Hearing without a tympanic ear

Sound represents an essential cue to animals by quickly conveying information about near and far mechanical disturbances around visual barriers and irrespective of light conditions. The natural environment is full of sounds that vary in their origin and relevance to the listener, and the vertebrate auditory system evolved to enable the detection and perceptual allocation of these sounds to their sources to inform a behavioral response (Bregman, 1990). Hearing organs evolved from receptors for detecting motion of the surrounding medium: sensory hairs that could be free-standing or coupled to accessory structures, such as gelatinous cupulae or otolithic masses (see Glossary) (Fritzsch and Straka, 2014). Owing to the higher density of the otoliths compared with the surrounding tissue, their gravitational pull will bend the sensory hairs depending on the orientation of the animal. Additionally, the inertia of the heavier otoliths will bend the sensory hairs when the animal is accelerated. Thus, in the aquatic ancestors to vertebrates, otolithic sensory organs had a dual function: they would register the animal's orientation and movement in space (as equilibrium organs – see Glossary) and concurrently respond to movement displacing the animal. This would be the basis of the ‘acoustical sense’, the ability to detect mechanical disturbances in the vicinity of the animal, including sound waves generated by biotic and abiotic events.

Sound waves are elastic longitudinal waves producing pressure and particle motion oscillations in a medium and represent one of the diverse types of acoustical signals that may be present in the environment (Christensen-Dalsgaard, 2019). In water, particle motion is easily transmitted through the animal to its inner ear because of the similar density of its tissue and water. In air, although the particle motion component of sound at a given sound pressure is nearly 4000 times larger than that in water – around 2.5 mm s⁻¹ Pa⁻¹ – a large part of sound energy is lost by reflection as a result of density differences between the medium and the body (i.e. an impedance mismatch). The evolution of structures that collect sound energy and transmit vibrations to the inner ear, the tympanic middle ears of the tetrapods (Box 1), emerged as an evolutionary novelty that, among other properties, radically changed sensitivity to airborne sound (Christensen-Dalsgaard and Carr, 2008; Clack, 1997).

The presence of a well-developed tympanic middle ear permits the detection of airborne sound across a broad range of frequencies; however, many extant vertebrates, including snakes, some lizards, salamanders, caecilians and ‘earless’ frogs, demonstrate some degree of reduction or loss of the various components of the tympanic middle ear (Pereyra et al., 2016; Toerien, 1963; Wever, 1973, 1985). In this Review, we describe the complex evolutionary history of the vertebrate tympanic ear and discuss the evolutionary processes that may have shaped the multiple instances of loss or reduction to the tympanic middle ear observed among extant vertebrates. We first focus on hearing without a tympanum in the transition from water to land, where recent work suggests early terrestrial tetrapods could have had similar hearing to modern lungfish and salamanders, and thus sensitivity to low-frequency airborne sound (Capshaw et al., 2020; Christensen-Dalsgaard et al., 2011; Christensen et al., 2015a,b). Next, we synthesize the results of recent and historical studies of hearing in atympanate species to discuss mechanisms for aerial sound detection in these species. Finally, we argue that extratympanic pathways for hearing are sufficient to enable detection and localization of sound pressure in the absence of an impedance-matching ear. We conclude with suggestions for future research, emphasizing the need to evaluate the mechanistic basis for directional processing in atympanate vertebrates.

The evolutionary history of the otic region (see Glossary) in early tetrapods shows initial variability followed by later stabilization, suggesting that early forms represent ‘experimentation’ with ear structure before gradual selection favored the tympanic middle ears of recent lineages (Clack and Anderson, 2016). The particle motion component of sound can potentially stimulate any mechanoreceptor, whereas sensitivity to the pressure component of sound entails specialized, compressible structures to transduce pressure to particle motion in the ear. Without such specialized structures, ancestral vertebrates most likely were sensitive to particle motion generated by sound sources, as in fish.

Early forms of the tetrapod stapes (or hyomandibula; see Glossary) functioned as a strut supporting the braincase, and it is likely that the hearing capabilities of these taxa were poor in air (Clack, 2015). Key rearrangements of the cranial skeleton resulted in the release of the hyomandibula from its structural role and, over time, its transition to a slender, rod-like bone positioned in the fenestra vestibuli (oval window) of the otic capsule (Clack, 1997, 2002). Tetrapods began to show distinct evidence of obligate terrestriality by the early Carboniferous approximately 365 million years ago, yet the appearance of a true otic notch housing a tympanum rather than the plesiomorphic spiracular notch does not appear until the lower Permian among the Temnospondyl ancestors to Lissamphibia (Bolt and Lombard, 1985; Robinson et al., 2005; Sigurdsen, 2008; Sigurdsen and Bolt, 2010). These species had perhaps the first true tympanic ear represented in the fossil record, characterized by a frog-like tympanic–stapedial complex. In amniotes, tympanic middle ears appear not to have evolved until the Triassic, around 120 million years after the first tetrapods and 70 million years after tetrapods became truly terrestrial (Fig. 1).

