Paul C. Sereno, ^* Allison L. Beck, Didier B. Dutheil, Boubacar Gado, Hans C. E. Larsson, Gabrielle H. Lyon, Jonathan D. Marcot, Oliver W. M. Rauhut, Rudyard W. Sadleir, Christian A. Sidor, David D. Varricchio, Gregory P. Wilson, Jeffrey A. Wilson

Fossils discovered in Lower Cretaceous (Aptian) rocks in the Ténéré Desert of central Niger provide new information about spinosaurids, a peculiar group of piscivorous theropod dinosaurs. The remains, which represent a new genus and species, reveal the extreme elongation and transverse compression of the spinosaurid snout. The postcranial bones include blade-shaped vertebral spines that form a low sail over the hips. Phylogenetic analysis suggests that the enlarged thumb claw and robust forelimb evolved during the Jurassic, before the elongated snout and other fish-eating adaptations in the skull. The close phylogenetic relationship between the new African spinosaurid and Baryonyx from Europe provides evidence of dispersal across the Tethys seaway during the Early Cretaceous.

P. C. Sereno, A. L. Beck, H. C. E. Larsson, J. D. Marcot, C. A. Sidor, J. A. Wilson, Department of Organismal Biology and Anatomy, University of Chicago, 1027 East 57th Street, Chicago, IL 60637, USA. D. B. Dutheil, Laboratoire de Paléontologie, Muséum National d'Histoire Naturelle, 8 rue du Buffon, 75005 Paris, France. B. Gado, Institut pour Recherche et Science Humaine, Republic of Niger. G. H. Lyon, Small Schools Workshop, 115 South Sangamon Street, University of Illinois, Chicago, IL 60607, USA. O. W. M. Rauhut, Department of Geology, University of Bristol, Wills Memorial Building, Queen's Road, Bristol BS8 1RJ, UK. R. W. Sadleir, Department of Earth and Environmental Sciences, University of Illinois at Chicago, 845 West Taylor Street, Chicago, IL 60607, USA. D. Varricchio, Old Trail Museum, 823 North Main Avenue, Choteau, MT 59422, USA. G. P. Wilson, University of California, Museum of Paleontology, 1101 Valley Life Sciences Building, Berkeley, CA 94720, USA.
^*   To whom correspondence should be addressed.

Additional bones of Spinosaurus-like predators have been discovered in Lower Cretaceous (Aptian or Albian) deposits in Niger and Brazil and in somewhat older (Barremian) rocks in Europe. The Nigerien fossils include peculiar arched snout tips and enormous manual unguals (6-8); the Brazilian remains consist of a single partial skull, Irritator (9, 10); and the European fossils, Baryonyx, include the first relatively complete spinosaurid skeleton (11-13).

Fossils were recovered recently from the Elrhaz Formation in Niger (Fig. 1). These Aptian-age rocks consist predominantly of fluvial channel deposits and are exposed in low outcrops amid dune fields in the Ténéré Desert (6, 14). The fossils found here include plant, invertebrate, and vertebrate remains; the vertebrate remains consist mainly of disarticulated bones and teeth in basal channel lag deposits. Dinosaurs are represented by at least three theropods, two sauropods, and three ornithopods (15).

The newly discovered fossils include a partial skull and skeleton of a new spinosaurid, Suchomimus tenerensis gen. nov. sp. nov. (16), which can be distinguished from other spinosaurids (17). An articulated snout (Fig. 2A and B) reveals its remarkably long, low, and narrow proportions. The elongation of the snout is the result of the hypertrophy of both the premaxilla and the anterior ramus of the maxilla. The premaxillae, which fuse early in growth, each contain alveoli for seven teeth. The subconical crowns are slightly recurved and have fine marginal serrations and textured enamel surfaces (Fig. 2E). The external nares are retracted posterior to the premaxillary teeth (Fig. 2, A and D), as in Baryonyx (14). The unusually long, plate-shaped anteromedial processes of the maxillae (18) are firmly held by the premaxillae (Fig. 2B). The medial wall of the antorbital fossa is confined to the anterior end of the antorbital fenestra, and a simple conical pneumatocoel extends anteriorly into the body of the maxilla. As in Baryonyx, the quadrate foramen is very large, and the distal condyles are very broad.

The new cranial bones indicate that the spinosaurid skull is considerably lower, narrower, and longer (Fig. 2D) than previously reconstructed (13, 19). In dorsal view, the snout is extremely narrow (Fig. 2C). In ventral view, the maxillae meet along the midline, displacing the internal nares and palatal complex (pterygoid, palatine, and ectopterygoid) toward the rear of the skull.

