Late Devonian vertebrates from Siberia: a synchrotron microtomography study of bone bed material

(1)

Late Devonian vertebrates from Siberia: a synchrotron microtomography study of bone bed material

Étienne Fortier-Dubois

Degree project inbiology, Master ofscience (2years), 2016 Examensarbete ibiologi 30 hp tillmasterexamen, 2016

Biology Education Centre and Department ofOrganismal Biology, Uppsala University

(2)

Abstract

This is an investigation of new vertebrate fossil material from the Late Devonian locality of Ivanovka, Uryup River, Siberia. This bone bed material, circa 375 million years in age, represents a unique opportunity to fill a gap in our understanding of Late Devonian di- versity, biogeography, and vertebrate evolution: Siberia, at the time, was an independent continent, and yet its fauna remains virtually unknown in comparison with the other paleocontinents, Euramerica and Gondwana. Using synchrotron microtomographic scanning, a non- destructive technique that has never, to our knowledge, been applied to bone bed material, we obtained 3D image stacks that were then modelled to yield triangle meshes representing the bones in three di- mensions. These meshes could then be identified, described, and interpreted. Many of the discovered bones belong to the poorly known genus Megistolepis Obruchev 1955, potentially allowing a radical increase in knowledge regarding this taxon. Other material includes lungfish and possible fragments of limbed tetrapods, though the evidence of the latter is scarce. A discussion of the advantages and disadvantages of synchrotron microtomography for the study of bone bed material con- cludes the paper.

(3)

Introduction

Background

The world of the Late Devonian, roughly 383 to 359 million years ago, contained three major paleocontinents (Fig. 1): Euramerica, comprising modern Europe, North America and Greenland; Gondwana, comprising South Amer- ica, Africa, Arabia, India, Antarctica and Australia; and Siberia, corresponding to parts of the current Russian region of that name. While the first two have yielded well-studied vertebrate assemblages, including the faunas of East Greenland [7], Canowindra, Australia [1] and the Gogo Formation, Australia [20], very little is known about the vertebrates of Late Devonian Siberia.

This lack of knowledge precludes complete understanding of Devonian biogeography. In the Early and Middle Devonian world, vertebrate faunas were highly regionalized; by the Late Devonian, the worldwide fauna had become homogenized, as evidenced by a wealth of non-marine and marginal marine fish fossils from both Gondwana and Euramerica [21]. Thus there must have been an increase in faunal interchange between these two paleocontinents, but whether this also happened in Siberia is an open question. This exploratory study of vertebrate material collected in Ivanovka, Southern Siberia, is an attempt to answer it.

Popularly dubbed the ‘Age of Fishes’, the Devonian period has special sig- nificance in the history of vertebrate life, as it is then that the first tetrapods evolved [10, 23], and the first complex land ecosystems, with recognizable

(5)

forests, developed [25]. The earliest body fossils of limb-bearing vertebrates are Late Devonian in age, with some trackway fossils suggesting that they first evolved in the Middle Devonian [23]. However, the first limbed tetrapods were immediately preceded by similar animals that still had fins rather than limbs; together, they form the ‘stem-group tetrapods’ [10]. A ‘stem-group’

consists of all extinct forms more closely related to its crown-group (here, the extant tetrapods) than to the closest extant group (here, the lungfish or Dip- noi). Thus, the stem-group tetrapods are extinct sarcopterygian (lobe-finned fish) forms that represent the transition between lungfish and tetrapods.

They diversified mostly during the Devonian and Carboniferous periods, and they draw considerable interest due to the precious data they can provide regarding the early evolution of tetrapods.

Phylogenetic hypotheses about the stem-group tetrapods abound, and typically change as new taxa are discovered and analyzed [10]. For instance, the order Osteolepiformes is now considered paraphyletic [2], although it has still been used in the past decade (e.g. [7, 10]). In a recent (2012) phylogenetic analysis, Swartz [26] found that the total-group (stem and crown) tetrapods can be classified into five clades: rhizodonts, canowindrids, megalichthyforms, tristichopterids, and elpistostegalians (a clade including all limbed tetrapods) (Fig. 2). In this phylogenetic context, Osteolepiformes corresponds roughly to canowindrids, megalichthyforms, and tristichopterids.

These phylogenetic considerations are of some importance because prelim- inary identification work on the Ivanovka material has revealed the presence of four main groups of fossils: antiarch placoderms (genus Bothriolepis Eich- wald 1840 [11]), lungfish, the osteolepiform Megistolepis Obruchev 1955, and possible limbed tetrapod material. Megistolepis, especially, is abundant in the material, which represents an opportunity to increase knowledge about this poorly studied taxon first found at Tuva, Russia, another Late Devonian Siberian locality. In 1955, Obruchev [24] described the genus as well as the species M. klementzi from scale remains; in 1977, Vorobyeva [28] described some new material and identified a second species, M. doroshkoji. She assigned the genus to the family Osteolepididae. In 1992, Young et al. [31]

assigned it to the family Megalichthyidae, but this has attracted criticism by Fox et al. [12] and recently by Witzmann and Schoch [30], who noted that such an assignment cannot be convincingly made until more material is described. Among the other material, the putative tetrapod remains could be of interest in building a picture of the early evolution of the group.

In addition to the value of the data itself, this study is also novel in terms of technique. The Ivanovka material forms a ‘bone bed’, an informal term for a layer of packed, mingled bones of several animals of different species, who likely died in a mass-death event followed by a period of decay and

(6)

Figure 2: Phylogenetic tree of total-group tetrapods, modified from [26]. ‘Osteolepiformes’

corresponds to the canowindrids (yellow), megalichthyiforms (blue), and tristichopterids (purple). The red arrows indicate the taxa that were used in this study to compare with the Megistolepis material.

(7)

disturbance, after which the bones were buried and fossilized. Traditional preparation, which consists of removing the rock matrix with tools, may be unsatisfactory in the case of bone beds, as the overlap between bones may prevent the exposition of one without the damaging of others. Moreover, the geochemistry of the Ivanovka material makes the use of acetic acid to dissolve the rock matrix impossible. The conventional non-destructive alternative to mechanical or chemical preparation is computed tomography (CT), using medical or laboratory scanners; however, the resolution of such scans tends to be too low for the study of small fossil material, and would result, in the case of the Ivanovka material, in data of poor quality.

Enter x-ray synchrotron microtomography, or SR-µCT. This technology, of relatively recent introduction in the field of paleontology, has been used for investigations of fossil material for little more than a decade [27], and it offers several advantages over mechanical preparation and CT scanning—

including monochromaticity and avoidance of beam hardening, high intensity and resolution, and high coherence allowing absorption constrast. (See for instance [8] for a comparison of CT and SR-µCT scans, in which the latter provides much more adequate resolution.) SR-µCT has never, however, seen use in the study of bone bed material. The Ivanovka fossils therefore represent an opportunity to apply the technique in a new way and to evaluate its advantages and shortcomings in the perspective of future use.

Locality information

The locality is situated on the Uryup River, 1.5 km downstream of the village of Ivanovka in the Kuznetsk Basin of Southern Siberia, Russian Federation (55^◦35’N, 88^◦49’E; Fig. 3). The fossils were collected from three individual localities: localities 1 and 2, which are close to each other, have an elevation of 286 m, and form part of the same fish horizon; and locality 3, approximately 10 m higher. These are part of the Kokhayskaya Formation, which is Upper (Late) Devonian in age and probably Upper Frasnian (circa 375 million years ago). At the time, it consisted of part of the Minusinskaya lake system, a freshwater environment on the Siberian continental margin. It contains plant and fish fossils, and is the source of the Megistolepis doroshkoji material described by Vorobyeva in 1977 [28]. Based on this, we can provisionally attribute the new osteolepiform material to this species.

