Main

Among the earliest and most enduring insights into the phylogenetic relationships of living birds was the recognition of the two deepest crown bird subclades based on palate structure: the reciprocally monophyletic Palaeognathae (‘ancient jaws’) and Neognathae (‘modern jaws’)1. The palate of palaeognaths is characterized by elongate basipterygoid processes arising from the caudal portion of the parasphenoid rostrum that serve as a buttress for the fused pterygoid–palatine complex2,4. By contrast, neognaths exhibit either reduced or absent basipterygoid processes and unfused pterygoid and palatine bones that meet at a mobile joint and confer a greater capacity for palatal kinesis3,5,16. The nature of the ancestral crown bird palate has long been assumed to have been palaeognathous3,5,6,7,17; however, the rarity of delicate fossil palate bones from early crown birds and crownward Mesozoic avialans has precluded direct evaluation of this assumption7,8,12,18. Here we address this question by reporting a complete, three-dimensionally preserved pterygoid from a 67-million-year-old relative of Ichthyornis, filling one of the few remaining gaps in the cranial osteology of ichthyornithine avialans—one of the most crownward groups of Mesozoic stem birds.

Systematic palaeontology

Avialae Gauthier, 1986 sensu Benito et al. 2022

Ornithurae Haeckel, 1866

Ichthyornithes Marsh, 1873b sensu Benito et al. 2022

Janavis finalidens

Etymology Janavis from the Roman god Janus and the Latin avis for bird. In Roman mythology, Janus is the god of beginnings, endings and transitions, reflecting transitional aspects of the morphology of Janavis (combining plesiomorphic features such as teeth with a neognath-like palate) as well as its temporal provenance (deriving from the uppermost Cretaceous, making it one of the youngest non-neornithine avialan fossils in the world). The specific epithet finalidens, from the Latin finalis (adj.), meaning ending or final, and dens, for teeth, reflects the fact that the specimen is among the latest-known toothed avialans, which appear to have died out in the end-Cretaceous mass extinction shortly after Janavis lived9.

Holotype Natuurhistorisch Museum Maastricht (NHMM) RD 271, a partial skeleton preserved in matrix, and associated elements extracted from the matrix by the collector, R. Dortangs, previously reported by Dyke et al.19 (Fig. 1a; see Supplementary Information for character information, measurements, additional description and discussion). Elements preserved within the main block include six cervical and four thoracic vertebrae, six ribs, a left scapula, a nearly complete left humerus, a right manual phalanx II:1, a proximal fragment of the right femur and several unidentifiable fragments (Extended Data Figs. 1 and 2). Elements fully extracted from the matrix include an isolated tooth, the complete left pterygoid preserved in two pieces, three thoracic vertebrae and a partial pedal phalanx (Extended Data Figs. 1 and 2). Several of these elements were left undescribed, unfigured or misidentified in the original report of the specimen19 (Fig. 1); see ‘Description’, ‘Materials’ and Supplementary Information for further details.

Fig. 1: Skeletal reconstruction, axial pneumaticity and phylogenetic position of J. finalidens.
figure 1

a, Skeletal reconstruction of J. finalidens (NHMM RD 271), including all preserved skeletal elements, compared with a composite skeleton of I. dispar illustrated at the same scale. Skeletal elements in yellow indicate previously unreported elements, and those in blue indicate elements reidentified in this study. b, CT cross-sections illustrating the apneumatic thoracic vertebrae of Ichthyornis compared with the extensively pneumatized thoracic vertebrae of J. finalidens. The red arrowheads indicate pneumatic foramina. c, Time-scaled phylogenetic tree showing the phylogenetic position and stratigraphic provenance of J. finalidens. The topology is simplified; see Extended Data Fig. 6 for full phylogenetic results. The orange branches denote non-ornithothoracine Avialae, the yellow branches denote Enantiornithes, the light blue branches denote paraphyletic ‘non-ornithurine Euornithes’ and the dark blue branches denote Ornithurae. The boxes at the tips of phylogenetic branches represent the temporal range of a taxon inclusive of stratigraphic uncertainty; the triangles at the tips of phylogenetic branches represent clades comprising more than one known species, with their length indicating the approximate temporal range of the known representatives of the collapsed clade. Illustrations of representative species for each lineage are depicted, from top to bottom: Eoconfuciusornis zhengi (Confuciusornithiformes), Eoalulavis hoyasi (Enantiornithes), Archaeorhynchus spathula, Longicrusavis houi (Hongshanornithidae), Abitusavis lii (‘Yanornithidae’), Iteravis huchzermeyeri, Gansus yumenensis, I. dispar and J. finalidens (Ichthyornithes), Brodavis varneri (Hesperornithes), Iaceornis marshi, and Asteriornis maastrichtensis (Neornithes). Background colours mark geological stages. Camp., Campanian; Cenom., Cenomanian; Con., Coniacian; Jur., Jurassic; Ma, million years ago; Maast., Maastrichtian; Sant., Santonian. The skeletal reconstruction of Ichthyornis was modified from Benito et al.22. Bird illustrations to the right of the phylogeny are courtesy of R. Olivé, used with permission.

