Skull

The skull of AMNH 30578 (fig. 5) lies on its right side with the left side exposed. The left tibiotarsus lies across the orbit, hiding details of the posterior portion of the skull and much of the posterior portion of the skull roof. The skull is mediolaterally compressed in such a way that the mandibular rami abut, hindering a ventral view of the palate. The overall length of the cranium (= 84 mm) is shorter than the humerus, and differs from the other known Green River Formation lithornithid (Pseudocrypturus cercanaxius, USNM 424078) in which the skull is longer than humerus. The rostrum of AMNH 30578, measured from the anteriormost portion of the lacrimal (= 46 mm), is slightly greater than half the skull length (= 42 mm).

The anterior portion of the mandible overlies much of the anterior portion of the beak. Within the naris, a thin sheet of mesethmoid is present (extending well anterior to the nasofrontal hinge) to form an internarial septum, as in other lithornithids (e.g., L. promiscuus, USNM 391983), tinamous, and Apteryx owenii (AMNH 18279). The premaxillary process of the nasal is dorsoventrally thin as in tinamous (e.g., Crypturellus undulatus), other lithornithids (e.g., Lithornis celetius, USNM 290601), and ratites. The ventral (= maxillary) process of the nasal is projected at an angle about 45° to horizontal. This mediolaterally compressed process widens slightly ventrally to meet the maxilla. The ventral process of the nasal is anteroposteriorly constricted, as in other lithornithids (Lithornis promiscuus, USNM 391983) and tinamous. Details of the articulations among the premaxillae, maxillae, and the palate cannot be discerned. The jugal bar is somewhat rectangular in cross section with its long axis oriented dorsoventrally, slightly thickened posteriorly, and is bowed slightly ventrally ventral to the orbital region. The posterior portion is broken and disarticulated, revealing a pitlike articular surface for the quadratojugal on the quadrate.

The lacrimal, exposed in lateral view, forms the anterior portion of the orbit. Its mediolaterally broad ventral process contacts the jugal. This flat process of the lacrimal is most similar to that of other lithornithids (e.g., Lithornis promiscuus, USNM 391983; Lithornis celetius, USNM 290601; Pseudocrypturus cercanaxius, USNM 336103) and contrasts with the anteroposteriorly restricted element in other palaeognaths (char. 13). It appears that the lacrimal contacted the ectethmoid, as in tinamous and in some lithornithids (e.g., Lithornis promiscuus, USNM 391983; Lithornis celetius, USNM 290601).

The large orbit contains a nearly intact sclerotic ring. The sclerotic ossicles are indistinguishable from each other in most cases, making a count of the number of ossicles impossible. No supraorbital bones are present in AMNH 30578, and they have not been described in other lithornithids (Houde, 1988); such elements are present in tinamous (Zusi, 1993). The postorbital process is somewhat broad, like that of Paracathartes howardae (USNM 361415). However, the apparent breadth of the postorbital process may be exaggerated by the crushing of the endocranial cavity by the tibiotarsus. A short and pointed zygomatic process covers the lateral side of the head of the quadrate. The cranium does not appear to have dendrite-shaped furrows on the frontoparietal area as in the Messel lithornithid (IRSNB Av82; Mayr, 2009) and Lithornis celetius. A robust paroccipital process with a nearly flat lateral surface forms the posterior portion of the tympanic cavity. The ventral extent of the paroccipital process terminates dorsal to the mandibular articulation of the quadrate. The posterodorsal portion of the tympanic cavity bears a well-defined fossa enclosing many tiny foramina within it, near the articulation with the head of the quadrate (= posterior tympanic recess).

The right quadrate is exposed in lateral view, but much of the lateral surface is covered by other elements. The orbital process is convex laterally, and tapers to an anteriorly rounded end. The bulbous anterior process is similar to that of other lithornithids (e.g., Lithornis promiscuus; USNM 336535), Dromaeus (TMM unnumbered), and tinamous. The posterodorsal portion of the quadrate head is partially exposed posteriorly where it meets the braincase. A disarticulated pterygoid is present anteroventral to the quadrate. The small pterygoid has an expanded, cuplike end for articulation with the quadrate. The lateral edge of the pterygoid thins into a ventrally directed flange. This thin flange indicates that a ventral pocket was present as in Pseudocrypturus cercanaxius (USNM 336103), Apteryx owenii, (USNM 18279), Lithornis promiscuus (USNM 391983), and Paracathartes howardae (USNM 391984) (Houde, 1988). The basipterygoid processes are not well exposed in this specimen, but from the morphology of the pterygoid, we deduced that the basipterygoid processes would have been posteriorly displaced unlike the more anteriorly shifted processes in galloanserines.

The left mandible is visible in lateral view (fig. 5), and the right mandible is partially visible in medial view. The height of the mandible increases posteriorly and is at its greatest height at its posterior termination. The mandibular rami are fused anteriorly in a short symphysis (~5% of mandible length). The anterior end of the mandible is laterally expanded along the dorsal margin to form a flat occlusal surface, as in all lithornithids and all other palaeognaths (Clarke, 2004). Small foramina (= foveae corpusculorum nervosorum) are present on the anterior end of the mandible, and a short anteroposteriorly oriented row lies in a shallow groove on the lateral side on the first third of the mandible. The lateral grooves extend posteriorly to the articulation with the lacrimal as a very shallow depression and are present in lithornithids and other palaeognaths (Houde, 1988). The anterior end of the mandible does not appear to be deflected ventrally as in Lithornis from Messel (IRSNB Av82; Mayr, 2009) and Lithornis promiscuus (USNM 336535). The dentary portion of the left mandible is disarticulated and rotated dorsally. However, the well-preserved surfaces of this element allow articulations to be discerned. The posterior portion of the dentary is forked and split into distinct dorsal and ventral processes. The ventral process is more posteriorly expanded and sits in a lateral groove on the angular, as in most stem Aves, lithornithids, and palaeognaths (Clarke, 2004). The angular continues anteriorly, medial to the dentary, for nearly half the length of the mandibular ramus. The anterior extent of the angular is not discernible, and it may be partially fused with the splenial. A small projection on the dorsal portion of the surangular is interpreted to be a small coronoid process. The surangular and angular are separated by a well-defined, recurved, open suture between the two elements anteriorly that becomes closed without a visible suture posteriorly. Two small foramina are present on the lateral side of the surangular (fig. 5). Medially, the splenial is unfused to the other mandibular elements as evidenced by a ventral margin that is well separated from the dentary and a posterior portion that medially overlaps the angular. Anteriorly, the splenial reaches the mandibular symphysis, although it is not clear whether the splenial is incorporated into the mandibular symphysis as in Apteryx owenii (USNM 18279).

The articular region of the mandible is visible only in lateral view. The lateral mandibular condyle of the quadrate sits in a well-defined notch of the lateral side of the mandible. The anterior wall of the notch terminates dorsally in a point, and the posterior wall expands into a distinct posterodorsally projected retroarticular process that expands well above the dorsal extent of the mandible. This process is continuous with the lateral side of the mandible and is distinctly separated from the medial process, and this process is present in all lithornithids (Lithornis promiscuus, USNM 391983; Pseudocrypturus cercanaxius, USNM 336103; MGUH 26770; Bourdon and Lindow, 2015) with the posterior portion of the mandible exposed. As noted by Mayr (2009), this process is hypothesized to be the attachment site for the aponeurosis mandibulae medialis. The mandibular ramus terminates directly posterior to this process.

Vertebral Column

The presacral vertebrae are split into three sections: an articulated series consisting of presacrals 1–12 that are in articulation with the skull (figs. 5, 6); a partially disarticulated series consisting of four presacral vertebrae partially obscured by the posterior portion of the sternum (fig. 5); and an articulated series consisting of the last six presacral vertebrae (fig. 7). A total count of 22 presacral vertebrae is preserved in AMNH 30578, but it is not clear how many presacrals were present in life; 23 were estimated for the Siber and Siber lithornithid specimen (USNM 424078; Houde, 1988). All presacral vertebrae have heterocoelous articulations.

The atlas lies under the left tibiotarsus (fig. 5). The axis is preserved in lateral view and is shorter than the third cervical. Small foramina, observable in lateral view, lie on the lateral side of the centrum just posterior to transverse foramina as in Chauna torquata (AMNH 1773). There are no winglike processes on the lateral side of the centra (Mayr and Clarke, 2003: char. 50), as there are in Crypturellus undulatus. The robust ventral process is well separated from the articular surface contacting the atlas. A deep pneumatic fossa is present on the ventral side of a lamina stretching anteriorly from the lateral side of the postzygapophysis. The neural spine is broad and low. Posteriorly, the neural arch bears slightly arched postzygapophyses. A large and rounded epipophysis (= torus dorsalis) extends posteriorly past the articular surface of the postzygapophysis.

The third cervical appears to have an osseous bridge from the transverse process to the postzygapophysis (Mayr and Clarke, 2003). The third cervical has a robust ventral process that spans the length of the ventral centrum body, whereas the fourth cervical has a low ventral process restricted to the anterior portion of the body. Ventral processes are absent in all the other preserved cervical vertebrae. The cervical vertebrae share a similar length, a depression on the posterior portion of the ventral surface, small tubercles on the ventral side of costal processes, and they lack fossae or foramina in the lateral sides of the centra. The costal processes are shorter than the length of the centrum and lack an osseous bridge connecting them with the midsection of the centrum of the vertebra (Mayr and Clarke, 2003: char. 53). The length of the costal processes in Calciavis is proportionally shorter than that of MGUH 26770 (referred to Lithornis vulturinus).

Six thoracic vertebrae exposed in ventral view (fig. 7) are loosely articulated with the synsacrum. The thoracic vertebrae do not form a notarium. A thin, anteroventrally expanded ventral process is present in the first of the articulated series (not necessarily the first thoracic vertebra). The remaining thoracic vertebrae have more abbreviated ventral processes. The transverse processes are well exposed in ventral view in the last three thoracic vertebrae. These expanded transverse processes terminate in straight lateral margins. Three deep fossae are located at the anterior, posterior, and ventral margins at the base of the transverse process. The parapophysis, located on the anterior edge of the body, is distinctly concave. Deep, anteroposteriorly elongated foramina (= central ovoid foramina of Clarke, 2004) are located in the middle of the body of the thoracic vertebrae (fig. 7), as in other lithornithids including Lithornis plebius (USNM 336534), Lithornis promiscuus (USNM 336535), Lithornis celetius (USNM 290601), MGUH 26770, and Paracathartes howardae (various USNM specimens). The foramina in lithornithids differ from the broad lateral fossae present in Ichthyornis dispar (YPM 1450C; Clarke, 2004) and Paraortygoides (Dyke and Gulas, 2002) in being much smaller and apparently perforating the body of the centra. The last thoracic vertebra of Calciavis is slightly shorter than that of the other thoracic vertebrae.

The synsacrum, exposed in ventral view (fig. 7), consists of 12 coossified vertebrae that can be divided into three distinct sections following Clarke (2004): anterior sacral vertebrae, a short series of wide vertebrae with dorsally directed transverse processes (= middle sacral vertebrae), and more posterior sacral vertebrae with a morphology similar to the anterior caudal vertebrae (fig. 7). The first section of the synsacrum consists of five vertebrae with laterally directed sacral transverse processes. The first sacral vertebra bears a free rib and has a wide anterior articulation surface for articulation with the last thoracic vertebra. The width of the bodies of the sacral vertebrae increases slightly across the first five vertebrae. A slight longitudinal depression is present on the ventral midline starting on the fourth sacral and continuing through the short series of midsacral vertebrae. The middle sacral vertebrae series consists of three elements with dorsally directed transverse processes. Weak furrows and paired sacral nerve foramina define the edges of the fused centra. Three morphologically distinct midsacral vertebrae are also present in other lithornithids (e.g., Lithornis plebius, USNM 336534; Lithornis promiscuus, USNM 336535; Lithornis celetius; USNM 290601) and in Ichthyornis dispar (YPM 1450, 1732), whereas there are four in ratites, tinamous, Apatornis celer (Clarke, 2004), and early diverging neognaths. The first sacral in the posterior series has posterolaterally directed transverse processes that are the longest of the series. These rodlike processes contact the ilium just posterior to the acetabulum as in other avialans (e.g., Lithornis promiscuus, USNM 336535, tinamous, and Ichthyornis dispar). The posterior sacral vertebrae progressively decrease in width and increase in length.

Calciavis (AMNH 30578) preserves the first complete series of caudal vertebrae in a lithornithid (fig. 7). In total, the series is short, measuring about two-thirds the length of the synsacrum. Seven free caudal vertebrae and an incompletely preserved pygostyle comprise the caudal series. No free chevrons are present. The first two caudal vertebrae are most similar to the last vertebrae in the synsacrum and are slightly larger both in width and length as compared with that of the other caudal vertebrae. The transverse processes of the first two caudal vertebrae are posterolaterally projected from the body like that of the last few sacral vertebrae. The next five caudal vertebrae are uniform in length and width. The anterior and posterior centrum ridges are poorly differentiated resulting in cylindrically shaped centra. Ventrally, there are no pneumatic foramina on the centrum as in some ratites (e.g., Dromaeus novaehollandiae). The pygostyle is narrow and approximately equal to the length of three of the diminutive posterior caudal vertebrae; it is proportionally longer than that of Crypturellus undulatus (AMNH 6480). The anterior portion of the pygostyle has a single short transverse process. Its ventral process is slightly clublike, as described for Lithornis promiscuus (USNM 336535). An anteriorly open foramen is present at the anteroventral margin of the pygostyle as in Lithornis promiscuus (USNM 336535), Lithornis plebius (USNM 336534), and Paracathartes howardae (USNM 361435). The posterior end, although broken, appears to terminate in a point as in Lithornis promiscuus (USNM 336535). In overall form, the tail is short, lightly built, and morphologically similar to those of tinamous (Houde, 1988) and other palaeognaths (Clarke, 2002).

Ribs

The ribs from the anterior portion of the thoracic series are disarticulated, and the ribs from the posterior portion of the thoracic series are loosely articulated with their respective vertebrae (fig. 7). The ribs from the posterior portion of the thoracic series broaden ventrally, are thickened at their distal ends, and bear uncinate processes (fig. 4). The elongated uncinate processes arc posterodorsally and taper to a point. The processes are not fused to their ribs, and ribs without articulated uncinate processes have a distinct scar for articulation with the element. Although Houde (1988) makes no mention of uncinate processes, the photograph of the Siber and Siber specimen (Houde, 1988: fig. 35) appears to preserve several incomplete uncinate processes. The remains of these processes appear consistent with smaller processes than those preserved in the holotype of the new species.

Pectoral Girdle

Sternum: The sternum, although partially crushed, is well preserved and exposed in left lateral view (fig. 4). The anterior portion of the sternum is covered. The overall proportions are similar to that of other lithornithids (e.g., Lithornis promiscuus, USNM 336535). The deep keel of the sternum is deepest anteriorly and terminates at the posterior extent of the element. The ventral margin of the keel is straight in lateral view. Anteriorly, the apex is short and rounded. The lateral surface of the body of the sternum is smooth, and there is no raised, paired intermuscular ridges visible parallel to the sternal midline. The posterior margin is slightly concave with no distinct posterior notches or trabeculae. The coracoidal sulci appear to cross slightly at the midline as in Lithornis promiscuus (USNM 336535) and there appear to be a minimum of four processes for articulation with the sternal ribs.

