Paraptenodytes, Ameghino, 1891

Bertelli, Sara, Giannini, Norberto P. & Ksepka, Daniel T., 2006, Redescription and Phylogenetic Position of the Early Miocene Penguin Paraptenodytes antarcticus from Patagonia, American Museum Novitates 3525 (1), pp. 1-36 : 3-24

THE SKULL OF PARAPTENODYTES ANTARTICUS

The preserved portions of the skull of Paraptenodytes antarcticus AMNH 3338 include the braincase, caudal and dorsal parts of orbit, occiput, left and right otic regions, basicranium, right quadrate, articular end of the right mandible, and a fragment of the left mandibular ramus. The exact size of the skull is uncertain because of incompleteness. The cranium is 80 mm in length as measured between the prominentia cerebellaris and the rostral preserved portion of the frontal. The distance between external borders of the fossae g. nasalis (interorbital area) is 16 mm. The width at the zygomatic processes is 62 mm.

FRONTAL REGION, ORBIT, AND TEMPORAL REGION: AMNH 3338 retains most of the interorbital region and frontals (fig. 1). In the caudal part of the frontal region, a shallow depression (depressio frontalis) separates the two slight prominences. In extant penguins, this depression is part of a continuous and shallow longitudinal concavity, whereas in Paraptenodytes , the caudal depression is clearly separated from the slightly depressed interorbital area. The supraorbital margins of the frontals preserve the salt-gland fossa (fossa glandulae nassalis), but as Simpson (1946) noted, they lack the supraorbital shelf of the frontal (supraorbital ridge of Pycraft, 1898) that bounds the salt gland laterally in some penguins. Ventrally, a small portion of the mesethmoid remains attached to the frontals (fig. 2). The foramina n. olfactorii open into grooves along the ventral surfaces of the frontals on either side of the mesethmoid (fig. 3). The caudal wall of the orbit is not perforated by fonticuli orbitocraniales. One of the most notable differences with respect to living penguins in this region of the skull is the configuration of the origin of the m. pseudotemporalis superficialis on the area muscularis aspera (orbital surface of the laterosphenoid; fig. 4). The origin in Paraptenodytes is clearly divided into two separate surfaces by a longitudinal crest, creating a ventral surface situated on the caudal wall of the orbital cavity and a dorsal surface bounded by the processus postorbitalis and the caudal margin of the saltgland fossa. In contrast, a single attachment surface occupies this entire region in extant taxa. The ventral portion of the attachment site is much deeper in the fossil than any living form. At the base of the caudal wall of the orbit, the borders of the foramen opticum are irregular in outline, and the foramina ethmoidale are intact.

Simpson (1946) observed that the temporal fossae (fossae temporales) were strongly developed, extending dorsally almost to the sagittal plane and squarely abutting a longitudinal crest (fig. 1). Strong development of the temporal fossa is seen in all known fossil penguin skulls (e.g., Marplesornis novaezealandiae [Marples, 1960], Waimanu tuatahi [ Slack et al. 2006], fossils assigned to the genus Palaeospheniscus [ Acosta Hospitaleche and Canto, 2005], an unamed Eocene fossil from Antarctica [Ksepka and Bertelli, in press], Spheniscus urbanai [ Stucchi, 2002], and S. megarhamphos [ Stucchi et al., 2003]). However, among living taxa, only Spheniscus approaches this degree of development. The left and right temporal crests (cristae temporales) closely approach one another between the temporal fossae, forming an extended sagittal crest (crista nuchalis sagitalis). The temporal crest reaches rostrally onto the base of the postorbital process (processus postorbitalis), preserved on the right side. This crest contacts the transverse nuchal crest (crista nuchalis transversa) at a nearly right angle. The transverse nuchal crest is nearly perfectly aligned to the coronal plane in Paraptenodytes , whereas in other living and fossil penguins this crest curves posterolaterally. Ventrally, the tranverse nuchal crest divides in two segments; the rostral segment extends to the zygomatic process (processus zygomaticus) and the caudal segment extends to the paroccipital process (processus paroccipitalis) (fig. 2). The subtemporal fossa (fossa subtemporalis, origin of m. depressor mandibulae) is located ventrally between the rostral and caudal divisions of the transverse nuchal crest (fig. 5). This fossa is deeply excavated in Paraptenodytes , contrasting with the nearly flat surface seen in living penguins.

