In extant elasmobranchs, the anterior and external semicircular canals are connected together, but the posterior canal forms an essentially separate circuit which is connected only to the vestibular region near the crista neglecta close to the base of the endolymphatic duct (de Beer, 1931, de Beer, 1937 and Retzius, 1881). A crus commune is therefore absent in extant elasmobranchs. The crus is also absent in the Pliocene otic region PV 4173 (Fig. 1) and in the two Early Cretaceous hybodont sharks scanned (Tribodus, Egertonodus); in all these forms, the posterior semicircular canal was found to be isolated as in extant elasmobranchs (Fig. 2). Data are unavailable for other Mesozoic and Paleozoic hybodonts, but it is possible that the posterior canal was isolated in most or even all of them. The anterior and posterior semicircular canals in the incertae sedis Triassic shark Acronemus (T/I 1289) appear to be widely separated, although the specimen investigated was too poorly preserved to provide many details of its skeletal labyrinth (Fig. 3). The phylogenetic relationships of Acronemus are uncertain; it has been classified as a ctenacanthiform on the basis of possibly plesiomorphic features of its fin spines (Rieppel, 1982), but presence of an isolated posterior semicircular canal supports a relationship with hybodonts and/or neoselachians. The anterior and posterior semicircular canals in Tristychius are said to be connected by a crus commune located below a laterally-directed branch of the dorsal otic ridge on the surface of the braincase (Dick, 1978). Whether the anterior and posterior canals actually meet within this ridge is unclear, but if this interpretation is correct the presence of a crus would represent an important difference from extant elasmobranchs and hybodonts, suggesting that Tristychius may be more distantly related to them than is Acronemus.

Unlike in extant elasmobranchs, there is a descending crus commune connecting the anterior and posterior canals in extant chimaeroids, osteichthyans, lampreys, as well as in Acanthodes, osteostracans, and galeaspids (Davis and Coates, 2009, Janvier, 1996, Maisey, 2001a and Miles, 1973). The semicircular canal arrangement in placoderms is apparently different from that seen in osteichthyans and has in the past been compared to that of elasmobranchs (Stensiö, 1950 and Stensiö, 1963), although there are important differences between the placoderm and elasmobranch skeletal labyrinth, and the placoderm posterior canal was probably never completely isolated from the others (Maisey, 2005). Among extinct chondrichthyans, a crus commune connecting the anterior and posterior semicircular canals has been identified in Pucapampella, Doliodus, Cladodoides, and symmoriiforms including “Cobelodus” (Maisey, 2001b, Maisey, 2005, Maisey, 2007, Maisey and Anderson, 2001 and Maisey et al., 2009). These early chondrichthyans therefore resemble osteichthyans, Acanthodes, and chimaeroids in retaining the crus commune (Fig. 4 and Fig. 5). Iniopterygian labyrinth morphology is probably specialized in having the anterior and posterior canals somewhat flattened in the horizontal plane (Pradel et al., 2009a), but a crus commune nevertheless seems to be present. It has been suggested that the anterior and external canals were united in Permian xenacanths, and that the xenacanth posterior canal was circular (and by inference isolated) as in extant elasmobranchs (Schaeffer, 1981). However the xenacanth endocast suggests that a crus commune was present although it does not reveal a pre-ampullary canal (see next section). Interpretation of a neoselachian-like arrangement of the posterior canal in xenacanths therefore seems doubtful, especially in view of other primitive features in the skeletal labyrinth shared by xenacanths and Cladodoides (Maisey, 2005). The semicircular canal arrangement has not yet been determined in the enigmatic Devonian gnathostome Ramirosuarezia (Pradel et al., 2009b).

In extant elasmobranchs, there is an elongated extension of the posterior semicircular canal anterior to the ampulla (de Beer, 1931, Retzius, 1881 and Schaeffer, 1981), termed the posterior pre-ampullary canal (Maisey, 2001a). The semicircular canal curves dorsally to complete its circuit and reach the perilymphatic fenestra. An identical arrangement has been found in the Pliocene carcharhinid otic region, as well as in the hybodonts Tribodus and Egertonodus (Lane, 2010b, Maisey, 2004b and Maisey, 2005) (Fig. 1 and Fig. 2). The posterior canal in Acronemus also appears to have an ascending pre-ampullary ramus (Fig. 3), although it is incompletely preserved in the material investigated.

