Palaeont. afr., 25,151-180 (1984) POSTCRANIAL REMAINS OF FABROSAURIDAE (REPTILIA: ORNITHISCHIA) FROM THE STORMBERG OF SOUTHERN AFRICA by A.P. Santa Luca Institute for the Study of Earth and Man, Southern Methodist University, Dallas, Texas 75275 ABSTRACT The postcranial skeletons of three fabrosaurids from the upper Elliot Formation "Red Beds" of the Stormberg Group in southern Africa are described. The material demonstrates details of fabrosaurid anatomy previously unknown, particularly a short, deep prepubic process which is undoubtedly primitive for the Ornithischia. Besides the short prepubis, fabrosaurids are characterized by 1) a reduced manus; 2) an ilium hav­ ing a lateral extension of the supra-acetabular margin and a deep nearly vertical brevis shelf; and 3) an elongated hindlimb. Postcranial morphology excludes the fabrosaurids from the ancestry of the contemporaneous heterodontosaurids. Neither can the fabro­ saurids be considered ancestral to the 'juvenile scelidosaurid' (BMNH R6 704) as has been suggested. On the contrary, the 'scelidosaurid' is more primitive in structure than fabrosaurids. The assignment of Nanosaurus agilis Marsh to the Fabrosauridae is not substantiated after morphological comparisons between the postcranial material of both. The taxonomic status of Scutellosaurus lawleri is regarded as uncertain. The fabro­ saurids are more similar to the Morrison Formation camptosaurids, than to Hypsilopho­ don. Finally, it is argued that ornithopods were not a basal stock for the phylogenesis of non-ornithopods but represent an independent radiation comparable to the other ornithi­ schian suborders. The fabrosaurids were an early development of the ornithopod radia­ tion itself. CONTENTS Page INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 DESCRIPTIOI'{ ............. . 153 DISCUSSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 7 ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176 REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176 APPENDICES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 7 8 151 INTRODUCTION For many years it was thought that ornithis­ chian dinosaurs appeared late in the Mesozoic since few unequivocal skeletal remains were known from earlier than the late Jurassic. This was in marked contrast to the saurischians which were relatively common in the later Triassic. The eventual recove­ ry of early ornithischian material from late Trias­ sic/early Jurassic deposits (Bonaparte 1976; Casa­ miquela 1967; Colbert 1981; Crompton and Charig 1962; Ginsburg 1964; Santa Luca, Crompton and Charig 1976; Simmons 1965; Thulborn 1970a, b, 1972. 1974) has disproven the notion oflateevol­ ving ornithischians. The next development in the assessment of early ornithischians was the recog­ nition that they could not all be accommodated within the conservative ornithopod family Hypsil­ ophodontidae· as a direct result the families Hete­ rodontosauridae (Kuhn 1966; Romer 1966) and Fabrosauridae (Galton 1972) were erected to con­ tain much of the new material. Finally, after a de­ tailed analysis of some of the heterodontosaurid material, it became apparent that the morphologi­ cal diversity displayed by the early ornithischians even went beyond the limits of the suborder Orni­ thopoda (Santa Luca Crompton and Charig 1976; Santa Luca 1980). MS accepted March 1983 At the present time most early omithischian di­ nosaurs known are classed either as Fabrosauridae or as Heterodontosauridae (the status of the 'juve­ nile scelidosaurid, BMNH R6 7 04, remains uncer­ tain). The fabrosaurids represent an early lineage in the ornithopod radiation; they were small, around a metre in length, and probably obligate bipeds in locomotion. They were also fairly widely distribu­ ted geographically: they are known from the Stormberg localities in southern Africa, from the Kayenta formation of western North America, from the Jurassic of Portugal (Thulborn 1973) and 152 perhaps the Jurassic-Cretaceous of England (Echi­ nodon). The heterodontosaurids represent an early ra­ diation of non-ornithopods; they were also small, the only complete postcranial skeleton giving a length of one metre. However, Heterodontosaurus tucki was certainly not an obligate biped and was capable of quadrupedal locomotion (Santa Luca 1980). The distribution of heterodontosaurids seems to be less widespread, from the Stormberg of southern Africa and perhaps from the Ischigua­ lasto of Argentina (Pisanosaurus: Casamiquela 1967; Bonaparte 1976). The contemporaneity of these families, and the 'juvenile scelid<;>saurid' to be discussed later, suggests that diverging trends had become established quite early in ornithischian phylogeny. CLASSIFICATION OF THE FABROSAURIDAE Ginsburg ( 1964) described the first fabrosaurid, Fabrosaurus australis, on the basis of a fragmentary right dentary with several teeth. This specimen was found in 1959 in the Upper Red Beds of the Storm­ berg series of Lesotho Because the teeth were tri­ angular in shape with small occlusal denticles and a basal cingulum he interpreted the specimen as a sce­ lidosaur, comparing it with Scelidosaurus harrisonii Owen 1861 and interpreting the latter as a member of the suborder Stegosauria.) Thulborn (1970) described the skull and denti­ tions of two more specimens from the Upper Red Beds of Lesotho which he believed were congene­ ric and conspecific with Ginsburg's specimens. One of these (University College, London B. 1 7) was collected from Likhoele Mountain, as was Gins­ burg's; the other (UCL B. 23) was also collected from the Upper Red Beds but about 40 miles (c. 60 km) distant from the former. Specimen UCL B. 17 comprises cranial, dental and postcranial ma­ terial but UCL B. 23 only cranial and dental mate­ rial. Thulborn (1970) disagreed with Ginsburg's as­ signment of F. australis to the Scelidosauridae and transferred both his and Ginsburg's specimens to the Hypsilophodontidae. The reason for doing so was the discovery that F._ australis had a toothed premaxilla which Thulborn considered distinctive of hypsilophodontids. For the same reason he in­ cluded Echinodon within the Hypsilophodontidae and interpreted it as an intermediate between Fa­ brosaurus and Hypsilophodon ( 19 70:429-430). Galton (1972) removed F. austral£s, both Gins­ burg's and Thulborn's specimens, from the Hypsi­ lophontidae; he reassigned them to the newly crea­ ted family Fabrosauridae which was defined as "having marginally positioned maxillary and den­ tary teeth" (1972:464). At the same time £chino­ don was transferred to this family because it also possessed this character, considered by Galton to be primitive for the Ornithischia. If in fact a mar­ ginal dentition is primitive, then the family Fabro­ sauridae would have been defined on the basis of a symplesiomorphy and the composition of the fami- ly (including Echinodon and Nanosaurus agilis) may be paraphyletic. In a discussion of the supposed generic syno­ nymy of Lycorhinus and Heterodontosaurus, Cha­ rig and Crompton (1974) briefly considered the dentition of the known fabrosaurids. In their opi­ nion the dentition of these specimens may be pri­ mitive or plesiomorphous for omithischians as a whole; the teeth would then not be diagnostic of a single genus or species but rather of a family or higher taxon. Consequently, they argued that the genus Fabrosaurus and the species F. australis were nomina dubia and that Ginsburg's and Thulbom's specimens were of an indeterminate genus and spe­ cies of fabrosaurid. Galton ( 197 8) partially revised the Fabrosauri­ dae to include Nanosaurus agilis Marsh 18 7 7 (Mar­ sh 1877, 1894) because the dentary is slender and had a marginal rather than inset dentition, the latter supposedly implying the presence of cheeks (Galton 1973). At the same time Galton removed Thulborn's University College, London specimens from F. aus­ tralis Ginsburg and created a new genus as well as a new species for thel!l, Lesothosaurus diagnosticus Galton 1978 . . He did not, however, follow Charig and Crompton (1974) regarding the invalidity of both the genus Fabrosaurus and the species F. aus­ tral£s, but accepted Ginsburg's assignment. In Gal­ ton's system then, F. australis refers only to the partial right dentary described by Ginsburg. The only existing description of the fabrosau­ rid postcranial skeleton comes from Thulbom (1972). The specimens he described were fragmen­ tary and many important structures were not pre­ served; in contrast the specimens to be described here, although also fragmentary, fortunately inclu­ de details of postcranial anatomy not available from the UCL specimens. These details clarify the relative position of fabrosaurids to heterodontosau­ rids, the character-state of some important structu­ ral features in ornithschian phylogeny, and add considerably to the functional understanding of fa­ brosaurid anatomy. The teeth of most early omithischians are re­ markably similar and primitive in character; in this case details of postcranial anatomy become all the more important in unravelling the evolutionary re­ lationships of early omithischians. Thus the goal of this paper is not merely to provide anatomical des­ criptions and functional interpretations of impor­ tant osteological material, but to evaluate the evo­ lutionary differences between the early omithis­ chians. Material The postcranial remains of three South African Museum specimens are described in this paper. SAM-K400 and K401 were recovered from Lik­ hoele Mountain in Lesotho, the site of Ginsburg's original discovery of fabrosaurid remains and of the UCL B.17 specimen (for map see Thulbom 1972). SAM-K1106 was recovered at Dangershoek. In addition to the postcr~nia some dental remains and fragmentary cranial material were recovered and are being described. An inventory of preserved material and a list of measurements are given in the Appendices. The fabrosaurid specimens have been compared with a wide range of omithischian postcranial mate­ rial. This included ankylosaurs (Coombs 1978; Eat­ on 1960; Maryanska 1977); ceratopsians (Brown and Schlaikjer 1940; Hatcher, Marsh and Lull 1907; Maryanska and Osm6lska 1975; Lull1933; Osborn 1923, 1924); ornithopods (Galton 1974a, b, 1977, 1981;GaltonandJensen 1973;Gilmore 1909,1915, 1924b, 1925; Hooley 1925; J anensch 1955; Lull and Wright 1942; Ostrom 1970; Marsh 1894; Parks 1920, 1926; Shepherd, Galton and Jensen _1977; Sternberg 1940); pachycephalosaurs (Gilmore 1924a Maryanska and Osm6lska 1974), stegosaurs (Gilmore 1914) andHeterodontosaurus tucki (San­ ta Luca 1980). B C 153 DESCRIPTION Vertebral Column It is still not possible to determine the total number of vertebrae or the number of vertebrae in a particular region of the fabrosaurid vertebral co­ lumn. Cervical vertebrae (fig.1-3). The description of the cervical vertebrae must begin with a consideration of the cervical vertebrae preserved in the UCL B.1 7 specimen. Thulbom (1972:33, fig. 3) illustrated a fragment containing parts of three cervical vertebrae (the anterior two almost complete, the third represented only by the pre zygapophyses) and attributed them to the middle cervical region but gave no evidence for this Fig. 1 SAM-K1106. A- C. Cervical centra. D- F. Dorsal centra. Right lateral view. Scale= 5 em. B D Fig. 2 SAM- K11 06. A- B. Cervical neural arches. C-E. Dorsal neural arches. Right lateral view. Scale= 5 em. c D Fig. 3 SAM- K1106. A- B. Cervical neural arches. C-E. Dorsal neural arches. Dorsal view. Scale = 5 em. 154 conclusion. On the basis of serial changes in cervical verte­ brae (e.g., see Hypsilophodon foxii, Galton 1974b or Heterodontosaurus tucki, Santa Luca 1980) those of the UCL B.l 7 specimen can only be ante­ rior cervicals First, the diapophyses are very small and set on the pedicles of the neural arch near the neurocentra! suture as in anterior cervical vertebrae; in middle cervicals the diapophysis would be at the end of a small transverse process. Second, the para­ pophyses on the centra are poorly developed and lie below the neurocentra! suture, also as in ante­ rior cervical centra; in middle cervicals the parapo­ physes would be larger and