Purpose of the Present Study

A reason for this lack of confidence in insectivoran relationships may be that basic information about the anatomy of several insectivoran-grade groups remains unknown. Even with the notable contributions of Roux (1947), Bugge (1974), and MacPhee (1981), details on cranial vasculature in the Tenrecidae has to date been adequately described in less than half of its constituent genera. Published information on the tenrecid nasal fossa is even more scarce, even though Broom (1915b) and Hofer (1982a) have made a good start. This lack of anatomical knowledge for tenrecs is surprising, given the fact that both regions have historically served as important sources of phylogenetic information in other mammalian groups (e.g., Kuhn, 1971; Maier, 1980; Wible, 1987; Sánchez-Villagra, 2001).

The present study makes a contribution to this shortfall by describing cranial arterial supply in Potamogale, Micropotamogale, Geogale, Microgale, Echinops, Tenrec the nontenrecid Erinaceus; and the anterior nasal region for these taxa plus the tenrec Setifer All of these animals are insectivoran grade and have traditionally been classified in the Lipotyphla sensu Haeckel (1866). After describing the cranial anatomy of these taxa, I distill this information into a limited number of discrete characters and code them for the preceding taxa plus 16 other mammals, including the marsupial Didelphis I then map these characters onto phylogenetic trees representing alternative hypotheses of insectivoran relationships, and note which hypothesis best explains their distribution.

A definitive solution to the phylogeny of insectivoran-grade mammals must await a simultaneous analysis of morphological and sequence data and sample relevant living and fossil taxa. The scope of this paper is much more limited, providing basic anatomical data that may be relevant for larger biological issues. In the short term, I use these data to evaluate previously existing phylogenetic hypotheses of African insectivoran phylogeny.

Taxonomic Terminology

The following taxonomic conventions will be used throughout this text: “Placental” mammals include crown group members of the Eutheria. “Eutheria” is more inclusive and encompasses members of the stem lineage leading to placental mammals (Novacek et al., 1997). The “Lipotyphla” denotes extant chrysochlorids, erinaceids, Solenodon soricids, talpids, tenrecids, the subfossil Nesophontes and several extinct Tertiary groups including (for example) apternodontids, dimylids, erinaceomorphs, and geolabidids (Carroll, 1988; Gould, 1995; McKenna and Bell, 1997; Asher, 1999). In the 19th and early 20th centuries, “Insectivora” commonly included lipotyphlans plus menotyphlans (i.e., scandentians and macroscelideans). During the last 50 years, “Insectivora” has most often been used interchangeably with “Lipotyphla”, and is so regarded here. However, I use “Insectivora” in an explicitly gradistic sense. “Insectivora” and “insectivoran” are taxonomic terms and do not imply any particular dietary strategy, although “insectivore” does (cf. “carnivoran” vs. “carnivore”). The “Afrotheria” refers to the group designated by Stanhope et al. (1998), that is, proboscideans, sirenians, hyracoids, tubulidentates, macroscelideans, tenrecids, and chrysochlorids. Proboscideans and sirenians are collectively referred to as the Tethytheria (Fischer, 1996), and these two plus hyracoids are referred to as the Paenungulata (Shoshani, 1993). Stanhope et al. (1998) proposed the name “Afrosoricida” to denote a clade containing extant chrysochlorids and tenrecids. However, 40 years earlier McDowell (1958) proposed the name “Tenrecoidea” as a superfamily with the same content. Other workers have used “Tenrecoidea” in a broader sense both formally (e.g., Gill, 1884) and informally (Asher, 1999) so as to include Solenodon and/or extinct apternodontids, in addition to tenrecs and golden moles. Here, I follow McDowell in using “Tenrecoidea” to indicate a tenrec-chrysochlorid clade (cf. Mouchaty, 1999).

Anatomical Terminology

I follow McDowell (1958), Novacek (1986), and MacPhee (1994) for osteological terminology; Bugge (1974), MacPhee (1981), and Wible (1984) for arterial terminology; and Sturm (1936), Kuhn (1971), and Maier (1980) for nasal terminology. Variations on anatomical terms used by these and other investigators are listed in table 1.

Morphological Character Coding

From the anatomical descriptions made possible by these methods, a number of discrete anatomical characters were constructed. No single method of transforming morphological observations into characters appropriate for phylogeny algorithms is without drawbacks (Strong and Lipscomb, 1999). Variable morphology coded into binary “presence/absence” characters maximally subjects primary hypotheses of homology (sensu De Pinna, 1991) to falsification (Pleijel, 1995; Lee and Bryant, 1999). With this coding method, a researcher's assumptions about evolutionary process (i.e., transformations between multiple character states) are not concealed within a single character. However, this method may spuriously overweight as “absent” characters that are in fact inapplicable (Strong and Lipscomb, 1999; Lee and Bryant, 1999). To continue an example used by Maddison (1993), organisms that lack a tail would be coded as “absent” for other characters regarding tail pigmentation. Depending on how many colors are observed in tailed taxa, those that are tailless would show multiple “homologies” of absent blue, absent yellow, and so on. This strict binary coding falsely treats certain states as biologically real, when in fact their existence is semantic; the character state “blue: absent” has not actually been observed in any organism and should not provide evidence with which to define a clade (Hawkins et al., 1997; Strong and Lipscomb, 1999).

Alternatively, variable morphology may be coded into multistate characters, with “missing” entries used for taxa that are inapplicable for a given character. This technique avoids the overweighting and reificationist pitfalls of binary coding described above, but of course requires assumptions about evolutionary process that binary coding lacks. In addition, the treatment of inapplicable characters as “missing” may yield odd state assignments to ancestral nodes, as when the common ancestor of tailless taxa is optimized as having possessed a tail (Platnick et al., 1991; Strong and Lipscomb, 1999).

Given these considerations, I have attempted in this study to maximally subject primary hypotheses of homology to falsification by favoring the use of binary characters, but have employed multistate characters in cases where binary coding would have resulted in overly redundant, morphologically unobserved, or (following Simmons and Geisler, 1998) uninformative characters.

Cranial Arterial Supply

Several authors have described what is hypothesized to be the primitive arterial pattern upon which modern placentals have acquired modifications to varying degrees (e.g., Bugge, 1974; Presley, 1979; Cartmill and MacPhee, 1980; Wible, 1984, 1987; Wible et al., 2001). The possibility that this pattern actually describes primitive states for tenrecid or other insectivoran groups remains to be demonstrated. Nevertheless, it is useful for descriptive purposes and is shown in figure 3. In addition to the proximal stapedial, internal, and external carotid branches discussed by the above authors as primitive for placental mammals, figure 3 depicts several terminal branches of the stapedial and external carotid arteries that are frequently observed among tenrecids.

In this primitive cranial arterial pattern, the common carotid artery ascends in the neck and divides into the internal and external carotid arteries. Along the ventrum of the pars cochlearis, the internal carotid gives off the proximal stapedial artery and continues anteriorly in a transpromontorial position (sensu Wible, 1986). After passing through the stapes, and still within the middle ear, the stapedial artery bifurcates into superior and inferior rami. The superior ramus leaves the tympanic cavity, running dorsad and forward into the sinus canal. The inferior stapedial ramus continues anteriorly along the ventrum of the tympanic roof. After leaving the middle ear, it bifurcates into the ramus infraorbitalis and ramus mandibularis ventral to foramen ovale (Wible, 1984: 40, 186; 1987: 121). The ramus infraorbitalis runs forward toward the infraorbital canal and face, and the ramus mandibularis supplies the jaw.

