Small-bodied ornithischians with the following features that may constitute heterodontosaurid synapomorphies in phylogenetic context: (1) three or fewer premaxillary teeth; (2) premaxillary teeth increase in size distally; (3) dentary caniniform tooth associated with an arched premaxilla-maxilla diastema; (4) nasal fossa, dorsomedian with rounded lateral margins; (5) jugal flange, ventral embayment of jugal-quadratojugal embayment; (6) jugal horn below orbit, laterally directed and dorsoventrally compressed; (7) postorbital body, arcuate fossa with raised anterior rim; (8) quadrate head included within laterotemporal fossa; (9) quadrate condyle, articular surface ventrolaterally inclined at approximately 30°; (10) quadratojugal T-shaped; (11) predentary processes (lateral, ventral) rudimentary; (12) dentary ramus stoutly proportioned, substantial depth at mid ramus compared to length; (13) fibular mid-shaft and distal end reduced.

The most inclusive clade containing Heterodontosaurus tucki Crompton and Charig 1962 but not Parasaurolophus walkeri Parks 1922, Pachycephalosaurus wyomingensis (Gilmore 1931), Triceratops horridus Marsh 1889, Ankylosaurus magniventris Brown 1908.

This stem-based phylogenetic definition (Sereno 2005b) includes, but does not reach beyond, all currently known heterodontosaurids under all proposed phylogenetic interpretations of the position of heterodontosaurids within Ornithischia (e.g., Sereno 1999; Butler et al. 2008). The first and only previous phylogenetic definition proposed for Heterodontosauridae (Sereno 1998: 61) is similar but lacks the negative specifiers of the present definition that stabilize its taxonomic content under alternative phylogenetic relationships.

Late Triassic (Norian) to Early Cretaceous (Barremian-Aptian), ca. 216–125 Ma (Gradstein and Ogg 2009; Martinez et al. 2011); global distribution includes northern localities (northern China, western North America, Europe) and southern localities (southern South America, southern Africa) (Fig. 1). The record of heterodontosaurids from the Late Triassic currently depends upon the interpretation of the poorly known Pisanosaurus mertii (Bonaparte 1976; Sereno 1991) and other fragmentary remains from Upper Triassic rocks elsewhere in Argentina (Báez and Marsicano 2001).

Kuhn (1966) is identified as the author of the taxon Heterodontosauridae, although Romer (1966) independently proposed the same taxon in the same year (synchronous publication noted by Kuhn 1967: 77, 122). In the literature, some cite Romer as the author of the taxon (e.g., Smith 1997; Sereno 1998; Sereno 2005b), some Kuhn (e.g., Norman et al. 2004, 2011), and some Kuhn and Romer with one author in parentheses (e.g., Steel 1969). Establishing priority by publication date in this case is no longer possible, and, unlike Romer, Kuhn also briefly diagnosed the family-level taxon. Here Kuhn is recognized as the author of Heterodontosauridae (P. Galton, pers. comm.).

Many of the cranial and postcranial apomorphies listed in the emended diagnosis were known previously only in Heterodontosaurus tucki but now are known in at least one other heterodontosaurid. When coded into a phylogenetic analysis, some of these features might be repositioned at nodes within Heterodontosauridae (under delayed transformation), given the large amount of missing data in known taxa. The list, nonetheless, attempts to capture as many skeletal modifications that are shared by Heterodontosaurus tucki and at least one other basal heterodontosaurid and may characterize the group. The features listed are discussed in more detail below (under Heterodontosaurid monophyly) and in Appendix I.

NHMUK 48209 (Figs 11, 13C, D, part of the left and right premaxillae, anterior part of left maxilla with the caniniform tooth and maxillary teeth 2 and 3, and an impression of the lateral aspect of the posterior ramus of the maxilla and maxillary teeth 4 and 5; NHMUK 48210 (Figs 13A, B, 14), posterior ramus of the maxilla with 6 alveoli and 5 complete or partial crowns, the ventral end of the left lacrimal, the anterior end of the left jugal, and most of the left ectopterygoid. Both specimens belong to the anterior end of a single, partially disarticulated snout embedded in a block that split between the maxillae during, or shortly after, its collection (Owen 1861: pl. 8, Figs 1, 2; Galton 1978: Figs 1A, B; Norman and Barrett 2002: pl. 1, Figs 1, 2).

NHMUK 48211 (Fig. 12), partial right maxilla with maxillary teeth 2-7 with the tip of the right jugal and part of the right palatine; NHMUK 48212, partial right maxilla with 6 teeth; NHMUK 48213 (Fig. 18), partial left dentary with 8 alveoli and 7 teeth; NHMUK 48214, partial right dentary without teeth; NHMUK 48215a, right dentary with 10 alveoli and 9 teeth; NHMUK 48215b (Figs 15–17), left dentary with 10 alveoli and 5 teeth.

Referred material. NHMUK 48229, jaw fragment; NHMUK 40723, dentary fragment; DORCM GS 1164-5, 1167, 1171, 1194, 1212-6, 1222-3, isolated teeth.

“Mammal Pit” located high on the coastal cliff section near the Zig Zag Path at Durlston Bay, Isle of Purbeck, Dorset, southern England (Fig. 1A); N50°35', W1°55' (Melville and Freshney 1982; Salisbury 2002).

Either the Marly or the Cherty Freshwater Member, Middle Purbeck Beds of the Purbeck Formation; Lower Cretaceous, Berriasian, ca. 146-140 Ma (Owen 1861; Clements 1993; Salisbury 2002; Gradstein and Ogg 2009). Although Galton (1978: 151) stated that the lectotype and paralectotype material came from the Mammal Bed (= “dirt bed”; Durlston Bay 83) of the Middle Purbeck Beds (Marly Freshwater Member), no evidence exists to link the fossils to that particular horizon. They may have come from the Feather Bed (Durlston Bay 108) slightly higher in the section (Cherty Freshwater Member).

The species name becklesii, coined by Owen (1861: 35)after Samuel H. Beckles who found the fossils, has been misspelled several ways in the literature (Norman and Barrett 2002), first as “becclesii” by Owen himself (1861: pl. 8) and later as “becklesi” (Lydekker 1888) and “becklessii” (Galton 1978). S. H. Beckles spelled his surname with a “k” (e.g., Beckles 1854), although variants on his surname have persisted as well (e.g., “Beccles”; Salisbury 2002).

Heterodontosaurid ornithischian characterized by the following six autapomorphies: (1) slender, nearly straight caniniform first maxillary tooth with unornamented anterior and posterior carinae; (2) edentulous anterior dentary margin (as long as two alveoli); (3) only 9 dentary teeth posterior to the caniniform tooth; (4) dentary crowns in the middle of the tooth row that are proportionately taller than opposing maxillary crowns (the apical 50% of middle dentary crowns are denticulate versus 25% of mid maxillary crowns); (5) anteroposteriorly elongate dentary symphysis (maximum length approximately 3 times maximum depth); (6) symphyseal flange ventral to primary dentary symphysis.

The original description of Echinodon becklesii is insightful and accurate in most regards (Owen 1861: 35–39, pl. 8). During the ensuing 150 years, the specimens have undergone further preparation and also have sustained some damage and loss (Norman and Barrett 2002: 173). Only a few new fragments and isolated teeth have come to light that may be tentatively referred to Echinodon becklesii (Norman and Barrett 2002), and so further information about this heterodontosaurid depends on the original materials. This review attempts to bring together published information and figures (Owen 1861; Galton 1978; Sereno 1991; Norman and Barrett 2002) and new observations on Echinodon in order to resolve conflicting statements and gain a better understanding of its morphology and status as a heterodontosaurid. The initial interpretation of Echinodon as a heterodontosaurid (Sereno 1991, 1997) was based on several of the observations documented below.

The ventral portions of the left and right premaxillae were originally preserved in mutual articulation in NHMUK 48209 (Owen 1861: pl. 8, fig. 1; Galton 1978: fig. 1A; Fig. 13C, D). Galton (1978: 140, fig. 1J) removed the left premaxilla (accidentally identifying it as the left “maxilla”) to expose the palate of the right premaxilla (also Norman and Barrett 2002: pl. 1, fig. 1). The premaxillae are figured here in their original position disarticulated from the associated left maxilla, both lacking anterodorsal and posterodorsal processes (Figs 11, 13C, D).

Despite two diagonal fractures and some crushing, several details of the left premaxilla have not been described previously. An anterior premaxillary foramen is present and split in two by a fracture with each half slightly separated (Fig. 11). The anterior premaxillary foramen is a good landmark for the anterior margin of the narial fossa, which is preserved as a broad depression extending ventrally from the foramen toward the a rugose rounded alveolar margin. The location of the anterior premaxillary foramen and ventral extension of the narial fossa is very similar to that in Heterodontosaurus but unlike the configuration in some other ornithischians such as Hypsilophodon (Galton 1974a). The rugose anterior portion of the alveolar margin is edentulous, such that the first premaxillary tooth is set in from the front margin of the premaxilla by a distance equal to two or three alveoli (Fig. 11). The edentulous margin was previously reconstructed (Galton 1978: fig. 2J) somewhat shorter in length (Fig. 19A). The longer edentulous margin more closely resembles the condition in Heterodontosaurus and some neornithischians than in basal ornithischians such as Lesothosaurus (Sereno 1991), in which the margin is only approximately one alveolus in length.

The posterior end of the alveolar margin is broken away along with part of the root of the third premaxillary tooth, as the specimen is now preserved (Fig. 11). When this portion of the premaxilla is preserved in other heterodontosaurids, it forms the anterior portion of an arched diastema with an inset medial wall. A similar recessed wall appears to be present on the left premaxilla in Owen’s figure (Fig. 13C, D), but this portion of the bone was broken away by the time Galton figured the specimen (1978: fig. 1A, A’; Fig. 11).

The dorsal portion of right and left premaxillae is broken away. Owen (1861: 36) remarked that part of the “boundary of the external nostril” (= external nares) is preserved, but this does not appear to be a natural margin (Fig. 11). The right premaxilla (exposed with removal of the left premaxilla) preserves a flat palatal surface as mentioned by Galton (1978) and Norman and Barrett (2002). A relatively flat secondary palate, however, is common among basal ornithischians and likely primitive (e.g., Lesothosaurus; Sereno 1991).

Finally, the position of the premaxilla relative to the maxilla is unknown, as the premaxillae are disarticulated and displaced anteroventral to the maxillae in the only partial cranium known for Echinodon (NHMUK 48209; Fig. 13C, D). The anteromedial process of the left maxilla (now only an impression) lies just above the palatal surface of the right premaxilla, not far from its original articulation. Tooth impressions figured by Owen suggest that a dentary was originally present immediately below the maxillary tooth row (Fig. 13C, D). None of the preserved dentary tooth rows have a crown configuration that matches these impressions. This missing dentary, if present, must have been lost during, or shortly after, collection of NHMUK 48209.

Thus it is likely that much of the anterior end of the snout was originally preserved in NHMUK 48209. In Heterodontosaurus, Abrictosaurus, and Tianyulong, the end of the snout is preserved intact, and the premaxillary alveolar margin is offset ventral to that of the maxilla. In an initial reconstruction of Echinodon, the alveolar margin of the premaxilla was drawn offset dorsally to that of the maxilla (Fig. 19A). The preserved location and form of the premaxilla in Echinodon (NHMUK 48209), to the contrary, suggests that there may have been some ventral offset of the premaxillary alveolar margin as in other heterodontosaurids (Fig. 19B).

The maxilla is known from three individuals, including the lectotypic left maxilla (NHMUK 48209, 48210; Figs 13, 14) and two partial right maxillae (NHMUK 48211, 48212; Fig. 12). These specimens allow a more complete description of this bone. Preserved sutural contacts of the maxilla include the premaxilla anteriorly, the lacrimal dorsally, the jugal posteriorly, and the palatine and ectopterygoid medially, (Figs 12–14).

A buccal emargination, or cheek embayment, is present along the entire length of the tooth row (Fig. 12). The emargination has been described as “shallow” (Galton 1978: 140), “gentle” (Sereno 199: 1761), and “very shallow” (Norman and Barrett 2002: 177). Galton (1978: 140, fig. 1G’) described and identified the upper border of the buccal emargination as a “slight horizontal ridge just above the tooth row”, and this is doubtless the eminence understood to define the upper boundary of the embayment by all three of the authors cited.

