Kupoupou sp.

Mayr, Gerald, De Pietri, Vanesa L., Proffitt, James, Blokland, Jacob C., Clarke, Julia A., Love, Leigh, Mannering, Al A., Crouch, Erica M., Reid, Catherine & Scofield, R. Paul, 2025, Multiple exceptionally preserved fossils from the Paleocene Waipara Greensand inform the diversity of the oldest stem group Sphenisciformes and the formation of their diving adaptations, Zoological Journal of the Linnean Society 204 (4) : -

? Kupoupou sp. ( Fig. 15)

Referred specimen: UC 22084 ( Fig. 15 KK–MM; right tarsometatarsus).

Locality and horizon: Waipara Greensand , Waipara River, Canterbury, New Zealand; from site S2; found lying loose; collected on 7.12.2010 by L. Love .

Measurements (in mm): See Table 1 for the tarsometatarsus length; tarsometatarsus, distal end, 23.7.

Remarks: In size and overall proportions, the specimen corresponds to Kupoupou stilwelli from the Late Paleocene of the Chatham Islands ( Fig. 15 NN, OO; Blokland et al. 2019). In its morphology, it also resembles this species, but the trochlea metatarsi IV is mediolaterally narrower and in distal view it does not project as much dorsally as it does in K. stilwelli .

Summary description of the skeletal morphology of the ‘ Muriwaimanu -like’ penguins from the Waipara Greensand In general, the osteology of the above species conforms well to that of Sequiwaimanu rosieae , which was described in detail by Mayr et al. (2018). In the following, we therefore focus on features that depart from the S. rosieae osteology, show variation within the new species, were previously unknown for stem group sphenisciforms, or are otherwise of taxonomic, phylogenetic, or evolutionary interest.

Substantial portions of the skull are known for Muriwaimanu tuatahi (CM zfa 33, UC 22078), cf. Waimanutaha kenlovei (UC 22079), Archaeodyptes waitahaorum (CM 2020.46.1), and Waiparadyptes gracilitarsus (CM 2013.27.1). Accounting for artefacts of preservation and diagenetic deformation, the specimens exhibit similar overall morphologies. In all species of which the upper beak is known (CM 2020.46.1, UC 22078, UC 22079), it is extremely long and pointed, measuring about 65% of the entire skull length ( Fig. 3). Unlike what has been reported for the equally long-beaked Icadyptes salasi Ksepka et al., 2008 from the Late Eocene of Peru, in which the premaxillaries are more extensively fused ( Ksepka et al. 2008, Chávez-Hoffmeister 2020), the premaxillary bones of the above species from the Waipara Greensand are separated along most of their length ( Figs 3C, 4B); the gap between them is narrower than in crown group Sphenisciformes . The nostrils are long and slit like ( Fig. 3D). Their caudal end is close to the nasofrontal hinge; rostrally, they extend near to the tip of the beak, but do not reach as far rostrally as the gap between the premaxillaries (in I.salasi the nostrils are proportionally shorter; Clarke et al. 2007, Ksepka et al. 2008). There is a pair of longitudinal furrows along the tip of the premaxilla and dorsal to the nostrils in UC 22078 ( M. tuatahi ), UC 22079 (cf. Wt. kenlovei ), and 2020.46.1 ( A. waitahaorum ) ( Figs 4A, C, D). Similar, albeit less pronounced rostral furrows were reported by Clarke et al. (2010; Supporting Information, Fig. S1) for Inkayacu paracasensis Clarke et al., 2010 from the Late Eocene of Peru and by Jadwiszczak (2011) for? Palaeeudyptes klekowskii Myrcha et al., 1990 from the Late Eocene of Antarctica. The internarial bar of M. tuatahi is mediolaterally narrower than in the equally long-beaked S. rosieae , and the beak altogether has a somewhat more gracile appearance.

