Cretaceous Research 30 (2009) Contents lists available at ScienceDirect. Cretaceous Research

Transcription

1 Cretaceous Research 30 (2009) Contents lists available at ScienceDirect Cretaceous Research journal homepage: Anurans from the Lower Cretaceous Crato Formation of northeastern Brazil: implications for the early divergence of neobatrachians Ana M. Báez a, *, Geraldo J.B. Moura b, Raúl O.Gómez a a Departamento de Geología, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Ciudad Universitaria, 1428 Buenos Aires, Argentina b Doutorado em Ciências Biológicas, Universidade Federal da Paraíba, João Pessoa, Brazil article info abstract Article history: Received 3 October 2008 Accepted in revised form 8 January 2009 Available online 19 January 2009 Keywords: Amphibia Hyloides Nobleobatrachia Archaeobatrachians Araripe Basin South America The upper Aptian-lower Albian lacustrine limestones of the Crato Formation of northeastern Brazil have yielded one of Gondwana s most important Cretaceous fossil assemblages. This assemblage includes a few articulated anuran remains that have been previously referred to a single neobatrachian taxon, Arariphrynus placidoi [Leal, M.E.C., Brito, P.M., Anura do Cretáceo Inferior da Bacia do Araripe, Nordeste do Brasil. In: Gallo, V., Brito, P.M., Silva, H.M.A., Figueiredo, F.J. (Eds.), Paleontología de Vertebrados. Grandes Temas e Contribuçoes Científicas. Interciencia, Rio de Janeiro, pp ]. Herein we redescribe these specimens, which document two additional genera and species, Eurycephalella alcinae and Cratia gracilis, as well as a possible pipoid. Although the monophyly of neobatrachians can be considered a well-corroborated hypothesis, neobatrachian interrelationships are still far from being satisfactorily resolved. In order to address the high-level relationships of the taxa to which these specimens belong, we conducted a phylogenetic analysis of a matrix of 42 taxa, including extant representatives of most of the higher groups of neobatrachians as well as non-neobatrachians, and 75 mostly osteological characters in TNT 1.1 under implied weights with different values of the concavity constant (k). As in other analyses based on morphological data, within Neobatrachia we recovered a monophyletic Ranoides but hyloid taxa appear as stem-ranoids. Our analysis consistently place A. placidoi and E. alcinae in nested positions among hyloid taxa, although the topology of the tree varies slightly, whereas C. gracilis appears to be a stem neobatrachian or have a basal position within crown Neobatrachia. Recent studies based on molecular data have estimated divergence times for several anuran clades and proposed the main radiation of hyloid neobatrachians, excluding the australobatrachians, as a Late Cretaceous-Paleogene event. The taxonomically diverse anurans from the Crato Formation show that some hyloid lineages might have diverged already by the mid Cretaceous and that the early history of neobatrachians is as yet not documented in the fossil record. Ó 2009 Elsevier Ltd. All rights reserved. 1. Introduction Anurans constitute one of the most morphologically distinct groups of vertebrates. Most living anurans belong to Neobatrachia (Reig, 1958), a group that has been recognized by traditional systematists but also recovered by recent phylogenetic analyses based on morphological (Haas, 2003), as well as DNA-sequence data (Frost et al., 2006; Roelants et al., 2007; Wiens, 2007a). The broad picture of neobatrachian evolution, however, is far from being resolved; the branching pattern is still a subject of intense disagreement (Wiens, 2007b; Frost et al., 2008). Notwithstanding, * Corresponding author. address: baez@gl.fcen.uba.ar (A.M. Báez). according to these studies many long-accepted neobatrachian groups of diverse systematic position, such as Leptodactylidae, Hylidae, Telmatobinae, and Bufo, are not monophyletic in their traditional concepts. Until recently, the earliest unquestionable neobatrachian was from the Turonian Santonian Araçatuba/Adamantina Formation of central-eastern Brazil. This record consists of a small collection of well-preserved specimens, which still remains undescribed, interpreted by Carvalho et al. (2003) as belonging to a neobatrachian taxon that shares features with some South American hyloids (sensu Frost et al., 2006). More recently, however, the presence of neobatrachians in the older Crato Formation of northeastern Brazil, a unit famous for containing one of Gondwana s most important Cretaceous Konservat Lagerstatten (Maisey, 1991), was reported by Leal and Brito /$ see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi: /j.cretres

2 830 A.M. Báez et al. / Cretaceous Research 30 (2009) (2006). These authors briefly described the fossil material and assigned it to a new genus and species, Arariphrynus placidoi, of the nonmonophyletic Leptodactylidae. Our examination of the same specimens revealed that several features have been misinterpreted by these authors and that the fossils belong to more than one anuran taxon. Moreover, the specimens attributed to Arariphrynus placidoi include a mixture of non-neobatrachian and neobatrachian material. Herein we redescribe these specimens and, in order to address the high-level relationships of the taxa to which they belong, we performed a parsimony analysis using 75 morphological characters scored in 39 extant anuran species. These species represent 2 costatan, 4 anomocoelan, and 22 neobatrachian families (Appendix 1) following the taxonomy of Frost (2008). 2. Geological setting and preservational aspects The Crato Formation is part of the infill of the Araripe Basin, a small intracratonic basin of north-eastern Brazil (Assine, 1992) that lies on the borders of the states of Ceará, Pernambuco, and Piauí (Fig. 1). The tectonic evolution of the Araripe Basin, as well as that of other basins in this region, was closely linked to the opening of the South Atlantic Ocean. Sequences composed of genetically related depositional systems in these basins have been correlated with major episodes in the tectonic history of the rifting involved in the final separation of Africa and South America (Medeiros et al., 2001). In this context, the mainly continental Santana Group, which comprises the Rio da Batateira, Crato, Ipubi, Romualdo, and Arajara formations, represents the postrift stage in the Araripe Basin (Martill, 1993; Neumann and Cabrera, 1999). However, there is still little consensus on the stratigraphic scheme and nomenclature of the Araripe Basin infill (See Martill, 2007a for a review). The frog-bearing sequence, the Crato Formation, is up to 100 m thick and comprises several carbonate units separated by sandstones and shales; each carbonate unit is mainly made up of laminated limestones interbedded between fine-to-medium grained siliciclastics (Viana and Neumann, 2002).The Crato Formation laterally interdigitates with the sandstones and siltstones of the upper part of the Rio da Batateira Formation according to Neumann (1999). The interpretation of the environmental setting of the laminated limestones is contentious. The Crato succession has been interpreted as deposits of an extensive paleolake (Neumann et al., 2002) situatedat10 15 S of the paleoequator (Martill, 2007b). Cyclical fluctuations in the productivity cycle, perhaps related to vast algal blooms, may have induced the formation of individual carbonate laminae (Heimhofer et al., 2006). It has also been proposed that deposition of the laminated carbonates occurred in a thermally stratified lagoon; rivers from the north and east flowed into waters of the lagoon, where freshwater conditions prevailed (Brito et al., 1998; Martill, 2007b; Martill et al., 2007a; Dietze, 2007). The climate during the deposition of the Crato Formation has been interpreted as semiarid with a marked dry season, based on palynological data (Lima, 1978). Recently, the study of some spore and pollen assemblages has suggested that plant communities supporting seasonally dry as well as generally moist conditions were developed during Crato time, but their chronological and geographical patterns of distribution can not be inferred on the basis of the few available samples (Batten, 2007). The age of the Crato Formation has been considered as late Aptian according to ostracods (Berthou et al., 1994) andpalynomorphs (Pons et al., 1990; Arai et al., 1997; Coimbra et al., 2002; Batten, 2007), although some palynological data suggest that it might be early Albian (Lima, 1978, 1980; Hashimoto et al., 1987). Outcrops of the Crato Formation along the northeastern margin of the basin have been mined extensively, especially between the towns of Santana do Cariri and Nova Olinda, and around Barbalha, where the frog remains were discovered (Fig. 1). Most of the fossils described from the limestones were collected by local workers; consequently, many lack detailed stratigraphic provenance data, as is the case with the material described herein. Exceptionally well preserved remains of algae, fungi, plants (ferns, gymnosperms, angiosperms), invertebrates (insects, diplurans, arachnids, myriapods, decapods), and vertebrates (fishes, turtles, crocodylomorphs, pterosaurs, dinosaurs, birds) have been recovered from the Crato succession (Moura et al., 2006; Martill et al., 2007b), an assemblage that shows what life was like in an Early Cretaceous low-latitude lacustrine system. However, the anuran remains are comparatively rare. They occur mostly as incomplete skeletons that retain some degree of three-dimensionality. Some of these specimens (e.g. MPSC-An 890, 891, and 893) are fully articulated and preserve the hind limbs, which, remarkably, keep even the delicate terminal phalanges. This completeness indicates that the remains were deposited under low-energy conditions and that they were neither Fig. 1. A, location of the Araripe Basin in northeastern Brazil. B, map of the Araripe Basin showing the location of the quarries of the Crato Formation from which the anurans were discovered.