Coracoid: The left coracoid is preserved in ventral view (fig. 4). The midshaft is narrow like that of other lithornithids (e.g., Lithornis promiscuus; USNM 336535) and rapidly expands into a broad sternal margin that is dorsoventrally flattened. A lateral intermuscular ridge is present stretching from near midshaft to the lateral 1/3 of the articular facet with the sternum. The lateral process tapers to a posterodorsally directed apex as in Lithornis promiscuus (USNM 336535), Lithornis plebius (USNM 336534), and Paracathartes howardae (USNM 361417). Proximally, a prominent scar for the m. acrocoracoidohumeralis ligament is visible in lateral view. A distinct rounded rim surrounds this scar, and a small pit is located at the posterolateral portion. The surface of the glenoid is flat. The shape of the procoracoid process and the presence or absence of a foramen for the supracoracoid nerve cannot be determined.

Scapula: The left scapula remains in articulation with the left coracoid (figs. 7, 8) whereas much of the right element is buried underneath ribs and vertebrae in the thoracic region. The scapula is at least three-quarters the length of the humerus. The proximal portion bears an elongated and oval glenoid facet. Most of the acromion is covered by other elements. The scapular shaft gradually decreases in dorsoventral width posteriorly, as in other lithornithids (e.g., Lithornis promiscuus, USNM 336535) and other Aves. The curvature of the shaft is more similar to that of Lithornis promiscuus (USNM 336535) and Lithornis plebius (USNM 336534) in that it is straighter than the more recurved scapula of Paracathartes howardae (USNM 361419). The distal tips of the scapulae are covered by other elements of the skeleton.

Forelimb

Humerus: The left humerus is exposed in posterior view (fig. 8) whereas the right humerus is mostly covered by other skeletal elements (fig. 4). The humerus is slightly longer than the ulna and the shaft of the humerus is sigmoid as described by Houde (1988) for other lithornithids. Proximally, the head of the humerus is rounded and well defined. The ventral tubercle is knoblike and located at the distal margin of the humeral head, as in other lithornithids (e.g., Paracathartes howardae, USNM 361420; Lithornis plebius, USNM 336534). The dorsal rim of the pneumotricipital fossa is present distal to the ventral tubercle. A wide, mediolaterally oriented capital groove separates the humeral head from the ventral tubercle. The capital groove is open in AMNH 30578 unlike the condition in tinamous (e.g., Crypturellus) and galliforms (e.g., Gallus), where the capital groove terminates distally in a tubercle (Clarke and Chiappe, 2001). The deltopectoral crest is obscured in both humeri.

The distal ends of the humeri are slightly crushed and broken. The m. humerotricipitalis sulcus is well defined and deep. The m. scapulotriceps groove is shallow to essentially absent in Calciavis, other lithornithids (Houde, 1988), tinamous, and most nonavian avialans (Clarke, 2004). The flexor process is weakly projected distally, but extends just slightly distal to the ventral condyle. A dorsal supracondylar tubercle is present at approximately the same proximal height as the dorsal condyle, as in other lithornithids and tinamous. In the flightless palaeognaths, this tubercle is unsurprisingly typically reduced along with the wing musculature. A small depression lies on the proximal edge of the dorsal supracondylar process.

Ulna and Radius: The left ulna is exposed in ventral view, and the proximal and distal ends of the right ulna are exposed in dorsal view (fig. 8). The bowed shaft of the ulna is slightly shorter than the humerus. This observation is consistent with most lithornithids (e.g., Lithornis plebius, USNM 336534; Lithornis promiscuus, USNM 336535; Siber and Siber specimen, USNM 424078; MGUH 26770). No conspicuous or raised feather papillae are present. However, these are preserved in the referred specimen (AMNH 30560). The olecranon process is short, but projects proximally to the cotylae. A low, elongate scar marks the attachment site for the m. biceps on the ulna. A thin rim defines the proximal and posterior portions of the brachial depression. However, the apparent depth of the depression may be the result of crushing of the specimen. The distal portion of the ulna bears a fingerlike carpal tuber that is well separated from the distal articular surfaces of the ulna. The intercondylar sulcus is shallow as in other lithornithids and tinamous, and a sharp ridge separates the ventral ulnar condyle and the carpal tuber.

The left radius is covered proximally by the humerus (fig. 8) and the pelvis covers the right radius distally. The slender radius has an essentially straight shaft. The diameter of the radius at its midpoint is about half that of the ulna. In proximal view, the articular surface is concave and round. A distinct bicipital tubercle is present, and a small, deep pit lies next to this tubercle.

Proximal carpals: Both the ulnare and radiale remain in articulation on the left side (fig. 8), but the wrist is disarticulated on the right side (fig. 4). The metacarpal incisure of the ulnare is deep, but it is not clear how the lengths of the rami compare to each other; in tinamous and other lithornithid specimens these rami are subequal in length (Clarke, 2002). A small tuber is present on the posterior side where the dorsal and ventral rami meet as in Galloanserae (e.g., Chauna torquata) and in lithornithids (Lithornis plebius, USNM 336534; Lithornis promiscuus, USNM 336535). The ventral ramus bears a deep tendinal groove as is present in other Aves and Ichthyornis dispar (SMM 2503; Clarke, 2004).

Carpometacarpus: The left carpometacarpus is exposed in ventral view (fig. 8), and the right carpometacarpus is disarticulated and exposed in dorsal view (fig. 7). The total length of the carpometacarpus is slightly more than half the length of the humerus. The slender carpometacarpus has a relatively wide metacarpal II (= major metacarpal) and straight metacarpal III (= minor metacarpal) circumscribing a narrow intermetacarpal space. The intermetacarpal space does not extend proximally to reach the distal end of metacarpal I. As in other lithornithids (Houde, 1988), metacarpal III is ventrally shifted relative to metacarpal II. Proximally, the posterior side of the carpal trochlea has a proximodistally elongated deep pit (= fovea carpalis caudalis). A similarly deep pit is also present in some Galloanserae (e.g., Gallus) and lithornithids (e.g., Lithornis plebius, USNM 336534; Lithornis promiscuus, USNM 336535; Paracathartes howardae, USNM 361445) but clearly absent in tinamous (e.g., Crypturellus undulatus, AMNH 6480) (Clarke and Chiappe, 2001).

The knoblike pisiform process is bordered by two fossae, one proximal and one anterior. The fossa proximal to the pisiform (fossa infratrochlearis) excavates the proximal edge of the pisiform process (Clarke and Chiappe, 2001; Clarke et al., 2006: char. 148). The pisiform process is connected to the ventral surface of metacarpal III by a distinct ridge. A tubercle is present at the proximoventral portion of metacarpal III. Most basal avians have a scar or slightly raised ridge here also. A distinct tubercle is present in AMNH 30578, Lithornis promiscuus (USNM 336535), Lithornis celetius (USNM 290554), and Lithornis plebius (336534), but not in Paracathartes howardae (USNM 361446). A deep pit just posterior to the pisiform process is present in AMNH 30578, other lithornithids (e.g., Lithornis plebius, L. promiscuus, L. celetius, and Paracathartes howardae), and tinamous (e.g., Crypturellus). There is no intermetacarpal process. A shallow tendinal sulcus is present on much of the dorsal surface of metacarpal II. Metacarpal III is slightly longer in distal extent than metacarpal II. A cup-shaped alular facet is present on the distal end of metacarpal I. The tip of extensor process conspicuously surpasses this facet for digit I-1 by approximately the width of the facet, producing a pronounced knob (Clarke et al., 2006: char. 144).

The length of the alular digit measures about half the length of the carpometacarpus. This digit in the Siber and Siber specimen referred to Pseudocrypturus is considerably shorter and is roughly one-third the length of the carpometacarpus (Houde, 1988: fig. 34). The proximal articular facet is the widest portion of the alular digit in Calciavis, and the element tapers distally. The distal end of digit I-1 has a wide facet for the alular ungual for a total of two phalanges. The ungual is slightly posteriorly recurved, tapers to a point, and bears a lateral groove that originates near the proximal end and terminates at the tip of the element.

The second manual digit is known from the combination of preserved portions of both the left and right sides. Calciavis does not have a proximally directed process on the anteroventral face of the proximal end as in tinamous, Ichthyornis dispar (Clarke, 2004), and some Galliformes. An internal index process is not present. Phalanx II-2 is a thin element that tapers distally. The proximal portion is rounded in cross section, and a small tuber is present on its anterodorsal edge. The body of the phalanx expands slightly posteriorly at approximately its midpoint where there is a depression on the dorsal surface. A similar expansion is present in the Siber and Siber specimen (USNM 424078), but is not present in Crax pauxi (AMNH 4719) or Crypturellus undulatus (AMNH 6480). Phalanx II-2 terminates in a blunt, poorly defined articulation surface. It is not clear whether an ungual was present. However, a tiny ungual is present on digit II in the exquisitely preserved lithornithid specimen (AMNH 30560) referred to Calciavis grandei (see description below).

The third manual digit consists of a single phalanx III-1. This digit is short, measuring less than half the length of phalanx II-1 as in tinamous (e.g., Crypturellus undulatus, AMNH 6480). The short phalanx terminates distally in a rounded point. The posterior side bears a distinct flexor tubercle giving phalanx III-1 a triangular shape. This shape is in contrast to the long, rodlike phalanx III-1 of MGUH 26770 (referred to Lithornis vulturinus; Bourdon and Lindow, 2015).

Pelvic Girdle

Pelvis: The ilia, ischia, and pubes are exposed in either ventral or medial views (fig. 7). The preacetabular portion of the ilium is longer than the postacetabular portion. The anterior margin of the preacetabular portion is slightly concave in lateral view. This shape is in contrast to the distinctly convex anterior margin of most palaeognaths including Lithornis celetius (PU 23485) and to the straight margin of MGUH 26770, referred to Lithornis vulturinus (Leonard et al., 2005; Bourdon and Lindow, 2015). The anterolateral corner of the ilium is rounded and not distinctly expanded as it is in tinamous (e.g., Crypturellus undulatus, AMNH 6480). Scars for the attachment of the sacral ribs are present in a longitudinally oriented row near the ventral margin of the preacetabular portion, just anterior to the acetabulum. It appears that the preacetabular portions of the ilia did not contact at the midline. The ventral portion of the postacetabular portion of the ilium bears a distinct, ventral pocket for the brevis musculature. A distinct ridge for the ilioischiadic membrane (Houde, 1998) is visible. It is also present in Lithornis promiscuus (USMN 336535) and tinamous. The posterior margin of the ilium is rounded.

The postacetabular portion of the ilium does not contact the ischium ventrally as in neognaths, and therefore, there is no enclosed ilioischiadic fenestra. The posterior margin of the ischium is straight, similar to the condition in Lithornis promiscuus, Apteryx, and tinamous, but unlike specimens of Lithornis plebius (USNM 336534) and a specimen that was referred to Pseudocrypturus cercanaxius (WN 80280; Houde, 1988). The obturator foramen is open posteriorly and framed posteriorly by a distinctly rounded and robust tuber on the ventral border of the ischium. The straight ventral and dorsal edges of the ischium slightly diverge from each other posteriorly resulting in a slight increase in height posteriorly relative to the anterior portion (fig. 7).

The right pubis is completely preserved in AMNH 30578 in medial view. The pectineal process is well developed as a knob that slightly projects anterior to its articulation with the ilium. The pectineal process is similar in shape to that of Lithornis plebius (USNM 336534) and Lithornis promiscuus (USNM 336535), but it is not as expanded anteriorly as in Lithornis celetius (USNM 290601) and tinamous (e.g., Crypturellus undulatus, AMNH 6480). The pubis is directed posteriorly paralleling the ischium, and is a third longer than the ischium. The shaft of the pubis does not contact the ischium throughout its length; it is straight, remains the same height and width posteriorly, and ends in a blunt termination. The proximal end does not appear to have a pectineal process on the ventral margin below the acetabulum as in Lithornis promiscuus (USNM 336535).

Hind Limb

Femur: The left femur is exposed in posterior view, and the right femur is exposed in anterior view (fig. 4). The largely straight shaft of the femur bows slightly anteriorly. The trochanteric crest is only weakly projected proximally, as in other lithornithids (Houde, 1988). The rounded head of the femur has a distinct neck. Laterally, the muscle scars for the m. ischiofemoralis and the m. iliotrochantericus are deep and well developed. Prominent anterior and posterior intermuscular lines extend the length of the femur. The popliteal fossa is shallow like that of other lithornithids (e.g., Lithornis promiscuus, USNM 336535; and Paracathartes howardae, USNM 361412). The distal condyles are approximately equal in distal extent. The medial condyle is set well apart from the shelflike lateral and tibiofibularis crests. The posterior surface of the medial condyle is nearly flat.

Tibiotarsus: The left tibiotarsus is exposed in posterolateral view (fig. 9), and the right tibiotarsus is in medial view (fig. 4). Proximally, the anterior and lateral cnemial crests are weakly projected proximally and anteriorly, respectively. The anterior cnemial crest is slightly longer than that of the lateral crest, but the lateral crest is more projected. The base of the anterior crest is slightly anteriorly expanded. A similar expansion is present and much more developed in Lithornis plebius (USNM 336534). As in Crypturellus undulatus, the base of the anterior cnemial crest of AMNH 30578 forms a distinct obtuse angle with the shaft. The lateral cnemial crest is thickened distally into a flat lateral surface of the tibia. The lateral articular surface is concave. The lateral side of the shaft bears a sharp, elongated fibular crest on the lateral side that extends for a length about one-fourth the length of the element. Distally, the medial condyle projects further anteriorly than the lateral condyle, but it is difficult to compare the relative size of the two condyles given their angle of preservation. The intercondylar sulcus is wide, shallow, and defined proximally. An extensor (= tendinal) groove is present and positioned proximomedial to the intercondylar space and exits nearly adjacent to the proximal edge of the medial condyle. There is no ossified supratendinal bridge as in most nonavian avialans (e.g., Clarke, 2004), lithornithids (Houde, 1988), some ratites (Cracraft, 1974), and owls. The position of the groove in Calciavis (AMNH 30578) and other lithornithids is similar to that of tinamous and other palaeognaths with a supratendinal bridge.

Fibula: The left fibula is exposed in lateral view (fig. 9). The element tapers distally and is approximately three-quarters the length of the tibiotarsus. Proximally, the fibula is asymmetrical in lateral view with the posterior portion is more pointed than the anterior margin. Its anterodorsal corner bears a deep pit. A similar scar also is present in Crypturellus undulatus (AMNH 6480), but is located more anteriorly. Another depression is located on the lateral side of the proximal end. The m. iliofibularis attachment is located on the posterior margin and is developed as a thin flange with a rugose distal end.

Tarsometatarsus: The right tarsometatarsus is preserved in medial view (fig. 10), and the left tarsometatarsus is in plantar view (fig. 11). The proximal tarsals are completely fused with the metatarsals without a visible suture. A small sesamoid is present proximal to the planter portion of the tarsometatarsus. The hypotarsus is proximodistally short and expanded plantarly with a weakly projected medial crest and no enclosed canals as in all known lithornithids (e.g., Lithornis promiscuus, USNM 336535; Lithornis plebius, USNM 336534; Paracathartes howardae, USNM 361406). Similarly sized, paired proximal vascular foramina exit at approximately the same proximodistal level on the plantar surface. Medial and lateral plantar crests frame a shallow sulcus extending much of the length of the tarsometatarsus. Metatarsal I is short and articulates with the medioplantar side of the tarsometatarsus at a very slight depression; it is proximally tapered with a blunt apex.

