The classification follows Near and Thacker (2024).

Actinopterygii Woodward, 1891

Teleostei Müller, 1846

Gobiiformes Bleeker, 1859

Gobioidei Bleeker, 1849

Gobiidae Cuvier, 1816

Amblygobius phalaena (Valenciennes, 1837)

INDONESIA – Sulawesi • Spermonde Archipelago, island of Barrang Lompo; 05°02.99'S, 119°19.77'E; November 2005, October 2006.

Austrolethops wardi Whitley, 1935

INDONESIA – Sulawesi • Spermonde Archipelago, island of Barrang Lompo; 05°02.99'S, 119°19.77'E; November 2005, October 2006.

Cryptocentrus cinctus (Herre, 1836)

INDONESIA – Sulawesi • Spermonde Archipelago, island of Barrang Lompo; 05°02.99'S, 119°19.77'E; October 2006.

Valenciennea sexguttata (Valenciennes, 1837)

INDONESIA – Sulawesi • Spermonde Archipelago, island of Barrang Lompo; 05°02.99'S, 119°19.77'E; November 2005.

For size, trophic guilds and registration numbers see Table 1.

NMW a traditional acronym for the Ichthyological collection of the Natural History Museum Vienna (NHMW).

ZMH-ICH Ichthyological collection of the “Museum der Natur, Hamburg”.

All CT-Scans were made at University of Vienna, Department of Evolutionary Biology using a Bruker/SkyScan 1272 micro-CT system. The X-ray source tungsten target was set at a voltage of 60 kV and a current of 166 µA, with a 0.25 mm Al beam filter. Projection images were collected over a half rotation. The reconstructed voxel sizes ranged from 3 to 6.5 µm (isotropic).

Reconstructions of all individuals were done in the program NRecon, Version: 2.2.0.6 by Brian Metscher. The images for this study were generated using the Dragonfly software (Object Research Systems, version 2022.1). The brush tool was used to color each structure individually and create segmentations. For this, the density differences of different structures were used. Later, counter meshes of the segmentations were created for visualization.

The following trophic structures were segmented in Dragonfly: oral jaws: premaxilla, maxilla, dentary, anguloarticular; pharyngeal jaws: ceratobranchial 5 and pharyngobranchial 2–4; first gill arch with gill rakers.

The nomenclature of the bones follows Springer (1983) except for the terminology of the processes of the head (anterior part) of the maxilla, which follows Rojo (1991: fig. 16). The nomenclature of the muscles follows Datovo and Vari (2013).

Two specimens of A. wardi were cleared and double stained for bone and cartilage following Thieme et al. (2021), a method combining the protocols of Dingerkus and Uhler (1977), Taylor and van Dyke (1985) and Schnell et al. (2016). The larger specimen was partly dissected and images were taken using a Canon R5 with a Canon MP-E 65 mm lens. Images were stacked using Helicon Focus 8.3.7 and processed in GIMP 3.0.4 and Inkscape 1.3.

Measurements were taken from the images using ImageJ software. How length measurements were taken is shown in the second figure (Left upper jaw in lateral view). Measurements for standard length (SL) were taken with a digital caliper to the nearest of 0.01 mm and rounded to the next decimal point.

The upper jaws of all four investigated specimens consist of two paired bones, the tooth-bearing premaxilla and the dorsolateral maxilla (Figs 1, 2, 3).

Premaxilla

The premaxilla is an elongated and laterally compressed bone. It is about twice as long as high (including the ascending process), except for A. wardi. In this species, it is about three times longer than high and its posterior end is bent ventrally and ends below the rows of teeth (Fig. 4). The premaxilla bears four processes, the ascending process, the articular process, the postmaxillary process and the caudal process (Figs 4, 5). Anteriorly located are the ascending and the articular process and posteriorly the postmaxillary and the caudal process.

The ascending process is distinct in all four species but relatively short in A. wardi (31.2% vs. 45.3–56.7% in premaxilla length) compared to the other three species (Fig. 2). It tapers towards its tip, roundish and massive in A. wardi but flattened and compressed in the other species, especially in the sand-sifting species A. phalaena and V. sexguttata. It is, relative to the shaft of the premaxilla, oriented dorsally in A. phalaena and in C. cinctus but is relatively inclined anteriorly in A. wardi and in V. sexguttata. The ascending process is fused with the articular process in A. wardi but both processes are separated in the other species.

