From the morphospace to the soundscape: Exploring the diversity and functional morphology of the fish inner ear, with a focus on elasmobranchs^a)

The inner ear is the basal organ responsible for hearing and three-dimensional (3D) positioning in vertebrates. Compared to other vertebrate taxa, there is an astonishing diversity in fish inner ear morphology (Ladich and Schulz-Mirbach, 2016; Retzius, 1881; Weber, 1820). The inner ears in fishes are typically composed of a membranous labyrinth, enclosed in an otic capsule. The labyrinth enfolds three semicircular canals (posterior, anterior, and horizontal) and three end organs (saccule, utricle, and lagena). In addition, some species also possess a diminutive fourth end organ, the macula neglecta (MN), typically found in cartilaginous fishes. Each canal and end organ house a sensory epithelium composed of hair cells (macula), linked to the auditory nerve. Furthermore, each end organ macula, except the MN, is loaded with a dense calcareous mass—an otolith or otoconia. The high variability in their inner ears relates to distinct morphological elements: size and area/volume of the end organs, semicircular canals, and sensory epithelia (relative to the size of the fish and its brain); presence, mass, and shape of otoliths or otoconia; and orientation patterns of the hair cells (Ladich and Schulz-Mirbach, 2016) (Fig. 1). The variability is also represented by the hearing abilities of fishes, which differs both spectrally (frequency ranges) and in sensitivity (intensity levels detected) (Ladich and Fay, 2013; Popper , 2019). Furthermore, while all fishes can detect the particle acceleration component of a sound, some bony fishes with gas-filled structures, such as swim bladders, may also indirectly detect acoustic pressure, typically increasing their overall hearing sensitivity (Popper , 2022). Other accessory hearing structures may assist bony fishes in detecting pressure, like extensions of the swim bladder coming directly in contact with the skull (e.g., Nelson, 1955; Schulz-Mirbach , 2012), and the so-called Weberian ossicles in the otophysan group that transmit vibrations of the swim bladder to the inner ears (Weber, 1819, 1820). These ancillary structures also reveal a large diversity and contribute to the complexity of describing the hearing structures in fish.

The evolutionary and functional implications of this particularly high degree of variability are largely unknown and has been presented as one of the biggest mysteries of fish sensory biology (Ladich and Schulz-Mirbach, 2016; Retzius, 1881; Schulz-Mirbach and Ladich, 2016). It is widely accepted that sensory system evolution occurs due to differential selection pressures associated with stimuli detection from a given environment and hetero- and conspecifics (e.g., Endler, 1993; Endler and McLellan, 1988). As such, the staggering variation of animal sensory system morphologies underscores how differently individual species perceive their external world. Therefore, it is expected that fish inner ear diversity serves in audition and balance, and that some functions are facilitated by each shape and/or pattern type.

Sensory system diversity evolves through a myriad of mechanisms. The sensory drive hypothesis, devised by Endler (1992), postulates that natural selection would favor signals that effectively stimulate the sensory systems of intended receivers, and that signal effectiveness is strongly influenced by transmission and noise properties of the environment. Signals and sensory systems would, thus, coevolve. However, most of the evidence for this hypothesis originates from the sense of vision (Seehausen , 2008) and its role in shaping acoustic systems is less well evidenced (Wilkins , 2013).

In 2013, Ladich presented a similar hypothesis, named the “eco-acoustical constraints” hypothesis, arguing that hearing evolved as an adaptation to the acoustical conditions in fishes' habitats (Ladich, 2013; Ladich and Schulz-Mirbach, 2016). It is grounded in the view that the development of any species' auditory system and sense of hearing is the analysis of the surrounding “auditory scene” (Bregman, 1990; Fay and Popper, 2000). Hearing sensitivities would, theoretically, be as low as possible, with acoustic specializations, so long as biologically relevant sounds are not masked by environmental background noise (Ladich, 2013). As the ambient background sound varies in aquatic environments, it would inevitably result in a large variety of hearing abilities in fishes. Low background levels would then facilitate the evolution of morphological structures for hearing enhancement and the detection of low-level sound, whereas high background levels make specialized structures meaningless. This hypothesis further postulates that advanced hearing sensitivity may have evolved independent of the ability to produce sounds for acoustic communication (Ladich, 2013). According to the eco-acoustical constraints hypothesis, fishes optimize their ability to pick up acoustic information emanating from abiotic (sounds generated by the physical environment, such as wind and waves) and biotic sources (sounds generated by animals, such as vocalising fish) under all prevailing acoustical conditions.