3 A.M. Báez et al. / Cretaceous Research 30 (2009) transported for a long distance nor were disturbed by scavengers. Postures in which the skeletons are preserved (crossed hind limbs, inwardly curled digits) suggest that some of them (MPSC-An 893, 1189) underwent subaereal exposure before burial (Moura and Barreto, 2006). Other specimens (e.g., MPSC-An 892, 894) are partially articulated and less complete, and might be allochtonous with respect to the depositional site. 3. Previous work on the anurans from the Crato Formation The first mention of anuran remains from the Crato Formation (as Crato Member of the Santana Formation) was by Kellner and Campos (1986) in a brief note. Subsequently, Maisey (1991) published the illustration of one specimen that preserves the articulated skeleton on two slabs of limestone as part and counterpart, in his atlas on the Santana fossils. The same specimen was also mentioned by Báez (1991), who commented on the poor preservation that makes it difficult to interpret its systematic placement. This specimen is now housed at the Direção NacionaldaProdução Mineral, in Rio de Janeiro, but we have been unable to locate it. Other anuran specimens were recovered from the Crato Formation and deposited in the Museum of Paleontology at Santana do Cariri (MPSC), near the locality of Crato, Estado de Ceará. Leal and Brito (2006) erected the new genus and species Arariphrynus placidoi for six of these specimens: MPSC-Ap 893 (holotype), 890 (paratype), 891, 892, 894, and 138. These authors briefly described the taxon and referred it to the Leptodactylidae, although no justification was given for this action. In addition, the diagnostic combination of characters proposed by these authors is insufficient to distinguish this taxon from many other neobatrachian taxa. Subsequently, Leal et al. (2007) published a note on A. placidoi and on a new specimen, MPSC-Ap 1189, in a review of the available information on the Crato lagerstätte (Martill et al., 2007b). No synapomorphy supports the pipoid affinities of MPSC-Ap 1189 suggested by Leal et al.; moreover, the well-developed mentomeckelians (Leal et al., 2007, figure 13.2 C) indicate that this referral is in error. Examination of the fossils described and assigned to Arariphrynus placidoi by Leal and Brito (2006) demonstrates that they belong to more than one taxon, as discussed below. Furthermore, several bones were misidentified or poorly described by these authors, making it necessary a thorough redescription to evaluate their phylogenetic placement. 4. Systematic palaeontology Amphibia Gray, 1825 Anura Fischer von Waldheim, 1813 Neobatrachia Reig, 1958 Arariphrynus Leal and Brito, 2006 (pars) Type species. Arariphrynus placidoi Leal and Brito, 2006, p. 146 Diagnosis. As for the type and only known species. Arariphrynus placidoi Leal and Brito, 2006, p.146 (pars) Figs. 2 and 3B Type horizon and locality. Crato Formation (Aptian Albian); northeastern fringe of the Araripe Basin, near the town of Crato, Estado de Ceará, Brazil. Type specimen. MPSC-Am 893, articulated, relatively complete adult specimen. Emended diagnosis. Small anuran that differs from all nonneobatrachians in having free palatines (¼neopalatines of Trueb (1993)); it further differs from leiopelmatids and discoglossids by lacking free ribs and transverse processes on the urostyle, from leiopelmatids also in having eight presacral vertebrae, and from pipoids in having a T-shaped parasphenoid and well-developed mentomeckelians. It can be distinguished from all other known neobatrachians in the following combination of characters: dentate maxilla with a smooth pars facialis; broad skull-roof in the orbital region, with its width diminishing posteriorly; robust pterygoid with a long medial ramus, the anteriorly-posteriorly expanded end of which contacts the parasphenoid ala; highly ossified crista parotica that extends laterally beyond the otic capsule; occipital condyles widely separate; atlas not fused to Vertebra II; atlantal centrum bearing a depressed posterior condyle; transverse processes on Vertebrae II, III, and IV slightly expanded distally and more robust and long than those on the posterior presacrals; transverse processes of last presacrals slightly anteriorly oriented and as long as the sacral diapophyses; sacral diapophyses narrow; articulation of sacrum and urostyle bicondylar; shaft of the scapula moderately long; proximal tarsals relatively slender, the fibulare specially so; Distal Tarsals II and III fused and articulated to Metatarsals II and III; terminal phalanx of Toe I distally bearing a hemispherical protuberance separated from the body of the phalanx by a distinct constriction. Description of the type specimen. This specimen consists of a single articulated adult individual preserved on a limestone slab (Fig. 2A). Almost the entire skeleton is exposed in ventral view although, because the olfactory capsules and the braincase in the orbital region are missing, some elements of the cranial roof, such as nasals and frontoparietals, are visible. The snout-vent length is about 39 mm. Cranial skeleton. The skull is wider than long; its estimated maximum width is 41% of the snout-vent length. The length of the preorbital region represents one third of the medial length of the skull. The articulation of the lower jaw lies posterior to the level of the occipital condyles (Fig. 2B and C). The nasals are paired laminar bones. The medial regions of both nasals are missing: this makes it difficult to determine whether they were in contact with one another. The irregular anterolateral margin of the left nasal indicates that the bone is abraded; by contrast, the smooth border of the contralateral element shows that the nasals had a slightly convex anterolateral margin, lacking distinct anteromedial processes. Laterally each nasal extends towards the maxillary arch and forms the anteromedial margin of the orbit. Posterolaterally, the nasals are in contact with the anterolateral edge of the frontoparietals. The well-ossified frontoparietals are exposed ventrally, but fragments of the sphenethmoid conceal their anterior portions. It is evident that these dermal bones formed conspicuous elements of the roof of the skull, which is wider anteriorly than posteriorly in the orbital region owing to their slightly convergent lateral margins (Fig. 2B and C). The frontoparietals might have been separated medially from each other along the anterior one third of their length, although the shape and size of this fontanelle are difficult to determine because the irregular medial edges of these bones suggest that part of them might have been weathered away. Posteriorly, the frontoparietals are in medial contact even though a suture is not clearly evident thus suggesting that they might be partially fused to one another. At the level of the anterior margin of the otic capsules, bilateral curved scars that reflect the morphology of the dorsal wall of the chondrocranium are visible on the ventral surface of the frontoparietals. The shape of these scars indicates that the taenia tecti medialis and the taenia tecti transversalis were at least partially developed; this suggests that the roof of the

4 832 A.M. Báez et al. / Cretaceous Research 30 (2009) Fig. 2. Arariphrynus placidoi Leal and Brito, holotype (MPSC-Am 893). A, general view of the specimen. B, close up of the skull and anterior postcranial region. C, interpretive drawing of the skull and anterior postcranial region. Abbreviations: asp, angulosplenial; cle, cleithrum; co,coracoid; cpar, crista parotica; d, dentary; fp, frontoparietal; m, maxilla; me, mentomeckelian bone; n, nasal; oc, occipital condyle; pa fen, impression of margins of parietal fenestra; pal, palatine; pro, prootic; ps, parasphenoid; pt, pterygoid; qj, quadratojugal; sc, scapula; sph, sphenethmoid; vo, vomer. chondrocranium was perforated by paired parietal fenestrae. Posteriorly, lateral extensions of the frontoparietals overlap the medial portion of the otic capsules. Most elements of the palatal region are weathered away, except the palatines (¼neopalatines) and an elongate piece of bone that we interpret as part of the postchoanal process of the vomer. The palatines are narrow, slender bones that lie near the antorbital margin on both sides of the skull. The left palatine is posteriorly rotated making it clear that it lacks an articulation with the maxilla. The ventral surfaces of these bones are slightly convex and lack ridges or odontoids (Fig. 2B and C). The preserved portion of the left vomer has a rounded, smooth margin that indicates that it corresponds to the distal end of the postchoanal process; it is evident that this process was not in contact with the palatine. The parasphenoid is only represented by the most distal parts of the posterolateral alae, which are oriented nearly at right angles to the longitudinal axis of the skull. The pterygoids are remarkably well ossified; the anterior, medial, and posterior rami are nearly equally long. The anterior ramus articulates with the maxilla at the level of the anterior margin of the orbit; laterally this ramus bears a groove for the pterygoid process of the palatoquadrate cartilage, whereas posterolaterally a small ventral flange is evident. The medial ramus forms a bony articulation with the anteroventral surface of the otic capsule; the proximal end of this ramus overlaps the corresponding posterolateral ala of the parasphenoid. The posterior ramus articulates with the pars articularis of the quadrate at about the level of the occipital condyles. We were unable to identify any portion of bone as part of the squamosals, except for an elongate fragment that might belong to the ventral ramus of the left squamosal. No components of the plectral apparatus are preserved. The premaxillae are missing. The maxillae are slightly rotated anticlockwise with respect to their longitudinal axes; this resulted in the partial exposure of the smooth external surfaces whereas the internal aspect is not visible in either of the two bones. In addition, this rotation makes the skull look wider than actually is. Each maxilla is a relatively long element that extends from its articulation with the premaxilla, anteriorly, to near the level of the occipital condyle, posteriorly. Whether this bone articulated with the quadratojugal is impossible to determine because only the most posterior part of the right quadratojugal is preserved. The pars