Close inspection of the single maxilla exposing this feature, however, casts doubt on this interpretation (NHMUK 48211; Fig. 12). A gently arched row of neurovascular foramina opens within the ornithischian buccal emargination on the maxilla. In Heterodontosaurus this row of foramina is present dorsally near the everted upper rim of the maxilla ventral to the antorbital fenestra. The row of large foramina in the maxilla in Echinodon represents the same neurovascular openings, which are located ventral to the margin of the antorbital fenestra. The maxilla thus has been flattened postmortem, reducing the depth of the buccal emargination. This transverse compression also has partially collapsed the internal canals associated with the row of foramina and reduced the eversion of the ventral margin of the antorbital fenestra. The opposing dentary emargination has deep proportions and includes a row of neurovascular foramina near the edge of the emargination (Figs 16, 17). The maxilla of Echinodon, in sum, appears to have had a buccal emargination (Fig. 19B) similar to that in Heterodontosaurus (Crompton and Charig 1974) and Lycorhinus (Gow 1990).

The anterior end of the same maxilla provides key evidence for the presence of an arched diastema to accommodate the apical end of a lower caniniform tooth (Fig. 13C, D). The laterally protruding, rounded, arched rim of the diastema is clearly preserved and is very similar to that in Heterodontosaurus. Ventral to that rim, the maxilla is inset and forms the posterior portion of a fossa within the diastema for reception of the tip of a dentary caniniform tooth. The margin of the maxilla within the fossa is not complete (contra Galton 1978: fig. 1G’). The short section of this margin that is preserved indicates that it arched above the maxillary tooth row as in Heterodontosaurus and Lycorhinus (NHMUK RU A100). The anterior end of the left maxilla (NHMUK 48210) may have originally included the region of the diastema (Owen 1861: pl. 8, fig. 1), but this region is broken away anterior to the first caniniform tooth (Norman and Barrett 2002: pl. 1, fig. 1).

Norman and Barrett (2002: 182) criticized my previous observation of an arched diastema in Echinodon (Sereno 1997), stating “it is impossible to determine whether the premaxilla-maxilla diastema was arched” and that a diastema of any kind is “absent from the available material of Echinodon”. Enough of this region is preserved in NHMUK 48211 to remove any doubt that an arched diastema is present in Echinodon (Fig. 12), despite the loss of bone fragments and crowns from two of the maxillae since Owen figured them.

Portions of both of these bones are preserved attached to two of the maxillae. In NHMUK 48210, the ventral ramus of the left lacrimal, including a portion of the orbital margin, is preserved posterodorsal to the left maxilla (Fig. 13A, B; Galton 1978: fig. 1G’). This portion of the lacrimal was identified previously as the “ascending process” of the maxilla (Norman and Barrett 2002: fig. 14A).

The anterior end of the jugal is preserved in articulation in two specimens. The first is preserved posterior to the antorbital fenestra (Fig. 12C, D; NHMUK 48211), and the second is located at the posterior end of the maxilla (Fig. 13A, B; NHMUK 48210). This portion of the jugal was identified previously as the “overhanging part of maxilla forming the base of lower temporal bar” (Galton 1978: fig. 1B’).

The right palatine is preserved in partial disarticulation from its lateral contact with the maxilla (Fig. 12C, D; NHMUK 48211). It appears to have rotated dorsally from its natural articulation exposing its ventral surface. The posterior margin has an embayment for a palatal fenestra.

In NHMUK 48210 the palatal bone posteromedial to the maxilla may be a left ectopterygoid. Immediately adjacent to this bone is an articular scar running across the maxilla-jugal suture. This is the lateral anchor for the ectopterygoid in many ornithischians (Fig. 13A, B). The posteromedial margin of the bone is preserved as an impression, the principal exposed surface is concave, and one of its margins is rounded and thickened as seen in cross-section. The position of the bone and its features best match that of an ornithischian ectopterygoid, a tentative identification. Galton (1978: 140, fig. 1B’) identified this bone as a right quadrate with a “twisted” shaft and “incomplete” distal condyles; Owen (1861: 37) and Norman and Barrett (2002: fig. 7B) identified the bone as the left pterygoid. The potential identifications of this bone could be tested by exposure of its opposing side, which is currently obscured by matrix, or via computed tomographic imaging.

Although the predentary is unknown in Echinodon, its presence is indicated by features on the anterior end of the dentary (Fig. 16A). A large anterior dentary foramen opens anteriorly and passes into an anterodorsally inclined, impressed vessel tract. In other ornithischians, this foramen provides passage for the principal vascular supply to the predentary (e.g., Lesothosaurus; Sereno 1991). The predentary in most ornithischians has lateral and ventral processes, and the vascular supply enters the bone near the junction between these processes. In Echinodon the vascular tract is deeply incised in a similar location, which is also the case in Fruitadens (Figs 9A, 17).

The anterior end of the dentary in Echinodon is intermediate between the condition typical of ornithischians (e.g., Lesothosaurus, Hypsilophodon; Sereno 1991; Naish and Martill 2001) and that in most heterodontosaurids. In the former, the end of the dentary is V-shaped, with articular surfaces for the lateral and ventral processes facing dorsally (or dorsomedially) and ventrolaterally, respectively. In most heterodontosaurids, in contrast, the end of the dentary is slightly expanded dorsoventrally and has a single well-defined, arcuate predentary articular surface that faces anterolaterally. In Echinodon the anterior end of the dentary is more rounded than in Lesothosaurus (Sereno 1991) or Hypsilophodon (Naish and Martill 2001) but lacks a well-defined, arcuate articular surface for the predentary.

The dentary does not have a well-defined surface dorsal to the foramen for articulating with the predentary, and so a projecting lateral predentary process probably was not present. The ventral aspect of the anterior end of the dentary has as a subtle smooth articular facet (Figs 15–17). It is not as well marked as in basal ornithischians, where the ventral edge of the dentary is strongly beveled for the median ventral process (e.g., Lesothosaurus; Sereno 1991). Nor is it comparable to that in some heterodontosaurids such as Heterodontosaurus, in which the articular surface for the predentary is well-defined and trough-shaped. Several small neurovascular foramina are present between dorsal and ventral articular areas for the predentary, an area that clearly would not have been covered by the predentary (Fig. 17). In sum, the predentary in Echinodon appears to have a fairly loose articular relation with the dentaries.

Previously Galton (1978: Figs 1C’, 2J) inferred the presence of a predentary in Echinodon but misinterpreted the anterior dentary foramen and its associated impressed vessel channel as an articular slot for a prong-shaped lateral predentary process (Fig. 19A). Although a lateral predentary process is present in Hypsilophodon (Galton 1974a: fig. 10) and other ornithischians, the dentary in Echinodon does not show a facet for such a process. Norman and Barrett (2002: 177) stated that they could not locate any articular surface for the ventral predentary process, but a smooth, flattened articular surface runs along the anteroventral edge at the anterior end of both right and left dentaries (Fig. 17).

The dentary is best known from paired right and left sides (Figs 15–17; NHMUK 48215a, b) and partial left and right dentaries (NHMUK 48213, 48214). Originally exposed on slabs of rock, all four dentaries were prepared free of matrix prior to their re-description by Galton (1978). Only a fragmentary dentary has survived from a fourth specimen, which originally included a portion of the left lower jaw (Owen 1861: pl. 8, Figs 6–8; NHMUK 48214).

The dentary in Echinodon has particularly stout proportions, even when compared to other heterodontosaurids. Its depth at mid-length is approximately 30% of its length (from anterior extremity to the anterior margin of the external mandibular fenestra) and approximately 40% of the length of the postcaniniform tooth row (Figs 15A, C, 16A, C). In lateral view, the anterior end of the dentary has a subtriangular shape with a strongly beveled anteroventral margin for contact with the predentary (Fig. 17). The middle section of the denary ramus is nearly parallel-sided, expanding only slightly in depth posteriorly. An arched row of relatively large neurovascular foramina opens along the ventral margin of a deep buccal emargination. As in other heterodontosaurids, this emargination tapers in depth near the alveolus for the caniniform tooth in advance of the anterior end of the dentary (Figs 15A, 16A, 17).

The coronoid process is distinctly expanded at mid-length, resulting in a diamond-shaped, rather than tapered, process (NHMUK 48215a, 48213; Owen 1861: pl. 8, fig. 8; Galton 1978: fig. 1D; Barrett 2002: pl. 2, fig. 3). The central axis of the coronoid process angles posterodorsally at about 45° to the long axis of the dentary, as best preserved in NHMUK 48215b and NHMUK 48213 (Fig. 19B). Postmortem crushing has increased the inclination of the coronoid process in one dentary (NHMUK 48215a). In basal ornithischians such as Lesothosaurus (Sereno 1991), in contrast, the coronoid process is narrow and tapered (finger-shaped) in lateral view, and the coronoid process is less strongly upturned, angling posterodorsally at about 30°.

The symphysis at the anterior end of the dentary is V-shaped (Figs 15B, 16B). The main articular surface is an oval, raised and textured platform with its long axis horizontal. This articular surface is located almost entirely anterior to the dentary caniniform tooth rather than directly ventral to the caniniform tooth as in Heterodontosaurus. A subtriangular fossa lies between the main symphyseal articulation and a secondary flat symphyseal surface, which is located along the ventral margin. This flat surface may have served as a median buttress or stop, as it is not textured for ligament attachment like the main symphyseal surface.

The anterior ends of the dentary are not inturned to form a spout shape as in most ornithischians. The symphysis in Echinodon, nevertheless, does not lie on the medial plane of the dentary ramus, but rather is elevated from that plane toward the midline. As a result, there is a narrow dorsoventrally concave surface dorsal to the symphysis and medial to the mesialmost teeth (Figs 15C, 16C). A narrow trough-shaped surface thus is present dorsal to the symphysis as in Heterodontosaurus. The symphysis also projects medially in Lycorhinus (NHMUK RU A100) and Heterodontosaurus (SAM-PK-K1332). The main derived features in Echinodon include the elongate symphyseal area, the anterior position of the symphysis relative to the dentary tooth row, and the accessory ventral symphyseal surface near the ventral margin of the dentary.

Galton (1978: 140, fig. 1C’) described and figured the buccal emargination of the dentary as a “small cheek” excluding the row of major neurovascular foramina. Norman and Barrett (2002: 177) also described the buccal emargination of both the dentaries and maxillae as “very shallow.” Transverse compression of several of the dentaries, however, has reduced the depth of the emargination, which must incorporate the principal row of neurovascular foramina. An area of impressed ornamentation is present just below the row of neurovascular foramina, suggesting the presence of tightly adhering integument below the buccal emargination (Figs 15A, 16A).

Galton (1978: 140) described the coronoid process as “prominent”, whereas Norman and Barrett (2002: 177) described it as “low”. One very appropriate comparison is the basal ornithischian Lesothosaurus (Sereno 1991: fig. 13F). Echinodon more closely resembles the more strongly upturned, transversely expanded process in Heterodontosaurus (see below) than the tapered process of shallow inclination in Lesothosaurus (Fig. 19B).

There are three premaxillary teeth in middle and posterior portions of the premaxilla, preceded by an edentulous margin several alveoli in length (Figs 11, 19B). The first and third crowns are preserved with some breakage and loss since Owen first figured them (Fig. 13C, D). Only small fragments of root and crown of the second tooth remain, the base of which was originally preserved. The crowns are slightly swollen, the mesial side of the crown base more bulbous than the distal side, and have smooth surfaces without denticles or serrations. The crowns are gently recurved with apices set slightly distal to the center of the crown base. As in other heterodontosaurids, the third premaxillary tooth is slightly larger than the first. In other ornithischians, premaxillary crowns tend to be subequal in size and have denticulate mesial and distal carinae (e.g., Lesothosaurus, Hypsilophodon; Naish and Martill 2001; Sereno 1991). Little else can be said about the premaxillary teeth given their state of preservation.