However, very few investigations have attempted to unpack the multifaceted drivers of sensory diversity or have empirically tested these hypotheses, likely due to a number of reasons:

1. There is still little understanding of the basic mechanism of fish hearing. For example, while fish inner ear contains three to four maculae, the role of each of these maculae in hearing function is partly uncertain. The saccule has long been assumed to be the main auditory end organ in modern bony fishes (Fay and Edds-Walton, 1997; v. Frisch and Stetter, 1932; Lu , 2002; Lu and Xu, 2002), accompanied by the MN in cartilaginous fishes (Corwin, 1981, 1989), but recent studies have demonstrated a hearing role from all end organs (saccule, lagena, and utricle) in the model teleost plainfin midshipman (Porichtys notatus) (Bhandiwad , 2017; Rogers and Sisneros, 2020; Vetter , 2019). Furthermore, the mechanism of otolith/otoconia motion and their role in stimulating the hair cells of the macula is still not fully resolved. This paucity of information hinders further assumptions about any structure-to-function link, making it difficult to present and test hypotheses.

2. Studies of hearing function in fishes have mainly focused on bandwidth (range of detected frequencies) and sensitivity (lowest detectable sound level) measurements. This restricted focus may not reflect the full hearing function in fishes. While considering fish behaviour within the “acoustic scene,” measuring an organism's capability to detect relevant signals (sound discrimination) and to determine their location (sound localisation) is as important, if not more, than measuring its hearing thresholds (Fay, 2009). The ability to localise sound sources is likely a primary selective pressure shaping hearing organs, as suggested in mammals (Masterton, 1974); however, very few studies have measured the relative performance of sound localisation across different fish species, and the underlying mechanisms of how any fish accomplishes this task have not yet been completely resolved (Hawkins and Popper, 2018). This is notably due to the lack of standardized measurements for assessing sound localisation in the large and diverse range of fish species: in mammals, this has been achieved by calculating the “minimum audible angle,” which is the smallest angular separation at which two sounds are perceived as coming from distinct sources (Heffner and Heffner, 2016).

3. There is a notable difficulty in collecting empirical data on the hearing system of live fishes, typically involving invasive physiological experiments [although note the non-invasive auditory brainstem recordings introduced in fish by Kenyon (1998)]. Furthermore, while gross dissections and histological techniques have been invaluable to build the current knowledge about fish inner ear microstructures, these methods are also usually invasive and can distort or destroy the relative positions of tissues, which increases the complexity of analysing the inner ear in situ.

4. The modular nature of the inner ear (in fishes and other vertebrates) makes its investigation more complicated. The inner ear is not only the primary structure serving audition, but also includes the vestibular system, which allow fishes to gain information about their body position and motion in the 3D space (balance and spatial orientation). Particularly, the semicircular canals and their gelatinous cupula, and the utricle, are thought to provide the basis of vestibular function (Kasumyan, 2004). As such, the evolution of “modules” within the inner ear structure may have allowed parts of the structure to specialize for different or both functions (vestibular and hearing). In fishes, the inner ear end organs seem to serve both senses at once, which has been called the “mixed function hypothesis” (Platt and Popper, 1981). Therefore, including this modularity as a potential mechanism underlying the morphological diversification of the inner ear in fishes is an added complexity.

5. Fishes may have undergone adaptive divergence in inner ear shapes as they evolved different “auditory modes” in different environments: it is logistically difficult to identify and untangle the factors responsible for these shifts in a phylogenetical context. A range of intrinsic and extrinsic influences may have triggered the development of pronounced differences in fish inner ear morphology and auditory capabilities. In addition to access to a new environmental niche or novel resources, diverse principles (developmental, constructional. and functional) ultimately determine the range of morphologies that can evolve (Wainwright, 2007). Most studies explore one single factor at a time, usually dictated by their own discipline, thus limiting the overview of the evolutionary process. A wider view of sensory ecology, on a broader interdisciplinary scale, may be beneficial to investigate organism–environmental interactions.

Here, we argue that we now have some of the tools and novel capabilities to test the eco-acoustical (and other) hypothesis. We need to design original studies to fully take advantage of current technologies and leverage open science and big data platforms. To simplify the system, we focus on the diversity present in the group of elasmobranch fishes (sharks, rays, and skates), a group of iconic cartilaginous fishes which are understudied and represent early stages of evolution of vertebrate hearing (Chapuis and Collin, 2022). Elasmobranchs represent a tractable system compared to other fish clades, as they are uniquely phylogenetically located and encompass more than 1000 species occupying all habitats of the world's oceans. They rarely use sound to communicate and lack any accessory hearing structures (e.g., swim bladders), making them ideal as a basal model to study the functional morphology of their hearing system.

Understanding the functional morphology of the inner ears in fishes is the missing scaffold for investigating many unresolved questions. In particular, the following questions summarise the knowledge gap:

• Which factors drive pronounced interspecific differences in inner ear morphology and hearing abilities in fishes?

• What elements or microstructure of the fish inner ear are involved in distinct functions (hearing bandwidth, sensitivity, discrimination, localisation, and/or vestibular functions)?

• How plastic and adaptable is the hearing system of fishes, both across ontogeny and over multiple generations, in an era characterised by anthropogenic changes (e.g., climate change and noise)?

