HARVARD UNIVERSITY LIBRARY OF THE Museum of Comparative Zoology \ ,1 * \ % * M ft — L_f P\ l\6i^ COMPARATIVE OSTEOLOGY AND EVOLUTION OF THE LUNGLESS SALAMANDERS, FAMILY PLETHODONTIDAE DAVID B. WAKE MEMOIRS OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOLUME 4 October 25, 1966 COMPARATIVE OSTEOLOGY AND EVOLUTION OF THE LUNGLESS SALAMANDERS, FAMILY PLETHODONTID AE Eurycea longicauda, an adult female from Deputy Cave, Jennings County, Indiana. COMPARATIVE OSTEOLOGY AND EVOLUTION OF THE LUNGLESS SALAMANDERS, FAMILY PLETHODONTIDAE DAVID B. WAKE MEMOIRS OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOLUME 4 October 25, 1966 MEMOIRS OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES The MEMOIRS of the Southern California Academy of Sciences is a series begun in 1938 and published on an irregular basis thereafter. It is intended that each article will continue to be of a monographic nature, and each will constitute a full volume in itself. Contributions to the MEMOIRS may be in any of the fields of science, by any member of the Academy. Acceptance of papers will be determined by the amount and character of new information and the form in which it is presented. Articles must not duplicate, in any substantial way, material that is published elsewhere. Manuscripts must conform to MEMOIRS style of volume 4 or later, and may be examined for scientific content by mem- bers of the Publications Committee other than the Editor or be sent to other competent critics for review. Because of the irregular nature of the series, authors are urged to contact the Editor be- fore submitting their manuscript. 1 3 10 ) 23 J 4 ) AM i pi h 7 44.-J o David K. Caldwell Editor Communications concerning the purchase of the Memoirs should be directed to the Treasurer, and communications concerning manuscripts should be directed to the Editor, both in care of the Los Angeles County Museum of Natural History, Exposition Park, Los Angeles, California 90007. Printed by Anderson, Ritchie & Simon, Los Angeles, California CONTENTS Abstract 1 Introduction 2 Acknowledgments 4 Materials and Methods 4 Non-osteological Family Characters 5 Analysis of Variation 7 Osteological Characters 7 Premaxilla 7 Maxilla 13 Septomaxilla 14 Nasal 15 Prefrontal 15 Route of the Nasolacrimal Duct 18 Vomer 20 Frontal 22 Parietal 23 Orbitosphenoid 25 Parasphenoid 25 Occipito-Otic 25 Occipital Condyles 26 Suspensorium 26 Palatopterygoid 27 Squamosal 27 Opercular Plate 27 Mandible 28 Hyobranchial Skeleton 28 Larval structure 28 Adult structure 28 Ceratohyal 29 Cornua 29 Lingual cartilage 30 First basibranchial 32 Proportions of articulated elements 32 Second basibranchial 34 Calcification 34 Synthesis of hyobranchial characters 35 Dentition 35 Vertebral Column 37 Regional differentiation 37 Cervical vertebra 39 Trunk vertebrae 39 Centrum structure 40 Neural arch structure 41 Transverse processes 41 Caudal vertebrae 43 Ribs 43 Limb Elements 44 Mesopodial Elements 44 Digits 46 Systematic Summary 47 Family Plethodontidae 47 Subfamily Desmognathinae 47 Desmognathus 48 Leurognathus 48 Phaeognathus 48 Subfamily Plethodontinae 49 Tribe Hemidactyliini 50 Gyrinophilus 50 Pseudotriton 50 Stereochilus 51 Eurycea 5 1 Typhlotriton 5 1 Typhlomolge 5 1 Haideotriton 5 1 Hemidactylium 5 1 Tribe Plethodontini 5 1 Plethodon 52 Ensatina 52 Aneides 52 Tribe Bolitoglossini 52 Supergenus Hydromantes 53 Hydromantes 53 Supergenus Batrachoseps 53 Batrachoseps 53 Supergenus Bolitoglossa 53 Bolitoglossa 53 Oedipina 54 Pseudo eurycea 54 Chiropterotriton 54 Lineatriton 54 Thorius 54 Parvimolge 55 Evolutionary Relationships and Trends 55 Relationships 55 Major Familial Divisions 55 Relationships of the Desmognathine Genera 58 Relationships of the Plethodontine Genera 60 Tribal relationships 60 Relationships of the hemidactyliines 62 Generic groupings 62 Status of the genus Manculus 64 Status of the genera Haideotriton and Typhlomolge 64 Relationships of the plethodonines 67 Relationships of the bolitoglossines 67 Trends 73 Ecology and Life History 73 The ancestral adaptive zone 73 Plethodontid origins 73 Evolution within the adaptive zone 74 Occupation of new adaptive zones and subzones 76 Paedomorphosis 79 Paedormorphic influence and plethodontid evolution 79 The evolutionary significance of paedomorphosis 8 1 Biogeography and Phylogeny 84 The Fossil Record 84 Historical Backgrounds and the Primary Dichotomy 85 The Desmognathines 86 The Plethodontines 87 Tribal Origins 87 History of the Hemidactyliines 89 The First Terrestrial Encounter: The History of the Bolitoglossines 93 The Second Terrestrial Encounter: The History of the Plethodonines 99 Conclusions and Summary 102 Literature Cited 104 COMPARATIVE OSTEOLOGY AND EVOLUTION OF THE LUNGLESS SALAMANDERS, FAMILY PLETHODONTIDAE David B. Wake1 Abstract: Lungless salamanders of the family Plethodontidae comprise the largest and most diverse group of tailed amphibians. An evolutionary morphological approach has been employed to elucidate evolutionary rela- tionships, patterns and trends within the family. Comparative osteology has been emphasized and skeletons of all twenty-three genera and three-fourths of the one hundred eighty-three species have been studied. A detailed osteological analysis includes consideration of the evolution of each element as well as the functional unit of which it is a part. Functional and developmental aspects are stressed. A new classification is suggested, based on osteological and other char- acters. The subfamily Desmognathinae includes the genera Desmogncithus, Leurognathus, and Phaeognathus. Members of the subfamily Plethodontinae are placed in three tribes. The tribe Hemidactyliini includes the genera Gyri- nophilus, Pseudotriton, Stereochilus, Eurycea, Typhlomolge, and Hemidac- tylium. The genera Plethodon, Aneides, and Ensatina comprise the tribe Pleth- odontini. The highly diversified tribe Bolitoglossini includes three super- genera. The supergenera Hydromantes and Batrachoseps include the nominal genera only. The supergenus Bolitoglossa includes Bolitoglossa, Oedipina, Pseudoeurycea, Chiropterotriton, Parvimolge, Lineatriton, and Thorius. Manculus is considered to be congeneric with Eurycea, and Magnadig- ita is congeneric with Bolitoglossa. Two species are assigned to Typhlomolge, which is recognized as a genus distinct from Eurycea. No new information is available concerning Haptoglossa. Recognition of a family Desmognathidae is rejected. All genera are defined and suprageneric groupings are defined and char- acterized. Range maps are presented for all genera. Relationships of all genera are discussed. Plethodontid salamanders have undergone an adaptive radiation char- acterized by structural, functional, and ecological parallelism. The relatively primitive desmognathines and hemidactyliines have remained close to the an- cestral habitat, mountain brooks and streams in eastern North America. Evo- lutionary trends in the direction of abandonment of the aquatic larval stages and attainment of direct development are evident in desmognathines. The tribes Bolitoglossini and Plethodontini have direct development in terrestrial habitats. Terrestriality has stimulated evolution in these two successful groups and both have extended far beyond the ancestral range. Paedomorphosis has played a significant role in influencing evolutionary patterns within the family. In hemidactyliines, paedomorphosis has led to paedogenesis in four genera. Paedogenesis is a highly specialized but regressive adaptation in these groups. Paedomorphosis has profoundly influenced ter- restrial groups by allowing them to achieve reproductive maturity at early structural stages. The phenomenon has led to evolutionary progress and may have been the key to the success of the terrestrial groups. Ancestral plethodontids probably arose from a stock close to that which gave rise to ambystomatids. The subfamily Desmognathinae is a specialized early derivative of this stock. Probable desmognathine remains from Creta- ceous formations suggest that subfamilial differentiation was a Mesozoic event. The tribe Hemidactyliini is the most generalized group, both in struc- ture and ecology, and is probably close to the ancestral familial stock. The tribe Bolitoglossini may have been the earliest derivative of the plethodontine stock. It ranges to western North America, Europe, and South America, and is the only salamander group that penetrates the tropics. In addition it is the only group of plethodontids that no longer occurs in eastern North America. The final derivative, the tribe Plethodontini, probably differentiated in early Tertiary times. The group has extended its range to western North America but still has relatively primitive representatives in Appalachia. i Department of Anatomy and The College, The University of Chicago, 1025 East 57th Street, Chicago, Illinois 60637 1 2 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 INTRODUCTION Despite considerable progress in studies of popula- tion genetics and the processes of speciation, biolo- gists have made little headway in understanding the mechanisms of macro- and megaevolution. The goal of this comparative osteological study of the salamander family Plethodontidae is to elucidate problems relating to evolution above the species level. Salamanders are generalized, evolutionarily conservative tetrapods. They were chosen for study because available evidence indicates that evolution- ary rates have been relatively slow in the group. The family Plethodontidae includes a large variety of adaptive types among its diverse species. Many intermediate adaptive stages are extant, thus facili- tating analysis of major evolutionary events. Hope- fully, this study will provide groundwork for comprehensive analyses of evolutionary patterns in the family Plethodontidae, as well as in other groups. The amphibian order Caudata includes nearly 300 species of salamanders. Almost two-thirds of these comprise the family Plethodontidae. The family is distinguished from all others by absence of lungs combined with presence of a cutaneous depression (the nasolabial groove) extending from each nostril to the upper lip of transformed indi- viduals. The most comprehensive treatment of the family is Dunn’s (1926) pioneer monograph, which is now, unfortunately, badly out of date. While papers by Piatt (1935), Taylor (1944), Soler (1950), Tanner (1952) and others have pro- vided additional information, the introduction to “The Salamanders of the Family Plethodontidae” (Dunn, 1926) has been the best source of informa- tion concerning structure and phylogeny. The most recent general reviews of theories of familial evolu- tion are given by von Wahlert (1957) and Martof (1962). A major adaptive radiation has been the out- standing feature of the phylogeny of the family. Primitive and generalized plethodontids live in semiaquatic habitats and pass through extended larval periods during their ontogeny. Some species are completely aquatic, but advanced species range from semiaquatic to terrestrial, with many of the intermediate stages still represented by living species. Many advanced species have abandoned the aquatic larval stage and undergo direct terres- trial development; they occupy a large variety of terrestrial habitats, from cold and barren highlands to heavily forested tropical lowlands. Some of the terrestrial species are arboreal, and a few are semi- fossorial, but most are surface dwellers that take daytime refuge in burrows or under surface cover. Primitive species of plethodontids occur in east- ern North America, primarily in Appalachia. Ad- vanced species have spread to western North America, Europe, and Central and South America. Because of the variety of adaptive types among plethodontid species, the family is veritably a “liv- ing laboratory.” It is an ideal group in which to study patterns of vertebrate evolution, particularly speciation, mechanisms and patterns of adaptive radiation, and the origins of higher taxa. This investigation has several objectives. Skeletal variation is analyzed, particularly to elucidate evolutionary relationships on generic and higher levels. Implications for taxonomy and phylogeny are emphasized and the biogeographic history of the family is reanalyzed on the basis of these data. A functional and evolutionary approach is em- ployed to focus attention on factors involved in the invasion of new adaptive zones and subsequent adaptive radiations, that is, on major features of evolution in the family Plethodontidae. The following brief sketches of the plethodontid genera are presented to provide general background information for the reader. Salamanders of the genus Desmognathus (eight species), the dusky salamanders, are found in the eastern portion of North America from Quebec and Nova Scotia to the Gulf Coastal Plain. The stocky, agile members of this genus occur in a variety of habitats, but most are aquatic to semiaquatic in habits. The most aquatic species are the largest, and the smallest species are terrestrial and even semi- arboreal. All are quick, alert forms, adept at leap- ing and climbing. An aquatic genus, Leurognatlxus (one species), resembles the larger species of Desmognathus in habitus. It occurs in mountain brooks in the south- ern Appalachian Mountains. A recently discovered genus, Phaeognathus (one species), is a primarily terrestrial form apparently restricted to a small area on the Coastal Plain of Alabama. The single species, an elongate form with very short legs, lives in burrows. Gyrinophilus (three species) is a group of large, stout salamanders with elongated trunks and short, stocky tails. The species occur in springs, seepages, rivulets, and cave waters from Quebec to northern Alabama and Georgia. One species is permanently larval. The strikingly colored (brownish to bright red, with dark spots) members of the genus Pseudo- triton (two species) are very stout, short-legged species which occur in springs and seepages in eastern North American from New York to Flor- ida, and west to Ontario, Ohio, and Louisiana. 1966 EVOLUTION OF LUNGLESS SALAMANDERS 3 The nearly aquatic salamanders of the genus Stereochilus (one species) have elongate trunks and relatively short, stout tails. The poorly known genus inhabits swampy and slow-moving waters on the Atlantic Coastal Plain from Virginia to Georgia. The slender, active members of the genus Eurycea (twelve species) range over most of east- ern United States, with isolated populations on the Edwards Plateau of Texas. The salamanders occur in a variety of habitats, from stream sides and swamps to springs, caves, and subterranean waters. Some species fail to metamorphose and achieve sexual maturity in a larval state. The genus Manculus (one species) is discussed separately in the data sections of this paper, but is considered by me to be congeneric with Eurycea. The unpigmented, blind, cave-dwelling members of the genus Typhlotriton (one species) inhabit subterranean waters and water edge habitats in caves of the Interior Highlands. Typhlomolge (two species) and Haideotriton (one species) are permanent larvae that are blind and unpigmented. The larger, spindly-legged Typh- lomolge occurs in localized underground water systems at the edge of the Balcones Escarpment in Texas. Haideotriton, smaller and with shorter legs, is found in underground waters of southern Geor- gia and northern Florida. Hemidactylium (one species) is a diminutive, highly distinctive inhabitant of swamps, sphagnum bogs, and neighboring woods. It has a basally con- stricted tail, only four toes, and is strikingly colored (clear white ventrally, with large black spots). The genus ranges from southern Canada to the southern Appalachian Mountains, with several disjunct southern and western populations. Plethodon (eighteen species) is a large genus of terrestrial salamanders found in forest floor and woodland habitats. The salamanders vary consider- ably interspecifically, from relatively large, stout, short-bodied and long-legged species to relatively small, slender, elongate and short-legged species. The genus occurs in many areas of eastern and northwestern North America, and in the Rocky Mountains. Ensatina (one species) is a stout-bodied, long- legged terrestrial salamander that ranges from southern British Columbia to southern California. The genus Aneides (five species) is a group of terrestrial to arboreal salamanders found in the Appalachians, the high mountains of New Mexico, and along the West Coast of North America. Ter- restrial members of the genus resemble generalized species of Plethodon and have cylindrical trunks and relatively short limbs and digits. Arboreal species have flattened bodies and relatively long limbs and digits. All species have enlarged jaw musculature. The relatively short-bodied species of Hydro- mantes (five species) occur in terrestrial habitats in limestone areas in the mountains of California and in Europe. The salamanders are most com- monly found under rocks and logs, and in caves. Batrachoseps (three species) is a genus of small, attenuate salamanders with four toes and very long tails. All are terrestrial and utilize underground burrows. The animals are found along the West Coast of North America from northern Oregon to northern Baja California. There may be isolated populations in southeastern Alaska and on the Nevado de Colima, Jalisco, Mexico. The seven remaining genera are collectively termed the neotropical genera in this paper. Boli- toglossa (fifty-two species) is a large, diverse, wide- spread group of terrestrial to arboreal species. Included are species that are among the largest (B. robusta) and smallest (B. rufescens, B. orestes ) of terrestrial salamanders. Highland species usually have unwebbed to partially webbed hands and feet, while webbing is more extensive in lowland forms. The species occur from northeastern Mexico through Central America to central Bolivia, the mouth of the Amazon River in Brazil, and possibly as far southeast as southern Minas Gerais, Brazil (Wake and Brame, 1966). The genus Oedipina (sixteen species) is a group of highly distinctive, elongate species with extreme- ly long tails, very short limbs, and attenuated bodies. All are terrestrial. The genus ranges from southern Mexico through Central America to west- ern Colombia and northern Ecuador. Pseudoeurycea (twenty species) includes a rather variable group of terrestrial to semiarboreal sala- manders found from northeastern Mexico to south- ern Guatemala. Although most of the species are of relatively moderate size, the genus includes species that are probably the largest terrestrial salamanders (P. bellii, P. gigantea). A variable group of relatively diminutive sala- manders is included in the genus Chiropterotriton (sixteen species). Members of the genus are largely arboreal, but some occur in forest floor habitats and others live mainly in caves. The genus is found from northeastern Mexico through northern Cen- tral America, and also in Costa Rica. Lineatriton (one species) is a poorly known form from central Veracruz, Mexico. This terrestrial animal is a small, elongate form with extremely shortened limbs and a long tail. Members of the genera Thorius (nine species) 4 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 and Parvimolge (three species) are the smallest species of salamanders. These diminutive, attenu- ated salamanders are found in terrestrial and ar- boreal habitats in wooded highland areas. Thorius occurs in southern Mexico, north of the Isthmus of Tehuantepec, and Parvimolge is found in Veracruz, Mexico and in Costa Rica. ACKNOWLEDGMENTS This work is a revised version of a doctoral disser- tation completed at the University of Southern California. My chief debt of gratitude is to Jay M. Savage, an inspiring teacher, who suggested the problem and offered constant encouragement and guidance. My wife, Marvalee H. Wake, has read the manuscript, offered constructive criticisms, and given her continuing moral support. Arden H. Brame, Jr., John L. Mohr and George B. Rabb have participated in stimulating discussions con- cerning portions of the work, and have made many helpful comments. Charles F. Walker loaned me a very important collection of cleared and stained salamanders housed at the Museum of Zoology, University of Michigan. Others who have con- tributed to the study include A. P. Blair, James P. Bogart, Charles M. Bogert, Ronald A. Brandon, William E. Duellman, Robert E. Gordon, Alice G. C. Grandison, W. R. Eleyer, Robert F. Inger, James Kezer, Jens W. Knudsen, Alan E. Leviton, James A. MacMahon, Bernard S. Martof, Hymen Marx, Hermano Niceforo Maria, James A. Peters, Philip J. Regal, William J. Riemer, Francis L. Rose, Rob- ert C. Stebbins, Russell H. Tuttle, Thomas M. Uzzell, Jr., Barry D. Valentine, Ernest E. Williams, and Richard G. Zweifel. The many courtesies shown me by Thelma Howell, Director, Highlands Biological Station, during a summer of research at that institution are gratefully acknowledged. Tech- nical assistance has been provided by Addye C. Brown, Les Siemens and James Everett. Mrs. Eleanor Willis of the Department of Anatomy, University of Chicago, typed the manuscript. Laboratory and library facilities were provided by the Allan Hancock Foundation of the University of Southern California. The National Science Foundation generously supported my studies in the form of three Cooperative Graduate Fellowships. In addition certain aspects of the study have been supported by grants of the Society of Sigma Xi (to D. B. Wake), National Institutes of Health, Institute of Arthritis and Metabolic Diseases, Grant A-3549 (to J. M. Savage), and National Science Foundation Grant GB 3868 (to D. B. Wake). Funds from the United States Public Health Serv- ice General Research Support Grant FR 5367 to the University of Chicago and from the Dr. Wal- lace C. and Clara A. Abbott Memorial Fund of the University of Chicago have aided terminal portions of the work. MATERIALS AND METHODS This study is based on a morphological analysis of plethodontid salamanders, with an emphasis on osteology. Materials have been prepared for oste- ological examination in a variety of ways. Pre- served specimens have been measured and sexed, then cleared in potassium hydroxide, stained with Alizarin Red-S, and stored in glycerine. Measure- ments of size included in the text are the distance in millimeters from tip of snout to posterior angle of vent. Cleared and stained material has provided most of the skeletons used in the study. Many skeletons have been prepared by maceration and disarticulation, and many others by dehydration and subsequent cleaning by dermestid beetle larvae and sodium hypochlorite solution. Skeletons of every plethodontid genus have been utilized in the study, and include the following (number skele- tonized, number cleared and stained; all measure- ments and locality information on file) : Desmog- nathus aeneus 0, 2; D. auriculatus 1, 6; D. fuscus 1, 6; D. monticola 5, 7; D. ochrophaeus 1, 5; D. ocoee 0, 4; D. quadramaculatus 4, 12; D. wrighti O, 1; Leurognathus marmoratus 0, 8; Phaeognathus hubrichti 1, 1; Gyrinophilus danielsi 1, 6; G. pal- leucus 0, 1; G. porphyriticus 2, 2; Pseudotriton montanus 2, 3; P. ruber 12, 10; Stereochilus margi- natus 0, 3; Eurycea aquatica 0, 3; E. bislineata 2, 12; E. longicauda 2, 13; E. lucifuga 0, 1 ; E. multi- plicata 0, 2; E. nana 0, 5; E. pterophila 1, 3; E. tynerensis 0, 1 ; E. troglodytes 0, 1; Manculus quadridigitatus 0, 9; Typhlotriton spelaeus 0, 8; Typhlomolge rathbuni 0, 1; T. tridentifera 0, 3; Haideotriton wallacei 0, 2; Hemidactylium scuta- tum 0, 6; Plethodon cinereus 2, 8; P. dorsalis 0, 3; P. dunni 3, 4; P. elongatus 2, 9; P. glutinosus 1,11; P. jordani 5, 9; P. larselli 0, 1; P. longicrus 0, 1; P. neomexicanus 0, 3; P. ouachitae 0, 1; P. richmondi 0, 7; P. vandykei 0, 7; P. vehiculum 2, 8; P. wehrlei 0, 1; P. welleri 0, 6; P. yonahlossee 3, 10; Ensatina eschscholtzii 6, 7; Aneides aeneus 3, 10; A. hardii 0, 11; A. ferreus 3, 25; A. flavipunctatus 2, 13 ; A. lugubris 5, 12; Hydromantes brunus 0, 5; H. genei 0, 1; H. italicus 0, 1 ; H. platycephalus 0, 5; H. shastae 0, 1; Batrachoseps attenuatus 3, 26; B. pad ficus 2, 25; B. wrighti 0, 10; B. sp. nov. (Monte- rey) 0, 22; B. sp. nov. (Bodfish) 0, 1; Bolitoglossa adspersa 0, 7; B. borburata 0, 1 , B. cerroensis 0, 2; 1966 EVOLUTION OF LUNG LESS SALAMANDERS 5 B. colonnea 0, 1; B. dunni 0, 1; B. engelhardti 0, 2; B. flaviventris 0, 1; B. helmrichi 0, 1; B. lignicolor 0, 1; B. marmorea 0, 2; B. mexicana 0, 2; B. mono 0, 1 ; B. nigroflavescens 0, 1 ; B. occidentals 0, 1 ; B. orestes 0, 1; B. platydactyla 0, 3; B. robusta 0, 1; B. rostrata 0, 2; B. rufescens 0, 3; B. salvinii 0, 2; B. savagei 0, 2; B. striatula 1, 1; B. subpalmata 6, 13; B. schizodactyla 0, 1; Oedipina bonitaensis 0, 4; O. complex 0, 1; O. cyclocauda 0, 1; O. gracilis O, 2; O. inusitata 0, 5; O. longissima 0, 2; O. par- vipes 0, 3; O. poelzi 0, 4; O. pacificensis 0, 1; O. syndactyla 0, 2; O. uniformis 0, 1; O. sp. nov. (Costa Rica) 0, 5; Pseudoeurycea altamontana 0, 1; P. bellii 0, 1; P. cephalica 0, 2; P. expectata 0, 1; P. gadovii 0, 1; P. goebeli 0, 1; P. leprosa 0, 3; P. rex 0, 1 ; P. robertsi 0, 1 ; P. scandens 0, 1 ; P. smithi 0, 2; P. unguidentis 0, 1; P. werleri 0, 2; P. sp. nov. (Mexico) 0, 1; Chiropterotriton abscondens 0, 3; C. arboreus 0, 1 ; C. bromeliacia 0, 2; C. chiropterus 0, 4; C. dimidiatus 0, 5; C. multidentatus 0, 5; C. nasalis 0, 1; C. priscus 0, 1; C. xolocalcae 0, 2; Parvimolge richardi 0, 1; P. townsendi 0, 2; Linea- triton lineola 0, 4; Thorius dubitus 0, 1; T. mac- dougalli 0, 3; T. pennatulus 0, 2; T. pulmonaris 0, 5. Skeletal material has also been available for the families Hynobiidae (Hynobius), Cryptobranchidae (Cryptobranchus, Megalobatrachus), Sirenidae (Si- ren), Ambystomatidae (Ambystoma, Dicamptodon, Rhyacotriton), Salamandridae (Chioglossa, Cynops, Notophthalmus, Paramesotriton, Salamandra, Tari- cha, Triturus), Necturidae (Necturus), Proteidae (Proteus), and Amphiumidae (Amphiuma). Many specimens of most plethodontid genera have been dissected in order to study musculature and duct and nerve pathways. These specimens have been stored in seventy per cent ethyl alcohol and dyed with borax carmine-picric acid (for muscles and nerves) and toluidine blue (for carti- lage). All observations have been made with a stereoscopic dissecting microscope with magnifica- tions of from ten to ninety diameters. Histological preparations of the tongues of Batrachoseps pacifcus, Bolitoglossa subpalmata, Chiropterotriton chiropterus, Ensatina esch- scholtzii, Eurycea bislineata, Plethodon cinereus, Pseudotriton ruber, and Typhlotriton spelaeus have been available. Materials were preserved in 8% formalin, Bouin’s solution, or Zenker’s solution, embedded in paraffin, and stained with Azan using standard techniques. X-ray radiographs, mostly stereoscopic, have been prepared for many species, particularly of those not well represented in collections and not available for direct skeletal examination. Plates were exposed for from thirty seconds to ten minutes at instrument readings of twenty-five to forty kilo- volts and five DC milliamperes. Much osteological information has been obtained in this manner. All radiographs are on file at the University of Southern California. Extensive field experience in the Appalachian Highlands and eastern United States, on the Ed- wards Plateau of Texas, in the mountains of New Mexico, in the Pacific Northwest, in California, in Costa Rica, and in Peru has provided me with valuable background information concerning sala- manders and habitats. NON-OSTEOLOGICAL FAMILY CHARACTERS Lunglessness was first reported among plethodontid genera by Wilder (1894). Subsequently it has been discovered that members of other salamander families (Hynobiidae, Ambystomatidae, Salaman- dridae) may have reduced lungs, and a few species of these families lack lungs, but lunglessness is apparently a universal character of plethodontid species. A small depression, the nasolabial groove, ex- tends from each nostril to the lip in all fully meta- morphosed plethodontids. Associated with the groove is a system of glands, which, in many ple- thodontid genera, forms a swollen protuberance. The swelling is usually more obvious in males than in females, and may be drawn into a slender cirrus that extends ventral to the lower jaw. The above characters, in combination, define the family Plethodontidae. The family is comprised of over one hundred and eighty species distributed among the following genera: Desmognathus, Leu- rognathus, Phaeognathus, Gyrinophilus, Pseudo- triton, Stereochilus, Eurycea, Typhlotriton, Typh- lomolge, Haideotriton, Hemidactylium, Plethodon, Aneides, Ensatina, Hydromantes, Batrachoseps, Bolitoglossa, Oedipina, Pseudoeurycea, Chiroptero- triton, Parvimolge, Lineatriton and Thorius. The last seven genera are collectively termed the neo- tropical genera in this report. The species Eurycea quadridigitata is retained as the sole representative of the genus Manculus in the data sections only. Plethodontid genera vary considerably in their ecological requirements. They may be grouped as follows (in part after Dunn, 1926; von Wahlert, 1957): 1. Primarily or strictly aquatic. Leurognathus, Gyrinophilus, Pseudotriton, Stereochilus, Eury- cea, Manculus, Typhlotriton, Typhlomolge, Haideotriton. 6 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 2. Semiaquatic or semiterrestrial. Desmognathus, Hemidactylium. 3. Terrestrial. Phaeognathus, Plethodon, Ensatina, Aneides, Hydromantes, Batrachoseps, Bolito- glossa, Oedipina, Pseudoeurycea, Chiroptero- triton, Parvimolge, Lineatriton, Thorius. The above classification is over-generalized and there are many exceptions. Desmognathus alone has species that may be aquatic (quadramaculatus), semiaquatic (monticola), semiterrestrial ( ochro - phaeus), or almost terrestrial (wrighti), and several species of Eurycea are semiaquatic. Life histories of only a few of the many pletho- dontid species are known, but the general mode of reproduction of most genera is understood (see Dunn, 1926; Noble, 1927 b, 1931; von Wahlert, 1957). Aquatic larvae are known to occur in Desmognathus, Leurognathus, Gyrinophilus, Pseu- dotriton, Stereochilus, Eurycea, Typhlotriton, Manculus, Typhlomolge, Haideotriton, and Hemi- dactylium, and all have been available for study. Desmognathus exhibits considerable interspecific variation in length of the aquatic larval period. According to Organ (1961 a) D. quadramaculatus has the longest larval period, and the period is suc- cessively shorter in D. fuscus, D. monticola, and D. ochrophaeus. Young of D. aeneus emerge from the egg in a semimetamorphosed condition, with gills, gill slits, nasolabial grooves, eyelids, and a greatly reduced dorsal tail fin (Valentine, 1963 a). Reduc- tion of gills begins immediately, and transforma- tion is completed in less than a week. Finally, Organ (1961 a, 1961 b) has shown that D. wrighti has no free living larval stage, and hatchlings lack gills and tail fins. Most of the aquatic and semiaquatic genera have stream larvae with low dorsal fins, reduced gills, and well developed hind limbs. Pond larvae with high dorsal fins that extend onto the trunk and slowly developing, reduced hind limbs are found in Hemidactylium (Blanchard, 1923; Bishop, 1941), Manculus (Goin, 1951), and Stereochilus (Schwartz and Etheridge, 1954). Paedogenesis, or retention of larval non-repro- ductive morphology throughout life, is character- istic of Typhlomolge, Haideotriton, Gyrinophilus palleucus, and certain species of Eurycea (latitans, nana, neotenes, pterophila, troglodytes, tynerensis). Terrestrial direct development with no aquatic larval stage is characteristic of Plethodon, Aneides, Ensatina, Hydromantes, Batrachoseps, Bolito- glossa, Pseudoeurycea, Chiropterotriton, and Par- vimolge, and is presumed to occur in the terrestrial genera Oedipina, Lineatriton, and Thorius. It is likely that Phaeognathus, a genus with terrestrial adults (Valentine, 1963 c; Brandon, 1965), also has direct development. It has long been stated as fact that Hydromantes and Oedipus (= the neo- tropical genera) are ovoviviparous or viviparous (Dunn, 1926; Noble, 1931). Recently Gorman (1956) has reported that H. shastae lays eggs, and has shown that earlier reports of ovoviviparity in the genus are questionable. Peters (1863) stated that Bolitoglossa adspersa bears living young, and he was followed by subsequent authors. Niceforo Maria (1958), however, found that B. adspersa lays eggs. Brame and Wake (1963) reviewed liter- ature reports and showed that there is no evidence of ovoviviparity or viviparity in neotropical sala- manders. Young males of all plethodontids have unilobed testes, but in certain genera an emptied lobe is moved forward and a second lobe appears caudally, separated from the first by a short, slender, non- functional region. A new lobe is added in the second, fourth, and sixth years of sexual activity in Desmognathus, with the new lobes becoming func- tional in the third, fifth, and seventh years (Humph- rey, 1922; Organ, 1961 a). The maximum number of lobes is five (Humphrey, 1922). Multiple testes are of common and apparently constant occurrence in old adult males of Desmog- nathus, Leurognathus, Phaeognathus, Batrachoseps, Bolitoglossa, Oedipina, Pseudoeurycea, Chiroptero- triton, Parvimolge, Lineatriton, and Thorius. Other than in the above genera multiple testes have been encountered only in single specimens of Typhlo- triton spelaeus, Eurycea pterophila, and Aneides ferreus. Multiple testes may be used with caution as an indicator of relationships, and genera in which the character consistently occurs fall into two groups: ( 1 ) Desmognathus and its allies, and (2) Batracho- seps and the neotropical genera. The remainder of this report deals primarily with osteological data. Detailed descriptions of the skeleton of Aneides, including illustrations of skulls and other features, have been presented previously (Wake, 1963). Original observations concerning osteological characters in additional genera are to be found in many papers, including in particular the following: Baird (1951), Cope (1859, 1866, 1869), Dunn (1926), Emerson (1905), Hansen and Tanner (1958), Hilton (1945 a, 1945 b, 1945 c, 1945 d, 1946, 1947 a, 1948, 1959), Joubert (1961), Kingsbury and Reed (1909), Moore (1900), Monath (1965), Noble (1921), Parker (1877, 1882), Piatt (1935), Reed (1920), Smith (1920), Soler (1950), Taylor (1944), Wieder- 1966 EVOLUTION OF LUNGLESS SALAMANDERS 7 sheim (1875, 1877), and Wilder (1925). Informa- tion presented here is limited to those elements that demonstrate significant intergeneric variation. ANALYSIS OF VARIATION Osteological Characters Premaxilla Cope (1866) first separated genera with two pre- maxillae from those in which the bones have fused, and these characters were adopted by most sub- sequent workers. Dunn (1926) concluded that premaxillae are primitively separated in pletho- dontids. His conclusion was probably based on the fact that separated premaxillae are the rule in fishes and early amphibians and are present in all mem- bers of the primitive salamander families Crypto- branchidae, Hynobiidae, and Ambystomatidae. In 1943 Grobman reported the appearance dur- ing the ontogeny of Gyrinophilus of a suture be- tween previously fused premaxillae. Larvae have a single premaxilla, but during metamorphosis a suture forms. In 1959 Grobman noted that a single premaxilla is present in the larvae of Gyrinophilus, Pseudotriton, and Eurycea, and stated “this may be considered primitive in the Salamandroidea al- though it is undoubtedly a specialized situation in comparison with the conditions prevailing in the hynobiids and remote ancestral stocks.” Grobman suggested that fused premaxillae be regarded as the ancestral condition for these genera, and that fusion be considered less specialized than separa- tion. The condition in Gyrinophilus was considered an advance over that of the other genera, in which the premaxillae remain fused. Martof and Rose (1962) suggested that a single larval premaxilla is the rule in plethodontids, as earlier suggested by Grobman. They examined very small larvae of Gyrinophilus, Pseudotriton, Eu- rycea, Desmognathus and Leurognathus and found a single element with no evidence of two centers of ossification. Rose and Bush (1963) accepted the reasoning of Grobman (1959) that a single element is primitive in plethodontids and rejected Dunn’s (1926) proposal that paired elements are primitive. They stated that fusion or separation of parts of the larval premaxilla, a single dentigerous rod with two frontal processes, should be considered ad- vanced. Thus Eurycea, which usually retains a larval-like structure, was considered more primitive than Pseudotriton, in which the frontal processes fuse, and Gyrinophilus, in which the premaxillae separate. Plethodontid larvae of all stages have a single premaxilla which consists of a curved, toothed pars dentalis, two tiny palatal processes, and two ascend- ing and posteriorly directed frontal processes (Fig. 1). The latter arise separately and embrace the internasal fontanelle, and in all small larvae are separated for their entire length. This condition has been found in all larvae of Desmognathus, Leurog- nathus, Gyrinophilus, Pseudotriton, Stereochilus, Manculus and Hemidactylium. The condition is also common in Typhlotriton and Eurycea, but ex- ceptional individuals of E. nana and E. neotenes have frontal processes that fuse posteriorly, behind the fontanelle. A single Typhlotriton larva has a single premaxilla with frontal processes that are broadly fused behind the fontanelle. The typical larval condition is found in all Haideotriton and Typhlomolge. The life history of Phaeognathus is unknown, and all remaining genera undergo direct development. Figure 1. Anterior cranial elements of a small (23.9 mm.) larval Gyrinophilus danielsi. Drawn from a cleared and stained individual. Despite numerous complexities and considerable variation two premaxillary conditions are recog- nizable in adult plethodontids. In most, a single premaxilla is present and the frontal spines may be separated, fused in front of the fontanelle, fused behind the fontanelle, fused in front of and behind the fontanelle, or fused for their entire length. This general situation is found in Desmognathus, Leurognathus, Phaeognathus, Pseudotriton, Stereo- chilus, Eurycea, Manculus, Typhlotriton, Typh- lomolge, Haideotriton, Aneides, B olito glossa, Oedipina, Pseudoeurycea, Chiropterotriton, Parvi- molge, Lineatriton, and Thorius. The other general condition in which two premaxillae are present in adults and separated for their entire lengths is found in Gyrinophilus, Hemidactylium, Plethodon, Ensatina, and Hydromantes. Grobman (1943) has shown that small Gyrinophilus have a single pre- maxilla, and Martof and Rose (1962) report a single premaxilla in small Plethodon and Hemidac- 8 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 tylium. One element is present in the smallest Hydromantes (27.5 mm. standard length) exam- ined, but it is apparent that a suture is forming. Two premaxillae are present in all Ensatina examined (as small as 23.1 mm.), with no sign of fusion. Exceptional conditions are found in some in- dividuals of Typhlotriton, Chiropterotriton, and Thor ius. A tendency toward premaxillary separa- tion seems evident in large, old adults of Typhlo- triton. Average adults invariably have a single element, but with increasing size the toothed bar narrows on the midline. In one large adult (55.3 mm.) a distinct suture is present and two premaxil- lae occur. Curiously, there seems also to be a tendency toward fusion of frontal processes in large Typhlotriton, although the individual with two premaxillae had separated frontal processes as well. A single adult Chiropterotriton xolocalcae has two premaxillae which are joined by a single tooth pedicel, a situation also rarely encountered in Plethodon, which normally has separated elements. One adult Thorius pennatulus has two distinctly separated premaxillae. Ontogenetic series of these species have not been available for study. Specialized species of Batrachoseps ( B . attenua- te, B. pacificus, two species nova ) have a single premaxilla. In B. attenuatus, B. pacificus, and one of the new species (Bodfish), the frontal processes arise separately, then fuse almost immediately and remain fused as far as the anterior margin of the fontanelle, where the processes separate and di- verge. The processes fall short of the posterior margin of the fontanelle. A small fontanelle, often triangular but sometimes reduced to a tiny circle, is present immediately above the pars dentalis and below the point of frontal process fusion. In the other new species (Monterey) eleven of fifteen in- dividuals have the condition discussed above, but the other four have frontal processes that remain separated. The most primitive member of the genus, B. wrighti, has a single premaxilla in small stages, but the frontal processes are separated. That there is ontogenetic variation is evident. Two small individuals (21.8 mm., 29.1 mm.) have a single premaxilla. A median depression is present in the toothed bar of slightly larger individuals (32.4 mm.), and at later stages (36.3 mm., 38.3 mm.) the single element of small individuals is separated by a suture into two distinct halves, connected by the pedicel of a single median tooth. As individuals become progressively larger (above 40 mm.), two premaxillae are present with no sign of fusion. The larval and adult premaxillary conditions must be considered separately if a cogent theory is to be advanced to explain observed facts. In the Ambystomatidae two premaxillae are present in larvae of all genera. Maxillae appear earlier during ontogeny than in the plethodontids, and ai;e moder- ately well developed before metamorphosis proper is under way. They are present in the permanently larval ambystomatids. In plethodontids maxillae do not form until late in metamorphosis, and no maxil- lae are present in the permanent larvae. In amby- stomatids the premaxillae are attached to the skull proper by means of articulations with the frontals, the vomers, and the maxillae. In plethodontids the premaxilla projects forward from the skull proper, to which it is attached only by the tips of the frontal processes and by a tenuous articulation with the vomers (Fig. 1). Two premaxillae projected an- teriorly from the skull proper, lacking lateral articu- lations, are considerably less rigid than a single structure resulting from fusion at the anterior ex- tremity of the formerly paired structures. I suggest that the premaxillary condition observed in pletho- dontid larvae is an example of larval evolution, or caenogenesis (see DeBeer, 1958), which has re- sulted from selection operative at early larval stages of the ancestral stock. The larvae of plethodontids are carnivorous; they seize and manipulate prey by the closing of the premaxilla and dentaries on each other. The stronger the premaxilla, the more efficient would be the feeding process, and indi- viduals having a single stronger structure would probably be favored over those with paired, weak- er elements. The attainment of the single larval premaxillary condition has doubtless been of im- portance in the success of plethodontids. Metamorphosis is a period of drastic molding and reorganization in plethodontids (Wilder, 1925). During metamorphosis several skeletal ele- ments are lost, others are gained, and the remaining skull elements are extensively remodeled. In con- sidering the evolution of plethodontids it is thus essential that the total metamorphic process be considered apart from the changes that take place in any given element. While this subject is con- sidered in detail elsewhere (see below), it bears directly on the question of premaxillary evolution and is thus raised here. It has long been customary to assume that one can find a more primitive condition in an earlier rather than a later developmental stage of a given species, and it is apparent that Grobman (1943) and Rose and Bush (1963) have followed this reasoning in their work. Such practices have been decried by DeBeer (1958) who states numerous objections and offers many alternative suggestions to explain commonly encountered phenomena. In the case of the plethodontid premaxilla it is clear 1966 EVOLUTION OF LUNGLESS SALAMANDERS 9 that the metamorphic process must be considered apart from the element itself. One must distinguish between a primitive metamorphic pattern, and the presumed primitive nature of a retained larval character. Metamorphosis in primitive salamanders (hyno- biids, ambystomatids) results in paired, separated premaxillae. It is clear that such a condition is the end result of a primitive metamorphic pattern and should be considered primitive regardless of the condition present in larvae. Plethodontid larvae may be considered advanced over larvae of other families in the fused condition of their premaxillae. Premaxillary fusion in adult plethodontids is not primitive even though it is the condition of the larvae. There is considerable evidence that the primitive metamorphic pattern is retained in sev- eral primitive plethodontids, with resultant pre- maxillary separation. My interpretation of the events transpiring during metamorphosis in Gyrin- ophilus, for example, is that the primitive meta- Figure 2. Trends in premaxillary evolution. A. Typical pattern of plethodontid larvae and embryos ( Gyrino - philus danielsi), B. Adult pattern of primitive genera (Gyrinophilus danielsi ), C.-G. Adult patterns of ad- vanced genera (C. Desmognathus monticola, D. Eury- cea bislineata, E. Pseudoeurycea leprosa, F. Stereo- chilus marginatus, G. Oedipina parvipes). Not drawn to scale. morphic pattern is retained (Fig. 2). The pattern is similar in Hemidactylium, Hydromantes, Pletho- don, and presumably Ensatina. The metamorphic process is slowed in Typhlotriton, and is drawn out over a long period of an organism’s life, but the premaxillae of the largest individuals may finally separate. In regard to premaxillae the exceptional Typhlotriton can be said to undergo delayed meta- morphosis. Additional support is gained from the genus Batrachoseps. Four of the five species are specialized and demonstrate numerous paedomor- phic tendencies, but the fifth, B. wrighti, is less specialized and more primitive. Delayed metamor- phosis occurs in B. wrighti which results, in old adults, in two premaxillae while the other species never undergo premaxillary metamorphosis. The evidence cited above concerning Typhlo- triton and Batrachoseps suggests that fused pre- maxillae in adults of these genera (and possibly in Chiropterotriton and Thorius, see above) is the result of paedomorphosis. Paedomorphosis also accounts for the fused premaxillae in Typhlomolge, Haideotriton, Gyrinophilus palleucus, Eurycea pterophila, E. nana, and E. tynerensis. It is prob- able that the fused premaxillae in Pseudoeurycea, Chiropterotriton, Bolitoglossa, Oedipina, Parvi- molge, Lineatriton, and Thorius are due in part to paedomorphosis. All are affected by paedomor- phosis in other ways (see below), and the pre- maxilla is relatively very small in the group. A simplification of parts has taken place in the neo- tropical genera, however, and it is possible that an evolutionary trend toward simplification has been operative in unison with a trend toward meta- morphic delay or failure. The skulls of Pseudotriton, Desmognathus, Leurognathus, Phaeognathus, Stereochilus, and Aneides appear to have been secondarily strength- ened. All have solidly articulated skulls and all may have fused frontal processes. Fusion of premaxillae in adults of these genera is probably the result of selection in favor of increased strength of skull elements. A similar explanation can be advanced in regard to Eurycea and its derivative, Manculus. Eurycea may have arisen from an ancestor with a single premaxilla and a solid skull (Fig. 2). Skull weakening in advanced Eurycea is apparently a secondary event. In the Plethodontidae it can be assumed that larval and embryonic premaxillary fusion is the rule. The primitive metamorphic pattern results in two premaxillae. I consider any modification of this primitive metamorphic pattern to represent an evolutionary advance, whether the result of paedo- morphic influence or of selective pressures favoring 10 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 the strengthened premaxilla which results from fusion. In this discussion I have limited the topic to fusion or non-fusion of the toothed portions of the premaxillae. Fusions of the frontal processes occur only in forms in which fusion of the toothed portions has occurred. Several groups are recognizable on the basis of metamorphic effects on premaxillary structure, and details of their premaxillary anatomy are discussed below. The larval premaxilla is retained in rela- tively unmodified form in Typhlomolge and Haide- otriton. The primitive metamorphic pattern is re- tained in Gyrinophilus, Hemidactylium, Plethodon, Ensatina, and Hydromantes. The final group in- cludes genera that have some modification of the primitive metamorphic pattern, and may be sub- divided as follows: (1) Desmognathus, Leurog- nathus, Phaeognathus, (2) Pseudotriton, Eurycea, Mane ulus, Typhlotriton, (3) Stereochilus, (4) Aneides, (5) Batrachoseps, (6) Bolitoglossa, Oedi- pina, Pseudoeurycea, Chiropterotriton, Parvimolge, Lineatriton, Thorius. The skulls of Haideotriton and Typhlomolge are strongly depressed and the premaxillae are modified accordingly. The single pars dentalis is relatively broader than in larvae, and is a straight structure, posteriorly recurved only at the lateral tips. Frontal processes arise separately and extend almost di- rectly posteriorly with little dorsal rise. The proc- esses are very broad in Typhlomolge, narrower in Haideotriton, and in both embrace the internasal fontanelle. Frontal processes remain separate in the genera but may articulate behind the fontanelle. Palatal portions are poorly developed. In Gyrinophilus, Hemidactylium, Plethodon, and Ensatina frontal processes embrace a moderate to large fontanelle but remain separated. The processes are longest in Gyrinophilus where they expand distally and articulate with each other be- hind the fontanelle. Post-fontanelle expansion is less well developed in the other genera. The proc- esses may articulate posterior to the fontanelle in some species of Plethodon (see Wake, 1963), but not in Hemidactylium or Ensatina. A unique condi- tion is encountered in Hydromantes in which the frontal processes are extremely short, especially in the American species. The slender processes abruptly terminate at about the anterior margin of the nasals, well anterior to the posterior end of the fontanelle. In addition the anterior portion of the snout is relatively deep in Hydromantes and the frontal processes are almost vertically oriented. The situation in Hydromantes, more than any other plethodontid, resembles that encountered in primi- tive salamander families (Hynobiidae, Ambysto- matidae), and may reflect the relative antiquity of the genus and its proximity to the familial ances- tral stock. It is not firmly established, however, that short frontal processes are primitive. In the above group palatal processes are generally moderately developed, but the palatal shelf of Gyrinophilus is relatively broad. The premaxillae of Desmognathus, Leurogna- thus, and Phaeognathus have a characteristic struc- ture, one not closely approximated by other pletho- dontids. Perhaps the most diagnostic features are the compactness, the close articulation of the pre- maxilla with surrounding elements, and the close association of the frontal processes and toothed bar. The skulls of all are depressed, especially Leurognathus, and it is impossible to distinguish between pars dentalis and pars frontalis where the two areas join. The lateral margins of the pars dentalis extend posteriorly and a little dorsally and become the lateral margins of either pars frontalis. The frontal processes of larvae resemble those of primitive plethodontid genera and have distinct shanks and blades. During ontogeny bone is added to the lateral margins of the shanks and the proc- esses become relatively very broad. The distance across the shanks is only slightly less than the width of the toothed bar (Fig. 2). A small internasal fontanelle is embraced by the frontal processes and is increasingly reduced in size during ontogeny. In old adult Leurognathus the fontanelle is reduced to a tiny opening or is entirely closed (Moore, 1899). Frontal processes fuse behind the fontanelle in Leurognathus, Phaeognathus, and some Desmog- nathus (fuscus, auriculatus, ochrophaeus, monti- cola), while in other Desmognathus (aeneus, ocoee, wrighti) the processes articulate but do not fuse. The relatively primitive D. quadramaculatus has unfused processes for a long time (fused in indi- viduals more than 80 mm.). The posterior tips of the frontal processes in this group of genera are lateral in position, and following fusion the poste- rior border is concave. The palatal shelves are relatively the broadest and best developed in the family. Pseudotriton and Eurycea illustrate a tendency for fusion of the frontal processes posterior to the fontanelle. In the smallest Pseudotriton larva avail- able (17.2 mm.) the frontal processes have already encircled the fontanelle and are in the process of fusing. In more advanced larvae and in all adults the frontal processes are solidly fused. Typical un- fused frontal processes are found in Manculus and most Eurycea (tynerensis, longicauda, lucifuga, multiplicata). In the other Eurycea a trend toward fusion of frontal processes is evident. A single 1966 EVOLUTION OF LUNGLESS SALAMANDERS 11 E. nana, a single E. pterophila, and all available E. aquatica have fused processes. Rose and Bush (1963) report that three of five transforming E. aquatica and eleven of twelve adults had fused processes. Wilder (1924) reported that 4.5 per cent of 109 larval E. bislineata from Massachusetts had fused frontal processes. All E. bislineata larvae ex- amined by me have unfused processes. Rose and Bush (1963) report unfused processes in eleven of twelve E. bislineata (size not specified) from Ala- bama, Mississippi, and Louisiana. My data indicate an ontogenetic trend toward process fusion in the species. Processes are fused in four of six adults (37.3 - 39.5 mm.) from North Carolina, but are separated in the two smallest individuals (36.9, 37.2 mm.). Frontal processes are partially to totally fused in three adults (37.4-44.4 mm.) from Indiana, but are not fused in one large adult (42.5 mm.). It is obvious that geographic and ontogenetic variation occurs. An unusual con- dition is evident in the frontal processes of Pseudo- triton in that, following fusion, secondary growth occurs and a single process, usually with a bifur- cated tip, grows posteriorly. Thus the postfontanelle portion of the frontal processes is large relative to other genera. Palatal shelves range from small (Manculus) to moderately large (Pseudotriton) in the group. The premaxilla of Typhlotriton closely resembles that of Eurycea. Typhlotriton may have paired premaxillae, however, and in a sense it joins this group of genera with Gyrinophilus, which it resembles in many other features. Perhaps Stereochilus should be grouped with Pseudotriton, Eurycea, and Manculus. The entire skull of Stereochilus is rather compressed and elongate, and the skull elements, including the premaxilla, are modified accordingly. In adults, frontal processes are fused and arise from the pars dentalis as a single element. The processes separate at a moderate distance from the toothed bar to encircle the relatively small fontanelle. In small adults the processes completely encircle the fontan- elle and articulate with each other posteriorly. The processes are in contact for a considerable distance posteriorly but separate near their tips. In larger adults, the fontanelle is almost completely closed over. The frontal processes are completely fused anterior and posterior to the fontanelle and separate only at the posterior exeremity (Fig. 2). Palatal shelves are moderately large. Hemidactylium has a larval premaxilla very simi- lar to that of larval Eurycea. Adult premaxillae have relatively long toothed portions and broad, stout palatal shelves. In these features they resemble Eurycea and related genera. Resemblance to Pleth- odon is superficial, related to premaxillary separa- tion in organisms of roughly similar size. A single premaxilla occurs in all species of Aneides, but in other regards the element resembles the paired premaxillae of Plethodon (Wake, 1963). A distinct trend in the direction of increased strength of skull elements is evident in Aneides, and is reflected by the premaxilla. A single element appears to represent a specialization within the Plethodon-Ensatina- Aneides assemblage. An addi- tional modification of the premaxilla occurs in large adult A. lugubris, possibly the most advanced and certainly the most specialized species. The frontal processes are very long, equaling or exceeding those in any other genus, and tend to fuse behind the fontanelle. Secondary bony growth in lugubris roofs the fontanelle. The advanced species of Batrachoseps have a pe- culiar fusion of the frontal processes above the pars dentalis but below the fontanelle; this condition is not encountered in any genus discussed so far. The frontal processes are fused from their origin to the fontanelle in Stereochilus, but in Batrachoseps the processes arise separately, fuse for a very short distance, then separate and diverge posterolaterally around the fontanelle. Such a situation is found elsewhere among plethodontids only in the neo- tropical genera. In most plethodontids the fonta- nelle tends to be enclosed on all sides by the frontal processes, and a trend for fusion of the frontal processes posterior to the fontanelle is evident. In Batrachoseps and the neotropical genera fusion of the processes occurs in front of the fontanelle, if at all, and the fontanelle tends to be pushed pos- teriorly in reference to the premaxilla. The pro- cesses themselves are relatively shorter (relative to over-all skull dimensions) than in the more primi- tive genera, and there is a trend toward decrease in the over-all size of the premaxilla. The neotropical genera resemble the advanced species of Batrachoseps in premaxillary structure more closely than do any other plethodontid genera. The most striking difference between the two groups is the great reduction seen in the neotropical species, particularly in the pars dentalis. While a general reduction in size of the premaxilla has taken place in Batrachoseps, the reduction is most evident in the palatal and frontal processes, and the pars dentalis remains a relatively large structure. In the neotropical species the pars dentalis is ob- viously very shortened relative to the total size of the premaxilla. This character separates the neo- tropical genera from all other plethodontids. Palatal processes tend to be poorly developed or absent in the entire group. 