On the cervical vertebrae of the

Transcription

1 ~oological Journal of the Linnean Society (1986) 88: With 12 figures On the cervical vertebrae of the Pterodactyloidea (Reptilia: Archosauria) S. C. B. HOWSE Department of Biology, Birkbeck College, Malet Street, London WClE 7HX Receiued Januav 1985, revised and accepted for publication October Within the Pterodactyloidea, the cervical vertebrae show considerable variation. These elements are also sufficiently common and contain enough anatomical information to make them taxonomically valuable. A survey of these vertebrae concludes that most known pterodactyloids fall into two groups: long-necked forms with attenuated cervical vertebrae that possess low neural spines, and tall-spined forms that possess relatively short neck vertebrae with tall neural spines. These two groups may represent natural taxonomic units. However, this is by no means conclusive and can only be tested by the study of other regions of pterodactyloid skeletons. KEY WORDS:- Arehosauria - Pterodactyloidea - cervical vertebrae - morphology - relationships. CONTENTS Introduction Description Pteranodon I0 Ornithocheirus and Nyctosaurus The Delaware specimen Quetzalcoatlus, Titanopteryx and Azhdarcho Doratorhynchus Santanadactyllrs The Jurassic pterodactyloids Other pterodactyloids Discussion and conclusion Acknowledgements References List of abbreviations INTRODUCTION The fossil remains of pterosaurs are relatively rare and the delicacy of their bones is such that they are often crushed and fragmentary. Associated or articulated material is thus even rarer and our understanding of the osteoanatomy of pterosaurs is limited to fewer than 40 genera, mostly monotypic. Apart from the widely agreed subdivision of the Pterosauria into the suborders Rhamphorhynchoidea and Pterodactyloidea, there is little consensus on the systematics of the group, and the intrinsic relationships of the Pterosauria remain unresolved /86/ $03.00/ The Linnean Society of London

2 308 S. C. B. HOWSE The rhamphorhynchoids are the first to appear in the geological record and the first to disappear. Their known history spans from the Upper Triassic to the Upper Jurassic. The pterodactyloids first appear in the Upper Jurassic as contemporaries of the later rhamphorhynchoids and survive into the Maastrichtian of the latest Cretaceous. Most authors consider the pterosaurs to have evolved from archosaurs of the thecodontian grade during the Triassic. However, Wild (1978) argues for an origin from more primitive diapsids of an eosuchian grade. It is uncontroversial that the Pterosauria represent a monophyletic taxon and that pterodactyloids evolved monophyletically from the rhamphorhynchoid grade. The rhamphorhynchoids are the more primitive of the two suborders. The skull is primitively shorter and deeper, with the exception of Rhamphorhynchus. The quadrate does not slant acutely forward to the same degree as in pterodactyloids, and so there is not the same disparity between the skull length and the length of the mandible. The antorbital opening is separated from the external naris by a narrow bar of bone. Teeth are usually present throughout the upper and lower jaws, though there is frequently topographic dimorphism. The anterior teeth are longer than the posterior teeth. The vertebral column contains primitive cervicals that have well-defined diapophyses and parapophyses. The atlas and axis are separate ossifications. The centra of these typically procoelous cervical vertebrae have deep anterior and posterior articular surfaces, but no exapophysial processes. There are up to 40 caudal vertebrae. The glenoid cavity is situated mainly or completely on the scapula. The carpus is less reduced in the number of elements it possesses and the wing metacarpal is less than half the length of the forearm. The pes has a long fifth toe. Most or all of these primitive features are present in Peteinosaurus, Eudimorphodon, Dimorphodon, Campylognathoides and Dorygnathus. Rhamphorhynchus has a relatively long, low skull with an acutely slanting quadrate and consequently there is a considerable disparity between the length of the skull and the lower jaw. Rhamphorhynchus is similar to the pterodactyloids in these respects. The pterodactyloid skeleton shows many additional derived characters. The skull is often very long and usually low and slender. The quadrate slants acutely forward creating a disparity between the length of the skull and the length of the mandible. As a result of the realignment of the quadrate, the infraorbital fenestra has migrated downwards and forwards under the orbit. The antorbital fenestra and external naris are confluent, though some forms retain an incomplete bar of bone marking the boundary between the two vacuities. The teeth are usually, but not always, small and morphologically similar. The teeth may be restricted to either the anterior (Ornithodesmus) or the posterior (Dsungaripteras) regions of the jaw or they may be evenly distributed. Some genera such as Pterodaustro possess extremely specialized teeth. Certain genera from the later Cretaceous are toothless (Pteranodon). The vertebral column contains variably modified cervicals while the tail is reduced to a few spool-shaped caudals. In some advanced forms the atlas and axis are completely fused into the co-ossified atlanto-axis. Cervical ribs appear to be absent in cervicals 1-7 (in Pteranodon the last two cervicals, 8 and 9 bear double-headed ribs). In some of the larger forms the anterior dorsals and associated ribs are extensively fused to form the notarium, which acts as a rigid