5 A.M. Báez et al. / Cretaceous Research 30 (2009) facialis of the maxilla is relatively high anteriorly; it forms a frontal process at the level of the anterior margin of the orbit although the shape and extent of this process cannot be described because it is not entirely visible. Posterior to the frontal process, the height of the pars facialis decreases abruptly. The ventral margin of the maxilla is abraded, but a series of regular, rounded cross sections of teeth are visible, at least along the anterior half of the bone (Fig. 2B and C). The sphenethmoid is poorly preserved. We interpret the fragments of bone that lie ventral to the nasals and frontoparietals as part of this element; they occur anterior to and between the anterior portions of the frontoparietals, thus suggesting that the sphenethmoid might have been dorsally exposed. The prootics and exoccipitals are not well preserved, but it is evident that they are synostotically fused to form the otic capsules and occipital condyles. The configuration of the anteromedial region of the floor of the otic capsule indicates that the prootics and the sphenethmoid were not fused; in contrast, the former bones appear to be fused to the frontoparietal. Dorsolateral to the inner ear region, each prootic forms the well-ossified, wide crista parotica which is visible behind the pterygoids on both sides of the skull. The skull is articulated to the vertebral column, but the position of the occipital condyles is evident because they stand out as two widely separated protuberances (Fig. 2B and C). The lower jaw is composed of angulosplenials, dentaries, and mentomeckelians. The angulosplenial extends from about the level of the posterior half of the antorbital region to the pars articularis quadrati. The poor preservation of the articular portion prevents description of its shape. The dentary invested the Meckel s cartilage laterally from the symphysial region to a level that coincides with the midlength of the orbit. Pieces of bone attached to the left dentary in the symphyseal region document the presence of mentomeckelian bones. Axial skeleton. The vertebral column is composed of eight presacral vertebrae, the sacrum, and the urostyle. The anterior margin of the atlas is not clearly visible although it is evident that the cervical cotyles are widely separated medially by a straight intercotylar region; the relatively flat atlantal centrum bears a solid posterior condyle; this indicates that the succeeding vertebral centrum was anteriorly concave. All the other presacral centra are weathered away; thus, Presacrals II to VII are represented only by the respective neural arches and transverse processes, which are exposed in ventral aspect (Fig. 2A). The neural arches of adjacent vertebrae are relatively narrow and not imbricate; their posterior margins project in moderately developed neural spines. The last presacral (VIII) preserves the posterior portion of the centrum articulated to that of the sacral vertebra. The transverse processes of Vertebra II are distally expanded and anteriorly directed; those of the two succeeding presacrals (III and IV) are perpendicular to the axis of the column, distally expanded, and wider and more robust than the transverse processes of the other presacrals. The following two transverse processes (borne by Vertebrae V and VI) have a similar development; the transverse processes of Vertebra V are horizontally oriented, whereas those of the following vertebra are directed slightly anteriorly. The last two presacrals bear anteriorly directed transverse processes. The width between the distal margins of contralateral transverse processes on all vertebrae nearly equals the distance between the lateral margins of the sacral diapophyses, except those between the transverse processes on Vertebrae III and IV, which are slightly greater. There is no evidence of free ribs. The sacrum is formed by Vertebra IX. The sacral diapophyses are narrow and deflected posteriorly. The sacral centrum posteriorly bears two well-separated condyles for the articulation with the urostyle (Fig. 2A). The urostyle lacks transverse processes; its length is estimated to have been equivalent to that of the last seven presacrals, even though the most posterior part of this element is obscured by the ilia. Pectoral girdle. The pectoral girdle is represented by the right scapula, the glenoidal end of the right coracoid, the distal part of the left scapula, and the left cleithrum (Fig. 2B and C). The scapular shaft is long and stocky, lacking an anterior lamina; its proximal end bears an anteroposteriorly expanded pars acromialis and a pars glenoidalis forming the lateral (dorsal) wall and floor of the glenoid, although the partes are not clearly exposed because they are crushed against the otic capsule. A medial notch separating the pars acromialis from the pars glenoidalis is not visible probably owing to the partial overlap of both partes. The proximal portion of the coracoid, the only part that this bone preserves, is laterally expanded. The right cleithrum is partially concealed by the lateral portion of the scapula of the same side, but the anterior prong that must have rested on the dorsal surface of the suprascapular cartilage and enclosed its leading edge is clearly visible. Forelimb. The humerus is only represented by its distal portion exposed in ventral view. The diaphysis of this bone is robust and distinctly distally expanded. Ventrally it bears a relatively large and well ossified ball flanked by the well-developed medial and lateral epicondyles; a deep fossa occurs anterior to the humeral ball. The radioulna is poorly preserved, although it is evident that its proximal end has a well-developed olecranon process whereas its distal end is expanded. The autopodium is not preserved. Pelvic girdle. The pelvic girdle is represented by the articulated ilia and ischia; the pubes seem to have remained cartilaginous. The preacetabular length of the iliac shaft equals the length of the seven last presacrals. A shallow groove is visible along the posterior two thirds of the length of the shaft; this indicates that a dorsal crest was present. The ventral acetabular expansions of both ilia are broken off and partially hidden by a fragment of the proximal part of the left femur; this precludes estimation of their extent. The ischia are fused synostotically; each of these bones forms the posterior part of the acetabulum and develops a posterior expansion, the length of which is one half that of the acetabulum. Hind limb. The femur is 47% of the snout-vent length and 87% the length of the tibiofibula. No crests are present. In turn, the length of the proximal tarsus is nearly 44% that of the femur. The tibiale is shorter than the fibulare; its proximal end overlaps the ventral surface of the proximal end of the fibulare and was probably fused to the latter bone (Fig. 3B 1 ). The distal ends of the tibiale and fibulare are expanded, that of the former more widely so and might have been fused; these bones leave a narrow oval window between them. A wide distal tarsal is preserved at the base of Metatarsals II and III on both feet; it corresponds to the fused Distal Tarsals 2 and 3. Lengths of the metatarsals decrease in the following order: IV, V, III, II, I. Although Metatarsal I is the shortest of these bones, it is distinctly robust and slightly sigmoid; its distal end is more widely expanded than those of the other metatarsals. The phalangeal formula is The only terminal phalanx well enough preserved to permit description of its distal end is that of Toe I. The body of this phalanx is straight and its pointed tip bears a single hemispheric protuberance that is separated from the rest of the phalanx by a distinct constriction (Fig. 3B 2 ). Terminal phalanges of Digits I and II are clearly shorter and more robust than those of the outer digits available for examination. No element that might be assigned to the prehallux is evident; thus, if present, this element might have been cartilaginous considering the completeness of other pedial parts. The apparent articulation of Metatarsal I with the distal end of the tibiale might be due to a postmortem shrinkage of soft tissues and cartilaginous structures.

