The skull and jaw musculature as guides to the ancestry of salamanders

Transcription

1 Zoological Journalofthe Linnean Society, 68: With 29 figures January 1980 The skull and jaw musculature as guides to the ancestry of salamanders ROBERT L. CARROLL AND ROBERT HOLMES Redpath Museam, M~Gill University, M~~trea~, Canada AccepledJor publication October 1978 ~I~lic lossil record provides no evidence supporting a unique common ancestry for frogs, salamanders arid apudans. The ancestors of the modern orders may have diverged trom one another as recently as 250 inillion years ago, or as long ago as 400 million years according to current theories of various authors. In order to evaluate the evolutiohary patterns of the modern orders it is necessary to determine whether their last common ancestor was a rhipidistian fish, a very primitive amphibian, a labyrinthodont or a lissamphibian. The broad cranial similarities of frogs and salamanders, especially the dominance of the braincase as a supporting element, can be associated with the small size of the skull in their immediate ancestors. Hynobiids show the most primitive cranial pattern known among the living salamander families andprovide a model for determining the nature of the.~~icestors of the entire order. Features expected in ancestral salamanders include: ( 1) Emargination 01 the cheek; (2) Movable suspensorium formed by the quadrate, squamosal and pterygoid; (3) Occipital condyle posterior to jaw articulation; (4) Distinct prootic and opisthotic; (5) Absence (11 otic notch; (6) Stapes forming a structural link between braincase and cheek. In the otic region, cheek and jaw suspension, the primitive salamander patternkesembles most closely the microsaurs among known Paleozoic amphibians, and shows no significant features in common with either ailcestral tkogs or the majority of labyrinthodonts. The basic pattern of the adductor jaw musculature is consistent within both frogs and salamanders, but major differences are evident between the two groups. The dominance of the adductor mandibulae externus in salamanders can be associated with the open cheek in all members of that order, and the small size of this musck in frogs can be associated with the large otic notch. The spread of different muscles over the otic capsule, the longus liead ol the adductor mandibulae posterior in frogs and the superficial head of the adductor inandibulae internus in salamanders, indicates that fenestration of the skull posterodorsal to the orbit occurred separately in the ancestors of the two groups. Reconstruction of the probable pattern 01 the jaw inusculature in Paleozoic amphibians indicates that frogs and salamanders might have evolved from a condition hypothesized for primitive labyrinthodonts, but the presence of a large otic iiutcli in dissorophids suggests specialization toward the anuran, not the urodele condition. The prcwnce ot either an einarginated cheek or an embayment of the lateral surface of the dentary and the absence 01 an otic notch in microsaurs indicate a salamander-like distribution of the adductor jaw muscles. The ancestors of frogs and salamanders probably diverged from one another in the early Cdmiiiterous, Frogs later evolved from small lpbyrinthodonts and salamanders from microsaurs. Feature, coilsidered typical of lissamphibians evolved separately in the two groups in the late Permian ailti.ri-iassic. KEY WORDS:- urodeles -jaw muscles - Paleozoic amphibians - modern amphibians - cranial iiiorphology - phylogeny - microsaws - labyrinthodonts. CONTENTS Introduction Cranial comparisons /80/ /$02.00/ The Linnean Society of London 1

2 2 R. L. CARROLL AND R. HOLMES Adductor jaw inuscles Adductorjawinusclesofurodeles..... Adductorinandibulaeexternus..... Adductor inandibulae posterior..... Adductormandibulaeinternus..... Adductorjaw inusclesofanurans..... Adductorinandibulaeexternus..... Adductor inandibulae posterior..... Adductorinandibulaeinternus..... Adductor jaw inusculature in Paleozoic amphibians Pedicellatr teeth Conclusions Atltleiitluin Acknowledgements References Abbreviations used in figures Abbreviations used for muscles I NTRO DUCT1 0 N Despite numerous recent publications concerning the origin and relationships of the modern amphibian orders-eaton (19591, Szarski (19621, Reig (19641, Estes (19651, Cox (1967), Shishkin (1970), Jurgens (197 1) and most influentially, Parsons & Williams (1963) - little can be said with confidence concerning the specific relationships of frogs, salamanders and apodans. Following the revival of the Lissamphibia concept by Parsons 8c Williams, many authors have accepted the assumption that the three orders are closely related, having descended from a common ancestor, itself distinguishable from other, more primitive, amphibians. Fossil evidence is not yet available that conclusively supports either this view or the alternative possibility of a distinct origin of the three groups. In view of the extensive work currently being undertaken on the behaviour and physiology of modern amphibians, it would be extremely useful to know the nature of their relationships more specifically. Were the characteristics by which we recognize frogs, salamanders and apodans largely achieved separately, or are they the result of inheritance from a common ancestor that would be readily recognizable as such? According to current theories of the origin of modern amphibians, the initial divergence of the three orders may have occurred as long as 400 million years ago or as recently as 250 million years ago. Jarvik (1942, 1954, 1963) has argued that anurans and urodeles evolved ultimately from different groups of rhipidistian fishes, already distinct at the base of the Devonian. The time of initial divergence may be assumed to have been late Silurian. Romer ( 19451, and many others previously, had suggested that modern amphibians had evolved from different groups of Paleozoic amphibians : frogs from labyrinthodonts and salamanders from lepospondyls. Such ancestry would suggest an initial divergence in the Lower Carboniferous or possibly Upper Devonian - approximately 340 million years ago. Noble (19311, Estes (1965) and others have proposed an origin of both frogs and salamanders from among small Paleozoic labyrinthodonts. If such a common origin does not imply a specificially lissamphibian ancestor, the point of divergence might have been about 300 million years ago. Parsons 8c Williams ( 1963) argued that all modern amphibians evolved from a common ancestor already specialized beyond the level of any know group of Paleozoic amphibians. If the Lower Permian and Carboniferous