Full size image

Locality and age CBR-Romontbos Quarry, Eben-Emael, Liège, Belgium. Valkenburg Member (67–66.9 million years ago20), Maastricht Formation, late Maastrichtian, Late Cretaceous. Additional details regarding the locality and stratigraphic setting are provided in the Supplementary Information.

Diagnosis Janavis is distinguished from other known Euornithes, and from Ichthyornis in particular, by the greater degree of pneumaticity of its thoracic vertebrae and ribs, especially the presence of large ventral pneumatic openings in the anterior thoracic vertebrae and fenestrated ventrolateral tubercles on the fifteenth presacral vertebra (Fig. 1b and Extended Data Figs. 1 and 3). It is also distinguished by the complete absence of an acromion process on the scapula21,22 (Extended Data Fig. 2d) and, most obviously, by its much larger size (maximum length of the Janavis humerus is 134.8 mm; maximum length of the longest known Ichthyornis humerus, YPM 1742, is 71.5 mm21). Additional character combinations from phylogenetic analyses that diagnose Janavis are presented in the Supplementary Information.

Description

Janavis is a large-bodied ichthyornithine with a mean estimated body mass based on humeral least shaft circumference (R2 = 0.955; Methods) of approximately 1,504 g—substantially heavier than the largest known Ichthyornis dispar specimens (mean body mass estimates of up to 486 g (ref. 22)) and slightly heavier than a typical adult magnificent frigatebird (Fregata magnificens; 1,465 g), grey heron (Ardea cinerea; 1,443 g) or turkey vulture (Cathartes aura; 1,430 g). Complete measurements of NHMM RD 271 are provided in the Supplementary Information, with further discussion of body size estimation provided in the Methods.

The limited skull material of Janavis, which, contrary to previous reports19, does not include jaws or portions of the zygoma, comprises a single isolated tooth, comparable in shape to those of Ichthyornis12,21,23,24 (Fig. 1a and Extended Data Fig. 2a), and a complete left pterygoid that until now had been identified as a partial coracoid19 (Fig. 2 and Extended Data Figs. 2b and 4). The pterygoid is preserved as two precisely matching fragments and is exceptionally well preserved despite its extremely thin bone walls. It exhibits a large, rostrally situated, ovoid and flat basipterygoid facet on its medial surface, very similar in shape, placement and proportions to those of extant galloanserans such as Chauna (Anseriformes) and Lophophorus (Galliformes), and highly unlike the minute facet situated approximately halfway along the pterygoid shaft found in neoavians exhibiting this structure, or the caudally situated shallow depressions for the basipterygoid processes found in palaeognaths4,5 (Fig. 2 and Extended Data Fig. 5). The rostral end of the pterygoid lacks any hint of the pronounced rostrally projecting process characteristic of Anseriformes and, as such, more closely resembles those of most Galliformes (Fig. 2 and Extended Data Fig. 4). The contact for the hemipterygoid–palatine complex is round, deep and cup-like, with a marked rim surrounding its dorsal and lateral edges, a morphology shared by extant galloanserans, such as Chauna, Lophophorus and Numida, and certain neoavians, such as Gavia and Puffinus5,10 (Fig. 2 and Extended Data Figs. 4 and 5). The midpoint of the medial surface of the pterygoid shaft is perforated by a very large foramen (2.55 mm in diameter); the bone is mostly hollow and exhibits extremely thin bone walls, suggesting that the foramen was pneumatic25,26.

