Bayesian divergence time analyses were used to simultaneously infer the phylogenetic relationships and date the major clades of snakes including several important fossils that have not previously been included in divergence dating analyses as terminal taxa. We also explored the effect of using fossilized birth–death (FBD) and uniform tree priors for divergence dating with terminal calibrations. Nonclock and relaxed clock analyses of the combined morphology and molecular data set supported previous molecular phylogenetic hypotheses for the major clades of snakes, including the paraphyly of the traditionally recognized Scolecophidia and Macrostomata. Tip-dating analyses using either a uniform tree prior or FBD prior that assume that all fossils are tips and that extant lineages are randomly sampled resulted in older ages than those inferred using a FBD prior assuming diversified sampling of extant lineages and those estimated by previous studies. We used Bayesian ancestral state reconstruction methods to map the evolution of the ability to consume large prey and the loss of limbs onto our inferred time-calibrated phylogeny. We found strong support for early evolution of the ability to consume large prey, indicating multiple independent losses of this ability. We also found strong support for retention of external hindlimbs until relatively late in snake evolution, indicating multiple independent losses of hindlimbs.

#### Fig. S1. Limb reconstruction on uniform clock phylogeny

Fig. S1. Ancestral state reconstructions from MultiState in BayesTraits showing the probability of fully-developed hind limbs (blue), reduced limbs (yellow), and completely lost limbs (red) for each node as well as the states coded for terminal taxa on the uniform clock phylogeny. As in Fig. 4, white areas of pies at internal nodes indicate the posterior probability that a node does not exist.

Fig_S1_limbs_uniform_clock.pdf

#### Fig. S2. Macrostomatan phenotype reconstruction on uniform clock phylogeny

Fig. S2. Ancestral state reconstructions from MultiState in BayesTraits showing the probability of the non-macrostomatan phenotype (blue) and macrostomatan phenotype (yellow) for each node as well as the states coded for terminal taxa on the uniform clock phylogeny. As in Fig. 4, white areas of pies at internal nodes indicate the posterior probability that a node does not exist.

Fig_S2_Gape_uniform_clock.pdf

#### Fig. S3. Limb reconstruction on constrained, fossil tip FBD phylogeny

Fig. S3. Ancestral state reconstructions from MultiState in BayesTraits showing the probability of fully-developed hind limbs (blue), reduced limbs (yellow), and completely lost limbs (red) for each node as well as the states coded for terminal taxa on the constrained, fossil tip FBD phylogeny. As in Fig. 4, white areas of pies at internal nodes indicate the posterior probability that a node does not exist.

Fig_S3_limbs_constrained_foss_tip_FBD.pdf.pdf

#### Fig. S4. Macrostomatan phenotype reconstruction on constrained, fossil tip FBD phylogeny

Fig. S4. Ancestral state reconstructions from MultiState in BayesTraits showing the probability of the non-macrostomatan phenotype (blue) and macrostomatan phenotype (yellow) for each node as well as the states coded for terminal taxa on the constrained, fossil tip FBD phylogeny. As in Fig. 4, white areas of pies at internal nodes indicate the posterior probability that a node does not exist.

Fig_S4_Gape_constrained_foss_tip_FBD.pdf

#### Fig. S5. Limb reconstruction on constrained, diversified FBD Tetrapodophis excluded phylogeny.

Fig. S5. Ancestral state reconstructions from MultiState in BayesTraits showing the probability of fully-developed hind limbs (blue), reduced limbs (yellow), and completely lost limbs (red) for each node as well as the states coded for terminal taxa on the constrained, diversified FBD Tetrapodophis excluded phylogeny. As in Fig. 4, white areas of pies at internal nodes indicate the posterior probability that a node does not exist. The phylogeny is plotted without Diablophis or Portugalophis because of issues visualizing branch lengths when these taxa were included, but they were included in the trees used for ancestral state reconstruction.

Fig_S5_Limbs_no_Tetrapodophis.pdf

#### Fig. S6. Macrostomatan phenotype reconstruction on constrained, diversified FBD Tetrapodophis excluded phylogeny

Fig. S6. Ancestral state reconstructions from MultiState in BayesTraits showing the probability of the non-macrostomatan phenotype (blue) and macrostomatan phenotype (yellow) for each node as well as the states coded for terminal taxa on the constrained, diversified FBD Tetrapodophis excluded phylogeny. As in Fig. 4, white areas of pies at internal nodes indicate the posterior probability that a node does not exist. The phylogeny is plotted without Diablophis or Portugalophis because of issues visualizing branch lengths when these taxa were included, but they were included in the trees used for ancestral state reconstruction.

Fig_S6_Gape_no_Tetrapodophis.pdf

#### Fig. S7. Limb reconstruction on constrained, diversified FBD Tetrapodophis and Caldwell et al. (2015) fossils excluded phylogeny

Fig. S7. Ancestral state reconstructions from MultiState in BayesTraits showing the probability of fully-developed hind limbs (blue), reduced limbs (yellow), and completely lost limbs (red) for each node as well as the states coded for terminal taxa on the constrained, diversified FBD Tetrapodophis and Caldwell et al. (2015) fossils excluded phylogeny. As in Fig. 4, white areas of pies at internal nodes indicate the posterior probability that a node does not exist.

Fig_S7_Limbs_no_Tetrapodophis_or_Caldwell_etal2015.pdf

#### Fig. S8. Macrostomatan phenotype reconstruction on constrained, diversified FBD Tetrapodophis and Caldwell et al. (2015) fossils excluded phylogeny

Fig. S8. Ancestral state reconstructions from MultiState in BayesTraits showing the probability of the non-macrostomatan phenotype (blue) and macrostomatan phenotype (yellow) for each node as well as the states coded for terminal taxa on the constrained, diversified FBD Tetrapodophis and Caldwell et al. (2015) fossils excluded phylogeny. As in Fig. 4, white areas of pies at internal nodes indicate the posterior probability that a node does not exist.

Fig_S8_Gape_no_Tetrapodophis_or_Caldwell_etal2015.pdf

#### Fig. S9. Macrostomatan phenotype reconstruction on constrained, diversified FBD phylogeny with madstoiids coded uncertain

Fig. S9. Ancestral state reconstructions from MultiState in BayesTraits showing the probability of the non-macrostomatan phenotype (blue) and macrostomatan phenotype (yellow) for each node as well as the states coded for terminal taxa. As in Fig. 4, white areas of pies at internal nodes indicate the posterior probability that a node does not exist. Madstoiids were coded as uncertain in this analysis and the analysis was performed on the phylogeny from the constrained, diversified FBD analysis.

Fig_S9_Gape_reconstruction_Madtsoiidae_uncertain.pdf

#### Files S1-S17. Supplemental tree files.

Files S1-S17. This zip file contains all of the supplemental tree files. Individual tree files are described in the ReadMe file and in a text file within the zip file.

Files_S1-S17_tree_files.zip

#### Files S18-S23. Zip file containing data matrices and MrBayes blocks

Files S18-S23. This zip file contains all of the data matrices and MrBayes blocks used for analyses. Individual files are described in the ReadMe file and in a text file within the zip file.

Files_S18-S23_matrices_MrBayes_blocks.zip

#### Table S1. Node calibrations included in divergence dating analyses.

Table_S1_calibrations.xlsx