Micromeryx? eiselei—A new moschid species from Steinheim am Albuch, Germany, and the first comprehensive description of moschid cranial material from the Miocene of Central Europe

Moschids are enigmatic pecoran ruminants whose phylogeny is still not fully understood. So far we know only little of the family’s early evolutionary history and the origin of the modern genus, Moschus. Here we present a comprehensive description of cranial material, including the ear region and the dentition, of fossil moschid material from the Middle Miocene locality Steinheim am Albuch (13.5 Ma; Germany). This study provides the first exhaustive dataset for the cranial osteology of Micromeryx flourensianus, the most likely oldest true moschid. It furthermore reveals the presence of a second, so far undescribed moschid species, we here name Micromeryx? eiselei, in the abundant material from the locality. The two taxa can be clearly distinguished by characters of the skull, the ear region, the dentition, as well as by size. This evidences the sympatric occurrence of two moschid species in the locality Steinheim am Albuch.


Introduction
One of the still intriguing questions in the evolutionary history of ruminants is the phylogeny of the family Moschidae and especially the origin of the modern genus Moschus Linnaeus, 1758. Moschidae are small pecoran ruminants with elongated upper canines in males lacking any cranial appendages. Moschus remains the only surviving genus of the family. It is restricted to montane woodlands in Asia [1], has been in decline for some decades [2], and was included in the IUCN Red List of Threatened Species [3]. The phylogenetic position of the family has been the subject of debate for a long time: based on morphologic, molecular, and behavioural data, the family was placed at the base of the Pecora [4], or considered as the sister group of either Cervidae (+ Antilocapridae) [5][6][7][8][9][10], or Bovidae (+Antilocapridae) [11][12][13][14][15][16] (see also discussions in [9,13,17,18] for this controversy). Although a sister group relationship with bovids appears most likely [11,12,14], the phylogeny of Moschidae is not fully understood with arguments still pointing to a closer relationship with Cervidae (see e.g. discussion in [17]). With only one modern genus, a focus on recent representatives provides a very limited data set a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 highly susceptible to homoplastic features. A better knowledge of the detailed morphology of extinct taxa can improve our understanding of the early evolution of the clade and thus help to better constrain its phylogeny, as also proposed by other authors (see e.g. discussion in [17]). For Miocene Moschidae a comprehensive analysis was undertaken on Spanish material, which showed a quite high diversity with at least five different species in two genera, and a rather common co-occurrence of two sympatric moschid species in one locality [15,19,20]. However, the possibly oldest moschid species, and the type species of the genus Micromeryx, Micromeryx flourensianus Lartet 1851, remains undescribed so far.
Traditionally dental material was the main source of information for smaller fossil mammals. However, with improved excavation and analysis techniques more data can also be gained on the skeletal morphology of this group. In particular, the area of the basicranium, including the ear region, is often considered highly relevant to resolve phylogenetic relationship, as it is assumed to be less susceptible to climatic and ecological changes [21]. Prior to non-invasive techniques, such as computed-tomography, studies of the ear region that did not partially destroy the skull were limited, and examinations including data of the inner ear of fossil specimens not possible at all. Consequently, few comparative works considered these data sets. However, during the last couple of years, studies on the petrosal and bony labyrinth have proven highly valuable for phylogenetic, developmental, and (palaeo-)ecologic questions in mammals (see e.g. [22][23][24][25]), and recently the great value of the ear region was demonstrated also for phylogenetic studies on ruminants [26,27]. A combination of data from this region with observation on other skull elements and the dental morphology can greatly improve our view on fossil taxa and help to better decipher ecomorphologic traits and phylogenetic signals.    [34] for details). Of these the upper trochiformis-to lower oxystoma-beds, roughly datable to the Middle Miocene (Serravallian; about 13.5 Ma [29]), are rich in mammal remains (114 vertebrate species described so far [31]).
The first palaeontological investigations on vertebrates from the locality Steinheim a. A. date back to the first half of the 19 th century [35]. They mainly focused on larger mammal remains found during commercial quarrying for sand in the community sandpit (former Pharion sandpit) on the west side of the central uplift of the crater. Scientific excavations carried out by one of us (EH) between 1969 and 1982 provided abundant material housed at the Naturhistorisches Museum Basel and the Staatliches Museum für Naturkunde Stuttgart. The most common large mammals found during these excavations are moschids, including several partial skeletons. The investigations presented here are mainly based on this material. Based on the specimens' sediment infilling, in addition to the field documentation, we can confirm that the material originated from the basal oxystoma-beds: an accumulation of the gastropod Gyraulus oxystoma is still preserved in the skull of NMB Sth.833 (Fig 2; pers. comm. M. Rasser (17.03.2017)) and it is also present in NMB Sth.834 and in NMB Sth.811.

Material and methods Material
The moschid material described in this work, including the new species Micromeryx? eiselei, comes from the Middle Miocene locality Steinheim am Albuch (Germany) and is housed in the Staatliches Museum für Naturkunde Stuttgart and the Naturhistorisches Museum Basel (the specimens referred to in this publication do not comprise the entire moschid material from Steinheim am Albuch, but are a selection with focus on cranial material; postcranial material exists and will be the focus of another study; for referred material see respective sections in the chapter Systematic Palaeontology). All necessary permits were obtained for the described study, which complied with all relevant regulations.
In order to understand the morphological traits in the moschid taxa from Steinheim a. A., different ruminant taxa were included in the comparison and the phylogenetic analysis.
For For SMNS 40500-1, SMNS 40252, SMNS 46123, data for the ear region were acquired by scanning with x-ray computed tomography at the Staatliches Museum für Naturkunde Stuttgart using a Bruker Skyscan 1272 equipped with a 20-100kV / 10W x-ray source.
3D reconstructions of inner ears and petrosals were achieved using the segmentation editor of software AVIZO1 7.0 and AMIRA1 6.2. Difference in resolution only has minor impact on the reconstructions of petrosals or on the segmentation of inner ears. Three-dimensional data are provided by [47].
