The Evolution of the Tetrapod Middle Ear in the Rhipidistian-Amphibian Transition

AM. ZOOLOGIST, 6-379-397(1966). The Evolution of the Tetrapod Middle Ear in the Rhipidistian-Amphibian Transition KEITH STEWART THOMSON Department, of Biology and Penbody Museum, Yale University New Haven, Connecticut SYNOPSIS: From a survey ot the structure of the skull in rhipidistian fishes and early labylinthodont Amphibia and of the mechanism of hearing in these two groups, an account of the evolution of the tetrapod middle ear is presented. The overall modification of the otic region o the skull during the rhipidistian-amphibian transition is analyzed in terms of changes in different organ systems in response to different selective pressures (affecting, for example, the feeding, respiratory, and locomotory mechanisms). These changes are seen to occur in a completely integrated pattern. Considerations of the different requirements for sound reception under water and in air, in connection with this correlated progression of evolutionary change in the otic region of the head, reveal the manner in which the hyomandibular, spiracular diverticulum, and operculum of rhipidistian fishes became modified to lorm the stapes, the tympanic cavity, and the outer portion of the tympanum, respectively, of tetrapods. The ear in vertebrates consists of two principal portions. The sensory structures of the inner ear associated with the brain and enclosed within the cranial cavity are essentially comparable in all vertebrates, although developed and modified to varying extents. The nature of the organ by which information about sound is conducted to the inner ear is different in fishes and in tetrapods, and this difference results from a basic difference in the problem of sound reception under water and on land. In fishes, sound (here we are concerned with wave-form pressure-disturbances of the medium, rather than seismic disturbances) passes more or less freely through the bodywater interface owing to the similarity of relative density, and thus acoustic impedance, of tissues and medium. In fishes, the receptor organ must include a structure which is opaque to sound waves, and generally this is an air-space enclosed within the body, usually associated with the airbladder (see van Bergeijk, 1966, and this Symposium). The range of different receptor organs in fishes is great, but one feature they have in common is that they need not be located near the surface of the body. In a terrestrial animal, the body-air interface is a barrier to the sound pressure-field. The receptor organ for sound in tetrapod vertebrates (the middle ear), therefore, must (379) include a specialized surface structure a flexible membrane (the tympanum) which is capable of reacting to the pressure field, a structure (the stapes), or structures, which can transmit the vibration of the tympanum to the inner ear organ, and also a special cavity behind the receptor membrane which allows the tympanum and stapes to move freely. This paper is concerned with the problem of establishing the evolutionary process by which the middle ear of tetrapods could have evolved from that of fishes. We must discover the morphological and functional continuity which links these apparently entirely distinct patterns of middle ear structure. Classical embryological and morphological studies have established beyond question the homology of the tetrapod stapes with the hyomandibular bone in fishes. This view is so well-known and universally accepted that no defense or bolstering is needed in this study. We may concentrate our attention upon examination of the combination of adaptive and selective factors controlling the evolutionary transformation of this principal bone in the fish-jaw suspension to an element in the tetrapod middle ear. The earliest tetrapods, the Amphibia, as is well known, evolved from ancestral cross-

380 KEITH STEWART THOMSON opterygian fish, members of the Rhipidistia, during Devonian time or slightly earlier. The principal characters in this study will thus be the Rhipidistia and the early Palaeozoic Amphibia. Much of the data presented here concerning the evolutionary history of the Rhipidistia is original and is culled from studies presently being pursued or prepared for publication. With respect to the Amphibia, I have mainly concentrated attention on a single, and hopefully not too atypical, genus, Palaeogyrinus, while Professor Olson's paper (this Symposium) will elaborate the amphibian aspect of the story. Essentially my aim is to discuss the sort of events which occurred during the rhipidistian-amphibian transition and the sorts of factors controlling them, rather than to attempt to discuss particular lineages. Our approach must be functional and morphological rather than simply systematic. THE RHIPIDISTIAN FISHES We may begin by reminding ourselves of the basic structure of the skull in rhipidistian fishes. The illustrations (Figs. 1 and 2) will serve in lieu of a lengthy verbal description. The braincase in Rhipidistia is formed in two separate portions, articulating with each other by a well-formed intracranial joint. The palate is attached to the braincase at only one point, behind the postnasal wall, where there is a fairly close connection of the two elements although a small amount of relative movement was possible. Otherwise the palate, and therefore also the lower jaws, which hinge on the quadrate portion of the palatoquadrate, are suspended from the skull only by the large separate hyomandibular bone. The skull is thus purely hyostylic: the only other interconnections of palate and braincase are through ligaments and the attachment of muscles. The proximal articulation of the hyomandibula onto the posterior portion of the braincase is double (Fig. 1) and the arrangement of the two articular surfaces is of the greatest importance in connection with the complicated kinetic mechanism of the rhipidistian skull which involves articulation of the intracranial joint (see below). The dermal roof of the skull is also arranged in two component portions. The anterior portion of the roof plus the cheek elements form one unit, and the posterior portion of the roof, covering the posterior division of the braincase, forms a second unit. The spiracular gill cleft (Fig. 1) opens at the surface between the dorsal margin of the squamosal and the lateral margin of the posterior roofing bones of the skull. There was probably a ligamentous binding of the hinge between cheek and skull roof, and in the region of the intracranial joint itself the intertemporal, supratemporal, postorbital, and sometimes also the squamosal are arranged with a complicated series of overlapping flanges the pattern of which is also of importance to the operation of the intracranial kinesis. Thus the rhipidistian skull is essentially composed of two units, an anterior unit of endocranium, dermal bones and palate, and a posterior unit of endocranium and dermal roofing bones. The two units are mechanically and functionally integrated principally by the hyomandibular bone. Onto the hyomandibular are also articulated the opercular series of bones, the lower jaw, and the strongly developed skeleton of the visceral arch, forming the third, fourth, and fifth components of the whole head. Close examination of the structure of the posterior portion of the endocranium shows that the double proximal articulation of the hyomandibular spans the lateral commissure of the lateral wall of the braincase. The jugular vein passes through the canal formed by the lateral commissure, and the truncus hyomandibularis of the seventh cranial nerve emerges from the braincase immediately medial to the lateral commissure. A small palatinus VII nerve passes through the lateral wall of the braincase near the anterior opening of the jugular canal, and the truncus hyomandibularis emerges through the posterior opening, curving laterally to enter a canal in the body of the hyomandibular bone. It gives off an oper-