By comparing auditory structures between individuals or species that differ in their ecology, a few studies have been exploring the relationship between morphology, ecology, and functions of the ear in bony fishes (i.e., structure–function relationships) (Aguirre and Lombarte, 1999; Lombarte and Popper, 2004; Norton , 1995). However, researchers have not yet been able to provide studies that definitively correlate hearing organ shape, size, and complexity to hearing performance in fishes. This suggests that the current available metrics (morphology, function, and/or ecological variables) have not yet allowed us to fully understand inter- and intra-specific differences. Excitingly, cutting edge technology and analytical tools at the forefront of a range of medical and scientific fields have emerged, thereby enabling a deeper investigation of relationships between the structure and the function of complex structures (Muñoz and Price, 2019), including the hearing organ of fishes. Here, we briefly describe novel techniques and opportunities, with the hope to motivate further studies exploring this important sensory modality in an equally important group of vertebrates.

Bioimaging techniques, such as computed tomography (CT) and magnetic resonance imaging (MRI) are unique in their ability to non-invasively acquire high-resolution, digital, 3D data. First, they render a geometrically accurate, 3D model of the whole anatomical structure. Second, the data are collected non-invasively, without dissection and therefore, without damaging the specimen, leaving all the microstructures in their original relative position (i.e., in situ). Finally, the resulting anatomical data can be processed into surfaces and volumes that allow for species comparative analyses and quantification and exportation for further mathematical modelling (Yopak , 2018).

Micro-CT (μCT) is a powerful, non-invasive technique for visualising an animal's internal structures in situ, with micrometre level resolution (∼1 μm). While the skeletal part of the fish inner ear (i.e., the skeletal labyrinth) has been successfully imaged in a few studies (e.g., Maisey, 2001; Maisey and Lane, 2010; Pfaff , 2019a; Schnetz , 2016; Staggl , 2022), until recently, soft-tissue imaging (such as that of the nervous system, including the membranous labyrinth of the inner ear) has been difficult to acquire with CT, due to the inherently low X-ray absorption of non-mineralised tissues. An aqueous solution of Lugol's iodine (I₂KI) has been demonstrated as a highly effective agent for rapidly differentiating many types of soft tissues, such as the brain and other aspects of the peripheral and central nervous systems (Gignac , 2016). This method has been successfully used in fishes (both bony and cartilaginous) to extract high-resolution 3D data from a range of neural structures, including the brain, cranial nerves, and inner ear structures (Camilieri-Asch , 2020b; Camilieri-Asch , 2020a; Schulz-Mirbach , 2012) (Fig. 2). An important limitation of using Lugol is that the staining process causes severe soft-tissue shrinkage (Vickerton , 2013), which can be reduced if using a buffered Lugol's solution (B Lugol) (Dawood , 2021). Another contrast agent, phosphotungstic acid (PTA), has also been successfully used to stain fishes for μCT, even allowing for the visualisation of sensory epithelia within the end organs (Schulz-Mirbach , 2013), and seems to provide reduced tissue shrinkage during staining than Lugol (Greef , 2015).

MRI is a technique that uses a magnetic field and radio frequency waves to create images of soft tissue structures. It allows investigators to capture high-resolution (∼10–100 μm) 3D data of soft tissue structures in situ in all animals, including fishes (Berquist , 2012). It has been used to investigate the nervous system of fishes, particularly in comparative brain morphology (Ullmann , 2010b, 2010a; Ullmann , 2011; Yopak , 2016; Yopak and Frank, 2009). Contrary to diffusible iodine‐based contrast‐enhanced computed tomography (diceCT), MRI does not seem to cause much tissue distortion or shrinking, allowing the use of the imaged tissue for further investigations (e.g., histology). However, MRI is less cost effective than CT and has only recently been used to visualize elasmobranch fishes' inner ears in situ (Sauer , 2023).

Open access repositories of 3D dataset for numerous fish species already exist and allow the use for any scientific or educational study, like on the MorphoSource online platform (MorphoSource, www.MorphoSource.org). It is also possible to combine MRI and CT datasets with image co-registration, a process of finding the mathematical transformation that aligns several different radiographic studies (e.g., Simões , 2012). While MRI and CT scans are usually acquired in different orientations and resolutions, and both generate different deformations and artefacts that complicates the co-registration, automatic algorithms are now available to overcome these challenges. In this way, high-resolution reconstructions of the whole nervous system of fishes in situ could be collected for a more comprehensive analysis.