Galton (1978) reconstructed the premaxillary tooth row with crowns of equal size (Fig. 19A), and Butler et al. (2012; 6) reported that crown size does not increase posteriorly in the premaxillary series. The preserved portions of pm1 and 3 (Figs 11, 19B), however, clearly show an increase in size as occurs in most other heterodontosaurids.

There are nine maxillary teeth (Figs 12–14), the first a nearly straight, transversely compressed caniniform tooth lacking any ornamentation on mesial or distal carinae (Fig. 11). Preserved only in the lectotypic left maxilla, the caniniform tooth has a relatively straight and slender crown that extends only a short distance below more distal maxillary crowns (Figs 12B, 13C, D). The maxillary caniniform tooth is preceded by an arched diastema for reception of the dentary caniniform tooth. The maxillary caniniform tooth is smaller than the dentary caniniform tooth (Fig. 19B), judging from the upper diastema and lack of an opposing lower diastema between the caniniform and third dentary tooth (Figs 15C, 16C). In Echinodon, thus, the upper caniniform tooth is positioned distal to a lower caniniform. In other heterodontosaurids, in contrast, the large third premaxillary crown is positioned mesial to a lower caniniform tooth (e.g., Lycorhinus, Heterodontosaurus).

Owen briefly described a second more mesial caniniform tooth in the maxilla based on a fragment and possible impression (Fig. 13C, D). Galton (1978) regarded the fragment as a basal piece of the relatively complete caniniform tooth. Norman and Barrett (2002: fig. 7A), however, added to the diagnosis the possibility of a second caniniform tooth based on Owen’s figures. Neither the tooth fragment nor the potential impression has survived. Available specimens, nonetheless, clearly indicate that only one caniniform is present at the anterior end of the maxilla. There is only a single empty alveolus for a caniniform tooth distal to the arched diastema in a right maxilla (Fig. 12), as correctly described by Galton (1978: 143).

There are seven or eight postcaniniform teeth. Owen described the two lectotypic specimens that represent part and counterpart of a single specimen that was split from a single block of matrix. He indicated how the tooth rows on the opposing pieces should be aligned (Fig. 13). In the anterior piece (NHMUK 48209), there are two postcaniniform teeth; in the posterior piece (NHMUK 48210), there are five teeth plus one empty alveolus documenting the distal end of the tooth row. Owen correctly identified one of the teeth in the anterior piece as a replacement crown erupting beneath the anteriormost tooth in the posterior piece. Following this alignment, the total number of teeth in this complete maxillary series is eight (one caniniform tooth followed by seven postcaniniform teeth) (Fig. 19B). Based on the same specimens, Galton (1978) suggested there are as many as 10 or 11 maxillary teeth, although his reasons for this higher number were not given. A second maxilla preserves the anteriormost alveolus for a caniniform tooth followed by six postcaniniform crowns (Fig. 12). A posterior piece with one and one-half crowns was originally present (Owen 1861: pl. 8, fig. 3), suggesting that this individual had one additional postcaniniform tooth. On the basis of the available collection, thus, Echinodon has eight or nine teeth in the maxilla, including a mesial caniniform tooth followed by seven or eight teeth with denticulate crowns (Fig. 19B).

The first postcaniniform tooth (second maxillary tooth) has taller crown proportions than succeeding maxillary teeth; the crown is narrower and the denticulate portion of the crown is approximately 45% total crown height (Fig. 12C, D). Although the distalmost crown is not preserved, all of the remaining maxillary crowns are very similar in size and shape. In other heterodontosaurids, crown size is more variable and often substantially larger toward the middle of the maxillary series. Except for marginal ridges along mesial and distal crown edges, there are no ridges on either labial or lingual crown surfaces. There is a rounded median eminence, which is low on both sides but possibly slightly stronger on the lingual side (Figs 12, 14). There are approximately 8 to 10 denticles to each side of the apical denticle in most crowns. This denticle count is best observed in a nonfunctioning replacement crown in the second alveolus (NHMUK 48209). Galton (1978) identified the median eminence as a “ridge” and suggested that there are fewer (only 5 or 6) denticles to each side of the crown apex.

The enamel is symmetrical on each side of the maxillary crowns as is visible in the cross-section of several crown tips. Wear facets are present on raised areas of the lingual face of the crowns of all fully erupted maxillary crowns that are well preserved and exposed in lingual view (Fig. 14B, C). These facets are more fully described below (see Discussion, Wear).

The dentary tooth row has 11 teeth, based on evidence from three dentaries with nearly complete alveolar margins (NHMUK 48214, 48215a, 48215b). The first dentary tooth must have had a very small peglike crown as in Lycorhinus, as the size of the alveolus is much smaller than any other in the dentary (Figs 15C, 16C). The root and base of the crown of this small tooth is preserved in one dentary (NHMUK 48214), and the small alveolus is preserved in the other two dentaries. The second dentary tooth, in contrast, is larger than all others, and judging from the elongate alveolus housed a caniniform tooth (Figs 15C, 16C). A substantial edentulous margin precedes both of these alveoli, a feature that distinguishes Echinodon from other heterodontosaurids and other basal ornithischians (Fig. 19B).

The small first dentary alveolus has never been described. Galton (1978: Figs 1C’, F’, 2J) figured the large second alveolus and indicated the presence of a more mesial alveolus in one of the specimens (NHMUK 48215b). He did not comment on the size of these teeth and reconstructed the dentary tooth row without caniniform or peglike anterior teeth (Fig. 19A). The initial suggestion that Echinodon has a caniniform dentary tooth (Sereno 1997) was criticized by Norman and Benton (2002: 182) who stated “no known specimen displays evidence of either in situ caniniforms or natural moulds thereof”. Technically speaking, of course, the caniniform tooth itself is not preserved in any available specimens. Nonetheless, a caniniform tooth was surely present in Echinodon, given the size and shape of the alveolus in three available dentaries and the evidence of an opposing arched diastema in the maxilla. Like the caniniform teeth in other heterodontosaurids (e.g., Fruitadens; Butler et al. 2010: fig. 3c, d), the alveolus and root are larger than adjacent crowns and angle ventrodistally rather than vertically. The enlarged alveolus in Echinodon, thus, should exhibit these features in a computed tomographic scan (Figs 15, 16).

There are eight and nine postcaniniform dentary teeth, respectively, in the left and right dentaries of NHMUK 48215. The left dentary, however, is incomplete posteriorly (Figs 15, 16) and appears to be missing the smaller distalmost tooth preserved on the right side. In the right series, dentary teeth 10 and 11 (postcaniniform teeth 8, 9) have progressively smaller crowns unlike the last two subequal alveoli on the left side. A total of 11 dentary teeth is also consistent with the teeth and alveoli in two other relatively complete dentaries (NHMUK 48213, 48214; Fig. 18). Galton’s (1978) estimate of 10 dentary teeth, therefore, is one too few (Fig. 19A), as he did not count the peg-shaped first dentary tooth.

The crowns of postcaniniform dentary teeth are well separated from their roots and have taller proportions than opposing maxillary crowns (Fig. 19B). All but the posterior three crowns are diamond-shaped, rather than subtriangular. The dorsal 50% of each of these crowns is denticulate, as opposed to approximately 25% in opposing maxillary crowns. Norman and Barrett (2002: 177) stated that the denticles of both dentary and maxillary teeth are “confined to the apical one-third of the tooth crown”, but this is not true for crowns in the middle of the dentary series. As in postcaniniform maxillary crowns, postcaniniform dentary crowns typically have about 8 to 10 denticles to each side of the apical denticle. Likewise, only a median eminence is present on labial and lingual crown faces, and the roots are subconical and tapered where they are exposed (NHMUK 48215a).

The enamel is symmetrically distributed on each side of the dentary crowns as in maxillary crowns. Enamel also appears to cover a flattened subtriangular area on the mesial and distal sides of the crown base between the crown faces and root. This surface, which is present only on the largest crowns, was described previously as exposed dentine and highlighted as unique to Echinodon (Owen 1861; Galton 1978). I am unable to verify the absence of enamel or the fact that this surface is a unique, diagnostic feature of Echinodon.

The reconstruction of the snout and dentition of Echinodon (Fig. 19B) is based on the original specimens and figures in Owen (1861) and differs from a previous reconstruction (Figure 19A) in several regards. The premaxilla is shown with its alveolar margin offset ventral to the maxillary tooth row and with a significant edentulous border anterior to the first premaxillary tooth. The maxilla has a shorter series of postcaniniform teeth, above which is a significantly deeper buccal emargination. Dentary tooth 1 is rudimentary and followed by a large caniniform tooth, the tip of which inserts into an inset, arched diastema between the premaxilla and maxilla. The dentary also has a deeper buccal emargination, which dissipates anteriorly as it passes near the base of the caniniform tooth. Finally, the anterior end of the dentary, which articulates with a reconstructed predentary lacking processes, has an anterior dentary foramen with an incised vascular canal, above which is a significant edentulous margin. The new reconstruction removes any reasonable doubt about the heterodontosaurid status of Echinodon.

LACM 115747, adult with partial maxillae and dentaries, cervical, dorsal, sacral and caudal vertebrae, proximal right femur, proximal and distal ends of the tibiae, and partial right metatarsal 1 (Butler et al. 2010: fig. 2b, e, i; Butler et al. 2012: Figs 1, 2–4, 8A, 9–12, 13A–E, 14G, H).

LACM 115727, adult partial postcranial skeleton with partial cervical, dorsal and caudal vertebrae, partial right and left femora, and an articulated distal left tibia and coossified astragalocalcaneum; LACM 120478, subadult with left humerus, distal left femur, and an articulated left tibia, fibula and coossified astragalocalcaneum; LACM 120602, distal caudal vertebra, left astragalocalcaneum, partial metatarsus and pes; LACM 128258, subadult with right premaxilla, partial left maxilla, partial left and right dentaries, and one dorsal and one distal caudal vertebra; LACM 128303, anterior left dentary (Butler et al. 2010, 2012).

Fruita Paleontological Area, approximately 10 kms southwest of Fruita, Mesa County, west-central Colorado, USA; approximately N39°10', W108°48'.

Just above the “clay change” near the base of the Brushy Basin Member and about 100 m from the base of the Morrison Formation (Kirkland et al. 2005: fig. 6; Kirkland 2006); Upper Jurassic (early Tithonian), ca. 153 Ma (Kirkland 2006; Gradstein and Ogg 2009). The boundary between the Salt Wash and Brushy Basin Members has undergone revision in the Fruita Paleontological Area (Kirkland 2006). Galton (2002) reported that the fossils were found in the Salt Wash Member but later correctly cited the overlying Brushy Basin Member as the unit of origin (Galton 2007).

Heterodontosaurid with (1) a discordantly small dentary tooth immediately distal to the caniniform dentary tooth and (2) a prominent anteromedial flange on the distal end of the tibia.

Butler et al. (2010: 376) listed nine features in the original diagnosis for Fruitadens haagarorum, indicating that some were primitive and others “autapomorphic” at different phylogenetic levels. Butler et al. (2012: 3) added one character, “small foramen on the anteroventral aspect of the medial dentary”, for a total of ten features.

This suite of features constitutes a differential diagnosis—a unique combination of the features that describes a monospecific genus rather than a set of autapomorphies hypothesized to have arisen in the immediate ancestry of the taxon (Sereno 1990). In the revised diagnosis above, a higher bar is applied that restricts listed features to those that are plausibly unique to Fruitadens or derived for Fruitadens within Heterodontosauridae. One of the features stands out as plausibly unique to Fruitadens and present in the holotype—the prominent anteromedial flange on the distal end of the tibia (Butler et al. 2010: fig. 2i). Comparison to other heterodontosaurids, however, is limited to Heterodontosaurus, which does not exhibit the condition. The small medial dentary foramen, which was added by Butler et al. (2012) as an autapomorphy, is not present in one of the two dentaries that preserve the region (Butler et al. 2012: fig. 5D).

One dental feature listed in the revised diagnosis may be an autapomorphy but is homoplasious among heterodontosaurids. The dentary tooth immediately distal to the caniniform tooth in Fruitadens is unusually small, its crown apparently somewhat smaller than the rudimentary first dentary tooth (Fig. 9A). Heterodontosaurus and a new taxon from southern Africa, Pegomastax gen. n. sp. n., are the only other heterodontosaurids with a discordantly small tooth immediately distal to the caniniform tooth.

