Comparison of acoustic particle acceleration detection capabilities in three shark species

Behavioural studies of free-swimming sharks have proven that sharks are well capable of directional orientation (Nelson, 1967; Nelson et al., 1969; Nelson and Johnson, 1972; Wisby and Nelson, 1964). Yet, there have been few studies investigating the physiological mechanisms of directional hearing abilities in sharks (Casper et al., 2012; Chapuis et al., 2019). Sound, whether in air or underwater, is a mechanical disturbance that produces a nondirectional pressure component and a directional particle motion component (Au and Hastings, 2008). Sharks and all other members of the elasmobranch clade (e.g. skates and rays) lack a swimbladder or any other known pressure-to-displacement transducer and therefore are thought to detect only the particle motion component of sound (Banner, 1967; Kelly and Nelson, 1975) via the mechanosensory receptor endorgans of the inner ears (Dijkgraaf, 1960). Sharks are most sensitive to particle motions at low frequencies (<300 Hz) and cannot detect frequencies much higher than 1 kHz (Chapuis and Collin, 2022; Mickle and Higgs, 2022).

The inner ear of elasmobranchs is composed of three semi-circular canals with associated sensory regions (canal cristae) and three otoconial endorgans: the utricle, saccule and lagena (Popper et al., 2021). Each of the otoconial endorgans contains a sensory epithelium (macula) and an overlying dense calcareous otoconia (Retzius, 1881). The maculae are composed of sensory hair cells and supporting cells (Roberts et al., 1988). The otoconia function as differential accelerometers, in short, the animal and the surrounding water are the same density, therefore the inner ear vibrates in sync with the particle motion component of the sound field (Popper and Hawkins, 2018; Schuijf and Buwalda, 1980). The denser otoconia lag these movements, which causes shearing of the hair cell bundles and gives rise to depolarizing receptor potentials that travel through the eighth nerve to the brain (Fay and Popper, 1980). A second, non-otoconial sound reception pathway in elasmobranchs is mediated by the macula neglecta, which is situated in the posterior canal duct and composed of one sensory macula in rays, and two opposing maculae in sharks (Corwin, 1978). The hair cells of the macula neglecta are embedded into a gelatinous cupula and are thought to be stimulated by displacements of the cupula and surrounding perilymphatic fluid of the posterior canal duct (Casper and Mann, 2007a; Fay et al., 1974; Tester et al., 1972). Parietal structures of the head, such as the parietal fossa, a dorsal indentation of the skull filled with loose connective tissue, and the fenestra ovalis, a membranous opening in the otic capsule just opposite the macula neglecta, are suspected to preferentially transmit displacements from above the sharks' head to the macula neglecta (Casper and Mann, 2007a; Corwin, 1981b; Fay et al., 1974; Lowenstein and Roberts, 1950; Tester et al., 1972). In species of the family Carcharhinidae, the macula neglecta is enlarged and has a strict ventral–dorsal hair cell alignment, which is not the case in other shark groups (Corwin, 1977, 1978). This led to the hypothesis that the carcharhinid sharks should be particularly sensitive to sounds coming from above the head and may possess better directional hearing abilities than sharks with a smaller, less well aligned macula neglecta (Corwin, 1978).

The maculae of the inner ear end-organs are composed of different hair cell orientation groups (Barber and Emerson, 1980; Fay and Popper, 1980; Lovell et al., 2007; Popper and Coombs, 1982; Sauer et al., 2022). Each group contains hair cells of similar polarization that are sensitive to sound (vibration) from a specific direction (Fay, 2005; Flock and Wersall, 1963). It is known that the directional responses of the auditory nerve fibres follow the polarization of the hair cells they innervate (Fay, 1984; Fay and Feng, 1987; Lanford et al., 2000). Together, the directional physiology of the hair cells and the formation of orientation patterns within the inner ear endorgans provide fish with the basic tools for directional hearing and sound localization (Hawkins and Popper, 2018; Popper et al., 2003; Schuijf and Buwalda, 1975; Sisneros and Rogers, 2016).