12 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 The pars dentalis is always relatively small in the neotropical genera, but in specialized species of several genera (Chiropterotriton, Bolitoglossa, Oedipina ) and in all Thorius the portion is ex- tremely reduced. The extreme is reached in O. parvipes in which the length is as great as the breadth and the structure is barely large enough to support a single tooth pedicel (Fig. 2). In all Parvimolge examined the frontal processes arise separately and remain separated for their en- tire length. In all other neotropical genera at least some fusion is present. Pseudoeurycea normally has separated frontal processes, but a single P. leprosa has processes that arise separately, then fuse, and finally separate to diverge at the fontanelle. In P. smithi the processes arise separately in small individuals, but are fused from the pars dentalis to the fontanelle in large adults (Fig. 2). Frontal processes arise and remain separated in all Chiropterotriton examined with the exception of the specialized C. abscondens and C. nasalis. The processes of abscondens and nasalis have a common origin and remain fused until the fontanelle is reached at which point they separate and diverge. Frontal processes are solidly fused from their origin on the pars dentalis to the somewhat pos- teriorly placed fontanelle in Lineatriton. At the anterior margin of the fontanelle the processes sep- arate. The fontanelle is very narrow anteriorly, and the processes remain very close for about one-half the distance from the point of separation to the terminus. Most of the fontanelle is located pos- teriorly and is bounded by the anterior margins of the frontals. The pars dentalis of Thorius is very small but gives rise to paired, extremely slender frontal pro- cesses. In most individuals the processes arise sep- arately and remain separated for about two-thirds of their length before diverging around the pos- teriorly placed fontanelle. The processes may be fused for short distances in front of the fontanelle in T. pulmonaris and T. macdougalli. The frontal processes of most Bolitoglossa are separated for their entire lengths, and are relatively long and well developed. In some species (robusta, borburata) the processes embrace the fontanelle, and are rather broadly expanded posterior to the fontanelle. In others the processes may arise sep- arately, fuse at a point, then diverge around the fontanelle (adspersa, orestes, savagei, subpalmata). The most specialized premaxillary conditions are achieved in the genus Oedipina. The fontanelle has moved posteriad in reference to the premaxilla, and only the extreme anterior margin of the fontanelle is formed by the frontal processes. It is instructive to follow the ontogenetic changes in O. poelzi. In juveniles the frontal processes arise separately and remain separated. The processes are very close to each other for most of their length and diverge slightly only near the terminus. In small adult 0. poelzi the processes are fused from their origin for about one-third of their length. The processes of old adults are fused from their origins for from one-half to three-fourths of their length. In general the frontal processes in all Oedipina are fused for one-third to three-fourths of their length, less in young and more in old individuals. O. parvipes illustrates the extreme condition in which a single, totally undivided frontal process arises from the midline of the pars dentalis and proceeds poster- iorly to the anterior margin of the fontanelle, where it is only slightly expanded (Fig. 2). In summary, primitive premaxillae that separate relatively early in ontogeny are found in five rela- tively primitive genera of plethodontids, and in the most primitive species of the advanced Batracho- seps. Premaxillary fusion occurs in forms with strengthened skulls; fusion thus may either be the result of a strengthening adaptation in the species, or of derivation from a stock in which such an adaptation had occurred previously. Secondary fusions of the frontal processes also occur in some of these forms. Fusions related to skull strengthen- ing have arisen separately in groups containing the following genera: 1. Pseudotriton, Eurycea, Manculus, Typhlotriton, Stereochilus. 2. Aneides. 3. Desmognathus, Leurognathus, Phaeognathus. Premaxillary fusion is characteristic of paedo- morphic forms (retention of the larval premaxilla). It is difficult to trace the parallelism, but fusion must have occurred at least as many as four times: 1. Gyrinophilus palleucus. 2. Typhlomolge, Haideotriton, and paedomorphic species of Eurycea. 3. Neotropical genera. 4. Advanced species of Batrachoseps. My ideas concerning morphological trends and relationships of premaxillary structure are illus- trated in Figure 2. On the basis of total premaxillary structure it is possible to place plethodontid genera in the following groups: 1 . Gyrinophilus, Pseudotriton, Eurycea, Manculus, Typhlotriton, Typhlomolge, Haideotriton, Ster- eochilus, Hemidactylium. 1966 EVOLUTION OF LUNGLESS SALAMANDERS 13 2. Hydromantes. 3. Plethodon, Ensatina, Aneides. 4. Desmognathus, Leurognathus, Phaeognathus. 5. Batrachoseps, Pseudoeurycea, Chiropterotriton, Parvimolge, Lineatriton, Thorius, Bolitoglossa, Oedipina. It is significant that each of the above groups except 4 has at least one genus that includes species with the proposed primitive condition — unfused premaxillae. The fossil relative of Desmognathus and allies, Prodesmodon, had unfused premaxillae (Estes, 1964) ; all modern genera of Group 4 have greatly strengthened skulls and it is likely that the condition in Prodesmodon was primitive. Maxilla Facial lobes arise from the second quarter (from anterior) of the maxillary pars dentalis in most plethodontid genera, and the bulk of each lobe is anterior to the midpoint of the maxilla. The lobes of Batrachoseps arise from the anterior one-third of the pars dentalis, and the structures are well in advance of the midpoint. This forward movement is greater than in any other plethodontid, and is probably related to the general shortening of the snout characteristic of the genus. No obvious an- terior movement of the lobes is seen in other species with greatly shortened snouts (e.g., Bolitoglossa orestes). The only other genera in which the an- terior margins of the facial lobes approach the anterior tip of the pars dentalis are Desmognathus, Leurognathus, and Phaeognathus, in which the lobes are very large and extend posteriorly beyond the midpoint of the element. Facial lobes of the remaining genera arise well posterior to the anterior tips of the maxillae. The lobes are primarily anterior to the midpoint of the pars dentalis and generally terminate anterior to that point in Gyrinophilus, Pseudotriton, Stereo- chilus, Eurycea, Manculus, Typhlotriton, and Hemidactylium. The lobes are of moderate size in most of these, but are relatively small in Eurycea and Manculus. In Plethodon, Ensatina and Aneides the lobes tend to arise from the central one-third of the pars dentalis, and usually extend past the midpoint. The lobes are relatively small in Ensatina, larger in Plethodon, and largest in Aneides. Hydro- mantes has a moderate-sized facial lobe lying an- teriorly and terminating at about the midpoint of the maxilla. The relatively small facial lobes of the neotropical genera terminate at about the midpoint of the pars dentalis. Desmognathus, Leurognathus, and Phaeognathus have greatly enlarged facial lobes. The genera lack prefrontals and the space normally occupied by the prefrontals has been invaded by the maxillae. Simi- lar instances of bones enlarging into areas vacated by other elements have been discussed by Parring- ton (1956). The nasals are very small in these genera; the result of maxillary enlargement is that the facial lobes greatly exceed the nasals in size, a situation not found elsewhere among the pletho- dontids. In primitive salamander families (Hynobiidae, Ambystomatidae) facial lobes arise from the an- terior portion of the pars dentalis and terminate before the mid-point. The moderate-sized elements in a pre-midpoint position encountered in Gyrino- philus and others are generalized, and probably resemble the ancestral condition. The conditions in Batrachoseps (anterior movement), Desmog- nathus and its allies (enlarged size with very broad bases), and Plethodon and its allies (posterior movement) appear to be derived, specialized situations. Nasals articulate firmly and extensively with the pars facialis of the maxillae only in Desmognathus, Leurognathus, Phaeognathus, and Aneides lugu- bris. Moderate articulation is found in Gyrino- philus, Typhlotriton, and Stereochilus, and rather limited articulations are found in a number of genera (Pseudotriton, Eurycea, Plethodon, Aneides, Hydromantes, Batrachoseps, Bolitoglossa, Pseudo- eurycea, Chiropterotriton, Parvimolge, Lineatri- ton). In many of these latter genera articulation is limited, and may consist simply of a meeting at a single point. Primitively the space between the nasals and the maxillae is occupied by prefrontals, lacrimals, or both (Ambystomatidae, Hynobiidae), and extensive articulation is a specialized condition. In adult Gyrinophilus, nasals and maxillae are in contact, but prefrontals extend to the anterior mar- gin of the nasals below the articulation. Nasals usually do not articulate with the maxillae in Pseudotriton, and prefrontals extend between the elements to the anterior margins of the nasals. This situation is also encountered in some Eurycea, and in Manculus, Typhlotriton, Hemidactylium, most Plethodon, and some Aneides and Ensatina; it is probably close to the ancestral plethodontid condition. The maxillae are long, extending beyond the posterior margins of the eyes in Desmognathus, Leurognathus, Phaeognathus, Stereochilus, and Typhlotriton, and terminating at about the posterior margins of the eyes in some Bolitoglossa (e.g., mex- icana). Maxillae are shorter in other genera and terminate between the centers and posterior mar- gins of the eyes. In primitive salamanders (Hyno- 14 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 biidae, Ambystomatidae) the standard condition is elongate maxillae which extend beyond the eyes. The elongate condition in plethodontids may be a reflection of this ancestral pattern. A process arises on the inner margin near the tip of each long pars dentalis of Phaeognathus and extends posteromedially toward the quadrate, ter- minating a short distance before the quadrate is reached. This process is large, stout and expanded, and has grown in the direction of the jugal liga- ments, which are very short, dorsoventrally ex- panded and stout. Precursors of the process are evident as tiny projections extending out into the jugal ligaments in Desmognathus and Leurogna- thus. It is well known that osseous processes often develop at tension points, and the development of the processes in all three genera, and particularly in Phaeognathus, is an indication of the relatively great tension in the jugal ligaments and of the strength and rigidity of the skulls. The skull par- tially raises to open the mouth in these three genera, a unique situation in urodeles, and the development of processes is doubtless related to the new tensions resulting from this drastic functional modification. Desmognathus, Leurognathus, and Phaeognathus have greatly enlarged palatal shelves that are broadly overlapped by the vomers. The shelves are considerably larger than in any other genus, and are approached in width only by Stereochilus. Gyrinophilus and Pseudotriton resemble Stereo- chilus, but have relatively narrower shelves. Palatal portions of the maxillae of all other genera are narrower than in the above genera. The shelves are subject to considerable ontogenetic and inter- specific variation, however, and in very large adults of several genera (e.g., Aneides, Plethodon) may be relatively broad. Primitive salamanders seem to have relatively narrow palatal shelves, but broad shelves are found only in relatively primitive plethodontids. Most specialized and advanced plethodontids, such as the neotropical genera, have narrow palatal shelves. Several unique maxillary specializations are en- countered in the genus Aneides. These will be dis- cussed more fully in a forthcoming paper (see also Wake, 1960; 1963). The American species of Hydromantes (brunus, platycephalus, shastae) have high, sharp-pointed facial lobes with relatively narrow bases, while the European species (genei, italicus) have low, round- ed, broad-based lobes. As maxillae appear during metamorphosis, they are, of course, absent from the paedogcnetic groups. On the basis of total maxillary structure it is apparent that the plethodontid genera fall into two major groups; the first contains the genera Desmog- nathus, Leurognathus, and Phaeognathus, the sec- ond contains the remaining genera. Septomaxilla Small paired septomaxillae are primitively pres- ent in plethodontids but are absent in some ad- vanced groups and in those species which fail to complete metamorphosis (Typhlomolge, Haideotri- ton, Gyrinophilus palleucus, and paedogenetic Eurycea). Septomaxillae are greatly reduced in Desmog- nathus, Leurognathus, and Phaeognathus, and ap- pear as tiny crescentic ossifications with relatively great dorsoventral dimensions. Maxillae and nasals are well separated from the septomaxillae. Other temperate genera have relatively large septo- maxillae, and the bones are particularly well de- veloped in Hemidactylium, Plethodon, Ensatina, and Batrachoseps. In Gyrinophilus, Pseudotriton, Stereochilus, Eu- rycea, Manculus, Typhlotriton, and Hemidactylium the septomaxillae are particularly well separated from the maxillary facial lobes, with the least separation in Hemidactylium. Septomaxillae may be slightly covered by lateral portions of the nasals in these genera. Maxillary facial lobes overlap the septomaxillae slightly to considerably in Plethodon, and exten- sively overlap the bones in Aneides. Septomaxillae are well separated from the facial lobes in Ensatina, but in Ensatina and in A . aeneus specialized lateral nasal processes may overlap the septomaxillae. Maxillary facial lobes tend to overlap slightly or just contact the posterior margins of the septo- maxillae of Hydromantes, but extensively overlap the septomaxillae of Batrachoseps. Septomaxillae are of rather irregular occurrence in the neotropical genera, but have been observed in at least some individuals of all genera. There is a well marked evolutionary trend toward reduction and loss of the bones in these genera. When present the septomaxillae are small and well separated from the maxillary facial lobes and the nasals. Septomaxillae are best developed in some species of Chiropterotriton, and Rabb (1956) stated that presence of septomaxillae distinguished Chiroptero- triton from Pseudoeurycea. In a later paper Rabb (1960) wrote that septomaxillae are present only in those species north of the Isthmus of Tehuante- pec, but my observations conflict somewhat with his. The bones are present in the following species (N=north of Isthmus; S=south) : C. arbor eus (N), C. bromeliacia (S), C. chiropterus (N), C. 1966 EVOLUTION OF LUNGLESS SALAMANDERS 15 multidentatus (N), and C. priscus (N). Septo- maxillae are absent in: C. abscondens (S), C. dimidiatus (N), C. nasalis (S), and C. xolocalcae (S). Septomaxillae, not previously reported in Pseudo- eurycea, are present in some individuals of P. cephalicus and P. werleri, and Rabb (in. litt.) says the bones occur in western populations of P. leprosa. Very small septomaxillae are present in Parvimolge townsendi, but the bones are absent in P. richardi. Septomaxillae are reduced to tiny splin- ters of bone in Lineatriton. Bolitoglossa normally lacks septomaxillae, but small remnants are present on either side of one B. rufescens, on one side of another, and on one side each in one individual each of B. mexicana, B. platydactyla, and B. rufescens. Small septomaxillae are present in some Oedipina inusitata, but are absent in other species. Septomaxillae are absent in most Thorius, but are present on one side in one T. pennatulus. Hilton ( 1946 b) reported some indication of septomaxillae in the same species. Generally septomaxillae are well developed in primitive plethodontids. The bones are either re- duced or lost in three separate evolutionary lines: (1) Desmognathus and its allies (always present, but reduced) ; (2) the paedogenetic species (absent because species fail to metamorphose); (3) the neotropical species (either true evolutionary reduc- tion and loss or the result of paedomorphosis) . Nasal The nasals of plethodontids are subject to con- siderable interspecific and intergeneric variation, but it is difficult to group genera on the basis of nasal structure. The primitive nasal condition is probably close to that seen in Gyrinophilus, Pletho- don, and other genera in which relatively large nasals only slightly overlap the frontal processes and usually only slightly, if at all, articulate with the maxillae. The bones are somewhat variable in shape, but are usually quadrangular or broadly tri- angular. Anterior borders of the bones slope pos- terolaterally. Posterior margins are irregular, but often terminate in poorly to well defined points. Similar conditions are encountered in the primitive Ambystomatidae. Since nasals appear during meta- morphosis they are absent in the paedogenetic species. Martof and Rose (1962) report the ap- pearance of nasals earlier in the ontogeny of Pseudotriton than Gyrinophilus. The elements also appear relatively early in Eurycea, at a stage ap- proximating that of Pseudotriton. Nasals of Des- mognathus and Leurognathus appear late during metamorphosis. Nasals in Desmognathus, Leurognathus, and Phaeognathus are small and narrow, with a charac- teristic shield-like shape. The nasals are much smaller than the maxillary facial lobes and slightly overlap the premaxillary frontal processes. Lateral portions of the nasals are in very close contact with the maxillary facial lobes. Pseudotriton adults have nasals that broadly overlap the premaxillary frontal processes and may meet in the midline. Grobman (1959) stated that nasals develop in contact during metamorphosis and tend to separate during ontogeny; my observa- tions conflict with his and support Martof and Rose (1962) who said in regard to the nasals “they are closer in adults than in earlier stages.” Examina- tion of eight cleared and stained P. ruber ranging in size from 44.8 to 75.6 mm. indicates that nasals arise far apart and tend to grow slightly mediad during ontogeny. Only one individual (73.0 mm.) had the nasals in median contact. Grobman (1959) considered the Pseudotriton condition to be prim- itive for the family, but such a situation is not seen in more primitive families. Nasals are in contact medially in hynobiids, but the bones are distinctly different from those of Pseudotriton and are over- lapped by the premaxillary frontal processes. Re- cent work by Lebednika (1964) indicates that medial portions of the nasals of hynobiids represent vestiges of the postfrontal bones of fishes. No evi- dence of postfrontal contribution to the nasals of plethodontids was found, and the condition in Pseudotriton is probably specialized. Skulls of Pseudotriton are relatively solid, rigid structures; secondary medial growth of the nasals is clearly related to a strengthening trend. A parallel to the condition in Pseudotriton is seen in large Ensatina, and to a lesser extent in Gyrino- philus. The nasals undergo medial growth and over- lap the frontal processes. Similar, but less extensive, medial growth is encountered in other genera as well. All anterior elements of Manculus are reduced, and those of Thorius are very small and variably shaped. The strikingly reduced nasals of Thorius are in the form of slender crescents lying far pos- teriorly at the anterior margins of the orbits. Some of the large-nostriled species of Chiropterotriton (e.g., nasalis) have tiny, crescentic nasals. Prefrontal Prefrontals are primitively present in pletho- dontids, but since the bones appear during meta- morphosis they are absent in Typhlomolge, Haideo- triton, and the paedogenetic species of Eurycea and Figure 3. Dorsal and ventral views of skull of adult female Phaeognathus hubrichti. Cartilage stippled. Jugal ligaments lined. Posterior patch of vomerine teeth outlined. Line equals 5 mm Figure 4. Dorsal and ventral views of skull of adult male Stereochilus marginatum. Cartilage stippled. Line equals 5 mm. Figure 5. Dorsal and ventral views of skull of adult male Plethodon jordani. Cartilage stippled. Posterior patch of vomerine teeth outlined. Line equals 5 mm. Figure 6. Dorsal and ventral views of skull of adult female Bolitoglossa subpalmata. Cartilage stippled. Posterior patch of vomerine teeth outlined. Line eauals 5 mm. 18 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 Gyrinophilus. The bones also have been lost in other genera including Desmognathus, Leurogna- thus, and Phaeognathus, in which the prefrontal area is invaded by the maxillary facial lobes and the expanded anterior portions of the frontals. All Hydromantes lack prefrontals; the prefrontal region is invaded by relatively large nasals and, in the American species, by enlarged maxillary facial lobes. Prefrontal loss in Batrachoseps and various neo- tropical genera is associated with trends toward reduction and simplification of general skull struc- ture as well as paedomorphic trends. The bones are absent in all species of Batrachoseps except B. wrighti. Prefrontals are absent in small B. wrighti, but appear during ontogeny and are present as very small dots of ossified tissue in adults over 45 mm. (large for the species). Enlarged nasals occupy some of the prefrontal region in Batrachoseps. Among the neotropical genera prefrontals are present in all Pseudoeurycea (moderate-sized), Parvimolge (small), and Lineatriton (small). Pre- frontals are inter- and intra-specifically variable in Thorius, and are represented by tiny slivers of bone in some, but appear to have fused with the nasals in others. Rabb (1960) stated that prefrontals are present in species of Chiropterotriton which occur north of the Isthmus of Tehuantepec, but absent in species south of the Isthmus. Prefrontals are present in all northern species that I have examined, but are also present in the southern C. nasalis, and present or absent in the southern C. abscondens. The bones are absent in the southern C. bromeliacia and C. xolocalcae. Prefrontals are present in primitive Bolitoglossa, but in a number of advanced species they may be lost (colonnea, orestes, robusta), or fused with the nasals (adspersa, cerroensis). Hansen and Tanner (1958) report greatly reduced prefrontals in B. occidentalis and B. rufescens, but the bones are absent in my material. Intraspecific variation occurs in some species (savagei, subpalmata), and the bones may either fuse with the nasals or remain discrete. Prefrontals are absent in all species of Oedipina, and the nasals and maxillary facial lobes have grown into the vacated area. The remaining plethodontid genera have rela- tively well developed prefrontals. The highest de- velopment is achieved in Aneides, in which the specialized members of the lugubris group have very large prefrontals which interlock with the maxillary facial lobes (see Wake, 1963). Route of the Nasolacrimal Duct Eye glands are associated with the lower eyelids of metamorphosed salamanders, and the secretions of the glands are conducted to the external nares by the nasolacrimal ducts (Noble, 1931). The se- cretion is apparently an eye lubricant, and the excess drains through the ducts (Noble, 1931). Since eyelids, eye glands, and nasolacrimal ducts develop during or just before metamorphosis, all are absent in Typhlomolge, Haideotriton, and the paedogenetic species of Gyrinophilus and Eurycea. Nasolacrimal ducts of hynobiid and primitive ambystomatid (Dicamptodon, Rhyacotriton) sala- manders pass through horizontally oriented fora- mina in the lacrimal bones, elongate, longitudinally oriented elements located between the nasals and the maxillary facial lobes. Near their anterior ends the ducts pass through evacuations or grooves in the septomaxillae, and open in the nasal cavities. Relations of the ducts to bony elements have changed in advanced ambystomatids, in salaman- drids, and in plethodontids. Larsen (1963) has sug- gested that intercalation of a lacrimal between the prefrontal and the maxilla reduces stability of the maxilla on the neurocranium and that the lacrimals have been lost in higher groups due to selective pressures acting to fix firmly the maxillae to sur- rounding elements. Lacrimal loss has led to rela- tively extensive prefrontal-maxillary articulations in primitive plethodontids. In addition the pre- frontals of genera such as Amby stoma have taken over the function of the lacrimals, and the ducts pass through the prefrontals. The primitive nasolacrimal duct route of pletho- dontids is best seen in Gyrinophilus, in which the ducts extend from the eyes through the tissue of the canthus rostralis to the anterior ends of the nasal capsules. Anteriorly the ducts proceed ven- trad and open in the nasal cavities. The ducts leave no or only slight impressions on the underlying bones. Similar routes are followed in Stereochilus and Typhlotriton. Ducts of the remaining plethodontid genera have wholly or in part abandoned the primitive dorsal integumentary routes for routes closer to the cranial elements. Each prefrontal of Pseudotriton has a narrow anterolateral projection, similar to that of Gyrinophilus, which extends between or slightly ventral to the maxillary facial lobe and the nasal. The prefrontals differ, however, in that those of Pseudotriton have tips which are slightly to con- siderably evacuated. In addition each prefrontal has a distinct, lengthwise dorsal groove or trough. Nasolacrimal ducts pass from the eyelids over the prefrontals and along the depressions or grooves 1966 EVOLUTION OF LUNGLESS SALAMANDERS 19 to the anterior margins of the prefrontals. The ducts proceed ventrad through foramina, the walls of which are formed by the evacuated prefrontal tips on three sides, and by integument on the fourth. Paths of the ducts in Plethodon, Ensatina, and Aneides (except A. lugubris ) are similar to those in Pseudotriton. Relatively large prefrontals, evacu- ated anteriorly, extend from the orbits nearly to the nares; the ducts extend dorsally and anteriad across the bones, then ventrad through the anterior evacuations. The ducts pass posterior and then ventral to characteristic lateral nasal lobes in Ensatina, and the foramina are formed of nasal and prefrontal material. An extensive maxilla-nasal articulation effectively separates the prefrontals from the narial region in Aneides lugubris. The an- terior cranial elements of the species are covered with dense osseous accretions, and the ducts enter foramina running lengthwise in the accretions and proceed to the anterior ends of the prefrontals. There the ducts enter grooves in the thickened lateral margins of the nasals, and enter the nasal cavities at the anterior ends of the bones. The prefrontals in Eurycea and Manculus are moderate to small and lie farther back than in the genera discussed above. The bones are primarily associated with the anterior orbital margins, and, although they extend between the maxillae and nasals, are walled off from the osseous nasal aper- tures by lateral processes of the nasals. In both genera the prefrontals have pronounced anterior concavities, and, in some Eurycea, distinct laterally placed foramina. The concavities and/or foramina may be associated with small lateral concavities of the nasals, mid-dorsal concavities of the maxillary facial lobes, or both (see also Wilder, 1925). The nasolacrimal ducts extend across the prefrontals, and, in some instances, dorsal to the posterior por- tions of the facial lobes, before entering the for- amina. The point at which the ducts run ventrad is relatively posterior to that found in Pseudotriton. Processes of the nasals articulate with the max- illary facial lobes in Hemidactylium, but the pre- frontals are similar to those of Eurycea and bear pronounced anterolateral concavities. The ducts pass over the prefrontals, through the prefrontal concavities, then under lateral nasal processes. Prefrontals play no role in the route of the nasolacrimal ducts in the remaining genera of middle lattitudes. Prefrontals of Batrachoseps wrighti are extremely reduced, and do not extend between the nasals and maxillae. The ducts pass across the maxillary facial lobes and through foramina in the interspace between the maxillae and nasals. Duct routes are similar in other species of Batrachoseps, but small posterolateral nasal con- cavities are usually associated with foramina which are otherwise integumentary. Prefrontals are absent in Hydromantes, and the ducts extend across the maxillary facial lobes, then through foramina formed by concavities in the posterolateral boundaries of the nasals and facial lobes. In Pseudoeurycea, Parvimolge, Lineatriton, and primitive species of Chiropterotriton and Bolito- glossa the ducts pass dorsally over the prefrontals and enter foramina formed by anterior prefrontal, dorsal maxillary, and posterolateral nasal con- cavities. Nasals are reduced in size in C. nasalis, and the prefrontals bear distinct, fully enclosed foramina on their anterior extremities. In those Chiropterotriton and Bolitoglossa which have lost prefrontals, and in all Oedipina, the nasals and maxillae usually articulate extensively, and the ducts pass dorsad initially, then move ventrad through concavities in the posterolateral nasal mar- gins. Slightly to moderately concave dorsal margins of the maxillary facial lobes also contribute to the foramina. An extreme of the trend described above is reached in Thorius, in which the prefrontals are greatly reduced and play no role in the nasolacrimal duct routes. The nasals have been forced posteriorly by the shortening of the snout and enlargement of the nostrils, and now form the anterior margins of the orbits. Nasal-maxillary articulation is slight. The ducts initially extend anteriad from the eyelids in a dorsal position, but very soon enter partially closed foramina in the posterior orbital margins of the nasals and proceed ventrad to the nasal capsules. Presumably the most specialized plethodontid condition is that encountered in Desmognathus and allied genera. Wilder (1913) was unable to find nasolacrimal ducts in Desmognathus, and the ducts appear to be absent in Phaeognathus and Leurog- nathus as well. On the basis of the arrangement of the antero- lateral cranial elements, and the route of the naso- lacrimal ducts, the following generic groupings are evident: 1. Gyrinophilus, Stereochilus, Typhlotriton. 2. Pseudotriton. 3. Plethodon, Ensatina, Aneides. 4. Eurycea, Manculus, Hemidactylium. 5. Hyclromantes, Batrachoseps, Bolitoglossa, Oedi- pina, Pseudoeurycea, Chiropterotriton, Par- vimolge, Lineatriton, Thorius. 6. Desmognathus, Leurognathus, Phaeognathus. 20 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 The above groupings reflect the general trends from primitive dorsal to advanced ventral naso- lacrimal duct routes. This is a morphological trend that may reflect phylogeny. It is likely that the ten- dency for posterior movement of the nasolacrimal duct foramina has proceeded in parallel along several phylogenetic lines, with parallelism most evident between groups 3, 4, and 5. Vomer Vomers of two distinct types are encountered in larval and paedogenetic plethodontids (Fig. 7): 1. Vomers in shape of inverted L’s with rounded corners and varying amounts of expansion; ar- ticulation between bilateral counterparts limited to extreme anterior tips; elements very widely separated posteriorly; small anterior processes attach vomers to premaxilla. Gyrinophilus, Pseudotriton, Stereochilus, Eurycea, Manculus, Typhlotriton, Haideotriton, Typhlomolge, Hemi- dactylium. 2. Vomers in shape of upper quarter circles, with small posteromedial projections; virtually com- plete articulation of bilateral counterparts along midline; no anteriorly directed processes, and no attachment of vomers to premaxilla. Desmognathus, Leurognathus. Figure 7. Vomers and palatopterygoids in larval plethodontids. A. Desmognathine pattern ( Desmognathus quadramaculatus) , B. Plethodontine pattern ( Gyrinophilus danielsi). Not drawn to scale. Larval ambystomatid vomers resemble condition 1, and it apparently is the more generalized and primitive of the two plethodontid conditions. The larval vomers of Desmognathus and Leurognathus articulate extensively with each other. The most plausible explanation for this articulation is that it is a strengthening compensation for the loss of suspen- sorial articulation of the palatopterygoids (which see), with which the larval vomers articulate posteriorly. Plethodontid genera may be placed in three groups on the basis of adult vomerine structure. Gyrinophilus, Pseudotriton, Stereochilus, Eurycea, Manculus, Typhlotriton, and H emidactylium form one group that is characterized by: (1) presence of sharply arched vomerine tooth series, (2) bony vomerine growth posteriolateral to the tooth series (Fig. 8). Fontanelles are open and of moderate size in all genera, and vaulting is moderately to well developed in all but Typhlotriton. Preorbital pro- cesses are present primitively. The vomerine tooth series originate on the processes and proceed an- teriomediad almost to the midline where the series sharply turn and proceed initially posteromediad, finally posterolaterad, to terminate below the pos- terior portions of the parasphenoid. Such a pattern is characteristic of Gyrinophilus, Pseudotriton, Stereochilus, and Typhlotriton. All four have well developed and stocky preorbital processes that extend beyond the lateral margins of the internal nares, but fall far short of the vomerine body margins. Anterior and posterior portions of the 1966 EVOLUTION OF LUNG LESS SALAMANDERS 21 tooth series are continuous in the four genera. The bilateral series are joined for a short distance at about the level of the anterior margins of the orbitosphenoids in Stereochilus. Lateral margins of the vomerine bodies are projected a little posteriad, below the preorbital processes, in Stereochilus and are drawn into spinous, posterolaterally projecting processes that are diagnostic of the genus; pre- orbital processes are also directed rather strongly posterolaterad, and thus are not overlapped by the body processes. Figure 8. Vomers of adult plethodontines. A. Pletho- donine and bolitoglossine pattern ( Plethodon jordani), B. Hemidactyliine pattern ( Pseudotriton ruber). Arrow indicates bony growth behind tooth series. Not drawn to scale. In contrast to Stereochilus with its compressed skull, Typhlotriton has a broad, flat skull. In old adults the vomerine series of teeth are very sharply arched and the posterior portion of each series curves back on itself as it leaves the vomer proper. Thus the series are very widely separated in the interorbital region, and in the largest individuals come to lie, for a short distance, along the orbito- sphenoids rather than the parasphenoid. The series are usually sinuous, however, and curve back below the parasphenoid, where they extend to the pos- terior end of the parasphenoid. At the latter point the series proceed medially and approach each other on the midline. The pattern in Typhlotriton is unique both in the degree of separation of the series and their posterior extent. The body of each vomer is also unusual in the genus and bears a posteriorly directed process that forms the lateral margins of the nares. Eurycea and Manculus have short preorbital processes that fall short of the lateral edges of the internal nares. Metamorphosing individuals have tooth patterns similar to those described above, but adults normally have the posterior patches separated by toothless gaps from the arched an- terior series (usually continuous in E. aquatica; Rose and Bush, 1963). Hemidactylium has slender, toothless preorbital processes that extend to the lateral margins of the internal nares. The anterior tooth series are arched anteromedially, but to a lesser extent than in other members of the group. In very small individuals (16 mm.) the anterior and posterior tooth series are continuous, but in adults there is a toothless gaP- The posterior portions of the tooth series, or the posterior tooth patches, are widely separated past the parasphenoid midpoints in all members of the group. The posterior teeth are in relatively narrow series or patches which contain from two to six diagonal rows at the point of maximum width. Plethodon, Ensatina, Aneides, Hydromantes, Batrachoseps, Bolitoglossa, Oedipina, Pseudoeury- cea, Chiropterotriton, Parvimolge, Lineatriton, and Thorius form a second group. The farthest anterior extent of the vomerine tooth series is at the extreme lateral extremity of each series, or slightly medial to that point, rather than near the midline. The an- terior tooth series extend along the lateral and pos- terior margins of the vomers, and there is no posterolateral vomerine growth beyond the series (Fig. 8). Anterior and posterior toothed areas are normally widely separated, particularly in adults. Fontanelles are normally moderate to very large, and vaulting is moderately to well developed. Rela- tively slender preorbital processes that extend to at least the lateral margins of the internal nares are present in all genera of this group except Batracho- seps, Thorius, some species of Aneides (ferreus, flavipunctatus, lugubris), and some Chiropterotri- ton (bromeliacia, dimidiatus, nasalis). Small pro- cesses are evident in some Batrachoseps, but are virtually absent in most. The longest preorbital processes occur in Ensatina, where they extend beyond the lateral margins of the vomerine bodies. The posterior tooth patches may be continuous with the anterior tooth series in very young indi- viduals ( e.g ., Ensatina at 23.1 mm.), but toothless gaps are normally present in adults. The posterior patches vary from very well separated in Hydro- mantes and Batrachoseps, to moderately or slightly separated in most genera, to closely appressed but unfused in Thorius and some species of Bolito- glossa. In genera such as Pseudoeurycea and Chiropterotriton the tooth patches may be in con- tact anteriorly but well separated posteriorly. In general, patches in genera of this group contain more teeth than those of the former group, and vary from three to four rows at maximum in some species of Pseudoeurycea, to as many as ten to twelve rows in some species of Bolitoglossa and Plethodon. 22 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 Parasphenoids in the first two groups of genera extend dorsally above the vomers proper to points equal to, or somewhat in advance of, the anterior orbitosphenoid margins. The area of overlap is relatively great, and at no point do the parasphen- oids and vomers articulate squarely. This is also the condition encountered in primitive salamander families. In the third group (Desmognathus, Leurognathus, Phaeognathus ) parasphenoids are greatly shortened in old adults, and articulate squarely with vomerine posterior margins at points near the orbitosphenoid midpoints. This articula- tion is considerably more rigid than in other groups, and is a specialized condition associated with the trends toward skull strengthening in the third group. The specialization involves not only para- sphenoid shortening, but also relative vomerine lengthening. Thin portions of the parasphenoids in Desmognathus may extend fairly far anteriorly, dorsal to the vomers. Vomerine tooth pattern is highly distinctive in the third group. In young Desmognathus and many adults the anterior teeth are in short, arched series that do not extend onto the preorbital processes. Anterior vomerine teeth are lost in the larger in- dividuals, especially males, of some species (e.g., D. monticola, D. quadramaculatus). Anterior teeth are in relatively long, straight series in Phaeog- nathus, with slight anterolateral curves; the series form relatively very small angles with the skull axes, and are located considerably more posteriorly than in other plethodontid genera. The slender pre- orbital processes of Phaeognathus are toothless. Teeth are usually absent in Leurognathus, but may be present in some populations (Martof, 1962). In toothed individuals of Desmognathus and Leurognathus, and in Phaeognathus, considerable amounts of bony posterolateral growth occur beyond the tooth series in a manner similar to that encountered in genera of the first group. Vomerine bodies are relatively small and very flattened in Phaeognathus and Leurognathus, but the somewhat larger elements of Desmognathus are slightly vaulted. The vomerine bodies are broadly overlapped dorsally by the palatal shelves of the premaxillae and maxillae in Desmognathus and Leurognathus, but the vomers overlap the pre- maxillary palatal shelves in Phaeognathus. Fon- tanelle closure was reported in Leurognathus by Moore (1899), and was employed as a generic character by Dunn (1926) and Noble (1931). Distinct fontanelles are present in young individuals and in small adults, however, and are progressively closed over with increasing size. Dunn (1926) has stated that fontanelle closure is related to the fact that the highly aquatic Leurognathus has lost the internasal glands, which aid in food capture in terrestrial habitats. Fontanelles in Desmognathus, and to a lesser extent in Phaeognathus, may be closed over secondarily by medial growth of the fontanelle borders. These genera differ further from Leurognathus in that dorsal growth of the vomerine bodies forms bony walls around the posterior and posterolateral portions of the fontanelles, while the processes are absent in Leurognathus because skull depression places the premaxillae and vomers essentially in contact. The internal nares have shifted laterally in Leurognathus, and the preorbital processes are very closely applied to the vomer bodies, with only narrow lateral narial slits. Nares are in the normal position in Desmognathus and Phaeognathus. Pre- orbital processes of Phaeognathus and, especially, of Leurognathus are long and extend laterally beyond the vomerine bodies. The processes are shorter in Desmognathus, and lateral margins of the bodies grow posterolaterad to extend below the tips of the processes. Preorbital processes tend to articulate with the antorbital portions of the frontals in all three genera, a situation not en- countered in the other generic groups. Posterior vomerine patches are well separated from the anterior series and slightly separated from each other in adults of the third generic group. The patches are well developed, and have five or six diagonal rows of teeth. The anterior series and the posterior patches are continuous in late larvae and metamorphosing individuals in D. quadramaculatus. At these stages the anterior series are highly arched, and the apices are antero- medial in position. From each apex the series slope laterally and terminate before the medial margins of the internal nares are reached; in the other direc- tion the series extend posteromediad, then postero- laterad, and finally terminate on the posterior, ventral surfaces of the parasphenoids. The general pattern closely resembles that seen in adult Stereochilus. Frontal Frontals primitively are large and occupy most of the interorbital area. Ambystomatid and hyno- biid frontals are long, and extend beyond the posterior margins of the orbitosphenoids, and such a condition is probably primitive in the pletho- dontids. The evolutionary trend in plethodontids appears to be toward a decrease in frontal size, associated with an increase in the relative size of the parietals. Plethodontid genera may be grouped as follows on the basis of this character: 1966 EVOLUTION OF LUNG LESS SALAMANDERS 23 1. Frontals long, extend beyond posterior margins of orbitosphenoids. Desmognathus, Leurogna- thus, Phaeognathus, Gyrinophilus, Pseudotriton, Stereochilus. 2. Frontals short, fall short of posterior margins of orbitosphenoids. Eurycea, Manculus, Typh- lotriton, Typhlomolge, Haideotriton, Hemidac- tylium, Plethodon, Ensatina, Aneides, Hydro- mantes, Batrachoseps, Bolitoglossa, Oedipina, Pseudoeurycea, Chiropterotriton, Parvimolge, Lineatriton, Thorius. The genera may also be grouped on the basis of presence or absence of distinct lateral processes on the posterior margins of the frontals: 1. Frontals with medially sloping posterolateral margins; no distinct lateral processes. Des- mognathus, Leurognathus, Phaeognathus, Gy- rinophilus, Pseudotriton, Stereochilus, Typhlo- molge, Haideotriton. 2. Frontals with relatively straight or only slightly medially sloping posterolateral margins; poorly to very well developed lateral processes which overlap parietals on posterior margins. Eurycea, Manculus, Typhlotriton, Hemidactylium, Pleth- odon, Ensatina, Aneides, Hydromantes, Bolito- glossa, Oedipina, Pseudoeurycea, Chiroptero- triton, Parvimolge, Lineatriton, Thorius. Many exceptions are found among species of the genera of the last group, but every genus has at least some species that fit the category. The best developed processes are found in Eurycea, Man- culus, Hemidactylium, and the neotropical genera, and the processes are apparently all that remains of the greatly reduced frontals of Batrachoseps. In Typhlotriton the posterolateral borders slope medi- ally, and the small posterior processes are located more medially than in other genera. There is no indication of lateral processes in ambystomatids or hynobiids, and the posterolateral borders slope medially. Lack of lateral processes appears to be primitive in plethodontids. The lateral processes may compensate for general skull weakening, since they tend to be better developed in forms with weakly developed skulls (e.g., Manculus). Frontals of Batrachoseps and Oedipina have well developed anteromedial spinous projections that extend beside and below the premaxillary frontal processes. Anterior expansion of the frontals is marked in Hydromantes, and is also evi- dent in various species of many other genera. Notable frontal reduction occurs in two genera. In Batrachoseps frontals are narrow and are very widely separated posteriorly. The elements are es- sentially in contact between the anterior margins of the orbits, but rapidly diverge posteriorly, and leave much of the dorsal portion of the brain un- covered by bone. The frontals of Thorius are also reduced, and leave middorsal gaps that are con- siderably smaller than those of Batrachoseps. The most distinct group of plethodontid genera based on frontal structure contains Desmognathus, Leurognathus, and Phaeognathus, in which the bones are dense, stout, relatively thick, and en- larged. Anterior expansion is great. Ventrolateral extensions of the expanded regions have invaded the antorbital regions between the maxillary facial lobes, and the maxillary palatal shelves and vomer- ine preorbital processes. This feature is one on which Cope’s (1866) family Desmognathidae was based. The antorbital processes are specializations associated with skull solidification and strengthen- ing, and are found in no other plethodontid genera. The expanded anterior portions of the frontals have grown into the area vacated by the prefrontals, and are extensively overlapped by the maxillary lobes in a firm, extensive articulation. Posteriorly, along the orbital margins, distinct lateral ridges are present. Parietal Generic groupings are recognizable on the basis of parietal structure. The first group includes Plethodon, Ensatina, Aneides, and Hydromantes. In these genera the parietals are relatively flattened with broad, concave, posterior depressions (some large An eides), slight depressions (Aneides, Pletho- don, Ensatina), or essentially no depressions (Hy- dromantes). In all four genera the orbital edges of the parietals are laterally rather than ventrally di- rected, and the bones have no lateral aspects. The orbitosphenoids are moderately to slightly over- lapped by the depressed posterior portions except in Hydromantes. The general condition in these genera is similar to that found in primitive sala- manders, and is perhaps close to the ancestral condition of the family. The second group includes Gyrinophilus, Pseu- dotriton, Stereochilus, Eurycea, Manculus, Typhlo- triton, Typhlomolge, Haideotriton, and Hemidac- tylium. Parietals of this group have posterior depressions that vary from marked to shallow, and the bones have distinct lateral aspects with ventrally directed edges that significantly overlap the orbit- osphenoids in metamorphosed individuals. Pos- terior depression is marked in Hemidactylium, Stereochilus, Eurycea, and Manculus, but is poorly developed in Gyrinophilus, Pseudotriton, and 24 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 Typhlotriton. The relatively smooth parietals of Typhlomolge and Haideotriton resemble those of larvae of the other genera, but are somewhat better developed. Lateral processes and distinct, obliquely oriented crests are present posteriorly in Typh- lomolge. The posterolateral margins of the parietals are raised to form portions of the parietal-otic crests in Pseudotriton, Gyrinophilus, Stereochilus, and some species of Eurycea (aquatica, bislineata). A third group includes Batrachoseps, Bolito- glossa, Oedipina, Pseudoeurycea, Chiropterotriton, Parvimolge, Lineatriton, and Thorius. The parietals of these genera are either smooth or have slight posterior depressions. Well developed, lateral parie- tal spurs are present in all genera, and are diagnostic of the group (this structure is well illustrated by Wiedersheim, 1877, Fig. 9b, in Batrachoseps ). The spurs are small lateral processes that are di- rected sharply ventrally over the area of the ascend- ing processes of the suspensoria. It is possible that the spurs may have arisen as compensations for the generally weakened condition of the dorsal roofing bones in members of this group. The spurs appear to be neomorphic developments in these genera, and are probably not morphologically related to the lateral portions of the parietals of Eurycea and Manculus which may, in very small species, re- semble the spurs. Genera of the final group (Desmognathus, Leurognathus, Phaeognathus ) have highly special- ized parietals that differ markedly from those of the other groups. Parietals are divisible into two sections. The anterior sections are high and very flat, and are on the same level as the frontals. Neither frontals nor anterior portions of parietals are covered by musculature. The anterior portions articulate very firmly with the frontals. Pronounced lateral ridges are present, and each bone proceeds ventrad and slightly mediad from the ridge; thus there is a distinct lateral aspect to each bone. The posterior portions of the parietals have been im- pressively molded by the strong ligaments that ex- tend from the atlas vertebrae to the mandibles (Fig. 9), and are in the form of deep, concave, almost U-shaped troughs that are directed anterolaterad, into the orbits. The parietals and anterior portions of the occipito-otics are drawn into anterolateral orbital extensions of the troughs. The distinction between anterior and posterior portions of the parietals is strong in Desmognathus and Leurog- nathus, but is less well marked in Phaeognathus, in which the anterior portions are relatively lesser proportions of the bones and the parietals form significantly greater amounts of the troughs. In Pseudotriton and some Aneides ( lugubris group) the parietals tend to fuse medially in large individuals, and a median crest forms. Parietals are separated by medial spaces in larvae and paedo- genetic adults. Separation in adults of non- paedogenetic species ( Hydromantes , Batrachoseps, Thorius, Eurycea multiplicata, Bolitoglossa platy- dactyla) may be due to paedomorphosis. A huge Figure 9. Skull, cervical vertebra and first trunk vertebra of adult female Desmognathus ochro- phaeus (46.7 mm.). Note atlas-mandibular ligament, shape of cervical vertebra, and pterygapo- physes extending posterolaterally above postzygapophyses of trunk vertebra; all are characteristic features of the subfamily Desmognathinae. 1966 EVOLUTION OF LUNGLESS SALAMANDERS 25 space separates the parietals of Batrachoseps (smaller in wrighti than in other species), and moderate spaces are present in Thorius. The state- ment by Cope (1889) that the parietal bones of Oedipina are “two small oval lateral scales” and are unossified is incorrect. The bones are well ossi- fied and completely roof the posterior portions of the brain in Oedipina. Orbitosphenoid Orbitosphenoids are absent in Typhlomolge and Haideotriton, and in small larvae of other genera. They are present in large larvae, and in the paedo- genetic species of Eurycea. Well developed orbito- sphenoids are present in adults of other genera, except in Thorius, in which the elements are small and do not articulate dorsally with the frontals and parietals. The optic fenestrae are always enclosed in bone and are located from near the midpoints of the bones (in some Aneides ), to very near the pos- terior margins. The fenestrae in Typhlotriton have been reduced to minute foramina which reflect the reduced nature of the visual system, including the drastic reduction of the optic nerves, in the genus. Parasphenoid Plethodontids may be placed in two groups on the basis of parasphenoid structure. The first group contains Desmognatlms, Leurognathus, and Phae- ognathus. The parasphenoids are shortened and fall short of the anterior margins of the orbito- sphenoids. In addition the anterior portions are truncate and strongly concave to U-shaped. The bones articulate along their anterior margins di- rectly with the vomers in moderately firm sutures. The posterior portions of the parasphenoids are very broad with large posterolateral processes that articulate directly with the enlarged quadrates. The parasphenoids are not notably shortened in the second group (all other plethodontid genera), and extend to the anterior margins of the orbito- sphenoids or beyond. The anterior portions are moderately concave (but never U-shaped) to al- most flat. Anteriorly the bones overlap the vomers, and posteriorly parasphenoids fail to articulate with the quadrates. Broad, posterolateral wings, which fall short of the quadrates, are present in Stereochilus, and somewhat smaller processes are present in Pseudotriton. In Gyrinophilus rounded processes are present, but they are less well de- veloped than in Pseudotriton. The processes are faintly indicated, or absent, in all other genera. Anterior portions of the parasphenoids are par- ticularly broad in Typhlotriton, and, while less broad, are noticeably expanded in Gyrinophilus, Pseudotriton, Stereochilus, Eurycea, Manculus, and Hemidactylium. Broad anterior parasphenoid tips are typical of plethodontid larvae, and are found in adult Typhlomolge and Haideotriton. A tendency toward tip narrowing is evident in the remainder of the genera, but the tips may be slightly expanded in some species of Pseudoeurycea and Chiroptero- triton, and relatively broad in large species of other genera (e.g., Bolitoglossa robusta, Plethodon yon- ahlossee). An extreme of reduction is achieved by Chiropterotriton abscondens, in which the anterior tips are narrowed to sharp, shortened points in front of which the orbitosphenoids articulate. Broadened posterolateral parasphenoid pro- cesses are found in primitive salamanders (hy- nobiids, ambystomatids) in which they do not ordinarily articulate with the quadrates. The con- dition encountered in Stereochilus appears to be the most primitive in the family, and two major types of modifications have occurred. Hypertrophy of the processes has taken place in the first group of genera, and reduction and eventual loss of the processes occurs in most members of the second group. The truncated anterior parasphenoid tips encountered in the first group are unlike those found in other salamander families, and the mod- erately long, overlapping tips of the second group of genera are apparently close to the primitive condi- tion. Furthermore, the broad or expanded para- sphenoid tips of some genera (e.g., Typhlotriton, Gyrinophilus ) resemble those characteristic of hynobiids, ambystomatids, and salamandrids, and probably represent the primitive plethodontid condition. Occipito-Otic Otic capsules of plethodontids are considerably variable in size, but genera cannot be grouped on that basis alone. In some the otic capsules are noticeably large (Ensatina); in others they are greatly reduced (Oedipina). Most genera have cap- sules of moderate size. Distinctive otic crests allow some groupings to be made. Desmognatlms, Leurognathus, and Phae- ognathus have posteromedial-anterolaterally orient- ed crests that extend along the posterior margins of the otic-parietal depressions. The crests are prominent and relatively high, and are formed ex- clusively of otic capsular material. No other genera have crests even remotely similar. Double crests, usually separated, are found in Gyrinophilus, Pseudotriton, Stereochilus, and some species of Eurycea (aquatica, bislineata). The an- 26 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 terior ciests are formed of parietal and otic projec- tions, and more posterior, lateral crests are formed of otic and squamosal projections. The crests are about equal in over-all height and extent, and are relatively isolated with rather limited bases. In all four genera the crests tend to be spine-like, and are directed laterally or slightly posterolaterally. Otic and parietal contribution is about equal in the an- terior crests of Gyrinophilus and Stereochilus, but parietal participation is sharply decreased in Pseu- dotriton and Eurycea. Otic and squamosal partici- pation is about equal in the lateral crests of Gyrinophilus and Stereochilus, while squamosal participation is sharply increased in Pseudotriton and Eurycea. The crests reach a maximum develop- ment in Pseudotriton and Gyrinophilus. In Gyri- nophilus one or two small projections may be present in the region between tbe two crests, and the crests may be joined to form high, curved lamina in large Pseudotriton ( circa 90 mm.). Distinctive high, wing-like otic crests are present in most species of Aneides (Wake, 1963). The crests are extensive and are oriented in a postero- lateral-anteromedial direction. Posteriorly the crests are continuous with long, low, posteriorly directed squamosal crests. Moderate crests similar to those in Aneides are present in Pseudoeurycea smithi. The crests have parietal rather than squamosal participation, and small, accessory, anterior crests proceed over the anterior semicircular canals perpendicular to the main crests. Slight otic crests are present in some other species of Pseudoeurycea (bellii, goebeli). Small anterolaterally oriented crests, or antero- laterally directed small projections, are found in some individuals of Plethodon, Ensatina, Batracho- seps, Chiropterotriton, and Bolitoglossa. Unique, laterally oriented and dorsoposteriorly directed crests are present in Plethodon elongatus (Wake, 1963). All crests are apparently specializations related to provision for additional space for the origin and insertion of enlarged amounts of mandibular ad- ductor musculature, and are most prominent in forms in which the jaw musculature is greatly enlarged. No otic crests have been observed in Manculus, Typhlomolge, Typhlotriton, Haideo- triton, Hemidactylium, Hydromantes, Oedipina, Parvimolge, Lineatriton, or Thorius, and crests are absent in most species of Batrachoseps, Eurycea, Chiropterotriton, and Bolitoglossa. Occipital Condyles Plethodontid genera may be placed in two groups on the basis of occipital condyle structure. Sessile condyles, attached directly to the lateral margins of the foramen magnum, are characteristic of primitive salamanders (hynobiids, ambystomatids), occur in most plethodontid genera, and are doubt- less primitive in plethodontids. Occipital condyles of Desmognathus, Leurognathus, and Phaeogna- thus are borne on the distal tips of moderately long, cylindrical pedicels or stalks. Stalked occipital condyles were first noted by Cope (1866) and were one of the characters on which his family Desmog- nathidae was based. Mandibles of these three gen- era are held largely immobile, and skulls move on condylar-atlas joints. The stalked condyles increase the distance between the skull proper and the vertebral column, and provide space for vertical skull movement. Suspensorium Plethodontids differ from other, more primitive salamanders in lacking bony pterygoids as adults. Pterygoids have been replaced by cartilaginous pterygoid processes of the suspensoria, which ex- tend anterolaterally toward the maxillary tips. The processes are shortest in those forms in which the maxillae are longest, and are particularly short in Desmognathus, Leurognathus, and Phaeognathus . Cartilaginous portions of the suspensoria are relatively large in most plethodontid genera, and the ossified portions (quadrates) are limited to distal regions and do not articulate with the occipito-otics or the parasphenoids. The quadrates communicate with the skull proper by means of large, cartilagin- ous, basal suspensorial portions. This situation is similar to that commonly encountered in primitive salamanders, and is evidently primitive in pletho- dontids. In large adults of some genera (e.g., Pseu- dotriton, Aneides ) there is a tendency for thin, bony, dorsomedial quadrate processes to develop, but the processes do not reach to the otic capsules. Ossified portions of the suspensoria are extensive in Desmognathus, Leurognathus, and Phaeogna- thus, but the cartilaginous portions are small. The quadrates extend below the squamosals to articu- late broadly and firmly with the otic capsules and with posterolateral parasphenoid processes. Crania raise in opening the mouths of these three genera, and forces between the crania proper and the quadrates are much greater than in more general- ized forms in which the quadrates serve simply as points of suspension of the mandibles. In most plethodontid genera the quadrates serve as the fulcra for mandibular motion; in Desmognathus and its allies the articulars serve as fulcra for cranial movement. 