The other features listed in the diagnosis by Butler et al. (2010, 2012) are demonstrably primitive within Heterodontosauridae or may be artifacts of preservation. The lack of a maxillary caniniform tooth, for example, is the common condition; only Echinodon has a caniniform tooth in the maxillary series among heterodontosaurids. Denticles that extend along more than one-half of the crown occur in Tianyulong and in ornithischian outgroups such as Lesothosaurus. Other features have broader distributions within Heterodontosauridae, such as the small peglike first dentary tooth, which is present in Echinodon, Lycorhinus, and Abrictosaurus. The dentary caniniform tooth was said to be shorter than in some heterodontosaurids and more nearly equal in depth to noncaniniform dentary crowns. The relative depth of individual crowns, however, depends to a great extent on the stage of replacement, which can be difficult to estimate in the case of caniniform teeth. The dentary caniniform tooth in the Kayenta heterodontosaurid, for example, has a similar relative depth to that in Fruitadens, although it is clearly undergoing eruption and would be considerably larger when fully functional (Fig. 9A, B). Most of the caniniform tooth in question (LACM 128258), furthermore, seems to have broken away by the time the specimen was described by Butler et al. (2010: fig. 2d; compare Galton 2007: fig. 2.7B). The pair of foramina on the anterior aspect of the astragalus can be compared only in Heterodontosaurus among heterodontosaurids, and there is some evidence the condition is present (SAM-PK-K1332).

The separation of the ascending process of the astragalus as a separate ossification was listed among the autapomorphies. The suture separating the distal tip of the astragalar ascending process, however, seems to continue laterally as a fracture line across the distal shaft of the fibula. The ascending process had been viewed as a separate ossification in the theropod Dilophosaurus (Welles 1984); review of this specimen, however, suggests that it also appears to be a postmortem fracture passing through vascular foramina. Persistence of such a suture in Fruitadens, in addition, seems unlikely in a taxon distinguished by coossification in this particular region of the limb skeleton (e.g., the tibiotarsus).

The six available specimens of Fruitadens haagarorum were collected between 1975 and 1980 in the Fruita Paleontological Area from four separate localities above a distinctive horizon (“clay change”) near the base of the Brushy Basin Member of the Morrison Formation (Kirkland et al. 2005; Kirkland 2006). The localities were not discovered in a single horizon but rather within a zone perhaps 10-15 m in thickness above the “clay change” (G. Callison, pers. comm.). Four specimens were described in the initial description (Butler et al. 2010); two additional fragmentary specimens were included more recently (Butler et al. 2012; LACM 120602, 128303).

Fruitadens is a small-bodied heterodontosaurid, the adult specimens of which appear to be slightly larger than two other small-bodied heterodontosaurids, Echinodon and Tianyulong (Table 3). The sole specimen known of the Kayenta heterodontosaurid is the smallest heterodontosaurid on record (skull length estimate of 53 mm), but histologic evidence suggests that it is a subadult roughly the same size as half-grown individuals of Heterodontosaurus (e.g., AMNH 24000; Fig. 2C).

The holotype of Fruitadens and three referred specimens were described as associated individuals, the supportive evidence limited to the lack of duplicate bones and consistent state of preservation within each specimen (Butler et al. 2010: suppl. info.; Butler et al. 2012). A first-hand account of the discovery of the holotype (LACM 115747) and a referred subadult specimen (LACM 120478) confirms their association as individuals. The first pieces of the holotype were surface-collected on a slope, which led to an in situ portion of the specimen that was recovered in a small field jacket (G. Callison, pers. comm.). The subadult specimen LACM 120478, which preserves the most complete long bone lengths, was also found in a confined space by quarrying at a nearby locality 5572 (“Main Callison Quarry”). A referred adult specimen was found at locality 5576 (“George’s Coelurosaur site”), and there is no specific site information available for a referred subadult with the most complete set of jaws (LACM 128258), except that it comes from the same general area (Fruita Paleontological Area).

The dentary in Fruitadens has a vascularized buccal emargination, presumably as an aid to the processing plant materials within the oral cavity (Fig. 9A). The anterior end of the dentary in Fruitadens is subrectangular, whereas in as Echinodon the ventral side of the anterior end of the dentary is strongly beveled. Both Fruitadens and Echinodon have a well-demarcated vessel tract passing from the anterior dentary foramen toward the predentary.

A jaw fragment housing three teeth was identified as a right premaxilla (Butler et al. 2010: fig. 2a). Only the middle tooth preserves the entire crown; only the base of the distal crown is preserved, and the first crown in the series appears to have been lost. Butler et al. (2010: 376) noted that all of these teeth have waisted, subtriangular crowns; there is no development of a caniniform crown. Butler et al. (2012), further, suggested that a small portion of the left premaxilla might be attached across the median palatal suture. A portion of an ascending ramus has been shown as preserved on the premaxilla (Fig. 9A), but no part of this ramus appears to be preserved on the actual specimen (Butler et al. 2012: fig. 7A). In other heterodontosaurids, the narial fossa extends close to the ventral margin of the premaxilla, whereas in Fruitadens the narial fossa is not evident in lateral view. The identification of this jaw fragment as a portion of the right premaxilla may be correct, but it exhibits several unusual features as compared to more completely known premaxillae in Echinodon, Tianyulong, Lycorhinus, and Heterodontosaurus.

The maxilla in Fruitadens has a deep margin between the antorbital fenestra and the maxillary tooth row, although this depth can appear greater with postmortem compression (Fig. 9A). As preserved it most closely resembles the condition in Echinodon and Lycorhinus. The buccal emargination is approximately one-half as deep in Tianyulong, Abrictosaurus, and Heterodontosaurus. Butler et al. (2010: 376, fig. 2c) correctly identified a partial left maxilla in a subadult (LACM 128258), whereas Galton (2007: Figs 2.6E, 2.7A) identified the same specimen as a right maxilla.

The location, size and shape of the anterior dentary teeth differ between Fruitadens and Echinodon. In Fruitadens (Fig. 9A) the first dentary tooth is closer to the anterior end of the dentary than in Echinodon, in which a short edentulous margin precedes the first dentary tooth (Galton 2007: fig. 2.7B; Butler et al. 2010: fig. 2d). Both Fruitadens (LACM 115747) and Echinodon have a small peg-shaped first dentary tooth. Although described as unique to Fruitadens (Butler et al. 2010: 376, 378), a rudimentary first dentary tooth (or tiny alveolus) is also known in Echinodon, Lycorhinus and Abrictosaurus. In some heterodontosaurids, including the Kayenta heterodontosaurid, Pegomastax gen n. sp. n., and Heterodontosaurus, there are no teeth mesial to the caniniform tooth.

The third dentary tooth in Fruitadens, or the first “cheek” tooth, is the smallest tooth in the dentary series (Fig. 9A). Although described as present in “other heterodontosaurids” (Butler et al. 2010: 378), such a diminutive tooth is known only in Pegomastax gen n. sp. n. and possibly in Heterodontosaurus. In other heterodontosaurids such as Echinodon, Lycorhinus, and Abrictosaurus, the first cheek tooth is subequal to successive crowns at the anterior end of the dentary tooth series.

The premaxillary teeth described by Butler et al. (2010) would be unusual in form for heterodontosaurids, but identification of the jaw fragment as a right premaxilla may be problematic as discussed above. Galton (2007: 28) reported the presence of five premaxillary teeth in Fruitadens, but there appears to be no evidence in support of this statement.

The largest maxillary crowns in the distal portion of the series have a bulbous cingulum with well-defined basal and apical edges (Butler et al. 2010: fig. 2b: Galton 2007: fig. 2.7C, F, G). The apical edge of the cingulum is maintained in the center of the crown base, where the median eminence joins the base of the crown (Fig. 9A). This well-defined cingulum is present only in the largest crowns and resembles the condition in the cheek teeth of thyreophoran ornithischians. A similar condition is present in Tianyulong. The apical edge of the cingulum merges with the crowns face in Echinodon and most other heterodontosaurids.

Computed-tomographic scans show active tooth replacement in Fruitadens (Butler et al. 2010: Figs 2f, 3d, 3a-c; Butler et al. 2012: Figs 3, 4). Wear facets from tooth-to-tooth occlusion, however, have not been identified. The presence or absence of wear facets is difficult to determine with so few available crowns. Some of these specimens, in addition, are from younger individuals, which may not show wear typical of adults. The low-angle wear facets in Echinodon are less obvious and sometimes absent on newly erupted crowns.

STMN 26-3, partial skeleton laying on its left side preserving most of the skull in left lateral view, the ventral portion of a skull and articulated skeleton lacking the mid and distal caudal vertebrae, right coracoid, left carpus, portions of the left manus, and portions of the right hindlimb (Table 3; Zheng et al. 2009).

IVPP V17090 (Fig. 20), partial skeleton laying on its left side preserving a nearly complete skull, an articulated portion of the column including the posterior dorsal vertebrae, sacral vertebrae, and proximal one-half of the tail with numerous, aligned epaxial and hypaxial ossified tendons, proximal portions of both scapulae, both coracoids, most of both forelimbs including an articulated left carpus and manus in ventral view, and both hindlimbs with right pes mostly in ventral view and left pes mostly in dorsal view (Tables 4, 5).

Jianchang County, Liaoning Province, PRC; collected privately but localities are probably in the vicinity N41°20', E119°40'.

Probably from the Lujiatun Beds of the Yixian Formation, Jehol Group; Lower Cretaceous (Barremian-Aptian), ca. 125 Ma (Xu and Norell 2006; Gradstein and Ogg 2009).

Heterodontosaurid with (1) only two premaxillary teeth, (2) rectangular dentary ramus with parallel dorsal and ventral margins, (3) extremely reduced forelimb that is less than 30% the length of the hindlimb, (4) manual digit III and metacarpal 3 shorter than manual digit II and metacarpal 2, respectively, (5) tail increased in length, (6) subtriangular chevrons in mid caudal vertebrae, and (7) numerous parallel ossified epaxial and hypaxial ossified tendons in the mid and distal regions of the tail.

The holotype is a mature skeleton as evidenced by fusion of sacral centra, fusion or tight articulation of the neural arch and centrum of all other preserved vertebrae, and fusion between the tibia and proximal tarsals and between the bases of the metatarsals. The stratigraphic origin and geological age of Tianyulong remain uncertain, because all currently known specimens were collected privately (X. Xu, pers. comm.). The initial description of Tianyulong only reported a general area (“Jianchang County, Liaoning Province”) and horizon (“Jehol Group, Early Cretaceous”) (Zheng et al. 2009: 333). Upper Jurassic formations underlying the Jehol Group also have yielded specimens with integument preservation and a similar taphonomic attributes (Hu et al. 2009), and so the assignment of Tianyulong to the Yixian Formation with a Lower Cretaceous (Barremian-Aptian) age awaits confirmation.

The following brief description is based on two skeletons, the holotype (STMN 26-3; Fig. 9C) and a referred specimen (IVPP V17090; 20–30). Four additional skeletons of varying completeness are catalogued in the collections of the Shandong Tianyu Museum of Nature. Further preparation and study of all of these specimens of Tianyulong is needed before attempting a reliable skull reconstruction. The present description focuses on the most important features, culminating in a skeletal reconstruction (Fig. 30) and a set of comparative measurements (Tables 3–5).

The cranium is well represented in the holotype and referred specimens, although several portions remain poorly understood (Figs 20–23). Little is known of the posterodorsal portion of the cranium, which is broken away in the holotype (STMN 26-3; Zheng et al. 2009; Fig. 9C) and disarticulated and only partially prepared in the referred specimen (Fig. 21). Likewise, virtually nothing is known about the palate and braincase. The shape of the skull is similar to that in Heterodontosaurus; both are subtriangular in lateral view with a gently concave roof over the antorbital region. Both crania also are preserved in a similar manner, with jaws in occlusion and the premaxilla and predentary in close approximation (Fig. 21).