6 834 A.M. Báez et al. / Cretaceous Research 30 (2009) Fig. 3. Comparison of the pedes of Eurycephalella alcinae n. gen et sp. and Arariphrynus placidoi Leal and Brito. A, E. alcinae, holotype (MPSC-Am 890), drawing of feet (1) and photograph of terminal phalanx of Toe I (2). B, A. placidoi Leal and Brito, holotype (MPSC-Am 893), drawing of feet (1) and photograph of terminal phalanx of toe I (2). Arrowheads mark the terminal phalanx of Toe I. Abbreviations: Dt1, Distal Tarsal 1; Dt2 þ 3, fused Distal Tarsals 2 and 3; fi, fibulare; ph1, prehallical element 1; ti, tibiale. Remarks. This specimen was described and figured by Leal and Brito (2006), but we disagree with several of their interpretations. Leal and Brito (2006, p. 147, fig. 1) probably misidentified some elements because, despite the ventral exposure of this specimen, some bones of the dorsal part of the skeleton are visible owing to the post-mortem destruction of ventral parts. Hence, the conspicuous nasals were identified as the palatines, whereas portions of the latter were not identified as these bones. In addition, the left cleithrum was thought to be the scapula and the neural arches, exposed in ventral aspect, were interpreted as vertebral centra. There is no evidence that the posterior terminus of the maxillae extended to the level of the occipital condyles; the stocky element labeled as quadratojugal on the left side of the skull by Leal and Brito (2006, p. 147, fig. 1) is probably the pars articularis of the quadrate. The piece of bone identified as the stapes is considered herein as an expansion of the medial ramus of the pterygoid crushed on the otic capsule; similar portions of bone are preserved on both sides of the skull in a comparable location and we were unable to distinguish them as elements separate from the pterygoids. Scrutiny of this specimen revealed that the atlantal cotyles are well separated from one another and, thus, they do not correspond to Type II of Lynch (1971), as Leal and Brito (2006) specified in the diagnosis of Arariphrynus, but to Type I of the same scheme. Because only the atlantal centrum is preserved, the types of articulation between adjacent vertebrae throughout the vertebral column cannot be determined, contrary to the statement by Leal and Brito (2006) that Arariphrynus has procoelous vertebrae. Although a bicondylar sacro-urostylar articulation was inferred for this taxon by these authors, they failed to identify the sacral condyles, as their interpretive drawing and restoration shows. The identification of the bone between the bases of the two inner metatarsals on the right foot as the (displaced?) Distal Tarsal 3 is rejected here; this is actually a portion of the expanded base of Metatarsal I. Instead, a wide distal tarsal between Metatarsals II and III, not identified by Leal and Brito (2006), probably represents the fused Distal Tarsals 2 and 3 (Fig. 3B). A monophyletic Neobatrachia was recovered in recent analyses based on DNA sequence data (Frost et al., 2006; Roelants et al., 2007; Wiens, 2007a); however, no exclusive adult osteological synapomorphies for the basal node have been identified. Notwithstanding, the combined occurrence of free palatines (neopalatines of Trueb, 1993), broadly separated occipital condyles, solid atlantal centrum bearing a condyle, narrow sacral diapophyses, absence of free ribs, bicondylar articulation between sacrum and urostyle, scapula with a proportionally long shaft that lacks an anterior lamina and with an expanded pars acromialis, and modified distal tips of toes suggests that MPSC Am-893 represents a neobatrachian taxon. In addition, the likely presence of both taeniae medialis and transversalis in the chondrocranium also points to this conclusion because their joint occurrence is known in neobatrachian groups only; moreover, presence of the taenia tecti medialis is one of the apomorphic features that support the neobatrachian clade, excluding the extant South African Heleophryne in the analysis by Haas (2003). Although fusion of Distal Tarsals 2 and 3 to one another has occurred independently in a few non-neobatrachian lineages (Fabrezi, 1993), a single distal tarsal at the base of Metatarsals II and III has been recorded in living representatives of most neobatrachian lineages, including those that diverged early in neobatrachian history according to recent studies. This suggests that fused Distal Tarsals 2 and 3 might be a plesiomorphic condition within Neobatrachia.

7 A.M. Báez et al. / Cretaceous Research 30 (2009) The unsculptured maxillae, nasals with a slightly convex anterolateral margin, widely separated cervical cotyles, and slender scapula lacking an anterior lamina, among other features, clearly indicate that MPSC-Ap 893 is not referable to Baurubatrachus, a putative neobatrachian genus from the Upper Cretaceous Marilia Formation of northern Brazil (Báez and Perí, 1989; Báez et al., 2005). Eurycephalella gen. nov. Derivation of name. Noun modified from the Greek eurys, meaning broad, and kephala, meaning head, in reference to the wide skull. Type species. E. alcinae sp. nov. Diagnosis. As for the type and only known species. Eurycephalella alcinae sp. nov. Figs. 3A and Arariphrynus placidoi Leal and Brito, p.146 (pars), fig. 2C; 2007 Arariphrynus placidoi Leal, Martill, and Brito, p.446 (pars), fig. 13.2a, b. Type horizon and locality. Crato Formation (Aptian Albian); northeastern fringe of the Araripe Basin, near the town of Crato, Estado de Ceará, Brazil. Type specimen. MPSC Am-890; incomplete, articulated adult specimen. Derivation of name.the specific name honours Alcina Barreto, paleobotanist and professor at the Universidade Federal de Pernambuco, Brazil, enthusiastic director of many projects to study the fossils from the Araripe Basin. Diagnosis. Small anuran that differs from all non-neobatrachians in having grooved distal tips of the terminal phalanges of feet, and from all non-neobatrachians except anomocoelans (sensu Frost et al., 2006) in the possession of a long scapula and clavicle that abuts, but does not anteriorly overlap, the pars acromialis of the scapula. It further differs from pipoids by possessing the T-shaped parasphenoid and well-developed mentomeckelians. It can be distinguished from all other known neobatrachians, including Arariphrynus placidoi, in the following combination of characters: maxillary arch complete; maxilla bearing a shelf-like pars palatina and pedicellate teeth; nasals separate from one another, lacking rostral processes and anterolateral concavities surrounding the external nares; skull-roof in the orbital region broad with parallel lateral margins; sphenethmoidal ossification invading the tectum nasi, which is dorsally exposed between nasals; squamosal with a broad zygomatic ramus; palatine tapering medially and bearing a fine ridge along the medial half of its length; robust pterygoid with a long medial ramus; atlas and second presacral not fused; atlantal cotyles narrowly separated; scapula with long shaft, lacking anterior lamina and bearing a widely expanded pars acromialis that is larger than the pars glenoidalis; clavicle robust and anteriorly curved; humerus with a moderate deltoid crest along the proximal half of its length; Finger II longer than Finger III; proximal tarsals expanded at both ends, leaving an ample intertarsal window; Distal Tarsals II and III fused, the compound element articulating with Metatarsals II and III; Distal Tarsal I present; Metatarsals III V with distal postaxial expansions; terminal phalanx of Toe I distally bearing a pair of rounded protuberances separated by a distinct indentation. Description of the type specimen. This specimen consists of an articulated skeleton exposed in ventral view, which belongs to an adult individual based on the complete ossification of the braincase and autopodia of fore- and hind limbs, and posterior position of the articulation of the lower jaw (Fig. 4A). Although the skull and limbs are fairly completely preserved, the vertebral column and pelvis are represented by fragments of bone and faint impressions of bone. In addition, impressions of some soft-tissue structures (e.g., eyes and masticatory musculature) are evident. The snout-vent length is estimated to have been about 38 mm. Cranial skeleton. The skull has a distinct rounded outline and is wider than long. The maximum width of the skull is 47% of the snout-vent length. The length of the preorbital region represents one-third of the medial length of the skull. The articulation of the lower jaw is located posterior to the level of the occipital condyles (Fig. 4B and C). The flat paired nasals are moderate in size; despite their incomplete preservation, the presence of an extensively mineralized tectum nasi indicates that they were not in contact with each other along the midline. The anterolateral margin of each nasal is slightly convex, lacking a rostral (or anterior) process as well as an excavation circumscribing the external nares. A distinct process that nearly reaches the maxilla projects laterally from the main body of the bone; this process forms the most anterior margin of the orbit and should have covered the post nasal wall. The overall shape and nature of the contact of the frontoparietals with one another cannot be described because these bones are concealed by the sphenethmoid and otic capsules. On both sides of the braincase, it is evident