The animals found in the bone bed itself probably suffered from a mass- death event, as evidenced by the high bone density in a rather thin rock layer.

However, it is almost impossible to tell the cause of death (for instance, the drying of a shallow pond of water), because the bones are not found exactly

(8)

where the animals died: some disturbance must have occurred between the moment of death and the moment of burial.

Figure 3: Map of the fossil locality, near the village of Ivanovka (Ивановка) in Kemerovo Oblast and the Russian Federation. From Jaroslav Gutak, personal communication.

Aims

This investigation is exploratory in nature. As one of the very few studies conducted so far on the Late Devonian vertebrates of Siberia, its primary goal is to provide an account of the Ivanovka fauna in the form of 3D models and morphological descriptions. Any further items are dependent on the outcome of the scanning and modelling process.

Since Megistolepis bones are abundant in the material, this study may allow for a more thorough description of the taxon, and a more precise posi- tioning within the stem-group tetrapod phylogeny. More generally, the information garnered on the Ivanovka fauna will allow a better understanding of Late Devonian biogeography, as well as early tetrapod evolution, especially if tetrapod material is found.

To our knowledge, this is the first study that uses synchrotron microtomography as a tool to study bone bed fossil material. As such, a discussion of the usefulness of the method is also an important output of this work.

(9)

Materials and Methods

In August 2011, Jaroslav Gutak, Per Ahlberg and colleagues collected 164 claystone blocks containing some 400-500 individual bones from the three localities in Ivanovka. The 164 specimens were catalogued and photographed, and their surface bones were identified whenever possible. A subset of 29 of the most promising blocks were then sent to the European Synchrotron Radi- ation Facility (ESRF) in Grenoble, France, to undergo scanning. Specifically, the material underwent propagation phase contrast microtomography on the ESRF ID19 beamline. This 145 m long beamline is devoted to microtomography and is currently being used mostly for paleontological imaging.

The selected samples were put in three 10 cm × 50 cm cylinders and scanned with 16 m of propagation. This allows for an excellent phase contrast effect—thereby increasing brightness contrast between the bone and the matrix—at very high energy (up to 250 keV) for a large field of view (10 cm).

To improve contrast at its maximum level, we used the absorption protocol originally developed for the scanning of hominid skulls [8].

The synchrotron scanning process yields a large number of grayscale image slices that can be concatenated into one image stack for each of the axial, coronal, and sagittal planes. Together, these stacks form a 3D image dataset with a particular voxel (3D pixel) size. In this study, the resolution is a voxel size of 52.15 µm. To obtain 3D models—that is, triangle meshes in the form of STL files—segmentation work (also called ‘modelling’) must be performed.

This consists in applying a mask to every slice in which a specimen is present, effectively reconstructing the three-dimensional structure. The point of phase contrast microtomography is that the bony material should contrast with the surrounding rock matrix in brightness, allowing the mask to be applied in a threshold fashion: only voxels with a grayscale value above (or below) a certain threshold are contained in the final model. In theory, little or no subjective judgement is necessary from the worker, as brightness alone differentiates between bone and stone; in practice, the Ivanovka material posed a special challenge in that the contrast was often very low, despite attempts at maximizing it in the scanning process. When differences in texture, but not in brightness, indicated the presence of a bone, it was necessary to resort to subjective segmentation.

The software used for modelling was Mimics Research 18.0.0.525 (Mate- rialise). I assigned a catalog number to each model as follows: the first digit indicates which dataset the model comes from (1, 2 or 3); the next two digits are arbitrary and identify the model within its dataset; and an optional let- ter differentiates specimens that occur together in the stone blocks and that

(10)

therefore share their catalog number (e.g. 305a and 305b). The rendering and display of the resulting STL files was done with the software Blender v2.76.

The analyses took the form of identification and interpretation of the modelled structures. Concerning the lobe-finned fish material, since very little has been published about Megistolepis, I compared the data with published descriptions of other osteolepiforms, such as the tristochopterid Eu- sthenopteron fordii [13], as well as the seemingly more closely related Gogo- nasus andrewsae [19], Medoevia lata [17], and the Megalichthyid Cladarosym- blema narrianense [12].

Results

A total of 91 specimens were modelled. Because the modelling process is labor-intensive and time was limited, this is by no means the equal to the total number of bones present in the scanned samples, which themselves contain but a fraction of the collected material. It is thus quite possible that valuable fossils have escaped inquiry.

The majority of the 91 specimens have turned out to be too small, incomplete or nondescript to yield any valuable information. Hereafter I describe and interpret the rest.

Probable Megistolepis elements

Ethmosphenoid unit

The dorsal part of the skull of osteolepiform fishes is divided into two units, linked by an intracranial joint: the ethmopshenoid unit, which is anterior to the joint and extends from the parietals to the premaxilla; and the otico- occipital unit, which is posterior to the joint. Both units comprise a skull roof made of dermal bone elements (dermal bone is derived from intramenbranous ossification), and a braincase, or endocranium, made of endoskeletal elements (which form from calcifying cartilage). The skull bones found in the Ivanovka material and described here almost all belong to the ethmosphenoid unit, and most are dermal bones, with the exception of a single endocranium.

Skull roofs

Four specimens contain partial ethmosphenoid skull roofs. The anterior part of the unit is visible in specimens 105 and 104 (Fig. 4), while specimens 202 (Fig. 5) and 305b (Fig. 6) show mostly the posterior part. Specimen 305b is

(11)

of particular interest due to its preserved attachment with the endocranium (see below).

Specimen 105 is the most complete, with perhaps 60% of the skull roof preserved (Fig. 4A-C), but its surface is heavily flattened. Its length is about 70 mm. The left side is mostly complete, while of the right side only the bones near the midline remain. It can be difficult to differentiate between true sutures and cracks in the rock; nevertheless, many individual bones can be distinguished. At the very anterior end, the premaxilla’s structures are not well defined, and the sutures between the bones of the snout may be difficult to see because of cosmine. Cosmine is an association of skeletal and soft tissues characteristic of the lobe-finned fishes, including most Osteolepiforms, on which it forms a layer covering the scales and skull bones [22]. When it conceals the outline of the bones, as it often does in the specimens shown here, it can make identification more difficult.

The labelling of specimen 105 on Fig. 4A is tentative: the numerous small bones found posteriorly to the premaxilla are probably nasals and possibly the rostral and postrostrals. On the lateral edge, the lateral rostral can be seen immediately adjacent to a depression that may correspond to the external naris. The anterior and posterior tectal bones are seen posteriorly to that.

At this point, a possible supraorbital is found, leading the way to the orbit, whose shape is not visible. The large, round gap next to the posterior nasal (easily visible in Fig. 4C) is a preservation artifact, and would have consisted of a part of the left parietal. The parietals are long, slightly less than the total length of the specimen, although it is uncertain whether the posterior end of the parietals is preserved. Between the parietals is an opening that is likely the pineal foramen. Its shape is unclear, but it appears larger than in Gogonasus.