In both skulls the alveolar margin of the premaxilla is tilted slightly anteroventrally and is positioned ventral to the maxillary tooth row (Figs 22, 23). A horizontal line along the maxillary crowns at mid height passes above the premaxillary alveolar margin and closer to the ventral margin of the external naris. Approximately 60% of the length of the premaxillary alveolar margin is edentulous, the posterior 40% of its length accommodating two premaxillary teeth. In both skulls the posterior end of the alveolar margin curves posterodorsally, exposing a portion of the root of the caniniform second premaxillary tooth. Dorsal to this root, the premaxilla forms the anterior portion of an inset, arched diastema for the large dentary caniniform tooth. The narial fossa extends ventrally near the alveolar margin, and the external naris is dorsoventrally elongate as in Heterodontosaurus. The posterolateral process of the premaxilla also is very similar in shape and articular contacts to that in Heterodontosaurus. This process expands in width above the caniniform tooth and then tapers to a narrow tip, which in the holotype appears to establish point contact with the anterior tip of the lacrimal (STMN 26-3; Zheng et al. 2009).

The subtriangular maxilla forms the posterior portion of the inset, arched diastema (Fig. 9C). Most of the lateral aspect of the maxilla is occupied by the subtriangular antorbital fossa, which is bordered ventrally by a sharp, slightly arched rim. The buccal emargination ventral to this rim is narrow compared to that in Echinodon and Lycorhinus. The jugal appears to lack the horn and flange that characterizes Heterodontosaurus and Manidens.

The predentary is a small, wedge-shaped bone that lacks discrete processes (Figs 22, 23). The predentary had been shown with a long ventral process (Zheng et al. 2009: fig. 1d; Fig. 9C), but a portion of this process is actually the ventral margin of the dentary. All but the posterodorsal corner of the predentary is positioned anterior to the premaxillary caniniform tooth.

The dentary ramus is straight and parallel-sided for most of its length (Fig. 21) in contrast to most heterodontosaurids, which exhibit a posterior deepening of the ramus. At its anterior end, a ventral protuberance is present in both the holotypic and referred skulls (Figs 22, 23). Anterior to the protuberance, the dentary end is strongly beveled as in Echinodon (Fig. 17). The ventral rim of the well developed buccal emargination is marked by diverging impressed vessel tracts (Figs 22, 23).

The sutures between the postdentary bones are poorly preserved in the two available specimens. The coronoid process rises well above the level of the dentary crowns as in Heterodontosaurus, but the jaw articulation is not dropped relative to the occlusal plane (Figs 9C, 21). As best seen in the holotypic skull, a line drawn through the zone of occlusion between the cheek tooth rows passes just ventral to the jaw articulation (Fig. 9C). The coronoid process of the dentary was shown as a deep ramus rather than a more slender process (Zheng et al. 2009), an interpretation that may have incorporated portions of the adjacent surangular (Fig. 9C). An external mandibular fenestra appears to have been present (Zheng et al. 2009). There is no evidence for the presence of an external mandibular fossa around the fenestra or for an incised vessel tract to a large anterior surangular foramen, both of which characterize Lycorhinus, Manidens and Heterodontosaurus.

There are two premaxillary teeth, the first a small tooth known only from its broken base and the second a large caniniform tooth (Figs 22, 23). The caniniform premaxillary tooth, the base of which is better preserved in the holotype skull, has a gentle posterior recurvature. Mesial and distal carinae are present, at least the latter with serrations. Only the larger distal premaxillary tooth was shown in the initial drawing of the holotypic skull (Fig. 9C). As in all heterodontosaurids and neornithischians, there is a substantial edentulous border preceding the first premaxillary tooth.

The total number of dentary teeth in Tianyulong cannot be established with certainty based on the holotype, although referred skulls suggest there are 10 dentary teeth. An enlarged caniniform is the first tooth in the dentary series; no trace of a leading rudimentary crown is present. The dentary caniniform tooth is followed by a postcaniniform crown without a significant intervening gap, similar to the condition in Echinodon and Fruitadens. The first postcaniniform tooth in the holotypic skull can be seen in the photographs slightly dislodged toward the caniniform tooth (Zheng et al. 2009: fig. 1e). It was omitted in a drawing of the holotypic skull (Fig. 9C), leaving the impression that a postcaniniform diastema may be present in Tianyulong. The first dentary tooth has a subconical crown that is slightly smaller than that of more distal dentary teeth. This differs from the substantially smaller tooth immediately distal to the caniniform tooth in Fruitadens (Fig. 9A). A referred specimen in the Shandong Tianyu Museum of Nature collections confirms the morphology, size and position of the first postcaniniform dentary tooth.

Successive dentary crowns 3-10 become slightly larger in the holotypic and referred skulls. In these postcaniniform dentary teeth, the bulbous cingulum has well-defined margins, a cingular ectoloph raised above the remainder of the crown surface including the primary ridge. The cingulum curves apically, terminating mesially and distally in prominent apically projecting denticles. Toward the posterior end of the dentary series, the prominence of the apical, mesial and distal basal denticles gives the crown a tricuspid appearance. The penultimate tooth, dentary tooth 9, is the largest in the series. The most distal tooth, dentary tooth 10, has a considerably smaller crown, as seen in several referred skulls.

The total number of maxillary teeth is currently unknown given the available evidence in the holotypic and referred skulls. In the right maxillary series of the holotypic skull, the crowns of the smallest teeth just behind the caniniform dentary tooth are broken at their bases (Fig. 9C). This pair is followed by three crowns, a gap for a missing sixth maxillary tooth, and a final set of three teeth, all of which have crowns that are progressively larger in size (Zheng et al. 2009: fig. 1d). Probably at least two additional teeth with progressively smaller crowns were present at the distal end of the maxillary series. Taking these two into account, a total of 11 maxillary teeth likely were present in the holotypic skull.

Maxillary and dentary crowns are subtriangular, the dentary crowns somewhat deeper than opposing maxillary crowns as in Echinodon. All have similar features in labial view (Fig. 24). The bulbous cingulum is well-defined from the remainder of the crown surface as in the largest crowns in Fruitadens, and the root is tapered rather than swollen. The cingulum curves apically as a prominent cingular ectoloph along the mesial and distal crowns edges, terminating in prominent basal denticles mesially and distally. Arching of the alveolar margins is not pronounced as in Abrictosaurus and Heterodontosaurus. The dentary alveolar margin is straight; the opposing maxillary margin is gently arched in both holotypic and referred skulls (Figs 22, 23).

Theholotype preserves the posterior one-half of the cervical series (C5-9), and much of the dorsal column is preserved in the referred skeleton. Centrum length is nearly constant, measuring approximately 5 mm (Fig. 30; Table 4). In Heterodontosaurus, in contrast, centrum length decreases in posterior cervical vertebrae by approximately 20% (Fig. 72, Table 7). Posterior dorsal centra have deeply concave sides and join as a gently arched series (Fig. 20). Their hatchet-shaped neural spines are longer than deep and nearly touch at their anterior and posterior extremities (IVPP V17090). In Heterodontosaurus, in contrast, the neural spines of the posterior dorsal vertebrae are deeper than long and well separated (Fig. 72). The sacral vertebrae are fused in IVPP V17090, and it is possible to measure only the centrum length of S1 (Table 4).

The proximal one-half of the caudal series is preserved in the holotypic skeleton and nearly as much in the referred skeleton (Fig. 25). The tail is very long, exceeding the length of the precaudal column by the twentieth caudal vertebra (Fig. 30). Caudal centra increase in length by approximately 40% from the first to the tenth caudal vertebra. By the seventh caudal vertebra, the neural arch is low and hatchet-shaped and the opposing chevrons subtriangular rather than strap-shaped (Fig. 25). At this point in the tail, a sheath of parallel ossified tendons are present spanning the neural spines and chevrons. By the thirteenth caudal vertebra, the neural spines are reduced to a ridge, and the chevrons are boat-shaped (Fig. 25). Although a few epaxial ossified tendons are present near the neural arches of the posterior dorsal, sacral and anterior caudal vertebrae (IVPP V17090), the sheath of ossified tendons starting around the seventh caudal vertebra would have stiffened mid and distal portions of the tail. The caudal structure in Tianyulong is remarkably similar to that in dromaeosaurid theropods (Ostrom 1969), the neural spines and chevrons changing in a similar manner distally and the parallel sheath of ossified tendons equivalent to the parallel, attenuated processes of the prezygapophyses and chevrons. Probably the more mobile tail base and stiffened mid and distal tail functioned in a similar manner as a balancing beam to enhance maneuverability.

Both scapulocoracoids are preserved in opposition in the referred skeleton (Fig. 20; IVPP V17090). The right forelimb is disarticulated dorsally, the humerus displaced from its articulation in the glenoid. The proximal end of the right humerus is exposed, the remainder of the bone damaged or embedded as it crosses a crack in the slab to a small corner of bone, the medial epicondyle. The right humerus has an estimated length of 28 mm. The left humerus, with an estimated length of 26 mm, lies in articulation with the glenoid of the scapulocoracoid proximally and the remainder of the left forelimb distally. The holotypic skeleton preserves both scapulocoracoids and humeri but little of the distal forelimb (Zheng et al. 2009: suppl. info.). Overlapping limb bone measurements suggest that the holotypic skeleton (STMN 26-3) is approximately 15% larger than the referred skeleton (IVPP V17090; Table 3).

Proximally, the scapula broadens gradually to the acromial process as in Heterodontosaurus (Santa Luca 1984). The neck is narrow as is most of the slender, straight strap-shaped blade (STMN 26-3; Zheng et al. 2009: suppl. info.). The distal end of the blade is not well exposed in either specimen. Expansion in width of the distal end of the blade, if present, would be proportionately less than in Heterodontosaurus. The coracoid has a subquadrate body and a prominent hook-shaped posterior process (STMN 26-3).

The humerus, which is best preserved in STMN 26-3, has a bulbous head, prominent deltopectoral crest, gently curved shaft and prominent distal condyles as in Heterodontosaurus. The olecranon process of the ulna, also best exposed in STMN 26-3, is very prominent proximally as in Heterodontosaurus. The radius tapers at mid-shaft and expands distally as in Heterodontosaurus to broadly contact a well ossified, but poorly preserved, carpus (Figs 26, 27).

The manus probably retained all five digits as in Heterodontosaurus, although only fragments of digits IV and V remain (Figs 26, 27; Table 5; IVPP V17090). The proportions of the manual digits are most unusual among ornithischians including Heterodontosaurus. Digit II and metacarpal 2 are longer than digit III and metacarpal 3. This mismatch in length is due to the shortening of all bones in digit III, as metacarpal 1 is also longer than metacarpal 3 (Figs 26, 27). The block-shaped bases and divided distal condyles of metacarpals 1-3 are well exposed. Metacarpal 1 has a particularly broad base as in Heterodontosaurus. The phalangeal formula 2-3-4-?-? is typical for the inner three digits of basal dinosaurs and archosaurs in general.

The two phalanges of manual digit I diverge medially from the others. The nonungual phalanges have proximal intercondylar processes articulating between paired distal condyles with well-formed collateral ligament pits as in Heterodontosaurus. The penultimate phalanges in digits II and III, in addition, are longer than the preceding phalanges, suggesting an enhanced grasping function for the manus in Tianyulong as in Heterodontosaurus. The unguals are transversely compressed and trenchant, that on digit I longer than those on digits II and III (Figs 26, 27).

Little can be said about the ilium, which is poorly preserved in both skeletons. The ischia and pubes are preserved in lateral view (STMN 26-3; Zheng et al. 2009: suppl. Info). The ischial shaft is gently dorsally bowed, a curve not present in Heterodontosaurus. The ischial shaft is transversely compressed and lacks any flanges or processes, including the lateral flange of Heterodontosaurus (Santa Luca 1984) or the obturator process present in many ornithischians.

Of the pubis, only the postpubic process is preserved (STMN 26-3; Zheng et al. 2009). Its shaft is rod-shaped and very slender as in Heterodontosaurus but only one-half its length. The process tapers to a slender tip just beyond the midpoint of the ischial shaft.

