First illustrated by Meckel (1826), the bill of the platypus, Ornithorhynchus anatinus, contains an unusual median bone anterior to the maxillae. As summarized by van Bemmelen (1901), as many as ten names were applied to this element, with most authors settling on either the os paradoxum (Albrecht 1883, cited in van Bemmelen 1901) or dumb-bell or dumb-bell-shaped bone (Turner 1884). The homologies of the os paradoxum have been the subject of long-term debate by some of the leading embryologists, comparative anatomists, and paleontologists of the times. Three principal views regarding it were expressed: (1) it is part of the premaxilla (e.g., Meckel 1826; Presley and Steel 1978); (2) it is a neomorph (e.g., Owen 1866; Flower 1876); and (3) it is the homologue of the vomer occurring in extant lizards (e.g., Broom 1895; De Beer 1937). Relevant to this debate are the homologies of the mammalian vomer, a midline bone, which is equated with two different bones in sauropsids, either with the paired vomer (e.g., Gaupp 1906; Starck 1967) or the midline parasphenoid (e.g., Broom 1902; De Beer 1937). Although some may argue, including this author, that these matters were convincingly laid to rest by Parrington and Westoll (1940), doubts about the homology of the os paradoxum continue to occur in the recent literature (e.g., Musser and Archer 1998; Sidor 2001; Musser 2013; Cheng et al. 2019). It is my goal here to present the history of study of the os paradoxum and evaluate its homologies, hopefully quelling continued debate. Included are relevant original observations on CT scans of Ornithorhynchus anatinus, the Miocene platypus Obdurodon dicksoni, and the didelphid marsupial Didelphis marsupialis.

To aid the reader in following the history of study regarding the os paradoxum, an overview of the anatomical domain in a placental and a lizard is presented.

Figure 1 shows fetal and young crania of a placental, the Philippine colugo, Cynocephalus volans, taken from Parker (1885b). The bony anterior palate is formed by the paired premaxilla (= intermaxilla of Meckel 1826; os incisivum of NAV 2017) lodging the incisors and the paired maxilla lodging the canines and postcanine dentition. The bodies of the premaxillae meet on the midline and each sends a narrow medial palatine process posteriorly (usually referred to as the palatine process in the older literature). Each medial palatine process abuts the paraseptal (Jacobson’s, vomeronasal) cartilage, which in turn shelters the vomeronasal (Jacobson’s) organ (Fig. 1A). The medial palatine process forms the medial border of the incisive foramen (Fig. 1B), which transmits the nasopalatine duct from the vomeronasal organ. At their posterior terminus (Fig. 1A), the medial palatine processes diverge from each other, lie dorsal to the maxillae, and contact a long midline bone, the vomer, which on its dorsal aspect lodges the nasal septum. A final bone of interest here found in the young colugo (Fig. 1B) is the submarine-shaped parasphenoid, a small midline bone positioned ventral to the basisphenoid.

Each of the elements named above exhibits considerable diversity across mammals. Regarding the adult premaxilla, Chiroptera, for example, shows a broad range of morphologies (Giannini and Simmons 2007): the medial palatine processes may be wholly absent, the left and right bodies may be separated by a midline gap, and the body may be separated from the maxilla. Although usually an outgrowth from the premaxillary body, the medial palatine process is reported to ossify from a separate center in, for example, the armadillo, Dasypus novemcinctus (= Tatusia novemcincta; Parker 1885a; Fawcett 1921). The paraseptal cartilage is generally much shorter than in the colugo (Fig. 1A), which has a remarkably long vomeronasal organ (Bhatnagar and Wible 1994). The paraseptal cartilage shows a variety of shapes in cross section and is reduced or absent in taxa lacking the vomeronasal organ (Wible and Bhatnagar 1996). The vomer generally arises from a single ossification, but paired centers are described in some taxa (De Beer 1837). The reports, for example, in Homo sapiens (Fawcett 1911; O’Rahilly and Gardner 1972) are supported by study of numerous ontogenetic stages. On the other hand, Gaupp (1908) observed what he interpreted as paired centers in one stage of the echidna Tachyglossus aculeatus (= Echidna aculeata) even though there was a weak median connector that he interpreted as a later addition; Kuhn (1971) cautioned that dual centers though likely in the echidna are not proven by this single stage. Among extant mammals, the parasphenoid is absent in monotremes (Kuhn 1971; Zeller 1989), has a spotty distribution in placentals but recently has been shown to be widely distributed in didelphid marsupials (Wible et al. 2018).