that the frontoparietals were not fused to the sphenethmoid in the orbital region. Additionally, it is clear that anterolaterally these roofing dermal bones were in contact with the nasals. Posterolaterally, each frontoparietal extends over the medial portion of the dorsal surface of the corresponding otic capsule. Unlike other elements of the palatine region, remnants of the palatines are preserved on both sides of the skull (Fig. 4B and C). Each palatine broadens laterally and extends across the antorbital plane from the palatine shelf of the maxilla to the anterolateral end of the bony sphenethmoid. Although the lateral terminus of the palatine is obscured by the articulated lower jaw and pars palatina of the maxilla, it is evident that is separate from the inner surface of the pars facialis of the latter bone. A fine ridge extends along the medial half of the palatine. Fragments of the parasphenoid are crushed on the ventral surface of the braincase. The shape and position of some of these fragments permit us to ascertain the presence of subotic alae that extended laterally at right angles from the cultriform process. The anterior terminus of the fore-mentioned process is difficult to determine on the available evidence. The pterygoids are remarkably well ossified; the anterior ramus seems to be shorter than the nearly equally long medial and posterior rami. The anterior ramus articulates with a modest, but distinct, pterygoid process of the maxilla at the level of the posterior third of the orbital length. The medial ramus is notably long and rests on the anterior wall of the otic capsule along the medial half of its length; this ramus may have contacted the parasphenoid (Fig. 4B and C). The robust posterior ramus articulates with the pars articularis of the quadrate cartilage at the level of the second presacral vertebra. The squamosals are only barely visible anterior and lateral to the otic capsules. It is evident that the zygomatic process is well developed and anterolaterally directed, although whether it contacted the maxilla is difficult to determine unequivocally. The long ventral ramus of the T-shaped squamosal, which must have invested the lateral surface of the palatoquadrate cartilage, is visible next to the posterior ramus of the left pterygoid. The premaxillae are relatively narrow and bear well-developed alary processes. Laterally, each bone articulates with the corresponding maxilla, the pars facialis of which seems to have overlapped the premaxilla. Most of the partes dentalis and palatina of this bone are missing and the few portions that are preserved are

8 836 A.M. Báez et al. / Cretaceous Research 30 (2009) Fig. 4. Eurycephalella alcinae n. gen. et sp., holotype (MPSC-Am 890). A, general view of the specimen. B, close up of the skull and anterior postcranial region. C, interpretive drawing of the skull and anterior postcranial region. Abbreviations: asp, angulosplenial; cl, clavicle; cle, cleithrum; co,coracoid; d, dentary; fp, frontoparietal; fpro, prootic foramen; h, humerus; m, maxilla; me, mentomeckelian bone; n, nasal; pal, palatine; pm, premaxilla; pro, prootic; ps, parasphenoid; pt, pterygoid; qj, quadratojugal; sc, scapula; sph, sphenethmoid; sq, squamosal; vo, vomer. obscured by the articulated lower jaw; thus, the detailed ventral morphology of this element is unknown. The maxilla is a relatively long bone that extends from near the premaxillary alary process, anteriorly, to about the level of the occipital condyles, posteriorly. The tapering posterior end of the bone articulates with the quadratojugal. The high pars dentalis bears around 50 small pedicellate teeth; only a few blunt crowns are preserved but the number of cuspids on each crown was not possible to determine with certainty. The tooth row extends posteriorly to a point slightly posterior to the midorbit length (Fig. 4B and C). The pars palatina is a moderately wide shelf, whereas the shape and extent of the pars facialis are not possible to be described in detail owing to the position in which both maxillae are preserved. However, partial lateral exposure of these bones indicates that their external surfaces are smooth. The quadratojugals are well-ossified splintlike bones. Anteriorly, each quadratojugal overlaps the medial surface of the maxilla, whereas posteriorly the bone reaches the pars articularis of the palatoquadrate cartilage. The sphenethmoid was extensively ossified, surrounding the braincase in the orbital region. Pieces of the ventral surface of this bone are visible on both sides of fragments of bone that we interpret as part of the parasphenoid. The sphenethmoidal ossification appears to have invaded the tectum nasi anterodorsally, thus contributing to the formation of the roof of the nasal capsules; the extensive anterior extent of this bone suggests that it was dorsally exposed between the nasals. The sphenethmoid forms short anterolateral processes which indicate that the orbitonasal foramina were surrounded by bone. The ventral part of the sphenethmoid is crushed against the frontoparietal and partially overlapped by fragments of the cultriform process of the parasphenoid; even so, it is evident that its posterior terminus lies at the level of the posterior one third of the orbits. Owing to the high degree of ossification of the specimen, prootics and exoccipitals must have been indistinguishably fused to form the otic capsules and occipital condyles. However, because the pectoral girdle is anteriorly shifted, most of the ventral surfaces of the otic capsules are not visible. On both sides, and medial to the corresponding medial ramus of the pterygoid, each prootic is

9 A.M. Báez et al. / Cretaceous Research 30 (2009) pierced by the large foramen for the trigeminal nerve. The configuration of the anteromedial region of the floor of the otic capsule indicates that the prootic was not synostotically fused to the sphenethmoid anteriorly; thus, these elements must have been separated by cartilage in life. Laterally, distal portions of the crista paroticae are preserved, furnishing evidence that these structures were well ossified. Additionally, it is clear that the cartilaginous solum synoticum intervened between the exoccipitals. Hence, the margin of the foramen magnum was not completely encircled in bone. The occipital condyles seem to have been narrowly separated. The lower jaw is formed by angulosplenials, dentaries, and mentomeckelians. The angulosplenial extends from the anterior third of the orbit to the pars articularis quadrati. Its posterior third is slightly sigmoid and bears a feeble coronoid process. The dentary laterally invested Meckel s cartilage along the anterior three quarters of its length; fragments of bone at the symphysis document the presence of mentomeckelian bones. Axial skeleton. The atlas is partially hidden by the coracoids, whereas the succeeding two vertebrae are only represented by the neural arches and broken transverse processes and pedicles; all these elements are exposed in ventral view. The shape of the anterior margin of the atlantal centrum indicates that the occipital cotyles were not broadly separated and that an intercotylar notch was not present. Incomplete preservation prevents description of the type of occipital-cervical and intervertebral articulations, as well as the total number of vertebrae. The rest of the axial skeleton is only represented by faint impressions; despite the poor preservation it is evident that the sacral diapophyses were not widely dilated (Fig. 4A). Pectoral girdle. The pectoral girdle is represented by scapulae, clavicles, and coracoids of both sides of the body; the shape and arrangement of these elements indicate that the pectoral girdle was of arciferal type. The scapular shaft is notably long and slender, and distinctly curved dorsoventrally (Fig. 4B and C). The proximal end of the scapula bears an anteroposteriorly expanded pars acromialis and a pars glenoidalis that forms the lateral (dorsal) wall and floor of the glenoid. A medial notch separating the pars acromialis from the pars glenoidalis is not clearly visible due to the ventral exposure of the scapula on both sides. The clavicle is well ossified and anteriorly curved, although its length cannot be determined accurately because the most anterior part is missing from the left bone and that part of the right side is not exposed. The lateral end of the clavicle is expanded and abuts the pars acromialis of the scapula but it does not overlap its anterior margin. The coracoid is expanded at both ends, its sternal end more notably so (Fig. 4B and C). The expansion of the sternal end is somewhat asymmetrical, as it is more pronounced anterior than posteriorly. Forelimb. The humerus is a relatively slender but slightly sigmoid bone; it is as long as the median length of the skull. A moderately developed deltoid crest originates from the wellossified humeral head and extends along the proximal half of the bone. The medial epicondyle, the only one exposed, reaches the distal end of the humerus. The ventral ball is not entirely exposed owing to the position of the bone, but it appears to have been well developed. The length of the radioulna is 75% that of the humerus. The olecranon process forms a wide angle with the articular facet of the radial portion of the bone. The left autopodium is disarticulated from the radioulna and part of it has flipped over