The articular process is large, roundish and laminar in C. cinctus, A. phalaena and V. sexguttata. It is positioned in an angle to the ascending process and separated by this process by a distinct gap. The articular process is distinctly reduced in size in A. wardi in comparison to the other species (4.7% vs. 22.5–33.8% in premaxilla length) (Fig. 2). Contrary to the former species, it is in line with the ascending process and fused to it, forming together a single branch, a combined ascending-articular process.

The postmaxillary process is long and very narrow in A. wardi, long in C. cinctus and shorter, higher and laminar in A. phalaena and V. sexguttata. In A. wardi and C. cinctus, the postmaxillary process reaches to the end of the premaxilla, but ends distant to it in A. phalaena and V. sexguttata (Figs 2, 3).

The caudal process is indistinct, contiguous with the long postmaxillary process in A. wardi and C. cinctus. It is distinct in A. phalaena and V. sexguttata. The caudal process is straight in all species except in A. wardi, where it is rounded and bent ventrally (Figs 2, 3).

Dentition of the premaxilla

The teeth of the premaxilla are similarly shaped caniniform in all four species. They are elongated, pointed and have a narrow base. The tips of all teeth are moderately curved towards the oral cavity in A. wardi, and more conspicuously curved in C. cinctus, A. phalaena and V. sexguttata. An apparent difference in the number of tooth rows across the species with most rows (six) in A. wardi, four rows in C. cinctus, two rows in A. phalaena and a single row in V. sexguttata. The number of rows on the premaxilla of A. wardi and C. cinctus narrows to two or to three rows posteriorly. The inner row is the shortest. The teeth of the external row are the largest, followed by the teeth of the internal row in A. wardi, C. cinctus and A. phalaena. The teeth in between are densely packed, of different sizes (small or medium sized) in the multi-rowed species A. wardi and C. cinctus.

The left premaxilla of A. wardi carries ~160 teeth, of C. cinctus ~120 teeth, of A. phalaena ~60 teeth and of V. sexguttata ~25 teeth. The anterior-most teeth of the external row are positioned ventral to the ascending process, the posterior most somewhat distant to the end of the premaxilla (Figs 4, 5).

Although the teeth shape of the premaxilla is in general similar in all four species, those of A. wardi are the most uniform (Figs 4A, 5A). The other three species have relatively long teeth in the external row anteriorly, a single enlarged tooth in C. cinctus at angle of the dentary, and four to five enlarged teeth in A. phalaena and V. sexguttata. The following teeth in this row are abruptly shorter (Figs 4B–D, 5B–D). Although the anterior most teeth of the external row of A. wardi are the longest, they are not distinctly longer than the following ones. The teeth of this row gradually decrease in size posteriorly (Figs 4A, 5A).

Maxilla

The maxilla is generally elongated, rod-like, semi-cylindrical anteriorly and laterally compressed posteriorly. It can be divided in an anterior cup-shaped head followed by a long and narrow shaft. It carries three processes, the internal (anteromedial) process and the external (anterolateral) process of the head and the maxillary process of the shaft (Figs 2, 3). With the internal and the external processes, the maxilla overlaps the articular process of the premaxilla. Both processes are roundish and laminar. The external process of the maxilla is medially deepened (palatine sulcus) allowing the contact with the maxillary process of the palatine. This is similar in all four species. However, the head and its two processes are smaller and less distinct in A. wardi (Figs 2, 3). Most notable is the absence of a maxillary process in A. wardi, a process which is present in the posterior third of the maxilla in the three other species (Figs 2, 3). The maxilla is longer than the premaxilla in C. cinctus, A. phalaena and V. sexguttata but both bones are of similar length in A. wardi.

The lower jaws of all four species consist of four pairs of bones (Figs 1, 6, 7), the tooth-bearing dentary, followed by the anguloarticular, the retroarticular and, medial, the coronomeckelian bone.

Dentary

The dentary is an elongate bone which continuously increases in depth posteriorly in C. cinctus, A. phalaena and V. sexguttata, but it is narrow anteriorly and increasing in width rapidly, resulting in a massive and large posterior part, in A. wardi. Therefore, its ventral portion is distinctly deeper and concave in A. wardi whereas it is more straight in C. cinctus, A. phalaena and V. sexguttata. The dentary has posteriorly two processes: the dorsal coronoid process somewhat rounded, well expressed in all four species, while the ventral process is pointed but relatively indistinct in A. wardi compared with the three other species (Figs 8, 9).

