t Current address: Institut fur Humangenetik, Universitat Wiirzburg, Biozentrum am Hubland, Wiirzburg, Germany.

Transcription

1 Journal of Vertebrate Paleontology 15(3): , September by the Society ofvertebrate Paleontology THE PECTORAL GIRDLE AND FORELIMB OF CARSOSAURUS MARCHESETTI (AIGIALOSAURIDAE), WITH A PRELIMINARY PHYLOGENETIC ANALYSIS OF MOSASAUROIDS AND VARANOIDS MICHAEL W. CALDWELL*, ROBERT L. CARROLL, AND HINRICH KAISERt Redpath Museum and Department of Biology, McGill University, 859 Sherbrooke Street West, Montreal, Quebec, Canada, H3A 2K6 ABSTRACT- The aigialosaur Carsosaurus marchesetti is represented by a nearly complete skeleton from the Upper Cretaceous (Cenomanian-Turonian) ofslovenia. Dermal girdle elements include portions of the clavicles and a small interclavicle with a short anterior process. Endochondral girdle elements include a small scapula and large unfenestrated coracoid. A mineralized sternum is also present. The carpus is anguid-like and consists of ten ossified elements. Reduction of the procoelous nature ofcentrum articular surfaces is restricted to the caudal series. The type of C. marchesetti is the largest aigialosaur specimen known (> 1.5 m) and has proportionally larger propodials than any other aigialosaur. Phylogenetic reconstruction places Carsosaurus within a polytomous clade composed of all known aigialosaurs. Aigialosaurs are the sister-group to Mosasauridae, forming the Mosasauroidea, and are within Anguimorpha but distinct from Varanoidea. Supposed synapomorphies of varanids and mosasauroids are identified as eitherplesiomorphiesofanguimorpha, oras misidentified homologs. Results of a principal component analysis of limb measurements suggest that aigialosaurs occupy a position on a functional continuum reflective ofterrestrial not aquatic animals. Morphometric criteria cannot validate the exclusion of Carsosaurus from the Aigialosauridae. INTRODUCTION The aigialosaur Carsosaurus marchesetti was described by Komhuber (1893) on the basis of a single specimen collected from a cave in the Trieste Limestone (Cenomanian-Turonian, U pper Cretaceous) near Komen, Slovenia (Calligaris, 1988). The more complete specimens assigned to the Aigialosauridae (Table 1) also come from the Trieste Limestone, though from different localities than C. marchesetti (Calligaris, 1988; Carroll and debraga, 1992). The only specimen of C. marchesetti is exposed in ventral view and is missing the skull, anterior cervicals, and a large part ofthe tail (Fig. 1). The remaining postcranial skeleton is nearly complete. The Aigialosauridae, exclusive of C. marchesetti, contains two monotypic genera (Table 1), each represented by a single specimen, and a single unnamed specimen referred to here as the Trieste Aigialosaur ("Trieste specimen" of Carroll and debraga, 1992). A series ofvertebrae, referred to Aigialosaurus novaki by Kramberger (1892), show no features that allow differentiation from other taxa. A skull fragment from the Upper Jurassic, referred to Proaigialosaurus hueni by Kuhn (1958), is now lost (Table 1). * Currentaddress: Department ofbiological Sciences, University of Alberta, Edmonton, Alberta T69 2E9, Canada. t Current address: Institut fur Humangenetik, Universitat Wiirzburg, Biozentrum am Hubland, Wiirzburg, Germany. Komhuber (1893) classified C. marchesetti as a member ofaigialosauridae on the basis ofpostcranial features. However, Carroll and debraga (1992) excluded C. marchesetti from their review of the Aigialosauridae because of limb bone ratios which they considered indicative ofsignificant differences between C. marchesetti and the other aigialosaurs. Carroll and debraga (1992) also suggested that A. dalmaticus and Opetiosaurus bucchichi might be conspecific, thereby making the Aigialosauridae monotypic. However, their assertion that C. marchesetti differs from other aigialosaurs only means that this species might not belong to the genus or species ofthe conspecific forms; it does not merit its exclusion from the family. It is therefore important to determine the conspecificity of A. dalmaticus and O. bucchichi and whether or not these taxa form a monophyletic group with C. marchesetti. The postcranium of C. marchesetti is an important source ofcharacters as it includes details ofthe shoulder girdle not preserved in other aigialosaurs. Conventional approaches to varanoid phylogeny consistently hypothesize that an aigialosaur-mosasaur clade is nested within the Varanoidea (McDowell and Bogert, 1954; Russell, 1967; Rieppel, 1980a; Borsuk-Bialynicka, 1984; Carroll and debraga, 1992; debraga and Carroll, 1993). However, it is difficult to account for the large number of reversals which result when polarity is determined using fossil or extant varanoids (sensu Pregill et ai., 1986) as outgroups. Gauthier (1982) and Pregill et ai. (1986) suggested that an aigialosaur-mosasaur clade might in fact lie 516

2 CALDWELL ET AL.-CARSOSAURUS PECTORAL GIRDLE 517 TABLE 1. List of taxa assigned to the family Aigialosauridae (Squamata: Mosasauroidea), with a short characterization of each specimen. Species assignment Aigialosaurus dalmaticus A. novaki Carsosaurus marchestetti Opetiosaurus bucchichi Proaigialosaurus hueni Trieste Aigialosaur Location Miinchen Trieste Trieste Vienna Lost Trieste Status Mostly complete 38 caudal vertebrae Complete except for skull Mostly complete Skull fragment Complete, associated, described but not named Reference Kramberger, 1892 Kramberger, 1892 Kronhuber, 1893 Komhuber, 1901 Kuhn, 1958 Carroll and debraga, 1992 outside of Varanoidea. This is an important and interesting suggestion which has had variable support, for very different reasons, from late nineteenth and early twentieth century squamate systematists (Fejervary, 1918 for a review). Hypotheses of a sistergroup relationship between aigialosaurs and mosasaurs appear well supported, making aigialosaur morphology important to polarizing mosasaur characters. In a general sense, the skull is almost indistinguishable from that of primitive mosausaurs (Carroll and debraga, 1992). However, as will be discussed below, characters in the postcranial skeleton show only minor differences from those of terrestrial anguimorphs. We present a detailed description of the reprepared pectoral girdle and forelimb of C. marchesetti along with a review of the remaining postcranial skeleton. From this description we have added characters to those used by Pregill et al. (1986) and Rieppel (1980a) in order to construct a character matrix for phylogenetic analysis of aigialosaurs, mosasaurs, and varanoids. We also present a preliminary phylogenetic hypothesis of mosasauroid-varanoid relationships using these characters. This analysis was motivated by the uncertain phylogenetic relationship of C. marchesetti with other aigialosaurs (Carroll and debraga, 1992). To investigate further the morphometrically-based assertion of Carroll and debraga (1992) that C. marchesetti should not be considered an aigialosaur, we have added measurements of limb and vertebral elements for Varanus salvator, Lanthanotus borneensis, Heloderma sp., and Clidastes propython (Table 2) to their data set, and conducted a principle component analysis (PCA). In association with qualitative morphological descriptions, our PCA results are interpreted in terms of function not phylogeny. MATERIALS AND METHODS The shoulder girdle and forelimb of Carsosaurus marchesetti were reprepared and drawn by RLC during the summer of Due to difficulties encountered in differentiating bone from matrix it was not possible to prepare the remainder ofthe skeleton. Reconstruction of C. marchesetti is based on drawings and photographs of the entire specimen. Institutional abbreviations are FMNH (Field Museum of Natural History, Chicago, Illinois, U.S.A.), RM (Redpath Museum, McGill University, Montreal, Quebec, Canada), KU (Museum of Natural History, University ofkansas, Lawrence, U.S.A.), BSP (Bayerische Staatssammlung fur Palaontologie und historische Geologie, Miinchen, Germany), MCSNT (Museo Civico di Storia Naturale, Trieste, Italy), and NMW (Naturhistorisches Museum, Wien, Austria). Phylogenetic Analysis Phylogenetic analysis ofnine mosasauroid-varanoid taxa was conducted using PAUP Version for the Macintosh (Swofford, 1993). A data matrix of sixtysix characters (sixty-one binary and five multistate) was constructed (Appendix I), using characters and character states from Pregill et al. (1986) and Rieppel (1980a), and from characters identified in this study (Appendix II). Some characters were modified to reflect ingroup character states. All multistate character transformations were unordered to avoid a priori assumptions of transformation vectors. Outgroup polarity was determined using characters ofgerrhonotus. Information on gerrhonotines was derived from an unnumbered skull ofgerrhonotus (=EIgaria) multicarinatus in the Redpath Museum collection, and from Meszoely (1970), Rieppel (1980a), Gauthier (1982), and Good (1987). Examination of hypotheses of anguimorph phylogeny indicate that gerrhonotines are among the least derived anguoids (Rieppel, 1980a) or anguids (Good, 1988). Gauthier (1982) and Estes et al. (1988) reported a basal anguimorphan trichotomy between varanoids, xenosaurids, and anguids. Estes et al. (1988:fig. 5C) also present a cladogram with snakes, dibamids, and amphisbaenids deleted, in which Anguidae is the closest sistergroup ofliving varanoids. A consensus ofthese phylogenetic hypotheses would seem to indicate that an anguid taxon is a reasonable choice for outgroup polarization of varanoid characters. Morphometric Analysis Measurements of upper and lower limb characters and average lengths of trunk vertebrae (Table 2) were taken from Varanus salvator (RM a), Lanthanotus borneensis (FMNH ), Heloderma sp.

