Modern phylogenetics place the Soricidae (shrews) into the order Lipotyphla, which belongs to the relatively new superorder clade Laurasiatheria. Their most derived skull feature is the unusual position and shape of the jaw articulation: Whereas in all other mammals the glenoid region of the squamosum is more or less tightly attached to the otic capsule or petrosal, respectively, in the soricids it is attached to the nasal capsule. This new position of the jaw articulation becomes possible by the posterior extension of the nasal capsule and the rostral shift of the glenoid fossa. By the study of dated postnatal ontogenetic stages of Crocidura russula and Sorex araneus, we show that the glenoid part of the squamosal becomes fixed to the nasal capsule by the ossified alae orbitalis and temporalis. The ala orbitalis is displaced laterally by the expanded cupula nasi posterior; this posterior expansion is well documented by the lamina terminalis, which incorporates parts of the palatinum and alisphenoid. Both alae consist largely of ‘Zuwachsknochen’ (‘appositional bone’) and are then named orbitosphenoid and alisphenoid. By the forward move of the pars glenoidea and of the alisphenoid, the foramen lacerum medium (‘fenestra piriformis’) also expands rostrally. Functionally, the forward shift of the jaw joint helps to keep the incisal biting force high. Biomechanically the jaws can be considered as a tweezer, and the rostral position of the jaw joints makes the interorbital pillar and the shell-like walls of the facial skull a lever for the highly specialized incisal dentition.

Crocidura russula, cupula nasi posterior, foramen lacerum medium, ontogeny, orbitosphenoid, Sorex araneus, squamoso-dental articulation

The morphology of the skull of Soricidae (shrews) is not well known. This is not too surprising, because the skulls are very small and are almost completely synostosed already in subadult stages. As will be demonstrated in this study, two distinctive features have been escaped notice so far: First, the progressive expansion of the posterior nasal capsule, which reaches the origin of the alae temporales; second, the rostral shift of the glenoid fossa and the mandibular joint. This results in the very peculiar position of the glenoid fossa of the squamosum to be attached at the sides of the nasal capsule (Fig. 1). In other mammals, the glenoid region of the squamosal is more or less closely attached to the petrosal (or parts of it such as the tegmen tympani) sometimes forming the compound os temporale (Mischknochen) as in Primates (Starck 1979).

Figure 1. 

A Virtual picture of the skull of an adult specimen of Sorex araneus (SMF 82598). The roof of the braincase is removed to show the internal cranial base. B This semi-transparent figure (pseudo x-ray) of the same skull shows that the glenoid fossa (vertical arrow) appears to be attached to the posterior part of the nasal capsule. The recessus ethmoturbinalis and the posterior cupula of the nasal capsule reach back to the carotid foramina (horizontal arrow). C Virtual cross section (scan no. 1205, at about the level of the vertical arrow in C) showing the fixation of the glenoid portion of the squamosal to the lateral wall of the nasal capsule. The lamina terminalis is broken in this specimen. D Histological section of an adult specimen of Sorex araneus (Coll. W. Maier) at about the same plane as the µCT scan. d1 32—3-3 means section series of neonate specimen: plate 32, row 3, and section 3. The fissura orbitalis superior is almost completely occupied by the ramus maxillaris of the trigeminal nerve and by the thin nervus opticus in its mediodorsal corner. The anterior notch of the foramen rotundum contains the ramus mandibularis of the trigeminal nerve. Note the pronounced dorsal protrusion of the posterior nasal capsule into the brain cavity and the peculiar structure of the jaw joint. The cross sections C and D are mirrored. Abbreviations: ali – alisphenoid, bol – bulbus olfactorius, dar – discus articularis, dnp – ductus nasopharyngeus, ett 3 – ethmoturbinal 3, fos – fissura orbitalis superior, fov – foramen ovale, lcr – lamina cribrosa, lt – lamina terminalis, nm – nervus maxillaris, osph – orbitosphenoid, pam – processus articularis mandiblae, pgi – processus glenoideus inferior, sn – septum nasi, squ – squamosal.

From this unique feature arises the question whether the nasal capsule expanded actively, and second, whether the glenoid fossa moved progressively from the otic to the nasal region – or whether both structural complexes changed. The peculiar relationship of the glenoid facet on the squamosal with the nasal capsule needs further explanation, combined with the question about the functional meaning of these complicated modifications of the soricid skull.