While directional hearing abilities in teleost fish have been experimentally addressed (Buwalda and Van der Steen, 1979; Buwalda et al., 1983; Chapman and Johnstone, 1974; Fay and Edds-Walton, 1997, 2000; King, 1985; Lu et al., 1996; Zeddies et al., 2010; Zeddies et al., 2012) sharks and other elasmobranchs have been largely overlooked. Behavioural studies from the 1970s and 80s clearly demonstrate that sharks are capable of orienting towards pulsed low-frequency sounds from underwater speakers (Davies et al., 1963; Myrberg et al., 1969, 1976; Nelson and Gruber, 1963). One way to test directional hearing, is to use a shaker table system where the ears can be stimulated in specific axes (Enger et al., 1973; Fay, 1984). The shaker table mimics the particle acceleration component of sound and provides the opportunity for measuring pure particle motion sensitivity of the auditory system without the presence of a pressure component (Dinh and Radford, 2021; Mooney et al., 2010; Radford et al., 2012). To date, there has been only one study investigating the directional hearing ability of sharks using a shaker table, finding that bamboo sharks (Chiloscyllium) detect accelerations from all directions equally well (Casper and Mann, 2007b). It is not known, however, whether that holds true for all sharks or whether some species have a preferred axis (e.g. more sensitive in a certain direction), which may result in different directional hearing abilities.

The aim of this experiment was to test if sharks are more sensitive to vertical or horizontal (head-to-tail) accelerations, and whether there are any differences between different species. Using a shaker table in combination with the auditory evoked potential (AEP) technique, we determined normalized acceleration detection thresholds to both the horizontal (x-axis) and the vertical (z-axis) axes in the New Zealand carpet shark (Cephaloscyllium isabellum), the rig shark, also known as spotted estuary smoothhound (Mustelus lenticulatus), and the school shark (Galeorhinus galeus). Finally, we discuss how directional hearing abilities may be linked to ecological differences in the three species.

Seven rig (Mustelus lenticulatus Phillipps 1932) and 15 school sharks (Galeorhinus galeus Linnaeus 1758) (Table 1) were caught using circle-hook and line in the Kaipara Harbour (New Zealand, North Island). Five carpet sharks (Cephaloscyllium isabellum Bonnaterre 1788) were collected by local commercial fishermen. Animals were housed in flow-through holding tanks, supplied with ambient seawater (mean water temperature 15.5±0.78°C, salinity range 35–36 ppt) and maintained on a mixed diet of squid and fish three times a week. Sharks were acclimated for at least 1 week prior to experimentation and no food was given for 24 h prior to AEP measurements. All procedures were conducted in accordance with ethics protocols #002066/#AEC23071, approved by the University of Auckland Animal Ethics Committee (AEC).

To compare pure acceleration sensitivities between the three shark species a custom-made moving coil shaker system was used to accommodate sharks of up to TL 75 cm (Fig. 1). A large Plexiglas tank (dimensions 123×123×123 cm L×W×H) was positioned on a pneumatic vibration control table (TMC, Ametek, Pennsylvania, USA). To achieve dominant stimulation in the x-axis (head to tail) two Brüel & Kjær mini-shakers (LDS V201; Nærum, Denmark) were connected at opposite ends of the length axis of the tank and operated in a push-and-pull motion. This configuration is referred to as X-shaker herein. For dominant stimulation in the z-axis (vertical) one larger Brüel & Kjær shaker (LDS V406 M4-CE; Nærum, Denmark) was connected to the centre bottom of the tank and is referred to as Z-shaker herein.

Signals were generated on a computer (Dell, OptiPlex) running Spike2^® CED (Cambridge Electronic Design Ltd, Cambridge, UK) v10 software. Stimuli consisted of acceleration bursts at 40, 80, 100, 200, 400, 800 and 1200 Hz and were smoothed with a Hanning window (Table 2). The signals were digitized (CED Micro3 1401), attenuated (CED 3505) and amplified (LDS LPA100; Nærum, Denmark) before being sent to either the X- or Z-shaker.