1966 EVOLUTION OF LUNG LESS SALAMANDERS 27 Palatopterygoid Palatopterygoids are present in larval and paedo- genetic plethodontids, but disappear during meta- morphosis. Larval plethodontids may be placed in two groups on the basis of palatopterygoid struc- ture. Palatopterygoids are large and elongate in Gyrinophilus, Pseudotriton, Stereo chillis, Eurycea, Manculus, Typhlotriton, Typhlomolge, Haideotri- ton, and Hemidactylium. The bones are expanded anteriorly, but are usually drawn into relatively narrow, elongate, posterior processes which articu- late directly, or almost articulate, with the quad- rates. The condition in these genera is similar to that encountered in primitive salamanders, and probably represents the primitive plethodontid condition. Palatopterygoids are small and teardrop-shaped, and have posterior points in Desmognathus and Leurognathus. The bones are separated from the quadrates by distances that approximate the total lengths of the palatopterygoids. This condition ap- pears to be advanced within the family. Squamosal The squamosals are most highly developed and strongest in those species of various genera in which head and jaw musculature is well developed ( e.g ., Desmognathus and allies, Aneides lugubris, Gyri- nophilus, Pseudotriton, Pseudoeurycea smithi) . Genera with reduced skeletal elements and head muscles (e.g., Manculus, Batrachoseps, Chiroptero- triton, Thorius, Oedipina) have thin, narrow, splint- like squamosals. Squamosal processes contribute to the otic-squamosal crests in Gyrinophilus, Pseu- dotriton, Stereochilus, and Eurycea, and the squa- mosals may have relatively large, dorsal raised re- gions in Desmognathus, Leurognathus, and Phaeog- nathus. Posterior squamosal crests are present in large Aneides. In general variation is closely related to function, and is more significant evolutionarily on the interspecific than on the intergeneric levels. The squamosals bear small, well-developed, slender, cylindrical processes in all Thorius (Tan- ner, 1952). The processes arise above the mid- points of the anteroventrally sloping, posterior squamosal edges and proceed almost directly pos- teriorly in horizontal planes. The quadratopector- alis musculature, which in other genera attaches to the quadrates, has shifted to the squamosals in Thorius, and the processes have apparently de- veloped in response to pressures resulting from the shift. These processes have been encountered else- where only in a single large individual of Oedipina complex. Opercular Plate The opercular apparatus of salamanders has re- cently been reviewed by Monath (1965). On the basis of an incomplete study of plethodontids he stated that primitive, aquatic, semi-terrestrial, and terrestrial plethodontid genera retain a well-de- veloped columella, but that columellar reduction has occurred independently in two lines (essentially the attached and free-tongued groups of von Wahlert, 1957). The major trend has been clear since Reed’s (1920) study. Although I disagree with the phylogenetic speculations of Monath con- cerning plethodontids, the overall trend is clearly as Reed, Dunn (1941), and others in addition to Monath have suggested. Desmognathus, Leurognathus, and Phaeognathus have large opercular plates which bear distinctive columellae on their anterolateral borders. The columellae are very short and stocky, and arise as broad-based, somewhat flattened processes. Columellae of the remaining plethodontid gen- era, when present, are tubular rods of small diameter. Salamander columellae are primitively rod-like (Reed, 1920), and the situation in these genera is probably primitive in the family. Columel- lae are present and well developed in Gyrinophilus, Pseudotriton, Stereochilus, Eurycea, Manculus, Typhlotriton, Typhlomolge, Haideotriton, Hemi- dactylium, Plethodon, Ensatina, Aneides, and Hydromantes. Columellae of Batrachoseps, re- ported to be absent by Reed (1920), Dunn (1926), and Piatt (1935), and to be present by Hilton (1945 a) and Monath (1965), are present in all individuals examined. Columellae are well de- veloped in B. wrighti and are only slightly reduced in the other species. Neotropical genera are characterized by a trend toward columellar reduction and loss. Primitive Chiropterotriton have the best developed columel- lae, but in advanced species (e.g., C. bromeliacia, C. xolocalcae, C. nasalis, C. abscond ens) columellae are lost. Very small columellae are present in Lineatriton. Pseudoeurycea has columellae that are usually vestigial and small. Vestigial, very short columellae are present in Bolitoglossa dunni and in one individual of Parvimolge townsendi, but are absent in all other species of both genera, and in all Thorius and Oedipina. When columellae are ab- sent the columellar processes of the suspensoria extend to the opercula. The opercula are essentially absent in some adult Chiropterotriton multidentatus, except for small amounts of tissue at the bases of the relatively well developed columellae. Reed (1920) stated that Typhlomolge should 28 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 not be included in the family Plethodontidae be- cause the columellae contribute considerably more to the plate portions than in other members of the family. Similar, but less extreme conditions are en- countered in Haideotriton and Eurycea pterophila, however, and appear to be related to the paedo- morphic nature of these groups. The situation cer- tainly does not warrant exclusion of the involved species from the Plethodontidae. Mandible The primitive salamander mandibular pattern is retained by the majority of plethodontid genera. Dentaries and prearticulars are of moderate to small size. Each mandible is curved primitively, so that the mandibular rami form broadly rounded arches. Meckelian grooves are open for much of the lengths of the dentaries, and at the anterior ends of the prearticulars. Specialized mandibles are found in Desmogna- thus, Leurognathus, and Phaeognathus. In these genera the dentaries are large, massive structures that are particularly broad ventrally. The dentaries are straighter than in other genera, and the resultant mandibular arches are considerably more pointed than in more primitive genera. The meckelian grooves are completely, or almost completely closed, and the anterior extensions of the prearticu- lars are entirely enclosed within the O-shaped dentaries. These specializations are related to func- tional changes in the three genera. The organisms force their way beneath rocks and into crevices, and the rigid, relatively straight, strengthened mandibles are presumably of significance in such behavior. The mandibular condition in Desmognathus, Leurognathus, and Phaeognathus is most closely approached by Stereochilus, in which the mandibu- lar arches are relatively elongate and pointed, and the anterior portions of the meckelian grooves are closed. The prearticulars in plethodontids are strongly inflected dorsomedially and shelves are small and directed medially in Gyrinophilus and Pseudotri- ton. In Stereochilus, Eurycea, Manculus, Hemi- dactylium, and Typhlotriton the shelves initially are inflected medially, then dorsomedially. Similar stiuations are encountered in Desmognathus, Leurognathus, and Phaeognathus, but the portions on which muscles insert are relatively larger than in other genera. The shelves are also directed dorso- medially in Plethodon, Ensatina, and Aneides, and the inflected portions are relatively very large in adult A. ferreus, A. flavipunctatus and A. luguhris. Prearticulars are very small in Ensatina. The rela- tively small prearticulars of Hydromantes, Batrach- oseps, Bolitoglossa, Oedipina, Pseudoeurycea, Chiropterotriton, Parvimolge, Lineatriton, and Thorius are dorsomedially inflected. Desmognathus, Leurognathus, and Phaeognathus have well developed, anterior, dorsal prearticular projections located just anterior to the prearticular shelves. The atlas-mandibular ligaments insert on the projections. Similar but considerably smaller projections are present in Stereochilus, some large Eurycea bislineata, and adult Typhlotriton. Slight indications of the processes are occasionally en- countered in other genera. Dentary bulk is relatively greatest in Desmogna- thus, Leurognathus, and Phaeognathus, and this bulk is approached, but not matched, by large, specialized species such as Aneides lugubris. The dentaries are relatively slender in most genera. Raised, pointed, dentary processes are located in the coronoid region in Typhlotriton, and about match the prearticulars in height. Laterally in- flected, flange-like processes are present in large Plethodon and Aneides. Other genera either have rather poorly developed or no dentary processes. Very small coronoids that bear teeth occur on the inner margins of the dentaries in larval Desmognathus, Leurognathus, Gyrinophilus, Pseu- dotriton, Stereochilus, Eurycea, Manculus, Typhlo- triton, and Hemidactylium, and in adults of the paedogenetic species. No indication of the element has been seen in embryos of species which undergo direct development. Hyobranchial Skeleton Larval structure Desmognathus and Leurognathus have four larval epibranchials, but three are present in larvae of all other genera, and in the paedogenetic adults. Bolitoglossa subpalmata embryos, removed from egg capsules, have adult configurations with a single pair of epibranchials, and the same is true of hatch- ling Plethodon and Batrachoseps. Dent (1942) has reported three epibranchials in an embryo (un- specified size) of Plethodon cinereus. Larval hyoids are found in adults of the paedo- genetic species, and all have branchial plates. The branchial plates of Typhlomolge are notably elon- gated and proportionately longer than in other genera. The ceratobranchial arms are swept pos- teriorly and make considerably smaller angles with each other than in other genera. Adult structure Several important papers (Piatt, 1935; Tanner, 1952) have dealt with tongue morphology, includ- 1966 EVOLUTION OF LUNGLESS SALAMANDERS 29 mg some aspects of the musculature. It is especially important that the attachment of the tongue to the anterior margin of the mouth be understood. Genioglossal muscles primitively extend as well developed strap-like structures from the area of the mandibular symphysis dorsally into the tongue. They bind the anterior border of the tongue and aid in returning the extended tongue into the mouth. The primitive condition is encountered in Desmognathus, Leurognathus, Phaeognathus, Plethodon and Amides (the muscles considered to be diagnostic of desmognathines by Tanner, 1952, are the genioglossals) . The fibers proceed for a short distance medially from their origin on the dentaries, then swing abruptly dorsally and insert in the tongue. The genioglossals are modified and a little elongated in Ensatina and Hemidactylium. Each muscle originates somewhat posteriorly, along the medial margin of the dentary, and proceeds almost directly medially, then abruptly moves dor- sally into the tongue. A bizarre arrangement is seen in Batrachoseps, in which the fibers originate near the articular end of the mandible and proceed anteriorly following the curve of the mandible. Near the anterior end of the mandible the fibers proceed medially, resembling the medially directed muscle in Ensatina and Hemidactylium, and finally move dorsally at the midline. Anterior portions of the tongue are held tightly in the mouth by the stout genioglossals in Desmog- nathus, Leurognathus, Phaeognathus, Plethodon, and Aneides. In Ensatina and Hemidactylium movement is possible, and tongues may be pro- pelled for some distance out of the mouth. The tongue of Batrachoseps may be protruded almost as far as in those genera which lack genioglossals: Gyrinophilus, Pseudotriton, Stereochilus, Eurycea, Manculus, Typhlotriton, Hydromantes, Bolito- glossa, Oedipina, Pseudoeurycea, Chiropterotriton, Parvimolge, Lineatriton, and Thorius. In Stereo- chilus and Typhlotriton, the tongues are attached anteriorly by fleshy, non-muscular tissue. Ceratohyal Anteriorly expanded ceratohyals are character- istic of most plethodontids, but the structures are attenuated anteriorly in Hydromantes, Oedipina, southern species of Chiropterotriton ( bromeliacia , abscondens, see Rabb, 1958), Lineatriton, Parvi- molge richardi, and Thorius (see Tanner, 1952). Attenuation occurs only in specialized groups, and is a specialized condition. The attenuated ceratohyals of Oedipina are diag- nostic of the genus (Tanner, 1952). The posterior cylindrical portions flatten slightly at about the level of the midpoint of the first basibranchial, and at that point small processes proceed antero- medially; the processes serve as the point of attachment of suprapeduncularis muscles. The ceratohyals extend well anterior to the processes as attenuated rods. Tanner (1952) has indicated that the ceratohyals of Bolitoglossa are noticeably shorter than those of other neotropical genera, and that the elements do not proceed much beyond the origin of the rela- tively broad suprapeduncularis muscles. Shortened ceratohyals are also present in Hydromantes, and both Bolitoglossa and Hydromantes lack sublingual folds, structures related to the anterior elongation of the ceratohyals. Sublingual folds occur in all neotropical genera except Bolitoglossa, and in several midlatitude genera (e.g., Pseudotriton), but the folds appear to be artifacts of the structure of the ceratohyals and of the adetoglossal condition. There is no evidence to support Tanner’s assurances that all folds are homologous structures. Cornua Paired cornua are associated with the anterior portions of the first basibranchials in all genera except Hydromantes, in which the tongue muscula- ture normally associated with the cornua is firmly attached to the anterior basibranchial tip. The re- mainder of the genera may be grouped as follows: 1. Cornua distinctly separated from the first basi- branchials, joined by connective tissue. Desmog- nathus, Leurognathus, Phaeognathus, Gyrin- ophilus, Pseudotriton, Stereochilus, Eurycea, Manculus, Typhlotriton, Hemidactylium, Pleth- odon, Aneides, Ensatina. 2. Cornua closely applied and broadly attached to the first basibranchials, often seemingly or actually continuous with the basibranchials. Batrachoseps, Bolitoglossa, Oedipina, Pseudo- eurycea, Chiropterotriton, Parvimolge, Linea- triton, Thorius. Cornual attachment is very clear for most of the genera. However, the cornua of Typhlotriton are relatively broad-based and superficially appear to be in direct continuity with the first basibranchial. Histological preparations reveal that the joint region contains densely packed cells similar to cartilage cells, but interstitial substance is absent in the joint region. The cartilages are not in direct continuity. Histological preparations of Batracho- seps and Chiropterotriton clearly show that the junction of the cornua and first basibranchial is an area containing cartilage cells and interstitial sub- 30 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 stance; there is no joint. The joint is sharp and clear in Enscitina, with the region between the cartilages consisting of densely packed fibroblasts and con- nective tissue fibers. The genera of group 1 may be placed in sub- groups. The narrow, elongate cornua of Desmogna- thus, Leurognathus, and Phaeognathus arise from narrow bases, and are about one-half or more of the total length of the first basibranchial. Those of Gyrinophilus, Pseudotriton, Stereochilus, and Hemi- dactylium are moderately broad-based structures one-fourth to one-half the length of the first basi- branchial; in contrast to those of the other three genera the cornua of Stereochilus are expanded distally rather than pointed or cylindrical. The small, tapering, pointed cornua of Eurycea, Manculus, and Typhlotriton are less than one-half the length of the first basibranchials. Old adults of Typhlotriton have cornua with slight distal expan- sion. The large, distally expanded, wing-like cornua of Plethodon, Ensatina, and Aneides are one-third (Ensatina) or from two-fifths to three-fourths the length of the first basibranchials. Expanded distal portions are extremely thin. Proximal portions are cylindrical. Batrachoseps has the longest cornua of the group 2 genera, and its long, thread-like cornua curve gracefully in 180° arcs with total lengths of from one-third to one-half the length of the first basi- branchial. The cornua of Bolitoglossa strikingly re- semble those of Batrachoseps, but are straighter and shorter (one-fifth to one-third times first basi- branchial length). Cornua of the remaining genera are successively shorter and inconspicuous; each is equal to less than one-fifth of the first basibranchial length in Pseudoeurycea, Chiropterotriton, and Oedipina, and less than one-tenth in some Chirop- terotriton (e.g., hromeliacia, abscondens), Par- vimolge, Lineatriton, and Thorius. Lingual cartilage Piatt (1935) reported lingual cartilages to be present in Gyrinophilus, Pseudotriton, Eurycea, and Manculus, found “forerunners” in Stereochilus and Typhlotriton, but found no trace in other plethodontids. He considered lingual cartilages to be “paleotelic” features, homologous with the “Sehnenplatte” of Salamandra, but his work was biased by his assumption that plethodontids had been derived from salamandrids. It is apparent that lingual cartilages are homo- logous with the anterior extensions of the first basibranchials of primitive salamanders, as sug- gested by Hilton (1947 a). Both serve as the sites of hyoglossal muscle insertion, and both occupy similar positions. They differ in that the lingual cartilages are squat (usually) rather than elongate, and separated from (usually) rather than con- tinuous with the first basibranchials. Plethodontids have long been grouped on the basis of tongue attachment. Uzzell (1961) intro- duced the useful terms deto- and adetoglossal to refer to presence and absence respectively of an- terior tongue attachment, or to the attached and free-tongued conditions. Detoglossal genera are Desmognathus, Leurognathus, Phaeognathus, Ster- eochilus, Typhlotriton, Hemidactylium, Plethodon, Aneides, Ensatina, and Batrachoseps, but of these the genioglossal muscles have been lost in Typhlo- triton and Stereochilus, and elongated from greater to lesser extents in Batrachoseps, Ensatina, and Hemidactylium. It is significant that anterior first basibranchial extensions are present as consistently well developed, distally knobbed structures only in genera in which the genioglossal muscles are short and strap-like (see Piatt, 1935), and in which es- sentially no movement of the anterior portions of the tongue out of the mouth is possible (Desmog- nathus, Leurognathus, Phaeognathus, Plethodon, Aneides). Ensatina, anatomically detoglossal but with a fairly protrusible tongue, has relatively smaller processes than the above genera, and the processes are pointed rather than distally expanded. Histological preparations show that the region just anterior to the cornual attachment in Ensatina is strongly depressed and contains densely arranged cartilage cells but notably small amounts of inter- stitial substance. This is in marked contrast with regions craniad and caudad where the cells are widely separated by large amounts of interstitial substance. This region is apparently flexible, and the anterior extension bends ventrally during tongue flipping. A joint apparently forms in the depressed region in some large individuals. Dissec- tions reveal that portions anterior to the attachment of the cornua are occasionally separated from the first basibranchials in large Ensatina, and form small, triangular, cartilaginous structures on which insert hyoglossal muscle fibers. These structures must be considered lingual cartilages, and their appearance in Ensatina demonstrates the relation- ship of the anterior processes or first basibranchial extensions and the lingual cartilages. Both anterior extensions and lingual cartilages are absent in Batrachoseps and Hemidactylium, the other genera that have genioglossal muscles. I am unable to cor- roborate Hilton’s (1959) observation that Hemi- dactylium has a lingual cartilage. Stereochilus and Typhlotriton have true lingual cartilages despite Piatt’s (1935) report that a 1966 EVOLUTION OF LUNGLESS SALAMANDERS 31 “Sehnenplatte” or membranous forerunner is pres- ent. In Typhlotriton lingual cartilage formation occurs rather late during ontogeny. In small adults the cartilages have a membranous appearance, as reported by Piatt. Microscopic sections of a tongue of an adult Typhlotriton (53.1 mm.) reveal that a well developed lingual cartilage is present and that it is attached to the anterior end of the first basibranchial at either lateral margin by dense uggregations of cartilage cells. On the midline only dense connective tissue fibers hold the vertically oriented cartilage to the horizontal basibranchial. Well developed lingual cartilages are present in Gyrinophilus, Pseuclotriton, Eurycea, and Mancu- lus. In these genera the structures are irregular, squat, moderately broad elements that resemble those of Stereochilus and Typhlotriton rather than those of Ensatina. The only remaining midlatitude genus, the truly adetoglossal Hydromantes, lacks both lingual car- tilages and anterior basibranchial extensions. Tanner (1952) corroborated Piatt’s observation that lingual cartilages are absent in neotropical genera. Rabb (1955), however, reported lingual cartilages to be present in some individuals of a newly described species, Chiropterotriton prisons, and I have found the cartilages in some Pseudo- eurycea ( bellii , cephalica), some Chiropterotriton (chiropterus, multident atus), and in Parvimolge townsendi. Small triangular structures similar to those of Ensatina are present in Pseudoeurycea werleri, and similarly shaped structures, but more definitely separated from the first basibranchials, are present in Chiropterotriton multidentatus. The cartilages of other species are rounded distally rather than pointed, and are short, squat structures. The cartilages are never as distinctly separated from the first basibranchials, nor as broad and flat, as those of Gyrinophilus, etc. The poorly defined, in- completely detached tissue found at the anterior ends of the first basibranchials in Pseudoeurycea goebeli, P. leprosa, Chiropterotriton dimidiatus, and C. bromeliacia may represent vestigial lingual cartilages. It is apparent that lingual cartilages are homo- logous with the anterior first basibranchial exten- sions of anatomically and functionally detoglossal genera. Separation of anterior extensions and sub- sequent formation of lingual cartilages appears to be associated with acquisition of a degree of tongue freedom. It may be assumed that retention of an- terior extensions is functionally undesirable in at- tainment of the adetoglossal condition, since well developed extensions are absent in all genera that have even a slight tendency toward the free-tongued condition. Lingual cartilages may be related to flip- ping the tongue pads, but they are lost in many accomplished tongue-flippers (e.g., Bolitoglossa) and are not functionally essential. Presence of lingual cartilages may represent a relatively ad- vanced, functionally significant stage in the loss of the anterior first basibranchial extensions. Evidence is strong for parallel gain and subsequent parallel loss of lingual cartilages as part of an over-all mor- phological trend which is as follows: 1. Well developed anterior extensions of the first basibranchials present; no indications of lin- gual cartilages. Desmognathus, Leurognathus, Phaeognathus, Plethodon, Aneides. 2. Anterior basibranchial extensions small, point- ed, flexible or partially detached; lingual carti- lage-like structures thus formed. Ensatina, some Pseudoeurycea. 3. Completely detached lingual cartilages; first basibranchial extensions absent. Gyrinophilus, Pseudotriton, Eurycea, Manculus, Stereochilus, some Chiropterotriton, Parvimolge townsendi, Typhlotriton,? some Hemidactylium. 4. Poorly defined tissue present at first basibran- chial tips which may represent lingual cartilage vestige or remnant; no anterior basibranchial extensions. Some Typhlotriton, some Pseudo- eurycea, some Chiropterotriton. 5. Both anterior basibranchial extensions and lin- gual cartilages absent. Hemidactylium, Hydro- mantes, Batrachoseps, some Pseudoeurycea, some Chiropterotriton, Lineatriton, Oedipina, Parvimolge richardi, Bolitoglossa, Thorius. The lingual cartilage of plethodontids is clearly not derived from the lingual process of the basi- branchial of porolepiform fishes, as suggested by Jarvik (1963), but is a new element that appears for the first time in plethodontids and is clearly related to the evolution of tongue flipping. Piatt (1935) has discussed the homologies of the lingual cartilage. He considered it to be the homologue of the “Sehnenplatte” of Salamandra. The “Sehnen- platte” is a ligamentous plate which serves as the site of insertion of the hyoglossal and abdomino- hyoideal (anterior extension of rectus abdominis, also called rectus cervicis profundus) musculature. This ligamentous plate is present as a well devel- oped structure in microscopic sections of the tongues of plethodontids dorsal to the first basi- branchial in the posteroventral portion of the tongue pad. Both lingual cartilage and ligamentous plate are present in such forms as Typhlotriton. The hyoglossal musculature originates on the lingual 32 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 cartilage and at least some of its posterior fibers insert on the plate. Clearly the cartilage and plate are functionally distinct, non-homologous struc- tures. The presence of lingual cartilages in pletho- dontids in no way supports Piatt’s theory of plethodontid origin from salamandrids. Since all evidence indicates that lingual cartilages occur only in plethodontids that have attained some degree of tongue freedom, Piatt was correct in stating that the lingual cartilage is not homologous with the otoglossal cartilages of ambystomatids, which serve as tongue pad supports. First basibranchial Plethodontid genera may be grouped on the basis of first basibranchial shape: 1. Major basibranchial expansion near midpoint. Desmognathus, Leurognathus, Phaeognathus, Hydromantes. 2. Major expansion slightly posterior to attachment of cornua. Gyrinophilus, Pseudotriton, Stereo- chilus, Eurycea, Manculus, Typhlotriton, Hemi- dactylium, Plethodon, Ensatina, Aneides. 3. No major expansion, but basibranchial broad- est at anterior end. Batrachoseps, Bolitoglossa, Oedipina, Pseudoeurycea, Chiropterotriton, Par- vimolge, Lineatriton, Thorius. Hydromantes has cylindrical basibranchials that have thin, flattened, flange-like expansions on either side. The greatest width is equal to about 3.5 to 5 times the least diameter (width ratio) of each basi- branchial. If the flanges are disregarded the first basibranchials resemble those of group 3. In the other genera of group 1 each first basibranchial is slightly flattened in the area of greatest expansion, but no flanges are present. Width ratios are from 1.5 to 3. Thin lateral extensions arise from the anterior one-third of each basibranchial in Eurycea, Man- culus, and Typhlotriton. Width ratios are from 5 to 6, the greatest in the family, and basibranchials are paddle-shaped. The basibranchials of Pseudo- triton and Gyrinophilus are rather flattened, and have width ratios of 2.5 to 3. Width ratios are less than 2.5 in Stereochilus, Hemidactylium, Pletho- don, Ensatina, and Aneides, but the expanded areas are much nearer the anterior ends in the first two. Plethodon, Ensatina, and Aneides differ from Hemidactylium in lacking sharply defined areas of expansion; Stereochilus is intermediate between the two conditions. The first basibranchials of Stereo- chilus and Hemidactylium resemble those of Gyri- nophilus and Pseudotriton more closely than those of other genera, despite their relatively slight expansion. The genera of group 3 have essentially con- tinuous cornua which arise from the anterior ends of the basibranchials, and give the impression, more apparent than real, that the anterior ends are ex- panded. When present, expansion is gradual, not sharply defined. Proportions of articulated elements Plethodontid genera may be grouped according to the relative proportions of the articulated ele- ments as follows: 1. Sequence of length from longest to shortest; ceratobranchial I-(ceratobranchial II, basibran- chial I, or epibranchial) . Desmognathus, Leu- rognathus, Phaeognathus, Plethodon, Aneides. 2. (Ceratobranchial I or epibranchial) -basibran- chial I-ceratobranchial II. Ensatina. 3. Epibranchial- (ceratobranchial I or basibran- chial I) -ceratobranchial II. Manculus. 4. Epibranchial-ceratobranchial I-(ceratobranchial II or basibranchial I). Eurycea, Batrachoseps wrighti. 5. Epibranchial-ceratobranchial I-ceratobranchial Il-basibranchial I. Gyrinophilus, Pseudotriton, Stereochilus, Typhlotriton, Hemidactylium. 6. Epibranchial-basibranchial I-ceratobranchial I- ceratobranchial II. Hydromantes, most Batra- choseps, Bolitoglossa, Oedipina, Pseudoeurycea (one species, P. hellii, has ceratobranchial I and basibranchial I reversed), Chiropterotriton, Parvimolge, Lineatriton, Thorius. The first ceratobranchials and the epibranchials are of nearly equal length in Ensatina, and the epi- branchials are the longest elements in all of the remaining genera, except those of group 1. Epi- branchial length is clearly associated with acquisi- tion of tongue freedom; the longer the epibranchials (relatively), the farther the tongue can be extended from the mouth. Significantly, short epibranchials (at least shorter than the first ceratobranchials) are found only in the five genera (group 1, above) which have no osteological or myological modifica- tions for even partially freeing the anterior tongue margins. The condition in which the first cerato- branchials are the longest elements and in which there is no tongue freedom is the condition in primi- tive families; it is clearly the primitive plethodontid condition. The first are longer than the second ceratobran- chials in all genera, and are longer than the first basibranchials in all but group 6. The first cerato- 1966 EVOLUTION OF LUNGLESS SALAMANDERS 33 branchiate are relatively shortest in Hydromantes (about one-half the length of the first basibran- chials), and longest in Stereochilus (about one and two-thirds times first basibranchials) . The genera may be grouped according to second ceratobranchial length: 1. Second ceratobranchials less than 0.8 times first basibranchials. Hydromantes, most Batracho- seps, Bolitoglossa, Oedipina, Pseudoeurycea, Chir optero triton, Parvimolge, Lineatriton, Thorius. 2. Second ceratobranchials more than 0.8 but less than 1.3 times first basibranchials. Batrachoseps wrighti, Desmognathus, Leurognathus, Phaeog- nathus, Gyrinophilus, Pseudotriton, Eurycea, Manculus, Typhlotriton, Plethodon, Aneides, Ensatina. 3. Second ceratobranchials more than 1.3 times first basibranchials. Hemidactylium, Stereo- chilus. The relative dimensions of the first and second ceratobranchials may also be used to group the genera: 1. First ceratobranchials of significantly greater bulk than second ceratobranchials; minimal diameter of first equal to or greater than maximal diameter of second. Desmognathus, Leurognathus, Phaeognathus, Gyrinophilus, Pseudotriton, Eurycea, Manculus, Stereochilus, Typhlotriton, Hemidactylium, Plethodon, En- satina, Aneides. 2. Second ceratobranchials of significantly greater bulk than first ceratobranchials; minimal di- ameter of second equal to or greater than maximal diameter of first. Hydromantes, Batra- choseps, Bolitoglossa, Oedipina, Pseudoeurycea, Chir opt er o triton, Parvimolge, Lineatriton, Thorius. First ceratobranchials that are stouter than sec- ond ceratobranchials are typical of primitive sala- manders, and such a condition is primitive in plethodontids. All of the genera of the latter group (2) are functionally free-tongued (although Batra- choseps is anatomically detoglossal) , and all have greatly elongated epibranchials. The subarcual rectus muscles envelop each epibranchial, and at- tach anteriorly to each ceratohyal. When these muscles are contracted, the epibranchials are moved forward. As a result the entire articulated hyobranchial skeleton, surrounding muscle, and connective tissue (i.e., the tongue) are extended from the mouth. For efficiency in transmission of force it is functionally advantageous for the lines of force between articulated elements to be rela- tively straight. In all plethodontids the second cera- tobranchials articulate directly with the posterior end of each first basibranchial, while the first cera- tobranchials are attached by ligaments to the side of the basibranchial. Primitively the first cerato- branchials are the more dominant force trans- mission elements, and the dominance is maintained in most midlatitude plethodontids. In the eastern North American adetoglossals epibranchials are not greatly elongated, but the first ceratobranchials have acquired specialized braces of two types: (1) attachment behind the specialized first basi- branchial anterior expansion (Manculus, Eurycea ), (2) cartilaginous struts attached to the expanded ends of the ceratobranchials (Gyrinophilus, Pseudo- triton). In Hydromantes, Batrachoseps, and the neotropical genera the primitively weak, relatively long, slightly sigmoid-shaped second ceratobran- chials have become strongly shortened and straight, apparently as a result of selection pressures associ- ated with attainment of the free-tongued condition and correlated lengthening of the epibranchials. In these genera the first ceratobranchials have been weakened and anteriorly bowed, and in some gen- era (e.g., Thorius ) have lost their direct articulation with the basibranchial. As a result of these changes there is a direct, rather straight line of force trans- mission from each epibranchial through the short, stout ceratobranchial to the first basibranchial. The mechanics of tongue action have recently been discussed by Uzzell (1961), who cites the enlarged second ceratobranchials and their tendency to cal- cify in certain neotropical genera (see below) as evidence that pressure applied at the posterior ends of the epibranchials is probably transmitted to the first basibranchial through the second rather than first ceratobranchials. As suggested above the length of the epibran- chials is highly significant, and the genera may be grouped as follows: 1. Epibranchials less than 1.25 times first basi- branchials. Desmognathus, Leurognathus, Phaeognathus, Plethodon, Ensatina, Aneides. 2. Epibranchials 1.5 to 2 times first basibranchials. Gyrinophilus, Pseudotriton, Stereochilus, Eury- cea, Manculus, Typhlotriton, Hemidactylium, Thorius. 3. Epibranchials 2 to 3.5 times first basibranchials. Hydromantes, Batrachoseps, Bolitoglossa, Oedi- pina, Pseudoeurycea, Chiropterotriton, Parvi- molge, Lineatriton. 34 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 Since the basibranchials in those genera which have anterior basibranchial extensions are propor- tionately longer than in those that lack the ex- tensions, the epibranchials have also been compared with the first ceratobranchials: 1. Epibranchials less than or almost equal to length of first ceratobranchials. Desmognathus, Leu- rognathus, Phaeognathus, Plethodon, Ensatina, Aneides. 2. Epibranchials more than 1 but less than 1.7 times first ceratobranchials. Gyrinophilus, Pseu- dotriton, Stereochilus, Eurycea, Manculus, Typhlotriton, Hemidactylium. 3. Epibranchials 1.8 to 6 times first ceratobran- chials. Hydromantes, Batrachoseps, Bolito- glossa, Oedipina, Pseudoeurycea, Chiropterotri- ton, Parvimolge, Lineatriton, Thorius. The epibranchials of group 3 genera are remark- ably long. The longest occur in Hydromantes, in which epibranchials may be 5.9 times the first ceratobranchials. Very long epibranchials are also found in Chiropterotriton (maximum, 4.6 times first ceratobranchials), Oedipina (4.0), Parvi- molge (3.7), Pseudoeurycea (3.2), Bolitoglossa (2.9), Batrachoseps (2.6), and Thorius (2.7). The shortest adult epibranchials in group 3 are found in Batrachoseps wrighti (1.8). Of the group 1 and 2 genera only Eurycea has epibranchials that are more than 1.4 times the first ceratobranchials (1.33-1.67). Second basibranchial Presence or absence of second basibranchials in adults serves as a character to group plethodontid genera: 1. Second basibranchials present in adults. Des- mognathus, Leurognathus, Phaeognathus, Gy- rinophilus, Pseudotriton, Stereochilus, Eurycea, Manculus, Typhlotriton, Hemidactylium, Hai- deotriton, Typhlomolge, Plethodon, Ensatina, Aneides. 2. Second basibranchials absent in adults. Hydro- mantes, Batrachoseps, Bolitoglossa, Oedipina, Pseudoeurycea, Chiropterotriton, Parvimolge, Lineatriton, Thorius. An extremely tiny second basibranchial, a small dot of ossified tissue, is present in the largest Pseudoeurycea smithi examined (73.2 mm.). Cope (1889) illustrated a second basibranchial in Batra- choseps, and Piatt (1935) reported a reduced sec- ond basibranchial, but Hilton (1947) and Uzzell (1961) say the structure is absent in their material. Second basibranchials are absent in seventy-seven Batrachoseps of five species examined by me. A very well developed but nonossified and noncalci- fied inscriptio tendinis is found in the longitudinal throat musculature of Batrachoseps in the same position as a second basibranchial, and less well developed structures are found in other genera of group 2. The element is ligamentous, not cartilagi- nous, and is considered a second basibranchial vestige by Tanner (1952). The genera of group 1 may be further grouped on the basis of second basibranchial structure: 1. Width of second basibranchials in adults greater than length of first basibranchials. Gyrinophilus, Pseudotriton, Eurycea, Manculus, Hemidactyl- ium. 2. Width of second basibranchials two-thirds to three-fourths times first basibranchials. Stereo- chilus, Typhlotriton. 3. Width of second basibranchials less than two- thirds times first basibranchials. Desmognathus, Leurognathus, Phaeognathus, Plethodon, En- satina, Aneides. Second basibranchials are very strongly triradiate with very long arms and long, medial, posteriorly extending processes in Gyrinophilus, Pseudotriton, Eurycea, Manculus, and Typhlotriton. The struc- tures are distinctly triradiate, but have shorter lateral and posterior processes than the above gen- era in Stereochilus, Typhlomolge, and Haideotri- ton. In the latter genera, and in Typhlotriton, second basibranchials are somewhat cruciform with arms strongly swept posteriorly. Second basi- branchials of remaining genera are usually bowed with arm-tips extended posterolaterally, and with no posteromedial projections. Small posteromedial enlargements may be present occasionally. Second basibranchials are notably reduced in size in Phaeognathus and some species of Plethodon (e.g., dunni). Calcification Calcifications of elements other than the second basibranchials occur in some genera. They are ap- parently not of universal occurrence in any single species or genus, and are more common in the larger members of a given population (Uzzell, 1961; Wake, 1963). The following list contains all observations of calcified elements known to me, and is a composite of my personal observations as well as those of Wiedersheim (1877), Piatt (1935), Hilton (1947a), and Uzzell (1961). Personal observations (*), unique observations (**): 1966 EVOLUTION OF LUNGLESS SALAMANDERS 35 1. Ceratohyals (calcification limited to region of posterior hook unless otherwise noted). Larval Eurycea (extensive, into expanded portions)**, Eurycea**, Typhlotriton** , Hemidactylium** , Batrachoseps** , Oedipina (extensive, at hook and into expanded portions)**. 2. First basibranchials (calcification usually lim- ited to portions of central cores) . Desmognathus (entire element)*, Leurognathus** , Manculus, Typhlotriton** , Hemidactylium**, Plethodon* , Aneides** , Lineatriton* , Thorius**. 3. Cornua of first basibranchials. Desmognathus** , Leurognathus (extreme proximal portions only) **. 4. First ceratobranchials (central portions). Leu- rognathus (proximal portions only)**, Man- culus, Hemidactylium**, Parvimolge town- sendi* , Lineatriton* . 5. Second ceratobranchials (central portions or complete). Leurognathus** , Manculus, Hemi- dactylium**, Parvimolge townsendi*, Linea- triton*. 6. Epibranchials. Larval Gyrinophilus* , larval Eurycea *, larval Stereochilus**, Batrachoseps (extreme distal tips)**, Parvimolge townsendi (proximal portions), Lineatriton (proximal portions) . Uzzell (1961) has recently reported sequence of calcification of hyobranchial elements in Parvi- molge and Lineatriton, and has postulated close relationship on the basis of sequence similarities, as Rabb (1955) had earlier suggested. Calcification sequences reported by Uzzell are: (1) Lineatriton, second ceratobranchials-epibranchials-first basi- branchials, (2) Parvimolge townsendi, second ceratobranchials-epibranchials, (3) Manculus, first basibranchials-second ceratobranchials-first cerato- branchials. In simple distribution Leurognathus and Hemidactylium resemble the condition report- ed in Manculus (all may have calcified first cera- tobranchials, none have calcified epibranchials) more closely than that reported in Parvimolge and Lineatriton (both may have calcified epibranchials, none have calcified first ceratobranchials). Uzzell (1961) reports that the second ceratobranchials calcify before the first in Manculus; in at least some Leurognathus the opposite occurs. Synthesis of hyobranchial characters A synthesis of all of the above information con- cerning hyobranchial skeletal structure results in the following generic groupings (see Figs. 10 and 11): 1. Desmognathus, Leurognathus, Phaeognathus. 2. Plethodon, Ensatina, Aneides. 3. Gyrinophilus, Pseudotriton, Stereochilus, Eury- cea, Manculus, Typhlotriton, Haideotriton, Typhlomolge, Hemidactylium. 4. Hydromantes, Batrachoseps, Bolitoglossa, Oedi- pina, Pseudo eurycea, Chiropterotriton, Parvi- molge, Lineatriton, Thorius. These groups correspond exactly with my pro- posed tribal grouping (see under systematic sum- mary). At least one genus in each of the four groups has genioglossal muscles and is thus tech- nically detoglossal (see Uzzell, 1961). Assuming the above groups to be natural, adetoglossy has arisen in parallel in groups 3 and 4 and may be evolving in group 2 (through Ensatina ). Piatt (1935) also concluded that free tongues had arisen in parallel in groups roughly corresponding to my groups 3 and 4. He based his conclusions on the premise that the lingual cartilage, a supposed paleotelic character, was not present in Hydro- mantes and the neotropical genera. Tanner (1952) has discussed the illogic of this argument, and it has been shown (see above) that Piatt’s data were incorrect. Nevertheless there is some congruence between Piatt’s generic groupings, based on gen- eralities concerning hyobranchial skeleton and four muscles, and mine. Dentition Each plethodontid tooth is composed of a crown and a pedicel (Parsons and Williams, 1962). In those genera with generally strengthened jaw clos- ing adaptations (Desmognathus, Leurognathus, Phaeognathus, Aneides ) the pedicels are particu- larly well developed. A remarkable situation is found in Phaeognathus in which the crowns are extremely shortened, very stocky, and blunted. The pedicels are very large and greatly strengthened, and are almost double the length of the crowns. Species with moderately to weakly developed skulls have relatively small, fragile pedicels. All larval teeth are unicuspid, and during meta- morphosis there is a gradual replacement with bicuspid crowns. Adult teeth are considerably vari- able as to shape and size, and both sexual and ontogenetic variation occur. Maxillary, dentary, and vomerine teeth are nor- mally bicuspid with large lingual cusps and small, crescentic labial cusps, but some exceptions occur (some Aneides, adult male Hydromantes, Eurycea ). Small individuals of all plethodontid species ex- cept Aneides lugubris have bicuspid premaxillary 36 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 Figure 10. Hyobranchial apparatus. Ceratohyals removed. A. Plethodon yonahlossee, B. Des- mognathus quadramaculatus, C. Typhlotriton spelaeus. Abbreviations: Bl, first basibranchial; B2, second basibranchial; CO, cornua of first basibranchial, Cl, first ceratobranchial; C2, second ceratobranchial; E, epibranchial; LC, lingual cartilage. Line equals one mm. 1966 EVOLUTION OF LUNGLESS SALAMANDERS 37 teeth, and bicuspid teeth occur in adults of most species. Males of the more primitive genera (Gyrinophilus, Pseuclotriton, Stereochilus, Typhlo- triton) have premaxillary teeth that are only slightly modified, but the teeth are elongated and directed anteroventrally in adult males of at least some species of Desmognathus, Leurognathus, Eurycea, Manculus, Hemidactylium, Plethodon, Aneides, Hydromantes, Batrachoseps, Bolitoglossa, Oedi- pina, Pseudoeurycea, Chiropterotriton, Lineatriton, Parvimolge, and Thorius. The elongated teeth may be unicuspid and spine-like, or the labial cusps may be greatly elongated and either projected antero- ventrally, or ventrally hooked (see also Noble, 1927 a). Hooked premaxillary teeth are common in male Plethodon, Pseudoeurycea, Chiropterotri- ton, and Bolitoglossa. Sexual dimorphism is more pronounced in advanced than in primitive genera, and elongated premaxillary teeth have unquestion- ably appeared more than once. There is no evidence that the character has disappeared and reappeared in the family, the views of Dunn (1927) and Noble (1927 a) to the contrary notwithstanding. Maxillary teeth are greatly reduced in number or absent in some Bolitoglossa (chica, colonnea, occidentalis, rufescens), some Oedipina (elongatus, alfaroi), and most Thorius. It has been stated (Taylor, 1944; Smith and Taylor, 1948) that Thorius lacks maxillary teeth, but they occur in some numbers in several populations (see Gehl- bach, 1959). Reduction in number of maxillary teeth also occurs in Aneides, especially in members of the lugubris species group (Wake, 1963). In general, it may be stated that reduction in maxillary dentition is a specialization encountered in highly specialized and advanced groups. Reduction in number coupled with hypertrophy in individual tooth size is also a specialized condition. Vertebral Column Regional differentation A single cervical vertebra, the atlas, occurs in all salamanders. Trunk vertebrae in salamanders vary from 12 to over 60, and from 13 to 24 in plethodontids. A single sacral vertebra is typical of all salamanders, except sirenids. From 2 to 3 caudosacral vertebrae are present in plethodontids, but caudal vertebrae are extremely variable, and are subject to interspecific, individual, ontogenetic, and sexual variation in number. Significant variation is encountered in the num- bers of trunk vertebrae of plethodontids. Range and modal numbers of trunk vertebrae are pre- sented in Table 1. The table is based primarily on my observations. Ranges of variation of trunk Table 1. Variation in number of trunk and caudosacral vertebrae. GENUS TRUNK VERTEBRAE CAUDOSACRAL RANGE MODE VERTEBRAE Desmognathus 15 15 2 Leurognathus 15 15 2 Phaeognathus 21-23 22 2 Gyrinophilus 18-20 19 3 Pseudotriton 17-18 18 3 Stereochilus 18-20 19 3 Eurycea 14-21 15 3 Manculus 17-18 18 3 Typhlotriton 17-20 18 3 Typhlomolge 13-14 14 3 Haideotriton 14 14 3 Hemidactylium 15 15 3 Plethodon 14-24 17 3 Ensatina 14-15 14 3 Aneides 15-18 16 3 Hydromantes 13-14 14 3 Batrachoseps 16-21 20 2-3 Bolitoglossa 14 14 2 Oedipina 18-22 20 2 Pseudoeurycea 14 14 2 Chiropterotriton 14 14 2 Parvimolge 14 14 2 Lineatriton 14 14 2 Thorius 14 14 2 38 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 vertebrae doubtless will be extended for some genera, and I have relied on the literature for cer- tain genera ( Phaeognathus data from Brandon, 1965; Typhlotriton and Gyrinophilus from Bran- don, 1966; Plethodon from Highton, 1962). A modal number of 14 is found in ten genera, four have modes of 15, one has 16, one has 17, three have 18, two have 19, two have 20, and one has 22. The genera thus cluster about the lower and higher extremes. A brief survey of the numbers of trunk vertebrae in relatively primitive families is instructive: Salamandra 14-15, Salamandrina 13, P achy triton 12, Chioglossa 13, Euproctus 13, Pleurodeles 15, Notophthabnus 13, Ambystoma 14-16, Rhyacotriton 16-17, Dicamptodon 14-15, Hynobius 15-16, Batrachuperus 17, Onychodac- tylus 17. A tendency for low or relatively low num- bers of trunk vertebrae in primitive groups is obvious. Numbers on the order of 13 to 17 are probably closer to the primitive plethodontid number than those from 18 to 24 range. Thus the relative constancy of the low vertebral number of most neotropical genera appears to be a primitive character in the otherwise specialized and advanced group. Elongation, whether associated with increases in numbers of vertebrae or not, and attenuation have arisen in parallel a number of times in plethodontid evolution. Trunk elongation has appeared inde- pendently at least seven different times: (1) Phaeognathus, (2) Gyrinophilus-Stereochilus, (3) Eurycea-Manculus, (4) Plethodon, (5) Batracho- seps, (6) Oedipina, (7) Lineatriton. Attenuation has appeared at least as many times, and is found to some extent in Phaeognathus, Stereochilus, Eurycea, Manculus, Typhlotriton, Plethodon, Aneides, Batrachoseps, Bolitoglossa, Lineatriton, and Oedipina. Attenuation correlated with over-all size diminution is apparent in Batrachoseps, Man- culus, Chiropterotriton, Parvimolge, and Thorius. The elongation achieved in each major group of plethodontids usually involves both trunk and tail, and in most instances the numbers of trunk verte- brae increase. Several exceptions occur. In Eurycea longicauda and E. lucifuga generalized low num- bers of body vertebrae are maintained, and the trunks are relatively non-elongated, but the num- bers of caudal vertebrae arc increased. In Linea- triton attenuation and elongation results in part from vertebral elongation. Elongation of Oedipina is accompanied by increases in the vertebral num- bers of all species. The primitive Batrachoseps wrighti has relatively low numbers of trunk verte- brae (16-17), while the highly specialized species (B. attenuatus, B. pacificus) have high numbers (19-20). In two genera, Eurycea and Plethodon, great intrageneric variation is encountered; gen- eralized species have low numbers of trunk verte- brae (e.g., E. bislineata, E. longicauda, E. lucifuga, 15-16; P. vandykei 15-16), but advanced, elongated species are also found ( E . multiplicata, E. tyner- ensis 20-21; P. cinereus, P. richmondi 18-24). Elongation in some groups is correlated with in- crease in the numbers of trunk and caudal verte- brae (Oedipina, Batrachoseps, Phaeognathus), but only numbers of trunk vertebrae are noticeably in- creased in Gyrinophilus and Stereochilus, and tails are relatively short. Most plethodontid genera have three caudosacral vertebrae, but two distinct generic groups have two: (1) Desmognathus, Leurognathus, Phaeognathus, (2) Pseudoeurycea, Chiropterotriton, Lineatriton, Parvimolge, Thorius, Bolitoglossa, Oedipina (i.e., the neotropical genera). Batrachoseps is the only genus that shows variation in numbers of caudo- sacral vertebrae. B. wrighti and an undescribed species may have either two or three, but all other species have only two. Despite this fact, basal tail breakage is more common between caudal verte- brae one and two than in front of caudal one, and one is easily misled as to the number of caudosacral vertebrae. Most primitive salamanders have three or four caudosacrals, but Cryptobranchus, some species of Ambystoma, Chioglossa, and probably other genera, have only two. Three seems to be the standard number in most hynobiids, salaman- drids, and ambystomatids, and is probably the primitive plethodontid number. Reduction in cau- dosacral numbers must have evolved in parallel in the family. Accurate information concerning numbers of caudal vertebrae is difficult to obtain because of the common occurrence of autotomy or breakage of the tail and the regeneration of true vertebrae. In general most adult plethodontids have between 25 and 50 caudal vertebrae, but more than 50 are found in at least five genera: Eurycea (lucifuga, longicauda), Manculus, Batrachoseps, Lineatriton, and Oedipina. Batrachoseps and Oedipina com- monly have more than 60, and Oedipina, with the longest tails in the family, often has more than 75 and may have 100 (O. longissima). Very short tails are present in Hydromantes and Parvimolge; maximum numbers of caudals in both are less than 25. Tail breakage is commonly encountered in all genera except Hydromantes and in some species of Aneides (aeneus, ferreus, lugubris), in which tails are used in locomotion. Tails are very stout proximally in Desmognathus, Leurognathus, Phaeognathus, Gyrinophilus, Pseudotriton, Stereo- 1966 EVOLUTION OF LUNGLESS SALAMANDERS 39 chilus, and in large species of other genera (Eurycea, Aneides, Plethodon), and breakage is usually limited to distal portions. Tails separate from the trunks at basal constrictions located just posterior to the last caudosacral vertebrae in Hemi- dactylium, Ensatina, Bolitoglossa, Oedipina, Pseu- doeurycea, Chiropterotriton, Parvimolge, Linea- triton, and Thorius, but breakage can occur distally as well. In other genera breakage can occur at almost any point along the tail. Cervical vertebra Atlases of two very different types occur in the plethodontid genera. Vertebrae similar to those of primitive salamander families occur in the majority of the genera, but highly specialized elements are present in Desmognathus, Leurognathus, and Phaeognathus. Primitive atlases have well developed, anteriorly directed, odontoid processes that bear large, laterally oriented facets that articulate with the inner walls of the foramen magnum. Anterior mar- gins of the process are slightly concave, and the processes are deeply concave dorsoventrally. Paired enlarged condylar articulating facets which meet the occipital condyles are borne on stout antero- lateral processes. The facets are almost flattened to slightly concave. Conical centra arise from the midpoints of the vertebrae and proceed posteriorly and just slightly ventrally. Neural arch pedicels arise on either side of the centra, meet dorso- medially, and form characteristic bony bosses where the mandibular adductor muscles attach. Hydromantes, Bolitoglossa, Oedipina, Pseudo- eurycea, Chiropterotriton, Parvimolge, and Linea- triton have weakly fused pedicels, with little or no dorsal boss formation. The pedicels fuse without forming bosses in Manculus, Batrachoseps (attenu- ate, pacificus), and Thorius, but some low ridges which serve as muscle attachments may be present. Median, anteriorly placed, raised nodes or nodule- like processes form in Aneides, Plethodon, Eurycea, Typhlotriton, Typhlomolge, Haideotriton, Batra- choseps (wrighti), Pseudoeurycea (unguidentis, smithi), Chiropterotriton (nasalis), and Bolitoglossa (adspersa, cerroensis, dunni). The nodules are crater-like in most, and especially in Haideotriton and Typhlomolge. Well developed, raised bosses are present in Gyrinophilus, Pseudotriton, Hemi- dactylium, Ensatina, and some Aneides. Well de- veloped, nodule-like bosses are present on the posterior portions of the neural arches in Stereo- chilus, and are associated with well developed, vertical crests. Highly specialized atlases are found in Desmog- nathus, Leurognathus, and Phaeognathus. Tension exerted by highly developed atlantal-mandibular ligaments (Fig. 9), holds the mandibles relatively rigid, and forces within the ligaments have been sufficiently great to result in the development of prearticular projections and marked posterior ele- vation of the atlas. The pedicels are fused in these genera, and very large, broad, dorsal bosses are present. The bosses are relatively low anteriorly, but rise steadily posteriorly. Precipitous drops occur posteriorly, and high, broad, bony cliffs are formed. The ligaments attach to posteriorly located, transverse ridges on top of the bosses. Because the mandibles are relatively fixed, the crania must be raised to open the mouths of members of these genera; such a situation is not encountered in other salamanders. Occipital condyles of Desmognathus, Leurognathus, and Phaeognathus are very convex, and articulate with enlarged, very concave, atlantal condylar facets. In addition the odontoid processes are very short. Ventral portions of the processes are greatly reduced, and the anterior margins are markedly concave and eroded. The lateral articu- lar facets of the processes, which primitively articu- late with the walls of the foramen magnum, are very small and fail to enter the foramen; rather they articulate with the inner, ventral portions of the condylar stalks. Because of these modifications, the cranio-atlantal joint is considerably more flex- ible in these genera than in other plethodontids. The atlases of Desmognathus, Leurognathus, and Phaeognathus are raised above the level of the trunk vertebrae, and the centra are thus directed posteroventrally. The angle is so severe in Desmog- nathus that the dorsal, posterior surfaces of the centra are on about the same level as the odontoid processes. In preserved individuals elevation of the atlases over the trunk vertebrae is increased, ap- parently as a result of shortening of the mandibular- atlantal ligaments, and the result is the character- istic “pose” of the genera noted by many authors. The atlases of other plethodontids are not obviously raised. Trunk vertebrae Problems relating to vertebral structure and evo- lution are being studied currently by the author. Much of the information presented here must be considered preliminary, and no new information will be presented at this time concerning several aspects of vertebral morphology. In particular questions relating to the fate of the notochordal canal, the ontogeny of the intervertebral articula- tions, evolution of the vertebral centrum, and mechanism of tail autotomy are deferred to a later 40 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 date. Reference is made to my earlier statements concerning these points (Wake, 1963). Centrum structure Central dimensions vary considerably in pletho- dontids, and are correlated to a large extent with elongation and relative thickness of trunk muscula- ture. Posterior central diameter of adult, mid- trunk vertebrae is from about 2 to 2.8 times the central length (centrum ratio) in Desmogncithus. Leurognathus and Phaeognathus. The smallest ratio is found in the elongated Phaeognathus in which the centra are relatively short and stout. Typhlomolge and Haideotriton are short bodied, relatively weakly muscularized, aquatic genera for which the water’s buoyancy provides the major body support, and in both the vertebral centra are small. Central ratios in both are about 3, and reflect the small size of the posterior portions of the centra rather than increased central length. Related genera (Gyrinophilus, Pseudotriton, Stereochilus, Eurycea, Manculus, Typhlotriton and Hemidactylium ) have central ratios of from 2.2 to 2.5, and have neither very stout nor very elongated centra. Relatively stout vertebral centra are present in Ensatina (centrum ratio 2.5) and Aneides (cen- trum ratio 1.8 to 2.4). Plethodon may be divided into two groups: (1) an eastern group with rela- tively stout centra and ratios of 2.5 to 2.8 (cinereus, dorsalis, glutinosus , jordani, neomexicanus, ouachi- tae, richmondi, wehrlei, welleri), (2) a western group with slenderer centra and ratios of 2.9 to 3.2 (dunni, elongatus, larselli, vandykei, vehiculum). I am unable to recognize the three major types of vertebrae in Plethodon proposed by Highton (1962). Hydromantes vertebrae have ratios of 3 to 3.9, and the centra are somewhat elongated. Central diameter is relatively small. The slender, elongated species of Batrachoseps have small, semi-elongated centra with ratios of 3 to 3.5. The neotropical gen- era also have rather small centra. Pseudoeurycea and Chiropterotriton have relatively the largest centra and the lowest central ratios (2.5 to about 3). Centra are somewhat slenderer and more elongated in Bolitoglossa (ratios 2.8 to 3.5). Very slender, elongated centra with high ratios are found in the remaining genera ( Parvimolge about 3, Thorius 3.4 to 3.7, Oedipina 3.0 to 3.5, Lineatriton 4.5) . Lineatriton is an exceptionally elongated form that has maintained a primitive low number of trunk vertebrae. Elongation has involved pro- nounced elongation of the vertebral centra, and thus of the vertebrae. All vertebrae of Leurognathus have ventro- medial hypapophyses which are better developed than in any other genus. Hypapophysial keels are well developed on the anterior vertebrae of Des- mognathus, but poorly developed posteriorly. Hy- papophyses are best developed in large species (D. quadramaculatus, D. monticola), poorly devel- oped in the small species (D. aeneus, D. wrighti). The best developed hypapophyses in the above genera are anteriorly placed, knob-like processes with small, posteriorly oriented keels, and small, posteriorly placed low keels. The keels may be con- tinuous on anterior vertebrae in Leurognathus. The hypapophyses of