would be bisected by the neurocentra! suture. The three vertebrae Thul­ bom ( 1972) illustrated are more likely to be either the axis, C3 and C4 or C3, C4 and C5, rather than middle cervicals. The three centra described here are from SAM­ Kll06 (fig. lA-C); each has a ventral keel and were thus considered to be cervicals. The two lar­ ger centra are asymmetrically constricted: concave above a diagonal from the middle of the ventral margin to the posterodorsal comer but convex or flat below this diagonal. This asymmetry seems to be related to the shape of the ventral keel which is wide, low and transversely rounded posteriorly but narrow and sharp anteriorly. The posterior part of the keel extends upward onto the lateral central surface; it has rugose, anteroposteriorly oriented striations probably associated with the anterior vertebral musculature. The remaining centrum (fig. lC) is smaller, not asymmetrically constricted and does not have an expanded and flattened ventral keel; posteriorly it does have a slight parallelogram-shaped outline. The anterior intercentral facet is circular and smal­ ler ( 7,3 mm high) than the p osteri.or which is oval (8, 7 mm high). The other two centra are almost rectangular in lateral outline with both the anterior and posterior facets circular. These cervical centra do not resemble those of Thulbom's (1972) fabrosaurid which are concave below a diagonal from the posteroventral comer to the parapophysis in the anterodorsal comer; they also lack the posterior expansion of the ventral keel. These differences may be due to serial chan­ ges, but at the present no data are available to set­ tle the question. The fabrosaurid cervical centra do not resemble those from any part of the cervical series of Hetero­ dontosaurus. In H. tucki the anterior and posterior central margins are marked by prominent vertical ridges, the anterior of which is joined to the para­ pophysis dorsally; the excavation of the lateral sur­ faces of the centrum is also considerably more pro­ nounced. The centra of SAM-Kll06 most nearly resemble those of Thescelosaurus neglectus (Gil­ more 1915; Galton 1974a), which have a similar asymmetric constriction of the centra. Serial changes in the two SAM-Kl106 cervi­ cal neural arches indicate that one probably comes from the posterior half but not the end of the cervi- cal series (fig. 2A, 3A); in this specimen only the uppermost part of the parapophysis is on the arch pedicle; the transverse process is about 8 mm long and arises just below the pre zygapophysis; the transverse axis of the zygapophyses is about 60° to the horizontal; the distance between the anterior and posterior margins of the zygapophyses is 15,4 mm; the postzygapophyses lie on the same level as the prezygapophyses; and the neural spine is a small nubbin. The other cervical neural arch pro­ bably comes from nearer the end of the cervical series (fig. 2B, 3B); here the parapophysis lies al­ most entirely above the neurocentra! suture; the transverse process is 9 mm long and arises at the level of the prezygapophyses; the transverse axis of the zygapophyses is about 45° ; the distance be­ tween the anterior and posterior margins of the zygapophyses is 16,7 mm; the pre- and postzyga­ pophyses lie on the same level; the neural arch, while broken off, is still quite small. Since these neural arches are from the posterior cervical region they are not comparable with those of UCL B.l7. However the arches are generally si­ milar to those of Hypsilophodon (Galton 1947b) and Iguanodon (Hooley 1925); they differ from Camptosaurus in that the cervicals of the latter do not have neural spines (Gilmore 1909) and their postzygapophyses are less posteriorly projecting than those of Thescelosaurus (Galton 194 7a). They differ from Heterodontosaurus only in that the posterior cervicals of the latter have larger neural spines and more anteriorly projecting, tongue­ shaped prezygapophyses. Dorsal vertebrae (fig. 1-3) The dorsal centra were identified by the absen­ ce of parapophyses and by matching the typical pattern of the neurocentra! suture on the base of the dorsal neural arches with the negative of that pattern on the centra. The anterior and posterior facets of the dorsal centra are nearly perpendicular to the A-P axis; the centra are also uniformly concave about the A-P axis. Two of the SAM-K401 centra have a faint ventral keel and are also relatively narrower than the non-keeled centra of SAM-K401. The centra are amphiplatyan or very slightly amphicoe­ lus in SAM-Kl106 but are more generally amphi­ coelus in SAM-K401, the centra of which are also larger than those of the former. No remarkable fea­ ture of the dorsal centra distinguishes them from those of other ornithopods such as Hypsilophodon or Camptosaurus. Neural arches are preserved from the anterior, middle and posterior dorsal regions in SAM-K401 and K1106. The anterior arches (fig. 2C,D; 3C,D) have stout transverse processes rising at about 20°- 300 above the horizontal; they are subcircular dis­ tally where the rib tubercle articulates. The neural arch pedicle is rather higher compared to the pedi­ cles of the posterior dorsals; the parapophysis lies just anterior to the root of the transverse process in SAM-K40 1 and immediately underneath the pro­ cess in SAM-K1106. In this feature the SAM­ K11 06 arches resemble the third dorsal of Campto­ saurus (Gilmore 1909:232, fig. 15). The transverse axis of the zygapophyses is also about 20°- 30° above the horizontal. The neural spine arises from a short base on the arch behind the prezygapophy­ ses. In the middle dorsal vertebrae (fig. 2E, 3E) the transverse processes are horizontal and rectangular in cross section at their distal end. The processes are also lower relative to the neurocentra! suture as the pedicles are shorter. The parapophysis i~ closer to the anterior margin of the centrum, antenor and just inferior to the root of the transverse process and lateral to the prezygapo_physi~. The base. of the neural spine is more extensive, since the distance between the pre- and postzygapophyses has increa­ sed. In SAM-K401 the anterior and posterior mar­ gins of the spine are somewhat divergent toward ~ ts extremity so the spine is wider at its top than at Its base. In Thulborn's (1972) fabrosaurid ~he n~ural spine was of consistent width throughout Its height. In the posterior dorsal neural arches the trans­ verse processes are shor!er but still horizontal_ and rectangular in cross sectwn. Tee transverse a~Is of the zygapophyses is about 10 above the honzon­ tal. The neural spine does not differ from that of the middle dorsal arches. None of these arches are from the posteriormost dorsals as the parapophy­ sis and diapophysis are still distinct. The morphology of the dorsal vertebrae seems to have been quite conservative within the Ornitho­ poda. Only two differences separate those just de­ scribed from the dorsals of Hypsilophodon: 1) the margins of the neural spine are divergent in the arches of SAM-K401 while the spines of Hypsilo­ phodon are consistently rectangular (Galton 1974b: 55, fig. 22); 2) in the fabrosaurid described here the transverse axis of the zygapophyses decreases from about 45° in the anterior arches to about 10° in the posterior but is consistently about 45u above the horizontal in Hypsilophodon. The third dorsal neural arch of Heterodontosaurus tucki differs from the anterior arches of the fabrosaurids in ha­ ving a small ridge which joins the transverse process and the parapophysis. However the fourth dorsal is a typical middle dorsal without a ridge. Sacral vertebrae The two centra of SAM-K1106 (fig. 4) are low, the anterior and posterior facets semicircular in shape. Ventrally one centrum 1s gently convex while the other has a small ventral ridge. All the sa­ cral centra of SAM-K401 have a small ventral rid­ ge but are not as low as those of SAM-K11 06. Thulborn (1972) noted that the sacral centra of the UCL B.1 7 fabrosaurid bore a distinct ventral keel; his drawing (1972:34, fig. 4) shows a struc­ ture much more pro min en t than that found on the centra described here though they all are similar in absolute size. · A c Fig. 4 155 B D SAM-Kll06. A-B. Sacral centra. C-D. Cau­ dal centra. Left lateral view. Scale = 5 em. The four sacral neural arches of SAM-K401 seem to represent a single series; the neural spines change from the dorsal type which are narrower at the base than distally, to the caudal type which are taller and uniform in anteroposterior width throu­ ghtout. A similar series of changes is noted in the level of the pre- and postzygapophyses. In the ante­ riormost arch the posterior zygapophyses are 2-3 mm higher than the anterior but in the last arch the posterior are about 7 mm above the level of the anterior. The orientation of the zygapophyseal fa­ cets cannot be determined because the articular surfaces have been heavily eroded. Each of these arches bore a true sacral rib. Caudal vertebrae The largest centra with facets for chevr_on arti­ culation are presumed to be from the antenor cau­ dal region. They are strongly convex along the _A-P axis with anterior and posterior facets much Wider than the centrum in mid-length. A block of matrix from SAM-K1106 (fig. 5) contains eight complete middle-to-posterior caudal vertebrae. These centra are approximately as long as the anterior caudal centra but are only about half as high (5-7 mm high versus 10-15 mm high for the anterior centra). A ridge runs anteroposte­ riorly midway between the dorsal and ventral mar- gin of the centrum; ventrally each centrum bears two long ridges between which the surface of the centrum is concave. Chevrons are present but their articular facets are poorly marked. The neural ar­ ches of these centra lack a neural spine and any tra­ ce of a transverse process; the arms of the zygapo­ physes lay 45° or less to the horizontal. These ver­ tebrate correspond in size to about the twenty­ fourth caudal of specimen BMNH R5830 of Hypsi­ lophodon foxii (Galton 1974:70, fig. 32). One further matrix fragment of SAM-K1106 contains three of the posteriormost caudal verte­ brae. They are only about 5 mm long and about 1,5 mm high; the zygapophyses are still borne on processes which rise above the centrum at about 10°, but there are no chevrons. Three partial caudal neural arches were identi­ fied among the SAM-K401 material. Only one has a complete neural spine which is narrow-based, high and vertical. The prezygapophyses extend in front of the arch pedicle on small horizontal pro- 156 Fig. 5 SAM-Kll 06. Matrix containing articulated caudal vertebrae. Left lateral view. Scale = 5 em. cesses just below the base of the spine; the poste­ rior zygapophyses are 5 mm above the level of the anterior. The transverse axis of the pre- and post­ zy~apophyses is about 45 ° in one arch and about 70 in the other two. Thulbom (1972:35) speculated that the post­ zygapophyses were weakly developed or absent in the single posterior caudal neural arch from the UCL B .1 7 specimen. This is highly unlikely since even the smallest caudals of SAM-K1106, much smaller than that illustrated by Thulbom (1972: 34, fig. 4K), bear the anterior and posterior arms for both sets of zygapophyses. The middle and pos­ terior caudals illustrated by Thulbom have no transverse process, as in the specimens here. In this they resemble the posterior caudals of Hypsiloph?