Aspidosaurus chiton Broili, 1904

Gee, Bryan M., Mann, Arjan & Sues, Hans-Dieter, 2025, A reassessment of Aspidosaurus chiton Broili, 1904 (Temnospondyli: Dissorophoidea) based on a new skeleton from the early Permian of Texas, Zoological Journal of the Linnean Society 204 (3) : -

Aspidosaurus chiton Broili, 1904

Revised diagnosis: The following features are considered autapomorphies for the genus: (i) double series of osteoderms in which the internal series is not exposed dorsally; (ii) external osteoderm articulation formed by overlap of each osteoderm onto the posterior margin of the osteoderm in front of it. The genus is further diagnosed by the following differential features: (i) preorbital region longer than postorbital region when measured along the midline, as with most cacopines; (ii) ornamentation relatively coarse, with nodes approaching the tubercular condition like those in highly nested cacopines (e.g. Anakamacops petrolicus Li & Cheng 1999 , Cacops aspidephorus , Cacops morrisi , Kamacops acervalis ) present in some regions of the cranial roof; (iii) postparietals are relatively equal in proportions in contrast to the markedly foreshortened postparietals of most cacopines (e.g. Cacops spp. , Conjunctio multidens , Parioxys ferricolus ); (iv) ornamentation forms an elevated transverse ridge (nuchal ridge) on the occipital margin of the postparietals, as in most cacopines (but not Scapanops neglectus ); (v) otic notch is posteriorly ‘open’ compared to dissorophid specimens of a similar size (e.g. Ca. aspidephorus , Ca. morrisi ) due to a lack of a ventral extension of the tabular horn and a rudimentary dorsal process of the quadrate; and (vi) external osteoderms form an inverted-V shape in anterior/posterior view, with a distinct apex and framing a tighter angle (approximately a right angle), rather than the gently dorsally convex osteoderms seen in cacopines and dissorophines and without an offset dorsal ridge as seen in S. neglectus . The absence of an internarial fontanelle is not definitive due to slight damage in this region, but if a fontanelle was present, it would have been much smaller than in most dissorophids with this opening, including Aspidosaurus binasser .

Neotype: USNM PAL 406323 , a nearly complete cranium with associated pectoral girdle, forelimb, axial column, pelvic girdle, and hindlimb material.

Type locality: 1.6 km south-southeast of the Kemp triangulation point shown on the 1966 United States Geological Survey Northeast of Lake Kemp 7.5’ Quadrangle, east side of East Coffee Creek, Wilbarger County, Texas, USA ( Fig. 2).

Type horizon: Lower Clear Fork Formation between the informally designated Craddock dolomite and Red Tank sandstone members of Nelson et al. (2013). Lower Permian (Cisuralian), Leonardian Series, Artinskian Stage.

Referred specimens: AMNH FARB 4756 ( holotype of ‘ Alegeinosaurus aphthitos ’) from East Coffee Creek , Texas, USA, a partial postcranial skeleton representing the anterior trunk region (pectoral girdle, forelimb, axial column) ; ROM VP 80069 from Richards Spur , Oklahoma, USA, pair of articulated osteoderms and neural spines .

Description

Cranium: The cranium is largely complete with the exception of the suborbital region in which portions of the maxilla, jugal, and possibly a lateral exposure of the palatine (LEP) and a lateral exposure of the ectopterygoid (LEE) have been lost on the dorsal surface, and portions of the palatine, ectopterygoid, and pterygoid have been lost on the palatal surface ( Fig. 3). It is compressed dorsoventrally and on the right side, transversely ( Figs 3, 4). It is thus unclear whether the cranial roof sloped continuously downwards towards the snout or whether it had a marked inflection like that in Cacops aspidephorus and Cacops morrisi (e.g. Gee and Reisz 2018a, Anderson et al. 2020). The left maxilla has also been partially folded underneath other cranial elements. As a result of this post-mortem compression, the occiput is also distorted ( Fig. 5), with the right exoccipital being posteriorly shifted, and the occipital condyles are dislodged into an unnatural articular facet.

The ornamentation of the cranial roof bears subcircular pits and grooves that are framed by shallow ridges ( Figs 3, 4). It is most strongly developed in the circumorbital region and along the dorsal margin of the otic notch where the ornamentation becomes more prominent and the pits become deeper and more pronounced. Broili (1904) similarly noted more pronounced ornamentation in these regions of the holotype of Aspidosaurus chiton . In the postorbital region, some of the ornamentation approaches the condition observed in cacopines like Cacops aspidephorus and Cacops morrisi in which the ridges intersect at nodes that are distinctly elevated (termed ‘tubercular’; e.g. Gee and Reisz 2018a, Anderson et al. 2020). However, the nodes are usually not elevated well above the ridges in USNM PAL 406323, even in comparison to smaller specimens of C. morrisi , so this may be of taxonomic significance. The ornamentation in the interorbital region and on the snout is more subdued, especially on the premaxilla and the maxilla, but the same subcircular pits and grooves persist ( Fig. 3). On the premaxilla, the surface is mostly smooth and marked by minute, shallow pores. Along the occipital margin, the ornamentation forms a semi-continuous transverse ridge (the nuchal ridge), which is slightly elevated above the postorbital region and markedly offset from the ventrally descending occipital flanges. The quadratojugal has a series of deeper pits in a laterally thickened longitudinal ridge that extends along the posterior half towards the quadrate ( Fig. 4). There are no strongly offset ridges on the remainder of the cranial roof like those seen in some cacopines (e.g. C. morrisi , Scapanops neglectus ; Reisz et al. 2009, Schoch and Sues 2013), but there is some topography in the preorbital region in which the nasal appears to be slightly dorsally convex (this is symmetrical on both sides and not thought to be the result of transverse compression on the right side) and some shallow depressions along the nasal-lacrimal and the nasal-prefrontal sutures ( Fig. 4). Some of these regions may have had more pronounced topography (e.g. a ridge originating on the prefrontal and bridging the naris and the orbit along the nasal’s lateral margin appears more subdued at present) when undistorted. Unlike in dissorophines (e.g. ‘ Broiliellus ’ brevis Carroll 1964a ; Dissorophus multicinctus Cope 1895 ; Carroll 1964a, DeMar 1968), no ‘eminence’ on the lateral surface of the jugal is present, nor are there elevated areas on the postorbital region (e.g. Diploseira angusta ; Dilkes 2020).