Such morphological analyses can only be performed with the help of suitable ontogenetic stages. Fortunately, we have available dated sectional series of both Sorex and Crocidura of the first two postnatal weeks of development. We owe these sectional series to the late Peter Vogel, who had prepared them himself (Vogel 1973). This invaluable material was generously presented to the first author many years ago. As we will show, the shrews are quite immature at birth, and important processes of differentiation take place within the first two postnatal weeks.

In the past, Parker (1885) studied two late fetal stages, a 10 to 12 days old postnatal stage, and an adult specimen of ‘Sorex vulgaris’ (= Sorex araneus). Late fetal stages of Sorex araneus based on serial sections were studied by DeBeer (1929). Very useful information on cranial anatomy of various soricids stem from Roux (1947) and Gaughran (1954). Masuda and Yohro (1994) analyzed the fetal skull of Suncus murinus by means of the staining and clearing technique. Vogel (1973) records several ossification sequences in postnatal stages of his soricids. More recently, Giere (2002) studied the orbitotemporal region of various ‘insectivores’, among them subadult and adult shrews; however, he did not systematically make use of ontogenetic stages.

While Crocidura is more derived in the reduction of the antemolar dentition, it seems to be somewhat more plesiomorphic in certain skull features. Crocidura russula has a pregnancy length of 30 days and Sorex araneus 20 days; therefore, Crocidura is born in a significantly more mature stage representing a derived pattern according to Vogel (1973), but we think it is the plesiomorphic one (see discussion). Therefore, in the following anatomical presentations, we always begin with the skull features of Crocidura.

Among the different components of the mammalian skull, the ear region has attracted by far the most interest. In recent years, with µCT technique available, the nasal region has won increasing importance (e.g., Van Valkenburg et al. 2014; Ruf 2014; Ruf et al. 2015; Martinez et al. 2020). The orbitotemporal region is somewhat more problematical, because it consists of heterogeneous components and it overlaps with the other two regions. First of all, it is usually defined by the orbit and the size of the eyes, but this aspect is less relevant for soricids due to their greatly reduced eyeballs. The medial wall of the orbit has been considered informative in systematical regard (cf. Novacek 1980; Giere 2002). All of these parts of the head skeleton house important sense organs (smell, vision, hearing and equilibrium (cf. Demes 1981; Preuschoft 1989; Witzel 2020). Reduction of the eyes comprises a natural experiment which, we assume, may better illuminate the mechanical role of the orbitotemporal pillar (Maier 1993a; Ross 2001).

We studied neonate to adult stages of Sorex araneus and Crocidura russula with histological serial sections and µCT scans; adult stages of further Soricidae were used for comparison. Details on the specimens, collections and scan parameters are presented in Table 1.

Histological serial sections of defined early postnatal age stages of Sorex araneus and Crocidura russula come from the collection of the late Peter Vogel (Lausanne, Switzerland). They were prepared in the 1960s by P. Vogel for his dissertation under the supervision of Prof. Adolf Portmann at the University of Basel, Switzerland. We assume that the thickness of the sections is 10 µm; the staining varies between ‘Azan’ and ‘Ladewig’. Histological serial sections of adult stages of Sorex araneus and Crocidura russula were prepared in the lab of the former Department of Zoology at the Eberhard-Karls-Universität Tübingen, Germany. All series comprise transversal sections except for a young adult specimen of Crocidura russula (Coll. W. Maier No. 2) that has been sectioned sagittally.

Mazerated skulls of adult soricids were scanned with the µCT scanner (Fraunhofer/ProConXray/Feinfocus) housed at the Senckenberg Forschungsinstitut und Naturmuseum Frankfurt, Frankfurt am Main, Germany. One wet specimen of Crocidura russula (Coll. W. Maier) with an articulated lower jaw was scanned with a Nikon XT H 320 at the Senckenberg Centre for Human Evolution and Palaeoenvironment (HEP), Eberhard-Karls-Universität Tübingen, Germany. Based on the µCT scans, virtual 3D models were automatically rendered and processed in Avizo 9.01 (Thermo Fisher Scientific FEI); the same software was used to produce pseudo X-ray images of the skulls based on the rendered 3D models.