ORIGIN OF TETRAPOD MIDDLE EAR 381 FIG. 1. Lateral view of the skull of Ectosteorhachis. A, dermal bones. B, dermal bones removed. C, dermal bones and palatoquadrate removed, bp, basipterygoid process; es, extrascapular; b, fossa bridgei; fh, facets for proximal hyomandibular articulation; h, hyomandibular; icj, intracranial joint; it, intertemporal; j, jugal; jc, jugular canal; 1, lachrymal; lc, lateral commissure; mx, maxilla; nc, nasal capsule; op, operculum; op pr, opetcular process; p, parietal; pm, premaxilla; po, postorbital; pop, preopercular; pp, postparietal; pq, palatoquadrate; q-j, quadrato-jugal; so, supraorbital; sop, subopercular; sp, spiracle; sq, squamosal; st, supratemporal; t, tabular; II, optic nerve foramen; V, fifth cranial nerve foramina; VII, seventh cranial nerve foramina.

382 KEJTH STEWART THOMSON rop FIG. 2. Lateral view of otic region of Ectosteorhdchis skull. A, dermal bones and palatoquadrate removed. B, showing inner ear cavity and outline of hyomandibular. esd, external opening of spiracular diverticulum; is, internal opening of spiracular cleft; 1, lagena; lat, foramen for ramus lateralis VH nerve; pal, ramus palatinus VII nerve; prf, foramen for profundus V nerve; rk, ramus hyoideus VII nerve; nn, ramus mandibularis VII nerve; rop, ramus opercularis VII nerve; s, sacculus; u, utriculus; VIIF, eighth cranial nerve. cular ramus through a foramen in the posterior wall of the hyomandibular, medial to the opercular articulation, and the hyoid and mandibular rami emerge together through a foramen in the anterior wall, distal to the opercular articulation (Fig. 2). The inner ear in Rhipidistia is quite well known from descriptions of Ectosteorhachis, Eusthenopteron, and Osteolepis. As shown in Figure 2, the three semicircular canals are perfectly formed, and saccular, utricular, and lagenar regions may also be distinguished. The inner ear is located in the midsection. of the posterior endocranium, and the ventral proximal hyomandibular articulation lies immediately outside the sacculus. We should note that a complete set of aquatic respiratory structures gills, operculum, spiracle, and spiracular sense organ is present in Rhipidistia. Thus, even though the Rhipidistia were almost surely able to breathe air (there is, of course, no direct evidence of this), it must be emphasized that they were primarily aquatic animals. In this connection the spiracular gill cleft bears close attention. It opens internally behind the parasphenoid, between the palate and hyomandibular. It passes dorsally between the last-named elements and opens as a slit on the surface of the skull as noted above. An unusual feature is that this external spiracular opening leads into a wide pouch the spiracular diverticulum (Jarvik, 1954) developed dorso-medial to the dorsal margin of the palatoquadrate and lying against the anterior surface of the hyomandibular (Fig. 2). From this spiracular diverticulum, a small blind pocket reaches anteromedially and surely contained a spiracular sense organ lying against the lateral wall of the braincase, as seen also in Polypterus, Amia, etc. Although the function of the spiracular sense organ (whose embryological derivation in Recent forms is from a modified neuromast of the lateral sensory system) is not known there is no reason to believe that it is directly associated with the development of the spiracular diverticulum, for the latter structure seems only to be developed in Rhipidistia. The spiracle itself, in Rhipidistia, probably served a normal function as the inhalant aperture for the water current during respiration. A possible function for the spiracular diverticulum is discussed below. As mentioned above, the hyomandibular is one of the most important elements in the hyostylic rhipidistian skull. It supports the upper and lower jaws (at the quadrate articulation), the skeleton of the visceral arch (hyoid articulation), and the operculum (opercular articulation) and has a double proximal articulation onto the braincase. The detailed nature of the intracranial kinetic mechanism of Rhipidistia, with which the hyomandibular is so intimately bound, is quite complex and forms the subject of a more comprehensive study