The premaxilla is a subtriangular element framing the external naris anteriorly, medially, and laterally ( Fig. 3). Long alary processes are dislodged slightly above the plane of the rest of the snout and form most of the medial narial margin. The length of the processes and the relative contribution to the medial narial margin are relatively distinct and most similar to Conjunctio multidens ( Schoch and Sues 2013) . There are at least 10 or 11 tooth positions that can be confidently attributed to the right premaxilla, with room for up to 13 positions. Whether an internarial fontanelle was present cannot be confidently determined. As preserved, there is an asymmetrical opening at the tip of the snout between the premaxillae and nasals ( Fig. 3), but this is exaggerated by deformation in this region and mostly to the left of the midline; on the right side of the cranium, there is a continuous surface along the midline, with only a minute off-centre perforation towards the tip. Because of imperfect preservation in the region, we do not make a definitive characterization of either absence or presence since it remains possible that a minute fontanelle was present at the very tip of the snout and left this character as unknown in the phylogenetic matrix. Schoch and Sues (2022b) reported a possible depression without a fontanelle in some specimens of Parioxys ferricolus , which could be another possibility in this specimen (perhaps predisposing the cranium to deformation in this region); Moustafa 1955a characterized the taxon as having a small ‘internasal pit’, which could be interpreted as either a depression or an actual opening (see also Carroll 1964a, who used the term to refer aternatively to an intervomerine fossa).

The maxilla is an elongate element that is largely incomplete on both sides ( Figs 3, 4). The most complete region of the element is the portion bridging the naris and the orbit on the right side, which expands dorsomedially from a slender process framing the naris laterally. Most of the suborbital portion of the maxilla is lost on both sides, and the posterior terminus is not definitively preserved on either side ( Fig. 4); on the left side ( Fig. 4A, B), it extends at least past the level of the posterior orbital margin to meet the quadratojugal, and the posterior extent can be more confidently inferred in palatal view at about the mid-length of the subtemporal fenestra ( Fig. 6). There is room for 31 or 32 tooth positions in the region preserved on each side, which suggests a complete tooth count around 55–65 positions based on the conserved small size of the teeth; the range is a result of the uncertainty over the posterior extent of the maxilla and of its tooth row. However, this count is undoubtedly much higher than in cacopines such as Cacops spp. and Parioxys ferricolus (e.g. Gee and Reisz 2018a, Schoch and Sues 2022b) and is more similar to counts in early-diverging taxa like Conjunctio multidens and Scapanops neglectus ( Schoch and Sues 2013) .

A septomaxilla is present in the left naris at the posterior end ( Figs 3, 4A, B). It is dorsally convex posteriorly and extends anteriorly to form part of the floor of the naris. While it reaches the height of the posterior narial margin, it is not integrated into the cranial roof.

The lacrimal is an elongate element that borders the external naris but that is separated from the orbit ( Figs 3, 4), which distinguishes it from a few dissorophids that either lack a lacrimal-naris contact (e.g. Broiliellus texensis Williston, 1914 ) or the majority that possess a lacrimal-orbit contact (e.g. Parioxys ferricolus ; Schoch and Sues 2022b). The absence of a lacrimal-orbit contact is shared with all three species of Cacops Williston, 1910 ( Gee and Reisz 2018a, Gee et al. 2019b, Anderson et al. 2020) and with Anakamacops petrolicus ( Liu 2018) . It has a relatively small contribution to the posterior narial margin. Although it appears to enter the right orbit ( Fig. 3), this region is markedly distorted ( Fig. 4C, D), and it is clearly excluded from the orbit on the less distorted left side, so we interpret the original condition to be that of the left side. Foramina for the nasolacrimal canal were not identified.

The presence of a lateral exposure of the palatine (LEP) is equivocal at best. This feature has not been universally reported in dissorophids, but because it was not known in any temnospondyl until the mid-twentieth century (e.g. Shishkin 1967), taxa described prior to this [e.g. Broiliellus texensis by Williston (1914)] may very well have a LEP that was either not identified or obscured by poor preservation. For every such taxon that has been redescribed, however, an LEP has been reported (e.g. Cacops aspidephorus , Conjunctio multidens , Dissorophus multicinctus , Parioxys ferricolus , Scapanops neglectus ; Schoch 2012, Schoch and Sues 2013, 2022b, Anderson et al. 2020). On the left side ( Fig. 4A, B, region labeled ‘pal/pf ’), it is notable that there is only a loss of the ornamented external surface at the anterolateral corner (compared to full loss of the maxilla in some of the suborbital region) because this is the precise location of the LEP in dissorophids with a known LEP. On the right side ( Fig. 4C, D, region labeled ‘l/pf’), this region is also mostly lost, and some elements are dislodged. The preserved portion of the palatine that is visible in dorsal view ( Fig. 3) is insufficient to determine whether there was a lateral exposure because the palatine may have been narrowly obscured laterally at early stages of ontogeny (e.g. Gee et al. 2019a). Interestingly, there is an isolated fragment of the same ornamentation as the cranium (not shown because it is largely encrusted in matrix) that is slightly laterally/dorsally concave. It is of the approximate size that would be appropriate for an LEP in this cranium, although it is encrusted with matrix and does not precisely fit (perhaps due to postmortem distortion), and the LEP is often concave (e.g. Cacops morrisi , P. ferricolus ; Gee and Reisz 2018a, Schoch and Sues 2022b). The presence or absence of a lateral exposure of the ectopterygoid (LEE) cannot be determined; this exposure is much rarer, usually smaller than the LEP, and may be ontogenetically transient (e.g. Reisz et al. 2009).

The nasal is a broadly rectangular element forming much of the dorsal surface of the snout ( Figs 3, 4). It is partly excluded from the medial narial margin by the premaxilla but has a major contribution to the posterior margin. The lateral margin with the lacrimal is gently sinusoidal, as is the posterior margin with the frontals such that there is a short posterior process wedging into the frontal. The suture with the prefrontal could thus be characterized as ‘stepped’ as the nasal widens anterior to the prefrontal.

The prefrontal is a triangular element framing the anteromedial corner of the orbit ( Figs 3, 4). The ventral process of the prefrontal (VPP) firmly abuts the palatine and is visible internally within the orbit ( Fig. 4C, D). The prefrontal is widely separated from the postfrontal, leading to a frontal-orbit contact.

