Main

Enantiornithes is the most species-rich and ubiquitously distributed clade of Mesozoic birds, known from every continent except Antarctica2. Nonetheless, as a result of taphonomic flattening associated with most enantiornithine-bearing localities (for example, the Jehol, Las Hoyas and Araripe Lagerstätten), complete, three-dimensionally preserved skulls are rare3,4,5,6,7. Virtually no well-preserved skulls bearing undistorted endocasts are known from taxa phylogenetically intermediate between Archaeopteryx (the earliest known Mesozoic bird) and some of the closest known relatives of crown birds (for example, Ichthyornis and Hesperornis), a gap exceeding 60 million years and encompassing most of the phylogenetic history of Mesozoic birds7,8,9,10,11. As a result, questions about the origins of the derived skulls and brains of crown birds remain among the most tantalizing knowledge gaps in vertebrate macroevolution1,7 and the detailed cranial osteology and endocranial morphology of enantiornithines remain poorly known3,12.

A bonebed contained in the Sítio Paleontológico ‘José Martin Suárez’ in Presidente Prudente (São Paulo State, southeastern Brazil) represents the richest Late Cretaceous fossil bird locality known and has exceptional potential to provide insight into the cranial morphology of enantiornithines. Unlike other quarries yielding an abundance of enantiornithine remains, such as those from the Early Cretaceous Jehol biota, this quarry is small in size (most fossils are contained in an area of about 6 m2 and in a sedimentary layer less than 50 cm deep), providing an unprecedented snapshot of a Late Cretaceous terrestrial avifauna. The quarry has thus far yielded hundreds of three-dimensionally preserved enantiornithine bones. Here we report on the only complete avian skull from this locality, which constitutes the best-preserved Mesozoic bird skull yet discovered, yielding insights into how and when characteristic features of the modern bird skull and central nervous system evolved. The remarkable state of preservation of the fossil may also clarify anatomical interpretations of previously known flattened enantiornithine cranial remains3.

Systematic palaeontology

Aves Linnaeus, 1758

Ornithothoraces Chiappe and Calvo, 1994

Enantiornithes Walker, 1981

Navaornis hestiae gen. et sp. nov.

Remarks. We use Aves to refer to all taxa descended from the most recent common ancestor of Archaeopteryx lithographica and crown birds13.

Etymology. Navaornis honours William Nava, who discovered the fossil locality in 2004 and the holotype specimen in 2016; the specific epithet hestiae alludes to Hestia, the Greek goddess of architecture, regarded as simultaneously the oldest and the youngest of the Twelve Olympians. Navaornis reflects this duality in that it belongs to an archaic lineage, yet its cranial geometry is essentially modern.

Holotype. MPM-200-1 (MPM, Museu de Paleontologia de Marília, Marília, Brazil). A complete skull (Fig. 1) articulated with the anterior-most cervical vertebrae, extracted from a block (MPM-200; Extended Data Fig. 1) from the William’s Quarry bonebed at Sítio Paleontológico ‘José Martin Suárez’. A cast of MPM-200 has been accessioned at the Dinosaur Institute, Natural History Museum of Los Angeles County.

Fig. 1: The holotype of the enantiornithine N. hestiae gen. et sp. nov. from the Late Cretaceous of Brazil.
figure 1

a,b, Photograph (a) and interpretive drawing (b) of the exposed side of the holotype of N. hestiae (MPM-200-1) in left lateral view. c, Micro-computed tomography rendering of MPM-200-1 in right ventral–lateral view. Scale bar, 10 mm.

Referred specimens. MPM-334-1, an isolated basicranium from William’s Quarry at Sítio Paleontológico ‘José Martin Suárez’12 whose morphology is identical to that preserved in MPM-200-1. MPM-200 includes a partially articulated postcranial skeleton, which is also referred to N. hestiae and is thought to belong to the same individual as the skull (Extended Data Fig. 1).

Locality and age. William’s Quarry is contained within the Adamantina Formation (Bauru Group, Bauru Basin); various lines of evidence14,15,16,17 indicate a late Santonian to early Campanian age (around 85–75 million years ago) for this site.

