The origin of the Ornithodira (Pterosauria+Dinosauria, see Benton 2004) is associated with a fundamental reorganization of the limbs. The sprawling stance characteristic of primitive tetrapods was abandoned as the limbs were rotated under the body to permit the development of a parasagittal gait. This reorganization was a necessary precondition for the appearance of bipedalism, which evolved in many early archosaurs, including the first dinosaurs. Although many later dinosaurs, such as sauropods, stegosaurs, ankylosaurs, and neoceratopsids reverted to a quadrupedal stance, none returned to the primitive sprawling posture, but rather evolved a stance roughly comparable to that seen in large mammalian quadrupeds.
In the first skeletal reconstruction of a neoceratopsid (Marsh 1891), the limbs of Triceratops were depicted in a more or less vertical orientation that resembled the condition in modern graviportal quadrupedal mammals. However, when skeletons were mounted for display, the front limbs were arranged in a sprawling posture, with the long axis of the humerus approximately horizontal, and projecting at a large angle from the parasagittal plane (e.g., Gilmore 1905; Sternberg 1927; Lull 1933; Osborn 1933; Erickson 1966). These reconstructions went unchallenged until the late 1960s, when Robert Bakker launched an extended campaign to revise our notion of dinosaur biology (e.g., Bakker 1968; 1975; 1986; 1987). He argued against the traditional concept of dinosaurs as essentially biologically larger versions of living reptiles, and in favour of the view of dinosaurs as active, endothermic animals. One of the consequences of this paradigm shift was a revision of our view of limb posture in neoceratopsids. The occurrence of sprawling forelimbs in an endothermic, cursorial animal was so incongruous that the possibility was rejected, and various versions of upright posture were proposed (e.g., Alexander 1991; Bakker 1986; Dodson and Farlow 1997; Ford 1997; Garstka and Burnham 1997; Paul and Christiansen 2000). As new neoceratopsid mounts were constructed and old ones refurbished, they were generally assembled with vertical, columnar forelimbs.
These views were not universally held, however. As noted by the early neoceratopsid workers, the structure of the neoceratopsid forelimb did not permit upright posture unless it was forced to do so (Gilmore 1905; Sternberg 1927; Erickson 1966). In an analysis of the biomechanics of an articulated forelimb of Torosaurus, Johnson and Ostrom (1995) noted that the glenoid condyle of the humerus is located on the external, rather than proximal, surface of the proximal expansion of the humerus as is the case in most other dinosaurs. They also observed that the glenoid of the scapulocoracoid faces more or less posteriorly rather than ventrally. When the proximal articular condyle of the humerus is placed into the glenoid, the humerus automatically assumed a horizontal attitude. If the humerus is rotated into a vertical position, any attempt to maintain the humeral head in the glenoid forces a rotation of the humerus so that the deltopectoral crest projects anterolaterally, making effective functioning of the pectoralis musculature problematic. Although the rib cage was not included in the model, it was argued that the medial side of the broad proximal humeral expansion (the lesser tubercle) of a parasagittal humerus would have projected into the rib cage. In addition, they pointed out that the articulations for the ulna and radius are "offset," causing the plane of elbow extension and flexion to deviate significantly from the long axis of the humerus, rendering a parasagittal gait in neoceratopsids unfeasible. Johnson and Ostrom (1995) concluded that the humerus of neoceratopsids must have projected horizontally from the trunk at close to right angles to the sagittal plane, with the elbow remaining strongly bent throughout the step cycle. Forward progression occurred as the humerus rotated about its long axis, operating rather like the axis of a wheel, with the epipodials representing one spoke in that wheel, transferring the locomotor force to the foot which functioned as a portion of the rim of the "wheel."
Resolving these diametrically opposed interpretations of neoceratopsid forelimb stance is difficult for a number of reasons. Articulated, complete specimens are rare, and most mounts are composites. The bones are large and fragile, making manipulation problematic. Johnson and Ostrom (1995) had the advantage of an articulated, well- preserved specimen, of which they could make casts. These were used to experiment with various limb positions. However, the scope of their conclusions was limited by the absence of the rib cage, carpus, metacarpus, and manus in this specimen.
In one recent analysis, Paul and Christiansen (2000) argue that the forelimb of neoceratopsids did not conform to either the sprawling or fully erect, essentially graviportal stereotype. Rather, the humerus was strongly retracted posteriorly, and was capable of swinging through a 70 degree arc that took it from approximately horizontal to quasi-vertical in orientation. The traditionally postulated anatomical impediments to parasagittal gait were countered with plausible arguments based on the authors’ observations of neoceratopsid osteology. In particular, it was argued that incorrect orientations of ribs and vertebral attachments have resulted in an inaccurate placement of the scapulocoracoid (and therefore glenoid orientation), and that insufficient use has been made of trackway data. A number of informative figures were produced, but it was not possible at the time to test their hypothesis with a working model.
In 1958, Dr. Wann Langston collected a specimen of Chasmosaurus near the town of Irvine, Alberta. The skull was later prepared, and a new species, Chasmosaurus irvinensis, was erected to accommodate the specimen (see Holmes et al. 2001 for details). The better part of the postcranial skeleton was preserved, including a largely complete but somewhat distorted rib cage, a complete right front scapulocoracoid, limb, and articulated manus. Anatomical data sets of comparable completeness and quality are rare, but not unknown in ceratopsid neoceratopsians (e.g., Brown 1917; Brown and Schlaikjer 1937; Lull 1933). However, these specimens have been rendered as display pieces, either as a panel mount or otherwise mounted so as to make manipulation of individual elements impossible. The specimen described here has been completely freed from the matrix, permitting access to all aspects of the preserved anatomy. This has provided a unique opportunity to fabricate and assemble an accurate, half-scale model of the girdle and limb to test current hypotheses of neoceratopsian forelimb stance.