Cretaceous theropods from India : A review of specimens described by Huene and Matley ( 1933 )

Abstraot: The Lato Cretaceous (Maas~richtian) Lameta Ihrmation of central India has yielded dissociated elements ofavariety ofpredatory dinosaurs, most ofthem comingfrom aquarry named the e~Carnosaur bed. The materials were described hv Huene and Matlev nearly 70 years ago. They recomized nine iheronod species, which they . " . . sorted out into the theropod subgroups 4!arnosaiirian and eaCoelurasauria~~, I-iuene and Matley also described a considerable amount of theropod hindlimb bones (e.g., femora, tibiae, metat,arsals, and pedal phalanges) that they oilld not refer to any of these species, hut vaguely interpreted as corresponding to ~~al losaurid~~ or ,~coelurosaurid,~ theropods. We reviewed the available collection of Cretaceous therapods from Bara Simla housed at the Geological Survey of India. Kolkata. avrivlne to the foliowine conclusions: 1) Indosuchus and Indosaurus are abelisaurids, as recognized by previous authors, hut avsilable information is not enough to judge whether they axire synonyms; 2) Lmt~oisuchus indicus is a small abclisauroid, related to Noasaurus and Masiahasaurus on the basis or their peculiar cervical vortcbrae: 3: the coiitroversial taxa *Clorn~sosuc~ius-. ~iDruolosauroidesn, ~Ornithornirnoides*. and

Reuista del Museo Argentino de Ciencias Naturales, n. s. 6 f l j , 2004 5a0 km Fig. 1.Map of India showing the Late Cretaceous (Maastrichtian) fossil site a t Jabalpur.
Ornithomimoides (?) barasimlensis were referred by Huene and Matley (1933) and later by Huene (1956) to the Ornithomimidae.But, most of the bones collected from the <<Carnosaur bed.(c.g., several skuil elements, t e e t h , vertebrae, liemapophyses, and pelvic and hindlimb bones) could not be referred to any of tho species listed above, and thus they were considered by Huene and Matley either as Allosauridae, Carnosauria or Coeiurosauria of indeterminable reidtionships.
Subsequent'taxonomic interpretations of the Indian theropods were seriously obstructed by the mixed condition of t h e numerous, diverse elements recovered from the '<Carnosaur bed.. Sorting out these skeletal elements into discrete specimens transformed into an impossible task, a problem that Huene and Matley and many other researchers (includingus) were unable to resolve.Moreover, original specimens were not described again in later studies (e.g., Romes 1956;Walker, 1964;Chatterjee, 1978;Molnar, 1990;Molnar et al., 1990;Norman, 1990) mainly because they were misplaced a t the collections of the Geological Survey of India, t h u s f u r t h e r delaying the elucidation of the phylogenetic relationships of the Indian theropods.
However, no major progress was made with regard to the Indian theropod postcranial material.I n two previous papers (Novas &  Bandyopadhyay, 1999, 20011, we studied the theropod pedal unguals of the Indian collection, recognizing their abelisaurid nature.This prompted some comprehensive ideas about the systematics of the Indian theropods as a whole.Probahiy the most relevant interpretations of our si~rvey were that Laeuisuchus is a n ahelisauroid Our primary goal is to recognize the presence diagnosable on the basis of its cervical vertebrae, of aheiisauroid features in the available GSI and that the controversial Laxa Compsosuchus, theropod collection, and secondarily to identify, if Dryptosauroides, Ornithomimoides, and possible, derived characters distinctive of the Jubbulpuria are based on isolated vertebrae Indian forms, To carry on this task, we have correspondingto different portions of the neckand studied mosL of the bones belonging to the GSI tail, which also exhibit abelisauroid featnres.that were described and illustrated by Huene and Through this "taxonomic purification", the Matley (19331, as well a s hones referred to phylogenetic relationships of the Indian theropods Indosuchus rapiorius housed at the AMNH (see can be addressedina clearer context.In agreement the Appendix for a complete list of theropod with our interpretations, Carrano el al. (2002) specimens considered in the present paper).recognized derived fkatures uniting Laevisuchus Available theropod materials are re-described with other small sized ahelisauroids (Noasauridael.below according to major anatomical regions (e.g., But the "Csrnosaur bed" also yielded the skull and jaw, vertebral column, and hindlimb remains of Lametasaurus indicus, originally bones), labeling each specimen (mostly individual described by Charles Matley (1923) a s a n hones) under current anatomical terminology.armoured ornithischian (i.e., a stegosaur) and However, for the sake of clarity, some of the consequently restudied by Huene and Matley specimens are re-described keeping their original (1933) in their section on ornithischian dinosaurs.generic and specific names (e.g., Indosuchus However, Lametasaur-us was reinterpreted by raptorius, Indosaurus matleyi, Conzpsosuchus Chakravarti (1935) as a member of the Theropoda, solus, Laevisuchus indicus, Jubbulpuria tenuis, and more recently, Wilson et al. (2003) found Coeluroides largus, Dryptosauroides grandis, evidence supporting that Lametasaurus indicus Ornithomimoides mohilis, and Ornithornimoides is an abelisaurid theropod.Thus, the "Carnosaur (?) barasimlensis).Several other bones that were bed" yielded remains of three nominated taxa of vaguely referred hy Huene and Matley (1993) as large abelisauroids: Indosuchus, Indosaurus and pertaining to <tallosaurids>> or ~~coelurosaurids.>Lametasaurus.
a r e also reviewed.T h e reason for this Our current knowledge of Gondwanan categorization is because the above mentioned abelisauroids, based on associated specimens from impossibility in identifying discrete individuals Argentina (Abelisaurus, Carnotaurus, Aucusaurus, represented by more than a single hone.We follow Ilokelesia, Noasaurus, Velocisaurus, Ligabueino), recent papers (e.g., Bonaparte, 1991b;Novas, Madagascar (Majungatholus, Masiakasaurus), and 1992;Carrano et ul., 2002;Wilson et al., 20031 in the new theropod material from India (Indosuchus considering Abelisauroidea as the node including raptorius andRajasaurus narmadensis; Chatterjec Aheiisauridae plus Noasauridae. & Rudra, 1996;Wilsonet al., 2003), invites areview ofthe theropod specimens first described by Huetie SPECIMEN STUDY and Matley in lm3.The aim of the present paper is to offer more information and some new I. Theropod taxa based on skull bones illustrations of such old specimens, comparing them with other ahelisaurs with the aim to test Two theropod taxa Pall within this category: previous interpretations.
Indosuchus raptorius.This taxon was MATERIALS AND METHODS coined on the basis of three basicrania belonging to a large theropod (GSI K201350, GSI K271685, Abbreviations: AMNH, American Museum of and GSI K271690).From these specimens, GSI Natural History, New York; FMNH PR, Field K271685 was designated as lectotype of Museum of Natural History, Chicago; GSI, indosuchus raptoriw by Chatterjee (1978).Since Geological Survey of India, Kolkata; ISI, Indian these specimens are currently lost or misplaced Statistical Institute, Kolkata; MACN-CH, Museo in the GSI collections, our available data source Argentino de Ciencias Naturales "Bernardino is restricted to the information originally offered Rivadavia", Paleontologia devertebrados (Colec-by Huene and Matley (1933).Skull roofs and ciirn Chuhut), Buenos Aires; MCA, Museo <<Car-basicrania referred to a s Indosuchus exhibit 10s Arneghinon, Cipoletti; UNPSJB-PY, Univer-interesting resemhlances with abelisaurid sidad Nacional de la Patagonia "S.J. Bosco", theropods in the fronto-lacrimal suture, the la.. Comodoro Rivadavia, Chubut; MPM, Museo Pa-crimal bone, the parietal crest, the parasphenoid, dre Manuel Molina, Rio Galiegos, Santa Cruz, and the orbitosphenoid (Fig. 2).In specimen GSI  1964).Abelisaurid traits are present in the sagital parietal crest of specimen GSI K271685: in dorsal "iew, the rostral half of this crest has a cup-shaped contour, beirig transversely narrow towards the rear (Fig. 2F).This condition is only documented in abelisaurids among Theropoda (e.g., Carnotaurus sastrei, Abelisaurus comahiiensis, Majungatholus atopus).Notable for specimen GSI m71685 is the presence of an inter-orbital wall (presumably made up by the parasphenoid bone), which is vertically hanging below the mid-frontal suture (Fig. 3C).Such inter-orbital wall, also seen in specimen GSI KZ71565 of Indosaurus matleyi (Figs.2B, 3B) is almost identical to t h a t of ahelisaurids Aklisaurzis comaizuens~s (Bonaparte & Novas, 1985), Carnolaurus sastrei (MACN-CII 894), and Majz~ngatholus alopus (Sampson et al., 1998).Moreover, in GSi K271685 the cranial half of the parasphenoid ends in a diamond-shaped struclure (eventuaily the orbitosphonoid; Currie & Zhao,19Y3), which bears a double-foramen for the exit of t h c olEactory nerve.Ossified parasphenoids tightly fused to tho skull roof and with a double exit for nerve I, are features not exclusi\,e for Abelisauridae, since they are present also in Ceratosaurus (Madsen & !Are1les, 20001, Acrocanthosaurus (Stovall & bangston, 1950), and some tymnnosaurids (Russell, 1970).Albeit such conditions for the parasphenoid and orbitosphenoid may be not syriapomorphic ibr Abelisauridae, at least their presence in the Indian basicrania is congruent with other ahelisaurid features.Sonie differences among the Indian basicrania and other ahelisaurid taxa are discernable on the basis of the figures given by Huene and Matley (1933).For example, in specimen GSI K201350 the fronto-nasal suture appears to be rostrally placed with respect to thelacrimais (Fig. 2G), in contrast to t,he remaining abelisaurids in which such suture constitute autapomorphic features of indo.~uclzusraptorius.