Calibration of both X- and Z-shaker was done daily with the calibration script in Spike2. The maximum shaker level for each frequency was calibrated using three single-axis Brüel & Kjær accelerometers (DeltaTron^® 4507 B 002) with sensitivities for x=96.30 mV ms⁻², y=97.35 mV ms⁻² and z=96.20 mV ms⁻². The accelerometers were fixed with plasticine on top of the head holder just above the shark's head, such that x was aligned in X-shaker direction, y with the left–right axis and z with the Z-shaker direction. The accelerometers were connected to a Brüel & Kjær signal conditioning amplifier (NEXUS Type 2690-OS4, Denmark) with the output (V_rms) measured on an oscilloscope (Tektronix DPO 2014).

Because the X- and Z-shaker signals were composed of motions in all three dimensions, to varying degrees and depending on frequency (see Fig. 2), particle acceleration was determined for all three dimensions and the acceleration magnitude was calculated as √(x²+y²+z²). However, it is known that the X-shaker signals were dominated by the x-axis component at 40, 80 and 100 Hz, and the Z-shaker signals were dominated by the z-axis component at all test frequencies, except 1200 Hz. The calculated acceleration magnitudes were normalized to range between 0 and 1 by: z_i = x_i – [min(x)/max(x)] – min(x), where z_i is the ith normalized value in the dataset; x_i: the ith value in the dataset, min(x) is the minimum value in the dataset and max(x) is the maximum value in the dataset.

Prior to experiments, the shark was anesthetized by immersion (∼10–15 min) in a saltwater bath of MS222 (ethyl 3-aminobenzoate methanesulfonate) and sodium hydrogen bicarbonate (ratio 1:2). The shark was closely monitored while immersed and the anaesthesia plane was tested by tail-pinching. When there was no response to tail-pinching, the shark was wrapped in a wet cotton cloth from pectoral fins to tail (gills were exposed for ventilation to occur) and positioned on foam padding inside the centre of the Plexiglas tank. The animal was held in place by a custom-made head clamp. Sharks were maintained during the experiment on a lower concentration of anaesthesia delivered by a mouthpiece connected to a circulating system. No muscle relaxants were needed for these experiments. The eyes were covered by small pieces of wet cloth to prevent them from drying out and seawater was dripped regularly over the skin with a plastic pipette to keep the animal wet.

Following anaesthesia, stainless steel reusable subdermal needle electrodes (27 gauge, 12 mm, Rochester Electromedical Inc.; Coral Springs, Florida, USA) were inserted to record shaker evoked responses. The recording electrode was inserted dorsally, underneath the skin at the border of the parietal fossa (where the parietal fossa has its largest lateral–medial diameter). The reference electrode was inserted into the cartilage at the tip of the snout, and the ground electrode was placed under the shark's body. The outputs from the electrodes were pre-amplified (×10), amplified again (×100) with an extracellular differential amplifier (Dagan EX4-400 Quad; Dagan Cooperation, Minneapolis, MN, USA), visualized and recorded in Spike2 (https://ced.co.uk/products/spkovan) using a modified version of the fishABR-script.

The presentation order of the frequencies was conducted randomly, where each frequency was presented 200 times at 0 deg and 200 times at 180 deg and averaged to eliminate stimulus artefacts. AEPs were first elicited using a shaker level above threshold and the levels were then attenuated in 3 dB steps for each frequency until an AEP could no longer be identified. Then, one to three additional measurements, at 5–15 dB below this roughly estimated threshold were made to ensure responses were not missed. The presence of an AEP was visually determined in Spike2 through: (1) observation of the characteristic wave visible above the background noise, and (2) observed peaks at twice the stimulus frequency in the power spectrum. This method is commonly used in fish AEP studies (Kenyon et al., 1998; Popper et al., 2005b; Stanley et al., 2020; Vetter et al., 2018) and is based on the theory that the opposed orientation of the hair cells in the saccule of the inner ear gives rise to the characteristic frequency response at twice the stimulus frequency (Fay, 1974).

