Holotype

EME PV. 313. Fragmentary skeleton found partly articulated and belonging to a single individual, including: 8 dorsal vertebrae (7 in articulated series); rib fragments; sacrum and associated caudosacral vertebrae; 4 proximal-middle caudal vertebrae; both left and right scapulocoracoids, humeri, ulnae, and radii; complete right carpometacarpus as well as left metacarpals I and II; complete right manus and complete digits I and II of left manus; articulated pelvis with acetabular region of both ilia together with fragmentary left and right pubes and ischia; left tibiotarsus and fibula, tarsometatarsus, and complete foot in articulation; and almost complete right tarsometatarsus. Associated sternal plates were observed and mapped in the field, but were recovered in a fragmentary state. Neurocentral sutures are closed in all dorsal and caudal vertebrae, suggesting that the holotype individual was mature or nearing skeletal maturity. This is corroborated by histological evidence: the presence of dense Haversian bone and substantially diminished spacing between lines of arrested growth (LAGs) near the periosteal surface suggests that this individual was somatically mature (see below for full explanation of histology), and no less than seven years of age (based on LAGs). A skeletal reconstruction of the holotype is presented in figure 1 and measurements of individual elements are presented in tables 1–2.

TABLE 1.

Measurements of dorsal and caudal vertebrae (in mm) of the holotype of Balaur bondoc (EME PV. 313)

TABLE 2.

Measurements of appendicular bones (in mm) of the holotype of Balaur bondoc (EME PV. 313) Asterisk indicates incomplete measurement due to damage.

TABLE 2.

Continued

TABLE 2.

Continued

TABLE 2.

Continued

Type Locality

The holotype was collected by M.V. (September 2009) from red floodplain mudstones of the lower-middle part of the Late Cretaceous (late early Maastrichtian) Sebeş Formation, exposed at the Sebeş-Glod locality (figs. 2–3). The Sebeş-Glod locality (abbreviated as SbG/A-D) encapsulates a suite of restricted outcrops located 1.5 to 3.0 km north of Sebeş city (SW Transylvanian Basin, Alba county, Romania), downstream along the Sebeş River (fig. 2) (Codrea et al., 2010; Csiki et al., 2010a). These outcrops expose a roughly 50 m thick profile from the mid-lower section of the Sebeş Formation (SBF), 300 to 350 m below the major middle-Miocene (Badenian) unconformity that caps the Sebeş Formation at its type section (Râpa Roşie/Red Cliff), and approximately 100 m above the top of the conformably underlying marine Bozeş Formation (late Santonian–lowermost Maastrichtian), which is well exposed at the nearby Petreşti-Arini (PT) section (Vremir, 2010). The Sebeş Formation succession is dominated by coarse, mainly cross-bedded channel fills (gravels, sandy gravels, cross-laminated sandstones) with occasional interbeds of finer grained red or brownish-red overbank and floodplain associations (fine laminated sandstones, silty claystones, massive mudstones), all of which was formed by a high-sinuosity fluvial system.

Fig. 2. 

Photographs of the type locality of Balaur bondoc, where the holotype specimen (EME PV.313) was discovered by MV in 2009 in Upper Cretaceous (Maastrichtian) red floodplain mudstones of the Sebeş Formation. This site, known as the Sebeş-Glod (SbG/A) locality, is located approximately 2.5 km north of Sebeş town, downstream and along the Sebeş River, in Alba County, Romania.

Fig. 3. 

Photographs of the holotype of Balaur bondoc (EME PV.313) as exposed in the field. Photograph of right forearm and pectoral girdle (A) and caudosacral region (B), with scale bars as indicated in the images. Abbreviations: CMc, carpometacarpus; co, ribs; Hu, humerus; Mc, metacarpal; Ph, phalanx; Ra, radius; Sc, scapulocoracoid; U, ulna. Numbers of individual phalanges denoted as phalanx number/digit number.

The SbG/B site is a small waterlogged riverbed outcrop, exhibiting a 6 m thick profile that dips slightly eastward (figs. 2, 4). The basal channel association is built up by fining upward, medium-to-coarse, occasionally pebbly, light pinkish-gray sandstones, grouped in progressively thinner trough cross-laminated sets (dunes) that show internal truncation (St/Sp), characterizing a lateral accretional facies (LA). The progressive abandonement and lateral migration of the channel is marked by fine, horizontally laminated, sand-silt-clay deposits, with occasional scour-fill features and flute marks, indicating minor crevasses (Sh/Fl), grading into a reddish sandy-silty claystone and mudstone dominated proximal floodplain association (OF). The fossiliferous layers are grouped into a 2.2 m thick dark-red, calcareous, silty-claystone sequence, which exhibits thinly laminated and lenticular sandy interbeddings (crevasses) likely related to periodic flooding events, and several pedogenic calcitic horizons containing nodules. Bone fragments are common troughout the whole overbank sequence, but more complete elements and associated and/or articulated skeletal parts (such as the Balaur holotype: fig. 3) are found immediately above the sandy crevasses, covered by fine sediments. The top of the overbank facies is represented by massive dark-red mudstone, which exhibits abundant tubular burrowings and rhysolites, marking the incipient topsoil horizon. A conglomerate channel fill complex, which is several meters thick and comprised of poorly sorted massive, plannar, or concoid cross-laminated multistoried conglomerates (Gmm, Gp/Gh), covers the whole sequence, bounded by a fifth-order flat lower surface. In the basal lag deposits, large cobble and boulder size, red claystone-mudstone intraformational subangular clasts are common, originating from the ancient undercut riverbank.

Fig. 4. 

Stratigraphic log of the holotype locality of Balaur bondoc (the Sebeş-Glod locality) (A) and quarry map showing how the individual elements of the holotype of (EME PV.313) were discovered in the field (B). Abbreviations for A are given in the text. The dromaeosaurid silhouette denotes the level where the Balaur holotype was recovered. Abbreviations for B: adv, anterior dorsal vertebra; c 1–9, caudal vertebrae; cmc, carpometacarpus; d, digit; dv, dorsal vertebrae; fi, fibula; hum, humerus; il, ilium; is, ischium; man ph, manual phalanx; mt, metatarsus; ph, phalanx; pub, pubis; r, rib; rad, radius; sac, sacrum; sc, scapulocoracoid; st, sternum; ti, tibia; ul, ulna; uph, ungual phalanx. The designations (l) and (r) denote left and right elements, respectively.

The multistory complex CH fills, dominated upward by truncated cross-bedded sets (Gh), may indicate mobile, broad, and shallow channels more common in braided fluvial settings. An unstable and dynamic environment, with frequent reactivations and flow-direction changes, is often present in these settings. Since its discovery in 1999 (Vremir, 2010), the SbG/B site has yielded a large number of vertebrate fossils, which originate entirely from the red overbank deposits. Most specimens are fragmentary isolated bones; complete bones, and especially closely associated and partially articulated skeletons, are comparatively rare. The vertebrate record from this site is comprised of various ornithopod (Zalmoxes, Telmatosaurus), sauropod (titanosaurs) and theropod (Balaur) dinosaurs, as well as abundant cryptodiran (Kallokibotion) and rare pleurodiran (“Muehlbachia”) turtles, small crocodylomorphs, rare pterosaurs (medium-sized azhdarchids), and birds.

The sedimentological and taphonomic evidence suggests a general attritional and trampled taphofacies of largely isolated and evenly distributed bone fragments. Many preserved bones suffered long-term subaerial biodegradation, disarticulation, effects of scavenging (possibly from crocodylomorphs), and prefossilization weathering (weathering stage 3–5), including occasional insect-related surficial modifications. Most insect marks may be assigned to coleopterans (Dermestidae?) and isopterans (termites), pointing to long-term subaerial exposure of bones, a low water table, a sparsely vegetated environment, and relatively dry conditions. Episodic flood-related fine sediment input then covered the drifted and partially degraded carcasses. As indicated by the bone map (fig. 4B), the partially decayed Balaur carcass (probably still bearing some soft tissues) was probably transported in a low-energy current during a waning flood, then subaerially exposed (weathering stage 1–2), partially disarticulated, and finally buried due to additional sediment input. As suggested by the quarry map, lithological profile, and sedimentological analysis, more extensive excavations at the site may lead to the recovery of some remaining bones of the Balaur holotype, particularly hind-limb elements, or eventually the neck region and skull. The distal caudal series, however, was destroyed and washed away by recent fluvial erosion.

Diagnosis

Dromaeosaurid theropod with the following autapomorphies (asterisk denotes autapomorphies unique among all theropods, other characters unique among dromaeosaurids): hypertrophied coracoid tubercle*; sinuous ridge on lateral surface of distal humerus extends for 1/3 of the length of the bone*; prominent ridge on medial surface of distal half of humerus*; anterior surface of ulna flattened and bisected by longitudinal ridge*; fused carpometacarpus; reduced, splintlike metacarpal III; metacarpal III contacting metacarpal II distally, buttressed by overhanging ridge on metacarpal II*; distal articular surface not extending onto plantar surfaces of metacarpals I and II; manual ungual II with Y-shaped lateral and medial grooves*; phalanges of manual digit III reduced to small nubbins without condylar articulations; ischial obturator tuberosity expressed as enlarged, thin flange that contacts or nearly contacts pubis ventrally*; tarsometatarsus substantially wider (1.5×) than distal tibiotarsus*; fused metatarsus (metatarsals II–V fused proximally and possibly along parts of their shafts); robust ridges on extensor surfaces of metatarsals II–IV*; metatarsals II and III not ginglymoid; articular region of metatarsals II–III narrower than entire distal end*; first digit of pes with enlarged phalanges but small metatarsal I*; and short, hooklike metatarsal V (emended from Csiki et al., 2010a).

Two pelvic characters that Csiki et al. (2010a) originally considered as autapomorphies of B. bondoc—a laterally everted pubic peduncle and the reorientation of the lateral surface of the pubis to face ventrally—are here reinterpreted as artifacts of crushing. Reexamination and additional preparation of the holotype pelvis indicates that the specimen has been subjected to extensive dorsoventral crushing, which has caused the ilium to break along the margins of the cuppedicus fossa (making it seem as if the pubic peduncle is laterally everted) and the pubis to artificially rotate ventrally. This crushing, which is described in further detail below, fully explains the presence of the two supposed autapomorphies, which we no longer consider valid. Furthermore, a third autapomorphy listed by Csiki et al. (2010a), extremely retroverted pubes and ischia whose long axes are nearly horizontal, is here considered only potentially valid because it may also be explained partially, or entirely, by crushing.

Additional Material

LPB (FGGUB) R.1580, incomplete left humerus; R.1581, left ulna; R.1585, incomplete left metacarpal II; R.1583, proximal fragment of right manual phalanx I-1; R.1582, proximal fragment of left? manual phalanx I-1 or II-1; and R.1584, distal fragment of left manual phalanx II-2, all from the same locality (see below) and possibly from the same individual. Histological analysis suggests that FGGUB R.1581 belongs to a skeletally mature individual, no less than six years of age. Measurements of the referred material is presented in table 3.

TABLE 3.

Measurements of appendicular bones (in mm) of additional referred specimens of Balaur sp. Asterisk indicates incomplete measurement due to damage.

Locality of Additional Material

The Tuştea (also referred to as Oltoane Hill) nesting site, north of the Tuştea locality, Haţeg Basin, Hunedoara County, Romania. The nesting site is located in a 1 m thick, red silty mudstone bed overlying a thin, medium-grained crevasse splay sandstone, and covered by a thick, poorly sorted conglomerate-sandstone unit locally showing tabular or trough cross-bedding, as well as graded bedding. According to the local stratigraphic framework, the nesting site is placed in the unnamed middle subunit (member) of the Densuş-Ciula Formation, a continental detritic-volcanoclastic unit of Maastrichtian age (Bojar et al., 2011); more detailed information on the sedimentology and paleoenvironmental setting of the Tuştea site is given by Bojar et al. (2005), Grigorescu et al. (1994, 2010), and Therrien (2005).

Besides egg clutches and hatchling bones, usually attributed to the basal hadrosaur Telmatosaurus transsylvanicus (see Grigorescu et al., 1994, 2010; but see Weishampel and Jianu, 2011, and discussion therein), a large number of skeletal remains were also recovered from the fossiliferous mudstone, representing several taxa. Although most of these fossils are found disarticulated, several instances of associated (even articulated) remains have also been recorded (e.g., Csiki et al., 2010b, Martin et al., 2010). All of the Balaur referred material was discovered during the same excavation season (1997), and all the bones were spread across a small area roughly along the same stratigraphic level within the fossiliferous bed. No futher postcranial remains referrable to Balaur or close relatives were subsequently discovered at the Tuştea site, despite continuous excavations. Moreover, all referred Balaur material belongs to the forelimb, all elements are commensurate in size, and all show the same peculiar surface texture, a characteristic feature of Balaur (see below). Based on these circumstances, it is probable that these elements belong to the same individual, whose skeleton was scattered and almost completely destroyed after its death and before its final burial in a well-drained floodplain environment undergoing pedogenesis.

Systematics of Additional Material

The referred specimens uniquely share characters with the holotype of B. bondoc that are not seen in any other theropod dinosaurs. First, all bones of the referred specimens possess the characteristic rugose surface texture of the Balaur holotype (this is also present in other isolated theropod bones from Romania). Second, the humerus exhibits a sinuous ridge on the lateral surface of the distal end, which extends for approximately 1/3 of the length of the bone, as well as a prominent ridge on medial surface of distal half. Third, the ulna possesses a flat anterior surface that is bisected by a longitudinal ridge. Fourth, metacarpal II possesses a distal ridge that laterally overhangs metacarpal III, as well as a distal articular surface that does not extend onto the plantar surface (the latter an autapomorphy of Balaur among dromaeosaurids but not among all theropods). Therefore, these specimens can be united with the holotype of B. bondoc based on synapomorphies, and for this reason, Csiki et al. (2010a) considered them to represent a referred specimen of B. bondoc.

Although these specimens share unique derived characters with the B. bondoc holotype, Csiki et al. (2010a) noted a peculiar fact: the referred material is substantially larger than the holotype. The referred humerus is 1.45 times the length of the holotype humerus, whereas the referred ulna is 1.65 times larger. Csiki et al. (2010a) suspected that the holotype of B. bondoc was a mature individual, because of extensive fusion across the skeleton, but were unable to provide a histological estimate of absolute age and growth stage because of extensive fungal damage to the interior regions of the long bones commonly used for these estimates. As part of this monograph we attempted to take further histological samples and were able to obtain samples from the holotype fibula and referred ulna. As detailed below, these samples indicate that both specimens were skeletally mature (i.e., full or nearly fully grown) and of similar numerical age.

It seems unusual that two mature and slow-growing adults of the same species would vary in size to the degree noted between the two specimens of Balaur. It is difficult to test whether this may be the case in other dinosaur species, because sample sizes are usually too low to permit histological examination of large numbers of individuals. Sander and Klein (2005) did find that fully grown individuals of the basal sauropodomorph Plateosaurus may differ in size by up to 50%, but it is worth noting that this determination is based on specimens from a single locality (not multiple localities like the two specimens of Balaur) and hinges on a single outlier specimen that is much larger than a tight cluster of several similarly sized mature individuals (Sander and Klein, 2005: fig. 3B). Furthermore, Plateosaurus is distantly related to dromaeosaurids, is hypothesized to have had a less active metabolism than the birdlike dromaeosaurids, and its great variability in adult body size is hypothesized to have been related to local environmental conditions (Sander and Klein, 2005). In summary, it is not clear whether the case study of Plateosaurus growth and variability is applicable to dromaeosaurids in general or the case of the two Balaur specimens in particular.

Because it seems unusual for two mature specimens to differ so greatly in size, and because there is no clear evidence that adult theropod dinosaurs varied in size to this degree, we take the conservative step and here remove the referred material from the species-level taxon Balaur bondoc. We emphasize, however, that this removal is based on a biological argument (the improbability of a biological species varying in adult size to this degree) and not an anatomical argument (a lack of synapomorphies). Indeed, the two specimens share several unique derived characters and are therefore hypothesized to be close relatives, and surely sister taxa among all currently known theropods. Because these shared characters are autapomorphies of Balaur, we now assign the referred material to Balaur sp., with the prediction that it belongs to a larger species closely related to B. bondoc. Therefore, the size differences between the holotype and referred specimen should not be used as evidence that a single dinosaur species can vary greatly in adult body size or continue growing after extensively fusing its skeleton. Fowler et al. (2011) recently used Balaur as an example of the latter phenomenon, and thus as support for their hypothesis that some tyrannosaurids may also undergo dramatic size changes after fusing portions of their skeleton. We strongly caution against this line of argument.

It is also possible that the extreme adult size difference between the two Balaur specimens is the result of sexual dimorphism, but this proposition is difficult to test because of low sample size and the difficulty of distinguishing male and female dinosaurs. Intriguingly, the recently extinct moa was sexually dimorphic to an extreme degree, with adult females up to 280% larger than adult males (Bunce et al., 2003). Such marked dimorphism cannot be ruled out for Balaur, and it is hoped that this possibility can be tested as more specimens become available. In the meantime, we consider removing the referred specimen from B. bondoc and referring to it as B. sp. to be the most conservative course of action.

Dorsal Vertebrae

Portions of eight dorsal vertebrae are preserved, including an isolated anterior dorsal (fig. 5) and an articulated series of seven mid-posterior dorsals (figs. 6–7). Because their exact position in the dorsal column is uncertain, these vertebrae are here referred to as dorsals A–H, with alphabetical notation corresponding to relative position on the trunk. The first vertebra in the articulated series (dorsal B) is represented by a small fragment, but the six remaining elements are well preserved, especially the two posteriormost vertebrae (dorsals G–H). The neurocentral suture is closed in all dorsal vertebrae in which this region is visible and well preserved.