Paleontological evidence supports independent acquisitions of the tympanic middle ear among mammals, lepidosaurs, archosaurs and turtles in which rearrangements of the jaw suspensorium led to different associations of the tympanum and hyomandibula with the lower jaw (Clack, 1997, 2002). Recent evo-devo analyses show distinct ontogenetic origins for the tympanum and middle-ear ossicles in birds and mammals, with only the stapes being homologous as a derivation of the second pharyngeal arch (Amin et al., 2007; Kitazawa et al., 2015; Thompson et al., 2012; Tucker et al., 2004; Urban et al., 2017). Thus, differences between various pharyngeal skeletal elements associated with the tympanic middle ear resulted in structural diversity and the acquisition of the tympanum in at least two distinct manners during amniote evolution (reviewed in Anthwal et al., 2012; Manley and Sienknecht, 2013; Tucker, 2017).

The aquatic ears of the earliest tetrapods lacked specializations to enable terrestrial hearing; therefore, the extent to which they could detect airborne sound remains an open question. A fundamental challenge to studies of sensory evolution is that direct examination of the ancestral state is often impossible, and inferences of functionality must be made using extant animal models. In the case of the ancestral tetrapod ear, our understanding of its configuration is restricted to structural data gleaned from fossils, which are often poorly preserved and incomplete, and rarely retain soft tissue characters. From these data, however, we may observe the presence or absence of structural features, and this can be an effective basis for selecting appropriate model organisms that reflect anatomical aspects of the ancestral ear. Such data may, therefore, inform our understanding of the physiology of the hypothesized ancestral state of the auditory system.

Lungfishes, the sister group to tetrapods, have no obvious adaptations for aerial hearing, i.e. they have no middle-ear cavity or functional equivalent to the tetrapod stapes through which sound can be conveyed to the inner ear. However, recent studies suggest that their atympanic auditory system confers some degree of sensitivity to aerial sound pressure (Christensen-Dalsgaard et al., 2011; Christensen et al., 2015a). It is possible that the physiology of the lungfish ear is indicative of the functionality of similar aquatically adapted ears during the early stages of the water-to-land transition (Clack, 2015). Additional studies of hearing in snakes and salamanders have shown that, even with an atympanic ear, animals can hear airborne sound, at least in the low-frequency range (Fig. 2) (Capshaw et al., 2020; Christensen et al., 2012). Although the auditory system of any extant species has been shaped by millions of years of natural selection and therefore does not replicate the early tetrapod ear, its physiology still reflects what an atympanic ear can do, albeit in a somewhat ideal situation (Christensen et al., 2016).

The ‘earless’ condition observed in most, if not all, extant atympanate tetrapods is secondarily obtained, i.e. it is the result of a reduction or loss of the tympanic middle ear present in a common ancestor [whether this is likely for urodeles and caecilians is contingent on the phylogenetic relationship of recent amphibians; see Maddin and Anderson (2012) and Marjanović and Laurin (2013) for alternative hypotheses of amphibian phylogeny]. Historically, the absence of a tympanic middle ear has been linked to ecological factors, including adaptation to aquatic or fossorial habitats in which acoustic sensing does not necessitate an impedance-matching ear. For example, loss of the tympanum was hypothesized to indicate specialization for fossoriality, particularly among squamates that use their heads to burrow through soil (Greer, 2002; Olson, 1966; Wever, 1978a, 1985). In these animals, a connection between the middle-ear ossicle and the lower jaw may provide an important route for transfer of acoustic energy to the inner ear. However, atympanate tetrapods occupy diverse ecological niches including many surface habitats. The selective pressures that underlie the evolutionary loss of the tympanic middle ear in terrestrial vertebrates may be as diverse as the species themselves, but studies using atympanate frogs reveal evolutionary correlates that may be associated with the loss and reduction of key structures of the tympanic ear. Interestingly, given the ubiquity of complete or partial loss of middle-ear structures in squamates and amphibians, such loss has (so far) not been reported for any wild-type mammal or bird species.

In amphibians, earlessness appears most strongly linked to protracted growth trajectories, developmental truncation and miniaturization (see below for a more detailed explanation of these terms). These biological phenomena may in fact reflect the varied, possibly interrelated, consequences of genome evolution (Hanken and Wake, 1993). Genome size is positively correlated with cell size; therefore, species with large genomes must make do with fewer, larger cells. This leads to an overall simplification of complex traits, including brain morphology and sensory structures (Roth and Walkowiak, 2015). Additionally, increased genome size slows cellular proliferation and development, in many cases leading to paedomorphosis (see Glossary) and miniaturization at the organismal level in vertebrates (Hanken and Wake, 1993; Pérez-Ben et al., 2018; Roth and Walkowiak, 2015). The link between genome size, paedomorphosis and simplification/loss of neural structures may be especially important for salamanders, which have among the largest genomes observed in vertebrates and are atympanate (Decena-Segarra et al., 2020; Roth et al., 1988).