Measurements of teeth were done with digital callipers. Precision of measurements is 0.2 mm.
See supporting information for dental measurements. The phylogenetic analysis presented here is based on a matrix of 50 characters, comprising 10 characters from the petrosal bone, 11 characters from the bony labyrinth, 19 characters from the cranium and mandibula, and 10 dental characters. We combined characters traditionally used to structure the basal nodes of the phylogenetic tree of ruminants with characters diagnostic for the moschid material from Steinheim a. A. The main purpose was to test the phylogenetic relationships of the two taxa from Steinheim a. A. with the extant moschid Moschus moschiferus. The analysis was performed using WinClada [48]. All characters were equally weighted without any ordering. All multistate characters were treated as unordered. We ran an exhaustive search which resulted in only one tree of 35 steps. See supporting information for the character list and character matrix used in the phylogenetic analysis.
Terminology for the cranium and mandibula follows [49] and [50], for petrosal and inner ear [24], [30], and [27]. For the dentition we refer to [51]. Due to the more common use in recent literature, anterior and posterior are used instead of mesial-distal and rostral-nuchal.
Nomenclatural acts The electronic edition of this article conforms to the requirements of the amended International Code of Zoological Nomenclature, and hence the new names contained herein are available under that Code from the electronic edition of this article. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix "http://zoobank.org/". The LSID for this publication is: urn:lsid:zoobank.org:pub:86EDF102-759B-4A77-83C7-7FAE99074C5A. The electronic edition of this work was published in a journal with an ISSN, and has been archived and is available from the following digital repositories: PubMed Central, LOCKSS. Description and comparison. Generally Micromeryx flourensianus from Steinheim a. A. possessed an elongated and low skull, with a broadened cranium and a facial part distinctly narrowing to anterior, similar to the modern genus Moschus. As the best preserved skull (NMB Sth.834) is strongly compressed estimations on the skull size are enhanced, but it appears to have been only slightly smaller than the modern Moschus moschiferus. A roughly rounded shape can be reconstructed for the orbita (observable in NMB Sth.834 and NMB Sth.811). Its relative size (in ratio to skull size) appears a little larger than in the modern Moschus.
Occipital (Figs 3 and 4). The partes laterales and the squama occipitalis of the occipital can be observed in NMB Sth.834, where they are obliquely compressed. A narrow, but distinct crista nuchae (as in Micromeryx azanzae [19]) appears to meet a well pronounced protuberantia occipitalis externa, comparable to the condition in Moschus and Hispanomeryx daamsi [15]. The area around the condyli occipitales and the shape of the foramen magnum are best preserved in SMNS 42920; however, due to compression the actual shape of the foramen cannot be reconstructed. A clear tuberculum nuchale is present on both sides of the dorsal rim lateral of a median incision. The processus jugularis is triangular and short as in Moschus but slightly more robust than in the latter. It is not as long and slender as e.g. in C. elaphus. Due to compression the size and position of the foramen jugulare cannot be reconstructed. There are two foramina for the nervus hypoglossus. The larger and posterior foramen is positioned more posterior in the fossa condylaris ventralis than it is generally the case in Moschus and C. elaphus. There is no foramen posterior to it in medial view as it is often the case in Moschus and C. elaphus.
The pars basilaris shows a smooth surface, except of the two concave surfaces for the muscle insertion with the distinct posteromedial tuberculi musculares and the posterior protuberantia at the suture to the partes laterales (Fig 3). We could not reconstruct the suture line with the basisphenoid. The tuberculi musculares are more posterior than they are usually in C. elaphus. Sphenoid (Figs 3 and 5). In NMB Sth.811 the ala ossis presphenoidalis forms the lower part of the medial wall of the orbita and meets the frontal here in an s-shaped suture line with the frontal reaching more ventral anteriorly. As in Moschus the ala ossis presphenoidalis is dorsoventrally higher than in C. elaphus, where the palatine more strongly intrudes into the medial wall of the orbita in dorsal direction. A relatively large foramen opticum (of similar size as in Moschus; relative to the skull size larger than in C. elaphus) is located posterior to two smaller foramina. The anteroposteriorly oriented foramen orbitorundum is observable in NMB Sth.811 (Fig 5). Posterolaterally the sphenoid meets the parietal and the temporal. The crest formed by sphenoid, vomer and pterygoid is preserved in ventral view. In NMB Sth.834 this area is heavily crushed, but the rests of the foramen orbitorundum and foramen ovale are still discernible median to a small crest (Fig 3). The latter could represent the weak crest described as diagnostic for Moschidae by Sánchez et al. [15] inside Bovoidea.
Temporal. Pars squamosa (Figs 3 and 4). The processus zygomaticus with fragments of the tuberculum articulare, the fossa mandibularis and the processus retroarticularis is preserved on the pars squamosa in NMB Sth.834. The processus retroarticularis comprises a well pronounced mediolateral crest, as in Moschus. The long and slender processus zygomaticus reaches as anterior as in Moschus. The suture with the jugal cannot be defined clearly as the respective part is heavily broken and covered with glue. The suture line to the parietal can be roughly followed in ventral view of NMB Sth.834, where the lateral part is partly shown due to the oblique compression (Fig 3), and in the lateral view of NMB Sth.811 (Fig 5). The suture line ascends from anterolaterally to posteromedially. It is more ventral than in C. elaphus, but similar to the condition in Moschus.
Pars mastoidea and tympanica (Fig 6). The processus mastoideus is well developed and the temporal forms at least the ventral half of the crista nuchae as in Moschus and Hispanomeryx daamsi [15]. The bulla tympanica in NMB Sth. 834 is more strongly inflated than in Moschus and slightly more than in M.? eiselei. Due to compression it cannot be verified if the bulla tympanica covered the foramen retroarticulare. The ventral surface is rounded. The external  auditory meatus is large and circular. The bulla is not strongly posteriorly expanded, the tympanohyal vagina appears slightly more posterior than in M.? eiselei (that this is influenced by different compression cannot be ruled out however). The lamina vaginalis encloses the processus styloideus at least proximally. Thus the thympanohyal vagina is not fully visible from lateral view. However, as the lamina is fragmented the full extent cannot be reconstructed. The processus styloideus is separated from the processus jugularis of the occipital, comparable to the situation in Moschus. In C. elaphus this feature was variable in the studied specimens. There appears to be no contact between the tympanic bulla and the occipital in M. flourensianus from Steinheim a. A.