ORIGIN OF TETRAPOD MIDDLE EAR 383 now in preparation by the author. However, the general nature of the mechanism must be summarized here for it is of prime importance in understanding the nature of the changes which occur in the otic region of the skull during the evolutionary transition between fish and tetrapod. The intracranial kinesis is best described as a specialization of the feeding and respiratory mechanism. Essentially it involves a dorsal and ventral flexure of the anterior portion of the endocranium upon the posterior portion. This flexure of the intracranial joint is operated by a complex series of muscles among which the subcephalic muscles are most important. The movements of the tip of the snout in the vertical plane are not accompanied by movements of the cheek in the same plane, but by movement of the whole suspensorium in the horizon Lai plane. As summarized in Figure 3, when the tip of the snout is raised the suspensorium is moved forwards and outwards, and when the snout is depressed the suspensorium is moved backwards and inwards. These lateral movements of the suspensorium are controlled and directed by the hyomandibular bone which rotates in an essentially horizontal arc around its proximal articulation. The orientation of the double proximal articulation ensures that this arc is horizontal, and the setting of the hyomandibular at an angle of about 45 to the antero-posterior axis of the body allows significant vertical displacement of the tip of the snout and change in the angle of the upper jaws to the horizontal to be achieved with surprisingly little anteroposterior movement of the suspensorium (Fig. 4). HEARING IN THE RHIPIDISTIA Dr. van Bergeijk has distinguished two different types of 'hearing' phenomena in fishes. First, all are equipped for reception of what van Bergeijk terms near-fie Id sound (water displacements) through the function of the lateral line system as a single unit. Secondly, many fishes are capable of farfield hearing (reception of pressure waves). Since the body tissues of fishes are of approximately the same relative density as the surrounding medium, the receptor organ for far-field sounds is always an enclosed air space connected with the swim-bladder/ lung or with accessory air-breathing organs (van Bergeijk, 1966, and this Symposium). Our problem here is to determine what evidence there is that the Rhipidistia were structurally adapted for the reception of far-field sound. It is possible that the Rhipidistia possessed some modification of the lungs, perhaps a series of ligaments connecting to the otic region of the skull in the manner of Recent ostariophysian fishes, that could have acted as a far-field receptor. However, there is no evidence for this from our fossil materials (the vestibular fontanelle in the lateral otic wall of Eusthenopteron described by Jarvik (1954) is seen from the account of Stensio (1963) not to be associated with the inner ear organs, and is certainly unknown in any other rhipidistian material; cf. van Bergeijk, 1966). In the absence of concrete information about receptor organs of this type it seems preferable to concentrate our attention on the hyomandibular bone and associated structures, for it is reasonable to assume that the transformation of the rhipidistian hyomandibular to the tetrapod stapes must have involved a certain continuity of function, that is, that the hyomandibular in Rhipidistia was connected in some way with a rudimentary far-field hearing function. At first sight, the important role that the hyomandibular plays in the hyostylic intercranial kinesis of Rhipidistia would seem to preclude such a function, except perhaps by direct seismic transmission of vibration if the fish were lying on the bottom. However, there are two possibilities which bear close examination; both concern the presumed air-breathing capabilities of the Rhipidistia. The first possibility is that an air bubble held under the medial surface of the operculum would be ideally situated so that oscillations in its volume in response to wave-form pressure-disturbances of the water could be transmitted as vibrations

384 KEITH STEWART THOMSON B

ORIGIN OF TETRAPOD MIDDLE EAR 385 FIG. 3. Mechanical arrangement of the rhipidistian skull. A, resting position, lateral view. B, dorsal flexure of snout (open lines) from resting position (solid lines), lateral view. C, ventral flexure of snout (open lines) from resting position (solid through the hyomandibular (onto which the operculum is hinged) and thus to the inner ear. In my laboratory I have observed that the brachyopterygian fish, Polypterus, (the general structure of which is remarkably similar to that of Rbipidistia) will very often release a small air bubble behind the operculum upon being disturbed with a glass rod. This is not to say that Polypterus habitually or facultatively keeps an air bubble behind the operculum (this might interfere with normal water passage over the gills) for the subject needs much closer attention. However, it is not difficult to see that even casual retention of an air bubble in this region might bestow a considerable immediate advantage in "hearing," and that this might easily become incorporated into the behavior of the fish if the selective advantage were also high. It will be seen that a similar mechanism might have application in the Dipnoi which otherwise seem ill-equipped- for far-field hearing. It may be noted that in both Dipnoi and Brachyopterygii the exhalant air current during aerial respiration seems normally to pass through the mouth, but once a new gulp of air has been taken, small bubbles (excess air?) always pass from behind the operculum/r suggestirig that the gill chamber is at least temporarily filled with air. This is, of ^course, conjectural and is presented solely with the aim of showing a possible situation, not necessarily r an actual one. The second'possibility concerns the spiracular diverticulum for, as van Bergeijk has recently indicated (1966), if this cavity were air-filled, its location next to the hyomandibular and the structures of the inner ear makes it theoretically an excellent receptor of far-field - sound. At first sight it is difficult to see how this cavity might become air-filled since it is essentially external in location and the spiracle infishesis normlines), lateral view, D, dorsal view of left side of skull in positions A, B, and C. x, angle of upper jaw in resting position; y, angle of upper jaw after dorsal flexure; z, angle of upper jaw after ventral flexure. ally associated with the respiratory water current. Again, observations on Polypterus give us a clue to what may have been the rhipidistian condition. As was first noted by Budgett, when Polypterus comes to the surface to take air it may occasionally be observed to do so through the spiracle, without the mouth breaking the surface. The selective advantage of this is obvious, for it allows the fish to take air with an almost imperceptible disturbance of the water surface. If this were the case in Rhipidistia, too, we might postulate that the spiracular diverticulum evolved as a sound receptor. Presumably at an early stage the spiracular cleft lacked, the diverticulum hut still possessed the small recess for the spiracular sense organ. Casual retention of air in the small cavity might have been of sufficient selective advantage to trigger a process of expansion of this chamber to a large lateral diverticulum within which the function of the spiracular sense organ, in its own small pocket, could be separate from the new function of a sound receptor. Dr. van Bergeijk has suggested an alternate pathway according to which the initial function of the diverticulum was as an accessory respiratory organ, similar to those of the recent anabantoids. This seems unlikely in a fish that already possessed lungs and gills. Whichever may be the case, there is strong indication that the spiracular diverticulum of Rhipidistia functioned as an air-filled receptor organ for. fartfield sound. Neither of the two possibilities just mentioned would seem to have been directly of significant value to the fish if it were to venture out of water, for the air space concerned would be masked from air-borne pressure waves by the tissue-air interface. This problem is discussed at length below. THE EAR REGION IN LABYRINTHODONTS In order to examine the manner in which

386 KEITH STEWART THOMSON psp FIG. 4. Lateral view of the skull ot Palaeogyrinus. A, dermal bones. B, dermal bones removed. C, dermal bones and palate removed. (Redrawn from Panchen, 1964.) at, anterior tectal; bo, basioccipital; eo, exoccipital; Co, foramen ovale; it, intertemporal; j, jugal; 1, lachrymal; mx, maxilla; n, nasal; op, opisthotic; p, parietal; ps, postfrontal; pm, premaxilla; po, postorbital; pit, prefrontal; pro, pro-otic; psp, parasphenoid; q, quadrate: qj, quadrato-jugal; sph, sphenethmoid; sq, squamosal; st, supratemporal; t, tabular; t fac, tabular facet; V pr, foramen for profundus V nerve; I, II, III, IV, V, VI, foramina for olfactory, optic, third, fourth, fifth, and sixth cranial nerves; VII pal, foramen for rainus palatinus VIT nerve.