­ don but differ from those of Heterodontosaurus, 1n which the posterior caudals bear a small nubbin of a transverse process. PECTORAL GIRDLE AND FORELIMB Two fragmentary scapular blades are preserved in the matrix block of SAM-K1106 (fig. 6). At their posterior and inferior margin they bear a pro­ minent expansion exactly like that found on the UCL B.1 7 scapula. This projection resembles that of Camptosaurus (except possibly C. medius Mar­ sh) more than that of either Hypsilophodon or Thescelosaurus in which the process is less pro­ nounced because the scapula of these latter expan­ ds in width gradually toward the vertebral border. An acromial process was present but its size is in­ determinate since the distal end has been broken away. The narrowest part of the scapular shaft is about 11 mm wide and about 6,5 mm thick; thus the shaft is rectangular and never becomes circu­ lar in cross section. The glenoid surface of each sca­ pula is covered with matrix. Two unfused coracoids of SAM-K11 06 are completely prepared (fig. 7). They are rectangular, about 27 mm dorsoventrally (from glenoid to acro­ mial end) and about 21 mm anteroposteriorly (from scapular to free end). A large foramen pier­ ces the external surface about 5 mm from the sutu­ ral attachment to the scapula. The course of the foramen is perpendicular to the plane of the cora­ coid and so pierces the deep surface of the bone al­ so about 5 m:r;n lateral to the scapular attachment. In the UCL B.1 7 fabrosaurid the coracoid surface of the scapula has a distinct notch whose mirror image in the coracoid would have created a fora­ men (the coracoids are not preserved in the UCL B.17 specimen). The coracoid is much thicker at the glenoid end than at the dorsal margin. The gle­ noid margin makes an angle of only about 20° with the scapular edge, creating a deep and narrow gle­ noid fossa. The coracoid of SAM-K1106 differs from that of Hypsilophodon BMNH R5830 (Gal­ ton 1974b) only in that the coracoid of Hypsilo­ phodon has a pronounced concave ventral border while that of the fabrosaurid is nearly straight, marked only by a groove near the glenoid lip. The scapulocoracoid of SAM-K1106 differs in several respects from that of Heterodontosaurus: 1) the hook-like process of the scapular blade at the posterior-ventral margin is much more promi­ nent; 2) the shaft of the fabrosaurid scapula be­ comes rectangular in cross section but not circular; 3) the glenoid fossa is probably deeper than in H. tucki. Fig. 6 SAM-K1106. Anterior portion of matrix block, left side. Note scapular blades and proximal humeral shaft. 157 Fig. 7 SAM-K1106. Coracoids. Lateral view. Scale= 5 em. Fig. 8 SAM-Kll 06. Left humerus. Anterior view. A cervical neural arch is still attached by matrix.Scale = 5 em. Fig. 9 SAM-K1106. Left radius (foreground) and ulna (back­ ground). Lateral view. Scale= 5 em. The proximal half of a right humerus (see fig. 6) and the entire left humerus (fig. 8) are preserved in SAM-K1106. The head is expanded (9 mm wide, considerably more than that of specimen UCL B.1 7) and convex both lateromedially and anteroposte­ riorly; it lies on the medialmost part of the proxi­ mal end and is not centrally placed. On the right humerus the medial side of the shaft below the hu­ meral head is thickened into a ridge which conti­ nues inferiorly to the lower level of the deltopecto­ ral crest opposite. A similar feature appears in Thul­ bom's (1972:38, fig. 7C) drawing of a humerus, but is not noted in the text. The surface of the left humerus is considerably damaged and consequent­ ly the ridge-like structure seems less well-developed. The deltopectoral crest projects from the humerus as a distinct process about 15 mm below the head; at its highest point is is about 5 mm above the hu­ meral shaft and it is only about 15 mm long. The radial condyle is anteroposteriorly longer than wide with the long axis directed anterolateral­ ly; the ulnar condyle is more oval in section, the long axis transversely oriented. Consequently, the .distal anteroposterior humeral width is greater late­ rally than medially. The distal posterior surface is slightly concave between the two condyles; the dis­ tal end of the shaft lacks both an ect- or entepicon­ dylar process. A left radius and ulna of SAM-K1106 (fig. 9) and a right radius of SAM-K400 were preserved but the articular surfaces and the cortical bone are so damaged that few details can be made out. The proximal end of the radius seems to be semicircular in cross-section, unlike the UCL B.17 radius which is kidney-shaped in proximal view, and the Hypsi­ lophodon radius, which is rectangular in section (Galton 1974: 79, fig. 40E). Distally the radius of both specimens is subcircular in section. The ulna of SAM-K1106 lacks an olecranon process; the shaft is triangular in section at both extremities. These forearm bones are larger than those of the UCL B.1 7 fabrosaurid, but the humeroradial index is nearly the same: 1,49 in SAM-K1106 and 1,58 in the UCL B.1 7 specimen. Two bones of the manus were found cemented by matrix to the radial margin of the left distal hu­ meral shaft. The larger is 9,6 mm long, the smaller 6,3 mm long. By its rounded rather than concave proximal end the larger is probably a metacarpal and by its small size it could have been of digit 4, the smaller bone is the first phalanx of this meta­ carpal. The SAM-K1106 humerus differs from that of the UCL B.1 7 specimen only in having a more con­ cave posterior surface just above the distal condyles. However the SAM-K1106 humerus does differ 158 considerably from that of Heterodontosaurus and most of the differences are probably related to more specialised forelimb use inHeterodontosaurus. In the latter the deltopectoral crest is considerably more robust and somewhat larger, the condyles are much less amorphous, and a large entepicondyle lies above the ulnar condyle. The ulna of H. tucki differs from the fabrosaurid ulna in having a distin­ ct olecranon process. PEL VIC GIRDLE AND HINDLIMB Ilium (fig. 10-17). Five fabrosaurid ilia were found in this material, but only one had been given a specimen number after preparation (the posterior process of the SAM-K1106 ilium). One other certainly belongs to SAM-K401 and another to SAM-K400, since their state of preservation and color match all the labelled SAM-K401 and SAM-400 material res­ pectively. The remaining two ilia cannot be asso­ ciated with any other postcranial material. Each of these ilia has the characteristic supra-acetabular flan~e which Thulbom (1972) described for his speCimen. The anterior iliac process is complete in only two specimens: in SAM-K401 (fig. 12) it is deflec­ ted somewhat ventrally and laterally and the tip is lateromedially compressed; in the other specimen (fig. 14) the anterior process is horizontal and approximately rectangular in cross-section at the tip. The difference in ventral deflection can proba­ bly be attributed to individual variation since Gal- ton (197 4b: 83) noted similar variation in the vari­ ous ilia of Hypsilophodon. The difference in cross­ sectional shape may also be individual variation; Galton's illustrations do not show comparable varia­ tion in this character but Gilmore's (1909) study of Camptosaurus showed that the tip of the anterior process had different shapes in the various species of that genus. In all specimens the pubic peduncle anteriorly is directed at about 3 5°- 4 5° to the horizontal long-axis of the ilium. The supra-acetabular flange is a laterally projecting ridge on the external aceta­ bular margin of the pubic peduncle; it reaches its maximum lateral extent just above the junction of the peduncle and iliac blade. Posterior to this the flange is reduced until it disappears at the postero­ dorsal acetabular comer. The ischiadic peduncle is rectangular in cross-section, not an expanded bul­ bous process as seen in Hypsilophodon. The posterior processes of the two more com­ plete ilia (fig. 12, 14) are superficially somewhat different because of compression fractures to, and breakage of, the external surface of the SAM-K40 1 ilium. The most striking feature of the posterior process is the large brevis shelf which curves media­ lly under the postacetabular process but then down ward to form a nearly vertical wall. The infe­ rior margin of the posterior process (the inferior margin of the brevis shelf) extends almost horizon- tally backwards from the distal end of the ischia­ dic peduncle; thus the posterior process is quite deep. The area posterior to the ischiadic peduncle was missing in Thulbom's specimen so the orienta­ tion and size of the brevis shelf in fabrosaurids was not previously known. The outline of the posterior iliac margin in SAM-K400 ? (fig. 10) and SAM-­ K401 is angular, but that of the other large ilium is smoothly convex. Careful examination of the former two under a low-power microscope indica­ tes that the fragile bony margin of the posterior process had probably been broken away several mi­ llimeters proximal to its termination and in its na­ tural condition may have been similar to the other large ilium (fig. 14). Each of the larger ilia shows four sacral rib at­ tachments; the most anterior is situated at the junc­ tion of the pubic peduncle with the anterior process rather than on the pubic peduncle itself as in the il­ ium of Hypsilophodon. A fifth sacral rib may have been attached to the posteriormost portion of the brevis shelf which is missing in each specimen. The smallest ilium is only about half the size of the others (fig. 16-17), but even so it has a charac­ teristic though . smaller supra-acetabular flange. The medial surface differs from the others in that two anterior sacral rib facets are found, one above the other, at the junction of the pubic peduncle and the anterior process. The ilia described here differ significantly from the UCL B.1 7 fabrosaur in the position of the sacral rib attachments. The latter has five facets beginning at the junction of the pubic peduncle and anterior iliac process. They extend backward and upward, each successive rib being attached closer to the dor­ sal iliac margin, the fifth rib being but about 7 mm below this edge. In all three specimens described here the fourth rib attached on the brevis shelf just posterior to the ischiadic peduncle, over 20 mm be­ low the dorsal margin. A line through the sacral rib facets begins anteriorly at the junction of the pubic peduncle and anterior process, as in the UCL B.1 7 specimen, but then runs inferiorly through succe­ ssive facets. This difference in sacral rib attachment finds no counterpart in the various Hypsilophodon specimens described by Galton ( 194 7b). In several important ways fabrosaurid ilia differ from that of Heterodontosaurus: in the latter, the anterior process is more robust; the pubic peduncle is more nearly perpendicular to the long axis of the ilium; the acetabular margin just above the ischiadic peduncle is not smooth but bears a strongly deve­ loped articular buttress; the posterior process is considerably shallower and the brevis shelf is nar­ row and horizontal. Ischium The proximal portions of both ischia, from the acetabular margin to the broken obturator process, are preserved in SAM-K401 (fig. 18) and SAM­ K1106 (fig. 19). The concave acetabular margin has a chord of about 16 mm in all specimens and a 159 Fig. 10 SAM-K400?. Left ilium, lateral view. Scale = 5 em. Fig. 11 SAM-K400?. Left ilium, medial view. Scale = 5 em. Fig. 12 SAM-K401. Right ilium, lateral view. Scale = 5 em. Fig. 13 SAM-K401. Right ilium, medial view. Scale= 5 em. 160 Fig. 14 SAM-K?. Right ilium of large unnumbered specimen, lateral view. Scale= 5 em. Fig. 15 SAM-K?. Right ilium of large unnumbered specimen, medial view. Scale= 5 em. Fig. 16 SAM-K?. Right ilium of small unnumbered specimen. Lateral view. Scale= 5 em . Fig. 17 SAM-K?. Right ilium of small unnumbered specimen. Medial view. Scale = 5 em. depth of about 5 mm. The iliac peduncle is thicker dorsally (about 8 mm) but thinner at the 'acetabu­ lar margin in the SAM-K1106 specimens; t~e SAM-K401 specimens have been crushed, but 1n the UCL B.1 7 ischium the iliac peduncle is not thicker dorsally. The length of the peduncle from dorsal. to acetabular margin is about 12 mm. The pubic peduncle is rectangular in cross se~tion, about 6 mm wide in both measurable speCimens and about 14 mm high in SAM-K1106 (left) and 18 mm high in SAM-K401 (right). The obturator process is anteriorly located in the SAM-K401 ischia (it is not visible in the SAM-K1106 specimens), beginning about 27 mm posterior to the acetabular margin. This position corresponds closely with the UCL B.1 7 ischium illustrated by Thulbom (1972:42, fig. 9). Most of the pubic peduncle is missing in Thulbom 's speci­ men but present in those described here; most of the ischial rod posterior to the obturator process is present in Thulbom's specimen but