Comparative morphology of the avian bony columella

In birds, the columella is the only bony element of the sound conducting apparatus, conveying vibrations of the cartilaginous extracolumella to the fluid of the inner ear. Although avian columellar morphology has attracted some attention over the past century, it nonetheless remains poorly described in the literature. The few existing studies mostly focus on morphological descriptions in relatively few taxa, with no taxonomically broad surveys yet published. Here we use observations of columellae from 401 extant bird species to provide a comprehensive survey of columellar morphology in a phylogenetic context. We describe the columellae of several taxa for the first time and identify derived morphologies characterizing higher‐level clades based on current phylogenies. In particular, we identify a derived columellar morphology diagnosing a major subclade of Accipitridae. Within Suliformes, we find that Fregatidae, Sulidae, and Phalacrocoracidae share a derived morphology that is absent in Anhingidae, suggesting a secondary reversal. Phylogenetically informed comparisons allow recognition of instances of homoplasy, including the distinctive bulbous columellae in suboscine passerines and taxa belonging to Eucavitaves, and bulging footplates that appear to have evolved at least twice independently in Strigiformes. We consider phylogenetic and functional factors influencing avian columellar morphology, finding that aquatic birds possess small footplates relative to columellar length, possibly related to hearing function in aquatic habitats. By contrast, the functional significance of the distinctive bulbous basal ends of the columellae of certain arboreal landbird taxa remains elusive.


| INTRODUCTION
A tympanic middle ear evolved multiple times among vertebrates as a means to compensate for the impedance mismatch between air and the fluid-filled inner ear (Clack, 2002;Lombard & Bolt, 1979;Tucker, 2017).In birds, this system consists of a tympanic membrane, a cartilaginous extracolumella, a bony columella, several ligaments, and one muscle (see Figure 1, Smith, 1904, andPohlman, 1921, for further details).
Some previous authors have used the term "columella" to refer to the composite structure comprising both the bony columella and the cartilaginous extracolumella (e.g., Pohlman, 1921;Starck, 1995), while the term "stapes" has been used to refer to the bony columella specifically.In this study, we use the term "columella" to refer only to the bony structure and do not consider the cartilaginous extracolumella.
The avian columella exhibits considerable morphological diversity.While several studies have been published detailing its morphology in individual bird species or clades, much morphological variation has remained undescribed, and few broad comparative studies exist.The most comprehensive studies to date were those of Krause and Feduccia.The study by Krause (1901) was the first attempt at a broad comparative survey, exploring the morphology of the columella in 70 species, attempting to categorize different types of variation.This study described numerous columellar morphologies for the first time, and provided detailed illustrations.The studies published by Feduccia (1974Feduccia ( , 1975aFeduccia ( , 1975bFeduccia ( , 1975cFeduccia ( , 1976Feduccia ( , 1977Feduccia ( , 1978)), were mainly focused on the phylogenetic utility of columellar morphology.Feduccia described a putatively ancestral columellar form that is common to most extant birds, and various derived forms that he took as evidence of phylogenetic relationships.Feduccia described the "primitive" columella as having a flat footplate, a shaft that "is a straight bony rod emanating approximately from the center of the footplate," and lacking struts at the basal end.The columella of Apteryx (see Figure 2 [3]) was taken as an exemplar typifying this morphology.
Feduccia described certain columellar forms that appeared to be diagnostic of various clades (such as Upupiformes, and Alcedinidae, Meropidae, Momotidae, and Todidae); however, his studies also led to several erroneous conclusions (e.g., placing Balaeniceps within Ciconiidae, or suggesting that Tyranni formed an exclusive clade with Alcedinidae).Mayr (2003) described similar columellar morphologies in Steatornis and Trogonidae, and suggested these similarities, along with other morphological traits, as evidence of close affinities between the clades, a conclusion that has not been supported by recent sequence-based phylogenies (e.g., Hackett et al., 2008;Jarvis et al., 2014;Kimball et al., 2019;Kuhl et al., 2021;Prum et al., 2015).
Relatively few studies have attempted to examine the functional morphology of the avian middle ear.Some of the earliest work was by Schwartzkopff (1955Schwartzkopff ( , 1968)).He suggested a relationship between the middle ear area ratio (the ratio of tympanic membrane to footplate area) and a species' "manner of life."He noted relatively large area ratios and longer columellae in Passeriformes and Strigiformes.In his 1968 publication, he also suggested that relatively large footplates and more robust columellae were characteristic of diving species.
A more quantitative investigation was undertaken by Peacock, Spellman, Tollin, and Greene (2020), which showed that a relationship exists between both the length and mass of the columella and the resonance frequency of the middle ear.Other quantitative studies of overall middle ear morphology have shown morphological differences between aquatic and terrestrial birds, with aquatic diving species being especially distinct (Zeyl et al., 2022).The relationship between middle ear morphology and hearing acuity has been explored in other studies (Claes, 2018;Peacock, Spellman, Tollin, & Greene, 2020;Zeyl et al., 2023), all of which showed some relationship between various morphological traits and features of the audiogram.However, these studies did not specifically examine the morphological diversity among avian columellae.
A study of columellar morphology in an allometric and phylogenetic context is essential to assess whether the columella displays adaptive morphological features.With this in mind, the goal of the present study is to survey the morphological diversity of avian columellae.We describe the variation in morphology within and between different clades and, for the first time, we assess columellar morphology in light of current phylogenetic hypotheses for extant birds.We therefore aim to evaluate whether avian columellar morphology exhibits phylogenetic diagnostic value that was overlooked by earlier authors, and whether updated phylogenetic hypotheses for extant birds help identify previously overlooked derived morphologies representing autapomorphies of certain clades.

| METHODS
Following the terminology of Feduccia (1975a), we divide the columella into four parts: the distal end, the shaft, the basal end, and the footplate (see Figure 3).The distal and basal ends are often flared, distinguishing them from the straight, central shaft.Each of these columellar components exhibits differences in form across avian phylogeny.The extracolumella is attached via a synchondrosis to the distal end, and it is at the basal end that the shaft joins to the tympanic side of the footplate.The other side of the footplate faces the fluid of the inner ear, which we term the perilymphatic side.
We examined columellae which had been extracted from the dried skulls of 401 species (40 order-and 146 family-level clades following IOC taxonomy; Gill et al., 2023) from museum collections.No birds were killed for this study.The vast majority of the specimens examined were wild birds and information on the sex, age, method of preparation, and so forth, is incomplete.A full list of species is provided in the Appendix S1.Specimens derived from the collections of the Denver Museum of Nature and Science, Denver (DMNS), and the Field Museum of Natural History, Chicago (FMNH).Two additional specimens were qualitatively examined: the first (Phodilus badius) from the University of Cambridge Museum of Zoology (UMZC) at the University of Cambridge, UK, was imaged using a Nikon XTEK H 225 ST MicroCT scanner at a resolution of 30 microns; the second (Rhynochetos jubatus) from the collections of the Senckenberg Research Institute, Frankfurt, Germany (SMF) was examined from photographs.For almost all species, we were only able to examine one columella due to limited availability of material.In only a few of the more common species were we able to examine columellae from multiple individuals; however, only measurements from one columella were used in any formal analysis.
Each specimen was examined under a Zeiss Stemi 2000-C Stereo Microscope (Carl Zeiss AG, Oberkochen, Germany).To facilitate more precise assessments of differences between clades, we also took several linear measurements using a graticule as follows: 1.The width of the distal and basal ends at their widest points (i.e., diameters of the major axes), and the widest diameters perpendicular to these axes (i.e., diameters of the minor axes).2. The diameter of the footplate along its major axis and the widest diameter perpendicular to this (its minor axis).
F I G U R E 2 Palaeognathae, Galloanserae, and Neoaves (excluding Phaethoquornithes and Telluraves): Drawings of the columella in 62 species to illustrate the diversity of forms.These are shown in different orientations so as to illustrate key features.Note that the color of the bones varies with the condition of the specimen and the conditions under which it was observed: no other meaning should be ascribed to these differences.Palaeognathae: 1. F I G U R E 3 A drawing of a columella with the major sections labeled.
3. The cross-sectional diameter of the shaft at its narrowest point and the maximum cross-sectional diameter at the same point along the distal-basal axis of the shaft.4. The total length of the columella.
From these measurements we were able to derive various other properties including: 5.The area of the footplate, calculated as the area of an ellipse.6.The area of the basal end, calculated as the area of an ellipse.7. The eccentricity of the footplate, calculated from the equation: , where D 1 is the major axis and D 2 the minor axis of the footplate.The eccentricity ranges from 0 to 1, with 0 denoting a perfect circle and 1 a line.8.The ratios of the above measurements to one another.
Since the widths of the distal end, basal end, the shaft and the area of the footplate scale with the overall length of the columella, we performed both ordinary least squares (OLS) regression analysis and phylogenetic generalized least squares (PGLS) regression analysis (Grafen, 1989) on log transformed data.PGLS analyses were conducted using the Caper package in R (Orme et al., 2013), with a maximum likelihood estimate of Pagel's λ (Pagel, 1999).The time-scaled phylogenetic topology we used for our PGLS analyses follows Cooney et al. (2017) and Prum et al. (2015), as performed by Burton et al. (2023).We used the y-axis residuals from these regressions to examine the relative proportions of these features.Thus, where relative sizes are mentioned in the text, they are relative to columellar length.
We observed some variation in, for example, the arrangement of struts around the basal end, as well as the size, location, and number of some fenestrae; however, the presence or absence of these sorts of features does not seem to vary within species.Likewise, the distal and basal ends do not appear to show intraspecific variability in their general shape and the overall form of the columella appears similar within a species.
We measured intraspecific variability (10 males and 10 females) in three species: Branta canadensis, Falco sparverius, and Buteo jamaicensis.Kolmogorov-Smirnov tests were used to determine whether any of the measures deviated from a normal distribution and T-tests were run to assess differences between males and females.No significant deviations from normality were found and no significant differences between sexes were found in any of the species.Among all these tests, the lowest p-value (0.19) was found between males and females of B. canadensis for the maximum diameter of the shaft at the narrowest point.

