Among extant odontocetes, members of the currently monotypic family Kogiidae (Kogia breviceps and K. sima, pygmy and dwarf sperm whales) are known as small-sized and in many respects enigmatic relatives of the great sperm whale, Physeter macrocephalus. In spite of their wide geographic distribution, the natural history of pygmy and dwarf sperm whales remains largely speculative, and most of our knowledge about them comes from the study of stranded individuals (e.g., Nagorsen, 1985; McAlpine, 2002; Bloodworth and Odell, 2008; Taylor et al., 2012a, b, and references therein). The still scanty fossil record of Kogiidae is mostly represented by isolated skulls and ear bones from Neogene deposits of the Northern Hemisphere (e.g., Barnes, 1973, 1998; Pilleri, 1987; Whitmore, 1994; Bianucci, 1996, 1997; Cigala Fulgosi, 1996; Luo and Marsh, 1996; Bianucci et al., 1998, 2011; Bianucci and Landini, 1999; Lambert, 2008; Whitmore and Kaltenbach, 2008; Velez-Juarbe et al., 2015, 2016), with the significant exception of Scaphokogia, a highly autapomorphic genus from the late Miocene marine deposits of the Pisco Formation exposed along the southern coast of Peru (de Muizon, 1988). The aim of this paper is the description of another kogiid cranium from the Pisco Formation referred to a new genus and species. The specimen was discovered in late Miocene beds exposed at the locality of Aguada de Lomas in the Sacaco Basin. We then undertake a phylogenetic analysis of the relationships of this new taxon and briefly discuss its paleobiological meaning.(ZooBank LSID for Koristocetus gen. nov.: urn:lsid:zoobank.org:act:790FADA0-A6C5-4345-95AE-7CD5CF00149D, record created by Alberto Collareta on 30 October 2017) (ZooBank LSID for Koristocetus pescei sp. nov.: urn:lsid:zoobank.org:act:805D6750-10CF-4FB0-ABDB-30160400A8D6, record created by Alberto Collareta on 30 October 2017)

Exposed along the southern coast of Peru, the Pisco Formation is a late Neogene shallow-marine sedimentary unit, known as one of the most impressive Cenozoic marine Fossil Lagerstätten worldwide. Indeed, the fossil record of vertebrates from Pisco Formation stands out for the exceptional abundance of specimens, high diversity and disparity, and remarkable quality of preservation (e.g., Hoffstetter, 1968; de Muizon, 1981, 1984, 1988, 1993; de Muizon and DeVries, 1985; Pilleri, 1989–1990; de Muizon and McDonald, 1995; de Muizon et al., 1999, 2002, 2003, 2004a, b; McDonald and de Muizon, 2002; de Muizon and Domning, 2002; Brand et al., 2004, 2011; Esperante et al., 2008, 2015; Lambert et al., 2008, 2009, 2010a, b, 2015, 2017; Ehret et al., 2009, 2012; Bianucci et al., 2010, 2016a, b, c; Bisconti, 2012; Lambert and de Muizon, 2013; Amson et al., 2014, 2015a, b, 2016; Collareta et al., 2015, 2017b; Esperante and Poma, 2015; Gariboldi et al., 2015; Stucchi et al., 2015, 2016; Gioncada et al., 2016; Marx and Kohno, 2016; Di Celma et al., 2017; Landini et al., 2017a, b; Marx et al., 2017; Ramassamy et al., 2017). One of the most significant paleontological sites of the Pisco Formation is Aguada de Lomas (Fig. 1; indicative geographic coordinates: 15∘30′ S; 74∘48′ W), about 8 km NNE of the village of Puerto Lomas, within the Neogene Sacaco Basin (de Muizon and DeVries, 1985). According to Lambert and de Muizon (2013), the thick sedimentary succession exposed at Aguada de Lomas includes four vertebrate-bearing levels; namely, ELJ (El Jahuay), AGL (Aguada de Lomas), MTM (Montemar), and SAS (Sud-Sacaco) as defined by de Muizon and DeVries (1985) and de Muizon (1988). The cetacean skull studied here (MUSM 888) was collected about 20 years ago by two of us (Christian de Muizon and Mario Urbina) from the Aguada de Lomas area; it comes from the late Miocene beds of the Pisco Formation located above the AGL vertebrate level of de Muizon and DeVries (1985) (personal observation by Mario Urbina). Therefore, MUSM 888 should be referred to either the MTM or SAS vertebrate levels. These two levels were originally dated to 6.0–5.5 Ma and ca. 5 Ma, respectively, based on the composition of the molluscan assemblages and radiometric dating of tuff layers (de Muizon and DeVries, 1985; de Muizon and Bellon, 1986; de Muizon, 1988). More recently, U–Pb dating on zircon grains from tuff layers and Sr-isotope analyses on marine mollusc shells provided new insights into the age of several fossiliferous exposures of the Pisco Formation in the Sacaco Basin (Ehret et al., 2012). In particular, for the SAS level, these geochemical methods yielded ages of 7 ± 1 Ma (U–Pb) and ca. 6.59 Ma (strontium ratio); such values are supported by Sr-isotope analyses on shells, collected at the site of Sud-Sacaco (West) from two mollusc-rich intervals located above the vertebrate-bearing tuff layer discussed by de Muizon and DeVries (1985), resulting in ages of 5.93 and 5.89 Ma, respectively (Ehret et al., 2012). Furthermore, Sr-isotope analyses yielded an age of ca. 7.30 Ma for the MTM vertebrate level, whereas the AGL vertebrate level was younger than 7.46 Ma, which in turn represents the estimated age of the beds just above the ELJ vertebrate level (Ehret et al., 2012) (see also Lambert and de Muizon (2013), Collareta et al. (2017a), and Lambert et al. (2017) for a discussion about the chronostratigraphic framework of the Pisco Formation cropping out in the Aguada de Lomas valley). Therefore, embracing these recent age estimates, and accounting also for the rather wide confidence intervals associated with the strontium-ratio analyses on shells provided by Ehret et al. (2012), it appears that MUSM 888 should be latest Tortonian or early to mid-Messinian in age.

