Chewing or not? Intraoral food processing in a salamandrid newt

Food processing involves any type of mechanical manipulation of food before swallowing and includes crushing, puncturing, shearing and grinding (Schwenk and Rubega, 2005). Mechanical processing of food facilitates chemical dissociation and nutrient resorption by the digestive tract and, thus, increases the efficiency of energy exploitation from a given food source (Bramble and Wake, 1985; Schwenk, 2000a,,b). Processing mechanisms differ substantially across vertebrate groups but coordinated rhythmic and cyclic movements of the jaw, skull and hyobranchial (tongue) system are common in cartilaginous and ray-finned fishes, lungfishes and amniotes (Bemis and Lauder, 1986; Dean et al., 2005; Gans et al., 1978; Gans and Vree, 1986; Gintof et al., 2010; Herrel et al., 1999; Sanford and Lauder, 1989; Schwenk and Rubega, 2005; Schwenk and Wake, 1993; Wainwright et al., 1989). Whereas some cartilaginous fishes, including sharks and rays, use rhythmic chewing to process food within their mandibular jaw systems (Dean et al., 2005; Kolmann et al., 2016), ray-finned fishes exhibit three ‘jaw systems’ for food processing: (i) raking, using the tongue–bite apparatus (Camp et al., 2009; Hilton, 2001; Konow et al., 2013; Konow and Sanford, 2008; Sanford and Lauder, 1989, 1990), (ii) grinding, using the pharyngeal jaw apparatus (referred to as ‘pharyngognathy’) (Gidmark et al., 2014; Liem and Greenwood, 1981; Wainwright, 2002; Wainwright et al., 1989) and (iii) chewing, using the mandibular jaw apparatus (Fernandez and Motta, 1997; Gintof et al., 2010; Konow and Sanford, 2008; Lauder, 1981). While raking and pharyngognathy are derived mechanisms that only occur in some ray-finned fish groups, chewing occurs in both fishes and amniotes (Gans et al., 1978; Gintof et al., 2010; Herring et al., 2001; Hiiemae and Crompton, 1985; Schwenk, 2000a; Schwenk and Rubega, 2005). By contrast to fishes, amniotes additionally rely on a derived anatomical feature for intraoral processing: their muscular and highly movable tongue (Iwasaki, 2002). However, the coordination of jaw and tongue movements across amniotes is strikingly similar and it has been suggested that cyclic intraoral processing shares a common origin associated with the tetrapod terrestrialization process (Reilly et al., 2001). If so, and considering that behaviours are genetically determined, with more closely related species generally showing more similarities than distantly related ones (Katz, 2011), we hypothesize that there are similar mechanisms for food processing among members of the extant sister group to amniotes: the lissamphibians.

The question whether lissamphibians process their food, however, remains virtually unaddressed as it has become widely accepted that lissamphibians nearly universally omit food processing and, with only a few exceptions, swallow their food whole and unreduced (Bemis et al., 1983; Lauder and Gillis, 1997; Reilly and Lauder, 1990; Schwenk and Rubega, 2005). Aside from examples of rudimentary processing such as powerful bites, prey shaking or spinning (Bemis et al., 1983; Deban and Wake, 2000; Fortuny et al., 2015; Lukanov et al., 2016; Measey and Herrel, 2006; O'Reilly, 2000; Summers and Wake, 2005; Tanner, 1971; Wake and Deban, 2000), it had been suggested that some salamanders might use palatal dentition and tongue movements to manipulate prey (Deban and Wake, 2000; Regal, 1966; Reilly, 1996). Still, the only elaborate processing mechanism involving complex and rhythmic movements demonstrated so far occurs in plethodontid salamanders from the genus Desmognathus. These salamanders employ cyclic head bobbing movements once prey is held between the mandibular jaws, which deliver a series of strong bites (Dalrymple et al., 1985; Deban and Richardson, 2017; Larsen and Beneski, 1988; Schwenk and Wake, 1993). We have observed a similarly elaborate behaviour that follows food capture in the salamandrid newt Triturus carnifex. The behaviour involves of the order of 9 sequential cycles of ‘head bobbing’, in concert with rhythmic movements of the jaw and tongue apparatus (Movie 1).