Diagnosis. Enantiornithine with a toothless skull and a combination of the following features: fully fused premaxillae with a convex dorsorostral surface, highly curved jugal, small, comma-shaped quadratojugal, diminutive lacrimal failing to separate the orbit from the antorbital fenestra, parasphenoidal rostrum perforated by a large ovoid fenestra, elongate basipterygoid processes, large and sinusoidal anterior semicircular canal excavating the dorsal margin of the supraoccipital, robust and prominent medial process of the mandible.

Phylogeny. Phylogenetic analyses (Methods) recovered the holotype of N. hestiae (MPM-200-1) within Enantiornithes, regardless of whether characters from the referred postcranium were included (Extended Data Fig. 2; see ‘Data availability’ for all phylogenetic tree topologies). Specifically, Navaornis clusters with Gobipteryx and Yuornis, other edentulous Late Cretaceous enantiornithines, although we note that most internal relationships within Enantiornithes remain poorly resolved18 and this last result may be influenced by a propensity for homoplastic tooth loss in Mesozoic birds19. The minimum age for the divergence between the most recent common ancestor of Enantiornithes and their sister clade Euornithes, which includes crown birds20,21 (the Ornithothoraces node), is defined by the oldest known, bona fide enantiornithines and euornithines from the lacustrine deposits of the Huajiying Formation (Jehol biota, northeastern China), dated to about 131 million years ago22,23.

Morphological description

The jaws of MPM-200-1 are completely toothless (Fig. 1). The premaxillae are fused into a single element, an unusual condition among enantiornithines, shared with Shangyang24 and other enantiornithines from William’s Quarry25, and the frontal process tapers caudally, terminating between the nasals without reaching the frontals, in contrast to the condition in Yuornis26 (Figs. 1 and 2 and Extended Data Figs. 26). The maxillary process of the premaxilla tapers caudally, inserting into a shallow trough on the lateral side of the maxilla (Figs. 1 and 2). The inverted T-shaped maxilla constitutes a prominent component of the facial margin (at least 1.5 times the length of the premaxilla at the tomial margin; Figs. 1 and 2). The maxillary rami are far more gracile than in many Early Cretaceous enantiornithines (for example, Bohaiornithidae27, Shenqiornis28 and Pengornis29), even when skeletally immature specimens are considered30,31. The premaxillary ramus of the maxilla is subequal in length to the jugal ramus and both are more robust than the nasal ramus; the last is long and directed caudomedially, underlying the maxillary process of the nasal and forming a medially directed lamina along its basal two-thirds (Figs. 1 and 2 and Extended Data Fig. 6). Medially, the premaxillary ramus expands into a sheet of bone which forms the palatal roof, as in Gobipteryx32 (Fig. 2 and Extended Data Figs. 2, 5 and 6). The confluence of the nasal and jugal rami form the rostral margin of the antorbital fenestra, which exhibits a funnel-like recess probably of pneumatic origin33 (Extended Data Fig. 6). The naris is elongate and tear-shaped, tapering caudodorsally; its caudal margin is formed by the maxilla and nasal, as in other Enantiornithes3 (Figs. 1 and 2 and Extended Data Figs. 2 and 5).

Fig. 2: Three-dimensional cranial reconstruction of N. hestiae.
figure 2

Composite three-dimensional reconstruction of the skull of N. hestiae from MPM-200-1 and referred braincase MPM-334-1. Scale bar, 10 mm.

The rostral two-thirds of the nasal comprises an elongate prong, which laterally lines the fused frontal process of the premaxilla; the postnarial portion of the nasal is proportionately shorter than in other enantiornithines (Figs. 1 and 2 and Extended Data Figs. 26). A minute, T-shaped bone is regarded as the lacrimal, an interpretation supported by its position between the orbit and the antorbital fossa and its overall morphology. Its ventral ramus curves slightly medially and, as in Yuornis, does not separate the orbit from the broad antorbital fenestra (Figs. 1c and 2 and Extended Data Figs. 2 and 5). The jugal is elongate and strongly bowed, with a long, tapering postorbital ramus and a greatly reduced quadratojugal ramus (Figs. 1 and 2 and Extended Data Figs. 2, 3, 5 and 6). This morphology is reminiscent of that seen in some other Enantiornithes30,31,34, although the postorbital process of Navaornis is proportionately longer and the quadratojugal ramus proportionately shorter. Another T-shaped bone, twice the size of the one interpreted as the lacrimal, is preserved in contact with the medial surface of the left jugal and is interpreted as a postorbital (Fig. 1c and Extended Data Figs. 2 and 4), an element previously identified in some enantiornithine taxa3,31,34. It exhibits a short squamosal ramus and a longer frontal ramus (Extended Data Fig. 6), presumably connecting to the postorbital process of the frontal (Fig. 2b). The length of the rami of the postorbital and jugal suggest that these bones would have fully separated the orbit from an infratemporal fenestra. A short, comma-shaped bone is interpreted as the quadratojugal (Figs. 1 and 2 and Extended Data Figs. 26); it differs from the flattened L-shaped quadratojugal known from some other enantiornithines31,35,36. Although disarticulated, the left and right quadratojugals are preserved in similar positions on both sides of the skull, implying that they are close to their natural positions (Fig. 1b,c), probably contacting the quadrate caudoventrally (Figs. 1 and 2 and Extended Data Fig. 5). Additionally, the quadratojugal exhibits a short anteroventral ramus which projects from the midpoint of the bone, a feature not observed in other enantiornithines31,35,36.