3 PTERODACTYLOID CERL ICAL VERTEBRAE 309 strut against which the scapula is braced (Wellnhofer, 1978). The glenoid facet is shared by both the scapula and the coracoid which may be fused into a single functional unit. The carpus is reduced to three elements. The wing metacarpal is longer than the forearm. There are two classifications of the pterodactyloids in the recent literature: one compounded by Romer (1956) and the other by Wellnhofer (1978). Romer subdivided the Pterodactyloidea into the two families Pterodactylidae and Ornithocheiridae, whereas Wellnhofer partitioned the Pterodactyloidea into nine families, namely Pterodactylidae, Germanodactylidae, Ctenochasmatidae, Pterodaustridae, Ornithodesmidae, Dsungaripteridae, Ornithocheiridae, Criorhynchidae and Pteranodontidae. Most of the families within Wellnhofer s classification comprise only one or two genera. These extremes of lumping by Romer and splitting by Wellnhofer reflect the lack of characters which vertebrate palaeontologists can use when considering suprageneric relationships within the order. Because of the usually fragmentary nature of the remains, insufficient morphological information has been available with which to attempt a more closely argued classification. The most structurally diverse components of the skeleton are the skull and mandibles but these are rarely preserved. The commonest remains are the long bones which are relatively uniform and hence bear little useful taxonomic information. Among the anatomically informative skeletal elements are the cervical vertebrae. Within the Pterodactyloidea, cervical vertebrae are known for the following genera: Pterodactylus (Tithonian), Gallodactylus (Tithonian), Germano- dactylus (Tithonian), Ctenochasma (Tithonian), Doratorhynchus (Tithonian-Berri- asian?), Pterodaustro (Lower cretaceous), Dsungaripterus (Lower Cretaceous), Noripterus (Lower Cretaceous), Ornithodesmus (Valanginian-Barremian), Ornithocheirus (Valanginian-Turonian), Santanadactylus (Aptian), Nyctosaurus (Albian- Turonian), Titanopteryx ( Albian-Turonian), Azhdarcho (Turonian-Coniacian), Pteranodon (Santonian), Quetzalcoatlus (Maastrichtian). The quality of the material and hence the amount of information available varies. In particular the Solnhofen and Niobrara fossils are usually crushed and interpretation is difficult. The following pterodactyloid cervical vertebrae were studied at first hand: (1) Material from the Sedgwick Museum, Cambridge (SM): (a) Ornithocheirus: atlanto-axis, B54305; postaxial cervicals, B54324, B54328, , B ; (b) Doratorhynchus: postaxial cervical, (2) Material from the Yorkshire Museum, York (YM): (a) Ornithocheirus postaxial cervical: 1983/383F. (3) Material from the Booth Museum of Natural History, Brighton: (a) Ornithocheirus: postaxial cervical, G (4) Material from the British Museum (Natural History), London (BMNH): (a) Quetzalcoatlus: casts of postaxial cervicals, R9323, R9325, R9326 (Texas Memorial Museum numbers TMM , TMM , TM ) ; (b) Ornithodesmus: postaxial cervical, R3877; (c) Pteranodon: postaxial cervicals, R2929, R4537; (d) Pterodaustro: cast, R (PVL 3860). The literature sources are: Arambourg (1959), Baird & Galton (1981), Bonaparte (1969), Currie & Russell (1982), de Buisonji. (1980), Eaton (1910),

4 310 S. C. B. HOWSE Figure I. Plcrunodon Marsh; restored atlanto-axis, left view, (after Eaton, 1910). Langston ( ), Lawson ( 1975), Nesov ( 1984), Romer ( 1956), Seeley ( 1875, 1901), Wellnhofer (1970, 1975, 1978), Wild (1978), Williston (1903), Young (1964). DESCRIPTION Pteranodon Eaton ( 1910) stated that Pterunodon possessed nine cervical vertebrae organized into three distinct types. For convenience these vertebrae can be described as belonging to the anterior, middle and posterior series. The first vertebra is the atlas, the second is the axis and vertebrae three to seven form the middle-series. Vertebrae eight and nine define the posterior series. Wellnhofer Figure 2. Nyctosuurus gracilis Marsh; atlanto-axis, right view (after Williston, 1903).

5 PTERODACTYLOID CERVICAL VERTEBRAE 31 1 Figure 3. Omithocheirus Seeley; atlanto-axis, right view, SM B54305 ( 1970) identified only seven cervicals in the Jurassic pterodactyloids, apparently treating Eaton s posterior series cervicals as anterior dorsals. The anterior series (the atlas and axis) of Pteranodon (Fig. 1) is completely COossified to form a single distinctive element, the atlanto-axis. Such a configuration is not found in the rhamphorhynchoids or in most pterodactyloids. However, a virtually identical atlanto-axis is found in Nyctosaurus (Williston, 1903) (Fig. 2) of the Niobrara Chalk and also in forms from the Cambridge Greensand, referred to Ornithocheirus (Fig. 3). The atlanto-axis of Pteranodon is procoelous, as are all the other cervicals, but has a short, deep-bodied centrum. The lateral surfaces of the centrum are excavated and these depressions house rounded pneumatic foramina. The anterior cotyle is deep and hemispherical and surrounded by a well-defined rim. The posterior condyle is a shallow oval structure and the posterior ventrolateral margins of the centrum are extended into thick-based, stoutly developed postexapophyses. There are no pre-exapophysial processes. The neural spine forms a high, narrow, thin blade-like structure, the leading edge of which curves backwards, and the whole blade strongly tapers towards its narrow truncated apex. High on the rear margins of the neural arch are well-developed postzygapophysial processes which terminate in tubercular extensions. Cervicals three to seven (the middle series) are all morphologically similar in Pteranodon (Fig. 4), except that there is an increase in length from cervical three to cervical five and then a decrease in length down to cervical seven (Eaton, 1910). For one specimen Eaton listed the centrum length for each of vertebrae three to seven as 61, 67, 83, 76 and 65 mm. This topographic disparity in relative size appears to be a typical, but not absolute, feature of pterodactyloid necks. All middle-series cervicals have relatively elongated and proportionally shallow centra. In each vertebra the anterior cotyle is wide and shallow and slightly crescentic in shape. The posterior condyle is a shallow, oval structure, which is slightly depressed on its upper surface. The lateral sides of the centrum are deeply excavated and these hollows possess the ovoid pneumatic foramina.

