CHAPTER XX

In this chapter it is proposed to treat of certain fossil "Fishes" which, from our ignorance of much that is essential to a proper estimate of their true relationships, cannot at present be referred to any of the recognised primary groups of Fishes.

I. Palaeospondylidae.

The interesting little fossil, Palaeospondylus gunni,^[615] discovered in the Lower Old Red Sandstone of Caithness, and first described by Traquair, represents the calcified endoskeleton of an elongated fish-like organism about an inch, or not exceeding two inches, in length. The vertebral column consists of a series of broad, calcified ring-like centra, destitute of ribs, but possessing neural arches and spines, and in the caudal region haemal arches and spines in addition. The skull, of which only the ventral surface is known, has a complete basis cranii, laterally expanded behind by periotic capsules, and in front by what seem to be bulging olfactory capsules. Anteriorly, the skull terminates in a ring of calcified cirri. Behind the skull there are two singular post-occipital plates, one on each side of the anterior section of the vertebral column. The tail was apparently furnished with a fringing caudal fin, supported dorsally by the long forked, neural spines, and below by the much shorter haemal spines. There is no trace of limbs, limb-girdles, jaws, or branchial arches, nor any evidence of the existence of scales, denticles, or other exoskeletal structures.

Palaeospondylus has been regarded as a Cyclostome, a view which derives its principal support from the resemblance of the cirri-encircled orifice at the anterior end of the skull to an unpaired nasal or naso-pituitary aperture, and perhaps some measure of credence from such purely negative evidence as the apparent want of limbs, and of any structures comparable to jaws. But even if it be admitted that there is some reason for this view, it is certain that Palaeospondylus obtained a far higher grade of specialisation in certain respects than any of the existing Cyclostomata; the presence of calcified vertebral centra and neural arches is conclusive on this point.^[616] Palaeospondylus has also been compared with a larval Arthrodiran and with a larval Dipnoid.^[617]

II. Ostracodermi.

The Palaeozoic fish-like forms, which, more as a matter of convenience than as an expression of real kinship, it has been customary to include in this group, are amongst the earliest Craniates of which we have any precise knowledge. Of the three subordinate groups or "Orders" into which they have usually been divided hitherto, two, the Heterostraci and the Osteostraci, may, with some show of reason, be considered as related forms, and although they are characterised by much specialisation on independent lines, there is yet some evidence of connecting links between the two. The organisms comprising the third group, the Antiarchi, stand upon a very different footing, and at present it cannot be said that they are in any way related to either the Heterostraci or the Osteostraci, or indeed to any other Craniates whatsoever. The association of the Ostracodermi with the Cyclostomata, a view which has received more influential support than it deserves, is based on the presumed absence of jaws and paired fins. The absence of jaws, which, if present, were almost certainly cartilaginous, has yet to be proved, and even in the latter group it is by no means certain that they do not possess structures which, morphologically if not functionally, are veritable jaws. Nor is it quite certain that the lateral lobes of some Ostracodermi are neither pectoral flaps nor lateral fin-folds, to say nothing of the lateral appendages of the Antiarchi. And to these objections there is the further difficulty that there is absolutely no evidence that the Ostracodermi are monorhinal in the sense in which this term is applied to the Cyclostomata.^[618] On these grounds it would seem more in accordance with our present knowledge to regard the Ostracodermi as an independent group whose exact position in the system has yet to be determined, including, however, besides the generally accepted orders Heterostraci and Osteostraci, the recently founded provisional order Anaspida, but excluding the Antiarchi as a separate and distinct section; rather than to crystallise in a definite system of classification views which are either purely conjectural or wholly unjustifiable. Even with this limitation the Ostracodermi are by no means easy to define, especially if we include those remarkable shark-like forms from the Upper Silurian rocks of the south of Scotland which have been so admirably described in the recent classical memoirs of Dr. Traquair. As a rule, the head and the anterior part of the body are laterally expanded, and more or less sharply defined from the rest of the body by prominent postero-lateral angles. The exoskeleton, which exhibits an extraordinary variety of structure in the different families, ranges from a uniform covering of dermal denticles to a condition in which the denticles fuse to form anteriorly a highly characteristic tessellated or continuous dorsal shield, while posteriorly they become replaced by a nearly typical rhombic squamation. The tail is heterocercal. Paired fins of the ordinary piscine type are absent. In some Ostracodermi it seems probable that the gill-clefts opened into a common branchial chamber on each side, with a single external aperture, but in others they may have been ventral. The endoskeleton, jaws, dentition, and the nostrils are unknown.

Order I. Heterostraci.

