Auditory evoked potentials were recorded in all test groups between 50 Hz and 6 kHz. Best hearing abilities were shown at frequencies between 0.3 and 1 kHz in all size groups, except the smallest group XXS. The largest group XL showed its lowest hearing threshold at 0.3 kHz (72.3 dB re 1 μPa), group XXS at 2 and 3 kHz (80.9 dB) (Figure 1, Table 1). Group XXS and XS showed poorest hearing abilities of all groups at 0.05 kHz. Group XXS showed lowest hearing abilities of all groups from 0.1 to 2 kHz, but showed best hearing abilities of all groups at the highest frequency tested (6 kHz). Group S had well-developed hearing abilities at low frequencies similar to larger size groups, but hearing sensitivity similar to that of group XXS and XS at the highest frequencies tested (5 and 6 kHz) (Figure 1, Table 1). Significant correlations between size and hearing thresholds existed at most frequencies. At lower frequencies (0.05 to 2 kHz), larger animals showed significantly better hearing, whereas at the highest frequencies tested (5 and 6 kHz) the opposite was the case: smaller animals had lower hearing thresholds. At 3 and 4 kHz no correlation was evident (Figure 2).

Stridulatory sounds were emitted during abduction (forward movement) and adduction (backward movement) of pectoral spines in all groups tested as soon as specimens were handled (Figure 3). Swimbladder drumming sounds could not be detected or recorded. Sound pressure level (SPL) increased with size up to 58 mm standard length (SL) whereas no further increase was observed in larger-sized animals. SPLs of animals up to 58 mm SL were significantly lower than in larger specimens (Mann-Whitney-U test: U = 62.5, N = 40, P < 0.01). In larger fish, no further increase in SPL was evident (Figure 4A). Duration of adduction and abduction sound as well as pulse period (PP) increased with size (Figures 3 and 4B-D). The dominant frequency of sounds decreased significantly with size in fish larger than 37 mm SL, while no correlation was found in smaller ones (Figure 4E). However, the dominant frequency of fish up to 37 mm SL was significantly higher than in larger fish (Mann-Whitney-U test: U = 5, N = 40, P < 0.01). While bandwidth decreased with increasing size in fish up to 73 mm SL, no further decrease of bandwidth was observed in larger fish. The main energies of stridulation sounds produced by smaller specimens were more broad-banded than those of larger ones (Mann-Whitney-U test: U = 44, N = 40, P < 0.01; Figures 3 and 4F).

All size groups showed best hearing abilities in frequency ranges where main energies of stridulation sounds were concentrated, and all size groups were able to detect conspecific sounds (Figure 5). Thus, group XL detected sounds produced by group XXS and vice versa.

Changes in hearing abilities have been reported in several fish taxa 15161819. The present study provides the first evidence that auditory thresholds change during ontogeny in an otophysine fish species (Figure 2), results that are in contrast with previous studies on two species of cypriniforms, namely the goldfish 11 and the zebrafish 14. Several potential explanations can be forwarded for this discrepancy among otophysines. First, different species have been investigated, which even belong to different orders, namely Cypriniformes and Siluriformes. Furthermore, the authors of those studies did not calculate regressions of hearing abilities including all data from each individual study specimen, but instead compared the mean results of different size groups. Furthermore, the size differences of the goldfish used as well as the range of frequencies tested might have been too small: the specimens were 45 to 48 mm SL and 110 to 120 mm SL, and the five test frequencies ranged from 100 to 2,000 Hz. However, goldfish are able to detect sounds at least up to 4 kHz 28. In contrast to the goldfish study, Higgs et al. measured early stages of zebrafish (10 mm) 14 and found an extension of the maximum frequency detectable from a 200 Hz upper limit in small specimens up to 4 kHz in large ones but no change in absolute thresholds. They argued that the development of the Weberian ossicles is responsible for this increase in the detectable frequency range. By contrast, Zeddies and Fay 12 found in their study on the development of startle response in zebrafish no change of stimulus thresholds and frequency bandwidth to which the zebrafish responded from five days post fertilization to adult. Similar to the observations of Zeddies and Fay 12 in zebrafish, we did not observe a change in the range of detectable frequencies. Based on Coburn and Grubach's 29 study on the ontogeny of the Weberian ossicles in several species of catfish we assume that all our tested animals possessed fully developed Weberian ossicles. Thus, the current study is the first to systematically demonstrate that auditory thresholds change with size in an otophysine fish (Figure 2), whereas the detectable frequency range does not change (Figures 1 and 2). Auditory sensitivity increased at lower frequencies up to 2 kHz and decreased at 5 and 6 kHz.

