Triassic stem caecilian supports dissorophoid origin of living amphibians

Living amphibians (Lissamphibia) include frogs and salamanders (Batrachia) and the limbless worm-like caecilians (Gymnophiona). The estimated Palaeozoic era gymnophionan–batrachian molecular divergence1 suggests a major gap in the record of crown lissamphibians prior to their earliest fossil occurrences in the Triassic period2–6. Recent studies find a monophyletic Batrachia within dissorophoid temnospondyls7–10, but the absence of pre-Jurassic period caecilian fossils11,12 has made their relationships to batrachians and affinities to Palaeozoic tetrapods controversial1,8,13,14. Here we report the geologically oldest stem caecilian—a crown lissamphibian from the Late Triassic epoch of Arizona, USA—extending the caecilian record by around 35 million years. These fossils illuminate the tempo and mode of early caecilian morphological and functional evolution, demonstrating a delayed acquisition of musculoskeletal features associated with fossoriality in living caecilians, including the dual jaw closure mechanism15,16, reduced orbits17 and the tentacular organ18. The provenance of these fossils suggests a Pangaean equatorial origin for caecilians, implying that living caecilian biogeography reflects conserved aspects of caecilian function and physiology19, in combination with vicariance patterns driven by plate tectonics20. These fossils reveal a combination of features that is unique to caecilians alongside features that are shared with batrachian and dissorophoid temnospondyls, providing new and compelling evidence supporting a single origin of living amphibians within dissorophoid temnospondyls.

Article by features found in referred specimens: co-ossified maxilla and palatine (compound maxillopalatine); palatal dentition of maxillopalatine terminated anteriorly by the lateral choanal margin*; maxillopalatine without osteological correlate of the tentacular organ*; absence of internal and retroarticular processes of the pseudoangular*; jaw articulation surface of pseudoangular formed by a subcircular flat pad; pseudoangular bearing a dorsally exposed adductor chamber occupying more than 30% of pseudoangular length*; three cranial nerve V insertions in pseudoangular*; femur present. Differential diagnosis in Supplementary Information, section 2.

Phylogenetic relationships
We tested the relationships of Funcusvermis gilmorei in a modified dataset 6 of 63 terminal taxa including stem tetrapods, stem and crown amniotes, and temnospondyl amphibians including stereospondyls and lissamphibians (Methods). Using both maximum parsimony and Bayesian inference optimality criteria (Methods), our phylogenetic analyses robustly support Funcusvermis as the earliest-diverging gymnophionomorph, sister taxon to the clade including Eocaecilia, Rubricacaecilia and Gymnophiona. All analyses unambiguously recovered a monophyletic Lissamphibia nested within amphibamiform dissorophoids, with Gerobatrachus and Doleserpeton annectens as successive outgroups to Lissamphibia ( Fig. 3 and Extended Data Figs. 5-7). Our parsimony analysis recovered Lissamphibia consisting of a polytomous trichotomy of Gymnophionomorpha, Batrachia and Albanerpetontidae (Extended Data Fig. 5), whereas our Bayesian analysis recovered Lissamphibia consisting of a sister group relationship between Batrachia and a clade comprised of a sister group relationship between Gymnophionomorpha and Albanerpetontidae (Extended Data Fig. 7). The varying position of albanerpetontids in these and other recent analyses 28