In contrast with the relatively well-preserved exterior surface, the ventral view of specimen 105 suffers from poor definition (Fig. 4B). The only obvious feature is a elongated, curvy process that extends from the roof toward the interior of the skull. It is probably a remnant of the endocranium, and perhaps one of the walls of the braincase. As discussed below, endoskeletal elements in the Ivanovka material pose a greater modelling challenge than dermal bone elements; this may explain why the internal face of the skull roof is so poorly defined. Specimen 105 is further unique in that certain parts of the bone, in the scanned images, are contrasted strongly with the matrix, and some other parts do not. The model includes mostly the high-contrast structures and has missed some of the lower contrasted regions. Yet, as a comparison of Figs. 4A and 4C shows, the resulting model seems to be faithful to the original fossil.

Specimen 104 (Fig. 4D,E) is also flattened, save for the anterior end of

(12)

Figure 4: A, specimen 105 (ethmosphenoid skull roof) in dorsal view. The dashed line indicates the approximate anteroposterior midline. B, same specimen in ventral view. C, optic photograph of specimen 105 embedded in its mineral matrix, for comparison with the 3D model. D-E, dorsal and anterodorsal views of specimen 104, another ethmosphenoid skull roof. Scales = 10 mm, except for C.

(13)

the premaxilla, where a few small teeth can be distinguished. The sutures are less evident than in specimen 105, but the preservation is superior: except for a chipped off section of the premaxilla, the unit is mostly complete anteriorly from the posterior tectals. Little remains from the posterior part of the unit, however, except for most of the left parietal and the median postrostral.

Small depressions (hardly visible on the figure) suggest the presence of the external nares on either side of the unit. The bones occurring posteriorly to the premaxilla are interpreted as the rostral and nasals, but their outlines are difficult to make out. The posterior tectals have well-defined shapes and protrude at the extreme left and right of the specimen, in both case making it evident that the supraorbital bones, and thus the orbital margins, are missing. Specimen 104 is smaller than 105, but when scaled to match its size, its overall shape and the position of its main bones correspond to those of 105. Thus, both skulls probably belong to the same taxon.

Moving to the posterior section of the ethmosphenoid skull roof, specimen 202 (Fig. 5) mostly shows the parietal and intertemporal bones, as well as part of the right orbital margin. Both the dorsal and ventral surfaces are almost featureless, with the sutures between these bones being completely invisible.

However, for this specimen I performed an analysis of the cross-section image slices, where slight changes in bone texture can reveal the presence of such sutures. The dashed lines in Fig. 5A,B indicate the resulting approxi- mations between the parietals and intertemporals. (More sutures must have been present, notably with the supraorbital, but were not observed in the cross-section analysis.) Besides showing the general shape of the bones, these suture lines draw attention to the fact that no pineal foramen is visible in this specimen. This is possibly an important difference with specimen 105.

Posteriorly, the intertemporals terminate with a pointy process that also extends downwardly into what would be the endocranium, most likely serving as attachments for the braincase, of which nothing is preserved. In posterior view (Fig. 5C), the edge of the unit, between the pointy processes, is the intracranial joint linking the ethmosphenoid unit to the otico-occipital unit. On the right end of the joint, a depression can be seen, in which the postparietals and supratemporals of the otico-occipital unit would fit.

The general outline of specimen 305b (Fig. 6) is difficult to establish:

the preserved parts correspond roughly to the middle and posterior part of a skull roof, and are missing the premaxilla, tectals and intertemporals.

Fortunately, the preserved endocranium beneath the roof provides some po- sitioning information. As is the case for the other specimens, skull roof 305b is mostly flattened, except for its left-anterior corner. The posterior margin of the unit (posterior end of the parietals) is preserved, and corresponds to the intracranial joint, but no features make a joint with the otico-occipital unit

(14)

Figure 5: Specimen 202a (posterior part of an ethmosphenoid skull roof) in A, dorsal; B, ventral; and C, posterior views. The dashed lines indicate approximate sutures between the parietals and intertemporals, as determined from the raw image slices. No dashed lines indicate the sutures with other bones, such as supraorbitals, even though such sutures must exist. Scale = 10 mm.

(15)

Figure 6: Skull roof (305a, beige) of specimen 305, also comprising an endocranium (305a, orange; see Fig. 7). A, dorsal view, showing the bones of the skull roof. The endocranium is partially visible behind it. B, ventral view of the skull roof, with the endocranium removed (Fig. 7A shows the same view with the endocranium present). Scale = 10 mm.

obvious. The large gap corresponds to a missing part of the right parietal and possibly the median postrostral. The postnasals and rostral seem to be well preserved, with sutures easily visible. On the internal side, which can be seen by removing the 3D model of the endocranium, rugose surfaces can be seen in areas where the endocranium seems to attach to the skull roof. It must be noted that in published descriptions, the internal face of the skull roof is seldom described, because it is concealed by the endocranium. This is a problem that SR-µCT studies do not encounter, but the resulting surface is rarely very informative, as the specimens described herein (internal surface of specimen 104 not shown) demonstrate.

Endocranium

In addition to the partial skull roof, specimen 305 comprises an endocranial braincase (305a; Fig. 7), with the relative positions of the roof and endocranium being probably preserved at least to some degree. It is immediately obvious that the endocranium displays much poorer definition than the skull roof: this is due to very low contrast in the image dataset, which has made modelling difficult. Moreover, the quality of the preservation is mediocre at best: the endocranium has evidently suffered oblique dorsoventral compres-

(16)

sion. Yet these difficulties in modelling and this poor preservation do not preclude the identification of the main features in the only available endocranium from the Ivanovka material.

Its general shape is reasonably well preserved, as is best seen in Fig. 7A.

Most ventrally, a thin parasphenoid runs along the middle half of the specimen, and rises towards the ethmosphenoid through the crista suspendens, which is, as with most lateral structures, fragile and poorly preserved. At the posterior end of the parasphenoid, thick bone makes the transition with the wide notochordal chamber (or notochordal pit), which, as a slanted oval, shows signs of compression (Fig. 7A,E), but remains the most recognizable feature of the specimen. Long et al. [19] note that a pit is visible on the roof of the notochordal chamber, directly below the opening the brain cavity, in both Gogonasus and Medoevia; in specimen 305, the presence or absence of such a pit is unclear due to the arrangement of the structures above the notochordal chamber (Fig 7E). This is also where the processus connectens is found ; it consists of two bumps that protrude posteriorly, forming the very posterior end of the ethmopsphenoid unit. Above it lies the opening of the brain cavity, also compressed dorsoventrally. Because of the compression, the brain cavity appears narrow, and opens directly dorsally on the skull roof (Fig. 7B), approximately where the suture between the left and right parietals is located. There is no evidence of a pineal foramen in specimen 305b, possibly due to the gap in the skull roof, but the position of the braincase below the interparietal suture is consistent with the existence of a pineal stalk rising from the brain into the roof.

On the right side of the endocranium (Fig. 7D), anteriorly from the notochordal chamber, is another thickening termed a ‘process’ in the literature, the basipterygoid process, which is the attachment point for the palatoquadrate on the side of the braincase. That such a structure be absent from the left side (Fig. 7C) is due to the overall poorer preservation of that side. On the right side, certain slight depressions are tentatively interpreted as attachment scars for the arcus palatini and subcranial muscles. Going anteriorly from the parasphenoid, we reach the palatal lamina—thin near the parasphenoid, and widening anteriorly towards what would be an attachment with the premaxilla if it were not missing from 305b.

