Feeding habits of the Middle Triassic pseudosuchian Batrachotomus kupferzellensis from Germany and palaeoecological implications for archosaurs

Bite traces on fossil bones are key to deciphering feeding ecology and trophic interactions of vertebrate past ecosystems. However, similarities between traces produced by different carnivorous taxa with similar dentitions, and misidentifications due to equifinality, hinder confident identifications of the bite makers. Here, we correlate bite traces with macroscopic wear and microanatomy of the teeth of the pseudosuchian archosaur Batrachotomus kupferzellensis from the Triassic Lower Keuper fossil lagerstätten (southern Germany), untangling its feeding habits and shedding light on the bite traces generated by ziphodont teeth (teeth with serrated carinae). Individually, bite traces reflect tooth morphology, whereas composite bite traces and their frequency are related to feeding behaviour and explain tooth macroscopic wear and microanatomy. Therefore the identification of the bite maker is possible by analysing composite bite traces, their location on bones, and their relative abundance. In addition, tooth macroscopic wear and microanatomy are proven as independent lines of evidence of feeding ecology. Comparing bite traces on fossil and present‐day bone assemblages, we observe that bone modifications by the crocodylomorph lineage (from Triassic pseudosuchian archosaurs to extinct and extant crocodylians) are strikingly similar, including taxa with and without ziphodont teeth. Such a set of features differs from bone modification assemblages produced by taxa with similar ziphodont teeth outside the pseudosuchian lineage, such as theropod dinosaurs and the Komodo monitor, suggesting phylogeny is a better predictor of feeding ecology among saurian reptiles than tooth morphology.

Abstract: Bite traces on fossil bones are key to deciphering feeding ecology and trophic interactions of vertebrate past ecosystems. However, similarities between traces produced by different carnivorous taxa with similar dentitions, and misidentifications due to equifinality, hinder confident identifications of the bite makers. Here, we correlate bite traces with macroscopic wear and microanatomy of the teeth of the pseudosuchian archosaur Batrachotomus kupferzellensis from the Triassic Lower Keuper fossil lagerst€ atten (southern Germany), untangling its feeding habits and shedding light on the bite traces generated by ziphodont teeth (teeth with serrated carinae). Individually, bite traces reflect tooth morphology, whereas composite bite traces and their frequency are related to feeding behaviour and explain tooth macroscopic wear and microanatomy. Therefore the identification of the bite maker is possible by analysing composite bite traces, their location on bones, and their relative abundance. In addition, tooth macroscopic wear and microanatomy are proven as independent lines of evidence of feeding ecology. Comparing bite traces on fossil and present-day bone assemblages, we observe that bone modifications by the crocodylomorph lineage (from Triassic pseudosuchian archosaurs to extinct and extant crocodylians) are strikingly similar, including taxa with and without ziphodont teeth. Such a set of features differs from bone modification assemblages produced by taxa with similar ziphodont teeth outside the pseudosuchian lineage, such as theropod dinosaurs and the Komodo monitor, suggesting phylogeny is a better predictor of feeding ecology among saurian reptiles than tooth morphology. D I R E C T evidence of feeding ecology and trophic interactions of fossil vertebrates mostly relies on the identification of bite traces on bones (Fiorillo 1991;Jacobsen 1998;Pobiner 2008;Drumheller et al. 2020), stomach contents and processed food in the form of coprolites and regurgitalites (Qvarnstr€ om et al. 2019;Gordon et al. 2020). Regarding bite traces, however, the identification of the producer (the bite maker) commonly remains impossible (Hone & Chure 2018) and can usually only be confidently identified at a major clade level (Mikul a s et al. 2006;Jacobsen & Bromley 2009;Pirrone et al. 2014;Augustin et al. 2020;Drumheller et al. 2020;Drymala et al. 2021). In addition, even if carcass utilization by carnivores is recognized, feeding habits may still remain uncertain (Hone & Rauhut 2010;Drumheller et al. 2014Drumheller et al. , 2020. This is particularly the case for ziphodont teeth (teeth with serrated carinae; Brink et al. 2015) because this dentition evolved convergently in different amniote clades (Whitney et al. 2020) but produces similar bite trace morphotypes (D'Amore & Blumenschine 2009). Taxa with ziphodont teeth include archosauromorphs (mostly purported terrestrial carnivores, but also phytosaurs and some marine crocodylomorphs), Varanus komodoensis (the extant Komodo monitor) and gorgonopsian therapsids (D'Amore & Blumenschine 2009Pirrone et al. 2014;Brink et al. 2015;Hone & Chure 2018;Whitney et al. 2020). Interpretations of feeding ecology and trophic interactions based on bite traces of extinct taxa with ziphodont teeth are often limited to non-avialan theropod dinosaurs (Fiorillo 1991;Jacobsen 1998;Jacobsen & Bromley 2009;Drumheller et al. 2020), with additional information obtained from bite traces produced by V. komodoensis (D'Amore & Blumenschine 2009. Here, based on the tetrapod collection from the Lower Keuper (Erfurt Fm., Ladinian, Middle Triassic) Kupferzell and Vellberg-Eschenau fossil lagerst€ atten, south-western Germany (Urlichs 1982;Schoch & Seegis 2016;Schoch et al. 2018), we reconstruct an exceptional palaeoecological scenario by combining multiple morphological and statistical analyses. We correlate the diverse bite trace assemblage with the macroscopic wear and microanatomy of the teeth of the up to five-metre-long pseudosuchian archosaur Batrachotomus kupferzellensis (Gower 1999;Gower & Schoch 2009), the bite maker of most of the traces studied herein. Since bite traces are confidently correlated to tooth macroscopic wear and therefore to microanatomical adaptations in tooth structure, our work shows that these can be independent lines of evidence to decipher feeding behaviours. With the aim of deciphering the feeding ecology behind bite traces produced by ziphodont teeth through morphological and statistical analyses, we show a linkage between bite makers' habits and phylogeny among archosaurs and, more broadly, saurian reptiles.

GEOLOGICAL SETTING
The fossil tetrapod bones studied in this work come from two different fossil lagerst€ atten, named Kupferzell and Vellberg-Eschenau (Fig. 1). These rich fossiliferous deposits correspond to the Untere Graue Mergel (UGM; lower Grey Marls), a unit within the upper part of the Lower Keuper facies from south-western Germany (Lettenkeuper, Erfurt Formation), of Ladinian (late Middle Triassic) age (Urlichs 1982; Schoch & Seegis 2016) ( Fig. 1A). In south-western Germany, the Lower Keuper mainly consists of an alternating carbonate-siliciclastic succession 20-25 m thick deposited in an epicontinental platform overlying the marine carbonate Muschelkalk facies (Brunner & Bruder 1981). This system is part of the Central European Basin, and during the Lower Keuper deposition it was occasionally flooded by marine transgressions from the Tethys sea, whereas the main sedimentary inputs came from the northern palaeogeographic highs (Etzold & Schweizer 2005;Nitsch 2015aNitsch , 2015b. The UGM is an overall succession of mudstones and marlstones, with different layers and bonebeds wellcharacterized at both the Kupferzell and Vellberg-Eschenau localities (Urlichs 1982;Nitsch 2015c;Hagdorn et al. 2015;Schoch & Seegis 2016). This unit is embedded between two dolostone banks, which may also contain fossils (Hagdorn et al. 2015;Schoch & Seegis 2016;Mujal & Schoch 2020).
The two localities are stratigraphically equivalent, though some lithological differences exist (Fig. 1B). The UGM layers in Kupferzell mostly present green, yellow and brown colours, whereas those from Vellberg-Eschenau are mainly grey-greenish and occasionally brown. Nevertheless, the palaeoenvironments for both localities are similar, being interpreted as lacustrine systems with marine influence and/or connection to marine environments, as evidenced by the presence of stratigraphically close marine carbonate units, and the presence of marine fossils, such as bivalves and sauropterygian reptiles within the UGM (Urlichs 1982;Nitsch 2015c;Hagdorn et al. 2015;Schoch & Seegis 2016;Mujal & Schoch 2020;DS, EM, RRS, pers. obs.) MATERIAL AND METHOD Institutional abbreviations. MHI, Muschelkalkmuseum Hagdorn, Ingelfingen, Germany; SMNS, Staatliches Museum f€ ur Naturkunde Stuttgart, Stuttgart, Germany.

