sensitivity analyses excluding species placed in the phylogeny the IOC or temporal families were analysed; (c) whether the tail without DNA sequence data measurements were included in the morphological analyses; or (d) whether the family Paradisaeidae (birds-of-paradise; BOP) was The morphological and phylogenetic data set analysed in the included, all of our results remain fully consistent with those pre- main text of the article contains 782 corvoid species, but the initial sented in the main text of the article. We continue to recover sig- species-level phylogeny generated by Jønsson et al. (2016) analysed nificantly negative correlations between the number of species in mitochondrial and nuclear loci for only 665 species. To generate a each family and the distance that those families occur from the cen- phylogeny for all 782 species, the remaining taxa for which DNA troid of morphospace (Table A1). Corvoid families that are currently sequences were unavailable at the time of the tree generation were KENNEDY et al. | 93 x axis) and y axis). Lines represent the least squares regression fit [Colour figure can be viewed at wileyonlinelibrary.com] those that account for the influence of phylogeny before their computation ( FIGURE A2 Scatterplots showing the correlations between the same principal component (PC) that axes were (PC1–PC10) generated in the absence of phylogenetic correction ( 94 | KENNEDY et al.

TABLE A1 Results from phylogenetic Family species richness Diversification rate least squares regression models 2 2 assessing the correlation between log10- Slope R p Slope R p transformed family species richness/ diversification rate [mean DR (Jetz Jønsson tree/IOC −0.59 0.13 .048 −0.03 0.12 .059 families et al., 2012) per family] as a factor of the distance that each family occurs from the Jønsson tree/IOC −0.73 0.18 .019 −0.04 0.17 .021 centroid of principal component space families/no tail Jønsson tree/IOC −0.79 0.21 .012 −0.04 0.18 .018 families/no BOP Jønsson tree/IOC −0.85 0.23 .007 −0.04 0.22 .009 families/no tail/no BOP Oliveros tree/IOC −0.59 0.13 .051 −0.03 0.12 .063 families Oliveros tree/IOC −0.71 0.17 .023 −0.04 0.17 .025 families/no tail Oliveros tree/IOC −0.78 0.21 .013 −0.04 0.18 .02 families/no BOP Oliveros tree/IOC −0.83 0.23 .009 −0.04 0.22 .011 families/no tail/no BOP Oliveros tree/temporal −0.62 0.15 .037 −0.03 0.11 .067 families Oliveros tree/temporal −0.74 0.2 .014 −0.04 0.17 .025 families/no tail Oliveros tree/temporal −0.81 0.23 .009 −0.04 0.18 .022 families/no BOP Oliveros tree/temporal −0.85 0.25 .005 −0.04 0.22 .011 families/no tail/no BOP

Note.: These analyses assessed the sensitivity of the models to: (a) using the Jønsson or Oliveros phylogenies to control for the non-independence of the family-level data; (b) following the IOC or temporal banding family classifications; (c) upon the exclusion of the tail measurements when assessing distance from the centroid of PC space; and (d) upon exclusion of the family Paradisaeidae (BOP) from the analysis.

added as polytomies based on taxonomic information. The branch information]. This resulted in a phylogeny containing 663 species that lengths subtending these species were subsequently derived using could be associated with our morphological data. the polytomy resolver method (Kuhn et al., 2011). Species without Results derived from analyses of the data set containing 663 DNA sequence data were generally placed in terminal positions within species are fully congruent with those assessing the same trends well-supported clades, and therefore, their phylogenetic position can among all 782 species. We found that species that are members be considered conservative with respect to taxonomy. However, the of the Australasian families are found at a greater distance from principle of conservative placement might not necessarily apply with the centroid of PC space and are also diversifying at lower rates regard to the morphological characters. Specifically, adding species (Table A3). These patterns remain consistent whether the 41 spe- to the phylogeny based solely on taxonomic information could lead cies of the family Paradisaeidae are included with the Australasian to a breakdown of the true patterns of phylogenetic trait structure. grouping or not (Table A3). Comparisons of the differences in the

