Scelidosaurus harrisonii from the Early Jurassic of Dorset, England: postcranial skeleton

Abstract Scelidosaurus fossils were first discovered during the commercial quarrying of the Liassic sea-cliffs between Charmouth and Lyme Regis in Dorset during the late 1850s. The original specimens included a well-preserved skull embedded in a block of argillaceous limestone (marlstone). Shortly after this skull was retrieved, a series of more-or-less contiguous marlstone slabs were recovered, containing most of the skeleton of the same animal (NHMUK R1111). After rudimentary (hammer and chisel) mechanical preparation, Owen published descriptions of this material (Owen, 1861, 1863). These two monographs have been the sole references pertaining to the anatomy of Scelidosaurus for >150 years. The skeleton of the lectotype of Scelidosaurus harrisonii (NHMUK R1111) has since been extracted from the surrounding matrix using an acid-immersion technique. Some additional specimens held in the collections of the Natural History Museum London, the Bristol City Museum and the Sedgwick Museum in Cambridge provide anatomical material that allows detailed description of this taxon, for which we have had, until now, a surprisingly poor understanding. Axial skeleton: The axial skeleton of Scelidosaurus comprises eight cervical, 16 dorsal, four sacral and > 40 caudal vertebrae. During ontogeny, the posterior centrum articular surface of the 16th dorsal vertebra develops a firm, ligament-bonded junction with the succeeding sacral centrum. Apart from the atlas rib, which is single headed, double-headed ribs are present throughout the presacral vertebral series, and none shows any indication of fusion to its associated vertebra. However, those ribs attached to cervical vertebrae 2–4 were evidently bound firmly by connective tissue to rugose diapophyses. The last two (presacral) dorsal ribs show merger of the capitulum and tuberculum, meaning that they are separated by only a step. The angulation and arching of the dorsal ribs suggest that these animals had a broad (barrel-like) torso. Intercostal uncinate plates were present, attached to the posterior margins of some of the largest dorsal ribs. Their attachment sites are clearly marked, and these plates might have been composed of calcified cartilage in larger individuals. The sacral vertebrae fuse progressively during ontogeny, in an anterior-to-posterior sequence. The sacral ribs are long and robust, and tilt the iliac blade outward dorsally. A sacricostal ‘yoke’ (created by the fusion of the distal ends of adjacent sacral ribs) never forms. The base of the tail has a unique ball-and-socket-style joint between the centra of caudal vertebrae 1 and 2 in only one skeleton. This might have permitted powerful, but controlled, movements of the tail as a defensive weapon (or increased flexibility at the base of the tail, which might have been necessary for reproduction). Caudal ribs are initially long, blade-shaped projections that gradually decrease in size and become stub-like remnants that persist as far back as the midtail (approximately caudal vertebra 25). Haemal arches (chevrons) disappear nearer to the distal end of the tail (approximately caudal vertebra 35). Ossified tendons are preserved as epaxial bundles that are clustered in the ‘axillary’ trough (between the neural spine and transverse processes on either side of the midline). Ossified tendons are restricted to the dorsal and sacral region. Flattened ossified tendons are fused to the sides of sacral neural spines. In life, the ossified tendons might have formed a low-angled trellis-like arrangement. Appendicular skeleton: The pectoral girdle comprises a long scapula, with a distally expanded blade. The proximal portion is expanded and supports an oblique promontory, forming an acromial process anteriorly and a thick, collar-like structure posteriorly above the glenoid. Between these two features is a shallow basin, bordered ventrally by a sutural edge for the coracoid. The scapula–coracoid suture remains unfused in large (5-m-long) individuals. The coracoid bears a discrete foramen and forms a subcircular dished plate, with the shallowest of embayments along its posterior edge. Clavicles are present as small fusiform bones attached to the acromial process of the scapulae and leading edge of each coracoid. A sternum was reported as ‘some partially ossified element of the endoskeleton’ Owen (1863: 13), but subsequent preparation of the skeleton has removed all trace of this material. The humerus is relatively long and has a prominent rectangular and proximally positioned deltopectoral crest. The ulna is robust and tapers distally, but there is no evidence of an olecranon process. The radius is more rod-like and terminates distally in an enlarged, subcircular and convex articular surface for the carpus. The carpus is represented by an array of five discoid carpals. The manus is pentadactyl and asymmetrical, with short, divergent