SEVEN
Therizinosauroidea
TERESA MARYA
SKA
In the decade since the publication of the first edition of this book several important discoveries have greatly expanded our understanding of the group previously known as Segnosauria and resolved the controversy over their relationships (table 7.1). Rather than their being relegated to a chapter between sauropodomorphs and ornithischians and considered Saurischia sedis mutabilis (Barsbold and Maryanska 1990), new evidence has reaffirmed their identity as theropods, as proposed by Perle and Barsbold in their original descriptions of this group. These important new discoveries include individuals with simple, feather-like integumentary structures (Xu et al. 1999a), the first record outside of Asia, and a tantalizing dentary that may extend the temporal range of the group back to the Early Jurassic (Zhao and Xu 1998; Xu et al. 2001). Furthermore, the poorly known Therizinosaurus is now considered to belong to this group (Russell and Dong 1993a), resulting in several important changes to the nomenclature that require a brief explanation.
Based on their allocation of Therizinosaurus to the group including Segnosaurus, Russell and Dong (1993a) revised the nomenclature in recognition of the priority of Therizinosauridae Maleev, 1954. They synonymized the Segnosauridae Perle, 1979, with the Therizinosauridae and, following the ICZN Principle of Coordination endowing the superfamily name with the same date of publication as the family name, synonymized the Segnosauria Barsbold et Perle, 1980, with the Therizinosauroidea, within which they included the Therizinosauridae and the new Family Alxasauridae. Although the replacement of a familiar name with a less familiar one can cause confusion, these changes are required by the ICZN as long as (1) Therizinosaurus is indeed a member of this taxon and (2) the name for the taxon is considered to be of the family group. The small amount of evidence placing Therizinosaurus within the taxon is not outweighed by evidence for its placement in any other group, and neither diversity within the group nor the position of the group within Theropoda requires a hierarchical position above the suprafamilial level, so Therizinosauroidea is considered here the valid name for the group.
The relationships of these therizinosauroids were poorly understood for many years because of their unusual combination of features. The pelvis is opisthopubic, a feature known otherwise only in birds and their closest relatives among coelurosaurian theropods. However, the first digit of the foot, including the proximal end of the first metatarsal, is fully formed, and the first four metatarsals are spread out, in contrast to the compact, tridactyl pes usually found in theropods. Furthermore, the skull is highly specialized, with an edentulous rostrum and many symmetrical, nonrecurved teeth.
Perle and Barsbold (Perle 1979; Barsbold and Perle 1980; Barsbold 1983a) considered the group they called Segnosauria to most probably represent a separate lineage of predatory dinosaurs within Theropoda. However, Perle (1981) suggested that Segnosauria did not show characters unequivocally justifying their assignment to Theropoda or any other known dinosaurian higher taxon. An alternative hypothesis by Paul (1984a) contended that segnosaurs are related to ornithischians and that prosauropods were the closest relatives of these two taxa. Gauthier (1986) considered them to be related to sauropodomorphs, as did Sereno (1989), and Gauthier presented evidence for the monophyly of Saurischia and Sauropodomorpha contradicting any special relationship between prosauropods and ornithischians.
The discovery and description of Alxasaurus elesitaiensis Russell et Dong, 1994, based on two articulated skeletons, one of which is nearly complete, revealed sufficient theropod features to shift the weight of evidence to this group. The three-digit manus with a short first metacarpal, slender third metacarpal, and semilunate distal carpal capping metacarpal I is particularly persuasive evidence. The redescription of the skull of Erlikosaurus (Clark et al. 1994) also highlighted several features allying it with theropods, such as extensive basicranial pneumaticity and an intramandibular joint. The description of the primitive therizinosauroid Beipiaosaurus inexpectus (Xu et al. 1999a), further revealed that the first metatarsal of this taxon is compressed proximally and does not reach the tarsus, as in coelurosaurian theropods and unlike the plesiomorphic dinosaurian condition retained in sauropodomorphs.
Therizinosauroidea currently is represented by 11 species in ten genera, all but one from Asia. The great generic diversity is due in part to the incompleteness of the material and the resulting poor understanding of species-level relationships, however. The ten definitive members of the group are from the Cretaceous, but Eshanosaurus deguchiianus, from the Lower Jurassic Lower Lufeng Series of China (Xu et al. 2001), is tentatively referred to Therizinosauroidea.
TABLE 7.1
Therizinosauroidea
For simplicity, those species in monospecific genera are usually referred to below by only their generic nomen.
Definition and Diagnosis
Therizinosauroidea is here defined as the least inclusive clade containing Therizinosaurus and Beipiaosaurus. This definition is problematic as it excludes Eshanosaurus, a taxon that is tentatively referred to this group but may prove to belong to another when it becomes better known. A stem-based definition might have certain advantages over this node-based definition, but uncertainties in the sister-group relationships of therizinosauroids preclude a stable stem-based name. Indeed, given the instability of recent phylogenetic hypotheses for coelurosaurs, phylogenetic definitions for higher-level coelurosaur taxa may be premature.
Therizinosauroidea is currently diagnosed by the following features (see the “Systematics and Evolution” section below for further details): a toothless premaxilla with a sharp, ventrally projecting rim (also in many avialans); a very elongate external naris bordered caudoventrally by the maxilla (also in troodontids); an antorbital fossa surrounded by a well-developed rim except caudally (also in some oviraptorids, but to a lesser extent); a palate with greatly elongate vomers and rostrally reduced pterygoids; a greatly enlarged and pneumatized basicranium constricting the external acoustic meatus ventrally (possibly homologous with the expanded parasphenoid of troodontids and ornithomimids but not strictly comparable); a dentary with a rim on the lateral surface forming a horizontal shelf; dentary teeth increasing in size rostrally; maxillary and distal dentary teeth laterally compressed, symmetrical in lateral view, with coarse serrations (except in Eshanosaurus) and cylindrical roots; a humerus with strongly expanded proximal and distal ends, distal condyles on the cranial surface (convergent with birds and alvarezsaurids), and a sharply pointed tubercle on the caudomedial surface midshaft; two large distal carpals capping metacarpals I and II, the medial one semilunate; a deep and long preacetabular process of the ilium with a pointed cranioventral extremity that flares outward at a right angle to the sagittal plane; a long obturator process of the ischium contacting the pubis; the ascending process of the astragalus covering only part of the distal end of the tibia; and short and broad metatarsals. Other distinctive features that may be diagnostic, depending on the relationships of the group, are an edentulous premaxilla (also in ornithomimosaurs, oviraptorosaurs, and Neornithes); a maxilla dorsal to the palatal surface and facing ventrolaterally (oviraptorids); a large prefrontal (reduced in all coelurosaurs except Ornithomimosauria and possibly Alvarezsauridae); the mesial end of the dentary downturned (also in ornithomimosaurs and some troodontids); the coronoid absent (highly reduced in oviraptorosaurs, absent in Avialae); the deltopectoral crest of the humerus one-third or more the length of the bone (also in oviraptorosaurs); and the opisthopubic condition of the pelvis (also in dromaeosaurids and avialans). The skull and mandible also may be small relative to the postcranial skeleton, but the currently known material is inadequate to demonstrate this.
Anatomy
Skull and Mandible
The following description of therizinosauroid cranial structures is based mainly on Erlikosaurus andrewsi (Clark et al. 1994). Skulls of Alxasaurus and Beipiaosaurus are also known but have not yet been described (Dong and Xu, pers. comm.).