The right lacrimal is preserved in situ in CM zfa 33 and is not fused to the frontal ( Fig. 4H); in all other specimens the lacrimals are detached from the skull. The skull of UC 22078 shows the rostral portions of the ossa palatina ( Fig. 3C). The robust jugal bars are straight ( Fig.3E), whereastheyaremarkedlycurvedincrowngroup Sphenisciformes . The septum interorbitale of CM 2013.27.1 ( Wp. gracilitarsus ) exhibits a large interorbital fenestra. The interorbital section of the frontals is mediolaterally narrow. The narrow fossae glandularum nasales run along the dorsal rims of the orbits ( Fig.5E). The neurocranium of CM 2013.27.1is dorsoventrally narrower than that of UC 22078 ( Fig. 4G, L), but it is unclear whether this represents the true condition or is an artefact of preservation. On the dorsal surface of the neurocranium, the cristae nuchales transversae and the cristae temporales almost meet and form a distinctive cruciform structure. The long and narrow processus postorbitales are laterally projected, which distinguishes them from the smaller and more ventrally pointing ones of crown group sphenisciforms. The fossae temporales are deeply incised and extend onto the caudal surfaces of the processus postorbitales; they almost reach the midline of the neurocranium but do not meet (the fossae temporales are somewhat deeper in CM 2013.27.1 than in UC 22078). The prominentia cerebellaris is strongly developed.The tubercula basilaria are prominent; rostrallytheyblendintoanarcuateridge, whichdelimits an essentially flat lamina parasphenoidalis.The large processus paroccipitalis is more caudally projected than in extant sphenisciforms; compared to crown group Sphenisciformes , it is also separated by a deepernotchfromtheprocessuszygomaticus, whichgivesthesquamosal portion of the otic region a bilobed appearance in dorsal view ( Fig. 5B). The foramen magnum is large, as is the hemispherical condylus occipitalis ( Fig. 3B). UC 22078 ( M. tuatahi ) shows well-developed, stalked processus basipterygoidei ( Fig.3C), which were also identified in CT scans of CM 2013.27.1 ( Wp.gracilitarsus , Fig.5B); these processes are absent in crown group Sphenisciformes and have not yet been reported from other fossil sphenisciform taxa.

A well-preserved isolated quadrate is associated with CM 2020.46.1 ( A. waitahaorum ; Fig. 6A–E). An isolated quadrate is also present in CM 2009.99.1 ( M. tuatahi ), but much of its morphology is obscured by adhering sediment; of the other quadrate of CM 2009.99.1 only the caudal portion is exposed ( Fig. 6F–J). The quadrate of Wp. gracilitarsus (CM 2013.27.1) is likewise embedded in matrix and surrounded by other bones ( Fig. 6K, L). The processus oticus is mediolaterally wider than in crown group sphenisciforms ( Fig. 6M–O), the incisura intercapitularis broader, and the capitulum squamosum is less globose than in crown group sphenisciforms. While it is not completely preserved, the processus orbitalis is longer than in crown group sphenisciforms; in A. waitahaorum it is dorsoventrally narrower than in M. tuatahi . The condylus pterygoideus is less rostrally and more medially projected than in crown group sphenisciforms.

Nearly complete mandibles are present in CM 2018.124.4 ( Wt. kenlovei ) and UC 22079 (cf. Wt. kenlovei ); the otherwise well-preserved mandible of CM 2013.27.1 ( Wp. gracilitarsus ) lacks the rostral portion. The rami run in parallel in the rostral section of the mandible and are closely adjacent ( Fig. 5H). The symphysis mandibulae is rostrocaudally long. On the medial surface of the midsection of the ramus there is a marked but dorsoventrally narrow fossa aditus canalis mandibulae ( Fig. 5F). The caudal end of the mandible of CM 2018.124.4 ( Fig. 5K) was described byMayr et al. (2020b). In the other species it appears to be similar, even though the processus medialis is more caudally directed in Wp. gracilitarsus (CM 2013.27.1; Fig. 5L). In Wp. gracilitarsus , the caudal surfaces of the rami mandibularum exhibit marked fossae, which can also be discerned in UC 22078. In extant sphenisciforms, these fossae are dorsally directed, and the caudal portions of the mandible form processus retroarticulares ( Fig. 5J), which are absent in Wt. kenlovei and Wp. gracilitarsus (a short processus retroarticularis was reported for S. rosieae and an unnamed sphenisciform from Chatham Island; Mayr et al. 2018, Blokland et al. 2019). A well-defined tuberculum pseudotemporale is visible in UC 22078 ( Fig. 5M).