Origins of the lissamphibian jaw apparatus
Funcusvermis indicates that many features of the lissamphibian mandibular ramus appeared initially in amphibamiform dissorophoids and were later lost or modified in batrachians, albanerpetontids and gymnophionomorphs (Fig. 3). All dentition in Funcusvermis ( Fig. 1) is pedicellate-the oldest known example of this distinctive tooth form in crown Lissamphibia-reinforcing hypotheses that pedicellate teeth are derived in amphibamiform dissorophoids 10 , conserved in gymnophionomorphs and batrachians 30 , and lost in albanerpetontids 29 . The rod-like pseudodentary of Funcusvermis resembles that of Eocaecilia 12 and the dentary of Doleserpeton 9 in the presence of tightly packed homodont tooth pedicels in parallel labial (dentary) and lingual (adsymphyseal) rows. The symphyseal foramen, of similar form and position to those of albanerpetontids 31 , suggests that the Meckel's cartilage never ossified at the mandibular symphysis, probably a conservation of the ancestral condition of temnospondyls 32 , and differing from the ossified condition of this element that forms a closed mandibular symphysis in batrachians and other gymnophionomorphs. As in Doleserpeton 9 , a vertical notch bisects the mandibular symphysis between the anterior termini of the dentary and adsymphyseal tooth rows forming medial and lateral processes in Funcusvermis ( Fig. 1  suggesting a transformation of the condition exhibited in Doleserpeton (5-7 teeth 9 ) through distal expansion via addition of new teeth. In living caecilians, the lingual tooth row forms embryonically on a distinct anlage that later ossifies to the medial surface of the dentary forming the medial part of the mandibular symphysis and the lingual tooth row of adult caecilians 33 . Recent identifications of a dorsally facing tooth-bearing adsymphyseal (equivalent to the parasymphyseal (plate)) medial to (and separate from) the dentary at the mandibular symphysis in early branching tetrapods 34,35 , juvenile temnospondyls 36 and dissorophoid temnospondyls 37 suggests that in taxa that appear to bear a lingual tooth row at the mandibular symphysis of the 'dentary' (for example, Doleserpeton, Funcusvermis and other gymnophionomorphs), the 'dentary' is actually composed of a tooth-bearing adsymphyseal (forming the lingual tooth row) co-ossified lingually to the dentary, and not a coronoid as previously thought 14,33 (Fig. 3 and Extended Data Fig. 8; see Supplementary Information, section 2 for discussion of adsymphyseal homology).
The pseudoangular of Funcusvermis is highly similar to the postdentary morphology of dissorophoids exemplified by the amphibamid Doleserpeton 9 (Fig. 3 and Extended Data Fig. 3); as in Doleserpeton, batrachians, and albanerpetontids, Funcusvermis lacks retroarticular and internal processes, suggesting their initial acquisition in the common ancestor of Eocaecilia and Gymnophiona. The absence of the retroarticular process and presence of a dorsally facing adductor fossa (insertion site of the m. adductor mandibulae complex (mAM)) occupying more than 30% of pseudoangular length (Supplementary Table 1) in the pseudoangular of Funcusvermis (Fig. 1) illuminate a major transformation from the ancestral lissamphibian condition to the unique musculoskeletal architecture of living gymnophionans. Jaw closure driven primarily by the mAM is ancestral for tetrapods, and is retained in batrachians, albanerpetontids and Funcusvermis, differing from the condition of all other gymnophionomorphs, which exhibit the distinctive caecilian dual jaw closure mechanism 15  Article DJCM is driven primarily by the hyobranchial muscle m. interhyoidus posterior (mIHP), and secondarily by the mAM; the mIHP inserts onto the ventral side of the retroarticular process and extends posteroventrally, acting as a first-order lever causing the anterior component of the lower jaw to pivot upwards with respect to the quadrate during jaw closure 15 . Acquisition of DJCM is hypothesized to be an adaptation for fossoriality: the mIHP contribution to bite force allows for reduction of the mAM and therefore compaction of the skull roof 15,16 , a suite of transformations shown to be acquired by the common ancestor of Eocaecilia and Gymnophiona. Although the skull roof of Funcusvermis is unknown, absence of the DJCM and our phylogenetic results suggest that it probably retained the plesiomorphic condition of cheek emargination (gymnokrotaphy, as in Gerobatrachus, batrachians, albanerpetontids and presumably the common ancestor of Lissamphibia) to accommodate the mAM, rather than a closed skull roof with large interpterygoid vacuities 13 (stegokrotaphy, as in dissorophoids). The obtuse angle of the orbital margin in the Funcusvermis maxillopalatine ( Fig. 1) may suggest the presence of large orbits as in dissorophoids, batrachians and albanerpetontids (differing from the reduced orbits of other gymnophionomorphs); however, the incomplete orbital margin in the single maxillopalatine specimen (PEFO 46481) prohibits conclusive assessment of this feature. The orbital margin of Funcusvermis lacks a tentacular fossa or aperture (osteological correlates for the chemosensory tentacle organ 18 ), suggesting its absence in early gymnophionomorphs and later derivation by the common ancestor of Eocaecilia and Gymnophiona 12 . The presence of a co-ossified maxilla and palatine (maxillopalatine) in Funcusvermis is shared with gymnophionans and differs from that of amphibamiforms, albanerpetontids and batrachians, evidence of maxillopalatine consolidation early in gymnophionomorph evolution; however, these bones are possibly separate in Eocaecilia 12 and Rubricacaecilia 23 . Ventrally, the maxillopalatine of Funcusvermis bears parallel maxillary and palatal rows of tightly packed pedicellate teeth of similar size to those in the pseudodentary, seemingly intermediate between the condition of these dentitions in Doleserpeton and Eocaecilia, sharing an anterior truncation of the palatal tooth row by the internal nares with the former, and mesiodistal distal extension (through addition of new teeth) of the palatal row with the latter. A comprehensive comparative description of the Funcusvermis skull and postcranial elements is included in Supplementary Information, section 2.