Bioimaging can also provide a direct, non-invasive view on the hearing structures “in motion” at a high spatiotemporal resolution. Pioneering studies provided the first multidimensional insights into the sound-induced motion of the diverse auditory structures (otoliths, as well as accessory hearing structures like auditory ossicles and the swim bladder walls) of freshly euthanised bony fishes (Maiditsch , 2022; Schulz-Mirbach , 2018; Schulz-Mirbach , 2020). Instead of using a conventional CT, in which the contrast is due to attenuation differences within the sample and the time of image acquisition can be slow, these studies used synchrotron phase microtomography, in which the contrast originates from the phase shift of the X-ray beam passing through the matter. Dynamic or high-speed imaging as a function of time is also possible due to the comparatively high X-ray flux. It thus allowed the visualization of the motion patterns of the otoliths inside the inner ear during sound presentations both in two-dimensional (2D) and in 3D (Maiditsch , 2022; Schulz-Mirbach , 2018; Schulz-Mirbach , 2020). Therefore, insights into the role of each end organ inside the fish inner ear may be possible using state-of-the-art bioimaging facilities, as well as in situ investigations at exceptionally high resolution. In humans, such high-resolution synchrotron imaging has revealed cytoarchitecture of the cochlea similar to histological investigations (Iyer , 2018).

The evolutionary history of vertebrates is characterised by extensive morphological changes, appearing in response to changes in environmental conditions and generally coupled with the reorganization of genetic pathways (Wainwright, 2007). Geometric morphometrics is the study of morphological forms and shapes: an analytical method of quantifying and comparing 2D or 3D objects. It is based on 2D or 3D landmark coordinates that represent biologically or geometrically corresponding point locations on the measured objects. Combined with multivariate statistics and phylogenetics, geometric morphometrics can be used to analyse the relationship between shape and a variety of evolutionary, developmental, ecological, and functional traits.

With the development of bioimaging, geometric morphometrics and its associated methods have been rapidly advancing over the last decade. They produce large sets of autocorrelated data, creating a mathematical representation of the biological form. These data are then usually dimensionally reduced to be compared among taxa, typically with principal component analysis (PCA), allowing the study of the evolutionary diversification. Changes in morphology are thus mapped onto phylogenetic trees and can be linked with ecological or behavioural observations.

The few studies of the geometric morphometric of the fish inner ear have so far focused on examining the diversity of otoliths shapes (Chollet-Villalpando , 2019; e.g., Monteiro , 2005; Ponton, 2006). However, the motivation for most of these studies was often taxonomic differentiation rather than explaining different phenotypes. Furthermore, although not using geometric morphometric methodologies per se, several studies initiated both qualitative and quantitative comparison of inner ear shapes and morphometrics, in an effort to explain differences in relation to ecological variables. Evangelista (2010) compared inner ear 2D shapes and measurements in sharks and rays, concluding that the variation could best be explained by considering phylogeny as well as feeding strategy. Similarly, Deng (2013) studied the interspecific variation of inner ear structures in a deep-sea fish family, highlighting the possibility that specialized morphologies may enhance the auditory function in deep waters. Finally, while the different morphotypes found in the inner ears of fishes have been exhaustively summarized (Ladich and Schulz-Mirbach, 2016; Schulz-Mirbach and Ladich, 2016), no studies have yet used the latest advancements in geometric morphometrics to try to untangle explanatory factors for this high morphological diversity. We argue that geometric morphometrics should and will become a more prevalent investigative tool in the future toolkit of fish bioacousticians, notably due to the prevalence of high-resolution, multidimensional shape data and analyses (see bioimaging above), the advances in phylogenetic comparative methods, and the increasing availability of better resolved fish phylogenies.

In other taxa, the combined power of 3D morphological datasets and geometric morphometrics has already proven fruitful. For example, this methodology was used to study extinct marine reptiles and crocodile relatives with divergent aquatic lifestyles, for which inner ear geometry was associated with locomotor mode and habitat (Neenan , 2017; Schwab , 2020). Later, Hanson (2021) investigated how the inner ear of a much larger sample of extant and extinct reptiles correlated with locomotor ability and hearing acuity. Similarly, Evers (2022) showed that the ear labyrinth size in turtles was correlated with ecology, whereby small labyrinths were associated with terrestrial habits. Similarly, in eulipotyphlan mammals (moles, shrew, and hedgehogs), middle ear morphology strongly reflected the degree of subterranean lifestyle (Koyabu , 2017). As the degree of inner ear diversification is considered to be even larger in fishes than in other vertebrates, we predict that, with enough species represented, the technique will be powerful to test key adaptive hypotheses.