Shaker evoked potential thresholds were estimated using the x-intercept method, an objective (statistical) AEP threshold determination method that has previously been used in hearing studies with marine mammals and crustaceans (Dinh and Radford, 2021; Jézéquel et al., 2021; Mooney et al., 2009, 2010). Using a custom-written MATLAB script, the 2048 point fast Fourier transform (FFT) power spectra was calculated for 5–12 averaged waveforms per frequency. As with fish AEPs, the spectra revealed peaks at twice the stimulus frequency at suprathreshold SPL and decreases in FFT peak amplitude corresponded with decreasing SPLs (Maruska and Sisneros, 2016). The maximum FFT value (peak value) was found across five FFT bins greater than, and five FFT bins less than twice the presented frequency and was plotted against the normalized signal level as shown in panel D of Figs 3–5. Then, as series of regressions were run, that included 3, then 4 and 5, to all peak values. The regression line that yielded the highest r²-value (best fitting line) was selected to calculate the x-intercept. The x-intercept of the best fitting line served as estimate of the animal's probable hearing threshold (Nachtigall et al., 2007).

All statistical analyses were performed in R (v4.1.1) (https://www.r-project.org/). To test for any effects of total length, and sex on AEP thresholds, linear mixed effects analyses, using ‘lmer’ in the ‘lme4’ package (Bates et al., 2015; https://CRAN.R-project.org/package=lme4) were firstly conducted. Normalized thresholds were fitted against total length, and sex as the main effects and random intercepts for each shark subject. Data for each species were analysed in separate models and checked for linear and quadratic curve relationships, respectively. There was no association between threshold and total length (carpet shark, t₃₂=−0.35, P=0.73; rig shark, t₇=1.69, P=0.14; school shark, t₁₃=1.08, P=0.30), and threshold and sex (carpet shark, t₃₂=−0.16, P=0.87; rig shark, t₅=−1.81, P=0.13; school shark, t₁₂=−0.46, P=0.66) (Fig. S1). Therefore, all data could be grouped to increase sample size and power of the analysis.

Secondly, a linear mixed-effects analysis was performed to (1) compare thresholds between X- and Z-shaker thresholds (within each species and each frequency) and (2) compare thresholds between species (within each shaker and each frequency). Normalized acceleration thresholds were fitted against a factor, termed ‘group’ as the main effect and random intercepts for each shark subject. The group factor represents all possible combinations of the variables species, frequency, and shaker, e.g. Carpet.Shark_40Hz_X-shaker, Carpet.Shark_40Hz_Z-shaker, Rig.Shark_40Hz_X-shaker, etc.). This factor was needed for the models to run properly, as there was not enough overlap in the distribution of frequencies, due to the carpet shark not showing any responses at 400, and 800 Hz. Post hoc pairwise comparisons were used to explore differences between threshold means between shaker (within species and frequency) and between species (within shaker and frequency) using the ‘emmeans’ (v1.6) package.). We only examined comparisons between shaker, within each species and each frequency (e.g. Carpet.Shark_80Hz_X-skaker vs. Carpet.Shark_80Hz_Z-shaker) and between species, within each shaker and each frequency (e.g. Carpet.Shark_80Hz_X-skaker vs. Rig.Shark_80Hz_X-shaker). All ‘nonsensical’ comparisons (e.g. Carpet.Shark_80Hz_X-shaker vs. Rig.Shark_100Hz_Z-shaker) were ignored. P-values were adjusted using the FDR method and all statistical tests were considered significant at P<0.05.

At the maximum shaker level (start level) the acceleration signals of the X-shaker were dominant in the x-axis at 40, 80 and 100 Hz (Fig. 2A), approximating the ideal scenario (see upper right inlet in Fig. 2A). At 200, 400, 800, and 1200 Hz the test conditions were not ideal, as signals contained strong resonances in the y- and z-axes. The acceleration magnitude of the Z-shaker was highest in the z-axis for all test frequencies and approximated the ideal scenario (see upper right inlet in Fig. 2B), except for 1200 Hz (Fig. 2B).

The auditory evoked potential (AEP) waveforms of the three shark species were similar in shape and time course and showed a sharp peak at twice the stimulus frequency in the FFT analysis at suprathreshold levels (Figs 3–5). A typical suprathreshold AEP response consisted of a series of downward and upward peaks superimposed over a slow negative deflection, that was generally followed by a slow positive deflection, as it is typically described for other fishes and lower vertebrates (Bullock and Corwin, 1979; Corwin et al., 1982; Kenyon et al., 1998).

