A New Fossil Species Supports an Early Origin for Toothed Whale Echolocation

LETTER doi:10.1038/nature13086

A new fossil species supports an early origin for toothed whale echolocation

Jonathan H. Geisler1, Matthew W. Colbert2 & James L. Carew3

Odontocetes (toothed whales, dolphins and porpoises) hunt and navi- of the facial muscles that modulate echolocation calls, which in turn gate through dark and turbid aquatic environments using echolo- led to marked, convergent changes in skull shape in the ancestors of cation; a key adaptation that relies on the same principles as sonar1. Cotylocara, and in the lineage leading to extant odontocetes. Among echolocating vertebrates, odontocetes are unique in pro- Cetacea ducing high-frequency vocalizations at the phonic lips, a constric- Odontoceti tion in the nasal passages just beneath the blowhole, and then using 2,3 Xenorophidae air sinuses and the melon to modulate their transmission .Allextant Cotylocara macei gen. et sp. nov. odontocetes seem to echolocate2,4; however, exactly when and how this complex behaviour—and its underlying anatomy—evolved is Etymology. Cotylo, for cavity, and cara, for head, both Greek, in refer- largely unknown. Here we report an odontocete fossil, Oligocene in ence to bilateral pits on the frontals. Species name recognizes Mace age (approximately 28 Myr ago), from South Carolina (Cotylocara Brown for his contributions to the founding of a Natural History Museum macei, gen. et sp. nov.) that has several features suggestive of echo- at the College of Charleston, South Carolina, USA. location: a dense, thick and downturned rostrum; air sac fossae; cranial Holotype. CCNHM-101. Nearly complete skull, partial dentaries, three asymmetry; and exceptionally broad maxillae. Our phylogenetic ana- cervical vertebrae (C2, C3 or C4, C6), and portions of at least seven ribs lysis places Cotylocara in a basal clade of odontocetes, leading us to found in close association and presumed to represent a single individual infer that a rudimentary form of echolocation evolved in the early (Figs 1 and 2 and Extended Data Figs 1, 2 and 3). Oligocene, shortly after odontocetes diverged from the ancestors of Locality and age. Drainageditch in the CollegePark Subdivision, Berke- filter-feeding whales (mysticetes). This was followed by enlargement ley County, South Carolina. Approximate coordinates: 33u 19 5099 N,

Dense Mx pf ip a c e La Porous Epoxy Px

Px Fr

Px dta f La rb rb ep

Mx Fr

rb dtp rb Mx g Px Section G an an Pl Mxrb rb Section F Px pop pop La Pl Na La Px Section E Mx Bs Mx zy pf Pt Pt gf Bo pgp Pa oc b d pap

Figure 1 | Holotype skull of Cotylocara macei (CCNHM-101), including anteriormost double-rooted tooth; dtp, posteriormost double-rooted tooth; cross-sections and bone density. a, b, CT model and photograph of skull in ep, embrasure pit; Fr, frontal; gf, glenoid fossa; ip, interparietal; La, lacrimal; dorsal view. c, d, Same as a, b, but in ventral view. e–g, CT cross-sections of Mx, maxilla; Na, nasal; oc, occipital; Pa, parietal; pap, paroccipital process; skull; locations indicated in b. Colours in a and c reflect the relative density of pf, postnarial fossa; pgp, postglenoid process; Pl, palatine; pop, postorbital bone, whereas in e–g, each bone is a separate colour as follows: blue, frontal; process; Pt, pterygoid; Px, premaxilla; rb, rostral basin; zy, zygomatic process. green, lacrimal; purple, palatine; red, premaxilla; white, interparietal; yellow, Scale bar is 5 cm. maxilla. an, antorbital notch; Bo, basioccipital; Bs, basisphenoid; dta,

1Department of Anatomy, New York Institute of Technology College of Osteopathic Medicine, Northern Boulevard, Old Westbury, New York 11568, USA. 2Jackson School of Geosciences, The University of Texas at Austin, Austin, Texas 78712, USA. 3Department of Geology and Environmental Geosciences, Natural History Museum, College of Charleston, Charleston, South Carolina 29424, USA.