The frontal is a rectangular element with an anterior terminus that is slightly anterior to the level of the anterior orbital margin and a posterior terminus that is slightly anterior to the level of the posterior orbital margin ( Fig. 3). It forms much of the medial orbital margin and contacts the parietal posteriorly in a mainly transverse contact. The anteriormost extent is approximately in line with the anteriormost extent of the prefrontals (as seen also in Aspidosaurus binasser ; Berman and Lucas 2003), which is unusual among dissorophids in which the frontals typically terminate well posterior to the prefrontals. In Parioxys ferricolus , there is a shift throughout ontogeny where the anteriormost extent of the frontals changes from being well posterior to that of the prefrontals to slightly anterior to it ( Schoch and Sues 2022b).

The parietal is a quadrangular element that is anterolaterally indented by the postfrontal, forming a marked ‘step’ in which the anterior half is distinctly narrower than the posterior half and the ‘step’ forms a right angle ( Fig. 3). A large pineal foramen is framed just anterior to about the mid-length of the parietals. The parietal contacts the supratemporal laterally and the postparietal posteriorly in straight sutures.

The postfrontal is incomplete on both sides but appears to have the same crescentic shape that is found in most other dissorophids, with a medially convex medial suture that indents the parietal, a broad posterior suture with the supratemporal that contributes to a separation of the postorbital and the parietal, and a lateral suture with the postorbital around the posterior orbital margin ( Figs 3, 4). It is widely separated from the squamosal, which is more common in non-cacopines (‘ Broiliellus ’ reiszi Holmes et al. 2013 , Dissorophus multicinctus ; Schoch 2012, Holmes et al. 2013) and early-diverging cacopines ( Scapanops neglectus ; Schoch and Sues 2013); these elements contact each other in Cacops aspidephorus and Cacops morrisi ( Reisz et al. 2009, Anderson et al. 2020).

The postorbital is incomplete ventrolaterally on the left side and mostly lost or damaged on the right side ( Figs 3, 4). Little more can be said beyond that it has a posteromedial suture with the supratemporal and a posterior suture with the squamosal. It appears to have had a posterior margin that forms more of a discrete tapering process (as in the majority of dissorophids) rather than being squared-off (as in Cacops morrisi ; Reisz et al. 2009).

The squamosal is inferred to be largely preserved on each side because it normally borders most of the otic notch, but its sutural margins are poorly defined due to extensive fracturing and compression in this region ( Fig. 4). Very little of the ornamented surface is preserved, and most of what is confidently identifiable is the unornamented supratympanic flange. On the left side ( Fig. 4A, B), the sutures are mostly only discernible dorsally. A small portion of the ornamented surface preserves the suture with the postorbital, and sutures with the supratemporal dorsally and the tabular posterodorsally are visible on the unornamented dorsal portion of the supratympanic flange. Conversely, on the right side ( Fig. 4C, D), the dorsal region is largely damaged or obscured, and the sutures are largely unclear. The otic notch itself is undoubtedly compressed dorsoventrally by crushing, but the ventral portion of the supratympanic flange slopes posteroventrally, as in other dissorophids and in contrast to some trematopids in which it is nearly horizontal (e.g. Ecolsonia cutlerensis, Mattauschia laticeps ( Fritsch, 1881); Berman et al. 1985, Milner 2018). The dorsal portion is oriented horizontally and has a nearly straight ventral margin beneath the tabular and supratemporal; it may be slightly angled due to the compression of the cranium. Thus, there does not appear to be a pronounced semilunar curvature, in contrast to many other dissorophids (‘ Broiliellus ’ olsoni , Cacops woehri , Dissorophus multicinctus ; DeMar 1967, 1968, Fröbisch and Reisz 2012).

The supratemporal is a rectangular element, slightly longer than it is wide ( Fig. 3). It has a minute ventrally directed semilunar flange that contributes to the supratympanic flange where it sutures to the squamosal ventrally and the tabular posteriorly ( Fig. 4A, B). It also forms a discrete supratympanic shelf that overhangs the supratympanic flange.

The postparietal is a rectangular element that is only slightly wider than it is long ( Fig. 3), similar to dissorophines and differing from most cacopines in which this element is markedly foreshortened [e.g. Cacops spp. , Conjunctio multidens , Parioxys ferricolus , Scapanops neglectus (polymorphic to some extent); Reisz et al. 2009, Schoch and Sues 2013, 2022b, Fröbisch et al. 2015]. The suture with the tabular is not clearly defined on either side on the dorsal surface, and it is completely indiscernible on the broad, sloping occipital flanges ( Fig. 5). The occipital margin is essentially straight, rather than posteriorly concave or biconcave, and then angles sharply posterolaterally around the position of the suture with the tabular. The occipital flanges contact the exoccipitals quite broadly, but this suture is not defined, and the contact may have been exaggerated by, or entirely the result of, the dorsoventral compression of the cranium.

The tabular is a subtriangular element dorsally that contributes to the dorsal margin of the otic notch and forms the entirety of the tabular horn ( Figs 3–5). The tabular horn is particularly rugose and forms a thickened structure, but it has no apparent ventral extension based on the complete right tabular, thereby contributing to a posteriorly open otic notch ( Fig. 4). Its contribution to the supratympanic flange is minute. It extends straight posteriorly from the cranial roof ( Fig. 6), rather than at an oblique posterolateral angle, and is symmetrical on both sides, indicating that this orientation is not the result of the crushing of the right side. As previously noted, the tabular should contribute to the occipital flange, but its relative contribution and its relationship to the exoccipital are unclear due to the lack of sutures ( Fig. 5).

The jugal is assuredly present on both sides since it consistently forms the posteroventral corner of the orbit, but it is markedly incomplete in all directions, and only small portions of its posterior and ventral sutures can be confidently discerned beyond part of the posterior margin that contacts the squamosal on the left side ( Fig. 4). Although it is exposed in ventral view on the left side ( Fig. 6), this is thought to be the result of damage to the maxilla ( Fig. 4A, B) and not the true biological condition (i.e. the maxilla should have continued beneath the entire jugal).

The quadratojugal is fairly complete on both sides (more so on the left side) but generally lacks defined sutural boundaries anteriorly with the jugal and dorsally with the squamosal, as previously noted ( Fig. 4). The clearest sutures are defined posteriorly where the quadratojugal overlaps the quadrate; these can be seen laterally and ventrally on the left side in particular ( Figs 4, 6). On the palate, the quadratojugal not only overlaps the lateral surface of the quadrate but also has a process extending medially towards the quadrate ramus of the pterygoid to partially overlap the quadrate anteriorly. Despite appearing to contact the broken terminus of the quadrate ramus of the pterygoid on the left side as figured, it does not contact the pterygoid and appears separated from it by the quadrate on the right side ( Fig. 6). The quadratojugal contributes to the ventral portion of the supratympanic flange ( Fig. 4), but the exact degree cannot be defined.