Otico-occipital unit

The only specimen that may belong to the otico-occipital unit is the fragmentary skull roof 213. (Fig. 8). The preservation and modelling of this specimen are quite good: on its internal surface, the radial growth patterns of the bones are visible, as are the sutures. A tentative interpretation is that they repre-

(17)

Figure 7: Endocranium (305a, orange) of specimen 305, also comprising a skull roof (305b, beige; see Fig. 6). A, ventral view, with the skull roof behind (Fig. 6B shows the same view without the endocranium). B, dorsal view with the skull roof removed. C-D, left and right lateral views with the skull roof removed. E, posterior view. Contrast the poor definition of the endocranium model with the better quality of the roof. Scale = 10 mm.

(18)

Figure 8: Specimen 213 (posterior otico-occipital skull roof) in A, ventral view, B, dorsal view, C, posterior view. The dashed line represents a possible midline. Note the well- preserved radial growth patterns and sutures on the internal surface in A. Scale = 10 mm.

sent the posterior part of the otico-occipital skull roof. If this is correct, we see the two postparietals separated by a midline suture, part of the tabu- lar (which curves down at the edge of the head), and, anteriorly, a small fragment of the supratemporal. The posterior edge of the specimen carries a somewhat deep groove that would in this case fit with the extrascapulars.

Palate

Two osteolepiform palatoquadrate complexes, comprising various palatal bones including partial dermopalatines and ectopterygoids, are available, as are some isolated palatal fanged bones: a dermopalatine, an ectopterygoid, and a vomer. These bones form the palate and part of the dentition of the upper jaw.

Palatoquadrate complexes

The first specimen that I interpret as a partial palatoquadrate complex is specimen 221 (Fig. 9). It includes two conspicuous fangs (teeth that are oversized compared to regular teeth, also called tusks in the literature), indicative of the dermopalatine and ectopterygoid bones, each of which typically pos- sesses a pair of one fang and one fang replacement pit. In 221, the anterior

(19)

fang belongs to the dermopalatine and is accompanied by its replacement pit, while the posterior fang belongs to the ectopterygoid, whose pit is not visible. A row of teeth runs along a ridge between the two fangs; it would undoubtedly have extended further than the fangs, but that region is not preserved. The external surface of the dermopalatine and ectopterygoid is lost; quite visible, on the other hand, is the suture that delimitates these two bones from the palatoquadrate itself. The general shape of the complex is difficult to determine, as most edges are broken. It can be said that the palatoquadrate, which forms the palatal wall, looms over the fanged bones in an oblique fashion, and terminates anteriorly with a pointy projection that I interpret as part of the processus autopalatinus. In lateral view, a ridge is seen running along the length of the palatoquadrate; more importantly, the dorsal part of the complex includes some sort of long flange that folds over the palatoquadrate. The external surface of that region is rugose, and may have been attached to the endocranium.

The second palatoquadrate complex, specimen 319 (Fig. 10), shows several differences with 221. One large fang is visible, belonging to the dermopalatine. This fang and its replacement pit rest on a ridge carrying a tooth row and forming the outer wall of the bone. Mesially, a thick platform constitutes the inner boundary of the dermopalatine, and the suture with the entopterygoid is clear. The dorsolateral and the anterior regions of the dermopalatine are lost, and whether an anterior process was present or not (as in specimen 101, see below and Fig. 11) is unknown. The tooth ridge, which is high in the dermopalatine region, gradually diminishes as it extends posteriorly towards the ectopterygoid. No suture is can be seen between the dermopalatine and ectopterygoid, and few features of the latter are visible, except for a rather deep fossa interpreted as the location where the coronoid fangs of the mandible would fit when the jaw is closed.

The rest of the complex differs markedly from specimen 221 in that it is not a smooth, clear surface. Immediately dorsally to the suture with the dermopalatine, the entopterygoid rises into a ridge that gets especially pronounced in a lip-like structure around the middle part of the specimen. Just behind this structure, a hole opens in the palatoquadrate, which is puzzling as no foramen is expected in the centre of a palatoquadrate complex. It may be an artifact of the modelling or preservation. The ridge presumably forms the end of the entopterygoid, and is mirrored on the lateral side by the end of the thick (but incomplete) region of the palatoquadrate. Dorsally to this, the thinner region of the palatoquadrate extends up; its dorsal edge is preserved, and the processus paratemporalis is visible.

Because neither 221 nor 319 are complete, and because they share few

(20)

Figure 9: Specimen 221 (sarcopterygian right palatoquadrate complex) in A, mesial; B, lateral; and C, posterior view. Scale = 10 mm.

(21)

Figure 10: Specimen 319 (sarcopterygian left palatoquadrate complex) in A, mesial; B, lateral; and C, anterior view. Scale = 10 mm.

(22)

clear common structures, comparison is difficult. They may or may not belong to the same taxon.

Dermopalatine

Possibly one of the specimens with the highest quality of preservation and modelling, specimen 101 is a dermopalatine missing only its posterior region, towards the suture with the ectopterygoid (Fig. 11). Unlike any of the other fanged bones, two fangs are present. It was clear, during segmentation, that one is attached to the bone, while the other is not. This free fang may have been in the process of shedding, or be new and not yet firmly attached to the bone. It was modelled as an independent 3D mesh, allowing it to be removed to reveal the replacement pit underneath (Fig. 11C). A shed fang that is still in location has taphonomic implications: it suggests that the bone must not have been subjected to heavy disturbance between the death of its owner and the burial and start of the fossilisation process. On the other hand, this dermopalatine is isolated from any other palatal bones, even though such bones are usually sutured strongly to each other, which must be due to disturbance of some sort. Perhaps the fang remained in position due to surrounding tissue as the bone was carried away from its skull.

On the lateral side of the fangs, a high outer wall ends in a ridge, at about half the height of the fangs, along which occurs a tooth row. The ridge is thicker where it is adjacent to the fangs; the teeth are small and do not vary much in size; and the tooth row is single, not multiple as it occurs in Medoevia and Gogonasus. The ridge ends and falls sharply anteriorly, where the specimen terminates with an anterior process similar, if a little shorter, to Medoevia’s, and more definite than Gogonasus’s. In these taxa, the anterior process protrudes into an open space, forming part of the edge of the choana. On the anterior region of the outer wall, a shallow depression occurs like in Gogonasus; Long et al. [19] note that this is where “the flange of the anteromesial face of the maxilla fits, to form the posterior edge of the choana”.

Most of the palatal surface, mesial to the outer wall, is occupied by a horizontal ‘mesial platform’ carrying the fangs and pit. Unlike in Gogonasus, the fangs and pit do not occupy the entire surface of the platform, which has a thin and smooth anterior region forming a flange slightly folding up.

The platform is thickest under the fangs and pit; the posterior part of the platform, however, is not preserved. The mesial platform and anterior process form a notch that houses the fang of the 1st coronoid when the jaw is shut. While this notch and the anterior process are not preserved in the dermopalatine of palatoquadrate complex 319 (Fig. 10), the rest of that bone

(23)

Figure 11: Specimen 101 (sarcopterygian left dermopalatine) in A, mesial; B, lateral; and C, ventral view. In C, the free fang has been removed to show the underlying pit. Scale = 10 mm.

(24)

Figure 12: Specimen 314 (sarcopterygian left ectopterygoid) in A, mesial; B, lateral; and C, ventral view. Scale = 10 mm.