absent in those described here. Consequently, more detailed comparisons are not possible. The obturator process is considerably more 161 Fig. 18 SAM- K401. Proximal parts of ischia. Lateral view. Scale = 5 em. Fig. 19 SAM-K1106. Posterior portion of matrix block; left side. Note prepubic process and proximal part of left ischium. posterior in Hypsilophodon, 53-63 mm behind the acetabular rim (in specimens comparable to SAM-K401 based on the size of the peduncles and of the ~cetabular margin). However, the posi­ tion of the obturator process is quite variable in or­ nithopods: it is anteriorly located in iguanodon~i~s, hadrosaurs, camptosaurs, Dryosaurus and Othnzelza; it is posteriorly located in Thescelosauru~ and f-!yp­ silophodon. The position of the process IS obvwus­ ly not a function of allometry since Camptosaur~s dispar had an ischium about 550 mm long while that of Thulborn's fabrosaurid was probably only about 100 mm long (based on the length ?f t?e postpubic rod which would h~ve .equ~lled t?e Isch~al rod in length). The fabrosaund Ischium diff~rs sig­ nificantly from that of Heterodontosaurus m that the latter does not have an obturator process. Pubis The left prepubic process of SAM-K1106 (fig. 19) and the obturator area with the pro.ximal part of the postpubic rod in SAM-K401 (fig. 20) a:e preserved. SAM-K1106 is the. only fabrosaund specimen described to date which has a complete pre-pubic process. Th?ugh the U~L B.1 7 fabrosau­ rid lacked the prepubis, the pelvis. has been recon­ structed once with a long prepubic process (Thul­ born 1972) and once with a short prepubic process (Thulborn 1971 ). The prepubis is a short stout pro­ cess, 23 mm long from the anterior margin of the obturator foramen and 11 mm deep just anterior to the iliac peduncle. The anterior margin of the process is convex, sloping upward and backward from the horizontal ventral margin. The lateral surface of the prepubic process lacks the muscular tubercles which were such a prominent feature of the Heterodontosaurus pre­ pubis. A small ridge courses anteriorly !rom the iliac peduncle just below the dorsal margin; above the ridge the lateral surface is convex and below the ridge it is shallowly concave. A very lo~ ridge parallels the inferior margin of the process; It con­ tinues posteriorly below the obturator foramen but fades away on the anterior part of the postpubic . process. The acetab~lar ma:gin is tr~I?-sver~ely thick in SAM-K401 (this area IS not VISible In SAM­ K1106); it turns downward posteriorly to close the obturator foramen. One small muscle tubercle lies just below the acetabulum and above the obturator foramen; a similarly positioned larger tubercle in Heterodontosaurus presumably marked the position of the accessorius muscles. The fabrosaurid pubis agrees with that of Hete­ rodontosaurus in having a short but deep prepubic process~ The presence of the same type of prepubis in these two as well as in the BMNH R6 7 04 'juve­ nile scelidosaurid' indicates that this is the primitive ornithischian pattern. The prepubis of Heterodon­ tosaurus is actually shorter (15,2 mm compared to 21 mm) than that of SAM-K1106, though total body size was probably not greatly different in the two specimens. 162 Fig. 20 SAM- K401. Fragmentary left pubis , lateral view. Scale = 5 em. Femur Several femoral fragments are preserved: the proximal and distal ends of a left femur from SAM-K11 06 (fig. 21A, 22, 23) and virtually the entire femur of an unnumbered specimen which matches the SAM-K400 material and will be re­ ferred to as the 'K400?' femur (fig. 24, 25). This latter femur is partially reconstructed dis­ tally but the lateral candy le is sufficiently comple­ te to give a length estimate of at least 14 7 mm (from the tip of the greater trochanter to the most distal point of the lateral condyle), considerably longer than the UCL B.17 femur (104 mm). Fur­ thermore, this femur is associated with a metatar­ sal (probably the fourth) 82 mm long, much larger than that of UCL B.17 (67 mm). The length of this femur and metatarsal corresponds quite closely with that of the femur (151 mm) and third meta­ tarsal (84 mm) of Hypsilophodon foxii (BMNH R196) (Galton 1974b) for which Galton gave an estimated body length of 1,36 metre. The 'K400' femur is also more curved than the UCL B.17 fe­ mur; measurement along the anterior surface of the former is 156 mm giving a curvature index (chord/ arc) of 0,94. In the 'K400?' femur a cleft separates the grea­ ter and lesser trochanters the long (dorsoventral) axis of the lesser trochanter makes an angle of 20u- 25°with the long axis of the greater trochan­ ter. The tip of the lesser trochanter is incomplete in 'K400?' but its preserved length is 21 mm from the base of the cleft. The lateral surfaces of both trochanters and the medial surface of the lesser are marked by strong vertical striations for muscle at­ tachments. A vertically oriented eminence just be­ low and behind the intertrochanteric cleft seems to be continuous with the muscle markings on the greater trochanter and thus may be associated with the insertion of the pubo-ischio-femoralis extemus 1. The proximal portion of the SAM-K11 06 fe­ mur (fig. 22A) retains part of the femoral head and the base of the trochanters. The head is 11 mm wide; its ventral outline makes about a 90° angle with the medial margin of the shaft . The fourth trochanter in 'K400?' is pendant but rises from the femoral shaft along a wide base; the tip is broken but the remaining portion is 21 mm long. The depression for the coccygeofemora­ lis longus muscle (on the medial femoral surface anterior to the fourth trochanter) seems to be less pronounced than in the UCL B.17 femur; but the degree of development of this depression varies in Galton's Hypsilophodon femora (1974:96). Distally the posterior intercondylar fossa of 'K400?' is much narrower and deepr than that of the UCL B.1 7 femur. The anteromedial condylar area is missing in ·K400?' but the distal end of the SAM-K1106 femur shows no anterior intercondy­ lar fossa (fig. 23); the condylar surface of the latter is transversely flat perpendicular to the femoral long axis, indicating that the femur was held in a parasagittal plane. Thulbom (1972:44) notes that 1n UCL B.17 the outer femoral condyle is fractio­ nally larger than the inner. The fabrosaurid femora differ from those of Hypsilophodon only in having a more divergent les- c Fig. 21 SAM-K1106. A. Proximal left femoral fragment, anterior view; note base of lesser trochanter. B. Proxi­ mal left tibial fragment, lateral view. C. Fibular fragment, anterior view. Scale= 5 em. Fig. 22 B A c SAM-Kll06. A. Proximal left femoral fragment, posterior view. B. Proximal left tibial fragment, medial view. C. Fibular shaft , posterior view. Scale = 5 em. Fig. 23 SAM-Kll06. Distal left femoral fragment. Scale= 5 em. Fig. 24 SAM-K401?. Left femur, lateral view. Scale = 5 em. Fig. 25 SAM-K401 ?. Left femur, medial view. Scale= 5 em. 163 164 ser trochanter. In this trait they resemble those of some larger ornithopods such as Camptosaurus and Thescelosaurus. However, both Hypsilophodon and the fabrosaurids lack an anterior intercondylar groove which is present in the latter two genera. The fabrosaurid femur difters considerably from that of Heterodontosaurus. In the latter 1) the lesser trochanter is not separated from the greater but is a strong vertical eminence developed com­ pletely on the anterior margin of the femoral shaft; 2) the greater trochanter does not protrude vertica­ lly above the dorsal margin of the femoral head; 3) the lateral femoral condyle is the smaller, con­ trary to the fabrosaurid condition (however, most ami thopods had a larger inner candy le); 4) H. tuc­ ki lacks the well-developed posterior intercondylar fossa typical of most omithischian femora; and 5) the terminal articular surface of the femur is obli­ quely oriented relative to the femoral long axis, giving the femur an abducted orientation. Signi­ ficantly, neither H. tucki nor the fabrosaurids have an anterior intercondylar groove, so this trait is presumably primitive for omithischians. Tibia-Fibula In SAM-K1106 a complete right tibia (fig. 26) with damaged condyles, a proximal left tibia (fig. 21B, 22B) and a single left fibula lacking the proxi­ mal end (fig. 21C, 22C) are preserved; the undama­ ged proximal portion of a left tibia and a distal right tibia were also found with the SAM-K400 material (fig. 27, 28). The only well-preserved proximal tibial shaft is that of the SAM-K400 tibia: at 21 mm wide across the condyles it is more lateromedially compressed than the tibial head of Hypsilophodon. The inner tibial condyle is more protracted posteriorly though the lateral femoral condyle seems to have been the larger in fabrosaurids. As usual the medial margin of the head is convex and the lateral margin is marked by a deep concavity behind the cnemial crest and in front of the outer condyle. The length of the outer condyle behind this concavity is 16 mm. The transverse width of the cnemial crest, from the concavity to the medial margin, is 14 mm. The posterior intercondylar fossa is V-shaped and 7,5 mm wide. The proximal articular surface is ho­ rizontal, perpendicular to the tibial long axis. Thulbom (1972:44) noted a torsion of 70°be- tween the long axes of the proximal and distal ti­ bial articulations. This is difficult to measure be­ cause the anterior and posterior faces of the distal articulation are not parallel. However, using the midpoint of the lateral and medial edges to give the distal transverse axis shows that the proximal and distal axes are perpendicular in SAM- K11 06. At midlength the shaft is ovoid in cross section, wider anteriorly; the lateromedial diameter is 12,2 mm and the anteroposterior 11,5 mm. A strong crest courses up the lateral margin from the anterolateral edge of the outer malleolus; a short crest about 7 mm long lies along the medial margin above the inner malleolus. In the middle of the anterior sur­ face a small vertical process divides the inner and outer malleolar areas The SAM-K1106 tibia (144 mm) is larger than that of UCL B.17 (129 mm) but they do not differ in any morphological feature. It is also generally si­ milar to that of Hypsilophodon except in having a n':lrrower and deeper posterior intercondylar groove and a more lateromedially constricted cnemial crest. In the fabrosaurid the lengths and widths of the articular extremities are greater, the outer con­ dyle is larger and the least shaft diameter is greater than they are in Heterodontosaurus. The most sig­ nificant difference, however, is that the tibia of Heterodontosaurus forms a functional tibiotarsus with the fused tibula. astragalus and calcaneum. The incomplete fibula of SAM-K11 06 is 119 mm long; the distal malleolar end is semicircular in shape, 8 mm in anteroposterior dimension (from the flattened surface in contact with the tibia to the apex of the anterior rounded surface) and 10 mm in lateromedial width. The fibular head of SAM-K401 is much more expanded anteroposte­ riorly than that of UCL B.1 7. Tarsus and Pes The astragalus (fig. 26), calcaneum and the first distal tarsal capping MT 3 are preserved from the right hindlimb of SAM-K1106. The astragalus is 19,2 mm wide along the anterior margin; it tapers somewhat toward the posterior surface along which it is 17,6 mm wide. The anteroposterior length along the medial margin is 1 7,5 mm and along the lateral margin 13,6 mm. A strong ascen­ ding process marks the anteromedial portion; the height from the inferior surface to the tip of the Fig. 26 SAM-K1106. Right tibia with astragalus anterior view. Note distal metatarsal fragments and proximal phalanges. Scale = 5 em. 165 Fig. 27 SAM- K401. A. Proximal left tibia, lateral view. B. Proximal left fibula, lateral view. C. Distal right tibia, anterior view. Scale = 5 em. Fig. 28 SAM-K401. A. Proximal left tibia, medial view. B. Proximal left fibula, medial view. C. Distal right tibia, posterior view. Scale = 5 em. ascending process is 17,6 mm. The astragalus is attached to but not ankylosed with the inner mal­ leolus. The calcaneum is crescentic in