Fossil sampling and analyses
All prepared fossil material from the Kupferzell and Vellberg-Eschenau fossil lagerst€ atten (south-western Germany) housed in the SMNS and MHI collections was examined using a magnifying glass, a dissecting microscope, and lighting in different orientations, to identify bite traces (e.g. Blumenschine et al. 1996;D'Amore & Blumenschine 2009;Drumheller & Brochu 2014;Drumheller et al. 2020). Bite traces were systematically classified according to their morphology (ichnotaxonomy) following Mikul a s et al. (2006), Jacobsen & Bromley (2009) and Pirrone et al. (2014). In addition, equivalences of the identified ichnotaxa with the taphonomic terms coined by Binford (1981) and extensively used in the literature (e.g. Pobiner 2008;D'Amore & Blumenschine 2009;Njau & Gilbert 2016;Drumheller et al. 2020) are provided. Associations between bite traces and their orientation and location on bones were also surveyed. All of this information can be found in Data S1 (and Mujal et al. 2022). In taphonomic studies, bite/tooth traces are often named bite/tooth 'marks' but, since they are biogenic structures (i.e. ichnofossils or trace fossils), the term 'trace' is preferred, with use of the term 'mark' restricted to nonbiogenic structures (Bertling et al. 2006;Mikul a s et al. 2006;Jacobsen & Bromley 2009;Vallon et al. 2015).
Bite traces produced by the dragging of denticles on the carinae of teeth (ichnotaxon Knethichnus parallelum) were measured on digital photos using ImageJ (v.1.52a) following D'Amore & Blumenschine (2012). Such measurements account for the maximum number of denticles per millimetre, allowing for the identification of the potential bite maker through comparisons with the tooth assemblage from the Lower Keuper (Schoch et al. 2018).
Since most of the bite traces were assigned to Batrachotomus kupferzellensis, teeth of this pseudosuchian archosaur were analysed as explained below.
Tooth crowns of Batrachotomus from Kupferzell (N = 258) and Vellberg-Eschenau (N = 56), all from the UGM, were analysed accounting for: (1) basic morphological classification by measuring (with a digital calliper) the apicobasal height of the crown (or crown height, CH), the mesiodistal depth and the labiolingual width of the crown at its base (or crown base length, CBL, and crown base width, CBW, respectively; for tooth parameters, see: Smith et al. 2005;Hendrickx et al. 2015); (2) identification of the position within the jaw, distinguishing premaxillary from non-premaxillary teeth (also distinguishing teeth from the dentaries and maxillae, when possible). In order to identify the position of teeth within the jaws, we examined the dentigerous elements of Batrachotomus still preserving attached teeth and also followed tooth descriptions from Gower (1999) and Schoch et al. (2018). Generally, premaxillary teeth display distinctive features allowing them to be distinguished from those of the maxillae and mandibles. The premaxillae of Batrachotomus held four teeth each, these with a characteristic slender shape (Gower 1999). Teeth from the premaxilla were almost cylindrical, with a nearly round or slightly oval cross-section, and an apical portion which is slightly (linguo-)distally curved; their basal mesial carina often lacks denticles (Data S2). Non-premaxillary teeth generally show diverse crown morphologies (Gower 1999;Schoch et al. 2018), with differences in: (1) the distal curvature (symmetry in labial/lingual view); (2) the symmetry of cross-section (more or less convexity of the labial side, though all are oval in cross-section); (3) curvature towards the lingual side; and (4) relative height of the crown with respect to the mesiodistal depth of the crown base. Both mesial and distal carinae possess denticles. However, due to the isolated nature of most of the teeth, in most cases it was not possible to identify whether the non-premaxillary teeth were from the maxillae (with 11 alveoli each) or the mandibles (with probably 12 alveoli each) or from which tooth position. Only teeth still attached to maxillae or mandibles, as well as some teeth with a recurved labial side of the crown and straight posterior edge that are characteristically from the anterior part of the maxilla (Gower 1999), can be more confidently positioned. According to Gower (1999, p. 38), teeth from the posterior part of the maxilla are more strongly convex anteriorly and have shorter crowns than those from the anterior part; however, these features seem to be present in some mandibular teeth. In addition, Schoch et al. (2018, p. 621) noted differences in tooth length across maxillae and dentaries, with the longest elements in the mid-portion of maxillae and anterior portion of dentaries and the smallest teeth in the anterior and posterior ends of maxillae. Nevertheless, the isolated nature of most of the teeth precludes their association to individuals, meaning that they could belong to either juvenile or mature/adult animals because teeth could be proportionally smaller or larger, respectively. Of note, all preserved dentigerous elements show that tooth replacement in Batrachotomus occurred at alternating tooth positions (Gower 1999).
Tooth macroscopic wear (macrowear) was classified in two ways (Table 1): qualitatively, distinguishing wear, notches, spalling and breakage, as well as noting the areas of the tooth crown with greater amounts of wear (lingual or labial, and mesial or distal); and (semi-)quantitatively, scoring the degree of tooth wear as none (0), low (1), high (2) or extreme (3). All tooth measurements and macrowear are integrated in Data S2 (see also Mujal et al. 2022). Based on the UGM tooth sample, we compared macrowear between premaxillary and non-premaxillary teeth in a box-plot by averaging the amount of macrowear (from none to extreme) of all teeth. In addition, we compared regional differences in macrowear (using averages) along the mesial and distal carinae and the tip in a box-plot. The number of occurrences (counts) of abrasion (wear and notches) for the tips and each portion of both mesial and distal carinae were counted and displayed as percentages for each region of the tooth crown, also differentiating between premaxillary and nonpremaxillary teeth.

Tooth histology
To complement the interpretations based on the macroscopic wear of Batrachotomus teeth (see Discussion, below), we analysed their microanatomy, also comparing them to other taxa with ziphodont teeth (Brink et al. 2015;Wang et al. 2015;Whitney et al. 2020). Histological thin sections were prepared from isolated Batrachotomus teeth originating from the UGM and showing different degrees of macroscopic tooth wear. Four teeth (premaxillary: SMNS 97009/8; non-premaxillary: SMNS 97009/9, 97009/10, 97009/16), with one thin section each, were prepared in longitudinal section capturing either the mesial or distal carina. The crowns of two nonpremaxillary teeth (SMNS 97009/11,97009/17), with five thin sections each, were prepared in cross-section at varying points along the apicobasal axis. Thin sections were prepared following standard palaeohistological methodology, with thicknesses ranging from 60 to 75 µm, and examined under a petrographic microscope (Leica DM750P). Photographs were taken using an attached Leica ICC50 W camera, and processed using Leica LAS EZ (v.3.1.0) software.