Here, we studied the kinematics of the rhythmic post-capture behaviour in T. carnifex. The rhythmicity of this intraoral behaviour, as well as the apparent similarities with intraoral processing behaviours seen in other gnathostomes, led us to hypothesize that T. carnifex might use a hitherto undescribed food-processing mechanism. We combined data from high-speed fluoroscopy, three-dimensional anatomical reconstructions by means of micro-computed tomography (µCT), and analyses of stomach content to describe the mechanism underlying the processing behaviour. Our comparisons of the mechanism seen in T. carnifex and processing mechanisms used by other gnathostomes seek to develop a better understanding of the diversity and evolution of food-processing and intraoral cyclic behaviours across gnathostome vertebrates.

Seven adult alpine crested newts, Triturus carnifex (Laurenti 1768), with snout–vent lengths of 80.5±10.6 mm and a mass of 10.4±2.6 g (mean±s.d.), were used in this study. The animals were collected in their aquatic phase between April and June 2011 and 2012 in Lower Austria, Austria, with collection permission (RU5-BE-18/022-2011) granted by the local government of Lower Austria. Animals were group-housed in large tanks with a water level of 20 cm and an easily accessible land area with piles of cork bark pieces. The water was permanently filtered by an external trickle filter and the top of the tanks was covered with a removable mosquito net to prevent newts from escaping. The animals were fed twice a week with a variety of red mosquito larvae (Chironomids), firebrats (Thermobia domestica), earthworms (Lumbricids) and maggots (Lucilia sp.). For the experiments, we fed maggots as standardized prey items and because dipteran larvae are part of the natural food source of T. carnifex (Romano et al., 2012). Like other newts, T. carnifex seasonally changes between aquatic and terrestrial lifestyles (Griffiths, 1997) but for the experiments described herein, all newts were in their terrestrial phase for at least 3 weeks prior to data collection. Preliminary experiments (data used for observation purposes only) were performed at the University of Antwerp, Belgium, and the main experimental part at the Friedrich-Schiller-University of Jena, Germany. Accordingly, husbandry and experiments were approved by the Ethical Commission for Animal Experiments of the University of Antwerp (code: 2010-36) and the Committee for Animal Research of the State of Thuringia, Germany (animal experiment codes: 02-042/14, 02-008/15, animal husbandry code: J-SHK-2684-05-04-05-07/14).

At the University of Antwerp, five newts were surgically implanted with radio-opaque metal markers on the skeletal structures of interest [following modified protocols of Herrel et al. (2000) and Manzano et al. (2008)]. The animals were anaesthetized with buffered (pH 7.2) aqueous 0.05% tricaine methanesulfonate (MS222) solution and markers were percutaneously implanted by using hypodermic needles on the basibranchial (‘tongue bone’) and in two animals on the snout tip (between the premaxillary upper jaw bones) and the lower jaw tip (in the region of the dentary symphysis). Immediately after implantation, marker placement was verified using X-ray images. All animals were given at least 3 days of post-surgery recovery before the start of X-ray recordings.