3 518 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 15, NO.3, 1995 TABLE 2. Measurements (in mm) for selected characters of mosasauroid and varanoid taxa used in a principal component analysis. Abbreviations are ADV (average length ofdorsal vertebrae), FEM D1ST (distal head width offemur), FEM prox (proximal head width offemur), HUM (humerus), HUM DIsT (distal head width ofhumerus), HUM prox (proximal head width ofhumerus, RAD (radius), TID (tibia), UL (ulna). Aigialo- Opetio- Varanus saurus saurus Trieste Carsosaurus exanthe- Clidastes Lanthanotus Heloderma dalmatieus bueehiehi Aigialosaur marehesetti matieus propython borneensis sp. HUM HUM prox HUM o1st UL RAD ADV FEM FEM prox FEM 01ST FIB TIB (RM unnumbered specimen); Clidastes prophython (KU 1022), Aigialosaurus dalmaticus (BSP ), Opetiosaurus bucchichi (NMW unnumbered specimen), the Trieste Aigialosaur (MCSNT 11430, 11431, and 11432), and Carsosaurus marchesetti (MCSNT unnumbered specimen). Several data points were missing for the limb elements of the Trieste Aigialosaur, necessitating a two-tiered analysis. For the initial PCA (Analysis I), we excluded the Trieste Aigialosaur to make use of ten characters, excluding only the length of the fibula, missing from Opetiosaurus. A second pea (Analysis II) included the Trieste Aigialosaur but reduced the number of variables to seven; the data of Aigialosaurus and Opetiosaurus were combined to retain the maximum number of variables. All measurements were standardized within each variable by conversion to z-scores before PCA. Statistical tests ofthe significance of the principle components could not be conducted due to the limited sample size. DESCRIPTION Pectoral Girdle and Left Limb (Figs. 1, 2A-C) The pectoral girdle, sternal cartilage, and left limb are preserved in articulation. The mineralized sternal cartilage is large and shield-shaped. The posterolateral margins of the sternal plate, beginning posteriorly at one-third ofits overall length, bear articular facets for five costosternal cartilages. There are spaces between the facets ofrib cartilages one to three, but not between the facets for ribs three to five. There are also facets for xiphisternal cartilages lateral to the midline at the posterior apex of the sternal shield. Small, paired cartilages are in articulation with these apical facets. The interclavicle is crucifix-shaped and small relative to the sternum. The small anterior process does not appear to be well ossified. The lateral processes forming the crossbar show cracks near the midpoint of their lateral extent. This may reflect the limit of overlap with the medial limb ofthe clavicles. The posterior extension of the interclavicle reaches a point level with the position ofthe facet for the first costosternal cartilage. The surface ofthe more posterior section is rougher than the anterior portion. The element identified as a clavicular fragment is small and pointed at its dorsolateral end. Medially the element is squared. The left coracoid is complete and not emarginate along its anteromedial border. It is approximately twothirds the length of the interclavicle and has two coracoid foramina in its anterolateral quadrant. The lateral margin of the coracoid contributes equally to the scapulacoracoid fenestra and the glenoid fossa. The suture with the scapula is clearly expressed. The scapula is small, approximately half the size of the coracoid, and is hatchet-shaped. It is broken along the anteromedial margin, where it forms the anterolateral border of the scapu1acoracoid fenestra. The scapular contribution to the glenoid fossa is directed laterally with only a minor ventral inflection. The medial and anteromedial areas ofthe scapulacoracoid are thickened in an arc around the glenoid fossa. The superior lip ofthe glenoid, located primarily on the scapula, overhangs the inferior lip. The humeral head is large and rugose indicating the presence ofa cartilaginous cap; there is the suggestion ofan epiphyseal suture between the head and shaft but this is not certain. The pectoral and deltoid crests on the medial tubercle of the humerus are very well developed, as are the radial condyle and trochlea. The latter are separated by an obvious radial groove. The trochlea is smaller than the radial condyle and is centered in the midline of the shaft. The radius and ulna both show proximal and distal epiphyses. The olecranon process on the ulna is present but its proximal extension is halfthat observed in varanoids. The radius and ulna, without the olecranon, are of similar length but are only three-quarters the length of the humerus (Table 1).

4 CALDWELL ETAL.-CARSOSAURUS PECTORAL GIRDLE 519 The carpus contains ten ossified elements. These are identified according to the conventions ofrenous-ucuru (1973) as the radiale, lateral centrale, intermedium, ulnare, pisiform, medial centrale, and distal carpals two through five. The radiale is an irregularly shaped element, compressed proximodistally, and expanded into small lobes on its medial and lateral margins. The medial expansion ofthe element contributes to the distal margin of the antebrachial space. The lateral centrale is an irregular hexagon and articulates proximally with the intermedium, medially with the radiale and medial centrale, laterally with the ulnare, and distally with distal carpals two to four. The ulnare is the largest element in the carpus. It articulates with both the lateral centrale and intermedium, as well as with distal carpals four and five. It likely articulated with the pisiform but this element appears to have rotated proximally out ofarticulation, making its exact morphology and position difficult to determine. Distal carpal four is the largest element in the distal carpal row. In order of largest to smallest in size, the remaining distal carpals are respectively three, two, and five. The distal portion of metacarpals one and two are incomplete, but metacarpals three through five are complete. Fragments ofdigits two and four are found in articulation with the respective metacarpals. The complete complement ofthree phalanges is present for digit five. The metacarpals all show epiphyseal plates at their proximal ends as do the phalangeal fragments for digits two and four, and all the phalanges of digit five. Overview of Remaining Axial Skeleton (Figs. 1, 3) The three most anterior vertebrae preserved are identified as the fifth, sixth, and seventh cervicals. Carroll and debraga (1992) identify seven cervicals in the Trieste Aigialosaur. The criteria used here to identify cervicals are the presence ofhypapophyseal peduncles on the fifth and sixth, and the absence ofa sternal rib on the seventh; small, short, pointed ribs are associated with these vertebrae (Hoffstetter and Gasc, 1967, 1969). The cervical centra are subequal or slightly shorter, than the trunk centra. There are twenty-one trunk vertebrae. The first to the fifth are preserved with their ventral, calcified, costosternal cartilages in place and in articulation with the calcified sternal cartilage. The ribs shorten quickly beginning at the fourteenth trunk rib. The ribs for the last trunk vertebra are missing. Kornhuber (1893) described two sacral ribs; however, the pelvic region is poorly preserved and this cannotbe confirmed. Twelve caudals are present. All vertebrae in this series have transverse processes and either haemal arches or their respective haemal facets. Pelvic Girdle and Limb (Figs. 1, 3) The iliac blade is directed posteriorly. Its posterior limit is approximately three vertebrae posterior to the FIGURE 1. Outline drawing ofholotype specimen ofcarsosaurus marchesetti; modified from Kornhuber (1893). The specimen is preserved in ventral view; note the relationship ofthe interclavicle to the calcified sternum and coracoid, and the varanoid-like nature ofcervicals and presacrals as compared to the caudals. Housed in the Museo Civico di Storia Naturale Trieste.