Most specimens preserve vertebrae, but these are often in a tangle with other bones, which impedes a study of their morphology. The atlas is exposed in UC 22079 (cf. Wt. kenlovei ), and CT scans show the atlas of CM 2009.99.1 ( M. tuatahi ) to be mediolaterally wider than that of CM 2010.108.1 (cf. Wt. kenlovei ) ( Fig. 7A, B). UC 22079 also includes the axis ( Fig. 7C–E), which is proportionally longer than in crown group Sphenisciformes and has broader zygapophyses (processus articulares) caudales; the processus spinosus is less dorsally projected than in S. rosieae ( Fig. 7F, G), with the processus spinosus of the latter being more like that of crown group sphenisciforms ( Fig. 7H, I). Unlike in crown group sphenisciforms, some of the cranial cervical vertebrae, presumably the third or fourth, exhibit large lateral foramina, which are formed by osseous bridges between the processus transversi and the zygapophyses caudales ( Fig. 7K–O). The cranial cervical vertebrae of extant penguins either lack lateral foramina ( Aptenodytes , Pygoscelis , Eudyptes , Eudyptula , Spheniscus ) or they are very small or incompletely closed ( Megadyptes antipodes ). The thoracic vertebrae of M. tuatahi have flat articular surfaces and, therefore, are acoelous (in extant penguins, the thoracic vertebrae are opisthocoelous, with convex cranial and concave caudal articular surfaces). Unlike in extant Sphenisciformes , their corpus bears deep lateral fossae (see Mayr 2021 concerning the occurrence of these fossae in neornithine birds). The synsacrum (CM 2018.124.4 and CM 2010.108.3) consists of 13 co-ossified vertebrae. Unlike in crown group Sphenisciformes , the cranial-most vertebra of the synsacrum has a flat rather than convex facies articularis cranialis, which is mediolaterally wider than it is dorsoventrally deep. The caudal vertebrae have long processus transversi. The pygostyle is present in CM 2020.46.1 ( A. waitahaorum ), UC 22078, and CM 2009.99.1 (both M.tuatahi ). The pygostyle of CM 2009.99.1 ( Fig. 7 CC, DD) differs from that of CM 2020.46.1 ( Fig. 7 HH–JJ) in that the ventral portion of the cranial-most co-ossified caudal vertebra is separated by a notch from the main body of the pygostyle, the dorsal margin of the bone exhibits a sigmoidal curvature, and the caudal tip tapers to a point (in CM 2020.46.1 the caudal tip is broadly rounded); the shape of the pygostyle of UC 22078 appears to be similar to that of CM 2009.99.1, but its morphology is obscured by adhering matrix ( Fig. 7 EE–GG). The lamina pygostyli is mediolaterally wide with convex—rather than flat—lateral surfaces.In all fossils, the bone is very different from the pygostyle of Kairuku grebneffi Ksepka, Fordyce et al., 2012 from the Late Oligocene of New Zealand ( Ksepka et al. 2012) and from that of extant sphenisciforms, in which the bone is proportionally longer and more elongated ( Fig. 7 KK, LL). Ribs are present in multiple specimens and as in extant Sphenisciformes they lack co-ossified uncinate processes.

The coracoid ( Fig. 8) is known from all species except D. primaevus and A. waitahaorum (for Wt. kenlovei it is only preserved in the tentatively referred specimens). The processus procoracoideus, the tip of which is broken or obscured in many specimens, is somewhat dorsally deflected in M. tuatahi (CM zfa 34, CM 2010.108.3, CM 2009.99.1), but not in UC 22079 (cf. Wt. kenlovei ). The cotyla scapularis is moderately concave and of circular outline. A foramen nervi supracoracoidei is absent in all specimens. The extremitas sternalis of CM 2010.108.3 exhibits a sterno-omally directed ridge in the lateral portion of its ventral surface ( Fig. 8G); this ridge also occurs in the holotype of S. rosieae (see Mayr et al. 2018). The angulus medialis is markedly medially projected in CM zfa 34 and UC 22078 (both M. tuatahi ); immediately next to it, the medial margin of the bone forms a convexity, which is broken or damaged in other specimens (e.g. CM 2013.27.1, CM 2010.108.3). The dorsal surface of the sternal extremity of the right coracoid of UC 22078 exhibits multiple shallow muscle striae. The processus lateralis is longest in UC 22079 (cf. Wt. kenlovei ; Fig. 8K).