Through these carotid branches, most of the head receives blood supply: the brain and eye via the internal carotid artery (supplemented by the vertebral arteries); the mandible and infraorbital region via the inferior stapedial ramus; the meninges, supraorbital, and part of the nose via the superior stapedial ramus; and the occiput, tongue, face, palatine, buccal, and temporal regions via the external carotid. Wible (1987) has noted that certain placentals (e.g., armadillos), and possibly the basal eutherian Prokennalestes (Wible et al., 2001), have a connection between the superior stapedial and occipital vasculature via the arteria diploëtica magna. Hence, this condition may also characterize the placental common ancestor; however, the relations of the arteria diploëtica magna were not recorded in this study.

Among the mammals described below, Echinops and Micropotamogale best approximate the pattern shown in figure 3, exhibiting variations only in supply to the eye. Otherwise, there are several deviations from this arterial pattern exhibited by living insectivorans in general and by tenrecids in particular. These are described below in six tenrecs (Potamogale, Micropotamogale, Geogale, Microgale, Echinops and Tenrec) and the hedgehog Erinaceus The descriptions proceed from proximal to distal, starting at the carotid bifurcation, then following the internal carotid, stapedial, and external carotid branches. In general, a common set of structures is described for each specimen. However, because the available histological specimens vary in quality, some specimens (e.g., the Echinops and 31-mm snout-vent length Erinaceus specimens) are described in slightly more detail than others. Descriptions are based on specimens listed after the taxonomic heading of each section. Using the technique described above, a graphic reconstruction of arteries present between the carotid bifurcation and the posterior jaw is presented for the tenrecs in figures 4–8.

Potamogale velox (ZIUT, HL 17 mm; fig. 4)

Micropotamogale lamottei (IZEA 939; fig. 5)

Geogale aurita (MCZ 45504, 45499; fig. 6)

Microgale dobsoni (MPIH 1964/103; fig. 7)

Echinops telfairi (ZIUT, HL 18 mm; fig. 2)

Tenrec ecaudatus (ZIUT, HL 20 mm; fig. 8)

Erinaceus europaeus (ZIUT, SVL 31 and 37 mm), E. concolor (ZIUT, HL 33 mm)

List of Arterial Characters

From the preceding descriptions, a number of characters can be used to summarize observed morphological variation. Admittedly, these characters do not represent all of the observed variation. Additional characters regarding, for example, the anastomosis of the stapedial system with an arteria diploëtica magna (Wible, 1987) may be of use in future studies. Nevertheless, the characters defined below reflect consistent and easily identifiable differences in arterial supply across the observed taxa.

Among humans it is well known that differences exist in arterial supply across individuals; for example, the facial and lingual arteries in humans variably arise independently from the external carotid or share a common trunk (Agur and Lee, 1991: 562). With this in mind, differences observed above regarding, for example, the origins of the ascending pharyngeal, anastomotic channel for the deep cervical, and sphenopalatine arteries are not used here to form phylogenetic characters. On the other hand, some characters may vary during ontogeny but appear to be consistent among older stages (e.g., involution of an intratympanic posterior stapedial ramus in Erinaceus). The apparent absence of an external carotid-supplied posterior auricular artery in potamogalines is worth examination in future studies. However, the distribution of this small artery is not incorporated into phylogenetic analyses presented here because of the difficulty of consistently verifying its presence in the current histological sample.

Original histological sections were not available for several taxa. Data on Bradypus and Dasypus are taken from Wible (1984), on Echinosorex and Tupaia from Cartmill and MacPhee (1980), on Procavia from Lindahl and Lundberg (1946), and on Talpa from Tandler (1899), Sicher (1913), and Bugge (1974).

Characters summarized below (see also table 3) represent the intertaxon variability observed in taxa described above and others listed in table 2. As is the case for other characters defined here, polarity is not defined a priori, but rather a posteriori by the process of rooting undirected trees (Farris, 1982; Nixon and Carpenter, 1993). Hence, character state numbers listed below have no necessary bearing on polarity.

1. Posterior stapedial ramus: When present in adult taxa, this artery arises from the proximal stapedial medial to the stapes and runs posterolaterally into the stapes fossa (e.g., Echinops state 0). In some taxa the proximal stapedial artery gives off no branches immediately medial to the stapes (e.g., Tenrec state 1).

2. Superior stapedial ramus: In adult Echinops the proximal stapedial bifurcates into superior and inferior rami after passing through the obturator foramen of the stapes. The superior ramus then courses in the sinus canal within the braincase along the surface of the lateral cranial wall (state 0). Several other taxa (e.g., Potamogale) show varying levels of involution of the superior stapedial ramus, all of which result in a small or absent superior ramus with supply of cranial structures such as eyes and meninges taken over by other arterial sources (state 1).

3. Inferior stapedial ramus: This inferior branch of the stapedial bifurcation, with target structures in the jaw, upper dentition, and/or face, is either present (e.g., Echinops state 0) or absent (e.g., Tenrec state 1).

4. Origin of ophthalmic artery: In Setifer the ophthalmic artery arises from the internal carotid (state 0) and is apparent passing through the bony optic canal accompanied by the optic nerve. Several taxa show an ophthalmic artery originating from the superior stapedial ramus as it leaves the sinus canal (e.g., Echinops state 1). Bugge (1974) described the ophthalmic artery in Talpa and Erinaceus as originating from the inferior stapedial-supplied ramus infraorbitalis (state 2). This pattern was also observed in the Erinaceus concolor specimen described above. Carnivorans such as Nandinia show an ophthalmic artery supplied by a branch of the external carotid (state 3). Polymorphisms are occasionally apparent in the origin of optic vasculature; hence, a given taxon may be coded as having arterial supply from multiple sources.

5. Origin of ramus mandibularis: In most taxa (e.g., Echinops), the arterial supply to the jaw is provided by the inferior stapedial ramus (state 0). Alternatively, the ramus mandibularis supplying the jaw may originate from the external carotid artery (e.g., Tenrec state 1).

6. Internal carotid relation to anterior cochlea: As the internal carotid artery approaches the point at which it enters the braincase, it may be medial (e.g., Microgale state 0; fig. 16) or lateral (Chrysochloris state 1; fig. 21) to the anterior pole of the pars cochlearis of the petrosal.

7. Bifurcation of stapedial artery: Typically (e.g., Echinops), the proximal stapedial artery bifurcates into inferior and superior rami after passing through the obturator foramen of the stapes, within the tympanic cavity (state 0). Geogale and Solenodon (fig. 22) show a stapedial bifurcation above the roof of the middle ear, within the braincase (state 1).

8. Proximal course of ramus infraorbitalis: Posterior to the exit of the ophthalmic and maxillary divisions of the trigeminal nerve from the cranium, the ramus infraorbitalis may travel within the cavum epiptericum, ventral to trigeminal nerve bundles (e.g., Echinops state 1), or it may be extracranial (e.g., Blarina state 0).

9. External carotid relation to thyropharyngeus muscle: In Echinops (fig. 18) the external carotid passes medial to a slip of thyropharyngeus muscle, just proximal to its division into lingual, facial, and temporal trunks (state 0). Other taxa (e.g., Micropotamogale; fig. 12) show the external carotid artery consistently lateral to thyropharyngeal musculature (state 1).