5 520 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 15, NO.3, 1995 Ie...ssc H -1cm c A B FIGURE 2. Front limb and pectoral girdle of Carsosaurus marchesetti. A, Forelimb and portions of pectoral girdle based on holotype; B, Reconstruction offorelimb and scapulacoracoid complex (with cartilages); C, Reconstruction ofpectoral girdle (ventral view). Abbreviations: C, clavicle; el, lateral centrale; em, medial centrale; Cor, coracoid; IC, interclavicle; iot, intermedium; pis, pisiform; rl-5, sternal cartilages ofpresacral ribs 1-5; R, radius; rad, radiale; scor, supracoracoid cartilage; ssc, suprascapular cartilage; sc, scapula; St, sternum; U, ulna; ul, ulnare; 2-5, distal carpals; i-v, metacarpals. acetabulum. The acetabulum appears well developed, with good delimitation ofthe margins. The ischia and pubi are poorly preserved and cannot be adequately described. The femur is longer than the humerus, while the tibia and fibula are shorter than the radius and ulna (Table 2). A small patellar element is present as well. The tarsus is poorly preserved and only fragments remain from the metatarsus and phalanges. It is not possible to refute or substantiate the descriptions ofthese elements as given by Komhuber (1893). Discussion and Comparison The condition of possessing short cervicals relative to the trunk vertebrae is observed in aigialosaurs.. mosasaurs, (Fig. 3A-D, F), and most anguimorphs (Hoffstetter and Gasc, 1967, 1969; Hecht and Costelli, 1969). Extant varanids are the exception, where the cervical centra are one-and-a-half times to almost twice the length of trunk centra (Fig. 3E). A lateral view ofthe axial skeleton of C. marchesetti is not possible. However, ventral examination of the condyle-cotyle articulations shows that the transitions in vertebral centrum morphology are similar to those of other aigialosaurs (Fig. 3A-D). The vertebral articulations show a progressive decrease in the concavity of the cotyle beginning in the caudal series (Fig. 1). The curvature of the condyle and the constriction of the centrum are reduced so that in ventral view the centra appear spool-shaped or amphicoelic, not procoelic. Mosasaurs show similar changes in centrum morphology, but the effect begins in the anterior dorsals. The more vertical orientation ofthe articular facets has been considered an important adaptation to anguilliform locomotion in these aquatically evolving animals (Russell, 1967; Carroll and debraga, 1992). It

6 CALDWELL ET AL.-CARSOSAURUS PECTORAL GIRDLE 521 is important to note that while the centra are approaching the mosasaur condition, the orientation of the zygapophyses of the cervicals and anterior trunk vertebrae in other aigialosaurs are varanoid-like (Fig. 3A-C), though the presence of this feature cannot be confirmed in C. marchesetti. Carsosaurus marchesetti retains a high number of elongate trunk ribs as in the Trieste Aigialosaur and O. bucchichi (Fig. 3B-D). This condition is plesiomorphic for squamates (Hoffstetter and Gasc, 1969) and suggests that these three species are more primitive than A. dalmaticus if Carroll and debraga's (1992) characterization of rib length is correct. Two other important features of C. marchesetti are the number ofsternal cartilages and the position in the trunk series of the corresponding vertebrae; there are five sternal cartilages articulating with trunk ribs one to five. Assuming seven cervicals, as in other aigialosaurs (Carroll and debraga, 1992), vertebrae eight through twelve articulate with the sternal ribs (the difficulty of defining the cervical-trunk transition makes it useful to compare positional relationships in both relative and absolute terms). In Varanus (Fig. 3F) there are only three ribs articulating with the sternum. These articulate with the third to the fifth trunk vertebrae, or vertebrae ten through twelve; trunk vertebrae one and two articulate with short, gravile ribs that do not reach the sternum. Adult Lanthanotus have nine cervicals and only two sternal ribs articulating with the first and second trunk vertebrae, or in absolute terms, vertebrae ten and eleven (Rieppel, 1980b). A cleared and stained juvenile Lanthanotus shows a third cartilaginous sternal rib, originating on the sternum anterior to the rib which articulates with vertebra ten (Rieppel, 1992). However, the ossified rib articulating with vertebrae nine, the putative terminal cervical, does not contact this cartilaginous structure. The condition seen in C. marchesetti may be primitive for anguimorphs, indicating that extant varanids have lost two anterior rib articulations with the sternum. It is reasonable to assume that the five ribs of the Trieste Aigialosaur, as restored by Carroll and debraga (1992), articulated with trunk vertebrae one to five as in C. marchesetti. However, such broad statements are complicated by other aspects of C. marchesetti (Fig. 2C). Its sternum bears paired xiphisternal cartilages found at the posterior apex of the sternal plate as well as the five sternal rib cartilages. In Varanus the paired xiphisternals bear ribs which articulate with trunk vertebra five. It is unfortunate that so little is known about the sternum in mosasaurs as there appear to be a high number of sternal ribs, with a maximum of ten known in Plotosaurus (Camp, 1942; Russell, 1967). Limb Girdles and Locomotion The limbs and girdles of C. marchesetti are fully ossified with well-defined articulating surfaces. The pectoral girdle is varanoid-like (Fig. 2C), most closely resembling that of Heloderma in terms of scapulocoracoid features and proportions (Lecuru, 1968a), and most closely with Varanus regarding sternal proportions (Lecuru, 1968b). The articulations ofthe clavicle, interclavicle, coracoid, and sternal elements indicate that the girdle functioned to increase the range ofmotion of the forelimb in a manner similar to that of extant varanids (Jenkins and Goslow, 1983). In mosasaurs, the sternum and interclavicle are present but rarely preserved, while clavicles are known only from Mosasaurus (Russell, 1967). Without further material it is very difficult to compare pectoral girdle structure and function between mosasaurs and other varanoids. The coracoids of C. marchesetti, o. bucchichi, Heloderma (Lecuru, 1968a), and tylosaurine mosasaurs (Russell, 1967) are not fenestrated. Varanids have anterior and posterior fenestra, while the remaining mosasaurs, anguids, and Lanthanotus have only the anterior one. Gilmore (1922) described only an anterior fenestra for the extinct varanid Saniwa ensidens. The polarity ofcoracoid fenestration is difficult to ascertian when mosasauroids are considered as members ofvaranoidea. Though mineralized cartilages are poorly known in mosasaurs, it appears that the small scapula and large suprascapular cartilage ofvaranoids, aigialosaurs, and mosasaurs (Fig. 3A-F) may be a synapomorphy of anguimorphs. The suprascapular cartilage of the xenosaur Shinisaurus is almost twice the length of the scapula proper (Costelli and Hecht, 1971). Examination of iguanids, lacertids, and geckos (RM unnumbered specimens) shows that the suprascapular cartilage is equal in length to, or much shorter than, the ossified portion of the scapula. The phylogenetic importance ofthis character is uncertain, but functionally it affects both the origin, insertion, and relative mass of shoulder and upper limb musculature. Functional aspects of the humeral-scapulocoracoid articulation in aigialosaurs appear similar to those of limbed terrestrial anguimorphs. Articular surfaces are well developed, unlike the mosasaur condition in which the articular surfaces are often rough and poorly finished. The superior lip of the glenoid fossa, located primarily on the scapula, overhangs the inferior lip as in Varanus and Heloderma. The fossa is constructed to resist vertical dislocation of the humerus by joint loading during the impact phase of terrestrial locomotion. The glenoid in C. marchesetti is directed laterally with only a slight ventral component. A minor posterior component exists, but, as in Varanus, this appears to result from the absence ofa posterior bony cup on the glenoid. Resistance to posterior humeral dislocation at the point of maximal excursion may have been provided by strong cruciate ligaments at the shoulder joint as in Varanus (Haines, 1952; Jenkins and Goslow, 1983). This idea is supported by the size ofthe deltoid and pectoral tuberosities on the humerus of C. marchesetti (sites of insertion for the dorsal cranial and caudal, and ventral cranial and caudal, cruciate ligaments). Primitive mosasaurs possess large tu-