Some prior data on other catfish species are in agreement with these findings. In eight specimens of the loricariid catfish Ancistrus ranunculus, the four smallest specimens tested detected sound stimuli at 5 kHz, whereas only one of the larger ones responded to 5 kHz stimuli at the maximum SPL tested (129 dB re 1 μPa) 4. This indicates that smaller individuals of this loricariid had higher auditory sensitivities at the highest frequency tested. A similar observation can be made when comparing results of three studies on the pimelodid catfish Pimelodus pictus 303132. The smallest specimens tested by Wysocki et al. 32 had better hearing abilities at the highest frequencies tested (mean hearing threshold at 4 kHz: 75.3 dB re 1 μPa) than the largest fish tested by Ladich 30 (81.3 dB).

These data indicate that smaller individuals within a given species of catfish may hear better at higher frequencies. One possible reason could be that smaller specimens produce sounds of higher frequencies and are adapted to detect sounds of similar-sized conspecifics. Ladich and Yan 33 argued that the pygmy gourami, the smallest species investigated in their comparative study on labyrinth fishes (family Osphronemidae), heard better at 3 to 5 kHz than the larger two species. The authors argued that higher sensitivity at higher frequencies in the smallest species may reflect the higher resonance frequencies of their smaller-sized accessory hearing structures, the so-called suprabranchial chambers (paired air-filled chambers close to the inner ears).

Studies on the ontogenetic development of hearing in Perciformes (Pomacentridae - damselfishes, Osphronemidae - gouramis, Sparidae - sea breams) and Batrachoidiformes (Batrachoididae - toadfishes) 1516171819 show a more consistent pattern. They reported mostly a moderate improvement of hearing with size, with one exception. Kenyon 15 observed a considerable 60 dB increase in hearing sensitivity in the bicolor damselfish Stegastes (syn. Pomacentrus) partitus at the most sensitive frequency (500 Hz) during ontogeny. Contrary to Kenyon's results, Egner and Mann 20 reported a decrease of hearing sensitivity in a representative of the same family. The sensitivity decreased in the sergeant major damselfish Abudefduf saxatilis with growth at 100 and 200 Hz. Iwashita et al. 16 observed an increase in hearing sensitivity in zero-, one- and two-year-old red sea bream Pagrus major, in particular between 100 and 300 Hz. In the croaking gourami Trichopsis vittata, different trends were found at different frequencies during development of hearing 17. Hearing improved with size in the frequency range from 800 to 2,000 Hz, where the main energies of sounds were concentrated. An opposite trend was observed at 4 and 5 kHz. This decrease in sensitivity at highest frequencies in T. vittata is in agreement with the catfish data of the current study. In both toadfish species tested so far, the plainfin midshipman 18 and the Lusitanian toadfish 19, a small increase of hearing sensitivity with size was evident. In the Lusitanian toadfish, hearing improved with growth at three out of seven frequencies tested.

Whereas different trends were found in the development of auditory sensitivities in different species, the development of sound characteristics shows more similar patterns among teleosts. These patterns agree with the current data in Synodontis schoutedeni.

The dominant frequency of sounds is negatively correlated with body size in representatives of all families investigated so far, for example, in pomacentrids - Stegastes partitus, Dascyllus albisella, Amphiprion akallopisos, A. clarkii, A. frenatus, A. ocellaris, 23263435, osphronemids - Trichopsis vittata, T. pumila and T. schalleri 172122, sciaenids - Cynoscion regalis 36, triglids - Eutrigla gurnardus 25, and toadfishes - Halobatrachus didactylus 19. Fine et al. 37 found a decrease of center frequency in the ictalurid catfish Ictalurus punctatus. In contrast, Ladich 38 reported such a correlation in only one out of six catfish species from three families. Only in the doradid Platydoras armatulus (formerly named P. costatus) did the dominant frequency decrease with size, while no such correlation was present in pimelodids and mochokids. The lack of a relationship might have been due to the difficulty of determining dominant frequencies in broad-band sounds and perhaps the small size differences of fish used in that study. In the present investigation, S. schoutedeni size differed more (1.02 g to 51.80 g vs. 1.7 to 5.6 g in 38). A significant correlation between dominant frequency of sounds and size was found in animals larger than 37 mm SL. The decrease in dominant frequency may be an effect of longer pulse periods within sounds, and thus based on size-related changes in morphology 36. However, these underlying structural changes seem not to apply to animals smaller than 37 mm in which dominant frequencies were significantly higher than expected by the linear regression. Smaller individuals of S. schoutedeni showed a more evenly distributed sound energy than large individuals. The frequency band 10 dB below the dominant frequency ranged from ca. 2,800 to 4,100 Hz in smallest fish down to ca. 370 to 1,500 Hz in largest ones (Figures 3 and 4F). The decrease in bandwidth was significant from the smallest size group up to 73 mm SL; no further change was observed in larger animals. No such relationship has been described in other species so far, although a closer examination of sounds of other species might yield similar results. The power spectra of sounds of three species of clownfishes Amphiprion spp. 26 indicate that such a relationship may also exist in other fish species.