The exoskeletal structures consist of dentine, or of a tissue resembling it, never of true bone. The orbits are marginal or lateral in position. With the exception of a caudal fin there are no median fins.

Fam. 1. Coelolepidae.^[619]—Head and anterior portion of the body flattened and expanded, with prominent lappet-like postero-lateral lobes, which may represent continuous lateral fin-folds or a very primitive type of pectoral fin. Nothing is known of the mouth, but it must have been ventral, nor of the position of the orbits. Branchial apertures unknown, but transverse markings on each side of the anterior part of Thelodus pagei may be indications of a branchial apparatus. The exoskeleton consists of a uniform covering of hollow pointed spines, devoid of a basal plate and open below (Lanarkia); or of minute shagreen-like tubercles (Thelodus). The tubercles or spines consist of dentine coated by ganoin. Of the only two known genera, Thelodus is a characteristic Upper Silurian genus from the Ludlow and Downtonian Beds of Lanarkshire. Detached scales are also known in the Upper Silurian of England. One species (Th. pagei) occurs in the Lower Old Red Sandstone of Forfarshire, and another (Th. tulensis) in the Upper Devonian of Russia. Lanarkia has only been found in the Downtonian Beds. None of the Coelolepidae exceed fourteen to fifteen inches in length.

Fam. 2. Drepanaspidae.—This family^[620] affords an interesting transition to the more highly specialised and carapaced Pteraspidae. The head and anterior part of the trunk now form a broad oblong shield, rounded in front and abruptly marked off from the tail by conspicuous rounded angles. The exoskeleton is no longer uniform. In the caudal region the scattered spines or shagreen tubercles of the Coelolepids have become transformed into tuberculated quadrangular scales, which are further differentiated along the dorsal and ventral margins into ridge scales or fulcra; and from a similar source by a process of basal fusion a series of larger or smaller dermal plates are formed as components of large dorsal or ventral shields. The dorsal shield (Fig. 314) is formed by a large central plate; the postero-lateral portions by two narrow falciform plates; and the anterior margin by a series of smaller rostral plates. Between the larger plates the shield is completed by numerous small polygonal plates. All the plates are superficially ornamented by small stellate tubercles. The ventral armature (Fig. 315) is similar to the dorsal. A large mental plate forms the hinder margin of the transverse slit-like mouth, the anterior limit of which is defined by the rostral plates already mentioned. Laterally may be seen a pair of small plates (x), each perforated by a small aperture, and probably indicating the position of some kind of sense-organ. Posteriorly there is a large median ventral plate, in relation with a pair of anterior and a pair of posterior ventral plates. The areas between the larger plates are filled in by numerous small polygonal plates. It is possible that there is a single external branchial aperture on each side, near the postero-lateral angle of the shield and behind the posterior ventro-lateral plate. The sole representative of the family is Drepanaspis gemündenensis, from the Lower Devonian of Gemünden in Rhenish Prussia. Large examples of this fossil must have exceeded two feet in length.

Fam. 3. Psammosteidae.—To this family are referred certain dermal plates occurring, in a more or less fragmentary condition, in the Old Red Sandstone and Devonian formations of Great Britain and Russia. In their size and shape, and in their stellate tubercles, these have been compared to the dorsal, postero-lateral, and ventral plates of Drepanaspis. That Psammosteus is closely allied to Drepanaspis seems certain, but for the present the two genera may be retained in separate families.

Fam. 4. Pteraspidae.^[621]—Until the recent inclusion of the three preceding families, the Pteraspidae were the only representatives of the Heterostraci. In the best known genus, Pteraspis, there is a marked reduction in the number of the component plates of the carapace, and only seven can now be distinguished (Fig. 316): (a) a large posterior dorsal plate, supporting behind a stout spine; (b) a conical rostral plate, covering the preorbital part of the head; (c) a pair of small marginal orbital plates, each with a small aperture, probably for the eye; (d) a pair of posterior lateral or cornual plates, each of which is perforated by a large oblique foramen, conjecturally an external branchial aperture; and (e) a large ventral plate. There is probably, also, a small median "parietal," or "pineal," plate, with a pit on its inner surface, situated between the rostral and posterior dorsal plates. Externally the plates are sculptured into fine ridges, which in their minute structure and their crenated free margins are suggestive of linear series of fused denticles. The tail appears to have been invested by imbricated rhombic scales. Pteraspis (Lower Old Red Sandstone of Scotland and England, and the Lower Devonian of Galicia); Cyathaspis (Upper Silurian and Lower Old Red Sandstone), known only by its dorsal and ventral shields; and Holaspis (Lower Old Red Sandstone of Monmouthshire, and the Upper Silurian of Pennsylvania), are the only genera.

