New fossil mousebird (Aves: Coliiformes) with feather preservation provides insight into the ecological diversity of an Eocene North American avifauna

Coliiformes (mousebirds) are represented by just six extant species. These species, restricted to sub-Saharan Africa, are all primarily frugivorous and are among the most sedentary of living birds. Previously described fossil Coliiformes preserving feather traces share the short, rounded wing shape of extant mousebirds. Along with osteological evidence, these observations have been proposed to support poor sustained ﬂight capabilities across the stem mousebird lineage. We report a new species of Coliiformes from the early Eocene (51.66 ± 0.09 Ma) Fossil Butte Member of the Green River Formation, represented by one of the comparatively few fossils from these deposits preserving carbonized traces of the wing and tail feathering. Feather traces indicate an elongate, tapering wing shape similar to that of some extant aerial insectivores, and suggestive of a capacity for sustained and agile open-air ﬂight. Traces of the rectrices reveal the tail accounted for approximately two-thirds of the total length of the bird, a proportion similar to that in living mousebirds. Phylogenetic analysis places the new species as a stem representative of Coliiformes, demonstrating for the ﬁrst time that the two major clades of Coliiformes – Sandcoleidae and Colii – co-occurred at Fossil Lake. Based on the recovered phylogeny, as well as the osteology and feathering of extant and fossil Coliiformes, the wing shape of the new species is interpreted as apomorphic. In addition to documenting unexpected morphological specialization within stem-lineage Coliiformes, the new species adds yet another taxon to the emerging reconstruction of the diverse Paleogene avifauna from the tightly dated and nearly synchronous fossiliferous deposits of the Fossil Butte Member.


INTRODUCTION
Coliiformes are a clade of small, arboreal birds. Extant representatives of Coliiformes (Coliidae) are primarily frugivorous, highly social, and capable of entering a temporary state of torpor (McAtee, 1947;Rowan, 1967;Bartholomew & Trost, 1970;Fry, Keith & Urban, 1988;de Juana, 2001). Morphologically, all extant species are characterized by a prominent feathered head crest, short and rounded wings, highly elongate tails, and a specialized foot. The plumage of these birds also shows several interesting modifications, including a lack of down, restriction of apteria to the head, and loose contour feather barbs that contribute to a hair-like quality (Rowan, 1967;de Juana, 2001). Modifications of the tarsometatarsus and phalanges facilitate rotation of the first and fourth digit, allowing the foot to attain facultatively anisodactyl, zygodactyl, and pamprodactyl configurations, which these birds employ adeptly in clinging, climbing, walking, and manipulating food (Rowan, 1967;Berman & Raikow, 1982). Extant Coliiformes have relatively poor sustained flight capacity, and frequently move through vegetation by 'creeping' using their specialized feet: a behaviour that accounts for the common name of mousebird.
Coliiformes provide a prime example of an extant group with a relict distribution. Considering only extant representatives, Coliiformes are one of the most species-depauperate of the traditional avian orders, comprising only six morphologically conservative species. All six living species are restricted to sub-Saharan Africa. By contrast, extinct stem Coliiformes were diverse and widespread, with 17 distinct named species and two morphologically distinct but as yet unnamed species, known from North America, Europe, and Africa (Ballmann, 1969;Rich & Haarhoff, 1985; Mourer-Chauviré, 1988; Houde & Olson, 1992;Mayr & Peters, 1998;Peters, 1999;Mayr, 2000aMayr, , 2005aKsepka & Clarke, 2009). A few problematic specimens may represent other distinct species, but are currently of uncertain status (see Ballmann, 1969, regarding the possible synonymy of the Miocene 'Colius' paludicola, 'Colius' archiaci, and 'Colius' consobrinus). The Eocene species Botauroides parvus and Eocolius walkeri were originally assigned to Coliiformes (Houde & Olson, 1992; Dyke & Waterhouse, 2001). However, Zelenkov & Dyke (2008) proposed that Botauroides parvus be recognized as Aves incerta sedis because of a lack of morphological character support for the referral, whereas Ksepka & Clarke (2009) proposed Eocolius walkeri would be best considered Aves incerta sedis for the same reason (also see the analyses below).
Coliiformes can be divided into two major subclades: Sandcoleidae (all species of which are extinct) and a second subclade (Colii in this paper) including extant mousebirds and fossil taxa more closely related to these living species than to Sandcoleidae ( ), and all the descendants of that ancestor, i.e. the coliiform crown clade. Keeping Coliidae as the name for the crown clade facilitates communication between paleontologists and neornithologists by providing a clear nomenclatural distinction between stem and crown mousebirds, and by restricting a family name originally defined with reference to extant taxa to the crown radiation. Importantly, this definition helps to avoid confusion, as the paleontological literature is increasingly accessed by neontologists for fossil calibration points used in molecular divergence dating analyses. Sandcoleidae is here applied to the clade uniting all taxa more closely related to Sandcoleus copiosus than to Coliidae, consistent with the usage in Mayr & Peters (1998) (Haarhoff, 1993), although these fossils have not yet been considered in a phylogenetic analysis. Both of these fossils occur within the geographic range of extant mousebirds (Fig. 1 Ballmann, 1969;Rich & Haarhoff, 1985), may eventually prove to represent crown taxa with better material, but at present no derived character states support these species belonging within crown Coliidae, or sharing a closer relationship with extant Colius than with extant Urocolius. Determining the phylogenetic position of these taxa would provide important insight into the biogeographic history of mousebirds.