Evolution of caecilian fossoriality
Given our phylogenetic results, the ecological habits of Funcusvermis may be transitional between terrestrial amphibamid dissorophoids and fossorial gymnophionans. The compound bones in the compact skull of fossorial gymnophionans are thought to withstand the forces associated with head-first burrowing 17 , and at least some are present in Funcusvermis (for example, maxillopalatine). Small pits covering the lateral surfaces of the pseudodentary and maxillopalatine in Funcusvermis are also found in Eocaecilia 12 , Rubricacaecilia 23 and gymnophionans 38 . External structure and internal microanatomy of these pits revealed by osteohistological sectioning of a Funcusvermis pseudodentary (PEFO 44432) show a marked resemblance to those of studied living caecilians (Extended Data Fig. 4), in which these pits act as anchor sites for collagen networks forming a tight skin-to-bone attachment and house glands that produce a lubricating mucus secretion, functions thought to aid in subterranean burrowing 39 . The dorsally flattened neural arch of the Funcusvermis postatlantal pleurocentrum (PEFO 45810; Extended Data Fig. 3) resembles those of Rubricacaecilia 23 , suggesting the acquisition of a tubular trunk, a feature crucial for underground locomotion in living caecilians 40 . These morphologies in Funcusvermis illustrate acquisition (by at least the Late Triassic) of some features that now facilitate fossoriality in living caecilians, later followed by acquisition of the DJCM and tentacular organ in Eocaecilia, and finally loss of the appendicular skeleton in gymnophionans.

Biogeography of early caecilians
The spatiotemporal occurrence of Funcusvermis empirically establishes lissamphibian geographic origins on the Pangaean supercontinent before its fragmentation 20 , and the similar palaeogeography of Eocaecilia 12 to Funcusvermis suggests the non-gymnophionan gymnophionomorph origin may lie in the early Mesozoic era of equatorial central Pangaea. The occurrence of Rubricacaecilia in the Early Cretaceous epoch of equatorial Gondwana may further support this hypothesis, suggesting non-gymnophionan gymnophionomorph distribution across both Laurasian and Gondwanan components of Pangaea in the early Mesozoic prior to its breakup 23 . The equatorial provenance of Funcusvermis adds to an exclusively equatorial pattern of gymnophionomorph distribution: all fossil occurrences fall between a minimum of approximately 16° N and 27° S ( Fig. 2 and Supplementary Table 2), and living caecilians are restricted to equatorial latitudes 19 between 27° N and 34° S. The tropical distribution of extant gymnophionans is notably disjunct from non-gymnophionan gymnophionomorph fossil occurrences in present-day western North America and Morocco (Fig. 2). Drift of the North American and African plates during the Mesozoic 41 may explain the extirpation of gymnophionomorphs from these areas later in the Phanerozoic as these previously humid palaeotropical regions moved north into the arid subtropics. Concurrently, the northern drift of Gondwana into the palaeotropics may have expanded suitable terrestrial habitats, consistent with molecular evidence of an early Mesozoic Gondwanan origin of gymnophionans 20 .
The earliest batrachians hail from the Triassic of southern 2 , equatorial 4,5 and northern 3,6 Pangaea (Supplementary Table 3), indicating extensive latitudinal dispersal by at least the Middle Triassic epoch; this pattern is further reflected in the subsequent batrachian fossil record and their extant distribution. Unlike in extant batrachians, evaporative water loss is found to be a critical physiological constraint in living caecilians, limiting their distribution to humid environments near the equator 19 . The contrasting spatiotemporal histories of batrachians and gymnophionomorphs suggest a divergence of physiological constraints linked to humidity prior to the Triassic; conserved physiological traits in these groups may explain subsequent patterns of dispersal reflected in present-day lissamphibian biogeography.