Phaeognathus differ sharply from those of Desmognathus and Leurognathus. Well developed, uninterrupted keels extend virtually the full length of the centra on about the first seven vertebrae, and progressively lower keels, or low ridges, are present on the more posterior vertebrae. Anterior knobs, so conspicuous in the other genera, are poorly developed and obvious only anteriorly. Remaining plethodontids lack well developed hypapophyses, but remnants occur in Pseudotriton, Stereochilus, Eurycea and very large Typhlotriton. Slight ridges are present on the posterior portions of the sixth vertebral centrum of Pseudotriton, and somewhat larger ridges are present on the posterior portions of centra three through six in Stereochilus. Small, keel-like projections are present on the posterior margins of centra four or five, or both, in Eurycea bislineata, and on centrum five in E. aquatica, but none occur in other species. Old Typhlotriton have a tiny vestige on centrum five. Long, well developed, very stout, paired basa- pophyses are present on the posterior central mar- gins in Desmognathus and Leurognathus, and the processes are only slightly less well developed in Phaeognathus. Basapophyses are well developed in many Eurycea, moderately developed in Typhlo- triton, present on a few anterior vertebrae, but variously developed posteriorly in Pseudotriton, Manculus, and Typhlomolge, small and variously developed in Hemidactylium and Stereochilus, and virtually absent in Haideotriton and Gyrinophilus. Ensatina lacks basapophyses, and the processes are absent or only slightly indicated in Plethodon and Aneides; freely projecting processes are never pres- ent. The processes are absent in Batrachoseps and Hydromantes, and are irregularly distributed in the neotropical genera. Basapophyses are absent in Oedipina, Pseudoeurycea, Parvimolge, Lineatriton, and Thorius. Some Chiropterotriton (some indi- viduals of chiropterus, priscus, dimidiatus ) have variously developed and irregularly distributed, small processes. A number of presumably advanced members of the genus Bolitoglossa (lignicolor, 1966 EVOLUTION OF LUNGLESS SALAMANDERS 41 platydactyla, flaviventris, striatula, mexicana, sal- vinii) have large, well developed, projecting basa- pophyses. Other Bolitoglossa (borburata, cerroen- sis, occidentalis, rufescens, subpalmata) have occasional small processes, and the remaining species have none. Neural arch structure Neural crests, primitively present in plethodon- tids, are subject to intergeneric, serial, and onto- genetic variation. A detailed discussion will be presented at a later date. Hyperapophyses are moderately to well devel- oped on the trunk vertebrae of all plethodontids, and two conditions are encountered: (1) hypera- pophyses arise united and remain united on all, or almost all, trunk vertebrae, (2) hyperapophyses arise united or separated; if united, only on the anterior vertebrae, and separated posteriorly. Many kinds of variation are encountered, but Leurog- nathus, Phaeognathus, Gyrinophilus, Pseudotriton ruber, Eurycea multiplicata, Ensatina, Hydroman- tes italicus, H. brunus, and H. genei have condition 1. The remainder fall in group 2, except for peculiar conditions encountered in some species of Desmog- nathus (monticola, quadramaculatus) and in Pseu- dotriton montanus in which the hyperapophyses are divided on a few anterior vertebrae, but united on most vertebrae. In addition the hyperapophyses of Typhlomolge and Haideotriton are united for most of their lengths, but are separated at their tips. Hyperapophyses tend to be united in hynobiid salamanders, and united on most vertebrae in am- bystomatids. In ambystomatids a tendency for divi- sion and elongation is apparent on the posterior vertebrae. It is apparent that united processes are more primitive than separated ones in salamanders in general, and probably in plethodontids as well. The genera of plethodontids in group 1 are rela- tively primitive, based on other characters. Desmognathus, Leurognathus, and Phaeogna- thus differ from all other plethodontids in the pos- session of pterygapophyses on all or some trunk vertebrae (Fig. 9). The posterior margins of the neural arches extend posterolaterally above and beyond the postzygapophyses, and form wing-like or spine-like processes. The processes are always largest and longest on the first vertebrae, and gradually become reduced in size posteriorly. All trunk vertebrae of large D. auriculatus bear ptery- gapophyses, but the processes become progressively reduced in size posteriorly. Pterygapophyses are present on about the first seven vertebrae in D. monticola, with at least rudiments on all trunk vertebrae. In most Desmognathus pterygapophyses are well developed on the first three to five verte- brae, and are apparent to vertebrae eight to eleven. The processes are well developed only on the first one or two vertebrae of D. quadramaculatus, and are not apparent past the sixth vertebrae. No ptery- gapophyses are present in D. wrighti. Well devel- oped processes that extend as spinous projections beyond the zygapophyses are present on the first four vertebrae of Phaeognathus, but the elements are reduced posteriorly and are not apparent past the ninth vertebra. Moderately developed ptery- gapophyses are present on the first vertebrae in Leurognathus, but they are very reduced by the fourth and only slightly indicated on more posterior vertebrae. The functional significance of pteryga- pophyses appears to lie in providing additional attachment sites for dorsal spinal muscles, which raise the skulls of the three genera to open their mouths. Transverse processes Diapophyses and parapophyses are long and well developed in Desmognathus, Leurognathus, and Phaeognathus, and extend far beyond the lateral zygapophysial margins. The posterolaterally di- rected processes are relatively straight. The centra of mid-trunk vertebrae are 0.7 to 0.9 times the distance across the parapophysial tips (centrum- parapophysial ratio). The parapophyses and dia- pophyses are entirely free, or are joined proximally only, in Desmognathus and Leurognathus, but are joined by thin bony webbing for their entire lengths in Phaeognathus. Parapophyses are directly below or just slightly anterior to the diapophyses. Alar expansions are usually present on the anterior margins of the parapophyses, especially on anterior and mid-trunk vertebrae. Alar processes are par- ticularly well developed with distinct anterior pro- jections in D. monticola, D. quadramaculatus, Leurognathus, and Phaeognathus. Well developed dorsal processes are present on the diapophyses and parapophyses of D. monticola, and slightly smaller processes are evident in D. quadramaculatus and Leurognathus. The processes correspond with raised processes on both rib heads. Centrum-parapophysial ratios are about 0.8 to 0.85 in Pseudotriton, Gyrinophilus, and Stereochi- lus, about 1 in Typhlomolge and Hemidactylium, and 1 to 1.4 in Eurycea, Manculus, Typhlo- triton, and Haideotriton. Parapophyses and dia- pophyses are moderately long and extend beyond the zygapophyses in Gyrinophilus, Pseudotriton, Stereochilus, Typhlomolge, and Hemidactylium, but are very shortened and do not extend (dia- pophyses) beyond the zygapophyses in Eurycea, 42 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 Manculus, Typhlotriton, and Haideotriton. The diapophyses and parapophyses are fused proxi- mally but free distally in Pseudotriton, Gyrinophi- lus, and Stereochilus, and are free for their entire lengths in Eurycea, Manculus, Typhlotriton, Typh- lomolge, Haideotriton, and Hemidactylium. Para- pophyses are slightly anterior to the diapophyses in Gyrinophilus, Pseudotriton, Stereochilus, and Hemidactylium, but are greatly in advance of the diapophyses in Eurycea, Manculus, Typhlotriton, Typhlomolge, and Haideotriton. In Gyrinophilus and allies the posterior margins of the parapophy- ses are located posterior to the anterior margins of the diapophyses. In Eurycea and allies the dia- pophyses are borne near the midpoints of centra, as in other plethodontids, but the parapophyses arise from the anterior one-third of the centra and are entirely in advance of the diapophyses. The ad- vanced position of the parapophyses and the very short transverse processes of Eurycea and allied genera are unique characters among the pletho- dontids. Centrum-parapophysial ratios of Plethodon and Aneides are from 0.85 to 1.2, and those of Ensatina are about 0.8. Diapophyses and parapophyses are moderately long and extend well beyond the zyga- pophyses. Transverse processes are separated completely in Ensatina, most Plethodon, and some individuals of Aneides; joined proximally but sep- arated distally in some Plethodon (xvehrlei, ciner- eus) and Aneides (aeneus, ferreus); and joined almost to their tips in some Plethodon (elongatus, neomexicanus, richmondi) and Aneides (flavipunc- tatus, hardii, lugubris). Parapophyses are either directly below or just slightly anterior to the dia- pophyses at their origins. Transverse processes are generally rather straight, and extend laterally with slight posterior angles (less than 30°) from the perpendicular with the vertebral axis. Alar expan- sions are generally absent, but relatively broad, shelf-like structures extend from the centra to points near the tips of the parapophyses both an- teriorly and posteriorly in A. lugubris, and, to a lesser extent, in A. flavipunctatus. Centrum-parapophysial ratios are 0.8 to 0.9 in Hydromantes, and the transverse processes are long, extending well beyond the zygapophyses. The transverse processes are initially straight, but may bend posteriorly past the midpoints of the centra. The processes are almost entirely separated in H. platycephalus and H. shastae, but may be joined for about one-half their lengths in H. brunus. The transverse processes of the European Hydro- mantes (italicus, genei) are joined proximally or are entirely separated on the first two vertebrae, joined to their tips on the third vertebra, and com- pletely fused to form cylindrical processes with single distal articular facets on succeeding verte- brae. Apparently the parapophyses contribute most to the single processes, and the diapophyses are lost. Centrum-parapophysial ratios are from 0.85 to 1.2 in Batrachoseps, and the transverse processes are long, extending beyond the zygapophyses. The processes are initially straight, but bend posteriorly just past their midpoints in a manner similar to Hydromantes. Transverse processes are separated for most of their lengths, but are joined for short distances proximally. As in Hydromantes the para- pophyses are almost directly below the diapophyses. Alar expansion in Batrachoseps and Hydromantes is virtually nonexistent. Centrum-parapophysial ratios are from 0.85 to 1.25 in Bolitoglossa, Oedipina, Pseudo eurycea, and Chiropterotriton, slightly more than 1 in Parvi- molge and Thorius, and about 1.3 in Lineatriton. The Lineatriton ratio reflects elongated centra rather than shortened transverse processes. Trans- verse processes of all neotropical genera extend slightly to well beyond the zygapophyses. Trans- verse processes are completely separated or joined proximally only in Chiropterotriton, Parvimolge, Thorius, most Pseudoeurycea, and many Bolito- glossa. The processes are joined for most of their lengths in some Pseudoeurycea (altamontana, smithi) and many Bolitoglossa (colonnea, dunni, engelhardti, marmorea, morio, orestes, platydac- tyla, rostrata, savagei, subpalmata). The processes are initially separated in Lineatriton, but are com- pletely joined past the third vertebra, and the diapophyses are lost by the fourth or fifth vertebrae; single processes are present on succeeding verte- brae. A similar situation is encountered in Oedipina in which single processes are present by the third or fourth vertebrae. Lineatriton and Oedipina re- semble the European species of Hydromantes in this regard, and the character is not found else- where in the family. The parapophyses are borne almost directly below, or slightly in advance of, the diapophyses in most neotropical species. The parapophyses of some groups (some Chiroptero- triton and especially Parvimolge ) may be distinctly anterior to the origin of the diapophyses. Trans- verse processes are usually initially straight, but are slightly to distinctly bent posteriorly past their midpoints. The processes extend primarily laterally, but are slightly to moderately swept posteriorly in most genera, and are very strongly swept posteriorly in Lineatriton and Oedipina. Exceptions are O. complex and O. parvipes in which the processes are 1966 EVOLUTION OF LUNG LESS SALAMANDERS 43 almost perpendicular to the vertebral axis. Slight anterior alar expansions are present on the para- pophyses of several Chiropterotriton (cibscondens, multidentatus, nasalis). Very large alar expansions that form moderately broad shelves are present in front of the parapophyses in Lineatriton, and much smaller expansions are present posteriorly. Alar expansions are evident on the anterior margins of the parapophyses of several Oedipina (inusitata, complex, bonitciensis, pacificensis, longissima), and may be drawn into anteriorly directed, pointed processes. Alar expansions are usually present but poorly developed in other Oedipina, and are poorly developed or absent in other neotropical genera than those discussed (see Tanner, 1950, for illustrations) . Caudal vertebrae Tails of plethodontid salamanders are of three general types: 1. Very stout proximally, tapering distally. Des- mognathus, Leurognathus, Phaeognathus, Gy- rinophilus, Stereochilus, Typhlomolge, Haideo- triton. 2. Relatively slender proximally, tapering some- what distally. Eurycea, Manculus, Typhlotri- ton, Plethodon, Aneides, Hydromantes. 3. Constricted proximally, expanding beyond con- striction, but tapering toward distal tips. Hemi- dactylium, Ensatina, Bolitoglossa, Oedipina, Pseudoeurycea, Chiropterotriton, Parvimolge, Lineatriton, Thorius. Batrachoseps has slight indications of tail con- striction, and rather bridges the gap between types 2 and 3. Some Eurycea (e.g., E. longicauda) bridge types 1 and 2. First caudal vertebrae of Ensatina and Hemi- dactylium, and to a lesser extent, the neotropical genera, are highly specialized structures of impor- tance in tail autotomy (see also Wake, 1963.) Detailed discussion is deferred to a later date. Midcentrally located transverse processes are borne on virtually all caudal vertebrae of Desmog- nathus, Leurognathus, and Phaeognathus, and the processes are relatively stout, particularly in Leurognathus. Transverse processes at midcentrum are mod- erately developed and present to about the tenth to eleventh caudal vertebrae of Gyrinophilus, Pseudotriton, Manculus, and Typhlomolge, to about the sixteenth of Stereochilus, and almost to the end of the tails of Eurycea and Typhlotriton. Small Manculus and Eurycea, and all Haideotriton, lack caudal transverse processes. Hemidactylium has transverse processes on about the first nineteen vertebrae, and all but the first are midcentral. Ensatina has rather poorly developed processes past the first vertebrae, and none occur past the tenth or eleventh vertebrae; all but the first few are located midcentrally. Transverse processes are somewhat better developed in Plethodon, and vir- tually all are midcentral. Midcentral processes are moderately developed and present on most caudal vertebrae of Aneides. Cylindrical, midcentral transverse processes are present on most caudal vertebrae of Hydromantes. Batrachoseps has rather complex, broad-based, strutted processes on most caudal vertebrae in intermediate positions; some processes are at the anterior ends of the centra, while others are slightly posterior but ahead of midcentrum. Bolitoglossa, Oedipina, Pseudoeurycea, Chiropterotriton, Par- vimolge, Lineatriton, and Thorius all have trans- verse processes on virtually all caudal vertebrae; the processes are located at the anterior ends of the centra and are swept moderately to strongly anteriorly. Unusual neural crests are found in Oedipina; usually only the first one or two caudal vertebrae (but as many as ten) have single medial crests. Commencing on about the second vertebra the crests divide and form double, parallel crests on either side of the midline. The crests are broadest basally and angle toward one another dorsally. Double caudal crests are constant, unique charac- ters of Oedipina. Ribs Ribs vary considerably in length in the family, and also vary serially within individuals. Mid-trunk ribs are from 0.6 to 0.75 times the transparapo- physial distance (rib-parapophysial ratio) in Des- mognathus and Leurognathus, but are relatively short in Phaeognathus (0.4). Ratios are from 0.65 to 0.8 in Gyrinophilus, Pseudotriton, Stereochilus, and Hemidactylium, about 0.9 in Typhlomolge, about 1 to 1.2 in Manculus, Typhlotriton, and Haideotriton, and from a little over 1 to about 1.5 in Eurycea. Ensatina has relatively short ribs (0.6), but ribs are longer in Plethodon (0.67 to about 1) and longer still in Aneides (0.85 to 1.25). Hydro- mantes and Batrachoseps have moderately short ribs with ratios of about 0.6 to 0.75. A distinct trend toward rib shortening is encountered in the neotropical genera. The trend is best exemplified by the genus Chiropterotriton in which relatively 44 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 primitive species have moderately long ribs ( prisons , ratio 1), but more specialized species have short ribs ( abscondens , ratio 0.35-0.4). Rib-parapo- physial ratios are about 0.75 to 0.8 in Thorius, 0.4 to 0.75 in Pseudoeurycea (lowest figure in alta- montana, highest in werleri), 0.4 to 0.67 in Boli- toglossa, (lowest figure in adspersa, highest in engelhardti) , 0.9 in Parvimolge townsendi, and 0.5 in P. richardi. Ribs are especially shortened in Oedipina (0.2-0. 3) and Lineatriton (0.2-0.25). Members of the latter two genera are elongate, attenuate forms, and the shortened and reduced ribs are probably related to the change in habitus. Rib heads are of moderate length and are well separated in most genera. The heads are greatly separated and rather long in Eurycea, Manculus, Typhlotriton, Typhlomolge, and Haideotriton. Rib heads are rather short in Batrachoseps. Fusion of the rib heads occurs in Aneides lugubris, Hydro- mantes brunus, Phaeognathus, some Batrachoseps attenuatus, Pseudoeurycea altamontana, Bolito- glossa orestes, and B. subpalmata. Usually the last one or two ribs of all species are unicipital. In Hydromantes genei, H. italicus, and in all species of Oedipina only the first three or four pairs of ribs are bicipital, and the remainder are unicipital. All but the first four or five pairs of ribs are unicipital in Lineatriton. Limb Elements Primitively plethodontid limb bones have car- tilaginous heads, but calcifications occur in all Thorius, Lineatriton, Parvimolge, and some indi- viduals of Oedipina. Apparently these calcifications are strengthening compensations for paedomorphic weakening. Relative limb length is highly variable, and description is made more complex by allometric growth and ontogenetic, sexual, geographic, inter- specific, and intergeneric variation. Certain of the attenuate genera (Phaeognathus, Stereochilus, Manculus, Batrachoseps, Oedipina, Lineatriton, Parvimolge, Thorius ) have obviously shortened limb elements. Long limbs tend to be the end result of trends toward increased scansorial activity in several forms (Eurycea lucifuga, Aneides lugubris, A. aeneus, A. ferreus). Moderately long limbs are present in Ensatina, Hydromantes, and primitive species of Plethodon and Bolitoglossa. Hind limbs are much longer and stouter than the fore limbs in Desmognathus, Leurognathus, and Phaeognathus. Fore limbs are a little shorter than the hind limbs in other genera, but the limbs are of about equal length in Ensatina. Femurs are always longer than humeri except in Ensatina where the situation is reversed. Limb length apparently increases as the species of Eurycea become increasingly troglobitic (see Mitchell and Reddell, 1965). The relatively longest limbs in the family are encountered in the extremely specialized troglobite, Typhlomolge. The elongated limbs of troglobitic species may be adap- tations which enable the salamanders to walk about quiet cave pools with their bodies elevated above the bottom debris. Distinct, well developed tibial spurs are present in most genera. The spurs are truncate and reduced in Typhlotriton, and are small and very slender in Typhlomolge and Haideotriton. Ensatina lacks tibial spurs, but well defined dorsal ridges are present. Very small tibial spurs, borne on the mid- sections of the bones and falling far short of the proximal ends, are found in Hydromantes. The spurs are present in most species of Batrachoseps, but are absent in most populations of B. attenuatus. A trend toward spur reduction is apparent in neo- tropical genera, reduced spurs being present in Lineatriton, Parvimolge townsendi, most Pseudo- eurycea, Chiropterotriton, and Thorius, and in some Bolitoglossa (dunni, engelhardti, helmrichi, rufescens). Spurs are absent in all Oedipina, most Bolitoglossa, some Chiropterotriton (bromeliacia, nasalis, xolocalcae), and some individuals of various species of Thorius and Pseudoeurycea. Mesopodial Elements Primitively eight carpals and nine tarsals are present (Desmognathus, Leurognathus, Phaeogna- thus, Gyrinophilus, Pseudotriton, Stereochilus, Eurycea, Typhlotriton, Haideotriton, Typhlomolge, Plethodon, Ensatina, Hydromantes, Pseudoeury- cea). All have primitive carpi with ulnars and centrals separated except Hydromantes and Pseu- doeurycea in which the elements are in contact. All of the above genera have primitive tarsi with fourth distal tarsals much larger than the fifth, which are the only tarsals that do not touch the centrals. Aneides differs from the above genera in that the ulnars articulate extensively with the centrals, and the fifth distals are the largest tarsals and articu- late broadly with the centrals. The intermedia and ulnars are fused in all carpi of A. hardii, and seven carpal elements are present. The remaining genera have reduced numbers of carpals, tarsals, or both, in some or all species. Batrachoseps, Manculus, and Hemidactylium have lost their fifth toes, and have also lost the fifth distal tarsals. Manculus, Hemidactylium, Batra- choseps wrighti, and some populations of B. pacifi- cus have eight carpals, but other Batrachoseps have 1966 EVOLUTION OF LUNGLESS SALAMANDERS 45 seven, the result of fusion of ulnars and intermedia. Tarsal patterns similar to those in Aneides are encountered in several Chiropterotriton (abscon- dens, arboreus, chiropterus, dimidiatus, multiden- tatus, priscus), but in two advanced species (xolo- calcae, bromeliacia) the primitive plethodontid pattern is found. Another advanced species, C. nasalis, has seven carpals and eight tarsals as a result of radial-intermedial and fourth and fifth tarsal fusions. C. abscondens may have seven or eight mesopodial elements as the result of ulnar- intermedial or third and fourth carpal, and fourth and fifth tarsal fusions. All ankles of three indi- viduals have seven tarsals; wrists usually have seven carpals, but one wrist each of two individuals has eight. The tarsal patterns unique to Aneides and the primitive species of Chiropterotriton apparently have functional and evolutionary significance. In the primitive plethodontid tarsus the lines of force from the first, second, third, and fourth digits are directed through the first three distal tarsal elements to the large centrals. Lines of force from the fifth digits are directed through the relatively small fourth distal elements (distal tarsal 5) to the fibu- lars. The relative weakness of this arrangement is illustrated by parallel loss of the fifth distal tarsals and digits in three genera. The arrangement in Aneides and Chiropterotriton is one in which all lines of force are relatively straight, and all are directed to a single pivotal point in each tarsus, the centrale (see illustrations in Wake, 1963). Since the lines of force emanate from single points, they may be swung or rotated about these points. Rela- tively greater spread across the outer digital tips than in the primitive arrangement is thus possible; the increased spread has obvious advantages for organisms with climbing propensities. Both Aneides and Chiropterotriton contain species that are adept climbers. Apparently their mesopodial arrangement is a prospective adaptation that has significant adaptive value in arboreal and scansorial but not terrestrial organisms. Parvimolge townsendi has six or seven carpal elements as a result of ulnar-intermedial and pos- sible fourth distal carpal-central fusions. Distal tarsals 4 and 5 have fused, and eight tarsal elements are present. A peculiar and unique situation is encountered in Lineatriton. Eight tarsal elements are present and distal tarsal 5 has fused with the fourth or has been lost; metatarsal 5 extends into the area seemingly vacated by the tarsal. Eight carpals are present. Ulnar-intermedial and fourth distal carpal- central fusions result in six carpals in Thorius. Distal tarsals 4 and 5, the intermedia and fibulars, and distal tarsals 4 and 5 and the centrals may fuse, and six to eight tarsals result. Intraspecific varia- tion is present in Thorius, and individuals may vary by one element from one side to the other. Eight tarsal elements are present in most Bolito- glossa as a result of fusion of distal tarsals 4 and 5. A tendency for elimination of the third distal car- pals from their articulation with the centrals, and subsequent reduction of carpal size, is apparent within the genus, especially in the group of species associated with B. dunni and B. engelhardti. Distal carpals 1 and 2 meet distal carpals 4 and distal carpals 3 are effectively walled-off from articulation with the centrals. The trend culminates in B. occi- dentalis and B. rufescens in which distal carpals 3 and 4 fuse to form single, strengthened, more efficient elements, and seven carpals result. A simi- lar situation is encountered in the tarsi of B. occi- dentals and B. rufescens, in which distal tarsals 3, 4, and 5 fuse, and seven tarsal elements are present. This condition is also encountered in one of three B. platydactyla examined. The most extreme mesopodial specializations occur in Oedipina. The limbs, hands and feet are greatly reduced in size, and are considerably less functional than in primitive members of the family. From five to seven carpal and from five to eight tarsal elements are present as the result of the following fusions (arranged in probable fusion sequence) : ulnar-intermedial, distal carpal 4- central, distal carpals 1, 2, and 3; distal tarsals 4 and 5 (fused in all species), distal tarsals 4, 5, and central, fibular-intermedial. Species of Oedipina vary as follows in the number of mesopodial fusions: bonitaensis 6 carpals, 7 tarsals; complex 6, 6; cyclocauda, uniformis 7, 7-8; gracilis 5, 5-7; inusitata 6, 6; longissima 6, 7; pacificensis 6, 7; poelzi 7-8, 7-8; syndactyla 6, 6; sp. nov. 7, 7-8. In Oedipina in which the third distal tarsals are not fused with some other elements, they are very small and resemble the condition in Thorius. Fur- ther similarities of the two genera are the reduction in size of distal carpals and tarsals 1 and 3, and shifts of the articulation of the first metacarpals and metatarsals from distal elements to centrals 1. The latter condition is also found in some special- ized Chiropterotriton (e.g., abscondens). Mesopodial calcifications of probably functional and phylogenetic significance occur in certain plethodontids. Cope (1869) stated that Thorius has ossified mesopodial elements, and differs in this character from other plethodontids. Rabb (1955) reported calcified mesopodials in Parvi- 46 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 molge townsendi. Uzzell (1961) found mesopodial calcifications in only twelve of thirty-four adult Thorius, thirteen of seventeen Parvimolge town- sendi, and three of thirteen Lineatriton. He found eight tarsals calcified in twelve ankles of seven in- dividuals of Thorius, with lesser numbers in other species. Uzzell suggested the following calcification sequences: (1) carpus, intermedium-ulnare; radiale, centrale 1, or carpals 1 and 2; carpal 3; (2) tar- sus. outer-most tarsal (either tarsals 4-5, or tarsals 4-5-centrale) ; intermedium; fibulare, or centrale; tibiale, or tarsals 1-2; tarsal 3, or centrale 1. Uzzell used slightly different terminology and calls cen- trals 1 of the carpi and tarsi the first carpals and tarsals, respectively. Sequence of calcification in Parvimolge and Lineatriton was not determinable, but Uzzell reported six elements with as many as eight centers of calcification in Parvimolge carpi, and eight tar- sal elements with as many as six calcifications. Lineatriton was reported to have from one to five carpal and from one to three tarsal calcifications. Uzzell (1961) suggested that fusion of the inter- media and ulnars, and close approximation of the intermedia and fibulars reflect concentration of locomotor forces. In both wrists and ankles the outer or postaxial mesopodials fuse before the inner or preaxial elements, and it is likely that these are the elements of greatest importance in force trans- mission. Mesopodial fusions are far more common in the preaxial than in the postaxial portions of the carpi and tarsi throughout the family. However, fusions (e.g., distal elements 1 and 2) are present in the preaxial portions of all plethodontids, and using Uzzell’s reasoning for postaxial regions, the preaxial area might be the most significant primi- tively with the postaxial portion gaining importance secondarily. Rabb (1955) interpreted several characters of Parvimolge townsendi to be the result of paedo- morphosis, and Uzzell (1961) suggested that calci- fied mesopodials may compensate for the paedo- morphic and presumably weak condition of the adult feet of Lineatriton, Parvimolge, and Thorius. Uzzell also suggested that calcified mesopodials of Parvimolge and Lineatriton reflect possible close relationship, but thought the conditions in Thorius might be due to parallelism. The only material available to me of Parvimolge townsendi has been that used by Uzzell. I am unable to add to his data. Some, all, or none of the carpals and tarsals of Thorius may be calcified. The maxi- mum numbers of calcifications observed by me in Parvimolge are five carpals (ulnare-intermedium, centrale, centrale 1, distal carpal 4, radiale) and seven tarsals (fibulare, intermedium, centrale, cen- trale 1, distal tarsals 1-2, 3, 4-5). I have not seen calcified mesopodials in Lineatriton. Calcifications occur in several additional genera. Mesopodial calcifications observed include: fibu- lars of one Desmognathus quadramaculatus; fibu- lars and tarsal intermedia of two Desmognathus ocoee (three ankles); carpal and tarsal intermedia and fibulars of a single Leurognathus; tarsal inter- media, fibulars, distal tarsals 5, one radiale, and one distal tarsal 1-2 of one Eurycea bislineata; tarsal intermedia and fibulars of one Typhlotriton; tarsal and carpal intermedia, ulnars, fibulars, tar- sal centrals 1, distal carpals 1-2, 3, 4, and distal tarsals 1-2, 3, 4, 5 of a single Ensatina; distal carpals 1-2 of a single Plethodon neomexicanus; distal carpals 1-2 and one distal tarsal 1-2 of a single Pseudoeurycea unguidentis; one ulnare, fibu- lars, one tibiale, distal carpals 4, one distal tarsal 1-2, and distal tarsals 4-5 of one Chiropterotriton abscondens; and ulnar-intermedia, fibular-inter- media, tibials, carpal centrals, centrals 1, radials, distal carpals, 1-2, distal tarsals 4-5, and one distal tarsal 1-2 of a single Oedipina sp. nov. It is apparent that mesopodial calcifications have appeared independently in several evolutionary lines, but calcification is certainly more common in Parvimolge townsendi and Thorius, and prob- ably in Lineatriton, than in the other plethodontid genera. Postaxial calcifications predominate, but preaxial calcifications appear first in some indi- viduals. The most serious objections to Uzzell’s (1961) suggestions concerning the functional sig- nificance of mesopodial fusions and calcifications are the preaxial distal elements 1 and 2 fusions universal in the family, and the tendency for loss of the fifth toes and fifth distal tarsals in the post- axial regions in several evolutionary lines. Digits Primitive plethodontids have four fingers and five toes, and the primitive phalangeal fomulae are 1 -2-3-2 for each hand and 1-2-3-3-2 for each foot. Such formulae are characteristic of all Desmogna- thus, Leurognathus, Phaeognathus, Gyrinophilus, Pseudotriton, Stereochilus, Eurycea, Typhlotriton, Haideotriton, Typhlomolge, Ensatina, Aneides, Hydromantes, Pseudoeurycea, Chiropterotriton, and Parvimolge. Hemidactylium, Manculus, and Batrachoseps have lost the fifth toes and one pha- lanx from each fourth toe; the foot formula is 1-2-3-2 (rarely 0-2-3-2 in B. attenuatus) . Single phalanges have been lost from the fifth toes of most individuals of two species of Plethodon (larselli, neomexicanus ) and the formula is 1-2-3-3-1; all 1966 EVOLUTION OF LUNGLESS SALAMANDERS 47 other Plethodon have the primitive formula. Some Lineatriton have the primitive formula, but reduc- tions may occur to 1 -2-3-1 in the hand and 1 -2-3-2- 1 in the foot. Thorius may have the primi- tive formulae, but intra- and interspecific variation occurs and formulae variations are: hand (0-1) -2- 3-( 1-2) , foot (0- 1 ) -2-3- (2-3 ) - ( 1 -2) . Most species of Bolitoglossa retain the primitive formulae, but B. occidentalis has lost one phalanx from each fourth toe (foot 1-2-3-2-2), and formulae varia- tions in B. rufescens are: hand (0-1 )-( 1-2) -(2-3)- (1-2), foot l-(l-2)-(2-3)-(2-3)-(l-2). Extensive phalangeal reduction occurs in Oedipina, in which only certain O. poelzi and O. sp. nov. retain the primitive formulae; formulae variations are: hand (0-1) -(1-2) -(2-3) -(1-2), foot (0-1 ) -2-(2-3 ) - (2- 3) -(1-2). The commonest reductions occur in digits one, four, and five of the feet, and digits one and three of the hands. A tendency for reduction of phalangeal elements and loss of digits in the postaxial portions of the hands and feet is evident. Reduction is most com- monly encountered in attenuate, short-limbed species. Uzzell (1961) has suggested that the post- axial portions of the carpi and tarsi are the areas of greatest transmission of forces from the lower limbs to the digits, but postaxial phalangeal and digital reductions suggest that other areas may be as important. SYSTEMATIC SUMMARY Family Plethodontidae Definition: adults lungless with nasolabial grooves, no bony pterygoids, no lacrimals, otic and occipital elements fused, columellae fused with opercula or absent, vomers with anterior laterally oriented tooth series and posterior multiple series borne below parasphenoids, no angulars; larvae present or absent, if present with three or four epibran- chials, fused premaxillae, teeth on premaxillae, vomers, palatopterygoids, dentaries, and coronoids. Range: eastern and western North America with some species in central regions; Middle America from Tamaulipas and Nuevo Leon in northeastern Mexico and Nayarit in western Mexico to eastern Brazil and central Bolivia; the Maritime Alpes of southeastern France, the Piedmont of Italy, and the island of Sardinia. Content: two subfamilies, twenty-three genera, about one hundred eighty-four species. Subfamily Desmognathinae Definition: plethodontid salamanders having four rather than three larval epibranchials, a unique mouth opening mechanism by means of which the mandibles are held relatively rigid and the skulls proper are raised, extensive skeletal and muscular modifications related to functional changes in mouth opening mechanics, and pterygapophyses and well developed hypapophysial keels on ante- rior trunk vertebrae. Characterization: (1) skulls rigid, depressed, streamlined, composed of dense, heavy bones, (2) larvae lack protruding gill rami, (3) larval skulls composed of ossified premaxillae, frontals, parietals, occipito-otics, opercula, vomers, pala- topterygoids, parasphenoids, quadrates, squamo- sals, dentaries, prearticulars, and coronoids, (4) four larval epibranchials, (5) larval vomers in broad median contact, (6) larval palatopterygoids well separated from quadrates, (7) unicuspid larval teeth, bicuspid adult teeth, (8) single larval and adult premaxilla, (9) premaxillae very broad and depressed, internasal fontanelles small or closed over, (10) frontal processes of premaxillae stout and broad, arise broadly from toothed portions, (11) maxillae long, extend beyond eyeballs, small to large posteromedially oriented jugal processes present, (12) maxillary facial lobes very large, articulate extensively with frontals, (13) palatal shelves of premaxillae and maxillae very broad, (14) septomaxillae present, but reduced, (15) nasals small, articulate firmly with maxillae and premaxillae, (16) prefrontals absent, (17) vom- erine preorbital processes toothless, (18) anterior vomerine teeth present or absent, (19) anterior vomerine teeth discontinuous, (20) posterior vom- erine teeth borne in dense, separated patches, (21) anterior portions of frontals very large, expanded, occupy prefrontal and antorbital regions, (22) frontals flat dorsally with sharp angular ridges along orbital borders, bones uncovered by muscu- lature, (23) deep parietal-otic troughs extend as bony processes into orbits, (24) well developed crests border posterolateral margins of parietal- otic troughs, (25) large posterolateral parasphe- noid processes, (26) parasphenoids relatively short, abut against posterior margins of vomers in simple articulations, (27) quadrates very large, articulate with parasphenoids and occipito-otics, cartilagin- ous portions of suspensoria relatively small, (28) orbitosphenoids large, well developed, quadrangu- lar, (29) squamosals large, solidly anchored to skulls, (30) occipital condyles stalked, (31) colu- mellae short, stocky, rather flattened, (32) denti- tion well developed, (33) meckelian grooves essentially closed, dentaries O-shaped anteriorly, (34) detoglossal; genioglossal muscles present, (35) cornua elongate, attaching to first basibran- 48 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 chials behind anterior extensions of latter, (36) lingual cartilages absent, (37) well developed first basibranchial anterior extensions, (38) no marked anterior first basibranchial expansion, (39) first ceratobranchials longest of articulated hyobran- chial elements, (40) first ceratobranchials consid- erably larger than second, (41) epibranchials less than 1.25 times first basibranchials, and shorter than first ceratobranchials, (42) second cerato- branchials short, less than two-thirds length of first basibranchials, (43) atlases large and stout, bear very large, markedly concave, condylar articular facets, (44) odontoid processes of atlases very small, (45) posterior margins of atlases very large, broad, markedly raised, (46) fifteen or twenty-one to twenty-three trunk vertebrae, (47) two caudo- sacral vertebrae, (48) larval vertebrae structurally amphicoelous, intervertebral cartilages become calcified in adults and resemble classical opistho- coelous condition, (49) hypapophysial keels well developed on at least the first few trunk vertebrae, or on all trunk vertebrae, (50) hyperapophyses well developed, tend to be united, but may be divided on anterior vertebrae, (51) well developed, stout basapophyses on posterior central margins, (52) pterygapophyses present on some anterior trunk vertebrae, (53) parapophyses and diapo- physes long, separated or fused, (54) centra shorter than distance across tips of parapophyses, (55) ribs usually absent on last trunk vertebrae, (56) ribs less than 0.75 times distance across parapo- physes, (57) tails stout proximally, no basal con- striction, (58) well developed tibial spurs, (59) hind limbs much larger than fore limbs, (60) fingers and toes long and slender with distinctive tendon arrangements and fleshy palms, (61) eight carpals and nine tarsals, (62) distal tarsals 5 do not articulate with central tarsals, (63) four fingers and five toes, (64) phalangeal formulae 1-2-3-2, 1-2-3-3-2, (65) adults aquatic to terrestrial, (66) aquatic larvae in most, but trends toward direct terrestrial development, (67) pigmented, lobed testes in old males, (68) small gularis muscles present posterior and external to very large quad- rato-pectoralis muscles. Content: three genera — Desmognathus, Leurog- nathus, Phaeognathus. Desmognathus Baird, 1850 Definition: desmognathine salamanders with fif- teen trunk vertebrae; small maxillary jugal proc- esses; internal nares open medially; vomerine preorbital processes well separated from vomers proper, and not extending laterally beyond vomer bodies; an internasal fontanelle; vomers and pre- maxillae joined by septum-like processes which form fontanelle borders; hypapophysial keels well developed on anterior trunk vertebrae, not present or only slightly indicated on posterior vertebrae; neural crests present only on first two or three trunk vertebrae; parapophyses and diapophyses separated for most of lengths; ribs more than one- half the distance across parapophyses; stout, stocky habitus; limbs moderately long; aquatic larvae, or direct development. Range: eastern North America, particularly in Appalachian Mountain region. Content: aeneus Brown and Bishop, 1947; auri- culatus (Holbrook, 1838); fuscus (Rafinesque, 1820); monticola Dunn, 1916; ochrophaeus Cope, 1859; ocoee Nicholls, 1949; quadramaculatus (Holbrook, 1840); wrighti King, 1936. Leurognathus Moore, 1899 Definition: desmognathine salamanders with fifteen trunk vertebrae; small maxillary jugal proc- esses; internal nares open laterally near lateral margins of vomers; vomerine preorbital processes slightly separated from vomers proper, extend laterally beyond vomer bodies; only in young an internasal fontanelle which becomes progressively closed during ontogeny; skulls strongly depressed, premaxillae and vomers in close proximity with no septum-like connecting processes; hypapophysial keels strongly developed on all trunk vertebrae, particularly anteriorly; neural crests present only on first two or three trunk vertebrae; parapophyses and diapophyses separated for most of lengths; ribs more than one-half the distance across the parapophyses; stout, stocky habitus; limbs moder- ately long; aquatic larvae and adults. Range: southern Appalachian Mountains of eastern United States. Content: marmoratus Moore, 1899. Phaeognathus Highton, 1961 Definition: desmognathine salamanders with twenty-one to twenty-three trunk vertebrae; very large, expanded maxillary jugal processes that extend almost to quadrates; internal nares open medially; vomerine preorbital processes well sep- arated from vomers proper, extend laterally beyond vomer bodies; an internasal fontanelle; vomers and premaxillae joined by septum-like processes which form fontanelle borders; anterior cranial elements sculptured; hypapophysial keels present on all trunk vertebrae, best developed anteriorly; neural crests present on all trunk vertebrae; parapophyses and diapophyses joined for virtually entire lengths; ribs less than one-half the distance across the 1966 EVOLUTION OF LUNGLESS SALAMANDERS 49 parapophyses; stout but elongated habitus with extremely long trunks and long, attenuated tails; limbs and digits markedly shortened; life history unknown, but adults are terrestrial burrowers. Range: known only from the pine woods of the Coastal Plain in south-central Alabama. Content: hubrichti Highton, 1961. Subfamily Plethodonlinae Definition: plethodontid salamanders having less than four larval or embryonic epibranchials, normal salamander mouth opening mechanisms in which the skulls proper remain relatively rigid and the mandibles are lowered, and trunk vertebrae which lack pterygapophyses and well developed hypapophyses. Characterization: (1) skulls variable in shape and rigidness, usually relatively broad, moderately high, and not strongly constructed; distinct tenden- cies in many groups for skull reduction and weak- ening, (2) larvae with protruding gill rami, (3) larval skulls comprised of ossified premaxillae, frontals, parietals, occipito-otics, opercula, vomers, palatopterygoids, parasphenoids, quadrates, squa- mosals, dentaries, prearticulars, and coronoids, (4) three larval epibranchials, (5) larval vomers in narrow anteromedial contact, well separated posteriorly, (6) larval palatopterygoids with long, posteriorly oriented processes that approach or articulate with the quadrates, (7) unicuspid larval teeth, bicuspid or specialized unicuspid adult teeth, (8) a single larval premaxilla, adults with single or paired elements, (9) premaxillae highly variable in shape, broad to narrow with a tendency for depression; internasal fontanelles usually large, tend to close during ontogeny of a few species, (10) premaxillary frontal processes highly vari- able, usually arise narrowly from toothed portions, (11) maxillae of variable lengths but usually fall short of posterior eyeball margins, no jugal proc- esses, (12) maxillary facial lobes moderate to small, frontal-maxilla articulation normally absent, (13) palatal shelves moderately broad in some species, narrow in most, very small in some, (14) septomaxillae primitively large, tend to be reduced or lost in one generic group, (15) nasals small to large, articulations highly variable from genus to genus, but articulation with maxillae usually neither firm nor extensive, (16) prefrontals primitively present, lost in a few genera, (17) vomerine pre- orbital processes present in most, absent or reduced in a few; teeth may or may not extend onto proc- esses, but do so primitively, (18) anterior vomer- ine teeth present, (19) anterior and posterior vom- erine teeth continuous in primitive members of one group, discontinuous in adults of all others, (20) posterior vomerine teeth borne sparsely or densely on narrow to broad, continuous or discontinuous posterior vomerine projections, (21) anterior por- tions of frontals moderately large and expanded, do not occupy antorbital regions and rarely occupy prefrontal regions, (22) frontals moderately rounded with no angular orbital ridges, frontals slightly to extensively covered by adductor mandi- bulae anterior musculature, (23) no parietal-otic troughs, but many have posterior parietal depres- sions, (24) well developed squamosal-otic and parietal-otic crests in some genera, others have various types of otic crests, or no crests, (25) posterolateral parasphenoid processes moderate to small, or absent, (26) parasphenoids relatively long, broadly overlap vomers, (27) quadrates moderate to small, do not articulate with either occipito-otics or parasphenoids, cartilaginous por- tions of suspensoria moderately large, (28) orbito- sphenoids large, moderate, or small, posteriorly constricted in some genera, (29) squamosals large to small, anchorage to skulls variable, (30) occipi- tal condyles sessile, (31) columellae present except in some neotropical genera, usually moderately long and cylindrical, (32) dentition poorly to well developed, (33) meckelian grooves open, dentaries C-shaped anteriorly, (34) detoglossal or adeto- glossal; genioglossal muscles present or absent, (35) cornua broad and long to short and attenuate, attaching to or continuous with first basibranchials, (36) lingual cartilages present or absent, (37) ante- rior first basibranchial extensions present only in some groups which lack lingual cartilages, (38) first basibranchials anteriorly expanded or not, (39) first ceratobranchials or epibranchials longest of articulated hyobranchial elements, (40) first ceratobranchials larger than second in most genera, smaller in one generic group, (41) epibranchials more than 1.5 times first basibranchials; epibran- chials shorter than first ceratobranchials in two genera, but longer in others, (42) second basi- branchials present or absent, width more or less than first basibranchial length, (43) moderate- sized atlases bear large condylar articular facets that are relatively flat to slightly concave, (44) large atlantal odontoid processes, (45) posterior atlantal margins relatively unmodified, (46) thir- teen to twenty-four trunk vertebrae, (47) two or three caudosacral vertebrae, (48) vertebrae mostly amphicoelous, but parallel tendencies toward a secondary opisthocoely, (49) hypapophysial keels usually absent, rudiments present on one or two vertebrae in a few species, (50) hyperapophyses usually well developed and divided, but united in 50 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 some species, (51) basapophyses present or absent, when present poorly to moderately developed, (52) no pterygapophyses, (53) parapophyses and diapophyses short to long, separated or united; diapophyses absent on posterior trunk vertebrae of three genera, (54) centra shorter or longer than distance across tips of parapophyses, (55) ribs present on last trunk vertebrae in some, absent in other genera, (56) ribs from 0.2 to 1.5 times dis- tance across parapophyses, (57) tails stout to slender proximally, basal tail constriction present or absent, (58) tibial spurs present or absent, (59) hind limbs only slightly larger than fore limbs, (60) fingers and toes highly variable in shape, no fleshy palm pads; digits long and slender with pointed to expanded tips in most, or short and stout, or joined, or united by webbing, (61) five to eight carpals, five to nine tarsals, (62) distal tarsals five do not articulate with central tarsals except in two genera, (63) four fingers and five toes, except for three genera which have four toes, (64) primitive phalangeal formulae 1-2-3-2; 1-2- 3-3-2; various reductions occur with extreme reduc- tion possible 0-1-2-2, 0-1-2-2-1 in forms with five toes, (65) adults aquatic, terrestrial, subterranean, or arboreal, (66) aquatic larvae in one group, direct terrestrial development in most species, (67) testes pigmented or unpigmented, lobed testes occur in some groups, (68) quadrato-pectoralis and gularis, or only gularis muscles. Content: three tribes, twenty genera — Gyri- nophilus, Pseudotriton, Stereochilus, Eurycea, Typhlotriton, Typhlomolge, Haideotriton, Hemi- dactylium, Plethodon, Aneides, Ensatina, Hydro- mantes, Batrachoseps, Bolitoglossa, Oedipina, Pseu- doeurycea, Chiropterotriton, Parvimolge, Lineatri- ton, Thorius. Tribe 1. Hemidactyliini Definition: plethodontine salamanders differing from all others by having aquatic larvae; differing from the Plethodontini in lacking anterior first basibranchial extensions, and from the Bolitoglos- sini by having large, ossified second basibranchials, and larger first than second ceratobranchials. Characterization: (1) premaxillae fused or sep- arate, (2) internasal fontanelles primarily enclosed by premaxillary frontal processes, or fontanelles secondarily closed, (3) maxillary facial lobes almost entirely anterior to midpoints of toothed portions, (4) septomaxillae and prefrontals present and well developed, (5) vomerine tooth series sharply arched anteriorly, continuous or not with posterior vomerine teeth, (6) vomers proper extend posterolaterally beyond anterior tooth series, (7) anterior portions of parasphenoids strongly expanded, (8) parietals depressed posteriorly, with distinct lateral aspects; no posterolateral parietal spurs, (9) otic crests, when present, double, involving parietals and squamosals; crests process- like, (10) columellae present and well developed, (11) detoglossal or adetoglossal; genioglossal muscles present or absent, (12) cornua distinctly separated from first basibranchials, (13) no anterior first basibranchial extensions, (14) first basibranchials moderately to greatly expanded an- teriorly, (15) first ceratobranchials of significantly greater diameter and bulk than second cerato- branchials, (16) epibranchials longest of articu- lated hyobranchial elements; 1.5 to 2 times the first basibranchials, longer than but less than 1.7 times the first ceratobranchials, (17) second basibran- chials with widths greater than two-thirds first basibranchial lengths, (18) thirteen to twenty-one trunk vertebrae, (19) three caudosacral vertebrae, (20) transverse processes may or may not extend beyond zygapophyses, (21) diapophyses and para- pophyses present on all but last one or two verte- brae, ribs all bicipital, (22) ribs 0.6 to 1.5 times distance across parapophyses, (23) tibial spurs well developed, (24) aquatic larvae with three epibranchials, some paedogenetic species remain aquatic, adults of other species aquatic to semi- terrestrial. Range: eastern North America with some species in central North America north of Mexico. Content: eight genera — Gyrinophilus, Pseudo- triton, Stereochilus, Eurycea, Typhlotriton, Typhlo- molge, Haideotriton, Hemidactylium. Gyrinophilus Cope, 1869 Definition: adults paedogenetic or fully trans- formed, eyes well developed and functional; pre- maxillae separate, fused in paedogenetic species; parietal-otic and squamosal-otic crests; well devel- oped orbitosphenoids; anterior and posterior vom- erine teeth continuous; adetoglossal with lingual cartilages and no genioglossal muscles; eighteen to twenty trunk vertebrae; tranverse processes of trunk vertebrae extend beyond zygapophyses; no basal tail constriction; four fingers and five toes. Range: eastern North America from southern Maine and Quebec, and Ohio, to northern Alabama and Georgia. Content: danielsi (Blatchley, 1901), palleucus McCrady, 1954; porphyriticus (Green, 1827). Pseudotriton Tschudi, 1838 Definition: adults fully transformed; eyes well developed and functional; premaxillae fused; parie- 1966 EVOLUTION OF LUNGLESS SALAMANDERS 51 tal-otic and squamosal-otic crests; well developed orbitosphenoids; anterior and posterior vomerine teeth continuous; adetoglossal with lingual cartil- ages and no genioglossal muscles; seventeen or eighteen trunk vertebrae; transverse processes of trunk vertebrae extend beyond zygapophyses; no basal tail constriction; four fingers and five toes. Range: eastern North America from New York and Ohio to Florida and Louisiana. Content: montanus Baird, 1849 \ ruber (Sonnini, 1802). Stereochilus Cope, 1869 Definition: adults fully transformed; eyes well developed and functional; premaxillae fused; pari- etal-otic and squamosal-otic crests; well developed orbitosphenoids; anterior and posterior vomerine teeth continuous; detoglossal with lingual cartilages and no genioglossal muscles; eighteen to twenty trunk vertebrae; transverse processes of trunk vertebrae extend beyond zygapophyses; no basal tail constriction; four fingers and five toes. Range: Atlantic Coastal Plain from Virginia to Georgia. Content: marginatus (Hallowell, 1857). Eurycea Rafinesque, 1822 Definition: adults paedogenetic or fully trans- formed; eyes well developed and functional; pre- maxillae fused; parietal-otic and squamosal-otic crests, or no crests; well developed orbitosphenoids; anterior and posterior vomerine teeth discontin- uous; adetoglossal with lingual cartilages and no genioglossal muscles; fifteen to twenty-one trunk vertebrae; transverse processes of trunk vertebrae do not extend beyond zygapophyses; no basal tail constriction; four fingers and four or five toes. Range: eastern North America from Quebec and New Brunswick to Florida, Oklahoma, and Texas. Content: aquatica Rose and Bush, 1963; bis- lineata (Green, 1818); latitans Smith and Potter, 1946; longicauda (Green, 1818); lucifuga Rafin- esque, 1822; multiplicata (Cope, 1869); nana Bishop, 1941; neotenes Bishop and Wright, 1937; pterophila Burger, Smith and Potter, 1950; quad- ridigitata (Holbrook, 1842); troglodytes Baker, 1957; tynerensis Moore and Hughes, 1939. Typhlotriton Stejneger,1892 Definition: adults essentially fully transformed, but eyes poorly developed and nonfunctional; pre- maxillae fused or separate; no otic crests; orbito- sphenoids with very small optic fenestrae; anterior and posterior vomerine teeth continuous; deto- glossal with lingual cartilages and no genioglossal muscles; seventeen to twenty trunk vertebrae; transverse processes of trunk vertebrae do not extend beyond the zygapophyses; no basal tail constriction; four fingers and five toes. Range: Interior Highlands of Missouri, Arkan- sas, Kansas, and Oklahoma. Content: spelaeus Stejneger, 1892. Typhlomolge Stejneger, 1896 Definition: adults blind, gilled, paedogenetic; premaxillae fused; no otic crests; no orbitosphe- noids; no posterior vomerine teeth; larval attached tongues; thirteen or fourteen trunk vertebrae; transverse processes of trunk vertebrae extend beyond zygapophyses; no basal tail constriction; four fingers and five toes. Range: Balcones Escarpment region of south- central Texas. Content: rathbuni Stejneger, 1896; tridentifera (Mitchell and Reddell, 1965). Haideotriton Carr, 1939 Definition: adults blind, gilled, paedogenetic; premaxillae fused; no otic crests; no orbitosphe- noids; no posterior vomerine teeth; larval attached tongues; fourteen trunk vertebrae; transverse proc- esses of trunk vertebrae do not extend beyond zygapophyses; no basal tail constriction; four fingers and five toes. Range: southern Georgia and north-central Florida. Content: wallacei Carr, 1939. H emidactylium Tschudi, 1838 Definition: adults fully transformed, eyes well developed and functional; premaxillae separate; no otic crests; orbitosphenoids well developed; anterior and posterior vomerine teeth discontin- uous; detoglossal with genioglossal muscles and no lingual cartilages; fifteen trunk vertebrae; trans- verse processes of trunk vertebrae extend beyond zygapophyses; basal tail constriction; four fingers and four toes. Range: eastern North America from Nova Scotia and Ontario to Arkansas, Louisiana, and Florida. Content: scutatum (Schlegel, 1838). Tribe 2. Plethodontini Definition: plethodontine salamanders which lack aquatic larvae; differing in addition from the Hemidactyliini by having anterior first basibran- chial extensions, and from the Bolitoglossini by having large, ossified second basibranchials and larger first than second ceratobranchials. Characterization: (1) premaxillae fused or sep- 52 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 arate, (2) internasal fontanelles primarily enclosed by premaxillary frontal processes, secondarily closed by osseous ornamentation in one species, (3) maxillary facial lobes arise from central one- third of toothed portions, (4) septomaxillae and prefrontals well developed, (5) vomerine tooth series gently curved anteriorly, not continuous with posterior vomerine teeth, (6) vomerine tooth series marks posterolateral extent of vomers proper, (7) anterior portions of parasphenoids not expanded, (8) parietals slightly or not depressed posteriorly, no lateral aspects or parietal spurs, (9) otic crests, when present, single involving otics only or otics and squamosals; crests long and continuous, (10) columellae well developed, (11) detoglossal, short genioglossal muscles present, (12) cornua dis- tinctly separated from first basibranchials, (13) anterior first basibranchial extensions well devel- oped, (14) first basibranchials moderately ex- panded just anterior to midpoints, (15) first ceratobranchials of significantly greater diameter and bulk than second ceratobranchials, (16) first ceratobranchials longest of articulated hyobran- chial elements in most, but epibranchials equal first ceratobranchials in one species; epibranchials less than 1.25 times first basibranchials, and shorter than or same length as first ceratobran- chials, (17) second basibranchials with widths less than two-thirds first basibranchial lengths, (18) fourteen to twenty-four trunk vertebrae, (19) three caudosacral vertebrae, (20) transverse proc- esses extend well beyond zygapophyses, (21) dia- pophyses and parapophyses present on all but last one or two vertebrae, ribs all bicipital, (22) ribs 0.6 to 1.25 times distance across parapophyses, (23) tibial spurs well developed, or absent, (24) direct terrestrial development with no aquatic lar- val stage, adults terrestrial to arboreal. Range: eastern and western North America with species in the Rocky Mountains, the Interior High- lands, and Texas. Content: three genera — Plethodon, Ensatina, Aneides. Plethodon Tschudi, 1838 Definition: adults differ from Aneides in having two premaxillae, and from Ensatina in having tibial spurs and lacking basal tail constriction. Range: eastern North America to the Interior Highlands and Texas, but particularly in the Appa- lachian region; western North America, primarily west of the Cascade-Sierra Nevada divide, but with populations in the Rocky Mountains of Idaho- Montana, New Mexico, and ? Utah. Content: caddoensis Pope and Pope, 1951; cin- ereus (Green, 1818); dorsalis Cope, 1889; dunni Bishop, 1934; elongatus Van Denburgh, 1916; glutinosus (Green, 1818); jordani Blatchley, 1901; larselli Burns, 1953; longicrus Adler and Dennis, 1962; neomexicanus