Indosaurus matleyi.This species was founded by Huenc on the basis of a single piece of skull, catalogued with the number GSI K2'7/565.Chatterjee (1978) later declared this specimen to be the holotypf of indosaurus matleyi.The specimen consists oftlle posterior part of the skull, the dorsal surface of which is partially damaged and includes tho right frontal bone, the temporal region, and the area for articulatiori with the postorbital (Fig. 213, El.In dorsal view the frontal is suhtrianplar, with an ar~terolateral notch for articulation with the lacrimal.Tile dorsal surface is slightly rugose, although not to the degree seen in Abelisuurus (Bonaparte & Novas, 1985).In Patcral view (Fig. 2B) the articular surface for the postorbital and lacrimal bones is rugose, being dorsoventrally deep in the postorbital portion (roaching 5 cm thick), and becoming shallower rostrally (nearly 3 cm thick).The posterior surface of the frontal is high, except for the surface hounding the suprate~nporal fbssa, w-hich is excavated.Both dorsai and posterior surfjces of the frontal are separated by a sharp border AIL interorbital wall, vertically hanging below tho mid-frontal suture and presumably made up by tho parasphenoid, is a character that Inilosuurus matle.yi(GSI K271565) shares with Indosuchus raptorius (GSI K271685) and other ahelisaurids (see above; Fig. 3).Bonaparte and Novas (1985) found similarities between slbelisauriis comah.uensisand Indosaurus matlqi based on the broad interorbital region, and Molnar (1990) noted that thisIndian taxon resembles Carnotaurus sastrei in the massive frontals and supraoccipital and markedly elevated sagittal crests oftheparietals.Later, Bonaparte (1991b) pointedout that the supratemporal openings of Indosau,ras are is more caudally placed, approximately a t level of anteropos&riorly short, resembling~belisaurus and the rostrolateral notch of t h e frontals for C~rmtaurm.Theabovementioncdauthorsobviouslv articulation with the lacrimals.GSI K201350 also exhibits on i t s caudal half a niedian suture between both frontals, as well as a clear frontoparietal suture.The presence of visible dorsal sutures in GSI K201350 is in agreement with the lack of fusion wit11 the parasphenoidal bone, thus exposing the ventral furrow for the olfactory canal (Huene & Matley, 1933).Both frontals and parasphenoid are completely fused in Abelisaurus co~izahuensis and Carnotaurus sastrei, as well as in specimens GSI K27168.5 ofIndosuchus raptorius and GSI K271565 of I n d o s a u r z ~s matleyi.Presumably the lack of ossification among the skull roof and braincase bones may be duo to concluded t h a t Indosaurus is a member of Abelisauridae, an interpretation also followed by Cllatterjee and Rudra (1996).
Indosuchus and Indosaurus shows some distinctions with respect to other abelisaurids.Tho Indian taxa lack, at least, the prominent central dome on frontal bones autapomorphic of Majungatholus atopus, or the paired frontal horns that characterize the Patagonian Carnotaurus.In this regard, the morphology of the skull roof of I~zdosuchus and Indosaurus is more conservative, and looks similar t o Abelisaurus in being dorsoventrally thick but without prominences above the skuli roof.askwhether they arevalidspecies.Since Huene's description, many authors have accepted the anatomical distinctions between indosuchus and h d o s a u r u s , supporting them a s valid taxa.Moreover, they were interpreted as belonging to quite different theropod clades: while Indosuchus was considered as a tyrannosaurid, Indosanrus was interpreted as representative of a lineage that inherited primitive features from Jurassic forms such as ~~megalosaurs~~ (Chatterjee, 1978;Walker, 1964).Huene and Matley (1933), and later Chatterjee (1978) and Chatterjee and Rudra specimen, and thus the validity of this fcature is here dismissed; preservation of the braincases does not prove t h e presence of horn-like tuberosities in indosaurus, nor a dorsally smooth postorbital in indosuchus.O t h e r possible distinctions recognized by previous authors between li~dosuchus and ir~dosaur-us concerning the thickness of the skull roof, the anteroposterior extension of supratemporal fbssa, the lusion of sutures, and the degree of development of rugosities on the skull bones, may reflect individual variations.In thoseregards, the frontal dome (19961, offered a list of anatomical distinctions ofMajungatholus shows a variety ofshapes, from between the hasicrania of both taxa.including being inflated in some soecimens (Sues & Taauet.differences in the transverse width of (he parietal sagital crest, the presence or absence of a "transverse crest" on the dorsal surface of the skull, the dorsoventrai thickness of the frontals, and the contour of the supratemporal fossa.However, i t is difficult t o evaluate such distinctions, not only because most of the basicrania were unavailable tbr t,he present study, but also because the preservation of the skulls is far from optimal.For example, in the available specimen ofindosaurus (GSI K271565) the dorsal surface of the braincase is eroded, thus no features of the frontal bones or sagital parietal crest are preserved; the purported presence of a ,ctransverse crest above arid behind the orbit., in indosaurus was not identified in our inspection of the ., 19797, to slightly dcveioped in others (Sampson etal., 1998).Such development of the frontaldome also affects the width and shape of the sagital parietal crest ofMajungatholus.This possihle case of individual variation in the Malagasy ahelisaur serves as an alert when distinctions between the poorly preserved skulls of indosuchus and Ii~dosaurus are evaluated.In sum, anatomical distinctions between Indosuchus andindosaurus arc doubtful, a t least.

S k u l l b o n e s o r i g i n a l l y d e s c r i b e d as <<carnosaurianx
Basioccipital Iluene and Matley (1993, pl X, XI) identified two different types of basioccipitals.I n GSI K271687 the exoccipitais presumably form t h e floor of t h e foramen' magnum, the neck is anteroposteriorly elongate and has a median ventral groove, the posterior surface of the basioccipital hears a douhle tubercle, and the basioccipital tubera seem to be ventrally bifurcated.Instead, in GSI K271628, the exoccipitals are excluded from the floor of the foramen magnum, and the neck is anteroposteriorly short.Unibrtunately, most of the ventral portion ofthe basioccipital is broken, so the peculiar traits enumerated above for GS1 K271687 (e.g., posterior surface of the basioccipital with a pair of tubercles, basioccipital t u b e r a ventrally bifurcated) remain unknown in GSI K271628.Chatterjee (1978) considered that basioccipital GSI K271687 belongs to indosaurus, and GSI K271 628 to Indosuchus, but such referal is untenable on the basis of current knowledge of these genera.However, it seems correct that two kinds of basioccipitals are present in the ~Carnosaur bed>,.Besides, it is difficult to discern abelisaurid traits in each of the basioccipitals.In Abelisaurus, Ilokelesia, Majungutholus and Carnotaurus the basioccipital condyle is rounded, robust and with a short neck, thus resembling GSI K271576.I-Iowever, in Ilokelesia the ventral surface of the neck is grooved, whereas in Majurzgatizolus it is keeled.However, because other cranial bones belong to Abelisauridae, we tentatively assign the occipital condyles to this group as well.