Timing of lissamphibian origins
Prior to the results of this study, the chronology of lissamphibian origins remained unresolved owing to the reliance of molecular clock estimates on different node minima derived from competing phylogenetic hypotheses that include extinct taxa 1 . Funcusvermis lends novel and strong support for a monophyletic origin of living amphibians within dissorophoid temnospondyls 30,42 (the 'classic' temnospondyl hypothesis), and thus the molecular clock estimates of caecilian-batrachian divergence using the temnospondyl hypothesis. Additionally, the recovery of Gerobatrachus as the sister taxon to Lissamphibia in our analysis suggests that taxon may not be a stem batrachian 7

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-022-05646-5. Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Gymnophionomorpha Marjanović and Laurin 2008
Remarks. Gymnophionomorpha is defined here as the total group consisting of Caecilia tentaculata and all taxa that share a more recent common ancestor with it than with Salamandra salamandra Linnaeus, 1758, Rana temporaria Linnaeus, 1758, and Albanerpeton inexpectatum, Estes and Hoffstetter 1976. This newly proposed stem-based definition of Gymnophionomorpha is modified after that originally proposed 43 .

Assignment of elements
Although all specimens assigned to Funcusvermis were found as isolated, dissociated elements, their assignment to a single gymnophionomorph taxon is supported by: (1)

Geological framework
The blue-coloured strata of the upper Blue Mesa Member of the Chinle Formation were deposited in a northwest-flowing fluviolacustrine system on the western margin of central Pangaea at a palaeolatitude of 5° to 15° N in a humid monsoonal climate 26 . Detrital zircon U-Pb radiometric ages provide robust geochronologic constraints on the Chinle Formation, bracketing deposition of the upper Blue Mesa Member 24,25 to ~223-218 Ma (Extended Data Fig. 2). The gymnophionomorph fossils described herein were collected from the Thunderstorm Ridge locality (PFV 456) near the Puerco River in PEFO, Arizona, USA (Extended Data Fig. 2). The fossiliferous unit is a 15-cm-thick, poorly sorted siltstone horizon, bearing a dense concentration of carbonate nodules, angular intraformational clasts, micro-and macrovertebrate bones and coprolites. PFV 456 has yielded a diverse assemblage of vertebrates including chondrichthyans, actinopterygians, dipnoans, coelacanths, metoposaurids, salentians 5 , drepanosauromorphs 44 , lepidosauromorphs, archosauromorphs 45 , pseudosuchian archosaurs 46 , dinosauromorphs 47 and cynodonts 48 . The lack of abrasion and polishing and the exceptional three-dimensional preservation of extremely delicate microvertebrate bones indicates initial deposition in a low-energy setting, followed by brief reworking and redeposition in a channel avulsion event that incorporated angular intraformational clasts and carbonate nodules into the fossiliferous layer. This sedimentological evidence in combination with the presence of abundant spinicaudatan exoskeletons, unionid bivalve steinkerns and obligate-aquatic, amphibious and fully terrestrial vertebrates indicates initial deposition in a marginal lacustrine palaeoenvironment occupied by a diverse vertebrate community.

