Zoological Journal of the Linnean Society, 2008, 154, 27–69. With 30 figures Three-dimensional geometry of a pterosaur wing skeleton, and its implications for aerial and terrestrial locomotion MATTHEW T. WILKINSON* Department of Zoology, University of Cambridge, Cambridge CB2 3EJ, UK Received 21 September 2007; accepted for publication 17 September 2007 This study reports on the three-dimensional spatial arrangement and movements of the skeleton of Anhanguera santanae (Pterodactyloidea: Ornithocheiridae), determined using exceptionally well-preserved uncrushed fossil material, and a rigid-body method for analysing the joints of extinct animals. The geometric results of this analysis suggest that the ornithocheirids were inherently unstable in pitch and yaw. As a result, pitch control would probably have been brought about by direct adjustment of the angle of attack of the wing, by raising or lowering the trailing edge from the root using the legs if, as is indicated in soft-tissue specimens of a number of unrelated pterosaur species, the legs were attached to the main wing membrane, or by using long-axis rotations at the shoulder or wrist to raise and lower the trailing edge from the wingtip. An analysis of the three-dimensional morphology of the wrist lends support to the idea that the pteroid – a long, slender wrist bone unique to pterosaurs that supported a membranous forewing – was directed forwards in flight, not towards the body. As a result, the forewing could have fulfilled the function of an air-brake and high-lift device, and may also have had an important role in pitch, yaw, and roll control. The joint analysis is consistent with a semi-erect quadrupedal model of terrestrial locomotion in the ornithocheirids. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69. ADDITIONAL KEYWORDS: Aerodynamics – arthrology – biomechanics – flight – functional morphology – morphometrics – Ornithocheiridae – pteroid – Pterosauria. INTRODUCTION It is now generally accepted that pterosaurs – the winged reptiles of the Mesozoic Era – were fully capable flyers (Padian, 1983; Padian & Rayner, 1993), although the larger species were almost certainly secondarily adapted for soaring (Hankin & Watson, 1914; Bramwell & Whitfield, 1974; Brower, 1983; Padian & Rayner, 1993). Nevertheless, controversy and uncertainty still surround many aspects of pterosaur locomotion, including their terrestrial ability (Bennett, 1997a, b; Unwin, 1997; Henderson & Unwin, 1999; Unwin & Henderson, 1999; Bennett, 2001; Chatterjee & Templin, 2004), the precise shape, aerodynamic behaviour, and possible control movements of their wings, and how the giant forms – the *E-mail: [email protected] largest flying animals that have ever lived – took off and landed (Alexander, 1998). In order to address these issues, an understanding of the functional morphology of the joints is vital in providing information about the range of movement and degrees of freedom of the limbs, and important flight parameters such as wing planform, span, and area. Unfortunately, because most pterosaurs had extremely thin-walled bones (Heptonstall, 1971; Bramwell & Whitfield, 1974; Wellnhofer, 1985, 1991a; Fastnacht, 2005), preserved skeletal remains are nearly always found in a crushed, flattened state. Clearly, such fossils cannot be used to reconstruct joint function or skeletal geometry with any real degree of accuracy: if bones are to be correctly articulated, they must be in their original three-dimensional condition. On a few rare occasions, however, pterosaur fossils in this exceptional state of preservation have been found. Notable among these are the thousands of pterosaur bone fragments © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 27 28 M. T. WILKINSON from the Cambridge Greensand (Lower Cretaceous) (Seeley, 1870; Hooley, 1914). These specimens were used as the basis of functional morphological studies by Hankin & Watson (1914) and Bramwell & Whitfield (1974), who manipulated incomplete articular ends of limb bones to determine the range of movement of the various joints of the wings. Unfortunately, owing to considerable post-mortem disturbance of the Cambridge Greensand material, its fragmentary nature, and the large number of individuals represented in the collection, it is almost certain that these attempts to directly articulate the joints used fragments from different individuals, probably of different size, and even of different species in some cases. In addition, because the shafts of the limb bones are missing, it is difficult to establish exactly how the orientation of the proximal end of a bone relates to that of its distal end, with the result that the three-dimensional spatial arrangement and range of movement of the whole wing skeleton cannot be determined with certainty. This limiting situation changed with the discovery of three-dimensionally preserved fossilized limb bones of single pterosaur individuals in calcareous nodules of the Santana Formation (Aptian–Albian) in the Chapada do Araripe, Ceará, Brazil (Price, 1971; de Buisonjé, 1980; Wellnhofer, 1985, 1991b; Kellner & Tomida, 2000; Veldmeijer, 2002, 2003). Most of these specimens have been assigned to the Ornithocheiridae sensu Unwin (2003), one of the dominant pterosaur groups of the Early Cretaceous (Wellnhofer, 1991a). Additionally, three-dimensionally preserved specimens of the Late Cretaceous azhdarchid pterosaur Quetzalcoatlus have been discovered in the Javelina Formation, Big Bend National Park, West Texas (Lawson, 1975; Padian, 1984). Wellnhofer (1985, 1991b), Hazlehurst & Rayner (1992), and, to a limited extent, Bennett (2001) used ornithocheirid fossils from the Santana Formation in their studies of the pterosaur wing skeleton, but manipulated, for the most part, incomplete bones when articulating joints. Thus, although the individual joints considered by these authors could be reconstructed with a far greater degree of confidence than was possible with the Cambridge Greensand material, the spatial arrangement and degrees of freedom of the complete wing could still not be accurately determined. In the last few years, however, a small number of nearcomplete skeletons have been discovered, offering an unprecedented opportunity for a comprehensive analysis of the functional morphology of the ornithocheirid pterosaurs. Chatterjee & Templin (2004) used one such skeleton in their study of pterosaur posture and locomotion, but went no further than previous authors, treating most of the joints in isolation. This study reports a thorough investigation of the range of movement, degrees of freedom, and spatial arrangement of the ornithocheirid wing skeleton, using the most complete three-dimensional specimens from the Santana Formation, and discusses the implications of this geometric information for aerial and terrestrial locomotion in this group. It is hoped that the techniques developed here will also provide a methodological basis for future studies of the joints of extinct animals. MATERIAL AND METHODS FOSSIL MATERIAL Ten three-dimensionally preserved ornithocheirid skeletons from the Santana Formation were used in the present analysis (for a list of the institutional abbreviations, see Appendix 1): Anhanguera santanae (AMNH 22555) (Wellnhofer, 1991b); Anhanguera sp. (IMCF 1053, SMNK 1136PAL); Brasileodactylus sp. (AMNH 24444); Coloborhynchus robustus (NSM-PV 19892) (Kellner & Tomida, 2000), originally classified as Anhanguera piscator, and reassigned by Unwin (2003), although two authors have implicitly rejected this reassignment (Codorniú & Chiappe, 2004; Kellner, 2004); C. robustus (SMNK 1133); Ornithocheiridae indet. (SMNK 1134PAL, SMNK 1135PAL); Santanadactylus pricei (AMNH 22552) (Wellnhofer, 1991b); and ?Santanadactylus sp. (SMNK 1250PAL). The wing skeletons of four of these specimens: Anhanguera sp. (SMNK 1136PAL), C. robustus (NSM-PV 19892, SMNK 1133) and S. pricei (AMNH 22552) are nearly complete, and provided the bulk of the morphological and arthrological information. Additional morphometric data were obtained from the crushed, near-complete articulated skeleton of the putative ornithocheirid Arthurdactylus conandoylei (SMNK 1132PAL) from the Crato Formation (Lower Cretaceous) of Brazil (Frey & Martill, 1994), and from the published description of the Santana ornithocheirid Coloborhynchus spielbergi (RGM 401 880) (Veldmeijer, 2003). MORPHOMETRIC ANALYSIS The specimens listed above include the most complete three-dimensionally preserved pterosaur skeletons yet discovered. Nevertheless, no single complete specimen exists. The available material was therefore used to determine the dimensions of a composite ornithocheirid, scaled to the size of A. santanae (AMNH 22555). This is the only specimen examined in which the vertebral column is preserved in its entirety. Hence, the dimensions of the axial skeleton can be determined by direct measurement. However, no complete long bones are preserved, so the dimen- © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE sions of the limb elements were estimated from measurements of the remaining specimens. Bones were measured in accordance with Bennett’s (2001) convention: the tubercula of the distal humerus, distal ulna, and proximal wing metacarpal were not included, and neither was the extensor tendon process of the first phalanx of the wing finger. The only element for which it was not possible to obtain an exact length measurement was the pteroid: where present, the bone was always broken, and the tip lost. The pteroid of C. robustus (SMNK 1133PAL) was the most complete of those examined, so its length was estimated on the assumption that only the very tip had broken off. With the exception of A. santanae (AMNH 22555), the ulna is preserved in all of the specimens examined. Log (bone length) was plotted against log (ulna length) to obtain linear scatter plots for each long bone, and a major axis regression analysis was carried out to determine their allometric relationships. An isometric relationship was assumed for any bone represented by only one specimen. The scapulae are complete in A. santanae (AMNH 22555), so the relative dimensions of the limb bones, expressed in terms of ulna length, were converted into expressions in terms of scapula length, and the skeletal dimensions of the composite A. santanae were estimated accordingly. JOINT ANALYSIS Three specimens were used in the joint analysis: C. robustus (NSM-PV 19892, SMNK 1133PAL) and S. pricei (AMNH 22552). These are largely free of any matrix and are the most complete three-dimensional skeletons that can be directly articulated. All the major joints of the wing skeleton, including those of the legs, are represented among them. When articulating fossil bones it is important to allow for an appropriate thickness of cartilage between them, for without this allowance the range of movement at a joint will be substantially underestimated. As a rule, the bone surfaces of articulating elements are not perfectly matched: typically, the radius of curvature of convex surfaces is less than that of the corresponding concave surfaces. However, the cartilage is usually moulded to the underlying bone in such a way that the true joint surfaces are congruent in at least one position – the so-called ‘close-packed’ position, which for hinge joints usually occurs at one end of the habitual articular movement, e.g. at the full extension of the human knee (Williams et al., 1989). Hence, in most cases, the thickness of cartilage should be equal to the difference between the radii of curvature of the corresponding bone surfaces when in the close-packed position. This difference can therefore be used to estimate the extent of 29 cartilage in fossil joints. For instance, a very close fit of the fossilized surfaces indicates that the layers of cartilage would have been very thin; conversely, if the curvatures of corresponding articular surfaces are very dissimilar, one can postulate that there may have been a greater thickness of cartilage between them in life. This is not always the case: in some joints, such as the knee joint of many large mammals, the curvatures of the articular surfaces differ radically, and an unfeasibly thick layer of cartilage would be required in order to make them congruent. Needless to say, such a thickness of cartilage is not actually observed, and full congruence is not attained at any stage of articular motion. Fortunately, the radii of curvature of the corresponding surfaces of all joints of the ornithocheirid wing skeleton are similar enough to permit the estimation of cartilage thickness by the method just described with a fair degree of confidence. The constraining actions of other soft tissues, such as the fibrous joint capsule, accessory ligaments, and muscles, are more difficult to ascertain in fossils than the thickness of cartilage. One must always bear in mind that the range of movement deduced from articular geometry alone is the maximum possible range: the range of movement in the living animal may have been more restricted. The task is made easier if there are bony stops that define one end of an articular excursion, such as the olecranon process of the ulna in the human elbow joint. In addition, joint surfaces occasionally show signs of pathological grooving associated with osteoarthritis (Bennett, 2001, 2003). The curvature and arc lengths of such grooves give a direct measure of the extent of articular movement that occurred, although these indicate pathological movements that may not be representative of normal joint function (Padian, 1984). The joint analysis procedure was as follows. To begin, each joint was classified by direct manipulation as immobile, uniaxial (with a single axis, e.g. hinge joints), biaxial (with two mutually perpendicular axes, e.g. saddle joints), or multiaxial (with three orthogonal axes, e.g. ball-and-socket joints). Long-axis rotations were additionally classified as either adjunct, i.e. independent of angulation (swinging motion), or conjunct, i.e. indissociable from angulation. Joint motion was then analysed further to identify the positions and orientations of the various joint axes. In clinical biomechanics, the analysis of articular motion classically involves placing markers on the fixed and moving limb elements, taking photographs or radiographs before and after angulation, and geometrically analysing the marker positions so obtained to find the axes of rotation and the range of movement around these axes. The analysis can either be two-dimensional, in which case a single camera is used (Frankel, Burstein & Brooks, 1971; Walker, Shoji & Erkman, 1972; Smidt, © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 30 M. T. WILKINSON 1973), or three-dimensional, in which case at least two cameras are used to give stereoscopic images (Blankevoort, Huiskes & de Lange, 1988; Boyd & Ronsky, 1998; Leardini et al., 1999). For the present study, a two-dimensional method was used. The procedure is illustrated using a simple hypothetical hinge joint, with a single degree of freedom and no conjunct rotation (Fig. 1). The proximal element of the joint was fixed and the distal element remained mobile. The elements were set up with respect to a single camera such that the joint axis was perpendicular to the photographic film. In the plane of the film, the axis was therefore manifest as a point – the centre of rotation (CR). A single marker – O – was placed on the proximal element to define the origin of an arbitrary three-dimensional Cartesian coordinate system. In Figure 1, the joint set-up is viewed in the y–z plane, and the joint axis lies parallel to the x-axis. At least two further markers – A and B – were placed on the distal element. The joint set-up was then photographed at least twice: once with the distal element at maximum extension (usually corresponding to the close-packed position), and a second time with the distal element at maximum flexion. z A number of methods exist for finding the CR from these marker coordinates. The method used here is the rigid-body method developed by Spiegelman & Woo (1987). This method is more accurate than older, graphical methods, whilst having fewer constraints in the placement of the markers (Panjabi, 1979; Panjabi, Goel & Walter, 1982). It uses rigid-body equations to find a clockwise rotation matrix M describing motion around the CR, and a translation vector c defining the position of the CR relative to the arbitrary origin located at marker O. The two markers on the moving bone – A and B – are denoted A′ and B′ after rotation of magnitude q (Fig. 1). Points A, A′, B, and B′ are defined by the position vectors p1, p2, p3, and p4, respectively: p1 = y1 j + z1 k, (1) p2 = y2 j + z2 k, (2) p3 = y3 j + z3 k, (3) p4 = y4 j + z4 k, (4) where j and k are unit vectors along the y- and z-axes, respectively. The angle of rotation q is given as follows: s′ s + t ′t q = cos −1⎛ 2 2 ⎞ , ⎝ s +t ⎠ A' (5) where θ B' B O c A CR s = z1 − z3, (6) s′ = z2 − z4, (7) t = y1 − y3, (8) t ′ = y2 − y4. (9) The position of the CR is given by: y Cz = z1 + Cy = y1 − Figure 1. A pair of articulating elements with a single joint axis, set up so that the axis runs perpendicular to the viewing plane, and is perceived as a point – the centre of rotation (CR). The proximal element is fixed and has a single marker O, defining the origin of an arbitrary coordinate system. The distal element is mobile and bears two markers A and B, denoted A′ and B′ after a rotation of magnitude q. The positions of the markers before and after angulation in the arbitrary coordinate system can be used to determine the position vector c of the CR, and the magnitude of rotation q. See text for details. ( y2 − u) sin q ( z2 − v) sin q − cos q ( y1 − u) , sin q (10) + cos q ( z1 − v) , sin q (11) where u= sin q ( z1 − z2 ) y1 + y2 , + 2 (1 − cos q ) 2 (12) v= z1 + z2 sin q ( y1 − y2 ) − . 2 2 (1 − cos q ) (13) This process requires a minimum of two markers (A and B) to compute the angle of rotation q and the © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE position of the CR, but more were used in practice, and mean values of the parameters were calculated to improve the accuracy. For most uniaxial joints, the rigid-body method was used in an unmodified form. The method was appropriate, even if the distal element underwent a conjunct rotation, provided the moving bone underwent a cardinal angulation, i.e. remained within a single plane. This is not always the case: conjunct rotation is often accompanied by rotation of the joint axis with respect to the proximal element, which causes the distal element to swing out of a single plane (Fig. 2). Such angulations are termed arcuate. The analysis of arcuate motion is described below. The analysis of biaxial joints was similar to that of uniaxial joints: each component axis was simply treated separately. In such joints, angulation occurs around two mutually perpendicular axes, primary and secondary, such that the maximum range of movement is not simply an angle, but a virtual cone. The shape of this cone is defined by the maximum ranges of angulation around the two axes. Here, the primary axis was defined as that around which the greatest range of angulation took place. The secondary axis was perpendicular to the primary axis. The CRs of these two axes and the ranges of angulation around them were found as described above. The long-axis rotation of a bone, whether conjunct or adjunct, was also measured using the rigid-body method. In this case, the bone in question was photographed in distal view, and markers were placed on the distal surface. Because the static proximal element was generally not in the field of view in this 31 case, another static element was placed near the distal surface of the rotating bone, and a marker was fixed to it in order to define the arbitrary origin, as before. Multiaxial joints were analysed using the methods employed for biaxial joints and long-axis rotations. As described above, arcuate swings occur if a joint axis itself rotates during the angulation of a bone. Arcuate motion was analysed by fixing the hinge axis, such that the distal element underwent a normal cardinal angulation. During angulation of the distal element, the proximal element was therefore rotated in order to maintain articular contact. This rotation – equal and opposite to that of the hinge axis, with respect to the proximal element – was measured using the rigid-body method by marking and photographing the proximal element in the usual way. Immobile joints were set up so that both elements were in the same plane, perpendicular to that of the photographic film, as before. The joint was then photographed in its life position and in a reference position, with proximal and distal elements in line. The CR and rotation angle q between this reference position and the life position were then calculated using the rigid-body calculations. Essentially, immobile joints were treated as if they were uniaxial, the life and reference positions corresponding to the limits of articular movement of a uniaxial joint. The various joint axes, found in isolation, were then placed in a single, common frame of reference, this being a three-dimensional Cartesian coordinate system, with its origin in the plane of symmetry of Figure 2. Diagrammatic representation of an articulating element undergoing a cardinal angulation (A) and an arcuate angulation (B). During a cardinal angulation, the joint axis remains fixed and the moving bone remains in a single plane. During an arcuate angulation, the joint axis rotates and the bone moves out of the plane. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 32 M. T. WILKINSON A wf pt ul ra hu co sc wm tt il fe B y x C z y Figure 3. Reconstruction of the axial skeleton and left wing of Anhanguera santanae in dorsal view (A), and a virtual fleshed-out reconstruction of the axial skeleton and right wing in ventral (B), and anterior (C) views. Principal limb bones are laid out end-to-end. For a list of anatomical/arthrological abbreviations, see Appendix 1. Scale bar: 500 mm. the pterosaur body. In order to visualize the threedimensional wing, a computer model was constructed to represent the reconstructed, fleshed-out skeleton of the composite A. santanae (AMNH 22555). Model dimensions were obtained from the morphometric analysis described above. The virtual limb elements were initially laid out end to end, along straight lines projecting from the shoulder and hip joints (Fig. 3). The x-axis passed through the plane of symmetry fore to aft, subparallel to the vertebral column; the y-axis ran left to right; and the z-axis ran ventral to dorsal. The procedure for placing and orienting the joints axes made use of the facts that: (1) in the analyses of the isolated joints, the joint axes were always oriented perpendicular to the photographic film, and (2) two consecutive joints along the arm or leg share a common skeletal element. This shared element generally had to be adjusted between joint set-ups in order to align the subsequent joint axis with the horizontal. The relative orientation of consecutive joint axes was found by quantifying the magnitude of this angular adjustment. By carrying out this procedure for each consecutive pair of joints, moving from root to tip, the orientation of all the joint axes was found relative to the body axis system. The joint axes were then mapped onto the limbs in locations corresponding to their respective CRs. Rotation matrices were used to rotate each limb element around the relevant CR within their respective ranges of angulation, working from tip to root. Moving the elements within their calculated ranges could generate any possible limb configuration. SKELETAL PROPORTIONS Bone dimensions are given in Appendix 2. The linear scatter plots of log (bone length) plotted against log © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE B C 2.50 2.65 2.15 2.45 2.60 2.10 2.05 2.00 1.95 1.90 1.85 2.40 2.45 2.50 2.55 2.60 2.40 2.35 2.30 2.25 2.20 2.15 2.35 2.65 2.40 Log ulna length (mm) 2.45 2.50 2.55 2.60 2.40 2.35 2.30 2.25 2.20 2.50 2.55 2.60 2.65 2.80 2.75 2.70 2.65 2.60 2.55 2.50 2.35 2.40 2.50 2.55 2.60 2.65 2.50 2.45 2.40 Log tibiotarsus length (mm) 2.55 2.40 2.30 2.20 2.10 2.35 2.45 2.50 2.55 2.60 2.65 2.50 2.55 2.60 2.65 2.60 2.65 2.60 2.65 2.80 2.75 2.70 2.65 2.60 2.55 2.50 2.45 2.35 2.40 2.45 2.50 2.55 2.60 2.65 2.60 2.45 Log ulna length (mm) 2.50 Log ulna length (mm) 2.40 I 2.70 Log femur length (mm) Log wing-finger phalanx 3 length (mm) 2.45 H G 2.40 2.35 Log ulna length (mm) Log ulna length (mm) 2.30 2.35 2.40 F Log wing-finger phalanx 1 length (mm) Log wing-metacarpal length (mm) 2.45 2.45 2.45 Log ulna length (mm) E 2.50 2.40 2.50 Log ulna length (mm) D 2.15 2.35 2.55 2.30 2.35 2.65 Log wing-finger phalanx 2 length (mm) 1.80 2.35 Log radius length (mm) 2.20 Log humerus length (mm) Log scapula length (mm) A 33 2.00 2.35 2.40 2.45 2.50 2.55 2.60 2.65 2.50 2.40 2.30 2.20 2.10 2.35 2.40 Log ulna length (mm) 2.45 2.50 2.55 Log ulna length (mm) Figure 4. Linear scatter plots of log-transformed length measurements of the long bones of selected ornithocheirid specimens from the Santana Formation. A, scapula. B, humerus. C, radius. D, wing-metacarpal. E, wing-finger phalanx 1. F, wing-finger phalanx 2. G, wing-finger phalanx 3. H, femur. I, tibiotarsus. (ulna length) are shown in Figure 4A–I. As Frey & Martill (1994) found in their morphometric analysis of the ornithocheirids, there was no significant departure from isometry for any of the long bones (regression ANOVA, P > 0.05). This can be seen clearly in Figure 5, which shows the relative lengths of all of the long bones, presented as percentages of the length of the ulna, plotted against the ulna length. There is no significant change in the relative length of any element with increasing size. This demonstrates that the ornithocheirids as a group were remarkably homogenous: despite differences in the size, taxonomic status, and ontogenetic stages of the specimens examined, the skeletal proportions remained essentially the same. Table 1 shows the relative lengths of the limb bones expressed in terms of ulna length and scapula length, and the estimated bone lengths of A. santanae. JOINT FUNCTIONAL MORPHOLOGY Table 2 lists the joints of the ornithocheirid wing skeleton, which specimens were used to analyse each joint, and how they were classified. Each is described below, and the reconstructions are compared with those of other workers. Only limited morphological information is given here – further details are given by Wellnhofer (1985, 1991b), Kellner & Tomida (2000), and Veldmeijer (2003). SCAPULA–NOTARIUM JOINT In the ornithocheirids, as in many large pterodactyloids, the scapula is directed medially, and articulates with a facet on the neural spine of the fourth dorsal vertebra, which in skeletally mature specimens is fused into the notarium, formed by the first © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 34 M. T. WILKINSON 160 sc hu ra wm wf-ph1 wf-ph2 wf-ph3 fe tt Relative bone length (%ulna length) 140 120 100 80 60 40 20 0 220 240 260 280 300 320 340 360 380 400 420 Ulna length (mm) Figure 5. Relative lengths of the long bones of ten selected ornithocheirid specimens, expressed as a percentage of the ulna length. The collection of points at each value of absolute ulna length represents measurements taken from a single specimen. Table 1. Allometric expressions for the lengths of the principal limb bones of ornithocheirid pterosaurs in terms of ulna length (ul) and scapula length (sc), and the estimated lengths of the corresponding limb bones of Anhanguera santanae (AMNH 22555). The number n of specimens used to derive each expression is indicated Bone n Length w.r.t. ulna Length w.r.t. scapula Estimated bone lengths of A. santanae (AMNH 22555) (mm) sc hu ul ra wm wf-ph1 wf-ph2 wf-ph3 wf-ph4 pt fe tt 8 10 12 6 8 7 3 4 1 1 4 2 0.13(ul)1.16 0.51(ul)1.05 – 1.00(ul)1.00 0.98(ul)0.94 1.20(ul)1.03 0.71(ul)1.11 0.30(ul)1.22 0.88(ul)1.00 0.47(ul)1.00 0.04(ul)1.49 0.04(ul)1.53 – 3.34(sc)0.91 6.03(sc)0.87 6.02(sc)0.86 5.29(sc)0.81 7.73(sc)0.90 5.23(sc)0.96 2.68(sc)1.05 2.82(sc)1.00 1.50(sc)1.00 0.51(sc)1.29 0.57(sc)1.32 90 198 296 293 205 435 395 308 254 136 170 216 five dorsal vertebrae and sometimes by the last cervical (Wellnhofer, Buffetaut & Gigase, 1983; Veldmeijer, 2003). This arrangement firmly braced the shoulder girdle on the vertebral column, creating a rigid base for the wing. The articular surfaces of the scapula–notarium joint are rather flat in the orni- thocheirids, being slightly concave on the notarium and convex on the scapula. In side view, the facets are roughly oval. The corresponding surfaces can be brought into near-congruence, indicating that there was only a thin layer of cartilage between them in life. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE 35 Table 2. Characterization of the principal joints of the ornithocheirid fore- and hindlimbs, and the specimens used to analyse each joint Joint Specimens used Joint type Degrees of freedom scno stco sh el ruc is mc cpt cm knu ip hp kne NSM-PV 19892 – left NSM-PV 19892 – right NSM-PV 19892 – left + SMNK 1133PAL – right SMNK 1133PAL – right SMNK 1133PAL – right SMNK 1133PAL – right SMNK 1133PAL – right SMNK 1133PAL – right SMNK 1133PAL – right NSM-PV 19892 – left + SMNK 1133PAL – left AMNH 22552 – right NSM-PV 19892 – right NSM-PV 19892 – right Immobile Constrained saddle Multiaxial Uniaxial hinge Uniaxial hinge with conjunct rotation Uniaxial hinge Uniaxial hinge Uniaxial hinge with conjunct rotation Uniaxial pivot Uniaxial hinge with conjunct rotation Immobile Ball-and-socket Uniaxial hinge 1 1 3 1 1 1 1 1 1 1 0 3 1 The opinion here is that the scapula–notarium joint was essentially immobile: i.e. that the joint did not permit muscle-controlled movements, but that it may have had a limited degree of flexibility (if no movement were permitted at all, one would expect fusion to have occurred). The flatness of the articular surfaces would, in my opinion, permit a maximum range of fore-and-aft angulation of only about 10° (incidentally, the curvature of the joint surfaces is similar to that of the interphalangeal joints, which are universally regarded as immobile, in the sense that they would not have permitted muscle-controlled movements – see below), and the proximity of the transverse processes of the vertebrae and the elongate shape of the facets would have prohibited significant long-axis rotation. The scapulocoracoids were additionally constrained by their articulations with the sternum (see below), which would have prohibited elevation or depression. Bennett (2001) and Wellnhofer (1991b) provided alternative interpretations to those just stated, proposing that significant voluntary movement was possible at the scapula–notarium joint. This is here regarded as unlikely, not only because the degree of movement suggested by these authors would have caused disarticulation, but also because the joint presumably evolved in order to provide a stable base for the humerus: a high degree of mobility would have been incompatible with this function. The orientation of the scapulocoracoid in life is indicated in Figure 6. The glenoid was unusually high in the ornithocheirids – nearly co-planar with the vertebral column, as has been noted before (Wellnhofer, 1991b; Frey, Buchy & Martill, 2003a). sc no gl co stp cs Figure 6. Diagrammatic anterior view of the left pectoral girdle of a typical ornithocheirid. For a list of anatomical/ arthrological abbreviations, see Appendix 1. STERNOCORACOID JOINT The articulation between the coracoid and the sternum, at the rear of the cristospine, is a saddle joint, concave in the fore-and-aft direction, but convex in the mediolateral direction. Such joints are usually biaxial; however, rotation of the sternum about a fore-and-aft axis would usually be prevented by the presence of the other coracoid, and hence rotation can generally only occur about a transverse axis. It is possible that asymmetric angulations of the left and © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 36 M. T. WILKINSON right scapulocoracoids, which could have occurred passively during walking for instance, may have caused small rotations of the sternum about a vertical axis passing between the two sternocoracoid joints, but it is unlikely that such a movement would have had functional significance. Rotation about a transverse axis causes the sternal plate to tilt up and down. The sternum of birds moves in a similar fashion during the ventilation cycle to expand and contract the air sacs (Zimmer, 1935), and it is possible that the movement of the sternum in pterosaurs had a similar ventilatory function, a view shared by Bennett (2001). Conversely, it has been suggested that the ribcage lacked sufficient mobility for such movement, and that pterosaurs instead used movements of the pelvis and abdominal wall, or maybe a crocodilian-type hepatic piston, for ventilation A sc SHOULDER The ornithocheirid shoulder joint is a deceptively complex structure. The glenoid fossa – the contribution of the scapulocoracoid to the shoulder joint – consists of a shallow, largely biconcave scapular part, which faces laterally, and a saddle-shaped coracoid part, which faces dorsolaterally and a little to the rear (Fig. 7A). The coracoid part is convex in the fore-andaft direction and concave in the mediolateral direction. The scapular part can itself be divided into a small anterior region and a large posterior region, with the two being separated by a shallow ridge. The posterior region extends aft beyond the articular B ri scgl cogl (Carrier & Farmer, 2000; Ruben, Jones & Geist, 2003; Claessens, Unwin & O’Connor, 2006). 1ax 2ax 1ax gl 2ax 3ax dpcr co C hu 1ax co dpcr D 2ax hu sc sc dpcr Figure 7. A, reconstructed articular surfaces of the right shoulder joint of Coloborhynchus robustus (SMNK 1133PAL). Elements are oriented as if articulated in their close-packed position: scapulocoracoid in lateral view and humerus in medial view. Scale bar: 50 mm. B, diagrammatic representation of (A), showing contact areas in the close-packed position (shaded) and the joint axes. C, right scapulocoracoid and humerus in the close-packed position in anterior aspect, viewed along the primary axis. D, (C) in dorsal aspect, viewed along the secondary axis. For a list of anatomical/arthrological abbreviations, see Appendix 1. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE surface of the coracoid by a distance equal to a little less than one half of the total width of the glenoid fossa. The head of the humerus is biconvex and slightly elongate in the fore-and-aft direction (Fig. 7A). In medial aspect it is bean-shaped, with a deeply excavated ventral border. The articular surface is asymmetrical, with a narrow anterior region, but expanded posterior region, corresponding to the similar asymmetry of the scapular part of the glenoid. In some specimens (e.g. SMNK 1133PAL, but not NSM-PV 19892) this articular surface is clearly delineated dorsally by a ridge of bone. The shoulder joint is multiaxial, but is not a true ball-and-socket joint, thanks to the partially saddleshaped glenoid fossa and the elongate nature of the head of the humerus, which restricts the range of protraction and retraction (nominal fore-and-aft angulation). In the close-packed position, the ridge on the head of the humerus (if present) closely follows the dorsal outline of the glenoid, and its ventral concavity fits closely around the articular surface of the coracoid. The estimated thickness of cartilage of specimen NSM-PV 19892 is 3 mm – about 7.5% of the width of the humeral articular surface. When in the close-packed position, and when the scapulocoracoid is oriented as in life, the shaft of the humerus lies in a horizontal plane and is directed backwards by about 10°. The contact areas in the close-packed position are shown in Figure 7B. The principal movement at the shoulder joint is a nominal elevation/depression of the humerus, with an estimated maximum range (that may have been reduced in life) of 95°. The primary axis passes through the head of the humerus, and is tilted back from the horizontal by about 35° (Fig. 7B, C), with the result that the humerus swings backwards a little when it is elevated, and forwards when it is depressed. The estimated maximum elevation from the horizontal is 70°. During elevation, the smaller, anterior part of the head of the humerus moves wholly onto the articular surface of the coracoid, whereas the larger posterior region of the head of the humerus rolls across the posterior region of the articular surface of the scapula. There is an indication that the humerus would usually have undergone a conjunct supination (backward long-axis rotation) of about 20° as it approached the closepacked position from above, as slight pronation (forward long-axis rotation) of the humerus during elevation maintains a greater area of articular contact. The humerus can also be depressed from its close-packed position by 25°, at which point the proximal end of the humeral shaft abuts onto the articular surface of the coracoid, preventing further angulation. 37 The secondary axis of the shoulder is at 90° to the primary, but is located medial to it, passing through the scapulocoracoid (Fig. 7B, D). Angulation around the secondary axis is predominantly manifest as protraction/retraction, but, thanks to the 35° backward tilt of the axis with respect to the vertical (corresponding to the equivalent tilt of the primary axis with respect to the horizontal), protraction is accompanied by a slight elevation, and retraction is accompanied by a slight depression. The scapular part of the glenoid is shallow in the fore-and-aft direction, and, as a result, the maximum range of angulation about the secondary axis is more limited than that about the primary axis: the humerus can be retracted from the close-packed position by no more than 40°, and can be protracted by only 10°. This asymmetry is caused by the nature of the fore-and-aft curvature of the coracoid part of the glenoid. Specifically, the radius of curvature progressively decreases from front to back. Hence, the congruence between the deeply curved ventral excavated region of the humeral head and the coracoid remains high throughout retraction. Beyond 40° retraction, contact with the coracoid is lost, and the contact area between the scapula and the humeral head is reduced nearly to zero. Because of the comparatively shallow curvature of the anterior region of the coracoid part of the glenoid, any substantial degree of protraction from the close-packed position causes the congruence between the articulating surfaces to diminish rapidly. The above description of secondary angulation applies only when the humerus is horizontal or nearhorizontal. The situation is different if the humerus is elevated. In this latter configuration the ventral excavated region of the head of the humerus is clear of the coracoid, allowing the humeral head to roll freely across it when protracted or retracted. The curvature of the coracoid part of the glenoid in the fore-and-aft direction is much deeper than that of the scapular part, so a greater range of angulation is possible from this position: up to 30° protraction and 50° retraction. In addition to the angulations described above, the humerus can also be rotated about its long axis. In medial aspect this axis is located very close to the centre of curvature of the ventral concavity of the humeral head (Fig. 7B). Hence, during long-axis rotation away from the close-packed position, the humerus does not simply spin in place, but slides around the articular surface of the coracoid. The estimated maximum range of supination is 50°, and the maximum range of pronation is 30°. Supination of the humerus from the close-packed position is topologically equivalent to elevation, followed by retraction, followed by depression back to the close-packed position (this phenomenon – known as a diadochal movement – can be demonstrated by raising one’s © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 38 M. T. WILKINSON Rayner, flapping could only have involved alternate supination and pronation, without the usual dorsoventral angulation. However, the specimen used by these authors – Santanadactylus brasilensis (M4894), housed in the Geological Institute of the University of Amsterdam, the Netherlands (de Buisonjé, 1980) – is incomplete, missing a large part of the scapular part of the glenoid. One must therefore conclude that contact areas, and therefore ranges of movement, were significantly underestimated in their study. arm to the side, swinging it forward, then depressing it: the arm will have been pronated by 90°). The reconstruction presented above is broadly in agreement with published accounts: most workers accept that the principal motion at the shoulder was elevation/depression, and that the range of elevation of the humerus above the horizontal was much larger than the range of depression below it (Hankin & Watson, 1914; Bramwell & Whitfield, 1974; Wellnhofer, 1985, 1991b). These authors have reconstructed a smaller range of secondary angulation than that presented here, but Bramwell & Whitfield (1974) noted that the range of protraction/retraction was greatly increased as the humerus was elevated. Bennett (2001), in contrast, believed that the humerus could be retracted to 65° behind the transverse axis, even at the bottom of the downstroke. I believe that this degree of retraction is possible, but only if the humerus is first supinated, in which case nominal retraction is actually an angulation around the primary axis, not the secondary. Retraction of the humerus by 65° from the close-packed position, as stated above, causes the humerus and coracoid to separate completely, and greatly reduces contact with the scapula. The only study in which the conclusions differ substantially from those presented above is an analysis of an isolated shoulder joint by Hazlehurst & Rayner (1992). It was concluded that the only possible movement at this joint was rotation around the humeral long axis, in marked contrast to the results of other workers, including myself. According to Hazlehurst & A rufac dpr ra ELBOW The elbow joint is uniaxial. Concave facets (shallow surfaces) and cotyles (deeper depressions) on the radius and ulna articulate with convex facets, and condyles, on the anterior side of the distal humerus (Fig. 8A). The axis is oriented 20° forward of the vertical (Fig. 8B), so that when flexed, the forearm swings forwards and a little downwards. At maximum extension (the close-packed position), a ridge on the proximal surface of the ulna fits into a groove in the distal surface of the humerus, and prevents further extension (Fig. 8A, B). In this position, the radius/ ulna and humerus make an angle of 160°, which agrees reasonably well with other estimates for the Ornithocheiroidea sensu Unwin (2003): 150° for Santanadactlyus (Wellnhofer, 1985) and Pteranodon (Bennett, 2001), and 145° for the basal ornithocheiroid Istiodactylus (Bramwell & Whitfield, 1974). There is considerable disagreement in the literature regarding the maximum range of flexion at B hufac pn ul hu racon gr ulcon C ri hucot ax ax ax hu ul ra dpcr Figure 8. A, reconstructed articular surfaces of the right elbow joint of Coloborhynchus robustus. Elements are oriented as in Fig. 7: humerus in lateral view, and radius and ulna in medial view. Scale bar: 50 mm. B, diagrammatic representation of (A), showing contact areas in the close-packed position and the joint axis. C, right humerus, radius, and ulna in the close-packed position in dorsal aspect, viewed along the joint axis. For a list of anatomical/arthrological abbreviations, see Appendix 1. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE the distal syncarpal and the wing-finger metacarpal. When analysing the radioulnocarpal joint, the proximal syncarpal was temporarily fixed with adhesive putty to the distal syncarpal, which was in turn fixed to the wing metacarpal, with the intersyncarpal and carpometacarpal joints placed in their close-packed positions. The joint axis and range of movement could then be more accurately determined thanks to the greater lengths of the articulating elements. Similarly, when analysing the intersyncarpal joint, the proximal syncarpal was fixed to the radius/ulna, and the distal syncarpal was fixed to the wing metacarpal. There are three additional splint-like metacarpals upon which the three short, clawed digits I–III articulate, but the arrangement of these bones has little bearing on the three-dimensional geometry of the wing. the elbow for the ornithocheiroids: estimates range from 30° (Bramwell & Whitfield, 1974) to 120° (Bennett, 2001); in other pterosaurs the morphology of the elbow joint is very different, and the maximum range of flexion is considerably greater (Wellnhofer, 1991a). The most likely cause of the large discrepancy is the different estimates of the extent of cartilage in the joint. I estimate that the thickness of cartilage in the elbow of C. robustus (SMNK 1133PAL) was approximately 5 mm – i.e. 6.5% of the width of the distal humeral articular surface. Taking this into account, the maximum range of flexion was estimated at 90°. WRIST The pterosaur wrist comprises four elements: the proximal syncarpal, formed by the fusion of the two proximal carpals (Bennett, 1993); the distal syncarpal, formed by the fusion of three distal carpals (Bennett, 1993); the block-like medial carpal (Padian, 1984), also termed the distal lateral (Wellnhofer, 1985) and preaxial carpal (Bennett, 2001); and the long, slender pteroid, which in life supported a membranous propatagium (forewing) in front of the arm (Wellnhofer, 1991a). There are five wrist joints: (1) the radioulnocarpal joint between the radius/ulna and the proximal syncarpal; (2) the intersyncarpal joint between the proximal and distal syncarpals; (3) the medial carpal joint between the distal syncarpal and the medial carpal; (4) the carpopteroid joint between the medial carpal and the pteroid; and (5) the carpometacarpal joint between A ul psfac rafac Radioulnocarpal joint The proximal surface of the proximal syncarpal bears two concave facets, which receive the convex facets of the radius and ulna, and a hemispherical tuberculum, which fits into a circular depression (fovea) in the distal surface of the ulna (Fig. 9A, B). The radioulnocarpal joint is uniaxial. As other workers have noted, flexion of the elbow automatically flexes the radioulnocarpal joint (Hankin & Watson, 1914; Bramwell & Whitfield, 1974). As the elbow is flexed, the radial facet and condyle on the distal humerus push against the radius, causing it to slide along the ulna. This movement is constrained to longitudinal translation by a pronounced ridge on the anterior surface of the ulna, near its proximal end, and another at its distal B ulfac ra psfac 39 ax ax ps tu fov C crax wm ul ax ra ds ps Figure 9. A, reconstructed articular surfaces of the right radioulnocarpal joint of Coloborhynchus robustus. Elements are oriented as in Fig. 7: radius and ulna in lateral view, and proximal syncarpal in medial view. Scale bar: 50 mm. B, diagrammatic representation of (A), showing contact areas in the close-packed position, the joint axis, and the conjunct rotation axis. C, right radius, ulna, syncarpals, and wing metacarpal in their respective close-packed positions in posterodorsal aspect, viewed along the radioulnocarpal joint axis. For a list of anatomical/arthrological abbreviations, see Appendix 1. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 40 M. T. WILKINSON Intersyncarpal joint The intersyncarpal joint is a sliding articulation. The articular surface of the proximal syncarpal bears a prominent ridge that runs anteroventrally, with concave facets to either side, resembling a short section of a left-handed corkscrew in the right-hand limb (Fig. 10A). The corresponding articular surface of the distal syncarpal fits very closely when the radius/ulna and wing metacarpal are in line, indicating that only a thin layer of cartilage was present. The joint axis is tilted back from the vertical by 40°, and the maximum range of angulation is about 25°. Flexion of the joint retracts the wing metacarpal by 20°, and depresses it by 15°, and is accompanied by a slight posteroventral translation of the distal syncarpal with respect to the proximal. The very tight fit between the proximal and distal syncarpals has led many workers to believe that it was incapable of muscle-controlled movement (Hankin & Watson, 1914; Bramwell & Whitfield, 1974; Padian, 1984). In this case, the joint would have functioned as a shock absorber, reducing the risk of breakage of the wing in turbulent conditions, for example. Opponents of this idea (Unwin, 1988; Bennett, 2001) have argued that the range of permitted movement is too large to be accounted for solely by nonvoluntary movement. The maximum theoretical range of 25° estimated for the intersyncarpal joint here would indeed be large for a passive joint, but it must be remembered that this range may have been restricted in life by the presence of ligaments. The evidence from the bones alone is clearly inconclusive, end, that together support the radius ventrally. As the radius slides along the ulna, its distal articular surface pushes on the radial facet of the proximal syncarpal, causing the latter to pivot around the tuberculum, thus flexing the joint. The radial facet occupies an anterodorsal position on the surface of the proximal syncarpal, above and slightly in front of the tuberculum, so that a push on the radial facet causes the wing metacarpal to swing backwards and downwards. Hence, the joint axis is tilted back from the vertical by 60° (Fig. 9B). As the proximal syncarpal pivots about the tuberculum, the ulnar facet – located behind the radial facet – slides across the convex articular surface of the ulna, causing the syncarpal to undergo a conjunct supination of 20° about the tuberculum over the full course of flexion. This conjunct rotation, which has been noted by other workers (Hankin & Watson, 1914; Bramwell & Whitfield, 1974), does not cause an equivalent rotation of the joint axis, which remains fixed relative to the ulna throughout flexion. At maximum extension – i.e. at maximum extension of the elbow – the angle between the radius/ulna and wing metacarpal is 175°. There is no evidence that the wrist could be hyperextended (i.e. angulated forwards beyond 180°), as suggested by Bramwell & Whitfield (1974). The maximum range of flexion is 50°, beyond which the radius fails to make contact with the radial facet of the proximal syncarpal. At maximum flexion, and with the radius/ulna oriented parallel to the transverse axis, the wing metacarpal is directed backwards by 35° and downwards by 50°. A gr dsfac ps B mcfac ax ds ri pn ax psfac ax C ul wm ds ps ra Figure 10. A, reconstructed articular surfaces of the right intersyncarpal joint of Coloborhynchus robustus. Elements are oriented as in Fig. 7: proximal syncarpal in lateral view and distal syncarpal in medial view. Scale bar: 50 mm. B, diagrammatic representation of (A), showing contact areas in the close-packed position and the joint axis. C, right radius, ulna, syncarpals, and wing metacarpal in their respective close-packed positions in posterodorsal aspect, viewed along the joint axis. For a list of anatomical/arthrological abbreviations, see Appendix 1. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE wm ds ps ten Figure 11. A, reconstructed articular surfaces of the right medial carpal joint of Coloborhynchus robustus. Elements are oriented as in Fig. 7: distal syncarpal in anterior view and medial carpal in posterior view. Scale bar: 25 mm. B, diagrammatic representation of (A), showing contact areas in the close-packed position and the joint axis. For a list of anatomical/arthrological abbreviations, see Appendix 1. and until a detailed reconstruction of these soft tissues is carried out, a definitive pronouncement on this matter cannot be made. Such a reconstruction is beyond the scope of this paper. Medial carpal joint The articular surfaces of the medial carpal joint are elliptical – i.e. are elongated in the dorsoventral direction (Fig. 11). The facet on the medial carpal is biconcave, and that on the distal syncarpal is biconvex. The joint permits a small range of movement (about 20°) in the mediolateral direction, but this movement may not have been under voluntary control. The facet of the medial carpal is nearly symmetrical about its dorsoventral and transverse axes, which makes it difficult to tell which way up the medial carpal should be articulated. This is important, because it affects the orientation of the concave distal cotyle of the medial carpal, within which the pteroid may have articulated (see below). In one restoration, the cotyle is directed anterodorsally (Wellnhofer, 1985; Bennett, 2001), in the other it points anteroventrally (Padian, 1984; Wellnhofer, 1991b). These two reconstructions have radically different implications for the function of the pteroid, which will be discussed further below. Carpopteroid joint The nature and function of the carpopteroid joint are highly contentious. Until relatively recently, there was a general consensus that the pteroid articulated in the distal cotyle of the medial carpal, and that it was directed medially, i.e. towards the body (Bramwell & Whitfield, 1974; Padian, 1984; Wellnhofer, 1985, 1991a). Both of these assertions have been questioned of late. Bennett (2001, 2006) pointed out that in many three-dimensionally preserved specimens a small oval sesamoid (sesamoid A) is located within the distal cotyle of the medial carpal, and that in no specimen is the pteroid preserved in articulation ses mc ul 41 ra pt Figure 12. Reconstruction of the right wrist of Coloborhynchus robustus in dorsal view according to descriptions provided by Bennett (2001, 2006), with a sesamoid bone within the distal cotyle of the medial carpal, and the pteroid articulating on the side of the medial carpal. The postulated trajectory of the wing-finger metacarpal extensor tendon, in which the sesamoid is embedded, is also shown. For a list of anatomical/arthrological abbreviations, see Appendix 1. Scale bar: 50 mm. there (Bennett, 2006). A sesamoid usually lies within a tendon at a point where it passes over a bony protuberance proximal to its insertion, and serves to increase the length of the tendon’s moment arm and to improve its mechanical advantage. Bennett concordantly argued that sesamoid A was embedded in the tendon of a wrist extensor muscle, where it ran through the distal cotyle of the medial carpal proximal to its insertion on the wing metacarpal, and that the pteroid articulated on the medial side of the medial carpal (Fig. 12). In this reconstruction, the medial carpal is oriented such that the cotyle faces anterodorsally. The sesamoid would have been essentially fixed with respect to the medial carpal, so that both would have acted as a single unit roughly analogous to the pisiform bone of the wrist of mammals and reptiles (Romer, 1970). I have two objections to this reconstruction. Firstly, although there are suggestive depressions and grooves on the medial side of the medial carpal of the azhdarchid Quetzalcoatlus (J. R. Cunningham, pers. comm.), there is no indication of an articular facet on the equivalent surface of any three-dimensionally preserved ornithocheirid medial carpals. The surface is nearly flat, and is therefore incongruent with the convex head of the pteroid. Secondly, Bennett’s proposed arrangement, in which the sesamoid is lodged within a deep concavity and is not directly associated with a joint, is to my knowledge unique and quite unlike equivalent systems in extant vertebrates, despite the fact that the proposed function of the sesamoid is typical of such bones. I propose instead that sesamoid A was embedded within the tendon of a pteroid extensor or flexor muscle, and that, as is generally the case with sesamoids, it was closely associated with a joint, i.e. the carpopteroid joint. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 42 M. T. WILKINSON I suggest that the common presence of the bone within the distal cotyle of the medial carpal is the result of post-mortem dislocation of the pteroid, which could have pulled sesamoid A into apparent articulation with the medial carpal, just as the sesamoids of the human hand and foot are sometimes pulled into their associated joint capsules by dislocation of the fingers or toes (Wood, 1984; Del Rossi, 2003). This raises the question of why such a dislocation of the pteroid is universally seen in three-dimensionally preserved specimens. One possibility relates to the fact that the pteroid points ventrally to some extent throughout its articular excursion, such that the pteroid tip is always some distance below the arm. Contact with the substrate following death would therefore push the tip of the pteroid dorsally, and could conceivably prise the bone out of the distal cotyle of the medial carpal. Contraction of the propatagium proximal to the wrist could bring about the same effect. I conclude from the above that the pteroid did articulate in the distal cotyle of the medial carpal, but I disagree with the prevailing opinion that the pteroid pointed towards the body, owing to the form of the corresponding joint surfaces. These are complex (Fig. 13). The cotyle of the medial carpal is shaped like a portion of the interior surface of a cone, and if the medial carpal is oriented such that the cotyle faces anteroventrally, as I believe is the case, for reasons given below, the apex of the cone points lateroventrally. The convex articular surface of the pteroid is elongated medially, and is similarly asymmetrical, with a broad, subtriangular medial condyle, and a narrow, roller-like lateral condyle. The head of the pteroid is offset ventrally from the shaft by 40°. Further details of pteroid morphology are given by Unwin et al. (1996). The following description of the functional morphology of the carpopteroid joint proceeds from the assumption that the pteroid articulated in the distal cotyle of the medial carpal. The joint is uniaxial in the ornithocheirids. Articular movement is illustrated in Figure 14. At maximum extension (Fig. 14A–C), the articular head of the pteroid is oriented horizontally, such that its shaft points anteroventrally, 15° beneath the horizontal plane. It is this configuration that defines what I believe to be the correct orientation of the medial carpal: flipping the medial carpal upside down, as in Bennett’s (2001, 2006) reconstruction, causes the pteroid to be directed anterodorsally. Given that the pteroid supported the propatagium, this orientation would have given the wing negative camber: a wholly unfeasible shape. At maximum extension, a shallow semicircular facet on the dorsal surface of the pteroid, just distal to the articular head (Unwin et al., 1996), fits tightly against the upper part of the articular surface of the medial carpal, which acts as a bony stop, preventing further elevation. As the joint is flexed, the large medial condyle of the head of the pteroid slides dorsally around the medial edge of the articular cotyle of the medial carpal, whereas the narrow, lateral condyle rolls in place. If one imagines that the cotyle of the medial carpal is a portion of the inside of a cone, the articular head of the pteroid sweeps the surface of this cone, such that the medial edge circumscribes its base, and the lateral edge is held in its apex. The 95° range of flexion is therefore accompanied by a conjunct lateral rotation of the joint axis by 55°, with the sense being anticlockwise for the right wrist if viewed from the front (Fig. 13B). This causes the pteroid to undergo an arcuate swing, with an initial depression (Fig. 14D–F), giving way to adduction as the limit of flexion is approached. At maximum flexion (Fig. 14G– I), the shaft of the pteroid points ventromedially 25° beneath the horizontal, and also posteriorly 5° behind the transverse axis. Further flexion is prevented by the ventral lip of the articular cotyle of the medial carpal, which acts as another bony stop. The above reconstruction of the carpopteroid joint is controversial, and will be discussed further below, Figure 13. A, reconstructed articular surfaces of the right carpopteroid joint of Coloborhynchus robustus. Elements are oriented as in Fig. 7: medial carpal in anterior view and pteroid in posterior view. Scale bar: 25 mm. B, Diagrammatic representation of (A), showing contact areas in the close-packed position and the joint axis. The axis rotates with respect to the medial carpal during angulation of the pteroid, and is indicated at maximum extension (ext) and maximum flexion (flex). For a list of anatomical/arthrological abbreviations, see Appendix 1. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE but for the moment it will be noted that, although a pure medial orientation of the pteroid (with the shaft parallel to the transverse axis) is possible, if the pteroid articulates in the distal cotyle of the medial carpal, articulating the bones in this way uses only a fraction of the available joint surfaces, primarily the medial condyle of the pteroid and the ventral region of the cotyle of the medial carpal. Carpometacarpal joint The carpometacarpal joint is uniaxial, permitting only long-axis rotation of the wing metacarpal. Rotation occurs about a point a short distance dorsal to the centre of a tuberculum, located midway along the medial border of the proximal end of the wing metac- A mc 43 arpal, which fits into a deep, well-defined fovea in the distal syncarpal (Fig. 15A, B). Behind and above the tuberculum are two shallow facets, separated by a step-like groove. These facets articulate in a sliding fashion with corresponding surfaces on the distal syncarpal. In the close-packed position, when the articular facets are completely congruent, the tuberculum of the wing metacarpal abuts onto the posterior border of the fovea in the distal syncarpal, preventing pronation of the metacarpal. From this position the wing metacarpal can be supinated by 20°, at which point the step between the facets on the distal syncarpal locks with the corresponding step on the distal surface of the wing metacarpal, and the tuberculum of the metacarpal makes contact with the D G E H F I pt ax ds B wm ds ul mc ra ps pt C wm mc ds ps ra ul pt Figure 14. Articular movement of the pteroid. A, lateral view of the right distal syncarpal, medial carpal, and pteroid of Coloborhynchus robustus, with the carpopteroid joint positioned at maximum extension (pteroid at maximum elevation), showing the position of the joint axis. The pteroid points forwards and downwards 15° below the horizontal. B, (A) in dorsal view, also showing the radius, ulna, proximal syncarpal, and wing metacarpal, with all joints in their respective close-packed positions. C, (A) in anterior view. D, lateral view of the wrist, with the carpopteroid joint partially flexed (pteroid at maximum depression). The pteroid points downwards 50° below the horizontal. E, dorsal view of (D). F, anterior view of (D). The pteroid is beginning to swing medially, towards the body. G, Lateral view of the wrist, with the carpopteroid joint at maximum flexion. H, dorsal view of (G), with the pteroid apparently pointing medially. I, anterior view of (G), showing the true ventromedial orientation of the pteroid at maximum flexion. The pteroid points downwards 25° below the horizontal. For a list of anatomical/arthrological abbreviations, see Appendix 1. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 44 M. T. WILKINSON Figure 15. A, reconstructed articular surfaces of the right carpometacarpal joint of Coloborhynchus robustus. Elements are oriented as in Fig. 7: distal syncarpal in lateral view and wing metacarpal in medial view. Scale bar: 50 mm. B, diagrammatic representation of (A), showing contact areas in the close-packed position and the joint axis. For a list of anatomical/arthrological abbreviations, see Appendix 1. anterior border of the fovea in the distal syncarpal. Both contacts prevent further supination. The range of movement estimated here closely matches an estimate made by Bennett (2003) on the basis of pathological concentric grooves, indicating osteoarthritis, on the articular surfaces of an ornithocheirid distal syncarpal from the Cambridge Greensand. As with the intersyncarpal joint, there is some disagreement as to whether the carpometacarpal joint was under direct muscular control (Bramwell & Whitfield, 1974; Bennett, 2001) or merely functioned to absorb sudden transient loads, in this case torsional loads (Padian, 1984). Again, the osteological evidence cannot categorically support either of these interpretations. KNUCKLE The knuckle joint of pterosaurs has a very characteristic form, often described as ‘pulley-like’. The distal end of the metacarpal bears two wheel-like condyles that fit into corresponding cotyles in the first phalanx of the wing finger (Fig. 16A, B). The ventral condyle is approximately circular, whereas the dorsal condyle is slightly elongate in the fore-and-aft direction. The close-packed position is attained at maximum extension. Hyperextension of the joint is prevented by the prominent process for the insertion of the wing-finger extensor tendon. In the extended position, the first phalanx of the wing finger and the wing metacarpal make an angle of 180° (Fig. 16C). The joint is uniaxial, with a flexural range of 160° (Fig. 16C, D). The joint axis is oriented vertically, and passes through the centre of curvature of the ventral distal condyle of the wing metacarpal (Fig. 16B). The wing finger undergoes a conjunct pronation of 20° over the full course of flexion, which allows the posterior part of the dorsal cotyle of the first phalanx to clear the metacarpal when the limit of flexion is approached (Fig. 16E, F). During this movement, the dorsal articular surfaces pull away from each other, whereas the ventral surfaces remain in close contact. Because the joint axis remains fixed relative to the ventral distal condyle of the wing metacarpal, angulation of the wing finger is cardinal, not arcuate. This interpretation of the knuckle joint agrees very closely with Bennett’s (2001) description of the knuckle joint of Pteranodon. INTERPHALANGEAL JOINTS The interphalangeal joints are almost universally believed to have been immobile, i.e. they were not under muscular control, but permitted passive deflection. This interpretation results mainly from the shallow nature of the simple articular surfaces. These are oval in outline, apart from the surfaces at the distal end of phalanx 3 and the proximal end of phalanx 4, which are circular. The distal articular surfaces of the phalanges are convex; the proximal surfaces are concave. The articular surfaces of all but the most distal interphalangeal joint are expanded posteriorly, which would have increased joint stiffness in the fore-and-aft direction. The angles between the life position and reference position of the first, second, and third interphalangeal joints (counting from root to tip) are, respectively, 6°, 4°, and 16°, and the corresponding axes about which this virtual rotation occurs are supinated from the vertical by 30°, 30°, and 40°. This backward tilt of the virtual axes means that successive phalanges are angled backwards and downwards, resulting in an overall backward sweep and downward curvature of the wing finger when the other joints of the arm are in their respective close-packed positions, as several other workers have noted previously (Short, 1914; Brower, 1983; Bennett, 2000, 2001). HIP The ornithocheirid hip is a standard multiaxial ball-and-socket joint, permitting elevation/depression, © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE 45 Figure 16. A, reconstructed articular surfaces of the right knuckle joint of Coloborhynchus robustus. Elements are oriented as in Figure 7: wing metacarpal in lateral view and wing-finger phalanx 1 in medial view. Scale bar: 50 mm. B, diagrammatic representation of (A), showing contact areas in the close-packed position and the joint axis. C, right wing metacarpal and wing-finger phalanx 1 in the close-packed position in dorsal aspect, viewed along the joint axis. D, Right wing metacarpal and wing-finger phalanx 1 at maximum flexion in dorsal aspect, viewed along the joint axis, which remains fixed with respect to the wing metacarpal throughout flexion. E, cross section X–X′ of the wing metacarpal viewed medially, with the proximal end of wing-finger phalanx 1 behind, and with the knuckle joint at maximum extension. F, Cross section X–X′ of the wing metacarpal viewed medially, with the proximal end of wing-finger phalanx 1 viewed from behind, with the knuckle joint at maximum flexion. The broken line indicates the position that wing-finger phalanx 1 would take if no conjunct rotation took place during flexion: this position is impossible, as the posterior part of the articular head of wing-finger phalanx 1 would overlap the shaft of the wing metacarpal. For a list of anatomical/ arthrological abbreviations, see Appendix 1. protraction/retraction, and supination/pronation about three mutually perpendicular axes that meet at the centre of the head of the femur. This head is nearly hemispherical, but not perfectly so, being slightly elongate in a direction perpendicular to the axis of the knee joint (see below). The acetabulum is similarly elongate, predominantly in the horizontal direction. Hence, one can define a close-packed orientation of the head of the femur about its long axis, when it fits most comfortably within the acetabulum. The primary axis of the hip (the joint axis around which the greatest range of angulation takes place) is oriented vertically, whereas the secondary axis (perpendicular to the primary) is oriented horizontally. Primary angulation is therefore equivalent to protraction/retraction, and secondary angulation is equivalent to elevation/depression. Without taking account of the constraining action of the soft tissues of the hip, the reconstruction of which is beyond the scope of this paper, the femoral shaft can be protracted 40° forward of the transverse axis, and retracted 70° behind it, and can be depressed 60° beneath the horizontal plane, and elevated 20° above it. There has been some disagreement in the literature as to the habitual orientation of the femur in flight. Bramwell & Whitfield (1974) and Wellnhofer (1985, 1991b) believed that the femur was usually directed back from the transverse axis by about 60°. Bennett (2001) noted that this orientation of the femur is very close to its limit of retraction, and he argued that habitually holding the femur at one end of its range would have restricted its ability to manipulate the wing membrane. He therefore postulated that the femur was usually held at a shallower angle from the transverse axis: between 25° and 30°. When the femur is so oriented, and additionally held at 20° below the horizontal, it is in its close-packed position, and is near the middle of its range of not only protraction/retraction but also elevation/depression. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 46 M. T. WILKINSON From this position, the leg would therefore have the maximum degree of control over the wing: evidence that this position was habitual. This line of reasoning presupposes, of course, that the main wing membrane was attached to the leg. This controversial aspect of pterosaur reconstruction will be discussed below. The maximum degree of depression of the femoral shaft from the horizontal has been the subject of considerable debate, thanks to the importance of this aspect of hip functional morphology in discussions of terrestrial locomotion. The range of depression is determined by two factors: the orientation of the acetabulum, which itself depends partly on the orientation of the pelvis on the vertebral column, and the range of angulation of the femur with respect to the pelvis. The controversy surrounding the orientation of the acetabulum has now been largely resolved, at least for the ornithocheiroids: three-dimensional pelvic material from the Santana Formation indicates that the acetabulum was directed posterolaterally (Bennett, 1990), not dorsolaterally, as suggested by Wellnhofer (1988). However, there is still some disagreement as to the maximum range of movement of the femur. As stated above, my estimate of the extent of femoral depression is 60°, because further angulation causes the head of the femur to begin to slide out of the acetabulum. I believe it is unlikely that the area of contact between the femoral head and the hip socket would be reduced in this way at the very time when adequate support for the pelvic girdle was most vital. Bennett (1997a, b) proposed that the femur could be depressed much further than 60°, perhaps even beyond the vertical. This inference was based partly on the type specimen of Dsungaripterus weii, in which the femora are preserved in articulation with the acetabula, and are adducted to such an extreme degree that their distal ends cross beneath the pelvis (Bennett, 1997a, 2001). I do not believe that this specimen provides useful information regarding this debate, because it only indicates that the bone geometry of the hip joint permits a very wide range of depression. This maximum range would have been reduced in life because of the presence of ligaments and the joint capsule. There are many instances in which fossil bones are preserved in an unnatural state of articulation precisely because the soft tissues of the joints had begun to rot away prior to burial, and there is no way to show conclusively that the hip of the type specimen of D. weii is not one of these cases. A study of a second wellpreserved pelvis and associated femora of a dsungaripterid from Oker, Germany (Upper Jurassic), indicated that the maximum degree of femoral depression was only 40°: any more caused disarticulation (Fastnacht, 2005). Bennett (1997a) also argued that poor joint contact at the hip would not necessarily have been problematic, citing the human shoulder as an example of a loose articulation that is stabilized with muscles and ligaments. This is indeed the case, but the human shoulder is not usually a weight-bearing joint (although it may be transiently used for this purpose by gymnasts, for example). In the vast majority of the terrestrial tetrapod skeletons I have examined, weight-bearing joints (specifically, joints where the weight is borne in compression, not tension as in brachiating primates) have good articular contact where the weight is transmitted from body to limb. In addition, the ornithocheirid hip was clearly not a loose joint: extensive contact was maintained between the femoral head and the acetabulum over a wide range of motion. Furthermore, Fastnacht (2005) argued that partial displacement of the hip throughout terrestrial locomotion, as proposed by Bennett (1997a, b), was highly unlikely from a consideration of the mechanical stability of the joint. Finally, Bennett (1997a, b, 2001) pointed out that the pterosaur knee and ankle joints would not have permitted a significant level of long-axis rotation, and therefore argued that the hindlimbs must have been swung back and forth parasagittally, and that the femora must have been nearly fully depressed (i.e. erect) in order to do so, as originally proposed by Padian (1983). I agree that the tibiotarsi would have moved in parasagittal or near-parasagittal planes, and that the distal end of the femur would have remained at an approximately constant distance from the midline throughout. However, this does not necessarily require the femora to have been completely erect. Even with a semi-erect posture, which is more in line with my estimate of the maximum degree of femoral depression, the hindlimbs could have moved parasagittally without the need for rotation at the knee or the ankle. The reasoning is as follows: for sprawling tetrapods, retraction of the femur is typically coupled with an equal-magnitude conjunct lateral rotation (technically supination) of the crus with respect to the pes, such that the knee axis rotates from an orientation roughly perpendicular to the direction of motion, to an orientation parallel to the direction of motion (Rewcastle, 1981). If the femur is adducted into a semi-erect stance, as in many lizards, this conjunct rotation is reduced in magnitude, and it can theoretically be eliminated altogether if the femur undergoes an equal and opposite adjunct medial rotation (Rewcastle, 1981). This holds the knee axis perpendicular to the direction of travel throughout hindlimb movement, and no ankle rotation is necessary. Considering the bones alone, this movement, hence a parasagittal hindlimb gait, would have been possible for pterosaurs, but further work will be necessary to assess © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE whether the hindlimb muscles were capable of executing the required rotations. KNEE The knee is a simple, uniaxial hinge joint. When the femur is in its close-packed orientation, the knee axis is oriented vertically. At maximum extension, the tibiotarsus and femur make an angle of 175°, and the range of angulation from this position is about 110°. THREE-DIMENSIONAL CONFIGURATION OF THE WING The ornithocheirid joint axes of the right arm and leg, mapped onto the virtual reconstruction of A. santanae, are shown in Figure 17 (humeral/femoral 47 long axes and the carpometacarpal axis have been omitted for clarity). The limbs are arranged along straight lines projecting from the shoulder and hip: this is obviously an unrealistic configuration, but is presented here to show with greater clarity the orientation of the joint axes, without any rotation having taken place around said axes. Table 3 shows the orientation of each joint axis when projected onto the xz (sagittal) plane and the yz (vertical frontal) plane when the limbs are arranged in straight lines, as in Figure 17, and gives the limits of articular movement of respective distal elements about each axis. In most cases, the joint axes lie in planes that are parallel to the xz plane (sagittal plane). The carpopteroid joint axis, carpometacarpal axis, and the long axes of the humerus and femur are the exceptions to this case: these lie parallel to the y-axis. Table 3. Orientations of the principal joint axes of the ornithocheirid fore- and hindlimbs when projected in the sagittal (xz) and vertical frontal (yz) planes, and the range of movement of the respective distal elements about each axis. Axis orientation angles denote anti-clockwise rotations from the x-axis in lateral view for the xz projection, or anti-clockwise rotations from the y-axis in anterior view for the yz projection, for the right limbs. An orientation angle is not given where the projected axis is a single point. Ranges of movement are given with respect to the straight-limbed configuration shown in Figures 3 and 16. Positive angles denote predominantly backward and/or downward angulation or supination of the relevant distal element, and negative angles denote forward and/or upward angulation or pronation of the distal element. Only a single value is given for immobile joints, for which the joint axis is a virtual axis. Angles in parentheses for the secondary axis of the shoulder (sh2) refer to the range of angulation when the humerus is fully elevated. Movement about a conjunct rotation axis (crax) is indissociably tied to the angulation about the relevant joint axis, so the articular limits of the former correspond to those of the latter Axis orientation Joint axis xz projection yz projection Range of movement scno sh1 sh2 sh3 el ruc ruc crax is mc cpt cpt crax cm knu knu crax ip1 ip2 ip3 hp1 hp2 hp3 kne 90° 35° 125° – 70° 150° – 130° 110° – 0° – 90° – 120° 120° 130° 90° 0° – 90° 90° 90° 90° 0° 90° 90° 0° 90° 90° 0° – 0° 90° 0° 90° 90° 90° 90° – 0° 90° 10° -70° -10° -30° -20° 5° 0° 0° -10° 15° 0° 0° 0° 0° 6° 4° 16° -40° -20° ? 5° © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 to 25° (-30°) to 40° (50°) to 50° to -110° to 55° to 20° to 25° to 10° to 110° to 55° to 20° to 160° to -20° to 70° to 60° to 115° 48 M. T. WILKINSON The entire wing, configured with all joints in their close-packed positions except the knee, is shown in Figure 18. In this configuration the wingspan is 4.3 m, which is slightly higher than the 4.15-m estimate of the wingspan of A. santanae made by Figure 17. Three-dimensional virtual model of Anhanguera santanae, showing the positions and orientations of the principal joint axes of the limbs. For a list of anatomical/arthrological abbreviations, see Appendix 1. Wellnhofer (1991b). Various outlines of the wing membranes have been added. Three possible trailing edges of the cheiropatagium are shown that differ in their proximal attachment position: at the distal end of the crus, near the proximal end of the crus, and the hip. There is now abundant evidence from fossilized wing membranes and wing membrane impressions that the trailing edge of the cheiropatagium ran to the distal end of the crus in a number of unrelated pterosaur species (Wellnhofer, 1987; Unwin & Bakhurina, 1994; Lu, 2002; Wang et al., 2002; Frey et al., 2003b). Figure 18 would appear to bear out this inference: a hip attachment results in an unfeasibly narrow cheiropatagium, which would give (1) very high minimum flight speeds owing to the small wing area, and (2) poor lift : drag ratios, on account of the large diameter of the wing spar when compared with the width of the cheiropatagium. The Vienna specimen of Pterodactylus kochi appears to show the trailing edge attaching directly to the femur, but Wellnhofer (1987) has convincingly argued that this apparent relationship is illusory: if the trailing edge of the cheiropatagium had indeed attached to the thigh, one would expect its fossilized impression to stop short some distance from the femur, because of the presence in life of the overlying musculature of the thigh. Wellnhofer (1987) proposed instead that the trailing edge of the cheiropatagium ran further down the leg, and that at the time of fossilization the Figure 18. Three-dimensional virtual model of Anhanguera, with all limb joints in their respective close-packed positions, except the knee, which is shown partially flexed so that the tibiotarsus is directed backwards. A, ventral view, with flight membranes. Three possible trailing edges of the cheiropatagium are shown: running from the wingtip to the distal end of the crus (solid line), the proximal end of the crus (broken line), and the hip (dotted line). Two possible leading edges of the propatagium are shown, corresponding to an anteroventral orientation of the pteroid (solid line) and a medial orientation of the pteroid (broken line – pteroid itself omitted in this case for clarity). Two possible trailing edges of the cruropatagium are shown: running from the tip of the tail to the distal end of the crus (solid line), and to the proximal end of the crus (broken line). See text for further explanation. B, anterior view, membranes omitted for clarity. For a list of anatomical/arthrological abbreviations, see Appendix 1. Scale bar: 500 mm. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE most proximal part of the cheiropatagium was folded underneath the crus, an idea developed further by Pennycuick (1988). Precisely the same arrangement is often seen in the folded wings of bats: in dorsal view, the trailing edge of the plagiopatagium can appear to run to the thigh, when in reality it runs to the ankle (Wellnhofer, 1987). Without direct evidence to the contrary in any pterosaur, it is most parsimonious to conclude that the trailing edge of the cheiropatagium ran at least part-way down the crus in the ornithocheirids. Two possible cruropatagium outlines are shown in Figure 18: the first shows the trailing edge running from the tail to the ankle, after Unwin (1999); the second shows the trailing edge running from the tail, embedded within the cruropatagium, to the knee joint, after Bennett (2001), who argued that removing the crura from the cruropatagium in pterodactyloids would have allowed the membrane to be more easily manipulated independently of the legs by the tail (Bennett, 2001). This condition contrasts sharply with that displayed by the ‘rhamphorhynchoid’ Sordes pilosus, in which the tail was free of the cruropatagium and situated dorsal to it (Unwin & Bakhurina, 1994). This does not necessarily mean that Bennett’s (2001) reconstruction is incorrect: within bats there is great variation in the extent of contact between the tail and the uropatagium, equivalent to the pterosaur cruropatagium (Hill & Smith, 1984). Two propatagium outlines are shown in Figure 18, corresponding to the two prevailing ideas regarding the orientation of the pteroid discussed above: forward-pointing and inward-pointing. In both reconstructions, it was assumed that the leading edge of the propatagium ran from the lower neck, as indicated in the Vienna and Munich specimens of P. kochi (Broili, 1938; Wellnhofer, 1987), to the knuckle. In accordance with published reconstructions, a medially-directed pteroid was assumed to form part of the leading edge of the propatagium itself, whereas a forward-pointing pteroid was assumed to intercept the leading edge only at its tip. These reconstructions of the propatagium will be discussed further below. Video sequences showing the possible joint movements of the three-dimensional reconstruction of A. santanae are available online (see Supporting Information). DISCUSSION The reconstructed three-dimensional geometry of the ornithocheirid wing presented above enables a number of inferences to be made about the group, particularly with regard to aerial and terrestrial locomotion. Much has been written on this subject 49 (e.g. Hankin & Watson, 1914; Short, 1914; von Kripp, 1943; Bramwell, 1971; Heptonstall, 1971; Bramwell & Whitfield, 1974; Brower, 1983; Pennycuick, 1988; Wellnhofer, 1988; Alexander, 1989; Padian & Rayner, 1993; Bennett, 1997a, b; Unwin, 1997; Bennett, 2000, 2001), and in some cases the present study merely confirms what has previously been supposed. This discussion will therefore be largely confined to the presentation of ideas that differ from those of previous accounts, or that clarify areas of disagreement or uncertainty, and will cover matters arising from the reconstruction of the pteroid: stability and control in gliding flight, and terrestrial locomotion. Flapping flight will not be covered, partly because so little is known about the wingbeat kinematics, or the material properties of the wing membrane, that only a cursory discussion can be attempted at present, but also because the ornithocheirids seem to have been