Although the Eocene-Oligocene fossil taxa Primocolius, Masillacolius, and Oligocolius were interpreted as part of 'crown-group' Coliidae by Zelenkov & Dyke (2008), these taxa were merely recovered as part of a polytomy with a 'Colius/Urocolius' composite crown clade terminal in that analysis. Placement of these fossil taxa relative to the crown clade was not phylogenetically tested because the use of a single composite terminal presumes the monophyly of extant mousebirds, to the exclusion of these or any other fossil taxa. Notably, all the taxa mentioned above as possible sister taxa for Coliiformes possess a foot modified for semi-zygodactyly (Strigiformes), zygodactyly (Piciformes and Psittaciformes), or heterodactyly (Trogoniformes). This suggests the unique foot of mousebirds could have been derived from an already specialized configuration, rather than the anisodactyl arrangement primitive for extant Aves. In light of recent support for a Coliiformes-Strigiformes clade (Hackett et al., 2008), it is interesting to note that the holotype humerus of the Eocene mousebird Eoglaucidium pallas was originally described as that of an owl (Fischer, 1987). However, no derived characters shared with Strigiformes can be readily observed on the humerus, and it does not appear that this misidentification reflects any putative underlying synapomorphies of a mousebird-owl clade.
Extant mousebirds consume a predominantly frugivorous diet, which they supplement with leaves, shoots, flowers, nectar, bark, and, on very rare occasions, insects or scraps of meat (Rowan, 1967;Downs, Wirminghaus & Lawes, 2000;de Juana, 2001). Morphological studies and direct evidence from gut contents have allowed some hypotheses of stem mousebird ecology to be forwarded. Houde & Olson (1992) proposed that the extinct Sandcoleidae were morphologically adapted for climbing and creeping in trees, and were poorly adapted for long periods of flight, based on specializations of the foot, the shallow carina of the sternum, and the short, round wing shape in these birds. Zelenkov & Dyke (2008) reiterated these observations in support of a clinging and/or climbing ecology for stem mousebirds, and further detailed the range of digital movement dictated by the conformation of the hindlimb, as well as noting that the shallow sternum known from many fossil Coliiformes would allow these birds to position themselves closer to tree trunks, and reduce the repulsive force in clinging. Gut contents preserved within the stomach area of Eoglaucidium (Sandcoleidae) and Selmes (Colii) preserve seeds, although it is unclear whether these taxa habitually fed on seeds per se or merely swallowed seeds incidentally during consumption of fruits (Mayr & Peters, 1998;Peters, 1999). Chascacocolius (Colii) exhibits an elongate beak and hypertrophied retroarticular process that have been interpreted as specializations for gaping, in association with prying under bark, breaking open fruits, or ground probing (Houde & Olson, 1992;Mayr, 2005a). In sum, previous studies of fossil taxa suggest most stem mousebirds shared a propensity for climbing through vegetation during foraging with extant mousebirds, but exhibited a wider range of strategies for obtaining food.
In this paper, we describe a new stem mousebird from the early Eocene Green River Formation of Wyoming, exhibiting a novel wing shape, suggesting a possible aerial foraging strategy not previously recognized in Coliiformes. The holotype specimen was collected at Lewis Ranch Site 1 ( Fig. 2; locality A in Grande & Buchheim, 1994). During the late early Eocene, three major lakes formed a large freshwater system, the Green River Lake Complex, stretching over parts of present-day Wyoming, Colorado, and Utah (McGrew & Casilliano, 1975;Grande, 1984Grande, , 1994Buchheim, 1994;Grande & Buchheim, 1994). Lewis Ranch Site 1 is situated within the paleoshoreline of Fossil Lake, the smallest of these three major lakes. Fossil Lake provides the best record of Paleogene birds in North America, having yielded a rich array of often exquisitely preserved avian fossils (Eastman, 1900;Brodkorb, 1970;Feduccia & Martin, 1976;Olson, 1977Olson, , 1987Olson, , 1992 Beds at Lewis Ranch Site 1 are part of the middle unit of the Fossil Butte Member (FBM) of the Green River Formation, which is bounded above by a K-spar tuff, dated by 40 Ar/ 39 Ar spectrometry to 51.66 ± 0.09 Ma (Smith, Carroll & Singer, 2008). Estimated rates of deposition for the well-mapped vertebrate fossil-producing deposits of the FBM support deposition of all fossils within a span of several thousand years or less (Grande & Buchheim, 1994). A predominantly freshwater lake paleoenvironment is indicated by the fish, invertebrate, and plant biota recovered from the fossiliferous layers of the middle unit of the FBM (Grande, 1994; Buchheim, 1998), and the terrestrial environment surrounding Fossil Lake appears to have consisted largely of paratropical lowland forest (MacGinitie, 1969;Buchheim, 1998;Cushman, 1999).
