Osteology of Leiopelma (Amphibia: Leiopelmatidae) and descriptions of three new subfossil Leiopelma species

Transcription

1 Journal of the Royal Society of New Zealand ISSN: (Print) (Online) Journal homepage: Osteology of Leiopelma (Amphibia: Leiopelmatidae) and descriptions of three new subfossil Leiopelma species Trevor H. Worthy To cite this article: Trevor H. Worthy (1987) Osteology of Leiopelma (Amphibia: Leiopelmatidae) and descriptions of three new subfossil Leiopelma species, Journal of the Royal Society of New Zealand, 17:3, , DOI: / To link to this article: Published online: 11 Dec Submit your article to this journal Article views: 629 Citing articles: 40 View citing articles Full Terms & Conditions of access and use can be found at

2 (c"' journal o/ihl' Ho)'al Sonel), of Xeli' Lealand" Volume 17,,",'umber}, ]W!7, /!/!" 2()]-:!.'j] Osteology of Leiopelma (Amphibia: Leiopelmatidae) and descriptions of three new subfossil Leiopelma species. Trevor H. Worthy* Three new species of leiopelmatid frog are described from New Zealand cave deposits: Leiopelma waitomoensis n,sp" an offshoot of the L. arcizeyill. izamiltoni lineage, and Leiopelma markizami n,sp, and Leiopelma auroraensis n,sp" which arc more closely related to L. izocizstetteri, Thc osteology of the extant species L.!wehr/etteri, L. izamlltoni and L. arcizeyi is described in greater detail than previously, The data do not support the contention of Stephenson (1960), that L. arcizeyi is neotenic in relation to L. izamiltoni, This idea was based only on the relative degree of ossification observed in these species; in fact both exhibit equal degrees of ossification, Detailed osteological studies of the Leiopelmatidae support the distinction of that family from the Ascaphidac, Kt,)'u)orril. Lciopclill<t, Leiopelmatiri(1(" O\{cu{ogy, nt'w )pecies, SU~f(),\sil, fl%rew',,v('lt' cula"r!, A~caphus truci. INTRODUCTION The leiopelmatid frogs are part of the "archaic" fauna of New Zealand, which survive here largely because of the unusual geological history and isolated geographical position of these islands, Other archaic elements include the ratite birds and the tuatara (Sphenodon punclalus), The lciopelmatid frogs were first discovered in the nineteenth century, and include three extant species; Leiopelma hochstetteri Fitzinger 1861, L arrheyi Turbott 1942, and L hamiltoni McCulloch 1919, All were recognised as primitive discoglossid h'ogs by both Fitzinger (1861) and McCulloch (1919). Noble (1924) considered that, together with the North American toad Ascaphus, these frogs were the most primitivt' known, and includt'd them all in the f~unily "Liopelmidae", although not everyone agrees that Leiopelrna and Ascaphus are that closely related (De Villiers, 1934a; De Vos, 19:)8a; Stephenson, 1960; Griffiths, 1963; Kluge and Farris, 1969; Green et al. 1980). Rt'ct'ntly Green et at. (1980) justified the separation of Ascaphus and Leiopelma into separate familit's on the basis of great karyotypic dissimilarity, long geographic isolation, unique life history and scarcity of shared derived characters, The distributions of the three extant Leiopelrna species have bet'n rt'duced during human occupation of New Zealand (Bull and Whitaker, 1975; Whitaker, 1978; Bell, 1982, 1985; Bell el ai., 1985), a parallel situation to that documented for New Zealand lizards (Whitaker, 1973, 197B; Crook, 1975; Towns el ai., 1985). L. hamiltom is restricted to Maud and Stephens Islands in the Marlborough Sounds; L archeyi is restricted to the Corolllandel Range; and L. hochstetteri is known only from isolated localities in tht' northern half of the North Island (Bell, 1982, 1985; Newman, 1980). There have been reductions in numbers of species too; subfossil remains show that there wt're formerly several more species ofleiopelmatids in New Zt'aland (Bull and Whitaker, 1975; Millcncr, 1981; Bell, 1985; Bell el at. 1985), Since the 1850's tht' main interest in subfossil deposits in New Zealand has been directed toward birds; smaller vertebrate remains have been overlooked or unrecognised. The earliest collection of subfossil It'iopelmatids still accounted for was made at Coonoor in 1914 (NMNZ Unreg.). Additional material has been deposited in New Zealand must'ums in greater quantity since the 1950's. Zoology Department. Victoria University of Wellington.

3 202 Journal oj the Royal Society of New Zealand, Volume 17, 1987 In G. S. Markham identified, for the first time, Leiopelma bones in the National Museum's collection. He separated the material into two unnamed species, which correspond with L. markhami n.sp. and L. waitomoensis n.sp. in this study. Records of subfossil frog bones, (including Patarau, Martinborough, Coonoor, Patoka, Hukanui and Waitomo), and the possible former existence of another, larger, extinct species, were first published by Bull and Whitaker (1975). They provided a distribution map based primarily on the National Museum collection and the records of P. M. Johns, G. S. Markham and C. J. Templer. Although the taxonomy of the sub fossil frogs has remained largely uninvestigated, new distribution records of subfossil deposits were included in Cody, 1979; Millener, 1981, 1984; Millener and Templer, 1982; and Worthy, 1981, 1984, 1985ab, Millener (1981) documented 14 King Country sites in which frogs were found and recognised that two or more species were represented. Worthy (1984) demonstrated that some unidentified species of Leiopelmatidae were prescnt in the King Country during the late Holocene. Bell et az. (1985) compiled an updated distribution of subfossil Leiopelma, and reported the tentative identification of three extinct species in the subfossil record. This study describes the bones of the three living and three presumably extinct species of Leiopelma from the subfossil fauna, and compares them with Ascaphus truei. De Vos (1938a) and Ritland (1955) have described and compared various bones of these two genera, but they were interested in the generic relationships between the "leiopelmatids" and other Anura, not interspecific differences within the Leiopelmatidae. Stephenson (1951, 1960) compared the extant species of Leiopelma, but not in sufficient detail to allow identification of disarticulated bones, such as are present in subjossil deposits. New Zealand Subfossil Deposits In a survey of the subf()ssil avifauna from dune, swamp, and cave deposits in the North Island, Millener (1981) listed all vertebrates found, although birds dominated in all types of deposit. Frog bones were recorded form 8.6 % of caves investigated by Millener (1981) but were absent from dune and swamp deposits. The study of subfossil frogs is therefore effectively confined to karst regions, i.e. those landscapes developed on limestones and marbles, characterized by underground drainage and the development of caves, potholes, grikes etc. Limestone is widely distributed in New Zealand (Fig. 1), but the best areas are (a) in the King Country centred on Waitomo; (b) east of the central ranges in the North Island, and (c) in northwest Nelson. The limestones east of the central ranges in the North Island cover a wide area, but generally do not form well developed karst landforms, because the beds are thin and the lime content is low. Hence Waitomo and northwest Nelson are the major sources of cave subfossils. Rare subfossils from small blocks of limestone in other areas such as Northland, Hawkes Bay, and Fiordland are of great distributional interest. To illustrate the general rarity of subfossil frog remains, Fig. 2 compares the known distribution of avian subfossils in the North Island (Millener, 1981) with that of the subfossil frogs recorded in this study. Numerous archaeological sites in New Zealand contain remains of animals collected by the human inhabitants of the original settlements, and sometimes also of animals that cohabited the site. Amongst these are occasional records of lizards, e. g. Gill (1985), but until now none of frogs. Recently, a pair of frog humeri, L. markhami n.sp. were found at a site on Lee Island in Lake Te Anau. The site is dated at about 200 years BP (Atholl Anderson, pers. comm. 1985); hence this presumably extinct species was at least temporarilly sympatric with man. Tbe Study Areas The study areas, all in karst country, were grouped into five regions (Fig. 1). Although there are many areas of karst in the eastern South Island, none has produced any subfossil frog remains. This is probably not because there are no suitable sites for preservation, as significant numbers of lizard bones have been found in this area. In the other regions, sites suitable for lizards are also suitable for frogs. However frog bones may be absent for other reasons, perhaps related to the low rainfall of the eastern South Island.

4 Worthy - Osteology oj Leiopelma 203 BAY 4 PATARAU COONOOR MARTINBOROUGH PUNAKAIKI 5 TE ANAU Fig. 1 - The study areas and their locations within the karst and limestone regions of New Zealand. After Williams (1982). Area 1, Northland; Area 2. Port Waikato and King Country; Area 3, southern Hawkes Bay and the Wairarapa; Area 4, northwest Nelson south to and including Punakaiki; Area 5, Fionlland. Caves were searched thmughout New Zealand, but the majority of sites containing frog bones were around WaitoIrlo or in northwest Nelson, particularly in the Honeycomb Hill Cave in the Oparara Valley. METHODS Frog material was examined in the collections of Auckland University Gcology Dcpartment, (AU); National Museum of New Zealand, (NMNZ); Waitomo Caves

5 204 Journal of the Royal Society of New Xealand, Volume WI I A B ~ Fig. 2 - (A) The distribution of subfossil remains of birds in tht:' North Island (aftn Milkncr, 1981), compared with (B) the distribution of subfossil frogs. Museum, (WO); Canterbury Museum, (CM) and Manawatu Museum. Whole individual skeletons of the presumably extinct "giant" forms allowed determination of the species characteristics and the correct association of disarticulated bones. Drawings of single bones were prepared using camera lucida apparatus or photography. Once each species had been described the subfossil assemblages were sorted into species. Limb boncs were the most commonly found, so only these were measured routinely. Measurements to the nearest 0.1 mm were made, with an eyepiece micrometer or vernier calipers, as follows. Tibiofibula: length, proximal and distal width. Humerus and radioulna: length, proximal width, distal width and shaft width. Femur and tibiale: length only (Fig. 3). Where available, Snout-Vent Length (SVL) of whole individuals helped to estimate the body size of sublossil specimens. Each bone was determined to have belonged to either an adult or a.iuvenile frog, according to the degree of ossification (Fox and Irving, 1950). Data for cach species were summarized by area, and the mean, range, standard deviation and variance for each parameter were calculated. Univariate analyses of variance confirmed the distinction of the nominate species (ie, those defined from shape and other non-metrical characters). The level of significance of any variation was tested using the Tukey Kramer Means Test, a suitable method for small, unequal sample sizes (Sokal and Rohlf, 1969). The data on the radioulna, humerus, and tibiofibula, the three bones most often found, were subjected to a multivariate discriminant function analysis, and the posterior probability of assigning the bones to the correct (nominate) species was calculated. The following text presents, first, osteological descriptions and diagnoses of the genus Leiopelma and of all six of its species: and then comparisons within the Leiopelmatidae and between the Leiopelmatidae and Ascaphus truei.