The scapula is completely preserved in UC 22078 ( M. tuatahi ; Fig. 9M, N) and nearly complete in CM 2013.27.1 ( Wp. gracilitarsus ; Fig. 9R, S). In these two specimens it has very different proportions. In contrast to previous assumptions about the shape of the scapula of M. tuatahi (see Mayr et al. 2018), the corpus of the bone is caudally widening in UC 22078, whereas it is narrow and has an equal width over its length in CM 2013.27.1. The scapula also widens caudally in CM 2010.108.1 (cf. Wt. kenlovei ; Fig. 9T), as it does in S. rosieae and more crownward sphenisciforms.

None of the specimens includes an intact furcula, but substantial portions of the bone are present in CM 2009.99.1 ( M.tuatahi ), where it is narrowly U-shaped ( Fig.9A–D). The extremitas omalis is widened and caudally deflected. The presence of a well-developed processus acrocoracoideus in M. tuatahi specimen CM zfa 34 ( Fig. 9E, F) is notable, but may be a pathological artefact (see above); all other specimens lack this process. In UC 22079 (cf. Wt. kenlovei ), the processus acromialis is proportionally shorter than in M. tuatahi and A. waitahaorum , and its tip is more rounded ( Fig. 9H). The extremitas sternalis is robust and evenly curved in M. tuatahi . Waiparadyptes gracilitarsus (CM 2013.27.1) differs from the other taxa described in this study as well as from most other extant and fossil sphenisciforms in that the extremitas sternalis bears a knob-like apophysis furculae ( Fig. 9K); the only other sphenisciform for which a well-developed apophysis furculae has been reported is In. paracasensis from the Late Eocene of Peru, a Palaeeudyptes species, and Platydyptes marplesi Simpson, 1971 from the Oligocene-Miocene of New Zealand (seeMarples 1952, Simpson 1971, Clarke et al. 2010), but a rudimentary apophysis is also present in an undetermined sphenisciform from the Late Eocene of Antarctica ( Jadwiszczak 2006). The extremitas sternalis of the furcula of S. rosieae has a rugose cranial surface and exhibits two small projections ( Mayr et al. 2018).

A largely complete sternum is present in CM 2009.99.1 ( M.tuatahi ; Fig. 10A–E); substantial portions of the corpus sterni are also preserved in CM 2010.108.3 ( M. tuatahi ; Fig. 10I) and UC 22079 (cf. Wt. kenlovei ; Fig. 10J). The sternum of CM 2009.99.1 exhibits a distinctive morphology in that the apex carinae is markedly cranially projected and bears a well-developed articular facet for the furcula ( Fig. 10D, correction of statement inMayr et al. 2018). This articular facet is also present in S. rosieae , in which it is markedly concave ( Mayr et al. 2018). Because the carina sterni is unknown from D. primaevus , Wp. gracilitarsus , and A. waitahaorum , it cannot be determined whether the morphology of CM 2009.99.1 is autapomorphic for M. tuatahi or whether it represents a feature of all sphenisciforms from the Waipara Greensand. Even though the latter assumption appears most likely based on the otherwise similar morphologies of the pectoral girdle, the furcula of Wp. gracilitarsus (CM 2013.27.1) exhibits an apophysis and its topologic relation to the tip of the carina sterni may therefore have been different from M. tuatahi and S. rosieae ( Fig. 10F). A cranially projected apex carinae also occurs in species of Kairuku Ksepka et al., 2012 (see Ksepka et al. 2012), in which an articular facet for the furcula is absent. The sternum of CM 2009.99.1 originally had a small but well-delimited and dorsally directed spina externa, which unfortunately is now broken and lost ( Fig. 10A, C). In crown group Sphenisciformes , the spina externa is variably developed ( Fig. 10G, H), but in most species except those of Spheniscus ( Fig. 10H) it is cranially directed; the condition in S. rosieae , species of Kairuku , and other early stem group sphenisciforms is unknown. There are five processus costales in CM 2009.99.1, but since the caudal portion of the margo costalis appears to be missing, this number is likely to have been higher. The caudal margin of the sternum is best preserved in CM 2009.99.1, where it shows one pair of incisions; the trabecula lateralis projects caudally beyond the trabecula mediana.