Only a single occurrence of Coliiformes has been previously reported from the Green River Formation. Houde & Olson (1992) described Anneavis anneae from an articulated skeleton lacking a skull, collected from the Warfield Springs locality (locality K, F-2 equivalent deposits of Grande & Buchheim, 1994) of the FBM. These authors also referred additional, less complete material from slightly older deposits of the early Eocene Willwood Formation to this species. Anneavis anneae has been placed within Sandcoleidae (  broadened, and (2) ulna proportionately more elongate than that of all other Coliiformes, except Oligocolius brevitarsus (pronounced elongation of the ulna evolved independently in these two taxa based on their inferred phylogeny). The ulna : humerus ratio is greater than 1.10 in C. acriala whereas in other Coliiformes including Sandcoleidae this ratio is typically less than 1.00. Within Colii, only Palaeospiza bella, Oligocolius brevitarsus, and extant Urocolius exhibit an ulna that is longer than the humerus.
Celericolius acriala differs from all other stem Coliiformes, but resembles extant Coliidae, in that pedal phalanx II-1 is subequal to phalanx II-2 in length. In other stem Coliiformes preserving the foot, phalanx II-1 is strongly abbreviated, measuring onehalf of the length (Palaeospiza and Masillacolius) or one-third of the length (Sandcoleidae and Selmes) of phalanx II-2. Because pedal phalanges remain unknown for several fossil taxa, it is currently equally parsimonious to interpret the elongate phalanx II-1 as a local autapomorphy of C. acriala acquired convergently in Coliidae, or as a synapomorphy of a clade uniting C. acriala and Coliidae, with a reversal to the abbreviated state in Palaeospiza bella.

DESCRIPTION AND COMPARISONS
In overall size, the holotype individual is approximately equal to the extant Urocolius indicus (Redfaced Mousebird), a bird with a body mass of~50 g (Downs et al., 2000). The holotype specimen is largely articulated and appears to be nearly complete, except for damage to the skull and the loss of two pedal unguals. The coracoids, right humerus, and left femur are either not preserved or are hidden beneath the remains of the sternum and pelvis (Fig. 4). The specimen has suffered postmortem degradation, and much of the skeleton is flattened and poorly preserved, with the exposed bone surface often lost, obscuring features such as muscle insertion scars. Some areas of the skull exhibit minor retouching. At the tip of the premaxilla and dorsal portions of the mandible, incomplete areas have been filled in by applying paint to the negative mold left in the slab by missing bone. Abbreviations: cmc, carpometacarpus, dI, pedal digit I; dII, pedal digit II; dIII, pedal digit III; dIV, pedal digit IV; ext, impression of processus extensorius; hyp, cristae hypotarsi; isch, ischium; mpI-1, manual phalanx I-1; mpII-1, manual phalanx II-1; mpIII-1, manual phalanx III-1; pc, processus costales; pi, processus intermetacarpalis; pt, processus transversus vertebrae; pu, pubis; py, pygostyle; rad, radiale; tmt, tarsometatarsus; uln, ulnare. Scale bars: 5 mm. As in extant Coliidae, the skull roof is domed and rounded, and a large fonticulus orbitocranialis is present. Unfortunately, fine details of the skull are nearly entirely lacking because of poor preservation. As noted above, the beak tip has been retouched. Because of this modification the precise shape and total length of the beak must be considered uncertain. The mandible is slender and straight near its midpoint, but the caudal and rostral portions are not preserved.
A few tracheal rings are scattered on the slab. Tracheal rings are also preserved in at least one specimen of the sandcoleid Eoglaucidium pallas from Messel (Mayr & Peters, 1998: fig. 3), but are not preserved in the articulated holotype skeleton of the sandcoleid Anneavis anneae, also from FBM deposits (Houde & Olson, 1992). However, because the Anneavis anneae holotype was collected from a different quarry (Warfield Springs, locality K), representing a more near-shore environment than Lewis Ranch Site 1 (Grande & Buchheim, 1994), it is possible that the lack of tracheal ring preservation in that specimen may be related to preservation in a distinct depositional environment.
The cervical and dorsal vertebrae are poorly preserved. In contrast, the caudal vertebrae and pygostyle are well-preserved and are exposed in ventral view. The processes transversus vertebrae of the caudal vertebrae are much wider anteroposteriorly than in extant Coliidae. Sandcoleus copiosus and a privately owned specimen referred to Chascacocolius (Mayr, 2005a) also possess narrow processes transversus vertebrae, comparable with those in Coliidae. As in extant mousebirds, the discus pygostyli is greatly expanded, forming a distinct broad, rounded plate (Fig. 5).
Much of the sternum is obscured by overlying elements, but several phylogenetically important features of the caudal margin can be observed. The incisura lateralis is shallow compared with extant Coliidae. In this respect, C. acriala shows a condition intermediate between that in Sandcoleidae, where the incisure lateralis is reduced such that the trabecula intermedia arises from the trabecula lateralis, and in Coliidae, where the incisura lateralis is very deep and the trabecula lateralis and trabecula intermedia are clearly separated. Distally, the trabecula lateralis shows slight mediolateral expansion, but does not approach the inverted T-shape developed in extant Urocolius. The scapula is shorter than the humerus, as in most stem mousebirds. In Coliidae and Oligocolius brevitarsus, the scapula exceeds the humerus in length. The corpus scapulae is fairly straight and maintains a nearly constant width, with no distal expansion.