Dentition of the dentary

The teeth of the dentary are similarly formed in all four species, with the exception of one or two additional large, caniniform teeth in C. cinctus, A. phalaena and V. sexguttata. All teeth are caniniform, elongated, pointed, have a narrow base and are curved inward to the oral cavity. A major difference between the species is the number of tooth rows, with the most rows, five, in A. wardi, four rows in C. cinctus, three rows in A. phalaena and one row in V. sexguttata. The external row is short in all four species and covers approximately the anterior quarter to third of the dentary in A. wardi, C. cinctus and V. sexguttata and approximately half of the denary in A. phalaena. The number of rows on the dentary of A. wardi, C. cinctus and A. phalaena narrows from two to one row posteriorly.

The left dentary of A. wardi carries altogether 145 teeth in six rows, 15 teeth in the external row, 40 in the internal row and 90 teeth in between. The left dentary of C. cinctus carries 103 teeth in four rows, 12 in the external row, 18 in the internal row and 73 teeth in between. The left dentary of A. phalaena carries in total 52 teeth in three rows, 13 in the external row and 10 in the internal row and 29 teeth in between. The left dentary of V. sexguttata carries altogether 26 teeth in two rows, 11 teeth in the external row and 15 in the internal row. Although the teeth of the dentary are generally similar in all four species (except for the large caniniform teeth of C. cinctus, A. phalaena and V. sexguttata), the dentition of A. wardi is the most uniform. The other three species have five to seven enlarged and longer teeth anteriorly in the external row, posteriorly bordered by a large caniniform tooth. The teeth following in this row are distinctly shorter (Figs 8B–D, 9B–D). Although the anterior teeth of the external row of A. wardi are the longest, they gradually decrease in size posteriorly (Figs 8A, 9A).

Anguloarticular

The anguloarticular is anteriorly bifurcated and the dorsal fork inserts into the dentary medially in all four species (Figs 6, 7). However, the dorsal fork is massive and deep in A. wardi and slender in the C. cinctus, A. phalaena and V. sexguttata.

Retroarticular

The retroarticular is similarly pyramidial-shaped in all four species, but most robust in A. wardi (Figs 6, 7). Together with the anguloarticular, it forms the posterior end of the lower jaw in A. wardi and C. cinctus (Fig. 7A, B). It is somewhat more anteriorly positioned on the anguloarticular in A. phalaena and in V. sexguttata (Fig. 7C, D). The posteroventral edge of the retroarticular has a projection for the attachment of the interopercular-mandibular ligament. This projection is most prominent in A. wardi than in all four species, followed by the projection of the retroarticular of C. cinctus (Fig. 7A, B).

Coronomeckelian

The small coronomeckelian is located on the medial side of the anguloarticular, immediately at the dorsal rim of the groove where the remnants of the Meckel’s cartilage extend (Fig. 7). It is elongate in A. wardi and V. sexguttata and roundish in C. cinctus and A. phalaena.

Upper pharyngeal jaw

The upper pharyngeal jaws consist of the most dorsal elements of the branchial arches 2–4, the pharyngobranchials 2–4 (Figs 10, 11). Each carries a tooth plate. As a unit, these plates form the dorsal biting surface of the pharyngeal jaws. These three tooth plates have approximately a triangular shape. The pharyngeal tooth plate three is the largest in all four species. The anterior and the posterior edge of the pharyngeal tooth plate three are slightly overlapping with the posterior and anterior edges of the pharyngeal tooth plates two and four, respectively. The three plates are most massive and pierced by foramina in A. wardi and C. cinctus (Fig. 11A, B) but delicate and lattice-like in the sand-sifting species A. phalaena and V. sexguttata (Fig. 11C, D). The teeth are relatively massive and similar in shape and number in A. wardi and C. cinctus and very numerous, thin, delicate and cushion-like arranged in the sand-sifting species A. phalaena and V. sexguttata. The teeth of all four species are long anteriorly and gradually decrease in length posteriorly in the second half of the pharyngeal jaw. They are longest in the external row, decreasing in length medially.