Trochlea II and IV are dorsoplantarly rounded. Trochlea III is ginglymoid, whereas trochlea II and IV have very weak to absent furrows. The articular surface of trochlea sal III is symmetrical in plantar view and poorly expanded proximally. In Lithornis plebius (USNM 336534), Lithornis promiscuus, (USNM 336535), and Paracathartes howardae (USNM 361406), trochlea III is slightly asymmetrical in plantar view. The distal extent of trochlea II and IV of Calciavis grandei are equal. The distal vascular foramen and distal interosseus canal are small, and exit plantarly and distally, respectively. Small subcircular structures adjacent to the posterior side of the right tarsometatarsus likely represent scales preserved as carbonized traces (fig. 10).

Pedal digits: The digits are articulated in the right pes and slightly disarticulated in the left pes (figs. 10, 11). Generally, the phalanges are elongated and gracile, and the distal ends are ginglymoid. The most proximal phalanges in each digit are the longest in the series. Phalanx digit I-1 has the smallest shaft diameter. Pedal phalanx VI-4 is longer than that of phalanx IV-2 and IV-3, and phalanx III-3 is slightly longer than phalanx III-2. The unguals have a curvature of ~30° measured from dorsal surface. There is a small flexor tubercle and deep lateral grooves are present. Remains of keratin indicate that the claws were ~30% longer than the bony portion of the unguals.

Soft Tissue

Feathering: The feathers are preserved as carbonized traces and impressions in the matrix (figs. 4, 7, 8, 10). Disorganized patches of feathers lie near the pectoral and pelvic girdles. The patches range in size from 20 to 40 mm in diameter. Both forelimbs have remiges associated with them. The left wing looks intact, but the impressions and traces have been damaged. The feather attached to manual phalanx I-1 and I-2 measures 76 mm and is parallel to the shaft of the phalanges. A primary feather associated with manual phalanx II-2 lies parallel to manual digit II. Other primary feathers are present, but further details cannot be discerned.

The preservation of the right wing feathering is better than that of the left and barbules can be discerned in several primaries. The number of primary feathers in this specimen is not clear. An unspecified number of primary and secondary feathers lie on top of each other and little can be discerned about individual feather lengths. It is unclear whether tail feathers are preserved because parts of the right wing feathering cover the caudal area (fig. 4).

Lithornithidae FMNH PA 729 Figure 16

FIG. 16.

The left forelimb of FMNH PA 729 in dorsal view with exquisite preservation of the covert and primary feathers (in dark gray). Scale = 1 cm. See anatomical abbreviations.

Locality: Thompson Ranch (locality H, F-2 deposits of Grande and Buchheim, 1994) within the Fossil Lake deposits of the Green River Formation, near Kemmerer, Wyoming.

Preservation: The specimens consist of a complete, articulated left forelimb with an articulated scapula and coracoid. Parts of the forelimb (e.g., metacarpal II, coracoid) are missing, but impressions of those missing segments still remain in the matrix. The specimen is crushed, but preserves faint impressions of the wing feathering.

Description: The left coracoid is preserved in dorsal view and, unfortunately, much of the glenoid region of the element is not well preserved. A lateral process is not preserved. The dorsal surface of the coracoid is concave. The medial edge tapers to a thin point. The scapula is preserved in medial view, although the proximal portion and the distal extremity are missing. The coracoidal tubercle is convex, as in other lithornithids. The shaft of the scapula gradually tapers like that of Lithornis promiscuus (USNM 336535) and AMNH 30560, rather than that of FMNH PA 739. The length versus width and arc of curvature of the shaft of the scapula is similar to that Lithornis promiscuus (USNM 336535) and AMNH 30560, rather than the more robust and curved shaft of Paracathartes (USNM 361419). The acromion, crushed under the coracoid, is pointed and recurved as in other lithornithids.

The humerus lies in posterodorsal view and has a sigmoid-shaped shaft. The proximal surface is well rounded, and there is broad capital incisure between the head of the humerus and the ventral tubercle. The deltopectoral crest is weakly expanded anterodorsally and distally grades smoothly into the shaft. Much of the distal portion of the humerus is covered by the ulna or matrix. Proximally, the ulna bears a low olecranon process. Distally, its dorsal condyle has a flat distal margin and a weakly developed tendinal pit. A paired series of slightly raised scars on the posterodorsal side of the ulna represent the anconal and inner papillae for the attachment of feathers. There appears to be 10 pairs of papillae that stretch from the proximal to the distal end (fig. 16). The proximal portion of the radius is mostly covered, but a pronounced bicipital tubercle is present near the proximal surface. The shaft of the radius is straight, and the shaft of the ulna has a low arc. A distinct muscular ridge is present on the posterodorsal edge of the radius for the length of the element.

The manus lies in dorsal view. The proximal carpals are largely buried in matrix. The carpometacarpus has a distinct pit (= supratrochlearis fossa) on the dorsal surface just distal to the carpal trochlea. Both metacarpals II and III have straight shafts enclosing a narrow intermetacarpal space, although much of metacarpal II is preserved only as an impression (fig. 16). Metacarpal III just surpasses metacarpal II in distal extent, and the shaft of metacarpal III is “twisted” (Houde, 1988). Phalanx I-1 is the same relative length as that in the holotype of Calciavis and articulates with a small ungual. Phalanx II-1 is poorly preserved and lacks an index process on the distal end. Like tinamous and lithornithids, phalanx III-1 is less than half the length of phalanx II-1, and is triangular, as a result of possessing a large flexor process. The shaft of phalanx II-3 tapers and has a posterior expansion at approximately its midpoint, but the expansion is not a great as that present in the Calciavis holotype and the referred specimen AMNH 30560.

Feathering: The feathers are preserved as faint impressions in the matrix without carbonized traces, as in all the other Green River Formation lithornithids. A detailed account of feathering in this specimen was presented in Torres and Clarke (2011).

Phylogenetic assignment: FMNH PA 729 was found within Lithornithidae in our phylogenetic analysis (see below). The poor preservation of the preserved elements inhibits the assignment of FMNH PA 729 to a species-level taxon. The limb proportions, absolute size, and observable morphology are consistent with other lithornithids especially AMNH 30578 (see table 1), but there are no unambiguous character states or a unique combination of character states linking the specimen to Calciavis grandei. Furthermore, phalanx III-1 of the manus is short and triangular like that of lithornithids, but this character state is also in other palaeognaths (e.g., tinamous). Therefore, we assign this specimen to Lithornithidae and this is supported in the phylogenetic analysis (see below).

FMNH PA 739 Figures 17–19

FIG. 17.

The incomplete skeleton of FMNH PA 739. Scale = 5 cm. See anatomical abbreviations.

FIG. 18.

The partial skull of FMNH PA 739 in dorsal view. Scale = 1 cm. See anatomical abbreviations.

FIG. 19.

The hind limbs of FMNH PA 739. The right pes is in anterior/dorsal view whereas the left pes is in medial view. Scale = 1 cm. See anatomical abbreviations.

Locality: Thompson Ranch (locality H, F-2 deposits of Grande and Buchheim, 1994), within the Fossil Lake deposits of the Green River Formation, near Kemmerer, Wyoming.

Preservation: The specimen consists of much of an articulated skeleton, missing the posterior two-thirds of the skull, much of the pectoral girdle, the crux region of the left wing, and much of the right wing. No feathers are preserved, but small traces of the ungual sheaths are preserved. Some regions are well preserved, but many of the long bones are badly crushed. Some portions of the skeleton partially deteriorated before fossilization (e.g., the pygostyle), and as a result some of the details cannot be discerned.

Description: Only parts of the back of the skull and the anterior third of the skull were recovered with FMNH PA 739 (fig. 18). The anterior portion of the skull is crushed dorsally where only the dorsal portions of the nasals, premaxillae, and the occlusal margin of the mandible are visible. The anteriorly tapering and robust nasal bar is similar to that of other palaeognaths. The lateral sides of the premaxillae bear furrows anterior to the external naris as in other palaeognaths (Mayr and Clarke, 2003). Small foramina in a single anteroposteriorly oriented row dot the flat occlusal margin of mandible (fig. 18), a character state present in lithornithids (e.g., Paracathartes), Apteryx (USNM 18279), and some neognaths (e.g., Gallus).

The cervical vertebrae are missing in the specimen, and the thoracic vertebrae are scattered near the pelvis and forelimb (fig. 17). The vertebrae have clear heterocoelous articulations and neural spines equal to the height of the centrum. A distinct foramen is present in the lateral side of the centrum as in other lithornithids (see above). A notarium is not present. Thoracic ribs are mediolaterally broad and bear small crenulations on the anterior and posterior surfaces, features associated with pneumaticity. The dorsoventrally flattened synsacrum is composed of a minimum of 10 vertebrae, and the anterior centrum bodies are longer than those of the more posterior ones. The articulations with the ilium are difficult to discern. The caudal vertebral series terminates in a small pygostyle. Individual centrum bodies are poorly preserved. However, they visibly decrease markedly in size posteriorly as in extant palaeognaths. The pygostyle is too poorly preserved to compare in detail to other lithornithids.

The sternum is crushed in left lateral view, and the anterior portion and left lateral edge are missing. A deep ventral keel is present, and a minimum of four sternal ribs attach to the sternum. It is not clear whether the posterior edge of the sternum was straight or concave. Only the distal end of the scapula is preserved. The scapula terminates in a blunt, gently rounded distal end; this character state is shared with AMNH 30560, but absent in the more tapered scapulae of FMNH PA 739 and the holotypes of Lithornis promiscuus and Lithornis plebius.

Both humeri are preserved, but severely crushed. The left humerus is in anterior view, whereas the remains of the right humerus are in posterior view. The base of rounded humeral head bears a shallow acrocoracoid ligament furrow on its anteroventral margin. Distally, the humeral condyles are about equal in size and well rounded. A distinct dorsal tubercle is present, as in other lithornithids (Houde, 1988; Clarke and Chiappe, 2001). The proximal ends of the left ulna and radius are in near articulation with the humerus (fig. 17). The olecranon process of the ulna is distinct but short. The radius has an articulation surface for the ulna distally. The proximal radius bears a distinct bicipital scar.

Parts of the left manus are preserved. Metacarpals II and III (major and minor, respectively) are parallel. There does not appear to be any expansion of the distal end of the articular surface of phalanx II-1 into an index process. Phalanx III-1 is triangular and less than half the length of phalanx II-1, as in tinamous and other lithornithids. Phalanx II-2 tapers anteriorly, but bears a distinct expansion, as in AMNH 30560. A small ungual is present on the tip of phalanx I-1.

The pelvic region is crushed. The preacetabular portion of the ilium is longer than the tapering posterior portion. The acetabulum is round in lateral view, and the antitrochanter is not expanded posterior of the posterior margin of the ischial peduncle, as in the holotype of Lithornis promiscuus (USNM 336535). The ischium and the ilium do not contact distally to enclose an ilioischiatic fenestra, as in other palaeognaths. The distal end of the pubis tapers to a point.

The femora are poorly preserved. The neck of the femur is poorly preserved, and the lateral surface of the proximal surface is not well expanded proximal to the head. The shaft of the femur is straight, and the distal ends are heavily crushed. The proximal ends of the tibiotarsi are either hidden or severely crushed. The distally tapering fibula is about three-quarters of the length of the tibiotarsus. Distally, an ossified supratendinal bridge appears to be absent. The pedes are well preserved (fig. 19) and have overall similar proportions to that of the other Green River Formation lithornithids (table 1). The plantar portion of the tarsometatarsus bears one proximodistally extensive lateral hypotarsal ridge and a relatively smaller medial hypotarsal ridge.

Phylogenetic assignment: FMNH PA 739 was found within Lithornithidae in our phylogenetic analysis (see below). Only a portion of the skull is present, so we are unable to estimate cranial or rostral length with confidence. Hence, we cannot confidently assign FMNH PA 739 to either Calciavis grandei or to Pseudocrypturus cercanaxius. The limb proportions are similar to that of the referred specimen (Siber and Siber) of Pseudocrypturus cercanaxius (table 1). FMNH PA 739 can be assigned only to Lithornithidae at present.

WSGS U1b-2001 Figures 20–22

FIG. 20.

Skeleton of WGS U1b-2001 (in white) with feather preserved (in dark gray). Scale = 5 cm. See anatomical abbreviations.

FIG. 21.

A, Distal end of the left tibiotarsus of WGS U1b-2001 in anterior view. B, Complete left pes of WGS U1b-2001 in anterior view. Scales = 1 cm. See anatomical abbreviations.

FIG. 22.

The well preserved wing feathers of WGS U1b-2001. The white arrows point to the rachis and the black arrow the folding of the feathers. Scale = 1 cm.

Locality: Ulrich Quarry within the Fossil Butte Member deposits of the Green River Formation (Locality E of Grande, 2013).

Preservation: This specimen consists of a largely disarticulated skeleton made up of the hind limbs, a partial right forelimb, thoracic ribs, parts of the synsacrum, and ilium, but the skeleton is missing much of the pectoral girdle, the vertebrae, and skull. Some of the bones have deteriorated prior to fossilization (e.g., pelvis, femur), and segments of some of the bones are missing, but impressions remain (fig. 20).

Description: Of the axial column, only thoracic ribs and part of the synsacrum are preserved. It appears that uncinate processes of the thoracic ribs were unfused to the ribs, but it is possible that they were cartilaginous. The poorly preserved coracoid is possibly the only remnant of the pectoral girdle, but there appears to be a small portion of the furcula near the preserved portion of the coracoid. The coracoid has a narrowly waisted shaft like that of other lithornithids (Houde, 1988). The sternal articulation is restricted to the distal surface, and there does not appear to be any distinct lateral or medial processes. Much of the right humerus is preserved except for the proximal end. The deltopectoral crest is distinctly rounded, and it has a low angle where the crest meets the shaft. The deltopectoral crest appears to be more anterodorsally expanded and proximally restricted in WSGS U1b-2001 than in Lithornis celetius (PU 23485), L. promiscuus (USNM 336535), and Paracathartes howardae (USNM 361420). The shape and size of the crest in WSGS U1b-2001 is similar to that of the Siber and Siber specimen (Houde, 1988: fig. 34) and tinamous, in contrast to other lithornithids (e.g., Lithornis promiscuus). Unfortunately, the other Green River Formation specimens are not comparable as the deltopectoral crest is covered or not preserved in all cases. Distally, the m. scapulotriceps groove on the posterior surface is not visible. The dorsal epicondyle and flexor processes are only weakly expanded in this specimen. The humerus is about the same length as the ulna (fig. 20). Most of the ulna and radius are preserved as impressions, and only the distal ends of both elements are preserved, but extensively crushed. The distal end of the radius is rounded, and the distal end of the ulna is compressed and rounded. The shaft of the radius is straight. A complete right ulnare is present, and the dorsal and ventral processes are about the same size. The ventral ramus has a very deep tendinal groove. The midshafts of metacarpals II and III are the only portions preserved of the carpometacarpus. Both are straight, and few other details are visible. Impressions of manual phalanges II-1, II-2, and I-2 are present, but uninformative.

The pelvis and the posterior portion of the synsacral vertebrae are poorly preserved. Both femora are present, but these elements are heavily crushed like the pelvic elements. The distal half of the right tibiotarsus lies in anterior view. There is a tendinal groove proximolateral to the medial condyle, and the ossified supratendinal bridge is absent as in other lithornithids, many ratites, and most proximal outgroups to Aves (Clarke, 2004; but see Iaceornis). The distal condyles are roughly the same size (fig. 21).