| RESULTS
In Sections 3.1-3.5,we describe the variation of the columella within and among different clades.In Section 3.6, we provide an overview of the general variability of the columella across all birds examined.broadening of the shaft into a rounded, unfenestrated distal end, but this is relatively undeveloped in all except the only kiwi examined, Apteryx australis.In that species the distal end is one third of the length of the columella compared to 10%-20% in all others, and A. australis is also the only species in which the distal end is broader than the basal.
Casuarius casuarius and Dromaius novaehollandiae share near identical columellae with broad basal ends and prominent struts forming large fenestrae.Struthio camelus is similar but with a wide curved sheet forming a broader basal end covering nearly all of the footplate.In Rhea americana the basal end is slightly less extensive, unfenestrated, and somewhat resembles that of grebes (Podicipedidae; Section 3.3.4)with a ladle-like structure.In contrast to all these larger species where the basal end covers 50%-100% of the footplate, we find much less extensive basal ends in Tinamiformes and A. australis (20% of footplate area).
A prominent feature of all palaeognaths, except Tinamiformes and S. camelus, is the presence of a thicker rim around one half of the footplate, giving it an asymmetrical bulge.It appears as if one edge of the footplate has curled towards the distal end, and in the larger species this joins with the basal end of the shaft.In A. australis the basal end is considerably smaller and does not join at the footplate's edge; nevertheless the footplate still appears to curl laterally, leaving a thicker rim at one side.This is smaller than in the other species.
The columellae in several palaeognaths were described in great detail by Starck (1995).These descriptions largely agree with our own observations except that Starck did not note a bulging footplate in either A. australis or D. novaehollandiae.The ontogeny of the columella of S. camelus was described by Frank and Smit (1976).

| Galloanserae: Galliformes and Anseriformes
The Galloanserae comprise the Galliformes (landfowl, Figure 2 [8-13]) and Anseriformes (waterfowl,).These share similar features in their columellae (Figure 5), which most closely resemble the "primitive" columellar form described by Feduccia (1975a).The shaft is rounded and narrow, with little to no broadening at the basal end.The basal end lacks prominent struts or fenestrae, is centered on the footplate, and in some is almost an unaltered extension of the shaft.The basal ends in Galloanserae have, on average, the smallest relative widths, and all species have a relative basal width below the average for all birds.The ratio of the area of the basal end to that of the footplate in Galloanserae is the lowest of all birds examined.While all Galloanserae share narrow basal ends, Anseriformes differ from Galliformes in exhibiting wider distal ends and smaller, more circular footplates.The distal end is wider than the basal end in over 81% of Anseriformes, but less than half of Galliformes share this trait.The anseriform with the smallest distal/basal ratio is Chauna chavaria (a member of Anhimidae, the sister group to all other Anseriformes) and thus is more similar to the galliform condition, but its columella does not appear particularly similar to either of the two examined Megapodiidae (the sister group to all other Galliformes).
Details of the middle ear structures in Galloanserae have been described in several previous publications (Claes, 2018;Krause, 1901;Mills, 1994;Saunders, 1985).However, none of these offer a comparative study across the clade.comparable in relative size to those of Columbimorphae (Section 3.3.3),but without showing the same eccentricity.All Rallidae except the largest specimen examined, Porphyrio porphyrio, have a simple sheetlike shaft, partially curled down its long axis, and broadening at either end.The distal end is flattened and pointed, and the basal end exhibits a number of mostly unfenestrated struts.P. porphyrio shares the same overall morphology except that it exhibits a rodlike shaft, and a somewhat more extensive basal end.An asymmetrical bulge is present in the footplate of all Rallidae and in the trumpeter Psophia leucoptera, but is absent in the limpkin (Aramidae) and cranes (Gruidae).
All three examined species of Grus, as well as Balearica regulorum, are similar in form to P. porphyrio, but differ in having a more fenestrated basal end.The distal ends in these species are also similar to those of Rallidae in being flattened and pointed.A notable outlier is Psophia leucoptera, in which the distal end appears more rounded, with a distal fossa, and still more complex basal fenestrae.
The columellae of Rallus aquaticus and Fulica atra were both illustrated in Krause (1901), and these illustrations do not differ substantially from our observations.
Musophagiformes all have straight shafts that are more or less centered on a flat footplate.The footplate is slightly bullate in Corythaeola cristata.In both Corythaixoides concolor and Corythaeola cristata the basal end is expanded and joins the footplate via a number of struts with small fenestrae.Musophaginae have all developed relatively thicker shafts than Corythaixoides concolor and Corythaeola cristata, but show variability in the basal end and footplate.In both Musophaga rossae and Musophaga violacea the basal end is larger, less fenestrated, and forms a more cavernous basal end open only on one side, while in Tauraco hartlaubi the basal end is much reduced, and is small and unfenestrated.Despite having the smallest basal end, T. hartlaubi has the largest relative footplate area observed in the clade.
The two otidiforms examined (Ardeotis kori and Otis tarda) have similar columellae with thick, robust, unfenestrated, rod-like shafts, and unfenestrated distal and basal ends.The distal and basal ends are subequal in width, and both have flat footplates.
Cuculiformes exhibit thick, hollow shafts with some of the greatest relative widths of all birds, and extensive basal ends that are also among the largest in relative width across all birds.All Cuculiformes have basal ends that extend to cover the whole of the footplate; a feature which is uncommon among the early diverging neoavian lineages examined (Figure 6), being shared only by some flamingos (Phoenicopteriformes; Section 3.3.4)and the oilbird Steatornis (Section 3.3.6).In Geococcyx californianus and Tapera naevia (both representing Neomorphinae, the sister group to all other cuckoos) the footplate is flat, and the struts of the basal end attach at various points around the rim.In most others, including Guira guira and Crotophaga ani (Crotophaginae), the footplate mostly appeared as a hollow, bulbous structure with a thick rim round its edge, often giving it an asymmetrical, wedge-like appearance.All species examined feature large fenestrae around their basal ends.

| Columbimorphae: Columbiformes and Pterocliformes
Columbiformes (doves and pigeons, Figure 2 [33-37]) and Pterocliformes (sandgrouse, Figure 2 [38]) share a similar columellar morphology.Compared to all other deeply diverging lineages of Neoaves examined, they have relatively large footplates, but their most striking feature is the high degree of eccentricity in their footplates (ranging from 0.81 to 0.88).The degree of eccentricity in Columbiformes is, on average, the greatest of all order-level clades after frogmouths (Podargiformes; Section 3.3.6)and mousebirds (Coliiformes; Section 3.5.3).The basal end is generally flat, curved, and sheet-like.The main exception is seen in the largest species examined, Caloenas nicobarica, where the shaft is more rod-like and the basal end is less sheet-like.All lack fenestrae at either the distal or the basal ends.No substantial differences could be identified between the sandgrouse Pterocles and Columbiformes.
The columella of Pterocliformes has not previously been described, but Columba livia and C. palumbus were included in Krause (1901) and Claes (2018).Neither of these appear different from our specimens.

| Mirandornithes: Phoenicopteriformes and Podicipediformes
Mirandornithes, consisting of Phoenicopteriformes (flamingos, Figure 2 [43, 44]) and Podicipediformes (grebes,Figure 2 [45,46]), are a highly varied group.Most have straight, rod-like shafts and some show slightly bullate footplates.They tend towards having smaller relative footplate areas.The major difference between Phoenicopteriformes and Podicipediformes is the form of the basal end, which is much larger in Phoenicopteriformes, and often extends to cover the entire footplate.
Except for Tachybaptus dominicus, all examined Podicipediformes have a columella and footplate that can be described as "ladle-like."The shaft sits in a position markedly offset from the center of the footplate, appearing near the footplate's edge, while the footplate itself is bulbous and bowl-shaped.The shaft is generally straight, but with a slight kink in Aechmophorus occidentalis.This is entirely absent in T. dominicus where the footplate is flat and the shaft more centered on the footplate.All species have a more rounded and near symmetrical distal end.The basal end is relatively small and unfenestrated, but with some struts emerging.
In Phoenicopteriformes, the distal end is larger, more asymmetrical, pointed, and sometimes fenestrated, while the basal end is more expansive and fenestrated.The basal end covers the entire footplate in Phoenicoparrus jamesi and Phoenicopterus chilensis, but not in Phoenicopterus ruber.In both species of Phoenicopterus the footplate was relatively flat, whereas the footplate in Phoenicoparrus jamesi was notably bulbous and somewhat resembled the ladle-like footplates of Podicipediformes.However, the shaft in Phoenicoparrus jamesi was the most centered of the clade, while it is notably offset in both species of Phoenicopterus (as in Podicipediformes).
The columella of Podiceps cristatus was illustrated by Krause (1901), while Phoenicopterus roseus and Phoenicopterus chilensis were described by Krause (1901) and Claes (2018) respectively.In both studies the columellae were illustrated and described as similar to our specimens.