In the postcranial skeleton, the cervical series arches upward (18) and has prominent epipophyses for muscle attachments. The neural spines increase in height rapidly in the middorsal vertebrae, forming a low median sail that is deepest over the sacral vertebrae (Fig. 3). This vertebral morphology, incipiently developed in Baryonyx (13), is distinct from that in Spinosaurus (1), in which the much deeper sail arches to an apex over the middorsal vertebrae (Fig. 4B).

Complete pectoral and pelvic bones show a deep subrectangular acromion on the scapula and a low obturator flange on the ischium. The distal end of the pubis is distinctive. Although there is a short, sagittal, symphyseal flange that may correspond with the pubic foot in other theropods, the flattened, subrectangular distal end of the pubis faces anteriorly, perpendicular to the sagittal plane (Fig. 3). The forearm is remarkably stout, and manual digit III is robust, as seen in the size of metacarpal III and the third ungual (Fig. 3 and Table 1). The femur has a blade-shaped anterior trochanter, and the ascending process of the astragalus is taller than that in Allosaurus (20).

Table 1. Length measurements (millimeters) and ratios in Suchomimus tenerensis (MNN GDF500), Baryonyx walkeri (13), and Allosaurus fragilis (20). Unguals are measured perpendicular to the articular end. Question mark indicates lack of preservation; I, II, and III are unguals of manual digits I, II, and III, respectively. Bone Suchomimus Baryonyx Allosaurus Lengths Humerus 560 483* 310 Radius 255 225 222 Metacarpal III 130 ? 101 I 190 173* 102 II 165 ? 92 III 120 120* 54 Femur 1075 ? 850 Tibia 945 ? 690 Ratios Radius/humerus 0.46 0.47 0.72 I/radius 0.75 0.77 0.46 II/I 0.87 ? 0.90 III/I 0.63 0.69 0.53 Tibia/femur 0.88 ? 0.81 * This measurement was taken by the authors and differs slightly from that reported previously (13).

Phylogenetic analysis (21, 22) links spinosaurids with torvosaurids (23) and places this clade (Spinosauroidea) as the sister group to Neotetanurae (Fig. 4A and Table 2). The derived features that are shared between spinosaurids and torvosaurids, including the short forearm and the enlarged manual digit I ungual (22), appear to have evolved by the Middle Jurassic (Fig. 4B). A hook-shaped coracoid characterizes Suchominus, Baryonyx, and neotetanurans but not torvosaurids (24). Thus, the hook-shaped coracoid either evolved convergently in neotetanurans and spinosaurids or originated as a tetanuran synapomorphy that was subsequently lost in torvosaurids (21, 22).

Table 2. Character-state matrix for two outgroups (Ceratosauria and Neotetanurae), six ingroups, and 45 characters (22) used in a phylogenetic analysis of spinosauroids (Fig. 4A). The holotypic specimens of Angaturama and Irritator were scored as one taxon (Irritator). X, unknown as a result of transformation; ?, not preserved. Taxa Characters 10 20 30 40 Ceratosauria 00000 00000 00000 00000 00000 00010 00000 000X0 00000 Neotetanurae 00000 00000 00000 00000 00000 00011 00000 000X0 00000 Eustreptospondylus 111?? 00000 00000 ?0000 00000 0000? 00?00 000X1 11111 Torvosaurus 11111 00000 00000 ?0000 00000 00000 00000 000X1 11111 Baryonyx 11111 11111 11111 11111 11111 11121 11110 0000? ?0??0 Suchomimus 11111 11111 1111? ?1111 11111 11121 11110 0000X 0??00 Irritator 11??? 11111 111?1 ?1?X1 1???? 0???? ????1 1111X 0???? Spinosaurus 1???? 1???? ???1? 111X? ?1000 0???? ????1 1111? ?????