The normalized AEP detection threshold curves for the X- and Z-shakers are similar in shape and slope within each species, with sensitivity maxima and minima around the same frequencies. In the carpet shark there were no differences between X- and Z-shaker thresholds at 40, 80, 100 and 200 Hz (Fig. 6A; Table S1). In the rig shark, the Z-shaker thresholds were significantly lower than X-shaker thresholds at 40, 80, 100 and 200 Hz. But there were no significant differences between X- and Z-shaker thresholds at 400 and 800 Hz (Fig. 6B). Similarly, in the school shark, Z-shaker thresholds were significantly lower than X-shaker thresholds at 40, 80, 100 and 200 Hz, but were similar at 400 and 800 Hz (Fig. 6C).

There were no differences in normalized X-shaker thresholds between the three species, except at 40 Hz where the carpet shark was more sensitive than the school shark (Fig. 7A; Table S2). For normalized Z-shaker thresholds there were no differences between the three species, except for the rig shark who was most sensitive between 100 Hz and 200 Hz, showing lower mean thresholds than the school shark at 100 Hz and lower mean thresholds than the carpet shark at 200 Hz (Fig. 7B; Table S3). At 200 Hz, the school shark also had a lower mean threshold than the carpet shark. The carpet shark had the narrowest hearing bandwidth (40–200 Hz), and was most sensitive to frequencies below 200 Hz, with a lowest mean threshold found at 40 Hz. All five carpet sharks still responded at 200 Hz and stopped responding at 400 Hz (Fig. 8). Four of 15 school sharks and three of seven rig sharks showed responses at 800 Hz and none of the sharks responded at 1200 Hz. The rig shark was most sensitive to frequencies below 400 Hz, with a lowest mean threshold at 100 Hz. Finally, the school shark was most sensitive to frequencies below 200 Hz, with lowest mean thresholds found at 100 Hz for the X-shaker and 80 Hz for the Z-shaker.

To our knowledge, this is the second study to investigate the gross directional response characteristics (x- and z-axis) of the auditory system in sharks using a shaker table. The three shark species tested in this study showed AEP responses to both horizontal and vertical accelerations, which is in general support of previous experiments using AEPs by Casper and Mann (2007b), suggesting that sharks possess omnidirectional ears that can receive sounds from all directions (Casper and Mann, 2007b).

Overall, acceleration detection bandwidth was narrowest for the carpet shark (40–200 Hz), and broader (40–800 Hz) for the rig and school shark. There was also variation between bands of greatest sensitivity, which were 40–80 Hz for the carpet shark, 100–200 Hz for the rig shark and 80–100 Hz for the school shark. The shaker-derived AEP threshold curves of the three species obtained in this study (e.g. shape, best sensitivity and upper frequency limit) resemble those previously obtained in response to an underwater speaker (Nieder et al., 2023). Unfortunately, it is not possible to directly compare acceleration thresholds from both experiments, because the shaker derived thresholds from the shaker experiment were normalized. It is important to note that AEP thresholds depend on electrode placement with respect to the neural tissue, and differences in head size and geometry of the skull may affect AEP detection thresholds (Lauridsen et al., 2021; Maruska and Sisneros, 2016). The carpet sharks used in this study were on average 18–20 cm larger than the school and rig sharks, and caution needs to be exercised in comparing the three species. Although similar sensitivity relationships between the three species were found in a recent study using speaker-derived AEP thresholds, where there was no significant size difference between the three species (Nieder et al., 2023).

We found that the carpet shark was equally sensitive to accelerations in both the vertical (z-axis) and the head-to-tail (x-axis) axes. In contrast, the school and rig sharks appeared to be significantly more sensitive to vertical accelerations than to head-to-tail acceleration at 40, 80, 100 and 200 Hz. These results suggest that there may be differences in directional hearing abilities between species. A caveat of this study is that the X-shaker only produced clean stimuli (dominant in the x-axis) at 40, 80 and 100 Hz; at the higher test frequencies, the X-shaker signals were conflated with motions in the y- and z-axis that originated from tank resonances. Therefore, comparisons between X-and Z-shaker directions need to be interpreted with caution with respect to directional sensitivity. This may explain why no differences in sensitivity between X- and Z-shaker were detected at 400 and 800 Hz for the rig and the school shark.