Both pedes are preserved in articulation in dorsal view. The tarsometatarsus bears a short, narrow intercotylar prominence on the proximal surface. Two similarly sized and small proximal vascular foramina occur at the same proximodistal level and lie in a shallow extensor sulcus. Trochlea II is slightly shorter than trochlea IV in distal extent. Metatarsal I is short and lies near the distal end of trochlea II. The phalanges are in articulation. Phalanx I-1 is slightly arched in lateral view (fig. 21). Digit III is the longest, and the phalanges are the most robust relative to the other digits. Like the other Green River Formation lithornithids, the first phalanx is longer than or equally long as any other phalanx in the same digit. The ungual on digit III is the largest, and the other unguals are nearly the same size.

Feathering: Feather preservation is exquisite in WSGS U1b-2001. Feather patches are present around the entire specimen, and long feathers of the wing (remiges) diverge from the partially preserved forelimb (fig. 22). The long feathers are generally well preserved and details of individual rachises and barbules are easily discernible. The rachises are preserved with darkened edges and lighter centers. Degradation that must have occurred in early diagenesis is visible in the flight feathers (fig. 22). The feathers are preserved as carbonized traces without textured impressions in the matrix.

Phylogenetic assignment: The assignment of WSGS U1b-2001 to a species-level taxon is difficult given the lack of a skull and other parts of the skeleton. However, the preserved morphology and proportions are consistent with an assignment to Lithornithidae, and this specimen was found within Lithornithidae in our phylogenetic analysis (see below). It lacks an ossified supratendinal bridge (fig. 21). The humerus has a short, rounded deltopectoral crest is similar to that of the Siber and Siber specimen referred to Pseudocrypturus cercanaxius. The absolute size and proportions of the limb bones are consistent with the Siber and Siber specimen (= Pseudocrypturus) except for the tarsometatarsus, which is ~10% larger than expected as compared to that of the tibiotarsus. Thus, we can refer WSGS U1b-2001 to the Lithornithidae, but do not find support for referral to a more specific taxon within the clade.

Taxon Sampling

Our data matrix includes a combination of extinct and extant taxa at the base of Aves (= crown) in order to evaluate the phylogenetic relationships among lithornithids. All extant and extinct taxa, with the exception of parts of the skeleton of Emeus crassus, were scored by directed observation. Exemplar species-level taxa for palaeognaths include Dinornithidae (Emeus crassus), Struthionidae (Struthio camelus), Tinamidae (Eudromia elegans, Crypturellus undulatus, Nothura maculosa, and Tinamus major), Apterygidae (Apteryx owenii), Rheidae (Rhea americana), and the emu (Dromaeus novaehollandiae) were used in order to circumvent problems associated with composite higher-level taxa. Casuarius was not included; in all recent molecular analyses Casuarius has been recovered as the sister taxa of Dromaius (e.g., Phillips et al., 2010; Mitchell et al. 2014; Baker et al., 2014). Evaluating the phylogenetic position of elephant birds (Aepyornithidae) was outside the scope of this analysis, and published illustrations of their anatomy are inconsistent.

Two extant exemplar species were used to represent Galliformes (Crax pauxi and Gallus gallus), and two to represent Anseriformes (Anas platyrhynchos and Chauna torquata). Gavia immer was included to sample non-Galloanserae neognaths, as the same genus (i.e., Gavia stellata) was utilized in the analyses of Phillips et al. (2010). Proposed stem members of crown subclades within Aves (i.e., Struthioniformes, Anatidae) include Gallinuloides wyomingensis, Anatalavis oxfordi, and Palaeotis weigelti. Specimen numbers or literature sources for each taxon are reported in table 2.

TABLE 2

Specimens and literature sources for the scoring of taxa in the phylogenetic analysis

See institutional abbreviations.

Each species-level taxon of lithornithid was scored independently in the phylogenetic analysis (appendix 1). Each specimen was first scored and then combined into these species-level terminal taxa for the “core” analysis. In the cases of Lithornis plebius and Calciavis grandei, the scores were based solely on the holotypes (USNM 336534 and AMNH 30578, respectively). For Lithornis celetius, the scores were based on the holotype (USNM 290601) and other specimens referred by Houde (1988) from the type locality (USNM 290554, USNM 336200, PU 23483, PU 23484, and PU 23485). The scorings of Lithornis promiscuus were based on the holotype (USNM 336535) and a skull (USNM 391983) referred by Houde (1988) found at the same quarry. The scores for Paracathartes howardae were based only on referred specimens of a minimum of five individuals (USNM 361402-361446, USNM 391984, USNM 404747-404806; USNM 361415) that were found in the same calcareous nodule (Houde, 1988). The scores of Pseudocrypturus cercanaxius derive from the holotype skull (USNM 336103) and a referred specimen (USNM 424078, cast of Siber and Siber lithornithid of Houde, 1988). MGUH 26770, an exquisitely preserved specimen recently referred to Lithornis vulturinus (Leonard et al., 2005; Bourdon and Lindow, 2015) was not seen by either of us and, therefore, scored from the description and additional pictures. The specimen was included in the “total analysis” but not the “core lithornithid” analysis because we were unable to confidently score some of the characters. More fragmentary lithornithids were excluded from this analysis because the scorings for those specimens were exactly the same as those for other lithornithid terminal taxa (e.g., Lithornis vulturinus, Lithornis nasi).

Outgroup selection for analyses of Aves remains an outstanding problem in investigations of avian relationships because of highly incomplete skeletons and specialized ecologies (e.g., full aquatic). Here, we used Apsaravis ukhaana Norell and Clarke, 2001, and Ichthyornis dispar Marsh, 1872 (sensu Clarke, 2004), as outgroups. Apsaravis ukhaana and Ichthyornis dispar have been found outside Aves in many phylogenetic analyses of avialan relationships (e.g., Chiappe, 2002; Clarke and Norell, 2002; Clarke, 2004; Clarke et al., 2006; You et al., 2006).

Character Sampling

The data matrix consists of 182 morphological characters scored for 38 terminal taxa. Of the original 182 characters, 18 were found as parsimony uninformative in the total analysis, and 20 characters were found as parsimony uninformative in the “core lithornithid” analysis. The parsimony-uninformative characters were kept in the data set to facilitate future expansion of character and taxonomic sampling, given that the parsimony-uninformative characters found here have supported clades in previous analyses (e.g., Clarke, 2004; Clarke et al., 2006). Characters 20, 25, 57, 59, 65, 68, 69, 70, 81, 83, 86, 104, 107, 118, 119, 137, 140, 154, 166, 168, 173, and 178 were ordered according to previous usage of the characters in other data sets and evidence of nested morphologies (Slowinski, 1993) for new characters. Our matrix combines characters from the morphological data sets of Cracraft (1974), Bledsoe (1988), Lee et al. (1997); Mayr and Clarke (2003), Clarke (2004), and Clarke et al. (2006) and new observations by the authors. The Bourdon et al. (2009) character set (128 characters) unfortunately included many composite character states (referencing multiple morphologies in one character set up as a binary with no description of the plesiomorphic state, e.g., “Ossa digiti IV, phalanges proximales et intermediate gradually shortening toward phalanx ungualis, so that phalanx intermedia just proximal to the latter is either wider than long or nearly square in shape: no (0); yes (1)”; Bourdon et al., 2009: 63, char. 126) that made it difficult to assimilate directly with other previously described characters. Many of these characters also were based on previous data sets (e.g., Cracraft, 1974; Bledsoe, 1988; Lee et al., 1997) that were considered at the time of data set construction. Another more recently published character (Worthy and Scofield, 2012, utilized in Mitchell et al., 2014; 179 characters) drew on many of these previous sets, especially revising Bourdon et al. (2009). This data set recovered a supraspecific lithornithid terminal taxon in an unresolved polytomy with the rest of Palaeognathae. Here, we were focused on adding characters relevant to relationships among lithornithids and consider our data set to address distinct aims from the valuable Worthy and Scofield (2012) data set. Future detailed work assimilating all of these character sets is clearly necessary.

Character descriptions are included below with character citations and modifications. The most recent author to use previously modified characters is cited in the character list; citations of the previous usage of the characters is not meant to be exhaustive (appendix 1).

Methods

The character-taxon matrix was assembled in Morphobank (O'Leary and Kaufman, 2012) and is published under project 340. We analyzed our data set using PAUP* 4.0b10 for Macintosh PPC (Swofford, 2002). Trees were searched for using the parsimony criterion implemented under the heuristic search option using tree bisection and reconnection (TBR) with 1000 random-addition sequence replicates. Zero-length branches were collapsed if they lacked support under any of the most parsimonious reconstructions. All characters were equally weighted. Bootstrap support was calculated from 1000 replicates using a heuristic search strategy with random taxon addition sequence and TBR branch swapping (Efron, 1979; Felsenstein, 1985). Bremer support was calculated manually in PAUP*4.0b10. Multistate scorings for terminals indicate polymorphism, which is treated as a “?” in our analysis.

We conducted two phylogenetic analyses without constraint trees that differed only by the inclusion of lithornithid taxa. In the “total” analysis, we included all lithornithids individuals as separate terminal taxa. For example, each Green River Formation lithornithid was its own terminal taxon (e.g., AMNH 30560), and the two specimens of Lithornis promiscuus were scored separately as the “Lithornis promiscuus holotype” (USNM 336535) and a referred skull (USNM 391983) as “Lithornis promiscuus skull.” We utilized this conservative strategy because: (1) the referral of some specimens to Lithornithidae is not always clear because of the lack of clearly defined apomorphies; and (2) the referral of some specimens to species-level lithornithid taxa from the same locality or same geological unit is not clear especially when there are few overlapping skeletal elements.

In the “core lithornithid” analysis, the number of terminal taxa was reduced to 25 taxa because we combined individual lithornithid specimens into species-level terminal taxa (labeled as “combined”). Specimens were combined in species-level terminal taxa after rediagnosing each of the lithornithid taxa utilized in this analysis (see appendix 2). In these cases, the individual specimens of lithornithids were almost always from the same locality and always from the same geological unit (see Houde, 1988).

Following the initial run of data set, we built various constraint trees for palaeognath relationships using the “core lithornithid” taxon sampling to explore the effect of alternative previously hypothesized topologies on the relationships among lithornithids. Further, we used these constraint trees to examine the effects on character support for Aves, Palaeognathae, and Lithornithidae (Houde, 1988; Dyke, 2003; Clarke, 2004; Livezey and Zusi, 2006; 2007; Leonard et al., 2005; Phillips et al., 2010; Johnston, 2011; Mitchell et al., 2014; Baker et al., 2014). Constraint trees were built in McClade (version 4.08; Maddison and Maddison, 2005) and then imported into PAUP. The constraint trees were loaded and then the option “enforce topological constraints” was chosen under the heuristic search settings. The proceeding analysis was then run as described above for the unconstrained analyses.

Results

In both unconstrained analyses, we recovered the Neognathae-Palaeognathae split at the base of Aves (figs. 23, 24) as previously recovered in morphological and molecular studies of avian relationships (e.g., Cracraft, 1988; Mayr and Clarke, 2003; Van Tuinen et al., 2000; Livezey and Zusi, 2006, 2007; Hackett et al., 2008; Jarvis et al., 2014). Ichthyornis dispar was found as the sister taxon of Aves as in all recent phylogenetic analyses of early diverging avian relationships. The relationships of neognaths are similar to those of morphological (Clarke, 2004; Livezey and Zusi, 2006, 2007) and recent molecular analyses (Hackett et al., 2008; Jarvis et al., 2014), whereas the palaeognath relationships are similar to those of the morphology-only phylogenetic hypotheses of Lee et al. (1997) and Bourdon et al. (2009). In contrast to nearly all molecular hypotheses that find Apteryx as the sister taxon to a Casuarius-Dromaeus clade (Lee et al., 1997; Cooper et al., 2001; Hackett et al., 2008; Phillips et al., 2010; Baker et al., 2014), we recover Apteryx as the sister taxon to the moa Emeus crassus at the base of ratites similar to the results in Cracraft (1974). Palaeotis weigelti was found outside a Struthio + Dromaius + Rhea clade. This result is identical to that of recent analysis of Palaeotis weigelti from an entirely different data set with similar taxon sampling (Mayr, 2015).

FIG. 23.

Strict consensus (TL = 470; MPTs = 2724; CI = 0.452; RI = 0.691; RC = 0.312) of the “total” analysis. Bremer supports located above branches.

FIG. 24.

Strict consensus (TL = 470; MPTs = 29; CI = 0.451; RI = 0.676; RC = 0.305) of the “core lithornithid” analysis. Bremer supports located above branches.

In both the “total” (fig. 23) and “core lithornithid” (fig. 24) analyses, we recovered a monophyletic Lithornithidae as the sister taxon of Tinamidae at the base of Palaeognathae. In both analyses, tinamous collapsed into a polytomy in the consensus tree. In the “total” analysis (tree length [TL] = 470; most parsimonious trees [MPTs] = 2724; consistency index [CI] = 0.452; retention index [RI] = 0.691; recalculated consistency index [RC] = 0.312), all of the taxa or specimens previously hypothesized to belong to Lithornithidae were found in a monophyletic group. All the lithornithids collapsed into a polytomy without any ingroup resolution in the strict consensus tree (fig. 23). The monophyly of Lithornithidae is supported by a combination of the following unambiguous character states (some of which represent the same states, presumably plesiomorphic states, present in outgroup taxa (= *)): quadrate with a small fossae on the posterior side of the body (C38:1); orbital process of the quadrate short, hatchet shaped, significantly shorter than the length of the body (C41:1*); centrally located ovoid foramina lateral surfaces of centra of the thoracic vertebrae (C67:2*); only three sacral vertebrae with dorsally directed parapophyses just anterior to the acetabulum (C70:1*); foramina present posteroventral surface to the hooked acrocoracoid process of the coracoid (C95:1); acromion of the scapula laterally hooked tip with small foramina on the posterior side (C104:2); tubercle present where this ramus joins the dorsal margin of the ulnare (C133:1); infratrochlearis fossa deeply excavating the proximal surface of the pisiform process on the ventral surface of the carpometacarpus (C136:1); deep groove at the distal end of the posterior side of the carpal trochlea of the proximal portion of the carpometacarpus (C138:1); and distinct tuber present on the ventral side of the proximal portion of metacarpal III (C143:1).

Our “core lithornithid” analysis (TL = 470; MPTs = 29; CI = 0.451; RI = 0.676; RC = 0.305) recovered a monophyletic Lithornithidae. However, all lithornithid taxa collapsed into a polytomy with no ingroup resolution in the strict consensus. The monophyly of Lithornithidae is supported by a combination of the following unambiguous character states: broad, flat ventral process of the lacrimal in lateral view (C13:1); quadrate with a small fossae on the posterior side of the body (C38:1*); orbital process of the quadrate short, hatchet shaped, significantly shorter than the length of the body (C41:1); short fingerlike retroarticular process just posterior to the articular facets of the mandible (C60:1); in the third cervical vertebra, the presence of an osseous bridge from the transverse process to the postzygapophysis so a foramen is formed in dorsal view (C64:1); centrally located ovoid foramina on the lateral surfaces of the centra of the thoracic vertebrae (C67:2*); number of ankylosed synsacral vertebrae numbering 11–14 (C69:1); three synsacral vertebrae with dorsally directed transverse processes in the area of the acetabulum (C70:1*); coracoidal sulci spacing on anterior edge of the sternum crossed on midline (C79:1); absence of trabeculae on the posterior margin of the sternum (C83:0); foramina present on the posteroventral surface of the hooked acrocoracoid process of the coracoid (C95:1); acromion of the scapula laterally hooked tip with small foramina on the posterior side (C104:2); infratrochlear fossa deeply excavating the proximal surface of the pisiform process on the ventral surface of the carpometacarpus (C136:1); distinct tuber present on the ventral side of the proximal portion of metacarpal III (C143:1).