ORIGIN OF TETRAPOD MIDDLE EAR 387 the various rhipidistian structures became modified during the transition from fish to amphibian, culminating in the evolution of the tetrapod middle ear and the evolution of the tetrapod faculty of "hearing" wave sounds due to aerial pressure, we must first review the structure of the otic region of the skull in a primitive Palaeozoic amphibian. We may use as an example the anthracosaur labyrinthodont, Palaeogyrinus. This has been chosen because its structure is well known, because it is in many ways extremely primitive (retaining a remnant of the rhipidistian cranial kinesis in the movable basipterygoid articulation and a loose connection of the posterior portion of the cheek to the skull roof), and because, being an anthracosaur, it is also most probably close in structure to the amphibian ancestor of the Reptilia. Figures 4 and 5 summarize the basic otic structure oi Palaeogyrinus. The large fenestra ovalis is situated rather towards the center of the inner ear. The rod-like stapes (restoration, Fig. 5) has a short dorsal process fitting into the fenestra ovalis, the latter probably being formed in two portions, an anterior process (homologous with the plectrum in Anura?) which butts onto or may be fused with the parasphenoid, and a posterior process (homologous with the operculum of Anura?) which fits into the fenestra ovalis. The stapes has a process inserting onto the tympanum (possibly by a cartilaginous extension) and possibly also quadrate and hyoid ligaments. These five points of attachment of the stegocephalian stapes have been homologized by Eaton (1939), Romer (1941), and Westoll (1943) with five processes of the rhipidistian hyomandibular. The back of the skull laterally is formed into an otic notch into which there is good evidence that the tympanum was inserted. The jugular vein and facial nerve pass backwards from the endocranial cavity between the dorsal and ventral processes, as may the stapedial (=orbital) artery which in many labyrinthodonts runs through the stapes via a stapedial foramen. The chorda tympani (=rhipidistian r. mandibularis VII) loops behind the stapes and then turns forward ventral and distal to the tympanic process, while the hyoid ramus remains behind the level of the quadrate ligament. The palate is fused to the squamosal and articulates with the braincase by a prominent movable basal articulation. All traces of the rhipidistian intracranial articulation are lost from the braincase (in the Upper Devonian Ichthostegalia a suture remains in this position) and from the skull roof, but the posterior portion of the cheek plate, as mentioned above, is only loosely attached to the skull roofing elements (see Fig. 4). Although we have no direct evidence of its presence, we may be sure that a tympanic cavity enclosed the stapes from the fenestra ovalis to the tympanum, and thus a true tetrapod middle ear was formed. THE EVOLUTION OF AMPHIBIAN STRUCTURES In structural and functional terms the origin of the Amphibia lrom the rhipidistian fishes presents a complex integrated mosaic. Several different adaptive trends in the evolution of the Rhipidistia, involving different structures and different functions, contributed to the development of the amphibian grade of organization. Our concern here is principally with the series of adaptive morphological changes affecting the otic region of the skull. A principal difference between labyrinthodont Amphibia and Rhipidistia lies in the overall proportions of the skull: in the evolution of the Amphibia there has been a considerable increase in the relative length of the anterior portion of the skull (that portion corresponding to the anterior unit of the rhipidistian skull). The adaptive significance of this change is probably that in this way the length of the tooth row, and thus the size of the dental battery, has been increased without an overall increase in the size of the whole skull. It also reflects an increasing development of what Olson (1961) terms the "kinetic inertial" system of jaw mechanics and is surely connected with the carnivorous/predatory mode of feeding. The Rhipidistia also used a basically kinetic inertial jaw mechanism

388 KEITH STEWART THOMSON (complicated by the intracranial kinesis; see Thomson, 1966), and trends towards relative elongation of the anterior unit of the skull may be seen in at least two groups of Rhipidistia, the Osteolepidae (cf. Osleolepis and Gyroptychius) and the Rhizodonticlae (for example, in the line, Tristicopterus, Eusthenopteron, Eusthenodon, and B FIG. 5. Lateral view of otic region of braincase of Palaeogyrinus. A, showing restored stapes, vessels, and cranial nerves. B, showing position of (restored) inner ear organ, apr, anterior process; ct, chorda tympani; dp, dorsal process; fo, foramen ovale; it, interteraporal; jv, jugular vein; oa, orbital (=stapedial) artery; on, otic notch; p, parietal; ppr, posterior process: ql, quadrate ligament; st, stapes; t, tabular; tmp, position of tympanum; tpr, tympanic process; VII, seventh cranial nerve.

ORIGIN OF TETRAPOD MIDDLE EAR 389 FIG. 6. Dorsal view of the skulls of A, Glyptopomus; B, Osteolepis; C, Gyroptychius; D, Eusthenopteron; E, Ichthyostega; F, Palaeogyrinus. (A-E, Rhizodus). These represent adaptive trends that either foreshadow, or more probably parallel, the evolutionary progression which gave rise to the Amphibia. Figure 6, by utilizing certain extremes in cranial proportions, illustrates the basic sort of phyletic progression that must have been involved. It is easy to see that an increase in the size of the dental battery and general feeding mechanism could not be achieved by relative enlargement of the posterior unit of the skull without drastic alteration of the proportions of the whole body. To maintain the hyostylic structure of the skull would necessitate a hyomandibular of large and extremely unwieldy size which might preclude any great increase in the absolute size of the fish. The constancy of overall bodily proportions in the rhipidistian fishes, except with respect to the depth of the trunk which may be relatively greater in some holotychioids, is quite remarkable. Predictably, this constancy of proportion redrawn from Jarvik; F, redrawn from Panchen) Postparietals shaded. imposes some restriction on possible changes in the cranial proportions and any such change will also affect the nature of the intracranial kinesis. Thus, in general, elongation of the anterior unit, if the overall proportions of the skull are maintained, leads to a corresponding reduction of the angle through which the anterior unit may be moved. This is shown in Figure 7 and considered in greater detail in my study of the whole system (Thomson, 1966). It is also demonstrated in a crude way by the progressive reduction in the size of the articular surface of the basipterygoid articulation in the rhipidistian genera, Rhizodopsis (short snout), Osteolepis (medium length snout), Eusthenopteron (long snout); see descriptions by Save-Soderbergh (1936), Thomson (1965), and Jarvik (1954), respectively. Obviously, there must have been a critical point beyond which elongation of the snout rendered obsolete the rhipidistian intracranial articulation, and this, to-