The newts were placed on a moistened tissue in a Plexiglas enclosure mounted on the experimental table of the X-ray setup. For the preliminary experiments performed at the University of Antwerp, we used a Tridoros-Optimatic 880 X-ray apparatus (Siemens, Erlangen, Germany); for the experiments at the University of Jena, a custom-build biplanar Neurostar setup (Siemens, Erlangen, Germany) was used. After acclimation, newts were fed maggots (29.8±5.1 mg, mean±s.d.) and in order to visualize the maggots in X-ray recordings, we glued small tantalum markers (diameter of 0.5 mm) to their cuticle. In total, 50 feeding events were recorded from which 106 processing cycles were extracted for statistical analyses described below (10, 21, 22, 24, 29 cycles for individuals 1–5, respectively). X-ray recordings were taken from the laterolateral and dorsoventral projections at 40 kV and 53 mA with a sampling frequency of 250 Hz. The dorsoventral recordings were performed to determine lateral movements of tongue and jaw systems during processing, but as no clear lateral movements were measured, they were excluded from further analyses. However, the dorsoventral image plane was used for the X-ROMM analyses (see below). Next, the resulting raw video recordings were filtered (e.g. gamma correction, contrast, sharpness) and the horizontal (x-axis) and vertical (y-axis) coordinates of previously defined landmarks (Fig. 1) were tracked frame by frame using SimiMotion software (SimiMotion Systems, Unterschleißheim, Germany). The 2D displacement of the landmarks was used to calculate the following movements: (1) jaw movements: angular displacement of the upper and lower jaw at the point ‘occipital’ (jaw joint was not always visible in the X-ray movies so jaw displacement was measured at the point ‘occipital’) (Fig. 1A); (2) head rotation: angular displacement between the two linear slopes connecting (i) the points ‘occipital’ and ‘snout tip’ and (ii) the points ‘first vertebra’ and ‘fifth vertebra’ (Fig. 1B); (3) longitudinal tongue movement: horizontal (i.e. parallel to the linear slope connecting the points ‘occipital’ and ‘snout tip’) displacement of the tongue relative to the point ‘occipital’; (4) vertical tongue movement: vertical displacement of the tongue relative to the linear slope connecting the points ‘occipital’ and ‘snout tip’; (5) longitudinal transport of the prey: horizontal (i.e. parallel to the linear slope connecting the points ‘occipital’ and ‘snout tip’) displacement of the prey relative to the point ‘occipital’; (6) vertical movement of the prey: displacement of the point ‘prey’ relative to the linear slope connecting the points ‘occipital’ and ‘snout tip’ (Fig. 1C).

From movements 1–4, we calculated the kinematic variables summarized in Table 1. To account for different head sizes between individuals, all displacement values for tongue movements were normalized as percentage of the respective cranial length. The cranial length was measured from the laterolateral X-ray recordings and defined as the distance between premaxillary and the occipital condyles (Fig. 1B,C). Calculations and graphic illustrations were performed using Microsoft Excel 2010, custom-written scripts for Matlab (MathWorks, Natick, MA, USA) and the open source software Inkscape.

From the descriptive kinematics (examples shown in Figs 2 and 3), we determined relationships between tongue, head and jaw movements and used bivariate correlations to compare coordination between movements (Wainwright et al., 2008). Specifically, we hypothesized tongue, head and gape cycles to be temporally linked. Furthermore, tongue protraction, tongue elevation and head depression on the one hand and tongue retraction, tongue depression and head elevation on the other hand were expected to show a high degree of temporal overlap. Based on these hypothesized links and temporal overlaps, we expected functional coordination that was quantitatively tested by performing bivariate correlations between the respective kinematic variables. We tested for (i) correlations between durations of total gape, head and tongue cycles (tongue: both horizontal and vertical movements) and (ii) correlations between tongue protraction, tongue elevation and head depression as well as between tongue retraction, tongue depression and head elevation. In the first approach (i), we only tested temporal variables while in the latter (ii), we tested temporal and magnitude variables (the latter measured as translations or rotations). To account for the multiple tests performed (18), the P-value was corrected after Bonferroni to P≤0.0028. All statistical analyses were performed using Microsoft Excel 2010 and SPSS Statistics 20 software package (IBM).

In order to (i) analyse the condition of processed and swallowed maggots and (ii) study the morphology of the skull with special emphasis on the dentition pattern, two metamorphosed adult newts (both males with snout–vent lengths of 62 and 70 mm) were used (they had not been used in the X-ray experiments). The newts were first fed maggots ad libitum: in total 19 (8 and 11, respectively) maggots were consumed by the two newts. The animals were then anaesthetized and subsequently euthanized by immersion in an aqueous solution of 0.5% MS222 buffered to pH 7.2 (Leary et al., 2013). The heads were removed post-mortem and fixed in 4% buffered formaldehyde solution. Next, the stomachs were removed and the maggots contained within were transferred into 70% ethanol solution for preservation. After 2 days, all maggots were analysed using a stereo-microscope and photographed to document punctures and lacerations caused by intraoral processing. As a control, we used (i) 10 unprocessed intact maggots and (ii) 10 unprocessed maggots that were pierced with a needle to visualize a puncture in the cuticle (to prevent misinterpretation of natural structures such as tracheal openings as punctures). Both controls were immersed in 70% ethanol for 2 days.