Premaxilla.The description of this cranial element will be primarily based on specimens GSI K271710 (a left premaxilla) and GSI K201619 (a right premaxilla), and a pair of premaxillae (AMNH 1763; Fig. 4) that Chatterjee (1978) interpreted as presumably corresponding to a sin.. gle individual.This author referred to AMNH 1753 as Indo.suchus,although there is no firm bases for such assumption, because the holotype of this species does not preserve premaxillary bones.Accordingly, we refer to these bones as Abelisauridae indet.The premaxilla is higher than rostrocaudally long (for example, in GSI K271710, the anteroposterior width is 6.2 cm, while its dorsoventral height is 8.5 cm), as characteristic of abelisaurids (Novas. 1997).The anterior surface of the ascending process is sharply defined.The lateral surface is decorated by numerons small foramina, which are particularly abundant on the ascending ramus.However, larger foramina exist along the alveolar margin.The medial symphysis for articulation with the opposite premaxilla is extended.Behind the nasal depression are two distinct, hook-like processes, which articulate with a pair of excavations on the medial premaxillary process of the maxilla.The Fig. 4. Right premaxilla of Indosuchus raptorius (AMNH 1753), in A, medial and B, lateral views.
AMNII 1'753 a shallow but distinct notch is identified on the posterior border of the premaxilla, corresponding with the subnarial foramen.The narial fossa is deep and well delimited, in cont,rast with Carnotaurus sastrei a n d Abelisaurus comahuensis.The external surface of the right premaxilla (K201619) is not decorated with the foramina and tuberosities present in the remaining abeiisaurids, and the ascending ramus looks more robust and complex than in AMNH 1753.
Maxilla.Chatterjee (1978,  The following description of the abelisaurid maxillae from India is based on observations made on AMNW 1955, AMNH 1753, GSI K271538, and GSI K271544 (it must be noted t h a t a t the collections of the GSI, specimen GSI K271538 is mistakenly labeled as GSI K271548, while specimen GSI K271544, not illustrated by Huene and Matley, is incorrectly labeled as GSI K271538; specimen GSI K271548 is missing).The maxilla GSI K271 538 (Fig. 5) is characteristically triangular, anteroposteriorly short, and bas a proportionally low ascending process, characters also present in AEielisaurus, Carnotauras andMajurigatholus.The articulation with the pre~naxilla is made through a strong medial premaxillary process, which is located high on the medial aspect of the bone, constituting another abelisaurid character.The lateral surface of the maxilla exhibits strong decoration that include foramina and grooves.The grooves, which a r e predominantly oriented dorsoventrally, split and join in a compiex pattern as occurs in other abelisaurids.The grooves are moro marked on the ascendingramus than in other regio~ls of the maxilla, specially in larger specimens (e.g., GSI K271538).The maxillary ascending ramus is almost verticaily oriented, with the rostra1 margin slightly convex in lateral view.The caudal margin of t h e ascending ramus is transversely wide and deeply excavated, and provided with a presumed promaxillary fenestra (hidden in side view).A maxillary fencstra is lacking.The dorsal margin of t h e maxilla is transversely convex and affected by deep, presumably pneumatic, excavations.0 1 1 the internal side is seen a roui of dental foramina along the contact between the dental plates and the remainder of the medial surface of the maxilla.
Two conspicuous abelisaurid synapomorphies are identified on the available Indian maxillae: a minute antorbital fossa, and dental plates dorsoventrally deep, strongly fused, and decorated by obliquely oriented striations.

M N H 19S5 was referred to the subfamily
Carnotaurinae by Lamanna et al. (2002), because it shares a promaxillary fenestra obscured by the lamina iateralis of the ascending ramus.Additionally, these authors suggested that a n anteroposteriorly short lnaxillary body with parallel dorsal a n d ventral margins is synapomorphic of this subfamily.EIowever, the diagnostic value of such characters is debatable, because t h e maxilla is unknown in other ahelisauroids (e.g.Ilokelesia) or it is incompletely preserved in others (e.g.AbeZisaunls).In other words, such features may exhihit a wider distribution among abelisaurids.
Several isolated teeth were recovered from the quarry.However, they a r e lost in t h e GSI collection, and the figures given by Huene (pl.XIII, figl-10) are not detailed enough.However, many of these dental pieces agree in general shape with t h e t e e t h of o t h e r ahelisaurids (e.g., Abelisaurus, Mqjungatholus) in t h e great transverse compression and degree of backward curvature.
Jugal.The following description of this hone is hased on specimens GSI K271577 and GSI K271 680.We could not access another two specimens (K271635 and GSI K271681) that were descrihed (but not illustrated) by Huene and Matiey (1933) as portions of right and left jugals.Specimen GSI K271577 was originally descrihed as a right lacrimal, but it matches well with theascendingramus of t h e left jugal of Carnotaurus and Majungatholus (Fig. 6).Besides, specimen GSI K271580 (Fig. 71, originally was interpreted by Huene a n d Matley a s a right postorbital, interpretation accepted hy Chatterjee (1978) who referred to the presumed postorbital (K271580) as Indosuchus.GSI K271580 is here reidentified as a portion of a right jugal (Fig. 7).
Specimen GSI N271580 has a triangular aspect in lateral view, with a slender and rod-like dorsal extremity.The ventral half of t h e bone is transversely narrow, and exhibits a concave lateral surface.Towards the dorsal end the hone becomes transversely thicker, constituting the most laterally projected portion of the jugal.Its lateral surface is decorated by grooves, being intensely sculptured on the posterior andventral orbital portions.A distinct oblique groove is present on its lateral surface.The medial surface of the lacrimal is smooth, with a caudal depression surrounding the infratemporal opening.The medial surface of the ascending process ofjugal forms a longitudinal prominence (Fig. 7).
Jugals GSI K271580 differs from Carnotaurus, Abelisaurus and A4ajungatholus mainly in the presence of a deep, rounded notch on the caudal margin of the ascending ramus.Specimen GSI K271577 exhibits rugosities with a different pattern than these seen in Carnotaurus and Majungatholus.
.~~Lacrimaln.Specimen GSI K271708 was interpreted hy Huene and Matley (1933 pl.XI, fig , 5) and later by Chatterjee (1978) a s Bonaparte et al., 1990).Not to scale.Abbreviations: asp, ascending process; ito, infratemporal opening.corresponding to the upper portion of a right lacrimal.However, the bone lacks the pattern of rugosities and the wide contact for the postorbital as seen in the lacrimal of Carnotaurus and Majungatholus.We a r e unable t o identify specimen GSI K271708.
Quadrate.This bone was originally described as a right astragalus (Huene & Matley, 1933,  pl.XIX, fig.1;GSI K27/684), but it corresponds in fact to a left quadrate.The specimen preserves the distal articular condyles, the base of the pterigoid ramus, and a rugose lateral surface for the attachment of quadratojugal.The anterior facet of the distal condyles is nearly flat, as it occurs in t h e abelisaurids Ilokelesiu, Majungatholus and Carnotaurus (see Wilson et al. 2003, character 53).
Dentary bones.The following specimens were studied: GSI K271550, GSI K271709, GSI K271529 (incorrectly catalogued as GSI K271527 in the GSI collections, a number corresponding to a left articular), and AMNH 1960 (a number that also applies to a caudal vertebrae).Huene and Matley (1933) listed, a l t h o u ~b did not des-ornamentations made up by loramina, grooves and prominences, resembling those of Carnotaurus and LMajungatholus.As in the latter two taxa, a clear separation exists between the strongly decorated ventral half relative to the smooth dorsal (or -labial,,) halfof the dentary The line defined by these two surfaces describes a dorsally concave curvature, which in Carnotaurus and Majungatholus is lined by a number of large foramina.The abovementioned resemblances clearly support the hypothesis that all of t h e tberopod deutaries recovered in the "Carnosaur bed" belong to Abelisauridae.
Huene and Matley (1933:50) and l a t e r Chatterjee (1978) cited some distinctions among these dentaries, but because of the fragmentary nature of the material, plus the impossibility of comparing them directly, we prefer do not address this aspect.
Surangular.Specimen GSI K271693 was originally described as a left articular (Huene &  Matley, 1933, pi.XII, fig.31, but it is identified here as a left surangular The dorsal surhce of this bone is almost flat and transversely wide,  continue forward and inward, perforating the anterior surface of t h e articular bone: The surangular resembles that of Carnotauras sastrei (Bonaparte et al., 1990) in the presence and position of the pair of fbramina near the glenoid cpvity.

Theropod tsxa based on vertebral elements
Seven theropod species were coined by Huene and Matley on the sole basis of vertebrae: C o m p s o s ~~c h u s solus, Laeuisz~chus indieus, J u b b a l p u ~i a tenuis, Coeluroides largus, Dryplosauroides grandis, Ornithomimoides mobilis, and Ornithornimoides (?) barasimlensis.Also, some isolated vertebrae were described as corresponding to "ailosauroids" or "coelurosaurs".These specimens are reviewed as follows: Compsosuchus solus.This taxon was described on the basis of a single axis with fused atlantal intercentrurn (GSI K271578; Fig. 10).