OCCIPUT, CRANIAL BASE, AND MIDDLE EAR: Simpson (1946) described the subcircular shape of the cranium in posterior view (fig. 5). This shape is caused primarily by the caudolateral arching of the dorsal and lateral borders of the transverse nuchal crest. The crest arches more sharply ventrally in living forms, giving the skull a more rhomboid outline. The paroccipital process is bifid, with a ventral and a dorsolateral projection. Simpson (1946) noted the uniqueness of this structure compared to the blunt, undivided paroccipital process of other fossil and living penguins. The borders of the process are thicker in Paraptenodytes than in extant taxa. Dorsolateral to this structure, and adjacent to the transverse nuchal crest, there is a small fossa that corresponds to the location of the foramen rami occipitalis arteriae ophthalmicae externae (best observed on the left side). The cerebellar prominence (prominentia cerebellaris) is relatively smaller in Paraptenodytes than in any living form ( Simpson, 1946). A slight median crest (crista nuchalis sagittalis) surmounts the dorsal surface of the cerebellar prominence, a feature also developed in extant Spheniscus . In Paraptenodytes , the crista nuchalis sagittalis is clearly confluent with the crista nuchalis tranversa. On either side of the prominence and on the posterior surface of the paroccipital process, there are depressed circular areas that we interpret as the surfaces for the attachment of the m. splenius capitis. Both are deeper than in any recent form (especially the surface on the paroccipital process). The venae occipitalis externae exits through a small foramen on either side of the foramen magnum, lateral to the paroccipital process (preserved on the left side). This foramen is expanded into a sulcus in some extant penguins.

The occipital condyle (condylus occipitalis) is circular in shape. Anterior to the condyle is a deep, sharply defined subcondylar fossa (fossa subcondylaris) ( Simpson, 1946). The opening for cranial nerve XII (foramen nervi hypoglossi) is preserved lateral to the condyle on both sides. The concave and rhombic parasphenoid plate (lamina parasphenoidalis) has a pair of blunt and weak basal tubercles (tuberculum basilare) at its caudal corners. Simpson (1946) described the tubercles as low, but both are incomplete, so their degree of development is uncertain. The medial process of the mandible articulates with a capping pad on the lateral process of the parasphenoid plate (basitemporal articulation of the mandible; Bock, 1960). The eustachian tubes (tubae auditivae) traverse the rostrolateral border of the parasphenoid plate enclosed in a bony canal and open at the base of the rostrum parasphenoidale. The opening of the eustachian tubes in the otic cavity is not visible because the middle ear region is filled with sediment (fig. 6). The squamosal and otic cotylae (cotylae quadratica squamosi et otici) are clearly separated from the foramen pneumaticum dorsale. Medial to the otical cotyle, a large foramen nervi mandibullaris is situated just anterior to the tympanic cavity. Pycraft (1898) and Saiff (1976) described the morphology of this region in living penguins. Posterior to the tympanic cavity, the openings for cranial nerves IX (foramen nervi glossopharyngealis) and X (foramen nervi vagi) are located in a poorly developed metotic process as described by Saiff (1976) in living penguins. The foramen nervi vagi is large and opens medial to the smaller foramen nervi glossopharyngealis. The canalis caroticicus cranialis and canalis ophthalmicicus externus are fully ossified, as in living penguins ( Saiff, 1976).