FIGURE 7.1. Skull and mandible of Erlikosaurus andrewsi: A–C, skull in A, ventral, B, left lateral, and C, dorsal views; D, E, left mandible in D, medial, and E, lateral views. Scale = 5 cm. (From Clark et al. 1994.)
Little of the postcranial skeleton of Erlikosaurus is preserved, but the size of the mandible relative to the postcranial skeleton in other taxa (Beipiaosaurus, Alxasaurus, Segnosaurus, Neimongosaurus) suggests that the skull was small (Russell and Russell 1993; Xu et al. 1999a) in comparison with that of many theropods.
The skull of Erlikosaurus is long and low (fig. 7.1), with large, elongate external nares surrounded by a horizontal narial shelf. It is delicately constructed, and the bone of the braincase and antorbital fossa regions is especially thin. The premaxillae form a wide, toothless beak that is heavily vascularized and was probably provided with a horny sheath during life. Its ventrolateral edge is sharp and vertical, unlike that of the otherwise similar ornithomimosaur premaxilla. The nasal process of the premaxilla is narrow and long, and the subnarial process is short. The maxilla forms the caudoventral margin of the external naris and bears a long jugal ramus caudal to the last maxillary tooth. The most rostral portion of the maxilla is edentulous, and the maxillary tooth row ends below the middle of the antorbital fenestra. The teeth are inset from the lateral edge of the maxilla, and the lateral surface of the maxilla beneath the antorbital fenestra faces ventrolaterally. There are only a few, large foramina along the dentigerous portion of the maxilla. The maxilla forms a well-developed, thin rim around the rostral edge of the antorbital fossa and a medial wall to the rostral third of the fossa. A promaxillary fenestra is apparently absent, unless a small one is present near the rostral or ventral edge of the fossa.
The prefrontal is large for a coelurosaurian, similar to that of ornithomimosaurians. Ventrally it contacts the dorsal surface of the T-shaped lacrimal. The long, rostrally narrow nasal forms two-thirds of the dorsal border of the elongate external naris but does not border it ventrally. The jugal is long and low rostrally and divided caudally. The caudal jugal process forms much of the ventral border of the infratemporal fenestra. The quadratojugal is shaped like an inverted T, overlapping rostrally the caudoventral process of the jugal, contacting the squamosal dorsally, and extending to the caudal end of the quadrate. The quadrate is vertical, with a strongly concave caudal margin, and a quadrate foramen is formed between it and the quadratojugal. The quadrate articulates only with the squamosal, well lateral to the side of the braincase.
The most rostral portion of the palate is formed by horizontal palatal processes of the premaxillae and maxillae. The choanae are bordered rostrally by the maxillae and vomer. Caudal to the palatal processes of the maxillae, the palate is highly vaulted along the midline. A small subsidiary palatine fenestra is present between the palatine and the pterygoid. The fused vomers form a long, vertical plate that is wedged rostrally between the palatal processes of the premaxillae and maxillae and reaches as far caudally as the lateral flange of the pterygoid. The large, long palatine is shifted slightly caudally.
The ectopterygoid bears a stout process ventrally, near the contact with the pterygoid, but much of the bone is missing. The palatal ramus of the pterygoid is short and lies close to the parasphenoid, and there is no interpterygoid vacuity. The quadrate ramus of the pterygoid is close to the braincase, and the basipterygoid processes are either small or absent. The narrow cultriform process is comparatively long and wedged between the vomers rostrally.
The basicranial and otic regions of the skull are swollen and strongly pneumatized (fig. 7.2). The expansion is caudal to the basipterygoid processes, unlike the parabasisphenoid bulla of troodontids and ornithomimosaurs. The semilunate orbitosphenoid is small, and the remaining bones of the braincase are co-ossified. The trigeminal opening of the braincase is divided into three passages laterally.
The parietals are fused and form a flat dorsal surface between the supratemporal fenestrae. The frontals are broad caudally and end abruptly dorsal to the caudal end of the antorbital fenestra. The postorbital and squamosal are both simple, triradiate bones, and the medial process of the postorbital is expanded and flattened (also in Beipiaosaurus). The infratemporal fenestra is oval with a caudodorsally directed long axis.
The occiput is broad below the occipital condyle and slopes rostrodorsally dorsal to the foramen magnum (fig. 7.2). The supraoccipital bears a vertical median ridge. The parietal invades the occiput but makes only a small contribution. The exoccipital portion of the co-ossified braincase is large, with a long, high, laterally directed, hollow paroccipital process. A large caudal tympanic recess opens into the rostral surface of the process. The ventrolateral expansion of the braincase has enclosed the passages of what may be auditory tubes and the internal carotid artery and greatly occludes the external acoustic meatus. Laterally the expansion encloses the entire otic region, creating a space similar to, but more extensive than, the lateral depression of troodontids.
FIGURE 7.2. Braincase of Erlikosaurus andrewsi: A, medial view of right side; B, occipital view. Scale = 5 cm. (From Clark et al. 1994.)
The lower jaw is generally shallow, and the rostral region is curved downward in all taxa in which it is known to be present. Downward bending is stronger in Segnosaurus, Beipiaosaurus, and Neimongosaurus than in Erlikosaurus, and in Eshanosaurus the symphysial region is more strongly expanded mediolaterally. The dentary is shallow for most of its length but deepens caudally to form a coronoid prominence, slight in Erlikosaurus and Eshanosaurus but stronger in Segnosaurus, Alxasaurus, and Beipiaosaurus. The most rostral portion of the dentary is toothless in Erlikosaurus, the tooth row being bordered laterally by a shelf starting at the fifth tooth position and extending to the end of the dentary. This shelf is also visible in Segnosaurus, where it starts at the fourteenth tooth position and continues backward for only half of the dentary. A shelf is also present in Neimongosaurus. The dentary of Segnosaurus is toothless rostrally, but the tooth row begins farther rostrally than in Erlikosaurus. The dentary of Alxasaurus, Beipiaosaurus, and Eshanosaurus has a similar shelf, but the teeth begin farther mesially than in therizinosaurids. In Alxasaurus and Beipiaosaurus the tooth row continues rostrally to the end of the mandible, but in Eshanosaurus the rostral-end part of the dentary is missing. In Eshanosaurus a round fenestra within the caudal part of the dentary may be an artifact.
The postdentary bones are known only in Erlikosaurus and Segnosaurus. The large surangular overlaps the caudal aspect of the dentary. There is no coronoid bone, and the retroarticular process is short and dorsally concave. The prearticular arches rostrodorsally along the caudal edge of the splenial. The splenial is triangular in medial view with a tall caudal end and is not visible on the lateral surface of the mandible. The mandibular groove is broadly exposed rostral to the splenial. A foramen aerum is present in the middle of the dorsal surface of the articular. The jaw articulation is set below the line of the tooth row. The articulation surface is oriented oblique to the long axis of the mandible, with its medial edge rostral to its lateral. The large, elongate external mandibular foramen is laterally bordered by the dentary, the surangular, and the angular.
The teeth of therizinosauroids are unusually small, although larger than in some troodontids, alvarezsaurids, and ornithomimosaurs. The 24 maxillary teeth are more or less uniform in shape in Erlikosaurus but diminish slightly in size from mesial to distal. Maxillary teeth are straight and narrow, pointed, and slightly flattened transversely. With the exception of the five mesial, the teeth are closely spaced. The mesial and distal edges of each crown are serrated; the serrations are coarse and inclined toward the edges. The roots are cylindrical, and the crown is constricted at its base. Interdental plates are visible lingually between distal teeth.