Fig. 2: Comparative pterygoid and palatal morphology of J. finalidens and representatives of the main neornithine lineages.
figure 2

The left pterygoids in dorsal (right) and medial (left) views. Skulls are in palatal view, with the main palatal elements coloured: quadrate (yellow), pterygoid (blue) and the hemipterygoid–palatine complex (green). The cladogram to the right of the skulls shows the phylogenetic relationships of the illustrated taxa; branch colours represent Galloanserae (pink), Neoaves (blue) and Palaeognathae (yellow). The ichthyornithine skull is a composite of several Ichthyornis specimens with the pterygoid of Janavis (Methods), scaled to the size of Janavis. Scale bars, 2.5 mm (isolated pterygoids) and 10 mm (skulls). †Extinct fossil taxon.

Full size image

The postcranial morphology of Janavis is remarkably similar to that of the Late Cretaceous ornithurine I. dispar in almost every regard19 (Fig. 1a,b and Extended Data Figs. 2 and 3), exhibiting several morphological features previously considered autapomorphic for Ichthyornis21,22,23. For a detailed description of the postcranial skeleton, see the Supplementary Information.

Phylogenetic analyses

Our analyses consistently recovered Janavis in an exclusive clade with Ichthyornis, a result that was robust to alternative methodological approaches and phylogenetic datasets (Fig. 1c and Extended Data Fig. 6). This clade is supported by two unambiguous synapomorphies and one ambiguous synapomorphy, all of which were previously considered to be autapomorphic for Ichthyornis21,22: amphicoelous cervical vertebrae (ambiguous), an acromion process projected less cranially than the scapular articulation surface for the coracoid (unambiguous) and the presence of an internal index process on manual phalanx II:1 (unambiguous).

The clade formed by Ichthyornis and Janavis (Ichthyornithes sensu Benito et al.22) was consistently recovered in a relatively crownward position within Euornithes, stemward only of Hesperornithes and Iaceornis among non-neornithine Avialae.

Geometric morphometric analyses

To assess the morphological similarity of the pterygoids of Janavis and those of crown birds quantitatively, and to evaluate the morphological variability of the crown bird pterygoid, we performed a three-dimensional geometric morphometric analysis of crown and stem bird pterygoids. Our analyses included 34 extant and fossil taxa, and a landmarking scheme comprising six landmarks and 386 pseudo-landmarks (Extended Data Fig. 7). The first three principal components (PCs) of the avian pterygoid morphospace, which explain 72.2% of the total variance of the dataset, illustrate clear geometric clusters for Galloanserae and Neoaves, and a wide distribution for Palaeognathae (Fig. 3 and Extended Data Figs. 8 and 9). Janavis is recovered near extant and fossil galloanserans on all three main PC axes: Janavis falls very close to Galliformes along PC1 and occupies a position roughly intermediate between Galliformes and Anseriformes on PC2 and PC3, and occupies a similarly intermediate position in the three-dimensional morphospace (Fig. 3). See Supplementary Information for additional interpretation of our geometric morphometric results.

Fig. 3: Three-dimensional geometric morphospace of avian pterygoid morphology.
figure 3

Proportion of variance explained by each of the first three PCs indicated on axis labels. The colours indicate phylogenetic affinities of the included taxa: Palaeognathae (yellow), Galliformes (burgundy), Anseriformes (pink) and Neoaves (blue); fossil taxa are indicated in grey. The illustrated pterygoids on the right correspond to (1) J. finalidens (mirrored left pterygoid), (2) Struthio camelus (Palaeognathae: Struthionidae), (3) Rhea americana (Palaeognathae: Rheidae), (4) Dasornis toliapica (Avialae incertae sedis: Pelagornithidae), (5) Anatalavis oxfordi (Neognathae: stem-Anseriformes), (6) Anas platyrhynchos (Anseriformes: Anatidae), (7) Lophophorus impejanus (Galliformes: Phasianidae), (8) Scolopax rusticola (Charadriiformes: Scolopacidae) and (9) Trogon collaris (Telluraves: Trogonidae). The cladogram to the right shows the phylogenetic relationships of the illustrated taxa; branch colours match those of the silhouettes that they correspond to; dashed line denotes the uncertain phylogenetic position of the clade.