Fig. 5. 

Dorsal vertebra A of Balaur bondoc (EME PV.313) in right lateral (A) and oblique dorsolateral (B) views. Abbreviations: gap, artificial gap between centrum and neural arch, caused by breakage; ipofos, infrapostzygapophyseal fossa; prez, prezygapophysis, posz, postzygapophysis; sw, dorsal swelling on neural spine. Scale bar equals 1 cm.

Fig. 6. 

Dorsal vertebrae B–H of Balaur bondoc (EME PV.313) in left lateral (A) and right lateral (B) views. Abbreviations: acdl, anterior centrodiapophyseal lamina; idfos, infradiapophyseal fossa; pcdl, posterior centrodiapophyseal lamina; pf, pneumatic foramen. dB–H designates the relative position of the vertebrae in sequence (see text for details). Scale bars equal 1 cm.

Fig. 7. 

Closeup of dorsal vertebrae G–H Balaur bondoc (EME PV.313) in left lateral view. Abbreviations: dia, diapophysis; pa, parapophysis; pf, pneumatic foramen; posz, postzygapophysis; prez, prezygapophysis. Scale bar equals 1 cm. Arrows point to internal pneumatic recesses exposed due to the erosion of the lateral surface of the vertebra.

The single anterior dorsal vertebra (dorsal A) is embedded in matrix and exposed in right lateral view (fig. 5). The right lateral side of the neural spine has been broken from the remainder of the vertebra and displaced dorsally, exposing a large artificial gap and exaggerating the dorsoventral height of the bone (fig. 5: gap). This specimen is the only dorsal vertebra that preserves large portions of the neural spine. The anterior margin of the spine is eroded, but the posterior and dorsal margins are swollen relative to the rest of the spine (fig. 5: sw). Such swelling is indicative of a rugose spine table at the apex of the neural spine, as is also seen in the velociraptorines Deinonychus (Ostrom, 1969), Velociraptor (Norell and Makovicky, 1999), and Saurornitholestes (TMP 67.20.36), as well as the unenlagiines Unenlagia (Novas and Puerta, 1997) and Austroraptor (Novas et al., 2009). Several laminae and fossae are also visible on this specimen. Anterior and posterior centrodiapophyseal laminae are present but mostly eroded ventrally, as is the infradiapophyseal fossa. The infrapostzygapophyseal fossa is deep and widely exposed in lateral view (fig. 5: ipofos), whereas the infraprezygapophyseal fossa is reduced in size, but very deep, and faces almost entirely anteriorly. A prespinal fossa is visible on the small region of the anterior surface of the neural spine that is exposed, bound laterally by the prespinal lamina. A spinoprezygapophyseal lamina connects the neural spine to the prezygapophysis, which is placed at approximately the same horizontal level as the postzygapophysis.

The series of articulated mid-posterior dorsal vertebrae (dorsals B–H) is embedded in matrix (fig. 6). Some vertebrae are split and exposed on a slab and counterslab; the anterior vertebrae are best exposed in right lateral view and the more posterior elements are exposed in left lateral view. The individual centra become progressively shorter anteroposteriorly and deeper dorsoventrally along the dorsal column (table 1). Where well preserved and visible, the anterior and posterior faces are ovoid, taller than wide, with subtly concave articular surfaces. They extend only slightly ventrally relative to the centrum midpoint and are slightly swollen such that they stand out laterally from the smooth and inset lateral surface of the centrum. The lateral and ventral surfaces of the centrum smoothly merge with each other and are not separated by a discrete angle. Anteriorly, the ventral surface is sharp and mediolaterally thin, but it funnels out to become mediolaterally wider as it continues posteriorly. Distinct ridges or grooves are not visible on any of the exposed ventral surfaces.

The line of fusion between the neural arches and centra is visible on most individual dorsals as a thin, slightly raised rim. The parapophysis is well preserved only on the two posteriormost dorsals (dorsals G–H; fig. 7: pa), but eroded surfaces mark its position on the five additional dorsals in the block. Anteriorly, in dorsals C–D, the parapophysis is located at the juncture of the centrum and neural arch, whereas in the posterior dorsals G–H it is located solely on the arch (fig. 7: pa). In these two posterior dorsals the parapophysis is a large structure offset laterally on a pronounced stalk, a characteristic dromaeosaurid feature (e.g., Norell and Makovicky, 1999; Turner et al., 2007a). Here, the lateral articular surface of the parapophysis is large and circular, with a diameter of approximately 4 mm, and deeply excavated as a smooth fossa.

The transverse processes are broken on all dorsal vertebrae, but their preserved bases indicate that they became more horizontally oriented posteriorly along the dorsal column. The anterior and posterior centrodiapophyseal laminae linking the transverse processes to the centrum are visible on most of the mid-posterior dorsals, and are best preserved on dorsal C (fig. 6: acdl, pcdl). Here, the two laminae are approximately equally developed, and the infradiapophyseal fossa between them is deep and triangular (fig. 5: idfos). In more posterior dorsals, it is clear that the posterior lamina becomes progressively thicker and more prominently offset than the anterior lamina, in concert with the dorsal migration of the parapophysis onto the neural arch. Unfortunately, details of the infrapre- and infrapostzygapophyseal fossae are visible only on dorsal C, where these depressions are much smaller than the infradiapophyseal fossa, only marginally visible in lateral view, and face primarily anteriorly and posteriorly, respectively.

The zygapophyses are eroded on most specimens but are well preserved on some mid-posterior vertebrae. The prezygapophysis is most complete on dorsal H, where it has a flat articular surface that expands anteriorly into a spatulate bulb (fig. 7: prez). This surface faces mostly dorsally, but is also oriented somewhat medially and posteriorly. Between the prezygapophysis, parapophysis, and diapophysis of dorsal H is a deep, mostly dorsally and posteriorly facing fossa, which is almost entirely concealed in lateral view. This is the infraprezygapophyseal fossa ( =  hapidocoel of Ostrom, 1969), which is a broader fossa in more anterior dorsals (e.g., dorsal C), but takes the form of a deep notch here because it occupies less space due to the dorsal migration of the parapophysis. The postzygapophysis is best preserved on dorsal G, where it is in near articulation with the prezygapophysis of dorsal H (fig. 7: posz). It is smaller than the adjoining prezygapophysis, has a flat articular surface, and is linked to the base of the transverse process by a sharp lamina. Spinoprezygapophyseal and spinopostzygapophyseal laminae are present at the base of the neural spines where dorsals G and H articulate, and here the lateral surfaces of the spine bases are smooth and slightly concave.

Pneumatic features are visible on all well-preserved dorsals, suggesting that pneumaticity extended throughout the entire dorsal column, unlike in Velociraptor (IGM 100/985, IGM 100/986) but similar to Deinonychus (Ostrom, 1969; see also Benson et al. 2012). Large pneumatic foramina (“pleurocoels”) are clearly visible on the lateral centrum surfaces of dorsals C, E, F, G, and H (figs. 6–7: pf). The size, shape, and position of the foramina are highly variable along the column and between the left and right sides of a single vertebra (dorsal F, the only vertebra exposed on both lateral sides). The left foramen of dorsal F is ovoid, much deeper, and larger (3.5 mm anteroposteriorly by 3 mm dorsoventrally) than the more circular right foramen (1.5 mm by 2 mm). Furthermore, the foramen on dorsal C is large and deep (3 mm anteriorposteriorly by 2 mm dorsoventrally), whereas that on dorsal D is a tiny circle (diameter of 0.75 mm), and some foramina are positioned immediately below the neurocentral suture (e.g., dorsals C, D) whereas others are positioned further ventrally (e.g., dorsal F). Pneumatic features are also present on the neural arch. Above the lamina linking the postzygapophysis and transverse process of dorsal G is a collapsed region that exposes large internal chambers, which would have been covered laterally in life (fig. 7, denoted by arrows). Other broken surfaces, especially on dorsals C, D, and H, also reveal extensive hollow internal cavities. Therefore, it is clear that both the centra and neural arches of all preserved dorsals were considerably pneumatized.

Sacral Vertebrae

A portion of the fused sacrum is preserved in articulation with the pelvis (fig. 8). Four sacral vertebrae are preserved, based on the presence of four clearly observable sacral ribs (fig. 8, denoted by arrows). However, the anteriormost preserved sacral has a broken anterior articular surface, and the rib of this vertebra reaches the broken anterior end of the ilium, making it difficult to determine whether additional sacral vertebrae may have been present anteriorly. As all other dromaeosaurids have at least five sacrals (e.g., Turner et al., 2007a), it is likely that at least some additional anterior sacrals were present in Balaur.

Fig. 8. 

The sacrum of Balaur bondoc (EME PV.313) in three slightly different anteroventral views. The arrows in part A denote the four preserved sacral ribs. Abbreviations: s1, first sacral vertebra. Scale bar in A equals 1 cm.

The length of the sacrum as preserved is approximately 100 mm. Although eroded, the anterior face of the first preserved sacral vertebra is approximately circular, with a diameter of 12 mm. Posteriorly, at the level of the third preserved sacral vertebra, the sacrum expands to 16 mm in mediolateral width. The lateral surface of the first preserved sacral bears a deep, smooth fossa. This depression, which is present on both sides of the vertebra, may be pneumatic in origin, but this remains equivocal since there is no pneumatic foramen (“pleurocoel”) that penetrates into the interior of the centrum. The lateral depressions are not present on more posterior sacral vertebrae, so if the sacrum is pneumatic then there is little external evidence of this morphology. This is not unexpected for a dromaeosaurid, as pneumatic foramina are also absent on the sacral vertebrae of Velociraptor (Norell and Makovicky, 1997), Buitreraptor (Makovicky et al., 2005), Unenlagia (Novas and Puerta, 1997), and Mahakala (Turner et al., 2007a). They are, however, present in an anterior sacral element of Saurornitholestes langstoni (TMP 67.20.36) and in the anterior sacrals of Adasaurus (IGM 100/20).

The ventral surface of the first preserved sacral is rounded and convex anteriorly, whereas those of the second, third, and fourth vertebrae are marked by a longitudinal groove. Longitudinal grooves are also present in most other dromaeosaurids, including Mahakala, Saurornitholestes, and Velociraptor (e.g., Norell and Makovicky, 1999; Turner et al., 2007a). The ventral groove starts on the second sacral in Velociraptor as well, but is present only on the third sacral in Bambiraptor (Burnham, 2004), whereas in Mahakala it is only feebly developed on the posterior half of the second sacral vertebrae to the posterior end of the fourth vertebra (Turner et al., 2011).

Unfortunately, the neural spines of all sacral vertebrae have been eroded, and details of the zygapophyses are not visible due to obscuring matrix.

The right sacral rib corresponding to the first preserved vertebra is visible in ventral view. It is a thin sheet that twists anteroventrally as it extends laterally, such that its anterior surface faces mostly ventrally where it meets the ilium along a broad contact. The rib expands in anterposterior length to form this expansive articulation. The second rib is not well preserved on the right side but is nearly complete on the left side. It extends laterally as a sheet, but unlike the first rib it does not twist anteroventrally where it contacts the ilium. Therefore, its anterior surface faces anteriorly across its entire length. The third rib, which is preserved on both sides, has a similar morphology as the second rib. Its anterior surface, which is well visible on the left side in anterior view, is smooth and concave; the concavity appears to be very deep, but this may be exaggerated by crushing. The third rib contacts the ilium immediately anterior to the level of the acetabulum and expands in anteroposterior length as it continues laterally to meet the ilium. The fourth rib is preserved on both sides and extends laterally, without twisting. It is clearly shallower dorsoventrally than the third rib, and hence is not as greatly expanded where it meets the ilium. The fourth rib is the longest of the sacral series in the mediolateral dimension, and as a result the sacral vertebrae and ilium are widely separated in this region, as in Velociraptor (Norell and Makovicky, 1999: fig. 17) and Microraptor zhaoianus (Hwang et al., 2002). In dorsal and ventral views there are large, ovoid fenestrae between the sacral ribs; these openings are bounded medially by the sacral vertebrae and laterally by the ilium.

There is a series of three articulated and coossified caudosacral vertebrae, which are fused to the sacrum (fig. 8). None of these preserve clearly observable sacral ribs, but there are broken regions on some of them that may represent the bases of ribs. We conservatively refer to these as caudosacrals, and note that if some or all of them are proper sacrals, then the sacrum of Balaur contained at least 5–7 vertebrae. Furthermore, there is an unfused single vertebra between the fused sacrum-plus-caudosacral block and the first caudal vertebra. We also refer to this as a caudosacral vertebra, based on comparison to Velociraptor, in which the first caudal is located between the postacetabular blades of the ilium and becomes incorporated into the sacrum late in ontogeny (Norell and Makovicky, 1999). In Balaur, this vertebra is morphologically similar to the first caudal and possesses clear zygapophyses, supporting its identity as a caudal vertebra.

Caudal Vertebrae

Five anterior caudals are present, including the first two caudals that are embedded in the pelvic block, the third caudal that was discovered in articulation with the first two caudals but was removed during preparation, and two more posterior caudals that come from the anterior portion of the tail but whose exact position is unknown (referred to here as caudals D–E) (figs. 8–9). These latter two vertebrae are identified as anterior caudals because they retain transverse processes, which disappear posterior to caudal 11 in Deinonychus (Ostrom, 1969) and Velociraptor (Norell and Makovicky, 1997, 1999). In comparison with Deinonychus (Ostrom, 1969), caudal E of Balaur appears to be approximately equivalent to caudal 8–9, due to the fact that the transverse process is reduced to a thin longitudinal ridge and the neural spine to a sharp, anteroposteriorly short ridge that also appears to have been dorsoventrally short in life.

Fig. 9. 

Caudal vertebrae of Balaur bondoc (EME PV.313). Caudal 3 (A–B), caudal D (C–F), caudal E (G–L) in dorsal (A, D, I), ventral (B, E, J), right lateral (C, H), posterior (F, L), left lateral (G), and anterior (K) views. Abbreviations: posz, postzygapophysis; prez, prezygapophysis; tvp, transverse process; tvpr, transverse process reduced to a ridge. Scale bar equals 1 cm.

The caudal centra lengthen posteriorly and their anterior and posterior articular faces, where well preserved, are concave. The ventral surfaces of caudals 3, D, and E are excavated by a longitudinal groove that is deepest at its center and shallows anteriorly and posteriorly before abruptly terminating at the articular face. Unfortunately, the ventral surfaces of caudals 1 and 2 are embedded in matrix and are not visible. The lateral surfaces of the caudal vertebrae are smoothly and slightly depressed, but there are no pneumatic foramina or other features of obvious pneumatic origin. The centrum and neural arch are coossified in all specimens and their suture has been obliterated.

The anteriormost four preserved caudal vertebrae possess large, winglike transverse processes that extend horizontally and posteriorly, and slightly expand distally into a spatulate swelling (fig. 9: tvp). In caudal E, however, the transverse process is reduced to a longitudinal ridge on the lateral surface of the neural arch (fig. 9: tvpr). This ridge extends along the entire anteroposterior length of the centrum, whereas the discrete transverse processes of the four more anterior caudals are centered on the centrum. The neural spine is incomplete on all caudal vertebrae and preserved only at its base. On caudals 1 and 2 the base is very thin, less than 1 mm in transverse width, but anteroposteriorly elongate. In contrast, the spine base of caudals 3, D, and E is shortened anteroposteriorly, suggesting that the neural spine was progressively reduced in size posteriorly along the tail before disappearing altogether, as in Deinonychus and Velociraptor (Ostrom, 1969; Norell and Makovicky, 1997, 1999).

The zygapophyses are poorly preserved on most caudal vertebrae. The postzygapophyses of the first caudal are preserved in articulation with the prezygapophyses of the second caudal, and this contact is unfused. In more posterior caudal vertebrae the zygapophyses are mostly broken, but the prezygapophysis is well preserved on both sides of caudal E (fig. 9: prez). These structures are elongate, extending approximately 3 mm anterior to the anterior articular face, and diverge from each other when seen in dorsal view. The anterior end of the prezygapophysis is swollen relative to its neck, and the articular surface is smooth, concave, and faces mostly medially but also slightly dorsally. The postzygapophysis is well preserved on the right side of caudal D. It is shorter than the prezygapophysis and extends only slightly posterior to the posterior articular face. The articular surface is smooth and flat, and faces mostly laterally but also slightly ventrally. There is a sharp lamina, preserved only on small regions of caudals D and E, that extends from the prezygapophysis to the postzygapophysis along the lateral surface of the vertebra. Because all five caudal vertebrae of Balaur are from the anterior region of the tail, none preserves the elongate prezygapophyses of the middle-posterior caudal vertebrae that are characteristic of derived dromaeosaurids (e.g., Ostrom, 1969; Norell and Makovicky, 1997, 1999; Hwang et al., 2002).

Sternum

Ossified sternal plates were discovered with the specimen but were so poorly preserved that they could not be conserved (see fig. 4B for a drawing of the specimen as it was excavated). As a result, few details of sternal morphology are apparent. However, it is clear that the two sternal plates are separate and unfused on the midline, as in Velociraptor (Norell and Makovicky, 1997, 1999) and most other dromaeosaurids for which sterna are known (Xu et al., 2003; Turner et al., 2007a), and that the articular facet for the coracoid is positioned along the anterior edge of the sternum, as is present in most coelurosaurs (Xu et al., 1999; Turner et al., 2007a).

