Caldwell, 2000b

720

Journal of Vertebrate Paleontology 20(4):720–735, December 2000q 2000 by the Society of Vertebrate Paleontology

ON THE AQUATIC SQUAMATE DOLICHOSAURUS LONGICOLLIS OWEN, 1850 (CENOMANIAN,UPPER CRETACEOUS), AND THE EVOLUTION OF ELONGATE NECKS IN SQUAMATES

MICHAEL W. CALDWELL*Paleobiology, Research Division, Canadian Museum of Nature, P.O. Box 3443, Station ‘D’, Ottawa, Ontario, Canada K1P 6P4

ABSTRACT—The marine squamate, Dolichosaurus longicollis, from the Upper Cretaceous (Cenomanian) Chalk de-posits of southeast England is redescribed. The elongate neck of D. longicollis is produced by an increased number ofcervical vertebrae. Cervical peduncles are elongate, curved and are not fused to the hypapophyses. There is no sca-pulocoracoid fenestra, the coracoid is not emarginated, and the scapula and coracoid are not fused. The splenial andangular articulate in a ball-and-socket joint similar to that of mosasaurs and Coniasaurus crassidens. The forelimb andpectoral girdle elements show evidence of reduction as compared to the pelvic girdle and rearlimb. Cladistic analysisof six mosasaur taxa, three ‘aigialosaur’ taxa, Coniasaurus crassidens, Coniasaurus gracilodens, and D. longicollis,using 66 characters, found 27 most parsimonious cladograms (MPCs): 122 steps; C.I. 0.648; H.I. 0.352; R.I. 0.669. AStrict Consensus Tree found support for the monophyly of the Mosasauridae and Aigialosauridae; sister-group rela-tionships between coniasaurs, Dolichosaurus, Aigialosauridae and Mosasauridae are an unresolved polytomy. A Ma-jority Rule Consensus Tree finds Dolichosaurus as sistergroup to (C. crassidens, C. gracilodens (Aigialosauridae(Mosasauridae))) in nine (33%) of the MPCs. Lack of support for a more inclusive Dolichosauridae composed ofDolichosaurus 1 (C. crassidens, C. gracilodens) is attributed to the incompleteness of the fossil remains of these threetaxa. Presence/absence of a pectoral girdle currently defines the presence/absence of a neck. This definition is insuffi-cient and hypapophyses are found more informative regarding taxic differences and transformational scenarios. Thepaleobiology of Dolichosaurus is reconstructed as similar to coniasaurs, nothosaurs, and modern sea snakes.

INTRODUCTION

Owen (1850) described two monotypic genera of marine liz-ards, Dolichosaurus longicollis and Coniasaurus crassidens,based on a number of specimens collected from the LowerChalk (Cenomanian; Upper Cretaceous) of southeast England.Owen used the exceedingly large number of cervical and dorsalvertebrae to diagnose Dolichosaurus, and tooth characteristicsto diagnose Coniasaurus. Owen’s diagnoses did not identifyany characteristics shared between these two taxa.

Nopcsa (1908) restudied Owen’s specimens and added sev-eral new specimens to the list of known coniasaurs and doli-chosaurs. Unfortunately, for Coniasaurus, Owen (1850) mis-identified the tooth-bearing element of the type specimen (amaxilla) as a dentary. Nopcsa’s (1908) re-descriptions built onthis error and further confounded the problem by not locatingand then incorrectly identifying the type specimen (the type wasin Brighton and Nopcsa only worked with specimens at the thenBritish Museum of Natural History). Recently, Caldwell andCooper (1999) located and redescribed the type maxilla andvertebrae of C. crassidens, and assigned a number of othertooth-bearing elements to Coniasaurus crassidens. Caldwell’s(1999a) description of Coniasaurus gracilodens, sp. nov. high-lighted several notable differences with C. crassidens such astooth shape, maxillary tooth number, and robustness of themaxilla (the maxilla of C. crassidens is much more elongatethan that of C. gracilodens). Both species of Coniasaurus areknown from disarticulated skulls with isolated vertebral ele-ments, but no more than fragmentary and disarticulated post-cranial remains.

In contrast, the type and referred specimens of Dolichosauruslongicollis (see below) are articulated postcranial skeletons withonly one very fragmentary skull and no preserved teeth.Owen’s (1850) original description included the type (the frag-mentary skull and articulated anterior part of the postcranium)

* Current address (Effective July 1, 2000): Department of Earth andAtmospheric Sciences, and Department of Biological Science, Univer-sity of Alberta, Edmonton, Alberta, Canada T6G 2E9.

and a referred specimen (the posterior dorsal, sacral, and caudalvertebral series, and pelvic girdle/hindlimb elements); he con-sidered these two pieces to be the front and back halves of oneindividual. His justification for this assumption was that bothcame from the same quarry and were collected at about thesame time (Owen, 1850).

Phylogenetically, coniasaurs and dolichosaurs have been con-sidered to be nested within basal mosasauroids (mosasaurs andaigialosaurs) (Nopcsa, 1908), or recently, as the sistergroup tomosasauroids (Caldwell, 1999b). The difficulty in resolvingthese problematic relationships arises from the disparate andnon-comparable fossil data forcing the creation and mainte-nance of two different genera. Dolichosaurus is known fromcomplete postcrania with few skull bones, while both speciesof Coniasaurus are known from skulls and a small number ofdisarticulated vertebrae.

Recent studies proposing a sistergroup relationship betweensnakes, coniasaurs, and mosasauroids, a group referred to as thePythonomorpha (Caldwell and Lee, 1997; Lee, 1997; Lee andCaldwell, 1998; 2000), have highlighted the phylogenetic im-portance of coniasaurs, dolichosaurs, and other similar Meso-zoic marine squamates. The Pythonomorpha excludes the fos-sorial amphisbaenids and dibamids, which are usually consid-ered to be the closest squamate relatives of snakes (see Bellairsand Underwood, 1951; Wu et al., 1996). This new phylogenetichypothesis, suggesting that snakes, mosasauroids, dolichosaurs,and coniasaurs may have a common aquatic ancestor, stands incontrast to the previous hypothesis of a fossorial snake origin.And, as a result of this new hypothesis of relationships, featuresof Dolichosaurus, i.e., the elongate neck, the possession of zyg-osphenes and zygantra, and the morphology of the intraman-dibular joint, are synapomorphies, and subsequently, by phy-logenetic inference, are homologies. As homologies, these char-acters are important starting points for reasoned inference andargumentation on the origins of snakes and related forms.Therefore, to address these important issues in squamate phy-logeny, this study re-examines the morphology and systematicsof Dolichosaurus longicollis Owen, 1850, and its interrelation-ships with other dolichosaurs, aigialosaurs, and mosasaurs.

721CALDWELL—UPPER CRETACEOUS AQUATIC SQUAMATE

METHODS AND MATERIALS

Osteological Analysis

The holotype specimen required minor preparation to exposethe partial skull, braincase, and cervical vertebrae. Drawingsand illustrations were made using a dissecting microscope andcamera lucida attachment. Photographs were taken by the au-thor and by the photographic department of The Natural HistoryMuseum (British Museum), London. Measurements were madeusing digital calipers.

Institutional Abbreviations: BMB, Booth Museum of Nat-ural History, Brighton, Sussex, England; BMNH, The NaturalHistory Museum (British Museum), London, England; BSP,Bayerische Staatssammlung fur Palaontologie und historischeGeologie, Munchen, Germany; FMNH, Field Museum of Nat-ural History, Chicago, Illinois, U.S.A.; GBA, Austrian Geolog-ical Survey, Wien, Austria; MCSNT, Museo Civico di StoriaNaturale, Trieste, Italy; NMW, Naturhistorisches Museum,Wien, Austria.

Cladistic Analysis

The phylogenetic relationships of six genera of mosasaur,three species of aigialosaur, two species of coniasaur, and Dol-ichosaurus longicollis, were examined by cladistic analysis ofa character matrix of 66 osteological characters (Appendices 1and 2). Justification for limiting the taxonomic scope of thisanalysis to these twelve terminal taxa is based on more detailedstudies of the interrelationships of these taxa with other squa-mates (Caldwell, 1999b), and with each other (Caldwell et al.,1995; Caldwell, 1996, 1999a; Bell, 1997). The conclusions ofprevious studies regarding the mosasauroid/pythonomorph af-finities of other dolichosaurs (Nopcsa, 1908, 1923; Caldwell,1999b; Lee and Caldwell, in press) are also accepted as a jus-tification for the taxonomic sample examined in this study.

Cladograms were generated by submitting the character ma-trix derived from study of the above squamates (Appendix 2)to Branch and Bound algorithms used in the computer softwareapplication PAUP Version 3.1.1 for the Macintosh (Swofford,1993). For determining character polarity, an outgroup codewas constructed from the character states of Pachyrhachis (Leeand Caldwell, 1998), Varanus salvator (unnumbered specimen,Redpath Museum), and Gerrhonotus leiocephalus (unnumberedspecimen, Redpath Museum). Outgroup choice was based onthe sistergroup relationships of mosasauroids as postulated byCaldwell (1999b) and Lee (1998). All characters were analyzedunordered and without character weight assignments. Characterstate transformations were ACCTRAN optimized so that char-acters appear on more internal nodes, i.e., lower in the tree.This procedure identifies synapomorphies (unequivocal andequivocal character states) for more inclusive clades, rather thanbeing optimized as apomorphies of more exclusive clades oreven terminal taxa (DELTRAN).