390 KEITH STEWART THOMSON FIG. 7. Diagram of the mechanics of the rhipidistian skull showing the effect of successive migration backward of the position of the intracranial joint gether with factors associated with increase in absolute size and the mechanical problems of suspending a disproportionately large anterior unit upon a small posterior unit of the skull, led to the abandonment of the hyostylic suspension of the jaw. Fusions of the palate to the cheek and the cheek to the skull roof then followed (in anthracosaurs the loose cheek-skull roof attachment was supplemented by dynamic support from the large depressor mandibulae muscles; see below; Panchen, 1964, and Thomson, 1966). Thus, we may see a sequence by which, from changes in the feeding mechanism, the hyomandibular bone was freed from part of its supportive function. Furthermore, once this had happened, the way became more clear for the position of the quadrate to be pushed backwards from its rhipidistian position, thus even further extending the size of the gap. The posterodorsal margin of the squamosal then comes to form the lower margin of the newly-forming otic notch (note that many authors seem to have assumed that the quadrate actually moved forward in the fish-amphibian transition). However, this posterior extension of the suspensorium could not occur without progressive modification of the respiratory mechanism, especially the operculum. Although changes in the respiratory mechanism involve many of the same fea- (1, 2, 3) upon the angle through which the snout may be deflected (1', 2', 3', lespectively) for a given movement of the hyomandibular. tures of the otic region of the skull as do changes in the feeding mechanism, they occur in response to an entirely different series of selective pressures, involving the adoption of aerial respiration as the principal method of respiration in adult Amphibia. Essentially, with increasing dependence on the lungs for respiration, the operculum, skeleton of the visceral arches, and the muscles associated with them, become superfluous. Their progressive modification or reduction removes the last supportive function of the hyomandibular. Absence of these structures also releases space for the reorganization of the otic anatomy, as noted below. There is little evidence that loss of the operculum occurs in response to a requirement for increasing lateral mobility of the head upon the trunk (Westoll, 1943); this seems to be a by-product which is not of significant utility until the amphibians have considerably less dependence upon the aquatic environment for locomotion and feeding. A most important factor in the reorganization of the otic anatomy necessitated by loss of the gills and operculum is that from the embryonic hyoid constrictor muscle sheath, from which the muscles of the hyomandibular and operculum of fishes are formed, new mandibular depressor muscles, which are necessary for the operation of the elongate and heavy lower jaws, may be formed. Dorsally, these

ORIGIN OF TETRAPOD MIDDLE EAR 391 new muscles have their origin in the epaxial fascia and on the posterior surface of the tabular bone. The posterior extension of the tabular for their support contributes in part to the formation of the dorsal margin of the otic notch, as does loss of the extratemporal and supracleithral dermal bones. Furthermore, from the modified hypobranchial muscles is formed the buccal pump by which the lungs are filled. The Dipnoi, which are also air-breathing, at least in part, show an exactly comparable reduction of gill structures and modification of the muscles. The structural and functional integration of the changes occurring in response to modifications of the feeding and respiratory requirements is most remarkable, and to it may be added changes in a third system, namely, the support of the head upon the trunk. In crossopterygian fishes this was not a great problem since the watermedium, and to some extent the notochord, supplied most of the vertical support necessary. Moreover, the otic region of the skull in Rhipidistia bears a pair of dorso-lateral excavations, the fossae bridgei, into which the epaxial musculature inserted. Presumably, this strong insertion, which undoubtedly contributed to the vertical support of the head, was connected primarily with the strong lateral forces developed during powerful swimming movements. In primarily aquatic amphibians such as Palaeogyrinus, these muscles have strong insertions on the medial faces of the tabular horns since the fore-shortening of the otic region of the skull essentially obliterates the fossae bridgei. In other Amphibia which seem to have been more terrestrial in habit, such as the Rhachitomi, lateral and vertical movement of the head upon the trunk would have been impeded by tabular horns, and the epaxial musculature is inserted on the occipital face of the skull in post-temporal fenestrae which seem directly homologous with the fossae bridgei of fishes. This matter of insertion of the epaxial muscles is,'another factor affecting the shape of the otic notch. From this series of integrated changes (which no doubt includes many factors yet to be recognized) the tetrapod ear emerged. The integration of these changes must have been complete, that is, it must have been structural, functional, and temporal (Fig. 8) in a system of correlated progression. The whole progression is more or less related to the occupation of a more "terrestrial" habitat or, rather, to a decreasing dependence upon the aquatic environment either for feeding or breathing. A part of the general selective pressure must have been toward the perfection of a hearing organ that could operate effectively both in the water and out. In the preceding paragraphs an attempt has been made to demonstrate the structural and functional continuity maintained during the transformation of the general otic anatomy during the rhipidistian to amphibian transition. It is now necessary to examine this process more closely to discover whether the same continuity may be seen in the evolution of the tetrapod middle ear (the stapes, tympanum, and tympanic cavity). The development of aerial hearing in tetrapods. As we have already noted, underwater hearing requires the presence of an airspace within the body cavity of the fish. In response to changes in pressure of sound waves propagated through the water and through the tissues of the fish, the volume of this air cavity varies. The physical displacement of the walls of the cavity is transmitted through the tissues of the fish (sometimes by special structures, such as the Weberian ossicles) and received by the inner ear organs. Thus, underwater hearing in a fish is a process depending upon the volume/pressure relationships of an enclosed air space which produces a particular type of mechanical displacement in response to a given sound. Once the fish leaves the water this situation is completely changed, and the air space is masked from the airborne sound waves by the surrounding body tissues (by virtue of the difference in acoustical impedance of tissue and air).