The quadrate is a triangular element mostly obscured dorsally and laterally by other elements ( Figs 4–6). It has a rudimentary dorsal process that barely rises above the level of the ventralmost margin of the otic notch and is similar to those in most dissorophids with the exception of some of the highly nested cacopines (e.g. Cacops morrisi , Cacops aspidephorus ; e.g. Reisz et al. 2009, Anderson et al. 2020) in which there is a marked dorsal process that forms a fan-like structure with a longitudinal orientation. Ventrally, the jaw articulation is formed by a well-ossified facet that is medially framed by a ventrally complex longitudinal ridge ( Fig. 6).

The palate is largely complete save for the loss of the lateralmost elements in the suborbital region (parts of the palatine, ectopterygoid, and pterygoid; Fig. 6). The compression of the right side has also resulted in the anterior portion of the right palate being dislodged medially to overlap onto some of the left palate. Sutures are largely undefined due to a combination of fracturing, overlapping elements, damage, and difficulty in further preparation of the specimen without obliterating denticles. In general, the palate resembles that of Cacops spp. ( Reisz et al. 2009, Gee and Reisz 2018a), with a large intervomerine depression at the anterior portion of the palate; it is about a third of the inferred length of the vomers and presumably mostly contained within the premaxillae ( Fig. 6). The choana is more defined on the left side; its proportions more closely resemble those in Cacops spp. ( Reisz et al. 2009, Gee and Reisz 2018a) than the markedly slit-like choanae of Parioxys ferricolus or Kamacops acervalis (Schoch 1999, Schoch and Sues 2022b). The interpterygoid vacuities are distorted, but the left one is nearly complete, oval in shape, and widest at the mid-length of the palatine ramus of the pterygoid, before tapering anteriorly ( Fig. 6). The proportions are most similar to Cacops spp. ( Reisz et al. 2009, Gee and Reisz 2018a) and unlike in P. ferricolus and K. acervalis (e.g. Schoch 1999, Schoch and Sues 2022b).

The vomers form large plate-like elements in the anterior palate ( Fig. 6). The palatines and ectopterygoids are largely lost or otherwise obscured by matrix; what is exposed is relatively featureless beyond the dentition ( Fig. 6). The pterygoids are largely complete, although only the suture forming the basicranial articulation with the parasphenoid is clearly defined among its inferred contacts ( Fig. 6). The quadrate ramus extends straight posteriorly and overlaps the medial surface of the quadrate on the right side ( Fig. 6). This ramus can be predicted to have an ascending flange that rises dorsally from the quadrate ramus (e.g. Gee and Reisz 2018a, for Cacops morrisi ), but this region is not exposed here. On the left pterygoid, a well-developed transverse flange extends posterolaterally into the subtemporal fenestra. The flange is angled ventrolaterally such that its edge lies well below the level of the rest of the palate. The basicranial articulation is fully intact, but there is no evidence for a tight interdigitating suture as seen in some other taxa, so the contact may have been overlapping with a degree of mobility permitted. The parasphenoid is essentially complete but badly distorted, especially the basal plate. The basal plate is rectangular, being wider than long. As preserved, it appears to have had two relatively large, central depressions that are separated by a ridge extending between the basipterygoid rami, but the more anterior depression may be distorted at the base of the cultriform process, which should be dorsally convex and thus appear recessed when viewed from below. The more posterior depression is inferred to be a biological feature like that found in other dissorophids, where it is interpreted to be for muscle attachment (e.g. Fröbisch and Reisz 2012). Posterolateral wings are present posterior to the basicranial articulation, but their full transverse extent is not entirely clear. Foramina for the carotid arteries were not evident. The cultriform process is narrow and ventrally convex. Its full anterior extent is unclear in the absence of clear sutures and lateral offset from the vomers.

The palatal dentition is like that of other dissorophids. There is no evidence for denticles on any part of the parasphenoid, but a shagreen is evident along the medial margin of the choana on the vomer and the palatine and on both the corpus and the transverse flange of the pterygoid ( Fig. 6). A lack of denticles on some areas of the vomer, palatine, and pterygoid may be attributed to clear taphonomic damage and preparation marks; the denticle field may have been more extensive as in other dissorophids (e.g. Anakamacops petrolicus , Cacops morrisi ; Reisz et al. 2009, Liu 2018). Along the posterolateral margin of the vomers is a pair of enlarged teeth (‘fangs’ or ‘tusks’). At the posterior end of the choana is another pair of enlarged teeth that should be situated on the palatine ( Fig. 6). A single enlarged tooth is present at the mid-length on both sides of the palate and is presumed to be on the ectopterygoid based on a relatively posterior position medial to the marginal tooth row ( Fig. 6).

Scattered throughout and around the vacuities are tiny plates that probably represent the palatal plates embedded in the soft tissue coverings of the vacuities ( Fig. 6; Gee et al. 2017). No denticles are preserved on these, although small spots on a few plates are interpreted as exposed pulp cavities. Large patches are found around the right choana, on the cultriform process, and at the anterior end of the left interpterygoid vacuity. Our interpretation of these as palatal plates rather than scleral plates is based on the sheer number, which far exceeds what would be expected from the sclerotic rings alone, and the marked variation in shape and size of the plates, which follows more definitive occurrences in other dissorophoids (e.g. Fröbisch and Reisz 2008, Gee et al. 2017). There is a patch of slightly smaller plates on a block of matrix attached to the left orbit ( Fig. 3), but it is unclear whether these are palatal plates or possibly scleral plates.

Elements of the braincase are partially exposed. The sphenethmoid is partially exposed dorsal to the cultriform process at about the level of the mid-length of the orbit ( Fig. 4C, D). The anterior and posterior extents are not fully defined, but it does not appear to have extended for the full length of the orbits, let alone beyond them anteriorly or posteriorly. Contact with the cranial roof is inferred but not positively identified and could have been altered by dorsoventral compression of the cranium. On the occiput, the right stapes is articulated, with the proximal end expanded for articulation with the fenestra vestibuli and the narrow and slender distal end ( Figs 5, 6). There is no curvature of the stapes in any direction. A stapedial foramen is present near the base of the stapes. A slightly dislodged opisthotic is interpreted to be present on the left side based on a bony surface extending dorsolaterally from the exoccipitals ( Fig. 5), but little can be said about its anatomy other than that it posteriorly and dorsally frames the fenestra vestibuli. Whether a prootic was present anteriorly, and if so, whether it was co-ossified with the opisthotic as in Cacops morrisi and Kamacops acervalis (Schoch 1999, Gee 2020), is unclear. A fragment in the correct position to be the right opisthotic is also present. The exoccipitals form the occipital condyles, which are deeply socketed and partially subdivided ( Fig. 5). From the condyle, each exoccipital expands broadly dorsally and transversely to form a broad buttress abutting the occipital flange; again, whether this contact is artificially introduced or exaggerated is unclear.