Statistical analysis
To investigate potential statistical differences regarding bite trace patterns on different bitten taxa (Mastodonsaurus vs Plagiosuchus vs Nothosaurus vs Batrachotomus), body regions (skull vs teeth vs free vertebrae vs sacrum vs ribs vs pectoral girdle vs pelvic girdle vs limbs) and bone regions (proximal end vs distal end vs side (widest area) of the shaft vs edge (narrowest area) of the shaft vs bone edge vs bone centre), we converted the data collected in the bite trace database (Data S1) into three separate datasets, scoring the absence or presence of each bite trace morphotype for all bones within each group. A small fraction of bones were not included in the analyses (Data S1) because the elements could not be identified due to their fragmentary nature or preservation. These datasets were analysed using a permutational multivariate analysis of variance (PERMANOVA). This method estimates the potential overlap between two or more groups by testing the significance of their distribution on the basis of permutation (9999 iterations) and the Euclidean distance as distance measures. In contrast to parametric tests, PERMANOVA does not require normal distribution of the data (Anderson 2001;Hammer & Harper 2006). The spatial relationship between groups is expressed by an F-value and a Bonferroni-corrected p-value. Following Wills et al. (1994), the three datasets were further transformed into Euclidean distance matrices and subjected to a principal coordinate analysis (PCOA). This method reduces multivariate data down to a new set of independent variables (principal coordinates) that are linear combinations of the original set with zero covariance (Hammer & Harper 2006), providing a 'biting space'. Afterwards, we ran a broken-stick method for each PCOA to obtain the number of principal coordinates (PCOs) that contain the relevant amount of total variation (De Vita 1979;Jackson 1993). For these PCOs we applied a linear discriminant analysis (LDA), which reduces the number of PCOs to a smaller set of dimensions by maximizing the separation between the given groups using the Mahalanobis distance. This distance measure is estimated from the pooled within-group covariance matrix, resulting in a linear discriminant classifier and an estimated group assignment for each species. These results were cross-validated using Jackknife resampling (Hammer & Harper 2006;Hammer 2020). Based on the confusion matrix (Stehman 1997), we estimated the error of correct identification.
Finally, given that the microanatomy of Batrachotomus teeth is strikingly similar to that of other hypercarnivores with ziphodont teeth (see Microanatomy of Batrachotomus teeth, below), we statistically compared the overall tooth morphology or morphometry of Batrachotomus kupferzellensis with that of other archosauromorphs by combining our measurements with the data from Hoffman et al. (2015) on theropod teeth. Because both datasets include different sets of parameters, we restricted the comparison to the crown height (CH), crown base length (CBL), crown base width (CBW), number of denticles on the mesial midcrown (MC) and number of denticles on the distal midcrown (DC). The latter two parameters were measured for a subset of 23 teeth from the Batrachotomus kupferzellensis sample (Data S2). In addition, we created three ratios between CBL and CBW (CBR), CH and CBL (CHR), and MC and DC (MDCR). Specimens with incomplete measurements were excluded from the final dataset. All measurements and ratios were log-transformed and analysed with LDA (including Jackknife resampling) and PERMANOVA (with 9999 iterations and Bonferroni-correction for p-values). In addition, we ran the LDA with all Batrachotomus teeth grouped as 'unknown' and classified them based on the training set, including the remaining specimens from the two original datasets. All statistical analyses and ordination methods were performed with the program PAST v.4.03 (Hammer et al. 2001).

Description of bite trace morphotypes on Lower Keuper bones
The presence of bite traces on Lower Keuper bones has previously been mentioned (Wild 1978(Wild , 1979(Wild , 1980Schoch & Seegis 2016)  Nihilichnus nihilicus, round/oval punctures produced by tooth tips; Linichnus serratus, linear grooves with serrated margins produced by serrated carinae; Knethichnus parallelum, regular and parallel grooves produced by denticles of carinae; Brutalichnus (two morphotypes), large holes reaching cancellous bone indicating strong bites; and Machichnus-like, multiple small punctures connected with short grooves produced by relatively small teeth. The 'dental hacks' described by Jacobsen & Bromley (2009) may also correspond to short Linichnus, and are likely to be equivalent to 'edge marks'. Nomenclatural differences rely on the extension of the bite trace (from a few mm to up to 2 cm long) and the incision into the bone (i.e. just deforming or completely removing the cortical bone), as occurs with Nihilichnus; thus, even if all of these linear grooves correspond to the same ichnotaxon the markedly shallower traces, although classified as Linichnus, are termed 'scores' (as in D'Amore & Blumenschine 2009). The traces of L. serratus within the Lower Keuper assemblage range from short (2-3 mm; 'simple dental hacks' of Jacobsen & Bromley 2009) to relatively long (c. 1 cm, and up to 2 cm) grooves that are usually straight to slightly curved; they can display short striations parallel to the main groove. This trace is formed by the serrated cutting edge of the tooth crown (i.e. from teeth possessing denticles on the carinae) and/ or the dragging of the ( (2016), even though their tooth carinae do not possess denticles); also, a single trace can be rectilinear in one part and become strongly curved in the other part (as do some 'striations with internal striae' produced by crocodylians; Njau & Gilbert 2016). They are in groups of 3 to up to 13 parallel striations, though this number can vary also in a single trace, reflecting the changing number of denticles in contact with the bone as a result of the curvature of the tooth carinae and/or the bone. This morphotype results from the transverse to oblique movement of the tooth along the bone (D'Amore & Blumenschine 2012), thus, the movement is perpendicular to oblique with respect to the longitudinal axis of the tooth crown (i.e. the movement is mostly perpendicular to that producing Linichnus serratus).
Ichnotaxon Brutalichnus Mikul a s et al., 2006. Traces corresponding to relatively large holes due to the removal of portions of both cortical and cancellous bone. This ichnogenus is recorded by two clearly differentiated morphotypes, here named morphotypes 1 and 2. The first one consists of relatively large and deep round to oval-shaped holes removing the cortical bone entirely and reaching the spongiosa (Fig. 2I). The second morphotype is represented by deep and elongated traces (also entirely removing the cortical bone) widening towards the edge of the bone, outlining a triangularor V-shaped trace (Fig. 2J). The margins are roughly occasionally result in a T-shaped trace. Some grooves display serrated margins, indicating that they were produced by (small) serrated teeth. In its diagnosis, this ichnotaxon was interpreted as a result of gnawing (Mikul a s et al. 2006) but, considering the specimens reported herein, it would more broadly represent repeated strokes on the bone. Interestingly, a 'gnawing-like' behaviour has recently been reported in a small-bodied theropod dinosaur (Brown et al. 2021); the bite traces reported herein (also associated with a small-bodied tetrapod, probably an archosauriform; see Identification of the bite maker, below) may represent a similar behaviour. In any case, considering the potential high disparity of behaviours (and wide range of potential bite makers), the bite traces of the Lower Keuper with this morphology are referred to as Machichnus-like, as the identification of this ichnotaxon is tentative.