Most of this vertebra is preserved, except for the the upper portion of its neural arch, which is broken.The axial centrum bears one large pleurocoel, and a pneumatic opening posteroventrally to the diapophysis.The anterior articular surface of the intercentrum is slightly convex and kidney-shaped, while the posterior one is slightly concave.The diapophyses are small and blunt.A sharp lamina extends obliquely from the diapophysis to the postzygapophysis.The neural arch is wide and low.Molnnr et al. (1990) found that the axis of Contpsosuchus resemhies ttiat ofAllosaurus in the similar position of the upper pleurocoel, the cylindrical aspect of she axial intercentrum in ventral view, the axial pieurocentrnm loss than twice the length of the axial intercentrum, find the broad condition of the neural canal.This lead Molnar et al. (1990) t o include Compsosuchus within Allosauridae.However, GSI K271578 exhibits t h e following resemblances with Carnotaurus: presence of a pneumatic pore posteroventrally to the diapophysis, and at least one large pleurocoel on the axial centrum, proportionally small and rod-like diapophyses, presence of a sharp lamina extending obliquely from the diapophysis to the postzygapophysis, and a neural arch laterally expanded and triangularshapedin dorsal view.s he axis that served as basis to create Comososuchus closelv resembles that of IS1 R9111, referred to Inclosaurus by Chatterjee and Rudra (1996).In sum, the generalmorphology of this cervical vertebra indicates that it pertains to an ahelisaurid theropod.Since there are no substantial differences with the axis of indosuchus ?nczusNaturales, n. s. 6 [li, 2004 Fig. 10.Compsosuchus solus (K271578), axis in A, left lateral, B, dorsal, and C, anterior views.Abbreviations: dp, diapophysis; ic, intercentrum; nc, neural canal; op, odontoid process; poz, postzigapophysis.
(IS1 R 9111) and no evident autapomorphies are recognized, we conclude that Compsosuchus is a nornen dubium.La,evisuchus indicus.This taxon was described (Huerre 1932;Huene & Matley 1933:60-61, pl.XX, figs. 2-5) on the basis of three cervicals (GSI K201613, GSI K201614, and GSI K271696) and one dorsal ~rertebra (K271588).Unfortunately, from these elements only a mid-cervical vertebra was located at the GSI collections (GSI K271696; Fig. 11).The vertebra presumably corresporlds with cervical 5.The centrum is dorsoventrally low (the cranial surface is 19 mm high), and long (nearly 42 mm), with a n almost flat and transversely wide ventral surface.The cranial articular surface olthe centrum is kidney-shaped, slightly concave, and.with rised borders.The caudal articular surface is also concave.A pair of pleurocoels are present on the sides of the centrunl, and a pneumatic depression exists more dorsally on the right side (this may correspond to the "third" pleurocoel cited by Hnene & Matley, 1933).The parapophyses are prominent.The neural arch is low and transversely wide (the distance between external margins of the prezygapophyses is 43 mm).Asharp dorsal margin connects the preaygapophyses with t h e epipophyses, thus bounding laterally the dorsal surface of the neural arch.The dorsal surface of the neural arch is concave between the lateral rnargin and the neural spine.T h e latter is pyramidal, low (7 mm height) and craniocaudally short (9 mm).Its cranial surface is damaged, but on the caudal surface exist ligament scars that do not reach to the top of the spine.On the cranial surface of the neural arch, and ventroniedial to the prezygapophyses, exist a pair of deep and elliptical pneumatic fossae.The articular surface of the prezygapophysis is smooth and slightly convex (transversely and craniocaudaily).The caudal surface of tho neural arch is deeply excavated between the postzygapophyses and diapophyses.The postzygapophyses are broken dorsally, and corisequently their respective epipophyses are incomplete.However, some inlormation about their morphology is still available: the epipaphyses are projected dorsally andlaterally (as seen from behind; Fig. 11A).They are craniocaudally extended, roughly representing 75% of t h e maximum diameter of the postzygapophyseal articular surface.The caudal portions of the epipophyses have not been preserved.However, the epipophyses lack the slender and conical cranial projections present in Noasaurus (Bonaparte & Powell, 1981), for example.
Huene and Matley (1933, pl.XX, figs. 2 a.nd 4) illustrated another two cervicals of Laeuisuchus (GSI K201613 and GSI K201614).Cervical GSI K201613 is remarkable fbr the extensive, iahleshaped dorsal surface of the neural arch, ciosoly resembling that of Noasaurus, Majungatholus and Carr~otaurus.Cervical GSI K201613 is here interpreted as more cranial in position than the proviously described cervical GSI K271696.Reassons supporting this include a centrum with diiYerently inclined cranial and caudal articular surfaces, a proportionally smaller centrum diameter with respect t o the n e u r a l arch (proportions that are also noticed in cranial cervical vertebrae of Carnotaurus, for example), the dorsal surface of the neural arch is wide, the lateral margin of the dorsal surface is straight in side view, the neural spine seems to he absent, and the epipophysis seems t o be well developed and dorsally projected.Norman (1990: 3021, following Huene andMatley (1933: 60-61),pointedout that thevertebra of Laeui.suchusresemhles that of "Aristosuchus" (junior synonym of Cala?nospondylw Fox, 1866).I-Iowever, the vertebra of Calamospondylus differs from Laeuisuchus in that only one pieurocoel is present, the cranial articular surface is convex, and the dorsal surface of the neural arch is not transversally wide and well defined as in 1,aeuisuchus.In sum, there are no coelurosaurian i'eatures in Laeuisuchus.On t h e contrary Laeuisuchus shows the following abelisauroid features: elongate epipophysis, pair of foramina on centrum, pyramid-shaped, low and transversely thick neural spines (Fig. 121.h i s ~~c h u s has cervicals that are proportiondly longer than in Carnotaurus and Majungatholus.Also, in Lueuisuchus the anterior articular surface is slightly concave, instead of being convex as in Carnotaurus.The articular surfaces of the prezygapophyses are anteroposteriorly \-vide in Lueuisuchus, opposite to the transversaily expanded ones of Carnotaurus and Majiingatholus.Laeui.suchus has large pneumatic cavities below the prezygapophysis, whereas in Carnotaurus and Majungutholus the cavities are smaller in diameter.
Within ahelisauroids, Laeuisuchus more closely resemblesNoasaurus and Masiahasaurus.Carrano et al. (2002) suggested that these ihree taxa could be included within Noasauridae because they shares cervical vertehrae with anteriorly placed neural spines and cervical epipophysis that are reduced posteriorly Laeuisuchus and Noasaurus Fig. 13.Omiihomirrtoides mobilis (K201614B1, dorsal vertebra in A, dorsal, and B, left lateral views.are similar in the development pneumatic cavities, the presurried abserice of neural spines on cranial cervicals, and in the position of the both pre-and postzygapopliyses. Laevisuchm m e r s fromNoasaurus in havingthe a~tediapophysial, postdiapphyi3ialanddiapophysial cavities shallower, the diapophysis are wider and less ventraUy directed, and the neural spine is less extendedanteroposteriorly Indorsalview Laeuisuchus has shorter prezygapophyses and the postzygapophyses are caudally rounded (not acute as in Noasaurus).Imuisuchus differs from Masiaizasauras in having the space between the postzygapophyses less excavated, the prezygapophyses are thinner, and the infrapostzygapophysial and infraprezygapophysial cavities shallowec Omitl~omimoides.Iluene & Mattey (1933, pi. YX, fig. 8-10) created Lhis genus (with a pair of species, 0 .mobilis and O.? barasimlensis! on the basis of several vertebrae t h a t this author interpreted as dorsals similar to those of the ornithomimids Orr~ithomimus andStruthiomimw.One of the species (0. mobilis! is represented hy iive large and elongate vertebrae (GSI K201610, GSI K20/614B, GSl K271586, GSI K271597, arid GSI K271600), and t h e second species (0.1 barasimlensis) by a set of smaller vertebrae (GSI K271531, GSI K271541, GSI II271604, and GSI K271 crest, the neural spine is anteroposteriorly extended, the base of the transverse processes is ventrally buttressed and excavated, and the apneumatic centrum is longer than deep.We did not recognize autapomorphic features diagnostic of Ornithomimoides.On the contrary, the caudal vertebrae referred to this taxon look closely similar to caudals of other abeiisauroids.I n sum, we follow previous authors (Norman, 1990) in considering Ornithomimoides as a nomen dubium, the set of vertebrae representing proximal caudals of an Abelisauroidea gen.et sp.indet.