Stebbins and Riemer, 1950; ouachitae Dunn and Heintze, 1933; richmondi Netting and Mittleman, 1938; stormi, Highton and Brame, 1965; vandykei Van Denburgh, 1906; vehiculum (Cooper, 1860); wehrlei Fowler and Dunn, 1917; welleri Walker, 1931; yonahlossee Dunn, 1917. Ensatina Gray, 1850 Definition: adults differ from Aneides and Ple- thodon in having basally constricted tails with highly specialized first caudal vertebrae, and in lacking tibial spurs. Range: far western United States from southern British Columbia to extreme southern California, west of the Cascade-Sierra Nevada divide. Content: eschscholtzii Gray, 1850. Aneides Baird, 1849 Definition: adults differ from Plethodon and Ensatina in having single premaxillae and special- ized tarsi in which all elements articulate with the centrals. Range: Applachian region of eastern North America, mountainous regions of south-central New Mexico, and far western North America west of the Cascade-Sierra Nevada divide from Van- couver Island and Oregon to California and north- ern Baja California. Content: aeneus (Cope and Packard, 1881); ferreus Cope, 1869; flavipunctatus (Strauch, 1870); hardii (Taylor, 1941); lugubris (Hallowell, 1849). Tribe 3. Bolitoglossini Definition: Plethodontine salamanders which differ from the Hemidactyliini and Plethodontini by lacking second basibranchials and having larger second than first ceratobranchials; differing in addition from the Hemidactyliini by lacking aquatic larvae, and from the Plethodontini by having epi- branchials that are much longer than the first cera- tobranchials. Characterization: (1) premaxillae fused or sep- arate, (2) anterior margins only of internasal fon- tanelles formed by premaxillary frontal processes, (3) maxillary facial lobes almost completely ante- rior to midpoints of toothed portions, (4) septo- maxillae and prefrontals present or absent, trends for reduction and loss of both, (5) vomerine tooth series gently curved anteriorly, not continuous with posterior vomerine teeth, (6) anterior vom- 1966 EVOLUTION OF LUNGLESS SALAMANDERS 53 erine tooth series marks posterolateral extent of vomers proper, (7) anterior portions of parasphe- noids slightly or not expanded, (8) parietals slightly or not depressed posteriorly; no lateral aspects, but usually with well developed posterolateral parietal spurs, (9) no otic crests, (10) columellae present in some, but definite trend for reduction and loss, (11) adetoglossal, no genioglossal muscles; modi- fied detoglossal with extremely specialized, elon- gate genioglossal muscles in one genus, (12) cornua present or absent, when present apparently continuous with first basibranchials, (13) no anterior first basibranchial extensions, (14) first basibranchials slightly or not at all expanded ante- riorly, (15) second ceratobranchials shorter, but of significantly greater diameter and bulk than first ceratobranchials, (16) epibranchials longest of articulated hyobranchial elements; 2 to 3.4 times the first basibranchials; 1.8 to 6 times the first ceratobranchials, (17) no second basibran- chials, (18) fourteen to twenty-two trunk verte- brae, (19) two or three caudosacral vertebrae, (20) transverse processes extend beyond zygapo- physes, (21) diapopbyses and parapophyses pres- ent on most trunk vertebrae of most genera, but diapophyses lost on most trunk vertebrae of sev- eral genera; ribs bicipital or unicipital, (22) ribs 0.2 to 1 times distance across parapophyses, (23) tibial spurs relatively small or absent, (24) direct terrestrial development with no aquatic larval stage, adults terrestrial to arboreal. Range: far western North America west of the Cascade-Sierra Nevada divide from ?Alaska and Oregon to northern Baja California, Mexico; the total neotropical (Mexico to Brazil and Bolivia) and European (Italy, France, Sardinia) range of the family. Content: three supergenera, nine genera — Hydromantes, Batrachoseps, Bolitoglossa, Oedi- pina, Pseudoeurycea, Chiropterotriton, Parvimolge, Lineatriton, Thorius. Supergenus Hydromantes Definition: plethodontine salamanders with sep- arated premaxillae; no prefrontals; no posterolat- eral parietal spurs; well developed columellae; adetoglossal, with no genioglossal muscles; no cornua; no omohyoideus muscles; fourteen trunk vertebrae; three caudosacral vertebrae; no basal tail constriction; five toes; and unilobed testes. Content: one genus — Hydromantes. Hydromantes Gistel, 1848 Definition: see above. Range: three species in California at from low to high elevations in the Cascade and Sierra Nevada regions; two species in Europe from north-central and western Italy to extreme southwestern France, and on the island of Sardinia. Content: brunus Gorman, 1954; genei (Schlegel, 1838); italicus Dunn, 1923; platycephalus (Camp, 1916); shastae Gorman and Camp, 1953. Supergenus Batrachoseps Definition: plethodontine salamanders with fused or separated premaxillae; extremely reduced or no prefrontals; posterolateral parietal spurs; some- what reduced columellae; highly modified deto- glossal tongues with extremely elongated and slender genioglossal muscles; relatively long slen- der cornua; omohyoideus muscles; sixteen to twenty-one trunk vertebrae; two or three caudo- sacral vertebrae; slight basal tail constriction; four toes; and multilobed testes. Content: one genus — Batrachoseps. Batrachoseps Bonaparte, 1839 Definition: see above. Range: western Oregon and California, and extreme northwestern Baja California; ? south- eastern Alaska and ? Jalisco, Mexico. Content: attenuatus (Eschscholtz, 1833); pacif- icus (Cope, 1865) ; wrighti (Bishop, 1937). Supergenus Bolitoglossa Definition: plethodontine salamanders with fused premaxillae; well developed, reduced, or no pre- frontals; posterolateral parietal spurs; columellae reduced or absent; adetoglossal, with no genio- glossal muscles; slender, sinuous, relatively short cornua; no omohyoideus muscles; fourteen or eighteen to twenty-two trunk vertebrae; two caudo- sacral vertebrae; slight to marked basal tail con- striction; five toes; and multilobed testes. Content: seven neotropical genera — Bolitoglossa, Oedipina, Pesudoeurycea, Chiropterotriton, Parvi- molge, Lineatriton, Thorius. Bolitoglossa Dumeril, Bibron, and Dumeril, 1854 Definition: salamanders with prefrontals present or absent; no posteriorly directed, rod-like squa- mosal processes; columellae extremely reduced or absent; no sublingual folds; fourteen trunk verte- brae; most ribs bicipital; distal tarsals 4 and 5 fused. Range: northeastern Mexico to eastern Brazil and central Bolivia. Content: adspersa (Peters, 1863); altamazonica (Cope, 1874); alvaradoi Taylor, 1954; arbores- candens Taylor, 1954; biseriata Tanner, 1962; bor- 54 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 burata Trapido, 1942; brevipes Bumzahem and Smith, 1955; capitana Brame and Wake, 1963; chica Brame and Wake, 1963; cerroensis Taylor, 1952; colonnea (Dunn, 1924); cuchamatana (Stuart, 1943); dofleini (Werner, 1903); dunni (Schmidt, 1933); engelhardti (Schmidt, 1936); epimela Wake and Brame, 1963; flavimembris (Schmidt, 1936); flaviventris (Schmidt, 1936); franklini (Schmidt, 1936); helmrichi (Schmidt, 1936); hypacra (Brame and Wake, 1962); ligni- color (Peters, 1873); lincolni (Stuart, 1943); macrinii (Lafrentz, 1930); marmorea (Tanner and Brame, 1961 ) ; mexicana Dumeril, Bibron, and Dumeril, 1854; morio (Cope, 1869); nicefori Brame and Wake, 1963; nigrescens (Taylor, 1949); nigrofiavescens Taylor, 1941; occidentals Taylor, 1941; omniumsanctorum (Stuart, 1952); orestes Brame and Wake, 1962; palmata (Werner, 1897); pandi Brame and Wake, 1963; peruviana (Boulen- ger, 1883); phalarosoma Wake and Brame, 1962; platydactyla (Gray, 1831); riletti Holman, 1964; robusta (Cope, 1894); rostrata (Brocchi, 1883); rufescens (Cope, 1869); salvinii (Gray, 1869); savagei Brame and Wake, 1963; schizodactyla Wake and Brame, 1966; schmidti (Dunn, 1924); sima (Vaillant, 1911); sooyorum Vial, 1963; striatula (Noble, 1918); subpalmata (Boulenger, 1896); vallecula Brame and Wake, 1963; veracru- cis Taylor, 1951 ; yucatana (Peters, 1882). Oedipina Keferstein, 1868 Definition: salamanders with no prefrontals; no posteriorly directed, rod-like squamosal processes; no columellae; sublingual folds; eighteen to twenty-two trunk vertebrae; most ribs unicipital; distal tarsals 4 and 5 fused. Range: southern Mexico through Central Amer- ica and western Colombia to north-central Ecuador. Content: alfaroi Dunn, 1921; bonitaensis Tay- lor, 1952; collaris (Stejneger, 1907); complex (Dunn, 1924); cyclocauda Taylor, 1952; elongata (Schmidt, 1936); gracilis Taylor, 1952; ignea Stuart, 1952; inusitata Taylor, 1952; longissima Taylor, 1952; pacificensis Taylor, 1952; parvipes (Peters, 1879); poelzi Brame, 1963; syndactyla Taylor, 1948; taylori Stuart, 1952; uniformis Kef- erstein, 1868. Pseudoeurycea Taylor, 1944 Definition: salamanders with prefrontals; no posteriorly directed, rod-like squamosal processes; columellae greatly reduced; sublingual folds; four- teen trunk vertebrae; most ribs bicipital; distal tar- sals 4 and 5 separated, 4 larger than 5, and 5 does not articulate with the centrale. Range: Nuevo Leon and Tamaulipas in north- eastern and Nayarit in western Mexico to Guate- mala. Content: altamontana (Taylor, 1938); bellii (Gray, 1849); brunnata Bumzahem and Smith, 1955; cephalica (Cope, 1865); expectata Stuart, 1954; frscheini Shannon and Werler, 1955; gado- vii (Dunn, 1926) ; galeanae (Taylor, 1941 );gigan- tea (Taylor, 1938); goebeli (Schmidt, 1936); leprosa (Cope, 1869); melanomolge (Taylor, 1941); nigromaculata (Taylor, 1941); rex (Dunn, 1921); robertsi (Taylor, 1938); scandens Walker, 1955; smithi (Taylor, 1938); sulcata (Brocchi, 1883); unguidentis (Taylor, 1941); werleri Darling and Smith, 1954. Chiropterotriton Taylor, 1944 Definition: salamanders with prefrontals pres- ent or absent; no posteriorly directed, rod-like squa- mosal processes; columellae present, reduced, or absent; sublingual folds; fourteen trunk vertebrae; most ribs bicipital; distal tarsals 4 and 5 separated or fused, when separated 5 usually larger than 4 and articulates with centrale, or, if 4 larger than 5, prefrontals are absent. Range: southern Nuevo Leon and southwestern Tamaulipas through eastern and east-central Mex- ico to Guatemala, El Salvador, Honduras, and Costa Rica. Content: abscondens Taylor, 1948; arbor eus (Taylor, 1941); barbouri (Schmidt, 1936); brom- eliacia (Schmidt, 1936); chiropterus (Cope, 1863); chondrostega (Taylor, 1941); dimidiatus (Taylor, 1939); lavae (Taylor, 1942); magnipes Rabb, 1965; megarhinus Rabb, 1960; mosauri (Woodall, 1941); multidentatus (Taylor, 1938); nasalis (Dunn, 1924); picadoi (Stejneger, 1911); priscus Rabb, 1956; xolocalcae (Taylor, 1941). Lineatriton Tanner, 1950 Definition: salamanders with prefrontals; no posteriorly directed, rod-like squamosal processes; columellae reduced; sublingual folds; fourteen trunk vertebrae; most ribs unicipital; no distal tarsal 5. Range: central Veracruz, Mexico. Content: lineola (Cope, 1865). Thorius Cope, 1869 Definition: salamanders with prefrontals ex- tremely reduced or absent; posteriorly directed, rod-like squamosal processes; no columellae; sub- lingual folds; fourteen trunk vertebrae; most ribs bicipital; distal tarsals 4 and 5 fused. Range: southern Veracruz and Puebla to Guer- 1966 EVOLUTION OF LUNG LESS SALAMANDERS 55 rero and Oaxaca, Mexico, north of the Isthmus of Tehuantepec. Content: dubitus Taylor, 1941; macdougalli Taylor, 1949; maxillabrochus Gehlbach, 1959; minutissimus Taylor, 1949; narisovalis Taylor, 1939; pennatulus Cope, 1869; pulmonaris Taylor, 1939; schmidti Gehlbach, 1959; troglodytes Tay- lor, 1941. Parvimolge Taylor, 1941 This genus, recognized by recent authors, is a poorly defined and apparently unnatural assem- blage of three distinct species, and diagnosis is not possible. The species may prove to be members of the genera Chiropterotriton and Pseudoeurycea, but at least one, P. townsendi, appears sufficiently distinct to warrant a separate genus. The two species which have been examined (P. richardi, P. townsendi ) have small prefrontals; no posteriorly directed, rod-like squamosal processes; extremely reduced or no columellae; sublingual folds present; fourteen trunk vertebrae; and bicipital ribs. Distal tarsals 4 and 5 are fused in P. townsendi. Range: central and southern Veracruz, Mexico, and north-central Costa Rica. Content: praecellens Rabb, 1955; richardi Tay- lor, 1949; townsendi (Dunn, 1922). EVOLUTIONARY RELATIONSHIPS AND TRENDS Relationships Major Familial Divisions The family Plethodontidae, as proposed by Gray (1850), included species now included in the fam- ilies Hynobiidae and Ambystomatidae as well as the following four groups: Plethodon; Desmognathus and Hemidactylium; Batrachoseps, Spelerpes (which included species now placed in Eurycea, Manculus, Pseudotriton, Gyrinophilus, and Pseu- doeurycea), Geotriton (= Hydromantes), and Oedi- pus (— Bolitoglossa ); Ensatina. Hallowell (1856) used family endings for three groups included in Gray’s single family: (1) Plethodontidae — Ple- thodon, Aneides, Desmognathus, (2) Bolitoglos- sidae — Pseudotriton, Spelerpes, Batrachoseps, Geo- triton, (3) Hemidactylidae — Hemidactylium. The groups were not precisely defined and never be- came established in the literature. Cope (1859) recognized two groups on the basis of tongue attachment; those with free tongues (including Batrachoseps ) were called Spelerpeae, those with attached tongues, Plethodontae. In 1866 Cope revised his views and recognized two families of present day plethodontids. A single genus, Desmognathus, was placed in the Desmog- nathidae, and the other genera in the Plethodonti- dae. Cope (1869) erected the family Thoriidae for the genus Thorius. Cope (1859, 1866, 1869) was the first to present osteological information of any significance and to characterize and discuss rela- tionships of salamander families. Desmognathids were said to differ from plethodontids in having stalked rather than sessile occipital condyles, opis- thocoelous rather than amphicoelous vertebrae, and a frontal process extending into the superpala- tal vacuity. Thoriids were said to have opisthocoe- lous vertebrae and ossified carpals and tarsals. Cope thought thoriids combined the characters of the Desmognathidae, Plethodontidae, and Amby- stomatidae. Boulenger (1882) reduced the three families of Cope to two subfamilies (thoriids included in Des- mognathinae) of the Salamandridae, but Cope (1889) rejected this work. Cope later modified his views (1894 a) and placed Thorius in the Desmog- nathidae. As a result of his discovery of Hapto- glossa, an elongate Costa Rican genus with an attached tongue and opisthocoelous vertebrae, Cope (1894 b) recognized a subfamily Thoriinae within the family Desmognathidae. The subfamily included Haptoglossa, Thorius, and Typhlotriton (described as a desmognathid by Stejneger, 1892). The unique type of Haptoglossa is lost and the status of the genus remains enigmatic (see also Taylor, 1944). The family Typhlomolgidae was erected for the single genus Typhlomolge by Stejneger and Bar- bour (1917), despite Emerson’s (1905) study which showed it to be closely related to Eurycea. The family received no support and Fowler and Dunn (1917) simply stated “Typhlomolge is a permanent larva of some plethodont.” On the basis of his studies of vertebral articula- tions, Moore (1900) suggested that desmognathids be included in the family Plethodontidae. He was followed by most subsequent workers (Stejneger and Barbour, 1917; Fowler and Dunn, 1917; Dunn, 1917 and 1926; Noble, 1931; Herre, 1935; Piatt, 1935; Schmidt, 1953;Conant, 1958). Dunn’s (1926) classical work, which has long been the standard authoritative reference on plethodontids, recognized no formal groupings below the family level. Recently there has been an attempt to revive the family Desmognathidae. Smith and Taylor (1948) recognized two families of lungless sala- manders, and two subfamilies (Thoriinae and Ple- thodontinae) of plethodontids. The Thoriinae included only Thorius. Soler (1950) presented evi- 56 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 dence of the distinctiveness of Desmognathus and Leurognathus, and argued for recognition of a family Desmognathidae. In a rather uncritical review based on the literature, von Wahlert (1957) recognized a superorder Plethodontoidea (sug- gested by Laurent, 1947) with two families, Des- mognathidae and Plethodontidae. Two subfamilies of plethodontids were recognized: Plethodontinae (Aneides, Bcitrachoseps, Ensatina, Hemidactylium, Plethodon, Stereochilus) and Thoriinae (remaining nondesmognathine plethodontids). Recent authors (Teege, 1956; Conant, 1958; Highton, 1961; Mar- tof, 1962; Wake, 1963; Estes, 1964 and 1965; Monath, 1965) have not recognized a family Desmognathidae. Plethodontid salamanders fall into two major assemblages that I consider to be of subfamilial rank. The subfamily Plethodontinae contains twenty of the twenty-three genera, and is the more generalized and progressive group. The more spe- cialized subfamily Desmognathinae is comprised of three genera. It is difficult to discuss the sub- families in terms of one being more primitive or advanced than the other. Primitive members of both groups have aquatic larvae, but desmogna- thine larvae have four epibranchials and pletho- dontines have three. In some respects ( e.g ., palatal elements, gill rami) plethodontine larvae are the more generalized, but several plethodontine lar- vae have secondary specializations (elongate snout and reduced eyes of Gyrinophilus, pond specializa- tion of Hemidactylium, Manculus and Stereochi- lus). Some plethodontines retain the primitive pre- maxillary metamorphic pattern, but others do not nor do the modern desmognathines. Primitive characters found in primitive plethodontines but not in desmognathines include presence of naso- lacrimal ducts, large septomaxillae, well developed prefrontals, toothed vomerine preorbital processes, continuous anterior and posterior vomerine tooth series, amphicoelous vertebrae, cartilaginous antor- bitals, convex frontals, and open ineckelian grooves. Other primitive characters (tibial spurs, mesopodial numbers and configuration) are shared by desmognathines and primitive plethodontines. Desmognathines have many unique morphological specializations, but subfamily-wide plethodontine specializations are not apparent. The desmogna- thines probably diverged from a pre-plethodontine stock at an early date (Mesozoic). Desmognathines have become highly specialized, but have retained a number of primitive characters. The Plethodon- tinae is a considerably more generalized and diverse group than the Desmognathinae, yet some pletho- dontine genera can be considered more advanced than any desmognathine genus (see below). The plethodontine genera are probably morphologi- cally more similar to the ancestral common stock than are the desmognathines, and the Plethodon- tinae must be considered the most primitive sub- family (see Fig. 12) . Figure 12. Diagrammatic representation of the relation- ships of the major groups of plethodontid salamanders. Functional specializations of four types in the desmognathines have led to drastic skeletal mod- ifications. These modifications sharply distinguish desmognathine from plethodontine genera, and include: (1) streamlining, a response to selective forces operative in the primitive, swift mountain brook environments, (2) skull strengthening, a response to selective forces correlated with use of heads as wedges, (3) new mouth opening mechan- isms, the result of selective forces in the primitive environment and related to peculiarities in use of the heads, (4) vertebral and limb modifications, the results of locomotor specializations. Streamlining is a common occurrence in verte- brates which inhabit aquatic environments, and especially those that occur in swift streams. Cer- tainly the remarkable desmognathine skull flatten- ing is the most important of the streamlining mod- ifications. Flattening imparts an elongate wedge- shape to the skulls, a factor of particular advantage to these organisms which not only inhabit rapidly flowing brooks, but also hide under rocks and in 1966 EVOLUTION OF LUNG LESS SALAMANDERS 57 rock crevices by forcing their way head first (Dunn, 1926; Martof, 1962). The anterior cranial elements of desmognathines are relatively dense, thickened, closely articulated bones, and the dorsal surfaces of the skulls are solid and relatively smooth. The frontals and anterior margins of the parietals are also flat and smooth. Head muscles that primitively originate on the dorsal surfaces of the frontals and parietals have shifted their origins lateroventrally, and arise from the lateral margins of the bones, below the ridge-like dorsolateral margins, and from the orbitosphenoids. Reduction or absence of vomerine vaulting is also directly associated with the streamlining and skull depression. Other features associated with streamlining include well developed eyes which are neither prominent nor strongly protuberant, fore limbs considerably smaller than hind limbs, and very smooth transitions from relatively small heads to cylindrical trunks and stout tails. Skulls of plethodontines remain relatively rigid in reference to the vertebral column, and the man- dibles are lowered to open the mouths. In contrast the mandibles of desmognathines are held rela- tively rigid and the skulls proper are raised some- what during mouth opening. The mandibles are far from immobile in Desmognathines (see Noble, 1931), but are much more so than in plethodon- tines. It has been suggested (Dunn, 1926; Conant, 1958) that the relative immobility of the lower jaws stiffens the anterior portions of the bodies so the salamanders can more readily force their way under objects. Functional modifications of the jaw musculature immobilize the mandibles. Slips of the adductor mandibulae musculature (adductor mandibulae externus) extend from the prearticu- lars to the atlas in all plethodontids, but desmog- nathines differ from plethodontines in that the slips contain stout, relatively inflexible ligaments that limit the ventral movement of the mandibles rela- tive to the trunk. Desmognathines have stalked occipital condyles that effectively move the skull away from the atlas and provide room for swinging the skull up and down to open and close the mouth. Further mod- ifications facilitating skull movement on the verte- bral column are the drastic reduction in size of the odontoid process and movement out of the foramen magnum. In addition the occipital con- dylar articular facets are strongly convex and are received by enlarged, strongly concave atlantal cotyles. Muscular modifications have evolved to enable skull elevation. The depressor mandibulae muscles lower the mandibles of plethodontines, but are of relatively minor importance in desmognathines; apparently the anterior dorsal spinal muscles have assumed the primary desmognathine mouth open- ing function. The dorsal spinal muscles are excep- tionally well developed in desmognathines and extend onto the skull to insert on raised otic crests, behind the parietal-otic troughs, and on the parie- tals in front of the troughs. The muscles originate on the dorsal surfaces of the anterior trunk verte- brae. Unique, moderately to well developed verte- bral pterygapophyses provide additional area for dorsal spinal muscle origin. Despite the fact that the adductor mandibulae musculature contains ligaments, it plays a role in lowering the desmognathine skull onto the mandi- bles. Additional musculature also is brought into play. The quadrato-pectoralis musculature is remarkably well developed in desmognathines, extending as stout bands from the quadrate region to the gular fold (see Piatt, 1935); contraction of the quadrato-pectoralis muscles pulls the skull down while contraction of the adductor mandibu- lae muscles raises the mandibles. The two sets of muscles effectively close the mouth. When the skull swings dorsally on the atlas, the mandibles are also raised. Mandibular elevation ceases, however, as the depressor mandibulae muscles contract and facilitate mouth opening by restricting dorsal mandibular motion. The skull swings dorsally and movement of the quadrates on the articulars occurs. Since the atlas is a relatively solid fulcrum and vertical motion of the mandibles is restricted, skull movement propels the mandibles forward. Movements of the quadrates on the articu- lars perhaps has served as a selective force that has led to strengthened quadrates. Desmognathine quadrates are enlarged and firmly articulated with the parasphenoids and occipito-otics, and anchored by strong jugal ligaments to the maxillae. The great tensions in the ligament are evidenced by the for- mation of moderate to very large maxillary jugal processes in the desmognathines. The squamosals are stout and articulate firmly with the quadrates and occipito-otics. The results of the above modifi- cations are that the desmognathine quadrates are solidly anchored to the brain case, and the pos- terior portions of the skull are as strong and rigid as the anterior portions. Primitive desmognathines are aquatic organisms living in swift streams. A number of morphological specializations are related to stream locomotion. Among these are the heavily muscularized trunks and stout, compressed tails. Development of opis- thocoely facilitates increased intervertebral move- ment, the result of increased trunk and tail flexure. 58 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 Desmognathine trunk vertebrae are also charac- terized by the presence of hypapophysial keels and basapophyses. Auffenberg (1961) and Estes (1964) suggest that such processes relate to the develop- ment of muscles aiding lateral undulation and to swimming efficiency. A final feature associated with aquatic habits and the particular behavior of desmognathines is the hypertrophy of the hind limbs, which are much larger than the fore limbs. The large hind limbs apparently play important roles in the locomotion of desmognathines, particularly in aiding the sala- manders to force their way under and between objects. There is no question as to the distinctiveness of the Desmognathinae, but if the differences which separate desmognathines and plethodontines are carefully examined, it is apparent that most result from adaptation to a particular habitat. Thus streamlining, skull strengthening, mouth opening modifications, vertebral modifications, and limb modifications are all directly related to way of life of the salamanders, and are specialized adaptations to the environment in which they occur. Desmognathines and plethodontines are closely related, and it is likely that they have had some common ancestor. The differences between the groups are sharp, but when analyzed carefully, are shown to be the results of relatively minor speciali- zations in the Desmognathines. So many character- istics are held in common by the plethodontines and desmognathines that recognition of a family Desmognathidae as proposed by Cope (1869) and recently advocated by Smith and Taylor (1948) and Soler (1950) appears unwarranted. Relationships of the Desmognathine Genera Desmognathus and Leurognathus are the best known of the three desmognathine genera, and similarities and differences of the two have been discussed to some extent by Moore (1899), Dunn (1926), and Martof (1962). The genera are basic- ally similar, and differences are primarily the result of the peculiar selective pressures within the strictly aquatic environment to which Leurognathus has adapted. Leurognathus has a more flattened skull and a sharper, more pointed snout than Desmogna- thus. Internal nares are reduced to small, laterally placed slits, and the internasal fontanelles of large Leurognathus are completely closed. The fontan- elles primitively contain glands that furnish lubri- cant utilized while swallowing by salamanders in terrestrial or semiterrestrial situations; Leurogna- thus has become so completely aquatic that such lubrication is unnecessary, and selection has favored dorsal and ventral fontanelle closure as parts of trends toward skull strengthening. Leurog- nathus is unique in having very long vomerine pre- orbital processes that extend beyond the vomerine bodies to articulate firmly with the maxillary palatal shelves. Leurognathus differs further in having better developed hypapophysial keels on more vertebrae, and fused hyperapophyses on all trunk vertebrae. One specialization found in Desmognathus but not in Leurognathus is the presence of large, calci- fied plaques within the mouths under the eyes (the “crushing plates” of Noble, 1931). The large, irregularly-shaped plaques are found in at least three species (D. quadramaculatus, D. monticola, D. ochrophaeus), and are apparently unique fea- tures of the genus. Since Moore’s (1899) description of Leurogna- thus the status of the genus has not been seriously questioned. Osteological data indicate that Leurog- nathus probably arose as a specialized branch of a Desmognathus ancestral stock that may have been very similar to modern D. quadramaculatus (see also Dunn, 1926). It is more difficult to separate D. quadramaculatus from L. marmoratus than from other Desmognathus. One might question the advisability of generic recognition of a single species so obviously closely related to a larger, rather diverse genus. I agree with Martof (1962) that recognition is justifiable because Leurognathus has entered upon a new way of life, one not fol- lowed by any species of Desmognathus, and is morphologically distinct from all Desmognathus. Leurognathus is the most aquatic of any of the metamorphosed plethodontids and is restricted to aquatic situations (Martof, 1962). The distinctive morphology, ecology, and life history of the species, L. marmoratus, warrant separation from Desmog- nathus and placement in a distinct genus. Highton (1961) announced the discovery of a remarkable new genus and species, Phaeognathus hubrichti, a very large, elongate, terrestrial bur- rower. Valentine (1963 b, 1963 c) and Brandon (1965) have recently provided some information that supports Highton’s suggestion that Phaeogna- thus is a desmognathine, and I have corroborated many of their observations; additional support has been presented above. Phaeognathus differs sharply and strikingly from Desmognathus and Leurogna- thus in many characters. Desmognathus and Leu- rognathus are rather stout, short-bodied forms with fifteen trunk vertebrae, but the slender, elongate Phaeognathus has twenty-one to twenty-three; limbs and digits of Phaeognathus are greatly shortened in comparison with the other genera. Phaeognathus 1966 EVOLUTION OF LUNG LESS SALAMANDERS 59 is a terrestrial burrower, a habit not encountered elsewhere in the Desmognathinae, and several characters are apparently related to burrowing. The skull is rather bullet-shaped and only slightly flattened. The head is small relative to body size. Skin covering the snout is co-ossified to the anterior cranial elements, which are sculptured and eroded. Particularly striking are the large, spatulate, max- illary jugal processes which extend almost to the quadrates. Size of the processes testifies to the extreme tension in the jugal ligaments, and to the over-all strength of the skulls. The mandibular adductor series of muscles of desmognathines consist of two major bundles, the adductor mandibulae anterior and the adductor mandibulae externus (terminology of Edgeworth, 1935) . The former extend from the medial margins of the prearticulars to the lateral walls of the brain- case, particularly the dorsal portions of the orbito- sphenoids and the ventrally directed lateral lips of the frontals. The adductor mandibulae externus are the muscles called temporalis by various authors ( e.g ., Soler, 1950) and extend from the atlases to the prearticulars. Phaeognathus has larger adductor mandibulae anterior and smaller adductor man- dibulae externus muscles than Desmognathus and Leurognathus. Unique, lateral, frontoparietal ridges of Phaeognathus serve as points of attach- ment for the enlarged adductor mandibulae ante- rior musculature. In addition the dorsal spinal musculature is better developed in Phaeognathus than in other desmognathines, and the points of insertion of most of the musculature are the sites of the raised terminal projections of the prominent fronto-parietal ridges. As a result of the relatively large amounts of musculature which insert on the parietals anterior to the parietal-otic troughs, the parietals are less flattened and the two parietal re- gions are less distinguishable in Phaeognathus than in Desmognathus or Leurognathus. The quadrato-pectoralis and especially the gularis muscles are considerably less developed in Phaeog- nathus than in the other desmognathine genera. The gularis muscles are barely perceptible and are of little functional importance, and the quadrato- pectoralis muscles are relatively slender. It is obvious that skull raising specializations are less well developed in Phaeognathus than in Desmog- nathus or Leurognathus. The strong development of the dorsal spinal musculature could as well be a burrowing as a mouth opening adaptation. Phaeognathus differs further from the other desmognathines in characters of the vertebrae and vomers (see above). The characters have no obvi- ous connection with ancestral conditions and may be related to new types of locomotion and feeding. There is no question concerning the distinctive- ness of Phaeognathus. Apparently it is a relatively advanced, specialized genus derived from a Desmognathus-Leurognathus ancestral stock at a relatively early date. Of the three desmognathines Desmognathus is the most generalized and primitive. Leurognathus is closely related to Desmognathus and has entered a new adaptive subzone — the aquatic environment. Phaeognathus is more closely related to Desmog- thus than to Leurognathus, and was probably de- rived from the common ancestral stock at an earlier date than was Leurognathus (see Fig. 13). Phaeog- nathus has entered a new adaptive zone — the terrestrial environment — in an opposite evolution- ary direction from that of Leurognathus. Desmog- nathus has primitive species which occupy the primitive semiaquatic niche (see below), but the genus is undergoing an adaptive radiation and in- cludes species which have successfully entered terrestrial niches. Desmognathus is the most suc- cessful and evolutionarily the most plastic of the three genera. 60 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 Relationships of the Plethodontine Genera Tribal relationships By far the largest number of species and genera of lungless salamanders are included within the subfamily Plethodontinae. These species may be placed conveniently in three natural subgroups or tribes on the basis of structural similarities. The generic assemblages are sufficiently distinct to in- dicate continuity of descent within each tribe, but not to warrant subfamilial recognition. Certain genera are somewhat intermediate between two tribes, and no tribe is as distinct a group as the subfamily Desmognathinae. The groupings proposed here differ from those of other authors, and some historical information is necessary as an introduction to the following discussion. Dunn’s (1926) comprehensive work recognized two major groups of non-desmogna- thine genera, the attached tongued Plethodon group (Plethodon, Ensatina, Aneides, Hemidactylium, Batrachoseps) and the free tongued Eurycea group (Eurycea, Manculus, Gyrinophilus, Pseudotriton, Hydromantes, Oedipus ) with Typhlotriton, Typhlo- molge, and Stereochilus intermediate. Most subse- quent authors have followed Dunn. Noble (1925, 1927 a, 1927 b, 1931) suggested that there might be a relationship between Hydromantes and Pletho- don. Herre (1935) saw an Oedipus-Hydromantes group arising from a Plethodon stock and derived the desmognathines from a Pseudotriton-like ances- tor, but otherwise followed Dunn. Piatt (1935) recognized Stereochilus ( Stereochilus , Typhlotriton, Typhlomolge) and Gyrinophilus ( Gyrinophilus , Pseudotriton, Eurycea, Manculus ) groups in one assemblage, and Plethodon ( Plethodon , Aneides, Ensatina, Hemidactylium) and Oedipus (Oedipus, Hydromantes, Batrachoseps) groups in another. Piatt’s arrangement was unique in that he pro- posed parallel evolution of free tongues. He stated “ Hyclromantes and Oedipus have no close affinities to the Gyrinophilus group but arose from early Plethodon stock by means of a form like Batra- choseps.” Piatt’s arguments were not particularly convincing and he has not been followed by sub- sequent workers. Tanner (1952) recognized two groups of free tongued genera which he thought were derived from a common free tongued ances- tor. His groups were: (1) Gyrinophilus, Pseudo- triton, Eurycea, Manculus and (2) Hydromantes, Bolitoglossa, Magnadigita, Pseudoeurycea, Chiro- pterotriton, Oedipina, Parvimolge, Lineatriton, Thorius. Tanner suggested that the relationships of Batrachoseps were with Plethodon and rejected Piatt’s proposal. Von Wahlert (1957) recognized two subfamilies (Plethodontinae: Aneides, Batra- choseps, Ensatina, Hemidactylium, Plethodon, Stereochilus; Thoriinae: Bolitoglossa, Chiroptero- triton, Eurycea, Gyrinophilus, Hydromantes, Lineatriton, Magnadigita, Manculus, Oedipina, Parvimolge, Pseudotriton, Pseudoeurycea, Thorius, Typhlotriton, lEladinea, IHaideotriton, ITyphlo- molge) . His treatment of Stereochilus was novel and without explanation. The only recent authors to recognize a subfamily Thoriinae have been Smith and Taylor (1948) and they clearly included only Thorius. The family group name Bolitoglos- sidae (Hallowell, 1856) has clear priority over Thoriidae (Cope, 1869) as used by von Wahlert (his Thoriinae). This was recognized by Monath (1965) who follows von Wahlert but refers to members of the latter’s subfamily Thoriinae as “bolitoglossines.” Brame (1960) has also referred to this group as the bolitoglossine genera. The tribe Hemidactyliini is restricted in distri- bution to eastern North America and contains eight genera: Gyrinophilus, Pseudotriton, Stereochilus, Eurycea, Typhlotriton, Typhlomolge, Haideotri- ton, and Hemidactylium. The tribe Plethodontini has a discontinuous distribution in eastern, central, and western North America, and contains three genera: Plethodon, Aneides, and Ensatina. Mem- bers of the tribe Plethodontini are called pletho- donines in this paper. The tribe Bolitoglossini occurs in western North America, Mexico, Central and South America, and Europe, and contains nine genera: Hydromantes, Batrachoseps, Bolitoglossa, Oedipina, Pseudoeurycea, Chiropterotriton, Parvi- molge, Lineatriton, and Thorius. The bolitoglos- sine group has the most genera, the most species, the most diversity, and the most specialization of the assemblages of the Plethodontinae. All members of the Hemidactyliini have aquatic larvae, but all members of the other two tribes have direct terrestrial development. This is a char- acter of major importance and one of the main reasons for considering the Hemidactyliini the most primitive of the three tribes. The hemidac- tyliine genera are aquatic to semiterrestrial as adults, while the other tribes are strictly terrestrial or arboreal. Hemidactyliine genera have larger and stronger second basibranchials than other plethodontids, and the character is apparently related to aquatic life. The bony second basibranchials are associated with the hyobranchial musculature, forming trans- verse septa in the rectus cervicis of adults. The second basibranchials presumably play a role in strengthening this musculature, which is of great importance in gill movements of larvae, and flota- 1966 EVOLUTION OF LUNGLESS SALAMANDERS 61 tion and submergence mechanisms of adults. Second basibranchials are present in plethodonines, but are relatively smaller than in hemidactyliines; the elements are absent in all bolitoglossines. Several additional characters associated with hyobranchial structure are among the most diag- nostic of the tribes (see above under hyobranchial skeleton). Several skull characters separate the tribes. Significant anterior parasphenoid expansion occurs only in the hemidactyliine genera. The anterior vomerine tooth rows mark the posterior margins of the vomers proper in plethodonines and bolito- glossines, but bony vomerine growth extends pos- terior to the tooth series, below the parasphenoids, in hemidactyliines. Anterior and posterior vomer- ine tooth series are continuous only in four hemi- dactyliine genera. Maxillary facial lobes arise from the central one-third of the maxillae in plethodon- ines, but the lobes are entirely anterior to the mid-points in the other tribes. The parietals of plethodonines and bolitoglossines are relatively flat, but those of hemidactyliines have distinct lateral aspects. Ribs tend to be shorter, and vertebral centra longer and slenderer in bolitoglossines than in the other two tribes. A number of specialized characters occur in at least some species in two or more tribes, and most or all are the result of parallel evolution. These include proximal tail constriction (Hemidactyliini, Plethodontini, Bolitoglossini), loss of fifth toes, (H,B), raised atlantal bosses (H,P), elongation, and increased numbers of trunk and tail vertebrae (H,P,B), otic crests (H,P), fused premaxillae (H,P,B), loss of vomerine preorbital processes (H,P,B), attainment of significant tongue freedom and ability to protrude tongues for significant dis- tances from the mouths (H,P,B), loss of genio- glossal muscles (H,B), reduction and loss of tibial spurs (P,B), and tendencies toward opisthocoely (H,P,B). The Bolitoglossini is obviously the most special- ized and advanced of the three tribes. Specialized and advanced conditions found only in some to all bolitoglossine genera include unicipital ribs and loss of the diapophyses on most trunk vertebrae, loss of prefrontals, loss of septomaxillae, loss of columellae, loss of maxillary teeth, loss of omo- hyoideus muscles, fusion of premaxillary frontal processes, extensive mesopodial fusions, second basibranchial loss, and presence of parietal spurs. Some bolitoglossines have reduced numbers of chromosomes (Kezer, in litt.). The above charac- ters are found in several to many bolitoglossine genera. Generalized and presumably primitive features of the bolitoglossines included paired pre- maxillae in primitive species, and low and invari- able numbers of trunk vertebrae in most genera. The Hemidactyliini and Plethodontini include relatively primitive genera, and are rather general- ized groups in comparison with the Bolitoglossini. Primitive features found in some or all hemidac- tyliines include aquatic larvae, primitive nasolac- rimal duct pathways, retention of genioglossal muscles, retention of all primitive plethodontid cranial elements, relatively long maxillae, con- tinuous anterior and posterior vomerine tooth series, and separated premaxillae. Certain hemidactyliines have specialized condi- tions which include loss of the fifth toes, presence of basal tail constriction, great first basibranchial expansion, genioglossal muscle loss (most genera), attainment of true adetoglossy, separation of the anterior and posterior vomerine tooth series, shortened transverse processes, presence of hypa- pophysial rudiments, presence of well developed basapophyses, attainment of paedogenesis, and premaxillary fusion. Plethodonines are primitive in having detoglossy and genioglossal muscles in all species. In addition they have all the primitive cranial elements, and two of the three genera have paired premaxillae. Specializations found in all species include separa- tion of anterior and posterior vomerine tooth series, advanced nasolacrimal duct routes, and direct terrestrial development. Certain species have other specializations including loss of tibial spurs and of vomerine preorbital processes, strengthened skull elements, fused premaxillae, specialized maxillae, specialized arrangements of mesopodial elements, and hypertrophy of quadrato- pectoralis musculature. All available evidence indicates that the hemi- dactyliine line is the most primitive and generalized of the three. A major consideration is the retention of an aquatic larval stage in all members of the tribe. Plethodonines have a more primitive tongue structure with retention of anterior attachments, strong genioglossal muscles, and relatively short epibranchials; however, one hemidactyliine genus, Hemidactylium, also retains both anterior attach- ment and genioglossal muscles. Features in which plethodonines are apparently more primitive than hemidactyliines are primarily those related to tongue structure, and in most other characters hemidactyliines are definitely more primitive. The plethodonine line is closer to the hemidac- tyliine line than is the bolitoglossine line (see (Fig. 12). The primary feature relating hemidac- 62 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 tyliines and bolitoglossines is the attainment of true adetoglossy, but evidence has been presented which indicates that parallelism likely has occurred. The above evidence indicates that hemidactyl- iines are probably closer to the ancestral pletho- dontid stock than any desmognathine, plethodo- nine, or bolitoglossine. Desmognathines were probably the earliest derivatives of the stock, fol- lowed by bolitoglossines and finally by plethodo- nines. Within the Plethodontinae the tribe Hemi- dactyliini is apparently the central, primitive group, the Plethodontini is somewhat more advanced, and the Bolitoglossini is the most spe- cialized and advanced of the tribes (see Fig. 12). Relationships of the hemidactyliines Generic groupings Three distinct generic assemblages of hemidac- tyliine salamanders are apparent: (1) Gyrino- philus, Pseudotriton, Stereochilus, (2) Eurycea, Typhlotriton, Haideotriton, Typhlomolge, (3) Hemidactylium (see Fig. 14). Gyrinophilus, Pseudotriton, and Stereochilus are relatively primitive, rather closely related groups, but all represent unquestionably distinct evolutionary lines worthy of generic recognition. The genera share a number of characters, includ- ing: (1) similar anterior cranial elements, in a broad sense, (2) broad palatal shelves, (3) parie- tal-otic and squamosal-otic crests (shared with some Eurycea ), (4) continuous anterior and pos- terior vomerine tooth series (shared with Typhlo- triton and some Eurycea ), (5) broadly similar hyobranchial structure, (6) lingual cartilages (shared with Eurycea), (7) no genioglossal muscles (shared with Eurycea and Typhlotriton) , (8) moderate first basibranchial expansion (shared with Hemidactylium), (9) moderate-sized cornua, (10) high modal numbers of vertebrae (shared with Typhlotriton) , (11) proximally fused diapo- physes and parapophyses, (12) long transverse processes (shared with Hemidactylium), (13) diapophysial origins just slightly in advance of parapophysial origins (shared with Hemidactyl- ium), (14) short ribs (shared with Hemidactyl- ium), (15) stout, relatively short vertebral centra, (16) long frontals. Pseudotriton and Gyrinophilus resemble each other more closely than either resembles Stereo- chilus. In addition to the above characters they share enlarged adductor mandibulae anterior muscles, which insert on top of the frontals and virtually cover the interorbital regions. This mus- Figure 14. Diagrammatic representation of the relationships of the hemidactyliine genera. 1966 EVOLUTION OF LUNGLESS SALAMANDERS 63 culature covers only the posterolateral frontal extremities in other hemidactyliines. Second basi- branchial width is greater than first basibranchial length in the two genera, but lesser in Stereochilus. Both Gyrinophilus and Pseudotriton have large, adetoglossal tongues, but Stereochilus has a rela- tively small tongue with a membranous anterior attachment. Grobman (1959) placed Gyrinophilus in the synonymy of Pseudotriton on morphological grounds, but Martof and Rose (1962) disagreed and reported that the two genera differ in skull bone density ( Gyrinophilus light, Pseudotriton heavy) snout length (G. long, P. short), frontal and toothed portions of premaxillae fused ( P .) or not (G.), and nasals elongate and contacting maxillae (G.), or broad and separated from maxil- lae (P.). My observations corroborate their state- ments in large part, but the nasals are in contact with the maxillae in some Pseudotriton. Other sig- nificant differences include presence of ventro- medial quadrate projections, nasolacrimal duct impressions on the anterodorsal prefrontal sur- faces, and broad and extensive overlap of the pre- maxillary frontal processes by the nasals in Pseu- dotriton but not in Gyrinophilus. Stereochilus resembles Pseudotriton in having fused premaxillary toothed and frontal processes, and hypapophysial remnants on some trunk verte- brae. Moderate maxilla-nasal articulation and no nasolacrimal duct impressions relate Stereochilus to Gyrinophilus. Numerous characters separate Stereochilus from both genera. The snouts and heads are greatly compressed laterally, and this compression is reflected in the extensive fusion of the premaxillary frontal processes anterior as well as posterior to the fontanelles, ontogenetic closure of the internasal fontanelles, and medial juncture of the vomerine tooth series below the anterior portions of the parasphenoids. Despite its specializations Stereochilus is a rela- tively primitive form, not far removed in most features from an idealized familial ancestral stock. One of the most primitive characters of Stereo- chilus is the presence in adults of a distinct, well developed lateral line system on the heads. The system is better developed than in any other plethodontid genus (see Hilton, 1947 b; 1950). Stereochilus diverged from the ancestral pletho- dontid stock at a somewhat later date than the stock which gave rise to the desmognathines, but it may have been one of the earliest hemidactyliine derivatives. Gyrinophilus, Stereochilus, and Pseudotriton form the most primitive generic group in the fam- ily. The larvae of Gyrinophilus are somewhat spe- cialized (elongate snout, reduced eyes) and adults are adetoglossal with elongated anterior cranial elements, but in most other features the genus is generalized; it may well be the most primitive plethodontid, as suggested by Dunn (1926). The Eurycea-Typhlotriton-Haideotriton-Typhlo- molge line appears to have been derived from a Gyrinophilus-Pseudotriton ancestral stock. Typhlo- triton is more primitive than Eurycea, and paired premaxillae, relatively long snouts, nasal-maxilla articulation, and absence of nasolacrimal duct impressions ally it with Gyrinophilus. Typhlotriton differs from other members of its line, and re- sembles Gyrinophilus, Pseudotriton, and Stereo- chilus in having continuous vomerine tooth series. Eurycea resembles Pseudotriton more closely than Gyrinophilus. Eurycea and Pseudotriton share several important characters, including fused pre- maxillae, relatively short snouts, marked nasolacri- mal duct impressions, and hypapophysial vertebral remnants. Lingual cartilages and parietal-otic and squamosal-otic crests, all present in Gyrinophilus, Pseudotriton, and Stereochilus, are present in Eurycea but absent in Typhlotriton. Typhlotriton and Eurycea are closely related, as is evident from the number of common characters, including: (1) similar hyobranchial structure, de- spite presence of membranous anterior attachments in Typhlotriton, (2) greatly expanded first basi- branchials, (3) short cornua, (4) well developed sublingual glands of unknown function which are poorly developed or absent in other genera, (5) extremely short transverse processes, (6) complete parapophysial and diapophysial separation, (7) parapophysial origin greatly in advance of dia- pophysial origin, (8) ribs longer than distance across parapophyses, (9) elongate centra, (10) well developed basapophyses. Apparently Typhlotriton is related to the Gyri- nophilus-Pseudotriton-Stereochilus assemblage, but it is closer to Eurycea. Many similarities between Typhlotriton and Eurycea suggest probable descent from an ancestral stock related to that from which Gyrinophilus and Pseudotriton were derived. Sep- aration of Typhlotriton and Eurycea from the more primitive genera is based primarily on the striking hyobranchial and vertebral differences. Hemidactylium is a member of the Hemidacty- liini but its relationships to other hemidactyliine genera are not clear. The outstanding primitive characters are retention of separated premaxillae, detoglossy, and genioglossal muscles. Hemidacty- lium has low numbers of trunk vertebrae, and is the only non-paedogenetic hemidactyliine that 64 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 does not have a tendency for opisthocoely. It re- sembles Gyrinophilus and allies in vertebral and hyobranchial characters, but resembles Eurycea in having separated anterior and posterior vomerine tooth series. The nasolacrimal ducts follow routes similar to those in Eurycea, and their paths are impressed on the prefrontals. Specialized features of H emidactylium include loss of the fifth toes and fifth distal tarsals, and presence of basally con- stricted tails with concomitant vertebral modifica- tions. H emidactylium has a puzzling mosaic of primi- tive and advanced characters indicating specializa- tion of a relatively primitive stock. It is closer to the plethodonine line than is any other hemidactyl- iine. Hemidactylium may be a specialized derivative of a primitive stock ancestral to all hemidactyliines. Despite the statements of Dunn (1926), Noble (1931), Piatt (1935), and others to the effect that Hemidactylium is a derivative or close rela- tive of Plethodon, there can be little question con- cerning its closer relationship to the hemidactyl- iine line. Bishop (1941) intimated that Hemi- dactylium was more distantly removed from Pletho- don than had formerly been suggested, but he did not mention possible relationship to other genera. The aquatic larval stage of Hemidactylium is clearly a primary, primitive character of great importance, not a derived feature as suggested by Noble (1927 a). The tribe Hemidactyliini presents a mosaic pat- tern of primitive and advanced characters. It is ob- vious that many ancestral and intermediate groups are no longer extant, and that parallel loss has occurred in regard to several characters. Status of the genus Manculus The genus Manculus, with a single species, M. quadridigitatus, has been recognized as forming a distinct group by most authors since 1869, when the genus was proposed by Cope. The sole excep- tion was Dunn (1923), who suggested that the single species be placed in the genus Eurycea. Dunn (1926) reiterated his position, stating, “I cannot think the genus Manculus worthy of recog- nition. The sole difference from Eurycea is the absence of the fifth toe.” Dunn’s suggestion was not followed and Noble (1927a) was soon recog- nizing Manculus; the genus has been recognized by all subsequent authors. No one has questioned the close affinity of Manculus and Eurycea. Noble (1931) frankly stated, “ Manculus is a dwarf form of Eurycea. . . It differs chiefly in the loss of the fifth (outer) toe.” Manculus differs from Eurycea in lacking fifth distal tarsals and toes, in being smaller, in having smaller anterior cranial elements, in having pond rather than stream larvae, and in being restricted to lowland habitats. The many striking similarities far outweigh the differences, however, and include similar hyobranchial structure, similar vertebrae, and similarly shaped and arranged cranial elements. Cranial similarities include presence of relatively short frontals and very distinct posterolateral frontal processes. Both Eurycea and Manculus lack well developed vomerine preorbital processes, but other hemidactyliines have the structures well developed. The posterior maxillary tips are rela- tively short and sinuous in both, and as a result, the subocular grooves intersect the lips below the eyes at the posterior maxillary tips. A connecting groove is present in many genera, but actual continuation of the subocular groove to the lip occurs elsewhere only in Thorius. Unquestionably Manculus is distinct from all other Eurycea, but when differences are weighed against similarities, it is apparent that Manculus is simply a specialized offshoot of the Eurycea line. No purpose is served by isolating the single species in a separate genus. Other genera which have lost fifth toes (Hemidactylium, Batrachoseps ) warrant generic recognition on the basis of marked differ- ences from other groups in numerous characters. It is recommended that the species quadridigitatus be placed in the genus Eurycea, and that Manculus be considered a subjective junior synonym of Eurycea. Status of the genera Haideotriton and Typhlomolge Following completion of this paper a new paedo- genetic and troglobitic species, Eurycea tridenti- fera, was described (Mitchell and Reddell, 1965). The authors considered the new species to be in- termediate between Typhlomolge and Eurycea and suggested that the former be placed in the syn- onymy of the latter. Through the kindness of James P. Bogart I have recently had an opportunity to examine three specimens of E. tridentifera and one of E. troglodytes. Information concerning these species is limited to this portion of the paper. My investigations, discussed below, indicate that the genus Typhlomolge is distinct from Eurycea and contains two species ( tridentifera and rathbuni) . Paedogenetic plethodontids occur in four general regions: 1) northern Alabama and southeastern Tennessee (Gyrinophilus palleucus), 2) southwest- ern Georgia and north-central Florida ( Haideo- triton wallacei), 3) eastern Oklahoma (Eurycea tynerensis), and 4) central and west-central Texas, along the Balcones Escarpment and on the Ed- 1966 EVOLUTION OF LUNG LESS SALAMANDERS 65 wards Plateau (Typhlomolge rcithbuni, T. tridenti- fera; Eurycea Icititans, E. nana, E. neotenes, E. pterophila, E. troglodytes). Evolution of paedo- genesis has apparently occurred independently in each of these areas. G. palleucus is clearly assign- able to Gyrinophilus, and is not a close relative of other paedogenes. Haideotriton shows as many similarities to larvae of such forms as E. bislineata and E. longicauda as to the Texan paedogenes. E. tynerensis seems to be more closely related to E. multiplicata of the Interior Highlands than to the Texas Eurycea. The Texas Eurycea form a quite distinct group and there is no clear evidence of close relationship to any single species of trans- formed Eurycea. The three epigeal species, E. nana, E. neotenes, and E. pterophila, are the least specialized of the Texan species and resemble generalized plethodon- tid larvae. Mitchell and Reddell (1965) have cor- rectly stated that specializations related to sub- terranean life become increasingly pronounced in the order E. latitans, E. troglodytes, T. tridentifera, and T. rathbuni. Specializations include such fea- tures as eye degeneration, skin depigmentation, limb attenuation and elongation, snout depression, truncation and elongation, and reduction in num- ber of trunk vertebrae. Mitchell and Reddell agreed with Baker (1957) that each species has evolved independently of the others since the original colonization of the subterranean habitat by the ancestral stock, and that present similarities of the species are likely due to convergent evolution. Unquestionably Typhlomolge rathbuni is the most highly specialized and distinctive species of the Texan group. The species is much larger, has longer limbs, more extremely specialized head and snout, more reduced eyes, and is more depigmented than any of the other species. In addition its dis- tinctive skull is greatly flattened and broadened. The toothed portion of the premaxilla is much broader and the frontal processes are broader and more depressed than in any of the other species. The palatopterygoids are very widely separated and are longer and more spatulate than in the other species. There are no orbitosphenoids. The pari- etals bear distinct crests, and the otic capsules are disproportionately large. Transverse processes are relatively long and extend beyond the lateral mar- gins of the zygapophyses. There are only 13 or 14 trunk vertebrae. The paedogenetic species of Eurycea (tridenti- fera excluded) all differ from T. rathbuni and re- semble nonpaedogenetic species of Eurycea in hav- ing more than 14 trunk vertebrae, diapophyses which extend no farther than the zygapophysial margins, and well developed orbitosphenoid bones. In addition all differ from T. rathbuni and resemble larval Eurycea in general skull dimensions and in the relative development of skull elements. All have very well developed columellae, thus resem- bling Typhlomolge rather than Eurycea. Mitchell and Reddell (1965), citing Burger, Smith and Potter (1956) and Baker (1957), state that the second basibranchial (posterior basibranchium) is continuous with the hyobranchial apparatus