The evolutionary loss of the tympanic ear in extant vertebrates may reflect convergent effects of developmental shifts on cranial structure. In many cases, developmental heterochronies, such as progenesis (early offset), post-displacement (early onset) and/or neoteny (decelerated development), result in ontogenetic truncation of late-forming cranial structures (Sánchez-Villagra et al., 2008; Schoch, 2006; Trueb and Alberch, 1985). In amphibians, the components of the tympanic middle ear are among the last cranial elements to develop – often achieving adult form late in metamorphosis and, in some cases, up to a year post-metamorphosis (Hetherington, 1987; Smirnov, 1991; Vorobyeva and Smirnov, 1987). Development of the anuran tympanic middle ear progresses in a medial-to-lateral sequence and heterochronic shifts appear to affect peripheral structures of the ear most strongly (Hetherington, 1987; Smirnov, 1991; Vorobyeva and Smirnov, 1987; Womack et al., 2018a, 2019). For example, differentiation of tympanic membrane tissue requires inductive factors generated by the developing tympanic annulus, the absence of which results in the concomitant absence of the tympanum (Gross and Hanken, 2008; Helff, 1928). Additionally, progenesis leads to degeneration of the developing stapes during late metamorphic stages of earless toads (Stynoski et al., 2021). Among extant anurans, loss of late-developing, peripheral structures of the tympanic ear reflects the loss of more medial structures, indicating that disruption of early stages of development may be a strong, proximate driver of earlessness.

Similarly, heterochrony has influenced cranial diversity among lepidosaurs, where loss of the tympanic ear may represent the outcome of selection for cranial kinesis to support prey capture and manipulation (Evans, 2016; Herrel et al., 2007; Irish, 1989; Rieppel, 1980). Selective pressures on the jaw suspensorium strongly influence the morphology and orientation of the quadrate, which serves as both the primary articulation point of the lower jaw and structural support for the tympanum in squamates (Palci et al., 2020). The absence of a tympanum may release mechanical constraints on the quadrate and enable selective pressures to shape novel jaw morphologies and configurations. Earlessness in these species could therefore reflect functional tradeoffs related to feeding biomechanics, especially among species with highly specialized feeding mechanisms such as snakes and chameleons (Iordansky, 2016; Moazen et al., 2009; Scanferla, 2016). The loss of the tympanum in fossorial species such as amphisbaenians and snakes, often combined with restructuring the middle ear to perceive substrate-borne vibration (Christensen-Dalsgaard and Manley, 2013), suggests that adaptations for fossorial life can also lead to earlessness.

Earlessness may also occur as a consequence of evolutionary body size reduction. In amphibians and squamates, miniaturization is commonly correlated with incomplete development, simplification, fusion and/or loss of anatomical structures, including those of the tympanic middle ear (Hanken, 1983; Hanken and Wake, 1993; Rieppel, 1996). Convergent changes to the skull of miniature squamates include a high level of ossification to the braincase, leading to increased compactness of the skull and, commonly, loss of the tympanum and enlargement of the stapes (Rieppel, 1985, 1996). However, because miniaturization in tetrapods often co-occurs with fossoriality, heterochrony and cranial consolidation, it is difficult to attribute the loss of the tympanum in miniature earless species to any single selective pressure.

Finally, extreme body size reduction imposes scaling limitations to tympanic function. Studies in gekkonid lizards and frogs indicate that large species have tympana with larger surface areas that are more responsive to sound pressure than those of smaller animals (Hetherington, 1992; Werner et al., 1998, 2002, 2008). The area ratio of the tympanum to the stapedial footplate area also decreases with decreasing body size, reducing the efficacy of this middle-ear system to transform sound energy to vibrations in the inner-ear fluids (Hetherington, 1992; Werner et al., 2005). Further, as adult body size decreases, the tympanum becomes less responsive to sound pressure compared with undifferentiated non-tympanic skin surfaces (Hetherington, 1992). This indicates that there may be a lower size limit for tympanic function, below which extratympanic pathways become more effective for sound reception. If extratympanic pathways are sufficient to enable aerial hearing, the functional significance of the tympanic ear may be reduced, thereby relaxing selective pressures to maintain these structures in small-bodied species.