The present study is not comparative in a strict sense, because meaningful comparisons have to begin with at least three taxa, whose phylogenetic systematic position is defined as precisely as possible. However, our two taxa Crocidura russula and Sorex araneus can at least be attributed to the two well established subfamilies of the Soricidae, the Crocidurinae and Soricinae, which are probably separated since the Eocene (Douady and Douzery 2003). More elaborate comparisons have not been possible at present, because serial sections of defined postnatal age have only been available from the two common shrews just mentioned. µCT scans of adult skulls of a few other soricids (Suncus, Myosorex, Anourosorex, Neomys) have been prepared for comparison of a few osteological features of adults. Hence, this study is only meant as an initial foray into the skull morphology of extant soricids and a basis for future studies of soricid craniology and systematics. Nevertheless, a simplified cladogram should help to understand the systematic outlines of this study (Fig. 19).

Figure 19. 

Simplified cladogramm of the Lipotyphla and Soricidae underlying this study. The relationship of the Soricidae to either the Talpidae or the Erinaceidae is still disputed in the literature (simplified from Dubey et al. 2007; Myosorex is ranged only at the generic level).

The phylogenetic systematics of mammals has experienced some dramatic changes during the last 25 years (Douady et al. 2002). One of the major insights was the invalidity of the old category Insectivora (Wagner 1855), which had already been suspected to be a waste basket. The Tenrecidae, Chrysochloridae, and Macroscelididae are now considered to be a member of the Afrotheria , and the remaining soricids, talpids, erinaceids and solenodontids are united in a monophylum Lipotyphla (Asher and Helgen 2010). The relationship of the Soricidae to either the Erinaceidae or the Talpidae is also disputed in the literature. Most cladistic studies based on molecular data favour erinaceids as sistergroup of shrews (Liu et al. 2001; Douady et al. 2002; Dubey et al. 2007, 2008; Bininda-Emonds et al. 2007; O’Leary et al. 2013) to name but a few. Especially Hylomys and Uropsilus need further study before morphology can contribute to the open systematic problems.

What is even more relevant for our study is the place of the Lipotyphla within the mammalian tree. Already the first published phylogenetic trees based on molecular data showed that the Lipotyphla are not a basal offshoot of eutherian mammals, but that they nest within the superorder Laurasiatheria – or are their sistergroup; more basal branches of the Placentalia are either the Xenarthra or the Afrotheria (e.g., Springer 1997; Stanhope et al. 1998; Mouchaty 1999; Madsen et al. 2001; van Dijk et al. 2001). These early studies have been essentially confirmed by all later publications. Contrary to the older studies based on morphology alone, which considered ‘insectivorans’ as models for basal eutherian mammals, the new phylogenetic data rather suggest that an ‘Insectivora--type’ of mammals evolved several times convergently. Of course, this conclusion does not preclude that they may resemble in some way an archaic adaptation-type (‘Lebensform-Typus’) (Weber 1904; Simpson 1945; Thenius 1979; Starck 1995). This adaptation-type can be characterized as small carnivore searching its animal food in the litter and undergrowth of forests and bushes. Among marsupials, tiny shrew-like species occur several times both in the Neotropic and Australian realms (Nowak 1999).

The most striking peculiarity of the cranial base is the different proportioning of some of its skeletal elements. Figure 3 shows schematic dorsal views of chondrocrania of three lipotyphlans. They clearly demonstrate that in Sorex the posterior part of the nasal capsule has expanded caudally. This progressive growth is most clearly shown by its relationships with the ala orbitalis: The narrow ala orbitalis is inclined rostrolaterally before it is fused with the taenia marginalis, which represents the primary sidewall of the braincase in the chondrocranium. This secondary expansion of the cupula nasi posterior almost completely pushes aside the roots of the ala orbitalis from the trabecular plate or septum nasi respectively. The cupula nasi posterior lies slightly above the level of the ala orbitalis. In the adult skull, the cupula nasi posterior projects into the cranial cavitiy, and it is medially fused with the sphenethmoid. This connection is somewhat different in Crocidura and Sorex (Figs 5, 8).