392 KEITH STEWART THOMSON hyomandibular suspensorium spiracular 'diverliculum operculum FIG. 8. Two-dimensional representation of the correlated changes leading to the development of the tetrapod middle ear. Thus, a new way must be found to develop a mechanical response to the pressure changes which constitute the sound wave. The problem is solved by greatly specializing a region of the body surface in such a way that it is physically capable of registering the tiny changes in air pressure in terms of slight mechanical deformations which are transmitted to the inner ear organs. Clearly, in addition to this specialized surface structure, a specialized transmitting organ is required the ear ossicle. The specialized surface membrane (the tympanum) and the ear ossicle (the stapes) must also be freely suspended within the body in order that the small movements developed in the tympanum may be transmitted to the inner ear with maximum mechanical efficiency. This is accomplished by enclosing the whole organ in an' air space the tympanic cavity. From what we have already seen of the structure of rhipidistian fishes and amphibians, it is obvious that the tetrapod stapes has evolved from the hyomandibular of fish, and the tympanic cavity must be related to the spiracular diverticulum. Further, if our attribution of an underwater hearing function to the hyomandibular is correct, then there is not only a morphological, but also a functional, continuity between the two sets of structures in fish and amphibia. van Bergeijk (this Symposium) has developed an ingenious theoretical model which seeks to interpret this continuity. According to his theory, the same pressure/ volume relationships which determine the characteristic response of the "middle ear" of fish are also of vital importance in the terrestrial vertebrates. However, since a sound in air "produces only 1/63 of the pressure developed by an underwater source of the same intensity" (van Bergeijk, this Symposium);, then,, he believes, some device has to be found by which the displacement of the wall of the air-space could be increased by 63 times. This, he suggests, could be accomplished by restricting the

ORIGIN OF TETRAPOD MIDDLE EAR 393 movable portion of the wall of the cavity to 1/63 of the total surface area. This restricted surface, necessarily coming to the body surface, becomes the tympanum. Unfortunately, closer examination of this theory shows that it is incompatible with most of what we know about the practical situation. First of all, van Bergeijk's model is made with the assumption that all the energy that would be available to the cavity if the whole of its wall were flexible (as is the case when the fish is under water) is still available in air when only 1/63 of the surface is movable. However, this is not the case. The tetrapod tympanum is not moved from within, it is moved from the outside. Thus, since the force available for deformation of the wall of the tympanic cavity is proportional to the surface area upon which the pressure acts (pressure = force per unit area), in the model only 1/63 of the force is available. Clearly then, it is desirable that the tympanum, the only place upon which the sound pressure can act, should really be as large as possible in order to produce the optimum response to a sound of given characteristics. Secondly, in his theoretical model the assumption is made that the tympanum is not loaded. However, in fact, the tympanum is connected to a large and relatively extremely heavy hyomandibular (especially in the earliest stages of the transition). Now according to the equation, F (force) = M (mass) X a (acceleration), the addition of the hyomandibular (stapes), which could scarcely have a mass less than 20 times that of the tympanum itself, to the tympanum will reduce the acceleration which may be produced by a corresponding proportion. This is a second factor favoring increase in the size of the tympanum. The next question is: suppose that the surface area of the tympanum were to be increased beyound the theoretical 1/63 proportion, would the change in volume which its displacement produces be greater than could be accommodated by the tympanic cavity, the walls of which (according to van Bergeijk) are essentially inflexible? Here we encounter a third objection to the proposed theoretical model. Simply consider the tympanic cavity of a large specimen of a rhipidistian fish such as Eusthenopteron: it might have had a volume of about 4 cm 3, that is, a volume equivalent to that of a sphere with radius of 1.0 cm. Now a typical movement of the tympanum of a tetrapod might be expected to be in the vicinity of 10 A (1 X 10" 7 cm). If the whole surface of the cavity were to be displaced by this amount, then the total change in volume would be only 2.57 X 10~ T cm3. which is an extremely small amount. However, if only 1/63 of the total surface area can move, then the change in volume is now only 4.08 X 10" 9 cm 3. It will be obvious that this extremely small change is unlikely to have been of great significance to the living animal (for example, the pulsing of the stapedial artery, which passes through the cavity, would have produced a change in volume of this order, or even greater). Clearly then, the surface area of the tympanum could be increased beyond the 1/63 limit many times without significantly affecting the pressure/volume relations of the tympanic cavity. We can only conclude that the direct relationship between the pressure and volume of the tympanic cavity and the displacement of the tympanum is not likely to have been the critical factor in determining the nature and course of evolution of the tetrapod ear. From the above considerations we may come to certain conclusions about the nature of the transition as it applies to the mechanics of the ear and the physics of hearing in air. From the two equations, P = F/A (pressure = force per unit area) and F = Ma (force = mass X acceleration), and from the fact that for a given sound the acceleration produced will be proportional to (the square of) the frequency, there will be an inverse relationship between the mass of the tympanumear ossicle system and the frequency of the sound that may be received (that is, that will produce a given movement of the ossicle). If the mass of the ear ossicle is large and the surface of the tympanum is very small, the range of possible frequency re-