Associations of bite traces from the Lower Keuper
All the previously described biting ichnotaxa are associated, with traces from the same and from different morphotypes, and show gradual transitions between types or are found in association with each other (being produced by the same taxon and/or individual; The most common association is a transition from Linichnus to Knethichnus, the long axes of each trace morphotype being perpendicular to oblique from each other (Fig. 2B, E). Transitions between Linichnus and one oblique to nearly perpendicular Knethichnus between them, outlining an N-or Z-shaped composite trace (Figs 2B, 3E, 5A, C, P). Occasionally, this association is the opposite, with Linichnus in-between Knethichnus traces (Fig. 5D, S).
Nihilichnus is most commonly associated with Linichnus (Data S1), although there are also Nihilichnus-Knethichnus associations (e.g. Fig. 5B, P, Q, S, T); in both cases, there is usually a relatively sharp transition between both trace morphotypes, though clearly produced by the same tooth. A peculiar association, found in different bones, is that of two gently curved/arched traces of Knethichnus outlining a fan or an open triangle (e.g. SMNS 81109, 97012); in some cases (e.g. SMNS 81210) this association consists of a Knethichnus trace in each outer end that changes to Linichnus in the middle of the composite trace, where the two traces join, or eventually the tooth only generates Linichnus (SMNS 80923) (Figs 2H, 5K, I, L; see below). Some (usually long) Knethichnus traces show sharp changes in orientation of the grooves (although direction of the grooves changes a few degrees), occasionally with a Linichnus and/or Nihilichnus trace in between (Fig. 5D, S), resembling the pivoting traces produced by crocodylians (Njau & Gilbert 2016).
Both Brutalichnus morphotypes are very commonly in association with Nihilichnus, but in some cases they are also in transition from Knethichnus and Linichnus (Figs 2I, J, 3B, D, 4G, 5T). Brutalichnus are mostly found on the end of long bones and bone edges (Figs 3-5; Data S1); when found on shafts, it is usually because the bone was broken from that part, most likely due to the bite that generated the Brutalichnus trace (e.g. Fig. 5G).
Several bones also display areas with dense clusters of diverse bite traces, in parallel or in multiple directions, and mostly including Nihilichnus and rectilinear to curved Linichnus and Knethichnus ( Several bones display clusters of (usually abundant) parallel to nearly parallel traces of Linichnus, with some associated Knethichnus as well, when they build up the aforementioned N-or Z-shaped composite traces (Figs 2B, 3A, E, 5C, D, P). Also, some bones, especially on the edge area of the shafts, preserve two or more traces of Linichnus in a V-shape (Figs 2C, D, 5A, E, F, I); sometimes they seem to form a bifurcated trace, corresponding to a single trace with an abrupt change of orientation with respect to the bone, as Some bones display surfaces completely covered by long traces of Knethichnus, with the different traces being roughly oriented in the same direction, and with only slight changes to the angle of the grooves among traces (Fig. 5P). Otherwise, relatively long but thin Knethichnus traces (i.e. a few denticles contacted the bone) are also relatively common (Figs 2G-I, 3B, E-F, 5A-B, Q-S). These traces are mostly rectilinear, sometimes with slight, though abrupt, changes of orientation. At the change of orientation point, Linichnus and/or Nihilichnus may appear (Fig. 5D, S).
On the edge of shafts (but also on some bone ends), deep traces are frequently observed. They are removed flakes of (mostly cortical) bone with Knethichnus extending obliquely within the entire pit and Linichnus often in one of the short margins of the removed flake ( Several bones also display bite traces on opposing surfaces; these are usually Nihilichnus against Nihilichnus, but also Nihilichnus against Brutalichnus, and Nihilichnus against Linichnus and/or Knethichnus (Figs 2A, 3-5). Notably, aligned sets of Nihilichnus also occur, outlining an arch (Figs 3D, 4A, H) representing a partial tooth row. These traces are reminiscent of the 'serial bite marks' (in the sense of Binford 1981) generated during a single biting event with multiple teeth impacting on the F I G . 7 . PCOA and LDA results from the bite traces. Biting spaces are compared for each bitten taxon (top), each skeletal region (limbs, pectoral girdle, pelvic girdle, ribs, sacrum, vertebrae, teeth and skull) (middle), and each bone region (proximal and distal ends, side and edge of the shaft, bone edge, and bone centre) (bottom) for the first two ordinates. Percentage of total variation for each ordinate is given in parentheses. Biplots visualizing the factor loadings of the LDA are shown next to the corresponding plot.
bone. Occasionally, such aligned traces are parallel grooves (i.e. Linichnus; Fig. 4J-K) indicating that the tooth row moved along the bone, and thus with teeth moving parallel to one another in a single bite.
Machichnus-like traces (Figs 2K, 6D-F) are the only ones not associated with any other biting ichnotaxon. Instead, they are only associated with (abundant) traces of the same morphotype. They are found as clusters of bite traces, being somewhat aligned with the long axis of each trace perpendicular to the major axis of the whole cluster. Therefore, they represent single bites or multiple, repeated strokes, or most likely both.

Distribution of bite traces along the Lower Keuper bones
As aforementioned, bite traces have been identified in five different tetrapod taxa. With the exception of a rib probably corresponding to a chroniosuchid reptile (Fig. 2H), all bones that were anatomically identified were included in the statistical analyses. Statistically, Mastodonsaurus bones show the largest anatomical distribution of bite trace types in PCOA 'biting space', followed by Batrachotomus and Plagiosuchus. All taxonspecific biting spaces show strong overlap (see LDA and PERMANOVA results: Fig. 7; Table 2; Tables S1-S3). The same associations or composite bite traces are found in different bone types and bone regions (Figs 3-5), suggesting that the bones were bitten in similar ways independently from their shape and body region (Figs 7, 8).
With the exception of the limbs and vertebrae, taxa show a similar distribution of bite traces across anatomical regions, and the location and identity of the bite traces on individual bones depend on the skeletal region (Figs 7, 8). Thus, all bitten taxa were similarly processed by the bite maker.
The distribution of bite traces along each type of bone (limb, rib, pectoral girdle, sacrum, pelvis, vertebral column, skull and teeth) and bone region (for elongated bones, proximal and distal end, and side and edge of the shaft; for all other bones, bone edge and bone centre) shows remarkable differences (Figs 7, 8; Table 2;  Tables S4-S9). The results of the PCOA and LDA (Error: 56%; Error-JK: 65.4%) indicate that bite trace patterns on bones from different body regions strongly overlap with each other, except for the limb bones. The PERMANOVA results (F = 2.894; p < 0.001) indicate no significant overlap of limbs and vertebrae with other bone regions. Nevertheless, the results show that bones from most body regions were processed by the bite maker in a similar way, as all ichnotaxona are observed on each bone type, skeletal region, and bone region of both Mastodonsaurus and Batrachotomus (Fig. 8). Each bone region shows a strong overlap in the biting space along PCO1 (Fig. 7). On the other hand, proximal ends, distal ends, side of shafts and edge of shafts of long bones are clearly separated from each other along PCO2. The PERMANOVA analysis also indicates a common biting pattern in different bone regions, which is also evidenced by the presence of equivalent composite bite traces in different bones and bone regions (see Associations of bite traces from the Lower Keuper, above; Figs 3-5; Data S1). Only the bone edge and bone centre were found to be significantly different from the other regions (Fig. 7), although this may result from the different shape of the corresponding bones (those that are non-elongated, such as vertebrae and Mastodonsaurus ischia and scapulacoracoids). However, in the LDA (Error: 43.75%; Error-JK: 54.17%) bone ends and shafts are well separated from each other along axis 1. Furthermore, proximal and distal ends as well as sides and edges of shafts are well separated along axis 2. Both the bone edge and centre regions cluster close to the point of origin (Fig. 7).
All in all, the results indicate that different bone regions seem to be more specifically processed by the bite maker when compared with the skeletal regions, but still show common patterns of bite traces. Such differential distribution of bite traces is also observed in the occurrence of each ichnotaxon on each bone type, skeletal region and bone region of Mastodonsaurus and Batrachotomus, which encompass the bulk of bones with bite traces, most of which are ribs ( Fig. 8; Data S1). Note that these counts are based on presence/absence, not the absolute number of each ichnotaxon (i.e. a given ichnotaxon on a bone region is counted as one regardless of whether there is a single or several traces of this same ichnotaxon). In Mastodonsaurus, Nihilichnus (punctures produced by tooth tips) are markedly more abundant The spatial overlap of biting spaces between taxa, types of bones and bone regions is indicated (see also Tables S1-S9). Bold p-values of the PERMANOVA indicate significant differences between biting spaces. LDA results are shown as errors in comparison to random sampling. based on count data than the other ichnotaxa on both proximal and distal bone ends, whereas in Batrachotomus this only occurs on the distal end (Fig. 8). Long bone ends frequently show traces (mostly Nihilichnus, though Brutalichnus presence is also notable) on both sides, resulting from a single bite using teeth from the lower and upper jaws (Figs 2A, 3A, C-E; Data S1). The side (the widest) and edge (the narrowest) areas of the shaft show similar distributions on long bones of both Mastodonsaurus and Batrachotomus (Fig. 8). On shafts, Linichnus dominates, followed by Knethichnus (which is often associated with the former); notably, the N-or Z-shaped traces, V-shaped traces, and the deep flakes covered with Knethichnus are often on the edge of the shaft (Figs 2C-D, F, 3B, E, 4A-C, E, 5C-E, P). Also, in several shafts, Linichnus traces are perpendicular (with respect to the bone axis) on the edge area and continue in oblique orientation to the side area. Nihilichnus is also relatively common on shafts, especially on ribs of Mastodonsaurus (Fig. 8); in fact, the uncinate processes of Mastodonsaurus ribs often show Nihilichnus traces, which occasionally are in transition to Linichnus and/or Knethichnus (Fig. 5A, G, H; Data S1). Both morphotypes of Brutalichnus are more commonly found on the ends of long bones and bone edges, which is consistent with the destructive nature of the trace. Nihilichnus is the dominant ichnotaxon on non-elongated bones (flat or polygonal elements and vertebrae).