Pars petrosa and bony labyrinth (Figs 7 and 8). The petrosal and bony labyrinth of M. flourensianus from Steinheim a. A. were both described in detail by [30]. The specimen NMB Sth.834 was not known back then but its morphology is highly similar to NMB Sth.828a, NMB Sth.865 and NMB Sth.866 with a more ventrally facing basicapsular groove and a bean shaped fossa for tensor tympani. The transpromontorial sulcus is not obvious on the promontorium of NMB Sth.834 but a slight groove towards the epitympanic wing may show it was present. The anterior process of the tegmen tympani is relatively shorter and blunter than in NMB Sth.828a, NMB Sth.865 and NMB Sth.866 and the tegmen tympani itself is slightly more expanded. The dorsomedial aspect is virtually identical. The bony labyrinth is also virtually identical with the ones published before with a relatively shorter vestibular aqueduct and endolymphatic sac than in M.? eiselei (see description below). The aqueduct is slightly curved along the common crus both on NMB Sth.834 and SMNS 46123. The cochlea has 2.25 to 2.  Frontal (Figs 3, 4, 5 and 9). The paired frontals form the interorbital part of the skull roof, as well as the dorsal and most of the medial wall of the orbitae. There are no indications for cranial appendages in any specimen. The supraorbital frontal convexity and the depression anterior to it cannot be reconstructed due to compression. However, based on NMB Sth. 811 they appear similar to the condition in Moschus, Micromeryx azanzae [19] and Hispanomeryx daamsi [15]. The foramen ethmoidale and the part of the skull with the anterior part of the frontal are not preserved in any of the specimens. Presence and extent of the fontanella nasolacrimalis (etmoidal vacuity in [15]; antorbital vacuity in [10]) cannot be reconstructed, therefore.
A well pronounced depression (oriented anteroposteriorly; posteriorly overlapping with the suture line to the sphenoid) is preserved in the orbita of NMB Sth.811. The dorsal rim of the orbita is more filigree than in M.? eiselei. The foramina supraorbitalia are not as distinct in M. flourensianus from Steinheim a. A. as they are in M.? eiselei. A weak sulcus supraorbitalis is preserved in M. flourensianus from Steinheim a. A. (see NMB Sth.811). The postorbital bar is preserved fragmentarily. As in Moschus the frontal seems to form the dorsal 2/3 of the bar. It appears slightly wider mediolaterally than in Moschus.
Ethmoid. Fragments of the ethmoid can be observed in NMS Sth.811 and in the CT scan of NMB Sth.834 though specific characteristics of the bone are not discernible. Nasal (Fig 9). Fragments of the nasal are preserved in NMB Sth.834 but the precise morphology cannot be reconstructed.
Lacrimal (Figs 3, 4, 5 and 9). A fragmented and compressed lacrimal is preserved in NMB Sth.834 and NMB Sth. 811. It forms the anterior rim of the orbita. The facies facialis is slightly concave, but, as in Moschus and Micromeryx azanzae [19], it does not form a distinct fossa lacrimalis. On the facies orbitalis a small fossa sacci lacrimalis houses one foramen lacrimale just inside the orbit, comparable to the condition in Moschus.
Jugal (Figs 3 and 4). Shape and size of the jugal cannot be fully reconstructed due to compression. It forms the ventral rim of the orbit. Like in Moschus and Micromeryx azanzae [19] the crista facialis is well marked. It is stronger than in C. elaphus. The participation of the jugal on the postorbital bar cannot be estimated reliably; we nonetheless found no indication that it exceeds more than 1/3 of the total bar in any of the specimens. The processus temporalis is preserved on the right side of NMB Sth.834. It appears to reach not more than the anterior 1/3 of the zygomatic arch and is laterally overlain by the temporal.
Maxilla (Figs 3, 4, 9 and 10). The anterior part of the maxilla is not preserved for M. flourensianus from Steinheim a. A. The posterior part of the facies facialis is smooth and slightly convex in NMB Sth.834. It meets the ventral side anterior of the tooth row forming a sharp, delicate crest. Due to compression it cannot be verified, if the crista facialis proceeds onto the maxilla. There appears to be a small foramen at the height of the posterior root of the P2  (observable in NMB Sth.834; Fig 4). It might be the infraorbital foramen. As it is rather small and as the area anterior to the P2 is fractured it cannot be ruled out, however, that a larger infraorbital foramen is located more anterior, as it is the case in Micromeryx? eiselei. The lacrimal-jugal suture lines are concealed due to compression; they roughly show an anterodorsal to posteroventral course. The facies pterygopalatina is laterally compressed in NMB Sth.834, but was not wider than 6-10 mm anterior of the P2. The maxillae meet along the midline and suture with the palatines at a curved line running from the midline at about the height of the P4 posterolaterally to the height of the M3 (well observable in SMNS 15776 (Fig 10); hardly in NMB Sth.834 (Fig 3)). Thus the palatine protrudes more anteriorly in this taxon than in Moschus, where the palatine reaches only as far as the M1.