Reduction or loss of the tympanic middle ear generally results in decreased sensitivity to airborne sound; however, the reduction to aerial hearing sensitivity is not as complete as one might expect for an ‘earless’ animal. All atympanate species studied to date can detect high-amplitude, low-frequency sound pressure (Fig. 2) and vibration (Capshaw et al., 2020; Christensen et al., 2012, 2015b; Lindquist et al., 1998; Womack et al., 2017, 2018b; Zeyl and Johnston, 2016). Therefore, extratympanic transmission of sound energy to the inner ear is sufficient to confer auditory sensitivity in air, even to fully aquatic vertebrates that lack specializations for aerial pressure detection (Christensen-Dalsgaard et al., 2011; Christensen et al., 2015a). It follows that these extratympanic pathways may represent a general mechanism for hearing that could support aerial sound sensitivity in the absence of a tympanic ear, and potentially could represent the earliest mechanism for terrestrial hearing in tetrapods. Extratympanic mechanisms also contribute to the low-frequency, directional auditory response in tympanate species (Christensen-Dalsgaard and Jørgensen, 1996; Jørgensen and Christensen-Dalsgaard, 1997a). Thus, extratympanic hearing mechanisms may provide a non-tympanic basis for sound localization that could relax selection on middle-ear structures to generate the varied forms of the atympanic ear observed in extant terrestrial vertebrates.

We normally associate ears and auditory structures with sound detection, but sound is only one of several kinds of mechanical disturbances that they can respond to. The proximate stimulus of the hair cells in the vertebrate inner ear is not sound, but fluid motion, and any stimulus that can excite such fluid motion, for example by vibrating the whole animal or its skull, can potentially stimulate the hair cells and be perceived equivalent to a sound stimulus. Thus, two different pathways exist for non-tympanic sound sensitivity: sound can vibrate the substrate and these substrate vibrations can stimulate the inner ear, or sound can vibrate the animal.

Many animals are highly sensitive to substrate vibration and use this sensitivity for communication (for example, by percussive impact with the substrate), predator avoidance and prey localization. Transmission of substrate vibration depends on the properties of the substrate, and the most biologically important types of vibrations in soil propagate in the surface of the substrate, such as Rayleigh and Lowe waves (see Glossary) (Narins, 1990). These types of waves propagate with dispersion and relatively low velocity depending on soil structure, approximately 100 m s⁻¹ in moist soil (Lewis and Narins, 1985; Narins, 1990), and mostly consist of low frequencies after propagation (a few hundred hertz). Other types of vibrations are prominent in thin structures, such as plant stems and leaves. Bending waves (see Glossary), for example, are used for both signaling and predator avoidance in many insect species (reviewed in Cocroft and Rodriguez, 2005) and in vertebrates, as demonstrated by the various vibration behaviors of tree frogs (Caldwell et al., 2010; Narins et al., 2018; Warkentin, 2005).

Alternatively, whereas vibration signals that are used for communication are usually generated by direct mechanical impact, ensonification (see Glossary) of the substrate will also produce vibrations. The utility of this pathway is probably limited for animals on soil where the energy loss as a result of reflection would be substantial and the induced vibrations therefore small, but for animals on less dense substrates, for example a small animal sitting on a leaf, the induced vibrations might be considerable. To our knowledge, only one recent study on spider webs as an accessory sound-receiving structure has investigated this possibility for hearing through the substrate (Zhou et al., 2022). The spider web collects sound energy over a larger area and transmits vibrations to the sensory hairs of spiders, analogous to the workings of the tympanic ear. Such a pathway through large leaves could also be relevant for treefrogs or small urodeles. Leaf vibrations, generated by calling frogs, can enhance sound radiation (Muñoz and Halfwerk, 2021) and influence calling behavior (Narins et al., 2018), but the induced vibrations in leaves could also enhance sound perception.

The general mechanism for non-tympanic hearing is that the head or body of the animal is vibrated by sound and that vibration is conveyed to the fluids in the inner ear. For sound pressure to generate translational motion (i.e. vibrations) in an animal, the acoustical size of the receiver must be small relative to the wavelength of the impinging sound (see Box 2). Therefore, low-frequency sounds are more effectively transmitted via extratympanic mechanisms. In humans, such extratympanic mechanisms are collectively referred to as ‘bone conduction’ but, evidently, there are several different pathways of non-tympanic stimulation of the inner ear. Human bone conduction is usually studied by applying local vibrations to the head, either by placing a vibrating tuning fork on the head or more recently by variable frequency bone vibrators. This mode of stimulation is not completely comparable to extratympanic stimulation with free-field sound (see Glossary), as sound interactions with the body are not evaluated; thus, it is not straightforward to relate bone conduction thresholds to thresholds to naturalistic, free-field sound. However, the thorough study of human bone conduction allows the characterization of three different frequency-dependent vibration modes (reviewed in Stenfelt, 2013). At frequencies below 400 Hz, the head is translated as a rigid body [ka<0.63 (where k is the acoustic wavenumber and a is the radius of the animal's head), assuming a head diameter of 0.17 m]. At intermediate frequencies (400–1000 Hz; 0.63<ka<1.6), the skull vibrates similarly to a mass–spring system where large components move in phase. Frequencies above 1000 Hz produce longitudinal waves in the base of the cranium and bending waves in the cranial vault (Stenfelt, 2011). From these data, we may infer that the low-frequency translational mode is probably very general for animals stimulated at ka<1, whereas the higher frequency modes are dependent on the structural properties of the cranium.