This issue has been suggested to lead to inaccurate results in phy- values of the log10-transformed morphological measurements and logenetic comparative analyses performed using trees generated in PC scores between the Australasian and widespread families are this manner (Rabosky, 2015). To assess the potential influence of this presented in Tables A4 and A5, respectively. In congruence with issue upon the results presented in the main text of this article, we the analysis of all 782 species, we find that of the original meas- repeated the relevant phylogenetic comparative analyses after exclu- urements, tarsus, hallux and hind claw show the greatest deviation sion of the taxonomically placed species [pruning the 119 species that between groups, with Australasian endemic species having the were added to the tree of Jønsson et al. (2016) based on taxonomic longest values (Table A4). However, these differences are driven KENNEDY et al. | 95

primarily by the inclusion of Paradisaeidae in the Austrasian group- ing (Table A4). Tarsus, hallux and hind claw, in addition to bill width, bill depth and outer rectrix, are the measurements that have the highest loadings on PC3. This was the only PC axis to be differenti- ated significantly between members of the Australasian and wide- 8.9 9.6 10.4 Centroid Centroid distance (no tail) 3 3.1 3.2 .008 .004 .003 spread families (Table A5), irrespective of the inclusion or exclusion of Paradisaeidae species.

3. Sensitivity analyses using phylogenetically corrected principal 7.9 8.1 8.7 Centroid Centroid distance 2.8 2.9 3 .007 .012 .007 component scores

Given that a major aim of this study was to assess the morpho- 37 45.7 53.8 DR −6.1 −6.8 −7.3 .001 .001 .001 logical positioning of clades containing different numbers of spe- cies, it is important to evaluate how the variation in clade richness could itself influence the relative trait loadings on the different PC axes. Specifically, the major axes of this variation are likely to be 28 31.2 24.7 Stem age – crown age 5.3 5.6 5 .001 .001 .001 dominated by the morphologies of species-rich groups that contain many closely related species (that probably possess comparatively similar morphologies), as opposed to lineages that are phyloge- netically isolated and species poor. Therefore, we assessed the influence of correcting for phylogenetic non-independence before 49 71.1 94.3 Excluding Paradisaeidae Excluding Log (species richness) −7 −8.44 −9.7 .001 .001 .001 computing our principal component scores. Specifically, we calcu- lated phylogenetic principal component (pPC) scores using the 10 original measurements for all 782 species in the R package phytools (Revell, 2012), with the correlation structure obtained by maximiz- ing the value of Pagel's λ. After standardization of the pPC such that all axes had equal variance, we repeated the calculation of 10.5 11.2 12 Centroid distance (no tail) 3.2 3.4 3.5 .008 .009 .004 the distance that each species occurred from the centroid of pPC space before also estimating the mean distance from the centroid for each family. We then repeated the relevant phylogenetic com- parative analyses to compare the results with those obtained using 9.7 9.9 10.3 Centroid Centroid distance 3.1 3.2 3.2 .005 .003 .003 the original PC scores. The loadings of the 10 original measurements onto the pPC axes are shown in Table A6. At the species level, considering all 10 27.8 33.6 38.6 DR −5.3 −5.8 −6.2 .001 .001 .001 measurements, values on the same pPC and PC axes are, in gen- eral, strongly correlated with one another (Figure A2). Although these relationships become progressively weaker on the higher PC axes, the strong correlations among the species-level values re- 21.8 24.3 19.6 Stem age– crown age 4.7 4.9 4.4 .001 .001 .001 flect similar loadings of the original measurements onto the same axes across both sets of analyses. Comparing the Australasian and widespread families, as with the PC that did not correct for phy- logeny before their computation, we found that members of the Australasian families occupy increasingly peripheral areas of pPC 34.1 46.1 58.6 All families All Log (species richness) −5.8 −6.8 −7.7 .001 .001 .001 space (all species: F = 136.7, t = 11.7, p = .01; excluding the BOP: F = 19.2, t = 4.4, p = .059). Furthermore, as with the phylogeneti- , t and p derived from phylogenetic ANOVAs assessing differences in the values of the species richness, stem–crown age, diversification rate (DR) and distance from F F F t t t p p p cally uncorrected PC scores, the Australasian and widespread spe- cies were significantly divergent on pPC3, but not on any further axes (Table A7). The highly similar nature of these results to those presented in the main text of the article reflect that tarsus, hallux, hind claw, bill depth, bill width and outer rectrix remain the indi- vidual measurements that have the greatest loadings on both PC3 and pPC3 (Table A6). IOC familiesIOC familiesIOC temporalfamilies Jønsson backbone/ Jønsson backbone/ Oliveros backbone/ Oliveros TABLE A2 TABLE Values of F non-independence. phylogenetic the centroid of principal component (PC) space (both including and excluding tail measurements in this computation) among families that are endemichave dispersed to Australasia, to other continental compared and with insular those landmasses that Note.: These analyses were performed using both the IOC and temporal banding family classifications, with both the Jønsson et al. (2016) and Oliveros et al. phylogenies (2019) used to correct for 96 | KENNEDY et al.