Scapulocoracoid

Portions of both the left and right scapulocoracoids are present, including a nearly complete left bone (fig. 10) and a small fragment of the glenoid region of the right element (fig. 11). The left scapulocoracoid is 94 mm long anteroposteriorly as preserved, but it is eroded posteriorly and would have been considerably longer in life. The scapula and coracoid are completely fused on both sides and the suture between them has been obliterated. Such fusion is also present in Velociraptor (Norell and Makovicky, 1999), Microraptor (Xu et al., 2000, 2003; Hwang et al., 2002), and Adasaurus (IGM 100/20), but the scapula and coracoid are unfused in most dromaeosaurids, including Achillobator (Perle et al., 1999), Buitreraptor (Makovicky et al., 2005), Deinonychus (Ostrom, 1969), Sinornithosaurus (Xu et al., 1999), and Unenlagia (Novas and Puerta, 1997). It is possible that scapulocoracoid fusion is ontogenetically variable, but its presence in multiple Velociraptor and Microraptor specimens of different sizes (Norell and Makovicky, 1999; Hwang et al., 2002) and in the diminutive paravian Pneumatoraptor (115Ősi et al., 2010) suggests that this feature is not solely characteristic of large-bodied taxa or adults.

Fig. 10. 

Left scapulocoracoid of Balaur bondoc (EME PV.313) in lateral (A); medial (B), dorsal (C), ventral (D), and anterior (E) views. Abbreviations: ac, acromion; ct, coracoid tubercle; cvp, coracoid ventral process; gl, glenoid (articular surface eroded on this specimen); sb, scapula blade; sgf, subglenoid fossa. Scale bar equals 1 cm.

Fig. 11. 

Right scapulocoracoid of Balaur bondoc (EME PV.313) in lateral (A, E), medial (B), dorsal (C), and ventral (D) views. E shows lateral view of specimen as it was discovered, with an additional piece of the distal scapular blade. Abbreviations: ac, acromion; cf, coracoid foramen (arrow pointing to location of foramen, which is obscured in lateral view by coracoid tubercle); ct, coracoid tubercle; cvp, coracoid ventral process; gl, glenoid; sgf, subglenoid fossa. Scale bar equals 1 cm.

The glenoid is formed by both the scapula and coracoid, but the relative contributions of each bone are uncertain because their suture is not visible. Unfortunately, the glenoid is preserved only on the right specimen, where it is heavily eroded (fig. 11: gl). However, it is clear that the glenoid faces mostly ventrally but also slightly laterally, as the smooth articular surface continues onto the lateral surface of the bone. This feature is characteristic of paravians, including other dromaeosaurids and avialans (e.g., Turner et al., 2007a, 2012), and allows increased humeral abduction, a necessary component of the avian flight stroke (Gauthier and Padian, 1985). The glenoid of Balaur is bounded anteriorly and posteriorly by enlarged, swollen rims.

The coracoid is incomplete in both specimens, and only the region near the glenoid and the base of the ventral process are preserved. As in paravians generally, the coracoid is L-shaped, due to a discrete ventral process that is inflected medially to articulate with the anterior margin of the sternum (e.g., Norell and Makovicky, 1999; Hwang et al., 2002; Turner et al., 2007a, 2012). The ventral process is large and robust, and measures 28 mm long as preserved on the more complete left coracoid, but its anterior edge is broken and it would have been larger in life (figs. 10–11: cvp). Anteroventral to the glenoid, on the posterolaterally facing surface of the ventral process, is a shallow subglenoid fossa (figs. 10–11: sgf). This triangular fossa is present in other paravians, as well as some ornithomimosaurs (e.g., Norell and Makovicky, 1999; Turner et al., 2007a, 2012). Compared to the morphology of other dromaeosaurids the subglenoid fossa of Balaur appears to be shallowly inset, but this may be an artifact of poor preservation. The fossa of Tsaagan is also described as shallow, however, so there may be phylogenetically informative variation in the depth of the fossa among dromaeosaurids (Norell et al., 2006).

The coracoid tubercle (sometimes referred to as the biceps tubercle) is autapomorphically enlarged in Balaur (figs. 10–11: ct) The tubercle is located immediately anterior to the glenoid and projects far laterally from the coracoid as a discrete, convex, rugose bulge. When the scapulocoracoid is held in life position the tubercle forms the anterolateral corner of the bone, which is most apparent in dorsal view, where it is seen to project as a pronounced, rounded corner. The large size of the tubercle is also evident in lateral view, as it rises above the floor of the coracoid foramen to obscure this small opening in lateral view. Although the coracoid tubercle of other dromaeosaurids is large (e.g., Velociraptor: Norell and Makovicky, 1999; Buitreraptor: Makovicky et al., 2005), Balaur is unusual in possessing a discrete, conical, rugose tuber that projects far laterally and obscures the floor of the coracoid foramen in lateral view.

Only the ventral margin of the coracoid foramen is present, and it is preserved as a smooth, concave notch that is 5 mm long anteroposteriorly (fig. 11: cf). The foramen is positioned 10 mm directly medial to the coracoid tubercle and is located at approximately the same dorsoventral level as the midpoint of the glenoid, as in other dromaeosaurids (e.g., Norell and Makovicky, 1999; Norell et al., 2006). Additionally, as in Tsaagan and Deinonychus, the foramen appears to form a small channel, which trends dorsomedially-ventrolaterally through the interior of the coracoid (Norell et al., 2006). In contrast, the coracoid foramen of most theropods is a larger aperture that opens straight laterally-medially (e.g., Madsen, 1976; Colbert, 1989; Madsen and Welles, 2000; Brochu, 2003; Brusatte et al., 2008). Although not explicitly mentioned by Norell et al. (2006) when they described this unusual feature, Velociraptor also appears to possess this morphology (IGM 100/986).

The scapula is much better preserved and more complete on the left specimen, but its blade is broken posteriorly at the point where it begins to expand in dorsoventral depth. The blade is 9 mm deep by 4 mm wide mediolaterally at its midshaft, and is 13 mm deep at the posterior broken margin. The acromion expands at least 10 mm dorsally relative to the blade, but its dorsal margin is broken. Because the acromion is a very thin and large structure in better preserved dromaeosaurid scapulae, it may have been much larger in life in Balaur.

The scapula blade is long and gracile, with approximately parallel dorsal and ventral edges that begin to diverge slightly from each other posteriorly (fig. 10: sb). The lateral surface of the blade is flat or shallowly convex across most of its length, but becomes more strongly convex anteriorly as the scapula curves medially ventral to the acromion to articulate with the coracoid. The medial surface, on the other hand, is flat anteriorly and becomes convex posteriorly, and there is no medial ridge as present in many theropods (e.g., Brochu, 2003). The acromion is broken dorsally, but the entire preserved portion has a smoothly concave lateral surface, as is characteristic of dromaeosaurids such as Velociraptor (Norell and Makovicky, 1999) (fig. 10: ac). Furthermore, the acromion is clearly everted, such that it projects anteromedially relative to the parasagittal long axis of the scapular blade, which is readily apparent in dorsal view. As a result, the lateral surface of the acromion faces not only laterally but also anteriorly as well. Such eversion is a characteristic feature of paravians and is present in other dromaeosaurids (e.g., Ostrom, 1969; Norell and Makovicky, 1999; Burnham et al., 2000; Hwang et al., 2002; Norell et al., 2006), troodontids (e.g., Xu et al., 2002), and avialans (Clarke and Norell, 2002; Zhou and Zhang, 2002, 2003; Clarke et al., 2006). The medial surface of the acromion of Balaur is also concave, but much shallower than the lateral surface concavity.

Humerus

Both the left and right humeri are known, and the right is nearly complete and better preserved (fig. 12; table 2). This bone is straight in anterior and posterior views, but sigmoidally curved, with a concave anterior surface and convex posterior surface, in lateral and medial views. Bowed humeri are also present in most dromaeosaurids, including Austroraptor (Novas et al., 2009), Deinonychus (Ostrom, 1969), Microraptor (Hwang et al., 2002), and Velociraptor (Norell and Makovicky, 1999). In contrast, the humeri of Buitreraptor (Makovicky et al., 2005) and Unenlagia (Calvo et al., 2004) are straight in lateral and medial view, with parallel anterior and posterior margins. Unfortunately, because both scapulae are incomplete, it is unclear whether the humerus of Balaur was longer than the scapula, as is the case in avialans and most, but not all, dromaeosaurids (Turner et al., 2007a, 2012).

Fig. 12. 

Humeri of Balaur bondoc (EME PV.313). Left humerus (A–F) and right humerus (G–L) in anterior (A, G), posterior (B, H), lateral (C, I), medial (D, J), proximal (E, K), and distal (F, L) views. Abbreviations: brid, ridge bisecting fossa on anterior surface of distal end; dfos, fossa on anterior surface of distal end; dpc, deltopectoral crest; ec, ectepicondyle; en, entepicondyle; et, external tuberosity; gr, groove between humeral head and internal tuberosity; hh, humeral head; it, internal tuberosity; lrid, ridge on lateral surface of distal end; mpc, medial crest on anterior surface of proximal end, demarcating proximal fossa; mrid, ridge on medial surface of distal end; pfos, fossa on anterior surface of proximal end; rc, radial condyle; slc, sinuous crest on lateral surface continuous with deltopectoral crest. Scale bars equal 1 cm. Smaller scale bar on left for anterior, posterior, lateral, and medial views. Larger scale bar on right for proximal and distal views.

The proximal portion of the humerus is expanded relative to the shaft. The proximal margin is approximately straight in anterior and posterior views, as is characteristic of paravians (e.g., Ostrom, 1969; Norell and Makovicky, 1999; Burnham et al., 2000; Currie and Dong, 2001; Hwang et al., 2002; Calvo et al., 2004), but contrasting with the highly convex or laterally sloped margins of more basal theropods (e.g., Madsen, 1976; Currie and Chen, 2001; Brochu, 2003; Carpenter et al., 2005; Zanno, 2006). The anterior surface of the proximal end is deeply excavated by a smooth fossa (fig. 12: pfos), bordered laterally by the deltopectoral crest and medially by the large internal tuberosity. A nearly identical fossa is also present in other dromaeosaurids (e.g., Ostrom, 1969; Burnham et al., 2000; Calvo et al., 2004) and troodontids (Currie and Dong, 2001). In Balaur, both the lateral and medial margins of the fossa are demarcated by a slight crest that separates the convex anterior surfaces of the deltopectoral crest and internal tuberosity from the fossa itself. The medial crest becomes thinner and sharper proximally where it merges with the base of the internal tuberosity, and its pronounced morphology may represent a unique feature of Balaur (fig. 12: mpc). However, similar, albeit weaker, crests are also present in Bambiraptor (Burnham et al., 2000), Deinonychus (Ostrom, 1969), and Velociraptor (IGM 100/985), and we hesitate to regard a subtle feature of the medial crest morphology as an autapomorphy of Balaur.

The proximal articular surface is somewhat eroded in both humeri, but several important details are apparent. The humeral head is missing on the right humerus and only half remains on the left, but it is clear that it is bulbous and strongly overhangs the posterior, but not the anterior, surface of the humerus (fig. 12: hh). This is also the case in other dromaeosaurids such as Bambiraptor (Burnham, 2004), Deinonychus (Ostrom, 1969), and Velociraptor (IGM 100/985). Medial to the head, and separated from it by a groove on the posterior surface, is an enlarged internal tuberosity (fig. 12: it). The elongate morphology of the internal tuberosity is characateristic of dromaeosaurids (e.g., Ostrom, 1969; Novas and Puerta, 1997; Burnham et al., 2000; Calvo et al., 2004; Makovicky et al., 2005) and is also present in other coelurosaurs (e.g., Choiniere et al., 2010). The enlarged nature of the tuberosity is readily apparent in medial view, in which the internal tuberosity appears as a convex, bulbous structure that is expanded anteriorly and posteriorly. Lateral to the head there is a small tuberosity on the proximal surface, which projects proximally as a discrete, conical structure, but is not expanded anteriorly or posteriorly (fig. 12: et). This tuber likely corresponds to the external tuberosity, but comparison with other dromaeosaurids is difficult due to poor preservation in most taxa.

The deltopectoral crest extends for approximately 30% of the length of the humerus, a comparable ratio to that seen in other dromaeosaurids (e.g., Ostrom, 1969; Novas and Puerta, 1997; Burnham et al., 2000; Calvo et al., 2004; Xu and Wang, 2004; Makovicky et al., 2005). The crest extends straight anteriorly, such that in proximal view it is approximately perpendicular to the long axis of the proximal surface of the humerus (fig. 12: dpc). This is also the case in most other dromaeosaurids, but the deltopectoral crests of the unenlagiines Austroraptor (Novas et al., 2009), Buitreraptor (Makovicky et al., 2005), and Unenlagia (Calvo et al., 2004) project anteriorly and laterally, and hence form an obtuse angle relative to the proximal surface. Proximally the crest is thin and sheetlike, with a mediolateral width of 2 mm, but distally it thickens and expands in width to 5 mm. The humeral shaft immediately distal to the deltopectoral crest is marked by a sinuous crest following the edge where the posterior and lateral surfaces meet (fig. 12: slc). This crest parallels the lateral margin of the bone for 22 mm. Medial to the crest, on the posterior surface of the humerus, is a deep and elongate groove. A nearly identical crest and groove are present in Bambiraptor (Burnham et al., 2000), Deinonychus (Ostrom, 1969), Linheraptor (Xu et al., 2010), and Velociraptor (Norell and Makovicky, 1999), and likely represents a derived character shared by these taxa. A ridge is also present in Buitreraptor (MPCA 245) and Unenlagia (Calvo et al., 2004), but it is located more proximally, such that it follows the trend of the deltopectoral crest itself, and does not demarcate a groove (Calvo et al., 2004: figs. 9–10).

The humeral shaft is circular in cross section at its midpoint. However, all four surfaces gradually flatten distally, especially the anterior and posterior surfaces. A subtle protuberance is present along the medial margin of the posterior surface, near the midshaft region of the bone. This delicate feature is difficult to assess in other dromaeosaurids but does not appear to be present in Velociraptor (IGM 100/985). A much larger tuberosity, referred to in the literature as a “spur,” is present in this region in some derived therizinosauroids (Zanno, 2006), and likely is a site of muscle attachment.

The distal end is expanded relative to the shaft and is excavated by a large triangular fossa on its anterior surface (fig. 12: dfos). The fossa, which extends approximately 25% of the proximodistal length of the humerus, is bordered laterally and medially by pronounced crests, which originate from the proximal tips of the ectepicondyle and entepicondyle, respectively (fig. 12: ec, en). The medial ridge is oriented straight proximodistally, extends to the midpoint of the humeral shaft, and gradually becomes sharper and more prominent as it continues proximally (fig. 12: mrid). At its most proximal position, before abruptly merging with the shaft, the ridge forms a sheetlike shelf that overhangs the medial surface. The lateral crest is oriented obliquely, proximomedially-distolaterally, and is shorter than the medial crest. Where it terminates proximally it merges with a third ridge, which extends straight proximodistally to bisect the fossa (fig. 12: brid). This ridge does not separate the fossa into two equal partitions, but is located closer to the medial margin and terminates distally at the proximal tip of the ulnar condyle. A deep fossa, demarcated by lateral and medial ridges, is present in other dromaeosaurids and troodontids (e.g., Ostrom, 1969; Burnham et al., 2000; Currie and Dong, 2001; Calvo et al., 2004). However, the prominent, shelflike morphology and extreme proximal extent of the medial ridge are autapomorphic for Balaur.

Much of the distal articular surface is eroded on both humeri of Balaur, including the entirety of the ulnar (medial) condyle. However, portions of the radial (lateral) condyle are present, and these poorly preserved regions indicate that this structure was located on the anterior surface of the humerus (fig. 12: rc). This is unique among dromaeosaurids, which otherwise have the normal theropod condition of a distally placed radial condyle, but is also seen in therizinosauroids and often is considered a diagnostic feature of that clade (e.g., Zanno, 2006). The ectepicondyle is mostly eroded on both humeri, but is clearly present as a discrete structure, separated from the radial condyle (fig. 12: ec). The entepicondyle is preserved on the right humerus and takes the form of a robust ridge that is oriented proximodistally and extends for a total length of 10 mm (fig. 12: en). It is separated from the base of the ulnar condyle by a groove on the anterior surface of the humerus; although the ulnar condyle is missing, the groove itself is well preserved. A subtle fossa is present immediately medial to the entepicondyle, but it does not extend further proximally than the entepicondyle itself. In general, the morphology of the distal humerus is extremely similar to Deinonychus in overall morphology (Ostrom, 1969).

An unusual feature is present on the lateral surface of the distal end. Beginning near the distal termination of the humerus is a thin, sharp, sinuous ridge that continues proximally for 1/3 of the length of the bone (fig. 12: lrid). This ridge is not present in Bambiraptor (Burnham et al., 2000), Buitreraptor (Makovicky et al., 2005), Deinonychus (Ostrom, 1969), Unenlagia (Calvo et al., 2005), or Velociraptor (IGM 100/985), and is therefore an autapomorphy of Balaur. A ridge is present on the lateral surface of the distal end in Bambiraptor, but this is simply the ridge defining the lateral margin of the anterior fossa, as described above, which is somewhat reoriented relative to other dromaeosaurids such that it is also visible in lateral view (Burnham et al., 2000: fig. 3).