The frontal is large and vaulted. It exhibits a straight interfrontal margin, clearly unfused to its counterpart, similar to the condition in Yuornis26, Zhouornis37 and other enantiornithines34, as well as isolated frontals recovered from the same outcrop; its caudal margin is similarly unfused to the parietal, which bears a distinct nuchal crest (Extended Data Figs. 2, 6 and 7). Caudolaterally, the frontal projects into a triangular postorbital process, also present in Zhouornis37 and the Montsec immature enantiornithine30. The postorbital process of the frontal seems to overlie a slightly larger, laterally projecting postorbital process of the laterosphenoid, as interpreted in Ichthyornis9 (Fig. 2). The internal surface of the laterosphenoid bears an ample fossa circumscribing part of the optic lobe, with the remainder of the depression extending caudally onto the parietal (Extended Data Fig. 4). The basicranium (including the vestibular region) is virtually identical to a previously reported braincase from the same quarry12 (for example, bearing large and slender basipterygoid processes and a long parasphenoidal rostrum perforated by an ovoid fenestra), which we here refer to N. hestiae (Extended Data Fig. 8). This referred specimen (MPM-334-1) bears a triradiate squamosal with a tapering zygomatic process similar to the Montsec immature enantiornithine30 which seems to be partially fused to the rest of the basicranium (Extended Data Fig. 8).

Information from the postorbital, jugal, quadratojugal, squamosal and laterosphenoid allows the skull of Navaornis to be reconstructed with a diapsid temporal configuration; the infratemporal fenestra is unbounded caudally as the squamosal does not seem to contact the quadratojugal (Fig. 2 and Extended Data Figs. 2 and 5). Varying degrees of closure of the upper and infratemporal fenestrae have been previously reported for Enantiornithes30,31,36,38 as well as both crownward9 and stemward39,40 Mesozoic birds. In Navaornis, although the postorbital probably contacted the caudal portion of the jugal, a continuous bony connection between the postorbital and the caudally positioned zygomatic process of the squamosal (incomplete or weathered in both the holotype and referred specimens) would have been very delicate or partially ligamentous (Figs. 1b and 2 and Extended Data Fig. 8).

Portions of the palate are infilled with radiopaque minerals, thus only the morphology of the left palatine and pterygoid can be unambiguously discerned from our computed tomography scans (Fig. 2 and Extended Data Figs. 36). The palatine is rostrocaudally elongate and mediolaterally narrow, with a rostrally directed maxillary ramus, a dorsally projecting flange at its rostrocaudal midpoint and a dorsomedially slanted pterygoid ramus; these features are reminiscent of those of Gobipteryx32, although the palatine of Navaornis is proportionately longer and more gracile. The pterygoid has a broad, lateroventrally slanted body and is roughly oval in shape. It has an expanded and biradiate caudal margin, including a prominent dorsolateral ramus which develops an ample contact with the orbital process of the quadrate (Extended Data Fig. 6). The morphology of the quadrate ramus of the pterygoid is comparable to that of a previously described juvenile enantiornithine31 and many non-avian theropods (for example, refs. 41,42) and the pterygoid lacks a mobile connection with the palatine/hemipterygoid as has been reported in the Cretaceous ornithurines Ichthyornis and Janavis11,43.