6 Worthy - Osteology oj Leiopelma 205 A. B. ~P..,j c. I L L L It-D~ D. r- P-1 E. r P" I L I \ \ \ \ -> r -I L ~D -J r- D --1 Fig. :) - Measurements recorded during this study. (A) Femur. (B) Tibiofibula. (e) Tibiale-fibularc. (D) Humerus. (E) Radioulna. (T" = length, P = proximal width, D=distal width, S=shaft width). OSTEOLOGICAL DIAGNOSES The available descriptions of the extant Leiopelma species emphasize non-osteological structures, or those parts of the skeleton (such as the carpus and tarsus) that are not recovered from subfossil deposits. For the purposes of the present study each Leiopelma species is therefore redescribed from osteological characters only. Generic Diagnosis of Leiopelma Nine presacral amphicoelous vertebrae; atlas lacks transverse processes; vertebrae 3 and 4 have associated ribs with characteristic uncinate processes (Figs_ 16, 17); sacral vertebrae narrower than preceding vertebrae, formed of two halves joined dorsally and ventrally by cartilage which may ossify in adults; no sacral post-zygopophyses, sacral diapophyses simple; maxilla dentate; paired dentate premaxillae; left and right prevomers with dentigerous processes meeting in a line across the roof of the buccal cavity, well developed anterior and posterior processes (Fig_ 11); no ossified quadratojugal; paired

7 206 Journal oj the Royal Sucil'ly oj Xw' Zealand, Volume 17, J987 frontoparietals with medial projections posteriorly and anteriorly j()rming two jc)j1tanellcs; pterygoid triradiate, (Fig. 6); angulosplenial as in Fig. 12; dentary a flat toothless blade with small mentomeckelians lijsed to it; parasphenoid triradiate with well devclopcdlateral alae and poorly developed posterior process; scapula not overlain by clavicle; coracoids as in Fig. 21; cleithrum present; ischia broad; ilia simple, sometimes with small dorsal prominence, oval cross-section; humerus with large more or less equally developed epicondyles, humeral ball may ossify and is relatively large; radioulna smaller than humerus, has olecranon process, shaft bilobed but boundary between radial and ulnar portions not evident in mid-shaft region; femur sinuous with well developed proximal crest; tibiofibula bilobed, tibial and fibular portions of equal length, boundary of fusion not evident over middle third of shaft; tibiale longer than fibulare, fused proximally and distally to enclose large interosseus space. Leiopelma hochstetteri Fitzinger, 1861 A small frog up to 50 mm SVL in adults. Skull: maxilla has characteristic notch in the pars facialis (Fig. 8); otic ramus of the squamosal shorter than zygomatic ramus; quadrate small, quadrangular, eventually ossifies in adults; frontoparietals with well developed medial projections (Fig. 5); cultriform process of the parasphenoid meets lateral alae at a distinctive narrowing (Fig. 13); prootic and exoccipital ossify and eventually fuse together in adults; prevomer with dentigerous process raised above plane of anterior and posterior process (Fig. 11); vertebral centrum about 1 mm in diameter; neural spine low, completely ossified, left and right parts of sacral vertebrae eventually ossify together, prezygopophyses oval large; urostyle down-curved, variable in form. Scapula: convergent anterior processes enclose glenoid foramen, dorsal surface expanded in a flange; procoracoid ossifies; clavicle equal in length to scapula, fused in older individuals. Humerus: relatively large, 64% of tibiofibula length, shaft to proximal end in straight line, lateral and medial ridges present. Radioulna: well developed olecranon process, ridged on the dorsal radial surface. Femur: length equal to, or slightly shorter than, the tibiofibula, sinuous in shape with a well developed proximal ridge. Tibiofibula with a small ridge variably present proximally on the tibia. Tibiale about 60 % of the length of the tibiofibula. Pelvis: ischial plate slightly developcd posteriorly but not dorsally, relatively wide (Fig. 31), pubis ossifies in older individuals, ilia with small dorsal prominence. Material examined: NMNZ AM 185, 201, 202, 203, 204, 205, 206. Leiopelma hamiltoni McCulloch, 1919 A small frog up to 50 mm SVL in adults. Skull: pars facialis of maxilla unnotched (Fig. 8); otic and zygomatic rami of squamosal similar in length; quadrate eventually ossifies; parasphenoid with distinctly wider neck where the cultriform process meets the lateral alae; prevomer with the dentigerous process not raised off the plane of the posterior/anterior processes; prootic and exoecipital calcify and eventually fuse in adults. Vertebrae: centrum 1 mm in diameter, neural spine moderately developed (Fig. 14); left and right halves of sacral vertebrae eventually fuse in adults; sacral prezygopophyses small and round; urostyle variable. Scapula: anterior processes divergent, no development of dorsal Bange; clavicle 25 % longer than scapula; procoracoid does not ossify. Humerus: proximal end curved from line of shaft, shaft round in cross-section, no medial/lateral ridges, about 58 % of the length of the tibiofibula. Radioulna: no lateral ridges. PelvzJ: ilia with no dorsal prominence; ischium relatively narrow, slight dorsal expansion; pubis ossifies unly rarely.

8 Worthy - OsteolO/!,y 0/ Leiopelma 207 Alaterialexamincd: NMNZ AM 177, 178, 179, 180, 181, 187, 188, 198, 199,200. Leiopelma archeyi Tmbott, 1942 A very small frog, up to 38 mm SVL in adults. Osteologically identical to L. hamiltoni in all aspects except size. Material examined: NMKZ AM 182, 183, 184, 186, 189, 190, 191, 192, 193, 194, 195, 196, 197, 207. Leiopelma markhami n.sp. A robust frog estimated to have measured mm SVL. Skull: Frontoparietals with well developed medial projections; maxilla with a prominant pars facialis, deeply notched; premaxilla with well developed horizontal lamina, teeth sockets, basal width of alae less than one quarter the length of the premaxilla; prevomers with 5 to 10 teeth on process that is on the plane of the anterior/posterior processes, posterior process longer and narrower than the anterior one with a marked constriction at its origin; parasphenoid - cultriforrn process longer than width of lateral alae and constricted at the junction; prootic and exoccipital calcified and may fuse; quadrate ossified in adults, and sleeved by the squamosal; sphenethymoid may ossify; squamosal- zygomatic ramus more robust and generally longer than the otic, length of the zygomatic ramus about half that of the ventral process. Vertebrae: neural spine low but completely ossified, height of the neural spine decreases posteriorly; transverse processes slightly forwardly directed on vertebrae 2, and backwardly directed and longer on vertebrae 2-9; centra circular in cross-section and mm in diameter; ribs on vertebrae 3 and 4 usually fused to the transverse processes, shape of the vertebrae vary but width across the zygopophyses roughly equals the height; left and right sides of the sacral vertebrae separated by cartilage, ossified in adults; sacral diapophyses simple and relatively unexpanded; sacral prezygopophyses oval, wider than long; urostyle rubust, down-curved and slightly dorso-ventrally compressed with 1 to 2 transverse processes arising from shaft at right angles then sharply angled posteriorly, fused nemal arch recuf\"cd posterio-dorsally in stepped fashion, foramen for spinal nerves round. Scapula: convergent anterior processes to enclose glenoid foramen, procoracoid ossified in adults, clavicle sometimes fused to the scapula, length of scapula equals that of the clavicle. Humerus: typically very robust, 73% length of tibiofibula; in adults the humeral ball completely ossifies to form a spherical structure with a diameter equal to half the distal diameter of the humerus; epicondyles large, especially the medial, well developed lateral, medial and ventral ridges often present, so the shaft is always angular in cross-section, a further small ridge (rugose lamina) present medially on the anterior end of the humeral shaft. Radioulna: has a well developed olecranon process, and occasionally a dorsal radial crest; structure of the carpus unknown. Pelvis: ilia with a small dorsal prominence, anterior ends of the ilia slightly flattened where they articulate with the sacral diapophyses, ilia short, i. e. the ratio of the length of the ilia (as measured from the mid-acetabular region) to the length of the vertebral column is less than one; pubis ossified in adults forming a broad triangular ventral expansion; ischium expanded posteriorly but not dorsally, and relatively wide; epipubis usually remains cartilaginous. Femur: about equal in length to the tibiofibula, markedly sinuous in shape and with a well developed proximal ridge. Tihiofibula: very robust (Fig. 36), distal end usually slightly wider than the proximal, tibia and fibula both circular and of equal size proximally; fibula somewhat flattened distally, proximal part of tibia may have small ridge 011 ventral surface; tibiale about 70 % of tibiofibular length; interosseus space between tibiale and hbulare only as wiele as the shaft of the tibiale; tarsus unknown.

9 208 Journal of the Royal Sonety of New Zealand, Volume 17, 1987 Type Data Holotype: NMNZ S23120 Honeycomb Hill Cave, Karamea. Coll. May 1985 by author. Paratypes: NMNZ S23121, NMNZ S23140, NMNZ S23141, NMNZ S23152 (same data as holotypc). W Haggas Hole, Waitomo. Coll. March 1984 by author. NMNZ S23414 Unnamed cave in the Fox River, Punakaiki. Further material identified but not accorded type status is listed in Appendix 1 of Worthy (1986). Distribution: Subfossil in North Island (Otangaroa, Waipu, Waitomo, Hawkes Bay, Coonoor, Martinborough) and in South Island (Takaka, Patarau, Karamea, Punakaiki, and Te Anau). Etymology: This species is named after Mr G. S. Markham, who pioneered the collection and identification of some subfossil Leiopelma in New Zealand. Leiopelma auroraensis n.sp. A very robust frog, estimated SVL about 60 mm. Forelimbs nearly equal in length to the hindlimbs. Skull: prootic and exoccipital fused; parasphenoid with cultriform process longer than combined width oflateral alae, cuitrif(jrm process has narrow junction with lateral alae; squamosal- zygomatic ramus twice length of otic ramus; maxilla with the pars facialis deeply notched; pterygoid triraclite (Fig. 4); vertebral centra 1.5 mm in diameter, neural spine low, totally ossified. Humeri: lateral and medial crests moderately developed, strongly developed ventral crest, humeral ball ossified, spherical, greater than 50'1<, width of epicondyles. Radioulna as in Fig. 4. Tibiofibula: very robust (Fig. 4); femur slightly longer than tibiofibula, very robust and sinuous with a strongly developed proximal crest; tibiale-fibulare very robust, fused, large interosseus space. Scapula: anterior processes convergent, enclosing a glenoid foramen, elavicles fused to scapula, ossified procoracoici, scapula to clavicle length ratio 1: 1. Pelvis: all elements fused, ossified pubis f(jrming large triangular ventral expansion, ischium wide, ilium with small dorsal prominence. The single known specimen has the f()llowing elements missing: left squamosal, left and right frontoparietals, left and right pl'emaxillae, left pterygoid, left angulosplenial, vertebrae 4 and 9, left femur, carpal and tarsal elements.. Type Data Holotype: NMNZ S23413 (Ex. W0207.8). Aurora Cave, Tunnel Rum, Lake Te AnelU, Fiordland. Coll. by author in February, Distribution: Known only from the holotype. Etymology: This species is named after the cave in which it was found. Leiopelma waitomoensis n.sp. The largest presently known Leiopelma, with estimated SVL about 100 mm. It had the greatest reduction in degree of ossification, so the terminal cartilages arc seldom calcified on long bones. Skull: prootic and exoccipital not known, presumably cartilaginous; frontoparietals unknown; maxilla with pars facialis unnotched; premaxilla has a relatively robust alae with width at base equal to a third the length of the premaxilla; squamosal- otic ramus shorter than zygomatic ramus. Vertebrae: centrae dorsoventrally compressed, about 2 mm in diameter, neural spines well developed but not completely ossified; ribs rarely fused to associated transverse processes; urostyle straight, transverse processes curved evenly posteriorly, fused neural arch recurved posteriodorsally evenly, not stepped, foramen for spinal nerve round. Pelvis: ilium with no dorsal prominence; ischium expanded dorsally, relatively narrow.

10 vi/ortiz)' - OsteologJ' of Leiopelma 209 Fig. 4-Reprcscntativc bones from N:\lNZ S23413, the type skeleton ofl. auriltlli'rlsi.l. Scale bar= 1 ern. Femur: distinctly smaller than tibiofibula, relatively straight. li'biofibula: sillall ridge sometimes present proximally on ventral side of tibia, relatively narrow. Tibiale-fibulare: fused, with relatively narrow interosseus space. Scapula: anterior processes divergent, no glenoid foramen; clavicle 25 % longer than scapula. Humerus: relatively small, 53 % length of the tibiofibula, lateral crests poorly developed so shaft circular in cross-section; humeral ball only partly ossified {{)fming a sub-spherical structure; proximal terminal cartilage only partly ossified resulting in a flattened appearance; Radioulna: no lateral crests. Type Data Holotype: NMNZ S23415 (Ex. W0315.5). Haggas Hole, Waitomo, Coil. by author March Paratypes: W0315.2, W0315.3, W same data as holotype. W Haggas Hole, Waitomo. ColI. by Mr C. J. Templer in 1960's. Further material identified but not included in the type series includes all that listed in Appendix 1 of Worthy (1986). Distribution. Subfossil in North Island only; Otangaroa and Waipu in Northland; Waitomo district; Hawkes Bay; Coonoor; Martinborough. Etymology. This species is named after the district from which the most remains have been recovered.