Character states for Aigialosaurus dalmaticus (BSP1902II501) were based on examination of latex peels and pho-tographs of the type specimen, and from reference to Kram-berger (1892), Carroll and DeBraga (1992), DeBraga and Car-roll (1993), and Bell (1997). Character states for Aigialosaurus(5Opetiosaurus) buccichi were obtained by examination of theholotype specimen (NMW unnumbered specimen) and fromGBA 1901/2, the recently rediscovered counterpart to the ho-lotype specimen, and from Kornhuber (1901). Character statesfor Carsosaurus marchesetti were obtained by examination ofthe holotype specimen (MCSNT unnumbered specimen) andfrom referred specimens MCSNT 11430, 11431, 11432 (inthree parts), and from Kornhuber (1893).

SYSTEMATIC PALAEONTOLOGY

SQUAMATA Oppel, 1811DOLICHOSAURIDAE Gervais, 1852

Genus DOLICHOSAURUS Owen, 1850

Type Species—Dolichosaurus longicollis Owen, 1850Range—Lower to Upper Cenomanian: Chalk Marl Fm.

through Plenus Marls. (Mantelliceras mantelli Zone through theAcanthoceras jukesbrowni Zone [equivalent bio-zones Actino-camax varius Zone through the Holaster subglobosus Zone](Fig.1).

Generic Diagnosis—At least nineteen cervical vertebraewith large posteriorly placed hypapophyses and unfused pedun-cles; at least thirty-two trunk vertebrae; long, slightly curvedribs, body laterally compressed; scapulocoracoid fenestra ab-sent; forelimb shorter and less robust than hindlimb; humerussubequal in length to two anterior dorsal vertebrae; ectepicon-dylar foramen absent; at least one pygal vertebra.

DOLICHOSAURUS LONGICOLLIS

(Figs. 2–12)

Diagnosis—As for genus.Type Locality and Horizon—Collected near Burham, Kent,

and is considered to be from the Lower Chalk (Abbotts CliffChalk Fm. [North Downs Nomenclature]—Zig Zag Chalk Fm.[South Downs Nomenclature]), Lower Cenomanian (Fig. 1).

Holotype—BMNH R 49002 (Figs. 3–7), a small chalk blockpreserving the articulated cranial and postcranial remains of oneindividual. The skull is represented by a partial braincase, andfragmentary right and left mandibles. The postcranium is com-posed of 17 articulated cervical vertebrae and 19 anterior dorsalvertebrae.

Referred Material—BMNH R49907 (Figs. 8A, B, 9) South-erham Pit, near Lewes, Sussex; purchased 1879, Capron Col-lection. BMNH R32268 (Fig. 10A, B) from Burham, Kent, pre-sented by P. de M.G. Egerton, 1856. BMB 0085687 (Fig. 11A,B, C), three prepared blocks, Southerham Pit, near Lewes, Sus-sex.

Remarks—BMNH R49002 was most recently mentionedand figured in part by Milner (1987:pl. 58, fig. 5) in a taxo-nomic list of reptiles of the English Chalk. In Milner’s descrip-tion BMNH R49002 and BMNH R49907 were identified asDolichosaurus longicollis along with BMNH R44141 (the latterhas been described as a new species of Coniasaurus by Cald-well, 1999a). The result of the description of BMNH R44141as a new species of Coniasaurus, is that now only the holotype,BMNH R49002, possesses any cranial material, and that D.longicollis is diagnosed by characters that are not comparableto the known osteology of Coniasaurus. In contrast, both Con-iasaurus crassidens (Caldwell and Cooper, 1999), and Coni-asaurus gracilodens (Caldwell, 1999a) are both known fromdiagnostic cranial remains. The associated postcranial elementscannot be differentiated from those of Dolichosaurus, but nei-ther are they diagnostic of Coniasaurus.

Dolichosaurus longicollis is diagnosed by an extremely elon-gate neck composed of at least seventeen cervical vertebrae (thecount is likely nineteen: 171atlas-axis), an elongate trunk withat least 32 dorsal vertebrae, and perhaps as high as 38 (Owen1850, considered BMNH R49002 and R32268 to be one indi-vidual), and the relative length of the humerus as compared totwo anterior dorsal vertebrae. It is this last character, drawnfrom the holotype specimen, that allows the referral of BMNHR49907 to Dolichosaurus longicollis. Because the assignmentof specimens to D. longicollis is dependent on the possessionof specific postcranial characters, and because it is possible thatsome of these specimens could be reassigned to one or the other

722 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 20, NO. 4, 2000

FIGURE 1. Lower Chalk stratigraphy for the North and South Downs of southeast England (Anglo-Paris Basin), and the localities and approx-imated stratigraphic positions of known specimens of Dolichosaurus longicollis. 1, BMNH 49002 and BMNH R32268 Lower Chalk, Burham,Kent; 2, BMNH R49907 and BMB 008567. Question mark (?), uncertainty of stratigraphic range of section. Dotted lines, uncertain horizon ofspecimens. Stratigraphic column compiled from: Owen (1987); Mortimore (1986; 1987; pers. comm.); Robinson (1986, 1987); Hancock et al.(1993); and Obradovich (1993). Abbreviations: CTBE, Cenomanian-Turonian Boundary Event; GM, Glauconitic Marl; MR, Melbourn Rock.

coniasaur species, each postcranial specimen is described in-dependent of the others.

The possibility remains that more complete specimens willindicate that either Coniasaurus crassidens or Coniasaurus gra-cilodens is a senior synonym of Dolichosaurus longicollis. InOwen’s (1850) original description, C. crassidens occurs beforeD. longicollis in the manuscript and therefore has generic pri-ority. Whatever the nomenclatorial outcome, it is certainly truethat there were at least two species (crassidens and gracilodens)of small marine squamates present in the seas that covered theAnglo-Paris Basin during the Cenomanian.

DESCRIPTION

Skull

Parietal (Fig. 6A, B) —The parietal is fragmentary and pre-served upside down as a hexagonal, concave plate of dermalbone. The posterior margin is present and preserves the pit forarticulation with the supraoccipital. The paired, symmetricalborders on either side of the pit represent the breakage lines forthe suspensorial rami of the parietal. It is not possible to deter-

mine the position of the parietal foramen as the anterior portionof the bone is broken away and turned into the matrix. Othermargins and suture lines with the frontal are not visible.

Braincase (Figs. 6A, B, 7) —The braincase is badly crushedand fragmentary, making exact identification of elements prob-lematic. As a result, useful comparisons with squamates suchas Varanus (see Bahl, 1938), mosasaurs (see Russell, 1967), orsnakes (Rieppel, 1979) are not possible. However, three inter-pretations and identifications of the various elements and fo-ramina are given; the presence of an exoccipital-opisthotic isconsistent in all three interpretations. Element and foramenidentities as given in Figure 7 follow Interpretation/Identifica-tion 1, but it is recognized that these identifications are tenta-tive.

Interpretation/Identification 1: the braincase is exposed inside view and is broken in saggittal section. If this were accu-rate, then the bones present would be the left portions of thebasioccipital and basisphenoid, a fragment of the left exoccip-ital-opisthotic and prootic. In most squamate families, includingthe Mosasauridae, the exoccipital and opisthotic are fused, butthe prootic is an independent element (Bahl, 1938; Russell,


FIGURE 2. Life reconstruction of Dolichosaurus longicollis. Head reconstructed using elements from Coniasaurus crassidens (Caldwell andCooper, 1999) and Coniasaurus gracilodens (Caldwell, 1999a).

FIGURE 4. Line drawing of dorsal view of holotype specimen ofDolichosaurus longicollis (BMNH R49002). Abbreviations: br, brain-case fragment composed of left prootic, opisthotic, basisphenoid, andbasioccipital; H, humerus; 11, 11th preserved cervical vertebra, figuredin Figure 12; 17, 17th cervical, likely C19. Scale bar equals 1 cm.

FIGURE 3. Dorsal view of holotype specimen of Dolichosaurus lon-gicollis (BMNH R49002). Scale bar equals 1 cm.

1967); in snakes, the exoccipital is free, but the opisthotic andprootic are fused. The foramina present on the exoccipital-op-isthotic might well correspond to the foramina for the eleventh(XI) and twelfth (XII) cranial nerves (Bahl, 1938).

Interpretation/Identification 2: the braincase is exposed indorsal view and is split about mid-height through the prooticsand exoccipitals-opisthotics. The right side of the lateral brain-case wall is missing as are the processes from the right side ofthe basioccipital. Fragments of the left supraoccipital, exoccip-ital and prootic have collapsed onto the floor of the basioccip-ital-basisphenoid.