6 312 S. C. B. HOWSE Figure 4. Pteranodon Marsh; restored middle-series cervical vertebrae. A, Sixth cervical, left lateral view; B, fourth cervical, anterior view; C, fourth cervical posterior view (after Eaton, 1910). Beneath the anterior cotyle, and developed from the median ventral surface of the centrum, is a keel-like hypapophysis with its apex situated immediately below the leading ventral margin of the centrum. Both anterior and posterior ventrolateral surfaces are developed into exapophysial processes. The exapophyses appear to be physically separate structures from the cotyle and condyle with which they are associated. The neural spines of the middle-series cervicals are long-based, thin blades which taper to narrow, truncated apices. The leading edge of each spine curves markedly backwards whereas the posterior edge is more or less vertically aligned. The leading edges of the neural arch are extended forwards into the prezygapophyses. Positioned somewhat higher on the trailing edges are the postzygapophysial processes, which in Eaton s restoration appear to overlap the limits of the centrum marked by the posterior condyle. Both anterior and posterior zygapophyses terminate in tubercular extensions of their processes. Linking the anterior to the posterior tubercles, on both sides of the vertebrae, are ridge-like transverse processes

7 PTERODACTYLOID CERVICAL VERTEBRAE 313 which curve upwards from the front to the back. No diapophysis is developed. The middle-series cervicals of Pteranodon do not bear ribs. All the middle-series cervicals in Pteranodon are very similar, according to Eaton. Their transverse processes have the same orientation relative to the long axis and this is possibly due to the presence of the pre-exapophyses which appear to be unique to Pteranodon. They are certainly not found in its near relatives Nyctosaurus and Ornithocheirus. The posterior-series cervicals, which comprise the eighth and ninth neck vertebrae, are morphologically distinct from the preceding elements. Like the anterior cervicals their centra are procoelous but the centrum bodies are much shorter. The length and depth are similar. These vertebrae have well-developed pre- and postzygapophysial processes and each appears to have both anterior and posterior exapophyses. The neural spine is narrow based and rises almost vertically from the apex of the neural arch. The truncated upper margin is approximately the same length as the base of the spine. The transverse processes are markedly orientated in an anteroventral-posterodorsal plane. Each process bears a diapophysis on its mid-section. There is also a capitular facet at the base of each pre-exapophysis. The eighth and ninth cervicals bear double-headed ribs. The posterior cervicals are structurally similar but not identical to the dorsals. The posterior dorsals have pronounced diapophyses but no capitular facets, and their neural spines are much narrower than those of the posterior cervicals. The centra of the dorsals bear no exapophyses and the anterior cotyle and posterior condyle are more hemispherical than in the cervicals. Ornithocheirus and Nyctosaurus An atlanto-axis, (SM B54305) from the Cambridge Greensand referred to Ornithocheirus (Fig. 3) displays a close resemblance to that of Pteranodon. It possesses a short, deep-bodied centrum. Its anterior margin is occupied by a hemispherical cotyle, the base of which bears a blunt and somewhat ill-defined hypapophysis. The posterior condyle is a shallow and ovoid structure; although it is damaged, its original shape would have been similar to those of the postaxial cervicals, The left postexapophysis is missing and only the base of the right process remains; however, this is very deep. The lateral sides of the centrum body house rounded pneumatic foramina. The postzygapophyses are high on the margins of the neural arch. Prezygapophyses are absent. The overall form of this element agrees with Williston s figure for Nyctosaurus (Fig. 2). The middle-series cervicals from the Cambridge Greensand (Fig. 5) have neural spines which broadly resemble those of Pteranodon and Nyctosaurus (Fig. 6), except that the leading edge of the spines of the Greensand vertebrae are more vertically aligned. The body of the centrum is long and shallow and very similar in general shape to that of Pteranodon. However, there are no preexapophysial processes and in this respect the Greensand material resembles flyctosaurus. Stout postexapophyses project divergently from the posterior margin of the centrum. The bases of these structures blend into the posterior condyle and are not physically separate as in Pteranodon (Eaton, 1910). An anteriorly tapering shallow excavation on the ventral surface of the centrum separates the diverging apophysial bases. The mid-section of the centrum body is

8 314 S. C. B. HOWSE Figure 5. Ornithocheirus Seeley; middle-series cervical vertebrae. A, Left lateral view of SM B54781; B, left lateral view of SM B54767; C, anterior view and D, posterior view of YM 1983/383F. conspicuously waisted and in this region the lateral sides are excavated and house ovoid pneumatic foramina. Only in the Greensand fifth cervicals are the transverse processes aligned in parallel fashion as in Pteranodon. The other cervical transverse processes are variably constructed, generally with an anteroventral-posterodorsal orientation. A notable difference between Eaton s restoration of the middle-series cervicals and the three-dimensional material from the Greensand is the shape of the anterior cotyle. In the Greensand specimens the upper rim of the cotyle curves upwards, not downwards as in Eaton s restoration. In some specimens there is a pronounced flange on the mid-section of the lower rim which marks the base of the hypapophysis. The cotyle is separated from the laterally positioned prezygapophyses by a trough, the margins of which are delineated by the cotylar rim on one side and the apophysial base on the other. Finally, unlike Eaton s restored figures of the Pteranodon middle series, the Cambridge Greensand vertebrae have centra which extend well beyond the postzygapophysial processes. The two cervical vertebrae of Pteranodon which were examined (BM(NH) R2929, R4537), both possess an anterior cotyle of similar