Within Lithornithidae, the interrelationships were completely unresolved in both analyses. However, there are a number of lithornithids that were closely related in the most of the MPTs, but these relationships are no better supported than others found in various MPTs. Calciavis grandei and Pseudocrypturus cercanaxius were almost always closely related in two positions: (1) as sister taxa as the most nested lithornithids; or (2) in a hard polytomy at the base of with each other and all other Lithornithidae. The sister-taxon relationship well nested in Lithornithidae is supported by two unambiguous character states in both analyses: anteroposterior straight medial margin of the pterygoid anterior to the articulation the basipterygoid process (C22:0) and distal extent of trochlea II and trochlea IV equal (C180:0). In this phylogenetic position, these two character states optimize as independently derived from the plesiomorphic condition found outside Lithornithidae among basal palaeognaths. In contrast, the same two character states represent plesiomorphies at Lithornithidae + Tinamidae and keep Calciavis grandei and Pseudocrypturus cercanaxius at the base of Lithornithidae. The other couplet typically found in most MPTs is a sister-taxon relationship between Paracathartes howardae and Lithornis promiscuus. This clade is found in nearly all positions among other lithornithids and is always supported by the height of shaft at the proximal end of the scapula greater than the height of the humeral glenoid (C105:1).

In the “total” unconstrained analysis, a Lithornithidae-Tinamidae clade is supported by the following unambiguous character states: maxillopalatine antrum absent (C4:0); frontal-parietal suture open (C12:0); extensive contact between the lacrimal and ectethmoid (C16:1); cluster of pneumatic foramina on the posterior surface of the otic head of the quadrate (C43:1); small horizontal shelf at the posterior portion of the mandibular symphysis (C48:1); corpus of the axis with pneumatic foramina on lateral sides (C62:1); acromion of the scapula projected anteriorly to surpass the articular surface for coracoid (tuberculum coracoidei of Baumel and Witmer, 1993) of the scapula (C103:0); pneumatic foramina within of pneumotricipital fossa of the humerus (C111:1); intermetacarpal space (between metacarpals II and III) terminates distal to end of metacarpal I (C141:1); manual digit III, phalanx 1, length less than 50% the length of digit II, phalanx 1 (C146:1); plantar surface of the tarsometatarsus trochlea II and IV expanded into distinct “wings” (C177:1). The character support for the Lithornithidae-Tinamidae clade in the “core lithornithid” analysis was nearly identical to that of the “total” analysis except that character state (C43:1) did not diagnose the clade in the “core lithornithid” analysis.

“Core lithornithid” analysis (fig. 24)

Constraint: None.

Support: Aves – C66:1, C69:2, C70:2, C71:1, C93:1, C102:0, C113:1, C114:0, C115:0, C117:1, C118:2, C129:0, C149:1, C165:0; Palaeognathae – C5:1, C9:1, C31:1, C36:0, C42:0, C51:1, C72:1, C137:1, C167:1, C180:0; Lithornithidae – C13:1, C38:1, C41:1, C60:1, C67:2, C69:1, C70:1, C79:1, C83:0, C95:1, C104:2, C136:1, C143:1.

Constraint Trees and Alternate Hypotheses of the Relationships of Lithornithidae

The impact on the inferred relationships among Lithornithidae species of imposing constraint trees to reflect: (1) previously proposed hypotheses of their relationships among basal avians (table 3); and (2) phylogenetic results from large-scale molecular sequence based analyses of Aves were explored. Constraint trees enforced placement of Lithornithidae as the sister taxon of Aves following Livezey and Zusi (2006; 2007), as a non-monophyletic group where Paracathartes was closer to ratites than other lithornithids following Houde (1986; 1988), and as the sister taxon to all other Palaeognathae or of ratites (Clarke, 2004).

TABLE 3

Constraint tree statistics compared

We also explored the affect on the inferred relationships of Lithornithidae of constraining the relationships of extant Palaeognathae to match the phylogenetic results obtained from diverse large molecular data sets by Phillips et al. (2010), Baker et al. (2014), and Mitchell et al. (2014). The relationships recovered in the “total” analysis described above are at odds with all recent molecular analyses of avian relationships (Harshman et al., 2008; Phillips et al., 2010; Smith et al., 2013a; Haddrath and Baker, 2012; Baker et al., 2014; Mitchell et al., 2014; Jarvis et al., 2014).

In the sections below, we present how the taxa were constrained, the resultant tree (figs. 25–31), tree statistics (table 3), and character support for recovered clades. The Tinamidae + Lithornithidae clade (fig. 23) is not recovered in these analyses with the exception of a constraint placing Paracathartes as the sister taxon of a monophyletic Ratitae (Houde, 1988).

FIG. 25.

Strict consensus when Lithornithidae is constrained as the sister taxon of Aves.

FIG. 26.

Strict consensus when Lithornithidae is constrained as the sister taxon of all other Palaeognathae.

FIG. 27.

Strict consensus when Lithornithidae is constrained as the sister taxon of ratites.

FIG. 28.

Strict consensus when Lithornithidae is constrained as not monophyletic and Paracathartes howardae is the sister taxon to all extant ratites.

FIG. 29.

Strict consensus when Palaeognathae relationships are constrained to the relationships found by Phillips et al. (2010).

FIG. 30.

Strict consensus when Palaeognathae relationships are constrained to the relationships found by Baker et al. (2014).

FIG. 31.

Strict consensus when Palaeognathae relationships are constrained to the relationships found by Mitchell et al. (2014).

Constraints based on previously proposed placements of the Lithornithidae

Lithornithidae as the sister taxon of Aves (fig. 25)

Constraint: Lithornithidae was constrained as the monophyletic sister taxon of all members of Aves found in the “core lithornithid” tree (fig. 24). All other taxa were unconstrained, and lithornithid interrelationships were unconstrained.

Support: Lithornithidae + Aves – C51:1, C66:1, C71:1, C93:1, C110:1, C113:1, C114:0, C115:0, C118:2, C129:0, C137:1, C140:2, C149:1, C165:0, C167:1, C181:1; Aves – C69:2, C70:2, C99:1, C117:1, C129:0, C136:0, C166:2; Lithornithidae – C13:1, C60:1, C64:1, C67:2, C83:0, C95:1, C104:2, C112:0, C143:1.

Results: With Lithornithidae outside Aves, Tinamidae was found as the sister taxon of Palaeognathae + Neognathae. Therefore, Palaeognathae did not include Tinamidae, and Emeus crassus + Apteryx owenii are the earliest diverging members of the clade. The relationships of Lithornithidae are fully resolved, but anseriform taxa form a polytomy. Within Lithornithidae, Lithornis is paraphyletic and the two Green River Formation taxa are sister taxa and well nested within the clade. At the base of Lithornithidae, Lithornis celetius shares a number of character states (e.g., elongated pectineal process of the pubis) with the tinamou clade, and thus these character states optimize as plesiomorphic for Aves. All other relationships were the same as the “core lithornithid” analysis.

Lithornithidae as the sister taxon of all other Palaeognathae (fig. 26)

Constraint: Lithornithidae was constrained as a monophyletic group (but interrelationships were unconstrained) as the sister taxon of all members of Palaeognathae found in the “core lithornithid” tree. All other taxa were unconstrained.

Support: Aves – C66:1, C71:1, C93:1, C102:0, C110:1, C113:1, C114:0, C115:0, C117:1, C118:2, C129:0, C149:1, C165:0, and C181:1; Palaeognathae – C5:1, C9:1, C31:1, C36:0, C42:0, C51:1, C135:1, C137:1, C139:1, and C167:1; Lithornithidae – C13:1, C38:1, C41:1, C60:1, C64:1, C67:2, C79:1, C83:0, C95:1, C104:2, C112:0, C136:1, and C143:1.

Results: With Lithornithidae constrained as the sister taxon of all other Palaeognathae, the Lithornithidae + Tinamou clade is no longer recovered; tinamous are the sister taxon of all ratites. Within a nearly unresolved Lithornithidae, Lithornis promiscuus and Paracathartes howardae were found as sister taxa. The relationships of the three anseriforms dissolved into a polytomy. All other relationships were the same as the “core lithornithid” analysis.

Lithornithidae as the sister taxon of ratites (fig. 27)

Constraint: Lithornithidae was constrained as the monophyletic sister taxon of all flightless ratites found in the “core lithornithid” tree. All other taxa unconstrained.

Support: Aves – C66:1, C69:2, C70:2, C71:1, C93:1, C110:1, C113:1, C114:0, C115:0, C117:1, C118:2, C129:0, C149:1, C165:0, and C181:1; Palaeognathae – C5:1, C9:1, C31:1, C36:0, C42:0, C51:1, C72:1, C135:1, C137:1, C139:1, C154:2, and C167:1; Lithornithidae + Ratites - None

Lithornithidae – C60:1, C67:2, C69:1, C70:1, C79:1, C83:0, C104:2, C112:0, and C143:1.

Results: With Lithornithidae constrained as the sister taxon of all ratites, there is very little effect on recovered relationships. However, a Lithornithidae + tinamou clade is no longer recovered. Species relationships in Lithornithidae are nearly completely resolved, and those in Tinamidae are completely unresolved. Lithornis is again paraphyletic and the Green River Formation lithornithids are again recovered as sister taxa.

Lithornithid Paracathartes howardae as the sister taxon to all crown ratites (fig. 28)

Constraint: Paracathartes howardae was constrained as the sister taxon of all crown ratites (one of the hypotheses of Houde, 1988). All other taxa were unconstrained.

Support: Aves – C66:1, C69:2, C70:2, C71:1, C93:1, C102:0, C110:1, C113:1, C114:0, C115:0, C117:1, C118:2, C129:1, C149:1, C165:0, and C181:1; Palaeognathae – C5:1, C9:1, C31:1, C36:0, C42:0, C51:1, C72:1, C135:1, and C167:1; Lithornithidae [no Paracathartes howardae] –- C4:0, C16:1, C111:1, C146:1, and C150:1. Paracathartes howardae + Ratites – C17:1.

Results: The relationships of the included Aves are nearly identical to that of the unconstrained analysis. All lithornithids except Paracathartes howardae were found in a clade as the sister taxon of Tinamidae. The interrelationships of Lithornithidae are completely resolved. Lithornis is recovered as monophyletic in this analysis, and the Green River Formation species are again sister taxa.

Constraints based on large-scale molecular phylogenies

Palaeognath relationships from Phillips et al. (2010) (fig. 29)

Constraint: Palaeognath relationships were constrained (as a backbone constraint in PAUP) to match the taxon sampling and results of Phillips et al. (2010). All other taxa were unconstrained.

Support: Aves – C66:1, C71:1, C93:1, C102:0, C113:1, C114:0, C115:0, C117:1, C118:2, C129:0, and C149:1; Palaeognathae – C5:1, C9:1, C31:1, C36:0, C42:0, C51:1, C72:1, C137:1, C139:1, C167:1, and C180:0; Lithornithidae – C4:0, C12:0, C16:1, C20:2, C60:1, C64:1, C67:2, C83:0, C103:0, C104:2, C112:0, C133:1, C141:1, C143:1, C150:1, and C177:1.

Results: Constraining the relationships among Palaeognathae to those recovered by Phillips et al. (2010) results in Lithornithidae being recovered as the basalmost clade of Palaeognathae and not as the sister taxon of Tinamidae. Tinamidae are highly nested within the clade (Phillips et al., 2010). The extinct Palaeotis weigelti was found as the sister taxon of extant palaeognaths. The interrelationships within Tinamidae are fully resolved, some of the interrelationships of lithornithids were resolved, and the interrelationships of the included anseriforms were unresolved. A monophyletic Lithornis + Paracathartes clade is recovered but the Green River Formation lithornithid species are not recovered as sister taxa.

Palaeognath relationships of from Baker et al. (2014: figure 3) (fig. 30)

Constraint: Palaeognath relationships were constrained (as a backbone constraint in PAUP) to match figure 3 of Baker et al. (2014). This is the tree of Haddrath and Baker (2012; based on 27 nuclear genes and 21 retroelements) with eight novel retroelement insertions mapped; Baker et al., 2014). All other taxa were unconstrained.

Support: Aves – C66:1, C71:1, C93:1, C102:0, C113:1, C114:0, C115:0, C117:1, C118:2, C129:0, and C149:1; Palaeognathae – C5:1, C9:1, C31:1, C36:0, C42:0, C51:1, C62:1, C72:1, C137:1, C139:1, C167:1, and C180:0; Lithornithidae – C4:0, C12:0, C13:1, C16:1, C20:2, C60:1, C64:1, C67:2, C83:0, C103:0, C104:2, C112:0, C133:1, C141:1, C143:1, C150:1, and C177:1.

Results: Lithornithidae again was found as the earliest diverging clade of the Palaeognathae, whereas Tinamidae is well nested within a paraphyletic Ratitae. The extinct Palaeotis weigelti was found as the sister taxon of all other Palaeognathae (exclusive of Lithornithidae). The resolution within Tinamidae and Anseriformes is lost, and taxa within Lithornithidae are only partially resolved with Calciavis and Pseudocrypturus placed as basal diverging members. The optimizations of the character states are nearly identical to that found in the Phillips et al. (2010) constraint.

Palaeognath relationships of Mitchell et al. (2014) (fig. 31)

Constraint: The palaeognath relationships were constrained (as a backbone constraint in PAUP) to match the results of Mitchell et al. (2014). Note, the sampling of Mitchell et al. (2014) is greater than ours in this analysis; we do not include the elephant bird or cassowary. All other taxa were unconstrained.

Support: Aves – C66:1, C69:1, C71:1, C93:1, C102:0, C113:1, C114:0, C115:0, C117:1, C129:0, and C149:1; Palaeognathae – C5:1, C9:1, C31:1, C36:0, C42:0, C51:1, C72:1, C137:1, C139:1, C167:1, and C180:0; Lithornithidae – C4:0, C12:0, C13:1, C16:1, C60:1, C64:1, C67:2, C69:1, C83:0, C103:0, C104:2, C112:0, C133:1, C141:1, C143:1, C150:1, and C177:1.

Results: As seen in the results from the analyses utilizing the Phillips et al. (2010) constraint and the Baker et al. (2014) constraint, Lithornithidae is not recovered with Tinamidae, but as basal within Palaeognathae as the sister taxon to all other included species. With this placement of the clade, the relationships among lithornithid species also are identical to those recovered the analyses using constraint trees based on the results of Phillips et al. (2010) and Baker et al. (2014).

Alternative hypotheses discussion: The relationships of nonpalaeognath Aves were generally similar to those recovered in the “core lithornithid” analysis. For example, neognath interrelationships remained nearly the same in each constraint tree and ratite interrelationships also remain identical to those of the “core lithornithid” analysis unless ratites were specifically constrained (e.g., Phillips et al., 2010; Mitchell et al., 2014; Baker et al., 2014, constraints). However, the recovered relationships among lithornithids, and placement of Lithornithidae within Palaeognathae, varied significantly in each of the constraint runs depending on their closest outgroups. In all analyses, enforcing previous hypothesis of lithornithid relationships, Lithornis is paraphyletic with respect to a monophyletic Green River Formation lithornithid clade. In all analyses, enforcing constraints based on molecular sequence based phylogenetic hypotheses, Lithornithidae is not recovered with Tinamidae. Furthermore, a monophyletic Lithornis + Paracathartes clade is recovered, whereas the Green River Formation lithornithid relationships are unresolved in these analyses.