Hemimandible: The hemimandible is only represented by fragments from the posterior half of each side that are identified by a pronounced ornamentation that is developed on the lateral and ventral surfaces of the angular ( Fig. 7). Given the incompleteness of the hemimandibles and the fracturing and dislodgement within the preserved fragments, it is difficult to confidently determine whether it formed a distinct keel that interrupts the continuity of the ventral margin, like that in ‘ Broiliellus ’ reiszi ( Holmes et al. 2013) and Dissorophus multicinctus ( DeMar 1968) , or is merely a region of coarser ornamentation, like that in Cacops aspidephorus ( Anderson et al. 2020) , Cacops morrisi ( Gee and Reisz 2018a, Gee et al. 2019b), and Parioxys ferricolus ( Moustafa 1955a) . A portion of the dentary is preserved but confers no other information beyond confirming the marginal teeth were relatively small, monocuspid, and non-pedicellate ( Fig. 7), like those of the cranium.

Postcranium: The postcranial skeleton is represented predominantly by the axial column (vertebrae, osteoderms, ribs), with portions of the pectoral girdle, forelimb, pelvic girdle, and hindlimb. The left pectoral girdle remains largely articulated, and there are several articulated series of vertebrae and osteoderms, but the remainder of the postcranial skeleton is disarticulated.

The osteoderms are the most informative part of the postcranium and represent 12 or 13 vertebral positions across four blocks ( Fig. 8). They are evenly covered with sub-circular pitting across the entire dorsal surface and markedly convex, though without a sharp ridge along the midline apex. In anterior and posterior views, the osteoderms form an inverted-V that differs from the less pronounced and more gradual dorsal convexity seen in most other dissorophids without hyperelongate osteoderms (e.g. Cacops ; Williston 1910, Dilkes and Brown 2007, Dilkes 2009, Gee et al. 2019b). The only other dissorophid with this marked inverted-V morphology is Scapanops neglectus , which has a discrete dorsal ridge along the midline keel; the holotype and only known specimen of this taxon is a skull about half the length of USNM PAL 406323 ( Schoch and Sues 2013). In USNM PAL 406323 and the holotype of Aspidosaurus chiton ( Fig. 1), there is no such ridge, only a shallow keel formed at the apex. Except for the anteriormost position ( Fig. 8A, B), the osteoderms are rectangular in profile, being about twice as wide as they are long in dorsal view and thus more similar to cacopines than to dissorophines (except ‘ Broiliellus ’ reiszi ; Holmes et al. 2013). The anteriormost osteoderm is triangular, as described and figured for the holotype of As. chiton by Broili (1904). In most other dissorophids in which the anteriormost osteoderm is known, the first osteoderm also differs from the remainder, but Dissorophus multicinctus has a broad shield that possibly represents fusion of multiple osteoderms, thus covering multiple vertebrae ( DeMar 1966b), and in Cacops aspidephorus , it is narrower than successive positions and much longer proportionately than in USNM PAL 406323 ( Dilkes and Brown 2007). In Anakamacops petrolicus , for which osteoderms are mostly unknown, a subtriangular osteoderm, like that of C. aspidephorus , was interpreted as the anteriormost osteoderm ( Liu 2018). In USNM PAL 406323, each osteoderm overlaps the posterior margin of the preceding osteoderm, although fractures near or along this contact sometimes give the impression that they are abutting ( Fig. 8A, B), rather than overlapping. The same articulation was described in the holotype by Broili (1904) and has never been reported in another dissorophid except for ‘ Alegeinosaurus aphthitos ’, now synonymized with Aspidosaurus (Gee 2018) . In ventrolateral view, it is possible to see the tapering posterolateral margin underlying the next osteoderm, confirming the overlapping nature of these osteoderms. This overlapping nature and the shape, proportions, and ornamentation all conform to the holotype of As. chiton as described and figured by Broili (1904) and to the morphology seen in ROM VP 80069 from Richards Spur, which was referred to cf. Aspidosaurus sp. ( Gee et al. 2019b). A hidden internal series of osteoderms in the Richards Spur specimen, elucidated only through CT analysis, is confirmed in a block of USNM PAL 406323 that has an osteoderm broken at the mid-length of the neural spine, revealing the same internal osteoderm ( Fig. 8I–J). As with ROM VP 80069, this osteoderm tightly adheres to the neural spine, without any clear sutural demarcation, is unornamented, is neither as anteroposteriorly long nor as transversely wide as the ornamented osteoderm capping it, and only covers one vertebra, rather than being situated between adjacent neural spines. It is not possible to say whether there is a ventral flange of the internal series, and there does not appear to be a distinct ventral flange on the external series. The presence of such a flange on the external series is a feature of cacopines with a double series that separates them from dissorophines with a double series (except for ' B.' reiszi ). In lateral view, the main osteoderm block has a distinct dorsal convexity, but this is most likely to be a postmortem artifact related to the unique articulation mode since the vertebrae are directed at various corresponding angles across a range that seems unnatural. The marked curvature forming the sail of Platyhystrix rugosa is formed largely by variation in the blade-like osteoderms, which widen sagittally and acquire some measure of curvature towards the apex ( Lewis and Vaughn 1965), and apparently by a reduction in presacral count.