Figure 2 shows the cranium of an embryo sand lizard, Lacerta agilis, taken from Gaupp (1906). A general correspondence regarding the elements highlighted above exists between the lizard and colugo, although there are differences. The sand lizard premaxilla, for example, has a prenasal (ascending) process that supports the egg-tooth and lacks a medial palatine process; the more substantial parasphenoid has large posterolateral wings covering much of the basicranium. The vomer is a paired element in the sand lizard in contrast to the single ossification in the colugo and most extant mammals. Each vomer generally forms from a single ossification in sauropsids (De Beer 1937), but there is variability. A recent study of the snake Natrix natrix (Sheverdyukova 2019) reported three ossification centers for each vomer. Turtles have an unpaired vomer in the adult (Gaffney 1979), but at least in some this condition results from fusion of paired ossifications during ontogeny (e.g., Kunkel 1912; Tokita et al. 2021). Lastly, the vomer in the sand lizard has an incisive incisure on its lateral margin marking the exit of the nasopalatine duct (Fig. 2).

Van Bemmelen (1901) provided a detailed literature review of the os paradoxum (his praevomer) to which the reader is referred. Here, I focus on the highlights prior to and after 1901 as they pertain to the three major views on the homologies of this bone.

Nearly 200 years ago, Johann Friedrich Meckel (1826) published his monographic descriptions of the platypus. His table IV has the earliest illustrations of the skeleton, with his figure I showing a ventral view of the cranium (Fig. 3). Within the bill, Meckel labeled two disjunct parts of the paired os intermaxillare, his internum and externum, which in current terminology are the medial palatine process and body of the paired premaxillae, respectively. Meckel recognized that although the os paradoxum was a midline element, it was a paired bone. As noted above, the premaxillae exhibit considerable diversity among extant mammals, but to my knowledge this separation between the bodies and palatine processes is unique to the platypus.

The first alternative view on the homologies of the os paradoxum came from Owen (1866) and Flower (1876) who equated it with the neomorphic prenasal bone of the pig (os rostrale of NAV 2017). However, the persuasive paper by Turner (1884) presenting details of the adult anatomy of both elements left little doubt that the two were not homologues. The os paradoxum is in the palate, bound to the premaxilla and maxilla by membrane, supports the nasal septum and the vomeronasal organ and cartilage, and forms the medial border of the incisive foramina; the unique prenasal bone of the pig lies anterior to the palate and in a plane dorsal to it. Flower (1885) abandoned his prior view, citing Turner (1884), and replaced it with Meckel’s identification of the bone as part of the premaxilla. A neomorphic origin for the os paradoxum has not been given serious consideration since Turner (1884), although it was recently posed by Sidor (2001) without additional justification.

After Turner (1884) came an interchange of papers documenting the anatomy of the anterior nasal cavity of adult Ornithorhynchus based on serial sections. The main controversy was whether or not posterior spurs on the os paradoxum were in the plane of the vomer. Wilson (1894) claimed the spurs were in the plane of the vomer, making the os paradoxum a vomerine element, which he called the anterior vomers. In contrast, Symington (1891, 1896) claimed they were not in the plane of the vomer and continued to support the os paradoxum as the medial palatine processes of the premaxillae.

