DISCUSSION
Freshwater sand gobies are considered important indicators for the conservation of Mediterranean inland aquatic ecosystems due to their wide range of habitats and high level of endemism (Vanhove, Kovačić & Zogaris, 2016). Our study investigated, for the first time, the sound production and reproductive intersexual behaviour of a freshwater endemic Mediterranean goby, the Neretva dwarf goby Orsinigobius croaticus, under laboratory conditions. In addition, we investigated multimodal signalling, using both acoustic and visual signals, in this protected and geographically restricted sand goby. The results of this study are consistent with prior research suggesting that gobies employ sound production, along with visual or chemical communication, as an important sensory modality for intraspecific communication (Lugli et al., 1995; Amorim & Neves, 2007; Malavasi et al., 2009; Amorim et al., 2013; Bolgan et al., 2013; Blom et al., 2016). O. croaticus is the ninth acoustically investigated European sand goby, excludingEconomidichthys pygmaeus , which was mute during the experiments. However, it is the only one with a high IUCN Red List classification status, listed as vulnerable (Crivelli, 2006, 2018).
Males of O. croaticus produced pulsatile sounds when interacting with females, during courtship, pre-spawning and spawning phases of the reproductive behaviour. Males did not produce sounds in all trials and calling rate varied between males and with female proximity. When males were in close contact with females or the prospective female approached/entered the nest, the calling rate would significantly increase from a few up to 10 sounds per min. Structurally, the pulsatile sounds in sand gobies are composed from a variable number (range 5 – 32) of pulses (organised in pulse trains), which are considered the fundamental units of this acoustic signal (Lindström & Lugli, 2000; Zeyl et al., 2016). Orsinigobius croaticus acoustic signals are short and low-frequency sounds (< 500 ms, ~ 140 Hz) composed from a short number of sound pulses with an average duration and period of around 15 ms and 32 Hz, respectively.
In this study, PD and PP differed significantly among males. In pulsed acoustic signals, PD can be related to body size and condition (Amorim et al., 2010) or temperature (Vicente et al., 2015), while PP is often dependent on temperature, but also reflects phylogenetic affinities in fish groups such as pomacentrids, cichlids, and sand gobies (Myrberg et al., 1978; Amorim et al., 2008, 2013; Vicente et al., 2015). Here, PP inO. croaticus increased with decreasing water temperature (~ 37 ms at 19°C vs . ~ 29 ms at 21.1°C). In P. pictus water temperature significantly influenced pulse period and explained 83% of its variability (Amorim et al., 2013). In ectothermic animals (such as fish), temperature-dependence in sound-producing central and peripheral mechanisms is corroborated, since it results from muscle activation (Bennett, 1985; Feher et al., 1998; Rome & Lindstedt, 1998; Kéver, Boyle, Parmentier, 2015; Vicente et al., 2015; Ladich, 2018).
Pulsatile sounds of O. croaticus males differed in all acoustic features except calling effort. These acoustic differences amongst soniferous males highlights the unique intraspecific acoustic variability of their reproductive sounds. In addition, several acoustic features were shown to be correlated to physical characteristics. We found an inverse effect of male size on sound frequency, since these two features were significantly and negatively related. This effect has previously been recognised in acoustic studies on sand gobies (Lindström & Lugli, 2000; Malavasi et al., 2008; Amorim et al., 2013). In terms of relationships, DUR and NP were highly and positively correlated in our study, whereas PRR and PP had a negative association. The strong correlation between DUR and NP suggests that the sound-producing mechanism is based on a fixed motor pattern (Parmentier & Lecchini, 2022).
In this study, O. croaticus males exhibited nine (visual) behavioural acts, confined to three distinct reproductive phases. The sound production in males was mostly associated with pre-spawning behaviours. Also, males exhibited courtship-related behaviours less frequently and for a shorter period then pre-spawning behaviours. These findings imply that the sound production is key in the mating process inO. croaticus and that it is likely efficient in transmitting information at only short-range distances (within one body length). Regarding the multimodal communication, soniferous O. croaticusmales differed in the frequency and occurrence of displayed behavioural categories (and their acts) when producing sounds and when they were mute, since most of the categories in the mute experiments were related to the courtship phase (outside the nest). Some behavioural acts, such as Pre-mating, Chase, Circling and Spawning, were completely absent from mute experiments. When producing sounds, Pre-mating and Nest display were the most frequent categories, indicating that males modulate their behaviour according to mate attraction investment. These findings could indicate that the multimodal signals, as produced by O. croaticusmales, could convey a wider set of information to the prospective breeding females, rather than using only one signal type. Indeed, males of different species, such as P. pictus , make a suite of signals from one or more modalities that females may use in mating decisions (Amorim & Neves, 2007; Amorim et al., 2013; Bro-Jørgensen, 2010). Multimodal signals, which are used by many species to communicate, contain components that can be analysed by multiple sensory channels (Otovic & Partan, 2009). Fish communicate through visual, chemical and acoustic signals often operating simultaneously to improve the chances of mating success, by indicating the physical quality or the motivation of the emitter (e.g., Levine, Lobel & MacNichol, 1980; Liley, 1982; Heuschele et al., 2009; Amorim et al., 2013). It has been suggested that this acoustic modality is highly advantageous for territorial species, in which the nest site is frequently hidden, and the male is out of sight from the prospective mate (Myrberg, 1981).
Another significant finding from the current study is that females entered the male’s territory, particularly the nest hollow, more frequently when accompanied by sound production than when the males were mute. In this study, the two males who received the most female entries were the largest. These two males exhibited the sounds with highest values of NP, FMi and PRR, suggesting that these acoustic features might be used to communicate important information to potential mates. Other studies suggest that different acoustic traits or morphological features could advertise male quality (genetic or phenotypic), serving as honest signals of different aspects of male quality in sand gobies (Knapp & Kovach, 1991; Amorim et al., 2013). According to Amorim et al. (2013), successful breeding P. pictus males produced more sounds and with a higher number of pulses than unsuccessful males.