Spinosaurids are characterized by numerous derived features, many of which are related to piscivory [including an unusually long snout with a long secondary palate; a terminal rosette of teeth in the upper and lower jaws (25); subcylindrical, spaced crowns; posteriorly displaced external nares; ventrally positioned basipterygoid articulation; and other features (8, 13, 22)]. Our analysis suggests that spinosaurids can be divided into two clades, the Baryonychinae and the Spinosaurinae (26), which diverged before the Barremian (Fig. 4B). The baryonychines Suchominus and Baryonyx are distinct (16, 27) but closely related, as evidenced by several derived features that include the small size and increased number of dentary teeth posterior to the terminal rosette and the deeply keeled anterior dorsal vertebrae (22). Many other similarities between these two taxa are ambiguous because they are not preserved in other spinosaurids. The spinosaurines Irritator and Spinosaurus are united on the basis of the straight unserrated crowns, the small first premaxillary tooth (4, 8), and the increased spacing of the teeth in the upper and lower jaws (8, 13, 22) (Fig. 4B). The posterior displacement of the external nares in Irritator (9) and the deep sail in Spinosaurus (Fig. 4B) may eventually characterize the Spinosaurinae, but these features are currently known in only one member.

Before the discovery of Suchomimus, the geographic distribution and relationships of spinosaurids matched the general pattern of continental fragmentation during the latter half of the Mesozoic and thus could be explained by large-scale vicariance. The split between the northern Baryonyx and the southern spinosaurines, in this hypothesis, could be attributed to the opening of the Tethyan seaway between Laurasia and Gondwanaland, and the divergence among spinosaurines could be the result of the subsequent opening of the Atlantic Ocean between South America and Africa. The discovery of Suchomimus on Africa in the mid-Cretaceous, however, complicates this scenario. Its closest relative is the European Baryonyx rather than the African Spinosaurus--a pattern of relationships that is inconsistent with the large-scale sequence of continental rifting described above.

One biogeographic hypothesis accounts most parsimoniously for the distribution of the four spinosaurids (28), assuming that we have correctly ascertained their phylogenetic relationships and accept the rifting sequence between the continental areas outlined above. Initially, spinosaurids may have had a distribution across Pangaea that was split by the opening of the Tethys; baryonychines evolved to the north (Europe, or Laurasia), and spinosaurines evolved on the southern landmass (South America and Africa, or Gondwanaland). A single dispersal event from Europe to Africa during the Early Cretaceous would account for the presence of Suchomimus in Africa. Alternative scenarios involve additional dispersal or extinction events to account for recorded distributions. The phylogenetic and biogeographic relationships of Suchomimus and other spinosaurids provide further evidence of dispersal across the Tethyan seaway during the Early Cretaceous (21, 29).