The femoral head is large and round, and the femoral shaft is robust with a proximally placed pendant fourth trochanter (IVPP V17090). Whether the anterior trochanter is separate as in the Kayenta heterodontosaurid and Abrictosaurus or fused with the greater trochanter as in Fruitadens cannot be determined. The tibia is extremely elongate, measuring more than 140% the length of the femur (Table 3). The cnemial crest is lower than in Heterodontosaurus, so that in medial view the anterior margin of the proximal end of the tibia is straight rather than convex (Fig. 20). The posterior condyles expand farther away from the central axis of the shaft. A fibular crest on the tibia supports the proximal end of the fibula, which has a swollen anterior trochanter (STMN 26-3; Zheng et al. 2009: suppl. info.). Distally, the fibula tapers to a slender rod. The tibia, fibula and proximal tarsals appear fused as a tibiotarsus as in most specimens of Heterodontosaurus. The distal tarsals are dislodged on one side of the referred skeleton and thus may not have coossified with the metatarsus.

In the pes, the proximal ends of metatarsal 1-4 appear to be coossified. Pedal digit I is very short, the tip of its ungual extending just beyond the condyles of metatarsal 2 (Figs 28, 29). That ungual, however, is quite large, equaling the length of the ungual on digit II and exceeding the length of the ungual on digit IV. Pedal digit III, in contrast, is slightly longer relative to pedal digits II and IV, a proportion that fits with other cursorial adaptations of the hindlimb. There is no trace of pedal digit V, which may owe its absence in Tianyulong to postmortem loss. The unguals are transversely compressed and taper to slender tips (Table 5). Several of the unguals preserve portions of the keratinous claw sheath (Figs 28, 29).

Tianyulong has very unusual skeletal proportions that differ from those in Heterodontosaurus and, to the extent that comparisons are possible, other heterodontosaurids (Figs 30, 72; Tables 3–9). The skull is proportionately very large and the hindlimbs are very long, whereas the neck and trunk are proportionately small and the forelimbs are reduced in length. Using Archaeopteryx for initial comparison (Wellnhofer 1993), the skull of Tianyulong is 25% longer than in the Eichstätt specimen, which has a nearly identical femoral length (Table 5). The tibiofemoral ratio in this specimen of Archaeopteryx is high (140%), although still slightly lower than in Tianyulong (143%). The humerus, in contrast, is approximately 48% the length of that in Archaeopteryx. As discussed further below (see Discussion, Body size and proportions), these unusual proportions give this small-bodied heterodontosaurid the appearance of a large-headed, short-armed, long-legged dwarf (Fig. 30) with very different proportions than in Heterodontosaurus (Fig. 72).

Small-bodied ornithischians with the following features that likely would constitute synapomorphies in phylogenetic context: (1) cheek tooth crowns that are taller than wide, (2) jaw joint set below the axis of occlusion between maxillary and dentary teeth.

The most inclusive clade containing Heterodontosaurus tucki Crompton and Charig 1962 but not Tianyulong confuciusi Zheng et al. 2009, Fruitadens haagarorum Butler et al. 2010, Echinodon becklesii Owen 1861. This stem-based phylogenetic definition for this new heterodontosaurid subgroup includes Heterodontosaurus tucki and other derived heterodontosaurid species but specifically excludes basal species with less modified teeth that retain subtriangular crowns enveloped in symmetrical enamel.

Earliest Jurassic (Hettangian-Sinemurian) to Middle Jurassic or earliest Late Jurassic (Aalenian-Callovian), ca. 197-165 Ma (Gradstein and Ogg 2009; Pol et al. 2011); distribution limited to southern localities including Argentina and southern Africa (Fig. 1B).

Authorship of family group names is given to the author of the first name coined within the family group (ICZN 1999), which in this case is Heterodontosauridae Kuhn 1966.

NHMUK RU B54, ventral portion of a skull and articulated skeleton lacking the mid and distal caudal vertebrae, right coracoid, left carpus, portions of the left manus, and portions of the right hindlimb (Table 1). The author and other researchers have examined fragmentary postcranial material of a second individual catalogued with the holotypic specimen before their transfer to the NHMUK collection. At present there is no evidence that this material pertains to Abrictosaurus consors (Norman et al. 2011: 236–237).

Stream-side exposure by the town Nosi (= “Noosi”), 8.2 km east of Whitehill, southern Lesotho; S30°03', E28°32' (Thulborn 1974; Kitching and Raath 1984) (Fig. 1B).

Top unit (dull red sandstone) of the Upper Elliot Formation; Lower Jurassic, Hettangian, ca. 202-197 Ma (Thulborn 1974; Knoll 2005; Gradstein and Ogg 2009).

Heterodontosaurid ornithischian characterized by the following autapomorphies: (1) premaxillary tooth 2 and 3 with tall, subcylindrical crowns, the latter possibly representing a reduced caniniform tooth; (2) dentary tooth 2 with subcylindrical crown, possibly representing a reduced caniniform tooth; (3) maxillary and dentary teeth with flat lateral and medial crown surfaces lacking discrete marginal or median ridges; and (4) maxillary crowns in the middle of the tooth row have deep parallel-sided crowns that do not expand in mesiodistal width toward their apex. Unlike other heterodontosaurids, there is no dentary or premaxillary caniniform teeth, although both the dentary and premaxillary have teeth with subcylindrical crowns that may be reduced caniniform teeth.

A small set of autapomorphies diagnose this genus and species, which otherwise closely resembles Heterodontosaurus tucki. The lengthy initial diagnosis for Abrictosaurus consors (as Lycorhinus consors; Thulborn 1974: 153–154) amounted to an abbreviated description. The only uniquely derived feature in the diagnosis (two premaxillary teeth) is erroneous, as the base of a third premaxillary tooth is present (Fig. 31). The revised diagnosis in Hopson (1975: 304) is problematic, as it draws on features from the holotype of Abrictosaurus consors as well as from a specimen (NHMUK RU A100) that is referred below to Lycorhinus angustidens (Table 2).

The recent revision of the diagnosis by Norman et al. (2011: 236) included five features. Two of these partially overlap those enumerated in the diagnosis above, namely the absence of enlarged caniniform teeth and reduced ornamentation on cheek tooth crowns. One feature listed by Norman and colleagues (12-14 dentary teeth) is primitive with a broader distribution than Abrictosaurus consors. Another feature (dorsoventrally expanded anterior end of the dentary) is regarded here as a synapomorphy for derived heterodontosaurids including Heterodontosaurus tucki. The absence of a projecting maxillary ridge and buccal emargination, the final feature cited in the revised diagnosis of Norman et al. (2011), is regarded here as an artifact of preservation. The neurovascular openings, everted rim on the maxilla, and form of the opposing depression on the dentary clearly indicate that Abrictosaurus consors had a buccal emargination comparable in depth to that in other heterodontosaurids (Figs 34A, 35). Postmortem transverse compression of the skull has brought the mandibles together and flattened prominent facial features.

Abrictosaurus consors is most similar in cranial and postcranial morphology to Heterodontosaurus tucki. They differ most noticeably in the absence of well-formed caniniform teeth in Abrictosaurus consors and in the shape and ornamentation of noncaniniform teeth. The presence and size of caniniform teeth may be the product of sexual dimorphism (Thulborn 1974), although the rarity of reduced caniniform teeth among the many specimens of heterodontosaurids now in collections argues against this interpretation (Norman et al. 2011).

The holotypic skeleton of Abrictosaurus consors (NHMUK RU B54), which is preserved for the most part in natural articulation, is the most complete specimen of a heterodontosaurid with the single exception of a nearly complete skeleton of Heterodontosaurus tucki (SAM-PK-K1332; Santa Luca et al. 1976; Santa Luca 1980). The following brief description of the holotypic specimen clarifies aspects of the morphology critical to the taxonomy and phylogenetic relationships of this important heterodontosaurid. The skull and skeleton require further preparation before a more detailed description is possible. The skull and femur of the single known specimen of Abrictosaurus consors are approximately 71% and 68% of the length of the skull and femur in Heterodontosaurus tucki (SAM-PK-K1332; Table 3), respectively. This suggests that Abrictosaurus consors probably grew to a comparable adult body size.

Preserved in two pieces, the skull is transversely compressed and better exposed in left lateral view (Fig. 31). Much of the dorsal skull roof, braincase, and posterior end of the lower jaws are broken away. The postcranial skeleton is preserved on separate blocks, one preserving an articulated right forelimb (Fig. 36) and the other the ilia, an ischium, sacrum and left hindlimb (Fig. 37). The phalanges of the right manus cross onto the skull block and are partially exposed near the scleral ring in the left orbit (Fig. 31).

Both premaxillae are partially preserved with the right shifted slightly ventral to the left. Portions that are broken on both sides include sections of the alveolar border, the internarial processes, and the distal end of both posterolateral processes (Fig. 31). The posterior end of the alveolar margin on the left side was originally shown as complete (Fig. 34A) but now is damaged (Fig. 31). The premaxillary tooth row is set below the maxillary tooth row, although the “overhanging and hood-like” positioning of the premaxilla (Thulborn 1974: 154; Fig. 34A) appears to have been enhanced by postmortem displacement. The form of the premaxilla is very similar to that in Heterodontosaurus. Although incomplete, the narial fossa is broad and extends toward the alveolar margin (Figs 31, 35).

The posterolateral process of the premaxilla narrows slightly posterior to the external naris and then broadens in transverse width distally, which in Heterodontosaurus is directly related to the anterior margin of the arched premaxilla-maxilla diastema (Fig. 59). In Abrictosaurus the region of a potential arched diastema, however, is incomplete on both sides (Figs 31, 32). The lower dentition would not have necessitated a diastema, as there is no development of the typical heterodontosaurid dentary caniniform tooth (Fig. 35).

Thulborn suggested, nonetheless, that caniniform crowns in heterodontosaurids might be sexually dimorphic, that NHMUK RU B54 may represent a female, and that a deep arched diastema may have been present in Abrictosaurus without an opposing lower caniniform tooth. Norman et al. (2011) presented a similar interpretation (Fig. 34B). Although sexual dimorphism remains a plausible hypothesis, details in the region of the diastema on both sides suggest that only a small diastema could have been present in Abrictosaurus. First, the alveolar margin of the right maxilla, which is preserved farther anteriorly than the left, extends anteriorly as a horizontal border as far as the anterior end of the dentary. There is no hint of an arched embayment on the right side. Second, the left dentary tooth 2, which is positioned below the proposed diastema, is truncated by an oblique wear facet (Fig. 32), indicating tooth-to-tooth contact with an opposing anterior maxillary crown (Figs 31, 32). Although no maxillary crowns are preserved in place above this tooth on either side, this planar wear facet suggests that a maxillary crown dorsal to this tooth was originally present on the anterior maxillary alveolar margin and that it has broken away (Fig. 35). It seems doubtful, finally, that the hypertrophied tooth in a caniniform-diastema complex would be reduced, while the recessed fossa that houses such a crown upon jaw closure would be maintained.

The maxilla has a dorsoventrally deep buccal emargination, which is bordered dorsally by an arched row of large neurovascular foramina and the everted and gently arched external rim of the antorbital fossa. The narrow width of the anterodorsal ramus brings the premaxilla in close proximity to the antorbital fossa as in Heterodontosaurus. At the base of this maxillary ramus, the anterior corner of the external opening of the antorbital fossa has a broader arc than in Heterodontosaurus. The external opening of the antorbital fossa is subtriangular (Fig. 35) as in Heterodontosaurus. As seen in lateral view (Figs 31, 35), the dorsally arched alveolar margin of the maxillary tooth row also resembles the condition in Heterodontosaurus.

The subtriangular lacrimal may be slightly broader dorsally than in Heterodontosaurus, its lateral surface lacking the subtriangular external fossa present in the latter genus. A slender, dorsoventrally flattened palpebral is preserved near its natural articulation with the lacrimal. In cross-section and length, the palpebral closely resembles that in Heterodontosaurus (Figs 31, 35). This bone was previously identified as the prefrontal (Fig. 34).

The partially preserved quadratojugal and ventral end of the quadrate form a T-shaped junction in Abrictosaurus as in Heterodontosaurus. The jaw joint appears to be preserved in natural articulation (Fig. 31). In Abrictosaurus the jaw articulation is set below the tooth row to the level of a horizontal line passing through the middle of the dentary ramus (Figs 31, 34).