9 PTERODACTYLOID CERVICAL VERTEBRAE 315 Figure 6. Nyctosuurus grucilis Marsh; middle-series cervical vertebrae. A, Fifth cervical, dorsal view; B, sixth or seventh cervical, left lateral view (after Williston, 1903). construction to that in the Cambridge Greensand material. The posterior condyles must be correspondingly similar. Furthermore, the posterior margins of the centrum extend beyond the postzygapophysial processes. Therefore, it would appear that in these features at least, Eaton s restorations are in error and that the Pteranodon cervicals resemble those of Ornithocheirus and Nyctosaurus as implied by Baird & Galton (1981). It is also possible that his restoration of the neural arch is incorrect. Eaton s concept of an arch with a broad rounded apex does not correspond to the configuration of the specimens which were studied during this work. Although these specimens are crushed dorsoventrally, the apex of the arch still forms an acute angle at the base of the neural spine. The Delaware specimen Baird & Galton (1981) described a large pterodactyloid cervical from the Merchantville Formation (early Campanian) of northern Delaware. Unlike the material from the Niobrara Chalk, this specimen is uncrushed and is relatively well preserved for a pterosaur. The specimen resembles Nyctosaurus and Ornithocheirus in its overall structure. Baird and Galton considered it to be either the fifth or the sixth cervical. The centrum has a long and shallow body, waisted in its mid-section and with excavated lateral surfaces containing the pneumatic

10 316 S. C. B. HOWSE foramina. The anterior cotyle is identical to those possessed by the Cambridge Greensand vertebrae; the upper rim curves upwards and the lower rim is flanged. The posterior condyle is a shallow ovoid. The prezygapophysial bases are level with the ventral surface of the centrum. The transverse processes linking the anterior and posterior zygapophyses are steeply inclined, approximately 40" from the horizontal. Like all pterosaur middle-series cervicals, this specimen once had a robust hypapophysis and has well-developed postexapophyses. The centrum body extends posteriorly well beyond the limits of the postzygapophysial processes. Unlike Pteranodon, but similarly to Nyctosaurus and Ornithocheirus, pre-exapophyses are absent. The neural spine is long-based, but most of this structure is missing. Anteriorly the neural canal penetrates the neural arch as a small tube, and on either side of the arch lateral foramina pierce the pedicel. the anterior apex of the neural arch was connected to the base of each prezygapophysis by a well-defined ridge. In the specimen only the left ridge is preserved and this rises steeply from the apophysial base to the apex of the arch at an angle of approximately 45" from the horizontal. The configuration is the same in the Cambridge Greensand vertebrae but unlike the low-vaulted condition of Eaton's restoration of Pteranodon. Thus, in several respects the Delaware cervical resembles the commonest cervical vertebrae from the Cambridge Greensand, but is markedly different from the Pteranodon middle series figured by Eaton, except in size. It has a centrum length of 67 mm and must have belonged to an animal much larger than any known specimen of Nyctosaurus, and perhaps comparable in size to Pteranodon longiceps (Baird & Galton, 1981). It demonstrates that forms related to Pteranodon attained a similar size to that genus. Quetzalcoatlus, Titanopteryx and Azhdarcho Lawson (1975) described the remains of a number of individuals belonging to a genus of large pterosaur, Quetzalcoatlus, discovered in sediments of Maastrichtian age in the Big Bend National Park of Texas. This genus is notable both for its size and for the peculiar nature of its cervical vertebrae (Fig. 7). Three such vertebrae have been discovered from presumed juvenile individuals (Langston, 1981) and these probably represent the third, fourth and fifth cervicals. They are unusually elongate, particularly the fifth cervical. The respective lengths for the centrum body of each are 14.5, 20.8 and 37.4 cm. All three vertebrae have long and very shallow centra. Although a certain amount of crushing and distortion has taken place they would still have originally been dorsoventrally depressed. Each centrum is pinched into a narrow waist about two-thirds of the distance along its length. The anterior cotyle of the longest vertebra is very similar to those in the Delaware and Cambridge Greensand specimens. The vertebrae possess well-developed pre- and postzygapophyses. The prezygapophyses form horn-like processes and these are only slightly divergent in the longest vertebra. The postzygapophysial processes are stoutly built structures with broad bases and are widely divergent. The centrum bears welldeveloped postexapophyses and, in accord with all known forms except Pteranodon, pre-exapophyses are absent. The bases of the postexapophysial processes blend into the ovoid condyle in a similar fashion to the Cambridge Greensand vertebrae and unlike the physically separate structures of Pteranodon.

11 PTERODACTYLOID CERk ICAL VERTEBRAE 317 Figure 7. Quelzalcoallus Lawson; fifth cervical vertebra of the middle series. Drawn in dorsal view from cast BMNH R9325 of original TMM The centrum extends posteriorly well beyond the limits of the postzygapophyses in the longest vertebrae but not to the same extent in the third. cervical. The condition of the fourth element is unknown. The fifth cervical has a neural arch which forms a low vaulted structure with a broad apex. In the third and fourth vertebrae the apex of the arch is, at its rear extremity, raised above the general plane of this structure. In the fifth cervical the arch is depressed in the middle. The neural spine is reduced in all three vertebrae but to the greatest extent in the fifth element. In this vertebra the spine is present as a low ridge towards the anterior and posterior ends of the neural arch. It is present as a thin, very slightly raised line in the middle section. The same condition is also true for the