The quadrate exhibits a broad orbital process (Figs. 1 and 2 and Extended Data Figs. 5 and 6), as is typical of non-ornithothoracine avians, Enantiornithes and stemward Euornithes (for example, Patagopteryx44). Its ventrolateral corner exhibits a small, circular quadratojugal cotyle (Fig. 2 and Extended Data Figs. 5 and 6); although similar cotyles of varying sizes are known in Cretaceous Euornithes (for example, Patagopteryx44, Parahesperornis45, Hesperornis46 and Ichthyornis9,47), this trait has never previously been reported in Enantiornithes.

The dentary occupies more than half the rostrocaudal length of the mandible (Fig. 1). It bears an obliquely oriented suture with the postdentary complex, as in other Enantiornithes. The robust medial process of the lower jaw projects dorsomedially, as in Yuornis26 and possibly Gretcheniao18. Most sutures in the postdentary complex are not distinguishable, although a triangular splenial and a broad pre-articular line the lingual margin of the caudal half of the mandible (Figs. 1 and 2), similar to the condition in some extant anseriforms.

Nearly complete endocranial reconstructions of the brain and inner ear of Navaornis were generated from the well-preserved frontals, parietals, laterosphenoids and basicranium of MPM-200-1 and MPM-334-1 (Methods; Fig. 3a and Extended Data Figs. 7 and 8). The brain is heart-shaped in dorsal view and exhibits a flexed configuration with the brainstem projecting ventrally to a degree heretofore seen only in crown birds12,48 (Fig. 3). The extent of separation between the telencephalic hemispheres remains somewhat ambiguous given that only the left frontal is well preserved. The hemispheres are vaulted and pyriform in dorsal view and are mediolaterally expanded to a degree exceeding that of Archaeopteryx7,49 and known non-avian pennaraptoran dinosaurs50,51 (Fig. 3b). However, the hemispheres do not envelop the dorsal surface of the optic lobes, as in Cerebavis and most crown birds, in which the lobes are obscured in dorsal view by extreme mediolateral expansion of the telencephalon52. Unlike the condition in Ichthyornis11 and most crown birds, the dorsal surface of the telencephalon is smooth (Fig. 3b), lacking a discernible Wulst—a thickening of the hyperpallium often visible on the telencephalic surface and an important structure involved in integrating visual and somatosensory information52. The optic lobe is subspherical and lies immediately ventral to the posterior portion of the telencephalon; both structures terminate caudally at the same point and probably have an extended contact area as a result of the ventral expansion of the telencephalon (Fig. 3). The optic lobe is more caudally positioned than in any previously known stem bird12, approaching the degree seen in many crown birds with moderately to strongly flexed brains52. The optic lobe sits dorsal to a greatly enlarged and sinusoidal anterior semicircular canal of the endosseous labyrinth whose morphology is indistinguishable from that of the referred specimen MPM-334-1 (ref. 12) (Extended Data Fig. 8). The cerebellum seems to be small, lacking the inflated morphology of Cerebavis and many crown birds, indicative of its volumetric expansion7,8 (Fig. 3). Indeed, the cerebellum of Navaornis seems to be even more weakly developed than those of Archaeopteryx7,49 and some non-avian dinosaurs (for example, Zanabazar53 and IGM 100/1126 (ref. 1)) and no distinct foliation of the cerebellar surface is discernible54,55. The endocast of the medulla is distorted in MPM-200-1 but its globular morphology matches that of the referred specimen MPM-334-1 (ref. 12).

Fig. 3: Endocranial anatomy of N. hestiae and brain evolution in stem birds.
figure 3

a, Three-dimensional reconstruction of the endocranial morphology of N. hestiae from MPM-200-1 and MPM-334-1. Portions deriving from MPM-200-1 and MPM-334-1, as well as the reconstruction process, are explained in the Methods and Extended Data Fig. 7. b, Evolution of endocranial morphology across Pennaraptora. Numbers in the coloured boxes refer to the degree of expansion of each of the main neuroanatomical and sensorial regions for each taxon. Brown arrows in b depict the orientation of the foramen magnum.