Finite element (FE) analysis is a numerical and engineering technique used to predict the performance of complex structures. In the past 15 years, the FE method has become a ubiquitous tool in the repertoire of evolutionary biologists (Kolston, 2000; Rayfield, 2007), although it has rarely been used to model the fish (or any other vertebrates') hearing system. FE analysis is notably often used in paleontology to predict the abilities of extinct taxa to withstand loads induced by a biomechanical function, like chewing, for example, or walking (e.g., Macho , 2005; Oldfield , 2012). As the extinct specimen in focus cannot be tested in vivo, digital versions of the morphology, materials, and loads can be modelled and analysed. In a FE model, the structure of focus (e.g., the hearing apparatus) is divided into discrete sub-regions of finite size—the “elements.” These elements are systems with linear algebraic equations and are bonded with each other at their vertices, called nodes. FE analysis uses this system of nodes to transform the model structure into a grid, termed mesh, which represents its geometry (Fig. 3). The mesh is assigned structural and material properties (e.g., Young's modulus and Poisson's ratio) which define the way the model will deform under stress. The “forces,” or load, are then applied to the FE model, which is to be solved by the software: in a hearing model, the forces would be represented by sound particle motion and/or sound pressure. FE analysis thus reconstructs stress (the amount of force per unit area experienced by tissues) and strains (the physical displacement of tissues) in a digital structure. The outcome is a series of nodal displacements, and the resulting strains and stresses are calculated and displayed. Their magnitudes reflect the mechanical behaviour of the structure and can be represented numerically and visually onto the model geometry. Most of the data used in FE analysis can both be based on representations or on real data: the mesh can be built from 3D data [e.g., from micro computed tomography (microCT)], the material properties can be measured from the real tissues if available, and the loads can be estimated realistically via experiments (in vivo or on the dead tissues) (Polly , 2016). Absolute values of loads can also be used depending on the question being asked, in a comparative context.

The FE method is used to estimate and compare biomechanical performance, implicated as selective factors in the evolution of morphological structures; thus, it addresses questions related to organismal morphology, function, and evolution. Both real and/or theoretical data and values can be used to feed the model, depending on the question. It is self-consistent and accurate, even when a number of different physical phenomena act simultaneously (Richmond , 2005). FE models are known to handle large numbers of variables in biological systems and have been successfully used to provide an understanding of how several marine mammal ear systems function (Aroyan, 2001; Cranford and Krysl, 2015; Maftoon , 2015; Tubelli , 2012; Tubelli , 2018; Wei , 2017; Wei , 2018) and, more recently, in lizards (Livens , 2019) and frogs (Fleming, 2021).

In teleost fishes, the first studies using FE analysis focused on the motions of otoliths, and how they respond to sounds from different directions and frequencies. The otolith was mimicked by a 3D structure embedded in viscoelastic matrix and exposed to pressure waves. Both translational and angular oscillation to low-frequency acoustic waves were observed, representing a more complex motion than simple back-and-forth oscillation (Krysl , 2008; Krysl , 2012; Schilt , 2011). Furthermore, the method was used to invent a fish-inspired underwater hearing device (Tse , 2014). More recently, 3D scans of fish inner ears were used to represent the otolith instead of a digitally constructed geometry, concluding a more accurate model of the acoustic response and directionality of fish otolith, for example, in the yellow croaker (Larimichthys crocea) (Zhang , 2021). Wei and McCauley (2022) used FE to simulate the otolith motion of the bight redfish (Centroberyx gerrardi) when exposed to signals arriving from the horizontal and vertical directions and showed different magnitudes of differential motion between the macula and otolith. The Weberian ossicles of the zebrafish (Danio rerio), a microstructure suggested to conduct sound pressure transmission from the swim bladder to the inner ear, were also reconstructed from μCT data and modelled to be stimulated by auditory signal vibrations (Marcé-Nogué and Liu, 2020). Salas (2019a,b) also used CT data of larval red drum (Sciaenops ocellatus) and a FE model to assess ontogenetic changes in sound pressure sensitivity.

FE models can lead to insights of function that are impossible to obtain from morphological studies or biomechanical experiments alone. Nowadays, FE models incorporate fine details of morphology, kinematics, and behaviour. They also have a powerful predictive utility, as some specific parameters can be easily changed, the model resolved, and the consequent functional difference mapped. Future models for fish hearing could include a more complete morphology (entire inner ear rather than only otolith/otoconia) to fully grasp the multidimensional functionality of hearing. Different signals could also be applied from different angles to understand frequency discrimination and source localisation. Furthermore, real life sounds, like anthropogenic noise sources, could be used as the driving signal to explore their potential effects.

While FE provides a simulation of a modelled function, it is not a strict representation of the reality. The geometry is divided into a finite number of elements, and while the model tries to minimize the error over the whole domain, exact solutions are only given at the nodes. Assumptions are also sometimes made to describe the materials and/or the loads, as not all physical properties can be measured. FE analysis thus yields an approximate solution, as does any modelling technique, while also being a computationally costly approach, depending on the complexity of the system. That said, the results of an FE model can be compared to experimental data obtained with more traditional in vivo experiments. Thus, the models can be corrected and updated as new data become available. For example, the results of dynamic imaging experiments with real tissues in which the motion can be visualized in 2D or 3D (Maiditsch , 2022; Schulz-Mirbach , 2018; Schulz-Mirbach , 2020), as depicted above, could ideally inform and correct FE models of the inner ears. It is also worth noting that the quality and accuracy of the 3D geometry sourced from the 3D scans (e.g., CT or MRI) is a key factor influencing outcomes of FE biomechanical simulations, especially when studying delicate and minute structures like fish inner ear microstructures (otoliths, end organs, maculae); scan resolution and contrast variations are known to affect the fidelity in 3D models and the simulation outcomes (Marcé-Nogué and Liu, 2020).

