This study is based on the right forelimb of Chasmosaurus irvinensis (CMN 41357). Half-scale models of the elements of the girdle and limb were sculpted in modeling clay and plastic foam, using proportional dividers to maintain accuracy. Information from the left scapulocoracoid, as well as from the undistorted scapulocoracoid of CMN 344 (Styracosaurus albertensis), was used as a reference during the reconstruction of the scapulocoracoid. Poly 74-30 RTV liquid silicon rubber moulds were taken from these models, and casts were made in thixotropic polyester resin. Thin, flexible wires were imbedded in each replica so that they could be held in place during the simulation of the step cycle. The rib cage, although somewhat distorted, is largely complete. Comparison with the less distorted rib cage of the type of Styracosaurus albertensis (CMN 344) indicates that, in both curvature and nature of their articulations to their respective vertebrae, the rib at each segment of the cervical and thoracic column is closely comparable in the two taxa. It was therefore possible, using CMN 344 as a reference, to reconstruct the shapes and orientations of the ninth (last) cervical, second, fifth, seventh, and ninth thoracic ribs with stiff copper wire. Another wire was bent to match the curvature of the presacral column of CMN 344, and the rib replicas were fixed to it. In this way, the approximate size and shape of the rib cage were reconstructed. This assembly was supported on two wooden dowels that were cut to lengths based on the assumption that an individual of Chasmosaurus would have stood approximately 1.5 metres at the hip (Dodson 1996). The ends of the dowels were inserted into metal brackets that allowed the entire assembly to stand upright, unsupported. The medial edge of the coracoid was placed close to the midline, and after considerable experimentation to find the best fit between the rib cage and scapula, the scapulocoracoid was wired into place.

Before the rest of the model was assembled, the various subunits (e.g., the ulna-radius and humerus) were articulated, and potential range of motion was investigated. This permitted us to position the elements as we assembled an initial pose.

The metacarpus and manus were assembled and wired together as a unit. In order to estimate the orientation of the digits, as well as the distance of the manus from the midline, published neoceratopsid trackway data (Lockley and Hunt 1995) were used. The trackway, probably made by the much larger (but similar with respect to pectoral girdle and limb anatomy) neoceratopsid Triceratops, was scaled down until the assembled manus could be fit inside the boundaries of the manus print. The scaled print was stencilled onto a base. The midline of the trackway, defined as the midpoint between the outer edges of the manus prints, was drawn in, and the rib cage/scapulocoracoid (hereafter referred to as torso) unit was placed so that the sagittal plane was directly over the midline of the track way.

The orientation of the manus was determined by aligning the digits with the individual toe prints preserved in the trackway. With the manus in place, the humerus, ulna, and radius were attached. The torso was moved "anteroposteriorly" along the midline of the track way until all bones attained a "best fit" with the limb in approximately mid-stance. From this reference point, the torso was advanced along the midline as the model was moved through a simulation of a step cycle. At all points, the orientations of all elements were dictated by the following parameters: 1) the digits of the manus were maintained in alignment with the toe marks of the manus print throughout the propulsive phase of the cycle; 2) articulations were maintained at all joints; and 3) the torso did not deviate from the midline (we are not suggesting that this could not have occurred, but simply that, since there is no way to estimate the extent of lateral movement that may have occurred, this potential variable should be held constant for the purposes of this simulation). Each point in the simulated step cycle was photographed in lateral, anterior , and dorsal views. The angular relations of the humerus, ulna, and radius to the parasagittal and horizontal (frontal) planes, as well as the angle formed by the epipodials on the humerus were taken directly from the images (Table 1). For the purpose of these measurements, the long axes of the bones were drawn through the centres of their shafts, in the humerus, extending from the centre of the trochlear notch distally to the lateral edge of the glenoid condyle, and in the ulna, passing through the centre of the semilunar notch and extending distally to bisect the distal expansion (Table 1).