Identification of the bite maker
Since most of the observed ichnotaxa are associated and build up regular composite traces with similar distribution in different bones ( , with a density of 2.8-5.4 denticles/mm on the mesial and 3-4.8 denticles/mm on the distal carina (Data S2). In fact, in the Lower Keuper tetrapod assemblage there are no other known taxa with the same denticle size and density (Schoch et al. 2018). Considering the association of Knethichnus with the other ichnotaxa (see Distribution of bite traces along the Lower Keuper bones, above), the attribution of most of the identified bite traces to Batrachotomus is therefore unambiguous (Data S1). In addition, the deep Nihilichnus traces found on some bones (e.g. Fig. 2A) Fig. 4H) draw a semicircular line that fits with the outline of the premaxillae of a medium-sized Batrachotomus (e.g. Gower 1999); each bite trace corresponds to a tooth present in the premaxillae, which would be four (two in each side) according to the alternating replacement of teeth (i.e. the first and third teeth in one premaxilla, and the second and the fourth in the other) as documented by Gower (1999) and observed in the paired premaxillae of SMNS 80260 (Data S2). The only bite traces most likely not to have been produced by Batrachotomus are: (1) small Nihilichnus traces on a Mastodonsaurus tusk (SMNS 91634) that are aligned and outline a curved arch with little spacing between them (Data S1); (2) the traces found on the ventral surface of two palates (SMNS 80704, 81310) and one mandible (SMNS 92128) of Mastodonsaurus ( Fig. 6A-C); and (3) all the Machichnus-like traces (SMNS 81210, 97014, 97013/5) (Figs 2K, 6D-F), because they are not associated with any other morphotype. Traces on the palates and mandible of Mastodonsaurus were all likely to have been produced by a conspecific, especially the paired large Nihilichnus on the palate SMNS 80704 and the mandible SMNS 92128 (Fig. 6A-B), which fit well with the morphology and position of the paired tusks of Mastodonsaurus mandibles (Schoch & Seegis 2016). Considering the small spacing between clusters of Machichnus-like traces (probably produced in single bites and/or repeated strokes on the same area), as well as the notable size and morphological differences with the other ichnotaxa, it can be asserted that the Machichnus-like traces were produced by a different taxon that could not be identified. In any case, the bite maker was most probably a relatively smallsized diapsid reptile with serrated teeth (possibly an archosauriform; Schoch et al. 2018), given the slightly serrated margins of some grooves in clusters of Machichnus-like traces (Fig. 6E). Within the Lower Keuper tooth assemblage, several tooth morphotypes (mostly of archosauriforms) that could fit with such bite traces have been identified, though some have not been associated with any particular taxon yet (Schoch et al. 2018). However, although unlikely considering the inferred behaviour for the formation of bite traces, the possibility that these bite traces were produced by a juvenile specimen of Batrachotomus cannot be totally excluded.

Morphology and macroscopic wear on Batrachotomus teeth
Given the frequency of tooth-bone contact suggested for Batrachotomus, we assessed the type and distribution of abrasive macroscopic wear on 82 premaxillary and 232 maxillary and dentary teeth of this pseudosuchian archosaur. In fact, the surveyed tooth macroscopic wear (including wear, notches, spalling and breakage; Table 1; Fig. 9; Data S2) shows patterns that closely fit with the morphology, associations and distribution of bite traces, as further shown in the Discussion, below. Teeth from Kupferzell (N = 258) and Vellberg-Eschenau (N = 56) show similar macrowear patterns, suggesting no ecological differences between the two localities despite (slight) sedimentological differences (Urlichs 1982; Schoch & Seegis 2016; DS, EM, RRS, pers. obs.) (Fig. 1B).
Box-plots and counts of macrowear show slight but statistically significant differentiation of the macroscopic wear from premaxillary and non-premaxillary teeth (Figs 10, 11), which indicate different usages of each tooth type (see Usage of teeth: correlating tooth wear and bite traces, below). Premaxillary teeth display relatively more wear on the tips than on the carinae, also accounting for a higher degree of abrasion than the nonpremaxillary teeth. Wear generally decreases from the tip to the base of both premaxillary and non-premaxillary teeth (e.g. Fig. 9C, D). The tips of premaxillary teeth possess more frequently severe (extreme) wear than those of maxillae and dentaries. Wear on non-premaxillary teeth is more common and severe on the mesial carina than on the distal carina, whereas premaxillary teeth show relatively homogeneous wear over both carinae, though the basal mesial carina (where denticles are often lacking) is significantly less worn than the other regions. Notches show much more differentiation in distribution. On premaxillary teeth, notches are more abundant and severe on the middle and basal distal carina, and a very few occur on the basal mesial carina. On non-premaxillary teeth, notches occur more frequently on the mesial carina; nonetheless, the middle distal carina accounts for the highest occurrence of notches, with a marked difference relative to the other regions. Interestingly, the counts of wear and notches are generally proportionally inverse on both premaxillary and non-premaxillary teeth: wear increases from the basal to the apical regions of both carinae, while notches show an opposite trend (with the exceptions mentioned above; Fig. 11).
Breakage and spalling most commonly occur on the tooth tips. Usually, the surfaces of broken tips are smoothed or polished due to abrasive wear ( Fig. 9A-C, G-I), indicating that breakage occurred during the tooth lifetime and thus was most likely to have happened during feeding. Interestingly, premaxillary teeth often show higher degrees of spalling (running from the tip) on the labial than on the lingual side (Data S2), thus indicating that such damage did not result from tooth-tooth contact. Some teeth display severe breakage, with the entire carina completely broken, starting from the tip. Notably, in some cases, only the tip and the apical distal carina display breakage, with a sudden stop of the damage on the carina (Fig. 9E-F).

Microanatomy of Batrachotomus teeth
Longitudinal thin sections along the carina reveal that denticles in Batrachotomus consist of both enamel and dentine, and thus are ziphodont following the definition of Brink et al. (2015) (see also Wang et al. 2015;Whitney et al. 2020). The tooth crown consists of enamel and bulk dentine, separated by a zone of globular mantle dentine (Fig. 9N-T). Enamel is thickest on the surface of the denticles (c. 80 µm), but also towards the mesial and distal carinae (up to 60 µm) (Fig. 9N, Q-S). Enamel thickness also decreases on both the carinae and the sides of the tooth in an apicobasal direction. The enamel of each denticle is separated from that of the adjacent denticle by a channel continuous with the interdental fold ( Fig. 9N-O) (Brink et al. 2015). Within a denticle, enamel thickness thins toward the interdental fold (Fig. 9N, Q, R). Enamel spindles are rare overall, but occasionally present in the denticles (Fig. 9P). Dentine tubules originating within the fold curve pulpward towards the base of the denticle, becoming more densely distributed between folds (as with plicidentine). The base of the interdental fold consists of globular dentine, but globular mantle dentine is thin to absent in the denticles themselves ( Fig. 9N-P, T). Sclerotic dentine was difficult to recognize: there are fewer dentine tubules surrounding the interdental fold, and this region appears to mineralize more heavily between the least worn and a heavily worn tooth. However, wear on individual denticles (Fig. 9Q, R) does not appear to influence the development of sclerotic dentine, nor does sclerotic dentine develop on the tips of denticles worn down to the bulk dentine.