The crista deltopectoralis of the humerus is strongly projected, as in other mousebirds. This crest extends for approximately one-third of the proximodistal length of the humerus, whereas it extends approximately one-fourth of the length of the humerus in Sandcoleidae and Coliidae. The humeral shaft is straight, more closely resembling the condition in extant Coliidae than the curved shaft of Sandcoleidae. Preservation at the distal end is poor, so it is not possible to determine whether C. acriala possessed the crescent-shaped depression proximal to the condylus dorsalis seen in some Colii. The ulna : humerus length ratio (Fig. 6) is higher than in any other representative of Coliiformes, with the possible exception of Oligocolius brevitarsus (Mayr, 2000c). Accounting for the slight damage to the proximal humerus, is it not possible to say with certainty whether C. acriala exceeded the ratio of 1.11 in Oligocolius brevitarsus. Few details of the radius and ulna can be ascertained, although it is clear that both elements are robust, as in other Coliiformes. The olecranon appears blunt and weakly differentiated, as visible in the right ulna. The left carpometacarpus is exposed in ventral view, and the right carpometacarpus is exposed in dorsal view. Regions of the ventral bone surface of the left element are missing, exposing the internal surface of the shaft for much of its length, but fortunately the bone is intact at the proximal end. A small processus intermetacarpalis, which does not contact metacarpal III, is preserved (Fig. 5). An impression of the processus extensorius of the left carpometacarpus indicates that this process was strongly projected. The spatum intermetacarpale is wide, although metacarpal III is less strongly bowed than in extant Coliidae. Although the distal end of metacarpal III has been lost from the slab on the right carpometacarpus, its impression indicates metacarpals II and III extended to the same level. The alular phalanx is fairly long. Distally, phalanx II-1 expands to a width equal to that of the carpometacarpus, giving the phalanx a hatchet-like shape. Phalanx II-2, although incomplete on the right side, was clearly elongate and straight, as indicated by the preserved portion. Phalanx III-1 bears a prominent, proximally positioned processus flexorius, and is slender and tapers distally, closely resembling the morphology in extant Urocolius.
Parts of the pelvis are exposed in ventral view, including the posterior portions of the right ischium and pubis. At the caudal portion of the synsacrum, distinct processus costales can be observed, a primitive feature for Coliiformes. As in extant mousebirds, the pelvis is mediolaterally wide. The foramen ilioischiadicum is elongate and narrow. The pubis is thin and rod-like, and does not approach or contact the ischium distally.
Although both legs are detached from the pelvis, the major hindlimb elements remain largely articu-lated. The hindlimb is relatively slender and elongate, but few details of the femur and tibiotarsus are discernable. The tarsometatarsus approaches the length of the humerus and is similar in proportions to the slender tarsometatarsus of Coliidae, as opposed to the stouter tarsometatarsus of Sandcoleidae. The cristae hypotarsi are proximodistally short. Unfortunately, no details of the foramina vascularia proximalia or canales hypotarsi can be discerned.
On the right foot, digit III is preserved overlying the other digits. Portions of phalanx IV-2 and all of phalanges IV-3, IV-4, and IV-5 are visible, whereas digits I and II are largely obscured. On the left foot, digits I and II are well exposed. Digit III is visible, but lacks the ungual, and digit IV is complete but largely obscured by overlying elements. Thus, digital proportions can be reconstructed by combining observations from the two feet. The ungual of digit I is oriented in the same direction as the remaining unguals on the left side, corresponding to the apparently pamprodactyl configuration of the digits. However, because phalanx I-1 is not in articulation with metatarsal I, this orientation is possibly an artifact of preservation. Phalanges II-1 and II-2 are elongate and subequal in size. In contrast, the proxi-mal three phalanges of digits III and IV are all strongly abbreviated compared with the penultimate phalanx. All pedal unguals are long and strongly curved. In the proportions of the pedal phalanges, C. acriala shares the derived pronounced abbreviation of the proximal three phalanges of digits III and IV with all other Coliiformes. The unabbreviated phalanx II-1 in C. acriala agrees with the condition in crown Coliidae. In contrast to C. acriala, other stem Coliiformes that preserve an intact foot show some degree of shortening in phalanx II-1, with the size of this element ranging from approximately one-third to onehalf of the length of phalanx II-2. As mentioned above, whether the phalangeal proportions in C. acriala represent convergence with Coliidae, or a synapomorphy of a larger clade including C. acriala and Coliidae, that is reversed in Palaeospiza bella, remains uncertain because of homoplasy and missing data.