FIGURE 7.3. Dentaries: A–C, left dentary of Eshanosaurus deguchiianus in A, medial, B, dorsal, and C, lateral views; D, E, right dentary of Alxasaurus elesitaiensis in D, medial, and E, lateral views. Scale = 1 cm (A–C), 2 cm (D, E). (A–C from Xu et al. 2001; D, E, from Russell and Dong 1993a.)
There are 31 dentary teeth in Erlikosaurus, most of which are similar in shape to those of the maxilla. The first 5 teeth are straight, markedly larger than the others, loosely spaced, and separated by interdental plates lingually. The distal teeth are closely packed and diminish in size toward the rear. They are also more mesiodistally expanded and transversely compressed. The crown is widest at the midpoint of its height. The labial surface of these distal teeth is convex. All teeth have cylindrical roots and are constricted at the base. The mesial and distal edges are coarsely serrated.
The dentary of Segnosaurus bears 24 sharply pointed teeth. The mesial teeth are recurved only along their apical halves. The last few teeth are straight and markedly smaller. The transverse compression of the teeth is greater on the larger, more mesial teeth. The labial surfaces are convex, whereas the lingual aspects are concave. The serrations of the mandibular crowns in Segnosaurus are more distinct than those in Erlikosaurus, and the serrations are slightly larger on the mesial edge in most teeth. The teeth in Segnosaurus are placed obliquely, so that the serrated mesial and distal edges are slightly lingual and labial, respectively. The entire dentition, with the exception of the first two teeth, is closely spaced. Wear surfaces are not observed on the teeth in Erlikosaurus, whereas they are weakly developed on the distal edges of some dentary teeth in S. galbinensis.
The dentary of Alxasaurus has approximately 40 tooth positions (fig. 7.3), and that of Beipiaosaurus has at least 37. The teeth are generally similar in shape to those of Erlikosaurus except that the serrations are larger. A replacement tooth in Beipiaosaurus is preserved within an oval resorption pit entirely within the root. The dentary teeth of Neimongosaurus are nearly symmetrical.
The dentary of Eshanosaurus (fig. 7.3) preserves positions for 32 teeth, and although it is incomplete rostrally, it could not have had more than 37 teeth. Unlike in all other therizinosauroids, the dentary teeth of Eshanosaurus have fine serrations similar to those of most other theropods except troodontids (Farlow et al. 1991). Unlike in ornithischian and sauropodomorph dinosaurs, these serrations are perpendicular to the edge of the tooth rather than deflected distally.
Postcranial Skeleton
AXIAL SKELETON
The vertebral column is known only from fragmentary or crushed material except in Nanshiungosaurus, Erliansaurus, and Neimongosaurus. The cervical vertebrae in Erlikosaurus, Segnosaurus, Nanshiungosaurus, Alxasaurus, Beipiaosaurus, Neimongosaurus, Erliansaurus, and Nothronychus are large and elongate and bear low neural spines and pleurocoels. The cervical ribs are fused to the vertebrae in the first three taxa but not in Alxasaurus, but as in several other coelurosaurs this may be an ontogenetic feature. The centra, where known, are amphiplatyan or shallowly amphicoelous except in Nanshiungosaurus brevispinus, in which the cranial cervicals are procoelous and the caudal cervicals are opisthocoelous. The centra of the dorsals, known only in Nanshiungosaurus, Alxasaurus, Beipiaosaurus, Neimongosaurus, Erliansaurus, and Nothronychus, have lateral depressions that may have been pneumatic, and the internal structure of Nanshiungosaurus dorsals also indicates pneumaticity (Makovicky 1995). Alxasaurus and Nanshiungosaurus brevispinus have five sacral vertebrae, and Segnosaurus and Neimongosaurus have six (fig. 7.6). Neural spines are low in the dorsal and sacral vertebrae, and the sacral spines do not extend above the top of the ilium. Ossified interspinous ligaments are present between the sacral vertebrae in Segnosaurus and Nanshiungosaurus but not in Alxasaurus. The tail is short in the only taxon in which it is nearly complete, Neimongosaurus, which has 23 caudals. The proximal caudal centra in Alxasaurus are significantly longer than the distal ones, changing after the fifth caudal, but in Neimongosaurus the change in size is more gradual. The neural spines and transverse processes are confined to the 13 proximal caudals in Alxasaurus.
APPENDICULAR SKELETON
Part of the pectoral girdle is preserved in Alxasaurus, Beipiaosaurus, Therizinosaurus, Neimongosaurus, Segnosaurus, Erliansaurus, and Nothronychus. The scapula and coracoid are fused in all but Beipiaosaurus and Nothronychus (unknown in Erliansaurus). The scapular blade is straight and narrow and expands only slightly distally, and it bears a well-developed, dorsally directed acromial process. The coracoid, where known, is subrectangular with a small coracoid foramen and a well-developed acromial process. The glenoid fossa in Segnosaurus, Neimongosaurus, and Nothronychus is oriented laterally, while in Therizinosaurus it faces caudoventrally. The furcula, known only in Beipiaosaurus and Neimongosaurus, is broadly U-shaped, robust, and lacking a hypocleideum.
The humerus is specialized in Erlikosaurus, Therizinosaurus, Erliansaurus, Segnosaurus, and Neimongosaurus but less so in Alxasaurus and Beipiaosaurus (fig. 7.4). In the first two genera it is massive with strongly expanded proximal and distal ends; in Erliansaurus, Segnosaurus, and Neimongosaurus it is similar but less massive. The humeral head is situated at about the middle of the proximal end and overhangs the caudal surface. This overhang is more pronounced in Erlikosaurus than in Segnosaurus. Medial to the humeral head an internal tubercle is well developed. The deltopectoral crest is massive and extends more than a third of the humeral length. A sharply pointed tubercle is present on the caudomedial surface of the humerus at the midlength of the bone. The radial and ulnar condyles, which are small, rounded, and separated by a narrow flexor groove, lie on the cranial edge of the bone rather than the distal. Lateral and medial epicondyles are well developed; the medial epicondyle is larger (fig. 7.4). In Therizinosaurus the humerus has expanded proximal and distal ends, a well-developed caudomedial tubercle, and a longer deltopectoral crest (extending two-thirds the length of the bone). The humerus of Alxasaurus is poorly preserved but more slender than in Segnosaurus, the deltopectoral crest is long but unexpanded, and a tubercle is absent on the caudomedial shaft. The humeri of Beipiaosaurus and Nothronychus are also slender with an unexpanded deltopectoral crest, and in Beipiaosaurus and Erliansaurus the flexor groove between the ulnar and radial condyles is shallower and wider.
FIGURE 7.4. Humerus and manus. A–C, Alxasaurus elesitaiensis: A, reconstruction of right humerus in ventral view; B, left distal carpals in proximal view; C, left metacarpals in dorsal view. D–G, Therizinosaurus cheloniformis: D, right humerus in ventral view; E, right distal carpals in proximal view; F, right metacarpals in dorsal view; G, phalanges of right digit II in medial view. H, Segnosaurus galbinensis, left humerus in ventral view. Scale = 5 cm (A, D–H), 2 cm (B, C). (A–C from Russell and Dong 1993a; D–G from Barsbold 1976c; H from Barsbold and Perle 1980.)
The antebrachium is known only in Beipiaosaurus, Alxasaurus, Therizinosaurus, Segnosaurus, Erliansaurus, and Neimongosaurus. The radius and ulna are stout in Segnosaurus and more slender in the other four taxa. The ulna bears a well-developed olecranon process. In Neimongosaurus the radius has a well-developed biceps tubercle.