In all cases, the constraint trees had a higher number of steps and lower CI, RI, and RC values compared to the unconstrained analysis. In the analyses where Lithornithidae was found as monophyletic, the number of added steps was comparable. By contrast, the nonmonophyly of Lithornithidae constraint added a significant number of steps (table 3). Constraining relationships to those obtained by Phillips et al., (2010), Baker et al. (2014), and Mitchell et al. (2014) led to an increase in the number of steps (41, 9% increase; 38, 8% increase; and 30, 6% increase, respectively) relative to the unconstrained tree. The number of MPTs was highest in the unconstrained “total” analysis.

The support (i.e., apomorphies) for a monophyletic Lithornithidae in the constrained analyses was rather similar to the support found in the unconstrained trees. A few apomorphies were gained or lost relative to the unconstrained tree, but there was a core character set (C13:1, C38:1, C60:1, C67:1, C95:1, C143:1) that supported the clade. In the iteration where nonmonophyly of Lithornithidae was forced (i.e., Paracathartes howardae closer to ratites), none of the core lithornithid characters supported the monophyly without Paracathartes howardae. Instead, the core lithornithid apomorphies were spread across the branches of the Palaeognathae tree. Moreover, the only character state that supported Paracathartes howardae closer to ratites than to other lithornithids was the reduction of the postorbital process of the frontal (C17:1), a character state suggest by Houde (1988) in support of nonmonophyly of Lithornithidae.

In the constraint analysis enforcing the relationships of Phillips et al. (2010), Lithornithidae was recovered as monophyletic and well supported (i.e., 18 unambiguous apomorphies). The support of this clade contained most, but not all of the core lithornithid apomorphies (4 total) of the unconstrained analysis. The additional apomorphies supporting the clade previously supported a Lithornithidae + Tinamidae clade in the unconstrained tree. Thus, part of the addition of the number of steps in the Phillips et al. (2010) constraint tree was the result of the independent origin of apomorphies of Lithornithidae and Tinamidae in the unconstrained analyses.

The character support for Aves and Palaeognathae varies throughout the alternative hypotheses iterations, but character support was rather constant when the relationships of the earliest member of the clade were not constrained. The biggest differences in character support for Aves and Palaeognathae relative to the unconstrained analysis occurred when Lithornithidae was constrained to a position outside Aves. Many of the character states that supported Aves in the unconstrained analysis supported a Lithornithidae + Aves clade in the constraint tree, and apomorphies that once supported Lithornithidae + Tinamidae, Lithornithidae + Tinamidae + ratites, and Palaeognathae in the unconstrained analysis were spread across the tree. Interestingly, two apomorphies that support Aves in both the unconstrained and “Lithornithidae outside Aves” are related to the increase of number of vertebrae in the synsacral series (C69:2 and C70:2) relative to the plesiomorphic condition in Avialae. Lithornithids, tinamous, and outgroups of Aves (e.g., Ichthyornis; Clarke, 2004) have 12 synsacral vertebrae whereas galloanserines and other palaeognaths generally have 15 or more.

Species Diversity of Green River Formation Lithornithids

The Fossil Butte Member of the Green River Formations has produced the largest sample of lithornithids skeletons to date. Here, we add five new specimens for a total of seven partial skeletons of lithornithids from near contemporaneous localities spanning both F1 and F2 deposits of Fossil Lake (of Grande and Buchheim, 1994). The morphological differences among the Green River Formation specimens are few, but this is not surprising given that Houde (1988) found little morphological variation across the skeleton of North American lithornithids spanning 10–15 million years. Here, we argue that there is a minimum of two species-level lithornithid taxa from the Green River Formation. We discuss the possible ontogenetic status of the specimens, as well as morphological and proportional differences to consider the possibility that one species-level taxon can explain all these data.

Ontogenetic Age: The ontogenetic stages represented by the Green River Formation lithornithids is important for considering their taxonomy. Histological analysis of long bones of avians can directly determine what part of the growth cycle an avian had reached at death (Chinsamy et al., 1995; Chinsamy and Elzanowski, 2001). Because, unfortunately, this method is destructive, it was not used here. Instead, we turned to a nondestructive method that correlates the surface texture of long bones to skeletal maturity in avians (e.g., “textural aging” of Tumarkin-Deratzian et al., 2006; 2007; Watanabe and Matsuoka, 2013). This method of evaluating skeletal maturity is determined by examining the surface texture of a long bone and its correlation with the histology of the same bone. For example, a fibrous external surface of a long bone correlates with the early faster growth (woven bone) of a skeletally immature avian whereas a smooth external surface of a long bone correlates with the later, zonal lamellar growth strategy of skeletally mature avians (Tumarkin-Deratzian et al., 2007). Furthermore, this method of accessing skeletal maturity appears to be valid in fast growing archosaurs (e.g., pterodactyloid pterosaurs, Bennett, 1993; possibly large theropods, Carr, 1999; early avialans, Erickson et al., 2009), but not in slower growing archosaurs (e.g., Alligator mississippiensis, Tumarkin-Deratzian et al., 2007).

Using this nondestructive method, we evaluated the skeletal maturity of Green River Formation lithornithids. The surface textures of long bones of both the holotype (AMNH 30578) and referred specimen (AMNH 30560) are nearly smooth. Given that these specimens were buried relatively quickly after death with finely preserved features and were prepared extremely carefully to preserve these features, it is clear that the smooth external surfaces were not taphonomic artifacts nor mechanically produced. Thus, it is clear that the Calciavis grandei specimens were skeletally mature at death. The ontogenetic age of the holotype of Pseudocrypturus cercanaxius (USNM 336102) cannot be determined using these methods confidently because no long bone elements are preserved. However, the same-sized referred specimen (Siber and Siber specimen, USNM 424078 [cast]) does preserve long bone elements. We assessed the skeletal maturity of Pseudocrypturus cercanaxius from a highly detailed plaster cast. The external textures of the long bones are as smooth as those of Calciavis grandei and thus, the referred specimen and likely the holotype were also skeletally mature. Note, however, it is possible that the mold was not detailed enough to pick up fine striations on the outside of long bones. Nevertheless, both the holotype and referred specimens of Pseudocrypturus cercanaxius have larger skulls than that of the skeletally mature Calciavis grandei and, therefore, Pseudocrypturus cercanaxius must have reached an absolute larger head size. The long bones of the other Green River Formation lithornithids (FMNH PA 739, FMNH PA 729, and WGS 41-2001) all have smooth external surfaces, and thus also appear to be skeletally mature.

As a part of this skeletally maturity evaluation, we also examined the maturity of Lithornis celetius (USNM 290554) and Paracathartes howardae (USNM 361407) because these specimens have both histological sections and long bones that can be evaluated for external texture. The carpometacarpus of Lithornis celetius (USNM 290554) has a smooth texture across the element. Similarly, the long bones of Paracathartes howardae (e.g., femur, USNM 361412) found with USNM 361407 also has a smooth texture across the element. Therefore, the external features of these specimens indicate they are from a skeletally mature individual. Furthermore, this age assessment is supported by histological sections of the tibiotarsi of both taxa (Houde, 1988). The histological sections show a distinct transition from woven bone to more parallel-fibered bone in the outer cortex of the section. Both individuals had nearly stopped growing at prior to death. The similar texture of the long bones of individuals of Lithornis celetius, Paracathartes howardae, Calciavis grandei, and Pseudocrypturus cercanaxius suggests that all these individuals were skeletally mature at death. Thus, Houde's (1988) original differentiation of species-level taxa of lithornithids based on size appears valid, and these species-level boundaries are backed up by morphology (see below).

Morphological differences: The first taxon recognized from the Green River Formation was Pseudocrypturus cercanaxius (Houde, 1988). Unfortunately, the well-preserved and prepared holotype (USNM 336102) consists of only a laterally compressed skull and partially preserved cervical series, thus making comparison to other lithornithids and referring incomplete lithornithids specimens from the Green River Formation difficult (fig. 32). Of the original diagnosis provided by Houde (1988), only the listed cranial characters (e.g., robust lateral nasal bar, small pterygoid fossa, palatine with posteriorly (= caudally) directed process, vomer that articulates with the premaxilla) can be used for comparison and for referring specimens to Pseudocrypturus cercanaxius. This combination of character states of the skull is unique (see Houde, 1988, and diagnoses below). The noncranial character states in the diagnosis of Pseudocrypturus cercanaxius comes from the referred specimen (Siber and Siber specimen, USNM 424078 [cast]) including: sternum with shallow notches in the posterior margin; coracoid with long slender shaft; scapula with straighter shaft and more pronounced acromion; tarsometatarsus with symmetrical trochlea III and winglike processes of trochleae II and IV; and short limbs. Most of the character states of the holotype skull of Pseudocrypturus cercanaxius cannot be evaluated in the referred specimen. The absolute length and proportions of the skull (table 4) and its individual elements are identical in the holotype and the referred specimen; therefore, we concur with Houde (1988) and refer the Siber and Siber specimen to Pseudocrypturus cercanaxius. Consequently, nearly the entire anatomy of Pseudocrypturus cercanaxius is visible and serves as a comparison to the other Green River Formation lithornithids from nearby localities.

TABLE 4

Ratios between different elements of the Green River Formation lithornithids

X = measurements that could not be made.

The holotype of Calciavis grandei has elements directly comparable to both the holotype and referred specimen of Pseudocrypturus cercanaxius. The morphological features of Calciavis grandei and Pseudocrypturus cercanaxius are nearly identical, and both taxa are about the same in body size. In our phylogeny, the character scores are identical (when scorable) between the holotype of Calciavis grandei and the holotype of Pseudocrypturus cercanaxius, and the holotype of Calciavis grandei and the referred specimen of Pseudocrypturus cercanaxius are scored for the same characters states except for character 18. In Pseudocrypturus cercanaxius, the skull is longer than the humerus, whereas in Calciavis grandei the humerus is clearly longer than the cranium. Furthermore, the skull is longer than the tibiotarsus in Pseudocrypturus cercanaxius, whereas the tibiotarsus is distinctly longer than the skull in Calciavis grandei.

The skeletal proportions of Pseudocrypturus cercanaxius and Calciavis grandei are different further supporting their identity as distinct species (table 1). For example, the length of the femur of the holotype of Calciavis grandei is longer than that of the referred specimen of Pseudocrypturus cercanaxius, but the length of the sternum of Pseudocrypturus cercanaxius is absolutely longer than that of the holotype of Calciavis grandei. By contrast, similar proportions (i.e., humerus longer than the cranium; in limb elements), as well as morphological features, led to our referral of AMNH 30560 to Calciavis grandei. However, AMNH 30560 is not exactly the same size as the holotype (~90% the size of the holotype), and the precise scaling proportions of some limb elements are not the same (e.g., the skull of the referred specimen is ~93% that of the holotype whereas the tibiotarsus length is 87% that of the holotype). It is not clear whether these minor proportional differences represent actual differences (e.g., individual variation of sexual dimorphism) or could be explained by taphonomic processes during fossilization such as compaction.

In sum, the differences in skull size relative to long bones in the skeletons easily differentiates Pseudocrypturus cercanaxius from Calciavis grandei given that the holotype and referred specimens represent skeletally mature adults. Differentiation of contemporaneous species-level taxa from proportional differences in the skeleton is practiced in avian studies of both extant and extinct (e.g., Olson and Matsuoka, 2005) taxa. As support, we examined 19 specimens from one population of the tinamou Eudromia elegans (tables 5, 6), a close relative of lithornithids. We measured individual elements and skull length to compare to Pseudocrypturus cercanaxius and Calciavis grandei. In this small sample Eudromia elegans, it was clear there was overall variation in size of each element. However, the range of lengths and the standard deviation of each element was within 5% the length of each element. Ergo, Eudromia elegans proportions differ little and show a much smaller range of variation of that of the Pseudocrypturus cercanaxius and Calciavis grandei specimens taken together.

TABLE 5

Selected measurements of an adult sample (n = 19) of the tinamous Eudromia elegans

Measurements rounded to nearest 1/10 mm. X = measurements that could not be made. All specimens are curated at USNM

TABLE 6

Ratios between different elements of Eudromia elegans

X = measurements that could not be made.

There appear to be only two presently diagnosable species from the Fossil Butte Member of the Green River Formation, Pseudocrypturus cercanaxius and Calciavis grandei. Yet there are a number of other specimens from the same unit that are far from complete and lack preservation of differentiating morphologies (see above) between the two recognized species-level taxa. FMNH PA 739 and WGS 41-2001 appear to share a more similar absolute size and similar proportions of elements with Pseudocrypturus cercanaxius than with Calciavis grandei. Likewise, FMNH PA 729 appears to share a similar absolute size and similar proportions of elements with Calciavis grandei than Pseudocrypturus cercanaxius. Nevertheless, all three specimens (FMNH PA 739, FMNH PA 729, WGS 41-2001) lack both a humerus and a skull and, as a result, they are not obviously referable to Calciavis grandei or Pseudocrypturus cercanaxius based on presently available morphological evidence. Furthermore, these three specimens are incomplete, and we cannot reject the notion that they represent different species-level taxa. Therefore, we assign FMNH PA 739, FMNH PA 729, and WGS 41-2001 only to Lithornithidae spp., following the results of our phylogenetic analyses.

Even though the Green River Formation lithornithids are scored nearly identically in our phylogenetic analysis, they do appear to share a few character states that differ from those of other lithornithids. For example, all the Green River Formation lithornithids have a scapular blade that terminates in a relatively broad tip (C106:0) rather than a finely tapered point. Additionally, tarsometatarsal trochlea II and IV are approximately equal in distal extent (C180:0), but other lithornithids have a metatarsal II that is shorter than trochlea IV (C180:1). Furthermore, the medial portion of the pterygoid anterior to the facet for the articulation of the basipterygoid process is straight (C22:0) in the Green River Formation lithornithids, but medially expanded in all other lithornithids.

Sexual dimorphism: Extant birds can show differences in plumage patterning, behavior, and size between the sexes (e.g., Owens and Hartley, 1998; Dunning, 2008). In the fossil record, detection of sexual dimorphism is almost entirely restricted to differences in absolute size at skeletal maturity because of the rarity of preservation of a large sample of plumage or evidence of behavior in fossils. The differences between the two Green River Formation species, Calciavis grandei and Pseudocrypturus cercanaxius, are only in the length of the skull relative to the lengths of long bones (e.g., humerus and tibiotarsus), and in proportional differences in the scaling of long bones (e.g., tibiotarsus versus tarsometatarsus) relative to size. We were not able to find any extant cases among avian of such a large difference of relative skull size between the sexes. For example, in the skeletally mature population of the tinamou Eudromia elegans we considered, the proportions of the skull versus long bones (e.g., tibiotarsus) did not differ more than a few percent compared to the total length (table 6). Unfortunately, the sex and weight of each specimen of Eudromia elegans was not recorded; thus, it is possible that sample contained maleonly or female-only members.