The sound pressure level (SPL) of vocalizations increases with size in species from different taxa such as in the tigerfish Therapon jarbua 39, the osphronemid T. vittata 17, the sciaenid C. regalis 36, the toadfish H. didactylus 19, and in the mochokid catfish S. schoutedeni (current study). Interestingly, while SPLs increased rapidly in animals below 58 mm SL, no further increase of SPLs occurred in the larger specimens (Figure 4A). The amplitude of sounds depends on anatomical constraints and on how long and hard fish press the dorsal process of the pectoral spine against the groove of the shoulder girdle 3740. This could cause intraindividual variation of SPLs and a lack of correlation in larger animals.

Temporal characteristics of sounds such as duration of abduction and adduction sounds and pulse period of sounds increase with size in S. schoutedeni. This can be explained by the growth of the dorsal process of the pectoral spine and the fact that a full pectoral sweep takes longer in larger fish than in a smaller one 3740. Increases with size have also been observed for pulse period in the Lusitanian toadfish 19, pulse duration in weakfish 36, and grunt duration in the grey gurnard 25.

In all tested size groups of S. schoutedeni, the main energies of stridulation sounds correspond to the most sensitive frequency range of hearing (Figure 5). Despite some uncertainty of sound level measurements in aquaria as compared to free field conditions our data show that all groups are apparently able to detect sounds produced by specimens of the other groups, which is in contrast to prior findings in other teleost species. In croaking gouramis and Lusitanian toadfish, individuals of the smallest size groups were probably unable to detect sounds produced by similar-sized conspecifics 1719 based on a comparison between absolute sound spectra and audiograms at a communication distance of 3 to 10 cm. The reason for this difference between S. schoutedeni and the latter two perciform species is two-fold: First, S. schoutedeni shows much higher hearing sensitivities than the other two species investigated, namely the croaking gourami and the Lusitanian toadfish 1719. Early stages of the mochokid catfish are more sensitive at the dominant frequencies of conspecific sounds than in the other two species. Secondly, the SPLs of juveniles' sounds are much higher than those of juveniles of the other two species. Thus, all squeaker catfish can detect sounds of conspecifics either uttered during agonistic intraspecific interactions or in a distress context 41.

For hearing measurements, fish were grouped into six size groups, XXS (SL = 21.9 to 36.5 mm, N = 12), XS (SL = 39.5 to 45.3 mm, N = 6), S (SL = 51.0 to 58.1 mm, N = 6), M (SL = 71.1 to 81.0 mm, N = 5), L (SL = 92.6 to 102.6 mm, N = 4) and XL (SL = 116.0 to 126.8 mm, N = 6). Hearing thresholds of group M were taken from Lechner and Ladich 4. For sound recordings, corresponding groups were defined. Because not every specimen vocalized during the sound recording procedures, minimum and maximum size ranges differed slightly: XXS: SL = 28.0 to 36.0 mm, N = 9; XS: SL = 39.0 to 45.0 mm, N = 6; S: SL = 46.7 to 58.1 mm, N = 7; M: SL = 61.8 to 81.1 mm, N = 10; L: SL = 92.6 to 102.2 mm, N = 3; XL: SL = 117.8 to 126.8 mm, N = 5.

Fishes of groups M (hearing tests) to XL were wild caught from Malebo Pool (Congo River, Democratic Republic of the Congo) and obtained from a tropical fish supplier (Transfish, Munich, Germany). After a quarantine of three weeks, fish were acclimated to our aquaria for at least two months prior to the first experiments. Specimens of groups XXS, XS, S and the specimens of group M used for sound recordings were aquarium reared. For a detailed description of the breeding procedure (breeding occurred spontaneously and was not induced by hormone injections) of the tested specimens see Drescher 42. Because breeding and rearing mochokid catfishes in aquaria is very difficult, we had only a small number of offspring and we were also interested in measuring fish of later stages, we did not test any specimens smaller than 22 mm SL.