3 ANCt5 I K? OF SALAMANDERS 62 I z 2 LL z W a - m x z 2 z 0 > w a z 4 - LL 12 m.(a Triadobatrachus Do1 eserpeton plausible ancestor of lissamphibians). Period during which microsaurs may have diverged from ancestors of labyrinthodonts. I I Occurence of distinct lineages leading to frogs and salamanders according to Jarvik. Period during which urodeles and frogs might Ilave diverged from common 1 i ssamphi bian ancestor. yeriod of time during which urodele ancestors might have diverged from - frog ancestors if lissamphibian hypothesis is incorrect Time scale in millions of years, from Romer Figut-e 1. Geological ranges of modern amphibian orders and possible times of divergence of the dncestorb ( ~ frogs t and salamanders according to various theories. A recently described larval salamander from Russia (Ivakhnenko, 1978) extends the range of urodeles back to the late Triassic. do provide a good record of the types of amphibians living during this time (Carroll, 19771, it is probable that a lissamphibian ancestor such as Parsons 8c Williams proposed evolved subsequent to about 250 million years ago. The possible time of divergence differs by approximately 150 million years according to the various theories (Fig. 1). Frogs and salamanders did unquestionably share a single common ancestor at some stage. Difficulties arise in establishing the probable time of original divergence, and in determining to what extent the anatomy of the ancestral group is shared by its living descendants. Our appreciation of the evolutionary position of frogs and salamanders would presumably differ considerably

4 4 R. L. CARROLL AND R. HOLMES depending on whether their last common ancestor was a rhipidistian fish, a very primitive Paleozoic amphibian, a labyrinthodont, or a lissamphibian. The earliest known frog, the LowerJurassicgenus Vieraellu (Estes 8c Reig, 1973) is known from a single, nearly complete skeleton which can be included in the most primitive family of living frogs, the Ascaphidae. All of the major groups of frogs may have differentiated by the end of the Jurassic. The only known fossil apodan is represented by a single vertebra from the Paleocene (Estes 8c Wake, 1973). It is placed in a distinct genus, but is closely related to the living forms. The earliest remains of salamanders consist of a cervical vertebra from the Middle Jurassic (Seiffert, 1969 : Estes 8c Hoffstetter, 1976) a complete and beautifidly preserved skeleton from the Upper Jurassic of Russia, a tiny lawal stage from the Upper Triassic (Ivakhnenko,I978) and a femur from the Upper Jurassic (Hecht 8c Estes, 1960). Fossils from the Cretaceous include some apparently primitive genera, such as Romonellus (Nevo 8c Estes, 1969j and Albanerpeton (Estes 8c Hoffstetter, 1976) and others (Estes, 1965, 1969 c, 1975) that clearly demonstrate the presence ofmodern families. The early fossils of frogs, salamanders and apodans provide little more anatomical information regarding the ultimate origin of these groups or their degree of interrelationship than do living genera. They simply establish a minimum time for the achievement of the definitive features of each group. In the absence of any fossils definitely linking the three modern orders with one another, it is necessary to appraise our knowledge of the origin of each group separately. The only fossil that establishes any connection between modern and Paleozoic amphibians is the genus Triadobatrachus [Protobutrachus] from the Lower Triassic of Madagascar (Piveteau, 1937). Estes 8c Reig (1973) in the most recent appraisal of the origin of frogs, accept this genus as belonging to the group from which the modern order evolved. Watson (1940) had previously stressed the affinities of Triadobatrachus with labyrinthodont amphibians. The discovery of a labyrinthodont amphibian with pedicellate teeth (Bolt, 1969, 1977) further supports the contention that this group may include the ancestors of frogs. It has recently been suggested that apodans might have evolved from a group of elongate, burrowing amphibians such as the lepospondyl microsaur family Goniorhynchidae (Carroll 8c Currie, 1975). Such an origin has not, however, been confirmed by the discovery of any genera intermediate in morphology between Paleozoic microsaurs and Tertiary apodans. There are no fossils linking salamanders with any of the Paleozoic amphibians, but arguments have been made for their relationship with both lepospondyls (Shishkin, 1970) and with labyrinthodonts (Estes, 1965). Derivation from labyrinthodonts implies relatively closer affinities with frogs (presuming frogs to be of labyrinthodont origins), although it does not necessitate a common lissamphibian ancestry. The ancestry of salamanders and the nature of their relationship with frogs remain among the most important unsolved questions associated with the problem of the affinities of the modern amphibian orders. Although continued reference to frogs will be made, the major effort of this paper is to determine what information the skull and jaw muscles of living salamanders may offer regarding the probable origin of urodeles. Although knowledge of all aspects of anatomy, physiology and behaviour would be useful in evaluating the degree of relationship of the modern amphibians, little more than the skeleton is preserved in fossil forms, and final assessment of ancestry must be determined primarily on the basis of that system.