A close relative of lithornithids, the recently extinct moas from New Zealand, had a unique system of sexual dimorphism only recently recognized (Bunce et al., 2003). Previously, three size classes of Dinornis were divided into three species (D. giganteus, D. novaezealandiae, and D. struthoides) with differences in cranial morphology as supporting evidence of species-level differences (Worthy, 1994). The recovery of mitochondrial data from each of the three size classes showed that: (1) all size classes belonged to the same species-level taxon; and (2) through sex determination using the same molecular data, the larger-size classes (D. giganteus and D. novaezealandiae) belonged to females and the smallest size class (D. struthoides) represented males (Bunce et al., 2003). Thus, Dinornis demonstrated sexual dimorphism in absolute body size. Yet, we were interested in the proportional differences between the skull and long bones (e.g., tibiotarsus and femur) among the size classes in Dinornis. Using data from skulls, tibiotarsi, and femora from all three size classes presented by Worthy (1994), we show that the differences in proportions across the three size classes were relatively similar (table 7), and far from the differences between the Green River Formation lithornithids Calciavis grandei and Pseudocrypturus cercanaxius (table 4). Therefore, the proportional differences among proposed lithornithids species-level taxa are different from the extreme sexual dimorphism proposed for the moa, Dinornis.

TABLE 7

Dinornis comparison

Data from Worthy (1994).

Both Calciavis grandei and Pseudocrypturus cercanaxius are from contemporaneous deposits (Grande and Buchheim, 1994), although both species have not been found at the exact same locality within the F-2 deposits of Fossil Lake. Houde (1988) noted that Lithornis promiscuus and Lithornis plebius are always found together at the same localities in the Willwood Formation (Clarkforkian) of Wyoming. Houde (1988) further entertained the hypothesis that these two taxa represent sexual dimorphs of one species-level taxon, but provided a number of morphological and size differences and hypothesized that the differences were diagnostic to specific taxa. If the differences in size (Lithornis plebius is about 75% the size of Lithornis promiscuus) represent dimorphic differences in a single species-level taxon, this dimorphism would not accord with that of Calciavis grandei and Pseudocrypturus cercanaxius, given that both proposed species-level taxa are about the same size but have slight proportional differences.

The lack of major morphological differences between skeletal elements across the body among males and females of a single population in extant and extinct avians provides support that the major differences in the skull length versus long bone length in Calciavis grandei and Pseudocrypturus cercanaxius represents taxonomic variation and not sexual dimorphism. Nevertheless, it remains possible that the proportional variation in the length of the skull versus other long bone elements could be a property of single species-level taxon of lithornithid or lithornithids as a whole. At this time, however, we hypothesize that the differences Calciavis grandei and Pseudocrypturus cercanaxius represent taxonomic differences.

Monophyly of Lithornithidae and Diagnostic Features of the Clade

Lithornithidae was previously considered a monophyletic assemblage or a basal stock of palaeognaths that gave rise to other palaeognath lineages (Houde, 1988). For the first time, we tested the monophyly of Lithornithidae based on scoring specimens and species-level taxa individually in a phylogeny, sampling the avian stem and extinct and extant members of Palaeognathae and basal Neognathae. We find support for this clade within Palaeognathae. Unfortunately, we failed to find any resolution within Lithornithidae with our data set. Thus, we were unable to test the monophyly of Lithornis in this analysis. Nevertheless, Lithornithidae is a diagnosable clade that is restricted to the Paleogene (late Paleocene to the late Eocene) of North America and Europe (see below).

Character support for Lithornithidae as a monophyletic clade is relatively high in both the “total” and “core lithornithid” analyses. However, some of the apomorphies supporting Lithornithidae in this analysis represent plesiomorphies in other phylogenetic analyses sampling other avian taxa (e.g., Mayr and Clarke, 2003). For example, in this analysis, the presence of an osseous bridge from transverse process to the postzygapophysis of the third cervical vertebra (C64:1) optimized as an apomorphy of Lithornithidae. This character state, however, occurs in a number of avians (e.g., basal neognath clades) at the base of the tree that were not sampled in this analysis (see Mayr and Clarke, 2003). Similarly, the presence of 12 synsacral vertebrae (C69:0) in lithornithids optimizes as an apomorphy of the clade. Among the palaeognaths sampled here and the structure of the interrelationships of those taxa, the presence of 12 synsacral vertebrae is apomorphic. However, the presence of 12 synsacral vertebrae is plesiomorphic for Aves (see Clarke, 2004), and thus, this character state in lithornithids likely represents a plesiomorphic character state in Aves.

Although, there are a number of character states supporting Lithornithidae that previously were found to represent plesiomorphies of either Palaeognathae or Aves in other analyses. Lithornithidae is supported as a monophyletic clade by a combination of apomorphies with high CIs in our analyses, and other character states that occur in other avian clades, but appear either to be independently evolved in lithornithids or to represent a unique combination of character states not present in any other avian group. In the following character list of lithornithid apomorphies, we focused on character states that: (1) were apomorphic in both the “total” and “core lithornithid” analyses; (2) were frequently apomorphic in the constraint trees (see Constraint Trees and Alternate Hypotheses of the Relationships of Lithornithidae)and (3) are not obviously plesiomorphic for Aves. The following list is also similar to the apomorphy list provided by Houde (1988) in his comprehensive study of lithornithids.

Lithornithid apomorphies:

FIG. 32.

The holotype skull and cervical vertebrae of Pseudocrypturus cercanaxius (USNM 336103) in A, left lateral and B, right lateral views. Scale = 1 cm. See anatomical abbreviations.

FIG. 33.

Nearly complete skull (USNM 391983) referred to Lithornis promiscuus in A, left lateral, B, right lateral, C, dorsal, and D, ventral views. Scale = 1 cm. See anatomical abbreviations.

FIG. 34.

A, Photograph and interpretative line drawing of the holotype skull of Lithornis celetius (USNM 290601) in right lateral view. B, Right quadrate of Lithornis promiscuus (USNM 336535) in posterior view. C, Left carpometacarpus of Lithornis promiscuus (USNM 336535) in ventral view. D, Isolated thoracic vertebra of Lithornis (NHMUK A5425) in left lateral view. E, Line drawing of the right half of the mandible of the holotype of Lithornis promiscuus (USNM 336535) in lateral view. F, Line drawing of the left coracoid of Paracathartes howardae (USNM 361417) in dorsal view. G, Left scapula of Paracathartes howardae (USNM 361419) in dorsal view.

FIG. 35.

Posterior portion of the left half of the mandible of the holotype of Lithornis promiscuus (USNM 336535) in A, lateral, B, medial, C, dorsal, and D, posterior views. Scale = 1 cm. See anatomical abbreviations.

Lithornithidae is diagnosed by the combination of apomorphies presented above in which few of the apomorphies represent nonlocal autapomorphies of the clade. With that said, we urge future workers to assign taxa to Lithornithidae only if they have a majority of all of the apomorphies in a partial or complete skeleton. Assignment of an isolated avian element to Lithornithidae based on a single apomorphy on the presented list will be poorly supported given the high rates of homoplasy present in avians (see Mayr and Clarke, 2003; Livezey and Zusi, 2007). Consequently, the assignment of isolated bones to Lithornithidae is not recommended. For example, the oldest possible members of Lithornithidae are represented by an isolated scapula from the Late Cretaceous/early Paleocene of New Jersey (Parris and Hope, 2002) and a partial coracoid named Fissuravis weigelti Mayr, 2007, from the Paleocene of Walbeck in Germany (Mayr, 2007). Both isolated bones may be the remains of Lithornithidae, but morphologies approaching the hooked acromion of the Lithornithidae are approached in several stem ornithurine birds in the Mesozoic Era (Clarke, 2004) and the isolated coracoid fragment does not appear to bear any lithornithid character state other than a similar size. In fact, the coracoid appears to be lacking small foramina present posteroventral surface to the hooked acrocoracoid process of the coracoid (C95:1), one of the character states listed above in the diagnosis. As such, isolated bones are nearly impossible to assign to Lithornithidae based on our phylogenetic hypothesis and further analysis.

Is Paracathartes howardae Part of Lithornithidae?

Houde (1988) hypothesized that Paracathartes howardae was not a lithornithid, but may be closer to ratites than to other lithornithids. He cited two pieces of evidence: (1) a reduced and rounded postorbital process of the frontal; and (2) a similar histologic growth pattern of Paracathartes howardae and ratites. In our analysis, we find Paracathartes howardae usually well nested (in the MPTs) within Lithornithidae, and therefore, the reduced and broad postorbital process shared by Paracathartes howardae and ratites (char. 17 here) likely is convergent. The growth pattern shared between Paracathartes howardae and ratites appears to be more similar to each other than to Lithornis (Houde, 1988; fig. 28) because Paracathartes howardae and ratites share a general growth pattern of concentric rings of vascularization throughout the cortex. However, the vascular canal pattern present in Paracathartes howardae is intermediate in morphology between ratites and Lithornis (Houde, 1988). The dominate orientation of the vascular canals in histological section of Paracathartes howardae is longitudinal like that of Lithornis, but the longitudinal vascular canals are arranged circumferentially (or plexiform; see Erickson et al., 2009) throughout the cortex like that of ratites. Yet, ratites appear to nearly completely lack longitudinally oriented canals. Additionally, recent work on body size and growth strategy in stem birds (Padian et al., 2004; Erickson et al., 2009; Werning et al., 2011), particularly vascularization patterns, has shown that body size and vascularization are correlated; smaller animals (e.g., the early avialan Archaeopteryx lithographica) have dominantly longitudinally oriented vascularization whereas larger animals (e.g., the oviraptorid theropod Citipati osmolskae) are dominated by circumferentially oriented (or plexiform) vascularization. Paracathartes howardae was estimated to be nearly twice as large as other lithornithids (Houde, 1988) and, therefore, body size may explain the similarity in circumferentially oriented vasculature in Paracathartes howardae and ratites, not homology or phylogeny.

We further tested Houde's (1988) hypothesis of the phylogenetic position of Paracathartes howardae in a constraint tree and found that if Paracathartes howardae is more closely related to ratites than to other lithornithids, one would need to add 17 more steps to our original unconstrained “core lithornithid” analysis. Moreover, there was almost no character support for this hypothesized relationship. Therefore, our data support Paracathartes howardae as a lithornithid.

What is the Basalmost Lithornithid Lineage?

We were unable to obtain any resolution of relationships within Lithornithidae in the most inclusive and unconstrained analysis (fig. 24). Calciavis grandei and Pseudocrypturus cercanaxius are found as sister taxa of all other lithornithids in a majority of the MPTs in the “core lithornithid” analysis. In the constrained analyses, when account relationships supported by molecular data (figs. 29–31) are taken into account, the position of the Green River Formation taxa relative to a monophyletic Paracathartes + Lithornis clade is again unresolved. By contrast, in the constrained analyses, when the position of a lithornithid clade is shifted, Lithornis is recovered as paraphyletic and the Green River Formation taxa are a well-nested subclade. None of these subclade relationships is robustly supported. However, Calciavis grandei, Pseudocrypturus cercanaxius, and all other Green River Formation lithornithid specimens not assignable to either taxon share character states that are not present in other lithornithids (e.g., C22:0, C106:0, C180:0). Pseudocrypturus cercanaxius exclusively shares small posteriorly directed processes of the palatine (C25:2) with Tinamidae, the sister taxon of Lithornithidae in our analysis. Inclusion of these taxa as separate terminals in future more-inclusive analyses may help resolve the placement of Lithornithidae and would reveal whether these character states are plesiomorphies or apomorphies for the clade.

In the analyses recovering a paraphyletic Lithornis (figs. 25, 27), Lithornis celetius is placed as the sister taxon of all other lithornithid species. The skull of Lithornis celetius is the only lithornithid to bear the plesiomorphic condition of a rounded posterolateral portion of the nasal (C7:0). Lithornis celetius also has a well-developed preacetabular pectineal process of the pubis (C154:2), a character state present in Tinamidae and other palaeognaths, but not in any other lithornithid. Lithornis celetius also bears a posteriorly elongated and tapering ischium (C156:1), a plesiomorphic character state found outside Aves, but not among any other lithornithid. To date, L. celetius is the oldest-named taxon (see appendix 2), but from our analysis, it is not clear whether the character states shared with Tinamidae are plesiomorphic or independently derived.

Variation within Lithornithidae

In an attempt to resolve the interrelationships of Lithornithidae, we added a number of characters that were variable among observed lithornithids. Some of these characters, although variable in lithornithids, were also variable across avian taxa included in the analysis, could not be scored in a number of lithornithid taxa, or represented autapomorphies (see appendix 2). Nevertheless, the added characters did not help resolve the interrelationships of Lithornithidae but may be useful for future more-inclusive analyses. With a few exceptions, lithornithid holotypes are mostly complete and well preserved relative to other extinct avian specimens from the Paleogene. However, most specimens still lack well preserved and three-dimensional cranial material, particularly of the palate. Additionally, as remarked by Houde (1988), lithornithids appear to display relatively conservative skeletal morphology across taxa more than five million years apart. With these caveats in mind, we present the following discussion of character states that are variable among lithornithid taxa for future analyses. Furthermore, the character states the some lithornithids share with nonlithornithid avians will also be helpful for further testing the relationships of lithornithids.

The posterodorsal portion of the external naris has a rounded margin bounded by a posterolateral process of the nasal in neognaths (Gallus gallus; USNM 19369) and within palaeognaths (Apteryx owenii; USNM 18279). In contrast, the posterodorsal portion of the external naris tapers to a point in lateral view in most lithornithids. In this state, the nasal is compressed dorsoventrally and lacks a significant posterolateral process. This state is observed in tinamous (Crypturellus undulatus; AMNH 6480) and also in Pseudocrypturus (USNM 336103), but is absent in the skull referred to Lithornis promiscuus (USNM 391983). Therefore, this character varies within Lithornithidae as well as among other palaeognaths.

The postorbital process takes two forms in early diverging avians, a small and distinctly tapered process, or a broader process that is little differentiated from the dorsal skull roof elements. The small and tapered process is present in the proximal outgroups to Aves (e.g., Archaeopteryx, Ichthyornis), in neognaths (e.g., Chauna torquata; AMNH 1773), in most tinamous (e.g., Crypturellus undulatus; AMNH 6480) examined by us, and in nearly all lithornithids with well-preserved skulls. The postorbital process is reduced to a poorly projecting knob in ratites (e.g., Struthio camelus, USNM 560081). As pointed out by Houde (1988), the largest lithornithid, Paracathartes howardae (USNM 361415) has a poorly developed, broad postorbital process like that of ratites.

The medial portion of the pterygoid between the basipterygoid process and the articulation with the palatines has a straight lateral margin in neognaths (e.g., Gallus gallus, USNM 19369), Apteryx owenii (USNM 18279), and in tinamous (Timamus major; USNM 621693). Within palaeognaths, the medial side of the pterygoid is clearly medially expanded as in Struthio camelus (USNM 560081) and Dromaeus novaehollandiae (TMM unnumbered). In these taxa, the medial expansion forms a depression that opens ventrally and medially. Unfortunately, the condition in the proximal outgroups to Aves is not known. Among lithornithids, expanded medial processes are present in Lithornis celetius (USNM 290601), a skull referred to Lithornis promiscuus (USNM 391983; fig. 33), a specimen referred to Lithornis vulturinus (MGUH 26770), and Paracathartes howardae (USNM 391984), but clearly are absent in all of the Green River Formation lithornithids (e.g., Pseudocrypturus cercanaxius and Calciavis grandei).

