Assessment of trophic ecomorphology in non‐alligatoroid crocodylians and its adaptive and taxonomic implications

Abstract

Although the establishment of trophic ecomorphology in living crocodylians can contribute to estimating feeding habits of extinct large aquatic reptiles, assessment of ecomorphological traits other than the snout shape has scarcely been conducted in crocodylians. Here, I tested the validity of the proposed trophic ecomorphological traits in crocodylians by examining the correlation between those traits and the snout shape (an established trophic ecomorphology), using 10 non‐alligatoroid crocodylian species with a wide range of snout shape. I then compared the ontogenetic scaling of trophic ecomorphology to discuss its adaptive and taxonomic significance. The results demonstrated that degree of heterodonty, tooth spacing, size of supratemporal fenestra (STF), ventral extension of pterygoid flange and length of lower jaw symphysis are significantly correlated with snout shape by both non‐phylogenetic and phylogenetic regression analyses. Gavialis gangeticus falls outside of 95% prediction intervals for the relationships of some traits and the snout shape, suggesting that piscivorous specialization involves the deviation from the typical transformation axis of skull characters. The comparative snout shape ontogeny revealed a universal trend of snout widening through growth in the sampled crocodylians, implying the existence of a shared size‐dependent biomechanical constraint in non‐alligatoroid crocodylians. Growth patterns of other traits indicated that G. gangeticus shows atypical trends for degree of heterodonty, size of STF, and symphysis length, whereas the same trends are shared for tooth spacing and ventral extension of pterygoid flange among non‐alligatoroid crocodylians. These suggest that some characters are ontogenetically labile in response to prey preference shifts through growth, but other characters are in keeping with the conserved biomechanics among non‐alligatoroid crocodylians. Some important taxonomic characters such as the occlusal pattern are likely correlated with ontogeny and trophic ecomorphology rather than are constrained by phylogenetic relationships, and careful reassessment of such characters might be necessary for better reconstructing the morphological phylogeny of crocodylians.

Introduction

Ecomorphology, in its broad sense, is the study of the interaction between the organismal design and environment (Bock, 1990; Wainwright & Reilly, 1994; Motta et al. 1995), and the correspondence between the two is made by linking the morphology with function and performance, and by measuring the effects of the latter on fitness (Arnold, 1983; Garland & Losos, 1994; Ricklefs & Miles, 1994; Wainwright, 1994). Because morphology can be rigorously measured based on the homologous structures and has a high repeatability of measurement, it has been used to derive ecological inferences including resource partitioning and niche relationships among taxa, community structure, and diversity within taxa (Ricklefs & Miles, 1994). Probably, the utility of ecomorphological analysis is most appreciated in paleontological researches, because direct measurements of performance and fitness are not possible in extinct animals, but the morphology of hard tissues is often available in fossils. By using morphology, we can estimate important biological aspects of extinct organisms such as feeding and locomotor habits and body size (Kubo & Benton, 2009; Zanno & Makovicky, 2011; Campione & Evans, 2012), and can get insights into the ecological diversity pattern, species or higher level interactions, and history of adaptation in an evolutionary time scale (Brusatte et al. 2008; Sookias et al. 2012; Larson et al. 2016).

Crocodylians are opportunistic ambush predators, feeding on a wide range of invertebrates and vertebrates, depending on the availability of their habitat (Pooley, 1989; Webb & Manolis, 1989). Generally, their feeding habits are thought to reflect the skull shapes, and most of its variation is explained by broad to slender‐snouted shape axis in extant as well as extinct forms (Brochu, 2001; Pierce et al. 2008; Okamoto et al. 2015). The extreme end‐member in the slender‐snouted taxa is Indian gharial (Gavialis gangeticus), which is considered as the most predominant fish‐eater in extant crocodylians (Whitaker & Basu, 1982; Trutnau & Sommerlad, 2006). Its long and narrow snout accounts for the high angular velocity of the jaw tip while lateral head sweeping (Thorbjarnarson, 1990) and jaw closure, thus favorable for catching small agile prey. However, this type of snout is mechanically weak for stress during biting (Busbey, 1995; McHenry et al. 2006; Pierce et al. 2008; Walmsley et al. 2013), and hydrodynamically costly as compared with the cranial morphology of generalist‐type taxa (e.g. Crocodylus porosus and Crocodylus niloticus; McHenry et al. 2006). On the other hand, experimental studies revealed that interspecific variation in crocodylian bite force is primarily explained by body mass if G. gangeticus is excluded, and slender‐snouted forms can generate equivalent bite forces with broader‐snouted forms, suggesting that slender‐snouted species may reduce jaw stresses by targeting on smaller prey such as fish and crustaceans (Erickson et al. 2012).

Although the snout shape has been particularly focused in the previous ecomorphological analyses of crocodylians, there are a number of other characters that have also been proposed as ecomorphological characters (Iordansky, 1973; Langston, 1973; Brochu, 2001). However, quantitative assessments of such potential ecomorphological traits have rarely been conducted (Walmsley et al. 2013; Gignac & Erickson, 2015). Crocodylians are known as the sole remnants of Pseudosuchia (the most inclusive clade containing C. niloticus but not Passer domesticus: Nesbitt, 2011), a diverse and successful group of reptilians that rivaled dinosaurs in the early Mesozoic (Brusatte et al. 2008), and are often regarded as modern analogs of extinct large‐bodied aquatic predators. Hence, establishment of trophic ecomorphology in living crocodylians can contribute to estimating feeding habits in extinct pseudosuchians and non‐pseudosuchian reptiles such as marine crocodyliforms, dinosaurs, phytosaurs and choristoderans (Erickson, 1972; Massare, 1987; Hunt, 1989; Sereno et al. 1998, 2001), and helps answer various ecological questions by integrating space, time and phylogenetic perspectives. Moreover, the finding of the ecological characters and their correlation (Sadleir & Makovicky, 2008) may help better understand the crocodylian phylogeny. It should be noted that a suite of potential feeding ecomorphological traits in rostrum is indeed consistent with molecular tree rather than supporting the morphological tree (Gatesy et al. 2003). However, removal or proper weighting of such non‐independent characters are necessary for rigorously comparing the fit of morphological data on morphological and molecular trees, and thus resolving the long‐standing discussion on the placement of G. gangeticus in the crocodylian phylogeny (Densmore, 1983; Norell, 1989; Poe, 1996; Brochu, 1997; Gatesy et al. 2003).

Meanwhile, as most of the animals do, crocodylian skulls change remarkably during the postnatal development (Mook, 1921b; Müller, 1927; Kälin, 1933; Dodson, 1975; Monteiro & Soares, 1997; Monteiro et al. 1997; Sadleir, 2009; Piras et al. 2010; Bona & Desojo, 2011; Watanabe & Slice, 2014; Fernandez‐Blanco et al. 2015; Foth et al. 2015). Previous studies demonstrated that growth trajectories of Crocodylus skulls can fall into some stages, and hypothesized that the breakpoints between the stages correspond to changes in diets (Hall & Portier, 1994; Radloff et al. 2012), whereas an experimental study suggested that positive allometry of bite force is conserved among crocodylians regardless of snout shape variation (Erickson et al. 2014). Comparisons of the developmental pathways of trophic ecomorphology among crocodylians would tell us whether dietary adaptation during growth occurs through certain morphological changes, and the role of developmental process in shaping the morphological and ecological diversity of crocodylians.