The morphology of the ala orbitalis of soricids is only properly understood when the early postnatal stages are considered. At this developemental period, the original cartilaginous structures are largely covered and replaced by ‘Zuwachsknochen’, which originates at the roots of the ala orbitalis (Figs 6, 7). Endochondral ossifications of the ala orbitalis are of minor importance and fuse with the ‘Zuwachsknochen’. Therefore, the orbitosphenoid of the juvenile and adult skulls consists mainly of ‘Zuwachsknochen’. There exist minor differences in the mode of ossification between Crocidura and Sorex (see Fig. 8).

Because it reaches so far posteriorly, the roots of the alisphenoid appear to originate from the lower end of the cupula nasi posterior in the juvenile and adult skull. The transitory connection of the pila metoptica with the trabecular plate may be considered as an example of ‘recapitulation’, but within the first postnatal days the ’Zuwachsknochen’ of the ala orbitalis fuses indistinguishingly with the remnants of the cupula nasi posterior. Generally, the origin of the ala orbitalis by two slender pilae probably has to be considered as a weak point of the biomechanical construction of the posterior facial skull, and therefore it seems plausible to assume that mechanical strengthening by ‘Zuwachsknochen’ begins in all mammals right at these pilae.

The progressive elongation of the pars posterior of the nasal capsule is also testified by the secondary elongation of the lamina terminalis. This horizontal septum, which separates the posterior recessus ethmoturbinalis from the nasopharyngeal duct, usually consists of the lamina transversalis posterior of the nasal capsule and the horizontal process of the vomer (Ruf et al. 2015, 2020; Ruf 2020). However, in soricids the lamina terminalis additionally includes the lamina transversalis posterior, the vomer, the ascending lamina of the palatine and the complete floor of the cupula nasi posterior. In the adult osteocranium, the heterogeneous components of the lamina terminalis fuse early into a thin and smooth bony plate, which forms an oblique roof and sidewall of the nasopharyngeal duct (cf. Fig. 2). This air passage of soricids is certainly adapted to high airstream velocities connected with the breathing frequencies up to 120 per minute (Flindt 2000).

The squamosum is relatively narrow dorsoventrally but stretched rostrocaudally. Near its anterior end it forms the fossa glenoidea with its prominent processus glenoideus inferior. The Nomina Anatomica officially uses fossa mandibularis, but Gegenbaur (1892), McDowell (1958) and Dötsch (1982) use cavitas or fossa glenoidea, and we adopt this term. The connections of the anterior end of the squamosum do not become clear in juveniles and adults, because the sutures are all fused. In the fetus of Sorex araneus, as shown by the plate reconstructions of DeBeer (1929), the slender squamosum does not yet reach the orbital process of the frontal bone, and the contact with the mandibular head does not yet exist either; however, his cross-sections (DeBeer 1929: figs 19-21) show that the differentiation of the secondary jaw articulation has already begun. In adult skulls of both Crocidura and Sorex, the dorsal suture with the parietal is retained, showing that in Crocidura the squamosum appears to be somewhat broader than in Sorex (cf. Fig. 5).

Figure 20. 

A Complex adductor muscles in Suncus murinus (after Dötsch 1983). B Ventral and lateral view of a skull and mandible of Sorex araneus (after Dötsch 1982). The arrow shows the positional shift of the jaw joint by about 2,2 mm. The solid beam represents the shortened loadlever (– 18 %) for the anteriormost biting point, by which the prominence of the incisors is compensated for. The jaw articulation is detached from the periotic ossification – a unique feature in mammals. The wide fenestra at the cranial base (blue asterisk) is interpreted as foramen lacerum medium (sensu Starck 1979). C Lateral view of a skull of Crocidura russula (Coll. W. Maier no. 3, virtual reconstruction of a µCT scan) with the molars in maximal occlusion. For the occlusion of the edge of the incisors, the mandible must be lowered at the secondary jaw joint. D The same specimen of Crocidura with a simplified drawing of the basic mechanic vectors and trajectories of the upper and lower jaws: Black colour – upper and lower jaws conceived as a pair of tweezers; blue colour – biting forces at the first incisors (K_in) and at the molars (K_mo) and jaw load K_mj. Red colour (F_mu) – estimated resultant of vertical adductor muscle forces with muscle-lever (L_mu). All partial forces were calculated in relation to F_mu = 1.

Figure 21. 

Skull size range of extant Soricidae as represented by Suncus murinus SMF 87406, Suncus etruscus SMF 26937, Myosorex varius SMF 55060, Sorex araneus SMF 82598 and Anourosorex squamipes SMF 48925. Suncus etruscus is the smallest living mammal. At this level of miniaturization, the skulls are relatively similar. All species are depicted in identical magnification (scale in Suncus murinus is 5 mm).