Comparison of Batrachotomus teeth with those of theropod dinosaurs and other archosauromorphs
According to our LDA analysis of dental measurements in different archosauromorphs, including non-avialan theropods (see Hendrickx et al. 2015;Hoffman et al. 2019), the teeth of Batrachotomus plot in between the area of non-spinosaurid Megalosauroidea and Spinosauridae, and are further surrounded by the tooth morphospaces of Allosauroidea, Ceratosauria, Tyrannosauroidea and Archosauromorpha morphotype A (Fig. 12). All these taxa represent apex predators with ziphodont teeth. If Spinosauridae are treated as an independent group, the LDA (Error: 43.05%; Error Resample : 46.03%) identifies 19 out of 23 Batrachotomus teeth correctly (  (Table 3). Treating Batrachotomus teeth as unknown, they are identified 16 times as Megalosauroidea (Table 3). Based on this grouping, the PERMANOVA found additional overlap with the morphospaces of Megalosauroidea and Silesauridae.

Usage of Batrachotomus teeth: correlating tooth wear and bite traces
Associated and composite bite traces, combined with a differential distribution of each bite trace morphotype in each skeletal and bone region ( Fig. 8; Data S1) and F I G . 1 0 . Box-plots of wear intensity on Batrachotomus teeth (1 = low; 2 = high; 3 = extreme). Above, intensity of the different types of macrowear. Below, intensity of macroscopic wear (including all four types) for each portion of the carinae, tip, and base of the crown. the macroscopic wear on Batrachotomus teeth, allow the movement of teeth on bones, and thus the most probable behaviours and usages of the carcasses, to be reconstructed.
As shown in 'Morphology and macroscopic wear on Batrachotomus teeth' above, tooth tips, especially from the premaxilla, showed the most wear (Figs 10, 11), correlating with the dominance of Nihilichnus traces, especially on the ends of ribs and limb bones (Fig. 8), and also on the uncinate processes of Mastodonsaurus ribs (Fig. 5A,  G, H). Such puncture traces are often aligned and found on opposing bone surfaces, outlining the tooth rows (Figs 3D-E, 4H). These features indicate a potential dismemberment of the carcasses in order to reach fleshy and nutrient-dense regions, as done by extinct and extant crocodylians (Njau & Blumenschine 2006;Njau & Gilbert 2016). Dismemberment and defleshing strokes would have been violent in some cases, as the Brutalichnus traces and the relatively high amount of breakage of tooth tips and carinae suggest. Brutalichnus is often associated with Nihilichnus, but also with Knethichnus and Linichnus (Figs 2-5; Data S1), indicating that there were pullbacks F I G . 1 1 . Percentage of abrasion (wear and notches) on Batrachotomus teeth. The degree of abrasion (none, low, high, extreme) is accounted. Numbers within colour bars represent counts (absolute occurrences) of each degree of abrasion in each tooth portion. Unknown values ('?' in Data S2) are excluded, thus the sum of values of each bar is lower than the total number of teeth (N) analysed.
F I G . 1 2 . LDA results of archosauromorph teeth. Plot comparing teeth measurements of Batrachotomus with that of different archosauromorphs, including various non-avialan theropod dinosaur groups. Spinosaurids are separated from Megalosauroidea; the dashed line shows the whole area occupied by megalosauroids, mostly overlapping that of Batrachotomus. Percentage of total variation for each ordinate is given in parentheses. The factor loadings of the LDA are visualized as a biplot. and/or repeated strokes associated with these more destructive bites.
The amount and distribution of wear and notches along the tooth carinae are consistent with the generation of most Linichnus, Knethichnus, and their associations (N/Z-, fan-and V-shaped composite traces, and clusters of Linichnus;  fig. 3B) and Linichnus, with tooth crowns perpendicular and oblique to the bone long axes, and resulted in the observed markedly higher number of notches on the distal carinae of premaxillary teeth and the middle distal carinae of non-premaxillary teeth (Fig. 11). The usual lack of denticles on the basal portion of the mesial carina of premaxillary teeth ( Fig. 9B; Data S2) probably responds to a low functionality of this region (D'Amore 2009), which is much less worn than the rest of the crown (Figs 10, 11). Thus, the observed pattern of macroscopic wear in the teeth of Batrachotomus provides complementary support for its identity as the bite maker.
In some fossil taxa, tooth-tooth contact (attritional wear) has been invoked to explain patterns of carinal wear and basal facets ( . We consider such explanations unlikely in Batrachotomous. The bite trace assemblage clearly supports extensive carina-bone contact as described above. Spalled teeth usually display damage on both labial and lingual sides (Fig. 9K, L); if such damage was generated due to occlusion, these teeth would have contacted the opposing teeth on both their lingual and labial sides, which is very unlikely or even biomechanically impossible. In fact, when the mouth of Batrachotomus was closed (see reconstruction by Gower & Schoch 2009), teeth from the upper jaw remain externally placed, thus only their lingual side could have contacted the labial side of teeth from the lower. Therefore not all spalling (if any) could result from tooth-tooth contact. Similarly, premaxillary teeth are usually more worn towards the labial side (Data S2), the side that would have never occluded with teeth from the mandible.

The role of microanatomy on the usage of Batrachotomus teeth
The microanatomy of Batrachotomus teeth (Fig. 9N-T) shows adaptations conferring resistance to carinal wear and breakage caused by tooth-bone contact seen in other hypercarnivores engaging in puncture and pull feeding (non-avialan theropod dinosaurs and gorgonopsian synapsids: D'Amore 2009; Brink et al. 2015;Wang et al. 2015;Whitney et al. 2020). The loss of globular mantle dentine at the tips of the denticles, combined with its presence at the enamel-dentine junction elsewhere along the tooth crown, has been interpreted as an adaptation for wear resistance in the denticles; in addition, elasticity elsewhere in the tooth has been inferred to confer resistance to breakage (Brink et al. 2015). The deep interdental folds would have helped to dissipate stress forces on the teeth (Brink et al. 2015;Gignac & Erickson 2017); globular mantle dentine as found at the bases of the folds has been linked to blocking crack propagation and increasing tissue elasticity (Wang et al. 2015). The sudden stop of breakage on the distal carina observed in Batrachotomus (Fig. 9E-F; Data S2) is consistent with this functional interpretation of deep interdental folds. Enamel thickness in Batrachotomus is quite thin relative to that in both therapsids and Tyrannosaurus (Whitney et al. 2020), but is within the reported range of 'rauisuchians' and other theropods (Sander 1999;Wang et al. 2015). Sander (1999) reported that 'rauisuchian' enamel was thinnest on the tops of the denticles; this is in contrast to what we observed in Batrachotomus (Fig. 9R, S), but might be attributable to carinal wear in the tooth sampled by Sander (1999) (see worn denticles in longitudinal and transverse section in Fig. 9Q, R). Here, considering the distribution and types of macrowear, variation in enamel thickness across the tooth crown in Batrachotomus is correlated with variation in wear, with enamel thickness greatest in those areas of the tooth most prone to contact with bone during feeding (Fig. 9Q-S). The overall tooth microstructure of Batrachotomus has been interpreted as a convergent character of hypercarnivores engaging in puncture and pull feeding (D'Amore 2009; Whitney et al. 2020), increasing wear resistance of the denticles and while maintaining overall tooth elasticity (Brink et al. 2015;Wang et al. 2015;Gignac & Erickson 2017). This functional interpretation is congruent with the wear and breakage patterns of Batrachotomus teeth described here (see Morphology and macroscopic wear on Batrachotomus teeth, above; Fig. 9; Data S2) indicating specializations for puncture and pull feeding by Batrachotomus on various large sympatric vertebrates.