or flat premaxillary sac ‘fossae’ have been identified in other Oligocene ec en odontocetes10,11, C. macei represents the first strong osteological evi- dc xn dence for air sinuses anterior to the bony nares in basal odontocetes. wf tf ab Similarly, we interpret the deep postnarial fossae as being excavated by pf air sinuses, a hypothesis supported by the thin median septum that or zy resembles the bony lamina in the pterygoid sinuses of many odonto- rb cete skulls12. We tentatively suggest that these fossae accommodated a c diverticulum of the inferior vestibule, possibly a homologue to the caudal sac of phocoenids13. In addition, the rostral basins and the postnarial fossae are formed by low-density bone, as in the premaxillary sac fossa of Tursiops (Fig. 1 and Extended Data Fig. 6). Cotylocara resembles extantodontocetes in having an expanded max- d e illa that covers most of the supraorbital process of the frontal, and this So expansion may be related to the fact that itserves as the proximal attach- 10,14 pop ment for the maxillonasolabialis muscle and homologues . In odontocetes the maxillonasolabialis is hypertrophied, split into several different muscles, and inserts onto the melon, nasal passages, and nasal diverticula2,9,13,15. If the site of attachment for the maxillonasolabialis is fm conserved, then a larger muscle would be coupled with a larger maxilla. zy It is generally assumed that some of the homologues of the maxillonasolabialis in odontocetes play key roles in the production of sound at Figure 2 | Holotype skull and dentition of Cotylocara macei (CCNHM- the phonic lips13, particularly because the nasal muscles in mysticetes, 101). a, b, Last maxillary tooth from the left side of the skull in labial and lingual which do not echolocate, are considerably smaller and less complex16. views. c, Skull in lateral view, downturned to indicate angle between rostrum Most extant odontocetes have asymmetric heads where parts of the and basicranial stem (marked by dashed line). d, CT-generated model of skull ‘median’ plane are shifted to the left side and some structures on the in anterior view with perspective showing counterclockwise torsion between 9 the face and braincase. e, Posterior view of skull. dc, denticles; ec, ectocingulum; right side are much larger than those on the left . Some researchers have en, entocingulum; fm, foramen magnum; or, orbit; pf, premaxillary fossa; suggested that this asymmetry is involved in the production of high- pop, postorbital process; rb, rostral basin; So, supraoccipital; tf, temporal fossa; frequency vocalizations9,15; for example, one nasal passage might be wf, wear facet; xn, external nares; zy, zygomatic process. Scale bar for a and b is primarily for vocalizations whereas the other only for respiration17.How- 5 mm, scale bars for c and e are 5 cm. ever, some studies suggest that both the right and left phonic lips generate sounds3, and others suggest that cranial asymmetry might allow for large 80u 59 51.499 W. Bed 2 of the Chandler Bridge Formation, late Oligo- prey to be swallowed whole18. Cotylocara represents one of the earliest cene in age5. and most definitive examples of odontocete cranial asymmetry. The left Diagnosis. The species differs from all known cetaceans (whales, dol- rostral basin is 140% wider and longer than its counterpart on the right phins, porpoises) in that it has deep, postnarial fossae on frontals that (Fig. 1b); the premaxillae are asymmetric where they border the ros- are separated by a median septum of the interparietal (Fig. 1b, e) and a tral basins and external bony nares; and portions of the internasal and vertical supraoccipital whose anterior face is overlapped by parietals interparietal sutures are to the left of the median plane. A recent study19 (Figs 1b and 2). Like the xenorophids Albertocetus meffordorum and identified directional asymmetry in the skulls of archaeocetes, the para- Xenorophus sloani, the lateral portion of the frontal is covered by the phyletic stem group of crown cetaceans, and speculated that this asym- ascending process of the lacrimal, the frontal has a window on its ven- metry was an adaptation for using high-frequency sound to locate prey, tral side that exposes strips of lacrimal and maxilla (Extended Data Fig. 4), as has been documented in owls20. In archaeocetes the internasal suture the premaxilla underlies the ascending process of the maxilla (Fig. 1e), is shifted towards the right side, and in anterior view, the rostrum is bilateral rostral basins are present, and the petrosal (Extended Data Fig. 2) twisted counterclockwise19, although clockwise rotation occursin a skull has an elongate lateral tuberosity that articulates with the squamosal6 of Cynthiacetus21. Interestingly, in Cotylocara the rostrum is also rotated (not preserved in Xenorophus). However, Cotylocara differs from these counterclockwise (Fig. 2b), but some sutures are shifted to the left, not taxa, as well as Archaeodelphis patrius, in having premaxillae that over- the right. The rostral torsion is probably not the result of post-mortem hang maxillae (Fig. 1f); maxillae overhanging the squamosal fossae; deformation because the Chandler Bridge Formation is unconsolidated thick asymmetrical nasals with transversely compressed crests at their and flat-lying5. Whether the unique pattern of asymmetry in Cotylo- anterolateral corners; and anterolaterally projecting, dorsoventrally deep cara is indicative of high-frequency sound production, sound recep- zygomatic processes of the squamosals6 (the last two characters are not tion, or the size of its prey can only be determined once the basis for preserved in Xenorophus). asymmetry in extant odontocetes is better understood. In addition to the phonic lips, odontocete echolocation is associated Computed tomography (CT) scans reveal that the premaxillae of with a suite of soft-tissue facial structures, including a fat-filled body C. macei are dense, including where they encircle the external bony called the melon, which is thought to focus vocalizations7; a series of nares and overhang the maxillae (Fig. 1a). A similarly dense and over- pneumatic diverticulae that allow for near continuous vocalizations while hanging premaxilla occurs in Cuvier’s beaked whale (Ziphius cavirostris), diving and may act as acoustic reflectors (for example, premaxillary and finite element modelling supports the hypothesis that the dense sinus, inferior vestibule)8; and a complex of muscles that change the premaxilla acts as an acoustic reflector in that taxon22. Likewise, the size and/or shape of the melon, the diverticulae, and nasal passages9. density contrast in C. macei could reflect much of the sound forward, Although fossils of these soft tissues have not been found, we contend particularly because the rostrum is deflected 20u downwards relative that some aspects of these tissues can be inferred from fossil skulls. For to the basicranial stem (Fig. 2 and Extended Data Fig. 1). The nasals example, C. macei has a pair of large, rostral basins immediately ven- and dorsal portions of the maxillae and lacrimals are also osteosclero- trolateral to the premaxillary sac fossae that breach the lateral walls of tic, raising the possibility that the large interorbital shield could be an the infraorbital canals (Extended Data Fig. 4), similar to exposure of acoustic reflector. the mandibular division of the trigeminal nerve by the pterygoid sinus Taken together, the cranial features described above make a compel- in extant cetaceans. We suggest that expansion of the premaxillary air ling case that Cotylocara could echolocate, and that has important impli- sinus excavated the rostral basin (Extended Data Fig. 5). Although shallow cations for the origin of this complex behaviour. Our phylogenetic analysis