Cetacea Brisson, 1762

Pelagiceti Uhen, 2008

Neoceti Fordyce and Muizon, 2001

Odontoceti Flower, 1867

Physeteroidea Gray, 1821

Kogiidae Gill, 1871

Kogiinae Gill, 1871

Koristocetus gen. nov.Type and only known species: Koristocetus pescei gen. et sp. nov.Etymology: The generic name originates from the combination of the Greek word koristos, meaning “with helmet” or “crested”, for the presence of a well-developed sagittal facial crest, and the Latin word cetus, meaning “whale”. Gender masculine.Known range: Late Miocene (latest Tortonian or early to mid-Messinian, following the chronostratigraphic framework provided by Ehret et al. (2012); see above for more details) of southern Peru.Differential diagnosis: Same as for the type species until other species are described.Koristocetus pescei sp. nov.Figs. 2–5Holotype and only known specimen: MUSM 888, a partial cranium.Etymology: The specific name honors the memory and heritage of Hugo Pesce Pescetto (1900–1969), Peruvian physician, politician, writer, and “hero of public health in Peru” (Burstein, 2003), who occupied a position of professor at the Universidad Nacional Mayor de San Marcos, Lima.Type locality: The holotype is from the locality of Aguada de Lomas, about 8 km NNE of the village of Puerto Lomas, Arequipa Province, southern coast of Peru (Fig. 1). Indicative geographic coordinates: 15∘30′ S, 74∘48′ W.Type horizon: Pisco Formation, late Miocene (latest Tortonian or early to mid-Messinian; see above).Differential diagnosis: Koristocetus is a small-sized physeteroid, similar in size to Kogia spp. and Nanokogia. It is recognized as a kogiid based on the following features: bizygomatic width smaller than 40 cm, presence of a sagittal facial crest, external nares greatly asymmetric, and nasals absent. Differs from Aprixokogia, Kogia, and Thalassocetus by the presence of a more delicate postorbital process of the frontal. Differs from Aprixokogia and Kogia by the smaller angle between the frontal–maxilla suture line and the horizontal plane. Differs from Aprixokogia, Kogia breviceps, and K. sima by having a supracranial basin that does not significantly extend onto the dorsal face of the rostrum. Differs from Aprixokogia, Nanokogia, and Praekogia by the foramen magnum being proportionally smaller. Differs from Aprixokogia, Scaphokogia, and Thalassocetus by the antorbital notches extending within the supracranial basin and by the presence of a deep and round notch for the posterior process of the tympanic bulla. Differs from Aprixokogia and Thalassocetus by the slit-like geometry of the antorbital notches. Differs from Kogia, Nanokogia, Praekogia, Scaphokogia, and Thalassocetus by the presence of a single, not subdivided right posterior dorsal infraorbital foramen. Differs from Kogia, Nanokogia, and Praekogia by presenting no lateral expansion of the postnarial eminence of the right premaxilla. Differs from Kogia, Nanokogia, and Thalassocetus by the larger, more anteroposteriorly elongated temporal fossa. Differs from Kogia and Nanokogia by presenting well-individualized dental alveoli on the maxilla. Differs from Kogia and Praekogia by the less posteroventrally elongated postorbital process of the frontal. Differs from Kogia by the proportionally smaller lacrimojugal complex. Differs from Kogia breviceps and K. sima by the proportionally longer rostrum, the supracranial basin not being laterally expanded, the more obliquely oriented frontal groove, and the thin and uninflated lateral maxillary crests. Differs from Kogia sima and Scaphokogia by the lack of a left premaxillary foramen. Differs from Kogia sima by the presphenoid which does not extend anteriorly within the mesorostral groove. Differs from Nanokogia by having a wider mesorostral groove in its posterior portion and a maxilla wider than the premaxilla all along the rostrum. Differs from Praekogia by having the left premaxilla that does not reach the sagittal facial crest and the presphenoid largely exposed on the ventral aspect of the skull. Differs from Scaphokogia by the rostrum having a dorsal face that is not semicylindrical, the postnarial eminence that significantly contributes to the sagittal facial crest, the sagittal facial crest that is not significantly dislocated towards the left side of the skull, the supracranial basin that is not spoon-shaped, and the supraoccipital shield that does not dip anteriorly.Description, comparisons, and remarks: MUSM 888 is an incomplete skull, lacking parts of the ventral and posterodorsal portions of the braincase, most of the auditory bones, both mandibles, all the teeth, and the left zygomatic process (Figs. 2–5). The rostrum is almost complete. The facial aspect of the skull is generally well preserved, although some surfaces are moderately abraded and the posterior part of the supracranial basin (including the vertex of the skull) is lacking. In ventral view, the surfaces of the squamosals and exoccipitals are poorly preserved. Locally, the skull is still encrusted by hardened sediment. An endocranial cast is preserved as a concretion.