the latter bone; owing to carpal torsion, the postaxial carpal elements are on top of the preaxial ones (Fig. 4A). The elements of the proximal row of the carpus are obscured by the distal ones; a large element that is in contact with the bases of Metacarpals IV and V can be identified as Distal Carpal 4 þ 5. On the preaxial side, and partially covered by Distal Carpal 4 þ 5, there appears another bone that we interpret as Element Y. The impression of a prepollex formed by two or three elements is visible distally to Element Y. Two additional distal carpals are preserved, one at the base of Metacarpal II and another at the base of Metacarpal III. The longest and more robust metacarpal is metacarpal II. Metacarpal V is longer than Metacarpal IV, whereas Metacarpal III is only represented by a proximal fragment. Only Digits II and III preserve the terminal phalanges; which appear to be slightly knobbed. Digit II is longer than Digit III. Pelvic girdle. The pelvic girdle is represented by fragments of the iliac shafts and the distal portion of the articulated ischia. The preacetabular length of the ilium is only slightly greater than the median length of the skull. Hind limb. The femora are slightly sigmoid; although the proximal portions of both bones are actually missing, it is possible to estimate that their length was nearly 39% the length of the snoutvent length due to the clear traces that these bones, in their natural position, left on the sediment. The femur represents 85% the length of the tibiofibula, whereas it is nearly twice as long as the tibiale and fibulare. Both ends of the latter bones are expanded, leaving a wide oval space between them (Fig. 3A 1 ). The proximal end of the tibiale ventrally overlaps that of the fibulare and was probably fused to it; in contrast, the distal ends of these bones lie next to each other and might have remained unfused. Other tarsal elements are preserved distal to the tibiale and fibulare. We identify the wide bone at the base of Metatarsals III and II as the fused Distal Tarsals 2 and 3, whereas the small bone articulating with Metatarsals II and I might be Distal Tarsal 1. Another round element lays medial to the left Metatarsal I; this might be Element Y or the basal element of the prehallux. The distal half of Metatarsals III and IV and the corresponding proximal end of the phalanges to which they are articulated are distinctly expanded; each of these metatarsals bear a post-axial keel along the expansion. The lengths of the metatarsals decrease in the following order: IV, V, III, II, I. The phalangeal formula is The terminal phalanx of Toe I bears a distinct groove that separates two lateral rounded protuberances (Fig. 3A 2 ). The distal ends of the terminal phalanges of the next three toes are slightly curved ventrally and also bear lateral processes that seem to be slightly longer and flatter than those of toe I, whereas the terminal phalanx of digit V is missing in both feet. Remarks. This specimen was selected as a paratype of Arariphrynus placidoi and, thus, it is supposedly conspecific with MPSC-893. From the foregoing description, however, it is clear that MPSC-Am 890 differs from the holotype of A. placidoi in several features, despite the similar size and limited number of structures that are preserved well enough to permit detailed comparisons between both specimens. The outline of the skull of MPSC-Am 890 is distinctly rounded and contrasts with the more triangular shape of the skull of A. placidoi. Unlike the latter, the articulation of the lower jaw is located far posteriorly with respect to the occipital condyles. The condyles are not broadly separated, thus the configuration of the atlantal cotyles differ from that in A. placidoi. Although few components of the pectoral girdle are preserved in both specimens, it is evident that the scapula of MPSC-Am 890 has a longer and more distinctly waisted shaft, than that of A. placidoi. The hind limbs also have relative proportions different from those of Araryphrynus placidoi; the femur and the tibiofibula are comparatively shorter with respect to the snout-vent length than they are in the latter species (39% vs 47% and 46% vs 54%, respectively). Specimen MPSC-Am 890 also differs from the holotype of A. placidoi in having a noticeably more expanded fibulare and a broader window between this bone and the tibiale (Fig. 3). The expanded distal halves of Metatarsals III and IV contrast with the slender metatarsals of A. placidoi. In addition, the morphology of the terminal phalanx of Toe I is strikingly different in these specimens, particularly that of the distal end. The shape of the terminal

10 838 A.M. Báez et al. / Cretaceous Research 30 (2009) phalanges has been considered taxonomically significant (e.g., Lynch, 1971; Clarke, 1981; Laurent, 1986), although some doubt has been casted on the phylogenetic utility of this character above the genus level (Fabrezi, 1996). Variation of the morphology may occur during development, with reduction of the lateral processes (Noble, 1917; Fabrezi, 1996), but no variation among adults of the same species has been reported. Apart from the shape of the terminal tip, in MPSC-Am 890 this phalanx is longer with respect to its basal width (1.6) than in A. placidoi (1.4). It is also noteworthy that among archeobatrachians the distal tips of the terminal phalanges are simple, with very few exceptions (Clarke, 2007); this condition has been considered plesiomorphic for anurans. All this evidence indicates that MPSC-Am 890 and 893 belong to different taxa. The joint presence of a T-shaped parasphenoid, long scapula, palatine bone not fused anterodorsally to the pars facialis of the maxilla, sacral diapophyses not widely expanded, clavicle that abuts the scapula and does not anteriorly overlie its forward edge, grooved distal tips of terminal phalanges, fused Distal Tarsals 2 and 3, suggests that this specimen does not belong to an archaeobatrachian taxon, and, conversely, that its placement may lie within Neobatrachia. The nasal lacking a rostral process and an extensive contact with the maxilla, the ample orbit, and the long and slender shaft of the scapula contrast with the respective conditions in the Late Cretaceous Baurubatrachus (Báez and Perí, 1989). Cratia gen. nov. Type species. C. gracilis sp. nov. Derivation of name. From Crato, the name of the stratigraphic unit that yielded the type and only known specimen. Cratia gracilis sp. nov. Fig Arariphrynus placidoi Leal and Brito, p.146 (pars) Type horizon and locality. Crato Formation (Aptian-Albian); northeastern fringe of the Araripe Basin, near the town of Crato, Estado de Ceará, Brazil. Type specimen. MPSC-Am 891, articulated postmetamorphic, but possibly juvenile, individual. Derivation of name. The specific name derives from the latin gratia, refering to the delicate skeleton. Diagnosis. Small anuran that differs from all leiopelmatids and pipoids in having a dermal ossification across the ventral surface of the planum antorbitale and from the latter also by the presence of well-developed mentomeckelian bones. It differs from anomocoelans in the possession of weakly dilated sacral diapophyses. It can be distinguished from all other known neobatrachians in having the following combination of characters: maxillary arcade complete; dentate maxilla with a smooth, relatively deep pars facialis; skull-roof narrow; sickle-shaped nasals medially separate from each other; bony sphenethmoid anteriorly surrounding the frontoparietal fenestra; otic capsule with prominent epiotic eminences; crista parotica moderately ossified, distally cartilaginous; T-shaped squamosal with free-ending, long zygomatic ramus; pterygoid with a long medial ramus; slightly dilated sacral diapophyses; shaft of the scapula relatively short, lacking an anterior lamina and bearing equally developed partes acromialis and glenoidalis that are separated by a wide, medially oriented notch; pars glenoidalis lacking a dorsal ridge; clavicle slender and anteriorly curved; coracoid with glenoidal end more expanded than sternal end. Description of the type specimen. This specimen consists of the skeleton of a small frog, the snout-vent length of which is 29 mm, although the head might be disarticulated and slightly shifted forward (Fig. 5A). It partially preserves the skull, pectoral girdle, pelvic girdle, and limbs in dorsal aspect; however, some skeletal parts, such as the right side of the skull, were weathered away exposing their ventral impression on the sediment. This is also the case of the vertebral column, represented only by faint impressions of the ventral surface of the most posterior vertebrae and the urostyle. The highly ossified otic capsules, mentomeckelians, and quadratojugals indicate that this specimen belongs to a postmetamorphic individual. Cranial skeleton. The skull has a rounded overall outline, being slightly wider than it is long. The width of the skull is 45% of the snoutvent length and its length 33% of the same measure. The preorbital region represents one third of the medial length of the skull. The articulation of the lower jaw lies at the level of the occipital condyles (Fig. 5B and C). Fragments and impressions of the nasals indicate that they were moderate-sized bones and were medially separated from one another. Anteromedially each bone extends almost to the alary process of the premaxilla, whereas