Referred Humerus

A left humerus, approximately 1.45 times the size of the holotype humeri, is among the referred specimens here assigned to Balaur sp. (LPB [FGGUB] R.1580) (fig. 13; table 3). It can be referred to Balaur based on the possession of autapomorphies, including the sinuous crest on the lateral surface of the distal end (fig. 13: lrid) and the elongate and shelflike medial crest demarcating the anterior fossa on the distal end (fig. 13: mrid). Additionally, it has a radial condyle located on the anterior surface of the distal end, which is also present in the holotype but otherwise unknown among dromaeosaurids (fig. 13: rc). Finally, it possesses dromaeosaurid features such as the groove distal to the deltopectoral crest and the deep fossa on the anterior surface of the distal end. Although this bone can be referred to Balaur it does exhibit some differences compared to the much smaller holotype humeri. For instance, on the posterior surface of the proximal end there is a distinct oblique crest that rises to the base of the head, which is absent in the holotype. Furthermore, on the anterior surface of the proximal end there is a ridge bisecting the deep fossa, whereas there is no median ridge in the holotype. Finally, the detailed morphology of the fossa differs between the two specimens: the entire fossa is excavated as a smooth concavity in the holotype, but in the referred specimen the region between the novel bisecting ridge and the medial ridge is convex.

Fig. 13. 

Referred humerus of Balaur sp. (LPB [FGGUB] R.1580). Left humerus in anterior (A), posterior (B), lateral (C), and medial (D) views. Abbreviations: dpc, region of deltopectoral crest (broken in this specimen); lrid, ridge on lateral surface of distal end; mrid, ridge on medial surface of distal end; rc, radial condyle. Scale bar equals 1 cm.

Ulna

Left and right ulnae are present in the holotype and both bones are generally well preserved (fig. 14). The ulna is bowed, as is usual for maniraptorans (Gauthier, 1986), and is of approximately the same gracile proportions as the radius (table 2). The ulna and radius are of approximately the same build in close relatives of Balaur, such as Bambiraptor (Burnham, 2004; Senter, 2006), Deinonychus (Ostrom, 1969; Senter, 2006), and Velociraptor (IGM 100/982). In contrast, the ulnae of more distantly related dromaeosaurids such as Microraptor (Hwang et al., 2002), Rahonavis (Forster et al., 1998), and Sinornithosaurus (Xu et al., 1999) are approximately twice as thick as the corresponding radius. The ulna of Balaur is relatively longer (compared to humeral length) than in close relatives such as Deinonychus and Saurornitholestes, but is more similar in proportions to the ulnae of the microraptorines Microraptor (CAGS 20-8-001; Hwang et al., 2002) and Graciliraptor (Xu and Wang, 2004) (table 4; see also figs. 42–43). In Bambiraptor (Burnham, 2004), and especially Mahakala (Turner et al., 2007a), the ulna is even more proportionally elongated than in Balaur.

Fig. 14. 

Ulnae of Balaur bondoc (EME PV.313). Left ulna (A–F) and right ulna (G–L) in lateral (A, G), medial (B, H), anterior (C, I), posterior (D, J), proximal (E, K), and distal (F, L) views. Abbreviations: arid, ridge on anterior surface; cp, coronoid process; cpr, ridge extending distally from coronoid process; dfos, fossa on the medial surface of the distal end, abutting the articular condyles; fos, fossa; ol, olecranon process; pfos, fossa on the medial surface of the proximal end. Scale bar equals 1 cm. Proximal end toward top in lateral, medial, anterior, and posterior views.

TABLE 4.

Measurements of forelimb elements (in mm) of dromaeosaurid theropods Asterisk indicates incomplete measurement due to damage or poor preservation.

The proximal end is triangular in proximal view. Its medial surface is excavated by a deep, ovoid fossa (fig. 14: pfos) and its lateral surface bears a large coronoid process that fits against the medial surface of the radius (fig. 14: cp). The coronoid process is continuous with a long, sharp crest that extends for 32 mm proximodistally and is only 2 mm thick anteroposteriorly at its midpoint (fig. 14: cpr). It is bordered both anteriorly and posteriorly by shallow fossae that parallel the crest across its entire length. A similar morphology is present in Deinonychus (Ostrom, 1969), but is difficult to assess in other dromaeosaurids because many specimens lack ulnae or because the ulna is often preserved flattened on slab specimens (especially in the Chinese Jehol material). The olecranon process of Balaur is mostly eroded (fig. 14: oc), but was clearly large, distinct, and conical, as in other dromaeosaurids (e.g., Ostrom, 1969; Forster et al., 1998; Norell and Makovicky, 1999; Hwang et al., 2002; Turner et al., 2007a).

The ulnar shaft is highly convex laterally and posteriorly, flat to slightly convex medially, and flattened anteriorly. As a result, it is somewhat D-shaped in cross section. Furthermore, the lateral and medial margins of the bone, where the flat anterior surface meets the more convex lateral and medial surfaces, are defined as sharp ridges due to the abrupt change of surface orientation. The flat anterior surface is bisected by an elongate, sharp ridge that is oriented proximodistally (fig. 14: arid). The ridge begins proximally at the level of the distal termination of the coronoid process and extends distally before terminating against an eroded surface 31 mm from the distal end of the bone. However, based on the morphology of the referred ulna (see below), it is likely that the ridge continued to the distal end of the bone. The flat anterior surface and bisecting ridge together comprise an autapomorphy of Balaur, as the anterior ulna is more convex and lacks a sharp, elongate ridge in other dromaeosaurids (e.g., Ostrom, 1969; Burnham, 2004). Unfortunately, surface preservation is not fine enough to assess the presence or absence of quill knobs for the attachment of secondary flight feathers, as are present in extant avialans and have been identified in the dromaeosaurids Rahonavis (Forster et al., 1998) and Velociraptor (Turner et al., 2007b). It is clear, however, that there was no pronounced bicipital tuberosity on the anterior surface of the proximal end of the shaft.

At approximately the midpoint of the anterior surface of the shaft there are distinct crests paralleling the lateral and medial margins of the bone, the latter of which is thicker and more prominent. Together these ridges enclose a subtle fossa. Farther distally, on the medial surface of the bone, is a deep fossa bordered anteriorly by the medial crest described above and distally by the raised, convex distal articular surface (fig. 14: dfos). The distalmost portion of the fossa is invaginated, as it is laterally enclosed by the distal articular surface itself. This fossa may be a unique feature of Balaur, as it is not present in some other dromaeosaurids (e.g., Ostrom, 1969; Burnham, 2004), but we hesitate to regard it as an autapomorphy here because comparable, well-preserved distal ulnae are not known for most dromaeosaurid taxa. The distal articular surface itself is mostly eroded in both ulnae, but preserved regions indicate that it was a convex, single surface and was not separated into multiple condyles.

Referred Ulna

A left ulna, approximately 60% larger than that of the holotype, is among the referred specimens assigned to Balaur sp. (LPB [FGGUB] R.1581) (fig. 15; table 3). It can be referred to Balaur based on the possession of the autapomorphic flat anterior surface bisected by a sharp proximodistal ridge (fig. 15: arid). The preserved portion of the referred ulna is 133 mm long proximodistally, but comparison with the holotype indicates that the complete bone would have measured approximately 163 mm. If the referred humerus and ulna belong to the same individual, as we argue above, then the referred ulna is proportionally larger relative to the humerus than in the holotype (table 4).

Fig. 15. 

Referred ulna of Balaur sp. (LPB [FGGUB] R.1581). Left ulna in lateral (A), medial (B), anterior (C), posterior (D), proximal (E), and distal (F) views. Abbreviations: arid, ridge on anterior surface; cp, coronoid process; fos, fossa; ol, olecranon process; pfos, fossa on the medial surface of the proximal end. Scale bar equals 1 cm. Proximal end toward top in lateral, medial, anterior, and posterior views.

Radius

Left and right radii, both generally well preserved, are present in the holotype (fig. 16; table 2). The radius is gracile and nearly straight, as is normal for dromaeosaurids (Ostrom, 1969; Forster et al., 1998; Norell and Makovicky, 1999; Hwang et al., 2002; Burnham, 2004; Xu et al., 2010). The proximal surface is ovoid in proximal view and smoothly concave to articulate with the humerus. The posterior region of the proximal surface is thickened, and although somewhat eroded, clearly raised above the level of the more anterior region.

Fig. 16. 

Radii of Balaur bondoc (EME PV.313). Left radius (A–F) and right radius (G–L) in lateral (A, G), medial (B, H), anterior (C, I), posterior (D, J), proximal (E, K), and distal (F, L) views. Abbreviations: fos, fossa on proximal end of medial surface. Scale bar equals 1 cm. Proximal end toward top in lateral, medial, anterior, and posterior views.

The slender shaft has a highly convex medial surface, which is so strongly convex at the midpoint of the bone that the entire surface resembles a swollen ridge. Here, the radius would have fit against the ulna. Proximally, there is an elongate fossa along the anterior edge of the medial surface (fig. 16: fos). Although somewhat exaggerated by erosion in the right radius, this fossa is undeformed on the left bone, where it measures 12 mm in proximodistal length and is demarcated posteriorly by a thin ridge. This smooth depression would have articulated against the coronoid process (and its associated elongate crest) of the ulna. The lateral surface of the shaft is flat proximally but becomes concave distally, where it is bisected by a proximodistally oriented ridge. This ridge, which measures 6 mm in length, is well preserved on both specimens. Two oblique muscle attachment ridges are present on the lateral surface proximal to midshaft.

The distal surface is ovoid and divided into two convex condyles to articulate with the semilunate carpal. The posterior condyle is much larger, and the two are separated by a smooth, concave trochlear surface. A double articular surface is normal for maniraptorans and is present in other dromaeosaurids, including Bambiraptor (Burnham, 2004) and Deinonychus (Ostrom, 1969). The medial surface of the distal end is concave for articulation with the distal ulna.

Carpometacarpus

The manus of Balaur is remarkably modified relative to other dromaeosaurids, and indeed, other dinosaurs in general (fig. 17). Notably, the three metacarpals (I–III) and at least some distal carpals are fused into a carpometacarpus (figs. 17–18), which is unknown among other dromaeosaurids and is otherwise present only in derived alvarezsauroids (e.g., Perle et al., 1994), perhaps in Avimimus (Vickers-Rich et al., 2002), and in derived avialans. The number of fused carpals is difficult to determine due to the extreme fusion of the carpometacarpus, which has obliterated sutures between individual bones. Similarly, the morphology of the carpals cannot be described in detail. The metacarpals themselves are fused proximally but diverge from each other distally, and much of their morphology is visible. Despite the heavy fusion, the first two metacarpals of Balaur are similar to those of other dromaeosaurids and show no affinity with the stocky metacarpals of alvarezsauroids (Perle et al., 1994; Choiniere et al., 2010) or the avialan condition in which metacarpals II and III are fused proximally and distally and metacarpal I is reduced to a nubbin. In the Balaur holotype, the right carpometacarpus is essentially complete (figs. 17–18), whereas only fragments of the left metacarpals I and II are present, broken and separated from each other (fig. 19). Fusion between metacarpals I and II proceeds farther distally than that between II and III, but these latter two metacarpals do contact each other distally at an unfused contact, which is an autapomorphy among dromaeosaurids.

Fig. 17. 

Right carpometacarpus and digits of Balaur bondoc (EME PV.313) in extensor view (unguals in medial view). Scale bar equals 1 cm.

Fig. 18. 

Right carpometacarpus of Balaur bondoc (EME PV.313) in extensor (A), flexor (B), medial (C), lateral (D), proximal (E), and distal (F) views. Abbreviations: ep, extensor pit; fg, flexor groove; mc I–III, individual metacarpals; ucII–III, distal unfused contact between metacarpals II and III; web, web of bone along the lateral edge of the extensor surface of metacarpal II. Scale bar equals 1 cm.

Fig. 19. 

Left carpometacarpus of Balaur bondoc (EME PV.313). Metacarpal II (A–F) and metacarpal I (G–H) in extensor (A, G), flexor (B, H), lateral (C), medial (D), proximal (E), and distal (F) views. Abbreviations: das, distal articular surface; ep, extensor pit; llp, lateral ligament pit; mlp, medial ligament pit. Scale bar equals 1 cm. Note that these two bones would have been fused in life but have been broken apart postmortem.

The metacarpus is shortened in Balaur, compared to the condition in most other dromaeosaurids (table 4; see also fig. 44). Metacarpal II represents about 35% of humeral length and 41% of ulnar length, compared to over 45% and 50%, respectively, in basal velociraptorines (Bambiraptor: Burnham, 2004), and in microraptorines (Microraptor: Hwang et al., 2002; Graciliraptor: Xu and Wang, 2004; Sinornithosaurus: IVVP V12811). Velociraptorines that are more closely related to Balaur, including Linheraptor (Xu et al., 2010) and Deinonychus (Ostrom, 1969), exhibit similar, but still larger, ratios than those recorded in Balaur.

Metacarpal I is small compared to metacarpals II and III, but nevertheless relatively longer (compared to metacarpal II) than in Velociraptor (IGM 100/982), more basal velociraptorines such as Linheraptor (Xu et al., 2010) and Bambiraptor (Burnham, 2004; AMNH FR 30556), and especially in microraptorines such as Microraptor (Hwang et al., 2002) and Sinornithosaurus (IVVP V12811). The relative degree of metacarpal I elongation in Balaur is most closely comparable to Deinonychus (YPM 5205), which has a metacarpal I that is slightly proportionally longer than in Balaur (table 4; see also fig. 44).

In Balaur, metacarpal I diverges medially relative to the remainder of the hand and has a lateral surface that abuts the medial surface of metacarpal II along a broad, flat contact, as is normal for theropod dinosaurs. Unfortunately, the proximal and lateral surfaces cannot be observed due to fusion, but the medial surface is subtly concave. Distally, there is no medial ligament pit, but the lateral ligament pit is large, deep, and ovoid. The flexor (ventral) surface is concave across its entire length, and this concavity is deepest distally where it smoothly merges with the flexor groove.

The distal articular surface is convex and asymmetrical, as is normal for theropods, allowing the phalanges to diverge medially relative to digit II. Posteriorly the distal surface divides into two condyles. The medial condyle is an elongate oval, with a long axis oriented ventromedially, which tapers posteriorly to a rounded point. The lateral condyle is smaller, more strongly convex distally, and has a long axis that is oriented closer to straight dorsoventrally (extensor-flexor direction). The condyles continue onto the flexor (ventral) surface and are separated by a deep flexor groove (fig. 18: fg), which excavates both the distal and flexor surfaces of the articular region. However, separate condyles do not extend onto the extensor (dorsal) surface, and in extensor view the distal margin appears as a slightly convex, continuously rounded margin. This is an autapomorphy among dromaeosaurids, and is associated with more limited manual extension relative to closely related taxa. In contrast, dromaeosaurids such as Deinonychus (Ostrom, 1969), Graciliraptor (Xu and Wang, 2004), and Velociraptor (Norell and Makovicky, 1999; IGM 100/982) possess a metacarpal I with distal condyles that extend onto the extensor surface, and are clearly apparent in extensor view as discrete structures separated by a concave margin. In Balaur, the reduced condyles are associated with the lack of an extensor pit on the extensor surface immediately proximal to the distal articular region, which is present in the other dromaeosaurids listed above.

Metacarpal II is the longest bone in the hand. Its extensor surface is flat proximally, becomes convex near the midpoint, and then is concave distally due to a large extensor pit (fig. 18: ep). Both medial and lateral ligament pits are present distally (fig. 19: mlp, llp). The distal articular surface is divided into two condyles of equal size, which are separated by an obtuse angle. As with metacarpal I, the condyles extend onto the flexor surface of the bone but are not apparent as discrete structures in extensor view. Furthermore, the groove between the condyles is present only on the flexor surface and does not extend onto the distal articular surface itself. However, the articular surface does extend onto the extensor surface, but does so as a single rounded convexity, and it is bordered by an extensor pit proximally (fig. 19: ep). Therefore, limited manual extension was possible. The flexor surface of the bone is concave proximally as a triangular fossa, then convex at midlength, and finally becomes concave again distally at the flexor groove. It is crossed obliquely by a weak ridge, stretching between the proximolateral corner, where it begins immediately distal to the fusion with metacarpal III and the edge of the medial condyle.

Unusually, there is a raised web of bone along the margin where the lateral and extensor surfaces meet (fig. 18: web). This begins proximally immediately distal to the fused contact of metacarpals II and III, and here it is located precisely at the dorsolateral corner of the bone. Further distally it curves ventrally (i.e., toward the flexor direction), and in doing so defines a flat, triangular facet dorsal to it (i.e., toward the extensor direction). This facet articulates against the distal end of metacarpal III. Between their proximal fused contact and distal unfused articulation (fig. 18: ucII–III), the two metacarpals parallel each other and are separated by approximately 2 mm. The web on metacarpal II is most pronounced at approximately one third of the length of the shaft, where it is visible in extensor view as a rugose crest that projects laterally, and in lateral view as a thickened ridge that is dorsally convex (toward the extensor direction). Although the web is prominent here, it directly articulates with metacarpal III only further distally, where it is weaker.

Metacarpal III is the most unusual element in the carpometacarpus, as it is extremely reduced in thickness relative to other dromaeosaurids such that it resembles a thin splint. In most theropods metacarpal III is more gracile than the first two metatarsals, but in Balaur it is only 30% as thick as metacarpal II. In contrast, metacarpal III is approximately 50% of the mediolateral thickness of metacarpal II in Velociraptor (IGM 100/982), and a similar ratio is seen in other dromaeosaurids (e.g., Ostrom, 1969; Hwang et al., 2002; Burnham, 2004; Xu and Wang, 2004; Senter, 2006). Therefore, the extremely reduced metacarpal III is an autapomorphy of Balaur. That said, metacarpal III of Balaur is not reduced in its anteroposterior length. In fact, it is slightly more elongated, compared to metacarpal II, than in closely related velociraptorines, and is comparable in relative length to the third metacarpals of microraptorines such as Microraptor (CAGS 20-7-004; Hwang et al., 2002), Graciliraptor (Xu and Wang, 2004), or Sinornithosaurus (IVVP V12811) (table 4; see also fig. 44).