| Opisthocomiformes (hoatzin)
The columella of Opisthocomus hoazin (Figure 2 [62])has not previously been described.It is fairly simple, having a straight, rod-like shaft, flat footplate with average eccentricity, and a distal end equal in width to the basal end.The entire columella lacks fenestrae.The shaft is centered on the footplate, is relatively thick, and sits at a notable angle with respect to the plane of the footplate.Apodiformes all lack fenestrae at any point on their columellae.They have straight, sheet-like shafts and flat footplates.A curve in the shaft towards the distal end is visible only in Streptoprocne rutila and to a far lesser extent in Collocalia esculenta.The shaft in C. esculenta is very flattened, sheet-like, and has a short, stubby appearance, while in all other species the shaft appears more elongated and thinner.It must be noted that C. esculenta was the smallest swift examined, which may at least partially account for this difference.Also of note was the shaft in the largest species of swift examined, Streptoprocne zonaris.The shaft here has a triangular aspect with an additional strut running along the shaft's length.Several additional Apodiformes were shown by Thomassen et al. ( 2007), and the general morphology of these does not appear to diverge significantly from ours.
The footplate of Podargus strigoides is notable for being among the most eccentric of all measured species.It also has one of the largest footplates relative to columellar length, which makes its eccentricity even more striking.The shaft is offset along the footplate and is both straight and sheet-like.There is a very large distal fossa extending more than one third of the shaft's length.The footplate is slightly wedge-shaped with the thicker part along the footplate's minor axis.
Steatornis caripensis possesses a relatively thick, hollow, tube-like shaft that expands towards the basal end to cover the whole of the footplate, which itself has become slightly bulbous.Among Strisores, such a wide basal end is unique to S. caripensis: in all others the basal/footplate area ratio is never more than 0.5, and is usually much lower.The S. caripensis columella is readily identifiable by the presence of a large fenestra that sits along the major axis of the footplate and runs the entire length of the columella from the basal to the distal end.This feature is present in the drawings by Krause (1901) and Mayr (2003).We have not observed an identical feature in any other species.
A transition from more sheet-like to rod-like shafts is evident across the Caprimulgiformes.The smallest species examined, Uropsalis segmentata, essentially has a similar columella to that of the larger Apodiformes studied.The shaft is thickened in parts but broadened to an unfenestrated sheet-like basal end.As we move to larger species the shafts become rod-like and form larger, more extensive hollow basal ends with a single fenestra.This is different in Chordeiles minor in which the shaft does not close to form a complete tube and thus is open along one side.The shaft is far more tube-like, and the basal end has multiple complex fenestrae in Caprimulgus carolinensis, which is the largest specimen examined.
Aegotheliformes and Nyctibiiformes are not represented in our sample as no specimens with columellae were available for study.Trochilidae were also not measured, due to the difficulties in extracting the columellae owing to their extremely small size.It is likely that the smallest avian columellae are found in this clade, so there remain questions as to how columellar morphology varies at very small sizes.We are not aware of any published studies of the trochilid ear.

| Charadriiformes (shorebirds)
Compared to all other birds, Charadriiformes  and Figure 8) skew towards wider distal ends, narrower basal ends, and smaller than average footplates.In almost all species, the footplates are flat and the shafts are very straight; the only notable exceptions are found in the bullate footplate of the skua Stercorarius pomarinus, and the slightly curved shaft of the thick-knee Burhinus bistriatus.The great majority of charadriiforms studied lack any fenestrae in their columellae: we only noted fenestrae in Rynchops niger, Glareola pratincola, and Rostratula benghalensis within our sample.However, many features of the columella show very divergent patterns between closely related species, and notable similarities with taxa that are more distantly related.Discerning consistent phylogenetic patterns was therefore challenging.
The most conspicuous form among the Charadriiformes is seen within Lari, which diverge morphologically from Charadrii and Scolopaci in having smaller and more circular footplates, basal ends that spread over a larger percentage of the footplate, and distal ends with greater relative widths.Laridae (specifically Larus) and Stercorarius pomarinus stand out in having very wide distal ends (some having among the greatest relative widths among all birds examined).This distal end takes the form of a thickened bony spine with a thin sheet emerging from one side.Larus occidentalis deserves particular mention for its distal end appearing extremely broad (close to 50% the length of its columella).These broad distal ends are characteristic of Laridae, including some, but not all, terns.A similar morphology is also observed in S. pomarinus.Both Haematopus ostralegus and Recurvirostra americana (Charadrii) also share a broadly similar appearance.The broad distal ends are not, however, shared by Alcidae, which instead exhibit some of the smallest relative distal widths among Charadriiformes.The distal ends in Alcidae range from very indistinct from the shaft in Uria aalge, to having notable width and structure in Alca torda.
Scolopacidae share a similar stumpy, sheet-like form to the shaft.There is a noticeable distal and basal end that are both wider than the shaft; however, the shaft itself tends towards the wider side (thus giving the "stumpy" appearance).There is usually a slightly curled aspect to the basal end and often to the distal end and shaft, but compared to many other birds they are remarkably flat.The exception within the group is Tringa flavipes, in which the shaft is notably narrower and the distal end notably wider.T. flavipes is the only representative of this family examined in which the distal end is wider than the basal end (and is significantly so at 1.7 times wider).This large distal end appears symmetrical about the shaft, and there is a symmetrical sheet-like basal end.A similar stumpy columella, with indistinct distal and basal ends, was observed in Anous stolidus (Lari).
Compared to Scolopacidae, Jacanidae have narrower shafts with a thicker spine on one edge.They share the same undeveloped distal and basal ends, but the latter noticeably spreads over a much smaller area of the footplate.
No comparative studies of the columella have examined Charadriiformes as a group, but several species from this clade were illustrated by Krause (1901), and detailed quantitative measurements have been published for Larus and some Alcidae (Anisimov, 2022;Anisimov & Barsova, 2010).

| Phaethoquornithes (Phaethontiformes, Eurypygiformes, and core water birds)
Most of the largest columellae examined were found among Phaethoquornithes (Figure 9), including the seven species with the absolute greatest overall length.As a clade they all tend towards relatively small footplates and relatively wide distal ends, although there are many exceptions: Sphenisciformes have among the narrowest distal ends, while Balaeniceps rex has a very large footplate (the absolute largest footplate area of all birds).Procellariimorphae have narrower distal ends, relatively smaller and more circular footplates, and narrower basal ends, than Pelecanimorphae, and also differ in having unfenestrated columellae.However, exceptions occur in Suliformes and the Pelecanidae, which share relatively small footplates and narrow basal ends.

| Phaethontimorphae: Phaethontiformes and Eurypygiformes
Phaethontimorphae is the clade including the Phaethontiformes (tropicbirds, Figure 10 [1]) and Eurypygiformes (sunbittern and kagu, Figure 10 [2]).The columellar structure of phaethontimorphs has not previously been reported.The three species examined showed very different morphologies, but they are very different in overall size (with Phaethon rubricauda being considerably larger and having a rod-like shaft compared to the shorter sheet-like shaft in Eurypyga helias).The footplate of P. rubricauda has, overall, the smallest relative area of all birds in this study, while in E. helias it is larger than average.The footplate is also notably circular in P. rubricauda, but of average eccentricity in E. helias.Both do share slightly bullate footplates, which are not seen in other Phaethoquornithes.
The basal end in P. rubricauda covers the whole of the footplate via branching struts.The shaft has a slight curve or bend halfway along its length.At the distal end, which remains symmetrical, there is a deep, circular distal fossa.E. helias has a straight sheet-like shaft that appears slightly curled along its long axis.This broadens and flattens towards the distal end.At the basal ends this expands to form a more cavernous structure, but does not extend to cover the whole of the footplate.It instead only covers around 70%, leaving a rim around the edge.In Rhynochetos jubatus, which was only examined from photographs, the shaft is thinner and more rod-like, and expands to form a hollow basal end with a single fenestra.This appears to extend over a considerably smaller area of the footplate than in E. helias.The distal end appears to be highly asymmetrical, with a long pointed process similar to that illustrated for Phoebastria nigripes (Figure 10 [19]).

| Aequornithes (core waterbirds)
Gaviiformes (loons) Both Gaviiformes have an expanded, fenestrated basal end and small, rounded distal end.In Gavia stellata the shaft ran straight but then showed a noticeable curve towards the distal end which was absent in Gavia immer.Both species have rod-like shafts, flat footplates, and shafts that were offset from the footplate's center.The columella of Gavia stellata as illustrated by Krause (1901) appeared similar to our specimen (Figure 10 [3]).
Pelecanimorphae: Ciconiiformes, Pelecaniformes, and Suliformes Ciconiiformes (storks).All Ciconiiformes (Figure 10 [4,5]) have straight shafts, flat footplates, as well as very broad distal and basal ends.As a group they have some of the relatively widest distal ends of all birds, rivaling Alcedinidae and Laridae.These distal ends often contain multiple, sometimes irregular fenestrae.Most species have fenestrated basal ends that cover all or most of the footplate.The basal end of Leptoptilos crumeniferus stands out among the group for its very numerous and complex fenestrae, while all others have only one or two large fenestrae (commonly with one positioned along the major and one along the minor axis of the footplate).The clade is distinctive for having very flattened shafts, particularly in the three examined species of Ciconia.The shaft resembles a flat sheet which is thickened at both edges: this becomes greatly extended towards the distal end.Towards the footplate the sheet begins to curl and forms the basal end with an additional strut emerging to create the second fenestra.The illustration of the columella of Ciconia ciconia by Krause (1901) is similar to the appearance of our specimen.