The carpus and manus are preserved in Alxasaurus, Beipiaosaurus, Erliansaurus, and Therizinosaurus. Three small proximal carpals are preserved in Alxasaurus and two in Beipiaosaurus. All three taxa have an unusual distal row of carpals comprising two large elements capping metacarpals I and II. In Alxasaurus and Therizinosaurus (fig. 7.4) the medial distal carpal is the larger of the two, and proximally it bears a well-developed trochlear surface with two ridges. In Therizinosaurus these carpals are partially fused to metacarpals I and II. The more lateral element is smaller and bears a less pronounced trochlear surface. Surprisingly, in Beipiaosaurus the position of the two bones on both sides of the partially disarticulated specimen is the reverse of that on the previous two taxa, and the larger element is preserved lateral to the smaller one (fig. 7.5).
FIGURE 7.5. Beipiaosaurus inexpectus: A, left manus in lateral view; B, right carpus and metacarpals; C, right ankle and pes. Scale = 10 cm (A), 2 cm (B, C). (From Xu et al. 1999a. Reprinted by permission of Nature [399:350–354], copyright 1999, Macmillan Publishers Ltd.)
The manus comprises three digits. The first metacarpal is approximately half the length of the second and is similar in width, whereas the third metacarpal is more slender and intermediate in length between metacarpals I and II. Metacarpal I is closely appressed to metacarpal II, and its lateral surface expands at the contact. The phalangeal formula is 2-3-4. The pits for the insertion of extensor ligaments on the distal ends of the phalanges are poorly developed in Therizinosaurus and Segnosaurus but present on Alxasaurus and Beipiaosaurus. Well-developed flexor tubercles are present on all phalanges of Segnosaurus, Erliansaurus, and Therizinosaurus, but they are weakly developed on the smaller Beipiaosaurus and Alxasaurus. As with the metacarpals, the phalanges of the third digit are much narrower than those of the first two. The ungual phalanges in most taxa are mediolaterally compressed and strongly curved with a slight lip over the proximal articular surface, although Nothronychus unguals lack a lip. In Therizinosaurus the unguals are weakly curved, greatly elongate (some more than 60 cm long), and massive.
The pelvis is more or less completely preserved in Segnosaurus, Nanshiungosaurus brevispinus, and Enigmosaurus and partially preserved in Alxasaurus, Beipiaosaurus, Neimongosaurus, Erliansaurus, and Nothronychus. In the taxa with well-preserved pelves it is opisthopubic, with widely separated ilia fused to the sacral diapophyses (fig. 7.6). The long preacetabular process of the ilium has a tall, pointed cranioventral end that is flared at right angles to the sagittal plane (not preserved in Beipiaosaurus). The postacetabular process is relatively long in Alxasaurus, Beipiaosaurus, and Neimongosaurus and short in Erliansaurus, Segnosaurus, Nanshiungosaurus, and Enigmosaurus, and in the latter three taxa it bears a strong knoblike lateral protuberance. The pubic peduncle is long; the ischial peduncle is much shorter and in Erliansaurus bears an antitrochanter. The shaft of the pubis is long and directed caudoventrally, with a distal apron and foot. The ischium is shorter than the pubis and bears a prominent obturator process that approaches or contacts the pubis. The ischium and pubis are flattened in Alxasaurus and Segnosaurus, and the ischium is flattened in Nothronychus, whereas these bones are more cylindrical in Beipiaosaurus, Nanshiungosaurus brevispinus, and Enigmosaurus.
The femur and crus are known in Segnosaurus, Alxasaurus, Beipiaosaurus, Erliansaurus, and Neimongosaurus, and only the crus is known in Nothronychus. The femur is weakly sigmoid except in Beipiaosaurus, and the femoral shaft is oval in cross section. The femoral head is turned in at an almost right angle and separated from the proximal shaft by a distinct neck. The greater trochanter is well developed, expanded craniocaudally, and separated from the cranial trochanter by a deep cleft. The fourth trochanter, in the form of a rugose crest, is placed above the middle of the femoral shaft but is weak or absent in Neimongosaurus and Erliansaurus. The distal condyles are well separated; the lateral condyle is larger than the medial. The extensor and flexor grooves are shallow.
The length of the tibia is more than 80% of the length of the femur in Segnosaurus and nearly equal in size in Beipiaosaurus. The tibia of Segnosaurus bears a well-developed cnemial crest, has an expanded distal end, and supports the distal end of the fibula (fig. 7.7). A fibular crest is present in Beipiaosaurus and Segnosaurus, and in Neimongosaurus it extends more than half the length of the tibia. Closely applied to the tibia for most of its length, the fibula is broad proximally but slender distally and contacts the calcaneum. In Beipiaosaurus, Segnosaurus, Erliansaurus, and Nothronychus, at least, the medial surface of the fibula is flat, lacking a pocket.
FIGURE 7.6. Pelves. A, B, Segnosaurus galbinensis: pelvis in A, dorsal, and B, left lateral views. C, Enigmosaurus mongoliensis, left pelvis in lateral view. D, Nanshiungosaurus brevispinus, left pelvis in lateral view. Scale = 10 cm (A), 20 cm (B–D). (A, B, from Perle 1979; C from Barsbold 1983a; D from Dong 1979.)
A tarsus comprising the calcaneum, the astragalus, and two distal tarsals (3 and 4) is preserved in Segnosaurus. The astragalus with reduced condyles covers only the medial part of the distal surface of the tibia and bears a tall, laterally curved and pointed ascending process. The astragalus of Beipiaosaurus and Neimongosaurus also does not cover the lateral part of the distal surface (fig. 7.5). The calcaneum of Segnosaurus is suboval, wider than high, and fits into a prominent socket on the astragalus. The distal tarsals are compressed, and the lateral element covers the proximal end of metatarsal IV, whereas the smaller medial element covers a portion of metatarsal III.
The pes of Erlikosaurus, Neimongosaurus, and Segnosaurus is short and broad; that of Beipiaosaurus is longer, but it is shorter than in other theropods. The pes of Alxasaurus is only partially preserved. The completely preserved metatarsus of Segnosaurus is composed of five elements, digit V lacking phalanges. Metatarsals I–IV are massive with widened proximal and distal extremities. Metatarsal I is the shortest and divergent from the rest; metatarsal III is the longest and proximally contacts II and IV only briefly. Metatarsal V is short and broad, less than one-third the length of metatarsal IV. The length of the metatarsus is approximately 30% of the femoral length.
In Erlikosaurus and Neimongosaurus the first metatarsal is not as expanded proximally as it is in Segnosaurus but has a proximal contact surface for the tarsus, and its lateral surface is flattened where it meets metatarsal II. In Neimongosaurus the proximal end of metatarsal III interlocks with those of II and IV, unlike the expanded proximal end of metatarsal III in Segnosaurus. In Beipiaosaurus the proximal end of metatarsal I is mediolaterally compressed and apparently does not reach the tarsus (fig. 7.5), similar to most other theropods. The proximal ends of metatarsals III and IV are more compressed mediolaterally than in Neimongosaurus.
The pedal phalangeal formula in Erlikosaurus is 2-3-4-5-0 (fig. 7.7). In this taxon the first digit is shortest, whereas digits II and III are the longest and subequal in length. The fourth digit is slightly shorter and the most slender of all. The phalanges of digits I–IV are short and robust, and in Neimongosaurus the proximal phalanges have a heel proximally. The third phalanx of digit IV is especially short in Erlikosaurus but not in Neimongosaurus. Pedal claws are large, recurved, and strongly compressed transversely in Erlikosaurus but less compressed in Segnosaurus, Alxasaurus, and Nothronychus.