proper in T. tridentifera and T. rathbuni and separated from it in the other species; the elements are also continuous in my E. nana, E. pterophila and E. troglodytes. T. tridentifera resembles the paedogenetic Texan Eurycea in general proportions of the anterior cran- ial elements, but in few other characters. Most of the skeletal similarities are in features common to all plethodontine larvae. The otic capsules are large and resemble those of T. rathbuni. Like T. rathbuni, the orbitosphenoids are absent, the diapo- physes are long and extend beyond the lateral margins of the zygapophyses, the numbers of trunk vertebrae are reduced (14 in two specimens, 13 in one), and the adpressed limbs overlap. Small alar processes are present on the anterior margins of the parapophyses as in T. rathbuni. Such proc- esses are not found in other paedogenetic Eurycea. Basapophyses are absent; the processes are very well developed in E. pterophila, and are present and usually well developed in the other Eurycea. Basapophyses are very poorly developed in T. rathbuni. Several unique features occur in tridenti- fera. The columellae are longer than in any other plethodontid and have a distinct bow (the columel- lae are also very long in T. rathbuni). In addition the columella may have a dorsolateral spur near the opercular-columellar junction. The columellar process of the suspensorium is ossified (or heavily calcified), a unique condition among plethodontids. There are only eight tarsals in the single tarsus available, a result of the fusion of distal tarsals four and five. Eye reduction reaches an extreme in T. rathbuni and Haideotriton wallacei, in which lenses are absent. Lenses are absent in about half the tridentifera known (Mitchell and Reddell, 1965), but are present in other Texan paedogenes. T. tridentifera is much smaller than T. rathbuni, has more pigmentation, lacks parietal crests, and has a less extremely specialized snout (althousffi snout flattening is more extreme than in any Eurycea). Larvae of non-paedogenetic plethodontids are all basically very similar with only minor sub- familial differences. Interpretation of the charac- 66 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 ters of paedogenetic species has been partially dis- cussed by Dundee (1962) who emphasized prob- lems of definitions of higher categories. Interpreta- tion is especially difficult in the case of the Texan paedogenes, between which greater differences are found than occur between the larvae of distantly related genera. The fact that morphoclines exist in regard to several characters with one of the less specialized Eurycea paedogenes on one end and T. rathbuni invariably on the other is not neces- sarily evidence that all have had a common ancestor or that these are chronoclines. No information concerning selection pressures in the various en- vironments is available, but it is known that the surrounding terrain is inhospitable to transformed hemidactyliines. In order to survive in the area any hemidactyliine would have been forced to remain in the larval habitat or to enter the permanent waters of springs and underground caverns. The Texan paedogenes are a diverse lot and need not have a common ancestor. It seems inadvisable to refer T. rathbuni to Eurycea until more information con- cerning many aspects of the biology of all the species involved becomes available. Mitchell and Reddell (1965) suggest that T. tridentifera is intermediate between T. rathbuni and the known paedogenetic Texan Eurycea. It is obvious from the above discussion that tridentifera is, in regard to most characters, much more similar to T. rathbuni than to any other species. Because the diversity within the Texan paedogene group is so great and because the vertebral and orbitosphe- noid characters enable the group to be conveniently divided, I choose to recognize the genus Typhlo- molge and refer to it the species rathbuni and tridentifera. The species E. troglodytes is the near- est to an intermediate but it is closely related to other Eurycea and is intermediate only in being largely depigmented, in having reduced eyes, and in having relatively elongate, attenuated limbs. In some ways the genus Haideotriton is more similar to the larvae of non-paedogenetic Eurycea than are either T. rathbuni or T. tridentifera. The vertebrae of Haideotriton, like Eurycea, have short diapophyses and lack parapophysial alar processes. Limbs are not nearly as long as in Typhlomolge. H. wallacei resembles Typhlomolge in having poor- ly developed basapophyses, 14 trunk vertebrae, lenseless eyes, in lacking orbitosphenoid bones, and in being greatly depigmented. The snout of H. wallacei is more truncate and the otic capsules are smaller than in Typhlomolge. The skull is rela- tively much larger, both longer and broader, than any Eurycea. The toothed portions of the premaxil- lae are much larger than those of any Eurycea or than T. tridentifera and approach those of T. rath- buni in size. The vomers are very broad and “L” shaped, and the palatopterygoids are large and spatulate; both elements resemble those of T. rath- buni. The skull presents an almost rectangular ap- pearance, unique in the family, because of the greatly expanded snout and relatively small otic capsules. Dundee (1957) reported some success in arti- ficially inducing transformation in T. rathbuni, but the species metamorphosed to a lesser extent than paedogenetic Eurycea. Atrophy of labial folds, gills, and fins was noted. The gill slits remained open. Maxillae, maxillary teeth, and a small un- paired anterior ossification (septomaxilla or nasal?) appeared. One of the specimens of T. tridentifera examined was kept alive by James P. Bogart for five months and was treated with thyroxin as well as with pituitary implants. The gills are greatly re- duced and only the anterior slit remains open. Some changes are apparent in the labial region. Skull shape is virtually identical to that of smaller, un- treated individuals despite the appearance of an elongate pars dentalis of the maxilla and a series of maxillary teeth. Small, paired nasals (?) are also present. Coronoids are present despite the fact that they are normally among the first of the larval elements to disappear during metamorphosis. The margins of the palatopterygoids are eroded. Appar- ently T. tridentifera transforms under the influence of metamorphosing agents to about the same degree as T. rathbuni, and to a much lesser degree than paedogenetic Eurycea ( E . nana, E. neotenes, E. tynerensis; Dundee, 1962). In contrast to the paedogenetic species of Gyri- nophilus and Eurycea, Haideotriton wallacei under- goes only a few minor integumentary changes, loses only a single bone (coronoid) and gains no new bony elements when treated with metamor- phosing agents (Dundee, 1962). Dundee concludes that Haideotriton, like Typhlomolge and Necturus, appears to be genetically resistant to thyroid secre- tions. He states further that H. wallacei is the most resistant to thyroxin of all plethodontids investi- gated. Haideotriton is clearly a close relative of Eurycea. Because the genus is biologically distinct from all others and because there is no evidence that it evolved from a direct ancestor either of Typhlomolge or of the paedogenetic Eurycea, the genus Haideotriton is maintained. Mitchell and Reddell (1965) have correctly noted the lack of information concerning evolu- tionary rates and selection pressures in subterra- nean habitats. Until such information becomes available it is not advisable to assume that all the 1966 EVOLUTION OF LUNGLESS SALAMANDERS 67 subterranean plethodontids of the Edwards Plateau entered the underground environments approxi- mately at the same time. It is more likely that those with the more extreme adaptations entered the environments first. Perhaps the ancestral Typhlo- molge stock entered the subterranean habitats at a relatively early date, and the Eurycea somewhat later. Anatomical evidence indicates that both gen- era evolved from the same general evolving ances- tral stock but that Typhlomolge was derived from a pr e-Eurycea ancestor and the other Texan paedo- genes from a Eurycea or very Eurycea-Yike ances- tor. Haideotriton seems not to be closely related to Typhlomolge. It probably originated from a Eurycea-Yike stock in the Southeast. Comments concerning Typhlomolge found else- where in this paper are based on study of T. rath- buni and do not necessarily refer to T. tridentifera. Relationships of the plethodonines Ensatina, Plethodon, and Aneides are a compact group of closely related genera which form the tribe Plethodontini. I have discussed their relation- ships recently (Wake, 1960; 1963), and only a re- view is presented here. The plethodonine genera appear to have been derived from a relatively primitive, common ances- tral stock. Plethodon is the central and most gen- eralized genus, and the other two are specialized offshoots. Ensatina is a primitive genus, and was probably derived from a Plethodon-Aneides an- cestral stock. It is more closely related to Plethodon than to Aneides. Aneides resembles Plethodon much more closely than it does Ensatina. Char- acters which relate Aneides to Ensatina are those characteristic of the tribe and are shared by Pletho- don as well. Aneides is the most specialized and advanced of the genera, and is closely related to Plethodon. My ideas concerning plethodonine re- lationships are illustrated in Figure 15. Primitive features of Ensatina include paired premaxillae (shared with Plethodon ), very high numbers of maxillary, dentary, and vomerine teeth, very long vomerine preorbital processes, low num- bers of trunk vertebrae, poorly developed otic crests, and relatively large otic capsules. Special- izations found in Ensatina include basally con- stricted tails, shortened first caudal vertebrae, elongated limb elements, especially fore limbs, ac- quisition of considerable tongue freedom, rather elongated genioglossal muscles, and possibly prim- ordial lingual cartilages. Ensatina differs further from Aneides and Plethodon in details of hyo- branchial construction (shorter cornua, larger epi- branchials and second basibranchials, small and unexpanded anterior first basibranchial extensions) and in vertebral characters (fused hyperapophy- ses, lack of caudal transverse processes, shorter ribs) . Plethodon and Ensatina resemble each other and differ from Aneides in sharing such generalized characters as paired premaxillae, unmodified maxillae, relatively weak prefrontal-maxilla articu- lations, relatively unmodified dentition, and primi- tive tarsal and carpal configurations. Plethodon has lower numbers of maxillary, dentary, and vomer- ine teeth than Ensatina, and has a great amount of vertebral number variation. Plethodon is a gen- eralized group, and only a few specializations oc- cur, including elongation of certain species (cinereus, elongatus, neomexicanus, richmondi), unusual otic crests (elongatus), and reduction in fifth toe phalangeal number (larselli, neomexi- canus). Aneides is so similar to Plethodon that it is tempting to postulate direct origin from Plethodon. Primitive species of Aneides (especially female hardii) differ from Plethodon only in having fused premaxillae and modified tarsal configurations. Al- most all of the many specializations which occur in the advanced species of Aneides are either feeding or climbing adaptations. Advanced species of Aneides differ sharply from Plethodon because of these specializations, which include reduction in number but increase in size of all teeth, sabre-like, enlarged, unicuspid maxillary and dentary teeth, enlarged otic crests that serve as points of origin of enlarged cephalo-hyo-mandibulare musculature, specialized interlocking maxilla-prefrontal articula- tions, loss of the vomerine preorbital processes, greatly elongated limbs, bifurcated and decurved terminal phalanges, greatly enlarged, edentulous posterior maxillae with cleaver-like cutting edges, hypertrophied quadrato-pectoralis musculature and prehensile tails. Despite the multitude of differences separating advanced species of Aneides from Plethodon, the presence of such primitive species as A. hardii clearly reveals close relationship of the two genera and suggests origin from either a common ancestral stock or of origin of Aneides from Plethodon. Relationships of the bolitoglossines Bolitoglossine salamanders form three groups sufficiently distinct to warrant recognition of super- genera. These groups are the supergenus Hydro- mantes with a single genus, the supergenus Bat- rachoseps with a single genus, and the supergenus Bolitoglossa with seven genera ( Bolitoglossa , Oedi- 68 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 Figure 15. Diagrammatic representation of the relationships of the plethodonine genera. 1966 EVOLUTION OF LUNG LESS SALAMANDERS 69 pina, Pseudoeurycea, Chiropterotriton, Parvimolge, Lineatriton, Thorius ) . The supergenera Bolitoglossa and Batrachoseps share several important characters that are absent in Hydromantes, including fused premaxillae (with fused frontal processes in some species of each group), similar basibranchial cornua, posterolat- eral parietal spurs, columellar reduction, and multilobed testes. James Kezer has told me {in litt.) that Batrachoseps and the supergenus Bolito- glossa (he has not examined Parvimolge ) have haploid chromosome numbers of thirteen, while Hydromantes and all other plethodontids examined have fourteen. One species of Batrachoseps ( B . wrighti ) has paired premaxillae in adults and is similar in that respect to Hydromantes, but the evolutionary trend in Batrachoseps is in the direction of premaxillary fusion, as in the supergenus Bolitoglossa. Hydro- mantes and Batrachoseps resemble one another, and differ from supergenus Bolitoglossa, in having three rather than two caudosacral vertebrae and in the universal presence of septomaxillae. How- ever, two caudosacral vertebrae occur in at least some individuals of all species of Batrachoseps, and in all members of three species. Hydromantes is very similar to certain members of the supergenus Bolitoglossa ( e.g ., Bolitoglossa ) in habitus and external morphology. In addition Hydromantes and the neotropical genera resemble each other, and differ from Batrachoseps, in having truly free tongues with no genioglossal muscles, low numbers of trunk vertebrae, tendencies for uni- cipital ribs and loss of diapophyses, loss of omo- hyoideus muscles, and five rather than four toes. Batrachoseps and Hydromantes are relatively primitive, and retain such primitive features as three caudosacral vertebrae (some individuals) and paired premaxillae (B. wrighti). Batrachoseps is highly specialized, however, and is elongated with increased numbers of trunk vertebrae, multi- lobed testes, tendencies for fusion of the premaxil- lary toothed and frontal processes, loss of pre- frontals, reduced frontals and parietals, well developed parietal spurs, and no fifth toes and fifth distal tarsals. It retains cornua, genioglossal mus- cles, and modified detoglossy, all relatively primi- tive features, but Hydromantes has lost both cornua and genioglossal muscles and has adetoglossy. Hydromantes is the most generalized of the three supergenera, retaining such primitive characters as unilobed testes, relatively large and separated premaxillae with relatively attenuated frontal processes, large columellae, and spurless parietals. Bolitoglossa is the most advanced of the three supergenera, and advanced characters found in at least several genera include septomaxillary, maxil- lary tooth, and columellar loss, tendencies toward opisthocoely, acquisition of basapophyses, distinct tail constrictions, phalangeal modifications, and mesopodial fusions and reductions. Advanced characters shared with one or the other of the bolitoglossine supergenera include bilobed testes, reduced numbers of chromosomes, tendencies for unicipital ribs, fused premaxillae, and adetoglossy. Hydromantes is apparently closer to the bolitog- lossine ancestral stock than are either of the other two supergenera. Both Hydromantes and Batra- choseps are more closely related to the advanced Bolitoglossa than to each other, but Batrachoseps may be closer to Bolitoglossa than is Hydromantes (see Fig. 16) . Figure 16. Diagrammatic representation of the relation- ships of the bolitoglossine supergenera. Tanner (1952) presented a diagrammatic family tree of the free-tongued plethodontids, and dis- cussed relationships of these genera. His study was based on throat musculature and hyobranchial cartilage structure, and he assumed that the free- tongued genera formed a natural group. Tanner suggested than an ancestral stock had given rise to two diverging generic groups, one containing Gyri- nophilus, Pseudotriton, Eurycea, and Manculus, and the other containing Hydromantes and the neo- tropical genera. Hydromantes was considered to be close to Bolitoglossa and Magnadigita, but the 70 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 other neotropical genera were placed in a separate line. Within the latter group three divergent evolu- tionary lines were recognized: (1) Thorius, (2) Oedipina, (3) Parvimolge, Pseudoeurycea, Chiro- pterotriton, Lineatriton. Tanner stated that he con- sidered association of Batrachoseps with Hydo- mantes and the supergenus Bolitoglossa, as had been proposed by Piatt (1935), to be entirely un- warranted. Although I strongly disagree with Tanner’s con- clusions concerning relationships of the neotropical genera to other plethodontids, our views of rela- tionships within the neotropical group are similar, with one notable exception in the case of Thorius. Salamanders of the supergenus Bolitoglossa pre- sent a bewildering mosaic of characters and an extremely complex evolutionary pattern. Three major generic groups may be recognized: (1) the Bolitoglossa group with one genus and about fifty species, (2) the Oedipina group with one genus and about sixteen species, (3) the Thorius group with five genera (Thorius, Pseudoeurycea, Chiro- pterotriton, Parvimolge, Lineatriton ) and about forty-six species. Bolitoglossa differs from the other genera of the supergenus Bolitoglossa in lacking sublingual folds. The folds are advanced characters and are related to the elongation of the ceratohyals into the an- terior portions of the mouths. Within the Bolito- glossini, Bolitoglossa, Hydromantes, and Batracho- seps are the only genera that lack the folds. Wake and Brame (1963a, 1963b) and Brame and Wake (1963) have presented evidence that the genera Magnadigita and Bolitoglossa in the sense of Taylor (1944) are not recognizable and have assigned all species to Bolitoglossa. It has been possible to examine skeletons of only about half of the species but all sections of the genus are represented. A number of osteological characters that vary interspecifically have been discovered, but none coincide with the old generic division based on amount of webbing (complete in Bolitoglossa, less extensive in Magnadigita) . Detailed discussion of osteology and evolution of Bolitoglossa is de- ferred to a later date, but it is clear that a mosaic distribution of characters occurs. Species groups which bridge the old genera are recognizable, but only on internal characters; none warrants generic rank. Oedipina differs from all other genera of the supergenus Bolitoglossa in having eighteen to twenty-two rather than fourteen trunk vertebrae. The marked elongation of the trunks and tails are specializations related to the evolution of special- ized behavior and ecology of Oedipina. Oedipina and Bolitoglossa are offshoots in dif- ferent directions from the ancestral stock that gave rise to the relatively generalized members of the Thorius group. Bolitoglossa is the more primitive of the two, and both contain highly specialized species. Specialized characters of Bolitoglossa in- clude loss of the septomaxillae and columellae by most species, and the prefrontals and tibial spurs by many, presence of basapophyses in many spe- cies, vomerine reduction in many species, tenden- cies for increased amounts of hand and foot web- bing, and presence of mesopodial fusions. Oedipina has a number of specialized and advanced char- acters, including high vertebral numbers, many unicipital ribs, no septomaxillae (most species), prefrontals, columellae, or tibial spurs, the most specialized and advanced premaxillae in the family, and extensive mesopodial fusions and phalangeal losses. The Thorius group of genera differs from Boli- toglossa and Oedipina in that distinct tendencies for intervertebral cartilage calcification are evident and transitional stages between amphicoely and opisthocoely occur. Included in the Thorius group are the most generalized and primitive species of the supergenus Bolitoglossa, but other species are among the most highly specialized and advanced in the family. Primitive Chiropterotriton retain such generalized and primitive features as septomaxillae, prefrontals, lingual cartilages, well developed colu- mellae, vomerine preorbital processes, unfused pre- maxillary frontal processes, tibial spurs, nine tar- sals, and eight carpals; but primitive species have advanced tarsal arrangements. The most advanced Chiropterotriton may lack septomaxillae, pre- frontals, columellae, lingual cartilages, vomerine preorbital processes, and tibial spurs, and may have fused premaxillary frontal processes, seven car- pals, and eight tarsals in a primitive configuration. All members of the genus retain bicipital ribs. Ad- vanced species have strongly opisthocoelous verte- brae. Primitive Pseudoeurycea have prefrontals, lin- gual cartilages, reduced columellae, vomerine pre- orbital processes, unfused premaxillary frontal processes, tibial spurs, eight carpals, nine tarsals, primitive tarsal arrangements, and bicipital ribs. These features are common to most species. Septo- maxillae are absent in most species. No Pseudo- eurycea have truly opisthocoelous vertebrae, but some intervertebral cartilage calcification occurs. Pseudoeurycea and Chiropterotriton are very closely related, and can be separated consistently only on the basis of tarsal structure. Primitive and most advanced species of Chiropterotriton have 1966 EVOLUTION OF LUNGLESS SALAMANDERS 71 specialized tarsi in which the fifth distal tarsals are larger than the fourth and articulate with the cen- trale. All Pseudoeurycea retain primitive tarsi in which the fourth distal tarsals are larger than the fifth, and articulation of the fourth tarsals with the fibulars eliminates the fifth tarsals from articu- lation with the centrals. Some advanced species of Chiropterotriton ( xolocalcae , bromeliacia) have the primitive configuration encountered in Pseudoeu- rycea but they can be distinguished from all Pseudoeurycea because they lack prefrontals. Prim- itive Chiropterotriton can usually be distinguished from Pseudoeurycea by having septomaxillae, but advanced species lack septomaxillae and some spe- cies of Pseudoeurycea may have the structures. Tanner (1952) found differences in the details of hyobranchial structure between the genera, but I cannot corroborate his observations. Despite the fact that primitive members of Chiropterotriton and Pseudoeurycea are very similar, it is apparent that two groups are represented and the two genera are retained. Tanner (1952) considered Pseudoeurycea to be the most primitive of the neotropical plethodon- tids. Pseudoeurycea and Chiropterotriton share several primitive and generalized characters, but each retains primitive characters reduced or lost in the other. As a genus Chiropterotriton has radiated and specialized to a greater extent than has Pseudo- eurycea, but primitive members of both genera are almost equally primitive. If the total species com- plements are considered, the largely ground-dwell- ing, morphologically relatively uniform Pseudoeu- rycea must be considered more primitive than the largely arboreal, morphologically diverse Chirop- terotriton. Primitive members of the two genera are very closely related and have diverged only slightly from common ancestral stocks. The three remaining genera of the Thorius group apparently diverged from a pr e-Pseudoeurycea- Chiropterotriton stock and have become highly specialized. Parvimolge and Lineatriton are prob- ably closer to each other than either is to any other genus, but the two are also relatively closely related to both Pseudoeurycea and Chiropterotriton. The single species of Lineatriton is an elongate form that has unicipital ribs, elongated vertebrae with only parapophyses on most vertebrae, and noticeably shortened limb elements. Septomaxillae have been lost. The genus is somewhat closer to Pseudoeurycea than to Chiropterotriton, but it is distinct from both. Resemblance to Oedipina is probably the result of parallel evolution. The genus Parvimolge is probably an unnatural assemblage. Adequate material has been unavail- able, and final assignment of the three species re- mains uncertain. I have not examined skeletal ma- terial of P. praecellens (known only from the unique type). P. townsendi has been considered to be very closely related to Pseudoeurycea by Tanner (1952) on the basis of similarities in throat muscu- lature, and Uzzell (1961) has suggested that the species is closely related to Lineatriton because of similarities in hyoid and mesopodial calcification patterns. P. townsendi resembles Chiropterotriton in having septomaxillae and primitive species of both Chiropterotriton and Pseudoeurycea in having such generalized features as columellae (reduced), bicipital ribs, prefrontals, vomerine preorbital proc- esses, and tibial spurs. The species differs from both genera in being dwarfed, in having distinc- tively shaped hands and feet (see Taylor, 1944), in having distinctive, large, dorsal glands, and in having the fourth and fifth distal tarsals fused. The species differs further from most Pseudoeurycea in having septomaxillae, and from all members of that genus in having eight rather than nine tarsals. The species richardi was placed in Parvimolge by Taylor (1949), presumably on the basis of hand and foot shape. Rabb (1955) described P. praecel- lens, including it in Parvimolge because it shared several features with P. townsendi, including simi- lar feet, coloration, conspicuous dorsal glands, and tail shape. Dorsal glands are absent in richardi, and Rabb suggested that absence of the glands, differ- ences in foot structure, and enormous range dis- junction ( richardi in Costa Rica, praecellens and townsendi in Veracruz, Mexico) imply that rich- ardi is not congeneric with praecellens and town- sendi. Septomaxillae, reduced columellae, lingual carti- lages, and mesopodial and hyobranchial calcifica- tions all occur in P. townsendi, but not in P. richardi. Nothing is known concerning the meso- podial pattern in richardi, but there is no reason why richardi cannot be placed in Chiropterotriton despite some differences in external foot morpho- logy. It is true that the characters lacking in richardi but present in townsendi are those often lost in southern species of Chiropterotriton. This fact might support the suggestion that richardi is a mem- ber of Chiropterotriton, but it is also possible that the same losses have occurred in parallel in Parvi- molge. Until more information concerning the spe- cies presently assigned to Parvimolge becomes available the problem is insoluble, and richardi is tentatively retained in the genus. Thorius probably represents the earliest diver- gent stock of the line that leads to Pseudoeurycea and Chiropterotriton (Fig. 17). Dwarfing is severe 72 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 in the genus, and the cranial elements are greatly reduced in size and modified in appearance. The genus is further characterized by such specializa- tions as loss of columellae and vomerine preorbital processes, variable loss of septomaxillae, high fre- quency of mesopodial calcifications, fusion of distal tarsals 4 and 5, intervertebral cartilage calcification, and presence of posteriorly directed squamosal processes. Primitive characters retained by the genus include bicipital ribs, fourteen trunk verte- brae, and tibial spurs. Highly modified prefrontals are usually present. This mosaic pattern of general- ized and specialized characters indicates that mod- ern Thorius is the specialized end point of a line derived from a stock ancestral to other genera of the Thorius group. Calcification and ossification of such elements as the ends of long limb bones, inter- vertebral cartilages, and mesopodials may be inter- preted as compensations for paedomorphic influ- ence that has led to general skeletal dwarfing and reduction (see below, Paedomorphosis). Calcified mesopodials have been considered diagnostic of the genus since the time of Cope (1869), but the character lacks reliability. The largest Thorius that I have examined (28.2 mm.) has uncalcified mesopodials (see also Uzzell, 1961). There is no justification in removing Thorius from other neo- tropical genera and placing it in its own subfamily, the Thoriinae, as Cope (1893), and more recently Smith and Taylor (1948), have suggested. I also consider the genus less distinct from other neo- tropical genera than did Tanner (1952), and think that Thorius is closer to Pseudoeurycea, Chiropter- otriton, Parvimolge, and Lineatriton than are either Oedipina or Bolitoglossa. To briefly summarize, the tribe Bolitoglossini Figure 17. Diagrammatic representation of relationships within the supergenus Bolitoglossa. contains three distinct lines. The ancestral stock has given rise to the relatively unspecialized Hydromantes. Two main specialized branches have resulted in the related supergenera Batracho- seps and Bolitoglossa (see Fig. 16). Within the supergenus Bolitoglossa there has been a further divergence, with the Thorius group of genera closest to the ancestral stock. Within the Thorius group Pseudoeurycea and Chiropterotriton are the most primitive genera, and Thorius is the most specialized and advanced. Thorius probably rep- resents the first divergence from the group ancestral stock, and Lineatriton and Parvimolge have di- verged somewhat more recently. Bolitoglossa and Oedipina represent offshoots in different evolu- tionary directions from the ancestral stock of the 1966 73 EVOLUTION OF LUNGLESS SALAMANDERS Thorius generic group. Oedipina is the most spe- cialized and advanced plethodontid genus. Trends Ecology and Life History The ancestral adaptive zone Simpson (1953: 199) states that at any instant in time environments and organisms of a given taxonomic level define a field or type of adaptation characteristic of the group. These adaptive types can be visualized if represented on paper as bands or zones that form sorts of grids (Fig. 18). Zone breadth is a reflection of general adaptation; the broader the zone, the more generally adapted are the organisms. Such zones involve abstraction and oversimplification but are useful in explaining evo- lutionary patterns and pathways. Simpson empha- sizes that such a zone is not geographic, physical, or even environmental in character, but “an adap- tive zone, representing a characteristic reaction and mutual relationship between environment and or- ganism, a way of life and not a place where life is lived.” I conceive of an adaptive zone as the sum total ecological niche relationships of a group of organisms with the same base level of general adaptation. Thus modern reptiles can be said to occupy a terrestrial adaptive zone in a broad sense, because of the general adaptational level associated with amniotic egg evolution. Birds, close reptilian relatives, have entered a new aerial adaptive zone as a result of a base level of general adaption (in the broadest sense) associated with the evolution of feathered flight. In order to establish the ances- tral adaptive zone of plethodontid salamanders it is essential to briefly consider the relationships of the family to other salamander families, and the nature of the ancestral plethodontid stock. Plethodontid origins Cope (1866, 1867, 1869, 1889) considered plethodontids to be closely related to and possibly derived from ambystomatid salamanders. Stejneger (1907) placed the families Ambystomatidae (in- cluding hynobiids), Salamandridae, and Pletho- dontidae (including desmognathines) in a super- family Salamandroidea, a group first proposed by Sarasin and Sarasin (1887-1890) for all salaman- ders except amphiumids. Dunn (1922) completed the transition from Cope’s ambystomatid theory by stating “it is probable that some primitive sala- mandrid . . . gave rise to the much degenerate Plethodontidae.” His statement was based on a study of the otic apparatus, and on the work of Wilder (1920) which demonstrated that pletho- dontid parasphenoid teeth were ontogenetically de- rived from the anterior vomerine teeth. The vom- erine arrangement of plethodontids, Dunn argued, could most easily have been derived from the pos- teriorly directed vomerine tooth series typical of salamandrids. In 1926 Dunn included six families (Proteidae, Ambystomatidae, Pleurodelidae, Sala- mandridae, Plethodontidae, Amphiumidae) in the suborder Salamandroidea. He declared that “the Plethodontidae can be most easily derived from the Salamandridae.” Dunn’s ideas have been wholly accepted by most recent workers ( e.g Noble, 1931; Piatt, 1935; Herre, 1935; Tanner, 1952; Auffenberg, 1961). Noble (1931) added one mod- ification — the ambystomatids were placed in a separate suborder, the Ambystomatoidea. Since that time salamandrids and plethodontids have continued to be associated in the same suborder, and ambystomatids have been placed in a separate suborder (Goin and Goin, 1962). Laurent (1947) thought the vomerine tooth pat- terns observed in salamandrids and plethodontids represented separate evolutionary derivations, and he suggested that both conditions could have evolved from an ambystomatid type. He also cited similarities in the vertebrae of plethodontids and ambystomatids (primitively amphicoelous versus opisthocoelous in salamandrids), presence of septo- maxillae and costal folds in both families (absent in salamandrids), and absence of paraoccipital processes and fronto-squamosal arches in both (both present in primitive salamandrids). Laurent concluded that the suborder Salamandroidea of recent workers was an artificial assemblage, and suggested that the Plethodontidae be housed in a third suborder, the Plethodontoidea. Laurent’s Am- bystomatoidea includes only the Ambystomatidae, the Plethodontoidea includes only the Plethodonti- dae, and the Salamandroidea includes the Sala- mandridae, Amphiumidae, Proteidae, and Sireni- dae. As an alternative Laurent suggests inclusion of all of the above families in a single suborder, the Salamandroidea. Teege (1956) found that plethodontid vertebrae were more similar to those of ambystomatids than to salamandrids. She was unable to find supporting data for Dunn’s theory. Larsen (1963) found little to: support salamandrid origin of plethodontids based on his study of cranial morphology, and suggested that plethodontids and ambystomatids were rather closely related. Both groups were thought to have arisen from a prohynobiid an- cestor. Recently Monath (1965) has shown that the opercular apparatus in plethodontids is basic- ally quite distinct from that of salamandrids. He 74 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 has suggested that plethodontids were not derived from salamandrids, but from a primitive sala- mander stock. During the course of this study I have had oc- casion to investigate salamandrid, hynobiid, and ambystomatid anatomy. In addition to data pre- sented by Laurent, Teege, Larsen and Monath, I have noted similarities in the throat musculature of salamandrids and hynobiids, and in plethodon- tids, ambystomatids, and hynobiids. Ambystoma- tids and plethodontids also resemble one another, and differ from salamandrids, in hyoid and meso- podial structure. Larval and adult palatal and vom- erine structure of plethodontids is much more similar to that observed in hynobiids and amby- stomatids than that in salamandrids. I recognize four suborders in the order Caudata. The suborder Cryptobranchoidea includes the fam- ilies Cryptobranchidae and Hynobiidae. Members differ from the other suborders in having external fertilization and in retaining distinct angulars in the lower jaws. The suborder Sirenoidea includes only the family Sirenidae, a group recently raised to ordinal level as the Order Trachystomata by Goin and Goin (1962). Sirenids differ from other salamanders in lacking hind limbs and in the pecu- liar structure of their vertebrae (see Goin and Auffenberg, 1958), but resemble salamanders in cranial morphology, particularly of the palatal ele- ments, and in hyoid morphology. I agree with Estes (1965) in considering ordinal status un- warranted. The suborder Ambystomatoidea in- cludes the families Ambystomatidae and Pletho- dontidae, and differs from the remaining suborder, the Salamandroidea (Salamandridae, Proteidae, Necturidae, and questionably Amphiumidae), in that primitive ambystomatoid genera have vomer- ine preorbital processes which bear laterally ori- ented series of teeth, and have septomaxillae. My hypothesis of the present state of salamander evolution is that the modern and progressive sub- orders Ambystomatoidea and Salamandroidea rep- resent separate evolutionary lines that have di- verged independently from an ancestral stock similar to modern hynobiids. Ambystomatoids dominate the New World, and salamandroids domi- nate the Old World, but representatives of both occur in both regions. The foregoing account has been necessary be- cause most previous theories concerning pletho- dontid ancestry have assumed origin from sala- mandrid stocks. Despite their apparent error in establishing plethodontid ancestry, the most ac- ceptable theory concerning the ancestral adaptive zone (their ancestral home) was advanced by Dunn (1917), Wilder and Dunn (1920), and Dunn (1926). All plethodontids are lungless and lack ypsiloid processes on the pelvis. These losses are presumed to be rheotropic adaptations. The reasoning is that lungs are hydrostatic organs and disadvantageous for stream bottom dwellers. Ypsi- loid processes aid in emptying the lungs, and are of little value in lungless forms. Wilder and Dunn (1920) reviewed literature reports and showed that lunglessness or greatly reduced lungs are asso- ciated with mountain brook environments in European salamandrid genera (Euproctus, Sala- mandrina, Chioglossa). The list of nonplethodontid lungless forms should also include the ambystoma- tid genus Rhyacotriton of northwestern North America, and the Asian hynobiid genus Onycho- dactylus; both are mountain brook forms. Because the entire family Plethodontidae is lungless, Wilder and Dunn reasoned that the species have had a common lungless ancestor, and that this ancestor likely inhabited mountain brook environments. The southern Appalachian region (Appalachia) contains the greatest diversity of plethodontids, including all of the mountain brook genera and the most primitive members of the family. Appalachia is a very ancient upland, and presumably this rela- tively stable area has been the region of origin, the primary distribution center, and the dominant area of survival of the family Plethodontidae. The ancestral adaptive zone can be characterized as a way of life in semiaquatic, montane, warm temperate niches such as probably existed in late Mesozoic times in what is now eastern North America. Evolution within the adaptive zone The Desmognathinae and the Hemidactyliini have remained primarily within the ancestral adap- tive zone, but the Bolitoglossini and Plethodontini both have occupied new terrestrial adaptive zones. In addition there are tendencies in desmognathine and hemidactyliine genera toward splitting the old zone into subzones and toward moving out of the zone (see Fig. 18). Within the genus Desmognathus the most im- portant trend obviously is the one leading in the direction of increased terrestriality (see below), but primitive species have remained in the ancestral adaptive zone. Phaeognathus is probably a terres- trial species that has left the old adaptive zone. Leurognathus has remained in the ancestral zone, but selection has resulted in special adaptation and in restriction to a specialized segment of the an- cestral environment — the mountain stream bot- toms. Martof (1962) has recently considered in 1966 75 EVOLUTION OF LUNGLESS SALAMANDERS some detail the ecology of Leurognathus. He states that the genus is restricted to cool mountain streams and that only during times of droughts do Leuro- gnathus and D. quadramaculatus (the most aquatic Desmognathus) utilize a common habitat. Leuro- gnathus should not be fitted into the predominant adaptive sequence toward terrestrialism; the evo- lutionary trend in the genus has been from semi- aquatic to strictly aquatic existence. The tribe Hemidactyliini has remained closer to Figure 18. Adaptive zones and subzones in the family Plethodontidae. la, semiaquatic zone; lb, paedogenetic subzone; lc, aquatic subzone; Id, swamp-bog subzone; 2a, terrestrial zone; 2b, arboreal subzone; 2c, semifossorial subzone. the ancestral adaptive zone than any other group of plethodontids. Evolution within the group has been characterized by movements out of the ances- tral mountain brook environment into closely ad- jacent environments and by restriction to special- ized aquatic environments. Gyrinophilus and Pseudotriton have remained very close to the an- cestral environment and occur in spring and stream habitats. A single species, Gyrinophilus palleucus, is a paedogenetic form in subterranean waters in Tennessee and Alabama. Eurycea has become somewhat diversified. Prim- itive species such as E. aquatica, E. bislineata, and E. multiplicata inhabit brooksides and springs, sit- uations that resemble the ancestral habitat. E. long- icauda inhabits streams and springs, but also the twilight zone of caves, and E. lucifuga is essentially limited to the twilight zones of limestone caves (Hutchison, 1958). Several species of Eurycea are paedogenetic and have adapted to specialized aquatic situations. E. tynerensis, E. nana, E. neo- tenes, and E. pterophila inhabit springheads and cold swift streams. E. latitans occurs in cave waters, and E. troglodytes occurs in the waters of a large solution cave on the Edwards plateau of Texas (Baker, 1957; 1961). These species occur only at the periphery of the generic range. Specialization into restricted portions of the ancestral habitat or into specialized adjacent habitats has allowed survi- val in areas beyond the general generic range. 76 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 Typhlomolge, Haideotriton, and Typhlotriton are all close relatives of Eurycea. All are found in areas peripheral to Appalachia in specialized aquatic habitats. All are unpigmented, blind, and subterranean and have apparently survived in rather unfavorable climatic regions following cli- matic change by adapting to specialized, relatively very stable, underground habitats. E. quadridigitata is more terrestrial than other Eurycea, and is found in low swampy areas, trickles from springs, and sphagnum bogs (Bishop, 1943; Carr, 1940). Hemidactylium, Stereochilus, and E. quadridigitata have entered pond and swamp habi- tats, and this movement is associated with occu- pancy of new adaptive subzones (Fig. 18). Appar- ently three separate and parallel evolutionary lines are represented. All breed in ponds and swamps in relatively quiet waters. The larvae resemble each other and differ from all other plethodontids, in- cluding sympatric species, in possessing relatively high dorsal fins that extend onto the trunks of the animals (Bishop, 1941; Goin, 1951; Schwartz and Etheridge, 1954). Goin compared the rudimentary hind limbs of hatchlings of Hemidactylium with those of E. quadridigitata and noted that both have toeless lobes at early stages. Both have four toes as adults, and that fact may be related to the slow rate of development of the hind limbs in larval stages. Hind limbs of Stereochilus larvae appar- ently are better developed than the others. Move- ment out of the stream and into pond environments has been accompanied by acquisition of pond-type larvae. Significantly all have ranges that are periph- eral to those of the primitive hemidactyliine genera which have stream-type larvae. E. quadridigitata is semiaquatic, Stereochilus is aquatic, and Hemi- dactylium is terrestrial in the adult stage. Occupation of new adaptive zones and subzones Simpson (1953) has discussed the biological basis of evolutionary or radiative movements into new adaptive zones, and the acquisition of new ways of life. In order for populations to enter new adaptive zones there are, according to Simpson, three essentials. The populations must have physi- cal access to the new zones. They must also have the necessary preadaptations which will be of sig- nificant adaptive value in the new way of life, that is they must have evolutionary access. Finally, com- petitive encounters in the new zones must be at a minimum so that the populations have ecological access to the zones. Only when populations have all types of access can occupation of new adaptive zones begin. Evolutionary trends within the subfamily Des- mognathinae are particularly instructive in eluci- dating how occupation of new adaptive zones may have occurred. Although primitive desmognathines remain in the primitive adaptive zone, a significant diversification has occurred within the subfamily. Occupation of a new aquatic adaptive subzone by Leurognathus has been discussed above. A ter- restrial trend is evident in Desmognathus. Organ (1961 a) recently completed a significant analysis of the ecology of five species of Desmognathus which occur in a variety of habitats on Whitetop Mountain, Virginia. The most aquatic and general- ized is D. quadramaculatus, a species that inhabits streams and stream banks. D. monticola represents the next evolutionary stage and is found on stream banks and in seepage areas, habitats more terrestrial than those of quadramaculatus. D. fuscus is slightly more terrestrial than monticola, and D. ochro- phaeus is unquestionably more terrestrial than fuscus. D. wrighti is the terminal and most ter- restrial species within the series; it is primarily a forest floor form with only breeding females and hibernating individuals ever found in aquatic sites. The species arranged in order of increasing ter- restriality are quadramaculatus-monticola-fuscus- ochrophaeus-wrighti. Arranged in order of decreas- ing larval period, the species are quadramaculatus- fuscus-monticola-ochrophaeus-wrighti; wrighti has direct terrestrial development with no free-living larvae. Organ presented survivorship curves which show that male survivorship increases in the series quadramaculatus - monticola - fuscus-ochrophaeus- wrighti, a trend identical to the aquatic-terrestrial series. Female survivorship in quadramaculatus and monticola is about equal to that of males of the same species, but mortality of mature females of fuscus, ochrophaeus, and wrighti is considerably higher than that of males of the species at the same ages. Organ has shown that brooding females of ochrophaeus and wrighti return to aquatic sites, and, as a group, have higher mortality rates than those parts of the populations that remain ter- restrial during the brooding season. In addition the more terrestrial the species, the higher is the pro- portion of mature males to mature females in the populations; this is another indication of the higher mortality of mature brooding females in aquatic nesting sites. Plethodontid evolution has been characterized by parallel invasion and occupation of terrestrial adaptive zones (see below). As a result of Organ’s study it is now known that sur- vivorship is increased in terrestrial habitats within a series of related species and within populations of a single species. Selection favors terrestrial over less terrestrial individuals within the popu- 1966 EVOLUTION OF LUNGLESS SALAMANDERS 77 lations. It is likely that the populations involved in crossing the adaptive threshold between the semi- aquatic and terrestrial zones (probably in early Tertiary) were under strong environmental selec- tion pressures, and that the populations evolved sequentially at very rapid, progressive rates. Although the trend in Desmognathus is toward increased terrestriality, the genus has not entered a new adaptive zone. One species, D. wrighti, has abandoned the aquatic larval stage and is essentially terrestrial, but it is a highly specialized form that is unlikely to give rise to an adaptive radiation. Even if the species has the necessary evolutionary plasticity, it is doubtful that successful occupation of the new adaptive zone will be possible, for the terrestrial niches in Appalachia are already largely occupied by species of Plethodon and Aneides and ecological access is not available. Available information indicates that Phaeog- nathus is a terrestrial species (Valentine, 1963 b; 1963 c; Brandon, 1965). Characters that suggest terrestriality include terrestrial adult habitat, co- ossified skin on anterior cranial elements and elongation as a result of the addition of trunk vertebrae (both adaptions related to terrestrial burrowing), and large, yolky oviducal eggs re- sembling those of terrestrial plethodontines. Phaeognathus apparently represents an invasion of the terrestrial adaptive zone by a specialized, dead-end group. All genera of the tribes Plethodontini and Bolitoglossini have achieved true terrestriality. All lack free living larvae, and have directly develop- ing embryos that are retained within the egg mem- branes; eggs are laid on land. The two assemblages are distinct morphologically, and terrestriality in the groups may have evolved in parallel. Movement to land is truly movement into a new adaptive zone and adoption of a new way of life for salamanders. Terrestriality has resulted in a significant elevation of the base level of general adaptation, which has been followed in parallel lines by significant evo- lutionary diversifications and adaptive radiations. The terrestrial environment borders the ancestral mountain brook habitat, the presumed home of plethodonine and bolitoglossine progenitors, and physical access to the new adaptive zone was no great problem. At the time the adaptive shift was proceeding (Cretaceous or early Tertiary periods) the ancestral environment was probably crowded with salamanders and competition was keen. By contrast the terrestrial environment adjoining was relatively unpopulated by potential competitors so that the pioneers had ecological access to the new zone. Today plethodontids have a wide variety of larval types ranging from permanent larvae (Typhlomolge) and species with long larval stages (Desmognathus quadramaculatus, Gyrinophilus) to species with brief larval stages (Desmognathus aeneus, Hemidactylium). The larvae may be stream or pond types. The variety of larvae in modern forms and the presence of distinct terrestrial trends in a single, primitively aquatic genus (Desmog- nathus) are indications of the evolutionary plasticity of plethodontid developmental processes. This plas- ticity might be considered the prospective adapta- tion which provided evolutionary access to the new terrestrial adaptive zone. Diversification usually follows occupation of new adaptive zones with the extent of subsequent adap- tive diversity roughly proportional to the distinc- tiveness of the new adaptive type and extent and diversity of the new territory (Simpson, 1953). The diversification is related to the number of new adaptive subzones to which the evolving popula- tions have access. In the case of the Plethodontini and Bolitoglossini diversification has been, relative to other salamander groups, very great. Within the Bolitoglossini terrestriality has been achieved and movement into several adaptive sub- zones has been initiated. The tribe as a whole has a high base level of general adaptation, but most genera are very specially adapted to specific environments. Hydromantes is associated with a rather re- stricted habitat in limestone areas. The species occur on the surface under rocks and other debris, but individuals also inhabit caverns. Survival of this relatively ancient group has been related to association with relatively stable microenviron- ments. Distinct evolutionary trends in the direction of trunk and tail elongation and general skeletal re- duction are found in Batrachoseps. The most primitive species (B. wrighti ) has relatively low numbers of trunk vertebrae, moderately long tails, and relatively long limbs, but the advanced species have many trunk vertebrae, very long tails, and very small limbs. All have lost fifth toes and fifth distal tarsals. These specializations indicate that Batra- choseps has undergone an evolutionary modifica- tion of both locomotor morphology and behavior, and these changes are apparently associated with occupancy of a new environment. Storer (1925) stated that Batrachoseps uses the excavations made by earthworms and insects, and interpreted elonga- tion of the body, increase in the number of trunk vertebrae, reduction in size of limbs and feet, and tail elongation as specializations for subterranean life. From personal acquaintance with Batracho- 78 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 seps in the field it is apparent to me that the south- ern California species spend considerably more time in subterranean refuges than they do on the surface and that morphological specializations are directly correlated with behavioral and ecological specializations. The genus is in the process of in- vading a new semi-fossorial or fossorial adaptive subzone. Many plethodontids have genetically fixed numbers of vertebrae. Apparently variability in the numbers of trunk vertebrae and attenuate habitus have given Bcitrachoseps evolutionary ac- cess to the new adaptive subzone. The supergenus Bolitoglossa is unique in being the only tropical group of salamanders. Within the neotropics the group has undergone a considerable and significant radiation. The supergenus comprises seven tropical genera and over 100 species — almost two-thirds of the plethodontid species. Pletho- dontids probably arose in what is now the Nearctic region, and the primary dispersal center has been northern; but a second center of dispersal is today the southern and eastern borders of the Mexican Plateau, and it is inappropriate to speak of the family Plethodontidae as a Holarctic or Nearctic group. Most tropical plethodontids occur at high eleva- tions in relatively cool habitats, and only two genera ( Bolitoglossa , Oedipina) have successfully invaded the tropical lowlands ( Lineatriton and Parvimolge may also occur in tropical lowlands but both have very limited distributions). Neotropical species occupy a variety of terrestrial niches. Lineatriton and Oedipina are greatly elon- gated with reduced limbs and feet and extremely long tails. Not much information is available con- cerning the habits of the two, but apparently both are subterranean to some extent and utilize burrows of earthworms, mammals, etc. (Dunn, 1928). Other genera (Pseudoeurycea, Thorius) are more or less surface litter forms found under rocks and logs and under the loose bark of fallen trees. Some Chiropterotriton are terrestrial and occur in mossy areas on banks and cliffs. Species of the genera Bolitoglossa, Chiroptero- triton, Pseudoeurycea, and Parvimolge are moving into a new adaptive subzone — the arboreal sub- zone. Chiropterotriton is a genus of primarily high- land, climbing species. The group has an adaptation which has apparently given it evolutionary access to the new subzone; this feature is the articulation of all distal tarsals with the centrals (see above, mesopodial elements). The mesopodial pattern is apparently of significance in facilitating climbing by allowing greater digital spread. Primitive species of Bolitoglossa occur at mod- erate to high elevations from southern Mexico to northwestern Colombia, and have little to mod- erate hand and foot webbing. Most primitive species are ground-dwellers living under rocks and surface debris in primarily unforested highlands. Some extend to moderate elevations and may occur in bromeliads as well as on the ground (e.g., B. subpalmata). Throughout its range Bolitoglossa has invaded the tropical lowlands and expanded its peripheral range. Invasions of lowlands have in- variably been accompanied by increased webbing, flattening of the hands and feet, and increased arboreality. Fully webbed hands and feet provide more surface area and are of great significance in forms such as Bolitoglossa, which depend on sur- face tension to support their weight while walking and climbing on smooth, moist leaf surfaces. Genetic potential for variation in hand and foot morphology and subsequent molding by selection are the most important factors providing evolu- tionary access to the lowland arboreal subzone. The problem of ecological access is particularly acute in tropical lowlands. The “salamander niches” of higher latitudes and elevations are filled — primarily by frogs and possibly by snakes and lizards. Ecological access has been available, how- ever, in arboreal habitats, the route taken by Bolitoglossa, and in semi-fossorial habitats, the route followed by Oedipina. The final group, tribe Plethodontini, probably left the semiaquatic adaptive zone at a later date than the Bolitoglossini, and as a group is more generally and less specially adapted. Species of Aneides and Plethodon still occur in Appalachia, but the bolitoglossines all occupy ranges far from the ancestral center. Two of the three plethodonine genera, Plethodon and Ensatina, are primarily ground-dwellers and are most frequently en- countered under rocks, logs, and other surface lit- ter, or out wandering about on the forest floor. Most or all species of the two genera spend con- siderable amounts of time below the surface. Ensatina is a rather stout-bodied form which uses burrows of larger organisms such as rodents as a means of getting below the surface (Stebbins, 1954). Species of Plethodon are more elongate than Ensatina. Slender attenuate forms such as P. cinereus, P. richmondi, and P. elongatus have reduced limbs and feet and are able to enter bur- rows of relatively small size. Some of the larger, longer-legged species are capable of some scansorial locomotion; I have encountered P. jordani at dis- tances one to three feet above ground on tree trunks. Other large species such as P. yonahlossee inhabit burrows of relatively large diameter. One 1966 EVOLUTION OF LUNGLESS SALAMANDERS 79 species, P. dunni, has secondarily returned to a semiaquatic habitat in the Pacific Northwest and is commonly encountered in rocky, moss-covered seepage areas. The primary habitat of both Pletho- don and Ensatina, however, is at the surface of the ground away from running water. The genus Aneides is an excellent example of a group entering a new adaptive subzone, the arboreal way of life. The most primitive species, A. hardii, is a Plethodon-Yike form that lives in, under, and on fallen logs. Three other species (A. aeneus, A. ferreus, A. lugubris) are arboreal to a large degree and have numbers of specializations related to arboreal life (depressed body, long limbs and digits, bifurcated terminal phalanges, prehensile tail). A detailed analysis of evolutionary patterns and trends in Aneides has been prepared and will be published elsewhere. A final species, A. flavipunctatus, is a short- legged salamander with a cylindrical body. It is often encountered in very moist situations such as brooksides (Myers and Maslin, 1948). It is similar to Plethodon dunni in this regard and has a much greater affinity for wet habitats than other Aneides. Return to semiaquatic habitats (but with retention of the terrestrial “way of life”) has oc- curred only in the Pacific Coast species of Pletho- don and Aneides, since such ecological niches are occupied by Desmognathus, Eurycea, etc., in the East. Paedomorphosis Paedomorphic influence and plethodontid evolution Paedomorphosis is a term used to denominate several morphological modes or patterns of evolu- tion. Paedomorphosis involves characters that are present or make their appearance in the young stage of an ancestral group and which in the ontogeny of a descendent appear: (1) in the young and adult stage, producing a substitution of a new adult condition for the old, resulting in progressive deviation in the ontogeny of the descendent from that of the ancestor; or (2) in the adult, by a rela- tive retardation of the development of the bodily structures as compared with the reproductive or- gans (DeBeer, 1958). Paedomorphosis is and has been of extreme importance in the evolution of plethodontid salamanders, and the general principle involved may be restated as the mode by which larval and youthful characters in the