Box 2. Calculation of the acoustical size of an animal

The interaction of acoustic energy with an animal is frequency dependent and may be determined by relating the acoustic wavenumber of a sound (i.e. the amount of phase change a sinusoidal sound wave undergoes per meter) to the size of the animal, i.e. ka where k is the acoustic wavenumber and a is the radius of the animal's head. The acoustic wavenumber is defined as:

where f is frequency, λ is wavelength, and c is the velocity of the propagating sound.

Models of the interaction of sound with a rigid body predict that the object will be translated with a vibration velocity proportional to sound pressure, but independent of frequency when ka<<1 (Morse, 1936; von Békésy, 1948; reviewed in Christensen-Dalsgaard et al., 2022). The vibration velocity depends mainly on the density and shape of the object and will be parallel and antiparallel to the direction of sound propagation. The velocity amplitude according to different models ranges from 3 to 6 µm s⁻¹ Pa⁻¹ for an object with the same density as water, assuming no friction. To relate this to tympanic hearing, the velocity amplitude of the tympanum in most sensitive terrestrial tetrapods is approximately 1 mm s⁻¹ Pa⁻¹ (Christensen-Dalsgaard and Manley, 2013), i.e. the rigid body translation is approximately 200 times (46 dB) lower, which is in the range of human conductive hearing loss (see Glossary; 40–60 dB; Reinfeldt et al., 2007). In small animals (ka<1), hearing by rigid body translation constitutes a ‘baseline’ low-frequency sensitivity that can be increased by resonant structures (for example in air-filled body cavities, described below). The resonant properties of these structures can enhance vibrations of the inner ear and would also decrease the density of the organism, thereby increasing its responsiveness to sound pressure.

Sound-induced body vibrations, transmitted either via the substrate or by direct interaction of the head and body with sound, are just the entry point for vibrations that must then be transmitted to the inner-ear sensory cells for them to be perceived. In inner-ear otolith organs, body vibrations will be transduced directly by inertial lag of the denser otolithic mass, leading to bending of the sensory hairs, but papillar end organs are stimulated indirectly through induced vibrations in the inner-ear fluids. Evidently, vibration of the cranial bones can travel to the otic capsule and generate inner-ear fluid vibrations, but the low-frequency translational mode may also stimulate the inner ear either by inertial movement of middle-ear structures (when present) or by fluid inertia. Fluid inertia, the lag of inner-ear fluid movement relative to the surrounding structures, depends on the pressure-release windows in the otic capsule and may contribute to directionality: movements parallel to the axis of the pressure-release windows will be larger than movements perpendicular to this axis. Interestingly, in humans, fluid inertia may dominate the cochlear response to bone-conducted sound, at least at low frequencies where inertial movement of the middle-ear ossicles is insignificant (Stenfelt, 2015).

A factor that might contribute to the importance of extratympanic sound reception in amphibians is that amphibians have a unique, additional movable element in the oval window. This is the operculum, connected via the opercularis muscle to the scapula (Fig. 3). There are different hypotheses for the function of the operculum (see Mason and Narins, 2002). Initially, it was proposed that substrate vibrations could travel through the front legs of amphibians to the scapula and then through the opercularis muscle to the operculum (Kingsbury and Reed, 1908). However, amphibians generally are in very close contact with the substrate and the wavelength of seismic waves at the frequencies of interest (usually below 200 Hz) is large compared with the size of the animal, so the whole body is probably being accelerated in phase. Alternatively, when the animal is translated by the sound wave, the head will move more freely than the rest of the body, which is restricted by friction. The optimal place to register this tension is between the scapula and operculum, i.e. via the opercularis muscle (Hetherington, 1985; Hetherington et al., 1986). The original function of the operculum may therefore have been sound reception, providing a pathway to transduce sound-induced body vibrations into fluid movement in the inner ear. This pathway may have originated before the tympanic apparatus, as suggested by the earlier occurrence of the operculum during ontogeny and its near ubiquity in the recent amphibians, and thus could constitute an early ‘experiment in hearing’.

Additional non-tympanic inputs to the auditory system include resonant air-filled body cavities that can transform sound pressure into particle motion. In ray-finned fishes, pressure sensitivity is enhanced by gas-filled chambers such as swimbladders or specialized air bubbles that are acoustically coupled to the inner ear. These cavities act as harmonic oscillators when ensonified and their pressure-to-displacement capability (see Glossary) extends aquatic sound detection into the far-field, lowers detection thresholds and increases the audible range of frequencies depending on the size and shape of the cavity (reviewed in Ladich and Schulz-Mirbach, 2016). Although the amplitude of particle displacement decreases with distance from these resonating structures, even species that lack specialized connections between their air cavities and inner ears are able to detect sound pressure under water (Christensen et al., 2015a; Jerkø et al., 1989). Air-filled body cavities may have been present in the common ancestor to actinopterygians and sarcopterygians and could have enhanced underwater hearing via pressure detection in a similar manner. These cavities were likely quite small prior to the evolution of paired lungs and would have resonated at high characteristic frequencies. However, this form of pressure sensitivity could have provided a substrate for the evolution of auditory papillae capable of transducing the high-frequency particle motion radiating from these air-filled cavities.