Integument
Several patches of simple, featherlike integumentary structures are known for Beipiaosaurus inexpectus (Xu et al. 1999a), from the deposits of the Yixian Formation at Sihetun, which is famous for soft-tissue preservation. They are preserved associated with the forelimb and pectoral girdle (fig. 7.5). They mainly comprise closely packed, parallel, filaments up to 70 mm long. Those associated with the ulna are preserved extending away from the lateral edge of the bone. Some of the fibers are lighter in the middle, suggesting a hollow core. The filaments are generally simple, but some branch distally.
PFIGURE 7.7. A, B, Segnosaurus galbinensis: A, right metatarsals in dorsal view; B, right tibia, fibula, astragalus, and calcaneum in cranial view. C, Erlikosaurus andrewsi, right partial pes. Scale = 10 cm (A, B), 5 cm (C). (A, B, from Perle 1979; C from Barsbold and Perle 1980.)
Eggs
Eggs with embryos from the Upper Cretaceous Nanchao Formation, Henan, China, have been identified tentatively as therizinosaurid (Currie 1996b; Manning et al. 1997, 2000). The embryos have not been described in detail, but an interesting pattern of tooth replacement has been described in which a slender, elongate tooth is replaced by a symmetrical, denticulate tooth (Manning et al. 2000). The embryos occur within spherical eggs (dendroolithid in the egg parataxonomy of Mikhailov et al. 1994). Dendroolithid eggs are common in some Upper Cretaceous deposits of Asia (Mikhailov et al. 1994) and have been found at Bayan Shiree (Ariunchimeg 1997; Watabe et al. 1997), the site of several therizinosaurid specimens. The outer surface of the skull illustrated by Manning et al. (1997) is generally similar to that of Erlikosaurus. Carpenter (1999:209) noted premaxillary teeth, which would be unique for a therizinosauroid, albeit expected in a primitive member, but Manning et al. (2000) describe the premaxilla as edentulous. A more definitive identification of this taxon awaits the detailed description of these specimens, which has been hampered by questions about the legality of their export from China. A second kind of egg—gigantic (as long as 50 cm), elongate eggs termed “elongatoolithid”—has been identified as belonging to therizinosauroids (Currie 1996b), and a reconstruction of a therizinosauroid embryo inside such an egg has been widely reprinted, but later reports have not confirmed this identification (Manning et al. 1997; Carpenter 1999).
Systematics and Evolution
As presently known, definitive members of Therizinosauroidea include ten species distributed in nine genera. An eleventh species in a monospecific genus is tentatively referred to the group. However, one Mongolian taxon, Enigmosaurus, does not share any elements with Erlikosaurus from the same formation that would allow differentiation, and the elements with diagnostic features of Nothronychus are unknown in some other named taxa. Thus, at present only eight valid species that are definitive members of Therizinosauroidea are unquestionably valid, along with one species tentatively referred to the group. Furthermore, the monophyly of the one genus with two species is questionable; it is hoped that the high level of monospecific genera in the group will decrease when supraspecific groups become better understood and some genera are synonymized.
Therizinosauroid discoveries, and the naming of taxa, can be divided roughly into three historical groups: (1) discoveries of enigmatic, fragmentary material by the first American and Soviet expeditions to the dinosaur deposits of Asia in the 1920s (Andrews 1932) and 1950s (Rozhdestvensky 1960b; Lavas 1993), respectively; (2) discoveries of more substantial specimens by Soviet-Mongolian expeditions in the 1970s (Lavas 1993; Kurochkin and Barsbold 2000), which signaled the presence of an endemic group of peculiar dinosaurs; and (3) discoveries of primitive members of the group in China (Russell and Dong 1993a; Xu et al. 1999a, 2000) since the 1970s, which have helped clarify relationships.
The first therizinosauroid to have a currently valid species name attached to it was Therizinosaurus cheloniformis, although the three unguals were considered by Maleev (1954) to be from a giant turtle. In the same paper Maleev erected a monospecific Therizinosauridae. Its theropod affinities were first suggested by Rozhdestventsky (1970b) and Osmólska and Roniewicz (1970), although the material was insufficient to indicate its relationships within the group. Barsbold (1976a, 1983) introduced Deinonychosauria, embracing the Deinocheiridae (Osmólska and Roniewicz 1970) and Therizinosauridae (with one genus, Therizinosaurus).
The material that established the concept of an endemic group of specialized dinosaurs was discovered by Soviet-Mongolian expeditions in the 1970s mainly at two sites in the eastern Gobi, Khar Hötöl and Bayanshiree. The first to be described was Segnosaurus galbinensis Perle, 1979, known from a dentary and most of its limb and girdle elements, for which Perle erected Segnosauridae and assigned it provisionally to Theropoda. Next, Erlikosaurus andrewsi was named and assigned to this taxon by Perle (in Barsbold and Perle 1980) on the basis of a skull, vertebrae, a humerus, and a foot discovered fortuitously beneath a larger dinosaur quarried from Bayanshiree. It was then described in detail by Perle (1981). In 1980 Barsbold and Perle erected Segnosauria, which contained Perle's Segnosauridae. In 1983 Barsbold and Perle (in Barsbold 1983a) also erected another taxon, Enigmosauridae, for a new monospecific genus, Enigmosaurus, based on a pelvis (described and illustrated in 1980 as segnosaurian indet.). At about the same time, Dong (1979) described two species from China, a supposed sauropod, Nanshiungosaurus brevispinus, and a supposed carnosaur, Chilantaisaurus zheziangensis, but their affinities with therizinosauroids were not recognized until later (Barsbold and Maryanska 1990). C. zheziangensis is here regarded as an indeterminate therizinosauroid.
The Sino-Canadian expeditions of the 1980s (Dong 1993; Grady 1993) collected two skeletons of a new therizinosauroid in the Alxa Desert of China, Alxasaurus elesitaiensis Russell and Dong (1993a). Russell and Dong also erected the monospecific Alxasauridae, and they revised the taxonomy based on their identification of Therizinosaurus as a relative of segnosaurs. A Sino-Japanese expedition in 1992 recovered remains of a new species, Nanshiungosaurus bohlini (Dong and Yu 1997). Remains of a therizinosauroid with integumentary structures quarried from the “feathered dinosaur” quarries in Liaoning were briefly described by Xu et al. (1999a) and named Beipiaosaurus inexpectus, placed within Therizinosauroidea following Russell and Dong (1993a). Xu et al. (2001) named and described a dentary from the Lower Jurassic Lower Lufeng Series of Yunnan, Eshanosaurus deguchiianus (described briefly by Xu et al. 1999a), and tentatively assigned it to Therizinosauroidea.
FIGURE 7.8. A, strict, and B, Adams consensus of 3,815 equally parsimonious cladograms of relationships among Therizinosauroidea resulting from a cladistic analysis of the data, excluding Nothronychus. Tree length = 63, CI = 0.746, RI = 0.754, RCI = 0.562. When Enigmosaurus is deleted from the analysis, the strict consensus resolves Alxasaurus outside the therizinosaurid group, as in the Adams consensus of the complete analysis.