Broom (1895) agreed with Wilson (1894) that the os paradoxum was a vomerine element, replacing the term anterior vomer with a new term, prevomer, which he also applied to the paired vomer in lizards (Fig. 2). Broom did not just equate the os paradoxum with the lizard prevomer but also with the medial palatine process of the premaxilla of marsupials and placentals; that is, Broom considered the medial palatine process of the premaxilla as a composite structure formed by fusion with the prevomer (Fig. 4F). He supported this with prior observations by Parker (1885a, 1885b) and others that the medial palatine process in some placentals (e.g., Dasypus, Erinaceus) forms from an ossification (Parker’s anterior vomer) separate from the rest of the premaxilla. Reinforcing this was Wilson’s (1901) observation of two ‘mammary foetuses’ of Ornithorhynchus. In the earlier stage (Fig. 5A), the bodies of the premaxillae are fused on the midline to support the egg-tooth (os carunculae) and there are short medial palatine processes. In the later stage (Fig. 5B), a pair of ossifications appear posterior to and separated by a broad gap from the stubbles of the medial palatine processes, suggesting origins distinct from the premaxillae. Following Broom (1895), Wilson (1901) identified these separate ossifications as prevomers. Broom (1897) added evidence from the fossil record, reporting the apparent incidence of separate prevomers in a non-mammalian cynodont, the tritylodontid Gomphognathus.

Five years later, Broom (1902) added another twist to the debate, repeating a homological issue raised previously by Sutton (1884). Broom accepted the prevailing view that the mammalian vomer, the median bone lodging the nasal septum in the nasal cavity (Fig. 1A), is homologous in monotremes, marsupials, and placentals. Yet, if the platypus os paradoxum is the homologue of the paired bone associated with the vomeronasal organ immediately behind the premaxilla in lizards (the prevomer of Broom 1895), then to what median bone in the lizard is the platypus and, therefore, the mammalian vomer homologous? According to Sutton (1884) and Broom (1902), the appropriate bone to equate the mammalian vomer with is the sauropsid parasphenoid (Fig. 4D–F). Although citing Parker (1885a, 1885b) in his paper, Broom (1902) did not mention Parker’s (1885b; Fig. 1B) description of a small parasphenoid in the Philippine colugo (Figs 1B, 4C), which if true would negate the proposed homology between the sauropsid parasphenoid and mammalian vomer. Not everyone was convinced by Broom regarding the homologies of the platypus os paradoxum and the mammalian vomer: notable among the skeptics was Gaupp (1905, 1906), considered to be the most learned student of the skull of the day.

In 1929, De Beer, a proponent of Wilson’s and Broom’s view on the os paradoxum, proposed the following three stages in the evolution of the prevomer and premaxilla: (1) forming from separate ossifications, as in sauropsids (Fig. 4D); (2) forming from separate ossifications but the premaxilla secondarily developing a medial palatine process to buttress the vomeronasal organ that may secondarily fuse with the prevomer (as in some placentals; Fig. 4F) or not (as in Ornithorhynchus; Fig. 4E); and (3) forming from a single ossification (as in most placentals). Green (1930) reported on an ontogenetic stage of Ornithorhynchus that was not entirely congruent with De Beer’s scenario. In Green’s specimen, which was younger than that with a separate os paradoxum observed by Wilson (1901; Fig. 5B), there was a continuous “thread of ossification” connecting the os paradoxum and the stubble of the medial palatine process of the premaxilla. Green interpreted this thread as the product of resorption from an earlier stage where the os paradoxum was merely the posterior end of the medial palatine process of the premaxilla. He concluded that the existence of a separate prevomer in Ornithorhynchus was yet to be proven.

The ontogeny of the platypus skull was notably documented by Watson (1916), but based on a series of only two specimens. To fill in the gaps, De Beer and Fell (1936) reported on an ontogenetic series of five Ornithorhynchus, with the largest specimen the one previously described by Green (1930). Contra Green’s observation of this specimen, De Beer and Fell observed no thread of ossification connecting the premaxilla and os paradoxum (Fig. 6B), although they admitted this was not entirely clear in the available transverse sections. However, they stated (p. 20) even if Green was correct that the os paradoxum was secondarily separated from the premaxilla during ontogeny, that was not opposed to a prevomer origin for the os paradoxum, “since it might be supposed that the centres of ossification of the prevomers had previously become fused with those of the premaxillae, and subsequently separated, perhaps in consequence of the changes which are associated with the precocious development of the premaxillæ, their fusion in the mid-line, and the carrying of the egg-tooth.” The views of De Beer and Fell (1936) were endorsed in De Beer’s (1937) highly influential book “The Development of the Vertebrate Skull,” as were the views of Sutton (1884) and Broom (1902) regarding the homology of the mammalian vomer and sauropsid parasphenoid.