Suliformes (cormorants, gannets, frigatebirds, and darters).
Suliformes )are notable for having a distinctive boomerang-like curve to their columellae.The feature has been previously mentioned for Morus bassanus (Mills & Zhang, 2006), however we found that Phalacrocorax auritus, Phalacrocorax penicillatus, and Fregata magnificens also possess a similarly curved columella.The specimen of Anhinga anhinga only displayed a very slight curve, and was thus not similar to the others (however, A. anhinga was also the smallest of the suliforms examined).The footplate of Suliformes, as well as exhibiting a relatively small area compared to most birds, is the most circular of all order-level clades examined.M. bassanus has the most circular footplate in the clade (and the most circular of all birds measured)  while A. anhinga was the least circular of the clade (although still far more circular than the average bird).
Pelecaniformes (pelicans, shoebill, hamerkop, ibises, spoonbills, herons, egrets, and bitterns).The columellae of various Pelecaniformes are illustrated in Figure 10 [6][7][8][9][10][11][12][13][14].We found that the columellae of Ardeidae and Threskiornithidae diverge from one another primarily in the relative size of their distal ends.The threskiornithid distal end is narrower than that of Ardeidae, but the basal end remains, on average, similar.Thus, the ratio of the width of the distal end to the basal end is greater in Ardeidae (mean 1) than in Threskiornithidae (mean 0.6).Otherwise, there were very few distinctive differences between the two clades.
Pelecanidae show the most divergent columellar morphology among Pelecaniformes.Both species examined have columellae similar in form, with the shaft being a thick, hollow, tube with multiple fenestrae along its length.The footplates of Pelecanidae have the smallest relative size of all Pelecaniformes by a considerable margin.These small footplates differ dramatically from Balaeniceps rex, which has the largest of all measured species, while Scopus umbretta is intermediate between the two.We see other large differences in the relative width of the shaft, with S. umbretta having one of the widest of all birds, B. rex one of the narrowest, and Pelecanidae being intermediate.The one feature which appears common to all three is a basal end which extends to cover all, or most of the footplate.This is not a common feature in other Pelecaniformes.
Columellar morphology in B. rex was previously described by Feduccia (1977), who also commented on the form in Ardeidae and Threskiornithidae (but not Pelecanidae).The columellae in four species were also illustrated by Krause (1901).None of these appears different from our specimens.
Procellariimorphae: Procellariiformes and Sphenisciformes Procellariiformes (albatrosses, storm petrels, petrels and shearwaters).The columellae of Procellariiformes ) all have relatively small, flat footplates.This is particularly noticeable in Phoebastria nigripes and Hydrobates socorroensis, which themselves have relatively small footplates compared to Procellariidae.The distal end, basal end, and shaft are unfenestrated in most species; however, the basal end in P. nigripes is split into two struts which join the footplate along its major axis.There are thus two fenestrae at each side of the footplate along its minor axes giving it a mammalian stapes-like appearance.A "kink" often occurs along the length of the shaft; this is most significant in Fulmarus glacialis where the kink is significant enough that one could almost describe the shaft as curved.
All species have wider distal than basal ends, with the exception of the two shearwaters (Calonectris diomedea and Puffinus nativitatis).The basal ends in all species are relatively narrow, while the distal ends are relatively wide with the exception of the shearwaters.
To our knowledge there are no previous publications describing the columella in Procellariiformes.
Sphenisciformes (penguins).The columella in all Sphenisciformes (Figure 10 [22-25]) appears as a solid, straight rod.This is especially the case for both species of Aptenodytes.The distal end is relatively narrow, and in Aptenodytes forsteri there barely appears to be any notable structure at the distal end.The basal end consists mainly of a flat, sheet-like strut that emerges from the shaft.In Pygoscelis papua, the solid columellar shaft flattens to become a curved, sheet-like structure which does not completely close to form a hollow basal end.All species have more circular footplates than average.These are particularly circular in Eudyptes chrysocome and P. papua, and in all species the shaft sits over the footplate's center.E. chrysocome in particular stands out for its unusually large footplate and wide basal end, extending to cover its entire surface.The footplate in E. chrysocome is noticeably triangular in form, a feature which is shared somewhat by other Sphenisciformes.
The columella in Spheniscus humboldti as described by Claes (2018) does appear different from our specimen.
3.5 | Telluraves (core land birds) 3.5.1 | Accipitriformes (New World vultures, secretarybird, osprey, hawks and allies) The columellae of the Accipitriformes are illustrated in Figure 11 [1-9].Cathartidae have unfenestrated distal and basal ends, with the distal end being asymmetrical and pointed.Cathartids have some of the most eccentric footplates among all birds, and the most eccentric of all measured bird was found in Vultur gryphus.In Sagittarius serpentarius, Pandion haliaetus, and Accipitridae, we see a transition towards narrower and more rounded distal ends, as well as more complex, fenestrated basal ends (see Figure 12).The columella in many accipitrids is easily identifiable by the presence of a distinctive oval distal fenestra.However, this is absent in Elanus caeruleus and Neophron percnopterus (and Gypaetus barbatus according to the drawings of Krause, 1901).These taxa represent Elaninae and Gypaetinae, which were shown to be deep phylogenetic branches within Accipitridae by sequence-based phylogenetic analyses (Lerner & Mindell, 2005;Mindell et al., 2018).The derived columellar morphology of the other accipitrids examined supports the phylogenetic hypothesis of Elaninae and Gypaetinae falling outside the rest of Accipitridae).
3.5.2| Strigiformes (owls) Krause (1901) first noted the disparity of the columellae of Strigiformes ), which was examined in more detail by Feduccia (1978).As noted by both these authors, the most striking feature of the Strigiformes is the presence of bulbous or bullate footplates in some species (Figure 13).Where present, these are asymmetrical in form, being expanded mostly around one side.This structure does not appear to entirely bulge into the inner ear fluid, but the expanded edge sits against other bony structures (see Figure 13b,c).This structure most prominent in Tyto alba, Phodilus badius, and Strix varia among taxa examined.Feduccia examined multiple species in the genus Strix, and found a great deal of variation in the size of the bullate footplate, including it being essentially flat in Strix nebulosa.We noted completely flat footplates in Bubo scandiaca, Bubo virginianus, Ninox philippensis, and Surnia ulula, while in all other taxa examined (Megascops asio, Otus longicornis, Asio flammeus, Athene cunicularia, and to a lesser extent Glaudicum perlatum) we noticed an asymmetrical bulge.However, this bulge was considerably smaller than that seen in T. alba, P. badius, or S. varia.
The shaft in most strigids is thick and hollow, with indistinct distal ends that often appear barely expanded from the shaft.Across all birds they are second only to the Cuculiformes in the relative thickness of the shaft.In all species this shaft joins to the footplate via a fenestrated basal end, normally made of many branching, spreading struts.The footplates in Strigidae exhibit some of the largest relative areas among Telluraves.T. alba shares a similar basal end and narrow distal end with Strigidae, but differs in having a relatively narrow shaft and smaller footplate area.

| Coraciimorphae: Coliiformes, Trogoniformes, Bucerotiformes, Coraciiformes, and Piciformes
Coraciimorphae (Figure 14) display a curiously diverse range of columellae with very distinctive forms, sometimes exhibiting large differences between closely related clades.The footplates tend towards more eccentric shapes, with only Picidae exhibiting an eccentricity below the average value for all measured birds.The distal ends tend towards being wider than other Telluraves.

Coliiformes (mousebirds)
The columella of Colius colius was figured by Mayr (2003).The two mousebirds, Urocolius macrourus and Colius striatus, appear distinct from one another (Figure 11 [15,16]).C. striatus has a much thicker, sheet-like shaft, compared to the thinner, more rod-like shaft of U. macrourus; however, the columella in C. striatus is also shorter.The shaft in U. macrourus, although overall rod-like, has a narrow groove running along its length and thus appears more as a sheet that has been curled along its long axis to be almost closed into a tube.The footplates of both species are among the most highly eccentric of all birds.