(fang, pit, outer wall and mesial platform) is similar enough to 101 to suggest that they belong to the same taxon.

Ectopterygoid

Specimen 314 (Fig. 12), an isolated ectopterygoid, is a model of much lower quality, but is fortunately complete and does not seem to have suffered compression or breakage. It is less than 20 mm long, and about three times as long as it is wide. Its size is thus much smaller than that of specimen 101, indicating that these two bones did not belong to the same individual; whether they belonged to the same taxon is uncertain. The fang and its replacement pit occur approximately halfway along the length of the mesial platform; on either side, at the anterior and posterior ends, are areas in which the fangs of the 2nd and 3rd coronoids fit. An outer wall, with teeth on its ridge, covers the entire lateral side of the bone. At least one tooth appears larger than the rest, but the tooth row is too poorly defined to say more. The bone terminates posteriorly with a pointy end, like in Medoevia but not Gogonasus, in which it is rounder.

(25)

Vomer

We have a single, complete sarcopterygian right vomer in specimen 312 (Fig.

13). A large, attached fang is clearly visible, as is, mesially, its replacement pit; they are bordered anteriorly by a high transverse ridge, which forms a wall on the anterior face of the bone. Two teeth, much smaller than the fang, are visible along this ridge. They are adjacent and lateral to the fang. More teeth were perhaps present, but were not captured by either the preservation process or the modelling, perhaps for being two small: in the Megalichthyid Cladarosymblema [12], the teeth (which Fox et al. call ‘denticles’) are numerous and vary in size, with the larger ones located at the corresponding position of the two teeth in specimen 312. The same is true of Medoevia.

Note, however, that specimen 101 (Fig. 11)also has fewer teeth on its ridge as compared to these taxa: it is thus possible that a single, simplified tooth row on the palatal fanged bones is a characteristic of the taxon to which both these bones belong.

Long et al. [19] note that Gogonasus, unlike Eusthenopteron and Me- galichthys, lacks pointed extensions on the posterior side of the vomer. Spec- imen 312 differs from Gogonasus in that it does carry a modest posterior extension, forming a small platform that possibly overlaps the dermopalatine;

however, it cannot be described as ‘pointy’, nor does it resemble the long, narrowing posterior end of the Eusthenopteron vomer [13]. Overall, the shape of the bone in ventral or dorsal view is almost star-like, with various flanges and extensions protruding in many directions. This is more pronounced here than in Gogonasus or Cladarosymblema, although the latter’s simple shape may be due to poor preservation [12]; in this respect, specimen 312 resembles the vomer of Medoevia more, except for its lack of a distinctive and pronounced antero-mesial process.

The vomer is one of the few Megistolepis klementzi bones drawn by Vorobyeva [28, Fig. 35Б,В]. Her figures of the ethmosphenoid unit show two vomers, each of which having an anterior tooth row, a pair of pits or fangs, and a moderate posterior extension that may be likened to the one seen in specimen 312. In posterior view, they appear to have a more pronounced pointy dorsal end than specimen 312. No other important differences are visible, and although Vorobyeva’s figures lack a high level of detail, they strengthen the possibility that specimen 312 does indeed belong to Megistolepis.

Mandibles

At least seven partial mandibular rami (halves of lower jaws) were found in the material, with varying degrees of completeness (Figs. 14-18).

(26)

Figure 13: Specimen 312 (right vomer). A, anterior view. B, posterior view. C, ventral view. D, dorsal view. Scale = 10 mm.

(27)

The most complete mandible is specimen 102 (Fig. 14A,B). It is almost complete, except for its posterior part, which is cut-off in the original material. The specimen is 84 mm in length and roughly 15 mm in height. It has three coronoids; on the first, there is a large and conspicuous fang, accompanied by a replacement pit. The fangs occurring on the second and third coronoid are smaller, though still larger than most teeth on the dentary tooth row. The exceptions are the two anterior-most teeth, which are roughly equal in size to the 2nd and 3rd coronoid fangs (similarly to Medoevia [17]). The tooth row extends from these teeth, at the anterior end of the mandible, to the cut-off area, which corresponds to the anterior end of the fossa for the adductor mandibulae muscle. It is not possible to comment on the length of the fossa. In general, the sutures between bones are not visible, and the coronoids are distinguished mainly from their fangs, although there an evident intercoronoid fossa between coronoids 1 and 2, in which we expect the dermopalatine fang of the upper jaw to interlock. Mesially, much of the posterior half of the specimen is occupied by the prearticular, which extends from the edges of the adductor fossa (where its height corresponds to that of the mandible), to the prearticular anterior process, directly mesial to the 1st fang (where it is much reduced in height). The canal extending from the adductor fossa runs between the prearticular and the external dermal bones.

Part of the articular must be visible posteriorly, but the boundary with the prearticular is unclear. Anteriorly, adjacent to the two oversized teeth, lies the mandibular symphysis—the point of contact with the other mandibular ramus. The external aspect of the mandible is covered with cosmine, compli- cating the identification of individual bones. A suture running laterally along the middle third of the specimen may indicate the suture between the dentary and infradentary bones. This right mandible appears overall similar to Medoevia, Gogonasus and Cladarosymblema, with a possibly slightly larger 1st fang than any of them.

Specimen 103, another right mandible, is much less complete and smaller in size, but overall similar to specimen 102 (Fig. 14C,D). Put to scale, the relative positions of the fangs, mandibular symphysis and adductor fossa are indeed almost the same, although fang 2 is somewhat larger, and there are possibly two fangs on coronoid 3. It is therefore likely that the two rami belong to two individuals of the same taxon, one slightly smaller than the other.

Though having evidently suffered less than ideal preservation, specimen 103 has the advantage of comprising the posterior part of the mandible, at least on its mesial side, up to the glenoid fossa. The dentary tooth row is lost, as is the lateral side of the adductor fossa and most of the external surface.

Anteriorly, almost nothing is preserved, except very fragmentary Meckelian bone possibly up to the mandibular symphysis.

(28)

Figure 14: A-B, mesial and lateral views of mandible 102. C-D, mesial and lateral views of mandible 103. Scale = 10 mm.

(29)

Figure 15: Specimen 201a (left mandibular ramus) in A, mesial and B, lateral view. Scale

= 10 mm.

Another fragmentary, but almost full-length (left) mandible is found in specimen 201a (Fig. 15). The external surface is almost all entirely lost, as are all of the coronoids, fangs, and endoskeletal bones. What remains is the prearticular running along the length of the entire specimen, with the exception of its posterodorsal end; a mesial, low part of the edge of the adductor fossa; and some well-preserved anterior structures. These include the parasymphysial plate, which is very well defined here, as well as the anterior end of the dentary tooth row. The general lack of preserved structures makes comparison with other specimens difficult, but the presence of the two oversized anteriormost teeth suggests this belongs to the same taxon as specimens 102 and 103. Specimen 201a appears slightly larger than 102, and its tooth row, after the two large teeth, rises gradually more than that of 102, but there is not sufficient data to detect any significant differences.

Specimens 110 (Fig. 16), 219, and 214 (Fig. 17) are not full-length: the first shows the anterior part of a left mandibular ramus, while the latter two represent the posterior part of other such rami. Specimen 110 has similar preservation to specimen 102, making them readily comparable. In most respects, they appear very much alike, to the point that they could have belonged to the same individual, though there is no way to prove this. The oversized anterior teeth do not show as well, but this is most likely a modelling artifact. The fang of the first coronoid is as large, and its replacement pit, the intercoronoid fossa, and the anterior part of the prearticular are all

(30)

visible. Externally, the cosmine cover is mostly gone, though this does not help identify the external bones since only the (partial) dentary is present.