A short posterior pre-ampullary canal is present in extant chimaeroids and Acipenser, but extends anteriorly only as far as the utricular chamber. There is no ascending ramus like that seen in extant elasmobranchs; instead, the crus commune extends downward to meet the utricular chamber, and perilymphatic openings are absent. A pre-ampullary canal is absent in other osteichthyans, as well as in galeaspids and osteostracans, but it is unknown whether one was present in Acanthodes. In placoderms, however, the posterior canal may form an almost complete circuit and there is a well-developed posterior pre-ampullary canal, which has been interpreted as forming a neoselachian-like “posterior utriculus” (e.g., Jagorina, Tapinosteus, Kujdanowiaspis) (Stensiö, 1969). This similarity to extant elasmobranchs has been credited with considerable phylogenetic importance, supposedly supporting the view that placoderms and elasmobranchs are closely related (e.g., as elasmobranchiomorphs) (Stensiö, 1963 and Stensiö, 1969). However, it is important to note that at the time that proposal was made no information was available concerning labyrinth morphology in early chondrichthyans; instead, it was assumed that they possessed the extant elasmobranch condition. A posterior pre-ampullary canal has not been recognized in Permian xenacanths. There may be a short pre-ampullary canal in Cladodoides, but there is no evidence of one in symmoriiforms (Maisey, 2005 and Maisey, 2007). The labyrinth in all the specimens of Pucapampella examined to date have failed to reveal details of the posterior semicircular canal below its ampulla, and the ampullae remain unknown in Doliodus. However, the skeletal labyrinths of both Pucapampella and Doliodus include a crus commune, showing that the posterior canal was not isolated (Fig. 5).

In extant elasmobranchs, this feature is related to isolation of the posterior semicircular canal, which is connected via the posterior pre-ampullary canal to the “posterior utriculus” containing the sensory crista neglecta (de Beer, 1931, de Beer, 1937 and Maisey, 2001a, Retzius, 1881 and Torrey, 1962). The “posterior utriculus” in Notorynchus forms an inflated chamber in the dorsal part of the posterior pre-ampullary canal, containing both the crista neglecta and a connection with the sacculus (see next section), and is therefore probably homologous with the posterior part of the utricular chamber in osteichthyans where the macula neglecta is also located. The elasmobranch “anterior utriculus” includes the entrance of the combined anterior and external semicircular canals, like the anterior part of the utriculus in other gnathostomes. The posterior pre-ampullary canal has been considered part of a “posterior utriculus” in chimaeroids and placoderms (Norris, 1929 and Stensiö, 1963). The posterior pre-ampullary canal lacks an inflated utricular chamber in these forms, but in chimaeroids the crista neglecta lies within the corresponding part of the posterior pre-ampullary canal, and perhaps this was also the case in placoderns. Additionally, the posterior pre-ampullary canal in some extant elasmobranchs (e.g., Torpedo) is not inflated into a distinct “posterior utriculus” around the crista, and the crista is located within the posterior canal duct instead of the posterior pre-ampullary canal (Maisey, 2001a).

In the Pliocene otic region (PV 4173; Fig. 1), the posterior pre-ampullary canal broadens dorsally, presumably forming an inflated chamber which was clearly separated from the sacculus by cartilage, although its outer wall is broken and its original size is uncertain. In the hybodonts Tribodus and Egertonodus, the dorsal part of the posterior pre-ampullary canal widens just below the perilymphatic fenestra (Fig. 2), suggesting the presence of an inflated “posterior utriculus” as in many extant sharks. The corresponding region of the Acronemus braincase was poorly resolved in the CT scan and its morphology could not be deduced. There is no evidence of an inflated pre-ampullary region in the xenacanth endocast, nor in endocasts of Cladodoides and “Cobelodus” sp. (Maisey, 2005, Maisey, 2007 and Schaeffer, 1981). This region is unknown in Pucapampella or Doliodus, but other features of the labyrinth resemble those of osteichthyans (in which a “posterior utriculus” is absent). Among gnathostomes, therefore, an inflated “posterior utriculus” within the posterior pre-ampullary canal is known at present only in neoselachians and hybodonts.

The posterior semicircular canal in extant elasmobranchs has a single opening into the vestibular region (posterior utriculo-saccular opening). The posterior canal duct which extends from this opening is one of the most obvious morphological features of the elasmobranch non-otolithic phonoreceptor (Bleckman and Hoffman, 1999, Corwin, 1978 and Corwin, 1981a). In chimaeroids and osteichthyans, however, there are two openings; the canal opens into the vestibular chamber ventrally and also shares a dorsal connection with the anterior canal. There is a single posterior utriculo-saccular opening in the Pliocene carcharhinid otic region PV 4173 (Fig. 1), and a single connection is also evident in Tribodus and Egertonodus (Fig. 2), but the arrangement in Acronemus is unknown. By contrast, in Cladodoides, “Cobelodus” sp., Pucapampella, and Doliodus, confluence of the anterior and posterior canals of the skeletal labyrinth resembles the osteichthyan arrangement, suggesting that the posterior canal had separate connections to the vestibular region dorsally and ventrally (Fig. 4). The nature of this connection in xenacanths has not yet been established.