Interpretation/Identification 3 (Susan Evans, pers. comm.): ifthis is a medial view of the left half of the braincase, then theelement labeled ‘‘ex-op’’ (Fig. 7) could be part of the supra-occipital. The foramen labeled XI might actually be the pointwhere the endolymphatic duct emerges from the supraoccipital(which better fits the angle of emergence of that foramen shownin Fig. 7). In that case, the large foramen labeled XII might be

that for the vagus nerve (CN-X) with the mass to the left of itbeing the bulk of the exoccipital.

Mandible

Dentary (Fig. 5A–C)—The right and left dentaries are ex-tremely fragmentary, appear to have been weathered throughthe middle of the bone on its long axis, and are fixed to thelateral/internal surface of the splenial. There are no teeth re-maining on either dentary even though in Owen’s (1850) pub-lication he does figure some small teeth in place on the dentary.

Splenial (Fig. 5A–C)—The right splenial is exposed in in-ternal view and is rotated vertically and out of articulation withthe postdentary and dentary elements. The posterior end isweathered. The splenial foramen is present just anterior to thearticular surface that would have formed the joint with the an-gular; on the left splenial the splenial foramen is located ap-proximately 4 mm anterior to the angular-splenial joint. The


FIGURE 5. Fragmentary skull elements of holotype specimen of Dolichosaurus longicollis (BMNH R49002): A, lateral view, fragmentary rightlower jaw elements; B, lateral view, fragmentary left lower jaw elements; C, detailed lateral view of fragmentary left lower jaw elements asshown in photograph for B. Abbreviations: Ang, angular; Cor, coronoid; Dent, dentary; Splen, splenial; sur, surangular. Scale bar equals 1 cm.

splenial is L-shaped in cross-section. The long side of the boneis exposed medially in a high ascending flange as in Coniasau-rus crassidens (see Caldwell and Cooper, 1999).

Coronoid (Fig. 5A–C)—The left coronid bears a deepgroove for the surangular. It appears to have a fairly long an-terior process and would have extended anteriorly along thelength of the surangular. The right coronoid is not preserved.

Surangular (Fig. 5A)—The right surangular is present butis weathered posteriorly, preserving no details of its contribu-tion to the glenoid facet. As preserved, the element is elongate,with a broad curving crest on the lateral surface that underliesthe coronoid process and continues anteriorly to the suture withthe dentary. The coronoid process is short but well-defined. Theanterior-most portion of the surangular is broken.

Angular (Fig. 5A)—The left angular is a small thin bonethat is ventral and medial to the surangular. On the left side thesplenial and angular are in articulation. The contact betweenthese two bones forms a ball and socket-like joint; the splenialhead, is slightly concave, while the angular head is slightlyconvex. The joint formed is identical to that of Coniasauruscrassidens (Caldwell and Cooper, 1999), all mosasauroid squa-mates, and snakes (Lee et al., 1999).

Axial Skeleton

Cervical Vertebrae (Figs. 3, 4, 12)—There are seventeencervical vertebrae preserved. However, this does not include theatlas and axis vertebrae as these two elements cannot be iden-


FIGURE 6. Fragmentary skull and braincase elements of holotype specimen Dolichosaurus longicollis (BMNH R49002): A, dorsal view ofhead. Note braincase elements; B, detail of braincase elements. Note left prootic, opisthotic, and fragmentary basioccipital-basisphenoid. Abbre-viations: Bo, Basioccipital; Cor, coronoid; Ex-Op, fused exocciptal-opisthotic; Par, paritetal; Pr, prootic; soc-p, pit for articulation with thesupraoccipital; Splen, splenial; Scale bar equals 1 cm.

tified from among the badly weathered bone fragments locatedbetween the mandibular elements. Therefore, the total count isat least nineteen and could be one or two vertebrae higher,particularly if the anterior-most dorsals did not bear ribs thatwere fixed to the sternum. This however, cannot be ascertainedfrom the specimen as preserved.

The individual vertebrae are low and flattened. The neuralspines are broken away but do not appear to have been verytall. The articulation of the pre- and postzygapophyses is nearlyhorizontal. Zygosphenes and zygantra are well developed andthe neural arch laminae are notched. In most modern snakesthe laminae are straight.

The cervicals are shorter and less robust than the dorsal ver-tebrae; the average length of a cervical vertebra is 9.4 mm asmeasured along the exposed portion of the neural arch. On av-erage the vertebrae are approximately 6mm wide measuredacross the top of the neural arch between the pre- and postzy-gapophyses.

Thin, short ribs are present from at least the fourth cervical

vertebrae. Details of the parapophyses of more anterior cervi-cals have weathered away as have any ribs that may have beenpresent. Intercentra are long, thin and directed posteriorly. Thehypapophyses are located on the posterior portion of the cen-trum.

Dorsal Vertebrae (Figs. 3, 4)—There are nineteen dorsalvertebrae preserved. It is not possible to determine the numberof ribs that articulated with the sternum, nor to establish whichribs and vertebra is actually the first dorsal as no articulationsand chondral cartilages are preserved.

The dorsal vertebrae are larger than the cervicals in bothlength and width. The average length of a dorsal vertebra is 9.6mm, measured along the neural lamina, and 10.4 mm, measuredalong the length of the centrum from the ventral lip of thecotyle to the end of the condyle. On average, the vertebrae areapproximately 6mm wide measured across the top of the neuralarch between the pre- and postzygapophyses. Vertebrae narrowimmediately behind the parapophyses and do not become wideror narrower at the condyle. The condyle is circular and the


FIGURE 7. Fragmentary left wall of braincase, holotype specimen ofDolichosaurus longicollis (BMNH R49002). The element not stippledis unknown, but may be a fragment of the decensus parietalis. Abbre-viations: Bo, Basioccipital; Ex-Op, fused exocciptal-opisthotic; fBa,foramen/canal for Basilar artery; Pr, prootic; XI, foramen for 11th cra-nial nerve; XII, foramen for 12th cranial nerve. Scale bar equals 5 mm.

cotyle is slightly elliptical in the horizontal plane. The articu-lation of the pre- and postzygapophyses are in the horizontalplane and the accessory articulations of the zygosphene-zyg-antrum are well developed.

The ribs of the dorsal vertebrae increase in size from anteriorto posterior. The largest rib is approximately 41 mm long andabout 2 mm wide at the proximal end.

Appendicular Skeleton (Figs. 3, 4)—Fragments of the ster-nal cartilages, coracoid, and scapula are present. However, de-tails of these elements are not visible. The only appendicularelement that is well preserved is the right humerus, a robustbone with well developed muscle attachments. The humerus is15 mm long, 6.1 mm wide across the proximal head, and 5.3mm wide at the distal end. The ratio of humerus length to thelength of an average dorsal vertebra is 1.44:1.

The distal epiphyses are absent and likely did not fuse to thebone. The proximal epiphysis appears fused to the bone as thisarea is rounded and presents a similar morphology to that ofother squamates. There is no evidence of either an ectepicon-dylar foramen or groove.

BMNH R49907

This specimen was collected at the Southerham Pit, nearLewes, Sussex. It was purchased by the Natural History Mu-seum, London, in 1879, as part of the Capron Collection.

Forelimb and Pectoral Girdle (Figs. 8, 9)—The right scap-ula and right coracoid are well preserved and in articulation.The elements are not co-ossified, but rather are sutured to eachother as in Coniasaurus gracilodens (Caldwell, 1999a) and theaigialosaur, Carsosaurus marchesetti (see Caldwell et al.,1995). The scapula is small and the scapular blade descendsinto the matrix. Nopcsa (1908) interprets this small scapula to

be a procoracoid element; however, there is no justification forthis identification. The principal difference between the scapu-locoracoid of Dolichosaurus longicollis and that of coniasaursor aigialosaurs, is the apparent lack of a scapulocoracoid fe-nestra; such a feature is strongly pronounced in aigialosaurs,more weakly manifest in Coniasaurus gracilodens.

The right coracoid is still in articulation with the supracora-coid cartilage although the exact margins of this calcified tissueare not obvious. The coracoid margin is not emarginate, andthe bone is perforated anteromedial to the glenoid fossa by asingle coracoid foramen.

There are three identifiable rib attachment points on the ster-nal cartilage in addition to the paired xiphisternal cartilagespresent at its extreme posterior tip. As compared to the aigialo-saur Carsosaurus marchesetti with five rib articulations and apair of xiphisternal processes (Caldwell et al., 1995), it is fairto conclude that Dolichosaurs longicollis shows reduction ofthe pectoral girdle elements.

The shafts of the right and left humeri are present, but theproximal and distal ends have been broken away and details oftheir articular surfaces are unknown. The right radius is narrowproximally, and widens distally to form a preaxial to postaxiallyelongate boot. The right and left ulnae appear to be slightlywider proximally (preaxial to postaxial) than distally. The ra-dius and ulna appear to be approximately equal in length. Car-pal elements are present on the left limb, though poorly pre-served. An elliptical ulnare and possible fragments of a pisiformare present. Metacarpals two and three are approximately halfthe length of the ulna, and several phalangeal elements are alsopreserved.

Dorsal Vertebrae (Figs. 8, 9) —There are six dorsals andthe fragments or impressions of three other poorly preservedmore posterior dorsals. There are no unique features of thesevertebrae that are not better preserved on other specimens.Length and width measurements were not taken.