Figure 16: Specimen 110 (left mandibular ramus, anterior part) in A, mesial, and B, lateral views. Scale = 10 mm.

Specimens 219 and 214, both posterior parts of left mandibular rami, are laterally compressed, as evidenced by their narrow adductor fossae. On specimen 219 (Fig. 17A-C), the fossa, although compressed, is complete, and bounded on its mesial side by the prearticular and articular bones. Posteri- orly, the glenoid fossa is visible, and heavily slanted mesially. The specimen posteriorly terminates in a post glenoid process that is not very pronounced.

On the anterior side of the fossa, the 3rd coronoid is visible and carries a medium-sized fang. Not much can be seen further anteriorly. The external

(31)

Figure 17: A-C, mesial, lateral and dorsal views of mandible 219; D-F, lateral, mesial and dorsal views of mandible 214. In both specimens, note the narrow adductor fossa and canal, indicating lateral compression. Scale = 10 mm.

side of the bone, however, is of some interest, as some sutures between the dermal bones are visible despite the smooth cosmine. The posterior part of the dentary is visible (compare with specimen 203a, below and Fig. 18), as is its suture with indradentaries 3 and 4. A conspicuous groove running almost parallel to this suture is interpreted as breakage. No teeth are visible on the dentary: this is partly due to the fact that the tooth row typically stops around the beginning of the adductor fossa, but the lack of teeth on the anterior-most section of the specimen is more likely due to poor preservation.

Overall, and taking the lateral compression into account, specimen 219 is not unlike specimen 102 and 103, and corresponds rougly in size to the latter.

Because of its more fragmentary nature, it is more difficult to make sense of specimen 214 (Fig. 17D-F). Some bones surrounding the adductor fossa are visible, as is the open canal extending anteriorly from the fossa. The external surface is preserved, but any grooves seen are due to breakage and do not indicate true sutures.

(32)

Figure 18: Specimen 203a (isolated dentary) in A, mesial; B, lateral; and C, anterior view. Scale = 10 mm.

The last mandibular specimen described here is specimen 203a (Fig. 18), an isolated dentary bone lacking its anterior part. Since the exterior of the mandibular rami found in the material is always covered in cosmine and complicates the interpretation of individual bones (except in specimen 219, which contains only the posterior part of the dentary), this specimen is valuable in showing the shape of an Ivanovka sarcopterygian dentary, though its missing anterior part limits this. The bone is less than 15 mm at its highest, anteriorly; this height in constant until the middle of the specimen, where it gradually decreases until it reaches the posterior-most part. There, a thin process extends, probably locking into the surangular. Externally, the smooth cosmine covering reveals little, except for a faint groove running from the broken anterior edge to the middle part of the bone. Small teeth (sometimes missing and revealing pits) are present on most of the length, along a ridge that is folded over the mesial side. This fold is more pronounced anteriorly and gradually recedes in the posterior half, with the posterior process being the continuation and end of the ridge. The posterior end of the bone, with its gradual decrease in height, fits the posterior end of the dentary in specimen 219 (Fig. 17B).

I conclude that, with the possible exception of specimen 214, which is too poorly preserved to allow interpretation, all the mandibular bones decribed in this section probably belong to the same taxon. This taxon is almost certainly Megistolepis. Specimens 102 and especially 110 are readily comparable with another of the rare drawings of M. klementzi provided by Vorobyeva, showing the anterior region of a mandible in mesial view [29, Fig. 5А]. The position

(33)

of the structures (fang and pit, intercoronoid fossa, parasymphysial plate, large anterior tooth) in her figure and specimens 102 and 110 are overall in agreement, supporting my conclusion. The sole obvious difference is that she has included two tooth rows where I see only one. It is unclear whether this additional tooth row is a misinterpretation by Vorobyeva, or an omission in specimen 214 due to a modelling or preservation artifact.

Isolated fangs and teeth

The Ivanovka material has yielded a few isolated fangs and teeth (specimens 207-210, Fig. 19B-E), and one fang that is attached to fragmentary bone (specimen 308, Fig. 19F). The fang of dermopalatine 101 is also shown as Fig. 19A for comparison. Fang 210 (Fig. 19B), which is slightly flattened at its base (see Fig. 20A), is roughly the same size as 101, and could be from a dermopalatine or vomer, or from the 1st coronoid of a mandible. Fangs 207 and 208 (Fig. 19C,D) are smaller and may come from the 2nd or 3rd coronoids; their 3D models show a hollow interior. Specimen 209 (Fig. 19E) is too small to be a fang and is probably a large tooth instead. Specimen 308 (Fig. 19F) is probably attached to either a dermopalatine or an ectopterygoid, due to the presence of a ridge with two teeth directly adjacent to it. Because it is considerably smaller than the dermopalatine 101 fang, I interpret it as an ectopterygoid tooth, but it may simply come from a smaller animal.

In the absence of attached bones, it is difficult to ascertain any information regarding these models, such as their taxonomic affinity. None of them shows folding at their base, though this may be due to poor (or absence) of base preservation—a cross-section of fang 210 shows some evidence of folding near the base (Fig. 20A). In any case, the folding is less pronounced than in Cladarosymblema, Medoevia, and Gogonasus. The cross-sections of fang 210 (Fig. 20) demonstrate that SR-µCT can help study the internal structures of the teeth and fangs, but this was not the focus here.

Scales

Several ostelolepiform scales are present, five of which are shown in Fig. 21.

In general, they are between 10 and 15 mm in length and approximately rhombic in shape, though shape does vary (and is expected to, depending on the part of the body they come from [12]). The scales shown in Fig. 21, with the possible exception of specimen 316 (Fig. 21F) for which it is unclear, show two distinctive regions, separated by a V-shaped gutter. The first region, which is exposed in the living indidivual, is covered with cosmine and raised slightly; the other is not exposed, being instead overlapped by the

(34)

Figure 19: Fangs and teeth. A, free fang of dermopalatine 101 (see Fig. 11). B-D, isolated fangs 210, 207 and 208. E, isolated tooth 209. F, specimen 308, comprising a fang still attached to a fragmentary dermopalatine or ectopterygoid. Scale = 10 mm.

Figure 20: A-B, two cross-sections of fang 210 as seen in the scans. A is more basal than B. C, model of fang 210 showing the approximate location and orientation of the cross-sections. Scale A,B = 1 mm.

(35)

two adjacent scales, and is not cosmine-covered (as seen in cross-section, Fig.

21G). The gutter is well preserved only in specimen 303 (Fig. 21A), taking the form of a slight groove. In specimen 224 (Fig. 21D), it is also depressed, but the scale appears to be flattened, with no difference in height between the cosmine-covered and overlapped areas. Specimen 303 is the model with the highest quality, especially on its inner face (Fig. 21B). Like in Cladarosym- blema and Gogonasus, it shows a ridge of raised bone ranging from the centre of the scale towards the overlapped area. Fox et al. interpret it as the remnant of an original peg-and-socket articulation [12]. The ridge is not seen in other specimens; this may be due to an artifact of modelling and preservation, or to the fact that, as in Cladarosymblema, it is “absent, or only poorly developed, on many scales” [12].