Comparative morphological studies of the inner ear across various species of elasmobranchs support the notion that habitat, foraging strategy and diet likely contribute to hearing differences between species (Corwin, 1978, 1989; Evangelista et al., 2010; Sauer et al., 2023). It has been inferred that differences in inner ear morphologies among elasmobranchs likely translate to differences in hearing capabilities, including directional hearing abilities (Corwin, 1977, 1978). However, the ecological factors that influence directional hearing abilities in sharks are currently not understood and may depend on species specific requirements (Chapuis and Collin, 2022). Feeding strategy may shape directional hearing abilities in some sharks. For instance, the carpet shark is a benthic ambush predator and an ability to detect sounds produced by vocalizations or swimming movements of their prey (Kalmijn, 1988; Richard, 1968) likely increases vigilance and feeding success and may explain why the carpet shark is equally sensitive to horizontal and vertical particle accelerations. Predation risk could potentially drive differences in directional hearing abilities in some active swimming medium-sized (TL_max ∼150–300 cm) mesopredatory sharks (Mourier et al., 2013; Roff et al., 2016), such as the rig and school shark. Both species are predated on by large predators, including sharks, teleosts and orcas (Francis, 2013; Lucifora et al., 2006; Shea et al., 2020). Orcas are vocal animals and likely attack their prey from above, therefore an increased sensitivity to sound coming from above or below would aid in detecting vocal predators. Further physiological, morphological and behavioural studies are needed to test these hypotheses.

In teleost fish, extracellular recordings from primary auditory afferents to directional stimulation via shaker table have shown that the axis of particle motion is encoded by hair cell orientation pattern of the inner ear sensory maculae (Fay, 1984; Fay and Edds-Walton, 2000; Lu et al., 1998; Sand, 1974). For instance, vertically oriented hair cells in the saccule were shown to selectively respond to vertical accelerations in some species of teleosts (Fay and Edds-Walton, 1997; Fay and Edds-Walton, 2000; Lu et al., 1998; Popper et al., 2003). Currently, information on hair cell orientation patterns in elasmobranchs only exists for very few species. In those investigated, the saccular epithelium is elongated and shows a vertical orientation pattern, composed of two groups of vertically oriented ciliary bundles (Barber and Emerson, 1980; Corwin, 1981b; Lovell et al., 2007; Lowenstein et al., 1964; Sauer et al., 2022). The sensory epithelia are not flat, but are curved and positioned in varying planes, so that enough hair cells can be effectively stimulated from all directions (Corwin, 1981a). There are currently no physiological data on the directional response characteristics of the individual otoconial endorgans from the inner ear in elasmobranchs. Several inferences have been made based on hair cell polarization maps and spatial orientation of the maculae in a few species. For instance, Lovell et al. (2007) found a large number of horizontally aligned utricular hair cells in the small-spotted catshark (Scyliorhinus canicula) and suggested that this benthic shark will be acutely aware of water displacements caused by sounds propagating in the horizontal plane (Lovell et al., 2007). In contrast, the saccule has fewer cells oriented in the vertical plane, because of its pronounced V-shape, which effectively decreases the maculae area with vertically oriented hair cells (Lovell et al., 2007). In the thornback ray (Raja clavata) (Lowenstein et al., 1964) and the winter skate (Raja ocellata) a seemingly large proportion of saccular hair cells show a vertical orientation with respect to the midline of the animal. It is therefore suggested that these batoid species may be particularly sensitive to particle motion in the vertical plane (Barber and Emerson, 1980; Casper and Mann, 2007b; Lowenstein et al., 1964). As shown in a recent morphological study by Sauer et al. (2022), the saccule of the school shark (Galeorhinus galeus) has a relatively wide medial part with ventral–dorsal polarized hair cells aligned in the vertical plane. As suggested for the two batoid species, this arrangement could indicate sensitivity in the vertical plane of the school shark too. Based on the physiological data presented here that suggest enhanced vertical motion sensitivity in both the rig and the school shark, one would expect to find an overall larger proportion of vertically oriented hair cells in these two species than in the carpet shark. Further quantitative morphological investigation of the structure and hair cell orientation patterns of the inner ear maculae in the three species are needed to better interpret the results of the present study.