Order II. Osteostraci.

While agreeing with the more specialised Heterostraci in the division of the body into an anterior carapaced portion and a free hinder part invested by a rhombic squamation, the Osteostraci are distinguished by the presence of bone as a histological component of the dermal hard parts; by the position of the orbits, which, instead of being marginal in position, are close together on the dorsal aspect of the carapace; and by the possession of a median dorsal fin.

Fam. 1. Ateleaspidae.^[622]—The general shape of the body is much the same as in the Coelolepidae, but the exoskeleton consists of numerous polygonal tuberculated plates in front of the postero-lateral lobes, and of sculptured rhombic scales behind. A pair of crescentic markings, placed close together about the middle of the dorsal surface of the head, probably indicate the outer margins of orbital recesses (Fig. 317). The only species at present known (Ateleaspis tessellata) occurs in the Downtonian beds.

Fam. 2. Cephalaspidae.^[623]—In this family the dorsal shield is rounded in front, strongly arched above, with its postero-lateral angles produced into highly characteristic cornua (Fig. 318). The shield consists of a single piece, but as the outer surface is ornamented by small tubercles arranged in polygonal areas, it is probable that it has been formed by the basal fusion of numerous primitively distinct polygonal plates (Fig. 319, A). Between the orbits there is a separately calcified but fixed plate, which bears a hollow prominence, probably for the reception of a parietal organ. In some genera certain of the anterior dorsal and ventral scales of the trunk fuse into a continuous plate. Internally to the postero-lateral cornua the middle layer of the shield is prolonged backwards into a pair of singular flap-like lobes, which have been variously interpreted as corresponding to the lateral lobes of the Coelolepidae, to pectoral fins, or to opercula. The scales of the trunk and tail are rhombic and imbricated; on the sides of the body they are remarkably high and narrow.

The best known genus is Cephalaspis. The earliest remains are found in the Ludlow Tilestones. The genus is also represented in the Ledbury Passage Beds, the Lower Old Red Sandstone of Scotland, and the Upper and Lower Devonian of Canada. Most of the species are of small size, but C. magnifica,^[624] from the Caithness Flagstones, the largest of all the Cephalaspids, has a shield 8½ inches long, and 12 inches across the widest part. Auchenaspis occurs in the Ludlow Tilestones and the Ledbury Passage Beds, and also in the Upper Silurian of the Isle of Oesel in the Baltic. Another genus, Didymaspis, has been found in the Lower Old Red Sandstone of Ledbury.

Fam. 3. Tremataspidae.—The interorbital plate is free, and hence it is often lost in the fossils. Several species of Tremataspis occur in the Upper Silurian of the Isle of Oesel.

As regards the origin and mutual relationships of the different families comprising the Heterostraci, it has been urged with great force by Dr. Traquair^[625] that they constitute a natural sequence of forms, beginning with organisms whose Elasmobranch ancestry is extremely probable, and leading to highly-specialised types, which, considered by themselves, possess little to justify any conclusions whatever as to their origin or kinship. The Coelolepidae form the starting-point, and in the light of their exoskeleton of dermal denticles, their derivation from some primitive Elasmobranch prototype seems a reasonable inference.^[626] From the Coelolepids the path of specialisation through the Drepanaspidae and Psammosteidae to the Pteraspidae is marked (i.) by the basal concrescence of isolated denticles to form, first, numerous small polygonal plates, and then larger and less numerous plates, as the constituent elements of a characteristic dorsal shield, leaving, however, the denticles of the rest of the body to become converted into a rhombic squamation; (ii.) by modifications in the "lateral fin-lobes," which may become enclosed in the developing dermal armour (e.g. Drepanaspis), or cease to be recognisable (e.g. Pteraspis). The affinities of the Osteostraci are very obscure, and their inclusion with the Heterostraci in the same group (Ostracodermi) has hitherto rested mainly on such negative evidence as the supposed absence of paired limbs, jaws, and teeth; in fact, it has been affirmed that "there is absolutely no reason for regarding Cephalaspis as allied to Pteraspis beyond that the two genera occur in the same rocks."^[627] It is possible, however, that in Ateleaspis we have an annectent form, which in some measure combines the structural peculiarities of the two groups. That this singular genus belongs to the Osteostraci is proved by the presence of bone lacunae in its dermal hard parts, a conclusion which is strengthened by the apparently dorsal position of the orbits and the presence of a dorsal fin. On the other hand, its close resemblance to the Coelolepids in the general contour of its laterally-lobed body, and the probability that its mosaic and tuberculated head-shield has been formed by the concrescence of Coelolepid denticles, is at least significant of a relationship to the more primitive Heterostraci. Little can be conjectured as to the habits of these ancient "Fishes." The form and regional proportions of the body, which in some respects often remind one of organisms so diverse as a King Crab, or a Loricaroid Teleost (such as Liposarcus), are strongly suggestive of a grovelling, bottom-feeding, sluggish habit of life, in sharp contrast to the more active and predaceous Fishes whose appearance is coincident with the extinction of the Ostracodermi at the close of the Devonian period. Habits such as these may well be associated with much structural degeneration, even, it may be, with the loss of paired fins, and hence it is not altogether improbable that the Ostracodermi are outcasts from the Elasmobranchs, a degenerate race which has sought safety in a sequestered life and a coat of mail.