In 1940, two paleontologists, Parrington and Westoll, revisited the evidence from embryology and the fossil record. Their synthesis convincingly debunked the homological hypotheses that the prevomer of lizards = the os paradoxum of the platypus = the medial palatine process of the premaxilla of therians (Broom 1895, 1935) and that the parasphenoid of lizards = the vomer of mammals (Sutton 1884; Broom 1902, 1935) (Fig. 4D–F). Parrington and Westoll (1940) noted recent discoveries in the fossil record of the parasphenoid in synapsids, including non-mammalian cynodonts that also have a vomer resembling that in mammals. With both the parasphenoid and vomer present in these fossils, the mammalian vomer is not homologous with the sauropsid parasphenoid. As additional support, they also noted the discovery of a parasphenoid in the extant placental Cynocephalus by Parker (1885b; Figs 1B, 4C) and the extant marsupial Didelphis by Fuchs (1910). Parrington and Westoll (1940) also showed the prevomers are paired in basal synapsids, such as Dimetrodon, and in the same position as the paired element in lizards. In the theriodont clade (p. 322), “The prevomers fuse, and come to support the secondary palate in a manner indistinguishable from that of the mammalian vomer; the palatine processes of the premaxillae become more important, and attain mammal-like proportions.” In other words, the paired prevomer of lizards equals the vomer of mammals, and the os paradoxum of Ornithorhynchus is the detached rear of the medial palatine processes of the premaxillae (Fig. 4A–C). The years after 1940 saw little appetite for the proposals of Broom (1895, 1902, 1935). For example, Gregory (1947) identified the platypus os paradoxum as part of the premaxilla, even though according to Broom (1935), Gregory was an early advocate of the alternate hypotheses.

In his monograph on the development of the skull of the echidna Tachyglossus aculeatus, Kuhn (1971) found a stage with a short, thin medial palatine process of the premaxilla in continuity with the body, which was not present in the stages studied by Gaupp (1908). This process must disappear in later stages as it is wholly absent in the adult echidna (van Bemmelen 1901). Kuhn (1971) strongly supported the homologies of the vomer in mammals and reptiles as well as the os paradoxum as part of the premaxilla.

In 1978, Presley and Steel reported on an ontogenetic series of ten stages of Ornithorhynchus, including some specimens studied previously by Green (1930) and De Beer and Fell (1936). In specimens with snout-tail lengths of 80, 122 (contra the observations of De Beer and Fell on this specimen; Fig. 6B), 170, and 200 mm, the os paradoxum was continuous anteriorly with the premaxilla, while in specimens of 140, 225, and 240 mm, the attenuated connection breaks down. From this, Presley and Steel (1978) concluded that the os paradoxum is a detached portion of the premaxilla (see also Green and Presley 1978). Additionally, they reiterated that the mammalian vomer is equivalent to the paired vomer (Broom’s prevomer) in sauropsids.

In his 1981 book “The Mammalian Skull,” Moore included a synopsis of the arguments for and against the hypotheses of Wilson and Broom, and he clearly settled on the ‘against’ side. He questioned the value of ossification centers as a guide to a bone’s phylogenetic history, and the case of the two centers in the human vomer I noted above is one of many appropriate examples of this issue. He noted that the embryological evidence used by Wilson and Broom and supported by De Beer and Fell (1936) that the os paradoxum develops independent of the premaxilla is flawed, citing Green and Presley (1978) and Presley and Steel (1978). Lastly, he observed the overwhelming paleontological evidence raised first by Parrington and Westoll (1940) against the homologies of the mammalian vomer and sauropsid parasphenoid.

Zeller (1989) published a monograph on the development of the skull in Ornithorhynchus, which because of its synthetic treatment and comprehensive index became widely used by researchers studying early mammal phylogeny. For example, Musser and Archer (1998) extensively cited Zeller (1989) in their descriptions of the cranium of the Miocene platypus, Obdurodon dicksoni. Regarding the os paradoxum, Musser and Archer (1998: p. 1066) concluded that its origins were unknown and wrote that Zeller “could not positively identify this bone as either a prevomer or as part of the premaxillae.” This was an unfortunate comment because Zeller (1989: p. 72) could not have been clearer about his view on this bone: “Morphologisch ist das Os paradoxum ein Teil des Praemaxillare.”