Despite the fact that the skeletal labyrinth of placoderms has been interpreted after a extant elasmobranch paradigm (Stensiö, 1950, Stensiö, 1963 and Stensiö, 1969), none of the placoderm labyrinth endocasts described so far has revealed whether the posterior canal had one or two openings into the vestibular region; moreover, the manner in which the semicircular canals are connected to the vestibular region is different (see below). We are thus reluctant to accept earlier interpretations of the placoderm posterior canal configuration based solely on a extant elasmobranch model.

In holocephalans and osteichthyans, the crista neglecta is located in the posterior part of the large utricular chamber, behind the utricular macula and the openings for the anterior and external ampullae. By contrast, in elasmobranchs, the crista neglecta is located adjacent to the posterior utriculo-saccular opening of the posterior semicircular canal, usually within the “posterior utriculus”. In batoids, the crista is either associated with the saccular end of the posterior canal duct or is located within the duct (Corwin, 1989 and Retzius, 1881). The crista neglecta is considerably larger in extant elasmobranchs than in most other craniates, but nevertheless displays a wide range of size and complexity (Corwin, 1978). The sensory crista is not preserved in fossils, so its former position and extent can only be inferred from other features of the labyrinth. It may well have been located close to the posterior utriculo-saccular opening in Tribodus and Egertonodus, but in Paleozoic chondrichthyans such as Cladodoides and “Cobelodus” it could have been located in the posterior part of the utriculus, as in extant osteichthyans and chimaeroids.

The crista neglecta in extant elasmobranchs varies in size and complexity, offering a potentially informative source of new phylogenetic data for extant elasmobranchs once a sufficient taxon sample has been investigated. For example, the macula in Mustelus, Ginglymostoma, Carcharhinus, Negaprion, and Notorynchus is divided into two sensory fields, located on opposite sides of the posterior canal duct (anteriorly and posteriorly) and separated by a longitudinal area lacking sensory hair cells; the posterior area is considerably larger than the anterior one (Corwin, 1978, Corwin, 1981a and Corwin, 1981b). By contrast, in Torpedo and Myliobatis there is a single macula with a central area lacking hair cells, potentially representing an important distinction between extant sharks and batoids. It is unknown whether the single sensory area in batoids is homologous with both or only one of the areas in sharks, but it corresponds topographically to the posterior area. In Carcharhinus, Negaprion, and Notorynchus, macular sensory hair cells are directed toward the perilymphatic fenestra in the anterior field and away from it in the posterior field, but in Mustelus and Ginglymostoma some of the cells are misaligned. Hair cells in the macula of batoids tend to be aligned medially and laterally. These limited observations suggest that there may be functionally-related phylogenetic patterns in macular structure and sensory hair cell orientation among extant elasmobranchs that would merit further investigation.

An extensive medial capsular wall is present in extant elasmobranchs, almost completely separating the endocranial and labyrinth spaces apart from foramina for the acoustic and glossopharyngeal nerves. By contrast, the medial wall of the otic capsule is membranous in extant holocephalans, actinopterygians and dipnoans (de Beer, 1937 and Romer, 1937). Among extant gnathostomes, therefore, only elasmobranchs have a chondrified medial capsular wall. A medial capsular wall is nevertheless well developed in placoderms, osteostracans, and galeaspids. In extant adult lampreys, the posterior part of the medial capsular wall is chondrified anterior to the acoustic nerve and fuses to the cartilaginous lateral wall of the cranial cavity (Jarvik, 1977), but the wall is unchondrified in hagfishes.

The medial capsular wall is fully developed in the Pliocene otic region PV 4173 (Fig. 1) and in both of the hybodont sharks investigated here (Tribodus, Egertonodus; Fig. 2) as well as in Acronemus (Fig. 3), but it is only partially chondrified in xenacanths (Schaeffer, 1981) and is mostly unchondrified in earlier chondrichthyans (Fig. 4); e.g., Cladodoides, symmoriiforms, Pucapampella, Doliodus. Within chondrichthyans, therefore, the presence of a chondrified medial capsular wall completely separating the saccular chamber from the endocranial space appears to be restricted to neoselachians, hybodonts, and the incertae sedis taxon Acronemus, although the condition in Tristychius and Paleozoic hybodonts (e.g., Hamiltonichthys, Onychoselache) needs further investigation. Collectively, these findings suggest that a medial capsular wall was primitively present in gnathostomes but was lost in the common ancestors of osteichthyans and chondrichthyans, only to be re-acquired within chondrichthyans after the divergence of elasmobranchs and chimaeroids but before the divergence of hybodonts and neoselachians.