Trogoniformes (trogons)
The distinctive columellae of Trogoniformes ) was first described by Feduccia (1975a).The trogon columella has a large, hollow, bulbous basal end that gives the whole columella a bell-like appearance.This structure makes up about half the length of the columella.The basal end broadens to cover the whole of the footplate, with one large basal fenestra that sits along the footplate's major axis.The shaft has a curve, and most have slightly bullate footplates.
F I G U R E 1 1 (Telluraves): Drawings of the columella in 59 species to illustrate the diversity of forms.These are shown in different orientations so as to illustrate key features.Note that the color of the bones varies with the condition of the specimen and the conditions under which it was observed: no other meaning should be ascribed to these differences.Accipitriformes: 1. Cathartes aura, 2. Vultur gryphus, 3. Sagittarius serpentarius, 4. Elanus caeruleus, 5. Neophron percnopterus, 6.

Bucerotiformes (hornbills, hoopoes, and woodhoopoes)
The suborders Upupi and Buceroti have very different columellar forms ).The columella of Upupi was described as "incus-like" by Feduccia (1975b).It has a sheet-like shaft with a wide distal end that is almost as wide as the columella is long (and all three species examined have the greatest relative distal widths of all measured birds).The distal end has a long pointed process; however, we did not find it to be as dramatically elongated as depicted by Krause (1901).We suspect the difference is that Krause may have mistakenly included a non-bony attachment to the columella in his illustration.Buceroti have straight, rod-like shafts with flat footplates.The basal end itself is fenestrated with multiple struts and trends towards becoming relatively wider with increasing columellar length.The distal end is more pointed than rounded although it appears to barely broaden from the shaft in the largest species.In smaller species the distal end is sheet-like but becomes increasingly thicker as columellar length increases.Buceroti thus show a trend towards relatively narrower distal ends as columellar length increases, thus small species are more Upupi-like than larger species.Buceroti are larger-bodied than Upupi, and all representatives of Buceroti examined in this study have longer columellae.Tockus jacksoni was the smallest hornbill studied and has a columella just under 1 mm longer than Phoeniculus purpureus.
The most distinctive feature is the basal end which is virtually identical in form to that described for Trogoniformes (Section 3.5.3).Like that of trogons, the alcedinid shaft has a visible curve, bending along the footplate's minor axis away from the basal fenestra, but this appears to have a greater curvature than in trogons.A nearly identical columella is present in Merops apiaster (Meropidae), except that its shaft lacks the distinctive curve of Alcedinidae and the distal end is narrower.
The large basal end of Alcedinidae is entirely absent in Coraciidae.The basal/footplate area ratio in the two coraciid species measured was 0.35 (compared to 1 in Alcedinidae and Meropidae).The basal end was slightly expanded from the shaft and contained some small fenestrae.The shafts appear straight and rod-like, and the footplates are flat.The distal end is similarly narrow to that of Meropidae and is still narrower than the basal end.
Momotidae were not sampled in this study due to difficulties in extracting the columella.Although multiple specimens were available for study, the shaft of the momotid columella emerges through an opening which was too narrow to bring the footplate through.Thus, the columella could not be extracted intact without damaging the skull.That being said, the motmot columella does appear identical in form to that of the Alcedinidae, excepting perhaps a smaller basal fenestra (as indicated in the drawing of Momotus momota in Mayr, 2003).
Piciformes (woodpeckers, barbets, honeyguides, puffbirds, jacamars, toucans) Picidae ) all display a similar, very unique columellar form.The shaft is curled into a tube which is closed at both ends, leaving one large fenestra along the length of the shaft (similar in some respects to the shaft in Steatornis caripensis; Section 3.3.6).The basal end in Picoides villosus is a hollow, bulging structure.This form is not shared by any other Piciformes.
Lybiidae all have sheet-like shafts that expand to form very wide distal ends, while simultaneously having relatively narrow basal ends; this gives Lybiidae the largest distal/basal ratios within Piciformes.The only other piciforms with wider distal than basal ends are the honeyguide Indicator variegatus and the puffbird Bucco macrodactylus (the distal and basal ends have equal width in Notharchus ordii).In these species the shaft is more rod-like but still has the expansive distal end.
Ramphastidae stand apart from other Piciformes in having footplates with greater relative widths, as well as wider, more extensive basal ends that extend to join the footplate at or near to its rim.Similarly wide basal ends were also seen in the jacamar Jacamerops aureus and the flicker Colaptes auratus.The basal end of the columella in Ramphastidae joins to the footplate at or near its rim via struts and leaves multiple fenestrae.The distal end is expanded from the shaft, mostly symmetrical and contains some fenestrae.Capito aurovirens is similar except that it is small, has an unfenestrated distal end, and larger fenestrae around the basal end.
Picus viridis and Ramphastos vitellinus were both illustrated by Krause (1901).While both columellae appeared similar to ours, the distal end of P. viridis was more complex than any of the four genera within the Picidae that we examined (all of which appeared relatively smaller and more rounded).

Cariamiformes (seriemas)
The columella of Cariamiformes has not previously been described.Here, we were only able to examine the columella of Cariama cristata (Figure 11 [38]).C. cristata has a relatively simple columella with a flat footplate, rodlike shaft, and unfenestrated distal and basal ends.The basal end is formed by a number of struts emerging from the shaft.Other measures such as the distal/basal ratio and footplate eccentricity are close to the average measured across all species.The shaft in C. cristata appears slightly more offset from the center of the footplate than in the average bird.

Falconiformes (falcons)
In Falconiformes (Figure 11 [39,40]), shafts were rod-like and straight, with footplates that are only slightly bullate.The shafts are thicker on average than in other members of Australaves.The distal end is unfenestrated and lacking in fenestrae, while the basal end is more extensive and expands to join the footplate with a hollow structure that has multiple struts and fenestrae.The basal end expands over the whole footplate in Falco peregrinus, F. sparverius and Polihierax semitorquatus, but not in any other; it remains similar in form across all species.The columella in three species of Falco have previously been reported (Claes, 2018;Krause, 1901); these do not appear to differ substantially from specimens we examined.

Psittaciformes (parrots)
Common features of Psittaciformes ) appear to be relatively circular footplates, narrow distal ends and strutted, fenestrated basal ends.Their basal ends and footplates have, on average, the smallest widths and areas of the examined representatives of Telluraves, and the footplates are some of the most circular of all birds.All have straight shafts although a slight kink was observable in Forpus passerinus.Footplates are generally flat, but a slight bulge can be seen in the largest species.The distal end is normally fairly symmetrical but in a few species, notably Ara macao and Deroptyus accipitrinus, it was more pointed and asymmetrical.In the largest species, Anodorhynchus hyacinthinus, we see a very long, straight, hollow, rod-like shaft, which has relatively narrow distal and basal ends containing multiple fenestrae.Four species were examined by Krause (1901), and Melopsittacus undulatus by Saunders (1985); the overall morphology of their columellae does not appear vastly different to the range of forms we observed.

Passeriformes (passerines)
The columellae of various passeriform taxa were studied by Feduccia (1974Feduccia ( , 1975aFeduccia ( , 1975cFeduccia ( , 1976) ) and by Krause (1901).In our study, Passeriformes ) are the most heavily sampled clade (102 species), but this represents only a very small percentage of their total diversity.Passerine columellar morphology is very diverse and, given that we have few representatives of any individual family, forming general conclusions about the phylogenetic distribution of forms across Passeriformes is difficult.We will therefore not discuss the clade in much detail.
As discussed by Feduccia (1975aFeduccia ( , 1975c)), the basal region in Tyranni is a hollow, bulbous structure similar to Alcedinidae and allies.However, the columella of Tyranni can be distinguished by the position of the basal fenestra, which sits along the footplate's minor axis as opposed to its major axis, and does not stretch to join to the footplate's rim.The shaft in these species is also straighter.The suboscines are the only clade within Passeriformes in which any species has a basal end that covers the whole of the footplate.The ratio of the distal/ basal width is lower in Tyranni than in Passeri; this appears to be a combination of Tyranni all exhibiting relatively wider basal ends, and also exhibiting relatively narrower distal ends on average.
Columellae from Corvides tend to be larger than those we examined from Passerides.The latter have stumpier, sheet-like shafts while the former are more slender and rod-like.Tangara cayana and Sporopipes frontalis stand out for having particularly wide distal ends, which in T. cayana also contained a large opening.
One of the most curious findings from all the passerines is the presence of a distinctly curved shaft in Cinclus mexicanus, which resembles that seen in the Suliformes; to our knowledge this has not been previously reported.The shaft in C. mexicanus also appears very offset from the footplate's center and appears to be one of the most offset of all species examined.
The shaft in different avian species appears to be either flattened and sheet-like, or more rounded and rodlike.Sheet-like shafts are never perfectly flattened; instead, they are always partially folded or curled down their long axes.The extent of this varies, and in some instances it becomes almost like a tube.Certain species possess columellae that sit somewhere between rod-and sheet-like forms.Whether the shaft is rod-like or sheetlike seems to depend on the columella's absolute length: sheet-like shafts dominate among those under 2 mm, and rod-like shafts dominate for longer columellar lengths (Figure 15b).A one-way ANOVA showed a significant difference between the mean lengths of the three columellar shaft types (P < 0.0001), although it should be noted that the classification of columellae into these three categories was somewhat subjective.
In most instances the shaft appears "twisted," as the major axis of the basal end and that of the distal end do not lie parallel to one-another.This is most striking and obvious in some species with sheet-like shafts.The cases where the major axes of the basal and distal ends do not align is more common than the case where they do align.The major axis of the basal end most often appears to align with the major axis of the footplate, while in almost no cases do the major axes of the distal end and that of the footplate align.
The shaft in most cases sits perpendicular to the plane of the footplate's tympanic side and appears to be relatively straight.In a few examples we find it sitting at a prominent angle, but there appears to be no pattern to the variation in this angle.The angle is greatest in Morus bassanus, where it leaves the footplate at an angle of around 45 .