394 KEITH STEWART THOMSON sponse is extremely limited. Frequency response could only be greater if the intensity of the sound were greater, that is if the pressure acting upon unit tympanic area were greater. However, we are talking here of an evolutionary process. All that we know about such processes requires us to expect that through even the very earliest stages of evolution the ear must have had significant functional value. In this case the animal must have been capable of responding to a useful (if small) range of frequencies, (e.g., up to 150-200 cps) at relatively modest intensities (an equivalent of ca. 60-80 decibels). An inescapable fact about the hyomandibular in the fish-amphibian transition is, as we have seen above, that it has important connections to the palatoquadrate bones, the lower jaw, the visceral skeleton, and to the operculum when present, together with various ligaments and muscles. Thus the effective mass of the ear ossicle in the transitional stages was extremely high, and the mechanical work that had to be performed in order to accelerate the stapes, even over the short distance of 10 A, is also considerable. Since the energy that is available to do the work is solely that which is "received" by the tympanum, we must expect the tympanum, at the earliest stages, to be of large size. This is especially true when we consider that in the earliest stages the tympanum may be expected to be relatively unspecialized in structure, van Bergeijk has pointed out to me (personal communication) that the problem of matching impedance is taken care of by matching the mass of the hyomandibular to that of the inner ear. In this connection it is interesting to note that the oval window is not perforate until the full amphibian condition is reached, at which time the hyomandibular (stapes) is relatively much less massive. There is, of course, a possibility that the movement of the hyomandibular was amplified by mechanical leverage as the bone was moved about its fulcrum at the dorsal proximal articulation. This amplification is unlikely to have been greater than a factor of 2. We may now turn to examine the morphology of the rhipidistian fishes to determine from what structure the amphibian tympanum has evolved. Arguing from the basis of his theoretical model (the validity of which has been questioned above) van Bergeijk has developed the theory (this Symposium) that the rhipidistian fish possessed a rudimentary tympanic membrane in the region of the spiracular opening, the membrane being formed by the pressing of the outer wall of the spiracular diverticulum against the ligament binding the squamosal to the tabular. There are certain practical difficulties in this theory, the principal one being that the connection between the skull roof and cheek plate must have been very close, the only significant gap being in the notch between the tabular and supratemporal where the spiracle opens. Most probably the squamosal-tabular contact was a simple hinge at which the squamosal was able to rotate upon the tabular while remaining in contact with it. The two bones were bound together by a ligamentous connection, but it is unlikely that, even during movements of the cheek associated with the cranial kinesis, the two bones were ever separated by more than a millimeter. In any case, if the fish were to venture on land, the cheek and operculum would surely be tightly adducted to prevent drying of the gills; in this position the squamosal and tabular would be pressed firmly together. Thus, it is unlikely that even a rudimentary eardrum could have developed in this position. Even if such a structure were present, the mechanical requirements of the hinge are such that the ratio of the thickness of ligament to the width of slit must have been very high (in the order of 0.6-0.9 : 1) and the ligament is, therefore, unlikely to have been physically suited for acting as a tympanum. Further, if such a system were present it would only be capable of functioning as a sound receptor device if the vibrations of the "tympanum" could be transmitted to the otic capsule through the hyomandibular. In the hypothetical tympanum in EusUienopteron postulated by

ORIGIN OF TETRAPOD MIDDLE EAR 395 van Bergeijk, the "eardrum" would be bounded solely by the dermal bones of the skull. The hyomandibular does not reach the opening of the diverticulum or any dermal bone except the operculum (this is shown clearly by Jarvik, 1954, Fig. 47). The only connection of the hypothetical eardrum to the hyomandibular is, thus, the flexible medial wall o the diverticulum which meets the "drum" at its attachment to the rim of the tabular bone. It does not seem likely that this could serve for an efficient transmission of "tympanic" vibration to the hyomandibular. The main objection to this theory is that such an "eardrum" would have been far too small to operate the massive hyomandibular in the manner that we have seen (above) would be necessary for effective functioning of the ear. The only serious candidate as the precursor of the tetrapod tympanum seems to be the opercular bone. We have noted previously that it might have a function in underwater hearing. It is directly connected to the hyomandibular bone, and comparative anatomical and embryological evidence suggests that the connection of the operculum to the hyomandibular in fishes is homologous with the tympanic process of the tetrapod stapes. Briefly, the hypothesis may be presented that in certain rhipidistian fishes the selective advantage of being able to "hear" far-field sound underwater would lead to an increase in the size of the spiracular diverticulum. Such an increase would be facilitated by the general changes in the proportions of the skull which, among other things, made space available for this expansion and involved a progressive decrease in the suspensorial functions of the hyomandibular. Eventually the spiracular diverticulum would be able essentially to surround the hyomandibular, and there would be developed a tympanic cavity of the basic tetrapod type. We should note that it would not be of advantage for the cavity to enclose the hyomandibular completely unless there were also developed some mechanical means whereby the oscillations in volume of the cavity could be transferred to the hyomandibular. At an earlier stage this was achieved through contact of the medial wall of the cavity with the hyomandibular. In the most advanced fish-stage this was best achieved through a connection of the outer wall of the cavity to a distal part of the hyomandibular, and this would occur at the opercular process which naturally limits the lateral extent of the cavity. It is theoretically possible that this stage of development of the tympanic cavity could be attained while a full set of gill structures was still present if the animal were concerned only with the underwater hearing and structural modification of the otic region of the skull had otherwise proceeded far enough to allow for expansion of the cavity. This might, then, closely approximate the "prototetrapod" stage that Parrington (1958) envisioned in the fish-tetrapod sequence, since the stapes would be only moderately modified from the rhipidistian condition and the otic notch would not be developed. However, the fully-correlated nature of all the adaptive changes occurring in the otic region of the skull and the nature of the selective pressures involved indicate that development of a tetrapod-type of tympanic cavity was proceeding in step with reduction of the whole aquatic respiratory system. Thus, we may suggest that precisely as the tympanic cavity expanded to enclose the hyomandibular and reached out laterally to the opercular process, the gills were fully reduced. When the fish "came on land" the former operculum, now lying in and being supported by the new otic notch, became a new and flexible structure which, by virtue of its shape and large size, was able to vibrate directly in response to air-borne pressure waves and to transfer this motion to the hyomandibular (now the stapes, lying in the tympanic cavity) by way of the opercular (tympanic) process. Since this mechanism must at the very outset also have been required to receive water-borne pressure waves, it is clear that the outer wall of the tympanic cavity must have been closely applied to the inner surface of the opercu-