REFERENCES AND NOTES

1. E. Stromer, Abh. Kgl. Bayer. Akad. Wiss. Math. Phys. Kl. 28, 1 (1915) .
2. S. Bouaziz, et al., Bull. Soc. Geol. Fr. 1988, 335 (1988) ; E. Buffetaut, Neues Jahrb. Geol. Palaeontol. Monh. 1989, 79 (1989) .
3. D. A. Russell, Bull. Mus. Natl. Hist. Nat. Paris Ser. 4 18, 349 (1996) .
4. P. Taquet and D. A. Russell, C. R. Acad. Sci. 327, 347 (1998) .
5. The holotypic skeleton of Spinosaurus aegyptiacus includes subconical teeth, dentaries with a squared distal end, and high-spined dorsal vertebrae (1); the association of these features is confirmed by additional remains from Morocco (3) and Algeria (4). Although the Moroccan and Algerian materials have been referred to a different species (S. maroccanus), its distinction from S. aegyptiacus (by the proportions of the centrum of an isolated cervical vertebra) and the basis for the referral of additional material are questionable. We regard S. maroccanus as a nomen dubium and provisionally refer all spinosaur material from Albian- and Cenomanian-age rocks in northern Africa to S. aegyptiacus. Stromer (30) described other postcranial remains from the Baharîya oasis as "Spinosaurus B," but these can be shown to overlap with a partial skeleton of the allosauroid Carcharodontosaurus saharicus from the same locality (21). Recently a new genus and species, Sigilmassasaurus brevicollis, was erected on the basis of isolated vertebrae from Cenomanian-age rocks in Morocco (3). We question its distinction from C. saharicus (by proportions of the centrum of an isolated cervical vertebra). We regard Sigilmassasaurus brevicollis as a subjective junior synonym of C. saharicus, to which we provisionally refer all carcharodontosaurid material from Albian- and Cenomanian-age rocks in northern Africa.
6. P. Taquet, Cah. Paleontol. 1976, 1 (1976) .
7. ___, C. R. Acad. Sci. 299, 217 (1984) .
8. A. W. A. Kellner and D. A. Campos, Neues Jarhb. Geol. Palaeontol. Abh. 199, 151 (1996) .
9. D. M. Martill, et al., J. Geol. Soc. London 153, 5 (1996) [Abstract/Free Full Text].
10. Shortly after Irritator challengeri was described (9), the anterior end of a spinosaurid snout was described from the same deposit as Angaturama limai (8), which may well pertain to the same taxon or possibly to the same specimen (13).
11. A. J. Charig and A. C. Milner, Nature 324, 359 (1986) [CrossRef] [Medline] .
12. ___, in Dinosaur Systematics: Perspectives and Approaches, K. Carpenter and P. J. Currie, Eds. (Cambridge Univ. Press, Cambridge, 1990), pp. 127-140; L. I. Viera and J. A. Torres, Munibe Cienc. Nat. 47, 57 (1995) .
13. A. J. Charig and A. C. Milner, Bull. Nat. Hist. Mus. 53, 11 (1997) .
14. The Tegama Group is composed of terrestrial rocks of middle to late Cretaceous age. Three formations (Tazolé, Elrhaz, and Echkar) have been recognized [ H. Faure, Mem. B.R.G.M. Paris 47, 1 (1966) ; J. Greigert and R. Pougnet, ibid. 48, 1 (1967)]. In the region southeast of the Aïr highlands, the Tegama Group was divided into eight Gadoufaoua (GAD) levels by the Center for Atomic Energy [ E. Molinas, Rapp. C.E.A. Marseille 1965, 1 (1965) ]. Horizons corresponding to GAD 5 have yielded all of the fossils in the present report and those described previously (6) (Fig. 1). GAD 5 appears to include the upper part of the Elrhaz Formation and the lower part of the Echkar Formation (6).
15. Theropods include the spinosaurid described here, an indeterminate tetanuran known primarily from teeth, and a small basal coelurosaur. Currently, there is no evidence to support the previous referral of small theropod remains from Gadoufaoua to Elaphrosaurus iguidensis (6) [ A. F. de Lapparent, Mem. Soc. Geol. Fr. 88A, 1 (1960) ]. Sauropods include a common, high-spined basal diplodocoid [formerly referred to as a dicraeosaurine (6)] and a rare titanosaur. Ornithopods include the dryosaurid Valdosaurus nigeriensis [ P. M. Galton and P. Taquet, Geobios 15, 147 (1982) ], the common "Iguanodon trapu" [S. Chablis, thesis, Université de Paris (1988)], and the high-spined Ouranosaurus nigeriensis (6). Nondinosaurian vertebrates that were not previously recorded include a long-snouted basal crocodyloid and an azhdarchid pterosaur.
16. Etymology: Souchos, crocodile (Greek); mimos, mimic (Greek); tenere, Ténéré Desert; ensis, from (Latin). Named for the low elongate snout and piscivorous adaptations of the jaws and for the region of the Sahara in which it was discovered. Holotype: Partial disarticulated skeleton (MNN GDF500) cataloged in the collections of the Musée National du Niger (MNN), Niamey, Republic of Niger. Referred material: Articulated premaxillae and maxillae (MNN GDF501), right quadrate (MNN GDF502), partial dentaries (MNN GDF503, GDF504, and GDF505), axis (MNN GDF506), posterior cervical vertebra (MNN GDF507), posterior dorsal vertebra (MNN GDF508), two caudal vertebrae (MNN GDF510 and GDF511), and many additional bones and teeth. Diagnosis: Spinosaurid characterized by an elongate posterolateral premaxillary process that nearly excludes the maxilla from the external naris; broadened and heightened posterior dorsal, sacral, and anterior caudal neural spines; robust humeral tuberosities; hypertrophied ulnar olecranon process that is offset from the humeral articulation; and hook-shaped radial ectepicondyle.