The functional meaning of the forward-inclination of the alae orbitalis and temporalis becomes clear when their relations with the glenoid region of the squamosum are considered. The rostral end of this membrane bone is tightly fused with other bone structures, namely the frontal, the orbitosphenoid and the alisphenoid. These close contacts are necessary, because this part of the squamosum is differentiated as ‘glenoid complex’, which is destined to withstand the pressure forces caused by the condyle of the mandible (Fearnhead et al. 1955). As shown in Figure 1, the glenoid fossa is fixed to the sidewall of the nasal capsule (lamina antorbitalis) and the nasal floor by means of two bony bars. The study of the early postnatal stages clearly proves, that the upper bar is the orbitosphenoid and the lower the alisphenoid. Both are tightly fused with the inner side of the squamosal, and thereby provide oblique pillars that can transmit pressure forces produced by the jaw muscles to the nasal capsule, the central stem or basisphenoid respectively.

The complicated fossa glenoidea has been carefully described by Fearnhead et al. (1955), Dötsch (1982) and others. From these studies we know that the articular fossa of the soricids has in fact two more or less separate compartments. The study of the morphogenesis of the joint clearly shows that the upper compartment develops first, and that the ventral joint cavity follows later (see Figs 15, 16). Only the upper joint develops a discus articularis. These morphogenetic observations are clearly in favour of the hypothesis to consider the upper compartment as the primary one, i.e., that it is homologous to the joint in the other mammals. The membrane bones squamosum and dentale develop secondary cartilage, which is functionally necessary in any diarthrotic joint.

For the lower compartment the squamosal develops a prominent process as counterbalance for the inferior condyle. Dötsch (1982) named this process simply ‘margo inferior‘ of the glenoid fossa, but by its peculiar morphogenesis and its functional importance for jaw mechanics, this structure deserves a name of its own, and we therefore call it ‘processus glenoideus inferior’, but we do not rule out that it may be homologous with the ‘postglenoid capitular facet’ of McDowell (1958: fig. 8). It obviously resists muscle forces pulling the mandible backwards and can guide it ventrally and rostrally. There exist differences in the modes of development of the processus between Crocidura and Sorex suggesting some degree of convergence.

Dötsch (1982) carefully described the movements of the mandibular head within the glenoid fossa, but she did not single out its extreme position at the occlusion of the leading edges of the procumbent first incisors. This rostral biting point is obviously reached when the mandible is shifted forward and downward. Figure 20 suggests that this movement is guided by the oblique dorsal plane of the processus glenoideus inferior. By this downward movement of the lower jaw, the articulation of the molars is unlocked. This assumed functional relationship between the secondary development of the lower compartment of the squamoso-dentary jaw joint (with its processus glenoideus inferior) and the specialized and apomorphic incisal biting, suggests that there exists a correlation between these two neomorphic features – a case of co-evolution of two heterogeneous structural units. In ontogeny, the differentiation of the jaw joint takes place prior to tooth eruption (Figs 15, 16). Whereas the joints have already reached maturity by postnatal day 15, tooth eruption begins in Crocidura russula just at this time, and in Sorex araneus only at day 22 or 23 (Vogel 1973). The function of the incisal bite seems to depend on the finished mandibular joint. Prior to biting, the mandibular jaw functions in suckling. The validity of this hypothesis may be tested by the fossil record. Heterosoricids from the Oligocene and Miocene already possess procumbent lower incisors and an elongated and oblique mandibular facet (Rabeder 1998). In Trimylus the dentary facets are widely divergent, whereas inn Domnina they are much more continuous (Asher 2005: fig. 5.3).

Figure 20 shows that the temporomandibular joint of soricids is shifted forward by at least 10 % of skull length, i.e., somewhat more than 2 mm in the two studied species. In other mammals the jaw joint is normally located directly adjacent to the meatus acusticus externus, i.e., anterolateral to the middle ear, but in the soricids it is shifted to the posterior part of the orbitotemporal fossa. Only cross sections show that the joint thereby comes to lie at the side of the expanded posterior nasal capsule (cf. Fig. 1). Our ontogenetic analysis shows that the connection of the glenoid region of the squamosal with the nasal capsule is therefore provided by the orbitosphenoid and alisphenoid. It was also shown above, that the final stages of this shift only take place in the first two postnatal weeks, and it was also demonstrated that the shift proceeds a little further in Sorex than in Crocidura (see also Fig. 12). Because we have no data available from the mass of the adductor muscles, we have made no efforts to calculate absolute measures of incisal bite force (Carraway and Verts 1994).