Here, using the postnatal growth series of extant crocodylians, I first assess the validity of proposed trophic ecomorphology of non‐alligatoroid crocodylians, with the special focus on piscivorous ecomorphology. I then compare the growth patterns of trophic ecomorphology among non‐alligatoroid crocodylians for finding differential and shared scaling patterns, and discuss their adaptive and taxonomic significance. Although the current study restricts the sample to non‐alligatoroid crocodylians, additional sampling of alligatoroid species may permit the generalization about the observed trends and patterns among crocodylians.

Materials and methods

Specimen sampling

For conducting species comparisons of snout shape and the associated skull characters among Crocodylia, multiple taxa with a wide range of snout shape are required. I sampled the postnatal growth series of skulls from 10 crocodylians (n = 236; seven species of Crocodylus, Mecistops cataphractus, Tomistoma schlegelii and G. gangeticus; Fig. 1). Alligatoroid species were not included in the sample because the current study specifically focuses on piscivorous specialization. Crocodylus is the most diverse and species‐rich genus in extant crocodylians, including 12 known species (Grigg & Kirshner, 2015) from blunt‐snouted type to slender‐snouted type. A recent molecular study by Oaks (2011) revealed the late Miocene divergence of this group, suggesting that Crocodylus snout shape is ecologically plastic and is not strongly constrained by the phylogenetic relationships. Three remaining non‐Crocodylus species are all slender‐snouted, yet they differ in the degree. In crocodylian systematics, the relationships among Crocodylus, and placement of T. schlegelii and G. gangeticus are still not fully settled (Densmore, 1983; Densmore & White, 1991; Brochu, 1997, 1999; Gatesy et al. 2003; McAliley et al. 2006; Roos et al. 2007; Oaks, 2011). However, in the current study, I chose to use the most recent and comprehensive time‐calibrated molecular tree by Oaks (2011) as a phylogenetic framework for statistical analyses because, contrary to the molecular analyses, previous morphological analyses (Brochu, 2000; Brochu et al. 2010; Brochu & Storrs, 2012) have not resolved the phylogenetic relationships among Crocodylus. Nearly all (232/236) samples used in this study are wild‐caught with locality information (Table S1). In 10 crocodylians used in this study, some species are known to have morphologically distinct cryptic populations [C. niloticus (Hekkala et al. 2011; Nestler, 2012); Crocodylus novaeguineae (Hall, 1989); Crocodylus palustris (Trutnau & Sommerlad, 2006) and M. cataphractus (Shirley et al. 2014)]. With this in mind, I sampled individuals from a limited geographical area for C. niloticus (Congo basin: Hekkala et al. 2011; Nestler, 2012) and for M. cataphractus (Congolian zone: Shirley et al. 2014), but other species are from all available localities to increase the sample size.

Figure 1.

A time‐calibrated molecular phylogeny of 10 sampled crocodylian species (based on Oaks, 2011).

Data acquisition

A list of previously proposed trophic ecomorphology in crocodylians that are functionally explained in relation with feeding behavior is shown in Table 1. Among the variables, I did not consider the tooth shape because teeth are often shed and their original positions are not available in the specimens. To assess validity of the proposed ecomorphological characters in crocodylians, I conducted regression analyses of the proposed ecomorphological characters against the established ecomorphological character (snout shape) to examine their correlations. In crocodylians, snout shape is considered as being correlated with their diet (Brochu, 2001), and is known to reflect the feeding function and performance (Taylor, 1987; Thorbjarnarson, 1990; McHenry et al. 2006). Therefore, if a character correlates with snout shape, then it can also be regarded as a trophic ecomorphological character.

Table 1.

Proposed trophic ecomorphological variables and their piscivorous states in crocodylians

Variable	Piscivorous state	Function and performance	Reference
Snout shape	Long and slender	Higher angular velocity of jaw tip during lateral head sweeping and jaw closing, thus favorable for catching small agile prey	Taylor, 1987; Thorbjarnarson, 1990; McHenry et al. 2006
Degree of heterodonty (tooth size disparity)	Less degree of heterodonty (coupled with less undulated jaw margin)	Efficient for detaining relatively docile prey, but inefficient for holding larger prey firmly	Iordansky, 1973; Busbey, 1995
Spacing of teeth	Widely spaced	Striking and impaling prey	Langston, 1973
Tooth count	Increased in number	Increase the chance of catching prey and facilitates transport of the food items towards the pharynx (while inertial feeding)	Busbey, 1995
Tooth shape	Long, slender and recurved	Striking and impaling prey, and maneuvering fish back toward gullet	Langston, 1973; Massare, 1987; Busbey, 1995
Size of STF	Increased	Increased mass of M. pseudotemporalis, which is assciated with rapid but weak jaw adduction	Iordansky, 1964; Endo et al. 2002
Ventral extension of pterygoid flange	Reduced	Reduced mass and moment arm of Mm. pterygoidei dorsalis et ventralis (strong jaw adductors), and reduction in maximal mouth opening	Iordansky, 1964; Endo et al. 2002; Piras et al. 2014
Development of basioccipital tubera	Highly developed and pendulous	Modified neck musculature (M. longissimus capitis profundus and M. rectus capitis anticus major; Tsuihiji 2007) advantageous for head motions during fish eating (e.g. lateral head sweeping)	Langston, 1973
Length of lower jaw symphysis	Elongated	Incur higher strain under loads used for feeding upon large prey	Walmsley et al. 2013

STF, supratemporal fenestra.

For the quantification of the snout shape, I performed two‐dimensional geometric morphometrics with seven landmarks (type 1 and 2 of Bookstein, 1991) and 38 sliding semi‐landmarks that were digitized from palates (Fig. 2A), using tpsdig v. 2.17 (Rohlf, 2013a). The current method is probably better in capturing snout shapes than previous studies that used the ratio of linear measurements from dorsal snouts (Hall & Portier, 1994; Erickson et al. 2012), which did not take into account subtle differences in snout outlines. Landmark and semi‐landmark coordinates of the pooled samples (n = 236) were superimposed by a generalized Procrustes analysis (Gower, 1975; Rohlf & Slice, 1990), and subjected to relative warp (RW) analysis using tpsrelw v. 1.53 (Rohlf, 2013b). Because much of the snout shape variation would be accounted for by the first axis of RWs, RW1 scores were extracted as the snout shape proxy.

Figure 2.