12 318 S. C. B. HOWSE fourth cervical. However, in the third member the neural spine reaches its greatest height in the mid-region of the arch. The transverse processes are nearly parallel to the long axis of the centrum in the fifth cervical. In the other two the processes are somewhat inclined from front to back. This condition is probably true for most pterodactyloids. However, the overall anatomy of these vertebrae is markedly different from the middle-series of Pteranodon, Jyctosaurus and Ornithocheirus. The description of another vertebra very similar to the fifth element from Texas predates Lawson s paper. Arambourg ( 1959) described a supposed metacarpal of a very large pterosaur, Titanopteryx, from the late Cretaceous (Albian-Turonian) of Jordan. This element is in fact a cervical vertebra (Lawson, 1975) and probably the longest cervical (i.e. the fifth) of the middle series. It appears to have nearly parallel transverse processes. This bone is notable for its length of 60cm (Arambourg, 1959), and it is therefore considerably longer than the fifth element from Texas. Otherwise, this attenuated vertebra with its reduced neural spine is almost identical to the fifth cervical of Quetzalcoatlus, only size differentiating the two. The recently described Azhdarcho (Nesov, 1984) from the Upper Cretaceous of the Kyzyl-kum region of the Uzbek S.S.R., has cervical vertebrae which closely resemble those of Quetzalcoatlus in every respect. Nesov associates this genus with Quetzalcoatlus and Titanopteryx in the subfamily Azhdarchinae, which is thus the senior family-level name for this group of long-necked pterodactyloids. Nesov also places this subfamily within the Pteranodontidae but it is an implicit conclusion of the work presented here that the tall-spined and long-necked forms are not closely related and probably have no common ancestor more recent than the Upper Jurassic. A much smaller, but anatomically identical, cervical from Alberta (Campanian?) has been described by Currie & Russell (1982). Only the anterior 38 mm of the specimen is preserved. Characters which suggest relationship to Quetzalcoatlus are the typical, horn-like, slightly divergent prezygapophyses, a suppressed neural spine, and a low-vaulted neural arch which is anteriorly broad and has a rounded apex. Doratorhynchus Seeley (1875) identified a vertebra (SM 55340) from the Purbeck Limestone of Langton Matravers near Swanage, Dorset (Tithonian-Berriasian), as a caudal. He referred this specimen and a mandible found at the same locality to the genus Doratorhynchus. His initial interpretation was based on the resemblance of the Purbeck vertebra to certain elements from the Cambridge Greensand which he had assumed to be pterosaur caudals. This cervical vertebra (Fig. 8) is similar in its morphology to the fourth or fifth cervicals of Quetzalcoatlus. It is crushed and distorted in such a way that the apex of the neural arch has been displaced to the left of the centrum midline. This distortion gives the impression of a much broader vertebra than would have existed in life. The centrum is 10.9 cm long; halfway along its right side a small oval pneumatic foramen can be identified. The extremities of the centrum are badly crushed and fractured, and as a result the condition of the anterior

13 PTERODACTYLOID CERVICAL VERTEBRAE 319 Figure 8. Doratorhynchus ualidus Seeley; middle-serics cervical vertebra in dorsal view. SM cotyle and posterior condyle is unknown. However, on the posterior left lateral margin a well-developed postexapophysis is indicated. The neural arch is low and broad at the ends. This is partly due to crushing but the uncrushed arch would still have been a low-vaulted and somewhat rounded structure. The whole arch tapers to a waist from both ends. The waist is narrowest some two-thirds of the distance from the anterior margin. The left anterior margin of the arch is extended out into a long, thin, horn-like prezygapophysis with an ovoid articular facet. The right prezygapophysis is missing. It appears that in the uncrushed vertebra the prezygapophyses would have been nearly parallel, or very slightly divergent. This condition is similar to that of the longest cervical of Quetzalcoatlus. In Doratorhynchus the remaining prezygapophysis is proportionally much longer than the same structure in the Texas specimen. The left postzygapophysis is present but is much more worn than the anterior process. It is a broad and blunt structure and apparently bears a pronounced tubercle. The neural spine is reduced to a thin rib of pencil-line thickness for most of its length. Fracturing at the anterior and posterior ends indicates that it was more developed in these regions and quite probably rose to a pronounced ridge. This condition resembles that in the longest element of

14 320 S. C. B. HOWSE Figure 9. Doratorhynchus validus Seeley; middle-series cervical vertebra, SM Restored in dorsal view. Quetzalcoatlus. When reconstructed (Fig. 9) the Doratorhynchus cervical is strikingly similar to the fifth cervical of Quetzalcoatlus. Seeley s initial misinterpretation of the Purbeck specimen as a caudal leads to an interesting reassessment of some of the Greensand material. Many of the vertebrae from the Cambridge Greensand that he identified as caudals are probably cervical vertebrae of small, long-necked pterosaurs. Although worn, many of these long, thin vertebrae resemble the larger elongated cervicals previously mentioned but do not resemble the caudal vertebrae of any known pterosaur. Santanadactylus Santanadactylus is a recently described pterodactyloid from the Aptian of Brazil. Only two cervical vertebrae are known for this genus but they yield enough information to be useful for comparative purposes. The vertebrae come from the lower part of the Santana Formation of the Chapada do Araripe and were described by de BuisonjC (1980). The elements are in contact between their zygapophyses and held together by matrix. The front half of the anterior