For µCT scanning, two newts (both males) were fixed in 4% formaldehyde for 1 month. Then, specimens were dehydrated in a graded series of ethanol and mounted in Falcon tubes. A scan of the whole head was acquired using a SkyScan 1174 (Bruker, Belgium) μCT scanner with a source voltage of 50 kV and an isovolumetric voxel resolution of 7.39 µm. After image acquisition, image stacks were imported into the 3D software package AMIRA 4 (FEI Visualization Sciences Group, Merignac Cedex, France). Based on tomographic image data, relevant structures were segmented by threshold segmentation and visualized using surface renderings.

The goal of our XROMM analyses (see Movie 2) was to animate and reconstruct 3D skeletal movements. We followed the standard protocol for Scientific Rotoscoping (Brainerd et al., 2010; Gatesy et al., 2010). In short, polygonal mesh models of the skull, lower jaw and bony hyobranchial elements (derived from µCT scans) were built and a digital avatar of the skeletal elements was constructed. Next, the avatar was aligned to the calibrated biplanar X-ray projections using the XROMM toolbox in Maya (Alias Systems Corporation, Toronto, ON, Canada) and animations were created (Brainerd et al., 2010).

Prey capture was always by the tongue and tongue retraction resulted in placement of the maggot prey (i) between the jaws (only a few cases) or (ii) directly behind the margins of the jaws as the gape was closed. After prey had been transported into the oral cavity, 8.8±3.4 (mean±s.d.) rhythmic cycles involving movements of the skull, jaw and tongue skeletal elements started. The head was rhythmically elevated and depressed, the jaw opened and closed (Movies 1–3) and the tongue moved in an elliptic loop in the lateral view (Figs 2 and 3). With the nose of the subjects pointing left, the tongue motion loop progressed in the counter-clockwise direction. The movement of the prey inside the oral cavity also progressed in a counter-clockwise loop; the prey was first moved dorsally and slightly anteriorly and then ventrally and posteriorly. Accordingly, the prey was first pressed against the roof of mouth during tongue protraction and then moved away from the mouth roof during tongue retraction. One processing cycle was defined as being from the start of tongue protraction until the completion of tongue retraction. Horizontal tongue movements were chosen as the reference, because in contrast to gape movements they could be clearly assigned to dorsal and ventral head rotation as well as tongue elevation and depression phases. A representative kinematic profile is shown in Fig. 2 and descriptive statistics (Table 1) reveal that during a given processing cycle, the tongue was protracted within 346±119 ms (mean±s.d.) to a peak of 28.7±7.9% cranial length and elevated within 300±94 ms to a maximum of 28.8±9.4% cranial length. Meanwhile, the head was depressed within 340±109 ms over an angle of 29.7±10.8 deg. After the tongue was maximally protracted, it was retracted within 161±41 ms to 29.1±8.1% cranial length and depressed within 208±76 ms to 29.4±9.3% cranial length. At about the same time as tongue retraction, the head was elevated within 168±64 ms over an angle of 29.6±9.7 deg. Gape opening (25.1±3.9 deg within 313±90 ms) and gape closing (25.1±4 deg within 207±91 ms) could not be categorically assigned to any phase of horizontal tongue movement as the tongue is partly protracted and retracted during both gape opening and closing phases. Gape phases only seemed to roughly correspond with vertical tongue movements.