Collection and preparation methods
The hypodigm and all referred specimens were collected by screenwashing fossiliferous matrix from PFV 456 (9 out of 11 fossil gymnophionomorph occurrences were recovered using screenwashing; Supplementary Table 2). Blocks of matrix weighing approximately 1.8-3.2 kg were individually disaggregated in water and subsequently washed through a series of wire mesh screens with a minimum screen opening of 0.5 mm (no. 35 mesh). Dividing the fossiliferous concentrate from each block into smaller fractions in this way accelerated the process of picking. The resulting concentrate fractions were picked using a dissecting microscope resulting in the identification and separation of all Funcusvermis specimens. Importantly, through processing individual blocks of matrix, Funcusvermis elements that fragmented into multiple parts during the screenwashing process could be re-associated after microscopic sorting. Elements found as multiple broken pieces were subsequently reassembled by adhering matching fractured surfaces using cyanoacrylate, typically a low viscosity PaleoBOND or Loctite brand. To facilitate rapidly and precisely adhering these miniscule fragments together we created a mechanism that combines aspects of a jeweler's block ball vice, and a hobbyist tool, sometimes called a third hand or helping hand. It combines a socket made of wood or closed cell polyethylene foam and a hemispherical wooden ball to create a pivot that can turn or tilt in all directions. This is topped with a small rectangle of wood with a small concave arch cut into it to provide a workspace. Insect pins are slid through channels in the wood filled with soft microcrystalline wax, which allows the pins freedom of movement, but the resistance needed to precisely position the fossil fragments. The fragments are temporarily adhered to the pin tips with more microcrystalline wax. Adhesive was applied to the joint between fragments as a microdroplet suspended on a single filament such as a cotton fibre and drawn into the joint via capillary action leaving a minimum of excess residue. Reassembly took place under a variety of Leica and Wild binocular microscopes, primarily MZ6, MZ12 and M8 models, varying in power from a maximum of ×40-×80 magnification.
To reveal the details of the pseudodentary dentition of Funcusvermis, matrix covering the dentition and other anatomy of PEFO 45800 was prepared through the following process. Melted cyclododecane (CDD) was poured into a shallow ceramic watch glass and allowed to harden. A small trench the size of the specimen was excavated, and the specimen was placed in the trench in the desired orientation. A Ukrainian kistky (a wax pen), was used to melt the CDD around the specimen and allowed it to adhere to and support the specimen. Then matrix was removed using a 1/32 inch (0.79375 mm) carbide-needle in a pin vice primarily under high magnification under a Leica MZ12 and MZ6 microscope. The point of the needle was ground to a superfine conical point at about 10°-15° parallel to the shaft and flattened briefly along one side to provide an edge to remove adhesives. Some of the softer clay particles were removed with a porcupine quill. When needed, the specimen was consolidated with a very dilute solution of polyvinyl butyral (Butvar B-76) in acetone; the solution was mixed by eye, applying a bit to another vertebrate bone fragment and looking for sheen upon drying. Any excess Butvar film was removed by abrasion with the porcupine quill. The specimen was rotated in the CDD by trenching around the specimen until it was loose, shifting it, and then remelting the resulting CDD powder with the kistky. After all matrix was removed, the specimen was trenched out a final time and set aside in the fume hood to allow the CDD to sublimate.

Digital photography methods
Photographs of PEFO 45800 in Fig. 1 were acquired using a Leica MZ67 stereomicroscope and a Sony NEX-5T digital camera. Image stacking was conducted in Adobe Photoshop CC (https://www.adobe.com/ products/photoshop.html).

Digital reconstruction methods
A composite reconstruction of a partial skull of Funcusvermis (Fig. 1) was produced using Meshmixer 3.5 (https://meshmixer.com). Digital 3D surface meshes representing the anterior (PEFO 43891; light pink in Fig. 1a-c) and posterior (PEFO 46284; dark pink in Fig. 1a-c) portions of a pseudodentary were scaled to the same dorsoventral height, and both specimens were overlapped to form a composite reconstruction of a complete pseudodentary. The pseudoangular (PEFO 46480) and maxillopalatine (PEFO 46481) were scaled to match the size of the reconstructed pseudodentary, and anatomically positioned relative to the pseudodentary to approximate their position in an articulated three-dimensional skull. A surface volume file (3D mesh) of the composite skull reconstruction is available for download under project 000382289 on Morphosource.org (https://www.morphosource.org/ projects/000382289?locale=en).

Osteohistology methods
PEFO 44432 (right pseudodentary) was embedded in clear epoxy (Castolite AP), cut into 1 mm sections, and then ground to a ~100 µm thickness in the Virginia Tech Fossil Preparation Lab. Images of the histologically sectioned pseudodentary slide used in Extended Data Fig. 4 were acquired using a Sony NEX-5T digital camera mounted on a Nikon OPTIPHOT-POL Polarizing microscope. Fracturing of the specimen occurred during osteohistological preparation, causing fracture planes apparent in histological imaging (Extended Data Fig. 4).

