The end-Devonian to mid-Mississippian time interval has long been known for its depauperate palaeontological record, especially for tetrapods. This interval encapsulates the time of increasing terrestriality among tetrapods, but only two Tournaisian localities previously produced tetrapod fossils. Here we describe five new Tournaisian tetrapods (Perittodus apsconditus, Koilops herma, Ossirarus kierani, Diploradus austiumensis and Aytonerpeton microps) from two localities in their environmental context. A phylogenetic analysis retrieved three taxa as stem tetrapods, interspersed among Devonian and Carboniferous forms, and two as stem amphibians, suggesting a deep split among crown tetrapods. We also illustrate new tetrapod specimens from these and additional localities in the Scottish Borders region. The new taxa and specimens suggest that tetrapod diversification was well established by the Tournaisian. Sedimentary evidence indicates that the tetrapod fossils are usually associated with sandy siltstones overlying wetland palaeosols. Tetrapods were probably living on vegetated surfaces that were subsequently flooded. We show that atmospheric oxygen levels were stable across the Devonian/Carboniferous boundary, and did not inhibit the evolution of terrestriality. This wealth of tetrapods from Tournaisian localities highlights the potential for discoveries elsewhere.
The term ‘Romer’s Gap’ was coined1,2 for a hiatus of approximately 25 million years (Myr) in the fossil record of tetrapods3, from the end-Devonian to the mid-Mississippian (Viséan). Following the end-Devonian, the earliest terrestrial tetrapod fauna was known from the early Brigantian (late Viséan) locality of East Kirkton near Bathgate, Scotland4,5. By that time, tetrapods were ecologically diverse, and were terrestrially capable. With five or fewer digits, some had gracile limbs6,7—unlike the polydactylous, predominantly aquatic, fish-like tetrapods of the Late Devonian8. Fossils representing transitional morphologies between these disparate forms were almost entirely lacking, limiting both understanding of the acquisition of terrestrial characteristics and the relationships between the diverse mid-Carboniferous taxa. Alternative hypotheses to explain the hiatus have included a low oxygen regime9 or lack of successful collecting in Tournaisian strata2.
Although isolated tetrapod limb bones, girdle elements and trackways are known from the Tournaisian of the Horton Bluff Formation at Blue Beach, Nova Scotia10,11, only a small fraction has been fully described12. The only other Tournaisian tetrapod material was the articulated skeleton of Pederpes finneyae, from the Tournaisian Ballagan Formation near Dumbarton, western Scotland13,14. More recently, new taxa from this formation in the Scottish Borders region were reported2. Further collecting from five localities (Supplementary Fig. 1) has since produced more data about the fauna, its environment and climatic conditions.
Our analysis shows that the Tournaisian included a rich and diverse assemblage of taxa, which included close relatives of some Devonian forms on the tetrapod stem and basal members of the amphibian stem. We diagnose, name and analyse five taxa (Figs 1,2,3,4,5), and summarize at least seven others that are distinct but undiagnosable at present (Fig. 6 and Supplementary Figs 2–6).
Tetrapods occupied a mosaic of juxtaposed microhabitats including ponds, swamps, streams and floodplains, the last of these with highly variable salinity and water levels in a sharply contrasting seasonal climate. Their fossils are most closely associated with palaeosols and the overlying sandy siltstones. These indicate exposed and vegetated land surfaces that were then flooded15,16 (Supplementary Fig. 7). This varied environment persisted over the 12 Myr of the Tournaisian3. In contrast to a previous study9, we show that atmospheric oxygen levels were stable across the Devonian/Carboniferous boundary and did not therefore compromise terrestrial faunal life.
Differential diagnoses below give the characters in which each taxa differs from all other tetrapods in its combination of autapomorphic and derived (relative to Devonian taxa) characters.
Tetrapoda Goodrich, 1930 indet.
Perittodus apsconditus gen. et sp. nov. Clack and Smithson T.R. (Fig. 1c–g).
Smithson et al., 2012, ‘new taxon A’.
Etymology. Genus from perittos (Greek) ‘odd’ and odus (Greek) ‘tooth’ referring to the unusual dentition of the mandible. Species from apsconditus (Latin) ‘covert, disguised, hidden, secret or concealed’, referring to the fact that key parts were only discovered by micro-computed tomography (micro-CT) scanning.
Holotype. UMZC 2011.7.2a and b. Cheek region of skull, lower jaw, and postcranial elements in part and counterpart.