384 | NATURE | VOL 508 | 17 APRIL 2014 ©2014 Macmillan Publishers Limited. All rights reserved LETTER RESEARCH hiacetus p Inia Auroracetus Albireo Atocetus Agorophius Xi Ziphius Zarhachis Archaeodelphis Albertocetus ChM PV2761 ChM PV4802 Squalodon Prosqualodon Squaloziphius Orycterocetus Physeter Kogia Lipotes Par. wilsoni Par. sternbergi Kentriodon Meherrinia Pliopontos Pontoporia Brachydelphis Protophocoena Stenasodelphis Monodontidae Phocoena Phocoenoides Leucopleurus Orcinus Tursiops Delphinus Orcaella Pseudorca Globicephala Grampus Berardius Mesoplodon Tasmacetus Ninoziphius Ischyrorhynchus Platanista Notocetus Patriocetus ChM PV4961 ChM PV2764 Waipatia Xeno. sp. Xeno. sloani ChM PV4178 ChM PV5852 Simocetus ChM PV4746 ChM PV4834 ChM PV5711 GSM 1098 ChM PV2758 Cotylocara Aetiocetus Janjucetus ChM PV5720 ChM PV4745 Mammalodon Chonecetus Eomysticetus Micromysticetus Diorocetus Pelocetus Balaenidae Caperea Eschrichtius Parabalaenoptera Megaptera Balaenoptera Zygorhiza Bos Sus Hippopotamidae Georgiacetus