The palate of palaeognaths has been extremely important for reconstructing relationships at the base of Aves and the interrelationships of Palaeognathae (Huxley, 1867; Bock, 1963; Cracraft, 1974; Houde, 1986, 1988; Lee et al., 1997; Livezey and Zusi, 2006; 2007; Bourdon et al., 2009), and the homology of these character states was previously hotly debated. The morphology of the posterior portion of the palatine where it meets the pterygoid varies across Lithornithidae, and character state changes within the clade also occur among other palaeognaths. Furthermore, the plesiomorphic condition for lithornithids is not clear given that the interrelationships of lithornithids are unresolved. State (0) appears to represent the plesiomorphic condition in Aves but is also present in some lithornithids and some other palaeognaths (e.g., Struthio camelus). Nearly all lithornithids (e.g., Lithornis promiscuus MGUH 26770) appear to have state (1) where the palatines posteriorly converge to a point and are completely connected with the pterygoid and this character state is also present in Palaeotis (Mayr, 2015). Pseudocrypturus cercanaxius (holotype), a referred specimen of Lithornis vulturinus (Bourdon and Lindow, 2015; MGUH 26770) and tinamous share character state (2) where the palatines posteriorly converge to a point as a small posteriorly projected process that is free from the rest of the palate. Houde (1986; 1988) used the presence of the posteriorly projected process to infer a close relationship between Pseudocrypturus cercanaxius and tinamous but hypothesized this was plesiomorphic for palaeognaths.

The distribution of this character is not clear given that many early diverging avian clades both lack and retain rather large basal tubera depending on the species-level taxa. Within palaeognaths, significant tubera appear absent in Rhea americana, Dromaeus novaehollandiae, and tinamous (e.g., Crypturellus undulatus; AMNH 6480), but are present in Apteryx owenii (USNM 18279), Emus crassus (Parker, 1895), and Struthio (USNM 560081). Within Lithornithidae, developed tubera are present in Lithornis celetius (USNM 290601) and a skull referred to Lithornis promiscuus (USNM 391983), are absent in Paracathartes (USNM 361415), and appear to be present in a specimen referred to Lithornis vulturinus (MGUH 26770).

A distinct foramen or a deep fossa on the posterior surface of the quadrate lies between the mandibular condyles and the otic process in Struthio camelus (USNM 560081) among palaeognaths and Gavia immer (TMM M-unnumbered) among neognaths sampled here, but it is clear the feature is present in some but not all galloanserines (Elzanowski et al., 2001). In lithornithids, a deep fossa is present in the same area in Lithornis celetius (USNM 290601), Lithornis plebius (USNM 336534), a skull referred to Lithornis promiscuus (USNM 391983), Pseudocrypturus cercanaxius (USNM 336103), and Calciavis grandei (AMNH 30560). A fossa is clearly absent in Paracathartes howardae (Houde, 1988).

Pygostyle morphology was previously used by Houde (1988) to differentiate among his new species of Lithornis, particularly L. promiscuus and L. plebius. We score three morphologies here: taxa with state (0) have no distinct ventral process on the proximoventral portion of the pygostyle; taxa with state (1) have an anteroventrally directed process that tapers well anteroventral of the articulation with the free caudal vertebrae; and those with state (2) have a ventral process that has a clublike anterodorsal expansion on the proximal termination of the process. In some taxa (e.g., Paracathartes howardae, Houde, 1988), the clublike process bears a foramen. The morphology of the ventral process of the pygostyle is thus variable within Lithornithidae, and similar variation is also present in the tinamous species examined for this analysis.

This character attempts to quantify the size of the articular surface of the scapula for the humerus versus the height of the base of the scapula blade. Among lithornithids, Paracathartes (USNM 361419) and Lithornis promiscuus (USNM 336535) have proportionally larger articulations for the humerus than the other lithornithids (e.g., Lithornis celetius; USNM 290554). Furthermore, the midportion of the scapula blade expands in dorsoventral height relative to the more proximal and distal portions in Paracathartes (USNM 361419) and Lithornis promiscuus (USNM 336535). A large articulation surface for the humerus of the scapula is also found in a variety of palaeognaths (tinamous, Rhea) and neognaths (e.g., Gallus gallus) sampled here, suggesting that this character is highly homoplastic.

The shape of the posterior termination of the scapula blade varies across the early diverging avians sampled here and throughout Aves. In most forms, the scapula blade gradually tapers posteriorly and terminates in a fine point. This is also true of nearly all the lithornithids sampled here. In contrast, all the Green River Formation lithornithids that could be scored have a scapula blade that terminates in a blunt and slightly rounded form.

The ventral surface of the proximal portion of metacarpal III bears a well-developed tuber (Houde, 1988) in most lithornithids, including Lithornis promiscuus (USNM 336535), Lithornis plebius (USNM 336534), Calciavis grandei (AMNH 30578), and a referred specimen of Lithornis celetius (Houde, 1988). A tuber is clearly absent in the same location in Paracathartes howardae (USNM 361446) and most other Aves with the exception of the extinct galliform Gallinuloides wyomingensis (WDC CGR–012). A similarly positioned, but poorly developed tuber is also present in Iaceornis marshi (YPM 1734; Clarke, 2004).

The lateral expansion of the antitrochanter of the ilium is variable across Aves. In taxa scored as state (0), the lateral apex of the antitrochanter projects laterally and does not expand posteriorly to obscure the junction between the ilium and the ischium in lateral view. In taxa scored as state (1), the lateral apex of the antitrochanter is shifted posteriorly, so nearly all of the ilium and ischium contact is covered in lateral view. Both character states are widely distributed across the taxa sampled in this analysis, thus suggesting these character states are homoplastic even within sampled subclades (e.g., tinamous). Nevertheless, the character is variable within lithornithids; Lithornis plebius (USNM 336534) and a specimen referred to L. vulturinus (MGUH 26770) are scored as state (0), whereas Lithornis promiscuus is scored as state (1).

The size of the preacetabular pectineal process of the pubis varies in early-diverging avians from a slight change in angle (considered absent here) to an elongated, anteroventrally directed fingerlike process. Among palaeognaths, a well-developed flange (state 2) is present in Struthio (USNM 560081), Apteryx owenii (USNM 18279), and all tinamous (e.g., Eudromia elegans USNM 345096). In most lithornithids, the pectineal process is present and developed as a distinct, but poorly expanded flange (e.g., Lithornis promiscuus, USNM 336535). However, in Lithornis celetius (USNM 290601), the pectineal process of the pubis is more elongate and more like that of some tinamous (Houde, 1988).

The ischium of most basal Aves sampled in this analysis is dorsoventrally tall and terminated abruptly distally. In lithornithids, the posterior portion of the ischium is rarely preserved, but in Lithornis plebius (USNM 336534), the ischium gradually tapers to a thin point posteriorly. In Calciavis grandei (AMNH 30578), the posterior portion of the ischium terminates abruptly like that of other palaeognaths.

The extent of the tarsometatarsal trochlea has been used widely in avian phylogenetics and character state changes are common among basal avians. Nearly all the lithornithids scored here have an asymmetrical distal portion of the tarsometatarsal where the trochlea II is shorter than that of trochlea IV. In contrast, all of the Green River Formation lithornithid specimens including Calciavis grandei and Pseudocrypturus cercanaxius have a symmetrical distal portion of the tarsometatarsus with trochleae II and IV subequal in distal extent.

Relationships of Lithornithidae

The relationships of Paleogene palaeognaths from North America were first tackled by Houde and Olson (1981) and Houde (1986). In a monograph that soon followed, Houde (1988) named three species of Lithornis (L. promiscuus, L. plebius, and L. celetius), named a closely related form Pseudocrypturus cercanaxius from the Green River Formation, referred a partial skeletons of a Lithornis-like taxon to Paracathartes howardae, revised Lithornis taxa from Europe, and hypothesized and detailed the relationships of these taxa to palaeognaths. Houde (1988) hypothesized that Lithornis, Pseudocrypturus cercanaxius, and Paracathartes howardae represented successive outgroups to extant ratites based on hand-drawn character optimizations (Tinamidae were considered neognaths in that analysis). Houde (1988) presented alternative hypotheses of the relationships of Lithornis, Pseudocrypturus cercanaxius, and Paracathartes howardae, but none included a monophyletic group consisting of all three taxa without any extant ratites. Houde (1988) elaborated, however, on the potential character optimization of features of lithornithids among basal avians. Later analyses of lithornithid relationships included supraspecific genus-level terminal taxa (e.g., Lithornis; Dyke, 2003; Clarke, 2004; Livezey and Zusi, 2006; 2007; Johnston, 2011). In all these analyses except that of Livezey and Zusi (2006; 2007), Lithornis was found within Palaeognathae with few palaeognaths sampled (Clarke, 2004), as the sister taxon of extant ratites (Dyke, 2003), and as a ratite as the sister taxon of Tinamidae (Johnston, 2011).

Other than Houde's (1988) analysis, all subsequent analyses did not specifically address the relationships of distinct species of lithornithids, and variation in character states among potential members of Lithornithidae was not addressed. The analysis of Houde (1988) included discussion of differences among lithornithids, specifically Lithornis, Pseudocrypturus cercanaxius, and Paracathartes howardae. However, the discussions of optimizations have drastically changed since Houde's (1988) study.

Relationships at the base of Aves, including a monophyletic Palaeognathae (containing ratites and Tinamidae) and Neognathae (containing Galloanserae and Neoaves) are well supported. Nevertheless, the relationships among Palaeognathae are far from resolved and continue to be a problem when reconstructing relationships from either morphology and/or molecular data (see Bourdon et al., 2009; Hackett et al., 2008; Phillips et al., 2010; Johnston, 2011; Haddrath and Baker, 2012; Baker et al., 2014; Mitchell et al., 2014). Moreover, nearly every possible relationship among palaeognaths has been suggested and supported by at least some data. In this context, we treat our limited osteology-only analyses with caution and understand it to be only one further contribution to an ongoing conversation on the evolution of morphology in Palaeognathae. Rather than proposing our results as definitive, we propose that only more-inclusive molecular, possibly genome-scale phylogenies of palaeognaths and more-inclusive analyses with all fossil taxa can approach this question with more confidence. Specifically, we focused our analysis at investigating the character optimizations among lithornithid taxa as presented by Houde (1988) using the morphological data extracted from the complete and partial specimens from the Green River Formation. Integration of these data resulted in a monophyletic Lithornithidae as the sister taxon of Tinamidae within Palaeognathae in the unconstrained analyses. A Lithornithidae-Tinamidae clade, placed at the base of Palaeognathae is supported in both the “total” and “core lithornithid” analysis by the following list of unambiguously optimized apomorphies:

Out of this list of apomorphies and the taxa sampled here, two apomorphies (extensive contact between the lacrimal and ectethmoid C16:1; small horizontal shelf at the posterior portion of the mandibular symphysis C48:1) were present exclusively in Lithornithidae and Tinamidae. Furthermore, two other apomorphies also are clearly present in Lithornithidae and Tinamidae among Palaeognathae (manual digit III, phalanx 1, length less than 50% the length of digit II, phalanx 1 [C146:1]; plantar surface of the tarsometatarsus trochlea II and IV expanded into distinct “wings” [C177:1] [also in Apteryx owenii]) but also present in some of the sampled neognaths. However, some unambiguous characters supporting this clade (C4:0, C12:0, C62:1, C103:0) likely represent plesiomorphies for Palaeognathae or even Aves based on other phylogenetic analyses of early avian lineages (e.g., Clarke, 2004). Furthermore, forelimb character states optimized as Lithornithidae-Tinamidae apomorphies (C111:1, C141:1, C146:1) are difficult to score in ratites with reduced forelimbs. In sum, the Bremer support of the Lithornithidae-Tinamidae clade is relatively high compared to other subclades, but on closer examination a number of the characters identified as supporting the relationship within the scope of our analysis are likely to be plesiomorphic or homoplastic at a broader scope.

Other evidence has been proposed to suggest a close relationship between Lithornithidae and Tinamidae. For example, an analysis examining tinamou eggshell microstructure (Grellet-Tinner and Dyke, 2005) and eggshell attributed to two lithornithids species (Lithornis vulturinus and Lithornis celetius) suggested that the microstructure of tinamous and lithornithid eggs are more similar to each other than those of ratites. Furthermore, Houde (1988) hypothesized that lithornithids exclusively shared a diminutive caudal series with tinamous, but at the time did not have any complete caudal series of any lithornithid. The complete caudal series of the holotype of Calciavis grandei (AMNH 30578) confirms Houde's (1988) statement. However, it is worth noting that the caudal series is weakly developed with diminutive transverse processes in all ratites; a similarly tiny pygostyle is also common and these morphologies are thus not unique to tinamous and lithornithids (Clarke, 2002; Bertelli et al., 2014). In an analysis of the relationships of palaeognaths based on the morphology of the palate, Johnston (2011) found a sister-taxon relationship between Lithornis and Tinamidae citing one apomorphy, palatine lateral border directed at midline rostral to basipterygoid articulation (Johnston, 2011: char. 17). Dyke and van Tuinen (2004) and Leonard et al. (2005) reported the placement of lithornithids (as the sister taxon to tinamous), and indicated this was recovered through analysis, but the associated data set yielding this result was never published. The data set of Dyke (2003) which appears to form the basis for these analyses does not yield the tree reported and scores characters from specimens not preserving the relevant morphologies (Mayr, 2009).

A Tinamou + Lithornithidae clade, if supported in more inclusive analyses, would have implications for the evolution of tinamous, the only comparatively species-rich clade of palaeognaths so far only represented by a small number extinct forms (Bertelli and Chiappe, 2005; Bertelli et al., 2014). The earliest lithornithids possibly appear in the Paleocene of Europe (e.g., Fissuravis) and are in North America by the Paleocene (Houde, 1988; Stidham et al., 2014), whereas the first tinamous appear in the Miocene of Argentina (Bertelli and Chiappe, 2005; Bertelli et al., 2014). These data have two important implications. First, it is possible that the tinamous total group including lithornithids originated in North America and retracted to their current Neotropical biogeographic distribution from the Paleogene to the Neogene. This pattern would be somewhat similar to other avian subclades with stem members in Paleogene of North America and Europe and a southern extant distribution across former parts of Gondwanan continents (e.g., Steatornithidae, Mayr, 2005; Coracii, Clarke et al., 2009; Coliiformes, Ksepka and Clarke, 2009; Podargiformes, Nesbitt et al., 2011). Second, tinamous may have diverged from other palaeognaths more recently and their crown diversification could be in the Miocene closer to their first fossil records (Bertelli and Chiappe, 2005).

Results constraining the analysis to take into account abundant data from molecular data from nuclear to mitochondrial gene sequence data and including assessment of retroelement insertions and ultraconserved elements (e.g., Phillips et al., 2010; Haddrath and Baker, 2012; Baker et al., 2014; Mitchell et al., 2014) are consistent with a different picture of early palaeognath evolution (figs. 29–31). In all these analyses, Lithornithidae is not recovered with Tinamidae (now well nested within the large flightless ratites) but is again placed at the base of the Palaeognathae. If this topology is correct, the unconstrained analysis may differ because that analysis is influenced by the flight-related characters in the clade, recovering a single loss in a monophyletic Ratitae of traits in the outgroups. Once tinamous are nested within a paraphyletic ratite assemblage in the constrained analyses, flight characters become ambiguously optimized (present in Neognathae, outgroups, and Palaeognathae). Recent phylogenies of extinct and extant palaeognaths (not sampling lithornithids) have painted a complex biogeographical picture largely inconsistent with a single loss of flight and rafting of already flightless taxa during Gondwanan breakup (e.g., Baker et al., 2014; Mitchell et al., 2014). Consistent placement of the volant Lithornithidae at the base of the clade suggests they could yield further insight into the plesiomorphic morphologies for a total group that has apparently lost flight repeatedly and dispersed broadly.