Locality. Willie’s Hole, Whiteadder Water near Chirnside.
Horizon. Ballagan Formation. Early mid-Tournaisian.
Diagnosis. Autapomorphies: unique adsymphysial and coronoid dentition—adsymphysial with two tusks and at least two smaller teeth, anterior coronoid with two or three larger tusks, middle coronoid with two larger and two or three smaller teeth, posterior coronoid row of small teeth; lozenge-shaped dorsal scales bearing concentric ridges centred close to one edge nearer to one end. Derived characters: deeply excavated jugal with narrow suborbital bar; lateral line an open groove on jugal.
Plesiomorphies and characters of uncertain polarity. No mesial lamina of postspenial (state of angular not known); 35 dentary teeth including spaces; 29 maxillary teeth including spaces; room for possibly six teeth on premaxilla; marginal teeth similar in size; short broad phalanges, rounded unguals longer than wide with ventral ridge.
Attributed specimen. UMZC 2016.1. Isolated dentary and adsymphysial (in micro-CT scan) from Burnmouth Ross end cliffs, 373.95 m above the base of the Ballagan Formation. Mid-Tournaisian.
Remarks. Lower jaw length 68 mm. Maxilla of holotype visible in micro-CT scan. UMZC 2016.1 is almost identical in size and dentition to the holotype. The pattern is most similar, but not identical, to that of the Devonian taxon Ymeria17. A distinct denticulated ridge on the prearticular is set off from the remainder of the bone by a ventral groove. Radius and ulna are of approximately equal length. A partial ischium reveals similarities to Baphetes18.
Koilops herma gen. et sp. nov. Clack and Smithson T.R. (Fig. 1a,b).
Smithson et al., 2012, ‘probable new taxon’.
Etymology. Genus from koilos (Greek) ‘hollow or empty’, and ops (Greek) ‘face’, referring to the skull mainly preserved as natural mould. Species from herma (Greek) ‘boundary marker, cairn, pile of stones’. The specimen, from the Scottish Borders region, has transitional morphology between Devonian and Carboniferous tetrapods.
Holotype. NMS G. 2013.39/14. Isolated skull mainly as a natural mould.
Locality. Willie’s Hole, Whiteadder Water near Chirnside.
Horizon. Ballagan Formation. Early mid-Tournaisian.
Diagnosis. Autapomorphies: fine irregular dermal ornament with conspicuous curved ridges around the parietal foramen and larger pustular ornament anterior to parietal foramen. Derived characters: deeply excavated jugal with narrow suborbital bar; large parietal foramen.
Plesiomorphies and characters of uncertain polarity. Orbit oval with slight anterior embayment; prefrontal–postfrontal contact narrow, anterior to orbit mid-length; about eight premaxillary teeth recurved, sharply pointed, ridged towards base; closed palate, denticulated pterygoid; vomers bearing tusks and smaller teeth, at least four moderately large teeth on palatine; short rounded snout, only slightly longer than maximum orbit length.
Remarks. Skull length 80 mm. The dermal bones are robust and well integrated so the individual was almost certainly not a juvenile.
Ossirarus kierani gen. et sp. nov. Clack and Smithson T.R. (Fig. 2).
Etymology. Genus from ossi (Latin) ‘bones’ and rarus (Latin) ‘scattered or rare’. Species to honour O. Kieran and B. Kieran, representing the Burnmouth community, who have supported us and encouraged local interest and cooperation.
Holotype. UMZC 2016.3. A single block containing scattered skull and postcranial remains.
Locality. Burnmouth Ross end cliffs.
Horizon. 340.5 m above the base of the Ballagan Formation. Mid-Tournaisian.
Diagnosis. Autapomorphies: tabular elongate triangle forming a conspicuous tabular horn with a convex lateral margin. Derived character: tabular-parietal contact; exoccipital separate from basioccipital.
Plesiomorphies and characters of uncertain polarity. Jugal with extensive posterior component, with anteriorly placed shallow contribution to orbit; lozenge-shaped interclavicle; humerus with elongate and oblique pectoralis process comparable with the ventral humeral ridge of elpistostegalians and Acanthostega; multipartite vertebrae with diplospondylous widely notochordal centra and neural arches as unfused bilateral halves.