Carbonized material adjacent to the wing bones and pygostyle is interpreted as integumentary traces. Contrast between these areas and the matrix is greater under ultraviolet light than under natural light; however, the carbonized material does not noticeably fluoresce. No structural details of the feathers (e.g. rachi and barbs) can be detected. Preservation is best along the leading edge of what is inferred as the tenth primary, and lessens in quality towards the trailing edge, so that it is unclear whether all primaries are represented. There is no indication of secondary feather preservation adjacent to the ulna. Traces of the leading primary extend 124.6 mm from the tip of manual phalanx II-2. This corresponds to approximately 60% of overall wing length (length of wing skeleton plus projecting feathers). Inferred wing shape is elongate and tapering, contrasting markedly with the short, rounded wing of extant mousebirds (Fig. 7). A short, rounded wing shape has also been reported in several stem Coliiformes with preserved feathering traces, including  Ingroup taxonomic sampling includes five of the six extant mousebird species (we were unable to obtain skeletal material of Colius leucocephalus for coding). We scored 13 fossil coliiform taxa, as well as two unnamed fossils that preserve informative combinations of character states. The first of these unnamed fossils, WDC-C-MG 148+149, represents a distinct but unnamed sandcoleid species from the middle Eocene Messel Formation (Mayr, 2000c). The second, MNH.Q.O.596, is an isolated tarsometatarsus from the Eocene-Oligocene Quercy fissure fills. This specimen may represent Selmes absurdipes (Mayr & Mourer-Chauviré, 2004), but comparisons with that species are limited by the state of preservation. Because a firm referral cannot be made at this time, we conservatively score MNH.Q.O.596 at the specimen level. Analyses were conducted both including and excluding three poorly known taxa represented by single-element holotypes: Eobucco brodkorbi (represented by a tarsometatarsus), Primocolius sigei (represented by a humerus), and Primocolius minor (represented by a tarsometatarsus). In previous analyses the inclusion of these taxa led to a large increase in the number of most parsimonious trees and low resolution in the strict consensus cladogram (Ksepka & Clarke, 2009). We consider referrals of a partial carpometacarpus to Primocolius sigei and partial humerus to Primocolius minor tentative, given the lack of association (Mourer-Chauviré, 1988), and did not incorporate codings from these elements. However, including codings from these referred specimens does affect the placement of Primocolius with respect to Coliidae.
Eocolius walkeri, a poorly known taxon proposed to represent a stem mousebird (Dyke & Waterhouse, 2001; Zelenkov & Dyke, 2008), was excluded from the primary analysis as this taxon lacks any synapomorphies supporting placement within Coliiformes (Ksepka & Clarke, 2009; Mayr, 2009). However, given the controversial nature of this taxon, we tested its phylogenetic position by including it in an additional analysis using the same data set and search strategy as used in the primary analysis.
Given the uncertainty regarding the closest extant relatives of Coliiformes (see above), we included multiple outgroup taxa representing proposed sister taxa for Coliiformes, including exemplars from Strigiformes (Hackett et al., 2008) Searches were conducted in PAUP*4.0b10 (Swofford, 2003) using the Branch and Bound algorithm, with all characters equally weighted and multistate codings used only for polymorphism. Branches with a minimum length of 0 were collapsed. Bootstrap support was calculated from 1000 replicates using a heuristic search strategy, with random taxon addition sequence and tree bisection and reconnection (TBR) branch swapping, and Bremer support was calculated manually in PAUP*4.0b10. Trees were rooted to the psittaciform Pseudasturides macrocephalus given the recovery of Psittaciformes as being more distantly related to Coliiformes than the remaining outgroup taxa in the large-scale phylogenomic analysis of Hackett et al. (2008).

RESULTS
In the analysis including all ingroup taxa, 1872 most parsimonious trees (MPTs) of 137 steps were recovered. In the strict consensus tree (Fig. 8A), C. acriala is recovered as a member of Colii, and placed in a large polytomy with other members of this clade. Although resolution within Colii is poor, monophyly of crown clade Coliidae to the exclusion of all fossil ingroup taxa is supported. Several Eocene-Oligocene taxa (Masillacolius, Oligocolius, and Primocolius) previously interpreted as part of crown group Coliidae (Zelenkov & Dyke, 2008) fall outside the crown clade.
The analysis excluding the poorly known taxa Eobucco brodkorbi, Primocolius sigei, and Primocolius minor (each known with certainty from a single bone), resulted in 20 MPTs of 137 steps. Resolution is markedly improved in the strict consensus tree (Fig. 8B). In this tree, C. acriala is placed in a polytomy with Masillacolius and the clade Palaeospiza + (Oligocolius + Coliidae). Phylogenetic relationships are congruent with those recovered by Ksepka & Clarke (2009) for the taxa common to these analyses.
Although none of the fossil taxa we included were recovered as parts of Coliidae, our results identify character states relevant to determining the stem/crown status of future fossil discoveries. Two character states are optimized as unambiguous synapomorphies of Coliidae: (8:1) indistinct processes costales of the synsacral vertebrae and (30:1) metacarpal III strongly bowed. This second feature, however, is also present in some stem mousebird taxa (i.e. Sandcoleidae), and so should not be considered diagnostic for incomplete specimens. Fourteen additional character states (state 1 of characters 2, 3, 5, 7, FOSSIL MOUSEBIRD WITH FEATHER PRESERVATION 695 9, 20, 21, 32, 33, 34, 41, and 44, and state 0 of characters 31 and 50) may represent synapomorphies for Coliidae, but are currently ambiguously optimized because of missing data in fossil taxa and/or because of equally parsimonious solutions for homoplastic distributions. We caution that the presence of character states optimized as synapomorphies of Coliidae are not in themselves sufficient to justify placement of fossils into the crown clade: these character states merely support placement closer to Coliidae than to Oligocolius. However, retention of plesiomorphic states for these characters does provide evidence that a fossil belongs outside of the crown clade.
The definitive placement of fossils into crown Coliidae requires recognition of unambiguous synapomorphies of the Colius and Urocolius lineages. For Colius, these character states are (12:1) elongate, strongly projected processus craniolateralis of the sternum, and (5:2) large, ovoid fenestra caudalis mandibulae. Ulna shorter than humerus (character state 26:0) is optimized as an unambiguous synapomorphy of Colius in our strict consensus tree, but is also present in many stem mousebirds. For Urocolius, unambiguous synapomorphies include: (4:1) sharp ventral deflection of the rostral end of the mandible, (11:1) a projecting, blade-like apophysis furculae, (14:1) inverted T-shaped expansion of the caudal ends of the lateral trabeculae of the sternum, and (19:1) acromion narrow and sharply projecting.