predominantly adapted for oceanic soaring, on account of their large size and high aspect ratio wings (Hankin & Watson, 1914; Bramwell & Whitfield, 1974; Brower, 1983). Furthermore, the inferences presented about stability and control should not be regarded as anything more than a working hypothesis at this stage. Further progress can be made only by carrying out wind tunnel and flight tests of accurate scale models, or computational fluid dynamic tests of accurate virtual models. The three-dimensional geometric data presented here can be regarded as a necessary first step towards this goal. Only limited aerodynamic background information is given here. Further details can be found in, for example, Prandtl & Tietjens (1957), Simons (1978), Vogel (1981), Norberg (1990), Barnard & Philpott (1995), and Etkin & Reid (1996). THE PTEROID AND PROPATAGIUM Three possible reconstructions of the carpopteroid joint were described above: in the first, the pteroid points towards the body, and articulates on the medial side of the medial carpal (Bennett, 2001, 2006); in the second, traditional reconstruction, the pteroid also points medially, but articulates in the distal cotyle of the medial carpal (Bramwell & Whitfield, 1974; Padian, 1984; Wellnhofer, 1985, 1991a); in the third, the pteroid articulates in the distal cotyle of the medial carpal, but can swing from an anteroventral orientation to a ventromedial orientation (Unwin et al., 1996; Wilkinson, Unwin & Ellington, 2006). Although controversial, it is the latter reconstruction that is preferred here for reasons that were given above. A forward-pointing orientation of the pteroid was previously suggested by Hankin (1912), Frey & Riess (1981), and Pennycuick (1988), but the idea was treated unfavourably, partly because, in many com- © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 50 M. T. WILKINSON plete, articulated fossils, the pteroid points medially (Padian, 1984; Wellnhofer, 1985). I suggest that this orientation is a taphonomic artefact: if the bone were fully flexed at death, it could easily have been elevated from its nominal ventromedial orientation to a fully medial orientation while remaining in contact with the substrate. As argued above, the same postmortem process, perhaps coupled with contraction of the propatagium proximal to the wrist, could have been responsible for the separation of the pteroid from the distal cotyle of the medial carpal, and the presence therein of sesamoid A (Bennett, 2006). In carrying out the full reconstruction of the wing, shown in Figure 18, I have assumed that the propatagium was attached to the entire length of the pteroid. It could be argued that this was not necessarily the case. The leading edge of the propatagium could conceivably have run from the shoulder to the wrist, excluding the pteroid or part of the pteroid, just as the bat forewing excludes the thumb. This is unlikely for two reasons. Firstly, propatagium impressions of articulated fossils, particularly those from the Solnhofen Limestone, clearly demonstrate its association with the pteroid: the tip of the bone often intercepts the leading edge of the propatagium precisely (Wellnhofer, 1970; Padian & Rayner, 1993). This would be most unlikely if there were no physical connection between the two. Secondly, it is difficult to envisage a function for the pteroid if it did not support the propatagium. It is true that bats can use their thumbs to climb, crawl, handle food, or jump (Hill & Smith, 1984; Schutt et al., 1997), but this is only possible because the thumb is clawed and robust: the slender pteroid would not have been suited to these activities. In reconstructing the propatagium associated with a forward-pointing pteroid, I have assumed that its leading edge ran distally from the tip of the pteroid to the knuckle. My reasons for doing so were as follows: firstly, if the propatagium had not continued beyond the pteroid there would have been a very abrupt reduction in the wing chord (the distance between the leading and trailing edges at the same spanwise station) at this point. This would have caused the creation of a longitudinally aligned vortex, of a kind usually seen only at the wingtips, part-way along the wing, created by the flow of air around the pteroid from the high-pressure region beneath the propatagium to the low-pressure region above it. This itself would have caused an increase in drag and a disruption of the airflow over the cheiropatagium immediately behind the propatagium. Secondly, if there was no membrane distal to the pteroid, tension in the propatagium proximal to the pteroid would have induced a severe bending moment at the carpopteroid joint, and there is no evidence that the distal cotyle of the medial carpal was buttressed to resist such loading. Some authors have argued that there is no evidence for a propatagial membrane distal to the wrist, but this claim is incorrect: the ‘Zittel wing’, assigned to Rhamphorhynchus, from the Solnhofen Limestone (Upper Jurassic) (von Zittel, 1882), and an azhdarchoid from the Crato Formation (Lower Cretaceous) of Brazil (Frey & Tischlinger, 2000) both bear traces of membrane distal to the wrist. The apparent lack of a distal propatagium in other Solnhofen specimens may be another taphonomic artefact: given that the pteroid is fully flexed in these fossils, the membrane distal to the wrist would have become tightly folded against the metacarpus, where it could easily have been obscured by the skeleton or removed during preparation. Frey & Riess (1981) argued that the distal propatagium extended even further along the wing than shown in Figure 18A, terminating near the second interphalangeal joint, and enveloping the short, clawed fingers. However, there is no evidence for a membrane associated with the fingers, even in those specimens with good soft-part preservation (Wellnhofer, 1985, 1987), so a more limited extent is preferred here. Indeed, extending the propatagium to the short digits would have had only a minor influence on its shape, given the small size of these digits in comparison with the pteroid (Fig. 18A). Nevertheless, a greater spanwise extent of the propatagium, with or without involvement of the clawed digits, would have led to a reduction in the compressive load, borne by a forward-pointing pteroid, and may in the fullness of time prove to be the correct reconstruction. An additional possibility is that there was a propatagial membrane distal to the wrist that terminated only part-way along the metacarpus. However, although the currently available soft tissue fossil evidence is of insufficient quality to identify the exact spanwise limit of the propatagium, a narrow distal section is unlikely within the paradigm of a forwardpointing pteroid, because of the concordant increased compressive loading of the pteroid and the rapid reduction of the chord, alluded to above. These arguments would not apply if the pteroid was directed medially. A comparison of Figure 14A–C and Figure 14G–I, which show the limits of angulation of the pteroid presuming it articulated within the distal cotyle of the medial carpal, appears to indicate that the tip of the pteroid moves a very substantial distance when the carpopteroid joint is fully flexed. This would require rather peculiar material properties of the propatagium, particularly the leading edge of the distal propatagium, for the membrane would have to retain sufficient tension in flight, but also be sufficiently elastic to permit furling. This requirement could be regarded as evidence against a forward- © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE pointing pteroid, for such properties would not be necessary if the pteroid pointed medially. However, measurements taken directly from the threedimensional virtual model show that the distance between the tip of the pteroid and the presumed distalmost point of the leading edge of the propatagium increases by only 16% when the carpopteroid joint is flexed from an anteroventral to a ventromedial orientation, i.e. the maximum strain of the leading edge of the distal propatagium was only 16%. This figure would be reduced if the maximum range of flexion of the pteroid were smaller. Nevertheless, it would still be an important stipulation of this reconstruction that propatagial tension was adjustable, which could have been realized by the existence of intrinsic propatagial muscle fibres, although there is as yet no evidence for such fibres in the propatagium. In functional terms, the chief difference between the reconstructions stipulating medially and anteroventrally oriented pteroids is the size of the propatagium. Clearly, it is much larger if the pteroid points forwards, with important consequences for the overall wing loading (weight divided by total wing area), and hence for speed: low wing loading gives lower gliding speeds. In all three reconstructions the pteroid would have had some ability to alter the chordwise camber in the proximal region of the wing (although this ability would be more definitely constrained within the forward-pointing paradigm, because the carpopteroid joint in this case would have had only a single degree of freedom). Hence, regardless of the reconstruction, the propatagium could have acted as a control surface for a number of flight manoeuvres (see below). Additionally, in the forwardpointing reconstruction paradigm, the pteroid can swing into a ventromedial orientation when fully flexed. It is possible that this latter state was the habitual orientation of the pteroid, and that the anteroventral orientation was only used to increase wing area, and hence to reduce wing loading during take-off, landing, or other manoeuvres requiring slow flight speeds, in a similar way to that proposed by Pennycuick (1988). Alternatively, the ventromedial orientation may only have been used to furl the propatagium when the pterosaur was on the ground. FLIGHT STABILITY Stability is defined here as the tendency to return to a given flight path, attitude, and velocity without active control, following a disturbance of one or more of these factors, e.g. by a gust. It has often been assumed that all pterodactyloids were inherently unstable, simply because they lacked any equivalent of the tailplane or tailfin of a conventional aircraft (Bramwell & Whitfield, 1974; Brower, 1983) [the situ- 51 ation is somewhat different for the ‘rhamphorhynchoids’, as they possessed long tails furnished with a vertically-oriented membrane at the tip (Marsh, 1882), but these will not be considered here]. This makes some sense from an evolutionary perspective: stability opposes control, so a high degree of manoeuvrability and a high degree of stability are mutually exclusive (Maynard Smith, 1952; Thomas & Taylor, 2001). The means of acquiring stability also tend to incur a drag cost, and can therefore reduce an aircraft’s aerodynamic efficiency (Thomas & Taylor, 2001) and maximum lift (Maynard Smith, 1952). However, tailless aircraft can be made to fly stably without active control (Weyl, 1945; Pennycuick, 1971a; Nickel & Wolfahrt, 1994), and the same may have been true of the pterodactyloids. Furthermore, pterodactyloids may not have required a high degree of manoeuvrability, and the neurological and mechanical costs of the continuous maintenance of a given flight attitude may have outweighed the drag saving associated with instability. Clearly, these cursory arguments are insufficient for providing any kind of understanding of stability in pterosaurs. A thorough assessment would require a complete dynamic analysis, which is beyond the scope of this paper. However, a qualitative appraisal of the 3D shape of the ornithocheirid wing does allow a number of important conclusions to be drawn. This section will only consider the stability of certain fixed-wing configurations, mainly the fully extended configuration shown in Figure 18: wing adjustments will be considered under flight control below. Pitch stability Pitch (longitudinal) stability is usually achieved by situating the centre of gravity (c.g.) ahead of the mean aerodynamic centre (m.a.c.) – this being the point through which the aerodynamic forces are conventionally taken to act; it usually lies roughly a quarter of the length of the mean chord from its anterior extent (Fig. 19A). The reasoning is simple to understand: a pitching motion in the nose-up sense (a positive moment) increases the angle of attack, and therefore increases the lift. Because this lift acts behind the centre of gravity, the result is a nose-down (negative) moment that restores the original attitude. Technically, this is static pitch stability: the restoring moment may cause the aircraft to overshoot its original attitude, leading to an oscillating motion that, if insufficiently damped, will render the aircraft dynamically unstable. Dynamic stability will not be considered here. For a statically unstable aircraft, the c.g. usually lies behind the m.a.c. The increased angle of attack caused by a nose-up disturbance therefore increases the pitching moment. Without active © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 52 M. T. WILKINSON wing lift A A weight B wing lift tail lift weight Figure 19. A, Stable but unbalanced wing profile, with the centre of gravity (c.g.) situated ahead of the mean aerodynamic centre (m.a.c.). B, stable and balanced configuration, with a small tailplane set at a negative incidence with respect to the main wing. control, the aircraft would diverge from its original attitude, and flight would be impossible. Although placing the c.g. ahead of the m.a.c. renders an aircraft statically stable, this is not in itself sufficient for flight. This is because positive lift in this case always gives rise to a nose-down pitching moment about the c.g. In fact, positively cambered wings (convex surface upwards) are inherently unbalanced: they always generate a nose-down pitching moment about the aerodynamic centre, regardless of the angle of attack (e.g. Simons, 1978; Nickel & Wolfahrt, 1994; Barnard & Philpott, 1995; Etkin & Reid, 1996; Thomas & Taylor, 2001). For these reasons, conventional stable aircraft typically require a second lifting surface to generate a nose-up moment to compensate. A small tailplane is the most common type. Being situated behind the main wing, the tailplane must generate negative lift to provide balance, and it is therefore set at a negative angle of attack, with respect to the main wing (Fig. 19B). The tailplane decreases the overall lift of an aircraft, but increases its drag; hence, aerodynamic efficiency is diminished in this case. In unstable aircraft, the nose-down moment of the cambered wings can be balanced by the aft position of the c.g. with respect to the m.a.c. In this case the lift provides a balancing nose-up moment directly. There are a number of ways in which a stable tailless aircraft can be made to generate a balancing B Figure 20. Methods of achieving longitudinal balance and positive stability without a tail. A, sweepback coupled with washout. The aft-situated tips are at a negative incidence with respect to the forward-situated wing root. B, reflex camber. nose-up pitching moment. If the wings are twisted such that the angle of attack of each wing decreases from root to tip – a geometry known as washout – and are additionally swept back, the aft-situated wingtips can fulfil the balancing function of the tailplane (Fig. 20A). The same effect can be achieved by sweeping the wings forward and twisting them so that the angle of attack increases from root to tip – a geometry known as washin. Alternatively, the wings can be reflex cambered, i.e. deflected upwards near the trailing edge (Fig. 20B), which reduces the magnitude of the inherent nose-down moment about the m.a.c. Evidence of these features in ornithocheirid wings could indicate that they were statically stable in pitch, whereas their absence would suggest static instability. It is impossible to tell directly whether or not the ornithocheirid wings were reflex cambered. The distal region of bat wings have been observed to be © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE upturned near the trailing edge in gliding flight (Pennycuick, 1971a), but this is due to upward deflection of the fourth and fifth digits, for which there is no analogue in pterosaur wings. However, bats can also facultatively reflex the inner panel of the wing membrane (the plagiopatagium) by contracting longitudinally arranged muscles there (Pennycuick, 1971a). It is possible that pterosaurs could do the same thing, for there is tentative evidence of muscles within the cheiropatagium (Frey & Tischlinger, 2000; Frey et al., 2003b). It is unlikely, however, that the wings were habitually reflex cambered, simply because the lift in this case would have been substantially reduced relative to that generated by positively-cambered wings of the same area. Hence, a larger area would be required to generate the same lift, which would also increase the drag and diminish the lift : drag ratio (Thomas & Taylor, 2001). Another way that static pitch stability could have been achieved was by using sweepback coupled with washout or sweepforward coupled with washin. It is most likely that the second of these options would not have been feasible, for there is no doubt that the pterosaur cheiropatagium was twisted in a nose-down sense from root to tip in flight: the aerodynamic load would have lifted the flexible trailing edge of the cheiropatagium until it was balanced by its spanwise tension. Furthermore, as Brower (1983) pointed out, this geometry would have been important for safe flight, because without it the tapered shape of the wings would have caused the aerodynamic angle of attack to increase from root to tip, thereby increasing the likelihood of a potentially dangerous tip stall (e.g. Simons, 1978; Marchaj, 1988; Chatterjee & Templin, 2004). Pitch stability in the ornithocheirids would therefore have depended on the wings being sufficiently swept back. Having said this, depression of the propatagium by the pteroid, if possible, would have reduced the angle of attack of the proximal part of the wing, and could therefore have conceivably caused washin, although this effect may have been cancelled out by the washout of the cheiropatagium. However, it is most unlikely that a strongly deflected propatagium was used habitually, as deflection of a leading edge flap at the relatively low angles of attack associated with cruising flight would have caused substantial leading-edge flow separation on the lower surface of the wing (Fullmer, 1947; Wilkinson et al., 2006). In the fully extended configuration depicted in Figure 18, the wing is slightly swept forward, which suggests pitch instability, based on the reasoning given above. However, it must be remembered that Figure 18 represents only one possible wing configuration that may not have been habitual. In fact, the knuckle joint could certainly not have been extended 53 to its full limit in flight, for the drag force acting on the wing finger would have caused some passive flexion. The greater the typical degree of flexion, the more likely it is that the ornithocheirids were habitually stable in pitch. I am of the opinion that the usual degree of knuckle flexion in flight was quite small, because the habitual position of a joint is usually very near its close-packed position (Williams et al., 1989), which is the position of full extension in this case. This would give only a small degree of sweepback, if any, which would require a rather extreme washout to achieve balance with stability. As with reflex camber, I believe that such a geometry would have compromised aerodynamic efficiency to an unacceptable level. The evidence therefore tentatively suggests that ornithocheirids were habitually unstable in pitch, which has a number of consequences for flight control (see below). Further work could assess whether estimates of the relative positions of the c.g. and m.a.c. support this statement: the prediction is that the m.a.c. was usually located in front of the c.g. (movements of the wings fore and aft could of course alter the relative positions of these points: such adjustments are discussed further below). It should be noted, however, that the inference that the ornithocheirids were unstable in pitch is based purely on a geometric consideration of the wing as a rigid structure. This is clearly an oversimplification: the pterosaur patagia were complex composite membranes. The limited literature on the stability of such wings indicates that compliance enhances pitch stability without the usual trade-off with aerodynamic efficiency (Fink, 1969; Sneyd, Bundock & Reid, 1982; Sneyd, 1984; Krus, 1997). Assessing the degree to which this effect applied to pterosaurs would require a coupled analysis of membrane mechanics and fluid flow. Such an analysis is beyond the scope of this paper. The remainder of this discussion will follow from the assumption that the ornithocheirids were habitually unstable in pitch. Roll stability Roll stability is generally obtained using dihedral: an upward tilt of the wings about the longitudinal axis. When an aircraft or flying animal with this geometry is disturbed about the roll axis, an opposing moment is generated that restores a level attitude (e.g. Simons, 1978; Nickel & Wolfahrt, 1994; Barnard & Philpott, 1995; Etkin & Reid, 1996; Thomas & Taylor, 2001). The anterior view of the extended configuration (Fig. 18B) shows that the dihedral angle was negative if the joints were in their respective closepacked positions: the wings exhibit a downward deflection or anhedral, which would reinforce a roll disturbance. By this criterion, one would conclude © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 