Remarks. Estimated skull length 50 mm on the basis of comparisons with Acanthostega, Ichthyostega and Greererpeton19,
Diploradus austiumensis gen. et sp. nov. Clack and Smithson T.R. (Fig. 3).
Etymology. Genus from diplo (Greek) ‘double’ and radus (Greek) ‘row’ referring to the double coronoid tooth row. Species from austium (Latin) ‘mouth of a river or stream’ referring to Burnmouth.
Holotype. UMZC 2015.30. Small disrupted skull with lower jaw, palate and skull roofing bones.
Locality. Burnmouth Ross end cliffs.
Horizon. 373.95 m above the base of the Ballagan Formation. Mid-Tournaisian.
Diagnosis. Autapomorphies: lower jaw with irregular double row of denticles along the coronoids; around 51 dentary teeth and spaces, with enlarged tusk at position 3 and the largest teeth in positions 8–13; parietals short, pineal foramen anteriorly placed; ?narrow curved pre- and postfrontals. Derived characters: deeply excavated jugal with narrow suborbital bar; parasphenoid with broad, flattened posterior portion with lateral wings, earliest known occurrence of a parasphenoid crossing the ventral cranial fissure, cultriform process flat, narrow.
Attributed specimen. UMZC 2016.4 a and b. The anterior end of a mandible from 341 m above the base of the Ballagan Formation at Burnmouth.
Plesiomorphies and characters of uncertain polarity. Unsutured junction between prearticular and splenial series; adductor fossa dorsally placed; adsymphysial plate possibly lacking dentition; closed, denticulated palate; broad pterygoid, quadrate ramus narrow with vertically orientated medial ascending lamina; ossified hyobranchial elements; maxilla and premaxilla with spaces for 35 and 10–12 teeth respectively; maxilla–premaxilla contact narrow, lacking interdigitations; dermal ornament with low profile, irregular on skull table, ridged on squamosal and quadratojugal.
Remarks. Lower jaw length 30 mm, superficially resembling that of Sigournea22, although a relationship is not supported by cladistic analysis. The thinness of the bones and their distribution suggest a juvenile.
Aytonerpeton microps gen. et sp. nov. Otoo, Clack and Smithson T.R. (Fig. 4).
Etymology. Genus name from Ayton, the parish in the Scottish Borders from which the specimen came, and erpeton (Greek) ‘crawler’ or ‘creeping one’. Species from micro (Greek) ‘small’ and ops (Greek) ‘face’.
Locality. Burnmouth Ross end shore exposure.
Horizon. 340.6 m above the base of the Ballagan Formation. Mid-Tournaisian.
Diagnosis. Autapomorphies: two enlarged premaxillary teeth plus one large tooth space at posterior end of premaxilla; five teeth on premaxilla; adsymphysial with a single tooth; coronoids apparently lacking shagreen; L-shaped lacrimal; vomer with at least one tooth, palatine with one large fang but lacking smaller teeth; ectopterygoid with at least two teeth and possible smaller teeth. Derived characters shared with colosteids: course of lateral line on maxilla and nasal; dentary teeth larger and fewer than upper marginal teeth; single large Meckelian fenestra; interpterygoid vacuities longer than wide; single large parasymphysial fang on dentary; ilium with a single strap-shaped iliac process.
Remarks. Reconstructed skull length about 50 mm. Other distinguishing features: short snout, approximately similar in length to orbit diameter; naris and choana both very large relative to skull size—larger than in Greererpeton. The enlarged premaxillary teeth prefigure those of more derived colosteids21, but the dentary lacks the corresponding reciprocal notch. This appears to be an early expression of a feature that becomes more elaborate in later taxa. All coronoids bear at least one tooth. Some colosteids lack coronoid teeth, and instead bear shagreen, a variable condition among individuals23. The small size of the skull but the strong integration of the lower jaw bones suggest a subadult or adult in which case the large orbit is unlikely to be a juvenile feature (see juvenile Greererpeton CMNH 11095)24. Its gracile limbs, metapodial bones and phalanges resemble Colosteus rather than Greererpeton. Clavicular ornament is similar to that of other colosteids25,26. The single iliac process is shared with other colosteids and with temnospondyls. This is the earliest known occurrence of this feature.
We performed parsimony and Bayesian analyses of a new data matrix (character list and data matrix in Supplementary Data) incorporating the new tetrapods. No taxon could be safely deleted27. Parsimony with all characters unordered and equally weighted produced 4,718 shortest trees, a poorly resolved strict consensus (Fig. 5 and Supplementary Fig. 8) and moderate branch support.