Figure 3 | Phylogeny supported by the present study, with evolution of from ref. 10 (Smithsonian Institution Scholarly Press) and ref. 29 (San Diego echolocation and skull shape. Skulls are not to scale; maxilla is in grey. Society of Natural History). C, Cetacea; CCNHM, College of Charleston Solid circles indicate convergent evolution of the roof to the temporal fossa Natural History Museum, South Carolina; ChM PV, The Charleston Museum, (see Supplementary Information). Aspects of the skull of Archaeodelphis were Vertebrate Paleontology Collection, South Carolina; GSM, Georgia Southern reconstructed, braincase of Xenorophus sloani based on CCNHM-168, and Museum; M, Mysticeti; O, Odontoceti; Par., Parapontoporia; X, Xenorophidae; three skulls redrawn from published figures10,29,30, adapted with permission Xeno., Xenorophus; W, crown Odontoceti. of a modified supermatrix of morphological and molecular data23–25 produced at the phonic lips. Although the inner ear labyrinth of Coty- resulted in one most parsimonious tree (Fig. 3). Cotylocara is nested in locara is not well preserved, anatomical studies on the inner ear of other Xenorophidae, the first clade to branch from the odontocete stem, and xenorophids should provide a robust test for an early evolution of odo- ismost closely related toAlbertocetusmeffordorumamong the described ntocete echolocation. taxa. If Cotylocara could echolocate, it is most parsimonious to infer that the ability to produce sounds at the phonic lips evolved in the common METHODS SUMMARY ancestor of all known odontocetes, shortly after they diverged from Our phylogeny (Fig. 3) is the single most parsimonious tree (41155.88 steps in mysticetes (that is, 34 to 30 Myr ago26). Our results are consistent with length) for a supermatrix25 (311 morphological and 60,851 molecular characters) earlier hypotheses, based on less evidence, that the early Oligocene odo- that was modified by adding Cotylocara, Albertocetus meffordorum, an undescribed ntocetes Simocetus rayi10 and two unnamed species11 could echolocate. xenorophid (GSM 1098), and a single morphological character (see Supplementary 28 6,27 Information). The tree was obtained using the computer application TNT , with Like other studies , we found Archaeodelphis patrius to be a basal 24,25 xenorophid, which is significant because the frontals and maxillae in default search parameters under a ‘New Technology Search’. As in previous studies , between-character scaling was achieved by down-weighting ordered, multistate, mor- that taxon do not roof over the temporal fossa. When xenorophids are phological characters so that they have the same minimum length as a binary character. arranged according to our phylogenetic tree, there is a trend towards The density model of the skull was based on CT data from the Medical Univer- posterior migration of the frontals, maxillae and premaxillae. This trend sity of South Carolina (cubic voxel size is 0.584 mm per side) and was generated in occurred in parallel along the stem of crown Odontoceti (Fig. 3, Extended Avizo v.7.1, with colours assigned according to grayscale values. Reconstructed Data Fig. 7 and Supplementary Information). As a result, Cotylocara portions were digitally isolated before applying the density-coded colour palette. looks superficially more like some crown odontocetes (for example, Current density should be a reasonable proxy for original density because exposed Mesoplodon) than it resembles basal xenorophids. Similarly, cranial breaks show no permineralization and the Chandler Bridge Formation is unlithified. asymmetry is likely to have evolved twice because basal xenorophids, Online Content Any additional Methods, Extended Data display items and Source like Archaeodelphis and Xenorophus, lack it. An extreme interpreta- Data are available in the online version of the paper; references unique to these tion of these patterns would be that echolocation evolved separately in sections appear only in the online paper. xenorophids, but we prefer the simpler explanation that a rudimentary form of echolocation was present in the common ancestor of xenoro- Received 7 October 2013; accepted 29 January 2014. phids and other odontocetes, and this shared behaviour facilitated the Published online 12 March 2014. convergent cranial evolution in both clades. In support of this view, we 1. Au, W. W. L. The Sonar of Dolphins (Springer-Verlag, 1993). corroborated that the lateral expansion of the maxillae over the orbits 2. Cranford, T. W., Amundin, M. & Norris, K. S. Functional morphology and homology is a synapomorphy of Odontoceti14,15,26,27, suggesting that hypertrophy in the odontocete nasal complex: implications for sound generation. J. Morphol. of the maxillonasolabialis and echolocation evolved before cranial con- 228, 223–285 (1996). 3. Cranford, T. W. et al. Observation and analysis of sonar signal generation in the vergence. Our hypothesis predicts that the inner ear of basal odonto- bottlenose dolphin (Tursiops truncatus): evidence for two sonar sources. J. Exp. Mar. cetes should be specialized to hear the echoes of high-frequency sounds Biol. Ecol. 407, 81–96 (2011).