ancestor can influence adult characters in the descendent (Gar- stang, 1922; De Beer, 1958). Conditions of two distinct types have resulted from paedomorphic influence in plethodontids. The first is paedogenesis in which sexual maturity is accelerated and all members of the species retain larval morphology (except reproductive system) throughout life. Paedogenesis is genetically fixed, leads to specialization and degeneration, and is of little significance as far as future phylogenetic progress is concerned. De Beer (1958) used the term neoteny to describe the permanent larval condition of salamanders, but he fails to distinguish between environmentally induced and genetically fixed conditions. The later (called permanent neoteny) perfectly fits his description of paedo- genesis. The other major condition which results from paedomorphosis is a phenomenon called ar- rested development or differential metamorphosis. Differential metamorphosis is a term that describes the paedomorphic pattern in which metamorphic processes are extended over a considerable period of time, with some elements (bones in the present case) completing metamorphosis early, others very late, and some not at all. This condition is more characteristic of terrestrial than aquatic groups of plethodontids and has been of great importance in providing an escape from specialization in the ancestral adaptive zone and access to new adaptive zones. Paedogenesis occurs only in the following species, all members of the tribe Hemidactyliini: Gyrinophilus palleucus (three distinct, allopatric races), Eurycea latitans, E. nana, E. neotenes, E. pterophila, E. troglodytes, E. tynerensis, Typhlo- molge rathbuni, T. tridentifera, and Haideotriton wallacei. Based on the morphological evidence pre- sented above, paedogenesis has probably evolved in parallel at least four, and possibly more times. All paedogenetic species are highly specialized, and their distribution is somewhat to greatly peripheral to the main tribal range. Paedogenesis may have arisen initially in response to selection pressures operative in marginal habitats during arid periods. Most paedogenetic species occupy subterranean waters exclusively or to a considerable degree. Cer- tainly aquatic larvae of ancestral populations had physical access to such environments, and the very fact that ancestral populations had aquatic larvae provided evolutionary access to the new aquatic or subterranean adaptive subzone. Subterranean waters are notably free of competitors, and the invaders also had ecological access. Differential metamorphosis has not been a major feature of the evolution of the desmognathines. Perhaps shortening of the larval periods and speed- ing of the metamorphic processes in advanced Desmognathus are related to paedomorphosis. Small adult size and resemblance in osteological 80 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 structure to young stages of primitive species characterize D. aeneus and D. wrighti. Examples are retention of anterior vomerine teeth in adults, relatively slight sexual dimorphism, and relatively thin, light cranial elements. These features are a result of paedomorphic influence, and sexual ma- turity is achieved at a morphological stage roughly equivalent to a juvenile stage of such species as D. quadramaculatus. Both species have distinct ter- restrial tendencies. Apart from the paedogenetic effect, paedo- morphosis has had some influence in the Hemidac- tyliini. Hemidactylium and Eurycea quadridigitata have only four toes. The larvae of both differ from those of other plethodontids in that the hind limbs develop slowly and are represented in hatchlings by undifferentiated lobes. Failure of the fifth toes to develop may be the result of differential initial growth rates and delayed limb development. Paedo- morphosis may also be responsible for the relatively weak cranial elements and tenuous cranial articula- tions of such species as E. quadridigitata and E. muitiplicata, diminutive size of E. quadridigitata, retention of a lateral line system on the head of adult Stereochilus, relatively small eyes of Stereo- chilus, and relatively poorly developed eyes, eye- lids, and nasolacrimal glands of Typhlotriton. Bolitoglossines have been influenced by paedo- morphosis to a greater degree than any other ter- restrial plethodontid group. Hydromantes has retained the primitive metamorphic pattern in re- gards to premaxillary structure, but paedomorpho- sis may be responsible for the failure of prefrontals to develop, the relatively weak skull articulations, and for the high numbers of unicipital ribs in the European species. Batrachoseps has been profoundly influenced by paedomorphosis, and differential metamorphosis is apparent in the modern species. Premaxillae are fused in the young of all species and in the adults of all except B. wrighti. Premaxillary metamorphosis is thus delayed relatively long in wrighti, and is never achieved in the other species. A further ex- ample of differential metamorphosis is the appear- ance in very large wrighti of tiny prefrontal bones, which are absent in other species. Other paedomor- phic effects evident in members of the genus include weak development of frontals and parietals, in- complete cranial roof formation, incomplete devel- opment of vomers with no preorbital processes, absence of second basibranchials, weak limb and foot development, and absence of fifth toes. Pre- frontals and vomerine preorbital processes are among the last elements to develop during primitive plethodontid metamorphosis. Neotropical salamanders (supergenus Bolitog- lossa ) have been affected by paedomorphosis more than any other plethodontid group with the excep- tion of the paedogenetic species. Examples of paedomorphic influence include fusion of premaxil- lae and absence of second basibranchials in all genera. Examples found in individual genera follow. Bolitoglossa: failure of prefrontals to develop, or irregular prefrontal development in a number of advanced species belonging to different species groups (rufescens, occidentalis; savagei, adspersa, orestes; cerroensis, robusta; colonnea); failure of septomaxillae to develop in most species; partially to fully webbed feet (a condition encountered in young or embryonic stages of other genera) . Oedipina: failure of prefrontals and septomaxil- lae to develop; weak development of limbs and feet, with reduced phalanges and variable numbers of mesopodial fusions; high percentage of unicipital ribs; extremely small premaxillae; weak skull articulations. Pseudoeurycea: failure of septomaxillae to ap- pear in all but a few species, which have irregular degrees of development. Lineatriton: failure of septomaxillae to become well developed; diminutive size; reduced limbs and feet. Thorius: variously developed septomaxillae; all cranial elements greatly reduced in size; diminutive size; reduced limbs and feet; enlarged nostrils of some species. Two genera, Chiropterotriton and Parvimolge, deserve special comment. Primitive Chiropterotri- ton are but slightly affected by paedomorphosis, but an evolutionary gradient related to paedomor- phosis exists in the genus on a roughly north to south axis (see also Rabb, 1960). The northern species usually have prefrontals, septomaxillae, tibial spurs, and vomerine preorbital processes, but the elements are absent in some of the southern species. In addition southern species demonstrate features considered by Rabb (1960) and me to be strengthening compensations in certain elements to paedomorphic weakening in others. These compen- sations include calcification of normally cartilagi- nous structures such as mesopodial elements, heads of limb bones, and intervertebral cartilages, fusion of premaxillary frontal processes, and mesopodial fusions. These compensations are also encountered in varying degrees in Batrachoseps, Lineatriton, Parvimolge, Oedipina, and Thorius. Table 2 is a representation of the distribution of paedomorphic characters and compensations to paedomorphosis in some species of Chiropterotriton. It is apparent that the southern, more advanced species have more 1966 EVOLUTION OF LUNGLESS SALAMANDERS 81 paedomorphic characters than the more primitive northern species. The first four species (priscus, chiropterus, multidentatus, dimidiatus) occur north of the Isthmus of Tehuantepec; the remainder occur south of the Isthmus. C. dimidiatus and C. multi- dentatus occur sympatrically at some localities, otherwise the chart is arranged on a geographical north-south axis. Some of the data in the table are at variance with information presented by Rabb (1960) , but the important point is that evolutionary specialization and advance in Chiropterotriton is associated with a geographic trend and with trends in the direction of increased paedomorphosis. Parvimolge townsendi of Veracruz, Mexico, is strongly affected by paedomorphosis (diminutive size, reduced limbs and feet, mesopodial fusions, calcified mesopodials, calcified intervertebral carti- lages, reduced septomaxillae and columellae). It resembles P. richardi of Costa Rica in some char- acters (see above) but differs in others. The two species probably represent different evolutionary lines, one of which ( townsendi ) may have arisen from a Pseudoeurycea-Chiropterotriton ancestor, the other from a Chiropterotriton stock. If placed in Table 2, richardi would have three paedomor- phically related characters (no septomaxillae, no columellae, calcified intervertebral cartilages), somewhat fewer than other southern Chiropterotri- ton. It is also possible that evolution is following similar patterns in Chiropterotriton and Parvi- molge, and richardi may be an advanced member of the latter. Further speculation must await in- formation concerning P. praecellens and the meso- podial structure of richardi. As a group the tribe Plethodontini has been little affected by paedomorphosis. The evolutionary significance of paedomorphosis The major role played by paedomorphosis in plethodontid evolution is demonstrated by different adaptive trends that relate to the phylogeny of two distinct groups. Paedogenesis permits survival of several species of hemidactyliines in specialized and restricted habitats. These habitats are found primarily in regions that are not favorable for the survival of related, non-paedogenetic species. Ex- treme specialization within such unusual habitats and consequential loss of evolutionary plasticity (potentiality for future evolutionary progress) characterize the paedogenetic forms. Differential metamorphosis, a pattern most apparent in the bolitoglossine genera, is associated with occupancy of and survival in terrestrial habitats. The bolito- glossine genera are all highly specialized, but they have avoided overspecialization through paedomor- Table 2. Paedomorphic characters and compensations in Chiropterotriton. Characters Species Septomaxillae Fail to Develop Prefrontals Fail to Develop Vomerine Preorbital Processes Fail to Develop Tibial Spurs Fail to Develop Columellae Fail to Develop Premaxillary Frontal Processes Fuse Enlarged Nostrils Mesopodial Fusions 1 Mesopodial Calcifications Intervertebral Cartilage Calcification Totals X priscus 0 h chiropterus 0 / multidentatus 0 dimidiatus X X 2 xolocalcae X X X X 4 hromeliacia X X X X X 5 X h nasalis X X X X X X X 7 X o abscondens X X X X X X X 7 82 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 phosis. This appears to be associated with and, in fact, may be responsible for a great increase in evolutionary plasticity. Paedogenetic plethodontids are degenerate spe- cies living in habitats that are secluded, difficult, or unfavorable from the standpoint of the major group, viz., the primitive hemidactyliine genera. These habitats are invariably on the margins of the present generic or tribal ranges. Such a distribution suggests that the species are descended from rem- nants of species which inhabited the regions when conditions were more favorable for their survival than at present. They probably entered their pres- ent habitats in response to strong eliminative pres- sures in the environments of the adults of ancestral populations. Such elimination may have been ran- dom and nonselective with the organisms virtually defenseless against such severe climatic factors as drought or extremes of heat or cold. Permanent occupancy of the larval habitat or of physically adjacent and accessible habitats, such as subter- ranean water systems, was necessary for survival. Under conditions of nonselective elimination there is selection for survival of progeny through in- creased fertility, and, particularly in salamanders, acceleration of the life cycle with earlier repro- ductive maturity. Accordingly, individuals do not complete ontogeny earlier than individuals of re- lated species, but certain systems ( e.g . reproduc- tive) mature while others do not. Most systems are characterized by larval morphology. Selection probably favored neotenic individuals over those that transformed, with eventual genetic fixation of neoteny, i.e., paedogenesis. The pre- mature maturation of reproductive organs linked with degeneration or underdevelopment, or both, of other organs is typical of this type of selection and characterizes the paedogenetic plethodontids. Results of such selection in other organisms have been summarized by Schmalhausen (1949). The second kind of paedomorphosis (differential metamorphosis) is associated with the evolutionary activity of the tribe Bolitoglossini. Terrestrial ple- thodontids lack aquatic larvae. Their metamorphic changes tend to be spread over a greater period of time and tend to be less extreme than in those species with larval stages that undergo complete transformation in a relatively short period of time. In terrestrial plethodonines most metamorphic changes occur before hatching. It is by differential metamorphosis’ in the bolitoglossincs that some of these changes (premaxillary separation, develop- ment of prefrontals, etc.) occur relatively late in life, or not at all. Differential metamorphosis has activated the variational potential of bolitoglossines and has led to the great differentiation of the group. Differential metamorphosis is most obvious in peripheral and marginal habitats where either selec- tion is very strong or entrance into a foreign en- vironment was difficult for the ancestral stock. The genus Batrachoseps provides an example of the significance of these paedomorphic effects. B. wrighti, an inhabitant of environments similar to those occupied by primitive plethodontids, is in- fluenced but slightly by paedomorphosis. The other species, all with peripheral or marginal distribu- tions, are strongly affected by paedomorphosis. Selection in the latter species probably favored populations or individuals in which reproductive maturity occurs at an early developmental stage. A similar example has been discussed earlier in re- gard to the southward spread of the genus Chirop- terotriton in the Neotropics. Metamorphosis and ontogenetic anatomical changes require energy expenditure beyond main- tenance levels. Thus, an additional factor in the evolution of paedomorphosis is the selection in marginal habitats of individuals that could repro- duce at relatively early overall developmental stages. Paedomorphic individuals expend relatively less energy in order to contribute to the gene pools than non-paedomorphs. Therefore the paedo- morphs could reproduce at higher rates than non- paedomorphs, and the populations would evolve in the direction of the former. Biotic selective pres- sures are strong in the bolitoglossines, particularly in such densely populated habitats as the tropical lowlands. These pressures have contributed to a substantial increase in the significance of paedo- morphosis in bolitoglossines, relative to more primitive plethodontids. Evolutionary stagnation has resulted from pae- domorphosis in the paedogenetic hemidactyliines. In contrast, paedomorphosis in the bolitoglossines has increased evolutionary plasticity and has ex- erted a profound directing influence in their phylo- geny. Bolitoglossines may have been the first plethodontid group to abandon the aquatic larval stage and to become truly terrestrial. Ancestors of this group were probably the first terrestrial pletho- dontids to extend their ranges out of Appalachia. Bolitoglossines must have begun specializing very early, so that they were no longer capable of ef- fectively competing with the relatively generally adapted and vigorous plethodonines when the latter invaded terrestrial habitats. Elimination of bolitoglossines in Appalachia resulted. Bolitoglos- sines in the West were forced to continue trends toward increased specialization as a means of avoid- ing competition. Only a fragment of a presumed 1966 EVOLUTION OF LUNGLESS SALAMANDERS 83 early radiation survived, and these were the an- cestors of the modern supergenera. Only one group, the tropical salamanders of the supergenus Bolito- glossa, has been successful, and all of its species are moderately to profoundly affected by paedomor- phosis. The tropical environment is foreign to sala- manders, and the association of paedomorphosis with the tropical salamander radiation suggests that it provided a means of evolutionary escape from overspecialization for ancestral stocks. The highly adaptive tropical radiation has resulted in the evo- lution of several genera and over 100 species of salamanders. Paedomorphosis seems to have been a means of releasing the evolutionary potential of the group. At the present time plethodontid evolutionary activity seems to be most intense in the wet tropical lowlands, a densely populated environment. In- vasion and radiation in such an environment by salamanders, which, except for this single group, are otherwise restricted to northern temperate en- vironments is highly significant. The vigor with which the radiation has proceeded and the high degree of species differentiation attest to the evolu- tionary plasticity of the group. The fact that all of the more advanced species of the various genera are increasingly affected by paedomorphosis dem- onstrates its dynamic, progressive influence in the group. The active role of paedomorphosis has been the dominant feature in the evolution and adaptive radiation of the tropical salamanders. Although little factual supportive evidence has been provided, paedomorphosis has been consid- ered by Garstang, DeBeer, Rensch, Schmalhausen, and other biologists to be an important evolutionary process. Few attempts have been made to reinter- pret the early work on paedomorphosis in the light of modern genetic and evolutionary theory. As a result paedomorphosis is often assigned only a minor role in theories of the evolution of higher levels of organization, and receives only passing mention in major treatises on evolution (e.g. Simp- son, 1953; Mayr, 1962). Recently additional facts have been made available. Manwell (1963) cites the almost identical nature of biochemical and physiological properties of the hemoglobins of adult brook lampreys (Entosphenus lamottei) and am- mocoete larvae of other species as evidence of paedomorphosis in the former. Several kinds of morphological evidence of paedomorphosis are described in the present paper. These and other data indicate that a reinvestigation of the general biological significance of paedomorphosis and other ontogenetic phenomena (DeBeer, 1958; Rensch, 1959) might have fruitful results. Paedomorphosis may result from a relatively simple genetic mechanism. There is some sugges- tion that control or regulatory genes, in some ways analogous to the operator genes of microorganisms, exist in vertebrates, and these may influence the activity of numerous structural genes or gene com- plexes (Ingram, 1961; Manwell, 1963). Cooper (1965) has demonstrated that cartilage inducing activity in embryonic stages of vertebrates is a stage-specific phenomenon, with no single “in- ducer” having a dominant role. Presumably there is an intricate pattern of gene activation and inac- tivation. Regulator genes, or their equivalent, may play very important integrating roles in these de- velopmental patterns. It is possible, however, that polygenic systems, operative at early stages of development, rather than single genes, are involved (see Waddington, 1956). Modification of these mechanisms could have very widespread effects, leading to differential metamorphosis and other evolutionary phenomena. Selection for favorable combinations of such genes or gene complexes could result in the fixation of new morphological arrangements. Favorable new arrangements, once introduced in populations, might play key roles in initiating major new evolutionary trends, especially if they provide access to a new habitat or a new adaptive zone. Genetic fixation of paedomorphic characters is closely associated with the evolution of compensa- tory adaptations. Selection for developmental ar- rythmia or partial developmental failure would not be successful without concomitant compensatory changes in the genetic background of the popula- tion as a result of counter selection. Examples are evident in some neotropical salamanders. In mem- bers of the paedomorphic genus, Thorius, diminu- tive size and weak bony and muscular development are compensated for and strengthened by calcifica- tions of normally uncalcified cartilaginous struc- tures such as the ends of long bones, the mesopodial elements, and the intervertebral cartilages. Accumulating evidence indicates that paedomor- phosis is of extreme importance in providing new opportunities to populations of organisms. Paedo- morphosis might initiate and even direct major evolutionary and phylogenetic trends, and provide a mechanism of rejuvenation for overspecialized or slowly evolving groups. Evolutionary break throughs which result may be involved in the adop- tion of new ways of life and may form the basis for subsequent adaptive radiations. Paedomorpho- sis clearly is a major feature of plethodontid evolu- tion and phylogeny, and investigations of the 84 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 significance of the phenomenon in other groups of organisms are suggested. BIOGEOGRAPHY AND PHYLOGENY The Fossil Record The only undoubted plethodontid fossils are Pleistocene remains of Plethodon glutinosus of essentially modern structure from the Wisconsin and Illinoian ages of Florida (Holman, 1958; 1959 a; 1959 b). Peabody (1959) reported track- ways of several salamanders, including Batra- choseps, from the Lower Pliocene of Tuolomne County in the Sierra Nevada of California. The Batrachoseps trackway was referred to B. pacificus on the basis of large size and body and limb dimen- sions, but it fits equally well an undescribed species which occurs near the fossil locality. These are the only New World Cenozoic plethodontid records. Herre and Lunau (1950) described a new genus and species of fossil salamander, Dehmiella schind- wolfi, from the Lower Miocene (Burdigalium) of southern Germany, and Herre (1950) described a new genus and species, Geyeriella mertensi, from the Paleocene of Germany. Both are known only from vertebrae and are said to resemble pletho- dontids because of the presence of well developed basapophyses. Such processes are well developed in desmognathines and some plethodontines, but not in salamandrids. Ambystomatids and hynobiids have either differently shaped processes at the front ends of the centra, or none at all. Dehmiella was considered to be a plethodontid, but Geyeriella was thought to be a forerunner of the plethodontids. Auffenberg (1961) suggested that Geyeriella may represent a side line of the main evolutionary sequence leading to the Plethodontidae and that Dehmiella represents the first “true plethodontid.” He assumed a close salamandrid-plethodontid re- lationship, and this assumption obviously influenced his judgment. Estes (1964) stated that Dehmiella is probably a plethodontid, but that the status of Geyeriella is uncertain. I have not examined the material upon which the genera Geyeriella and Dehmiella are based, but it is doubtful that either is a plethodontid. The character upon which the genera are based (basa- pophyses) is an adaptive feature that has arisen a number of times within the Plethodontidae. It is apparent from the descriptions and illustrations that both fossil genera differed in a number of structural features from modern plethodontids. Geyeriella differed in the extreme posterior place- ment and great length of the transverse processes, and in the relative neural arch length. It apparently had opisthocoelous vertebrae. Dehmiella had stout neural crests that were considerably higher than the posterior neural arch margins, a condition not encountered in plethodontids. Transverse processes were much straighter than in most plethodontids and were almost perpendicular to the central axis. The vertebrae of Dehmiella were apparently some- what broader than those of any modern pletho- dontid, but were amphicoelous. Certainly the ver- tebrae of the two genera bear little resemblance to those of Hydromantes, the only modern European plethodontid, or to other plethodontids. Experience has shown that parallelism is common in salaman- der evolution, and basapophyses, found only in a few plethodontids, are not sufficient reason for as- signment of fossil forms to the family Plethodonti- dae. Certain Mesozoic species may be plethodontids or close relatives. Auffenberg (1961) described a new genus and species, Opisthotriton kayi, from vertebrae collected from the Lance Formation, Upper Cretaceous, of eastern Wyoming. Opistho- triton was tentatively assigned to the Salamandridae on the basis of very long maxillae and opisthocoel- ous vertebrae. Hypapophysial keels and basapophy- ses led Auffenberg to consider the genus to be close to plethodontids. Estes (1964) examined much more material of Opisthotriton than was available to Auffenberg. He showed that material (a maxilla) on which pro- posed salamandrid relationship was based is refer- able to Scapherpeton, a fossil cryptobranchid. Estes considered the vertebrae of Opisthotriton to be definitely opisthocoelous. On the basis of vertebral characters, including presence of pterygapophyses, Opisthotriton is questionably referred to the sub- family Desmognathinae. Opisthotriton differed from all modern pletho- dontids in lacking internasal fontanelles, and had hour glass-shaped palatopterygoids unlike those of any plethodontid. It further differed from desmog- nathines in having two premaxillae and open meck- elian grooves, but both are primitive characters that might be expected in an ancestral group. Transverse processes were considerably longer than those of any modern plethodontid, but otherwise the figures of the vertebrae are similar to those of modern desmognathines. The maxillae are strikingly differ- ent from those of all plethodontids. Estes thinks Opisthotriton was a larval or semilarval form. Estes (1964) described a new salamander genus and species, Prodesmodon copei, from the Upper Cretaceous of Eastern Wyoming. The genus is also known from the early Cretaceous of Texas (Hecht, 1963). The species was considered by Estes to be a 1966 EVOLUTION OF LUNGLESS SALAMANDERS 85 desmognathine on the basis of disarticulated bony elements. Important characters of the genus in- clude: opisthocoelous vertebrae with strongly pro- truding condyles, strongly developed hypapophy- ses, pterygapophyses, and basapophyses; large, concave, atlantal condylar articulating facets; small odontoid processes; alar processes; paired premaxil- lae; no internasal fontanelles; closed meckelian grooves; sculptured anterior cranial elements; sinu- ous tooth bearing dentary margins; small maxillary facial lobes; interlocking mandibular symphyses; non-pedicellated teeth; no ribs. This is a mosaic of primitive and advanced characters unlike any pat- tern found in modern plethodontids. Lack of ribs and internasal fontanelles, and presence of non- pedicellated teeth and interlocking mandibular symphyses distinguish Prodesmodon from all ple- thodontid genera. Most of the characters suggest a relationship to desmognathines, but modern des- mognathines have a single premaxilla and large maxillary facial lobes. Paired premaxillae are to be expected in primitive desmognathines, however, and on the basis of the information presented by Estes, Prodesmodon apparently is a specialized (no ribs, vertebral specializations) offshoot of a primi- tive desmognathine stock. Research on fossil salamanders is actively being pursued by Estes, and further comments are de- ferred until results are made available. Historical Backgrounds and the Primary Dichotomy Origin of plethodontid salamanders in what is now called the Appalachian Highland physiogra- phic province (Atwood, 1940), or Appalachia, was first discussed by Wilder and Dunn (1920), and the proposal was expanded and detailed by Dunn (1926). Since that time the theory has gained wide acceptance. Three of the four major groups of plethodontids occur in Appalachia, including all of the primitive genera. Wilder and Dunn (1920) and Dunn (1926) developed the thesis that mountain brooks were the ancestral plethodontid habitats; mountain brook plethodontids occur only in Ap- palachia. Evolution of the family requires an area that has been relatively stable geologically and cli- matically for a very long time, at least since mid- Mesozoic; Appalachia is such a region. Thornbury (1954) states that mountain building affected Appalachia as early as the Ordovician and contin- ued intermittently throughout the Paleozoic. The Appalachian revolution at the close of the Paleozoic was apparently the culmination of mountain build- ing trends. Erosion, sedimentation, and peneplana- tion have since modified the landscape, but Ap- palachia has provided a relatively stable environ- ment for the development of the biota since the end of Permian (Graham, 1964). It was stated in rather unequivocal terms by Dunn (1926) that plethodontids have been derived from a salamandrid stock in the general present- day Appalachian region during late Eocene to Mio- cene. Evidence has been presented above to show that the plethodontid-salamandrid relationship is not close, and that it is highly unlikely that the two groups have had a common ancestor. The report of highly specialized desmognathine salamanders in early Cretaceous (Estes, 1964, 1965; Hecht, 1963) suggests that family characteristics of the Plethodontidae were established during Mesozoic. The probability that the specialized Prodesmodon occupied swampy environments beyond the west- ern periphery of Appalachia in Cretaceous indi- cates that plethodontids had then been in existence for some time, and suggests that if plethodontids did arise in what is present-day Appalachia, they may have originated in mid-Mesozoic or earlier. Based on morphological similarities, plethodon- tids may have arisen from an ambystomatid or pre- ambystomatid ancestral stock, which in turn was probably the New World derivative of the ancestors of the Old World Hynobiidae. The family Ambysto- matidae is an endemic North American family which is more primitive than the family Pletho- dontidae. Modern ambystomatids apparently have less evolutionary plasticity than plethodontids. but the family has undergone a moderate diversifica- tion. Significantly, one genus, Rhyacotriton, occu- pies mountain brook habitats in northwestern North America, and it is virtually lungless. It is postulated that a similarly adapted population, but one with great genetic and evolutionary plasticity and a higher base level of general adaptation, suc- cessfully occupied the mountain brooks of Appala- chia during Mesozoic times and gave rise to the plethodontid radiation. Because the fossil record is scanty and incom- plete it is impossible to date the primary dichotomy of the Plethodontidae. Divergence of the desmog- nathine line from the more generalized stock that gave rise to the present plethodontines must have taken place in Mesozoic. Certain characters ( e.g ., separated premaxillae) of Prodesmodon indicate that it was relatively close to the ancestral stock. The closed meckelian grooves, opisthocoelous ver- tebrae, and shape of cranial elements indicate that the primary plethodontid dichotomy had already occurred, and that the Cretaceous Prodesmodon 86 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 was a specialized desmognathine offshoot which had left the ancestral habitat. Tertiary events are responsible for the sub- familial evolutionary patterns seen in the pletho- dontids. Because the fossil record is scanty, paleo- botanical and paleoclimatic data must be used to infer the past history of the Plethodontidae. The evolution of flowering plants has been re- viewed by Axelrod (1960) who states that close relationship between modern plants and fossils of the later Oligocene makes it possible to reconstruct the habitats and climates under which Tertiary floras lived, to outline the belts of Tertiary vegeta- tion, and to explain the evolution of modern pat- terns of plant distributions. Three broad, world- wide belts of distinctive vegetation were present in early Tertiary: the Arcto-Tertiary Geoflora in the North, the Antarcto-Tertiary Geoflora in the South, and the very broad Tropical-Tertiary Geoflora in between. Many modifications occurred during Ter- tiary, but the modern floristic patterns can be de- rived from these three belts. Plants are the domi- nant members of biotic communities, which in turn are evolutionarily conservative units. Information about modern plant-animal associations allows one to make hypotheses concerning past events. Tertiary events resulted in numerous reorganiza- tions of the early Tertiary pattern. The tropical vegetational belt gave rise in the New World to the Neotropical-Tertiary Geoflora which is ancestral to modern neotropical forest formations. A major geoflora, the Madro-Tertiary, developed in situ in southwestern North America in early Tertiary in response to increasing cooling and aridity which forced the old communities out. This geoflora was derived primarily from Neotropical-Tertiary floral elements. The Madro-Tertiary Geoflora expanded steadily in the increasingly arid Southwest from late Eocene throughout Tertiary, and has given rise to the modern semiarid woodland, chaparral, thorn forest, arid scrub, desert grassland, and desert vege- tation of southwestern North America. Of con- siderable interest to the present study are the modern forest derivatives of the Arcto-Tertiary Geoflora. The more important of these in North America are the Eastern Deciduous, Hemlock- Hardwood, Boreal, Rocky Mountain, Pacific, and Southeast Evergreen Forests. In addition disjunct elements of the Geoflora survive today interspersed with Neotropical-Tertiary derivatives in the high- lands of Mexico, Guatemala, and other Central American countries (Braun, 1950; Martin and Harrell, 1957; Axelrod, 1960). Plethodontid sala- manders are today primarily members of commu- nities dominated by Arcto-Tertiary Geofloral de- rivatives. Salamanders and Arcto-Tertiary deriva- tives are both relatively ancient groups. Their as- sociation is probably ancient and the communities may have evolved rather slowly throughout Terti- ary. Savage (1960) has suggested that correlation of geologic and geofloral history forms a firm basis for development of concepts concerning historical bigeography of animal groups; this approach is followed in the present work. Historical elements of the North American her- petofaunas have been discussed by Savage (1960). Of particular interest is the concept of an Old Northern Herpetofaunal Element first proposed by Dunn (1931) and further developed by Savage. The Old Northern Element as defined by Savage consists of descendents of temperate circumpolar groups found in the far northern continental areas in early Tertiary but forced southward as the tropical limits were restricted by long-term Tertiary climatic trends. Savage emphasized the historical association of the Old Northern Element with the Arcto-Tertiary Geoflora, and his analysis aptly fits the family Plethodontidae. Savage further subdi- vides the Old Northern Element into four historical zoogeographical complexes. These complexes and the plethodontid genera I consider to be members of each are: 1. Eastern American Complex. Desmognathus, Leurognathus, Gyrinophilus, Pseudotriton, Stereochilus, Eurycea, Typhlotriton, Typhlo- molge, Hemidactylium. 2. Western American Complex. Hydromantes, Batrachoseps, Ensatina. 3. Central American Complex. Bolitoglossa, Oedipina, Pseudoeurycea, Chiropterotriton, Lineatriton, Parvimolge, Thorius. 4. Southeastern American Complex. Phaeogna- thus, Haideotriton. Species groups of Aneides and Plethodon are mem- bers of both 1 and 2. In the following sections the probable history of these genera is reviewed. The Desmognathines Desmognathines have remained close to the plethodontid source region. Some species of Des- mognathus today occupy what might well be the primitive habitat. A moderate diversification has occurred in the subfamily, however, and some groups have left Appalachia (Fig. 19). Prodesmodon represents the earliest known di- vergence from the generalized desmognathine stock. It left the ancestral habitat and familial 1966 EVOLUTION OF LUNGLESS SALAMANDERS 87 source region, became morphologically specialized, and occupied a peripheral swampy habitat by early Cretaceous; divergence must have occurred at a relatively very early date. Phaeognathus may represent a somewhat later, but still very early divergence (Fig. 13). Ancestors of Phaeognathus left Appalachia, became terres- trial, and occupied the Coastal Plain at the peri- phery of Appalachia. The Coastal Plain slowly emerged from the sea starting in early Tertiary and has been in existence since Eocene (see also Gra- ham, 1964). Today Phaeognathus appears to be a Coastal Plain relict, a specialized survivor of a group that was probably more widespread during Tertiary than at present (see Valentine, 1963 a, for further speculations concerning the genus). Another divergent desmognathine line is repre- sented by Leurognathus, a southern Appalachian endemic. It has adapted to a specialized, strictly aquatic environment, and it is impossible to date its divergence. Adaptation to the new adaptive sub- zone has taken place very slowly and over a long period of time. The modern genus Leurognathus is probably the culmination of long Tertiary trends, and divergence from the ancestral stock has been more recent and less extreme than that of Phaeo- gnathus. Desmognathus has remained relatively close to the ancestral desmognathine stock, both morpho- logically and ecologically. The genus is centered in Appalachia, but has entered the Coastal Plain, the Interior Highlands, and adjacent areas. Movement into peripheral lowland areas by species such as D. auriculatus and D. fuscus has been accompanied by adaptation to slower moving waters than in the ancestral habitat, and invasion of the Interior High- lands has probably been via lowland routes. Several montane species (D. ochrophaeus, D. aeneus, D. wrighti) have become increasingly terrestrial. Major trends in Desmognathus are clearly in the direction of dwarfing and increased terrestriality. Desmo- gnathus is ecologically the most diverse plethodon- tid genus, but it is remarkable uniform osteologi- cally. Most of the osteological differences are re- lated to size and development stage. Evolution within the Desmognathinae and par- ticularly within Desmognathus has proceeded in situ in southern Appalachia. The course of evolu- tion within the subfamily has involved a successive occupation of distinct ecological niches in the same general environmental and geographic region (Dunn, 1917; 1926; Hairston, 1949; Organ, 1961 a) . Radiation within the subfamily is a miniature of that in the family as a whole. The Plethodontines Tribal Origins The plethodontines are more generalized mor- phologically than are the desmognathines and have remained closer to the ancestral plethodontid stock. The plethodontine stock gave rise relatively early to three major evolutionary lines (Fig. 12). The first probably diverged from the generalized stock at the beginning of Tertiary. The immediate factors in the divergence of this group, the fore- runners of the Bolitoglossini, were the loss of aqua- tic larvae, acquisition of terrestrial habits, and oc- cupation of terrestrial niches by generally adapted populations with genetic and evolutionary plas- ticity. Occupation of the new adaptive zone pro- vided impetus for the subsequent diversification and adaptive radiation of the Bolitoglossini. Several types of evidence support the theory of early origin of the bolitoglossines. The group today contains three distinct, specialized subgroups. It has differentiated to a greater degree than either of the other two tribes. The high degree of diversification suggests a long history, and morphological differ- ences suggest an ancient separation from the an- cestral stock. Bolitoglossines are distantly peri- pheral to the source and dispersal center of the family and are the only plethodontid group that no longer occurs either in Appalachia or its border- lands. It is the only group that has invaded Europe and the Tropics. Finally, a Batraehoseps trackway from the Mio-Pliocene border of California (Pea- body, 1959) indicates that the genus was of essen- tially modern construction at that early date. The above facts suggest a very early origin of the Boli- toglossini and an early spread with subsequent dis- placement by other plethodontids. The second evolutionary line, which has given rise to the Plethodontini, diverged from the gen- eralized plethodontid stock at a later date than the bolitoglossine ancestors. This divergence also in- volved entrance into a new adaptive zone with con- comitant loss of aquatic larvae and occupation of terrestrial niches by generally adapted populations. There are several reasons for thinking the pletho- donine divergence followed rather than preceded the bolitoglossine evolution. Members of tbe group left Appalachia and reached the Rocky Mountains and the Pacific Coast, but the group as a whole is not as peripherally distributed as tbe bolitoglos- sines, and two of the three genera are still repre- sented by Appalachian species. Modern pletho- donines are more primitive morphologically than the bolitoglossines, but this fact may be due to the more generalized nature of the plethodonines and ' SCALE 400 600 125 120 Desmoy Leurognathus Phaeognathus Ensatina Hydromantes Figure 19. Distribution of Desmogncithus, Leurognathus, Phaeognathus, and Ensatina. Figure 20. Distribution of Hydromantes in California. Figure 21. Distribution of Hydromantes in Europe. 1966 EVOLUTION OF LUNG LESS SALAMANDERS 89 their more recent association with the generalized plethodontine ancestral stock. The final group, the Hemidactyliini, has remain- ed primarily in Appalachia and its borderlands. It is closer morphologically to the ancestral stock than is any other plethodontid group. Evolution and differentiation has been centered in eastern United States. History of the Hemidactyliines The history of the hemidactyliine genera is closely tied to the Cenozoic history of the eastern United States. Only one genus, Gyrinophilus, is now limited in its distribution to the general Appa- lachian region, but two other genera, Eurycea and Pseudotriton, occur widely in Appalachia (Fig. 22, 24, 25). Appalachia is the center of dispersal of the tribe. Dispersal has been centrifugal in all directions from Appalachia. Gyrinophilus is morphologically the most primi- tive hemidactyliine and it has remained close to the ancestral home. Today the generic range cor- responds to the approximate limits of ancient Ap- palachia. Gyrinophilus is found in mountain brook habitats, but may occur in springs and woodland seepage areas. Paedogenetic populations that occur in caves in southeastern Tennessee and northern Alabama have been assigned to Gyrinophilus (G. palleucus) on the basis of their resemblances to larvae of Gyrinophilus porphyriticus and G. danielsi. The salamanders have been induced to metamorphose by Dent, Kirby-Smith, and Craig ( 1955) , Dent and Kirby-Smith (1963), and Blair (1961). These authors state that G. palleucus will undergo com- plete metamorphosis in the laboratory, and that metamorphosed individuals closely resemble other species of Gyrinophilus; but Blair (1961) reports that the premaxillae remain fused, a larval feature. Further, the vomers retain larval configuration. It is apparent from X-ray photographs presented by Dent and Kirby-Smith (1963) that fused pre- maxillae and larval vomers are also retained in their supposedly completely metamorphosed specimens. In most characters G. palleucus does resemble oth- er Gyrinophilus, but generic assignment remains tentative. It could also be a remnant of a primitive Gyrinophilus-Yike stock. The distribution of G. palleucus is that expected of a paedogenetic species — on the periphery of the range of its non-paedo- genetic congeners. Stereochilus is the derivative of an early hemi- dactyliine stock that invaded and successfully adapted to the swampy environments of the Atlan- tic Coastal Plain. Numerous primitive features plus several specializations ( e.g skull compression, habitat) suggest a relatively early divergence from the tribal stock. Stereochilus divergence and dif- ferentiation probably dates from the establishment of the Coastal Plain in early Tertiary. Atwood (1940) divides the Appalachian High- lands Physiographic Province into two parts: the northeastern division (New England-Acadian) and the southwestern division (Hudson Valley to cen- tral Alabama). Gyrinophilus is found throughout both divisions, but Pseudotriton is associated only with the latter. Pseudotriton has a more extensive, but more southerly range than Gyrinophilus (Fig. 22 and 24). Although Pseudotriton occurs in springs and seeps at high elevations and may be sympatric with Gyrinophilus, it can survive in the warm waters of low elevations. Pseudotriton has successfully invaded the Coastal Plain from Louisi- ana to the mouth of the Hudson River, and P. mon- tanus, in particular, has adapted to very muddy, mucky areas. Thus in its ecology and distribution, as well as in its morphology, Pseudotriton demon- strates evolutionary advance over Gyrinophilus. Eurycea ranges more widely than any of the above genera and has successfully invaded the Coastal Plain, the Interior Highlands, and the Balcones Escarpment-Edwards Plateau areas of Texas in addition to occupying the entire Appa- lachian Highland region (Fig. 25). Presumably the genus arose relatively early in Tertiary. Eurycea has undergone a moderate evolutionary diversifica- tion. The three widest ranging species, E. longi- cauda, E. lucifuga, and E. bislineata, are all found in the Appalachian Highlands, and E. aquatica is found on the southern borders of Appalachia in northern Alabama. Both longicauda and bislineata also occupy favorable microhabitats (seepages and swamps near springs and rivers) on the Coastal Plain but retain primitive stream larvae. E. quad- ridigitata has adapted (pond larvae) to the swamps of the Coastal Plain. The ranges of E. lucifuga and E. longicauda ex- tend from Appalachia through a narrow corridor across northwestern Kentucky and southern Illinois and Indiana to the Interior Highlands Physiograph- ic Province (Atwood, 1940) in Missouri, Arkansas, eastern Oklahoma, and extreme southeastern Kan- sas. Two distinct species of Eurycea (tynerensis and multiplicata) and the genus Typhlotriton, a Eurycea relative, are confined to the Interior High- lands. Other plethodontines found in the Interior Highlands include the endemic species, Plethodon ouachitae and P. caddoensis, and P. glutinosus ranges southwest from Appalachia in a pattern sim- Hemidactyli um Typhlomolge Typhlotriton Pseudotriton 1000 1200 1400 KILOMETERS Gyrinophi lus Stereochilus Haideotriton Eurycea Figure 22. Distribution of Pseudotriton. Figure 23. Distribution of Hemidcictylium, Typhlomolge, and Typhlotriton. Figure 24. Distribution of Gyrinophilus, Stereochilus, and Haideotriton. Figure 25. Distribution of Eurycea. 1966 EVOLUTION OF LUNGLESS SALAMANDERS 91 ilar to that of E. longicauda and E. lucifuga. Dis- junct western and southwestern populations of P. ciner eus, P. dorsalis, and Hemidactylium scutatum also occur in the Interior Highlands. Five species of paedogenetic Eurycea (latitans, nana, neotenes, pterophila, troglodytes ) and the paedogenetic genus Typhlomolge, derived from a pr e-Eurycea stock, are endemic to the Balcones Escarpment-Edwards Plateau region of central Texas. Disjunct populations of P. glutinosus also occur in the region. For biogeographic purposes % the entire plethodontine fauna of the Interior Highlands and of Texas is discussed as a unit, al- though both hemidactyliine and plethodontine spe- cies are involved. The problem of disjunct populations of Appa- lachian organisms in the Interior Highlands and on the Edwards Plateau has been discussed by num- erous authors. The works of Braun (1950), Smith (1957), Dowling (1958), Blair (1958), and Con- ant (1960) are pertinent to this discussion. Dowling discussed the geology and paleoclimatology of the Interior Highlands, and stated that the region has been above sea level since Pennsylvanian times. Dating the time of isolation of the Interior High- lands is controversial and probably impossible with the data at hand, but most authors agree that it was a Cenozoic event; Atwood (1940) suggests mid-Tertiary as the time of isolation from eastern highlands. The modern Interior Highlands Physiographic Province, as described by Atwood (1940), includes the St. Francois Mountains and Salem Plateau, the Ozark Plateau, and the Boston Mountains, all north of the Arkansas River Valley, and the Ouachita Mountains to the south of the valley. The province is bounded on the north and west by the Prairie Province, on the east by the Mississippi Lowland, and on the south by the Coastal Plain. Braun (1950) has discussed the significance of disjuncts of Appalachian forest plants which occur in the Interior Highlands and canyons in the Ed- wards Plateau. She interprets these disjuncts as relicts of a former more widespread mesophytic forest (the East American Element of the Arcto- Tertiary Geoflora; Axelrod, 1960). Today the In- terior Highlands are drier than Appalachia and Oak-Hickory communities dominate. Braun sug- gests that a mixed mesophytic forest was continu- ous across eastern North America and into the Mexican Highlands in Early Tertiary. Interior regions were drying in middle Tertiary and the mixed forest was forced to retreat northward and eastward. At this time the Oak-Hickory Associa- tion evolved in the Interior Highlands, and drier western and southwestern plants (derivatives of the Madro-Tertiary Geoflora; see below) invaded the Edwards Plateau. Mesophytic species of the mixed forest survive as relicts in only a few spots west of the Mississippi River. Humid corridors were re- established during Pleistocene pluvials below the glacial borders. The primary corridor between the Appalachians and the Interior Highlands was through the Hill Section of the Western Mesophytic Forest Region of Braun (1950), which extends from the Knob region of western Kentucky through the Norman and Crawford uplands of southern In- diana, and the Ozark Hills of southern Illinois. This corridor has been especially important for plethodontine distribution. Plethodontines in the Interior Highlands and the Edwards Plateau represent three levels of divergence from the ancient Appalachian popula- tions. The earliest level includes the two endemic genera, Typhlotriton of the Interior Highlands and Typhlomolge of the Edwards Plateau borders. Ancestors of these genera may have entered the regions in early Tertiary when favorable climatic conditions prevailed and relatively uniform forest communities extended from Appalachia into Mex- ico and to the West Coast. They have survived drier surroundings than are generally considered suitable for survival of plethodontids by occupying subterranean aquatic habitats. Paedomorphosis has been the key to the success and survival of both genera. Typhlomolge is paedogenetic and is restricted to underground water systems. Typhlo- triton, which is probably not closely related to Typhlomolge, occurs in surface creeks in the larval state, but the adults are restricted to caverns. The second level of differentiation is represented by the endemic species of Appalachian genera which occur in the two regions. These include Eurycea latitans, E. nana, E. neotenes, E. ptero- phila and E. troglodytes of the Edwards Plateau, and E. multiplicata, E. tynerensis, Plethodon ouach- itae, and P. caddoensis of the Interior Highlands. Six of the seven species of Eurycea are paedoge- netic, and the seventh (multiplicata) is sometimes neotenic (Conant, 1958). P. ouachitae and P. cad- doensis are much better adapted to dry conditions than P. glutinosus (Pope and Pope, 1951). Dowling (1958) has suggested that isolation and differentia- tion of the Interior Highland endemics dates from Miocene to Pliocene on the generic level and the late Pliocene to early Pleistocene on the specific level. Blair (1958) attempted to explain the distri- bution of the relict species of Plethodon and Eury- cea on the basis of Pleistocene events. These times seem much too recent. Typhlomolge and Typhlotri- 92 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 ton probably were differentiated when Tertiary drying was beginning to drive plethodontids east- ward. They survived in specialized microhabitats to which they became increasingly more highly adapted. The species of Eurycea are probably de- scendents of populations “trapped” in isolated favorable microhabitats as increasing Tertiary aridity forced the genus as a whole eastward. Paedomorphosis has been a universal means of survival for these endemics. They have survived by remaining for all (the paedogenetic species) or much (multiplicata) of their lives in waters that are surrounded by terrain largely or wholly unsuited for urodele terrestrial or semiterrestrial existence. In these niches they survived the late Pliocene and Pleistocene droughts. Plethodon caddoensis and P. ouachitae have been considered to be members of the yonahlossee species group, the most primitive of the nine groups of Plethodon recognized by Highton (1962). The other members of the group, P. longicrus and P. yonahlossee, have limited ranges in the southern Appalachians. Highton suggests that the prototype of the group was once widely distributed in the eastern United States. The primitive nature of P. caddoensis and P. ouachitae, and the fact that they are adapted to live in relatively dry environments (Pope and Pope, 1951; Dowling, 1958) suggest a rather early divergence and an in situ adaptation to the increasing dryness of the Interior Highlands from mid-Tertiary to Recent. The final level of differentiation is represented by populations of primarily Appalachian species that are subspecifically distinct from the main popula- tions to the east. These include the Balcones Es- carpment populations of Plethodon glutinosus, the Ouachita Mountain populations of P. cinereus, the Ozark Plateau populations of P. dorsalis, and the Interior Highland populations of Eurycea longi- cauda. Undifferentiated Interior Highland popula- tions of Eurycea lucifuga and Hemidactylium scutatum, and Balcones Escarpment and Interior Highland populations of P. glutinosus also may be included with this group. Some of these popula- tions are truly disjunct (Balcones Escarpment P. glutinosus; Interior Highland P. cinereus, P. dor- salis, and H. scutatum) ; ranges of the other species are continuous with those of more easterly popula- tions. I agree with Dowling (1958), Smith (1957), and Blair (1958) that the distribution of this final group is the result of Pleistocene and Recent events. It is likely that invasion of the regions oc- curred from the east during pluvial periods, possibly with the differentiated populations entering rela- tively early, the undifferentiated entering later. The Interior Highland populations of Hemidactylium are relicts of a former more southerly distribution (see below). The corridor by means of which Pleistocene in- vasion of the Interior Highlands occurred extends from the Appalachians through the hill sections of western Kentucky and southern Indiana, the Ozark Hills of southern Illinois, and the St. Francois Mountains and Salem Plateau of eastern and south- eastern Missouri. The entire corridor region is oc- cupied today by Eurycea longicauda, E. lucifuga, and Plethodon glutinosus. Plethodon cinereus ex- tends south to the Ohio River and west to the Illinois border in Indiana and has populations on the Salem Plateau. P. dorsalis extends as far west as the Mississippi River, and Hemidactylium has a relict Salem Plateau population. To the north of the corridor region the distribution of plethodontids is limited by prairies, and to the south by the Coastal Plain. The corridor is most constricted in southern Illinois. There is a striking correlation between the northern limits of the ranges of several of the species, particularly E. longicauda and E. lucifuga, and the maximum southern extent of continental glaciation during the Kansan glacial period (the farthest southern glacial penetration) as mapped by Smith (1961) for the salamanders, and Flint and co-workers (1958) for the glaciers. Glaciated terrain does not provide suitable habitat for most plethodontids. The highly specialized, paedogenetic Haideo- triton occurs in limestone caverns and underground water systems of the Coastal Plain in southwestern Georgia and northern Florida and was probably derived from a Eurycea-\ike ancestral stock rela- tively early. Invasion of the coastal plain by Haideotriton ancestors may have occurred as early as Eocene. The relationship of Hemidactylium to other hemidactyliine genera is not clear, but the com- bination of primitive and specialized characters in- dicates a relatively early origin from a hemi- dactyliine stock. Hemidactylium has departed from the primitive way of life and is rather terrestrial in habits; it breeds in sphagnum bogs and has pond larvae. Hemidactylium has followed favorable en- vironments to the north and northwest and extends considerably farther to the northwest (western Wisconsin) than any other hemidactyliine (Fig. 23). Disjunct relict populations of the genus occur in sphagnum areas in the southern Appalachians, northern Florida, eastern Louisiana, the Ouachita Mountains of Arkansas, and the Salem Plateau of southeastern Missouri. It is apparent that Hemi- dactylium occupied more southerly regions during 1966 EVOLUTION OF LUNGLESS SALAMANDERS 93 periods of Pleistocene glacial maxima than it does now, and that following the last glacial retreat it has spread rapidly into formerly glaciated regions (e.g., Michigan). The relicts of lower latitudes are the remnants of populations that once occupied Pleistocene refugia (see Smith, 1957). Conant (1960) thinks the disjunctness of Hemiclactylium in Missouri and Arkansas, of Plethodon glutinosus in Texas, and of other amphibians and reptiles may be attributed to climatic changes during the Xero- thermic Interval following Climatic Optimum in post-Wisconsin times. The First Terrestrial Encounter — the History of the Bolitoglossines Divergence of the bolitoglossine ancestral stock proceeded very rapidly following movement of the primitive bolitoglossine stock across the adaptive threshold between the aquatic and terrestrial adap- tive zones. Establishment of the group probably occurred very early, possibly in late Cretaceous or early Tertiary. The group spread out from Appa- lachia into terrestrial communities dominated by elements of the Arcto-Tertiary Geoflora, which ex- tended in a broad belt across Holarctica as a mix- ture of conifers and deciduous hardwoods. The Arcto-Tertiary Geoflora linked western Europe, the alpine belt, northeastern Asia, western North America, and eastern North America from early through middle Tertiary, and provided the con- tinuous forest corridor that has presumably been essential for plethodontid dispersal. Bolitoglossines probably reached the West Coast of North America some time in early Tertiary via terrestrial forest corridor routes. Subsequent di- vergence of the plethodonine terrestrial species may have brought plethodonines and bolitoglos- sines into competition, with resultant extinction of the less well adapted bolitoglossines in Appalachia and surrounding eastern North America. Absence of bolitoglossine species from the Great Plains and Rocky Mountains is almost certainly related to Tertiary drying and cooling, and inhospitable Quaternary climates. Three distinct groups of bolitoglossines survive today, the supergenera Hydromantes, Bolitoglossa, and Batrachoseps. Of these Hydromantes is the most primitive and may be relatively little changed from the ancestral bolitoglossine stock (Fig. 16). The present range of Hydromantes is a classic example of disjunct distribution and relict popula- tions (Figs. 20 and 21). It is apparent that a once widespread, primitive, but rather specially adapted ancestral group was unable to adapt to the Tertiary climatic, physiographic, and biotic changes in its former habitat, and survived only in a few rather isolated, specialized habitats in widespread seg- ments of its former range. Two species, H. italicus of Italy and France and H. genei of Sardinia, are the sole living European plethodontids. The route by which Hydromantes entered Europe, probably in early Tertiary, was the Bering Land Bridge between northeastern Asia and northwestern North Amer- ica. Simpson (1947) and others think the bridge was above sea level for most of Tertiary, and was submerged only for considerable lengths of time in middle Eocene and middle to late Oligocene. It is likely that Hydromantes reached Europe as early as early Eocene through an Arcto-Tertiary Forest corridor on the Bering Land Bridge. Absence of Hydromantes from northwestern North America and northern Eurasia may be the result of extinc- tion of most populations during unfavorable cli- matic periods of late Pliocene and Pleistocene. Modern species of Hydromantes are all limited to specialized and rather isolated limestone areas, and all species, both European and American, have relatively small ranges. Several authors have proposed theories in con- flict with that suggested above for Hydromantes dispersal. Dunn (1926) suggested Hydromantes reached Europe in late Miocene or Pliocene. Noble (1931) thought plethodontids reached Europe by way of Greenland “at a time when this northern region enjoyed a warmer climate.” Schmidt (1946) suggested that the relation between west European and west North American faunas is based on a later Tertiary, Pleistocene, and Recent dispersal across the Bering Land Bridge. The genus Batrachoseps is an ancient group of highly specialized salamanders. The most primitive and generalized species is B. wrigliti, which occu- pies forests derived from the Arcto-Tertiary Geo- flora in the Cascade Mountains in northern and central Oregon. The species is the only one re- stricted to the presumed ancestral environment. B. attenuatus ranges from southern Oregon to southern California, and occurs in a variety of habi- tats— in forests derived from the Arcto-Tertiary Geoflora to areas dominated by derivatives of the Madro-Tertiary Geoflora. B. pacificus of the low- lands of Southern California and northern Baja California is restricted to coastal plant communi- ties derived from Madro-Tertiary elements. Two and possibly more undescribed species occupy dis- junct, relict ranges in areas of Arcto-Tertiary de- rived plant communities in the Santa Lucia Moun- tains and southern Sierra Nevada of California, and the high Sierra San Pedro Martir of northern Baja 94 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 California. This distributional pattern (Fig. 26) suggests a once far-ranging group which has been restricted to relatively small areas in what are for plethodontid salamanders specialized environ- ments. Evidence that the genus once had a more extensive range is based on two marginal records. The first is the collection of the types of Batracho- seps caudatus from southeastern Alaska (Cope, 1889). No further individuals of the species have been found, and subsequent authors (Dunn, 1926; Stebbins, 1951) have questioned the authenticity of the locality data. There has been little collecting in Alaska and there is little reason to question the record. The second is the collection of a single im- mature individual of Batrachoseps from the north- ern slope of the Nevado de Colima at about 7,000 feet elevation in the state of Jalisco, Mexico. The collector was Hans Gadow, a noted biologist, who discussed conditions of the collection in some detail in two separate publications (1905, 1908). He was aware of the significance of his discovery from the first. The specimen, in the British Museum, has been examined by Stebbins and Lowe (1949), Hen- drickson ( 1954), Arden H. Brame, Jr. and me. All concur that it is a Batrachoseps. The Mexican rec- ord makes good biogeographic sense for a bolito- glossine genus, and I consider it to be authentic. It is possible that other populations of Batrachoseps occur as relicts in the Sierra Madre Occidental of Mexico. Failure of collectors to obtain additional material may be due to the relative lack of intensive collecting during the right seasons in the critical areas. The entrance of the supergenus Bolitoglossa into Mexico has recently been discussed in some detail (Brame and Wake, 1963), and only a review will be presented here. It is proposed that a bolito- glossine stock ancestral to the supergenus Bolito- glossa was associated in early Tertiary with elements of the Arcto-Tertiary Geoflora at relatively high latitudes. Axelrod (1960) has stated that a mixing of elements of the Arcto-Tertiary and Madro- Tertiary Geofloras started in middle Cretaceous and formed what he calls a broad “ecotone.” The “ecotone” extended southward along low mountains into Mexico during Cretaceous and early Eocene. Species of plants, or paired-species, which now oc- cur disjunctly in Appalachia and the Mexican and Guatemalan highlands, probably entered the area at this time (see also Braun, 1950; Martin and Harrell, 1957); these plants are relicts of the old world-wide Arcto-Tertiary Geoflora, and similar species occur today in eastern Asia. Plethodontid salamanders may also have entered the “ecotone” along with their Arcto-Tertiary associates, both plant and animal. The entire assemblage shifted to the south, especially at high elevations, as the tropical border continued to move southward in Oligocene and Miocene in response to long range Tertiary cooling trends. The Madro-Tertiary Geoflora was developing in the increasingly arid southwestern United States and northern Mexico from Eocene throughout Tertiary. Increased lowland aridity favored expan- sion of the Geoflora between the main portions of the Arcto-Tertiary belt to the north and the Arcto- Tertiary -Neotropical-Tertiary “ecotone” isolated in the highlands to the south. Presumably the long- range Tertiary climatic trends resulted in effective separation of the ancestral populations of the super- genus Bolitoglossa from other bolitoglossine popu- lations to the north. It is apparent that Appalachia is the primary dispersal center of plethodontid salamanders, but it is equally apparent that the moist borders of the Mexican Plateau have acted as a secondary center of evolution and dispersal. Seven genera and over one hundred species of bolitoglossine salamanders have been derived from the ancestral Tertiary popu- lations. A number of species remain in the ancient “ecotonal” situation, but several lines have suc- cessfully adapted to other habitats where they are associated with derivatives of the Neotropical Tertiary Geoflora. Significantly the more struc- turally primitive and generalized species have remained in the highland “ecotone.” Three distinct lines of tropical bolitoglossine evo- lution are apparent (Fig. 17). Perhaps the earliest divergence from the ancestral stock was by Bolito- glossa. The genus is found in the Mexican High- lands but has also extended into the Central Ameri- can and South American highlands as far south as Ecuador and Venezuela. Throughout its range various species have invaded and adapted to the lowland habitats where many species have become arboreal. Lowland species of Bolitoglossa range from Mexico through Central America to Bolivia and Brazil (Figs. 27 and 29). Invasion of the low- lands and acquisition of arboreal habits has ap- parently occurred in parallel several times in the genus. Highland species are usually both more ^primitive and more generalized than lowland species. With only a few exceptions (e.g., rufescens, occidentalis) the more northerly species of any habitat type are more primitive and generalized 4han their more southerly ecological equivalents. Stuart (1954) has suggested that Magnadigita ( = primitive highland Bolitoglossa ) is autochthonous in Central America, and has spread north and south from the Guatemalan Highlands. What may be the 1966 EVOLUTION OF LUNGLESS SALAMANDERS 95 Figure 26. Distribution of Batrachoseps. 