Metacarpal III is thickest anteroposteriorly at its proximal fused end, but abruptly thins distally into a splint with an ovoid cross section. On the flexor surface, the proximal fused region is projected below the remainder of the metacarpus and extends to approximately the same level as the lateral edge of metacarpal I. Together, these two ventrally projected regions define a deep depression between them, which covers the entirety of the flexor surface of the proximal portion of metacarpal II and much of the flexor surfaces of the proximal regions of metacarpals I and III. Distally, the articular surface of metacarpal III is a single, rounded structure that is not divided into separate condyles. This is an unusual condition, departing from the double condyles present in other dromaeosaurids (e.g., Ostrom, 1969; Norell and Makovicky, 1999; Burnham, 2004; Xu and Wang, 2004), and is related to the atrophied morphology of the phalanges of the third digit (see below).

Referred Manual Metacarpal

A distal fragment of left metacarpal II (LPB [FGGUB] R.1585) is included within the referred material of Balaur sp. (fig. 20). The shaft and distal articular end are essentially complete and well preserved, but the proximal region, where the bone was likely coossified with metacarpals I and III, is missing. Furthermore, the extensor margin of the distal articular surface is largely abraded, as is the distalmost extremity of the flexor surface of the medial condyle, exposing cancellous bone texture.

Fig. 20. 

Referred metacarpal II of Balaur sp. (LPB [FGGUB] R.1585). Left metacarpal II in extensor (A), flexor (B), lateral (C), medial (D), and distal (E) views. Scale bar equals 1 cm. Proximal end toward top in lateral, medial, anterior, and posterior views.

Despite its fragmentary nature, this bone exhibits several features shared with the holotype of B. bondoc (EME PV.313). These include: the restriction of the intercondylar groove to the flexor surface of the distal end, so that the distal articular end is straight in extensor view; the widely divergent distal condyles on the flexor surface; and, especially, the unique raised web that would have overhung metacarpal III distally. As this referred specimen is disarticulated from the remainder of the (presumably) fused carpometacarpus, it also reveals some details of the lateral and medial surfaces that are hidden in the holotype by the closely appressed metacarpals I and III.

On the lateral edge of the extensor surface, the proximal part of the overhanging web is broken off, but its preserved distal part suggests it was also prominent, as in the holotype of B. bondoc, possibly even more so. Distally, it thins and becomes less pronounced, until it finally merges with the body of the bone near the distal end. Distal to the web, there is a small lateral extension, corresponding to the area of contact with the distal part of metacarpal III, as seen in the holotype of B. bondoc. Ventral to the web (in the flexor direction), the lateral edge of the flexor surface the shaft is marked by a less well developed, but still clearly marked, ridge, which extends from the area of sutural contact with metacarpal III to the proximal edge of the flexor surface of the lateral condyle distally. The area between this ridge and the overhanging dorsal web—the lateral surface of metacarpal II—is unevenly concave and rugose, and is distally excavated by a teardrop-shaped depression that marks the contact with the medial margin of the distal metacarpal III.

The flexor surface of the shaft is widely exposed in lateral view proximally, but reorients to face only ventrally (i.e., in the flexor direction) distally. The laterally exposed portion is positioned between the lateral edge of the flexor surface and an oblique ridge crossing the flexor surface. This ridge corresponds to, but is significantly more pronounced than, the slight longitudinal prominence present on the holotype metacarpal II of B. bondoc. A thin, longitudinal ridge dissects this laterally facing part of the flexor surface, extending somewhat less than halfway along the preserved length of the shaft. Distal to this point, the lateral exposure of the flexor surface diminishes rapidly, so that the flexor surface is only ventrally exposed, while also becoming widely concave transversely.

The medial edge of the extensor surface of the shaft is slightly raised as a ridge, and, as a result, the extensor surface is slightly depressed between the elevated lateral and medial edges. The raised medial edge flattens out at midlength of the shaft, but continues as a barely visible, thin ridge onto the distal portion of the medial surface of the bone. On the medial surface, a short, flat, triangular area faces ventromedially along the proximal region of the bone; it is bordered ventrally (in the flexor direction) by an oblique and subtly pronounced ridge ending at approximately the same level as the ridge that bisects the portion of the flexor surface that is exposed laterally (see above). The subtle ridge on the medial surface begins in a ventral position proximally and extends toward the midline of the medial surface as it continues distally, before eventually terminating in a small protuberance. Based on comparisons with the holotype metacarpus, the flat triangular facet on the medial surface is an articulation site for metacarpal I. Due to the ventromedial orientation of this articular surface, the line of contact between the two metacarpals would have been more laterally displaced on the flexor surface, as metacarpal II has a wider exposure in extensor view than in flexor view; this condition is also seen in the articulated carpometacarpus of the holotype of B. bondoc. Distal to the triangular contact facet for metacarpal I, the medial surface of metacarpal II is mildly convex with a small prominence distally, near the proximal edge of the flexor surface of the medial condyle.

In distal view, the articular surface is comprised of two widely separated condyles that diverge from each other ventrally (i.e, in the flexor direction). The lateral condyle is slightly narrower than the medial condyle and has a laterally concave outline in distal view, due to the presence of a well-developed lateral ligament pit that deeply excavates the lateral margin of the condyle and is opened distally. The medial ligament pit, on the other hand, is less developed, fully closed distally, and placed at approximately midheight of the condyle. As in the holotype of B. bondoc, the distal articular surface is continuous across the two condyles (i.e., not divided by a trochlear groove), and is essentially flat in flexor and extensor views. The condyles are proximally separated on the flexor surface by a wide and deep flexor groove, which is clearly separated from the articular surface itself. It is not possible to ascertain whether the condyles were also separate on the extensor surface, due to abrasion in this region. However, the presence of an extensor pit or groove is hinted at by a slight depression in the damaged area.

Manual Digits

The complete array of phalanges is preserved in both the left and right hands of the holotype (figs. 17, 21–Fig. 22. Fig. 23. Fig. 24. 25). As preserved, the phalangeal formula is 2-3-1, which is autapomorphic among dromaeosaurids due to the atrophied digit III (see below). In general, the lateral ligament pits are deeper and larger than the medial pits on all phalanges, and deep flexor grooves and the great extent of the distal condyles onto the flexor surface of the phalanges is indicative of substantial manual flexion abilities. On the contrary, the lack of deep extensor pits and the absence of any migration of the distal articular surface onto the extensor surface of the phalanges suggest that manual extension was limited relative to that of other dromaeosaurids (e.g., Ostrom, 1969; Norell and Makovicky, 1999). However, the ability to extend the individual elements of the manus may have become progressively greater distally along each digit, judging by the increasingly deeper extensor pits (although this is a tentative interpretation, because some extensor muscles do not have bony insertions). This is especially evident in digit II, as metacarpal II lacks an extensor pit, phalanx II-1 has a small extensor pit, and phalanx II-2 has a larger and deeper extensor pit, although it is still quite shallow compared to the condition in closely related taxa.

Fig. 21. 

Left manual digit I of Balaur bondoc (EME PV.313). Left phalanx I-1 (A–F) and left phalanx I-2 (G–H) in extensor (A), flexor (B), lateral (C, G), medial (D, H), proximal (E), and distal (F) views. Abbreviations: for, foramen; ft, flexor tubercle; gr, groove separating flexor tubercle from ventral surface of ungual. Scale bar equals 1 cm.

Fig. 22. 

Right manual digit I of Balaur bondoc (EME PV.313). Right phalanx I-1 (A–F) and right phalanx I-2 (G–H) in extensor (A), flexor (B), lateral (C, H), medial (D, G), proximal (E), and distal (F) views. Abbreviations: for, foramen; ft, flexor tubercle. Scale bar equals 1 cm.

Fig. 23. 

Left manual digit II of Balaur bondoc (EME PV.313). Left phalanx II-1 (A–F), left phalanx II-2 (G–L), and left phalanx II-3 (M–N) in extensor (A, G), flexor (B, H), lateral (C, I, M), medial (D, J, N), proximal (E, K), and distal (F, L) views. Abbreviations: ygr, proximal bifurcations of the Y-shaped groove. Scale bar equals 1 cm.

Fig. 24. 

Right manual digit II of Balaur bondoc (EME PV.313). Right phalanx II-1 (A–F), right phalanx II-2 (G–L), and right phalanx II-3 (M–N) in extensor (A, G), flexor (B, H), lateral (C, I, N), medial (D, J, M), proximal (E, K), and distal (F, L) views. Scale bar equals 1 cm.

Fig. 25. 

Manual digit III of Balaur bondoc (EME PV.313). Single phalanx of the right (A) and left (B–E) manual digit III. The right phalanx is figured in extensor view, with its proximal articular surface to the left. The left phalanx is figured in flexor (B), extensor (C), proximal (D), and distal (E) views. Scale bar equals 1 cm.

Manual Digit I

Digit I is approximately 72% of the length of digit II, but the two digits are essentially equally robust (table 2) (figs. 21–22). Phalanx I-1 is the largest phalanx in the hand, and has a triangular proximal articular surface that is depressed and saddle shaped to receive the distal condyles of the first metacarpal (figs. 21–22, A–F). The saddle is asymmetrical to match the condyles of the metacarpal: the lateral portion is deeper to accept the more convex lateral condyle of metacarpal I, and the two lateral and medial portions are oriented obliquely to each other. The long axis of the proximal surface is dorsomedially oriented, such that it appears twisted relative to the long axis of the phalanx itself. There are no flexor or extensor tubercles on the proximal surface, but there is a pronounced lip surrounding the entire proximal region, which is especially prominent ventrally (in the flexor direction).

The flexor surface of phalanx I-1 is concave, defined laterally and medially by longitudinal ridges on the edges of the shaft. The concavity is most pronounced proximally and distally adjacent to the articular ends. The extensor surface of the phalanx is convex and the lateral and medial surfaces are subtly convex but nearly flat. Distally there are ligament pits on both sides, unlike metacarpal I, which possesses only a lateral pit. The pits are extensively visible in extensor view because the distal end is expanded relative to the shaft.

The distal end is divided into two pronounced condyles posteriorly, which are approximately equal in size and extend to the same posterior ( =  flexor) level. They are separated by a deep groove and form an acute angle to each other. The groove continues ventrally as the flexor groove and also slightly onto the extensor surface, unlike metacarpal I which autapomorphically lacks any continuation of the groove dorsally (toward the extensor direction). Therefore, unlike the condition in the metacarpal, the separate condyles of phalanx I-1 are visible in extensor view, separated by the concave trace of the flexor groove.

The ungual of digit I, phalanx I-2, is the largest claw in the hand (table 2) (figs. 21–22, G–H). It is recurved, although not as strongly as in Deinonychus (Ostrom, 1969) and Velociraptor (IGM 1000/982), and is excavated by a single groove on the lateral and medial sides. The groove begins at the midpoint of the phalanx, curves as it continues distally, and terminates at the distal margin of the bone, slightly dorsal to the distal tip. The proximal articular surface is saddle shaped to match the distal condyles of phalanx I-1 and is surrounded by a pronounced lip, which is especially prominent along the lateral and medial edges. Ventrally there is a large, swollen flexor tubercle (fig. 21: ft), whose proximal base is penetrated by two small foramina, one on each side of the midline (fig. 21: for). The foramina are approximately 1.5 mm in diameter and are visible in both lateral and flexor views. It is possible that these are unique features of Balaur because they are absent in Velociraptor (IGM 100/982), but we hesitate to regard them as autapomorphic since the corresponding ungual of most other dromaeosaurids is not known or, if known, not well preserved. Anterior to the foramina, and separated from them, is a curved groove that divides the flexor tubercle from the flexor surface of the ungual (fig. 21: gr). This deep groove is also absent in Velociraptor and may be a unique feature of Balaur.

Manual Digit II

Digit II is the longest in the hand (table 2) (figs. 23–24). Phalanx II-1 has a proximal surface that is triangular in shape and only weakly depressed to articulate with the distal condyles of metacarpal II (figs. 23–24, A–F). A pronounced lip surrounds the entire proximal surface and is especially prominent ventrally (i.e., toward the flexor direction), where it extends much farther proximally than the dorsal lip. As a result, the ventral lip is visible in extensor view. The flexor surface of the phalanx is concave, which is deepest distally as a large flexor pit. The distal articular surface is divided into two condyles, of which the lateral is larger, more bulbous, and extends further distally. It is nearly circular in shape, with a long axis that is approximately dorsoventrally (extensor-flexor) oriented, whereas the smaller medial condyle is aligned ventromedially. The two condyles are separated by an obtuse angle and a deep groove, which continues onto the extensor and flexor surfaces of the bone. The lateral ligament pit is large and deep, but the medial pit is absent and the corresponding surface is flat and smooth.

The proximal surface of phalanx II-2 is saddle shaped and asymmetric to match the condyles of II-1 (figs. 23–24, G–L). The flexor surface is concave, but much more subtly than in phalanx II-1 or I-1. However, as in these other phalanges, the concavity on the flexor surface is deepest proximally and distally. The distal end is separated into two condyles of approximately the same size. Both are remarkably thin and sheetlike, as a result of the deep and wide groove between them. This groove forms an enormous trochlea and extends far onto the extensor and flexor surfaces of the bone. The extensor (dorsal) extent is especially extensive relative to all other phalanges of the hand and would have permitted more extension than most other joints of the manus. However, since the groove is raised above the rest of the extensor surface of the phalanx and is not bordered by an extensor pit proximally, the amount of extension was still quite limited. Lateral and medial ligament pits are present and both are placed along the extensor surface of the bone, unlike the more central pits of the other phalanges. The lateral pit is larger and deeper.

The ungual, phalanx II-3, is broken distally but appears to be slightly smaller than the ungual of the first digit (figs. 23–24, M–N). The flexor tubercle is smaller than that of the first ungual, and the presence or absence of the ventral foramina cannot be assessed due to breakage and poor preservation. One notable difference with the first ungual is that the lateral and medial grooves are Y-shaped, as the single groove that is present across most of the length of the bone bifurcates proximally into two short grooves that are oriented obliquely to each other (fig. 23: ygr). This feature is not present in Deinonychus (Ostrom, 1969) or Velociraptor (IGM 100/982; Norell and Makovicky, 1999), which possess a standard single groove on each surface, and is therefore an autapomorphy of Balaur.

Manual Digit III

Digit III is autapomorphically reduced in Balaur, both in the size and shape of individual elements and the total number of phalanges (figs. 17, 25). Theropods generally possess four phalanges in digit III, but Balaur preserves only a single small phalanx (phalanx III-1) on both the left and right hands, the latter of which is complete. This bone is reduced to a vestigial nubbin and indicates that digit III was probably nonfunctional in life, which is also suggested by the atrophied, splintlike metacarpal III. Proximally phalanx III-1 is ovoid in cross section, with a dorsoventrally (extensor-flexor) oriented long axis. The articular surface itself is a shallow, simple concave pit, which matches the single bulbous distal articular surface of metacarpal III. The flexor surface of the bone is concave and penetrated by two small foramina, one proximal to the other. The phalanx tapers distally and then expands slightly at its terminus, where the distal articular surface is a suboval, single convexity. It is possible that another small phalanx may have articulated here, since the distal surface does appear to be articular. However, if present, this phalanx must have been very small, and it may not have been ossified.

Referred Phalanges

Three incomplete phalanges were recovered from the same site that yielded the referred humerus, ulna, and metacarpal (fig. 26). These phalanges are referred to Balaur sp. based on general morphological similarity with the holotype of B. bondoc, the likelihood that they belong to the same individual as the other referred forelimb bones (which possess clear autapomorphies of Balaur), and due to the presence of wrinkled surface texture, an autapomorphy of Balaur well expressed in the holotype. Based on comparison with the complete hand of the holotype of B. bondoc, these referred phalanges can be identified as the right? phalanx I-1 (LPB [FGGUB] R.1583) (fig. 23A–E), left? phalanx II-1 (LPB [FGGUB] R.1582), and left phalanx II-2 (LPB [FGGUB] R.1584) (fig. 23F–J), respectively. Both R.1582 and R.1583 are proximal phalangeal fragments. Of these, R.1582 is the more complete, whereas R.1583 is better preserved. Fortunately, in R.1582 the dimensions and the shape of the flexor aspect of the distal articular end can be reliably reconstructed using the recovered impression of the distal, missing part on the matrix. The same holds for R.1584, a distal phalangeal fragment whose proximal extent and shape of its flexor surface are revealed by a similar impression. All phalanges are relatively long, slender, weakly arched dorsally (in the extensor direction), and internally hollow (due to a large medullary cavity).

Fig. 26. 

Referred manual digits of Balaur sp. ?Right manual digit I-1 (LPB [FGGUB] R.1583) in extensor (A), flexor (B), lateral (C), medial (D), and proximal (E) views. Left manual phalanx II-2 (LPB [FGGUB] R.1584) in extensor (F), flexor (G), lateral (H), medial (I), and distal (J) views. Scale bar equals 1 cm. Proximal end toward top in lateral, medial, anterior, and posterior views.