The perilymphatic fenestrae or fenestrae vestibuli are paired openings in the dorso-medial capsular wall in extant elasmobranchs that expose small sections of the posterior semicircular canal (de Beer, 1931, de Beer, 1937, Holmgren, 1940, Holmgren, 1941 and Howes, 1883). During ontogeny, their position is defined in part by the margins of the taenia medialis forming the floor of the endolymphatic (parietal) fossa (see below). Each of the fenestrae lies medial to a posterior utriculo-saccular opening and is therefore located directly opposite the crista neglecta. The perilymphatic fenestrae act as a sound-transmitting device into the non-otolithic component of the phonoreceptor system (Corwin, 1977, Fay et al., 1974 and Tester et al., 1972) and are closed by a specialized membrane in some elasmobranchs (Norris, 1929)

The perilymphatic fenestrae are usually visible on the dorsal surface of an extant elasmobranch braincase as paired openings just behind the endolymphatic fenestrae. Similar openings have been identified in some fossils (Brito and Seret, 1996) and were probably present in many if not all extinct neoselachians. Cartilage is not preserved adjacent to the expected position of the perilymphatic fenestra in the Pliocene otic region PV 4173, but the fenestrae may be represented by expanded areas above each posterior canal in CT reconstitutions of the labyrinth in Tribodus and Egertonodus (Fig. 2). Perilymphatic fenestrae seem to be present in Acronemus although these were not readily observed in the reconstitution. Vestibular fenestrae are said to be present in Tristychius (Dick, 1978). Perilymphatic fenestrae were not observed in any of the scanned Paleozoic sharks investigated here and are absent in extant chimaeroids, osteichthyans, lampreys and hagfishes. They have also not been found in placoderms, Ramirosuarezia, or fossil agnathans. Collectively, these findings suggest that perilymphatic fenestrae have a very restricted occurrence within craniates, and their presence is interpreted here as a potential synapomorphy of neoselachians, hybodonts, Acronemus, and Tristychius.

In the membranous labyrinth of extant elasmobranchs, the distal (i.e. non-ampullary) end of the external semicircular canal meets the descending distal part of the anterior canal, forming a crus that meets the anterior utricular region. Unfortunately, this union is located deep within the vestibular chamber and is consequently not evident in endocasts of the skeletal labyrinth (Maisey, 2004b and Schaeffer, 1981). The labyrinth in some placoderms is said to possess an elasmobranch-like crus between the anterior and external semicircular canals (Goujet, 1984 and Stensiö, 1963), although these are situated superficial to the sacculus rather than deep within it as in extant elasmobranchs. Furthermore, the connection between the anterior and external semicircular canals in some placoderms (e.g., Tapinosteus, Macropetalichthys) was located dorsal to the loop of the posterior canal instead of ventral to it as in elasmobranchs (Stensiö, 1969). In others (e.g., Kujdanowiaspis, Dicksonosteus), the external semicircular canal apparently met the posterior one, and the combined canal then continued forward to meet the anterior canal (Goujet, 1984 and Stensiö, 1969). The labyrinth of Jagorina has been restored using an extant elasmobranch paradigm (Stensiö, 1950), but its endocast could just as readily have included a crus commune. Thus the posterior semicircular canal in placoderms was probably not isolated as in extant elasmobranchs, and in some forms a crus commune may have been present. While some placoderms may have possessed a crus between the external and anterior semicircular canals, the arrangement apparently differed from that found in extant elasmobranchs and therefore provides less morphological support for a phylogenetic relationship than was previously supposed. There is also no evidence to suggest that the odd semicircular canal arrangement in placoderms was associated with an elasmobranch-like semi-directional phonoreceptor system.

In lateral view, the external canal of extant elasmobranchs passes posteriorly from its ampulla within the lateral wall of the sacculus, then curves upwards posteriorly as it approaches the anterior canal (Maisey, 2001a). The inner and outer parts of the canal do not therefore lie in the same plane. This feature can also be observed in the Pliocene otic region PV 4173 and in the skeletal labyrinths of Tribodus and Egertonodus (Fig. 1 and Fig. 2). However, in Acronemus, Cladodoides, symmoriiforms, Doliodus and Pucapampella, the external canal lies almost entirely within a plane passing through the mid-region of the saccular region, as in osteichthyans (Maisey, 2005 and Maisey, 2007). In extant chimaeroids, the canal may climb slightly out of the mid-saccular plane. In some placoderms, the external canal passes around the sacculus much as in osteichthyans, although its distal end may be slightly elevated where it passes over the saccular chamber (it does not pass deeper within the chamber as in extant elasmobranchs). The external semicircular canal in symmoriiform sharks has an unusual anterodorsal inclination with respect to the vertebral axis (Maisey, 2007), but this affects the entire canal and not merely its distal part. These observations suggest that out-of-plane orientation of the external canal may unite neoselachians and hybodonts (and perhaps chimaeroids), but would exclude Acronemus and many earlier chondrichthyans. From its rather disjunct distribution, this feature is probably unrelated to LFSDP in elasmobranchs.