The additional analysis including Eocolius walkeri resulted in 12 168 MPTs of 138 steps. In the strict consensus tree (Fig. 8C), Coliiformes are monophyletic to the exclusion of Eocolius walkeri. Relationships within Coliiformes are the same as those in the primary analysis, although relationships between the outgroup taxa are less well resolved. Excluding Eobucco brodkorbi, Primocolius sigei, and Primocolius minor results in fewer MPTs (130) and more resolution within Coliiformes, but does not affect the placement of Eocolius walkeri or outgroup taxa. Three character states are optimized as unique synapomorphies of Coliiformes: (21:0) fossa tricipitalis of humerus not pneumatized (reversed in Coliidae), (26:0) ulna shorter than humerus (reversed in several taxa), and (27:1) cotyla ventralis of ulna expanded and extending onto olecranon. A fourth character state is present only in Coliiformes and one outgroup species, the psittaciform Pseudasturides macrocephalus: (36:1) crista trochanteris of femur does not project proximal to level of femoral head. None of these character states is present in Eocolius walkeri.
Three character states were optimized as synapomorphies of Coliiformes with Eocolius as basally divergent within that clade in the only phylogenetic analysis to recover this proposed position (Zelenkov & Dyke, 2008). These characters were included in our data set, but none support an Eocolius + Coliiformes clade as optimized in our results. The single proposed unambiguous synapomorphy, ). Of the two character states proposed as ambiguous synapomorphies, (23:0) curved shaft of humerus is also present in the exemplars of Strigiformes that we examined (Zelenkov & Dyke, 2008, coded the humerus as 'straight' in a composite Strigiformes terminal, although they did not list the species examined). The second, (30:1) metacarpal III bowed, cannot in our opinion be coded in Eocolius, as this part of the carpometacarpus is almost completely lacking in the holotype. Indeed, Zelenkov & Dyke (2008: 1418 noted that 'the shape of the bone is extrapolated'. Regardless of coding decisions, both curvature of the humerus and bowing of the third metacarpal are widely distributed in Aves. We recover no phylogenetic support for placing Eocolius walkeri in Coliiformes, and conclude that further material is necessary to resolve the relationships of this taxon within Aves.

DISCUSSION
The FBM deposits have yielded the most diverse and well-preserved Eocene avifauna from North America. Further insights into the taxonomic and ecological characteristics of this fauna are still actively emerging. Two representatives of Coliiformes, C. acriala and the sandcoleid Anneavis anneae, are now known to occur within the FBM. These two species differ dramatically in wing shape, increasing the observed ecological and species-level diversity both within the FBM avifauna and within the clade Coliiformes.
The elongate wing shape in C. acriala is unexpected given the morphology of living mousebirds and our previous knowledge of fossil representatives of this clade. Support for this unique wing shape comes from two lines of evidence. First, the proportions of the major wing bones in C. acriala indicate a higher ulna : humerus ratio than any living mousebird (Fig. 6). Second, the preserved outline of the wing integument indicates a much more elongate and tapering wing than reported in any other member of Coliiformes (Fig. 7). Together, these morphologies imply a unique flight style not previously described for any extinct or extant mousebird. Although the wing bone proportions are notable, it is predominantly the length of the primary feathers that contribute to the long, pointed shape of the wing (Fig. 7). In this regard, C. acriala resembles the oscine Hirundinidae (swallows), in which the primary feathers account for approximately 60% of the wing length (Kaiser, 2007).

FOSSIL MOUSEBIRD WITH FEATHER PRESERVATION 697
Long, pointed wings result in higher aspect ratios and lower loading (Saville, 1957;Rayner, 1988). This wing shape is associated with agile open-air flight and is typical of small, aerial insectivores, such as tyrannid flycatchers and swallows. By contrast, the short, rounded wing shape seen in extant Coliidae is optimized for fast, short bursts of energetically expensive flight (Saville, 1957;Rayner, 1988). This wing morphology is advantageous for maneuvering between trees in closed forested environments over short distances, but is poorly suited for longer periods of sustained flight (Kaiser, 2007). Living mousebirds show relatively poor capacity for sustained flight, and are among the most sedentary of living birds (de Juana, 2001). Previously available feather traces indicate that Sandcoleidae, part of the basal divergence within Coliiformes, possessed a wing similar in aspect to extant Coliidae (Houde & Olson, 1992;Mayr, 2000c). Thus, a short, rounded feathered wing can be inferred through phylogenetic bracketing and proportions of the wing bones as present across most of stem Colii. Intriguingly, one other stem representative of Colii, Oligocolius brevitarsus, possesses elongate distal wing bones, also consistent with a relatively long wing (Mayr, 2000c). Although feather preservation is lacking for that taxon, precluding the determination of wing shape, it should be considered that wing elongation, and associated ecological correlates, may be discovered to be a trend in multiple stem mousebird lineages.