Our findings indicate there are anatomical similarities in the musculo-skeletal system of the pectoral girdle between the previously studied Pomatoschistus gobies and O. croaticus (Adriaens et al., 1993; Parmentier et al., 2017). Our study provided the first anatomical dissections and μCT scans of the O. croaticus pectoral girdle and neurocranium. However, it is hypothesised that the Bauplan of soniferous gobies does not show deep significant modifications, meaning that the anatomy of soniferous species appears to be comparable to that of their mute relatives (Parmentier & Fine, 2016). To investigate the anatomy of the sound producing mechanism in gobies, Parmentier et al. (2013, 2017) undertook two empirical studies in two European gobies, gobiid Gobius paganellus (Gobiidae) and sand goby P. pictus (Gobionellidae), with the goal of testing the hypothesis of contraction of the pectoral girdle muscles. These multidisciplinary studies suggested strong similarities between the two gobies, and that sounds might be generated by the contraction of the levator pectoralis muscle. These results suggested that the pectoral girdle is most likely involved in sound production. It is worth noting that sound production was coupled with nodding in G. paganellus or with lateral head movements in P. pictus (Parmentier et al., 2013, 2017). However, this does not mean head movements are responsible for the sound production. In this study, the pectoral girdle of O. croaticus consists of three functional osseous parts, with main elements present as in other dissected sand gobies (Adriaens et al., 1993; Parmentier et al., 2013, 2017). In addition, the levator pectoralis muscles, divided into two bundles (pars lateralis andpars medialis ), were also found in O. croaticus , originating on the neurocranium and inserting onto the pectoral girdle. Four large radial bones were also present, forming the shoulder plate inO. croaticus . Lastly, the males performed lateral head movements during sound emission. Some authors suggest that certain sound characteristics are positively correlated with temperature if pulses are directly based on sonic muscle contractions (Ladich, 2018). Although our study did not include methodologies such as muscle histology, high-speed video, or electromyography to fully corroborate the findings from earlier research, we believe there is sufficient evidence to hypothesise that the assumed sound producing mechanism in O. croaticus could be related with the contractions of the levator pectoralis(pars lateralis and medialis ) muscles. Our assumptions are based on: 1) the observed anatomical similarities (i.e., muscle organization) between O. croaticus and other tested sand gobies, 2) prominent temperature-dependence of the peripheral (muscular) part of the sonic mechanism (as seen from the correlation of acoustic features with water temperature, and PP variation), and 3) head lateral movements observed during sound emission. Interestingly, in some situations, males were observed to perform body movements (lateral movements, head uplift, erection of fins), but without sound production, indicating that sound production requires more than just body movements. This supports the hypothesis that sounds are intentional and not only a by-product of other activities such as breathing, feeding or swimming.
Sand gobies are highly similar morphologically (Kovačić, 2008) and frequently live in sympatry (Miller, 1986), making their discrimination difficult. Several discrimination techniques have previously been proposed for gobioids, such as mitochondrial/nuclear DNA markers (Agorreta et al., 2013; Vanhove et al., 2012; Thacker et al., 2018), otoliths in the inner ear (Lombarte et al., 2018) and behaviour (Malavasi et al., 2012). Recently, the sounds (and their acoustic features) have become a useful parameter in determining the phylogenetic relationships in fish (Rice & Bass, 2009; Parmentier et al., 2009; Mélotte et al., 2016; Bolgan et al., 2020), particularly in gobies (Malavasi et al., 2008; Horvatić et al., 2021). The aim of this study was not to infer the phylogenetic relationships between sand gobies, but rather to investigate how the species can be separated according to their acoustic features, and how well the sounds can be classified for each taxon. However, qualitatively, the pulsatile sounds of soniferous sand gobies from this study are similar in that they are composed of a series of pulses (Figure 9 ), though when examined quantitatively (using a multivariate approach), they were discriminated according to their spectral and temporal acoustic parameters. In the present study, we found interspecific differences among the sand gobies species based on acoustic properties. The LDA assigned each sound produced by sand gobies to the correct species with a discrimination rate of 86%. On the scatterplot, acoustic variables NP and FM contributed to the separation of species in the negative direction, while DUR, PF and PRR separated species in the positive direction. Note, however, that species were recorded at different temperatures and results should be taken with caution. The observed interspecific differences, although based on a limited dataset, shed light on the taxonomic position and affinities of the genera Orsinigobius andNinnigobius , relative to the rest of the sand gobies.Ninnigobius canestrinii and K. panizzae , along withP. pictus , were the species most separated from the other taxa on the LDS bi-plot. Some authors have opposed the separation of O. croaticus and O. punctatissimus into the genusOrsinigobius , and the isolation of N. canestrinii from the genus Pomatoschistus (Thourgard et al., 2021). On the LDS bi-plot, the two Orsinigobius taxa were closely situated.Pomatoschistus taxonomy is currently complicated, but P. minutus from our study was in the close proximity of the twoOrsinigobius taxa. Interestingly, the hulls of the two populations of P. marmoratus overlapped in LDA, despite the fact they encompass individuals from a wide geographic area (the Po River delta in Italy and Parede/Arrábida in Portugal). However, the Italian population appeared partially isolated from the rest of the species. When applying the reduced dataset, the classification rate in LDA decreased from 86% to 69%, which is not an unacceptable outcome, though it implies that interspecific discrimination becomes more difficult without certain acoustic features, such as temperature-dependent DUR and PRR in our case.