Sex differences in frequencies in a species with modest sexual size dimorphism

ABSTRACT Sex differences in vocalizations are found across the animal kingdom, which may be due to different vocal apparatus, call function, and context. Rock hyraxes (Procavia capensis) of both sexes are vocal, but the sexes differ in repertoire size, call usage, and amplitude. In this study we examined sex differences in vocalization frequencies and predicted that they will be low and that frequency ranges will overlap since sexual size dimorphism in this species is modest (~ 15%). We utilized two datasets: recordings of captive hyraxes using a condenser microphone; and audio files of wild-living individuals recorded via a miniature recorder mounted on a collar. With the exclusion of two proximate call types recorded in the wild, all calls featured an ultrasonic component. However, in females there was an effect of duration on minimal frequency. Warning trills, which are heard by humans far away, featured the highest maximal visible harmonic in both datasets. No relationship was found between calling distance and the maximal harmonic in males, while in wild females, distant calls featured higher frequencies. Our results show sex differences in hyrax vocalization frequencies. Exploring the information encoded in frequencies beyond the human hearing range may expand our understanding of animal communication.


Introduction
Differences between male and female vocalisations in acoustic variables and frequencies can be observed in many mammalian species. Even where sexual dimorphism is small, there may be substantial sex differences in frequencies due to differences in vocal muscles length (Lenell and Johnson 2017) and reduction or absence of vocalisation apparatuses (Benz et al. 1990). Unique frequencies may also be used for sex-specific roles, for example in mother-offspring communication (Nakagawa et al. 2012). However, there are also generalities that apply to all vocalisations, regardless of sex. For example, to communicate over long distances, low frequencies that bypass obstacles are used, in contrast to high frequencies that are more readily absorbed in the environment (Titze 2000). To communicate between group members over short distances, high frequencies may have advantages. For example, in Richardson's ground squirrel (Spermophilus richardsonii), warning calls featured mainly ultrasonic frequencies when the source of threat was far away, as most of their predators cannot hear ultrasonic vocalisation (USVs; i.e. >20 kHz) (Wilson and Hare 2006). In Muridae both males and females use USVs for advertising their sexual and mating state (Sewell 1970;Musolf et al. 2010;Hanson et al. 2012), as well as their social status and territory (Moles and D'Amato 2000). However, some bats that use USVs for orientation, feeding, and predator identification (Møhl and Miller 1976), use low frequencies for socialising (Bohn et al. 2008;Behr et al. 2009).
Generally, a species' vocalisation and hearing range depend on its physiology and anatomy (Heffner and Heffner 1980;Ryan and Brenowitz 1985;Titze 2000), and an individual's vocalisation is shaped by its sex (Mendoza et al. 1996), individual traits (Ryan and Brenowitz 1985), arousal state (e.g. aggression levels (Koren et al. 2008;Demartsev et al. 2019b)), and habitat (Morton 1975;Ryan and Brenowitz 1985;Sugiura 2007). For example, individuals with larger body size usually have longer vocal tracts that produce and amplify lower frequencies (Titze 2000). Thus, in species where there is a pronounced sexual size dimorphism, sex differences are expected in vocal frequencies and amplitudes. The social context also influences sound frequencies, including the proximity to group members or other species (Sugiura 2007;Iacobucci et al. 2015). Piglets (Sus scrofa domesticus) produced more grunts (low frequency) and screams (high frequency) when close to the mother, but if far from her they produced mostly screams (Iacobucci et al. 2015). In Japanese macaques (Macaca fuscata), the use of high-frequency calls was mostly between group members when they dispersed (Sugiura 2007).
Here we test whether there are sex differences in vocalisation frequencies in the monomorphic rock hyrax (Procavia capensis), which shows a modest size dimorphism (15%). Rock hyrax are very vocal, and although there are no sex differences in their social status or testosterone levels (Koren et al. 2006;Koren and Geffen 2009a), there are sex differences in their vocal communication (Demartsev et al. 2019a). Proximate calls (i.e. growls, coos, whines, howls, twitters) are mostly used by females (Demartsev et al. 2019b) to communicate with group members (<2-3 metres away), while mostly males sing loud (80 dB) long-range songs (Koren et al. 2008). To date, rock hyrax calls have only been studied within the human hearing range (i.e. <20 kHz; (Fourie 1977;Koren et al. 2008). In this study, we recorded rock hyrax sonic and ultrasonic vocalisations. Considering the manifestation of USVs across taxa (Morton 1975;Moss 1988;Sales et al. 1988;Kalcounis-Rueppell et al. 2006;Behr et al. 2009;Musolf et al. 2010;Brinkløv et al. 2017;Cobo-Cuan et al. 2020), as well as the physical properties of high frequencies and energy-saving strategies that may relate to the low energetic cost of male hyrax calls (Ilany et al. 2013;Demartsev et al. 2019a), and modest sexual size dimorphism, we hypothesised that male and female rock hyrax vocalisations would feature similar frequencies.