In extant elasmobranchs, the membranous anterior and external canals share a single anterior utriculo-saccular opening (Torrey, 1962), whereas in osteichthyans and holocephalans these canals open separately into the utriculus. The scan of the Pliocene otic region PV 4173 reveals that the anterior and external ampullae make contact before entering the utricular recess and share an opening into the utricular recess (which is connected to the saccular chamber dorsally via a single opening; Fig. 6). In Notorynchus, the chambers of the anterior and external ampullae meet above the utricular recess and their openings are confluent, with a single opening into the utricular recess and sacculus (Fig. 7). In endocasts of Tribodus, Egertonodus, Cladodoides, symmoriforms, the anterior and external ampullary chambers lie close together but probably opened separately into the utricular recess, although there is still a single utricular entrance into the sacculus (Fig. 2 and Fig. 4). The arrangement was probably similar in Pucapampella, but has not yet been determined in Acronemus and Doliodus.

In placoderms, the anterior and external ampullae are widely separated and clearly have separate connections with the utriculus; in Jagorina the ampullae lie on opposite sides of a small, indistinct utricular recess, and in other placoderms (e.g., Kujdanowiaspis, Tapinosteus) the anterior ampulla is separated from the utricular recess by a pre-ampullary extension of the anterior canal (Stensiö, 1950 and Stensiö, 1969). No placoderm is known in which the anterior and external ampullae meet as in elasmobranchs. Pre-ampullary extensions of the anterior and external canals are present in some batoids (e.g., Torpedo) but these typically unite before entering the utricular recess via a single opening (Maisey, 2001a), unlike in placoderms. Such pre-ampullary canals are absent in the fossil chondrichthyans examined here. A common utricular opening for the anterior and external ampullae may therefore be an apomorphic character of neoselachians, although its distribution needs further investigation.

Instead of solid otoliths, extant chondrichthyans possess an otolithic mass within the saccular chamber, consisting primarily of endogenous crystalline calcium carbonate otoconia bound together by an organic matrix (Mulligan and Gauldie, 1989). Additional exogenous grains of sand (quartz, feldspar) and ferruginous material (magnetite, hematite) are sometimes present within the otolithic mass of elasmobranchs (Fänge, 1982, Norris, 1929 and Stewart, 1906) although it is still unknown whether these also occur in chimaeroids. Exogenous sand grains may have been present within the otic labyrinth of osteostracans, acanthodians, and putative fossil chondrichthyans from the Early Devonian of Canada (Sahney and Wilson, 2001), suggesting a widespread distribution among early vertebrates including many forms which probably lacked an elasmobranch-like phonoreceptor. It is unlikely that exogenous sand grains play a role in LFSDP of elasmobranchs, and are instead probably related to the inner ear's mechanosensory function (Hanson et al., 1990). Mechanoreception using assimilated exogenous sand grains is most parsimoniously interpreted as a primitive feature of gnathostomes that has been conserved by elasmobranchs. Although exogenous sand grains were not recognized in scans of the fossil chondrichthyans investigated, they may have been present originally and either were lost or simply not detected.

The endolymphatic fossa in extant elasmobranchs is a depression in the cranial roof between the otic capsules, filled with a gelatinous subcutaneous fluid with an acoustic impedance matching that of sea water (Corwin, 1989). The fossa is associated with paired perilymphatic fenestrae (see above), plus small paired muscles connected with the endolymphatic ducts. During ontogeny, the anterior margin of the fossa is defined by the embryonic tectum synoticum and it becomes floored by cartilage derived from the embryonic taenia medialis, which extends between the otic capsules and represents either a posterior extension of the synotic tectum or an independent bridge of cartilage (de Beer, 1931 and Holmgren, 1940). The embryonic otico-occipital fissure (a space between the otic capsules and occipital arch) becomes closed during ontogeny, leaving the posterior dorsal fontanelle open as the parietal fossa. However, closure of the fossa posteriorly in extant elasmobranchs is brought about in various ways, for example by direct fusion of the occipital pilae to the otic capsules (e.g., Squalus), or by posterior extension of the synotic tectum (which forms the floor of the fossa; e.g., Heterodontus). Thus, the posterior tectum behind the fossa can originate either from cartilage of the otic region or from the occipital arch.