metacarpals and digits that terminate in small, arched and pointed unguals on digits 1–3 (only). The phalangeal formula of the manus is 2-3-4-3-2. The pelvis is dominated by a long ilium; the preacetabular process is arched, transversely broad, and curves laterally. In juveniles, this process is short and horizontal, but during ontogeny it increases considerably in length and becomes arched. The iliac blade is tilted laterally, meaning that its dorsal blade partly overhangs the femur. The acetabulum forms a partial cupola, and there is a curtain-like medial wall that reduces the acetabular fenestra to a comparatively low, triangular opening between the pubis and ischium. The postacetabular portion of the ilium is long and supports a brevis shelf. The ischium has a long, laterally compressed shaft that hangs almost vertically beneath the ilium, and there is no obturator process. The pubis has a long, narrow shaft and a relatively short, deep, laterally compressed prepubic process that twists laterally (its distal end lies almost perpendicular to the long axis of the ilium). The articular pad on the pubis for the femoral head faces posteriorly. The obturator foramen is not fully enclosed within the pubis, but its foramen is closed off posteriorly by the pubic peduncle of the ischium. The femur is stout and has a slightly medially offset femoral head, and the greater trochanter forms a sloping shoulder continuous with, and lateral to, the femoral head. The anterior (lesser) trochanter is prominent and forms a thick, thumb-shaped projection on the anterolateral corner of the femoral shaft. The fourth trochanter is pendent and positioned at midshaft. In larger individuals, it appears to become thickened and reinforced by becoming coated with metaplastic bone derived from the tendons attached to its surface. The distal end of the femoral shaft is slightly curved and expands to form condyles. There is a deep and broad posterior intercondylar groove, but the anterior intercondylar groove is barely discernible in juveniles and not much better developed in subadults. The tibia and fibula are shorter than the femur. The tibia is structurally dominant, and the shorter fibula is comparatively slender and bowed. The proximal tarsals are firmly bound by connective tissue to the distal ends of the tibia and fibula. The distal end of the tibia is stepped, which aids the firm interlock between the crus and proximal tarsals. There appear to be two roughly discoid tarsals (distal tarsals 3 and 4), and a rudiment of distal tarsal 5 appears to be sutured to the lateral margin of distal tarsal 4. Five metatarsals are preserved, but the fifth is a splint of bone attached to the proximal end of metatarsal 4. Metatarsals 2–4 are dominant, long and are syndesmotically interlocked proximally, but their shafts splay apart distally. Metatarsal 1 is much shorter than the other three, but it retains two functional phalanges (including a short, pointed ungual). The foot is anatomically tetradactyl but functionally tridactyl. The pedal digit formula is 2-3-4-5-0. The digits diverge, but each appears to curve medially along its length, creating the impression of asymmetry. This asymmetry is emphasized, because the three principal unguals are also twisted medially. The ungual of digit 2 is the largest and most robust of the three, whereas that of digit 4 is the smallest and least robust. The general girth of the torso and the displacement of the abdomen posteriorly (a consequence of the opisthopubic pelvic construction in this dinosaur) constrained the excursion of the hindlimb during the protraction phase of the locomotor cycle. The anterolateral displacement of the hindlimb during protraction is in accord with the freedom of motion that is evident at the acetabulum, the susceptibility of the hindlimb to torsion between and within its component parts, and the asymmetry of the foot. It is probable that thyreophorans (notably, ankylosaurs) used a similar oblique-parasagittal hindlimb excursion to accommodate their equally large and wide abdomens. This surmise accords with the structure of the pelves and hindlimbs of ankylosaurs. Derived stegosaurs might have obviated this ‘problem’, in part, because their hindlimbs were longer and their torsos and abdomens narrower and capable of being ‘stretched’ vertically to a greater extent. Nevertheless, the structure of their acetabula and hindlimbs indicates that the oblique-parasagittal style of hindlimb excursion remained a possibility and might be an evolutionary remnant of the locomotor style of basal, shorter-limbed stegosaurs. A reconstruction of the endoskeleton of Scelidosaurus is presented on the basis of this updated description. Although quadrupedal, this animal was only facultatively so, judged by its forelimb-to-hindlimb proportions and structure; it therefore betrays bipedality in its ancestry.PRISM FUND TRINITY COLLEGE CAMBRIDGE ST JOHN'S COLLEGE OXFORD SMITHSONIAN INSTITUTION (WASHINGTON DC