Even though the humeri exhibit a similar overall morphology, they distinctly differ in some morphological features, which is particularly true for the length of the tuberculum dorsale, the shape and position of the condylus dorsalis, and the sizes of the ridges on the distal end of the bone ( Fig. 11). The tuberculum dorsale (‘crista musculi supracoracoidei’ of other authors) of D. primaevus is the shortest and hardly reaches as far distally as the proximal terminus of the crus dorsale fossae ( Fig. 2F). The tuberculum dorsale is of intermediate length in A. waitahaorum , in which it reaches distally slightly beyond the proximal terminus of the crus dorsale fossae ( Fig. 2D), and it is longest in the M. tuatahi holotype (OU 12651; Fig. 2A) and paratype CM zfa 33 ( Fig. 11D), where it extends to the level of the distal terminus of the tuberculum ventrale; however, unlike in post-Paleocene sphenisciforms it still does not reach as far distally as the distal terminus of the crista bicipitalis in these taxa. In some of the specimens referred to M. tuatahi (CM 2008.145.3, CM zfa 34, CM 2009.99.1), the tuberculum dorsale is proportionally somewhat shorter and wider than in the holotype and terminates at the level of the crus dorsale fossae. The crista bicipitalis forms a marked distal convexity in M. tuatahi , which is less developed in Wt. kenlovei and absent in A. waitahaorum and D. primaevus (the proximal end of the humerus of Wp. gracilitarsus is unknown); a similar convexity occurs in other Paleogene and Early Eocene stem group sphenisciforms, including Kumimanu biceae, Petradyptes stonehousei , and Kaiika maxwelli . The ridge, which separates the incisura capitis from the sulcus transversus is proportionally wider in M. tuatahi than in A. waitahaorum ( Fig. 2C, E). The impressio coracobrachialis is a marked elongate fossa in most specimens, but it is less sharply delimited in UC 22081 ( D. primaevus ). None of the humeri exhibits a distinct sulcus for the coracobrachialis nerve, which occurs in In. paracasensis from the Late Eocene of Peru ( Clarke et al. 2010). The shaft of the bone is somewhat wider in Wt. kenlovei than it is in M. tuatahi . The dorsal margin just above the condylus dorsalis forms a tab-like protrusion ( Fig. 2G). The condylus dorsalis itself is situated in the dorsal-most section of the bone in M. tuatahi and the new taxa, but does not project as much dorsally as it does in S. rosieae , where it is situated on the dorsal margin of the bone ( Fig. 2O–Q). In Wt. kenlovei and M. tuatahi , the elongate condylus ventralis is only weakly distally and cranially protruding, but the condylus ventralis of D. primaevus and Wp. gracilitarsus is more globose than that of the other taxa. In Wp. gracilitarsus (CM 2013.27.1), the distal end of the humerus is furthermore markedly oblique in cranial view and the ventrodistal portion is strongly projected. In the other taxa, the distal end has a broadly rounded ventrodistal portion, which in A. waitahaorum , M. tuatahi , and Wt. kenlovei forms three marked ridges, that are generally termed ventral, intermediate, and dorsal ones (strictly spoken, and based on the orientation of the humerus in Paleocene stem group sphenisciforms, they would better be designated as cranial, intermediate, and caudal ridges). In A. waitahaorum , the distal margin of the sulcus scapulotricipitalis forms a convex protrusion ( Fig. 2K). In A. waitahaorum , M. tuatahi , Wt. kenlovei , and S. rosieae , the ventral and intermediate trochlear ridges have a similar ventral extent ( Fig. 2H, J, L, Q). Daniadyptes primaevus is characterized by a shorter intertrochlear ridge, which does not reach as far ventrally as the ventral ridge ( Fig. 2N). Unlike in other sphenisciforms, and in distal view of the humerus, the condylus dorsalis projects only slightly further dorsally than the dorsal margin of the condylus ventralis. Furthermore, and unlike in other stem group sphenisciforms, the dorsal trochlear ridge of D. primaevus is inconspicuous; this may be a plesiomorphic feature.