The skull of Koristocetus is small, distinctly asymmetric, with a rather short rostrum (ratio between length of the rostrum and condylobasal length ca. 0.45). Nevertheless, the skull appears anteroposteriorly elongated and drop-shaped in dorsal view (Fig. 2) with respect to Recent kogiids (ratio between condylobasal length and skull width across the postorbital processes ca. 1.7, versus 1.0–1.2 in extant Kogia spp.), thus recalling Nanokogia in this regard (ratio ca. 1.8 in Nanokogia). Sutures between adjacent cranial bones are fused in MUSM 888, thus indicating that it belonged to a full-grown (i.e., physically adult) individual. Measurements of the cranium of Koristocetus are reported in Table 1.

Premaxilla. In dorsal view, the lateral margins of the premaxillae are slightly concave laterally and weakly convergent towards the apex of the rostrum, which in turn presents a significant constriction at mid-length (Fig. 2). Medial to each premaxilla, the mesorostral groove opens starting from the base of the rostrum. It widens progressively towards the anterior tip of the skull as the medial margins of the premaxillae weakly diverge forward, reaching a transverse width of 23 mm at about two-thirds of the rostrum length (measured from its base). The posterior half of the mesorostral groove is wider than in Nanokogia, recalling instead the condition observed in Aprixokogia and Kogia breviceps. In the posterior half of the rostrum, anterior to the bony nares, the dorsal surface of the right premaxilla is convex transversely, whereas the dorsal surface of the left premaxilla is distinctly concave transversely; both premaxillae progressively flatten towards the apex of the rostrum. No premaxillary foramina are observable on the right and left premaxilla; however, the dorsal surface of the right premaxilla is moderately eroded at the base of the rostrum, and the possibility that a tiny right premaxillary foramen was present cannot be completely ruled out. By contrast, the unambiguous absence of the left premaxillary foramen is a feature shared with Aprixokogia, Kogia breviceps, K. pusilla, and Nanokogia, whereas this foramen is present (although reduced in size) in Kogia sima and Scaphokogia. The external nares are observed at the base of the rostrum (Fig. 2). They are strongly asymmetrical and displaced towards the left side of the skull (a feature shared by all Physeteroidea). The left naris is large and oval-shaped; it is anteroposteriorly elongated and is situated in a similarly oriented funnel-shaped depression (a character also observed in Kogia spp.); its transverse diameter is 24 mm. The right naris is significantly smaller than the left one, having a transverse diameter of 9 mm. The left premaxilla presents a lobe-like posterior end which forms the anterior and lateral walls of the left naris (a feature reminiscent of Aprixokogia, Kogia, and Nanokogia); as also observed in all Kogiidae except presumably Praekogia, the posterior end of the left premaxilla does not take part in building the sagittal facial crest. In turn, the right premaxilla forms the anterior, lateral, and posterior walls of the right naris. The right premaxilla is constricted transversely at the level of the right naris; posterior to this level, the right premaxilla is significantly expanded posteromedially as a stripe of bone that joins the upturned ventral face of the medial margin of the left maxilla to form the sagittal facial crest (Fig. 2). Such a posterior expansion of the right premaxilla (defined by Whitmore and Kaltenbach, 2008, as the postnarial eminence) is present, more or less developed, in all Kogiidae. The postnarial eminence significantly contributes to build the sagittal facial crest in all Kogiidae except Scaphokogia, in which the right premaxilla constitutes only a small fraction of the supracranial basin and sagittal facial crest, being exposed only at the anterior end of the latter feature. In contrast with Kogia, Nanokogia and, to a lesser extent, Praekogia, no lateral expansion of the postnarial eminence posterior to the right naris is observed in Koristocetus (a feature somewhat reminiscent of Aprixokogia). On this portion of the right premaxilla, which dips anterolaterally, a very shallow depression (premaxillary fossa of Barnes, 1973) occupies the area posterior to the right naris (Fig. 2). This premaxillary fossa is oval-shaped and roughly anteroposteriorly elongated; it measures 37 mm in length and 27 mm in width. Several other Kogiidae exhibit a more or less shallow depression located in the right part of the supracranial basin and affecting the right premaxilla (Kogia spp. and Thalassocetus) or both the right premaxilla and maxilla (Praekogia). In extant kogiids, this depression corresponds to the spermaceti chamber hosting the case in which the spermaceti organ is located (Thornton et al., 2015). The posterior half of the facial sagittal crest is not preserved in MUSM 888. In ventral view, the premaxillae are not exposed at the apex of the rostrum (a character shared with Kogia, Nanokogia, and Scaphokogia; for the other fossil kogiids the anterior portion of the rostrum is not known); as such, they bear no teeth.