Four parsimony analyses with implied weighting, each using a different value (3, 4, 5, 10) of the concavity constant28 (K) produced many fewer trees (Fig. 5a,b), with novel topologies and increased stability for most of the new taxa. In these analyses, the relative positions of Ossirarus, Perittodus and Diploradus remain unaltered (Methods and Supplementary Fig. 8). Except in the analysis with K = 10, Koilops and Aytonerpeton emerge as stem amphibians29,
We also performed a Bayesian analysis (Fig. 5c). The results were largely similar to the parsimony analysis, except for the position of Ossirarus. In the Bayesian analysis, Ossirarus appears as a stem amniote, whilst Perittodus, Diploradus, Koilops and Aytonerpeton are stem tetrapods.
Despite inconsistencies, these results imply a substantial reshuffling of the branching sequence of Carboniferous stem tetrapods relative to previous studies29,
Geology and environment
The Ballagan Formation (Inverclyde Group) underlies much of the Midland Valley of Scotland and the northern margin of the Northumberland Basin. At Burnmouth, the vertically dipping strata probably span the entire Tournaisian2,34. Environmental interpretation was based on a 490 m core from a borehole through the formation, a complete logged succession at centimetre-scale intervals through 520 m at Burnmouth and an 8 m section at Willie’s Hole (Fig. 6, Methods and Supplementary Fig. 7).
Perittodus apsconditus occurs within a 6 cm thick laminated grey siltstone16 that contains a network of cracks filled with sandy siltstone identical to that of the overlying bed. Occurring within laminated siltstones, this may record an autochthonous lake dweller. Associated fossils comprise plants, actinopterygians, myriapods and ostracods. Koilops occurs within a unit comprising four beds of alternating black and green siltstone in which abundant palaeosol clasts indicate erosion and transport of land-surface sediment during flooding events.
Diploradus occurs in a 40 cm thick, bedded, black sandy siltstone that lies between pedogenically modified grey siltstones. Associated fossils comprise fish scales, abundant plant fragments, megaspores and shrimp and scorpion cuticle.
Ossirarus and Aytonerpeton occur within a complex 15 cm thick grey-black sandy siltstone that overlies a gleyed palaeosol and grades upwards into a laminated grey siltstone with brecciation cracks (Fig. 6, Methods and Supplementary Fig. 7). Ossirarus occurred just above the palaeosol in a light grey clay-rich sandy siltstone, whereas Aytonerpeton occurred within an overlying black sandy siltstone with abundant plant material. Associated fauna comprise abundant plants, megaspores, unusually abundant rhizodont bones and scales, actinopterygians, chondrichthyans (Ageleodus, gyracanthids), dipnoans, eurypterids and ostracods.
An association between wetland palaeosols and tetrapod-bearing facies has emerged from our studies, which is important because those horizons indicate a vegetated land surface (Fig. 6)15,16. The floodplain environments of semi-permanent water bodies, marsh, river banks and areas of dry land with trees were laid down at a time of change in the land plant flora of the Mississippian, following the end-Devonian extinctions. The new flora initiated a change in fluvial and floodplain architecture35,
Atmospheric oxygen levels in the Tournaisian
To address the low oxygen hypothesis9, we examined fossil charcoal (fusinite) in the Ballagan Formation to compare atmospheric oxygen levels in the Tournaisian with the Late Devonian and later Mississippian.
Charcoal, either as microscopic dispersed organic matter (DOM) or visible in hand specimens, is relatively common at Burnmouth and Willie’s Hole. Although charcoal is reported from the Tournaisian Horton Bluff Formation, Nova Scotia38, as indicating O2 concentrations above 16%, no quantitative study to validate this result has been undertaken.
We analysed DOM from 73 rock samples from Burnmouth shore and Willie’s Hole. For comparison with wildfire activity before and after Romer’s Gap, we also analysed 42 samples from the Viséan of East Fife, Scotland (Strathclyde Group) and nine samples from the Famennian of Greenland (Stensiö Bjerg Formation) (Supplementary Fig. 9 and Supplementary Table 1). All were found to contain fusinite, with a mean abundance relative to total phytoclasts of 2.2%, 2.3% and 2.6% for the Famennian, Tournaisian and Viséan, respectively. We also analysed 12 samples from Willie’s Hole, which had a mean value of 2.0% (Supplementary Table 1). These results mean that fire activity persisted through Romer’s Gap and indicate that atmospheric O2 did not fall below 16%; they also show that there was no substantial change in charcoal production compared with the Famennian and Viséan (Supplementary Fig. 9). This strongly suggests that atmospheric O2 was stable across this time interval, directly refuting hypoxia9 as an explanation for Romer’s Gap.