15 PTERODACTYLOID CERVICAL VERTEBRAE 32 1 vertebra is missing. The posterior vertebra is much more complete and is the longer of the two. It has a centrum length of 8.8 cm. De Buisonjt estimated the length of the anterior centrum to be 7.3 cm. The longer of the two vertebrae is regarded as the fourth or fifth cervical and the shorter one as the third or fourth. The centrum of the larger of the two vertebrae is long, shallow and deeply excavated. Again, there is the typical waist at a point approximately midway along its length. Above the waist, on both sides, the centrum bears two relatively small, ovoid foramina, the anterior one situated slightly higher than the posterior one. The neural arch is a low, broad structure penetrated by a restricted neural canal, on either side of which are situated the lateral foramina. The neural spine runs the full length of the neural arch. It is low and, as restored by de Buisonjt, of uniform height. Its upper boundary is strengthened by a ridge. The neural spine of the anterior vertebra is somewhat higher. The spine rises to its greatest height at a point slightly behind the posterior margin of the neural arch. Then, according to de Buisonjt s reconstruction, it gradually drops to the precipice of its truncated front margin. The vertebrae of Santanadactylus resemble the cervicals of the long-necked Plerodactylus species in general form (see below) and in the strong similarity of the neural spine in the two genera. The major difference between the two genera is that Santanadactylus has postexapophyses. The Jurassic pterodactyloids The European Jurassic pterodactyloids are readily divisible into two groups: those with long, relatively slender cervical vertebrae and those with short, stocky cervicals. Pterodactyloids with long cervicals include Pterodactylus antiquus, Pterodactylus longicollum and Ctenochasma gracile (all Lower Tithonian). Forms with short cervicals include Pterodactylus kochi (type specimen), Pterodactylus elegans, Plerodactylus micronyx, Germanodactylus rhamphastinus (all Lower Tithonian) and Gallodactylus canjuersensis (Middle Tithonian). In all the examples mentioned above, the elements of the skeleton are to some extent crushed and therefore the information content of the cervical vertebrae is limited. The atlas and axis are primitively separate structures. The middle-series cervicals of Pterodactylus antiquus (Wellnhofer, 1970) (Fig. 10) possess long, shallow centra without postexapophysial processes. The posterior condyle of the centrum is situated well beyond the limit of the postzygapophysial processes. A hypapophysis is developed on the anteroventral midline of the centrum. The neural spine forms a low, flange-like structure. Its leading edge is curved backwards and the spine attains its greatest height somewhat behind the front apex of the neural arch. The upper border drops very slightly towards the rounded trailing edge. The overall impression is of a low neur,al spine of uniform height. A similar morphology appears in Pterodactylus longicollum and Ctenochasma gracile. The middle-series cervicals of Pterodactylus antiquus display a strong similarity to the cervicals of Quetzalcoatlus and Doratorhynchus, although Pterodactylus lacks postexapophyses. The short-necked pterodactyloids of the Jurassic, such as the holotype of Pterodactylus kochi, possess shorter but more robust cervicals with relatively

16 322 S. C. B. HOWSE Figure 10. Pterodactylus antiquus Soemmerring; middle-series cervical vertebra in left lateral view restored with position of pneumatic foramen indicated (after Wellnhofer, 1970). prominent neural spines. In Pterodactylus kochi (holotype) the hypapophysis of each vertebra is exaggerated, its leading edge extending forwards in the form of a spine-like process. The prezygapophyses are short, wide-based structures with extensive articular surfaces for the reception of equally well-built postzygapophyses. The middle-series cervicals of Germanodactylus rhamphastinus and Gallodactylus canjuersensis probably possessed short, deep-bodied centra and prominent neural spines. The spinous hypapophyses of Pterodactylus kochi may be a specialized structure peculiar to this species. Wellnhofer (1970) listed the various lengths of the middle series vertebrae for the holotype of Pterodactylus antiquus, a specimen of Pterodactylus longicollum (both long-necked forms), the neotype of Pterodactylus micronyx and the holotype of Germanodactylus rhamphastinus (both short-necked forms). These are given in Table 1. With the exceptions of Pterodact_ylus kochi and Pterodactylus elegans, the known Jurassic forms possess the same pattern of relative length for the middle-series cervicals described for Pteranodon. The fifth cervical is always the longest and the third, or seventh, is the shortest. Pterodactylus kochi and Pterodactylus elegans are unusual in that they appear to possess middle-series cervicals of uniform length. In this respect they are similar to the more primitive rhamphorhynchoids in their organization. Other pterodactyloids The other South American pterodactyloid whose cervical vertebrae are preserved is Pterodaustro, described by Bonaparte ( 1969). The remains are crushed, but the neck vertebrae have a superficial resemblance to those of Santanadactylus and the longer-necked Jurassic forms from southern Germany. Table 1. Lengths of middle-series vertebrae (mm) (from Wellnhofer, 1970) 3rd 4th 5th 6th 7th Pterodactylus antiquus Pterodactylus longicohm Pterodactylus microryx Germanodactylus rhamphastinus