Praemaxilla. The praemaxilla is not recorded for M. flourensianus from Steinheim a. A. Palatine (Figs 3 and 10). Anteriorly bordering the maxilla the lamina horizontalis of the palatine forms the posterior part of the bony palate including the anterior rim of the choanae. As mentioned above the palatine reaches further anterior than in Moschus. A fairly large foramen palatinum majus (observable in SMNS 15776 (Fig 10)) is positioned at the anterior suture line with the maxilla, as is the case in Moschus (often reaching further anterior on the maxilla in this genus, however). This area is strongly fractured in NMB Sth.834 but a smaller foramen median to the foramen palatinum majus can still be seen. Several small foramina on the palatine are preserved in SMNS 15776. In NMB Sth.834 and SMNS 15776 the posterior rim of the lamina horizontalis is rounded (concave to anterior). Laterally it meets the maxilla at the height of the M3 (see description of maxilla for general course of suture line), medially the lamina perpendicularis penetrates posteriorly to about the height of the foramen opticum, forming the lateral wall of the choanae. As in Moschus (observable in NMB Sth.811; Fig 5) it is less high than in C. elaphus, where it reaches further dorsal than in moschids. Fragments of the fossa pterygopalatina are preserved posteromedially of the tooth row in NMB Sth.834.
Pterygoid. Fragments of the pterygoid are preserved in NMB Sth. 834. No statement can be given on extent and morphology however.
Vomer. In NMB Sth.811 possible fragments of the vomer are still preserved. Statements on shape and posterior extent cannot be given.
Mandibula (Fig 11). Micromeryx flourensianus from Steinheim a. A. possesses a slender corpus mandibulae with a long convex ventral side as in Moschus and C. elaphus (see e.g. SMNS 46123). SMNS 12966 shows a fragmented pars incisiva widening to anterior. The symphysis mandibulae is preserved only fragmentarily; its length is about 30 to 40% of the length of the lower post canine tooth row. The relative length of the margo interalveolaris appears shorter than in Moschus (line measured from anterior rim of p2 to posterior rim of symphysis about 40 to 50% of the lower post canine tooth row in M. flourensianus from Steinheim a. A., while it is usually about 50 to 60% in Moschus). The main foramen mentale, opening anteriorly, is large and situated at the height of the posterior rim of the symphysis mandibulae or slightly posterior to it on the facies labialis, as in Moschus. M. flourensianus from Steinheim a. A. often possesses a second, smaller foramen mentale situated at the anterior part of the premolar row or just anterior to it (observable in NMB Sth. 804, SMNS 12966, SMNS 46123, and SMNS 46116). In Moschus, this smaller foramen is, when present, positioned more anteriorly, at the diastema just posterior to the larger one. The incisura vasorum facialium is fragmented in every specimen. Shape and extent of the angulus mandibulae is comparable to the condition in Moschus. The ramus mandibulae is relatively long in anteroposterior direction, as is the processus coronoideus, in comparison to Moschus. The processus condylaris possesses a short collum mandibulae and thus the incisura mandibulae is shallow. The articulation facet is often preserved obliquely due to compression. Nevertheless, it was stronger inclined than in Moschus. In respect to the processus coronoideus the processus condylaris reaches further laterally than in C. elaphus and in Hispanomeryx [15], as is also the case in Moschus. Fossa pterygoidea and foramen mandibulae are not preserved in any of the specimen.
Dentition (Figs 3, 11 and 12). Micromeryx flourensianus from Steinheim a. A. shows a typical moschid dentition with a closed or nearly closed anterior valley in the triangular lower p4 and a bicuspid third lobe with a high entoconulid in the lower m3. The tooth formula is 0133/ 3133. In size the teeth are in the upper range of M. flourensianus from Sansan, or slightly larger, but smaller than M.? eiselei.
The upper canines are elongated and saber like in males (Fig 12E), and smaller and teardrop shaped in females. The upper canines of the males are smaller than in M.? eiselei and clearly shorter than in the modern Moschus. The anterior upper premolars are elongated triangular with two lingual cones and relatively pronounced labial styles and ribs (Fig 12A). The P3 is often shorter and generally wider than the P2. In comparison to M. flourensianus from Sansan there is a tendency in the reduction of the lingual crown elements. The P4 is shorter and wider than the anterior premolars and possesses a relatively flat labial wall with moderately pronounced elements comparable to the condition in Sansan. In most of the specimens in  [15]. The upper molars (Fig 12A and 12C) are selenodont, trapezoid to subquadratic in shape and possess four main cusps. The morphology is comparable to M. flourensianus from Sansan. A tendency to increased complexity can be observed with the occurrence of additional crests in many specimens, however. The lingual walls in M. flourensianus from Steinheim a. A. are stronger inclined than in Hispanomeryx and the ribs and styles are not column like, as it is usually the case in the latter [15]. The rib at the metacone and the metastyle are weak.
There is a disparity in the incisor arcade with a clearly widened i1, a slightly widened i2 and less widened i3 and c (visible in the isolated i-c of NMB Sth.804 (Fig 12I)) in contrast to the modern Moschus, where the incisor arcade is more homogenous. The lower canine is incisiform in M. flourensianus from Steinheim a. A.
The lower premolar tooth row is less shortened in M. flourensianus from Steinheim a. A. than in Moschus and Hispanomeryx [15]. In comparison to M. flourensianus from the type  [15,19,20] and personal observation). The tooth crown height, best observable in the lower molars, is slightly increased in M. flourensianus from Steinheim a. A. in comparison to the material from Sansan; however, the differences are gradual in this case as well and there is no distinct difference between the two assemblages. The lingual wall of the lower molars in M. flourensianus from Steinheim a. A. is less aligned and more bulky than in Hispanomeryx. Furthermore, the lingual elements are more pronounced than in the latter [15]. The anterior cingulid is weaker and lower than in Moschus and not fused with the mesostylid as in the latter. For more details on the differences in the dentition to M.? eiselei see below in description of this species.