Maxilla. In dorsal view, the maxilla reaches the anterior end of the skull (Fig. 2). As in Nanokogia, the dorsal exposure of the maxilla tapers transversely towards the apex of the rostrum. However, in contrast to Nanokogia, the maxilla is transversely wider than the premaxilla all along the rostrum. Medial and anteromedial to the antorbital notches, the dorsal surface of the maxilla is transversely convex (almost crest-like on the left side of the skull); then it flattens progressively forwards. Both antorbital notches are very narrow, slit-like (as in Kogia, Nanokogia, and Scaphokogia); they enter the supracranial basin (thus resembling Kogia, Nanokogia, and, at least for the right side, Praekogia). Lateral to the antorbital notch, the lateral maxillary crest is sharp and not transversely inflated on both sides (thus differing from those of Kogia breviceps and K. sima, resembling instead those of Aprixokogia, K. pusilla, and Scaphokogia). The right and left lateral maxillary crests slightly converge anteriorly and form part of the lateral borders of the supracranial basin. When viewed dorsally, the supracranial basin appears roughly trapezoidal (Fig. 2) as in Kogia pusilla, Nanokogia, and Praekogia, whereas it appears as more laterally and posteriorly expanded (thus appearing roughly semicircular) in K. breviceps, K. sima, and Scaphokogia (the outline of the supracranial basin is poorly preserved in Praekogia and Thalassocetus). As in Kogia pusilla and Scaphokogia, the supracranial basin does not significantly extend onto the dorsal surface of the rostrum (Fig. 2); in turn, in Aprixokogia, Kogia breviceps, and K. sima, the anterior part of the supracranial basin is open into the region above the rostrum at both sides of the bony nares. Posterior to the antorbital notches, the maxillae form the major part of the floor of the preserved part of the supracranial basin, whose deepest region is located on the right maxilla, posterolateral to the right external naris (a condition reminiscent of what is observed in Aprixokogia and Nanokogia). In extant kogiids, the posterodorsal surface of the right maxilla hosts a wide fossa that accommodates the vocal chamber, in which the vocal cap and the phonic lips (two soft-tissue structures comprising part of the kogiid echolocation system) are situated (Thornton et al., 2015). The upturned median margin of the left maxilla contributes to build the sagittal facial crest, being exposed on the right side of this crest along its dorsal (subhorizontal) edge. The sagittal facial crest is bent leftwards, thus partially hiding the left part of the supracranial basin in dorsal view (a condition reminiscent of Aprixokogia and Kogia pusilla). Two dorsal infraorbital foramina are present on the left maxilla (Fig. 2). The posterior foramen is located posterolateral to the left external naris, roughly at mid-length between the level of the preorbital process and the level of the postorbital process of the frontal; it continues posterolaterally as a short and shallow groove. The anteriormost left dorsal infraorbital foramen is located at about one-third of the length of the rostrum near the maxilla–premaxilla suture; it is anteroposteriorly elongated, with no groove extension. Three dorsal infraorbital foramina are observed on the right maxilla (Fig. 2). The posterior foramen is situated roughly at the same level of the left posterior maxillary foramen; it is large and is followed posterolaterally for about 25 mm by an elliptical groove. The two anterior right dorsal infraorbital foramina are very small-sized and oval-shaped, located slightly posterior to the level of the left dorsal infraorbital foramen along the maxilla–premaxilla suture. In Kogia, the posterior dorsal infraorbital foramina are commonly subdivided (e.g., duplicated), and a similar condition is observed in Nanokogia, Praekogia, Scaphokogia, and Thalassocetus; in turn, in Aprixokogia, a