Anterior Nasal Capsule

In a number of papers during the late 19th and early 20th centuries, Robert Broom emphasized the importance of the anterior mammalian nasal capsule for systematic purposes. Based on nasal anatomy, Broom (1897) proposed two high-level groups of mammals: Archaeorhinata (including edentates and rodents) and Caenorhinata (including most other placental mammals). Further work (1898, 1902, 1909, 1915a, 1915b) led Broom to add aardvarks, lagomorphs, tree shrews, elephant shrews, and golden moles to his archaeorhinate grade. As this name implies, he regarded these taxa as representing a primitive grade of organization and argued that they shared several features of nasal anatomy with monotremes and marsupials. Most consistently, Broom's archaeorhinates lack cartilaginous support for the nasopalatine canal. Caenorhinates, in contrast, possess a nasopalatine canal well supported by cartilage (Broom, 1898: 712). Broom (1898) also implied that archaeorhinates often show a vomeronasal duct that opens directly into the nasal fossa. However, further study (e.g., Sánchez-Villagra, 1998, 2001) has shown that most marsupials possess a vomeronasal duct that opens directly into the nasopalatine canal, as in caenorhinates.

By focusing on the nasal region, Broom came up with taxonomic conclusions that, 100 years later, have received support from contemporary morphologists (e.g., MacPhee and Novacek, 1993) and/or molecular biologists (e.g., Mouchaty et al., 2000). These conclusions include the removal of macroscelidids, scandentians, and chrysochlorids from the “Insectivora” (Broom, 1915a, 1915b) and the basal position of rodents within Placentalia (Broom, 1897).

A number of other researchers have documented anatomical diversity in the nasal region across mammals (e.g., Sweet 1904; Ärnbäck Christie-Lindie, 1914; De Beer, 1929; Sturm, 1936; Reinbach, 1952a, 1952b; Kuhn, 1971; Hofer, 1977, 1982a, 1982b; Maier, 1980, 1991, 1997; Wöhrmann-Repenning, 1984, 1987, 1991; Wible and Bhatnagar, 1996; Frahnert, 1998; Asher, 1998; Mess, 1999; Göbbel, 2000; Sánchez-Villagra, 2001) and have to varying degrees noted phylogenetically informative differences. However, except for three unpublished academic dissertations (Schunke, 1996; Behrens, 1998; Asher, 2000), the anatomy of the anterior nasal capsule among tenrecs has been known only for tenrecine genera (Broom, 1915b; Roux, 1947; Hofer, 1982a); information on potamogalines, oryzorictines, and Geogale has until now been unavailable in the published literature. Here, I describe the anterior nasal capsule in Potamogale, Micropotamogale, Geogale and Microgale as well as Echinops, Setifer, Tenrec, Solenodon and Erinaceus

The following descriptions trace relevant anatomical structures as they occur from posterior to anterior. For readers who have access to some or all of the original histological material used here (table 2), section numbers are referred to throughout the descriptions in the format “slide#.column#.row#”, as described above in the section on cranial arteries. Sections from the MPIH now located in Düsseldorf and New York are numbered by distance in 0.01 mm units from the anterior point of sectioning (typically the anterior rostrum). For some taxa, individuals representing multiple ontogenetic stages are represented (table 2). Morphological characters that vary ontogenetically will be detailed at the end of the descriptions. Each description focuses on a single individual identified after the taxon heading. Depending on availability, identifications use catalog numbers, specimen measurements, or simply institutional provenance. In all cases, details of the specimens listed in table 2 include unique aspects (e.g., section thickness, type of stain, embedding medium), which facilitate identification of the original specimen.

Maier (1993) has presented a model for the cranial anatomy of the common ancestor of placental and marsupial mammals, including information on nasal structures. Sánchez-Villagra (2001) has focused on the anatomy of the anterior nose among marsupials; and Giere et al. (1999) have discussed how one aspect of anterior nasal morphpology (the anterior opening of the vomeronasal organ) may have appeared in the therian common ancestor. In general, however, and unlike the case for cranial arteries, a widely accepted model for the anterior nasal anatomy of early placental mammals does not yet exist. Thus, the following descriptions focus on the major elements of nasal anatomy (fig. 23) as discussed by Broom (numerous papers cited above), Maier (1980), and Sánchez-Villagra (2001).

A general outline of nasal structures discussed in the following pages is as follows (see also fig. 23). The bilateral vomeronasal organ is situated anteriorly on the nasal floor, on either side of the nasal septum. A paraseptal cartilage supports the vomeronasal organ along its length. Anteriorly, the vomeronasal organ typically empties via the vomeronasal ducts into the nasopalatine ducts, which connect the nasal and oral cavities through the incisive foramina. A cartilaginous “outer bar” variably extends lateral to the vomeronasal duct, connecting medial and ventral components of the paraseptal cartilage. As the nasopalatine ducts empty into the oral cavity, they define a palatine papilla, situated ventral to the incisive foramina. Within this papilla, a spherical or tubelike papillary cartilage is sometimes present. The nasopalatine duct cartilage is usually evident along the lateral and posterior aspect of the nasopalatine ducts. Depending largely on the ontogenetic stage of the individual, the floor of the nasal capsule anterior to the premaxilla, or the “anterior transverse lamina”, may be continuous with the medial aspect of the paraseptal cartilage (fig. 23B). A “zona annularis” is said to be present when the anterior transverse lamina is continuous with the sidewall and roof of the nasal capsule. The anteriormost nasal capsule consists of two nasal cupulae on either side of the nasal septum. Posterior and lateral to the cupulae are the cartilaginous superior alar processes. These provide attachments for small muscles of the rhinarium and variably show connections with the nasal sidewall, anterior transverse lamina, and cupular processes (fig. 23A). Variations on this general anatomical theme are described on a taxon-by-taxon basis below.

Potamogale velox (ZIUT, HL 17 mm; fig. 23)

As in other mammals, the bilateral vomeronasal organ of Potamogale lies on either side of the nasal septum on the nasal floor. In this specimen with a head length of 17 mm, it starts from a region below the anterior cribriform plate and stretches about 2.6 mm anteriorly to the incisive foramina. In the anteroposterior dimension, the vomeronasal organ is 38% as long as the nasal fossa (6.9 mm) and 15% of the head length (17 mm). The vomeronasal organ in Potamogale extends farther posteriorly (i.e., ventral to the cribriform plate) than that of a specimen of Cynocephalus described by Bhatnagar and Wible (1994: 47) as “one of the relatively longest … reported thus far in mammals”. The large vomeronasal organ of Potamogale corresponds with its unusually large accessory olfactory bulb (Stephan et al., 1991). No single major blood vessel is apparent traveling with the vomeronasal organ; however, numerous small vessels are evident along its length.

Along its entire medial and ventral aspects, the vomeronasal organ is supported by a J-shaped paraseptal cartilage (e.g., slice 16.2.4). At no point is the vomeronasal duct enclosed laterally by cartilage; however, immediately before it reaches the nasopalatine duct, an incomplete cartilaginous strut is present dorsal to the vomeronasal organ, extending laterally from the midpoint of the paraseptal cartilage (slice 12.1.6; fig. 24). Anteriorly, the organ grades into a narrow vomeronasal duct, which empties on each side directly into the nasopalatine ducts (e.g., slice 12.5.6), which in turn connect the nasal fossa and mouth via two openings on either side of the palatine papilla (slice 12.1.6; fig 24). The nasopalatine duct is laterally supported by a cartilage that runs ventromedially to the lateral rims of the incisive foramina; this nasopalatine duct cartilage is not connected to the paraseptal cartilage.