Histologically (Fig. 21G), three layers are visible, in agreement with Gog- onasus. The thin cosmine layer is noticeably absent from the overlapped area, where the vesicular bone layer is exposed. A few vesicles are seen, though the resolution is not high enough to distinguish microscopic structures with an excellent level of detail. Underneath, the basal lamellar layer appears thicker than that of Gogonasus scales. As for the fangs, the brightness contrast between the different structures of the scales is good, suggesting that more detailed study of these scans could be possible.

Lungfish material

Shoulder girdle

While no undoubtedly osteolepiform shoulder girdle elements have been found, there are several that belong to lungfish (dipnoans). Specimens 204 and 315a are lungfish cleithra and are similar in morphology to the cleithrum of the Late Devonian lungfish Scaumenacia curta [13, Fig. 335]; specimen 315a is also associated with clavicle 315b.

Specimen 204 (Fig. 22), a right side cleithrum, shows in lateral aspect two regions separated by a prominent ridge: the external portion and the postbranchial lamina. In Scaumenacia, the opercular overlaps the postbranchial lamina, which forms the rear wall of the gill chamber, while the external portion is exposed and is covered with grooves and foramina. Some foramina are perhaps visible in specimen 204, but they could also be the result of poor preservation or modelling (and indeed, the postbranchial lamina appears to bear many more holes, proably as an artifact of modelling). At the ventral end, the postbranchial lamina folds gently upward to form an arched crest that would contact with a similar structure in the clavicle.

(36)

Figure 21: Osteolepiform scales. A-B, external and internal views of specimen 303. C-F, external view of specimens 320, 224, 223, and 316. G, cross-section of specimen 223 as seen in the scans, corresponding to the dashed line in E, and showing basic histology. Scales A-F = 10 mm, G = 1 mm.

(37)

Figure 22: Specimen 204 (lungfish right cleithrum) in A, lateral, and B, mesial views.

Scale = 10 mm.

(38)

Figure 23: A-B, lateral and mesial views of specimen 315a (lungfish left cleithrum). C, mesial view of specimen 315b (lungfish right clavicle). D, crude reconstruction of the right shoulder girdle of a lungfish, showing clavicle 315b in lateral view and a mirror-image of cleihtrum 315a. Scale = 10 mm.

Specimen 315a is a left side cleithrum and is overall very similar, showing a distinct ridge, the postbranchial lamina and the external region (Fig.

23A,B). In the material, it is found very close to a right clavicle (315b, Fig.

23C,D) that may well have belonged to the same individual: when one of the bones is mirror-imaged, the two approximately fit together, with the slight folds of their arched crests touching (Fig. 23D). The postbranchial lamina of the clavicle is perpendicular to the external portion, which extends further ventrally than the lamina and terminates in a rectangular fashion, seemingly unlike the clavicle of Scaumenacia, which is rather pointy. As in Scaumenacia, the clavicle is only a little shorter than the cleithrum.

Opercular

Specimen 220 (Fig. 24) is a lungfish opercular bone. It is approximately circular, and has an short process on its anterodorsal corner, similarly to Soederberghia simpsoni [3]. It is approximately flat, but does display a bump on the anterior region of its external face, corresponding to a depression on the internal surface. This depression is most likely where the hyomandibular bone was attached. The concentric rings seen in the figure are an artifact of the imaging process: each ring represents an image slice, and their thick-

(39)

Figure 24: Specimen 220 (lungfish opercular) in A, external, B, internal, and C, ventral view. Scale = 10 mm.

ness corresponds to the resolution of the scans. There is no cosmine cover.

Lungfish typically possess large operculars [15] , which are often almost circular, as is the case, for instance, in the Devonian taxa illustrated by Jarvik [13, Figs. 337-339], including Scaumenacia, Dipterus, Griphognathus, Rhyn- chodipterus, and Fleurantia, or the Australian dipnoans described by Long [18] Howidipterus and Barwikcia. Thus, this specimen is not very diagnostic of any particular Devonian lungfish taxon.

Toothplate

Specimen 218 (Fig. 25) is the only lungfish toothplate found in the material. It is made of seven blade-like ridges gradually decreasing in size from the anteriormost, which is much larger than the rest, to the posteriormost.

Lungfish have very diverse and unique dentitions within Osteichthyes [4], but a very large number of lungfish taxa share the morphology seen in specimen 218 (e.g. [16]). Similarly to the opercular bone above, taxonomic inferences are almost impossible from this toothplate alone.

(40)

Figure 25: Specimen 218 (lungfish toothplate) in A, occlusal, and B, anti-occlusal view.

Scale = 10 mm.

Ribs

Four fragmentary ribs were found (Fig. 26). Three of them (304a,b,c) occur adjacent to each other and may have belonged to the same animal. The two longest show some degree of curvature at what it probably their basal end.

It is suspected that they are lungfish ribs, as the ribs of other Devonian vertebrates tend to be different: ribs of osteolepiforms, for instance, are very short [5, 9], whereas those of tetrapods like Acanthostega [9] and Ichthyostega [14] have a broad, spatulate proximal end. Long, curvy ribs like specimens 304a,b,c are thus probably lungfish. More ribs were noticed in the datasets, but were not modelled as it was thought that they would provide little new information.

Other bones

Other than the Megistolepis and lungfish material, a few specimens merit mention. Their interpretation is uncertain.

Two other bones are interesting because of their curvy shape. Specimen 205 (Fig. 27A,B) is a thin bone with a relatively strong curvature forming an almost perfect circular arc. It may be part of the ring of a vertebra. Specimen 231 (Fig. 27C,D) is thick, less than 10 mm in length, and its cross-section

(41)

Figure 26: Lungfish ribs. A-C, specimens 304a, b and c. D, specimen 317. Scale = 10 mm.

shape is roughly triangular. Its curvature is less pronounced than in 205.

Tentatively, it may be interpreted as part of the orbital arch of a tetrapod, the postfrontal bone. If this is correct, it would be unique evidence of the presence of Devonian tetrapods in Siberia. However, with so little material, it is not possible to make any definite conclusions.

Specimen 225 (Fig. 28) is certainly related to the shoulder girdle, and is probably a cleithrum, with part of the scapulocoracoid attached and forming an ‘arch’. Unlike the vast majority of the Ivanovka specimens, its external face is ornamented. It is unclear to which group this bone belongs.

Discussion

The Ivanovka fauna

The bone bed nature of the material means that the specimens shown and described here are but a sample of the total wealth of the Ivanovka locality.

Only a subset of the modelled bones were informative enough to be included here; only a fraction of the bones present in the scanned samples were modelled; only a few of the collected blocks were scanned; and the collected blocks by no means represent the total material present at the locality. With this in

(42)

Figure 27: Two curvy bones with uncertain identification. A-B, specimen 205. C-D, specimen 231. Scale = 10 mm.

mind, the present study at least gives a first picture of the faunal composition of this Siberian locality.

Most of the modelled material is lobe-finned fish, probably Megistolepis, as discussed in more detail in the next section. Whether other lobe-finned fish taxa are represented is a difficult question to answer. The interpretation for some specimens remain uncertain, such as the possible palatoquadrate complex 221 (Fig. 9), some mandibles, and the shoulder girdle 225 (Fig. 28).