Order III. Anaspida.

This group has been instituted by Traquair^[628] for the provisional reception of two remarkable genera, which, owing to the absence of precise knowledge of the histology of their exoskeletal structures, cannot at present be referred either to the Heterostraci or the Osteostraci, and for which, as their discoverer remarks, no place can be found in the system unless they are admitted to the Ostracodermi.

Fam. 1. Birkeniidae.—Body fusiform and fish-like. Head bluntly rounded, without a cranial shield. Caudal fin bilobate and heterocercal A median row of scales with recurved spines arranged along the ventral surface. Orbits, jaws, teeth, paired fins, and endoskeleton unknown.

In Birkenia (Fig. 320) the body is invested by longitudinal rows of narrow scales arranged in oblique transverse rows, which are replaced on the head by much smaller, peculiarly disposed, spindle-shaped scutes. On the side of the hinder part of the head there is an oblique row of small apertures, possibly branchial. A small remote dorsal fin, invested by the trunk scales, is present. Birkenia elegans, the only species known, does not exceed 3½ inches in length. Less is known about the second genus, Lasanius, of which there are two species. Except for the mid-ventral series of spiny scutes, and a row of slender, parallel, rod-like structures, the body appears to have been naked (Fig. 321). The two genera belong to the remarkable series of fossil Fishes from the Silurian rocks of Lanarkshire. Rare in the Ludlow series, Birkenia is by far the most common of the Fishes of the over-lying Downtonian Beds. Lasanius is confined to the latter horizon. Euphanerops, from the Upper Devonian of Canada, is probably related to this family, but lateral branchial apertures are not known.^[629]

III. Antiarchi.

The organisms comprising this group^[630] resemble the Ostracodermi in possessing a well-developed carapace of bony plates and a heterocercal tail, as well as in many of the purely negative features which are characteristic of the latter group.

The remarkable dorsal shield is divided into a small cephalic portion and a much larger hinder part investing the greater part of the trunk, both of which are strongly arched above and flattened ventrally, with a movable articulation between the two. The cephalic shield is formed by numerous symmetrically-disposed tuberculated plates, suturally connected with one another, and, like the other exoskeletal structures, containing bone lacunae (Fig. 322).^[631] The orbits are close together, near the middle of the dorsal surface, and between them there is a small median interorbital plate, with a deep pit on its inner surface, possibly for a parietal organ. A small lateral plate (e.l.), evidently free behind, suggests the presence of an operculum. Nothing is certainly known about the jaws or the nostrils. The mouth is situated just behind the anterior margin of the cephalic shield on the ventral surface, and in front of it there are two plates, which in Bothriolepis canadensis have their oral margins fringed by small "denticles"; it is possible that these plates represent the components of a secondary upper jaw. The dorsal armature of the trunk is shown in Fig. 322. Ventrally it is completed by a pair of anterior ventro-lateral plates and a pair of posterior ventro-lateral plates with a small median plate between the two pairs. Articulating with the anterior ventro-lateral plates by means of a complex hinge joint there is a pair of pectoral appendages of a kind entirely without parallel in any other vertebrated animals. Each appendage is completely encased by numerous suturally connected plates, and about the middle of its length there is a second movable joint. The appendages are hollow, and their cavities probably contained the muscles by which the limbs were moved, and the blood-vessels and nerves for their nutrition and innervation. A lateral line system of the normal type is present in Pterichthys, consisting of a lateral groove along the side of the trunk, and of supra-orbital and infra-orbital grooves, and post-temporal and infra-orbital commissures, on the head. The free portion of the body and the tail are invested by imbricated and finely tuberculated scales, which form fulcra in front of and behind the small dorsal fin. There are no pelvic fins. The caudal fin is heterocercal.