Dryptosauroides grandis.Dryptosauroides was recognized by Huene and Matley, (1933; pl.XXII, figs. 1, 2, 3 and 4) on the basis of six dorsal vertebrae (K201334, GSI K201609, GSI K271549, GSI K271601, GSI K271626, and GSI K27!602), but they also referred to this taxon a cervical vertebra (K271555) and several dorsal ribs (GSI K201615, GSI K271547, GSI K271623, GSI K271624, and GSI K271625).This set ~Evertebrae does not belong, in fact, to the dorsal, but to the caudal region.Among them we have only accessed specimen GSI K201609 (Fig. 14) which exhibits almost the same morphology as caudal vertebra GSI K201610 of Ornithomimoides mobilis (Fig. 14).Caudal vertebrae of Dryptosauroides match well with the proximal caudals ofMajnngatholus @ers.obs.).As 682).Review of these specimens indicate that they is the case for Ornithomimoides the vertebrae of are not dorsals but caudalvertebrae (Fie.13).their Drvotosauroides corresaond to the nroximal ." , ., .morphology corresponding t o those of caudals of a n indeterminate abelisauroid.Majungatholus (pers.obs.).Asit occursin theiatter Consequently, Dryptosauroides g r a n d i s is abelisaurid, the prezygapophyses are close each considered as anomen dubium, followingprevious other, they lack of the ventral projections present authors (Molnar, 1990).The size of the caudals in dorsal vertebrae of neoceratosaurs (see Fig. 18 indicate the presence of a very large animal, foraniUustrationofsuchprojections),theprespinal surpassing the size of Carnotaurus (MACN-CH depression is deep and divided by a tiny sagital 8941, for example.Jubbulpuria tennis.This taxon was crected on the basis of two small vertebrae (K201612 and GSI K271614), identified by Huene and Matley (1933, pl. XX, figs. ti and 7) as corresponding to the dorsal region.Review of' available specimen GSI K201612 indicates that it is not a dorsal but a distal caudal vertebra (Fig. 15).The centrum is low and elongate, the transverse processes are expanded and dorsally cxcavated (as in Coeluroides largus GSI K271562), and the neural spine is anteroposteriorly extended.The spine is represented by a tiny axial crest between the prezygapophyses, but towards the rear it becomes transversely s t o u t e r a n d was probably dorsoventrally higher.The postzygapophyses (Huene & Matley, 1933, pl.XX, fig, tia,bj were laterally facing, as it occurs in mid to distalcaudals of o t h e r theropods (e.g., Majungatholus, Allosaurus, Tyrannosaurus).
The caudal vertebra described as Jubbulpuria has wing-shaped transverse processes, which look well developed for such a distal caudal.Distal caudals of abelisaurids (e.g., Majungatholus, pers.obs.jlack well developed transverse processes.Also, its dorsal surkce is excavated, different fro111 t h e dorsally flattened of abelisaurids.However, the Neocomian basal abelisauroid Ligahueino andesi (Bonaparte, 1996) shows similarly developed transverse processes on distal caudals, supporting referal of vertebrae of Jubbulpuria as to Abelisauroidea.Romer (1956) agreed with Huene and Prlatley (1933) in t h a t Jubbulpuria is a member of <Coelnrosauria~, but Norman (1990) considered Jubbulpuria tenuis as a nomen dubium.We follow this last interpretation.
Coeluroides largus.This taxon was coined on the basis several isolated vertebrae (K 271562, GSI K271574, GSI K27i595) that Huene and Matley (1933) erroneously interpreted a s corresponding to the dorsal region.They all belong to the caudal region, as already recognized by Welles (1984) and Molnar (1990).Specimen GSI K271595 (Huene & Matley, 1933, pl.XXI fig.5) is a neural arch that closely resembles proximal caudals of Majungatholus (pers.obs.) a n d "0rnithomimoide.s"(Fig. 13).However, specimens GSI K 271562 and GSI K27i574 of Coeluroides largus show distinctive fcatures that merit more detailcd consideration.Caudal GSI K 271562 (Fig. 16) is distinguished by i t s wide, almost horizontally oriented and wcll separated pre-and postzygapophyses; also, the transverse processes are notably expanded and triangular-shaped in dorsal view, with their dorsal surface deeply excavated, thus resulting tho anterior margins of the transverse proccsses beingraised.The ncural spine is broken, but its basc is axially extended and transversely r o b u s t Zygapophyseal rnorphology of GSI K271562 suggests that it is a mid-caudal vertebra.The peculiar morphology described above is also seen in AMNH 1957 (Fig. 171, a caudal vertebra catalogued as Indosuci&us mploriils, which also has an elongate and low, apneumatic centrum.Moreover, specimens GSI K271562 and AMNH 1957 are similar to t h e fragmentary caudal vertebra GSI K201612 referred t o Jubbulpuria (Fig. 15) in t h e morphology of the transverse processes (e.g., extensive, triangular shaped, a n d dorsally excavated).Interestingly, GSI K 271562 (Coeluroides largus), AMNI-I 1957 (catalogued as Indosuchus raptoriusj and GSI K201612 (Jz~bbulpuria teni~isj share a similar set of features that contrasts with the caudal morphology of Majungatholus, Carnotaurus, Ilohelesia, Aucasaz~rus, as well as other abelisaurid caudal vertebrae of the Indian collections (o.g., GSI K271 Fig. 15.Jubbulpuria tenuis (K20/612), distal caudal vertebra, in A, right lateral, and B, dorsal views.Abbreviations: ns, neural spine; pre, prezygapophysis; tp, transverse process.595, GSI K201610.GSI K20/614B, GSI K271614, GSI K271586, GSI K271597, GSI K271600).It is iniportant to note that mid-caudal AVNH 1957 is not only different in morphology from proximal caudal AMNH 1960 (also catalogued as Indosuchus raptorius), but it is iarger than the latter, thus indicating that they do not belong to a same individual (and presumably pertain to different species).
Possibly Coeluroides largus may represent a valid taxon of a n indeterminate ahelisauroid theropod.Mid and distal caudals of Coeluroides retained notably developed neural spines.Comparing AMNH 1957 with the similarly elongated and low caudal centra of caudals of Majungatholus, it becomes evident that in AMNH 1957 the transverse processes are well developed, aliform structures, whereas in Majungatholus they a r e absent or represented by a faint longitudiilal ridge.The evidence is not enough to evaluate whether Coeluroides l a r g u s and Jubbulpuria tenuis are synonyms, but their caudal vertebrae may represent theropod lineages different from Abelisauroidea.Similarities noted above between distal caudals of Jubbulpuria tenuis and Ligabuieno andesi (Bonaparte, 1996) argues in favor that other candais with deltashaped transverse processes (e.g., Coeluroides largus, AMNH 1957) also belong t o Abelisauroidea.
IV Axial skeleton remains originally referred to as "allosaunid" and "coelurosawid" "Allosaurid cervical vertebra" (K 271590) This was described as a cervical vertebra (Huene & Matiey, 1933, pi.XIV, fig.I), hut it is here reinterpreted as a dorsal vertebra because the prezygapophyses are close each other, and the parapophyses are projected outwards occupying a high position on the neural arch (Fig. 18).GSI K271590 is similar to dorsal 9 of Sinraptor (Currie & Zhao, 1993) in t h e morphology of the prezygapophyses, with pendant ventral processes, and t h e pattern of laminae connecting the diapophysis with the parapophysis.Also, the prezygapophysis and diapophysis are connected by a ridge that is dorsally convex in lateral view, a deep pneumatic cavity is located between prezygapophysis and parapophysis, and an " Yshaped crest connects the parapophysis with the diapophysis.The same description also applies to dorsal 5 of Carnotaurus and dorsai 7 of Ceratosaurus (Welies & Madsen, 2000), but in these twoiaxa and GSI K271590 the parapophyses are more laterally prominent than in Sinraptor.Also, the prespinal cavity is large and deep, a synapomorphic trait shared by all neoceratosaurs (Holtz, 2000).In sum, dorsal GSI K271590 exhibits neoceratosaurian features, and because its morphology is congruent with that of abelisaurids and it was found in association with abelisaurid bones, we refer this specimen to Abelisauroidea indet.