The absence of a pineal foramen in some specimens and its presence in others (also discussed in the next section), though unconclusive, could perhaps indicate different taxa, but generally speaking, there are no specimens that are clearly different from what is expected of Megistolepis while at the same time showing decidedly ‘osteolepiform’ or stem-group tetrapod morphology.

The lungfish material does not give any indication of belonging to more than one taxon (although such a thing would be possible); and it does not provide sufficient information to assign it to any specific taxon. For instance, the toothplate 218 (Fig. 25) resembles that of a large number of extant and fossil lungfishes. The two cleithra (204 and 315a, Figs. 22-23), at least, are very similar and most likely belong to the same taxon. Hence, little more can be said than this: there are lungfish in the Ivanovka fauna. This is not surprising considering that this group was very diverse and widespread in the Devonian.

(43)

Figure 28: Specimen 225 (shoulder girdle element?) in A, internal; B, external; and C, anterior or posterior view. Scale = 10 mm.

(44)

Upon undertaking this study, there was hope that some limbed tetrapod material would be found, casting some new light on the early evolution of this group and the fin-to-limb transition. Unfortunately, no bones convincingly belonging to limbed tetrapods were found, with the possible exception of the putative postfrontal 231 (Fig. 27C,D). Some tetrapod material may have been collected, but is not present in the scans analyzed here.

I mentioned in the introduction the presence of the placoderm Bothri- olepis in the material, yet none was described in the results. This is partly due to the fact that none of the Bothriolepis material was chosen to be scanned, since the anatomy of this genus is well known and was not prior- itized. However, it is also true that the blocks of vertebrate bone bed that were selected for scanning do not seem to contain any placoderm material: at the location, Bothriolepis occurs mostly in another layer, about half a metre below the main bone bed.

These findings do not reveal an extremely clear view of the fauna of the Ivanovka locality, but they may contribute to understanding the general biogeographical patterns of Late Devonian Siberia. Based on the material presented here, these patterns appear consistent with a general homogeniza- tion of the worldwide fauna. Further work on the material may be helpful in confirming what is at the moment but a hypothesis.

Megistolepis

It is difficult to positively assign bone fragments to a taxon of which only a small number of (mostly different) fragments have been found and described.

Since so little data exists on Megistolepis, the only direct morphological com- parisons that are possible are with the anterior region of the mandible [29, Fig. 5А] and the ethmosphenoid unit [28, Fig. 35] drawn by Vorobyeva. These, especially the mandible, do appear to support the proposition that at least some of the material presented here belongs to Megistolepis.

The case of the pineal foramen is an interesting matter. The skull roof specimen 202 (Fig. 5) clearly does not show any pineal opening between the two parietals, although it is unclear whether the specimen includes the region where such an opening should be found if it were there. Specimen 105, on the other hand, seemingly possessed one (Fig. 4). The two specimens may therefore belong to different taxa, but this conclusion cannot be safely made, for the opening in specimen 105 is not clearly defined. The presence or absence of a pineal foramen could be useful is deciding whether Megistolepis should be assigned to the family Megalichthyidae, as it is one of the characteristics of that group [12]. The megalichthyid Cladarosymblema does not have one

(45)

(see [12, Fig. 7,8]), whereas the non-megalichthyid osteolepiforms Gogonasus and Medoevia do.

Other characters of Megalichthyids [12] include: a fang or tusk on the premaxilla (not observed in Megistolepis, neither here nor by Vorobyeva [28]); a vomer that is “transverse in outline, lacking posterior process, but with mesial process meeting, or almost meeting, its fellow in midline” (specimen 312 does have both a posterior and a short mesial process; Fig. 13); a “well-developped, deep anterior mandibular fossa” (such fossae do not appear particularly deep in Megistolepis); and rhombic scales lacking a peg-and-socket articulation (as is the case here). The Ivanovka material does not allow a conclusive state- ment on the potential assignment of Megistolepis to Megalichthyidae, but overall does not seem particularly supportive of it. Even if the genus is not part of the family, its position within the broader taxa Megalichthyiformes, to which Gogonasus and Medoevia belong, remains undisputed—the Ivanovka specimens have more in common with these taxa than, for instance, the tris- tichopterid Eusthenopteron.

Synchrotron microtomography as a tool for the study of bone bed material

While several paleontological studies have made use of synchrotron microtomography in the past two decades, we believe the present work to be the first in which the technique was applied to bone bed material. The rationale for this was that the high density of bones, including many small specimens, makes mechanical preparation difficult and destructive. What follows is a comment on the usefulness and shortcomings of the method in this particular context.

The models obtained and described in the results are overall of good quality, and allow good visualization of the features of the bones. They are not, however, perfect. Modelling artifacts are commonplace, and the quality in some regions is poor. Much of the challenge came from the low brightness contrast, in the scans, between the matrix and the bones. The use of propagation phase contrast microtomography, which not only detects differences in transparency to x-rays, but also refraction effects within the sample, was meant to maximize contrast; yet, even with the technology that probably allows the highest possible contrast at this time (while providing adequate resolution), the contrast was still far from ideal.

Due to this, certain modelled objects suffered from a tradeoff between high precision and short working time: obtaining a precise shape can be very time-consuming and allow less time to be devoted to modelling more

(46)

of the numerous specimens. For an exploratory investigation, this is a limitation. Conversely, modelling a large number of specimens in finite time results in specimens that, while readily identifiable and describable, may not be suited for high-precision morphological analysis. Some features may have been missed in this way, such as small foramina. Foramina can carry important physiological information (for example, as openings for nerves), but were rarely seen in the models here, possibly because of the difficulties in outlining the bones with maximal precision. In addition, dermal bones and endoskeletal bones appear to react differently to synchrotron scanning: compare, for instance, in Figs. 6 and 7, the quality of the dermal skull roof and the endocranium. In the scans, the brightness contrast was arduously low for much of the endocranial elements; moreover, most of the models shown here are dermal bones. Studying endoskeletal elements in detail may require other methods than SR-µCT, for which low contrast may be a limitation in paleontological studies, depending on the particular geology of the samples.

Less of an issue is the resolution. In the image stacks used here, the resolution may be slightly too low for detailed histological analysis (see for instance Figs. 20 and 21, in which the smallest details are not perfectly visualized; compare with the histological optical photographs in [19]), but this was not a goal of this study. Furthermore, higher resolution can be achieved with SR-µCT if the histology of a particular specimen is deemed worthy of investigation.

On a related note, SR-µCT has the advantage of giving access to the internal structure of the specimens without damaging them. It was noticed in many instances that the contrast between different regions of a bone in cross-section was high enough to allow the creation of several models, such as one for the external outline and one for the internal vascularization or porosity. I did not explore this possibility here, but certain bones—the vomer 312 (Fig. 13) comes to mind) could lend themselves to this type of analysis.

Another positive point is that 3D models are easier to study separately. In prepared material, it may not be possible to observe the internal face of a skull roof if the endocranium was not removed; this is easily done with the models.

It is difficult, in the study of bone bed material, to prepare out small (5-15 mm) elements; it is trivial to model them. The focus of this investigation was often on large, complex and informative bones such as mandibles and skull roofs, but small elements could have been studied in more detail if desired.

One drawback of an exploratory study is that the material is spread over a large number of stone blocks: many of these have to be scanned and pro- cessed in order to get a satisfying number of models. This is expensive in both time and money, and as such, much of the Ivanovka material is still

Late Devonian vertebrates from Siberia: a synchrotron microtomography study of bone bed material