Animal maintenance and recording in captivity
Ten hyraxes (5 females, 5 males) were kept in an outdoor enclosure (5.42 × 5.55 × 2.24 m; net density 9.5 × 1.4 cm) at Bar Ilan University (32.067778°, 34.8425°), Ramat-Gan, Israel as part of a long-term study of hyrax social behaviour. The enclosure included a big rockery in the middle and lairs for hiding. Two plywood sheds were placed in opposite ends of the enclosure, and fresh cabbage supplemented with bell peppers, carob (Ceratonia siliqua) leaves and pods, Greek strawberry tree (Arbutus andrachne) leaves and fruits and occasionally Nerium (Nerium oleander) leaves was supplied each morning on a concrete platform. The captive hyrax colony at Bar Ilan University included three wild-caught adult females (captured as adults with a male at Ein Ya'akov; 32.066°, 34.839° in the winter of 2016), and their sexually mature offspring that were born in captivity (two females and five males aged 15-20 months).
Hyraxes in the enclosure were trapped at the beginning of this study using Tomahawk live traps baited with fruit and vegetables and anesthetized using 0.1 ml/Kg ketamine HCl. Following physical inspection that included body measurements, individuals were marked with coloured collars, RFIDs (T-VAS bio-glass microchips; Datamars, Switzerland), and hair-dye (Protein colour, Henkel, Germany). The enclosure was monitored by two closed-circle TV cameras (Hikvision, model: DS-7104HQHI-F1/N) 24/7. Video recordings were examined for behavioural and contextual information. All behavioural occurrences were documented (Altmann 1974).
Hyrax vocalisations were recorded by the first author (GF) over 6 months (June -November 2019) using a single CM16/CMPA-P48 microphone with a USGH condenser, Ultrasound Gate 116HnbM recording system, and 5-metre-long XLR-5 microphone extension cable (Avisoft bioacoustics, Germany) at a frequency rate of 125 kHz, 16 bit and 3-seconds pre-record buffer. The microphone was secured inside the enclosure, while the observer sat in a fenced area outside the enclosure. Vocalisations were either spontaneous or induced via playbacks of pup screams that were recorded in our field site in Ein Gedi (Ilany et al. 2013;Goll et al. 2022). Recorded pup screams were broadcasted via a remote-controlled speaker (Foxpro Inferno, USA). Recording was activated upon the observer (GF) hearing a vocalisation.

Recordings of free-ranging hyrax
Rock hyraxes have been extensively studied in the Ein Gedi Nature Reserve for over 20 years (Koren et al. 2002;Koren and Geffen 2009b), and the field protocols have been published in previous studies (e.g. Geffen 2009b, 2011). Using our long-term dataset, we deduced that sexual size dimorphism in our study population is modest (mean male weight ± SD for N = 570 was 2.67 ± 0.45 Kg, while mean female weight ± SD for N = 733 was 2.3 ± 0.35 Kg). Fifteen marked hyraxes (ten females and five males) were recorded during the spring months of 2017 and 2018 by YG using custom-made micro electromechanical system (MEMS) sensors and microphones (Vesper loggers, Alexander Schwarts Developments, Israel) mounted on a collar (logger weight 1.85 gr, total collar weight 70 gr). The sampling rate was 150 kHz, at 8-bit format, and 50 Hz recording threshold. Each sensor continuously recorded all vocalisations for an average of 5.2 ± 2.4 days per individual hyrax (male average: 4.2 ± 3.8 days, female average: 5.6 ± 1.4 days). The collars were removed following recapture. Recordings were saved as binary files and converted into wav files using VesperStudio (version v6.3.30.0; Alexander Schwarts Developments, Israel). The files were examined and those with a high signal-to-noise ratio were analysed.