Although an extinction event at the end of the Devonian saw the demise of many archaic fish groups39, our studies provide new perspectives on the recovery and diversification of surviving groups that went on to found the basis of modern vertebrate diversity40,41.
The new tetrapods show no close relationship to each other, exhibiting different combinations of plesiomorphic and derived characters. Some taxa cluster with Devonian forms, suggesting a possible relict fauna, whereas others appear more crownward, even clustering near the base of the crown group. They imply an early radiation of tetrapods during the Tournaisian and, at the same time, suggest a blurring of the Devonian/Carboniferous boundary, in respect of tetrapod evolution—a feature also noted in tetrapod remains from Nova Scotia12.
If confirmed, our results imply a deep split between stem amphibians and stem amniotes in the earliest Carboniferous. This accords with most molecular dates42,43 for the split, which place it at an average of 355 Myr ago, a date only 4 Myr after the end-Devonian. It suggests that the origin of the tetrapod crown group occurred soon after the extinction event as tetrapods began to recover. Their radiation into a range of new taxa parallels that of lungfish40 and chondrichthyans41 as they adapted to a post-extinction world.
The occurrence of probable plesiomorphic members of the Crassigyrinidae2 and Colosteidae indicates an inception 20–24 Myr earlier than the Late Mississippian, as previously considered44. Other tetrapod material of uncertain attribution are distinct and increase known tetrapod diversity in the Tournaisian (Fig. 6 and Supplementary Figs 2–6).
The preponderance of small animals throughout the sequence is unusual, particularly the very small tetrapod found in a horizon 33 m above the Devonian/Carboniferous boundary, around 1 Myr after the extinction event (Fig. 6). None of the five taxa described above has a skull length of more than 80 mm. This could indicate preservational or collector bias, but they occur throughout different lithologies, horizons and localities (Fig. 6 and Supplementary Figs 2–6). Larger tetrapod taxa are found at Willie’s Hole, about one quarter of the way up the sequence, probably representing about 3 or 4 Myr after the Devonian/Carboniferous boundary. Larger sizes seem to have appeared relatively rapidly in the Tournaisian, as also documented by trackways38; this challenges suggestions of a prolonged period of reduced body size in vertebrates following the DC extinction event45.
The tetrapods of the Ballagan Formation lived in a mosaic of floodplain environments. Some were under water for long periods, others alternated between land surface and standing water. A recent study of the development of Polypterus shows how, early in life, their skeletons can be differentially modified in response to exposure to water-based or land-based conditions46. Such skeletal flexibility might have contributed to the origin of tetrapod terrestrial morphology in the varied environments of the Ballagan Formation.
The wealth and diversity of tetrapod taxa from the Tournaisian refutes the proposal of a depauperate Tournaisian stage, and our charcoal studies show that atmospheric oxygen levels, stable from the Famennian to the Viséan, were not a causal factor of the apparent gap. We emphasize the importance of exploring or re-exploring non-marine Tournaisian sites elsewhere in the world, and examining previously overlooked lithologies.
Specimens UMZC 2016.3 Ossirarus, NMS G. 2013.39/14 Koilops and UMZC 2011.7.2a Perittodus were prepared mechanically with mounted needles; some matrix was removed from Ossirarus with a brush and water, and consolidated where necessary with Paraloid B72. Specimens UMZC 2011.7.2a Perittodus and UMZC 2015.46b Aytonerpeton were scanned at the Cambridge Tomography Centre with a Nikon XTH225 ST scanner. For Perittodus the settings used were: isotropic voxel size 0.0444 mm, 1,080 projections, 0.25 mm Cu filter, X-ray160 kV 70 μA, 1,647 slices, 1,000 milliseconds per slice exposure, 24 dB gain. For Aytonerpeton the following were used: isotropic voxel size 0.0609 mm, 1,080 projections, no filter, X-ray 120 kV, 125 μA, 1,789 slices, 1,000 milliseconds per slice exposure, 24 dB gain.