"Allosaurid cervicalvertebra" (K271572).This is a large vertebra (16 cm height), wish an opisthocoelous centrum, albeit the cranial articular surface is almost flat (Huene & Matley, 1933, lam.XIV fig.2).Huene cites t h a t a single pleurocoel is present below the diapophysis.This cervical does not resemble that of Aheiisauridae in the shape of the neural spine (which is axially extented and transversely narrow, instead of craniocaudally short and transversely wide as in abelisaurids), and t h e apparently poor development of the epipophyses (in contrast with the high and craniocaudally extended epipophyses of abelisauroids).Unfortunately, specimen GSI K271572 is lost at the GSI collections, and first Sacral vertebrae.H u e n e and Matley illustrated some portions of fused sacral.vertebraeunder the numbers GSI K271.554 (twopieces), GSI K27/533 (two pieces) and GSI K271571 (Fig. 19).More recently, Bonaparte (1991b) referred specimens GSI K271533 and GSI K271554 to Abelisauridae because they are fused into a sin-gle, rod-like s t r u c t u r e similar t o t h a t of Carnotaurus.Although we do not dismiss that specimens GSI K271533 and GSI K271554 belong to Abelisauridae, their morpholog~ more closely resembles that of tainetasaurus (Matley, 1923), R a j a s a u r u s (Wilson et al., 2003) a n d iWasiakasaurus (Csrraro etal., 2002) in that each sacral element is transversely broad and the contact between succesive vertebra is well marked.I n Carnotaurus, instead, the sacral cen-Fig.18. Dorsal vertebra of a n indeterminated abelisaurid (JC.271590) in A, lateral, B, anterior, and C, dorsal views.Abbreviations: nc, neural canal; pp, parapophysis; prz, prezygapophysis; psf, prespinal fossa; pvp, pendant ventral process.tra are strongly reduced in transverse diameter and the contacts among succesive sacrais are slightly marked.Notably, specimen GSI K271571 (Eirstly interpreted by Huene and Matley as coelurosaurian) responds to the Carnotaurus morphotype, thus suggesting the presence of two different large abelisauroids in t h e fossil assembiange.
There is a large isolated centrum (GSI Pi271 598; 19C,E) with a strong constriction a t midlenght, but with highly expanded articular facets.This morphology remember that of sacral 1 of Rajasaurus Wilsonet al., 2003), thus we interpret specimcn GSI K27i598 as a probable sacral 1.
Proximal caudal vertebra (AMNH 19601.This proximal caudal is represented by a neural arch with elongate transverse processes, which a r e n o t entirely preserved a t their extremities (Fig. 20).General morphology of AMNH 1960 is congruent with t h a t of the proximal caudals of Majungatholus, Carnotauras and Abelisauridae indet.MPM 99.However, derived traits of Abelisauridae (e.g., distally fanshaped transverse processes, and presence of a slender anterior projection on the transverse processes t h a t contacts with the transverse process of t h e contiguous anterior caudal; Martinez et al., in press)  Medium and distal caudals.Some distal caudals (Fig. 22) are characterized by a po1yg.onalcentrum in cross-section, transverse processes represented by thick ridges overiapping both sides of the centrum, flat ventral surface of the centrum,     (t~coelurosaurid~~, Huene & Matley, 1933, pl. XXIII, fig. 3) and GSI K271705 exhibit a similar pattern, although their prezygapophyses and centra are longer and thus they correspond to the distal end of the tail.
Tlie kind of caudais described previously contrasts with another group in which the transverse processes are well developed, deltashaped (i.e., caudoiateraily expanded), and have elongate centra and prezygapophyses.This group of vertebrae includes caudals of Jubbulpuria.An isolated caudal vertebra GSI K271599 (Fig. 21), assigned by Huene and Matley (1933) to a coelurosaurid, resembles Masiakasaurus (Carrano et al., 2002) in the general shape and principally in the shape of the transverse processes and elongate prezygapophyses.The latter suggests this is amid-to distal caudal which retained well developed transverse processes.This combination of features sharply differs from the condition seen in other caudals with equally longcentra (Fig. 221, but with short prezygapophyses and nearly absent transverse processes.This suggests the presence of abelisauroids with different kinds of caudal processes Obviously, more work needs to be done on the caudal anatomy of abelisauroids in order t o resolve the allocation of isolated vertebrae, recognize the morphological variations alongthe tail series, and evaluate the taxonomic significance.

Haemal arches. Several isolated haemal arches
Except for GSI K271680, the haemal arches are elongate and rod-like, and lack a distal expansion, resembling the condition seen in other ahelisanrids and Ceratosaurus Wilson et al. 2003).In contrast with Carnotaurus, at least, the Indian chewrons posses a haemal canal that is proximally open.
V Pelvie and hind limb bones originally referred to as "allosaurid" and "coelurosaurid" With the exception of two fragmentary ischia, no other pelvic bone was describcd from the Cranial is to the left."Carnosaur bed".However, several hind limb elements (femora, tibia, metatarsals, a n d phalanges) have been recovered and they offer a good source of anatomical information.Huorie and Matley (1933) distinguished two types oi' femora, one of a robust animal and another more slender one.IInene associated the robust kind of femora with a single, equally stout tibia, and referred them to the eallosanrids.>.Walker (1964) considered t h e slender kind of femora a s belonging to Inilosaurus and the stouter type to Indosuchus.
Ischiurn.Two fragmentary proximal ischia were described by IIuene (K271686 and GSI K271 546;pl. XVI, figs. 7, 8).They a r e poorly informative, and their morphology matches with most basal theropods.The distance between the Carnotaurw, and that both bones would not have belonged to the larger forms found in t h e "Carnosaur bed".
Femur.Most of the femora described by Huene and Matley (1933) are between 60 cm and 74 cm in length (Fig. 23), thus belonging t o animals of large size.We did not locate the great majority oi'the femora a t the GSI collections, and our comments will mostly rely on Huene and Matley 's i:lustrations.The exception is a proximal portion of a left femoral shaft newly cataloged with number GSI 296 which lost the corresponding numbers of the GSI K series (Fig. 24A).We presume that it may belong to any of the left femora (either GSI K271564 or GSI K271563) cited, but not figured, by Iluene and Matley (1933:55).
As mentioned before, Hnene and Matley distal extremity of the obturator process and the sorted out the available 9 theropod femora from than 4 cm.This is in contrast to the holotype of femora (GSI K271560, GSI K271563, GSI K271564, Carnotaurus, for example, in which the distal tip GSI K271569, GSI K271621, GSI K271627).The of the obturator process is 30 cm from the iliac robust femoraare characterized by their relatively pedicle, and the craniocaudal diameter of the shaft short and robust shafts, thus looking sharply (immediately distal to the obturator process) is 7 diff'erent from t h e remaining non-avian cm.This indicates that ischia GSI K271686 and theropods.They may belong to a single taxon (e.g., GSI K271546 belonged to animals smaller than species) characterized by stout hindlimbs, although some distinctions are observed among them, for exaniple GSI K271570 appears to exhibit a well-developed (i.e., proximally projected) anterior trochanter, whereas in GSI Ii271558 this trochanter is smaller.This last specimen, a t least, exhibits features resembling Xenotarsosaurus (UNPSJB-PV 184-6121 and Indosuchus (IS1 R911 I), including: rounded femoral head, anterior trochanter low with respect to the femoral head, 4'" trochanter convex in side view, and presence of a prominent mediodistal crest.The slender specimen GSI 296 (Fig. 24A) exhibits t h e following resembidnces withXenotarsosaurus and Carnotaurus: the anterior trochanter is cranially convex in lateral view, the trochanteric shelf is prominent and located a t level of the distal end of the anterior trochanter, the 4"' trochanter is also convex in side view, and a conspicuous pit fbr attachment of the M. caudofemoralis is present on the medial surface of femur, cranial to the 4l" trochanter.All of t h e 9 femora discovered in the "Carnosaur bed" are morphologically congruent with the femora of other abelisaurids (e.g.Xenotarsosaurus, Carnotaurus, and IS1 specimens of Indosuchus).Our iriterpretation is that the femora described by Huene and Matley belong to Abelisauroidea, a conclusion that is in agreement with the whole bone assemblage, mostly (if not entirely) made up by abelisauroid hones.