PTERYGOID: The pterygoid of Paraptenodytes (fig. 7) is so different from that of living penguins that Simpson (1946:12) noted that, ‘‘[i]f the pterygoid … had been found isolated it would never have been considered spheniscid but perhaps procellariiform’’ (cf. fig. 7 and fig. 7 in Bertelli and Giannini, 2005). The most obvious difference between Paraptenodytes and living penguins is the lack of expansion of the anterior portion of the pterygoid (pes pterygoidei) in the fossil. The pes pterygoidei is expanded laterally into a thin plate in all living penguins, giving the bone a subtriangular shape. Marplesornis also exhibits widened pes pterygoidei (Marples, 1960), but the degree of expansion is much less than that in extant penguins. In Paraptenodytes , a tablike process projects medially about one-third of the bone’s length from the anterior end. This feature is not seen in living penguins or Marplesornis . Simpson (1946) noted that the position of this process would suggest it was part of a basipterygoid articulation, were it not for the absence of a corresponding articular surface on the parasphenoid rostrum. He speculated that this process possibly represented a nonosseus connection between the pterygoid and parasphenoid. A shallow, subcircular fossa is located at the posterior border of the base of the process. Two small articular surfaces, one for the parasphenoid rostrum (facies articularis parasphenoidalis) and one for the palatine (facies articularis palatina) are present at the anterior end of the pes pterygoidei. At the caudal end of the pterygoid, a rounded cotyla (facies articularis quadratica) serves as the articulation with the condylus pterygoideus of the quadrate. Dorsal to this cotyle, a facet forms a second articulation with the quadrate at the base of the orbital process.

QUADRATE: The right quadrate is complete (figs. 8 and 9). The quadrate was reattached to the mandible following Simpson’s (1946) description, but most of the morphology remains observable. Simpson (1946: 12) provided only a few brief remarks on the general structure of the quadrate, noting it is ‘‘more stoutly constructed with a longer otic process than in any recent species.’’ On the articular surface of the otic process (processus oticus), a sligthly larger squamosal capitulum (capitulum squamosum) projects farther dorsally than the otic capitulum (capitulum oticum). The intercapitular incisure (incisura intercapitularis) is shallow (fig. 8), whereas in the living penguins the capitula are separated by a deep incisure. In lateral view (fig. 9), the squamosal capitulum extends ventrally in a prominent hooklike tubercle for the attachment of the pars profunda of the m. adductor mandibulae externus ( Hofer, 1950). In living penguins, this process is clearly separated from the squamosal capitulum. As noted by Simpson (1946), the orbital process (processus orbitalis) is longer than the otic process, the reverse of the condition found in living penguins. The lateral crest (crista lateralis of Elzanowski et al., 2000) for attachment of the tympanic membrane is sharp and extends from the lateral border of the squamosal capitulum to the quadratojugal cotyla (cotyla quadratojugalis) (fig. 8). Adjacent to this crest, an impression located on the lateral side of the otic process near its transition to the body of the quadrate marks the probable attachment of the medial (deep) layers of the m. protractor pterygoidei et quadrati ( Elzanowski et al., 2000). There is no evidence of a medial crest ( Fuchs, 1954; Elzanowski, 1987) or a tympanic crest on the medial or caudal surface of the otic process. The orbital process is concave medially and tapers toward its distal end. In medial aspect, the concavity deepens ventrally into the basiorbital fossa ( Elzanowski et al., 2000), accommodating the insertion of the lateral (superficial) layers of the m. protractor pterygoidei et quadrati ( Elzanowski et al., 2000) (fig. 9). In lateral view, the base of the process is ventrally connected to the quadratojugal articulation by a distinctive orbitocotylar crest ( Elzanowski et al., 2000) ending on the surface of the quadrate body just before the quadratojugal cotyla (fig. 8). The orbitocotylar crest and the ventral margin of the orbital process enclose a deep groove. The pterygoid condyle (condylus pterygoideus) is a round, knoblike process with a constricted base; it projects more anteromedially than the medial condyle (condylus medialis) (fig. 9). A pterygoid facet is located medially at the base of the orbital process (fig. 9). The quadratojugal articulation is a deep, rounded socket (fig. 8). Its caudal margin projects laterally well past the other surface of the rim. There is no sign of pneumatic foramina on the quadrate. The medial and caudal mandibular condyles (condylus medialis and condylus caudalis) are obscured by the attached mandible.