17. In contrast to the specimens described here, previously known spinosaurid material from Niger has been limited to fragmentary disarticulated bones that are attributable to an as yet indeterminate spinosaurid (6-8). Recently, however, a new spinosaurid, Cristatusaurus lapparenti, was named on the basis of material from Gadoufaoua (4). The holotypic specimen consists of portions of the premaxillae, maxilla, and dentary, the association of which was not established. The authors state that the material differs from Baryonyx walkeri by the "brevirostrine condition of premaxilla." However, no distinguishing features or proportions are apparent to us or to previous authors (13), who attributed the premaxillae to an indeterminate species of Baryonyx. We therefore regard C. lapparenti as a nomen dubium.
18. We offer alternative identifications for several cranial elements in the holotypic specimen of Baryonyx walkeri. We regard the bones that were identified as the left postorbital, left jugal, right atlantal neural arch, and left angular (13) as the posterior portion of the right surangular, right prearticular, central body of the left pterygoid, and right angular, respectively. These are repositioned accordingly in our cranial reconstruction (Fig. 2, C and D). The plate-shaped anteromedial process of the maxilla was formerly identified as the vomer (13). We regard the deeper proportions of the occiput as reconstructed in B. walkeri (13) as an artifact of unnatural ventral displacement of the quadrate. The cranium in Baryonyx was probably as low, long, and narrow as in Suchomimus. The cervical series in both Baryonyx and Suchomimus shows a dorsal offset of the anterior articular surfaces.
19. E. Buffetaut, Neues Jahrb. Geol. Palaeontol. Monh. 1992, 88 (1992) .
20. C. W. Gilmore, Bull. U.S. Natl. Mus. 110, 1 (1920) .
21. P. C. Sereno, et al., Science 272, 986 (1996) [Abstract] .
22. The following 45 synapomorphies (optimized with delayed transformation) correspond with the scored character states [(0) or (1)] (Table 2) that were used in the analysis of the spinosaurid relationships presented in Fig. 4A. Synapomorphies 27 through 34 uniting Baryonyx and Suchomimus cannot be observed in other spinosaurids because of incomplete preservation. Spinosauroidea: 1, anterior ramus of maxilla, length: 70% (0) or 100% or more (1) of maximum depth; 2, lacrimal anterior ramus, length: more (0) or less (1) than 65% of the ventral ramus; 3, humeral deltopectoral crest, length: less (0) or more (1) than 45% of humeral length; 4, radial (forearm) length: more (0) or less (1) than 50% of humeral length; 5, manual digit I-ungual, length: 2.5 (0) or 3 (1) times the depth of the proximal end. Spinosauridae: 6, anterior end of upper and lower jaws, form: convergent (0); expanded into a premaxillary/dentary rosette (1); 7, snout length: less (0) or more (1) than three times the length of the antorbital fenestra; 8, external nares positioned entirely posterior to the premaxillary tooth row; 9, antorbital fossa, size: larger (0) or smaller (1) than the orbit; 10, interpremaxillary suture, form: open (0); fused (1) at maturity; 11, premaxillary-maxillary articulation, form: scarf or butt joint (0); interlocking (1); 12, maxillary anteromedial process, shape: fluted prong (0); plate (1); 13, maxillary anteromedial process, anterior extension: as far as (0) or far anterior to (1) the anterior margin of the maxilla; 14, paradental laminae: present (0); absent (1); 15, lacrimal anterior and ventral rami, angle of divergence: 75° to 90° (0); 30° to 45° (1); 16, splenial foramen, size: small (0); large (1); 17, midcrown cross section: elliptical (0); circular (1); 18, crown striations: absent (0); present (1); 19, premaxillary tooth count: 3 to 4 (0); 6 to 7 (1); 20, maxillary crowns, spacing: adjacent (0); with intervening space (1); 21, distal root shape: broad (0); strongly tapered (1). Baryonychinae (Suchomimus and Baryonyx): 22, anterior dorsal centra, depth of ventral keel: weak (0); blade-shaped (1); 23, maximum height of dorsal neural spines: less (0) or more (1) than 2.5 times the centrum height; 24, posterior dorsal neural spines, basal webbing: absent (0); present (1); 25, posterior dorsal neural spines, accessory centrodiapophyseal lamina: absent (0); present (1); 26, dentary tooth count: ~15 (0); ~30 (1); 27, quadrate head, shape: oval (0); subquadrate (1); 28, quadrate foramen, size: foramen (0); broad fenestra (1); 29, coracoid posterior process, shape: low and rounded (0); crescentic (1); 30, humeral trochanters, size: low and rounded (0); hypertrophied (1); 31, humeral deltopectoral crest, orientation of apex: anterior (0); lateral (1); 32, humeral internal tuberosity, size: low and rounded (0); hypertrophied (1); 33, radial external tuberosity and ulnar internal tuberosity, size: low and rounded (0); hypertrophied (1); 34, pubic foot, size: moderate to large (0); reduced to a small flange (1). Spinosaurinae (Irritator and Spinosaurus): 35, crown recurvature: present (0); very reduced or absent (1); 36, crown serrations: present (0); absent (1). 