outline; the proximal surface for articulation with the fibula is flat, 8 mm long anteroposteriorly, 5,5 mm wide. Neither the astragalus nor the calcaneum are pre­ served in UCL B.17. The astragalus of SAM­ K11 06 most closely resembles small hypsilopho­ dontids and Pisanosaurus in the presence of a well developed ascending process; this process is absent from larger ornithopods like Thescelosaurus and Camptosaurus but also the smaller Othnielia (= Nanosaurus rex) or Laosauru~ co~sors). The calcaneum is more elongated proximodistally than it is in small h ypsilophodon tids. . Fabrosaurids possessed only two distal tarsals, one capping MT 3 and the other MT 4. Distal tarsal 1 in SAM-K1106 is elliptical in shape, 14,4 mm long anteroposteriorly and 1_0 m~ wide laterome­ dially; it is flattened proxrmod1st~ly and only about 3,5 mm thick. Distal tarsal1In UCL B.1 7 Is disc-shaped and not elongated along any axis. T~e proximal surface of the metatarsals descends _m stepwise fashion from MT2 to MT 3 to ~T ~. Dis­ tal tarsal 1 is just sufficiently thick to bnng Its up­ per surface level with the upper surface of MT 2. The lateral distal tarsal which capped MT 4 must then have been somewhat thicker to bring it level with MT 2 and distal tarsal 1. The fabrosaurid dis­ tal tarsals generally resemble camptosaurids in that distal tarsal 1 covers only MT 3; in small ornitho­ pods such as Hypsilophodon, Othnielia (Nanosau­ rus), Laosaurus and Thescelosaurus, distal tarsal 1 partially covers MT 2 as well as all of MT 3. The fa­ brosaurids differ considerably from H. tucki in tar­ sal structure. In the latter 1) the astragalus and cal­ caneum are indistinguishably fused to each other and to the tibia and fibula creating a bird-like tibia­ tarsus; 2) there are three distal tarsals, the medial caps both MT1 and MT 2 (the homologous bone in fabrosaurids was absent or present only in a cartila­ ginous state), the middle caps MT 3 and the lateral caps MT 4; and 3) the distal tarsals are fused to each other and to the metatarsals creating a tarso­ metatarsus. Right metatarsals 1-4 and most phalanges are preserved from SAM-K11 06, right metatarsals 2-4 (fig. 29,30) and most phalanges from SAM­ K401 (fig. 31, 32). The following description of MT 2-4 applies to both specimens unless other­ wise noted. Though MT 1 had become reduced in many ornithischians, fabrosaurids are unusual in that it is a very small splint of bone applied to the upper medial side of MT 2 (and completely fused there in SAM-K1106). The distal end is expanded somewhat into a small flat articulation with a sin­ gle condyle for the first phalanx. Proximally ~T 2 is lateromedially compressed and anteropostenorly elongated but the lower ~wo-thirds is expanded _and subcircular in cross- sectiOn. The plane of the distal articular surface is inclined upward from the ante- 166 Fig. 29 SAM-K401. Right metatarsals 2,3,4. Oblique view. Scale= 5 em. Fig. 30 SAM-K401. Right metatarsals 2,3,4. Posterior view. Scale= 5 em. rior (dorsal) edge to the posterior (ventral) edge and in this it differs from MT 3 and 4 which are horizontal anteroposteriorly. The transverse axis of the metatarso-phalangeal joint deviated medially in digit 2. Proximally MT 3 is also lateromedially compre­ ssed and anteroposteriorly elongated, though less so than MT 2; the anterior edge is wider than the posterior so that the shaft is triangular in cross-sec­ tion. Metatarsal 1 and 3 were joined in their upper two-thirds but not in their lower third. The long axis of MT 2 (and of MT 4) courses slightly poste­ rior to that of MT 3 which then lies anterior to the other two. Unlike MT 2 and MT 3 the uppermost part of MT 4 is lateromedially expanded and anteroposte­ riorly compressed; in cross-section it is approxima­ tely triangular with the apex lateral. In SAM-K401 only the upper part of MT 4 is joined medially with the shaft of MT 3 but distally it deviates laterally to lie about 6 mm from the end of MT 3; in SAM­ K11 06 the long axis of MT 4 also curves laterally, but less so. In cross section the distal end of MT 4 resembles a right triangle, the base posterior and the hypotenuse, which is convex, lateral. In comparison with the metatarsals of SAM -­ K40 1 and SAM-K11 06 the first metatarsal of UCL B.17 is more robust and not completely fused along its entire length to MT 2, and the shaft of MT 4 is straight, contacting MT 3 along all its med­ ial border. However, the lateral deviation of MT 4 in SAM-K401 and SAM-K1106 may be due to in­ dividual variation since the BMNH R5 83 0 Hypsilo­ phodon specimen described by Galton (1974b:99, fig. 57K) has a divergent MT 4 while that of BMNH R196 rests against MT 3 and is not divergent at all (1974b:100, fig. 58). The fabrosaurid phalangeal formula is 2-3-4-5-? No identifiable digit 5 was found in either speci­ men described here but some indication of its at­ tachment to the posterior upper surface of MT 4 was found in SAM-K401 (see fig. 30). The phalan­ ges of digit 1 are reduced in size but are not atro­ phied or vestigial. The proximal surface of phalanx 1 bears a shallow subcircular depression for the flat, triangular-shaped distal end of MT 1. The distal end of phalanx 1 has a normal trochlea which accepts a small ungual phalanx. In the remaining digits none of the first pha­ langes bears a projecting dorsal process but all the other phalanges do except the unguals. Phalanx 1 of digit 2 is longer and more slender than the first phalanges of digits 3 and 4; the proximal articular surface is very slightly concave. Several features of both phalanx 1 and 2 are associated with the medial deviation of digit 2: 1) the lateral condyle of the distal trochlea is larger than the medial; 2) the pits for the lateral collateral ligaments are much larger than those for the medial collateral ligaments; 3) phalangeal length is greater along the outer margin than along the inner margin. No ungual phalanx can definitely be associated with digit 2 in either SAM- K40 1 or SAM-K11 06. The phalanges of digit 3 are not asymmetric like those of digit 2, indicating that digit 3 followed the midline long axis. Phalanx 1 of digit 3 is only 10% longer than that of digit 2 (22 mm to 20 mm) but it is 33% wider (12 mm to 9 inm); phalanx 2 of digit 3 is also wider than that of digit 2. The plantar surface of the ungual phalanx is a little fla­ ttened and a small ridge projects laterally below the grooves for the claw. The lateral divergence of digit 4, necessary to produce the typical divergent three-toed stance, is accomplished primarily by the deviation of MT 4 from the midline axis. The lateral and medial pha­ langeal lengths are equal and the plane of the inter­ phalangeal joints is flat, not oblique. The phalanges decrease rapidly in length: the fourth is only 6 mm long, though in a comparably sized Hypsilophodon (Galton 1974b:14-15, Table 111) it is 12 mm. The phalanges of UCL B.1 7 were all dissociated and most had damaged articular surfaces. This may account for the major discrepancy between the specimens described here and the former: in SAM­ K401 and SAM-K1106 the anterior surface of the phalanges of digit 1 face anteriorly just as in the other digits, but Thulbom reconstructed the first digit of UCL B.1 7 with the lateral surface of the phalanges anteriorly (1972:47, fig. 12R). The data from the specimens here shows that orientation to be incorrect. The pedal phalanges of the fabrosaurids resem­ ble those of small hypsilophodontids. Existing dif­ ferences are minor: for example, the asymmetry in 167 medial-lateral phalangeal length is not present in Hypsilophodon. The phalanges of digit 1 are not reduced compared to Hypsilophodon, though MT 1 is. Even by Stormberg times reduction of digit 1 in fabrosaurids had progressed beyond that seen in most ornithopods but had not reached the point seen in Iguanodon and Dryosaurus altus (both with rudimentary MT 1). The fabrosaurid phalanges differ from H. tucki in that those of the latter had highly developed trochleas whose articular surfaces extended com­ pletely onto the dorsal (anterior) surface and per­ mitted considerable hyper-extension of the inter­ phalangeal joints. Thus the pes in the two Storm­ berg omithischians with adequately known post­ crania is functionally and structurally different: the fabrosaurid structure being fund amen tally that of small hypsilophodontid ornithopods specialized in the reduction of digit 1; that of the heterodon­ tosaurids being adapted to maximize metatarsal ri­ gidity and hyperextension of the weight-bearing digits. DISCUSSION Diagnosis of the Fabrosauridae Based on the postcranial material from sou­ them Africa the following traits characterize the Fabrosauridae: 1) pelvis in which ischium and pubis are para­ llel and retroverted and in which the ilium has an elongated anterior process (features defining fabrosaurids as omithischians); 2) an obturator process on the ischium (a fea­ ture defining fabrosaurids as ornithopods); c B A Fig. 31 SAM-K401. Phalanges of right pes, anterior view. A. Digit 1. B. Digit 3. C. Digit 4. Scale= 5 em. ... _.A Fig. 32 SAM-K401. Phalanges of right pes, lateral view. A. Digit 1. B. Digit 3. C. Digit 4. Scale= 5 em. 168 3) ilium with a brevis shelf which first turns medially and then downwards (probably a familial trait); , 4) ilium with a supra-acetabular flange over the anterior half of the acetabulum (a fea­ ture of uncertain character-state and fami­ lial validity since it is also possessed by the 'juvenile scelidosaurid' which is not a fabro­ saurid); 5) a short, deep prepubic process (a feature common to all early ornithischians in whi­ ch this area has been preserved); 6) considerably reduced first metatarsal (pro­ bably a generic rather than familial trait; variation in degree of reduction may cha­ racterize different species; has parallels in genera of other ornithopod families); 7) reduced metacarpals and phalanges (lack of specimens prevents determining whether this is a generic or familial trait). It has been difficult to judge the level of signifi­ cance which should be attached to variations be­ tween the four fabrosaurid specimens available. When samples from other genera such as Hypsilo­ phodon or Camptosaurus have been available for comparison, most of the differences between the various fabrosaurid specimens were resolvable in to individual variation. Only in the case of the sacral rib facets does this not hold up; here the UCL B.1 7 specimen differs in an unusual way from the four South African Museum ilia. This warrants generic or at least specific distinction, much in the same way as variation in sacral articulation led Gilmore (1909:281) to distinguish species of Camptosau­ rus. However, no new genera or species of fabro­ saurid will be named on the basis of this postcra­ nial material until the related dental material is de­ scribed and a total morphological assessment can be made. The diagnosis given above, especially the pelvic features, clearly distinguishes fabrosaurids from ot­ her small ornithopods such as Hypsilophodon and Dryosaurus. It also provides the criteria against which proposed assignment to the Fabrosauridae should be judged. Later sections of this discussion will examine those specimens that have postcranial material and have been assigned to the Fabrosauri­ dae. This includes Nanosaurus agilis Marsh 18 7 7 and the 'juvenile scelidosaurid; unfortunately, Echinodon cannot be considered in this context. The only other specimen assigned to the Fabro­ sauridae has been Scutellosaurus lawleri from the Kayenta formation of North America (Colbert, 1981). Of the diagnostic criteria given above, it has only the supra-acetabular flange which by itself is not diagnostic for the family Fabrosauridae. Its taxonomic status is thus uncertain. Functional morphology of the fabrosaurid postcrania The interpretation of the fabrosaurid postcrania Is based on a joint consideration of anatomical structure and relative proportions of skeletal ele­ ments. Table 1 presents a sample of ornithischian limb proportions for comparison. The general impression is that fabrosaurids had a diminutive forelimb, but this assumption can be