Manual phalanx I-1 (LPB [FGGUB] R.1583) is elongate, with a roughly square-shaped proximal articular surface that is excavated by a single (undivided) depression for the reception of the distal condyles of metacarpal I (fig. 26A–E). The ventral lip delimiting the proximal articular surface is well developed and extends slightly more proximally on the left side; a similar asymmetry is also evident on the dorsal lip. In flexor view, the phalanx is transversely concave between the raised lateral and medial margins. The concavity is especially deep proximally and distally. Far proximally, underneath the ventral lip of the articular surface, the concavity grades into a nearly flat platform. This platform is bounded laterally and medially by small, moundlike prominences, which are the proximal terminations of the raised lateral and medial edges of the flexor surface. Immediately above the medial prominence there is a smooth, teardrop-shaped pit that excavates the medial surface of the phalanx. The medial surface is marked by an angular ridge along its midline; this ridge begins dorsal to the teardrop-shaped pit (i.e., further in the extensor direction) and extends longitudinally to approximately the midpoint of the shaft, where it is most prominent. The extensor surface is marked by another angular longitudinal ridge that begins at the halfway point of the dorsal lip of the proximal articular surface and flattens as it extends distally to a point slightly beyond the middle of the bone. The more proximally extended and thickened dorsolateral corner of the dorsal lip gives the proximal end of the bone a markedly asymmetrical shape, and also makes it appear as if the proximal region is offset laterally from the longitudinal midline of the shaft. Distally, despite poor preservation, it is apparent that similarly sized distal condyles are divided by a flexor groove that appears to extend onto the distal articular surface, identical to the condition in the holotype phalanx I-1 of B. bondoc.

The preserved distal portion of manual phalanx II-2 (LPB [FGGUB] R.1584) is similar in most respects to those of the B. bondoc holotype (fig. 26: F–J). Both the referred and holotype material exhibit thin, sheetlike distal condyles; a well-developed groove separating the condyles in flexor, distal, and extensor views; and dorsally displaced ligament pits. In the referred material, the lateral ligament pit is deeper but the medial pit larger in area. In medial and lateral views the distal articular surface is conspicuously asymmetrical: the articular surface of the condyles extends far proximally on the flexor surface and cleanly merges with the shaft proximally, whereas on the extensor surface the condyles rise high above the shaft but their articular surfaces end abruptly, and in a more distal position than those on the flexor surface. There is no flexor pit, but a small extensor pit is notched between the proximal ends of the condyles.

Pelvic Crushing

The block containing the holotype pelvis and sacrum of Balaur is dorsoventrally crushed (fig. 27). Additional preparation after the original description of the specimen (Csiki et al., 2010a), along with detailed reexamination undertaken as part of this monograph project, reveals that this crushing is more extensive than originally appreciated. On both sides, crushing has cracked the pelvis along the dorsal and posterior bounding ridges of the cuppedicus fossa on the pubic peduncle (these pronounced ridges seem to have acted as local strong points). The deformation appears to be stronger on the left side, however, as there are several noticeable breaks on the left ilium and pubis, and because the left sacral rib is noticeably displaced ventrally relative to the right sacral rib (both ribs are in articulation with the ilium). This extreme deformation explains two supposed autapomorphies described by Csiki et al. (2010a), which we here consider invalid.

First, the cuppedicus fossa, which would have faced laterally in life, now faces almost entirely ventrally because of the crushing. The crushing is so extensive that Csiki et al. (2010a) mistook the broken dorsal bounding ridge of the cuppedicus fossa (which in life also defines the ventral margin of the iliac preacetabular process; fig. 27: dcfr) to be the ventral margin of the pubic peduncle. As a result, they misidentified the preacetabular process as the pubic peduncle, and the fossa covering it as the cuppedicus fossa, and regarded the laterally projecting nature of this part of the ilium to be autapomorphic. Indeed, a laterally (instead of ventrally) projecting pubic peduncle would be unusual, but the ventral portion of the preacetabular process does project laterally as a sheetlike flange in other dromaeosaurids (e.g., Velociraptor: IGM 100/986). Therefore, the condition in Balaur is similar to that in close relatives and is not autapomorphic. Csiki et al. (2010a) interpreted the supposed autapomorphy of Balaur as an attachment site for increased femoral extensor musculature, but our reinterpretation confirms that there was nothing unusual about the size or position of this muscle attachment site (at least compared to closely related dromaeosaurids such as Velociraptor and Adasaurus).

Second, on the left side of the pelvis the crushing of the posterior bounding ridge of the cuppedicus fossa (fig. 27: pcfr) was so extreme that the pubis was plastically deformed and snapped in two along the pubic peduncle. The break appears to occur slightly proximal to where the pubic peduncle of the ilium and the pubis are fused together. After the pubis was detached, it then rotated somewhat ventrally, such that its lateral surface now faces primarily ventrally. Csiki et al. (2010a) considered this to be a genuine feature of the anatomy, and hence an autapomorphy, but we here reinterpret it as an artifact of crushing, due to the obvious break in the pubis on the left side and the more traditional orientation of the pubis (lateral surface facing more strongly laterally) on the right side.

Ilium

Only the acetabular region and the ventral portions of the pre- and postacetabular processes are present on both left and right ilia of the holotype, the former of which is better preserved and more complete. The left ilium is 75 mm long anteroposteriorly as preserved, but would have been much longer in life because it is broken both anteriorly and posteriorly (table 2). The preserved regions are enough to indicate that the two ilia did not approach each other and make contact with the sacral neural spine apron, but rather diverged from each other laterally, as is the case in all dromaeosaurids (Norell and Makovicky, 1997; Turner et al., 2007a).

The most prominent feature of the acetabular region of the ilium is the laterally extensive and swollen antitrochanter. This is eroded laterally and consequently would have been even more pronounced in life (fig. 27, 30: anti). A large antitrochanter is normal for dromaeosaurids (Norell and Makovicky, 1997; Turner et al., 2007a; Brougham and Brusatte, 2010; contra Alexander et al., 2010). Anterior to the antitrochanter is a thin supraacetabular crest (fig. 27: sac), which is subtle and overhangs the acetabulum only slightly, as is normal for dromaeosaurids and many other coelurosaurs (e.g., Norell and Makovicky, 1997, 1999; Turner et al., 2007a). The delicate morphology of the supraacetabular crest has been described as permitting a greater ranger of femoral abduction (Longrich and Currie, 2009; Alexander et al., 2010), but the large antitrochanter of Balaur and other dromaeosaurids would in fact have precluded extreme lateral movement of the femur (Brougham and Brusatte, 2010). In lateral view, the supraacetabular crest of Balaur is nearly horizontal, and is not as arched as in Bambiraptor (Burnham et al., 2000), Deinonychus (Ostrom, 1969), Velociraptor (Norell and Makovicky, 1997, 1999), and other dromaeosaurids (e.g., Novas and Puerta, 1997; Hwang et al., 2002; Turner et al., 2007a; Longrich and Currie, 2009). In concert with the shape of the crest, the acetabulum of Balaur has more of an anteroposteriorly elongated oval shape, compared with the more circular acetabula of other dromaeosaurids. This may be an unusual feature of Balaur, but we suspect that it is largely, or entirely, an artifact of dorsoventral crushing, and do not interpret it as an autapomorphy.

Only a small region of the lateral surface of the ilium immediately above the acetabulum is preserved on the left side. Most of this region is smooth and slightly depressed, but posteriorly there is a pronounced oblique ridge extending anterodorsally from the antitrochanter; this ridge is mostly eroded on the specimen and only its ventral base is preserved (figs. 27, 30: obrid). This ridge is extremely well developed and much larger than similar oblique ridges present in Mahakala (Turner et al., 2011), Unenlagia (MCF PVPH-78), and Rahonavis (US 8656). In these dromaeosaurids, the oblique ridge terminates dorsally in a large supratrochanteric process. Therefore, given the size of the ridge in Balaur, it is likely that an enlarged supratrochanteric process (relative to closely related taxa such as Velociraptor and Adasaurus) would have been present at the eroded dorsal end of the ridge.

Posterior to the acetabulum and antitrochanter is the shallow brevis fossa, which faces completely ventrally and is entirely concealed in lateral view (fig. 30: bf). The medial shelf defining this fossa, the brevis shelf, projects nearly straight medially, and as a result the fossa is broad mediolaterally but is not particularly deeply inset ventrally. Unfortunately, fine details of the pubic and ischial peduncles are not observable due to poor preservation, crushing, and complete fusion of the pelvic bones. It is clear, however, that the anterior region of the acetabulum is not open, but rather formed by an enlarged, anterposteriorly extensive fossa, which would have formed part of the articulation with the femur. This fossa is located on the posterior portion of the pubic peduncle, as in Velociraptor and other dromaeosaurids (e.g., Norell and Makovicky, 1997, 1999; Novas and Puerta, 1997; Turner et al., 2007a, 2012). It is separated from the more anterior cuppedicus fossa (fig. 27: cf) by the dorsoventrally trending ridge on the lateral surface of the pubic peduncle (the “posterior bounding ridge” of the cuppedicus fossa, described above; fig. 27: pcfr). Unfortunately, the posterior fossa is largely covered by unremovable matrix on both sides of the pelvis, so it is not visible in the figures.

Anterior to the acetabulum, the lateral surface of the pubic peduncle is excavated by a deep and anteroposteriorly expansive cuppedicus fossa, which artificially faces mostly ventrally due to crushing (as described above) (fig. 27: cf). An expansive cuppedicus fossa that faces mostly laterally is present in other dromaeosaurids and other coelurosaurs more generally, contrasting with the deeply inset fossae of basal tetanurans that face strongly anteriorly and ventrally and are restricted to the corner where the preacetabular process and pubic peduncle meet (Norell and Makovicky, 1997, 1999; Hutchinson, 2001). The dorsal ridge defining the cuppedicus fossa is dorsally convex and extensive (fig. 27: dcfr), and it continues posteriorly to become confluent with the supraacetabular crest, as in Rahonavis (Forster et al., 1998), Unenlagia (Novas and Puerta, 1997), Velociraptor (Norell and Makovicky, 1997, 1999), and Adasaurus (IGM 100/20).

Pubis

At first glance, the most remarkable feature of the pubis, which is also true of the ischium, is the extreme degree of retroversion (fig. 27). With the iliac blade held horizontal for reference, the pubis and ischium are oriented at more than 55° from the vertical, and the ischium in particular appears to nearly parallel the long axis of the ilium. Although all dromaeosaurids and other basal paravians have a retroverted pubis, it is usually oriented at 10°–20° from vertical (e.g., Ostrom, 1969; Novas and Puerta, 1997; Forster et al., 1998; Longrich and Currie, 2009). Velociraptor exhibits the most extreme retroversion of any other dromaeosaurid, with a pubis backswept at a 35° angle (Norell and Makovicky, 1997, 1999), but this is still far less extreme than the condition in Balaur. Therefore, the pronounced retroversion of the pubis and ischium is considered an autapomorphy of Balaur, although some degree of increased retroversion may be shared by Balaur and Velociraptor, which are sister taxa, due to common ancestry. However, we consider this autapomorphy only tentative, because it is likely that the extreme degree of retroversion in Balaur is at least partially due to dorsoventral crushing. That said, we do not completely dismiss this character as artifactual. Two lines of evidence suggest that at least some degree of greater-than-average retroversion is present: (1) the pubis is still extremely retroverted on the less deformed right side; and (2) the shape of the left ischium (also highly retroverted) appears to be less affected by crushing.

The more complete right pubis is 146 mm long as preserved, but the posterior margin is eroded, and comparison with close relatives suggests that several centimeters are missing. Therefore, based on these comparisons, the pubis would probably have been at least 170 mm long in life (table 2). Distal to the acetabular region the pubis tapers into an elongate, thin shaft. At its midsection the shaft is ovoid in cross section and measures 11 mm in mediolateral width by 9 mm in dorsoventral depth. The shaft is strongly curved when seen in dorsal or ventral view, with a convex lateral margin and a concave medial surface. The two pubes eventually meet on the midline and together define a wide region between them. Anteriorly, there is a pronounced ridge on the dorsal surface of the shaft, which dissipates posteriorly as the bone becomes wider mediolaterally and thinner dorsoventrally (fig. 28: arid). Farther distally, approximately 20 mm before the two pubes meet, another ridge emerges (fig. 28: prid), which becomes more prominent as it continues distally and forms the lateral bounding wall of the dorsally concave pubic apron (fig. 28: pa; fig. 31).

Only the anterior portion of the pubic apron is preserved, and this region corresponds well with the morphology of Velociraptor (Norell and Makovicky, 1997, 1999) and Adasaurus (IGM 100/20). The apron is separated from the shaft by a distinct change of curvature, and its ventral surface is nearly horizontal. Further distally, however, the ventral surface curves upward such that it also faces posteriorly, and based on the shape of the broken posterior surface it is likely that the distalmost portion of the ventral pubis would have faced almost entirely posteriorly. The apron is wider than the remainder of the pubis and the actual contact between the pubes is made by two discrete ridges of bone. The medial surface of the shaft is marked by a thin ridge, which begins immediately posterior to the acetabulum and becomes thinner as it continues distally, before meeting its counterpart to form the apron (fig. 28: mdrid). This ridge is centered on the medial surface proximally, but migrates further dorsally as it extends distally. On the ventral surface, there is a ridge that begins at approximately midshaft, becomes more prominent distally, and eventually curves medially and migrates dorsally to join with the medial ridge forming the apron (fig. 28: mvrid).

The space between the pubes is proportionally wider than in close relatives such as Velociraptor (IGM 100/985, 986; Norell and Makovicky, 1997, 1999), and is indicative of an autapomorphically wider-barreled abdominal and pelvic region (somewhat similar to the condition in therizinosauroids). The greatest interpubic width occurs slightly posterior to the acetabulum, where the two bones are separated by 72 mm. The intervening space remains wide for a considerable distance distally, in concert with the strongly laterally bowed pubes. Because each pubis is bowed to such a pronounced degree, the interpubic distance only begins to narrow at approximately 2/3 of the length of the bone. The narrowing occurs abruptly once it begins, unlike the condition in Velociraptor where the narrowing between the two pubes is more gradual over a long distance (IGM 100/986; Norell and Makovicky, 1999: fig. 17).

Csiki et al. (2010a) described the pubic tubercle as located directly below the acetabulum, as a result of the supposed rotation of the pubis. As described above, this rotation is entirely due to postmortem crushing and does not reflect original morphology. The structure Csiki et al. (2010a) originally identified as the pubic tubercle is in fact a broken and swollen portion of the posterior bounding ridge of the cuppedicus fossa on the pubic peduncle of the ilium, proximal to the pubis itself. Reexamination of the specimen indicates that the true pubic tubercle is a subtle structure, visible only in ventral view due to crushing, whose size, shape, and position are similar to that in Velociraptor (IGM 100/986).

Ischium

The ischium is nearly complete on the left side, and only a small portion is missing posteriorly (table 2) (fig. 27). It is 70 mm long as preserved and 15 mm deep dorsoventrally near its midpoint. Although both ischia are broken posteriorly, it can be inferred that they did not make broad contact at a symphysis, as there is no room for such a contact due to the intervening caudal vertebrae. This is also the case in most other maniraptorans, whereas in more basal theropods the ischia meet each other across a broad contact to form a symphysis (e.g., Madsen, 1976; Xu et al., 2006; Brusatte et al., 2008).

The long axis of the ischium parallels that of the pubis and the two are separated by a 12 mm gap at the midpoint of the ischium. The ischium is dorsoventrally deep proximally, where it forms the posteroventral corner of the acetabulum, due to dorsal and ventral expansions. The ventral expansion is formed by a greatly enlarged ischial tubercle, whose size and morphology are autapomorphic for Balaur (figs. 27, 30: it). A tubercle is present in Deinonychus (Hutchinson, 2001) and Velociraptor (Norell and Makovicky, 1997, 1999), and is a synapomorphy that unites Balaur and these genera as a dromaeosaurid subclade. However, in these other taxa, the tuber is expressed as a small, discrete, rugose bulge (e.g., Norell and Makovicky, 1997, 1999). In Balaur, in contrast, it takes the form of a large, thin flange that projects ventrally and nearly contacts the pubis. Since the ventral surface of the tuber is broken, it is possible that this flange actually did contact the pubis in life, which would comprise a further autapomorphy of Balaur. The dorsal expansion of the proximal region is formed by a separate flange, which is thin and concave laterally (fig. 27: df). It is not present in Velociraptor, but is seen in an undescribed dromaeosaurid (Norell and Makovicky, 1999: fig. 24; Montanari and Norell, in prep.). Between the dorsal and ventral expansions, the lateral surface of the ischium is deeply convex and smooth.

In dorsal view, the lateral margin of the ischium has a sinusoidal shape: it is laterally convex anteriorly, then concave, and then sweeps out again posteriorly. The posterior expansion is due to a reorientation of the lateral surface, which here curves ventrally such that it now faces both ventrally and laterally. A similar, although less striking, twisting of the lateral surface is present in Velociraptor (IGM 100/986). The middle and posterior region of the ischium has a strongly rugose lateral surface, which is bisected by a crest that begins on the dorsal surface of the ischium near its midpoint, migrates posteroventrally up to a sharp inflection point, and then extends posteriorly along a nearly horizontal trend (figs. 27, 30: ic). The crest is most prominent at the inflection point, which forms a discrete corner. This ridge is present in many, but not all, dromaeosaurids, including Deinonychus and Velociraptor (Norell and Makovicky, 1997, 1999; Turner et al., 2007a), Buitreraptor (Makovicky et al., 2005), and Sinornithosaurus (Xu et al., 1999).