Full size image

Discussion

Evolution of Ichthyornithes

Our phylogenetic analyses support the original interpretation of Janavis as a member of Ichthyornithes19 following the updated definition of the clade21,22.

The postcranial morphology of Janavis closely resembles Ichthyornis despite stratigraphic separation of roughly 20 million years. Indeed, the most notable morphological differences between these taxa, such as the proportionally shorter deltopectoral crest of Janavis, are seemingly allometric features related to its much larger size (Fig. 1a). Notably, Janavis exhibits several morphological features previously considered autapomorphic for Ichthyornis, including amphicoelous or biconcave cervical vertebrae, and the presence of an internal index process on manual phalanx II:1 (refs. 21,22). Our analyses recover these features instead as synapomorphies of the expanded Ichthyornithes clade. The shape of the acromion process—minute and not extending beyond the coracoid articular facet of the scapula—was considered a diagnostic feature of Ichthyornis21,22. This feature represents one of the clearest morphological differences between Ichthyornis and Janavis, in which the acromion process is entirely absent (Extended Data Fig. 2c,d), a condition only observed in Anhimidae among crown birds21. Other elements bearing diagnostic features for Ichthyornis are not preserved among the holotype remains of Janavis. Despite their otherwise comparable vertebral morphology, Janavis is distinguished from Ichthyornis by the extensive pneumatization of its thoracic vertebrae and ribs (Fig. 1b and Extended Data Figs. 1, 2e and 3).

The marked overall similarity between Ichthyornis and Janavis suggests a notable degree of morphological stasis within Ichthyornithes over a time span of nearly 20 million years21,22,27. Although fragmentary remains showing affinities with Ichthyornis have previously been reported from uppermost Cretaceous rocks9, thus far, Janavis constitutes the only confidently identified and non-fragmentary Maastrichtian ichthyornithine known, and its large size and presumed extinction at the end-Cretaceous are congruent with hypothesized selection favouring relatively small-bodied birds across the Cretaceous–Paleogene (K–Pg) boundary28,29. Indeed, Janavis greatly outweighs the only other co-occurring avialan known from the Maastricht Formation: the early neornithine Asteriornis maastrichtensis, the holotype of which is estimated to have weighed in the range of 395 g (ref. 27). Relative stasis in the overall skeletal morphology of Ichthyornithes preceding the end-Cretaceous mass extinction event contrasts markedly with current hypotheses of rapid morphological evolution among crown birds in the immediate aftermath of the K–Pg mass extinction, emphasizing the hypothesized role of the K–Pg transition as a pivotal event in avian evolutionary history9,28,30,31,32.

Avian pterygoid evolution

Recent research has revealed several components of the ichthyornithine cranial kinetic system such as the quadrate and jugal24, as well as the palatine and hemipterygoid12. The details of the Janavis pterygoid described here represent the last crucial piece for understanding key aspects of the morphology and function of the palate in Ichthyornithes and other crownward stem-Euornithes. Scaling the pterygoid of Janavis to half its true size (roughly the same scaling factor exhibited by the humeri and scapulae of Ichthyornis and Janavis) reveals a remarkably precise fit within a nearly complete composite reconstruction of the Ichthyornis skull12,24 (Fig. 2 and Extended Data Fig. 10). Indeed, the articular facets for the quadrate and hemipterygoid–palatine complex on the Janavis pterygoid closely match the corresponding articular surfaces on those elements in Ichthyornis, consistent with the morphological similarities exhibited throughout the rest of the skeleton of these taxa (Extended Data Fig. 10c).