7 522 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 15, NO.3, 1995 A 1t;{:{N4'2' B c E

8 CALDWELL ETAL.-CARSOSAURUS PECTORAL GIRDLE 523 bercles and crests on the humeri, while more derived forms have reduced tuberosities (Russell, 1967). The posterior and ventral direction of the glenoid in mosasaurs suggests a very different function associated with aquatic locomotion, due largely to the absence of dislocative joint loading encountered during terrestrial locomotion. This is not the case in aigialosaurs. The potential terrestrial nature ofaigialosaurs is also supported by interpretations ofosteology and function in the lower portion of the forelimb. The large olecranon process of the ulna would have stabilized the elbow during both pronation of the lower limb, and the downward and forward thrust phases of forward propulsion. The olecranon is not as well developed as that seen in extant varanoids. Due to major changes in the limbs the olecranon is often difficult to identify in mosasaurs. The carpi ofall aigialosaurs appear similar and show no obvious structural differences from the carpi ofextant anguimorphs (Renous-Ucuru, 1973). However, differences between the carpi of aigialosaurs and mosasaurs are profound (Fig. 3A-D, F). In mosasaurs the carpus is altered by reduction of the number of elements and the degree to which they are ossified. Reduction may have occurred by terminal deletions during development, or by chondrogenic-osteogenic fusions. The phalangeal count of the manus in aigialosaurs is plesiomorphic for squamates, though in C. marchesetti only the three phalanges of digit five are preserved. Phalangeal count is important as it indicates that neither reduction nor multiplication ofphalangeal elements has occurred in C. marchesetti. Unlike mosasaurs, the ungual ofdigit five indicates that C. marchesetti had claws as did Aigialosaurus, Opetiosaurus, and other terrestrial anguimorphs (Fig. 3A-C). The presence of two sacral vertebrae in C. marchesetti is plesiomorphic for squamates. However, the posterior limits of the iliac blade are slightly further posterior along the vertebral column, extending past the second sacral and onto the first caudal. In other aigialosaurs the iliac blade extends only to the posterior margin of the second sacral (Fig. 3A-D). In Varanus, Heloderma, and Lanthanotus, the iliac blade covers almost three vertebral centra beyond the position of the acetabulum. The elongate posterior process is likely related functionally to terrestrial locomotion in large anguimorphs. A longer process would allow more muscle mass at the point of origin, as well as increase the effective angle of contraction for the large flexor muscles. Results PHYLOGENETIC ANALYSIS Branch-and-Bound search options found nine equally parsimonious cladograms during analysis ofthe data matrix (Appendix I). Tree length is 107 steps, with a consistency index (CI) of Both Strict Consensus (Fig. 4A) and 50 % Majority-Rule Consensus (Fig. 4B) trees show 100 % support for all major dichotomies within varanoids (Varanus + Heloderma + Estesia + Lanthanotus), and between varanoids and mosasauroids (Mosasauridae + Aigialosauridae). While the Strict Consensus (Fig. 4A) shows no resolution within the mosasauroids, the 50 % Majority-Rule Consensus (Fig. 4B) supports a resolved Mosasauridae-Aigialosauridae dichotomy in 89 %, or eight out of nine, of the most parsimonious trees. Neither consensus provides any resolution within the Aigialosauridae. Only a single character (Character 58 [2]: elongate ribs present only as far posteriorly as trunk vertebrae seven to nine) supports the tree topology responsible for the mosasauroid polytomy in the Strict Consensus tree (Fig. 4A). Because of such tenuous support for a single tree topology out of nine, the data set was also tested with Character 58 excluded. This produced eight most parsimonious trees (103 steps; CI = 0.66). The topology of the Strict Consensus tree for this reduced data set was identical to the 50 % Majority-Rule Consensus obtained for the full data set (Fig. 4B), thereby strengthening hypotheses of monophyly for the Mosasauridae and Aigialosauridae as independent mosasauroid clades. Overall mosasauroid (Aigialosauridae + Mosasauridae) monophyly is supported in both the first and second (reduced data set) analyses. Varanoidea (sensu Pregill et ai., 1986), including Estesia sinensis, is a robust clade with consistent support at all levels of our analysis and an identical topology to that hypothesized by Norell et al. (1992). The relationships of mosasauroids with other anguimorphs (xenosaurids and anguids) outside of Varanoidea, are uncertain at this point. As Gauthier (1982) stated in discussing the uncertainty of xenosaurid-anguid-varanoid relationships, this analysis shows only that mosasauroids are non-varanoid anguimorphs. The use of Gerrhonotus to polarize character transformations in this analysis serves as a tool in testing conventional hypotheses of relationship within varanoids. The trees we have presented (Fig. 4A, B) were intentionally rooted at an internal basal polytomy to communicate our uncertainty ofbasal anguimorph relationships. We acknowledge the paraphyly ofsuch an FIGURE 3. Reconstructions and skeletal drawings of fossil and extant varanoids. A, holotype ofaigialosaurus dalmaticus; B, holotype of Opetiosaurus bucchichi; C, Trieste Aigialosaur; D, holotype of Carsosaurus marchesetti; E, skeleton of Varanus salvator; F, reconstruction of Clidastes sternbergi.

9 524 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 15, NO.3, 1995 A Strict B Majority rule 0 eu CI) CI) CI) CI) cd 2 2 '6'0 :::3 cd :::3 :z ::: :.( CI) Q Q cd 0 0 eu cd :::3 ti) 0 "0...c:: Q m cd ca 0 ti) -E Q e B '6"0 ' eu "i) cd cd ::I:...:l ;> :.( 0 E5 d analysis through exclusion of some fossil taxa (necrosaurids, Saniwa, Cherminotus), as well as the possible polyphyly resulting from the absence of xenosaurids and other anguids. However, this last point is derived from a posteriori recognition of clade status for mosasauroids apart from varanoids. As our goal was a preliminary analysis of the conventional taxon Varanoidea, new phylogenetic hypotheses, based on recognition of our results here, can now be tested. Discussion 0 eu CI) CI) CI) cd B e ca cd B ] '6'0 Q CI) Q a 0 eu cd 0 '1:l 0 -E 0 m "i) Q B 0 '6"0 eu eu cd Our goal with this analysis was not to review or revise phylogenetic hypotheses regarding the Anguimorpha, but rather to assess the relationships of aigialosaurs. We do not therefore, intend to discuss at I-< CI) :.( os ca ' cd d ::r:...:l ;> ::E :.( E5 U FIGURE 4. Consensus cladograms of mosasauroid-varanoid interrelationships based on nine most parsimonius trees (107 steps; CI = 0.654) derived from the data matrix in Appendix I. A, Strict Consensus tree ofrelationships. B, 50% Majority-Rule Consensus tree. It is important to note that in 89% (8 out of 9) of the most parsimonious trees that the Mosasauridae is resolved relative to the aigialosaur polytomy. Only one character in one topology, Character 58 (0), places the Mosasauridae in a clade with Aigialosaurus dalmaticus. I-< ca CI) 2 0 length the character transformation series implied by our results for all characters. These will be considered in forthcoming studies ofthe Anguimorpha. However, there are a number of characters and character states (Characters 3-6, 11-13, 16-17, concerned with narial retraction) which can be discussed confidently in terms of homology and transformation between varanoids and mosasauroids. Nares and Narial Retraction (Fig. 5A-D): The degree ofnarial retraction has been cited as a synapomorphy ofvaranids and mosasauroids without attention to the actual homologous structures involved. It cannot be stressed strongly enough that an open space (fenestra), created by the differential spatial interaction of numerous developmental units (bones, soft tissues), cannot be treated as one independent character. Each element involved in such a structure must be considered not in terms of the fenestra itself, but in terms of its topological relations with other developmental units. This erroneous notion ofhomolog characterization has misconstrued notions of synapomorphy and relationships within the conventional taxon Varanoidea (for a review Fejervary, 1918; McDowell and Bogert, 1954). In Varanus (Fig. 5A) (Bahl, 1938; Rieppel, 1980a) perceived narial retraction results from the following: elongation ofthe narial and vomerine processes ofthe premaxilla; anterior elongation and narrowing of the nasal bone and medial retraction ofthe nasal from the prefrontal and maxilla; elongation ofthe anterolateral processes ofthe frontal around the posterior processes of the nasal; shortening of the prefrontal; exposure of the septomaxilla; elongation of the vomer anterior to the vomerine aperture; and anterior elongation of the maxilla, with medial expansion and elongation anteriorly. In aigialosaurs, and in less derived mosasaurs (Fig. 5D) such as Halisaurus and some mosasaurines (Russell, 1967; debraga and Carroll, 1993), narial retraction results from quite different structural relationships: extreme elongation of the narial process of the premaxilla; elongation of the maxilla and development of an anterodorsal maxillary process; extreme elongation and narrowing of the septomaxilla (Camp, 1942); loss or fusion of the nasal bone with the anterior process of the frontal with no evidence of nasal elongation contributing to the internarial bar as in Varanus (in some mosasaurs there is evidence of nasals, but these are small and lie at or near the junction of the narial bar and frontal [Russell, 1967]); extreme elongation ofthe frontals anterior to the anterior orbital margin (except Plotosaurus bennisoni; Camp [1942]); no size decrease in the anterior portion of the prefrontals (in Ectenosaurus clear contact is maintained between the maxilla, prefrontal and frontal, preventing apparent narial emargination to a point superior to the maxillary-lacrimal suture as in Varanus). In short, the narial opening has not retracted at all. The frame of reference for elongation should not be from the snout backwards, but rather from the orbit-