The vertebrae are largely obscured by the osteoderms, ribs, and supporting matrix ( Fig. 8), but the stereotypical structure of dissorophids is observed in that the neural spines are moderately tall, about as long as the width between the transverse processes; there are stout transverse processes with circular to oval cross-sectional ends that are about as laterally extensive as the osteoderms ( Fig. 8E, F); and centra are rhachitomous ( Fig. 8C–H). The neural spine expands transversely towards the apex, giving a ‘swollen’ profile when viewed along the sagittal axis ( Fig. 8I, J), which was also noted in the holotype of Aspidosaurus chiton by Broili (1904). Although Broili highlighted only the transverse expansion, there is a ridge on the anterior and posterior faces as well that is figured for some osteoderms ( Broili 1904; Fig. 1). However, it is more clearly defined by a vertical groove separating it from the transverse expansion, rather than being a markedly projecting feature, and the cross-section at the tip of the neural spine is distinctly wider than it is long. The same morphology is seen in USNM PAL 406323, AMNH FARB 4756 (the holotype of ‘ Alegeinosaurus aphthitos ’; Gee 2018), and ROM VP 80069 ( Gee et al. 2019b). A similar morphology is observed in cacopines such as Anakamacops petrolicus , Cacops aspidephorus , and Cacops morrisi (e.g. Dilkes 2009, Gee and Reisz 2018a, Liu 2018, Gee et al. 2019b). In these taxa, the ridges are symmetrical such that the cross-section at the tip is a symmetrical plus sign (+), and outward expansion towards the tip is less pronounced and does not result in the ‘swollen’ morphology of As. chiton . In dissorophines, there is some measure of transverse expansion (e.g. Dilkes 2009), but the articulation of the neural spine with ventral flanges of the internal osteoderms is formed by grooves in the tip of the spine and thus precludes the development of comparable ridges along the anterior and posterior surfaces. The intercentra are either lost or obscured by matrix for most positions, but the few that are exposed appear to be of the typical rhachitomous wedge-shaped morphology.

Ribs are partially articulated with some of the articulated vertebral positions, but only the proximal ends are preserved ( Fig. 8E–H). Various exposures of other parts of the ribs are preserved in partial articulation on some of the other pieces associated with this specimen, but a complete rib was not identified among any of them. The proximal ends are broadly expanded and appear to have been relatively flat before transitioning to a more slender and more cylindrical shaft. Well-developed spikelike uncinate processes extend posterodorsally from the shaft a short distance distal to the proximal head and overlap the succeeding rib ( Fig. 8G, H). Other than the presence of these processes, which occur in many (‘ Aspidosaurus ’ novomexicanus , Cacops morrisi , Diploseira angusta , Dissorophus multicinctus ; Carroll 1964a, DeMar 1968, Gee and Reisz 2018a, 2018b, Dilkes 2020) but not all dissorophids (e.g. Cacops aspidephorus , Parioxys ferricolus ; Williston 1910, Moustafa 1955a), the ribs confer no other information.

The pectoral girdle consists mostly of the right half contained within a single highly fragmented block ( Fig. 9). There are several other exposed, disarticulated elements, most if not all of which are ribs. The cleithrum is of the typical dissorophid morphology (e.g. DeMar 1968, Dilkes and Brown 2007, Dilkes 2009, Holmes et al. 2013, Gee 2018, Gee and Reisz 2018a, b), with a ventrally descending process that forms a discrete ridge along the anterior margin of the scapulocoracoid and a dorsal process that is flat and more oval in profile, overlapping the dorsal margin of the scapulocoracoid. The ventral process of the cleithrum articulates with the similarly narrow, but flatter, dorsal process of the clavicle, which is incomplete at its mid-length ( Fig. 9). The ventral portion of the clavicle, flat and oval in profile, partially overlaps the interclavicle. The interclavicle is mostly complete but lacks a clear posterior process, which could have been broken off, and thus is diamond-shaped ( Fig. 9; as preserved in Cacops morrisi and Dissorophus multicinctus and unlike in Cacops aspidephorus ; Williston 1910, DeMar 1968, Gee and Reisz 2018a). The paucity of complete, undamaged interclavicles complicates efforts to separate taxonomic and ontogenetic signals. The scapulocoracoid is heavily fractured and partially covered by other elements, thus being mostly exposed only laterally ( Fig. 9). What can be observed resembles the scapulocoracoids of other dissorophids ( Williston 1910, Moustafa 1955a, DeMar 1968, Gee and Reisz 2018a, 2018b) in being a well-ossified structure that is substantially dorsoventrally taller than it is anteroposteriorly long. There is no pronounced ornamentation on any of the pectoral elements. Because this piece remains partially unprepared due to the disarticulation and variable orientation of ribs, it is possible that the remaining pectoral elements are preserved in some form.

The forelimb is represented only by a nearly complete left humerus ( Fig. 10A–D) and a partial right humerus. The complete humerus ( Fig. 10A–D) differs little from those reported in other dissorophids such as ‘ Broiliellus ’ reiszi , Cacops aspidephorus , Cacops morrisi , and Dissorophus multicinctus : the shaft is relatively short; the articular ends are set at about 60° relative to each other; and a supinator process is absent (e.g. Broili 1904, Williston 1910, DeMar 1968, Holmes et al. 2013, Gee and Reisz 2018a). It is notably shorter than that of the enigmatic dissorophine noted by Gee and Reisz (2018b) and Gee et al. (2019b). The shaft (which is partially covered by some matrix and consolidant) is oblate in cross-section, the ends are broadly expanded, and the proximal head has a broad convex articular surface ( Fig. 10A–D). The radial condyle appears relatively small, but the distal end is clearly damaged. The only notable feature that could differentiate this specimen from some other dissorophids is that the posterior margin leading from the shaft to the entepicondyle is strongly curved ( Fig. 10A, C) compared to that in most other dissorophids for which humeri are known, being most similar to Parioxys ferricolus ( Moustafa 1955a) . It differs little from the humerus figured for the holotype of Aspidosaurus chiton ( Broili 1904) beyond a slightly shorter shaft and a slightly larger entepicondyle, but these attributes may be a mere difference in figured perspective. Compared to the humerus in the holotype of ‘ Alegeinosaurus aphthitos ’, which is embedded in a block and thus not exposed on all sides, the humerus of USNM PAL 406323 appears to be slightly less torqued (although we have not taken precise measurements) and with a more slender shaft. Overall, it is important to note the general paucity of forelimb data for the majority of dissorophids and the near total absence of ontogenetic data.

There are two partial limb ends, possibly from a single element, that could pertain to a radius ( Fig. 8E–J). Both are relatively featureless and have oval cross-sectional articular surfaces and that narrow to a cylindrical shaft. There is no curvature, likely precluding a fibula from consideration, nor any evidence for an olecranon, likely precluding an ulna, and the ends are barely wider than the preserved shaft, precluding a tibia. One fragment bears a straight, narrow ridge running down the shaft ( Fig. 10F, H).

The pelvic girdle is represented only by a ventral portion of the ischium ( Fig. 11E–H). The lateral surface is gently concave, but no remnant of the acetabulum is preserved, and it offers no useful data for comparison.

The hindlimb is definitively represented only by the distal and proximal ends of a right femur ( Fig. 11A–D). As with the humerus, there are no substantial differences from the femora of other dissorophids such as Cacops aspidephorus and Parioxys ferricolous ( Williston 1910, Moustafa 1955a). The element appears to have been relatively long; there is a prominent adductor crest; and a narrow trochanter is present at the proximal end of the adductor crest ( Fig. 11A–D). It is unclear how long the element would have been because the shaft appears to have reached a consistently narrow circumference at the point of breakage in both fragments.