54 M. T. WILKINSON that the ornithocheirids were unstable about the roll axis when in this configuration. However, Figure 18B shows the wings in an unloaded condition. The dorsoventral stiffness of the wing spar is unknown, but it is likely that an aerodynamic load would cause an upward deflection of the distal wing-finger phalanges about the interphalangeal joints, particularly given the lack of any dorsoventral buttressing of these joints. A pronouncement regarding roll stability should therefore be postponed until a mechanical analysis of the wing spar is carried out. Elevating or depressing the wings using primary angulation of the shoulder could, of course, have been used to increase or decrease dihedral and roll stability in a facultative manner. Sweeping the wings back could also have conferred roll stability (e.g. Weyl, 1945; Nickel & Wolfahrt, 1994; Barnard & Philpott, 1995; Thomas & Taylor, 2001). Yaw stability Yaw stability is conventionally provided by situating a vertical tailfin some distance behind the c.g. A yaw in one direction inclines the fin with respect to the airflow, such that it generates an opposing moment about the c.g., restoring the original attitude. Neither the ornithocheirids nor any other pterodactyloids had such a surface. The head may have acted in a similar way to a fin (this would certainly have been the case for elaborately crested forms), but because it was situated ahead of the c.g., a sideforce produced in response to a yaw would have been destabilizing. A second means of achieving yaw stability is to deflect the wingtips downward about an oblique axis, so that they act in a similar way to wingtip fins (Weyl, 1945; Pennycuick, 1971a). These so-called diffuser wingtips can generate small sideforces to oppose a yaw disturbance. The washout of the cheiropatagium would have given a similar geometry, and the greater the degree of twist in the membrane, the more significant this effect would have been. In terms of performance, however, an excessively twisted membrane would have been disadvantageous, and for this reason the diffuser wingtip effect may not have been important. In addition, the effect would have depended on there being only a small degree of aeroelastic deflection of the distal wing finger, which may not have been the case. FLIGHT CONTROL The extended configuration shown in Figure 18 represents only one of many possible wing configurations that would have been used during gliding or soaring flight to turn, adjust the flight speed, or change the angle of descent. Additionally, adjustments would have been required in order to respond to routine changes of mass, and the position of the centre of mass, such as those caused by feeding. Furthermore, given that the ornithocheirids were probably unstable in pitch and yaw, any such deviation caused by an atmospheric disturbance would have to have been corrected actively. The following section will describe the basic requirements for flight control, and examine how these requirements may have been realized by the ornithocheirids, given the geometric constraints of their wing skeleton. Speed/pitch control During steady (i.e. equilibrium) flight, the weight must be balanced by the total aerodynamic force produced by the wings and tailplane (if present). To a first approximation, the magnitude of the resultant aerodynamic force of the wings depends on their angle of attack, camber, and area, and also on the relative airspeed. Hence, by altering one or more of the wing configuration factors symmetrically (i.e. identically for both wings), a new equilibrium will be assumed with an aerodynamic force of the same magnitude, but at a different speed. A discussion of speed control in the ornithocheirids is therefore a discussion of how wing camber, area, and angle of attack could be altered in order to establish a new equilibrium. The equilibrium angle of attack – i.e. the angle of attack at which the overall pitching moment about the c.g. is zero – essentially depends on two factors: the pitching moment when the lift is zero, M0, and the change in pitching moment with angle of attack, dM/da. Stability requires that dM/da must be negative, and hence M0 must be positive (nose up) for M to be zero at a positive angle of attack; when unstable, dM/da is positive and M0 must be negative (nose down) to balance at a positive angle of attack (Fig. 21A). The sign and magnitude of dM/da depends on the separation of the c.g. and m.a.c.: it increases as the m.a.c. moves forwards with respect to the c.g., and is positive if the m.a.c. lies ahead of the c.g. (dM/da also depends on the flexibility of the wing, but this phenomenon will not be considered here). M0 depends partly on wing camber: as camber increases, M0 decreases, i.e. becomes increasingly negative, as a result of the increased inherent nose-down moment of the wing profile. Conversely, M0 is increased by the addition of a negative-incidence tail, washout with sweepback, etc. The stability paradigm has a profound influence on pitch control, as the following example will illustrate. For conventional, stable aircraft, the equilibrium angle of attack is usually increased by raising the elevators, situated on the tailplane, which increases M0, thereby causing a nose-up pitching moment that brings the aircraft to the new equilibrium (Fig. 21B). For unstable aircraft, raising the elevators still increases M0, but this © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE A A M unstable 55 M 1 2 α α stable B B M M 2 1 1 α C M 2 1 α Figure 21. A, typical plots of pitching moment M against angle of attack a (measured with respect to the zero-lift angle of attack) for stable and unstable aircraft. Both aircraft are shown balancing at the same equilibrium angle of attack, at which point the pitching moment is zero. B, the effect of raising the elevators (1) on the longitudinal balance of a stable aircraft: the pitching moment is increased, as is the equilibrium angle of attack, so the aircraft pitches up (2) until equilibrium is restored. C, the effect of raising the elevators (1) on the longitudinal balance of an unstable aircraft: the pitching moment is increased, but the equilibrium angle of attack is reduced, so the aircraft now pitches up away from equilibrium (2). causes a decrease in the equilibrium angle of attack, giving rise to a nose-up moment (Fig. 21C): a new equilibrium is set, but cannot be reached using the elevators alone. For the ornithocheirids, the equilibrium angle of attack could have been adjusted by shifting the m.a.c. with respect to the c.g., which would have altered dM/da, or by changing the camber of the wings, which would have altered M0. For example, sweeping the wings backwards would have decreased dM/da (Fig. 22A), and increasing the camber would have decreased M0 (Fig. 22B). The situation is actually 2 α Figure 22. Increasing the equilibrium angle of attack of an unstable aircraft can be brought about by decreasing the derivative dM/da, where M is the pitching moment and a is the angle of attack, by sweeping the wings back (A), or by decreasing the zero-lift pitching moment M0, by depressing the pteroids for example (B). These adjustments can theoretically be used in response to an unstable nose-up pitch (1), thus establishing a new equilibrium (2). more complicated than this thanks to the washout of the wings: sweeping the twisted wings backwards would increase M0 as well as decrease dM/da; also, the magnitude of twist itself may have been affected by the sweep, camber, and camber distribution, but these effects will not be considered yet. Assuming pitch instability, a decrease in dM/da and M0 would have increased the equilibrium angle of attack, as shown in Figure 22. However, neither would have brought the pterosaur to the new equilibrium – in fact, both would have given rise to a nose-down moment. In order to pitch up, the wings could have been swept forwards, but this would have led to a decrease in the equilibrium angle of attack, and a consequent loss of balance (this could have been useful for certain transitory manoeuvres, such as landing, but would have been inappropriate for steady flight). Alternatively, the angle of attack of the wings could have been increased directly. This direct control of the angle of attack would not be effective in a stable paradigm, for any increase in the angle of attack would be followed by a nose-down pitch: only when unstable does an increased angle of attack increase the pitching moment. In birds and bats, the angle of attack of the entire wings can be adjusted by pronation or supination at © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 56 M. T. WILKINSON the shoulder. This is only possible, however, because there are rigid beams cantilevered on the wing spar that run aft to the trailing edge, these being the elongate fourth and fifth digits in bats, and the feather shafts of birds. Pterosaur wings possessed no such structures. There was a system of fine structural fibres (actinofibrils) within the distal cheiropatagium that radiated in a posterodistal pattern, but these did not attach to the wing spar, and had insufficient bending stiffness to act as cantilevers (Bennett, 2000). Nevertheless, because the wing finger is curved backwards, supination at either the shoulder or the carpometacarpal joint (if possible) would have caused depression of the trailing edge near the wingtip, and would therefore increase the angle of attack there. The magnitude of this effect, and the extent to which it propagated towards the root, would have depended on the stiffness and tension of the actinopatagium, the latter being controllable to an extent through wing retraction. Another means of altering the angle of attack was elevation and depression of the leg at the hip (Fig. 23A, B). This may only have affected the inner wing: given the high aspect ratio, the actinopatagium may have been too remote to be controlled in this fashion. If the trailing edge of the cheiropatagium ran to the ankle, the angle of attack of the inner wing could also have been adjusted by pronation and supination of the femur (Fig. 23A, C), because these movements would have elevated and depressed the distal end of the tibiotarsus, respectively. Because the tibiotarsus projects backwards further than the femur, this would have had a more significant effect on the angle of attack than femoral elevation or depression. Interestingly, pterosaurs appear to have had almost independent control of the angle of attack of the proximal and distal regions of the wings: the legs could have controlled the former, whereas pronation or supination of either the entire wing spar or the wing finger and metacarpus would have controlled the latter. Directly altering the angle of attack of the wings would have resulted in a loss of balance in an unstable paradigm, and its restoration would have Figure 23. Methods of changing the angle of attack of the inner wing. A, left lateral view of Anhanguera configured as in Fig. 18. The broken line indicates the chord line. The angle of attack of the section is zero. B, as in (A), but with the left leg depressed at the hip by 30°. The geometric angle of attack a is indicated. C, as in (A), but with the leg supinated at the hip by 30°. The leg movements shown in (B) and (C) increase both the angle of attack and the camber of the inner wing. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE depended upon the establishment of a new equilibrium angle of attack. The pitch derivative dM/da could have been decreased to increase the equilibrium angle of attack in response to a nose-up pitch by sweeping the wings backwards (Fig. 22A), using secondary angulation at the shoulder or flexion at the intersyncarpal joint or knuckle, perhaps coupled with a compensatory retraction of the leg to preserve membrane tension. The opposite movements could have decreased the equilibrium angle of attack in response to a nose-down pitch. However, a cautionary note should be sounded, because, as mentioned above, the washout of the wings complicates matters considerably: a backward sweep increases M0, whereas a forward sweep decreases M0, which would have had the opposite effect on the equilibrium angle of attack to the corresponding changes in dM/da. Furthermore, sweeping the wings back may have increased the washout resulting from a reduction of spanwise tension that, if uncompensated for, would have increased M0 still further. Without quantitative windtunnel data, it is therefore very difficult to predict the net effect of sweeping the wings backwards or forwards, but the use of these movements in the way just described is counter-intuitive. One would expect a protraction of the wings to be used to pitch up (albeit in an unbalanced fashion), and not to respond to a pitch down. Alternatively, M0 could have been decreased in order to establish a higher angle of attack in response to a pitch up (Fig. 22B), by depressing the cruropatagium using the tail and/or increasing the camber of the wings. Conveniently, this would also have increased the maximum lift, thereby contributing directly to the speed reduction brought about by the increased equilibrium angle of attack. A camber 57 increase would in fact have occurred automatically if depression of the legs were used to increase the angle of attack, because the leading edge was fixed by the presence of the propatagium (Fig. 23B, C). An increase in wing camber could also have been readily achieved by depressing the pteroid at the carpopteroid joint (Fig. 24). It is important to bear in mind that this movement would not only have increased the camber: the accompanying depression of the leading edge of the inner wing with respect to its trailing edge would also have decreased its angle of attack relative to the incoming airflow (Fig. 24B). By this token, one might presume that depressing the pteroid alone was not an effective means of reducing speed. However, the increased camber caused by pteroid depression would also have decreased the zero-lift angle of attack: hence the effective angle of attack, measured relative to the zero-lift angle, might have been increased by this movement (a more rigorous analysis is required to say for sure, but this is beyond the scope of the present work). The possibility also exists that the increased drag force associated with deflection of the propatagium could have caused a nose-up moment, if it acted above the c.g. Nevertheless, control authority would doubtless have been improved if pteroid depression were coupled with or preceded by leg depression. As a further complicating factor, propatagium deflection alone would have decreased the washout of the wing, for the resulting decrease in the angle of attack would only have occurred in the proximal part of the wing, because the propatagium was absent from the distal part. It is possible, however, that the accompanying leg depression, which would have increased the angle of attack of the proximal part of the wing, negated this effect. A B α Figure 24. Effects of pteroid depression. A, left lateral view of Anhanguera as in Fig. 23. B, as in (A), but with the left pteroid depressed by 30°. The camber is increased and the angle of attack, a, is reduced by this movement. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 58 M. T. WILKINSON Camber may also have been increased by retracting the femur, as suggested by Bennett (2000), and/or flexing the tibiotarsus (if the cheiropatagium ran to the crus). Both of these movements would have shifted the rear proximal margin of the cheiropatagium medially. The distance between the hip and the wingtip would, however, have remained constant (assuming, for the sake of simplicity, no other wing adjustments took place): hence, the spanwise strain would have increased from the mid-chord region to the trailing edge. Because the increased spanwise tension would have been largely limited to the posterior part of the cheiropatagium, the result would have been a decreased washout and increased camber [this latter point was not noted by Short (1914), who believed that leg retraction would have decreased the camber]. However, the reduced washout may also have increased the risk of tip stall, because tapered, untwisted wings exhibit a spanwise increase in the effective angle of attack caused by diminished downwash at the tips. Hence, this method of camber control may not have been used. It is more likely that leg retraction and protraction were used in a compensatory manner in response to secondary angulation at the shoulder, as described above. Flight speed could have been increased by flexing the elbow, wrist, and knuckle to reduce the area of the wings. Such an adjustment has been documented in birds (Pennycuick, 1960, 1988; Parrott, 1970; Tucker & Parrott, 1970; Pennycuick, 1971b; McGahan, 1973; Tucker & Heine, 1990; Rosén & Hedenström, 2001) and bats (Pennycuick, 1971a), but the degree to which it could have been used by pterosaurs is uncertain. In birds, wing flexion is straightforward because the feathers can overlap; hence, a significant reduction in area is possible with no disruption of the flight surface. Bats cannot flex their wings to the same degree in flight (Pennycuick, 1971a), because the elastic wing membranes would slacken and flutter, making flight impossible, or at least very inefficient. Hence, the minimum wingspan in flight is no less than about 85% of the maximum (Pennycuick, 1971a). Without the additional membrane-supporting digits of bat wings, it is logical to conclude that pterosaurs were even more restricted in this regard, although the stiffening actinofibrils, acting in conjunction with the contraction of intrinsic membrane muscles, may have offered some scope for wing flexion (Bennett, 2000). However, it is unlikely that the maximum degree of wing retraction in pterosaurs was greater than that in bats, for the reasons given above. I therefore regard the bat limit of 85% as a minimum estimate of the ornithocheirid limit. Flexion of the elbow, radioulnocarpal, and knuckle joints all by 25° brings the wingspan to this value (Fig. 25). It has been suggested that the pteroid (on the assumption that it articulated in the distal cotyle of the medial carpal) could ‘snap’ from a medial orientation to an anteroventral orientation, thereby transforming the wing from a shallow-cambered, small-area state to a deep-cambered, large-area state (Pennycuick, 1988). This switch would have instantly changed the wing from a high-speed to a low-speed configuration. However, this adjustment is not supported by the fossil evidence. As stated above, when fully flexed (and articulating in the distal cotyle of the medial carpal) the pteroid points ventromedially, not medially. The pteroid must first be disarticulated to permit a medial orientation. Therefore, this rather abrupt form of speed control could not have been used. To summarize, an increase in speed could have been brought about by elevating the legs and pronating the wing spar at the shoulder, and/or carpometacarpal joint, to decrease the angle of attack, and by raising the pteroid to decrease the camber of the wings – an effect augmented by the elevated legs – to establish a new equilibrium. Alternatively, or additionally, the wing area could have been reduced by flexion at the elbow, wrist, and knuckle. Sweeping the Figure 25. Ventral view of Anhanguera, with the elbow, radioulnocarpal, and knuckle joints flexed by 25° from their close-packed positions. The wingspan is 85% of the maximum. Scale bar: 500 mm. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE wings backwards would have caused a nose-down moment and a concordant increase in speed, but in an unbalanced fashion, unless the decrease in dM/da caused by the sweep-back alone was more than compensated by an increase in M0 caused by the sweepback plus washout, or decreased camber. A decrease in speed was probably effected by lowering the legs and cruropatagium (using the tail), supinating the wing, and depressing the pteroid. Sweeping the wings forwards would have caused a nose-up moment, but, again, balance would have been lost unless M0 was decreased to compensate. When coming in to land, the angle of attack must have been very high, because the legs would have been held quite far below the wing spar in order to contact the ground (Fig. 26). At such a time, the pteroid (assuming it articulated with the distal cotyle of the medial carpal) could have been depressed to align the propatagium with the airflow, like a leading edge flap. Wind-tunnel tests of model wing profiles that incorporated a broad propatagium demonstrated that the angle of attack at which stalling took place could be increased by this means (Wilkinson et al., 2006). As a result, the maximum lift coefficient (the lift force standardized with respect to relative air velocity and wing area) was greatly increased, which, coupled with the increased area of the broad propatagium, would have substantially reduced the pterosaur’s minimum flight speed. The high-lift effect associated with a forward-pointing pteroid may have been important in facilitating landing in giant pterosaurs. It should be noted that the wind-tunnel model from which high lift coefficients were obtained, some 45% higher than for a model without a propatagium, was not a model of the Figure 26. Left lateral view of Anhanguera in a possible landing configuration. The body is pitched up by 20°, the femur is supinated by 30° and depressed by 30°, and the pteroid is depressed by 30°. The geometric angle of attack of the wing section is 40°. 