Analyses
We used phylogenetic comparative methods, using a tree pruned from Pyron & Burbrink (2014). If species were not present in this tree, we used the closest relative (Supplementary material 2). We tested whether tooth curvature length, the mean degree of curvature, and the maximum degree of curvature were correlated with one another using Pearson correlation tests (function cor.test from the ‘stats’ package). We then tested whether these measurements were associated with our factors using the function phylANOVA of the package ‘phytools’ using 1000 simulations and a Holm correction for the post hoc pairwise t-tests. The normality of the data distribution was checked using a Shapiro test and variables were log-transformed to ensure normality when necessary. We used the function procD.pgls in ‘geomorph’ (Adams, 2014) to run phylogenetic ANCOVAs and ANOVAs to test the link between tooth shape and our different dietary constraints. For ANCOVAs, we used the log-transformed centroid size of the tooth as a covariate. Since our predictive factors are not entirely independent from one another, we performed one model per factor. We compared the fit of our different models using the function model.comparison in the ‘RRPP’ package (Collyer & Adams, 2022), that calculated the log-likelihood of each model. Statistical significance was tested by performing 10000 permutations. We performed subsequent post-hoc pairwise comparisons using the pairwise function in ‘RRPP’ on the best models. All geometric morphometric, statistical analyses and visualizations were performed in R version 3.4.4 (R Core Team, 2018) (R code and data available in Supplementary Material), except the landmark acquisition performed in MorphoDig (Lebrun, 2017). To highlight shape differences between ecological groups, we used mesh deformation of the template specimen toward the mean landmark configuration for each group, using the plotRefToTarget function of the ‘geomorph’ package. We also used the function spheres3d from the ‘rgl’ package that allows to visualize both the inner and outer surfaces of the teeth, providing information on teeth thickness (Fig. 5).