refined by more complete data. The scapula (with­ out coracoid) is about 78% of iliac length in the UCL B.17 specimen; this indicates no reduction in length relative to any of the later ornithopods. Fa­ brosaurid scapular morphology is relatively robust: the acromial process is well developed, minimum shaft width is large and the coracoids are 25% of the scapular length. Morphology and proportions both show that the scapula was not in any way re­ duced. The proportion of fabrosaurid humeral to sea~ pular length also shows no sign of forelimb re_duc­ tion. The fabrosaurid ratios are all near 90%, slight­ ly smaller than that for Hypsilophodon and Thes­ celosaurus (Galton 1974a), in which the humerus is as large as or larger than the scapula, but larger than that of Iguanodon, camptosaurids and hadro­ saurids. Morphologically the humerus shows. no signs of reduction: the deltopectoral crest nses about 5 mm above the humeral surface; a tubercle for muscle attachment lies opposite the base of the deltopectoral crest on the medial margin of the shaft, perhaps associated with the insertion of_ the coracobrachialis muscle (for humeral protractwn) or the origin of the humero-radialis (brachialis) muscle (for forearm flexion). In the forearm the fa­ brosaurid radius is about 60% of scapular length, smaller than in the other ornithopods except cam­ tosaurids and Iguanodon. Morphologically the ra­ dius shows no signs of robusticity as the humerus and scapula do. The size of the manus can only be compared using metacarpal 3 since no single digit has all the phalanges and only UCL B.1 7 has a partial manus. The fabrosaurid third metacarpal is only 20% of scapular length; this is shorter than that of mos.t ornithopods except Iguanodon, but that of Hypsz­ lophodon is so close that the difference is probab­ ly insignificant. Compared to the radius, MC 3 is relatively larger than that of other ornithopods; but this is misleading because the fabrosaurid ra­ dius itself is relatively smaller as noted above. Compared to a group of ornithopods the fabro­ saurid forelimb is reduced in only its distal elemen­ ts; the scapula and humerus are comparable in rela­ tive size or even larger than those of many later or­ nithopods. Only compared with H. tucki does ~he fabrosaurid forelimb show the degree of reductiOn already undergone in the ornithopod lineage by the late Triassic/early Jurassic. Each segment of the fa­ brosaurid anterior extremity is relatively shorter; culmulatively this results in a much shorter fore­ limb, both relatively and absolutely. When the measurements of humerus, radius and longest metacarpal (MC 2 in H. tucki) are added, the UCL B.17 specimen (107 mm) is just two-thirds that of H. tucki (154 mm). The difference would be grea­ ter if the entire fabrosaurid hand were preserved since phalanx 1 of its digit .3 is only 5 mm long but TABLE 1. RELATIVE PROPORTIONS OF SKELETAL ELEMENTS IN EARLY ORNITHISCHIANS AND VARIOUS ORNITHOPODS Sc/11 H/Sc R/Sc R/H MC3/Sc MC3/R Tr4 T/F T+MT3/F MT3/F MT3/T Fabrosaurids UCLB.17 (1) 0,78 0,88 0,56 0,64 0,19 0,34 0,37 1,24 1,89 0,64 0,52 SAM-Kll06 -- 0,91 0,62 0,68 -- -- (K400) -- -- -- 0,49 0,39 H etero dontosaurus tucki (2) 0,89 0,97 0,68 0,70 0,27 0,40 0,41 1,29 1,90 0,61 0,47 Hypsilopho don foxii (3) BM~NH~ R 196 0,73 1,0 0,79 0,79 0,23 0,29 0,43 1,18 1 '73 0,56 BM NH R 5830 -- -- -- -- -- -- 0,42 1,17 1,79 0,62 0,53 Kritosaurus incurvimanus ( 4,3) 0,77 0,81 0,72 0,88 0,29 0,41 0,54 0,90 1,24 0,35 0,38 Parksosaurus warreni ( 5,3) 0,70 0,95 -- -- -- -- 0,49 1,18 1, 74 0,56 0,47 Camptosaurus nanus (6,3) 0,77 0,77 0,24 0,31 -- -- 0,54 0,95 1,31 0,40 0,42 Micro cera tops gobiensis (7) -- 0,95 0,67 0,70 -- -- 1,16 1, 72 0,56 0,48 Iguanodon atherfieldensis (8,3) 0,85 0,67 0,47 0,71 0,20 0,42 0,48 0,88 1,23 0,35 0,40 Abbreviations: F, femur; H, humerus; Il, ilium; MC3, third metacarpal; MT3, third metatarsal; R, radius; Sc, scapula; T, tibia; Tr4, fourth trochanter index (length from femoral head to root of trochanter/femoral length). Sources of measurements: 1, Thulborn 1972; 2, Santa Luca 1980; 3, Gal- ton 1974; 4, Parks 1920; 5, Parks 1926; 6, Gilmore 1925; 7, Maryanska and Osm6lska 1975; 8, Hooley 1925. ...... Ol l.O 170 phalanx 1 of digit 2 (the longest digit) in H. tucki is 16 mm long. Thus the shortness of the phalanges of the fabrosaurid manus indicates even more than metacarpal length that fabrosaurid and heterodon­ tosaurid hands had quite different functions. The fabrosaurid ilium differs in three essential respects from almost all other ornithopod ilia. The depth of the postacetabular process is relatively great due to the more vertical orientation of the brevis shelf; the supra-acetabular flange makes a bony hood over the acetabulum; and the medial acetabular border formed by the ilium was partially ossified. The depth of the postacetabular process (mea­ sured at its midpoint) is 28% of iliac length, twice that of Hypsilophodon and more than three times that of H. tucki (8,3%). The postacetabular process is deep in camptosaurids, in which the brevis shelf also curves downwards to become more vertical, especially in C. dispar and C. medius (Gilmore 1909: Plates 15-16). Discounting the brevis shelf, the fabrosaurid postacetabular process is no deeper than that of Hypsilophodon. The more vertically disposed brevis shelf does not necessarily mean a different fiber direction for the caudifemoralis ( coccygeofemoralis) brevis mus­ cle which attaches there. It does however give alar­ ger attachment area for this muscle; the shelf is at least 13-15 mm in extent. If such a large shelf were horizontally disposed, then the medialmost fibers of caudifemoralis brevis would have a strong line of action medially (i.e. adduction) as well as posteriorly (i.e. retraction). The brevis shelf of H. tucki is horizontal and only 4-6 mm wide though the ilia of both are about the same absolute length. The supra-acetabular flange must have affected femoral orientation, at least to the extent that the femur would be placed in some way so that the flange did not interfere with muscular attachments to the lesser trochanter. The femoral head may have articulated underneath the flange which would have transmitted weight between body and limb. Evidence for this is the allometric increase in lateral extent and in thickness of the flange with increasing iliac size. The distal femoral articular surface is horizontal, indicating a vertical orienta­ tion of the femoral shaft. The medial acetabular wall formed by the ilium is more extensively ossified than in all other orni­ thopods, except perhaps camptosaurids. Normally the acetabulum is open medially to the level of the lateral upper acetabular margin. From the upper la­ teral margin the fabrosaurid acetabular surface turns inward to the cup-shaped undersurface of the flange and then down wards into a vertical wall; the ossification reaches almost to the lower ends of the ischiadic and pubic peduncles. This seems to be si­ milar to the drawing of the 'juvenile scelidosaurid, given by Charig ( 1972:123, fig. 2). The partial ossi­ fication of the medial acetabular wall probably pre­ vented the femoral head from seating directly un­ derneath the body of the ilium; the supra-acetabu­ lar flange may have compensated for this by exten- ding the acetabular articular surface laterally and allowing the direct transmission of weight and for­ ces between the femur and the ilium. Thus the three special morphological features of the ilium may all be functionally interrelated. The incompletely perforate acetabulum displaced a vertically oriented femoral shaft lateral to the main mass of the iliac blade; a supra-acetabular flange transmitted forces through the more laterally pla­ ced femoral shaft and a vertical brevis shelf kept the mass of the caudifemoralis brevis muscle more lateral and in the same parasagittal plane as the fe­ mur much more than a horizontal shelf would. Wh­ ile none of this can be proven, it is one logically and functionally consistent explanation of the morphological peculiarities seen in the fabrosaurid ilium. The hindlimb ratios of fabrosaurids are in the high end of the range usually associated with rapid bipedal progression in ornithischians (see Table 1 ). The extent to which the distal hindlimb elements exceed the femur in length presumably indicates the extent to which the demand for rapidity of hindlimb stroke exceeds the demand for strength of hindlimb stroke. The 4th trochanter is located higher on the fe­ moral shaft than in any other ornithopod, in a po­ sition more advantageous for rapid than for power­ ful femoral retraction. The tibia/femur and tibia + MT3/femur ratios are also higher than in any other ornithopod; in the latter the tibia accounts for most of the difference and is relatively the most elongated while the relative length of MT 3 compares well with that of Hypsilophodon and does not demonstrate a greater degree of elonga­ tion. The relative length of the tibia is even greater in H. tucki than in fabrosaurids, presumably indica­ ting more highly developed bipedal capabilities. This is all the more remarkable since the forelimb of H. tucki is also relatively elongated and must have had a specialised function. The tibial ratios of Microceratops gobiensis also do not differ from those of Hypsilophodon, and the former must be considered to be as well adapted for rapid bipedal progression as the later. Thus both H. tucki and M. gobiensis demonstrate the existence of cursorial trends in non-ornithopod lineages. Fabrosaurids and the ancestry of early ornithischians The most important question for ornithischian phylogeny is whether a fabrosaurid could be ances- tral to the early ornithischians Heterodontosaurus and the 'juvenile scelidosaurid'. A brief comparison of the postcrania of fabrosaurids and H. tucki ap­ peared in the description of the latter specimen (Santa Luca 1980: 199-200). The conclusions rea­ ched there, that fabrosaurids and H. tucki represent independent lineages and that the fabrosaurid line­ age could not be ancestral to heterodontosaurids, are reinforced by further knowledge of fabrosaurid postcranial anatomy. Differences in structure are 171 TABLE 2. POSTCRANIAL SKELETAL DIFFERENCES BETWEEN FABROSAURIDS AND HETERODONTOSA UR US TUCK/ SKELETAL ELEMENT cervical centra dorsal neural arch ossified tendons scapula, distal end scapula, shaft glenoid cavity deltopectoral crest distal humerus ulna metacarpals ungual phalanges joints of manus ilium anterior iliac process acetabulum ischium prepubic process proximal trochanters greater trochanter distal candy les intercondylar fossa transverse axis tibia-fibula as tragal us-calcaneum distal tarsals BP - M F ABROSAURIDS (UCL B.17, SAM-K400, K401 and K1106) asymmetric constriction of lateral surface small ventral keel wider posteriorly spine wider at top than at base (in K401); constant in width (in UCL B.1 7) in dorsal and caudal region hook-like process at ventral comer rectangular in X-section above glenoid deep (9 mm) gracile, smaller en tepicondy le ·absent anterior and posterior intercondylar grooves no olecranon process I-III small, IV-V reduced unspecialized simple trochlea supra-acetabular flange brevis shelf deep and nearly vertical no accessory articular surface at ischiadic peduncle about 1/3 total length, simple spine-like process medial iliac ossification obturator process present relatively longer cleft separates greater and lesser extends,above femoral head lateral larger posterior only distal almost horizontal separate separate 2, unfused, flat H. tucki (SAM-K1332) strong vertical ridges anteriorly and posteriorly prominent keel, ventral surface deeply concave spine rectangular limited to dorsal region smaller, attenuated process at ventral corner circular in X-section above glenoid shallow (4 mm) robust, larger entepicondyle present no posterior intercondylar groove large olecranon process 1-111 large, IV-V reduced prominent flexor tubercles specialized trochlea permitting hyperextension of digit I-III no supra-acetabular flange brevis shelf small and horizontal accessor 'avian-like antitrochan­ ter' at ischiadic peduncle