Carbonized traces of the rectrices indicate that the tail of C. acriala was greatly elongated, approaching two-thirds of the bird's total length based on the preserved length of the rectrices. The distal tip of the tail is poorly preserved, so whether this region was square, graduated, forked, or ornamented in some fashion remains unknown. An elongate tail is present in all extant mousebirds and has been documented in multiple fossil taxa (Anneavis, Eoglaucidium, and the unnamed sandcoleid WDC-C-MG 148+149). At present, no fossil coliiform preserves evidence of a short tail, although feather traces are unpreserved for many species. Given this distribution, an elongate tail is currently optimized as ancestral for Coliiformes, and is retained throughout the clade.
Extant mousebirds possess distinctly stiffened and graduated tails. Tail shape differs only slightly between the extant taxa Urocolius and Colius. Colius possesses a comparatively broad tail compared to Urocolius, although the overall shape is still narrow relative to other birds. Tail function has been little commented on in extant mousebirds, and the possible functions of the elongate tails of extinct mousebirds have likewise received little speculation. Modifications of the tail serve many functions in birds, including lift generation, improving maneuverability, and enhancing intraspecific display (Fitzpatrick, 1999). Enhancing aerodynamic properties, balance, or display visibility are all plausible, and not necessarily mutually exclusive, functional explanations for the elongation of the tail in Coliiformes.
An aerodynamic function seems the least likely. Although enlarging the tail can increase airfoil size, elongation beyond the total width of the tail increases drag without further increasing lift (Evans & Thomas, 1992; Balmford, Thomas & Jones, 1993). Whereas the elongate tail streamers of some Hirundinidae enhance maneuverability (Norberg, 1994), these streamers are formed only by the outermost pair of rectrices. Although the shape of the distal tip of the tail in C. acriala is uncertain, a relatively broad shape is indicated for most of the tail area, and there is no evidence for 'streamers'. The tail of the sandcoleid Eoglaucidium is likewise relatively broad (Mayr & Peters, 1998).
A non-aerodynamic locomotor function could involve the use of the tail in propping the bird upright against vertical inclines, such as tree trunks during climbing, a behaviour exhibited by Picidae (woodpeckers). In Picidae, the inner rectrices are stiffened to provide support during this propping behaviour. Unfortunately, fine details of the individual rectrices remain unavailable for most fossil representatives of Coliiformes, although rectrices of Anneavis and Eoglaucidium preserve a broad proximal shaft (Houde & Olson, 1992;Mayr & Peters, 1998). Although the rectrices are stiffened in Coliidae (Rowan, 1967), this propping behaviour does not appear to occur regularly in extant mousebirds. Nonetheless, a propping function was proposed for the tail in at least one previously described fossil mousebird. Zelenkov & Dyke (2008) advocated scansorial abilities for Chascacocolius oscitans (stem Colii) in part based on two putative features of the pelvis that are present in climbing woodpeckers, but are unknown in other Coliiformes. These features, a short ala preacetabularis ilii and well-developed processus terminalis ilii (origin of m. flexor cruris lateralis and m. iliofibularis), were illustrated in a reconstruction of the incomplete pelvis of the holotype (and only known) specimen of Chascacocolius oscitans (Houde & Olson, 1992: fig. 11). However, re-examination of the holotype, upon which this reconstruction was based, revealed that neither the cranial margin of the ala preacetabularis ilii nor the processus terminalis ilii are preserved (see Houde & Olson, 1992: plate II.1), so the proportions of the ilia and the shape of the processus terminalis ilii remain unknown. We find no evidence that the pelvis of Chascacocolius or any other mousebird was substantially modified for vertical climbing, casting doubt on a proposed propping function of the rectrices.
A third possibility is that the tail served as a display structure. This hypothesis may best explain why the elongate tail is present in stem and extant coliiform taxa showing morphologies consistent with markedly different flight styles and feeding strategies. Extant mousebirds are highly social birds, but have relatively simple sexual display behaviours, consisting primarily of repeated 'bouncing' from a fixed perch (Rowan, 1967;de Juana, 2001). However, no described display behaviours for Coliidae actively utilize the tail, and tail length is not noticeably sexually dimorphic, with the exception of slight differences in Urocolius macrourus (de Juana, 2001).
Further study of extant mousebird behaviour and life history is necessary to advance the investigation of possible tail functions in fossil species. Aerodynamic and propping functions seem unlikely given the present data. However, balance-related or species recognition functions remain plausible.
When the acquisition of key morphological features are optimized on the coliiform cladogram, specializations of the feet and tail are shown to evolve early in the clade's history, and specializations of the wing and beak are shown to evolve later. Specializations of the foot correlated with the ability to rotate at least digit IV into a facultatively zygodactyl position are optimized as present at the base of Coliiformes. However, because the higher level relationships of Coliiformes are not yet firmly established, it remains ambiguous whether many specializations of the foot (e.g. plantarly-projecting wing-like flange on trochlea metatarsi II and IV and shortened pedal phalanges) might be synapomorphies of Coliiformes or of a larger clade. For example, if Coliiformes and Strigiformes are sister taxa, as recovered by Hackett et al. (2008), the aforementioned specializations of the foot are most parsimoniously interpreted as synapomorphies of a Coliiformes + Strigiformes clade.