Lower pharyngeal jaw

The ceratobranchial 5 forms the lower pharyngeal tooth plate of jaw. In A. wardi and C. cinctus ceratobranchial 5 is similar and has approximately the shape of a boomerang (Figs 10, 12A, B), while in A. phalaena and V. sexguttata it is almost equilaterally triangular (Fig. 12C, D). The ventral ridge and the well-developed ventromedial process extend close to the external margin of the lower pharyngeal jaw in all four species. This process is roundish with a wide base and is directed ventrally in A. wardi and C. cinctus, but is narrower, rectangular and projects anteroventrally in A. phalaena and V. sexguttata.

In all four species, the teeth are similar in size and shape on ceratobranchial 5 to those on the dorsal pharyngeal jaw. Four rows of teeth extend from anterior to posterior in A. wardi and C. cinctus, narrowing anteriorly to one row and posteriorly to two rows. The teeth of the internal row are the largest and most massive in these two species. The internal row is the longest in A. wardi and teeth of this row are the anterior most on the lower pharyngeal jaw. In contrast, the external row is the longest in C. cinctus and teeth of this row are the anteriormost. Furthermore, the size of the teeth of the other rows of A. wardi gradually decreases from the inside to the outside, with the external-most row having the shortest teeth. To the contrary, the teeth of the most external row of C. cinctus are longer than the teeth between the external and the internal row. They are shorter and distinctly less massive than the teeth of the internal row (Fig. 12A, B). The lower pharyngeal jaw of A. phalaena and V. sexguttata is covered by numerous narrow and long teeth that are cushion-like arranged with the largest teeth on the posteromedial edge. Especially dense and uniform is the dentition of this bone in A. phalaena (Fig. 12C).

Gill-rakers

Fig. 13

The gill-rakers are of similar size and shape in A. wardi and C. cinctus. They are ossified, blade-like with a pointed tip. Their curved surface is directed from ventro-laterally to dorso-medially and has serrations similar to a saw blade, while the other side is smooth. The gill-rakers increase in size from ventral to dorsal (Fig. 13A1–B) but decrease in A. phalaena (Fig. 13C). The ceratobranchial 1 carries 12 gill-rakers in a row in A. wardi, eight in C. cinctus and five in A. phalaena. No bony gill-rakers were developed in V. sexguttata (Fig. 13D). The gill-rakers extend spiraled ventrally from the caudal border of the ceratobranchial dorsally onto the rostral border in A. wardi, but extend in a more or less straight row on the rostral border in C. cinctus and A. phalaena.

Gut analysis (Liu et al. 2008) and stable isotope analyses (Kneer et al. 2008) showed that A. wardi relies on seagrass as food collected by the host shrimp and deposited in its burrow. Seagrass made up 94% of the gut content (Liu et al. 2008) and, based on the analysis of stable isotopes (¹³C and ¹⁵N), Kneer et al. (2008) stated that this goby is a primary consumer of detrital seagrass. Epiphytes on the seagrass leaves made up 3% of the ingested food, others (copepods, ostracodes, foraminifera) made up 1% (Liu et al. 2008; Kneer et al. 2008). These authors placed A. wardi in the category of browsers, herbivorous fishes which bite off pieces of plants, because the gut of this goby contained relatively big pieces of seagrass leaves and roots (Liu et al. 2008; Kneer et al. 2008: fig. 5C–H). The shrimp does not leave the burrow to collect seagrass actively but it is waiting in the entrance of its burrow to collect drifting detrital seagrass leaves. These leaves are shredded into pieces (up to several millimeters) and are stored in chambers or get incorporated into the burrow wall. The lining of the burrow wall with seagrass pieces was aided by agglutination of mucus excreted by mucus glands of the shrimp (Kneer et al. 2008).