Most recently, three new monospecific genera have been briefly described. Nothronychus mckinleyi was described by Kirkland and Wolfe (2001) from the Moreno Hill Formation of New Mexico, the first definitive record of the group from outside of Asia. This material was initially thought to belong to the ceratopsian Zuniceratops christoferi (Wolfe and Kirkland 1998). Also in 2001, Neimongosaurus yangi was described by Zhang et al. from the Iren Dabasu Formation of Inner Mongolia, the strata from which the first therizinosauroid remains were collected. The two specimens preserve much of the axial skeleton and most long bones. From the same formation, Xu et al. (2003) described Erliansaurus bellamanus based upon a partial postcranial skeleton, including most of a forelimb and hindlimb.
The taxonomic nomenclature of therizinosauroids depends largely on the relationships of Therizinosaurus cheloniformis, which unfortunately remain tentative. The unusual manual unguals first described by Maleev as from a turtle are easily distinguished by their great length and slight curvature, but they offer little information about the affinities of the species. Barsbold (1976c) described a shoulder girdle and forearm (fig. 7.4) from Hermiin Tsav that he assigned to Therizinosaurus on the basis of its similar claws and its occurrence in the same strata as the original material, the Nemegt Svita (later termed the Nemegt Formation in the western literature). He noted its theropod features, such as a manus with three digits and a short first metacarpal, and pointed out similarities and differences between it and the equally enigmatic Deinocheirus, from the same formation. In 1982 Perle described a fragmentary hindlimb (including parts of a femur, a tibia, an astragalus, a calcaneum, metatarsals, and pedal digits) from near the site of the material described by Barsbold (1976c) and referred it to Therizinosaurus mainly on the basis of its locality (since no hindlimb elements of Therizinosaurus are known). Because of its similarities with segnosaurs, especially in having a wide, short foot with four functional toes, Perle included Therizinosaurus in Segnosauria (along with Deinocheirus). However, Barsbold and Maryanska (1990) considered the material described by Perle to belong to a segnosaur but not Therizinosaurus and excluded Therizinosaurus (and Deinocheirus) from the group. Norman (1990a) considered Therizinosaurus to be Theropoda incertae sedis but did not comment on the material described by Perle (1982). Russell and Dong (1993a) pointed out similarities the specimen described by Barsbold (1976c) shares with segnosaurs, especially the large distal carpals and the shallow pits for extensor ligaments on the phalanges, and argued that Therizinosaurus is a member of the same clade.
The postulated relationship of Therizinosaurus to segnosaurs therefore rests primarily on the features of the material described by Barsbold (1976c) that are shared with this group. The scapula and coracoid of Therizinosaurus are generally similar to those of more basal theropods and unlike those of segnosaurs in having the glenoid oriented caudoventrally rather than laterally. The humerus of Therizinosaurus shares with those of Erlikosaurus, Alxasaurus, and Segnosaurus an unusually long deltopectoral crest, although this is not unique to this group and it is more elongate in Therizinosaurus. A distinctive tubercle preserved on the caudomedial side of the humeral shaft in Erlikosaurus, Segnosaurus, and Therizinosaurus (Maryanska 1997) is absent in Alxasaurus, Nothronychus, and other coelurosaurs. In Therizinosaurus, Erlikosaurus, and Segnosaurus the radial and ulnar condyles on the distal end of the humerus are positioned on the cranial, rather than the distal, surface, a feature otherwise found only in birds and the enigmatic alvarezsaurids among theropods. The most compelling evidence is the structure of the carpus and manus, which are nearly identical in Therizinosaurus and Alxasaurus and probably in Beipiaosaurus (unless the distal carpals are truly reversed in the latter). The large, semilunate medial distal carpal is unique in having only a small contact with metacarpal II while bearing a double-ridged trochlear surface proximally. The three-fingered manus has similar proportions among all three genera, and as in segnosaurs, the first metacarpal broadly abuts the second and is expanded at their contact. The ligament pits on the distal ends of the phalanges are poorly developed or absent in both Therizinosaurus and Segnosaurus. The manual unguals of Therizinosaurus are less curved than those of segnosaurs and thus more similar to those of ornithomimids and alvarezsaurids, but there is otherwise little similarity among the unguals of these groups.
There is therefore a small but significant amount of evidence supporting a relationship between Therizinosaurus and segnosaurs, and the features in which they differ (e.g., the length of the deltopectoral crest and the shape of unguals) are generally specializations of one group rather than similarities shared with other groups. The glenoid fossa of Therizinosaurus is similar to that of primitive theropods in facing caudoventrally, unlike that of Segnosaurus, Neimongosaurus, and Nothronychus, which face laterally as in birds (avialans) and dromaeosaurids, but the derived condition may have evolved separately in therizinosauroids as Alxasaurus also has the primitive theropod condition. It is unfortunate that the taxonomy of the group is affected so greatly by the validity of a relationship so tenuous, but unless evidence for a different relationship for Therizinosaurus emerges, the taxonomy should reflect that current evidence favoring a relationship between this genus and segnosaurs.
The definition and diagnosis of Therizinosauroidea given at the beginning of this chapter considers Eshanosaurus to be a member of this clade. If only those features that can be determined in Beipiaosaurus are considered, then the diagnosis is restricted to a sharp, ventrally projecting rim on the premaxilla, elongate vomers that contact rostrally reduced pterygoids, an enlarged pneumatic basicranium, a horizontal rim lateral to the dentary dentition, a gentle rise at the caudal end of the dentary, large serrations on the dentary teeth, straight and only slightly flattened mesial dentary teeth, procoelous cranial cervical vertebrae, and opisthocoelous caudal cervical vertebrae. If Eshanosaurus is indeed a primitive member of the group, then only elongate vomers that contact rostrally reduced pterygoids, an enlarged pneumatic basicranium, a horizontal rim lateral to the dentary dentition, procoelous cranial cervical vertebrae, and opisthocoelous caudal cervical vertebrae can be determined as diagnostic of the group. Russell and Dong (1993a) first erected Therizinosauroidea to include Therizinosauridae and Alxasauridae. In describing Beipiaosaurus, Xu et al. (1999a) included it within Therizinosauroidea, leaving the group comprising Alxasaurus and Therizinosauridae unnamed. Sereno (1999a) proposed a revised, phylogenetic definition of Therizinosauridae that would include Alxasaurus and Beipiaosaurus but based it on Erlikosaurus rather than Therizinosaurus. He defined the group as “all maniraptoriforms closer to Erlikosaurus than to Ornithomimus, Oviraptor, or Neornithes” (Sereno 1999a:2147), which is invalid under the ICZN (and Phylocode, www.ohiou.edu/phylocode) because the family is not based on the genus for which it was named.
Characters varying among therizinosauroids are modified from Xu et al. (1999a), and the results of a cladistic analysis of these data are presented in figure 7.8. Forty characters among ten ingroup taxa were analyzed, but Nothronychus was excluded because its relationships were highly unresolved. Five outgroup taxa were also included, chosen on the basis of several recent analyses (Makovicky and Sues 1998; Sereno 1999a; Xu et al. 1999a), but no attempt was made to determine relationships among them or to determine the sister taxon of Therizinosauroidea. Erliansaurus appeared too late for inclusion in this analysis, and Nothronychus was excluded due to its incompleteness.