Diagnosis. With a closed anterior valley in the triangular lower p4 and a bicuspid third lobe with a high entoconulid in the lower m3 Micromeryx? eiselei shows a typical moschid dentition. The species can be defined by a unique combination of morphometrical characters: It has a large size, possesses a well pronounced external postprotocristid, a distinct lateral extent of the processus coronoideus, and a relatively wide lower p4. It differs from M. flourensianus from Sansan and from Steinheim a. A. in a larger size, generally slightly higher tooth crowns, a slightly shorter premolar row in relation to the molar tooth row, a stronger relief in the labial wall of the upper P4, a more elongated posterior part of the upper P3, and more pillar-like elements on the labial wall of the upper molars. From M. flourensianus from Steinheim a. A. it furthermore differs in a crest on the pars basilaris of the occipital, a more slender and more elongated processus jugularis, a less rounded tympanic bulla, as well as in the morphology of the petrosal and of the inner ear. M.? eiselei clearly differs from M. azanzae, M. soriae and M. mirus in size and the presence of a strong external postprotocristid. From M. styriacus the species differs by the relatively wider lower p4. From species of the genus Hispanomeryx it differs in the lateral extent of the processus coronoideus, a less shortened premolar row, the presence of an external postprotocristid, and the more bulky labial wall of the lower molars as well as in the shape of p4. M.? eiselei differs from Moschus in the arrangement of the foramina for the nervus hypoglossus, the morphology of bulla tympanica, the morphology of the petrosal, the course of the suture line between maxilla and palatine, the position of the foramina mentalia, a less shortened premolar row, the morphology of the p4, the presence of an external postprotocristid in the lower molars, a lower anterior cingulid, and the less increased tooth crown height. Description and comparison. Micromeryx? eiselei possessed an elongated and low skull shape, with a broadened cranium and a facial part distinctly narrowing to anterior similar to Micromeryx flourensianus from Steinheim a. A. and the modern genus Moschus. It was only slightly smaller than the modern Moschus moschiferus.
Occipital (Figs 13-18). Parts of the squama occipitalis and of the partes laterales are preserved in NMB Sth.833. A relatively well marked crista sagittalis externa meets the protuberantia occipitalis externa, which is fragmented in NMB Sth.833. As in M. flourensianus from Steinheim a. A. a narrow, but distinct crista nuchae proceeds from this point ventrolaterally. Although it is difficult to verify due to fragmentation, the dorsal surface of the condyli occipitales appears slightly more inclined anterodorsally-posteroventrally than in Moschus. As in Moschus and Hispanomeryx daamsi [15] the ventral plane of the condyli is extended more ventrally than the pars basilaris of the occipital. A clearly distinct tuberculum nuchale is positioned on each side of the dorsal rim of the rhomboid shaped foramen magnum in NMB Sth.833. The small foramen jugulare is situated posteromedial of the bulla and not covered by it. As in M. flourensianus from Steinheim a. A. two foramina are observed in the fossa condylaris ventralis (the latter is quite deep in M.? eiselei; Fig 16D). And as in M. flourensianus from Steinheim a. A., the posterior foramen for the nervus hypoglossus is the larger one and is positioned more posterior than in Moschus. The processus jugularis is slightly fragmented, but appears to be more slender and elongated than in M. flourensianus from Steinheim a. A. and stronger developed than in Moschus. The surfaces for the muscle insertion are more posterior than it is usually the case in C. elaphus. The posteromedial tuberculi musculares and the posterior protuberances at the suture to the partes laterales are stronger developed than in M. flourensianus from Steinheim a. A. (clearly observable in NMB Sth.833). We could not reconstruct the suture line with the basisphenoid. A well pronounced median crest runs along the midline of the pars basalis in M.? eiselei as in adult Moschus specimens (juveniles with deciduous dentition (NMB Sth.5111 and 5731) do not show this crest). As this can be observed in both specimens of M.? eiselei (SMNS 40010: young female; NMB Sth.833: old male) the feature can neither be ascribed to sexual dimorphism nor to ontogenetic change but appears to be diagnostic for M.? eiselei in comparison to M. flourensianus from Steinheim a. A., in specimens with permanent dentition. Sphenoid (Fig 13). This part of the skull is heavily fragmented in all specimens of M.? eiselei and no detailed diagnosis can be given on the morphology. The position of the foramen orbitorundum and ovale seem to be roughly similar to the condition in M. flourensianus from Steinheim a. A.; the presence or absence of a distinct crest in this region cannot be verified, however.
Temporal. Pars squamosa (Figs 13-16). The pars squamosa of the temporal is heavily fragmented and fractured in NMB Sth.833 and SMNS 40010. The course of the suture line to parietal and occipital on the lateral and posterior part of the skull can be reconstructed only arbitrarily; the first ascends from anterolaterally to posteromedially as in M. flourensianus Pars mastoidea and tympanica (Figs 6 and 16). The mastoid process is well pronounced and the temporal forms at least the ventral half of the crista nuchae as in M. flourensianus from Steinheim a. A., Moschus and Hispanomeryx daamsi [15]. As in M. flourensianus from Steinheim a. A. the bulla in NMB Sth.833 is more strongly inflated than in Moschus, the ventral surface is rounded, laterally the surface is plainer than in M. flourensianus from Steinheim a. A. The foramen retroarticulare is covered by the inflated bulla tympanica in NMB Sth.833. As the specimen is compressed, the original situation is difficult to reconstruct, however. In any case, the foramen is quite large, as can be observed on the right side, where the bulla is not preserved in situ. The external auditory meatus is large and subcircular. There is no significant posterior expansion of the bulla, and the tympanohyal vagina is located subposteriorly. The lamina vaginalis does not enclose the processus styloideus and thus the thympanohyal vagina is visible from lateral view. As the lamina is in general quite delicate, we cannot exclude loss due to taphonomic processes, however. The processus styloideus is separated from the processus jugularis of the occipital, comparable to the situation in Moschus. As far as it can be reconstructed there is no contact between the tympanic bulla and the occipital. In NMB Sth.833 the anterior wall of the right bulla is preserved as an isolated fragment (Fig 6C). It allows the observation of the inner structure of the bulla. The opening to the external auditory meatus can be observed dorsal of the sulcus tympanicus. Ventrally the hypotympanicum, the tuba eustachii, and the processus muscularis are still preserved. The structure of the hypotympanicum is not cellular.