Austrolethops wardi is dependent on seagrass collected by its host shrimp. Shredded seagrass leaves deposited by the shrimp in chambers (Kneer et al. 2008) do not require a special biting system to take up pieces of it. However, more force is needed to take up seagrass used by the shrimp for lining of the burrow walls. The shrimp does not collect fresh but detrital seagrass and the nutrition of this senescent seagrass depletes with maceration (Kneer et al. 2008). However, the seagrass used for the lining of the burrow wall is agglutinated by the mucus secreted by mucus glands of the shrimp and gives the lining a high organic content (Kneer et al. 2008). The lining of shrimp burrows is settled by microbes, meiofauna and foraminifera (Koller et al. 2006). It is very likely that it is the same in the lining of the burrow wall of the shrimp. Possibly, this organic enriched lining based on shredded seagrass is the primary food of A. wardi. It might be a superior food source as it had lower C/N ratios compared to seagrass material (Kneer et al. 2008). On the other hand, it would be only accessible for the goby by biting.

The basal system for jaw protrusion in acanthomorphs is characterized by a long ascending process of the premaxilla which is separated from the articular process and by a lateral expansion of the oral cavity during suction feeding (Gosline 1987; Kotrschal 1988; Clifton and Motta 1998). A more specialized feeding mode is characterized by a shortening of the ascending process and the fusion of the ascending and articular processes, which results in the reduction of the lateral expansion of the oral cavity. In such species with a single ascending-articular process as A. wardi, suction as primary feeding technique is replaced by biting, although suction is not lost (Gosline 1987; Kotrschal 1988). This increased the ability of biting instead of relying on suction feeding, opened up new trophic niches and the possibility to feed on sedentary animals and on plants (Gosline 1987).

The fusion of the articular process with the ascending process has evolved repeatedly and independently in several teleost groups. It is often accompanied by a reduction in size of the articular process to a more or less distinct shoulder (Gosline 1981, 1987; Linde et al. 2004), like in A. wardi. Separated ascending and articular processes are the plesiomorphic condition for Gobioidei. Both processes are separated in the species of the most basal gobioid families, in Rhyacichthys guilberti Dingerkus & Seret, 1992 (Reichenbacher et al. 2020: fig. 4A1, B1) and Protogobius attiti Watson & Pöllabauer, 1998 (Reichenbacher et al. 2020: fig. 5A1) (Rhyacichthyidae Jordan, 1923) and Perccottus glenii Dybowski, 1877 (Voskoboinikova and Pavlov 2006: fig. 3, Reichenbacher et al. 2020: fig. 5B) (Odontobutidae Hoese & Gill, 1993; Bauer et al. 2025: fig. 4H). The ascending and articular process of the premaxilla are also separated in Eleotridae (Gosline 1981: 22, Springer 1983: fig. 7, Reichenbacher et al. 2020, figs 3A, 5C), Thalasseleotrididae Gill & Mooi, 2012 (Goatley and Tornabene 2022: fig. 10), Gobiidae (e.g., Birdsong 1975: fig. 7E, Murdy 1985: figs 5–6, Gill 1993: Fig., 152, 9B, Reichenbacher et al. 2020: figs 3B, 5D) and Oxudercidae Günther, 1861 (e.g. Larson 2001: figs 36, 37, 79, 129, 210, Kindermann et al. 2007: figs 9C–D, 34A–B). The fusion and the position of the articular process in line with the ascending process appears to have evolved repeatedly and independently in Gobiidae (e.g., Huie et al. 2019: fig. 2A–C, Peinemann et al. 2024: fig. 4A, Vaz et al. 2025: figs 5A–B, 6A–D) and Oxudercidae (Larson 2001: figs 39A–B, 70, 197, 201–202, Michel et al. 2014: fig. 4) as well.

A reduced articular process is morphologically and functionally included into the ascending process in A. wardi supporting its massive base, likely to cope with the compressive forces generated by biting (Linde et al. 2004). Additionally, the ascending process of A. wardi is relatively short compared to other carnivorous gobiid fishes and is inclined anteriorly vs. vertical positioned in carnivorous Gobiidae (e.g., Huie et al. 2019: figs 1–2, Reichenbacher et al. 2020: figs 3–5) and Oxudercidae (Kindermann et al. 2007: figs 7–8, 9C–D, 33, 34A–B, Larson 2001: figs 51, 53, 61, 64, 129). Jaw protrusion in acanthomorph fishes is determined by the length of the ascending process (Gosline 1987; Cooper et al. 2017; Hulsey and Gorman 2023). In acanthomorphs with a biting feeding mode, upper jaw protrusion diminishes to improve the transmission of the biting force. This transmission is optimized by shortening of the ascending process and by increasing the angle between the ascending process and the dentigerous shaft of the premaxilla by inclining this process anteriorly (Bouton et al. 1999; Sibbing and Witte 2005), which is the case in A. wardi.