Eshanosaurus deguchiianus is placed within Therizinosauroidea only tentatively at present because it shares the following features with therizinosauroids: a lateral shelf on the dentary, dentary teeth becoming larger rostrally, and teeth mediolaterally compressed and symmetrical with cylindrical roots. It is distinguished by the smaller size of the serrations on its teeth and its broader symphysial region. The small serrations on the teeth may be a primitive condition for the group, depending on its relationships. Although the species bears some general similarities to the dentary of prosauropods, there are few detailed similarities (see Xu et al. 2001), and its teeth do not have a vertical ridge similar to that on prosauropod teeth (contra Lamanna, in Kirkland and Wolfe 2001:412). A large, circular fenestra in the caudal part of the dentary is a unique feature but may be an artifact. E. deguchiianus's relationships with other therizinosauroids are unclear, but the analysis presented here indicates that, surprisingly, it might group with therizinosaurids on the basis of the number of dentary teeth. However, Dromaeosauridae is the only outgroup taxon included in this analysis that has a small number of teeth, and if other outgroup taxa with few teeth are invoked (e.g., Ornitholestes, Tyrannosauridae), then a more basal position for Eshanosaurus may be equally parsimonious.
Beipiaosaurus inexpectus as presently known combines a unique suite of features—a longer dentary than other therizinosauroids, the preacetabular and postacetabular parts of the pelvis similar in size, the preacetabular part of the pelvis unexpanded, the ischial shaft cylindrical, the first metatarsal mediolaterally compressed proximally, the proximal part of metatarsals II and III compressed—but those in which it differs from Alxasaurus and Therizinosauridae are primitive for Therizinosauroidea rather than being unequivocal autapomorphies. Its more bulbous teeth may also diagnose the species, but this depends on which group of coelurosaurians is the closest relative of Therizinosauroidea.
Nothronychus mckinleyi is known from fragmentary material, but it preserves the autapomorphic features of denticles on cheek teeth reaching to the base of the crown and manual unguals without a dorsal lip proximally. It has a slender humerus, as in Alxasaurus but not Therizinosauridae. Its relationships within Therizinosauroidea are unclear due to the lack of material.
The group comprising Alxasaurus, Enigmosaurus, and Therizinosauridae currently lacks a name but is diagnosed by several features, including the preacetabular portion of the ilium ventrally expanded and longer than the postacetabular portion and the ischial shaft mediolaterally compressed (except in Enigmosaurus). The dentaries of Alxasaurus and Segnosaurus may be shorter relative to body size than in Beipiaosaurus. The condition of the proximal end of the first metatarsal is unknown in Alxasaurus, so it is unclear whether it is expanded to converge on the primitive dinosaurian condition as in therizinosaurids.
Alxasaurus elesitaiensis as presently known has a unique combination of characters, including those that diagnose the unnamed group but are primitive for Therizinosauridae, but it lacks unequivocal autapomorphies.
The analysis presented here was unable to resolve relationships among all other therizinosauroids, in part due to the ambiguous relationships of Enigmosaurus. Adams consensus (fig. 7.8) shows that a clade excluding Alxasaurus and Enigmosaurus is stable, and when Enigmosaurus is excluded this clade is found in the strict consensus. The instability of Enigmosaurus is due not only to missing data but also to its unflattened pubic shaft (a flattened shaft is present in Alxasaurus and Segnosaurus). This is contradicted by the short postacetabular portion of the ilium, unlike in Alxasaurus but as in Segnosaurus and possibly N. brevispinus.
Enigmosaurus mongoliensis is characterized by a small, shallow obturator process on the ischium and a caudal process on the pubic foot. This condition differs from that in Segnosaurus, in which the distal foot has an ellipsoidal shape and the obturator process is deep and subquadrate. The synonymy of E. mongoliensis with Erlikosaurus andrewsi, from the same stratigraphic beds but at a different locality, cannot be ruled out until elements shared by both taxa are discovered.
In the cladistic analysis presented here, Therizinosauridae—comprising Segnosaurus, Erlikosaurus, Therizinosaurus, N. brevispinus, and Neimongosaurus—is stable in the Adams consensus of the complete data matrix and in the strict consensus when Enigmosaurus is excluded. Thus, the absence of this group from the strict consensus must be due to the uncertain relationships of Enigmosaurus. This group is difficult to diagnose due to the incompleteness of most of the taxa, so that derived characters shared by several taxa cannot be determined in the others. The name Therizinosauridae is here applied to the group defined as the least inclusive clade including Therizinosaurus and Erlikosaurus, although the unstable relationships of Enigmosaurus render it ambiguous for this data set.
Evidence for some groupings of these taxa exists, such as groups comprising at least Therizinosaurus and Segnosaurus, diagnosed by the reduced or absent ligament pits on the distal ends of the manual phalanges; at least Segnosaurus and Erlikosaurus, diagnosed by the smaller number of dentary teeth (no more than 31) and an edentulous symphysial region, also found in Neimongosaurus; Segnosaurus, Erlikosaurus, and Therizinosaurus, diagnosed by the large tubercle on the caudomedial part of the humeral shaft; at least Segnosaurus and Nanshiungosaurus brevispinus, diagnosed by the much shorter postacetabular portion of the ilium with a knoblike process (also found in Enigmosaurus); and at least Segnosaurus and Nanshiungosaurus brevispinus, diagnosed by ossified interspinous ligaments above the sacrum. A group comprising Segnosaurus and Neimongosaurus is resolved on the Adams consensus, but there are no characters shared uniquely by these two taxa.
Therizinosaurus cheloniformis is characterized by a long deltopectoral crest on the humerus (more than half the length of the bone), distal carpals partially fused to metacarpals I and II, and elongate, only slightly curved unguals of the manus.
Segnosaurus galbinensis is characterized by, among other features, mesial mandibular teeth that are markedly flattened and slightly recurved, only 24 dentary teeth, and six sacral vertebrae.
In Erlikosaurus andrewsi the 31 small mandibular teeth are straight and only slightly flattened, the coronoid process is low, and the pedal claws are strongly compressed transversely.
Nanshiungosaurus brevispinus was originally described by Dong (1979) as a sauropod in part due to its large size. N. brevispinus has an opisthopubic pelvis that is similar to that of Segnosaurus except that the cranial portion of the preacetabular process of the ilium is shallower in Nanshiungosaurus than in Segnosaurus and the pubis and ischium have a more extensive contact beneath the acetabulum. The structure of cervical vertebrae in Nanshiungosaurus also is generally similar to the incomplete material of Segnosaurus and Erlikosaurus. Unfortunately, other parts of the skeleton of the Mongolian therizinosaurids are not duplicated in Nanshiungosaurus, and the shape of the preacetabular process of the ilium, the more extensive contact between the pubis and the ischium, and the primitive condition of five sacral vertebrae are the only features differentiating them at present.
Nanshiungosaurus bohlini was described recently (Dong and Yu 1997) as a new species within this genus based on large cervical and thoracic vertebrae (i.e., with a centrum up to 18.4 in length). However, its relationships are unclear since the axial skeleton is not well known in most therizinosauroids. The two species of this genus are similar in being larger than other therizinosauroids and in most aspects of their vertebral anatomy, but they differ significantly in that all of the vertebrae of N. bohlini are amphiplatyan with low, axially elongate spines, whereas in N. brevispinus the cranial cervicals are procoelous, the caudal cervicals are opisthocoelous, and their spines are transversely broad.
Neimongosaurus yangi is characterized by a circular fossa under the transverse process of the proximal caudal vertebrae, a prominent biceps tuberosity on the radius, well-developed heels on the proximal pedal phalanges, a fibular crest more than half the length of the tibia, and caudal vertebrae with widely divergent prezygapophyses (Zhang et al. 2001). It shares with therizinosaurids such as Segnosaurus the twisting of the preacetabular portion of the ilium to face dorsally and a laterally facing glenoid fossa on the scapulacoracoid.