Cavity resonance may also influence aerial sound pressure detection in earless tetrapods. In snakes, the lung is hypothesized to play a role in transmitting body vibrations to the ear (Hartline, 1971), although this has not been empirically demonstrated. In lizards and frogs, the lateral body surface overlying the lungs is highly responsive to sound, especially in response to low frequencies (Hetherington, 2001; Hetherington and Lindquist, 1999). However, recent studies using salamanders indicate that cavity resonance is not necessary for aerial sound pressure detection (Capshaw et al., 2020; Zeyl and Johnston, 2017). In these small species, enclosed body cavities likely resonate far beyond the upper frequency limit of their auditory end organs. This does not mean that cavity resonance is irrelevant for hearing – in earless sooglossid frogs, modeling indicates that the mouth cavity resonates near the dominant frequency of its call, and the close proximity of the mouth to the inner ear in this species suggests that it may effectively convert pressure to particle motion within the inner ear (Boistel et al., 2013). Additionally, the resonant frequency of the lungs in the earless pumpkin toadlet closely matches its vocalizations at approximately 3–4 kHz (Goutte et al., 2017). Although evolutionary degeneration of the basilar papilla and its innervation has rendered the pumpkin toadlet largely insensitive to high frequencies (Goutte et al., 2017), other earless species that retain a functional basilar papilla may benefit from cavity resonance. For example, earless and partially eared harlequin frogs demonstrate enhanced sensitivity to frequencies greater than 2 kHz (Fig. 2) that may be mediated by cavity resonance and transduced by well-developed auditory epithelia (Lindquist et al., 1998; Womack et al., 2018b).

In tympanate frogs, lung vibrations are transmitted to the ear via the mouth, which is broadly confluent with the middle-ear cavity. Through this pathway, vibrations of the ensonified lungs can interact with the inner surface of the tympanum to influence its sound-induced oscillation. Although initially this pathway was hypothesized to enhance hearing, lung vibrations appear to cancel out certain sound frequencies at the eardrum to provide a natural bandpass filter (see Glossary) at the peripheral level of the anuran auditory system (Christensen-Dalsgaard et al., 2020; Lee et al., 2021).

Traditional views of aerial hearing emphasize the evolution of a dedicated pressure-relief window (see Glossary) as a prerequisite for acute sensitivity to high-frequency sound pressure. This is because the inner-ear fluids are considered incompressible and thus require compliant input/output fenestrations of the otic capsule (the oval and round windows, respectively) for acoustic energy. Despite this, clear evidence for dedicated pressure-relief windows appears to lag the emergence of tympanic middle ears in the fossil record by nearly 50 million years (Müller et al., 2018). For example, the stem-archosaurs Euparkeria and Prolacerta are inferred to have been tympanate, yet both lacked an ossified otic fenestration that could be considered homologous to the round window of crown group archosaurs (Clack, 1997; Gower and Weber, 1998; Sobral et al., 2016). In early tympanate tetrapods, pressure relief may have occurred via the metotic foramen and perhaps through the largely unossified medial wall of the otic capsule.

Extant vertebrates that lack dedicated pressure-relief windows, such as salamanders, caecilians, Sphenodon, snakes, chameleons and many burrowing lizards, may rely on alternative pathways to reduce impedance to fluid motion within the ear. In salamanders, for example, the perilymphatic foramen in the medial wall of the otic capsule may provide a compliant pathway through which pressure can dissipate into the endocranial cavity (Box 1) and potentially through to the opposite ear (Wever, 1978b). The close proximity of these foramina to the amphibian papillae supports a fluid circuit that generates high-amplitude displacements at the location of the sensory epithelia (Smith, 1968). In caecilians and squamates that lack round windows, fluid pressure relief may occur via a re-entrant pathway in which vibrations in the perilymph make a complete circuit within the otic capsule and dissipate at the lateral surface of the stapes (Baird, 1970; Wever, 1973, 1978a). This re-entrant pathway also provides crucial fluid pressure relief within the otic capsules of turtles that lack round windows despite having tympanic middle ears (Wever, 1978a).