Discussion

Combined with recently published material25, Navaornis provides evidence of co-occurring toothless and toothed enantiornithines from a single site, hinting at a considerable degree of sympatric ecological disparity in some Cretaceous bird communities. The toothless condition of its skull notwithstanding, the plesiomorphic nature of most individual elements in the skull of Navaornis (Fig. 2 and Extended Data Figs. 5 and 6) supports the suggestion that enantiornithines retained an akinetic skull with a diapsid temporal configuration30,36. Nonetheless, the geometry of the three-dimensionally preserved and undistorted skull of Navaornis is similar to that of crown birds overall. This interpretation is supported by our quantitative morphometric comparisons which show that Navaornis falls comfortably in a densely occupied region of extant avian cranial morphospace defined by the three principal components of cranial shape variation (Fig. 4a and Extended Data Fig. 10). Navaornis is closest in overall cranial geometry to an array of extant taxa belonging to disparate crown bird groups (for example, Chauna, Psophia, Chunga and Corvus; Extended Data Fig. 9).

Fig. 4: Navaornis combines a geometrically crown bird-like skull with a central nervous system morphologically intermediate between Archaeopteryx and crown birds.
figure 4

a, Three-dimensional principal component morphospace of neornithine skulls (PC1 versus PC2). Warped models illustrate extremes along PC1 and PC2. b, Three-dimensional principal component morphospace (PC1 versus PC3) of endocrania from Aves and relevant non-avian taxa. Navaornis falls in crown bird morphospace along PC3 but not along PC2 (Extended Data Fig. 10). Landmarks are colour-coded for high (warm colours) to low (cold colours) per-landmark variances. Brown arrows in b depict the foramen magnum orientation.

Navaornis provides a long-awaited illustration of the morphology of the brain and inner ear within a critically undersampled phylogenetic interval of the avian stem lineage, capturing a previously unknown combination of plesiomorphic and derived traits which clarify the pattern by which the distinctive central nervous system of living birds arose (Fig. 3b). Our quantitative geometric comparisons demonstrate that Navaornis exhibits a brain morphology intermediate between Archaeopteryx and crown birds along the main axis of endocranial shape variation (PC1), which accounts for nearly half of total neuroanatomical shape variation in birds and closely related theropod dinosaurs (Fig. 4b and Extended Data Fig. 10). Specifically, Navaornis suggests that the expansions of the telencephalon and, to a lesser extent, cerebellum characteristic of crown birds7,56 seem to have arisen at a comparatively crownward point in avian evolutionary history (Figs. 3b and 4b). The characteristic inflated morphology of the cerebellum in crown birds can be concealed by ventral folding (flexion) of the brain in small-sized neoavian birds (for example, hummingbirds, woodpeckers and passerines)52, potentially leading to underestimates of cerebellar size from endocranial surface reconstructions. However, although the endocast of Navaornis is more ventrally flexed than that of many extant representatives of Palaeognathae and Galloanserae57, it is substantially less flexed than in most extant neoavian birds12, suggesting that our reconstruction of a comparatively flat cerebellum in Navaornis is unlikely to be meaningfully influenced by brain flexion.

The presence of a ventrally flexed brain configuration, caudally displaced optic lobes and an enlarged sinusoidal labyrinth in Navaornis implies that the origins of these ‘advanced’ traits often associated with crown birds11,58 either predated the origin of Ornithothoraces or evolved convergently among both Enantiornithes and crownward Euornithes. Regardless of which of these evolutionary scenarios is supported by future findings, aspects of the neuroanatomy of Navaornis are conspicuously similar to those of late-stage extant bird embryos and hatchlings (for example, Gallus and Ficedula59). For example, the optic tectum of crown birds attains adult-like proportions in early ontogeny along with a ventrally oriented brainstem in species with a ventrally flexed brain, yet the telencephalon and cerebellum only acquire their more expanded forms later in development, echoing other instances of ontogenetic and phylogenetic covariation across the dinosaur–bird transition53,60,61. Altogether, the morphology of the endocast of Navaornis shows an intermediate stage in the evolutionary history of the unique avian brain and its combination of crown bird-like and plesiomorphic attributes is congruent with hypotheses of modular evolution whereby different anatomical regions evolve quasi-independently at varying rates60.