TABLE A3 Values of F, t and p derived F t p from phylogenetic ANOVAs assessing differences in the distance from the Distance from the centroid (all All species/IOC 158 12.6 .004 centroid of principal component (PC) measurements) families space and diversification rates (DR) No BOP/IOC families 17.9 4.2 .075 among species that are members of All species/temporal 153.7 12.4 .002 the Australasian and widespread family families groupings No BOP/temporal 16.7 4.1 .059 families Distance from the centroid (no All species/IOC 143.4 12 .004 tail measurements) families No BOP/IOC families 25.9 5.1 .027 All species/temporal 138.2 11.8 .003 families No BOP/temporal 23.7 4.9 .025 families Diversification rate (DR) All species/IOC 48.3 −6.9 .062 families No BOP/IOC families 118.9 −10.9 .001 All species/temporal 51.7 −7.2 .072 families No BOP/temporal 125.4 −11.2 .001 families

Note.: In addition to assessing the sensitivity of these results to the inclusion or exclusion of the tail measurements, we also determined the influence of following the IOC or temporal banding family units and of excluding the family Paradisaeidae (BOP) from the analysis.

TABLE A4 Values of F, t and p derived from phylogenetic ANOVAs assessing differences in the values of the 10 log10-transformed morphological measurements between species that are members of the Australasian families and those that have dispersed to other continental and insular landmasses

IOC families Temporal families

Excluding BOP (623 All 663 species Excluding BOP (623 species) All 663 species species)

Measurement F t p F t p F t p F t p

Tarsus 49.8 7.1 .079 3.6 1.9 .376 46.2 6.8 .081 3.6 1.9 .395 Hallux 71.2 8.4 .031 0.1 0.2 .907 66.1 8.1 .045 0.1 0.2 .93

Hallux + mid claw 55 7.4 .063 0.2 0.5 .85 51.1 7.2 .063 0.2 0.5 .835 Bill length 18.2 4.3 .286 3.6 −1.9 .437 15.8 4 .316 3.6 −1.9 .406 Bill depth 2.6 1.6 .659 1.6 −1.2 .564 1.7 1.3 .735 1.6 −1.3 .58 Bill width 0.1 −0.3 .925 9.7 −3.1 .175 0.5 −0.7 .857 9.7 −3.1 .174 Outer rectrix 0.9 1 .81 2.2 −1.5 .516 0.7 0.8 .83 2.2 −1.5 .504 Longest rectrix 9.8 3.1 .422 1.2 −1.1 .634 9.1 3 .451 1.2 −1.9 .666 Primaries 6.5 2.6 .509 4.7 −2.2 .343 5.7 2.3 .566 4.7 −2.2 .334 Secondaries 23.8 4.9 .195 2.3 −1.5 .471 22 4.7 .229 2.3 −1.5 .514