MANDIBLE: Three portions of the mandible are preserved in AMNH 3338: the articular ends of the right and left mandible and a fragment of the left mandibular ramus (figs. 9, 10, and 11). Simpson (1946) commented on very few aspects of the mandibular morphology. He described the quadrate articulation, noting its resemblance to Pygoscelis among living penguins. Also, he described the morphology of the ‘‘postarticular’’ bone, in particular the medial and ‘‘posterior’’ processes, as the greatest distinction of the fossil mandible. Finally, he mentioned that because the mandibular ramus was so crushed, its anatomy and identification were uncertain. However, the incomplete mandible does provide significant information and further comment is warranted.

The lateral and medial cotylae (cotylae lateralis et medialis) are well preserved on the articular area of the left mandible (fig. 10). The positions and shapes of the cotylae are in general similar in dorsal view to those in living penguins; we attribute apparent differences of the medial cotyla to breakage along the anteromedial margin. The most notable difference in Paraptenodytes is that the anterior and posterior margins of the lateral cotyla are more sharply projected. The fossa adjacent to the lateral cotyla is also deeper than in living penguins. As noted by Simpson (1946), there is no clear separation of the medial and retroarticular processes of the articular (processus medialis mandibulae and processus retroarticularis). The medial process is not projected and lacks the hooklike shape typical of all living penguins. The retroarticular process is wider and less projecting than in extant taxa, and the caudal fossa (fossa caudalis) is shallower. In living penguins, the medial and retroarticular processes are connected by a thin sheet of bone. The edge of this sheet of bone is a thin crest in living taxa, but is greatly thickened in Paraptenodytes . A wide groove separates the medial process from the medial cotyla. The depression for the insertion of the m. pterygoideus on the medial surface is extremely deep, much more so than in any living taxon. The features discussed above are also clear on the right mandible, except the cotylae are obscured by the articulated quadrate.

The lateral surface of the mandibular ramus is broken (fig. 11). In medial aspect, portions of the surangular (os supra-angulare), and prearticular (os prearticulare) are preserved. The caudal mandibular fossa (fossa aditus canalis neurovascularis) is large, as in other penguins. The caudal mandibular fenestra (fenestra caudalis mandibulae) is identifiable, but its borders are enlarged by breakage, so its true size is uncertain. The coronoid process (processus coronoideus) is weakly developed. The process must have been either posterior to or level with the caudal mandibular fenestra, though the exact position is uncertain. The prearticular lines the medial surface of the mandible, but the regions adjacent to the surangular and the caudal mandibular fossa are broken away. Dorsally the prearticular is broken away, exposing the surangular in medial view.

SCORING OF PARAPTENODYTES

Thirty osteological characters from Bertelli and Giannini (2005) and one additional character from Mayr (2005) could be scored in Paraptenodytes antarcticus AMNH 3338. A discussion of this scoring follows.

Character 73: Basioccipital, subcondylar fossa (os basioccipitale, fossa subcondylaris): absent or shallow (0); deep (1). All extant penguins lack a distinct subcondylar fossa. The alternative condition—a deep fossa—is present in most outgroup taxa. Paraptenodytes exhibits a deep subcondylar fossa (fig. 4). Simpson (1946:11) noted this character state in AMNH 3338. This is one of the characters that contribute to place this form at the base of the penguin subtree (see below).