This unique shift of the jaw is also reflected by the peculiar elongated shape of the squamosum, and by the elongated foramen lacerum medium (see asterisk in Fig. 20 B). The name of this foramen is contentious in the literature. Our ontogenetic approach has shown that this foramen is posteriorly bound by the tegmen tympani and rostrally by the alisphenoid, and that it is therefore homologous with the foramen lacerum medium. The foramen lacerum medium is defined as the gap between the alisphenoid and the ear capsule or petrosum (Starck 1979, p. 361). According to Gaupp (1911) the foramen lacerum medium is a secondary cranial exit (‘Nervenöffnung zweiter Ordnung’) for the mandibular branch of the nervus trigeminus (V₃). The foramen lacerum medium must not be mistaken for the foramen prooticum or foramen sphenoparietale, which lies at a deeper level and means the primary opening between the pila antotica and the capsula otica of the chondrocranium (“Nervenöffnung erster Ordnung”, Gaupp 1911; Kuhn and Zeller 1987). The nervus mandibularis (V₃) may be enclosed by an incisura or foramen ovale at the posterior margin of the alisphenoid. The textbook of Gray (1858) mentioned a foramen lacerum medium; Gegenbaur (1892) used foramen lacerum anterius. Versluys (1927) and Stadtmüller (1936) named the ‘prootische Spalte’ also foramen lacerum anterius. Gregory (1910) had named the relatively large opening in Solenodon paradoxus (his fig. 18) “f.l.m.” (i.e., foramen lacerum medium). Maier (1987) used the term fissura sphenopetrosa. The nomenclature of this skull opening was confused by McDowell (1958), who named the opening ‘fenestra pyriforme’ without acknowledging or discussing the problems of terminological priority or homology. Nevertheless, most authors subsequently have used this new term since then (MacPhee 1981; Novacek 1993; Wible 2008; Wible and Shelley 2020).

In neonate soricids the foramen lacerum medium is still fairly narrow, and it is mediorostrally framed by the cartilage of the ala temporalis. In postnatal day 5, the gap has become distinctly wider, but is still partly bordered anteriorly by the cartilaginous ala temporalis. In the adult skull of Crocidura and Sorex this foramen of the basicranium is very wide, and it is closed by a membrane, which extends from the ventral side of the inner tegmen tympani to the free ventral edge of the squamosal (Maier et al. in press). Anteriorly it is attached to the ala temporalis or, later, to the alisphenoid; posteriorly it is attached to the anterior rim of the tegmen tympani and the tip of the pars cochlearis of the petrosal. Therefore, it is homologous to the posterior sphenobturate membrane (Maier 1987). The mandibular branch of the trigeminal nerve becomes enclosed by the alisphenoid, forming the foramen ovale.

We assume that the wide foramen lacerum medium is causally connected with the rostral shift of the mandibular joint. Its size can best be interpreted as saving of bone tissue of the alisphenoid, i.e. it is a feature of economization. Therefore, Anourosorex, which has a broad alisphenoid, possesses a small foramen lacerum medium (see Fig. 19 and McDowell 1958: fig. 38D) and Sorex araneus an excessively large one (McDowell 1958: figs 12, 19). In other taxa with a conspicuous foramen lacerum medium (Solenodon, Microgale; cf. MacPhee 1981: fig. 57), the size of the foramen lacerum is defined by the posterior margin of the alisphenoid. In the soricids its posterior margin is also defined by the tiny pars cochlearis of soricids (Ekdale 2013, 2015), which is partly compensated for by the expanded tegmen tympani.