11 21 () Journal of the Royal Society of Nel ' Zealand, Volume 17, J 98 7 OSTEOLOGICAL COMPARISON OF LEIOPELMATID SPECIES In this section the elements of the skeleton are described in detail, and the distinguishing features of each spccies are summarised. The tcrminology follows Trueb (1973).. Skull Prootic and exoccipital These clements arc primarily cartilaginous but may become calcified, to equal degrees, in adults of extant spccies: both elements become fused, although the resulting structure is not very robust. In comparison, calcification ;s much stronger in L. markhami and L. auroraensis. In L. waitomoenjl~\, these elements have never been pr'eservcd, despite favourable conditions, which suggests that they may not have been calcified at all. A B C D E ~ Fig. 5 - Right frontoparictals of subf(jssil specimens, dorsal view. (A) L. fwch.ltetteri, Honeycomb Hill Cave (NM;\IZ S231(6). (B) L hamiltoni (NMNI S23111). Left frontoparictals of recent specimens. (e) L hoc/l.imleri from Corornanclel. (D) L. hamil/oni. (E) L archeyi from Cmomandcl. Scale bar in' mrn. i'hejrontoparietal (Fig. 5) All cxtant species have paircd frontoparietals, variable in shape (Figs. 5A and C, B and D), separated by cartilage. At a point equal to the most anterior process of the prootic, there are medial projections on the frcmtopar'ietals, better developed in L hochstelteri than in other species. The latcraj edge of thc frontoparietals curve ventrally, more so in L. hochstelteri and L. markhami than in J~. hamiltoni. Frontoparietals of L. auroraensis and L. waitomocnsis are unknown. Pterygoids (Fig. 6). Similar in all species except in size. Squamosal (Fig. 7). This bone connects the exocciptal to the quadrate and the pterygoid. It is essentially 'T' shaped, with the posterior ventrally-directed process equal in length to the combined len1-,rth of the otic ramus and the zygomatic ramus. The quadrate is sleeved by the squamosal on its lateral surface, but the two bones do nol fuse. Those of the three extant species arc similar in size, and identical in L. archeyi and L. hamiltoni with equal sized zygomatic and otic rami; in L. floc/lstpueri the otic ramus is shorter than the zygomatic ramus and flares out in it very thin flange of bone. Roth L. markhami and L. waitomoensis have similar fragile flared otic rami, which is nearly always broken in subfossil deposits. The squamosal fragment pr-esent in the only known skelelon of L. (1I1roraensis is similar to those of othn Leioprlma species but is br okcll.

12 a. h. Worthy - Osteology of Leiopdma 211 c. d. e.~ )! 1"..)/ Fig. 6 - Left ptcrygoids of sub/ossil specimens. (a) L. rnarkizarnl (W0315). (b) L walto1lwensls (NMNZ S23415). Left ptcrygoids of recent specimens. (c) L arcizeyl. (d) L izarnlltoni (NMNZ AM198). (e) L. Iwdlstetteri (NMNZ AM201). Quadrate The quadrate is a small bone which ossifies late in the sequence of bone formation (Stephenson, 1960). It was not ossified in the specimens of L. archeyi examined by Stephenson, but it was in several new specimens collected recently and examined during this study. Ossified quadrates were found in extant and subfossil L. hamiltoni and in L. markhami, but not in L. auroraensis, L. waitomoensis or L. hochstetteri. Nasals and Sphenethmoid Neither of these thin, fragile bones was positively identified from any of the subfossil material, even though all extant species have them. The Hyoid Only the posterior median process and the central parahyoid bone are ossified: the rest of the hyoid is cartilaginous in leiopelmatids and so not recoverable from subfossil deposits. A small parahyoid element is present in each of the extant species, and the two most complete skeletons of L. markhami (NMNH S23120 and S23121) have an element which may be the parahyoid. The Maxilla (Figs. 8, 9) This is the main dentate bone of leiopelmatids. The pars palatina abuts the pterygoid posteriorly, and anteriorly is greatly expanded dorsally to meet with the nasal in the extant species. The maxillae of L. archeyi, L. hamiltoni and L. waitomoensis are similar, though that of L. waitomoensis is larger than any of the other species. The

13 212 a. Joumal l!ftlii' Hoyai Society oj NeIL' Zealand, ~olwl1c J /, /987 c. d. e. / \ ~J Fig. 7 - Right (a-c) and left (d-c) squamosal of (a) L Im.hql. (b) L hlllllliton1 (NM;\IZ AM 1 ')g). (e) L hocill/eiterz (NM;\IZ AM2(1). (d) L lnuiomocmis (NMNZ S2:l415). (e) L. markhaml (Site 52, N I'vlN:L U nrcg.). ot = otic rami, zr = zygomatic rami. Scale bar in mm. maxilla of L. hochstetteri, L. markhami and L. auromensis has a marked notch in the pars facialis slightly anterior to the greatest dorsal expansion; in L hochstetteri, the smallest of the three, the posterior end is rdatively larger and Illore pronounced. The Premaxilla (Fig. 10). This is a small dentate bone anterior to the maxilla which exhibits considerable intraspecific variation. Up to 15 teeth or teeth sockets may be present in premaxillae of the extant species. The upper section of the alary process varies in shape and sweeps back later ally. The base of the alary process is proportionally wider in

14 A..., Worthy - Osteology of Leiopelma 213.., B..., pf pp ~.., Fig. 3 - Right maxillae of subf()ssil specimens from Honeycomb Hill Cave. (A) L. floch,iettm (;\IMNZ S231(6). (8) L. hamlltoni (NMNZ S2:3 111). pf= pars facialis, pd = pars dentalis, pp = pars palatina. Scale bar in mm. Fig. 9-Maxillae of (A) L waitomoensis NMNZ S (B) L. auroraensis NMNZ S (C) L. markhami NMNZ S23120 Scale bar = 1 em

15 214 Journal o/the Royal Society 0/ New Zealand, Volume 17, 1 (Iii 7 a. "' ;-ap ~'::;:;:l- pp b. c. " Fig. 10- Premaxillae of subfossil specimens. (a) L. hamiltoni (NMNZ S23111 ). (b) L. hockl/cltcri (N:vINZ 52:{1(5). (c) L. waitomoensi.\ (NMNZ S23415). (d) L. markhami. ap =alary process, pel = pars clentalis, pp = pars palatina. Scale bar in mm. L. waitomoensis than in L. rnarkharni, and hence the alary process is less robust in L. rnarkhami. The pars palatina projects beyond the pars dentalis to a greater degree in L. hochstetteri than in either L. arch~yi or L. hamiltoni. The Preuorner (Fig. 11). In all Leiopelrna species this bone has three distinct elements: a medial dentigerous ridge, an anterior process, and a posterior process. In all extant Leiopelma species the dentigerous ridge, carrying two interlocking rows of teeth, is aligned across the roof of the mouth, with the left and the right side not quite meeting in the middle. The teeth on the prevomer are successively replaced in the same way as those on the maxilla and premaxilla. The number of prevomerine teeth varies from 3-6 in the extant species, 6 in L. rnarkharni, up to 11 in L. waitornoensis. The structure of the prevomer of L. hochstl/tteri is unique in that the dentigerous ridge is on a long "neck", raising it above the plane of the other two processes. The shape of the anterior and posterior processes are similar in all species. Both processes are flat and in the same plane; the anterior one is much broader but less robust, whereas the posterior process arises as a thin neck before expanding to twice the width then tapering to a point.

16 Worthy - Osteology oj Leiopclma 215 a. h. c. d. e. ~~ ~. ~4 Fig. ll-prevomers from recent specimens. (a) L anheyi. (b) L. hamiltonz (NMNZ AM198). (c) L. hochstelleri (NMNZ AM2(1). Prevomers from subfossil specimens. (d) L. markhami (Ruakokapatuna NMNZ Unreg.). (e) L waitomoensi.1 (NMNZ S23415). Scale bar in mm. ap = anterior process, pp = posterior process. a. b. \\ \~\.~ \ \ Fig. 12-Angulosplenials ofreeent specimens. (a) L archqi. (b) L. hoc/llittteri (NMNZ AM201). (c) L hamiltoni (NMNZ AM198). Scale bar in lillll.

17 216 Journal 0/ the Royal Soaety of New Zealand, Volume 17, 1987 Angulosplenzal (Fig. 12). Similar in al\ species except in size. It articulates with the quadrate and anteriorly narrows to a point medial to the dentary. Dentary and Mentomeckelian These two bones are fused with the left and the right mentomeckclians, abutting but not fusing in the midline of the lower jaw. There was no observable interspecific variation other than size. a. b. ~fo -cp c. d. e. {;~~ II \J f. c~~~~ ( \ Fig. 13-Parasphenoids. (a, b) Recent L. archeyi. (c) recent L. hamiltoni NMNZ AM198. (d) subfossil L. hamiltoni (NMNZ S23114, with associated fused prootic and exoccipital region). (e) recent L. hochs/eueri (NMNZ AM201). (I) subfossil L. hochstetteri (NMNZ S23105). la 0= lateral alae, cp 0= cultrii()rm process. Scale in mm. Parasphenoid (Fig. 13). This triradiate bone forms the base of the skull. The cultriforrn process extends nearly to the prevomers. Posteriorly there are left and right lateral processes (alae), which lie ventral to the prootic and exoccipitals and which have a combined width slightly less than the length of (he cultriform process. Where the cultriform process meets the alae there is a distinct neck, varying in relative width between species. This width

18 Worthy - Osteology oj Leiopelma 217 can conventiently be expressed as a percentage of the width of the alae (Table 1). The neck in L. hochstetteri is proportionally narrower than in either L. archeyi or L. hamiltoni. This difference is accentuated by the different width of the alae. Table 1 - Mean measurements of parasphenoids in species of Leiopelma SPECIES LENGTH WIDTH NECK W PROP wt N archevi hamiltani hochsteueri markhami 13.7* auroraensls waitomoensis t PROP W=NECK wx 100/WIDTH. * N=2 There are few subfossil specimens of this bone. Those known for L. markhami are similar in shape to L. hochstetteri; the one complete parasphenoid of L. waitomoensis and L. auroraensis each has a relatively wide neck. Vertebral Column (Figs ) Leiopelmatids have amphicoelous vertebrae, i. e. those in which the vertebral centra are ectochordal, slightly biconcave or flat terminally and the intervertebral cartilage is contiguous to, and not subdivided between, successive presacral vertebrae (Trueb, 1973). According to Stephenson (1960) all three extant species have a similar vertebral column, differing only in L. archeyi in which it is finer and more delicate. There are, however, several differences between extant species. L. archeyi and L. hamiltoni share identical form and differ only in size, whereas L. hochstetteri has vertebrae that are typically more dorsoventrally flattened because the neural spines are not as pronounced (Fig. 14). The neural arch is ossified to the same extent in adults of each of the extant species. In L. hamiltoni, L. markhami and L. waitomoensis, but not L. hochstetteri, both sides of the sacral vertebrae are commonly fused together. Furthermore the sacral vertebrae of L. hochstetteri differ in shape from those of L. hamiltoni. On the antero-dorsal surface of the sacral vertebrae there are prezygapophyses, with shapes characteristic of each species (Fig. 15). In L. hamiltoni they are small and tend to be circular, while in L. hochstetteri they arc relatively larger and tend to be oval in a transverse direction. Also, in L. hochstetteri the sacral diapophyses are relatively more expanded than in L. hamiltoni. The vertebrae of the extinct species can be distinguished on size alone. The vertebral centra of each species, circular in cross-section, measure about 1 mm diameter in L. markhami, 1.4 mm in L. auroraensis, and about 2.0 mm in L. waitomoensis. L. markhami and L. auroraensis have vertebrae with a low neural arch on which the neural spine is wholly ossified (Fig. 16). L. waitomoensis tends to have a raised neural spine, which never completely ossifies. In all species, ribs are present on vertebrae 3 and 4, each with an uncinate process, best developed on the rib associated with vertebra 3. Fusion of these ribs to the associated transverse process of the vertebrae appears to be common in L. markhami and L. auroraensis but not in L. waitomoensis (Fig. 17). Abnormalities such as fusion of adjacent vertebrae (synostosis) were recorded by Stephenson (1952, 1960) and was observed in specimens of L. hamiltoni and L. markhami from sites at Honeycomb Hill Cave in northwest Nelson. An extreme example of this was seen in a specimen of L. markhami collected from Waipuna Cave at Waitomo (AU 7459). There is fusion of vertebrae 2-5 inclusive, and, furthermore, vertebrae 3-5 have fused ribs, each with well developed uncinate processes. In a recent specimen of L. archeyi, an adult of SVL = 31.6 mm, collected from the Tapu-Coroglen summit, Coromandel,

19 218 Journal of the Royal Society of New Zealand, Volume 17, 1987 A. B Fig Selected vertebrae. (A) L hamiltani (NM S23111). (B) L hochstetteri (NM S23105). Al = atlas; A2 = vertebrae 3-4; A3 = vertebrae 5-6; A4 = vertebra 7; A5 = vertebra 3; A6=vertebra 9. HI = vertebra 2; H2=vertebra 3; B3=vertebra 4; B4=vertcbra 5; B5=vertebrae 6-7; B6=vertebrae 3-9. Scale bar in mm. the sacral diapophyses had developed on vertebra 9 on the left side and on vertebra 10 on the right side. In each specimen the opposite side had only a small transverse process.