The dentary has a deep, arched buccal emargination that dissipates anteriorly near the subconical second tooth (Figs 31, 35). An anterior dentary foramen is located farther ventrally under the same tooth as in Heterodontosaurus (Figs 31, 35). The anterior end of the dentary is more strongly expanded dorsoventrally than in Heterodontosaurus, although both are similar in form. The articular surface for the predentary, as in Heterodontosaurus and Pegomastax gen. n. sp. n., is saddle-shaped; in lateral view the articular surface is dorsoventrally convex and transversely concave (Fig. 32). In addition, a rugose swelling here termed a dentary boss is present on the lateral aspect of the anterior end (Figs 32, 59).

The posterior margin of the dentary is not well preserved; the notched posterior margin of the right dentary suggests that an external mandibular fenestra of moderate size may have been present as in Heterodontosaurus tucki (contra Thulborn 1974: 158). An internal mandibular foramen is also present between the splenial and prearticular in medial view of the right mandibular ramus (contra Thulborn 1974: 158; Fig. 31). The splenial tapers anteriorly before reaching the symphysis, where a section of Meckel’s canal is exposed. The canal is represented by a narrow trough running just above the ventral margin of the dentary (Fig. 31) as in Echinodon, Heterodontosaurus and several other heterodontosaurids.

There are three premaxillary crowns as in other heterodontosaurids (rather than two, contra Thulborn 1974: 15) (Fig. 31, 32). The first and smallest of the premaxillary teeth is preserved on the left side, its crown broken away flush with the alveolar margin (Figs 31, 32). This tooth (pm1) is inset a good distance from the anterior end of the premaxilla. Pm2 and 3 are subconical and slightly recurved, their apical ends broken away. Judging from the preserved portions of the crowns, more of the crown tip of pm3 is broken away than in pm2. Their smooth crowns are bounded mesially and distally by low, but distinct, edges. The crown merges with the root without an intervening neck (Fig. 10A), unlike the premaxillary teeth in Echinodon and the first two premaxillary crowns in Heterodontosaurus. In Abrictosaurus and other heterodontosaurids, pm3 is the largest tooth in the premaxillary series, although in Abrictosaurus the size differential is minor. Pm3 is not transversely compressed, and there is no development of serrations on mesial or distal edges of the crown (Fig. 31).

A complete maxillary series probably included 14 teeth, as estimated from the 12 preserved maxillary teeth in the left maxilla and the space between these and the anterior end of the maxilla on the right side (Figs 31, 35). At least two small teeth may have been present at the mesial end of the series, indicating there were more than 12 maxillary teeth (contra Thulborn 1974: 158). The teeth are numbered accordingly, accounting for two missing anterior crowns. In Abrictosaurus, Heterodontosaurus and Echinodon, the tooth rows are nearly straight in apical view, whereas in Lycorhinus and “Geranosaurus” they are bowed medially (Gow 1990; Broom 1911). Opposite to the condition in Echinodon, the maxillary teeth are slightly taller than opposing dentary teeth along the tooth row (Figs 31, 35).

Abrictosaurus has very distinctive maxillary crowns, best exemplified in the middle portion of the tooth rows. The maxillary crowns are extremely tall, their height more than twice their maximum width in the anterior center of the tooth row (teeth 5-10). Crown proportions (height versus width) decrease distally toward the end of the tooth row. Some of the extra crown height of the central teeth is accommodated by an upward arching of the alveolar margin as in Heterodontosaurus, which keeps the axis of occlusion horizontal (Fig. 35). In Echinodon and Lycorhinus, in contrast, the crowns in the tooth row are more similar in size and the alveolar margins less arched.

The cingulum is reduced in Abrictosaurus with the crown in most teeth expanding gently from the root. The best-developed cingulum occurs in smaller teeth at the distal end of maxillary and dentary tooth rows. In these teeth, as in Heterodontosaurus, the crown expands more abruptly from the root (Figs 31, 32).

In labial view, mesial and distal crown edges are nearly parallel, with no development of marginal ridges like those in Echinodon, Lycorhinus, and Heterodontosaurus. The denticulate margin near the top of the crown has approximately five to six denticles to each side. These margins slope at a low angle, about 30 degrees from the perpendicular to the crown axis. The denticulate portion of the crown, thus, is confined to the apical 25% of crown height. In cross-section, the crowns are blade-shaped and transversely narrower than in Lycorhinus and Heterodontosaurus. The enamel may have an asymmetrical distribution, thicker on the labial side of the crown, but this is not well established.

The maxillary crowns are canted at a slight angle to their roots, directing the crown lingually. Low-angle wear facets graze the lingual side of the maxillary crowns. As in Heterodontosaurus more than one facet is present on a single crown, as opposing cheek teeth are not aligned one-to-one. Contrary to Thulborn (1974: 154, 159), the wear facets do not lie in a single plane and two wear facets, rather than only one, occur adjacent to one another on several crowns. Maxillary and dentary teeth did not occlude one-to-one as implied by Thulborn, nor do they occlude in strict alternation as suggested by Weishampel (1984: 53).

There is indisputable evidence for active tooth replacement in Abrictosaurus. The base of the penultimate maxillary tooth on the right side has broken away to expose an erupting crown (Fig. 33). Thulborn (1974: 159) stated there is no evidence of active tooth replacement in Abrictosaurus, suggesting that the entire maxillary and dentary tooth rows were replaced simultaneously during aestivation (Thulborn 1978). The evidence from the dentition does not support this scenario.

The left dentary has 14 teeth, which may comprise a complete tooth row (Figs 31, 35). A small distalmost dentary tooth, however, is preserved on the right side that may constitute dentary tooth 15, which could be obscured by matrix on the left side. The first two dentary teeth have crowns with an atypical shape (Figs 32, 35), and these may correspond to the peglike tooth and caniniform tooth in some heterodontosaurids such as Lycorhinus. The small first tooth has a smooth subconical crown that is swollen slightly above the root. The larger second crown, which also shows some swelling above the root, has a convex mesial margin and low unornamented distal carina (Fig. 32). This tooth may represent a reduced caniniform tooth, as it occupies a similar position at the anterior end of the buccal emargination.

The remaining dentary teeth have diamond-shaped crowns that are shorter than opposing maxillary crowns (Fig. 31). The first five dentary tooth crowns do not overlap one another. The remainder of the dentary crowns and all of the maxillary crowns are closely spaced or in contact. As much as 50% of the crown is bordered by mesial and distal denticulate margins, which are more steeply inclined than the denticulate margins of opposing maxillary crowns. Echinodon also exhibits a similar differential in the inclination of the denticulate margins between maxillary and dentary crowns, the latter also more steeply inclined.

The reconstruction of the partial skull and dentition of Abrictosaurus (Fig. 35) differs in several regards from previous reconstructions (Fig. 34; Thulborn 1974; Norman et al. 2011). In the present reconstruction, the premaxilla is rotated in a clockwise direction to restore its position relative to the maxilla and bring the premaxillary tooth row, now with three teeth rather than two, closer to a horizontal orientation. Based on evidence from the right side of the skull, the anterior end of the maxilla is restored, the diastema reduced, and two small teeth added to the anterior end of the tooth row (Fig. 35). The alveolar margins of the dentary and maxillary tooth rows are arched ventrally and dorsally, respectively, as preserved on the skull and as in Heterodontosaurus. Using the better-preserved skulls of Heterodontosaurus as a guide, the palpebral is rotated clockwise to align it with the dorsal margin of the snout. The external opening of the antorbital fenestra is enlarged, following preserved margins on the maxilla and lacrimal. Finally, the ventral margins of the predentary and anterior end of the dentary are reduced and expanded, respectively, as preserved (Fig. 34).

The articulated cervical series is only partially exposed. The atlas and axis are incomplete, and C3-9 are exposed in dorsal view. Judging from their neural arches, length decreases markedly from C6-9 as in Heterodontosaurus. The neural arches of C8 and C9 are only approximately one-half that of C3.

In C3 the neural spine is low and ridgelike and the postzygapophyses have smooth dorsal surfaces lacking any development of epipophyseal processes. In Heterodontosaurus, in contrast, a well developed epipophysis projects as a subconical process from the postzygapophysis of C3. The longer, arched postzygapophyses of C5 and shorter postzygapophyses of C6 also lack discrete epipophyseal processes. A low crest at the base of each of these postzygapophyses joins the low neural spine. In C6 the tab-shaped neural spine is located at the anterior end of the neural arch. Although the form of these vertebrae is similar to that in Heterodontosaurus, important differences include the absence of the unusual anterodorsally inclined neural spines in C5 and C6.

Ossified tendons are associated with the neural arches of the dorsal and sacral vertebrae, but as in Heterodontosaurus few of these are in natural position. Ossified tendons appear to be limited to dorsal and sacral regions of the vertebral column, and none is present in the proximal portion of the caudal series (Fig. 37).

The scapula, the only bone of the pectoral girdle that is exposed, shows a prominent acromial expansion and strap-shaped proximal scapular blade. Little else can be said without further preparation.

The right forelimb is nearly complete with most of its bones in articulation or near their natural location and with the manus in ventral view overlapping the left orbit of the skull (Fig. 36). The humerus, exposed in anterior view, is poorly preserved proximally and distally. The deltopectoral crest projects strongly from the proximal shaft. The length of the crest is about 34% of humeral length (Table 6). In Heterodontosaurus, the proximal end of the crest arches more abruptly from the shaft than in Abrictosaurus, and the relative length of the crest is greater (approximately 42% of the humerus) (Table 8).

The radius and ulna are preserved on both sides but are incomplete distally and only partially exposed (Fig. 36). In Heterodontosaurus the ulnar shaft thickens on its ventral aspect prior to the strong proximally projecting olecranon process. Although that process is not preserved or exposed in Abrictosaurus, the right ulna shows the ventral thickening of the proximal shaft.

The carpus is composed of many elements, which may have been assembled in articulation with a compact arrangement as in Heterodontosaurus. In the right carpus, three carpals are preserved proximal to metacarpals 3 through 5 (Fig. 36). The large ovoid bone distal to the ulna and closest to the base of metacarpal 5 may represent the ulnare, which in Heterodontosaurus is even larger and more subrectangular in shape. The pair of smaller subspherical elements associated with the bases of metacarpals 3 and 4 may represent distal carpals. In Heterodontosaurus the distal carpals in the lateral side of the carpus appear more lenticular in shape with an overlapping arrangement (Fig. 67C).

The metacarpals are proportionately elongate relative to either the bones of the forearm or humerus as in Heterodontosaurus (Fig. 36). Metacarpal 2, the most complete of the series, is approximately 32% of humeral length (Table 6). The same proportion in Heterodontosaurus is only 27% (Table 8), and thus the hand in Abrictosaurus appears to be proportionately longer. An unusual feature of the metacarpus in Heterodontosaurus is that metacarpal 2 is slightly longer and more robust than metacarpal 3 (Table 8). Their distal condyles, nevertheless, are nearly aligned, as the base of metacarpal 2 is inset slightly into the carpus relative to metacarpal 3 (Fig. 36). The slightly inset position of the base of metacarpal 2 relative to metacarpal 3 also is preserved on both sides in the articulated skeleton of Heterodontosaurus (Fig. 65). Metacarpals 1-3 are more robust and longer than metacarpals 4 and 5, the former measured from the left manus. Metacarpal 5 is the shortest metacarpal with a shorter length than in Heterodontosaurus (Tables 6, 8).

The proximal ends of the metacarpals show edges that appear to square the base of the bone as in Heterodontosaurus and the Kayenta heterodontosaurid. The squared bases articulate against each other in Heterodontosaurus, and a similar condition may hold for at least the medial metacarpals in Abrictosaurus (Fig. 36). In Heterodontosaurus, however, the metacarpus is exposed only in dorsal view (Figs 65, 67). In Abrictosaurus, in contrast, the metacarpus is currently exposed only in ventral view. The distal ends of metacarpals 1-3 are expanded, and thus it seems likely that distal extensor pits would be present dorsally for hyperextension of the phalanges as in Heterodontosaurus.