Embedded in a small block are five small rounded rectangular elements that represent part of the carpus or tarsus ( Fig. 11I). However, the exposed portions exhibit no particular identifying features, and their proportions are relatively equant and consistent among all five. Based on this, they most likely represent the distalmost carpal or tarsal elements (i.e. centrale 1–3 and distal carpals or tarsals). An isolated elongate bone of unclear relationship to this particular block also pertains to the carpus or tarsus. The largest surfaces are covered in finished bone, but the remaining surfaces are largely unfinished and presumably represent facets. Two distinct facets at one end are divided by a V-shaped groove on the finished bone surface. It most closely resembles the right fibulare as described by Dilkes (2015), specifically in taxa where the facets for the intermedium and centrale 4 are very narrowly separated (e.g. Dissorophus multicinctus ).

Phylogenetic results: The analysis with all 30 taxa (iteration 1) recovered 22 MPTs with a length of 330 steps (CI [consistency index] = 0.397; RI [retention index] = 0.554). The strict consensus topology ( Fig. 12A) is mostly unresolved and generally accords with results from the similar taxon sample analyses of Gee (2021). Olsoniformes is recovered as monophyletic, in contrast to some analyses of previous versions of the matrix in which relatively incomplete taxa like Reiszerpeton renascentis fall outside of the clade. Within Olsoniformes, there is little resolution beyond a monophyletic Trematopidae , a monophyletic Cacops , and the pairing of Anakamacops petrolicus and Kamacops acervalis ; Dissorophidae is not recovered as a clade of all of its nominal members.

As previously noted, this matrix samples several poorly known dissorophids that are typically excluded in previous analyses due to their incompleteness: Brevidorsum profundum , ‘ Broiliellus ’ arroyoensis , and Reiszerpeton renascentis . These were removed for iteration 2 ( Fig. 12B). Conjunctio multidens and Scapanops neglectus are also considered wildcards in some studies (e.g. Schoch 2012), but these are retained for iteration 2 because they historically cluster with cacopines, Co. multidens in particular has a superficially similar cranium to USNM PAL 406323 (and cacopines), and the entire cranium and their dorsal sutures are known (e.g. Schoch and Sues 2013). Iteration 2 (27 OTUs) recovered 13 MPTs with a length of 319 steps (CI = 0.411; RI = 0.578). The resolution of the strict consensus ( Fig. 12B) increased slightly within Olsoniformes, but nominal dissorophids still do not comprise a clade. However, this analysis recovered an expanded Cacopinae ( Cacops spp. + Anakamacops petrolicus , Aspidosaurus binasser , Kamacops acervalis , Parioxys ferricolus , and USNM PAL 406323) and a Dissorophinae comprising its nominal members ( Broiliellus spp. , Diploseira angusta , Dissorophus multicinctus ). Conjunctio multidens , S. neglectus , and Platyhystrix rugosa form early diverging branches of the olsoniform polytomy.

The third iteration removes taxa historically treated as wildcards ( Conjunctio multidens , Scapanops neglectus ) and those with little to no detailed information on the cranium (‘ Broiliellus ’ olsoni , Diploseira angusta , Kamacops acervalis ). Iteration 3 (22 OTUs) recovered 51 MPTs with a length of 294 steps (CI = 0.446; RI = 0.635). Despite the removal of these five taxa, the strict consensus ( Fig. 13A) is nearly identical to that of iteration 2, with the exception that a Dissorophidae comprised of Cacopine and Dissorophinae, to the exclusion of Platyhystrix rugosa , was recovered. This iteration is the first analysis in the history of this matrix where progressive taxon removal did not result in significant improvement to the strict consensus topology.

Based on the relatively consistent findings that USNM PAL 406323 clusters with cacopines, we performed a final iteration with a cacopine-focused taxon sample. The final iteration mirrors the dissorophid sample used by Schoch and Sues (2022b) in their analysis of Parioxys ferricolus , which also includes Anakamacops petrolicus , Broiliellus texensis , Cacops (all three species are included here), Conjunctio multidens , Dissorophus multicinctus , Kamacops acervalis , and Scapanops neglectus . Iteration 4 (21 OTUs) recovered just two MPTs with a length of 287 steps (CI = 0.456; RI = 0.650) and a well-resolved strict consensus ( Fig. 13B). This topology is discordant with the results of Schoch and Sues (2022b) but more similar to previous studies in which An. petrolicus + K. acervalis are the sister group to Cacops spp. Here, USNM PAL 406323 is the sister group to this core set of cacopines, followed by Co. multidens and then P. ferricolus ; the latter was recovered in a higher nested position by Schoch and Sues (2022b). Scapanops neglectus is not recovered within Olsoniformes and instead clusters with the two amphibamiforms. If Aspidosaurus binasser is re-added (22 OTUs), the analysis recovers five MPTs with a length of 293 steps (CI = 0.447; RI = 0.637), and the strict consensus topology ( Fig. 13C) has lost resolution in Cacopinae (which includes As. binasser ), with a monophyletic Cacops and the pairing of An. petrolicus + K. acervalis recovered. Other relationships are unchanged.

While adding scores for Parioxys ferricolus , we referenced the matrix of Schoch and Sues (2022a), which contains many of the same characters. In addition to noting a number of problematic scores, we were also unable to reproduce the reported results (five MPTs with a length of 291 steps; Fig. 14A). This may relate in part to the largely unreported analytical parameters; only the use of the New Technology Search (NTS); the treatment of multitstate characters; and the outgroup were specified by Schoch and Sues (2022b), so it is unclear if all other parameters were left as default (which would almost certainly be insufficient to explore the full tree space) or if parameters were modified. We re-analyzed the matrix published in the Dryad repository ( Schoch and Sues 2022a) using the NTS with parameters similar to those reported in other studies (e.g. Kligman et al. 2023): 1000 random addition sequences (RAS); sectorial search with 1000 rounds of constrained sectorial search; ratchet with 1000 iterations; tree fusing with 100 rounds; all multistate characters ordered; and ‘ Dendrysekos ’ as the outgroup. All other parameters were left as the default. A second round of TBR was run on the recovered MPTs using the ‘traditional’ heuristic search, which did not increase the number of MPTs. This re-analysis recovered nine MPTs with a length of 326 steps and a distinctly different strict consensus ( Fig. 14B) than that reported by Schoch and Sues, which include notable departures such as: (i) the nesting of Acanthostomatops vorax ( Credner, 1883) and Apateon pedestris Meyer, 1844 , within Olsoniformes; (ii) the clustering of Micromelerpeton credneri ( Bulman & Whittard, 1926) , with the non-branchiosaurid amphibamiforms; (iii) the clustering of Perryella olsoni with small-bodied dissorophoids, forming a sister clade to Olsoniformes; and (iv) different relationships among cacopines, including an earlier diverging position of Pa. ferricolus . Altering the analytical parameters (e.g. treating all multistate characters as unordered instead of ordered) was unsuccessful in even closely approaching, let alone reproducing, the reported results. This is unrelated to the inclusion of Reiszerpeton renascentis , which despite having been included in a variant analysis of Schoch and Sues (2022b), is not included in the data matrix (the resultant consensus of the trees in this variant analysis was not provided).