10 CALDWELL ET AL.-CARSOSAURUS PECTORAL GIRDLE 525 B 0 c FIGURE 5. Dorsal view ofnarial regions in three anguimorph diapsids. A, Varanus salvator (RM a). B, Gerrhonotus multicarinatus (unnumbered RM specimen). C, dorsolateral view ofgerrhonotus multicarinatus to illustrate the size and shape ofthe septomaxilla. D, Plotosaurus bennisoni (UCMP 32778), narial bar ofpremaxilla has been cut away (redrawn from Camp, 1942). Abbreviations: f, frontal; lac, lacrimal; mx, maxilla; 0, nasals; pi, palatine; pm, premaxilla; prf, prefrontal; sm, septomaxilla; v, vomer. braincase region forward. In relative proportions, this region is equivalent in the skulls of Varanus (Fig. 5A), Heloderma, and Gerrhonotus (Fig. 5B, C) we examined (see also the illustrations provided by Rieppel [1980a]), and in aigialosaurs and mosasaurs (Fig. 5D). Changes in the topological relation ofthe nasals with the maxillae and prefrontals appear to be the major effector of narial enlargement (Heloderma) compared to the plesiomorphic condition (Gerrhonotus). The elongate narial bar is a result of snout elongation and should not be thought ofas the effector ofthat change. The elongate frontals ofaigialosaurs and mosasaurs may be a result of ontogenetic fusion with the nasals. The elongate frontals of mosasauroids extend past the orbits above the point where the orbital process of the maxilla expands dorsally; this is equivalent to the position ofthe nasal-premaxillary bar contact in Varanus (Fig. 5A). The only mosasauroid exception is Plotosaurus bennisoni (Fig. 5D) where the frontal is short, the nasals are present and small, and the prefrontals are excluded from the narial margin. In terms of sutural relationships, the insertion ofthe narial bar into the frontal in mosasauroids is identical to the insertion of the bar into the nasals in Varanus. Together, these points strongly suggest that ontogenetically, mosasauroids possessed nasals which fused to the frontals, but maintained the plesiomorphic topology with the prefrontals, and variably, the maxillae. This condition is in direct contrast to Varanus. Snout elongation is therefore accomplished convergently in mosasauroids and varanoids. Results MORPHOMETRIC ANALYSIS In both morphometric analyses, principal component (PC) 1 and PC2 explained over 98% of the total variance in the data set (Table 3). The percentage of total variance explained by PC3 and PC4 combined was negligible and these components were thus not considered further. Loadings for PCl in both cases were all positive, identifying this component as an indicator ofsize variation between taxa, while PC2 loadings included both positive and negative values, characterizing it as a shape component (Table 3). Analysis I-Loadings ofpc1 are all nearly equal in magnitude. Conversely, loadings of PC2 emphasize relative lengths of ulna, femur, and the heads of the humerus (Table 3). Graphic representaiton ofpc1 and pe2 (Fig. 6A) placed Carsosaurus in a position nearly equidistant to either Varanus or the aigialosaurs Opetiosaurus and Aigialosaurus, along both the size (PCl) and shape (PC2) components. The distance from these taxa to the mosasaur Clidastes is greater along both size and shape axes. While Clidastes and Varanus occupy positions of maximum size and shape, respectively, Opetiosaurus and Aigialosaurus cannot be differentiated confidently along either axis (Fig. 6A). All taxa except the aquatic Clidastes fit well (r 2 = 0.955) along a line indicating proportionality (slope = 1.11) of size and shape. Morphometric change for both size and shape occurs gradually from Lanthanotus at one extreme to Varanus at the other, with the three aigi-

11 526 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 15, NO.3, 1995 TABLE 3. Prinicipal component loadings (PCL) for a principal component analysis of measurements from varanoids and mosasauroids. Analyses were conducted in a stepwise fashion to circumvent data limitations. The first analysis (I) included ten variables and omitted the Trieste aigialosaur. The second analysis (II) included seven variable and combined data for Aigialosaurus and Opetiosaurus. Characters with the greatest contribution to each PCL (assigned arbitrarily) are indicated by asterisk (*). Abbreviations of measurements are as listed in Table 2. 2 N U 0 -I A lanthanotus Varanus 0 PCL 1 PCL2 PCL 1 PCL2 HUM HUM prox * HUM D1sT * UL * * RAD * ADV FEM * FEM prox 0.980* * FEM D1ST FIB * TIB /0 variation alosaurs and Heloderma occupying intermediate positions (Fig. 6A). AnalysisII-Loadingsfor PC1 (size axis) in this case are influenced nearly equally by the measurements, while the shape axis is greatly influenced by the fibular and ulnar lengths (Table 3). Graphically, Clidastes lies at the high extreme for both size and shape axes (Fig. 6B). Both Clidastes and Varanus hold positions nearly equidistant from an "aigialosaur morphospace," as defined by the presence ofaigialosaurus (=Opetiosaurus), Carsosaurus, and the Trieste Aigialosaur. The latter is in a position intermediate to Carsosaurus and Aigialosaurus (=Opetiosaurus). As in the previous analysis, all taxa, except Clidastes, lie closely along a straight line (r 2 = 0.885; slope = -1.02) with Lanthanotus and Varanus at the upper and lower extremes, respectively. Discussion The most important conclusions to be drawn from this approach reflect on functional aspects oflimb proportions relative to body size. Since the position of Carsosaurus is equivocal when compared to Lanthanotus, Varanus, and Clidastes (near the centroid of an equilateral triangle), consideration of the relative influence of variables and their component loadings becomes important. The selective, though necessary, exclusion of variables in these two analyses results in a stepwise operation ofsorts, which assists in determining some functional relationships between variables. While size is generally ofgreat influence on structure, the dual analysis conducted here shifts the importance of the simple length measurements from size in the larger data set, to shape in the more exclusive data set II Clidastes I 0 2 PC 1 2 N U 0 -I B 0 Clidastes OCarsosaurus Varanus _---L...I..- -J -2 -I 2 o PC 1 FIGURE 6. Principlecomponentanalyses (PCA) using data for Aigalosaurus dalmaticus, Opetiosaurus bucchichi, Trieste Aigialosaur, Carsosaurus marchesetti, Varanus salvator, Lanthanotus borneensis, Heloderma sp., and Clidastes propython (Table 2). A, PCA (Analysis I), exclusion ofthe Trieste Aigialosaur and the variable "length ofthe fibula." Slope of regression line for all taxa except Clidastes is 1.11 (r 2 = 0.955). B, PCA (Analysis II) including the Trieste Aigialosaur but with a reduction in the number of variables to seven, combining the data ofaigialosaurus and Opetiosaurus. Slope ofregression line for all taxa except Clidastes is (r 2 = 0.885). (Table 3). Thus, in defining limb and body characteristics, aspects of size variability and functionality, as inferred from shape, can be incorporated. Conventional varanoid taxonomy can easily be derived from the groupings shown here (Fig. 6A, B), suggesting that past hypotheses have been based on measures ofoverall similarity. Such studies have attempted to represent a phylogenetic continuum from morphometric information we interpret as a functional continuum (McDowell and Bogert, 1954; Borsuk-Bialynicka, 1984; Carroll and debraga, 1992). We have interpreted our PCA results to indicate two functionally disparate groups. The first is represented by one unique and disparate taxon, the aquatic mosasaur Clidastes. The second is a series of taxa from Lanthanotus to Varanus, arranged linearly along a functional continuum, indicating adaptive trends from