17 FTERODACTYLOID CERVICAL VERTEBRAE 323 The known cervical vertebrae attributed to Ornithodesmus from the Wealden of southern England are very fragmentary and distorted but the visible structure of the centrum resembles that in Pteranodon, Nyctosaurus and Ornithocheirus. Young ( 1964) described two cervicals of the genus Dsungaripterus (Lower Cretaceous) as having a similar morphology to those of Ornithodesmus but, as the latter are poorly known, little can be made of this statement. DISCUSSION AND CONCLUSION Pterosaurs appear to have a uniform number of neck vertebrae, a total of seven or nine depending on whether the last two vertebrae are considered to be cervicals or dorsals. The postaxial cervical vertebrae, defined in this paper as the middle series, are more massive structures than the dorsal vertebrae which follow them. This is especially true of Rhamphorhynchus and the pterodactyloids. The cervical vertebrae are procoelous throughout the order and all appear to possess pneumatic foramina. The latter are housed in lateral excavations of the centrum body immediately beneath the pedicel of the neural arch. In rhamphorhynchoids the postaxial cervicals bear double-headed ribs which articulate with well-developed diapophyses and parapophyses. The anterior cotyle and posterior condyle are hemispherical. There is no hypapophysis developed on the anterior ventral surface of the centrum and, as in the Jurassic pterodactyloids, there are no exapophyses. The transverse processes of the postaxial cervicals appear to have near-parallel alignment throughout the series, though this is not conclusive. The postzygapophyses extend beyond the posterior limits of the centrum. In the more primitive rhamphorhynchoids the individual members of the middle series all have an equal length and the centrum body is short and deep. In Rhamphorhynchus there is a disparity in the size of the various members of the middle-series which is dependent on sequential position of each element. The fifth cervical is the largest and the third or seventh is the smallest (Wellnhofer, 1975), and the centrum body in Rhamphorhynchus (unlike those of primitive rhamphorhynchoids) is relatively long and shallow. Therefore, the postaxial cervicals of Rhamphorhynchus have certain features in common with most known genera of the Pterodactyloidea, while Pterodactylus kochi and Pterodactylus elegans display the condition of the primitive rhamphorhynchoids. Where known, the atlas and axis vertebrae of rhamphorhynchoids are separate structures. The resemblance of the postaxial cervicals of Rhamphorhynchus to those of the pterodactyloids (excluding Pterodactylus kochi and Pterodactylus elegans) is probably the result of convergent evolution, although it must be noted that it is possible that an evolutionary reversal has occurred and that these two Pterodactylus species have reverted to the primitive condition of equal-length vertebrae. The pterodactyloids, with the exception of Pterodactylus kochi and Pterodactylus elegans, possess middle-series cervicals with a disparity in length in which the fifth is always the longest. The centra, like those of Rhamphorhynchus, are long and shallow but bear no ribs, and the transverse processes vary in alignment. The anterior cotyle is crescentic with the crescent directed upwards. A welldeveloped hypapophysis is associated with this structure. The posterior condyle is ovoid and the centrum body extends posteriorly well beyond the

18 324 S. C. B. HOWSE postzygapophysial processes. The centrum body has a conspicuous waist formed by the extensive lateral excavations typical of pterodactyloid cervicals, especially those of the Cretaceous genera. The Jurassic pterodactyloids, with the exception of Pterodactylus kochi, possess no unique features of the cervical vertebrae and hence provide no derived characters which might define the genus Pterodactylus. The long-necked forms are anatomically similar to the later (Cretaceous) long-necked pterodactyloids. Those Jurassic pterodactyloids with shorter neck vertebrae are similar to the ornithocheirids in that they possess high, blade-like neural spines. All Jurassic forms are essentially differentiated from most Cretaceous pterodactyloids by the absence of exapophyses. Pterodactylus kochi is unique in that its middle-series vertebrae have exaggerated spine-like hypapophyses. Most advanced Cretaceous pterodactyloids readily fall into two groups defined on the basis of their neck vertebrae. These are the long-necked forms (e.g. Doratorhynchus, Azhdarcho, Titanopteryx and Quetzalcoatlus) and the tall-spined forms (e.g. Omithocheirus, Nyctosaurus, the Delaware material and Pteranodon). Both groups possess the crescent-like anterior cotyle; its lower rim may be flanged in mid-section. The flange forms the base of the hypapophysis. The ovoid posterior condyle is intimately associated with postexapophyses which blend into the former structure s lateral margins. The transverse processes of the fifth cervical vertebra are more or less aligned parallel to the long axis of the centrum body. The transverse processes of the other vertebrae are variously aligned according to their position within the series. The major characteristics of the cervicals of the long-necked Cretaceous forms are: (1) extreme length and attenuated shape; (2) low-vaulted neural arch with broad, rounded apex; (3) neural spine reduced to a low ridge, which is only expressed at its anterior and posterior ends in the fifth element; and (4) relatively slender, horn-like prezygapophyses only slightly divergent from each other. The shorter neck vertebrae of the tall-spined forms differ in that they possess (besides the difference in length): ( 1) high, tapering, blade-like neural spines, each with a long base and a narrowly truncated upper border; (2) high-vaulted neural arch with an acutely angled apex; and (3) relatively short, robust and divergent pre- and postzygapophysial processes. It can be argued that the long-necked Jurassic pterodactyloids (e.g. Pterodactylus antiquus and Pterodactylus longicollurn) are related to the long-necked Cretaceous pterosaurs because they share relatively long middle-series cervicals, the low aspect of the neural spine and the low-vaulted, arcuate neural arch. Ornithocheirus, Nyctosaurus and Pteranodon share a general conformity of their middle-series cervicals. They also share a distinctive, completely fused atlantoaxis, the morphology of which is essentially the same in all three genera. This element differs from the separate atlas and axis of the rhamphorhynchoids and Jurassic pterodactyloids. The condition of the atlas and axis in the Cretaceous long-necked pterosaurs is unknown. Pteranodon is unique in that its postaxial cervicals bear pre-exapophyses. On the basis of the evidence outlined above most pterosaur genera can be organized into eight groups. These are as follows: (1 ) Primitive rhamphorhynchoids with postaxial cervicals of equal length (Eudirnorphodon, Dirnorphodon, Scaphognathus) ;