(A) Rostral landmarks and semi‐landmarks used in this analysis. Seven landmarks (white markers) are taken at: 1: anterior end of rostrum; 2, 3: both lateral edges of premaxillary–maxillary contact; 4, 5: both lateral edges of maxillary alveolus #5; 6, 7: both lateral edges of rostrum at the level of posterior end of last maxillary alveolus. A total of 38 semi‐landmarks (gray markers) are taken on the outline of six palatal outline segments. (B) Measurement of lower jaw symphysis length. (C) Nineteen post‐rostral landmarks taken at: 1: orbit‐prefrontal‐lacrimal contact; 2: orbit‐lacrimal‐jugal contact; 3: orbit‐prefrontal‐frontal contact; 4: orbit‐frontal‐postorbital contact; 5: anteroventral end of postorbital bar; 6: posteroventral end of postorbital bar; 7: anterodorsal end of postorbital bar; 8: posterodorsal end of postorbital bar; 9: infratemporal fenestra‐jugal‐quadratojugal contact; 10: lateral edge of skull table at postorbital‐squamosal contact; 11: supratemporal fenestra (STF)‐postorbital‐squamosal contact; 12: triple junction of frontal‐parietal‐postorbital; 13: STF‐parietal‐postorbital contact; 14: STF‐parietal‐squamosal contact; 15: frontal‐parietal contact on midline; 16: supraoccipital‐parietal contact on midline; 17: posterior edge of skull table at parietal‐squamosal contact; 18: posteromedial corner of quadrate jaw joint; 19; posterior end of quadrate‐quadratojugal contact. (D) Landmarks and semi‐landmarks for calculating area of STF. Four landmarks (white markers) are taken at: 1: anterior edge of STF at postorbital‐parietal contact; 2: lateral edge of STF at postorbital‐squamosal contact; 3: medial‐most point of STF along parietal; 4: posterior edge of STF at parietal‐squamosal contact. A total of 12 semi‐landmarks (gray markers) are taken on the outline of STF to calculate the STF area including its smooth floor. (E) Measurements of pterygoid flange depth (PFD) and basioccipital tubera width.

The degree of heterodonty (tooth size disparity) was calculated as the coefficient of variation (ratio of standard deviation to mean) of mesiodistal diameters of all right/left maxillary alveoli (Table S2). The spacing of teeth was represented as the ratio of the sum of mesiodistal diameters of all right/left maxillary alveoli to the maxillary tooth row length (Fig. 2A), which was measured in juvenile–adult specimens showing interalveolar bones. The tooth count was taken from left and right maxillae, because tooth counts in maxillae are more variable than those of premaxillae among crocodylians (Iordansky, 1973). These tooth size, tooth spacing and tooth count patterns in the upper jaw can almost certainly be generalized for the lower jaw (Brown et al. 2015). The size of supratemporal fenestra (STF) was measured as an area enclosed by landmarks and semi‐landmarks (Fig. 2D) using tpsutil v. 1.58 (Rohlf, 2013c), based on the photos of posterior skulls taken from dorsal view, with the scale on the flat part of skull table. The measurements of other metrics are shown in Fig. 2B,E. As the body size proxy, I used the centroid size (CS) of superimposed dorsal post‐rostrum landmarks (Fig. 2C) of each species obtained from tpsrelw, because the post‐rostrum shape is more conservative compared with the rostrum (Piras et al. 2014). To obtain the size information for the digitized landmark data, samples were scaled with reference to maximum skull width at jaw joint in tpsdig. Linear measurements were taken by Tresna point digital caliper (Series SC02, ID 111–203, resolution = 0.01 mm). Photos of palatal and dorsal aspects of the skulls for geometric morphometrics were taken by positioning Nikon D3100 camera (AF‐S DX NIKKOR 18–55 mm f/3.5–5.6G VR, 14.2 Mp resolution) perpendicular to the palatal and skull table surface, respectively, consistently centering specimens in the field of view, and keeping enough distance between the lens and specimens.

Assessment of the proposed trophic ecomorphology in non‐alligatoroid crocodylians

To size‐normalize the dimensioned variables, linear measurements were divided by CS, and area measurements were square rooted and then divided by CS. However, even after the size‐removal, various allometric trends in ecomorphological variables need to be controlled. A common approach for correcting allometry is to regress variables on size metrics using least squares regression and then to take the residuals (Piras et al. 2014; Watanabe & Slice, 2014; Foth et al. 2015), but crocodylian snout shapes are known to show non‐linear growth trajectories (Hall & Portier, 1994; Radloff et al. 2012; this study), thus use of such an approach would be inappropriate. Therefore, I selected individuals within the arbitrarily set thresholds of log10 transformed CS [log(CS)], which correspond to subadult–adult stages for all species. For examining the correlations between proposed ecomorphological variables and snout shape (RW1 score) within the set size range, I first conducted ordinary least squares (OLS) regression with whole samples. Maxillary tooth count was not analyzed in the whole‐sample OLS. Secondly, species comparison was performed by phylogenetic generalized least squares (PGLS) regression, in which average species ecomorphological variable values were regressed on average species snout shape (RW1 score) within the set size range, on an assumption of Brownian motion model of trait evolution. For maxillary tooth count, mode instead of mean was used for the regression. Because PGLS regression requires a time‐calibrated tree, a phylogenetic tree of 10 sampled crocodylian species (Fig. 1) was constructed with reference to a molecular study by Oaks (2011), and divergence time estimates were adopted from the species tree of the original study with a 90 myr upper limit on the root age. Additionally, a morphological tree (Fig. S1; Appendix S1) was used for testing the sensitivity of the PGLS results to tree shape. PGLS regression was conducted using software r with caper package (Orme et al. 2013), and a phylogenetic signal parameter (Pagel's λ) was optimized and took into account in the analysis.

Comparative growth trajectories of trophic ecomorphology in non‐alligatoroid crocodylians

Prior to analyses, linear and area measurements were log10 transformed, yet dimensionless variables (snout shape, degree of heterodonty and spacing of teeth) were left untransformed. The growth trajectories of ecomorphological variables were visualized using a bivariate plot of log(CS) of the post‐rostrum and each variable. Because the trajectories of snout shape were non‐linear with breakpoint(s), piecewise linear models, where two straight lines are joined at an unknown knot (breakpoint), were fit to find the breakpoint. The analysis was conducted using r with the sizer package (Sonderegger, 2012), and the breakpoint estimate was reported with 95% confidence intervals. In the model fitting, very small individuals were excluded by setting an arbitrary size threshold for all species so as not to include individuals from relative skull growth phase I (Hall & Portier, 1994). Therefore, the first and second segments in the model correspond to the skull growth phases II and III (Hall & Portier, 1994). Because some species do not grow larger than the assumed breakpoint or have small sample size beyond the breakpoint, the analysis was performed only in large‐bodied well‐sampled taxa (six species: Crocodylus acutus, C. niloticus, C. palustris, C. porosus, G. gangeticus and T. schlegelii). After the mean snout shape breakpoint of six species was obtained, I additionally tested if snout shapes differ between growth phase II (smaller than mean breakpoint) and phase III (larger than mean breakpoint) in the six species using Welch's t‐test.