Many extinct chondrichthyans are known to have the parietal fossa completely surrounded by cartilage as in neoselachians (Maisey, 1983 and Maisey et al., 2004), but a chondrified floor to the fossa is known only in neoselachians, hybodonts and a few incertae sedis taxa such as Acronemus and Tristychius. This floor helps define discrete perilymphatic openings, and thus contributes to the channeling of sound waves into the posterior canal duct. The otico-occipital fissure is closed in all these forms, suggesting that there may be a developmental and/or functional correlation between closure of the fissure and the formation of a chondrified floor to the fossa between the otic capsules in elasmobranchs. The posterior tectum and the floor of the fossa in hybodonts and Acronemus probably included cartilage derived from the synotic tectum, the taenia medialis, and/or the occipital arch. There is no direct evidence that the fossa in hybodonts or A. tuberculatus contained impedance-matching gel or muscles closing the endolymphatic ducts, although this seems a reasonable assumption since other neoselachian-like features are evident in their labyrinths. Where the floor of the fossa is unchondrified, the otico-occipital fissure is typically persistent whether the fossa is closed by a cartilaginous tectum posteriorly (e.g., xenacanths, Tamiobatis, symmoriiforms, Cladodoides) (Maisey, 2005 and Schaeffer, 1981) or is confluent with the otico-occipital fissure (e.g., Pucapampella) (Maisey, 2001b and Maisey and Anderson, 2001). It is unlikely that the posterior tectum in early chondrichthyans was derived ontogenetically from the synotic tectum farther anteriorly because the intervening floor of the fossa is unchondrified. It seems equally implausible that the occipital arch made a contribution to the posterior tectum in forms where the two were separated by the otico-occipital fissure. It is therefore concluded that the posterior tectum in Paleozoic chondrichthyans probably represents a separate embryonic taenia tecti medialis situated at the posterior end of the fossa.

An endolymphatic fossa is said to be absent in extant chimaeroids (Schaeffer, 1981), but there is an unpaired “endolymphatic duct” in the corresponding position, housing the paired endolymphatic canals and their associated sacs (de Beer and Moy-Thomas, 1935 and Jollie, 1962). A small unpaired endolymphatic opening has also been found in extinct holocephalimorphs (e.g., Squaloraja, Debeerius, iniopterygians) (de Beer and Moy-Thomas, 1935, Grogan and Lund, 2000 and Pradel et al., 2009b). This small canal is probably homologous with the much larger fossa of extant elasmobranchs although it lacks a chondrified floor in holocephalans. The ontogeny of the otic region in chimaeroids is poorly known, but there is no otico-occipital fissure in the adult; the arch is already fused to the otic region in a 6-month Chimaera embryo, but is not at all evident in a 60 mm Callorynchus (de Beer and Moy-Thomas, 1935). It is concluded that an endolymphatic fossa is primitively present in crown-group chondrichthyans and became secondarily reduced early in holocephalan evolution (for a discussion of its possible homology with the osteichthyan posterior dorsal fontanelle, see reference Maisey (2005)). The endolymphatic fossa in symmoriiforms is also comparatively small and lacks a chondrified floor (Maisey, 2007), and probably housed both endolymphatic ducts. Presence of this small median opening is therefore a potential synapomorphy of holocephalans and symmoriiforms.

A posterior dorsal fontanelle may have been present in Acanthodes, although its interpretation as forming a neoselachian-like endolymphatic fossa is considered highly speculative (Maisey, 2005). Endolymphatic ducts were present in placoderms, and in Brindabellaspis these follow a tortuous path reminiscent of the ducts in elasmobranchs, with expanded upper parts that have been interpreted as forming an endolymphatic fossa (Young, 1980), although that interpretation has been questioned and it is doubtful whether any placoderm possessed a posterior dorsal fontanelle comparable to that of osteichthyans or chondrichthyans (Maisey, 2001a). There is no evidence of a fontanelle or endolymphatic foramina in Ramirosuarezia (Pradel et al., 2009b). The posterior dorsal fontanelle is therefore considered a plesiomorphic feature of crown-group gnathostomes associated with paired endolymphatic ducts and primitively continuous with the otico-occipital fissure. Since there is no evidence of a persistent otico-occipital fissure in placoderms, osteostracans or galeaspids, this feature may also represent another apomorphic feature of crown-group gnathostomes (possibly related to the evolution of the posterior dorsal fontanelle). Isolation of the fontanelle from the otico-occipital fissure apparently occurred early in chondrichthyan evolution, prior to closure of the fissure and the development of a continuous cartilaginous floor by the taenia medialis.