Distally, the ischium thins such that its posterior preserved edge is only 1.5 mm thick. There is no clear obturator process along the ventral margin, which is straight anteriorly and changes to a posteroventral orientation posteriorly at an inflection point. However, it is possible that the obturator process was present but located far distally on the bone, and thus is missing due to erosion. Along the dorsal margin there is clearly no median posterior process, which is present in Rahonavis (Forster et al., 1998) and several other basal dromaeosaurids (Turner et al., 2007a).

Tibiotarsus

Although fused, the tibia and fibula are recognizable as distinct bones proximally, but distally the tibia, fibula, and proximal tarsals are indistinguishably fused and sutures between them are impossible to distinguish. The tibiotarsus has a straight shaft (figs. 32–33). In contrast, the shaft is somewhat curved, due to a concave medial margin, in Velociraptor (IGM 100/986; Norell and Makovicky, 1999). Comparisons between the hind-limb ratios of Balaur and other dromaeosaurids are provided in table 6 (see also figs. 45–46).

The proximal surface of the tibia is heavily eroded, but much of the cnemial crest is preserved (fig. 33: cn). The crest is prominent, 10 mm wide mediolaterally by 10 mm long anteroposteriorly as preserved, and curves laterally as it continues anteriorly, as is normal for dinosaurs (Gauthier, 1986; Novas, 1996; Brusatte et al., 2010; Nesbitt, 2011). The proximal surface of the crest slopes laterally, and as a result the medial region of the crest is higher than the lateral region, suggesting that the lateral condyle of the femur extended further distally than the medial condyle. Distally, the cnemial crest gradually merges with the shaft, and in doing so twists medially such that it terminates along the anteromedial edge of the shaft. Posteriorly, the proximal surface separates into lateral and medial condyles, both of which are heavily eroded. The incisura tibialis, the notch between the cnemial crest and lateral condyle along the lateral surface of the tibia, is visible as a concave margin in proximal view. This notch is moderate in size, 6 mm long anteroposteriorly and inset 2 mm into the lateral surface. Both the cnemial crest and lateral condyle project to the same level laterally, and as a result the incisura tibialis opens only laterally and is not visible in anterior or posterior views.

On the medial surface of the proximal end is a distinct rounded crest, which is thick and prominent before gradually merging with the shaft distally (fig. 33: cr). The crest begins at, or immediately ventral to, the proximal articular surface and curves slightly anteriorly, while expanding in width, to nearly merge with the anterior surface of the cnemial crest. Between this crest and the cnemial crest, and facing anteriorly and medially, is a smooth fossa, which dissipates distally as both crests merge with the shaft (fig. 33: fo). The medial crest and associated fossa are unusual features that also appear to be present, albeit much more subtly, in Deinonychus (Ostrom, 1969), Velociraptor (IGM 100/986), and Rahonavis (UA 8656), and may be present in other taxa (but perhaps sometimes are imperceptible due to the delicate nature of the crest or poor preservation in this area). The medial crest may be homologous at some level with the medial cnemial crest of alvarezsaurids, which is a unique character present in derived members of this clade (Perle et al., 1994; Turner et al., 2007a). However, the alvarezsaurid crest is positioned at the anteromedial corner of the cnemial crest itself, not on the medial surface of the tibia as in Balaur. Furthermore, basal alvarezsaurids lack a medial cnemial crest, as do other coelurosaurs, suggesting that the crests of derived alvarezsaurids and dromaeosaurids evolved independently (Choiniere et al., 2010).

The fibular crest is a rugose flange on the lateral surface of the proximal tibia, widely visible in anterior and posterior views (fig. 33: fc). It begins proximally immediately below the lateral condyle and extends anterodistally for 56 mm before abruptly terminating. The anterior surface of the crest is smoothly concave, and is essentially a ventral extension of the smooth incisura tibialis. This concavity faces solely laterally on the proximal portion of the fibular crest, but twists distally such that it terminates on the anterior surface of the tibial shaft. The medial surface of the shaft bordering the fibular crest is damaged, but a region of collapsed bone indicates that there may have been a fossa here, as in Velociraptor (IGM 100/986). Further distally on the shaft it is unclear how much contact there was between the fibula and the tibia, as the former bone is largely eroded in this area. It is clear that the fibula is separated and unfused to the tibia immediately distal to the fibular crest (up to about 2/3 of the length of the tibiotarsus), and then further distally it appears as if the two bones abut each other and fuse into a common element (with indistinguishable sutures).

The fibula is expanded proximally and tapers into a thin shaft distally, which is ovoid in cross section at its midpoint. The medial surface of the proximal end is smoothly concave and lacks the discrete, deep fossa present in more basal theropods, including basal coelurosaurs such as tyrannosauroids and ornithomimosaurs (e.g., Madsen, 1976; Brochu, 2003; Makovicky et al., 2004; Brusatte et al., 2008). Although the proximal region is somewhat eroded, it appears that the fibula was more expanded posteriorly than anteriorly. As seen in lateral view, the anterior margin of the shaft is approximately straight, whereas the posterior margin is concave and curves strongly posteriorly as it continues proximally, in concert with the more expanded posterior end of the proximal fibula. The iliofibularis tubercle is proximally positioned, beginning approximately 41 mm distal to the proximal end of the bone. It is expressed as a slight, convex bulge on the posterior surface of the shaft, not a large rugosity as in some basal coelurosaurs, especially tyrannosauroids (e.g., Brochu, 2003). The morphology of the distal fibula is not clear, as the tibia, fibula, and proximal tarsals are strongly fused here (fig. 34). There is a pronounced lateral tubercle on the distal end of the tibiotarsus (fig. 35: lt), which is in the same approximate position as the distal end of the fibula in Velociraptor (IGM 100/986) and therefore may represent the distalmost extent of the fibula.

The astragalus and calcaneum are fused to the tibia and fibula and their individual shapes are obscured due to this fusion (fig. 34). Complete fusion between the proximal tarsals and the tibia and fibula is rare among coelurosaurs, present otherwise only in Buitreraptor (Makovicky et al., 2005) and Graciliraptor (Xu and Wang, 2004) among dromaeosaurids (see Norell and Makovicky, 1997; Turner et al., 2007a). Because fusion is so extreme in this region, it is possible to describe only the various surfaces of the distal tibiotarsus without reference to the individual bones.

The distal surface of the tibiotarsus is saddle shaped, comprised of medial and lateral condyles that are separated by a wide and deep sulcus, which continues onto the posterior surface of the bone as a flexor groove. The anterior surface is deeply notched at its center as an extensor groove, which is visible in distal view as an inset, concave margin. This groove is inset approximately 8 mm into the bone as preserved, but was likely deeper since the anterior surfaces of the lateral and medial condyles, which define its anterior extent, are eroded. Although deep, the groove is only 5 mm wide mediolaterally. The prominent and deep extensor groove is an unusual feature among dromaeosaurids, present otherwise in Buitreraptor (Makovicky et al., 2005). Other dromaeosaurids, including Velociraptor (IGM 100/986; Makovicky and Norell, 1999) have a much shallower groove, which is considerably wider than deep. A deep groove is present, however, in some alvarezsaurids, troodontids, and basal avialans (Turner et al., 2007a), and therefore is homoplastic within coelurosaurs.

Approximately 10 mm proximal to the distal end of the bone, the lateral surface of the tibiotarsus is marked by a pronounced, curved ridge, which is oriented anteroposteriorly (fig. 35: lrid). The ridge is most pronounced anteriorly, where it terminates at a discrete rounded tubercle (fig. 35: lt) located at the point where the lateral condyle merges with the shaft of the tibiotarsus. This ridge may represent the line of fusion between the calcaneum and tibia, as its shape and position approximates the trace of this suture in many theropods (e.g., Velociraptor: IGM 100/986). If this is the case, then the fibula would contact these bones on, or near, the discrete tubercle, suggesting that the tubercle itself may represent the distal end of the fibula. Ventral to the ridge, the lateral surface of the tibiotarsus is smoothly concave. However, the ridge is kinked at its midpoint (fig. 35: vk), forming another small tuber that projects ventrally. Additionally, there is a sharp ridge along the anterior margin of the lateral surface, which forms the anterolateral corner of the bone (fig. 35: arid). This ridge begins at the proximal tip of the large tuberosity and extends proximally until it terminates against a broken surface. It is most pronounced distally and becomes less prominent and thinner as it continues proximally. It is likely that this crest represents the portion of the fibula that tapers distally before slightly expanding to articulate with the tarsus.

The medial surface of the distal tibiotarsus (presumably formed by the medial condyle of the astragalus) is marked by a deep, circular pit (fig. 35: mp). This feature is generally prevalent among dromaeosaurids, as it is present in Velociraptor (IGM 100/986), Deinonychus (YPM 5226), Neuquenraptor (MCF PVPH 77), Microraptor (IVPP V13352), and Mahakala (IGM 100/1013). The presence or absence of this feature in Adasaurus cannot be determined because the relevant portion of the astragalus is not sufficiently prepared. The especially deep morphology of the pit of Balaur may be unusual among dromaeosaurids, but the depth of this depression is variable in other taxa, and is sometimes even variable between left and right sides in individual skeletons (e.g., Velociraptor IGM 100/986).

Tarsometatarsus

The distal tarsals form a fused cap over the metatarsals, but individual tarsals are not apparent and it is difficult to determine how many tarsals were present. Fusion between the distal tarsals and metatarsals, as well as between the proximal ends of the metatarsals, is rare among nonavialan coelurosaurs. It is present in the oviraptorosaurs Elmisaurus (Osmólska, 1981) and Avimimus (Vickers-Rich et al., 2002), and the alvarezsaurid Albinykus (Nesbitt et al., 2011); in dromaeosaurids it is also present in Velociraptor (Norell and Makovicky, 1997, 1999), Adasaurus (IGM 100/21), and Microraptor (Hwang et al., 2002), albeit to a much less extensive degree, as in these taxa the sutures between elements of the tarsometatarsus are often visible. In Balaur, only the line of fusion between the tarsal cap and the metatarsus is visible, and it is clear that the tarsals are thickened proximodistally. The proximal surface of the tarsus is convex and smooth (fig. 36–37).

Metatarsals II–V are heavily fused to each other proximally and sutures between individual bones are not clear (figs. 36–37: fus). The fused region is approximately 8 mm in proximodistal length, and further distally the individual metatarsals separate from each other. However, it is possible that fusion continued distally, as the individual metatarsals are in close contact throughout their lengths. Furthermore, in both the left and right feet the regions between metatarsals II–IV are poorly preserved. Each metatarsal thins dramatically in dorsoventral (extensor-flexor) depth toward its lateral and medial margins, and these surfaces are largely broken. Therefore, it is possible that the broken surfaces may have contacted each other in life. Regardless, it is clear that the metatarsals were separated at their extreme distal ends, as the individual distal articular surfaces are unfused and diverge from each other.

The three central metatarsals are closely bunched to form a stout, paddlelike metatarsus. The metatarsus is 62 mm long proximodistally at its longest point, along metatarsal III, and 32 mm in mediolateral width proximally. Therefore, the metatarsus is less than twice as long as it is wide, contrasting with dromaeosaurids such as Velociraptor, in which the central metatarsus is more than three times longer than wide (IGM 100/985). The short and wide nature of the metatarsus is also reflected in the fact that it is more than 1.5 times as wide as the distal end of the tibiotarsus, a remarkable condition compared to other dromaeosaurids in which these two elements are approximately coequal in width (e.g., Ostrom, 1969; Norell and Makovicky, 1997, 1999; Xu et al., 1999; Hwang et al., 2002). In sum, the metatarsus of Balaur is autapomorphically shorter and wider compared to not only other dromaeosaurids, but all other theropods.

Although the remainder of the metatarsus and pes is heavily modified, metatarsal I is similar in position and morphology to the corresponding bone in other dromaeosaurids (fig. 38). As in most other theropods, including all other dromaeosaurids, this bone is a small, wedgelike element with a flat lateral surface that articulates along the midpoint of the medial surface of metatarsal II (e.g., Ostrom,1969; Forster et al., 1998; Norell and Makovicky, 1997, 1999; Hwang et al., 2002; Novas and Pol, 2005). Although metatarsal I supports large and functional phalanges (see below; fig. 36), it does not reach the tarsus, as is the case in derived therizinosauroids that independently evolved a large pedal digit I (e.g., Perle, 1979; Barsbold and Perle, 1980). The enlarged nature of the first digit of Balaur is reflected in the size of metatarsal I, however, which although still smaller than the other metatarsals is proportionally larger than in all other dromaeosaurids (e.g., Deinonychus: MCZ 4371, Ostrom, 1976; Velociraptor: IGM 100/986, Norell and Makovicky, 1999). Indeed, while in these velociraptorines the length of metatarsal I varies between 25%–28% of that of metatarsal III, in Balaur it reaches 60% (table 6; see also fig. 47). This ratio is well below 25% in other derived dromaeosaurids and drops below 20% in the microraptorines and unenlagiines (table 6; see also fig. 47).

One of the most prominent features of metatarsal I is a pronounced ridge along the lateral margin of the extensor surface, which helps form the articulation with metatarsal II (fig. 38: rid). A very subtle ridge is present in one specimen of Velociraptor (IGM 100/986), but absent in others. Adasaurus (IGM 100/21) also has a slight ridge in this region, but in this taxon and in the specimen of Velociraptor the ridge is much smaller than that in Balaur. Therefore, the pronounced ridge may be an autapomorphy of Balaur, but we hesitate to regard it as a formal autapomorphy because metacarpal I is not preserved for most other dromaeosaurid species.

Metatarsal II is dorsoventrally (extensor-flexor direction) thickened at its proximal end and remains thick distally along its medial margin. Its lateral margin, on the other hand, thins into a platelike flange that may have made contact with metatarsal III (see above). As a result of this morphology, there appears to be a strong ridge, oriented proximodistally, along the medial edge of the bone in extensor (dorsal) view (fig. 37: rid). This ridge is not a discrete structure, but rather reflects the difference between the thick medial edge and thin lateral half of the bone. Similar ridges are present on metatarsals III and IV, also a result of the thinning of the metatarsals toward one or both edges, and are autapomorphic of Balaur. In contrast, the metatarsals of other dromaeosaurids, including Deinonychus (Ostrom, 1969), Microraptor (Hwang et al., 2002), Rahonavis (Forster et al., 1998), Unenlagia (Novas and Pol, 2005), and Velociraptor (IGM 100/985) do not thin toward their edges, and hence do not appear to possess large ridges on their extensor surfaces. Velociraptor has an oblique ridge on metatarsal III, but this does not extend across the length of the shaft and is not associated with thin edges of the bone, nor with similar ridges on metatarsals II and IV (Norell and Makovicky, 1999). Furthermore, in some dromaeosaurids, including Velociraptor (IGM 100/986) and Adasaurus (IGM 100/21), a tubercle is present on the extensor surface just proximal to midshaft that likely corresponds to the insertion of M. tibialis cranialis. No such tubercle is present in Balaur.

On the extensor surface of the distal end is an enormous extensor pit that is expressed as a broad, deep fossa, one of several features that enables hyperextension of the second pedal digit. It is not a discrete pit, with distinct edges, as is usual for theropods, but rather occupies the entire extensor surface of the distal end. It is most deeply inset distally, where it is invaginated underneath the distal condyles, which extend onto the extensor surface to slightly overhang the pit. The anterior margin of the pit is smoothly convex and its lateral and medial edges are formed by thin ridges that continue proximally from the corner of each distal condyle. The flexor (ventral) surface of metatarsal II is collapsed proximally on the right metatarsus in which this surface is visible, indicating that a fossa was present in this region. Further distally, around midlength, the flexor surface flattens out, then becomes slightly convex, and eventually is concave as a small flexor cleft where it terminates against the distal condyles. The lateral ligament pit is much larger and deeper than the medial pit, which is reduced to a small, circular foramen.

The distal articular region is large and convex, and divides into separate condyles on the flexor surface. The condyles are not apparent as discrete structures in extensor view, and as a result, metatarsal II is not ginglymoid as in all other dromaeosaurids (e.g., Norell and Makovicky, 1997, 1999; Turner et al., 2007a, 2012). This morphology is indicative of less extreme hyperextensive abilities compared to close relatives, but several troodontids also are capable of hypextension without a ginglymoid metatarsal II (e.g., Xu et al., 2002; Turner et al., 2007a). The distal condyles are, however, separated on the flexor surface of the metatarsal. Here, the medial condyle, which is nearly circular, projects further ventrally (toward the flexor direction) than the lateral condyle, which is ovoid, larger than the medial condyle, and extends farther proximally and distally. The long axis of both condyles projects mediodistally, and as a result, digit II curves medially away from digit III, as is normal for dromaeosaurids (e.g., Ostrom, 1969; Norell and Makovicky, 1997, 1999). Unusually, the width of the distal articular surface is less than the width of the entire distal end of the metatarsal. This is especially evident in extensor view, where a large nonarticular region is present both lateral and medial to the smooth articular surface. This is not the case in other dromaeosaurids, including Deinonychus (Ostrom, 1969) and Velociraptor (Norell and Makovicky, 1997, 1999), and therefore is an autapomorphy of Balaur.

Metatarsal III has a pronounced ridge extending proximodistally at the center of the extensor surface, as a result of the lateral and medial thinning of the bone as described above. The ridge is a single structure across most of its length, but proximally it divides into separate ridges that are oblique to each other and distally, at approximately 2/3 of the length of the shaft, it bifurcates into two ridges that eventually merge with the lateral and medial edges of the bone. The ridge is most pronounced at the center of the shaft, and lateral and medial to it are smooth fossae that slope away from the long axis of the bone. Distally, the portion of the extensor surface between the bifurcated ridges is depressed as a shallow extensor fossa, which is deepest distally adjacent to the distal articular surface. As in metatarsal II, the fossa is invaginated due to the overhanging distal articular surface, and the articular surface itself is slightly narrower than the total mediolateral width of the distal end. Both lateral and medial ligament pits are large and deep, and positioned dorsally (toward the extensor surface) such that they abut the smooth articular surface.