single posterior dorsal infraorbital foramen is present on each maxilla at the level of the external nares. The palatal exposure of the maxilla is significantly transversely convex in its rostral portion (Fig. 3). Both maxillae bear circular dental alveoli starting from the posterior third of the rostrum. They are shallow, but clearly individualized where the bone surface is nicely preserved; their mean diameter is about 6 mm. Six dental alveoli have been detected on each maxilla; due to poor preservation of the ventral aspect of the anterior half of the rostrum, however, a few more alveoli could have been present. The presence of discrete dental alveoli on the maxillae of MUSM 888 is shared with Aprixokogia and Scaphokogia, whereas just a faint alveolar groove is present in Kogia and Nanokogia (condition unknown in Praekogia and Thalassocetus). We regard the dental alveoli on the maxillae of Koristocetus as presumably bearing functional teeth. The maxillae contact each other medially for about 15 mm posterior to the posteriormost termination of the left alveolar row. Anterior and posterior to this area, the vomer is exposed between the maxillae. No palatine foramina have been detected on the ventral surface of the maxilla near the contact with the palatine. Posterior to the base of the rostrum, each maxilla projects between the palatine and the lacrimal as a lobe-shaped extension that forms the medial walls of the ventral infraorbital foramen (Fig. 3). This foramen is rather large and oval-shaped; it is elongated along a posteromedial to anterolateral axis, posterolateral to the choanae (i.e., anteromedial to the frontal groove). Anterior to the ventral infraorbital foramen, two shallow depressions are present on each maxilla (Fig. 3); similar features have also been observed in Kogia and Nanokogia and could be related to the anteriormost portion of the pterygoid sinus complex. This interpretation would imply a development of the pterygoid sinus beyond the main fossa outlined in Kogia breviceps by Fraser and Purves (1960: plate 16).

Nasal. As in all Kogiidae known to date, the nasals are absent.

Palatine. Because of the loss of the palatal part of the pterygoid, the ventral surface of the left palatine is clearly observable in MUSM 888; in turn, the right palatine is largely missing, but its outline is nonetheless inferable from the observation of the corresponding suture with the overlying right maxilla. The palatines are located posteromedially to the maxillae and anterolaterally to the choanae; they are anterolaterally bent (i.e., roughly kidney-shaped), and the left palatine is distinctly transversely convex (Fig. 3). A cusp-like projection of the maxillae is present between the palatines (a condition reminiscent of that observed in Scaphokogia). The left palatine contributes to the anterolateral border of the left choana.

Lacrimal + Jugal. As in all kogiids and, more generally, in all Physeteroidea, the boundary between the lacrimal and the jugal is not discernible due to complete fusion. In lateral view, the lacrimojugal complex is roughly triangular (Fig. 4a) and smaller than in Kogia. As observed in Kogia, Nanokogia, and Scaphokogia, the lacrimal contacts the maxilla forming a sigmoidal suture line. The posterodorsal corner of the lacrimal is wedged between the frontal and the maxilla. This portion of the lacrimal is posterodorsally projected as in Kogia, Nanokogia, and Praekogia, whereas it is distinctly posteriorly projected in Scaphokogia; it is shorter than in Kogia and Praekogia (where it terminates just anterior to the apex of the orbit) but longer than in Nanokogia and Scaphokogia. The ventralmost tip of the lacrimojugal complex is missing on both sides of the skull. When viewed ventrally, the lacrimojugal complex is posteromedially to anterolaterally elongated and tapers caudally (Fig. 3), thus recalling the condition observed in Nanokogia.