Spearman's Rho correlation revealed 18 significant correlations (see Figs 4–6). We tested for correlations between (i) total cycle duration and (ii) sub-movements. Regarding total cycle duration, all variables correlated with each other. Specifically, the duration of the horizontal tongue cycle correlated significantly with the duration of the vertical tongue (r_s=0.72; P<0.001; Fig. 4A), head (r_s=0.96; P<0.001; Fig. 4B) and gape cycles (r_s=0.83; P<0.001; Fig. 4C). The duration of the vertical tongue cycle correlated with the duration of the head (r_s=0.76; P<0.001; Fig. 4D), gape (r_s=0.82; P<0.001; Fig. 4E) and horizontal tongue cycles (see above). The duration of the head cycle correlated with gape cycle duration (r_s=0.86; P<0.001; Fig. 4F), as well as horizontal and vertical tongue cycle duration (see above). When testing the single movement phases, we found significant correlations between the following variables: duration of tongue protraction correlated significantly with duration of tongue elevation (r_s=0.85; P<0.001; Fig. 5A) and head depression (r_s=0.90; P<0.001; Fig. 5B); duration of tongue elevation correlated significantly with duration of tongue protraction (see above) and head depression (r_s=0.84; P<0.001; Fig. 5C); duration of tongue retraction correlated significantly with duration of tongue depression (r_s=0.52; P<0.001; Fig. 5D) and head elevation (r_s=0.56; P<0.001; Fig. 5E); duration of tongue depression correlated significantly with duration of tongue retraction (see above) and head elevation (r_s=0.82; P<0.001; Fig. 5F). The magnitude of tongue protraction correlated significantly with the magnitude of tongue elevation (r_s=0.68; P<0.001; Fig. 6A) and head depression (r_s=0.68; P<0.001; Fig. 6B). The magnitude of tongue elevation correlated significantly with the magnitude of tongue protraction (see above) and head depression (r_s=0.56; P<0.001; Fig. 6C). Similarly, the magnitude of tongue retraction correlated significantly with the magnitude of tongue depression (r_s=0.74; P<0.001; Fig. 6D) and head elevation (r_s=0.74; P<0.001; Fig. 6E). The magnitude of tongue depression correlated significantly with the magnitude of tongue retraction (see above) and head elevation (r_s=0.69; P<0.001; Fig. 6F).

Salamandrid skull morphology is described in detail elsewhere (e.g. Francis, 1934; Ivanović and Arntzen, 2017; Trueb, 1993); we focus here on observations relevant to food processing. Teeth are found on both the upper (premaxilla and maxilla) and lower (dentary) jaw bones, as well as on the roof of the mouth, specifically the vomerine bones (Fig. 7A). The vomers are flattened bony plates positioned anteriorly in the oral roof between the premaxillae and maxillae. From these flattened plates, a tooth bearing posterior vomerine process extends caudally up to the level of the squamosal base (Fig. 7A) where the rod-like vomerine process overlies the large parasphenoid. The vomerine teeth are arranged in an arc-like fashion at the interface between the flattened vomer and posterior vomerine process. The dentition extends posteriorly along the vomerine process to form two parallel rows of teeth. The vomerine teeth are about half the size of the teeth found on the jaws, but densely arranged, sharply pointed and slightly posteriorly recurved (Fig. 7A). Medially to the vomerine tooth rows lie two additional rows of small denticles (Fig. 7A).

Both the newts that were fed maggots for further stomach contents analyses showed the characteristic – presumably processing – behaviour (described in the kinematics section) after prey capture. Maggots were captured by the tongue and transported directly beyond the jaws, so puncturing of the maggots by the closing jaws can be excluded in these experiments. Microscopic examinations revealed clear lesions characterized by a distinct outline all over the surface of the processed maggots (Fig. 7B). In contrast, the control maggots only showed the puncture that was manually caused with a needle (Fig. 7C). Apart from that manually induced puncture, no further lesions were evident. The lesions caused by processing were characteristically small roundish perforations with a diameter of 30–50 µm or elongated incisions of up to 500 µm length. On average, the 19 processed maggots showed 21.6±11.6 (mean±s.d.) lesions.