For the remaining variables, I performed reduced major axis (RMA) regression of ecomorphological variables against log(CS) with whole samples using r with smatr package (Warton et al. 2012). Dimensionless variables were examined if there are significant relationships between the variables and size [log(CS)], and dimensioned variables were examined if RMA slopes for the relationships of variables (log10 transformed) and size [log(CS)] are different from the expected isometric slopes (1.0 and 2.0 for linear and area measurement variables, respectively). The growth of the variables was considered either positively or negatively allometric if 95% confidence intervals of the slopes do not overlap the expected isometric slopes. Because maxillary tooth count is different in its data type, it was separately subjected to Spearman rank correlation to test the correlation between tooth count and size. All statistical tests were carried out with r language and environment (R Core Team, 2014).

Results

Assessment of the proposed trophic ecomorphology

Because body size [log(CS)] of sampled crocodylians (10 species, n = 236) ranges from ~1.6 to 2.6, and snout shape (RW1 score) largely changes beyond log(CS) of ~ 2.4 for all taxa (Fig. 4), I selected individuals from the arbitrary set size range [2.1 ≤ log(CS) ≤ 2.4], which covers a relatively stable size range of the snout shape, and secures 110–123 individuals for each variable. A summary of the whole‐sample OLS regression and species PGLS regression between proposed trophic ecomorphological variables and snout shape (RW1 score) are found in Fig. 3 and Table 2. Among the proposed ecomorphological variables, CV of maxillary alveolus diameter, maxillary tooth spacing, relative STF area, relative pterygoid flange depth (PFD) and relative lower jaw symphysis length are significantly correlated with snout shape in both whole‐sample OLS and species PGLS regressions, thus verified as trophic ecomorphology. The use of an alternative tree (morphological tree) for PGLS did not significantly change the results (Table S3). These indicate that generally, tooth size variability becomes smaller, maxillary teeth are more spaciously positioned, STF size increases, PFD decreases and lower jaw symphysis length extends in accord with the snout narrowing in the sampled crocodylians. On the contrary, relative basioccipital tubera width and mode species maxillary tooth count were not significantly correlated with snout shape in either whole‐sample OLS or species PGLS regression.

Figure 4.

Bivariate plot of log(CS) and snout shape (RW1 scores). Area with gray gradient represents the estimated range of log(CS) breakpoints of snout shape for six species (C. acutus, C. niloticus, C. palustris, C. porosus, G. gangeticus and T. schlegelii), and the dashed line is their mean.

Figure 3.

(A,B,D–G) Bivariate plots of snout shape (RW1 score) and proposed ecomorphological variables for individuals within 2.1 ≤ log(CS) ≤ 2.4. Ordinary least squares (OLS) regression lines are fit on the plots. (C) Bivariate plot of mean species RW1 score and mode species maxillary tooth count for individuals within 2.1 ≤ log(CS) ≤ 2.4. CV, coefficient of variation. Dashed lines are 95% prediction intervals (PIs).

Table 2.

Whole‐sample and species regressions of proposed ecomorphological variables against snout shape (RW1 score). Individuals with 2.1 ≤ log(CS) ≤ 2.4 were used for the analyses

Variable by snout shape (RW1 score)	n	R 2	P	Slope	Intercept	Pagel's λ
Whole‐sample analysis with ordinary least squares (OLS) regression
CV of maxillary alveolus diameter	111	0.668	< 0.001	−0.242	0.151	NA
Maxillary tooth spacing	111	0.860	< 0.001	−0.676	0.593	NA
Relative STF area [(STF area)1/2/CS]	123	0.315	< 0.001	0.069	0.134	NA
Relative PFD (PFD/CS)	120	0.376	< 0.001	−0.247	0.564	NA
Relative basioccipital tubera width (BTW/CS)	122	0.009	0.311	0.015	0.184	NA
Relative symphysis length (SL/CS)	110	0.667	< 0.001	2.233	0.567	NA
Species analysis with phylogenetic generalized least squares (PGLS) regression
CV of maxillary alveolus diameter	10	0.863	< 0.001	−0.232	0.155	0.000
Maxillary tooth spacing	10	0.972	< 0.001	−0.646	0.591	0.000
Mode maxillary tooth count	10	0.054	0.519	3.154	16.352	1.000
Relative STF area [(STF area)1/2/CS]	10	0.509	0.021	0.098	0.137	0.000
Relative PFD (PFD/CS)	10	0.634	0.006	−0.237	0.541	1.000
Relative basioccipital tubera width (BTW/CS)	10	0.130	0.306	0.053	0.187	0.000
Relative symphysis length (SL/CS)	10	0.676	0.004	1.476	0.835	1.000

CS, centroid size; CV, coefficient of variation; PFD, pterygoid flange depth; STF, supratemporal fenestra. Mean species RW1 scores and mean variable values were used for species analysis (PGLS regression). The molecular tree of Oaks (2011) was used for PGLS regression.

Growth trajectories of trophic ecomorphology

An observation of the bivariate plot of snout shape (RW1 score) and log(CS) (Fig. 4) revealed a universal trend of snout widening through growth. In most of the species, the snout shape is isometric or somewhat positively allometric in its width until reaching log(CS) of ~2.4, but beyond this size the snout grows increasingly wider. Although a rigorous comparison of species trajectories was not attempted because of the difficulty in comparing non‐linear trajectories, this pattern seems global and unrelated with interspecific snout shape differences. The result of the snout shape breakpoint estimation by piecewise linear model fitting is shown in Fig. 5 (Table S4). In this analysis, individuals with log(CS) ≥ 2.0 were used for excluding very small forms that might belong to relative skull growth phase I (Hall & Portier, 1994). The estimated log(CS) breakpoints of snout shape growth trajectories are 2.31–2.46 in six species (C. acutus, C. niloticus, C. palustris, C. porosus, G. gangeticus and T. schlegelii), and their mean is 2.40. Individuals with log(CS) smaller and larger than the mean snout shape breakpoint were considered as members of growth phase II and III (Hall & Portier, 1994), respectively, and Welch's t‐test confirmed that snout shapes are significantly different between two phases (Table 3).

Figure 5.

Log(CS) breakpoints of snout shape growth trajectories in six species estimated by piecewise linear model fitting. Individuals with log(CS) ≥ 2.0 were used for this analysis. Dots are the estimated values, and bars represent their 95% confidence intervals (CIs).

Table 3.