12 CALDWELL ETAL.-CARSOSAURUS PECTORAL GIRDLE 527 fossoriality to active locomotor pursuit. This suggests that aigialosaurs were more like Heloderma in terms oftheir locomotor capabilities and limb function. Further, aigialosaurs cannot be assumed to have been aquatic reptiles on the basis of limb morphometrics. Their limbs do not show any size or shape changes indicative ofaquatic adaptation anymore than do Heloderma and Lanthanotus. The lateral compression of the tail and changes in caudal centrum size and articular surfaces support hypotheses ofaquatic habits, but cannot prove them. There is no evidence that having a mosasaurian skull is an aquatic adaptation. In the context ofthe phylogenetic hypotheses presented (Fig. 4A, B) and the results of PCA (Fig. 6A, B), it is better stated that mosasaurs possess aigialosaurian cranial and caudal characteristics. The placement of Carsosaurus in "aigialosaur morphospace," as defined by the two most important principal components (Fig. 6A, B), and despite limitations ofthe data set, also provides some additional support for our preliminary phylogenetic hypotheses (Fig. 4A, B). Morphometrically, the intermediate position of Carsosaurus in both analyses (Fig. 6A, B) confirms gross morphological impressions that it is clearly closer in size and shape to other aigialosaurs than it is to nonaigialosaurs. The distance of Carsosaurus from Aigialosaurus (=Opetiosaurus) and the Trieste Aigialosaur along the size axis (PC1) may be explained by its larger size, a factor not removed by PCA. This analysis shows that Carsosaurus cannot be excluded from the Aigialosauridae on the basis ofminor size and shape ratios (Carroll and debraga, 1992). CONCLUSIONS Carsosaurus marchesetti is an important source of information for structure, function, and evolution of the postcranial skeleton during phylogenesis of early mosasauroid anguimorphs. Phylogenetic analysis indicates that C. marchesetti is an aigialosaur, and that mosasauroids (mosasaurs + aigialosaurs) are a monophyletic assemblage distinct from Varanoidea. The exact position of mosasauroids within Anguimorpha is uncertain but it is tentatively accepted here that membership is within that clade. Osteological information indicates that the skeleton ofcarsosaurus is very similar to that ofsome anguids, lanthanotids, and helodermatids (in carpus structure, large number of sternal ribs, absence of coracoid fenestra, and relative limb to body proportions). This alone suggests a terrestrial habit and terrestrial locomotor capabilities. The position of aigialosaurs, between Lanthanotus and Varanus, in a PCA comparing the proportions and shape of limbs to the body (Fig. 6A, B), can be easily interpreted in terms of function. Aigialosaurs are clearly intermediate in a linear series that describes active terrestrial locomotion to fossorial locomotor habits, further supporting the conceptualization of aigialosaurs as terrestrial animals. This information, in association with lithofacies analysis and interpretations of depositional environments from aigialosaur fossil localities (Calligaris, 1988), suggests that it is reasonable to conclude that aigialosaurs lived along or near coastal margins. They may in fact have had aquatic habits, but there is little osteological evidence to support speculations that aigialosaurs were obligatorily aquatic. Postcranially, they are plesiomorphic compared to their aquatic sistergroup (Mosasauridae). However, by virtue oftheir discovery in marine deposits, and cranial synapomorphies shared with mosasaurs, it has been concluded that these animals were at least facultatively aquatic. From the observations presented here, the only osteological feature which supports such an interpretation is the most posterior portion of the tail. The anterior caudals show the same vertical aspect of the neural spines and haemal arches as do terrestrial varanids. The relative proportions of the transverse processes are also similar to both varanids and helodermatids. Only the posterior caudals show both an absence of transverse processes and laterally compressed centra. Cranial characters, such as the purported aquatic adaptation ofa calcified tympanum, a synapomorphy of mosasaurs and aigialosaurs, is symplesiomorphic with their unknown terrestrial mosasauroid common ancestor. From this perspective, mosasaurian innovations for successful adaptation to aquatic environments were entirely postcranial. The primitive anguoid-like postcranium of Carsosaurus is significant in terms of limb evolution. In a global sense, the morphology of the terrestrial lacertilian limb has changed very little from that of early diapsids (Carroll and Currie, 1991; Caldwell, 1994). Only flying and limb-assisted swimming require structural modifications which significantly alter the bony patterns and proportions of the limb (limb loss is not considered in this statement). Swimming reptiles, such as mosasaurs and ichthyosaurs, reduced theirlimb proportions and evolved locomotor styles where propulsive forces were generated by muscles supported by the axial skeleton, not the limb skeleton (Carroll, 1985). This sort of shift in the limbs of Carsosaurus cannot be derived from either the qualitative descriptions we have given, nor from the results of quantitative analysis (PCA). The available postcranial material of other aigialosaurs shows no major differences in anatomy which would suggest family-level distinction for C. marchesetti. Phylogenetic relationships indicate that C. marchesetti is allied with other aigialosaurs and not with Varanoidea or Mosasauridae. Our analysis qualitatively (phylogenetic) and quantitatively (PCA) supports Carroll and debraga's (1992) suggestion that Opetiosaurus and Aigialosaurus may belong to the same genus; the posterior shortening of the ribs is the only equivocal character. If both are members ofthe same genus, then Aigialosaurus Kramberger, 1892 would have priority. Carsosaurus and the Trieste Aigialosaur cannot be differentiated qualitatively from Opetiosaurus on the basis of postcranial material, but we retain