Note.: These analyses consider only: (a) the 663 species sampled with DNA sequence in the phylogeny of Jønsson et al. (2016); or (b) further excluding the 40 sampled species of Paradisaeidae (BOP). KENNEDY et al. | 97

TABLE A5 Values of F, t and p derived from phylogenetic ANOVAs assessing differences in the values of the species scores on the 10 principal component (PC) axes between species that are members of the Australasian families and those that have dispersed to other continental and insular landmasses

IOC families Timeslice families

All 663 species Excluding BOP (623 species) All 663 species Excluding BOP (623 species)

F t p F t p F t p F t p

PC1 15.6 4 .325 1.4 −1.2 .586 13.6 3.7 .339 2.2 −1.5 .495

PC2 0.9 1 .813 < 0.1 < −0.1 .988 0.5 0.7 .862 0.2 −0.4 .832 PC3 284.7 −16.9 .001 67.4 −8.2 .001 289.6 −17 .001 71.2 −8.4 .001

PC4 < 0.1 0.2 .958 0.2 0.4 .865 0.2 0.5 .912 0.7 0.8 .685 PC5 13.4 3.7 .367 20.7 −4.6 .042 14.8 3.8 .305 17.3 −4.2 .07 PC6 21.9 4.7 .215 1.1 −1.1 .615 20.5 4.5 .258 1.5 −1.2 .589 PC7 26.8 −5.2 .21 3.3 −1.8 .416 27.2 −5.2 .189 3.6 −1.9 .397 PC8 1.1 −1 .791 0.8 −0.9 .668 0.5 −0.7 .862 0.2 −0.4 .849 PC9 10.5 3.2 .416 3 −1.7 .439 11.4 3.4 .394 2.2 −1.5 .486 PC10 6.6 −2.6 .51 0.1 0.4 .864 5.9 −2.4 .526 0.3 0.5 .793

Note.: These analyses consider only: (a) the 663 species sampled with DNA sequence in the phylogeny of Jønsson et al. (2016); or (b) further excluding the 40 sampled species of Paradisaeidae (BOP).

TABLE A6 Loadings of the 10 log10-transformed morphological measurements on the individual phylogenetic principal component (pPC) axes

Measurement PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC8 PC9 PC10

Tarsus −0.78 −0.1 −0.41 0.09 0.04 −0.03 −0.44 0.07 0.02 0.03

Hallux −0.84 −0.28 −0.35 0.14 0.1 −0.08 0.23 < 0.01 0.01 0.1 Hallux + mid claw −0.87 −0.23 −0.36 0.14 0.06 −0.06 0.1 0.02 −0.01 −0.15 Bill length −0.86 −0.25 0.12 0.04 −0.41 −0.13 −0.01 −0.04 0.01 < 0.01 Bill depth −0.83 −0.36 0.3 0.05 0.17 < 0.01 −0.06 −0.23 0.01 < 0.01 Bill width −0.76 −0.35 0.41 0.09 0.09 0.01 0.01 0.34 < 0.01 < 0.01 Outer rectrix −0.82 0.29 0.01 −0.44 0.08 −0.19 0.01 0.02 < 0.01 < 0.01 Longest rectrix −0.74 0.63 0.08 0.21 <0.01 0.01 0.01 −0.01 < 0.01 < 0.01 Longest primary feather −0.89 −0.02 −0.09 −0.22 −0.07 0.34 0.05 0.01 0.14 < 0.01 First secondary feather −0.92 −0.04 −0.08 −0.15 −0.06 0.25 −0.01 −0.01 −0.21 0.01

Variance explained 0.67 0.14 0.06 0.04 0.02 0.02 0.02 0.02 < 0.01 < 0.01 98 | KENNEDY et al.