A new database of 46 taxa coded for 214 osteological characters (170 cranial, 43 postcranial) was subjected to maximum parsimony analyses. It was designed to include representative early tetrapods. Characters were drawn up to capture the features of the new taxa as far as possible in the context of the range of early tetrapods that are available for comparison. Most were drawn from recent analyses14,29,
The data matrix was subjected to maximum parsimony analyses in TNT v. 1.1 (ref. 49). Several experiments of taxon and character manipulation were carried out, as detailed below, with identical search protocols throughout. Given the size of the matrix, tree searches relied on heuristic algorithms, following a simple series of steps under the ‘Traditional search’ option in the ‘Analyze’ menu in TNT. Before each search, we modified memory requirements under the ‘Memory’ option in the ‘Settings’ menu. One hundred megabytes of general RAM were allocated and a total of 50,000 trees were selected as the maximum size of tree space for the exploration of alternative tree topologies. In the initial part of the ‘Traditional search’ (‘Wagner trees’ box ticked), we chose 10,000 replicates (random stepwise addition sequences of taxa), keeping a maximum of five trees at the end of each replicate, using the bisection–reconnection algorithm for tree branch swapping and retaining all trees found at the end of all replicates. A new round of branch swapping was then applied to all trees retained from the initial search (‘trees from RAM’ box ticked). For each set of experiments, where applicable, we summarized the results in the form of a strict consensus (a 50% majority-rule consensus).
Using the search settings explained above, we carried out three types of parsimony analysis. The first parsimony analysis, employing all taxa and characters from the original matrix, treated all characters as having equal unit weight (default TNT option). The second analysis, again using all taxa and characters, was based on implied character reweighting28, briefly described as follows. Given a character, its implied weight (W) is given by K / (K + M − O), where M and O represent, respectively, the greatest number of character-state changes and the observed number of character-state changes for that character. The constant of concavity (K) is an integer, the value of which determines the most parsimonious trees as those trees for which W is maximized across all characters. As the selection of K is arbitrary, we experimented with increasing values (K = 3, 4, 5 and 10) (Fig. 5 and Supplementary Fig. 8). We did not report details of searches with other K values, as our goal was to establish whether the Tournaisian taxa showed stable positions within a minimal range of implied weighting increments. However, we ran analyses with values varying between 6 and 10, with mixed outcomes. In some cases, the Tournaisian taxa were heavily reshuffled, in others the branching sequence of other groups revealed implausible arrangements that, we feel, were dictated by varying amounts of homoplasy in the data. However, a proper characterization of this phenomenon requires further testing. Topologies with K = 10 are reported as an example.
In the third analysis, characters were reweighted by the maximum value (best fit) of their rescaled consistency indexes, obtained from the first analysis.
Statistical branch support was evaluated through character resampling via bootstrap (resampling with replacement50) and jackknife (resampling without replacement, with 33% of characters removed51), using 1,000 replicates in each case and collapsing nodes with less than 50% support.
Of the new Tournaisian taxa, only Diploradus appeared in a maximum agreement subtree (a taxonomically pruned tree showing only taxa for which all most parsimonious trees agree on relationships).
As for the implied weighting analysis, we found stable mutual arrangements for most Tournaisian taxa with K = 3, 4 and 5. With K = 10, the branching sequence of Tournaisian taxa differed from those found with smaller K values. In addition, slightly different branching patterns emerged for various early tetrapod taxa/groups following different implied weighting searches. Below, we highlight key differences among various tree topologies.