20 Worthy - Osteology of Leiopelma 219 a. h. ~c.~ Q==J Fig. 15-Sublossil specimens of right sacral vertebrae. (a, b) L. hochstetteri (NMNZ S23106, NMNZ S23105). (c) L. hamlltoni (NMNZ S23111). Scale bar in 11l11l. D {\ 0 :" ".. ~, v " \',. ~-- -_/ A E B :~ o \KJ... ~~..., F c G H =- Fig. 16-Vertebrae ofl. markhami (NMNZ S23120). (A) anterior atlas. (D) posterior atlas. (C) left lateral atlas. (D) posterior view vertebra 3. (E) dorsal view vertebra 3. (F) dorsal view vertebra 5. (G) dorsal view vertebra 10. (H) posterior view vertebra 10. (I) anterior view vertebra 10. Scale bar in mm.

21 220 Journal of the Royal Soriety of New Zealand, Volume 17, 1987 a. d. :~~ f. 0"' :\:_~ g. J. ~ k. /' '>.. / ). ~\ \ 'r-< ( I. h. <:: I. Fig V t'rtcbrac of L. wailomoensis (,"IM NZ S23415). Scale bar in mrn. (a-c) atlas. (d-c) vertebra 3. (f-g) vertebra 4. (h) vertebra 5. (i) vertebra 9. (j-i) vertebra 10. In Leiopelma the coccyx or urostyle is extremely variable; either one, two or three transverse processes may be present, and these vary in relative stoutness. However the urostyles of L. markhami and L. waitomoensis can be distinguished on several characters. (1) The long posterior section is distinctly convex in L. markhami and straight in L. waitomoensis. (2) The fused neural arch (fna) anteriorly tends to rise abruptly and at right angles to the rest of the bone in L. markhami whereas in L. waitomoensis it slopes upwards in a distinct posterior direction. The first pair of transverse processes are initially at right angles to the length of the urostyle, then are angled sharply in a posterior direction in L. markhami. In L. waitomoensis these are evenly tapered and curved posteriorly (Fig. 18). (3) The foramen for the exit of the twelfth spinal nerve, which is shielded by the first transverse process, is circular in L. markhami and oval in L. waitomoensis. L. auroraensis, which has a more

22 Worthy - Osteology of I -t::iopelma 221 robust skeleton than L. markhami exhibits a combination of the characters of L. markhami and L. waitomoensis, its relatively slender urostylc is straight, has backward curved transverse processes, an abruptly arising fused neural arch and an oval foramen for the exit of the twelfth spinal nerve. Individuals of L. markhami exhibit progressively greater skeletal robustness in southerly latitudes. The urostyle of specimens from northwest Nelson are very robust and tend to have well developed lateral ridges along the posterior rod. The specimen of L. auroraensis has the least robust urostyle observed in any of the extinct species from anywhere in New Zealand. a. b. c. d. e. f. g fit T1irJ) o~f(./\ r<:::i n u..;~! f,',- " 'V Fig. 18-(a-d) Urostyles of L waitomoensi." (e-g) Urostyles of L markhami. Lateral (upper) and dorsal (lower) views. Scale bar 1 cm. fna = fused neural arch; f12 = twelfth spinal nerve, U rostyles of the extant species arc also variable but differ in the shape of the first transverse process. In L. hochstetteri this is shorter and curved posteriorly likc those of L. waitomaensis, whereas those of L. hamiltani and L. archeyi arc like L. markhami. The Pectoral Girdle (Figs ) The clements of the pectoral girdle of L. anheyi and L. hamiltoni do not differ except in size. In L. hochstetteri the scapula is distinctly different in form (Fig. 19): the midsection is expanded dorsally to form a thin flange, and the two anterior processes do not diverge. L. hochstetteri is the only extant species in which the procoracoid is ossified (Stephenson 1960): in older specimens the clavicle may become fused to the scapula. This condition was observed only once in L. hamiltoni, in a specimen from Honeycomb Hill Cave. The scapula and the coracoid arc the most robust and more frequently found girdle elements: the others are small and tend to be under-represented in the subfossil record. Scapulae of L. markhami also show ossification of th(> procoracoid, and often fusion of the scapula and clavicle (Fig. 20). The two interior processes do not diverge, and arc

23 222 Journal of the Royal Society of New Zealand, Volume 17, 1987 a. c. b. d. ;~ Fig. 19-(a-b) Right scapulae of subfossil specimens of L. hochstetteri. (a) exterior, (b) interior views of NMNZ S (c-d) Right scapula of sub fossil L. hamlltonz (NMNZ S23111). Scale bar in mm. often fused anteriorly to form a glenoid foramen. In contrast to L. hochstetteri, there is no extensive development of a dorsal flange, so the scapula retains a distinctly concave upper profile. The scapula of L. auroraensis is similar in all respects to that of L. markhami. Scapulae of L. waitomoensis do not have an ossified procoracoid, so the scapula and clavicle are never fused. The two anterior processes of the scapula are divergent, and the dorsal surface is deeply concave. The size of the scapula relative to the clavicle varies with species. In L. markhami, L. auroraensis and L. hochstetteri the scapula and the clavicle are equal in overall length; in L. waitomoensis the clavicle is about 30 % longer; in L. hamiltoni and L. archeyi it is about 25 % longer. The cleithrum in L. waitomoensis is unknown, but in L. markhami it is slightly longer than either the scapula or the clavicle. Coracoids differ only in size between all six species (Fig. 21). The Forelimb Humerus (Figs ). Humeri of L. archeyi and L. hamiltoni differ only in size. In L. hochstetteri the humeral shaft is more robust and the proximal end is straight, not curved medially; and lateral ridges on the dorsal margin make the dorsal surface of the shaft squared towards the distal end, rather than rounded, in L. hamiltoni (Fig. 22). In living L. hochstetteri the males have more muscular forearms (Bell, 1978); hence they also have more pronounced lateral ridges on the humeri than females (Figs. 22, 24). Disassociated humeri of L. markhami and L. waitomoensis are relatively difficult to identify. The lateral crests and rugose lamina are much more pronounced in L. markhami than L. waitomoensis, although there is considerable variation, and the terminal cartilages are

24 Worthy - Osteology of Leiopelma 223 a. b. ~.... ~~ d. c. ~... "',." '. ;."..... '.< :"'" e. Fig. 20-Lateral views of left scapulae. (a) L. markhami (NMNZ S23120). (b) L. archeyi. (c) L. hamiltani (NMNZ AMI98). (d) L. hochstelten (NMNZ AM2(1). (e) L. waitomoensis (NMNZ S23415). (a) and (d) have fused clavicles. Scale bar in mm. a. b. Fig (a) Coracoid, and (b) clavicle, of L. waitomoenlzs (;\!MNZ S23415), typical of Leiopelma species. Scale bar in mm.

25 224 Juurnal of the Royal Suciety oj New Zealand, Volume 17, 1987 Fig Ventral views of humeri. (A) L ho(hstetteri (NMNZ AM201). (E) L ar(hcyi (NM:\fZ AM189). (e) L hamiltoni (NMNZ AM198). Scale bar=().5 elll. Fig. 23-Humeri in ventral aspect. (A-C) L. markhaml (NMNZ S23120, 2314(), 23252). (D) L auroraensi.\ (NMNZ S2:H13). (E,F) L waitolnoensis (W0328, NMNZ S:B415). Scale bar=l elll.

26 Worthy - Osteology oj Leiopelma 225 a. b. c. Fig. 24-Subfossil humeri from Honeycomb Hill Cave, Karamea, NW Nelson, in ventral, anterior, dorsal and posterior views. (a) L. hamiltani (NMNZ S231.'i7). (b) L. hochstetteti (NMNZ S23138). (c) L. hochstetteti (NMNZ S23105). Scale bar in film.

27 226 Journal of the Royal Society of New Zealand, Volume 17, 1987 distinctively ossified in adult specimens; the capitate eminence is usually a spherical body, equal in diameter to half the distal width of the humerus, and calcification at the proximal end results in a rounded appearance. Humeri of L. waitomoensis are more robust, and the range of measurements for the shaft diameter almost excludes those of L. markhami (Table 2); the capitate eminence is only partly calcified, as a sunken hollow on the surbce, and the proximal end, although sealed, has a truncated rather than rounded appearance (Fig. 23). Table 2 - Length statistics for adult humeri of Leiopelma species, by area. SPECIES AREA t MEAN S.E. RANGE cv K markhami :) : auroraenslj waitomoensis anheyi E hamiltoni E !l hochstetteri E B t Area 1 = Northland; Area 2 = Waitomo; Area 3 = Wairarapa/Hawkes Bay; Area 4 = NW Nelson; Area 6 = Fiordland. E = Extant populations Table 3 - The posterior probability of correctly classifying humeri into the appropriate nominate spccies from measurement data alone, showing the number of observations and relative percentages. For example: 112 humeri of L. markhami (or %) were correctly classified back to this species from measuremcnt data alone. KOMINATE SPECIES markhami lj.jaitornocnsis archeyi hamiltoni!lochsteueri TOTAL markharni :) !l8.8'! auroraensls waitomoensis 1: B archeyi izarniltoni U hochstetteri (l.oo 13.:n loo.no Within a given geographical area each species has a narrow size rangc, which may overlap that of other species to some extent (Table 2); but a univariate analysis of the data confirmed that the nominate species, that is, those defined from shape characters, accounted for a high proportion of the observed statistical variation: 80% of length, 81 'Ir)

28 Worthy-Osteology of Leiopelma 227 of proximal width, 80% of distal width, and 74% of shaft width. 1\1ultivariate techniques were therefore employed to see if each species could be identified from statistical data alone. Plots of canonical variables, derived from discriminant function analyses, showed (Fig. 25) that the six species tended to form distinct clusters. The posterior probability of membership to the nominate species, determined from all four available variables (Table 3), confirmed that the measurements do characterize each nominate species. The only exception is that L. hamiltoni was misclassified as L. hochstetteri 25 % of the time. This is understandable, since they are of similar size and robustness; but they can be easily separated from the osteological characters described. 'S '; ~ ~ S5!i 5 5" 5 ~ ~~ 5s~B:i5~55n 5 '>55 S 'i 'is '> '> ~ 5 '> 55 S 44 ~s.,, 6, ] 1 1) 11 J 1 11 J ",',11 1 \' Jl 1 l t 11 J ) 1 1 jj ; ] j " -, -3_ ,., u Fig. L5 - A canonical plot of two variables of humeri per species (length and relative robustness). 1 = L. marklzami; 2 = L. auroraenszs; :) = L. waitomoen.,i.,; 4 = L. archeyi; 5 = L. Izamiitoni; 6 = L. Izodzsteltrri. Radioulna (Fig. 26). Radioulnae of L. hochstetteri and L. hamiltoni arc very similar but in L hochstetteri the shaft is proportionally wider; the dorsal radial surface is angled into a crest and the cross section of the mid shaft is not circular but ridged on one side; and the olecranon process arises from the articular surface of the radius at a steeper angle. These differences arc accentuated in males. The ranges of length in L. archeyi and L. hamiltoni are mutually exclusive, although the sample sizes are small; the ranges oflengths in L. hochstetteri and L. hamiltoni overlap. Radioulnae of extinct species are difiicult to identify if not associated with more diagnostic elements. However identification is usually possible from length, since within a given area the size ranges of adult bones of L. markhami and L. waitomoensis are mutually exclusive Cfable 4). The height of the olecranon process is also less than the width of the radius in L. waitomoensis but greater in L. markhami. A univariate analysis of variance for the variables length, proximal width, distal width and shaft width showed that the nominate species accounted for a high proportion of the variation, (85 %, 85 %, 81 %, 80 %, respectively) and the differences were significant (P=0.05) for all variables in L. markhami and L. waitomoensi.1 (Table 5). In the extant species the measurements were not significantly different except for length in L. archeyi compared with L. hamiltani. The univariate analysis indicated highly significant differences between species in several of the measured variables, so a discriminant function analysis was carried out to see whether a multiple factor approach might improve the statistical separation of nominate species. In a plot of two of the canonical variables (Fig. 27) all species except L. hochstetteri and L. hamiltoni separated out into clusters. The posterior probability analysis verified this, and confirmed the correct classification of bones to their nominate species most of the time (Table 6). The radioulnae of L. hachstetteri and L. hamiltoni can easily be distinguished from their shape.

