Life reconstruction of enantiornithine birds feeding. Longipteryx (left), Bohaiornis (centre), and Pengornis (right) are pictured in the Early Cretaceous forests of northeastern China, roughly 120 million years ago. Bohaiornis is depicted feeding on cypress (Cupressaceae, [150]) leaves after the findings in this work. Longipteryx is depicted feeding on the mayfly Epicharmeropsis hexavenulosus [151] after [8]. Pengornis is depicted feeding on the fish Lycoptera davidi [152] after [9].

Bohaiornithid skull reconstructions used for MA and FEA calculations in this study. Reconstructions are of Bohaiornis (A), Longusunguis (B), Parabohaiornis (C), Shenqiornis (D), Sulcavis (E), and Zhouornis (F). Different colours indicate elements taken from different individual specimens. All sclerotic rings are based on Longipteryx specimen BMNHC Ph-930B. See the Methods section for more details on reconstruction. Scale for each reconstruction is based on the individual which makes up the largest portion of the reconstruction.

Masses for bohaiornithid taxa based on the regression equations of [26]. Most masses were previously reported in [2], though masses for the juvenile bohaiornithid CUGB P1202 [13], Beiguornis khinganensis MHGU-F307/8 [22], and cf. Sulcavis BMNHC-Ph1204 [21] are newly-calculated in this study from literature images.

Posterior probabilities predicting bohaiornithid diet by FDA from MA and functional indices of extant bird jaws. Values with green backgrounds are more likely, values with red backgrounds are less likely. All bohaiornithids have high affinity with generalists and low affinity with husking granivores, with other affinities varying by taxon. Diet abbreviations: GranivoreH, Husking Granivore; GranivoreS, Swallowing Granivore; Tetra Hunt, Tetrapod Hunter.

Posterior probabilities predicting bohaiornithid diet by FDA from FEA following the intervals method [46]. Values with green backgrounds are more likely, values with red backgrounds are less likely. Bohaiornithid affinities varying considerably between taxa, only universally not resembling tetrapod hunters. Diet abbreviations: GranivoreH, Husking Granivore; GranivoreS, Swallowing Granivore; Tetra Hunt, Tetrapod Hunter

Violin plots of bird mass by diet, arranged in order of ascending mean mass. Masses were tested for significant differences via phylogenetic HSD. Diets marked with the same letter are not significantly different from one another. Cut-off points between significantly different mass groups (black lines, with 95% CIs as grey lines) were calculated by optimizing the Youden index and plotted. Note that, unlike in other diet treatments herein, granivores are not separated into husking and swallowing granivores. Mean bohaiornithid mass estimates are plotted for comparison, see Table 1. Diet abbreviations: Tetra Hunt, Tetrapod Hunter. Fossil taxon abbreviations: Be, Beiguornis khinganensis MHGU-F307/8; Bi, Bohaiornithidae indet. CUGB P1202; Bo1, Bohaiornis LPM B00167; Bo2, Bohaiornis IVPP V17963; L1, Longusunguis IVPP V17964; L2, Longusunguis IVPP V18693; P1, Parabohaiornis IVPP V18691; P2, Parabohaiornis IVPP V28398; Sh, Shenqiornis DNHM D2950/1; Su, Sulcavis BMNH Ph-805; cSu, cf. Sulcavis BMNHC-Ph1204; Z1, Zhouornis CNUVB-0903, Z2, Zhouornis BMNHC Ph 756.

Posterior probabilities predicting bohaiornithid pedal ecology by FDA and pFDA from TM of extant bird claws. Values with green backgrounds are more likely, values with red backgrounds are less likely. Bohaiornithid affinities vary considerably by taxon and between FDA and pFDA. Diet abbreviations: GranivoreH, Husking Granivore; GranivoreS, Swallowing Granivore; Tetra Hunt, Tetrapod Hunter. Category abbreviations: Large Raptor, raptor taking prey which does not fit in the foot; Small Raptor, raptor taking prey which can fit in the foot.

Quantitative ancestral states of the common ancestor of Enantiornithines. Mean values are the average of 10,000 random tree permutations where polytomies are randomly resolved. 95% confidence intervals are given as the 2.5% and 97.5% quantiles of the permutations, though upper and lower bounds for all values are identical to seven significant figures. As these values are based on multiple possible topologies of the enantiornithine tree, we consider them more valid than those visualised in Figs. 12–16.

Posterior probability of diet in the common ancestor of Enantiornithes based on MA and functional indices. Generalist feeding is the most likely, followed by piscivory and invertivory. Diet abbreviations: GranivoreH, Husking Granivore; GranivoreS, Swallowing Granivore; Tetra Hunt, Tetrapod Hunter.