Broom's “inferior septal ridge” (1896: 593), consisting in most taxa of an anteroposteriorly running emargination of mucosa on the inferolateral aspect of the nasal septum and occasionally supported by anterior paraseptal cartilage or the adjacent anterior transverse lamina, is present in Potamogale far anterior to the vomeronasal organ. It lacks cartilaginous support.

There is a broad connection between the medial component of the paraseptal cartilage and the anterior transverse lamina (fig. 23B). DeBeer (1929: figs. 5–8) also noted continuity between these structures in Sorex araneus; in fact, he referred to both paraseptal cartilage (supporting the vomeronasal organ) and anterior transverse lamina (forming the nasal capsule floor anterior to the vomeronasal duct) by the single nomen “anterior paraseptal cartilage”. Anterior to the premaxilla of Potamogale the nasal cartilages form a complete, uninterrupted ring surrounding the nasal cavity, connecting the nasal septum, lamina transversalis anterior, and nasal sidewall and roof (fig. 23A). This cartilaginous ring is referred to here as a “zona annularis” (Gaupp, 1908). Kuhn (1971: 28) has recommended abandoning this term because of inconsistent usage; nevertheless, I retain “zona annularis” as defined above.

A remarkable aspect of Potamogale is the absence of the nasolacrimal duct—a common structure among mammals that usually lies in an anteroposteriorly running sulcus along the lateral nasal capsule (see figs. 25 and 26 for an illustration of this structure in Solenodon and Erinaceus). The nasolacrimal duct is also absent in the older Potamogale specimen described by Behrens (1998). This corresponds with the absent lacrimal gland in this individual, and is consistent with the observation that a lacrimal foramen is absent in adult Potamogale skulls (Asher, 1999).

The most conspicuous structures of the Potamogale anterior cartilaginous nose are the superior alar processes, located immediately posterior to the external nares (fig. 23A; slice 5.5.4). These provide a large surface area for attachment of facial muscles and connective tissue (slice 5.5.4). Dorsal to each of the anterior nasal cupulae lies a large, superiorly projecting “spur” (fig. 23A; slice 4.5.6; see Roux [1947: fig. 55] for a comparable structure in Setifer), dorsal to which runs a prominent alar-cupular muscle, which in turn originates from the dorsal aspect of the superior alar process (slice 6.3.2). The dorsal spur provides this muscle with a greater moment arm with which to act upon its insertion point in the superior rhinarium (slice 4.1.1). The superior alar processes are broadly connected to the floor of the anterior nose. In addition, a narrow alicupular commissure connects the superior alar process and cupular process (fig. 23a; slice 4.3.3). In contrast to the individual of Setifer figured by Roux (1947: fig. 55), the superior nasal fenestrae (located just posterior to the cupular spurs) in Potamogale are occluded by its large superior alar processes, and are not widely exposed dorsally.

Micropotamogale lamottei (IZEA 939)

The vomeronasal organ of Micropotamogale is approximately 3.5 mm in length and does not reach as far posteriorly as the first ethmoturbinal; in an adult specimen (IZEA 939) it extends only as far back as a frontal plane posterior to the canines and anterior to the anteriormost premolars (slice 48.3.2). In this specimen (IZEA 939) the vomeronasal organ comprises about 13% of nasal fossa length (ca. 26 mm) and 7% of head length (ca. 50 mm). Fetal specimens (i.e., AIG 2273–1, 2273–3) show a more elongate vomeronasal organ, but still shorter than that seen in a Potamogale of comparable age (table 2). No major blood vessels are apparent coursing along the length of the vomeronasal organ; instead, numerous smaller vessels are visible.

Most of the support for the vomeronasal organ along its length is ossified, although some paraseptal cartilage remains ventral and lateral to the organ. In this regard, Micropotamogale resembles the specimens of Geogale described below. This is due to the advanced ontogenetic stage represented by the IZEA 939 Micropotamogale and all Geogale individuals (all showing fully erupted teeth).

Anteriorly, the cartilaginous outer bar does not cover the vomeronasal organ laterally. The vomeronasal ducts are apparent at the anterior extreme of the vomeronasal organ, consisting of tubes on either side of the nasal septum that empty into each nasopalatine duct, separated from the nasal fossa proper (e.g., slice 31.3.2). Elongate, thin cartilages support each nasopalatine duct laterally, and line the posterior margins of each incisive foramen (e.g., slice 32.1.2; fig. 27). These cartilages are not connected to other nasal floor or septal cartilages. The palatine papilla of Micropotamogale is dorsoventrally elongate, defined by each nasopalatine duct laterally (slice 32.1.2; fig. 27).

Anterior to the incisive foramina and dorsal to the premaxilla, the anterior transverse lamina is evident (slice 21.3.2) and shows no connection with paraseptal cartilages.

Immediately anterior to the premaxilla, the nasal cartilages form an uninterrupted ring, or zona annularis, connecting the nasal sidewall, floor, and roof (slice 17.3.1).

As in Potamogale and in concordance with the observation that a lacrimal foramen is reduced in adult Micropotamogale skulls (Asher, 1999), Micropotamogale shows no nasolacrimal duct. Both potamogalines do, however, possess a duct for the lateral nasal gland that travels from the posterolateral nasal fossa anteriorly, ventral to the nasoturbinate (e.g., slice 35.3.2; see fig. 26 for the same structure in Erinaceus).

Well anterior to the nasopalatine ducts and the posterior margin of the anterior transverse lamina, the superior and inferior septal ridges are apparent (slice 10.3.1). Neither pair of ridges shows cartilaginous support.

The superior alar processes of Micropotamogale are smaller than those of Potamogale and do not extend far posterior or lateral to the nasal cupula. They originate posteriorly from the lateral margin of the anterior nasal floor (slice 7.1.2). No dorsal projections are evident on the anterior cartilaginous nose in the available adult specimens (IZEA 939), but they are evident in embryos (AIG 2273–1, 2273–3).

Geogale aurita (MCZ 45499; fig. 28)

The vomeronasal organ of Geogale is small compared to that of Potamogale Its anteriormost point is directly over the posterior margin of the palatine papilla. Instead of traversing a large portion of the nasal floor as in Potamogale the vomeronasal organ tapers to a posterior terminus at about the same transverse plane as the third upper incisor—roughly 1 mm posterior to its starting point. In the postnatal MCZ 45499, it comprises about 10% of nasal capsule length (about 10 mm) and 8% of head length (14 mm). Proportions are similar in the larger ZIUT individual (SVL 43 mm): the 1.2-mm vomeronasal organ comprises 9% of nasal capsule length (14.3 mm) and 6% of head length (21 mm).

The vomeronasal organ itself is kidney-shaped and laterally concave. Blood vessels are conspicuous dorsolateral and ventromedial to the vomeronasal organ. The paraseptal cartilages are most evident ventral to the vomeronasal organ along its length and are reduced medially. Relative to other animals examined here, there is very little space between the vomeronasal organ on each side of the septum. In other taxa, a dorsally extending flange of the paraseptal cartilage contributes to the space separating each half of the paired vomeronasal organ, whereas in Geogale bony elements of the nasal septum are evident medial to the vomeronasal organ (compare figs. 24 and 29 or slices 14.2.4 of Potamogale and 8.5.5 of Geogale). This may be due to the advanced ontogenetic stage and consequent resorption of cartilage in the available Geogale specimens.