13 528 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 15, NO.3, 1995 C. marchesetti as a valid taxon pending discovery of associated cranial and postcranial material. We conclude with the following systematic treatment of the aigialosaur taxa examined in this study Infraorder ANGUIMORPHA Fiirbringer, 1900 Superfamily MOSASAUROIDEA Camp, 1923 Family AIGIALOSAURIDAE Kramberger, 1892 Genus AIGIALOSAURUS Kramberger, 1892 Type Species-Aigialosaurus dalmaticus Kramberger, 1892, PI. 3. Referred Genus- Opetiosaurus Komhuber, Holotype-BSP Diagnosis-as in Kramberger (1892). Assigned Species-A. bucchichi (Komhuber, 1901), (Komhuber, 1901, PI. 1-2). Holotype-NMW unnumbered specimen. Diagnosis-as in Komhuber (1901). Genus CARSOSAURUS Komhuber, 1893 Type Species- Carsosaurus marchesetti Komhuber, 1893, PI. 1. Assigned Materials-Trieste specimen (MCSNT, 11430,11431, and [three parts]); Carroll and debraga (1992:figs. 4, 8). Holotype-MCSNT specimen figured by Komhuber (1893, PI. 1). Syntype-MCSNT, 11430, 11431,and 11432(three parts), figured by Carroll and debraga (1992:figs. 4,8). Diagnosis-as in Komhuber (1893). ACKNOWLEDGMENTS We thank P. Gaskill for illustrating five of the reconstructions shown in Figure 3. Data on mosasaurs was obtained by the first author while visiting the Museum of Paleontology, University of Kansas. The assistance of L. Martin and J. Chom is greatly appreciated. R. L. Carroll gratefully acknowledges the assistance of Dr. R. Calligaris and the staff at the Museo Civico di Storia Naturale, Trieste, Italy, while preparing, drawing, and measuring the specimen of Carsosaurus marchesetti. O. Rieppel and S. Evans are gratefully acknowledged for making useful comments on the manuscript. This research was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) Post Graduate Scholarship to M.W. Caldwell, and by NSERC operating grants to R. L. Carroll. LITERATURE CITED Bahl, K. N Skull of Varanus monitor (Linn.). Records ofthe Indian Museum (Journal ofindian Zoology) 39: Borsuk-Bialynicka, M Anguimorphans and related lizards from the Late Cretaceous of the Gobi Desert, Mongolia. Palaeontologica Polonica 46: Caldwell, M. W Developmental constraints and limb evolution in Permian and modern Lepidosauromorph diapsids. Journal of Vertebrate Paleontology 14: Calligaris, R I rettili fossili degli "Strati calcariei ittilitici di comeno" e dell'isola di lesina. Atti del Museo Civico di Storia naturale Trieste 41 : Camp, C. L California Mosasaurs. University of California Memoirs 13: Carroll, R. L Evolutionary constraints in aquatic diapsid reptiles. Special Papers in Palaeontology 33: and P. J. Currie The early radiation ofdiapsid reptiles; pp in H.-P. Schultze and L. Trueb (eds.), Origins ofthe Higher Groups oftetrapods. Comstock Publishing Associates, Ithaca. --- and M. debraga Aigialosaurs: Mid-Cretaceous varanoid lizards. Journal of Vertebrate Paleontology 12: Costelli, J., Jr., and M. K. Hecht The postcranial osteology of the lizard Shinisaurus: the appendicular skeleton. Herpetologica 27: debraga, M., and R. L. Carroll The origin of mosasaurs as a model of macroevolutionary patterns and processes. Evolutionary Biology 27: Estes, R., K. De Quieroz, and J. Gauthier Phylogenetic relationships within Squamata; pp in R. Estes and G. Pregill (eds.), Phylogenetic Relationships ofthe Lizard Families. Stanford University Press, Stanford. Fejervary, G. J. de Contributions to a monography on fossil Varanidae and Megalanidae. Budapesch, Annales Historico-Naturales Musei Nationalis Hungarici 16: Gauthier, J. A Fossil xenosaurid and anguid lizards from the early Eocene Wasatch Formation, southeast Wyoming, and a revision of Anguoidea. Contributions to Geology, University of Wyoming 21:7-54. Gilmore, C. W A newdescriptionofsaniwa ensidens Leidy, an extinct varanid lizard from Wyoming. Proceedings of the United States Museum 60: Good, D. A A phylogenetic analysis of cranial osteology in the gerrhonotine lizards (Squamata: Anguidae). Journal of Herpetology 21: The phylogenetic position offossils assigned to the Gerrhonotinae (Squamata: Anguidae). Journal of Vertebrate Paleontology 8: Haines, R. W The shoulder joint oflizards and the primitive reptilian shoulder mechanism. Journal of Anatomy 86: Hecht, M. K., and J. Costelli, Jr The postcranial osteology ofthe lizard Shinisaurus 1. The vertebral column. American Museum Novitates 2378:1-21. Hoffstetter, R., and J.-P. Gasc Observations sur Ie squelette cervical et specialement sur les hypapophyses des sauriens Varanoides actuels et fossiles. Bulletin du Museum D'Histoire Naturelle, Deuxieme Serie 39: and Chapter 5. Vertebrae and ribs of modern reptiles; pp in C. Gans (ed.), Biology of the Reptilia, Vol. 1. Academic Press, New York. Jenkins, F. A., and G. E Goslow The functional anatomy ofthe shoulder ofthe Savannah monitor lizard (Varanus exanthematicus). Journal ofmorphology 175: Kornhuber, A. G Carsosaurus marchesetti, ein neuer fossiler Lacertilier aus den Kreideschichten des Karstes

14 CALDWELL ET AL.-CARSOSAURUS PECTORAL GIRDLE 529 bei Komen. Abhandlungen der geologischen Reichsanstalt Wien 17: Opetiosaurus bucchichi, eine neue fossile Eidechse aus der unteren Kreide von Lesina in Dalamtien. Abhnadlungen der geologischen Reichsanstalt Wien 17: Kramberger, K. G Aigialosaurus, eine neue Eidechse aus den Kreideschiefern der Insel Lesina mit Riicksicht aufdie bereits beschriebenen Lacertiden von Comenund Lesina. Glasnik huvatskoga naravolosovnoga derstva (Societas historico-matulis croatica) u Zagrebu 7: Kuhn, O Ein neuer Lacertilier aus dem frankischen Lithographieschiefer. Neues Jahrbuch fur Geologie und Palaontologie, Monatshefte 1958: Lecuru, S. 1968a. Remarques sur Ie scapulo-coracoldes des Lacertiliens. Annales des Sciences Naturelles, Zoologie, Paris, Douzieme Serie, 10: b. Etude des variations morphologiques du sternum, des clavicules et de l'interclavicule des Lacertiliens. Annales des Sciences Naturelles, Zoologie, Paris, Douzieme Serie, 10: McDowell, S. B., Jr., and C. M. Bogert The systematic position of Lanthanotus and the affinities of the anguimorph lizards. Bulletin ofthe American Museum of Natural History 105: Meszoely, C. A. M North American fossil anguid lizards. Bulletin ofthe Museum ofcomparative Zoology 139: Norell, M. A., M. C. McKenna, and M. J. Novacek Estesia mongoliensis, a new fossil varanoid from the Late Cretaceous ofthe Barun Goyot Formation ofmongolia. American Museum Novitates 3045: Pregill, G. K., J. A. Gauthier, and H. W Greene The evolution of helodermatid squamates, with description of a new taxon and an overview of Varanoidea. Transactions ofthe San Diego Society ofnatural History 21: Renous-Lecuru, S Morphologie comparee du carpe chez les Lepidosauriens actuels (Rhyncocephales, Lacertiliens, Amphisbeniens). Gegenbaurs morphologisches Jahrbuch, Leipzig 119: Rieppel, O. 1980a. The Phylogeny ofanguinomorph Lizards. Birkhauser Verlag, Basel, 86 pp b. The postcranial skeleton of Lanthanotus borneensis (Reptilia, Lacertilia). Amphibia-Reptilia 1: The skeleton ofajuvenile Lanthanotus (Varanoidea). Amphibia-Reptilia 13: Russell, D. A Systematics and morphology ofamerican mosasaurs. Peabody Museum of Natural History, Yale University Bulletin 23: Swofford, D. L PAUP: Phylogenetic Analysis Using Parsimony, Version Computer program distributed by the Illinois Natural History Survey, Champaign, Illinois. Received 9 September 1993; accepted 14 July APPENDIX I. Data Matrix Gerrhonotus Heloderma Lanthanotus Varanus Estesia OlIO? Mosasauridae 10? ? Aigialosaurus????? 1??00?OOO? 0010??OO?? 0100? Opetiosaurus 10????1????OO?? OO????OO?? O?OOO Trieste Aigialosaur?????????????????????????????? Carsosaurus?????????????????????????????? Gerrhonotus Heloderma Lanthanotus Varanus Estesia?OOOO 1111? 111?0?1??0 10?????????????? Mosasauridae Aigialosaurus 100??????????????? ?0??1?0 0 Opetiosaurus 100???????????????11 010?? 010?? 101?? 0 Trieste Aigialosaur 1000??????????????? ?????? Carsosaurus???????????????????????1? Character state abbreviations:? = not preserved/unknown.