Acoustic and statistical analyses
The most frequent vocalisation types (Demartsev et al. 2019a) were analysed ( Figure 1): trills (alarm calls) and song elements (wails, chucks, and snorts) that are heard by humans from far (>100 metres), and sounds that are heard by humans only if they are within 2-3 metres of the hyrax (i.e. proximate calls): growls (aggressive), howls (aggressive), twitters (affiliative), and squeaks (submissive) (Demartsev et al. 2019a). We quantified ten samples for each vocalisation type per individual when available (87% of all recorded calls).
Sound analysis was performed using the Avisoft SASLab (pro version 5.2.07; Avisoft bioacoustics, Germany) at 256 fast Fourier transformation (FFT) length, 50% frame size, using the Hamming window, and 93.75% overlap with 1523 Hz bandwidth in captivity and FFT at 512 lengths, 100% frame size, Hamming window, and an overlap of 93.75%, bandwidth of 381 Hz, and 293 Hz resolution in the wild. The categorisation of vocal calls and the distance that they were heard were based on Demartsev et al. (2019a). Since the signal-to-noise ratio varied between recordings in the field, the gain was manually adjusted to produce a clear spectrogram. Using the cursors, we manually measured the start and end times of the fundamental frequency, as well as the minimal fundamental frequency (Fig. S1). The frequencies of the maximal visible harmonic was measured from the highest most consistent harmonic (Fig. S1). There was no difference in the variance (i.e. CV) between the frequencies of the minimal and the maximal visible harmonic (t 12 = 0.84, P = 0.21). Each call was measured five times by the first author (GF) and the mode (i.e. most frequent measurement) was used. Sound files with inconsistent measurements were excluded.
We used general linear mixed models (GLMM) to calculate the effect of sex, vocal proximity, and call duration on the minimal and maximal call frequencies that were visible. We set individual identity as a random effect and P-values were calculated using randomisations. All statistical analyses were conducted via JMP version 15 (SAS institute, USA).

Recordings from captive hyraxes
In the captive enclosure, 734 vocalisations were recorded in total, but we could only identify the performers of 245 calls. Thus, we omitted sex or identity data ( Figure 2); and no statistical analysis was performed on those data since we could not properly correct for pseudoreplication.
Although we recorded from a short distance, only 280 out of the 734 vocalisations featured an ultrasonic component (i.e. >20 kHz maximal frequency), and 15 of those had a sonic fundamental frequency accompanied with solely ultrasonic harmonics. While the highest fundamental frequencies were in twitters (5.3-14.6 kHz), the highest frequencies of the maximal visible harmonic were measured in trills (4.1-58.8 kHz; Figure 2, Table  S1). Both twitters and trills were characterised by a wide frequency range. Song elements (i.e. wails, chucks) had the lowest minimal frequencies (wail: 1.2-9.5 kHz; chuck: 1.4-6.5 kHz; Figure 2, Table S1).

Recordings from free-ranging hyraxes
Overall, 740 calls were analysed from the field: 506 by females and 234 by males. Our model (Table 1) revealed sex differences in both minimum fundamental frequencies (i.e. minimal frequencies), and frequencies of the maximal visible harmonic (i.e. maximal frequencies), and that female vocalisations featured higher minimal and maximal frequencies (i.e. related harmonics). However, the minimal frequency was mostly affected by call duration, with shorter calls featuring higher minimal frequencies. No other factor, including the calling distance (as previously categorised in (Demartsev et al. 2019a) or the interaction between sex and calling distance, had any significant effect on the minimal frequency.
The fundamental frequency of all calls was in the sonic range, with the highest fundamental frequencies recorded in twitters (9.9-13.7 kHz) and squeaks (8.7-9.6 kHz; Figure 3, Table S2). The highest visible maximal frequencies were in trills and twitters (i.e. >75 kHz). Calling distance category and its interaction with sex were also significant, as well as call duration. Male vocal call characteristics were not related to their distance category, but in females the highest measured frequencies were associated with the call distance category: the far (i.e. long-distance trills) featured higher visible maximal frequencies than the proximate calls (i.e. growl, howl, squeak, twitter; Table S2).
The frequencies of the maximal visible harmonic in howls (males: n = 35; females: n = 72) and growls (males: n = 36; females: n = 84) were in the sonic range ( Figure 3, Table  S2). The frequencies of the maximal visible harmonic in all other call types featured ultrasonic frequencies. Twitter, especially in females, featured the highest frequencies, with more than 77% (152/195) containing ultrasonic harmonics. In comparison, only 55.1% (43/78) of female squeaks and 3.7% (1/27) of male squeaks had an ultrasonic component. Long-distance (i.e. far) communication calls (101/183, 55.2%) contained more ultrasonic components than proximate calls (220/557, 39.5%). Table 1. The effect of sex, vocal proximity, and call duration on the minimum of the fundamental frequency and the frequencies of the maximal visible harmonic of free-ranging hyrax. Individual identity was set as random effect. The importance of each predictor was accessed by the total effect. P-values were calculated by randomisations, and significant relationships are in bold.