Pars petrosa and bony labyrinth (Figs 8 and 19). No isolated petrosal is attributed to M.? eiselei. Its morphology was reconstructed from CT scans of NMB Sth.833 and SMNS 40010. As SMNS 40010 is heavily broken, most of the information comes from NMB Sth. 833. With a general elongated shape and a pointed apex on the epitympanic wing the petrosal is similar in morphology to M. flourensianus from Steinheim a. A., though slightly less wide in ventromedial-dorsolateral direction. The promontorium has a biconvex hemiellipsoid shape. In contrast to the condition in M. flourensianus from Steinheim a. A. no weak transpromontorial sulcus can be observed in NMB Sth.833 across the promontorium. Although it is rather unlikely, we cannot fully rule out that the missing of the shallow sulcus could be related to its taphonomical history, as the bone surface appears quite smoothed in the specimen in general. The fenestra cochleae is oval shaped and more than twice as large as the fenestra vestibuli. The crista interfenestralis is of similar width as in M. flourensianus from Steinheim a. A. The fossa for the tensor tympani is large, deep, elongated (but not as long and narrow as in Moschus; see [30]), and   The bony labyrinth of NMB Sth.833 is described in [30] and ascribed to M. flourensianus. Better knowledge of variability in this structure acquired since the first description shows that it is clearly different from M. flourensianus from Steinheim a. A. described above. The main differences are found in the vestibular aqueduct, which is here longer, straighter, and surmounted by a furrow shaped and long endolymphatic sac, a very different situation than the shorter, curved aqueduct and short pouch shaped sac in M. flourensianus from Steinheim a. A. The cochlear aqueduct is short, massive and slightly more anteriorly positioned than in M. flourensianus from Steinheim a. A.
Parietal (Figs 13-15 and 18). The parietal is preserved fragmentarily in NMB Sth.833 and SMNS 40010. The lineae temporales are stronger than in M. flourensianus from Steinheim a. A. In NMB Sth. 833 the lineae temporales converge posteriorly forming a strong crista sagittalis externa, which proceeds on the occipital. The crista is shorter than in and not as distinct as in Dremotherium (see discussion and Fig 2 in [52] for details) but comparable to the condition in Moschus, when a crista is present in the latter. However, presence, extent and strength of this structure varies in Moschus (a clearly distinct sagittal crest was only observed in SMNS 1238). Interestingly Sánchez and Morales [19] describe a well distinct crista sagittalis in a juvenile male of Micromeryx azanzae, whereas in Hispanomeryx daamsi a crest is only present in terms of the protuberantia occipitalis externa [15].
Frontal (Figs 14, 15 and 18). The paired frontal bones form the interorbital part of the skull roof, as well as the dorsal and most of the medial wall of the orbitae. As in M. flourensianus from Steinheim a. A., there are no indications for cranial appendages in any specimen. The supraorbital frontal convexity and the depression anterior to it cannot be reconstructed in detail due to compression. As in M. flourensianus from Steinheim a. A. they appear similar to the condition in Moschus, Micromeryx azanzae [19], and in Hispanomeryx daamsi [15]. The foramen ethmoideum and the part of the skull with the anterior part of the frontal are not preserved in any of the specimen. Therefore, the presence and extent of the fontanella nasolacrimalis (etmoidal vacuity in [15]; antorbital vacuity in [10]) cannot be reconstructed here as in M. flourensianus from Steinheim a. A.
The dorsal rim of the orbit is more bulgy in M.? eiselei (observable in NMB Sth.833) than in M. flourensianus from Steinheim a. A. As NMB Sth.833 is an older individual, it cannot be excluded that this feature results from its advanced age. The postorbital bar (Fig 16C) is clearly wider than in Moschus and M. flourensianus from Steinheim a. A. The depression posterior to the postorbital bar, ventral to the origin of the linea temporalis, is mediolaterally wider than in Moschus. There are two foramina supraorbitalia. The anterior one is smaller and forms a short oblique canal leading into the orbita, while the posterior one is larger and directly opens into the orbita. A distinct sulcus supraorbitalis as in M. flourensianus from Steinheim a. A. cannot be observed in M.? eiselei.
Ethmoid. Fragments of the ethmoid can be seen in the CT scans of NMB Sth.833. Specific characteristics of the bone cannot be described. Nasal (Fig 18). Fragments of the nasal are preserved in SMNS 40010. They are elongated and increase in lateromedial width to anterior. Otherwise, not much can be said on the morphology.
Lacrimal (Figs 14, 15 and 18). The facies facialis of the lacrimal is smooth and slightly concave. It forms the anterior rim of the orbita (preserved in NMB Sth.833). Due to compression and fragmentation the morphology of the facies orbitalis cannot be reconstructed nor can the position and number of foramina lacrimales be verified. However, there is no indication for the presence two foramina on the orbital rim as in cervids.
Jugal (Figs 13-15, 17 and 18). As in M. flourensianus from Steinheim a. A. the detailed shape and size of the jugal cannot be fully reconstructed due to compression. It forms the ventral rim of the orbit. Like in Moschus, Micromeryx azanzae [19], and M. flourensianus from Steinheim a. A. the crista facialis is well marked. It is stronger than in C. elaphus. It reaches posteriorly as far as the processus temporalis as in M. flourensianus from Steinheim a. A. The jugal does not form more than 1/3 of the postorbital bar in M.? eiselei (NMB Sth.834). As only a small part of the processus temporalis of the jugal is preserved, its share on the formation of the temporal bar cannot be reconstructed in detail.
Maxilla (Figs 13-15, 17 and 18). The anterior part of the maxilla is preserved in SMNS 40010 comprising both canines. As SMNS 40010 is a female, the canines are smaller and shorter than in NMB Sth. 833. In ventral view, the anterolateral suture line with the praemaxilla is as well observable in SMNS 40010. Anteriorly the maxillae protrude in between the pro- Praemaxilla (Figs 17 and 18). This element is only preserved in SMNS 40010. In ventrolateral view, it meets the maxilla in a straight line running anteromedial to posterolateral at the height of the canines. It forms the anterior and medial wall of the alveole for the canine and the lateral and medial wall of the fissura palatina. Only fragments of the ventromedial processi palatini are preserved.