In extant elasmobranchs, the parachordals are expanded laterally below the otic capsules, forming the hypotic lamina (de Beer, 1931). The presence of this lamina has been regarded as an apomorphic character uniting non-holocephalan chondrichthyans, despite the apparent presence of an expanded parachordal plate in some placoderms (Schaeffer, 1981). However, the lamina may be secondarily absent in chimaeroids if holocephalans are derived from shark-like chondrichthyans (Coates and Sequeira, 2001 and Ginter, 2005) (see Discussion below).

The metotic fissure represents a ventrolateral continuation of the otico-occipital fissure, and forms a transient embryonic space between the floor of the otic capsule and the hypotic lamina (a lateral extension of the parachordal cartilage in extant elasmobranchs). The fissure closes during ontogeny apart from the glossopharyngeal canal, which passes below the otic capsule and opens posteroventrally in extant elasmobranchs, hybodonts, Acronemus and Tristychius. A persistent metotic fissure has been documented in numerous Paleozoic chondrichthyans (e.g., xenacanths, Tamiobatis, Cladodoides, symmoriiforms, Pucapampella) and one may also be present in Doliodus and Antarctilamna (Maisey, 2001b, Maisey, 2001c, Maisey, 2004b, Maisey, 2005, Maisey, 2007, Maisey and Anderson, 2001, Maisey et al., 2009 and Schaeffer, 1981). In all these sharks, however, the glossopharyngeal nerve probably passed through the metotic fissure between the otic capsule and a hypotic lamina, and the course of the nerve is sometimes marked by a local widening of the fissure in a similar position to the glossopharyngeal canal in extant elasmobranchs (e.g., Cladodoides, “Cobelodus”; Fig. 4). It could therefore be argued these forms possess both a persistent metotic fissure and a glossopharyngeal canal.

Closure of the metotic and otico-occipital fissures in the common ancestors of neoselachians and hybodonts probably represents an important precursor state for the evolution of LFSDP, since the medial wall of the posterior semicircular canal (forming part of the non-otolithic phonoreceptor component in elasmobranchs) also forms the outer surface of the otico-occipital fissure both in extant elasmobranchs prior to its closure and in Paleozoic sharks with a persistent fissure. Chimaeroids are thought to lack an expanded hypotic lamina because their glossopharyngeal foramen in situated ventrally as in osteichthyans (de Beer, 1937). The positions of the vagal and glossopharyngeal foramina in Ramirosuarezia resemble those of chimaeroids (Jollie, 1962 and Pradel et al., 2009b), suggesting that it also lacked a hypotic lamina. A metotic fissure is present in some early actinopterygians (Gardiner, 1984) and may be a primitive feature of crown-group gnathostomes. Persistent metotic and otico-occipital fissures are typically absent in placoderms. No vertebrate is known with a persistent metotic fissure and a closed otico-occipital one.

This large paired process represents an extension of the posterolateral wall of the otic capsule and encloses the outer loop of the posterior semicircular canal in xenacanths, Tamiobatis, and Ctenacanthus from the Cleveland Shale (Schaeffer, 1981). The lateral walls of the otic capsules are broken in the only known braincase of Cladodoides, exposing the loops of the posterior semicircular canals (Fig. 8D). Consequently, the lateral otic processes are missing in the specimen and their original extent can only be surmised (Maisey, 2005). Lateral otic processes are absent in symmoriiform sharks (Fig. 8E) and in Ramirosuarezia (Maisey, 2007 and Pradel et al., 2009a). Chimaeroids and osteichthyans also lack such a process, and the placoderm posterior postorbital process is probably not homologous with the lateral otic process of sharks (Schaeffer, 1981). However, in a three-dimensional iniopterygian braincase, the posterior semicircular canal appears to bulge into the base of a small process on the posterolateral wall of the otic capsule (Pradel et al., 2009a), although it apparently did not articulate with a hyomandibula as in xenacanths or Tamiobatis.

It has been suggested that a lateral otic process occurs in many extant and fossil elasmobranchs (Coates and Sequeira, 2001, Maisey, 2001a and Schaeffer, 1981), although the process as originally defined (Schaeffer, 1981) is absent in neoselachians, hybodonts, and certain other chondrichthyans (Maisey, 2005). Neoselachians possess a post-otic process (Fig. 8A: pstpr) containing the glossopharyngeal canal (Holmgren, 1941). A similar process is present in Acronemus (Fig. 3). By contrast, the glossopharyngeal nerve in xenacanths and Tamiobatis apparently passed mesial and ventral to the lateral otic process rather than through it. The supposed lateral otic process of Egertonodus (Maisey, 1983) is probably not homologous with either the lateral otic process of Paleozoic sharks or the post-otic process of neoselachians, since it does not contain either the posterior semicircular canal or the glossopharyngeal canal. The posterior semicircular canal is not enclosed by a process in symmoriiforms, Pucapampella, Doliodus, neoselachians, hybodonts, or Acronemus. The “otic process” of Tristychius has been compared with the lateral otic process of Tamiobatis and xenacanths (Dick, 1978), although it did not form part of the hyomandibular articulation (which is apparently located farther anteriorly) and it resembles the neoselachian post-otic process in being located immediately above the glossopharyngeal foramen.