Discussion
In this study we recorded and analysed the frequencies present in male and female rock hyrax vocalisations both in captivity and in the wild. We found a difference between male and female vocalisation frequencies in free-ranging hyrax. This might be due to differences in laryngeal structure (Lenell and Johnson 2017) or functional sex differences (Smirnova et al. 2016). Generally, we found that calls that are involved in communicating with nearby conspecifics (i.e. growls, howls, squeaks, and twitters) featured lower fundamental and visible maximal frequencies than calls used to communicate to a greater distance (i.e. trills), regardless of sex. The detection of higher harmonics in long-range calls does not appear to be related to the fact that they are louder, since other loud, long-range calls (i.e. male song elements) do not contain ultrasonic frequencies. Perhaps males can vocalise song elements with better control than trills, which are emitted in times of stress. These trills might be selected to be heard from afar and perceived across species and age categories (Ancillotto et al. 2014), to both assemble or drive away conspecifics, as well as other species (e.g. as in Mexican free-tailed bat (Tadarida brasiliensis) 'protest calls' (Bohn et al. 2008); Pygmy marmosets (Cebuella pygmaea) 'long calls' (de la Torre and Snowdon 2002)). Although proximate calls were expected to include high frequencies, half of them did not contain an ultrasonic harmonic in the wild; and, in males, only one proximate call type (i.e. twitter) consistently included ultrasonic frequencies in the maximal visible harmonic. It is possible that USVs in rock hyraxes are non-adaptive biproducts of the vocal system. In females, contrary to what is predicted (Titze 2000), proximate calls featured lower frequencies in the maximal visible harmonic than more distant calls, and the minimal fundamental frequency was associated mainly with call duration. This strong interaction between sex and proximity could be attributed to the higher sociality of females and of males that reside with female groups (i.e. residents), as such males are also more vocal than bachelor males (Demartsev et al. 2019a(Demartsev et al. , 2019b. Our results revealed that call duration was related to vocal frequency. Higher frequencies may be harder to sustain while changing amplitude, since high-amplitude oscillations are dependent on an ongoing reduced damping pressure and increased stiffness of the vocal folds (Titze 2000). Thus, the greater physiological effort required might limit the duration of loud and highly modulated calls such as trills (Demartsev et al. 2019a). Our study in captivity allowed us to observe the context in which calls were used, and indeed we found that calls with low fundamental frequencies and narrow bandwidths (e.g. growls, squeaks, howls) were used in the context of agonistic interactions. These may have evolved to be 'loud and clear', while their narrow bandwidth may be selected for close-range communication. Aggressive call characteristics are generally relatively low and rough, due to hoarseness, or physiological or vocal effort (Poole et al. 1988;Titze 2000;Slocombe and Zuberbühler 2007;Bohn et al. 2008;Ancillotto and Russo 2016;Demartsev et al. 2019b;Weissman et al. 2019). Using very low frequencies may not be favourable as small animals usually cannot hear low frequencies (Heffner and Heffner 1980;Ryan and Brenowitz 1985). Similar constraints likely apply to high frequencies as well.
The relationship between the acoustic characteristics of a habitat and species adaptations has been well studied (Morton 1975;de la Torre and Snowdon 2002). However, very little attention has been focused on how males and females differ in their use of the acoustic space. Our results suggest that there are sex differences in hyrax vocalisations in multiple dimensions, including sound frequencies. Finally, the findings from this study indicate the importance of deepening our understanding of the soundscapes that surround us to include frequencies that are beyond the human hearing range (Moss 1988;Moles and D'Amato 2000;Kalcounis-Rueppell et al. 2006;Behr et al. 2009;Hanson et al. 2012;Nakagawa et al. 2012;Ramsier et al. 2012).