29 228 Journal of the Royal Society of New Zealand, Volume 17, 1987 a. b. c. o o Fig. 26-Radioulnae. (a) L. archeyl. (b) L (NMNZ AM201). Scale bar in mm. hamiltoni (:--.IMNZ AMI(8). (c) L. IlOch.,let/en Table 4 - Length statistics for adult radioulnac of Leiopelma species, by area. SPECIES AREA t MEAN S.D. RANGE CV N markhaml SI waitorrwensi S lS :) :) anheyi E " :) hamlltoni E B IlOchstelleri E S auroraensls t Area 1 = Northland; Area 2 = Waitorno; Area:3 = Wairarapa/Hawkes Bay; Area 4 = NW Nelson; Area 6 = Fiordlancl. E = Extant populations The carpus of the extant New Zealand species has been described by Stephenson (1952, 1960) but since subfossil material is always disassociated, the configuration of the carpal elements of. the extinct species is unknown. The relative proportions of the humerus and the tibiofibula in each species show (Table 7) that the forelimb is reduced to differing degrees in each. L. auroraensis is an extremely stout frog with relatively massive forelimbs, while at the other end of the scale is L. waitomoensis with forelimbs reduced to half the length of the hind legs.

30 Worthy - Osteology of Leiopelma , II ~ :5 \,,,,,,, '" I 1 1 1,'1,33 " I 3 t , 3,',1,1 1 3, JJ JJ,, JJ 3 ) 3 33,, Clll '.5 Fig. 27-A canonical plot of two variables of radioulna per species (length and relative robustness). 1 = L. markhami; 2 = L. autoraensi.l; 3 = L. waitomoensi.l; 4 = L. atcheyi; 5 = L. hamiltoni; 6 = L. hochstetteri. Table 5 - A pairwise comparison of the relative significance of the difference between species for the variables: length, proximal width, distal width and shaft width of radioulnae. SPECIES t LENCTH PROX W DIST W SHAFT W 3-1 *** *** 3-2 *** NS.'is :)-5 *** *** *** 3-6 *** *** 1-2 NS.'is NS 1-5 *** *** NS 1-6 *** *** 1-4 *** *** *** 2-3 *** NS NS 2-1.'is NS NS NS 2-5 NS *** *** 2-6 ***.'is NS 2-4 *** 5-3 *** *** 5 1 *** *** *** 5-2 NS NS 5-6 NS.'is.'is NS 5-4.'is.'is NS 6-3 *** *** *** 6-1 *** *** 6-2 *** NS NS 6 5 NS NS NS NS 6-4.'J"S NS NS NS 4-3 *** 4-1 *** *** 4-2 *** *** NS 4-5 NS NS NS 4-6 NS NS NS NS t Species 1 = markhami; 3 = waitomoen.lis; 4 = archeyi; 5 = hamiltoni; 6 = hochstettri. * * * Significant at P=D.05; NS Not significant.

31 230 Journal oj the Royal Society of New Zealand, Volume 17, 1987 Table 6 - The posterior probability of correctly classifying radioulnae into the appropriate nominate species from measurement data alone, showing the number of observations and rciati\'c percentages. See Table 3.!'JOMI!'JATE SPECIES markhami waitomaensis archeyi hamiltani hochstetteri TOTAL markhami {luroraenszs D.OO uwitomoensis : loo.oo archeyi hamiltani 0 () : hochstetteri :l Table 7 - The length of the humerus expressed as a percentage of the length of the tibiofibula. SPECIES % AREA I{AKGE N hamiltoni extant :l archeyi extant hochstetteri extant markhami Waitorno G markhami 7:).84 NW Nelson 6S.B-81.S 7 {luroraenslj Fiordland 1 lfjaitomoenjis Waitomo The Pelvic Girdle Ilium (Fig. 28). The dorsal acetabular rim is on a level with the dorsal surface of the shaft., and the ventral acetabular expansion is only slightly developed and a dorsal protuberance and an iliac crest are lacking. There is a dorsal prominence only on ilia of L. hochstetteri, L markhami and L. auroraensis. Immediately posterior to the dorsal prominence, and dorsal to the anterior margin of the acetabulum, there is a small depression in all species. The shaft is more ventrally curved in L. markhami (and L. aururaensis) than L waitomoensis, and ilia of L. markhami arc smaller: the length of the ischial contact ranges from 3.2 nun to 5.0 mm, mean = 4.47 mm, n = 12 in L. markhami, but 4.7 mm to 8.0 mm, mean = 6.45 mm, n = 32 in L. waitomoensis. 1schium (Figs ). The ischia of all Leiopelma species other than L archeyi and L. hamiltoni are distinct (Fig. 29), espccially when viewed in lateral profile or Ji'om the dorsal aspect. Pubi were ossified and fused to the ventral edge of the ischium in all species. The ischium of L. hochstetteri is proportionally broader than either of the other two extant species. The width, expressed as a percentage of height, is % in L. hochstetteri, but about 50 % in L hamiltoni. In lateral aspect the ischial flange is developed dorsally more extensively in L. hamiltoni. The ischia of the extinct species are significantly larger than those of the extant species. Ischia of L. markhami and L. auroraensis are identical, and together are easily separated from the much larger L. waitomoensis (Fig. 29). The ischial flange in L. markhami and L. auroraensis is similar to that of L hochstetteri, while that of L. waitomoensis is much more developed in a posterio-dorsal direction, as in L. hamiltoni, only more so. In older animals, in which ossification was extensive bef(xe death, the ilium-ischium complex may survive as a single element if the preservation was good (Fig. 30).

32 Worthy - Osteology of Leiopelma 231 a. ~~=-~ b. ~~ ~-~J d. C. /_---{\ / -r-l '~d ~ ))/c Fig. ~W - Ilia. (a) L (mhe)'! (b) L ho(h.,ltllm (\1:\1 \iz A:\1:Z0 1). (c) L hillllllioll! (\I!\10!Z AM19B). (d) L markham! (:\,:\1\;Z L!nn'g.). (e) L /NuIOlIlOi'll.'!' C''';:\1 NZ S2:l415). Scale bar in lilll!. a. c. " ~~~J ~>j b. d. 0~~) \ \. Fig \'ostltiur and latltal profiles o\' ischia. (a) 1. IIWrUlIlIl/i (\ \'():\ \~). (b) L /i'a!ioil/ol'1/'!' (WO:l:22.2) (c) L I""hllttlm \\;:\1\1Z S2:l1(5). \ti) L hamillon! (\lmnz S:Z:11:i7). Scale bar in 11l1l!.

33 232 Journal oj the Royal Society oj New Zealand, Volume 17, J 987 a. b. Fig. 30-A complete ilium-ischium complex (pelvis) of L. markhami. (a) ventral. (b) dorsal (NM S23120). Scale bar in mm. a. b. c. d.. '-~(fj') ~::":'.,;. ;..'...:...:'~ ~....,-. ':.',':',',... Fig. 31- Right lateral views of the acetabular region of pelves. (a-c) L markhami (A-W0304, B-W0320, C-NMNZ S23120). (d) L waitomoensis (W030S.2). Scale bar in mm.

34 Worthy - Osteology of Leiopelma 233 The ossified pubis has prominent pubic eminences, and ventrally a broad triangular expansion. In L. markhami there is intraspecific variation in the degree of ossification around the acetabular fossa (Fig. 31), but the size and profile of the ischial flange is fairly constant. In L. waitomoensis the shape of the acetabular fossa is similar to that in L. markhami, but the ischial flange is markedly more expanded dorsally. The pubis supports the epipubis via connective tissue only, presumably to allow for movement. Fig Epipubis and two unidentified bone elements, possibly either post pubal rods or abdominal ribs, from L. markhami (NMNZ S23120). Scale bar=0.5 cm. Epipubis (Fig. 32). Only one subfossil epipubis is known, from a L. markhami from Honeycomb Hill Cave (Fig. 32). It is very similar in outline shape to that of L. hochstetteri (Stephenson, 1952) and is obviously calcified cartilage. However the posterior portion of this epipubis has two projections, like the base of a pair of hollow rods, suggesting that L. markhami had post pubal rods or Nobelian bones. Associated with the skeleton with which the epipubis was found were two rod-like structures (Fig. 32) whose function can be interpreted in either of two ways. Their shape and basal diameter matches the projections on the epipubis, so that they could be post pubal rods. However there is a dear affinity between L. markhami and L. hochstetteri, which has cartilaginous abdominal ribs known to calcify and in which there is no evidence of Nobelian bones (Stephenson, 1952); hence there is the possibility that these are a pair of abdominal ribs. Skeletons of recent specimens of L. hochstetteri were examined, and series of X-ray plates of further specimens. The first abdominal rib is markedly shorter and more robust than the posterior three, and similar in form to the two elements figured here. The available evidence on the identity of the epipubic projections is therefore inconclusive, but supports the suggestion that L. markhami may have had Nobelian bones. The Hind Limb In subfossil deposits the tarsus and the pes do not survive in articulation, and the tibiale-

35 234 Journal of the Royal Society of New Zealand, Volume 17, 1987 fibulare are rarely recovered. No character could be identified from the extant species which might allow them to be recognised to species. Fig Femurs. (A) L. hochstetteri (NMNZ AM201). (B) L. archeyi (NMNZ AMI89). (C) L. hamiltani (NMNZ AMI98). Scale bar=o.) ern. Fig. 34-Femurs. (A-C) L. markhami (NMNZ S23120, 23140, 23152). (D) L. auroraenri., (NMNZ S23413). (E,F) L. wailomoensis (W0323, NMNZ S23415). Scale bar= 1 em.

36 Worthy - Osteology of Leiopelma 235 Table 8 - Length statistics for femurs of the Leiopelma species by area. SPECIES AREA t MEAN S.E. RANGE GV N hochstetteri E archeyi E hamiltoni E markhami auroraensis waitomoensis t Area 1 = Northland; Area 2 = Waitomo Region; Area 3 = Hawkes Bay-Wairarapa; Area 4 = NW Nelson; Area 5 = Fiordland. E = Extant populations Femur (Figs. 33, 34). The femur is slightly shorter than the tibiofibula in all species and there is virtually no variation in form other than that of absolute size (Figs. 33, 34). A Leiopelma femur is easily distinguished from those of the introduced Litoria species, which are also found in caves, because Leiopelma femurs are more obviously "sigmoid" in shape, the ends are much greater in diameter than the shaft, and the femur ridge is much better developed. The statistics on femur lengths in the Leiopelma species (Table 8), show that (1) there is regional variation in some of the extinct forms; (2) the size ranges in the extant species overlap to some extent; (3) size alone distinguishes L. waitomoensis from L. markhami and L. auroraensis, although the latter also tend to be more robust and have a more curved form (Fig. 34); (4) the larger specimens of L. hochstetteri and L. hamil/ani fall within the range for L. markhami, but are easily distinguished by their much slighter build. Fig. 35-Tibiofi.bulac. (A) L. hochsteltrri (NMNL: AM2(1). (B) L (mhtyi (N~NZ AM189). (e) L hamiltani (NMNL: AM198). Scale bar=o:i CIll.