about 1/2 total length; thicken­ ed, terminating in knob-like pro­ cess completely perforate obturator process absent relatively shorter greater and lesser continuous level with femoral head medial larger none oblique fused tibiotarsus fused; strong lateral and medial flanges 3, fused, lateral and medial flan­ ges 172 found in all areas of the axial and appendicular ske­ leton; Table 2 presents a summary of these diffe­ rences which, taken together, substantiate the exis­ tence of divergent structural patterns in early orni­ thischians. Some features listed in Table 2 do not preclude a fabrosaurid ancestry for heterodonto­ saurids and the condition in the latter could be de­ rived from that of the former. For example, the olecranon process, entepicondyle, specialized inter­ phalangeal joints and thickened anterior iliac pro­ cess of heterodontosaurids could all be derived from the fabrosaurid condition. Of course, the he­ terodontosaurid features are probably all develo­ ped from primitive conditions and so do not prove a special phylogenetic relationship between fabro­ saurids and heterodontosaurids. However, primitive features aside, the structure of fabrosaurids is so ornithopod in its specializa­ tions that it cannot serve as a precursor for hetero­ dontosaurids. The principal features excluding them from heterodontosaurid ancestry are: 1) the reduced manus; 2) the presence of an obturator process; 3) the presence of only two distal tarsals; and 4) the extreme reduction of metatarsal 1. If the Fabrosauridae, as the earliest well known or­ nithopod family, cannot be ancestral to the only other well known early ornithischian family, the Heterodontosauridae, then the concept of ornitho­ pods as the basal stock for all other ornithischians radiations is questionable. Because the hypothesis has been so important in discussions of ornithis­ chian evolution, it will be treated separately in a later section. While the differences listed in Table 2 illustrate the existence of ornithopod and ·non-ornithopod lineages in Stormberg times, the similarities betwe­ en fabrosaurids and H. tucki point out the most likely ancestral features for ornithischians as a whole. In the postcranial skeleton these are the sca­ pular acromial process, a projecting spine-like ante­ rior iliac process, a short but deep prepubis and a straight postpubic rod equal in length to the ischia­ dic rod. The presence of the same pelvic features in the 'juvenile scelidosaurid' (BMNH R6 704; Charig 1972: 138, Plate VIA) reinforces their interpreta­ tion as ancestral traits. Fabrosaurids have also been compared (Thul­ born 1977) with the undescribed but illustrated (Newman 1968; Rixon 1968; Charig 1972) remains of the 'juvenile scelidosaurid' from the Lower Lias of England. Its systematic position is still uncertain but Thulborn (1977) argued that it is actually a cursorial ornithopod unrelated to Scelidosaurus harrisonii Owen 1861 (interpreted by Thulborn as an ankylosaur) and a direct descendant of the Stormberg fabrosaurids of southern Africa. Thul­ born's method seriously biased the results of the comparisons since he only used two extremes in as­ sessing the relationship of the 'juvenile scelidosau­ rid': the UCL fabrosaurids, and ankylosaurs. The fact that the 'juvenile scelidosaurid' did not resem­ ble ankylosaurs at all is neither necessary nor suffi­ cient reason to classify it in the only other group used for comparison. Thulborn also made the erro­ neous assumption that all bipedal ornithischians must be classified in the suborder Omithopoda; this prejudiced the systematic and phylogenetic assessment of the specimen since many bipedal or­ nithischians are demonstrably not ornithopods (Heterodontosaurus, Microceratops, pachycephalo­ saurs, Protiguanodon and Psittacosaurus). Thulborn cited eight traits shared by the UCL fabrosaurid and the 'juvenile scelidosaurid' to sup­ port his hypothesis that the former was ancestral to the latter: 1) scapula slender and constricted in midshaft; 2) femur with a pendent fourth trochan­ ter , 3) greater and lesser trochan ters divided by a cleft; 4) extensive postacetabular process; 5) ilium with a simple dorsal margin; 6) a distinctive swell­ ing above the perforate acetabulum (presumably the supra-acetabular flange; 7) a short prepubis; and 8) postpubis equals length of ischial rod. Most of these characters are not definitive for a systematic assignment and are widely distributed throughout the Ornithischia as primitive retentions. Character 1 is typical of small ornithischians and equally true for the non-ornithopods Microceratops and Heterodontosaurus as for fabrosaurids. Charac­ ters 2 and 3 are not restricted to ornithopods eith­ er but are present in, for example, Leptoceratops, Protoceratops, Psittacosaurus and Protiguanodon. Character 4 is present in pachycephalosaurs and the Ceratopsia as well as the Ornithopoda. Charac­ ter 5, presumably meaning that the dorsal margin is not everted as in ankylosaurs, is also found in Stegoceras and the less modified ceratopsians such as Leptoceratops. Character 6 does not appear in other known omithischians besides these two. Cha­ racters 7 and 8 are primitive for the order Ornithis­ chia, characteristic of Heterodontosaurus as well as of primitive ornithopods. Therefore, of these 8 traits only one has any significance in comparing fabrosaurids with the 'juvenile scelidosaurid'. Thulborn also noted four differences between the two specimens but dismissed them as irrelevant compared to the many similarities he listed. Since only one ·similarity of any consequence exists then these differences should be more carefully evalua­ ted: 1) dermal armor, present in the 'juvenile sceli­ dosaurid', but presumably absent in fabrosaurids; 2) sacral vertebrae, four in the 'scelidosaurid, five in the UCL fabrosaurid; 3) ischial rod and postpu­ bis, concave ventrally in the 'scelidosaurid', convex ventrally in the fabrosaurid; and 4) obturator pro­ cess, absent in the 'scelidosaurid', present in fabro­ saurids. Character 1 may not be a difference at all if the North American specimens (Colbert, 1981), which do indeed retain dermal armour, are actually fabro­ saurids; otherwise it would be difficult to accept the reappearance of armour once it had been lost. Character 2 means very little since the number of sacral vertebrae, determined by the number of sac­ ral ribs, is seen to vary when sufficient specimens of a single genus, or even species, are known. Cha­ racter 3 is equivocal in meaning; a ventrally conca- ve ischium and postpubis is common in quadrupe­ dal ornithischians (Ankylosaurus, Leptoceratops, Brachyceratops, Monoclonius) but not universally (these elements are straight in ~tego_saurs ). C?arac­ ter 4 is given much importance In this paper; In any single instance t?e.loss o! the obt~rator proce_ss du­ ring phylogenesis IS possible. But Its absence In ~he 'juvenile scelidosaurid' takes on. anothe: ~~aning once it is accepted that absence IS the pn~rutlve or­ nithischian condition. Without strong evidence of derivation from the fabrosaurids (and one shared character of unknown character-state and functio­ nal significance is not strong evidence), absence could equally well mean retention of_ a primitive condition rather than a case of reduction and loss. Two differences not noted by Thulborn make the derivation of the 'scelidosaurid' from the fabro­ saurid less likely. First, the scapula is relatively lar­ ger in the 'scelidosaurid'; scap~la len&t~Jfemor_al length is about 90% in the 'scehdosaund , 63% In the smaller UCL fabrosaurid. Second, the fourth trochanter is more proximally located in fabro­ saurids; the position index (length from level of femoral head to base of 4th trochanter...;- femoral length) is 36,5% in the smaller UCL fabrosaurid, but 50% in the 'juvenile scelidosaurid'. These in­ dices signify somewhat different adaptive strate­ gies and it seems unlike!y that the 'j~venile sceli­ dosaurid' would be denved from a lineage com­ mitted to forelimb reduction and a rapid hind­ limb stroke. The only significant similarity which can be demonstrated between fabrosaurids and the 'ju­ venile scelidosaurid' is iliac structure, the everted dorsal acetabular margin and the dorsoventrally deep postacetabular process. To ~his is oppo~ed differences in the ischium, prepubis, scapular size and fourth trochanter position. The data only par­ tially resolve the question of fabrosaurid relation­ ship to the 'juvenile scelidosaurid'. The only infe­ rence supported by the data is that the latter diff­ ers sufficiently from the former so as not to be cla­ ssified within the Fabrosauridae (the absence of the obturator process is sufficient for that). Before concluding this discussion, one viable alternative hypothesis deserves consideration. In­ stead of the Thulborn (1977) scenario discussed above, it is possible that the 'juvenile scelidosaurid' was ancestral to the fabrosaurids. The 'juvenile sce­ lidosaurid' is a better structural precursor for the fabrosaurid than vice versa: the smaller prepubis (not the result of reduction but the absence of de­ rived elongation common in ornithopods), the ab­ sence of the obturator process and the relatively larger scapula are probably all primitive relative to the fabrosaurids. The relatively lower 4th trochan­ ter is also expected in the fabrosaurid ancestor, which had not yet become so specialized for bipe­ dal locomotion. The supposed discrepancy in time between Stormberg and Lower Lias probably acted to formulate the hypothesized relationship as fab­ rosaurid - 'juvenile scelidosaurid'. However, the Lower Lias may not be any more recent than the 173 Upper Red Beds (now Elliot Formation) of the Stormberg. Long ago Broom (1911:307) consider­ ed the Cave Sandstone (now Clarens Formation) to be Lower Jurassic in age and recently Olsen and Galton (1977; see also Olsen and Galton, this volu­ me) have revised the correlations of the Newark, the Glen Canyon group and the Stormberg. They offered vertebrate paleontological, palynological and ichnological evidence that the upper part of the Newark (zone 3), the Glen Canyon group and the Upper Stormberg were all Liassic in age (Het­ tangian and Sinemurian). Since the scelidosaurids, both 'juvenile' and the 1861 type, were found in the Sinemurian of England and the fabrosaurids in the Upper Elliot Formation of the Stormberg they were probably contemporaneous. Thus there is no chronological or stratigraphic reason to reject an inversion of the hypothesized phylogenetic rela­ tionship and certainly no morphological reason to reject such a possibility. Review of these three early ornithischians with good postcranial material has demonstrated a grea­ ter degree of evolutionary divergence than can be subsumed within the family Fabrosauridae and even within the suborder Ornithopoda. The 'juve­ nile scelidosaurid' cannot be assigned to the Fabro­ sauridae for the reasons noted above; since its pel­ vic structure differs from all other known ornithi­ schians including Heterodontosaurus it probably belongs to an entirely new family. Its sub ordinal status is also indeterminate because, unlike the fa­ brosaurids, it cannot be associated with any late Jurassic/Cretaceous ornithischian suborder. Its sta­ tus as an ornithopod is questionable since it lacks an obturator process. As with Heterodontosaurus its subordinal assignment must await further dis­ coveries of early ornithischians. H. tucki cannot be assigned to the Fabrosauridae either for the many reasons noted above, nor can it be assigned to the Ornithopoda for reasons of pelvic structure. The relationship between fabrosaurids and later ornithopods The fabrosaurids have been offered as a model of the primitive, archetypical ornithopod from which all later ornithopods, particularly hypsilo­ phodontids, could be derived (Galton 1978; Thul­ born 1971. 