Palaeoecological implications of the feeding habits and taphonomy of Batrachotomus
Feeding ecology and taphonomy. The distribution and frequency of bite traces is suggestive of different behaviours by Batrachotomus, including scavenging, cannibalism and predation. We found a strong correlation between the frequency of bone types in a Mastodonsaurus skeleton and the frequency with which each bone is bitten (Fig. 13), suggesting a systematic and virtually complete exploitation of carcasses by Batrachotomus. Different bones of Mastodonsaurus and Batrachotomus from Kupferzell were found in clusters reliably referred to single individuals (Wild 1980;Gower 1999;Schoch 1999;Gower & Schoch 2009; Data S1). These associations, together with the lack of bone sorting, indicate that there was little-to-no transport of carcasses, at least in Kupferzell. In fact, in Vellberg-Eschenau, where bite traces are notably lower in number than in Kupferzell (Data S1), evidence of bone sorting has been reported (Schoch & Seegis 2016). This would indicate a higher washout of carcasses in Vellberg-Eschenau, further noting slight sedimentological differences (Urlichs 1982; DS, EM, RRS, pers. obs.; Fig. 1B). Nonetheless, between Kupferzell and Vellberg-Eschenau there are no differences in the location of bite traces on bones nor in the macroscopic wear of Batrachotomus teeth (Data S1, S2). These similarities suggest that, as a whole, this bite trace assemblage is not an unusual scenario under specific environmental conditions, but the record of the usual feeding behaviour of the bite maker. Hypothetically, this is also indicated by the predictable distribution of composite bite traces in different taxa, different bones, and different bone regions 7,8) suggesting that, during feeding, Batrachotomus modified bones in a regular, monotonous manner.
The distribution and frequency of bite traces produced by Batrachotomus across bone regions (Fig. 8) (2006) associated the presence of relatively deep traces to disarticulation attempts, which was also likely to have been the case for the bones bitten by Batrachotomus (see Usage of Batrachotomus teeth: correlating tooth wear and bite traces, above). Otherwise, also like in crocodylians (Drumheller & Brochu 2014), the shafts of long bones from the Lower Keuper more commonly (though not exclusively) display relatively shallow grooves (Linichnus and Knethichnus) that usually do not remove the entire bone cortex and are oriented oblique to perpendicular with respect to long bone axis (e.g. Fig. 3D, E; Data S1). Similarly, spalled flakes of bone, often covered with Knethichnus, are mainly found on the shaft of long bones (e.g. Figs 2D, F, 3A-C), also reminiscent of those produced by crocodylians (Drumheller & Brochu 2014;Njau & Gilbert 2016). Such bite traces are likely to be associated with defleshing rather than disarticulation, as has been suggested for crocodylians (Njau & Blumenschine 2006). All in all, the differential distribution of bite traces supports the hypothesis of different usage for each bone region, in agreement with the statistical results ( Fig. 7) and further pointing to conserved feeding behaviour in pseudosuchians (Fig. 14).The incompleteness of skeletons would suggest habitats roamed by scavengers (Hungerb€ uhler 1998), tearing carcasses and thus facilitating the loss of specific parts that otherwise would have been preserved. Similarly, the high abundance of shed teeth within mudstones also indicates a low energy environment inhabited by several individuals (Hungerb€ uhler 1998;Augustin et al. 2020). Further evidence of scavenging is provided by the presence of bite traces produced by much smaller organisms than Batrachotomus. The Machichnus-like traces suggest abandoned corpses consumed by small-sized scavengers (probably archosauriforms; see teeth in Schoch et al. 2018) that could reach carcasses when relatively high nutrient-value regions (such as the rib cage) were still available (Drumheller et al. 2020). Therefore, even though there may have been periods of stressed environments (e.g. droughts) as indicated by histological data from Batrachotomus bones (Klein et al. 2017), scavenging was not an unusual habit. F I G . 1 3 . Frequency of bitten bones of Mastodonsaurus and Batrachotomus. Numbers of bones (N) account for the bones that are known for each taxon and skeletal region (e.g. forelimb includes six bones: two humeri, two radii and two ulnae); teeth, phalanges, tarsals, carpals and vertebrae are excluded because their total amount per skeleton is unknown. Sacral ribs (differentiated in the statistical analyses) are here counted together with thoracic and caudal ribs. Restored skeletons are coloured according to the skeletal regions shown in the table and in Figures 7 and 8. Note the high percentage of bitten bones from the pelvic girdle of Batrachotomus, being nearly three times the percentage of these bones per skeleton (highlighted in bold).
Remarkably, Batrachotomus also fed on conspecifics (Figs 2C-E, G, 3F, 4I-K, 5E, N, 8; Data S1), representing one of the few cases of cannibalism in the archosaur fossil record, previously only reported in non-avialan theropods (Rogers et al. 2003;Longrich et al. 2010;Drumheller et al. 2020), but well-known in present-day crocodylians (Cott 1961). Most of the bite traces on Batrachotomus bones are concentrated on ribs and pelvic elements (Fig. 13), which give access to the nutrient-rich viscera and muscles of the tail and hindlimbs. Considering their location on the bones and the lack of evidence of healing, these traces could not have been produced in a living specimen; therefore, they indicate feeding rather than non-trophic intraspecific interactions such as social behaviour. This latter behaviour is known from both fossil pseudosuchians (including crocodylians) (Abel 1922;Buffetaut 1983;Avilla et al. 2004;Katsura 2004) and present-day crocodylians (K€ alin 1936;Cott 1961;Webb & Manolis 1983), and mostly results in wounds on skull bones (either healed or not) and/or traumatic pathologies on regions of the skeleton that can be reached in a living individual. The Batrachotomus teeth preserving Knethichnus and Linichnus traces on their labial surfaces (Data S1) provide conclusive evidence of intraspecific interaction, possibly competition, among living individuals and are consistent with aggressive interactions seen in fossil pseudosuchians and extant crocodylians (Abel 1922;Cott 1961;Buffetaut 1983). This also agrees with purported evidence for gregariousness in Triassic pseudosuchians (Franc ßa et al. 2011;Nesbitt et al. 2013Nesbitt et al. , 2020. Batrachotomus was less frequently bitten than Mastodonsaurus ( Fig. 8; Data S1), being a more occasional food source, and possibly representing evidence of scavenging behaviour. Thus, under unstable environmental conditions, Batrachotomus possibly adopted cannibalism. Nonetheless, the relatively high number of Batrachotomus bones with bite traces suggests that cannibalism was not a totally unusual feeding habit. The frequency of bitten bones with respect to their number in a skeleton of Batrachotomus (Fig. 13) suggests that carcasses of this taxon were not completely exploited and, as the Machichnus-like traces on Batrachotomus ribs suggest, small-sized scavengers did have access to them (Drumheller et al. 2020) (see above).
The large Nihilichnus found on opposing surfaces of Mastodonsaurus femora (Fig. 3C-E), even preserving the outline of the tooth ( Fig. 2A), indicate active predation, as they are strikingly similar to those produced by large Triassic predatory archosaurs (Nied zwiedzki et al. have not been found. Thus, if hunting did occur, all the evidence indicates a predominance of successful attacks. Further interaction between Mastodonsaurus and Batrachotomus is recorded by two Mastodonsaurus tusks showing Linichnus and Knethichnus traces (Data S1). Since no bite traces on Mastodonsaurus skull bones were produced by Batrachotomus (Fig. 6A-C), it is unlikely that these bite traces on teeth represent feeding; rather they probably represent a face-to-face interaction.
Considering that Mastodonsaurus most commonly roamed ( Palaeoecological implications for archosaurs. The aforementioned morphological similitudes between the dentaries of Batrachotomus and ichthyophagous theropods are in accordance with the similarities between the teeth of Batrachotomus and megalosauroids (see Comparison of Batrachotomus teeth with those of theropod dinosaurs and other archosauromorphs, above; Fig. 12; Table 3), a theropod clade that evolved piscivorous adaptations (Charig & Milner 1997;Sadleir et al. 2008). At first glance, this is unexpected, given the evidence for hypercarnivory in Batrachotomus documented herein. However, as discussed above, such similitudes suggest a potentially broad diet not reflected in the bite-trace record. Ticinosuchus, another Triassic pseudosuchian, is also known to have included fish in its diet (Nesbitt 2011). Thus, Batrachotomus and Ticinosuchus highlight the potential for a broad dietary spectrum in basal pseudosuchians. In this regard, some regurgitalites including skeletal remains from Vellberg-Eschenau (Schoch & Seegis 2016) could have been produced by Batrachotomus, as suggested for Triassic 'rauisuchians', representing another potential behaviour shared with crocodylians (Gordon et al. 2020) and further showing conserved feeding ecology in pseudosuchians (Fig. 14). Multidisciplinary studies including bite traces, tooth macrowear and microanatomy, as well as comparisons with taxa with similar dentition, allow for the better identification of feeding habits and ecological roles of extinct taxa. In the same way, studies specifically focused on tooth macrowear and/or microanatomy may already provide information of the feeding ecology of the analysed group (e.g. D 'Amore 2009;Brink et al. 2015;Wang et al. 2015). This is in line with the functionality study of denticulated (ziphodont) teeth by D'Amore (2009). As pointed out in that work, the mesial carina of theropod teeth tend to be less denticulated, suggesting that the mesial carina was used less than the distal carina during feeding. In Batrachotomus, this lack of denticles is only observed in the basal portion of the mesial carina of some premaxillary teeth ( Fig. 9B; Data S2) and, in contrast to what is observed and expected in theropods (cf. Fiorillo 1991), Batrachotomus teeth show higher degrees of abrasion (wear) on the mesial carina (Figs 10, 11), indicating that these parts were more used during feeding. Such wear can be linked to the relatively high frequency of Knethichnus (Fig. 8), especially also when this morphotype outlines fan-shaped traces and when it is associated with Linichnus in clusters of traces (e.g. Figs 2B, H, 3D, 4K-L). Such traces represent repetitive biting strokes and show changes in the direction of movement of teeth contacting the bone (probably indicating defleshing attempts; see also Feeding ecology and taphonomy, above). Knethichnus is a bite trace morphotype exclusively linked to denticulated (ziphodont) teeth (Jacobsen & Bromley 2009) also produced by theropod dinosaurs and V. komodoensis. However, in these groups, this bite trace morphotype is much less frequent than others, such as This notable difference is more likely to be explained by different bone usages and feeding behaviours between clades, showing that Batrachotomus and pseudosuchians, including extant crocodylians, process the bones more extensively than theropods and V. komodoensis (Fig. 14). Fiorillo (1991) noticed that isolated theropod teeth are common in dinosaur fossil localities and that few of them show signs of wear, explaining the low frequency of bone tracing of the clade (and possibly the lack of denticulation of some teeth: D'Amore 2009). Indeed, low levels of tooth wear are in contrast with our findings from the Batrachotomus tooth assemblage (Figs 9-11; Data S2), including only a few pristine tooth crowns probably corresponding to unerupted teeth. Fiorillo (1991) concluded that theropod dinosaurs did not usually crush bones (potentially with a few exceptions, see Hone & Rauhut 2010) and considered that this ecological role was unoccupied during the Mesozoic. While agreeing with a low level of bone crushing activity by theropods, considering the high frequency of bone tracing by pseudosuchians, we suggest that this ecological role was occupied by the pseudosuchians. Potential evidence for this hypothesis is found in the Lower Cretaceous vertebrate lagerst€ atte from France analysed by Rozada et al. (2021). These authors reported numerous bite traces only associated with crocodylomorphs (see also Gônet et al. 2019), even though megalosauroid theropods are also present in the locality (mostly represented by teeth; Rozada et al. 2021). Notably, our analysis of archosauromorph teeth shows that those of Batrachotomus are mostly similar to those of megalosauroids ( Fig. 12; Table 3). Therefore, although Batrachotomus and megalosauroids probably shared some feeding habits (see above), it is likely that they also had different ecological roles, again suggesting a broad diet for Batrachotomus.
All in all, the relatively lower frequency of bonetracing by theropods with respect to pseudosuchians (including present-day crocodylians) may relate to different behaviours in the manipulation of prey or carcasses; the former groups would mostly engage in pullback strokes (possibly similar to the behaviour of V. komodoensis; e.g. D'Amore 2009; D'Amore & Blumenschine 2009Blumenschine , 2012Snively et al. 2013) resulting in a lower likelihood of tooth-bone contact than that of pseudosuchians (Fig. 14), which usually engage in violent and abrupt moves to tear prey and carcasses (e.g. Drumheller & Brochu 2014Njau & Gilbert 2016). Recently, Drumheller et al. (2020) documented a bone assemblage with unusually high frequencies of theropod bite traces. These authors linked this to long exposure times for the remains and late access to carcasses (scavenging, including cannibalism) in a stressed ecosystem. Nonetheless, Drumheller et al. (2020) also highlighted potential collecting biases against remains with bone surface modifications (see also Pobiner 2008), which could be the case for tyrannosaurs, probably overrepresented in the literature (see also Hone & Rauhut 2010). Even if collecting biases exist in dinosaur localities (as shown by Drumheller et al. 2020), in the case of the Lower Keuper, many complete (or nearly complete) bones display bite traces (Figs 3-5; Data S1). This indicates that biases against collection and preparation of incomplete bones do not necessarily preclude the collection of bitten elements, and supports a higher frequency of bone tracing by Batrachotomus and pseudosuchians than by theropod dinosaurs, probably resulting from the deep phylogenetic split between these two archosaur clades (Fig. 14).