Welch's t‐test of snout shape (RW1 score) between skull growth phases II and III. Individuals with log(CS) ≥ 2.0 were used for the analysis

Taxon	n	t	df	P
C. acutus	31	−3.591	5.958	0.012
C. niloticus	17	−5.777	13.607	< 0.001
C. palustris	21	−3.819	2.665	0.039
C. porosus	23	−12.821	5.068	< 0.001
G. gangeticus	16	−5.696	12.045	< 0.001
T. schlegelii	25	−4.821	17.774	< 0.001

Bivariate plots of log(CS) and trophic ecomorphological variables are provided in Fig. 6, and a summary of the RMA regression between ecomorphological variables and log(CS) for whole samples is found in Table 4. Maxillary tooth size variability is significantly (P < 0.05) and near significantly (P < 0.1) correlated with size in seven species, excluding G. gangeticus, Crocodylus johnsoni and M. cataphractus, all showing a positive trend through growth. Maxillary tooth spacing is significantly and near significantly correlated with size in six species, and all of which show a positive trend. A strong positive correlation (R ²> 0.6) is found in two most longirostral taxa (G. gangeticus and T. schlegelii), indicating a significant decrease in the relative spacing of maxillary teeth through growth. The relationship of relative STF area is negatively allometric (95% CI < 2) in four species, isometric (95% CI includes 2) in five species, and positively allometric (95% CI > 2) in G. gangeticus. The slopes for relative PFD are positively allometric for all taxa, and those for relative basioccipital tubera widths vary among 10 species. The relationship of relative symphysis length is positively allometric and isometric in nine species, but negatively allometric in G. gangeticus.

Figure 6.

(A–G) Bivariate plots of log(CS) and ecomorphological variables. Reduced major axis (RMA) regression lines are fit on the plots (A,B,D–G).

Table 4.

RMA regressions of ecomorphological variables against size [log(CS)]

Taxon	n	R 2	P	Slope	Lower 95% CI	Upper 95% CI	Allometric trend
CV of maxillary alveolous diameter
C. acutus	29	0.401	< 0.001	0.203	NA	NA	NA
C. intermedius	11	0.329	0.065	0.313	NA	NA	NA
C. johnsoni	17	0.082	0.264	0.296	NA	NA	NA
C. niloticus	15	0.221	0.077	0.195	NA	NA	NA
C. novaeguineae	17	0.303	0.022	0.158	NA	NA	NA
C. palustris	19	0.296	0.016	0.189	NA	NA	NA
C. porosus	23	0.503	< 0.001	0.188	NA	NA	NA
G. gangeticus	14	0.030	0.554	−0.086	NA	NA	NA
M. cataphractus	31	0.079	0.127	0.122	NA	NA	NA
T. schlegelii	22	0.690	< 0.001	0.279	NA	NA	NA
Maxillary tooth spacing
C. acutus	29	0.198	0.015	0.399	NA	NA	NA
C. intermedius	11	0.000	0.982	0.322	NA	NA	NA
C. johnsoni	17	0.078	0.279	0.524	NA	NA	NA
C. niloticus	15	0.551	0.002	0.361	NA	NA	NA
C. novaeguineae	17	0.000	0.993	0.273	NA	NA	NA
C. palustris	19	0.018	0.584	0.343	NA	NA	NA
C. porosus	23	0.211	0.027	0.294	NA	NA	NA
G. gangeticus	14	0.875	< 0.001	0.592	NA	NA	NA
M. cataphractus	31	0.103	0.079	0.248	NA	NA	NA
T. schlegelii	22	0.606	< 0.001	0.557	NA	NA	NA
log(STF area)
C. acutus	33	0.958	< 0.001	1.895	1.758	2.043	=
C. intermedius	12	0.981	< 0.001	1.852	1.682	2.039	=
C. johnsoni	16	0.947	< 0.001	2.044	1.792	2.332	=
C. niloticus	19	0.964	< 0.001	1.659	1.506	1.828	−
C. novaeguineae	19	0.954	< 0.001	1.799	1.613	2.007	=
C. palustris	22	0.969	< 0.001	1.816	1.673	1.971	−
C. porosus	30	0.992	< 0.001	1.916	1.850	1.984	−
G. gangeticus	18	0.991	< 0.001	2.118	2.015	2.225	+
M. cataphractus	34	0.978	< 0.001	1.877	1.779	1.981	−
T. schlegelii	24	0.977	< 0.001	1.955	1.828	2.090	=
log(PFD)
C. acutus	31	0.993	< 0.001	1.231	1.193	1.270	+
C. intermedius	12	0.996	< 0.001	1.267	1.209	1.328	+
C. johnsoni	17	0.977	< 0.001	1.222	1.125	1.327	+
C. niloticus	18	0.994	< 0.001	1.241	1.193	1.292	+
C. novaeguineae	18	0.998	< 0.001	1.225	1.193	1.258	+
C. palustris	20	0.994	< 0.001	1.236	1.190	1.283	+
C. porosus	30	0.997	< 0.001	1.261	1.236	1.286	+
G. gangeticus	18	0.992	< 0.001	1.221	1.164	1.281	+
M. cataphractus	38	0.989	< 0.001	1.191	1.149	1.234	+
T. schlegelii	27	0.994	< 0.001	1.215	1.175	1.255	+
log(basioccipital tubera width)
C. acutus	31	0.980	< 0.001	1.151	1.091	1.215	+
C. intermedius	12	0.962	< 0.001	1.063	0.927	1.220	=
C. johnsoni	17	0.840	< 0.001	0.990	0.796	1.231	=
C. niloticus	18	0.973	< 0.001	1.147	1.052	1.251	+
C. novaeguineae	19	0.950	< 0.001	0.878	0.783	0.985	−
C. palustris	20	0.920	< 0.001	0.977	0.850	1.123	=
C. porosus	30	0.978	< 0.001	1.060	1.001	1.122	+
G. gangeticus	17	0.986	< 0.001	1.176	1.102	1.256	+
M. cataphractus	38	0.954	< 0.001	1.055	0.981	1.134	=
T. schlegelii	27	0.982	< 0.001	1.173	1.111	1.239	+
log(symphysis length)
C. acutus	30	0.985	< 0.001	1.186	1.130	1.244	+
C. intermedius	12	0.970	< 0.001	1.061	0.939	1.198	=
C. johnsoni	17	0.897	< 0.001	1.112	0.933	1.325	=
C. niloticus	18	0.992	< 0.001	1.070	1.022	1.121	+
C. novaeguineae	13	0.989	< 0.001	1.200	1.118	1.287	+
C. palustris	20	0.992	< 0.001	1.102	1.053	1.153	+
C. porosus	24	0.995	< 0.001	1.195	1.157	1.234	+
G. gangeticus	17	0.991	< 0.001	0.825	0.784	0.868	−
M. cataphractus	37	0.976	< 0.001	1.149	1.089	1.212	+
T. schlegelii	24	0.968	< 0.001	0.936	0.864	1.013	=

CV, coefficient of variation; PFD, pterygoid flange depth; STF, supratemporal fenestra.

+, positive allometry; −, negative allometry; =, isometry.

The maxillary tooth count is stable through growth in all sampled taxa (Fig. 6). In nine out of 10 species, maxillary tooth count varies within each species, but it is not significantly correlated with size, as confirmed by Spearman rank correlation (Table 5).

Table 5.