The ulna is preserved in specimens of M. tuatahi ( Figs 2S, 12B, C), Wt. kenlovei , A. waitahaorum ( Figs 2T, 12I), and (albeit only partially) Wp. gracilitarsus , and closely resembles the ulna of S. rosieae , which was described in detail by Mayr et al. (2018). As detailed by these authors, the stem group sphenisciforms from the Waipara Greensand are characterized by the plesiomorphic presence of a processus cotylaris dorsalis (this process was mistakenly termed processus supracondylaris dorsalis in the material and method sections of Mayr et al. 2018, 2020b and in the character list of Blokland et al. 2019 —it was referred to its correct name in the main text of these papers). Unlike in S. rosieae , the dorsal surface of the proximal end of the bone lacks an ‘edge-like jut’ ( Mayr et al. 2018: fig.10O). As in S. rosieae , the condyles of the distal end are better developed than in extant sphenisciforms.

The radius of M. tuatahi ( Fig. 12A, D–F) and Wt. kenlovei ( Fig. 2V) resembles that of S. rosieae ( Fig. 2U) in its proportions. In A. waitahaorum , the proximal end of the bone is more strongly caudally deflected ( Figs 2W, 12H). The proximal end of the radius of Wt. kenlovei exhibits a notch, which also occurs in S. rosieae , but is less developed in A. waitahaorum .

Complete carpometacarpi are known from CM zfa 34, UC 22078, and CM 2009.99.1 ( M. tuatahi : Fig. 12K–T), and CM 2018.124.4 ( Wt. kenlovei ; Fig. 12Y); the holotype of A. waitahaorum (CM 2020.46.1) includes the proximal portion of the bone ( Fig. 12U). In all specimens, the carpometacarpus exhibits a similar morphology, even though it is craniocaudally somewhat narrower in CM 2018.124.4 than in the other two specimens. The morphology of the bone corresponds well to that of the carpometacarpus of S. rosieae ( Fig. 12W, X), but the os metacarpale alulare is proportionally shorter—which is likely to be a plesiomorphic character—and the facies articularis alularis is more convex. The processus pisiformis is poorly developed but more prominent than in post-Eocene and extant sphenisciforms. The caudal margin of the proximal end forms a concavity ( Fig. 12O), which is obscured by sediment in CM zfa 34. Unlike in an indeterminate species of Anthropornis Wiman, 1905 and Palaeeudyptes gunnari ( Wiman, 1905) from the Late Eocene of Antarctica ( Jadwiszczak 2012), the species of Kairuku from the Late Oligocene of New Zealand ( Ksepka et al. 2012), and crown group sphenisciforms, the ventral surface of the proximal end of the os metacarpale minus of M.tuatahi and A. waitahaorum bears a small tubercle ( Fig. 12O, U).

The os carpi ulnare is present as an isolated element in CM 2009.99.1 ( M. tuatahi ; Fig. 12 CC). Overall, the ossicle resembles the os carpi ulnare of non-sphenisciform birds and the crus longum and crus breve are well differentiated. Its shape is very different from the os carpi ulnare of crown group Sphenisciformes , which is a large, plate-like bone of subtriangular shape that bears no resemblance to the ulnar carpal bone of other birds ( Fig. 12 DD). The os carpi ulnare of P. stonehousei from the Late Paleocene of New Zealand was described as ‘greatly expanded’ and with a ‘disclike shape’ ( Ksepka et al. 2023, p. 443). In the Late Eocene Anthropornis sp. (see Jadwiszczak 2012), the os carpi ulnare has an intermediate morphology between that of M.tuatahi and crown group sphenisciforms, whereas the shape of the os carpi ulnare of the Late Oligocene Kairuku waitaki Ksepka et al., 2012 approaches that of extant penguins ( Ksepka et al. 2012).

The os carpi radiale is likewise present as an isolated element in CM 2009.99.1 ( M. tuatahi ; Fig. 12Z); an os carpi radiale is also associated with UC 22079 (cf. Wt. kenlovei ; Fig. 12 AA). The os carpi radiale of CM 2009.99.1 is distinguished from that of crown group Sphenisciformes ( Fig. 12 BB) in that it exhibits a notch for the tendon of m. ulnometacarpalis ventralis; this notch is absent in UC 22079 and extant penguins ( Mayr 2014).