Palatine (Figs 13 and 17). Anteriorly bordering the maxilla, the lamina horizontalis of the palatine forms the posterior part of the bony palate including the anterior rim of the choanae. As mentioned above the palatine reaches further anterior than in Moschus (see chapter on maxilla for general course of suture line). The suture line with the maxilla meets the concave posterior rim of the bony palate at the height of the M3. In posterior view the fossa pterygopalatina can be observed dorsally of the palate; position and relative size are similar to the condition in Moschus.
Vomer (Fig 13). Possible fragments of the vomer are preserved in NMB Sth.833. Statements on shape and posterior extent cannot be given. Mandibula (Fig 11). A fragmented mandibula is preserved of M.? eiselei (specimen: SMNS 40010). As in Moschus, C. elaphus, and M. flourensianus from Steinheim a. A., the corpus mandibulae is slender; the ventral side is slightly less convex than in M. flourensianus from Steinheim a. A., Moschus and C. elaphus. Only a fragmented pars incisiva is preserved of M.? eiselei. As in M. flourensianus from Steinheim a. A. it shows a widening of the pars incisiva to anterior. The length of the symphysis mandibulae and of the margo interalveolaris cannot be estimated. As in M. flourensianus from Steinheim a. A., a smaller foramen mentale is situated just anterior of the p2, distinguishing this taxon as well from Moschus, where this foramen is more anterior, when present. The incisura vasorum facialium seems to have been shallower then in Moschus. The angulus mandibulae is fragmented. Shape and extent appear similar as in Moschus, however. As in M. flourensianus from Steinheim a. A. the articulation facet is inclined stronger than in Moschus. The processus condylaris reaches further laterally than in C. elaphus and Hispanomeryx [15], as is the case in M. flourensianus from Steinheim a. A. and Moschus. Fossa pterygoidea, foramen mandibulae and processus coronoideus are not preserved in any specimen assigned to M.? eiselei.
Dentition. Micromeryx? eiselei possesses a typical moschid dentition with a closed or nearly closed anterior valley in the triangular lower p4 and a bicuspid third lobe with a high entoconulid in the lower m3. The tooth formula is 0133/3133. The teeth of M.? eiselei are larger than in M. flourensianus from Steinheim a. A. and from the type locality Sansan (Fig 20). Furthermore, the dentition of the female in M.? eiselei is larger than the one of the male. This indicates a sexual dimorphism with a larger size in females, as it often occurs in ruminants with a small body size. The strong attrition of the male dentition superposes a clear signal, however. With the description of the postcranial skeleton we hope to gain a better idea about sexual size dimorphism in M.? eiselei.
As in M. flourensianus from Steinheim a. A. the upper canines are elongated and saber like in the males (Fig 12E), and smaller and teardrop shaped in the females (Figs 17 and 18). They are shorter than in Moschus but longer than in M. flourensianus from Steinheim a. A.
The anterior upper premolars are elongated triangular with two lingual cones and pronounced labial styles and ribs (Fig 12B). The P3 is often shorter and generally wider than the P2. In the P3 the posterolabial cristid is more elongated in M.? eiselei than in M. flourensianus from Steinheim a. A. The P4 is shorter and wider than the anterior premolars, shows only one distinct lingual cone, and due to strong labial styles and a weak rib at the labial cone a concave labial wall. This distinguishes M.? eiselei from M. flourensianus from Steinheim a. A., which shows a more flat labial wall and a tendency to split the lingual wall in anterior and posterior cone. The upper premolars are more elongated than in Moschus and Hispanomeryx [15]. As in M. flourensianus from Steinheim a. A. the upper molars (Fig 12B and 12D) are selenodont and possess a trapezoid to subquadratic shape with four main cusps. The lingual walls are inclined more strongly than in Hispanomeryx and the labial ribs and styles are stronger than it is usually the case in the latter [15]. As in other Micromeryx taxa the enamel on the labial wall of the lingual elements is well pronounced, while it is much weaker in the genus Hispanomeryx. In contrast to M. flourensianus from Steinheim a. A. the styles at the labial wall of the upper molars are more pillar-like in M.? eiselei and the rib at the metacone as well as the metastyle are stronger.
In the fragment of the symphysis of SMNS 40010 it can be observed that the i1 is clearly widened, i2, i3, and c are less wide (Fig 12H). This distinguishes this taxon from the modern Moschus, where the incisor arcade is more homogenous, as mentioned above. The lower canine is incisiform in M.? eiselei, as it is in M. flourensianus from Steinheim a. A.
Although the premolar tooth row is slightly more shortened in M.? eiselei than in M. flourensianus from Steinheim a. A., it is not as shortened as in Hispanomeryx [15] and in Moschus.  [15,19,20] and personal observation). The lingual walls are less flattened and less aligned than in Hispanomeryx. The metastylid is slightly stronger than in M. flourensianus from Steinheim a. A. and in comparison to the latter the anterior cingulid is stronger on the lingual wall. As in M. flourensianus from Steinheim a. A. the anterior cingulid is weaker and lower than in Moschus and not fused with the mesostylid, however.

Phylogenetic analysis
In order to test the phylogenetic relationships of the two taxa from Steinheim a. A. with the extant moschid Moschus, we performed a preliminary phylogenetic analysis with a focus on these taxa (Fig 21). This resulted in a single most parsimonious tree (CI = 0.79; RI = 0.8; for each node, the list of unambiguous synapomorphies is given in Fig 21). Comparable to the results in [27], the monophyly of the clade Ruminantia comprising the tragulids as the sistergroup of the Pecora was confirmed.The Pecora are represented by a cervid clade and a "hornless ruminant clade", composed by Moschus and the taxa form Steinheim a. A. This "hornless ruminant clade" could represent the moschid family. The two taxa from Steinheim a. A. and Moschus share the following synapomorphies: an elongated and bean-shaped fossa for the tensor tympani muscle in the petrosal; a less developed dorsal extent of the palatine and a higher sphenoid; a lower parieto-temporal suture; a stronger inclined articulation facet of the mandibula. The two taxa from Steinheim a. A. are monophyletic, being the sister-group of the modern genus Moschus, and defined by the following characters: anterior process of the tegmen tympani broad and triangular; maxilla-palatine-suture line on bony palate reaching as anterior as or more anterior than P4; a weaker metastylid.