4. Ketten, D. R. in Marine Mammal Sensory Systems (eds Thomas, J.A., Kastelein, R.A. & 22. Cranford, T. W., Krysl, P. & Hildebrand, J. A. Acoustic pathways revealed: simulated Supin, A.Y.) 53–75 (Plenum Press, 1992). sound transmission and reception in Cuvier’s beaked whale (Ziphius cavirostris). 5. Sanders, A. E., Weems, R. E., Lemon, E. M. & Jr.. Chandler Bridge Formation—a new Bioinspir. Biomim. 3, 016001 (2008). Oligocene stratigraphic unit in the lower coastal plain of South Carolina. US Geol. 23. McGowen, M. R., Spaulding, M. & Gatesy, J. Divergence date estimation and a Surv. Bull. 1529, 105–124 (1982). comprehensive molecular tree of extant cetaceans. Mol. Phylogenet. Evol. 53, 6. Uhen, M. D. A new Xenorophus-like odontocete cetacean from the Oligocene of 891–906 (2009). North Carolina and a discussion of the basal odontocete radiation. J. Syst. 24. Geisler, J. H., McGowen, M. R., Yang, G. & Gatesy, J. A supermatrix of genomic, Palaeontology 6, 433–452 (2008). morphological, and paleontological data from crown Cetacea. BMC Evol. Biol. 11, 7. McKenna, M. F., Cranford, T. W., Berta, A. & Pyenson, N. Morphology of the 112 (2011). odontocete melon and its implications for acoustic function. Mar. Mamm. Sci. 28, 25. Geisler, J. H., Godfrey, S. J. & Lambert, O. A new genus and species of late Miocene 690–713 (2012). inioid (Cetacea, Odontoceti) from the Meherrin River, North Carolina, U.S.A. J. 8. Reidenberg, J. S. & Laitman, J. T. Sisters of the sinuses: cetacean air sacs. Anat. Rec Vertebr. Paleontol. 32, 198–211 (2012). (Hoboken) 291, 1389–1396 (2008). 26. Gatesy, J. et al. A phylogenetic blueprint for a modern whale. Mol. Phylogenet. Evol. 9. Mead, J. G. Anatomy of the external nasal passages and facial complex in the 66, 479–506 (2013). Delphinidae (Mammalia: Cetacea). Smithson. Contrib. Zool. 207, 1–35 (1975). 27. Geisler, J. H. & Sanders, A. E. Morphological evidence for the phylogeny of Cetacea. 10. Fordyce, R. E. in Cenozoic Mammals of Land and Sea: Tributes to the Career of Clayton J. Mamm. Evol. 10, 23–129 (2003). E. Ray (ed. Emry, R. J.) 185–222 (Smithsonian Institution Scholarly Press, 2002). 28. Goloboff, P. A., Farris, J. S. & Nixon, K. C. TNT, a free program for phylogenetic 11. Barnes, L. G., Goedert, J. L. & Furusawa, H. in Western Association of Vertebrate analysis. Cladistics 24, 774–786 (2008). Paleontologists with Mesa Southwest Museum and Southwest Paleontological 29. Fordyce, R. E. in Contributions in Marine Mammal Paleontology Honoring Frank Society: First Meeting of the New Millenium Vol. 8 (eds McCord, R. D. & Boaz, D.) Whitmore Jr (eds Berta, A. & Deme´re´, T. A.) 147–176 (San Diego Society of Natural 91–100 (Southwest Paleontological Society, 2001). History, 1994). 12. Fraser, F. C. & Purves, P. E. Hearing in cetaceans, evolution of the accessory air sacs 30. Dubrovo, I. A. & Sanders, A. E. A new species of Patriocetus (Mammalia, Cetacea) and the structure and function of the outer and middle ear in recent cetaceans. from the late Oligocene of Kazakhstan. J. Vertebr. Paleontol. 20, 577–590 (2000). Bull. Brit. Mus. Nat. Hist. Zool. 7, 1–140 (1960). 13. Huggenberger, S., Rauschmann, M. A., Vogl, T. J. & Oelschla¨ger, H. H. Functional Supplementary Information is available in the online version of the paper. morphology of the nasal complex in the harbor porpoise (Phocoena phocoena L.). Anat. Rec. (Hoboken) 292, 902–920 (2009). Acknowledgements This research was supported by funds from the New York Institute 14. Fordyce, E. & de Muizon, C. in Secondary Adaptation of Tetrapods to Life in Water of Technology, College of Charleston, and M. Brown. During this project we benefitted (eds Mazin, J.-M. & de Buffreńil, V.) 169–234 (Verlag Dr. Friedrich Pfeil, 2001). from discussions with B. Beatty, P. Gingerich, W. Hillenius, M. Mihlbachler, A. Sanders 15. Heyning, J. E. Comparative facial anatomy of beaked whales (Ziphiidae) and a and N. Solounias. We thank A. Sanders and The Charleston Museum for access to systematic revision among the families of extant Odontoceti. Nat. Hist. Mus. Los specimens under their care. B. Dulguun drew the skulls in Fig. 3. We acknowledge the Angeles Cty. Contrib. Sci. 405, 1–64 (1989). Deptartment of Radiology, Medical University of South Carolina, and T. Holden in 16. Carte, A. & Macalister, A. On the anatomy of Balaenoptera rostrata. Phil. Trans. R. Soc. particular, for conducting CT scans of Cotylocara. We thank T. Rowe for access to the Lond. 158, 201–261 (1868). Tursiops scan data, which were scanned using his NSF digital libraries initiative grant 17. Madsen, P. T., Wisniewska, D. & Beedholm, K. Single source sound production and (IIS-987-4781). dynamic beam formation in echolocating harbor porpoises (Phocoena phocoena). Author Contributions J.H.G. designed the research plan and conducted the J. Exp. Biol. 213, 3105–3110 (2010). phylogenetic analysis. M.W.C. generated three-dimensional models from CT data and 18. Macleod, C. D. et al. Breaking symmetry: the marine environment, prey size, and conducted related analyses on this data. J.H.G. wrote the paper with discussion from the evolution of asymmetry in cetacean skulls. Anat. Rec. (Hoboken) 290, 539–545 J.L.C and M.W.C. (2007). 19. Fahlke, J. M., Gingerich, P. D., Welsh, R. C. & Wood, A. R. Cranial asymmetry in Author Information Supermatrix used for phylogenetic analysis deposited at Eocene archaeocete whales and the evolution of directional hearing in water. Proc. Morphobank (http://www.morphobank.org/) project P888. The new taxa Natl Acad. Sci. USA 108, 14545–14548 (2011). (new genus and new species) have been registered with Zoobank (LSID: 20. Norberg, R. A. in Ecology and Conservation of Owls (eds Newton, I., Kavanagh, R., urn:lsid:zoobank.org:pub:415D35BC-3306-4361-B34D-B7702649B425). Olsen, J. & Taylor, I.) 329–342 (Csiro Publishing, 2002). Reprints and permissions information is available at www.nature.com/reprints. The 21. Martıńez-Caćeres, M. & de Muizon, C. A new basilosaurid (Cetacea, Pelagiceti) authors declare no competing financial interests. Readers are welcome to comment on from the late Eocene to early Oligocene Otuma Formation of Peru. C. R. Palevol 10, the online version of the paper. Correspondence and requests for materials should be 517–526 (2011). addressed to J.H.G. ([email protected]).