Burda (1979), who did not use histology and microscopic anatomy, misinterpreted the dorsal membrane of the tympanic cavity: “Its dorsal wall is formed by the dura mater separating the tympanic cavity from the brain” (p. 8). He may not have been aware of the existence of the cavum epiptericum and its main contents, the huge trigeminal ganglion, which lies between the tympanic cavity and the brain cavity - at least in its medial part. Primarily, the alisphenoid and the sphenobturate membrane constitute the lateral wall of the cavum epiptericum (Maier 1987). Because of the expansion of the brain, this lateral wall is in the adult pushed into a horizontal position below the trigeminal ganglion. However, in its lateral part, the membrane may indeed include a bit of the dura mater of the brain cavity. Krapp and Niethammer (1990) characterized the big opening in the basicranium of soricids as follows: „Bei den Spitzmäusen dagegen ist der Schädelboden über dem Ectotympanicum membranös begrenzt und nicht von den Knochen Squamosum and Perioticum wie bei Igeln und Maulwürfen“ (p.16). („In shrews, the floor of the basicranium above the ectotympanic is membraneous and not from the squamosal and periotic bones as in hedgehogs and moles”; transl. WM). In terms of functional morphology, this expanded foramen lacerum medium may be analogous to the maxillary-palatine vacuities found in many small mammals.

Functionally, the rostral shift of the mandibular joint most probably is correlated with the acquisition of the procumbent first incisors, which put the anterior biting point forward. The upper and lower jaw can be conceived as a tweezer, where the upper arm consists of the interorbital pillar and the shell-like wall of the facial skeleton. The biting force at the incisors (F_in) produced by the adductor muscles, depends on the length of the load lever (l_in) of the mandible (Fig. 20D) – and is therefore increased by a shortening of the jaw length. The same holds true, of course, for the biting force at the molars. The functionally less important unicuspid antemolar teeth, which do not occlude, could become easily reduced to various degrees in different taxa of shrews; this part of the tooth rows may be considered as an initial diastema. The homology of these teeth is difficult to determine, because the ante-molar milk deciduous dention is reduced prior to eruption completely (Ärnbäck-Christie-Linde 1912; Kindahl 1959; Dannelid 1998; Yamanaka et al. 2010).

This unique loss of a complete dental generation – molars, which belong to the first tooth generation, excepted - to us is the strongest proof of miniaturization of the whole family Soricidae (see below). That the similarity of the front dentition of rodents and soricids is analogous, is simply proven by the fact that the incisors of the Glires belong to the first dental generation (Luckett 1985), and that of the shrews to the second. The zygomatic arch has become mechanically unimportant and was reduced. The pressure forces acting on the glenoid facets are therefore completely transferred by the bony trajectories of the interorbital pillar and the solid walls of the snout, which seems to have the characteristics of a mechanical shell (Figs 2, 12, 18, 21). The anterior end of the skull, where the upper incisors are implanted, became fortified by fusion of the maxillaries and premaxillaries mainly. We therefore hypothesize that the complicated mandibular jaw joint and mandibular apparatus is highly determined by the special requirements of the incisal and molar biting respectively. The ‘incisal complex’ will be studied in a separate paper (Maier, in prep.).

It is an open question whether soricids should be considered as an example of miniaturization, i.e., a product of evolution toward extremely small body size (Hanken and Wake 1993). Recently Polilov (2016) has shown examples of extreme miniaturization in insects, but Hanken and Wake (1993) have pointed out that vertebrates never became so small, that dramatic effects in their body organization resulted. Suncus etruscus is known as one of the smallest living mammals, and therefore it is likely that the whole family is miniaturized at specific mammalian conditions. We find a considerable span of body size between Suncus etruscus and Suncus murinus, but with no significant change in skull morphology; most soricid species range in the size category of Crocidura and Sorex (Fig. 21). However, the strongest argument of size reduction of the group as a whole is to us the suppression of the complete premolar dentition (Kindahl 1959). Therefore, soricids can be characterized as very small predators which have highly specialized their first incisors, jaw joint and jaw muscles. They have almost completely lost their eyesight, and rely instead on olfactory and tactile senses. It is an unsettled problem, whether soricids make use of their acoustic and vibration senses (Siemers et al. 2009; Zaytseva et al. 2015; Zsebök et al. 2015, Maier et al. in press). At present, it seems most appropriate, to define shrews as tiny insectivorous and carnivorous eutherian mammals, which search for their prey on microhabitats of forest and shrub floors as well as in superficial burrows dug by moles and rodents.