Fish were kept in planted aquaria with a sand bottom equipped with roots and clay or bamboo tubes as shelters. Only external filters were used. In order to provide a quiet environment, we did not use any internal filters or air stones. Temperature was kept at 25 ± 1°C and a 12 h: 12 h L: D cycle was maintained. Fish were fed frozen chironomid larvae and artificial food (granulate, flakes and tablets); the small specimens of groups XXS, XS and S were also fed Cyclop-Eeze^® (freeze-dried copepods, Argent Chemical Laboratories, Redmond, WA, USA). Since fry and juveniles grow very unequally despite identical conditions of husbandry 1343, we classified the tested specimens as different size groups rather than age groups. Standard length (SL) was measured as standard length 2 following Holcik et al. 44. Using total length or body mass instead of SL for analyses did not change the results.

All experiments were performed with the permission of the Austrian Federal Ministry for Education, Science and Culture (GZ 66.006/2-BrGT/2006).

Hearing thresholds were obtained using the AEP recording technique developed by Kenyon et al. 28 and modified by Wysocki and Ladich 4546. Only a brief description of the technique is given here. Test subjects were mildly immobilized with Flaxedil (gallamine triethiodide; Sigma-Aldrich, Vienna, Austria) diluted in a Ringer solution. The dosage applied (3.5 to 5.4 μg g^-1) allowed fish to still perform slight opercular movements but not to initiate significant myogenic noise that could interfere with the AEP recordings. All auditory measurements were carried out in a bowl-shaped plastic tub (diameter 33 cm, water depth 13 cm, 1 cm layer of gravel), which was lined inside with acoustically absorbent material (air-filled packing wrap) to decrease resonances and reflections 47. The tub was positioned on an air table (TMC Micro-g 63-540, Technical Manufacturing Corporation, Peabody, MA, USA), which rested on a vibration-isolated plate of concrete. A sound-proof chamber, constructed as a Faraday cage (interior dimensions: 3.2 m × 3.2 m × 2.4 m), enclosed the whole setup. The subjects were placed at the water surface in the center of the tub. The contacting points of the electrodes were maximally 1 to 2 mm above the water surface. Tissue paper (Kimwipes^®, Kimberly-Clark Corporation, Irving, TX, USA) was placed on the fish head to keep it moist and ensure proper contact of electrodes. Respiration was achieved through a temperature-controlled (25 ± 1°C), gravity-fed water circulation system using a pipette inserted into the subject's mouth. The AEPs were recorded by using silver wire electrodes (0.38 mm diameter) pressed firmly against the skin: the recording electrode was placed over the region of the medulla and the reference electrode cranially between the nares. Shielded electrode leads were attached to the differential input of an AC preamplifier (Grass P-55, Grass Technologies, West Warwick, RI, USA, gain 100 x, high-pass at 30 Hz, low-pass at 1 kHz), with a ground electrode placed in the water near the fish's body. A hydrophone (Brüel and Kjaer 8101, Naerum, Denmark; frequency range 1 Hz to 80 kHz ± 2 dB; voltage sensitivity -184 dB re 1 V μPa^-1) was placed close to the head on the right side of the animals (approximately 1 cm away) in order to determine absolute stimulus SPLs. A custom-built preamplifier was used to boost the hydrophone signal. Both presentation of sound stimuli and AEP waveform recording were achieved using a modular rack-mount system (Tucker-Davis Technologies (TDT) System 3, Gainesville, FL, USA) controlled by a PC containing a TDT digital signal processing board and running TDT BioSig RP software (Tucker-Davis Technologies, Gainesville, FL, USA).

Hearing thresholds were determined at 0.05, 0.1, 0.3, 0.5, 0.8, 1, 2, 3, 4, 5 and 6 kHz. The duration of sound stimuli increased from two cycles at 50 Hz and 100 Hz up to eight cycles at 4 kHz and above. Rise and fall times increased from one cycle at 50 to 300 Hz up to three cycles at frequencies from 2 to 6 kHz. All bursts were gated using a Blackman window. Data for group M were taken from Lechner and Ladich 4, and this group was not tested at 6 kHz. For each test condition, one thousand stimuli were presented at opposite polarities, that is, 90° and 270°, and were averaged together by the BioSig RP Software, yielding a 2000-stimulus trace to eliminate any stimulus artefact. The SPL was reduced in 4-dB steps. Close to hearing threshold, this procedure was performed twice and the AEP traces were overlaid to visually check if they were repeatable. The lowest SPL at which a repeatable AEP trace could be obtained, as determined by overlaying replicate traces, was defined as the threshold [see also 48]. Sound stimuli waveforms were created using TDT SigGen RP software (Tucker-Davis Technologies, Gainesville, FL, USA). Tone-bursts were presented through two speakers (Fostex 256 PM-0.5 Sub and PM-0.5 MKII, Fostex Corporation, Tokyo, Japan). These were positioned 0.5 m above the water surface.