5 ANCESTRY OF SALAMANDERS Postcranially, the dif erences between frogs and salamanders are very great, reflecting strong habitus divergence whatever the ultimate relationship. Only the general nature of the vertebrae suggests a common pattern (Wake, 1970). In contrast, the skulls of frogs and salamanders are similar, at least in so far as they differ from those of other vertebrate groups. There are, however, some consistent differences which suggest that further study would be profitable. It is surprising that relatively little attention has been paid to the question of the origin of the skull pattern in either frogs or salamanders, aside from assuming general reduction in ossification from that of Paleozoic amphibians. The structure of the skull, particularly of the temporal region and the area of the jaw musculature, has proved of extreme importance in elucidating the pattern of evolution in bony fishes (Schaeffer 8c Rosen, 19611, reptiles (Romer, 1956) and the origin of mammals (Barghusen, 1972; Crompton & Parker, 1978). Because of the readily visualized importance of this area, and the generally conservative nature of the jaw musculature, it is probable that study of this system would also cast some light on the origin of the modern amphibian groups. 5 CRANIAL COMPARISONS There are several significant features of the general cranial anatomy that unite frogs and salamanders and distinguish them from all Paleozoic amphibians. The primitive amphibian skull (Fig. 21, like that of primitive reptiles and early bony fish, is composed of a nearly continuous covering of dermal bone, with only orbital, narial and pineal openings dorsally, and the internal nares and openings for the adductor jaw musculature on the palatal surface. Inside the dermal box is suspended the endochondral braincase. In frogs and salamanders, the dermal ossification of the skull roof and palate is much reduced, and the braincase, which incorporates dermal bone along the midline, forms a major structural element of the skull. It serves as a longitudinal beam, extending from the occiput to the snout, with the otic capsules extending laterally to support the cheek dorsally and the pterygoid ventrally. The relatively large size of the braincase and its structural significance can be associated with the absolutely small size of the skull. It is almost certain that the initiation of the cranial patterns of frogs and salamanders was associated with small skull size. The few larger members of these orders are almost certainly specialized in this feature. The minimum functional size of the semicircular canals (Jones & Spells, 1963) and the brainstem require that the braincase occupy much of the skull in small vertebrates. With a skull less than about 2 cm in length and width, the otic capsules extend close to the cheeks and the braincase approaches the snout anteriorly. Such relative expansion of the braincase compared with the remainder of the skull permits a different pattern of the surrounding support structures than that seen in most Paleozoic amphibians. Another aspect of small skull size is the relative expansion of the orbits and nasal capsules. Similarities of the skull in the ancestors of frogs and salamanders that can be attributed to small size do not, of themselves, imply common ancestry, since grossly similar patterns might have evolved in descendants of different Paleozoic amphibian groups with primitively solidly roofed skulls. Watson ( 1940), for example, illustrated both labyrinthodonts and a lepospondyl (termed Hylonomus geinitzi) as possible anuran ancestors.

6 6 R. L. CARROLL AND R. HOLMES A L D Figure 2. The skulls of Paleozoic and modern amphibians. A. Rana catesbeiana. 8. Ambystoma manrlatum. C. The labyrinthodont Dendrerpelon. D. The lepospondyl microsaur Asaphestra. Although both ma'or groups of Paleozoic amphibians exhibit considerable diversity, these genera are representative oflthe basic patterns from which frogs and salamanders may ultimately have evolved. Other features that are common to primitive frogs and salamanders include a movable articulation between the base of the braincase and the pterygoid, and large palatal vacuities. These characteristics are known among both labyrinthodont and lepospondyl amphibians. Within this general pattern, common to the two groups, there are also consistent differences that distinguish the skulls of all urodeles from those of anurans. In both groups there is considerable diversity. This is particularly true of the frogs (Trueb, 1973). The fossil record of frogs (Estes 8c Reig, 1973) does, however, provide a reasonable basis for establishing the primitive pattern for the group. The fossil record of salamanders does not give much evidence as to the primitive skull pattern of the group, but the anatomy of the living families is sufficiently conservative to suggest a similar basic pattern for the common ancestral stock. Frogs (Figs 2, 31, including the ancestral form Triadobatruchus (Fig. 27), in so far as it is known, have similar cranial specializations. All have a fused frontoparietal

7 ANCES rky OF SALAMANDERS a Figure 3. Representative skulls of the most primitive frog family Ascaphidae. A. The living genus Ascaphus in dorsal, palatal, lateral and occipital views. B. Vieradla, the oldest known frog, Lower Jurassic (From Estes & Reig, 1973). In Ascaphus the otico-occipital region is not fully ossified and tlii-w distinct centers of ossification are evident. This genus lacks the stapes, the tympanum and the iiiiddle ear cavity. In these features it is specialized over the condition seen in Jurassic ascaphids. bone and a well defined otic notch formed by the squamosal. The palatal margin of the skull is nearly always continuous, with the quadrate characteristically joined to the maxilla by a quadratojugal. The quadrate is characteristically at the posterior margin of the skull and the gape of the jaw (primitively defined by the length of the tooth row) is very great, with the jaw angle behind the rear margin of the orbit. The otic capsule is always ossified as a single unit in the adult. The stapes is a rod-shaped structure, characteristically activated by a tympanum. As in salamanders, there is frequently a second ear ossicle, the operculum, making up a portion of the otic capsule adjacent to the base of the stapes. Among salamanders (Figs 4-12), there is never an otic notch and the frontals and parietals are not fused. The jaw articulation is generally well anterior to the level of the occipital condyle and the gape of the jaw and the tooth row are comparatively short. Of particular importance is the cheek region. There is only rarely an independent quadratojugal, and there is always a gap between the quadrate and the maxilla. (A quadratojugal is present in TyEototrzton and the closely related Oligo-Miocene salamandrid Chelotriton (Estes, pers. comm.). This bone ossifies late in ontogeny and its general absence in urodeles may be