The scales in our histological figures indicate that almost all structures of the soricid skulls treated here measure only a few millimeters - even in adult animals. This means, that we are dealing with a mesoscopic size dimension (Maier 2021). This size category is situated between the microscopic and macroscopic approach, and it is difficult to handle technically. Therefore, morphological problems of this sort have been neglected in the past. However, modern µCT technique will help to solve some of the methodical problems at least as far as the osseous skeleton of adult mammals is concerned. Fetal and juvenile age stages will have to be studied by means of ‘microtome histology’ also in the future (Bargmann 1956).

Vogel (1973) concluded from his ontogenetic studies that Sorex has a more ‘primitive’ ontogeny than Crocidura. According to him, this difference is not only expressed by a much shorter gestation length, i.e., 20 vs. 30 days, but also by a different speed of development. He named Sorex a ‘primitver Nesthocker’ (“primitive and remaining in natal group”) and Crocidura an ‘evoluierter Nesthocker’ (“evolved and remaining in natal group”). He even stated: ”Die Geburtsrteife von Sorex und Neomys liegt dem Geburtszustand der Marsupialier wesent­lich näher als demjenigen der zur Zeit bekannten Eutheriennesthocker“ (p. 1314). (“The maturity of neonates of Sorex and Neomys is closer to that of marsupials than to any other known eutherians”; transl. by WM). At that time the understanding of mammalian phylogeny was quite undeveloped, and evolutionary constraints were not considered to be important factors. Because lipotyphlans are now generally associated with the superorder Laurasiatheria, it is very unlikely that they represent a morphotype from the lowermost base of placental mammals – and shrews are even the most specialized and derived members of the extant lipotyphlans. Contrary to Vogel (1973) we consider Sorex to be somewhat more derived than Crocidura in several features, but the problem of character polarity is not finally settled. Such questions of polarity are difficult to solve, and the answer normally depends on a reliable cladogram and on the reconstruction of the grundplan of different ancestral stages as well as on well-defined outgroup comparisons (Hennig 1982). Later on, Peter Vogel became very active in phylogenetic systematics based on molecular data (Dubey et al. 2007, 2008), but to our knowledge he did not revise his primary ontogenetic model of shrews.

The skull of shrews is not well studied, because it is too small for macroscopic and too big for microscopic investigations, i.e., it represents a ‘mesoscopic’ type of biological organization. µCT technique helps a lot to solve this problem, at least as far as bony structures are concerned. However, traditional ‘microtome histology’ will remain essential at least for the study of earlier ontogenetic stages and for the study of soft tissue structures.

The present study has shown that soricids are not only highly specialized in their dentition (complete loss of their anterior milk dention; procumbent first incisors with highly effective biting edge) but also in their double mandibular joint. We postulate that the rostral shift of the glenoid fossa compensates for the prominence of the procumbent incisors, and that the lower extension of the jaw joint co-evolved with the specialized incisal bite. As a result of this shift, the glenoid part of the squamosum became largely detached from the otic or petrosal and instead attached to the nasal capsule by means of the orbitosphenoid and the alisphenoid. As a by-product of the rostral move of the glenoid fossa and of the alisphenoid, the soricid foramen lacerum medium increased in size and was covered by the posterior sphenobturate membrane.

Because in soricids most skull elements fuse within the first 2–3 weeks of postnatal life, as many as possible ontogenetic stages are necessary to identify and homologize skeletal elements of the head skeleton – comparative anatomy necessarily has to become comparative ontogeny (Vergleichende Entwicklungsgeschichte). Moreover, it is mandatory for comparisons that the objects of study are defined within a suitable phylogenetic-systematical framework, and hypotheses on functional adaptations are necessary to understand evolutionary transformations. The present study barely reaches these methodological standards. Future research should widen the number of soricid taxa in order to better understand the existing differences. The ‘key innovation’ appears to be incisal biting, which explains both the permanent dentition as well as the jaw apparatus. Of course, the comparative ontogenetic study of other lipotyphlans, for example Uropsilus and Hylomys, is necessary to better understand soricids.

We thank Manfred Drack (Eberhard-Karls-Universität Tübingen) for his advice in problems of biomechanics and Katrin Krohmann and Juliane Eberhardt (both Senckenberg Forschungsinstitut und Naturmuseum Frankfurt) for technical support. Many thanks go to Ross D. E. MacPhee and Robert J. Asher who helped to improve the manuscript.