Considering these morphological specializations of the premaxilla, we assume that the feeding mode of A. wardi has majority shifted from suction to biting. This is supported by the lack of a maxillary process on the dorsal rim of the maxilla, the area where the dorsal part of the musculus adductor mandibulae (A1 of Motta 1984; Gosline 1987, maxillaris muscle of Gosline 1981, pars malaris of the adductor mandibulae of Datovo and Vari 2013; Cooper et al. 2017) inserts. Strong forces act on this part of the maxilla during suction feeding and may favor the formation of a process on the dorsal rim (e.g., Motta 1984: fig. 6, Gosline 1987: fig. 1E, Evans et al. 2024), also in gobioid fishes (Hoese and Gill 1993: 419, Larson 2001: fig. 18, Maie 2007: fig. 2.1.A). As the adductor mandibulae muscle is activated, it initiates the anterior shift and the rotation of the maxilla which again initiates the motion of the premaxilla (Gosline 1981; Motta 1984; Roberts et al. 2021; Hulsey and Gorman 2023). This rotation needs to be fast during suction feeding to capture elusive prey but not during biting to collect sedentary prey or plants (Gosline 1987; Maie et al. 2009). Such a maxillary process is missing in A. wardi but is developed in A. phalaena, C. cinctus and V. sexguttata.

Fish species with a biting feeding mode have stronger jaws than suction feeding species (Linde et al. 2004). We did not test the thickness of the jaw bones. However, the two larger bones of the lower jaw, especially the dentary, are massive and the latter is very deep in A. wardi compared with other gobiids, which increases force efficiency during biting (Sibbing and Witte 2005; Olivier et al. 2021). In a ventral view, both bones, the dentary and the angulo-articular, appear to be more massive and wider in A. wardi than in the other investigated species. Typically, gobiids with a vertical ascending process and separated from the articular process of the premaxilla have elongated and relatively slender lower jaws (e.g., Birdsong 1975: fig. 7F, Scsepka et al. 1999: fig. 4, Larson 2001: figs 36–37, 39, 64, 70, Kindermann et al. 2007: figs 7–9, 32–34, Reichenbacher et al. 2020, figs 5–6). Such jaws are generally adapted for the intake of elusive prey.

In contrast to the shape of the premaxilla, the dentition of A. wardi does not significantly differ from the dentition of C. cinctus and most carnivorous Gobioidei, except by the number of teeth. The teeth of the oral and pharyngeal jaws are caniniform, which is the ancestral tooth shape of teleost fishes (Bellwood et al. 2014; Davis et al. 2016). It has been retained in many carnivorous and omnivorous species (Jackman et al. 2013; Davis et al. 2016) and is the typical tooth shape of most Gobioidei (e.g., Birdsong 1975; Murdy 1985, 1989; Scsepka et al. 1999; Herler et al. 1999; Hoese and Reader 2001; Kramer et al. 2009, 2012; Reichenbacher et al. 2020). The tooth shape of Teleostei just increased in complexity with the shift to herbivorous feeding (Bellwood et al. 2014; Streit et al. 2015; Davis et al. 2016). The teeth of herbivore fishes are adapted to various methods to extract plants from the substratum and of various shapes, e.g., bicuspid, tricuspid, peg-shaped, flattened or chisel-like (Kido 1996; Streit et al. 2015; Davis et al. 2016). However, the teeth of A. wardi are not more complex than those of carnivorous gobies. The only difference in tooth morphology we found is that the tips of the oral teeth of A. wardi are less distinctly recurved than those of C. cinctus and other carnivorous Gobiidae (e.g., Kramer et al. 2009, 2012; Huie et al. 2019; Reichenbacher et al. 2020).

Unlike A. wardi and C. cinctus, the pharyngeal and oral dentition of A. phalaena and V. sexguttata is more specialized. The oral teeth are arranged in few rows (1–3) and are relatively robust, the teeth of the pharyngeal jaws are very numerous, thin and cushion-like arranged, which is obviously a derived character and an adaptation to a sand-sifting feeding mode (Murdy 1989: 62, figs 68, 70). The pharyngeal jaws of A. wardi and C. cinctus are elongated and not robust, typical for fishes with a carnivorous feeding mode (Ronco and Salzburger 2021).