Character 74: Supraoccipital, paired grooves for the exit of the v. occipitalis externa (os supraoccipitale, s ulci venae occipitalis externae): absent or poorly developed (0); deeply excavated (1). This structure is apparently not preserved in the right side of AMNH 3338. In the left side (fig. 5), the groove is inconspicuous (state 0), and the foramen for the v. occipitalis externa is placed lateral and adjacent to the foramen magnum.

Character 75: Frontal, salt-gland fossa (os frontale, fossa glandulae nasalis), lateral supraorbital shelf of bone: absent (0); present (1). Paraptenodyptes lacks the lateral supraorbital shelf of the frontal (state 0; fig. 1) that bounds the salt gland laterally in some other penguins (state 1). The fossil AMNH 3338 allows for a positive scoring of this character, as the postorbital process on the left side appears complete and without the rostral extension that is associated with the shelf in the forms that possess it. Also, the salt-gland fossa (supraorbital groove of Simpson, 1946) is very narrow instead of expanded medially and laterally as when the shelf is present.

Character 76: Squamosal, temporal fossa (os squamosum, fossa temporalis), size: less extensive, both fossae separated by considerable cranial surface (at least the width of the cerebellar prominence) (0); more extensive, fossae meeting or nearly meeting at midline of the skull (1). Paraptenodytes resembles Spheniscus in that the temporal fossa is deep and extends dorsally to midsagittal line (state 1; Simpson, 1946). The fossa is so excavated dorsally that a tall sagittal line is present (fig. 1).

Character 77: Squamosal, temporal fossa (os squamosum, fossa temporalis), depth of posterior region: flat (0); shallowly depressed (1); greatly deepened (2). Paraptenodytes shares a temporal fossa greatly deepened posteriorly (state 2; fig. 2) with several extant genera (except Aptenodytes and Pygoscelis ; state 1).

Character 79: Orbit (orbita), fonticuli orbitocraniales: small or vestigial (0); broad and conspicuous (1). These openings in the caudal wall of the orbit communicate with the cranial cavity. They are invariably present and comparatively very large in Aptenodytes (state 1). In Paraptenodytes , the fonticuli are missing altogether, as in several old specimens of extant genera that possess small fonticuli (state 0).

Character 86: ‘‘Basitemporal plate’’ (lamina parasphenoidalis), dorsoventral position with respect to the occipital condyle: ventral to the level of the condyle (0); at the level of the condyle (1); dorsal to the level of the condyle, surface depressed (2). Paraptenodytes , as all extant penguins, exhibits a depressed lamina parasphenoidalis (state 2).

Character 87: Basipterygoid process: absent (0); vestigial or poorly developed (1); well developed (2). Paraptenodytes , as all penguins, lacks a basipterygoid process (state 0). Each side of the rostrum parasphenoidale, where the basipterygoid process would be located if present, is well preserved in AMNH 3338, so the absence of this process could be established positively (fig. 4). Simpson (1946) commented on the presence of an anterior medial facet (his pseudo-basipterygoid facet; see description above).

Character 88: Eustachian tubes (tuba auditiva): open or with very little bony covering near the medial end of the tube (0); mostly enclosed by bone (1). AMNH 3338 shows well-preserved left and right eustachian tubes (fig. 4). The tubes are long (state 1), reaching laterally to the otic cotylae (cotylae quadratica otici, the articulations for the otic capitulum of the left and right quadrate). The medial openings of both tubes (on the ventral aspect of the base of the rostrum parasphenoidale) are also preserved in AMNH 3338. These openings are identical to the openings observed in extant penguins.

Character 89: Pterygoid (os pterygoideum), shape: elongated (0); broad, triangular-shaped (1). Unlike any other penguin for which the pterygoid is known, Paraptenodytes exhibits an elongated, approximately cylindrical pterygoid (state 0; fig. 7). Penguins in extant genera possess a pterygoid with a wide bladelike rostrolateral flange (state 1). Simpson (1946) emphasized the distinctiveness of this bone in Paraptenodytes , much like the pterygoid of procellariiforms (most of our outgroup taxon set).