METHODS initial taxon addition sequences, trees then subjected to sectorial searches and three rounds of tree fusing, and the process repeated until the most parsimonious trees Our phylogenetic hypothesis was based on a modified version of a supermatrix were found 1,000 times. The memory was setso thatup to 1,000 mostparsimonious compiled previously25. This supermatrix (311 morphological and 60,851 molecular characters) was enlarged by adding one morphological character (character 305, trees could be saved, although the analysis did not approach that limit. We found a see Supplementary Information) and three taxa: Cotylocara macei (based on CCNHM- single most parsimonious tree (Fig. 3) that is 41,155.88 steps in length. Characters 101), Albertocetus meffordorum (based on USNM 525001), and an unnamed xeno- related to echolocation and/or soft tissue characters of the face were mapped onto rophid (GSM 1098). We also made one correction: the coding of character 65 (that the single most parsimonious tree in the application WinClada v.10.0008. In map- is, presence of rostral basin) was changed from present to unknown for Archa- ping our characters, we focused on unambiguous optimizations, although we also eodelphis patrius. The entire supermatrix was loaded into the application TNT28, explored fast (that is, ACCTRAN) and slow (that is, DELTRAN) optimizations. but before it wasopened, the option that treatsgaps in molecular sequences as miss- A total of 923 CT slices were obtained using Siemens Somaton Sensation in the ing data was activated (format . data format . read gaps as missing). The super- Department of Radiology and Radiological Sciences at the Medical University of matrix was quite large, thus the general random access memory was raised to at South Carolina, with each slice separated by 0.6 mm. The slices had sufficient reso- least 400 megabytes in TNT (settings .memory). As in previous studies24,25,between- lution to create cubic voxels of 0.584 mm per side, and the computer application character scaling was achieved by down-weighting ordered, multistate, morpho- Avizo v.7.1 was used to create the three-dimensional model. Relative density was logical characters so that they have the same minimum length as a binary character. visualized by assigning colours to each voxel based on their grayscale values. Parts Unordered multistate morphological characters and nucleotide characters were of the skull reconstructed with epoxy putty could easily be recognized in the CT not down-weighted. All weights are included in the datamatrix, which is available slices by their lack of internal structure and consistent density, and reconstructed for download. Heuristic searches were used to find the most parsimonious tree(s) areas identified through CT scans were confirmed by close inspection of external for the entire supermatrix. Specifically, the following default settings were used surfaces. Once verified, reconstructed portions were isolated digitally before col- under a ‘New Technology Search’: trees obtained under a driven search with five ours were applied to voxels of the three-dimensional model.