396 KEITH STEWART THOMSON lum in order that volume changes of the air-filled tympanic cavity would also be transmitted to the stapes via the tympanic process. According to this view, the tetrapod tympanum may be considered to have been derived from two sources, the outer portion from the former opercular bone and the inner portion from the wall of the spiracular diverticulum of the fish ancestor. It might seem that the radical nature of these structural modifications, especially in the respiratory system, might present serious problems in the ontogenetic mechanics of the earliest amphibian larvae. This would not have been so if, as seems most probable, the larvae of Rhipidistia possessed external gills. Indeed, the presence of larval external gills in the ancestral stages is virtually a prerequisite for the whole evolutionary sequence. POSTSCRIPT One point of general theoretical importance emerges from this evolutionary study of the rhipidistian and amphibian otic anatomy. Many authors from Darwin onwards have observed that selection of a particular character is frequently accompanied by correlated side-effects which are not directly related to the immediate selective pressure or the particular structure upon which the selection is apparently acting. These responses have been called "correlated responses" or "correlated effects" (see Mayr, 1963, pp. 287, 290, 607). The sequence of events which I have called "correlated progression" above is clearly related to these effects but in a curious and special way. In a correlated progression of evolutionary change, such as we have seen to have been associated with the origin of the tetrapod middle ear, correlated changes occur in related structures in response to different selective pressures. The progress of each modification is dependent upon the simultaneous progress of the other modifications and the whole system may be considered to act in response to an overall selection which is the sum of the series of separate pressures and responses (concerning feeding, respiration, locomotion, etc.). In the example considered here, this overall selection is concerned with decreasing dependence upon the aquatic habitat and has resulted in the origin of an entirely new organ, the tetrapod middle ear. The implication of this sort of analysis of palaeontological data is considerable, having bearing upon such subjects as preadaptation and the mechanism of the origin of major groups, but consideration of these matters must be made elsewhere than in the present paper. Finally, on a phylogenetic note, it will be recognized that in the preceding pages I have been concerned principally with the evolution of the labyrinthodont middle ear. My study of this leads me to the conclusion that from the outset an important factor in the evolutionary complex is the development of an otic notch and also a relatively dorsal position of the stapes. Thus, if the Reptilia have evolved from a labyrinthodont ancestor, the stapes must have migrated ventrally from an originally dorsal position (see Parrington, 1958, 1959; Hotton, 1959, 1960). If the Amphibia are truly monophyletic, then all early forms must have possessed an otic notch and dorsal stapes. However, it is perfectly possible that various Palaeozoic Lepospondyli have had a separate origin (cf. Thomson, 1964; Romer, 1964; Schaeffer, 1965) and could have developed an essentially homologous middle ear by a slightly different pathway; in this line, modification of the feeding mechanism may not have involved a posterior migration of the quadrate. ACKNOWLEDGMENTS The final version of this paper was prepared after the Symposium for which it was originally written, and after extended discussion of the topic with Dr. W. A. van Bergeijk whose own contribution presents a different viewpoint on the origin of the tympanum. I wish to thank Dr. van Bergeijk and also Dr. J. A. Hopson for valuable discussions on this subject. Thanks are also due to my wife and to Ivfrs. Barbara E. Moss for assistance with the preparation of the manuscript. REFERENCES Eaton, T. H. 1939. The crossopterygian hyomandibular and the tetrapod stapes. J. Washington

ORIGIN OF TETRAPOD MIDDLE EAR 397 Acad. Sci. 29:109-117. Hotton, N. 1959. The pelycosaur tympanum and early evolution o the middle ear. Evolution 13:99-121. Hotton, N. 1960. The chorda tympani and middle ear as guides to origin and divergence of reptiles. Evolution 14:194-211. Jarvik, E. 1954. On the visceral skeleton in Eusthenopteron with a discussion of the parasphenoid and palatoquadrate in fishes. Kungl. Svenska Vetenskapsakad. Handl. (4) 5: (1)1-104. Mayr, E. 1963. Animal species and evolution. Harvard Univ. Press, Cambridge, xiv + 797 p. Olson, E. C. 1961. Jaw mechanisms: rhipidistians, amphibians, reptiles. Am. Zoologist 1:205-215. Panchen, A. L. 1964. The cranial anatomy of two Coal Measure anthracosaurs. Phil. Trans. Roy. Soc. (London), B, 247:593-637. Parrington, F. R. 1958. The problem of the classification of reptiles. J. Linn. Soc. (London). Zool. 44:99-115. Parrington, F. R. 1959. A note on the labyrinthodont middle ear. Ann. Mag. Nat. Hist. (13) 2:24-28. Romer, A. S. 1941. Notes on the crossopterygian hyomandibular and braincase. J. Morphol. 69: 141-160. Romer, A. S. 1964. Problems in early amphibian history. J. Animal Morphol. Physiol. 2:1-20. Save-S6derbergh, G. 1936. On the morphology of Triassic stegocephalians from Spitzbergen, and the interpretation of the endocranium in Labyrinthodontia. Kungl. Svenska Vetenskapsakad. Handl. (3) 16: (1)1-181. Schaelfer, B. 1965. The rhipidislian-amphibian transition. Am. Zoologist 5:267-276. Stensio, E. A. 1963. The brain and the cranial nerves in fossil, lower craniate vertebrates. Skr. Norske Videnskaps-Akad. Oslo, I Mat-Naturw. Klasse. Ny Serie. no. 13:5-120. Thomson, K. S. 1964. The ancestry of the tetrapods. Sci. Progress 52:451-458. Thomson, K. S. 1965. The endocranium and associated structures in the Middle Devonian rhipidistian fish Osteolepis. Proc. Linn. Soc. (London) 176:181-195. Thomson, K. S. 1966. Mechanisms of intracranial kinetics in fossil rhipidistian fishes (Crossopterygii) and their relatives. Proc. Linn. Soc. (London). (In press) van Bergeijk, W. A. 1966. The evolution of vertebrate hearing. Contrib. Sensory Physiol. 2: (In press) Westoll, T. S. 1943. The hyomandibular o Eusthenopteron and the tetrapod middle ear. Proc. Roy. Soc. (London), B, 131:393-414.