At the anterior margin of the vomeronasal organ of Geogale just dorsal to the nasopalatine duct, a small outer bar connects the ventrolateral and ventromedial components of the paraseptal cartilage, lying dorsal to the vomeronasal duct (e.g., slice 8.1.5; figs. 28, 29).

Posterior to the anterior margin of the first upper incisors and continuing through the length of the premaxilla, the cartilaginous nasal septum consists of a single anteroposteriorly running cylinder, surrounded by ossifications of the premaxilla (slice 11.2.3). In this region the septum lacks a connection to the nasal roof. This condition may simply be a result of resorption of cartilage in a relatively well-ossified postnatal individual, an interpretation supported by the observation that other adult specimens (e.g., IZEA 939 and AIG 1227 Micropotamogale and ZIUT Sorex) also lack a cartilaginous septum-roof connection in parts of the nasal fossa.

At its anterior margin, the vomeronasal organ gives off a well-defined vomeronasal duct (slice 8.2.3; fig. 29), which enters the nasopalatine canal dorsally as the latter opens into the nasal fossa proper. As in most other taxa described here, the nasopalatine canal is paired, opening on either side of a well-defined palatine papilla (slice 7.5.5). A very small nasopalatine duct cartilage is evident along the lateral margin of the incisive foramina (slice 8.5.1), which shows no connection to other nasal cartilages.

Within the palatine papilla is a cartilaginous body (slice 8.1.4; fig. 29), ending anteriorly in two points. Although ubiquitous among marsupials (Sánchez-Villagra, 2001), this papillary cartilage has been observed in relatively few placental taxa, such as Elephantulus and Tupaia (Wöhrmann-Repenning, 1984, 1987).

Just anterior to the rostral openings of the vomeronasal ducts is the inferior septal ridge, a laterally projecting flap of epithelial tissue, closely associated with the posterior margin of the anterior transverse lamina (slice 7.2.4). This structure extends ventrolaterally out from the posterior tip of the medial arm of the anterior transverse lamina, and supports both glandular and olfactory tissue.

The lack of prominent medial processes of the paraseptal cartilage contributes to the absence of a connection between it and the anterior transverse lamina. That is, the nasal cavity floor anterior to the vomeronasal duct is not broadly continuous with the paraseptal cartilage.

Geogale also displays a complete zona annularis, that is, continuity between the nasal septum, anterior transverse lamina, and nasal sidewall for a short distance immediately anterior to the first upper incisor (e.g., slice 4.4.3).

Geogale exhibits relatively large superior alar processes, located just posterior to the external nares, that serve as the site of origin of several facial muscles (e.g., slice 3.4.4). The alar processes are broadly connected to the nasal floor ventrally and show a narrow alicupular commissure anteriorly.

Microgale dobsoni (MPIH 1964/104)

The vomeronasal organ of Microgale is about 1.2 mm in length and extends posteriorly to a point close, but still anterior, to the first ethmoturbinal. It comprises roughly 14% of nasal fossa length (8.4 mm) and 8% of head length (15 mm). A large vein courses lateral to the vomeronasal organ on each side and forms the axis around which the organ has a kidney-shaped appearance (e.g., slice 280).

A J-shaped paraseptal cartilage supports the vomeronasal organ throughout its length, but at no point does the outer bar completely enclose the vomeronasal organ laterally. Anteriorly, the paired vomeronasal organ empties into the superior margin of each nasopalatine duct (slice 244; fig. 30). The nasopalatine canal is paired and connects the oral and nasal cavities on either side of the palatine papilla (slice 240). Lateral to the nasopalatine duct, an isolated nasopalatine duct cartilage is present (slice 244; fig. 30), showing no connection to other cartilages.

Immediately anterior to the nasopalatine duct, the inferior septal ridge is evident and is supported in part by the cartilage of the anterior transverse lamina (slice 220). The latter cartilage has a very narrow posterior connection to the paraseptal cartilage, dorsal to the frontal plane in which the nasopalatine ducts connect the oral and nasal cavities (slice 240).

Anterior to the premaxilla, a complete zona annularis is evident, connecting the nasal floor, sidewalls, and roof (slice 140). The superior alar cartilages of the external nose are evident in slice 80 and show a connection to the anterior nasal floor. Unlike the external nose of Potamogale (fig. 23A), no superiorly projecting cartilaginous spur is evident dorsally.

Echinops telfairi (ZIUT, HL 18 mm)

The posterior margin of the paraseptal cartilage in Echinops is coincident with the anterior margin of the first ethmoturbinal (slice 15.2.4); the vomeronasal organ itself appears a fraction of a millimeter farther anteriorly (slice 14.2.2). Rostrally, the vomeronasal organ terminates over the posterior incisive foramina. The vomeronasal organ (approximately 1.6 mm long) comprises 17% of nasal capsule length (9.5 mm) and 9% of head length (18 mm). Two venous channels are evident dorsolateral and ventromedial to the vomeronasal organ on each side (e.g., slice 11.3.4), each of which gives off several small branches toward the inferior part of the nasal septum.

The anterior paraseptal cartilage curves around the vomeronasal organ laterally (slice 10.5.4), but does not possess an outer bar passing dorsal to the vomeronasal organ or vomeronasal duct. The vomeronasal duct empties directly into the nasopalatine duct, well separated from the nasal cavity proper (slice 9.3.6). Farther anteriorly, the paired nasopalatine duct appears very atypical, as the conduits on each side coalesce with one another into a single channel that opens into the oral cavity (slice 8.3.6; fig. 31). As a result, the typical mammalian inverted mushroom-shaped palatine papilla is absent. This condition has been previously documented in Echinops and Setifer by Hofer (1982a). Cartilages supporting the nasopalatine duct are small; only a minor cartilaginous center is evident along the lateral margin of the incisive foramen. The paraseptal cartilage does not show a significant anterior continuity with the anterior transverse lamina (slice 9.5.5).

Just anterior to the incisive foramina and dorsal to part of the nasopalatine duct, anteroposteriorly running ridges on either side of the septum are evident (slice 9.2.5). The posterior margin of the anterior transverse lamina provides cartilaginous support for these inferior septal ridges. Farther anteriorly, bulges on each side of the nasal septum comprise the “superior” septal ridges, which have no cartilaginous support and contain glandular tissue (slice 5.1.5). Also in this region of the anterior nose, a complete zona annularis is evident.

Adjacent to the anterior termination of the nasolacrimal duct, the superior alar processes originate from the nasal sidewall and extend dorsolaterally (slice 3.6.4). No connection with the anterior nasal cupula is apparent.

Setifer setosus (ZIUT, celloidin-embedded individual)

The vomeronasal organ of this adult Setifer specimen extends approximately 5 mm posteriorly beyond the premaxilla, to the level of the first premolars (slice 75.2.1). Unfortunately, length estimates for the nasal fossa and head of this specimen are unavailable. One particularly large blood vessel lies lateral to the vomeronasal organ along most of its length; numerous other vessels of moderate caliber are also evident scattered throughout the organ (e.g., slice 64.1.1).

The paraseptal cartilage borders the vomeronasal organ ventrally and medially throughout its length. In some sections it extends lateral to the vomeronasal organ (e.g., slice 57.2.1), but the paraseptal cartilage does not give off an outer bar that passes dorsal to the vomeronasal organ or the vomeronasal duct.