Various otic fenestrations may also influence hearing in tympanate tetrapods by providing ‘third windows’, in addition to the oval and round windows, through which acoustic energy can interact with the fluid-filled labyrinth. Similar to the amphibian perilymphatic foramen, the cochlear aqueduct connects the inner-ear perilymph to the endocranial cavity in mammals and archosaurs. The size of the cochlear aqueduct is highly variable among diverse species and is notably hypertrophied among odontocetes, pinnipeds, cetaceans and waterbirds (Baird, 1974; Kohllöffel, 1984; March et al., 2016). For these species, large cochlear aqueducts may boost bone conduction hearing by acting as a low-impedance pathway to conduct pressure waves between the fluid-filled endocranial cavity and otic capsules and/or by facilitating stimulation via fluid inertia (Freeman et al., 2000; Tonndorf, 1966). Models of pathological human cochleae (e.g. those with semicircular canal dehiscence – see Glossary) indicate that additional compliant windows in the otic capsule can improve sensitivity to frequencies below 1 kHz by up to 14 dB (Stenfelt, 2015). These ‘third windows’ may play a particularly important role for low-frequency hearing and, potentially, infrasound detection.

Detection of infrasound – sounds below 20 Hz – has been demonstrated in some mammal and bird species and may be used for navigation in birds (Zeyl et al., 2020). The actual mechanism of infrasound reception is unknown, but it has been proposed that whole-body vibration of birds could be more important than tympanum vibration, which is limited by membrane stiffness at low frequencies (Zeyl et al., 2020). Possibly, the mechanisms for infrasound sensitivity are quite similar to the extratympanic mechanisms for aerial sound sensitivity described in this Review. Whole-body vibration is assumed to be transmitted to the inner ear, either special regions of the basilar papilla or the otolith organs, via a translational mode of stimulation. It is hypothesized that pressure-release windows, for example the cochlear aqueduct or various other ‘third windows’ in the otic capsule, could increase sensitivity to infrasound by reducing overall impedance to fluid motion (Zeyl et al., 2020).

A general hypothesis proposed in this Review is that separation of sound and vibration stimuli at the sensory level is difficult. In anurans, fibers from both the low-frequency hearing organ (amphibian papilla) and the vibration-sensitive organ (sacculus) have a dual sensitivity to low-frequency sound and vibration (Christensen-Dalsgaard and Narins, 1993; Moffat and Capranica, 1976; Yu et al., 1991). However, fibers from the two organs can still be distinguished based on their relative sensitivity to substrate vibrations (i.e. the difference between thresholds to substrate vibrations and to sound-induced vibration) (Christensen-Dalsgaard and Jørgensen, 1996; Christensen-Dalsgaard and Narins, 1993). Separation between sound and vibration could be based on the relative excitation of the different inner-ear organs, and upon differential projections to the central nervous system. In frogs, for example, otolith organs and auditory papillae project to different nuclei with very different projection patterns in the brain, e.g. to the dedicated nucleus saccularis and the dorsal medullary nucleus (McCormick, 1999). For animals communicating with sound, substrate vibrations will usually be lower frequency than communication frequencies; thus, these differences in processing and projections may reflect the different behavioral significance of substrate-borne vibration and sound.

Where it is advantageous for an animal to discriminate between airborne sound and substrate vibrations, for example to detect predators, stimulus direction might prove useful. The most prominent substrate vibrations (Rayleigh waves – see Glossary) generally will vibrate the animal in a dorso-ventral direction, which matches the orientation of the saccular macula in amphibians (Christensen-Dalsgaard and Jørgensen, 1988), whereas sound-induced vibrations generally will occur along the horizontal plane. In amphibians, where sensitivity to sound-induced body vibrations is lower for sounds from frontal directions (Capshaw et al., 2021; Jørgensen and Christensen-Dalsgaard, 1997a,b), auditory fibers might similarly be less sensitive to sound coming from dorsal or ventral directions, although this has not been investigated.

In fish otolith organs, directionally polarized sensory hair cells are organized to allow inherent sensitivity to multiple sound source directions, albeit with 180 deg ambiguity (Fig. 4) (Edds-Walton and Fay, 2009; Walton et al., 2017; Zeddies et al., 2012). It is reasonable to suppose that the first (amphibious) tetrapods were also sensitive to directional cues when they were in water. When these early tetrapods moved onto land, it is likely that they would have retained neural circuits sensitive to sound source direction, which would have worked well in water but less so in air (Carr and Christensen-Dalsgaard, 2016).

According to our hypothesis of extratympanic hearing via translation by sound, the response to sound-induced vibrations would be quite similar in air and water in an otolith-based ear. The inner-ear response in a terrestrial animal with non-otolithic papillae, however, would reflect the fluid inertia movement and not the actual movement of the body, necessitating a different processing of sound direction. Recent work using salamanders found that directional interaction of airborne sound pressure with the salamander's head generated a figure-eight pattern of sensitivity, with greatest responses to sound presented along the mediolateral axis, or in-plane with the location of the input/output fenestrations of the otic capsule (Capshaw et al., 2021). These findings from atympanate salamanders are reinforced by data from frogs, where extratympanic sensitivity produces similar directional responses in low-frequency auditory neurons (Fig. 4) (Jørgensen and Christensen-Dalsgaard, 1997a,b).