As a final exploratory test, we checked for possible frameshifts in the strings for each taxon because this has occurred previously for an earlier derivate of this matrix ( Schoch and Milner 2021) in which some strings are simply incomplete (i.e. there are fewer scores for some taxa than there are characters in the matrix). Close manual examination of this matrix revealed a frameshift in one taxon: all scores for Apateon pedestris are down-frameshifted by one position (e.g. the score for character 1 should be the score for character 2), which is visually discernible when comparing scores to those of other branchiosaurids. Re-analysis of the adjusted matrix (with state 1 scored for character 1, as in previous derivates; all existing scores up-shifted by one; and the missing data entry for the final character removed) and the same settings as the previous reanalysis recovered one MPT with a length of 295 steps ( Fig. 14C), which is much closer to the original results reported by Schoch and Sues (2022b), as is the topology. The most perplexing positions recovered in our original re-analysis no longer persist (e.g. Acanthostomatops vorax and Ap. pedestris ). However, there are still notable differences from the original results, which suggest that our corrected version of the matrix is still not the same as the original matrix analyzed by Schoch and Sues (2022b) or was analyzed with significantly different parameters than those listed in the original publication and that were used here, with Parioxys ferricolus being recovered in an earlier diverging position towards the base of Cacopinae (as with our analyses of the modified matrix fromGee 2021). Other peculiar and discordant results include the Late Carboniferous trematopids Actiobates peabodyi Eaton, 1973 , and Mattauschia laticeps, and the indeterminate olsoniform Palodromeus bairdi clustering at the base of Dissorophidae (though these taxa have been recovered outside of Trematopidae in previous studies that used some form of this matrix; see Schoch and Werneburg 2023a). A possible explanation is the presence of numerous problematical scores; some examples include: Parioxys being scored as having an entepicondylar foramen; Dissorophus being scored as having the trematopid nostril; Conjunctio and Scapanops being scored as having no internarial fenestra; Broiliellus being scored as having bicuspid teeth; Ecolsonia being scored as lacking osteoderms (its ossifications were not alternatively treated as bony scales); and intercentra being scored as absent in Cacops morrisi . Because we were unable to find evidence for a frameshift in taxa with these clearly erroneous scores, it remains unclear whether these errors were present in the version of the matrix that Schoch and Sues (2022b) utilized or whether they were introduced in the unnecessary conversion of the.tnt input file to.txt format ( Schoch and Sues 2022a). These errors and others are detailed in Appendix 2 (Gee 2025), but given our inability to reproduce the original reported results or to readily identify the full source(s) of these discrepancies and the fact that this matrix has already been superseded by other derivates, we did not attempt to fully revise it for re-analysis [see Gee (2021) for comments on propagated issues of character construction and scoring approaches, some of which characterize the most recent derivates].

An additional consideration that is not tested here is the search method used in TNT. As noted above, Schoch and Sues (2022b) used the NTS option, which allows for significantly more parameterization than the ‘traditional’ heuristic search, but that study did not specify most parameters. The NTS was designed for large genomic matrices at the scale of hundreds of OTUs and thousands of characters ( Goloboff et al. 2008) and is unnecessary for phenotypic matrices of the size that most temnospondyl workers and palaeontologists work with ( Torres et al. 2022). In both our reanalyses of the matrix of Schoch and Sues (2022a) and the modified matrix of Gee (2021), the NTS search recovered the same MPTs as a ‘traditional’ search, provided that a sufficiently large number of replicates were specified. Regardless, if the search parameters used by Schoch and Sues (2022b) were entirely or largely left as the defaults, this could have been insufficient to comprehensively explore the tree space (as would occur if using only the default parameters for any heuristic search in any software) and could either lead to finding no MPTs or only some MPTs. It should be noted that the algorithms underlying the NTS do not necessarily recover all MPTs; they only seek to recover a certain (arbitrarily defined) amount necessary to estimate a consensus (e.g. Goloboff 1999, 2002, Giribet 2005, Meier and Ali 2005; the ‘quick consensus’ of Goloboff and Farris 2001).

Comparison of the matrix of Schoch and Sues (2022a) with other derivates in the same family (e.g. Schoch and Milner 2021, Schoch 2022, Schoch and Werneburg 2023b, Werneburg et al. 2023, So et al. 2024) did not recover evidence for the same frameshift problem in Apateon pedestris (although we have not tested those matrices to see if the reported results are reproducible), which does suggest the potential for errors that occur only in one study. A different set of frameshift issues were identified in another set of derivates ( Schoch and Werneburg 2023b, Werneburg et al. 2023) by So et al. (2024). Finally, the derivate of Schoch and Milner (2021) is characterized by a third type of frameshift issue: incomplete strings (i.e. the string contains fewer than the required 109 positions for that matrix) for several taxa ( Broiliellus , Dendrerpeton Owen 1853 , Fedexia Berman et al. 2010 , and Perryella Carlson, 1987 ) in which the missing scores for the four former taxa are within the string, not at the end (i.e. not truncation of character 1 or 109). This more severe formatting issue is only possible because the strings were converted to PDF format and appended to the end of the article, rather than being provided as a stand-alone file in the disciplinepreferred NEXUS or TNT format. It remains possible that other formatting/frameshift issues that we did not detect account for why the data file for the Parioxys study fails to reproduce the reported results. Many of these errors seem to have been propagated into other matrices (e.g. Schoch and Werneburg 2023b, So et al. 2024) but did not originate with the source matrix ( Schoch 2019) or early derivates (e.g. Gee and Reisz 2020). We believe their first occurrence is in the matrix of Schoch et al. (2021).