Extended Data Figure 1 | Images of the holotype skull of Cotylocara macei display similar angles. b, Medial view of the right dentary. Bo, basioccipital; (CCNHM-101). a, Parasagittal section of the skull generated from CT slices Bs, basisphenoid; cc, cranial cavity; co, mandibular condyle; cp, coronoid demonstrating the anatomical basis for klinorhynchy (downturned face and process; ip, interparietal; mf, mandibular foramen; ms, surface for mandibular rostrum). Red lines follow the axis of the basicranial stem and the rostrum; symphysis; Mx, Maxilla; Na, nasal; n-1, penultimate tooth; n-5, fifth from the together they form an angle of approximately 20u. This slice is situated ,10 mm last tooth; Pa, parietal; pf, postnarial fossa; Px, premaxilla; So, supraoccipital; to the right of the median plane. Other slices, medial and lateral to this one, xn, external bony nares.

Extended Data Figure 2 | Petrosal and tympanic bones of the holotype skull fenestra rotunda; fp, falciform process of squamosal; fi, fossa incudis; fm, of Cotylocara macei (CCNHM-101). a, Ventrolateral view of right petrosal mallear fossa; fo, fenestra ovalis; gf, glenoid fossa; he, epitympanic hiatus; articulated with the rest of the skull. Anterior is towards the upper left corner. iv, involucrum; mf, median furrow; opp, outer posterior prominence; b–d, Ventral, dorsal and lateral views of tympanic bulla. e–h, Ventrolateral, pap, paroccipital process; pc, pars cochlearis; pgp, postglenoid process; dorsomedial, dorsolateral and ventromedial views of right petrosal (periotic). pp, posterior process; sf, fossa for stapedial muscle; smf, suprameatal fossa; Scale bars are 1 cm. ap, anterior process; bc, basioccipital crest; ca, aperture for spf, fossa for sigmoid process; tc, tympanic cavity; tf, articular facet for posterior cochlear aqueduct; ch, cranial hiatus; ctp, caudal tympanic process; dc, dorsal process of tympanic; va, aperture for vestibular aqueduct; vlt, ventrolateral crest; eam, external acoustic meatus; fc, proximal opening of facial canal; fer, tuberosity.

Extended Data Figure 3 | Postcrania of holotype of Cotylocara macei anterior and posterior views. j, Middle cervical (C3 or C4) vertebra in anterior (CCNHM-101). a, b, Posteromedial views of anterior left ribs. c, d, Middle left view. k, Sixth cervical vertebra in anterior view. Scale bar for a–g is 5 cm; those ribs. e, Posterior left rib. f, g, Middle right ribs. h, i, Axis vertebra (C2) in for h–k are 1 cm.