As mentioned above, geometric morphometrics is a correlational approach, and lacks the mechanistic explanation of the functional consequences of specific changes in morphology. FE analyses can provide this explanatory power, enriching the description of form and function. FE is now often used in combination with geometric morphometrics and both use 3D representations issued from bioimaging data. FE analysis returns detailed information about the biomechanics of individual biological structures, whereas geometric morphometrics allows shape differences across a sample to be quantified and analysed. They both can be combined within a framework of quantitative evolutionary theory to test hypotheses about the role of functional factors in the evolution of morphological form (Polly , 2016) (Fig. 4). For example, a study using 3D geometric morphometrics of teeth shape in snakes highlighted that the diversification of tooth morphologies were linked to the diet of each species, a hypothesis confirmed by an FE analysis, showing that some teeth were better suited and suffered less generated stresses for biting hard prey (Rajabizadeh , 2021).

Similar hypotheses could be tested concerning the hearing function of both bony and cartilaginous fishes in relation to the shape of their inner ear microstructures. Such questions could include:

• Do larger end organs or maculae (relative to the fish's body and brain size) convey a greater range of motion for the overlying otolith/otoconia?

• Does the density of the otolith/otoconia influence its differential movement?

• Does the motion of the otolith/otoconia change with different frequencies and does the shape and orientation of each end organ influence that motion, potentially reflecting a specialization of each end organ for some frequency ranges?

• Do enlarged end organs facilitate low-frequency sensitivity?

Two different approaches could be taken to answer these questions. In one case, the forces (or loads) of the FE models can be changed (e.g., different sound intensities, different frequencies) to test the stress and strains of one specific morphology. In the other case, the morphology can be changed (i.e., in a hypothetical form), while applying the same forces/load to the system, to understand the impact of morphological variance on mechanical function. For example, models that represent extremes of morphological variation within a taxon may be created (due to the results of the geometric morphometrics analyses), and then used in a simulation with the FE environment (O'Higgins , 2019). While hypothetical shapes may be used for these tests, some of these “extreme” specializations can actually be observed in some species, like the very large saccular otolith in gobies (Popper, 1981) or very large utricular otolith in some catfishes (Popper and Tavolga, 1981). These particular morphologies could be prioritized in bioimaging, geometric morphometrics, and FE studies.

As mentioned above, the fish hearing apparatus can be seen as separate “modules,” and some with separate functions that may or may not combine. Understanding how the inner ear microstructures and its modules integrate into a coherent, functionally apt structure is central towards advancing knowledge about auditory and vestibular functions in the inner ear. While these modules are likely evolving semi-independently from one another (Köppl, 2010), they stay interconnected to perform a particular function. Once the different modules of the inner ear and their array of diversification have been identified, their association with ecological shifts and changes in patterns and magnitude of the functions can be investigated.

Ancillary auditory structures, like the swim bladder or any other gas-filled organ and Weberian ossicles which physically link the swim bladder to the inner ear, can be seen as extra modules that are sometimes present, or not. These separated modules may couple for enhanced function and thus have potentially coevolved. For example, Schulz-Mirbach (2014) showed that, in the cichlid Etroplus maculatus, improved audition was associated not only with inner ear morphology modifications, but also with altered swim bladder shapes, which enhanced sound pressure detection. However, in such a coevolution scenario, trade-offs tied to an inability to simultaneously optimize alternative functions appear (Wainwright, 2007). Therefore, species with ancillary structures may present a more “conservative” inner ear diversification than other species, where multifunctionality discourages the incorporation of novel functions into existing repertoires (Corn , 2021; Farina , 2019). However, this pattern has not been observed in fishes. For example, Popper and Coombs (1982) found that there was a correlation between having more complex cell orientation patterns in the saccule end organ and the close proximity between the swim bladder and the inner ear. Understanding how different anatomical regions interconnect to achieve a particular performance outcome could provide a deeper understanding about the covariances in morphological structures.

This process is also complicated by the fact that some species have adaptations that are thought to enhance hearing (such as an extension of the swim bladder to the ear), yet they still have relatively poor hearing abilities (Popper , 2022). However, as discussed in and highlighted in Popper (2022), hearing performance and the definition of its “functions” are not yet fully understood. Hearing sensitivity (relative thresholds), hearing bandwidth, particle motion detection, sound pressure detection, signal discrimination, and sound localisation capabilities are all vital functions of hearing in all vertebrates. However, only the hearing sensitivity and often, pressure detection, is usually considered and measured. Vestibular functions seem a bit more conserved, as there is less variation in the canal labyrinth in fishes and across all vertebrates, as least in the species studied to date (Platt, 1988; Ramprashad , 1986). The superior part of the inner ear is generally considered as contributing to the vestibular system, composed of the three semicircular canals, which detect angular acceleration, and the utricule end organ, which detects linear acceleration (Pfaff , 2019b; Retzius, 1881). Still, the results in the description of “functionality: of the inner ear is rather uncomplete. Biomechanical reconstruction of each of these functions (both for vestibular and hearing performance) may help to decipher the complexity of modality in the inner ear and its ancillary structures.