37 236 Journal of the Royal Society of New Zealand, Volume 17, 1987 Fig. 36-Tibiofibulae. (A-C) L markhami (NMNZ S21320, 23140, 23152). (D) L auroraensl.l. (E,F) L wailomoensis (WO:)28, NMNZ S23415). Scale bar= 1 ettl. Tibiofibula (Figs. 35, 36). This is a composite bone formed by the fusion of the tibia and the fibula. At the proximal end both bones are of equal size and circular in crosssection, but at the distal end the tibia is somewhat flattened, so this end of the bone is wider. There is a distinct suture on the terminal third of each end of the bone. There is little intraspecific variation in the form of the adult bone, except the presence or absence of a small tibial ridge on the ventral surface of the proximal end. In a sample of 56 tibiofibulae from a single site in Honeycomb Hill Cave, 61 % were ridged, the rest not so; these were associated with humeri of which 85 % were referable to L. hamiltoni. ] uvenile tibiofibula are easily distinguished in all species because the non-calcified terminal cartilages fall oft, leaving hollow ends. Calcification of cartilage increases with age, more so in L. markhami than in L. waitomoensis, but equally in the extant species. Tibiofibulae of the extinct species are easily identified from either size or shape. L. waitomoensis has an exclusive size range (Table 9), and a distinct form (the lateral edge of the mid shaft is sharply angled in this species but not in others); L. markhami and L. auroraensis are each easily separated from other species by relative width, e.g. the proximal end is % and 22 % of length respectively. The tibiofibulae of extant species are not separable on any character, and only L. archeyi has a distinct size range. Since the tibiofibula is the most frequently recovered bone, and many of these are juvenile, it is important to know the probability of assigning any given bone to the correct species. Data for juvenile and adult bones were analysed and compared separately. A series of univariate analyses was made to see how mueh of the total observed variation was explained by the nominate species (Table 10), using the following variables; length, proximal width, distal width, and the ratio of proximal width/length. Relative width was included because it negates the effect of absolute size, so it may be suitable fc)r comparing species of different sizes. The results (Table 11) suggest that there are real differences between the nominate species as f()llows. (1) L. waitomoensis can be separated from all other species, either as

38 Worthy - Osteology oj Leiopelma 237 Table 9 - Length statistics for adult tibiofibulae of each Leiopelma species hy area. SPECIES AREA t MEAN S.D. RANGE cv N markhaml noo waitomoen5.ij al1roraenslj hochstetteri E archeyi E hamiltani E t Area 1 == Northland; Area 2 == Waitomo Region; Area 3 == Hawkes Bay-Wairarapa; Area 4 == NW Nelson; Area.5 == Fiordland. E == Extant populations. Table 10 - The R 2 Values from a univariate analysis in whieh the proportion of observed variation ascribable to the nominate species was determined for juvenile and adult tibiofibulae. LENGTH PROX WIDTH DIST WIDTH PROX W/LENGTH Adults Juveniles adult or juvenile, by length or proximal width with confidence. (2) Adult bones of L. markhami are significantly smaller than L. waitomoensis and larger than other species except L. auroraensis, although the size ranges of juvenile bones significantly overlap those of other species. The relative width and robustness of the tibiofibula in both adults and juveniles distinguishes this species from all others although the difference is not significant when comparing L. auroraensis since there is only one specimen. (3) L. auroraensis is similar to L. markhami, but is relatively broader. (4) L. archeyi can be distinguished from all other species on size of adult bones alone (there were no juvenile bones of L. archeyi available); the mean proximal and distal widths were significantly different for all except L. hochstette1i. (5) Tibiofibula of L. hamiltoni and L. hochstetteri are not significantly different from each other but together are different from all other species. There is some suggestion that the tibiofibulae of L. hochstetteri are less robust than L. hamiltoni. The majority of the tibiofibulae described as "indeterminate extant species" are probably one of these two species. A discriminant function analysis to produce canonical functions gave good rcsolution of species groups (Fig. 37). The posterior probability of a bone belonging to its nominate species was calculated (Table 12). The multivariate analysis of the tibiofibula statistics confirms the original identifications of these bones except that the single specimen of L. auroraensis was elassified as L. markhami. Relationships between the Leiopelma Species Table 13 defines the characters which may vary within the genus, and the interspecific differences are summarized in Table 14, and converted into a matrix of similarity in Table 15. The six species fall into two groups, whose members share 80-90% of characters, but only 5-20% of characters are shared between groups. L. markhami and L. auroraensis are similar to L. hochstelteri and L. waitomoensis to L. hamiltoni and L. archeyi. These relationships are shown in the phenogram, Fig. 38.

39 238 Journal of the Royal ")ociet}' of New /;ca/rmd, Volume 17, 1 (}If 7 Table 11 - The relative significance of the dill'crence between species for j\l\'('niks and adults for length, proximal width, distal width and proximal width/length of tibiofibulae. * * * = significant': different P = 0.05, NS = Not Significant, = no comparison possible I.ENGTH PROX W IlIST W PROX wit. Species t JUV. adult JU\,. adult JU\. adult JU\,. adult 3-1 *** *** *** * * * *** *** * * * 3-2 NS 0JS * * * 3-5 *** *** ** * ;\!S 0JS 3-6 * ** *** *** *** * * * NS NS 3-4 *** NS 3-7 *** *** *** NS 1-2 ;\!S ;'\IS ;'\IS * * * 1-3 *** *** *** *** 1-5 NS *** * * * *** * * * 1-6 NS *** *** *** *** * * * 1-4 *** * ** 1-7 *** *** *** * ** * * * 5-3 *** *** NS ;'\IS 5-1 ;'\IS *** ** * *** * ** 5-2 NS *** *** 5-6 NS NS NS NS NS NS NS NS 5-4 ;\!S 5-7 ;'\IS NS NS NS NS NS ;'\IS NS 6-3 *** *** ** * *** NS NS 6-1 NS *** * ** *** *** 6-2 ;'\IS *** *** 6-5 NS ;'\IS NS NS NS NS NS NS 6-4 NS NS ;'\IS NS 6-7 ;'\IS NS ;\!S NS NS NS ;\!S NS 4-:) *** NS 4-2 ;'\IS *** *** 4-1 *** *** *** 4-5 *** NS 4-6 * ** NS NS NS 4-7 ;\!S 2-:) *** NS ;'\IS * ** 2-1 NS NS NS * * * 2-6 NS ** * 2-5 NS *** 2-4 ;\!S *** *** *** 2-7 NS *** t Species 1 = markhami; :3 = waitomoerili,; 4 = archeyi; 5 = hamiltoni; 6 = hachstelleri; 7 = Leiapelma extant species indeterminate. DISCUSSION Is there a case for neoteny in L. archeyi and L. hamiltoni? Neoteny or heterochrony is the condition in which the gonads develop to maturity before other parts of the body have reached adult size or character. As a group, the Anurans display neotenic arrest of ossification (Trueb, 1973). In comparison with more primitive amphibia, e.g. urodeles, the neurocranium is incompletely ossified, the quadrate and the pubis usually remain cartilaginous, and the dermal roofing bones are reduced. Stephens(ln (1960) suggested that the relative degree of neoteny could be used to distinguish L. archeyi from L. hamiltoni, on the following grounds. (1) In L. archeyi the exoccipital and the prootic are widely separated by wholly or (in one individual) largely uncalcified cartilage. In L. hamiltoni, these bones always fuse. (2) In L. archeyi, but not

40 Worthy - Osteolagy af Leiopelma ,,, III '1.,5'> /5 " ] ] ] ] 1 1 ] ] ] -5-5 Fig A canonical plot of two variables of the tibiofibula (X = length and Y = relative robustness). 1 = L. markhami; 2 = L. aurnram.li.l; :3 = L. waitomoensis; 4 = L. archeyi; 5 = L. hamiltani; 6 = L. horh.l/etteri. CUI Table 12 - The posterior probability of correctly classifying tibiofibulae into their correct nominate species from measurement data alone, showing the number of observations and relative percentages. Sec Table 3. NOMINATE SPECIES markhami auroraenslj waitomoensis archeyi hamilloni hachstetteri markhami o 0.00 o 0.00 o 0.00 o 0.00 waitomoensis o 0.00 o o 0.00 o 0.00 o 0.00 archeyi o 0.00 o 0.00 o o 0.00 o 0.00 hamiltani o 0.00 o 0.00 o hochstetteri o 0.00 o 0.00 o 0.00 o TOTAL in L. hamiltani, the quadrate and the articular remain cartilaginous. (3) In L. archeyi, ossification of the sphenethymoid was rare except in occasional large and relatively old specimens. In L. hamiltani, the sphenethymoid was always well ossified. (4) The relative degree of ossification of the vertebral column and other bones generally was greater in L. hamiltani, although there were no differences in structure of the head and skeleton. Trueb (1973) agreed that, although the two sexes of one species could display neoteny to different extents, no other interspecific differences such as reported by Stephenson are known. During this study further skeletal material of both species was examined, and the results suggest that the evidence for neoteny as a specific character should be reviewed. The degree of ossification in L. archeyi was the same as that observed in L. hamiltani (Fig. 39). The explanation of Stephenson's mistake may be the assumption that final skeletal ossihcation is achieved at sexual maturity. Observations of X-ray photographs of both L. hamiltani and L. hachstetteri, and dissections made before the skeleton was prepared, showed that sexual maturity precedes total skeletal ossihcation in both species. I see no reason why this pattern of development should not be repeated in L. archeyi. Perhaps

41 240 Journal of the Royal Socze{J! 0/ Nne' Zealand, Volume 17, 1987 Table 13-A list of os teo log-i cal characters that arc known to vary between species. (I), (0) label alternative character states for usc in the f()llowing phenetic analysis. They do not imply prilllitin' or derived conditions. ELEMENT CH/\R. :-JO. maxilla 1 premaxilla 2 frontoparietals :) squamosal 4 quadrat", 5 prevomer 6 parasphenoid 7 vert vert urostyle scapula 12 scapula:clavicle 13 proeoracoid 14 hulllerous radioulna 18 humerus/tibia 19 ilia 20 ischium 21 tibiofibula 22 2:l CHARACTER f(jrjti of the anterior lamina notched (1), unnotched (0) pars facialis central (1), anteriorly offset (0) lateral Hange well developed (1), not so (0) Length posterior process = Length anterior process (0) Length posterior process < Length anterior process (1) ossified (1), unossified (0) teeth on raised structure (0), not so (1) relative width "neck" narrow (1), wide (0) neural spine low (1), high (0) prezygapophyse large, oval (1), slllall and round (0) fna at right angles to shaft (1) sloped (0) lateral processes angled (1), evenly sloped (0) robust, dorsal flange present, anterior processes not divergent (1); less robust, no dorsal Hang-e, ant. processes divergent (0) 1:1 (1),1:1.25 (0) ossified (1) unossified (0) shaft anteriorly curved (0), straight (1) shaft cfoss-section angular (1), round (0) widest point distal end is above humeral ball (1), widest point distal end at midpoint of humeral ball (0) shaft ridged (1), unridged (0) 50-60% (0), 60-75% (1), >75% (2) distinct dorsal prominancc (1), not so (0) relatively wiele (1), narrow (0) shaft round (1), angled (0) ends relatively broad (1), narrow (0) Table 14 - Distribution of character states from Table 13 in Lez'opelma species. CHAR. No. markhaml fluroraenjls wailomoensz's archeyi hamdtonz' hochstelleri () () 0 I I I I () () 0 1 <) 1 0 () () 1 I () I I () () () I 1 () I 1 () () 0 I () lb I 2 () I () I 0 1 I () 0