Sacral Vertebrae—Two poorly preserved sacral vertebraeand ribs are present (Fig. 8A). The anterior rib is slender andangled posteriorly, while the second sacral rib is more robustand is angled anteriorly. What may be a fragment of the iliumlies on top of the vertebral centra.

Caudal Vertebrae—There are ten caudals with transverseprocesses and partial neural spines (Fig. 8A). Unfortunately, itis not possible to determine the presence of pygal vertebrae.

BMNH R 32268 (Fig. 10)

This specimen was found near Burham, Kent, in the samequarry as the holotype.

Rear Limb and Girdle Elements—Portions of the girdleand limb elements are present for the both left and right sidesof the body. The ilium is posteriorly elongate as is common inmost squamates with the exception of mosasaurs and snakes.In mosasaurs the posterior iliac crest and process are lost, andthe anterior, superior iliac process is greatly enlarged (see Rus-sell, 1967); in snakes that still possess pelvic girdle rudiments,the ilium is often anteriorly elongate (pers. obs.).

The sacral ribs are stout, elongate elements with broadlyrounded articular heads. The foot of the second rib is muchmore expanded than the first. The second is angled slightlyanterior, while the first is angled slightly posterior. The articularheads appear to touch each other at the contact with the ilium.

The left femur is badly crushed and the proximal portion ismissing; the element is approximately 1.5 cm long as preserved.The distal end of the right femur is well preserved and verybroad distally with well-developed articular surfaces. The shaftof the right femur is missing.

Dorsal Vertebrae—Nineteen posterior dorsal vertebrae arepreserved. They differ in no visible way from the more anterior


FIGURE 8. Dolichosaurus longicollis (BMNH R49907), Lower Chalk, Southerham Pit, Lewes, Sussex: A, ventral view of complete specimen;B, detail view of pectoral girdle and anterior dorsal vertebrae. Abbreviations: s1, 1st sacral rib and vertebra; s2, 2nd sacral rib and vertebra.Scale bar equals 1 cm.

dorsals preserved on the holotype. The average centrum lengthis 9.5 mm with the vertebrae decreasing in length approachingthe sacrum. The average width of the neural arch is 5.4 mmand the average width across the centrum, posterior to the par-apophysis, is 4.5 mm.

Sacral Vertebrae—There are two sacral vertebrae eachbearing very well developed sacral ribs. The more anterior ribis slightly more slender than the posterior one. The posterior-most rib is expanded distally into a wide and slightly flattenedend. The sacral vertebrae are much shorter than the dorsals(approximate 6.5 mm in length).

Caudal Vertebrae—Four caudal vertebrae are preserved.Transverse processes are present on the most anterior vertebra,but are broken on the other three. Facets for the haemal ribs/

arches are found on the second caudal but not on the first. Thiswould indicate the presence of at least one pygal vertebra inDolichosaurus. Haemal facets are present on the most posteriorportions of long, raised crests that run the length of the ventralsurface of the centrum. Between these crests is a flat shallowtriangle-shaped region. The presence of facets cannot be eval-uated on the two more distal caudals as the ventral surfaces arenot preserved. The two caudals for which lengths are obtainedare longer than the sacrals, but shorter than the dorsals (7.2 mmand 7.8 mm).

BMB 008567 (Fig. 11)

This specimen was collected at the Southerham Pit, nearLewes, Sussex, and was prepared in three blocks, the anterior


FIGURE 9. Dolichosaurus longicollis (BMNH R49907), LowerChalk, Southerham Pit, Lewes, Sussex: detailed line drawing of pectoralgirdle, left forelimb, and anterior dorsal vertebrae. Small arrows indicatethe large and well preserved sternal cartilages. Abbreviations: cl, clav-icle; Co, coracoid; S, scapula; cCc, calcified supracoracoid cartilage;stc, calcified sternal cartilages; ul, ulnare? Scale bar equals 1 cm.

FIGURE 10. Dolichosaurus longicollis (BMNH R32268), LowerChalk, Burham, Kent: A, ventral view of specimen; B, detail view ofpelvic girdle and posterior dorsal vertebrae. Abbreviations: lF, left fe-mur; rF, right femur; ril, right ilium; P, pygal vertebra; s1, 1st sacralrib and vertebrae; s2, 2nd sacral rib and vertebra. Scale bars equal 1cm.

block in dorsal view, the middle block in ventral view, and theposterior block in dorsal view.

Rear Limb and Girdle Elements—Portions of both theright and left girdle and limb elements are present (Fig. 11B,C). There are remnants of seven phalangeal elements, the distaland proximal portions of the femur, tibia, fibula and left iliumare present on the block 2. Block 3 has a very complete femur(approximately, 1.5 cm in length) that has a definite bend inthe mid-shaft of the bone. This is not an artifact of either pres-ervation or preparation, but is instead a real feature of the bone.It may represent a paleopathology. There is also a fragment ofthe distal epiphysis, and a number of unidentifiable fragmentsof the phalanges and metatarsals.

Dorsal Vertebrae—There are thirty-two dorsal vertebraepreserved (Fig. 11A, B). These vertebrae differ in no visibleway from the more anterior dorsals preserved on either theholotype or the other referred specimens. The individual ver-tebrae narrow immediately behind the parapophyses and do notbecome wider or narrower at the condyle. The condyle is cir-cular and the cotyle slightly elliptical. Measured along the mid-line of the centrum, the length appears to average around 8.1mm. The vertebrae become shorter towards the sacrum. Theaverage width of the neural arch is 4.9 mm and the averagewidth across the centrum, posterior to the parapophysis is 3.6mm. The width across the parapophyses is 8.7 mm.

Sacral Vertebrae—There are two sacral vertebrae eachbearing very well developed sacral ribs (Fig. 11B). The moreanterior rib is slightly more slender than the posterior one. Theposterior rib is expanded distally into a wide and slightly flat-tened end. Near the base of the sacral rib and along the posteriorborder there is a small recess and groove for the lymphaticsystem of the pelvic region. The sacral vertebrae are muchshorter than the dorsals (approximately 6.5 mm in length).

Caudal Vertebrae (Fig. 11B, C)—There are at least thirty-two articulated caudal vertebrae preserved on the second andthird block; more may be present but are badly weathered (Fig.11B, C). The first caudal vertebra also bears a stout pair oftransverse processes and there are no caudal peduncles visible;this vertebra is considered to be a pygal. It is possible thatseveral vertebrae were lost between the second and third blockwhen the specimen was excavated. Broad, flat, transverse pro-cesses, preserved on all caudals, become gradually shorter andslimmer posteriorly. Facets for the haemal ribs/arches havebeen destroyed by weathering. Small facets may be present onthe first but this is difficult to ascertain as the posterior regionis broken. The caudals are longest just behind the sacrum (6.8mm) and shortest at the most posterior extent of the tail (4.6mm).

PHYLOGENETIC ANALYSIS

Results

Cladistic analysis of the data matrix (Appendix 2) resultedin 27 most parsimonious cladograms (122 steps) with a Con-sistency Index (C.I.) of 0. 648, a Homoplasy Index (H.I.) of0.352, and a Retention Index (R.I.) of 0.669.


FIGURE 11. Dolichosaurus longicollis (BMB 008567), Lower Chalk, Burham, Kent. Most complete postcranial skeleton currently known, in 3blocks. A, dorsal view of anterior dorsal vertebrae and ribs; B, ventral view of mid to posterior dorsal vertebrae and pelvic girdle; C, dorsal viewof caudal vertebrae and elements of right hindlimb. Abbreviations: rF, right femur; lil, left ilium; P, pygal vertebra; s1, 1st sacral rib andvertebrae; s2, 2nd sacral rib and vertebra. Scale bar equals 1 cm.

The topology of the Strict Consensus Tree (Fig. 13A) sup-ports the monophyly of the Mosasauridae (Russell, 1967; Bell,1997; Caldwell, 1996): (Halisaurus (Ectenosaurus (Clidastes,Mosasaurus)), Tylosaurus, Platecarpus)). The Aigialosauridaeis also reconstructed as a distinct clade, though their ingrouprelationships are unresolved. No support is found for sugges-tions of aigialosaur paraphyly (Bell, 1997) as the clade is wellsupported and differentiated in all cladograms relative to theMosasauridae, Dolichosaurus, and coniasaurs. Dolichosaurusand both coniasaurs form an unresolved five-branch polytomywith the Mosasauridae and Aigialosauridae.

A Majority Rule Consensus Tree (Fig. 13B) shows that infifty-six percent of the trees, aigialosaurs and mosasaurs arefound to form a clade distinct from Dolichosaurus and coni-asaurs (Fig. 13B); this clade is conventionally referred to as theMosasauroidea. In thirty-three percent of the trees Aigialosau-

rus dalmaticus and Aigialosaurus [5Opetiosaurus] buccichiform a monophyletic genus; as was noted by Caldwell et al.(1995), generic characters separating A. dalmaticus from A.[5O.] buccichi cannot be identified although the species areidentifiable. Dolichosaurus is reconstructed as the most basalmember of the clade, with coniasaurs ([Coniasaurus crassidens,Coniasaurus gracilodens] in forty-four percent of the trees) inthe sistergroup position to the Mosasauroidea (thirty-three per-cent of the trees).