Tibia.Huene and Matley (19333 described three theropod tibiae corresponding to large theropods, none of which was available for tho present study a t the GSI collection.Only GSI K271568 was illustrated by Hueno and Matley (Fig. 25A).As earlier suggested by Bonaparte (1991b1, this  specimen resembles Abelisarlridac in havinga welldeveloped cnemial crest, and apoorly differentiated outer condyle on the proximal end, which is located at almost the same level as the inner condyle.The tibia bears an elonbwte cnemiai crest as usual in neoceratosaurs, but the shaft is remarkably short and stout, being clearly different from other theropods, including most ahelisauroids (e.g., Aucasaurus, Xenotar.sosaurus,Majungatholus, Masiakuswrus).Theonly exceptionsare thelndian Larnetasuurus (Matley, 1923; Fig. 25B) and the Braziiian Pycnonenzosaurus (Kellner & Campos, 2002), in which the tibiais proportionally short.This peculiar condition of the tibia is not due to deformation, loss of its distal portion, or a pathological case, and therefore it constitutes a Huene and Matley (1933) also described some other tibiae i n t e r p r e t e d a s belonging t o aCoelurosaurian (GSI K271526, GSI K271670, GSI K271552, GSI Ii271556, GSI K271662, and GSI K27/669), but none of them could be located at the GSI collections, with the exception of GSI K271 669.We doubt that specimens GSI K271526 and GSI K271669 were correctly identified as tibiae, and we prefer identify them as indeterminate limb bones.
Fihula.An incomplete left fibula (K271620; Fig. 26Aj resembles that of abelisaurids (e.g.Xenotarsosaurus, 12ajasaurus) in having a prominent iliolibularis tubercle and a well excavated fossa on the medial surface of the proximal end.Reserriblances between GSI K271 620 and the fibula of Rajasar~rus (Wilson et al., 2003) include the subtriangular contour of the proximal fossa, which is bounded by strong cranial and caudal ridges.Distal to the fossa, both ridges join to form asingle, prominent longitudinal ridge extending along the fibular shaft.GSI K271620 and Rajasaurus appear to lack the longitudinal groove present in tile allosauroids Sinraptor and Allosaurus (Madsen, 1976).The fibula GST K271 620 and that of Rajasaurus differ from that of the abelisauroid Deltadromeus (Wilson ei al., 20031, because in the latter taxon the fossa, albeit deep, is not subtrianylar but proximodistally elongate.In sum, GSI K271620 is recognized as an abelisauroid fibula. Astragalus and caleaneum.As mentioned in previous pages, the purported astragalus (K271 684;Huene & Matley, 1933, pl Metatarsals.Soveral isolated metat,arsals were recovered in the "Carnosaur-bed".Review of these elements (either in the GSI collection or on the basis ol Huene & Matley's illustrations) indicate that: 1) the elen~ents originally thought as helouging to the manus correspond in fact to the pes; 21 available metatarsals correspond to metatarsal 11, 111 or IV; and 3) all these uieces exhibit abelisauroid leatures.
probably loft metatarsal I").All these hones exhibit deep and transversely compressed distal ends, with asymmetrically developed articular condyles ii.e., the inner condyle is more developed than the outer one), features that also apply to metatarsal IV of the ahelisauroids Masiaka.sauras(Carrano et ul. 20021, Deltadromeus (Sereno et al. 1996), Aucasaurus (Coria et al., 20021, and Abelisauridae indet (MCA 56).As far as Huene & Matley's figures suggest (pl.XIX, figs.5 and 6, and pl.XXIX fig.41, two kinds of metatarsal IV may be recognized: one in which t h e distal ginglyrnus is relatively robust (K271539) and others with a transversely narrower girrglymus (K271659 and GSI K271666).Also, the shaft exhibits adifferent contour in transverse section: in GSI K271539 it is trapezoidal-shaped wiih the longest side dorsal, instead in GSI K271659 the transverse section is subtriangular, with the longest side ventral (Fig. 29).Such differences may correspond to two kinds of pes within Abelisauroidea, one in which metatarsals and their respective phalanges are robust, and another kind in which side metatarsals (11 and especially IV) are more slender, as well as their respective phalanges, a s i t occurs in Velocisaurus (Bonaparte, 1991a).phalanges (e.g, GSI K2716.51,GSI K271652, GSI B phalanges, we have only accessed GSI K20/626B, GSI K271648, GSI K271524 (this last number has ~2 7 m been also applied to a pedal ungual; Huene & Matley, 1933, pl.XIX, fig.13).
With the aim of determining their tentative positions in the pes, we have sorted out the phalanges illustated by Huene and Matley (1933, pls.XIX a n d XXIV) on the basis of their morphology, size, and relative proportions, identifying themas belonging to digits 11,111 and IV This task was also supported by comparing this set of pedal elements with phalanges of other theropods !e.g., Allosaurus, Sinraptor, Velocisaurus.Aucasaurus).It is clear that the phalanges correspond to an~mals ofd~fferent slze and robustness The phalanges of digit I1 are represented by specimens GSI K271654 (Huene & Matley's "allosaurid pedal digit I") and GSI K271524 (Huene & Matley's "coelurosaurid manual digit I").The morphology of these phalanges matches well with that of pedal phalanx 1 of digit 11 of Velocisaurus (Bonaparte, 1991aj in being elongate, strongly asymmetrical, with a dorsoventrally deep proximal end, and a pair of well developed proximoventral longitudinal ridges (Fig. 30, A,B).However, specimens GSI K271654 (8 cm long) and GSI K271524 ( 6 cm long) correspond to an animal considerably larger than Velocisaurus.in which nhalanx 111-1 reaches 2.3 cm long.
The phalanges of digit I11 are represented by the following GSI specimens (Fig. 30, C-F K2716541, of which only GSI K271654 has been illustrated (Huene & Matley, 1933, pl.XIX, fig.7).They are relatively large, measuring between 8 and 7 cm long.Specimen GSI K271654 was originally interpreted as belonging to digit I, hut we interpret it as corresponding to phalanx 11-1.In addition, 18 non-unguals phalanges were also listed by the same authors as belonging to smaller theropods ("coelurosaurids").From this set of symmetrical and d o r s b v e n t r a ~l ~ depressed phalanges, which are more robust than the remaining phalanges.Their proximal ends are laterally and medially flared for articulation with t h e corresponding metatarsal 111 or t h e preceeding phalanx.I n proximal aspect they are crescent-shaped.Available phalanges of digit I11 are characterized by the presence of a low and wide proximal articular surfice, in congruence with the subrectangular distal condyle of metatarsal III..In lateral view, the dorsal margin of these phalanges is more or less straight, and the ginglymus lacks a dorsally expanded articular facet, being slightly more depressed than thc rest of tho dorsal margin of the bone, a condition contrasting with most other theropods (e.g., Allosaurus and Sinraptor).Specimen GSI K271 653 (Fig. 30, D) is interpreted here a s a pedal phalanx 111-2 and GSI K271646 (Fig. 30, E) probably corresponds to pedal phalanx 111.1.The latter one resembles Aucasaurus in its robust proportions, albeit its size (3.6cmlong and 2.6cm wide proximally) indicates t h a t it probably corresponds to a juvenile individual of a robust abelisauroid.I n addition, GSI K271525 is considered to be pedal phalanx 111-1 (it measures 7.6cm long and 5cnl wide proximally), and GSI K271644 as pedal phalanx 111-2 (4.6cm long and approximately 2.3cm wide proximally).I n sum, two subsets of digit 111 phalanges seem to be distinguished by their relative proportions: GSI K271653 and GSI 11271646 are proportionally robust, whereas GSI K271525 and GSI K271644 are of more slender proportions, in particular the latter specimen which resembles Velocisaurus (Bonaparte, 1991a).
Specimen GSI K201626B was interpreted by Huone and Matley (1933, pl. XXIV, fig. 7) as a "coelurosaurid manual phalanx".However, this is a pedal element that closely resembles pedal phalanx IV-1 of Velocisaurus (Bonaparte, 1991a).As in t h e l a t t e r taxon, GSI K201626B is proximodistally s h o r t h u t transversely compressed (Fig. 30, G-I).GSI K201626B is pcdal phalanx IV-1, and GSI K271648 (Fig. 30, I) is in being transversely compressed, with a distal ginglymus asymmetrically developed (i.e., the inner condyle is wider and deeper than the outer one), and with a deep dorsoventral groove on distal ginglymus.They also exhibit well excavated collateral and extensor ligament pits.The proximal end ofboth phalanges is triangular, with the long axis oriented dorsoventrally This condition resembles the proximal phalanx of digit TI, and may lead to confusion regarding the encias Naturales, n. s. 6 fli, 2004 identification of such pedal elements, but the phalanges of digit IV lack of the proximovcntral ridges characteristic of the proximal phalanges of digit 11.Nso, the medial surface of digit IV phalanges is high and almost flat, and exists on its proximoventral corner a deep excavation.Although specimens GSI K20162fiB and GSI K271 648 exhibit a similar morphology, they appear to belong to different individuals: GSI K201626B is 5cm long and 2.2cm wide, but GSI K271648 is 2 .6 ~~1 long and 1.2cm wide, suggesting that individuals in different.growth stages, and presumably belonging to a same species (e.g., a gracile ahelisauroidj, are represented in the "Carnosaur bed". Huene and Matley say (1933:67) that phalanx GSI K201626B ~crfectlvarticulates with the distal halfof a bone that they thought as a "metacarpal of a coelurosaur" (K271666, here reindentitied as distal end of metatarsal IV).