The phalanx proximalis digiti majoris is preserved in CM 2009.99.1, UC 22078 (both M. tuatahi ; Fig. 12K–N), CM 2020.46.1 ( A. waitahaorum ; Fig. 12V), and CM 2018.124.4 ( Wt. kenlovei ; Fig. 12Y). The bone exhibits a shape more similar to the condition found in non-sphenisciform birds than to corresponding phalanx of extant penguins, which is much more dorsoventrally flattened.

The preservation of the pelvis is best in CM 2009.99.1 ( Fig. 13B, C) and CM 2020.46.1 ( Fig. 13I). The ilium and ischium have a similar shape to those of Waimanu manneringi ( Fig. 13A). As in extant penguins, the ilium is not co-ossified with the synsacrum. The cranial portion of the ala praeacetabularis ilii is less horizontally oriented than in crown group sphenisciforms. As preserved, the ala praeacetabularis ilii appears to be shorter in M. tuatahi (CM 2009.99.1) and A. waitahaorum (CM.2020.46.1) than in W. manneringi , but we consider this to be an artefact of breakage in the two aforementioned specimens. In CM 2009.99.1, the postacetabular portions of the ilium show remarkably different shapes, and on the left side there is a distinct sulcus, which is not present on the right side; the left fossa renalis is dorsally bulging and forms a marked dorsal convexity ( Fig. 13C), the dorsal rim of the left ala postacetabularis ilii projects as a ledge-like dorsal convexity. This morphology is here interpreted as a pathological feature of the specimen. The left ilium of CM 2009.99.1 exhibits a distinct, spine-like spina dorsolateralis ilii. In CM 2020.46.1, the pubis is complete and longer than it is in crown group Sphenisciformes ; a foramen obturatum is absent. The foramen ilioischiadicum is proportionally larger than in crown group Sphenisciformes and exceeds the size of the foramen acetabuli.

The femur of Wt. kenlovei ( Fig. 14E) is proportionally shorter than that of M. tuatahi ( Fig. 14A–D). The proximal margin of the caput femoris, i.e. the lateral margin of the fovea ligament capitis, forms a proximally directed protrusion in M. tuatahi ( Fig. 14D), Wt. kenlovei ( Fig. 14E), and A. waitahaorum ( Fig. 14F), which is absent in S. rosieae ( Fig. 14H). The distal end of the bone is mediolaterally proportionally narrower than in S. rosieae , and the lateral rim of the trochlea fibularis protrudes less strongly laterally. The distal end of the femur of CM 2020.46.1 ( A. waitahaorum ) exhibits a distinct medial projection ( Fig. 14F), which is absent in M. tuatahi but also occurs in S. rosieae ( Mayr et al. 2018: fig. 11A; designated as ‘crista supracondylaris medialis’ in this figure).

A complete tibiotarsus is preserved in CM 2020.46.1 ( A. waitahaorum ; Fig. 14N, O), CM 2010.108.3 ( Fig. 14I), UC 22077 and UC 22078 (both M. tuatahi ; Fig. 14J–M), and UC 22082 (cf. D. primaevus ; Fig. 14P). The cristae cnemiales project proximally beyond the articular facets of the bone to a similar degree as they do in crown group Sphenisciformes . The morphology of the distal end of the bone corresponds to the distal tibiotarsi of W. manneringi and S. rosieae (see Mayr et al. 2018). As in other Paleocene sphenisciforms but unlike in more crownward taxa, the sulcus extensorius ( Fig. 14N) is plesiomorphically located in the medial portion of the bone (centrally situated in post-Paleocene sphenisciforms). In UC 22078 ( M. tuatahi ), there is a small foramen on the lateral side of the distal end. The long fibula is preserved in situ in UC 22078 ( M. tuatahi ), CM 2018.124.4 ( Wt. kenlovei ), and CM 2020.46.1 ( A. waitahaorum ), and reaches to the level of the pons supratendineus of the tibiotarsus.