In trees generated with K = 3, 4 and 5, Ossirarus, Perittodus and Diploradus emerged as increasingly crownward taxa, in that sequence, along the tetrapod stem group, whereas Aytonerpeton and Koilops were placed among stem amphibians and were thus part of the tetrapod crown group. Ossirarus was crownward of a (Ventastega + Ichthyostega) clade, with Ossinodus placed either immediately anti-crownward of (K = 3), in a polytomy with (K = 4), or immediately crownward of Ossirarus (K = 5). Perittodus was the sister taxon to the Devonian Ichthyostega-like taxon Ymeria, and the (Perittodus + Ymeria) clade formed the sister group to Pederpes. Diploradus was immediately crownward of a (Whatcheeria + Occidens) clade, which in turn occurred crownward of (Pederpes + (Perittodus + Ymeria)). However, the branching sequence of Carboniferous stem tetrapods more crownward than Diploradus varied. Thus, in trees with K = 3, the branching sequence included Crassigyrinus, Doragnathus, (Megalocephalus + Baphetes) and Loxomma. In trees with K = 4, the sequence included only Crassigyrinus and Doragnathus, whereas all baphetids formed a clade on the amphibian stem (Megalocephalus + (Loxomma + Baphetes)). In trees with K = 5, the baphetid clade was, once again, on the amphibian stem, but the sequence of stem tetrapods crownward of Diploradus differed substantially and included (Eucritta + Doragnathus), Sigournea and Crassigyrinus. In trees from K = 3 and 4, the (Aytonerpeton + Sigournea) clade formed the sister group to a (Koilops + (Tulerpeton + (Greererpeton + Colosteus))) clade. In turn, this wider group joined temnospondyls on the amphibian stem, with Caerorhachis as a more immediate sister taxon. In trees from K = 5, Aytonerpeton was collapsed in a trichotomy with temnospondyls and the (Koilops + (Tulerpeton + (Greererpeton + Colosteus))) clade. With K = 10, the results matched those from the second set of parsimony analyses (reweighting).
As for other tetrapod groups, the amniote stem underwent little reshuffling in trees derived from different K values. The most noticeable difference among such trees was the placement of Silvanerpeton and Gephyrostegus, both of which were immediately crownward of the ‘anthracosauroids’ (Eoherpeton + (Pholiderpeton + Proterogyrinus)) but swapped their positions as the first and second most crownward plesion after anthracosauroids.
With characters reweighted by the maximum value of the rescaled consistency index, we found three trees differing only in the relative positions of Whatcheeria, Pederpes and Occidens, all of which formed a clade. In those trees, all new Tournaisian taxa appeared on the tetrapod stem. In particular, Aytonerpeton and Perittodus were sister taxa, and together they joined Ymeria. In crownward order, the sequence of stem tetrapods included: Acanthostega, Ossinodus, Ventastega, Ichthyostega, Ossirarus, the (Ymeria (Aytonerpeton + Perittodus)) clade, the (Whatcheeria, Pederpes, Occidens) clade, Diploradus, Doragnathus, Sigournea, a (Koilops + (Tulerpeton + (Greererpeton + Colosteus))) clade, Crassigyrinus and a baphetid clade. Caerorhachis and Eucritta appeared as the earliest diverging plesions on the amphibian and amniote stem groups, respectively.
Sedimentological and environmental interpretation
The borehole was located at Norham near Berwick-Upon-Tweed, British National Grid Reference (BNGR) 391589, 648135, and the Burnmouth section was at BNGR 396000, 661000.
The stratigraphical position of the succession at Willie’s Hole was inferred from a nearby borehole (Hutton Hall Barns, BGS Registered number NT85SE1). The exact stratigraphical position of the Willie’s Hole section was uncertain within the overall succession. No direct correlation with the succession recorded in the Hutton Hall Barns borehole was possible because the borehole was old and the level of detail insufficient; in addition, distinctive markers are not present in the Ballagan Formation. However, that borehole proved 142.5 m of Ballagan Formation strata; the log is good enough to define precisely where the base is, resting on Kinnesswood Formation. The proximity of Willie’s Hole to the borehole allowed us to infer that the Willie’s Hole section lies approximately 150 m above the base of the Ballagan Formation. The palynological samples from Willie’s Hole contained Umbonatisporites distinctus, a spore that was only found in the lower part of our borehole core. We argue that the Willie’s Hole section belongs to the lower part of the Ballagan Formation. We indicated some uncertainty in the figure and gave an approximate range.
The dominance of actinopterygians and rhizodonts within these lakes indicated brackish–freshwater salinity levels52,53. Diverse palaeosols15 and palynology suggest habitats including forest, low-growing and creeping flora, wetland and desiccating pools traversed by rivers (predominantly meandering channels) and saline–hypersaline lakes depositing cementstones and evaporites (Fig. 6 and Supplementary Fig. 7)27,
DOM was extracted by standard palynological demineralization techniques64. Measurement of maceral reflectance in oil was by means of a Zeiss UMSP 50 Microspectrophotometer, housed in the School of Ocean and Earth Science, National Oceanography Centre Southampton, University of Southampton Waterfront Campus. Measurements were made under standard conditions, as defined by the International Committee for Coal Petrology65.