Phylogenetic methods
See 'Code availability' to access and download phylogenetic matrix and analysis scripts.

Taxon sampling
Recent analyses recovered gymnophionomorphs at variable positions within Tetrapoda dependent on character and taxon sampling, including: (1) as 'microsaur' 'lepospondyls' 49 (note that taxa formerly included in 'Lepospondyli' are now understood as polyphyletic 50 ); (2) as stereospondyl temnospondyls forming the sister group to C. jenkinsi 14 ; (3) as 'microsaurian' or aïstopod 'lepospondyls' 51 ; and [4] as amphibamiform dissorophoid temnospondyls forming the sister group to batrachians 6,13 . The matrix of Schoch et al. (2020), recently used to hypothesize the phylogenetic position of the stem salamander Triassurus sixtelae and the origin of lissamphibians, was selected to test the phylogenetic relationships of F. gilmorei given its comprehensive sampling of taxa proposed to be sister groups to Gymnophionomorpha including stem and crown amniotes, stereospondyl and dissorophoid temnospondyl amphibians, batrachians, gymnophionomorphs and albanerpetontids. F. gilmorei was coded into the modified Schoch et al.  Table 9). The final matrix includes 355 morphological characters (Full character list in Supplementary  Information, section 8; see 'Code availability' to access and download phylogenetic matrix and analysis scripts).

Maximum parsimony and Bayesian analysis
All characters were equally weighted and unordered in both analyses following previous versions 6,14 . The character-taxon matrix was first analysed in the phylogenetic analysis software package TNT 1.5 (ref. 52 ) using New Technology Search options with the following parameters: ratchet (1,000 iterations), sectoral search (1,000 rounds), tree fusing (100 rounds), and random additional sequence (1,000 replicates). A total of 71 most parsimonious trees of 1,468 steps each were recovered (consistency index = 0.287; retention index = 0.675). A strict consensus tree calculated from the most parsimonious trees is presented in Extended Data Fig. 5. Bootstrap support values were obtained using TNT 1.5, and a strict consensus topology of trees produced via 1,000 bootstrap replicates resampled with replacement is presented in Extended Data Fig. 6. A Bayesian inference analysis of the charactertaxon matrix was conducted in the phylogenetic software package MrBayes v.3.2.6 (ref. 53 ) with the Mkv 54 model and gamma rate variation and the following parameters: four runs (six Markov chain Monte Carlo chains each), sampled every 1,000 generations, for 10 million generations with a relative burn-in of 0.25. Convergence of independent runs was assessed using Tracer v.1.76.1 (http://beast.bio.ed.ac.uk/Tracer). A consensus cladogram with mapped posterior probability values is presented in Extended Data Fig. 7.

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

Data availability
The holotype, paratypes and referred specimens of F. gilmorei are catalogued and available for study to qualified researchers at PEFO. Computed tomographic scan data, including surface volume files (3D meshes) and raw CT data of Funcusvermis specimens mentioned in the main text and extended data figures (including the holotype, paratypes and referred specimens), as well as a surface volume file of the composite skull reconstruction of Funcusvermis are available for download under project 000382289 on Morphosource.org (https:// www.morphosource.org/projects/000382289?locale=en).

Code availability
Code for TNT and MrBayes scripts used in the phylogenetic analyses conducted herein are available in Supplementary Information     Extended Data Fig. 9 | Phylogenetic distribution and comparative morphology of a lateral exposure of the palatine (LEP) and the loss of the lacrimal in Temnospondyli. a, Phylogenetic topology (derived from refs. 14,55 ) with variable presence of the lacrimal and the LEP (See Supplementary  Information 3 for further discussion of this feature). Comparative silhouettes of skulls in dorsal b-h and lateral i-k views; the prefrontal is coloured yellow here as it is the main element to positionally compensate for an absent lacrimal. b, The dvinosaur Thabanchuia oomie 64 (Illustration reproduced with permission from Cambridge University Press from ref. 64 ). c, The amphibamid Doleserpeton annectens 9 (Illustration adapted from ref. 9