It has been suggested that the macula neglecta will not be stimulated by the shaker table because it is not mass loaded and would not be sensitive to linear accelerations (Meyer et al., 2012). The density differences between the tissues of the macula neglecta and the cartilaginous otic capsule are thought to be too small to create an inertial lag that would bend the hair cells and effectively stimulate the macula neglecta (Kalmijn, 1988). However, Corwin (1981a,b, 1998) felt that it would be an oversimplification to assume that all tissues of the head, except the otoconia, are the same density as water, hence the same sound conducting properties. The tissue characteristics, known to affect sound transmission (e.g. density, elasticity, viscosity) should be assessed, to better understand how different tissues in the shark head are affected by sound (Corwin, 1989). Casper and Mann (2007a) obtained similar results when moving a dipole stimulus, which was positioned a few centimetres above the head (but not touching the skin) of the white-spotted bamboo shark (Chiloscyllium plagiosum) and the horn shark (Heterodontus francisci). AEP responses were strongest when the dipole was positioned directly over and slightly posterior to parietal fossa. The authors speculated that the localised water movements produced by the dipole above the parietal fossa resulted in the stimulation of the macula neglecta (Casper and Mann, 2007a,b). However, this theory remains to be tested, as it is not possible to distinguish between contributions of individual endorgans using whole-brain auditory evoked potentials. Thus, the adequate stimulus (e.g. displacement, velocity, acceleration, pressure) and exact mechanism that would cause the relative movement between the gelatinous cupula and the hair cells of the macula neglecta remain to be resolved (Rogers and Cox, 1988; Rogers and Zeddies, 2008). To identify the adequate stimulus for the macula neglecta, future studies should characterize electrophysiological responses directly from the ramus neglectus using various modes of stimulation (e.g. shaker table, underwater monopole and dipole), along different axes or from different locations around the fish. In addition, experiments measuring physiological and behavioural responses to acoustic stimuli before and after selective ablation of sensory hair cells in the macula neglecta will be crucial to uncover the function of the macula neglecta in elasmobranch hearing.

Further studies with an improved shaker table system that generates clean directional acceleration stimuli across a broader range of frequencies are required to further investigate directional sensitivities in these species. Additionally, quantitative analyses of the size, total number of hair cells and orientation patterns of the four sensory maculae in these species are needed to better interpret the physiological data provided here. It is important that in addition to the hair cell orientation patterns, the orientation of the epithelia in space and with respect to the animal's midline are known, for example, through 3D modelling, to predict potential directional response characteristics of the inner ear endorgans (Hawkins and Popper, 2018). Finally, examination of the size, mass and density of the saccular, utricular and lagenar otoconia may provide additional insight into how otoconia contribute to auditory and directional abilities in elasmobranchs fishes (Boyle and Herrel, 2018; Popper et al., 2005a; Schulz-Mirbach et al., 2019).

In summary, the results of this study add further support to the previous evidence that sharks can detect sounds from all directions. The data suggest that the bottom-dwelling carpet shark is equally sensitive to accelerations in the head-to-tail and vertical axes. In contrast, both benthopelagic rig and school sharks appeared to be more sensitive to vertical accelerations at frequencies up to 200 Hz. An enhanced vertical acceleration sensitivity may present an adaptative advantage for sharks that spend most of their time swimming and foraging in the water column. Directional hearing abilities in sharks may be influenced by multiple ecological factors, including feeding strategy and predation risk. Further physiological, morphological and behavioural studies are needed to test these hypotheses.

The authors would like to thank Errol Murray, Gavin Perry, Derek Sauer, Stefan Spreitzenbarth and Stefano Schoene from the Leigh Marine Lab for their help with the animal collections. Many thanks to Maria Mugica and Peter Brown for lab space assistance and fixing the shaker table setup. We would like to express our gratitude to Jason Dinh for his help with the MATLAB script for the x-intercept threshold determination method.

This article has an associated ECR Spotlight interview with Carolin Nieder.