Model-based estimates of atmospheric oxygen concentrations during the early Tournaisian vary from 10–20%, with more recent models favouring the higher figure66,
By focusing on DOM extracted from sedimentary rocks other than coal, fusinite content can be measured through stratigraphic successions in which coals are rare or absent. The values reported here represent the proportion of fusinite within the organic matter isolated from each 5 g shale sample, on the basis of examination of 500 organic (that is, plant derived) macerals. This indicates the proportion of plant-derived material in the sample which has been burned at high temperatures, and is therefore independent of sediment supply.
The specific Famennian and Viséan sampling localities chosen were selected because, apart from being of the required age:
The stratigraphic context of the sampled formations is well understood, with well-established biozonation (Supplementary Table 1).
Thermal maturity in these successions is low. This is essential, because with increasing thermal maturity the reflectance of non-pyrolitic macerals (most notably vitrinite) increases, eventually rendering them indistinguishable from fusinite.
Both localities represent largely terrestrial environments, containing a succession of fluviodeltaic, lacustrine or nearshore marine deposits (Supplementary Table 1). Sediments deposited in such environments represent an accumulation point for river-transported organic material derived from the wider region; this mitigates the distorting effect of local fire activity.
The organic maceral fusinite is considered synonymous with charcoal and can be distinguished from other maceral types by its reflectance under incident light76; we focused solely on fusinite for this study because, although most other members (semi-fusinite) of the inertinite group are also accepted as pyrolitic in origin77, their reflectance forms a continuum between that of vitrinite and fusinite, and forms the bulk of the organic matter. This makes the per cent sum of semi-fusinite and fusinite very large (>90%) and less reliable.
Supplementary Table 1b gives the samples taken from Famennian sites, Burnmouth, Willie’s Hole and Viséan sites. These were analysed for charcoal content. Mean abundance was 2.0%, which is within the margin of error of the data obtained from Burnmouth Shore; this suggests that the contribution from local fire activity (if any) was similar at both sites (Supplementary Table 1 and Supplementary Fig. 9).
Specimen information is available from the respective housing institutions. Micro-CT scan data will be placed in the NERC National Geoscience Data Centre. This published work and the nomenclature act it contains have been registered in Zoobank: http://www.zoobank.org/urn:lsid:zoobank.org:pub:4BFFB544-7B08-4F2F-80EC-11226C0FDAAB. The names Perittodus apsconditus, Koilops herma, Ossirarus kierani, Diploradus austiumensis, and Aytonerpeton microps have been deposited in the Zoobank database under LSIDs urn:lsid:zoobank.org:act:69DB72E5-F9BD-49C6-B471-CD8E03767732, urn:lsid:zoobank.org:act:8C43E66A-3822-49B4-B3B5-E43C79FA9C70, urn:lsid:zoobank.org:act:FC9FAB5C-CC3E-4D0D-B7D7-8030FBAA4F0C, urn:lsid:zoobank.org:act:268DDD4F-289D-4F83-8172-1A18A1007B7C and urn:lsid:zoobank.org:act:E1E094A8-FAC0-4A2A-A13D-487D7775FBE1.
How to cite this article: Clack, J. A. et al. Phylogenetic and environmental context of a Tournaisian tetrapod fauna. Nat. Ecol. Evol. 1, 0002 (2016).
We acknowledge funding from NERC consortium grants NE/J022713/1 (Cambridge), NE/J020729/1 (Leicester), NE/J021067/1 (BGS), NE/J020621/1 (NMS) and NE/J021091/1 (Southampton). We thank the following for their support and contributions: the late S. Wood and M. Wood for discovery of and access to collections; O. Kieran and B. Kieran and the Burnmouth community for support for the project; M. Browne for field assistance and information on stratigraphy; M. Lowe for access to UMZC collections; S. Finney for field assistance, conservation advice and preparation of Koilops; V. Carrió for conservation and preparation of NMS specimens; J. Sherwin for stratigraphy and field assistance; and S. Akbari (Southampton) for contribution to palynological processing. T.I.K. and D.M. publish with the permission of the Executive Director, British Geological Survey (NERC). A. Brown and C. MacFadyen of Scottish Natural Heritage gave permission to collect at sites in their care and P. Bancks, from The Crown Estates Office in Edinburgh, gave permission to collect on Crown land. PRISM, the Isaac Newton Trust Fund (Trinity College, Cambridge), the Crotch Fund (UMZC) and an anonymous donor provided funding for the purchase of specimens. This is a contribution to IGCP project 596.