There is another group of phalanges of digit IV (Fig. 30, J-M) which are very short, deep and transversely wide, showing a more conservative morphology similar to that present in other theropods (e.g., Sinraptor, Allosaurus, ornithomimids).Digit IV phalanges of the robust kind are similar to those of the abelisaurid Aucasaurus (Coria et al., 2002), and they can he sorted out on the basis of their size: GST K201337B (probably a pedal phalanx IV-2; it is 2.4cm long and 1.2cm wide proximally) and GSI K271647 (interpreted as a pedal phalanx IV-4) may correspond to a single specimen of small size.Instead, GSI K271638 (identified here as phalanx IV-3, being 3.8cm long and 2.6cm wide proximally), and GSI K271!337(interpreted as phalanx IV-4; it is 2.8cm long and 2.4cm wide proximally), are short, wide and deep pedal digit IV phalanges, responding to an Aucasaurus kind of h o t but belonging to a larger specimen.Summing up, available pedal digits indicate that: 1) they are congruent with abclisauroid anatomy; 2) differences with the set of transversely narrow digit TI1 and IV phalanges described before may reveal the presence of more than one type of abelisauroid species in the quarv, i.e., an Aucasaurus-like foot with more robust phalanges on digits 111 and especially IV (Figs. 29,321, and a Velocisaurus-like foot with slender phalanges (Fig. 31).
There were many more phalanges (Haene & Matley, 1933:67), originally described as belonging to a single foot, but such pedal elements were not illustrated with the exception of GSI K271646, GSI K271647 and GSI K291337B (see Figs. 30 and 321, so there are no possibilities to evaluate such an association.1991a), in dorsal view.C-E, matatarsal IV (K271666) articulated with phalanx 1.IV (K20/626B), in dorsal (C), distal, proximal, and side views; F, composite reconstruction of pedal digit IV in side view, based on different pedal elemcnts (indicated on the figure), and compared with same digit of' Velocisaurus (from Bonaparte, 1991a).

Unguals.
As analyzed elsewhere (Novas & Bandyopadhyay, 2001) the set of ungual phalanges figured by Huene and Matley (1933) correspond to the pes, thus dissmising interpretations of these authors that at least some unguals belong to the hand.Also, these pedal unguals exhibit the same morphological pattern (e.g., presence of proximally bifurcated grooves, rounded bump on the lateral side of pedal unguals, and ventral surface excavated or witb a narrow deep furrow), described for other abelisauroids (Abelisauridae indet.MCA 56, Masiahasaurus; Novas & Bandyopadhyay, 2001;Carrano et al., 2002).For a more exhaustive review of the pedal unguals, see Novas and Bandyopadhyay (2001 ).

CONCLUSIONS
The review of the theropod hones collected at t h e "Carnosaur bed" demonstratcs t h a t differences in shape, size and proportions of postcranial and cranial bones support the presence of individuals a t different growth stages.Cranial and postcranial elements exhibit ceratosaurian, abelisauroid or abelisaurid traits, or they are n~orphologically congruent witb abelisauroids.In other words, most, if not all, of the theropod specimens ducumented in the quarry correspond to Abelisauroidea.The only possible exception may be a large cervical vertebra (K271 Jubbulpuria tenuis, Coeluroides Eargus) are here Most of the taxaoriginally described by Huene considered nomina duhia, in agreement with and Matley (1933) on the basis of vertebral previous authors (Molnar, 1990;Norman, 1990;remains (Con~psosuchus solus, Dryptosauroides Welles, 1984).The axis of Compsosuchw solus may belong t o a n abelisauroid t h e s i z e o f Carnotaurus, and it is not illogical to refer this vertebra to Indosuchus raplorius because of its similarities with this taxon.Besides, the proximal caudals of llryptosauroides g r a n d i s and Orfiithomimoides barasimler~sis a r e morphologically congruent with the proximal caudals of abelisaurids (e.g., Majungatholus).The caudals of Jubbulpuria tenuis and Coeluroides largus may also helongto Abelisauroidea (because of their resemblance with the Patagonian Ligabueilao), albeit they seem different from the caudals of abelisaurids, thus suggesting that the fossil assemblange a t t h e "Carnosaur bed" includes the remains of individuals corresponding to different abelisauroid species.
Sorting o u t t h e available cranial and postcranial materials into discrete individuals and taxa is currently not possible, but some postcranial bones (sacral and caudal vertebrae and pedal bones) show contrasting morphological patterns.Among equally elongate distal caudals, some bear s h o r t prezygapophyses a n d almost absent transverse processes, instead other caudals bear elongate prezygapophyses and well developed, triangular-shaped transverse processes.We also recognize two different pedal types (i.e., transversely wide vs. transversely narrow pedal digit IV) and two types of sacrum (i.e., a conservative one in which sacral centra -although fused-remain distinct from each other, vs. a Carnotaurus-like sacrum in which the centra are rod-like, with smooth i~~tervertebral contacts).Such distinctions in sacral, caudal, and pedal morphologies probably reflect the presence of two main abelisauroid clades, informally large abelisaurids (represented in the quarry by Indosuchus and I n d o s a u r u s j aiid smaller noasaurids (represented in the quarry a t least by Laeuisuclzusj.Because Laeuisuchu.~ is referred to as the Noasauridae, and since some metatarsals and pedal phalanges resemble these of the noasaurids Velocisaurus and Masiakasuurus, we tentatively associate the slender foot bones with Noasauridae or Laeuisuchus.Caudals with well developed transverse processes a r e also tentatively referred to Noasauridae.Instead, distal caudals devoid of transverse processes and having short prezygapophyses, and short and robust pedal digits, more probably belong to abelisaurids.To this list of robust abelisaurid bones, we add all the hindlimb bones (femora, tibiae, metatarsals, phaianges) of large size as well as those of small size but with robust proportions.With respect to the sacral vertebrae, the slender and rod-like kind is documented s o far in Carnotaurw among abelisauroids, but the more conservative type (in which each sacral eleinent is transversely broad and the contact between succesive vertebrae is well m a r k e d ) are documented among abelisauroids hot11 in Ahelisauridae (e.g., Rajasaurus) and Noasauridae (e.g., Masiakasaurusj.For this reason, it is probable that the "Carnosaur bed" includes remains of different forms ofAbelisauridae.In this regard, we concur with Huene and Matley about the distinctions in femoral proportions ("slender" vs "robust") see11 iii the femora of large abelisaurids.However, referral of this set of hindlimb bones either to Indosuchus o r Indosaurus is not warranted a t the moment.
As already mentioned in the Introduction, t h e ahelisaurid Lametasaurus indicus was found in the "Carnosaur b e d .Since this taxon characterizes by a stout and transversely wide tibia, we interpret the short and stout femora (GSI K271558, GSI K271570, GSI K271618) and tibiae (GSI K2715681 discovered in the same quarry as probably belonging to Larnetasaurus indicus.Moreover, it would not be dismissed that Lametasaurus indicus may constitutes a senior synomym of the also robust Indosaurus rnatleyi and Rajusaurus narmudensis. The fragmentary n a t u r e of t h e Indian abelisaurids, including Lametasaurus, Rajusaums, Indosaurus and Indosuchus (except specimen IS1 RBI11 referred to as Indosuchi~s raptor-ius, still awaiting a detailed description), obstructs easy recognition of the taxonomic validity of each of these taua.Eventual solution of their respective taxonomic validity will need direct comparisons among all specimens as well as new, more complete discoveries.We keep the names Indosuchus, Indosaurus, Rajasaurus and Larnetasaurus pending new studies or more discoveries that might clarify whether these taxa can be diagnosed on the basis of autapomorphies.
Although our knowledge on the anatomy and phylogeny of Indian abelisauroids is far from being settled, it seems clear that these theropods were numerically dominant and taxonomically diverse in the Late Cretaceous of India, as documented in different fossil sites of the Lameta Formation (e. g., Bara Simia, Rahiolij, a view that is in concert with the information available from other Gondwanan localities of Late Cretaceous age (Madagascar, Patagonia).

ACILWOWLEDGEMENTS
The senior author received support from Agencia Nacional de Promoci6n Cientifica y Thcnica, The National Geographic Society, CONICEX, Appendix.List of theropod specimens described by Huene and Matley (19331,