| The distal and basal ends
At both ends, the shaft normally becomes noticeably broader, forming distinct distal and basal structures.There is appreciable variation in the size of both the distal and basal ends.In most species the basal end is wider than the distal end (64.3% of species in our sample have a wider basal end, 32.5% have a wider distal end, and the remaining species were measured with equal width distal and basal ends), while those with wider distal ends are restricted to a few clades.The major clades generally characterized by wider distal ends are Ciconiiformes (86%), Anseriformes (81%), Bucerotiformes (75%), Procellariiformes (71%), Suliformes (60%), and Charadriiformes (53%).This condition was also observed in Apteryx australis and Phaethon rubricauda.
Where and how the basal end joins the footplate varies considerably.Many species have very wide basal ends that extend to join the footplates at its edge, thus covering the footplate's entire surface.We find examples of this in species covering almost the entire range of footplate areas and columellar lengths.In others we find much narrower basal ends that join the footplate near the center or offset to one side.The distribution of values for this ratio is shown in Figure 15c.The mean ratio is 0.42 (standard deviation: 0.31).Having a basal end that covers the whole of the footplate is a trait found in Alcedinidae, Cuculiformes, Meropidae, Pelecanidae, Phaethontiformes, Phoeniculiformes, Ramphastidae, Steatornithiformes, Struthioniformes, Trogoniformes, and Tyranni.
The maximum width of the distal end varied from 0.24 mm (Charadrius montanus) to 2.84 mm (Leptoptilos crumeniferus), and the maximum width of the basal end varied from 0.16 mm (Dicaeum cruentatum) to 3.01 mm (Balaeniceps rex).There is a significant ( p < 0.0001), positive correlation between the columellar length and the widths of both the distal and basal ends, as well as the minimum width of the shaft (Figure 15a).For the distal end, the OLS regression results in a line with the equation y = 0.41x 0.56 (R 2 = 0.45, P < 0.0001), and PGLS regression gives y = 0.33x 0.67 (R 2 = 0.48, P < 0.0001, λ = 0.72).For the basal end, the OLS regression results in a line with the equation y = 0.5x 0.60 (R 2 = 0.48, P < 0.0001), and PGLS regression gives y = 0.5x 0.58 (R 2 = 0.45, P < 0.0001, λ = 0.79).
As well as the maximum width, we also measured the minimum width of both the distal and basal ends.The mean ratio of the maximum to minimum diameter of the distal end is 3.81 (standard deviation: 1.76); this ranges from 1.2 (Odontophorus stellatus and Aptenodytes forsteri) to 10.4 (Lybius vieilloti).For the basal end the mean is 2.05 (standard deviation: 0.93), with the smallest ratio being 1 (Cereopsis novaehollandiae) and the largest being 7 (Gallinago gallinago).There is more variation in the distal end than in the basal end, with the distal ends tending to be much flatter and the basal ends more rounded.

| The footplate
The area of the footplate in our sample ranges from 0.16 mm 2 (Dicaeum cruentatum) to 7.16 mm 2 (Balaeniceps rex).There is a significant correlation between footplate area and columellar length (Figure 15d).The OLS regression results in a line with the equation y = 0.40x 0.95 (R 2 = 0.74, P < 0.0001), while the PGLS regression results in a line with the equation y = 0.43x 0.94 (R 2 = 0.7, P < 0.0001, λ = 0.78).The footplate is not circular but elliptical, with an eccentricity that ranges from 0.32 (Morus bassanus) to 0.91 (Vultur gryphus).The mean eccentricity is 0.71 (standard deviation: 0.10).Across all birds, eccentricity does not appear to be related to other traits.Differences in the degree of eccentricity do, however, appear between different clades.We ran a Mann-Whitney U test to compare the eccentricity of each order-level clade with the whole population minus the clade in question.The following clades showed significantly (P < 0.05) more circular footplates: Anseriformes, Suliformes, Piciformes, Psittaciformes, Sphenisciformes, and Procellariiformes; by contrast, significantly more eccentric footplates appear in Bucerotiformes, Columbiformes, Coliiformes, and Cuculiformes.The footplate's surface is never completely flat, but instead we find ridges and valleys in various configurations (see Krause, 1901 for further descriptions of these F I G U R E 1 5 Variation in some of the quantitative measurements across species.Figure 15a shows the maximum widths of the distal (magenta) and basal (cyan) ends as well as the minimum widths of the shaft (green), plotted against columellar length on a logarithmic scale.The solid lines show the results of the ordinary least squares regressions and the dashed lines the PGLS regressions.Figure 15b shows a histogram illustrating the relationship between columellar length and the form of the shaft: sheet-like (blue), rod-like (red) or somewhere intermediate (rodsheet, yellow).The figure shows the number of species with each shaft-type, divided into bins based on columellar length, and plotted on a logarithmic scale.Translucency in the colors is used for bins containing multiple columellar types.Figure 15c shows a histogram illustrating the number of species with different basal/ footplate area ratios.Figure 15d plots footplate area against columellar length on logarithmic scales.The results of the regression analyses are shown as a solid black line for OLS and a dashed line for PGLS.features).In several taxa the perilymphatic side of the footplate has a more convex appearance, bulging into the inner ear space, while in others the opposite occurs, and the footplate has a more concave appearance.The former is more common and is characteristic of certain clades (in particular Strigiformes and Cuculiformes), while the latter does not appear diagnostic of any clade in particular.While in some species this footplate bulges out from its center to become bullate in shape, in others it appears that only one edge of the footplate has expanded, creating a thicker rim at one side and an asymmetrical bulge.In all cases, the bullate footplate is hollow and fenestrated.
The footplates of most birds appear to be thin even at the edge, lacking the thick labrum seen in mammals.There are a few exceptions, such as in some Palaegnoathae, Cuculiformes, and Strigiformes mentioned in the preceding paragraphs, where the footplate does thicker (although often only around one part).As our specimens were all taken from dried skulls, we are unable to comment on the structure of the annular ligament and its relationship with the thickness of the footplate's rim.

| DISCUSSION
4.1 | The effects of allometry Krause (1901) noted that very differently sized birds may have columellae of similar dimensions.In our study, the largest bird (Struthio camelus) has only the 11th longest columella, with longer columellae appearing to be characteristic of Phaethoquornithes.An investigation on scaling of the avian middle ear showed that, while most auditory structures show negative allometry with body and head mass, columellar length shows positive allometry (Zeyl et al., 2023).The same study also reported that head mass in birds shows negative allometry with body mass.Thus, larger birds have relatively smaller heads, but larger heads have relatively longer columellae.An isometric relationship was reported between columellar length and interaural distance (Peacock, Spellman, Greene, & Tollin, 2020), with Pelecanus erythrorhynchos being an outlier for having unusually long columellae.
A larger, wider head will require a longer columella to span the gap from the tympanic membrane and extracolumella to the oval window.However, various other factors such as the volume of the brain, the orientation of the tympanic membrane and the length of the extracolumella may influence the size of this gap.Further quantitative analyses will be necessary to better reveal the factors that influence columellar length and other features of middle ear morphology.