At its anterior margin the vomeronasal organ grades into the vomeronasal duct, which enters the nasopalatine duct ventral to and separate from the nasal fossa (slice 48.1.1). The nasopalatine canal of Setifer leaves the nasal fossa paired; farther ventrally, however, Setifer resembles Echinops in that the canals on each side join with each other to form a single canal that opens into the oral cavity (slice 38.1.1; see also Hofer, 1982a). A very small nasopalatine duct cartilage is evident connected with the paraseptal cartilage (e.g., 48.3.1).

Anterior to the vomeronasal organ, the paraseptal cartilage ends, showing no continuity with the anterior transverse lamina. The posterior margin of the anterior transverse lamina supports the inferior septal ridge, extending out from the nasal septum, just dorsal and anterior to the nasal openings of the nasopalatine ducts (slice 47.2.1).

Anterior to the premaxilla, the nasal floor descends relative to the roof, greatly increasing the volume within the nasal cupula (compare slices 35.2.1 and 25.2.1). In this region, Setifer shows a broad continuity, or zona annularis, between the nasal septum, floor, and sidewall.

The superior alar processes are relatively narrow, showing only a short connection with the anterior nasal floor and no connection with the anterior nasal cupula. However, the alar processes do provide attachments for small muscles of the rhinarium (e.g., slice 13.2.1).

Tenrec ecaudatus (ZIUT, HL 20 mm; fig. 32)

Posteriorly, the vomeronasal organ and paraseptal cartilages extend to the upper canines. The vomeronasal organ itself measures approximately 1.5 mm in length. This makes up 14% of nasal fossa length (10.2 mm) and 7.5% of head length (20 mm). Several small blood vessels are apparent running parallel to the vomeronasal organ.

The paraseptal cartilage lies medial to the vomeronasal organ along its entire length and does not exhibit a complete outer bar laterally (fig. 32). The vomeronasal ducts are narrow, anterior continuations of each vomeronasal organ tube, emptying into the nasopalatine canal at a point well-separated from the nasal fossa (slice 7.5.1). A large palatine papilla is evident medial to the nasopalatine ducts as they open into the oral cavity.

A nasopalatine duct cartilage is present lateral to the nasopalatine ducts and is connected to both the paraseptal cartilage dorsomedially and the “nasopalatine duct” cartilage medially (slice 8.2.4; fig. 33). Kuhn (1971: 30) noted that confusion exists regarding the definitions of the “palatine” and “nasopalatine duct” cartilages; in some taxa (including Tenrec), the two are continuous anteriorly. As defined in previous studies (e.g., Maier, 1980), they are at least distinguishable posteriorly; the palatine cartilage exists lateral, and the nasopalatine duct cartilage medial, to the nasopalatine duct itself.

Anterior to the nasopalatine canal is the inferior septal ridge, supported medially by the superior arm of the anterior transverse lamina and containing olfactory epithelium (slice 6.1.1). The medial arm of the anteriormost paraseptal cartilage has a narrow connection with the anterior transverse lamina. The connection between the nasal septum, floor, and sidewall in Tenrec is constricted, but present, at the point just anterior to the premaxilla and where the nasolacrimal duct enters the nasal fossa (slice 4.4.5).

A feature of the external nose that Tenrec shares with Setifer is the abrupt inferior extension of the anterior nasal floor, greatly enlarging the surface area available for olfactory epithelium in the anterior nose (slice 5.3.1). In Tenrec the narrow superior alar processes originate from this ventrally extended nasal floor, and show no commissura alicupularis connecting them to the anterior nasal cupulae.

Solenodon paradoxus (ZIUT)

Despite the enormous length of the Solenodon anterior nose, the vomeronasal organ is not elaborate. The vomeronasal organ is in fact quite small, failing to extend farther posteriorly than the row of upper incisors. Unfortunately, the available Solenodon specimen is only partially sectioned through the midrostrum, leaving no means of precisely inferring the nasal capsule or head length in this specimen. However, it is likely that the vomeronasal organ of Solenodon comprises a very small percentage of nasal capsule and head length, given the great size of the latter two structures in this genus. No major blood vessels are evident coursing adjacent to the vomeronasal organ.

The paraseptal cartilages are relatively small J-shaped structures that, as in most other mammals described here, course medially along most of the length of the vomeronasal organ (e.g., slice 110.3.1). At its anterior end, Solenodon shows a small outer bar which does not enclose the vomeronasal organ laterally, but ends in a free process. At the anterior margin of the vomeronasal organ, the vomeronasal duct opens ventrally, medial to the inferior septal ridge, and empties directly into the nasal fossa (slice 100.3.1; fig. 34) anterior to the superior margin of the nasopalatine duct (slice 106.3.2). Unlike other animals described here, and like some “archaeorhinate” taxa described by Broom (1897), the nasopalatine and vomeronasal ducts do not directly communicate (see also Menzel, 1979; Hofer, 1982b). However, Solenodon differs from Broom's “archaeorhinates” in possessing cartilaginous support for the nasopalatine duct. A connection is evident between the inferior arm of the paraseptal cartilage and the cartilage supporting the nasopalatine duct (slice 101.2.1; fig. 25). The latter cartilage is present at the superior margin of the nasopalatine duct, and extends lateral to the duct to the lateral rim of the bony incisive foramina. The nasopalatine ducts run posteroinferiorly to empty into the oral cavity adjacent to the palatine papilla (slice 108.2.2).

The relatively tiny medial arm of the paraseptal cartilage has no connection with the anterior transverse lamina, which lies far anterior to the vomeronasal organ and its cartilages. Although the nasal septum and floor are continuous throughout most of the elongate external nose of Solenodon the connection between the nasal floor and sidewall is never complete, as Solenodon shows an anteroposteriorly running suture between nasal floor and roof adjacent to the nasolacrimal duct.

The superior alar processes of Solenodon are relatively small, consisting of an extended nasal sidewall that ends freely, with no connection to the anterior nasal cupula (e.g., slice 11.1.3).

Erinaceus europaeus (ZIUT, SVL 37 mm)

Posteriorly, the vomeronasal organ begins at a point slightly posterior to the anterior tip of the first ethmoturbinal (e.g., slice 21.3.2) and ends 1.6 mm anteriorly, over the incisive foramina. The vomeronasal organ comprises about 22% of nasal capsule length (7.3 mm) and 8% of head length (19 mm). No major blood vessels accompany the vomeronasal organ along its length.

The paraseptal cartilages are J-shaped (e.g., slice 17.2.3). Anteriorly, the vomeronasal duct is completely enclosed by the outer bar (slice 15.1.4), which begins posteriorly from the inferior arm of the paraseptal cartilage, runs dorsally over the vomeronasal duct, and joins the medial arm of the paraseptal cartilage as the duct empties into the nasopalatine canal. Erinaceus displays a paired vomeronasal duct, leaving the vomeronasal organ anteriorly and connecting directly into the nasopalatine duct (slice 14.5.1). The paired nasopalatine duct empties into the mouth on either side of a well-defined palatine papilla (e.g., slice 14.3.3; fig. 26). A small nasopalatine duct cartilage, continuous with the inferior portion of the paraseptal cartilage, is evident medially and anterolaterally to the nasopalatine canal (slice 14.2.4; fig. 26).

Anterior to the nasopalatine ducts, both inferior and superior septal ridges are apparent, without cartilaginous support (slice 12.5.2). As in Potamogale the medial processes of the paraseptal cartilages are continuous anteriorly with the anterior transverse lamina. This lamina is continuous with the nasal sidewall, roof, and septum throughout a large portion of the external nose anterior to the premaxilla (e.g., slice 9.2.3).