8 8 R. L. CARROLL AND R. HOLMES attributed to paedomorphism.) Among some hynobiids, ambystomatids and proteids, the otic capsule is made up of distinct opisthotic and prootic ossifications. In other genera, it is ossified as a unit. The stapes, in contrast to that of frogs, has a very large foot-plate and a short stem. There is never a middle ear cavity or tympanum. The similarities and differences between the skulls of frogs and salamanders may be interpreted as resulting from simple divergence from a common ancestral pattern, or as resulting from a long period of convergence and parallelism from clearly distinct ancestral stocks. Current evidence strongly suggests evolution of the frog cranial pattern from that of small labyrinthodonts such as the dissorophid genus Doleserpeton (Bolt, 1977). The question of salamander ancestry can be evaluated by considering whether or not their cranial anatomy supports derivation from the same stock. In order to investigate this problem in detail, it is necessary to determine the primitive cranial configuration for salamanders as a group. Unfortunately, the fossil record of urodele skulls is very incomplete, and most adequately preserved specimens from the Mesozoic show specific similarities only with the more specialized modern families - Sirenidae, Proteidae and Amphiumidae. It is commonly accepted that hynobiids are the most primitive of living salamanders in terms of both their general anatomy and biology (Hecht & Edwards, 1977). Examination of a number of species (Fig. 4) shows a relatively constant structure that, compared with skulls of all other salamander families (fossil and living), appears closest to that seen in Paleozoic amphibians. The general configuration of the skull resembles that of small labyrinthodonts and lepospondyls in the large extent of dermal bone, with relatively limited dorsal exposure of the endochondral braincase. Of the bones present in most Paleozoic amphibians, only those at the posterior orbital margin (postfrontal, jugal, postorbital and quadratojugal) and at the posterior margin of the skull table (postparietal, tabular and supratemporal) have been lost, together with the ectopterygoid on the palatal surface. The palatine bone appears during development in most salamander groups, but is lost in most adults except sirenids, the axolotl and other neotenic forms. In hynobiid larvae (Fox, 1959) the palatine occupies a position comparable to that bone in most Paleozoic amphibians. As in several microsaurian lepospondyls (Carroll & Gaskill, 1978) and some labyrinthodonts (Romer, 1947) it bears a row of teeth parallel with those of the jaw margin. The parietal, frontal, prefrontal, nasal, lacrimal, premaxilla, septomaxilla, maxilla and vomer retain their primitive configuration and relationships. The pterygoid is small, but has a well-defined articulation with the base of the braincase. The base of the epipterygoid is closely integrated with the pterygoid, but in Hynobius naeviw, the stem retains its primitive character as a narrow rod, extending vertically alongside the braincase, just anterior to the prootic foramen (Fig. 4). The major differences from Paleozoic amphibians are in the area of the cheek and the posterior portion of the braincase. The squamosal is in contact with the parietal dorsally, without intervention of any of the primitive temporal series (tabular or supratemporal). The otic capsule underlies the squamosal anteriorly and together with the parietal forms a surface for its attachment which appears to allow the distal end of the squamosal to swing in a very restricted medio-lateral

9 ANCESTKY OF SALAMANDERS 9 A B P Figure 4. Representative skulls of the most primitive family of living salamanders, the Hynobiidae. A. Hynobivr lsvensrs (British Museum no ), skull roof and posterior palatal surface, bones of right suspensorium (squamosal, quadrate and pterygoid) are dotted. B. Batrachuperus sinasis (British Museum no ) skull roof and palate (lateral view of this specimen shown in Fig. 15 B). C. Hynobius naevius (Paris Museum no loti), skull in dorsal and lateral views and in sagittal section (occiput of this specimen shown in Fig. 15 E). Cartilage and connective tissue coarsely stippled.

10 10 R. L. CARROLL AND R. HOLMES I m Figure 5. Skull of Cryptobranchus allegheniewis in dorsal, palatal, lateral and occipital views. The cryp tobranchid pattern is generally considered to be directly derived from that of hynobiids. The skull is much inore open and flattened. The squamosal retains a broad contact with the parietal. Ventrally, the pterygoid has a very broad contact with the parasphenoid and sphenethmoid, precluding movement of the suspensoriurn. This family is neotenic and an operculum is not developed. arc. Anteroposterior movement would appear to be precluded by the hinge-like nature of the joint. In hynobiids, the squamosal forms a closely integrated unit with the quadrate and pterygoid. The pterygoid is free to move across the articulating surface of the basicraniurn as the squamosal swings on the parietal and prootic. On the basis of manipulation of dissected specimens and observations of dried skulls, it seems possible that the movement is limited to one or two millimetres, or a few degrees of arc. Confirmation of skull movement requires quantitative study of living material. Acceptance of this configuration of the squamosal as primitive for salamanders as a group is based on broad similarities of its shape and position relative to other bones in all salamander families. Some support for the assumption that there was a movable suspensorium in ancestral salamanders is provided by the presence in all salamanders, in contrast with frogs, of a large muscle, the interhyoideus posterior (Fig. 61, connecting the base of the jaw suspension with the midline and the gular fold. In addition to other functions, this muscle could act to adduct the suspensorium, an action very unlikely in frogs, in which mediolateral movement of the squamosal would interfere with the function of the middle ear. In species of Batrachuperus and Hynobius, the otic capsule is composed of two clearly separate ossifications. Anteriorly, the prootic appears to have incorporated the basisphenoid area of more primitive amphibians and like that unit, bears a well-developed facet for articulation with the pterygoid. Posteriorly,