Frontal. The frontal is mostly exposed in lateral and ventral views, the dorsal face of this bone being largely covered by the maxilla. In lateral view, the suture line between the frontal and the maxilla is inclined at about 30∘ with respect to the horizontal plane (Fig. 4a), an angle smaller than in Aprixokogia and Kogia but similar to that observed in Nanokogia. When viewed dorsally, the lateral edge of the supraorbital process is slightly S-shaped, being bent medially in its anterior portion and laterally in its posterior portion (Fig. 2). In lateral view, the roof of the orbit is less concave downwards than in Kogia and Nanokogia. The postorbital process is short and slender; it projects posteroventrolaterally. The postorbital process is less massive than in Aprixokogia, Kogia, and Thalassocetus; furthermore, it is less posteroventrally elongated than in Kogia and Praekogia. The frontal constitutes the upper part of the roof of the large temporal fossa. This fossa is large (Figs. 4a, 5a), being distinctly anteroposteriorly longer than the distance between the preorbital process of the maxilla and the anteriormost wall of the temporal fossa; with a ratio between the greatest length of the temporal fossa and the condylobasal length being ca. 0.30, MUSM 888 has a fossa longer than in Nanokogia (ratio in Nanokogia ca. 0.23), contrasting even more with the anteroposteriorly shortened temporal fossa of Kogia and Thalassocetus and resembling instead that of Aprixokogia and Praekogia. The dorsalmost end of the fossa was most likely located near the lost vertex of the skull. The ventral surface of the frontal is only partly preserved in MUSM 888. It displays a wide frontal groove following anterolaterally the optic canal (Fig. 3). The frontal groove widens and flattens anterolaterally; it is more obliquely oriented with respect to the transverse plane than in extant Kogia, thus resembling Nanokogia and other fossil Kogiidae.

Vomer. The vomer reaches the anterior tip of the skull (Figs. 2, 3, 4a, 5a, c). In dorsal view, it forms the floor of the mesorostral groove (Fig. 2). In ventral view, the vomer is exposed along the rostrum as a narrow slit, running anteriorly between the maxillae from the point where they contact each other anterior to the palatines (Fig. 3). Posterior to this level, the vomer covers ventrally the anterior portion of the presphenoid, thus forming the anteromedial segment of the border of the inner nares; it also covers the lateral sides of this bone projecting posteriorly as a pair of slice-like bony projections. This condition recalls that observed in most Kogiidae (Velez-Juarbe et al., 2015: fig. 12) except for Praekogia, in which the vomer partially drapes the ventral surface of the presphenoid.

Sphenoid. In dorsal view, as in all Kogiidae, the presphenoid forms the septum that separates the external nares medially (Fig. 2). It is transversely thin, and its dorsal edge is sharp. The presphenoid does not extend anteriorly within the mesorostral groove, thus differing from many adult specimens of Kogia sima and some specimens of Kogia breviceps. In ventral view, the presphenoid is partially exposed, being bordered laterally and anteroventrally by the vomer. Posterior to the inner nares, the presphenoid is robust and expands posterolaterally, thus recalling the shape of an upturned T (Fig. 3). The orbitosphenoid is probably present in MUSM 888 as a thin, elongated bony slice lying anterior to the anteromedial margin of the optic canal/frontal groove. The orbitosphenoid runs posteromedially–anterolaterally, from the posterolateral corner of the presphenoid towards the posteromedial termination of the lacrimal (Fig. 3). Among Kogiidae, the orbitosphenoid is visible in Aprixokogia, Kogia, and possibly Scaphokogia; apparently, it has not been observed in Nanokogia and Praekogia. Only shreds of the posteriormost portion of the alisphenoid are preserved on both sides of the skull. The basisphenoid is not preserved in MUSM 888.

Pterygoid. Most of the pterygoid is lost on both sides of the skull and only a shred of that bone is tentatively detectable in MUSM 888. It consists of a thin slice of bone located posterolateral to the right bony naris and lateral to the posterior projection of the vomer; it also contacts the frontal (anteriorly), the orbitosphenoid (posterolaterally), and the presphenoid (posteriorly).