Our experiments on T. carnifex reveal a previously undescribed processing mechanism for lissamphibians that involves rhythmic and cyclic movements of the skull, jaw and tongue (hyolingual) elements. Below, we discuss how food processing is achieved using these element movements, and how this behaviour compares with food-processing behaviours in other vertebrate groups.

How and where does food processing take place inside the mouth of Triturus? Like most other post-metamorphic salamandrids, T. carnifex has two parallel-running lateral rows of vomerine teeth on the roof of the mouth (Trueb, 1993). Our kinematics data reveal that as the tongue moves anteriorly and dorsally, the skull is depressed, and when the tongue moves posteriorly and ventrally, the skull is elevated. In a cyclic context, the effect of this motion pattern is an anteriorly and dorsally directed movement of the tongue to translate food across the palate, which is adorned with dentition (Figs 2, 3 and 7A). Accordingly, prey is cyclically pressed against and translated across the needle-like vomerine teeth, causing the prey to be pierced. The coordination of anterior tongue movement with respect to head depression (Figs 2 and 5) may increase the mechanical resistance between the protracting tongue and the palate, which in turn is likely to increase rasping efficiency. Our observations of processed maggots from newt stomachs revealed multiple cuticle perforations that are lacking in control maggots that have not undergone food processing (Fig. 7B,C).

The rhythmic and cyclic oral behaviour observed in T. carnifex results in mechanical processing of food. This is a distinct behaviour from the intraoral transport of food described for ambystomatid salamanders (Reilly and Lauder, 1990, 1991). Ambystomatid intraoral behaviours are rhythmic, but supposedly only serve to move food away from the mouth aperture and towards the oesophagus and so do not process food. The transport cycles described for ambystomatid salamanders are different from the processing cycles in T. carnifex. For instance, the ambystomatid tongue is retracted and depressed during gape opening and the first part of gape closing, stays relatively stationary during the second part of gape closing and only slowly starts protracting and elevating after gape closure. In T. carnifex, during the gape-opening phase, the tongue is first protracted and elevated and then retracted and depressed. During the gape-closing phase, the tongue first continues retracting and depressing after which it starts protracting and elevating (Fig. 2). Therefore, there are obvious differences in the coordination of tongue and jaw movements between ambystomatid transport and salamandrid processing. However, it remains unclear how the mechanics of food processing in T. carnifex compare with processing mechanisms in other lissamphibians.

There have only been a few descriptions of mechanisms for food processing in lissamphibians and these mechanisms appeared to differ from that of T. carnifex. Food processing in T. carnifex is different because of the involvement of rhythmic head and jaw movements in concert with cyclic tongue movements to reduce food intraorally. However, there are some superficial similarities with the processing mechanism of desmognathine salamanders (Schwenk and Wake, 1993). Both taxa use rhythmic ‘head bobbing’ in concert with gape cycles but the mechanisms also involve obvious differences: in Desmognathus, the head is elevated during gape opening, followed by rapid depression of the skull and gape closure. Skull depression places the massive, pulley-like atlantomandibular ligaments (connections between the cervical vertebra and lower jaw) under tension to transmit force from head flexion to assist the jaw adductor muscles with gape closure and amplify bite force (Dalrymple et al., 1985; Deban and Richardson, 2017; Schwenk and Wake, 1993). Our anatomical observations reveal that T. carnifex lacks the atlantomandibular ligament. Head bobbing in Desmognathus applies strong bites to food contained between the mandibular arch elements (Deban and Richardson, 2017). By contrast, our data suggest that T. carnifex processes food by rasping it against the palatal dentition with its tongue and not between mandibular arch elements, resulting in a fundamentally different mechanism for food processing.

How does the mechanism of food processing in T. carnifex compare with processing mechanisms across gnathostomes? To address this question, we focus on two aspects: (i) coordination between tongue and gape cycles, because tongue movements are traditionally associated with – and interpreted relative to – jaw movements in gnathostomes and (ii) use of similar mechanical systems where the tongue rasps against rough palatal surfaces. The hyobranchial system in fishes and the hyolingual system in tetrapods (tongue) are here considered homologous structures (Reilly and Lauder, 1988) and to simplify the interpretations, both hyobranchial and hyolingual systems are henceforth simply referred to as ‘tongue’.