In Navaornis, this intermediate stage of brain evolution is combined with a skull retaining numerous plesiomorphic features in the main cranial elements, which serve as building blocks from which a geometrically modern cranial configuration is constructed. This degree of geometric convergence between Enantiornithes and crown birds suggests that developmental constraints responsible for canalizing the general shape of the bird skull may have been present throughout much of avian evolutionary history, predating both the phylogenetic divergence between Enantiornithes and Euornithes more than 130 million years ago as well as the evolutionary acquisition of several apomorphic characteristics of crown bird skull and brain morphology. The exceptionally well-preserved skull of Navaornis emphasizes the necessity of hitherto elusive undistorted Mesozoic bird skulls for illuminating the complex sequence by which the unique brains and skulls of modern birds arose.

Methods

Preparation and imaging of specimens

MPM-200-1 was scanned using a Bruker SkyScan 1173 computed tomography scanner at the Instituto de Petróleo e dos Recursos Naturais (Laboratório de Sedimentologia e Petrologia) of the Pontifícia Universidade Católica do Rio Grande do Sul (PUCRS), Porto Alegre, Rio Grande do Sul, Brazil. Scanning parameters were as follows: 10.71 µm voxel size, 60 kV, 133 µA, exposure time 850 ms, averaging two frames, no 360° rotation, using an aluminium filter of 1.0 mm. Slices were reconstructed using the software NRecon v.1.7.4.6. Volumes were digitally reconstructed and segmented using VGSTUDIOMAX 3.4.0 (VolumeGraphics; see Extended Data Figs. 3 and 4 for all segmented elements in situ).

Three-dimensional skull and endocast reconstruction

Three-dimensional meshes of each recognizable segmented cranial element and endocranial surface were exported from VGSTUDIOMAX 3.4.0 to Blender 3.3.0, in which they were re-articulated to reconstruct the skull and endocranial anatomy of N. hestiae (Figs. 2 and 3a and Extended Data Figs. 5 and 7). Owing to the lack of distortion of most cranial elements, only the left frontal bone required slight retrodeformation following established best practices62. This was accomplished using the Lattice function in Blender, in which mediolateral compression of this element was corrected to match the geometry of the dorsal rim of the complete and mostly undistorted left parietal bone (Extended Data Fig. 7). The same degree of retrodeformation was then applied to the endocast surface of the left frontal, that is, the left hemisphere of the telencephalon (Extended Data Fig. 7). The remaining surfaces composing the endocranium of N. hestiae were: ventral (derived from the basicrania of both the holotype and referred specimen MPM-334-1), anterior (derived from the right and left laterosphenoids of the holotype), dorsal and lateral surfaces of the left optic lobes (derived from the parietal of the holotype), cerebellum (derived from the parietal of the holotype and basicranium of both the holotype and referred specimen MPM-334-1) and medulla (derived from both the holotype and referred specimen MPM-334-1; Extended Data Fig. 7). Left and right surfaces were mirrored from the best-preserved element/endocranial surface.

Three-dimensional geometric morphometrics of the skull and endocast

Landmark-based geometric morphometrics were used to quantitatively compare the exocranial and endocranial anatomy of Navaornis with crown birds and a selection of non-avian dinosaurs and Mesozoic birds. First, the three-dimensional geometry of our exocranial reconstruction of Navaornis (Fig. 2 and Extended Data Figs. 5 and 7) was captured by adapting a previously proposed landmarking scheme designed for crown birds63 to enantiornithine cranial morphology (Extended Data Fig. 10), enabling us to explore the striking similarities in exocranial geometry between crown birds and Navaornis despite their anatomical differences across several regions of the skull. For instance, although the maxilla represents one of the main elements of the rostrum in enantiornithines, it is greatly reduced in crown birds, drawing the cranial end of the jugal rostrally such that it is essentially level with the rostral-most point of the antorbital fenestra. To accommodate these anatomical differences between crown birds and enantiornithines, we moved the landmark located at the rostral end of the jugal bar in crown birds from the original landmarking scheme to the rostral-most point of the antorbital fenestra on the ventrolateral rim of the maxilla in Navaornis (Extended Data Fig. 10). Likewise, we redefined the original landmark at the caudoventral-most point of the maxilla and placed it just rostral to the aforementioned modified landmark in Navaornis. These modifications affected the semilandmark curves defining the ventrolateral rims of the upper jaw and jugal bar, as these strings of semilandmarks are anchored to the aforementioned landmarks. These semilandmarks were redigitized ensuring that they were evenly spaced along the redefined curves. Finally, the landmark defining the basioccipital–parabasisphenoidal rostrum contact in the original landmarking scheme was placed just caudal to the recess representing the caudal-most point of the parasphenoidal rostrum in Navaornis. Following these modifications to the landmarking scheme, we incorporated the three-dimensional coordinates of Navaornis into a broad three-dimensional avian exocranial morphological dataset encompassing 228 extant species representing the phylogenetic breadth of crown birds64.