42 Worthy - Osteology of Leiopelma ~ i;'t~,q,~' ~(:> 6fl1~,~ (j:~ 0<":<::-f? ~ 6v~Ov ~.~O ~C> ~. \ r:h~ Q,~' ~O 0 <...,0 0:.:j.& i;,f? > l- e = 60 E til ~ o 0/0 similarity Fig. 38-A phenogram based on the similarity matrix given in Table 15. Table 15 - A matrix of similarity for the six species of Leiopelma known from either subfossil or recent material. markhaml al1roraensl.\ waitomoensis archeyi hamiltani hachstetteri markhami 100 auroraen.hs BB waitomoensis archcyi hamiltani B hock,tetteri B B the specimens examined by Stephenson were sexually mature, but not adult in terms of skeletal ossification. There is therefore no evidence to support the idea that L. archeyi is neotenic in relation to L. hamiltuni. In all aspects of breeding, morphology and physiology these two species arc l:ssl:ntially similar. Thl: main difference is in the absolute sizl: of adult specimens (Stl:phenson, 1951; Bell, 1978; 1982; Cree, 1985). Interspecific relationships within the genus Leiopelma Stephenson (1955, 1960) was the first to consickr interspecific relationships in the genus Leiopelma. She concluded that L. archeyi was very similar to L. hamiltoni, and that each differed from L. hochstetteri to a similar degree. This arrangement was confirmed

43 242 Journal of the Royal Society oj New Zealand, Volume 17, 1987 Fig (A) Dorsal view of the skull and vertebrae of L. ardllyi (NMNZ AM 189) collected from the Tapu-Coroglen summit, Coromandel, showing extreme ossification. (B) Ventral view oftl1(' same specimen. Seale bars = (). 5 ern. by various studies on, ecology and reproduction (Bell, 1978, 1985), biochemistry (Daugherty et al. 1981; Daugherty e/ al. 1982), physiology (Cree, 1985), karyology (Stephenson e/ al. 1972, 1(74), and in this study (Fig. 38). Osteological comparison between Ascaphus and Leiopelma Ascaphus and Leiopelma share many primitive character states, and the basic arrangement of their cranial bones are the same, so they have often been included in one famiiy, "Ascaphidae". Detailed osteological studies have been largely confined to A scaplz us, and the conditions described have been accepted as, firstly, representative of the family, and secondly, as primitive. However the present study has revealed that the condition in Ascaplzus does not reflect that found in Leiopelma. The Skull Much attention has been given to the description of the chondrocranium in these genera (Pusey, 194:-); De Villiers, 1934a; Stephenson, 1951, 1955;), but relatively little to the ossified cranial skeleton. Noble (1931) provided the only available figure of Aswphus, a ventral view oj'the skull, so few comparisons of the shape of the cranial dements can be made from the literature. Leiopelma has paired frontoparietals that do not meet medially. The gap between them is constricted by a medial projection of the frontoparietal at a point equal to the anterior edge of the prootic. A similar condition is described in AscaphuJ (De Villiers, 1934a), but the two Jurassic ascaphids, Vieraella and Notobatrac/zus, have paired frontoparietals which meet medially, posterior to the anterior edge of the prootic (Estes and Rieg, 1973). In Leiopelma there is considerable intraspecific variation in the degree of ossification on the lateral and medial edges of the frontoparietal, which affects the outline shape of the bone and hence the extent of the gap, but there are always two gaps or fontanelles, the anterior one open anteriorally. Such variation is present in other frogs, e. g. Hyla lanez/ormis (Trueb, 1977). In frogs gcnerally, the persistence of these fontanelles is regarded, f()r no specific reason, as a pacdomorphic feature (Tyler, 1982). lfso, it implies tbat Leiopc!ma

44 Worthy - Osteology of Leiopelma 243 and the Jurassic species exhibit the primitive condition. But since there is a general trend among the Anura towards reduced ossification, the condition in Leiopelma is more likely to be the primitive state. No details of the structure of the premaxilla, maxilla, pterygoid and squamosal of Ascaphus are available. Both genera have an ossified quadrate. The parasphenoid of Leiopelma has lateral ala that extend the full width of the prootic to adjoin the pterygoid, whereas Noble (1931) figured Ascaphus with ala that underlie only half the width of the prootic. The cultriform process extends up to the prevomers in Leiopelma, but in Ascaphus it extends between them. In all Leiopelma species the dentigerous processes of the prevomers are aligned at right angles to the medial line and nearly adjoin. They have well differentiated anterior and posterior ala. The prevomers of Ascaphus appear to have very broad ala with only slight differentiation of anterior and posterior regions. Moreover the dentigerous processes are aligned quite differently and are well separated medially. Noble (1931) stated: Palates of modern amphibia are remarkable for the variability of the bones that occur there.... The prevomers may entirely disappear in some frogs while in others, such as Bombina, they may fuse to form a single element. Similarly Tyler and Davies (1979) recorded great variation and diversity of form for this element throughout the Anura. However within such diverse groups as Nyctimystes (Tyler and Davies, 1979), Litoria (Tyler and Davies, 1978), and Uperoleia (Tyler, Davies and Martin, 1981) the basic form of the prevomer and the orientation of the dentigerous ridges vary little. The number of teeth on the dentigerous ridge varies within a group according to body size. Leiopelma follows this trend, in that all species have the prevomers with the same orientation and basic shape. Variation is observed only on the dentigerous ridge. The Vertebral Column The vertebral column of Ascaphus was described by Ritland (1955). Ascaphus is most similar to L. hochstetteri in that the vertebrae have a low poorly defined neural spine and the sacral diapophyses are slightly expanded: L. hamiltoni and L. archeyi have a well defined neural spine and the sacral diapophyses are not at all expanded. The coccyx is too variable to make comparisons useful (Ritland, 1955). Pelvic Girdle The ilium in Ascaphus and Leiopelma is similar in form (Ritland, 1955); the ischium depicted by Ritland is relatively wide, similar to L. hochstetteri or L. markhami. The epipubis in Ascaphus is a dense fibrous connective tissue with no evidence for either calcification or ossification (Ritland, 1955). This, if true in Leiopelma, explains its near absence in deposits. However, the only one found did appear to be calcified tissue similar to that found in the prootic or exoccipial. Stephenson (1952, 1960) reported that the epipubis ultimately ossified in both L. hamiltoni and L. hochstetteri. Nobelian bones are considered to be one of the distinctive features of Ascaphus, absent in Leiopelma, so the discovery of an epipubis of L. markhami with structures suggesting that it supported Nobelian bones (or some similar bone) is very interesting. It suggests that L. markhami did in fact have structures comparable to Nobelian bones, and that these have been lost in the extant species. Hence this is probably a primitive feature which, along with others, indicates an ancient lineage although not necessarily any close affinity. Pectoral Girdle. Ascaphus and L. hochstetteri were compared by De Vos (1938a). He found that the ossification of the procoracoid and the fusion of this cartilage bone with the clavicula has proceeded further in Ascaphus than in Leiopelma. But in my material L. markhami and L. auroraensis are ossified to the same extent observed in Ascaphus (Ritland, 1955), although much less so in other species. De Vos (1938a) also found that the clavicula is a much larger bone in Ascaphus (clavicle/scapula ratio 3:1) than in Leiopelma and has developed its own marrow cavities only in Ascaphus. Subsequently, long clavicles have been assumed to represent the primitive state in the Anura (Kluge and Farris, 1969; Trueb, 1973). But this study has shown that Leiopelma has relatively short clavicles: L. hochstelteri, L. markhami and L. auroraensis have a clavicle/scapula ratio of 1: 1, and the other species, 1.25: 1. This study has documented several important variations in the shoulder girdle among Leiopelma species, and between Leiopelma and Ascaphus. (1) In Ascaphus De Vos (1938a)

45 244 Journal of Ihl' Royal SOCIl'ty of,vci ' Zealand, Volume] 7, 1987 and Ritland (1955) found that the scapula was uncleft and overlain dorsally by the clavicle, and subsequent authors, e. g. Trueb (1973), accepted both conditions as primitive: but Kluge and Farris (1969) found that there was too much intraspecific variation to interpret this condition, especially since some known Jurassic species had cleft scapulae. In Leiopelma, L. hochstetteri, L. markhami and L. auroraensis have the uncleft condition, and in the other three species it is cleft. The scapula does not overlap the clavicle in Leiopelma, Vieraella or Notobatrachus species, only abuts against it. (2) My data do not support Trueb's (1973) statement, that the clavicle does not articulate with the scapula in L. hochstetteri, because an ossified procoracoid intrudes between them: I found no evidence of this in the present study. In L. hoch,tetteri, L. markhami and L. auroracnsis the clavicle is often fused to the scapula in adults, although this fusion is strengthened ventrally by an ossified procoracoid. In the remaining Leiopelma species fusion of the clavicle and the scapula was never observed, nor was an ossified procoracoid. In Ascaphus, some of the discoglossids, Pipidae, and Rhinophrynidae, the distal end of the clavicle overlaps the scapula (Kluge and Farris, 1969). Phyletic relationships between Ascaphus and Leiopelma and within Leiopelma The phenetic similarity of the genera Ascaphus and Leiopelma was formally recognised by Noble (1924) when he included both in the family "Liopelmidae" because of shared unique but primitive characters. The relationship between Ascaphus and the Leiopelma species was not discussed, nor was the subject tackled in ensuing years. Most workers e.g. De Villiers (1934a, b), De Vos (1938a, b) and Stephenson (1951) were content to verify the similarity between them and thereby support the proposed familial relationship. Any attempt to determine phylogenetic relationships using cladistic techniques meets problems. No paraphyletic taxa are permitted, although Charig (1981) states that "to recognise a paraphyletic taxon is a very convenient way in which to categorize a 'stem group' or 'ancestral group' ". Paraphyly can not be ruled out, since there could be other recently extinct species, still unknown; but to forgo a cladistic analysis for this reason is nonsensical. As defined the Ascaphidae (sensu Trueb, 1973) is a family erected to include species having the most primitive character states known in the Anura, and species with the characteristics of this family are envisaged to have given rise to the rest of the Anura. If it is accepted that the Anura evolved from forms similar to present day amphibians, then the various characters that define the modern Anura are undoubtedly "derived" compared with those that define the other classes. However, evolution has proceeded to the point where the great majority of the Anura exhibit derived character states compared with the ascaphids and leiopelmatids. Therefore the extant leiopelmatid and ascaphid species represent taxa that have survived with the retention of relatively primitive anuran characters, and Ascaphus, Leiopelma, Notobatrachus and Viera ella comprise a paraphyletic group, defined by Trueb (1973) as the Ascaphidae. But I contend that there are good grounds for placing Leiopelma in its own family, the Leiopelmatidae. The most important consideration is that Leiopelma differs significantly from Ascaphus in the form of the shoulder girdle. Noble (1931) outlined an evolutionary trend wherein more derived groups, within the Anura, exhibit reduced ossification. At the family level there is also a trend toward overlap of the scapula by the clavicle (Kluge and Farris, 1969). This and the results of the present study lead me to suggest the following evolutionary trends within the pectoral girdle. (a) clavicle abutting scapula--> overlapping; (b) relatively long clavicle --> relatively short clavicle; (c) cleft scapula --> uncleft scapula. On these grounds, Ascaphus is derived for characters (a) and (c). Within Leiopelma, all species exhibit the primitive condition of character (a). But all are derived for (b) though within Leiopelma the L. hochstetterill. markhamzjl. auroraensis group has the greatest reduction. For character (c) this group is derived, but the L. hamiltonzjl. archeyill. waitomoensis group is not. However since these three characters are undoubtedly part of a single functional complex, the trends should be analysed together. The most parsimonious hypothesis is that initially the Anura had a cleft scapula with a long abutting clavicle. The trend is towards a reduced length of a functional clavicle, either by reduction in size of the clavicle (in Leiopelma) or by overlap of the scapula (in Ascaphus). Hence the structure of the pectoral girdle in the L. hamiltoni group is the most primitive as shown below.