Spearman rank correlation of maxillary tooth count and size [log(CS)]

Taxon	n (right)	n (left)	Mode	Mean	Deviation from mode (total)	Mode tooth count frequency (%) (total)	Spearman ρ (right)	P (right)	Spearman ρ (left)	P (left)
C. acutus	33	33	14	13.94	4	93.9	−0.147	0.415	−0.187	0.298
C. intermedius	12	12	14	13.92	2	91.7	−0.194	0.545	NA	NA
C. johnsoni	17	17	14	14.18	6	82.4	0.085	0.746	0.112	0.669
C. niloticus	19	19	13	13.21	10	73.7	0.358	0.133	0.589	0.008
C. novaeguineae	19	19	14	14.00	0	100.0	NA	NA	NA	NA
C. palustris	20	20	14	13.98	1	97.5	NA	NA	−0.099	0.677
C. porosus	30	30	14	13.98	1	98.3	−0.225	0.231	NA	NA
G. gangeticus	19	19	24	23.74	10	73.7	0.269	0.266	0.165	0.500
M. cataphractus	38	38	13	13.01	1	98.7	−0.037	0.823	NA	NA
T. schlegelii	27	27	16	15.98	1	98.1	NA	NA	−0.302	0.126

Discussion

Non‐alligatoroid crocodylian trophic ecomorphology and the uniqueness of G. gangeticus

The crocodylian skulls are featured by the high level of character convergence that is thought to reflect trophic convergence (Brochu, 2001; Sadleir & Makovicky, 2008). The non‐phylogenetic and phylogenetic methods in testing the correlations between the proposed trophic ecomorphology and snout shape indicated that with narrowing of the snout, maxillary tooth size disparity increases (lateral maxillary margin more festooned), maxillary teeth less packed, STF enlarged, pterygoid flange becomes shallow and lower jaw symphysis extended. These characters are intercorrelated with each other, and supposedly transformed concurrently in the trophic specialization of non‐alligatoroid crocodylians. Other variables (tooth count and relative basioccipital tubera width) yielded non‐significant correlation with snout shape. However, it is possible that the character transformation of those two variables is stepwise, and increase in tooth count and enlargement of basioccipital tubera occur once snout is narrowed to a certain degree, perhaps as much as G. gangeticus, the most predominant fish‐eater in extant crocodylians. Previous studies demonstrated that skull of G. gangeticus is known to occupy unique morphospace among extant adult crocodylians (Pierce et al. 2008; Sadleir, 2009), and its developmental pathway begins and ends in a distinct region of the morphospace (Piras et al. 2010). Its bite force performance is also different from other species, because G. gangeticus falls outside the 95% confidence interval of the OLS regression of bite force and size (Erickson et al. 2012). The current result is consistent with the previous notion that G. gangeticus is morphologically and functionally atypical. In the whole‐sample OLS regressions of ecomorphological variables against snout shape (RW1 score), G. gangeticus is placed outside of the 95% prediction interval for relative STF area, relative symphysis length and relative PFD among the verified trophic ecomorphology (Fig. 3). These suggest that specialization for true piscivory involves the deviation from the conserved transformation axis of crocodylian skull characters. However, the results of the current analyses should be interpreted carefully. Inclusion of extant alligatoroids and some fossil forms, especially with relatively narrow snouts such as Caiman crocodilus apaporiensis, Orthogenysuchus (Mook, 1924) and Procaimanoidea (Gilmore, 1946) in the analyses may have a significant impact on the results.

Until now, adaptation for feeding on small aquatic prey has been suggested for many taxa of large aquatic reptiles, including thalattosuchian (Langston, 1973; Massare, 1987; Hua, 1994; Young et al. 2010) and tethysuchian (Denton et al. 1997; Sereno et al. 2001; Hastings et al. 2011; Schwarz‐Wings, 2014), crocodyliforms, phytosaurs (Hunt, 1989; Renesto & Paganoni, 1998) and choristoderans (Erickson, 1972; Evans & Hecht, 1993), mostly on the grounds of long and narrow snouts as well as slender and pointed teeth. Establishment of trophic ecomorphology in crocodylians, and a close comparison of above‐listed forms with crocodylians will reveal a degree to which the adaptation to similar ecology is driven by convergence among large aquatic reptiles.

Comparative snout shape ontogeny

Using 2D geometric morphometrics of palate shapes, I demonstrated that the snout widening is a universal trend in non‐alligatoroid crocodylians, regardless of the interspecific snout shape differences. Moreover, growth trajectories of sampled crocodylians are non‐linear, and beyond some size, the snout grows increasingly wider. This is evident when the trajectories of extremely slender‐snouted taxa (G. gangeticus and T. schlegelii), moderately slender‐snouted taxa (e.g. Crocodylus intermedius) and generalist‐type taxa (e.g. C. porosus and C. niloticus) are compared (Fig. 4). Statistical tests supported that the mean snout shape threshold in six species is log(CS) of 2.40, and snout shape is significantly wider beyond the threshold in all six species (Fig. 7). The log(CS) of post‐rostrum landmarks is strongly correlated with log(skull width) for all taxa [n = 236, log(skull width) = 1.185 log(CS) − 0.459, R ²= 0.993], and the skull width predicted from the mean snout shape threshold [2.40 log(CS)] is 242 mm, with the lowest and highest estimate values of six species range 191–287 mm. In other words, extant non‐alligatoroid crocodylians begin significant snout shape changes when their skull widths are about 242 mm during growth. Meanwhile, larger variance in the snout shape of C. acutus might be derived from greater geographical variation, because C. acutus samples were collected from Florida (USA), Mexico, and Caribbean, Central American and Northern South American countries.

Figure 7.

Significant snout shape changes across the skull growth phases II and III. (A) C. acutus (from top: FMNH 20159; FMNH 59070). (B) C. intermedius (from top: FMNH 75659; SMF 28139). (C) C. niloticus (from top: IRSNB 5007; IRSNB 5012). (D) C. palustris (from top: BMNH 1848.2.5.9; BMNH 1868.4.9.11). (E) C. porosus (from top: FMNH 63744; FMNH 14071). (F) G. gangeticus (from top: YPM R10514; ZSM 28.1912). (G) T. schlegelii (from top: BMNH 1899.1.31.1; ZSM 395.1907). Ca, C. acutus; Ci, C. intermedius; Cni, C. niloticus; Cpa, C. palustris; Cpo, C. porosus; Gg, G. gangeticus; Ts, T. schlegelii. All scale bars are 10 cm.

The results are partly congruent with previous studies on comparative skull shape allometry (Piras et al. 2010; Watanabe & Slice, 2014), yet differ in showing a universal trend of snout widening in non‐alligatoroid crocodylians. A non‐linearity of the snout growth in some species of crocodylians was mentioned by Hall & Portier (1994) and Watanabe & Slice (2014). Both studies described U‐shaped trajectories, in which snout narrowing occurs in early stages, followed by snout widening in late stages. Hall & Portier (1994) defined three general growth phases of skulls in C. novaeguineae, each characterized by snout narrowing (phase I: ≤ 60–70 mm skull length), stationary snout shape (phase II: 60–70 mm ≤ skull length ≤ 350 mm) and snout widening (phase III: ≥ 350 mm skull length), respectively. Although the boundaries of the three growth phases are ambiguous, they hypothesized that the phase transitions are concomitant with dietary shifts during ontogeny. This idea was examined in C. niloticus using stable carbon and nitrogen isotope analyses of scute keratin, finding some support for the correlated changes of skull morphometrics and the isotopic ratios (Radloff et al. 2012). The estimated snout shape breakpoint in the current samples (191 mm ≤ skull width ≤ 287 mm) is close to Hall & Portier's (1994) phase II–III boundary (estimated skull width of 188 mm), which corresponds 350 mm skull length in the current C. novaeguineae samples. However, because the pattern of snout shape change seems homogeneous among sampled taxa that display various dietary patterns, distinct growth phases of the skull imply the existence of a shared size‐dependent biomechanical constraint, rather than the morphological response to the dietary shift in non‐alligatoroid crocodylians.