The right metacarpus, exposed in ventral view, provides the best information regarding the disposition of the digits (Fig. 36). Metacarpal 1 diverges medially only slightly from an axis established on the bones of the forearm. As in Heterodontosaurus (Fig. 67A), the distal condyles are strongly asymmetrical, which cants the phalanges of digit I medially. Digit V, preserved only in the right manus, diverges at nearly a right angle laterally as in Heterodontosaurus and other bipedal ornithischians such as Hypsilophodon (Galton 1974a). The articulation of the base of metacarpal 5 with the lateral aspect of the ulnare, which is also present in Heterodontosaurus, seems to be the natural disposition of this short digit (Figs 36, 67).

The phalangeal formula is poorly known in Abrictosaurus. In the left manus, digit IV has two phalanges, and in the right manus digit V has one (Fig. 36). The distal ends of phalanges IV-2 and V-1 are rounded, and so the presence of a diminutive nonungual terminal phalanx in each digit, as occurs in Heterodontosaurus (Fig. 67), cannot be ruled out. The phalanges of digits I-III are not well exposed. The proximal phalanges of digits I-III show a similar decrease in length to the lateral side as in Heterodontosaurus. The distal ends of phalanges I-1 and II-1 are preserved on the skull block over the left orbit (Fig. 31). It is possible that the remaining phalanges of digits I-III are embedded in matrix within the orbit. The edge of an extremely long recurved element may represent the edge of an ungual. Detailed preparation of the skull block is warranted.

The ilium, preserved in part on both sides, is very similar to that in Heterodontosaurus (Fig. 37, Table 6). The gently everted dorsal margin arches between the slender preacetabular process and the proportionately long postacetabular process. Although incomplete, the anterior end of the preacetabular process appears to taper gradually rather than terminating with a gentle lobe-shaped expansion as in Heterodontosaurus (Santa Luca 1980: fig. 18A). The acetabulum, also similar to that in Heterodontosaurus, is completely open with no development of a supraacetabular rim. The pubic peduncle is longer and narrower than the more robust ischial peduncle, which does not project laterally as prominently as in Heterodontosaurus (Fig. 68). Only the proximal shaft of the right ischium is preserved. No obturator process is visible, but the shaft is not completely exposed (Fig. 37).

The general form of the femur does not depart significantly from that in Heterodontosaurus (Santa Luca 1980, fig. 18B). The shaft is bowed with a pendant fourth trochanter located proximal to mid-shaft (Fig. 37) as in Heterodontosaurus. There is no development of an anterior intercondylar groove. The anterior trochanter, which extends to the level of the greater trochanter, is separated from the shaft of the femur by a deep cleft, as seen in lateral view (Fig. 37). In Heterodontosaurus, in contrast, the lesser trochanter projects dorsally alongside the shaft, to which it is fully coossified (Fig. 68). The greater trochanter in Abrictosaurus is proportionately narrow, subequal in anteroposterior width to the anterior trochanter (Fig. 37). In Heterodontosaurus and neornithischians in general, the greater trochanter is always broader than the lesser trochanter in lateral view. The pendant fourth trochanter is located on the proximal half of the femoral shaft as in Heterodontosaurus.

In Heterodontosaurus several features in the crus arose in parallel among coelurosaurian theropods, including an articular crest protruding from the shaft of the proximal end of the tibia to support the reduced shaft of the fibula, reduction of the fibular shaft to a narrow rod, and coossification of the distal ends of the tibia and fibula with the proximal tarsals. Distal coossification of the tibia, fibula, and proximal tarsals (as a “tibiofibulotarsus”), however, may be variable in Heterodontosaurus. A referred right tibia, fibula, and astragalocalcaneum do not exhibit obliterating coossification between the crus and tarsus (Fig. 70). In Abrictosaurus most of the fibular shaft is similarly reduced to a rod, but there does not appear to be any development of a lateral supporting flange on the tibial shaft or distal fusion between the crus and tarsus (Fig. 37).

The tarsus includes the astragalus, calcaneum, and lateral and medial distal tarsals, which do not appear to be coossified (Fig. 37). The flat, platelike medial distal tarsal is preserved in articulation over metatarsals 2 and 3, whereas the smaller cuboid lateral distal tarsal is displaced a short distance from its articulation over metatarsal 4.

The metatarsal shafts are not coossified, and the pes as preserved is less compact than in Heterodontosaurus (Fig. 37). The distal end of metatarsal 3 diverges laterally, and the shaft of metatarsal 4 follows a S-shaped curve. These plesiomorphic features are only weakly expressed in Heterodontosaurus. In Abrictosaurus metatarsal 1 tapers toward its proximal end as an extremely thin splint, which is very narrow both transversely and dorsoventrally. Metatarsal 3 is the longest, with metatarsals 2 and 4 subequal in length. An extensor pit above the distal condyles occurs only on metatarsal 3 rather than on both metatarsals 3 and 4 as in Heterodontosaurus. Extensor pits are also developed above the distal condyles of the proximal phalanges of digits I-III. Only pedal digits I and IV preserve complete phalangeal series, which are consistent with the primitive ornithischian phalangeal formula (2-3-4-5-0).

SAM-PK-K337, nearly complete, articulated skull.

SAM-PK-K1332, articulated skull and skeleton lacking only a few mid and distal caudal vertebrae; SAM-PK-K10487, anterior portion of a juvenile skull including the left orbit with palpebral and anterior portion of the lower jaws with the predentary; SAM-PK-K1328, partial postcranial skeleton including dorsal and caudal vertebrae, forelimbs and hindlimbs; SAM-PK-K1334, left maxilla with six teeth and portions of adjacent bones preserving the antorbital fenestra, the ventral end of the lacrimal, and the anterior end of the jugal; AMNH 24000, posteroventral portion of skull with articulated lower jaws preserving the posterior half of maxillary and dentary tooth rows, parts of the right jugal, quadratojugal and quadrate, posterior half of the right lower jaw, and traces of the anterior cervical vertebrae.

On the mountainside behind the trading store in Tyinindini at an altitude of 1890 m, Transkei (Herschel) District, Cape Province, South Africa; S30°32', E27°32' (Crompton and Charig 1962; Kitching and Raath 1984) (Fig. 1B).

Upper Elliot Formation and Clarens Formation (Smith 1990; Knoll 2005); Lower Jurassic, Hettangian to Sinemurian, ca. 200-190 Ma (Crompton and Charig 1962; Santa Luca et al. 1976; Gradstein and Ogg 2009).

Heterodontosaurid ornithischian characterized by the following autapomorphies: (1) cheek tooth crowns subrectangular in cross-section; (2) prominent crown margins and primary ridge resulting in mesial and distal paracingular fossae on labial and lingual faces of maxillary and dentary crowns, respectively; (3) asymmetrical enamel on maxillary and dentary crowns (reduced in thickness on lingual and labial sides of maxillary and dentary crowns, respectively); (4) lacrimal with shallow lateral fossa; (5) jugal with extension of antorbital fossa onto the orbital ramus; (6) jugal flange posteroventrally inclined; (7) trapezoidal anterior surangular foramen; (8) axis and C3 neural spine with lateral flange; (9) C5 and C6 neural spines project anterodorsally; (10) C3-C7 with subcylindrical parapophyses; (11) mid dorsal vertebrae (D6-D10) with Y-shaped transverse processes (for di- and parapophyses); (12) scapulocoracoid foramen absent; (13) humeral epicondyles present; (14) lobe-shaped distal expansion of the iliac preacetabular process; (15) ischial peduncle narrow, columnar; (16) ischial shaft with laterally-directed crescentic flange at mid-length; (17) femoral anterior and greater trochanters coossified; (18) tibiofibulotarsus coossification (possibly variable); (19) tarsometatarsus coossification.

The initial diagnosis of Heterodontosaurus tucki by Crompton and Charig (1962: 1075) listed 10 features, none of which now stand as autapomorphies particular to the species. Some of the listed features, such as an edentulous predentary, are plesiomorphies of broad distribution, whereas others (presence of three premaxillary teeth, arched diastema) now characterize other heterodontosaurids. In a review of ornithischian taxa, Steel (1969: 7-8) listed more than a dozen cranial features in a revised diagnosis, but again none are regarded here as autapomorphies given the diversity of heterodontosaurids now known. The reduction of the cingulum and the presence of a jugal flange, for example, are now known in the recently described South American heterodontosaurid Manidens (Pol et al. 2011). In his description of the postcranial skeleton, Santa Luca (1980: 197) presented a revised diagnosis based solely on postcranial features. Of the 20 features listed, 4 are regarded here as autapomorphies that are currently known in other heterodontosaurids (autapomorphies 13, 17-19; humeral epicondyles; femoral anterior and greater trochanters coossified, tibiotarsus and tarsometatarsus coossification).

The recent revision of the diagnosis in Norman et al. (2011: 187) added 19 cranial and dental autapomorphies to a revised list of postcranial features from Santa Luca for a total of 44 characters. The authors marked 30 of these as potential autapomorphies. Many of these features, however, are problematic as autapomorphies. Of the 12 cranial features cited by Norman et al. (2011), 4 are retained in the revised diagnosis above (autapomorphies 2, 5–7). Some of the features cited by these authors but excluded here are known in other heterodontosaurids, such as the everted rim on the maxilla, presence of a jugal horn and flange, T-shaped quadratojugal, the reduction of the cingulum, and the presence of an external mandibular fossa (Thulborn 1974; Pol et al. 2011). The form of the basal tubera and basipterygoid processes also appear to be similar to that in Manidens (Pol et al. 2011). One of the cited features, “two fingerlike rami” of the surangular, is based on a damaged surangular as discussed below. One cranial feature new to the diagnosis above is the shallow subtriangular fossa on the lateral aspect of the lacrimal (autapomorphy 4) (Fig. 59).

Two of the seven dental features cited by Norman et al. (2011) are comparable to autapomorphy 2 in the diagnosis above. Other dental features cited by these authors are difficult to differentiate from the condition in several other heterodontosaurids, such as the distribution of serrations on caniniform teeth and reduction of the cingulum. The presence of extensive tooth wear, another cited feature, is not regarded here as diagnostic, given the presence of broad wear facets in Lycorhinus, Abrictosaurus and Pegomastax gen. n. sp. n.Dental features new to the revised diagnosis (autapomorphies 1, 3) include the subrectangular cross-section of the cheek teeth (Fig. 49C) and asymmetrical enamel on maxillary and dentary crowns. The distribution and thickness of enamel on the crowns of Heterodontosaurus tucki and other heterodontosaurines is poorly established, and so this may eventually characterize other species. Basal heterodontosaurids such as Echinodon and Fruitadens do not appear to have asymmetrical enamel on the crowns of the cheek teeth.

Six of the 10 postcranial features cited by Norman et al. (2011) as autapomorphic are reformulated above (as autapomorphies 13, 16–19). Other postcranial features cited by these authors are difficult to defend as autapomorphies, such as the number of precaudal or sacral vertebrae (see discussion below), the presence of epipophyses in anterior cervical vertebrae (present in Lesothosaurus), and a narrow scapular blade (present in other heterodontosaurids). Postcranial features new to the revised diagnosis (autapomorphies 8-12, 14, 15) include unusual features of the axial skeleton, the absence of the scapulocoracoid foramen, and the shape of processes in the pelvic girdle.

The aim of the following descriptive comments is to correct and extend where needed available accounts of the skull and skeleton of the best-known heterodontosaurid, Heterodontosaurus tucki. Initial accounts of the skull of Heterodontosaurus were based on two specimens, the holotype (SAM-PK-K337; Crompton and Charig 1962) and a better preserved referred specimen (SAM-PK-K1332; Charig and Crompton 1974; Weishampel 1984). Since that time, additional specimens have been prepared including an adult skull fragment containing a left maxilla (SAM-PK-K1334; Norman et al. 2011: Figs 30-33), the snout end of a subadult skull (SAM-PK-K10487; Butler 2008; Figs 38, 39), and the posterior half of another subadult skull (AMNH 24000; Figs 2C, 40-52, 61). All of these specimens except AMNH 24000 were considered in a recent more detailed description of the skull (Norman et al. 2011). The comments below extend that description and discuss aspects of cranial morphology where the interpretation here (Fig. 59) differs from either their text or figures.