Discussion on the taxonomical attribution of the Steinheim a. A. Moschidae
The comparison of the moschids from Steinheim a. A. with other Miocene Moschidae is based primarily on dental material, as comparative descriptions of cranial elements and data from the ear region are limited.

M. flourensianus from Steinheim a. A.
Taking into consideration data from the literature [3,19,20,46,[53][54][55] and personal observation, M. flourensianus from Steinheim a. A. differs from M. flourensianus from the type locality Sansan by an increased tooth size, shorter and more bulky premolars, a more complex lower p4 and upper P4, a tendency towards a more complex pattern in the upper molars, as well as a tendency to increase tooth crown height. However, there is size overlap between the two localities, and the morphology of the dentition is variable within the population from Steinheim a. A. independent of tooth size. With the pending revision of the type material (in preparation) and based on the gradual and indistinct nature of the observed differences in the dentition between M. flourensianus from Sansan and from Steinheim a. A. we currently cannot justify a separation of the material from Steinheim a. A. from the species M. flourensianus from Sansan. We therefore assign the specimens from Steinheim a. A. to an "evolved" morphotype of the type species M. flourensianus for the moment. There are no cranial elements recorded for the type locality.
In any case, M. flourensianus from Steinheim a. A. differs by size and morphology from the other Micromeryx-species. M. mirus is smaller, possesses higher crowned dentition, and shows a strongly reduced to non-existent external postprotocristid (personal observation and data from [53] and [55]). M. styriacus is larger (the validity of this taxon is still unclear [19,53]). M. azanzae does not possess an external postprotocristid [19]. M. soriae is similar in size to M. flourensianus from Steinheim a. A. but has a broader external postprotocristid, so far uniquely described for the genus [20]. M. flourensianus from Steinheim a. A. is smaller than Hispanomeryx, and also differs in the presence of an external postprotocristid (not present in any species of the genus Hispanomeryx), and the stronger lateral extent of the processus coronoideus [15].

Defining Micromeryx? eiselei
In summary the morphologic differences between M. flourensianus from Steinheim a. A. and M.? eiselei in the cranium, mandibula, dentition, and especially in the ear region are too large to ascribe them to intraspecific variability. Both taxa are furthermore represented by both sexes and different ontogenetic stages. These differences are sufficient to clearly distinguish between the two species.
The main differences are: Micromeryx? eiselei from Steinheim a. A. cannot be assigned to any known Micromeryx-or Hispanomeryx-species. It possesses a unique combination of characters in cranial, mandibular, and dental morphology: large size of teeth, the presence of a strongly pronounced external postprotocristid, a distinct lateral extent of the processus coronoideus, and a wider p4 than that observed in M. styriacus. Therefore, we consider it reasonable to erect a new species, although the postcranial material has not been analysed yet. M.? eiselei clearly differs from Moschus in the morphology of the cranium and the dentition. Interestingly, the characters shared by M.? eiselei and Moschus are only characters of the ear region and the basicranium, which are considered to be less plastic than e.g., dental data. However, considering the evolution of Pecora, some characters of the ear region are plesiomorphic [27], and thus non-informative for the relative position of these two species in the phylogenetic tree. These features are still an indication, however, that with more data of different Moschidae from the basicranium and ear region, combined with data on the complete skeleton, the sister group relationship of M. flourensianus from Steinheim a. A. and M.? eiselei might prove less stable than it now appears. However, the revision of the type material of M. flourensianus as well as a full analyses of the postcranial and the ear regions of more fossil Moschidae is in progress, but not yet completed. Therefore, we can assign M.? eiselei to a new species but not to a new genus, and thus decided with reservations to leave it in the genus Micromeryx at present.

Conclusion
The small ruminant remains from the Middle Miocene locality Steinheim a. A. can be assigned to Miocene Moschidae closely related to the living genus Moschus. Dental and cranial characters as well as features of the ear region reveal the presence of at least two species at the locality.
The smaller of the two, Micromeryx flourensianus from Steinheim a. A., is similar to the type material of M. flourensianus from Sansan, but differs by the presence of an "evolved" dentition. The larger taxon, Micromeryx? eiselei clearly differs from all Miocene Moschidae and is therefore referred to a new species.
The two moschid taxa described here from Steinheim a. A. originate from the same layer in the stratigraphic column based on the biostratigraphy of the embedded gastropods. The fossil assemblage shows a high percentage of semi-articulated partial skeletons and is embedded in lacustrine sediments, it did not result from a fluvial accumulation and represents an autochthonous taphocoenosis. Thus, independent of the further taxonomic status, there are two sympatric moschid species recorded from Steinheim a. A. This confirms once more that the cooccurrence of two moschid species is rather a more common than an exceptional phenomenon for the Miocene. However, it is the first time that this observation is made with unambiguous remains for European material outside of Spain (for the locality Gratkorn (late Middle Miocene, Austria) fragmentary teeth were tentatively assigned to? Hispanomeryx due their size and described as sympatric with Micromeryx flourensianus by Aiglstorfer et al. [46]; with the here published remains from Steinheim am Albuch we show that there are more large sized moschids in the Miocene of Europe; thus the cautiousness given in the cited publication is fully justified and the teeth could represent a different taxon after all).
The exhaustive description of the morphology of the ear region combined with data on cranial, mandibular, and dental material represented here is the first step towards a better understanding of the possibly oldest moschid species, M. flourensianus and the Central/Western European record of the family. Combining the data published here with the study of the type material, and postcranial remains, as well as with data for the ear region of more Miocene Moschidae will improve the knowledge on the early evolutionary history of the family and the origin of the recent genus Moschus.