The three most recent views on the os paradoxum that I have found in the literature cover the possible hypotheses about this bone except for the one settled on by most researchers in the wake of Parrington and Westoll (1940). Sidor (2001) described it as a possible neomorph; Musser (2013) equated it with the prevomer; and Cheng et al. (2019) called it “mysterious.” The last moniker suggests to me that the authors consider the os paradoxum of unknown origin. Rather than a neomorph, a prevomer, or mysterious, a review of the history of the os paradoxum clearly identifies it as part of the premaxillae.

CT scans of the following specimens were studied, with structures of interest segmented in Avizo 2020.3 (FEI SAS a part of Thermo Fisher Scientific).

1) The head, neck, and thorax of male Ornithorhynchus anatinus, MVZ MAMM 32885 [http://arctos.database.museum/guid/MVZ:Mamm:32885]. Museum of Vertebrate Zoology, University of California, Berkeley provided access to these data, the collection of which was funded by oVert TCN, NSF DBI-1702442, and NSF DBI-1701714. The files were downloaded from www.MorphoSource.org, Duke University. The CT image series included 1,878 tiff images [https://www.morphosource.org/concern/media/000059596?locale=en; ark:/87602/m4/M59596]. X, Y, and Z spacing is 0.0768222 mm.

2) The cranium of the Miocene ornithorhynchid Obduodon dicksoni, QM F20568. This specimen from the Riversleigh Formation, the only known cranium of a fossil ornithorhynchid, was scanned by Dr. Richard Ketcham at the University of Texas High resolution X-ray Computed Tomography Facility on 12–13 November 1998 along the coronal axis for a total of 612 512×512 pixel slices with each slice 0.26 mm thick, with an interslice spacing of 0.22 mm. For more information, visit http://www.digimorph.org/specimens/Obdurodon_dicksoni. For permission to access the scans, I acknowledge Dr. Michael Archer, University of New South Wales, and Dr. Ted Macrini, St. Mary’s University. The cranium of this specimen is described in Archer et al. (1992, 1993) and Musser and Archer (1998) prior to the CT scans, and the endocast of this specimen is described in Macrini et al. (2006) based on the scans.

3) The cranium of Didelphis marsupialis, du baa 0164. Duke University provided access to these data, the collection of which was funded by NSF BCS 1552848 (to D.M. Boyer) and NSF DBI 1458192 (to G.F. Gunnell). The CT image series included 1,910 tiff images [doi: 10.17602/M2/M58076]. X, Y, and Z spacing is 0.0690778 mm.

As with any anatomical contribution, it is important to acknowledge the source of terminology. The substance of this paper concerns a near 200-year debate concerning the homologies of bony elements of the mammalian rostrum. The terminology employed is wholly dependent on the side of the debate taken. Here, I follow the terms used by the main proponents of the debate. I usually try to conform to the Nomina Anatomica Veterinaria (NAV 2017), using anglicized versions of the Latin terms. However, NAV (2017) has a limited number of terms for the anterior rostrum.

Institutional Abbreviations: du baa, Duke University, Biological Anthropology and Anatomy, Durham, North Carolina, USA; MVZ, Museum of Vertebrate Zoology, University of California, Berkeley, California, USA; QM, Queensland Museum, Brisbane, Australia.

The mammalian skull has ossifications widely considered as evolutionary novelties. Well-known examples include entotympanics, independent elements in the auditory bulla (e.g., MacPhee 1979, 2014; Maier 2013) and the rostral or prenasal bone of the pig (e.g., Herring 1972; Hou et al. 2014; NAV 2017). More obscure examples include the os proboscidis of Solenodon paradoxus (Wible 2008) and the internasal bone of the two-toed sloth Choloepus (Gaudin et al. 2021). The origin of another bone in the anterior rostrum remains controversial, the xenarthran os nariale, considered a neomorph by some (e.g., Wible et al. 1990; Wible and Gaudin 2004) and a transformed septomaxilla by others (e.g., Zeller et al. 1993). Recently, a hypothesis that the therian premaxilla results from fusion with the septomaxilla, traceable back to Gaupp (1905, 1906), has received support from comparative embryological studies distinguishing the therian premaxilla from that in monotremes and other tetrapods (Higashiyama et al. 2021). If true, this hypothesis supports the element in xenarthrans as a novelty and may also explain why the neomorph examples above, with the exception of the entotympanics, are from the tip of the snout, an anatomical domain transformed in therians (Higashiyama et al. 2021).