DISCUSSION

Phylogeny and Origins

Caldwell (1999b), in an analysis of higher-level squamatephylogeny, found the clade coniasaurs1Mosasauroidea (Aigi-alosauridae and Mosasauridae) in all eighteen shortest clado-


FIGURE 12. Left lateral view of 11th cervical vertebrae Dolichosau-rus longicollis (BMNH R49002): Abbreviations: hy, hypapophysis; p,peduncle; poz, postzygapophysis; prz, prezygapophysis; zg, zygantrum.

FIGURE 13. Concensus Trees of 27 cladograms (122 steps) showing ingroup relationships of 12 fossil squamate taxa using 66 osteologicalcharacters. A, Strict Consensus Tree. B, Majority Rule Consensus Tree. Numbers on branches indicate the percentage of most parsimoniuscladograms showing that particular branching pattern.

grams, and snakes to be the sistergroup of that clade (Serpentes(Coniasaurus, Mosasauroidea)) in twelve of those eighteenshortest cladograms. A similar treatment of terminal taxa andclades was also presented by Caldwell and Lee (1997) and Leeand Caldwell (1998) in their examination of the relationshipsof Pachyrhachis problematicus Haas, 1979, where they referredto the snake-mosasauroid clade as the Pythonomorpha (see alsoLee, 1997, 1998).

Despite the recent efforts devoted to including fossils in anal-yses of squamate phylogeny, many theoretical problems remain

concerning the interrelationships of mosasauroids, coniasaurs,and snakes. For example, Lee (1998), in contrast to Caldwell(1999b) and Caldwell et al. (1995), finds mosasauroids, and byextension coniasaurs and dolichosaurs (see Caldwell, 1999a;Caldwell and Cooper, 1999, and this study), to be nested withinVaranoidea. Such a phylogenetic position for dolichosaurs andother mosasauroids indicates a very different evolutionary his-tory for basal varanoids and their sistertaxa than is currentlyrecognized from the fossil record (see Norell and Gao, 1997).The degree of difference between Caldwell and Lee’s hypoth-eses requires further examination of other dolichosaurs (see Leeand Caldwell, 2000), other fossil varanoids, basal anguids, andthe varanoid or non-varanoid characteristics of snakes.

A far more contentious issue is the snakes-as-pythonomorphshypothesis supported by Caldwell (1999b), Lee (1997, 1998),and Lee and Caldwell (1998, 2000). Recently, Zaher (1998),and Zaher and Rieppel (1999) have argued that Pachyrhachisis a derived macrostomatan snake rather than a primitive snake(contra Caldwell and Lee, 1997). Zaher (1998) and Zaher andRieppel (1999) base these assertions on their reinterpretation ofthe skeletal anatomy of Pachyrhachis. Zaher and Rieppel(1999) support the idea that snakes are the sistergroup to am-phisbaenids and dibamids, not mosasauroids and coniasaurs.Zaher (1998) has been addressed in detail by Caldwell (2000)and will not be examined further here. However, it is importantto examine the data and ideas put forward by Zaher and Rieppel(1999) that are relevant to the phylogenetic relationships andanatomy of Dolichosaurus.

As was shown by Caldwell (2000) regarding Zaher (1998),the salient problem with Zaher and Rieppel’s (1999) study isthat they provide no evidence (data matrices, characters, etc.)in support of their phylogenetic hypothesis. Zaher and Rieppel(1999:834; fig. 2) list synapomorphies for their snake clades,but oddly enough the synapomorphy list contains none of the‘undoubted’ macrostomatan characters of Pachyrhachis thatthey stated were misinterpretations by Caldwell and Lee (1997),


Lee and Caldwell (1998), and Lee (1998). Does this mean thesecharacters were not found as true synapomorphies in a cladisticanalysis (for which there is no evidence of having been under-taken)? Or perhaps the undoubted macrostomatan features ofPachyrhachis, which were described by Lee and Caldwell(1998:1550) were found homoplastic, were not included in theunpublished data matrices? If not, and why not? In either caseit is impossible to recover this information as the data is absent.As with Zaher (1998), until character lists, state assignments,and data matrices are available such inductively derived phy-logenetic conclusions must be deemed as primary hypotheticalstatements, and not deductive statements worthy of tests of fal-sification (see Kluge, 1997).

In relation to testable statements of phylogeny, Dolichosau-rus longicollis presents a suite of morphological characteristicsthat should prove informative when analyzed for more globalcharacter congruence with snakes, mosasaurs, and other squa-mates (see Lee and Caldwell, 2000). However, the current pau-city of good Dolichosaurus fossils that possess both skulls andjaws, is frustrating when compared to the skulls and jawsknown for two species of Coniasaurus. It is equally problematicthat the reverse condition exists for the postcranium: absenceof postcrania for Coniasaurus versus complete postcrania forDolichosaurus. The possibility remains that these taxa are con-generic. Synonomizing the two genera would reduce the threespecies to one genus and two species. Dolichosaurus/Coni-asaurus would then be a small, very elongate, limb reduced,marine squamate, of Cenomanian age, with a head very similarto an aigialosaur, an intramandibular joint synapomorphic withmosasauroids and snakes, and an elongate neck following asimilar pattern to that found in snakes (neck elongation is dis-cussed below). The polytomy (Fig. 13) currently obscuring re-constructions of dolichosaur-coniasaur relationships would dis-appear. The remaining question is whether or not such a taxonwould be reconstructed as a basal mosasauroid, or in a sister-group position to a clade containing crown-group snakes (seeLee and Caldwell, 2000).

Neck Elongation and Elongate Squamates

The long neck of Dolichosaurus is unique among non-snakesquamates. This aquatic lizard has doubled the number of cer-vical vertebrae from the primitive squamate count of eight, andincreased the number of dorsal vertebrae as well (32 to 38 ifthe association of BMNH R49002 and R32268 is correct asOwen [1850] supposed). The question is whether or not an in-creased number of cervicals and dorsals are informative onbroader problems of squamate phylogeny and evolution.

Elongation of the squamate body is accomplished by chang-ing the length of any or all of the three body regions: neck,trunk, and tail. Each region appears able to independentlychange the number, and to a lesser degree, the size of its re-spective vertebrae in order to effect elongation. The terminol-ogy for axial variation uses a number of landmarks or referencepoints to define the beginning and end of a particular region.

Cervicals are vertebrae anterior to the first vertebra with arib contacting a sternal cartilage (see below for more detail).Dorsals are vertebrae immediately posterior to the last cervicaland anterior to the first sacral. Presacrals are all vertebrae (cerv-icals and dorsals) anterior to the first sacral vertebrae; this termis applied to most squamates, but is most accurate when appliedto those with no sternal attachments but a definite pelvic girdle(e.g., pygopodid geckos). The term precloacals is applied toamphisbaenids and snakes as there are little or no remnants ofthe pectoral girdle, and the sacral vertebrae have been lost; pre-cloacals are all vertebrae anterior to the cloaca. Sacrals are ver-tebrae with transverse processes contacting the ilium. Caudalsare all vertebrae posterior to the last sacral. Mosasaurs have no

sacral vertebrae, but possess a unique pygal vertebra series lo-cated between the last dorsal, and the first caudal. Pygals aredifferentiated from dorsals and caudals by the absence of freeribs and haemal arches.

For typical limbed squamates (typical teiids, varanids, lac-ertids, etc.), the cervical count is 8 to 9, while the dorsal countis from 20 to 36. Aigialosaurids have 7 to 8 cervicals and ap-proximately 20 dorsals; mosasaurs average 7 to 8 cervicals and22 to 35 dorsals (Carroll and deBraga, 1992; Caldwell et al.,1995; Holmes, 1996). The only non-snake squamates with com-parable numbers of presacrals are the limbless dibamids, am-phisbaenids, anguids, pygopodids, and scincids (including fey-linids and acontines). However, none of these groups show anincreased number of cervicals. Instead, these taxa reduce thenumber of cervicals, ranging from four in Dibamus and four tofive in most amphisbaenids, to seven in feylinids (Hoffstetterand Gasc, 1969). For limbless or limb-reduced non-snake squa-mates, the dorsal count increases while the cervical count de-creases; in limb-reduced anguids the dorsal count is 68, whilein amphisbaenids it reaches 160. For snakes, the precloacal andpostaxial counts begins at 120 and reaches 320 (some fossilforms are over 400; for more detail see Hoffstetter and Gasc,1969).

The details of some specific examples are informative onmodes of elongation. Neck elongation in extant Varanus resultsfrom an increase in the relative size of the cervical vertebrae(nearly twice as long as the anterior dorsals), not an increasednumber. In comparison, neck elongation in Dolichosaurus re-sults from the addition of at least 11 vertebrae to the primitivecount of 8, for a total of at least 19. Dibamus has only fourcervicals, produced either by loss in absolute number, or theanterior translocation of the pectoral girdle.