| Phylogenetic implications
As we have shown in the preceding sections, and as noted by earlier authors (Feduccia, 1975a(Feduccia, , 1975b(Feduccia, , 1977(Feduccia, , 1978;;Krause, 1901), the columella shows significant variation within closely related taxa.However, even though it can be difficult to identify derived morphologies that characterize particular clades, some features appear to carry a clear phylogenetic signal.Feduccia (1975a) hypothesized a columellar morphology with a rod-like shaft and narrow basal end, centered on a flat footplate, to be plesiomorphic for neornithine birds.We found that this description most closely resembles the columella of Galloanserae, which have a narrow shaft and basal end that are distinctive enough to diagnose the clade.Although Feduccia (1975a) used the columella of Apteryx to illustrate what he took to be the primitive condition for extant birds, this form is not characteristic of all Palaeognathae, especially with regards to differences in the form of the basal end in the larger species.That being said, we agree with Feduccia that a columella with a rod-like shaft centered on a simple, flat footplate is likely to be plesiomorphic for neornithine birds.These features appear common among non-avian archosaurs (Colbert & Ostrom, 1958;Colbert & Russell, 1969), and are found in multiple groups of early diverging crown birds (Galloanserae and Apteryx as already mentioned, as well as Tinamiformes), as well as in phylogenetically disparate neognathous clades that are deeply nested within Neornithes.Morphologies that diverge from this form tend to show some phylogenetic signal.It is less straightforward to characterize major clades within Neornithes based on columellar morphology and there appears to have been much homoplasy in columellar morphology, with distinctive morphologies having evolved multiple times independently.
Cuculiformes exhibit a number of distinctive derived morphologies, including very thickened shafts and large basal ends, that are atypical of their close relatives and instead more closely resemble the columellae of Telluraves.The phylogenetic relationships of cuckoos have long been disputed.Different studies have placed them within Otidimorphae, as the sister group of Musophagiformes (Jarvis et al., 2014) or Otidiformes (Prum et al., 2015), or as the sister group of Columbiformes (Kuhl et al., 2021).The columella of Pterocles exustus is indistinguishable from that of the Columbiformes, thus if the hypothesis of Kuhl et al. (2021) is correct, it would seem likely that P. exustus and Columbiformes retain the plesiomorphic columbimorph columella, and the columellar morphology of the Cuculiformes is autapomorphic.The columella in Cuculiformes displays features that are not shared by Musophagiformes and Otidiformes, and overall it does not immediately suggest affinities with either.
Fregatidae, Sulidae, and Phalacrocoracidae share a derived morphology which is absent in Anhingidae.This conflicts with current phylogenies, which support a sister group relationship between the Phalacrocoracidae and Anhingidae.Among Aequornithes, straight shafts are likely to represent the plesiomorphic condition, and curved shafts are apomorphic for Suliformes.Only very slight curves were seen in other Aequornithes, notably in the derived columellae of Pelecanidae.Accordingly, we hypothesize that the straightened shaft of Anhinga anhinga represents an autapomorphic reversal towards the primitive condition.
We for the first time identify a derived type of columella, which characterizes a subclade of Accipitridae.This morphology, with a distinctive oval distal fenestra, is found in all studied accipitrids except Elaninae and Gypaetinae, which are hypothesized to represent deeply diverging accipitrid lineages in sequence-based phylogenetic analyses (Lerner & Mindell, 2005;Mindell et al., 2018), outside the clade formed by the remaining accipitrids examined.This columellar morphology is present among accipitrids with widely differing ecologies, and thus appears to represent a synapomorphy of this major accipitrid subclade.
The derived columellar form of Upupiformes was described by Krause (1901) and Feduccia (1975aFeduccia ( , 1975b)), and was used by Feduccia as evidence of common ancestry between the Upupidae and Phoeniculidae (a conclusion that remains widely supported).The functional significance of this morphology remains elusive in light of the disparate ecologies of extant Phoeniculidae and Upupidae.However, we note that upupiform birds are adapted to gaping, that is, opening of the beak within a substrate.As an adaptation to this feeding behavior, the caudal portion of their mandible exhibits a long retroarticular process and other peculiarities such as specializations of the jaw musculature (see e.g., Burton, 1984, for some anatomical descriptions).We hypothesize that changes in skull geometry necessary to accommodate their gaping behavior may have led to the evolution of their derived columellar shape.The absence of this morphology in Buceroti, which is the sister taxon of the Upupi and lack this specialized morphology of the jaw apparatus, adds weight to a possible link between mandibular function and columella morphology.
As first detailed by Feduccia (1975a), a characteristic derived columellar morphology with a bulbous basal end characterizes Trogoniformes as well as Meropidae, Alcedinidae, Todidae, and Momotidae (Figure 14).This morphology was taken as evidence for close affinities of these taxa by Feduccia (1975a).At that time, the interrelationships of neornithine birds were poorly resolved, and the assumption of close affinities of trogons and kingfishers, bee-eaters, and allies was considered reasonable.However, according to sequence-based phylogenies (Kuhl et al., 2021;Prum et al., 2015), the bullate columellae of Trogoniformes and those of the aforementioned coraciiform lineages are likely the result of homoplasy.Recent phylogenetic studies reject a close phylogenetic relationship between trogons and coraciiforms, and also support a topology within Coraciiformes whereby Alcedinidae, Todidae, and Momotidae form a clade sister to Meropidae and Coracii (which do not exhibit a bulbous basal end of the columella).Thus, a considerable degree of convergence in the morphology of the basal end of the columella is required to explain the observed patterns in these groups: if a bullate columella characterized the last common ancestor of a trogon + Coraciiformes clade, it must have been lost independently in Bucerotiformes, Coracii (Coraciidae and Brachypteraciidae), and Piciformes.Alternatively, it may have evolved multiple times independently in trogons and Coraciiformes.The differing conditions in Meropidae and Coracii imply either an additional acquisition of this distinctive morphology in Meropidae, or a reversal of this condition in Coracii.
For trogons, kingfishers, and allies, multiple convergent origins of the derived bullate columella may be more likely than a reversal to the primitive condition in some taxa.Because Trogoniformes, Meropidae, Alcedinidae, Todidae, and Momotidae are closely associated in current phylogenies, their similar columellae may be the result of parallel evolution owing to shared soft tissue structures or other anatomical or physiological features that may have facilitated the convergent acquisition of similar columellar morphologies.We note another apparent example of parallel origins of similar derived columellar morphologies in Strigiformes.In owls, the derived columellar type, with a markedly bulging footplate, almost certainly evolved at least twice independently: once in Tytonidae (Tyto and Phodilus) and once in the strigid taxon Strix (Feduccia, 1978).
Passeriformes exhibit great variation in columellar morphology, and Feduccia (1975a) already noted that Tyranni are characterized by a distinctive bullate columella.Feduccia (1975a) used the similarity between the basal end of the columella in Tyranni and Alcedinidae to suggest that it was "probable or possible that the alcediniform birds and suboscines are monophyletic."However, the monophyly of Passeriformes is well established by other morphological and molecular evidence, and the morphological similarity between the Tyranni, trogons, as well as kingfishers and allies (the "Alcediniformes," sensu Feduccia, 1975a), is thus most likely convergent.
At the time of Feduccia's pioneering studies, the interrelationships of passerines were poorly resolved.This is no longer the case, with numerous well-supported phylogenies having been published in recent decades (e.g., Barker et al., 2002;Ericson et al., 2002;Oliveros et al., 2019).All of these studies supported monophyly of the Tyranni (the suboscines), and the derived morphology of the columella clearly represents a synapomorphy of these birds.As noted in Section 3.5.4,columella morphology in the Tyranni is readily distinguishable from that of trogons, kingfishers and allies by its differently positioned basal fenestra.However, some similar convergent forms have appeared in the closely related Falconiformes, as well as in the more distantly related Cuculiformes and the oilbird Steatornis.
Sequence-based analyses have also shown that Acanthisittidae is the sister taxon of all other extant passerines.The morphology of the columella of acanthisittids was described by Feduccia (1975c) and may represent the primitive columellar type for passerines.Similar columellae also occur in early diverging taxa of Passeri (oscines, the sister group of Tyranni) such as Menuridae, Corvidae, and Laniidae.Within oscines, various taxa exhibit derived columellar morphologies, which may characterize particular oscine subclades.For example, the columellae of Tangara cayana, in which the tip of the very wide shaft bears a large opening, and Cinclus mexicanus, in which the shaft has a boomerang-like curve, exemplify autapomorphic variation in columella morphology within this hyperdiverse passerine subclade.A more thorough study of Passeriformes will be necessary to better characterize columellar diversity within the clade and identify apomorphies of subclades.