1972). This proposition was supported primarily on the basis of shared primitive features of the skull and dentition but without detailed consideration of the postcranial material. The fabrosaurids clearly belong to the ornitho­ pod lineage but several features suggest that they were on an adaptive pathway different from that of the ancestors of the late Jurassic/Cretaceous or­ nithopods. Three features distinguish them from al­ most all later ornithopods: the supra-acetabular flange, the short prepubic process, and the partial acetabular ossification of the ilium. The short pre­ pubic process is primitive for the Ornithischia, so its absence in later ornithopods does not automa­ tically eliminate the fabrosaurids from their ances- 174 try. The phylogenetic significance of the supra-ace­ tabular flange is more difficult to evaluate. Mor­ phologically it is an accentuation of a ridge on the anterior acetabular margin which is typically pre­ sent in many omithischians. The flange could as easily be derived from the ridge as the ridge be de­ rived by reduction of the flange. The flange seems to be present in the 'juvenile scelidosaurid' (Charig 1972:123, fig. 2,1972:138, Plate VIA) but is clear­ ly absent in Heterodontosaurus. The flange also seems to be present in the various species of camp­ tosaurids. Gilmore ( 1909:25 7) noted that the pu­ bic peduncle of the ilium was quite wide. His pho­ tographs (Gilmore, 1909:256, fig. 29; 257, fig. 30; Plate 16) show a flaring process over the anterior half of the acetabulum, though perhaps not as pro­ minent as that seen in the fabrosaurids. The one remaining feature of the fabrosaurid pelvis not easily matched in later ornithopods is the structure of the brevis shelf. In fabrosaurids it is deep and curves downward to become nearly ver­ tical, and does not terminate by forming a horizon­ tal shelf under the postacetabular process. The ab­ sence of a fabrosaurid-like brevis shelf in most later ornithopods seems to isolate the fabrosaurid linea­ ge from any other except perhaps camptosaurids. Finally, the first digit of the fabrosaurid pes is considerably reduced, much more so than that of Hypsilophodon, Parksosaurus and Thescelosaurus, but comparable to Camptosaurus. Such reduction by late Triassic/early Jurassic time may take known fabrosaurids outside the ancestral lineage of the three former ornithopods. The pes in Dryosau­ rus/Dysalotosaurus seems to have an even more re­ duced first metatarsal Qanensch 1955; Galton 1977), although Galton (1981) restored a large MT 1 to Dry osaurus let to wvorb ec ki. On the basis of this postcranial survey fabrosau­ rids can be separated from the ancestry of some late Jurassic/Cretaceous ornithopods notably Hyp­ silophodon. In other cases the data are somewhat equivocal, especially when the structural differen­ ces pertain to probable ancestral traits in the fabro­ saurids. However, fabrosaurids show more parallels with the various camptosaurids; iliac morphology in the latter is more similar to that of fabrosaurids (including the supra-acetabular flange and the bre­ vis shelf) than to that of Hypsilophodon, which differs by having a horizontal brevis shelf and no development of a supra-acetabular flange. The most reasonable conclusion to be drawn is that fa­ brosaurids display some trends present in the later camptosaurids but the postcranial data are not suf­ ficient to support the contention that known fa­ brosaurids were the common ancestors of all later ornithopods. Fabrosaurids and Nanosaurus agilis Marsh The only postcranial material outside of south­ em Africa assigned to the Fabrosauridae is that de­ scribed as Scutellosaurus lawler£ by Colbert ( 1981) and the very fragmentary remains of Nanosaurus agilis Marsh 1877 (Galton 1978). Amongstthelat- ter is a matrix slab containing postcranial material, Yale Peabody Museum 1913 b-g, associated with a partial dentary in the same piece of matrix, YPM 1913a. Galton interpreted the dentary as a fabro­ saurid since the dentition was marginally placed, a characteristic of his family Fabrosauridae (1972, 1973 ). The dentary teeth do not preseYve any sur­ face morphology and only their outline remains. The postcranial material of N. agilis referred to comprises an ilium, two femora, two tibiae and a fibula, but none of it can unequivocally be assig­ ned to the Fabrosauridae. The Nanosaurus ilium presumably had a well developed supra-acetabular flange, but it differs in structure from that of the fabrosaurids in which the flange is a lateral exten­ sion of the acetabular margin along the pubic pe­ duncle and anterior half of the dorsal acetabular margin. The Nanosaurus fragment has a reflected margin over the entire dorsal acetabular area and not just the anterior half. In addition, nothing in­ dicates that the presumed anterior iliac process was deflected inferiorly; in fact it is straight as figured by Huene and Lull ( 1908: fig. 4 ). The dor­ sal margin of the ilium seems to be broken, so it cannot be accurately reconstructed. Even as it is drawn in Huene and Lull ( 1908) and Galton (1978:148, fig. 4A), the depth of the ilium above the acetabular margin is deeper than in any of the known fabrosaurid ilia . Several features of the posterior iliac process al­ so separate Nanosaurus from the fabrosaurids. In the former an oblique ridge courses backwards and upwards to the dorsal comer of the process ; in fa­ brosaurids this ridge, which divides the brevis shelf from the remainder of the posterior iliac process, is directed almost straight backwards and only slightly upwards, so that its highest point never rea­ ches above the level of the superior acetabular margin. Finally, the posterior acetabular process of Nanosaurus is truncated, unlike any known orni­ thopod, comprising only 20% of iliac length; the process in fabrosaurids (the horizontal distance from the posterior edge of the ischiadic peduncle to the end of the ilium) is 33% of total iliac length. The two Nanosaurus femora are equally inde­ terminate in structure. The lesser trochanter does not seem to be separated by a cleft from the grea­ ter trochanter even though enough of the proximal end is preserved in both femora to have shown the cleft had it been there. The Nanosaurus tibiae and fibula also lack any diagnostic fabrosaurid charac­ ter. Only one articular surface is preserved, the dis­ tal end of a tibia, and this is not diagnostic for fa­ brosaurids; the remaining bones are parts of the shaft and are not distinctive in any way. In summary, the postcranial skeleton of Nano­ saurus agilis offers no definitive evidence of fabro­ saurid affinities; it does, on the other hand, have several features incommensurate with fabrosaurid postcranial anatomy: the structure of the posterior iliac process and the absence of a cleft between greater and lesser trochanters. Therefore, the post­ cranial material of Nanosaurus agilis Marsh 1877 should be removed from •the family Fabrosauridae and classified £ncertae sed£s. ORNITHISCHIAN EVOLUTION It is generally assumed that ornithopods for­ med a basal ornithischian stock from which the ot­ her ornithischian suborders were derived. However, this hypothesis needs to be examined for several reasons: first of all, the definition of the Ornitho­ poda focuses on primitive ornithischian traits; se­ cond, since the Microceratopsidae were bipedally adapted and cranially similar to the Ceratopsia, there is no need to invoke the ornithopods as the structurally conservative ancestor of the ceratop­ sians; and last, the early ornithischians, the fabro­ saurids and heterodontosaurids, are contempora­ neous but structurally quite divergent, which poin­ ts to a division in ornithischian evolution even at this early point. For all practical purposes most bipedal orni­ thischians have been called ornithopod. Yet, pla­ cing all bipedal ornithischians within a single cate­ gory makes the unwarranted assumption that bipe­ dalism appeared only once in ornithischian phy­ logeny and that all bipedal forms are closely rela­ ted. However, if bipedalism was the rule for the an­ cestral ornithischians, then bipedalism is a primitive character. And as a primitive character its appea­ rance in different ornithischians is neither nece­ ssary nor sufficient evidence of phylogenetic rela­ tionship, nor is it evidence for the erection of a category based on this character. Perhaps the best example of the difficulties caused by defining all bipedal ornithischians as or­ nithopods is presented by the genus M£croceratops from Mongolia. This little ornithischian has a small cervical collar, or frill, characteristic of the Ceratop­ sia, but the skull is associated with a hindlimb adap­ ted for bipedal locomotion. In this case, using the bipedal definition of ornithopods would lead to the inconsistency of classifying a form clearly rela­ ted to the Ceratopsia in its specializations to the Ornithopoda, with which it shared the primitive or­ nithischian form of locomotion. To classify and to determine phylogenetic relationship on the basis of such a symplesiomorphy is certainly not valid. Another reason for questioning the hypothesis that ornithopods form a basal ornithischian stock comes from the known early ornithischians them­ selves. If ornithopods were a basal stock, then the early ornithischians should show clear affinity to them; however, only fabrosaurids do so. But Hete­ rodontosaurus is so divergent in structure that it cannot be derived from fabrosaurids or be ances­ tral to later ornithopods. At this historical point in the study of the ornithischian phylogeny it seems necessary to re­ vise the definition of the Ornithopoda to avoid these inconsistencies. A feature, or group of fea­ tures, must be found which unites closely related forms--fabrosaurids, hypsilophodontids, thescelo­ saurids, camptosaurids, iguanodontids and the had­ rosaurids--yet at the same time separates other forms which share with the former only primitive 175 features such as bipedalism. The one postcranial feature which meets this criterion is the presence or absence of the obturator process of the ischium. This flange-like process which projects from the ventral surface of the ischium has a very limited distribution throughout the Ornithischia. It is pre­ sent only in hypsilophodon tids, thescelosaurids, camptosaurids, iguanodontids and hadrosaurids of the later ornithopods. That is, the process is pre­ sent only in those forms which have always been grouped together. The process is absent, however, in every other ornithischian group: it is absent in all the other bipedal forms like Prot£guanodon, psi­ ttacosaurids, pachycephalosaurs and Heterodonto­ saurus, and absent in all stegosaurs and ceratop­ sians. Consequently, ornithopods should be rede­ fined on the basis of this character and should in­ clude only those ornithischians that have an ob­ turator process on the ischium. The real importance of this new definition of ornithopods depends, of course, on whether the obturator process is a primitive or derived feature. If it were primitive then nothing would be clari­ fied in the analysis of ornithischian phylogeny sin­ ce one primitive character would have been substi­ tuted for another and a paraphyletic grouping would have resulted again. The data suggest that the obturator process is probably a derived feature while the absence of the process is primitive. Most importantly, the obtura­ tor process itself only appears in a group of closely related forms, those ornithopods noted above; thus, the feature seems to be circumscribed within a sin­ gle lineage. Primitive traits, on the other hand, ge­ nerally have a rather random distribution through­ out unrelated genera. For example, the prin1itive and simple ornithischian tooth form is found in some ornithopods, stegosaurs and ankylosaurs. So the distribution of the obturator process conforms more closely to that expected of a derived trait. The pelvic morphology of the early ornithischians also lends support to this view. Pelvic structure is known well only in the early ornithischians discu­ ssed in this paper and only the fabrosaurids have an obturator process. Neither Heterodontosaurus nor the 'juvenile scelidosaurid' has one; the absence of the process in these two forms which are not themselves closely related, matches the distribution expected of a primitive trait. Similarly, a primitive ornithischian pelvic feature such as the s