CONCLUSION
By analysing bite traces on bones produced by the pseudosuchian Batrachotomus kupferzellensis together with tooth morphology, microanatomy and macrowear, and comparing them with those of other pseudosuchians (including extinct and extant crocodylians), theropod dinosaurs, and Varanus komodoensis, we suggest that phylogeny can be a better predictor of feeding behaviour than tooth morphology (Fig. 14). Even if tooth-bone contact during feeding was mostly unintentional in Batrachotomus, as suggested for non-avialan theropods and V. komodoensis (Fiorillo 1991; D'Amore & Blumenschine 2009), the high frequency of bitten bones, the predictable distribution of traces within the Lower Keuper assemblage, and the severe damage of many teeth indicate that Batrachotomus did not actively avoid such contact. This is comparable to the feeding behaviour of crocodylians , a likely synapomorphy of this long-lived hypercarnivorous group. This points to a dichotomy in the interpretation of bite traces: on the one hand, trace morphologies correlate with morphology of teeth and their movement along the bone; on the other hand, the association, location and frequency of bite traces reflect the behaviour of the bite maker, probably differing between carnivorous clades (Fig. 14). Furthermore, our results show that tooth gross morphology alone may not reflect the complete dietary spectrum of a given taxon; this can be overcome by combining analyses of macroscopic wear and microanatomy, which provide independent lines of evidence of dietary preference and feeding behaviour.
Our multidisciplinary analyses of an exceptional association of bite traces, tooth wear patterns, and tooth microanatomy from the Lower Keuper fossil lagerst€ atten allow the feeding ecology of a 240-million-year-old pseudosuchian reptile to be understood in detail. Despite morphological affinities of Batrachotomus teeth with those of megalosauroid theropod dinosaurs, the behaviours underlying the produced bite traces share greater similarities to those produced by members of the crocodylomorph lineage. This suggests evolutionary conservation of feeding behaviour in this clade, indicating that this successful feeding ecology extended from Triassic pseudosuchians until present-day crocodylians. Concurrently, longterm behavioural and ecological separation between pseudosuchians and theropod dinosaurs was probably the result of the deep phylogenetic split between them.