11 ANCESTRY OF SALAMANDERS A Figure 6. A. Skull of Ambystoma maculutum in dorsal, palatal, lateral and occipital views. Amhystornatids are considered among the more advanced salamanders because of their practice of internal fertilization (in contrast with hynobiids and cryptobranchids), but the skull is basically similar tu that of hynobiids in the configuration and distribution of the individual bones. The otic capsule is typically ossified as a single unit, and the squamosal is not in contact with the parietal. The pterygoid typically retains a well developed area of articulation with the base of the otic capsule. B. Interniandibular musculature of Ambystoma tzgrinum (from Larsen & Guthrie, 1975). This pattern is characteristic of primarily terrestrial salamanders. The interhyoideus posterior is not present in adult murans, but is an important muscle in all urodeles. Abbreviations used only in this figure: ap, aponeurosis; gh, geniohyoideus; gm, genioglossus, medial division; ih, interhyoideus; im, intermandibularis posterior; ip, interhyoideus posterior. a single ossification incorporates areas that were more primitively recognizable as separate exoccipital and opisthotic. In the fusion of these elements, hynobiids might be considered advanced above the condition seen in living proteids (Fig. 9) where they are separate, although it is usually considered that proteids retain a larval condition rather than one that was expressed in primitive adult salamanders (see Hecht & Edwards, 1977 for discussion). Fox (1959) has demonstrated the incorporation of the exoccipitals into the opisthotic during development in Hynobius. As in frogs, the occipital condyles in all salamanders are formed by posterior projections of the exoccipitals, and the basioccipital (present in the majority of Paleozoic amphibians) does not ossify. Primitive

12 12 R. L. CARROLL AND R. HOLMES salamanders (Fox, 1959) retain a distinct XIIth cranial nerve, exiting through the exoccipital, as in most Paleozoic amphibians. As in all other salamanders, hynobiids have no tympanum or middle ear cavity. The stapes has a wide foot-plate and a short stem, attached by a ligament to a posterior process of the squamosal. Embryological evidence presented by Fox ( 1959) shows a close connection between the stapes and the palatoquadrate in a variety of salamanders. This may be interpreted as a retention of a primitive condition, such as that seen in rhipidistians and early reptiles (Heaton, 1978) in which the hyomandibular-stapes supports the braincase against the cheek. Structurally, the stapes continues to serve as a link between the cheek and the braincase in the adults of many salamander groups. Hynobiids show a variety of patterns in the area of the operculum (Monath, 1965). Salarnandrella keyserlingz and Onychodactylus japonicus have no muscular attachment from the fenestral plate to the shoulder girdle and no element that can be considered an operculum. Batrachuperus pinchonii lacks a distinct operculum, but has a cartilaginous area of the otic capsule ventral to the fenestra vestibuli to which is attached the levator scapulae (the opercularis muscle). In a variety of species of Hynobius, there is a distinct, but cartilaginous, operculum to which is attached the levator scapulae. Selection of certain hynobiids as representing the most primitive cranial configuration among living salamanders is based primarily on broad similarities with Paleozoic amphibians. It is supported by surprisingly strong similarities among some genera within the families Ambystomatidae and Salamandridae (Figs 6, 8) suggesting divergence from a close common ancestor. In most ambystomatids and salamandrids, the lacrimal is lost and the otic capsule is Figure 7. Skull of Phaeognathur hubrichti in dorsal and palatal views (from Wake, 1966). The plethodontids are the most numerous and varied of the modern salamander families. The pattern of the skull varies widely from genus to genus. It is generally accepted (Regal, 1966: Hecht & Edwards, 1977) that plethodontids were derived from the base of the ambystomatid stock. They show a continuation of trends evident between hynobiids and ambystomatids to reduce the extent of dermal ossification, with the otic capsule, in particular, forming a very substantial portion of the skull. Dorsally, the pattern of the suspensorium can be seen as an extension of that seen in ambystomatids, but ventrally it differs greatly in the loss of the pterygoid.

13 A oper W P C Figure 8. Skulls otsalamandridae. A. Sdamandra aka, from Wiedersheim (187 7). in dorsal and ventral views. In its general anatomy, the skull of Sdamandra atra resembles those of ambystomatids. According to Regal ( 1966) the embryological development of the vomer is fundamentally different, iridicatiiig d separate evolution from a hynobiid level. 8. Nolophthdmur urridescenr in dorsal, palatal, lateral and occipital views. Other members of the family Salamandridae show an increase in ossilication from that of Sdamandra (which is presumed to be primitive for the family on the basis of siniilarities with hynobiids). The frontal has grown back to reach the squamosal, forming a bar separating the posterior extent of the adductor mandibulae internus superficialis from the adductor riiandibulae externus. The suspensorium is immobilized. C. Tylototrilon, palatal view, from Noble (1951). Shows extension of the maxilla nearly to the quadrate. This condition was considered primitive by Noble, but in view of the secondary elaboration of the frontal and squamosal. the great length of the masilla is more likely to be a specialized feature. The specialized nature of the frontosquamosal arc is discussed by Naylor 11978).