Character 96: Quadrate, otic process (os quadratum, proc. oticus), ventral border, process for attachment of the m. adductor mandibulae externus, pars profunda ( Hofer, 1950): absent (0); present as a ridge (1); present as a tubercle or short process (2). The process is located on the rostrolateral surface of the otic process of the quadrate, immediately ventral to the squamosal articulation (capitulum squamosum). The right quadrate of Paraptenodytes exhibits a prominent, hooklike process (state 2; fig. 8).

Character 100: Mandible, caudal fenestra (mandibula, fenestra mandibulae caudalis): open, can be seen through from the medial or lateral aspects (0); nearly or completely concealed by the os spleniale medially, that is, fenestra not visible in the medial aspect (1). The preserved (left) mandibular body of AMNH 3338 (fig. 11) shows a distinct mandibular fenestra (state 0).

Character 104: Mandible, articular, medial process (os articulare, proc. medialis mandibulae): not hooked (0); hooked (1). The process is curved rostromedially in all extant penguins. As Simpson (1946) noted, Paraptenodytes lacks a hooked process (fig. 10).

Character 105: Mandible, angular, retroarticular process (mandibula, os angulare, proc. retroarticularis), aspect in dorsal view in relation to the articular area with the quadrate (area between the lateral condyle [condylus lateralis] and medial condyle [condylus medialis]): broad, approximately equal to the articular area (0); moderately long, narrower than the articular area (1); very long, longer and narrower than the articular area (2). Paraptenodytes exhibits a broad retorarticular process (state 0) that Simpson (1946) deemed similar to Pygoscelis (fig. 10). The outgroup taxa lack retroarticular process altogether (see character 106), and therefore were coded as noncomparable.

Character 106: Mandible, angular (mandibula, os angulare), aspect in dorsal view: sharply truncated caudally (0); caudally projected, forming the retroarticular process (proc. retroarticularis) (1). In Paraptenodytes , as in all penguins, the angular projects caudally, forming a retroarticular process (state 1), whose variation in shape was discussed in character 105 (fig. 10). This character supports the monophyly of Sphenisciformes .

Character 107: Mandible, caudal fossa (mandibula, fossa caudalis): shallow (0); deep (1). In Paraptenodytes , the caudal fossa of the mandible is shallow (state 0).

Character 108: Atlas (atlas), proc. ventralis corporis: absent or sligthly developed (0); well developed, high with a prominent ridge (crista ventralis corporis) on the ventral surface of the corpus atlantis (1). Paraptenodytes clearly exhibits character state 1 (fig. 12) and shares it with Aptenodytes (and Gavia among the outgroup taxa), as noted by Simpson (1946).

Character 120: Coracoid, foramen nervi supracoracoidei: present (0), absent (1). This character was described by Mayr (2005), who discussed the homology of the conditions present in penguins and outgroup taxa. As in extant penguins, the foramen is absent in Paraptenodytes (fig. 16).

Character 122: Forelimbs (ossa alae), strongly flattened: absent (0); present (1). The humeri of AMNH 3338 are similar to those of other penguin forms, that is, strongly flattened (state 1; fig. 17).

Character 123: Humerus, aspect of pneumatic fossa (humerus, fossa pneumotricipitalis): small with pneumatic foramina (0); without pneumatic opening, moderate size (1); without pneumatic opening, great size and deep fossa (2). All penguin taxa, including Paraptenodytes , show a large pneumatic fossa without pneumatic opening (state 2; fig. 18).

Character 124: Humerus, head (humerus, caput humeri) large and reniform in shape, ventrally directed: absent (0); present (1). The shape of the humerus head in Paraptenodytes is typical of penguins (state 1, fig. 18).