The significant skull shape change in very large size classes of crocodylians has been known (Mook, 1921b; Müller, 1927; Kälin, 1933), and previous studies on extant crocodylians have been concerned with the effect of ontogenetic allometry in species or larger group comparisons of skull shapes (Pierce et al. 2008; Sadleir & Makovicky, 2008; Piras et al. 2009, 2014). However, perhaps more susceptible to allometry are analyses of fossil forms, because most of the species are represented by single individuals, and their growth patterns and maximum sizes are not known. If the size‐related shape change is large and samples from different size classes are used simultaneously without size correction, ecological diversity inferences such as morphological disparity metrics are spuriously calculated, and higher level analyses of disparity integrating time, phylogeny and geography are confounded with various allometric effects. Figure 8 illustrates the effect of size range on the rostrum shape disparity of crocodylians using the empirical dataset, by randomly sampling 10 individuals of each 10 species from a series of size ranges. While the mean variance seems unrelated with size range, 95% CI of variance is more than doubled with the increase in the size range. A significant widening of the 95% CI of snout shape variance occurs where log(CS) > 2.4, which corresponds to the skull growth phase III. Given that most of the skull shape variation is related with rostrum shape in extant crocodylians (Pierce et al. 2008), analysis of skull shape disparity of fossil taxa (Young et al. 2010; Wilberg, 2012; Hastings & Hellmund, 2017) may need to limit the size range of samples.

Figure 8.

Effect of size ranges on snout shape disparity tested by the empirical dataset. Ten individuals from each 10 species were randomly subsampled (10 000 replicates) without replacement, and the snout shape (RW1 scores) variance of 10 species was calculated for series of size ranges (2.0 ≤ log(CS) ≤ x, 2.30 ≤ x ≤ 2.60), where x sequentially changes by 0.01. The means and 95% confidence intervals of the variance for the series of size ranges were shown by a line chart. The subsampling and variance calculation were performed using the software r.

Allometry of trophic ecomorphology in non‐alligatoroid crocodylians

As indicated by regression slopes of ecomorphological variables against log(CS), the allometric patterns of the variables differ among species, possibly related to trophic specialization during growth. The relative maxillary tooth size variation increases with size in seven species, yet the variation remains smaller throughout life in G. gangeticus, and possibly in C. johnsoni and M. cataphractus (Fig. 6; Table 4). Accordingly, the lateral maxillary margin is more ‘festooned’ in adults of non‐Gavialis taxa, including T. schlegelii (Fig. 9). The increased degree of tooth size disparity in large adults of non‐Gavialis taxa is most likely advantageous for holding larger prey firmly (Iordansky, 1973). On the other hand, G. gangeticus is arguably a predominant fish‐eater, although they occasionally feed on prey other than fish (Whitaker & Basu, 1982; Trutnau & Sommerlad, 2006). Thus, it is intuitive that they maintain near‐homodont dentition through growth. While most of the sampled taxa acquire more heterodont dentition toward adults, interalveolar space of maxillary dentition decreases during growth in six species, in which two longirostrine taxa (G. gangeticus and T. schlegelii) exhibit the most significant trends (Fig. 6; Table 4). These, combined with constant tooth counts through growth, imply that at least some of the maxillary teeth exhibit positive allometric growth relative to the maxillary tooth row length in non‐alligatoroid crocodylians. As expected, a qualitative observation revealed that teeth positioned near the top of the convexities of the maxillary margin grow with positive allometry (Fig. 9). A previous study agrees with this observation by showing positive allometry in the diameter of ‘caniniform’ tooth relative to tooth row length (maxillary tooth #5 in C. porosus and #4 in Alligator mississippiensis), which is positioned at the apex of the ‘first wave’ of maxillary tooth row (Brown et al. 2015).

Figure 9.

Comparative growth of snout shape (right palate) in sampled crocodylians. (A) C. palustris (from top: ZSM 523.1911; ZSM 542.1911; BMNH 1897.12.31.1). (B) C. acutus (from top: USNM 52336; FMNH 11038; USNM 268760). (C) T. schlegelii (from top: ZSM 202.1907; SMF 28135; ZSM 375.1907). (D) G. gangeticus (from top: BMNH 46.1.7.3; ZSM 29.1912; BMNH 1974.3009). All scale bars are 5 cm.

Relative STF area and PFD are associated with the relative mass of jaw adductor muscles. Growth trends for relative STF area are isometric or negatively allometric in most of the sampled taxa, which agree with the STF scaling in A. mississippiensis (Dodson, 1975), but only G. gangeticus exhibits a positive trend (Fig. 6; Table 4). Because enlarged STF area is correlated with the hypertrophied Musculus pseudotemporalis, which helps weak but rapid jaw closure (Iordansky, 1964; Endo et al. 2002) and hence is advantageous for catching agile prey, G. gangeticus might be specialized for fish eating through growth. By contrast, all taxa show a positive allometric trend for relative PFD (Figs 6 and 10; Table 4), as suggested in A. mississippiensis (Dodson, 1975), which indicates that relative mass of Mm. pterygoidei dorsalis et ventralis, strong jaw adductor muscles (Iordansky, 1964), increases through growth in all sampled species. This is consistent with the study of adductor muscle mass and physiological cross‐sectional areas (PCSA) scaling in A. mississippiensis (Gignac & Erickson, 2016). This study demonstrated that the mass and PCSA of both dorsal and ventral pterygoid muscles are positively allometric, in which the ventral one shows greatest relative growth of all adductor muscles. Moreover, experimental studies of bite force scaling among Crocodylia showed that a positive allometry of bite force is conserved among extant crocodylians regardless of phylogenetic relatedness and snout shape (Erickson et al. 2003, 2014). The other aspect of relative ventral extension of pterygoid flange is its correlation with maximal mouth opening (Piras et al. 2014). It could be possible that crocodylians, regardless of their snout shapes, achieve greater maximal mouth opening during growth.

Figure 10.

Comparison of occipital part of skulls in sampled crocodylians. (A) C. palustris (from left: ZSM 523.1911; ZSM 517.1911; BMNH 1897.12.31.1). (B) C. acutus (from left: AMNH 15182; AMNH 43299; USNM 268760). (C) T. schlegelii (from left: ZSM 202.1907; SMF 28135; ZSM 375.1907). (D) G. gangeticus (from left: BMNH 1896.7.7.4; ZSM 29.1912; AMNH 7138). All scale bars are 5 cm.