Although the mammalian skull includes neomorphic structures, the usual first course of action for neontologists and paleontologists is to identify and name structures in the context of elements already known in other taxa, with the implicit assumption that creating novelties is less parsimonious. The platypus os paradoxum is an instructive example in that early invocations of novelty by Owen (1866) and Flower (1876) were abandoned when detailed anatomy (e.g., Turner 1884; Wilson 1901) supported linkage to pre-existing bones in related forms. The controversy was that two camps promoted different bony origins using the same ontogenetic database. The Wilson-Broom-De Beer camp (Fig. 4D–F) determined the os paradoxum to be a separate paired ossification, which was used to support its homologies with the paired sauropsid vomer (Broom’s prevomer), with the corollary that the unpaired mammalian vomer was the equivalent of the unpaired sauropsid parasphenoid. This camp held sway for the first forty years of the nineteenth century, although acceptance was not universal (e.g., Gaupp 1905, 1906). The second camp (Fig. 4A–C), traceable back to Meckel (1826) and Turner (1884), determined the os paradoxum to be part of the premaxillae ontogenetically through the studies of Green (1930), Green and Presley (1978), and Presley and Steel (1978). Moreover, this camp incorporated new discoveries in the fossil record that the parasphenoid is present in non-mammalian synapsids and that the paired vomer in early synapsids gave way in later taxa to an unpaired vomer resembling that in extant mammals (Parrington and Westoll 1940). Wible et al. (2018) further strengthened this view with new observations of the parasphenoid in extant marsupials and the suggestion, bolstered here, of a parasphenoid in the Miocene platypus Obdurodon (Fig. 9). The presence of a parasphenoid in the mammalian lineage complicates the Wilson-Broom-De Beer camp by debunking the hypothesis that the mammalian vomer is homologous with the sauropsid parasphenoid (see also Atkins and Franz-Odendaal 2016). With the mammalian vomer the homologue of the sauropsid vomer (Broom’s prevomer), the os paradoxum cannot be the prevomer.

Ontogeny has identified the os paradoxum as forming as part of the medial palatine processes of the premaxillae. Anatomical comparison of the elements in adult Ornithorhynchus (Fig. 7) and Didelphis (Fig. 8) reinforces the correspondence in structure, with concavities for the paraseptal cartilages (and vomeronasal organs) and V-shaped septal processes supporting the cartilaginous nasal septum. Recent fossil discoveries reveal that this arrangement predates the origin of Mammalia. For most non-mammalian eutheriodonts, the location of the vomeronasal organ is interpreted to be in concavities on the vomer (e.g., Maier et al. 1996; Hillenius 2000; Crompton et al. 2017; Pusch et al. 2019). However, Ruf et al. (2014) showed that the Late Triassic advanced non-mammalian cynodont Brasilitherium, widely considered to be the sister group to Mammaliaformes (e.g., Soares et al. 2014; Martinelli et al. 2017), had medial palatine processes resembling those in Didelphis with concavities for the vomeronasal organs and septal processes. Not surprisingly, the absence of medial palatine processes in adult Tachyglossus and the isolation of the os paradoxum from the rest of the premaxillae in adult Ornithorynchus are derived conditions likely related to their specialized lifestyles. It is not known if Obdurodon had an os paradoxum; Musser and Archer (1998) reconstructed one, noting the remarkable similarities to the bony rostrum of Ornithorhynchus. One clear point is that the premaxillae in Obdurodon are not as reduced as in Ornithorhynchus because its right and left premaxillary bodies approximated each other on the midline (Fig. 9A). It is my conclusion that the os paradoxum of Ornithorhynchus should not be labeled as mysterious, a novelty, or the prevomers but as a retained portion of the premaxillae that were undergoing reduction in the monotreme lineage.