Parietal. The parietal is not clearly identifiable but still present in MUSM 888. It likely constitutes most of the posterodorsal portion of the medial wall of the temporal fossa (see description of the frontal bone above).

Squamosal. The zygomatic process is incompletely preserved on the right side of the skull (Figs. 2, 3, 4b, 5a, c). Its proximal portion is triangular in lateral view, and its transverse section is roughly triangular. The supramastoid crest, which represents the anterolateral continuation of the temporal crest, appears as well defined and rather salient. The mandibular fossa is shallow, anteroposteriorly elongated, and facing anteroventrally (Fig. 3). Posteromedial to the mandibular fossa the tympanosquamosal recess is a shallow fossa in which the middle sinus probably resided (Fraser and Purves, 1960: plate 16; see also Barnes, 1973); this depression is observed in Nanokogia and Praekogia and is seemingly less well defined in Kogia. In lateral view, the postglenoid process is robust and triangular in outline (Figs. 4b, 5a), with a long, anteroventrally directed posterior wall; it is followed medially by the sharp, spine-like anterior meatal crest. This crest is gently sigmoidal and anterolaterally–posteromedially elongated in ventral view, whereas it is generally more transversely oriented in other Kogiidae for which this feature is known. A narrow groove is observed posterior and parallel to the anterior meatal crest; it is here interpreted as the vestigial external auditory meatus (Fig. 3). Posterior and medial to this level the ventral surface of the squamosal is poorly preserved, and features such as the falciform process and the sutures between the squamosal, the exoccipital, the parietal, and the basioccipital cannot be defined. The posteroventral portion of the squamosal and the anteroventral portion of the paroccipital process of the exoccipital (see below) are shaped as a wide and deep round notch (Figs. 4b, 5a), as in Kogia breviceps, K. sima, Nanokogia, and Praekogia. A similar feature is present but distinctly shallower in Aprixokogia and Scaphokogia, whereas it is not developed in Thalassocetus; as in extant Kogiidae, this notch hosts the inflated posterior process of the tympanic bulla (see below).

Occipital. The basioccipital crests are known only for their posterior portions; they are salient and markedly posterolaterally diverging (Fig. 3), as observed in other Kogiidae. The foramen magnum is roughly oval-shaped (Fig. 5b), higher than wider, and proportionally smaller than in Aprixokogia, Nanokogia, and Praekogia. The occipital shield is only partially preserved: as observed in most Kogiidae, it forms an angle of about 60–70∘ with the horizontal plane. The occipital condyles are dorsoventrally elongated and separated ventrally by a shallow furrow. Dorsolateral to the condyles, the shallow dorsal condyloid fossae appear to be present. Lateral and ventrolateral to the condyles, the exoccipitals are broad and flare ventrolaterally to form the paroccipital processes. A reduced jugular notch separates the paraoccipital process of the exoccipital from the basioccipital crest. The preservation of an endocranial cast (Figs. 2, 3, 4a) allows for the reconstruction of the general shape of the braincase of MUSM 888. The two cerebral hemispheres appear as separated dorsally by a wide but shallow longitudinal sulcus, strongly reminiscent of the interhemispheric cleft.

Tympanic bulla. Part of the posterior process of the left tympanic bulla is preserved in MUSM 888 (Figs. 3, 4a). It is broken and its original shape cannot be ascertained. As in extant Kogiidae, the posterior process of the tympanic bulla is enlarged and occupies the wide and deep notch present on the posteroventral portion of the squamosal and the anteroventral portion of the paroccipital process of the exoccipital by firmly attaching to the neurocranium.

Our phylogenetic analysis resulted in six most parsimonious trees having a consistency index (CI) of 0.5379 and a retention index (RI) of 0.7185. The strict-consensus tree is shown in Fig. 6. The base of Physeteroidea is marked by a polytomy involving (i) Eudelphis, (ii) a clade embracing the macroraptorial physeteroid genera (i.e., Acrophyseter, Brygmophyseter, Livyatan, and Zygophyseter), and (iii) a clade formed by “Aulophyseter” rionegrensis plus the crown Physeteroidea. Differing form previous phylogenetic analyses of physeteroids (e.g., Lambert et al., 2010, 2017; Boersma and Pyenson, 2015; Velez-Juarbe et al., 2015), our consensus tree recovers all the middle to late Miocene macroraptorial sperm whales (as defined by Lambert et al., 2017) as a monophyletic group, with Livyatan as a sister group of Acrophyseter spp. Crown physeteroids consist of two clades corresponding to the families Physeteridae and Kogiidae. In agreement with Lambert et al. (2017), but differing from other recent analyses (Boersma and Pyenson, 2015; Velez-Juarbe et al., 2015), Orycterocetus and Placoziphius are here recognized as Physeteridae.