The superior alar processes show most continuity with the anterior floor of the nasal capsule (slice 9.1.4); anteriorly, they bifurcate around the external nares.

List of Nasal Characters

Several morphological differences are evident in specimens of different age. For example, like other cranial sense organs (e.g., eyes), the vomeronasal organ grows with negative allometry relative to body mass and is proportionately larger in younger specimens. Thus, a fetal Micropotamogale specimen (AIG 22733–3) with a head length of about 9 mm shows a posterior margin of the vomeronasal organ overlapping the anterior extent of the first ethmoturbinial. The vomeronasal organ in an adult Micropotamogale (IZEA 939) with a head length of about 50 mm, on the other hand, ends posteriorly near the upper canines, as described above. Even so, the vomeronasal organ in an embryo of Potamogale is considerably longer than that of similarly aged Micropotamogale, Echinops, Tenrec and Microgale individuals examined in this study.

Adult specimens (e.g., ZIUT Sorex AIG 1227 and IZEA 939 Micropotamogale) show resorption of certain regions of nasal cartilage not evident in fetal representatives of these taxa. This includes the medial wing of the paraseptal cartilage, the connection between the nasal septum and roof, and the connection between the paraseptal cartilage and anterior transverse lamina. The connection between the paraseptal and nasopalatine duct cartilages appears less ontogenetically variable, as both structures persist intact postnatally, and two relatively older specimens (Setifer and Solenodon) with erupted deciduous teeth show a connection between the two. Finally, as described above, tubercle-like “spurs” on the dorsum of the nasal cupulae are evident in the fetal Potamogale (fig. 23A) and two fetal Micropotamogale specimens. However, these are absent in the two available adult Micropotamogale specimens, and must therefore be presumed to disappear during the course of ontogeny.

There is of course nothing wrong with attempting to extract phylogenetic information from characters that are expressed only during a certain point in ontogeny. The problem arises when a given sample contains taxa representing different ontogenetic stages. Such is presently the case; therefore, I attempt to use characters that are consistently expressed in individuals of various ages. The remaining morphological differences described above are constant in those taxa represented in this study by adult and fetal specimens (table 2). This enables comparisons between taxa represented by only postnatal (Geogale) or fetal (Potamogale) specimens.

Some of the morphological variation described above (i.e., relations and cartilaginous support for the inferior septal ridge and connections of the superior alar cartilage and nasal cupulae) may provide valuable character information for phylogenetic purposes. However, based on the present sample, it is difficult to unambiguously refine the observed morphology into discrete character states. Hence, these characters are for present purposes not considered in the following pages.

Table 4 includes character data for several taxa that are not described above. This information is based on observations of specimens listed in table 2, as well as the work of authors cited below. Specifically, data on nasal structure in Bradypus is taken from Schneider (1955); in Dasypus from Broom (1897) and Reinbach (1952a, 1952b); in Echinosorex from Broom (1915a) and Wöhrmann-Repenning (1984); in Orycteropus from Broom (1909) and W. Maier (personal commun.); in Procavia from Broom (1898), Lindahl (1948), and Weisser (1992); in Talpa from Fischer (1901) and Broom (1915b); and in Tupaia from Broom (1915a), Maier (1980), and Wöhrmann-Repenning (1984). Not all characters relevant to this study are described by these authors; hence, where information on a given character is lacking, I code it as “missing”.

The following characters are numbered after the nine arterial characters described above, so as to correspond with the matrix presented in table 4 (see also table 3). As detailed previously, a “0” character state assignment does not necessarily imply primitiveness.

10. Anterior relation of vomeronasal duct: Solenodon and Dasypus differ from other animals examined here in possessing a vomeronasal duct that empties directly into the nasal fossa (state 1) rather than into the nasopalatine duct (e.g., Tenrec state 0).

11. Paraseptal cartilage and anterior vomeronasal organ (“outer bar”): Among marsupials, a strut of paraseptal cartilage consistently encloses the vomeronasal organ anterolaterally (e.g., Didelphis state 0; Sánchez-Villagra, 2001). This is variably present among placental mammals; in many taxa the paraseptal cartilage is laterally open along its anterior half (e.g., Micropotamogale state 1).

12. Oral opening of nasopalatine duct: The nasopalatine duct opens into the oral region via a single, unpaired channel in Echinops and Setifer (state 1; Hofer, 1982a). In other taxa, the nasopalatine duct is paired and opens into the mouth on either side of the palatine papilla (e.g., Tenrec state 0).

13. Presence of nasopalatine duct cartilage: Broom (e.g., 1898) noted that marsupials consistently lack cartilaginous support for the nasopalatine duct as it passes ventrally from the nasal fossa into the oral cavity (e.g., Didelphis state 0). Some placentals, on the other hand, often show a cartilage medial and/or lateral to the nasopalatine duct (e.g., Tenrec state 1). Many authors assign the medial (nasopalatine) and lateral (palatine) components different names (e.g., Maier, 1980); however, as mentioned above, Kuhn (1971) noted that these cartilages may be continuous and difficult to distinguish from one another. Due to this ambiguity, all cartilaginous support for the nasopalatine duct is here labeled “nasopalatine duct cartilage”.

14. Connection of nasopalatine duct and paraseptal cartilages: When present, the nasopalatine duct cartilage may show a connection to the paraseptal cartilage posterior to the nasopalatine duct (e.g., Tenrec state 1). Alternatively, it may appear adjacent to the duct with no connection to the paraseptal cartilage (e.g., Microgale state 0).

15. Presence of nasolacrimal duct: Most taxa show a glandular duct running from the anteromedial margin of the eye, along the sidewall of the cartilaginous nose ventral to the maxilloturbinal, and opening anteriorly within the nasal cupula adjacent to the external nares (e.g., Tenrec state 0). Potamogalines, on the other hand, show no trace of any such duct (state 1).

16. Lateral cover of nasolacrimal duct: The extent to which the nasolacrimal duct is shielded laterally by cartilage varies (W. Maier, personal commun.). In Echinops there is no cartilaginous barrier lateral to the duct at any point except for its anterior extreme, when it opens into the nasal cupula (state 0). Erinaceus on the other hand, shows a lateral extension of the anterior transverse lamina dorsal to the anterior margin of the premaxilla that shields the nasolacrimal duct laterally, well before it empties anteriorly into the nasal cupula (state 1).

17. Papillary cartilage: Some taxa exhibit a cartilaginous structure within the palatine papilla (e.g., Geogale state 0). Broom (1896) and Sánchez-Villagra (2001) noted the presence of a papillary cartilage in most marsupials; however, such a structure is lacking in a Didelphis individual examined in this study (table 2) and in those examined by Wöhrmann-Repenning (1984), although it is present in others (Sánchez-Villagra, personal commun.). Most placentals lack a papillary cartilage (e.g., Tenrec state 1).

18. Vomeronasal organ blood vessels: In several taxa, a prominent blood vessel travels anteroposteriorly along with the vomeronasal organ tucked into its lateral side, giving the vomeronasal organ a kidney-shaped appearance when viewed coronally (e.g., Didelphis state 0). Other animals show blood vessels scattered throughout the vomeronasal epithelium that do not travel in any single, well-defined fossa in the vomeronasal organ (e.g., Potamogale state 1).

19. Shape of nasal septum: Sánchez-Villagra (2001) noted that a parallel-sided nasal septum (when viewed coronally) is ubiquitous among marsupials (e.g., Didelphis state 0), and contrasts with the ventrally ovoid septum present in many placental mammals (e.g., Echinops state 1).