The experiments were performed in a test tank (50 cm × 27 cm × 30 cm; length × width × height; water depth 25 cm) whose bottom was covered with sand. The aquarium walls, except for the front glass, were lined on the inside with air-filled packing wrap in order to reduce resonances and reflections. Prior empirical tests using white noise and pulsed sounds had shown that lining reduced artefacts such a high frequency resonances, which are a known phenomenon of small tank acoustics in this setup 17. Akamatsu et al. [49, see their Figure 7c] and our own recordings (Figure 5) revealed resonance frequencies at 4 kHz and higher frequencies. Video and audio signals were stored synchronously on the hard disc of a DVD hard disc recorder (Panasonic DMR-EX95V, Panasonic Corporation, Osaka, Japan). A hydrophone (Brüel & Kjaer 8106, sensitivity -174 dB re 1 V per l Pa) was positioned about 5 cm off the center of the aquarium and connected to a microphone power supply (Brüel & Kjaer 2804) whose output was sent to the hard disc recorder. Simultaneous video recordings of fin movements were carried out by a video camera (Sony CCD-VX1E, Sony Corporation, Shinagawa, Tokyo, Japan) connected to the DVD hard disc recorder. Sound pressure levels (SPLs) were measured in parallel with the sound recordings, with the hydrophone power supply additionally connected to a sound level meter (Brüel & Kjaer Mediator 2238). Fish were held about 5 cm away from the hydrophone in the middle of the tank, with one pectoral fin fixed either by the frame of a fish net or the fingers of the testing person to avoid overlap of sounds produced by the left and right pectoral fins. According to Akamatsu et al. 49 a short recording distance reduces the effects of tank acoustics on SPLs and signal distortions. Sounds produced by both fins were recorded in each fish. All experiments were carried out in a walk-in soundproof room, which was constructed as a Faraday cage. Test tanks were placed on a vibration-isolated floor.

The VLC Media Player (VideoLAN, Club VIA, Châtenay Malabry, France, released under the GNU General Public License) was used to assign single sounds to right and left fins and to either abduction (of the body) or adduction (to the body) movement of the fins. All sounds were sampled at 44.1 kHz. Temporal characteristics of sounds were analysed using Raven Pro 1.3 for Windows (Bioacoustics Research Program, Cornell Laboratory of Ornithology, Ithaca, NY, USA) and spectral characteristics were analysed using STx 3.7 (Acoustics Research Institute, Austrian Academy of Sciences, Vienna, Austria).

Ten abduction sounds and 10 adduction sounds were analysed per individual with respect to total duration and pulse period (except for four specimens of group XS, four specimens of group S and one specimen of group M, where only one to five abduction and adduction sounds could be analysed because of a lack of vocalising activity in these individuals) (Figure 6). The dominant frequency of sounds for each individual was determined from the sound power spectrum calculated from all stridulatory sounds emitted by one specimen. A sound file made up of vocalisations emitted by all specimens of a size group was created separately to calculate group-specific cepstrum-smoothed sound spectra 50. Absolute sound spectra of the recordings were calculated as described previously 4651. A value bandwidth -10 dB was calculated for each size group, characterising the frequency range of sounds at a level 10 dB below the spectral sound level of the dominant frequency (Figure 7).

All data were tested for normal distribution utilising Shapiro-Wilk's test. When data were normally distributed, parametric statistical tests were applied. Mean hearing thresholds were determined for each size group and at each frequency, and audiograms were drawn using SigmaPlot 10.0 (Systat Software/Cranes Software Inc., Bangalore, India and San Jose, CA, USA). Means of sound characteristics were calculated for each fish and used for further analyses. Relationships between fish size (SL or logSL) and hearing thresholds or sound characteristics were determined by Pearson's correlation coefficients and linear regressions. The statistical tests were performed with the software SPSS 17.0 (SPSS Inc., Chicago, IL, USA). If it was evident that data points in a graph showed two different distribution patterns, segmented linear regressions and breakpoints were calculated with the software SegReg (R. J. Oosterbaan, Wageningen, The Netherlands).