Functional phylomorphospace of MA and functional index data, grouped by diet. Grey lines indicate phylogenetic relationships. Line drawings of skulls for selected taxa are provided for reference. Data are presented with PCA (A) and FDA (B). See Fig. S2 for character weights and Table 2 for FDA posterior predictions. Diet abbreviations: GranivoreH, Husking Granivore; GranivoreS, Swallowing Granivore; Tetra Hunt, Tetrapod Hunter. Fossil taxon abbreviations: B, Bohaiornis; L, Longusunguis; P, Parabohaiornis; Sh, Shenqiornis; Su, Sulcavis; Z, Zhouornis.

Violin plots of MWAM strain of FEA models, organized by diet. Extant diets ascend in average MWAM strain from left to right. MWAM strains were tested for significant differences via phylogenetic HSD. Diets marked with the same letter are not significantly different from one another. Diet abbreviations: GranivoreH, Husking Granivore; GranivoreS, Swallowing Granivore; Tetra Hunt, Tetrapod Hunter. Fossil taxon abbreviations: B, Bohaiornis; L, Longusunguis; P, Parabohaiornis; Sh, Shenqiornis; Su, Sulcavis; Z, Zhouornis.

Phylogenetic strain-space of total maximum in-plane principal strain intervals for extant and fossil bird lower jaw finite element models, grouped by diet. MWAM strain is mapped overtop the data. Grey lines indicate phylogenetic relationships. Contour plots for selected taxa are provided for reference. Data are presented with PCA (A) and FDA (B). In PCA (A) overall strain increases along PC1, and strain heterogeneity (i.e. lower areas of intermediate strain) increases along PC2. In FDA (B), DA1 and DA2 have loadings of various similar low-strain intervals, with high-strain intervals clustering near the origin. See Table 3 for FDA posterior predictions. Diet abbreviations: GranivoreH, Husking Granivore; GranivoreS, Swallowing Granivore; Tetra Hunt, Tetrapod Hunter. Fossil taxon abbreviations: B, Bohaiornis; L, Longusunguis; P, Parabohaiornis; Sh, Shenqiornis; Su, Sulcavis; Z, Zhouornis.

Phylomorphospace of extant and fossil bird claw shape from pedal TM, grouped by pedal ecological category. Grey lines indicate phylogenetic relationships. Line drawings of claws for selected taxa are provided for reference. Data are presented with PCA (A), FDA (B), and pFDA (C). See Fig. l for character weights and Table 4 for FDA and pFDA posterior predictions. Category abbreviations: Large Raptor, raptor taking prey which does not fit in the foot; Small Raptor, raptor taking prey which can fit in the foot. Fossil taxon abbreviations: B1, Bohaiornis LPM B00167; B2, Bohaiornis IVPP V17963; L, Longusunguis IVPP V18693; P1, Parabohaiornis IVPP V18690; P2, Parabohaiornis IVPP V18691; Su, Sulcavis BMNH Ph-805; Z1, Zhouornis CNUVB-0903, Z2, Zhouornis BMNHC Ph 756.

Summary table of interpretations of each line of evidence used herein. Body mass, MA, and FEA inform diet. Pedal TM informs the use of the pes or lack thereof in feeding. See relevant Discussion sections for additional details Bolded diets are agreed upon by all available diet proxies. Bolded pedal ecologies are either help discriminate diet (carnivory in Longusunguis) or are supported over other possibilities by diet information (non-raptorial perching in Bohaiornis). Given the uncertain application of skull TM to bohaiornithids, diets from this proxy are not bolded.

Phylomorphospace of extant and fossil bird skull proportions, grouped by diet. Grey lines indicate phylogenetic relationships. Line drawings of skulls for selected taxa are provided for reference. The data presented are modified from [10], see Fig. S4 for data more directly comparable to that study. Diet abbreviation: Tetra Hunt, Tetrapod Hunter. Fossil taxon abbreviations: B, Bohaiornis; L, Longusunguis; P, Parabohaiornis; Sh, Shenqiornis; Su, Sulcavis; Z, Zhouornis.

Ancestral state reconstruction of enantiornithine diet. Phylogeny is presented as not time-scaled for node visibility. All enantiornithine taxa with dietary hypotheses are included, as well as enantiornithines complete enough to create robust mass estimates [2, 8, 9, 26] and the non- ornithothoracine pygostylian Sapeornis as an outgroup. Taxa with bold names have diet assigned based on preserved meals (Eoalulavis [153], Sapeornis [3]) or quantitative diet proxies [8, 9]. Remaining diets assignments are based on qualitative morphology and depositional setting (Table S5). The diet of the common ancestor of Enantiornithes remains obscure, though many late-diverging enantiornithines are recovered as ancestrally invertivorous.