All the phylogenetic analyses of Physeteroidea including all extant and fossil kogiids published to date (Bianucci and Landini, 2006; Lambert, 2008; Lambert et al., 2008, 2010, 2017; Boersma and Pyenson, 2015; Velez-Juarbe et al., 2015) agreed on the existence of three groups among Kogiidae: (i) a paraphyletic group including early branching forms such as Aprixokogia and Thalassocetus; (ii) a lineage (subfamily Scaphokogiinae) consisting of the sole bizarre genus Scaphokogia; (iii) a clade (subfamily Kogiinae), sister group to Scaphokogia, whose members are Kogia spp., Nanokogia, and Praekogia. Our cladistic analysis confirms this phylogenetic structure for Kogiidae but identifies a sister group to Scaphokogia in MUSM 3291 and MUSM 3405, which are thus recognized as representatives of a new form of Scaphokogiinae. Although the osteo-anatomical differences that separate the four genera of Kogiinae known to date are rather minor, our cladistic analysis is able to fully resolve the phylogenetic relationships among Kogia, Nanokogia, Praekogia, and Koristocetus. The latter genus is here recognized as the earliest branching member of the subfamily Kogiinae, whereas the late Tortonian Nanokogia from Panama, the Messinian Praekogia from Baja California, Mexico, and the late Pliocene to Recent, globally distributed Kogia represent subsequent branches.

Therefore, the late Miocene genus Koristocetus represents the first representative of the subfamily Kogiinae from the Southern Hemisphere, and it retains a series of archaic traits, including the presence of well-individualized dental alveoli on the upper jaws. Extant dwarf and pygmy sperm whales primarily forage on deep-water cephalopods and, secondarily, on fish and crustaceans; these odontocetes exhibit a strong suction feeding specialization that relies on a peculiar craniomandibular architecture (featuring the proportionally shortest rostrum among Recent cetaceans) and on a highly derived hyoid apparatus (e.g., Caldwell and Caldwell, 1989; McAlpine, 2002; Bloodworth and Marshall, 2005; Werth, 2006; Bloodworth and Odell, 2008; Hocking et al., 2017). As reported above, Koristocetus pescei mainly differs from extant Kogia spp. by displaying a longer rostrum, a larger and more anteroposteriorly elongated temporal fossa, and a presumably functional maxillary dentition. Coupled with a rather robust and not ventrally deflected zygomatic process, these characters suggest that Koristocetus retained some degree of raptorial feeding ability, contrasting with the obvious suction feeding specialization observed in extant dwarf and pygmy sperm whales. The discovery of new specimens of Koristocetus, and especially of remains of the mandibles and hyoid apparatus, will hopefully contribute to clarify the ecology of this fossil kogiid.

During the late Miocene Koristocetus shared the southern coast of present-day Peru with members of the genus Scaphokogia, whose unique pachyostotic and downwards-deflected semicylindrical rostrum evokes a peculiar feeding/foraging specialization still to be understood. Whereas at least five skulls of Scaphokogia – either collected and deposited in museum collections or still in the field – are known from the late Miocene coastal marine deposits exposed at Aguada de Lomas (Alberto Collareta, personal observation), a single specimen of Koristocetus has been found in the same deposits (and, more generally, from the whole Pisco Formation) to date. Such an uneven distribution could reflect an intense ecospace use of the coastal area of Aguada de Lomas by members of the genus Scaphokogia; in turn, Koristocetus could have exhibited more open-ocean habits, thus recalling the habitat preferences of extant Kogia in this aspect (McAlpine, 2002). However, since skulls of Scaphokogia are much more robust and possibly less prone to degradation than those of other small-sized odontocetes such as Koristocetus, a preservation bias cannot be ruled out.

In conclusion, Koristocetus evokes a long history of high diversity, morphological disparity, and sympatric habits in fossil kogiids (as already highlighted elsewhere; Velez-Juarbe et al., 2015, 2016), thus suggesting that our comprehension of the evolutionary history of pygmy and dwarf sperm whales is still far from being exhaustive, especially considering the broad geographical distribution of extant species.