To be able to untangle the complexity of different modules, different shapes, different functions, and different performance, focusing first on a simplified system may be a recommended strategy. We argue that, in the hope to reduce complexity in the system, a focus on a group of “non-specialist” species of fishes may be preferred, i.e., fishes without known specializations that could contributed to enhanced sensitivity (Popper , 2022). For example, the elasmobranchs (sharks, rays, and skates) may have several advantages over most of the teleost fishes. While the taxon still exhibits a well-developed and largely unexplained inner ear diversity (Evangelista , 2010), it is one of the most basally positioned vertebrate groups, possessing an inner ear that has undergone 400 × 10⁶ years of evolution. There are more than c. 1000 species of elasmobranchs, occupying a range of underwater habitats and with different trophic ecologies, all over the world. They rarely use sound to communicate (although see Fetterplace , 2022), do not possess any ancillary hearing structures (no swim bladder), and thus are thought to only detect the sound particle motion (Chapuis and Collin, 2022). We therefore argue that elasmobranch inner ears offer suitable basis to construct FE models and a good starting place to explore the diversification. While the functional data (e.g., audiograms) on this taxa is arguably limited compared to some groups of bony fishes (like the otophysines, for example), recent work on their hearing abilities and their inner ear morphology are providing valuable insights (Nieder , 2023; Sauer , 2022a,b).

The FE analysis can follow two different approaches: it can be deductive, where close relationships between form and function are assumed, or inductive, where the aim is to study this relationship (Panagiotopoulou, 2009; Rayfield, 2007). The deductive approach begins with simplified models of the structure of interest, to which representative boundary conditions (i.e., force/loads) are applied. Then, a series of analyses can be run, where unstressed elements are sequentially removed, resulting in a fully optimized model in which all elements are stressed in response to the forces and can be compared to the real structure. The inductive approach has the benefit of being able to test the assumption that form and function are tightly linked. Thus, it investigates form without a set of predetermined assumptions, and is usually preferred by biologists. Inductive studies attempt to represent accurate geometry, estimate loading conditions and realistic material properties, and can be validated by experimental data. With a validated FE model, the function of the structure can then be determined and dependable predictions can be made as to the influence of the structural form in the evolution. While both approaches have their strength, an inductive approach may be favored in priority to study the morphology-to-function link of the fish inner ear. It is best to first determine the combination of biomechanical processes involved in auditory function, before identifying the importance of each structural module in the function.

We plan to undertake a combination of the methodologies discussed above (Fig. 4), using preserved tissues from diverse species of elasmobranchs. We aim to apply both geometric morphometrics and FE analysis (inductive approach) on the datasets, and correct and improve our models with phylogenetic and functional data as available (e.g., Naylor , 2012; Nieder , 2023).

Although the discussed methodologies may bring a new light on some processes, some overarching challenges remain and new concerns may appear.

The eco-acoustical constraints hypothesis proposes that the acoustical conditions of a fish's environment, i.e., its “soundscape,” are the key factors in the evolution of its acoustic sensitivity. The soundscape encompasses all the sounds emitted across the landscape in acoustic space, a collection of biological (biophony), geophysical (geophony), and human-produced (anthrophony) sounds. It contains a wealth of information that animals can use to interpret and respond to environmental conditions, given their auditory capabilities. The soundscape is seen as a crucial element of a species' habitat. The abilities that an animal has to receive, interpret, and respond to information in the soundscape are contributory adaptations that may drive the natural selection process. Measuring and quantifying soundscapes is, however, a complex endeavor.

Recent developments in acoustic sensors are enabling long term, non-invasive collection of soundscape data underwater (Lamont , 2022). However, effectively retrieving information from a soundscape, and the integration of this information into ecological models as a dependent factor is a novel, yet challenging, enterprise. To “summarize” the information included in the soundscape and, thus, characterise underwater soundscapes, metrics based on sound levels exist (Erbe , 2016). Another approach is to use mathematical representations of the soundscape contents, known as acoustic indices. However, the use of these metrics and indices to extract meaningful ecological information is controverted, especially for aquatic soundscapes, for which they often fail to discriminate ecological gradients (Alcocer , 2022).

Artificial intelligence, and methods in deep learning, may bring new ways to digest complicated and information-rich soundscapes into key elements that can be used as ecological factors (Stowell, 2022). For example, pre-trained neural networks, purposed for general audio classification, and fed with recordings from different ecosystems, can generate arbitrary acoustic features. These abstracted representations of each soundscape act as “fingerprints,” usable for further multivariate analyses incorporating the soundscape as a factor. While this approach has been applied successfully for terrestrial studies (Sethi , 2020), there is a need to explore their use for aquatic soundscapes.