19 PTERODACTYLOID CERVICAL VERlEBRAE 325 (2) Advanced rhamphorhynchoids with postaxial cervicals, the lengths of which vary according to their position within the series (Rhamphorhynchus); (3) Pterodactyloids with postaxial cervical vertebrae equal in length throughout the series in which the centra are short and deep (Pterodactylus kochi, Pterodactylus elegans). The following categories (4-8) possess postaxial cervicals of variable length: (4) Pterodactyloids with short and deep postaxial cervicals without postexapophyses (Pterodactylus micronyx, Germanodactylus rhamphastinus, Gallodactylus canjuersensis); (5) Pterodactyloids with long and shallow postaxial cervicals without postexapophyses (Pterodactylus antiquus, Pterodactylus longicollum, Ctenochasma gracile) ; (6) Advanced pterodactyloids with very long, shallow, postaxial cervicals with postexapophyses (Doratorhynchus, Azhdarcho, Titanopteryx, Quetzalcoatlus) ; (7) Advanced pterodactyloids with relatively short, shallow postaxial cervicals which have postexapophyses and prominent neural spines (Omithocheirus, Nyctosaurus) ; (8) Advanced pterodactyloids with postaxial cervicals which are like those in (7) but with pre-exapophyses (Pteranodon). This analysis of cervical vertebrae alone is not sufficient to justify a change in classification because of the possibility of evolutionary convergence within a single system. Hopefully the information outlined in this paper will be useful in connection with the study of other regions of pterosaur osteoanatomy. To summarize, during the Cretaceous there were two major groups of advanced pterodactyloids, recognizable by the construction of their cervical vertebrae. The resemblances among the long cervical vertebrae of Doratorhynchus, Azhdarcho, Titanopteryx and Quetzalcoatlus are sufficient to link the four genera in a single taxonomic group for which the senior family-level name is Azhdarchidae Nesov, While the Jurassic long-necked pterodactyloids possess no unique features, the overall anatomy of the cervical vertebrae of such types as Pterodactylus antiquus and Pterodacolus longicollum indicate a relationship to the previously mentioned Cretaceous forms. Santanadactylus appears to be little modified from the long-necked Jurassic genera except for its possession of postexapophyses. Pteranodon, Nyctosaurus and Ornithocheirus belong to the second Cretaceous group for which the senior family-level name is Ornithocheiridae Seeley, Their middle-series cervical vertebrae share a number of characteristics with both the long-necked Jurassic and Cretaceous pterodactyloids. The shallow state of each centrum, and its posterior limits situated well beyond the postzygapophyses, are obvious features that all three groups have in common. Where known, the anterior cotyles and posterior condyles of the ornithocheirids are virtually identical to those of the Cretaceous long-necked forms. The presence of postexapophyses likewise appears to link the two Cretaceous groups. The presence, on each side of the neural arch, of a long ridge-like transverse process is another character shared by all three groups. The major differences lie in the length of the vertebrae, the form of the neural arch and the height and shape of the neural spine. The relatively short cervical vertebrae with tall neural spines, belonging to the ornithocheirids, are similar to those of the short-necked

20 326 S. C. B. HOWSE Tall-spined 0 e 2 t 2 Figure 11. Cladogram of the pterodactyloid relationships based only on the morphology of the cervical vertebrae. Key to numbered nodes: 1, middle-series cervicals, rib-bearing and of the same length throughout the series; 2, middle-series cervicals, rib-bearing but with variation in length throughout the series; 3, loss of ribs in middle-series cervicals but equal or subequal length throughout the series; 4, elongated and spinous hypapophysis; 5, variation in length of middle-series cervicals throughout the series (convergent with 2); 6, presence of postexapophyses on cervicals (convergent with 9), completely fused atlanto-axis; 7, presence of pre-exapophyses on postaxial cervicals; 8, elongate middle-series cervicals, with low neural spine; 9, presence of postexapophyses on cervicals (convergent with 6); 10, neural spine greatly reduced and exaggerated length of vertebral body. Jurassic pterodactyloids. The long cervical vertebrae with low neural spines and arches, belonging to the long-necked Jurassic group, are very similar in overall form to the long-necked Cretaceous group. On the evidence that has been presented there would appear to be two positions which the tall-spined ornithocheirids could occupy within pterodactyloid phylogeny. They could be placed with the short-necked Jurassic pterodactyloids (Fig. 11). In this case, postexapophyses must have evolved twice. Unfortunately the state of the anterior cotyles, and posterior condyles, of the short-necked Jurassic forms is not known and therefore offers no assistance in resolving this problem. An alternative is to ally the ornithocheirids with Suntunadactylus (Fig. 12). The cervical vertebrae of this genus are very similar to those of Pterodactylus antiquus and Pterodactylus longicollum. The main difference lies in the possession of postexapophysial processes by Suntunadactylus. Its cervical vertebrae are also closest (of the long-necked pterodactyloids) in overall morphology to the tallspined forms. If the latter position for the ornithocheirids is correct, then the postexapophysial processes evolved once only, but an evolutionary reversal,

21 PTERODACTYLOID CERVICAL VERTEBRAE 327 Figure 12. Alternative, equally parsimonious cladogram of pterodactyloid relationships based on the cervical vertebrae. Key to numbered nodes: I, 2, 3, 4, 5 and 10-see Fig. 11; 6, elongated middle-series cervicals with low neural spine; 7, presence of postexapophyses; 8, completely fused atlanto-axis; 9, presence of pre-exapophyses on post-axial cervicals. namely the shortening of the cervical vertebrae, is implied. If the ornithocheirids did evolve from a Jurassic long-necked form, then the foreshortening of the cervical vertebrae would necessitate an increase in the height of the neural spines (to give ample area for the origin of the neck musculature). Whatever the evolutionary status of the ornithocheirids, it would appear (from what is known) that by the middle Cretaceous the pterodactyloids had diverged into two groups composed mainly of large animals, and readily identifiable by the nature of their cervical vertebrae. The two groups would appear to have had very long separate histories extending back to the late Jurassic. ACKNOWLEDGEMENTS I have pleasure in recording my gratitude to Dr Angela Milner, British Museum (Natural History), London, Mr John Cooper, Booth Museum of Natural History, Brighton, Miss Barbara Pyrah, Yorkshire Museum, York, and Mrs F. Etchells-Butler, Sedgwick Museum, Department of Earth Sciences, University of Cambridge, for their kind assistance and the loan of material discussed in this paper. Dr Andrew Milner, Department of Biology, Birkbeck College, University of London, is especially thanked for his guidance and