Extended Data Figure 4 | Holotype of skull of Cotylocara macei (CCNHM- itc, infratemporal crest; nc, nuchal crest; npp, nasal process of premaxilla; 101). a, Dorsal view. b, Ventral view. Colours represent individual bones pap, paroccipital process; pf, postnarial fossa; pgp, postglenoid process; of the skull: turquoise, occipital; brown, nasals; grey, epoxy putty; light plp, palatine process of premaxilla; pop, postorbital process of frontal; green, lacrimal; olive, sphenoid; orange, parietals; pink, premaxillae; purple, ppf, preorbital process of frontal; pr, postorbital ridge; rb, rostral basin; palatines (ventral side only); red, petrosals; light blue, frontals; dark ps, palatine sulcus; rc, reconstructed tooth; rif, reentrant infraorbital foramen; blue, squamosals; yellow, maxillae; lime green, pterygoids; white, interparietal. spf, fossa for sigmoid process of tympanic; sqf; squamosal fossa; tif, thin lamina alv, alveolus; an, antorbital notch; ap, anterior process of petrosal; apl, of frontal; vlt, ventrolateral tuberosity of petrosal; V3, path of mandibular ascending process of lacrimal; apm, ascending process of maxilla; asf?, possible division of trigeminal nerve; zy, zygomatic process. Boundaries between bones air sinus fossa; cc, cranial cavity; df, deep fossa within the broader periotic fossa; are based on external sutures and CT data. Synchondrosis between basioccipital dif, dorsal infraorbital foramina; dta, anteriormost double-rooted tooth; dtp, and basisphenoid is completely closed, thus the boundary shown here is posteriormost double-rooted tooth; ep, embrasure pit; fp, falciform process of speculative. The frontal–parietal suture is damaged and is our best squamosal; fw, frontal window; gf, glenoid fossa; gpf, greater palatine foramen; interpretation. Scale bar is 5 cm.

Extended Data Figure 5 | Fossae in the holotype skull of Cotylocara inferior vestibule as well as the route by which it connected to the soft tissue macei (CCNHM-101) that are likely to have been filled with air sinuses. nasal passages. Red, premaxillary air sinus. La, lacrimal; Mx, maxilla; Na, nasal; a, b, CT-generated model of skull in oblique dorsolateral view. c, d,Sameas Pa, parietal; pf, postnarial fossa; Px, premaxilla; rb, rostral basin; So, a, b, but from a more dorsal perspective, anterior is towards the lower left supraoccipital; Sq, squamosal; zy, zygomatic process. corner. Light blue, excavations of an air sinus that may be homologous to the

Extended Data Figure 6 | Comparison between the relative densities of dorsal and ventral sides of the skull. c, d, Dorsal and ventral views of a bones in the holotype skull of Cotylocara macei (CCNHM-101) and the three-dimensional digital model of Tursiops truncatus (SDSNH 21212). modern bottlenose dolphin Tursiops truncatus (SDSNH 21212). Note the porous bone in the premaxillary sac fossa and the similar densities of a, b, Dorsal and ventral views of a three-dimensional digital model of the the dorsal and ventral sides of the skull. pf, postnarial fossa; psf, premaxillary holotype skull of Cotylocara macei (CCNHM-101), based on CT data. Densest sac fossa; Px, premaxilla; rb, rostral basin; SDSNH, San Diego Society of bones are in bright orange, most porous bones are in purple, and reconstructed Natural History. CT scans of Tursiops were performed at the University of portions are in grey. Note the dense rostrum, the less dense bone flooring Texas High-Resolution X-ray CT Facility (data courtesy of T. Rowe). The voxel the rostral basin and postnarial fossae, and the density difference between the size for these data was 0.298 mm 3 0.298 mm 3 0.9 mm.

Extended Data Figure 7 | Evolution of skeletal features likely associated to positions of the maxilla, as illustrated on the right side of a, with changes with echolocation. a, Skull of Georgiacetus vogtlensis in dorsal view: to the left between character states (states 0 through 5) labelled below for clarity. Dashed are conditions of characters mapped in d–h and to the right are the different terminal branches indicate that data are missing for that fossil taxon. Character possible character states for posterior migration of the maxilla (yellow) across evolution was inferred using parsimony on the entire phylogeny, although all taxa. b, c, Skulls of Cotylocara macei and Simocetus rayi in dorsal view, the d–i only include those taxa needed to convey the pattern of character evolution latter redrawn10, with their respective states for characters mapped in at the base of Odontoceti. All evolutionary reconstructions are unequivocal d–h. d–h, Evolution of characters that are likely to be associated with except for f, which is an ACCTRAN optimization. An alternative, but equally echolocation on our phylogenetic tree (Fig. 3). i, Summary of all changes parsimonious reconstruction for f, is discussed in the Supplementary Information. detailed in d–h with convergent characters in green and non-convergent An arrow in d–i indicates where echolocation is inferred to have evolved based on changes in black; numbers refer to characters described in the Supplementary its occurrence in bolded taxa: osteological evidence in Cotylocara macei and Information. In a–f and h red is the primitive state, blue is the most derived observational evidence in the crown group of Odontoceti. More details on the state, and purple is an intermediate derived state. Shades of grey in g correspond characters and their states can be found in the Supplementary Information.