We then applied a previously published landmarking scheme56 to the endocranial reconstruction of Navaornis (Extended Data Fig. 10) to capture and compare its three-dimensional endocranial anatomy with a representative sample of crown birds, Archaeopteryx, closely related non-avian theropods and Alligator, representing Crocodylia, the extant sister group to birds. Exocranial landmarks were digitized in Avizo Lite 2019.3 (Thermo Fisher Scientific) and endocranial landmarks were digitized in Landmark Editor65 following previously described procedures56,64. Thereafter, the landmark datasets were imported into the R statistical environment v.4.1.2 (ref. 66), in which all downstream analyses were conducted.

Generalized Procrustes analyses were performed on both sets of landmark coordinates to separate shape data from size and other confounding factors and the minimum bending energy criterion67 was used to slide curve (exocranium and endocranium) and patch (endocranium) semilandmarks following previously described procedures56,64, using the function gpagen in the R package geomorph v.4.0.5 (ref. 68).

Principal components analyses were carried out on the exocranial and endocranial Procrustes coordinates to visualize shape variation using the function gm.prcomp in geomorph. To determine the extant species geometrically closest to Navaornis in exocranial shape, we determined the Procrustes distances between Navaornis and all extant taxa in our dataset (Extended Data Fig. 9) using Euclidean distances with the function dist from the R package stats v.4.1.2 (ref. 66). Changes associated with main axes of exocranial and endocranial shape variation were illustrated as deformations warped from the three-dimensional surface of the exocranium and endocranium of the individual species closest to the mean shape in both samples. Specifically, this three-dimensional surface and the mean shape from the sample were projected onto the scores representing the 0.05 and 0.95 quantiles for each principal component axis by means of thin-plate spline deformation69 using the function tps3d from the package Morpho v.2.10 (ref. 70) and shape.predictor from geomorph. We also plotted the respective landmark configurations onto the deformed meshes using shape.predictor and coloured these landmark constellations according to per-landmark variances from each dataset using the hot.dots function (freely available following this link: https://zenodo.org/record/3929193)71.

Phylogenetic analysis

Heuristic parsimony analyses were applied to a previously published and expanded (adding Navaornis and Yuornis) dataset18 using TNT v.1.6 (ref. 72) under equal and implied weights (K = 3, 9 and 12). Twenty-four multistate characters (1, 3, 8, 23, 39, 44, 48, 51, 52, 68, 77, 84, 134, 143, 149, 162, 173, 180, 181, 183, 187, 190, 195 and 200) were treated as additive (or ‘ordered’), according to the recommendation that multistate characters be analysed as additive when they represent morphoclines (for example, small–medium–large)73. We analysed a character matrix in which the scorings for Navaornis were based exclusively on the holotype (MPM-200-1) and another matrix with further scorings from the referred specimen MPM-200 (a postcranial skeleton contained in the same block; Extended Data Fig. 1). Analyses were performed with 10,000 replicates, using tree bisection–reconnection branch swapping, retaining ten trees per replicate. Support was obtained by calculating bootstrap values set at 10,000 replicates. The equal weights analysis using only cranial scorings resulted in 102 most parsimonious trees of 909 steps each; the strict consensus tree resulted in a large polytomy among enantiornithines. By contrast, the analysis using K = 3 (fit = 88.25) resulted in two most parsimonious trees, whereas those performed under K = 9 (fit = 46.21) and K = 12 (fit = 37.53) resulted in one tree each with nearly identical topologies. The consensus tree obtained from the K = 3 analysis has some polytomies in Enantiornithes, whereas the relationships in this clade in the K = 9 and 12 trees are fully resolved. The analysis of the character matrix with further postcranial scorings resulted in two most parsimonious trees of 916 steps and yielded a nearly identical topology to the analysis including only the holotype skull. In all most parsimonious trees, Navaornis is nested in the enantiornithine clade. Trees resulting from all the analyses are freely available at ref. 74.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.