The morphology of the Janavis pterygoid is notably similar to that of extant galloanserans, combining features present in both Galliformes and Anseriformes, including the presence of large, well-defined and flat basipterygoid facets positioned near the rostral end of the element, a ball-and-socket palatine–pterygoid articular facet and a simple ovoid quadrate articular facet reminiscent of the condition in extant Anhimidae (Anseriformes)2,3,5 (Fig. 2 and Extended Data Fig. 4). Similar palatine and quadrate articular surfaces are found in some surveyed neoavians; however, the neoavian basipterygoid facet, when present, differs markedly from that of Janavis and galloanserans: it is proportionally minute and is generally situated approximately halfway along the caudal–rostral length of the pterygoid shaft. The functional implications of these differing basipterygoid positions and morphologies among major neognath lineages are unclear33, but we hypothesize a similar mechanical function for the pterygoids of Janavis and galloanserans given their marked geometric similarities. Indeed, our quantitative analysis of pterygoid geometry illustrates that Janavis falls extremely close to extant galloanserans along PC1–3, occupying a roughly intermediate position between Galliformes and Anseriformes in the three-dimensional morphospace (Fig. 3 and Extended Data Fig. 8). By contrast, palaeognaths exhibit wide variance in pterygoid shape, yet fall outside the range of shape variation exhibited by neognaths in the three-dimensional morphospace. See Supplementary Information for further discussion of the pterygoid and basisphenoid of extant birds and Mesozoic avialans.

Ancestral neornithine palate morphology

The pterygoid is a key component of the neornithine cranial kinetic system, forming a link between the mobile, caudally positioned quadrate and the rostrally positioned palatine2,3,5,7,34. This system transmits rostrally directed forces from the quadrate to the upper bill, which may result in dorsally directed bending of the upper jaw in taxa with adequately thin flexion zones, or discrete hinges formed between functional units of the bill5,33,35,36. The nature of the ichthyornithine pterygoid has never previously been described, and with the exception of highly autapomorphic pterygoids known from the hesperornithines Hesperornis23,34,37,38 and Parahesperornis11, that of Janavis represents the crownward-most stem bird pterygoid thus far reported.

The kinetic or akinetic nature of the ancestral neornithine palate has been the subject of substantial attention, and several studies have posited a ‘palaeognathous’ and weakly kinetic condition for the ancestral crown bird, implying that the acquisition of a more mobile palate represents a synapomorphy of Neognathae2,6,7. In the absence of complete palatal material known from crownward Mesozoic avialans, this hypothesis has rested on the nature of the fused palatal elements of non-avialan theropods, early avialans and enantiornithines7,39,40. The bony palates of these stemward avialans, characterized by elongate basipterygoid processes, fused pterygoid–palatine complexes and low, mediolaterally expanded pterygoids41,42,43,44 approximate long-standing interpretations of the ‘palaeognathous’ condition, differing substantially from the morphologies exhibited by extant neognaths. As noted by Torres et al.12, discussions on the early evolution of the neornithine palate have occasionally referenced hesperornithine palatal morphology; however, the highly apomorphic morphology of the hesperornithine pterygoid (along with the rest of the hesperornithine skeleton) impedes clear inferences regarding the plesiomorphic condition of the crown bird palate10. In addition, we acknowledge that the pterygoid constitutes only one of the major elements of the avian palate, and although the ball-and-socket intrapterygoid or pterygoid–palatine contact in Janavis might suggest a similar function to those of extant taxa exhibiting similar morphologies, the hemipterygoid–palatine complex differs in stem ornithurines with respect to that of extant neognaths. Indeed, although the palatine and the hemipterygoid are fused in adult neognaths3, these elements remained unfused in Hesperornithes11,38 and in Ichthyornis12, which may have had implications for palate function in these taxa. A detailed assessment of palate function in stem ornithurines is beyond the scope of this work but represents a promising avenue for future research.