Character 125: Humerus, pneumatic fossa (humerus, fossa pneumotricipitalis) subdivision into cavities: undivided (0); divided (1). In the previous analysis of this character, Bertelli and Giannini (2005) considered three states for the conditions observed in the pneumatic fossa— single, partially divided, and bipartite. The intermediate state was present only in Eudyptula minor . We reconsidered this coding and included just two states, the presence/ abscence of a divided pneumatic fossa. Paraptenodytes , as noted by Simpson (1946), lacks a distinct division of the pneumatic fossa in either the left or right humerus, and therefore it was scored 0 (fig. 18).

Character 126: Humerus, development of dorsal supracondylar process (humerus, proc. supracondylaris dorsalis): absent (0); compact tubercle (1); very long process (2). Paraptenodytes shares with all extant penguins the lack of supracondylar process (state 0; fig. 17).

Character 127: Humerus, deltoid crest, area of attachment of pectoral muscle (humerus, crista deltopectoralis, impressio m. pectoralis): shallow groove (0); deep, oblong fossa (1). The left and right humerus heads are preserved in AMNH 3338. As in extant penguins, the impression for the attachment of the pectoral muscle in Paraptenodytes is a deep fossa (state 1; fig. 17).

Character 128: Humerus, distal end, ventral border with ‘‘trochlear process’’ (caudalmost crest at the epicondylus ventralis caudally bordering the sulcus humerotricipitalis): present (0); absent (1). Paraptenodytes exhibits a trochlear process, a feature diagnostic of penguins (state 1; fig. 17).

Character 129: Humerus, proximalmost ‘‘trochlear process’’: extends beyond the humeral shaft (0); does not extend beyond the humeral shaft (1). The proximal trochlear process does not extend beyond the humeral shaft in Eudyptula and Spheniscus (state 1) and in Paraptenodytes . In the remainder of extant genera, the proximal trochlear process does extend beyond the humeral shaft (state 0; fig. 17).

Character 138: Tarsometatarsus: slender, proximodistal length much greater than mediolateral width (0); very stout, mediolateral width nearly equal to proximodistal length (1). In AMNH 3338, the left and right tarsometatarsus are not complete. However, the preserved parts (fig. 21) permit a positive scoring of Paraptenodytes , which exhibits a typical sphenisciform tarsometatarsus (state 1).

Character 139: Tarsometatarsus, blood vessel foramen located on fossa para hypotarsalis medialis: absent (0); present (1). Both the left and right tarsometatarsus of AMNH 3338 lack this vascular foramen (state 0; fig. 21).

Character 140: Tarsometatarsus, medial proximal vascular foramen (foramina vascularia proximalia): absent (0); present (1). The proximal foramen opens lateral to medial crest (crista medialis hypotarsi). The left and right tarsometatarsus of AMNH 3338 exhibit this foramen visible in the plantar side (state 1; fig. 21).

Character 141: Tarsometatarsus, hypotarsus, tendinal canals (hypotarsus, canales hypotarsi): present (0); absent (1). All penguins, including Paraptenodytes , lack tendinal canals in the proximal end of the tarsometatarsus (state 1; fig. 21).

Character 142: Tarsometatarsus, tuberositas m. tibialis cranialis: flat (0); raised (1). A raised tuberositas tibialis is present in all penguins (state 1) except Aptenodytes (state 0). AMNH 3338 shows a raised tubercle in both the left and right tarsometatarsus (fig. 21).

TOPOLOGIES

The analysis including only the osteological characters initially resulted in 425 most parsimonious trees; an extra TBR round on optimal trees elevated this number to a total of 2004 equally optimal trees at 148 steps (strict consensus in fig. 22) The consensus recovered Procellariiformes (unresolved) and Sphenisciformes , but overall resolution and support was poor. Absolute Bremer support values (hereafter ABS) were low, as ABS = 2 in all resolved nodes (calculated on the basis of a sample of 16,000 suboptimal trees). Paraptenodytes antarcticus was the sister to all extant penguins. Eudyptula , a monophyletic Spheniscus , and a clade containing the remaining genera formed a trichotomy. Within the last group, two clades appeared, one including Pygoscelis papua , a clade containing the other two species of Pygoscelis , and Aptenodytes , and another clade formed by Megadyptes and Eudyptes . Relationships within the polytypic genera were unresolved.