14 14 R. L. CARROLL AND R. HOLMES A Figure 9. Representative skulls of the neotenic family Proteidae. Necturus and Proteus are among a number ofsalamanders that have a highly specialized skull and whose relationships with other families are uncertain. In the living genera, the maxillae are lost and functionally replaced by denticulate palatopterygoids. A. Necturus in dorsal, palatal, lateral and occipital views; this genus appears primitive in the retention of three distinct centers of ossification in the area of the otic-occipital: exoccipital, opisthotic and prootic. A distinct area of articulation is evident between the prootic and the palatopterygoid, but unlike the condition in hynobiids, ambystomatids and primitive salamandrids, the pterygoid is integrated with the remainder of the palate by an anterior contact with the vonier and parasphenoid; movement of the suspensorium may be possible, however. B. Opistotriton; this skull from the Paleocene shows general resemblances with Necturw (Estes, 1975) but is primitive in retaining contact of the parietal and squamosal above the otic capsule and in having a well developed maxilla. As in hynobiids and cryptobranchids, proteids retain a columellar process of the squamosal.

15 ANCESI KY OF SALAMANDERS 15 Figure 10. Skull of Amphiuma in dorsal, palatal, lateral and occipital views. Amphiuma is the only living representative of a further family of specialized aquatic salamanders. Unlike Necturus, the nidxilla is a large element, but the pterygoid is small and articulates with the base of the braincase. Tlic squainosal is firmly integrated with the otic capsule. The otic capsule is almost completely covered doi-sally by the parietal. This genus has a ligament joining the stem of the stapes with the quadrate. An anterior portion of an amphiurnid skull from the Upper Cretaceous, described by Estes i 1969~) indicates the early establishment of this basic configuration. ossified as a unit. Functionally the dermal skull roof remains basically the same. In most species, the otic capsule comes to be exposed between the parietal and the squamosal and the squamosal now attaches solely to the unitary otic capsule. One major change seen between hynobiids and even the most conservative of salamandrids and ambystomatids is in the pattern of the vomerine teeth. In ambystomatids, they form a transverse series, nearly reaching the maxillae. In salamandrids, they extend posteriorly, across the surface of the parasphenoid to the level of the basicranial articulation (Regal, 1966). The configuration of the skull seen in plethodontids (Fig. 7) is a further extension of the specialization noted in ambystomatids. The other salamander families (Figs 9, 10-12) depart to a greater degree from the configuration seen in hynobiids, primitive salamandrids and ambystomatids, but all retain a broadly similar suspensorium, formed by the squamosal and quadrate. Considering all currently known living and fossil salamanders, hynobiids appear to show almost entirely primitive character states for the characteristics here emphasized. No features have been recognized in hynobiids as a group that

16 16 K. L. CARROLL AND R. HOLMES 0 A Figure 11. Skulls of Sirenidae. A. Siren in dorsal, palatal, lateral and occipital views. The sirenids are among the most specialized ofliving salamanders. The maxilla in the living genera is reduced to a tiny remnant and the premaxilla has lost its teeth. B. The Late Cretaceous genus Habrosaurus, palatal view, from Estes (1965), shows a much more normal arrangement of the marginal tooth-bearing elements. The pterygoid is completely lost, but unlike other adult salamanders, there appears to be a discrete palatine bone. The squamosal is firmly united with the otic capsule. The exoccipital appears as a distinct ossification. Figure 12. Skull ofalbanel.peton, in dorsal and lateral view, from Estes & Hoffstetter (1976). The only adequately known skull of the extinct family Prosirenidae, Miocene. The skull configuration is advanced over that of the hynobiids in the fusion of the frontals and the great dorsal exposure of the otic capsule. In contrast with most other salamanders, this family does not have pedicellate teeth.

17 ANCESTRY OF SALAMANDERS 17 would preclude salamanders with this morphological pattern having given rise to all other living salamander families (Hecht 8c Edwards, 1977). Hynobiid cranial structure seems to provide a valid basis for comparison with Paleozoic amphibians. The following cranial characteristics are considered of particular importance in primitive salamanders, and would be expected in the immediate ancestors of the group: ( 1 ) Emargination of the cheek (2) Jaw suspensorium formed by quadrate, squamosal and pterygoid. Pterygoid movable on the basicranial articulation and squamosal hinged to otic capsule and parietal (3) Otic capsule made up of distinct opisthotic and prootic (4) Occipital condyle posterior to jaw articulation (5) Double occipital condyle, loss of the basioccipital (6) Absence ofa tympanum (7) Stapes with a large foot-plate and forming a structural link between the braincase and the cheek Items 1, 2, 4, and 5 are definitely specializations over the condition in most Paleozoic amphibians, although 4 is a common feature of larval labyrinthodonts. Item 3 is a primitive character, but a more specialized condition occurs in some Paleozoic amphibians. It is not possible to ascertain with assurance what the primitive character state was in regard to the stapes and the otic notch among plausible lissamphibian ancestors. One of the most primitive adequately known amphibians, the Lower Carboniferous labyrinthodont Greererpeton (Carroll, 1980) lacks an otic notch and has a large h omandibular supporting the braincase against the cheek. This may represent Xe primitive character state for amphibians as a group. The embryological association of the stapes and the palatoquadrate in salamanders shown by Fox (1959) can be interpreted as a direct inheritance from this condition in primitive amphibians. Salamanders differ markedly from labyrinthodonts in the emargination of the cheek. All labyrinthodonts maintain an unbroken palatal margin. Salamanders differ from all but a few labyrinthodonts in lacking an otic notch. The dissorophid labyrinthodonts, which most closely resemble frogs (Bolt, ), have a well developed notch that almost certainly supported a tympanum, and have a small stapes that could have functioned like that of modern frogs. In contrast to primitive salamanders, the otic capsule in dissorophids and their close relatives (Olson, 1941; Sawin, 1941; Carroll, 1964) is formed primarily by a single large ossification which, as in frogs, forms a stout lateral process of the posterior portion of the braincase. It is conceivable that salamanders evolved from such labyrinthodonts, but this derivation would require considerable reorganization of the skull. Much closer resemblances can be noted between primitive salamanders and certain genera of lepospondyl amphibians included in the order Microsauria (Carroll & Gaskill, 1978). Primitive microsaurs (Fig. 2) have a solidly roofed skull, but members of two otherwise quite distinct families have evolved a deep emargination of the cheek (Fig. 13). A condition most like that of salamanders is seen in the family Hapsidopareiontidae. In the genus Liistro&s, the squamosal is reduced to a narrow vertical rod that extended medially to the level of the otic L