Although the possibility that the palaeognathous palate was derived from a ‘neognath-like’ precursor has been previously suggested based on developmental observations4,45, more recent developmental research has rejected this hypothesis46. Nonetheless, the notably galloanseran-like pterygoid of Janavis, together with recent evidence of a neognath-like hemipterygoid–palatine complex in Ichthyornis12, rejects the hypothesis that the plesiomorphic condition of the neornithine palate was palaeognathous. Instead, these observations support the hypothesis that the extant palaeognathous condition is derived and convergent with the superficially similar morphologies found in stemward avialans and non-avialan theropods. Indeed, the pronounced morphometric disparity among extant palaeognath pterygoids (Fig. 3 and Extended Data Figs. 8 and 9) further emphasizes that the ‘palaeognathous condition’ may be little more than an anachronistic conceptual wastebasket, oversimplifying the substantial morphological variability within this taxonomically depauperate extant clade, and perhaps reflecting outdated assumptions that palaeognaths represent ‘primitive avian stock’ carrying over from the early days of systematic ornithology4. The recognition of a neognathous and probably galloanseran-like palate as the plesiomorphic condition for Neornithes suggests that extant Galloanserae may provide the best analogues among living birds of ancestral neornithine palate function and development.

Implications for Cenozoic avian oddities

Several disparate lineages of early Cenozoic fossil birds, such as Pelagornithidae14, Dromornithidae and Gastornithidae47, have been interpreted as representatives of total-clade Galloanserae, in part due to palatal features such as pterygoid morphology and the presence of sessile basipterygoid processes13,14,15,17,47. Recognition that these features of the palate may constitute neornithine plesiomorphies instead of galloanseran apomorphies may force reconsideration of the phylogenetic placement of some of these bizarre fossil birds, and the possibility that some of these groups might represent early diverging lineages of crown or near-crown stem birds instead of total-clade galloanserans cannot be dismissed15. Further research into the cranial morphology of these early Cenozoic lineages will be necessary to reevaluate their phylogenetic position and determine whether some of their purported galloanseran features could instead represent retained neornithine symplesiomorphies. Casting light on the evolutionary origins of modern birds and their characteristic features remains dependent on new avian discoveries from the Upper Cretaceous and lowermost Cenozoic, as exemplified by the new insights into the evolution of avian palate morphology provided by Janavis. Discoveries such as this one emphasize that even the most long-standing assumptions about the origins of crown birds are provisional until illuminated by direct fossil evidence.

Methods

Materials

NHMM RD 271, an unprepared, partial skeleton from the upper Maastrichtian of Eben Emael, province of Liège, north-east Belgium, was reported by Dyke et al.19, who suggested possible ichthyornithine affinities19,48 in a brief report. The specimen was deemed worthy of detailed restudy shortly after its original publication21, and its single isolated tooth was included in a comparative study of ornithurine tooth morphology49. Otherwise, the specimen has eluded further attention.

In a brief description based on the unprepared specimen, Dyke et al.19 noted several skeletal elements that, with the benefit of additional preparation and µCT scanning, we show were misidentified. For instance, the reported lower jaws, partial jugals and a possible quadrate19 all correspond to a portion of the thorax comprising two thoracic vertebrae and six associated ribs (Extended Data Fig. 2f). The reported partial right ulna19 instead corresponds to the caudal end of the right scapular blade, whereas the scapula itself was not identified (Extended Data Fig. 2c). A reported tarsometatarsus19 also appears to be absent. Among the elements fully extracted from the matrix by the collector, Dyke et al.19 reported a proximal coracoid, which is instead identified here as the rostral portion of the left pterygoid (Fig. 2 and Extended Data Figs. 2b and 4), and a proximal tarsal, which most probably corresponds to a partial pedal phalanx (Extended Data Fig. 2m).

Here we thoroughly reinvestigated NHMM RD 271. The specimen was further mechanically prepared at the Sedgwick Museum of Earth Sciences and µCT scanned at the Cambridge Biotomography Centre using a Nikon 49 Metrology XT H 225 ST high-resolution CT scanner. Several focused scans of the different regions of the main block and isolated elements were performed; scan parameters and further details are provided in the Supplementary Information. Scans were also obtained of a broad phylogenetic sample of crown birds from the collections of the University of Cambridge Museum of Zoology (see Supplementary Information for complete details). Scans were reconstructed and digitally segmented using VGSTUDIOMAX 3.30 and 3.40 (Volume Graphics) and Avizo 2019.3 (Thermo Fisher Scientific), to generate and export 3D-surface meshes of individual elements. Skeletal models of all elements in anatomical connection and morphological comparative plates were built in Autodesk Maya 2022.