15 530 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 15, NO.3, 1995 APPENDIX II SPECIMENS AND CHARACTER DESCRIPTIONS Descriptions ofcharacters and character states identified in the varanoids and mosasauroids examined. Numbered characters followed by parentheses containing a number and an 'R' are found in Rieppel (1980a); those with a 'P' are found in Pregill et al. (1986). In several cases characters were recoded to reflect the character states of taxa considered for this analysis. Examination ofthe following specimens and published descriptions were used in association with Pregill et al. (1986) and Rieppel (1980a): Varanus salvator RM a, and Bahl (1938), McDowell and Bogert (1954), Norelletal. (1992); Lanthanotus, FMNH , and McDowell and Bogert (1954), Rieppel (1980b; 1992), Norell et al. (1992); Heloderma sp., RM specimen, and McDowell and Bogert (1954), Gauthier (1982), Norell et al. (1992); Estesia, Norelletal. (1992); Clidastes prophython, KU 1022 and Russell (1967); Aigialosaurus dalmaticus, BSP , and Kramberger (1892), Carroll and debraga (1992); Opetiosaurus buccichi, NMW specimen, and Kornhuber (1901), Carroll and debraga (1992); the Trieste Aigialosaur, MCSNT, 11430, 11431, and (three parts), and Carroll and debraga, (1992); and Carsosaurus marchesetti, MCSNT unnumbered specimen, and Kornhuber (1893). Cranium 1(IR) 2 (23P) 3 (2R) 4 (2P) 5 (3--4P) 6 (5P) 7 (3R) 8 (6R) 9 (28-29P) 10 (30P) 11 (lor) 12 (13R) 13 (14R) Numberofpremaxillary teeth: more than 6 (0); less than 6 (1). Premaxillary teeth, size: same size or larger than maxillary teeth (0); abruptly smaller than maxillary teeth (1). Nasal process ofpremaxilla elongate entering deeply between nasal bones: no (0); yes (1). Nasal and prefrontal bones: little or no contact (0); broad contact (1). Nasal and maxillaries: contact (0); no contact (1). Nasal process of maxilla: arises from middle ofmaxilla (0); arises on posterior aspect of maxilla (1). Premaxillary foramen: absent (0); present (1). Vomerine process of premaxilla: short and blunt (0); elongate (1). Maxillary tooth number: greater than twelve (0); less than twelve (1). Venom groove on maxillary teeth: absent (0); present (1). Nasal bones: paired (0); fused (1). Frontal bones: fused (0); paired (1). Contribution offrontal to dorsal margin oforbit: enter the dorsal margin (0); excluded from margin (1). 14 (16R) 15 (17R) 16 (lop) 17 (20R) 18 (22R) 19 (23R) 20 (24R) 21 (25R) 22 (27R) 23 (30R) 24 (32R) 25 (33R) 26 (35R) 27 (14P) 28 (13P) 29 (18P) 30 (37R) 31 (38R) 32 (39R) 33 (20P) 34 (41R) Palate 35 (42R) 36 (44R) Descensus frontalis development: poorly developed with no sutural contact below olfactory tract (0); well developed and contacts in a suture below olfactory tract (1). Descensus frontalis separation of orbitae: does not separate orbits (0); separates orbits (1). Prefrontal contact with postfrontal: no contact (0); contact or closely approaches (1). Postfrontal and postorbital: separate (0); fused (1). Lacrimal bone lateral exposure: extensive (0); narrow (1). Lacrimal spine on caudal edge lateral to lacrimal foramen: no spine (0); spine present (1). Lacrimal separation of prefrontal from jugal: yes (0); no (1). Foramen pro ductus lacrimalis: single (0); double (1). Jugal has distinct triangular projection on caudo-ventral edge: yes (0); no (1). Pineal foramen: present (0); absent (1). Supratemporal process of parietal contact: contacts distal end of paroccipital process (0); separate from paroccipital process (1). Contact between supraoccipital and parietal: narrow (0); broad contact (1). Upper temporal arch: complete (0); interrupted (1). Temporal musculature insertion on parietal table: ventral (0); dorsal (1). Squamosal size and contact with postorbital process: large and contacts postorbital (0); small and reduced (1). Supratemporal process ofparietal in dorsal aspect: broad (0); narrow (1). Squamosal contact with supratemporal process ofparietal: no contact (0); in front of supratemporal (1). Posterior tip of squamosal articulation with quadrate: in a socket on dorsal surface ofquadrate (0); notch on laterodorsal edge ofquadrate (1). Anterodorsal edge of quadrate: broad transversely (0); narrow and pointed (1). Quadrate outer conch: large (0); reduced (1). Mesial crest of quadrate: distinct (0); weakly developed (1). Palate: neochoanate (0); palaeochoanate (1). Length ofchoanal groove: more than half

16 CALDWELL ETAL.-CARSOSAURUS PECTORAL GIRDLE 531 the length of body of palatine (0); less than half the length (1). 37 (46R) Length of mesial palatine process articulating with dorsal surface of pterygoid: elongate (0); short (1). 38 (48R) Palatal shelves of vomers: present (0); absent (1). 39 (33P) Palatines size: longer than wide (0); width roughly equal to length (1). 40 (35P) Pterygoid teeth: present (0); absent (1). Brain Case 41 (49R) Lateral head on distal end of ectopterygoid: concealed behindjugal (0); exposed laterally (1). 42 (36P) Ectopterygoid contact with palatine anteriorly: no contact (0); contact (1). 43 (50R) Length and articulation ofproximal head of ectopterygoid: short and lies along dorsal edge of pterygoid (0); long and contacts body of pterygoid (1). 44 (52R) Spine on mesial tip of medial pterygoid ridge which articulates basipterygoid process: absent (0); present (1). 45 (55R) Crista prootica development: well developed (0); weakly developed (1). 46 (57R) Basisphenoid participation in sphenooccipital tubercle: excluded (0); included (1). 47 (59R) Foramina for facial nerve; single (0); double (1). 48 (16P) Hypoglossal foramen size: not enlarged (0); enlarged (1). Mandible 49 (38N) Coronoid and surangular processes of dentary: well developed (0); reduced (1). 50 (63R) Suture between angular and splenial: angular, interdigitating (0); vertical and straight (1). 51 (64R) Medial exposure ofangular compared to lateral exposure: lateral greater (0); medial greater (1). 52 (65R) 53 (68R) Anterior extension ofsurangular beyond the anterior process of coronoid: does (0); does not (1). Splenial contact with the dentary: contact in front ofinferior alveolar foramen to tip of splenial (0); no contact and Meckels groove open anteriorly (1). Axial and Appendicular Skeleton 54 Length ofcervical centra relative to trunk centra: twice as long (0); centra equal (1); centra subequal or shorter (2). 55 (49P) Number ofcervical vertebrae: eight (0); nine (1); seven (2). 56 (53P) Peduncles on cervical and caudal vertebrae: short (0); long (1). 57 (51 P) Number of presacral vertebrae: greater than 30 (0); less than 30 (1). 58 Elongate trunk ribs posterior on vertebral column: to trunk vertebrae (0); to vertebrae (1); to trunk vertebrae 7-9 (2). 59 (61P) Number ofrib attachments on sternum: more than three pairs (0); three or fewer pairs (1). 60 Posterior iliac process: extends the length of two vertebrae (0); extends the length of three vertebrae (1); no posterior extension (2). 61 Anterior coracoid fenestra: present (0); absent (1). 62 Posterior coracoid fenestra: absent (0); present (1). 63 Proximal portion of olecranon: large process (0); moderate sized process (1); process absent (2). 64 (59P) Size ofinterclavicle anterior process: long (0); short or absent (1). 65 Intermedium ossification of carpus: present (0); absent (1). 66 Ulna length: approximately % length of humerus (0); approximately the same length (1).