The eco-acoustical constraints hypothesis implies that auditory sensitivities of any fish evolved in parallel to the soundscape of its habitat, i.e., its ambient background noise (Ladich, 2013). Therefore, the fish hearing abilities could be measured in the presence of various background noise types: the natural background noise of each species should not affect the hearing capability of fishes while any other soundscapes may be masking the tested signals. For example, the hearing abilities of three species of bony fishes barely changed under quiet lab conditions and ambient noise, as measured through electrophysiological experiments (auditory evoked potentials), while there was a considerable increase in the auditory thresholds under boat noise conditions (Codarin , 2009). More studies testing the masking effects of different soundscapes on the physiological and behavioural hearing abilities of fish are required to further verify the hypothesis.

Studying functional morphology of the inner ear and comparative acoustic neuroecology across the fish clade will instigate global questions about the ecological and evolutionary mechanisms driving the development of hearing. However, to fully understand variations in morphologies, the interface between hearing and other individual sensory systems need to be mapped. Similar to the modulatory view of the inner ear structures discussed above, trade-offs and correlations between sensory systems may affect their respective morphology and functional performance. This is even more relevant when the sensory systems are closely related, in space (e.g., vestibular and auditory functions in the inner ear), or in functions (e.g., inner ear and lateral line may both detect near field particle motion). A question arises whether a species can evolve one sensory system over another in a certain environment.

The study of sensory “extreme” environments may be useful to highlight these sensory trade-offs. For example, do deep-sea fishes, living in perpetual darkness, show particularly well-developed auditory apparatus and functions, at the cost of a poor visual system? The reality seems more complex, as a lot of deep-sea fishes exhibit unexpected adaptations to the dark environment; for example, many deep-sea fishes have enlarged eyes relative to their body size and a high number of rod photoreceptors to maximize photon capture (Collin, 1997). Similarly, biofluorescence, a process well known in deep-sea organisms, has been observed in cartilaginous and bony fishes living in shallow water, like in the swell shark (Cephaloscyllium ventriosum) or the midshipman (Porichthys notatus) (Sparks , 2014). Still, Deng (2013) observed substantial and unique (so far) inner ear variations and apparent specializations in the deep-sea fish family Melamphaidea, and larger maculae in deeper living species of Macrouridae (ranging from 200 to 5000 m deep), compared to shallower species equipped with visual adaptations (Deng , 2017). More research is needed to identify trade-offs and correlations between different sensory modalities, and with the brain, which also highlights the need for collaboration of experts in different fields, with the hopes to consider the broader, multimodal evolution of fish sensory systems.

Aquatic soundscapes are changing, due to the continuous addition of anthrophony and the modifications of habitats by climate change (Duarte , 2021). It is now evident that fishes face increased anthrophony levels and reduced acoustic communication space (Popper and Hawkins, 2019). While adaptation involves the selection for genetic variation that increases organismal fitness, acclimatization (or habituation) relies on plastic responses in morphology, physiology, or behaviour to new environmental conditions. Behavioural (Holmes , 2017; e.g., Nedelec , 2016) and physiological (Johansson , 2015) acclimatization from short term noise exposure has been observed in fishes but, plastic morphological responses of the auditory apparatus to anthropogenic conditions are yet to be investigated. However, the plasticity of inner ear structures in fish has otherwise been demonstrated; for example, in the seasonally breeding plainfin midshipman (Porichthys notatus), hair cell density in the saccular maculae increase in reproductive females and correlates to a higher auditory sensitivity, likely to better detect and locate vocalising males (Coffin , 2012). Therefore, the relationship between morphology and auditory performance likely constrains inner ear shape development along environmental gradients and ecological shifts, and potentially along anthropogenic axes as well. A focus on functional morphology in fish sensory systems, in particular in the inner ear, could be used to identify general patterns of variation and make better predictions of the responses of fish communities to anthropogenic disturbances.

In conclusion, we identify two approaches to further understand how the diverse morphology of the fish inner ear is linked to its physiology and function. The first way is to contribute more experimental data about hearing in fishes (not only auditory thresholds and hearing bandwidth, but also frequency discrimination, directional hearing, etc.). Another novel, but equally pertinent, methodology is by modelling the auditory system, facilitated by state-of-the-art techniques coming from bioengineering and computer science (geometric morphometrics, FE analyses). Both approaches will complement each other and initiate progress. As such, sophisticated physiological measurements, in a comparative context, can not only improve our fundamental knowledge of the function of the inner ear, but also act as experimental test proofing data to inform, test, and/or correct digital models. In turn, the quantification and modelling of the discovered variations in a complex biological form like the inner ear will allow us to test hypotheses regarding what factors impact its morphological diversification.

This project has received funding from the European Union Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie agreement (Grant No. 897218) to L.C., and Marsden Grant No. UOA1808 from the Royal Society of New Zealand Te Apārangi to K.E.Y. and C.A.R.