Fam. 1. Asterolepidae.—The best known genera are Pterichthys from the Lower Old Red Sandstone of Scotland and the Devonian of Eifel, and Bothriolepis, a more widely distributed genus which occurs in the Upper Old Red of Scotland and Shropshire, and in the Upper Devonian of Russia and Canada. Two other genera, Asterolepis and Microbrachius, are also found in the Old Red Sandstone of Scotland.^[632]

Beyond an uncertain and shadowy relationship to the Ostracodermi, and perhaps some points of resemblance to the Arthrodira, the Antiarchi stand alone among Craniates. Nothing is known of their origin; no intermediate forms link them to any other groups, and the high specialisation they have attained is sufficient to negative any idea that they can "be credited with any share in the evolution of the Fishes of more recent periods."

IV. Arthrodira

This group has been instituted for the reception of a number of remarkable armoured Fishes of uncertain relationships which flourished in Europe during the Devonian and Old Red Sandstone periods, and in North America from the Devonian to the Lower Carboniferous. The head (e.g. in Coccosteus)^[633] is invested dorsally by a series of median and lateral symmetrically-disposed tuberculated plates (Fig. 323). Two of the lateral plates are notched for the orbits, and between them there is an interorbital plate which either has a pit on its inner surface or is perforated by an open funnel, as in Dinichthys, possibly for a parietal or a pineal organ. Some of the bones present some analogy, to say the least, to certain of the dermal bones of a typical Teleostome, apparently representing such elements as paired parietals and frontals, a dermal mesethmoid, and toothless premaxillae and maxillae (Fig. 324, A). As in the Antiarchi, the anterior portion of the trunk is also armoured, above by a dorsal shield, formed by median and lateral plates, and below by a similarly constructed ventral shield (Fig. 324, B). A huge joint connects the head and trunk shields: hence the term Arthrodira or "joint-neck." The rest of the body is naked.

Pectoral fins are unknown, but pelvic fins, each supported by a stout basal plate or basipterygium, and with traces of radials, are present. There is a small dorsal fin. Little is known of the primary cranium, but in the trunk and tail it is evident that there are well-developed and partially calcified neural and haemal arches associated with a persistent notochord. It is possible that the skull is autostylic. Gill-arches are not known. A pair of plates (Fig. 324, A, j) at the postero-lateral angles of the cephalic shield may perhaps be opercula. The teeth are conical. Those in the upper jaw are supported by two pairs of plates, probably vomers and palatines. In the lower jaw there are two series of teeth, one in front near the symphysis, and the other behind, supported by a single bone in each ramus. There is a well-developed lateral line system, indicated by surface markings on the head and trunk shields.

Fam. 1. Coccosteidae.—Coccosteus occurs in the Devonian of Europe and North America, and includes species of relatively small size, not exceeding half a metre in length. C. decipiens, the best known species, is a characteristic fossil in the Old Red Sandstone of Scotland. Phlyctaenaspis^[634] is found in the Lower Devonian of Canada, England, and Poland. A larger Arthrodiran, with slender toothless jaws, Homosteus,^[635] is met with in the Lower Old Red Sandstone of the North of Scotland, and in the Devonian of Germany and Russia. The Old World Arthrodira must yield, however, to those of the New World for variety in size and shape, and in the character of their dentition.^[636] Some of the North American genera (e.g. Dinichthys) probably attained a length of ten feet, or even, as in Titanichthys, a much greater size. Some are fusiform in shape, but Mylostoma is flattened and Ray-like, and, judging from the dentition, their food and habits must have been equally varied. Mylostoma has tritoral plates not unlike those of Neoceratodus or Chimaera. In others the teeth are single, and conical or pointed; in Titanichthys the front teeth in both jaws are beak-shaped structures. It is highly probable that Titanichthys, Mylostoma, and perhaps other genera, are types of distinct families.

The Arthrodira have been regarded as armoured Dipneusti, a view which is mainly based on their supposed autostylism and the nature of the dentition. But this autostylism has yet to be verified, and, if proved, the possibility that it may be a secondary feature, associated with the evolution of a peculiar dentition, must not be forgotten. Much more may be said for their claim to be regarded as a highly specialised race of primitive Teleostomi. Besides a well-developed lower jaw, bones comparable to the elements of a secondary upper jaw are known, and in a general way the disposition of the cranial roofing bones, and the arrangement of the endoskeletal elements of the pelvic fins, tend to conform to the normal Teleostome type. In fact, Dr. Traquair has expressed the opinion that the Arthrodira are Teleostomi and Actinopterygii.^[637]