59 entire wing, but of a cross section taken between the wrist and knuckle (Wilkinson et al., 2006). Nevertheless, the high-lift effect would still have been significant for the complete wing, because, as explained above, the propatagium probably extended from the wing root to the knuckle, a region of the wing that, in ornithocheirids, represented over half its projected wing area (Wilkinson, 2002). The wind-tunnel tests described above also indicated that the drag coefficient was very high when in a landing configuration (Wilkinson et al., 2006), but this would have been a distinct advantage, as the high drag would have decelerated the pterosaur and steepened the approach path, thereby allowing a safer, more controlled touchdown. Indeed, it is known that vultures lower their legs when coming in to land for precisely this purpose (Pennycuick, 1971c; Tucker, 1988). Essentially, the forward-pointing pteroid, if achievable, would have enabled the propatagium to act not only as a very effective leading-edge flap, but also as an air brake. The need to attain very high angles of attack when landing is perhaps the most persuasive argument against inherent pitch stability in the ornithocheirids, and possibly in all pterodactyloid pterosaurs. If stable, any attempt to lower the legs directly without other adjustments would have led to an immediate nose-down pitch, so an alternative means of increasing the equilibrium angle of attack would have been required. Increasing the camber of the wings would have been useless (in fact, detrimental) for this purpose, as the resulting decrease in M0 would have caused the equilibrium angle of attack to decrease in a stable paradigm. Given sufficient washout, sweeping the wings far back may have been helpful, as this adjustment would have increased M0, and with it the equilibrium angle of attack, but it would also have decreased dM/da, which would have had the opposite effect. The substantial washout would also have compromised lift production. Finally, one might suppose that the cruropatagium and tail could have been used as an elevator. According to Bennett (2001), the tail was embedded within the cruropatagium, and so could indeed have elevated the membrane to increase M0, although the control authority of this small membrane, situated as it was very close to the c.g. and in the wake of the body, cannot have been great. If, in contrast to Bennett (2001), the tail was not embedded within the cruropatagium, but instead lay dorsal to it, as in the ‘rhamphorhynchoid’ S. pilosus (Unwin & Bakhurina, 1994), the tail could only have been used to lower the membrane, and thus decrease M0. Roll control In order to turn in flight, the wings must be banked so that the lift has a horizontal component to provide © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 60 M. T. WILKINSON the required centripetal force. The animal must therefore roll in the direction of the turn until the correct angle of bank is attained. This could have been achieved in the ornithocheirids by increasing the lift of the outside wing by depressing the leg, supinating the wing at the shoulder and/or carpometacarpal joint (if possible), and depressing the pteroid (if possible), and/or decreasing the lift of the inside wing by elevating the leg, pronating the wing, or flexing the elbow, wrist, and knuckle, as described above. The most effective means of generating a large rolling moment would have been by using the distal parts of the wings, as these were furthest from the c.g. Hence, the carpometacarpal joint, which could have pronated or supinated the outer wing largely independently of the inner wing, may have been particularly useful for this purpose. Yaw control Yawing rotations are not usually required for the direct initiation of a turn: as described above, a change of direction in flight is always preceded by a roll, unless manoeuvring very close to the ground or water, when banking could cause one wing to collide with the surface. Yaw control is still important for normal turning flight, however, because roll and yaw are coupled, and any undesirable yaw movements caused by a roll (adverse yaw) would need to have been corrected. Specifically, the increased lift of the ascending wing would have been accompanied by an increased drag on the same wing, which if uncorrected would have caused the pterosaur to yaw in the opposite direction to the roll. There are two ways by which an ornithocheirid pterosaur could have produced a yaw moment: by turning its head or increasing the drag of one wing with respect to the other. Even though the ornithocheirids lacked the bizarre sail-like crests of some Cretaceous pterosaurs (Frey & Tischlinger, 2000; Frey et al., 2003b), their heads would still have produced significant sideforces when inclined to the flight path, as suggested by Bramwell & Whitfield (1974). Because the head was situated ahead of the c.g., turning it to the left would have caused a yaw to the left. Hence, a pterosaur may have been able to correct an adverse yaw during a roll simply by looking where it was going. Failing this, a yaw moment could have been produced by carefully manipulating the angle of attack and camber of the proximal wing in opposite directions using the leg and pteroid, which could conceivably have caused the lift on the downgoing wing to be held constant but the drag to be increased. Alternatively, if a roll were initiated by retraction of one wing at the elbow, wrist, and knuckle to decrease its lift, the accompanying decrease in spanwise tension would, if left uncompen- sated, have increased the drag of the same wing through increased twist, and maybe even aeroelastic flutter. In this latter case, adverse yaw would have been immediately addressed without additional control movements. TERRESTRIAL LOCOMOTION The way pterosaurs moved on the ground has been one of the most contentious aspects of their palaeobiology for many years (e.g. Padian, 1983; Wellnhofer, 1988; Padian, 1991; Bennett, 1997a, b; Unwin, 1997; Bennett, 2001). We are now approaching a consensus with regard to the terrestrial locomotion of smaller pterodactyloids, thanks mainly to the discovery of many well-preserved trackways, assigned to the ichnogenus Pteraichnus, that are stratigraphically and morphologically consistent with this kind of pterosaur (Stokes, 1957; Lockley et al., 1995; Mazin et al., 1995, 1997, 2003; Bennett, 1997b; Unwin, 1997). These show that the pterosaur trackmakers were quadrupeds with plantigrade pedes, digitigrade manus, and erect/semi-erect hindlimbs. Unfortunately, the track record for large pterodactyloids is much less extensive than that for the smaller forms. The best-preserved traces, from the Purbeck Limestone Formation (Lower Cretaceous) of southwestern England (Wright et al., 1997) and the Uhangri Formation (Upper Cretaceous) of southwestern Korea (Hwang et al., 2002), have been assigned, respectively, to two ichnotaxa: Purbeckopus and Haenamichnus, although the pterosaurian origin of the first of these has been called into question (Padian, 2003). These trackways, like Pteraichnus, show that the makers were moving quadrupedally. From this limited ichnological evidence one would conclude that large pterosaurs, like their smaller relatives, were predominantly erect/semi-erect plantigrade quadrupeds, excepting the period immediately after landing, when a bipedal posture must have been transiently adopted while the forelimbs were still deployed for flight, as has now been confirmed by a trackway (Padian, Mazin & Billon-Bruyat, 2003), and possibly before take-off when the same may have been true. However, the precise nature of the quadrupedal stance is still under debate. Wellnhofer (1988) proposed that the limbs were semi-erect, and that as a result the hands were more widely separated than the feet. This model is qualitatively consistent with the Purbeckopus tracks, in which the manus prints often lie outside the pes prints (Wright et al., 1997), although these tracks may not be pterosaurian (Padian, 2003). This is not the case for Haenamichnus, however: in these trackways the manus and pes prints are nearly in line (Hwang et al., 2002). This latter track morphology is consistent with recent © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE models of quadrupedal locomotion developed by Henderson & Unwin (1999) and Chatterjee & Templin (2004). In these, both fore- and hindlimbs are held nearly erect, and swing back and forth parasagittally. The reconstruction of Chatterjee & Templin (2004) is rather similar to a model developed for small pterodactyloids on the basis of the Crayssac tracks (Mazin et al., 2003): the spine is held at an angle of about 45° to the horizontal, and the humeri are directed downwards. By contrast, Henderson & Unwin (1999) propose that the spine was elevated to a greater degree, and that the humeri were retracted, not depressed. All three of the quadrupedal models described above were developed using ornithocheirid material, and it would therefore be instructive to see which are supported by the three-dimensional reconstruction of the present work. Two possible quadrupedal stances of A. santanae, both consistent with the limitations imposed by its joints, are shown in Figures 27 and 28. In both postures the elbow, radioulnocarpal, and knuckle joints are at their respective limits of flexion, the intersyn- 61 carpal joints are extended, and the carpometacarpal joints are fully supinated. The femora are depressed by 60°, and the pedes are separated by c. 0.2 m. In Figure 27 the shoulders are close-packed, whereas in Figure 28 the humeri are maximally depressed and protracted, which has the effect of reducing the manus separation from c. 0.6 m (3 ¥ pes separation) to c. 0.3 m (1.5 ¥ pes separation). Of the two postures, I regard Figure 27 as the more feasible, because adduction of the forelimbs as shown in Figure 28 causes the wing fingers to splay outwards, which would have made them more vulnerable to damage. It is therefore likely that the manus were usually quite widely separated, although an intermediate degree of adduction cannot be definitely ruled out. Pronation at the carpometacarpal joints has a similar effect on the position of the wing fingers as depressing and protracting the humeri. Of the three quadrupedal models described above, the semi-erect model presented by Wellnhofer (1988) is the most similar to that presented here, in that the manus are widely separated and the spine subhori- Figure 27. Dorsal (A), anterior (B), and left lateral (C) views of Anhanguera in a semi-erect quadrupedal stance. The shoulder and intersyncarpal joints are in their close-packed positions, the carpometacarpal joints are supinated, and the elbow, radioulnocarpal, knuckle, and carpopteroid joints are maximally flexed. The femora are depressed by 60°. Scale bar: 500 mm. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 62 M. T. WILKINSON zontal. The models of Chatterjee & Templin (2004) and Henderson & Unwin (1999), are problematic from an osteological perspective. Chatterjee & Templin (2004) thought that the humerus could be depressed below the horizontal by 60°, but an angulation of this magnitude causes complete disarticulation of the joint. The fully retracted position of the humerus reconstructed by Henderson & Unwin (1999), on the other hand, lies within the range of shoulder motion proposed in the present work. However, the humerus can only take this position if it is supinated, which also supinates the elbow axis. When the elbow is flexed, the forearm therefore points in entirely the wrong direction for manus–ground contact (Fig. 29). This emphasizes the importance of correctly assessing how the orientations of neighbouring joint axes relate to each other when carrying out whole-skeleton reconstructions. It would be useful to compare the reconstructed terrestrial stance with the ichnological evidence. Unfortunately, the dimensions of the prints of Purbeckopus, even if these are pterosaurian, and Haenamichnus are not consistent with an ornithocheirid trackmaker. Of particular importance in drawing this conclusion is the small size of the ornithocheirid pedes. Of the specimens I examined, C. robustus (NSM-PV 19892) has the best-preserved pes, with complete digits II and III, and their respective metatarsals. The estimated pes length is only 100 mm, which is very small given the estimated wingspan of 5.7 m (assuming it was geometrically similar to A. santanae, as the morphometric analysis indicates: see Appendix 2). The pes is also smaller than the manus: in C. robustus (NSM-PV 19892) manual digit III alone is 10 mm longer than the entire pes. The largest pes prints of Purbeckopus and Haenamichnus measure 225 mm and 350 mm in length, respectively. If it is assumed that these prints were roughly similar in length to the pedes that made them, ornithocheirid trackmakers geometrically similar to A. santanae would have had wingspans of 13 and 20 m, respectively, which seem excessive. In addition, the manus prints of Purbeckopus and Haenamichnus are usually similar in length or slightly shorter than associated pes prints (Wright et al., 1997; Hwang et al., 2002): the manus prints of an ornithocheirid track would be expected to be significantly larger than the pes prints. I would therefore argue that, in all likelihood, verifiable ornithocheirid tracks have yet to be discovered. The idea that different pterosaur taxa could have Figure 28. As in Fig. 27, but with the humeri at the limit of depression and protraction, and the elbows set 10° short of the limit of flexion. Scale bar: 500 mm. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE 63 Figure 29. Dorsal (A), anterior (B), and left lateral (C) views of Anhanguera in an upright bipedal stance. The humeri are retracted by 65°, requiring a maximal supination of 50°, and the other arm joints are maximally flexed. made tracks of different gross form is interesting, and demonstrates that care must be taken when attempting to fit three-dimensional reconstructions to trackways. It is not universally accepted that large pterosaurs were habitually quadrupedal. Some workers, most notably Bennett (1990, 1997b, 2001), have argued that the great disparity between the forelimb and hindlimb lengths in large forms would have precluded quadrupedal locomotion. Bennett (1990) therefore proposed that large pterodactyloids were bipedal with erect hindlimbs, as had been argued in the heyday of the bird-like/bat-like controversy (Padian, 1983, 1991). Additionally, Bennett (1990) suggested that the spine was elevated 60° from the horizontal, arguing, as had Pennycuick (1988), that if the spine had been subhorizontal as suggested by Padian (1983), the long neck and skull typical of large pterodactyloids would have caused them to topple forwards. Figure 30 shows A. santanae in such a posture, with bird-like subhorizontal femora, the shoulders in their closepacked positions, the elbow, radioulnocarpal, and knuckle joints at their respective limits of flexion, the intersyncarpal joints extended, and the carpometacarpal joints fully supinated. The objections raised by Pennycuick (1988) and Bennett (1990) are borne out: the pterosaur would be markedly front-heavy in this posture. The upright bipedal model proposed by Bennett (1990) is of course inconsistent with trackways that indicate quadrupedal locomotion, although this ichnological evidence only indicates that certain large pterosaurs were quadrupedal some of the time, and not all pterosaurs all of the time. Bennett’s (1990, 1997b, 2001) chief objection may also be unfounded: it does not necessarily follow that a disparity in length between the fore- and hindlimbs should make quadrupedal locomotion prohibitively difficult. There are or were quadrupeds with longer forelimbs than hindlimbs (the sauropod Brachiosaurus and small pterodactyloids, for example), and a human with a pair of crutches can walk quadrupedally quite easily, despite the great difference between the lengths of crutch and leg. Most importantly, however, the full retraction of the wings indicated in the bipedal models of Padian (1983, 1991) and Bennett (2001) is unfeasible, as Figure 29 clearly shows. The wings could not have been fully furled against the body, making habitual bipedal progression awkward. CONCLUSIONS On the basis of my analysis of the three-dimensional geometry and range of movement of the ornithocheirid skeleton, I propose that: © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 64 M. T. WILKINSON or some other camber-adjustment mechanism. Sweeping the wings fore and aft at the shoulder and/or knuckle would also have pitched the pterosaur up and down, respectively, but in an unbalanced fashion. Additionally, speed could have been increased by flexing the wing at the elbow, wrist, and/or knuckle. 4. Roll control was probably brought about by altering the angle of attack of the distal part of one wing by rotating at the shoulder and/or carpometacarpal joint, and/or changing the area of one wing using flexion/extension at the elbow, wrist, and knuckle joints. 5. Yaw control was probably achieved by turning the head, or by careful manipulation of the pteroid and hindlimb or membrane tension to cause a drag asymmetry. 6. When on the ground, it is likely that the ornithocheirids adopted a semi-erect quadrupedal posture, with subhorizontal spine and humeri. Typical manus separation is predicted to be about three times the pes separation. ACKNOWLEDGEMENTS Figure 30. Left lateral view of Anhanguera in a bird-like bipedal posture, with subhorizontal femora. The shoulder and intersyncarpal joints are in their close-packed positions, the carpometacarpal joints are supinated, and the elbow, radioulnocarpal, knuckle, and carpopteroid joints are maximally flexed. The femora are depressed by 60°. Scale bar: 500 mm. 1. The pteroid was directed forwards in flight, not towards the body. The propatagium would therefore have functioned as an adjustable leading edge flap and high-lift device, delaying aerodynamic stall during landing, when the angle of attack was unavoidably high because of the attachment of the main wing membrane to the legs. 2. The ornithocheirids were, in all likelihood, inherently unstable in pitch and yaw, largely because of the minimal backward sweep or even slight forward sweep of the wings when extended, and also because of the need to attain very high angles of attack when landing. 3. Pitch and therefore speed control were probably brought about by changing the angle of attack of the inner region of the wings, using the hindlimbs and/or the outer region of the wings, using pronation and supination of the shoulder and carpometacarpal joints. Increasing and decreasing the angle of attack would have caused a nose-up and nose-down pitch, respectively. Following such an adjustment, longitudinal balance could have been restored using depression/elevation of the pteroid, I thank Makoto Manabe (NSM), Satoru Nabana (IMCF), Yuji Takakuwa (Gunma Prefectural Museum, Japan), Eberhard ‘Dino’ Frey (SMNK), Mark A. Norell, and Eugene Gaffney (AMNH) for granting access to the fossil material, and further thank M. Manabe, E. Frey, and David M. Unwin (Museum für Naturkunde der Humboldt-Universität, Berlin, Germany) for the loan of specimens and casts. I also thank the above, and additionally Yoko Kakegawa, Hiroya Ogata, and Yukimitsu Tomida of the NSM, and Steve Salisbury, Alex Anders, and Rene Kastner of the SMNK for their generous assistance during my museum visits, and also the Perez family for providing accommodation for the duration of my stay in New York. I am also very grateful to Charles P. Ellington, D. Unwin, Colin Palmer, and James Cunningham for their helpful comments on the manuscript, and for many interesting discussions about pterosaur functional morphology and flight. Funding for the research and travel was provided by the Biotechnology and Biological Sciences Research Council (BBSRC), Queens’ College, University of Cambridge and Clare College, University of Cambridge. I am currently supported by a Junior Research Fellowship at Clare College, University of Cambridge. REFERENCES Alexander RM. 1989. Mechanics of fossil vertebrates. Journal of the Geological Society 146: 41–52. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE Alexander RM. 1998. All-time giants: the largest animals and their problems. Palaeontology 41: 1231–1245. Barnard RH, Philpott DR. 1995. Aircraft flight: a description of the physical principles of aircraft flight, 2nd edn. Harlow: Longman. Bennett SC. 1990. A pterodactyloid pterosaur pelvis from the Santana formation of Brazil: implications for terrestrial locomotion. Journal of Vertebrate Paleontology 10: 80–85. Bennett SC. 1993. The ontogeny of Pteranodon and other pterosaurs. Paleobiology 19: 92–106. Bennett SC. 1997a. The arboreal leaping theory of the origin of pterosaur flight. Historical Biology 12: 265–290. Bennett SC. 1997b. Terrestrial locomotion of pterosaurs: a reconstruction based on Pteraichnus trackways. Journal of Vertebrate Paleontology 17: 104–113. Bennett SC. 2000. Pterosaur flight: the role of actinofibrils in wing function. Historical Biology 14: 255–284. Bennett SC. 2001. The osteology and functional morphology of the Late Cretaceous pterosaur Pteranodon. Palaeontographica A 260: 1–153. Bennett SC. 2003. A survey of pathologies of large pterodactyloid pterosaurs. Palaeontology 46: 185–198. Bennett SC. 2006. Articulation and function of the pteroid bone of pterosaurs. Journal of Vertebrate Paleontology 26: 41A. Blankevoort L, Huiskes R, de Lange A. 1988. The envelope of passive knee joint motion. Journal of Biomechanics 9: 705–720. Boyd SK, Ronsky JL. 1998. Instantaneous moment arm determination of the cat knee. Journal of Biomechanics 31: 279–283. Bramwell CD. 1971. Aerodynamics of Pteranodon. Biological Journal of the Linnaean Society 3: 313–328. Bramwell CD, Whitfield GR. 1974. Biomechanics of Pteranodon. Philosophical Transactions of the Royal Society of London B 267: 503–581. Broili F. 1938. Beobachtungen an Pterodactylus. Sitzungberichte der Bayerischen Akademie der Wissenschaften, Mathematisch-Naturwissenschaftliche Abteilung 1938: 139– 154. Brower JC. 1983. The aerodynamics of Pteranodon and Nyctosaurus, two large pterosaurs from the Upper Cretaceous of Kansas. Journal of Vertebrate Paleontology 3: 84–124. de Buisonjé PH. 1980. Santanadactylus brasilensis nov. gen., nov. sp., a long-necked, large pterosaurier from the Aptian of Brasil. Proceedings, Koninklijke Nederlandse Akademie van Wetenschappen B 83: 145–172. Carrier DR, Farmer CG. 2000. The evolution of pelvic aspiration in archosaurs. Paleobiology 26: 271–293. Chatterjee S, Templin RJ. 2004. Posture, locomotion, and paleoecology of pterosaurs, Special Paper 376. Boulder, CO: Geological Society of America. Claessens L, Unwin DM, O’Connor P. 2006. Evolution of the respiratory apparatus and breathing mechanisms in Pterosauria. Journal of Vertebrate Paleontology 26: 50A. Codorniú L, Chiappe LM. 2004. Early juvenile pterosaurs (Pterodactyloidea: Pterodaustro guinazui) from the Lower 65 Cretaceous of central Argentina. Canadian Journal of Earth Sciences 41: 9–18. Del Rossi G. 2003. Great toe pain in a competitive tennis athlete. Journal of Sports Science and Medicine 2: 180–183. Etkin B, Reid LD. 1996. Dynamics of flight: stability and control. New York: John Wiley & Sons, Inc. Fastnacht M. 2005. The first dsungaripterid pterosaur from the Kimmeridgian of Germany and the biomechanics of pterosaur long bones. Acta Palaeontologica Polonica 50: 273–288. Fink MP. 1969. Full-scale investigation of the aerodynamic characteristics of a sailwing of aspect ratio 5.9. NASA TN D-5047. Frankel VH, Burstein AH, Brooks DB. 1971. Biomechanics of internal derangement of the knee. Journal of Bone and Joint Surgery 53A: 945. Frey E, Buchy M-C, Martill DM. 2003a. Middle- and bottom-decker Cretaceous pterosaurs: unique designs in active flying vertebrates. In: Buffetaut E, Mazin J-M, eds. Evolution and palaeobiology of pterosaurs, Special Publications 217. London: Geological Society, 267–274. Frey E, Tischlinger H, Buchy M-C, Martill DM. 2003b. New specimens of Pterosauria (Reptilia) with soft parts with implications for pterosaurian anatomy and locomotion. In: Buffetaut E, Mazin J-M, eds. Evolution and palaeobiology of pterosaurs, Special Publications 217. London: Geological Society, 233–266. Frey E, Martill DM. 1994. A new pterosaur from the Crato Formation (Lower Cretaceous, Aptian) of Brazil. Neues Jahrbuch für Geologie und Paläontologie Abhandlungen 194: 379–412. Frey E, Riess J. 1981. A new reconstruction of the pterosaur wing. Neues Jahrbuch für Geologie und Paläontologie Abhandlungen 161: 1–27. Frey E, Tischlinger H. 2000. Weichteilanatomie der Flugsaurierfüße und Bau der Scheitelkämme: Neue Pterosaurierfunde aus den Solnhofener Schichten (Bayern) und der Crato-Formation (Brasilien). Archaeopteryx 18: 1–16. Fullmer FF. 1947. Two-dimensional wind-tunnel investigation of the NACA 641-012 airfoil equipped with two types of leading-edge flap. NACA TN 1277. Hankin EH. 1912. The development of animal flight. Aeronautical Journal 16: 24–40. Hankin EH, Watson DMS. 1914. On the flight of pterodactyls. Aeronautical Journal 18: 324–335. Hazlehurst GA, Rayner JMV. 1992. An unusual flight mechanism in the Pterosauria. Palaeontology 35: 927– 941. Henderson DM, Unwin DM. 1999. Mathematical and computational model of a walking pterosaur. Journal of Vertebrate Paleontology 19: 50A. Heptonstall WB. 1971. An analysis of the flight of the Cretaceous pterodactyl Pteranodon ingens (Marsh). Scottish Journal of Geology 7: 61–78. Hill JE, Smith JD. 1984. Bats: a natural history. London: British Museum (Natural History). Hooley RW. 1914. On the ornithosaurian Ornithocheirus, with a review of the specimens from the Cambridge Green- © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 66 M. T. WILKINSON sand in the Sedgwick Museum, Cambridge. Annals and Magazine of Natural History Series 8: 529–557. Hwang K-G, Huh M, Lockley MG, Unwin DM, Wright JL. 2002. New pterosaur tracks (Pteraichnidae) from the Late Cretaceous Uhangri Formation, southwestern Korea. Geological Magazine 139: 421–435. Kellner AWA. 2004. The ankle structure of two pterodactyloid pterosaurs from the Santana Formation (Lower Cretaceous), Brazil. Bulletin of the American Museum of Natural History 285: 25–35. Kellner AWA, Tomida Y. 2000. Description of a new species of Anhangueridae (Pterodactyloidea) with comments on the pterosaur fauna from the Santana Formation (AptianAlbian), Northeastern Brazil. National Science Museum Monographs 17: 1–135. von Kripp D. 1943. Ein Lebensbild von Pteranodon ingens auf flugtechnischer Grundlage. Nova Acta Leopoldina 12: 217–246. Krus P. 1997. Natural methods for flight stability in birds. AIAA Paper AIAA 97-5653. Lawson DA. 1975. Pterosaur from the latest Cretaceous of west Texas: discovery of the largest flying creature. Science 187: 947–948. Leardini A, O’Connor JJ, Catani F, Giannini S. 1999. Kinematics of the human ankle complex in passive flexion; a single degree of freedom system. Journal of Biomechanics 32: 111–118. Lockley MG, Logue TJ, Moratalla JJ, Hunt AP, Schultz RJ, Robinson JW. 1995. The fossil trackway Pteraichnus is pterosaurian, not crocodilian: implications for the global distribution of pterosaur tracks. Ichnos 4: 7–20. Lu J-C. 2002. Soft tissue in an early Cretaceous pterosaur from Liaoning province, China. Memoir of the Fukui Prefectural Dinosaur Museum 1: 19–28. Marchaj CA. 1988. Aero-hydrodynamics of Sailing, 2nd edn. London: Adlard Coles Nautical. Marsh OC. 1882. The wings of pterodactyles. American Journal of Science 23: 251–256. Maynard Smith J. 1952. The importance of the nervous system in the evolution of animal flight. Evolution 6: 127– 129. Mazin J-M, Billon-Bruyat J-P, Hantzpergue P, Lafaurie G. 2003. Ichnological evidence for quadrupedal locomotion in pterodactyloid pterosaurs: trackways from the Late Jurassic of Crayssac (southwestern France). In: Buffetaut E, Mazin J-M, eds. Evolution and palaeobiology of pterosaurs, Special Publications 217. London: Geological Society, 281– 296. Mazin J-M, Hantzpergue P, Bassoullet J-P, Lafaurie G, Vignaud P. 1997. The Crayssac site (Lower Tithonian, Quercy, Lot, France): discovery of dinosaur trackways in situ and first ichnological results. Comptes Rendus de l’Academie des Sciences, Serie ii, Fascicule a 325: 733– 739. Mazin JM, Hantzpergue P, Lafaurie G, Vignaud P. 1995. Pterosaur trackways from the Tithonian-of-Crayssac (Quercy, France). Comptes Rendus de l’Academie des Sciences, Serie ii, Fascicule a 321: 417–424. McGahan J. 1973. Gliding flight of the Andean condor in nature. Journal of Experimental Biology 58: 225–237. Nickel KLE, Wolfahrt M. 1994. Tailless aircraft in theory and practice. Trans. by Brown EM. London: Edward Arnold. Norberg UM. 1990. Vertebrate Flight. Berlin Heidelberg: Springer-Verlag. Padian K. 1983. A functional analysis of flying and walking in pterosaurs. Paleobiology 9: 218–239. Padian K. 1984. A large pterodactyloid pterosaur from the Two Medicine Formation (Campanian) of Montana. Journal of Vertebrate Paleontology 4: 516–524. Padian K. 1991. Pterosaurs: were they functional birds or functional bats? In: Rayner JMV, Wootton RJ, eds. Biomechanics in evolution. Cambridge: Cambridge University Press, 145–160. Padian K. 2003. Pterosaur stance and gait and the interpretation of trackways. Ichnos 10: 115–126. Padian K, Mazin JM, Billon-Bruyat JP. 2003. How pterosaurs landed: the first evidence from footprints. Integrative and Comparative Biology 43: 823. Padian K, Rayner JMV. 1993. The wings of pterosaurs. American Journal of Science 293A: 91–166. Panjabi MM. 1979. Centers and angles of rotation of body joints: a study of errors and optimization. Journal of Biomechanics 12: 911–920. Panjabi MM, Goel VK, Walter SD. 1982. Errors in kinematic parameters of a planar joint: guidelines for optimal experimental design. Journal of Biomechanics 15: 537– 544. Parrott GC. 1970. Aerodynamics of gliding flight of a black vulture Coragyps atratus. Journal of Experimental Biology 53: 363–374. Pennycuick CJ. 1960. Gliding flight of the fulmar petrel. Journal of Experimental Biology 37: 330–338. Pennycuick CJ. 1971a. Gliding flight of the dog-faced bat Rousettus aegyptiacus observed in a wind tunnel. Journal of Experimental Biology 55: 833–845. Pennycuick CJ. 1971b. Gliding flight of the white-backed vulture Gyps africanus. Journal of Experimental Biology 55: 13–38. Pennycuick CJ. 1971c. Control of gliding angle in Rüppell’s griffon vulture Gyps rüppellii. Journal of Experimental Biology 55: 39–46. Pennycuick CJ. 1988. On the reconstruction of pterosaurs and their manner of flight, with notes on vortex wakes. Biological Reviews of the Cambridge Philosophical Society 63: 299–331. Prandtl L, Tietjens OG. 1957. Applied hydro- and aeromechanics. Trans. by Hartog JPD. New York: Dover Publications, Inc. Price LI. 1971. A presença de Pterosauria no Cretáceo Inferior da Chapada do Araripe Brasil. Anais do Academia Brasileira de Ciências 43: 452–461. Rewcastle SC. 1981. Stance and gait in tetrapods: an evolutionary scenario. Symposia of the Zoological Society of London 48: 239–267. Romer AS. 1970. The vertebrate body, 4th edn. Philadelphia, PA: W. B. Saunders Company. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE Rosén M, Hedenström A. 2001. Gliding flight in a jackdaw: a wind tunnel study. Journal of Experimental Biology 204: 1153–1166. Ruben JA, Jones TD, Geist NR. 2003. Respiratory and reproductive paleophysiology of dinosaurs and early birds. Physiological and Biochemical Zoology 76: 141–164. Schutt WA Jr, Altenbach JS, Chang YH, Cullinane DM, Hermanson JW, Muradali F, Bertram JEA. 1997. The dynamics of flight-initiating jumps in the common vampire bat Desmodus rotundus. Journal of Experimental Biology 200: 3003–3012. Seeley HG. 1870. The Ornithosauria: an elementary study of the bones of Pterodactyls. Cambridge: Cambridge University Press. Short GH. 1914. Wing adjustments of pterodactyls. Aeronautical Journal 18: 336–343. Simons M. 1978. Model aircraft aerodynamics. Watford: Model and Allied Publications, Argus Books Ltd. Smidt GL. 1973. Biomechanical analysis of knee flexion and extension. Journal of Biomechanics 6: 79–92. Sneyd AD. 1984. Aerodynamic coefficients and longitudinal stability of sail aerofoils. Journal of Fluid Mechanics 149: 127–146. Sneyd AD, Bundock MS, Reid D. 1982. Possible effects of wing flexibility on the aerodynamics of Pteranodon. The American Naturalist 120: 455–477. Spiegelman JJ, Woo SLY. 1987. A rigid-body method for finding centers of rotation and angular displacements of planar joint motion. Journal of Biomechanics 20: 715–721. Stokes WL. 1957. Pterodactyl tracks from the Morrison Formation. Journal of Paleontology 31: 952–954. Thomas ALR, Taylor GK. 2001. Animal flight dynamics I. Stability in gliding flight. Journal of Theoretical Biology 212: 399–424. Tucker VA. 1988. Gliding birds: descending flight of the white-backed vulture, Gyps africanus. Journal of Experimental Biology 140: 325–344. Tucker VA, Heine CE. 1990. Aerodynamics of gliding flight in a Harris’ hawk, Parabuteo unicinctus. Journal of Experimental Biology 149: 469–489. Tucker VA, Parrott GC. 1970. Aerodynamics of gliding flight in a falcon and other birds. Journal of Experimental Biology 52: 345–367. Unwin DM. 1988. New remains of the pterosaur Dimorphodon (Pterosauria: Rhamphorhynchoidea) and the terrestrial locomotion of early pterosaurs. Modern Geology 13: 57–68. Unwin DM. 1997. Pterosaur tracks and the terrestrial ability of pterosaurs. Lethaia 29: 373–386. Unwin DM. 1999. Pterosaurs: back to the traditional model? Trends in Ecology and Evolution 14: 263–268. Unwin DM. 2003. On the phylogeny and evolutionary history of pterosaurs. In: Buffetaut E, Mazin J-M, eds. Evolution and palaeobiology of pterosaurs, Special Publications 217. London: Geological Society, 139–190. Unwin DM, Bakhurina NN. 1994. Sordes pilosus and the nature of the pterosaur flight apparatus. Nature 371: 62– 64. 67 Unwin DM, Frey E, Martill DM, Clarke JB, Riess J. 1996. On the nature of the pteroid in pterosaurs. Proceedings of the Royal Society of London B 263: 45–52. Unwin DM, Henderson DM. 1999. Testing the terrestrial ability of pterosaurs with computer-based models. Journal of Vertebrate Paleontology 19: 81A. Veldmeijer AJ. 2002. Pterosaurs from the Lower Cretaceous of Brazil in the Stuttgart collection. Stuttgarter Beiträge zur Naturkunde B 327: 1–27. Veldmeijer AJ. 2003. Description of Coloborhynchus spielbergi sp. nov. (Pterodactyloidea) from the Albian (Lower Cretaceous) of Brazil. Scripta Geologica 125: 35–139. Vogel S. 1981. Life in moving fluids. Princeton, NJ: Princeton University Press. Walker PS, Shoji H, Erkman MJ. 1972. The rotational axis of the knee and its significance to prosthesis design. Clinical Orthopaedics and Related Research 89: 160–170. Wang X-L, Zhou Z-H, Zhang F-C, Xu X. 2002. A nearly completely articulated rhamphorhynchoid pterosaur with exceptionally well-preserved wing membranes and ‘hairs’ from Inner Mongolia, northeast China. Chinese Science Bulletin 47: 226–230. Wellnhofer P. 1970. Die Pterodactyloidea (Pterosauria) der Oberjura-Plattenkalke Süddeutschlands. Abhandlungen der Bayerischen Akademie der Wissenschaften zu München, mathematisch-naturwissenschaftlichen Klasse 141: 1–133. Wellnhofer P. 1985. Neue Pterosaurier aus der SantanaFormation (Apt) der Chapada do Araripe, Brasilien. Palaeontographica A 187: 105–182. Wellnhofer P. 1987. Die Flughaut von Pterodactylus (Reptilia, Pterosauria) am Beispiel des Wiener Exemplares von Pterodactylus kochi (Wagner). Annalen des Naturhistorischen Museums in Wien A 88: 149–162. Wellnhofer P. 1988. Terrestrial locomotion in pterosaurs. Historical Biology 1: 3–16. Wellnhofer P. 1991a. The illustrated encyclopedia of pterosaurs. London: Salamander Books Ltd. Wellnhofer P. 1991b. Weitere Pterosaurierfunde aus der Santana-Formation (Apt) der Chapada do Araripe, Brasilien. Palaeontographica A 215: 43–101. Wellnhofer P, Buffetaut E, Gigase G. 1983. A pterosaurian notarium from the Lower Cretaceous of Brazil. Paläontologische Zeitschrift 57: 146–157. Weyl AR. 1945. Stability of tailless aeroplanes. Aircraft Engineering 17: 73–81, 103–111. Wilkinson MT. 2002. Flight of the ornithocheirid pterosaurs. PhD thesis, University of Cambridge. Wilkinson MT, Unwin DM, Ellington CP. 2006. High lift function of the pteroid bone and forewing of pterosaurs. Proceedings of the Royal Society of London B 273: 119–126. Williams PL, Warwick R, Dyson M, Bannister LH, eds. 1989. Gray’s anatomy, 37th edn. Edinburgh: Churchill Livingstone. Wood VE. 1984. The sesamoid bones of the hand and their pathology. Journal of Hand Surgery (British and European Volume) 9: 261–264. Wright JL, Unwin DM, Lockley MG, Rainforth EC. 1997. Pterosaur tracks from the Purbeck Limestone Formation of © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 68 M. T. WILKINSON Dorset, England. Proceedings of the Geologists Association 108: 39–48. Zimmer K. 1935. Beiträge zur Mechanik der Atmung bei den Vögeln in Stand und Flug. Zoologica 33: 1–69. von Zittel KA. 1882. Ueber Flugsaurier aus dem lithographischen Schiefer Bayerns. Palaeontographica 29: 47–80. APPENDIX 1 Institutional abbreviations: AMNH, American Museum of Natural History, New York, USA; IMCF, Iwaki Coal and Fossil Museum, Iwaki, Japan; NSM, National Science Museum, Tokyo, Japan; RGM, National Natuurhistorisch Museum, Leiden, the Netherlands; SMNK, Staatliches Museum für Naturkunde, Karlsruhe, Germany; SMNS, Staatliches Museum für Naturkunde, Stuttgart, Germany. Anatomical/arthrological abbreviations: (1,2,3)ax, (primary, secondary, tertiary = long) axis; ch, cheiropatagium; cm, carpometacarpal joint; co, coracoid; cogl, coracoid part of the glenoid fossa; cpt, carpopteroid joint; cr, cruropatagium; CR, centre of rotation; crax, conjunct rotation axis; cs, cristospine; dpcr, deltopectoral crest; dpr, dorsal process; ds, distal syncarpal; dsfac, distal syncarpal facet; el, elbow joint; etpr, extensor tendon process; ettr, extensor tendon track; fe, femur; fov, fovea; gl, glenoid fossa; gr, groove; hp(1,2,3), hip joint (primary axis, secondary axis, tertiary = long axis); hu, humerus; hucot, humeral cotyle; hufac, humeral facet; il, ilium; ip(1,2,3), (first, second, third) interphalangeal joint; is, intersyncarpal joint; kne, knee joint; knu, knuckle joint; lcon, lateral condyle; mc, medial carpal/medial carpal joint; mcfac, medial carpal facet; mcon, medial condyle; no, notarium; phcon, phalangeal condyle; phfac, phalangeal facet; pn, pneumatic foramen; pro, propatagium; ps, proximal syncarpal; psfac, proximal syncarpal facet; pt, pteroid; ptcot, pteroid cotyle; ra, radius; racon, radial condyle; rafac, radial facet; ri, ridge; ruc, radioulnocarpal joint; rufac, radial/ulnar facet; sc, scapula; scgl, scapular part of the glenoid fossa; scno, scapulonotarium joint; ses, sesamoid; sh(1,2,3), shoulder joint (primary axis, secondary axis, tertiary = long axis); stco, sternocoracoid joint; stp, sternal plate; ten, tendon; tt, tibiotarsus; tu, tuberculum; ul, ulna; ulcon, ulnar condyle; ulfac, ulnar facet; wf, wing finger; wf-ph(1,2,3,4), (first, second, third, fourth) wing-finger phalanx; wm, wing metacarpal; wmcot, wing-metacarpal cotyle; wmfac, wing-metacarpal facet. Other abbreviations: a, angle of attack; c.g., centre of gravity; M, pitching moment; M0, zero-lift pitching moment; m.a.c., mean aerodynamic centre. APPENDIX 2 Bone measurements (in mm) of selected ornithocheirid specimens from the Santana Formation and Crato Formation (in the case of SMNK SMNK 1132PAL) of Brazil, given to the nearest mm, with estimated wingspans (WS) in m. Each specimen represents a single individual. Asterisks indicate measurements taken from the literature. Parentheses indicate an incomplete element. Wingspans were estimated on the assumption that the ornithocheirid specimens were geometrically similar, with wingspan : ulna length ratios equal to those estimated for A. santanae. Specimen sc co hu ul ra ps+ds pt wm wf–ph1 wf–ph2 wf–ph3 wf–ph4 fe tt WS AMNH 22552 AMNH 22555 AMNH 24444 IMCF 1053 NSM–PV 19892 RGM 401 880* SMNK 1132PAL SMNK 1133PAL SMNK 1134PAL SMNK 1135PAL SMNK 1136PAL SMNK 1250PAL – 90 73 – 106 130 80 137 – 96 112 122 – 126 90 151 – 175 115 185 – 121 158 – 170 – – 220 254 290 230 295 – 172 225 230 243 – 242 341 390 410 312 397 241 263 380 353 240 – – – 385 401 – 391 238 – 380 – 20 – 25 – 41 – – 47 37 – – – – – – – – – – (178) – – – – 172 – 170 – 255 – 227 – 165 179 254 248 355 – 353 – – – 445 – 357 383 577 515 324 – – – – – 402 – – – 532 – 252 – – – – – 312 470 – – 393 – – – – – – – 275 – – – – – – – – – 234 285 190 271 – – – – – – – – – 355 234 – – – – – 3.5 4.3 3.5 5.0 5.7 6.0 4.5 5.8 3.5 3.8 5.5 5.1 © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69 3D PTEROSAUR WING SHAPE 69 SUPPORTING INFORMATION Additional Supporting Information may be found in the online version of this article: Video clips S1–S3. AVI films of a three-dimensional reconstruction of Anhanguera santanae, showing the overall effects of full flexion at the elbow, radioulnocarpal, carpopteroid, and knuckle joints: the movements that would have been used to furl the wing. Flexion of the elbow joint was indissociably coupled with that of the radioulnocarpal joint, but the movements at the other joints could have occurred independently. These films also show movement of the hindlimb from its close-packed position (20° below and 30° behind the transverse y-axis) to a position suitable for a terrestrial stance (60° below and 20° ahead of the the transverse y-axis). The trajectory of the hindlimb shown is one of very many possibilities. The relevant joints axes are indicated in red. Video clip S1, dorsal aspect. Video clip S2, anterior aspect. Video clip S3, anterodorsolateral aspect. Gridlines are spaced at intervals of 500 mm. Video clips S4–S6. AVI films of the same sequence of movements shown in Video clips S1–S3, with the addition of full flexion of the intersyncarpal joint. Video clip S4, dorsal aspect. Video clip S5, anterior aspect. Video clip S6, anterodorsolateral aspect. Video clips S7–S9. AVI films showing the various possible shoulder movements in the following sequence: (A) elevation/retraction of 25° about the primary axis from the position of maximum depression to the close-packed position; (B) supination of 50° about the tertiary axis; (C) primary angulation of 70°, manifest as almost pure retraction caused by the supination of the primary axis in step (B); (D) protraction/elevation of 80° to the position of maximum elevation, with no angulation about the primary axis; (E) depression of 70° about the primary axis, following which the arm is maximally pronated as a result of diadochal movement that occurred in step (D); (F) supination of 30° about the tertiary axis to the close-packed position; (G) positive secondary angulation of 10°, manifest predominantly as protraction, but also elevation resulting from the 35° tilt of the secondary axis; (H) negative secondary angulation of 50°, manifest predominantly as retraction, but also as depression. These video clips do not illustrate realistic wing kinematics, but show the maximum theoretical range of movement about the different shoulder joint axes estimated from a consideration of the bones alone. Video clip S7, dorsal aspect. Video clip S8, anterior aspect. Video clip S9, antero-dorso-lateral aspect. Please note: Blackwell Publishing are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article. © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 154, 27–69
© Copyright 2025 Paperzz