Erliansaurus bellamanus is characterized by an enlarged nutrient foramen on the anterior caudal vertebrae, a crestlike posterior trochanter on the humerus bordered medially by an oval depression, a rugose swelling dorsal to the ischial peduncle on the lateral surface of the ilium, the posterior margin of the proximal end of the fibula significantly higher than the anterior margin, and a hypertrophied anterior trochanter located distally on the fibula (Xu et al. 2003). Xu et al. (2003) also inferred that it and Neimongosaurus are closer to Therizinosauridae than is Alxasaurus but that it falls outside the family.
Stratigraphy, Paleoecology, and Biogeography
The oldest taxon tentatively identified as a therizinosauroid is Eshanosaurus, from the lower part of the Lower Jurassic Lower Lufeng Series of China (Xu et al. 2001). The age of the Lower Lufeng Series is poorly constrained, and age estimates are based solely on the vertebrate fauna. Two vertebrate assemblages are recognized within the Lower Lufeng Series, and the stratigraphically lower of the two has been tentatively correlated with the Hettangian stage (Luo and Wu 1994).
The oldest definitive therizinosauroids are from the Early Cretaceous of China. Beipiaosaurus is from the shales from which “feathered dinosaurs” have been collected in the lower part of the Yixian Formation, in Liaoning. The age of these beds has been disputed as either Late Jurassic or Early Cretaceous, but the most recent Ar/Ar radiometric dates from volcanics interbedded with the fossil-bearing horizon indicate an age correlative with the Albian marine invertebrate stage (Swisher et al. 1999). The age of the Bayin-Gobi Formation in the Alxa Desert, from which Alxasaurus derives, is poorly constrained, but it is tentatively dated as Albian by Russell and Dong (1993a) based mainly on the vertebrate fauna. The deposits in Gansu from which Nanshiungosaurus bohlini was collected also appear to be correlative with the Albian (You, pers. comm.).
The stratigraphic nomenclature of the Mongolian localities from which most of the Late Cretaceous therizinosauroids derive is based on the Russian svita concept rather than formations. Formational names are now used (Jerzykiewicz and Russell 1991), albeit usually informally. Here we follow the terminology of Benton (2000).
The age of the Bayanshiree Svita, from which Segnosaurus, Erlikosaurus, and Enigmosaurus were collected, is poorly constrained biostratigraphically (Martinsson 1975; Jerzykiewicz and Russell 1991). Age estimates based on the vertebrate and molluscan fauna are from Cenomanian to early Campanian, but paleomagnetic evidence (Hicks et al. 1999) suggests that it is no younger than Santonian, and radiometric dates place some deposits as no younger than Cenomanian (Shuvalov 2000). The dinosaur-bearing Iren Dabasu Formation of Inner Mongolia, from which the first therizinosaurids (Mader and Bradley 1989) and Neimongosaurus (Zhang et al. 2001) and Erliansaurus were collected, is similar in age to the Bayanshiree Svita on the basis of the vertebrate fauna. The age of the Nemegt Svita, or Nemegt Formation, from which Therizinosaurus is known, is also poorly constrained but is probably late Campanian or early Maastrichtian (Jerzykiewicz and Russell 1991).
The depositional regimes of the sediments in which therizinosauroids have been found range from fluvial sandstones and conglomerates to lacustrine shales, but they are not known from sediments indicative of more arid paleoenvironments. Eshanosaurus was collected from a marl (Zhao and Xu 1998; Xu, pers. comm.), an unusual lithology for vertebrate fossils in the Lower Lufeng Series. Alxasaurus and Beipiaosaurus both are from thin-bedded lacustrine sediments (Russell and Dong 1993a; Xu, et al. 1999a), the latter along with a diverse fauna of aquatic invertebrates and vertebrates and rare terrestrial vertebrates such as psittacosaurids. The remains of Segnosaurus and Erlikosaurus were found in poorly cemented, gray sands with intraformational conglomerates, gravels, and gray claystones, interpreted by Tsybin and Kurzanov (1979) as deposits in small lakes but by Hicks et al. (1999) as a meandering fluvial system. In the same sediments, skeletons of the theropods Alectrosaurus olseni and Garudimimus brevipes, the ankylosaur Talarurus plicatospineus, and abundant hadrosaurs and turtles were discovered. The bones of Enigmosaurus mongoliensis are from red beds, which also yielded bones of sauropods, theropods, and stems of Taxodiaceae in situ (Martinsson 1975). Nanshiungosaurus brevispinus comes from red clays and sandstones in which dinosaur bones, footprints, and eggs are abundant (Dong 1987). Therizinosaur remains are also known from sandstones and sandy clays of the Iren Dabasu Formation in northern China (Mader and Bradley 1989; Zhang et al. 2001), which in addition to an abundant dinosaur fauna also preserve sharks, rays, and plesiosaurs indicative of an estuarine environment (Currie and Eberth 1993). The Nemegt Svita deposits from which Therizinosaurus specimens have been collected are interpreted as representing a broad, meandering alluvial system (Gradzínski 1970).
Therizinosauroids currently are not known from the Baruungoyot and Djadokhta formations of Mongolia and China, although they are present in formations above and below these. The Djadokhta and Baruungoyot formations include eolian sandstones indicative of drier habitats than, for example, the overlying Nemegt Formation (Gradzínski and Jerzykiewicz 1974; Jerzykiewicz et al. 1993), and the absence of therizinosauroids from the eolian beds suggests that they did not frequent the more arid environments present in the Late Cretaceous of the Gobi region.
The unusual anatomy of therizinosauroids prompted speculation that their behavior may likewise have been unusual. Rozhdestvensky (1970b) speculated that the large unguals of Therizinosaurus were used for opening termite mounds or gathering fruit from trees. Barsbold and Perle (1980) suggested that therizinosauroids were amphibious carnivores, preying on fish. In contrast, Paul (1984a) suggested that these dinosaurs had a herbivorous diet, consistent with his hypothesis of their phylogenetic relationships with ornithischians. Russell and Russell (1993; see also Russell 1997) noted the similarity of therizinosauroids’ body proportions and pelvis to those of chalicotheres, unusual clawed ungulates, and giant ground sloths. They suggested that therizinosauroids may have had a lifestyle similar to that reconstructed for the herbivorous chalicotheres and giant ground sloths, in which the animal presumably sat on the ground and used the forelimbs for gathering food. Nessov (1995) suggested arboreality and a diet of wasp nests and presented a reconstruction of a segnosaur hanging from a tree limb in the manner of a sloth.
Although therizinosauroids’ dentition and peculiar postcranial anatomy suggest a feeding strategy different from that of most nonavian theropod dinosaurs, at present there is no strong evidence, such as stomach contents or trace fossils, for their diet or for other aspects of their behavior. The definitive identification of dendroolithid eggs as therizinosauroid would reveal something of their nesting biology, but that conclusion must await the full description and identification of the embryonic material.
The geographic distribution of therizinosauroids extends through much of eastern Asia, from Liaoning in northeastern China westward through the Gobi Desert to Gansu and the Ordos Basin and southward to Zhejiang and perhaps Yunnan, China. Fragmentary therizinosauroid remains are also reported farther west, in Uzbekistan and Kazakhstan (Nessov 1995). Outside of Asia the only definitive occurrence at present is from New Mexico (Kirkland and Wolfe 2001). This distribution tentatively suggests an Asian origin for the group with dispersal to North America, but the relationships of the single North American form, Nothronychus, are poorly understood due to the incompleteness of the material. Kirkland and Wolfe (2001) suggested that Nothronychus is more closely related to therizinosaurids than is Alxasaurus on the basis of the lateral orientation of its glenoid fossa, but when it is included in the analysis here its relationships are completely unresolved.