It is clear that Dolichosaurus is unique among limbed squa-mates as none of the latter have a presacral/precloacal countincluding 19 or more cervicals. The current definition of cer-vical vertebrae restricts its usage and identification to all ver-tebrae anterior to the first ‘dorsal’ vertebra, with the first dorsalbeing identified as the first vertebra that articulates with a ster-nal rib (Hoffstetter and Gasc, 1969). Therefore, the presence/absence and position of the sternal rib, and the pectoral girdleof which it is a part, dictates the presence/absence of a neck.For Varanus, dolichosaurs, and other elongate non-snake squa-mates such as Dibamus, it is comparatively easy to recognizethe mode of neck elongation (increased number and/or size ofcervical vertebrae) because they have either a complete or par-tial pectoral girdle. Squamates, in particular snakes, that no lon-ger possess any remnant of the pectoral girdle, are thereforewithout a neck and cervical vertebrae. In short, the neck ceasesto exist because a language is created that abstracts it to thepoint of erasing it. Such definitions are undesirable and prob-lematic as they do not define cervical vertebrae based on theircharacteristics, but rather in terms of their association, or lackof association, with a suite of appendicular elements. Forsnakes, quite apart from comparisons with Dolichosaurus, thissemantic obfuscation makes it impossible to discuss neck evo-lution, or to discover the cladogenic distribution of charactersof the neck and cervical vertebrae with the matching region inputative sistergroups, i.e., with dolichosaurs (this study), or am-phisbaenids (Zaher and Rieppel, 1999). In light of the sister-group relationships of snakes and mosasauroids as hypothesizedby Caldwell (1999b) and Lee (1998), it is important that somestandard of comparison between the ‘necks’ of snakes and Dol-ichosaurus be established.

A neck is a body region that contains a variety of organsystems such as the hyoid apparatus, the esophagus, the trachea,a discrete series of arteries, veins, motor nerves, and so on. Assnakes still possess all of these structures, it can only be con-cluded that snakes have not lost their necks, only their pectoral


girdles and forelimbs. Redefining the neck in terms of featurespossessed by the constituent vertebra resolves this problem, andallows snake evolution to be better understood in the contextof neck elongation and forelimb loss.

Squamate cervical vertebrae have hypapophyses and inter-centra. I recommend here, that their presence, position, mor-phology, and number be reconsidered as informative on thebeginning and end of the cervical series in limbed and limblesssquamates. This position is contrary to Hoffstetter and Gasc(1969) who argue that hypapophyses and their characteristicsare so variable that they are of no utility in characterizing thepresence of a neck and/or cervical region. A survey of hypa-pophyseal variation is therefore necessary to support my posi-tion.

Not counting the atlas-axis vertebrae, hypapophyses are pre-sent on one to three vertebrae in Heloderma, and on seven oreight vertebrae in Varanus (sometimes extending onto the firstdorsal [pers. obs.]). Typical limbed squamates show a count ofsix to seven vertebrae with hypapophyses. According to Hof-stetter and Gasc (1969) Feylinia has nine while Anniella hasten to eleven; Acontias and Dibamus have seven while the limb-less anguids have six. It is interesting to note that counting thenumber of hypapophysis-bearing vertebra shows a much closercorrespondence to the primitive squamate cervical count ofeight, than the corresponding cervical number derived by ref-erence to the pectoral girdle (see above). The question then iswhich is more reliable. The features of cervical vertebrae, orthe position of the pectoral girdle. Let us weigh the evidenceby examination of Dibamus. Dibamus has four cervicals basedon the position of the pectoral girdle, it seven cervicals derivedfrom counting the hypapophyses. If the primitive squamatecount is approximately eight cervicals with hypapophyses, andthe first sternal attachment is on the ninth vertebra, then it canbe hypothesized that primitive dibamids possessed eight cerv-icals, and that along the way lost the expression of the hypa-pophysis on the eighth; the pectoral girdle has been anteriorlytranslocated providing mechanical stability for the neck duringburrowing. Dibamus retains a nearly primitive cervical count,while showing a derived state for pectoral girdle positioning.For an animal that burrows with its head a short neck is afunctional advantage. The same reasoning suggests that the lowhypapophyseal count of scolecophidians and other burrowingsnakes is also derived.

For squamates with long necks, indexed by the presence ofa large number of hypapophyses, the opposite conclusion isreached, and the opposite scenario applies. The number of hy-papophysis-bearing vertebrae increase, but not as a result ofburrowing adaptations. An elongate neck functions differentlythan a short neck. Modern snakes present the ‘elongate neckcondition’, not the shortened, primitive neck of Dibamus. Thenumber of vertebrae with hypapophyses ranges from all pre-cloacal vertebrae to only the first 4 or 5 immediately posteriorto the occipital condyle (Hoffstetter and Gasc, 1969); the prim-itive snake Pachyrhachis has at least 18 (see Lee and Caldwell,1998) vertebrae with hypapophyses, similar to the number inDolichosaurus (contra Zaher and Rieppel’s (1999) recent dis-missal of any notable cervical zonation in Pachyrhachis).

Comparing the number of hypapophysis-bearing vertebrae ofDolichosaurus with Pachyrhachis and other snakes highlightsthe limitations and deficiencies of the current definition of cer-vical vertebrae. If such a definition is used, Pachyrhachis andall other snakes must be coded ‘not applicable’ for Character65 (Appendix 1). As I have argued, it is the definition thatcreates the ‘not applicable’ state assignment, and not the mor-phology. In short, if cervical synapomorphies are present be-tween Dolichosaurus and snakes they are impossible to detectnot because they do not exist, but because of the language ap-plied to the character state distinction.

Under the freedom of a more accurate definition, snakes canbe seen as having dramatically elongated the neck, rather thanhaving lost it. For most tetrapods the neck is relatively mobileand usually capable of a wider range of rotational movementthan other axial regions, particularly if the pre- and postzyga-pophyseal surfaces are horizontal. Horizontal zygapophyses arepresent throughout the column in most modern snakes. It is alsoimportant to note that snakes possess zygosphene and zygantralarticulations in association with horizontal zygapophyseal fac-ets. The only other squamates sharing this character complexare dolichosaurs. The increased number of cervicals and atten-dant hypapophyses pass the test of similarity for snakes andDolichosaurus. For snakes, the unique evolutionary feature isthe cervicalization of a significant proportion of the presacralcolumn thus creating a very long neck. The next step is the testof congruence as outlined by Patterson (1982) (see Lee andCaldwell, 2000).

Dolichosaur Paleobiology

Dolichosaurus was a small marine squamate, likely no morethan 0.5–1.0 meters in total length. As is likely the case forDolichosaurus and other putative dolichosaurs (see Nopcsa,1908, 1923; Caldwell, 1999a; Caldwell and Cooper, 1999),these lizards had long narrow heads, and elongate necks, bod-ies, and tails (Dal Sasso and Pinna, 1997; Dal Sasso and Re-nesto, 1999). As such, Cenomanian dolichosaurs were elongate,limb-reduced marine lizards with small, thin heads, that mayhave served them very well as predators feeding in crevicesand narrow spaces such as might exist in coral reefs and onrocky shores. In this role, dolichosaurs would have occupied aniche similar to that of the earlier nothosaurs (see discussion inCaldwell and Cooper, 1999), to that occupied by extant seasnakes (Greene, 1995), and to the one proposed for the Ceno-manian marine snakes Pachyrhachis (Caldwell and Lee, 1997;Lee and Caldwell, 1998; Scanlon et al., 1999), and Pachyophis(Lee et al., 1999). The degree of limb reduction (pectoral versuspelvic, and overall reduction in limb size), suggests that doli-chosaurs spent much of their time in water, and may have reliedvery heavily on lateral undulations of the body to generate forcefor anguilliform locomotion. Their limbs appear to have beentoo small to have generated any significant motive force eitheron land, or in the water. While of no obvious adaptive advan-tage, the reduction of the forelimb and girdle as compared tothe rearlimb, may be a common feature shared with earlysnakes (Caldwell and Lee, 1997; Lee and Caldwell, 1998).

ACKNOWLEDGMENTS

For assistance while gathering data, I wish to thank S. Chap-man, J. Cooper, A. Currant, J. Evans, and C. Price. For assis-tance in editing the manuscript I wish to thank A. Nicholsonand R. Holmes. I also thank S. Evans and R. Carroll for ex-tremely helpful suggestions and for declining anonymity in thereview process. In particular I wish to acknowledge the pho-tographs taken, and prints provided to me, by the PhotographicsDepartment at the Natural History Museum (London). This re-search was supported by a Natural Sciences and EngineeringResearch Council of Canada (NSERC) Postdoctoral Fellowshipand by a Canandian Museum of Nature, Research AdvisoryCouncil Grant (No. RS34) to the author.

LITERATURE CITED

Bahl, K. N. 1938. Skull of Varanus monitor (Linn.). Records of theIndian Museum, Journal of Indian Zoology 39:133–174.

Bell, G. L. 1997. A phylogenetic revision of North American and Adri-atic Mosasauroidea; pp. 293–332 in J. M. Callaway and E. L. Nich-olls (eds.), Ancient Marine Reptiles, Academic Press, San Diego.