37, dentary crowns, spacing: adjacent (0); with intervening space (1); 38, premaxillary tooth 1, size: slightly smaller (0) or much smaller (1) than crowns 2 and 3; 39, diastemata within the premaxillary rosette: narrow (0); broad (1). Torvosauridae: 40, antorbital fossa, width of ventral margin: more (0) or less (1) than 30% of the maximum depth of the posterior (principal) ramus; 41, subcircular depression in the anterior corner of the antorbital fossa: absent (0); present (1); 42, lacrimal foramen, position: near the base (0) or at midheight (1) on the ventral process; 43, jugal posterior ramus, depth: less (0) or more (1) than that of the orbital ramus; 44, postorbital ventral process, cross section of distal half: subcircular (0); U-shaped (1); 45, puboischial fenestra: broadly open (0); closed or nearly closed (1).
23. F. von Huene, Rev. Mus. La Plata 29, 35 (1926) ; S.-Y. Hu, Vertebr. Palasiat. 8, 42 (1964) ; P. M. Galton and J. A. Jensen, BYU Geol. Stud. 26, 1 (1979) ; J. A. Jensen, Great Basin Nat. 45, 710 (1985) ; B. B. Britt, BYU Geol. Stud. 37, 1 (1991) .
24. A previous cladistic analysis placed spinosaurids as the sister taxon to Neotetanurae (13). Notably, synapomorphies linking spinosaurids and torvosaurids were simply ignored in that analysis, and spinosaurids and neotetanurans were joined by one character with an ambiguous optimization (a hook-shaped coracoid).
25. The expanded terminal rosette has a very specific structure in spinosaurids that is probably related to the manner in which its teeth articulate. Seven premaxillary teeth are opposed by five dentary teeth. Three diastemata are present in the upper rosette (between teeth 3 and 4, between teeth 5 and 6, and between tooth 7 and the maxillary teeth).
26. Baryonychinae (11) is defined here as all spinosaurids that are more closely related to Baryonyx than to Spinosaurus; this clade currently includes Baryonyx and Suchomimus. Spinosaurinae (1) is defined here as all spinosaurids that are more closely related to Spinosaurus than to Baryonyx; this clade currently includes Spinosaurus and Irritator (= Angaturama).
27. Revised diagnosis for Baryonyx walkeri: Spinosaurid characterized by fused nasals with a median crest terminating posteriorly in a cruciate process, a solid subrectangular lacrimal horn, a marked transverse constriction of the sacral or anterior caudal centra, a well-formed peg-and-notch articulation between the scapula and coracoid, an everted distal margin of the pubic blade, and a very shallow fibular fossa.
28. Biogeographic hypotheses were optimized with dispersal-vicariance analysis [ F. Ronquist, Syst. Biol. 46, 195 (1997) [CrossRef] [ISI] ], which counts the minimum number of dispersal or extinction events that is required to account for the observed distributions. There is no cost associated with vicariance. In our example, there is only one hypothesis that requires a single event (dispersal from Europe to Africa during the Early Cretaceous), if one accepts the general pattern of continental breakup as described in the text.
29. E. Buffetaut, Terra Nova 1, 69 (1989) ; J. Le Loeuff, Cretaceous Res. 12, 93 (1991) [CrossRef].
30. E. Stromer, Abh. Bayer. Akad. Wiss. Math. Naturwiss. Abt. N. F. 22, 1 (1934) .
31. A. G. Smith, D. G. Smith, B. M. Funnell, Atlas of Mesozoic and Cenozoic Coastlines (Cambridge Univ. Press, Cambridge, 1994), p. 41.
32. D. L. Swofford, PAUP 3.1 (Illinois Natural History Survey, Champaign, IL, 1993).
33. W. B. Harland et al., A Geologic Time Scale (Cambridge Univ. Press, Cambridge, 1989).
34. Supported by the David and Lucile Packard Foundation, National Geographic Society, Pritzker Foundation, and the Women's Board of the University of Chicago. We thank K. Bainbridge, A. Boldizar, J. Bradshaw, J.-P. Cavigelli, J. Ogradnick, and F. Stroik for participation in field excavation; C. Abraczinskas for drawing from the original specimens and executing the final drafts of Figs. 1A and 2 through 4; B. Strack (Field Museum) for assistance with microphotography; Q. Cao and E. Dong for directing fossil preparation and casting; and J. Hopson, F. Lando, R. Molnar, and H.-D. Sues for reviewing an earlier draft of the paper. We gratefully acknowledge the assistance of I. Kouada of the Ministère de L'Enseignement Supérieur de la Recherche et de la Technologie (Niger). For permission to conduct fieldwork, we are indebted to the Republic of Niger.