How do vertebrates disambiguate these (nearly) symmetrical directional responses? In ray-finned fishes, binaural neural comparisons of phase differences generated by direct stimulation and indirect stimulation of the ear through secondary inputs (e.g. acoustic energy re-radiated from resonant air cavities) may enable the animal to resolve sound source locations (Hawkins and Popper, 2018). These secondary inputs, combined with differences in rotation of the otoliths in either ear, can potentially generate large phase differences between the two ears. The lung and mouth cavities may play a similar role as secondary inputs to the auditory system of terrestrial species, although this remains to be tested. In amphibians, the perilymphatic foramen may also serve as a pathway for acoustic energy to stimulate both ears via the endocranial space, one directly and the other indirectly (Wever, 1978b). This may allow disambiguation of sound source location via binaural computation of phase and timing differences across the two ears in a manner similar to that observed in tympanate species with internally coupled ears.

If the central nervous system of early tetrapods was organized in a similar fashion to that of fishes, then binaural neurons in the medulla like those found in toadfish could have acted to further sharpen directional responses from the periphery (Carr and Christensen-Dalsgaard, 2016; Edds-Walton, 2016). Similar binaural interactions in the anuran dorsal medullary nucleus convert the figure-eight pattern of the low-frequency response in the auditory nerve to an ovoidal pattern that allows disambiguation of sound source location (Christensen-Dalsgaard and Kanneworff, 2005), although the neural mechanism underlying this directional sharpening is not resolved. This model of directional hearing differs from the de novo computation of binaural cues found in birds and mammals; nevertheless, it may represent an ancestral mechanism for localization that preceded the water-to-land transition.

Although the aerial auditory sensitivity of an atympanic ear is reduced relative to its tympanic counterpart, insight from extant atympanate species indicates that the early tetrapod ear should have been able to detect low-frequency acoustic cues and potentially localize sound sources within its surroundings. Shared mechanisms for sound detection among extant atympanate taxa indicate that sound pressure sensitivity of the ancestral tetrapod ear was probably mediated by bone conduction of acoustic energy, where sound-induced head vibrations could have been detected by the auditory organs of the inner ear. Extratympanic pathways would have thus conferred a rudimentary ability to detect acoustic energy on land prior to the evolution of the specialized pressure-transducing tympanic middle ear. This translational mode of sound detection is analogous to the well-studied mechanism for bone conduction in humans, a special case of extratympanic hearing that can provide sensitivity to airborne sound that bypasses the tympanic ear and is used in assisted hearing devices, such as bone conduction hearing aids.

Extratympanic pathways for sound generate vibrations in the animal that vary in amplitude with the incident angle of the sound source, and so this mechanism for hearing confers a peripheral basis for directionality in an atympanic ear. The ancestral tetrapod ear probably possessed acoustically sensitive end organs with directionally polarized hair cells similar to those found in aquatically adapted vertebrates. On land, the location and orientation of these hair cells relative to the pressure-relief windows of the otic capsule could enable maximal stimulation of the inner ear by a pressure wave traveling in-plane with these compliant windows.

The particle motion response of the fish saccule and the low-frequency auditory response of frogs and salamanders all form a figure-eight pattern of directionality at the peripheral level of the auditory system. Further study of how these taxa resolve the 180 deg directional ambiguity problem is crucial to understanding sound source localization in these species. Future experimental work should focus on the underlying neural correlates for sharpening the directional response in atympanate species. Anatomical studies such as tract tracing could be useful to reveal bilateral projections between the medullary nuclei that may serve to refine the directional response. It is possible that contralateral inhibition, as observed in ray-finned fishes and frogs, is an ancestral neural mechanism for transforming the ambiguous figure-eight pattern of directionality conferred at the peripheral level to a more ovoidal pattern in higher processing centers. Identification of similar neural pathways in other atympanate species such as salamanders would provide support for an early origin of binaural projections between brainstem nuclei. This could potentially reveal a key ancestral feature of the vertebrate ear enabling sound source localization in the early terrestrial tetrapods. Further, behavioral studies of sound localization in atympanate species, for example those leveraging phonotaxis to airborne sound pressure in earless frogs, are an important next step for understanding the functionality of bone conduction hearing in terrestrial vertebrates.

The massive convergent evolution of tympanic ears, an evolutionary novelty in the tetrapods, emphasizes the benefit of the impedance-transformed ear – providing not only an increase in aerial sensitivity by up to 50 dB but also new mechanisms for directional hearing through coupling of the tympana. The atympanic ears of recent species represent a useful proxy for the predecessors of tympanic ears and may provide an improved understanding of the structures and mechanisms that gradual natural selection could transform into this decisive novelty.