posterolaterally it nearly reaches the maxilla. Prominent promontories of sediment where the nasals were lying indicate that the posteromedial part of these bones was distinctly concave ventrally. Although the anterior extent of the frontoparietals is not possible to be determined with certainty, the nasals did not have an extensive contact with these bones, a feature that suggests that the dorsal surface of the sphenethmoid was partially exposed anterior to the frontoparietals and between the nasals. Pieces of a perforated small bone that lie anterior to the left nasal are identified as part of the left septomaxilla. In addition, clear traces of bone at the anterior margin of the orbits that lay ventral to the posterior margin of the nasals and dorsal to the lower jaw, and thus ventral to the post nasal wall in life, might correspond to the palatines. The frontoparietals are not preserved. Remnants of the squamosal on both sides of the skull allow us to ascertain that it was T- shaped. It has a free-ending, long, laterally oriented zygomatic process. The otic process is shorter than the zygomatic process and articulates with the anterolateral corner of the crista parotica. The ventral process is not completely preserved but the position of the pars articularis of the palatoquadrate cartilage suggests that it was relatively long. Obliteration of the right otic capsule revealed fragments and impressions of the pterygoid on this side of the skull, whereas on the left side only a few fragments of this bone are preserved (Fig. 5B and C). The pterygoid has a long medial ramus that forms a wide angle with the anterior ramus. The latter ramus articulates with the maxilla at midorbital level, although traces of bone medial to the maxilla might indicate that it extended anteriorly, at least along part of the anterior half of the orbit length. The posterior ramus is relatively short. The premaxilla bears a long, posterodorsally directed alary process, which is incompletely preserved on both sides. Other parts of the premaxilla are not visible. The maxilla has a high pars facialis that lacks dermal ornamentation. Posterior to the articulation with the premaxilla, the pars facialis of the maxilla increases in height dorsally for a short distance, and its keeps this level up to the anterior margin of the orbit. Further back, the bone is broken off but it is evident that it extends posteriorly as a low wall below the orbit. The posterior end of the maxilla terminates at the level of the midlength of the otic capsule, where it articulates with the quadratojugal. On the right side, partial loss of the pars facialis of the maxilla reveals an impression of the internal aspect of this bone which shows the presence of numerous teeth. It is evident that the maxillary tooth-row extended from the anterior end of the bone up to the level of the anterior margin of the subtemporal

11 A.M. Báez et al. / Cretaceous Research 30 (2009) Fig. 5. Cratia gracilis n.gen et sp., holotype (MPSC-Am 891). A, general view of the specimen. B, close up of the skull and anterior postcranial region. C, interpretive drawing of the skull and anterior postcranial region. Abbreviations: cl, clavicle; co,coracoid; d, dentary; h, humerus; l, lower jaw; m, maxilla; me, mentomeckelian bone; n, nasal; pm, premaxilla; pro, prootic; pt, pterygoid; q, pars articularis quadratum; qj, quadratojugal; sc, scapula; sm, septomaxilla; sph, sphenethmoid; sq, squamosal. fenestra; the shape of individual teeth can not be described in detail, but they are relatively high. The quadratojugal is a splintlike bone, the anterior lateral surface of which overlaps the medial surface of the most posterior part of the maxilla. Posteriorly, the quadratojugal extends to the pars articularis of the palatoquadrate cartilage. The sphenethmoid is only represented by impressions; it is relatively short forming the narrow braincase in the orbital region. Anteriorly, the sphenethmoidal ossification does not extend much into the septum nasi, medially, or the post nasal walls, laterally, whereas dorsally it surrounds the anterior margin of the frontoparietal fenestra. Its posterior terminus apparently reached the level of the midorbital length. The prootics form the conspicuous otic capsules that bear prominent epiotic eminences and extend laterally into rather wellossified parotic crests, although only the left otic capsule is completely preserved. Evidence from this side of the skull indicates that the exoccipitals were fused to the prootics and formed the narrowly, but distinctly, separated occipital condyles. Each half of the lower jaw is mainly composed of a relatively narrow angulosplenial and dentary. A conspicuous cylindrical ossification fused to the symphyseal portion of the left dentary is identified as a mentomeckelian bone (Fig. 5B and C). Axial skeleton. The total number of presacral vertebrae and their type of articulation can not be determined. The last four presacrals left faint impressions of their ventral surfaces; the cylindrical groove in the sediment left by this series of vertebrae indicates that the ventral surfaces of the centra were distinctly convex. At least the most posterior three presacral vertebrae bear short, laterally oriented transverse processes; the width between the distal ends of contralateral processes of the last three presacrals is shorter than the distance between those of sacral diapophyses. The sacral vertebra has narrow or slightly dilated diapophyses that are somewhat deflected posteriorly, although their distal ends are obscured by the impressions of the iliac shafts. The impression of the sacral centrum is too poor to ascertain the type of articulation with the urostyle. The latter has an expanded anterior end that seems to bear two shallow cotyles. It is not possible to ascertain whether the urostyle bore transverse processes because this element is not well exposed dorsally. The urostyle is shorter than the length of the presacral portion of the vertebral column.

12 840 A.M. Báez et al. / Cretaceous Research 30 (2009) Pectoral girdle. All preserved components of the pectoral girdle are exposed in dorsal aspect. The right scapula is entirely preserved (Fig. 5B and C). Its medial end bears nearly equally developed partes acromialis and glenoidalis that are separated from each other by a distinct notch. The pars glenoidalis lacks a dorsal ridge. The relatively short, symmetrical scapular shaft is waisted and slightly expanded distally. A groove for the articulation with the suprascapular cartilage is clearly visible along its distal edge. Anterior to the right scapula lies the strongly arcuate clavicle, indicating that it was not fused to the former bone. Both ends of the clavicle are missing; the incomplete preservation together with the lack of exposure of the articular facet of the pars acromialis of the scapula prevents detailed description of the articulation between the elements of the pectoral girdle. The coracoid is robust, having a glenoidal end more expanded than the sternal end. Forelimb. The humerus and radioulna are poorly preserved. The former is a relatively long bone bearing a moderate deltoid crest along the proximal one third of its length. Two phalanges in articulation with an incomplete metacarpal are visible on the right side of the skeleton. Pelvic girdle. This skeletal region is represented by impressions and portions of the articulated ilia and articulated ischia. The extended dorsal contact between the left and right ilia, the ends of which nearly reach the posterior margins of the acetabula, indicates that these bones had well-developed supracetabular expansions. Hind limb. The femur is as long as the tibiofibula. Its length is about 48% of the snout-vent length. The tibiale/fibulare is 50% the length of the femur. Incomplete right Metatarsals II and III with a few phalanges are preserved. Remarks. Specimen MPSC-Am 891 was ascribed to Arariphrynus placidoi by Leal and Brito (2006), although the close examination and comparisons with the holotype of this species and with MPSC- Am 890 demonstrated taxonomically significant differences that cannot be attributed to ontogenetic or individual variation and, conversely, indicate that it represents a third taxon. Among the few skeletal elements that MPSC-Am 891 preserves well enough to allow meaningful comparisons, the braincase and the scapula stand out as notably different. It is evident that this specimen has a much narrower braincase in the orbital region than those of A. placidoi and Eurycephalella alcinae, whereas the scapula has a lessdeveloped shaft, less expanded pars acromialis, which has a similar proportion than the pars glenoidalis, and a more medial orientation of the notch relative to those of the former two taxa. Additionally, MPSC-Am 891 differs from E. alcinae in having a premaxilla with a mediolateral length that represents nearly 25% the length of the maxilla instead of about 16%, more anterior position of the lower jaw articulation, coracoid with the glenoid end more expanded than the sternal end, and hind limb with different stylopodial, zeugopodial, and autopodial proportions, as described above. The poor preservation of many features that traditionally have been used to characterize anuran taxa above the genus level, such as the total number of presacral vertebrae, the configuration of vertebral centra, the presence of ribs, and the type