The most commonly occurring mechanism for food processing across gnathostomes is grinding or puncturing of food between the occlusal surfaces of mandibular arch dentition via a repetitive series of bites, also known as chewing. These rhythmic bites serve to crush, grind or puncture food whereas a carefully coordinated and cyclic movement loop of the tongue system serves to reposition food items in between chew cycles. This pattern is recognizable for most gnathostomes but the mechanisms for repositioning food differ between aquatic and terrestrial forms. Aquatic gnathostomes such as sharks, rays, bony fishes and lungfishes use the action of their tongue system to generate water flow to move and reposition food (Bemis and Lauder, 1986; Dean et al., 2005; Gintof et al., 2010; Lauder, 1985). Among fishes that chew, the tongue system moves in the caudal and ventral direction during gape opening so food is transported inwards and repositioned as the gape is closed (Konow and Sanford, 2008; Reilly and Lauder, 1990). Furthermore, in lungfishes (Bemis and Lauder, 1986) and probably also in some ray-finned fishes (Lauder, 1981; Van Wassenbergh et al., 2016) the tongue system can alternatively be elevated and protracted during gape opening to induce a posterior–anteriorly directed flow if food items have to be transported from back to front. Hence, coordination of jaw and tongue-system cycles appears to depend on functional requirements. In T. carnifex, the coordination of tongue and gape cycles shows some overlap with that of bony fishes and lungfishes but also some differences. The newt tongue initially remains stationary and is then depressed during gape opening and is elevated during most of gape closure. Furthermore, the tongue is partly protracted and retracted during both gape opening and gape closing. In other words, the relationship between vertical tongue movements and gape movements is overall similar to that of fishes in the first phase of the gape cycle, whereas horizontal tongue movements show a phase shift in relation to the gape cycle, when compared with fishes (see also Figs 2 and 3).

In some ray-finned fishes, a further processing behaviour that superficially resembles the tongue–palate rasping in newts is raking. In raking, food is stabilized between the mandibular jaws while the skull is elevated and the tongue apparatus forcefully retracted, causing food shredding by dentition on the mouth roof and tongue tip (basihyal) (Camp et al., 2009; Konow and Sanford, 2008; Sanford and Lauder, 1989, 1990). Raking in ray-finned fishes and tongue–palate rasping in newts appear similar but the coordination of tongue, jaw and skull movements differs. In raking, the tongue loops in the reverse, clockwise direction (rostrum facing left) and although the tongue shreds the food against palatal dentition in both systems, the power stroke in raking involves tongue retraction versus tongue protraction in the newt system. However, some groups of spiny-rayed (acanthomorph) fishes, including anabantoids (Konow et al., 2013; Liem and Greenwood, 1981), may have evolved a raking analogue with the same tongue motion pattern as seen in the newt.