The location of the postorbital arcade on the side walls of the braincase in extant elasmobranchs seems to be highly constrained ontogenetically (Holmgren, 1940), but its position is slightly more variable in extinct chondrichthyans (Maisey, 2005). The arcade is positioned entirely anterior to the otic labyrinth in Cladodoides, xenacanths, Tamiobatis sp., symmoriiforms, Doliodus, and Pucapampella (Fig. 5 and Fig. 8), whereas in hybodonts and neoselachians the postorbital arcade or process is positioned farther posteriorly, either lateral to the anterior ampulla (e.g., Notorynchus; Fig. 8A) or just behind the ampulla (e.g., Tribodus, Egertonodus; Fig. 8B, C). The anterior ampulla is located medial to the postorbital arcade in Acronemus, providing additional support to the hypothesis that it is phylogenetically related to hybodonts and/or neoselachians. The anterior part of the otic capsule is located level with the postorbital wall in iniopterygians (Pradel et al., 2009a), but the capsule in extant chimaeroids is located entirely behind the postorbital wall (Jollie, 1962). The ampulla is positioned behind the level of the postorbital wall in early osteichthyans (e.g., Kansasiella) and in Latimeria (Millot and Anthony, 1958 and Poplin, 1974), and this arrangement has been considered primitive for crown-group gnathostomes (Maisey, 2001a). The arrangement in hybodonts and neoselachians has been interpreted as a derived condition (Maisey, 2005), resulting from a posterior shift in the attachment point of the primary postorbital process formed in the posterior part of the embryonic supraorbital cartilage (Holmgren, 1940). Nevertheless, in Kujdanowiaspis the anterior ampulla is located medial to the anterior postorbital process as in neoselachians and hybodonts (Stensiö, 1969), possibly offering some evidence of homology between this process and the postorbital process of elasmobranchs. The position of the anterior ampulla in Ramirosuarezia is unknown and there is no obvious postorbital process although there is a slight bulge in the anterolateral surface of the otic capsule (Pradel et al., 2009b).

It is unlikely that the relative positions of the anterior ampulla and postorbital arcade had any bearing on the evolution of phonoreception in elasmobranchs (which primarily involves the posterior part of the otic region).

The occipital region is located behind the otic capsules in osteostracans, placoderms, and Acanthodes (Janvier, 1996), and this may be the primitive gnathostome condition (Schaeffer, 1981). In extant elasmobranchs, however, the occipital arch extends anteriorly during ontogeny, ultimately becoming wedged between the posterior parts of the otic capsules (Johnels, 1948). The extant elasmobranch arrangement can also be observed in Mesozoic hybodonts, as well as in Acronemus and Tristychius. Even in those Paleozoic chondrichthyans whose occipital region extends a considerable distance posteriorly (e.g., xenacanths, Tamiobatis, Ctenacanthus, Cladodoides), the anterior part of the occipital arch is wedged between the otic capsules (Maisey, 2001c, Maisey, 2005, Maisey, 2007 and Schaeffer, 1981). The occipital arch also extends a short distance between the otic capsules in Paleozoic symmoriifom sharks and Cladoselache (Maisey, 2007). However, the occipital arch in Pucapampella and Doliodus does not intrude at all (Maisey and Anderson, 2001 and Maisey et al., 2009). Much of the occipital arch in chimaeroids is situated anteriorly (between the posterior parts of the otic capsules), apart from paired lateral extensions forming an articulation with the synarcual (Jollie, 1962). The arch in iniopterygians intrudes slightly between the capsules (Pradel et al., 2009a) but the condition is uncertain in Ramirosuarezia (Pradel et al., 2009b).

Based on these observations, wedging of the occipital region between the otic capsules seems to represent a synapomorphy of a large clade of chondrichthyan fishes including neoselachians, holocephalans, hybodonts, Acronemus, Tristychius, xenacanths, and Paleozoic Ctenacanthus-like cladodont sharks (i.e., crown-group chondrichthyans plus many stem chondrichthyans), and excludes only a few early chondrichthyans such as Doliodus and Pucapampella.