Growth patterns of relative lower jaw symphysis length show ontogenetic convergence through growth (Fig. 6; Table 4), where patterns of G. gangeticus and T. schlegelii are negatively allometric (95% CI < 1) or almost negatively allometric (upper 95% CI slightly exceeds 1), respectively, whereas that of other species are isometric or positively allometric. Because shorter symphysis incurs lower strain under equivalent loads (Walmsley et al. 2013), slender‐snouted taxa such as G. gangeticus and T. schlegelii might be able to tolerate higher relative loads through growth. Overall, it is conceivable that non‐alligatoroid crocodylian skull characters react differently towards the trophic specialization: some of the characters respond to the change in prey preference during growth, yet other characters are in keeping with the conserved relationship of maximum bite force and body size among crocodylians (Erickson et al. 2012, 2014).

Taxonomic implication

The occlusal pattern has been traditionally used for the classification of two large groups of crocodylians, Alligatoroidea and Crocodyloidea: maxillary teeth of living alligatoroids occlude buccal to dentary teeth, but maxillary and dentary teeth of living crocodyloids interfinger. While fossils suggested the presence of the intermediate character state, and the character transformation history has been understood (Brochu, 1999, 2003), the occlusal pattern is still recognized as a popular identification key for crocodylians in academic and public scenes. The observation of the ontogenetic series of non‐alligatoroid crocodylians revealed that the occlusal pattern is continuous in nature rather than discrete, and is strongly associated with growth stages and snout shape.

As the consequence of increased tooth size disparity and positive allometry of teeth at maxillary margin convexity through ontogeny, the first five maxillary teeth and last five–six teeth are densely packed in large adults of generalist‐type taxa (e.g. C. acutus; C. niloticus; C. palustris). This is also a characteristic of large adults of slender‐snouted taxa, such as C. intermedius and T. schlegelii (Fig. 9). By contrast, in all taxa maxillary teeth #6–8 are widely spaced even in large adults, to occlude with large dentary teeth that are positioned at a convex margin of the dentary. The ‘packed’ anterior and posterior maxillary teeth lead to an unavoidable result: overbite occlusion at anterior and posterior maxilla. This is clear from the observation of live individuals (M. I. personal observation), and occlusal pits occasionally left on palates, that are placed lingual to maxillary alveoli, even in slender‐snouted taxa like T. schegelii (Fig. 11A). Yet still, in‐line occlusion is maintained at the concavity of maxillary margin (maxillary tooth # 6–8). The current study also suggested that degree of heterodonty and spacing of maxillary teeth are strongly correlated with snout shape: in taxa with wider snout, maxillary teeth are more heterodont and more packed. Considering these, it can be supposed that short‐snouted taxa like C. palustris are prone to exhibit ‘semi‐overbite’ (overbite at anterior and posterior maxilla) occlusion, and achieve it in younger age (with smaller size) compared with slender‐snouted taxa. It might also be worthwhile to mention that a few specimens of P. palustris (e.g. BMNH 1848.2.5.9; BMNH 1852.5.9.45; ZSM 565/1911) examined here bear a pit between premaxilla and maxilla for reception of the fourth dentary tooth (Fig. 12), although its lateral wall is not fully formed. Formation of such pit is often associated with overbite occlusion in crocodylians (Brochu, 2003). When fossils are taken into account, the acquisition of ‘semi’ and ‘full’ overbite occlusion of maxillary and dentary teeth are common in stem Crocodylidae (Brochu, 1997, 1999), also occurred repeatedly among Crocodylidae, mainly in large specimens. In subfossil remains of osteolaemine Voay robustus from Madagascar (Mook, 1921a; Brochu, 2007), and the Pleistocene remain of Crocodylus rhombifer from Cuba (Varona, 1984), the interalveolar space of maxillary teeth # 6–8 is much narrower than that in C. palustris. As a result, occlusal pits are placed lingually to maxillary teeth # 6–8, achieving a near full overbite occlusion (Fig. 11B,C), while in smaller individuals of V. robustus, more in‐line occlusion is recognized (Mook, 1921a; Brochu, 2007). Likewise, a various degree of overbite occlusion is observed in mekosuchine crocodiles (Willis et al. 1990; Megirian et al. 1991; Willis, 1993), probably caused by densely packed teeth. Although further studies are necessary to understand the interaction of the occlusal pattern with snout shape, tooth spacing and tooth count, there seem to be a few barriers in the transition of ‘interfingering’ and ‘overbite’ patterns in crocodyloids, mostly in short‐snouted forms.

Figure 11.

Occlusal pits left on palates. (A) T. schlegelii (BMNH 94.2.21.1). (B) Voay robustus (AMNH 3101). (C) Crocodylus rhombifer (AMNH 6181). Arrows indicate occlusal pits placed medial to maxillary tooth row. M5, maxillary alveolus #5; SOF, suborbital fenestra. All scale bars are 10 cm.

Figure 12.

A pit between premaxilla and maxilla for reception of the fourth dentary tooth (C. palustris: BMNH 1852.5.9.45). The arrow indicates the pit. Scale bar: 5 cm.

Taxonomic characters that are related with trophic ecomorphology can be found in other parts of the skull. Associated with the ventral extension of the pterygoid flange is the extent to which basisphenoid exposes below median Eustachian opening. In the sampled taxa, it is longest in C. palustris (Fig. 13), the most wide‐snouted taxon in genus Crocodylus, and shortest in slender‐snouted G. gangeticus and T. schlegelii. Consequently, the posterior pterygoid process is also tallest in C. palustris (Fig. 13), as opposed to the shortest condition in G. gangeticus and T. schlegelii. These characters (chars. #98 and #119 in Brochu, 1999) are highly variable even within the identical character states, and rather correlated with the PFD, and hence with the snout shape. The problem of ecomorphological character correlation has been debated in morphological phylogenetic analyses of crocodylians (Gatesy et al. 2003; Sadleir & Makovicky, 2008). Although improved taxon and character sampling is key for the better phylogenetic reconstruction, and rooting and the choice of outgroups have a great impact on the placement of ingroup taxa (Wilberg, 2015), recognition of ecomorphological characters and their inter‐correlations, and evaluation of such characters in datasets will be of further help in accurately reconstructing the phylogenetic relationships of crocodylians.

Figure 13.

Basioccipital region of the skull in C. palustris (BMNH 1897.12.31.1). Scale bar: 3 cm.

Supporting information

Fig. S1. A time‐calibrated morphological phylogeny of 10 sampled crocodylian species.

Table S1. Specimens, locality information and raw data used in the current study.

Table S2. Measurements of mesiodistal diameters of all right/left maxillary alveoli.

Table S3. Phylogenetic generalized least squares (PGLS) regressions of proposed ecomorphological variables against snout shape (RW1 score), using an alternative (morphological) tree (Fig. S1).

Table S4. Complete results of piecewise linear model fitting for the relationship of snout shape (RW1 score) and size [log(CS)].

Appendix S1. Literature cited.