18 I8 R. L. CARROLL AND R. HOLMES Figure 13. Skulls of rnicrosaurs. A. Mimuroter. B. Oslodolepis. As in urodeles, there is a large ernbayment of the cheek and the occipital condyle is situated behind the level of the jaw articulation. Members of this family have a large pleurosphenoid ossification that would not be expected in a urodele ancestor. C. The gymnarthrid Curdzocephalus. The cheek shows little ernargination, but the lateral surface of the dentary is excavated for the insertion of a large external adductor. The occipital condyle is far posterior to the level of the jaw articulation, exposing the middle ear in a manner very similar to that of urodeles. D. Lower jaw of Euryodu~, another gyrnnarthrid, showing a very strongly developed area for insertion of the external adductor. E. Llzstrophus, family Hapsidopareiontidae. This genus has a jaw suspensorium like that of primitive salamanders, with a narrow, upright squamosal exteiiditigltnedially to the otic capsule and the parietal. It might have articulated with the braincase. The quadratojugal is very small. Unlike the condition in most salamanders, the occipital condyle is iiot known to extend behind the level of the jaw articulation in this family. Scale for each skull is one inin. capsule and parietal (Fig. 13C). A narrow postorbital bar is retained, but there is a large gap between the maxilla and the jaw suspensorium. As in salamanders, several families of microsaurs (although not known members of the Hapsidopareiontidae) have the occipital condyle well behind the level of the jaw articulation. All microsaurs lack an otic notch. As in salamanders, the stapes (Figs 14, 15) has a broad foot-plate, and a short stem, extending toward the squamosal or quadrate. In so far as the braincase is known, the otic capsule is always ossified from separate opisthotic and prootic elements and the exoccipital is not fused to the opisthotic (Fig. 15). Microsaurs retain a basioccipital, but the exoccipitals are clearly separated and form double condyles in Hupsidopureion. A striking resemblance between a variety of microsaurs and primitive salamanders is in the area of the otic capsule adjacent to the fenestra ovalis. In

19 B Figure 14. Palates of microsaurs. A. Gornorhynchuj. B. Mitraroler. C. Cardzocephalus. These genera Iiavr a staprs with a large footplate adjacent to an unossified area of the otir capsule that may have Iwu~ed ail operr.uluin or connective tissue that was linked to the shoulder girdle by an1 opercularis iiiuscle. The stein of the stapes is directed towards the quadrate and the occipital condyle is behind tlir jaw articulation. As in larval and neotenic salamanders, a row of palatal teeth parallels the jaw iiia~-gin. Figure 15. A, B. Lateral views of the skulls of the microsaur Llislrophus and the hynobiid Batrachuperus wmi) rlie postorbital bar of the rnicrosaur is represented by a dashed line. The braincase and ptcrygoid havr been restored on the basis of other members of the family Hapsidopareiontidae. The rpiptrrygd is based on the structure common to other microsaur genera. C, D, E. Occiputs of.the niici-osaurs Cardzocephalus (C) and Hapsidopareion (D) compared with the salamander Hynobius naevius (El. Specializations of the salamander include the fusion of the exoccipitals and otic capsule and the loss ot the derinal bones at the skull margin: the postparietals and the tabulars. The scale is 1 mm. members of the families Ostodolepidae, Gymnarthridae and Goniorhynchidae (Figs 14, 16) there is an unossified area medial to the base of the stapes. Osteologically, the configuration is almost identical with that seen in Batrachuperus sinensis and Hynobius naevius, in which there is a muscular connection between the otic capsule and the shoulder girdle, but in which the operculum is not a distinct ossification. Taken by themselves, these numerous points of resemblance between salamanders and microsaurs might be considered to support close relationship.

20 20 R. L. CARROLL AND R. HOLMES Figure 16. Palatal views of the skulls of A, the primitive salamander Hynobius naevius, showing the relationship of the opercularis and adjacent hypaxial musculature; 8, the microsaur Mimaroter with these muscles restored according to the pattern in Hynobiw. Although no one microsaur that has so far been described has all the characteristics expected in a salamander ancestor, such an animal would, nevertheless, readily fit within the criteria established for the group as a whole, and might be included within either the family Gymnarthridae or Hapsidopareion tidae. In conjunction with the differences in the bony anatomy of the skull of frogs and salamanders, there is also a consistent difference in the pattern of the jaw musculature. Although there is always some question of the validity of restoring musculature in extinct forms, the configuration in living frogs and salamanders is so consistent and so distinctive in each that it seems logical to assume the same pattern in their immediate ancestors in the early Mesozoic in which an essentially modern skull morphology had already evolved. The jaw musculature may provide additional clues as to the significance of the differences in the surrounding skull bones. ADDUCTOR JAW MUSCLES The terminology of the adductor jaw musculature developed by Luther (1914) and Lakjer (19261, and used by Save-Soderbergh (1945), Haas (1973) and Schumacher ( , together encompassing all the non-mammalian tetrapods, allows simple and consistent comparisons throughout these groups. Three major units of the adductor jaw musculature can be recognized by their position relative to branches of the trigeminal nerve (Fig. 17). The adductor mandibulae externus lies lateral to the maxillary (V,) and rostral to the mandibular (V,) branches of trigeminal nerve. The adductor mandibulae internus lies medial to the maxillary (V,) and rostral to the mandibular (V,) branches of trigeminal nerve, and the adductor mandibulae posterior lies lateral to the maxillary (V,) and posterior to the mandibular (V,) branches of trigeminal nerve. Each of these three masses is variably subdivided in different tetrapod groups.