TABLE A7 Values of F, t and p derived from phylogenetic ANOVAs assessing differences in the values of the species scores on the 10 phylogenetic principal component (pPC) axes between taxa that are members of the Australasian families and those that have dispersed to other continental and insular landmasses

IOC families Temporal families

All species Excluding BOP All species Excluding BOP

F t p F t p F t p F t p

PC1 17 4.1 .293 1.7 −1.3 .595 15.1 3.9 .365 2.4 −1.6 .514

PC2 7.6 −2.7 .513 < 0.1 0.2 .926 5.8 −2.4 .532 0.5 0.7 .765 PC3 312.3 −17.7 .001 76.7 −8.8 .001 322.4 −18 .001 83.5 −9.1 .001 PC4 42.5 6.5 .118 8.2 2.9 .21 39.6 6.3 .108 6.6 2.6 .255 PC5 12.5 −3.5 .355 9.6 3.1 .194 13.6 −3.7 .325 7.6 2.8 .22 PC6 0.1 0.3 .957 10.5 −3.2 .165 0.1 0.3 .953 10 −3.2 .173 PC7 0.5 0.7 .876 8.3 −2.9 .217 0.7 0.8 .849 7.1 −2.7 .23 PC8 12.6 −3.6 .337 15.9 −4 .086 14 −3.7 .343 18 −4.2 .056 PC9 59.1 −7.7 .052 1.2 −1.1 .628 60.2 −7.8 .06 1.6 −1.3 .601 PC10 18.2 4.2 .28 0.5 −0.7 .774 17.4 4.2 .263 0.6 −0.8 .697

Note.: These analyses consider the data set of: (a) all 782 species; or (b) excluding species of the Paradisaeidae (BOP).

Kuhn, T. S., Mooers, A. Ø. & Thomas, G. H. (2011). A simple polytomy resolver REFERENCES for dated phylogenies. Methods in Ecology and Evolution, 2, 427–436. Gill, F. & Donsker, D. (2010). IOC world bird names (v.2.7.). Retrieved Moyle, R. G., Oliveros, C. H., Andersen, M. J., Hosner, P. A., Benz, B. from http://www.world​birdn​ames.org W., Manthey, J. D., … Faircloth, B. C. (2016). Tectonic collision and Jetz, W., Thomas, G. H., Joy, J. B., Hartmann, K. & Mooers, A. O. (2012). uplift of Wallacea triggered the global songbird radiation. Nature The global diversity of birds in space and time. Nature, 491, 444–448. Communications, 7, 12709. Jønsson, K. A., Fabre, P. H., Kennedy, J. D., Holt, B. G., Borregaard, M. Oliveros, C. H., Field, D. J., Ksepka, D. T., Barker, F. K., Aleixo, A., K., Rahbek, C., … Fjeldså, J. (2016). A supermatrix phylogeny of cor- Andersen, M. J., … Faircloth, B. C. (2019). Earth history and the pas- void passerine birds (Aves: Corvides). Molecular Phylogenetics and serine superradiation. Proceedings of the National Academy of Sciences Evolution, 94, 87–94. USA, 116, 7916–7925. Kennedy, J. D., Borregaard, M. K., Jønsson, K. A., Holt, B., Fjeldså, J. & Rabosky, D. L. (2015). No substitute for real data: a cautionary note on Rahbek, C. (2017). Does the colonization of new biogeographic re- the use of phylogenies from birth–death polytomy resolvers for gions influence the diversification and accumulation of clade richness downstream comparative analyses. Evolution, 69, 3207–3216. among the Corvides (Aves: Passeriformes)? Evolution, 71, 38–50. Revell, L. J. (2009). Size-correction and principal components for inter- Kennedy, J. D., Borregaard, M. K., Marki, P. Z., Machac, A., Fjeldså, J. specific comparative studies. Evolution, 63, 3258–3268. & Rahbek, C. (2018). Expansion in geographical and morphological Revell, L. J. (2012). phytools: an R package for phylogenetic compara- space drives continued lineage diversification in a global passerine tive biology (and other things). Methods in Ecology and Evolution, 3, radiation. Proceedings of the Royal Society B: Biological Sciences, 285, 217–223. 20182181.