46 ... hamiltoni ~hochstetteri type ~ Ascaphus -+ Other frogs Worthy - Osteology oj Leiopelma 245 This hypothesis partially contradicts the one erected on evidence of the operculum and the spiracle of larva (Kluge and Farris, 1969), in which L. hochstetteri is inferred to be the stem form from which one lineage leads to L. archeyi and L. hamiltoni, and a second to A.lcaphus and other frogs. The primary difference between these hypotheses is in which Leiopelma group is taken as representative of the primitive form, since both agree that Ascaphus is derived in comparison to Leiopelma, and is on the lineage to other frogs. Other evidence comes from the details of the cranial anatomy, particularly the form of the prevomer and the parasphenoid, and from the vertebrae. The fossil forms are unique among ascaphids in that Vieraella has a quadrateojugal and Notobatrachus has a middle ear. The frontoparietals of both these species have only an anterior fontanelle (Estes and Rieg, 1983). Leiopelma has been included as a member of the Ascaphidae primarily because it has nine presacral amphicoelous vertebrae, and has retained primitive tail-wagging muscles in the adult. But Leiopelma also has several characters which define it as a monophyletic group excluding Ascaphus: (1) frontoparietals with a central plus anterior fontanelle; (2) parasphenoid in which the width of the alae is equal to the length of the cultriform process; (3) prevomers which meet medially with teeth aligned at right angles to medial line; (4) absence of a quadrateojugal. Another problem with the cladistic method is the identification of polarity for characters. This is usually resolved by outgroup comparison. Since the Ascaphidae (sensu Trueb, 1973) has been defined by a set of characters which are supposed to be primitive (Griffiths, 1963; Kluge and Farris, 1969) there can be no justification in believing that the various included taxa are any more a sister group to Leiopelma than are any of the other families in the Anura. Hence there is no suitable taxon for use as an out group for Leiopelma, and no means of determining polarity of characters by this method. But some predictions can be made which could be tested if an ancestral species is ever discovered in New Zealand. Fossiliferous terrestrial deposits from the Tertiary (E. Fordyce, pers. comm. 1985) have recently been found, so perhaps this will only be a matter of time. Table 16 - Postulated polarity of characters in Lelopelrna and AscaphllS CHARACTER NO. PRIMITIVE DERIVED REASON Procoracoid Present Absent General trend to reduced ossification (Noble, 1931) Ischium 2 Wide Narrow Wide in discoglossids and other primitive groups (Kluge & Farris, 1969) (Ritland, 1955) Ilium :) Cylindrical Embellished Trueb, (1973) Clavicle 4, 5 Long, abutting short, over-lapping sec p. 244 Scapula 6 Cleft Uncleft see p. 244 Prevomer 7 Teeth and other paired, nearly meet sec p. 214 processes on same in midline, teeth on plane raised structure. Parasphenoid 8 Reduced ala well developed ala see p. 243 The characters, summarised in Table 16, can be coded using binary nomenclature with (1) representing the derived condition (Table 17), and from this a cladogram can be constructed (Fig. 40). L. arclteyi and L. hamiltoni are associated together, rather than either"with L. waitomoensis, because osteologically they are inseparable from each other

47 246 Journal oj the Royal Society oj New Zealand, Volume 17, 1WJ7 Table 17 - The distribution and polarity of characters among the various taxa of I"ciopdma (1) derived condition; (0) primitive condition. CHAR. No. markhami auroraensl,) u'aitomoensis archey1 hamiltoni hochs/cttfti () but each difkrs from L /L'aitulllol'm!, in the samc Illanner. This dariogralll therc/ore illlplil" a phylogeny which is mirrored bv the o\)stt\('(1 phenetic silllilarit\. Recent biwhelllic;t\ work has suggested that L. arclieri and L. 11IlIllil/llII/ separatnl thn-c to jellir lllillioll n'drs ago, and that the lineage leading to L Iwrli,/(//cri di\erged about 1 J lllillioll \TarS ag') (Daugherty r:i ai. 19i1 1; Daugherty ti Ill. 1 (Jil2). If the latter is a reliable date. thell ;111 the known Leiopelma species hah' radiated in the last 1') million years. i (' since til\" Miocene. Any f()ssil j<jriils f(llllld in T\'e\\ Zealand which arc ofgre<lftt age should e"lribit unikjrmly "primitin'" state.'i fell' characters I-J in Table 17. I conclude that the lilmilv Ascaphidae. sensu Trueh (En:;). is a paraphvktic assl'lllbbgc of primitive anurans. \IVithin this group the LeiojJl'lma specics arc it 1lI0lH)ph\lctic group. :\'0 dose relationship with Asm/JIIII., can be dclilonstrat('(1. Osteological colilpari'ioll'i support the suggestion that l"(i0/jc/iii(i should 1)(' retained ill the disti net Lim il v "Leiopellllat idal," (Green ei at. )l)ilo). as defined in {he generic diagnosis abo\"('. ~ -;<::-0 ~ ~O :5 0<::- 0' '(' 0,:<;:-0 ~",'ClJ,0 cy ~ 0" ClJ<::- ~o.,0 -:<::-ClJ~'.~O~ o <.,0 ~'" '('0 Fig A cladograrn depicting {he rclationshi ps of the Lrioprima species based on synapnlllorphics shown in Table 17 (characters l-il). 8

48 Worthy ~ Osteology of Leiopelma 247 The Ascaphidae were redefined from osteological criteria by Lynch (1973). The following points of his definition were either wrong, or exclude Leiopelma. "Sacral dispophyses broadly dilated." Not so in Leiopelma. "Scapula overlaid anteriorly by clavicle". Not so in Leiopelma. "Tarsal bones free". Fused in Leiopelma. "Quadratojugal lost in Ascaphus, not in Leiopelma". Lost in both. Zoogeography of Leiopelma and Ascaphus Most zoogeographers consider Leiopelma to be an archaic remnant of the New Zealand fauna, which originated in Gondwana (Fleming, 1979; Daugherty et al. 1981). Fleming (1979) argues that the fossils Vieraella and Notobatrachus from Patagonia are evidence for this. However, in New Zealand there is a remarkable lack of other members of the characteristic amphibian fauna of the constituent land masses of Gondwana, either fossil or recent. Other Gondwana fragments have much more in common. For example, Casamiquela (1961) noted the similarity of the Cretaceous pipid fauna of South America and Africa, and Warren (1984) the affinities of the Triassic labyrinthodont fauna of South America and Australia; the Hylidae and the Leptodactylidae of South America and Australia are considered to be sister groups (Tyler, 1984; White, 1984). This is not to mention non-amphibian groups, e.g. marsupials, fish, reptiles and birds (Cracraft, 1975). There are numerous other terrestrial faunal links; e.g. Onychophoridae (Fleming, 1979), midges (Brundin, 1966), and earthworms (Stephens, 1980). Australia has a rich fossil amphibian fauna, which dates back till at least the Triassic (White, 1984), but it does not include leiopelmatids. The lack of fossils is usually attributed to the poor fossil record, but it is harder to ignore the lack of other "Gondwanan" Anuran groups in New Zealand's fauna. There is now increasing evidence for the former existence of a continent termed Pacifica, first proposed by Croizat (1958). Geological evidence for it was presented by Nur and Avraham (1977), who suggested that it broke up in the Jurassic and frat,'ltients of it became accreted to Asia, North America, South America and New Zealand. Although New Zealand is generally considered to be derived soley from Gondwana (Fleming, 1979; Stephens, 1980), there are some advocates of a composite geological structure. Howell (1978, 1980) postulated a former source landmass, eastward of New Zealand, for one of the accreted terranes of New Zealand. Biogeographers, e.g. Craw (1983, 1985), also advocate the composite nature of New Zealand, from biogeographic evidence. They suggest that New Zealand has both Gondwana and Pacifica affinities, so for example the Gondwana distribution of the ratites (Fleming, 1979; Craw, 1985), is not in dispute. The distribution of Pacifica remnants and leiopelmatids (including Ascaphus) are essentially the same, so it might be more parsimonious to argue that leiopelmatids originated in Pacifica, not Gondwana; but since Ascaphus and Leiopelma are related only by common retention of primitive characters, such a suggestion is unfounded. There are however, other faunal relationships with a primarily Pacific distribution. There is increasing evidence that the New Zealand gecko fauna is very ancient (Towns et al., 1985). These geckos belong to the tribe Carphodactylini, which has a distribution oriented the western Pacific, in northern 'Australia, New Caledonia and New Zealand (Kluge, 1967). The New Zealand skink fauna, long considered to be of Pliocene/Pleistocene origin Crowns, 1974; Hardy, 1977), is now considered to be much older than that (Towns et al. 1985). The genus Leiolopisma has essentially a western Pacific distribution, which has been established "a very long time" (Gibbons, 1985). Henderson (1985) found that most of the New Zealand trichopteran fauna, which dates back to the Mesozoic, has Pacific-oriented relationships. Biogeographical evidence therefore supports the hypothesis that New Zealand is a composite land mass whose biota is derived from both Pacifica and Gondwana. Until the relationships between the ascaphids and leiopelmatids and the rest of the Anura are known, their interrelationships will be unresolvable. If (a) they are rdicts of a primitive amphibian fauna, then Leiopelma could equally well have originated in Pacifica

49 248 Journal of the Royal Society of New Zealand, Volume 17, 1987 or Gondwana. But if (b) they represent a monophyletic offshoot from the main stem of amphibian evolution, with present connections only at the family level, then a Pacifica origin is the best hypothesis. ACKNOWLEDGEMENTS I am primarily indebted to Dr Ben Bell, Victoria University of Wellington, who provided reference material and constructive comment and guidance throughout this study. The aid and cooperation of the Waitomo Caves Museum, the National Museum of New Zealand, and the Manawatu Museum was most appreciated. For advice and encouragement, I thank Dr Graham Hardy, Mr Richard Cassels, and Mr Sandy Bartle. Professor] ack Grant-Mackie of Auckland University Geology Department made available the extensive collection housed there. The Wildlife Service made available reference material for skeletonizing. The Lands and Survey Department permitted collecting in lands under their jurisdiction, and provided financial and other support for investigation of caves in Fiordland. The New Zealand Forest Service permitted me to collect from Honeycomb Hill Cave in Northwest Nelson State Forest Park. I am most grateful to the numerous friends and colleagues from the caving fraternity have helped me during this project, either by accompaniment on collection trips, or in making collections. Special thanks go to Mr Chris Templer who gave his collection of subfossil herpetological material to me. I am pleased to acknowledge the assistance in the preparation of the manuscript, of Mr Ross Renner in the statistical manipulation of data, Professor Jack Garrick for commenting on the systematics section, and John Casey for preparation of the photographs. Lastly this study would not have been possible without the assistance from my wife, Cathy. REFERENCES Bell, B. D., Observations of the Ecology and Reproduction of the New Zealand Leiopelmid Frogs. Herpetologica 34: New Zealand Frogs. Herpetojauna 14: Conservation Status of the Endemic New Zealand Frogs. In G. Grigg, R. Shine and H. Ehmann (Eds): The Biology oj Australasian Frogs and Reptiles, pp Royal Zoological Society of New South Wales, Sydney. Bell, B. D., Newman, D. G., and Daugherty, C. H., The Ecological Biogeography of the Archaic New Zealand Herpetofauna (Leiopelmatidae, Sphenodontidae). In G. Grigg, R. Shine, and H. Ehmann (Eds): The Biology oj Australasian Frogs and Reptiles, pp Royal Zoological Society of New South Wales, Sydney. Bull, P. C., and Whitaker, A. H., The Amphibians, Reptiles, Birds and Mammals. In G. Kuschel (Ed): Biogeography and Ecology in New Zealand, pp Junk, The Hague. Brundin, L., Transantarctic Relationships and Their Significance as Evidence by Chironimid Midges. Kungliga Sven.lka Vetenskapsakademiens Handlingen 4( 11): Casamiquela, R. M., Un Pipoideo Fosil de Patagonia. Revista del Museo dela Plata, Seccion Paleontologia, nueva Seria 4: Charig, A., Cladistics: A Different Point of View. Biologist 28: Cody, A., Ratite Hole, Canaan. NZ. Speleological Bulletin 6(109): Cracraft, J., Mesozoic Dispersal of Terrestrial Faunas Around the Southern end of the World. Memoires due Museum National dristoire Naturelle 88: Craw, R. C., Panbiogeography and Vicariance Cladistics. Are They Truly Different? Systematic Zoology 32: Classic Problems of Southern Hemisphere Biogeography Reexamined. Panbiogeographic Analysis of the New Zealand Frog, Leiopelma-The Ratite Birds and Nothojagus. Zeitschrift juer Zoologische Systematik und Evolutionsjorschung 23: 1-10.