21 ANCESTRY OF SALAMANDERS MAM E MAMP b igure 17. Subdivision of the adductor jaw musculature giving the nomenclature of Save-Siiderbergh (1945). The jaw musculature of modern amphibians has been described systematically and in detail by Lubosch (1914) and Luther (1914) and reviewed more recently by Save-Soderbergh ( 1945). Additional dissection in the course of this study has confirmed details and provided a tangible basis for comparison with Paleozoic amphibians. Fundamental differences in the anuran and urodele pattern can be seen from the following descriptions (Figs 18-24). The terms primitive and advanced or specialized are used in this description as they apply to the pattern of the adductor jaw muscles and are not to be considered as reflecting general taxonomic position of the genera. The term primitive is used for the condition considered closest to that prevailing in small Paleozoic amphibians in which the jaw musculature was entirely confined to the inside of a continuous dermal skull roof. Fenestration of the skull roof was certainly a fundamental feature of the origin of urodeles, whatever their relationship to frogs. Hynobiids, ambystomatids and primitive salamandrids retain a relatively primitive pattern of the adductor jaw muscles with a low degree of differentiation of separate heads within each of three major muscle units. Specialization or advancement in the adductor jaw musculature is measured by the degree to which the muscles have extended beyond the confines of the original dermal skull roof. The aquatic, and variably neotenic families, Cryptobranchidae, Proteidae, Amphiumidae and Sirenidae have much modified the general pattern of the skull roof and the adductor jaw muscles are much specialized. Adductor mandibulae externus Adductor jaw muscles of urodeles This muscle is always large in urodeles. It typically originates from the anterior surfaces of the squamosal and quadrate (Hynobius, Ambystoma). In larger, aquatic forms, its origin has migrated posteriorly and dorsally to attach to the dorsal fascia of the depressor mandibulae (Cryptobranchus), the prootic and parietal

22 22 K. L. CARROLL AND R. HOLMES Figure 18. Patterns ofjaw musculature of a representative salamander (A), Ambytoma muculdum, and a trog (B) AJCU~~UJ Iruei. These patterns are little modified in common representatives of major groups within each order. Ranids have a small MAME originating on the squamosal and tympanic ring, but this muscle is absent in most anurans. (Necturus, Amphiuma) and even to the fascia of the epaxial trunk musculature (Siren). The adductor mandibulae externus is a single muscle in Hynobius and Ambystoma. In forms with more specialized skulls, it often forms two heads (Cryptobranchus), or even three heads (Necturus), which have adjacent points of origin. The adductor mandibulae externus inserts on the dorsal and/or lateral surface of the lower jaw. In Hynobius and Ambystoma, the insertions are fleshy. This may be a primitive feature, but could also be related to small size. In the more derived (and larger) urodeles, the anterior part of the muscle usually inserts by a tendon (Amphiuma. Necturus). The muscle is clearly pinnate in Siren. Adductor mandibulae posterior Unlike the condition in anurans (see below), the adductor mandibulae posterior of urodeles is a relatively small, usually poorly differentiated muscle

23 ANCESTRY OF SALAMANDERS , Figure 19. Pattern of the adductor jaw muscles in Hynobius naeuius. The quadrato-maxillaly ligament is a feature in common with other salamanders but is omitted trom the subsequent drawings to show better the area of muscle insertions. Areas of muscle insertion on lower jaw shown in dorsal view. MAMI NAME (Pro) Figure 20. Adductor jaw musculature of Crypfobranchus aflegheniensis. All of the major muscles masses are enlarged and subdivided relative to the primitive condition. The MAME is very large and subdivided. The MAMI (sup post) attaches by a long tendon.

24 24 R. L. CARROLL AND R. HOLMES Figure 2 I. The adductor jaw musculature in the salamandrid Notophthalmus viridescms. Despite the bony connection between the frontal and squamosal, the configuration of the muscles remains siniilar to that of' the hynobiids. Two divisions of the MAMP can be recognized. mass. Luther (1914) identified three heads in many urodeles and homologized them with the subexternus, longus and articularis heads of the adductor mandibulae posterior of anurans. Dissections suggest the existence of these three heads in Cvptobranchus, Siren, Amphiuma and Necturus, although identification is sometimes questionable. The adductor mandibulae posterior is not clearly divided in Hynobius retardatus. In Hynobius naevius, there is a clearly defined head inserting on the articular and more diffuse fibers inserting into the Meckelian fossa. All heads originate from the squamosal and quadrate, medial to the origin of the adductor mandibulae externus. Where it can be differentiated, the subexternus head is the most lateral, and inserts on the coronoid process and/or Figure 22. The adductor jaw musculature of Nectuuncr maculosur. This species shows extensive development of tendons from the MAME and MAMI (sup). The MAME has several discrete heads.