Bellairs, A. d’A., and G. Underwood. 1951 The origin of snakes. Bi-ological Reviews 26:193–237.

Caldwell, M. W. 1996. Ontogeny and phylogeny of the mesopodialskeleton in mosasauroid reptiles. Zoological Journal of the LinneanSociety 116:407–436.

1999a. Description and phylogenetic relationships of a new spe-cies of Coniasaurus Owen, 1850 (Squamata), Journal of VertebratePaleontology 19:438–455.

1999b. Squamate phylogeny and the relationships of snakes andmosasauroids. Zoological Journal of the Linnean Society 125:115–147.

2000. On the phylogenetic relationships of Pachyrhachis withinsnakes: A response to Zaher (1998). Journal of Vertebrate Pale-ontology 20:181–184.

, R. L. Carroll, and H. Kaiser. 1995. The pectoral girdle andforelimb of Carsosaurus marchesetti (Aigialosauridae), with a pre-liminary phylogenetic analysis of mosasauroids and varanoids.Journal of Vertebrate Paleontology 15:516–531.

, and J. Cooper. 1999. Redescription, palaebiogeography, andpalaeoecology of Coniasaurus crassidens Owen, 1850 (Squamata)from the English Chalk (Cretaceous; Cenomanian). ZoologicalJournal of the Linnean Society 127:423–452.

, and M. S. Y Lee. 1997. A snake with legs from the marineCretaceous of the Middle East. Nature 386:705–709.

Carroll, R. L., and M. Debraga. 1992. Aigialosaurs: Mid-Cretaceousvaranoid lizards. Journal of Vertebrate Paleontology 12:66–86.

Cope, E. D. 1869. On the reptilian order Pythonomorpha and Strepto-sauria Proceedings of the Boston Society of Natural History 12:250–261.

Dal Sasso, C., and G. Pinna. 1997. Aphanizocnemus libanensis n. gen.n. sp., a new dolichosaur (Reptilia, Varanoidea) from the UpperCretaceous of Lebanon. Paleontologia Lombarda della Societa It-aliana di Scienze Naturali e del Museo Civico di Storia Naturaledi Milano, new series, 7:3–31.

, and S. Renesto. 1999. Aquatic varanoid reptiles from the Cen-omanian (upper Cretaceous) lithographic limestones of Lebanon.Rivista della Museo Civico di Scienze Naturale, ‘‘E. Caffi,’’ Ber-gamo 20:63–69.

Debraga, M., and R. L. Carroll. 1993. The origin of mosasaurs as amodel of macroevolutionary patterns and processes. EvolutionaryBiology 27:245–322.

Estes, R., K. De Queiroz, and J. Gauthier. 1988. Phylogenetic relation-ships within Squamata; pp. 119–281 in R. Estes and G. Pregill(eds.), Phylogenetic Relationships of the Lizard Families. StanfordUniversity Press, Stanford.

Gervais, P. 1852. Zoologie et Paleontologie generale.Greene, H. W. 1995. Snakes. The Evolution of Mystery in Nature. Uni-

versity of California Press, Berkeley.Haas, G. 1979. On a new snakelike reptile from the Lower Cenomanian

of Ein Jabrud, near Jerusalem. Bulletin de la Museum Nationaled’Histoire Naturelle, Paris, Ser. 4, 1: 51–64.

1980a. Pachyrhachis problematicus Haas, snakelike reptilefrom the Lower Cenomanian: ventral view of the skull. Bulletin dela Museum Nationale d’Histoire Naturelle, Paris, Ser. 4, 2: 87–104.

1980b. Remarks on a new ophiomorph reptile from the LowerCenomanian of Ein Jabrud, Israel, pp. 177–192, in L. L. Jacobs(ed.), Aspects of Vertebrate History, in Honor of E. H. Colbert.Museum of Northern Arizona Press, Flagstaff.

Hancock, J. M., W. J. Kennedy, and W. A. Cobban. 1993. A correlationof the Upper Albian to Basal Coniacian sequences of NorthwestEurope, Texas and the United States Western Interior; pp. 453–476in W. G. E. Caldwell and E. G. Kauffman (eds.), Evolution of theWestern Interior Basin. Geological Association of Canada, Ottawa,Volume 39.

Hoffstetter, R., and J.-P. Gasc. 1969. Chapter 5. Vertebrae and ribs ofmodern reptiles; pp. 201–310 in C. Gans (ed.), Biology of the Rep-tilia, Vol. 1. Academic Press, New York.

Holmes, R. 1996. Plioplatecarpus primaevus (Lacertilia, Mosasauridae)from the Bearpaw Formation (Campanian, Upper Cretaceous) ofthe Western Interior Seaway (North America). Journal of VertebratePaleontology 16: 673–687.

Kluge, A. 1997. Testability and the refutation and corroboration of cla-distic hypotheses. Cladistics 13:81–96.

Kornhuber, A. 1893. Carsosaurus marchesetti, ein neuer fossiler Lac-

ertilier aus den Kreideschichten des Karstes bei Komen. Abhan-dlungen der geologischen Reichsanstalt Wien 17:1–15.

1901. Opetiosaurus bucchichi, eine neue fossile Eidechse ausder unteren Kreide von Lesina in Dalamtien. Abhandlungen dergeologischen Reichsanstalt Wien 17:1–24.

Kramberger, K. G. 1892. Aigialosaurus, eine neue Eidechse aus denKreideschiefern der Insel Lesina mit Rucksicht auf die bereits bes-chriebenen Lacertiden von Comen und Lesina. Glasnik huvatskoganaravolosovnoga derstva (Societas historico-matulis croatica) u Za-grebu 7:74–106.

Lee, M. S. Y. 1997. The phylogeny of varanoid lizards and the affinitiesof snakes. Philosophical Transactions of the Royal Society, Lon-don: Biological Sciences 352:53–91.

1998. Convergent evolution and character correlation in bur-rowing reptiles: towards a resolution of squamate phylogeny. Bi-ological Journal of the Linnean Society 65:369–453.

, G. L. Bell, and M. W. Caldwell. 1999. The origins of snakefeeding. Nature 400:655–659.

, and M. W. Caldwell. 1998. Anatomy and relationships of Pa-chyrhachis, a primitive snake with hindlimbs. Philosophical Trans-actions of the Royal Society, London: Biological Sciences 353:1521–1552.

, and M. W. Caldwell. 2000. Adriosaurus and the affinities ofmosasaurs, dolichosaurs, and snakes. Journal of Paleontology 74:915–937.

, M. W. Caldwell, and J. S. Scanlon. 1999. A second primitivemarine snake: Pachyophis woodwardi Nopcsa. Journal of Zoology248: 509–520.

Milner, A. 1987. 12. Reptiles; pp. 266–280 in A. B. Smith (ed.), Fossilsof the Chalk, Palaeontological Association Field Guides to Fossils:Number 2. The Palaeontological Association. University PrintingHouse, Oxford.

Mortimore, R. N. 1986. Stratigraphy of the Upper Cretaceous WhiteChalk of Sussex. Proceedings of the Geologists Association 97:97–139.

1987. Upper Cretaceous Chalk in the North and South Downs,England: A correlation. Proceedings of the Geologists Association98:77–86.

Nopcsa, F. 1908. Zur kenntnis der fossilen Eidechsen. Beitrage zur Pa-laontologie und Geologie Osterreich-Ungarns und des Orients 21:33–62.

1923. Eidolosaurus und Pachyophis. Zwei neue Neocomen-Reptilien. Palaeontographica 65:97–154.

Norell, M.A., and K. Gao. 1997. Braincase and phylogenetic relation-ships of Estesia mongoliensis from the Late Cretaceous of the GobiDesert and the recognition of a new clade of Lizards. AmericanMuseum Novitates 3211:1–25.

Patterson, C. 1982. Morphological characters and homology; pp. 21–74in K. A. Joysey and A. E. Friday (eds.), Problems of PhylogeneticReconstruction, Vol. 21, Academic Press, New York.

Obradovich, J. D. 1993. A Cretaceous Time Scale; pp. 379–396 in W.G. E. Caldwell and E. G. Kauffman (eds.), Evolution of the West-ern Interior Basin. Geological Association of Canada, Ottawa, Vol-ume 39.

Oppel, M. 1811. Die Ordnungen, Familien, und Gattungen der Repti-lien. Munchen.

Owen, E. 1987. 1. Introduction; pp. 9–14 in A. B. Smith (ed.), Fossilsof the Chalk, Palaeontological Association Field Guides to Fossils:Number 2. The Palaeontological Association. University PrintingHouse, Oxford.

Owen, R. 1850. Description of the Fossil Reptiles of the Chalk For-mation; pp. 378–404 in F. Dixon (ed.), The Geology and Fossilsof the Tertiary and Cretaceous Formations of Sussex. Longman,Brown, Green and Longman, London.

Rieppel, O. 1979. The evolution of the basicranium in the Henophidia(Reptilia: Serpentes). Zoological Journal of the Linnean Society 66:411–431.

Robinson, N. D. 1986. Lithostratigraphy of the Chalk Group of theNorth Downs, southeast England. Proceedings of the GeologistsAssociation 97:141–170.