| Functional significance
The goal of this study was to provide a survey of columellar morphology; the nature of the morphometric data collected precludes a thorough, detailed functional analysis.This is particularly the case since the data fails to capture features such as curved shafts and bullate footplates that may carry functional significance.A geometric morphometrics approach would capture the anatomical morphospace more fully, provided homologous features can be identified, and could additionally take into account the morphology of functionally important soft tissue structures such as the extracolumella.No definite functional significance has been demonstrated for any of the columella's variable features, nor for any of the derived morphologies previously described.Nevertheless, we provide some comments on possible function in this section.Mills and Zhang (2006) speculated that the distinctive columella in Morus bassanus might be adapted to prevent trauma from plunge diving and from diving to great depths.It is notable that the Alcediniade, which are also plunge-diving species, possess curved shafts, and the solitary curved shaft noted in Passeriformes was found in Cinclus (dippers), the only group of aquatic diving and swimming passerines-although dippers do not submerge to any significant depth.However, Trogoniformes also share shafts that curve similarly to those in the Alcedinidae (although not to quite as extreme a degree), while the shaft in Anhinga anhinga has only a very slight curve, and Sphenisciformes have very straight shafts.If the shaft is straight between the distal and basal ends, we would expect a large static force applied to the distal end to be directed straight to the oval window; however, if the shaft is curved this may be directed towards a rocking rather than a piston-like footplate movement, or flexing of the bone.Whether and how columellae with curved shafts deform under static pressure loads has not been examined experimentally, but could potentially be imaged with CT data as has been done for other species (e.g., Claes, Muyshondt, Dirckx, & Aerts, 2018).Various adaptations to prevent barotrauma are known to be present in the ears of diving mammals such as the hooded seal (Cystophora cristata, Stenfors et al., 2001), and it is likely that diving birds also show adaptations in ear structures besides the ossicles.
Similar deformation studies may also help explore the functional significance of the columella's overall length on shaft shape.As seen in Figure 15b, the shafts of smaller columellae tend to resemble curled sheets, while those of longer columellae tend to resemble rods.The shape of the shaft may well be optimized to prevent twisting and bending while simultaneously minimizing mass.The presence of many fenestrae, particularly in the larger specimens, and larger fenestrae in those with larger basal ends may also be related to mass reduction.The mass of the columella will influence the resonance frequency of the system, with a lower mass shifting the resonance higher, while the reduction in inertia may also augment higher frequency sound transmission (Mason, 2016).
A wider distal end of the shaft in principle presents a greater surface area for extracolumellar attachment.It is unclear if the connection with the cartilaginous extracolumella does vary with the distal end's width: most of our specimens were from skeletons which no longer had an extracolumella preserved.However, if the two are correlated then a relatively wide distal end could imply a more robust attachment to the extracolumella and, since a wider attachment is also flatter, it may also work to restrict motion along its major axis, thus changing the rotational motion of the system.This could then alter the manner in which energy is transferred to the columella, as well as the manner in which the ossicular system both bends and provides mechanical advantage.Various studies have examined the behavior of the columella-extracolumella joint via manipulations and by applying static pressure loads (see e.g., Claes, Muyshondt, Van Assche, et al., 2018;Mills & Zhang, 2006;Norberg, 1978;Starck, 1995).Further such studies, via measurements and modeling, would be valuable.
The structure of the basal end of the shaft will determine how a force applied to the distal end is distributed across the footplate.In most mammals, the force is directed to the footplate via two crura near to the footplate's edge, and thus the force is applied close to where the annular ligament is attached.Fleischer (1978) describes this as providing greater stability compared to the avian condition where the force is directed "more or less on the center of the clipeolus."As described in the preceding sections, this morphology (where the shaft has a narrow basal end that sits in the center of the footplate) is not common to all birds.Whether or not directing the force to the rim, as opposed to the center, results in a different type of footplate motion has not been examined experimentally.Some measurements of the threedimensional motion of the columella have been described, but only in a limited number of species (e.g., Struthio in Muyshondt et al., 2018).Thus, any effect of basal end morphology on mechanics remains essentially unexplored.
As noted in Section 3.6.3,the rim of the footplate in birds appears to be much thinner than that of the mammalian stapes (with a few exceptions including, e.g., Palaeognathae, Cuculiformes, and Strigiformes).It is at the footplate's rim that the annular ligament is attached; thus, a thinner rim may be indicative of a thinner ligament.This is consistent with studies showing larger footplate displacements in birds as compared to mammals (Claes, Muyshondt, Van Assche, et al., 2018).Experimental studies of middle ear transfer functions in a range of birds (Peacock, Spellman, Tollin, & Greene, 2020) suggest that the response of the avian columella to a given sound pressure may also prove to be greater than the equivalent stapes response in mammals (e.g., gerbils, Ravicz et al., 2008;chinchillas, Ravicz & Rosowski, 2013;humans, Aibara et al., 2001).Other studies have noted the asymmetrical nature of the annular ligament in some species, reporting that it is broadest at its anterior edge (Gaudin, 1968), which could influence the mode of footplate motion.
The occurrence of distinctive bulbous basal ends in the passerine Tyranni (suboscines) is of particular interest as recent studies have described differences in the frequency range of vocalizations in the clade compared to other passerines (Goller et al., 2021).It is therefore possible that the derived morphology of the columella in these birds is related to a particular hearing range.The occurrence of similar columellae in Trogoniformes, Alcedinidae, Meropidae, Momotidae, and Todidae may provide a basis to test this hypothesis.
However, given the ecomorphological disparity of suboscines, it also remains possible that their distinctive columellar morphology is unrelated to any specific functional specializations.The columella of Steatornis caripensis shares some similarities in having a bulbous form, which differs markedly from all other Strisores.S. caripensis is well known as an auditory specialist due its unique echolocation capabilities (Konishi & Knudsen, 1979;Suthers & Hector, 1985); however, previous studies have found no evidence of any special adaptations in the middle ear of the echolocating species of Apodidae as compared to non-echolocating species (Thomassen et al., 2007).
Studies in mammals have failed to reveal any obvious correlation between bullate footplates and "any particular habit or ecology" (S anchez-Villagra & Nummela, 2001).In birds, the bullate footplate in Tyto alba was first noted by Krause (1901), who hypothesized that the convex portion may provide an explanation for the barn owl's very sensitive hearing since it is able to "emit oscillations radially," 1 while Schwartzkopff (1955) suggested that the owl footplate worked to prevent the formation of eddies within the fluid of the inner ear.However, these conclusions were not made with reference to any measurements or modeling.Whether the bullate footplate is an adaptation to improve hearing, or evolved as a result of other changes to skull morphology, remains unknown.
Aquatic birds (Anseriformes, Charadriiformes, Mirandornithes, and Phaethoquornithes) all appear to have footplates that are small in area relative to columellar length.We see further divergence to still smaller relative footplate areas in Pelecanidae, Phaethontiformes, Procellariiformes, Sphenisciformes, Suliformes, and the charadriiform suborder Lari.These clades contain many marine species, and they dominate among those with the smallest relative footplate areas; a finding which is not consistent with Schwartzkopff's (1968) contention that diving species generally have larger footplates (this was given, however, in reference to the middle ear area ratio [tympanic membrane/footplate area], which was not examined in this study).The functional significance of this, if there is one, may be related to differences in hearing function in aquatic species potentially including adaptation to underwater hearing, or from the general acoustic environment in aquatic habitats.Marine environments in particular, with wide open spaces and constant acoustic background noise (surf and microbaroms), place different demands on hearing than terrestrial environments.The relationship between the size of the footplate and the oval window is unknown, and any adaptive function to smaller footplates may lie in concomitant changes to size and flexibility of the annular ligament.
The columella needs to bridge the gap between the tympanic membrane/extracolumella and the oval window, and thus much of the columella's variation could be due to variation in the overall morphology of the ear region of the skull.Different skull morphologies may demand changes in the columella's length and orientation and it remains possible that much of its overall form is not immediately related to auditory capabilities but constrained by geometric properties of the skull, reflecting a level of morphological integration within the avian skull that has repeatedly been noted to influence avian endocranial morphology (Chiappe et al., 2022;Felice & Goswami, 2018).The large variation between species may be indicative of some features having no constraining physiological function.

| Translational applications
As has been noted by several authors (e.g., Arechvo et al., 2013;Manley, 2021;Mills, 1994), the avian columella resembles prostheses used in reconstructing the human middle ear in cases of conductive hearing loss.However, these prostheses are very simple in appearance, being somewhat similar to Feduccia's primitive columellar form (Feduccia, 1975a).If any of the diverse morphologies we see in the avian columella carry functional significance, then continued exploration of the functional morphology of the avian columella may prove profitable in improving the design of middle ear prostheses.

| Limitations of this study
In this study, we could only present a cursory analysis of intraspecific variability including sexual dimorphism, while we were unable to examine changes in the columella due to ontogeny.Although we cannot be certain that all of our specimens were full adults, studies of the ontogeny of the columella in Gallus gallus have shown the columella to be fully developed at 74 days (Cohen et al., 1992).Thus, if similar developmental patterns hold for other birds, then the columella becomes adult size relatively early in development.Another study has reported sex differences in columellar morphology in G. gallus (Claes et al., 2017).This study found longer columellae in males than in females, but did not report other significant morphological differences between sexes.That being said, the example male and female columellae illustrated do not appear identical, particularly in the basal end.
We have worked on the assumption that intraspecific variability in columellar morphology is small compared to interspecific variability.Significantly, closely-related species tend to show similar columellar features, with particular morphologies being diagnostic of certain clades, which is consistent with the notion that intraspecific variability is relatively small.Comparable studies to ours, which have looked at the scaling of middle ear morphology across a wide range of mammal or bird species, have also used very small intraspecific sample sizes (e.g., Nummela, 1995;Nummela & S anchez-Villagra, 2006;Zeyl et al., 2023), so their findings have been based on the same assumption.Differences in columellar morphology based on age and sex are, however, certainly worthy of further study.

| CONCLUSIONS
We examined the morphology of the avian columella in 401 phylogenetically widespread species, and described how this varies within and between clades.Phylogenetic signal of columella morphology was noted in several clades with disparate morphologies; however, we also find that morphology of the columella exhibits considerable homoplasy, and significant morphological differences exist between closely related clades.The functional significance of variations in columellar structure in birds remains largely unknown; many of the suggestions that may be found here and in other morphological studies require experimental investigation.

F
I G U R E 1 2 Left is the phylogeny of Accipitiformes following Mindell et al. (2018), with columellar types of higher-level taxa shown.The columellae are from Vultur gryphus (Cathartidae), Sagittarius serpentarius (Sagittariidae), Neophron percnopterus (Gypaetinae), Trigonoceps occipitalis (Aegypiinae), Aquila chrysaetos (Aquilinae), and Buteo swainsoni (Buteoninae).Right shows V. gryphus, S. serpentarius, N. percnopterus, and T. occipitalis with the differences in the morphology labeled.F I G U R E 1 3 (a) Shows the phylogeny of Strigiformes following Wink and Sauer-Gürth (2021), with columellar types of higher-level taxa shown.The columellae are from Tyto alba (Tytonidae), Otus longicornis, Bubo scandiaca, and Strix varia.(b) and (c) are drawings of the columella of Tyto alba at perpendicular angles to illustrate how it sits with the surrounding bone; the columella is cut through the midline of the shaft in each case.(b) shows the columella from a more medial perspective, while (c) is depicted from a more dorsal view.While the shaft sits perpendicular to the tympanic side of the footplate, it makes an angle of 50 with the major part of the footplate's perilymphatic side.