20. Zona annularis: Anterior to the premaxilla, the nasal cartilages may form a complete, uninterrupted ring of cartilage around the nasal fossa, joining the anterior transverse lamina with the nasal sidewall and roof (e.g., Echinops state 0). Alternatively, these structures may not be fully continuous at any point anterior to the premaxilla (e.g., Solenodon state 1).

Differences Between Trees Based on Soft-Tissue Dataset

Figure 35 shows the soft-tissue characters described in the previous section optimized on these trees. The phylogeny of Asher (1999: set no. 1) requires 43 steps (CI = 0.55) and that of Stanhope et al. (1998: fig. 1) 50 steps (CI = 0.45). A parametric significance test of this difference (Kishino and Hasegawa, 1989; calculated using PAUP* version 4.0b8, Swofford, 2001) yields a two-tailed P-value of 0.0692. Similarly, a nonparametric Wilcoxon Rank Sum test (Templeton, 1983; also calculated using PAUP* version 4.0b8) yields a T-value of 2; given n = 6 degrees of freedom this value indicates two-tailed significance between alpha levels 0.05 and 0.10 (Rohlf and Sokal, 1995: 136). Two-tailed statistical tests are appropriate since both of these trees were selected a priori, with no necessary expectation that one would better explain the soft-tissue dataset than the other (Swofford et al., 1996: 505). Hence, the phylogeny of Asher (1999) better explains the soft-tissue data discussed above than that of Stanhope et al. (1998), but without a decisive level of significance.

Possible Evolution of Cranial Characters Among Insectivoran-Grade Mammals

In order to better explore the possible evolution of the soft-tissue characters described above, I have mapped them (fig. 36) onto a more taxonomically complete tree produced by the osteological and dental dataset of Asher (1999: set no. 1). If this topology is correct regarding, for example, the placement of Solenodon and monophyly of the Lipotyphla, then a number of previously underemphasized morphological synapomorphies are worth discussing.

Affinities of Solenodon

First, contrary to the view outlined by McDowell (1958) that evidence for a tenrec-Solenodon clade was restricted to the dentition, one of the most unique soft-tissue character transformations recorded during the course of this study was observed only among most tenrecs and Solenodon In these taxa, the external carotid artery runs medial to a slip of the thyropharyngeus muscle, immediately prior to the origin of the artery's facial and lingual branches (character 9, state 0; figs. 14, 16, 18, and 20). In most other mammals, including potamogalines and Hemicentetes the external carotid remains lateral to the thyropharyngeus muscle throughout its course (fig. 12).

As reconstructed assuming accelerated transformations (Swofford and Begle, 1993: 20) on set no. 1 of Asher (1999), the node containing Solenodon plus tenrecs and Chrysochloris shows the medial course of the external carotid relative to thyropharyngeus to be a synapomorphy. It is equally parsimonious to reconstruct this character state as evolving in parallel in Solenodon and certain tenrecs, depending on the nature of the unresolved Geogale-Microgale-potamogaline node. In any event, this character varies within tenrecs; as noted, Hemicentetes and potamogalines show the putatively primitive state with the external carotid consistently lateral to the constrictor musculature throughout the artery's course (fig. 12). Chrysochloris shows this more general condition as well.

Another notable character shared by Solenodon and at least one tenrec (Geogale) is the supratympanic bifurcation of the stapedial artery (character 7, state 1). Figure 22A shows the undivided proximal stapedial artery of Solenodon passing through the tympanic roof via a foramen in the petrosal, anterior to the promontory, medial to the epitympanic recess, and lateral to the origin of the tensor tympani muscle. Figure 22B shows a slightly more anterior slice in which the intracranial proximal stapedial divides into superior and inferior rami, above the tegmen tympani's anterior margin. After its origin superior to the petrosal, the inferior stapedial leaves the cranial cavity anteroventrally through the piriform fenestra. The path and bifurcation of the stapedial artery in Solenodon closely resemble those discussed above for Geogale but have not been observed in any other mammal, including tenrecs, examined during the course of this study. All topologies recovered by Asher (1999) optimize this character state as evolving homoplastically in the two taxa.

Lipotyphlan Monophyly

Several arterial characters traditionally regarded as primitive for placental mammals (fig. 3), and thus of allegedly no value for recognizing monophyletic taxa within placentals, are in fact optimized as synapomorphies for lipotyphlans in figures 35A and 36. Of the taxa included in figure 35, only Erinaceus, Sorex, Talpa, Solenodon most tenrecs, and Chrysochloris possess patent posterior, superior, and inferior rami of the stapedial artery (characters nos. 1–3, state 0), in addition to a ramus mandibularis that arises from the inferior stapedial ramus (character 5, state 0). However, the relatively high consistency of these characters is somewhat artefactual, as it is known that other mammals not discussed here (e.g., tree shrews, some rodents, rabbits, bats, and elephant shrews), and one that is (Orycteropus), possess one or both stapedial rami with their potential tributaries (Bugge, 1974; MacPhee, 1981; Wible, 1987). Similarly, the dense taxon sample of tenrecs depicted in figure 36 indicates considerable homoplasy, as Potamogale shows a reduced superior stapedial ramus, both Tenrec and Hemicentetes lack an inferior ramus, and adult specimens of Tenrec, Sorex, Erinaceus and Chrysochloris lack patent posterior stapedial rami.

Nevertheless, support for the Lipotyphla from these allegedly primitive characters does not disappear completely with the recognition that they occur in other parts of the mammalian tree. “Primitive” and “derived” are relative terms, and the appearance of a certain character state at a basal node in any phylogeny does not rule out the possibility that subsequent recurrences of that state are derived for a different group. Hence, patent inferior, superior, and posterior stapedial rami (for example) would fail to comprise synapomorphies for a monophyletic Lipotyphla only if its immediate sister group(s) also possessed these characters, and this has yet to be demonstrated.

Morphological Support for Afrotheria

Osteological or dental support for an Afrotheria including tenrecs and golden moles is lacking (Asher, 1999). However, two soft-tissue characters present interesting distributions that could be consistent with monophyly of the Afrotheria. First, golden moles and Orycteropus both have an internal carotid artery that enters into the braincase lateral to the anterior pole of the cochlea (character 6, state 1; fig. 21). Although data from elephant shrews have not been incorporated into this study, this character state has also been observed in serial sections of Elephantulus at the Zoologisches Institut, Tübingen. In contrast, tenrecs and Procavia show the more widely distributed condition of an internal carotid artery entering the braincase medial to the anterior pole of the cochlea (fig. 16; see also Lindahl and Lundberg, 1946), indicating that if this character is a synapomorphy, it either does not define Afrotheria as a whole or that some homoplasy has occurred during its evolution. When considering a larger mammalian sample, the latter seems to be the case, as some shrews (Blarina and Crocidura) and the carnivoran Nandinia have been observed to possess an internal carotid artery lateral to the anterior cochlea as it enters the braincase (Asher, 2000).

A second potential Afrotherian morphological synapomorphy concerns vomeronasal organ vasculature. Procavia, Chrysochloris potamogalines, and Setifer show small vomeronasal blood vessels dispersed throughout the organ (character 18, state 1); this contrasts with the more typical condition of a single, large vessel running lateral to the organ and forming an axis around which the organ takes a kidney-shape. Most other taxa observed during this study show the latter character state. However, like certain African taxa, and again indicative of homoplasy, Solenodon, Canis and Crocidura show the former state.