The arrangement and the shape of the oral and pharyngeal teeth of A. wardi and of C. cinctus is similar and it is also similar to the dentition of the generalized shape of carnivorous Gobioidei (e.g., Murdy 1989; Scsepka et al. 1999; Herler et al. 1999, 2006; Kramer et al. 2009, 2012). Typically, the teeth of these fishes just differ in size, with the largest teeth of the multirowed species in the external row and smaller teeth in the internal row of the oral jaws but reverse in the pharyngeal jaws (e.g., Winterbottom and Burridge 1992; Gill 1993; Larson 2001; Hoese and Larson 2004). This is also characteristic for A. wardi. A distinct difference in the oral dentition of A. wardi compared to C. cinctus, A. phalaena and V. sexguttata is the size of the anterior teeth of the external tooth row of both jaws. These teeth are long but the following teeth are abruptly shorter in these three species. On the contrary, the dentition of A. wardi is more uniform in both oral jaws in that the anterior teeth of the external row are also the longest, but that there is no abrupt change in tooth length, as the teeth gradually decrease in size posteriorly. There is also no abrupt change in tooth length in many other gobioids, but contrary to A. wardi, most of them have some somewhat larger caniniform teeth in the external tooth row (e.g., Kindermann et al. 2007: figs 9C, D, 34A, Larson 2001: figs 39A, 70, 79, 94, 102, Kramer et al. 2009: figs 2A, C, 3A, C, 4A, C, Kramer et al. 2012 figs 2, 3, Huie et al. 2019: fig. 2B, C).

It is not clear if the uniform dentition of the oral jaws of A. wardi is directly linked to its herbivorous feeding mode. However, Orlandi Bonato et al. (2017) found that fishes with a uniform dentition along the external row of the jaws consumed higher proportions of plants. It should be considered, that the very large caniniform teeth on the dentary of C. cinctus, A. phalaena and V. sexguttata are likely not a trophic but an antagonistic character as it has been shown for blenniid fishes (Kotrschal and Goldschmid 1992). We assume that the only slightly curved front teeth of the oral jaws of A. wardi enable successful uptake of the seagrass fragments and that the large number of teeth (160 in the premaxilla and 145 in the dentary) may make it easier to grasp them.

In general, however, oral and pharyngeal jaws of many acanthomorph species function independently, the oral jaws perform prey capture and the pharyngeal jaws manipulate and process the prey (Roberts-Hugghis et al. 2025 and authors cited therein). This is seemingly the case in the two sand-sifting species A. phalaena and V. sexguttata with large pharyngeal jaws and, especially in A. phalaena, densely packed with narrow, elongated teeth that are cushion-like arranged, but not in A. wardi. We found no transformation of the pharyngeal jaws in A. wardi compared with the pharyngeal teeth and their pattern resembles those of C. cinctus and other carnivorous gobioids (Herler et al. 2006; Kindermann et al. 2007; Kramer et al. 2009, 2012). Seemingly, the morphology of the pharyngeal jaws is less affected by the type of ingested food than the morphology of the oral jaws. This is supported by observations of Kido (1996) who studied gobioid fishes with different feeding ecology, from carnivores to herbivores. This author (Kido 1996) found only little differences of the morphologies of the pharyngeal tooth plates. The teeth of the plates were similar in all species, pointed and slightly curved. Only the density of the dentition differed slightly between the investigated species.

Gill-rakers are important in food particle retention. Their number and size select the size of the food (Magnuson and Heitz 1971; Alsafy et al. 2025). The gill-rakers of fishes which predate on zoobenthos and zooplankton are relatively wide-spaced, those of herbivorous fishes often have additionally small projections (Alsafy et al. 2025). Kido (1996) found bony gill rakers only in gobioid species with the most generalized, carnivorous, feeding type. The herbivorous species had cartilaginous and cartilaginous/fleshy gill rakers. The gill rakers of A. wardi are strong, blade-like and bony, similar in size and shape to those of C. cinctus and other carnivorous, predatory gobiids (Hoese and Allаn 1977) and differ from those of herbivorous and sand-sifting species (Kido 1996, Hoese and Allаn 1997). Seemingly, their morphology reflects a carnivorous ancestry.