of articulation between scapula and clavicle, obscures the systematic position of Cratia gracilis. Comparisons with the extant Ascaphus and Leiopelma show that the bases of the alary processes of the premaxillae are closer to each other than in these basal genera suggesting a different arrangement of the prenasal cartilages; also, the quadratojugals are well developed but are absent in these taxa. The conspicuous mentomeckelian bones, moderate-sized otic capsules, weakly dilated sacral diapophyses, and symmetrical scapula with a relatively short margin for the suprascapular cartilage rule out pipoid affinities. Xenoanura? Savage, 1973 Pipimorpha? Ford and Cannatella, 1993 Genus and species indeterminate Fig Arariphrynus placidoi Leal and Brito, p (pars) Horizon and locality. Crato Formation (Aptian Albian); northeastern fringe of the Araripe Basin, near the town of Crato, Estado de Ceará, Brazil. Referred material:mpsc-am 892, incomplete, partially articulated individual. Description. Specimen MPSC-Am 892 consists of a partially articulated skeleton that preserves some skull elements and the incomplete left forelimb, pelvis, and hind limbs (Fig. 6A). The snout-vent length is not possible to be determined. Although the superposition of the bones of the skull makes the interpretation of this region difficult, it is evident that the lower jaw is relatively short, suggesting that the articulation with the skull was anterior to the level of the occipital condyles. The humerus has a distinctly straight diaphysis in lateral view and a little expanded medial epicondyle that rests on a well-ossified ventralball(fig. 6B). The radioulna has a developed olecranon. The carpal bones are displaced over and next to the distal end of the radioulna; two large elements, each of which has a proximal oval articular facet, are tentatively identified as the radiale and ulnare. Distal to the latter bones there are other carpal elements but their identity is difficult to establish; remains of four long metacarpals and phalanges are also preserved (Fig. 6B). The pelvic girdle is represented by the left ilium. The latter bone has a long shaft with a round cross section, seemingly lacking a dorsal crest. The posterior end of the bone is sectioned, revealing an expanded symphyseal region. The femur is slightly sigmoid and shorter than the tibiofibula. The tibiale and fibulare are relatively long, their length represents about 56% that of the tibiofibula. The fibulare has a stouter diaphysis and more expanded ends than the tibiale; these bones seem to be synostotically fused to one another proximally and distally (Fig. 6C). Some distal tarsals are preserved; one large element, visible on both feet, occurs at the bases of Metatarsals II and III and, thus, we identify it as the fused Distal Tarsals 2 and 3. Another relatively large bone is visible between Metatarsals I and II on the right foot, which is preserved in dorsal aspect; the same element is preserved as a small bone at the base of the same metatarsals on the left foot, preserved in plantar aspect; we consider this element as Distal Tarsal 1 (Fig. 6C ).Ontheleftfoot there is a second element medial to Distal Tarsal 1 which might be Element Y. The metatarsals, mostly represented by actual bones, are nearly 73% the length of the proximal tarsals. The digits are preserved as impressions; their distal phalanges are straight and long. Remarks. This specimen was also referred to Arariphrynus placidoi by Leal and Brito (2006, p. 146). The relative proportions and morphology of the bones of the hind limb differ from those of A. placidoi. The proximal tarsals are distinctly longer than those of the latter species, where their length is less than one half the length of the tibiofibula and both ends of the fibulare are more widely expanded. In addition, the metatarsals are not as proximally and distally expanded as they are in Arariphrynus and are notable because Metatarsal IV is not longer than Metatarsals III and V, whereas in A. placidoi, aswellasineurycephalella alcinae, Metatarsal IV is markedly longer than the other metatarsals as in most anurans. The available evidence clearly indicates that MPSC-Am 892 is not referable to A.

13 A.M. Báez et al. / Cretaceous Research 30 (2009) Fig. 6. Pipoidea? (MPSC-Am 892). A, general view of the specimen. B, interpretive drawing of the left forelimb. C, interpretive drawing of the left pes in plantar view. Abbreviations: Dt1, Distal Tarsal; Dt 2 þ 3, Distal Tarsal 2 þ 3; fi, fibulare; h, humerus; Mc, metacarpal; Mt, metatarsal, ra, radiale; ru, radioulna; ti, tibiale; Y, Element Y. placidoi or to E. alcinae. It also differs from Cratia gracilis in having the femur shorter than the tibiofibula. The relatively long and straight metapodials, the straight diaphysis in lateral view and little expanded epicondyles of the humerus, and the presence of an interiliac symphysis suggest that it probably belongs to a pipimorph taxon. 5. Phylogenetic relationships Each of the taxa described above, Arariphrynus placidoi, Eurycephalella alcinae n. gen and sp., and Cratia gracilis n.gen and sp, possesses a suite of characters that in combination are consistent with neobatrachian affinity, as commented on previously. However, to test the allocation of these taxa to Neobatrachia and discuss their placements on the tree is troublesome because the available specimens are incompletely preserved and few morphology-based broad-scale parsimony analyses of anurans have been conducted. The analysis of Haas (2003) relies primarily on larval evidence that is unknown for the extinct taxa described herein. The latter author (2003) also included 13 adult morphological characters, although about half of them refer to soft tissue structures and hence are inapplicable to estimate the affinities of fossil specimens. In total, only four characters among those used by Haas can be coded unambiguously for the taxa from Crato. Most of the transformations series of morphology used by Haas (2003) were subsequently included together with DNA sequence data in the comprehensive analysis of Frost et al. (2006). Other analyses that consider osteological data have been focused on the evolutionary relationships of selected groups of neobatrachian taxa (e.g., hylids (Wiens et al., 2005); ranoids (Scott, 2005)). Fabrezi (2006) coded 81 morphological characters, including osteological features, for 62 anuran species representing 42 genera to address the relationships of ceratophrynes, a clade of hyperossified neobatrachians. Her taxon sampling, however, is insufficient to investigate the higherlevel relationships of the taxa described from Crato because this goal requires a more extensive sampling of Anomocoela, recovered as the sister group of neobatrachians by Frost et al. (2006) and Wiens (2007a), as well as of the putative basal groups of neobatrachians. Evans et al. (2008) used the same matrix as Fabrezi (2006) in their analysis to assess the phylogenetic position of the Upper Cretaceous Beelzebufo from Madagascar; although they expanded the sampling of taxa, including more anomocoelan and hyloid taxa as well as extinct putative neobatrachians, their terminal taxa are genera. In our analysis, ingroup terminals include 39 extant species representative of major acosmanuran clades recovered by Frost et al. (2006) (Appendix 1). In many cases the species were selected by the availability of dry skeletons and/or clear-and-stained specimens for personal examination, although a few species were coded from descriptions and illustrations in the literature (See Appendix 1). In light of Frost et al. s (2006) and Wiens s (2007a) results, predominantly based on molecular data, and that of Haas (2003), we used the costates Bombina variegata, Alytes obstetricans, and Discoglossus pictus as outgroups. With regard to the characters, we coded 75 adult morphological, mostly osteological, characters (Appendix 2). The majority of the characters that we coded are those listed in the analyses by Fabrezi (2006) and Evans et al. (2008), although these characters were previously used in other studies (e.g., Lynch, 1971, 1973, 1978; Clarke, 1981; Mendelson et al., 2000; Wiens et al., 2005; Scott, 2005) and many date back to classic works of the XIX Century. Some states were modified with respect to those analyses because they were not clearly delimited and to make more explicit the way they were coded herein; also, other states were added to include variation within the taxon sample. Autapomorphies of terminal taxa and invariant states among the sampled taxa are not phylogenetically informative and, thus, were excluded from the analysis. The dataset (Appendix 3) was analyzed using the heuristic search of TNT v. 1.1 (Goloboff et al., 2003a, 2008). Each search round consisted of 500 random-addition sequence Wagner builds followed by treebisection reconnection (TBR) branch swapping with a hold of 10. Assumption of equal weight resulted in 290 equally parsimonious trees of 372 steps (RCI ¼ 0.154) in which Acosmanura was consistently recovered, as well as a few subclades within the latter (Microhylidae, Ceratophryinae). In all trees Arariphrynus placidoi, Eurycephalella alcinae, and Cratia gracilis are within Acosmanura. In order to improve resolution additional searches were performed under implied weights with different values of the concavity constant k (1 15, 20, 30). Support for individual branches was