Phylogenetic analyses

We tested the phylogenetic position of Janavis by incorporating it into two alternative morphological matrices originally published by Wang et al.50 and Torres et al.12, most recently updated and corrected by Benito et al.22. To further test the relationships among Janavis, Ichthyornis, other Mesozoic ornithurines and Neornithes, we expanded the sampling of Hesperornithes in both datasets and updated the scorings of hesperornithine taxa on the basis of recent work11,51,52,53. Details of the changes made to these matrices can be found in the Supplementary Information.

Phylogenetic analyses were conducted under both parsimony and Bayesian analytical frameworks to account for potential differences introduced by alternative optimality criteria. Parsimony analyses were conducted using TNT 1.5 (ref. 54) (made available with the sponsorship of the Willi Hennig Society). We performed an unconstrained heuristic search with equally weighted characters, with 1,000 replicates of random stepwise addition using the tree bisection reconnection algorithm. Ten trees were saved per replicate, and all most parsimonious trees (MPTs) were used to calculate a strict consensus. Bremer decay indices were calculated in TNT using tree bisection reconnection from existing trees. Bootstrap analyses were performed using a traditional search and 1,000 replicates, with outputs saved as absolute frequencies. We conducted Bayesian analyses with MrBayes55 using the CIPRES Science Gateway56, under the Mkv model57. Gamma-distributed rate variation was assumed to allow for variation in evolutionary rates across different characters. Analyses were conducted using four chains and two independent runs, with a tree sampled every 4,000 generations and a burn-in of 25%. Analyses were run for 30,000,000 generations, and analytical convergence was assessed using standard diagnostics provided in MrBayes (average standard deviation of split frequencies < 0.02, potential scale reduction factors = 1 and effective sample sizes > 200). Results obtained from independent runs of the same analyses were summarized using the sump and sumt commands in MrBayes. We exported the recovered tree topologies into TNT, and morphological synapomorphies were optimized onto those topologies under parsimony.

We recovered Janavis in an exclusive clade with Ichthyornis in all of our analyses, irrespective of the methods or matrices used. This clade, which fits the phylogenetic definition of Ichthyornithes21,22, occupies a relatively crownward position within Euornithes, but is stemward of Hesperornithes and Iaceornis, a position previously recovered for Ichthyornis in multiple studies22,24,58,59,60,61. Janavis represents the only well-known member of Ichthyornithes yet recognized other than Ichthyornis itself22. However, fragmentary remains from the Cenomanian of Russia62 and Egypt63, as well as the Maastrichtian of North America9 have been hypothesized to represent close relatives of Ichthyornis, and further work is needed to fully reveal the diversity, stratigraphic range and geographical distribution of this clade.

Geometric morphometric analyses

We performed three-dimensional geometric morphometric analyses of the pterygoids of Janavis and representatives of the three major extant neornithine subclades: Palaeognathae (five species), Galloanserae (eight species of Galliformes; eight species of Anseriformes, including the fossil stem anseriform A. oxfordi64) and Neoaves (11 species), as well as the fossil D. (‘Odontopteryx’) toliapica65,66 as a representative of the extinct clade Pelagornithidae, whose phylogenetic affinities remain uncertain13,14,15 (see Supplementary Information for the complete list of specimens included). CT data were obtained from specimens housed in the collections of the University of Cambridge Museum of Zoology (see ‘Materials’ section) and were supplemented by data from Project TEMPO on the online repository MorphoSource. Pterygoids were segmented in VGSTUDIOMAX 3.30 and 3.40 and Avizo 2019.3.