Waiparadyptes gracilitarsus has an exceptionally slender tarsometatarsus ( Fig. 15 EE–JJ), whereas the tarsometatarsus of UC 22084 (? Kupoupou sp. ) is short and stocky ( Fig. 15 KK–MM); the tarsometatarsi of the other species exhibit similar proportions. In all specimens the hypotarsus forms a deep sulcus for the tendon of m. flexor digitorum longus, which is bordered by a plantarly prominent crista medialis ( Fig. 15G). Lateral to this sulcus is a plantar embossment, which bears a shallow sulcus for the tendon of m. flexor hallucis longus. A low intermediate crest separates the sulcus for m. flexor digitorum longus from a shallow sulcus for the tendon of m. flexor hallucis longus. The foramina vascularia proximalia have similar sizes and are situated on approximately the same level; distal to the lateral foramen vasculare proximale there is a small but well defined tuberositas musculi tibialis cranialis (CM 2020.46.1). In CM 2020.46.1 ( A. waitahaorum ), the fossa metatarsi I is a distinct concavity on the medial surface of the midsection of the shaft ( Fig. 2X); by contrast, a fossa metatarsi I is absent or indistinct in CM zfa 34, CM 2009.99.1, and CM 2010.108.3 ( M. tuatahi ) and UC 22084 (? Kupoupou sp. ). A fossa metatarsi I is also visible in CM 2013.27.1 ( Wp. gracilitarsus ; Fig. 2Y) and UC 22078 ( M. tuatahi ). The lateral margin of the plantar surface of the shaft of CM 2020.46.1 exhibits a ridge-like embossment, which is not present in the other tarsometatarsi. The foramen vasculare distale is large and situated at the end of a short and mediolaterally wide sulcus extensorius (in many post-Eocene sphenisciforms the foramen vasculare distale is reduced); a canalis interosseus distalis is present. The configuration of the trochleae is similar in all taxa of which the tarsometatarsus is known. In distal view, the trochlea metatarsi II exhibits a medial notch ( Fig. 15H, K, DD), which is absent in crown group Sphenisciformes but characterizes the tarsometatarsus of suliform birds ( Smith 2010). Waimanu manneringi ( Fig. 15A–D) is distinguished from other stem group Sphenisciformes from the Waipara Greensand in that the plantar articular surface of the trochlea metatarsi III tapers proximally ( Mayr et al. 2018) and the trochlea metatarsi II is more plantarly deflected.

Two of the M. tuatahi specimens (CM 2009.99.1 and UC 22078) for the first time preserve the os metatarsale I of a Paleocene sphenisciform ( Fig. 16A, B). This ossicle is proportionally larger than and not as flattened as in crown group sphenisciforms. The trochlea metatarsi I is cylindrical, the processus articularis tarsometatarsalis is long and slender.

Pedal phalanges are present in several of the specimens, but often they are in a tangle with other bones, which impedes a close examination. The hind toe is present in CM 2009.99.1 ( Fig. 16A) and CM 2020.46.1 ( Fig. 16C) and is distinctly longer than in crown group Sphenisciformes ( Fig. 16D). Compared to the other phalanges, the first phalanx of the hind toe is considerably narrower. The phalanges are proportionally longer than in more crownward fossil sphenisciforms such as Anthropornis grandis ( Wiman 1905) (seeAcosta Hospitaleche et al. 2019) and Kairuku waitaki (seeKsepka et al. 2012), and the first phalanx of the second toe is the longest phalanx. The tubercula flexoria of the ungual phalanges are only weakly developed.

CM 2009.99.1 ( M. tuatahi ) preserves an unusual plate-like ossicle, which defies a straightforward identification, but may be an intertarsal ossification ( Fig. 17A–C). In its shape, this bone is strikingly similar to the uncinate bone (os uncinatum) of the Rheidae ( Mayr 2022b: fig. 9f), but uncinate bones are unknown from other fossil sphenisciforms and extant penguins, and CM 2009.99.1 otherwise does not preserve cranial remains. Alternatively, and perhaps more likely, the ossicle may be a sesamoid bone of the wing (the bone is too flat to be a patella). Another small ossicle associated with CM 2009.99.1 ( Fig. 17E) likewise defies an identification. This bone may actually be a small patella, with the marked incision on one of its sides possibly being for m. ambiens (see Mayr 2005: fig. 5).

Some pelvis fragments of UC 22078 are associated with several rounded stones, which we interpret as gastroliths (the fine-grained Waipara Greensand is usually devoid of larger stones). One such stone is also attached to a caudal vertebra ( Fig. 17F).