The distal articular surface is not ginglymoid, which along with the same condition in metatarsal II is a unique feature among dromaeosaurids (see above; Turner et al., 2007a). However, the nonginglymoid condition of metatarsal III is more extreme than that in metatarsal II, as separate condyles are not apparent in both extensor and distal views. The articular region divides into medial and lateral condyles on the flexor surface, but these are much smaller than the corresponding condyles in metatarsal II and are only visible in flexor view. The condyles are symmetrical, extend to the same level distally, have long axes that are oriented straight proximodistally, and continue proximally as thin ridges which merge with the lateral and medial edges of the bone. They are separated by a narrow flexor cleft. Further proximally, the flexor surface flattens out and then eventually becomes concave adjacent to the proximal articular surface. The flat portion is bisected by a subtle, proximodistally extending ridge, which abruptly ends distally against the flexor cleft. Distally, the distal articular end of metatarsal II abuts the medial edge of the shaft of metatarsal III, and this contact is reinforced by a small, convex tuberosity on metatarsal III.

Metatarsal IV is approximately the same length as metatarsal II, and both are slightly shorter than metatarsal III (figs. 36–37). As in the other central metatarsals, the extensor surface is ornamented with a pronounced ridge. Here, the ridge begins proximally at the center of the bone and trends laterally as it continues distally, such that it extends along the lateral margin of the shaft. Shallow fossae border the ridge both medially and laterally, the latter of which is small and narrows in width distally in concert with the lateral migration of the ridge. There is also a thin ridge on the proximal portion of the lateral surface, which dissipates distally. The flexor surface is concave proximally and then flattens distally. Unfortunately, the distal articular region is broken in both specimens, precluding observation of the condyles and extensor pit.

Metatarsal V is a small bone that resembles a stout hook (figs. 36–37). It is vestigial as in most other theropods and does not support any phalanges. It does exhibit a unique shape among dromaeosaurids, however, as all other members of this clade possess a longer, straighter, and more gracile metatarsal V (e.g., Ostrom, 1969; Norell and Makovicky, 1997, 1999). The hooklike morphology of this bone is the result of a concave extensor margin and convex flexor edge. Moreover, metatarsal V of Balaur is uniquely shortened among dromaeosaurids, at less than 30% of metatarsal III length, whereas this ratio exceeds 40% in all derived dromaeosaurids and approaches 50% in the more basal microraptorines (table 6; fig. 47). In articulation, it is positioned only slightly lateral to metatarsal IV and is displaced ventrally (toward the flexor direction) relative to the level of the other metatarsals, and as a result there is a deep notch between metatarsals IV and V. The lateral surface of the bone is smooth and the muscle attachment sites that are so apparent in many large theropods (e.g., Currie and Zhao, 1993; Brusatte et al., 2008) are not present. There does, however, appear to be a slightly convex tuber on the distal portion of the flexor margin.

Pedal Digits

The digits are complete and well preserved on the left foot (figs. 32, 26). The phalangeal formula is 2-3-4-5-x and most remarkably, digit I is large and functional. Both digit I and II are modified for hyperextension and possess large unguals, and both are preserved in the stereotypical flexed position seen in the fossils of many deinonychosaurs. Digit III is the longest in the foot, although digit IV is nearly as long.

The most unusual feature of the foot is the large digit I, which, unlike in all other dromaeosaurids and most other theropod dinosaurs, was fully functional and faced forward in concert with the remaining digits (figs. 32, 36). The large size of the first digit is due to enlarged individual phalanges, which are approximately the same size as the phalanges of digits II–IV. In particular, the proximal phalanx of digit I is the longest nonungual phalanx in the foot and the ungual is nearly as large as the ungual of digit II (exceeding 80%). This departs remarkably from the morphology of other dromaeosaurids, best exemplified by Velociraptor, in which the nonungual phalanx of digit I is approximately 30% of the length of the longest nonungual phalanx in the foot and the ungual is approximately 31% of the length of the ungual of digit II (IGM 100/985; Norell and Makovicky, 1997, 1999) (table 7). Within dromaeosaurids, only Adasaurus (IGM 100/21) shows a similar I2/II3 ratio to that of Balaur, but this is due to the autapomorphically reduced sickle claw of digit II instead of the enlargment of the ungual of digit I. Furthermore, digit I of Balaur exhibits the same hyperextensive capabilities of digit II, most notably a robust extensor lip overhanging the dorsal (extensor) margin of the proximal surface of the ungual. The proximal phalanx has two large distal condyles that are separated by a deep flexor groove and the ungual has a small, but discrete, flexor tubercle. The ungual is slightly less recurved than that of the second digit, but more recurved than the claws on digits III and IV, and has a single groove on its lateral and medial surfaces.

Digit II is modified for hyperextension, as in other dromaeosaurids and troodontids (Ostrom, 1969; Norell and Makovicky, 1997, 1999; Manning et al., 2006). Its ungual is the largest and most recurved in the foot, and the distal articular surfaces of individual phalanges (especially the penultimate phalanx) are ginglymoid, with enlarged and highly convex distal condyles that extend onto the extensor surface of the bone. Lateral and medial ligament pits are deep on the nonungual phalanges, which also possess deep extensor pits. The ungual has a robust flexor tubercle, which is larger and more rugose than the tubercles of the remaining unguals, and a prominent extensor lip dorsally. Compared with Deinonychus and Velociraptor, however, the second pedal digit of Balaur is less modified for hyperextension, as the ungual is less strongly recurved and possesses a smaller and less rugose flexor tubercle (Ostrom, 1969; Norell and Makovicky, 1997: fig. 6).

Digit III is the primary weight-bearing digit of the foot. The nonungual phalanges have extensor pits, but these are not as deep as in the hyperextendable second digit. Where observable, the distal condyles are not ginglymoid and do not extend for a great distance onto the extensor surface of the bone. Lateral and medial ligament pits are large and deep where they are exposed. The ungual is broken distally but is clearly not as large as the unguals of the first two digits. The flexor tubercle is essentially absent and the extensor lip is reduced. Notably, the distal segment of digit III is elongated relative to the proximal segment or to metatarsal III. Both phalanx III-3 and, especially, ungual III-4 are about 1.5–2× as long relative to the length of metatarsal III than in most other derived dromaeosaurids, as well as in the basal microraptorines and unenlagiines (table 7).

Digit IV is similar to digit III in the morphology of individual phalanges. Extensor pits are present, but shallow, on the nonungual phalanges, and the distal articular surface extends only slightly onto the extensor surface of the bone. Some individual phalanges are ginglymoid, but the extensor groove between the distal condyles is shallow. The ungual is broken distally but appears to be the smallest and straightest ungual in the foot. It has a single groove on both the lateral and medial surfaces, as do the other unguals, but unfortunately its extensor lip and flexor tubercle cannot be observed due to breakage and concealing matrix, respectively.

Holotype Specimen

The fibula of EME PV.313 shows dense Haverisan remodeling thoughout the inner two-thirds of the cortex (fig. 40A). The endosteal surface is lined with lamellar bone (fig. 40A). Because of the extensive fungal damage, remodeling, and the hollow nature of the bone, the primary cortical bone types deposited early in ontogeny were indeterminable. The outer third of the cortex is composed of parallel-fibered matrix with sparse longitudinal vascularization. The matrix is interrupted by lines of arrested growth (LAGs). At least seven are present and their spacing diminishes nearer the periosteal surface. The deepest LAGs show negligible folding, whereas the outermost ones follow the contours of the external bone sculpturing manifest at the perisoteal surface (fig. 40A).

The results from the LAG count suggest that EME PV.313 was no less than seven years of age at the time of death. Based on the distance from the innermost preserved LAG to the centroid of the section, and the width of the deepest zone, it was likely several years older. The presence of dense Haversian bone, diminished spacing between LAGs near the periosteal surface, parallel-fibered bone deposited just prior to death, and negligible vascularization suggest that this individual is a skeletally mature individual (i.e., nearly full grown at the time of death). The extensive folding seen only in the outermost LAGs shows that the peculiar scultured texture of B. bondoc long bones is not a result of diagenesis or taphonomic alteration (e.g., chemical or biological erosion), but rather is a natural osteogenic developmental feature, one that manifested itself very late in the life of EME PV.313. Whether this feature is a nondebilitating or debilitating product of senescence, a response to disease, or even a functional attribute is indeterminable at this time, but is worthy of further study as new specimens are found and comparative histological datasets (of both extinct and extant taxa) are expanded.

Referred Specimen

The ulna sample from LPB (FGGUB) R.1581 shows dense Haverisan remodeling thoughout the inner third of the cortex (fig. 40B). Endosteal bone is absent. Despite extensive fungal damage, remodeling, and the hollow medullary cavity, a small section of primary bone is locally preserved deep in the cortex (fig. 40C). This shows fibro-lamellar matrix with circumferential vascularization. The outer two-thirds of the cortex is composed of parallel-fibered matrix with sparse longitudinal vascularization (fig. 40B). The matrix is interrupted by LAGs. There are at least six of these and their spacing substantially diminishes nearer the periosteal surface. As with the fibula from the holotype, the deepest LAGs show negligible folding whereas the outermost ones follow the sculpturing seen at the perisoteal surface (fig. 40B, D)

The results from the LAG count suggest that LPB (FGGUB) R.1581 was no less than six years of age at the time of death. However, based on the distance from the innermost preserved LAG to the centroid, and the sample's distance from midshaft, this individual is likely several years older. The presence of dense Haversian bone and the attenuated spacing of the outermost LAGs suggest that this individual is (like the B. bondoc holotype EME PV.313) skeletally mature (i.e., full or nearly full grown at the time of death). The extensive folding of only the outermost LAGs, as was also seen in EME PV.313 (see above), suggests that the peculiar sculptured texture of Balaur long bones is not a result of diagenesis. It is a natural late developmental osteogenic feature present on both specimens, which as discussed by Csiki et al. (2010a) are of different body size and were discovered at different localities that may be slightly different in geological age.

Romanian Specimens

Prior to the discovery of Balaur, the best-known, best-preserved, and most diagnostic theropod taxon from the Late Cretaceous of Europe was Elopteyx nopcsai, a small theropod named by Andrews (1913) based on two proximal femoral fragments (and possibly a distal tibiotarsus) from the Sânpetru Formation of the Haţeg Basin of Transylvania, Romania. Subsequently, various authors assigned additional material to this taxon, which usually could not be compared directly with the fragmentary type material (e.g., Lambrecht, 1929; Grigorescu and Kessler, 1981). Later workers also disagreed on the taxonomic affinities of both the Elopteyx type and referred material, with some authors suggesting derived avian affinities (e.g., Lambrecht, 1929; Harrison and Walker, 1975) and others arguing that some, or all, of this material belonged to small nonavian coelurosaurian theropods (e.g., Martin, 1983; Grigorescu, 1984; Le Loeuff et al., 1992; Csiki and Grigorescu, 1998; Naish and Dyke, 2004; Kessler et al., 2005; Weishampel and Jianu, 2011).

A more detailed account of the tortuous taxonomic history of Elopteryx is provided by Csiki et al. (2010a) and will not be repeated here. For the sake of this paper, however, it is important to point out that the holotype of Balaur cannot be compared to the holotype of Elopteyx because there is no overlapping material between the two specimens. It may be noteworthy that both Balaur (holotype and additional referred specimens) and the specimens referred to Elopteryx, as well as other isolated bones from the Late Cretaceous of Romania, possess a similar woven and rugose external bone texture (Andrews, 1913; Le Loeuff et al., 1992; Kessler et al., 2005; Csiki et al., 2010a). This may turn out to be an autapomorphy of a single species (which includes the holotypes of both Balaur and Elopteryx), but it could also represent a synapomorphy of a larger group, or perhaps an environmentally controlled feature unique to theropods living on the Late Cretaceous European archipelago. Therefore, we do not currently consider the presence of this single character as strong evidence for the synonymy of Balaur and Elopteyx.

A number of Romanian specimens besides the holotype of Elopteryx are preserved in collections of the Natural History Museum in London. Locality information for much of this material is uncertain (many are simply labeled “Transylvania”), and some of these are femora that, because femora are not preserved in the Balaur holotype, disallow direct comparison with Balaur. However, it is interesting to note that all these bones preserve the very unusual surface texture seen in both Balaur (holotype and referred specimens) and the holotype of Elopteryx. Although most of these specimens are referred to Elopteryx (both in print and in the museum's catalog) there is significant size and morphological differences among them. This suggests that there is probably more than one theropod taxon represented in this collection and that the unusual surface texture of the bones may be present on all, or many, of these small, island-dwelling theropods.

Additional theropod material from the Late Cretaceous of Romania includes isolated teeth and fragmentary cranial and postcranial elements. Various teeth have been referred to dromaeosaurids, troodontids, and enigmatic tooth genera such as Richardoestesia, Paronychodon, and Euronychodon (Grigorescu, 1984; Grigorescu et al., 1985; Csiki and Grigorescu, 1998; Codrea et al., 2002; Smith et al., 2002; Vasile, 2008). Unfortunately, these teeth cannot be compared to the dentition of the Balaur holotype, which lacks the entire skull and dentition. Similarly, a set of putative dromaeosaurid frontals, which were described by Weishampel and Jianu (1996) and identified as belonging to a Saurornitholestes-like taxon, cannot be compared to Balaur. Finally, a sacrum mentioned in passing by Nopcsa (1915, 1923) and described in detail by Ősi and Főzy (2007) may represent an animal closely related to Balaur, but it is difficult to compare the two specimens (because Nopcsa's specimen is fragmentary), and our comparisons indicate that they do not share any unique derived characters. We find it likely that many of the isolated theropod remains from the Late Cretaceous of Romania do indeed belong to the same taxon as Balaur, but this is often difficult to test due to the lack of overlapping material, poor preservation, and the lack of clear synapomorphies.

Hungarian Specimens

Along with isolated teeth that cannot be directly compared with Balaur (Ősi, 2004), two peculiar specimens from the Santonian (middle Late Cretaceous) of Hungary belong to derived coelurosaurian theropods that may be close relatives of Balaur. First, Ősi et al. (2010a) named a new taxon, Pneumatoraptor fodori, based on a fused scapulocoracoid. Although Balaur also possesses fusion between the scapula and coracoid, these bones in the two taxa differ in detail. Most prominently, Balaur lacks the large pneumatic foramen on the coracoid that is an autapomorphy of Pneumatoraptor, whereas Pneumatoraptor lacks the hypertrophied coracoid tubercle that is autapomorphic of Balaur. It is also worth noting that other dromaeosaurids, such as Velociraptor and Microraptor, also possess fusion between the scapula and coracoid (Norell and Makovicky, 1999; Turner et al., 2007a), so it is possible that Pneumatoraptor may be a dromaeosaurid closely related to Balaur. Second, Ősi and Rabi (2006) and Ősi (2008) described a fused tarsometatarsus, which they referred to an enantiornithine bird. The various characters used to unite this specimen with enantiornithines are also present in Balaur, which may cast doubt on this taxonomic assignment. Regardless, it is clear that the Hungarian specimen is distinct from Balaur for several reasons. First, the tarsometatarsus is much shorter proximodistally relative to its mediolateral width in Balaur. Second, the metatarsals are more closely appressed along their shafts in Balaur. Third, the Hungarian specimen lacks the autapomorphic ridges on the extensor surface of metatarsals II–IV in Balaur. Fourth, and finally, the Hungarian specimen lacks a metatarsal V that is fused to metatarsal IV as in Balaur. Therefore, none of the Hungarian specimens can be assigned to Balaur, which is not surprising, given that these specimens are about 15–20 million years older.

French Specimens

Coelurosaurian theropod specimens have been reported from both the Cenomanian (earliest Late Cretaceous) and Campanian-Maastrichtian (latest Late Cretaceous) of France. Isolated teeth described by Vullo et al. (2007) from the Cenomanian cannot be compared directly to Balaur, nor can teeth from the Campanian-Maastrichtian (e.g., Buffetaut et al., 1986). However, two more complete and potentially diagnostic dromaeosaurid specimens from France can be compared to Balaur. The first, Pyroraptor olympius, is different from Balaur because it lacks the flat anterior surface of the ulna that is autapomorphic of Balaur. Furthermore, Balaur lacks the deep muscle attachment site on the lateral surface of the proximal ulna that Allain and Taquet (2000) described as autapomorphic of Pyroraptor. The second dromaeosaurid, Variraptor mechinorum, is also distinct from Balaur (Le Loeuff and Buffetaut, 1998). The humerus of Variraptor is more robust and straighter than that of Balaur, and it exhibits a prominent posteromedial tuber level with the distal end of the deltopectoral crest, which is absent in Balaur. Furthermore, the sacrum of Variraptor is marked by a ventral groove on only the fourth vertebra, whereas in Balaur these grooves are present on more anterior sacrals. Finally, it is worth noting that Balaur can be differentiated from two more fragmentary French specimens: a manual ungual recently described by Chantasit and Buffetaut (2009) lacks the autapomorphic Y-shaped lateral and medial grooves of Balaur, and a caudosacral vertebra reported by Le Loeuff et al. (1992) possesses a dorsolaterally oriented transverse process that is unlike the more laterally oriented process of Balaur.