1987. Upper Cretaceous Chalk in the North and South Downs,England: a reply. Proceedings of the Geologists Association 98:87–93.

Russell, D. A. 1967. Systematics and morphology of American mosa-


saurs. Peabody Museum of Natural History, Yale University Bul-letin 23:1–241.

Scanlon, J. D., M. S. Y. Lee, M. W. Caldwell, R. Shine. 1999. Thepalaeoecology of the primitive snake Pachyrhachis. Historical Bi-ology 13:127–152.

Swofford, D. L. 1993. PAUP: Phylogenetic Analysis Using Parsimony,Version 3.1.1. Laboratory of Moelcular Systematics, SmithsonianInstitution, Washington D.C.

Wu, X. -C., D. B. Brinkman, and A. P. Russell. 1996. Sineoamphisbaenahexatabularis, an amphisbaenian (Diapsida: Squamata) from theUpper Cretaceous redbeds at Bayan Mandahu (Inner Mongolia,People’s Republic of China), and comments on the phylogeneticrelationships of the Amphisbaenia. Canadian Journal of Earth Sci-ences 33:541–577.

Zaher, H. 1998. The phylogenetic position of Pachyrhachis withinsnakes (Squamata, Serpentes). Journal of Vertebrate Paleontology18:1–3.

, and O. Rieppel. 1999. The phylogenetic relationships of Pa-chyrhachis problematicus, and the evolution of limblessness insnakes (Lepidosauria, Squamata). Comptes rendus de l’academiedes sciences, Sciences de la terre et des planetes 329:831–837.

Received 27 August 1999; accepted 7 June 2000.

APPENDIX 1

Character and character state descriptions.

1. Bony predental rostrum on premaxilla: short and obtuse (0); dis-tinctly protruding (1).

2. Size of premaxillary foramen on rostrum: small (0); large (1).3. Dorsal keel on internarial bar of premaxilla: absent (0); present (1).4. Nasal bones: present (0); absent (1).5. Internarial process of frontal: not constricted (0); very constricted

(1).6. Frontal: broad and short (0); long and narrow (1).7. Frontal forms part of margin of nares: not on margin (0); forms

part of margin (1).8. Frontal with sagittal crest: absent (0); low (1); high, and well de-

veloped (2).9. Shape of frontal ala: narrow and pointed (0); broad, rounded (1).

10. Fronto-parietal suture: interlocking sutures (0); overlapping flanges,no sutures (1).

11. Parietal, dorsal surface: short (0); elongate (1).12. Parietal table: rectangular to trapezoidal, sides convergent (0); tri-

angular, straight sides contact anterior to suspensorial rami (1).13. Parietal foramen size: small (0); large (1).14. Parietal foramen position: near to center of parietal table (0); close

to suture (1); touching suture (2); crosses suture including frontal(3).

15. Posterior shelf of parietal: distinct shelf, projects between suspen-sorial rami (0); shelf absent (1).

16. Parietal suspensorial ramus, greatest width: vertical or oblique (0);horizontal (1).

17. Prefrontal, suborbital process: absent or very small (0); large over-hanging wing (1).

18. Prefrontal contact with postorbital frontal: no contact (0); contact(1).

19. Postorbitofrontal, transverse dorsal ridge: absent (0); present (1).20. Maxillary tooth count: 20 to 24 (0); 17–19 (1); 15–16 (2); 12–14

(3).21. Posterior terminus of maxillo-premax suture: with first to fourth

maxillary tooth (0); between fourth and ninth tooth (1); even withor posterior to ninth tooth (2).

22. Ascending process of maxilla: recurved wing of maxilla dorsolat-erally overlaps anterior end of prefrontal (0); process absent (1).

23. Posteroventral process of jugal: absent (0); present (1).

24. Pterygoid tooth row: teeth arise from main shaft of pterygoid (0);teeth arise from thin pronounced ridge (1).

25. Length of quadrate stapedial process: short (0); moderate length(1); long (2).

26. Constriction of quadrate stapedial process: distinct (0); none (1).27. Fusion of quadrate stapedial process to ventral process: absent (0);

present (1).28. Quadrate tympanic rim size: large, almost as high as quadrate (0);

smaller, 50–65% of quadrate height (1).29. Quadrate anteroventral condyle modification: no upward deflection

of anterior edge of condyle (0); distinct deflection (1).30. Pterygoid process of basisphenoid: process relatively narrow with

articular surface facing mostly anterolaterally (0); process thinner,more fan-shaped with posterior extension of articular surface (1).

31. Length of basioccipital tuber: short (0); long (1).32. Dentary tooth number: .19 (0); 18–15 (1); 14–13 (2); ,12 (3).33. Sub-dental shelf of dentary: shelf low (0); shelf raised to median

height (1); equal in height to lateral wall (2).34. Coronoid shape: slight dorsal curvature, posterior wing not wide

fan-shape (0); very concave above, posterior wing expanded (1).35. Coronoid medial wing: does not reach angular (0); contacts angular

(1).36. Surangular-coronoid buttress: low and thick (0); high and thin (1).37. Surangular-articular suture position: behind condyle (0); middle of

glenoid (1).38. Foramina on lateral aspect of retroarticular process: none (0); one

to three (1).39. Tooth surface, medial aspect: finely striate (0); not striated (1).40. Tooth facets: absent (0); present (1).41. Tooth carinae: weak (0); strong and elevated (1).42. Atlas neural arch: notch in anterior order (0); no notch (1).43. Atlas synapophysis: extremely reduced (0); large and elongate (1).44. Condyle inclination, dorsal vertebrae: inclined (0); vertical (1).45. Condyle shape: depressed (0); slightly depressed (1); rounded (2).46. Number of sacral vertebrae: two (0); one (1).47. Caudal neural spines: uniformly short (0); several dorsally elongate

mid-tail region (1).48. Haemal arch articulations: unfused (0); fused (1).49. Scapulocoracoid size: bone approximately equal in proximo-distal

length (0); scapula about half length of coracoid (1).50. Scapula width: no anteroposterior widening (0); distinct fan-shaped

expansion (1); extreme widening (2).51. Scapula posterior emargination: gently concave (0); deeply concave

(1).52. Scapulocoracoid fusion: bones fused (0); not fused (1).53. Humerus length relative to distal width: elongate, 3 to 4 (0); short-

ened, 1.5 to 2 (1); length and width equal (2); distal width greater(3).

54. Humerus postglenoid process: absent or small (0); distinctly en-larged (1).

55. Humerus deltopectoral crest: single ridge (0); two separate insertionareas (1).

56. Humerus pectoral crest: located anteriorly (0); located medially (1).57. Humerus entepicondyle: absent (0); present as a prominence (1).58. Radius shape: radius not expanded (0); slightly expanded (1);

broadly expanded (2).59. Ulna contact with centrale: excluded by broad ulnare (0); contacts

centrale (1).60. Radiale size: large and broad (0); reduced or absent (1).61. Carpal reduction: six or more (0); five or less (1).62. Pisiform: present (0); absent (1).63. Metacarpal I expansion: spindle shaped, elongate (0); broadly ex-

panded (1).64. Tooth crown inflation: not inflated (0); crown moderate to thick

inflation [bulbous] (1).65. Number of cervical vertebrae: 7–9 (0); 10–20 (1).66. Zygosphenes and zygantra: where present, consistently expressed

throughout length of axial skeleton (0); present, but show reductionfrom caudal region towards head (1).


APPENDIX 2

Character matrix with character state distributions.

Taxon

Character

5 10 15 20 25 30 35

outgroupA. dalmaticusA. buccichiCarsosaurusHalisaurusClidastesTylosaurusMosasaurusEctenosaurusPlatecarpusConiasaurus crassidensConiasaurus gracilodensDolichosaurus longicollis

00000??0100???0?????010011001121111111111001100011?????????1?????

110001110011100?????111001011101211001111011101011?????11100?????

000001?10110?0??????0010010011100111001111100??101?????0??0??????

000000000?0?0???????000?01100210113111031001110013????1??0?2?????

00000?00?20?0?2????220?12100111110110?101110100102?0???00?0??????

00100?0000?00??000??0110?0011?101110010?011?11001????????????????

000000?00??0000???????1101121103200122110221003200?00?0????????00

40 45 50 55 60 65

outgroupA. dalmaticusA. buccichiCarsosaurusHalisaurusClidastesTylosaurusMosasaurusEctenosaurusPlatecarpusConiasaurus crassidensConiasaurus gracilodensDolichosaurus longicollis

0001001??00??0000???0101010011000011001100101001010??00???000?0??

000001?0??1??0??1?010?10010112100?1101121??1?110110??010??01???01

000000??00000000?00011111?1101???111?101???01???01????????0000000

000000?00?0100?01000121010221112101?2???1221?12201?????01???01000

00000?000000000000000010101210101????210?12011?201??????????00000

000010000000000000001100000100???000?1000100011000???1????1???0?1

01111111110?0

Caldwell, 2000b

Documents

number of specimens

elongate neck of

restudied owens specimens

aigialosaur taxa

mosasaur taxa

type specimen amaxilla

pelvic girdle

several new specimens