We developed a landmark and semi-landmark scheme to compare the shape of the pterygoid of Janavis with a broad phylogenetic assemblage of crown birds. We identified six quantifiable morphological characters of the pterygoid, focusing on the connection between the pterygoid and its neighbouring bones (that is, the quadrate and the hemipterygoid–palatine complex), the dorsal, medial and lateral ridges running along the pterygoid shaft, and the basipterygoid facet. Given the absence of basipterygoid processes in several neoavian lineages and the inherent difficulty of quantifying the shape of missing structures (see Supplementary Information for a detailed description and discussion of our landmarking scheme), only taxa exhibiting facets for the basipterygoid processes were included in our analyses. Shape variation was quantified using six discrete landmarks and six circular or curved pseudo-landmarks (see Extended Data Fig. 7 and Supplementary Information for a detailed description of each character).

To generate a matrix of equally spaced pseudo-landmarks, preliminary pseudo-landmarks were first placed on all specimens using Avizo 2019.3. Pseudo-landmarks were used to calculate the optimal number and position of circular pseudo-landmarks with equal spacing on each specimen using the geomorph package67 in the R statistical environment68,69. Landmark configurations for six landmarks and 386 pseudo-landmarks were included in a generalized Procrustes analysis using the ‘gpagen’ function in geomorph67, and the covariance matrix of Procrustes coordinates was subjected to PC analysis to produce a pterygoid shape morphospace using the ‘plotTangentSpace’ function in geomorph67.

Unlike most similar studies, which outline curved structures as a series of semi-landmarks to which a sliding algorithm is then applied69, we treated equivalent series of landmarks as pseudo-landmarks70, and sliding was not applied71. During our preliminary morphometric analyses, in which we treated all curves as composed of regular sliding semi-landmarks, we noticed that the sliding process heavily distorted the Procrustes shapes of certain taxa, notably flightless palaeognaths such as Struthio and Rhea, which exhibit highly divergent pterygoid morphologies with respect to the remainder of our sample (see main text). We identified the main cause of this artefactual distortion as the interaction between the landmarks and sliding semi-landmarks characterizing the shape of the medial crest of the pterygoid and the position of the basipterygoid process; when the medial crest was excluded from our morphometric analyses, very little distortion occurred. We attempted four variations of our morphometric analyses to address this issue: outlining both the medial crest and the basipterygoid facet with pseudo-landmarks (that is, sliding was not applied), whereas all other semi-landmark sequences were slid; excluding the medial crest and basipterygoid facets from our analyses; and combinations of both these approaches. Our overall results were robust to all of these analytical variations, with the major neognath clades (Galliformes, Anseriformes and Neoaves) forming discrete clusters on all three main PC axes. Janavis was recovered closest to Galloanserae in all alternative analyses, and palaeognaths were always found to exhibit a very wide morphospace distribution. In general, these alternative analyses returned very similar results, with the exception of the specific morphospace positions of the most morphologically divergent palaeognaths. We found that inclusion of the medial crest was important for characterizing the morphological difference between Galliformes and Anseriformes, as it incorporates the rostral process characteristic of Anseriformes. We therefore chose to retain the medial crest in the primary analyses reported in the main text, and treat all curves as sequences of pseudo-landmarks to avoid distortion caused by sliding. The results of our preliminary analyses treating all curves as composed of semi-landmarks, and those of all four alternative analytical variations described above, are provided as part of the Supplementary Information, including 2D plots for the first three PC axes and Procrustes distance matrices.

Given the unexpected distortion caused by the interaction between highly variable pterygoid morphologies and the semi-landmark sliding process noted above, we urge geometric morphometricians to exercise caution when interpreting morphospace distributions of particularly complex morphological elements using sliding semi-landmarks, and we recommend exploring alternative landmark treatments to test the robustness of morphometric results and the possible impact of distortion artefacts.

Estimation of body size

Measurements of the minimum humerus shaft diameter in cranial view (7.39 mm) and the minimum circumference of the humerus shaft (25.47 mm) were obtained from high-resolution µCT scans of NHMM RD 271. These measurements were incorporated into published scaling equations for estimating the body masses of volant birds72, yielding mean estimates of body size of 1,120 g and 1,504 g, respectively. These results were compared with similar estimates published for I. dispar22. Measurements of mass from extant comparative taxa, such as those mentioned in the main text, were obtained from Dunning Jr73.

Reporting summary

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