In amniotes, cyclic movements of the tongue system during a bout of chewing first help position the food between the teeth and later help gradually transport food towards the oesophagus in preparation for swallowing (Bramble and Wake, 1985; Herrel et al., 2008; Herrel and De Vree, 1999; Hiiemae and Crompton, 1985; Smith, 1984). Although feeding mechanics differ substantially between amniote groups, a general pattern has been described using a simplified model (Hiiemae and Crompton, 1985; Bramble and Wake, 1985; Reilly and Lauder, 1990): during most of the mouth-opening phase, the tongue is first protracted and elevated and then partly retracted and depressed. During mouth closure, the tongue is mostly retracted and depressed. With the snout oriented to the left, the tongue describes a counter-clockwise loop. In newts, the tongue also loops in a counter-clockwise direction, yet the coordination of this loop with the gape cycle differs from the amniote pattern. During mouth opening, the newt tongue is first stationary and is then depressed (versus elevated and then partly depressed in the amniote model), while during mouth closing, the newt tongue is mostly elevated (versus depressed in the amniote model). As the gape is opened, the newt first protracts and then retracts its tongue (similar pattern to that in amniotes) and during gape closure, the tongue first continues to retract and then starts protracting (versus mostly retraction in amniotes). Compared with amniotes, the newt therefore shows a phase shift of vertical and partly horizontal tongue movements relative to the gape cycle, yet the direction of the loop is the same. The phase shifts might be due to different mechanisms underlying food processing: amniotes use their tongue to move food into jaw occlusion, which leads to a requirement for tightly coordinated tongue and jaw movements (Alfaro and Herrel, 2001; Hiiemae and Crompton, 1985; Lund, 1991). In contrast, the tongue–palate rasping mechanism in newts appears to require tight coordination of tongue and head movements, more so than tongue and jaw movements. Still, the ubiquitous counter-clockwise loop motion of the tongue suggests that amniotes also use their tongue to move food over the palate (Hiiemae, 2000; Palmer et al., 1997; Reilly et al., 2001; Schwenk and Rubega, 2005). Might this tongue–palate interaction play a role in food reduction in amniotes too? In many groups, mechanical reduction of food might be a minor element of the tongue sliding forward during the chewing cycle but in certain groups like sea cows, the rough keratinized palate is significantly involved in food processing (Werth, 2000). In the echidna and platypus, which are basal, edentulous mammals, tongue–palate interaction replaces tooth function and the tongue is adorned with keratinous spines that interact with similar palatal structures to grind food into a viscous slurry (Doran and Baggett, 1972; Griffiths, 1978; Schwenk and Rubega, 2005). Tongue–palate rasping mechanisms in amniotes remain poorly understood and the kinematics are relatively unstudied but we predict that the movement patterns are similar to those involved in food processing in the newt.

Despite the apparent differences outlined above, intraoral food-processing systems where the tongue rasps food against rough palatal structures are found in different gnathostome groups, opening up the question of whether the underlying motor patterns might have evolved convergently or have a common neuromechanics ancestry. Salamanders may be promising models to begin testing associated motor control hypotheses on the evolution of processing mechanisms in tetrapods because (i) lissamphibians are the only extant anamniote tetrapods that might have retained many ancestral tetrapod features and (ii) contrary to earlier assumptions (for reviews, see Deban and Wake, 2000; Schwenk and Rubega, 2005), salamanders do use rhythmic processing involving tongue, head and jaw systems.

The detection of tongue–palate rasping systems is complicated by movements of the tongue system being only partially visible to the eye, because they occur deep within the oral cavity. This makes X-ray, or alternatively invasive measurement techniques like sonomicrometry, necessary for analysis. A clear distinction between intraoral transport (sensu Gillis and Lauder, 1994; Reilly, 1996; Reilly and Lauder, 1990, 1991) and tongue–palate rasping might not always be possible because the transition from one to the other function is often continuous. For example, cyclic intraoral behaviours in lissamphibians have been suggested to only involve food transport. However, only a few lissamphibian taxa have been studied so far and we think it is highly likely that the electromyographic and light videography approaches used by earlier authors (Gillis and Lauder, 1994; Reilly, 1996; Reilly and Lauder, 1990, 1991) might have caused food-processing behaviours to go unnoticed. Data from lissamphibians are critical for unravelling and reconstructing the evolution of food-processing systems because, aside from being the only extant anamniote tetrapod clade that might have retained many ancestral tetrapod features, salamanders also permit studies of food processing across aquatic and terrestrial environments given the many semi-terrestrial species. Lissamphibians also allow observations of processing changes across the transformation of a hyobranchial to a hyolingual system as they metamorphose from a gill-bearing larva to a tongue-bearing post-metamorphic animal. Taken together, lissamphibians have many traits that make them suitable analogues of early tetrapods that similarly had to undergo structural and functional changes of their oropharyngeal system during terrestrial evolution.

We thank Christian Proy for his help in collecting newts, Simon Baeckens for performing µCT scans, Sam Van Wassenbergh and Krijn Michel for their help during preliminary X-ray experiments in Antwerp and Laura Porro, Rainer Schoch and Martin Fischer for scientific input along with two anonymous reviewers for their constructive criticism.