96 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 Figure 27. Distribution of Bolitoglossa and Thorius in Middle America. Figure 28. Distribution of Oedipina in Middle America. The genus also has been reported from Chiapas, Mexico. Figure 29. Distribution of Bolitoglossa and Oedipina in South America. 1966 EVOLUTION OF LUNGLESS SALAMANDERS 97 most primitive species is a Guatemalan highland form, B. dunni, which retains columellae, pre- frontals, primitive vertebrae, long limbs, slightly webbed hands and feet, primitive numbers of carpals and tarsals, tibial spurs, and primitive vomers. The most specialized species appear to be B. rufescens of the northern lowlands and mem- bers of the South American palmata and altama- zonica groups (Brame and Wake, 1963) . A second phylogenetic line contains the highly specialized, greatly elongated species of Oedipina. These bizarre organisms are attenuate forms with elongated trunks, reduced limbs, and extremely long tails (often over seventy-five per cent of total length). The high degree of specialization, adapta- tion to lowland environments, and penetration of South America all suggest early divergence from the common tropical bolitoglossine ancestral stock. Two major sections are evident. One contains rela- tively short-bodied forms with few or no maxillary teeth and relatively large, essentially fully webbed feet (elongatus, parvipes, complex ); they are found over the entire range of the genus from Mexico, British Honduras, and Guatemala to north-central Ecuador. The other section includes relatively long- bodied species with many maxillary teeth and rather small, syndactylous feet (all other species) which range from Guatemala to western Panama. Oedi- pina is the only plethodontid genus found primarily in tropical lowlands and is the only tropical genus that does not occur in the highlands along or near the southern and eastern margins of the Mexican Plateau. Costa Rica is the present dispersal center and apparently has been the center of evolution of this, the most specialized and advanced pletho- dontid genus (Figs. 28 and 29). The third and final line includes those species now assigned to the genera Pseudoeurycea, Chi- ropterotriton, Thorius, Parvimolge, and Linea- triton. All are found in mountains in and around the southern margins of the Mexican Plateau (Figs. 27, 30, 31). Thorius, Lineatriton, and probably Parvimolge (depending on the assignment of the Costa Rican species, richardi) are limited to the area north of the Isthmus of Tehuantepec. The primitive species of both Pseudoeurycea and Chi- ropterotriton are also found north of the Isthmus. The mountainous southern and eastern margins of the Mexican Plateau have been the center of evolution of the group. Pseudoeurycea and Chiropterotriton extend far- ther to the northeast than any other bolitoglossines, occurring in the mountains of southwestern Tamaulipas and southern Nuevo Leon. These northeastern populations are separated from the main populations to the south by dry gaps ( e.g San Louis Potosian Gap). Martin (1958 a) refers these populations to a humid montane group of the Northeast Madrean Herpetofaunal Component. The salamanders of this component must be con- sidered one of several in the Central American Complex of the Old Northern Herpetofaunal Ele- ment of Savage (1960). Martin (1958 a) suggests that the northeastern isolates reflect post-pluvial confinement of a habitat formerly widespread. Sev- eral species (C. multidentatus, P. bellii, P. cepha- lica) occur on both sides of the ecological barrier and the suggestion seems reasonable. Both Pseudoeurycea and Chiropterotriton have crossed the Isthmus of Tehuantepec and entered the mountains of southern Mexico and Guatemala. Chiropterotriton extends even farther to the south into northeastern Costa Rica. The increasing spe- cialization of the southern Chiropterotriton has been previously discussed. Stuart (1957) has sug- gested that Pseudoeurycea moved southward over the high plateaus of Chiapas and Guatemala after crossing the low, hot Isthmus of Tehuantepec in the Pleistocene. Stuart (1954) thinks that Pseudoeury- cea is unquestionably a recent entrant into Nuclear Central America from Mexico, and the slight penetration of Pseudoeurycea into Central America supports this suggestion. Stuart (1954) also sug- gests that Chiropterotriton reached Guatemala by relatively low Pacific mountain-slope routes. In view of the great differentiation and greater south- ward penetration of Chiropterotriton, it is possible that the genus may have crossed the Isthmus of Tehuantepec at an earlier date than Pseudoeurycea. Numerous biogeographic problems of consider- able interest and importance to the local and regional distribution of the genera Pseudoeurycea, Chiropterotriton, Bolitoglossa, and Oedipina in Guatemala are discussed by Stuart (1935, 1943, 1948, 1950, 1951, 1954, 1957, 1958) and Schmidt (1936). To summarize, the history of the neotropical plethodontids has involved mainly southward movements along the Central American axis. Major movements have probably been via highland routes. Two genera, Bolitoglossa and Oedipina, have been in Nuclear Central America for a long time and may have arisen there. Others (Pseudoeurycea, Chiropterotriton, IParvimolge) are probably more recent invaders. It seems likely that northward movements of Bolitoglossa and southward move- ments of Pseudoeurycea across the Isthmus of Tehuantepec by highland routes may have occurred during Pleistocene. Stuart (1954) has stated that geographic continuity existed during Pleistocene, Pseudoeurycea | Parvimolge i i Figure 30. Distribution of Chiropterotriton and Lineatriton. Figure 31. Distribution of Pseudoeurycea and Parvimolge. 1966 EVOLUTION OF LUNGLESS SALAMANDERS 99 when a temperature drop of about 5° C would have permitted the majority of highland species to move across Nuclear Central America. While the degree of southward distribution is dependent in part upon the time of entrance into Nuclear Central America, the vagility of the group must also be considered (Stuart, 1954). Bolitoglossa and Oedipina are clearly early derivatives of the ancestral stock, but the greater southward penetration of Bolitoglossa may be more a reflection of greater vagility than of greater antiquity. The Second Terrestrial Encounter — the History of the Plethodonines The second terrestrial group, the Plethodontini, is found today in eastern and western North America, with several Rocky Mountain isolates. The ancestral stock probably diverged from a primitive plethodontine group in what is now Appa- lachia later than the divergence of the bolitoglossine stock. Neither terrestrial group is ancestral to the other. Although bolitoglossines may have arisen earlier, both groups were probably well established by Oligocene. There are three genera of plethodonines. Aneides has one species (aeneus) in southern Appalachia, one (hardii) in the high mountains of south-central New Mexico, and three (ferreus, flavipunctatus, lugubris) on the Pacific Coast (Fig. 32). Plethodon has eleven species in eastern North America of which six (welleri, richmondi, wehrlei, jordani, longicrus, yonahlossee ) are primarily Appalachian species, two (ouachitae, caddoensis) are endemic to the Interior Highlands, two (cinereus, dorsalis) oc- cur in both areas, and one (glutinosus) occurs in Appalachia, the Coastal Plain, the Interior High- lands, and the Edwards Plateau of Texas. Five species of Plethodon (elongatus, dunni, larselli, stormi, vehiculum) are restricted to the Pacific Coast of northwestern United States and southern British Columbia, one ( vandykei ) occurs in western Washington, northern Idaho, and northwestern Montana, and one (neomexicanus) occurs in the highlands of north-central New Mexico (Fig. 33). The single species of Ensatina (eschscholtzii) ranges from British Columbia to extreme southern Califor- nia along the Pacific Coast and the slopes of the Cascade-Sierra Nevada system (Fig. 19). Lowe (1950) suggested that the history of Aneides is related to the Cenozoic history of the Arcto-Tertiary Geoflora in North America. It is assumed by Lowe and by me that the present day association of Aneides with Arcto-Tertiary floral elements is an ancient one and that much can be learned about the history of Aneides by studying the history of the geoflora. All plethodonine genera are now members of communities in v/hich floral elements derived from the Arcto-Tertiary Geo- flora dominate or play important roles, so that Lowe’s suggestions concerning Aneides are at least in part applicable to the tribe as a whole. The history of the Arcto-Tertiary Geoflora in North America has been outlined in broad form by Chaney, Condit, and Axelrod (1944), Axelrod (1948, 1950, 1956, 1957, 1960), Braun (1950, 1955), and MacGinitie (1958). In early Tertiary the Tropical-Tertiary Geoflora formed a broad belt which in North America extended as far north as southeastern Alaska on the West Coast and pos- sibly to Nova Scotia on the East Coast (Axelrod, 1960). Above the tropical border was the circum- polar Arcto-Tertiary Geoflora. Tertiary cooling and drying trends resulted in a southward shift of the tropical border and a concomitant southward shift of the Arcto-Tertiary Geoflora. A broad “ecotone” formed where the two major floral groups met. As a result of climatic changes during middle and late Tertiary a number of segregated elements evolved, two of which, the East American Element and the West American Element (Chaney, Condit, and Axelrod, 1944; Axelrod, 1960), are of primary importance to this discussion. Ancestral plethodonines invaded and adapted to the terrestrial environment in the Arcto-Tertiary forests of Appalachia and began to move westward in the favorable habitats provided by the forests, probably sometime in early Tertiary. Secular cool- ing and drying trends reinforced by the rainshadow following the uplift of the Rocky Mountains dis- rupted the Arcto-Tertiary belt in central North America. Forests were replaced by savannah and possibly some grassland. It is difficult to date the disjunction, but Miocene is my best estimate. In an important discussion of late Tertiary biogeography of the Great Basin, Shotwell (1961) reviewed the distribution of fossil horses from late Miocene to the end of Tertiary. He records Merychippus ( Protohip pus ) from as far north as northwestern Montana as early as the Barstovian faunas (late Miocene). This is an indication of savannah and possibly grassland habitats (see Shotwell, 1961; 1963, for correlations of fossil horse genera and habitats). He also indicates that central North America was covered by dry facies of Arcto- Tertiary vegetation during late Miocene; such a situation was probably unsuitable for continuous plethodontid occupation. Pliohippus is present in Clarendonian faunas (early Pliocene) in many areas of central United States, coinciding with the Aneides 1000 MILES 200 400 600 800 1000 1200 1400 KILOMETERS Plethodon Figure 32. Distribution of Aneides. Figure 33. Distribution of Plethodon 1966 EVOLUTION OF LUNGLESS SALAMANDERS 101 expansion of open grassland (Shotwell, 1961). However, MacGinitie (1962) doubts that large grassland areas existed until Pleistocene or Recent times. MacGinitie has shown that what is now northern Nebraska was a savannah region in late Miocene with Neotropical-Tertiary relicts, pine-oak forest elements (Madro-Tertiary), and a swamp cypress component of the East American Element of the Arcto-Tertiary Geoflora. All such habitats are now unsuitable for plethodontid dispersal. Thus at some time from middle Miocene to early Pliocene the environment of central North America became inhospitable for plethodontids, and the species in the east and in the west were separated. The theory proposed here for dating of the dis- junction of the ranges of plethodonine genera is in conflict with that of Dunn (1926), who did not have the benefit of the important paleobotanical literature of the past thirty years. My theory is in fairly close agreement with that of Lowe (1950), which placed disjunction of Aneides in late Mio- cene with complete disjunction during early Pliocene. Plethodonine genera were well established by the time of the disjunction. Two of the three genera (Aneides, Plethodon) occur on both sides of the grassland barrier, and forest corridors are not known to have existed across the barrier. Dating of the disjunction of the ranges of Pletho- don and Aneides provides some indirect evidence on the relative age of the genera. Divergence of Aneides from a Plethodon-Yike ancestral stock must have occurred at an early date. By the time of the range disjunctions both genera were widespread, perhaps with transcontinental distributions. Generic differentiation may date from Oligocene or early Miocene. Aneides and Ensatina may have arisen inde- pendently from Plethodon in western North Amer- ica. While the most primitive species of Plethodon are found in eastern North America (Highton, 1962), the most primitive Aneides ( hardii , see Wake, 1963) is found in the west (New Mexico) and Ensatina is limited to the Pacific Coastal re- gion. Four of the five Aneides are western species and the fifth ( aeneus of Appalachia) is a morpho- logically specialized species that occupies a highly specialized arboreal-rock crevice habitat (Gordon, 1952; Wake, 1963). Plethodonines were probably widespread in western United States in Miocene. Elimination of the group from most of the area east of the Cascade- Sierra Nevada divide and west of the Great Plains occurred subsequent to midcontinental disjunction. The researches of Axelrod (1957, 1958), Axelrod and Ting (1960, 1961), and Shotwell (1961, 1963) are particularly pertinent to this discussion. De- velopment of the xeric Madro-Tertiary Geoflora in the interior southwestern United States began in early Tertiary. During Tertiary there was a general drying of western North America favoring expan- sion of the Madro-Tertiary elements at the expense of the Arcto-Tertiary elements. On the basis of geofloral distribution and the distribution of fossil horses, Shotwell (1961, 1963) has made the fol- lowing suggestions concerning the history of the Great Basin in late Tertiary. In Barstovian times the mesic Arcto-Tertiary elements extended throughout the Great Basin but were mixed with drier Madro-Tertiary elements in the south. Shot- well states that by Clarendonian times the Madro- Tertiary Geoflora extended northward, and Arcto- Tertiary elements had largely been displaced from the central Great Basin except for moister local areas (high elevations, stream borders). Pliohippus has been found in Clarendonian deposits in much of the Great Basin. This fact indicates that savan- nah or grassland was present. Both habitats are unfavorable for plethodontids. By Hemphillian times (middle to late Pliocene) increasing cold stopped the northward extension of the Madro- Tertiary elements, but the present semiarid charac- ter of the Great Basin was already established. No woodland elements were present in Hemphillian times in the northern Great Basin (Shotwell, 1963 ) . Later Tertiary and Quaternary changes in the flora of the Great Basin were the result of secular drying and cooling reinforced by Pleistocene warp- ing and faulting which raised the Sierran crest from five thousand to nine thousand feet (Axelrod, 1950; 1957; Axelrod and Ting, 1960; 1961). This uplift, according to Axelrod (1957), would have ac- counted for a ten to fifteen inch reduction in rainfall to the east. Increasing continentality of climate, rain shadow effect of the Sierran uplift, and severe cooling of Pleistocene glacial periods all con- tributed to the disappearance of plethodontids over most of the Rocky Mountain and interior range of the group. The distribution of the West Coast species was further severely limited by the major Sierra Nevada uplift of Pleistocene times (see Axel- rod and Ting, 1961 ) . Three species of plethodonines occur today in the Rocky Mountains. Plethodon vandykei of coastal Washington has disjunct populations in northern Idaho and Montana. Several other am- phibians are also found in both areas (Ascaphus truei, Dicamptodon ensatus, Taricha granulosa) and at least one lizard (Gerrhonotus coeruleus) is distributed continuously across intervening moun- 102 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 tain ranges in southern British Columbia. Disjunct distribution of P. vandykei is probably the result of Pleistocene or Recent events. P. neomexicanus and A. hardii of New Mexican highlands are apparently relicts of the mid-Tertiary Rocky Mountain fauna as suggested by Lowe (1950). Murray (1957) thinks final isolation may have occurred as late as late Pleistocene. Martin (1958 b) states that 4,000 to 4,500 foot altitudinal displacement of biotic zones occurred during full- glacial periods in the Southwest. It is possible that final southern displacement of the species may have occurred at that time. Survival of the two species is related to the fact that they have been able to adapt to favorable microhabitats in the relatively inhospitably cold Rocky Mountains. Blair (1958) has suggested that P. neomexicanus may have arisen from populations that dispersed directly across Oklahoma in Wisconsin times, and that A. hardii may have been connected with the eastern population as recently as late Pleistocene. The New Mexican highland flora as a whole is clearly derived from the West American Element of the Arcto-Tertiary Geoflora. This indicates a western, not eastern origin for the salamanders as well (see Chaney in Chaney, Condit, and Axelrod, 1944). Both species are isolated at altitudes above 8,000 feet. They are well defined species, not closely related to any others. It is likely that they are dis- placed southern relicts of the ancient Rocky Moun- tain fauna, and their evolutionary history is not related strictly to Pleistocene events. Pacific Coast species have undergone a moderate amount of diversification. Ensatina eschscholtzii is the most primitive in morphology, the most gen- erally adapted in ecology, and the most widespread of the western plethodonines. Other species have more restricted ranges, and are generally more specialized (see Peabody and Savage, 1958, for detailed biogeographic analyses of western species). To summarize, the plethodonines occurred all across the continent in early Tertiary. Middle and late Tertiary events are responsible for the present generic disjunctions. Plethodonines are more suc- cessful than other high latitude terrestrial genera and were apparently the victors in competitive en- counters with the bolitoglossines in the east. West- ern plethodonines have also had notable success, and the bolitoglossine genera have survived the plethodonine invasion only by specializing. The failure of plethodonines to reach the tropics and the Old World is probably the result of relatively recent derivation from the plethodontine ancestral stock. CONCLUSIONS AND SUMMARY The large, diverse family Plethodontidae con- tains more than two-thirds of the living species of salamanders. The morphology, ecology, and be- havior of the species have shown them to be the most progressive and advanced group within the order Caudata. The general adaptive level of the family is high, relative to other salamanders, and a successful radiation of specially adapted species has occurred. Because so many of the adaptive types have survived, the family has been ideal for the study of major features of evolution, especially the morphological bases for adoption of new ways of life and the ensuing adaptive radiations that typically characterize such events. A functional and evolutionary morphological analysis has been employed to outline the evolu- tionary patterns of individual bony elements, func- tional anatomical units, and groups of organisms. This analysis has led to a new classification of the family and has formed the basis of theories con- cerning major features of evolution within the family. The twenty-three genera and one hundred eighty- four species of plethodontids may be grouped in two subfamilies based on morphological evidence. The more specialized subfamily Desmognathinae is the smaller and has the more restricted geographic range. Its three genera (ten species) live in eastern North America, centered in the southern Appa- lachian mountains, an area presumed to be the an- cestral home of the family. The larger and more generalized subfamily Plethodontinae includes twenty genera grouped in three tribes. The most primitive tribe, the Hemidactyliini (eight genera, twenty-four species) is concentrated in eastern North America. The tribe Plethodontini (three genera, twenty-four species) occurs in both eastern and western North America. The remaining nine genera are included in the most advanced tribe, the Bolitoglossini (one hundred twenty-six species), a group that ranges from western North America into the Neotropics, and has two species in Europe. It is the only plethodontid group that lacks repre- sentatives in eastern North America. The two subfamilies of lungless salamanders differ markedly in the functional morphology of the head. In the Desmognathinae the lower jaw can be lowered only until a ligament extending from the mandible to the atlas vertebra becomes taut. The skull is then raised to open the mouth further. Morphological modifications directly and indirectly associated with this functional shift are great and the desmognathine salamanders differ considerably 1966 EVOLUTION OF LUNGLESS SALAMANDERS 103 from the more generalized plethodontines. The three tribes of plethodontines differ mainly in anatomical details, such as hyobranchial configura- tion, form of cranial elements, and vertebral differences. Each of the four major groups of plethodontids has undergone a secondary adaptive radiation. As a result structural and ecological parallelism are common phenomena in the family, and include the following examples. Tongues are primitively at- tached to the front of the mouth in salamanders, including the Desmognathinae and Plethodontini. In one genus each of the Hemidactyliini (Hemidactylium) and Bolitoglossini (Batracho- seps) tongues are attached, although in a modi- fied form, to the anterior margins of the mouth. Parallel evolution of free tongues has occurred in these two tribes and free tongues are found in nearly all species of the groups. In addition there is a trend in the direction of tongue freedom in some Plethodontini (Ensatina). Primitive members of the family have aquatic larvae and are aquatic to semiaquatic as adults. There are distinct parallel trends in the direction of terrestrialism with loss of aquatic larvae. All species of the Bolitoglossini and Plethodontini are terrestrial and there are definite tendencies in the direction of terrestrialism in cer- tain species of the other groups as well (Desmogna- thus aeneus and wrighti; Hemidactylium scutatum). Elongated trunks have evolved in parallel in each of the four groups and within genera of a single group. Basal tail constriction and associated special- izations have arisen independently in the three tribes. The fifth toe has been lost in three separate genera. Many additional instances of parallelism have been cited. Parallelism is the rule rather than the exception in the evolution of the family. One genus dominates each group in terms of numbers of species and size of geographic range (Desmognathus, Eurycea, Plethodon, Bolitoglossa). Within each of these genera a tertiary level of adap- tive radiation has occurred. Diversity of size, habi- tat, behavior, and coloration characterizes each of the genera. Paedomorphosis has been a very important morphological mode of evolution in the family. Species of four genera of hemidactyliines are paedogenetic. This developmental phenomenon has permitted survival of specialized forms in perma- nently aquatic and underground habitats in regions unsuitable for their more generalized relatives. Paedogenetic species are so highly specialized that they are probably evolutionary dead-ends. By con- trast the phenomenon of differential metamorphosis has given bolitoglossines access to new ways of life in tropical environments and has contributed to a relatively large scale adaptive radiation and evolu- tionary diversification. The evolution of a great variety of special adap- tive features has been associated with the adaptive radiation of the lungless salamanders. Many ex- amples have been cited, among them new feeding mechanisms in desmognathines, skull strengthening adaptations in forms that force their way head first under rocks or make burrows (desmognathines) and in forms that have adopted new feeding tech- niques (Aneides), limb shortening in burrowers of several genera, toe loss, basal tail constriction and adaptations for tail autotomy, specialized climbing adaptations in the limbs, hands and feet, eye re- duction and loss in cave dwellers, size diminution, gigantism, and others. The radiation has been ex- tensive and complex, with the adaptations appear- ing in parallel and multiple parallel lines found within the family and familial subgroups. The re- sult is an evolutionary pattern that is characterized by the mosaic distribution of primitive and ad- vanced, generalized and specialized characters on all taxonomic levels. Knowledge of the late Mesozoic and Cenozoic history of the Northern Hemisphere, fossil history of the family, ecology and distribution of the living species, and evolutionary relationships within the family has formed the basis for a theoretical con- sideration of the phylogeny and biogeography of the family. Ancestors of the family were probably derived from a stock close to that which gave rise to ambystomatid salamanders. Appalachia is the present center of dispersal and earlier suggestions that this region is the source area are accepted. Divergence of the desmognathine line from the main line of plethodontid evolution occurred very early, probably by the beginning of Cretaceous times. Plethodontine tribal divergence probably dates from earliest Tertiary. The first plethodontine group to differentiate may have included the ter- restrial ancestors of the modern bolitoglossines. The plethodonines are thought to have arisen after the bolitoglossines because they have a less extensive range, have undergone a less extensive adaptive radiation and diversification, and they have living representatives in eastern North America. Possibly competition with the more vigorous plethodonines led to elimination of the older and more specialized bolitoglossines from eastern North America. The Hemidactyliini represent the diversification of the ancestral stock. Some of the genera (e.g., Hemidac- tylium) diverged from the stock very early, perhaps as early as the ancestors of the other tribes. The group is thus less compact than the other two tribes. 104 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 There is evidence that the family remains evolu- tionarily active. Entrance into the tropics has given the salamanders access to new ways of life, and such genera as Thorius, Oedipina, and Bolitoglossa may be initiating new phylogenetic trends. Activity of these groups is attested by the facts that forty per cent of all living species of salamanders are found within a single supergenus that ranges from Mexico into South America, and more than twenty- five per cent of the species of plethodontid sala- manders are members of the genus Bolitoglossa. Literature Cited Atwood, W. W. 1940. The physiographic provinces of North America. Boston: Ginn and Co., 536 pp. Auffenberg, W. 1961. A new genus of fossil salamander from North America. Amer. Midi. Nat., 66(2): 456-465. Axelrod, D. I. 1948. Climate and evolution in western North America during middle Pliocene time. Evo- lution, 2(2): 127-144. 1950. Evolution of desert vegetation in western North America. Carnegie Inst. Wash. Publ., 590: 215-306. 1956. Mid-Pliocene floras from west-central Ne- vada. Univ. California Publ. Geol. Sci., 33: 1-322. 1957. Late Tertiary floras and the Sierra Nevadan uplift. Bull. Geol. Soc. Amer., 68: 19-46. 1958. Evolution of the Madro-Tertiary Geoflora. Bot. Rev., 24: 433-509. 1960. The evolution of flowering plants. In Tax, S. (ed.), Evolution after Darwin. Chicago: the University of Chicago Press, Vol. 1, The evolution of life, pp. 227-305. Axelrod, D. I. and W. S. Ting 1960. Late Pliocene floras east of the Sierra Ne- vada. Univ. California Publ. Geol. Sci., 39(1): 1-118. 1961. Early Pleistocene floras from the Chagoopa surface, southern Sierra Nevada. Univ. Cali- fornia Publ. Geol. Sci., 39(2) : 119-194. Baird, I. L. 1951. An anatomical study of certain salamanders of the genus Pseudoeurycea. Univ. Kansas Sci. Bull., 34, pt. 1(6): 221-265. Baker, J. K. 1957. Eurycea troglodytes : a new blind cave sala- mander from Texas. Texas Jour. Sci., 9(3) : 328-336. 1961. Distribution of and key to the neotenic Eurycea of Texas. Southwestern Nat., 6(1): 27-32. Bishop, S. C. 1941. The salamanders of New York. New York State Mus. Bull., 324: 1-365. 1943. Handbook of salamanders. Ithaca, New York: Comstock Publ. Co., Inc., 555 pp. Blair, A. P. 1961. Metamorphosis of Pseudotriton palleucus with iodine. Copeia, 1961(4) : 499. Blair, W. F. 1958. Distributional patterns of vertebrates in the southern United States in relation to past and present environments. In Hubbs, C. L. (ed.), Zoogeography. Washington: Amer. Assn. Adv. Sci., publ. 51(17): 433-468. Blanchard, F. N. 1923. The life history of the four-toed salamander. Amer. Nat., 62: 156-164. Boulenger, G. A. 1882. Catalogue of the Batrachia Gradientia s. Caudata in the collection of the British Museum, ed. 2. London: British Museum, 127 pp. Brame, A. H., Jr. 1960. The salamander genus Oedipina in North- ern Central America. Bull. So. Calif. Acad. Sci., 59(3): 153-162. Brame, A. H., Jr. and D. B. Wake 1963. The salamanders of South America. Los Angeles Co. Mus. Contrib. Sci., 69: 1-72. Brandon, R. A. 1965. Morphological variation and ecology of the salamander Phaeognathus hubrichti. Copeia, 1965(1): 67-71. 1966. A re-evaluation of the status of the sala- mander Typhlotriton nereus Bishop. Copeia, 1966(3): 555-561. Braun, E. L. 1950. Deciduous forests of eastern North Ameri- ca. Philadelphia and Toronto: Blakiston, 596 pp. 1955. The phytogeography of the unglaciated east- ern United States and its interpretation. Bot. Rev., 21: 297-375. Burger, W. L., H. M. Smith, and F. E. Potter 1956. Another neotenic Eurycea from the Ed- 1966 EVOLUTION OF LUNGLESS SALAMANDERS 105 wards Plateau. Proc. Biol. Soc. Wash., 63 : 51-58. Carr, A. F., Jr. 1940. A contribution to the herpetology of Flori- da. Univ. Florida Publ. Biol. Sci. Ser., 3(1): 1-118. Chaney, R. W., C. Condit, and D. I. Axelrod 1944. Pliocene floras of California and Oregon. Carnegie Inst. Washington Publ., 553: 1- 407. Conant, R. 1958. A field guide to reptiles and amphibians of the United States and Canada east of the 100th meridian. Boston: Houghton Mifflin, 366 pp. 1960. The queen snake, Natrix septemvittata, in the interior highlands of Arkansas and Mis- souri, with comments upon similar disjunct distributions. Proc. Acad. Nat. Sci., Phila- delphia, 112(2): 25-40. Cooper, G. W. 1965. Induction of somite chondrogenesis by car- tilage and notochord. A correlation between inductive activities and specific stages of cytodifferentiation. Devel. Biol., 12: 185- 212. Cope, E. D. 1859. On the primary divisions of the Salaman- dridae with descriptions of two new species. Proc. Acad. Nat. Sci., Philadelphia, 1859: 122-128. 1866. On the structure and distribution of the genera of the arciferous Anura. J. Acad. Nat. Sci., Philadelphia, ser. 2, vol. 6, pt. 1(2): 67-112. 1867. A review of the species of the Ambysto- midae. Proc. Acad. Nat. Sci., Philadelphia, 19: 166-211. 1869. A review of the species of the Plethodon- tidae and Desmognathidae. Proc. Acad. Nat. Sci., Philadelphia, 1869: 93-118. 1889. The Batrachia of North America. Bull. U.S. Natl. Mus., 34: 1-525. 1894a. On a collection of Batrachia and Reptilia from southwest Missouri. Proc. Acad. Nat. Sci., Philadelphia, 1893: 383-385. 1894b. Second addition to the knowledge of the Batrachia and Reptilia of Costa Rica. Proc. Amer. Philos. Soc., 31(142): 333-347. De Beer, G. 1958. Embryos and ancestors, ed. 3. London, Ox- ford Univ. Press, 197 pp. Dent, J. N. 1942. The embyronic development of Plethodon cinereus as correlated with the differentia- tion and functioning of the thyroid gland. J. Morph., 71: 577-601. Dent, J. N. and J. S. Kirby-Smith 1963. Metamorphic physiology and morphology of the cave salamander, Gyrinophilus pal- leucus. Copeia, 1963(1): 119-130. Dent, J. N., J. S. Kirby-Smith, and D. L. Craig 1955. Induction of metamorphosis in Gyrinophi- lus palleucus. Anat. Rec., 121-129. Dowling, H. G. 1958. Geographic relations of Ozarkian amphib- ians and reptiles. Southwestern Nat., 1(4): 174-189. Dundee, H. A. 1957. Partial metamorphosis induced in Typhlo- molge rathbuni. Copeia, 1957(1): 52-53. 1962. Response of the neotenic salamander Haide- otriton wallacei to a metamorphic agent. Science, 135: 1060-1061. Dunn, E. R. 1917. The salamanders of the genera Desmogna- thus and Leurognathus. Proc. U.S. Natl. Mus., 53: 393-433. 1922. The sound-transmitting apparatus of sala- manders and the phylogeny of the Caudata. Amer. Nat., 56: 418-427. 1923. Mutanda herpetologica. Proc. New England Zool. Club, 8: 39-40. 1926. The salamanders of the Family Plethodon- tidae. Northampton, Mass.: Smith College, vii + 441 pp. 1927. On the relationships of certain plethodont salamanders. Copeia, 165: 102-106. 1928. The habitats of plethodontidae. Amer. Nat. 680: 236-248. 1931. The herpetological fauna of the Americas. Copeia, 1931(3): 106-119. 1941. The “opercularis” muscle of salamanders. J. Morph., 69(2): 207-215. Edgeworth, F. H. 1935. The cranial muscles of vertebrates. London: Cambridge Univ. Press, 493 pp. Emerson, E. T., II 1905. General anatomy of Typhlomolge rathbuni. Proc. Boston Soc. Nat. Hist. 32(3) : 42-76. Estes, R. 1964. Fossil vertebrates from the late Cretaceous 106 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 Lance formation, Eastern Wyoming. Univ. California Publ. Geol. Sci., 49: 1-187. 1965. Fossil salamanders and salamander origins. Amer. Zoologist, 5: 319-334. Flint, R. F., R. B. Colton, R. P. Goldthwait, and H. B. Willman 1959. Glacial map of the United States east of the Rocky Mountains, ed. 1. New York: Geol. Soc. America. Fowler, H. W. and E. R. Dunn 1917. Notes on salamanders. Proc. Acad. Nat. Sci., Philadelphia, 1917: 7-28. Gadow, H. 1905. The distribution of Mexican amphibians and reptiles. Proc. Zool. Soc. London, 2: 191- 224. 1908. Through Southern Mexico. London: Witherby and Co., xvi + 527 pp. Garstang, W. 1922. The theory of recapitulation: a critical re- statement of the biogenetic law. J. Linn. Soc. London, Zoology, 35: 81-101. Gehlbach, F. R. 1959. New plethodontid salamanders of the genus Thorius from Puebla, Mexico. Copeia, 1959(3): 203-206. Goin, C. J. 1951. Notes on the eggs and early larvae of three more Florida salamanders. Ann. Carnegie Mus., 32(2): 253-263. Goin, C. J. and W. Auffenberg 1958. New salamanders of the family Sirenidae from the Cretaceous of North America. Fieldiana-Geology, Chicago, 10(33): 449- 459. Goin, C. J. and O. B. Goin 1962. Introduction to herpetology. San Francisco: Freeman, 341 pp. Gordon, R. E. 1952. A contribution to the life history and ecol- ogy of the plethodontid salamander Aneides aeneus (Cope and Packard). Amer. Midi. Nat., 47(3): 666-701. Gorman, J. 1956. Reproduction in plethodont salamanders of the genus Hydromantes. Herpetologica, 12(4): 249-259. Graham, A. 1964. Origin and evolution of the biota of south- eastern North America: evidence from the fossil plant record. Evolution, 18: 571-585. Gray, J. E. 1850. Catalogue of the specimens of Amphibia in the collection of the British Museum, ed. 1. Part II. Batrachia Gradientia, etc. 72 pp. Grobman, A. B. 1943. The appearance during ontogeny of a suture between previously fused dermal bones. Anat. Rec., 87: 211-214. 1959. The anterior cranial elements of the sala- manders Pseudotriton and Gyrinophilus. Copeia, 1959(1): 60-63. Hairston, N. G. 1949. The local distribution and ecology of the plethodontid salamanders of the southern Appalachians. Ecol. Monog., 19: 47-73. Hallowell, E. 1856. Description of several species of Urodela, with remarks on the geographical distribu- tion of the Caducibranchiate division of these animals and their classification. Proc. Acad. Nat. Sci., Philadelphia, 8: 6-11. Hansen, A. M. and W. W. Tanner 1958. A comparative osteological study of certain species belonging to the genus Bolitoglossa (Amphibia). Great Basin Nat., 18(3-4): 85-100. Hecht, M. 1963. A reevaluation of the early history of the frogs. Part II. Systematic Zool., 12: 20-35. Hendrickson, J. R. 1954. Ecology and systematics of the salamanders of the genus Batrachoseps. Univ. California Publ. Zool., 54(1): 1-46. Herre, W. 1935. Die Schwanzlurche der mitteleocanen (ober- lutetischen) Braunkohle des Geiseltales und die Phylogenie der Urodelen unter Ein- schluss der fossilen Formen. Zoologica, 33 (87): 1-85. 1950. Schwanzlurche aus dem Paleocan von Wal- beck. In Neue Ergebnisse und Probleme der Zoologie (Klatt-Festschrift). Leipzig: Geest und Portig, 286-301. Herre, W. and H. Lunau 1950. Neue fossile Schwanzlurche aus dem Burdi- galium. Neuen Jahrb. Geol. Palaon., 8: 247- 252. Highton, R. 1961. A new genus of lungless salamanders from the Coastal Plain of Alabama. Copeia, 1961(1): 65-68. 1966 EVOLUTION OF LUNGLESS SALAMANDERS 107 1962. Revision of North American salamanders of the genus Plethodon. Bull. Florida State Mus., Biol. Sci., 6(3) : 235-367. Hilton, W. A. 1945a. The skeleton of Batrachoseps. J. Ent. Zool., 37(1): 8-17. 1945b. The skeletons of Aneides, Plethodon, Eurycea, Manculus qucidridigitatus, Hemi- dactylium, Ensatina, Gyrinophilus, and Pseudotriton. J. Ent. Zool., 37(2) : 43-68. 1945c. The skeletons of Desmognathus, Leurogna- thus, Typhlotriton, Stereochilus, and Ple- thopsis. J. Ent. Zool., 37(3): 75-84. 1945d. The skeletons of Hydromantes, Typhlo- molge, and Haideotriton. J. Ent. Zool., 37 (4): 98-102. 1946. Skeletons of Mexican and Central American salamanders of the family Plethodontidae. J. Ent. Zool., 38(1): 1-8. 1947a. The hyobranchial skeleton of Plethodon- tidae. Herpetologica, 3(6): 191-194. 1947b. Lateral line sense organs in salamanders. Bull. So. Calif. Acad. Sci., 46(3): 97-110. 1948. The vertebrae of salamanders. J. Ent. Zool., 40(3): 47-65. 1950. Dermal sense organs of Stereochilus. Her- petologica, 6(5) : 135-136. 1959. The tongue in Urodela. Herpetologica, 15 (3) : 137-140. Holman, J. A. 1958. The Pleistocene herpetofauna of Sabertooth Cave, Citrus County, Florida. Copeia, 1958 (4) : 276-280. 1959a. Amphibians and reptiles from the Pleis- tocene (Illinoian) of Williston, Florida. Copeia, 1959(2): 96-102. 1959b. A Pleistocene herpetofauna near Orange Lake, Florida. Herpetologica, 15(3): 121- 125. Humphrey, R. R. 1922. The multiple testis in urodeles. Biol. Bull., 43(1): 45-67. Hutchison, V. H. 1958. The distribution and ecology of the cave salamander, Eurycea lucifuga. Ecol. Monog., 28: 1-20. Ingram, V. M. 1961. Gene evolution and the haemoglobins. Na- ture, 189: 704-708. Jarvik, E. 1963. The composition of the intermandibular di- vision of the head in fish and tetrapods and the diphyletic origin of the tetrapod tongue. Kungl. Svenska Vetenskapakad. Hand., 9: 5-74. Joubert, P. J. 1961. Contributions to the cranial morphology of Pseudotriton ruber ruber (Sonnini). Ann. Univ. Stellenbosch, 36(8): 389-418. Kingsbury, B. F. and H. D. Reed 1909. The columella auris in Amphibia. J. Morph., 20(4): 549-628. Larsen, J. H., Jr. 1963. The cranial osteology of neotenic and trans- formed salamanders and its bearing on in- terfamilial relationships. Ph. D. Thesis, Uni- versity of Washington, Seattle, 205 pp. Laurent, R. 1947. La disposition des dents vomeriennes chez les urodeles superieurs et son importance phylogenetique. Bull. Mus. Roy. Hist. Nat. Belgique, 23(11): 1-4. Lebednika, N. S. 1964. Development of the nasal bones in the caudate amphibia. Doklady Akademii Nauk SSSR, 159: 219-222. Lowe, C. H., Jr. 1950. The systematic status of the salamander Plethodon hardii, with a discussion of bio- geographical problems in Aneides. Copeia, 1950(2): 92-99. Macginitie, H. D. 1958. Climate since the late Cretaceous. In Hubbs, C. L. (ed.), Zoogeography. Washington: Amer. Assn. Adv. Sci., publ. 51(2): 61-79. 1962. The Kilgore flora: a late Miocene flora from northern Nebraska. Univ. California Publ. Geol. Sci., 35(2): 67-158. Manwell, C. 1963. The blood proteins of cyclostomes. A study in phylogenetic and ontogenetic biochemis- try. In Brodal, A. and Fange, R. (eds), The Biology of Myxine. Oslo: Scandinavian University Books, 372-455. Martin, P. S. 1958a. A biogeography of reptiles and amphibians in the Gomez Farias region, Tamaulipas, Mexico. Misc. Publ. Mus. Zool. Univ. Mich- igan, 101: 1-102. 1958b. Pleistocene ecology and biogeography of North America. In Hubbs, C. L. (ed.), Zoo- geography. Washington: Amer. Assn. Adv. Sci., publ. 51(15): 375-420. 108 MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 Martin. P. S. and B. E. Harrell 1957. The Pleistocene history of temperate biotas in Mexico and eastern United States. Ecol- ogy, 38(3): 468-480. Martof, B. S. 1962. Some aspects of the life history and ecology of the salamander Leurognathus. Amer. Midi. Nat., 67(1): 1-35. Martof, B. S. and F. L. Rose 1962. The comparative osteology of the anterior cranial elements of the salamanders Gyrino- philus and Pseudotriton. Copeia, 1962(4): 727-732. Mayr, E. 1962. Animal species and evolution. Cambridge, Mass.: Harvard University Press, 797 pp. Mitchell, R. W. and J. R. Reddell 1965. Eurycea tridentifera, a new species of trog- lobitic salamander from Texas and a re- classification of Typhlomolge rathbuni. Texas J. Sci., 17: 12-27. Monath, T. 1965. The opercular apparatus of salamanders. J. Morph. 116: 149-170. Moore, J. P. 1899. Leurognathus marmorata, a new genus and species of salamander of the family Des- mognathidae. Proc. Acad. Nat. Sci., Phila- delphia, 1899: 316-323. 1900. Post-larval changes in the vertebral articula- tions of Spelerpes and other salamanders. Proc. Acad. Nat. Sci., Philadelphia, 1900: 613-622. Murray, K. F. 1957. Pleistocene climate and the fauna of Burnet Cave, New Mexico. Ecology, 38(1): 129- 131. Niceforo Maria 1958. Section Herpetologia. Bol. Inst. La Salle, Bogota, 45(198): 1-16. Noble, G. K. 1921. The anterior cranial elements of Oedipus and certain other salamanders. Bull. Amer. Mus. Nat. Hist., 44(1): 1-6. 1925. An outline of the relation of ontogeny to phylogeny within the Amphibia II. Amer. Mus. Novit., 166: 1-10. 1927a. The plethodontid salamanders; some as- pects of their evolution. Amer. Mus. Novit., 249: 1-26. 1927b. The value of life history data in the study of the evolution of the Amphibia. Ann. New York Acad. Sci., 30: 31-128. 1931. The biology of the Amphibia. New York: McGraw-Hill, 577 pp. Organ, J. A. 1961a. Studies of the local distribution, life his- tory, and population dynamics of the sala- mander genus Desmognathus in Virginia. Ecol. Monog., 31: 189-220. 1961b. Life history of the pigmy salamander, Des- mognathus wrighti, in Virginia. Amer. Midi. Nat., 66(2): 384-390. Parker, W. K. 1877. On the structure and development of the skull in the urodelous Amphibia. Part I. Philos. Trans. Roy. Soc. London, 167(2): 529-597. 1882. On the morphology of the skull in the Am- phibia Urodela. Trans. Linn. Soc. London, Zoology, ser. 2, 2(3): 165-212. Parrington, F. R. 1956. The patterns of dermal bones in primitive vertebrates. Proc. Zool. Soc. London, 127 (3): 389-411. Parsons, T. S. and E. E. Williams 1962. The teeth of Amphibia and their relation to amphibian phylogeny. J. Morph., 110: 375- 390. Peabody, F. S. 1959. Trackways of living and fossil salamanders. Univ. California Publ. Zool., 63(1) : 1-72. Peabody, F. S. and J. M. Savage 1958. Evolution of a Coast Range corridor in California and its effect on the origin and dispersal of living amphibians and reptiles. In Hubbs, C. L. ( ed. ) , Zoogeography. Wash- ington: Amer. Assn. Adv. Sci., publ. 51(8) : 159-186. Peters, W. 1863. Fernere Mittheilungen fiber neue Batrachier. Monats. Konig. Preuss Akad. Wiss., Berlin, 1863: 445-470. Piatt, J. 1935. A comparative study of the hyobranchial apparatus and throat musculature in the Plethodontidae. J. Morph., 57(1): 213-251. Pope, C. H. and S. H. Pope 1951. A study of the salamander Plethodon oua- chitae and the description of an allied form. Bull. Chicago Acad. Sci., 9: 129-152. 1966 EVOLUTION OF LUNG LESS SALAMANDERS 109 Rabb, G. B. 1955. A new salamander of the genus Parvimolge from Mexico. Breviora, Mus. Comp. Zool., Harvard Coll., 42: 1-9. 1956. A new plethodontid salamander from Nuevo Leon, Mexico. Fieldiana-Zoology, Chicago, 39(3): 11-20. 1958. On certain salamanders of the plethodontid genus Chiropterotriton. Occ. Pap. Mus. Zool. Univ. Michigan, 587: 1-37. 1960. A new salamander of the genus Chiroptero- triton from Chiapas, Mexico, with notes on related species. Copeia, 1960(4): 304-311. Reed, H. D. 1920. The morphology of the sound-transmitting apparatus in caudate Amphibia and its phy- logenetic significance. J. Morph., 33(2): 325-387. Rensch, B. 1954. Die phylogenetische Abwandlung der On- togenese. In Heberer, G. (ed.), Die Evolu- tion der Organismen. Stuttgart: Fischer, 103-130. Rose, F. L. and F. M. Bush 1963. A new species of Eurycea (Amphibia: Cau- data) from the southeastern United States. Tulane Stud. Zool., 10(2): 121-128. Sarasin, P. and F. Sarasin 1887- 1890. Zur Entwicklungsgeschichte und Anatomie der ceylonesischen Blindwiihle Ichthyophis glutinosus. In Ergebnisse Naturwissenschaft- licher Forschungen auf Ceylon. Wiesbaden: Kreidels, 2: 153-263. Savage, J. M. 1960. Evolution of a peninsular herpetofauna. Systematic Zool., 9(3): 184-212. Schmalhausen, 1. 1. 1949. Factors of Evolution (American transla- tion). Philadelphia and Toronto: Blakiston, xiv + 327 pp. Schmidt, K. P. 1936. Guatemalan salamanders of the genus Oedipus. Zool. Ser. Field Mus. Nat. Hist., Chicago, 20(17): 135-166. 1946. On the zoogeography of the Holarctic re- gion. Copeia, 1946(3): 144-152. 1953. A check list of North American amphibians and reptiles, ed. 6. Chicago: Amer. Soc. Ichthy. Herpt., 280 pp. Schwartz, A. and R. E. Etheridge 1954. New and additional records from the North Carolina Coastal Plain. Herpetologica, 10 (3): 167-171. Shotwell, J. A. 1961. Late Tertiary biogeography of horses in the northern Great Basin. J. Paleon. 35(1): 203-217. 1963. Pliocene mammalian communities of the Juntura Basin. In Shotwell, J. A., R. G. Bowen, W. L. Gray, D. C. Gregory, D. E. Russell, D. Taylor, The Juntura Basin: studies in earth history and paleoecology. Trans. Amer. Philos. Soc., new ser., 53(1) : 7-21. Simpson, G. G. 1947. Evolution, interchange, and resemblance of the North American and Eurasian Cenozoic mammalian faunas. Evolution, 1(3): 218- 220. 1953. The major features of evolution. New York: Columbia Univ. Press, xx + 434 pp. Smith, H. M. and E. H. Taylor 1948. An annotated checklist and key to the am- phibia of Mexico. Bull. U.S. Natl. Mus., 194: 1-118. Smith, L. 1920. The hyobranchial apparatus of Spelerpes bislineatus. J. Morph., 33: 527-583. Smith, P. W. 1957. An analysis of post-Wisconsin biogeography of the prairie peninsula region based on dis- tributional phenomena among terrestrial vertebrate populations. Ecology, 38(2): 205-218. 1961. The amphibians and reptiles of Illinois. Bull. Illinois Nat. Hist. Surv., 28(1): 1-298. Soler, E. I. 1950. On the status of the family Desmognathidae. Univ. Kansas Sci. Bull., vol. 33, pt. 2(12) : 459-480. Stebbins, R. C. 1951. Amphibians of western North America. Berkeley and Los Angeles: Univ. California Press. 539 pp. 1954. Natural history of the salamanders of the plethodontid genus Ensatina. Univ. Califor- nia Publ. Zool., 54(2) : 47-124. Stebbins, R. C. and C. H. Lowe, Jr. 1949. The systematic status of Plethopsis with a discussion of speciation in the genus Botra- choseps. Copeia, 1949(2): 116-129. no MEMOIR OF THE SOUTHERN CALIFORNIA ACADEMY OF SCIENCES VOL. 4 Stejneger, L. H. 1892. Preliminary description of a new genus and species of blind cave salamander from North America. Proc. U.S. Natl. Mus., 15: 115-117. 1907. Herpetology of Japan and adjacent terri- tory. Bull. U.S. Natl. Mus., 58: 1-577. Stejneger, L. H. and T. Barbour 1917. A check list of North American amphibians and reptiles, ed. 1. Cambridge: Harvard Univ. Press, 125 pp. Storer, T. I. 1925. A synopsis of the Amphibia of California. Univ. California Publ. Zool., 27: 1-342. Stuart, L. C. 1935. A contribution to a knowledge of the her- petology of a portion of the savanna region of central Peten, Guatemala. Misc. Publ. Mus. Zool. Univ. Michigan, 29: 1-56. 1943. Taxonomic and geographic comments on Guatemalan salamanders of the genus Oedi- pus. Misc. Publ. Mus. Zool. Univ. Michigan, 56: 1-33. 1948. The amphibians and reptiles of Alta Vera- paz, Guatemala. Misc. Publ. Mus. Zool. Univ. Michigan, 69: 1-109. 1950. A geographic study of the herpetofauna of Alta Verapaz, Guatemala. Contrib. Lab. Vert. Biol. Univ. Michigan, 45: 1-77. 1951. The herpetofauna of the Guatemalan Pla- teau, with special reference to its distribu- tion on the southwestern highlands. Contrib. Lab. Vert. Biol. Univ. Michigan, 49: 1-71. 1954. Herpetofauna of the southeastern highlands of Guatemala. Contrib. Lab. Vert. Biol. Univ. Michigan, 68: 1-65. 1957. Herpetofaunal dispersal routes through northern Central America. Copeia, 1957 (2): 89-94. 1958. A study of the herpetofauna of the Uaxac- tun-Tikal area of northern El Peten, Guate- mala. Contrib. Lab. Vert. Biol. Univ. Michi- gan, 75: 1-30. Tanner, W. W. 1950. A new genus of plethodontid salamander from Mexico. Great Basin Nat., 10: 37-44. 1952. A comparative study of the throat muscula- ture of the Plethodontidae of Mexico and Central America. Univ. Kansas Sci. Bull., 34, pt. 2(10): 583-677. Taylor, E. H. 1940. New salamanders from Mexico, with a dis- cussion of certain known forms. Univ. Kansas Sci. Bull., 26(12): 407-439. 1944. The genera of plethodont salamanders in Mexico, pt. I. Univ. Kansas Sci. Bull., 30, pt. 1(12): 189-232. 1949. New salamanders from Costa Rica. Univ. Kansas Sci. Bull., 33: 279-288. Teege, M-J. Laurinat 1956. Studien zur Entwicklung und Gestalt der Urodelenwirbel. Zeit. wissen. Zoologie, 160: 95-163. Thornbury, W. D. 1954. Principles of Geomorphology. New York and London: Wiley, 618 pp. Uzzell, T. M., Jr. 1961. Calcified hyoid and mesopodial elements of plethodontid salamanders. Copeia, 1961(1) : 78-86. Valentine, B. D. 1963a. Notes on the early life history of the Ala- bama salamander, Desmognathus aeneus chermocki Bishop and Valentine. Amer. Midi. Nat., 69(1): 182-188. 1963b. The plethodontid salamander Phaeogna- thus: external morphology and zoogeogra- phy. Proc. Biol. Soc. Washington, 76: 153- 158. 1963c. The plethodontid salamander Phaeogna- thus : collecting techniques and habits. J. Ohio Herpt. Soc., 4(1-2): 49-54. Waddington, C. H. 1956. Genetic assimilation of the bithorax pheno- type. Evolution, 10: 1-13. Wahlert, G. von 1957. Biogeographische und Oekologische Tatsa- chen zur phylogenie amerikanischer Schwanzlurche. Zool. Jb. Syst., 85(3): 253- 282. Wake, D. B. 1960. The osteology and relationships of the sala- mander genus Aneides Baird. M. Sc. Thesis, University of Southern California, Los An- geles, 181 pp. 1963. Comparative osteology of the plethodontid salamander genus Aneides. J. Morph., 113 (1): 77-118. Wake, D. B. and A. H. Brame, Jr. 1963a. The status of the plethodontid salamander genera Bolitoglossa and Magnadigita. Copeia, 1963(2): 382-387. 1963b. A new species of Costa Rican salamander, genus Bolitoglossa. Rev. Biol. Trop., 11(1): 63-73. 1966 EVOLUTION OF LUNG LESS SALAMANDERS 111 1966. Notes on South American salamanders, genus Bolitoglossa. Copeia, 1966(2): 360-363. Wiedersheim, R. 1875. Salamandrinci perspicillata und Geotriton fuscus. Versuch einer vergleichenden Anat- omie du Salamandriden. Genua: Instituts der Sordo-Muti, 207 pp. 1877. Das Kopfskelet der Urodelen. Gegen. Morph. Jb., 3: 352-548. Wilder, H. H. 1894. Lungenlosen Salamandriden. Anat. Anz., 9: 216-220. Wilder, I. L. Whipple 1913. The life history of Desmognathus fusca. Biol. Bull., 24(5): 293-343. 1920. The urodele vomer. Anat. Rec., 18: 349. 1924. Variations in the premaxillary of Eurycea bislineata. Amer. Nat., 58: 538-543. 1925. The morphology of amphibian metamor- phosis. Northampton, Mass.: Smith College, xviii -f- 161 pp. Wilder, I. L. Whipple and E. R. Dunn 1920. The correlation of lunglessness in salaman- ders with a mountain brook habitat. Copeia, 84: 63-68.