The analysis including all nonmolecular characters (morphological and behavioral traits) yielded two optimal trees of 393 steps (strict consensus in fig. 23). The tree was almost fully resolved, the consensus showing just a trichotomy within Spheniscus . Within procellariiforms, albatrosses formed a clade sister to all other procellariforms. In the latter clade, the successive sister groups were Macronectes, Daption , and a clade formed by shearwaters ( Procellaria + Puffinus ), Pterodroma , and a group composed of an oceanitid clade ( Oceanites + Oceanodroma ) and a composite clade including the prion Pachyptila and the diving petrel Pelecanoides . In the penguin subtree, Paraptenodytes was the sister of all extant forms, Spheniscidae (sensu Clarke et al., 2003) . The topology within Spheniscidae is the same as that recovered by the morphological analysis of Bertelli and Giannini (2005). Eudyptula + Spheniscus form the sister group to a clade containing the remainder of the genera. The latter clade is split into two groups, one including Aptenodytes and Pygoscelis , the other including Megadyptes and Eudyptes . In Spheniscus , the Pacific forms S. humboldti and S. mendiculus grouped together. Within Pygoscelis , P. antarctica was sister to the other two species. Eudyptes pachyrrhynchus was sister to the other congeners, and successive sister groups were ( E. robustus + E. sclateri ), ( E. chrysocome chrysocome + E. chrysocome moseleyi ), and ( E. chrysolophus + E. schlegeli ). ABS values, calculated from a sample of 10,520 suboptimal trees up to seven steps longer than the optimals, were higher for recovered genera (3 ≤ ABS ≤ 7) and for orders (ABS for Procellariiformes and Sphenisciformes = 7). Relative Bremer support values (RBS) followed a pattern similar to absolute Bremer values; that is, backbone nodes exhibited the highest conflict (RBS as low as 23), recovered genera (RBS ≥ 40) and Sphenisciformes (RBS as high as 100) the lowest conflict.

The combined analysis (molecular + nonmolecular data) yielded two trees of 2329 steps (strict consensus in fig. 24). Procellariiformes and Sphenisciformes were monophyletic. The former clade was fully resolved; the successive sister clades recovered were Oceanitidae , Diomedeidae , ( Pterodroma + Puffinus ), ( Pachyptila + Pelecanoides ), Procellaria , and the fulmarine petrels (Daption + Macronectes ). Support values were moderate to high (4 ≤ ABS ≤ 17), except for the sister group to Oceanitidae (ABS = 1). The penguin subtree was fully resolved except for a trichotomy in Spheniscus . Paraptenodytes appeared as sister to all extant forms. Successive sister groups within the crown group of extant penguins were Aptenodytes , Pygoscelis , and clade E of Bertelli and Giannini (2005), composed of ( Eudyptula + Spheniscus ) and ( Megadyptes + Eudyptes ). Intrageneric relationships were as follows. Within Pygoscelis , unlike the nonmolecular analysis, P. adeliae was sister to the other two species. Relationships remained unchanged within Spheniscus and Eudyptes with respect to the nonmolecular analysis. Lowest support values occurred within Eudyptes (1 ≤ ABS ≤ 3). The backbone of the penguin subtree was moderately supported (4 ≤ ABS ≤ 5), taking into account the fact that this arrangement of suprageneric groups strongly contradicts the nonmolecular analysis. The conflict between the two trees reflects the differential rooting of the same basic network favored by the molecular (on Aptenodytes ) and nonmolecular (on Eudyptula + Spheniscus ) partitions, as discussed by Bertelli and Giannini (2005). All genera were recovered and well supported (6 ≤ ABS ≤ 8).