Presence of an elongated tail is optimized as an unambiguous synapomorphy of Coliiformes, regardless of the higher level relationships of the clade. The elongate tail is indeed one of the most distinctive features of both living and fossil mousebirds. Yet, as discussed above, we currently have little understanding of its function. Finer preservation yielding insight into properties such as feather stiffness, shape of the distal tip of the tail, or coloration pattern in basal mousebirds could potentially improve our understanding of why the distinct coliiform tail evolved initially, and why it has been conserved across coliiform phylogeny, and over > 55 Ma of the known temporal range of the clade.
Specializations of the wing departing from the short, rounded morphology reconstructed as ancestral for Coliiformes occurred later in the phylogenetic history of the clade. Celericolius acriala provides the most striking example of wing specialization, and at least some degree of elongation is implied by wing bone proportions in Oligocolius brevitarsus (Mayr, 2000a). Extant Coliidae are inferred to have inherited the short, rounded wing shape seen in other stem mousebirds preserving feather traces from their common ancestor. The specialized short, heavy beak present in extant Coliidae appears to be one of the latest features to arise in the phylogenetic history of the clade. Sandcoleidae and Selmes have significantly longer, more generalized beaks (Houde & Olson, 1992;Mayr, 2001), and Chascacocolius exhibits an elongate beak specialized for gaping (Houde & Olson, 1992;Mayr, 2005a). Unfortunately, most fossil representatives of Colii do not preserve the complete skull, and so it remains uncertain whether the shortened beak of Coliidae arose within some part of stem Colii or only characterizes the crown clade.

CONCLUSIONS
Celericolius acriala adds to the extensive species diversity and morphological disparity recognized within stem Coliiformes. We tentatively propose an aerial foraging strategy for C. acriala from the similarities in wing shape to extant small, insectivorous birds. Future recovery of a specimen with an intact beak could help refine our understanding of feeding ecology in this species. More diverse feeding specializations may explain the co-occurrence of multiple stem coliiform taxa at many Paleogene localities. Chascacocolius, with its highly modified beak adapted for breaking open fruit, prying under bark, or ground probing (Houde & Olson, 1992;Mayr, 2005a), provides another example of diverse foraging habits in stem mousebirds. Both Celericolius and Chascacocolius co-occur with taxa interpreted as more general feeders at their respective localities.
We have demonstrated the co-occurrence of two coliiform species in a restricted area (Fossil Lake) and horizon (FBM) within the Green River Formation, but even greater mousebird diversity is known from other deposits. A cladogram of Coliiformes calibrated to the fossil record is presented in Figure 9, and the global distribution of fossil mousebird localities is illustrated in Figure 1 Peters, 1998;Peters, 1999;Mayr, 2005a, b). The Messel Coliiformes demonstrate that mousebird diversity at a single local environment (Lake Messel) was at least equal to the current worldwide species diversity. Standing global diversity is, of course, uncertain for any time interval, given the nature of sampling in the fossil record. Nonetheless, it is most likely that global diversity was several times higher than today throughout the Eocene, especially considering the scarcity of Eocene terrestrial localities suitable for preserving small birds over vast geographic regions.
Passeriformes, which include insectivorous radiations such as Tyrannidae and Hirundinidae, are thus far absent from the Eocene of North America. Aerial insectivores of the clade Apodiformes are relatively common components of the Eocene London Clay and Messel avifaunas (e.g. Harrison, 1984;Mayr & Peters, 1999;Mayr, 2005c). However, no fossils assignable to this clade have yet been identified in our review of over 100 FBM specimens curated in museum collections. A privately owned fossil from the Tipton Member at Lake Gosiute (approximately equivalent in age to the FBM of Fossil Lake; Smith et al., 2008) is the only reported record of Apodiformes from the Green River Formation (see Grande, 1984;Feduccia, 1999), but this specimen remains undescribed precluding more detailed comparisons. One stem representative of Coracii (rollers) from the Green River Formation possesses modifications of the Figure 9. Phylogeny of the Coliiformes calibrated to the stratigraphic record. Character states reconstructed as unambiguous synapomorphies in the strict consensus tree from Figure 8B are listed for each branch. Additional character states that are reconstructed as unambiguous synapomorphies in all most parsimonious trees combined in this consensus (but not reconstructed as such in the strict consensus tree because of soft polytomies) are listed in italics. Characters that exhibit homoplasy within Coliiformes are indicated with an asterisk. As a result of the lack of phylogenetically constrained Neogene mousebird fossils, the timing of divergences within Coliidae is uncertain, but we have placed the Urocolius/Colius divergence in the Pliocene, reflecting probable African representatives of at least the extant Colius lineage (Haarhoff, 1993). Stratigraphic ranges of European taxa are in light grey; North American taxa in are in dark grey. All extant Coliiformes are restricted to Africa. Life reconstructions by Kristin Lamm. rostrum and gape consistent with aerial insect capture (Clarke et al., 2009), although the wing proportions of this much larger taxon are consistent with a flight style distinct from that inferred for C. acriala.
Competitive exclusion by Passeriformes has been hypothesized as a major factor in the retreat of arboreal clades such as Coliiformes, Coracii, and Upupiformes from niches now occupied by Passeriformes (Harrison, 1979;Mayr, 2005cMayr, , 2009). The arrival of Passeriformes in North America is as yet poorly temporally constrained, as are the disappearances of groups such as Coliiformes and Coracii from the continent. Further revision of the Green River avifauna in concert with data from younger deposits will contribute significantly to investigating these hypotheses.